From ace9429bb58fd418f0c81d4c2835699bddf6bde6 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Thu, 11 Apr 2024 10:27:49 +0200 Subject: Adding upstream version 6.6.15. Signed-off-by: Daniel Baumann --- drivers/gpu/drm/nouveau/nvkm/engine/Kbuild | 25 + drivers/gpu/drm/nouveau/nvkm/engine/bsp/Kbuild | 2 + drivers/gpu/drm/nouveau/nvkm/engine/bsp/g84.c | 44 + drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild | 12 + drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/com.fuc | 864 +++++ .../gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3 | 2 + .../drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h | 607 ++++ .../gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3 | 2 + .../drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h | 621 ++++ drivers/gpu/drm/nouveau/nvkm/engine/ce/ga100.c | 92 + drivers/gpu/drm/nouveau/nvkm/engine/ce/ga102.c | 45 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gf100.c | 69 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gk104.c | 101 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gm107.c | 43 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gm200.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gp100.c | 102 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c | 44 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gt215.c | 83 + drivers/gpu/drm/nouveau/nvkm/engine/ce/gv100.c | 64 + drivers/gpu/drm/nouveau/nvkm/engine/ce/priv.h | 16 + drivers/gpu/drm/nouveau/nvkm/engine/ce/tu102.c | 41 + drivers/gpu/drm/nouveau/nvkm/engine/cipher/Kbuild | 2 + drivers/gpu/drm/nouveau/nvkm/engine/cipher/g84.c | 133 + drivers/gpu/drm/nouveau/nvkm/engine/device/Kbuild | 7 + drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c | 58 + drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.h | 9 + drivers/gpu/drm/nouveau/nvkm/engine/device/base.c | 3268 +++++++++++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.c | 212 ++ drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.h | 13 + drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c | 1702 ++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/device/priv.h | 61 + drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c | 341 ++ drivers/gpu/drm/nouveau/nvkm/engine/device/user.c | 430 +++ drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild | 34 + drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c | 394 +++ drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.c | 275 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.h | 135 + drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.c | 95 + drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.h | 33 + drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.c | 853 +++++ drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.h | 102 + drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c | 348 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/g94.c | 378 +++ drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c | 150 + drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c | 1261 ++++++++ drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c | 332 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/gk110.c | 59 + drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c | 112 + drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c | 187 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c | 84 + drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c | 200 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/gt200.c | 109 + drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c | 229 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c | 1254 ++++++++ drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.c | 84 + drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.h | 16 + drivers/gpu/drm/nouveau/nvkm/engine/disp/head.c | 67 + drivers/gpu/drm/nouveau/nvkm/engine/disp/head.h | 63 + drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.c | 72 + drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h | 196 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c | 72 + drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c | 86 + drivers/gpu/drm/nouveau/nvkm/engine/disp/nv04.c | 130 + drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c | 1645 ++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c | 338 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h | 98 + drivers/gpu/drm/nouveau/nvkm/engine/disp/priv.h | 86 + drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c | 236 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/uconn.c | 213 ++ drivers/gpu/drm/nouveau/nvkm/engine/disp/udisp.c | 114 + drivers/gpu/drm/nouveau/nvkm/engine/disp/uhead.c | 127 + drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c | 379 +++ drivers/gpu/drm/nouveau/nvkm/engine/disp/vga.c | 205 ++ drivers/gpu/drm/nouveau/nvkm/engine/dma/Kbuild | 14 + drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c | 116 + drivers/gpu/drm/nouveau/nvkm/engine/dma/gf100.c | 37 + drivers/gpu/drm/nouveau/nvkm/engine/dma/gf119.c | 37 + drivers/gpu/drm/nouveau/nvkm/engine/dma/gv100.c | 35 + drivers/gpu/drm/nouveau/nvkm/engine/dma/nv04.c | 37 + drivers/gpu/drm/nouveau/nvkm/engine/dma/nv50.c | 37 + drivers/gpu/drm/nouveau/nvkm/engine/dma/priv.h | 19 + drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c | 138 + drivers/gpu/drm/nouveau/nvkm/engine/dma/user.h | 21 + .../gpu/drm/nouveau/nvkm/engine/dma/usergf100.c | 150 + .../gpu/drm/nouveau/nvkm/engine/dma/usergf119.c | 132 + .../gpu/drm/nouveau/nvkm/engine/dma/usergv100.c | 119 + drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv04.c | 134 + drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv50.c | 157 + drivers/gpu/drm/nouveau/nvkm/engine/falcon.c | 355 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild | 30 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c | 387 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.c | 252 ++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.h | 75 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c | 480 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h | 78 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.c | 111 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.h | 25 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c | 231 ++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/g98.c | 70 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga100.c | 611 ++++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c | 45 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c | 969 ++++++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c | 833 +++++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk110.c | 132 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk208.c | 77 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk20a.c | 52 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c | 156 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm200.c | 66 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c | 139 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/gv100.c | 492 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c | 543 ++++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv10.c | 113 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c | 142 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c | 252 ++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c | 398 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h | 209 ++ .../gpu/drm/nouveau/nvkm/engine/fifo/regsnv04.h | 133 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.c | 430 +++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.h | 131 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.c | 45 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.h | 31 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/tu102.c | 286 ++ drivers/gpu/drm/nouveau/nvkm/engine/fifo/ucgrp.c | 125 + drivers/gpu/drm/nouveau/nvkm/engine/fifo/uchan.c | 409 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild | 67 + drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c | 205 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxga102.c | 77 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c | 1587 ++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h | 276 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c | 109 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c | 807 +++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c | 357 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c | 308 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c | 527 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c | 1002 ++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c | 867 +++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c | 107 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c | 572 ++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c | 90 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c | 992 ++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm200.c | 130 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c | 91 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c | 151 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c | 125 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp104.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp107.c | 58 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgv100.c | 222 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c | 693 ++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h | 131 + drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c | 3347 ++++++++++++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxtu102.c | 94 + drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/com.fuc | 335 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc | 492 +++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h | 532 ++++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h | 539 ++++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h | 539 ++++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h | 539 ++++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h | 475 +++ .../drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5 | 42 + .../drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h | 607 ++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hub.fuc | 699 ++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h | 1049 ++++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h | 1049 ++++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h | 1046 ++++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h | 1046 ++++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h | 918 ++++++ .../drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5 | 40 + .../drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h | 918 ++++++ .../gpu/drm/nouveau/nvkm/engine/gr/fuc/macros.fuc | 261 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/os.h | 9 + drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c | 198 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/ga102.c | 354 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c | 2659 ++++++++++++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h | 448 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c | 161 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c | 160 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c | 136 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c | 203 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c | 227 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c | 506 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.h | 55 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c | 402 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c | 154 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c | 211 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c | 346 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c | 446 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c | 296 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c | 188 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/gp100.c | 165 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gp102.c | 161 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gp104.c | 102 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gp107.c | 91 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gp108.c | 98 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gp10b.c | 103 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c | 49 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/gr/gv100.c | 342 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c | 48 + drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c | 1426 +++++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c | 1221 +++++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h | 14 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c | 59 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c | 59 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c | 376 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h | 35 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c | 135 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c | 126 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c | 200 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c | 137 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c | 137 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c | 476 +++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h | 49 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c | 108 + drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c | 796 +++++ drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h | 37 + drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h | 47 + drivers/gpu/drm/nouveau/nvkm/engine/gr/regs.h | 275 ++ drivers/gpu/drm/nouveau/nvkm/engine/gr/tu102.c | 220 ++ drivers/gpu/drm/nouveau/nvkm/engine/mpeg/Kbuild | 6 + drivers/gpu/drm/nouveau/nvkm/engine/mpeg/g84.c | 44 + drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.c | 298 ++ drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.h | 32 + drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv40.c | 83 + drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv44.c | 216 ++ drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv50.c | 136 + drivers/gpu/drm/nouveau/nvkm/engine/mpeg/priv.h | 17 + drivers/gpu/drm/nouveau/nvkm/engine/msenc/Kbuild | 2 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/Kbuild | 6 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/base.c | 31 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/g98.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gf100.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gk104.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gt215.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/mspdec/priv.h | 12 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/Kbuild | 5 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/base.c | 31 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/g98.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/gf100.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/gt215.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/msppp/priv.h | 10 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/Kbuild | 7 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/base.c | 31 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/g98.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/gf100.c | 50 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/gk104.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/gt215.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/mcp89.c | 42 + drivers/gpu/drm/nouveau/nvkm/engine/msvld/priv.h | 12 + drivers/gpu/drm/nouveau/nvkm/engine/nvdec/Kbuild | 4 + drivers/gpu/drm/nouveau/nvkm/engine/nvdec/base.c | 61 + drivers/gpu/drm/nouveau/nvkm/engine/nvdec/ga102.c | 61 + drivers/gpu/drm/nouveau/nvkm/engine/nvdec/gm107.c | 58 + drivers/gpu/drm/nouveau/nvkm/engine/nvdec/priv.h | 19 + drivers/gpu/drm/nouveau/nvkm/engine/nvenc/Kbuild | 3 + drivers/gpu/drm/nouveau/nvkm/engine/nvenc/base.c | 62 + drivers/gpu/drm/nouveau/nvkm/engine/nvenc/gm107.c | 52 + drivers/gpu/drm/nouveau/nvkm/engine/nvenc/priv.h | 19 + drivers/gpu/drm/nouveau/nvkm/engine/pm/Kbuild | 11 + drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c | 868 +++++ drivers/gpu/drm/nouveau/nvkm/engine/pm/g84.c | 165 + drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c | 243 ++ drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.h | 20 + drivers/gpu/drm/nouveau/nvkm/engine/pm/gf108.c | 66 + drivers/gpu/drm/nouveau/nvkm/engine/pm/gf117.c | 80 + drivers/gpu/drm/nouveau/nvkm/engine/pm/gk104.c | 184 ++ drivers/gpu/drm/nouveau/nvkm/engine/pm/gt200.c | 157 + drivers/gpu/drm/nouveau/nvkm/engine/pm/gt215.c | 138 + drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.c | 123 + drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.h | 15 + drivers/gpu/drm/nouveau/nvkm/engine/pm/nv50.c | 175 + drivers/gpu/drm/nouveau/nvkm/engine/pm/priv.h | 105 + drivers/gpu/drm/nouveau/nvkm/engine/sec/Kbuild | 2 + .../gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s | 698 ++++ .../drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h | 585 ++++ drivers/gpu/drm/nouveau/nvkm/engine/sec/g98.c | 81 + drivers/gpu/drm/nouveau/nvkm/engine/sec2/Kbuild | 6 + drivers/gpu/drm/nouveau/nvkm/engine/sec2/base.c | 164 + drivers/gpu/drm/nouveau/nvkm/engine/sec2/ga102.c | 197 ++ drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c | 318 ++ drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp108.c | 43 + drivers/gpu/drm/nouveau/nvkm/engine/sec2/priv.h | 35 + drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c | 86 + drivers/gpu/drm/nouveau/nvkm/engine/sw/Kbuild | 10 + drivers/gpu/drm/nouveau/nvkm/engine/sw/base.c | 109 + drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.c | 91 + drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.h | 30 + drivers/gpu/drm/nouveau/nvkm/engine/sw/gf100.c | 147 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nv04.c | 139 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nv10.c | 68 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.c | 142 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.h | 21 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.c | 86 + drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.h | 22 + drivers/gpu/drm/nouveau/nvkm/engine/sw/priv.h | 22 + drivers/gpu/drm/nouveau/nvkm/engine/vic/Kbuild | 2 + drivers/gpu/drm/nouveau/nvkm/engine/vp/Kbuild | 2 + drivers/gpu/drm/nouveau/nvkm/engine/vp/g84.c | 43 + drivers/gpu/drm/nouveau/nvkm/engine/xtensa.c | 191 ++ 309 files changed, 76593 insertions(+) create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/bsp/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/bsp/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/com.fuc create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/ga100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gm200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gp100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/gv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/ce/tu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/cipher/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/cipher/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/device/user.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/g94.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gk110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gt200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/head.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/head.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/nv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/uconn.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/udisp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/uhead.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/disp/vga.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/gf119.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/gv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/nv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/user.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf119.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/usergv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/falcon.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/g98.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk208.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk20a.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/gv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv10.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/regsnv04.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/tu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/ucgrp.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/uchan.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxtu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/com.fuc create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hub.fuc create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/macros.fuc create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/os.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp108.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gp10b.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/gv100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/regs.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/gr/tu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv40.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv44.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mpeg/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msenc/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/g98.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/mspdec/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/g98.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msppp/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/g98.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/mcp89.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/msvld/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvdec/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvdec/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvdec/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvdec/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvdec/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvenc/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvenc/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvenc/gm107.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/nvenc/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gf108.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gf117.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gk104.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gt200.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/gt215.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/pm/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec/g98.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/ga102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp108.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/base.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/gf100.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nv04.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nv10.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/sw/priv.h create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/vic/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/vp/Kbuild create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/vp/g84.c create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/xtensa.c (limited to 'drivers/gpu/drm/nouveau/nvkm/engine') diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/Kbuild new file mode 100644 index 0000000000..c6dfed18f3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/Kbuild @@ -0,0 +1,25 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/falcon.o +nvkm-y += nvkm/engine/xtensa.o + +include $(src)/nvkm/engine/bsp/Kbuild +include $(src)/nvkm/engine/ce/Kbuild +include $(src)/nvkm/engine/cipher/Kbuild +include $(src)/nvkm/engine/device/Kbuild +include $(src)/nvkm/engine/disp/Kbuild +include $(src)/nvkm/engine/dma/Kbuild +include $(src)/nvkm/engine/fifo/Kbuild +include $(src)/nvkm/engine/gr/Kbuild +include $(src)/nvkm/engine/mpeg/Kbuild +include $(src)/nvkm/engine/msenc/Kbuild +include $(src)/nvkm/engine/mspdec/Kbuild +include $(src)/nvkm/engine/msppp/Kbuild +include $(src)/nvkm/engine/msvld/Kbuild +include $(src)/nvkm/engine/nvenc/Kbuild +include $(src)/nvkm/engine/nvdec/Kbuild +include $(src)/nvkm/engine/pm/Kbuild +include $(src)/nvkm/engine/sec/Kbuild +include $(src)/nvkm/engine/sec2/Kbuild +include $(src)/nvkm/engine/sw/Kbuild +include $(src)/nvkm/engine/vic/Kbuild +include $(src)/nvkm/engine/vp/Kbuild diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/bsp/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/bsp/Kbuild new file mode 100644 index 0000000000..b596b99051 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/bsp/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/bsp/g84.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/bsp/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/bsp/g84.c new file mode 100644 index 0000000000..39f6db269c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/bsp/g84.c @@ -0,0 +1,44 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Ilia Mirkin + */ +#include + +#include + +static const struct nvkm_xtensa_func +g84_bsp = { + .fifo_val = 0x1111, + .unkd28 = 0x90044, + .sclass = { + { -1, -1, NV74_BSP }, + {} + } +}; + +int +g84_bsp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_xtensa_new_(&g84_bsp, device, type, inst, + device->chipset != 0x92, 0x103000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild new file mode 100644 index 0000000000..8bf1635ffa --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/Kbuild @@ -0,0 +1,12 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/ce/gt215.o +nvkm-y += nvkm/engine/ce/gf100.o +nvkm-y += nvkm/engine/ce/gk104.o +nvkm-y += nvkm/engine/ce/gm107.o +nvkm-y += nvkm/engine/ce/gm200.o +nvkm-y += nvkm/engine/ce/gp100.o +nvkm-y += nvkm/engine/ce/gp102.o +nvkm-y += nvkm/engine/ce/gv100.o +nvkm-y += nvkm/engine/ce/tu102.o +nvkm-y += nvkm/engine/ce/ga100.o +nvkm-y += nvkm/engine/ce/ga102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/com.fuc b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/com.fuc new file mode 100644 index 0000000000..6226bcd98c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/com.fuc @@ -0,0 +1,864 @@ +/* fuc microcode for copy engine on gt215- chipsets + * + * Copyright 2011 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#ifdef GT215 +.section #gt215_ce_data +#else +.section #gf100_ce_data +#endif + +ctx_object: .b32 0 +#ifdef GT215 +ctx_dma: +ctx_dma_query: .b32 0 +ctx_dma_src: .b32 0 +ctx_dma_dst: .b32 0 +#endif +.equ #ctx_dma_count 3 +ctx_query_address_high: .b32 0 +ctx_query_address_low: .b32 0 +ctx_query_counter: .b32 0 +ctx_src_address_high: .b32 0 +ctx_src_address_low: .b32 0 +ctx_src_pitch: .b32 0 +ctx_src_tile_mode: .b32 0 +ctx_src_xsize: .b32 0 +ctx_src_ysize: .b32 0 +ctx_src_zsize: .b32 0 +ctx_src_zoff: .b32 0 +ctx_src_xoff: .b32 0 +ctx_src_yoff: .b32 0 +ctx_src_cpp: .b32 0 +ctx_dst_address_high: .b32 0 +ctx_dst_address_low: .b32 0 +ctx_dst_pitch: .b32 0 +ctx_dst_tile_mode: .b32 0 +ctx_dst_xsize: .b32 0 +ctx_dst_ysize: .b32 0 +ctx_dst_zsize: .b32 0 +ctx_dst_zoff: .b32 0 +ctx_dst_xoff: .b32 0 +ctx_dst_yoff: .b32 0 +ctx_dst_cpp: .b32 0 +ctx_format: .b32 0 +ctx_swz_const0: .b32 0 +ctx_swz_const1: .b32 0 +ctx_xcnt: .b32 0 +ctx_ycnt: .b32 0 +.align 256 + +dispatch_table: +// mthd 0x0000, NAME +.b16 0x000 1 +.b32 #ctx_object ~0xffffffff +// mthd 0x0100, NOP +.b16 0x040 1 +.b32 0x00010000 + #cmd_nop ~0xffffffff +// mthd 0x0140, PM_TRIGGER +.b16 0x050 1 +.b32 0x00010000 + #cmd_pm_trigger ~0xffffffff +#ifdef GT215 +// mthd 0x0180-0x018c, DMA_ +.b16 0x060 #ctx_dma_count +dispatch_dma: +.b32 0x00010000 + #cmd_dma ~0xffffffff +.b32 0x00010000 + #cmd_dma ~0xffffffff +.b32 0x00010000 + #cmd_dma ~0xffffffff +#endif +// mthd 0x0200-0x0218, SRC_TILE +.b16 0x80 7 +.b32 #ctx_src_tile_mode ~0x00000fff +.b32 #ctx_src_xsize ~0x0007ffff +.b32 #ctx_src_ysize ~0x00001fff +.b32 #ctx_src_zsize ~0x000007ff +.b32 #ctx_src_zoff ~0x00000fff +.b32 #ctx_src_xoff ~0x0007ffff +.b32 #ctx_src_yoff ~0x00001fff +// mthd 0x0220-0x0238, DST_TILE +.b16 0x88 7 +.b32 #ctx_dst_tile_mode ~0x00000fff +.b32 #ctx_dst_xsize ~0x0007ffff +.b32 #ctx_dst_ysize ~0x00001fff +.b32 #ctx_dst_zsize ~0x000007ff +.b32 #ctx_dst_zoff ~0x00000fff +.b32 #ctx_dst_xoff ~0x0007ffff +.b32 #ctx_dst_yoff ~0x00001fff +// mthd 0x0300-0x0304, EXEC, WRCACHE_FLUSH +.b16 0xc0 2 +.b32 0x00010000 + #cmd_exec ~0xffffffff +.b32 0x00010000 + #cmd_wrcache_flush ~0xffffffff +// mthd 0x030c-0x0340, various stuff +.b16 0xc3 14 +.b32 #ctx_src_address_high ~0x000000ff +.b32 #ctx_src_address_low ~0xffffffff +.b32 #ctx_dst_address_high ~0x000000ff +.b32 #ctx_dst_address_low ~0xffffffff +.b32 #ctx_src_pitch ~0x0007ffff +.b32 #ctx_dst_pitch ~0x0007ffff +.b32 #ctx_xcnt ~0x0000ffff +.b32 #ctx_ycnt ~0x00001fff +.b32 #ctx_format ~0x0333ffff +.b32 #ctx_swz_const0 ~0xffffffff +.b32 #ctx_swz_const1 ~0xffffffff +.b32 #ctx_query_address_high ~0x000000ff +.b32 #ctx_query_address_low ~0xffffffff +.b32 #ctx_query_counter ~0xffffffff +.b16 0x800 0 + +#ifdef GT215 +.section #gt215_ce_code +#else +.section #gf100_ce_code +#endif + +main: + clear b32 $r0 + mov $sp $r0 + + // setup i0 handler and route fifo and ctxswitch to it + mov $r1 #ih + mov $iv0 $r1 + mov $r1 0x400 + movw $r2 0xfff3 + sethi $r2 0 + iowr I[$r1 + 0x300] $r2 + + // enable interrupts + or $r2 0xc + iowr I[$r1] $r2 + bset $flags ie0 + + // enable fifo access and context switching + mov $r1 0x1200 + mov $r2 3 + iowr I[$r1] $r2 + + // sleep forever, waking for interrupts + bset $flags $p0 + spin: + sleep $p0 + bra #spin + +// i0 handler +ih: + iord $r1 I[$r0 + 0x200] + + and $r2 $r1 0x00000008 + bra e #ih_no_chsw + call #chsw + ih_no_chsw: + and $r2 $r1 0x00000004 + bra e #ih_no_cmd + call #dispatch + + ih_no_cmd: + and $r1 $r1 0x0000000c + iowr I[$r0 + 0x100] $r1 + iret + +// $p1 direction (0 = unload, 1 = load) +// $r3 channel +swctx: + mov $r4 0x7700 + mov $xtargets $r4 +#ifdef GT215 + // target 7 hardcoded to ctx dma object + mov $xdbase $r0 +#else + // read SCRATCH3 to decide if we are PCOPY0 or PCOPY1 + mov $r4 0x2100 + iord $r4 I[$r4 + 0] + and $r4 1 + shl b32 $r4 4 + add b32 $r4 0x30 + + // channel is in vram + mov $r15 0x61c + shl b32 $r15 6 + mov $r5 0x114 + iowrs I[$r15] $r5 + + // read 16-byte PCOPYn info, containing context pointer, from channel + shl b32 $r5 $r3 4 + add b32 $r5 2 + mov $xdbase $r5 + mov $r5 $sp + // get a chunk of stack space, aligned to 256 byte boundary + sub b32 $r5 0x100 + mov $r6 0xff + not b32 $r6 + and $r5 $r6 + sethi $r5 0x00020000 + xdld $r4 $r5 + xdwait + sethi $r5 0 + + // set context pointer, from within channel VM + mov $r14 0 + iowrs I[$r15] $r14 + ld b32 $r4 D[$r5 + 0] + shr b32 $r4 8 + ld b32 $r6 D[$r5 + 4] + shl b32 $r6 24 + or $r4 $r6 + mov $xdbase $r4 +#endif + // 256-byte context, at start of data segment + mov b32 $r4 $r0 + sethi $r4 0x60000 + + // swap! + bra $p1 #swctx_load + xdst $r0 $r4 + bra #swctx_done + swctx_load: + xdld $r0 $r4 + swctx_done: + xdwait + ret + +chsw: + // read current channel + mov $r2 0x1400 + iord $r3 I[$r2] + + // if it's active, unload it and return + xbit $r15 $r3 0x1e + bra e #chsw_no_unload + bclr $flags $p1 + call #swctx + bclr $r3 0x1e + iowr I[$r2] $r3 + mov $r4 1 + iowr I[$r2 + 0x200] $r4 + ret + + // read next channel + chsw_no_unload: + iord $r3 I[$r2 + 0x100] + + // is there a channel waiting to be loaded? + xbit $r13 $r3 0x1e + bra e #chsw_finish_load + bset $flags $p1 + call #swctx +#ifdef GT215 + // load dma objects back into TARGET regs + mov $r5 #ctx_dma + mov $r6 #ctx_dma_count + chsw_load_ctx_dma: + ld b32 $r7 D[$r5 + $r6 * 4] + add b32 $r8 $r6 0x180 + shl b32 $r8 8 + iowr I[$r8] $r7 + sub b32 $r6 1 + bra nc #chsw_load_ctx_dma +#endif + chsw_finish_load: + mov $r3 2 + iowr I[$r2 + 0x200] $r3 + ret + +dispatch: + // read incoming fifo command + mov $r3 0x1900 + iord $r2 I[$r3 + 0x100] + iord $r3 I[$r3 + 0x000] + and $r4 $r2 0x7ff + // $r2 will be used to store exception data + shl b32 $r2 0x10 + + // lookup method in the dispatch table, ILLEGAL_MTHD if not found + mov $r5 #dispatch_table + clear b32 $r6 + clear b32 $r7 + dispatch_loop: + ld b16 $r6 D[$r5 + 0] + ld b16 $r7 D[$r5 + 2] + add b32 $r5 4 + cmpu b32 $r4 $r6 + bra c #dispatch_illegal_mthd + add b32 $r7 $r6 + cmpu b32 $r4 $r7 + bra c #dispatch_valid_mthd + sub b32 $r7 $r6 + shl b32 $r7 3 + add b32 $r5 $r7 + bra #dispatch_loop + + // ensure no bits set in reserved fields, INVALID_BITFIELD + dispatch_valid_mthd: + sub b32 $r4 $r6 + shl b32 $r4 3 + add b32 $r4 $r5 + ld b32 $r5 D[$r4 + 4] + and $r5 $r3 + cmpu b32 $r5 0 + bra ne #dispatch_invalid_bitfield + + // depending on dispatch flags: execute method, or save data as state + ld b16 $r5 D[$r4 + 0] + ld b16 $r6 D[$r4 + 2] + cmpu b32 $r6 0 + bra ne #dispatch_cmd + st b32 D[$r5] $r3 + bra #dispatch_done + dispatch_cmd: + bclr $flags $p1 + call $r5 + bra $p1 #dispatch_error + bra #dispatch_done + + dispatch_invalid_bitfield: + or $r2 2 + dispatch_illegal_mthd: + or $r2 1 + + // store exception data in SCRATCH0/SCRATCH1, signal hostirq + dispatch_error: + mov $r4 0x1000 + iowr I[$r4 + 0x000] $r2 + iowr I[$r4 + 0x100] $r3 + mov $r2 0x40 + iowr I[$r0] $r2 + hostirq_wait: + iord $r2 I[$r0 + 0x200] + and $r2 0x40 + cmpu b32 $r2 0 + bra ne #hostirq_wait + + dispatch_done: + mov $r2 0x1d00 + mov $r3 1 + iowr I[$r2] $r3 + ret + +// No-operation +// +// Inputs: +// $r1: irqh state +// $r2: hostirq state +// $r3: data +// $r4: dispatch table entry +// Outputs: +// $r1: irqh state +// $p1: set on error +// $r2: hostirq state +// $r3: data +cmd_nop: + ret + +// PM_TRIGGER +// +// Inputs: +// $r1: irqh state +// $r2: hostirq state +// $r3: data +// $r4: dispatch table entry +// Outputs: +// $r1: irqh state +// $p1: set on error +// $r2: hostirq state +// $r3: data +cmd_pm_trigger: + mov $r2 0x2200 + clear b32 $r3 + sethi $r3 0x20000 + iowr I[$r2] $r3 + ret + +#ifdef GT215 +// SET_DMA_* method handler +// +// Inputs: +// $r1: irqh state +// $r2: hostirq state +// $r3: data +// $r4: dispatch table entry +// Outputs: +// $r1: irqh state +// $p1: set on error +// $r2: hostirq state +// $r3: data +cmd_dma: + sub b32 $r4 #dispatch_dma + shr b32 $r4 1 + bset $r3 0x1e + st b32 D[$r4 + #ctx_dma] $r3 + add b32 $r4 0x600 + shl b32 $r4 6 + iowr I[$r4] $r3 + ret +#endif + +// Calculates the hw swizzle mask and adjusts the surface's xcnt to match +// +cmd_exec_set_format: + // zero out a chunk of the stack to store the swizzle into + add $sp -0x10 + st b32 D[$sp + 0x00] $r0 + st b32 D[$sp + 0x04] $r0 + st b32 D[$sp + 0x08] $r0 + st b32 D[$sp + 0x0c] $r0 + + // extract cpp, src_ncomp and dst_ncomp from FORMAT + ld b32 $r4 D[$r0 + #ctx_format] + extr $r5 $r4 16:17 + add b32 $r5 1 + extr $r6 $r4 20:21 + add b32 $r6 1 + extr $r7 $r4 24:25 + add b32 $r7 1 + + // convert FORMAT swizzle mask to hw swizzle mask + bclr $flags $p2 + clear b32 $r8 + clear b32 $r9 + ncomp_loop: + and $r10 $r4 0xf + shr b32 $r4 4 + clear b32 $r11 + bpc_loop: + cmpu b8 $r10 4 + bra nc #cmp_c0 + mulu $r12 $r10 $r5 + add b32 $r12 $r11 + bset $flags $p2 + bra #bpc_next + cmp_c0: + bra ne #cmp_c1 + mov $r12 0x10 + add b32 $r12 $r11 + bra #bpc_next + cmp_c1: + cmpu b8 $r10 6 + bra nc #cmp_zero + mov $r12 0x14 + add b32 $r12 $r11 + bra #bpc_next + cmp_zero: + mov $r12 0x80 + bpc_next: + st b8 D[$sp + $r8] $r12 + add b32 $r8 1 + add b32 $r11 1 + cmpu b32 $r11 $r5 + bra c #bpc_loop + add b32 $r9 1 + cmpu b32 $r9 $r7 + bra c #ncomp_loop + + // SRC_XCNT = (xcnt * src_cpp), or 0 if no src ref in swz (hw will hang) + mulu $r6 $r5 + st b32 D[$r0 + #ctx_src_cpp] $r6 + ld b32 $r8 D[$r0 + #ctx_xcnt] + mulu $r6 $r8 + bra $p2 #dst_xcnt + clear b32 $r6 + + dst_xcnt: + mulu $r7 $r5 + st b32 D[$r0 + #ctx_dst_cpp] $r7 + mulu $r7 $r8 + + mov $r5 0x810 + shl b32 $r5 6 + iowr I[$r5 + 0x000] $r6 + iowr I[$r5 + 0x100] $r7 + add b32 $r5 0x800 + ld b32 $r6 D[$r0 + #ctx_dst_cpp] + sub b32 $r6 1 + shl b32 $r6 8 + ld b32 $r7 D[$r0 + #ctx_src_cpp] + sub b32 $r7 1 + or $r6 $r7 + iowr I[$r5 + 0x000] $r6 + add b32 $r5 0x100 + ld b32 $r6 D[$sp + 0x00] + iowr I[$r5 + 0x000] $r6 + ld b32 $r6 D[$sp + 0x04] + iowr I[$r5 + 0x100] $r6 + ld b32 $r6 D[$sp + 0x08] + iowr I[$r5 + 0x200] $r6 + ld b32 $r6 D[$sp + 0x0c] + iowr I[$r5 + 0x300] $r6 + add b32 $r5 0x400 + ld b32 $r6 D[$r0 + #ctx_swz_const0] + iowr I[$r5 + 0x000] $r6 + ld b32 $r6 D[$r0 + #ctx_swz_const1] + iowr I[$r5 + 0x100] $r6 + add $sp 0x10 + ret + +// Setup to handle a tiled surface +// +// Calculates a number of parameters the hardware requires in order +// to correctly handle tiling. +// +// Offset calculation is performed as follows (Tp/Th/Td from TILE_MODE): +// nTx = round_up(w * cpp, 1 << Tp) >> Tp +// nTy = round_up(h, 1 << Th) >> Th +// Txo = (x * cpp) & ((1 << Tp) - 1) +// Tx = (x * cpp) >> Tp +// Tyo = y & ((1 << Th) - 1) +// Ty = y >> Th +// Tzo = z & ((1 << Td) - 1) +// Tz = z >> Td +// +// off = (Tzo << Tp << Th) + (Tyo << Tp) + Txo +// off += ((Tz * nTy * nTx)) + (Ty * nTx) + Tx) << Td << Th << Tp; +// +// Inputs: +// $r4: hw command (0x104800) +// $r5: ctx offset adjustment for src/dst selection +// $p2: set if dst surface +// +cmd_exec_set_surface_tiled: + // translate TILE_MODE into Tp, Th, Td shift values + ld b32 $r7 D[$r5 + #ctx_src_tile_mode] + extr $r9 $r7 8:11 + extr $r8 $r7 4:7 +#ifdef GT215 + add b32 $r8 2 +#else + add b32 $r8 3 +#endif + extr $r7 $r7 0:3 + cmp b32 $r7 0xe + bra ne #xtile64 + mov $r7 4 + bra #xtileok + xtile64: + xbit $r7 $flags $p2 + add b32 $r7 17 + bset $r4 $r7 + mov $r7 6 + xtileok: + + // Op = (x * cpp) & ((1 << Tp) - 1) + // Tx = (x * cpp) >> Tp + ld b32 $r10 D[$r5 + #ctx_src_xoff] + ld b32 $r11 D[$r5 + #ctx_src_cpp] + mulu $r10 $r11 + mov $r11 1 + shl b32 $r11 $r7 + sub b32 $r11 1 + and $r12 $r10 $r11 + shr b32 $r10 $r7 + + // Tyo = y & ((1 << Th) - 1) + // Ty = y >> Th + ld b32 $r13 D[$r5 + #ctx_src_yoff] + mov $r14 1 + shl b32 $r14 $r8 + sub b32 $r14 1 + and $r11 $r13 $r14 + shr b32 $r13 $r8 + + // YTILE = ((1 << Th) << 12) | ((1 << Th) - Tyo) + add b32 $r14 1 + shl b32 $r15 $r14 12 + sub b32 $r14 $r11 + or $r15 $r14 + xbit $r6 $flags $p2 + add b32 $r6 0x208 + shl b32 $r6 8 + iowr I[$r6 + 0x000] $r15 + + // Op += Tyo << Tp + shl b32 $r11 $r7 + add b32 $r12 $r11 + + // nTx = ((w * cpp) + ((1 << Tp) - 1) >> Tp) + ld b32 $r15 D[$r5 + #ctx_src_xsize] + ld b32 $r11 D[$r5 + #ctx_src_cpp] + mulu $r15 $r11 + mov $r11 1 + shl b32 $r11 $r7 + sub b32 $r11 1 + add b32 $r15 $r11 + shr b32 $r15 $r7 + push $r15 + + // nTy = (h + ((1 << Th) - 1)) >> Th + ld b32 $r15 D[$r5 + #ctx_src_ysize] + mov $r11 1 + shl b32 $r11 $r8 + sub b32 $r11 1 + add b32 $r15 $r11 + shr b32 $r15 $r8 + push $r15 + + // Tys = Tp + Th + // CFG_YZ_TILE_SIZE = ((1 << Th) >> 2) << Td + add b32 $r7 $r8 + sub b32 $r8 2 + mov $r11 1 + shl b32 $r11 $r8 + shl b32 $r11 $r9 + + // Tzo = z & ((1 << Td) - 1) + // Tz = z >> Td + // Op += Tzo << Tys + // Ts = Tys + Td + ld b32 $r8 D[$r5 + #ctx_src_zoff] + mov $r14 1 + shl b32 $r14 $r9 + sub b32 $r14 1 + and $r15 $r8 $r14 + shl b32 $r15 $r7 + add b32 $r12 $r15 + add b32 $r7 $r9 + shr b32 $r8 $r9 + + // Ot = ((Tz * nTy * nTx) + (Ty * nTx) + Tx) << Ts + pop $r15 + pop $r9 + mulu $r13 $r9 + add b32 $r10 $r13 + mulu $r8 $r9 + mulu $r8 $r15 + add b32 $r10 $r8 + shl b32 $r10 $r7 + + // PITCH = (nTx - 1) << Ts + sub b32 $r9 1 + shl b32 $r9 $r7 + iowr I[$r6 + 0x200] $r9 + + // SRC_ADDRESS_LOW = (Ot + Op) & 0xffffffff + // CFG_ADDRESS_HIGH |= ((Ot + Op) >> 32) << 16 + ld b32 $r7 D[$r5 + #ctx_src_address_low] + ld b32 $r8 D[$r5 + #ctx_src_address_high] + add b32 $r10 $r12 + add b32 $r7 $r10 + adc b32 $r8 0 + shl b32 $r8 16 + or $r8 $r11 + sub b32 $r6 0x600 + iowr I[$r6 + 0x000] $r7 + add b32 $r6 0x400 + iowr I[$r6 + 0x000] $r8 + ret + +// Setup to handle a linear surface +// +// Nothing to see here.. Sets ADDRESS and PITCH, pretty non-exciting +// +cmd_exec_set_surface_linear: + xbit $r6 $flags $p2 + add b32 $r6 0x202 + shl b32 $r6 8 + ld b32 $r7 D[$r5 + #ctx_src_address_low] + iowr I[$r6 + 0x000] $r7 + add b32 $r6 0x400 + ld b32 $r7 D[$r5 + #ctx_src_address_high] + shl b32 $r7 16 + iowr I[$r6 + 0x000] $r7 + add b32 $r6 0x400 + ld b32 $r7 D[$r5 + #ctx_src_pitch] + iowr I[$r6 + 0x000] $r7 + ret + +// wait for regs to be available for use +cmd_exec_wait: + push $r0 + push $r1 + mov $r0 0x800 + shl b32 $r0 6 + loop: + iord $r1 I[$r0] + and $r1 1 + bra ne #loop + pop $r1 + pop $r0 + ret + +cmd_exec_query: + // if QUERY_SHORT not set, write out { -, 0, TIME_LO, TIME_HI } + xbit $r4 $r3 13 + bra ne #query_counter + call #cmd_exec_wait + mov $r4 0x80c + shl b32 $r4 6 + ld b32 $r5 D[$r0 + #ctx_query_address_low] + add b32 $r5 4 + iowr I[$r4 + 0x000] $r5 + iowr I[$r4 + 0x100] $r0 + mov $r5 0xc + iowr I[$r4 + 0x200] $r5 + add b32 $r4 0x400 + ld b32 $r5 D[$r0 + #ctx_query_address_high] + shl b32 $r5 16 + iowr I[$r4 + 0x000] $r5 + add b32 $r4 0x500 + mov $r5 0x00000b00 + sethi $r5 0x00010000 + iowr I[$r4 + 0x000] $r5 + mov $r5 0x00004040 + shl b32 $r5 1 + sethi $r5 0x80800000 + iowr I[$r4 + 0x100] $r5 + mov $r5 0x00001110 + sethi $r5 0x13120000 + iowr I[$r4 + 0x200] $r5 + mov $r5 0x00001514 + sethi $r5 0x17160000 + iowr I[$r4 + 0x300] $r5 + mov $r5 0x00002601 + sethi $r5 0x00010000 + mov $r4 0x800 + shl b32 $r4 6 + iowr I[$r4 + 0x000] $r5 + + // write COUNTER + query_counter: + call #cmd_exec_wait + mov $r4 0x80c + shl b32 $r4 6 + ld b32 $r5 D[$r0 + #ctx_query_address_low] + iowr I[$r4 + 0x000] $r5 + iowr I[$r4 + 0x100] $r0 + mov $r5 0x4 + iowr I[$r4 + 0x200] $r5 + add b32 $r4 0x400 + ld b32 $r5 D[$r0 + #ctx_query_address_high] + shl b32 $r5 16 + iowr I[$r4 + 0x000] $r5 + add b32 $r4 0x500 + mov $r5 0x00000300 + iowr I[$r4 + 0x000] $r5 + mov $r5 0x00001110 + sethi $r5 0x13120000 + iowr I[$r4 + 0x100] $r5 + ld b32 $r5 D[$r0 + #ctx_query_counter] + add b32 $r4 0x500 + iowr I[$r4 + 0x000] $r5 + mov $r5 0x00002601 + sethi $r5 0x00010000 + mov $r4 0x800 + shl b32 $r4 6 + iowr I[$r4 + 0x000] $r5 + ret + +// Execute a copy operation +// +// Inputs: +// $r1: irqh state +// $r2: hostirq state +// $r3: data +// 000002000 QUERY_SHORT +// 000001000 QUERY +// 000000100 DST_LINEAR +// 000000010 SRC_LINEAR +// 000000001 FORMAT +// $r4: dispatch table entry +// Outputs: +// $r1: irqh state +// $p1: set on error +// $r2: hostirq state +// $r3: data +cmd_exec: + call #cmd_exec_wait + + // if format requested, call function to calculate it, otherwise + // fill in cpp/xcnt for both surfaces as if (cpp == 1) + xbit $r15 $r3 0 + bra e #cmd_exec_no_format + call #cmd_exec_set_format + mov $r4 0x200 + bra #cmd_exec_init_src_surface + cmd_exec_no_format: + mov $r6 0x810 + shl b32 $r6 6 + mov $r7 1 + st b32 D[$r0 + #ctx_src_cpp] $r7 + st b32 D[$r0 + #ctx_dst_cpp] $r7 + ld b32 $r7 D[$r0 + #ctx_xcnt] + iowr I[$r6 + 0x000] $r7 + iowr I[$r6 + 0x100] $r7 + clear b32 $r4 + + cmd_exec_init_src_surface: + bclr $flags $p2 + clear b32 $r5 + xbit $r15 $r3 4 + bra e #src_tiled + call #cmd_exec_set_surface_linear + bra #cmd_exec_init_dst_surface + src_tiled: + call #cmd_exec_set_surface_tiled + bset $r4 7 + + cmd_exec_init_dst_surface: + bset $flags $p2 + mov $r5 #ctx_dst_address_high - #ctx_src_address_high + xbit $r15 $r3 8 + bra e #dst_tiled + call #cmd_exec_set_surface_linear + bra #cmd_exec_kick + dst_tiled: + call #cmd_exec_set_surface_tiled + bset $r4 8 + + cmd_exec_kick: + mov $r5 0x800 + shl b32 $r5 6 + ld b32 $r6 D[$r0 + #ctx_ycnt] + iowr I[$r5 + 0x100] $r6 + mov $r6 0x0041 + // SRC_TARGET = 1, DST_TARGET = 2 + sethi $r6 0x44000000 + or $r4 $r6 + iowr I[$r5] $r4 + + // if requested, queue up a QUERY write after the copy has completed + xbit $r15 $r3 12 + bra e #cmd_exec_done + call #cmd_exec_query + + cmd_exec_done: + ret + +// Flush write cache +// +// Inputs: +// $r1: irqh state +// $r2: hostirq state +// $r3: data +// $r4: dispatch table entry +// Outputs: +// $r1: irqh state +// $p1: set on error +// $r2: hostirq state +// $r3: data +cmd_wrcache_flush: + mov $r2 0x2200 + clear b32 $r3 + sethi $r3 0x10000 + iowr I[$r2] $r3 + ret + +.align 0x100 diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3 new file mode 100644 index 0000000000..36f0a99ac7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3 @@ -0,0 +1,2 @@ +#define GF100 +#include "com.fuc" diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h new file mode 100644 index 0000000000..96d934f816 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gf100.fuc3.h @@ -0,0 +1,607 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gf100_ce_data[] = { +/* 0x0000: ctx_object */ + 0x00000000, +/* 0x0004: ctx_query_address_high */ + 0x00000000, +/* 0x0008: ctx_query_address_low */ + 0x00000000, +/* 0x000c: ctx_query_counter */ + 0x00000000, +/* 0x0010: ctx_src_address_high */ + 0x00000000, +/* 0x0014: ctx_src_address_low */ + 0x00000000, +/* 0x0018: ctx_src_pitch */ + 0x00000000, +/* 0x001c: ctx_src_tile_mode */ + 0x00000000, +/* 0x0020: ctx_src_xsize */ + 0x00000000, +/* 0x0024: ctx_src_ysize */ + 0x00000000, +/* 0x0028: ctx_src_zsize */ + 0x00000000, +/* 0x002c: ctx_src_zoff */ + 0x00000000, +/* 0x0030: ctx_src_xoff */ + 0x00000000, +/* 0x0034: ctx_src_yoff */ + 0x00000000, +/* 0x0038: ctx_src_cpp */ + 0x00000000, +/* 0x003c: ctx_dst_address_high */ + 0x00000000, +/* 0x0040: ctx_dst_address_low */ + 0x00000000, +/* 0x0044: ctx_dst_pitch */ + 0x00000000, +/* 0x0048: ctx_dst_tile_mode */ + 0x00000000, +/* 0x004c: ctx_dst_xsize */ + 0x00000000, +/* 0x0050: ctx_dst_ysize */ + 0x00000000, +/* 0x0054: ctx_dst_zsize */ + 0x00000000, +/* 0x0058: ctx_dst_zoff */ + 0x00000000, +/* 0x005c: ctx_dst_xoff */ + 0x00000000, +/* 0x0060: ctx_dst_yoff */ + 0x00000000, +/* 0x0064: ctx_dst_cpp */ + 0x00000000, +/* 0x0068: ctx_format */ + 0x00000000, +/* 0x006c: ctx_swz_const0 */ + 0x00000000, +/* 0x0070: ctx_swz_const1 */ + 0x00000000, +/* 0x0074: ctx_xcnt */ + 0x00000000, +/* 0x0078: ctx_ycnt */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: dispatch_table */ + 0x00010000, + 0x00000000, + 0x00000000, + 0x00010040, + 0x0001019f, + 0x00000000, + 0x00010050, + 0x000101a1, + 0x00000000, + 0x00070080, + 0x0000001c, + 0xfffff000, + 0x00000020, + 0xfff80000, + 0x00000024, + 0xffffe000, + 0x00000028, + 0xfffff800, + 0x0000002c, + 0xfffff000, + 0x00000030, + 0xfff80000, + 0x00000034, + 0xffffe000, + 0x00070088, + 0x00000048, + 0xfffff000, + 0x0000004c, + 0xfff80000, + 0x00000050, + 0xffffe000, + 0x00000054, + 0xfffff800, + 0x00000058, + 0xfffff000, + 0x0000005c, + 0xfff80000, + 0x00000060, + 0xffffe000, + 0x000200c0, + 0x000104b8, + 0x00000000, + 0x00010541, + 0x00000000, + 0x000e00c3, + 0x00000010, + 0xffffff00, + 0x00000014, + 0x00000000, + 0x0000003c, + 0xffffff00, + 0x00000040, + 0x00000000, + 0x00000018, + 0xfff80000, + 0x00000044, + 0xfff80000, + 0x00000074, + 0xffff0000, + 0x00000078, + 0xffffe000, + 0x00000068, + 0xfccc0000, + 0x0000006c, + 0x00000000, + 0x00000070, + 0x00000000, + 0x00000004, + 0xffffff00, + 0x00000008, + 0x00000000, + 0x0000000c, + 0x00000000, + 0x00000800, +}; + +static uint32_t gf100_ce_code[] = { +/* 0x0000: main */ + 0x04fe04bd, + 0x3517f000, + 0xf10010fe, + 0xf1040017, + 0xf0fff327, + 0x12d00023, + 0x0c25f0c0, + 0xf40012d0, + 0x17f11031, + 0x27f01200, + 0x0012d003, +/* 0x002f: spin */ + 0xf40031f4, + 0x0ef40028, +/* 0x0035: ih */ + 0x8001cffd, + 0xf40812c4, + 0x21f4060b, +/* 0x0041: ih_no_chsw */ + 0x0412c4ca, + 0xf5070bf4, +/* 0x004b: ih_no_cmd */ + 0xc4010221, + 0x01d00c11, +/* 0x0053: swctx */ + 0xf101f840, + 0xfe770047, + 0x47f1004b, + 0x44cf2100, + 0x0144f000, + 0xb60444b6, + 0xf7f13040, + 0xf4b6061c, + 0x1457f106, + 0x00f5d101, + 0xb6043594, + 0x57fe0250, + 0x0145fe00, + 0x010052b7, + 0x00ff67f1, + 0x56fd60bd, + 0x0253f004, + 0xf80545fa, + 0x0053f003, + 0xd100e7f0, + 0x549800fe, + 0x0845b600, + 0xb6015698, + 0x46fd1864, + 0x0047fe05, + 0xf00204b9, + 0x01f40643, + 0x0604fa09, +/* 0x00c3: swctx_load */ + 0xfa060ef4, +/* 0x00c6: swctx_done */ + 0x03f80504, +/* 0x00ca: chsw */ + 0x27f100f8, + 0x23cf1400, + 0x1e3fc800, + 0xf4170bf4, + 0x21f40132, + 0x1e3af053, + 0xf00023d0, + 0x24d00147, +/* 0x00eb: chsw_no_unload */ + 0xcf00f880, + 0x3dc84023, + 0x090bf41e, + 0xf40131f4, +/* 0x00fa: chsw_finish_load */ + 0x37f05321, + 0x8023d002, +/* 0x0102: dispatch */ + 0x37f100f8, + 0x32cf1900, + 0x0033cf40, + 0x07ff24e4, + 0xf11024b6, + 0xbd010057, +/* 0x011b: dispatch_loop */ + 0x5874bd64, + 0x57580056, + 0x0450b601, + 0xf40446b8, + 0x76bb4d08, + 0x0447b800, + 0xbb0f08f4, + 0x74b60276, + 0x0057bb03, +/* 0x013f: dispatch_valid_mthd */ + 0xbbdf0ef4, + 0x44b60246, + 0x0045bb03, + 0xfd014598, + 0x54b00453, + 0x201bf400, + 0x58004558, + 0x64b00146, + 0x091bf400, + 0xf4005380, +/* 0x0166: dispatch_cmd */ + 0x32f4300e, + 0xf455f901, + 0x0ef40c01, +/* 0x0171: dispatch_invalid_bitfield */ + 0x0225f025, +/* 0x0174: dispatch_illegal_mthd */ +/* 0x0177: dispatch_error */ + 0xf10125f0, + 0xd0100047, + 0x43d00042, + 0x4027f040, +/* 0x0187: hostirq_wait */ + 0xcf0002d0, + 0x24f08002, + 0x0024b040, +/* 0x0193: dispatch_done */ + 0xf1f71bf4, + 0xf01d0027, + 0x23d00137, +/* 0x019f: cmd_nop */ + 0xf800f800, +/* 0x01a1: cmd_pm_trigger */ + 0x0027f100, + 0xf034bd22, + 0x23d00233, +/* 0x01af: cmd_exec_set_format */ + 0xf400f800, + 0x01b0f030, + 0x0101b000, + 0xb00201b0, + 0x04980301, + 0x3045c71a, + 0xc70150b6, + 0x60b63446, + 0x3847c701, + 0xf40170b6, + 0x84bd0232, +/* 0x01da: ncomp_loop */ + 0x4ac494bd, + 0x0445b60f, +/* 0x01e2: bpc_loop */ + 0xa430b4bd, + 0x0f18f404, + 0xbbc0a5ff, + 0x31f400cb, + 0x220ef402, +/* 0x01f4: cmp_c0 */ + 0xf00c1bf4, + 0xcbbb10c7, + 0x160ef400, +/* 0x0200: cmp_c1 */ + 0xf406a430, + 0xc7f00c18, + 0x00cbbb14, +/* 0x020f: cmp_zero */ + 0xf1070ef4, +/* 0x0213: bpc_next */ + 0x380080c7, + 0x80b601c8, + 0x01b0b601, + 0xf404b5b8, + 0x90b6c308, + 0x0497b801, + 0xfdb208f4, + 0x06800065, + 0x1d08980e, + 0xf40068fd, + 0x64bd0502, +/* 0x023c: dst_xcnt */ + 0x800075fd, + 0x78fd1907, + 0x1057f100, + 0x0654b608, + 0xd00056d0, + 0x50b74057, + 0x06980800, + 0x0162b619, + 0x980864b6, + 0x72b60e07, + 0x0567fd01, + 0xb70056d0, + 0xb4010050, + 0x56d00060, + 0x0160b400, + 0xb44056d0, + 0x56d00260, + 0x0360b480, + 0xb7c056d0, + 0x98040050, + 0x56d01b06, + 0x1c069800, + 0xf44056d0, + 0x00f81030, +/* 0x029c: cmd_exec_set_surface_tiled */ + 0xc7075798, + 0x78c76879, + 0x0380b664, + 0xb06077c7, + 0x1bf40e76, + 0x0477f009, +/* 0x02b7: xtile64 */ + 0xf00f0ef4, + 0x70b6027c, + 0x0947fd11, +/* 0x02c3: xtileok */ + 0x980677f0, + 0x5b980c5a, + 0x00abfd0e, + 0xbb01b7f0, + 0xb2b604b7, + 0xc4abff01, + 0x9805a7bb, + 0xe7f00d5d, + 0x04e8bb01, + 0xff01e2b6, + 0xd8bbb4de, + 0x01e0b605, + 0xbb0cef94, + 0xfefd02eb, + 0x026cf005, + 0x020860b7, + 0xd00864b6, + 0xb7bb006f, + 0x00cbbb04, + 0x98085f98, + 0xfbfd0e5b, + 0x01b7f000, + 0xb604b7bb, + 0xfbbb01b2, + 0x05f7bb00, + 0x5f98f0f9, + 0x01b7f009, + 0xb604b8bb, + 0xfbbb01b2, + 0x05f8bb00, + 0x78bbf0f9, + 0x0282b600, + 0xbb01b7f0, + 0xb9bb04b8, + 0x0b589804, + 0xbb01e7f0, + 0xe2b604e9, + 0xf48eff01, + 0xbb04f7bb, + 0x79bb00cf, + 0x0589bb00, + 0x90fcf0fc, + 0xbb00d9fd, + 0x89fd00ad, + 0x008ffd00, + 0xbb00a8bb, + 0x92b604a7, + 0x0497bb01, + 0x988069d0, + 0x58980557, + 0x00acbb04, + 0xb6007abb, + 0x84b60081, + 0x058bfd10, + 0x060062b7, + 0xb70067d0, + 0xd0040060, + 0x00f80068, +/* 0x03a8: cmd_exec_set_surface_linear */ + 0xb7026cf0, + 0xb6020260, + 0x57980864, + 0x0067d005, + 0x040060b7, + 0xb6045798, + 0x67d01074, + 0x0060b700, + 0x06579804, + 0xf80067d0, +/* 0x03d1: cmd_exec_wait */ + 0xf900f900, + 0x0007f110, + 0x0604b608, +/* 0x03dc: loop */ + 0xf00001cf, + 0x1bf40114, + 0xfc10fcfa, +/* 0x03eb: cmd_exec_query */ + 0xc800f800, + 0x1bf40d34, + 0xd121f570, + 0x0c47f103, + 0x0644b608, + 0xb6020598, + 0x45d00450, + 0x4040d000, + 0xd00c57f0, + 0x40b78045, + 0x05980400, + 0x1054b601, + 0xb70045d0, + 0xf1050040, + 0xf00b0057, + 0x45d00153, + 0x4057f100, + 0x0154b640, + 0x808053f1, + 0xf14045d0, + 0xf1111057, + 0xd0131253, + 0x57f18045, + 0x53f11514, + 0x45d01716, + 0x0157f1c0, + 0x0153f026, + 0x080047f1, + 0xd00644b6, +/* 0x045e: query_counter */ + 0x21f50045, + 0x47f103d1, + 0x44b6080c, + 0x02059806, + 0xd00045d0, + 0x57f04040, + 0x8045d004, + 0x040040b7, + 0xb6010598, + 0x45d01054, + 0x0040b700, + 0x0057f105, + 0x0045d003, + 0x111057f1, + 0x131253f1, + 0x984045d0, + 0x40b70305, + 0x45d00500, + 0x0157f100, + 0x0153f026, + 0x080047f1, + 0xd00644b6, + 0x00f80045, +/* 0x04b8: cmd_exec */ + 0x03d121f5, + 0xf4003fc8, + 0x21f50e0b, + 0x47f101af, + 0x0ef40200, +/* 0x04cd: cmd_exec_no_format */ + 0x1067f11e, + 0x0664b608, + 0x800177f0, + 0x07800e07, + 0x1d079819, + 0xd00067d0, + 0x44bd4067, +/* 0x04e8: cmd_exec_init_src_surface */ + 0xbd0232f4, + 0x043fc854, + 0xf50a0bf4, + 0xf403a821, +/* 0x04fa: src_tiled */ + 0x21f50a0e, + 0x49f0029c, +/* 0x0501: cmd_exec_init_dst_surface */ + 0x0231f407, + 0xc82c57f0, + 0x0bf4083f, + 0xa821f50a, + 0x0a0ef403, +/* 0x0514: dst_tiled */ + 0x029c21f5, +/* 0x051b: cmd_exec_kick */ + 0xf10849f0, + 0xb6080057, + 0x06980654, + 0x4056d01e, + 0xf14167f0, + 0xfd440063, + 0x54d00546, + 0x0c3fc800, + 0xf5070bf4, +/* 0x053f: cmd_exec_done */ + 0xf803eb21, +/* 0x0541: cmd_wrcache_flush */ + 0x0027f100, + 0xf034bd22, + 0x23d00133, + 0x0000f800, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3 new file mode 100644 index 0000000000..07bda93cfd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3 @@ -0,0 +1,2 @@ +#define GT215 +#include "com.fuc" diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h new file mode 100644 index 0000000000..d3fbd4ab5e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/fuc/gt215.fuc3.h @@ -0,0 +1,621 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gt215_ce_data[] = { +/* 0x0000: ctx_object */ + 0x00000000, +/* 0x0004: ctx_dma */ +/* 0x0004: ctx_dma_query */ + 0x00000000, +/* 0x0008: ctx_dma_src */ + 0x00000000, +/* 0x000c: ctx_dma_dst */ + 0x00000000, +/* 0x0010: ctx_query_address_high */ + 0x00000000, +/* 0x0014: ctx_query_address_low */ + 0x00000000, +/* 0x0018: ctx_query_counter */ + 0x00000000, +/* 0x001c: ctx_src_address_high */ + 0x00000000, +/* 0x0020: ctx_src_address_low */ + 0x00000000, +/* 0x0024: ctx_src_pitch */ + 0x00000000, +/* 0x0028: ctx_src_tile_mode */ + 0x00000000, +/* 0x002c: ctx_src_xsize */ + 0x00000000, +/* 0x0030: ctx_src_ysize */ + 0x00000000, +/* 0x0034: ctx_src_zsize */ + 0x00000000, +/* 0x0038: ctx_src_zoff */ + 0x00000000, +/* 0x003c: ctx_src_xoff */ + 0x00000000, +/* 0x0040: ctx_src_yoff */ + 0x00000000, +/* 0x0044: ctx_src_cpp */ + 0x00000000, +/* 0x0048: ctx_dst_address_high */ + 0x00000000, +/* 0x004c: ctx_dst_address_low */ + 0x00000000, +/* 0x0050: ctx_dst_pitch */ + 0x00000000, +/* 0x0054: ctx_dst_tile_mode */ + 0x00000000, +/* 0x0058: ctx_dst_xsize */ + 0x00000000, +/* 0x005c: ctx_dst_ysize */ + 0x00000000, +/* 0x0060: ctx_dst_zsize */ + 0x00000000, +/* 0x0064: ctx_dst_zoff */ + 0x00000000, +/* 0x0068: ctx_dst_xoff */ + 0x00000000, +/* 0x006c: ctx_dst_yoff */ + 0x00000000, +/* 0x0070: ctx_dst_cpp */ + 0x00000000, +/* 0x0074: ctx_format */ + 0x00000000, +/* 0x0078: ctx_swz_const0 */ + 0x00000000, +/* 0x007c: ctx_swz_const1 */ + 0x00000000, +/* 0x0080: ctx_xcnt */ + 0x00000000, +/* 0x0084: ctx_ycnt */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: dispatch_table */ + 0x00010000, + 0x00000000, + 0x00000000, + 0x00010040, + 0x00010160, + 0x00000000, + 0x00010050, + 0x00010162, + 0x00000000, + 0x00030060, +/* 0x0128: dispatch_dma */ + 0x00010170, + 0x00000000, + 0x00010170, + 0x00000000, + 0x00010170, + 0x00000000, + 0x00070080, + 0x00000028, + 0xfffff000, + 0x0000002c, + 0xfff80000, + 0x00000030, + 0xffffe000, + 0x00000034, + 0xfffff800, + 0x00000038, + 0xfffff000, + 0x0000003c, + 0xfff80000, + 0x00000040, + 0xffffe000, + 0x00070088, + 0x00000054, + 0xfffff000, + 0x00000058, + 0xfff80000, + 0x0000005c, + 0xffffe000, + 0x00000060, + 0xfffff800, + 0x00000064, + 0xfffff000, + 0x00000068, + 0xfff80000, + 0x0000006c, + 0xffffe000, + 0x000200c0, + 0x00010492, + 0x00000000, + 0x0001051b, + 0x00000000, + 0x000e00c3, + 0x0000001c, + 0xffffff00, + 0x00000020, + 0x00000000, + 0x00000048, + 0xffffff00, + 0x0000004c, + 0x00000000, + 0x00000024, + 0xfff80000, + 0x00000050, + 0xfff80000, + 0x00000080, + 0xffff0000, + 0x00000084, + 0xffffe000, + 0x00000074, + 0xfccc0000, + 0x00000078, + 0x00000000, + 0x0000007c, + 0x00000000, + 0x00000010, + 0xffffff00, + 0x00000014, + 0x00000000, + 0x00000018, + 0x00000000, + 0x00000800, +}; + +static uint32_t gt215_ce_code[] = { +/* 0x0000: main */ + 0x04fe04bd, + 0x3517f000, + 0xf10010fe, + 0xf1040017, + 0xf0fff327, + 0x12d00023, + 0x0c25f0c0, + 0xf40012d0, + 0x17f11031, + 0x27f01200, + 0x0012d003, +/* 0x002f: spin */ + 0xf40031f4, + 0x0ef40028, +/* 0x0035: ih */ + 0x8001cffd, + 0xf40812c4, + 0x21f4060b, +/* 0x0041: ih_no_chsw */ + 0x0412c472, + 0xf4060bf4, +/* 0x004a: ih_no_cmd */ + 0x11c4c321, + 0x4001d00c, +/* 0x0052: swctx */ + 0x47f101f8, + 0x4bfe7700, + 0x0007fe00, + 0xf00204b9, + 0x01f40643, + 0x0604fa09, +/* 0x006b: swctx_load */ + 0xfa060ef4, +/* 0x006e: swctx_done */ + 0x03f80504, +/* 0x0072: chsw */ + 0x27f100f8, + 0x23cf1400, + 0x1e3fc800, + 0xf4170bf4, + 0x21f40132, + 0x1e3af052, + 0xf00023d0, + 0x24d00147, +/* 0x0093: chsw_no_unload */ + 0xcf00f880, + 0x3dc84023, + 0x220bf41e, + 0xf40131f4, + 0x57f05221, + 0x0367f004, +/* 0x00a8: chsw_load_ctx_dma */ + 0xa07856bc, + 0xb6018068, + 0x87d00884, + 0x0162b600, +/* 0x00bb: chsw_finish_load */ + 0xf0f018f4, + 0x23d00237, +/* 0x00c3: dispatch */ + 0xf100f880, + 0xcf190037, + 0x33cf4032, + 0xff24e400, + 0x1024b607, + 0x010057f1, + 0x74bd64bd, +/* 0x00dc: dispatch_loop */ + 0x58005658, + 0x50b60157, + 0x0446b804, + 0xbb4d08f4, + 0x47b80076, + 0x0f08f404, + 0xb60276bb, + 0x57bb0374, + 0xdf0ef400, +/* 0x0100: dispatch_valid_mthd */ + 0xb60246bb, + 0x45bb0344, + 0x01459800, + 0xb00453fd, + 0x1bf40054, + 0x00455820, + 0xb0014658, + 0x1bf40064, + 0x00538009, +/* 0x0127: dispatch_cmd */ + 0xf4300ef4, + 0x55f90132, + 0xf40c01f4, +/* 0x0132: dispatch_invalid_bitfield */ + 0x25f0250e, +/* 0x0135: dispatch_illegal_mthd */ + 0x0125f002, +/* 0x0138: dispatch_error */ + 0x100047f1, + 0xd00042d0, + 0x27f04043, + 0x0002d040, +/* 0x0148: hostirq_wait */ + 0xf08002cf, + 0x24b04024, + 0xf71bf400, +/* 0x0154: dispatch_done */ + 0x1d0027f1, + 0xd00137f0, + 0x00f80023, +/* 0x0160: cmd_nop */ +/* 0x0162: cmd_pm_trigger */ + 0x27f100f8, + 0x34bd2200, + 0xd00233f0, + 0x00f80023, +/* 0x0170: cmd_dma */ + 0x012842b7, + 0xf00145b6, + 0x43801e39, + 0x0040b701, + 0x0644b606, + 0xf80043d0, +/* 0x0189: cmd_exec_set_format */ + 0xf030f400, + 0xb00001b0, + 0x01b00101, + 0x0301b002, + 0xc71d0498, + 0x50b63045, + 0x3446c701, + 0xc70160b6, + 0x70b63847, + 0x0232f401, + 0x94bd84bd, +/* 0x01b4: ncomp_loop */ + 0xb60f4ac4, + 0xb4bd0445, +/* 0x01bc: bpc_loop */ + 0xf404a430, + 0xa5ff0f18, + 0x00cbbbc0, + 0xf40231f4, +/* 0x01ce: cmp_c0 */ + 0x1bf4220e, + 0x10c7f00c, + 0xf400cbbb, +/* 0x01da: cmp_c1 */ + 0xa430160e, + 0x0c18f406, + 0xbb14c7f0, + 0x0ef400cb, +/* 0x01e9: cmp_zero */ + 0x80c7f107, +/* 0x01ed: bpc_next */ + 0x01c83800, + 0xb60180b6, + 0xb5b801b0, + 0xc308f404, + 0xb80190b6, + 0x08f40497, + 0x0065fdb2, + 0x98110680, + 0x68fd2008, + 0x0502f400, +/* 0x0216: dst_xcnt */ + 0x75fd64bd, + 0x1c078000, + 0xf10078fd, + 0xb6081057, + 0x56d00654, + 0x4057d000, + 0x080050b7, + 0xb61c0698, + 0x64b60162, + 0x11079808, + 0xfd0172b6, + 0x56d00567, + 0x0050b700, + 0x0060b401, + 0xb40056d0, + 0x56d00160, + 0x0260b440, + 0xb48056d0, + 0x56d00360, + 0x0050b7c0, + 0x1e069804, + 0x980056d0, + 0x56d01f06, + 0x1030f440, +/* 0x0276: cmd_exec_set_surface_tiled */ + 0x579800f8, + 0x6879c70a, + 0xb66478c7, + 0x77c70280, + 0x0e76b060, + 0xf0091bf4, + 0x0ef40477, +/* 0x0291: xtile64 */ + 0x027cf00f, + 0xfd1170b6, + 0x77f00947, +/* 0x029d: xtileok */ + 0x0f5a9806, + 0xfd115b98, + 0xb7f000ab, + 0x04b7bb01, + 0xff01b2b6, + 0xa7bbc4ab, + 0x105d9805, + 0xbb01e7f0, + 0xe2b604e8, + 0xb4deff01, + 0xb605d8bb, + 0xef9401e0, + 0x02ebbb0c, + 0xf005fefd, + 0x60b7026c, + 0x64b60208, + 0x006fd008, + 0xbb04b7bb, + 0x5f9800cb, + 0x115b980b, + 0xf000fbfd, + 0xb7bb01b7, + 0x01b2b604, + 0xbb00fbbb, + 0xf0f905f7, + 0xf00c5f98, + 0xb8bb01b7, + 0x01b2b604, + 0xbb00fbbb, + 0xf0f905f8, + 0xb60078bb, + 0xb7f00282, + 0x04b8bb01, + 0x9804b9bb, + 0xe7f00e58, + 0x04e9bb01, + 0xff01e2b6, + 0xf7bbf48e, + 0x00cfbb04, + 0xbb0079bb, + 0xf0fc0589, + 0xd9fd90fc, + 0x00adbb00, + 0xfd0089fd, + 0xa8bb008f, + 0x04a7bb00, + 0xbb0192b6, + 0x69d00497, + 0x08579880, + 0xbb075898, + 0x7abb00ac, + 0x0081b600, + 0xfd1084b6, + 0x62b7058b, + 0x67d00600, + 0x0060b700, + 0x0068d004, +/* 0x0382: cmd_exec_set_surface_linear */ + 0x6cf000f8, + 0x0260b702, + 0x0864b602, + 0xd0085798, + 0x60b70067, + 0x57980400, + 0x1074b607, + 0xb70067d0, + 0x98040060, + 0x67d00957, +/* 0x03ab: cmd_exec_wait */ + 0xf900f800, + 0xf110f900, + 0xb6080007, +/* 0x03b6: loop */ + 0x01cf0604, + 0x0114f000, + 0xfcfa1bf4, + 0xf800fc10, +/* 0x03c5: cmd_exec_query */ + 0x0d34c800, + 0xf5701bf4, + 0xf103ab21, + 0xb6080c47, + 0x05980644, + 0x0450b605, + 0xd00045d0, + 0x57f04040, + 0x8045d00c, + 0x040040b7, + 0xb6040598, + 0x45d01054, + 0x0040b700, + 0x0057f105, + 0x0153f00b, + 0xf10045d0, + 0xb6404057, + 0x53f10154, + 0x45d08080, + 0x1057f140, + 0x1253f111, + 0x8045d013, + 0x151457f1, + 0x171653f1, + 0xf1c045d0, + 0xf0260157, + 0x47f10153, + 0x44b60800, + 0x0045d006, +/* 0x0438: query_counter */ + 0x03ab21f5, + 0x080c47f1, + 0x980644b6, + 0x45d00505, + 0x4040d000, + 0xd00457f0, + 0x40b78045, + 0x05980400, + 0x1054b604, + 0xb70045d0, + 0xf1050040, + 0xd0030057, + 0x57f10045, + 0x53f11110, + 0x45d01312, + 0x06059840, + 0x050040b7, + 0xf10045d0, + 0xf0260157, + 0x47f10153, + 0x44b60800, + 0x0045d006, +/* 0x0492: cmd_exec */ + 0x21f500f8, + 0x3fc803ab, + 0x0e0bf400, + 0x018921f5, + 0x020047f1, +/* 0x04a7: cmd_exec_no_format */ + 0xf11e0ef4, + 0xb6081067, + 0x77f00664, + 0x11078001, + 0x981c0780, + 0x67d02007, + 0x4067d000, +/* 0x04c2: cmd_exec_init_src_surface */ + 0x32f444bd, + 0xc854bd02, + 0x0bf4043f, + 0x8221f50a, + 0x0a0ef403, +/* 0x04d4: src_tiled */ + 0x027621f5, +/* 0x04db: cmd_exec_init_dst_surface */ + 0xf40749f0, + 0x57f00231, + 0x083fc82c, + 0xf50a0bf4, + 0xf4038221, +/* 0x04ee: dst_tiled */ + 0x21f50a0e, + 0x49f00276, +/* 0x04f5: cmd_exec_kick */ + 0x0057f108, + 0x0654b608, + 0xd0210698, + 0x67f04056, + 0x0063f141, + 0x0546fd44, + 0xc80054d0, + 0x0bf40c3f, + 0xc521f507, +/* 0x0519: cmd_exec_done */ +/* 0x051b: cmd_wrcache_flush */ + 0xf100f803, + 0xbd220027, + 0x0133f034, + 0xf80023d0, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga100.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga100.c new file mode 100644 index 0000000000..315a69f7fd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga100.c @@ -0,0 +1,92 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +#include + +static irqreturn_t +ga100_ce_intr(struct nvkm_inth *inth) +{ + struct nvkm_subdev *subdev = container_of(inth, typeof(*subdev), inth); + + /*TODO*/ + nvkm_error(subdev, "intr\n"); + return IRQ_NONE; +} + +int +ga100_ce_nonstall(struct nvkm_engine *engine) +{ + struct nvkm_subdev *subdev = &engine->subdev; + struct nvkm_device *device = subdev->device; + + return nvkm_rd32(device, 0x104424 + (subdev->inst * 0x80)) & 0x00000fff; +} + +int +ga100_ce_fini(struct nvkm_engine *engine, bool suspend) +{ + nvkm_inth_block(&engine->subdev.inth); + return 0; +} + +int +ga100_ce_init(struct nvkm_engine *engine) +{ + nvkm_inth_allow(&engine->subdev.inth); + return 0; +} + +int +ga100_ce_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_subdev *subdev = &engine->subdev; + struct nvkm_device *device = subdev->device; + u32 vector; + + vector = nvkm_rd32(device, 0x10442c + (subdev->inst * 0x80)) & 0x00000fff; + + return nvkm_inth_add(&device->vfn->intr, vector, NVKM_INTR_PRIO_NORMAL, + subdev, ga100_ce_intr, &subdev->inth); +} + +static const struct nvkm_engine_func +ga100_ce = { + .oneinit = ga100_ce_oneinit, + .init = ga100_ce_init, + .fini = ga100_ce_fini, + .nonstall = ga100_ce_nonstall, + .cclass = &gv100_ce_cclass, + .sclass = { + { -1, -1, AMPERE_DMA_COPY_A }, + {} + } +}; + +int +ga100_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&ga100_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga102.c new file mode 100644 index 0000000000..461b73c7e2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/ga102.c @@ -0,0 +1,45 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +static const struct nvkm_engine_func +ga102_ce = { + .oneinit = ga100_ce_oneinit, + .init = ga100_ce_init, + .fini = ga100_ce_fini, + .nonstall = ga100_ce_nonstall, + .cclass = &gv100_ce_cclass, + .sclass = { + { -1, -1, AMPERE_DMA_COPY_A }, + { -1, -1, AMPERE_DMA_COPY_B }, + {} + } +}; + +int +ga102_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&ga102_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gf100.c new file mode 100644 index 0000000000..b9cc395659 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gf100.c @@ -0,0 +1,69 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "fuc/gf100.fuc3.h" + +#include + +static void +gf100_ce_init(struct nvkm_falcon *ce) +{ + nvkm_wr32(ce->engine.subdev.device, ce->addr + 0x084, ce->engine.subdev.inst); +} + +static const struct nvkm_falcon_func +gf100_ce0 = { + .code.data = gf100_ce_code, + .code.size = sizeof(gf100_ce_code), + .data.data = gf100_ce_data, + .data.size = sizeof(gf100_ce_data), + .init = gf100_ce_init, + .intr = gt215_ce_intr, + .sclass = { + { -1, -1, FERMI_DMA }, + {} + } +}; + +static const struct nvkm_falcon_func +gf100_ce1 = { + .code.data = gf100_ce_code, + .code.size = sizeof(gf100_ce_code), + .data.data = gf100_ce_data, + .data.size = sizeof(gf100_ce_data), + .init = gf100_ce_init, + .intr = gt215_ce_intr, + .sclass = { + { -1, -1, FERMI_DECOMPRESS }, + {} + } +}; + +int +gf100_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(inst ? &gf100_ce1 : &gf100_ce0, device, type, inst, true, + 0x104000 + (inst * 0x1000), pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gk104.c new file mode 100644 index 0000000000..27f29eb049 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gk104.c @@ -0,0 +1,101 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include + +#include + +static const struct nvkm_enum +gk104_ce_launcherr_report[] = { + { 0x0, "NO_ERR" }, + { 0x1, "2D_LAYER_EXCEEDS_DEPTH" }, + { 0x2, "INVALID_ARGUMENT" }, + { 0x3, "MEM2MEM_RECT_OUT_OF_BOUNDS" }, + { 0x4, "SRC_LINE_EXCEEDS_PITCH" }, + { 0x5, "SRC_LINE_EXCEEDS_NEG_PITCH" }, + { 0x6, "DST_LINE_EXCEEDS_PITCH" }, + { 0x7, "DST_LINE_EXCEEDS_NEG_PITCH" }, + { 0x8, "BAD_SRC_PIXEL_COMP_REF" }, + { 0x9, "INVALID_VALUE" }, + { 0xa, "UNUSED_FIELD" }, + { 0xb, "INVALID_OPERATION" }, + {} +}; + +static void +gk104_ce_intr_launcherr(struct nvkm_engine *ce, const u32 base) +{ + struct nvkm_subdev *subdev = &ce->subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x104f14 + base); + const struct nvkm_enum *en = + nvkm_enum_find(gk104_ce_launcherr_report, stat & 0x0000000f); + nvkm_warn(subdev, "LAUNCHERR %08x [%s]\n", stat, en ? en->name : ""); + nvkm_wr32(device, 0x104f14 + base, 0x00000000); +} + +void +gk104_ce_intr(struct nvkm_engine *ce) +{ + struct nvkm_subdev *subdev = &ce->subdev; + struct nvkm_device *device = subdev->device; + const u32 base = subdev->inst * 0x1000; + u32 mask = nvkm_rd32(device, 0x104904 + base); + u32 intr = nvkm_rd32(device, 0x104908 + base) & mask; + if (intr & 0x00000001) { + nvkm_warn(subdev, "BLOCKPIPE\n"); + nvkm_wr32(device, 0x104908 + base, 0x00000001); + intr &= ~0x00000001; + } + if (intr & 0x00000002) { + nvkm_warn(subdev, "NONBLOCKPIPE\n"); + nvkm_wr32(device, 0x104908 + base, 0x00000002); + intr &= ~0x00000002; + } + if (intr & 0x00000004) { + gk104_ce_intr_launcherr(ce, base); + nvkm_wr32(device, 0x104908 + base, 0x00000004); + intr &= ~0x00000004; + } + if (intr) { + nvkm_warn(subdev, "intr %08x\n", intr); + nvkm_wr32(device, 0x104908 + base, intr); + } +} + +static const struct nvkm_engine_func +gk104_ce = { + .intr = gk104_ce_intr, + .sclass = { + { -1, -1, KEPLER_DMA_COPY_A }, + {} + } +}; + +int +gk104_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gk104_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm107.c new file mode 100644 index 0000000000..c3c476592c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm107.c @@ -0,0 +1,43 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +static const struct nvkm_engine_func +gm107_ce = { + .intr = gk104_ce_intr, + .sclass = { + { -1, -1, KEPLER_DMA_COPY_A }, + { -1, -1, MAXWELL_DMA_COPY_A }, + {} + } +}; + +int +gm107_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gm107_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm200.c new file mode 100644 index 0000000000..d2db618653 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gm200.c @@ -0,0 +1,42 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +static const struct nvkm_engine_func +gm200_ce = { + .intr = gk104_ce_intr, + .sclass = { + { -1, -1, MAXWELL_DMA_COPY_A }, + {} + } +}; + +int +gm200_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gm200_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp100.c new file mode 100644 index 0000000000..a4f08a4472 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp100.c @@ -0,0 +1,102 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include + +#include + +static const struct nvkm_enum +gp100_ce_launcherr_report[] = { + { 0x0, "NO_ERR" }, + { 0x1, "2D_LAYER_EXCEEDS_DEPTH" }, + { 0x2, "INVALID_ALIGNMENT" }, + { 0x3, "MEM2MEM_RECT_OUT_OF_BOUNDS" }, + { 0x4, "SRC_LINE_EXCEEDS_PITCH" }, + { 0x5, "SRC_LINE_EXCEEDS_NEG_PITCH" }, + { 0x6, "DST_LINE_EXCEEDS_PITCH" }, + { 0x7, "DST_LINE_EXCEEDS_NEG_PITCH" }, + { 0x8, "BAD_SRC_PIXEL_COMP_REF" }, + { 0x9, "INVALID_VALUE" }, + { 0xa, "UNUSED_FIELD" }, + { 0xb, "INVALID_OPERATION" }, + { 0xc, "NO_RESOURCES" }, + { 0xd, "INVALID_CONFIG" }, + {} +}; + +static void +gp100_ce_intr_launcherr(struct nvkm_engine *ce, const u32 base) +{ + struct nvkm_subdev *subdev = &ce->subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x104418 + base); + const struct nvkm_enum *en = + nvkm_enum_find(gp100_ce_launcherr_report, stat & 0x0000000f); + nvkm_warn(subdev, "LAUNCHERR %08x [%s]\n", stat, en ? en->name : ""); +} + +void +gp100_ce_intr(struct nvkm_engine *ce) +{ + struct nvkm_subdev *subdev = &ce->subdev; + struct nvkm_device *device = subdev->device; + const u32 base = subdev->inst * 0x80; + u32 mask = nvkm_rd32(device, 0x10440c + base); + u32 intr = nvkm_rd32(device, 0x104410 + base) & mask; + if (intr & 0x00000001) { //XXX: guess + nvkm_warn(subdev, "BLOCKPIPE\n"); + nvkm_wr32(device, 0x104410 + base, 0x00000001); + intr &= ~0x00000001; + } + if (intr & 0x00000002) { //XXX: guess + nvkm_warn(subdev, "NONBLOCKPIPE\n"); + nvkm_wr32(device, 0x104410 + base, 0x00000002); + intr &= ~0x00000002; + } + if (intr & 0x00000004) { + gp100_ce_intr_launcherr(ce, base); + nvkm_wr32(device, 0x104410 + base, 0x00000004); + intr &= ~0x00000004; + } + if (intr) { + nvkm_warn(subdev, "intr %08x\n", intr); + nvkm_wr32(device, 0x104410 + base, intr); + } +} + +static const struct nvkm_engine_func +gp100_ce = { + .intr = gp100_ce_intr, + .sclass = { + { -1, -1, PASCAL_DMA_COPY_A }, + {} + } +}; + +int +gp100_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gp100_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c new file mode 100644 index 0000000000..180d497a95 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gp102.c @@ -0,0 +1,44 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include + +#include + +static const struct nvkm_engine_func +gp102_ce = { + .intr = gp100_ce_intr, + .sclass = { + { -1, -1, PASCAL_DMA_COPY_B }, + { -1, -1, PASCAL_DMA_COPY_A }, + {} + } +}; + +int +gp102_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gp102_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gt215.c new file mode 100644 index 0000000000..c9bf6305c3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gt215.c @@ -0,0 +1,83 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "fuc/gt215.fuc3.h" + +#include +#include +#include +#include + +#include + +static const struct nvkm_enum +gt215_ce_isr_error_name[] = { + { 0x0001, "ILLEGAL_MTHD" }, + { 0x0002, "INVALID_ENUM" }, + { 0x0003, "INVALID_BITFIELD" }, + {} +}; + +void +gt215_ce_intr(struct nvkm_falcon *ce, struct nvkm_chan *chan) +{ + struct nvkm_subdev *subdev = &ce->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 base = subdev->inst * 0x1000; + u32 ssta = nvkm_rd32(device, 0x104040 + base) & 0x0000ffff; + u32 addr = nvkm_rd32(device, 0x104040 + base) >> 16; + u32 mthd = (addr & 0x07ff) << 2; + u32 subc = (addr & 0x3800) >> 11; + u32 data = nvkm_rd32(device, 0x104044 + base); + const struct nvkm_enum *en = + nvkm_enum_find(gt215_ce_isr_error_name, ssta); + + nvkm_error(subdev, "DISPATCH_ERROR %04x [%s] ch %d [%010llx %s] " + "subc %d mthd %04x data %08x\n", ssta, + en ? en->name : "", chan ? chan->id : -1, + chan ? chan->inst->addr : 0, + chan ? chan->name : "unknown", + subc, mthd, data); +} + +static const struct nvkm_falcon_func +gt215_ce = { + .code.data = gt215_ce_code, + .code.size = sizeof(gt215_ce_code), + .data.data = gt215_ce_data, + .data.size = sizeof(gt215_ce_data), + .intr = gt215_ce_intr, + .sclass = { + { -1, -1, GT212_DMA }, + {} + } +}; + +int +gt215_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(>215_ce, device, type, -1, + (device->chipset != 0xaf), 0x104000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/gv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gv100.c new file mode 100644 index 0000000000..44021d1395 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/gv100.c @@ -0,0 +1,64 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include +#include + +#include + +static int +gv100_ce_cclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, int align, + struct nvkm_gpuobj **pgpuobj) +{ + struct nvkm_device *device = object->engine->subdev.device; + u32 size; + + /* Allocate fault method buffer (magics come from nvgpu). */ + size = nvkm_rd32(device, 0x104028); /* NV_PCE_PCE_MAP */ + size = 27 * 5 * (((9 + 1 + 3) * hweight32(size)) + 2); + size = roundup(size, PAGE_SIZE); + + return nvkm_gpuobj_new(device, size, align, true, parent, pgpuobj); +} + +const struct nvkm_object_func +gv100_ce_cclass = { + .bind = gv100_ce_cclass_bind, +}; + +static const struct nvkm_engine_func +gv100_ce = { + .intr = gp100_ce_intr, + .cclass = &gv100_ce_cclass, + .sclass = { + { -1, -1, VOLTA_DMA_COPY_A }, + {} + } +}; + +int +gv100_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&gv100_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/ce/priv.h new file mode 100644 index 0000000000..0be72c463b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/priv.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_CE_PRIV_H__ +#define __NVKM_CE_PRIV_H__ +#include + +void gt215_ce_intr(struct nvkm_falcon *, struct nvkm_chan *); +void gk104_ce_intr(struct nvkm_engine *); +void gp100_ce_intr(struct nvkm_engine *); + +extern const struct nvkm_object_func gv100_ce_cclass; + +int ga100_ce_oneinit(struct nvkm_engine *); +int ga100_ce_init(struct nvkm_engine *); +int ga100_ce_fini(struct nvkm_engine *, bool); +int ga100_ce_nonstall(struct nvkm_engine *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/ce/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/ce/tu102.c new file mode 100644 index 0000000000..9563c01751 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/ce/tu102.c @@ -0,0 +1,41 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +static const struct nvkm_engine_func +tu102_ce = { + .intr = gp100_ce_intr, + .cclass = &gv100_ce_cclass, + .sclass = { + { -1, -1, TURING_DMA_COPY_A }, + {} + } +}; + +int +tu102_ce_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&tu102_ce, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/cipher/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/cipher/Kbuild new file mode 100644 index 0000000000..ffad341f42 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/cipher/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/cipher/g84.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/cipher/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/cipher/g84.c new file mode 100644 index 0000000000..caca4f6398 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/cipher/g84.c @@ -0,0 +1,133 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include +#include + +#include +#include +#include + +#include + +static int +g84_cipher_oclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, + align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; +} + +static const struct nvkm_object_func +g84_cipher_oclass_func = { + .bind = g84_cipher_oclass_bind, +}; + +static int +g84_cipher_cclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + return nvkm_gpuobj_new(object->engine->subdev.device, 256, + align, true, parent, pgpuobj); + +} + +static const struct nvkm_object_func +g84_cipher_cclass = { + .bind = g84_cipher_cclass_bind, +}; + +static const struct nvkm_bitfield +g84_cipher_intr_mask[] = { + { 0x00000001, "INVALID_STATE" }, + { 0x00000002, "ILLEGAL_MTHD" }, + { 0x00000004, "ILLEGAL_CLASS" }, + { 0x00000080, "QUERY" }, + { 0x00000100, "FAULT" }, + {} +}; + +static void +g84_cipher_intr(struct nvkm_engine *cipher) +{ + struct nvkm_subdev *subdev = &cipher->subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_chan *chan; + u32 stat = nvkm_rd32(device, 0x102130); + u32 mthd = nvkm_rd32(device, 0x102190); + u32 data = nvkm_rd32(device, 0x102194); + u32 inst = nvkm_rd32(device, 0x102188) & 0x7fffffff; + unsigned long flags; + char msg[128]; + + chan = nvkm_chan_get_inst(cipher, (u64)inst << 12, &flags); + if (stat) { + nvkm_snprintbf(msg, sizeof(msg), g84_cipher_intr_mask, stat); + nvkm_error(subdev, "%08x [%s] ch %d [%010llx %s] " + "mthd %04x data %08x\n", stat, msg, + chan ? chan->id : -1, (u64)inst << 12, + chan ? chan->name : "unknown", + mthd, data); + } + nvkm_chan_put(&chan, flags); + + nvkm_wr32(device, 0x102130, stat); + nvkm_wr32(device, 0x10200c, 0x10); +} + +static int +g84_cipher_init(struct nvkm_engine *cipher) +{ + struct nvkm_device *device = cipher->subdev.device; + nvkm_wr32(device, 0x102130, 0xffffffff); + nvkm_wr32(device, 0x102140, 0xffffffbf); + nvkm_wr32(device, 0x10200c, 0x00000010); + return 0; +} + +static const struct nvkm_engine_func +g84_cipher = { + .init = g84_cipher_init, + .intr = g84_cipher_intr, + .cclass = &g84_cipher_cclass, + .sclass = { + { -1, -1, NV74_CIPHER, &g84_cipher_oclass_func }, + {} + } +}; + +int +g84_cipher_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_engine_new_(&g84_cipher, device, type, inst, true, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/device/Kbuild new file mode 100644 index 0000000000..293c57678d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/Kbuild @@ -0,0 +1,7 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/device/acpi.o +nvkm-y += nvkm/engine/device/base.o +nvkm-y += nvkm/engine/device/ctrl.o +nvkm-y += nvkm/engine/device/pci.o +nvkm-y += nvkm/engine/device/tegra.o +nvkm-y += nvkm/engine/device/user.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c new file mode 100644 index 0000000000..c948a0dc9e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.c @@ -0,0 +1,58 @@ +/* + * Copyright 2014 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "acpi.h" + +#include +#include + +#ifdef CONFIG_ACPI +static int +nvkm_acpi_ntfy(struct notifier_block *nb, unsigned long val, void *data) +{ + struct nvkm_device *device = container_of(nb, typeof(*device), acpi.nb); + struct acpi_bus_event *info = data; + + if (!strcmp(info->device_class, "ac_adapter")) + nvkm_clk_pwrsrc(device); + + return NOTIFY_DONE; +} +#endif + +void +nvkm_acpi_fini(struct nvkm_device *device) +{ +#ifdef CONFIG_ACPI + unregister_acpi_notifier(&device->acpi.nb); +#endif +} + +void +nvkm_acpi_init(struct nvkm_device *device) +{ +#ifdef CONFIG_ACPI + device->acpi.nb.notifier_call = nvkm_acpi_ntfy; + register_acpi_notifier(&device->acpi.nb); +#endif +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.h b/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.h new file mode 100644 index 0000000000..1d3c5cf7c3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/acpi.h @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DEVICE_ACPI_H__ +#define __NVKM_DEVICE_ACPI_H__ +#include +struct nvkm_device; + +void nvkm_acpi_init(struct nvkm_device *); +void nvkm_acpi_fini(struct nvkm_device *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c new file mode 100644 index 0000000000..1c81e5b34d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c @@ -0,0 +1,3268 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "acpi.h" + +#include + +#include +#include + +static DEFINE_MUTEX(nv_devices_mutex); +static LIST_HEAD(nv_devices); + +static struct nvkm_device * +nvkm_device_find_locked(u64 handle) +{ + struct nvkm_device *device; + list_for_each_entry(device, &nv_devices, head) { + if (device->handle == handle) + return device; + } + return NULL; +} + +struct nvkm_device * +nvkm_device_find(u64 handle) +{ + struct nvkm_device *device; + mutex_lock(&nv_devices_mutex); + device = nvkm_device_find_locked(handle); + mutex_unlock(&nv_devices_mutex); + return device; +} + +int +nvkm_device_list(u64 *name, int size) +{ + struct nvkm_device *device; + int nr = 0; + mutex_lock(&nv_devices_mutex); + list_for_each_entry(device, &nv_devices, head) { + if (nr++ < size) + name[nr - 1] = device->handle; + } + mutex_unlock(&nv_devices_mutex); + return nr; +} + +static const struct nvkm_device_chip +null_chipset = { + .name = "NULL", + .bios = { 0x00000001, nvkm_bios_new }, +}; + +static const struct nvkm_device_chip +nv4_chipset = { + .name = "NV04", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv04_devinit_new }, + .fb = { 0x00000001, nv04_fb_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv04_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv04_fifo_new }, + .gr = { 0x00000001, nv04_gr_new }, + .sw = { 0x00000001, nv04_sw_new }, +}; + +static const struct nvkm_device_chip +nv5_chipset = { + .name = "NV05", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv05_devinit_new }, + .fb = { 0x00000001, nv04_fb_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv04_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv04_fifo_new }, + .gr = { 0x00000001, nv04_gr_new }, + .sw = { 0x00000001, nv04_sw_new }, +}; + +static const struct nvkm_device_chip +nv10_chipset = { + .name = "NV10", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv04_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .gr = { 0x00000001, nv10_gr_new }, +}; + +static const struct nvkm_device_chip +nv11_chipset = { + .name = "NV11", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv11_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv10_fifo_new }, + .gr = { 0x00000001, nv15_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv15_chipset = { + .name = "NV15", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv04_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv10_fifo_new }, + .gr = { 0x00000001, nv15_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv17_chipset = { + .name = "NV17", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv17_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv18_chipset = { + .name = "NV18", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv17_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv1a_chipset = { + .name = "nForce", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv1a_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv04_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv10_fifo_new }, + .gr = { 0x00000001, nv15_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv1f_chipset = { + .name = "nForce2", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv1a_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv17_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv20_chipset = { + .name = "NV20", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv20_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv20_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv25_chipset = { + .name = "NV25", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv25_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv25_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv28_chipset = { + .name = "NV28", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv25_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv25_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv2a_chipset = { + .name = "NV2A", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv25_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv2a_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv30_chipset = { + .name = "NV30", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv30_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv30_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv31_chipset = { + .name = "NV31", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv30_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv30_gr_new }, + .mpeg = { 0x00000001, nv31_mpeg_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv34_chipset = { + .name = "NV34", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv10_devinit_new }, + .fb = { 0x00000001, nv10_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv34_gr_new }, + .mpeg = { 0x00000001, nv31_mpeg_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv35_chipset = { + .name = "NV35", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv04_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv35_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv35_gr_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv36_chipset = { + .name = "NV36", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv04_clk_new }, + .devinit = { 0x00000001, nv20_devinit_new }, + .fb = { 0x00000001, nv36_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv04_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv04_pci_new }, + .timer = { 0x00000001, nv04_timer_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv17_fifo_new }, + .gr = { 0x00000001, nv35_gr_new }, + .mpeg = { 0x00000001, nv31_mpeg_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv40_chipset = { + .name = "NV40", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv40_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv40_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv40_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv41_chipset = { + .name = "NV41", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv41_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv40_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv42_chipset = { + .name = "NV42", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv41_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv40_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv43_chipset = { + .name = "NV43", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv41_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv40_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv44_chipset = { + .name = "NV44", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv44_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv45_chipset = { + .name = "NV45", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv40_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv46_chipset = { + .name = "G72", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv46_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv46_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv47_chipset = { + .name = "G70", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv47_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv49_chipset = { + .name = "G71", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv49_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv4a_chipset = { + .name = "NV44A", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv44_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv04_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv4b_chipset = { + .name = "G73", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv49_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv17_mc_new }, + .mmu = { 0x00000001, nv41_mmu_new }, + .pci = { 0x00000001, nv40_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv40_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv4c_chipset = { + .name = "C61", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv46_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv4c_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv4e_chipset = { + .name = "C51", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv4e_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv4e_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv4c_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv50_chipset = { + .name = "G80", + .bar = { 0x00000001, nv50_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv50_bus_new }, + .clk = { 0x00000001, nv50_clk_new }, + .devinit = { 0x00000001, nv50_devinit_new }, + .fb = { 0x00000001, nv50_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, nv50_gpio_new }, + .i2c = { 0x00000001, nv50_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, nv50_mc_new }, + .mmu = { 0x00000001, nv50_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, nv46_pci_new }, + .therm = { 0x00000001, nv50_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv50_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, nv50_fifo_new }, + .gr = { 0x00000001, nv50_gr_new }, + .mpeg = { 0x00000001, nv50_mpeg_new }, + .pm = { 0x00000001, nv50_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nv63_chipset = { + .name = "C73", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv46_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv4c_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv67_chipset = { + .name = "C67", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv46_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv4c_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv68_chipset = { + .name = "C68", + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv31_bus_new }, + .clk = { 0x00000001, nv40_clk_new }, + .devinit = { 0x00000001, nv1a_devinit_new }, + .fb = { 0x00000001, nv46_fb_new }, + .gpio = { 0x00000001, nv10_gpio_new }, + .i2c = { 0x00000001, nv04_i2c_new }, + .imem = { 0x00000001, nv40_instmem_new }, + .mc = { 0x00000001, nv44_mc_new }, + .mmu = { 0x00000001, nv44_mmu_new }, + .pci = { 0x00000001, nv4c_pci_new }, + .therm = { 0x00000001, nv40_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, nv04_disp_new }, + .dma = { 0x00000001, nv04_dma_new }, + .fifo = { 0x00000001, nv40_fifo_new }, + .gr = { 0x00000001, nv44_gr_new }, + .mpeg = { 0x00000001, nv44_mpeg_new }, + .pm = { 0x00000001, nv40_pm_new }, + .sw = { 0x00000001, nv10_sw_new }, +}; + +static const struct nvkm_device_chip +nv84_chipset = { + .name = "G84", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv50_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, nv50_gpio_new }, + .i2c = { 0x00000001, nv50_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g84_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, g84_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, g84_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nv86_chipset = { + .name = "G86", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv50_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, nv50_gpio_new }, + .i2c = { 0x00000001, nv50_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g84_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, g84_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, g84_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nv92_chipset = { + .name = "G92", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, nv50_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, nv50_gpio_new }, + .i2c = { 0x00000001, nv50_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g92_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, g84_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, g84_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nv94_chipset = { + .name = "G94", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, g94_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, g84_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nv96_chipset = { + .name = "G96", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, g94_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, g84_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nv98_chipset = { + .name = "G98", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g98_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g98_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, g94_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, g84_gr_new }, + .mspdec = { 0x00000001, g98_mspdec_new }, + .msppp = { 0x00000001, g98_msppp_new }, + .msvld = { 0x00000001, g98_msvld_new }, + .pm = { 0x00000001, g84_pm_new }, + .sec = { 0x00000001, g98_sec_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nva0_chipset = { + .name = "GT200", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, g84_clk_new }, + .devinit = { 0x00000001, g84_devinit_new }, + .fb = { 0x00000001, g84_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, nv50_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g84_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .bsp = { 0x00000001, g84_bsp_new }, + .cipher = { 0x00000001, g84_cipher_new }, + .disp = { 0x00000001, gt200_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g84_fifo_new }, + .gr = { 0x00000001, gt200_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .pm = { 0x00000001, gt200_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, + .vp = { 0x00000001, g84_vp_new }, +}; + +static const struct nvkm_device_chip +nva3_chipset = { + .name = "GT215", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, gt215_clk_new }, + .devinit = { 0x00000001, gt215_devinit_new }, + .fb = { 0x00000001, gt215_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, gt215_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .pmu = { 0x00000001, gt215_pmu_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .ce = { 0x00000001, gt215_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, gt215_gr_new }, + .mpeg = { 0x00000001, g84_mpeg_new }, + .mspdec = { 0x00000001, gt215_mspdec_new }, + .msppp = { 0x00000001, gt215_msppp_new }, + .msvld = { 0x00000001, gt215_msvld_new }, + .pm = { 0x00000001, gt215_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nva5_chipset = { + .name = "GT216", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, gt215_clk_new }, + .devinit = { 0x00000001, gt215_devinit_new }, + .fb = { 0x00000001, gt215_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, gt215_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .pmu = { 0x00000001, gt215_pmu_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .ce = { 0x00000001, gt215_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, gt215_gr_new }, + .mspdec = { 0x00000001, gt215_mspdec_new }, + .msppp = { 0x00000001, gt215_msppp_new }, + .msvld = { 0x00000001, gt215_msvld_new }, + .pm = { 0x00000001, gt215_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nva8_chipset = { + .name = "GT218", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, gt215_clk_new }, + .devinit = { 0x00000001, gt215_devinit_new }, + .fb = { 0x00000001, gt215_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, gt215_mc_new }, + .mmu = { 0x00000001, g84_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .pmu = { 0x00000001, gt215_pmu_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .ce = { 0x00000001, gt215_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, gt215_gr_new }, + .mspdec = { 0x00000001, gt215_mspdec_new }, + .msppp = { 0x00000001, gt215_msppp_new }, + .msvld = { 0x00000001, gt215_msvld_new }, + .pm = { 0x00000001, gt215_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nvaa_chipset = { + .name = "MCP77/MCP78", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, mcp77_clk_new }, + .devinit = { 0x00000001, g98_devinit_new }, + .fb = { 0x00000001, mcp77_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g98_mc_new }, + .mmu = { 0x00000001, mcp77_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, mcp77_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, gt200_gr_new }, + .mspdec = { 0x00000001, g98_mspdec_new }, + .msppp = { 0x00000001, g98_msppp_new }, + .msvld = { 0x00000001, g98_msvld_new }, + .pm = { 0x00000001, g84_pm_new }, + .sec = { 0x00000001, g98_sec_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nvac_chipset = { + .name = "MCP79/MCP7A", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, mcp77_clk_new }, + .devinit = { 0x00000001, g98_devinit_new }, + .fb = { 0x00000001, mcp77_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, g98_mc_new }, + .mmu = { 0x00000001, mcp77_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .therm = { 0x00000001, g84_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .disp = { 0x00000001, mcp77_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, mcp79_gr_new }, + .mspdec = { 0x00000001, g98_mspdec_new }, + .msppp = { 0x00000001, g98_msppp_new }, + .msvld = { 0x00000001, g98_msvld_new }, + .pm = { 0x00000001, g84_pm_new }, + .sec = { 0x00000001, g98_sec_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nvaf_chipset = { + .name = "MCP89", + .bar = { 0x00000001, g84_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, g94_bus_new }, + .clk = { 0x00000001, gt215_clk_new }, + .devinit = { 0x00000001, mcp89_devinit_new }, + .fb = { 0x00000001, mcp89_fb_new }, + .fuse = { 0x00000001, nv50_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, gt215_mc_new }, + .mmu = { 0x00000001, mcp77_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, g94_pci_new }, + .pmu = { 0x00000001, gt215_pmu_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, nv40_volt_new }, + .ce = { 0x00000001, gt215_ce_new }, + .disp = { 0x00000001, mcp89_disp_new }, + .dma = { 0x00000001, nv50_dma_new }, + .fifo = { 0x00000001, g98_fifo_new }, + .gr = { 0x00000001, mcp89_gr_new }, + .mspdec = { 0x00000001, gt215_mspdec_new }, + .msppp = { 0x00000001, gt215_msppp_new }, + .msvld = { 0x00000001, mcp89_msvld_new }, + .pm = { 0x00000001, gt215_pm_new }, + .sw = { 0x00000001, nv50_sw_new }, +}; + +static const struct nvkm_device_chip +nvc0_chipset = { + .name = "GF100", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf100_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000003, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf100_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvc1_chipset = { + .name = "GF108", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf108_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf106_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000001, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf108_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf108_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvc3_chipset = { + .name = "GF106", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf106_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000001, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf104_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvc4_chipset = { + .name = "GF104", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf100_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000003, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf104_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvc8_chipset = { + .name = "GF110", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf100_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000003, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf110_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvce_chipset = { + .name = "GF114", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf100_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000003, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf104_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvcf_chipset = { + .name = "GF116", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, g94_gpio_new }, + .i2c = { 0x00000001, g94_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf106_pci_new }, + .pmu = { 0x00000001, gf100_pmu_new }, + .privring = { 0x00000001, gf100_privring_new }, + .therm = { 0x00000001, gt215_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000001, gf100_ce_new }, + .disp = { 0x00000001, gt215_disp_new }, + .dma = { 0x00000001, gf100_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf104_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf100_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvd7_chipset = { + .name = "GF117", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gf119_gpio_new }, + .i2c = { 0x00000001, gf117_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf106_pci_new }, + .privring = { 0x00000001, gf117_privring_new }, + .therm = { 0x00000001, gf119_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf117_volt_new }, + .ce = { 0x00000001, gf100_ce_new }, + .disp = { 0x00000001, gf119_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf117_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf117_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvd9_chipset = { + .name = "GF119", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gf100_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gf100_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gf119_gpio_new }, + .i2c = { 0x00000001, gf119_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gf100_ltc_new }, + .mc = { 0x00000001, gf100_mc_new }, + .mmu = { 0x00000001, gf100_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gf106_pci_new }, + .pmu = { 0x00000001, gf119_pmu_new }, + .privring = { 0x00000001, gf117_privring_new }, + .therm = { 0x00000001, gf119_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .volt = { 0x00000001, gf100_volt_new }, + .ce = { 0x00000001, gf100_ce_new }, + .disp = { 0x00000001, gf119_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gf100_fifo_new }, + .gr = { 0x00000001, gf119_gr_new }, + .mspdec = { 0x00000001, gf100_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gf100_msvld_new }, + .pm = { 0x00000001, gf117_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nve4_chipset = { + .name = "GK104", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk104_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk104_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk104_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk104_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk104_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk104_fifo_new }, + .gr = { 0x00000001, gk104_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .pm = { 0x00000001, gk104_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nve6_chipset = { + .name = "GK106", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk104_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk104_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk104_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk104_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk104_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk104_fifo_new }, + .gr = { 0x00000001, gk104_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .pm = { 0x00000001, gk104_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nve7_chipset = { + .name = "GK107", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk104_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk104_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk104_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk104_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk104_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk104_fifo_new }, + .gr = { 0x00000001, gk104_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .pm = { 0x00000001, gk104_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvea_chipset = { + .name = "GK20A", + .bar = { 0x00000001, gk20a_bar_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk20a_clk_new }, + .fb = { 0x00000001, gk20a_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .imem = { 0x00000001, gk20a_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gk20a_mmu_new }, + .pmu = { 0x00000001, gk20a_pmu_new }, + .privring = { 0x00000001, gk20a_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk20a_volt_new }, + .ce = { 0x00000004, gk104_ce_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk20a_fifo_new }, + .gr = { 0x00000001, gk20a_gr_new }, + .pm = { 0x00000001, gk104_pm_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvf0_chipset = { + .name = "GK110", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk110_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk104_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk110_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk110_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk110_fifo_new }, + .gr = { 0x00000001, gk110_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nvf1_chipset = { + .name = "GK110B", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk110_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk104_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk110_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk110_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk110_fifo_new }, + .gr = { 0x00000001, gk110b_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv106_chipset = { + .name = "GK208B", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk110_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk208_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk110_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk208_fifo_new }, + .gr = { 0x00000001, gk208_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv108_chipset = { + .name = "GK208", + .bar = { 0x00000001, gf100_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gf100_devinit_new }, + .fb = { 0x00000001, gk110_fb_new }, + .fuse = { 0x00000001, gf100_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gk104_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gk208_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gk104_therm_new }, + .timer = { 0x00000001, nv41_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gk104_ce_new }, + .disp = { 0x00000001, gk110_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gk208_fifo_new }, + .gr = { 0x00000001, gk208_gr_new }, + .mspdec = { 0x00000001, gk104_mspdec_new }, + .msppp = { 0x00000001, gf100_msppp_new }, + .msvld = { 0x00000001, gk104_msvld_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv117_chipset = { + .name = "GM107", + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gm107_devinit_new }, + .fb = { 0x00000001, gm107_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gm107_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gm107_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gm107_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000005, gm107_ce_new }, + .disp = { 0x00000001, gm107_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm107_fifo_new }, + .gr = { 0x00000001, gm107_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv118_chipset = { + .name = "GM108", + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gk104_clk_new }, + .devinit = { 0x00000001, gm107_devinit_new }, + .fb = { 0x00000001, gm107_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gk110_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gm107_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gk104_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gm107_pmu_new }, + .privring = { 0x00000001, gk104_privring_new }, + .therm = { 0x00000001, gm107_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000005, gm107_ce_new }, + .disp = { 0x00000001, gm107_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm107_fifo_new }, + .gr = { 0x00000001, gm107_gr_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv120_chipset = { + .name = "GM200", + .acr = { 0x00000001, gm200_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fb = { 0x00000001, gm200_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gm200_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gm200_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gm200_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gm200_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gm200_ce_new }, + .disp = { 0x00000001, gm200_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm200_fifo_new }, + .gr = { 0x00000001, gm200_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000003, gm107_nvenc_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv124_chipset = { + .name = "GM204", + .acr = { 0x00000001, gm200_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fb = { 0x00000001, gm200_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gm200_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gm200_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gm200_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gm200_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gm200_ce_new }, + .disp = { 0x00000001, gm200_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm200_fifo_new }, + .gr = { 0x00000001, gm200_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000003, gm107_nvenc_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv126_chipset = { + .name = "GM206", + .acr = { 0x00000001, gm200_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fb = { 0x00000001, gm200_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .iccsense = { 0x00000001, gf100_iccsense_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gm200_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gm200_mmu_new }, + .mxm = { 0x00000001, nv50_mxm_new }, + .pci = { 0x00000001, gk104_pci_new }, + .pmu = { 0x00000001, gm200_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gm200_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gk104_volt_new }, + .ce = { 0x00000007, gm200_ce_new }, + .disp = { 0x00000001, gm200_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm200_fifo_new }, + .gr = { 0x00000001, gm200_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv12b_chipset = { + .name = "GM20B", + .acr = { 0x00000001, gm20b_acr_new }, + .bar = { 0x00000001, gm20b_bar_new }, + .bus = { 0x00000001, gf100_bus_new }, + .clk = { 0x00000001, gm20b_clk_new }, + .fb = { 0x00000001, gm20b_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .imem = { 0x00000001, gk20a_instmem_new }, + .ltc = { 0x00000001, gm200_ltc_new }, + .mc = { 0x00000001, gk20a_mc_new }, + .mmu = { 0x00000001, gm20b_mmu_new }, + .pmu = { 0x00000001, gm20b_pmu_new }, + .privring = { 0x00000001, gk20a_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .volt = { 0x00000001, gm20b_volt_new }, + .ce = { 0x00000004, gm200_ce_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gm200_fifo_new }, + .gr = { 0x00000001, gm20b_gr_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv130_chipset = { + .name = "GP100", + .acr = { 0x00000001, gm200_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp100_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp100_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gm200_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000003f, gp100_ce_new }, + .dma = { 0x00000001, gf119_dma_new }, + .disp = { 0x00000001, gp100_disp_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp100_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000007, gm107_nvenc_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv132_chipset = { + .name = "GP102", + .acr = { 0x00000001, gp102_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000000f, gp102_ce_new }, + .disp = { 0x00000001, gp102_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp102_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000003, gm107_nvenc_new }, + .sec2 = { 0x00000001, gp102_sec2_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv134_chipset = { + .name = "GP104", + .acr = { 0x00000001, gp102_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000000f, gp102_ce_new }, + .disp = { 0x00000001, gp102_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp104_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000003, gm107_nvenc_new }, + .sec2 = { 0x00000001, gp102_sec2_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv136_chipset = { + .name = "GP106", + .acr = { 0x00000001, gp102_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000000f, gp102_ce_new }, + .disp = { 0x00000001, gp102_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp104_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, gp102_sec2_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv137_chipset = { + .name = "GP107", + .acr = { 0x00000001, gp102_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000000f, gp102_ce_new }, + .disp = { 0x00000001, gp102_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp107_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000003, gm107_nvenc_new }, + .sec2 = { 0x00000001, gp102_sec2_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv138_chipset = { + .name = "GP108", + .acr = { 0x00000001, gp108_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gm200_devinit_new }, + .fault = { 0x00000001, gp100_fault_new }, + .fb = { 0x00000001, gp102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gp100_mmu_new }, + .therm = { 0x00000001, gp100_therm_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x0000000f, gp102_ce_new }, + .disp = { 0x00000001, gp102_disp_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp108_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .sec2 = { 0x00000001, gp108_sec2_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv13b_chipset = { + .name = "GP10B", + .acr = { 0x00000001, gp10b_acr_new }, + .bar = { 0x00000001, gm20b_bar_new }, + .bus = { 0x00000001, gf100_bus_new }, + .fault = { 0x00000001, gp10b_fault_new }, + .fb = { 0x00000001, gp10b_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .imem = { 0x00000001, gk20a_instmem_new }, + .ltc = { 0x00000001, gp10b_ltc_new }, + .mc = { 0x00000001, gp10b_mc_new }, + .mmu = { 0x00000001, gp10b_mmu_new }, + .pmu = { 0x00000001, gp10b_pmu_new }, + .privring = { 0x00000001, gp10b_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .ce = { 0x00000001, gp100_ce_new }, + .dma = { 0x00000001, gf119_dma_new }, + .fifo = { 0x00000001, gp100_fifo_new }, + .gr = { 0x00000001, gp10b_gr_new }, + .sw = { 0x00000001, gf100_sw_new }, +}; + +static const struct nvkm_device_chip +nv140_chipset = { + .name = "GV100", + .acr = { 0x00000001, gv100_acr_new }, + .bar = { 0x00000001, gm107_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, gv100_devinit_new }, + .fault = { 0x00000001, gv100_fault_new }, + .fb = { 0x00000001, gv100_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, gv100_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, gv100_vfn_new }, + .ce = { 0x000001ff, gv100_ce_new }, + .disp = { 0x00000001, gv100_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, gv100_fifo_new }, + .gr = { 0x00000001, gv100_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000007, gm107_nvenc_new }, + .sec2 = { 0x00000001, gp108_sec2_new }, +}; + +static const struct nvkm_device_chip +nv162_chipset = { + .name = "TU102", + .acr = { 0x00000001, tu102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, tu102_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, tu102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, tu102_vfn_new }, + .ce = { 0x0000001f, tu102_ce_new }, + .disp = { 0x00000001, tu102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, tu102_fifo_new }, + .gr = { 0x00000001, tu102_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, tu102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv164_chipset = { + .name = "TU104", + .acr = { 0x00000001, tu102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, tu102_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, tu102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, tu102_vfn_new }, + .ce = { 0x0000001f, tu102_ce_new }, + .disp = { 0x00000001, tu102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, tu102_fifo_new }, + .gr = { 0x00000001, tu102_gr_new }, + .nvdec = { 0x00000003, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, tu102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv166_chipset = { + .name = "TU106", + .acr = { 0x00000001, tu102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, tu102_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, tu102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, tu102_vfn_new }, + .ce = { 0x0000001f, tu102_ce_new }, + .disp = { 0x00000001, tu102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, tu102_fifo_new }, + .gr = { 0x00000001, tu102_gr_new }, + .nvdec = { 0x00000007, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, tu102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv167_chipset = { + .name = "TU117", + .acr = { 0x00000001, tu102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, tu102_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, tu102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, tu102_vfn_new }, + .ce = { 0x0000001f, tu102_ce_new }, + .disp = { 0x00000001, tu102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, tu102_fifo_new }, + .gr = { 0x00000001, tu102_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, tu102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv168_chipset = { + .name = "TU116", + .acr = { 0x00000001, tu102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .bus = { 0x00000001, gf100_bus_new }, + .devinit = { 0x00000001, tu102_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, tu102_fb_new }, + .fuse = { 0x00000001, gm107_fuse_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .gsp = { 0x00000001, gv100_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, gp102_ltc_new }, + .mc = { 0x00000001, gp100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .pmu = { 0x00000001, gp102_pmu_new }, + .privring = { 0x00000001, gm200_privring_new }, + .therm = { 0x00000001, gp100_therm_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, gk104_top_new }, + .vfn = { 0x00000001, tu102_vfn_new }, + .ce = { 0x0000001f, tu102_ce_new }, + .disp = { 0x00000001, tu102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, tu102_fifo_new }, + .gr = { 0x00000001, tu102_gr_new }, + .nvdec = { 0x00000001, gm107_nvdec_new }, + .nvenc = { 0x00000001, gm107_nvenc_new }, + .sec2 = { 0x00000001, tu102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv170_chipset = { + .name = "GA100", + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga100_fb_new }, + .gpio = { 0x00000001, gk104_gpio_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x000003ff, ga100_ce_new }, + .fifo = { 0x00000001, ga100_fifo_new }, +}; + +static const struct nvkm_device_chip +nv172_chipset = { + .name = "GA102", + .acr = { 0x00000001, ga102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga102_fb_new }, + .gpio = { 0x00000001, ga102_gpio_new }, + .gsp = { 0x00000001, ga102_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, ga102_ltc_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x0000001f, ga102_ce_new }, + .disp = { 0x00000001, ga102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, ga102_fifo_new }, + .gr = { 0x00000001, ga102_gr_new }, + .nvdec = { 0x00000001, ga102_nvdec_new }, + .sec2 = { 0x00000001, ga102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv173_chipset = { + .name = "GA103", + .acr = { 0x00000001, ga102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga102_fb_new }, + .gpio = { 0x00000001, ga102_gpio_new }, + .gsp = { 0x00000001, ga102_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, ga102_ltc_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x0000001f, ga102_ce_new }, + .disp = { 0x00000001, ga102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, ga102_fifo_new }, + .gr = { 0x00000001, ga102_gr_new }, + .nvdec = { 0x00000001, ga102_nvdec_new }, + .sec2 = { 0x00000001, ga102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv174_chipset = { + .name = "GA104", + .acr = { 0x00000001, ga102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga102_fb_new }, + .gpio = { 0x00000001, ga102_gpio_new }, + .gsp = { 0x00000001, ga102_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, ga102_ltc_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x0000001f, ga102_ce_new }, + .disp = { 0x00000001, ga102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, ga102_fifo_new }, + .gr = { 0x00000001, ga102_gr_new }, + .nvdec = { 0x00000001, ga102_nvdec_new }, + .sec2 = { 0x00000001, ga102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv176_chipset = { + .name = "GA106", + .acr = { 0x00000001, ga102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga102_fb_new }, + .gpio = { 0x00000001, ga102_gpio_new }, + .gsp = { 0x00000001, ga102_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, ga102_ltc_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x0000001f, ga102_ce_new }, + .disp = { 0x00000001, ga102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, ga102_fifo_new }, + .gr = { 0x00000001, ga102_gr_new }, + .nvdec = { 0x00000001, ga102_nvdec_new }, + .sec2 = { 0x00000001, ga102_sec2_new }, +}; + +static const struct nvkm_device_chip +nv177_chipset = { + .name = "GA107", + .acr = { 0x00000001, ga102_acr_new }, + .bar = { 0x00000001, tu102_bar_new }, + .bios = { 0x00000001, nvkm_bios_new }, + .devinit = { 0x00000001, ga100_devinit_new }, + .fault = { 0x00000001, tu102_fault_new }, + .fb = { 0x00000001, ga102_fb_new }, + .gpio = { 0x00000001, ga102_gpio_new }, + .gsp = { 0x00000001, ga102_gsp_new }, + .i2c = { 0x00000001, gm200_i2c_new }, + .imem = { 0x00000001, nv50_instmem_new }, + .ltc = { 0x00000001, ga102_ltc_new }, + .mc = { 0x00000001, ga100_mc_new }, + .mmu = { 0x00000001, tu102_mmu_new }, + .pci = { 0x00000001, gp100_pci_new }, + .privring = { 0x00000001, gm200_privring_new }, + .timer = { 0x00000001, gk20a_timer_new }, + .top = { 0x00000001, ga100_top_new }, + .vfn = { 0x00000001, ga100_vfn_new }, + .ce = { 0x0000001f, ga102_ce_new }, + .disp = { 0x00000001, ga102_disp_new }, + .dma = { 0x00000001, gv100_dma_new }, + .fifo = { 0x00000001, ga102_fifo_new }, + .gr = { 0x00000001, ga102_gr_new }, + .nvdec = { 0x00000001, ga102_nvdec_new }, + .sec2 = { 0x00000001, ga102_sec2_new }, +}; + +struct nvkm_subdev * +nvkm_device_subdev(struct nvkm_device *device, int type, int inst) +{ + struct nvkm_subdev *subdev; + + list_for_each_entry(subdev, &device->subdev, head) { + if (subdev->type == type && subdev->inst == inst) + return subdev; + } + + return NULL; +} + +struct nvkm_engine * +nvkm_device_engine(struct nvkm_device *device, int type, int inst) +{ + struct nvkm_subdev *subdev = nvkm_device_subdev(device, type, inst); + if (subdev && subdev->func == &nvkm_engine) + return container_of(subdev, struct nvkm_engine, subdev); + return NULL; +} + +int +nvkm_device_fini(struct nvkm_device *device, bool suspend) +{ + const char *action = suspend ? "suspend" : "fini"; + struct nvkm_subdev *subdev; + int ret; + s64 time; + + nvdev_trace(device, "%s running...\n", action); + time = ktime_to_us(ktime_get()); + + nvkm_acpi_fini(device); + + list_for_each_entry_reverse(subdev, &device->subdev, head) { + ret = nvkm_subdev_fini(subdev, suspend); + if (ret && suspend) + goto fail; + } + + nvkm_therm_clkgate_fini(device->therm, suspend); + + if (device->func->fini) + device->func->fini(device, suspend); + + nvkm_intr_unarm(device); + + time = ktime_to_us(ktime_get()) - time; + nvdev_trace(device, "%s completed in %lldus...\n", action, time); + return 0; + +fail: + list_for_each_entry_from(subdev, &device->subdev, head) { + int rret = nvkm_subdev_init(subdev); + if (rret) + nvkm_fatal(subdev, "failed restart, %d\n", ret); + } + + nvdev_trace(device, "%s failed with %d\n", action, ret); + return ret; +} + +static int +nvkm_device_preinit(struct nvkm_device *device) +{ + struct nvkm_subdev *subdev; + int ret; + s64 time; + + nvdev_trace(device, "preinit running...\n"); + time = ktime_to_us(ktime_get()); + + nvkm_intr_unarm(device); + + if (device->func->preinit) { + ret = device->func->preinit(device); + if (ret) + goto fail; + } + + list_for_each_entry(subdev, &device->subdev, head) { + ret = nvkm_subdev_preinit(subdev); + if (ret) + goto fail; + } + + ret = nvkm_devinit_post(device->devinit); + if (ret) + goto fail; + + ret = nvkm_top_parse(device); + if (ret) + goto fail; + + ret = nvkm_fb_mem_unlock(device->fb); + if (ret) + goto fail; + + time = ktime_to_us(ktime_get()) - time; + nvdev_trace(device, "preinit completed in %lldus\n", time); + return 0; + +fail: + nvdev_error(device, "preinit failed with %d\n", ret); + return ret; +} + +int +nvkm_device_init(struct nvkm_device *device) +{ + struct nvkm_subdev *subdev; + int ret; + s64 time; + + ret = nvkm_device_preinit(device); + if (ret) + return ret; + + nvkm_device_fini(device, false); + + nvdev_trace(device, "init running...\n"); + time = ktime_to_us(ktime_get()); + + nvkm_intr_rearm(device); + + if (device->func->init) { + ret = device->func->init(device); + if (ret) + goto fail; + } + + list_for_each_entry(subdev, &device->subdev, head) { + ret = nvkm_subdev_init(subdev); + if (ret) + goto fail_subdev; + } + + nvkm_acpi_init(device); + nvkm_therm_clkgate_enable(device->therm); + + time = ktime_to_us(ktime_get()) - time; + nvdev_trace(device, "init completed in %lldus\n", time); + return 0; + +fail_subdev: + list_for_each_entry_from(subdev, &device->subdev, head) + nvkm_subdev_fini(subdev, false); +fail: + nvkm_device_fini(device, false); + + nvdev_error(device, "init failed with %d\n", ret); + return ret; +} + +void +nvkm_device_del(struct nvkm_device **pdevice) +{ + struct nvkm_device *device = *pdevice; + struct nvkm_subdev *subdev, *subtmp; + if (device) { + mutex_lock(&nv_devices_mutex); + + nvkm_intr_dtor(device); + + list_for_each_entry_safe_reverse(subdev, subtmp, &device->subdev, head) + nvkm_subdev_del(&subdev); + + if (device->pri) + iounmap(device->pri); + list_del(&device->head); + + if (device->func->dtor) + *pdevice = device->func->dtor(device); + mutex_unlock(&nv_devices_mutex); + + kfree(*pdevice); + *pdevice = NULL; + } +} + +/* returns true if the GPU is in the CPU native byte order */ +static inline bool +nvkm_device_endianness(struct nvkm_device *device) +{ +#ifdef __BIG_ENDIAN + const bool big_endian = true; +#else + const bool big_endian = false; +#endif + + /* Read NV_PMC_BOOT_1, and assume non-functional endian switch if it + * doesn't contain the expected values. + */ + u32 pmc_boot_1 = nvkm_rd32(device, 0x000004); + if (pmc_boot_1 && pmc_boot_1 != 0x01000001) + return !big_endian; /* Assume GPU is LE in this case. */ + + /* 0 means LE and 0x01000001 means BE GPU. Condition is true when + * GPU/CPU endianness don't match. + */ + if (big_endian == !pmc_boot_1) { + nvkm_wr32(device, 0x000004, 0x01000001); + nvkm_rd32(device, 0x000000); + if (nvkm_rd32(device, 0x000004) != (big_endian ? 0x01000001 : 0x00000000)) + return !big_endian; /* Assume GPU is LE on any unexpected read-back. */ + } + + /* CPU/GPU endianness should (hopefully) match. */ + return true; +} + +int +nvkm_device_ctor(const struct nvkm_device_func *func, + const struct nvkm_device_quirk *quirk, + struct device *dev, enum nvkm_device_type type, u64 handle, + const char *name, const char *cfg, const char *dbg, + bool detect, bool mmio, u64 subdev_mask, + struct nvkm_device *device) +{ + struct nvkm_subdev *subdev; + u64 mmio_base, mmio_size; + u32 boot0, boot1, strap; + int ret = -EEXIST, j; + unsigned chipset; + + mutex_lock(&nv_devices_mutex); + if (nvkm_device_find_locked(handle)) + goto done; + + device->func = func; + device->quirk = quirk; + device->dev = dev; + device->type = type; + device->handle = handle; + device->cfgopt = cfg; + device->dbgopt = dbg; + device->name = name; + list_add_tail(&device->head, &nv_devices); + device->debug = nvkm_dbgopt(device->dbgopt, "device"); + INIT_LIST_HEAD(&device->subdev); + + mmio_base = device->func->resource_addr(device, 0); + mmio_size = device->func->resource_size(device, 0); + + if (detect || mmio) { + device->pri = ioremap(mmio_base, mmio_size); + if (device->pri == NULL) { + nvdev_error(device, "unable to map PRI\n"); + ret = -ENOMEM; + goto done; + } + } + + /* identify the chipset, and determine classes of subdev/engines */ + if (detect) { + /* switch mmio to cpu's native endianness */ + if (!nvkm_device_endianness(device)) { + nvdev_error(device, + "Couldn't switch GPU to CPUs endianness\n"); + ret = -ENOSYS; + goto done; + } + + boot0 = nvkm_rd32(device, 0x000000); + + /* chipset can be overridden for devel/testing purposes */ + chipset = nvkm_longopt(device->cfgopt, "NvChipset", 0); + if (chipset) { + u32 override_boot0; + + if (chipset >= 0x10) { + override_boot0 = ((chipset & 0x1ff) << 20); + override_boot0 |= 0x000000a1; + } else { + if (chipset != 0x04) + override_boot0 = 0x20104000; + else + override_boot0 = 0x20004000; + } + + nvdev_warn(device, "CHIPSET OVERRIDE: %08x -> %08x\n", + boot0, override_boot0); + boot0 = override_boot0; + } + + /* determine chipset and derive architecture from it */ + if ((boot0 & 0x1f000000) > 0) { + device->chipset = (boot0 & 0x1ff00000) >> 20; + device->chiprev = (boot0 & 0x000000ff); + switch (device->chipset & 0x1f0) { + case 0x010: { + if (0x461 & (1 << (device->chipset & 0xf))) + device->card_type = NV_10; + else + device->card_type = NV_11; + device->chiprev = 0x00; + break; + } + case 0x020: device->card_type = NV_20; break; + case 0x030: device->card_type = NV_30; break; + case 0x040: + case 0x060: device->card_type = NV_40; break; + case 0x050: + case 0x080: + case 0x090: + case 0x0a0: device->card_type = NV_50; break; + case 0x0c0: + case 0x0d0: device->card_type = NV_C0; break; + case 0x0e0: + case 0x0f0: + case 0x100: device->card_type = NV_E0; break; + case 0x110: + case 0x120: device->card_type = GM100; break; + case 0x130: device->card_type = GP100; break; + case 0x140: device->card_type = GV100; break; + case 0x160: device->card_type = TU100; break; + case 0x170: device->card_type = GA100; break; + default: + break; + } + } else + if ((boot0 & 0xff00fff0) == 0x20004000) { + if (boot0 & 0x00f00000) + device->chipset = 0x05; + else + device->chipset = 0x04; + device->card_type = NV_04; + } + + switch (device->chipset) { + case 0x004: device->chip = &nv4_chipset; break; + case 0x005: device->chip = &nv5_chipset; break; + case 0x010: device->chip = &nv10_chipset; break; + case 0x011: device->chip = &nv11_chipset; break; + case 0x015: device->chip = &nv15_chipset; break; + case 0x017: device->chip = &nv17_chipset; break; + case 0x018: device->chip = &nv18_chipset; break; + case 0x01a: device->chip = &nv1a_chipset; break; + case 0x01f: device->chip = &nv1f_chipset; break; + case 0x020: device->chip = &nv20_chipset; break; + case 0x025: device->chip = &nv25_chipset; break; + case 0x028: device->chip = &nv28_chipset; break; + case 0x02a: device->chip = &nv2a_chipset; break; + case 0x030: device->chip = &nv30_chipset; break; + case 0x031: device->chip = &nv31_chipset; break; + case 0x034: device->chip = &nv34_chipset; break; + case 0x035: device->chip = &nv35_chipset; break; + case 0x036: device->chip = &nv36_chipset; break; + case 0x040: device->chip = &nv40_chipset; break; + case 0x041: device->chip = &nv41_chipset; break; + case 0x042: device->chip = &nv42_chipset; break; + case 0x043: device->chip = &nv43_chipset; break; + case 0x044: device->chip = &nv44_chipset; break; + case 0x045: device->chip = &nv45_chipset; break; + case 0x046: device->chip = &nv46_chipset; break; + case 0x047: device->chip = &nv47_chipset; break; + case 0x049: device->chip = &nv49_chipset; break; + case 0x04a: device->chip = &nv4a_chipset; break; + case 0x04b: device->chip = &nv4b_chipset; break; + case 0x04c: device->chip = &nv4c_chipset; break; + case 0x04e: device->chip = &nv4e_chipset; break; + case 0x050: device->chip = &nv50_chipset; break; + case 0x063: device->chip = &nv63_chipset; break; + case 0x067: device->chip = &nv67_chipset; break; + case 0x068: device->chip = &nv68_chipset; break; + case 0x084: device->chip = &nv84_chipset; break; + case 0x086: device->chip = &nv86_chipset; break; + case 0x092: device->chip = &nv92_chipset; break; + case 0x094: device->chip = &nv94_chipset; break; + case 0x096: device->chip = &nv96_chipset; break; + case 0x098: device->chip = &nv98_chipset; break; + case 0x0a0: device->chip = &nva0_chipset; break; + case 0x0a3: device->chip = &nva3_chipset; break; + case 0x0a5: device->chip = &nva5_chipset; break; + case 0x0a8: device->chip = &nva8_chipset; break; + case 0x0aa: device->chip = &nvaa_chipset; break; + case 0x0ac: device->chip = &nvac_chipset; break; + case 0x0af: device->chip = &nvaf_chipset; break; + case 0x0c0: device->chip = &nvc0_chipset; break; + case 0x0c1: device->chip = &nvc1_chipset; break; + case 0x0c3: device->chip = &nvc3_chipset; break; + case 0x0c4: device->chip = &nvc4_chipset; break; + case 0x0c8: device->chip = &nvc8_chipset; break; + case 0x0ce: device->chip = &nvce_chipset; break; + case 0x0cf: device->chip = &nvcf_chipset; break; + case 0x0d7: device->chip = &nvd7_chipset; break; + case 0x0d9: device->chip = &nvd9_chipset; break; + case 0x0e4: device->chip = &nve4_chipset; break; + case 0x0e6: device->chip = &nve6_chipset; break; + case 0x0e7: device->chip = &nve7_chipset; break; + case 0x0ea: device->chip = &nvea_chipset; break; + case 0x0f0: device->chip = &nvf0_chipset; break; + case 0x0f1: device->chip = &nvf1_chipset; break; + case 0x106: device->chip = &nv106_chipset; break; + case 0x108: device->chip = &nv108_chipset; break; + case 0x117: device->chip = &nv117_chipset; break; + case 0x118: device->chip = &nv118_chipset; break; + case 0x120: device->chip = &nv120_chipset; break; + case 0x124: device->chip = &nv124_chipset; break; + case 0x126: device->chip = &nv126_chipset; break; + case 0x12b: device->chip = &nv12b_chipset; break; + case 0x130: device->chip = &nv130_chipset; break; + case 0x132: device->chip = &nv132_chipset; break; + case 0x134: device->chip = &nv134_chipset; break; + case 0x136: device->chip = &nv136_chipset; break; + case 0x137: device->chip = &nv137_chipset; break; + case 0x138: device->chip = &nv138_chipset; break; + case 0x13b: device->chip = &nv13b_chipset; break; + case 0x140: device->chip = &nv140_chipset; break; + case 0x162: device->chip = &nv162_chipset; break; + case 0x164: device->chip = &nv164_chipset; break; + case 0x166: device->chip = &nv166_chipset; break; + case 0x167: device->chip = &nv167_chipset; break; + case 0x168: device->chip = &nv168_chipset; break; + case 0x172: device->chip = &nv172_chipset; break; + case 0x173: device->chip = &nv173_chipset; break; + case 0x174: device->chip = &nv174_chipset; break; + case 0x176: device->chip = &nv176_chipset; break; + case 0x177: device->chip = &nv177_chipset; break; + default: + if (nvkm_boolopt(device->cfgopt, "NvEnableUnsupportedChipsets", false)) { + switch (device->chipset) { + case 0x170: device->chip = &nv170_chipset; break; + default: + break; + } + } + + if (!device->chip) { + nvdev_error(device, "unknown chipset (%08x)\n", boot0); + ret = -ENODEV; + goto done; + } + break; + } + + nvdev_info(device, "NVIDIA %s (%08x)\n", + device->chip->name, boot0); + + /* vGPU detection */ + boot1 = nvkm_rd32(device, 0x0000004); + if (device->card_type >= TU100 && (boot1 & 0x00030000)) { + nvdev_info(device, "vGPUs are not supported\n"); + ret = -ENODEV; + goto done; + } + + /* read strapping information */ + strap = nvkm_rd32(device, 0x101000); + + /* determine frequency of timing crystal */ + if ( device->card_type <= NV_10 || device->chipset < 0x17 || + (device->chipset >= 0x20 && device->chipset < 0x25)) + strap &= 0x00000040; + else + strap &= 0x00400040; + + switch (strap) { + case 0x00000000: device->crystal = 13500; break; + case 0x00000040: device->crystal = 14318; break; + case 0x00400000: device->crystal = 27000; break; + case 0x00400040: device->crystal = 25000; break; + } + } else { + device->chip = &null_chipset; + } + + if (!device->name) + device->name = device->chip->name; + + mutex_init(&device->mutex); + nvkm_intr_ctor(device); + +#define NVKM_LAYOUT_ONCE(type,data,ptr) \ + if (device->chip->ptr.inst && (subdev_mask & (BIT_ULL(type)))) { \ + WARN_ON(device->chip->ptr.inst != 0x00000001); \ + ret = device->chip->ptr.ctor(device, (type), -1, &device->ptr); \ + subdev = nvkm_device_subdev(device, (type), 0); \ + if (ret) { \ + nvkm_subdev_del(&subdev); \ + device->ptr = NULL; \ + if (ret != -ENODEV) { \ + nvdev_error(device, "%s ctor failed: %d\n", \ + nvkm_subdev_type[(type)], ret); \ + goto done; \ + } \ + } else { \ + subdev->pself = (void **)&device->ptr; \ + } \ + } +#define NVKM_LAYOUT_INST(type,data,ptr,cnt) \ + WARN_ON(device->chip->ptr.inst & ~((1 << ARRAY_SIZE(device->ptr)) - 1)); \ + for (j = 0; device->chip->ptr.inst && j < ARRAY_SIZE(device->ptr); j++) { \ + if ((device->chip->ptr.inst & BIT(j)) && (subdev_mask & BIT_ULL(type))) { \ + ret = device->chip->ptr.ctor(device, (type), (j), &device->ptr[j]); \ + subdev = nvkm_device_subdev(device, (type), (j)); \ + if (ret) { \ + nvkm_subdev_del(&subdev); \ + device->ptr[j] = NULL; \ + if (ret != -ENODEV) { \ + nvdev_error(device, "%s%d ctor failed: %d\n", \ + nvkm_subdev_type[(type)], (j), ret); \ + goto done; \ + } \ + } else { \ + subdev->pself = (void **)&device->ptr[j]; \ + } \ + } \ + } +#include +#undef NVKM_LAYOUT_INST +#undef NVKM_LAYOUT_ONCE + + ret = nvkm_intr_install(device); +done: + if (device->pri && (!mmio || ret)) { + iounmap(device->pri); + device->pri = NULL; + } + mutex_unlock(&nv_devices_mutex); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.c new file mode 100644 index 0000000000..ce774579c8 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.c @@ -0,0 +1,212 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctrl.h" + +#include +#include + +#include +#include +#include +#include + +static int +nvkm_control_mthd_pstate_info(struct nvkm_control *ctrl, void *data, u32 size) +{ + union { + struct nvif_control_pstate_info_v0 v0; + } *args = data; + struct nvkm_clk *clk = ctrl->device->clk; + int ret = -ENOSYS; + + nvif_ioctl(&ctrl->object, "control pstate info size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(&ctrl->object, "control pstate info vers %d\n", + args->v0.version); + } else + return ret; + + if (clk) { + args->v0.count = clk->state_nr; + args->v0.ustate_ac = clk->ustate_ac; + args->v0.ustate_dc = clk->ustate_dc; + args->v0.pwrsrc = clk->pwrsrc; + args->v0.pstate = clk->pstate; + } else { + args->v0.count = 0; + args->v0.ustate_ac = NVIF_CONTROL_PSTATE_INFO_V0_USTATE_DISABLE; + args->v0.ustate_dc = NVIF_CONTROL_PSTATE_INFO_V0_USTATE_DISABLE; + args->v0.pwrsrc = -ENODEV; + args->v0.pstate = NVIF_CONTROL_PSTATE_INFO_V0_PSTATE_UNKNOWN; + } + + return 0; +} + +static int +nvkm_control_mthd_pstate_attr(struct nvkm_control *ctrl, void *data, u32 size) +{ + union { + struct nvif_control_pstate_attr_v0 v0; + } *args = data; + struct nvkm_clk *clk = ctrl->device->clk; + const struct nvkm_domain *domain; + struct nvkm_pstate *pstate; + struct nvkm_cstate *cstate; + int i = 0, j = -1; + u32 lo, hi; + int ret = -ENOSYS; + + nvif_ioctl(&ctrl->object, "control pstate attr size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(&ctrl->object, + "control pstate attr vers %d state %d index %d\n", + args->v0.version, args->v0.state, args->v0.index); + if (!clk) + return -ENODEV; + if (args->v0.state < NVIF_CONTROL_PSTATE_ATTR_V0_STATE_CURRENT) + return -EINVAL; + if (args->v0.state >= clk->state_nr) + return -EINVAL; + } else + return ret; + domain = clk->domains; + + while (domain->name != nv_clk_src_max) { + if (domain->mname && ++j == args->v0.index) + break; + domain++; + } + + if (domain->name == nv_clk_src_max) + return -EINVAL; + + if (args->v0.state != NVIF_CONTROL_PSTATE_ATTR_V0_STATE_CURRENT) { + list_for_each_entry(pstate, &clk->states, head) { + if (i++ == args->v0.state) + break; + } + + lo = pstate->base.domain[domain->name]; + hi = lo; + list_for_each_entry(cstate, &pstate->list, head) { + lo = min(lo, cstate->domain[domain->name]); + hi = max(hi, cstate->domain[domain->name]); + } + + args->v0.state = pstate->pstate; + } else { + lo = max(nvkm_clk_read(clk, domain->name), 0); + hi = lo; + } + + snprintf(args->v0.name, sizeof(args->v0.name), "%s", domain->mname); + snprintf(args->v0.unit, sizeof(args->v0.unit), "MHz"); + args->v0.min = lo / domain->mdiv; + args->v0.max = hi / domain->mdiv; + + args->v0.index = 0; + while ((++domain)->name != nv_clk_src_max) { + if (domain->mname) { + args->v0.index = ++j; + break; + } + } + + return 0; +} + +static int +nvkm_control_mthd_pstate_user(struct nvkm_control *ctrl, void *data, u32 size) +{ + union { + struct nvif_control_pstate_user_v0 v0; + } *args = data; + struct nvkm_clk *clk = ctrl->device->clk; + int ret = -ENOSYS; + + nvif_ioctl(&ctrl->object, "control pstate user size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(&ctrl->object, + "control pstate user vers %d ustate %d pwrsrc %d\n", + args->v0.version, args->v0.ustate, args->v0.pwrsrc); + if (!clk) + return -ENODEV; + } else + return ret; + + if (args->v0.pwrsrc >= 0) { + ret |= nvkm_clk_ustate(clk, args->v0.ustate, args->v0.pwrsrc); + } else { + ret |= nvkm_clk_ustate(clk, args->v0.ustate, 0); + ret |= nvkm_clk_ustate(clk, args->v0.ustate, 1); + } + + return ret; +} + +static int +nvkm_control_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + struct nvkm_control *ctrl = nvkm_control(object); + switch (mthd) { + case NVIF_CONTROL_PSTATE_INFO: + return nvkm_control_mthd_pstate_info(ctrl, data, size); + case NVIF_CONTROL_PSTATE_ATTR: + return nvkm_control_mthd_pstate_attr(ctrl, data, size); + case NVIF_CONTROL_PSTATE_USER: + return nvkm_control_mthd_pstate_user(ctrl, data, size); + default: + break; + } + return -EINVAL; +} + +static const struct nvkm_object_func +nvkm_control = { + .mthd = nvkm_control_mthd, +}; + +static int +nvkm_control_new(struct nvkm_device *device, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + struct nvkm_control *ctrl; + + if (!(ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL))) + return -ENOMEM; + *pobject = &ctrl->object; + ctrl->device = device; + + nvkm_object_ctor(&nvkm_control, oclass, &ctrl->object); + return 0; +} + +const struct nvkm_device_oclass +nvkm_control_oclass = { + .base.oclass = NVIF_CLASS_CONTROL, + .base.minver = -1, + .base.maxver = -1, + .ctor = nvkm_control_new, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.h b/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.h new file mode 100644 index 0000000000..9f6d7f23af --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/ctrl.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DEVICE_CTRL_H__ +#define __NVKM_DEVICE_CTRL_H__ +#define nvkm_control(p) container_of((p), struct nvkm_control, object) +#include + +struct nvkm_control { + struct nvkm_object object; + struct nvkm_device *device; +}; + +extern const struct nvkm_device_oclass nvkm_control_oclass; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c new file mode 100644 index 0000000000..abccb2bb68 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/pci.c @@ -0,0 +1,1702 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include +#include "priv.h" + +struct nvkm_device_pci_device { + u16 device; + const char *name; + const struct nvkm_device_pci_vendor *vendor; +}; + +struct nvkm_device_pci_vendor { + u16 vendor; + u16 device; + const char *name; + const struct nvkm_device_quirk quirk; +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0189[] = { + /* Apple iMac G4 NV18 */ + { 0x10de, 0x0010, NULL, { .tv_gpio = 4 } }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_01f0[] = { + /* MSI nForce2 IGP */ + { 0x1462, 0x5710, NULL, { .tv_pin_mask = 0xc } }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0322[] = { + /* Zotac FX5200 */ + { 0x19da, 0x1035, NULL, { .tv_pin_mask = 0xc } }, + { 0x19da, 0x2035, NULL, { .tv_pin_mask = 0xc } }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_05e7[] = { + { 0x10de, 0x0595, "Tesla T10 Processor" }, + { 0x10de, 0x068f, "Tesla T10 Processor" }, + { 0x10de, 0x0697, "Tesla M1060" }, + { 0x10de, 0x0714, "Tesla M1060" }, + { 0x10de, 0x0743, "Tesla M1060" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0609[] = { + { 0x106b, 0x00a7, "GeForce 8800 GS" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_062e[] = { + { 0x106b, 0x0605, "GeForce GT 130" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0649[] = { + { 0x1043, 0x202d, "GeForce GT 220M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0652[] = { + { 0x152d, 0x0850, "GeForce GT 240M LE" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0654[] = { + { 0x1043, 0x14a2, "GeForce GT 320M" }, + { 0x1043, 0x14d2, "GeForce GT 320M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0655[] = { + { 0x106b, 0x0633, "GeForce GT 120" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0656[] = { + { 0x106b, 0x0693, "GeForce GT 120" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06d1[] = { + { 0x10de, 0x0771, "Tesla C2050" }, + { 0x10de, 0x0772, "Tesla C2070" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06d2[] = { + { 0x10de, 0x088f, "Tesla X2070" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06de[] = { + { 0x10de, 0x0773, "Tesla S2050" }, + { 0x10de, 0x082f, "Tesla M2050" }, + { 0x10de, 0x0840, "Tesla X2070" }, + { 0x10de, 0x0842, "Tesla M2050" }, + { 0x10de, 0x0846, "Tesla M2050" }, + { 0x10de, 0x0866, "Tesla M2050" }, + { 0x10de, 0x0907, "Tesla M2050" }, + { 0x10de, 0x091e, "Tesla M2050" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06e8[] = { + { 0x103c, 0x360b, "GeForce 9200M GE" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06f9[] = { + { 0x10de, 0x060d, "Quadro FX 370 Low Profile" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_06ff[] = { + { 0x10de, 0x0711, "HICx8 + Graphics" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0866[] = { + { 0x106b, 0x00b1, "GeForce 9400M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0872[] = { + { 0x1043, 0x1c42, "GeForce G205M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0873[] = { + { 0x1043, 0x1c52, "GeForce G205M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a6e[] = { + { 0x17aa, 0x3607, "Second Generation ION" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a70[] = { + { 0x17aa, 0x3605, "Second Generation ION" }, + { 0x17aa, 0x3617, "Second Generation ION" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a73[] = { + { 0x17aa, 0x3607, "Second Generation ION" }, + { 0x17aa, 0x3610, "Second Generation ION" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a74[] = { + { 0x17aa, 0x903a, "GeForce G210" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a75[] = { + { 0x17aa, 0x3605, "Second Generation ION" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0a7a[] = { + { 0x1462, 0xaa51, "GeForce 405" }, + { 0x1462, 0xaa58, "GeForce 405" }, + { 0x1462, 0xac71, "GeForce 405" }, + { 0x1462, 0xac82, "GeForce 405" }, + { 0x1642, 0x3980, "GeForce 405" }, + { 0x17aa, 0x3950, "GeForce 405M" }, + { 0x17aa, 0x397d, "GeForce 405M" }, + { 0x1b0a, 0x90b4, "GeForce 405" }, + { 0x1bfd, 0x0003, "GeForce 405" }, + { 0x1bfd, 0x8006, "GeForce 405" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0dd8[] = { + { 0x10de, 0x0914, "Quadro 2000D" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0de9[] = { + { 0x1025, 0x0692, "GeForce GT 620M" }, + { 0x1025, 0x0725, "GeForce GT 620M" }, + { 0x1025, 0x0728, "GeForce GT 620M" }, + { 0x1025, 0x072b, "GeForce GT 620M" }, + { 0x1025, 0x072e, "GeForce GT 620M" }, + { 0x1025, 0x0753, "GeForce GT 620M" }, + { 0x1025, 0x0754, "GeForce GT 620M" }, + { 0x17aa, 0x3977, "GeForce GT 640M LE" }, + { 0x1b0a, 0x2210, "GeForce GT 635M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0dea[] = { + { 0x17aa, 0x365a, "GeForce 615" }, + { 0x17aa, 0x365b, "GeForce 615" }, + { 0x17aa, 0x365e, "GeForce 615" }, + { 0x17aa, 0x3660, "GeForce 615" }, + { 0x17aa, 0x366c, "GeForce 615" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0df4[] = { + { 0x152d, 0x0952, "GeForce GT 630M" }, + { 0x152d, 0x0953, "GeForce GT 630M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0fd2[] = { + { 0x1028, 0x0595, "GeForce GT 640M LE" }, + { 0x1028, 0x05b2, "GeForce GT 640M LE" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_0fe3[] = { + { 0x103c, 0x2b16, "GeForce GT 745A" }, + { 0x17aa, 0x3675, "GeForce GT 745A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_104b[] = { + { 0x1043, 0x844c, "GeForce GT 625" }, + { 0x1043, 0x846b, "GeForce GT 625" }, + { 0x1462, 0xb590, "GeForce GT 625" }, + { 0x174b, 0x0625, "GeForce GT 625" }, + { 0x174b, 0xa625, "GeForce GT 625" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1058[] = { + { 0x103c, 0x2af1, "GeForce 610" }, + { 0x17aa, 0x3682, "GeForce 800A" }, + { 0x17aa, 0x3692, "GeForce 705A" }, + { 0x17aa, 0x3695, "GeForce 800A" }, + { 0x17aa, 0x36a8, "GeForce 800A" }, + { 0x17aa, 0x36ac, "GeForce 800A" }, + { 0x17aa, 0x36ad, "GeForce 800A" }, + { 0x705a, 0x3682, "GeForce 800A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_105b[] = { + { 0x103c, 0x2afb, "GeForce 705A" }, + { 0x17aa, 0x36a1, "GeForce 800A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1091[] = { + { 0x10de, 0x088e, "Tesla X2090" }, + { 0x10de, 0x0891, "Tesla X2090" }, + { 0x10de, 0x0974, "Tesla X2090" }, + { 0x10de, 0x098d, "Tesla X2090" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1096[] = { + { 0x10de, 0x0911, "Tesla C2050" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1140[] = { + { 0x1019, 0x999f, "GeForce GT 720M" }, + { 0x1025, 0x0600, "GeForce GT 620M" }, + { 0x1025, 0x0606, "GeForce GT 620M" }, + { 0x1025, 0x064a, "GeForce GT 620M" }, + { 0x1025, 0x064c, "GeForce GT 620M" }, + { 0x1025, 0x067a, "GeForce GT 620M" }, + { 0x1025, 0x0680, "GeForce GT 620M" }, + { 0x1025, 0x0686, "GeForce 710M" }, + { 0x1025, 0x0689, "GeForce 710M" }, + { 0x1025, 0x068b, "GeForce 710M" }, + { 0x1025, 0x068d, "GeForce 710M" }, + { 0x1025, 0x068e, "GeForce 710M" }, + { 0x1025, 0x0691, "GeForce 710M" }, + { 0x1025, 0x0692, "GeForce GT 620M" }, + { 0x1025, 0x0694, "GeForce GT 620M" }, + { 0x1025, 0x0702, "GeForce GT 620M" }, + { 0x1025, 0x0719, "GeForce GT 620M" }, + { 0x1025, 0x0725, "GeForce GT 620M" }, + { 0x1025, 0x0728, "GeForce GT 620M" }, + { 0x1025, 0x072b, "GeForce GT 620M" }, + { 0x1025, 0x072e, "GeForce GT 620M" }, + { 0x1025, 0x0732, "GeForce GT 620M" }, + { 0x1025, 0x0763, "GeForce GT 720M" }, + { 0x1025, 0x0773, "GeForce 710M" }, + { 0x1025, 0x0774, "GeForce 710M" }, + { 0x1025, 0x0776, "GeForce GT 720M" }, + { 0x1025, 0x077a, "GeForce 710M" }, + { 0x1025, 0x077b, "GeForce 710M" }, + { 0x1025, 0x077c, "GeForce 710M" }, + { 0x1025, 0x077d, "GeForce 710M" }, + { 0x1025, 0x077e, "GeForce 710M" }, + { 0x1025, 0x077f, "GeForce 710M" }, + { 0x1025, 0x0781, "GeForce GT 720M" }, + { 0x1025, 0x0798, "GeForce GT 720M" }, + { 0x1025, 0x0799, "GeForce GT 720M" }, + { 0x1025, 0x079b, "GeForce GT 720M" }, + { 0x1025, 0x079c, "GeForce GT 720M" }, + { 0x1025, 0x0807, "GeForce GT 720M" }, + { 0x1025, 0x0821, "GeForce 820M" }, + { 0x1025, 0x0823, "GeForce GT 720M" }, + { 0x1025, 0x0830, "GeForce GT 720M" }, + { 0x1025, 0x0833, "GeForce GT 720M" }, + { 0x1025, 0x0837, "GeForce GT 720M" }, + { 0x1025, 0x083e, "GeForce 820M" }, + { 0x1025, 0x0841, "GeForce 710M" }, + { 0x1025, 0x0853, "GeForce 820M" }, + { 0x1025, 0x0854, "GeForce 820M" }, + { 0x1025, 0x0855, "GeForce 820M" }, + { 0x1025, 0x0856, "GeForce 820M" }, + { 0x1025, 0x0857, "GeForce 820M" }, + { 0x1025, 0x0858, "GeForce 820M" }, + { 0x1025, 0x0863, "GeForce 820M" }, + { 0x1025, 0x0868, "GeForce 820M" }, + { 0x1025, 0x0869, "GeForce 810M" }, + { 0x1025, 0x0873, "GeForce 820M" }, + { 0x1025, 0x0878, "GeForce 820M" }, + { 0x1025, 0x087b, "GeForce 820M" }, + { 0x1025, 0x087f, "GeForce 820M" }, + { 0x1025, 0x0881, "GeForce 820M" }, + { 0x1025, 0x0885, "GeForce 820M" }, + { 0x1025, 0x088a, "GeForce 820M" }, + { 0x1025, 0x089b, "GeForce 820M" }, + { 0x1025, 0x0921, "GeForce 820M" }, + { 0x1025, 0x092e, "GeForce 810M" }, + { 0x1025, 0x092f, "GeForce 820M" }, + { 0x1025, 0x0932, "GeForce 820M" }, + { 0x1025, 0x093a, "GeForce 820M" }, + { 0x1025, 0x093c, "GeForce 820M" }, + { 0x1025, 0x093f, "GeForce 820M" }, + { 0x1025, 0x0941, "GeForce 820M" }, + { 0x1025, 0x0945, "GeForce 820M" }, + { 0x1025, 0x0954, "GeForce 820M" }, + { 0x1025, 0x0965, "GeForce 820M" }, + { 0x1028, 0x054d, "GeForce GT 630M" }, + { 0x1028, 0x054e, "GeForce GT 630M" }, + { 0x1028, 0x0554, "GeForce GT 620M" }, + { 0x1028, 0x0557, "GeForce GT 620M" }, + { 0x1028, 0x0562, "GeForce GT625M" }, + { 0x1028, 0x0565, "GeForce GT 630M" }, + { 0x1028, 0x0568, "GeForce GT 630M" }, + { 0x1028, 0x0590, "GeForce GT 630M" }, + { 0x1028, 0x0592, "GeForce GT625M" }, + { 0x1028, 0x0594, "GeForce GT625M" }, + { 0x1028, 0x0595, "GeForce GT625M" }, + { 0x1028, 0x05a2, "GeForce GT625M" }, + { 0x1028, 0x05b1, "GeForce GT625M" }, + { 0x1028, 0x05b3, "GeForce GT625M" }, + { 0x1028, 0x05da, "GeForce GT 630M" }, + { 0x1028, 0x05de, "GeForce GT 720M" }, + { 0x1028, 0x05e0, "GeForce GT 720M" }, + { 0x1028, 0x05e8, "GeForce GT 630M" }, + { 0x1028, 0x05f4, "GeForce GT 720M" }, + { 0x1028, 0x060f, "GeForce GT 720M" }, + { 0x1028, 0x062f, "GeForce GT 720M" }, + { 0x1028, 0x064e, "GeForce 820M" }, + { 0x1028, 0x0652, "GeForce 820M" }, + { 0x1028, 0x0653, "GeForce 820M" }, + { 0x1028, 0x0655, "GeForce 820M" }, + { 0x1028, 0x065e, "GeForce 820M" }, + { 0x1028, 0x0662, "GeForce 820M" }, + { 0x1028, 0x068d, "GeForce 820M" }, + { 0x1028, 0x06ad, "GeForce 820M" }, + { 0x1028, 0x06ae, "GeForce 820M" }, + { 0x1028, 0x06af, "GeForce 820M" }, + { 0x1028, 0x06b0, "GeForce 820M" }, + { 0x1028, 0x06c0, "GeForce 820M" }, + { 0x1028, 0x06c1, "GeForce 820M" }, + { 0x103c, 0x18ef, "GeForce GT 630M" }, + { 0x103c, 0x18f9, "GeForce GT 630M" }, + { 0x103c, 0x18fb, "GeForce GT 630M" }, + { 0x103c, 0x18fd, "GeForce GT 630M" }, + { 0x103c, 0x18ff, "GeForce GT 630M" }, + { 0x103c, 0x218a, "GeForce 820M" }, + { 0x103c, 0x21bb, "GeForce 820M" }, + { 0x103c, 0x21bc, "GeForce 820M" }, + { 0x103c, 0x220e, "GeForce 820M" }, + { 0x103c, 0x2210, "GeForce 820M" }, + { 0x103c, 0x2212, "GeForce 820M" }, + { 0x103c, 0x2214, "GeForce 820M" }, + { 0x103c, 0x2218, "GeForce 820M" }, + { 0x103c, 0x225b, "GeForce 820M" }, + { 0x103c, 0x225d, "GeForce 820M" }, + { 0x103c, 0x226d, "GeForce 820M" }, + { 0x103c, 0x226f, "GeForce 820M" }, + { 0x103c, 0x22d2, "GeForce 820M" }, + { 0x103c, 0x22d9, "GeForce 820M" }, + { 0x103c, 0x2335, "GeForce 820M" }, + { 0x103c, 0x2337, "GeForce 820M" }, + { 0x103c, 0x2aef, "GeForce GT 720A" }, + { 0x103c, 0x2af9, "GeForce 710A" }, + { 0x1043, 0x10dd, "NVS 5200M" }, + { 0x1043, 0x10ed, "NVS 5200M" }, + { 0x1043, 0x11fd, "GeForce GT 720M" }, + { 0x1043, 0x124d, "GeForce GT 720M" }, + { 0x1043, 0x126d, "GeForce GT 720M" }, + { 0x1043, 0x131d, "GeForce GT 720M" }, + { 0x1043, 0x13fd, "GeForce GT 720M" }, + { 0x1043, 0x14c7, "GeForce GT 720M" }, + { 0x1043, 0x1507, "GeForce GT 620M" }, + { 0x1043, 0x15ad, "GeForce 820M" }, + { 0x1043, 0x15ed, "GeForce 820M" }, + { 0x1043, 0x160d, "GeForce 820M" }, + { 0x1043, 0x163d, "GeForce 820M" }, + { 0x1043, 0x165d, "GeForce 820M" }, + { 0x1043, 0x166d, "GeForce 820M" }, + { 0x1043, 0x16cd, "GeForce 820M" }, + { 0x1043, 0x16dd, "GeForce 820M" }, + { 0x1043, 0x170d, "GeForce 820M" }, + { 0x1043, 0x176d, "GeForce 820M" }, + { 0x1043, 0x178d, "GeForce 820M" }, + { 0x1043, 0x179d, "GeForce 820M" }, + { 0x1043, 0x2132, "GeForce GT 620M" }, + { 0x1043, 0x2136, "NVS 5200M" }, + { 0x1043, 0x21ba, "GeForce GT 720M" }, + { 0x1043, 0x21fa, "GeForce GT 720M" }, + { 0x1043, 0x220a, "GeForce GT 720M" }, + { 0x1043, 0x221a, "GeForce GT 720M" }, + { 0x1043, 0x223a, "GeForce GT 710M" }, + { 0x1043, 0x224a, "GeForce GT 710M" }, + { 0x1043, 0x227a, "GeForce 820M" }, + { 0x1043, 0x228a, "GeForce 820M" }, + { 0x1043, 0x22fa, "GeForce 820M" }, + { 0x1043, 0x232a, "GeForce 820M" }, + { 0x1043, 0x233a, "GeForce 820M" }, + { 0x1043, 0x235a, "GeForce 820M" }, + { 0x1043, 0x236a, "GeForce 820M" }, + { 0x1043, 0x238a, "GeForce 820M" }, + { 0x1043, 0x8595, "GeForce GT 720M" }, + { 0x1043, 0x85ea, "GeForce GT 720M" }, + { 0x1043, 0x85eb, "GeForce 820M" }, + { 0x1043, 0x85ec, "GeForce 820M" }, + { 0x1043, 0x85ee, "GeForce GT 720M" }, + { 0x1043, 0x85f3, "GeForce 820M" }, + { 0x1043, 0x860e, "GeForce 820M" }, + { 0x1043, 0x861a, "GeForce 820M" }, + { 0x1043, 0x861b, "GeForce 820M" }, + { 0x1043, 0x8628, "GeForce 820M" }, + { 0x1043, 0x8643, "GeForce 820M" }, + { 0x1043, 0x864c, "GeForce 820M" }, + { 0x1043, 0x8652, "GeForce 820M" }, + { 0x1043, 0x8660, "GeForce 820M" }, + { 0x1043, 0x8661, "GeForce 820M" }, + { 0x105b, 0x0dac, "GeForce GT 720M" }, + { 0x105b, 0x0dad, "GeForce GT 720M" }, + { 0x105b, 0x0ef3, "GeForce GT 720M" }, + { 0x10cf, 0x17f5, "GeForce GT 720M" }, + { 0x1179, 0xfa01, "GeForce 710M" }, + { 0x1179, 0xfa02, "GeForce 710M" }, + { 0x1179, 0xfa03, "GeForce 710M" }, + { 0x1179, 0xfa05, "GeForce 710M" }, + { 0x1179, 0xfa11, "GeForce 710M" }, + { 0x1179, 0xfa13, "GeForce 710M" }, + { 0x1179, 0xfa18, "GeForce 710M" }, + { 0x1179, 0xfa19, "GeForce 710M" }, + { 0x1179, 0xfa21, "GeForce 710M" }, + { 0x1179, 0xfa23, "GeForce 710M" }, + { 0x1179, 0xfa2a, "GeForce 710M" }, + { 0x1179, 0xfa32, "GeForce 710M" }, + { 0x1179, 0xfa33, "GeForce 710M" }, + { 0x1179, 0xfa36, "GeForce 710M" }, + { 0x1179, 0xfa38, "GeForce 710M" }, + { 0x1179, 0xfa42, "GeForce 710M" }, + { 0x1179, 0xfa43, "GeForce 710M" }, + { 0x1179, 0xfa45, "GeForce 710M" }, + { 0x1179, 0xfa47, "GeForce 710M" }, + { 0x1179, 0xfa49, "GeForce 710M" }, + { 0x1179, 0xfa58, "GeForce 710M" }, + { 0x1179, 0xfa59, "GeForce 710M" }, + { 0x1179, 0xfa88, "GeForce 710M" }, + { 0x1179, 0xfa89, "GeForce 710M" }, + { 0x144d, 0xb092, "GeForce GT 620M" }, + { 0x144d, 0xc0d5, "GeForce GT 630M" }, + { 0x144d, 0xc0d7, "GeForce GT 620M" }, + { 0x144d, 0xc0e2, "NVS 5200M" }, + { 0x144d, 0xc0e3, "NVS 5200M" }, + { 0x144d, 0xc0e4, "NVS 5200M" }, + { 0x144d, 0xc10d, "GeForce 820M" }, + { 0x144d, 0xc652, "GeForce GT 620M" }, + { 0x144d, 0xc709, "GeForce 710M" }, + { 0x144d, 0xc711, "GeForce 710M" }, + { 0x144d, 0xc736, "GeForce 710M" }, + { 0x144d, 0xc737, "GeForce 710M" }, + { 0x144d, 0xc745, "GeForce 820M" }, + { 0x144d, 0xc750, "GeForce 820M" }, + { 0x1462, 0x10b8, "GeForce GT 710M" }, + { 0x1462, 0x10e9, "GeForce GT 720M" }, + { 0x1462, 0x1116, "GeForce 820M" }, + { 0x1462, 0xaa33, "GeForce 720M" }, + { 0x1462, 0xaaa2, "GeForce GT 720M" }, + { 0x1462, 0xaaa3, "GeForce 820M" }, + { 0x1462, 0xacb2, "GeForce GT 720M" }, + { 0x1462, 0xacc1, "GeForce GT 720M" }, + { 0x1462, 0xae61, "GeForce 720M" }, + { 0x1462, 0xae65, "GeForce GT 720M" }, + { 0x1462, 0xae6a, "GeForce 820M" }, + { 0x1462, 0xae71, "GeForce GT 720M" }, + { 0x14c0, 0x0083, "GeForce 820M" }, + { 0x152d, 0x0926, "GeForce 620M" }, + { 0x152d, 0x0982, "GeForce GT 630M" }, + { 0x152d, 0x0983, "GeForce GT 630M" }, + { 0x152d, 0x1005, "GeForce GT820M" }, + { 0x152d, 0x1012, "GeForce 710M" }, + { 0x152d, 0x1019, "GeForce 820M" }, + { 0x152d, 0x1030, "GeForce GT 630M" }, + { 0x152d, 0x1055, "GeForce 710M" }, + { 0x152d, 0x1067, "GeForce GT 720M" }, + { 0x152d, 0x1092, "GeForce 820M" }, + { 0x17aa, 0x2200, "NVS 5200M" }, + { 0x17aa, 0x2213, "GeForce GT 720M" }, + { 0x17aa, 0x2220, "GeForce GT 720M" }, + { 0x17aa, 0x309c, "GeForce GT 720A" }, + { 0x17aa, 0x30b4, "GeForce 820A" }, + { 0x17aa, 0x30b7, "GeForce 720A" }, + { 0x17aa, 0x30e4, "GeForce 820A" }, + { 0x17aa, 0x361b, "GeForce 820A" }, + { 0x17aa, 0x361c, "GeForce 820A" }, + { 0x17aa, 0x361d, "GeForce 820A" }, + { 0x17aa, 0x3656, "GeForce GT620M" }, + { 0x17aa, 0x365a, "GeForce 705M" }, + { 0x17aa, 0x365e, "GeForce 800M" }, + { 0x17aa, 0x3661, "GeForce 820A" }, + { 0x17aa, 0x366c, "GeForce 800M" }, + { 0x17aa, 0x3685, "GeForce 800M" }, + { 0x17aa, 0x3686, "GeForce 800M" }, + { 0x17aa, 0x3687, "GeForce 705A" }, + { 0x17aa, 0x3696, "GeForce 820A" }, + { 0x17aa, 0x369b, "GeForce 820A" }, + { 0x17aa, 0x369c, "GeForce 820A" }, + { 0x17aa, 0x369d, "GeForce 820A" }, + { 0x17aa, 0x369e, "GeForce 820A" }, + { 0x17aa, 0x36a6, "GeForce 820A" }, + { 0x17aa, 0x36a7, "GeForce 820A" }, + { 0x17aa, 0x36a9, "GeForce 820A" }, + { 0x17aa, 0x36af, "GeForce 820A" }, + { 0x17aa, 0x36b0, "GeForce 820A" }, + { 0x17aa, 0x36b6, "GeForce 820A" }, + { 0x17aa, 0x3800, "GeForce GT 720M" }, + { 0x17aa, 0x3801, "GeForce GT 720M" }, + { 0x17aa, 0x3802, "GeForce GT 720M" }, + { 0x17aa, 0x3803, "GeForce GT 720M" }, + { 0x17aa, 0x3804, "GeForce GT 720M" }, + { 0x17aa, 0x3806, "GeForce GT 720M" }, + { 0x17aa, 0x3808, "GeForce GT 720M" }, + { 0x17aa, 0x380d, "GeForce 820M" }, + { 0x17aa, 0x380e, "GeForce 820M" }, + { 0x17aa, 0x380f, "GeForce 820M" }, + { 0x17aa, 0x3811, "GeForce 820M" }, + { 0x17aa, 0x3812, "GeForce 820M" }, + { 0x17aa, 0x3813, "GeForce 820M" }, + { 0x17aa, 0x3816, "GeForce 820M" }, + { 0x17aa, 0x3817, "GeForce 820M" }, + { 0x17aa, 0x3818, "GeForce 820M" }, + { 0x17aa, 0x381a, "GeForce 820M" }, + { 0x17aa, 0x381c, "GeForce 820M" }, + { 0x17aa, 0x381d, "GeForce 820M" }, + { 0x17aa, 0x3901, "GeForce 610M" }, + { 0x17aa, 0x3902, "GeForce 710M" }, + { 0x17aa, 0x3903, "GeForce 710M" }, + { 0x17aa, 0x3904, "GeForce GT 625M" }, + { 0x17aa, 0x3905, "GeForce GT 720M" }, + { 0x17aa, 0x3907, "GeForce 820M" }, + { 0x17aa, 0x3910, "GeForce GT 720M" }, + { 0x17aa, 0x3912, "GeForce GT 720M" }, + { 0x17aa, 0x3913, "GeForce 820M" }, + { 0x17aa, 0x3915, "GeForce 820M" }, + { 0x17aa, 0x3983, "GeForce 610M" }, + { 0x17aa, 0x5001, "GeForce 610M" }, + { 0x17aa, 0x5003, "GeForce GT 720M" }, + { 0x17aa, 0x5005, "GeForce 705M" }, + { 0x17aa, 0x500d, "GeForce GT 620M" }, + { 0x17aa, 0x5014, "GeForce 710M" }, + { 0x17aa, 0x5017, "GeForce 710M" }, + { 0x17aa, 0x5019, "GeForce 710M" }, + { 0x17aa, 0x501a, "GeForce 710M" }, + { 0x17aa, 0x501f, "GeForce GT 720M" }, + { 0x17aa, 0x5025, "GeForce 710M" }, + { 0x17aa, 0x5027, "GeForce 710M" }, + { 0x17aa, 0x502a, "GeForce 710M" }, + { 0x17aa, 0x502b, "GeForce GT 720M" }, + { 0x17aa, 0x502d, "GeForce 710M" }, + { 0x17aa, 0x502e, "GeForce GT 720M" }, + { 0x17aa, 0x502f, "GeForce GT 720M" }, + { 0x17aa, 0x5030, "GeForce 705M" }, + { 0x17aa, 0x5031, "GeForce 705M" }, + { 0x17aa, 0x5032, "GeForce 820M" }, + { 0x17aa, 0x5033, "GeForce 820M" }, + { 0x17aa, 0x503e, "GeForce 710M" }, + { 0x17aa, 0x503f, "GeForce 820M" }, + { 0x17aa, 0x5040, "GeForce 820M" }, + { 0x1854, 0x0177, "GeForce 710M" }, + { 0x1854, 0x0180, "GeForce 710M" }, + { 0x1854, 0x0190, "GeForce GT 720M" }, + { 0x1854, 0x0192, "GeForce GT 720M" }, + { 0x1854, 0x0224, "GeForce 820M" }, + { 0x1b0a, 0x20dd, "GeForce GT 620M" }, + { 0x1b0a, 0x20df, "GeForce GT 620M" }, + { 0x1b0a, 0x210e, "GeForce 820M" }, + { 0x1b0a, 0x2202, "GeForce GT 720M" }, + { 0x1b0a, 0x90d7, "GeForce 820M" }, + { 0x1b0a, 0x90dd, "GeForce 820M" }, + { 0x1b50, 0x5530, "GeForce 820M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1185[] = { + { 0x10de, 0x106f, "GeForce GTX 760" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1189[] = { + { 0x10de, 0x1074, "GeForce GTX 760 Ti OEM" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1199[] = { + { 0x1458, 0xd001, "GeForce GTX 760" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_11e3[] = { + { 0x17aa, 0x3683, "GeForce GTX 760A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1247[] = { + { 0x1043, 0x212a, "GeForce GT 635M" }, + { 0x1043, 0x212b, "GeForce GT 635M" }, + { 0x1043, 0x212c, "GeForce GT 635M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_124d[] = { + { 0x1462, 0x10cc, "GeForce GT 635M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1290[] = { + { 0x103c, 0x2afa, "GeForce 730A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1292[] = { + { 0x17aa, 0x3675, "GeForce GT 740A" }, + { 0x17aa, 0x367c, "GeForce GT 740A" }, + { 0x17aa, 0x3684, "GeForce GT 740A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1295[] = { + { 0x103c, 0x2b0d, "GeForce 710A" }, + { 0x103c, 0x2b0f, "GeForce 710A" }, + { 0x103c, 0x2b20, "GeForce 810A" }, + { 0x103c, 0x2b21, "GeForce 810A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1299[] = { + { 0x17aa, 0x369b, "GeForce 920A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1340[] = { + { 0x103c, 0x2b2b, "GeForce 830A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1341[] = { + { 0x17aa, 0x3697, "GeForce 840A" }, + { 0x17aa, 0x3699, "GeForce 840A" }, + { 0x17aa, 0x369c, "GeForce 840A" }, + { 0x17aa, 0x36af, "GeForce 840A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1346[] = { + { 0x17aa, 0x30ba, "GeForce 930A" }, + { 0x17aa, 0x362c, "GeForce 930A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1347[] = { + { 0x17aa, 0x36b9, "GeForce 940A" }, + { 0x17aa, 0x36ba, "GeForce 940A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_137a[] = { + { 0x17aa, 0x2225, "Quadro K620M" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_137d[] = { + { 0x17aa, 0x3699, "GeForce 940A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1391[] = { + { 0x17aa, 0x3697, "GeForce GTX 850A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_1392[] = { + { 0x1028, 0x066a, "GeForce GPU" }, + { 0x1043, 0x861e, "GeForce GTX 750 Ti" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_139a[] = { + { 0x17aa, 0x36b9, "GeForce GTX 950A" }, + {} +}; + +static const struct nvkm_device_pci_vendor +nvkm_device_pci_10de_139b[] = { + { 0x1028, 0x06a3, "GeForce GTX 860M" }, + { 0x19da, 0xc248, "GeForce GTX 750 Ti" }, + {} +}; + +static const struct nvkm_device_pci_device +nvkm_device_pci_10de[] = { + { 0x0020, "RIVA TNT" }, + { 0x0028, "RIVA TNT2/TNT2 Pro" }, + { 0x0029, "RIVA TNT2 Ultra" }, + { 0x002c, "Vanta/Vanta LT" }, + { 0x002d, "RIVA TNT2 Model 64/Model 64 Pro" }, + { 0x0040, "GeForce 6800 Ultra" }, + { 0x0041, "GeForce 6800" }, + { 0x0042, "GeForce 6800 LE" }, + { 0x0043, "GeForce 6800 XE" }, + { 0x0044, "GeForce 6800 XT" }, + { 0x0045, "GeForce 6800 GT" }, + { 0x0046, "GeForce 6800 GT" }, + { 0x0047, "GeForce 6800 GS" }, + { 0x0048, "GeForce 6800 XT" }, + { 0x004e, "Quadro FX 4000" }, + { 0x0090, "GeForce 7800 GTX" }, + { 0x0091, "GeForce 7800 GTX" }, + { 0x0092, "GeForce 7800 GT" }, + { 0x0093, "GeForce 7800 GS" }, + { 0x0095, "GeForce 7800 SLI" }, + { 0x0098, "GeForce Go 7800" }, + { 0x0099, "GeForce Go 7800 GTX" }, + { 0x009d, "Quadro FX 4500" }, + { 0x00a0, "Aladdin TNT2" }, + { 0x00c0, "GeForce 6800 GS" }, + { 0x00c1, "GeForce 6800" }, + { 0x00c2, "GeForce 6800 LE" }, + { 0x00c3, "GeForce 6800 XT" }, + { 0x00c8, "GeForce Go 6800" }, + { 0x00c9, "GeForce Go 6800 Ultra" }, + { 0x00cc, "Quadro FX Go1400" }, + { 0x00cd, "Quadro FX 3450/4000 SDI" }, + { 0x00ce, "Quadro FX 1400" }, + { 0x00f1, "GeForce 6600 GT" }, + { 0x00f2, "GeForce 6600" }, + { 0x00f3, "GeForce 6200" }, + { 0x00f4, "GeForce 6600 LE" }, + { 0x00f5, "GeForce 7800 GS" }, + { 0x00f6, "GeForce 6800 GS" }, + { 0x00f8, "Quadro FX 3400/Quadro FX 4000" }, + { 0x00f9, "GeForce 6800 Ultra" }, + { 0x00fa, "GeForce PCX 5750" }, + { 0x00fb, "GeForce PCX 5900" }, + { 0x00fc, "Quadro FX 330/GeForce PCX 5300" }, + { 0x00fd, "Quadro FX 330/Quadro NVS 280 PCI-E" }, + { 0x00fe, "Quadro FX 1300" }, + { 0x0100, "GeForce 256" }, + { 0x0101, "GeForce DDR" }, + { 0x0103, "Quadro" }, + { 0x0110, "GeForce2 MX/MX 400" }, + { 0x0111, "GeForce2 MX 100/200" }, + { 0x0112, "GeForce2 Go" }, + { 0x0113, "Quadro2 MXR/EX/Go" }, + { 0x0140, "GeForce 6600 GT" }, + { 0x0141, "GeForce 6600" }, + { 0x0142, "GeForce 6600 LE" }, + { 0x0143, "GeForce 6600 VE" }, + { 0x0144, "GeForce Go 6600" }, + { 0x0145, "GeForce 6610 XL" }, + { 0x0146, "GeForce Go 6600 TE/6200 TE" }, + { 0x0147, "GeForce 6700 XL" }, + { 0x0148, "GeForce Go 6600" }, + { 0x0149, "GeForce Go 6600 GT" }, + { 0x014a, "Quadro NVS 440" }, + { 0x014c, "Quadro FX 540M" }, + { 0x014d, "Quadro FX 550" }, + { 0x014e, "Quadro FX 540" }, + { 0x014f, "GeForce 6200" }, + { 0x0150, "GeForce2 GTS/GeForce2 Pro" }, + { 0x0151, "GeForce2 Ti" }, + { 0x0152, "GeForce2 Ultra" }, + { 0x0153, "Quadro2 Pro" }, + { 0x0160, "GeForce 6500" }, + { 0x0161, "GeForce 6200 TurboCache(TM)" }, + { 0x0162, "GeForce 6200SE TurboCache(TM)" }, + { 0x0163, "GeForce 6200 LE" }, + { 0x0164, "GeForce Go 6200" }, + { 0x0165, "Quadro NVS 285" }, + { 0x0166, "GeForce Go 6400" }, + { 0x0167, "GeForce Go 6200" }, + { 0x0168, "GeForce Go 6400" }, + { 0x0169, "GeForce 6250" }, + { 0x016a, "GeForce 7100 GS" }, + { 0x0170, "GeForce4 MX 460" }, + { 0x0171, "GeForce4 MX 440" }, + { 0x0172, "GeForce4 MX 420" }, + { 0x0173, "GeForce4 MX 440-SE" }, + { 0x0174, "GeForce4 440 Go" }, + { 0x0175, "GeForce4 420 Go" }, + { 0x0176, "GeForce4 420 Go 32M" }, + { 0x0177, "GeForce4 460 Go" }, + { 0x0178, "Quadro4 550 XGL" }, + { 0x0179, "GeForce4 440 Go 64M" }, + { 0x017a, "Quadro NVS 400" }, + { 0x017c, "Quadro4 500 GoGL" }, + { 0x017d, "GeForce4 410 Go 16M" }, + { 0x0181, "GeForce4 MX 440 with AGP8X" }, + { 0x0182, "GeForce4 MX 440SE with AGP8X" }, + { 0x0183, "GeForce4 MX 420 with AGP8X" }, + { 0x0185, "GeForce4 MX 4000" }, + { 0x0188, "Quadro4 580 XGL" }, + { 0x0189, "GeForce4 MX with AGP8X (Mac)", nvkm_device_pci_10de_0189 }, + { 0x018a, "Quadro NVS 280 SD" }, + { 0x018b, "Quadro4 380 XGL" }, + { 0x018c, "Quadro NVS 50 PCI" }, + { 0x0191, "GeForce 8800 GTX" }, + { 0x0193, "GeForce 8800 GTS" }, + { 0x0194, "GeForce 8800 Ultra" }, + { 0x0197, "Tesla C870" }, + { 0x019d, "Quadro FX 5600" }, + { 0x019e, "Quadro FX 4600" }, + { 0x01a0, "GeForce2 Integrated GPU" }, + { 0x01d0, "GeForce 7350 LE" }, + { 0x01d1, "GeForce 7300 LE" }, + { 0x01d2, "GeForce 7550 LE" }, + { 0x01d3, "GeForce 7300 SE/7200 GS" }, + { 0x01d6, "GeForce Go 7200" }, + { 0x01d7, "GeForce Go 7300" }, + { 0x01d8, "GeForce Go 7400" }, + { 0x01da, "Quadro NVS 110M" }, + { 0x01db, "Quadro NVS 120M" }, + { 0x01dc, "Quadro FX 350M" }, + { 0x01dd, "GeForce 7500 LE" }, + { 0x01de, "Quadro FX 350" }, + { 0x01df, "GeForce 7300 GS" }, + { 0x01f0, "GeForce4 MX Integrated GPU", nvkm_device_pci_10de_01f0 }, + { 0x0200, "GeForce3" }, + { 0x0201, "GeForce3 Ti 200" }, + { 0x0202, "GeForce3 Ti 500" }, + { 0x0203, "Quadro DCC" }, + { 0x0211, "GeForce 6800" }, + { 0x0212, "GeForce 6800 LE" }, + { 0x0215, "GeForce 6800 GT" }, + { 0x0218, "GeForce 6800 XT" }, + { 0x0221, "GeForce 6200" }, + { 0x0222, "GeForce 6200 A-LE" }, + { 0x0240, "GeForce 6150" }, + { 0x0241, "GeForce 6150 LE" }, + { 0x0242, "GeForce 6100" }, + { 0x0244, "GeForce Go 6150" }, + { 0x0245, "Quadro NVS 210S / GeForce 6150LE" }, + { 0x0247, "GeForce Go 6100" }, + { 0x0250, "GeForce4 Ti 4600" }, + { 0x0251, "GeForce4 Ti 4400" }, + { 0x0253, "GeForce4 Ti 4200" }, + { 0x0258, "Quadro4 900 XGL" }, + { 0x0259, "Quadro4 750 XGL" }, + { 0x025b, "Quadro4 700 XGL" }, + { 0x0280, "GeForce4 Ti 4800" }, + { 0x0281, "GeForce4 Ti 4200 with AGP8X" }, + { 0x0282, "GeForce4 Ti 4800 SE" }, + { 0x0286, "GeForce4 4200 Go" }, + { 0x0288, "Quadro4 980 XGL" }, + { 0x0289, "Quadro4 780 XGL" }, + { 0x028c, "Quadro4 700 GoGL" }, + { 0x0290, "GeForce 7900 GTX" }, + { 0x0291, "GeForce 7900 GT/GTO" }, + { 0x0292, "GeForce 7900 GS" }, + { 0x0293, "GeForce 7950 GX2" }, + { 0x0294, "GeForce 7950 GX2" }, + { 0x0295, "GeForce 7950 GT" }, + { 0x0297, "GeForce Go 7950 GTX" }, + { 0x0298, "GeForce Go 7900 GS" }, + { 0x0299, "Quadro NVS 510M" }, + { 0x029a, "Quadro FX 2500M" }, + { 0x029b, "Quadro FX 1500M" }, + { 0x029c, "Quadro FX 5500" }, + { 0x029d, "Quadro FX 3500" }, + { 0x029e, "Quadro FX 1500" }, + { 0x029f, "Quadro FX 4500 X2" }, + { 0x02e0, "GeForce 7600 GT" }, + { 0x02e1, "GeForce 7600 GS" }, + { 0x02e2, "GeForce 7300 GT" }, + { 0x02e3, "GeForce 7900 GS" }, + { 0x02e4, "GeForce 7950 GT" }, + { 0x0301, "GeForce FX 5800 Ultra" }, + { 0x0302, "GeForce FX 5800" }, + { 0x0308, "Quadro FX 2000" }, + { 0x0309, "Quadro FX 1000" }, + { 0x0311, "GeForce FX 5600 Ultra" }, + { 0x0312, "GeForce FX 5600" }, + { 0x0314, "GeForce FX 5600XT" }, + { 0x031a, "GeForce FX Go5600" }, + { 0x031b, "GeForce FX Go5650" }, + { 0x031c, "Quadro FX Go700" }, + { 0x0320, "GeForce FX 5200" }, + { 0x0321, "GeForce FX 5200 Ultra" }, + { 0x0322, "GeForce FX 5200", nvkm_device_pci_10de_0322 }, + { 0x0323, "GeForce FX 5200LE" }, + { 0x0324, "GeForce FX Go5200" }, + { 0x0325, "GeForce FX Go5250" }, + { 0x0326, "GeForce FX 5500" }, + { 0x0327, "GeForce FX 5100" }, + { 0x0328, "GeForce FX Go5200 32M/64M" }, + { 0x032a, "Quadro NVS 55/280 PCI" }, + { 0x032b, "Quadro FX 500/FX 600" }, + { 0x032c, "GeForce FX Go53xx" }, + { 0x032d, "GeForce FX Go5100" }, + { 0x0330, "GeForce FX 5900 Ultra" }, + { 0x0331, "GeForce FX 5900" }, + { 0x0332, "GeForce FX 5900XT" }, + { 0x0333, "GeForce FX 5950 Ultra" }, + { 0x0334, "GeForce FX 5900ZT" }, + { 0x0338, "Quadro FX 3000" }, + { 0x033f, "Quadro FX 700" }, + { 0x0341, "GeForce FX 5700 Ultra" }, + { 0x0342, "GeForce FX 5700" }, + { 0x0343, "GeForce FX 5700LE" }, + { 0x0344, "GeForce FX 5700VE" }, + { 0x0347, "GeForce FX Go5700" }, + { 0x0348, "GeForce FX Go5700" }, + { 0x034c, "Quadro FX Go1000" }, + { 0x034e, "Quadro FX 1100" }, + { 0x038b, "GeForce 7650 GS" }, + { 0x0390, "GeForce 7650 GS" }, + { 0x0391, "GeForce 7600 GT" }, + { 0x0392, "GeForce 7600 GS" }, + { 0x0393, "GeForce 7300 GT" }, + { 0x0394, "GeForce 7600 LE" }, + { 0x0395, "GeForce 7300 GT" }, + { 0x0397, "GeForce Go 7700" }, + { 0x0398, "GeForce Go 7600" }, + { 0x0399, "GeForce Go 7600 GT" }, + { 0x039c, "Quadro FX 560M" }, + { 0x039e, "Quadro FX 560" }, + { 0x03d0, "GeForce 6150SE nForce 430" }, + { 0x03d1, "GeForce 6100 nForce 405" }, + { 0x03d2, "GeForce 6100 nForce 400" }, + { 0x03d5, "GeForce 6100 nForce 420" }, + { 0x03d6, "GeForce 7025 / nForce 630a" }, + { 0x0400, "GeForce 8600 GTS" }, + { 0x0401, "GeForce 8600 GT" }, + { 0x0402, "GeForce 8600 GT" }, + { 0x0403, "GeForce 8600 GS" }, + { 0x0404, "GeForce 8400 GS" }, + { 0x0405, "GeForce 9500M GS" }, + { 0x0406, "GeForce 8300 GS" }, + { 0x0407, "GeForce 8600M GT" }, + { 0x0408, "GeForce 9650M GS" }, + { 0x0409, "GeForce 8700M GT" }, + { 0x040a, "Quadro FX 370" }, + { 0x040b, "Quadro NVS 320M" }, + { 0x040c, "Quadro FX 570M" }, + { 0x040d, "Quadro FX 1600M" }, + { 0x040e, "Quadro FX 570" }, + { 0x040f, "Quadro FX 1700" }, + { 0x0410, "GeForce GT 330" }, + { 0x0420, "GeForce 8400 SE" }, + { 0x0421, "GeForce 8500 GT" }, + { 0x0422, "GeForce 8400 GS" }, + { 0x0423, "GeForce 8300 GS" }, + { 0x0424, "GeForce 8400 GS" }, + { 0x0425, "GeForce 8600M GS" }, + { 0x0426, "GeForce 8400M GT" }, + { 0x0427, "GeForce 8400M GS" }, + { 0x0428, "GeForce 8400M G" }, + { 0x0429, "Quadro NVS 140M" }, + { 0x042a, "Quadro NVS 130M" }, + { 0x042b, "Quadro NVS 135M" }, + { 0x042c, "GeForce 9400 GT" }, + { 0x042d, "Quadro FX 360M" }, + { 0x042e, "GeForce 9300M G" }, + { 0x042f, "Quadro NVS 290" }, + { 0x0531, "GeForce 7150M / nForce 630M" }, + { 0x0533, "GeForce 7000M / nForce 610M" }, + { 0x053a, "GeForce 7050 PV / nForce 630a" }, + { 0x053b, "GeForce 7050 PV / nForce 630a" }, + { 0x053e, "GeForce 7025 / nForce 630a" }, + { 0x05e0, "GeForce GTX 295" }, + { 0x05e1, "GeForce GTX 280" }, + { 0x05e2, "GeForce GTX 260" }, + { 0x05e3, "GeForce GTX 285" }, + { 0x05e6, "GeForce GTX 275" }, + { 0x05e7, "Tesla C1060", nvkm_device_pci_10de_05e7 }, + { 0x05ea, "GeForce GTX 260" }, + { 0x05eb, "GeForce GTX 295" }, + { 0x05ed, "Quadroplex 2200 D2" }, + { 0x05f8, "Quadroplex 2200 S4" }, + { 0x05f9, "Quadro CX" }, + { 0x05fd, "Quadro FX 5800" }, + { 0x05fe, "Quadro FX 4800" }, + { 0x05ff, "Quadro FX 3800" }, + { 0x0600, "GeForce 8800 GTS 512" }, + { 0x0601, "GeForce 9800 GT" }, + { 0x0602, "GeForce 8800 GT" }, + { 0x0603, "GeForce GT 230" }, + { 0x0604, "GeForce 9800 GX2" }, + { 0x0605, "GeForce 9800 GT" }, + { 0x0606, "GeForce 8800 GS" }, + { 0x0607, "GeForce GTS 240" }, + { 0x0608, "GeForce 9800M GTX" }, + { 0x0609, "GeForce 8800M GTS", nvkm_device_pci_10de_0609 }, + { 0x060a, "GeForce GTX 280M" }, + { 0x060b, "GeForce 9800M GT" }, + { 0x060c, "GeForce 8800M GTX" }, + { 0x060d, "GeForce 8800 GS" }, + { 0x060f, "GeForce GTX 285M" }, + { 0x0610, "GeForce 9600 GSO" }, + { 0x0611, "GeForce 8800 GT" }, + { 0x0612, "GeForce 9800 GTX/9800 GTX+" }, + { 0x0613, "GeForce 9800 GTX+" }, + { 0x0614, "GeForce 9800 GT" }, + { 0x0615, "GeForce GTS 250" }, + { 0x0617, "GeForce 9800M GTX" }, + { 0x0618, "GeForce GTX 260M" }, + { 0x0619, "Quadro FX 4700 X2" }, + { 0x061a, "Quadro FX 3700" }, + { 0x061b, "Quadro VX 200" }, + { 0x061c, "Quadro FX 3600M" }, + { 0x061d, "Quadro FX 2800M" }, + { 0x061e, "Quadro FX 3700M" }, + { 0x061f, "Quadro FX 3800M" }, + { 0x0621, "GeForce GT 230" }, + { 0x0622, "GeForce 9600 GT" }, + { 0x0623, "GeForce 9600 GS" }, + { 0x0625, "GeForce 9600 GSO 512" }, + { 0x0626, "GeForce GT 130" }, + { 0x0627, "GeForce GT 140" }, + { 0x0628, "GeForce 9800M GTS" }, + { 0x062a, "GeForce 9700M GTS" }, + { 0x062b, "GeForce 9800M GS" }, + { 0x062c, "GeForce 9800M GTS" }, + { 0x062d, "GeForce 9600 GT" }, + { 0x062e, "GeForce 9600 GT", nvkm_device_pci_10de_062e }, + { 0x0630, "GeForce 9700 S" }, + { 0x0631, "GeForce GTS 160M" }, + { 0x0632, "GeForce GTS 150M" }, + { 0x0635, "GeForce 9600 GSO" }, + { 0x0637, "GeForce 9600 GT" }, + { 0x0638, "Quadro FX 1800" }, + { 0x063a, "Quadro FX 2700M" }, + { 0x0640, "GeForce 9500 GT" }, + { 0x0641, "GeForce 9400 GT" }, + { 0x0643, "GeForce 9500 GT" }, + { 0x0644, "GeForce 9500 GS" }, + { 0x0645, "GeForce 9500 GS" }, + { 0x0646, "GeForce GT 120" }, + { 0x0647, "GeForce 9600M GT" }, + { 0x0648, "GeForce 9600M GS" }, + { 0x0649, "GeForce 9600M GT", nvkm_device_pci_10de_0649 }, + { 0x064a, "GeForce 9700M GT" }, + { 0x064b, "GeForce 9500M G" }, + { 0x064c, "GeForce 9650M GT" }, + { 0x0651, "GeForce G 110M" }, + { 0x0652, "GeForce GT 130M", nvkm_device_pci_10de_0652 }, + { 0x0653, "GeForce GT 120M" }, + { 0x0654, "GeForce GT 220M", nvkm_device_pci_10de_0654 }, + { 0x0655, NULL, nvkm_device_pci_10de_0655 }, + { 0x0656, NULL, nvkm_device_pci_10de_0656 }, + { 0x0658, "Quadro FX 380" }, + { 0x0659, "Quadro FX 580" }, + { 0x065a, "Quadro FX 1700M" }, + { 0x065b, "GeForce 9400 GT" }, + { 0x065c, "Quadro FX 770M" }, + { 0x06c0, "GeForce GTX 480" }, + { 0x06c4, "GeForce GTX 465" }, + { 0x06ca, "GeForce GTX 480M" }, + { 0x06cd, "GeForce GTX 470" }, + { 0x06d1, "Tesla C2050 / C2070", nvkm_device_pci_10de_06d1 }, + { 0x06d2, "Tesla M2070", nvkm_device_pci_10de_06d2 }, + { 0x06d8, "Quadro 6000" }, + { 0x06d9, "Quadro 5000" }, + { 0x06da, "Quadro 5000M" }, + { 0x06dc, "Quadro 6000" }, + { 0x06dd, "Quadro 4000" }, + { 0x06de, "Tesla T20 Processor", nvkm_device_pci_10de_06de }, + { 0x06df, "Tesla M2070-Q" }, + { 0x06e0, "GeForce 9300 GE" }, + { 0x06e1, "GeForce 9300 GS" }, + { 0x06e2, "GeForce 8400" }, + { 0x06e3, "GeForce 8400 SE" }, + { 0x06e4, "GeForce 8400 GS" }, + { 0x06e5, "GeForce 9300M GS" }, + { 0x06e6, "GeForce G100" }, + { 0x06e7, "GeForce 9300 SE" }, + { 0x06e8, "GeForce 9200M GS", nvkm_device_pci_10de_06e8 }, + { 0x06e9, "GeForce 9300M GS" }, + { 0x06ea, "Quadro NVS 150M" }, + { 0x06eb, "Quadro NVS 160M" }, + { 0x06ec, "GeForce G 105M" }, + { 0x06ef, "GeForce G 103M" }, + { 0x06f1, "GeForce G105M" }, + { 0x06f8, "Quadro NVS 420" }, + { 0x06f9, "Quadro FX 370 LP", nvkm_device_pci_10de_06f9 }, + { 0x06fa, "Quadro NVS 450" }, + { 0x06fb, "Quadro FX 370M" }, + { 0x06fd, "Quadro NVS 295" }, + { 0x06ff, "HICx16 + Graphics", nvkm_device_pci_10de_06ff }, + { 0x07e0, "GeForce 7150 / nForce 630i" }, + { 0x07e1, "GeForce 7100 / nForce 630i" }, + { 0x07e2, "GeForce 7050 / nForce 630i" }, + { 0x07e3, "GeForce 7050 / nForce 610i" }, + { 0x07e5, "GeForce 7050 / nForce 620i" }, + { 0x0840, "GeForce 8200M" }, + { 0x0844, "GeForce 9100M G" }, + { 0x0845, "GeForce 8200M G" }, + { 0x0846, "GeForce 9200" }, + { 0x0847, "GeForce 9100" }, + { 0x0848, "GeForce 8300" }, + { 0x0849, "GeForce 8200" }, + { 0x084a, "nForce 730a" }, + { 0x084b, "GeForce 9200" }, + { 0x084c, "nForce 980a/780a SLI" }, + { 0x084d, "nForce 750a SLI" }, + { 0x084f, "GeForce 8100 / nForce 720a" }, + { 0x0860, "GeForce 9400" }, + { 0x0861, "GeForce 9400" }, + { 0x0862, "GeForce 9400M G" }, + { 0x0863, "GeForce 9400M" }, + { 0x0864, "GeForce 9300" }, + { 0x0865, "ION" }, + { 0x0866, "GeForce 9400M G", nvkm_device_pci_10de_0866 }, + { 0x0867, "GeForce 9400" }, + { 0x0868, "nForce 760i SLI" }, + { 0x0869, "GeForce 9400" }, + { 0x086a, "GeForce 9400" }, + { 0x086c, "GeForce 9300 / nForce 730i" }, + { 0x086d, "GeForce 9200" }, + { 0x086e, "GeForce 9100M G" }, + { 0x086f, "GeForce 8200M G" }, + { 0x0870, "GeForce 9400M" }, + { 0x0871, "GeForce 9200" }, + { 0x0872, "GeForce G102M", nvkm_device_pci_10de_0872 }, + { 0x0873, "GeForce G102M", nvkm_device_pci_10de_0873 }, + { 0x0874, "ION" }, + { 0x0876, "ION" }, + { 0x087a, "GeForce 9400" }, + { 0x087d, "ION" }, + { 0x087e, "ION LE" }, + { 0x087f, "ION LE" }, + { 0x08a0, "GeForce 320M" }, + { 0x08a2, "GeForce 320M" }, + { 0x08a3, "GeForce 320M" }, + { 0x08a4, "GeForce 320M" }, + { 0x08a5, "GeForce 320M" }, + { 0x0a20, "GeForce GT 220" }, + { 0x0a22, "GeForce 315" }, + { 0x0a23, "GeForce 210" }, + { 0x0a26, "GeForce 405" }, + { 0x0a27, "GeForce 405" }, + { 0x0a28, "GeForce GT 230M" }, + { 0x0a29, "GeForce GT 330M" }, + { 0x0a2a, "GeForce GT 230M" }, + { 0x0a2b, "GeForce GT 330M" }, + { 0x0a2c, "NVS 5100M" }, + { 0x0a2d, "GeForce GT 320M" }, + { 0x0a32, "GeForce GT 415" }, + { 0x0a34, "GeForce GT 240M" }, + { 0x0a35, "GeForce GT 325M" }, + { 0x0a38, "Quadro 400" }, + { 0x0a3c, "Quadro FX 880M" }, + { 0x0a60, "GeForce G210" }, + { 0x0a62, "GeForce 205" }, + { 0x0a63, "GeForce 310" }, + { 0x0a64, "Second Generation ION" }, + { 0x0a65, "GeForce 210" }, + { 0x0a66, "GeForce 310" }, + { 0x0a67, "GeForce 315" }, + { 0x0a68, "GeForce G105M" }, + { 0x0a69, "GeForce G105M" }, + { 0x0a6a, "NVS 2100M" }, + { 0x0a6c, "NVS 3100M" }, + { 0x0a6e, "GeForce 305M", nvkm_device_pci_10de_0a6e }, + { 0x0a6f, "Second Generation ION" }, + { 0x0a70, "GeForce 310M", nvkm_device_pci_10de_0a70 }, + { 0x0a71, "GeForce 305M" }, + { 0x0a72, "GeForce 310M" }, + { 0x0a73, "GeForce 305M", nvkm_device_pci_10de_0a73 }, + { 0x0a74, "GeForce G210M", nvkm_device_pci_10de_0a74 }, + { 0x0a75, "GeForce 310M", nvkm_device_pci_10de_0a75 }, + { 0x0a76, "Second Generation ION" }, + { 0x0a78, "Quadro FX 380 LP" }, + { 0x0a7a, "GeForce 315M", nvkm_device_pci_10de_0a7a }, + { 0x0a7c, "Quadro FX 380M" }, + { 0x0ca0, "GeForce GT 330" }, + { 0x0ca2, "GeForce GT 320" }, + { 0x0ca3, "GeForce GT 240" }, + { 0x0ca4, "GeForce GT 340" }, + { 0x0ca5, "GeForce GT 220" }, + { 0x0ca7, "GeForce GT 330" }, + { 0x0ca8, "GeForce GTS 260M" }, + { 0x0ca9, "GeForce GTS 250M" }, + { 0x0cac, "GeForce GT 220" }, + { 0x0caf, "GeForce GT 335M" }, + { 0x0cb0, "GeForce GTS 350M" }, + { 0x0cb1, "GeForce GTS 360M" }, + { 0x0cbc, "Quadro FX 1800M" }, + { 0x0dc0, "GeForce GT 440" }, + { 0x0dc4, "GeForce GTS 450" }, + { 0x0dc5, "GeForce GTS 450" }, + { 0x0dc6, "GeForce GTS 450" }, + { 0x0dcd, "GeForce GT 555M" }, + { 0x0dce, "GeForce GT 555M" }, + { 0x0dd1, "GeForce GTX 460M" }, + { 0x0dd2, "GeForce GT 445M" }, + { 0x0dd3, "GeForce GT 435M" }, + { 0x0dd6, "GeForce GT 550M" }, + { 0x0dd8, "Quadro 2000", nvkm_device_pci_10de_0dd8 }, + { 0x0dda, "Quadro 2000M" }, + { 0x0de0, "GeForce GT 440" }, + { 0x0de1, "GeForce GT 430" }, + { 0x0de2, "GeForce GT 420" }, + { 0x0de3, "GeForce GT 635M" }, + { 0x0de4, "GeForce GT 520" }, + { 0x0de5, "GeForce GT 530" }, + { 0x0de7, "GeForce GT 610" }, + { 0x0de8, "GeForce GT 620M" }, + { 0x0de9, "GeForce GT 630M", nvkm_device_pci_10de_0de9 }, + { 0x0dea, "GeForce 610M", nvkm_device_pci_10de_0dea }, + { 0x0deb, "GeForce GT 555M" }, + { 0x0dec, "GeForce GT 525M" }, + { 0x0ded, "GeForce GT 520M" }, + { 0x0dee, "GeForce GT 415M" }, + { 0x0def, "NVS 5400M" }, + { 0x0df0, "GeForce GT 425M" }, + { 0x0df1, "GeForce GT 420M" }, + { 0x0df2, "GeForce GT 435M" }, + { 0x0df3, "GeForce GT 420M" }, + { 0x0df4, "GeForce GT 540M", nvkm_device_pci_10de_0df4 }, + { 0x0df5, "GeForce GT 525M" }, + { 0x0df6, "GeForce GT 550M" }, + { 0x0df7, "GeForce GT 520M" }, + { 0x0df8, "Quadro 600" }, + { 0x0df9, "Quadro 500M" }, + { 0x0dfa, "Quadro 1000M" }, + { 0x0dfc, "NVS 5200M" }, + { 0x0e22, "GeForce GTX 460" }, + { 0x0e23, "GeForce GTX 460 SE" }, + { 0x0e24, "GeForce GTX 460" }, + { 0x0e30, "GeForce GTX 470M" }, + { 0x0e31, "GeForce GTX 485M" }, + { 0x0e3a, "Quadro 3000M" }, + { 0x0e3b, "Quadro 4000M" }, + { 0x0f00, "GeForce GT 630" }, + { 0x0f01, "GeForce GT 620" }, + { 0x0f02, "GeForce GT 730" }, + { 0x0fc0, "GeForce GT 640" }, + { 0x0fc1, "GeForce GT 640" }, + { 0x0fc2, "GeForce GT 630" }, + { 0x0fc6, "GeForce GTX 650" }, + { 0x0fc8, "GeForce GT 740" }, + { 0x0fc9, "GeForce GT 730" }, + { 0x0fcd, "GeForce GT 755M" }, + { 0x0fce, "GeForce GT 640M LE" }, + { 0x0fd1, "GeForce GT 650M" }, + { 0x0fd2, "GeForce GT 640M", nvkm_device_pci_10de_0fd2 }, + { 0x0fd3, "GeForce GT 640M LE" }, + { 0x0fd4, "GeForce GTX 660M" }, + { 0x0fd5, "GeForce GT 650M" }, + { 0x0fd8, "GeForce GT 640M" }, + { 0x0fd9, "GeForce GT 645M" }, + { 0x0fdf, "GeForce GT 740M" }, + { 0x0fe0, "GeForce GTX 660M" }, + { 0x0fe1, "GeForce GT 730M" }, + { 0x0fe2, "GeForce GT 745M" }, + { 0x0fe3, "GeForce GT 745M", nvkm_device_pci_10de_0fe3 }, + { 0x0fe4, "GeForce GT 750M" }, + { 0x0fe9, "GeForce GT 750M" }, + { 0x0fea, "GeForce GT 755M" }, + { 0x0fec, "GeForce 710A" }, + { 0x0fef, "GRID K340" }, + { 0x0ff2, "GRID K1" }, + { 0x0ff3, "Quadro K420" }, + { 0x0ff6, "Quadro K1100M" }, + { 0x0ff8, "Quadro K500M" }, + { 0x0ff9, "Quadro K2000D" }, + { 0x0ffa, "Quadro K600" }, + { 0x0ffb, "Quadro K2000M" }, + { 0x0ffc, "Quadro K1000M" }, + { 0x0ffd, "NVS 510" }, + { 0x0ffe, "Quadro K2000" }, + { 0x0fff, "Quadro 410" }, + { 0x1001, "GeForce GTX TITAN Z" }, + { 0x1004, "GeForce GTX 780" }, + { 0x1005, "GeForce GTX TITAN" }, + { 0x1007, "GeForce GTX 780" }, + { 0x1008, "GeForce GTX 780 Ti" }, + { 0x100a, "GeForce GTX 780 Ti" }, + { 0x100c, "GeForce GTX TITAN Black" }, + { 0x1021, "Tesla K20Xm" }, + { 0x1022, "Tesla K20c" }, + { 0x1023, "Tesla K40m" }, + { 0x1024, "Tesla K40c" }, + { 0x1026, "Tesla K20s" }, + { 0x1027, "Tesla K40st" }, + { 0x1028, "Tesla K20m" }, + { 0x1029, "Tesla K40s" }, + { 0x102a, "Tesla K40t" }, + { 0x102d, "Tesla K80" }, + { 0x103a, "Quadro K6000" }, + { 0x103c, "Quadro K5200" }, + { 0x1040, "GeForce GT 520" }, + { 0x1042, "GeForce 510" }, + { 0x1048, "GeForce 605" }, + { 0x1049, "GeForce GT 620" }, + { 0x104a, "GeForce GT 610" }, + { 0x104b, "GeForce GT 625 (OEM)", nvkm_device_pci_10de_104b }, + { 0x104c, "GeForce GT 705" }, + { 0x1050, "GeForce GT 520M" }, + { 0x1051, "GeForce GT 520MX" }, + { 0x1052, "GeForce GT 520M" }, + { 0x1054, "GeForce 410M" }, + { 0x1055, "GeForce 410M" }, + { 0x1056, "NVS 4200M" }, + { 0x1057, "NVS 4200M" }, + { 0x1058, "GeForce 610M", nvkm_device_pci_10de_1058 }, + { 0x1059, "GeForce 610M" }, + { 0x105a, "GeForce 610M" }, + { 0x105b, "GeForce 705M", nvkm_device_pci_10de_105b }, + { 0x107c, "NVS 315" }, + { 0x107d, "NVS 310" }, + { 0x1080, "GeForce GTX 580" }, + { 0x1081, "GeForce GTX 570" }, + { 0x1082, "GeForce GTX 560 Ti" }, + { 0x1084, "GeForce GTX 560" }, + { 0x1086, "GeForce GTX 570" }, + { 0x1087, "GeForce GTX 560 Ti" }, + { 0x1088, "GeForce GTX 590" }, + { 0x1089, "GeForce GTX 580" }, + { 0x108b, "GeForce GTX 580" }, + { 0x1091, "Tesla M2090", nvkm_device_pci_10de_1091 }, + { 0x1094, "Tesla M2075" }, + { 0x1096, "Tesla C2075", nvkm_device_pci_10de_1096 }, + { 0x109a, "Quadro 5010M" }, + { 0x109b, "Quadro 7000" }, + { 0x10c0, "GeForce 9300 GS" }, + { 0x10c3, "GeForce 8400GS" }, + { 0x10c5, "GeForce 405" }, + { 0x10d8, "NVS 300" }, + { 0x1140, NULL, nvkm_device_pci_10de_1140 }, + { 0x1180, "GeForce GTX 680" }, + { 0x1183, "GeForce GTX 660 Ti" }, + { 0x1184, "GeForce GTX 770" }, + { 0x1185, "GeForce GTX 660", nvkm_device_pci_10de_1185 }, + { 0x1187, "GeForce GTX 760" }, + { 0x1188, "GeForce GTX 690" }, + { 0x1189, "GeForce GTX 670", nvkm_device_pci_10de_1189 }, + { 0x118a, "GRID K520" }, + { 0x118e, "GeForce GTX 760 (192-bit)" }, + { 0x118f, "Tesla K10" }, + { 0x1193, "GeForce GTX 760 Ti OEM" }, + { 0x1194, "Tesla K8" }, + { 0x1195, "GeForce GTX 660" }, + { 0x1198, "GeForce GTX 880M" }, + { 0x1199, "GeForce GTX 870M", nvkm_device_pci_10de_1199 }, + { 0x119a, "GeForce GTX 860M" }, + { 0x119d, "GeForce GTX 775M" }, + { 0x119e, "GeForce GTX 780M" }, + { 0x119f, "GeForce GTX 780M" }, + { 0x11a0, "GeForce GTX 680M" }, + { 0x11a1, "GeForce GTX 670MX" }, + { 0x11a2, "GeForce GTX 675MX" }, + { 0x11a3, "GeForce GTX 680MX" }, + { 0x11a7, "GeForce GTX 675MX" }, + { 0x11b4, "Quadro K4200" }, + { 0x11b6, "Quadro K3100M" }, + { 0x11b7, "Quadro K4100M" }, + { 0x11b8, "Quadro K5100M" }, + { 0x11ba, "Quadro K5000" }, + { 0x11bc, "Quadro K5000M" }, + { 0x11bd, "Quadro K4000M" }, + { 0x11be, "Quadro K3000M" }, + { 0x11bf, "GRID K2" }, + { 0x11c0, "GeForce GTX 660" }, + { 0x11c2, "GeForce GTX 650 Ti BOOST" }, + { 0x11c3, "GeForce GTX 650 Ti" }, + { 0x11c4, "GeForce GTX 645" }, + { 0x11c5, "GeForce GT 740" }, + { 0x11c6, "GeForce GTX 650 Ti" }, + { 0x11c8, "GeForce GTX 650" }, + { 0x11cb, "GeForce GT 740" }, + { 0x11e0, "GeForce GTX 770M" }, + { 0x11e1, "GeForce GTX 765M" }, + { 0x11e2, "GeForce GTX 765M" }, + { 0x11e3, "GeForce GTX 760M", nvkm_device_pci_10de_11e3 }, + { 0x11fa, "Quadro K4000" }, + { 0x11fc, "Quadro K2100M" }, + { 0x1200, "GeForce GTX 560 Ti" }, + { 0x1201, "GeForce GTX 560" }, + { 0x1203, "GeForce GTX 460 SE v2" }, + { 0x1205, "GeForce GTX 460 v2" }, + { 0x1206, "GeForce GTX 555" }, + { 0x1207, "GeForce GT 645" }, + { 0x1208, "GeForce GTX 560 SE" }, + { 0x1210, "GeForce GTX 570M" }, + { 0x1211, "GeForce GTX 580M" }, + { 0x1212, "GeForce GTX 675M" }, + { 0x1213, "GeForce GTX 670M" }, + { 0x1241, "GeForce GT 545" }, + { 0x1243, "GeForce GT 545" }, + { 0x1244, "GeForce GTX 550 Ti" }, + { 0x1245, "GeForce GTS 450" }, + { 0x1246, "GeForce GT 550M" }, + { 0x1247, "GeForce GT 555M", nvkm_device_pci_10de_1247 }, + { 0x1248, "GeForce GT 555M" }, + { 0x1249, "GeForce GTS 450" }, + { 0x124b, "GeForce GT 640" }, + { 0x124d, "GeForce GT 555M", nvkm_device_pci_10de_124d }, + { 0x1251, "GeForce GTX 560M" }, + { 0x1280, "GeForce GT 635" }, + { 0x1281, "GeForce GT 710" }, + { 0x1282, "GeForce GT 640" }, + { 0x1284, "GeForce GT 630" }, + { 0x1286, "GeForce GT 720" }, + { 0x1287, "GeForce GT 730" }, + { 0x1288, "GeForce GT 720" }, + { 0x1289, "GeForce GT 710" }, + { 0x1290, "GeForce GT 730M", nvkm_device_pci_10de_1290 }, + { 0x1291, "GeForce GT 735M" }, + { 0x1292, "GeForce GT 740M", nvkm_device_pci_10de_1292 }, + { 0x1293, "GeForce GT 730M" }, + { 0x1295, "GeForce 710M", nvkm_device_pci_10de_1295 }, + { 0x1296, "GeForce 825M" }, + { 0x1298, "GeForce GT 720M" }, + { 0x1299, "GeForce 920M", nvkm_device_pci_10de_1299 }, + { 0x129a, "GeForce 910M" }, + { 0x12b9, "Quadro K610M" }, + { 0x12ba, "Quadro K510M" }, + { 0x1340, "GeForce 830M", nvkm_device_pci_10de_1340 }, + { 0x1341, "GeForce 840M", nvkm_device_pci_10de_1341 }, + { 0x1344, "GeForce 845M" }, + { 0x1346, "GeForce 930M", nvkm_device_pci_10de_1346 }, + { 0x1347, "GeForce 940M", nvkm_device_pci_10de_1347 }, + { 0x137a, NULL, nvkm_device_pci_10de_137a }, + { 0x137d, NULL, nvkm_device_pci_10de_137d }, + { 0x1380, "GeForce GTX 750 Ti" }, + { 0x1381, "GeForce GTX 750" }, + { 0x1382, "GeForce GTX 745" }, + { 0x1390, "GeForce 845M" }, + { 0x1391, "GeForce GTX 850M", nvkm_device_pci_10de_1391 }, + { 0x1392, "GeForce GTX 860M", nvkm_device_pci_10de_1392 }, + { 0x1393, "GeForce 840M" }, + { 0x1398, "GeForce 845M" }, + { 0x139a, "GeForce GTX 950M", nvkm_device_pci_10de_139a }, + { 0x139b, "GeForce GTX 960M", nvkm_device_pci_10de_139b }, + { 0x139c, "GeForce 940M" }, + { 0x13b3, "Quadro K2200M" }, + { 0x13ba, "Quadro K2200" }, + { 0x13bb, "Quadro K620" }, + { 0x13bc, "Quadro K1200" }, + { 0x13c0, "GeForce GTX 980" }, + { 0x13c2, "GeForce GTX 970" }, + { 0x13d7, "GeForce GTX 980M" }, + { 0x13d8, "GeForce GTX 970M" }, + { 0x13d9, "GeForce GTX 965M" }, + { 0x1401, "GeForce GTX 960" }, + { 0x1617, "GeForce GTX 980M" }, + { 0x1618, "GeForce GTX 970M" }, + { 0x1619, "GeForce GTX 965M" }, + { 0x17c2, "GeForce GTX TITAN X" }, + { 0x17c8, "GeForce GTX 980 Ti" }, + { 0x17f0, "Quadro M6000" }, + {} +}; + +static struct nvkm_device_pci * +nvkm_device_pci(struct nvkm_device *device) +{ + return container_of(device, struct nvkm_device_pci, device); +} + +static resource_size_t +nvkm_device_pci_resource_addr(struct nvkm_device *device, unsigned bar) +{ + struct nvkm_device_pci *pdev = nvkm_device_pci(device); + return pci_resource_start(pdev->pdev, bar); +} + +static resource_size_t +nvkm_device_pci_resource_size(struct nvkm_device *device, unsigned bar) +{ + struct nvkm_device_pci *pdev = nvkm_device_pci(device); + return pci_resource_len(pdev->pdev, bar); +} + +static int +nvkm_device_pci_irq(struct nvkm_device *device) +{ + return nvkm_device_pci(device)->pdev->irq; +} + +static void +nvkm_device_pci_fini(struct nvkm_device *device, bool suspend) +{ + struct nvkm_device_pci *pdev = nvkm_device_pci(device); + if (suspend) { + pci_disable_device(pdev->pdev); + pdev->suspend = true; + } +} + +static int +nvkm_device_pci_preinit(struct nvkm_device *device) +{ + struct nvkm_device_pci *pdev = nvkm_device_pci(device); + if (pdev->suspend) { + int ret = pci_enable_device(pdev->pdev); + if (ret) + return ret; + pci_set_master(pdev->pdev); + pdev->suspend = false; + } + return 0; +} + +static void * +nvkm_device_pci_dtor(struct nvkm_device *device) +{ + struct nvkm_device_pci *pdev = nvkm_device_pci(device); + pci_disable_device(pdev->pdev); + return pdev; +} + +static const struct nvkm_device_func +nvkm_device_pci_func = { + .pci = nvkm_device_pci, + .dtor = nvkm_device_pci_dtor, + .preinit = nvkm_device_pci_preinit, + .fini = nvkm_device_pci_fini, + .irq = nvkm_device_pci_irq, + .resource_addr = nvkm_device_pci_resource_addr, + .resource_size = nvkm_device_pci_resource_size, + .cpu_coherent = !IS_ENABLED(CONFIG_ARM), +}; + +int +nvkm_device_pci_new(struct pci_dev *pci_dev, const char *cfg, const char *dbg, + bool detect, bool mmio, u64 subdev_mask, + struct nvkm_device **pdevice) +{ + const struct nvkm_device_quirk *quirk = NULL; + const struct nvkm_device_pci_device *pcid; + const struct nvkm_device_pci_vendor *pciv; + const char *name = NULL; + struct nvkm_device_pci *pdev; + int ret, bits; + + ret = pci_enable_device(pci_dev); + if (ret) + return ret; + + switch (pci_dev->vendor) { + case 0x10de: pcid = nvkm_device_pci_10de; break; + default: + pcid = NULL; + break; + } + + while (pcid && pcid->device) { + if (pciv = pcid->vendor, pcid->device == pci_dev->device) { + while (pciv && pciv->vendor) { + if (pciv->vendor == pci_dev->subsystem_vendor && + pciv->device == pci_dev->subsystem_device) { + quirk = &pciv->quirk; + name = pciv->name; + break; + } + pciv++; + } + if (!name) + name = pcid->name; + break; + } + pcid++; + } + + if (!(pdev = kzalloc(sizeof(*pdev), GFP_KERNEL))) { + pci_disable_device(pci_dev); + return -ENOMEM; + } + *pdevice = &pdev->device; + pdev->pdev = pci_dev; + + ret = nvkm_device_ctor(&nvkm_device_pci_func, quirk, &pci_dev->dev, + pci_is_pcie(pci_dev) ? NVKM_DEVICE_PCIE : + pci_find_capability(pci_dev, PCI_CAP_ID_AGP) ? + NVKM_DEVICE_AGP : NVKM_DEVICE_PCI, + (u64)pci_domain_nr(pci_dev->bus) << 32 | + pci_dev->bus->number << 16 | + PCI_SLOT(pci_dev->devfn) << 8 | + PCI_FUNC(pci_dev->devfn), name, + cfg, dbg, detect, mmio, subdev_mask, + &pdev->device); + + if (ret) + return ret; + + /* Set DMA mask based on capabilities reported by the MMU subdev. */ + if (pdev->device.mmu && !pdev->device.pci->agp.bridge) + bits = pdev->device.mmu->dma_bits; + else + bits = 32; + + ret = dma_set_mask_and_coherent(&pci_dev->dev, DMA_BIT_MASK(bits)); + if (ret && bits != 32) { + dma_set_mask_and_coherent(&pci_dev->dev, DMA_BIT_MASK(32)); + pdev->device.mmu->dma_bits = 32; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/device/priv.h new file mode 100644 index 0000000000..24faaac158 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/priv.h @@ -0,0 +1,61 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DEVICE_PRIV_H__ +#define __NVKM_DEVICE_PRIV_H__ +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +int nvkm_device_ctor(const struct nvkm_device_func *, + const struct nvkm_device_quirk *, + struct device *, enum nvkm_device_type, u64 handle, + const char *name, const char *cfg, const char *dbg, + bool detect, bool mmio, u64 subdev_mask, + struct nvkm_device *); +int nvkm_device_init(struct nvkm_device *); +int nvkm_device_fini(struct nvkm_device *, bool suspend); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c new file mode 100644 index 0000000000..87caa4a729 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c @@ -0,0 +1,341 @@ +/* + * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include +#ifdef CONFIG_NOUVEAU_PLATFORM_DRIVER +#include "priv.h" + +#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU) +#include +#endif + +static int +nvkm_device_tegra_power_up(struct nvkm_device_tegra *tdev) +{ + int ret; + + if (tdev->vdd) { + ret = regulator_enable(tdev->vdd); + if (ret) + goto err_power; + } + + ret = clk_prepare_enable(tdev->clk); + if (ret) + goto err_clk; + ret = clk_prepare_enable(tdev->clk_ref); + if (ret) + goto err_clk_ref; + ret = clk_prepare_enable(tdev->clk_pwr); + if (ret) + goto err_clk_pwr; + clk_set_rate(tdev->clk_pwr, 204000000); + udelay(10); + + if (!tdev->pdev->dev.pm_domain) { + reset_control_assert(tdev->rst); + udelay(10); + + ret = tegra_powergate_remove_clamping(TEGRA_POWERGATE_3D); + if (ret) + goto err_clamp; + udelay(10); + + reset_control_deassert(tdev->rst); + udelay(10); + } + + return 0; + +err_clamp: + clk_disable_unprepare(tdev->clk_pwr); +err_clk_pwr: + clk_disable_unprepare(tdev->clk_ref); +err_clk_ref: + clk_disable_unprepare(tdev->clk); +err_clk: + if (tdev->vdd) + regulator_disable(tdev->vdd); +err_power: + return ret; +} + +static int +nvkm_device_tegra_power_down(struct nvkm_device_tegra *tdev) +{ + int ret; + + clk_disable_unprepare(tdev->clk_pwr); + clk_disable_unprepare(tdev->clk_ref); + clk_disable_unprepare(tdev->clk); + udelay(10); + + if (tdev->vdd) { + ret = regulator_disable(tdev->vdd); + if (ret) + return ret; + } + + return 0; +} + +static void +nvkm_device_tegra_probe_iommu(struct nvkm_device_tegra *tdev) +{ +#if IS_ENABLED(CONFIG_IOMMU_API) + struct device *dev = &tdev->pdev->dev; + unsigned long pgsize_bitmap; + int ret; + +#if IS_ENABLED(CONFIG_ARM_DMA_USE_IOMMU) + if (dev->archdata.mapping) { + struct dma_iommu_mapping *mapping = to_dma_iommu_mapping(dev); + + arm_iommu_detach_device(dev); + arm_iommu_release_mapping(mapping); + } +#endif + + if (!tdev->func->iommu_bit) + return; + + mutex_init(&tdev->iommu.mutex); + + if (device_iommu_mapped(dev)) { + tdev->iommu.domain = iommu_domain_alloc(&platform_bus_type); + if (!tdev->iommu.domain) + goto error; + + /* + * A IOMMU is only usable if it supports page sizes smaller + * or equal to the system's PAGE_SIZE, with a preference if + * both are equal. + */ + pgsize_bitmap = tdev->iommu.domain->pgsize_bitmap; + if (pgsize_bitmap & PAGE_SIZE) { + tdev->iommu.pgshift = PAGE_SHIFT; + } else { + tdev->iommu.pgshift = fls(pgsize_bitmap & ~PAGE_MASK); + if (tdev->iommu.pgshift == 0) { + dev_warn(dev, "unsupported IOMMU page size\n"); + goto free_domain; + } + tdev->iommu.pgshift -= 1; + } + + ret = iommu_attach_device(tdev->iommu.domain, dev); + if (ret) + goto free_domain; + + ret = nvkm_mm_init(&tdev->iommu.mm, 0, 0, + (1ULL << tdev->func->iommu_bit) >> + tdev->iommu.pgshift, 1); + if (ret) + goto detach_device; + } + + return; + +detach_device: + iommu_detach_device(tdev->iommu.domain, dev); + +free_domain: + iommu_domain_free(tdev->iommu.domain); + +error: + tdev->iommu.domain = NULL; + tdev->iommu.pgshift = 0; + dev_err(dev, "cannot initialize IOMMU MM\n"); +#endif +} + +static void +nvkm_device_tegra_remove_iommu(struct nvkm_device_tegra *tdev) +{ +#if IS_ENABLED(CONFIG_IOMMU_API) + if (tdev->iommu.domain) { + nvkm_mm_fini(&tdev->iommu.mm); + iommu_detach_device(tdev->iommu.domain, tdev->device.dev); + iommu_domain_free(tdev->iommu.domain); + } +#endif +} + +static struct nvkm_device_tegra * +nvkm_device_tegra(struct nvkm_device *device) +{ + return container_of(device, struct nvkm_device_tegra, device); +} + +static struct resource * +nvkm_device_tegra_resource(struct nvkm_device *device, unsigned bar) +{ + struct nvkm_device_tegra *tdev = nvkm_device_tegra(device); + return platform_get_resource(tdev->pdev, IORESOURCE_MEM, bar); +} + +static resource_size_t +nvkm_device_tegra_resource_addr(struct nvkm_device *device, unsigned bar) +{ + struct resource *res = nvkm_device_tegra_resource(device, bar); + return res ? res->start : 0; +} + +static resource_size_t +nvkm_device_tegra_resource_size(struct nvkm_device *device, unsigned bar) +{ + struct resource *res = nvkm_device_tegra_resource(device, bar); + return res ? resource_size(res) : 0; +} + +static int +nvkm_device_tegra_irq(struct nvkm_device *device) +{ + struct nvkm_device_tegra *tdev = nvkm_device_tegra(device); + + return platform_get_irq_byname(tdev->pdev, "stall"); +} + +static void * +nvkm_device_tegra_dtor(struct nvkm_device *device) +{ + struct nvkm_device_tegra *tdev = nvkm_device_tegra(device); + nvkm_device_tegra_power_down(tdev); + nvkm_device_tegra_remove_iommu(tdev); + return tdev; +} + +static const struct nvkm_device_func +nvkm_device_tegra_func = { + .tegra = nvkm_device_tegra, + .dtor = nvkm_device_tegra_dtor, + .irq = nvkm_device_tegra_irq, + .resource_addr = nvkm_device_tegra_resource_addr, + .resource_size = nvkm_device_tegra_resource_size, + .cpu_coherent = false, +}; + +int +nvkm_device_tegra_new(const struct nvkm_device_tegra_func *func, + struct platform_device *pdev, + const char *cfg, const char *dbg, + bool detect, bool mmio, u64 subdev_mask, + struct nvkm_device **pdevice) +{ + struct nvkm_device_tegra *tdev; + unsigned long rate; + int ret; + + if (!(tdev = kzalloc(sizeof(*tdev), GFP_KERNEL))) + return -ENOMEM; + + tdev->func = func; + tdev->pdev = pdev; + + if (func->require_vdd) { + tdev->vdd = devm_regulator_get(&pdev->dev, "vdd"); + if (IS_ERR(tdev->vdd)) { + ret = PTR_ERR(tdev->vdd); + goto free; + } + } + + tdev->rst = devm_reset_control_get(&pdev->dev, "gpu"); + if (IS_ERR(tdev->rst)) { + ret = PTR_ERR(tdev->rst); + goto free; + } + + tdev->clk = devm_clk_get(&pdev->dev, "gpu"); + if (IS_ERR(tdev->clk)) { + ret = PTR_ERR(tdev->clk); + goto free; + } + + rate = clk_get_rate(tdev->clk); + if (rate == 0) { + ret = clk_set_rate(tdev->clk, ULONG_MAX); + if (ret < 0) + goto free; + + rate = clk_get_rate(tdev->clk); + + dev_dbg(&pdev->dev, "GPU clock set to %lu\n", rate); + } + + if (func->require_ref_clk) + tdev->clk_ref = devm_clk_get(&pdev->dev, "ref"); + if (IS_ERR(tdev->clk_ref)) { + ret = PTR_ERR(tdev->clk_ref); + goto free; + } + + tdev->clk_pwr = devm_clk_get(&pdev->dev, "pwr"); + if (IS_ERR(tdev->clk_pwr)) { + ret = PTR_ERR(tdev->clk_pwr); + goto free; + } + + /** + * The IOMMU bit defines the upper limit of the GPU-addressable space. + */ + ret = dma_set_mask(&pdev->dev, DMA_BIT_MASK(tdev->func->iommu_bit)); + if (ret) + goto free; + + nvkm_device_tegra_probe_iommu(tdev); + + ret = nvkm_device_tegra_power_up(tdev); + if (ret) + goto remove; + + tdev->gpu_speedo = tegra_sku_info.gpu_speedo_value; + tdev->gpu_speedo_id = tegra_sku_info.gpu_speedo_id; + ret = nvkm_device_ctor(&nvkm_device_tegra_func, NULL, &pdev->dev, + NVKM_DEVICE_TEGRA, pdev->id, NULL, + cfg, dbg, detect, mmio, subdev_mask, + &tdev->device); + if (ret) + goto powerdown; + + *pdevice = &tdev->device; + + return 0; + +powerdown: + nvkm_device_tegra_power_down(tdev); +remove: + nvkm_device_tegra_remove_iommu(tdev); +free: + kfree(tdev); + return ret; +} +#else +int +nvkm_device_tegra_new(const struct nvkm_device_tegra_func *func, + struct platform_device *pdev, + const char *cfg, const char *dbg, + bool detect, bool mmio, u64 subdev_mask, + struct nvkm_device **pdevice) +{ + return -ENOSYS; +} +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c new file mode 100644 index 0000000000..9b39ec3416 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/user.c @@ -0,0 +1,430 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define nvkm_udevice(p) container_of((p), struct nvkm_udevice, object) +#include "priv.h" +#include "ctrl.h" + +#include +#include +#include +#include + +#include +#include +#include + +struct nvkm_udevice { + struct nvkm_object object; + struct nvkm_device *device; +}; + +static int +nvkm_udevice_info_subdev(struct nvkm_device *device, u64 mthd, u64 *data) +{ + struct nvkm_subdev *subdev; + enum nvkm_subdev_type type; + + switch (mthd & NV_DEVICE_INFO_UNIT) { + case NV_DEVICE_HOST(0): type = NVKM_ENGINE_FIFO; break; + default: + return -EINVAL; + } + + subdev = nvkm_device_subdev(device, type, 0); + if (subdev) + return nvkm_subdev_info(subdev, mthd, data); + return -ENODEV; +} + +static void +nvkm_udevice_info_v1(struct nvkm_device *device, + struct nv_device_info_v1_data *args) +{ + if (args->mthd & NV_DEVICE_INFO_UNIT) { + if (nvkm_udevice_info_subdev(device, args->mthd, &args->data)) + args->mthd = NV_DEVICE_INFO_INVALID; + return; + } + args->mthd = NV_DEVICE_INFO_INVALID; +} + +static int +nvkm_udevice_info(struct nvkm_udevice *udev, void *data, u32 size) +{ + struct nvkm_object *object = &udev->object; + struct nvkm_device *device = udev->device; + struct nvkm_fb *fb = device->fb; + struct nvkm_instmem *imem = device->imem; + union { + struct nv_device_info_v0 v0; + struct nv_device_info_v1 v1; + } *args = data; + int ret = -ENOSYS, i; + + nvif_ioctl(object, "device info size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v1, 1, 1, true))) { + nvif_ioctl(object, "device info vers %d count %d\n", + args->v1.version, args->v1.count); + if (args->v1.count * sizeof(args->v1.data[0]) == size) { + for (i = 0; i < args->v1.count; i++) + nvkm_udevice_info_v1(device, &args->v1.data[i]); + return 0; + } + return -EINVAL; + } else + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "device info vers %d\n", args->v0.version); + } else + return ret; + + switch (device->chipset) { + case 0x01a: + case 0x01f: + case 0x04c: + case 0x04e: + case 0x063: + case 0x067: + case 0x068: + case 0x0aa: + case 0x0ac: + case 0x0af: + args->v0.platform = NV_DEVICE_INFO_V0_IGP; + break; + default: + switch (device->type) { + case NVKM_DEVICE_PCI: + args->v0.platform = NV_DEVICE_INFO_V0_PCI; + break; + case NVKM_DEVICE_AGP: + args->v0.platform = NV_DEVICE_INFO_V0_AGP; + break; + case NVKM_DEVICE_PCIE: + args->v0.platform = NV_DEVICE_INFO_V0_PCIE; + break; + case NVKM_DEVICE_TEGRA: + args->v0.platform = NV_DEVICE_INFO_V0_SOC; + break; + default: + WARN_ON(1); + break; + } + break; + } + + switch (device->card_type) { + case NV_04: args->v0.family = NV_DEVICE_INFO_V0_TNT; break; + case NV_10: + case NV_11: args->v0.family = NV_DEVICE_INFO_V0_CELSIUS; break; + case NV_20: args->v0.family = NV_DEVICE_INFO_V0_KELVIN; break; + case NV_30: args->v0.family = NV_DEVICE_INFO_V0_RANKINE; break; + case NV_40: args->v0.family = NV_DEVICE_INFO_V0_CURIE; break; + case NV_50: args->v0.family = NV_DEVICE_INFO_V0_TESLA; break; + case NV_C0: args->v0.family = NV_DEVICE_INFO_V0_FERMI; break; + case NV_E0: args->v0.family = NV_DEVICE_INFO_V0_KEPLER; break; + case GM100: args->v0.family = NV_DEVICE_INFO_V0_MAXWELL; break; + case GP100: args->v0.family = NV_DEVICE_INFO_V0_PASCAL; break; + case GV100: args->v0.family = NV_DEVICE_INFO_V0_VOLTA; break; + case TU100: args->v0.family = NV_DEVICE_INFO_V0_TURING; break; + case GA100: args->v0.family = NV_DEVICE_INFO_V0_AMPERE; break; + default: + args->v0.family = 0; + break; + } + + args->v0.chipset = device->chipset; + args->v0.revision = device->chiprev; + if (fb && fb->ram) + args->v0.ram_size = args->v0.ram_user = fb->ram->size; + else + args->v0.ram_size = args->v0.ram_user = 0; + if (imem && args->v0.ram_size > 0) + args->v0.ram_user = args->v0.ram_user - imem->reserved; + + snprintf(args->v0.chip, sizeof(args->v0.chip), "%s", device->chip->name); + snprintf(args->v0.name, sizeof(args->v0.name), "%s", device->name); + return 0; +} + +static int +nvkm_udevice_time(struct nvkm_udevice *udev, void *data, u32 size) +{ + struct nvkm_object *object = &udev->object; + struct nvkm_device *device = udev->device; + union { + struct nv_device_time_v0 v0; + } *args = data; + int ret = -ENOSYS; + + nvif_ioctl(object, "device time size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "device time vers %d\n", args->v0.version); + args->v0.time = nvkm_timer_read(device->timer); + } + + return ret; +} + +static int +nvkm_udevice_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + nvif_ioctl(object, "device mthd %08x\n", mthd); + switch (mthd) { + case NV_DEVICE_V0_INFO: + return nvkm_udevice_info(udev, data, size); + case NV_DEVICE_V0_TIME: + return nvkm_udevice_time(udev, data, size); + default: + break; + } + return -EINVAL; +} + +static int +nvkm_udevice_rd08(struct nvkm_object *object, u64 addr, u8 *data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + *data = nvkm_rd08(udev->device, addr); + return 0; +} + +static int +nvkm_udevice_rd16(struct nvkm_object *object, u64 addr, u16 *data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + *data = nvkm_rd16(udev->device, addr); + return 0; +} + +static int +nvkm_udevice_rd32(struct nvkm_object *object, u64 addr, u32 *data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + *data = nvkm_rd32(udev->device, addr); + return 0; +} + +static int +nvkm_udevice_wr08(struct nvkm_object *object, u64 addr, u8 data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + nvkm_wr08(udev->device, addr, data); + return 0; +} + +static int +nvkm_udevice_wr16(struct nvkm_object *object, u64 addr, u16 data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + nvkm_wr16(udev->device, addr, data); + return 0; +} + +static int +nvkm_udevice_wr32(struct nvkm_object *object, u64 addr, u32 data) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + nvkm_wr32(udev->device, addr, data); + return 0; +} + +static int +nvkm_udevice_map(struct nvkm_object *object, void *argv, u32 argc, + enum nvkm_object_map *type, u64 *addr, u64 *size) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + struct nvkm_device *device = udev->device; + *type = NVKM_OBJECT_MAP_IO; + *addr = device->func->resource_addr(device, 0); + *size = device->func->resource_size(device, 0); + return 0; +} + +static int +nvkm_udevice_fini(struct nvkm_object *object, bool suspend) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + struct nvkm_device *device = udev->device; + int ret = 0; + + mutex_lock(&device->mutex); + if (!--device->refcount) { + ret = nvkm_device_fini(device, suspend); + if (ret && suspend) { + device->refcount++; + goto done; + } + } + +done: + mutex_unlock(&device->mutex); + return ret; +} + +static int +nvkm_udevice_init(struct nvkm_object *object) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + struct nvkm_device *device = udev->device; + int ret = 0; + + mutex_lock(&device->mutex); + if (!device->refcount++) { + ret = nvkm_device_init(device); + if (ret) { + device->refcount--; + goto done; + } + } + +done: + mutex_unlock(&device->mutex); + return ret; +} + +static int +nvkm_udevice_child_new(const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + struct nvkm_udevice *udev = nvkm_udevice(oclass->parent); + const struct nvkm_device_oclass *sclass = oclass->priv; + return sclass->ctor(udev->device, oclass, data, size, pobject); +} + +static int +nvkm_udevice_child_get(struct nvkm_object *object, int index, + struct nvkm_oclass *oclass) +{ + struct nvkm_udevice *udev = nvkm_udevice(object); + struct nvkm_device *device = udev->device; + struct nvkm_engine *engine; + u64 mask = (1ULL << NVKM_ENGINE_DMAOBJ) | + (1ULL << NVKM_ENGINE_FIFO) | + (1ULL << NVKM_ENGINE_DISP) | + (1ULL << NVKM_ENGINE_PM); + const struct nvkm_device_oclass *sclass = NULL; + int i; + + for (; i = __ffs64(mask), mask && !sclass; mask &= ~(1ULL << i)) { + if (!(engine = nvkm_device_engine(device, i, 0)) || + !(engine->func->base.sclass)) + continue; + oclass->engine = engine; + + index -= engine->func->base.sclass(oclass, index, &sclass); + } + + if (!sclass) { + if (index-- == 0) + sclass = &nvkm_control_oclass; + else if (device->mmu && index-- == 0) + sclass = &device->mmu->user; + else if (device->fault && index-- == 0) + sclass = &device->fault->user; + else if (device->vfn && index-- == 0) + sclass = &device->vfn->user; + else + return -EINVAL; + + oclass->base = sclass->base; + oclass->engine = NULL; + } + + oclass->ctor = nvkm_udevice_child_new; + oclass->priv = sclass; + return 0; +} + +static const struct nvkm_object_func +nvkm_udevice_super = { + .init = nvkm_udevice_init, + .fini = nvkm_udevice_fini, + .mthd = nvkm_udevice_mthd, + .map = nvkm_udevice_map, + .rd08 = nvkm_udevice_rd08, + .rd16 = nvkm_udevice_rd16, + .rd32 = nvkm_udevice_rd32, + .wr08 = nvkm_udevice_wr08, + .wr16 = nvkm_udevice_wr16, + .wr32 = nvkm_udevice_wr32, + .sclass = nvkm_udevice_child_get, +}; + +static const struct nvkm_object_func +nvkm_udevice = { + .init = nvkm_udevice_init, + .fini = nvkm_udevice_fini, + .mthd = nvkm_udevice_mthd, + .sclass = nvkm_udevice_child_get, +}; + +static int +nvkm_udevice_new(const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + union { + struct nv_device_v0 v0; + } *args = data; + struct nvkm_client *client = oclass->client; + struct nvkm_object *parent = &client->object; + const struct nvkm_object_func *func; + struct nvkm_udevice *udev; + int ret = -ENOSYS; + + nvif_ioctl(parent, "create device size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, "create device v%d device %016llx\n", + args->v0.version, args->v0.device); + } else + return ret; + + /* give priviledged clients register access */ + if (args->v0.priv) + func = &nvkm_udevice_super; + else + func = &nvkm_udevice; + + if (!(udev = kzalloc(sizeof(*udev), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(func, oclass, &udev->object); + *pobject = &udev->object; + + /* find the device that matches what the client requested */ + if (args->v0.device != ~0) + udev->device = nvkm_device_find(args->v0.device); + else + udev->device = nvkm_device_find(client->device); + if (!udev->device) + return -ENODEV; + + return 0; +} + +const struct nvkm_sclass +nvkm_udevice_sclass = { + .oclass = NV_DEVICE, + .minver = 0, + .maxver = 0, + .ctor = nvkm_udevice_new, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild new file mode 100644 index 0000000000..e1aecd3fe9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/Kbuild @@ -0,0 +1,34 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/disp/base.o +nvkm-y += nvkm/engine/disp/chan.o +nvkm-y += nvkm/engine/disp/conn.o +nvkm-y += nvkm/engine/disp/dp.o +nvkm-y += nvkm/engine/disp/hdmi.o +nvkm-y += nvkm/engine/disp/head.o +nvkm-y += nvkm/engine/disp/ior.o +nvkm-y += nvkm/engine/disp/outp.o +nvkm-y += nvkm/engine/disp/vga.o + +nvkm-y += nvkm/engine/disp/nv04.o +nvkm-y += nvkm/engine/disp/nv50.o +nvkm-y += nvkm/engine/disp/g84.o +nvkm-y += nvkm/engine/disp/g94.o +nvkm-y += nvkm/engine/disp/gt200.o +nvkm-y += nvkm/engine/disp/mcp77.o +nvkm-y += nvkm/engine/disp/gt215.o +nvkm-y += nvkm/engine/disp/mcp89.o +nvkm-y += nvkm/engine/disp/gf119.o +nvkm-y += nvkm/engine/disp/gk104.o +nvkm-y += nvkm/engine/disp/gk110.o +nvkm-y += nvkm/engine/disp/gm107.o +nvkm-y += nvkm/engine/disp/gm200.o +nvkm-y += nvkm/engine/disp/gp100.o +nvkm-y += nvkm/engine/disp/gp102.o +nvkm-y += nvkm/engine/disp/gv100.o +nvkm-y += nvkm/engine/disp/tu102.o +nvkm-y += nvkm/engine/disp/ga102.o + +nvkm-y += nvkm/engine/disp/udisp.o +nvkm-y += nvkm/engine/disp/uconn.o +nvkm-y += nvkm/engine/disp/uoutp.o +nvkm-y += nvkm/engine/disp/uhead.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c new file mode 100644 index 0000000000..73104b59f9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/base.c @@ -0,0 +1,394 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "conn.h" +#include "dp.h" +#include "head.h" +#include "ior.h" +#include "outp.h" + +#include +#include +#include +#include + +#include +#include +#include +#include + +static void +nvkm_disp_vblank_fini(struct nvkm_event *event, int type, int id) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), vblank); + struct nvkm_head *head = nvkm_head_find(disp, id); + if (head) + head->func->vblank_put(head); +} + +static void +nvkm_disp_vblank_init(struct nvkm_event *event, int type, int id) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), vblank); + struct nvkm_head *head = nvkm_head_find(disp, id); + if (head) + head->func->vblank_get(head); +} + +static const struct nvkm_event_func +nvkm_disp_vblank_func = { + .init = nvkm_disp_vblank_init, + .fini = nvkm_disp_vblank_fini, +}; + +void +nvkm_disp_vblank(struct nvkm_disp *disp, int head) +{ + nvkm_event_ntfy(&disp->vblank, head, NVKM_DISP_HEAD_EVENT_VBLANK); +} + +static int +nvkm_disp_class_new(struct nvkm_device *device, + const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + return nvkm_udisp_new(oclass, data, size, pobject); +} + +static const struct nvkm_device_oclass +nvkm_disp_sclass = { + .ctor = nvkm_disp_class_new, +}; + +static int +nvkm_disp_class_get(struct nvkm_oclass *oclass, int index, + const struct nvkm_device_oclass **class) +{ + struct nvkm_disp *disp = nvkm_disp(oclass->engine); + if (index == 0) { + oclass->base = disp->func->root; + *class = &nvkm_disp_sclass; + return 0; + } + return 1; +} + +static void +nvkm_disp_intr(struct nvkm_engine *engine) +{ + struct nvkm_disp *disp = nvkm_disp(engine); + disp->func->intr(disp); +} + +static int +nvkm_disp_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_disp *disp = nvkm_disp(engine); + struct nvkm_conn *conn; + struct nvkm_outp *outp; + + if (disp->func->fini) + disp->func->fini(disp); + + list_for_each_entry(outp, &disp->outps, head) { + nvkm_outp_fini(outp); + } + + list_for_each_entry(conn, &disp->conns, head) { + nvkm_conn_fini(conn); + } + + return 0; +} + +static int +nvkm_disp_init(struct nvkm_engine *engine) +{ + struct nvkm_disp *disp = nvkm_disp(engine); + struct nvkm_conn *conn; + struct nvkm_outp *outp; + struct nvkm_ior *ior; + + list_for_each_entry(conn, &disp->conns, head) { + nvkm_conn_init(conn); + } + + list_for_each_entry(outp, &disp->outps, head) { + nvkm_outp_init(outp); + } + + if (disp->func->init) { + int ret = disp->func->init(disp); + if (ret) + return ret; + } + + /* Set 'normal' (ie. when it's attached to a head) state for + * each output resource to 'fully enabled'. + */ + list_for_each_entry(ior, &disp->iors, head) { + ior->func->power(ior, true, true, true, true, true); + } + + return 0; +} + +static int +nvkm_disp_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_disp *disp = nvkm_disp(engine); + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_bios *bios = subdev->device->bios; + struct nvkm_outp *outp, *outt, *pair; + struct nvkm_conn *conn; + struct nvkm_head *head; + struct nvkm_ior *ior; + struct nvbios_connE connE; + struct dcb_output dcbE; + u8 hpd = 0, ver, hdr; + u32 data; + int ret, i; + + /* Create output path objects for each VBIOS display path. */ + i = -1; + while ((data = dcb_outp_parse(bios, ++i, &ver, &hdr, &dcbE))) { + if (ver < 0x40) /* No support for chipsets prior to NV50. */ + break; + if (dcbE.type == DCB_OUTPUT_UNUSED) + continue; + if (dcbE.type == DCB_OUTPUT_EOL) + break; + outp = NULL; + + switch (dcbE.type) { + case DCB_OUTPUT_ANALOG: + case DCB_OUTPUT_TV: + case DCB_OUTPUT_TMDS: + case DCB_OUTPUT_LVDS: + ret = nvkm_outp_new(disp, i, &dcbE, &outp); + break; + case DCB_OUTPUT_DP: + ret = nvkm_dp_new(disp, i, &dcbE, &outp); + break; + case DCB_OUTPUT_WFD: + /* No support for WFD yet. */ + ret = -ENODEV; + continue; + default: + nvkm_warn(subdev, "dcb %d type %d unknown\n", + i, dcbE.type); + continue; + } + + if (ret) { + if (outp) { + if (ret != -ENODEV) + OUTP_ERR(outp, "ctor failed: %d", ret); + else + OUTP_DBG(outp, "not supported"); + nvkm_outp_del(&outp); + continue; + } + nvkm_error(subdev, "failed to create outp %d\n", i); + continue; + } + + list_add_tail(&outp->head, &disp->outps); + hpd = max(hpd, (u8)(dcbE.connector + 1)); + } + + /* Create connector objects based on available output paths. */ + list_for_each_entry_safe(outp, outt, &disp->outps, head) { + /* VBIOS data *should* give us the most useful information. */ + data = nvbios_connEp(bios, outp->info.connector, &ver, &hdr, + &connE); + + /* No bios connector data... */ + if (!data) { + /* Heuristic: anything with the same ccb index is + * considered to be on the same connector, any + * output path without an associated ccb entry will + * be put on its own connector. + */ + int ccb_index = outp->info.i2c_index; + if (ccb_index != 0xf) { + list_for_each_entry(pair, &disp->outps, head) { + if (pair->info.i2c_index == ccb_index) { + outp->conn = pair->conn; + break; + } + } + } + + /* Connector shared with another output path. */ + if (outp->conn) + continue; + + memset(&connE, 0x00, sizeof(connE)); + connE.type = DCB_CONNECTOR_NONE; + i = -1; + } else { + i = outp->info.connector; + } + + /* Check that we haven't already created this connector. */ + list_for_each_entry(conn, &disp->conns, head) { + if (conn->index == outp->info.connector) { + outp->conn = conn; + break; + } + } + + if (outp->conn) + continue; + + /* Apparently we need to create a new one! */ + ret = nvkm_conn_new(disp, i, &connE, &outp->conn); + if (ret) { + nvkm_error(subdev, "failed to create outp %d conn: %d\n", outp->index, ret); + nvkm_conn_del(&outp->conn); + list_del(&outp->head); + nvkm_outp_del(&outp); + continue; + } + + list_add_tail(&outp->conn->head, &disp->conns); + } + + if (disp->func->oneinit) { + ret = disp->func->oneinit(disp); + if (ret) + return ret; + } + + /* Enforce identity-mapped SOR assignment for panels, which have + * certain bits (ie. backlight controls) wired to a specific SOR. + */ + list_for_each_entry(outp, &disp->outps, head) { + if (outp->conn->info.type == DCB_CONNECTOR_LVDS || + outp->conn->info.type == DCB_CONNECTOR_eDP) { + ior = nvkm_ior_find(disp, SOR, ffs(outp->info.or) - 1); + if (!WARN_ON(!ior)) + ior->identity = true; + outp->identity = true; + } + } + + i = 0; + list_for_each_entry(head, &disp->heads, head) + i = max(i, head->id + 1); + + return nvkm_event_init(&nvkm_disp_vblank_func, subdev, 1, i, &disp->vblank); +} + +static void * +nvkm_disp_dtor(struct nvkm_engine *engine) +{ + struct nvkm_disp *disp = nvkm_disp(engine); + struct nvkm_conn *conn; + struct nvkm_outp *outp; + struct nvkm_ior *ior; + struct nvkm_head *head; + void *data = disp; + + nvkm_ramht_del(&disp->ramht); + nvkm_gpuobj_del(&disp->inst); + + nvkm_event_fini(&disp->uevent); + + if (disp->super.wq) { + destroy_workqueue(disp->super.wq); + mutex_destroy(&disp->super.mutex); + } + + nvkm_event_fini(&disp->vblank); + + while (!list_empty(&disp->conns)) { + conn = list_first_entry(&disp->conns, typeof(*conn), head); + list_del(&conn->head); + nvkm_conn_del(&conn); + } + + while (!list_empty(&disp->outps)) { + outp = list_first_entry(&disp->outps, typeof(*outp), head); + list_del(&outp->head); + nvkm_outp_del(&outp); + } + + while (!list_empty(&disp->iors)) { + ior = list_first_entry(&disp->iors, typeof(*ior), head); + nvkm_ior_del(&ior); + } + + while (!list_empty(&disp->heads)) { + head = list_first_entry(&disp->heads, typeof(*head), head); + nvkm_head_del(&head); + } + + return data; +} + +static const struct nvkm_engine_func +nvkm_disp = { + .dtor = nvkm_disp_dtor, + .oneinit = nvkm_disp_oneinit, + .init = nvkm_disp_init, + .fini = nvkm_disp_fini, + .intr = nvkm_disp_intr, + .base.sclass = nvkm_disp_class_get, +}; + +int +nvkm_disp_new_(const struct nvkm_disp_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_disp **pdisp) +{ + struct nvkm_disp *disp; + int ret; + + if (!(disp = *pdisp = kzalloc(sizeof(**pdisp), GFP_KERNEL))) + return -ENOMEM; + + disp->func = func; + INIT_LIST_HEAD(&disp->heads); + INIT_LIST_HEAD(&disp->iors); + INIT_LIST_HEAD(&disp->outps); + INIT_LIST_HEAD(&disp->conns); + spin_lock_init(&disp->client.lock); + + ret = nvkm_engine_ctor(&nvkm_disp, device, type, inst, true, &disp->engine); + if (ret) + return ret; + + if (func->super) { + disp->super.wq = create_singlethread_workqueue("nvkm-disp"); + if (!disp->super.wq) + return -ENOMEM; + + INIT_WORK(&disp->super.work, func->super); + mutex_init(&disp->super.mutex); + } + + return nvkm_event_init(func->uevent, &disp->engine.subdev, 1, ARRAY_SIZE(disp->chan), + &disp->uevent); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.c new file mode 100644 index 0000000000..d5e18daed7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.c @@ -0,0 +1,275 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "chan.h" + +#include +#include + +#include + +static int +nvkm_disp_chan_rd32(struct nvkm_object *object, u64 addr, u32 *data) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_device *device = chan->disp->engine.subdev.device; + u64 size, base = chan->func->user(chan, &size); + + *data = nvkm_rd32(device, base + addr); + return 0; +} + +static int +nvkm_disp_chan_wr32(struct nvkm_object *object, u64 addr, u32 data) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_device *device = chan->disp->engine.subdev.device; + u64 size, base = chan->func->user(chan, &size); + + nvkm_wr32(device, base + addr, data); + return 0; +} + +static int +nvkm_disp_chan_ntfy(struct nvkm_object *object, u32 type, struct nvkm_event **pevent) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_disp *disp = chan->disp; + + switch (type) { + case 0: + *pevent = &disp->uevent; + return 0; + default: + break; + } + + return -EINVAL; +} + +static int +nvkm_disp_chan_map(struct nvkm_object *object, void *argv, u32 argc, + enum nvkm_object_map *type, u64 *addr, u64 *size) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u64 base = device->func->resource_addr(device, 0); + + *type = NVKM_OBJECT_MAP_IO; + *addr = base + chan->func->user(chan, size); + return 0; +} + +struct nvkm_disp_chan_object { + struct nvkm_oproxy oproxy; + struct nvkm_disp *disp; + int hash; +}; + +static void +nvkm_disp_chan_child_del_(struct nvkm_oproxy *base) +{ + struct nvkm_disp_chan_object *object = container_of(base, typeof(*object), oproxy); + + nvkm_ramht_remove(object->disp->ramht, object->hash); +} + +static const struct nvkm_oproxy_func +nvkm_disp_chan_child_func_ = { + .dtor[0] = nvkm_disp_chan_child_del_, +}; + +static int +nvkm_disp_chan_child_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(oclass->parent); + struct nvkm_disp *disp = chan->disp; + struct nvkm_device *device = disp->engine.subdev.device; + const struct nvkm_device_oclass *sclass = oclass->priv; + struct nvkm_disp_chan_object *object; + int ret; + + if (!(object = kzalloc(sizeof(*object), GFP_KERNEL))) + return -ENOMEM; + nvkm_oproxy_ctor(&nvkm_disp_chan_child_func_, oclass, &object->oproxy); + object->disp = disp; + *pobject = &object->oproxy.base; + + ret = sclass->ctor(device, oclass, argv, argc, &object->oproxy.object); + if (ret) + return ret; + + object->hash = chan->func->bind(chan, object->oproxy.object, oclass->handle); + if (object->hash < 0) + return object->hash; + + return 0; +} + +static int +nvkm_disp_chan_child_get(struct nvkm_object *object, int index, struct nvkm_oclass *sclass) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_device *device = chan->disp->engine.subdev.device; + const struct nvkm_device_oclass *oclass = NULL; + + if (chan->func->bind) + sclass->engine = nvkm_device_engine(device, NVKM_ENGINE_DMAOBJ, 0); + else + sclass->engine = NULL; + + if (sclass->engine && sclass->engine->func->base.sclass) { + sclass->engine->func->base.sclass(sclass, index, &oclass); + if (oclass) { + sclass->ctor = nvkm_disp_chan_child_new; + sclass->priv = oclass; + return 0; + } + } + + return -EINVAL; +} + +static int +nvkm_disp_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + + chan->func->fini(chan); + chan->func->intr(chan, false); + return 0; +} + +static int +nvkm_disp_chan_init(struct nvkm_object *object) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + + chan->func->intr(chan, true); + return chan->func->init(chan); +} + +static void * +nvkm_disp_chan_dtor(struct nvkm_object *object) +{ + struct nvkm_disp_chan *chan = nvkm_disp_chan(object); + struct nvkm_disp *disp = chan->disp; + + spin_lock(&disp->client.lock); + if (disp->chan[chan->chid.user] == chan) + disp->chan[chan->chid.user] = NULL; + spin_unlock(&disp->client.lock); + + nvkm_memory_unref(&chan->memory); + return chan; +} + +static const struct nvkm_object_func +nvkm_disp_chan = { + .dtor = nvkm_disp_chan_dtor, + .init = nvkm_disp_chan_init, + .fini = nvkm_disp_chan_fini, + .rd32 = nvkm_disp_chan_rd32, + .wr32 = nvkm_disp_chan_wr32, + .ntfy = nvkm_disp_chan_ntfy, + .map = nvkm_disp_chan_map, + .sclass = nvkm_disp_chan_child_get, +}; + +static int +nvkm_disp_chan_new_(struct nvkm_disp *disp, int nr, const struct nvkm_oclass *oclass, + void *argv, u32 argc, struct nvkm_object **pobject) +{ + const struct nvkm_disp_chan_user *user = NULL; + struct nvkm_disp_chan *chan; + union nvif_disp_chan_args *args = argv; + int ret, i; + + for (i = 0; disp->func->user[i].ctor; i++) { + if (disp->func->user[i].base.oclass == oclass->base.oclass) { + user = disp->func->user[i].chan; + break; + } + } + + if (WARN_ON(!user)) + return -EINVAL; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + if (args->v0.id >= nr || !args->v0.pushbuf != !user->func->push) + return -EINVAL; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + *pobject = &chan->object; + + nvkm_object_ctor(&nvkm_disp_chan, oclass, &chan->object); + chan->func = user->func; + chan->mthd = user->mthd; + chan->disp = disp; + chan->chid.ctrl = user->ctrl + args->v0.id; + chan->chid.user = user->user + args->v0.id; + chan->head = args->v0.id; + + if (chan->func->push) { + ret = chan->func->push(chan, args->v0.pushbuf); + if (ret) + return ret; + } + + spin_lock(&disp->client.lock); + if (disp->chan[chan->chid.user]) { + spin_unlock(&disp->client.lock); + return -EBUSY; + } + disp->chan[chan->chid.user] = chan; + spin_unlock(&disp->client.lock); + return 0; +} + +int +nvkm_disp_wndw_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + + return nvkm_disp_chan_new_(disp, disp->wndw.nr, oclass, argv, argc, pobject); +} + +int +nvkm_disp_chan_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + + return nvkm_disp_chan_new_(disp, disp->head.nr, oclass, argv, argc, pobject); +} + +int +nvkm_disp_core_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + + return nvkm_disp_chan_new_(disp, 1, oclass, argv, argc, pobject); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.h new file mode 100644 index 0000000000..398336ffb6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/chan.h @@ -0,0 +1,135 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_CHAN_H__ +#define __NVKM_DISP_CHAN_H__ +#define nvkm_disp_chan(p) container_of((p), struct nvkm_disp_chan, object) +#include +#include "priv.h" + +struct nvkm_disp_chan { + const struct nvkm_disp_chan_func *func; + const struct nvkm_disp_chan_mthd *mthd; + struct nvkm_disp *disp; + + struct { + int ctrl; + int user; + } chid; + int head; + + struct nvkm_object object; + + struct nvkm_memory *memory; + u64 push; + + u32 suspend_put; +}; + +int nvkm_disp_core_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +int nvkm_disp_chan_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +int nvkm_disp_wndw_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); + +struct nvkm_disp_chan_func { + int (*push)(struct nvkm_disp_chan *, u64 object); + int (*init)(struct nvkm_disp_chan *); + void (*fini)(struct nvkm_disp_chan *); + void (*intr)(struct nvkm_disp_chan *, bool en); + u64 (*user)(struct nvkm_disp_chan *, u64 *size); + int (*bind)(struct nvkm_disp_chan *, struct nvkm_object *, u32 handle); +}; + +void nv50_disp_chan_intr(struct nvkm_disp_chan *, bool); +u64 nv50_disp_chan_user(struct nvkm_disp_chan *, u64 *); +extern const struct nvkm_disp_chan_func nv50_disp_pioc_func; +extern const struct nvkm_disp_chan_func nv50_disp_dmac_func; +int nv50_disp_dmac_push(struct nvkm_disp_chan *, u64); +int nv50_disp_dmac_bind(struct nvkm_disp_chan *, struct nvkm_object *, u32); +extern const struct nvkm_disp_chan_func nv50_disp_core_func; + +void gf119_disp_chan_intr(struct nvkm_disp_chan *, bool); +extern const struct nvkm_disp_chan_func gf119_disp_pioc_func; +extern const struct nvkm_disp_chan_func gf119_disp_dmac_func; +void gf119_disp_dmac_fini(struct nvkm_disp_chan *); +int gf119_disp_dmac_bind(struct nvkm_disp_chan *, struct nvkm_object *, u32); +extern const struct nvkm_disp_chan_func gf119_disp_core_func; +void gf119_disp_core_fini(struct nvkm_disp_chan *); + +extern const struct nvkm_disp_chan_func gp102_disp_dmac_func; + +u64 gv100_disp_chan_user(struct nvkm_disp_chan *, u64 *); +int gv100_disp_dmac_init(struct nvkm_disp_chan *); +void gv100_disp_dmac_fini(struct nvkm_disp_chan *); +int gv100_disp_dmac_bind(struct nvkm_disp_chan *, struct nvkm_object *, u32); + +struct nvkm_disp_chan_user { + const struct nvkm_disp_chan_func *func; + int ctrl; + int user; + const struct nvkm_disp_chan_mthd *mthd; +}; + +extern const struct nvkm_disp_chan_user nv50_disp_oimm; +extern const struct nvkm_disp_chan_user nv50_disp_curs; + +extern const struct nvkm_disp_chan_user g84_disp_core; +extern const struct nvkm_disp_chan_user g84_disp_base; +extern const struct nvkm_disp_chan_user g84_disp_ovly; + +extern const struct nvkm_disp_chan_user g94_disp_core; + +extern const struct nvkm_disp_chan_user gt200_disp_ovly; + +extern const struct nvkm_disp_chan_user gf119_disp_base; +extern const struct nvkm_disp_chan_user gf119_disp_oimm; +extern const struct nvkm_disp_chan_user gf119_disp_curs; + +extern const struct nvkm_disp_chan_user gk104_disp_core; +extern const struct nvkm_disp_chan_user gk104_disp_ovly; + +extern const struct nvkm_disp_chan_user gv100_disp_core; +extern const struct nvkm_disp_chan_user gv100_disp_curs; +extern const struct nvkm_disp_chan_user gv100_disp_wndw; +extern const struct nvkm_disp_chan_user gv100_disp_wimm; + +struct nvkm_disp_mthd_list { + u32 mthd; + u32 addr; + struct { + u32 mthd; + u32 addr; + const char *name; + } data[]; +}; + +struct nvkm_disp_chan_mthd { + const char *name; + u32 addr; + s32 prev; + struct { + const char *name; + int nr; + const struct nvkm_disp_mthd_list *mthd; + } data[]; +}; + +void nv50_disp_chan_mthd(struct nvkm_disp_chan *, int debug); + +extern const struct nvkm_disp_mthd_list nv50_disp_core_mthd_base; +extern const struct nvkm_disp_mthd_list nv50_disp_core_mthd_sor; +extern const struct nvkm_disp_mthd_list nv50_disp_core_mthd_pior; +extern const struct nvkm_disp_mthd_list nv50_disp_base_mthd_image; + +extern const struct nvkm_disp_chan_mthd g84_disp_core_mthd; +extern const struct nvkm_disp_mthd_list g84_disp_core_mthd_dac; +extern const struct nvkm_disp_mthd_list g84_disp_core_mthd_head; + +extern const struct nvkm_disp_chan_mthd g94_disp_core_mthd; + +extern const struct nvkm_disp_mthd_list gf119_disp_core_mthd_base; +extern const struct nvkm_disp_mthd_list gf119_disp_core_mthd_dac; +extern const struct nvkm_disp_mthd_list gf119_disp_core_mthd_sor; +extern const struct nvkm_disp_mthd_list gf119_disp_core_mthd_pior; +extern const struct nvkm_disp_chan_mthd gf119_disp_base_mthd; + +extern const struct nvkm_disp_chan_mthd gk104_disp_core_mthd; +extern const struct nvkm_disp_chan_mthd gk104_disp_ovly_mthd; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.c new file mode 100644 index 0000000000..fbdae11378 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.c @@ -0,0 +1,95 @@ +/* + * Copyright 2014 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "conn.h" +#include "outp.h" +#include "priv.h" + +#include + +#include + +void +nvkm_conn_fini(struct nvkm_conn *conn) +{ +} + +void +nvkm_conn_init(struct nvkm_conn *conn) +{ +} + +void +nvkm_conn_del(struct nvkm_conn **pconn) +{ + struct nvkm_conn *conn = *pconn; + if (conn) { + kfree(*pconn); + *pconn = NULL; + } +} + +static void +nvkm_conn_ctor(struct nvkm_disp *disp, int index, struct nvbios_connE *info, + struct nvkm_conn *conn) +{ + static const u8 hpd[] = { 0x07, 0x08, 0x51, 0x52, 0x5e, 0x5f, 0x60 }; + struct nvkm_gpio *gpio = disp->engine.subdev.device->gpio; + struct dcb_gpio_func func; + int ret; + + conn->disp = disp; + conn->index = index; + conn->info = *info; + conn->info.hpd = DCB_GPIO_UNUSED; + + CONN_DBG(conn, "type %02x loc %d hpd %02x dp %x di %x sr %x lcdid %x", + info->type, info->location, info->hpd, info->dp, + info->di, info->sr, info->lcdid); + + if ((info->hpd = ffs(info->hpd))) { + if (--info->hpd >= ARRAY_SIZE(hpd)) { + CONN_ERR(conn, "hpd %02x unknown", info->hpd); + return; + } + info->hpd = hpd[info->hpd]; + + ret = nvkm_gpio_find(gpio, 0, info->hpd, DCB_GPIO_UNUSED, &func); + if (ret) { + CONN_ERR(conn, "func %02x lookup failed, %d", info->hpd, ret); + return; + } + + conn->info.hpd = func.line; + } +} + +int +nvkm_conn_new(struct nvkm_disp *disp, int index, struct nvbios_connE *info, + struct nvkm_conn **pconn) +{ + if (!(*pconn = kzalloc(sizeof(**pconn), GFP_KERNEL))) + return -ENOMEM; + nvkm_conn_ctor(disp, index, info, *pconn); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.h new file mode 100644 index 0000000000..a0600e72b0 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/conn.h @@ -0,0 +1,33 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_CONN_H__ +#define __NVKM_DISP_CONN_H__ +#include "priv.h" + +#include +#include + +struct nvkm_conn { + struct nvkm_disp *disp; + int index; + struct nvbios_connE info; + + struct list_head head; + + struct nvkm_object object; +}; + +int nvkm_conn_new(struct nvkm_disp *, int index, struct nvbios_connE *, + struct nvkm_conn **); +void nvkm_conn_del(struct nvkm_conn **); +void nvkm_conn_init(struct nvkm_conn *); +void nvkm_conn_fini(struct nvkm_conn *); + +#define CONN_MSG(c,l,f,a...) do { \ + struct nvkm_conn *_conn = (c); \ + nvkm_##l(&_conn->disp->engine.subdev, "conn %02x:%02x%02x: "f"\n", \ + _conn->index, _conn->info.location, _conn->info.type, ##a); \ +} while(0) +#define CONN_ERR(c,f,a...) CONN_MSG((c), error, f, ##a) +#define CONN_DBG(c,f,a...) CONN_MSG((c), debug, f, ##a) +#define CONN_TRACE(c,f,a...) CONN_MSG((c), trace, f, ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.c new file mode 100644 index 0000000000..b8ac66b4a2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.c @@ -0,0 +1,853 @@ +/* + * Copyright 2014 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "dp.h" +#include "conn.h" +#include "head.h" +#include "ior.h" + +#include + +#include +#include +#include +#include + +#include + +/* IED scripts are no longer used by UEFI/RM from Ampere, but have been updated for + * the x86 option ROM. However, the relevant VBIOS table versions weren't modified, + * so we're unable to detect this in a nice way. + */ +#define AMPERE_IED_HACK(disp) ((disp)->engine.subdev.device->card_type >= GA100) + +struct lt_state { + struct nvkm_outp *outp; + + int repeaters; + int repeater; + + u8 stat[6]; + u8 conf[4]; + bool pc2; + u8 pc2stat; + u8 pc2conf[2]; +}; + +static int +nvkm_dp_train_sense(struct lt_state *lt, bool pc, u32 delay) +{ + struct nvkm_outp *outp = lt->outp; + u32 addr; + int ret; + + usleep_range(delay, delay * 2); + + if (lt->repeater) + addr = DPCD_LTTPR_LANE0_1_STATUS(lt->repeater); + else + addr = DPCD_LS02; + + ret = nvkm_rdaux(outp->dp.aux, addr, <->stat[0], 3); + if (ret) + return ret; + + if (lt->repeater) + addr = DPCD_LTTPR_LANE0_1_ADJUST(lt->repeater); + else + addr = DPCD_LS06; + + ret = nvkm_rdaux(outp->dp.aux, addr, <->stat[4], 2); + if (ret) + return ret; + + if (pc) { + ret = nvkm_rdaux(outp->dp.aux, DPCD_LS0C, <->pc2stat, 1); + if (ret) + lt->pc2stat = 0x00; + + OUTP_TRACE(outp, "status %6ph pc2 %02x", lt->stat, lt->pc2stat); + } else { + OUTP_TRACE(outp, "status %6ph", lt->stat); + } + + return 0; +} + +static int +nvkm_dp_train_drive(struct lt_state *lt, bool pc) +{ + struct nvkm_outp *outp = lt->outp; + struct nvkm_ior *ior = outp->ior; + struct nvkm_bios *bios = ior->disp->engine.subdev.device->bios; + struct nvbios_dpout info; + struct nvbios_dpcfg ocfg; + u8 ver, hdr, cnt, len; + u32 addr; + u32 data; + int ret, i; + + for (i = 0; i < ior->dp.nr; i++) { + u8 lane = (lt->stat[4 + (i >> 1)] >> ((i & 1) * 4)) & 0xf; + u8 lpc2 = (lt->pc2stat >> (i * 2)) & 0x3; + u8 lpre = (lane & 0x0c) >> 2; + u8 lvsw = (lane & 0x03) >> 0; + u8 hivs = 3 - lpre; + u8 hipe = 3; + u8 hipc = 3; + + if (lpc2 >= hipc) + lpc2 = hipc | DPCD_LC0F_LANE0_MAX_POST_CURSOR2_REACHED; + if (lpre >= hipe) { + lpre = hipe | DPCD_LC03_MAX_SWING_REACHED; /* yes. */ + lvsw = hivs = 3 - (lpre & 3); + } else + if (lvsw >= hivs) { + lvsw = hivs | DPCD_LC03_MAX_SWING_REACHED; + } + + lt->conf[i] = (lpre << 3) | lvsw; + lt->pc2conf[i >> 1] |= lpc2 << ((i & 1) * 4); + + OUTP_TRACE(outp, "config lane %d %02x %02x", i, lt->conf[i], lpc2); + + if (lt->repeater != lt->repeaters) + continue; + + data = nvbios_dpout_match(bios, outp->info.hasht, outp->info.hashm, + &ver, &hdr, &cnt, &len, &info); + if (!data) + continue; + + data = nvbios_dpcfg_match(bios, data, lpc2 & 3, lvsw & 3, lpre & 3, + &ver, &hdr, &cnt, &len, &ocfg); + if (!data) + continue; + + ior->func->dp->drive(ior, i, ocfg.pc, ocfg.dc, ocfg.pe, ocfg.tx_pu); + } + + if (lt->repeater) + addr = DPCD_LTTPR_LANE0_SET(lt->repeater); + else + addr = DPCD_LC03(0); + + ret = nvkm_wraux(outp->dp.aux, addr, lt->conf, 4); + if (ret) + return ret; + + if (pc) { + ret = nvkm_wraux(outp->dp.aux, DPCD_LC0F, lt->pc2conf, 2); + if (ret) + return ret; + } + + return 0; +} + +static void +nvkm_dp_train_pattern(struct lt_state *lt, u8 pattern) +{ + struct nvkm_outp *outp = lt->outp; + u32 addr; + u8 sink_tp; + + OUTP_TRACE(outp, "training pattern %d", pattern); + outp->ior->func->dp->pattern(outp->ior, pattern); + + if (lt->repeater) + addr = DPCD_LTTPR_PATTERN_SET(lt->repeater); + else + addr = DPCD_LC02; + + nvkm_rdaux(outp->dp.aux, addr, &sink_tp, 1); + sink_tp &= ~DPCD_LC02_TRAINING_PATTERN_SET; + sink_tp |= (pattern != 4) ? pattern : 7; + + if (pattern != 0) + sink_tp |= DPCD_LC02_SCRAMBLING_DISABLE; + else + sink_tp &= ~DPCD_LC02_SCRAMBLING_DISABLE; + nvkm_wraux(outp->dp.aux, addr, &sink_tp, 1); +} + +static int +nvkm_dp_train_eq(struct lt_state *lt) +{ + struct nvkm_i2c_aux *aux = lt->outp->dp.aux; + bool eq_done = false, cr_done = true; + int tries = 0, usec = 0, i; + u8 data; + + if (lt->repeater) { + if (!nvkm_rdaux(aux, DPCD_LTTPR_AUX_RD_INTERVAL(lt->repeater), &data, sizeof(data))) + usec = (data & DPCD_RC0E_AUX_RD_INTERVAL) * 4000; + + nvkm_dp_train_pattern(lt, 4); + } else { + if (lt->outp->dp.dpcd[DPCD_RC00_DPCD_REV] >= 0x14 && + lt->outp->dp.dpcd[DPCD_RC03] & DPCD_RC03_TPS4_SUPPORTED) + nvkm_dp_train_pattern(lt, 4); + else + if (lt->outp->dp.dpcd[DPCD_RC00_DPCD_REV] >= 0x12 && + lt->outp->dp.dpcd[DPCD_RC02] & DPCD_RC02_TPS3_SUPPORTED) + nvkm_dp_train_pattern(lt, 3); + else + nvkm_dp_train_pattern(lt, 2); + + usec = (lt->outp->dp.dpcd[DPCD_RC0E] & DPCD_RC0E_AUX_RD_INTERVAL) * 4000; + } + + do { + if ((tries && + nvkm_dp_train_drive(lt, lt->pc2)) || + nvkm_dp_train_sense(lt, lt->pc2, usec ? usec : 400)) + break; + + eq_done = !!(lt->stat[2] & DPCD_LS04_INTERLANE_ALIGN_DONE); + for (i = 0; i < lt->outp->ior->dp.nr && eq_done; i++) { + u8 lane = (lt->stat[i >> 1] >> ((i & 1) * 4)) & 0xf; + if (!(lane & DPCD_LS02_LANE0_CR_DONE)) + cr_done = false; + if (!(lane & DPCD_LS02_LANE0_CHANNEL_EQ_DONE) || + !(lane & DPCD_LS02_LANE0_SYMBOL_LOCKED)) + eq_done = false; + } + } while (!eq_done && cr_done && ++tries <= 5); + + return eq_done ? 0 : -1; +} + +static int +nvkm_dp_train_cr(struct lt_state *lt) +{ + bool cr_done = false, abort = false; + int voltage = lt->conf[0] & DPCD_LC03_VOLTAGE_SWING_SET; + int tries = 0, usec = 0, i; + + nvkm_dp_train_pattern(lt, 1); + + if (lt->outp->dp.dpcd[DPCD_RC00_DPCD_REV] < 0x14 && !lt->repeater) + usec = (lt->outp->dp.dpcd[DPCD_RC0E] & DPCD_RC0E_AUX_RD_INTERVAL) * 4000; + + do { + if (nvkm_dp_train_drive(lt, false) || + nvkm_dp_train_sense(lt, false, usec ? usec : 100)) + break; + + cr_done = true; + for (i = 0; i < lt->outp->ior->dp.nr; i++) { + u8 lane = (lt->stat[i >> 1] >> ((i & 1) * 4)) & 0xf; + if (!(lane & DPCD_LS02_LANE0_CR_DONE)) { + cr_done = false; + if (lt->conf[i] & DPCD_LC03_MAX_SWING_REACHED) + abort = true; + break; + } + } + + if ((lt->conf[0] & DPCD_LC03_VOLTAGE_SWING_SET) != voltage) { + voltage = lt->conf[0] & DPCD_LC03_VOLTAGE_SWING_SET; + tries = 0; + } + } while (!cr_done && !abort && ++tries < 5); + + return cr_done ? 0 : -1; +} + +static int +nvkm_dp_train_link(struct nvkm_outp *outp, int rate) +{ + struct nvkm_ior *ior = outp->ior; + struct lt_state lt = { + .outp = outp, + .pc2 = outp->dp.dpcd[DPCD_RC02] & DPCD_RC02_TPS3_SUPPORTED, + }; + u8 sink[2], data; + int ret; + + OUTP_DBG(outp, "training %dx%02x", ior->dp.nr, ior->dp.bw); + + /* Select LTTPR non-transparent mode if we have a valid configuration, + * use transparent mode otherwise. + */ + if (outp->dp.lttpr[0] >= 0x14) { + data = DPCD_LTTPR_MODE_TRANSPARENT; + nvkm_wraux(outp->dp.aux, DPCD_LTTPR_MODE, &data, sizeof(data)); + + if (outp->dp.lttprs) { + data = DPCD_LTTPR_MODE_NON_TRANSPARENT; + nvkm_wraux(outp->dp.aux, DPCD_LTTPR_MODE, &data, sizeof(data)); + lt.repeaters = outp->dp.lttprs; + } + } + + /* Set desired link configuration on the sink. */ + sink[0] = (outp->dp.rate[rate].dpcd < 0) ? ior->dp.bw : 0; + sink[1] = ior->dp.nr; + if (ior->dp.ef) + sink[1] |= DPCD_LC01_ENHANCED_FRAME_EN; + + ret = nvkm_wraux(outp->dp.aux, DPCD_LC00_LINK_BW_SET, sink, 2); + if (ret) + return ret; + + if (outp->dp.rate[rate].dpcd >= 0) { + ret = nvkm_rdaux(outp->dp.aux, DPCD_LC15_LINK_RATE_SET, &sink[0], sizeof(sink[0])); + if (ret) + return ret; + + sink[0] &= ~DPCD_LC15_LINK_RATE_SET_MASK; + sink[0] |= outp->dp.rate[rate].dpcd; + + ret = nvkm_wraux(outp->dp.aux, DPCD_LC15_LINK_RATE_SET, &sink[0], sizeof(sink[0])); + if (ret) + return ret; + } + + /* Attempt to train the link in this configuration. */ + for (lt.repeater = lt.repeaters; lt.repeater >= 0; lt.repeater--) { + if (lt.repeater) + OUTP_DBG(outp, "training LTTPR%d", lt.repeater); + else + OUTP_DBG(outp, "training sink"); + + memset(lt.stat, 0x00, sizeof(lt.stat)); + ret = nvkm_dp_train_cr(<); + if (ret == 0) + ret = nvkm_dp_train_eq(<); + nvkm_dp_train_pattern(<, 0); + } + + return ret; +} + +static int +nvkm_dp_train_links(struct nvkm_outp *outp, int rate) +{ + struct nvkm_ior *ior = outp->ior; + struct nvkm_disp *disp = outp->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_bios *bios = subdev->device->bios; + u32 lnkcmp; + int ret; + + OUTP_DBG(outp, "programming link for %dx%02x", ior->dp.nr, ior->dp.bw); + + /* Intersect misc. capabilities of the OR and sink. */ + if (disp->engine.subdev.device->chipset < 0x110) + outp->dp.dpcd[DPCD_RC03] &= ~DPCD_RC03_TPS4_SUPPORTED; + if (disp->engine.subdev.device->chipset < 0xd0) + outp->dp.dpcd[DPCD_RC02] &= ~DPCD_RC02_TPS3_SUPPORTED; + + if (AMPERE_IED_HACK(disp) && (lnkcmp = outp->dp.info.script[0])) { + /* Execute BeforeLinkTraining script from DP Info table. */ + while (ior->dp.bw < nvbios_rd08(bios, lnkcmp)) + lnkcmp += 3; + lnkcmp = nvbios_rd16(bios, lnkcmp + 1); + + nvbios_init(&outp->disp->engine.subdev, lnkcmp, + init.outp = &outp->info; + init.or = ior->id; + init.link = ior->asy.link; + ); + } + + /* Set desired link configuration on the source. */ + if ((lnkcmp = outp->dp.info.lnkcmp)) { + if (outp->dp.version < 0x30) { + while ((ior->dp.bw * 2700) < nvbios_rd16(bios, lnkcmp)) + lnkcmp += 4; + lnkcmp = nvbios_rd16(bios, lnkcmp + 2); + } else { + while (ior->dp.bw < nvbios_rd08(bios, lnkcmp)) + lnkcmp += 3; + lnkcmp = nvbios_rd16(bios, lnkcmp + 1); + } + + nvbios_init(subdev, lnkcmp, + init.outp = &outp->info; + init.or = ior->id; + init.link = ior->asy.link; + ); + } + + ret = ior->func->dp->links(ior, outp->dp.aux); + if (ret) { + if (ret < 0) { + OUTP_ERR(outp, "train failed with %d", ret); + return ret; + } + return 0; + } + + ior->func->dp->power(ior, ior->dp.nr); + + /* Attempt to train the link in this configuration. */ + return nvkm_dp_train_link(outp, rate); +} + +static void +nvkm_dp_train_fini(struct nvkm_outp *outp) +{ + /* Execute AfterLinkTraining script from DP Info table. */ + nvbios_init(&outp->disp->engine.subdev, outp->dp.info.script[1], + init.outp = &outp->info; + init.or = outp->ior->id; + init.link = outp->ior->asy.link; + ); +} + +static void +nvkm_dp_train_init(struct nvkm_outp *outp) +{ + /* Execute EnableSpread/DisableSpread script from DP Info table. */ + if (outp->dp.dpcd[DPCD_RC03] & DPCD_RC03_MAX_DOWNSPREAD) { + nvbios_init(&outp->disp->engine.subdev, outp->dp.info.script[2], + init.outp = &outp->info; + init.or = outp->ior->id; + init.link = outp->ior->asy.link; + ); + } else { + nvbios_init(&outp->disp->engine.subdev, outp->dp.info.script[3], + init.outp = &outp->info; + init.or = outp->ior->id; + init.link = outp->ior->asy.link; + ); + } + + if (!AMPERE_IED_HACK(outp->disp)) { + /* Execute BeforeLinkTraining script from DP Info table. */ + nvbios_init(&outp->disp->engine.subdev, outp->dp.info.script[0], + init.outp = &outp->info; + init.or = outp->ior->id; + init.link = outp->ior->asy.link; + ); + } +} + +static int +nvkm_dp_train(struct nvkm_outp *outp, u32 dataKBps) +{ + struct nvkm_ior *ior = outp->ior; + int ret = -EINVAL, nr, rate; + u8 pwr; + + /* Retraining link? Skip source configuration, it can mess up the active modeset. */ + if (atomic_read(&outp->dp.lt.done)) { + for (rate = 0; rate < outp->dp.rates; rate++) { + if (outp->dp.rate[rate].rate == ior->dp.bw * 27000) + return nvkm_dp_train_link(outp, ret); + } + WARN_ON(1); + return -EINVAL; + } + + /* Ensure sink is not in a low-power state. */ + if (!nvkm_rdaux(outp->dp.aux, DPCD_SC00, &pwr, 1)) { + if ((pwr & DPCD_SC00_SET_POWER) != DPCD_SC00_SET_POWER_D0) { + pwr &= ~DPCD_SC00_SET_POWER; + pwr |= DPCD_SC00_SET_POWER_D0; + nvkm_wraux(outp->dp.aux, DPCD_SC00, &pwr, 1); + } + } + + ior->dp.mst = outp->dp.lt.mst; + ior->dp.ef = outp->dp.dpcd[DPCD_RC02] & DPCD_RC02_ENHANCED_FRAME_CAP; + ior->dp.nr = 0; + + /* Link training. */ + OUTP_DBG(outp, "training"); + nvkm_dp_train_init(outp); + + /* Validate and train at configuration requested (if any) on ACQUIRE. */ + if (outp->dp.lt.nr) { + for (nr = outp->dp.links; ret < 0 && nr; nr >>= 1) { + for (rate = 0; nr == outp->dp.lt.nr && rate < outp->dp.rates; rate++) { + if (outp->dp.rate[rate].rate / 27000 == outp->dp.lt.bw) { + ior->dp.bw = outp->dp.rate[rate].rate / 27000; + ior->dp.nr = nr; + ret = nvkm_dp_train_links(outp, rate); + } + } + } + } + + /* Otherwise, loop through all valid link configurations that support the data rate. */ + for (nr = outp->dp.links; ret < 0 && nr; nr >>= 1) { + for (rate = 0; ret < 0 && rate < outp->dp.rates; rate++) { + if (outp->dp.rate[rate].rate * nr >= dataKBps || WARN_ON(!ior->dp.nr)) { + /* Program selected link configuration. */ + ior->dp.bw = outp->dp.rate[rate].rate / 27000; + ior->dp.nr = nr; + ret = nvkm_dp_train_links(outp, rate); + } + } + } + + /* Finish up. */ + nvkm_dp_train_fini(outp); + if (ret < 0) + OUTP_ERR(outp, "training failed"); + else + OUTP_DBG(outp, "training done"); + atomic_set(&outp->dp.lt.done, 1); + return ret; +} + +void +nvkm_dp_disable(struct nvkm_outp *outp, struct nvkm_ior *ior) +{ + /* Execute DisableLT script from DP Info Table. */ + nvbios_init(&ior->disp->engine.subdev, outp->dp.info.script[4], + init.outp = &outp->info; + init.or = ior->id; + init.link = ior->arm.link; + ); +} + +static void +nvkm_dp_release(struct nvkm_outp *outp) +{ + /* Prevent link from being retrained if sink sends an IRQ. */ + atomic_set(&outp->dp.lt.done, 0); + outp->ior->dp.nr = 0; +} + +static int +nvkm_dp_acquire(struct nvkm_outp *outp) +{ + struct nvkm_ior *ior = outp->ior; + struct nvkm_head *head; + bool retrain = true; + u32 datakbps = 0; + u32 dataKBps; + u32 linkKBps; + u8 stat[3]; + int ret, i; + + mutex_lock(&outp->dp.mutex); + + /* Check that link configuration meets current requirements. */ + list_for_each_entry(head, &outp->disp->heads, head) { + if (ior->asy.head & (1 << head->id)) { + u32 khz = (head->asy.hz >> ior->asy.rgdiv) / 1000; + datakbps += khz * head->asy.or.depth; + } + } + + linkKBps = ior->dp.bw * 27000 * ior->dp.nr; + dataKBps = DIV_ROUND_UP(datakbps, 8); + OUTP_DBG(outp, "data %d KB/s link %d KB/s mst %d->%d", + dataKBps, linkKBps, ior->dp.mst, outp->dp.lt.mst); + if (linkKBps < dataKBps || ior->dp.mst != outp->dp.lt.mst) { + OUTP_DBG(outp, "link requirements changed"); + goto done; + } + + /* Check that link is still trained. */ + ret = nvkm_rdaux(outp->dp.aux, DPCD_LS02, stat, 3); + if (ret) { + OUTP_DBG(outp, "failed to read link status, assuming no sink"); + goto done; + } + + if (stat[2] & DPCD_LS04_INTERLANE_ALIGN_DONE) { + for (i = 0; i < ior->dp.nr; i++) { + u8 lane = (stat[i >> 1] >> ((i & 1) * 4)) & 0x0f; + if (!(lane & DPCD_LS02_LANE0_CR_DONE) || + !(lane & DPCD_LS02_LANE0_CHANNEL_EQ_DONE) || + !(lane & DPCD_LS02_LANE0_SYMBOL_LOCKED)) { + OUTP_DBG(outp, "lane %d not equalised", lane); + goto done; + } + } + retrain = false; + } else { + OUTP_DBG(outp, "no inter-lane alignment"); + } + +done: + if (retrain || !atomic_read(&outp->dp.lt.done)) + ret = nvkm_dp_train(outp, dataKBps); + mutex_unlock(&outp->dp.mutex); + return ret; +} + +static bool +nvkm_dp_enable_supported_link_rates(struct nvkm_outp *outp) +{ + u8 sink_rates[DPCD_RC10_SUPPORTED_LINK_RATES__SIZE]; + int i, j, k; + + if (outp->conn->info.type != DCB_CONNECTOR_eDP || + outp->dp.dpcd[DPCD_RC00_DPCD_REV] < 0x13 || + nvkm_rdaux(outp->dp.aux, DPCD_RC10_SUPPORTED_LINK_RATES(0), + sink_rates, sizeof(sink_rates))) + return false; + + for (i = 0; i < ARRAY_SIZE(sink_rates); i += 2) { + const u32 rate = ((sink_rates[i + 1] << 8) | sink_rates[i]) * 200 / 10; + + if (!rate || WARN_ON(outp->dp.rates == ARRAY_SIZE(outp->dp.rate))) + break; + + if (rate > outp->info.dpconf.link_bw * 27000) { + OUTP_DBG(outp, "rate %d !outp", rate); + continue; + } + + for (j = 0; j < outp->dp.rates; j++) { + if (rate > outp->dp.rate[j].rate) { + for (k = outp->dp.rates; k > j; k--) + outp->dp.rate[k] = outp->dp.rate[k - 1]; + break; + } + } + + outp->dp.rate[j].dpcd = i / 2; + outp->dp.rate[j].rate = rate; + outp->dp.rates++; + } + + for (i = 0; i < outp->dp.rates; i++) + OUTP_DBG(outp, "link_rate[%d] = %d", outp->dp.rate[i].dpcd, outp->dp.rate[i].rate); + + return outp->dp.rates != 0; +} + +/* XXX: This is a big fat hack, and this is just drm_dp_read_dpcd_caps() + * converted to work inside nvkm. This is a temporary holdover until we start + * passing the drm_dp_aux device through NVKM + */ +static int +nvkm_dp_read_dpcd_caps(struct nvkm_outp *outp) +{ + struct nvkm_i2c_aux *aux = outp->dp.aux; + u8 dpcd_ext[DP_RECEIVER_CAP_SIZE]; + int ret; + + ret = nvkm_rdaux(aux, DPCD_RC00_DPCD_REV, outp->dp.dpcd, DP_RECEIVER_CAP_SIZE); + if (ret < 0) + return ret; + + /* + * Prior to DP1.3 the bit represented by + * DP_EXTENDED_RECEIVER_CAP_FIELD_PRESENT was reserved. + * If it is set DP_DPCD_REV at 0000h could be at a value less than + * the true capability of the panel. The only way to check is to + * then compare 0000h and 2200h. + */ + if (!(outp->dp.dpcd[DP_TRAINING_AUX_RD_INTERVAL] & + DP_EXTENDED_RECEIVER_CAP_FIELD_PRESENT)) + return 0; + + ret = nvkm_rdaux(aux, DP_DP13_DPCD_REV, dpcd_ext, sizeof(dpcd_ext)); + if (ret < 0) + return ret; + + if (outp->dp.dpcd[DP_DPCD_REV] > dpcd_ext[DP_DPCD_REV]) { + OUTP_DBG(outp, "Extended DPCD rev less than base DPCD rev (%d > %d)\n", + outp->dp.dpcd[DP_DPCD_REV], dpcd_ext[DP_DPCD_REV]); + return 0; + } + + if (!memcmp(outp->dp.dpcd, dpcd_ext, sizeof(dpcd_ext))) + return 0; + + memcpy(outp->dp.dpcd, dpcd_ext, sizeof(dpcd_ext)); + + return 0; +} + +void +nvkm_dp_enable(struct nvkm_outp *outp, bool auxpwr) +{ + struct nvkm_gpio *gpio = outp->disp->engine.subdev.device->gpio; + struct nvkm_i2c_aux *aux = outp->dp.aux; + + if (auxpwr && !outp->dp.aux_pwr) { + /* eDP panels need powering on by us (if the VBIOS doesn't default it + * to on) before doing any AUX channel transactions. LVDS panel power + * is handled by the SOR itself, and not required for LVDS DDC. + */ + if (outp->conn->info.type == DCB_CONNECTOR_eDP) { + int power = nvkm_gpio_get(gpio, 0, DCB_GPIO_PANEL_POWER, 0xff); + if (power == 0) { + nvkm_gpio_set(gpio, 0, DCB_GPIO_PANEL_POWER, 0xff, 1); + outp->dp.aux_pwr_pu = true; + } + + /* We delay here unconditionally, even if already powered, + * because some laptop panels having a significant resume + * delay before the panel begins responding. + * + * This is likely a bit of a hack, but no better idea for + * handling this at the moment. + */ + msleep(300); + } + + OUTP_DBG(outp, "aux power -> always"); + nvkm_i2c_aux_monitor(aux, true); + outp->dp.aux_pwr = true; + + /* Detect any LTTPRs before reading DPCD receiver caps. */ + if (!nvkm_rdaux(aux, DPCD_LTTPR_REV, outp->dp.lttpr, sizeof(outp->dp.lttpr)) && + outp->dp.lttpr[0] >= 0x14 && outp->dp.lttpr[2]) { + switch (outp->dp.lttpr[2]) { + case 0x80: outp->dp.lttprs = 1; break; + case 0x40: outp->dp.lttprs = 2; break; + case 0x20: outp->dp.lttprs = 3; break; + case 0x10: outp->dp.lttprs = 4; break; + case 0x08: outp->dp.lttprs = 5; break; + case 0x04: outp->dp.lttprs = 6; break; + case 0x02: outp->dp.lttprs = 7; break; + case 0x01: outp->dp.lttprs = 8; break; + default: + /* Unknown LTTPR count, we'll switch to transparent mode. */ + WARN_ON(1); + outp->dp.lttprs = 0; + break; + } + } else { + /* No LTTPR support, or zero LTTPR count - don't touch it at all. */ + memset(outp->dp.lttpr, 0x00, sizeof(outp->dp.lttpr)); + } + + if (!nvkm_dp_read_dpcd_caps(outp)) { + const u8 rates[] = { 0x1e, 0x14, 0x0a, 0x06, 0 }; + const u8 *rate; + int rate_max; + + outp->dp.rates = 0; + outp->dp.links = outp->dp.dpcd[DPCD_RC02] & DPCD_RC02_MAX_LANE_COUNT; + outp->dp.links = min(outp->dp.links, outp->info.dpconf.link_nr); + if (outp->dp.lttprs && outp->dp.lttpr[4]) + outp->dp.links = min_t(int, outp->dp.links, outp->dp.lttpr[4]); + + rate_max = outp->dp.dpcd[DPCD_RC01_MAX_LINK_RATE]; + rate_max = min(rate_max, outp->info.dpconf.link_bw); + if (outp->dp.lttprs && outp->dp.lttpr[1]) + rate_max = min_t(int, rate_max, outp->dp.lttpr[1]); + + if (!nvkm_dp_enable_supported_link_rates(outp)) { + for (rate = rates; *rate; rate++) { + if (*rate > rate_max) + continue; + + if (WARN_ON(outp->dp.rates == ARRAY_SIZE(outp->dp.rate))) + break; + + outp->dp.rate[outp->dp.rates].dpcd = -1; + outp->dp.rate[outp->dp.rates].rate = *rate * 27000; + outp->dp.rates++; + } + } + } + } else + if (!auxpwr && outp->dp.aux_pwr) { + OUTP_DBG(outp, "aux power -> demand"); + nvkm_i2c_aux_monitor(aux, false); + outp->dp.aux_pwr = false; + atomic_set(&outp->dp.lt.done, 0); + + /* Restore eDP panel GPIO to its prior state if we changed it, as + * it could potentially interfere with other outputs. + */ + if (outp->conn->info.type == DCB_CONNECTOR_eDP) { + if (outp->dp.aux_pwr_pu) { + nvkm_gpio_set(gpio, 0, DCB_GPIO_PANEL_POWER, 0xff, 0); + outp->dp.aux_pwr_pu = false; + } + } + } +} + +static void +nvkm_dp_fini(struct nvkm_outp *outp) +{ + nvkm_dp_enable(outp, false); +} + +static void +nvkm_dp_init(struct nvkm_outp *outp) +{ + nvkm_dp_enable(outp, outp->dp.enabled); +} + +static void * +nvkm_dp_dtor(struct nvkm_outp *outp) +{ + return outp; +} + +static const struct nvkm_outp_func +nvkm_dp_func = { + .dtor = nvkm_dp_dtor, + .init = nvkm_dp_init, + .fini = nvkm_dp_fini, + .acquire = nvkm_dp_acquire, + .release = nvkm_dp_release, + .disable = nvkm_dp_disable, +}; + +int +nvkm_dp_new(struct nvkm_disp *disp, int index, struct dcb_output *dcbE, struct nvkm_outp **poutp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_bios *bios = device->bios; + struct nvkm_i2c *i2c = device->i2c; + struct nvkm_outp *outp; + u8 hdr, cnt, len; + u32 data; + int ret; + + ret = nvkm_outp_new_(&nvkm_dp_func, disp, index, dcbE, poutp); + outp = *poutp; + if (ret) + return ret; + + if (dcbE->location == 0) + outp->dp.aux = nvkm_i2c_aux_find(i2c, NVKM_I2C_AUX_CCB(dcbE->i2c_index)); + else + outp->dp.aux = nvkm_i2c_aux_find(i2c, NVKM_I2C_AUX_EXT(dcbE->extdev)); + if (!outp->dp.aux) { + OUTP_ERR(outp, "no aux"); + return -EINVAL; + } + + /* bios data is not optional */ + data = nvbios_dpout_match(bios, outp->info.hasht, outp->info.hashm, + &outp->dp.version, &hdr, &cnt, &len, &outp->dp.info); + if (!data) { + OUTP_ERR(outp, "no bios dp data"); + return -EINVAL; + } + + OUTP_DBG(outp, "bios dp %02x %02x %02x %02x", outp->dp.version, hdr, cnt, len); + + mutex_init(&outp->dp.mutex); + atomic_set(&outp->dp.lt.done, 0); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.h new file mode 100644 index 0000000000..9a6be43916 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/dp.h @@ -0,0 +1,102 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_DP_H__ +#define __NVKM_DISP_DP_H__ +#include "outp.h" + +int nvkm_dp_new(struct nvkm_disp *, int index, struct dcb_output *, + struct nvkm_outp **); +void nvkm_dp_disable(struct nvkm_outp *, struct nvkm_ior *); +void nvkm_dp_enable(struct nvkm_outp *, bool auxpwr); + +/* DPCD Receiver Capabilities */ +#define DPCD_RC00_DPCD_REV 0x00000 +#define DPCD_RC01_MAX_LINK_RATE 0x00001 +#define DPCD_RC02 0x00002 +#define DPCD_RC02_ENHANCED_FRAME_CAP 0x80 +#define DPCD_RC02_TPS3_SUPPORTED 0x40 +#define DPCD_RC02_MAX_LANE_COUNT 0x1f +#define DPCD_RC03 0x00003 +#define DPCD_RC03_TPS4_SUPPORTED 0x80 +#define DPCD_RC03_MAX_DOWNSPREAD 0x01 +#define DPCD_RC0E 0x0000e +#define DPCD_RC0E_AUX_RD_INTERVAL 0x7f +#define DPCD_RC10_SUPPORTED_LINK_RATES(i) 0x00010 +#define DPCD_RC10_SUPPORTED_LINK_RATES__SIZE 16 + +/* DPCD Link Configuration */ +#define DPCD_LC00_LINK_BW_SET 0x00100 +#define DPCD_LC01 0x00101 +#define DPCD_LC01_ENHANCED_FRAME_EN 0x80 +#define DPCD_LC01_LANE_COUNT_SET 0x1f +#define DPCD_LC02 0x00102 +#define DPCD_LC02_TRAINING_PATTERN_SET 0x0f +#define DPCD_LC02_SCRAMBLING_DISABLE 0x20 +#define DPCD_LC03(l) ((l) + 0x00103) +#define DPCD_LC03_MAX_PRE_EMPHASIS_REACHED 0x20 +#define DPCD_LC03_PRE_EMPHASIS_SET 0x18 +#define DPCD_LC03_MAX_SWING_REACHED 0x04 +#define DPCD_LC03_VOLTAGE_SWING_SET 0x03 +#define DPCD_LC0F 0x0010f +#define DPCD_LC0F_LANE1_MAX_POST_CURSOR2_REACHED 0x40 +#define DPCD_LC0F_LANE1_POST_CURSOR2_SET 0x30 +#define DPCD_LC0F_LANE0_MAX_POST_CURSOR2_REACHED 0x04 +#define DPCD_LC0F_LANE0_POST_CURSOR2_SET 0x03 +#define DPCD_LC10 0x00110 +#define DPCD_LC10_LANE3_MAX_POST_CURSOR2_REACHED 0x40 +#define DPCD_LC10_LANE3_POST_CURSOR2_SET 0x30 +#define DPCD_LC10_LANE2_MAX_POST_CURSOR2_REACHED 0x04 +#define DPCD_LC10_LANE2_POST_CURSOR2_SET 0x03 +#define DPCD_LC15_LINK_RATE_SET 0x00115 +#define DPCD_LC15_LINK_RATE_SET_MASK 0x07 + +/* DPCD Link/Sink Status */ +#define DPCD_LS02 0x00202 +#define DPCD_LS02_LANE1_SYMBOL_LOCKED 0x40 +#define DPCD_LS02_LANE1_CHANNEL_EQ_DONE 0x20 +#define DPCD_LS02_LANE1_CR_DONE 0x10 +#define DPCD_LS02_LANE0_SYMBOL_LOCKED 0x04 +#define DPCD_LS02_LANE0_CHANNEL_EQ_DONE 0x02 +#define DPCD_LS02_LANE0_CR_DONE 0x01 +#define DPCD_LS03 0x00203 +#define DPCD_LS03_LANE3_SYMBOL_LOCKED 0x40 +#define DPCD_LS03_LANE3_CHANNEL_EQ_DONE 0x20 +#define DPCD_LS03_LANE3_CR_DONE 0x10 +#define DPCD_LS03_LANE2_SYMBOL_LOCKED 0x04 +#define DPCD_LS03_LANE2_CHANNEL_EQ_DONE 0x02 +#define DPCD_LS03_LANE2_CR_DONE 0x01 +#define DPCD_LS04 0x00204 +#define DPCD_LS04_LINK_STATUS_UPDATED 0x80 +#define DPCD_LS04_DOWNSTREAM_PORT_STATUS_CHANGED 0x40 +#define DPCD_LS04_INTERLANE_ALIGN_DONE 0x01 +#define DPCD_LS06 0x00206 +#define DPCD_LS06_LANE1_PRE_EMPHASIS 0xc0 +#define DPCD_LS06_LANE1_VOLTAGE_SWING 0x30 +#define DPCD_LS06_LANE0_PRE_EMPHASIS 0x0c +#define DPCD_LS06_LANE0_VOLTAGE_SWING 0x03 +#define DPCD_LS07 0x00207 +#define DPCD_LS07_LANE3_PRE_EMPHASIS 0xc0 +#define DPCD_LS07_LANE3_VOLTAGE_SWING 0x30 +#define DPCD_LS07_LANE2_PRE_EMPHASIS 0x0c +#define DPCD_LS07_LANE2_VOLTAGE_SWING 0x03 +#define DPCD_LS0C 0x0020c +#define DPCD_LS0C_LANE3_POST_CURSOR2 0xc0 +#define DPCD_LS0C_LANE2_POST_CURSOR2 0x30 +#define DPCD_LS0C_LANE1_POST_CURSOR2 0x0c +#define DPCD_LS0C_LANE0_POST_CURSOR2 0x03 + +/* DPCD Sink Control */ +#define DPCD_SC00 0x00600 +#define DPCD_SC00_SET_POWER 0x03 +#define DPCD_SC00_SET_POWER_D0 0x01 +#define DPCD_SC00_SET_POWER_D3 0x03 + +#define DPCD_LTTPR_REV 0xf0000 +#define DPCD_LTTPR_MODE 0xf0003 +#define DPCD_LTTPR_MODE_TRANSPARENT 0x55 +#define DPCD_LTTPR_MODE_NON_TRANSPARENT 0xaa +#define DPCD_LTTPR_PATTERN_SET(i) ((i - 1) * 0x50 + 0xf0010) +#define DPCD_LTTPR_LANE0_SET(i) ((i - 1) * 0x50 + 0xf0011) +#define DPCD_LTTPR_AUX_RD_INTERVAL(i) ((i - 1) * 0x50 + 0xf0020) +#define DPCD_LTTPR_LANE0_1_STATUS(i) ((i - 1) * 0x50 + 0xf0030) +#define DPCD_LTTPR_LANE0_1_ADJUST(i) ((i - 1) * 0x50 + 0xf0033) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c new file mode 100644 index 0000000000..23ae451ba4 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/g84.c @@ -0,0 +1,348 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" + +#include + +static void +g84_sor_hdmi_infoframe_vsi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe vsi; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x61653c + hoff, 0x00010001, 0x00010000); + if (!size) + return; + + pack_hdmi_infoframe(&vsi, data, size); + + nvkm_wr32(device, 0x616544 + hoff, vsi.header); + nvkm_wr32(device, 0x616548 + hoff, vsi.subpack0_low); + nvkm_wr32(device, 0x61654c + hoff, vsi.subpack0_high); + /* Is there a second (or up to fourth?) set of subpack registers here? */ + /* nvkm_wr32(device, 0x616550 + hoff, vsi.subpack1_low); */ + /* nvkm_wr32(device, 0x616554 + hoff, vsi.subpack1_high); */ + + nvkm_mask(device, 0x61653c + hoff, 0x00010001, 0x00010001); +} + +static void +g84_sor_hdmi_infoframe_avi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe avi; + const u32 hoff = head * 0x800; + + pack_hdmi_infoframe(&avi, data, size); + + nvkm_mask(device, 0x616520 + hoff, 0x00000001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x616528 + hoff, avi.header); + nvkm_wr32(device, 0x61652c + hoff, avi.subpack0_low); + nvkm_wr32(device, 0x616530 + hoff, avi.subpack0_high); + nvkm_wr32(device, 0x616534 + hoff, avi.subpack1_low); + nvkm_wr32(device, 0x616538 + hoff, avi.subpack1_high); + + nvkm_mask(device, 0x616520 + hoff, 0x00000001, 0x00000001); +} + + +static void +g84_sor_hdmi_ctrl(struct nvkm_ior *ior, int head, bool enable, u8 max_ac_packet, u8 rekey) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 ctrl = 0x40000000 * enable | + 0x1f000000 /* ??? */ | + max_ac_packet << 16 | + rekey; + const u32 hoff = head * 0x800; + + if (!(ctrl & 0x40000000)) { + nvkm_mask(device, 0x6165a4 + hoff, 0x40000000, 0x00000000); + nvkm_mask(device, 0x616500 + hoff, 0x00000001, 0x00000000); + return; + } + + /* Audio InfoFrame */ + nvkm_mask(device, 0x616500 + hoff, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x616508 + hoff, 0x000a0184); + nvkm_wr32(device, 0x61650c + hoff, 0x00000071); + nvkm_wr32(device, 0x616510 + hoff, 0x00000000); + nvkm_mask(device, 0x616500 + hoff, 0x00000001, 0x00000001); + + + nvkm_mask(device, 0x6165d0 + hoff, 0x00070001, 0x00010001); /* SPARE, HW_CTS */ + nvkm_mask(device, 0x616568 + hoff, 0x00010101, 0x00000000); /* ACR_CTRL, ?? */ + nvkm_mask(device, 0x616578 + hoff, 0x80000000, 0x80000000); /* ACR_0441_ENABLE */ + + /* ??? */ + nvkm_mask(device, 0x61733c, 0x00100000, 0x00100000); /* RESETF */ + nvkm_mask(device, 0x61733c, 0x10000000, 0x10000000); /* LOOKUP_EN */ + nvkm_mask(device, 0x61733c, 0x00100000, 0x00000000); /* !RESETF */ + + /* HDMI_CTRL */ + nvkm_mask(device, 0x6165a4 + hoff, 0x5f1f007f, ctrl); +} + +const struct nvkm_ior_func_hdmi +g84_sor_hdmi = { + .ctrl = g84_sor_hdmi_ctrl, + .infoframe_avi = g84_sor_hdmi_infoframe_avi, + .infoframe_vsi = g84_sor_hdmi_infoframe_vsi, +}; + +static const struct nvkm_ior_func +g84_sor = { + .state = nv50_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, + .hdmi = &g84_sor_hdmi, +}; + +int +g84_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&g84_sor, disp, SOR, id, false); +} + +static const struct nvkm_disp_mthd_list +g84_disp_ovly_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x6109a0 }, + { 0x0088, 0x6109c0 }, + { 0x008c, 0x6109c8 }, + { 0x0090, 0x6109b4 }, + { 0x0094, 0x610970 }, + { 0x00a0, 0x610998 }, + { 0x00a4, 0x610964 }, + { 0x00c0, 0x610958 }, + { 0x00e0, 0x6109a8 }, + { 0x00e4, 0x6109d0 }, + { 0x00e8, 0x6109d8 }, + { 0x0100, 0x61094c }, + { 0x0104, 0x610984 }, + { 0x0108, 0x61098c }, + { 0x0800, 0x6109f8 }, + { 0x0808, 0x610a08 }, + { 0x080c, 0x610a10 }, + { 0x0810, 0x610a00 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +g84_disp_ovly_mthd = { + .name = "Overlay", + .addr = 0x000540, + .prev = 0x000004, + .data = { + { "Global", 1, &g84_disp_ovly_mthd_base }, + {} + } +}; + +const struct nvkm_disp_chan_user +g84_disp_ovly = { + .func = &nv50_disp_dmac_func, + .ctrl = 3, + .user = 3, + .mthd = &g84_disp_ovly_mthd, +}; + +static const struct nvkm_disp_mthd_list +g84_disp_base_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x0008c4 }, + { 0x0088, 0x0008d0 }, + { 0x008c, 0x0008dc }, + { 0x0090, 0x0008e4 }, + { 0x0094, 0x610884 }, + { 0x00a0, 0x6108a0 }, + { 0x00a4, 0x610878 }, + { 0x00c0, 0x61086c }, + { 0x00c4, 0x610800 }, + { 0x00c8, 0x61080c }, + { 0x00cc, 0x610818 }, + { 0x00e0, 0x610858 }, + { 0x00e4, 0x610860 }, + { 0x00e8, 0x6108ac }, + { 0x00ec, 0x6108b4 }, + { 0x00fc, 0x610824 }, + { 0x0100, 0x610894 }, + { 0x0104, 0x61082c }, + { 0x0110, 0x6108bc }, + { 0x0114, 0x61088c }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +g84_disp_base_mthd = { + .name = "Base", + .addr = 0x000540, + .prev = 0x000004, + .data = { + { "Global", 1, &g84_disp_base_mthd_base }, + { "Image", 2, &nv50_disp_base_mthd_image }, + {} + } +}; + +const struct nvkm_disp_chan_user +g84_disp_base = { + .func = &nv50_disp_dmac_func, + .ctrl = 1, + .user = 1, + .mthd = &g84_disp_base_mthd, +}; + +const struct nvkm_disp_mthd_list +g84_disp_core_mthd_dac = { + .mthd = 0x0080, + .addr = 0x000008, + .data = { + { 0x0400, 0x610b58 }, + { 0x0404, 0x610bdc }, + { 0x0420, 0x610bc4 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +g84_disp_core_mthd_head = { + .mthd = 0x0400, + .addr = 0x000540, + .data = { + { 0x0800, 0x610ad8 }, + { 0x0804, 0x610ad0 }, + { 0x0808, 0x610a48 }, + { 0x080c, 0x610a78 }, + { 0x0810, 0x610ac0 }, + { 0x0814, 0x610af8 }, + { 0x0818, 0x610b00 }, + { 0x081c, 0x610ae8 }, + { 0x0820, 0x610af0 }, + { 0x0824, 0x610b08 }, + { 0x0828, 0x610b10 }, + { 0x082c, 0x610a68 }, + { 0x0830, 0x610a60 }, + { 0x0834, 0x000000 }, + { 0x0838, 0x610a40 }, + { 0x0840, 0x610a24 }, + { 0x0844, 0x610a2c }, + { 0x0848, 0x610aa8 }, + { 0x084c, 0x610ab0 }, + { 0x085c, 0x610c5c }, + { 0x0860, 0x610a84 }, + { 0x0864, 0x610a90 }, + { 0x0868, 0x610b18 }, + { 0x086c, 0x610b20 }, + { 0x0870, 0x610ac8 }, + { 0x0874, 0x610a38 }, + { 0x0878, 0x610c50 }, + { 0x0880, 0x610a58 }, + { 0x0884, 0x610a9c }, + { 0x089c, 0x610c68 }, + { 0x08a0, 0x610a70 }, + { 0x08a4, 0x610a50 }, + { 0x08a8, 0x610ae0 }, + { 0x08c0, 0x610b28 }, + { 0x08c4, 0x610b30 }, + { 0x08c8, 0x610b40 }, + { 0x08d4, 0x610b38 }, + { 0x08d8, 0x610b48 }, + { 0x08dc, 0x610b50 }, + { 0x0900, 0x610a18 }, + { 0x0904, 0x610ab8 }, + { 0x0910, 0x610c70 }, + { 0x0914, 0x610c78 }, + {} + } +}; + +const struct nvkm_disp_chan_mthd +g84_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = 0x000004, + .data = { + { "Global", 1, &nv50_disp_core_mthd_base }, + { "DAC", 3, &g84_disp_core_mthd_dac }, + { "SOR", 2, &nv50_disp_core_mthd_sor }, + { "PIOR", 3, &nv50_disp_core_mthd_pior }, + { "HEAD", 2, &g84_disp_core_mthd_head }, + {} + } +}; + +const struct nvkm_disp_chan_user +g84_disp_core = { + .func = &nv50_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &g84_disp_core_mthd, +}; + +static const struct nvkm_disp_func +g84_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = nv50_sor_cnt, .new = g84_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,G82_DISP }, + .user = { + {{0,0,G82_DISP_CURSOR }, nvkm_disp_chan_new, &nv50_disp_curs }, + {{0,0,G82_DISP_OVERLAY }, nvkm_disp_chan_new, &nv50_disp_oimm }, + {{0,0,G82_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,G82_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g84_disp_core }, + {{0,0,G82_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, & g84_disp_ovly }, + {} + }, +}; + +int +g84_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&g84_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/g94.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/g94.c new file mode 100644 index 0000000000..67ef889a0c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/g94.c @@ -0,0 +1,378 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +#include + +void +g94_sor_dp_watermark(struct nvkm_ior *sor, int head, u8 watermark) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + + nvkm_mask(device, 0x61c128 + loff, 0x0000003f, watermark); +} + +void +g94_sor_dp_activesym(struct nvkm_ior *sor, int head, + u8 TU, u8 VTUa, u8 VTUf, u8 VTUi) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + + nvkm_mask(device, 0x61c10c + loff, 0x000001fc, TU << 2); + nvkm_mask(device, 0x61c128 + loff, 0x010f7f00, VTUa << 24 | VTUf << 16 | VTUi << 8); +} + +void +g94_sor_dp_audio_sym(struct nvkm_ior *sor, int head, u16 h, u32 v) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + + nvkm_mask(device, 0x61c1e8 + soff, 0x0000ffff, h); + nvkm_mask(device, 0x61c1ec + soff, 0x00ffffff, v); +} + +void +g94_sor_dp_drive(struct nvkm_ior *sor, int ln, int pc, int dc, int pe, int pu) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + const u32 shift = sor->func->dp->lanes[ln] * 8; + u32 data[3]; + + data[0] = nvkm_rd32(device, 0x61c118 + loff) & ~(0x000000ff << shift); + data[1] = nvkm_rd32(device, 0x61c120 + loff) & ~(0x000000ff << shift); + data[2] = nvkm_rd32(device, 0x61c130 + loff); + if ((data[2] & 0x0000ff00) < (pu << 8) || ln == 0) + data[2] = (data[2] & ~0x0000ff00) | (pu << 8); + + nvkm_wr32(device, 0x61c118 + loff, data[0] | (dc << shift)); + nvkm_wr32(device, 0x61c120 + loff, data[1] | (pe << shift)); + nvkm_wr32(device, 0x61c130 + loff, data[2]); +} + +void +g94_sor_dp_pattern(struct nvkm_ior *sor, int pattern) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + u32 data; + + switch (pattern) { + case 0: data = 0x00001000; break; + case 1: data = 0x01000000; break; + case 2: data = 0x02000000; break; + default: + WARN_ON(1); + return; + } + + nvkm_mask(device, 0x61c10c + loff, 0x0f001000, data); +} + +void +g94_sor_dp_power(struct nvkm_ior *sor, int nr) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 loff = nv50_sor_link(sor); + u32 mask = 0, i; + + for (i = 0; i < nr; i++) + mask |= 1 << sor->func->dp->lanes[i]; + + nvkm_mask(device, 0x61c130 + loff, 0x0000000f, mask); + nvkm_mask(device, 0x61c034 + soff, 0x80000000, 0x80000000); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c034 + soff) & 0x80000000)) + break; + ); +} + +int +g94_sor_dp_links(struct nvkm_ior *sor, struct nvkm_i2c_aux *aux) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 loff = nv50_sor_link(sor); + u32 dpctrl = 0x00000000; + u32 clksor = 0x00000000; + + dpctrl |= ((1 << sor->dp.nr) - 1) << 16; + if (sor->dp.ef) + dpctrl |= 0x00004000; + if (sor->dp.bw > 0x06) + clksor |= 0x00040000; + + nvkm_mask(device, 0x614300 + soff, 0x000c0000, clksor); + nvkm_mask(device, 0x61c10c + loff, 0x001f4000, dpctrl); + return 0; +} + +const struct nvkm_ior_func_dp +g94_sor_dp = { + .lanes = { 2, 1, 0, 3}, + .links = g94_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = g94_sor_dp_pattern, + .drive = g94_sor_dp_drive, + .audio_sym = g94_sor_dp_audio_sym, + .activesym = g94_sor_dp_activesym, + .watermark = g94_sor_dp_watermark, +}; + +static bool +g94_sor_war_needed(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + + if (sor->asy.proto == TMDS) { + switch (nvkm_rd32(device, 0x614300 + soff) & 0x00030000) { + case 0x00000000: + case 0x00030000: + return true; + default: + break; + } + } + + return false; +} + +static void +g94_sor_war_update_sppll1(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_ior *ior; + bool used = false; + u32 clksor; + + list_for_each_entry(ior, &disp->iors, head) { + if (ior->type != SOR) + continue; + + clksor = nvkm_rd32(device, 0x614300 + nv50_ior_base(ior)); + switch (clksor & 0x03000000) { + case 0x02000000: + case 0x03000000: + used = true; + break; + default: + break; + } + } + + if (used) + return; + + nvkm_mask(device, 0x00e840, 0x80000000, 0x00000000); +} + +static void +g94_sor_war_3(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + u32 sorpwr; + + if (!g94_sor_war_needed(sor)) + return; + + sorpwr = nvkm_rd32(device, 0x61c004 + soff); + if (sorpwr & 0x00000001) { + u32 seqctl = nvkm_rd32(device, 0x61c030 + soff); + u32 pd_pc = (seqctl & 0x00000f00) >> 8; + u32 pu_pc = seqctl & 0x0000000f; + + nvkm_wr32(device, 0x61c040 + soff + pd_pc * 4, 0x1f008000); + + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c030 + soff) & 0x10000000)) + break; + ); + nvkm_mask(device, 0x61c004 + soff, 0x80000001, 0x80000000); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c030 + soff) & 0x10000000)) + break; + ); + + nvkm_wr32(device, 0x61c040 + soff + pd_pc * 4, 0x00002000); + nvkm_wr32(device, 0x61c040 + soff + pu_pc * 4, 0x1f000000); + } + + nvkm_mask(device, 0x61c10c + soff, 0x00000001, 0x00000000); + nvkm_mask(device, 0x614300 + soff, 0x03000000, 0x00000000); + + if (sorpwr & 0x00000001) + nvkm_mask(device, 0x61c004 + soff, 0x80000001, 0x80000001); + + g94_sor_war_update_sppll1(sor->disp); +} + +static void +g94_sor_war_2(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + + if (!g94_sor_war_needed(sor)) + return; + + nvkm_mask(device, 0x00e840, 0x80000000, 0x80000000); + nvkm_mask(device, 0x614300 + soff, 0x03000000, 0x03000000); + nvkm_mask(device, 0x61c10c + soff, 0x00000001, 0x00000001); + + nvkm_mask(device, 0x61c00c + soff, 0x0f000000, 0x00000000); + nvkm_mask(device, 0x61c008 + soff, 0xff000000, 0x14000000); + nvkm_usec(device, 400, NVKM_DELAY); + nvkm_mask(device, 0x61c008 + soff, 0xff000000, 0x00000000); + nvkm_mask(device, 0x61c00c + soff, 0x0f000000, 0x01000000); + + if (nvkm_rd32(device, 0x61c004 + soff) & 0x00000001) { + u32 seqctl = nvkm_rd32(device, 0x61c030 + soff); + u32 pu_pc = seqctl & 0x0000000f; + nvkm_wr32(device, 0x61c040 + soff + pu_pc * 4, 0x1f008000); + } +} + +void +g94_sor_state(struct nvkm_ior *sor, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 coff = sor->id * 8 + (state == &sor->arm) * 4; + u32 ctrl = nvkm_rd32(device, 0x610794 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = LVDS; state->link = 1; break; + case 1: state->proto = TMDS; state->link = 1; break; + case 2: state->proto = TMDS; state->link = 2; break; + case 5: state->proto = TMDS; state->link = 3; break; + case 8: state->proto = DP; state->link = 1; break; + case 9: state->proto = DP; state->link = 2; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x00000003; + nv50_pior_depth(sor, state, ctrl); +} + +static const struct nvkm_ior_func +g94_sor = { + .state = g94_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, + .war_2 = g94_sor_war_2, + .war_3 = g94_sor_war_3, + .hdmi = &g84_sor_hdmi, + .dp = &g94_sor_dp, +}; + +static int +g94_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&g94_sor, disp, SOR, id, false); +} + +int +g94_sor_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = (nvkm_rd32(device, 0x610184) & 0x0f000000) >> 24; + return 4; +} + +static const struct nvkm_disp_mthd_list +g94_disp_core_mthd_sor = { + .mthd = 0x0040, + .addr = 0x000008, + .data = { + { 0x0600, 0x610794 }, + {} + } +}; + +const struct nvkm_disp_chan_mthd +g94_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = 0x000004, + .data = { + { "Global", 1, &nv50_disp_core_mthd_base }, + { "DAC", 3, &g84_disp_core_mthd_dac }, + { "SOR", 4, &g94_disp_core_mthd_sor }, + { "PIOR", 3, &nv50_disp_core_mthd_pior }, + { "HEAD", 2, &g84_disp_core_mthd_head }, + {} + } +}; + +const struct nvkm_disp_chan_user +g94_disp_core = { + .func = &nv50_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &g94_disp_core_mthd, +}; + +static const struct nvkm_disp_func +g94_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = g94_sor_cnt, .new = g94_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,GT206_DISP }, + .user = { + {{0,0, G82_DISP_CURSOR }, nvkm_disp_chan_new, & nv50_disp_curs }, + {{0,0, G82_DISP_OVERLAY }, nvkm_disp_chan_new, & nv50_disp_oimm }, + {{0,0,GT200_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,GT206_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g94_disp_core }, + {{0,0,GT200_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, >200_disp_ovly }, + {} + }, +}; + +int +g94_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&g94_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c new file mode 100644 index 0000000000..52099b75f5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ga102.c @@ -0,0 +1,150 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +#include + +static int +ga102_sor_dp_links(struct nvkm_ior *sor, struct nvkm_i2c_aux *aux) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 loff = nv50_sor_link(sor); + u32 dpctrl = 0x00000000; + u32 clksor = 0x00000000; + + switch (sor->dp.bw) { + case 0x06: clksor |= 0x00000000; break; + case 0x0a: clksor |= 0x00040000; break; + case 0x14: clksor |= 0x00080000; break; + case 0x1e: clksor |= 0x000c0000; break; + case 0x08: clksor |= 0x00100000; break; + case 0x09: clksor |= 0x00140000; break; + case 0x0c: clksor |= 0x00180000; break; + case 0x10: clksor |= 0x001c0000; break; + default: + WARN_ON(1); + return -EINVAL; + } + + dpctrl |= ((1 << sor->dp.nr) - 1) << 16; + if (sor->dp.mst) + dpctrl |= 0x40000000; + if (sor->dp.ef) + dpctrl |= 0x00004000; + + nvkm_mask(device, 0x612300 + soff, 0x007c0000, clksor); + + /*XXX*/ + nvkm_msec(device, 40, NVKM_DELAY); + nvkm_mask(device, 0x612300 + soff, 0x00030000, 0x00010000); + nvkm_mask(device, 0x61c10c + loff, 0x00000003, 0x00000001); + + nvkm_mask(device, 0x61c10c + loff, 0x401f4000, dpctrl); + return 0; +} + +static const struct nvkm_ior_func_dp +ga102_sor_dp = { + .lanes = { 0, 1, 2, 3 }, + .links = ga102_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gm107_sor_dp_pattern, + .drive = gm200_sor_dp_drive, + .vcpi = tu102_sor_dp_vcpi, + .audio = gv100_sor_dp_audio, + .audio_sym = gv100_sor_dp_audio_sym, + .watermark = gv100_sor_dp_watermark, +}; + +static void +ga102_sor_clock(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + u32 div2 = 0; + + if (sor->asy.proto == TMDS) { + if (sor->tmds.high_speed) + div2 = 1; + } + + nvkm_wr32(device, 0x00ec08 + (sor->id * 0x10), 0x00000000); + nvkm_wr32(device, 0x00ec04 + (sor->id * 0x10), div2); +} + +static const struct nvkm_ior_func +ga102_sor = { + .route = { + .get = gm200_sor_route_get, + .set = gm200_sor_route_set, + }, + .state = gv100_sor_state, + .power = nv50_sor_power, + .clock = ga102_sor_clock, + .hdmi = &gv100_sor_hdmi, + .dp = &ga102_sor_dp, + .hda = &gv100_sor_hda, +}; + +static int +ga102_sor_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 hda = nvkm_rd32(device, 0x08a15c); + + return nvkm_ior_new_(&ga102_sor, disp, SOR, id, hda & BIT(id)); +} + +static const struct nvkm_disp_func +ga102_disp = { + .oneinit = nv50_disp_oneinit, + .init = tu102_disp_init, + .fini = gv100_disp_fini, + .intr = gv100_disp_intr, + .super = gv100_disp_super, + .uevent = &gv100_disp_chan_uevent, + .wndw = { .cnt = gv100_disp_wndw_cnt }, + .head = { .cnt = gv100_head_cnt, .new = gv100_head_new }, + .sor = { .cnt = gv100_sor_cnt, .new = ga102_sor_new }, + .ramht_size = 0x2000, + .root = { 0, 0,GA102_DISP }, + .user = { + {{-1,-1,GV100_DISP_CAPS }, gv100_disp_caps_new }, + {{ 0, 0,GA102_DISP_CURSOR }, nvkm_disp_chan_new, &gv100_disp_curs }, + {{ 0, 0,GA102_DISP_WINDOW_IMM_CHANNEL_DMA}, nvkm_disp_wndw_new, &gv100_disp_wimm }, + {{ 0, 0,GA102_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gv100_disp_core }, + {{ 0, 0,GA102_DISP_WINDOW_CHANNEL_DMA }, nvkm_disp_wndw_new, &gv100_disp_wndw }, + {} + }, +}; + +int +ga102_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&ga102_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c new file mode 100644 index 0000000000..a48e9bdf4c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gf119.c @@ -0,0 +1,1261 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" +#include "outp.h" + +#include +#include + +#include + +static void +gf119_sor_hda_device_entry(struct nvkm_ior *ior, int head) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 hoff = 0x800 * head; + + nvkm_mask(device, 0x616548 + hoff, 0x00000070, head << 4); +} + +void +gf119_sor_hda_eld(struct nvkm_ior *ior, int head, u8 *data, u8 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 soff = 0x030 * ior->id + (head * 0x04); + int i; + + for (i = 0; i < size; i++) + nvkm_wr32(device, 0x10ec00 + soff, (i << 8) | data[i]); + for (; i < 0x60; i++) + nvkm_wr32(device, 0x10ec00 + soff, (i << 8)); + nvkm_mask(device, 0x10ec10 + soff, 0x80000002, 0x80000002); +} + +void +gf119_sor_hda_hpd(struct nvkm_ior *ior, int head, bool present) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 soff = 0x030 * ior->id + (head * 0x04); + u32 data = 0x80000000; + u32 mask = 0x80000001; + + if (present) { + ior->func->hda->device_entry(ior, head); + data |= 0x00000001; + } else { + mask |= 0x00000002; + } + + nvkm_mask(device, 0x10ec10 + soff, mask, data); +} + +const struct nvkm_ior_func_hda +gf119_sor_hda = { + .hpd = gf119_sor_hda_hpd, + .eld = gf119_sor_hda_eld, + .device_entry = gf119_sor_hda_device_entry, +}; + +void +gf119_sor_dp_watermark(struct nvkm_ior *sor, int head, u8 watermark) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616610 + hoff, 0x0800003f, 0x08000000 | watermark); +} + +void +gf119_sor_dp_audio_sym(struct nvkm_ior *sor, int head, u16 h, u32 v) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616620 + hoff, 0x0000ffff, h); + nvkm_mask(device, 0x616624 + hoff, 0x00ffffff, v); +} + +void +gf119_sor_dp_audio(struct nvkm_ior *sor, int head, bool enable) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = 0x800 * head; + const u32 data = 0x80000000 | (0x00000001 * enable); + const u32 mask = 0x8000000d; + + nvkm_mask(device, 0x616618 + hoff, mask, data); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x616618 + hoff) & 0x80000000)) + break; + ); +} + +void +gf119_sor_dp_vcpi(struct nvkm_ior *sor, int head, u8 slot, u8 slot_nr, u16 pbn, u16 aligned) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616588 + hoff, 0x00003f3f, (slot_nr << 8) | slot); + nvkm_mask(device, 0x61658c + hoff, 0xffffffff, (aligned << 16) | pbn); +} + +void +gf119_sor_dp_drive(struct nvkm_ior *sor, int ln, int pc, int dc, int pe, int pu) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + const u32 shift = sor->func->dp->lanes[ln] * 8; + u32 data[4]; + + data[0] = nvkm_rd32(device, 0x61c118 + loff) & ~(0x000000ff << shift); + data[1] = nvkm_rd32(device, 0x61c120 + loff) & ~(0x000000ff << shift); + data[2] = nvkm_rd32(device, 0x61c130 + loff); + if ((data[2] & 0x0000ff00) < (pu << 8) || ln == 0) + data[2] = (data[2] & ~0x0000ff00) | (pu << 8); + + nvkm_wr32(device, 0x61c118 + loff, data[0] | (dc << shift)); + nvkm_wr32(device, 0x61c120 + loff, data[1] | (pe << shift)); + nvkm_wr32(device, 0x61c130 + loff, data[2]); + + data[3] = nvkm_rd32(device, 0x61c13c + loff) & ~(0x000000ff << shift); + nvkm_wr32(device, 0x61c13c + loff, data[3] | (pc << shift)); +} + +static void +gf119_sor_dp_pattern(struct nvkm_ior *sor, int pattern) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + u32 data; + + switch (pattern) { + case 0: data = 0x10101010; break; + case 1: data = 0x01010101; break; + case 2: data = 0x02020202; break; + case 3: data = 0x03030303; break; + default: + WARN_ON(1); + return; + } + + nvkm_mask(device, 0x61c110 + soff, 0x1f1f1f1f, data); +} + +int +gf119_sor_dp_links(struct nvkm_ior *sor, struct nvkm_i2c_aux *aux) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 loff = nv50_sor_link(sor); + u32 dpctrl = 0x00000000; + u32 clksor = 0x00000000; + + clksor |= sor->dp.bw << 18; + dpctrl |= ((1 << sor->dp.nr) - 1) << 16; + if (sor->dp.mst) + dpctrl |= 0x40000000; + if (sor->dp.ef) + dpctrl |= 0x00004000; + + nvkm_mask(device, 0x612300 + soff, 0x007c0000, clksor); + nvkm_mask(device, 0x61c10c + loff, 0x401f4000, dpctrl); + return 0; +} + +const struct nvkm_ior_func_dp +gf119_sor_dp = { + .lanes = { 2, 1, 0, 3 }, + .links = gf119_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gf119_sor_dp_pattern, + .drive = gf119_sor_dp_drive, + .vcpi = gf119_sor_dp_vcpi, + .audio = gf119_sor_dp_audio, + .audio_sym = gf119_sor_dp_audio_sym, + .watermark = gf119_sor_dp_watermark, +}; + +static void +gf119_sor_hdmi_infoframe_vsi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe vsi; + const u32 hoff = head * 0x800; + + pack_hdmi_infoframe(&vsi, data, size); + + nvkm_mask(device, 0x616730 + hoff, 0x00010001, 0x00010000); + if (!size) + return; + + /* + * These appear to be the audio infoframe registers, + * but no other set of infoframe registers has yet + * been found. + */ + nvkm_wr32(device, 0x616738 + hoff, vsi.header); + nvkm_wr32(device, 0x61673c + hoff, vsi.subpack0_low); + nvkm_wr32(device, 0x616740 + hoff, vsi.subpack0_high); + /* Is there a second (or further?) set of subpack registers here? */ + + nvkm_mask(device, 0x616730 + hoff, 0x00000001, 0x00000001); +} + +static void +gf119_sor_hdmi_infoframe_avi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe avi; + const u32 hoff = head * 0x800; + + pack_hdmi_infoframe(&avi, data, size); + + nvkm_mask(device, 0x616714 + hoff, 0x00000001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x61671c + hoff, avi.header); + nvkm_wr32(device, 0x616720 + hoff, avi.subpack0_low); + nvkm_wr32(device, 0x616724 + hoff, avi.subpack0_high); + nvkm_wr32(device, 0x616728 + hoff, avi.subpack1_low); + nvkm_wr32(device, 0x61672c + hoff, avi.subpack1_high); + + nvkm_mask(device, 0x616714 + hoff, 0x00000001, 0x00000001); +} + +static void +gf119_sor_hdmi_ctrl(struct nvkm_ior *ior, int head, bool enable, u8 max_ac_packet, u8 rekey) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 ctrl = 0x40000000 * enable | + max_ac_packet << 16 | + rekey; + const u32 hoff = head * 0x800; + + if (!(ctrl & 0x40000000)) { + nvkm_mask(device, 0x616798 + hoff, 0x40000000, 0x00000000); + nvkm_mask(device, 0x616730 + hoff, 0x00000001, 0x00000000); + nvkm_mask(device, 0x6167a4 + hoff, 0x00000001, 0x00000000); + nvkm_mask(device, 0x616714 + hoff, 0x00000001, 0x00000000); + return; + } + + /* ??? InfoFrame? */ + nvkm_mask(device, 0x6167a4 + hoff, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x6167ac + hoff, 0x00000010); + nvkm_mask(device, 0x6167a4 + hoff, 0x00000001, 0x00000001); + + /* HDMI_CTRL */ + nvkm_mask(device, 0x616798 + hoff, 0x401f007f, ctrl); +} + +static const struct nvkm_ior_func_hdmi +gf119_sor_hdmi = { + .ctrl = gf119_sor_hdmi_ctrl, + .infoframe_avi = gf119_sor_hdmi_infoframe_avi, + .infoframe_vsi = gf119_sor_hdmi_infoframe_vsi, +}; + +void +gf119_sor_clock(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + u32 div1 = sor->asy.link == 3; + u32 div2 = sor->asy.link == 3; + + if (sor->asy.proto == TMDS) { + const u32 speed = sor->tmds.high_speed ? 0x14 : 0x0a; + nvkm_mask(device, 0x612300 + soff, 0x007c0000, speed << 18); + if (sor->tmds.high_speed) + div2 = 1; + } + + nvkm_mask(device, 0x612300 + soff, 0x00000707, (div2 << 8) | div1); +} + +void +gf119_sor_state(struct nvkm_ior *sor, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 coff = (state == &sor->asy) * 0x20000 + sor->id * 0x20; + u32 ctrl = nvkm_rd32(device, 0x640200 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = LVDS; state->link = 1; break; + case 1: state->proto = TMDS; state->link = 1; break; + case 2: state->proto = TMDS; state->link = 2; break; + case 5: state->proto = TMDS; state->link = 3; break; + case 8: state->proto = DP; state->link = 1; break; + case 9: state->proto = DP; state->link = 2; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x0000000f; +} + +static const struct nvkm_ior_func +gf119_sor = { + .state = gf119_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gf119_sor_hdmi, + .dp = &gf119_sor_dp, + .hda = &gf119_sor_hda, +}; + +static int +gf119_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&gf119_sor, disp, SOR, id, true); +} + +int +gf119_sor_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + *pmask = (nvkm_rd32(device, 0x612004) & 0x0000ff00) >> 8; + return 8; +} + +static void +gf119_dac_clock(struct nvkm_ior *dac) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 doff = nv50_ior_base(dac); + nvkm_mask(device, 0x612280 + doff, 0x07070707, 0x00000000); +} + +static void +gf119_dac_state(struct nvkm_ior *dac, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 coff = (state == &dac->asy) * 0x20000 + dac->id * 0x20; + u32 ctrl = nvkm_rd32(device, 0x640180 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = CRT; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x0000000f; +} + +static const struct nvkm_ior_func +gf119_dac = { + .state = gf119_dac_state, + .power = nv50_dac_power, + .sense = nv50_dac_sense, + .clock = gf119_dac_clock, +}; + +int +gf119_dac_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&gf119_dac, disp, DAC, id, false); +} + +int +gf119_dac_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + *pmask = (nvkm_rd32(device, 0x612004) & 0x000000f0) >> 4; + return 4; +} + +static void +gf119_head_vblank_put(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x800; + nvkm_mask(device, 0x6100c0 + hoff, 0x00000001, 0x00000000); +} + +static void +gf119_head_vblank_get(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x800; + nvkm_mask(device, 0x6100c0 + hoff, 0x00000001, 0x00000001); +} + +void +gf119_head_rgclk(struct nvkm_head *head, int div) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + nvkm_mask(device, 0x612200 + (head->id * 0x800), 0x0000000f, div); +} + +static void +gf119_head_state(struct nvkm_head *head, struct nvkm_head_state *state) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = (state == &head->asy) * 0x20000 + head->id * 0x300; + u32 data; + + data = nvkm_rd32(device, 0x640414 + hoff); + state->vtotal = (data & 0xffff0000) >> 16; + state->htotal = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x640418 + hoff); + state->vsynce = (data & 0xffff0000) >> 16; + state->hsynce = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x64041c + hoff); + state->vblanke = (data & 0xffff0000) >> 16; + state->hblanke = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x640420 + hoff); + state->vblanks = (data & 0xffff0000) >> 16; + state->hblanks = (data & 0x0000ffff); + state->hz = nvkm_rd32(device, 0x640450 + hoff); + + data = nvkm_rd32(device, 0x640404 + hoff); + switch ((data & 0x000003c0) >> 6) { + case 6: state->or.depth = 30; break; + case 5: state->or.depth = 24; break; + case 2: state->or.depth = 18; break; + case 0: state->or.depth = 18; break; /*XXX: "default" */ + default: + state->or.depth = 18; + WARN_ON(1); + break; + } +} + +static const struct nvkm_head_func +gf119_head = { + .state = gf119_head_state, + .rgpos = nv50_head_rgpos, + .rgclk = gf119_head_rgclk, + .vblank_get = gf119_head_vblank_get, + .vblank_put = gf119_head_vblank_put, +}; + +int +gf119_head_new(struct nvkm_disp *disp, int id) +{ + return nvkm_head_new_(&gf119_head, disp, id); +} + +int +gf119_head_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + *pmask = nvkm_rd32(device, 0x612004) & 0x0000000f; + return nvkm_rd32(device, 0x022448); +} + +static void +gf119_disp_chan_uevent_fini(struct nvkm_event *event, int type, int index) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), uevent); + struct nvkm_device *device = disp->engine.subdev.device; + nvkm_mask(device, 0x610090, 0x00000001 << index, 0x00000000 << index); + nvkm_wr32(device, 0x61008c, 0x00000001 << index); +} + +static void +gf119_disp_chan_uevent_init(struct nvkm_event *event, int types, int index) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), uevent); + struct nvkm_device *device = disp->engine.subdev.device; + nvkm_wr32(device, 0x61008c, 0x00000001 << index); + nvkm_mask(device, 0x610090, 0x00000001 << index, 0x00000001 << index); +} + +const struct nvkm_event_func +gf119_disp_chan_uevent = { + .init = gf119_disp_chan_uevent_init, + .fini = gf119_disp_chan_uevent_fini, +}; + +void +gf119_disp_chan_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00000001 << chan->chid.user; + if (!en) { + nvkm_mask(device, 0x610090, mask, 0x00000000); + nvkm_mask(device, 0x6100a0, mask, 0x00000000); + } else { + nvkm_mask(device, 0x6100a0, mask, mask); + } +} + +static void +gf119_disp_pioc_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_disp *disp = chan->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + nvkm_mask(device, 0x610490 + (ctrl * 0x10), 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x00030000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d fini: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); + } +} + +static int +gf119_disp_pioc_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_disp *disp = chan->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* activate channel */ + nvkm_wr32(device, 0x610490 + (ctrl * 0x10), 0x00000001); + if (nvkm_msec(device, 2000, + u32 tmp = nvkm_rd32(device, 0x610490 + (ctrl * 0x10)); + if ((tmp & 0x00030000) == 0x00010000) + break; + ) < 0) { + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +gf119_disp_pioc_func = { + .init = gf119_disp_pioc_init, + .fini = gf119_disp_pioc_fini, + .intr = gf119_disp_chan_intr, + .user = nv50_disp_chan_user, +}; + +int +gf119_disp_dmac_bind(struct nvkm_disp_chan *chan, struct nvkm_object *object, u32 handle) +{ + return nvkm_ramht_insert(chan->disp->ramht, object, chan->chid.user, -9, handle, + chan->chid.user << 27 | 0x00000001); +} + +void +gf119_disp_dmac_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* deactivate channel */ + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00001010, 0x00001000); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000003, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x001e0000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d fini: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); + } + + chan->suspend_put = nvkm_rd32(device, 0x640000 + (ctrl * 0x1000)); +} + +static int +gf119_disp_dmac_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x610494 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x610498 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61049c + (ctrl * 0x0010), 0x00000001); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), chan->suspend_put); + nvkm_wr32(device, 0x610490 + (ctrl * 0x0010), 0x00000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +gf119_disp_dmac_func = { + .push = nv50_disp_dmac_push, + .init = gf119_disp_dmac_init, + .fini = gf119_disp_dmac_fini, + .intr = gf119_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = gf119_disp_dmac_bind, +}; + +const struct nvkm_disp_chan_user +gf119_disp_curs = { + .func = &gf119_disp_pioc_func, + .ctrl = 13, + .user = 13, +}; + +const struct nvkm_disp_chan_user +gf119_disp_oimm = { + .func = &gf119_disp_pioc_func, + .ctrl = 9, + .user = 9, +}; + +static const struct nvkm_disp_mthd_list +gf119_disp_ovly_mthd_base = { + .mthd = 0x0000, + .data = { + { 0x0080, 0x665080 }, + { 0x0084, 0x665084 }, + { 0x0088, 0x665088 }, + { 0x008c, 0x66508c }, + { 0x0090, 0x665090 }, + { 0x0094, 0x665094 }, + { 0x00a0, 0x6650a0 }, + { 0x00a4, 0x6650a4 }, + { 0x00b0, 0x6650b0 }, + { 0x00b4, 0x6650b4 }, + { 0x00b8, 0x6650b8 }, + { 0x00c0, 0x6650c0 }, + { 0x00e0, 0x6650e0 }, + { 0x00e4, 0x6650e4 }, + { 0x00e8, 0x6650e8 }, + { 0x0100, 0x665100 }, + { 0x0104, 0x665104 }, + { 0x0108, 0x665108 }, + { 0x010c, 0x66510c }, + { 0x0110, 0x665110 }, + { 0x0118, 0x665118 }, + { 0x011c, 0x66511c }, + { 0x0120, 0x665120 }, + { 0x0124, 0x665124 }, + { 0x0130, 0x665130 }, + { 0x0134, 0x665134 }, + { 0x0138, 0x665138 }, + { 0x013c, 0x66513c }, + { 0x0140, 0x665140 }, + { 0x0144, 0x665144 }, + { 0x0148, 0x665148 }, + { 0x014c, 0x66514c }, + { 0x0150, 0x665150 }, + { 0x0154, 0x665154 }, + { 0x0158, 0x665158 }, + { 0x015c, 0x66515c }, + { 0x0160, 0x665160 }, + { 0x0164, 0x665164 }, + { 0x0168, 0x665168 }, + { 0x016c, 0x66516c }, + { 0x0400, 0x665400 }, + { 0x0408, 0x665408 }, + { 0x040c, 0x66540c }, + { 0x0410, 0x665410 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +gf119_disp_ovly_mthd = { + .name = "Overlay", + .addr = 0x001000, + .prev = -0x020000, + .data = { + { "Global", 1, &gf119_disp_ovly_mthd_base }, + {} + } +}; + +static const struct nvkm_disp_chan_user +gf119_disp_ovly = { + .func = &gf119_disp_dmac_func, + .ctrl = 5, + .user = 5, + .mthd = &gf119_disp_ovly_mthd, +}; + +static const struct nvkm_disp_mthd_list +gf119_disp_base_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x661080 }, + { 0x0084, 0x661084 }, + { 0x0088, 0x661088 }, + { 0x008c, 0x66108c }, + { 0x0090, 0x661090 }, + { 0x0094, 0x661094 }, + { 0x00a0, 0x6610a0 }, + { 0x00a4, 0x6610a4 }, + { 0x00c0, 0x6610c0 }, + { 0x00c4, 0x6610c4 }, + { 0x00c8, 0x6610c8 }, + { 0x00cc, 0x6610cc }, + { 0x00e0, 0x6610e0 }, + { 0x00e4, 0x6610e4 }, + { 0x00e8, 0x6610e8 }, + { 0x00ec, 0x6610ec }, + { 0x00fc, 0x6610fc }, + { 0x0100, 0x661100 }, + { 0x0104, 0x661104 }, + { 0x0108, 0x661108 }, + { 0x010c, 0x66110c }, + { 0x0110, 0x661110 }, + { 0x0114, 0x661114 }, + { 0x0118, 0x661118 }, + { 0x011c, 0x66111c }, + { 0x0130, 0x661130 }, + { 0x0134, 0x661134 }, + { 0x0138, 0x661138 }, + { 0x013c, 0x66113c }, + { 0x0140, 0x661140 }, + { 0x0144, 0x661144 }, + { 0x0148, 0x661148 }, + { 0x014c, 0x66114c }, + { 0x0150, 0x661150 }, + { 0x0154, 0x661154 }, + { 0x0158, 0x661158 }, + { 0x015c, 0x66115c }, + { 0x0160, 0x661160 }, + { 0x0164, 0x661164 }, + { 0x0168, 0x661168 }, + { 0x016c, 0x66116c }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +gf119_disp_base_mthd_image = { + .mthd = 0x0020, + .addr = 0x000020, + .data = { + { 0x0400, 0x661400 }, + { 0x0404, 0x661404 }, + { 0x0408, 0x661408 }, + { 0x040c, 0x66140c }, + { 0x0410, 0x661410 }, + {} + } +}; + +const struct nvkm_disp_chan_mthd +gf119_disp_base_mthd = { + .name = "Base", + .addr = 0x001000, + .prev = -0x020000, + .data = { + { "Global", 1, &gf119_disp_base_mthd_base }, + { "Image", 2, &gf119_disp_base_mthd_image }, + {} + } +}; + +const struct nvkm_disp_chan_user +gf119_disp_base = { + .func = &gf119_disp_dmac_func, + .ctrl = 1, + .user = 1, + .mthd = &gf119_disp_base_mthd, +}; + +const struct nvkm_disp_mthd_list +gf119_disp_core_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x660080 }, + { 0x0084, 0x660084 }, + { 0x0088, 0x660088 }, + { 0x008c, 0x000000 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +gf119_disp_core_mthd_dac = { + .mthd = 0x0020, + .addr = 0x000020, + .data = { + { 0x0180, 0x660180 }, + { 0x0184, 0x660184 }, + { 0x0188, 0x660188 }, + { 0x0190, 0x660190 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +gf119_disp_core_mthd_sor = { + .mthd = 0x0020, + .addr = 0x000020, + .data = { + { 0x0200, 0x660200 }, + { 0x0204, 0x660204 }, + { 0x0208, 0x660208 }, + { 0x0210, 0x660210 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +gf119_disp_core_mthd_pior = { + .mthd = 0x0020, + .addr = 0x000020, + .data = { + { 0x0300, 0x660300 }, + { 0x0304, 0x660304 }, + { 0x0308, 0x660308 }, + { 0x0310, 0x660310 }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +gf119_disp_core_mthd_head = { + .mthd = 0x0300, + .addr = 0x000300, + .data = { + { 0x0400, 0x660400 }, + { 0x0404, 0x660404 }, + { 0x0408, 0x660408 }, + { 0x040c, 0x66040c }, + { 0x0410, 0x660410 }, + { 0x0414, 0x660414 }, + { 0x0418, 0x660418 }, + { 0x041c, 0x66041c }, + { 0x0420, 0x660420 }, + { 0x0424, 0x660424 }, + { 0x0428, 0x660428 }, + { 0x042c, 0x66042c }, + { 0x0430, 0x660430 }, + { 0x0434, 0x660434 }, + { 0x0438, 0x660438 }, + { 0x0440, 0x660440 }, + { 0x0444, 0x660444 }, + { 0x0448, 0x660448 }, + { 0x044c, 0x66044c }, + { 0x0450, 0x660450 }, + { 0x0454, 0x660454 }, + { 0x0458, 0x660458 }, + { 0x045c, 0x66045c }, + { 0x0460, 0x660460 }, + { 0x0468, 0x660468 }, + { 0x046c, 0x66046c }, + { 0x0470, 0x660470 }, + { 0x0474, 0x660474 }, + { 0x0480, 0x660480 }, + { 0x0484, 0x660484 }, + { 0x048c, 0x66048c }, + { 0x0490, 0x660490 }, + { 0x0494, 0x660494 }, + { 0x0498, 0x660498 }, + { 0x04b0, 0x6604b0 }, + { 0x04b8, 0x6604b8 }, + { 0x04bc, 0x6604bc }, + { 0x04c0, 0x6604c0 }, + { 0x04c4, 0x6604c4 }, + { 0x04c8, 0x6604c8 }, + { 0x04d0, 0x6604d0 }, + { 0x04d4, 0x6604d4 }, + { 0x04e0, 0x6604e0 }, + { 0x04e4, 0x6604e4 }, + { 0x04e8, 0x6604e8 }, + { 0x04ec, 0x6604ec }, + { 0x04f0, 0x6604f0 }, + { 0x04f4, 0x6604f4 }, + { 0x04f8, 0x6604f8 }, + { 0x04fc, 0x6604fc }, + { 0x0500, 0x660500 }, + { 0x0504, 0x660504 }, + { 0x0508, 0x660508 }, + { 0x050c, 0x66050c }, + { 0x0510, 0x660510 }, + { 0x0514, 0x660514 }, + { 0x0518, 0x660518 }, + { 0x051c, 0x66051c }, + { 0x052c, 0x66052c }, + { 0x0530, 0x660530 }, + { 0x054c, 0x66054c }, + { 0x0550, 0x660550 }, + { 0x0554, 0x660554 }, + { 0x0558, 0x660558 }, + { 0x055c, 0x66055c }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +gf119_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = -0x020000, + .data = { + { "Global", 1, &gf119_disp_core_mthd_base }, + { "DAC", 3, &gf119_disp_core_mthd_dac }, + { "SOR", 8, &gf119_disp_core_mthd_sor }, + { "PIOR", 4, &gf119_disp_core_mthd_pior }, + { "HEAD", 4, &gf119_disp_core_mthd_head }, + {} + } +}; + +void +gf119_disp_core_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + /* deactivate channel */ + nvkm_mask(device, 0x610490, 0x00000010, 0x00000000); + nvkm_mask(device, 0x610490, 0x00000003, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490) & 0x001e0000)) + break; + ) < 0) { + nvkm_error(subdev, "core fini: %08x\n", + nvkm_rd32(device, 0x610490)); + } + + chan->suspend_put = nvkm_rd32(device, 0x640000); +} + +static int +gf119_disp_core_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x610494, chan->push); + nvkm_wr32(device, 0x610498, 0x00010000); + nvkm_wr32(device, 0x61049c, 0x00000001); + nvkm_mask(device, 0x610490, 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000, chan->suspend_put); + nvkm_wr32(device, 0x610490, 0x01000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "core init: %08x\n", + nvkm_rd32(device, 0x610490)); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +gf119_disp_core_func = { + .push = nv50_disp_dmac_push, + .init = gf119_disp_core_init, + .fini = gf119_disp_core_fini, + .intr = gf119_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = gf119_disp_dmac_bind, +}; + +static const struct nvkm_disp_chan_user +gf119_disp_core = { + .func = &gf119_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &gf119_disp_core_mthd, +}; + +void +gf119_disp_super(struct work_struct *work) +{ + struct nvkm_disp *disp = container_of(work, struct nvkm_disp, super.work); + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_head *head; + u32 mask[4]; + + nvkm_debug(subdev, "supervisor %d\n", ffs(disp->super.pending)); + mutex_lock(&disp->super.mutex); + + list_for_each_entry(head, &disp->heads, head) { + mask[head->id] = nvkm_rd32(device, 0x6101d4 + (head->id * 0x800)); + HEAD_DBG(head, "%08x", mask[head->id]); + } + + if (disp->super.pending & 0x00000001) { + nv50_disp_chan_mthd(disp->chan[0], NV_DBG_DEBUG); + nv50_disp_super_1(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_1_0(disp, head); + } + } else + if (disp->super.pending & 0x00000002) { + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_2_0(disp, head); + } + nvkm_outp_route(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00010000)) + continue; + nv50_disp_super_2_1(disp, head); + } + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_2_2(disp, head); + } + } else + if (disp->super.pending & 0x00000004) { + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_3_0(disp, head); + } + } + + list_for_each_entry(head, &disp->heads, head) + nvkm_wr32(device, 0x6101d4 + (head->id * 0x800), 0x00000000); + + nvkm_wr32(device, 0x6101d0, 0x80000000); + mutex_unlock(&disp->super.mutex); +} + +void +gf119_disp_intr_error(struct nvkm_disp *disp, int chid) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x6101f0 + (chid * 12)); + u32 type = (stat & 0x00007000) >> 12; + u32 mthd = (stat & 0x00000ffc); + u32 data = nvkm_rd32(device, 0x6101f4 + (chid * 12)); + u32 code = nvkm_rd32(device, 0x6101f8 + (chid * 12)); + const struct nvkm_enum *reason = + nvkm_enum_find(nv50_disp_intr_error_type, type); + + nvkm_error(subdev, "chid %d stat %08x reason %d [%s] mthd %04x " + "data %08x code %08x\n", + chid, stat, type, reason ? reason->name : "", + mthd, data, code); + + if (chid < ARRAY_SIZE(disp->chan)) { + switch (mthd) { + case 0x0080: + nv50_disp_chan_mthd(disp->chan[chid], NV_DBG_ERROR); + break; + default: + break; + } + } + + nvkm_wr32(device, 0x61009c, (1 << chid)); + nvkm_wr32(device, 0x6101f0 + (chid * 12), 0x90000000); +} + +void +gf119_disp_intr(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_head *head; + u32 intr = nvkm_rd32(device, 0x610088); + + if (intr & 0x00000001) { + u32 stat = nvkm_rd32(device, 0x61008c); + while (stat) { + int chid = __ffs(stat); stat &= ~(1 << chid); + nv50_disp_chan_uevent_send(disp, chid); + nvkm_wr32(device, 0x61008c, 1 << chid); + } + intr &= ~0x00000001; + } + + if (intr & 0x00000002) { + u32 stat = nvkm_rd32(device, 0x61009c); + int chid = ffs(stat) - 1; + if (chid >= 0) + disp->func->intr_error(disp, chid); + intr &= ~0x00000002; + } + + if (intr & 0x00100000) { + u32 stat = nvkm_rd32(device, 0x6100ac); + if (stat & 0x00000007) { + disp->super.pending = (stat & 0x00000007); + queue_work(disp->super.wq, &disp->super.work); + nvkm_wr32(device, 0x6100ac, disp->super.pending); + stat &= ~0x00000007; + } + + if (stat) { + nvkm_warn(subdev, "intr24 %08x\n", stat); + nvkm_wr32(device, 0x6100ac, stat); + } + + intr &= ~0x00100000; + } + + list_for_each_entry(head, &disp->heads, head) { + const u32 hoff = head->id * 0x800; + u32 mask = 0x01000000 << head->id; + if (mask & intr) { + u32 stat = nvkm_rd32(device, 0x6100bc + hoff); + if (stat & 0x00000001) + nvkm_disp_vblank(disp, head->id); + nvkm_mask(device, 0x6100bc + hoff, 0, 0); + nvkm_rd32(device, 0x6100c0 + hoff); + } + } +} + +void +gf119_disp_fini(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + /* disable all interrupts */ + nvkm_wr32(device, 0x6100b0, 0x00000000); +} + +int +gf119_disp_init(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_head *head; + u32 tmp; + int i; + + /* The below segments of code copying values from one register to + * another appear to inform EVO of the display capabilities or + * something similar. + */ + + /* ... CRTC caps */ + list_for_each_entry(head, &disp->heads, head) { + const u32 hoff = head->id * 0x800; + tmp = nvkm_rd32(device, 0x616104 + hoff); + nvkm_wr32(device, 0x6101b4 + hoff, tmp); + tmp = nvkm_rd32(device, 0x616108 + hoff); + nvkm_wr32(device, 0x6101b8 + hoff, tmp); + tmp = nvkm_rd32(device, 0x61610c + hoff); + nvkm_wr32(device, 0x6101bc + hoff, tmp); + } + + /* ... DAC caps */ + for (i = 0; i < disp->dac.nr; i++) { + tmp = nvkm_rd32(device, 0x61a000 + (i * 0x800)); + nvkm_wr32(device, 0x6101c0 + (i * 0x800), tmp); + } + + /* ... SOR caps */ + for (i = 0; i < disp->sor.nr; i++) { + tmp = nvkm_rd32(device, 0x61c000 + (i * 0x800)); + nvkm_wr32(device, 0x6301c4 + (i * 0x800), tmp); + } + + /* steal display away from vbios, or something like that */ + if (nvkm_rd32(device, 0x6100ac) & 0x00000100) { + nvkm_wr32(device, 0x6100ac, 0x00000100); + nvkm_mask(device, 0x6194e8, 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x6194e8) & 0x00000002)) + break; + ) < 0) + return -EBUSY; + } + + /* point at display engine memory area (hash table, objects) */ + nvkm_wr32(device, 0x610010, (disp->inst->addr >> 8) | 9); + + /* enable supervisor interrupts, disable everything else */ + nvkm_wr32(device, 0x610090, 0x00000000); + nvkm_wr32(device, 0x6100a0, 0x00000000); + nvkm_wr32(device, 0x6100b0, 0x00000307); + + /* disable underflow reporting, preventing an intermittent issue + * on some gk104 boards where the production vbios left this + * setting enabled by default. + * + * ftp://download.nvidia.com/open-gpu-doc/gk104-disable-underflow-reporting/1/gk104-disable-underflow-reporting.txt + */ + list_for_each_entry(head, &disp->heads, head) { + const u32 hoff = head->id * 0x800; + nvkm_mask(device, 0x616308 + hoff, 0x00000111, 0x00000010); + } + + return 0; +} + +static const struct nvkm_disp_func +gf119_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .dac = { .cnt = gf119_dac_cnt, .new = gf119_dac_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gf119_sor_new }, + .root = { 0,0,GF110_DISP }, + .user = { + {{0,0,GF110_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GF110_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GF110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GF110_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gf119_disp_core }, + {{0,0,GF110_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gf119_disp_ovly }, + {} + }, +}; + +int +gf119_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gf119_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c new file mode 100644 index 0000000000..876a21a0ce --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk104.c @@ -0,0 +1,332 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" + +#include + +void +gk104_sor_hdmi_infoframe_vsi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe vsi; + const u32 hoff = head * 0x400; + + pack_hdmi_infoframe(&vsi, data, size); + + /* GENERIC(?) / Vendor InfoFrame? */ + nvkm_mask(device, 0x690100 + hoff, 0x00010001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x690108 + hoff, vsi.header); + nvkm_wr32(device, 0x69010c + hoff, vsi.subpack0_low); + nvkm_wr32(device, 0x690110 + hoff, vsi.subpack0_high); + /* Is there a second (or further?) set of subpack registers here? */ + nvkm_mask(device, 0x690100 + hoff, 0x00000001, 0x00000001); +} + +void +gk104_sor_hdmi_infoframe_avi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe avi; + const u32 hoff = head * 0x400; + + pack_hdmi_infoframe(&avi, data, size); + + /* AVI InfoFrame */ + nvkm_mask(device, 0x690000 + hoff, 0x00000001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x690008 + hoff, avi.header); + nvkm_wr32(device, 0x69000c + hoff, avi.subpack0_low); + nvkm_wr32(device, 0x690010 + hoff, avi.subpack0_high); + nvkm_wr32(device, 0x690014 + hoff, avi.subpack1_low); + nvkm_wr32(device, 0x690018 + hoff, avi.subpack1_high); + + nvkm_mask(device, 0x690000 + hoff, 0x00000001, 0x00000001); +} + +void +gk104_sor_hdmi_ctrl(struct nvkm_ior *ior, int head, bool enable, u8 max_ac_packet, u8 rekey) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 ctrl = 0x40000000 * enable | + max_ac_packet << 16 | + rekey; + const u32 hoff = head * 0x800; + const u32 hdmi = head * 0x400; + + if (!(ctrl & 0x40000000)) { + nvkm_mask(device, 0x616798 + hoff, 0x40000000, 0x00000000); + nvkm_mask(device, 0x690100 + hdmi, 0x00000001, 0x00000000); + nvkm_mask(device, 0x6900c0 + hdmi, 0x00000001, 0x00000000); + nvkm_mask(device, 0x690000 + hdmi, 0x00000001, 0x00000000); + return; + } + + /* ??? InfoFrame? */ + nvkm_mask(device, 0x6900c0 + hdmi, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x6900cc + hdmi, 0x00000010); + nvkm_mask(device, 0x6900c0 + hdmi, 0x00000001, 0x00000001); + + /* ??? */ + nvkm_wr32(device, 0x690080 + hdmi, 0x82000000); + + /* HDMI_CTRL */ + nvkm_mask(device, 0x616798 + hoff, 0x401f007f, ctrl); +} + +const struct nvkm_ior_func_hdmi +gk104_sor_hdmi = { + .ctrl = gk104_sor_hdmi_ctrl, + .infoframe_avi = gk104_sor_hdmi_infoframe_avi, + .infoframe_vsi = gk104_sor_hdmi_infoframe_vsi, +}; + +static const struct nvkm_ior_func +gk104_sor = { + .state = gf119_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gk104_sor_hdmi, + .dp = &gf119_sor_dp, + .hda = &gf119_sor_hda, +}; + +int +gk104_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&gk104_sor, disp, SOR, id, true); +} + +static const struct nvkm_disp_mthd_list +gk104_disp_ovly_mthd_base = { + .mthd = 0x0000, + .data = { + { 0x0080, 0x665080 }, + { 0x0084, 0x665084 }, + { 0x0088, 0x665088 }, + { 0x008c, 0x66508c }, + { 0x0090, 0x665090 }, + { 0x0094, 0x665094 }, + { 0x00a0, 0x6650a0 }, + { 0x00a4, 0x6650a4 }, + { 0x00b0, 0x6650b0 }, + { 0x00b4, 0x6650b4 }, + { 0x00b8, 0x6650b8 }, + { 0x00c0, 0x6650c0 }, + { 0x00c4, 0x6650c4 }, + { 0x00e0, 0x6650e0 }, + { 0x00e4, 0x6650e4 }, + { 0x00e8, 0x6650e8 }, + { 0x0100, 0x665100 }, + { 0x0104, 0x665104 }, + { 0x0108, 0x665108 }, + { 0x010c, 0x66510c }, + { 0x0110, 0x665110 }, + { 0x0118, 0x665118 }, + { 0x011c, 0x66511c }, + { 0x0120, 0x665120 }, + { 0x0124, 0x665124 }, + { 0x0130, 0x665130 }, + { 0x0134, 0x665134 }, + { 0x0138, 0x665138 }, + { 0x013c, 0x66513c }, + { 0x0140, 0x665140 }, + { 0x0144, 0x665144 }, + { 0x0148, 0x665148 }, + { 0x014c, 0x66514c }, + { 0x0150, 0x665150 }, + { 0x0154, 0x665154 }, + { 0x0158, 0x665158 }, + { 0x015c, 0x66515c }, + { 0x0160, 0x665160 }, + { 0x0164, 0x665164 }, + { 0x0168, 0x665168 }, + { 0x016c, 0x66516c }, + { 0x0400, 0x665400 }, + { 0x0404, 0x665404 }, + { 0x0408, 0x665408 }, + { 0x040c, 0x66540c }, + { 0x0410, 0x665410 }, + {} + } +}; + +const struct nvkm_disp_chan_mthd +gk104_disp_ovly_mthd = { + .name = "Overlay", + .addr = 0x001000, + .prev = -0x020000, + .data = { + { "Global", 1, &gk104_disp_ovly_mthd_base }, + {} + } +}; + +const struct nvkm_disp_chan_user +gk104_disp_ovly = { + .func = &gf119_disp_dmac_func, + .ctrl = 5, + .user = 5, + .mthd = &gk104_disp_ovly_mthd, +}; + +static const struct nvkm_disp_mthd_list +gk104_disp_core_mthd_head = { + .mthd = 0x0300, + .addr = 0x000300, + .data = { + { 0x0400, 0x660400 }, + { 0x0404, 0x660404 }, + { 0x0408, 0x660408 }, + { 0x040c, 0x66040c }, + { 0x0410, 0x660410 }, + { 0x0414, 0x660414 }, + { 0x0418, 0x660418 }, + { 0x041c, 0x66041c }, + { 0x0420, 0x660420 }, + { 0x0424, 0x660424 }, + { 0x0428, 0x660428 }, + { 0x042c, 0x66042c }, + { 0x0430, 0x660430 }, + { 0x0434, 0x660434 }, + { 0x0438, 0x660438 }, + { 0x0440, 0x660440 }, + { 0x0444, 0x660444 }, + { 0x0448, 0x660448 }, + { 0x044c, 0x66044c }, + { 0x0450, 0x660450 }, + { 0x0454, 0x660454 }, + { 0x0458, 0x660458 }, + { 0x045c, 0x66045c }, + { 0x0460, 0x660460 }, + { 0x0468, 0x660468 }, + { 0x046c, 0x66046c }, + { 0x0470, 0x660470 }, + { 0x0474, 0x660474 }, + { 0x047c, 0x66047c }, + { 0x0480, 0x660480 }, + { 0x0484, 0x660484 }, + { 0x0488, 0x660488 }, + { 0x048c, 0x66048c }, + { 0x0490, 0x660490 }, + { 0x0494, 0x660494 }, + { 0x0498, 0x660498 }, + { 0x04a0, 0x6604a0 }, + { 0x04b0, 0x6604b0 }, + { 0x04b8, 0x6604b8 }, + { 0x04bc, 0x6604bc }, + { 0x04c0, 0x6604c0 }, + { 0x04c4, 0x6604c4 }, + { 0x04c8, 0x6604c8 }, + { 0x04d0, 0x6604d0 }, + { 0x04d4, 0x6604d4 }, + { 0x04e0, 0x6604e0 }, + { 0x04e4, 0x6604e4 }, + { 0x04e8, 0x6604e8 }, + { 0x04ec, 0x6604ec }, + { 0x04f0, 0x6604f0 }, + { 0x04f4, 0x6604f4 }, + { 0x04f8, 0x6604f8 }, + { 0x04fc, 0x6604fc }, + { 0x0500, 0x660500 }, + { 0x0504, 0x660504 }, + { 0x0508, 0x660508 }, + { 0x050c, 0x66050c }, + { 0x0510, 0x660510 }, + { 0x0514, 0x660514 }, + { 0x0518, 0x660518 }, + { 0x051c, 0x66051c }, + { 0x0520, 0x660520 }, + { 0x0524, 0x660524 }, + { 0x052c, 0x66052c }, + { 0x0530, 0x660530 }, + { 0x054c, 0x66054c }, + { 0x0550, 0x660550 }, + { 0x0554, 0x660554 }, + { 0x0558, 0x660558 }, + { 0x055c, 0x66055c }, + {} + } +}; + +const struct nvkm_disp_chan_mthd +gk104_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = -0x020000, + .data = { + { "Global", 1, &gf119_disp_core_mthd_base }, + { "DAC", 3, &gf119_disp_core_mthd_dac }, + { "SOR", 8, &gf119_disp_core_mthd_sor }, + { "PIOR", 4, &gf119_disp_core_mthd_pior }, + { "HEAD", 4, &gk104_disp_core_mthd_head }, + {} + } +}; + +const struct nvkm_disp_chan_user +gk104_disp_core = { + .func = &gf119_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &gk104_disp_core_mthd, +}; + +static const struct nvkm_disp_func +gk104_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .dac = { .cnt = gf119_dac_cnt, .new = gf119_dac_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gk104_sor_new }, + .root = { 0,0,GK104_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GK104_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GK104_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gk104_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gk104_disp_ovly }, + {} + }, +}; + +int +gk104_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gk104_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk110.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk110.c new file mode 100644 index 0000000000..1704aa381e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gk110.c @@ -0,0 +1,59 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +static const struct nvkm_disp_func +gk110_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .dac = { .cnt = gf119_dac_cnt, .new = gf119_dac_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gk104_sor_new }, + .root = { 0,0,GK110_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GK110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GK110_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gk104_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gk104_disp_ovly }, + {} + }, +}; + +int +gk110_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gk110_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c new file mode 100644 index 0000000000..b4d8e86861 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm107.c @@ -0,0 +1,112 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +void +gm107_sor_dp_pattern(struct nvkm_ior *sor, int pattern) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + u32 mask = 0x1f1f1f1f, data; + + switch (pattern) { + case 0: data = 0x10101010; break; + case 1: data = 0x01010101; break; + case 2: data = 0x02020202; break; + case 3: data = 0x03030303; break; + case 4: data = 0x1b1b1b1b; break; + default: + WARN_ON(1); + return; + } + + if (sor->asy.link & 1) + nvkm_mask(device, 0x61c110 + soff, mask, data); + else + nvkm_mask(device, 0x61c12c + soff, mask, data); +} + +static const struct nvkm_ior_func_dp +gm107_sor_dp = { + .lanes = { 0, 1, 2, 3 }, + .links = gf119_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gm107_sor_dp_pattern, + .drive = gf119_sor_dp_drive, + .vcpi = gf119_sor_dp_vcpi, + .audio = gf119_sor_dp_audio, + .audio_sym = gf119_sor_dp_audio_sym, + .watermark = gf119_sor_dp_watermark, +}; + +static const struct nvkm_ior_func +gm107_sor = { + .state = gf119_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gk104_sor_hdmi, + .dp = &gm107_sor_dp, + .hda = &gf119_sor_hda, +}; + +static int +gm107_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&gm107_sor, disp, SOR, id, true); +} + +static const struct nvkm_disp_func +gm107_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .dac = { .cnt = gf119_dac_cnt, .new = gf119_dac_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gm107_sor_new }, + .root = { 0,0,GM107_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GK110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GM107_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gk104_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gk104_disp_ovly }, + {} + }, +}; + +int +gm107_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gm107_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c new file mode 100644 index 0000000000..562ebae57d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gm200.c @@ -0,0 +1,187 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" +#include "outp.h" + +#include + +void +gm200_sor_dp_drive(struct nvkm_ior *sor, int ln, int pc, int dc, int pe, int pu) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 loff = nv50_sor_link(sor); + const u32 shift = sor->func->dp->lanes[ln] * 8; + u32 data[4]; + + pu &= 0x0f; + + data[0] = nvkm_rd32(device, 0x61c118 + loff) & ~(0x000000ff << shift); + data[1] = nvkm_rd32(device, 0x61c120 + loff) & ~(0x000000ff << shift); + data[2] = nvkm_rd32(device, 0x61c130 + loff); + if ((data[2] & 0x00000f00) < (pu << 8) || ln == 0) + data[2] = (data[2] & ~0x00000f00) | (pu << 8); + + nvkm_wr32(device, 0x61c118 + loff, data[0] | (dc << shift)); + nvkm_wr32(device, 0x61c120 + loff, data[1] | (pe << shift)); + nvkm_wr32(device, 0x61c130 + loff, data[2]); + + data[3] = nvkm_rd32(device, 0x61c13c + loff) & ~(0x000000ff << shift); + nvkm_wr32(device, 0x61c13c + loff, data[3] | (pc << shift)); +} + +const struct nvkm_ior_func_dp +gm200_sor_dp = { + .lanes = { 0, 1, 2, 3 }, + .links = gf119_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gm107_sor_dp_pattern, + .drive = gm200_sor_dp_drive, + .vcpi = gf119_sor_dp_vcpi, + .audio = gf119_sor_dp_audio, + .audio_sym = gf119_sor_dp_audio_sym, + .watermark = gf119_sor_dp_watermark, +}; + +void +gm200_sor_hdmi_scdc(struct nvkm_ior *ior, u8 scdc) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(ior); + const u32 ctrl = scdc & 0x3; + + nvkm_mask(device, 0x61c5bc + soff, 0x00000003, ctrl); + + ior->tmds.high_speed = !!(scdc & 0x2); +} + +const struct nvkm_ior_func_hdmi +gm200_sor_hdmi = { + .ctrl = gk104_sor_hdmi_ctrl, + .scdc = gm200_sor_hdmi_scdc, + .infoframe_avi = gk104_sor_hdmi_infoframe_avi, + .infoframe_vsi = gk104_sor_hdmi_infoframe_vsi, +}; + +void +gm200_sor_route_set(struct nvkm_outp *outp, struct nvkm_ior *ior) +{ + struct nvkm_device *device = outp->disp->engine.subdev.device; + const u32 moff = __ffs(outp->info.or) * 0x100; + const u32 sor = ior ? ior->id + 1 : 0; + u32 link = ior ? (ior->asy.link == 2) : 0; + + if (outp->info.sorconf.link & 1) { + nvkm_mask(device, 0x612308 + moff, 0x0000001f, link << 4 | sor); + link++; + } + + if (outp->info.sorconf.link & 2) + nvkm_mask(device, 0x612388 + moff, 0x0000001f, link << 4 | sor); +} + +int +gm200_sor_route_get(struct nvkm_outp *outp, int *link) +{ + struct nvkm_device *device = outp->disp->engine.subdev.device; + const int sublinks = outp->info.sorconf.link; + int lnk[2], sor[2], m, s; + + for (*link = 0, m = __ffs(outp->info.or) * 2, s = 0; s < 2; m++, s++) { + if (sublinks & BIT(s)) { + u32 data = nvkm_rd32(device, 0x612308 + (m * 0x80)); + lnk[s] = (data & 0x00000010) >> 4; + sor[s] = (data & 0x0000000f); + if (!sor[s]) + return -1; + *link |= lnk[s]; + } + } + + if (sublinks == 3) { + if (sor[0] != sor[1] || WARN_ON(lnk[0] || !lnk[1])) + return -1; + } + + return ((sublinks & 1) ? sor[0] : sor[1]) - 1; +} + +static const struct nvkm_ior_func +gm200_sor = { + .route = { + .get = gm200_sor_route_get, + .set = gm200_sor_route_set, + }, + .state = gf119_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gm200_sor_hdmi, + .dp = &gm200_sor_dp, + .hda = &gf119_sor_hda, +}; + +static int +gm200_sor_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 hda; + + if (!((hda = nvkm_rd32(device, 0x08a15c)) & 0x40000000)) + hda = nvkm_rd32(device, 0x101034); + + return nvkm_ior_new_(&gm200_sor, disp, SOR, id, hda & BIT(id)); +} + +static const struct nvkm_disp_func +gm200_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .dac = { .cnt = gf119_dac_cnt, .new = gf119_dac_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gm200_sor_new }, + .root = { 0,0,GM200_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GK110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GM200_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gk104_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gk104_disp_ovly }, + {} + }, +}; + +int +gm200_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gm200_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c new file mode 100644 index 0000000000..7f1eb43320 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp100.c @@ -0,0 +1,84 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +static const struct nvkm_ior_func +gp100_sor = { + .route = { + .get = gm200_sor_route_get, + .set = gm200_sor_route_set, + }, + .state = gf119_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gm200_sor_hdmi, + .dp = &gm200_sor_dp, + .hda = &gf119_sor_hda, +}; + +int +gp100_sor_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 hda; + + if (!((hda = nvkm_rd32(device, 0x08a15c)) & 0x40000000)) + hda = nvkm_rd32(device, 0x10ebb0) >> 8; + + return nvkm_ior_new_(&gp100_sor, disp, SOR, id, hda & BIT(id)); +} + +static const struct nvkm_disp_func +gp100_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gf119_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gp100_sor_new }, + .root = { 0,0,GP100_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gf119_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gf119_disp_oimm }, + {{0,0,GK110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gf119_disp_base }, + {{0,0,GP100_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gk104_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gk104_disp_ovly }, + {} + }, +}; + +int +gp100_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gp100_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c new file mode 100644 index 0000000000..07e9aeec5e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gp102.c @@ -0,0 +1,200 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +#include + +static int +gp102_disp_dmac_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x611494 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x611498 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61149c + (ctrl * 0x0010), 0x00000001); + nvkm_mask(device, 0x610490 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), chan->suspend_put); + nvkm_wr32(device, 0x610490 + (ctrl * 0x0010), 0x00000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490 + (ctrl * 0x10)) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d init: %08x\n", user, + nvkm_rd32(device, 0x610490 + (ctrl * 0x10))); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +gp102_disp_dmac_func = { + .push = nv50_disp_dmac_push, + .init = gp102_disp_dmac_init, + .fini = gf119_disp_dmac_fini, + .intr = gf119_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = gf119_disp_dmac_bind, +}; + +static const struct nvkm_disp_chan_user +gp102_disp_curs = { + .func = &gf119_disp_pioc_func, + .ctrl = 13, + .user = 17, +}; + +static const struct nvkm_disp_chan_user +gp102_disp_oimm = { + .func = &gf119_disp_pioc_func, + .ctrl = 9, + .user = 13, +}; + +static const struct nvkm_disp_chan_user +gp102_disp_ovly = { + .func = &gp102_disp_dmac_func, + .ctrl = 5, + .user = 5, + .mthd = &gk104_disp_ovly_mthd, +}; + +static const struct nvkm_disp_chan_user +gp102_disp_base = { + .func = &gp102_disp_dmac_func, + .ctrl = 1, + .user = 1, + .mthd = &gf119_disp_base_mthd, +}; + +static int +gp102_disp_core_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x611494, chan->push); + nvkm_wr32(device, 0x611498, 0x00010000); + nvkm_wr32(device, 0x61149c, 0x00000001); + nvkm_mask(device, 0x610490, 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000, chan->suspend_put); + nvkm_wr32(device, 0x610490, 0x01000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610490) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "core init: %08x\n", + nvkm_rd32(device, 0x610490)); + return -EBUSY; + } + + return 0; +} + +static const struct nvkm_disp_chan_func +gp102_disp_core_func = { + .push = nv50_disp_dmac_push, + .init = gp102_disp_core_init, + .fini = gf119_disp_core_fini, + .intr = gf119_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = gf119_disp_dmac_bind, +}; + +static const struct nvkm_disp_chan_user +gp102_disp_core = { + .func = &gp102_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &gk104_disp_core_mthd, +}; + +static void +gp102_disp_intr_error(struct nvkm_disp *disp, int chid) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mthd = nvkm_rd32(device, 0x6111f0 + (chid * 12)); + u32 data = nvkm_rd32(device, 0x6111f4 + (chid * 12)); + u32 unkn = nvkm_rd32(device, 0x6111f8 + (chid * 12)); + + nvkm_error(subdev, "chid %d mthd %04x data %08x %08x %08x\n", + chid, (mthd & 0x0000ffc), data, mthd, unkn); + + if (chid < ARRAY_SIZE(disp->chan)) { + switch (mthd & 0xffc) { + case 0x0080: + nv50_disp_chan_mthd(disp->chan[chid], NV_DBG_ERROR); + break; + default: + break; + } + } + + nvkm_wr32(device, 0x61009c, (1 << chid)); + nvkm_wr32(device, 0x6111f0 + (chid * 12), 0x90000000); +} + +static const struct nvkm_disp_func +gp102_disp = { + .oneinit = nv50_disp_oneinit, + .init = gf119_disp_init, + .fini = gf119_disp_fini, + .intr = gf119_disp_intr, + .intr_error = gp102_disp_intr_error, + .super = gf119_disp_super, + .uevent = &gf119_disp_chan_uevent, + .head = { .cnt = gf119_head_cnt, .new = gf119_head_new }, + .sor = { .cnt = gf119_sor_cnt, .new = gp100_sor_new }, + .root = { 0,0,GP102_DISP }, + .user = { + {{0,0,GK104_DISP_CURSOR }, nvkm_disp_chan_new, &gp102_disp_curs }, + {{0,0,GK104_DISP_OVERLAY }, nvkm_disp_chan_new, &gp102_disp_oimm }, + {{0,0,GK110_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &gp102_disp_base }, + {{0,0,GP102_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gp102_disp_core }, + {{0,0,GK104_DISP_OVERLAY_CONTROL_DMA}, nvkm_disp_chan_new, &gp102_disp_ovly }, + {} + }, +}; + +int +gp102_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gp102_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt200.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt200.c new file mode 100644 index 0000000000..6f69c4e3ad --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt200.c @@ -0,0 +1,109 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +static const struct nvkm_disp_mthd_list +gt200_disp_ovly_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x6109a0 }, + { 0x0088, 0x6109c0 }, + { 0x008c, 0x6109c8 }, + { 0x0090, 0x6109b4 }, + { 0x0094, 0x610970 }, + { 0x00a0, 0x610998 }, + { 0x00a4, 0x610964 }, + { 0x00b0, 0x610c98 }, + { 0x00b4, 0x610ca4 }, + { 0x00b8, 0x610cac }, + { 0x00c0, 0x610958 }, + { 0x00e0, 0x6109a8 }, + { 0x00e4, 0x6109d0 }, + { 0x00e8, 0x6109d8 }, + { 0x0100, 0x61094c }, + { 0x0104, 0x610984 }, + { 0x0108, 0x61098c }, + { 0x0800, 0x6109f8 }, + { 0x0808, 0x610a08 }, + { 0x080c, 0x610a10 }, + { 0x0810, 0x610a00 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +gt200_disp_ovly_mthd = { + .name = "Overlay", + .addr = 0x000540, + .prev = 0x000004, + .data = { + { "Global", 1, >200_disp_ovly_mthd_base }, + {} + } +}; + +const struct nvkm_disp_chan_user +gt200_disp_ovly = { + .func = &nv50_disp_dmac_func, + .ctrl = 3, + .user = 3, + .mthd = >200_disp_ovly_mthd, +}; + +static const struct nvkm_disp_func +gt200_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = nv50_sor_cnt, .new = g84_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,GT200_DISP }, + .user = { + {{0,0, G82_DISP_CURSOR }, nvkm_disp_chan_new, & nv50_disp_curs }, + {{0,0, G82_DISP_OVERLAY }, nvkm_disp_chan_new, & nv50_disp_oimm }, + {{0,0,GT200_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,GT200_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g84_disp_core }, + {{0,0,GT200_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, >200_disp_ovly }, + {} + }, +}; + +int +gt200_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(>200_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c new file mode 100644 index 0000000000..506ffbe7b8 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gt215.c @@ -0,0 +1,229 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" + +#include + +#include + +static void +gt215_sor_hda_eld(struct nvkm_ior *ior, int head, u8 *data, u8 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 soff = ior->id * 0x800; + int i; + + for (i = 0; i < size; i++) + nvkm_wr32(device, 0x61c440 + soff, (i << 8) | data[i]); + for (; i < 0x60; i++) + nvkm_wr32(device, 0x61c440 + soff, (i << 8)); + nvkm_mask(device, 0x61c448 + soff, 0x80000002, 0x80000002); +} + +static void +gt215_sor_hda_hpd(struct nvkm_ior *ior, int head, bool present) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + u32 data = 0x80000000; + u32 mask = 0x80000001; + if (present) + data |= 0x00000001; + else + mask |= 0x00000002; + nvkm_mask(device, 0x61c448 + ior->id * 0x800, mask, data); +} + +const struct nvkm_ior_func_hda +gt215_sor_hda = { + .hpd = gt215_sor_hda_hpd, + .eld = gt215_sor_hda_eld, +}; + +void +gt215_sor_dp_audio(struct nvkm_ior *sor, int head, bool enable) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 data = 0x80000000 | (0x00000001 * enable); + const u32 mask = 0x8000000d; + + nvkm_mask(device, 0x61c1e0 + soff, mask, data); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c1e0 + soff) & 0x80000000)) + break; + ); +} + +static const struct nvkm_ior_func_dp +gt215_sor_dp = { + .lanes = { 2, 1, 0, 3 }, + .links = g94_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = g94_sor_dp_pattern, + .drive = g94_sor_dp_drive, + .audio = gt215_sor_dp_audio, + .audio_sym = g94_sor_dp_audio_sym, + .activesym = g94_sor_dp_activesym, + .watermark = g94_sor_dp_watermark, +}; + +static void +gt215_sor_hdmi_infoframe_vsi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe vsi; + const u32 soff = nv50_ior_base(ior); + + pack_hdmi_infoframe(&vsi, data, size); + + nvkm_mask(device, 0x61c53c + soff, 0x00010001, 0x00010000); + if (!size) + return; + + nvkm_wr32(device, 0x61c544 + soff, vsi.header); + nvkm_wr32(device, 0x61c548 + soff, vsi.subpack0_low); + nvkm_wr32(device, 0x61c54c + soff, vsi.subpack0_high); + /* Is there a second (or up to fourth?) set of subpack registers here? */ + /* nvkm_wr32(device, 0x61c550 + soff, vsi.subpack1_low); */ + /* nvkm_wr32(device, 0x61c554 + soff, vsi.subpack1_high); */ + + nvkm_mask(device, 0x61c53c + soff, 0x00010001, 0x00010001); +} + +static void +gt215_sor_hdmi_infoframe_avi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe avi; + const u32 soff = nv50_ior_base(ior); + + pack_hdmi_infoframe(&avi, data, size); + + nvkm_mask(device, 0x61c520 + soff, 0x00000001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x61c528 + soff, avi.header); + nvkm_wr32(device, 0x61c52c + soff, avi.subpack0_low); + nvkm_wr32(device, 0x61c530 + soff, avi.subpack0_high); + nvkm_wr32(device, 0x61c534 + soff, avi.subpack1_low); + nvkm_wr32(device, 0x61c538 + soff, avi.subpack1_high); + + nvkm_mask(device, 0x61c520 + soff, 0x00000001, 0x00000001); +} + +static void +gt215_sor_hdmi_ctrl(struct nvkm_ior *ior, int head, bool enable, u8 max_ac_packet, u8 rekey) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 ctrl = 0x40000000 * enable | + 0x1f000000 /* ??? */ | + max_ac_packet << 16 | + rekey; + const u32 soff = nv50_ior_base(ior); + + if (!(ctrl & 0x40000000)) { + nvkm_mask(device, 0x61c5a4 + soff, 0x40000000, 0x00000000); + nvkm_mask(device, 0x61c53c + soff, 0x00000001, 0x00000000); + nvkm_mask(device, 0x61c520 + soff, 0x00000001, 0x00000000); + nvkm_mask(device, 0x61c500 + soff, 0x00000001, 0x00000000); + return; + } + + /* Audio InfoFrame */ + nvkm_mask(device, 0x61c500 + soff, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x61c508 + soff, 0x000a0184); + nvkm_wr32(device, 0x61c50c + soff, 0x00000071); + nvkm_wr32(device, 0x61c510 + soff, 0x00000000); + nvkm_mask(device, 0x61c500 + soff, 0x00000001, 0x00000001); + + nvkm_mask(device, 0x61c5d0 + soff, 0x00070001, 0x00010001); /* SPARE, HW_CTS */ + nvkm_mask(device, 0x61c568 + soff, 0x00010101, 0x00000000); /* ACR_CTRL, ?? */ + nvkm_mask(device, 0x61c578 + soff, 0x80000000, 0x80000000); /* ACR_0441_ENABLE */ + + /* ??? */ + nvkm_mask(device, 0x61733c, 0x00100000, 0x00100000); /* RESETF */ + nvkm_mask(device, 0x61733c, 0x10000000, 0x10000000); /* LOOKUP_EN */ + nvkm_mask(device, 0x61733c, 0x00100000, 0x00000000); /* !RESETF */ + + /* HDMI_CTRL */ + nvkm_mask(device, 0x61c5a4 + soff, 0x5f1f007f, ctrl); +} + +const struct nvkm_ior_func_hdmi +gt215_sor_hdmi = { + .ctrl = gt215_sor_hdmi_ctrl, + .infoframe_avi = gt215_sor_hdmi_infoframe_avi, + .infoframe_vsi = gt215_sor_hdmi_infoframe_vsi, +}; + +static const struct nvkm_ior_func +gt215_sor = { + .state = g94_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, + .hdmi = >215_sor_hdmi, + .dp = >215_sor_dp, + .hda = >215_sor_hda, +}; + +static int +gt215_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(>215_sor, disp, SOR, id, true); +} + +static const struct nvkm_disp_func +gt215_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = g94_sor_cnt, .new = gt215_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,GT214_DISP }, + .user = { + {{0,0,GT214_DISP_CURSOR }, nvkm_disp_chan_new, & nv50_disp_curs }, + {{0,0,GT214_DISP_OVERLAY }, nvkm_disp_chan_new, & nv50_disp_oimm }, + {{0,0,GT214_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,GT214_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g94_disp_core }, + {{0,0,GT214_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, & g84_disp_ovly }, + {} + }, +}; + +int +gt215_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(>215_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c new file mode 100644 index 0000000000..4ebc030e40 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/gv100.c @@ -0,0 +1,1254 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "hdmi.h" +#include "head.h" +#include "ior.h" +#include "outp.h" + +#include +#include +#include +#include + +#include +#include + +static void +gv100_sor_hda_device_entry(struct nvkm_ior *ior, int head) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 hoff = 0x800 * head; + + nvkm_mask(device, 0x616528 + hoff, 0x00000070, head << 4); +} + +const struct nvkm_ior_func_hda +gv100_sor_hda = { + .hpd = gf119_sor_hda_hpd, + .eld = gf119_sor_hda_eld, + .device_entry = gv100_sor_hda_device_entry, +}; + +void +gv100_sor_dp_watermark(struct nvkm_ior *sor, int head, u8 watermark) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616550 + hoff, 0x0c00003f, 0x08000000 | watermark); +} + +void +gv100_sor_dp_audio_sym(struct nvkm_ior *sor, int head, u16 h, u32 v) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x616568 + hoff, 0x0000ffff, h); + nvkm_mask(device, 0x61656c + hoff, 0x00ffffff, v); +} + +void +gv100_sor_dp_audio(struct nvkm_ior *sor, int head, bool enable) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = 0x800 * head; + const u32 data = 0x80000000 | (0x00000001 * enable); + const u32 mask = 0x8000000d; + + nvkm_mask(device, 0x616560 + hoff, mask, data); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x616560 + hoff) & 0x80000000)) + break; + ); +} + +static const struct nvkm_ior_func_dp +gv100_sor_dp = { + .lanes = { 0, 1, 2, 3 }, + .links = gf119_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gm107_sor_dp_pattern, + .drive = gm200_sor_dp_drive, + .audio = gv100_sor_dp_audio, + .audio_sym = gv100_sor_dp_audio_sym, + .watermark = gv100_sor_dp_watermark, +}; + +static void +gv100_sor_hdmi_infoframe_vsi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe vsi; + const u32 hoff = head * 0x400; + + pack_hdmi_infoframe(&vsi, data, size); + + nvkm_mask(device, 0x6f0100 + hoff, 0x00010001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x6f0108 + hoff, vsi.header); + nvkm_wr32(device, 0x6f010c + hoff, vsi.subpack0_low); + nvkm_wr32(device, 0x6f0110 + hoff, vsi.subpack0_high); + nvkm_wr32(device, 0x6f0114 + hoff, 0x00000000); + nvkm_wr32(device, 0x6f0118 + hoff, 0x00000000); + nvkm_wr32(device, 0x6f011c + hoff, 0x00000000); + nvkm_wr32(device, 0x6f0120 + hoff, 0x00000000); + nvkm_wr32(device, 0x6f0124 + hoff, 0x00000000); + nvkm_mask(device, 0x6f0100 + hoff, 0x00000001, 0x00000001); +} + +static void +gv100_sor_hdmi_infoframe_avi(struct nvkm_ior *ior, int head, void *data, u32 size) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + struct packed_hdmi_infoframe avi; + const u32 hoff = head * 0x400; + + pack_hdmi_infoframe(&avi, data, size); + + nvkm_mask(device, 0x6f0000 + hoff, 0x00000001, 0x00000000); + if (!size) + return; + + nvkm_wr32(device, 0x6f0008 + hoff, avi.header); + nvkm_wr32(device, 0x6f000c + hoff, avi.subpack0_low); + nvkm_wr32(device, 0x6f0010 + hoff, avi.subpack0_high); + nvkm_wr32(device, 0x6f0014 + hoff, avi.subpack1_low); + nvkm_wr32(device, 0x6f0018 + hoff, avi.subpack1_high); + + nvkm_mask(device, 0x6f0000 + hoff, 0x00000001, 0x00000001); +} + +static void +gv100_sor_hdmi_ctrl(struct nvkm_ior *ior, int head, bool enable, u8 max_ac_packet, u8 rekey) +{ + struct nvkm_device *device = ior->disp->engine.subdev.device; + const u32 ctrl = 0x40000000 * enable | + max_ac_packet << 16 | + rekey; + const u32 hoff = head * 0x800; + const u32 hdmi = head * 0x400; + + if (!(ctrl & 0x40000000)) { + nvkm_mask(device, 0x6165c0 + hoff, 0x40000000, 0x00000000); + nvkm_mask(device, 0x6f0100 + hdmi, 0x00000001, 0x00000000); + nvkm_mask(device, 0x6f00c0 + hdmi, 0x00000001, 0x00000000); + nvkm_mask(device, 0x6f0000 + hdmi, 0x00000001, 0x00000000); + return; + } + + /* General Control (GCP). */ + nvkm_mask(device, 0x6f00c0 + hdmi, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x6f00cc + hdmi, 0x00000010); + nvkm_mask(device, 0x6f00c0 + hdmi, 0x00000001, 0x00000001); + + /* Audio Clock Regeneration (ACR). */ + nvkm_wr32(device, 0x6f0080 + hdmi, 0x82000000); + + /* NV_PDISP_SF_HDMI_CTRL. */ + nvkm_mask(device, 0x6165c0 + hoff, 0x401f007f, ctrl); +} + +const struct nvkm_ior_func_hdmi +gv100_sor_hdmi = { + .ctrl = gv100_sor_hdmi_ctrl, + .scdc = gm200_sor_hdmi_scdc, + .infoframe_avi = gv100_sor_hdmi_infoframe_avi, + .infoframe_vsi = gv100_sor_hdmi_infoframe_vsi, +}; + +void +gv100_sor_state(struct nvkm_ior *sor, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 coff = (state == &sor->arm) * 0x8000 + sor->id * 0x20; + u32 ctrl = nvkm_rd32(device, 0x680300 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = LVDS; state->link = 1; break; + case 1: state->proto = TMDS; state->link = 1; break; + case 2: state->proto = TMDS; state->link = 2; break; + case 5: state->proto = TMDS; state->link = 3; break; + case 8: state->proto = DP; state->link = 1; break; + case 9: state->proto = DP; state->link = 2; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x000000ff; +} + +static const struct nvkm_ior_func +gv100_sor = { + .route = { + .get = gm200_sor_route_get, + .set = gm200_sor_route_set, + }, + .state = gv100_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gv100_sor_hdmi, + .dp = &gv100_sor_dp, + .hda = &gv100_sor_hda, +}; + +static int +gv100_sor_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 hda; + + if (!((hda = nvkm_rd32(device, 0x08a15c)) & 0x40000000)) + hda = nvkm_rd32(device, 0x118fb0) >> 8; + + return nvkm_ior_new_(&gv100_sor, disp, SOR, id, hda & BIT(id)); +} + +int +gv100_sor_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = (nvkm_rd32(device, 0x610060) & 0x0000ff00) >> 8; + return (nvkm_rd32(device, 0x610074) & 0x00000f00) >> 8; +} + +static void +gv100_head_vblank_put(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + nvkm_mask(device, 0x611d80 + (head->id * 4), 0x00000004, 0x00000000); +} + +static void +gv100_head_vblank_get(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + nvkm_mask(device, 0x611d80 + (head->id * 4), 0x00000004, 0x00000004); +} + +static void +gv100_head_rgpos(struct nvkm_head *head, u16 *hline, u16 *vline) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x800; + /* vline read locks hline. */ + *vline = nvkm_rd32(device, 0x616330 + hoff) & 0x0000ffff; + *hline = nvkm_rd32(device, 0x616334 + hoff) & 0x0000ffff; +} + +static void +gv100_head_state(struct nvkm_head *head, struct nvkm_head_state *state) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = (state == &head->arm) * 0x8000 + head->id * 0x400; + u32 data; + + data = nvkm_rd32(device, 0x682064 + hoff); + state->vtotal = (data & 0xffff0000) >> 16; + state->htotal = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x682068 + hoff); + state->vsynce = (data & 0xffff0000) >> 16; + state->hsynce = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x68206c + hoff); + state->vblanke = (data & 0xffff0000) >> 16; + state->hblanke = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x682070 + hoff); + state->vblanks = (data & 0xffff0000) >> 16; + state->hblanks = (data & 0x0000ffff); + state->hz = nvkm_rd32(device, 0x68200c + hoff); + + data = nvkm_rd32(device, 0x682004 + hoff); + switch ((data & 0x000000f0) >> 4) { + case 5: state->or.depth = 30; break; + case 4: state->or.depth = 24; break; + case 1: state->or.depth = 18; break; + default: + state->or.depth = 18; + WARN_ON(1); + break; + } +} + +static const struct nvkm_head_func +gv100_head = { + .state = gv100_head_state, + .rgpos = gv100_head_rgpos, + .rgclk = gf119_head_rgclk, + .vblank_get = gv100_head_vblank_get, + .vblank_put = gv100_head_vblank_put, +}; + +int +gv100_head_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + if (!(nvkm_rd32(device, 0x610060) & (0x00000001 << id))) + return 0; + + return nvkm_head_new_(&gv100_head, disp, id); +} + +int +gv100_head_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = nvkm_rd32(device, 0x610060) & 0x000000ff; + return nvkm_rd32(device, 0x610074) & 0x0000000f; +} + +const struct nvkm_event_func +gv100_disp_chan_uevent = { +}; + +u64 +gv100_disp_chan_user(struct nvkm_disp_chan *chan, u64 *psize) +{ + *psize = 0x1000; + return 0x690000 + ((chan->chid.user - 1) * 0x1000); +} + +static int +gv100_disp_dmac_idle(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 soff = (chan->chid.ctrl - 1) * 0x04; + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x610664 + soff); + if ((stat & 0x000f0000) == 0x00040000) + return 0; + ); + return -EBUSY; +} + +int +gv100_disp_dmac_bind(struct nvkm_disp_chan *chan, + struct nvkm_object *object, u32 handle) +{ + return nvkm_ramht_insert(chan->disp->ramht, object, chan->chid.user, -9, handle, + chan->chid.user << 25 | 0x00000040); +} + +void +gv100_disp_dmac_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 uoff = (chan->chid.ctrl - 1) * 0x1000; + const u32 coff = chan->chid.ctrl * 0x04; + nvkm_mask(device, 0x6104e0 + coff, 0x00000010, 0x00000000); + gv100_disp_dmac_idle(chan); + nvkm_mask(device, 0x6104e0 + coff, 0x00000002, 0x00000000); + chan->suspend_put = nvkm_rd32(device, 0x690000 + uoff); +} + +int +gv100_disp_dmac_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 uoff = (chan->chid.ctrl - 1) * 0x1000; + const u32 poff = chan->chid.ctrl * 0x10; + const u32 coff = chan->chid.ctrl * 0x04; + + nvkm_wr32(device, 0x610b24 + poff, lower_32_bits(chan->push)); + nvkm_wr32(device, 0x610b20 + poff, upper_32_bits(chan->push)); + nvkm_wr32(device, 0x610b28 + poff, 0x00000001); + nvkm_wr32(device, 0x610b2c + poff, 0x00000040); + + nvkm_mask(device, 0x6104e0 + coff, 0x00000010, 0x00000010); + nvkm_wr32(device, 0x690000 + uoff, chan->suspend_put); + nvkm_wr32(device, 0x6104e0 + coff, 0x00000013); + return gv100_disp_dmac_idle(chan); +} + +static void +gv100_disp_wimm_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00000001 << chan->head; + const u32 data = en ? mask : 0; + nvkm_mask(device, 0x611da8, mask, data); +} + +static const struct nvkm_disp_chan_func +gv100_disp_wimm_func = { + .push = nv50_disp_dmac_push, + .init = gv100_disp_dmac_init, + .fini = gv100_disp_dmac_fini, + .intr = gv100_disp_wimm_intr, + .user = gv100_disp_chan_user, +}; + +const struct nvkm_disp_chan_user +gv100_disp_wimm = { + .func = &gv100_disp_wimm_func, + .ctrl = 33, + .user = 33, +}; + +static const struct nvkm_disp_mthd_list +gv100_disp_wndw_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0200, 0x690200 }, + { 0x020c, 0x69020c }, + { 0x0210, 0x690210 }, + { 0x0214, 0x690214 }, + { 0x0218, 0x690218 }, + { 0x021c, 0x69021c }, + { 0x0220, 0x690220 }, + { 0x0224, 0x690224 }, + { 0x0228, 0x690228 }, + { 0x022c, 0x69022c }, + { 0x0230, 0x690230 }, + { 0x0234, 0x690234 }, + { 0x0238, 0x690238 }, + { 0x0240, 0x690240 }, + { 0x0244, 0x690244 }, + { 0x0248, 0x690248 }, + { 0x024c, 0x69024c }, + { 0x0250, 0x690250 }, + { 0x0254, 0x690254 }, + { 0x0260, 0x690260 }, + { 0x0264, 0x690264 }, + { 0x0268, 0x690268 }, + { 0x026c, 0x69026c }, + { 0x0270, 0x690270 }, + { 0x0274, 0x690274 }, + { 0x0280, 0x690280 }, + { 0x0284, 0x690284 }, + { 0x0288, 0x690288 }, + { 0x028c, 0x69028c }, + { 0x0290, 0x690290 }, + { 0x0298, 0x690298 }, + { 0x029c, 0x69029c }, + { 0x02a0, 0x6902a0 }, + { 0x02a4, 0x6902a4 }, + { 0x02a8, 0x6902a8 }, + { 0x02ac, 0x6902ac }, + { 0x02b0, 0x6902b0 }, + { 0x02b4, 0x6902b4 }, + { 0x02b8, 0x6902b8 }, + { 0x02bc, 0x6902bc }, + { 0x02c0, 0x6902c0 }, + { 0x02c4, 0x6902c4 }, + { 0x02c8, 0x6902c8 }, + { 0x02cc, 0x6902cc }, + { 0x02d0, 0x6902d0 }, + { 0x02d4, 0x6902d4 }, + { 0x02d8, 0x6902d8 }, + { 0x02dc, 0x6902dc }, + { 0x02e0, 0x6902e0 }, + { 0x02e4, 0x6902e4 }, + { 0x02e8, 0x6902e8 }, + { 0x02ec, 0x6902ec }, + { 0x02f0, 0x6902f0 }, + { 0x02f4, 0x6902f4 }, + { 0x02f8, 0x6902f8 }, + { 0x02fc, 0x6902fc }, + { 0x0300, 0x690300 }, + { 0x0304, 0x690304 }, + { 0x0308, 0x690308 }, + { 0x0310, 0x690310 }, + { 0x0314, 0x690314 }, + { 0x0318, 0x690318 }, + { 0x031c, 0x69031c }, + { 0x0320, 0x690320 }, + { 0x0324, 0x690324 }, + { 0x0328, 0x690328 }, + { 0x032c, 0x69032c }, + { 0x033c, 0x69033c }, + { 0x0340, 0x690340 }, + { 0x0344, 0x690344 }, + { 0x0348, 0x690348 }, + { 0x034c, 0x69034c }, + { 0x0350, 0x690350 }, + { 0x0354, 0x690354 }, + { 0x0358, 0x690358 }, + { 0x0364, 0x690364 }, + { 0x0368, 0x690368 }, + { 0x036c, 0x69036c }, + { 0x0370, 0x690370 }, + { 0x0374, 0x690374 }, + { 0x0380, 0x690380 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +gv100_disp_wndw_mthd = { + .name = "Window", + .addr = 0x001000, + .prev = 0x000800, + .data = { + { "Global", 1, &gv100_disp_wndw_mthd_base }, + {} + } +}; + +static void +gv100_disp_wndw_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00000001 << chan->head; + const u32 data = en ? mask : 0; + nvkm_mask(device, 0x611da4, mask, data); +} + +static const struct nvkm_disp_chan_func +gv100_disp_wndw_func = { + .push = nv50_disp_dmac_push, + .init = gv100_disp_dmac_init, + .fini = gv100_disp_dmac_fini, + .intr = gv100_disp_wndw_intr, + .user = gv100_disp_chan_user, + .bind = gv100_disp_dmac_bind, +}; + +const struct nvkm_disp_chan_user +gv100_disp_wndw = { + .func = &gv100_disp_wndw_func, + .ctrl = 1, + .user = 1, + .mthd = &gv100_disp_wndw_mthd, +}; + +int +gv100_disp_wndw_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = nvkm_rd32(device, 0x610064); + return (nvkm_rd32(device, 0x610074) & 0x03f00000) >> 20; +} + +static int +gv100_disp_curs_idle(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 soff = (chan->chid.ctrl - 1) * 0x04; + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x610664 + soff); + if ((stat & 0x00070000) == 0x00040000) + return 0; + ); + return -EBUSY; +} + +static void +gv100_disp_curs_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00010000 << chan->head; + const u32 data = en ? mask : 0; + nvkm_mask(device, 0x611dac, mask, data); +} + +static void +gv100_disp_curs_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 hoff = chan->chid.ctrl * 4; + nvkm_mask(device, 0x6104e0 + hoff, 0x00000010, 0x00000010); + gv100_disp_curs_idle(chan); + nvkm_mask(device, 0x6104e0 + hoff, 0x00000001, 0x00000000); +} + +static int +gv100_disp_curs_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + nvkm_wr32(device, 0x6104e0 + chan->chid.ctrl * 4, 0x00000001); + return gv100_disp_curs_idle(chan); +} + +static const struct nvkm_disp_chan_func +gv100_disp_curs_func = { + .init = gv100_disp_curs_init, + .fini = gv100_disp_curs_fini, + .intr = gv100_disp_curs_intr, + .user = gv100_disp_chan_user, +}; + +const struct nvkm_disp_chan_user +gv100_disp_curs = { + .func = &gv100_disp_curs_func, + .ctrl = 73, + .user = 73, +}; + +static const struct nvkm_disp_mthd_list +gv100_disp_core_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0200, 0x680200 }, + { 0x0208, 0x680208 }, + { 0x020c, 0x68020c }, + { 0x0210, 0x680210 }, + { 0x0214, 0x680214 }, + { 0x0218, 0x680218 }, + { 0x021c, 0x68021c }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +gv100_disp_core_mthd_sor = { + .mthd = 0x0020, + .addr = 0x000020, + .data = { + { 0x0300, 0x680300 }, + { 0x0304, 0x680304 }, + { 0x0308, 0x680308 }, + { 0x030c, 0x68030c }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +gv100_disp_core_mthd_wndw = { + .mthd = 0x0080, + .addr = 0x000080, + .data = { + { 0x1000, 0x681000 }, + { 0x1004, 0x681004 }, + { 0x1008, 0x681008 }, + { 0x100c, 0x68100c }, + { 0x1010, 0x681010 }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +gv100_disp_core_mthd_head = { + .mthd = 0x0400, + .addr = 0x000400, + .data = { + { 0x2000, 0x682000 }, + { 0x2004, 0x682004 }, + { 0x2008, 0x682008 }, + { 0x200c, 0x68200c }, + { 0x2014, 0x682014 }, + { 0x2018, 0x682018 }, + { 0x201c, 0x68201c }, + { 0x2020, 0x682020 }, + { 0x2028, 0x682028 }, + { 0x202c, 0x68202c }, + { 0x2030, 0x682030 }, + { 0x2038, 0x682038 }, + { 0x203c, 0x68203c }, + { 0x2048, 0x682048 }, + { 0x204c, 0x68204c }, + { 0x2050, 0x682050 }, + { 0x2054, 0x682054 }, + { 0x2058, 0x682058 }, + { 0x205c, 0x68205c }, + { 0x2060, 0x682060 }, + { 0x2064, 0x682064 }, + { 0x2068, 0x682068 }, + { 0x206c, 0x68206c }, + { 0x2070, 0x682070 }, + { 0x2074, 0x682074 }, + { 0x2078, 0x682078 }, + { 0x207c, 0x68207c }, + { 0x2080, 0x682080 }, + { 0x2088, 0x682088 }, + { 0x2090, 0x682090 }, + { 0x209c, 0x68209c }, + { 0x20a0, 0x6820a0 }, + { 0x20a4, 0x6820a4 }, + { 0x20a8, 0x6820a8 }, + { 0x20ac, 0x6820ac }, + { 0x2180, 0x682180 }, + { 0x2184, 0x682184 }, + { 0x218c, 0x68218c }, + { 0x2194, 0x682194 }, + { 0x2198, 0x682198 }, + { 0x219c, 0x68219c }, + { 0x21a0, 0x6821a0 }, + { 0x21a4, 0x6821a4 }, + { 0x2214, 0x682214 }, + { 0x2218, 0x682218 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +gv100_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = 0x008000, + .data = { + { "Global", 1, &gv100_disp_core_mthd_base }, + { "SOR", 4, &gv100_disp_core_mthd_sor }, + { "WINDOW", 8, &gv100_disp_core_mthd_wndw }, + { "HEAD", 4, &gv100_disp_core_mthd_head }, + {} + } +}; + +static int +gv100_disp_core_idle(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x610630); + if ((stat & 0x001f0000) == 0x000b0000) + return 0; + ); + return -EBUSY; +} + +static u64 +gv100_disp_core_user(struct nvkm_disp_chan *chan, u64 *psize) +{ + *psize = 0x10000; + return 0x680000; +} + +static void +gv100_disp_core_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00000001; + const u32 data = en ? mask : 0; + nvkm_mask(device, 0x611dac, mask, data); +} + +static void +gv100_disp_core_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + nvkm_mask(device, 0x6104e0, 0x00000010, 0x00000000); + gv100_disp_core_idle(chan); + nvkm_mask(device, 0x6104e0, 0x00000002, 0x00000000); + chan->suspend_put = nvkm_rd32(device, 0x680000); +} + +static int +gv100_disp_core_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + nvkm_wr32(device, 0x610b24, lower_32_bits(chan->push)); + nvkm_wr32(device, 0x610b20, upper_32_bits(chan->push)); + nvkm_wr32(device, 0x610b28, 0x00000001); + nvkm_wr32(device, 0x610b2c, 0x00000040); + + nvkm_mask(device, 0x6104e0, 0x00000010, 0x00000010); + nvkm_wr32(device, 0x680000, chan->suspend_put); + nvkm_wr32(device, 0x6104e0, 0x00000013); + return gv100_disp_core_idle(chan); +} + +static const struct nvkm_disp_chan_func +gv100_disp_core_func = { + .push = nv50_disp_dmac_push, + .init = gv100_disp_core_init, + .fini = gv100_disp_core_fini, + .intr = gv100_disp_core_intr, + .user = gv100_disp_core_user, + .bind = gv100_disp_dmac_bind, +}; + +const struct nvkm_disp_chan_user +gv100_disp_core = { + .func = &gv100_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &gv100_disp_core_mthd, +}; + +#define gv100_disp_caps(p) container_of((p), struct gv100_disp_caps, object) + +struct gv100_disp_caps { + struct nvkm_object object; + struct nvkm_disp *disp; +}; + +static int +gv100_disp_caps_map(struct nvkm_object *object, void *argv, u32 argc, + enum nvkm_object_map *type, u64 *addr, u64 *size) +{ + struct gv100_disp_caps *caps = gv100_disp_caps(object); + struct nvkm_device *device = caps->disp->engine.subdev.device; + *type = NVKM_OBJECT_MAP_IO; + *addr = 0x640000 + device->func->resource_addr(device, 0); + *size = 0x1000; + return 0; +} + +static const struct nvkm_object_func +gv100_disp_caps = { + .map = gv100_disp_caps_map, +}; + +int +gv100_disp_caps_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + struct gv100_disp_caps *caps; + + if (!(caps = kzalloc(sizeof(*caps), GFP_KERNEL))) + return -ENOMEM; + *pobject = &caps->object; + + nvkm_object_ctor(&gv100_disp_caps, oclass, &caps->object); + caps->disp = disp; + return 0; +} + +void +gv100_disp_super(struct work_struct *work) +{ + struct nvkm_disp *disp = container_of(work, struct nvkm_disp, super.work); + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_head *head; + u32 stat, mask[4]; + + mutex_lock(&disp->super.mutex); + stat = nvkm_rd32(device, 0x6107a8); + + nvkm_debug(subdev, "supervisor %d: %08x\n", ffs(disp->super.pending), stat); + list_for_each_entry(head, &disp->heads, head) { + mask[head->id] = nvkm_rd32(device, 0x6107ac + (head->id * 4)); + HEAD_DBG(head, "%08x", mask[head->id]); + } + + if (disp->super.pending & 0x00000001) { + nv50_disp_chan_mthd(disp->chan[0], NV_DBG_DEBUG); + nv50_disp_super_1(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_1_0(disp, head); + } + } else + if (disp->super.pending & 0x00000002) { + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_2_0(disp, head); + } + nvkm_outp_route(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00010000)) + continue; + nv50_disp_super_2_1(disp, head); + } + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_2_2(disp, head); + } + } else + if (disp->super.pending & 0x00000004) { + list_for_each_entry(head, &disp->heads, head) { + if (!(mask[head->id] & 0x00001000)) + continue; + nv50_disp_super_3_0(disp, head); + } + } + + list_for_each_entry(head, &disp->heads, head) + nvkm_wr32(device, 0x6107ac + (head->id * 4), 0x00000000); + + nvkm_wr32(device, 0x6107a8, 0x80000000); + mutex_unlock(&disp->super.mutex); +} + +static void +gv100_disp_exception(struct nvkm_disp *disp, int chid) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x611020 + (chid * 12)); + u32 type = (stat & 0x00007000) >> 12; + u32 mthd = (stat & 0x00000fff) << 2; + const struct nvkm_enum *reason = + nvkm_enum_find(nv50_disp_intr_error_type, type); + + /*TODO: Suspect 33->41 are for WRBK channel exceptions, but we + * don't support those currently. + * + * CORE+WIN CHIDs map directly to the FE_EXCEPT() slots. + */ + if (chid <= 32) { + u32 data = nvkm_rd32(device, 0x611024 + (chid * 12)); + u32 code = nvkm_rd32(device, 0x611028 + (chid * 12)); + nvkm_error(subdev, "chid %d stat %08x reason %d [%s] " + "mthd %04x data %08x code %08x\n", + chid, stat, type, reason ? reason->name : "", + mthd, data, code); + } else { + nvkm_error(subdev, "chid %d stat %08x reason %d [%s] " + "mthd %04x\n", + chid, stat, type, reason ? reason->name : "", mthd); + } + + if (chid < ARRAY_SIZE(disp->chan) && disp->chan[chid]) { + switch (mthd) { + case 0x0200: + nv50_disp_chan_mthd(disp->chan[chid], NV_DBG_ERROR); + break; + default: + break; + } + } + + nvkm_wr32(device, 0x611020 + (chid * 12), 0x90000000); +} + +static void +gv100_disp_intr_ctrl_disp(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x611c30); + + if (stat & 0x00000007) { + disp->super.pending = (stat & 0x00000007); + queue_work(disp->super.wq, &disp->super.work); + nvkm_wr32(device, 0x611860, disp->super.pending); + stat &= ~0x00000007; + } + + /*TODO: I would guess this is VBIOS_RELEASE, however, NFI how to + * ACK it, nor does RM appear to bother. + */ + if (stat & 0x00000008) + stat &= ~0x00000008; + + if (stat & 0x00000080) { + u32 error = nvkm_mask(device, 0x611848, 0x00000000, 0x00000000); + nvkm_warn(subdev, "error %08x\n", error); + stat &= ~0x00000080; + } + + if (stat & 0x00000100) { + unsigned long wndws = nvkm_rd32(device, 0x611858); + unsigned long other = nvkm_rd32(device, 0x61185c); + int wndw; + + nvkm_wr32(device, 0x611858, wndws); + nvkm_wr32(device, 0x61185c, other); + + /* AWAKEN_OTHER_CORE. */ + if (other & 0x00000001) + nv50_disp_chan_uevent_send(disp, 0); + + /* AWAKEN_WIN_CH(n). */ + for_each_set_bit(wndw, &wndws, disp->wndw.nr) { + nv50_disp_chan_uevent_send(disp, 1 + wndw); + } + } + + if (stat) + nvkm_warn(subdev, "ctrl %08x\n", stat); +} + +static void +gv100_disp_intr_exc_other(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x611854); + unsigned long mask; + int head; + + if (stat & 0x00000001) { + nvkm_wr32(device, 0x611854, 0x00000001); + gv100_disp_exception(disp, 0); + stat &= ~0x00000001; + } + + if ((mask = (stat & 0x00ff0000) >> 16)) { + for_each_set_bit(head, &mask, disp->wndw.nr) { + nvkm_wr32(device, 0x611854, 0x00010000 << head); + gv100_disp_exception(disp, 73 + head); + stat &= ~(0x00010000 << head); + } + } + + if (stat) { + nvkm_warn(subdev, "exception %08x\n", stat); + nvkm_wr32(device, 0x611854, stat); + } +} + +static void +gv100_disp_intr_exc_winim(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + unsigned long stat = nvkm_rd32(device, 0x611850); + int wndw; + + for_each_set_bit(wndw, &stat, disp->wndw.nr) { + nvkm_wr32(device, 0x611850, BIT(wndw)); + gv100_disp_exception(disp, 33 + wndw); + stat &= ~BIT(wndw); + } + + if (stat) { + nvkm_warn(subdev, "wimm %08x\n", (u32)stat); + nvkm_wr32(device, 0x611850, stat); + } +} + +static void +gv100_disp_intr_exc_win(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + unsigned long stat = nvkm_rd32(device, 0x61184c); + int wndw; + + for_each_set_bit(wndw, &stat, disp->wndw.nr) { + nvkm_wr32(device, 0x61184c, BIT(wndw)); + gv100_disp_exception(disp, 1 + wndw); + stat &= ~BIT(wndw); + } + + if (stat) { + nvkm_warn(subdev, "wndw %08x\n", (u32)stat); + nvkm_wr32(device, 0x61184c, stat); + } +} + +static void +gv100_disp_intr_head_timing(struct nvkm_disp *disp, int head) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x611800 + (head * 0x04)); + + /* LAST_DATA, LOADV. */ + if (stat & 0x00000003) { + nvkm_wr32(device, 0x611800 + (head * 0x04), stat & 0x00000003); + stat &= ~0x00000003; + } + + if (stat & 0x00000004) { + nvkm_disp_vblank(disp, head); + nvkm_wr32(device, 0x611800 + (head * 0x04), 0x00000004); + stat &= ~0x00000004; + } + + if (stat) { + nvkm_warn(subdev, "head %08x\n", stat); + nvkm_wr32(device, 0x611800 + (head * 0x04), stat); + } +} + +void +gv100_disp_intr(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x611ec0); + unsigned long mask; + int head; + + if ((mask = (stat & 0x000000ff))) { + for_each_set_bit(head, &mask, 8) { + gv100_disp_intr_head_timing(disp, head); + stat &= ~BIT(head); + } + } + + if (stat & 0x00000200) { + gv100_disp_intr_exc_win(disp); + stat &= ~0x00000200; + } + + if (stat & 0x00000400) { + gv100_disp_intr_exc_winim(disp); + stat &= ~0x00000400; + } + + if (stat & 0x00000800) { + gv100_disp_intr_exc_other(disp); + stat &= ~0x00000800; + } + + if (stat & 0x00001000) { + gv100_disp_intr_ctrl_disp(disp); + stat &= ~0x00001000; + } + + if (stat) + nvkm_warn(subdev, "intr %08x\n", stat); +} + +void +gv100_disp_fini(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + nvkm_wr32(device, 0x611db0, 0x00000000); +} + +static int +gv100_disp_init(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_head *head; + int i, j; + u32 tmp; + + /* Claim ownership of display. */ + if (nvkm_rd32(device, 0x6254e8) & 0x00000002) { + nvkm_mask(device, 0x6254e8, 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x6254e8) & 0x00000002)) + break; + ) < 0) + return -EBUSY; + } + + /* Lock pin capabilities. */ + tmp = nvkm_rd32(device, 0x610068); + nvkm_wr32(device, 0x640008, tmp); + + /* SOR capabilities. */ + for (i = 0; i < disp->sor.nr; i++) { + tmp = nvkm_rd32(device, 0x61c000 + (i * 0x800)); + nvkm_mask(device, 0x640000, 0x00000100 << i, 0x00000100 << i); + nvkm_wr32(device, 0x640144 + (i * 0x08), tmp); + } + + /* Head capabilities. */ + list_for_each_entry(head, &disp->heads, head) { + const int id = head->id; + + /* RG. */ + tmp = nvkm_rd32(device, 0x616300 + (id * 0x800)); + nvkm_wr32(device, 0x640048 + (id * 0x020), tmp); + + /* POSTCOMP. */ + for (j = 0; j < 6 * 4; j += 4) { + tmp = nvkm_rd32(device, 0x616100 + (id * 0x800) + j); + nvkm_wr32(device, 0x640030 + (id * 0x20) + j, tmp); + } + } + + /* Window capabilities. */ + for (i = 0; i < disp->wndw.nr; i++) { + nvkm_mask(device, 0x640004, 1 << i, 1 << i); + for (j = 0; j < 6 * 4; j += 4) { + tmp = nvkm_rd32(device, 0x630050 + (i * 0x800) + j); + nvkm_wr32(device, 0x6401e4 + (i * 0x20) + j, tmp); + } + } + + /* IHUB capabilities. */ + for (i = 0; i < 4; i++) { + tmp = nvkm_rd32(device, 0x62e000 + (i * 0x04)); + nvkm_wr32(device, 0x640010 + (i * 0x04), tmp); + } + + nvkm_mask(device, 0x610078, 0x00000001, 0x00000001); + + /* Setup instance memory. */ + switch (nvkm_memory_target(disp->inst->memory)) { + case NVKM_MEM_TARGET_VRAM: tmp = 0x00000001; break; + case NVKM_MEM_TARGET_NCOH: tmp = 0x00000002; break; + case NVKM_MEM_TARGET_HOST: tmp = 0x00000003; break; + default: + break; + } + nvkm_wr32(device, 0x610010, 0x00000008 | tmp); + nvkm_wr32(device, 0x610014, disp->inst->addr >> 16); + + /* CTRL_DISP: AWAKEN, ERROR, SUPERVISOR[1-3]. */ + nvkm_wr32(device, 0x611cf0, 0x00000187); /* MSK. */ + nvkm_wr32(device, 0x611db0, 0x00000187); /* EN. */ + + /* EXC_OTHER: CURSn, CORE. */ + nvkm_wr32(device, 0x611cec, disp->head.mask << 16 | + 0x00000001); /* MSK. */ + nvkm_wr32(device, 0x611dac, 0x00000000); /* EN. */ + + /* EXC_WINIM. */ + nvkm_wr32(device, 0x611ce8, disp->wndw.mask); /* MSK. */ + nvkm_wr32(device, 0x611da8, 0x00000000); /* EN. */ + + /* EXC_WIN. */ + nvkm_wr32(device, 0x611ce4, disp->wndw.mask); /* MSK. */ + nvkm_wr32(device, 0x611da4, 0x00000000); /* EN. */ + + /* HEAD_TIMING(n): VBLANK. */ + list_for_each_entry(head, &disp->heads, head) { + const u32 hoff = head->id * 4; + nvkm_wr32(device, 0x611cc0 + hoff, 0x00000004); /* MSK. */ + nvkm_wr32(device, 0x611d80 + hoff, 0x00000000); /* EN. */ + } + + /* OR. */ + nvkm_wr32(device, 0x611cf4, 0x00000000); /* MSK. */ + nvkm_wr32(device, 0x611db4, 0x00000000); /* EN. */ + return 0; +} + +static const struct nvkm_disp_func +gv100_disp = { + .oneinit = nv50_disp_oneinit, + .init = gv100_disp_init, + .fini = gv100_disp_fini, + .intr = gv100_disp_intr, + .super = gv100_disp_super, + .uevent = &gv100_disp_chan_uevent, + .wndw = { .cnt = gv100_disp_wndw_cnt }, + .head = { .cnt = gv100_head_cnt, .new = gv100_head_new }, + .sor = { .cnt = gv100_sor_cnt, .new = gv100_sor_new }, + .ramht_size = 0x2000, + .root = { 0, 0,GV100_DISP }, + .user = { + {{-1,-1,GV100_DISP_CAPS }, gv100_disp_caps_new }, + {{ 0, 0,GV100_DISP_CURSOR }, nvkm_disp_chan_new, &gv100_disp_curs }, + {{ 0, 0,GV100_DISP_WINDOW_IMM_CHANNEL_DMA}, nvkm_disp_wndw_new, &gv100_disp_wimm }, + {{ 0, 0,GV100_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gv100_disp_core }, + {{ 0, 0,GV100_DISP_WINDOW_CHANNEL_DMA }, nvkm_disp_wndw_new, &gv100_disp_wndw }, + {} + }, +}; + +int +gv100_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&gv100_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.c new file mode 100644 index 0000000000..1ccfc83148 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.c @@ -0,0 +1,84 @@ +// SPDX-License-Identifier: MIT +#include "hdmi.h" + +void pack_hdmi_infoframe(struct packed_hdmi_infoframe *packed_frame, + u8 *raw_frame, ssize_t len) +{ + u32 header = 0; + u32 subpack0_low = 0; + u32 subpack0_high = 0; + u32 subpack1_low = 0; + u32 subpack1_high = 0; + + switch (len) { + /* + * "When in doubt, use brute force." + * -- Ken Thompson. + */ + default: + /* + * We presume that no valid frame is longer than 17 + * octets, including header... And truncate to that + * if it's longer. + */ + case 17: + subpack1_high = (raw_frame[16] << 16); + fallthrough; + case 16: + subpack1_high |= (raw_frame[15] << 8); + fallthrough; + case 15: + subpack1_high |= raw_frame[14]; + fallthrough; + case 14: + subpack1_low = (raw_frame[13] << 24); + fallthrough; + case 13: + subpack1_low |= (raw_frame[12] << 16); + fallthrough; + case 12: + subpack1_low |= (raw_frame[11] << 8); + fallthrough; + case 11: + subpack1_low |= raw_frame[10]; + fallthrough; + case 10: + subpack0_high = (raw_frame[9] << 16); + fallthrough; + case 9: + subpack0_high |= (raw_frame[8] << 8); + fallthrough; + case 8: + subpack0_high |= raw_frame[7]; + fallthrough; + case 7: + subpack0_low = (raw_frame[6] << 24); + fallthrough; + case 6: + subpack0_low |= (raw_frame[5] << 16); + fallthrough; + case 5: + subpack0_low |= (raw_frame[4] << 8); + fallthrough; + case 4: + subpack0_low |= raw_frame[3]; + fallthrough; + case 3: + header = (raw_frame[2] << 16); + fallthrough; + case 2: + header |= (raw_frame[1] << 8); + fallthrough; + case 1: + header |= raw_frame[0]; + fallthrough; + case 0: + break; + } + + packed_frame->header = header; + packed_frame->subpack0_low = subpack0_low; + packed_frame->subpack0_high = subpack0_high; + packed_frame->subpack1_low = subpack1_low; + packed_frame->subpack1_high = subpack1_high; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.h new file mode 100644 index 0000000000..fb1c3e3c5d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/hdmi.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_HDMI_H__ +#define __NVKM_DISP_HDMI_H__ +#include "ior.h" + +struct packed_hdmi_infoframe { + u32 header; + u32 subpack0_low; + u32 subpack0_high; + u32 subpack1_low; + u32 subpack1_high; +}; + +void pack_hdmi_infoframe(struct packed_hdmi_infoframe *packed_frame, + u8 *raw_frame, ssize_t len); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.c new file mode 100644 index 0000000000..7f5d13d13c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.c @@ -0,0 +1,67 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "head.h" + +#include + +#include +#include + +struct nvkm_head * +nvkm_head_find(struct nvkm_disp *disp, int id) +{ + struct nvkm_head *head; + list_for_each_entry(head, &disp->heads, head) { + if (head->id == id) + return head; + } + return NULL; +} + +void +nvkm_head_del(struct nvkm_head **phead) +{ + struct nvkm_head *head = *phead; + if (head) { + HEAD_DBG(head, "dtor"); + list_del(&head->head); + kfree(*phead); + *phead = NULL; + } +} + +int +nvkm_head_new_(const struct nvkm_head_func *func, + struct nvkm_disp *disp, int id) +{ + struct nvkm_head *head; + if (!(head = kzalloc(sizeof(*head), GFP_KERNEL))) + return -ENOMEM; + head->func = func; + head->disp = disp; + head->id = id; + list_add_tail(&head->head, &disp->heads); + HEAD_DBG(head, "ctor"); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.h new file mode 100644 index 0000000000..856252bf55 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/head.h @@ -0,0 +1,63 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_HEAD_H__ +#define __NVKM_DISP_HEAD_H__ +#include +#include "priv.h" + +struct nvkm_head { + const struct nvkm_head_func *func; + struct nvkm_disp *disp; + int id; + + struct list_head head; + + struct nvkm_head_state { + u16 htotal; + u16 hsynce; + u16 hblanke; + u16 hblanks; + u16 vtotal; + u16 vsynce; + u16 vblanke; + u16 vblanks; + u32 hz; + + /* Prior to GF119, these are set by the OR. */ + struct { + u8 depth; + } or; + } arm, asy; + + struct nvkm_object object; +}; + +int nvkm_head_new_(const struct nvkm_head_func *, struct nvkm_disp *, int id); +void nvkm_head_del(struct nvkm_head **); +struct nvkm_head *nvkm_head_find(struct nvkm_disp *, int id); + +struct nvkm_head_func { + void (*state)(struct nvkm_head *, struct nvkm_head_state *); + void (*rgpos)(struct nvkm_head *, u16 *hline, u16 *vline); + void (*rgclk)(struct nvkm_head *, int div); + void (*vblank_get)(struct nvkm_head *); + void (*vblank_put)(struct nvkm_head *); +}; + +int nv50_head_cnt(struct nvkm_disp *, unsigned long *); +int nv50_head_new(struct nvkm_disp *, int id); +void nv50_head_rgpos(struct nvkm_head *, u16 *, u16 *); + +int gf119_head_cnt(struct nvkm_disp *, unsigned long *); +int gf119_head_new(struct nvkm_disp *, int id); +void gf119_head_rgclk(struct nvkm_head *, int); + +int gv100_head_cnt(struct nvkm_disp *, unsigned long *); +int gv100_head_new(struct nvkm_disp *, int id); + +#define HEAD_MSG(h,l,f,a...) do { \ + struct nvkm_head *_h = (h); \ + nvkm_##l(&_h->disp->engine.subdev, "head-%d: "f"\n", _h->id, ##a); \ +} while(0) +#define HEAD_WARN(h,f,a...) HEAD_MSG((h), warn, f, ##a) +#define HEAD_DBG(h,f,a...) HEAD_MSG((h), debug, f, ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.c new file mode 100644 index 0000000000..e420bf2e43 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.c @@ -0,0 +1,72 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ior.h" + +static const char * +nvkm_ior_name[] = { + [DAC] = "DAC", + [SOR] = "SOR", + [PIOR] = "PIOR", +}; + +struct nvkm_ior * +nvkm_ior_find(struct nvkm_disp *disp, enum nvkm_ior_type type, int id) +{ + struct nvkm_ior *ior; + list_for_each_entry(ior, &disp->iors, head) { + if (ior->type == type && (id < 0 || ior->id == id)) + return ior; + } + return NULL; +} + +void +nvkm_ior_del(struct nvkm_ior **pior) +{ + struct nvkm_ior *ior = *pior; + if (ior) { + IOR_DBG(ior, "dtor"); + list_del(&ior->head); + kfree(*pior); + *pior = NULL; + } +} + +int +nvkm_ior_new_(const struct nvkm_ior_func *func, struct nvkm_disp *disp, + enum nvkm_ior_type type, int id, bool hda) +{ + struct nvkm_ior *ior; + if (!(ior = kzalloc(sizeof(*ior), GFP_KERNEL))) + return -ENOMEM; + ior->func = func; + ior->disp = disp; + ior->type = type; + ior->id = id; + ior->hda = hda; + snprintf(ior->name, sizeof(ior->name), "%s-%d", nvkm_ior_name[ior->type], ior->id); + list_add_tail(&ior->head, &disp->iors); + IOR_DBG(ior, "ctor"); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h new file mode 100644 index 0000000000..da1b1a626e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/ior.h @@ -0,0 +1,196 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_IOR_H__ +#define __NVKM_DISP_IOR_H__ +#include "priv.h" +struct nvkm_i2c_aux; + +struct nvkm_ior { + const struct nvkm_ior_func *func; + struct nvkm_disp *disp; + enum nvkm_ior_type { + DAC, + SOR, + PIOR, + } type; + int id; + bool hda; + char name[8]; + + struct list_head head; + bool identity; + + struct nvkm_ior_state { + struct nvkm_outp *outp; + unsigned rgdiv; + unsigned proto_evo:4; + enum nvkm_ior_proto { + CRT, + TV, + TMDS, + LVDS, + DP, + UNKNOWN + } proto:3; + unsigned link:2; + unsigned head:8; + } arm, asy; + + /* Armed DP state. */ + struct { + bool mst; + bool ef; + u8 nr; + u8 bw; + } dp; + + /* Armed TMDS state. */ + struct { + bool high_speed; + } tmds; +}; + +struct nvkm_ior_func { + struct { + int (*get)(struct nvkm_outp *, int *link); + void (*set)(struct nvkm_outp *, struct nvkm_ior *); + } route; + + void (*state)(struct nvkm_ior *, struct nvkm_ior_state *); + void (*power)(struct nvkm_ior *, bool normal, bool pu, + bool data, bool vsync, bool hsync); + int (*sense)(struct nvkm_ior *, u32 loadval); + void (*clock)(struct nvkm_ior *); + void (*war_2)(struct nvkm_ior *); + void (*war_3)(struct nvkm_ior *); + + const struct nvkm_ior_func_hdmi { + void (*ctrl)(struct nvkm_ior *, int head, bool enable, u8 max_ac_packet, u8 rekey); + void (*scdc)(struct nvkm_ior *, u8 scdc); + void (*infoframe_avi)(struct nvkm_ior *, int head, void *data, u32 size); + void (*infoframe_vsi)(struct nvkm_ior *, int head, void *data, u32 size); + } *hdmi; + + const struct nvkm_ior_func_dp { + u8 lanes[4]; + int (*links)(struct nvkm_ior *, struct nvkm_i2c_aux *); + void (*power)(struct nvkm_ior *, int nr); + void (*pattern)(struct nvkm_ior *, int pattern); + void (*drive)(struct nvkm_ior *, int ln, int pc, + int dc, int pe, int tx_pu); + void (*vcpi)(struct nvkm_ior *, int head, u8 slot, + u8 slot_nr, u16 pbn, u16 aligned); + void (*audio)(struct nvkm_ior *, int head, bool enable); + void (*audio_sym)(struct nvkm_ior *, int head, u16 h, u32 v); + void (*activesym)(struct nvkm_ior *, int head, + u8 TU, u8 VTUa, u8 VTUf, u8 VTUi); + void (*watermark)(struct nvkm_ior *, int head, u8 watermark); + } *dp; + + const struct nvkm_ior_func_hda { + void (*hpd)(struct nvkm_ior *, int head, bool present); + void (*eld)(struct nvkm_ior *, int head, u8 *data, u8 size); + void (*device_entry)(struct nvkm_ior *, int head); + } *hda; +}; + +int nvkm_ior_new_(const struct nvkm_ior_func *func, struct nvkm_disp *, + enum nvkm_ior_type type, int id, bool hda); +void nvkm_ior_del(struct nvkm_ior **); +struct nvkm_ior *nvkm_ior_find(struct nvkm_disp *, enum nvkm_ior_type, int id); + +static inline u32 +nv50_ior_base(struct nvkm_ior *ior) +{ + return ior->id * 0x800; +} + +int nv50_dac_cnt(struct nvkm_disp *, unsigned long *); +int nv50_dac_new(struct nvkm_disp *, int); +void nv50_dac_power(struct nvkm_ior *, bool, bool, bool, bool, bool); +int nv50_dac_sense(struct nvkm_ior *, u32); + +int gf119_dac_cnt(struct nvkm_disp *, unsigned long *); +int gf119_dac_new(struct nvkm_disp *, int); + +static inline u32 +nv50_sor_link(struct nvkm_ior *ior) +{ + return nv50_ior_base(ior) + ((ior->asy.link == 2) * 0x80); +} + +int nv50_sor_cnt(struct nvkm_disp *, unsigned long *); +void nv50_sor_state(struct nvkm_ior *, struct nvkm_ior_state *); +void nv50_sor_power(struct nvkm_ior *, bool, bool, bool, bool, bool); +void nv50_sor_clock(struct nvkm_ior *); + +int g84_sor_new(struct nvkm_disp *, int); +extern const struct nvkm_ior_func_hdmi g84_sor_hdmi; + +int g94_sor_cnt(struct nvkm_disp *, unsigned long *); + +void g94_sor_state(struct nvkm_ior *, struct nvkm_ior_state *); +extern const struct nvkm_ior_func_dp g94_sor_dp; +int g94_sor_dp_links(struct nvkm_ior *, struct nvkm_i2c_aux *); +void g94_sor_dp_power(struct nvkm_ior *, int); +void g94_sor_dp_pattern(struct nvkm_ior *, int); +void g94_sor_dp_drive(struct nvkm_ior *, int, int, int, int, int); +void g94_sor_dp_audio_sym(struct nvkm_ior *, int, u16, u32); +void g94_sor_dp_activesym(struct nvkm_ior *, int, u8, u8, u8, u8); +void g94_sor_dp_watermark(struct nvkm_ior *, int, u8); + +extern const struct nvkm_ior_func_hdmi gt215_sor_hdmi; +void gt215_sor_dp_audio(struct nvkm_ior *, int, bool); +extern const struct nvkm_ior_func_hda gt215_sor_hda; + +int gf119_sor_cnt(struct nvkm_disp *, unsigned long *); +void gf119_sor_state(struct nvkm_ior *, struct nvkm_ior_state *); +void gf119_sor_clock(struct nvkm_ior *); +extern const struct nvkm_ior_func_dp gf119_sor_dp; +int gf119_sor_dp_links(struct nvkm_ior *, struct nvkm_i2c_aux *); +void gf119_sor_dp_drive(struct nvkm_ior *, int, int, int, int, int); +void gf119_sor_dp_vcpi(struct nvkm_ior *, int, u8, u8, u16, u16); +void gf119_sor_dp_audio(struct nvkm_ior *, int, bool); +void gf119_sor_dp_audio_sym(struct nvkm_ior *, int, u16, u32); +void gf119_sor_dp_watermark(struct nvkm_ior *, int, u8); +extern const struct nvkm_ior_func_hda gf119_sor_hda; +void gf119_sor_hda_hpd(struct nvkm_ior *, int, bool); +void gf119_sor_hda_eld(struct nvkm_ior *, int, u8 *, u8); + +int gk104_sor_new(struct nvkm_disp *, int); +extern const struct nvkm_ior_func_hdmi gk104_sor_hdmi; +void gk104_sor_hdmi_ctrl(struct nvkm_ior *, int, bool, u8, u8); +void gk104_sor_hdmi_infoframe_avi(struct nvkm_ior *, int, void *, u32); +void gk104_sor_hdmi_infoframe_vsi(struct nvkm_ior *, int, void *, u32); + +void gm107_sor_dp_pattern(struct nvkm_ior *, int); + +void gm200_sor_route_set(struct nvkm_outp *, struct nvkm_ior *); +int gm200_sor_route_get(struct nvkm_outp *, int *); +extern const struct nvkm_ior_func_hdmi gm200_sor_hdmi; +void gm200_sor_hdmi_scdc(struct nvkm_ior *, u8); +extern const struct nvkm_ior_func_dp gm200_sor_dp; +void gm200_sor_dp_drive(struct nvkm_ior *, int, int, int, int, int); + +int gp100_sor_new(struct nvkm_disp *, int); + +int gv100_sor_cnt(struct nvkm_disp *, unsigned long *); +void gv100_sor_state(struct nvkm_ior *, struct nvkm_ior_state *); +extern const struct nvkm_ior_func_hdmi gv100_sor_hdmi; +void gv100_sor_dp_audio(struct nvkm_ior *, int, bool); +void gv100_sor_dp_audio_sym(struct nvkm_ior *, int, u16, u32); +void gv100_sor_dp_watermark(struct nvkm_ior *, int, u8); +extern const struct nvkm_ior_func_hda gv100_sor_hda; + +void tu102_sor_dp_vcpi(struct nvkm_ior *, int, u8, u8, u16, u16); + +int nv50_pior_cnt(struct nvkm_disp *, unsigned long *); +int nv50_pior_new(struct nvkm_disp *, int); +void nv50_pior_depth(struct nvkm_ior *, struct nvkm_ior_state *, u32 ctrl); + +#define IOR_MSG(i,l,f,a...) do { \ + struct nvkm_ior *_ior = (i); \ + nvkm_##l(&_ior->disp->engine.subdev, "%s: "f"\n", _ior->name, ##a); \ +} while(0) +#define IOR_WARN(i,f,a...) IOR_MSG((i), warn, f, ##a) +#define IOR_DBG(i,f,a...) IOR_MSG((i), debug, f, ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c new file mode 100644 index 0000000000..841e3b69fc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp77.c @@ -0,0 +1,72 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +static const struct nvkm_ior_func +mcp77_sor = { + .state = g94_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, + .hdmi = &g84_sor_hdmi, + .dp = &g94_sor_dp, +}; + +static int +mcp77_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&mcp77_sor, disp, SOR, id, false); +} + +static const struct nvkm_disp_func +mcp77_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = g94_sor_cnt, .new = mcp77_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,GT206_DISP }, + .user = { + {{0,0, G82_DISP_CURSOR }, nvkm_disp_chan_new, & nv50_disp_curs }, + {{0,0, G82_DISP_OVERLAY }, nvkm_disp_chan_new, & nv50_disp_oimm }, + {{0,0,GT200_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,GT206_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g94_disp_core }, + {{0,0,GT200_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, >200_disp_ovly }, + {} + }, +}; + +int +mcp77_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&mcp77_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c new file mode 100644 index 0000000000..f96ba47526 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/mcp89.c @@ -0,0 +1,86 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" + +#include + +static const struct nvkm_ior_func_dp +mcp89_sor_dp = { + .lanes = { 3, 2, 1, 0 }, + .links = g94_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = g94_sor_dp_pattern, + .drive = g94_sor_dp_drive, + .audio = gt215_sor_dp_audio, + .audio_sym = g94_sor_dp_audio_sym, + .activesym = g94_sor_dp_activesym, + .watermark = g94_sor_dp_watermark, +}; + +static const struct nvkm_ior_func +mcp89_sor = { + .state = g94_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, + .hdmi = >215_sor_hdmi, + .dp = &mcp89_sor_dp, + .hda = >215_sor_hda, +}; + +static int +mcp89_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&mcp89_sor, disp, SOR, id, true); +} + +static const struct nvkm_disp_func +mcp89_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = g94_sor_cnt, .new = mcp89_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0,0,GT214_DISP }, + .user = { + {{0,0,GT214_DISP_CURSOR }, nvkm_disp_chan_new, &nv50_disp_curs }, + {{0,0,GT214_DISP_OVERLAY }, nvkm_disp_chan_new, &nv50_disp_oimm }, + {{0,0,GT214_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, & g84_disp_base }, + {{0,0,GT214_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, & g94_disp_core }, + {{0,0,GT214_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, & g84_disp_ovly }, + {} + }, +}; + +int +mcp89_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&mcp89_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv04.c new file mode 100644 index 0000000000..e4cf11a339 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv04.c @@ -0,0 +1,130 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "head.h" + +#include + +static void +nv04_head_vblank_put(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + nvkm_wr32(device, 0x600140 + (head->id * 0x2000) , 0x00000000); +} + +static void +nv04_head_vblank_get(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + nvkm_wr32(device, 0x600140 + (head->id * 0x2000) , 0x00000001); +} + +static void +nv04_head_rgpos(struct nvkm_head *head, u16 *hline, u16 *vline) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + u32 data = nvkm_rd32(device, 0x600868 + (head->id * 0x2000)); + *hline = (data & 0xffff0000) >> 16; + *vline = (data & 0x0000ffff); +} + +static void +nv04_head_state(struct nvkm_head *head, struct nvkm_head_state *state) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x0200; + state->vblanks = nvkm_rd32(device, 0x680800 + hoff) & 0x0000ffff; + state->vtotal = nvkm_rd32(device, 0x680804 + hoff) & 0x0000ffff; + state->vblanke = state->vtotal - 1; + state->hblanks = nvkm_rd32(device, 0x680820 + hoff) & 0x0000ffff; + state->htotal = nvkm_rd32(device, 0x680824 + hoff) & 0x0000ffff; + state->hblanke = state->htotal - 1; +} + +static const struct nvkm_head_func +nv04_head = { + .state = nv04_head_state, + .rgpos = nv04_head_rgpos, + .vblank_get = nv04_head_vblank_get, + .vblank_put = nv04_head_vblank_put, +}; + +static int +nv04_head_new(struct nvkm_disp *disp, int id) +{ + return nvkm_head_new_(&nv04_head, disp, id); +} + +static void +nv04_disp_intr(struct nvkm_disp *disp) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 crtc0 = nvkm_rd32(device, 0x600100); + u32 crtc1 = nvkm_rd32(device, 0x602100); + u32 pvideo; + + if (crtc0 & 0x00000001) { + nvkm_disp_vblank(disp, 0); + nvkm_wr32(device, 0x600100, 0x00000001); + } + + if (crtc1 & 0x00000001) { + nvkm_disp_vblank(disp, 1); + nvkm_wr32(device, 0x602100, 0x00000001); + } + + if (device->chipset >= 0x10 && device->chipset <= 0x40) { + pvideo = nvkm_rd32(device, 0x8100); + if (pvideo & ~0x11) + nvkm_info(subdev, "PVIDEO intr: %08x\n", pvideo); + nvkm_wr32(device, 0x8100, pvideo); + } +} + +static const struct nvkm_disp_func +nv04_disp = { + .intr = nv04_disp_intr, + .root = { 0, 0, NV04_DISP }, + .user = { {} }, +}; + +int +nv04_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + int ret, i; + + ret = nvkm_disp_new_(&nv04_disp, device, type, inst, pdisp); + if (ret) + return ret; + + for (i = 0; i < 2; i++) { + ret = nv04_head_new(*pdisp, i); + if (ret) + return ret; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c new file mode 100644 index 0000000000..be81168029 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/nv50.c @@ -0,0 +1,1645 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "head.h" +#include "ior.h" +#include "outp.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +static void +nv50_pior_clock(struct nvkm_ior *pior) +{ + struct nvkm_device *device = pior->disp->engine.subdev.device; + const u32 poff = nv50_ior_base(pior); + + nvkm_mask(device, 0x614380 + poff, 0x00000707, 0x00000001); +} + +static int +nv50_pior_dp_links(struct nvkm_ior *pior, struct nvkm_i2c_aux *aux) +{ + int ret = nvkm_i2c_aux_lnk_ctl(aux, pior->dp.nr, pior->dp.bw, pior->dp.ef); + if (ret) + return ret; + + return 1; +} + +static const struct nvkm_ior_func_dp +nv50_pior_dp = { + .links = nv50_pior_dp_links, +}; + +static void +nv50_pior_power_wait(struct nvkm_device *device, u32 poff) +{ + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61e004 + poff) & 0x80000000)) + break; + ); +} + +static void +nv50_pior_power(struct nvkm_ior *pior, bool normal, bool pu, bool data, bool vsync, bool hsync) +{ + struct nvkm_device *device = pior->disp->engine.subdev.device; + const u32 poff = nv50_ior_base(pior); + const u32 shift = normal ? 0 : 16; + const u32 state = 0x80000000 | (0x00000001 * !!pu) << shift; + const u32 field = 0x80000000 | (0x00000101 << shift); + + nv50_pior_power_wait(device, poff); + nvkm_mask(device, 0x61e004 + poff, field, state); + nv50_pior_power_wait(device, poff); +} + +void +nv50_pior_depth(struct nvkm_ior *ior, struct nvkm_ior_state *state, u32 ctrl) +{ + /* GF119 moves this information to per-head methods, which is + * a lot more convenient, and where our shared code expect it. + */ + if (state->head && state == &ior->asy) { + struct nvkm_head *head = nvkm_head_find(ior->disp, __ffs(state->head)); + + if (!WARN_ON(!head)) { + struct nvkm_head_state *state = &head->asy; + switch ((ctrl & 0x000f0000) >> 16) { + case 6: state->or.depth = 30; break; + case 5: state->or.depth = 24; break; + case 2: state->or.depth = 18; break; + case 0: state->or.depth = 18; break; /*XXX*/ + default: + state->or.depth = 18; + WARN_ON(1); + break; + } + } + } +} + +static void +nv50_pior_state(struct nvkm_ior *pior, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = pior->disp->engine.subdev.device; + const u32 coff = pior->id * 8 + (state == &pior->arm) * 4; + u32 ctrl = nvkm_rd32(device, 0x610b80 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + state->rgdiv = 1; + switch (state->proto_evo) { + case 0: state->proto = TMDS; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x00000003; + nv50_pior_depth(pior, state, ctrl); +} + +static const struct nvkm_ior_func +nv50_pior = { + .state = nv50_pior_state, + .power = nv50_pior_power, + .clock = nv50_pior_clock, + .dp = &nv50_pior_dp, +}; + +int +nv50_pior_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&nv50_pior, disp, PIOR, id, false); +} + +int +nv50_pior_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = (nvkm_rd32(device, 0x610184) & 0x70000000) >> 28; + return 3; +} + +void +nv50_sor_clock(struct nvkm_ior *sor) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const int div = sor->asy.link == 3; + const u32 soff = nv50_ior_base(sor); + + nvkm_mask(device, 0x614300 + soff, 0x00000707, (div << 8) | div); +} + +static void +nv50_sor_power_wait(struct nvkm_device *device, u32 soff) +{ + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c004 + soff) & 0x80000000)) + break; + ); +} + +void +nv50_sor_power(struct nvkm_ior *sor, bool normal, bool pu, bool data, bool vsync, bool hsync) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 shift = normal ? 0 : 16; + const u32 state = 0x80000000 | (0x00000001 * !!pu) << shift; + const u32 field = 0x80000000 | (0x00000001 << shift); + + nv50_sor_power_wait(device, soff); + nvkm_mask(device, 0x61c004 + soff, field, state); + nv50_sor_power_wait(device, soff); + + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61c030 + soff) & 0x10000000)) + break; + ); +} + +void +nv50_sor_state(struct nvkm_ior *sor, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 coff = sor->id * 8 + (state == &sor->arm) * 4; + u32 ctrl = nvkm_rd32(device, 0x610b70 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = LVDS; state->link = 1; break; + case 1: state->proto = TMDS; state->link = 1; break; + case 2: state->proto = TMDS; state->link = 2; break; + case 5: state->proto = TMDS; state->link = 3; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x00000003; +} + +static const struct nvkm_ior_func +nv50_sor = { + .state = nv50_sor_state, + .power = nv50_sor_power, + .clock = nv50_sor_clock, +}; + +static int +nv50_sor_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&nv50_sor, disp, SOR, id, false); +} + +int +nv50_sor_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = (nvkm_rd32(device, 0x610184) & 0x03000000) >> 24; + return 2; +} + +static void +nv50_dac_clock(struct nvkm_ior *dac) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 doff = nv50_ior_base(dac); + + nvkm_mask(device, 0x614280 + doff, 0x07070707, 0x00000000); +} + +int +nv50_dac_sense(struct nvkm_ior *dac, u32 loadval) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 doff = nv50_ior_base(dac); + + dac->func->power(dac, false, true, false, false, false); + + nvkm_wr32(device, 0x61a00c + doff, 0x00100000 | loadval); + mdelay(9); + udelay(500); + loadval = nvkm_mask(device, 0x61a00c + doff, 0xffffffff, 0x00000000); + + dac->func->power(dac, false, false, false, false, false); + if (!(loadval & 0x80000000)) + return -ETIMEDOUT; + + return (loadval & 0x38000000) >> 27; +} + +static void +nv50_dac_power_wait(struct nvkm_device *device, const u32 doff) +{ + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x61a004 + doff) & 0x80000000)) + break; + ); +} + +void +nv50_dac_power(struct nvkm_ior *dac, bool normal, bool pu, bool data, bool vsync, bool hsync) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 doff = nv50_ior_base(dac); + const u32 shift = normal ? 0 : 16; + const u32 state = 0x80000000 | (0x00000040 * ! pu | + 0x00000010 * ! data | + 0x00000004 * ! vsync | + 0x00000001 * ! hsync) << shift; + const u32 field = 0xc0000000 | (0x00000055 << shift); + + nv50_dac_power_wait(device, doff); + nvkm_mask(device, 0x61a004 + doff, field, state); + nv50_dac_power_wait(device, doff); +} + +static void +nv50_dac_state(struct nvkm_ior *dac, struct nvkm_ior_state *state) +{ + struct nvkm_device *device = dac->disp->engine.subdev.device; + const u32 coff = dac->id * 8 + (state == &dac->arm) * 4; + u32 ctrl = nvkm_rd32(device, 0x610b58 + coff); + + state->proto_evo = (ctrl & 0x00000f00) >> 8; + switch (state->proto_evo) { + case 0: state->proto = CRT; break; + default: + state->proto = UNKNOWN; + break; + } + + state->head = ctrl & 0x00000003; +} + +static const struct nvkm_ior_func +nv50_dac = { + .state = nv50_dac_state, + .power = nv50_dac_power, + .sense = nv50_dac_sense, + .clock = nv50_dac_clock, +}; + +int +nv50_dac_new(struct nvkm_disp *disp, int id) +{ + return nvkm_ior_new_(&nv50_dac, disp, DAC, id, false); +} + +int +nv50_dac_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + struct nvkm_device *device = disp->engine.subdev.device; + + *pmask = (nvkm_rd32(device, 0x610184) & 0x00700000) >> 20; + return 3; +} + +static void +nv50_head_vblank_put(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + + nvkm_mask(device, 0x61002c, (4 << head->id), 0); +} + +static void +nv50_head_vblank_get(struct nvkm_head *head) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + + nvkm_mask(device, 0x61002c, (4 << head->id), (4 << head->id)); +} + +static void +nv50_head_rgclk(struct nvkm_head *head, int div) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + + nvkm_mask(device, 0x614200 + (head->id * 0x800), 0x0000000f, div); +} + +void +nv50_head_rgpos(struct nvkm_head *head, u16 *hline, u16 *vline) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x800; + + /* vline read locks hline. */ + *vline = nvkm_rd32(device, 0x616340 + hoff) & 0x0000ffff; + *hline = nvkm_rd32(device, 0x616344 + hoff) & 0x0000ffff; +} + +static void +nv50_head_state(struct nvkm_head *head, struct nvkm_head_state *state) +{ + struct nvkm_device *device = head->disp->engine.subdev.device; + const u32 hoff = head->id * 0x540 + (state == &head->arm) * 4; + u32 data; + + data = nvkm_rd32(device, 0x610ae8 + hoff); + state->vblanke = (data & 0xffff0000) >> 16; + state->hblanke = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x610af0 + hoff); + state->vblanks = (data & 0xffff0000) >> 16; + state->hblanks = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x610af8 + hoff); + state->vtotal = (data & 0xffff0000) >> 16; + state->htotal = (data & 0x0000ffff); + data = nvkm_rd32(device, 0x610b00 + hoff); + state->vsynce = (data & 0xffff0000) >> 16; + state->hsynce = (data & 0x0000ffff); + state->hz = (nvkm_rd32(device, 0x610ad0 + hoff) & 0x003fffff) * 1000; +} + +static const struct nvkm_head_func +nv50_head = { + .state = nv50_head_state, + .rgpos = nv50_head_rgpos, + .rgclk = nv50_head_rgclk, + .vblank_get = nv50_head_vblank_get, + .vblank_put = nv50_head_vblank_put, +}; + +int +nv50_head_new(struct nvkm_disp *disp, int id) +{ + return nvkm_head_new_(&nv50_head, disp, id); +} + +int +nv50_head_cnt(struct nvkm_disp *disp, unsigned long *pmask) +{ + *pmask = 3; + return 2; +} + + +static void +nv50_disp_mthd_list(struct nvkm_disp *disp, int debug, u32 base, int c, + const struct nvkm_disp_mthd_list *list, int inst) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int i; + + for (i = 0; list->data[i].mthd; i++) { + if (list->data[i].addr) { + u32 next = nvkm_rd32(device, list->data[i].addr + base + 0); + u32 prev = nvkm_rd32(device, list->data[i].addr + base + c); + u32 mthd = list->data[i].mthd + (list->mthd * inst); + const char *name = list->data[i].name; + char mods[16]; + + if (prev != next) + snprintf(mods, sizeof(mods), "-> %08x", next); + else + snprintf(mods, sizeof(mods), "%13c", ' '); + + nvkm_printk_(subdev, debug, info, + "\t%04x: %08x %s%s%s\n", + mthd, prev, mods, name ? " // " : "", + name ? name : ""); + } + } +} + +void +nv50_disp_chan_mthd(struct nvkm_disp_chan *chan, int debug) +{ + struct nvkm_disp *disp = chan->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + const struct nvkm_disp_chan_mthd *mthd = chan->mthd; + const struct nvkm_disp_mthd_list *list; + int i, j; + + if (debug > subdev->debug) + return; + if (!mthd) + return; + + for (i = 0; (list = mthd->data[i].mthd) != NULL; i++) { + u32 base = chan->head * mthd->addr; + for (j = 0; j < mthd->data[i].nr; j++, base += list->addr) { + const char *cname = mthd->name; + const char *sname = ""; + char cname_[16], sname_[16]; + + if (mthd->addr) { + snprintf(cname_, sizeof(cname_), "%s %d", + mthd->name, chan->chid.user); + cname = cname_; + } + + if (mthd->data[i].nr > 1) { + snprintf(sname_, sizeof(sname_), " - %s %d", + mthd->data[i].name, j); + sname = sname_; + } + + nvkm_printk_(subdev, debug, info, "%s%s:\n", cname, sname); + nv50_disp_mthd_list(disp, debug, base, mthd->prev, + list, j); + } + } +} + +static void +nv50_disp_chan_uevent_fini(struct nvkm_event *event, int type, int index) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), uevent); + struct nvkm_device *device = disp->engine.subdev.device; + nvkm_mask(device, 0x610028, 0x00000001 << index, 0x00000000 << index); + nvkm_wr32(device, 0x610020, 0x00000001 << index); +} + +static void +nv50_disp_chan_uevent_init(struct nvkm_event *event, int types, int index) +{ + struct nvkm_disp *disp = container_of(event, typeof(*disp), uevent); + struct nvkm_device *device = disp->engine.subdev.device; + nvkm_wr32(device, 0x610020, 0x00000001 << index); + nvkm_mask(device, 0x610028, 0x00000001 << index, 0x00000001 << index); +} + +void +nv50_disp_chan_uevent_send(struct nvkm_disp *disp, int chid) +{ + nvkm_event_ntfy(&disp->uevent, chid, NVKM_DISP_EVENT_CHAN_AWAKEN); +} + +const struct nvkm_event_func +nv50_disp_chan_uevent = { + .init = nv50_disp_chan_uevent_init, + .fini = nv50_disp_chan_uevent_fini, +}; + +u64 +nv50_disp_chan_user(struct nvkm_disp_chan *chan, u64 *psize) +{ + *psize = 0x1000; + return 0x640000 + (chan->chid.user * 0x1000); +} + +void +nv50_disp_chan_intr(struct nvkm_disp_chan *chan, bool en) +{ + struct nvkm_device *device = chan->disp->engine.subdev.device; + const u32 mask = 0x00010001 << chan->chid.user; + const u32 data = en ? 0x00010000 << chan->chid.user : 0x00000000; + nvkm_mask(device, 0x610028, mask, data); +} + +static void +nv50_disp_pioc_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_disp *disp = chan->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + nvkm_mask(device, 0x610200 + (ctrl * 0x10), 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x00030000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d timeout: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); + } +} + +static int +nv50_disp_pioc_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_disp *disp = chan->disp; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + nvkm_wr32(device, 0x610200 + (ctrl * 0x10), 0x00002000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x00030000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d timeout0: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); + return -EBUSY; + } + + nvkm_wr32(device, 0x610200 + (ctrl * 0x10), 0x00000001); + if (nvkm_msec(device, 2000, + u32 tmp = nvkm_rd32(device, 0x610200 + (ctrl * 0x10)); + if ((tmp & 0x00030000) == 0x00010000) + break; + ) < 0) { + nvkm_error(subdev, "ch %d timeout1: %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +nv50_disp_pioc_func = { + .init = nv50_disp_pioc_init, + .fini = nv50_disp_pioc_fini, + .intr = nv50_disp_chan_intr, + .user = nv50_disp_chan_user, +}; + +int +nv50_disp_dmac_bind(struct nvkm_disp_chan *chan, struct nvkm_object *object, u32 handle) +{ + return nvkm_ramht_insert(chan->disp->ramht, object, chan->chid.user, -10, handle, + chan->chid.user << 28 | chan->chid.user); +} + +static void +nv50_disp_dmac_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* deactivate channel */ + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00001010, 0x00001000); + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00000003, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x001e0000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d fini timeout, %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); + } + + chan->suspend_put = nvkm_rd32(device, 0x640000 + (ctrl * 0x1000)); +} + +static int +nv50_disp_dmac_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ctrl = chan->chid.ctrl; + int user = chan->chid.user; + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x610204 + (ctrl * 0x0010), chan->push); + nvkm_wr32(device, 0x610208 + (ctrl * 0x0010), 0x00010000); + nvkm_wr32(device, 0x61020c + (ctrl * 0x0010), ctrl); + nvkm_mask(device, 0x610200 + (ctrl * 0x0010), 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000 + (ctrl * 0x1000), chan->suspend_put); + nvkm_wr32(device, 0x610200 + (ctrl * 0x0010), 0x00000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200 + (ctrl * 0x10)) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "ch %d init timeout, %08x\n", user, + nvkm_rd32(device, 0x610200 + (ctrl * 0x10))); + return -EBUSY; + } + + return 0; +} + +int +nv50_disp_dmac_push(struct nvkm_disp_chan *chan, u64 object) +{ + chan->memory = nvkm_umem_search(chan->object.client, object); + if (IS_ERR(chan->memory)) + return PTR_ERR(chan->memory); + + if (nvkm_memory_size(chan->memory) < 0x1000) + return -EINVAL; + + switch (nvkm_memory_target(chan->memory)) { + case NVKM_MEM_TARGET_VRAM: chan->push = 0x00000001; break; + case NVKM_MEM_TARGET_NCOH: chan->push = 0x00000002; break; + case NVKM_MEM_TARGET_HOST: chan->push = 0x00000003; break; + default: + return -EINVAL; + } + + chan->push |= nvkm_memory_addr(chan->memory) >> 8; + return 0; +} + +const struct nvkm_disp_chan_func +nv50_disp_dmac_func = { + .push = nv50_disp_dmac_push, + .init = nv50_disp_dmac_init, + .fini = nv50_disp_dmac_fini, + .intr = nv50_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = nv50_disp_dmac_bind, +}; + +const struct nvkm_disp_chan_user +nv50_disp_curs = { + .func = &nv50_disp_pioc_func, + .ctrl = 7, + .user = 7, +}; + +const struct nvkm_disp_chan_user +nv50_disp_oimm = { + .func = &nv50_disp_pioc_func, + .ctrl = 5, + .user = 5, +}; + +static const struct nvkm_disp_mthd_list +nv50_disp_ovly_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x0009a0 }, + { 0x0088, 0x0009c0 }, + { 0x008c, 0x0009c8 }, + { 0x0090, 0x6109b4 }, + { 0x0094, 0x610970 }, + { 0x00a0, 0x610998 }, + { 0x00a4, 0x610964 }, + { 0x00c0, 0x610958 }, + { 0x00e0, 0x6109a8 }, + { 0x00e4, 0x6109d0 }, + { 0x00e8, 0x6109d8 }, + { 0x0100, 0x61094c }, + { 0x0104, 0x610984 }, + { 0x0108, 0x61098c }, + { 0x0800, 0x6109f8 }, + { 0x0808, 0x610a08 }, + { 0x080c, 0x610a10 }, + { 0x0810, 0x610a00 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +nv50_disp_ovly_mthd = { + .name = "Overlay", + .addr = 0x000540, + .prev = 0x000004, + .data = { + { "Global", 1, &nv50_disp_ovly_mthd_base }, + {} + } +}; + +static const struct nvkm_disp_chan_user +nv50_disp_ovly = { + .func = &nv50_disp_dmac_func, + .ctrl = 3, + .user = 3, + .mthd = &nv50_disp_ovly_mthd, +}; + +static const struct nvkm_disp_mthd_list +nv50_disp_base_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x0008c4 }, + { 0x0088, 0x0008d0 }, + { 0x008c, 0x0008dc }, + { 0x0090, 0x0008e4 }, + { 0x0094, 0x610884 }, + { 0x00a0, 0x6108a0 }, + { 0x00a4, 0x610878 }, + { 0x00c0, 0x61086c }, + { 0x00e0, 0x610858 }, + { 0x00e4, 0x610860 }, + { 0x00e8, 0x6108ac }, + { 0x00ec, 0x6108b4 }, + { 0x0100, 0x610894 }, + { 0x0110, 0x6108bc }, + { 0x0114, 0x61088c }, + {} + } +}; + +const struct nvkm_disp_mthd_list +nv50_disp_base_mthd_image = { + .mthd = 0x0400, + .addr = 0x000000, + .data = { + { 0x0800, 0x6108f0 }, + { 0x0804, 0x6108fc }, + { 0x0808, 0x61090c }, + { 0x080c, 0x610914 }, + { 0x0810, 0x610904 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +nv50_disp_base_mthd = { + .name = "Base", + .addr = 0x000540, + .prev = 0x000004, + .data = { + { "Global", 1, &nv50_disp_base_mthd_base }, + { "Image", 2, &nv50_disp_base_mthd_image }, + {} + } +}; + +static const struct nvkm_disp_chan_user +nv50_disp_base = { + .func = &nv50_disp_dmac_func, + .ctrl = 1, + .user = 1, + .mthd = &nv50_disp_base_mthd, +}; + +const struct nvkm_disp_mthd_list +nv50_disp_core_mthd_base = { + .mthd = 0x0000, + .addr = 0x000000, + .data = { + { 0x0080, 0x000000 }, + { 0x0084, 0x610bb8 }, + { 0x0088, 0x610b9c }, + { 0x008c, 0x000000 }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +nv50_disp_core_mthd_dac = { + .mthd = 0x0080, + .addr = 0x000008, + .data = { + { 0x0400, 0x610b58 }, + { 0x0404, 0x610bdc }, + { 0x0420, 0x610828 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +nv50_disp_core_mthd_sor = { + .mthd = 0x0040, + .addr = 0x000008, + .data = { + { 0x0600, 0x610b70 }, + {} + } +}; + +const struct nvkm_disp_mthd_list +nv50_disp_core_mthd_pior = { + .mthd = 0x0040, + .addr = 0x000008, + .data = { + { 0x0700, 0x610b80 }, + {} + } +}; + +static const struct nvkm_disp_mthd_list +nv50_disp_core_mthd_head = { + .mthd = 0x0400, + .addr = 0x000540, + .data = { + { 0x0800, 0x610ad8 }, + { 0x0804, 0x610ad0 }, + { 0x0808, 0x610a48 }, + { 0x080c, 0x610a78 }, + { 0x0810, 0x610ac0 }, + { 0x0814, 0x610af8 }, + { 0x0818, 0x610b00 }, + { 0x081c, 0x610ae8 }, + { 0x0820, 0x610af0 }, + { 0x0824, 0x610b08 }, + { 0x0828, 0x610b10 }, + { 0x082c, 0x610a68 }, + { 0x0830, 0x610a60 }, + { 0x0834, 0x000000 }, + { 0x0838, 0x610a40 }, + { 0x0840, 0x610a24 }, + { 0x0844, 0x610a2c }, + { 0x0848, 0x610aa8 }, + { 0x084c, 0x610ab0 }, + { 0x0860, 0x610a84 }, + { 0x0864, 0x610a90 }, + { 0x0868, 0x610b18 }, + { 0x086c, 0x610b20 }, + { 0x0870, 0x610ac8 }, + { 0x0874, 0x610a38 }, + { 0x0880, 0x610a58 }, + { 0x0884, 0x610a9c }, + { 0x08a0, 0x610a70 }, + { 0x08a4, 0x610a50 }, + { 0x08a8, 0x610ae0 }, + { 0x08c0, 0x610b28 }, + { 0x08c4, 0x610b30 }, + { 0x08c8, 0x610b40 }, + { 0x08d4, 0x610b38 }, + { 0x08d8, 0x610b48 }, + { 0x08dc, 0x610b50 }, + { 0x0900, 0x610a18 }, + { 0x0904, 0x610ab8 }, + {} + } +}; + +static const struct nvkm_disp_chan_mthd +nv50_disp_core_mthd = { + .name = "Core", + .addr = 0x000000, + .prev = 0x000004, + .data = { + { "Global", 1, &nv50_disp_core_mthd_base }, + { "DAC", 3, &nv50_disp_core_mthd_dac }, + { "SOR", 2, &nv50_disp_core_mthd_sor }, + { "PIOR", 3, &nv50_disp_core_mthd_pior }, + { "HEAD", 2, &nv50_disp_core_mthd_head }, + {} + } +}; + +static void +nv50_disp_core_fini(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + /* deactivate channel */ + nvkm_mask(device, 0x610200, 0x00000010, 0x00000000); + nvkm_mask(device, 0x610200, 0x00000003, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200) & 0x001e0000)) + break; + ) < 0) { + nvkm_error(subdev, "core fini: %08x\n", + nvkm_rd32(device, 0x610200)); + } + + chan->suspend_put = nvkm_rd32(device, 0x640000); +} + +static int +nv50_disp_core_init(struct nvkm_disp_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->disp->engine.subdev; + struct nvkm_device *device = subdev->device; + + /* attempt to unstick channel from some unknown state */ + if ((nvkm_rd32(device, 0x610200) & 0x009f0000) == 0x00020000) + nvkm_mask(device, 0x610200, 0x00800000, 0x00800000); + if ((nvkm_rd32(device, 0x610200) & 0x003f0000) == 0x00030000) + nvkm_mask(device, 0x610200, 0x00600000, 0x00600000); + + /* initialise channel for dma command submission */ + nvkm_wr32(device, 0x610204, chan->push); + nvkm_wr32(device, 0x610208, 0x00010000); + nvkm_wr32(device, 0x61020c, 0x00000000); + nvkm_mask(device, 0x610200, 0x00000010, 0x00000010); + nvkm_wr32(device, 0x640000, chan->suspend_put); + nvkm_wr32(device, 0x610200, 0x01000013); + + /* wait for it to go inactive */ + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x610200) & 0x80000000)) + break; + ) < 0) { + nvkm_error(subdev, "core init: %08x\n", + nvkm_rd32(device, 0x610200)); + return -EBUSY; + } + + return 0; +} + +const struct nvkm_disp_chan_func +nv50_disp_core_func = { + .push = nv50_disp_dmac_push, + .init = nv50_disp_core_init, + .fini = nv50_disp_core_fini, + .intr = nv50_disp_chan_intr, + .user = nv50_disp_chan_user, + .bind = nv50_disp_dmac_bind, +}; + +static const struct nvkm_disp_chan_user +nv50_disp_core = { + .func = &nv50_disp_core_func, + .ctrl = 0, + .user = 0, + .mthd = &nv50_disp_core_mthd, +}; + +static u32 +nv50_disp_super_iedt(struct nvkm_head *head, struct nvkm_outp *outp, + u8 *ver, u8 *hdr, u8 *cnt, u8 *len, + struct nvbios_outp *iedt) +{ + struct nvkm_bios *bios = head->disp->engine.subdev.device->bios; + const u8 l = ffs(outp->info.link); + const u16 t = outp->info.hasht; + const u16 m = (0x0100 << head->id) | (l << 6) | outp->info.or; + u32 data = nvbios_outp_match(bios, t, m, ver, hdr, cnt, len, iedt); + if (!data) + OUTP_DBG(outp, "missing IEDT for %04x:%04x", t, m); + return data; +} + +static void +nv50_disp_super_ied_on(struct nvkm_head *head, + struct nvkm_ior *ior, int id, u32 khz) +{ + struct nvkm_subdev *subdev = &head->disp->engine.subdev; + struct nvkm_bios *bios = subdev->device->bios; + struct nvkm_outp *outp = ior->asy.outp; + struct nvbios_ocfg iedtrs; + struct nvbios_outp iedt; + u8 ver, hdr, cnt, len, flags = 0x00; + u32 data; + + if (!outp) { + IOR_DBG(ior, "nothing to attach"); + return; + } + + /* Lookup IED table for the device. */ + data = nv50_disp_super_iedt(head, outp, &ver, &hdr, &cnt, &len, &iedt); + if (!data) + return; + + /* Lookup IEDT runtime settings for the current configuration. */ + if (ior->type == SOR) { + if (ior->asy.proto == LVDS) { + if (head->asy.or.depth == 24) + flags |= 0x02; + } + if (ior->asy.link == 3) + flags |= 0x01; + } + + data = nvbios_ocfg_match(bios, data, ior->asy.proto_evo, flags, + &ver, &hdr, &cnt, &len, &iedtrs); + if (!data) { + OUTP_DBG(outp, "missing IEDT RS for %02x:%02x", + ior->asy.proto_evo, flags); + return; + } + + /* Execute the OnInt[23] script for the current frequency. */ + data = nvbios_oclk_match(bios, iedtrs.clkcmp[id], khz); + if (!data) { + OUTP_DBG(outp, "missing IEDT RSS %d for %02x:%02x %d khz", + id, ior->asy.proto_evo, flags, khz); + return; + } + + nvbios_init(subdev, data, + init.outp = &outp->info; + init.or = ior->id; + init.link = ior->asy.link; + init.head = head->id; + ); +} + +static void +nv50_disp_super_ied_off(struct nvkm_head *head, struct nvkm_ior *ior, int id) +{ + struct nvkm_outp *outp = ior->arm.outp; + struct nvbios_outp iedt; + u8 ver, hdr, cnt, len; + u32 data; + + if (!outp) { + IOR_DBG(ior, "nothing attached"); + return; + } + + data = nv50_disp_super_iedt(head, outp, &ver, &hdr, &cnt, &len, &iedt); + if (!data) + return; + + nvbios_init(&head->disp->engine.subdev, iedt.script[id], + init.outp = &outp->info; + init.or = ior->id; + init.link = ior->arm.link; + init.head = head->id; + ); +} + +static struct nvkm_ior * +nv50_disp_super_ior_asy(struct nvkm_head *head) +{ + struct nvkm_ior *ior; + list_for_each_entry(ior, &head->disp->iors, head) { + if (ior->asy.head & (1 << head->id)) { + HEAD_DBG(head, "to %s", ior->name); + return ior; + } + } + HEAD_DBG(head, "nothing to attach"); + return NULL; +} + +static struct nvkm_ior * +nv50_disp_super_ior_arm(struct nvkm_head *head) +{ + struct nvkm_ior *ior; + list_for_each_entry(ior, &head->disp->iors, head) { + if (ior->arm.head & (1 << head->id)) { + HEAD_DBG(head, "on %s", ior->name); + return ior; + } + } + HEAD_DBG(head, "nothing attached"); + return NULL; +} + +void +nv50_disp_super_3_0(struct nvkm_disp *disp, struct nvkm_head *head) +{ + struct nvkm_ior *ior; + + /* Determine which OR, if any, we're attaching to the head. */ + HEAD_DBG(head, "supervisor 3.0"); + ior = nv50_disp_super_ior_asy(head); + if (!ior) + return; + + /* Execute OnInt3 IED script. */ + nv50_disp_super_ied_on(head, ior, 1, head->asy.hz / 1000); + + /* OR-specific handling. */ + if (ior->func->war_3) + ior->func->war_3(ior); +} + +static void +nv50_disp_super_2_2_dp(struct nvkm_head *head, struct nvkm_ior *ior) +{ + struct nvkm_subdev *subdev = &head->disp->engine.subdev; + const u32 khz = head->asy.hz / 1000; + const u32 linkKBps = ior->dp.bw * 27000; + const u32 symbol = 100000; + int bestTU = 0, bestVTUi = 0, bestVTUf = 0, bestVTUa = 0; + int TU, VTUi, VTUf, VTUa; + u64 link_data_rate, link_ratio, unk; + u32 best_diff = 64 * symbol; + u64 h, v; + + /* symbols/hblank - algorithm taken from comments in tegra driver */ + h = head->asy.hblanke + head->asy.htotal - head->asy.hblanks - 7; + h = h * linkKBps; + do_div(h, khz); + h = h - (3 * ior->dp.ef) - (12 / ior->dp.nr); + + /* symbols/vblank - algorithm taken from comments in tegra driver */ + v = head->asy.vblanks - head->asy.vblanke - 25; + v = v * linkKBps; + do_div(v, khz); + v = v - ((36 / ior->dp.nr) + 3) - 1; + + ior->func->dp->audio_sym(ior, head->id, h, v); + + /* watermark / activesym */ + link_data_rate = (khz * head->asy.or.depth / 8) / ior->dp.nr; + + /* calculate ratio of packed data rate to link symbol rate */ + link_ratio = link_data_rate * symbol; + do_div(link_ratio, linkKBps); + + for (TU = 64; ior->func->dp->activesym && TU >= 32; TU--) { + /* calculate average number of valid symbols in each TU */ + u32 tu_valid = link_ratio * TU; + u32 calc, diff; + + /* find a hw representation for the fraction.. */ + VTUi = tu_valid / symbol; + calc = VTUi * symbol; + diff = tu_valid - calc; + if (diff) { + if (diff >= (symbol / 2)) { + VTUf = symbol / (symbol - diff); + if (symbol - (VTUf * diff)) + VTUf++; + + if (VTUf <= 15) { + VTUa = 1; + calc += symbol - (symbol / VTUf); + } else { + VTUa = 0; + VTUf = 1; + calc += symbol; + } + } else { + VTUa = 0; + VTUf = min((int)(symbol / diff), 15); + calc += symbol / VTUf; + } + + diff = calc - tu_valid; + } else { + /* no remainder, but the hw doesn't like the fractional + * part to be zero. decrement the integer part and + * have the fraction add a whole symbol back + */ + VTUa = 0; + VTUf = 1; + VTUi--; + } + + if (diff < best_diff) { + best_diff = diff; + bestTU = TU; + bestVTUa = VTUa; + bestVTUf = VTUf; + bestVTUi = VTUi; + if (diff == 0) + break; + } + } + + if (ior->func->dp->activesym) { + if (!bestTU) { + nvkm_error(subdev, "unable to determine dp config\n"); + return; + } + + ior->func->dp->activesym(ior, head->id, bestTU, bestVTUa, bestVTUf, bestVTUi); + } else { + bestTU = 64; + } + + /* XXX close to vbios numbers, but not right */ + unk = (symbol - link_ratio) * bestTU; + unk *= link_ratio; + do_div(unk, symbol); + do_div(unk, symbol); + unk += 6; + + ior->func->dp->watermark(ior, head->id, unk); +} + +void +nv50_disp_super_2_2(struct nvkm_disp *disp, struct nvkm_head *head) +{ + const u32 khz = head->asy.hz / 1000; + struct nvkm_outp *outp; + struct nvkm_ior *ior; + + /* Determine which OR, if any, we're attaching from the head. */ + HEAD_DBG(head, "supervisor 2.2"); + ior = nv50_disp_super_ior_asy(head); + if (!ior) + return; + + outp = ior->asy.outp; + + /* For some reason, NVIDIA decided not to: + * + * A) Give dual-link LVDS a separate EVO protocol, like for TMDS. + * and + * B) Use SetControlOutputResource.PixelDepth on LVDS. + * + * Override the values we usually read from HW with the same + * data we pass though an ioctl instead. + */ + if (outp && ior->type == SOR && ior->asy.proto == LVDS) { + head->asy.or.depth = outp->lvds.bpc8 ? 24 : 18; + ior->asy.link = outp->lvds.dual ? 3 : 1; + } + + /* Handle any link training, etc. */ + if (outp && outp->func->acquire) + outp->func->acquire(outp); + + /* Execute OnInt2 IED script. */ + nv50_disp_super_ied_on(head, ior, 0, khz); + + /* Program RG clock divider. */ + head->func->rgclk(head, ior->asy.rgdiv); + + /* Mode-specific internal DP configuration. */ + if (ior->type == SOR && ior->asy.proto == DP) + nv50_disp_super_2_2_dp(head, ior); + + /* OR-specific handling. */ + ior->func->clock(ior); + if (ior->func->war_2) + ior->func->war_2(ior); +} + +void +nv50_disp_super_2_1(struct nvkm_disp *disp, struct nvkm_head *head) +{ + struct nvkm_devinit *devinit = disp->engine.subdev.device->devinit; + const u32 khz = head->asy.hz / 1000; + HEAD_DBG(head, "supervisor 2.1 - %d khz", khz); + if (khz) + nvkm_devinit_pll_set(devinit, PLL_VPLL0 + head->id, khz); +} + +void +nv50_disp_super_2_0(struct nvkm_disp *disp, struct nvkm_head *head) +{ + struct nvkm_outp *outp; + struct nvkm_ior *ior; + + /* Determine which OR, if any, we're detaching from the head. */ + HEAD_DBG(head, "supervisor 2.0"); + ior = nv50_disp_super_ior_arm(head); + if (!ior) + return; + + /* Execute OffInt2 IED script. */ + nv50_disp_super_ied_off(head, ior, 2); + + /* If we're shutting down the OR's only active head, execute + * the output path's disable function. + */ + if (ior->arm.head == (1 << head->id)) { + if ((outp = ior->arm.outp) && outp->func->disable) + outp->func->disable(outp, ior); + } +} + +void +nv50_disp_super_1_0(struct nvkm_disp *disp, struct nvkm_head *head) +{ + struct nvkm_ior *ior; + + /* Determine which OR, if any, we're detaching from the head. */ + HEAD_DBG(head, "supervisor 1.0"); + ior = nv50_disp_super_ior_arm(head); + if (!ior) + return; + + /* Execute OffInt1 IED script. */ + nv50_disp_super_ied_off(head, ior, 1); +} + +void +nv50_disp_super_1(struct nvkm_disp *disp) +{ + struct nvkm_head *head; + struct nvkm_ior *ior; + + list_for_each_entry(head, &disp->heads, head) { + head->func->state(head, &head->arm); + head->func->state(head, &head->asy); + } + + list_for_each_entry(ior, &disp->iors, head) { + ior->func->state(ior, &ior->arm); + ior->func->state(ior, &ior->asy); + } +} + +void +nv50_disp_super(struct work_struct *work) +{ + struct nvkm_disp *disp = container_of(work, struct nvkm_disp, super.work); + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_head *head; + u32 super; + + mutex_lock(&disp->super.mutex); + super = nvkm_rd32(device, 0x610030); + + nvkm_debug(subdev, "supervisor %08x %08x\n", disp->super.pending, super); + + if (disp->super.pending & 0x00000010) { + nv50_disp_chan_mthd(disp->chan[0], NV_DBG_DEBUG); + nv50_disp_super_1(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(super & (0x00000020 << head->id))) + continue; + if (!(super & (0x00000080 << head->id))) + continue; + nv50_disp_super_1_0(disp, head); + } + } else + if (disp->super.pending & 0x00000020) { + list_for_each_entry(head, &disp->heads, head) { + if (!(super & (0x00000080 << head->id))) + continue; + nv50_disp_super_2_0(disp, head); + } + nvkm_outp_route(disp); + list_for_each_entry(head, &disp->heads, head) { + if (!(super & (0x00000200 << head->id))) + continue; + nv50_disp_super_2_1(disp, head); + } + list_for_each_entry(head, &disp->heads, head) { + if (!(super & (0x00000080 << head->id))) + continue; + nv50_disp_super_2_2(disp, head); + } + } else + if (disp->super.pending & 0x00000040) { + list_for_each_entry(head, &disp->heads, head) { + if (!(super & (0x00000080 << head->id))) + continue; + nv50_disp_super_3_0(disp, head); + } + } + + nvkm_wr32(device, 0x610030, 0x80000000); + mutex_unlock(&disp->super.mutex); +} + +const struct nvkm_enum +nv50_disp_intr_error_type[] = { + { 0, "NONE" }, + { 1, "PUSHBUFFER_ERR" }, + { 2, "TRAP" }, + { 3, "RESERVED_METHOD" }, + { 4, "INVALID_ARG" }, + { 5, "INVALID_STATE" }, + { 7, "UNRESOLVABLE_HANDLE" }, + {} +}; + +static const struct nvkm_enum +nv50_disp_intr_error_code[] = { + { 0x00, "" }, + {} +}; + +static void +nv50_disp_intr_error(struct nvkm_disp *disp, int chid) +{ + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 data = nvkm_rd32(device, 0x610084 + (chid * 0x08)); + u32 addr = nvkm_rd32(device, 0x610080 + (chid * 0x08)); + u32 code = (addr & 0x00ff0000) >> 16; + u32 type = (addr & 0x00007000) >> 12; + u32 mthd = (addr & 0x00000ffc); + const struct nvkm_enum *ec, *et; + + et = nvkm_enum_find(nv50_disp_intr_error_type, type); + ec = nvkm_enum_find(nv50_disp_intr_error_code, code); + + nvkm_error(subdev, + "ERROR %d [%s] %02x [%s] chid %d mthd %04x data %08x\n", + type, et ? et->name : "", code, ec ? ec->name : "", + chid, mthd, data); + + if (chid < ARRAY_SIZE(disp->chan)) { + switch (mthd) { + case 0x0080: + nv50_disp_chan_mthd(disp->chan[chid], NV_DBG_ERROR); + break; + default: + break; + } + } + + nvkm_wr32(device, 0x610020, 0x00010000 << chid); + nvkm_wr32(device, 0x610080 + (chid * 0x08), 0x90000000); +} + +void +nv50_disp_intr(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 intr0 = nvkm_rd32(device, 0x610020); + u32 intr1 = nvkm_rd32(device, 0x610024); + + while (intr0 & 0x001f0000) { + u32 chid = __ffs(intr0 & 0x001f0000) - 16; + nv50_disp_intr_error(disp, chid); + intr0 &= ~(0x00010000 << chid); + } + + while (intr0 & 0x0000001f) { + u32 chid = __ffs(intr0 & 0x0000001f); + nv50_disp_chan_uevent_send(disp, chid); + intr0 &= ~(0x00000001 << chid); + } + + if (intr1 & 0x00000004) { + nvkm_disp_vblank(disp, 0); + nvkm_wr32(device, 0x610024, 0x00000004); + } + + if (intr1 & 0x00000008) { + nvkm_disp_vblank(disp, 1); + nvkm_wr32(device, 0x610024, 0x00000008); + } + + if (intr1 & 0x00000070) { + disp->super.pending = (intr1 & 0x00000070); + queue_work(disp->super.wq, &disp->super.work); + nvkm_wr32(device, 0x610024, disp->super.pending); + } +} + +void +nv50_disp_fini(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + /* disable all interrupts */ + nvkm_wr32(device, 0x610024, 0x00000000); + nvkm_wr32(device, 0x610020, 0x00000000); +} + +int +nv50_disp_init(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_head *head; + u32 tmp; + int i; + + /* The below segments of code copying values from one register to + * another appear to inform EVO of the display capabilities or + * something similar. NFI what the 0x614004 caps are for.. + */ + tmp = nvkm_rd32(device, 0x614004); + nvkm_wr32(device, 0x610184, tmp); + + /* ... CRTC caps */ + list_for_each_entry(head, &disp->heads, head) { + tmp = nvkm_rd32(device, 0x616100 + (head->id * 0x800)); + nvkm_wr32(device, 0x610190 + (head->id * 0x10), tmp); + tmp = nvkm_rd32(device, 0x616104 + (head->id * 0x800)); + nvkm_wr32(device, 0x610194 + (head->id * 0x10), tmp); + tmp = nvkm_rd32(device, 0x616108 + (head->id * 0x800)); + nvkm_wr32(device, 0x610198 + (head->id * 0x10), tmp); + tmp = nvkm_rd32(device, 0x61610c + (head->id * 0x800)); + nvkm_wr32(device, 0x61019c + (head->id * 0x10), tmp); + } + + /* ... DAC caps */ + for (i = 0; i < disp->dac.nr; i++) { + tmp = nvkm_rd32(device, 0x61a000 + (i * 0x800)); + nvkm_wr32(device, 0x6101d0 + (i * 0x04), tmp); + } + + /* ... SOR caps */ + for (i = 0; i < disp->sor.nr; i++) { + tmp = nvkm_rd32(device, 0x61c000 + (i * 0x800)); + nvkm_wr32(device, 0x6101e0 + (i * 0x04), tmp); + } + + /* ... PIOR caps */ + for (i = 0; i < disp->pior.nr; i++) { + tmp = nvkm_rd32(device, 0x61e000 + (i * 0x800)); + nvkm_wr32(device, 0x6101f0 + (i * 0x04), tmp); + } + + /* steal display away from vbios, or something like that */ + if (nvkm_rd32(device, 0x610024) & 0x00000100) { + nvkm_wr32(device, 0x610024, 0x00000100); + nvkm_mask(device, 0x6194e8, 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x6194e8) & 0x00000002)) + break; + ) < 0) + return -EBUSY; + } + + /* point at display engine memory area (hash table, objects) */ + nvkm_wr32(device, 0x610010, (disp->inst->addr >> 8) | 9); + + /* enable supervisor interrupts, disable everything else */ + nvkm_wr32(device, 0x61002c, 0x00000370); + nvkm_wr32(device, 0x610028, 0x00000000); + return 0; +} + +int +nv50_disp_oneinit(struct nvkm_disp *disp) +{ + const struct nvkm_disp_func *func = disp->func; + struct nvkm_subdev *subdev = &disp->engine.subdev; + struct nvkm_device *device = subdev->device; + int ret, i; + + if (func->wndw.cnt) { + disp->wndw.nr = func->wndw.cnt(disp, &disp->wndw.mask); + nvkm_debug(subdev, "Window(s): %d (%08lx)\n", disp->wndw.nr, disp->wndw.mask); + } + + disp->head.nr = func->head.cnt(disp, &disp->head.mask); + nvkm_debug(subdev, " Head(s): %d (%02lx)\n", disp->head.nr, disp->head.mask); + for_each_set_bit(i, &disp->head.mask, disp->head.nr) { + ret = func->head.new(disp, i); + if (ret) + return ret; + } + + if (func->dac.cnt) { + disp->dac.nr = func->dac.cnt(disp, &disp->dac.mask); + nvkm_debug(subdev, " DAC(s): %d (%02lx)\n", disp->dac.nr, disp->dac.mask); + for_each_set_bit(i, &disp->dac.mask, disp->dac.nr) { + ret = func->dac.new(disp, i); + if (ret) + return ret; + } + } + + if (func->pior.cnt) { + disp->pior.nr = func->pior.cnt(disp, &disp->pior.mask); + nvkm_debug(subdev, " PIOR(s): %d (%02lx)\n", disp->pior.nr, disp->pior.mask); + for_each_set_bit(i, &disp->pior.mask, disp->pior.nr) { + ret = func->pior.new(disp, i); + if (ret) + return ret; + } + } + + disp->sor.nr = func->sor.cnt(disp, &disp->sor.mask); + nvkm_debug(subdev, " SOR(s): %d (%02lx)\n", disp->sor.nr, disp->sor.mask); + for_each_set_bit(i, &disp->sor.mask, disp->sor.nr) { + ret = func->sor.new(disp, i); + if (ret) + return ret; + } + + ret = nvkm_gpuobj_new(device, 0x10000, 0x10000, false, NULL, &disp->inst); + if (ret) + return ret; + + return nvkm_ramht_new(device, func->ramht_size ? func->ramht_size : + 0x1000, 0, disp->inst, &disp->ramht); +} + +static const struct nvkm_disp_func +nv50_disp = { + .oneinit = nv50_disp_oneinit, + .init = nv50_disp_init, + .fini = nv50_disp_fini, + .intr = nv50_disp_intr, + .super = nv50_disp_super, + .uevent = &nv50_disp_chan_uevent, + .head = { .cnt = nv50_head_cnt, .new = nv50_head_new }, + .dac = { .cnt = nv50_dac_cnt, .new = nv50_dac_new }, + .sor = { .cnt = nv50_sor_cnt, .new = nv50_sor_new }, + .pior = { .cnt = nv50_pior_cnt, .new = nv50_pior_new }, + .root = { 0, 0, NV50_DISP }, + .user = { + {{0,0,NV50_DISP_CURSOR }, nvkm_disp_chan_new, &nv50_disp_curs }, + {{0,0,NV50_DISP_OVERLAY }, nvkm_disp_chan_new, &nv50_disp_oimm }, + {{0,0,NV50_DISP_BASE_CHANNEL_DMA }, nvkm_disp_chan_new, &nv50_disp_base }, + {{0,0,NV50_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &nv50_disp_core }, + {{0,0,NV50_DISP_OVERLAY_CHANNEL_DMA}, nvkm_disp_chan_new, &nv50_disp_ovly }, + {} + } +}; + +int +nv50_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&nv50_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c new file mode 100644 index 0000000000..6094805fbd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c @@ -0,0 +1,338 @@ +/* + * Copyright 2014 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "outp.h" +#include "dp.h" +#include "ior.h" + +#include +#include +#include + +void +nvkm_outp_route(struct nvkm_disp *disp) +{ + struct nvkm_outp *outp; + struct nvkm_ior *ior; + + list_for_each_entry(ior, &disp->iors, head) { + if ((outp = ior->arm.outp) && ior->arm.outp != ior->asy.outp) { + OUTP_DBG(outp, "release %s", ior->name); + if (ior->func->route.set) + ior->func->route.set(outp, NULL); + ior->arm.outp = NULL; + } + } + + list_for_each_entry(ior, &disp->iors, head) { + if ((outp = ior->asy.outp)) { + OUTP_DBG(outp, "acquire %s", ior->name); + if (ior->asy.outp != ior->arm.outp) { + if (ior->func->route.set) + ior->func->route.set(outp, ior); + ior->arm.outp = ior->asy.outp; + } + } + } +} + +static enum nvkm_ior_proto +nvkm_outp_xlat(struct nvkm_outp *outp, enum nvkm_ior_type *type) +{ + switch (outp->info.location) { + case 0: + switch (outp->info.type) { + case DCB_OUTPUT_ANALOG: *type = DAC; return CRT; + case DCB_OUTPUT_TV : *type = DAC; return TV; + case DCB_OUTPUT_TMDS : *type = SOR; return TMDS; + case DCB_OUTPUT_LVDS : *type = SOR; return LVDS; + case DCB_OUTPUT_DP : *type = SOR; return DP; + default: + break; + } + break; + case 1: + switch (outp->info.type) { + case DCB_OUTPUT_TMDS: *type = PIOR; return TMDS; + case DCB_OUTPUT_DP : *type = PIOR; return TMDS; /* not a bug */ + default: + break; + } + break; + default: + break; + } + WARN_ON(1); + return UNKNOWN; +} + +void +nvkm_outp_release(struct nvkm_outp *outp, u8 user) +{ + struct nvkm_ior *ior = outp->ior; + OUTP_TRACE(outp, "release %02x &= %02x %p", outp->acquired, ~user, ior); + if (ior) { + outp->acquired &= ~user; + if (!outp->acquired) { + if (outp->func->release && outp->ior) + outp->func->release(outp); + outp->ior->asy.outp = NULL; + outp->ior = NULL; + } + } +} + +static inline int +nvkm_outp_acquire_ior(struct nvkm_outp *outp, u8 user, struct nvkm_ior *ior) +{ + outp->ior = ior; + outp->ior->asy.outp = outp; + outp->ior->asy.link = outp->info.sorconf.link; + outp->acquired |= user; + return 0; +} + +static inline int +nvkm_outp_acquire_hda(struct nvkm_outp *outp, enum nvkm_ior_type type, + u8 user, bool hda) +{ + struct nvkm_ior *ior; + + /* Failing that, a completely unused OR is the next best thing. */ + list_for_each_entry(ior, &outp->disp->iors, head) { + if (!ior->identity && ior->hda == hda && + !ior->asy.outp && ior->type == type && !ior->arm.outp && + (ior->func->route.set || ior->id == __ffs(outp->info.or))) + return nvkm_outp_acquire_ior(outp, user, ior); + } + + /* Last resort is to assign an OR that's already active on HW, + * but will be released during the next modeset. + */ + list_for_each_entry(ior, &outp->disp->iors, head) { + if (!ior->identity && ior->hda == hda && + !ior->asy.outp && ior->type == type && + (ior->func->route.set || ior->id == __ffs(outp->info.or))) + return nvkm_outp_acquire_ior(outp, user, ior); + } + + return -ENOSPC; +} + +int +nvkm_outp_acquire(struct nvkm_outp *outp, u8 user, bool hda) +{ + struct nvkm_ior *ior = outp->ior; + enum nvkm_ior_proto proto; + enum nvkm_ior_type type; + + OUTP_TRACE(outp, "acquire %02x |= %02x %p", outp->acquired, user, ior); + if (ior) { + outp->acquired |= user; + return 0; + } + + /* Lookup a compatible, and unused, OR to assign to the device. */ + proto = nvkm_outp_xlat(outp, &type); + if (proto == UNKNOWN) + return -ENOSYS; + + /* Deal with panels requiring identity-mapped SOR assignment. */ + if (outp->identity) { + ior = nvkm_ior_find(outp->disp, SOR, ffs(outp->info.or) - 1); + if (WARN_ON(!ior)) + return -ENOSPC; + return nvkm_outp_acquire_ior(outp, user, ior); + } + + /* First preference is to reuse the OR that is currently armed + * on HW, if any, in order to prevent unnecessary switching. + */ + list_for_each_entry(ior, &outp->disp->iors, head) { + if (!ior->identity && !ior->asy.outp && ior->arm.outp == outp) { + /*XXX: For various complicated reasons, we can't outright switch + * the boot-time OR on the first modeset without some fairly + * invasive changes. + * + * The systems that were fixed by modifying the OR selection + * code to account for HDA support shouldn't regress here as + * the HDA-enabled ORs match the relevant output's pad macro + * index, and the firmware seems to select an OR this way. + * + * This warning is to make it obvious if that proves wrong. + */ + WARN_ON(hda && !ior->hda); + return nvkm_outp_acquire_ior(outp, user, ior); + } + } + + /* If we don't need HDA, first try to acquire an OR that doesn't + * support it to leave free the ones that do. + */ + if (!hda) { + if (!nvkm_outp_acquire_hda(outp, type, user, false)) + return 0; + + /* Use a HDA-supporting SOR anyway. */ + return nvkm_outp_acquire_hda(outp, type, user, true); + } + + /* We want HDA, try to acquire an OR that supports it. */ + if (!nvkm_outp_acquire_hda(outp, type, user, true)) + return 0; + + /* There weren't any free ORs that support HDA, grab one that + * doesn't and at least allow display to work still. + */ + return nvkm_outp_acquire_hda(outp, type, user, false); +} + +void +nvkm_outp_fini(struct nvkm_outp *outp) +{ + if (outp->func->fini) + outp->func->fini(outp); +} + +static void +nvkm_outp_init_route(struct nvkm_outp *outp) +{ + struct nvkm_disp *disp = outp->disp; + enum nvkm_ior_proto proto; + enum nvkm_ior_type type; + struct nvkm_ior *ior; + int id, link; + + /* Find any OR from the class that is able to support this device. */ + proto = nvkm_outp_xlat(outp, &type); + if (proto == UNKNOWN) + return; + + ior = nvkm_ior_find(disp, type, -1); + if (!ior) { + WARN_ON(1); + return; + } + + /* Determine the specific OR, if any, this device is attached to. */ + if (ior->func->route.get) { + id = ior->func->route.get(outp, &link); + if (id < 0) { + OUTP_DBG(outp, "no route"); + return; + } + } else { + /* Prior to DCB 4.1, this is hardwired like so. */ + id = ffs(outp->info.or) - 1; + link = (ior->type == SOR) ? outp->info.sorconf.link : 0; + } + + ior = nvkm_ior_find(disp, type, id); + if (!ior) { + WARN_ON(1); + return; + } + + /* Determine if the OR is already configured for this device. */ + ior->func->state(ior, &ior->arm); + if (!ior->arm.head || ior->arm.proto != proto) { + OUTP_DBG(outp, "no heads (%x %d %d)", ior->arm.head, + ior->arm.proto, proto); + + /* The EFI GOP driver on Ampere can leave unused DP links routed, + * which we don't expect. The DisableLT IED script *should* get + * us back to where we need to be. + */ + if (ior->func->route.get && !ior->arm.head && outp->info.type == DCB_OUTPUT_DP) + nvkm_dp_disable(outp, ior); + + return; + } + + OUTP_DBG(outp, "on %s link %x", ior->name, ior->arm.link); + ior->arm.outp = outp; +} + +void +nvkm_outp_init(struct nvkm_outp *outp) +{ + nvkm_outp_init_route(outp); + if (outp->func->init) + outp->func->init(outp); +} + +void +nvkm_outp_del(struct nvkm_outp **poutp) +{ + struct nvkm_outp *outp = *poutp; + if (outp && !WARN_ON(!outp->func)) { + if (outp->func->dtor) + *poutp = outp->func->dtor(outp); + kfree(*poutp); + *poutp = NULL; + } +} + +int +nvkm_outp_new_(const struct nvkm_outp_func *func, struct nvkm_disp *disp, + int index, struct dcb_output *dcbE, struct nvkm_outp **poutp) +{ + struct nvkm_i2c *i2c = disp->engine.subdev.device->i2c; + struct nvkm_outp *outp; + enum nvkm_ior_proto proto; + enum nvkm_ior_type type; + + if (!(outp = *poutp = kzalloc(sizeof(*outp), GFP_KERNEL))) + return -ENOMEM; + + outp->func = func; + outp->disp = disp; + outp->index = index; + outp->info = *dcbE; + outp->i2c = nvkm_i2c_bus_find(i2c, dcbE->i2c_index); + + OUTP_DBG(outp, "type %02x loc %d or %d link %d con %x " + "edid %x bus %d head %x", + outp->info.type, outp->info.location, outp->info.or, + outp->info.type >= 2 ? outp->info.sorconf.link : 0, + outp->info.connector, outp->info.i2c_index, + outp->info.bus, outp->info.heads); + + /* Cull output paths we can't map to an output resource. */ + proto = nvkm_outp_xlat(outp, &type); + if (proto == UNKNOWN) + return -ENODEV; + + return 0; +} + +static const struct nvkm_outp_func +nvkm_outp = { +}; + +int +nvkm_outp_new(struct nvkm_disp *disp, int index, struct dcb_output *dcbE, + struct nvkm_outp **poutp) +{ + return nvkm_outp_new_(&nvkm_outp, disp, index, dcbE, poutp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h new file mode 100644 index 0000000000..4e7f873f66 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.h @@ -0,0 +1,98 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_OUTP_H__ +#define __NVKM_DISP_OUTP_H__ +#include "priv.h" + +#include +#include +#include +#include + +struct nvkm_outp { + const struct nvkm_outp_func *func; + struct nvkm_disp *disp; + int index; + struct dcb_output info; + + struct nvkm_i2c_bus *i2c; + + struct list_head head; + struct nvkm_conn *conn; + bool identity; + + /* Assembly state. */ +#define NVKM_OUTP_PRIV 1 +#define NVKM_OUTP_USER 2 + u8 acquired:2; + struct nvkm_ior *ior; + + union { + struct { + bool dual; + bool bpc8; + } lvds; + + struct { + struct nvbios_dpout info; + u8 version; + + struct nvkm_i2c_aux *aux; + + bool enabled; + bool aux_pwr; + bool aux_pwr_pu; + u8 lttpr[6]; + u8 lttprs; + u8 dpcd[DP_RECEIVER_CAP_SIZE]; + + struct { + int dpcd; /* -1, or index into SUPPORTED_LINK_RATES table */ + u32 rate; + } rate[8]; + int rates; + int links; + + struct mutex mutex; + struct { + atomic_t done; + u8 nr; + u8 bw; + bool mst; + } lt; + } dp; + }; + + struct nvkm_object object; + struct { + struct nvkm_head *head; + } asy; +}; + +int nvkm_outp_new_(const struct nvkm_outp_func *, struct nvkm_disp *, int index, + struct dcb_output *, struct nvkm_outp **); +int nvkm_outp_new(struct nvkm_disp *, int index, struct dcb_output *, struct nvkm_outp **); +void nvkm_outp_del(struct nvkm_outp **); +void nvkm_outp_init(struct nvkm_outp *); +void nvkm_outp_fini(struct nvkm_outp *); +int nvkm_outp_acquire(struct nvkm_outp *, u8 user, bool hda); +void nvkm_outp_release(struct nvkm_outp *, u8 user); +void nvkm_outp_route(struct nvkm_disp *); + +struct nvkm_outp_func { + void *(*dtor)(struct nvkm_outp *); + void (*init)(struct nvkm_outp *); + void (*fini)(struct nvkm_outp *); + int (*acquire)(struct nvkm_outp *); + void (*release)(struct nvkm_outp *); + void (*disable)(struct nvkm_outp *, struct nvkm_ior *); +}; + +#define OUTP_MSG(o,l,f,a...) do { \ + struct nvkm_outp *_outp = (o); \ + nvkm_##l(&_outp->disp->engine.subdev, "outp %02x:%04x:%04x: "f"\n", \ + _outp->index, _outp->info.hasht, _outp->info.hashm, ##a); \ +} while(0) +#define OUTP_ERR(o,f,a...) OUTP_MSG((o), error, f, ##a) +#define OUTP_DBG(o,f,a...) OUTP_MSG((o), debug, f, ##a) +#define OUTP_TRACE(o,f,a...) OUTP_MSG((o), trace, f, ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/disp/priv.h new file mode 100644 index 0000000000..ec5292a8f3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/priv.h @@ -0,0 +1,86 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DISP_PRIV_H__ +#define __NVKM_DISP_PRIV_H__ +#define nvkm_udisp(p) container_of((p), struct nvkm_disp, client.object) +#include +#include +struct nvkm_head; +struct nvkm_outp; +struct dcb_output; + +int nvkm_disp_ctor(const struct nvkm_disp_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_disp *); +int nvkm_disp_new_(const struct nvkm_disp_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_disp **); +void nvkm_disp_vblank(struct nvkm_disp *, int head); + +struct nvkm_disp_func { + int (*oneinit)(struct nvkm_disp *); + int (*init)(struct nvkm_disp *); + void (*fini)(struct nvkm_disp *); + void (*intr)(struct nvkm_disp *); + void (*intr_error)(struct nvkm_disp *, int chid); + + void (*super)(struct work_struct *); + + const struct nvkm_event_func *uevent; + + struct { + int (*cnt)(struct nvkm_disp *, unsigned long *mask); + int (*new)(struct nvkm_disp *, int id); + } wndw, head, dac, sor, pior; + + u16 ramht_size; + + const struct nvkm_sclass root; + + struct nvkm_disp_user { + struct nvkm_sclass base; + int (*ctor)(const struct nvkm_oclass *, void *argv, u32 argc, + struct nvkm_object **); + const struct nvkm_disp_chan_user *chan; + } user[]; +}; + +int nv50_disp_oneinit(struct nvkm_disp *); +int nv50_disp_init(struct nvkm_disp *); +void nv50_disp_fini(struct nvkm_disp *); +void nv50_disp_intr(struct nvkm_disp *); +extern const struct nvkm_enum nv50_disp_intr_error_type[]; +void nv50_disp_super(struct work_struct *); +void nv50_disp_super_1(struct nvkm_disp *); +void nv50_disp_super_1_0(struct nvkm_disp *, struct nvkm_head *); +void nv50_disp_super_2_0(struct nvkm_disp *, struct nvkm_head *); +void nv50_disp_super_2_1(struct nvkm_disp *, struct nvkm_head *); +void nv50_disp_super_2_2(struct nvkm_disp *, struct nvkm_head *); +void nv50_disp_super_3_0(struct nvkm_disp *, struct nvkm_head *); + +int gf119_disp_init(struct nvkm_disp *); +void gf119_disp_fini(struct nvkm_disp *); +void gf119_disp_intr(struct nvkm_disp *); +void gf119_disp_super(struct work_struct *); +void gf119_disp_intr_error(struct nvkm_disp *, int); + +void gv100_disp_fini(struct nvkm_disp *); +void gv100_disp_intr(struct nvkm_disp *); +void gv100_disp_super(struct work_struct *); +int gv100_disp_wndw_cnt(struct nvkm_disp *, unsigned long *); +int gv100_disp_caps_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); + +int tu102_disp_init(struct nvkm_disp *); + +void nv50_disp_dptmds_war_2(struct nvkm_disp *, struct dcb_output *); +void nv50_disp_dptmds_war_3(struct nvkm_disp *, struct dcb_output *); +void nv50_disp_update_sppll1(struct nvkm_disp *); + +extern const struct nvkm_event_func nv50_disp_chan_uevent; +void nv50_disp_chan_uevent_send(struct nvkm_disp *, int); + +extern const struct nvkm_event_func gf119_disp_chan_uevent; +extern const struct nvkm_event_func gv100_disp_chan_uevent; + +int nvkm_udisp_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +int nvkm_uconn_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +int nvkm_uoutp_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +int nvkm_uhead_new(const struct nvkm_oclass *, void *, u32, struct nvkm_object **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c new file mode 100644 index 0000000000..f5242a6722 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/tu102.c @@ -0,0 +1,236 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "chan.h" +#include "priv.h" +#include "head.h" +#include "ior.h" + +#include +#include + +#include + +void +tu102_sor_dp_vcpi(struct nvkm_ior *sor, int head, u8 slot, u8 slot_nr, u16 pbn, u16 aligned) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 hoff = head * 0x800; + + nvkm_mask(device, 0x61657c + hoff, 0xffffffff, (aligned << 16) | pbn); + nvkm_mask(device, 0x616578 + hoff, 0x00003f3f, (slot_nr << 8) | slot); +} + +static int +tu102_sor_dp_links(struct nvkm_ior *sor, struct nvkm_i2c_aux *aux) +{ + struct nvkm_device *device = sor->disp->engine.subdev.device; + const u32 soff = nv50_ior_base(sor); + const u32 loff = nv50_sor_link(sor); + u32 dpctrl = 0x00000000; + u32 clksor = 0x00000000; + + clksor |= sor->dp.bw << 18; + dpctrl |= ((1 << sor->dp.nr) - 1) << 16; + if (sor->dp.mst) + dpctrl |= 0x40000000; + if (sor->dp.ef) + dpctrl |= 0x00004000; + + nvkm_mask(device, 0x612300 + soff, 0x007c0000, clksor); + + /*XXX*/ + nvkm_msec(device, 40, NVKM_DELAY); + nvkm_mask(device, 0x612300 + soff, 0x00030000, 0x00010000); + nvkm_mask(device, 0x61c10c + loff, 0x00000003, 0x00000001); + + nvkm_mask(device, 0x61c10c + loff, 0x401f4000, dpctrl); + return 0; +} + +static const struct nvkm_ior_func_dp +tu102_sor_dp = { + .lanes = { 0, 1, 2, 3 }, + .links = tu102_sor_dp_links, + .power = g94_sor_dp_power, + .pattern = gm107_sor_dp_pattern, + .drive = gm200_sor_dp_drive, + .vcpi = tu102_sor_dp_vcpi, + .audio = gv100_sor_dp_audio, + .audio_sym = gv100_sor_dp_audio_sym, + .watermark = gv100_sor_dp_watermark, +}; + +static const struct nvkm_ior_func +tu102_sor = { + .route = { + .get = gm200_sor_route_get, + .set = gm200_sor_route_set, + }, + .state = gv100_sor_state, + .power = nv50_sor_power, + .clock = gf119_sor_clock, + .hdmi = &gv100_sor_hdmi, + .dp = &tu102_sor_dp, + .hda = &gv100_sor_hda, +}; + +static int +tu102_sor_new(struct nvkm_disp *disp, int id) +{ + struct nvkm_device *device = disp->engine.subdev.device; + u32 hda = nvkm_rd32(device, 0x08a15c); + + return nvkm_ior_new_(&tu102_sor, disp, SOR, id, hda & BIT(id)); +} + +int +tu102_disp_init(struct nvkm_disp *disp) +{ + struct nvkm_device *device = disp->engine.subdev.device; + struct nvkm_head *head; + int i, j; + u32 tmp; + + /* Claim ownership of display. */ + if (nvkm_rd32(device, 0x6254e8) & 0x00000002) { + nvkm_mask(device, 0x6254e8, 0x00000001, 0x00000000); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x6254e8) & 0x00000002)) + break; + ) < 0) + return -EBUSY; + } + + /* Lock pin capabilities. */ + tmp = 0x00000021; /*XXX*/ + nvkm_wr32(device, 0x640008, tmp); + + /* SOR capabilities. */ + for (i = 0; i < disp->sor.nr; i++) { + tmp = nvkm_rd32(device, 0x61c000 + (i * 0x800)); + nvkm_mask(device, 0x640000, 0x00000100 << i, 0x00000100 << i); + nvkm_wr32(device, 0x640144 + (i * 0x08), tmp); + } + + /* Head capabilities. */ + list_for_each_entry(head, &disp->heads, head) { + const int id = head->id; + + /* RG. */ + tmp = nvkm_rd32(device, 0x616300 + (id * 0x800)); + nvkm_wr32(device, 0x640048 + (id * 0x020), tmp); + + /* POSTCOMP. */ + for (j = 0; j < 5 * 4; j += 4) { + tmp = nvkm_rd32(device, 0x616140 + (id * 0x800) + j); + nvkm_wr32(device, 0x640680 + (id * 0x20) + j, tmp); + } + } + + /* Window capabilities. */ + for (i = 0; i < disp->wndw.nr; i++) { + nvkm_mask(device, 0x640004, 1 << i, 1 << i); + for (j = 0; j < 6 * 4; j += 4) { + tmp = nvkm_rd32(device, 0x630100 + (i * 0x800) + j); + nvkm_mask(device, 0x640780 + (i * 0x20) + j, 0xffffffff, tmp); + } + nvkm_mask(device, 0x64000c, 0x00000100, 0x00000100); + } + + /* IHUB capabilities. */ + for (i = 0; i < 3; i++) { + tmp = nvkm_rd32(device, 0x62e000 + (i * 0x04)); + nvkm_wr32(device, 0x640010 + (i * 0x04), tmp); + } + + nvkm_mask(device, 0x610078, 0x00000001, 0x00000001); + + /* Setup instance memory. */ + switch (nvkm_memory_target(disp->inst->memory)) { + case NVKM_MEM_TARGET_VRAM: tmp = 0x00000001; break; + case NVKM_MEM_TARGET_NCOH: tmp = 0x00000002; break; + case NVKM_MEM_TARGET_HOST: tmp = 0x00000003; break; + default: + break; + } + nvkm_wr32(device, 0x610010, 0x00000008 | tmp); + nvkm_wr32(device, 0x610014, disp->inst->addr >> 16); + + /* CTRL_DISP: AWAKEN, ERROR, SUPERVISOR[1-3]. */ + nvkm_wr32(device, 0x611cf0, 0x00000187); /* MSK. */ + nvkm_wr32(device, 0x611db0, 0x00000187); /* EN. */ + + /* EXC_OTHER: CURSn, CORE. */ + nvkm_wr32(device, 0x611cec, disp->head.mask << 16 | + 0x00000001); /* MSK. */ + nvkm_wr32(device, 0x611dac, 0x00000000); /* EN. */ + + /* EXC_WINIM. */ + nvkm_wr32(device, 0x611ce8, disp->wndw.mask); /* MSK. */ + nvkm_wr32(device, 0x611da8, 0x00000000); /* EN. */ + + /* EXC_WIN. */ + nvkm_wr32(device, 0x611ce4, disp->wndw.mask); /* MSK. */ + nvkm_wr32(device, 0x611da4, 0x00000000); /* EN. */ + + /* HEAD_TIMING(n): VBLANK. */ + list_for_each_entry(head, &disp->heads, head) { + const u32 hoff = head->id * 4; + nvkm_wr32(device, 0x611cc0 + hoff, 0x00000004); /* MSK. */ + nvkm_wr32(device, 0x611d80 + hoff, 0x00000000); /* EN. */ + } + + /* OR. */ + nvkm_wr32(device, 0x611cf4, 0x00000000); /* MSK. */ + nvkm_wr32(device, 0x611db4, 0x00000000); /* EN. */ + return 0; +} + +static const struct nvkm_disp_func +tu102_disp = { + .oneinit = nv50_disp_oneinit, + .init = tu102_disp_init, + .fini = gv100_disp_fini, + .intr = gv100_disp_intr, + .super = gv100_disp_super, + .uevent = &gv100_disp_chan_uevent, + .wndw = { .cnt = gv100_disp_wndw_cnt }, + .head = { .cnt = gv100_head_cnt, .new = gv100_head_new }, + .sor = { .cnt = gv100_sor_cnt, .new = tu102_sor_new }, + .ramht_size = 0x2000, + .root = { 0, 0,TU102_DISP }, + .user = { + {{-1,-1,GV100_DISP_CAPS }, gv100_disp_caps_new }, + {{ 0, 0,TU102_DISP_CURSOR }, nvkm_disp_chan_new, &gv100_disp_curs }, + {{ 0, 0,TU102_DISP_WINDOW_IMM_CHANNEL_DMA}, nvkm_disp_wndw_new, &gv100_disp_wimm }, + {{ 0, 0,TU102_DISP_CORE_CHANNEL_DMA }, nvkm_disp_core_new, &gv100_disp_core }, + {{ 0, 0,TU102_DISP_WINDOW_CHANNEL_DMA }, nvkm_disp_wndw_new, &gv100_disp_wndw }, + {} + }, +}; + +int +tu102_disp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_disp **pdisp) +{ + return nvkm_disp_new_(&tu102_disp, device, type, inst, pdisp); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/uconn.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uconn.c new file mode 100644 index 0000000000..3249e5c1c8 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uconn.c @@ -0,0 +1,213 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define nvkm_uconn(p) container_of((p), struct nvkm_conn, object) +#include "conn.h" +#include "outp.h" + +#include +#include +#include +#include + +#include + +static int +nvkm_uconn_uevent_aux(struct nvkm_object *object, u64 token, u32 bits) +{ + union nvif_conn_event_args args; + + args.v0.version = 0; + args.v0.types = 0; + if (bits & NVKM_I2C_PLUG) + args.v0.types |= NVIF_CONN_EVENT_V0_PLUG; + if (bits & NVKM_I2C_UNPLUG) + args.v0.types |= NVIF_CONN_EVENT_V0_UNPLUG; + if (bits & NVKM_I2C_IRQ) + args.v0.types |= NVIF_CONN_EVENT_V0_IRQ; + + return object->client->event(token, &args, sizeof(args.v0)); +} + +static int +nvkm_uconn_uevent_gpio(struct nvkm_object *object, u64 token, u32 bits) +{ + union nvif_conn_event_args args; + + args.v0.version = 0; + args.v0.types = 0; + if (bits & NVKM_GPIO_HI) + args.v0.types |= NVIF_CONN_EVENT_V0_PLUG; + if (bits & NVKM_GPIO_LO) + args.v0.types |= NVIF_CONN_EVENT_V0_UNPLUG; + + return object->client->event(token, &args, sizeof(args.v0)); +} + +static bool +nvkm_connector_is_dp_dms(u8 type) +{ + switch (type) { + case DCB_CONNECTOR_DMS59_DP0: + case DCB_CONNECTOR_DMS59_DP1: + return true; + default: + return false; + } +} + +static int +nvkm_uconn_uevent(struct nvkm_object *object, void *argv, u32 argc, struct nvkm_uevent *uevent) +{ + struct nvkm_conn *conn = nvkm_uconn(object); + struct nvkm_device *device = conn->disp->engine.subdev.device; + struct nvkm_outp *outp; + union nvif_conn_event_args *args = argv; + u64 bits = 0; + + if (!uevent) { + if (conn->info.hpd == DCB_GPIO_UNUSED) + return -ENOSYS; + return 0; + } + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + list_for_each_entry(outp, &conn->disp->outps, head) { + if (outp->info.connector == conn->index) + break; + } + + if (&outp->head == &conn->disp->outps) + return -EINVAL; + + if (outp->dp.aux && !outp->info.location) { + if (args->v0.types & NVIF_CONN_EVENT_V0_PLUG ) bits |= NVKM_I2C_PLUG; + if (args->v0.types & NVIF_CONN_EVENT_V0_UNPLUG) bits |= NVKM_I2C_UNPLUG; + if (args->v0.types & NVIF_CONN_EVENT_V0_IRQ ) bits |= NVKM_I2C_IRQ; + + return nvkm_uevent_add(uevent, &device->i2c->event, outp->dp.aux->id, bits, + nvkm_uconn_uevent_aux); + } + + if (args->v0.types & NVIF_CONN_EVENT_V0_PLUG ) bits |= NVKM_GPIO_HI; + if (args->v0.types & NVIF_CONN_EVENT_V0_UNPLUG) bits |= NVKM_GPIO_LO; + if (args->v0.types & NVIF_CONN_EVENT_V0_IRQ) { + /* TODO: support DP IRQ on ANX9805 and remove this hack. */ + if (!outp->info.location && !nvkm_connector_is_dp_dms(conn->info.type)) + return -EINVAL; + } + + return nvkm_uevent_add(uevent, &device->gpio->event, conn->info.hpd, bits, + nvkm_uconn_uevent_gpio); +} + +static int +nvkm_uconn_mthd_hpd_status(struct nvkm_conn *conn, void *argv, u32 argc) +{ + struct nvkm_gpio *gpio = conn->disp->engine.subdev.device->gpio; + union nvif_conn_hpd_status_args *args = argv; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + args->v0.support = gpio && conn->info.hpd != DCB_GPIO_UNUSED; + args->v0.present = 0; + + if (args->v0.support) { + int ret = nvkm_gpio_get(gpio, 0, DCB_GPIO_UNUSED, conn->info.hpd); + + if (WARN_ON(ret < 0)) { + args->v0.support = false; + return 0; + } + + args->v0.present = ret; + } + + return 0; +} + +static int +nvkm_uconn_mthd(struct nvkm_object *object, u32 mthd, void *argv, u32 argc) +{ + struct nvkm_conn *conn = nvkm_uconn(object); + + switch (mthd) { + case NVIF_CONN_V0_HPD_STATUS: return nvkm_uconn_mthd_hpd_status(conn, argv, argc); + default: + break; + } + + return -EINVAL; +} + +static void * +nvkm_uconn_dtor(struct nvkm_object *object) +{ + struct nvkm_conn *conn = nvkm_uconn(object); + struct nvkm_disp *disp = conn->disp; + + spin_lock(&disp->client.lock); + conn->object.func = NULL; + spin_unlock(&disp->client.lock); + return NULL; +} + +static const struct nvkm_object_func +nvkm_uconn = { + .dtor = nvkm_uconn_dtor, + .mthd = nvkm_uconn_mthd, + .uevent = nvkm_uconn_uevent, +}; + +int +nvkm_uconn_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + struct nvkm_conn *cont, *conn = NULL; + union nvif_conn_args *args = argv; + int ret; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + list_for_each_entry(cont, &disp->conns, head) { + if (cont->index == args->v0.id) { + conn = cont; + break; + } + } + + if (!conn) + return -EINVAL; + + ret = -EBUSY; + spin_lock(&disp->client.lock); + if (!conn->object.func) { + nvkm_object_ctor(&nvkm_uconn, oclass, &conn->object); + *pobject = &conn->object; + ret = 0; + } + spin_unlock(&disp->client.lock); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/udisp.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/udisp.c new file mode 100644 index 0000000000..0268d1d758 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/udisp.c @@ -0,0 +1,114 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "conn.h" +#include "head.h" +#include "outp.h" + +#include +#include + +static int +nvkm_udisp_sclass(struct nvkm_object *object, int index, struct nvkm_oclass *sclass) +{ + struct nvkm_disp *disp = nvkm_udisp(object); + + if (index-- == 0) { + sclass->base = (struct nvkm_sclass) { 0, 0, NVIF_CLASS_CONN }; + sclass->ctor = nvkm_uconn_new; + return 0; + } + + if (index-- == 0) { + sclass->base = (struct nvkm_sclass) { 0, 0, NVIF_CLASS_OUTP }; + sclass->ctor = nvkm_uoutp_new; + return 0; + } + + if (index-- == 0) { + sclass->base = (struct nvkm_sclass) { 0, 0, NVIF_CLASS_HEAD }; + sclass->ctor = nvkm_uhead_new; + return 0; + } + + if (disp->func->user[index].ctor) { + sclass->base = disp->func->user[index].base; + sclass->ctor = disp->func->user[index].ctor; + return 0; + } + + return -EINVAL; +} + +static void * +nvkm_udisp_dtor(struct nvkm_object *object) +{ + struct nvkm_disp *disp = nvkm_udisp(object); + + spin_lock(&disp->client.lock); + if (object == &disp->client.object) + disp->client.object.func = NULL; + spin_unlock(&disp->client.lock); + return NULL; +} + +static const struct nvkm_object_func +nvkm_udisp = { + .dtor = nvkm_udisp_dtor, + .sclass = nvkm_udisp_sclass, +}; + +int +nvkm_udisp_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_disp(oclass->engine); + struct nvkm_conn *conn; + struct nvkm_outp *outp; + struct nvkm_head *head; + union nvif_disp_args *args = argv; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + spin_lock(&disp->client.lock); + if (disp->client.object.func) { + spin_unlock(&disp->client.lock); + return -EBUSY; + } + nvkm_object_ctor(&nvkm_udisp, oclass, &disp->client.object); + *pobject = &disp->client.object; + spin_unlock(&disp->client.lock); + + args->v0.conn_mask = 0; + list_for_each_entry(conn, &disp->conns, head) + args->v0.conn_mask |= BIT(conn->index); + + args->v0.outp_mask = 0; + list_for_each_entry(outp, &disp->outps, head) + args->v0.outp_mask |= BIT(outp->index); + + args->v0.head_mask = 0; + list_for_each_entry(head, &disp->heads, head) + args->v0.head_mask |= BIT(head->id); + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/uhead.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uhead.c new file mode 100644 index 0000000000..f072cec160 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uhead.c @@ -0,0 +1,127 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define nvkm_uhead(p) container_of((p), struct nvkm_head, object) +#include "head.h" +#include + +#include + +#include + +static int +nvkm_uhead_uevent(struct nvkm_object *object, void *argv, u32 argc, struct nvkm_uevent *uevent) +{ + struct nvkm_head *head = nvkm_uhead(object); + union nvif_head_event_args *args = argv; + + if (!uevent) + return 0; + if (argc != sizeof(args->vn)) + return -ENOSYS; + + return nvkm_uevent_add(uevent, &head->disp->vblank, head->id, + NVKM_DISP_HEAD_EVENT_VBLANK, NULL); +} + +static int +nvkm_uhead_mthd_scanoutpos(struct nvkm_head *head, void *argv, u32 argc) +{ + union nvif_head_scanoutpos_args *args = argv; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + head->func->state(head, &head->arm); + args->v0.vtotal = head->arm.vtotal; + args->v0.vblanks = head->arm.vblanks; + args->v0.vblanke = head->arm.vblanke; + args->v0.htotal = head->arm.htotal; + args->v0.hblanks = head->arm.hblanks; + args->v0.hblanke = head->arm.hblanke; + + /* We don't support reading htotal/vtotal on pre-NV50 VGA, + * so we have to give up and trigger the timestamping + * fallback in the drm core. + */ + if (!args->v0.vtotal || !args->v0.htotal) + return -ENOTSUPP; + + args->v0.time[0] = ktime_to_ns(ktime_get()); + head->func->rgpos(head, &args->v0.hline, &args->v0.vline); + args->v0.time[1] = ktime_to_ns(ktime_get()); + return 0; +} + +static int +nvkm_uhead_mthd(struct nvkm_object *object, u32 mthd, void *argv, u32 argc) +{ + struct nvkm_head *head = nvkm_uhead(object); + + switch (mthd) { + case NVIF_HEAD_V0_SCANOUTPOS: return nvkm_uhead_mthd_scanoutpos(head, argv, argc); + default: + return -EINVAL; + } +} + +static void * +nvkm_uhead_dtor(struct nvkm_object *object) +{ + struct nvkm_head *head = nvkm_uhead(object); + struct nvkm_disp *disp = head->disp; + + spin_lock(&disp->client.lock); + head->object.func = NULL; + spin_unlock(&disp->client.lock); + return NULL; +} + +static const struct nvkm_object_func +nvkm_uhead = { + .dtor = nvkm_uhead_dtor, + .mthd = nvkm_uhead_mthd, + .uevent = nvkm_uhead_uevent, +}; + +int +nvkm_uhead_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + struct nvkm_head *head; + union nvif_head_args *args = argv; + int ret; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + if (!(head = nvkm_head_find(disp, args->v0.id))) + return -EINVAL; + + ret = -EBUSY; + spin_lock(&disp->client.lock); + if (!head->object.func) { + nvkm_object_ctor(&nvkm_uhead, oclass, &head->object); + *pobject = &head->object; + ret = 0; + } + spin_unlock(&disp->client.lock); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c new file mode 100644 index 0000000000..fc283a4a15 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/uoutp.c @@ -0,0 +1,379 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define nvkm_uoutp(p) container_of((p), struct nvkm_outp, object) +#include "outp.h" +#include "dp.h" +#include "head.h" +#include "ior.h" + +#include + +static int +nvkm_uoutp_mthd_dp_mst_vcpi(struct nvkm_outp *outp, void *argv, u32 argc) +{ + struct nvkm_ior *ior = outp->ior; + union nvif_outp_dp_mst_vcpi_args *args = argv; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + if (!ior->func->dp || !ior->func->dp->vcpi || !nvkm_head_find(outp->disp, args->v0.head)) + return -EINVAL; + + ior->func->dp->vcpi(ior, args->v0.head, args->v0.start_slot, args->v0.num_slots, + args->v0.pbn, args->v0.aligned_pbn); + return 0; +} + +static int +nvkm_uoutp_mthd_dp_retrain(struct nvkm_outp *outp, void *argv, u32 argc) +{ + union nvif_outp_dp_retrain_args *args = argv; + + if (argc != sizeof(args->vn)) + return -ENOSYS; + + if (!atomic_read(&outp->dp.lt.done)) + return 0; + + return outp->func->acquire(outp); +} + +static int +nvkm_uoutp_mthd_dp_aux_pwr(struct nvkm_outp *outp, void *argv, u32 argc) +{ + union nvif_outp_dp_aux_pwr_args *args = argv; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + outp->dp.enabled = !!args->v0.state; + nvkm_dp_enable(outp, outp->dp.enabled); + return 0; +} + +static int +nvkm_uoutp_mthd_hda_eld(struct nvkm_outp *outp, void *argv, u32 argc) +{ + struct nvkm_ior *ior = outp->ior; + union nvif_outp_hda_eld_args *args = argv; + + if (argc < sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + argc -= sizeof(args->v0); + + if (!ior->hda || !nvkm_head_find(outp->disp, args->v0.head)) + return -EINVAL; + if (argc > 0x60) + return -E2BIG; + + if (argc && args->v0.data[0]) { + if (outp->info.type == DCB_OUTPUT_DP) + ior->func->dp->audio(ior, args->v0.head, true); + ior->func->hda->hpd(ior, args->v0.head, true); + ior->func->hda->eld(ior, args->v0.head, args->v0.data, argc); + } else { + if (outp->info.type == DCB_OUTPUT_DP) + ior->func->dp->audio(ior, args->v0.head, false); + ior->func->hda->hpd(ior, args->v0.head, false); + } + + return 0; +} + +static int +nvkm_uoutp_mthd_infoframe(struct nvkm_outp *outp, void *argv, u32 argc) +{ + struct nvkm_ior *ior = outp->ior; + union nvif_outp_infoframe_args *args = argv; + ssize_t size = argc - sizeof(*args); + + if (argc < sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + if (!nvkm_head_find(outp->disp, args->v0.head)) + return -EINVAL; + + switch (ior->func->hdmi ? args->v0.type : 0xff) { + case NVIF_OUTP_INFOFRAME_V0_AVI: + ior->func->hdmi->infoframe_avi(ior, args->v0.head, &args->v0.data, size); + return 0; + case NVIF_OUTP_INFOFRAME_V0_VSI: + ior->func->hdmi->infoframe_vsi(ior, args->v0.head, &args->v0.data, size); + return 0; + default: + break; + } + + return -EINVAL; +} + +static int +nvkm_uoutp_mthd_release(struct nvkm_outp *outp, void *argv, u32 argc) +{ + struct nvkm_head *head = outp->asy.head; + struct nvkm_ior *ior = outp->ior; + union nvif_outp_release_args *args = argv; + + if (argc != sizeof(args->vn)) + return -ENOSYS; + + if (ior->func->hdmi && head) { + ior->func->hdmi->infoframe_avi(ior, head->id, NULL, 0); + ior->func->hdmi->infoframe_vsi(ior, head->id, NULL, 0); + ior->func->hdmi->ctrl(ior, head->id, false, 0, 0); + } + + nvkm_outp_release(outp, NVKM_OUTP_USER); + return 0; +} + +static int +nvkm_uoutp_mthd_acquire_dp(struct nvkm_outp *outp, u8 dpcd[DP_RECEIVER_CAP_SIZE], + u8 link_nr, u8 link_bw, bool hda, bool mst) +{ + int ret; + + ret = nvkm_outp_acquire(outp, NVKM_OUTP_USER, hda); + if (ret) + return ret; + + memcpy(outp->dp.dpcd, dpcd, sizeof(outp->dp.dpcd)); + outp->dp.lt.nr = link_nr; + outp->dp.lt.bw = link_bw; + outp->dp.lt.mst = mst; + return 0; +} + +static int +nvkm_uoutp_mthd_acquire_tmds(struct nvkm_outp *outp, u8 head, u8 hdmi, u8 hdmi_max_ac_packet, + u8 hdmi_rekey, u8 hdmi_scdc, u8 hdmi_hda) +{ + struct nvkm_ior *ior; + int ret; + + if (!(outp->asy.head = nvkm_head_find(outp->disp, head))) + return -EINVAL; + + ret = nvkm_outp_acquire(outp, NVKM_OUTP_USER, hdmi && hdmi_hda); + if (ret) + return ret; + + ior = outp->ior; + + if (hdmi) { + if (!ior->func->hdmi || + hdmi_max_ac_packet > 0x1f || hdmi_rekey > 0x7f || + (hdmi_scdc && !ior->func->hdmi->scdc)) { + nvkm_outp_release(outp, NVKM_OUTP_USER); + return -EINVAL; + } + + ior->func->hdmi->ctrl(ior, head, hdmi, hdmi_max_ac_packet, hdmi_rekey); + if (ior->func->hdmi->scdc) + ior->func->hdmi->scdc(ior, hdmi_scdc); + } + + return 0; +} + +static int +nvkm_uoutp_mthd_acquire_lvds(struct nvkm_outp *outp, bool dual, bool bpc8) +{ + if (outp->info.type != DCB_OUTPUT_LVDS) + return -EINVAL; + + outp->lvds.dual = dual; + outp->lvds.bpc8 = bpc8; + + return nvkm_outp_acquire(outp, NVKM_OUTP_USER, false); +} + +static int +nvkm_uoutp_mthd_acquire(struct nvkm_outp *outp, void *argv, u32 argc) +{ + union nvif_outp_acquire_args *args = argv; + int ret; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + if (outp->ior) + return -EBUSY; + + switch (args->v0.proto) { + case NVIF_OUTP_ACQUIRE_V0_RGB_CRT: + ret = nvkm_outp_acquire(outp, NVKM_OUTP_USER, false); + break; + case NVIF_OUTP_ACQUIRE_V0_TMDS: + ret = nvkm_uoutp_mthd_acquire_tmds(outp, args->v0.tmds.head, + args->v0.tmds.hdmi, + args->v0.tmds.hdmi_max_ac_packet, + args->v0.tmds.hdmi_rekey, + args->v0.tmds.hdmi_scdc, + args->v0.tmds.hdmi_hda); + break; + case NVIF_OUTP_ACQUIRE_V0_LVDS: + ret = nvkm_uoutp_mthd_acquire_lvds(outp, args->v0.lvds.dual, args->v0.lvds.bpc8); + break; + case NVIF_OUTP_ACQUIRE_V0_DP: + ret = nvkm_uoutp_mthd_acquire_dp(outp, args->v0.dp.dpcd, + args->v0.dp.link_nr, + args->v0.dp.link_bw, + args->v0.dp.hda != 0, + args->v0.dp.mst != 0); + break; + default: + ret = -EINVAL; + break; + } + + if (ret) + return ret; + + args->v0.or = outp->ior->id; + args->v0.link = outp->ior->asy.link; + return 0; +} + +static int +nvkm_uoutp_mthd_load_detect(struct nvkm_outp *outp, void *argv, u32 argc) +{ + union nvif_outp_load_detect_args *args = argv; + int ret; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + ret = nvkm_outp_acquire(outp, NVKM_OUTP_PRIV, false); + if (ret == 0) { + if (outp->ior->func->sense) { + ret = outp->ior->func->sense(outp->ior, args->v0.data); + args->v0.load = ret < 0 ? 0 : ret; + } else { + ret = -EINVAL; + } + nvkm_outp_release(outp, NVKM_OUTP_PRIV); + } + + return ret; +} + +static int +nvkm_uoutp_mthd_acquired(struct nvkm_outp *outp, u32 mthd, void *argv, u32 argc) +{ + switch (mthd) { + case NVIF_OUTP_V0_RELEASE : return nvkm_uoutp_mthd_release (outp, argv, argc); + case NVIF_OUTP_V0_INFOFRAME : return nvkm_uoutp_mthd_infoframe (outp, argv, argc); + case NVIF_OUTP_V0_HDA_ELD : return nvkm_uoutp_mthd_hda_eld (outp, argv, argc); + case NVIF_OUTP_V0_DP_RETRAIN : return nvkm_uoutp_mthd_dp_retrain (outp, argv, argc); + case NVIF_OUTP_V0_DP_MST_VCPI: return nvkm_uoutp_mthd_dp_mst_vcpi(outp, argv, argc); + default: + break; + } + + return -EINVAL; +} + +static int +nvkm_uoutp_mthd_noacquire(struct nvkm_outp *outp, u32 mthd, void *argv, u32 argc) +{ + switch (mthd) { + case NVIF_OUTP_V0_LOAD_DETECT: return nvkm_uoutp_mthd_load_detect(outp, argv, argc); + case NVIF_OUTP_V0_ACQUIRE : return nvkm_uoutp_mthd_acquire (outp, argv, argc); + case NVIF_OUTP_V0_DP_AUX_PWR : return nvkm_uoutp_mthd_dp_aux_pwr (outp, argv, argc); + default: + break; + } + + return 1; +} + +static int +nvkm_uoutp_mthd(struct nvkm_object *object, u32 mthd, void *argv, u32 argc) +{ + struct nvkm_outp *outp = nvkm_uoutp(object); + struct nvkm_disp *disp = outp->disp; + int ret; + + mutex_lock(&disp->super.mutex); + + ret = nvkm_uoutp_mthd_noacquire(outp, mthd, argv, argc); + if (ret <= 0) + goto done; + + if (outp->ior) + ret = nvkm_uoutp_mthd_acquired(outp, mthd, argv, argc); + else + ret = -EIO; + +done: + mutex_unlock(&disp->super.mutex); + return ret; +} + +static void * +nvkm_uoutp_dtor(struct nvkm_object *object) +{ + struct nvkm_outp *outp = nvkm_uoutp(object); + struct nvkm_disp *disp = outp->disp; + + spin_lock(&disp->client.lock); + outp->object.func = NULL; + spin_unlock(&disp->client.lock); + return NULL; +} + +static const struct nvkm_object_func +nvkm_uoutp = { + .dtor = nvkm_uoutp_dtor, + .mthd = nvkm_uoutp_mthd, +}; + +int +nvkm_uoutp_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = nvkm_udisp(oclass->parent); + struct nvkm_outp *outt, *outp = NULL; + union nvif_outp_args *args = argv; + int ret; + + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + list_for_each_entry(outt, &disp->outps, head) { + if (outt->index == args->v0.id) { + outp = outt; + break; + } + } + + if (!outp) + return -EINVAL; + + ret = -EBUSY; + spin_lock(&disp->client.lock); + if (!outp->object.func) { + nvkm_object_ctor(&nvkm_uoutp, oclass, &outp->object); + *pobject = &outp->object; + ret = 0; + } + spin_unlock(&disp->client.lock); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/vga.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/vga.c new file mode 100644 index 0000000000..8bff95c634 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/vga.c @@ -0,0 +1,205 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include + +u8 +nvkm_rdport(struct nvkm_device *device, int head, u16 port) +{ + if (device->card_type >= NV_50) + return nvkm_rd08(device, 0x601000 + port); + + if (port == 0x03c0 || port == 0x03c1 || /* AR */ + port == 0x03c2 || port == 0x03da || /* INP0 */ + port == 0x03d4 || port == 0x03d5) /* CR */ + return nvkm_rd08(device, 0x601000 + (head * 0x2000) + port); + + if (port == 0x03c2 || port == 0x03cc || /* MISC */ + port == 0x03c4 || port == 0x03c5 || /* SR */ + port == 0x03ce || port == 0x03cf) { /* GR */ + if (device->card_type < NV_40) + head = 0; /* CR44 selects head */ + return nvkm_rd08(device, 0x0c0000 + (head * 0x2000) + port); + } + + return 0x00; +} + +void +nvkm_wrport(struct nvkm_device *device, int head, u16 port, u8 data) +{ + if (device->card_type >= NV_50) + nvkm_wr08(device, 0x601000 + port, data); + else + if (port == 0x03c0 || port == 0x03c1 || /* AR */ + port == 0x03c2 || port == 0x03da || /* INP0 */ + port == 0x03d4 || port == 0x03d5) /* CR */ + nvkm_wr08(device, 0x601000 + (head * 0x2000) + port, data); + else + if (port == 0x03c2 || port == 0x03cc || /* MISC */ + port == 0x03c4 || port == 0x03c5 || /* SR */ + port == 0x03ce || port == 0x03cf) { /* GR */ + if (device->card_type < NV_40) + head = 0; /* CR44 selects head */ + nvkm_wr08(device, 0x0c0000 + (head * 0x2000) + port, data); + } +} + +u8 +nvkm_rdvgas(struct nvkm_device *device, int head, u8 index) +{ + nvkm_wrport(device, head, 0x03c4, index); + return nvkm_rdport(device, head, 0x03c5); +} + +void +nvkm_wrvgas(struct nvkm_device *device, int head, u8 index, u8 value) +{ + nvkm_wrport(device, head, 0x03c4, index); + nvkm_wrport(device, head, 0x03c5, value); +} + +u8 +nvkm_rdvgag(struct nvkm_device *device, int head, u8 index) +{ + nvkm_wrport(device, head, 0x03ce, index); + return nvkm_rdport(device, head, 0x03cf); +} + +void +nvkm_wrvgag(struct nvkm_device *device, int head, u8 index, u8 value) +{ + nvkm_wrport(device, head, 0x03ce, index); + nvkm_wrport(device, head, 0x03cf, value); +} + +u8 +nvkm_rdvgac(struct nvkm_device *device, int head, u8 index) +{ + nvkm_wrport(device, head, 0x03d4, index); + return nvkm_rdport(device, head, 0x03d5); +} + +void +nvkm_wrvgac(struct nvkm_device *device, int head, u8 index, u8 value) +{ + nvkm_wrport(device, head, 0x03d4, index); + nvkm_wrport(device, head, 0x03d5, value); +} + +u8 +nvkm_rdvgai(struct nvkm_device *device, int head, u16 port, u8 index) +{ + if (port == 0x03c4) return nvkm_rdvgas(device, head, index); + if (port == 0x03ce) return nvkm_rdvgag(device, head, index); + if (port == 0x03d4) return nvkm_rdvgac(device, head, index); + return 0x00; +} + +void +nvkm_wrvgai(struct nvkm_device *device, int head, u16 port, u8 index, u8 value) +{ + if (port == 0x03c4) nvkm_wrvgas(device, head, index, value); + else if (port == 0x03ce) nvkm_wrvgag(device, head, index, value); + else if (port == 0x03d4) nvkm_wrvgac(device, head, index, value); +} + +bool +nvkm_lockvgac(struct nvkm_device *device, bool lock) +{ + bool locked = !nvkm_rdvgac(device, 0, 0x1f); + u8 data = lock ? 0x99 : 0x57; + if (device->card_type < NV_50) + nvkm_wrvgac(device, 0, 0x1f, data); + else + nvkm_wrvgac(device, 0, 0x3f, data); + if (device->chipset == 0x11) { + if (!(nvkm_rd32(device, 0x001084) & 0x10000000)) + nvkm_wrvgac(device, 1, 0x1f, data); + } + return locked; +} + +/* CR44 takes values 0 (head A), 3 (head B) and 4 (heads tied) + * it affects only the 8 bit vga io regs, which we access using mmio at + * 0xc{0,2}3c*, 0x60{1,3}3*, and 0x68{1,3}3d* + * in general, the set value of cr44 does not matter: reg access works as + * expected and values can be set for the appropriate head by using a 0x2000 + * offset as required + * however: + * a) pre nv40, the head B range of PRMVIO regs at 0xc23c* was not exposed and + * cr44 must be set to 0 or 3 for accessing values on the correct head + * through the common 0xc03c* addresses + * b) in tied mode (4) head B is programmed to the values set on head A, and + * access using the head B addresses can have strange results, ergo we leave + * tied mode in init once we know to what cr44 should be restored on exit + * + * the owner parameter is slightly abused: + * 0 and 1 are treated as head values and so the set value is (owner * 3) + * other values are treated as literal values to set + */ +u8 +nvkm_rdvgaowner(struct nvkm_device *device) +{ + if (device->card_type < NV_50) { + if (device->chipset == 0x11) { + u32 tied = nvkm_rd32(device, 0x001084) & 0x10000000; + if (tied == 0) { + u8 slA = nvkm_rdvgac(device, 0, 0x28) & 0x80; + u8 tvA = nvkm_rdvgac(device, 0, 0x33) & 0x01; + u8 slB = nvkm_rdvgac(device, 1, 0x28) & 0x80; + u8 tvB = nvkm_rdvgac(device, 1, 0x33) & 0x01; + if (slA && !tvA) return 0x00; + if (slB && !tvB) return 0x03; + if (slA) return 0x00; + if (slB) return 0x03; + return 0x00; + } + return 0x04; + } + + return nvkm_rdvgac(device, 0, 0x44); + } + + return 0x00; +} + +void +nvkm_wrvgaowner(struct nvkm_device *device, u8 select) +{ + if (device->card_type < NV_50) { + u8 owner = (select == 1) ? 3 : select; + if (device->chipset == 0x11) { + /* workaround hw lockup bug */ + nvkm_rdvgac(device, 0, 0x1f); + nvkm_rdvgac(device, 1, 0x1f); + } + + nvkm_wrvgac(device, 0, 0x44, owner); + + if (device->chipset == 0x11) { + nvkm_wrvgac(device, 0, 0x2e, owner); + nvkm_wrvgac(device, 0, 0x2e, owner); + } + } +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/dma/Kbuild new file mode 100644 index 0000000000..a0e551b92c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/Kbuild @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/dma/base.o +nvkm-y += nvkm/engine/dma/nv04.o +nvkm-y += nvkm/engine/dma/nv50.o +nvkm-y += nvkm/engine/dma/gf100.o +nvkm-y += nvkm/engine/dma/gf119.o +nvkm-y += nvkm/engine/dma/gv100.o + +nvkm-y += nvkm/engine/dma/user.o +nvkm-y += nvkm/engine/dma/usernv04.o +nvkm-y += nvkm/engine/dma/usernv50.o +nvkm-y += nvkm/engine/dma/usergf100.o +nvkm-y += nvkm/engine/dma/usergf119.o +nvkm-y += nvkm/engine/dma/usergv100.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c new file mode 100644 index 0000000000..425cde35f1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/base.c @@ -0,0 +1,116 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include +#include + +#include + +static int +nvkm_dma_oclass_new(struct nvkm_device *device, + const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + struct nvkm_dma *dma = nvkm_dma(oclass->engine); + struct nvkm_dmaobj *dmaobj = NULL; + int ret; + + ret = dma->func->class_new(dma, oclass, data, size, &dmaobj); + if (dmaobj) + *pobject = &dmaobj->object; + return ret; +} + +static const struct nvkm_device_oclass +nvkm_dma_oclass_base = { + .ctor = nvkm_dma_oclass_new, +}; + +static int +nvkm_dma_oclass_fifo_new(const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + return nvkm_dma_oclass_new(oclass->engine->subdev.device, + oclass, data, size, pobject); +} + +static const struct nvkm_sclass +nvkm_dma_sclass[] = { + { 0, 0, NV_DMA_FROM_MEMORY, NULL, nvkm_dma_oclass_fifo_new }, + { 0, 0, NV_DMA_TO_MEMORY, NULL, nvkm_dma_oclass_fifo_new }, + { 0, 0, NV_DMA_IN_MEMORY, NULL, nvkm_dma_oclass_fifo_new }, +}; + +static int +nvkm_dma_oclass_base_get(struct nvkm_oclass *sclass, int index, + const struct nvkm_device_oclass **class) +{ + const int count = ARRAY_SIZE(nvkm_dma_sclass); + if (index < count) { + const struct nvkm_sclass *oclass = &nvkm_dma_sclass[index]; + sclass->base = oclass[0]; + sclass->engn = oclass; + *class = &nvkm_dma_oclass_base; + return index; + } + return count; +} + +static int +nvkm_dma_oclass_fifo_get(struct nvkm_oclass *oclass, int index) +{ + const int count = ARRAY_SIZE(nvkm_dma_sclass); + if (index < count) { + oclass->base = nvkm_dma_sclass[index]; + return index; + } + return count; +} + +static void * +nvkm_dma_dtor(struct nvkm_engine *engine) +{ + return nvkm_dma(engine); +} + +static const struct nvkm_engine_func +nvkm_dma = { + .dtor = nvkm_dma_dtor, + .base.sclass = nvkm_dma_oclass_base_get, + .fifo.sclass = nvkm_dma_oclass_fifo_get, +}; + +int +nvkm_dma_new_(const struct nvkm_dma_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_dma **pdma) +{ + struct nvkm_dma *dma; + + if (!(dma = *pdma = kzalloc(sizeof(*dma), GFP_KERNEL))) + return -ENOMEM; + dma->func = func; + + return nvkm_engine_ctor(&nvkm_dma, device, type, inst, true, &dma->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf100.c new file mode 100644 index 0000000000..99a1e07fa2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf100.c @@ -0,0 +1,37 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "user.h" + +static const struct nvkm_dma_func +gf100_dma = { + .class_new = gf100_dmaobj_new, +}; + +int +gf100_dma_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_dma **pdma) +{ + return nvkm_dma_new_(&gf100_dma, device, type, inst, pdma); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf119.c new file mode 100644 index 0000000000..fd1d1fc22d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gf119.c @@ -0,0 +1,37 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "user.h" + +static const struct nvkm_dma_func +gf119_dma = { + .class_new = gf119_dmaobj_new, +}; + +int +gf119_dma_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_dma **pdma) +{ + return nvkm_dma_new_(&gf119_dma, device, type, inst, pdma); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/gv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gv100.c new file mode 100644 index 0000000000..a5af0df306 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/gv100.c @@ -0,0 +1,35 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "user.h" + +static const struct nvkm_dma_func +gv100_dma = { + .class_new = gv100_dmaobj_new, +}; + +int +gv100_dma_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_dma **pdma) +{ + return nvkm_dma_new_(&gv100_dma, device, type, inst, pdma); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv04.c new file mode 100644 index 0000000000..ea5a889f60 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv04.c @@ -0,0 +1,37 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "user.h" + +static const struct nvkm_dma_func +nv04_dma = { + .class_new = nv04_dmaobj_new, +}; + +int +nv04_dma_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_dma **pdma) +{ + return nvkm_dma_new_(&nv04_dma, device, type, inst, pdma); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv50.c new file mode 100644 index 0000000000..6e8f796600 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/nv50.c @@ -0,0 +1,37 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "user.h" + +static const struct nvkm_dma_func +nv50_dma = { + .class_new = nv50_dmaobj_new, +}; + +int +nv50_dma_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_dma **pdma) +{ + return nvkm_dma_new_(&nv50_dma, device, type, inst, pdma); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/dma/priv.h new file mode 100644 index 0000000000..d403bedb48 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/priv.h @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DMA_PRIV_H__ +#define __NVKM_DMA_PRIV_H__ +#define nvkm_dma(p) container_of((p), struct nvkm_dma, engine) +#include + +struct nvkm_dmaobj_func { + int (*bind)(struct nvkm_dmaobj *, struct nvkm_gpuobj *, int align, + struct nvkm_gpuobj **); +}; + +int nvkm_dma_new_(const struct nvkm_dma_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_dma **); + +struct nvkm_dma_func { + int (*class_new)(struct nvkm_dma *, const struct nvkm_oclass *, + void *data, u32 size, struct nvkm_dmaobj **); +}; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c new file mode 100644 index 0000000000..797131ed7d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.c @@ -0,0 +1,138 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "user.h" + +#include +#include +#include + +#include +#include + +static const struct nvkm_object_func nvkm_dmaobj_func; +struct nvkm_dmaobj * +nvkm_dmaobj_search(struct nvkm_client *client, u64 handle) +{ + struct nvkm_object *object; + + object = nvkm_object_search(client, handle, &nvkm_dmaobj_func); + if (IS_ERR(object)) + return (void *)object; + + return nvkm_dmaobj(object); +} + +static int +nvkm_dmaobj_bind(struct nvkm_object *base, struct nvkm_gpuobj *gpuobj, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nvkm_dmaobj *dmaobj = nvkm_dmaobj(base); + return dmaobj->func->bind(dmaobj, gpuobj, align, pgpuobj); +} + +static void * +nvkm_dmaobj_dtor(struct nvkm_object *base) +{ + return nvkm_dmaobj(base); +} + +static const struct nvkm_object_func +nvkm_dmaobj_func = { + .dtor = nvkm_dmaobj_dtor, + .bind = nvkm_dmaobj_bind, +}; + +int +nvkm_dmaobj_ctor(const struct nvkm_dmaobj_func *func, struct nvkm_dma *dma, + const struct nvkm_oclass *oclass, void **pdata, u32 *psize, + struct nvkm_dmaobj *dmaobj) +{ + union { + struct nv_dma_v0 v0; + } *args = *pdata; + struct nvkm_object *parent = oclass->parent; + void *data = *pdata; + u32 size = *psize; + int ret = -ENOSYS; + + nvkm_object_ctor(&nvkm_dmaobj_func, oclass, &dmaobj->object); + dmaobj->func = func; + dmaobj->dma = dma; + + nvif_ioctl(parent, "create dma size %d\n", *psize); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, true))) { + nvif_ioctl(parent, "create dma vers %d target %d access %d " + "start %016llx limit %016llx\n", + args->v0.version, args->v0.target, args->v0.access, + args->v0.start, args->v0.limit); + dmaobj->target = args->v0.target; + dmaobj->access = args->v0.access; + dmaobj->start = args->v0.start; + dmaobj->limit = args->v0.limit; + } else + return ret; + + *pdata = data; + *psize = size; + + if (dmaobj->start > dmaobj->limit) + return -EINVAL; + + switch (dmaobj->target) { + case NV_DMA_V0_TARGET_VM: + dmaobj->target = NV_MEM_TARGET_VM; + break; + case NV_DMA_V0_TARGET_VRAM: + dmaobj->target = NV_MEM_TARGET_VRAM; + break; + case NV_DMA_V0_TARGET_PCI: + dmaobj->target = NV_MEM_TARGET_PCI; + break; + case NV_DMA_V0_TARGET_PCI_US: + case NV_DMA_V0_TARGET_AGP: + dmaobj->target = NV_MEM_TARGET_PCI_NOSNOOP; + break; + default: + return -EINVAL; + } + + switch (dmaobj->access) { + case NV_DMA_V0_ACCESS_VM: + dmaobj->access = NV_MEM_ACCESS_VM; + break; + case NV_DMA_V0_ACCESS_RD: + dmaobj->access = NV_MEM_ACCESS_RO; + break; + case NV_DMA_V0_ACCESS_WR: + dmaobj->access = NV_MEM_ACCESS_WO; + break; + case NV_DMA_V0_ACCESS_RDWR: + dmaobj->access = NV_MEM_ACCESS_RW; + break; + default: + return -EINVAL; + } + + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.h b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.h new file mode 100644 index 0000000000..9c72ee214b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/user.h @@ -0,0 +1,21 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_DMA_USER_H__ +#define __NVKM_DMA_USER_H__ +#define nvkm_dmaobj(p) container_of((p), struct nvkm_dmaobj, object) +#include "priv.h" + +int nvkm_dmaobj_ctor(const struct nvkm_dmaobj_func *, struct nvkm_dma *, + const struct nvkm_oclass *, void **data, u32 *size, + struct nvkm_dmaobj *); + +int nv04_dmaobj_new(struct nvkm_dma *, const struct nvkm_oclass *, void *, u32, + struct nvkm_dmaobj **); +int nv50_dmaobj_new(struct nvkm_dma *, const struct nvkm_oclass *, void *, u32, + struct nvkm_dmaobj **); +int gf100_dmaobj_new(struct nvkm_dma *, const struct nvkm_oclass *, void *, u32, + struct nvkm_dmaobj **); +int gf119_dmaobj_new(struct nvkm_dma *, const struct nvkm_oclass *, void *, u32, + struct nvkm_dmaobj **); +int gv100_dmaobj_new(struct nvkm_dma *, const struct nvkm_oclass *, void *, u32, + struct nvkm_dmaobj **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf100.c new file mode 100644 index 0000000000..ef7ac36010 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf100.c @@ -0,0 +1,150 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define gf100_dmaobj(p) container_of((p), struct gf100_dmaobj, base) +#include "user.h" + +#include +#include +#include + +#include +#include + +struct gf100_dmaobj { + struct nvkm_dmaobj base; + u32 flags0; + u32 flags5; +}; + +static int +gf100_dmaobj_bind(struct nvkm_dmaobj *base, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct gf100_dmaobj *dmaobj = gf100_dmaobj(base); + struct nvkm_device *device = dmaobj->base.dma->engine.subdev.device; + int ret; + + ret = nvkm_gpuobj_new(device, 24, align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, dmaobj->flags0); + nvkm_wo32(*pgpuobj, 0x04, lower_32_bits(dmaobj->base.limit)); + nvkm_wo32(*pgpuobj, 0x08, lower_32_bits(dmaobj->base.start)); + nvkm_wo32(*pgpuobj, 0x0c, upper_32_bits(dmaobj->base.limit) << 24 | + upper_32_bits(dmaobj->base.start)); + nvkm_wo32(*pgpuobj, 0x10, 0x00000000); + nvkm_wo32(*pgpuobj, 0x14, dmaobj->flags5); + nvkm_done(*pgpuobj); + } + + return ret; +} + +static const struct nvkm_dmaobj_func +gf100_dmaobj_func = { + .bind = gf100_dmaobj_bind, +}; + +int +gf100_dmaobj_new(struct nvkm_dma *dma, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_dmaobj **pdmaobj) +{ + union { + struct gf100_dma_v0 v0; + } *args; + struct nvkm_object *parent = oclass->parent; + struct gf100_dmaobj *dmaobj; + u32 kind, user, unkn; + int ret; + + if (!(dmaobj = kzalloc(sizeof(*dmaobj), GFP_KERNEL))) + return -ENOMEM; + *pdmaobj = &dmaobj->base; + + ret = nvkm_dmaobj_ctor(&gf100_dmaobj_func, dma, oclass, + &data, &size, &dmaobj->base); + if (ret) + return ret; + + ret = -ENOSYS; + args = data; + + nvif_ioctl(parent, "create gf100 dma size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, + "create gf100 dma vers %d priv %d kind %02x\n", + args->v0.version, args->v0.priv, args->v0.kind); + kind = args->v0.kind; + user = args->v0.priv; + unkn = 0; + } else + if (size == 0) { + if (dmaobj->base.target != NV_MEM_TARGET_VM) { + kind = GF100_DMA_V0_KIND_PITCH; + user = GF100_DMA_V0_PRIV_US; + unkn = 2; + } else { + kind = GF100_DMA_V0_KIND_VM; + user = GF100_DMA_V0_PRIV_VM; + unkn = 0; + } + } else + return ret; + + if (user > 2) + return -EINVAL; + dmaobj->flags0 |= (kind << 22) | (user << 20) | oclass->base.oclass; + dmaobj->flags5 |= (unkn << 16); + + switch (dmaobj->base.target) { + case NV_MEM_TARGET_VM: + dmaobj->flags0 |= 0x00000000; + break; + case NV_MEM_TARGET_VRAM: + dmaobj->flags0 |= 0x00010000; + break; + case NV_MEM_TARGET_PCI: + dmaobj->flags0 |= 0x00020000; + break; + case NV_MEM_TARGET_PCI_NOSNOOP: + dmaobj->flags0 |= 0x00030000; + break; + default: + return -EINVAL; + } + + switch (dmaobj->base.access) { + case NV_MEM_ACCESS_VM: + break; + case NV_MEM_ACCESS_RO: + dmaobj->flags0 |= 0x00040000; + break; + case NV_MEM_ACCESS_WO: + case NV_MEM_ACCESS_RW: + dmaobj->flags0 |= 0x00080000; + break; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf119.c new file mode 100644 index 0000000000..c068cee345 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergf119.c @@ -0,0 +1,132 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define gf119_dmaobj(p) container_of((p), struct gf119_dmaobj, base) +#include "user.h" + +#include +#include +#include + +#include +#include + +struct gf119_dmaobj { + struct nvkm_dmaobj base; + u32 flags0; +}; + +static int +gf119_dmaobj_bind(struct nvkm_dmaobj *base, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct gf119_dmaobj *dmaobj = gf119_dmaobj(base); + struct nvkm_device *device = dmaobj->base.dma->engine.subdev.device; + int ret; + + ret = nvkm_gpuobj_new(device, 24, align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, dmaobj->flags0); + nvkm_wo32(*pgpuobj, 0x04, dmaobj->base.start >> 8); + nvkm_wo32(*pgpuobj, 0x08, dmaobj->base.limit >> 8); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_wo32(*pgpuobj, 0x10, 0x00000000); + nvkm_wo32(*pgpuobj, 0x14, 0x00000000); + nvkm_done(*pgpuobj); + } + + return ret; +} + +static const struct nvkm_dmaobj_func +gf119_dmaobj_func = { + .bind = gf119_dmaobj_bind, +}; + +int +gf119_dmaobj_new(struct nvkm_dma *dma, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_dmaobj **pdmaobj) +{ + union { + struct gf119_dma_v0 v0; + } *args; + struct nvkm_object *parent = oclass->parent; + struct gf119_dmaobj *dmaobj; + u32 kind, page; + int ret; + + if (!(dmaobj = kzalloc(sizeof(*dmaobj), GFP_KERNEL))) + return -ENOMEM; + *pdmaobj = &dmaobj->base; + + ret = nvkm_dmaobj_ctor(&gf119_dmaobj_func, dma, oclass, + &data, &size, &dmaobj->base); + if (ret) + return ret; + + ret = -ENOSYS; + args = data; + + nvif_ioctl(parent, "create gf119 dma size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, + "create gf100 dma vers %d page %d kind %02x\n", + args->v0.version, args->v0.page, args->v0.kind); + kind = args->v0.kind; + page = args->v0.page; + } else + if (size == 0) { + if (dmaobj->base.target != NV_MEM_TARGET_VM) { + kind = GF119_DMA_V0_KIND_PITCH; + page = GF119_DMA_V0_PAGE_SP; + } else { + kind = GF119_DMA_V0_KIND_VM; + page = GF119_DMA_V0_PAGE_LP; + } + } else + return ret; + + if (page > 1) + return -EINVAL; + dmaobj->flags0 = (kind << 20) | (page << 6); + + switch (dmaobj->base.target) { + case NV_MEM_TARGET_VRAM: + dmaobj->flags0 |= 0x00000009; + break; + case NV_MEM_TARGET_VM: + case NV_MEM_TARGET_PCI: + case NV_MEM_TARGET_PCI_NOSNOOP: + /* XXX: don't currently know how to construct a real one + * of these. we only use them to represent pushbufs + * on these chipsets, and the classes that use them + * deal with the target themselves. + */ + break; + default: + return -EINVAL; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergv100.c new file mode 100644 index 0000000000..39eba9fc82 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usergv100.c @@ -0,0 +1,119 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define gv100_dmaobj(p) container_of((p), struct gv100_dmaobj, base) +#include "user.h" + +#include +#include +#include + +#include +#include + +struct gv100_dmaobj { + struct nvkm_dmaobj base; + u32 flags0; +}; + +static int +gv100_dmaobj_bind(struct nvkm_dmaobj *base, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct gv100_dmaobj *dmaobj = gv100_dmaobj(base); + struct nvkm_device *device = dmaobj->base.dma->engine.subdev.device; + u64 start = dmaobj->base.start >> 8; + u64 limit = dmaobj->base.limit >> 8; + int ret; + + ret = nvkm_gpuobj_new(device, 24, align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, dmaobj->flags0); + nvkm_wo32(*pgpuobj, 0x04, lower_32_bits(start)); + nvkm_wo32(*pgpuobj, 0x08, upper_32_bits(start)); + nvkm_wo32(*pgpuobj, 0x0c, lower_32_bits(limit)); + nvkm_wo32(*pgpuobj, 0x10, upper_32_bits(limit)); + nvkm_done(*pgpuobj); + } + + return ret; +} + +static const struct nvkm_dmaobj_func +gv100_dmaobj_func = { + .bind = gv100_dmaobj_bind, +}; + +int +gv100_dmaobj_new(struct nvkm_dma *dma, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_dmaobj **pdmaobj) +{ + union { + struct gf119_dma_v0 v0; + } *args; + struct nvkm_object *parent = oclass->parent; + struct gv100_dmaobj *dmaobj; + u32 kind, page; + int ret; + + if (!(dmaobj = kzalloc(sizeof(*dmaobj), GFP_KERNEL))) + return -ENOMEM; + *pdmaobj = &dmaobj->base; + + ret = nvkm_dmaobj_ctor(&gv100_dmaobj_func, dma, oclass, + &data, &size, &dmaobj->base); + if (ret) + return ret; + + ret = -ENOSYS; + args = data; + + nvif_ioctl(parent, "create gv100 dma size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, + "create gv100 dma vers %d page %d kind %02x\n", + args->v0.version, args->v0.page, args->v0.kind); + kind = args->v0.kind != 0; + page = args->v0.page != 0; + } else + if (size == 0) { + kind = 0; + page = GF119_DMA_V0_PAGE_SP; + } else + return ret; + + if (kind) + dmaobj->flags0 |= 0x00100000; + if (page) + dmaobj->flags0 |= 0x00000040; + dmaobj->flags0 |= 0x00000004; /* rw */ + + switch (dmaobj->base.target) { + case NV_MEM_TARGET_VRAM : dmaobj->flags0 |= 0x00000001; break; + case NV_MEM_TARGET_PCI : dmaobj->flags0 |= 0x00000002; break; + case NV_MEM_TARGET_PCI_NOSNOOP: dmaobj->flags0 |= 0x00000003; break; + default: + return -EINVAL; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv04.c new file mode 100644 index 0000000000..5159d5df20 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv04.c @@ -0,0 +1,134 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define nv04_dmaobj(p) container_of((p), struct nv04_dmaobj, base) +#include "user.h" + +#include +#include +#include + +#include + +struct nv04_dmaobj { + struct nvkm_dmaobj base; + bool clone; + u32 flags0; + u32 flags2; +}; + +static int +nv04_dmaobj_bind(struct nvkm_dmaobj *base, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nv04_dmaobj *dmaobj = nv04_dmaobj(base); + struct nvkm_device *device = dmaobj->base.dma->engine.subdev.device; + u64 offset = dmaobj->base.start & 0xfffff000; + u64 adjust = dmaobj->base.start & 0x00000fff; + u32 length = dmaobj->base.limit - dmaobj->base.start; + int ret; + + if (dmaobj->clone) { + struct nvkm_memory *pgt = + device->mmu->vmm->pd->pt[0]->memory; + if (!dmaobj->base.start) + return nvkm_gpuobj_wrap(pgt, pgpuobj); + nvkm_kmap(pgt); + offset = nvkm_ro32(pgt, 8 + (offset >> 10)); + offset &= 0xfffff000; + nvkm_done(pgt); + } + + ret = nvkm_gpuobj_new(device, 16, align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, dmaobj->flags0 | (adjust << 20)); + nvkm_wo32(*pgpuobj, 0x04, length); + nvkm_wo32(*pgpuobj, 0x08, dmaobj->flags2 | offset); + nvkm_wo32(*pgpuobj, 0x0c, dmaobj->flags2 | offset); + nvkm_done(*pgpuobj); + } + + return ret; +} + +static const struct nvkm_dmaobj_func +nv04_dmaobj_func = { + .bind = nv04_dmaobj_bind, +}; + +int +nv04_dmaobj_new(struct nvkm_dma *dma, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_dmaobj **pdmaobj) +{ + struct nvkm_device *device = dma->engine.subdev.device; + struct nv04_dmaobj *dmaobj; + int ret; + + if (!(dmaobj = kzalloc(sizeof(*dmaobj), GFP_KERNEL))) + return -ENOMEM; + *pdmaobj = &dmaobj->base; + + ret = nvkm_dmaobj_ctor(&nv04_dmaobj_func, dma, oclass, + &data, &size, &dmaobj->base); + if (ret) + return ret; + + if (dmaobj->base.target == NV_MEM_TARGET_VM) { + if (device->mmu->func == &nv04_mmu) + dmaobj->clone = true; + dmaobj->base.target = NV_MEM_TARGET_PCI; + dmaobj->base.access = NV_MEM_ACCESS_RW; + } + + dmaobj->flags0 = oclass->base.oclass; + switch (dmaobj->base.target) { + case NV_MEM_TARGET_VRAM: + dmaobj->flags0 |= 0x00003000; + break; + case NV_MEM_TARGET_PCI: + dmaobj->flags0 |= 0x00023000; + break; + case NV_MEM_TARGET_PCI_NOSNOOP: + dmaobj->flags0 |= 0x00033000; + break; + default: + return -EINVAL; + } + + switch (dmaobj->base.access) { + case NV_MEM_ACCESS_RO: + dmaobj->flags0 |= 0x00004000; + break; + case NV_MEM_ACCESS_WO: + dmaobj->flags0 |= 0x00008000; + fallthrough; + case NV_MEM_ACCESS_RW: + dmaobj->flags2 |= 0x00000002; + break; + default: + return -EINVAL; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv50.c new file mode 100644 index 0000000000..6a85b5dea6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/dma/usernv50.c @@ -0,0 +1,157 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define nv50_dmaobj(p) container_of((p), struct nv50_dmaobj, base) +#include "user.h" + +#include +#include +#include + +#include +#include + +struct nv50_dmaobj { + struct nvkm_dmaobj base; + u32 flags0; + u32 flags5; +}; + +static int +nv50_dmaobj_bind(struct nvkm_dmaobj *base, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nv50_dmaobj *dmaobj = nv50_dmaobj(base); + struct nvkm_device *device = dmaobj->base.dma->engine.subdev.device; + int ret; + + ret = nvkm_gpuobj_new(device, 24, align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, dmaobj->flags0); + nvkm_wo32(*pgpuobj, 0x04, lower_32_bits(dmaobj->base.limit)); + nvkm_wo32(*pgpuobj, 0x08, lower_32_bits(dmaobj->base.start)); + nvkm_wo32(*pgpuobj, 0x0c, upper_32_bits(dmaobj->base.limit) << 24 | + upper_32_bits(dmaobj->base.start)); + nvkm_wo32(*pgpuobj, 0x10, 0x00000000); + nvkm_wo32(*pgpuobj, 0x14, dmaobj->flags5); + nvkm_done(*pgpuobj); + } + + return ret; +} + +static const struct nvkm_dmaobj_func +nv50_dmaobj_func = { + .bind = nv50_dmaobj_bind, +}; + +int +nv50_dmaobj_new(struct nvkm_dma *dma, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_dmaobj **pdmaobj) +{ + union { + struct nv50_dma_v0 v0; + } *args; + struct nvkm_object *parent = oclass->parent; + struct nv50_dmaobj *dmaobj; + u32 user, part, comp, kind; + int ret; + + if (!(dmaobj = kzalloc(sizeof(*dmaobj), GFP_KERNEL))) + return -ENOMEM; + *pdmaobj = &dmaobj->base; + + ret = nvkm_dmaobj_ctor(&nv50_dmaobj_func, dma, oclass, + &data, &size, &dmaobj->base); + if (ret) + return ret; + + ret = -ENOSYS; + args = data; + + nvif_ioctl(parent, "create nv50 dma size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, "create nv50 dma vers %d priv %d part %d " + "comp %d kind %02x\n", args->v0.version, + args->v0.priv, args->v0.part, args->v0.comp, + args->v0.kind); + user = args->v0.priv; + part = args->v0.part; + comp = args->v0.comp; + kind = args->v0.kind; + } else + if (size == 0) { + if (dmaobj->base.target != NV_MEM_TARGET_VM) { + user = NV50_DMA_V0_PRIV_US; + part = NV50_DMA_V0_PART_256; + comp = NV50_DMA_V0_COMP_NONE; + kind = NV50_DMA_V0_KIND_PITCH; + } else { + user = NV50_DMA_V0_PRIV_VM; + part = NV50_DMA_V0_PART_VM; + comp = NV50_DMA_V0_COMP_VM; + kind = NV50_DMA_V0_KIND_VM; + } + } else + return ret; + + if (user > 2 || part > 2 || comp > 3 || kind > 0x7f) + return -EINVAL; + dmaobj->flags0 = (comp << 29) | (kind << 22) | (user << 20) | + oclass->base.oclass; + dmaobj->flags5 = (part << 16); + + switch (dmaobj->base.target) { + case NV_MEM_TARGET_VM: + dmaobj->flags0 |= 0x00000000; + break; + case NV_MEM_TARGET_VRAM: + dmaobj->flags0 |= 0x00010000; + break; + case NV_MEM_TARGET_PCI: + dmaobj->flags0 |= 0x00020000; + break; + case NV_MEM_TARGET_PCI_NOSNOOP: + dmaobj->flags0 |= 0x00030000; + break; + default: + return -EINVAL; + } + + switch (dmaobj->base.access) { + case NV_MEM_ACCESS_VM: + break; + case NV_MEM_ACCESS_RO: + dmaobj->flags0 |= 0x00040000; + break; + case NV_MEM_ACCESS_WO: + case NV_MEM_ACCESS_RW: + dmaobj->flags0 |= 0x00080000; + break; + default: + return -EINVAL; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/falcon.c b/drivers/gpu/drm/nouveau/nvkm/engine/falcon.c new file mode 100644 index 0000000000..d619b40a42 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/falcon.c @@ -0,0 +1,355 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include + +#include +#include +#include +#include + +static int +nvkm_falcon_oclass_get(struct nvkm_oclass *oclass, int index) +{ + struct nvkm_falcon *falcon = nvkm_falcon(oclass->engine); + int c = 0; + + while (falcon->func->sclass[c].oclass) { + if (c++ == index) { + oclass->base = falcon->func->sclass[index]; + return index; + } + } + + return c; +} + +static int +nvkm_falcon_cclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + return nvkm_gpuobj_new(object->engine->subdev.device, 256, + align, true, parent, pgpuobj); +} + +static const struct nvkm_object_func +nvkm_falcon_cclass = { + .bind = nvkm_falcon_cclass_bind, +}; + +static void +nvkm_falcon_intr(struct nvkm_engine *engine) +{ + struct nvkm_falcon *falcon = nvkm_falcon(engine); + struct nvkm_subdev *subdev = &falcon->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 base = falcon->addr; + u32 dest = nvkm_rd32(device, base + 0x01c); + u32 intr = nvkm_rd32(device, base + 0x008) & dest & ~(dest >> 16); + u32 inst = nvkm_rd32(device, base + 0x050) & 0x3fffffff; + struct nvkm_chan *chan; + unsigned long flags; + + chan = nvkm_chan_get_inst(engine, (u64)inst << 12, &flags); + + if (intr & 0x00000040) { + if (falcon->func->intr) { + falcon->func->intr(falcon, chan); + nvkm_wr32(device, base + 0x004, 0x00000040); + intr &= ~0x00000040; + } + } + + if (intr & 0x00000010) { + nvkm_debug(subdev, "ucode halted\n"); + nvkm_wr32(device, base + 0x004, 0x00000010); + intr &= ~0x00000010; + } + + if (intr) { + nvkm_error(subdev, "intr %08x\n", intr); + nvkm_wr32(device, base + 0x004, intr); + } + + nvkm_chan_put(&chan, flags); +} + +static int +nvkm_falcon_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_falcon *falcon = nvkm_falcon(engine); + struct nvkm_device *device = falcon->engine.subdev.device; + const u32 base = falcon->addr; + + if (!suspend) { + nvkm_memory_unref(&falcon->core); + if (falcon->external) { + vfree(falcon->data.data); + vfree(falcon->code.data); + falcon->code.data = NULL; + } + } + + if (nvkm_mc_enabled(device, engine->subdev.type, engine->subdev.inst)) { + nvkm_mask(device, base + 0x048, 0x00000003, 0x00000000); + nvkm_wr32(device, base + 0x014, 0xffffffff); + } + return 0; +} + +static void * +vmemdup(const void *src, size_t len) +{ + void *p = vmalloc(len); + + if (p) + memcpy(p, src, len); + return p; +} + +static int +nvkm_falcon_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_falcon *falcon = nvkm_falcon(engine); + struct nvkm_subdev *subdev = &falcon->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 base = falcon->addr; + u32 caps; + + /* determine falcon capabilities */ + if (device->chipset < 0xa3 || + device->chipset == 0xaa || device->chipset == 0xac) { + falcon->version = 0; + falcon->secret = (falcon->addr == 0x087000) ? 1 : 0; + } else { + caps = nvkm_rd32(device, base + 0x12c); + falcon->version = (caps & 0x0000000f); + falcon->secret = (caps & 0x00000030) >> 4; + } + + caps = nvkm_rd32(device, base + 0x108); + falcon->code.limit = (caps & 0x000001ff) << 8; + falcon->data.limit = (caps & 0x0003fe00) >> 1; + + nvkm_debug(subdev, "falcon version: %d\n", falcon->version); + nvkm_debug(subdev, "secret level: %d\n", falcon->secret); + nvkm_debug(subdev, "code limit: %d\n", falcon->code.limit); + nvkm_debug(subdev, "data limit: %d\n", falcon->data.limit); + return 0; +} + +static int +nvkm_falcon_init(struct nvkm_engine *engine) +{ + struct nvkm_falcon *falcon = nvkm_falcon(engine); + struct nvkm_subdev *subdev = &falcon->engine.subdev; + struct nvkm_device *device = subdev->device; + const struct firmware *fw; + char name[32] = "internal"; + const u32 base = falcon->addr; + int ret, i; + + /* wait for 'uc halted' to be signalled before continuing */ + if (falcon->secret && falcon->version < 4) { + if (!falcon->version) { + nvkm_msec(device, 2000, + if (nvkm_rd32(device, base + 0x008) & 0x00000010) + break; + ); + } else { + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, base + 0x180) & 0x80000000)) + break; + ); + } + nvkm_wr32(device, base + 0x004, 0x00000010); + } + + /* disable all interrupts */ + nvkm_wr32(device, base + 0x014, 0xffffffff); + + /* no default ucode provided by the engine implementation, try and + * locate a "self-bootstrapping" firmware image for the engine + */ + if (!falcon->code.data) { + snprintf(name, sizeof(name), "nouveau/nv%02x_fuc%03x", + device->chipset, falcon->addr >> 12); + + ret = request_firmware(&fw, name, device->dev); + if (ret == 0) { + falcon->code.data = vmemdup(fw->data, fw->size); + falcon->code.size = fw->size; + falcon->data.data = NULL; + falcon->data.size = 0; + release_firmware(fw); + } + + falcon->external = true; + } + + /* next step is to try and load "static code/data segment" firmware + * images for the engine + */ + if (!falcon->code.data) { + snprintf(name, sizeof(name), "nouveau/nv%02x_fuc%03xd", + device->chipset, falcon->addr >> 12); + + ret = request_firmware(&fw, name, device->dev); + if (ret) { + nvkm_error(subdev, "unable to load firmware data\n"); + return -ENODEV; + } + + falcon->data.data = vmemdup(fw->data, fw->size); + falcon->data.size = fw->size; + release_firmware(fw); + if (!falcon->data.data) + return -ENOMEM; + + snprintf(name, sizeof(name), "nouveau/nv%02x_fuc%03xc", + device->chipset, falcon->addr >> 12); + + ret = request_firmware(&fw, name, device->dev); + if (ret) { + nvkm_error(subdev, "unable to load firmware code\n"); + return -ENODEV; + } + + falcon->code.data = vmemdup(fw->data, fw->size); + falcon->code.size = fw->size; + release_firmware(fw); + if (!falcon->code.data) + return -ENOMEM; + } + + nvkm_debug(subdev, "firmware: %s (%s)\n", name, falcon->data.data ? + "static code/data segments" : "self-bootstrapping"); + + /* ensure any "self-bootstrapping" firmware image is in vram */ + if (!falcon->data.data && !falcon->core) { + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, + falcon->code.size, 256, false, + &falcon->core); + if (ret) { + nvkm_error(subdev, "core allocation failed, %d\n", ret); + return ret; + } + + nvkm_kmap(falcon->core); + for (i = 0; i < falcon->code.size; i += 4) + nvkm_wo32(falcon->core, i, falcon->code.data[i / 4]); + nvkm_done(falcon->core); + } + + /* upload firmware bootloader (or the full code segments) */ + if (falcon->core) { + u64 addr = nvkm_memory_addr(falcon->core); + if (device->card_type < NV_C0) + nvkm_wr32(device, base + 0x618, 0x04000000); + else + nvkm_wr32(device, base + 0x618, 0x00000114); + nvkm_wr32(device, base + 0x11c, 0); + nvkm_wr32(device, base + 0x110, addr >> 8); + nvkm_wr32(device, base + 0x114, 0); + nvkm_wr32(device, base + 0x118, 0x00006610); + } else { + if (falcon->code.size > falcon->code.limit || + falcon->data.size > falcon->data.limit) { + nvkm_error(subdev, "ucode exceeds falcon limit(s)\n"); + return -EINVAL; + } + + if (falcon->version < 3) { + nvkm_wr32(device, base + 0xff8, 0x00100000); + for (i = 0; i < falcon->code.size / 4; i++) + nvkm_wr32(device, base + 0xff4, falcon->code.data[i]); + } else { + nvkm_wr32(device, base + 0x180, 0x01000000); + for (i = 0; i < falcon->code.size / 4; i++) { + if ((i & 0x3f) == 0) + nvkm_wr32(device, base + 0x188, i >> 6); + nvkm_wr32(device, base + 0x184, falcon->code.data[i]); + } + } + } + + /* upload data segment (if necessary), zeroing the remainder */ + if (falcon->version < 3) { + nvkm_wr32(device, base + 0xff8, 0x00000000); + for (i = 0; !falcon->core && i < falcon->data.size / 4; i++) + nvkm_wr32(device, base + 0xff4, falcon->data.data[i]); + for (; i < falcon->data.limit; i += 4) + nvkm_wr32(device, base + 0xff4, 0x00000000); + } else { + nvkm_wr32(device, base + 0x1c0, 0x01000000); + for (i = 0; !falcon->core && i < falcon->data.size / 4; i++) + nvkm_wr32(device, base + 0x1c4, falcon->data.data[i]); + for (; i < falcon->data.limit / 4; i++) + nvkm_wr32(device, base + 0x1c4, 0x00000000); + } + + /* start it running */ + nvkm_wr32(device, base + 0x10c, 0x00000001); /* BLOCK_ON_FIFO */ + nvkm_wr32(device, base + 0x104, 0x00000000); /* ENTRY */ + nvkm_wr32(device, base + 0x100, 0x00000002); /* TRIGGER */ + nvkm_wr32(device, base + 0x048, 0x00000003); /* FIFO | CHSW */ + + if (falcon->func->init) + falcon->func->init(falcon); + return 0; +} + +static void * +nvkm_falcon_dtor(struct nvkm_engine *engine) +{ + return nvkm_falcon(engine); +} + +static const struct nvkm_engine_func +nvkm_falcon = { + .dtor = nvkm_falcon_dtor, + .oneinit = nvkm_falcon_oneinit, + .init = nvkm_falcon_init, + .fini = nvkm_falcon_fini, + .intr = nvkm_falcon_intr, + .fifo.sclass = nvkm_falcon_oclass_get, + .cclass = &nvkm_falcon_cclass, +}; + +int +nvkm_falcon_new_(const struct nvkm_falcon_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, bool enable, u32 addr, + struct nvkm_engine **pengine) +{ + struct nvkm_falcon *falcon; + + if (!(falcon = kzalloc(sizeof(*falcon), GFP_KERNEL))) + return -ENOMEM; + falcon->func = func; + falcon->addr = addr; + falcon->code.data = func->code.data; + falcon->code.size = func->code.size; + falcon->data.data = func->data.data; + falcon->data.size = func->data.size; + *pengine = &falcon->engine; + + return nvkm_engine_ctor(&nvkm_falcon, device, type, inst, enable, &falcon->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild new file mode 100644 index 0000000000..5a074b9970 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild @@ -0,0 +1,30 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/fifo/base.o +nvkm-y += nvkm/engine/fifo/cgrp.o +nvkm-y += nvkm/engine/fifo/chan.o +nvkm-y += nvkm/engine/fifo/chid.o +nvkm-y += nvkm/engine/fifo/runl.o +nvkm-y += nvkm/engine/fifo/runq.o + +nvkm-y += nvkm/engine/fifo/nv04.o +nvkm-y += nvkm/engine/fifo/nv10.o +nvkm-y += nvkm/engine/fifo/nv17.o +nvkm-y += nvkm/engine/fifo/nv40.o +nvkm-y += nvkm/engine/fifo/nv50.o +nvkm-y += nvkm/engine/fifo/g84.o +nvkm-y += nvkm/engine/fifo/g98.o +nvkm-y += nvkm/engine/fifo/gf100.o +nvkm-y += nvkm/engine/fifo/gk104.o +nvkm-y += nvkm/engine/fifo/gk110.o +nvkm-y += nvkm/engine/fifo/gk208.o +nvkm-y += nvkm/engine/fifo/gk20a.o +nvkm-y += nvkm/engine/fifo/gm107.o +nvkm-y += nvkm/engine/fifo/gm200.o +nvkm-y += nvkm/engine/fifo/gp100.o +nvkm-y += nvkm/engine/fifo/gv100.o +nvkm-y += nvkm/engine/fifo/tu102.o +nvkm-y += nvkm/engine/fifo/ga100.o +nvkm-y += nvkm/engine/fifo/ga102.o + +nvkm-y += nvkm/engine/fifo/ucgrp.o +nvkm-y += nvkm/engine/fifo/uchan.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c new file mode 100644 index 0000000000..5db37247dc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/base.c @@ -0,0 +1,387 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" +#include "runq.h" + +#include +#include +#include +#include + +#include +#include + +bool +nvkm_fifo_ctxsw_in_progress(struct nvkm_engine *engine) +{ + struct nvkm_runl *runl; + struct nvkm_engn *engn; + + nvkm_runl_foreach(runl, engine->subdev.device->fifo) { + nvkm_runl_foreach_engn(engn, runl) { + if (engn->engine == engine) + return engn->func->chsw ? engn->func->chsw(engn) : false; + } + } + + return false; +} + +void +nvkm_fifo_pause(struct nvkm_fifo *fifo, unsigned long *flags) +{ + return fifo->func->pause(fifo, flags); +} + +void +nvkm_fifo_start(struct nvkm_fifo *fifo, unsigned long *flags) +{ + return fifo->func->start(fifo, flags); +} + +void +nvkm_fifo_fault(struct nvkm_fifo *fifo, struct nvkm_fault_data *info) +{ + return fifo->func->mmu_fault->recover(fifo, info); +} + +static int +nvkm_fifo_class_new(struct nvkm_device *device, const struct nvkm_oclass *oclass, + void *argv, u32 argc, struct nvkm_object **pobject) +{ + struct nvkm_fifo *fifo = nvkm_fifo(oclass->engine); + + if (oclass->engn == &fifo->func->cgrp.user) + return nvkm_ucgrp_new(fifo, oclass, argv, argc, pobject); + + if (oclass->engn == &fifo->func->chan.user) + return nvkm_uchan_new(fifo, NULL, oclass, argv, argc, pobject); + + WARN_ON(1); + return -ENOSYS; +} + +static const struct nvkm_device_oclass +nvkm_fifo_class = { + .ctor = nvkm_fifo_class_new, +}; + +static int +nvkm_fifo_class_get(struct nvkm_oclass *oclass, int index, const struct nvkm_device_oclass **class) +{ + struct nvkm_fifo *fifo = nvkm_fifo(oclass->engine); + const struct nvkm_fifo_func_cgrp *cgrp = &fifo->func->cgrp; + const struct nvkm_fifo_func_chan *chan = &fifo->func->chan; + int c = 0; + + /* *_CHANNEL_GROUP_* */ + if (cgrp->user.oclass) { + if (c++ == index) { + oclass->base = cgrp->user; + oclass->engn = &fifo->func->cgrp.user; + *class = &nvkm_fifo_class; + return 0; + } + } + + /* *_CHANNEL_DMA, *_CHANNEL_GPFIFO_* */ + if (chan->user.oclass) { + if (c++ == index) { + oclass->base = chan->user; + oclass->engn = &fifo->func->chan.user; + *class = &nvkm_fifo_class; + return 0; + } + } + + return c; +} + +static int +nvkm_fifo_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_fifo *fifo = nvkm_fifo(engine); + struct nvkm_runl *runl; + + nvkm_inth_block(&fifo->engine.subdev.inth); + + nvkm_runl_foreach(runl, fifo) + nvkm_runl_fini(runl); + + return 0; +} + +static int +nvkm_fifo_init(struct nvkm_engine *engine) +{ + struct nvkm_fifo *fifo = nvkm_fifo(engine); + struct nvkm_runq *runq; + struct nvkm_runl *runl; + u32 mask = 0; + + if (fifo->func->init_pbdmas) { + nvkm_runq_foreach(runq, fifo) + mask |= BIT(runq->id); + + fifo->func->init_pbdmas(fifo, mask); + + nvkm_runq_foreach(runq, fifo) + runq->func->init(runq); + } + + nvkm_runl_foreach(runl, fifo) { + if (runl->func->init) + runl->func->init(runl); + } + + if (fifo->func->init) + fifo->func->init(fifo); + + nvkm_inth_allow(&fifo->engine.subdev.inth); + return 0; +} + +static int +nvkm_fifo_info(struct nvkm_engine *engine, u64 mthd, u64 *data) +{ + struct nvkm_fifo *fifo = nvkm_fifo(engine); + struct nvkm_runl *runl; + struct nvkm_engn *engn; + int ret; + + ret = nvkm_subdev_oneinit(&fifo->engine.subdev); + if (ret) + return ret; + + switch (mthd) { + case NV_DEVICE_HOST_CHANNELS: *data = fifo->chid ? fifo->chid->nr : 0; return 0; + case NV_DEVICE_HOST_RUNLISTS: + *data = 0; + nvkm_runl_foreach(runl, fifo) + *data |= BIT(runl->id); + return 0; + case NV_DEVICE_HOST_RUNLIST_ENGINES: + runl = nvkm_runl_get(fifo, *data, 0); + if (runl) { + *data = 0; + nvkm_runl_foreach_engn(engn, runl) { +#define CASE(n) case NVKM_ENGINE_##n: *data |= NV_DEVICE_HOST_RUNLIST_ENGINES_##n; break + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_DMAOBJ: + break; + CASE(SW ); + CASE(GR ); + CASE(MPEG ); + CASE(ME ); + CASE(CIPHER); + CASE(BSP ); + CASE(VP ); + CASE(CE ); + CASE(SEC ); + CASE(MSVLD ); + CASE(MSPDEC); + CASE(MSPPP ); + CASE(MSENC ); + CASE(VIC ); + CASE(SEC2 ); + CASE(NVDEC ); + CASE(NVENC ); + default: + WARN_ON(1); + break; + } +#undef CASE + } + return 0; + } + return -EINVAL; + case NV_DEVICE_HOST_RUNLIST_CHANNELS: + if (!fifo->chid) { + runl = nvkm_runl_get(fifo, *data, 0); + if (runl) { + *data = runl->chid->nr; + return 0; + } + } + return -EINVAL; + default: + break; + } + + return -ENOSYS; +} + +static int +nvkm_fifo_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_subdev *subdev = &engine->subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_fifo *fifo = nvkm_fifo(engine); + struct nvkm_runl *runl; + struct nvkm_engn *engn; + int ret, nr, i; + + /* Initialise CHID/CGID allocator(s) on GPUs where they aren't per-runlist. */ + if (fifo->func->chid_nr) { + ret = fifo->func->chid_ctor(fifo, fifo->func->chid_nr(fifo)); + if (ret) + return ret; + } + + /* Create runqueues for each PBDMA. */ + if (fifo->func->runq_nr) { + for (nr = fifo->func->runq_nr(fifo), i = 0; i < nr; i++) { + if (!nvkm_runq_new(fifo, i)) + return -ENOMEM; + } + } + + /* Create runlists. */ + ret = fifo->func->runl_ctor(fifo); + if (ret) + return ret; + + nvkm_runl_foreach(runl, fifo) { + RUNL_DEBUG(runl, "chan:%06x", runl->chan); + nvkm_runl_foreach_engn(engn, runl) { + ENGN_DEBUG(engn, ""); + } + } + + /* Register interrupt handler. */ + if (fifo->func->intr) { + ret = nvkm_inth_add(&device->mc->intr, NVKM_INTR_SUBDEV, NVKM_INTR_PRIO_NORMAL, + subdev, fifo->func->intr, &subdev->inth); + if (ret) { + nvkm_error(subdev, "intr %d\n", ret); + return ret; + } + } + + /* Initialise non-stall intr handling. */ + if (fifo->func->nonstall) { + if (fifo->func->nonstall_ctor) { + ret = fifo->func->nonstall_ctor(fifo); + if (ret < 0) { + nvkm_error(subdev, "nonstall %d\n", ret); + return ret; + } + } else { + ret = 1; + } + + ret = nvkm_event_init(fifo->func->nonstall, &fifo->engine.subdev, 1, ret, + &fifo->nonstall.event); + if (ret) + return ret; + } + + /* Allocate USERD + BAR1 polling area. */ + if (fifo->func->chan.func->userd->bar == 1) { + struct nvkm_vmm *bar1 = nvkm_bar_bar1_vmm(device); + + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, fifo->chid->nr * + fifo->func->chan.func->userd->size, 0, true, + &fifo->userd.mem); + if (ret) + return ret; + + ret = nvkm_vmm_get(bar1, 12, nvkm_memory_size(fifo->userd.mem), &fifo->userd.bar1); + if (ret) + return ret; + + ret = nvkm_memory_map(fifo->userd.mem, 0, bar1, fifo->userd.bar1, NULL, 0); + if (ret) + return ret; + } + + return 0; +} + +static void +nvkm_fifo_preinit(struct nvkm_engine *engine) +{ + nvkm_mc_reset(engine->subdev.device, NVKM_ENGINE_FIFO, 0); +} + +static void * +nvkm_fifo_dtor(struct nvkm_engine *engine) +{ + struct nvkm_fifo *fifo = nvkm_fifo(engine); + struct nvkm_runl *runl, *runt; + struct nvkm_runq *runq, *rtmp; + + if (fifo->userd.bar1) + nvkm_vmm_put(nvkm_bar_bar1_vmm(engine->subdev.device), &fifo->userd.bar1); + nvkm_memory_unref(&fifo->userd.mem); + + list_for_each_entry_safe(runl, runt, &fifo->runls, head) + nvkm_runl_del(runl); + list_for_each_entry_safe(runq, rtmp, &fifo->runqs, head) + nvkm_runq_del(runq); + + nvkm_chid_unref(&fifo->cgid); + nvkm_chid_unref(&fifo->chid); + + nvkm_event_fini(&fifo->nonstall.event); + mutex_destroy(&fifo->mutex); + return fifo; +} + +static const struct nvkm_engine_func +nvkm_fifo = { + .dtor = nvkm_fifo_dtor, + .preinit = nvkm_fifo_preinit, + .oneinit = nvkm_fifo_oneinit, + .info = nvkm_fifo_info, + .init = nvkm_fifo_init, + .fini = nvkm_fifo_fini, + .base.sclass = nvkm_fifo_class_get, +}; + +int +nvkm_fifo_new_(const struct nvkm_fifo_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_fifo **pfifo) +{ + struct nvkm_fifo *fifo; + + if (!(fifo = *pfifo = kzalloc(sizeof(*fifo), GFP_KERNEL))) + return -ENOMEM; + + fifo->func = func; + INIT_LIST_HEAD(&fifo->runqs); + INIT_LIST_HEAD(&fifo->runls); + /*TODO: Needs to be >CTXSW_TIMEOUT, so RC can recover before this is hit. + * CTXSW_TIMEOUT HW default seems to differ between GPUs, so just a + * large number for now until we support changing it. + */ + fifo->timeout.chan_msec = 10000; + spin_lock_init(&fifo->lock); + mutex_init(&fifo->mutex); + + return nvkm_engine_ctor(&nvkm_fifo, device, type, inst, true, &fifo->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.c new file mode 100644 index 0000000000..ea53fb3d5d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.c @@ -0,0 +1,252 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" +#include "priv.h" + +#include +#include + +static void +nvkm_cgrp_ectx_put(struct nvkm_cgrp *cgrp, struct nvkm_ectx **pectx) +{ + struct nvkm_ectx *ectx = *pectx; + + if (ectx) { + struct nvkm_engn *engn = ectx->engn; + + if (refcount_dec_and_test(&ectx->refs)) { + CGRP_TRACE(cgrp, "dtor ectx %d[%s]", engn->id, engn->engine->subdev.name); + nvkm_object_del(&ectx->object); + list_del(&ectx->head); + kfree(ectx); + } + + *pectx = NULL; + } +} + +static int +nvkm_cgrp_ectx_get(struct nvkm_cgrp *cgrp, struct nvkm_engn *engn, struct nvkm_ectx **pectx, + struct nvkm_chan *chan, struct nvkm_client *client) +{ + struct nvkm_engine *engine = engn->engine; + struct nvkm_oclass cclass = { + .client = client, + .engine = engine, + }; + struct nvkm_ectx *ectx; + int ret = 0; + + /* Look for an existing context for this engine in the channel group. */ + ectx = nvkm_list_find(ectx, &cgrp->ectxs, head, ectx->engn == engn); + if (ectx) { + refcount_inc(&ectx->refs); + *pectx = ectx; + return 0; + } + + /* Nope - create a fresh one. */ + CGRP_TRACE(cgrp, "ctor ectx %d[%s]", engn->id, engn->engine->subdev.name); + if (!(ectx = *pectx = kzalloc(sizeof(*ectx), GFP_KERNEL))) + return -ENOMEM; + + ectx->engn = engn; + refcount_set(&ectx->refs, 1); + refcount_set(&ectx->uses, 0); + list_add_tail(&ectx->head, &cgrp->ectxs); + + /* Allocate the HW structures. */ + if (engine->func->fifo.cclass) + ret = engine->func->fifo.cclass(chan, &cclass, &ectx->object); + else if (engine->func->cclass) + ret = nvkm_object_new_(engine->func->cclass, &cclass, NULL, 0, &ectx->object); + + if (ret) + nvkm_cgrp_ectx_put(cgrp, pectx); + + return ret; +} + +void +nvkm_cgrp_vctx_put(struct nvkm_cgrp *cgrp, struct nvkm_vctx **pvctx) +{ + struct nvkm_vctx *vctx = *pvctx; + + if (vctx) { + struct nvkm_engn *engn = vctx->ectx->engn; + + if (refcount_dec_and_test(&vctx->refs)) { + CGRP_TRACE(cgrp, "dtor vctx %d[%s]", engn->id, engn->engine->subdev.name); + nvkm_vmm_put(vctx->vmm, &vctx->vma); + nvkm_gpuobj_del(&vctx->inst); + + nvkm_cgrp_ectx_put(cgrp, &vctx->ectx); + if (vctx->vmm) { + atomic_dec(&vctx->vmm->engref[engn->engine->subdev.type]); + nvkm_vmm_unref(&vctx->vmm); + } + list_del(&vctx->head); + kfree(vctx); + } + + *pvctx = NULL; + } +} + +int +nvkm_cgrp_vctx_get(struct nvkm_cgrp *cgrp, struct nvkm_engn *engn, struct nvkm_chan *chan, + struct nvkm_vctx **pvctx, struct nvkm_client *client) +{ + struct nvkm_ectx *ectx; + struct nvkm_vctx *vctx; + int ret; + + /* Look for an existing sub-context for this engine+VEID in the channel group. */ + vctx = nvkm_list_find(vctx, &cgrp->vctxs, head, + vctx->ectx->engn == engn && vctx->vmm == chan->vmm); + if (vctx) { + refcount_inc(&vctx->refs); + *pvctx = vctx; + return 0; + } + + /* Nope - create a fresh one. But, context first. */ + ret = nvkm_cgrp_ectx_get(cgrp, engn, &ectx, chan, client); + if (ret) { + CGRP_ERROR(cgrp, "ectx %d[%s]: %d", engn->id, engn->engine->subdev.name, ret); + return ret; + } + + /* Now, create the sub-context. */ + CGRP_TRACE(cgrp, "ctor vctx %d[%s]", engn->id, engn->engine->subdev.name); + if (!(vctx = *pvctx = kzalloc(sizeof(*vctx), GFP_KERNEL))) { + nvkm_cgrp_ectx_put(cgrp, &ectx); + return -ENOMEM; + } + + vctx->ectx = ectx; + vctx->vmm = nvkm_vmm_ref(chan->vmm); + refcount_set(&vctx->refs, 1); + list_add_tail(&vctx->head, &cgrp->vctxs); + + /* MMU on some GPUs needs to know engine usage for TLB invalidation. */ + if (vctx->vmm) + atomic_inc(&vctx->vmm->engref[engn->engine->subdev.type]); + + /* Allocate the HW structures. */ + if (engn->func->bind) { + ret = nvkm_object_bind(vctx->ectx->object, NULL, 0, &vctx->inst); + if (ret == 0 && engn->func->ctor) + ret = engn->func->ctor(engn, vctx); + } + + if (ret) + nvkm_cgrp_vctx_put(cgrp, pvctx); + + return ret; +} + +static void +nvkm_cgrp_del(struct kref *kref) +{ + struct nvkm_cgrp *cgrp = container_of(kref, typeof(*cgrp), kref); + struct nvkm_runl *runl = cgrp->runl; + + if (runl->cgid) + nvkm_chid_put(runl->cgid, cgrp->id, &cgrp->lock); + + mutex_destroy(&cgrp->mutex); + nvkm_vmm_unref(&cgrp->vmm); + kfree(cgrp); +} + +void +nvkm_cgrp_unref(struct nvkm_cgrp **pcgrp) +{ + struct nvkm_cgrp *cgrp = *pcgrp; + + if (!cgrp) + return; + + kref_put(&cgrp->kref, nvkm_cgrp_del); + *pcgrp = NULL; +} + +struct nvkm_cgrp * +nvkm_cgrp_ref(struct nvkm_cgrp *cgrp) +{ + if (cgrp) + kref_get(&cgrp->kref); + + return cgrp; +} + +void +nvkm_cgrp_put(struct nvkm_cgrp **pcgrp, unsigned long irqflags) +{ + struct nvkm_cgrp *cgrp = *pcgrp; + + if (!cgrp) + return; + + *pcgrp = NULL; + spin_unlock_irqrestore(&cgrp->lock, irqflags); +} + +int +nvkm_cgrp_new(struct nvkm_runl *runl, const char *name, struct nvkm_vmm *vmm, bool hw, + struct nvkm_cgrp **pcgrp) +{ + struct nvkm_cgrp *cgrp; + + if (!(cgrp = *pcgrp = kmalloc(sizeof(*cgrp), GFP_KERNEL))) + return -ENOMEM; + + cgrp->func = runl->fifo->func->cgrp.func; + strscpy(cgrp->name, name, sizeof(cgrp->name)); + cgrp->runl = runl; + cgrp->vmm = nvkm_vmm_ref(vmm); + cgrp->hw = hw; + cgrp->id = -1; + kref_init(&cgrp->kref); + INIT_LIST_HEAD(&cgrp->chans); + cgrp->chan_nr = 0; + spin_lock_init(&cgrp->lock); + INIT_LIST_HEAD(&cgrp->ectxs); + INIT_LIST_HEAD(&cgrp->vctxs); + mutex_init(&cgrp->mutex); + atomic_set(&cgrp->rc, NVKM_CGRP_RC_NONE); + + if (runl->cgid) { + cgrp->id = nvkm_chid_get(runl->cgid, cgrp); + if (cgrp->id < 0) { + RUNL_ERROR(runl, "!cgids"); + nvkm_cgrp_unref(pcgrp); + return -ENOSPC; + } + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.h new file mode 100644 index 0000000000..5f6abd59a6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/cgrp.h @@ -0,0 +1,75 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_CGRP_H__ +#define __NVKM_CGRP_H__ +#include +struct nvkm_chan; +struct nvkm_client; + +struct nvkm_vctx { + struct nvkm_ectx *ectx; + struct nvkm_vmm *vmm; + refcount_t refs; + + struct nvkm_gpuobj *inst; + struct nvkm_vma *vma; + + struct list_head head; +}; + +struct nvkm_ectx { + struct nvkm_engn *engn; + refcount_t refs; + refcount_t uses; + + struct nvkm_object *object; + + struct list_head head; +}; + +struct nvkm_cgrp { + const struct nvkm_cgrp_func { + void (*preempt)(struct nvkm_cgrp *); + } *func; + char name[64]; + struct nvkm_runl *runl; + struct nvkm_vmm *vmm; + bool hw; + int id; + struct kref kref; + + struct list_head chans; + int chan_nr; + + spinlock_t lock; /* protects irq handler channel (group) lookup */ + + struct list_head ectxs; + struct list_head vctxs; + struct mutex mutex; + +#define NVKM_CGRP_RC_NONE 0 +#define NVKM_CGRP_RC_PENDING 1 +#define NVKM_CGRP_RC_RUNNING 2 + atomic_t rc; + + struct list_head head; +}; + +int nvkm_cgrp_new(struct nvkm_runl *, const char *name, struct nvkm_vmm *, bool hw, + struct nvkm_cgrp **); +struct nvkm_cgrp *nvkm_cgrp_ref(struct nvkm_cgrp *); +void nvkm_cgrp_unref(struct nvkm_cgrp **); +int nvkm_cgrp_vctx_get(struct nvkm_cgrp *, struct nvkm_engn *, struct nvkm_chan *, + struct nvkm_vctx **, struct nvkm_client *); +void nvkm_cgrp_vctx_put(struct nvkm_cgrp *, struct nvkm_vctx **); + +void nvkm_cgrp_put(struct nvkm_cgrp **, unsigned long irqflags); + +#define nvkm_cgrp_foreach_chan(chan,cgrp) list_for_each_entry((chan), &(cgrp)->chans, head) +#define nvkm_cgrp_foreach_chan_safe(chan,ctmp,cgrp) \ + list_for_each_entry_safe((chan), (ctmp), &(cgrp)->chans, head) + +#define CGRP_PRCLI(c,l,p,f,a...) RUNL_PRINT((c)->runl, l, p, "%04x:[%s]"f, (c)->id, (c)->name, ##a) +#define CGRP_PRINT(c,l,p,f,a...) RUNL_PRINT((c)->runl, l, p, "%04x:"f, (c)->id, ##a) +#define CGRP_ERROR(c,f,a...) CGRP_PRCLI((c), ERROR, err, " "f"\n", ##a) +#define CGRP_TRACE(c,f,a...) CGRP_PRINT((c), TRACE, info, " "f"\n", ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c new file mode 100644 index 0000000000..b7c9d6115b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.c @@ -0,0 +1,480 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "chan.h" +#include "chid.h" +#include "cgrp.h" +#include "chid.h" +#include "runl.h" +#include "priv.h" + +#include +#include +#include + +#include + +const struct nvkm_event_func +nvkm_chan_event = { +}; + +void +nvkm_chan_cctx_bind(struct nvkm_chan *chan, struct nvkm_engn *engn, struct nvkm_cctx *cctx) +{ + struct nvkm_cgrp *cgrp = chan->cgrp; + struct nvkm_runl *runl = cgrp->runl; + struct nvkm_engine *engine = engn->engine; + + if (!engn->func->bind) + return; + + CHAN_TRACE(chan, "%sbind cctx %d[%s]", cctx ? "" : "un", engn->id, engine->subdev.name); + + /* Prevent any channel in channel group from being rescheduled, kick them + * off host and any engine(s) they're loaded on. + */ + if (cgrp->hw) + nvkm_runl_block(runl); + else + nvkm_chan_block(chan); + nvkm_chan_preempt(chan, true); + + /* Update context pointer. */ + engn->func->bind(engn, cctx, chan); + + /* Resume normal operation. */ + if (cgrp->hw) + nvkm_runl_allow(runl); + else + nvkm_chan_allow(chan); +} + +void +nvkm_chan_cctx_put(struct nvkm_chan *chan, struct nvkm_cctx **pcctx) +{ + struct nvkm_cctx *cctx = *pcctx; + + if (cctx) { + struct nvkm_engn *engn = cctx->vctx->ectx->engn; + + if (refcount_dec_and_mutex_lock(&cctx->refs, &chan->cgrp->mutex)) { + CHAN_TRACE(chan, "dtor cctx %d[%s]", engn->id, engn->engine->subdev.name); + nvkm_cgrp_vctx_put(chan->cgrp, &cctx->vctx); + list_del(&cctx->head); + kfree(cctx); + mutex_unlock(&chan->cgrp->mutex); + } + + *pcctx = NULL; + } +} + +int +nvkm_chan_cctx_get(struct nvkm_chan *chan, struct nvkm_engn *engn, struct nvkm_cctx **pcctx, + struct nvkm_client *client) +{ + struct nvkm_cgrp *cgrp = chan->cgrp; + struct nvkm_vctx *vctx; + struct nvkm_cctx *cctx; + int ret; + + /* Look for an existing channel context for this engine+VEID. */ + mutex_lock(&cgrp->mutex); + cctx = nvkm_list_find(cctx, &chan->cctxs, head, + cctx->vctx->ectx->engn == engn && cctx->vctx->vmm == chan->vmm); + if (cctx) { + refcount_inc(&cctx->refs); + *pcctx = cctx; + mutex_unlock(&chan->cgrp->mutex); + return 0; + } + + /* Nope - create a fresh one. But, sub-context first. */ + ret = nvkm_cgrp_vctx_get(cgrp, engn, chan, &vctx, client); + if (ret) { + CHAN_ERROR(chan, "vctx %d[%s]: %d", engn->id, engn->engine->subdev.name, ret); + goto done; + } + + /* Now, create the channel context - to track engine binding. */ + CHAN_TRACE(chan, "ctor cctx %d[%s]", engn->id, engn->engine->subdev.name); + if (!(cctx = *pcctx = kzalloc(sizeof(*cctx), GFP_KERNEL))) { + nvkm_cgrp_vctx_put(cgrp, &vctx); + ret = -ENOMEM; + goto done; + } + + cctx->vctx = vctx; + refcount_set(&cctx->refs, 1); + refcount_set(&cctx->uses, 0); + list_add_tail(&cctx->head, &chan->cctxs); +done: + mutex_unlock(&cgrp->mutex); + return ret; +} + +int +nvkm_chan_preempt_locked(struct nvkm_chan *chan, bool wait) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + + CHAN_TRACE(chan, "preempt"); + chan->func->preempt(chan); + if (!wait) + return 0; + + return nvkm_runl_preempt_wait(runl); +} + +int +nvkm_chan_preempt(struct nvkm_chan *chan, bool wait) +{ + int ret; + + if (!chan->func->preempt) + return 0; + + mutex_lock(&chan->cgrp->runl->mutex); + ret = nvkm_chan_preempt_locked(chan, wait); + mutex_unlock(&chan->cgrp->runl->mutex); + return ret; +} + +void +nvkm_chan_remove_locked(struct nvkm_chan *chan) +{ + struct nvkm_cgrp *cgrp = chan->cgrp; + struct nvkm_runl *runl = cgrp->runl; + + if (list_empty(&chan->head)) + return; + + CHAN_TRACE(chan, "remove"); + if (!--cgrp->chan_nr) { + runl->cgrp_nr--; + list_del(&cgrp->head); + } + runl->chan_nr--; + list_del_init(&chan->head); + atomic_set(&runl->changed, 1); +} + +void +nvkm_chan_remove(struct nvkm_chan *chan, bool preempt) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + + mutex_lock(&runl->mutex); + if (preempt && chan->func->preempt) + nvkm_chan_preempt_locked(chan, true); + nvkm_chan_remove_locked(chan); + nvkm_runl_update_locked(runl, true); + mutex_unlock(&runl->mutex); +} + +void +nvkm_chan_insert(struct nvkm_chan *chan) +{ + struct nvkm_cgrp *cgrp = chan->cgrp; + struct nvkm_runl *runl = cgrp->runl; + + mutex_lock(&runl->mutex); + if (WARN_ON(!list_empty(&chan->head))) { + mutex_unlock(&runl->mutex); + return; + } + + CHAN_TRACE(chan, "insert"); + list_add_tail(&chan->head, &cgrp->chans); + runl->chan_nr++; + if (!cgrp->chan_nr++) { + list_add_tail(&cgrp->head, &cgrp->runl->cgrps); + runl->cgrp_nr++; + } + atomic_set(&runl->changed, 1); + nvkm_runl_update_locked(runl, true); + mutex_unlock(&runl->mutex); +} + +static void +nvkm_chan_block_locked(struct nvkm_chan *chan) +{ + CHAN_TRACE(chan, "block %d", atomic_read(&chan->blocked)); + if (atomic_inc_return(&chan->blocked) == 1) + chan->func->stop(chan); +} + +void +nvkm_chan_error(struct nvkm_chan *chan, bool preempt) +{ + unsigned long flags; + + spin_lock_irqsave(&chan->lock, flags); + if (atomic_inc_return(&chan->errored) == 1) { + CHAN_ERROR(chan, "errored - disabling channel"); + nvkm_chan_block_locked(chan); + if (preempt) + chan->func->preempt(chan); + nvkm_event_ntfy(&chan->cgrp->runl->chid->event, chan->id, NVKM_CHAN_EVENT_ERRORED); + } + spin_unlock_irqrestore(&chan->lock, flags); +} + +void +nvkm_chan_block(struct nvkm_chan *chan) +{ + spin_lock_irq(&chan->lock); + nvkm_chan_block_locked(chan); + spin_unlock_irq(&chan->lock); +} + +void +nvkm_chan_allow(struct nvkm_chan *chan) +{ + spin_lock_irq(&chan->lock); + CHAN_TRACE(chan, "allow %d", atomic_read(&chan->blocked)); + if (atomic_dec_and_test(&chan->blocked)) + chan->func->start(chan); + spin_unlock_irq(&chan->lock); +} + +void +nvkm_chan_del(struct nvkm_chan **pchan) +{ + struct nvkm_chan *chan = *pchan; + + if (!chan) + return; + + if (chan->func->ramfc->clear) + chan->func->ramfc->clear(chan); + + nvkm_ramht_del(&chan->ramht); + nvkm_gpuobj_del(&chan->pgd); + nvkm_gpuobj_del(&chan->eng); + nvkm_gpuobj_del(&chan->cache); + nvkm_gpuobj_del(&chan->ramfc); + + nvkm_memory_unref(&chan->userd.mem); + + if (chan->cgrp) { + nvkm_chid_put(chan->cgrp->runl->chid, chan->id, &chan->cgrp->lock); + nvkm_cgrp_unref(&chan->cgrp); + } + + if (chan->vmm) { + nvkm_vmm_part(chan->vmm, chan->inst->memory); + nvkm_vmm_unref(&chan->vmm); + } + + nvkm_gpuobj_del(&chan->push); + nvkm_gpuobj_del(&chan->inst); + kfree(chan); +} + +void +nvkm_chan_put(struct nvkm_chan **pchan, unsigned long irqflags) +{ + struct nvkm_chan *chan = *pchan; + + if (!chan) + return; + + *pchan = NULL; + spin_unlock_irqrestore(&chan->cgrp->lock, irqflags); +} + +struct nvkm_chan * +nvkm_chan_get_inst(struct nvkm_engine *engine, u64 inst, unsigned long *pirqflags) +{ + struct nvkm_fifo *fifo = engine->subdev.device->fifo; + struct nvkm_runl *runl; + struct nvkm_engn *engn; + struct nvkm_chan *chan; + + nvkm_runl_foreach(runl, fifo) { + nvkm_runl_foreach_engn(engn, runl) { + if (engine == &fifo->engine || engn->engine == engine) { + chan = nvkm_runl_chan_get_inst(runl, inst, pirqflags); + if (chan || engn->engine == engine) + return chan; + } + } + } + + return NULL; +} + +struct nvkm_chan * +nvkm_chan_get_chid(struct nvkm_engine *engine, int id, unsigned long *pirqflags) +{ + struct nvkm_fifo *fifo = engine->subdev.device->fifo; + struct nvkm_runl *runl; + struct nvkm_engn *engn; + + nvkm_runl_foreach(runl, fifo) { + nvkm_runl_foreach_engn(engn, runl) { + if (fifo->chid || engn->engine == engine) + return nvkm_runl_chan_get_chid(runl, id, pirqflags); + } + } + + return NULL; +} + +int +nvkm_chan_new_(const struct nvkm_chan_func *func, struct nvkm_runl *runl, int runq, + struct nvkm_cgrp *cgrp, const char *name, bool priv, u32 devm, struct nvkm_vmm *vmm, + struct nvkm_dmaobj *dmaobj, u64 offset, u64 length, + struct nvkm_memory *userd, u64 ouserd, struct nvkm_chan **pchan) +{ + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_chan *chan; + int ret; + + /* Validate arguments against class requirements. */ + if ((runq && runq >= runl->func->runqs) || + (!func->inst->vmm != !vmm) || + ((func->userd->bar < 0) == !userd) || + (!func->ramfc->ctxdma != !dmaobj) || + ((func->ramfc->devm < devm) && devm != BIT(0)) || + (!func->ramfc->priv && priv)) { + RUNL_DEBUG(runl, "args runq:%d:%d vmm:%d:%p userd:%d:%p " + "push:%d:%p devm:%08x:%08x priv:%d:%d", + runl->func->runqs, runq, func->inst->vmm, vmm, + func->userd->bar < 0, userd, func->ramfc->ctxdma, dmaobj, + func->ramfc->devm, devm, func->ramfc->priv, priv); + return -EINVAL; + } + + if (!(chan = *pchan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + + chan->func = func; + strscpy(chan->name, name, sizeof(chan->name)); + chan->runq = runq; + chan->id = -1; + spin_lock_init(&chan->lock); + atomic_set(&chan->blocked, 1); + atomic_set(&chan->errored, 0); + INIT_LIST_HEAD(&chan->cctxs); + INIT_LIST_HEAD(&chan->head); + + /* Join channel group. + * + * GK110 and newer support channel groups (aka TSGs), where individual channels + * share a timeslice, and, engine context(s). + * + * As such, engine contexts are tracked in nvkm_cgrp and we need them even when + * channels aren't in an API channel group, and on HW that doesn't support TSGs. + */ + if (!cgrp) { + ret = nvkm_cgrp_new(runl, chan->name, vmm, fifo->func->cgrp.force, &chan->cgrp); + if (ret) { + RUNL_DEBUG(runl, "cgrp %d", ret); + return ret; + } + + cgrp = chan->cgrp; + } else { + if (cgrp->runl != runl || cgrp->vmm != vmm) { + RUNL_DEBUG(runl, "cgrp %d %d", cgrp->runl != runl, cgrp->vmm != vmm); + return -EINVAL; + } + + chan->cgrp = nvkm_cgrp_ref(cgrp); + } + + /* Allocate instance block. */ + ret = nvkm_gpuobj_new(device, func->inst->size, 0x1000, func->inst->zero, NULL, + &chan->inst); + if (ret) { + RUNL_DEBUG(runl, "inst %d", ret); + return ret; + } + + /* Initialise virtual address-space. */ + if (func->inst->vmm) { + if (WARN_ON(vmm->mmu != device->mmu)) + return -EINVAL; + + ret = nvkm_vmm_join(vmm, chan->inst->memory); + if (ret) { + RUNL_DEBUG(runl, "vmm %d", ret); + return ret; + } + + chan->vmm = nvkm_vmm_ref(vmm); + } + + /* Allocate HW ctxdma for push buffer. */ + if (func->ramfc->ctxdma) { + ret = nvkm_object_bind(&dmaobj->object, chan->inst, -16, &chan->push); + if (ret) { + RUNL_DEBUG(runl, "bind %d", ret); + return ret; + } + } + + /* Allocate channel ID. */ + chan->id = nvkm_chid_get(runl->chid, chan); + if (chan->id < 0) { + RUNL_ERROR(runl, "!chids"); + return -ENOSPC; + } + + if (cgrp->id < 0) + cgrp->id = chan->id; + + /* Initialise USERD. */ + if (func->userd->bar < 0) { + if (ouserd + chan->func->userd->size >= nvkm_memory_size(userd)) { + RUNL_DEBUG(runl, "ouserd %llx", ouserd); + return -EINVAL; + } + + ret = nvkm_memory_kmap(userd, &chan->userd.mem); + if (ret) { + RUNL_DEBUG(runl, "userd %d", ret); + return ret; + } + + chan->userd.base = ouserd; + } else { + chan->userd.mem = nvkm_memory_ref(fifo->userd.mem); + chan->userd.base = chan->id * chan->func->userd->size; + } + + if (chan->func->userd->clear) + chan->func->userd->clear(chan); + + /* Initialise RAMFC. */ + ret = chan->func->ramfc->write(chan, offset, length, devm, priv); + if (ret) { + RUNL_DEBUG(runl, "ramfc %d", ret); + return ret; + } + + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h new file mode 100644 index 0000000000..85b94f6991 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chan.h @@ -0,0 +1,78 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_CHAN_H__ +#define __NVKM_CHAN_H__ +#include +struct nvkm_dmaobj; +struct nvkm_engn; +struct nvkm_runl; + +extern const struct nvkm_event_func nvkm_chan_event; + +struct nvkm_cctx { + struct nvkm_vctx *vctx; + refcount_t refs; + refcount_t uses; + + struct list_head head; +}; + +struct nvkm_chan_func { + const struct nvkm_chan_func_inst { + u32 size; + bool zero; + bool vmm; + } *inst; + + const struct nvkm_chan_func_userd { + int bar; + u32 base; + u32 size; + void (*clear)(struct nvkm_chan *); + } *userd; + + const struct nvkm_chan_func_ramfc { + const struct nvkm_ramfc_layout { + unsigned bits:6; + unsigned ctxs:5; + unsigned ctxp:8; + unsigned regs:5; + unsigned regp; + } *layout; + int (*write)(struct nvkm_chan *, u64 offset, u64 length, u32 devm, bool priv); + void (*clear)(struct nvkm_chan *); + bool ctxdma; + u32 devm; + bool priv; + } *ramfc; + + void (*bind)(struct nvkm_chan *); + void (*unbind)(struct nvkm_chan *); + void (*start)(struct nvkm_chan *); + void (*stop)(struct nvkm_chan *); + void (*preempt)(struct nvkm_chan *); + u32 (*doorbell_handle)(struct nvkm_chan *); +}; + +int nvkm_chan_new_(const struct nvkm_chan_func *, struct nvkm_runl *, int runq, struct nvkm_cgrp *, + const char *name, bool priv, u32 devm, struct nvkm_vmm *, struct nvkm_dmaobj *, + u64 offset, u64 length, struct nvkm_memory *userd, u64 userd_bar1, + struct nvkm_chan **); +void nvkm_chan_del(struct nvkm_chan **); +void nvkm_chan_allow(struct nvkm_chan *); +void nvkm_chan_block(struct nvkm_chan *); +void nvkm_chan_error(struct nvkm_chan *, bool preempt); +void nvkm_chan_insert(struct nvkm_chan *); +void nvkm_chan_remove(struct nvkm_chan *, bool preempt); +void nvkm_chan_remove_locked(struct nvkm_chan *); +int nvkm_chan_preempt(struct nvkm_chan *, bool wait); +int nvkm_chan_preempt_locked(struct nvkm_chan *, bool wait); +int nvkm_chan_cctx_get(struct nvkm_chan *, struct nvkm_engn *, struct nvkm_cctx **, + struct nvkm_client * /*TODO: remove need for this */); +void nvkm_chan_cctx_put(struct nvkm_chan *, struct nvkm_cctx **); +void nvkm_chan_cctx_bind(struct nvkm_chan *, struct nvkm_engn *, struct nvkm_cctx *); + +#define CHAN_PRCLI(c,l,p,f,a...) CGRP_PRINT((c)->cgrp, l, p, "%04x:[%s]"f, (c)->id, (c)->name, ##a) +#define CHAN_PRINT(c,l,p,f,a...) CGRP_PRINT((c)->cgrp, l, p, "%04x:"f, (c)->id, ##a) +#define CHAN_ERROR(c,f,a...) CHAN_PRCLI((c), ERROR, err, " "f"\n", ##a) +#define CHAN_TRACE(c,f,a...) CHAN_PRINT((c), TRACE, info, " "f"\n", ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.c new file mode 100644 index 0000000000..23944d95ef --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.c @@ -0,0 +1,111 @@ +/* + * Copyright 2020 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "chid.h" + +void +nvkm_chid_put(struct nvkm_chid *chid, int id, spinlock_t *data_lock) +{ + if (id >= 0) { + spin_lock_irq(&chid->lock); + spin_lock(data_lock); + chid->data[id] = NULL; + spin_unlock(data_lock); + clear_bit(id, chid->used); + spin_unlock_irq(&chid->lock); + } +} + +int +nvkm_chid_get(struct nvkm_chid *chid, void *data) +{ + int id = -1, cid; + + spin_lock_irq(&chid->lock); + cid = find_first_zero_bit(chid->used, chid->nr); + if (cid < chid->nr) { + set_bit(cid, chid->used); + chid->data[cid] = data; + id = cid; + } + spin_unlock_irq(&chid->lock); + return id; +} + +static void +nvkm_chid_del(struct kref *kref) +{ + struct nvkm_chid *chid = container_of(kref, typeof(*chid), kref); + + nvkm_event_fini(&chid->event); + + kvfree(chid->data); + kfree(chid); +} + +void +nvkm_chid_unref(struct nvkm_chid **pchid) +{ + struct nvkm_chid *chid = *pchid; + + if (!chid) + return; + + kref_put(&chid->kref, nvkm_chid_del); + *pchid = NULL; +} + +struct nvkm_chid * +nvkm_chid_ref(struct nvkm_chid *chid) +{ + if (chid) + kref_get(&chid->kref); + + return chid; +} + +int +nvkm_chid_new(const struct nvkm_event_func *func, struct nvkm_subdev *subdev, + int nr, int first, int count, struct nvkm_chid **pchid) +{ + struct nvkm_chid *chid; + int id; + + if (!(chid = *pchid = kzalloc(struct_size(chid, used, nr), GFP_KERNEL))) + return -ENOMEM; + + kref_init(&chid->kref); + chid->nr = nr; + chid->mask = chid->nr - 1; + spin_lock_init(&chid->lock); + + if (!(chid->data = kvzalloc(sizeof(*chid->data) * nr, GFP_KERNEL))) { + nvkm_chid_unref(pchid); + return -ENOMEM; + } + + for (id = 0; id < first; id++) + __set_bit(id, chid->used); + for (id = first + count; id < nr; id++) + __set_bit(id, chid->used); + + return nvkm_event_init(func, subdev, 1, nr, &chid->event); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.h new file mode 100644 index 0000000000..2a42efb184 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/chid.h @@ -0,0 +1,25 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_CHID_H__ +#define __NVKM_CHID_H__ +#include + +struct nvkm_chid { + struct kref kref; + int nr; + u32 mask; + + struct nvkm_event event; + + void **data; + + spinlock_t lock; + unsigned long used[]; +}; + +int nvkm_chid_new(const struct nvkm_event_func *, struct nvkm_subdev *, + int nr, int first, int count, struct nvkm_chid **pchid); +struct nvkm_chid *nvkm_chid_ref(struct nvkm_chid *); +void nvkm_chid_unref(struct nvkm_chid **); +int nvkm_chid_get(struct nvkm_chid *, void *data); +void nvkm_chid_put(struct nvkm_chid *, int id, spinlock_t *data_lock); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c new file mode 100644 index 0000000000..6b229a3fbd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c @@ -0,0 +1,231 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "runl.h" + +#include +#include + +#include + +static void +g84_chan_bind(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x002600 + (chan->id * 4), chan->ramfc->addr >> 8); +} + +static int +g84_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + const u32 limit2 = ilog2(length / 8); + int ret; + + ret = nvkm_gpuobj_new(device, 0x0200, 0, true, chan->inst, &chan->eng); + if (ret) + return ret; + + ret = nvkm_gpuobj_new(device, 0x4000, 0, false, chan->inst, &chan->pgd); + if (ret) + return ret; + + ret = nvkm_gpuobj_new(device, 0x1000, 0x400, true, chan->inst, &chan->cache); + if (ret) + return ret; + + ret = nvkm_gpuobj_new(device, 0x100, 0x100, true, chan->inst, &chan->ramfc); + if (ret) + return ret; + + ret = nvkm_ramht_new(device, 0x8000, 16, chan->inst, &chan->ramht); + if (ret) + return ret; + + nvkm_kmap(chan->ramfc); + nvkm_wo32(chan->ramfc, 0x3c, 0x403f6078); + nvkm_wo32(chan->ramfc, 0x44, 0x01003fff); + nvkm_wo32(chan->ramfc, 0x48, chan->push->node->offset >> 4); + nvkm_wo32(chan->ramfc, 0x50, lower_32_bits(offset)); + nvkm_wo32(chan->ramfc, 0x54, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->ramfc, 0x60, 0x7fffffff); + nvkm_wo32(chan->ramfc, 0x78, 0x00000000); + nvkm_wo32(chan->ramfc, 0x7c, 0x30000000 | devm); + nvkm_wo32(chan->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) | + (4 << 24) /* SEARCH_FULL */ | + (chan->ramht->gpuobj->node->offset >> 4)); + nvkm_wo32(chan->ramfc, 0x88, chan->cache->addr >> 10); + nvkm_wo32(chan->ramfc, 0x98, chan->inst->addr >> 12); + nvkm_done(chan->ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +g84_chan_ramfc = { + .write = g84_chan_ramfc_write, + .ctxdma = true, + .devm = 0xfff, +}; + +const struct nvkm_chan_func +g84_chan = { + .inst = &nv50_chan_inst, + .userd = &nv50_chan_userd, + .ramfc = &g84_chan_ramfc, + .bind = g84_chan_bind, + .unbind = nv50_chan_unbind, + .start = nv50_chan_start, + .stop = nv50_chan_stop, +}; + +static void +g84_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->cgrp->runl->fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u64 start = 0, limit = 0; + u32 flags = 0, ptr0, save; + + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_GR : ptr0 = 0x0020; break; + case NVKM_ENGINE_VP : + case NVKM_ENGINE_MSPDEC: ptr0 = 0x0040; break; + case NVKM_ENGINE_MPEG : + case NVKM_ENGINE_MSPPP : ptr0 = 0x0060; break; + case NVKM_ENGINE_BSP : + case NVKM_ENGINE_MSVLD : ptr0 = 0x0080; break; + case NVKM_ENGINE_CIPHER: + case NVKM_ENGINE_SEC : ptr0 = 0x00a0; break; + case NVKM_ENGINE_CE : ptr0 = 0x00c0; break; + default: + WARN_ON(1); + return; + } + + if (!cctx) { + save = nvkm_mask(device, 0x002520, 0x0000003f, BIT(engn->id - 1)); + nvkm_wr32(device, 0x0032fc, chan->inst->addr >> 12); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x0032fc) != 0xffffffff) + break; + ); + nvkm_wr32(device, 0x002520, save); + } else { + flags = 0x00190000; + start = cctx->vctx->inst->addr; + limit = start + cctx->vctx->inst->size - 1; + } + + nvkm_kmap(chan->eng); + nvkm_wo32(chan->eng, ptr0 + 0x00, flags); + nvkm_wo32(chan->eng, ptr0 + 0x04, lower_32_bits(limit)); + nvkm_wo32(chan->eng, ptr0 + 0x08, lower_32_bits(start)); + nvkm_wo32(chan->eng, ptr0 + 0x0c, upper_32_bits(limit) << 24 | + lower_32_bits(start)); + nvkm_wo32(chan->eng, ptr0 + 0x10, 0x00000000); + nvkm_wo32(chan->eng, ptr0 + 0x14, 0x00000000); + nvkm_done(chan->eng); +} + +const struct nvkm_engn_func +g84_engn = { + .bind = g84_ectx_bind, + .ramht_add = nv50_eobj_ramht_add, + .ramht_del = nv50_eobj_ramht_del, +}; + +static void +g84_fifo_nonstall_block(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + nvkm_mask(fifo->engine.subdev.device, 0x002140, 0x40000000, 0x00000000); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +static void +g84_fifo_nonstall_allow(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + nvkm_mask(fifo->engine.subdev.device, 0x002140, 0x40000000, 0x40000000); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +const struct nvkm_event_func +g84_fifo_nonstall = { + .init = g84_fifo_nonstall_allow, + .fini = g84_fifo_nonstall_block, +}; + +static int +g84_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl; + + runl = nvkm_runl_new(fifo, 0, 0, 0); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_SW, 0); + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_DMAOBJ, 0); + nvkm_runl_add(runl, 1, fifo->func->engn, NVKM_ENGINE_GR, 0); + nvkm_runl_add(runl, 2, fifo->func->engn, NVKM_ENGINE_MPEG, 0); + nvkm_runl_add(runl, 3, fifo->func->engn, NVKM_ENGINE_ME, 0); + nvkm_runl_add(runl, 4, fifo->func->engn, NVKM_ENGINE_VP, 0); + nvkm_runl_add(runl, 5, fifo->func->engn, NVKM_ENGINE_CIPHER, 0); + nvkm_runl_add(runl, 6, fifo->func->engn, NVKM_ENGINE_BSP, 0); + return 0; +} + +static const struct nvkm_fifo_func +g84_fifo = { + .chid_nr = nv50_fifo_chid_nr, + .chid_ctor = nv50_fifo_chid_ctor, + .runl_ctor = g84_fifo_runl_ctor, + .init = nv50_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .nonstall = &g84_fifo_nonstall, + .runl = &nv50_runl, + .engn = &g84_engn, + .engn_sw = &nv50_engn_sw, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, G82_CHANNEL_GPFIFO }, &g84_chan }, +}; + +int +g84_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&g84_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g98.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g98.c new file mode 100644 index 0000000000..c6ca050c38 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g98.c @@ -0,0 +1,70 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "runl.h" + +#include + +static int +g98_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl; + + runl = nvkm_runl_new(fifo, 0, 0, 0); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_SW, 0); + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_DMAOBJ, 0); + nvkm_runl_add(runl, 1, fifo->func->engn, NVKM_ENGINE_GR, 0); + nvkm_runl_add(runl, 2, fifo->func->engn, NVKM_ENGINE_MSPPP, 0); + nvkm_runl_add(runl, 3, fifo->func->engn, NVKM_ENGINE_CE, 0); + nvkm_runl_add(runl, 4, fifo->func->engn, NVKM_ENGINE_MSPDEC, 0); + nvkm_runl_add(runl, 5, fifo->func->engn, NVKM_ENGINE_SEC, 0); + nvkm_runl_add(runl, 6, fifo->func->engn, NVKM_ENGINE_MSVLD, 0); + return 0; +} + +static const struct nvkm_fifo_func +g98_fifo = { + .chid_nr = nv50_fifo_chid_nr, + .chid_ctor = nv50_fifo_chid_ctor, + .runl_ctor = g98_fifo_runl_ctor, + .init = nv50_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .nonstall = &g84_fifo_nonstall, + .runl = &nv50_runl, + .engn = &g84_engn, + .engn_sw = &nv50_engn_sw, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, G82_CHANNEL_GPFIFO }, &g84_chan }, +}; + +int +g98_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&g98_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga100.c new file mode 100644 index 0000000000..c56d2a839e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga100.c @@ -0,0 +1,611 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" +#include "runq.h" + +#include +#include +#include + +#include + +static u32 +ga100_chan_doorbell_handle(struct nvkm_chan *chan) +{ + return (chan->cgrp->runl->doorbell << 16) | chan->id; +} + +static void +ga100_chan_stop(struct nvkm_chan *chan) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + + nvkm_wr32(runl->fifo->engine.subdev.device, runl->chan + (chan->id * 4), 0x00000003); +} + +static void +ga100_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + struct nvkm_device *device = runl->fifo->engine.subdev.device; + const int gfid = 0; + + nvkm_wr32(device, runl->chan + (chan->id * 4), 0x00000002); + nvkm_wr32(device, runl->addr + 0x0090, (gfid << 16) | chan->id); /* INTERNAL_DOORBELL. */ +} + +static void +ga100_chan_unbind(struct nvkm_chan *chan) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + + nvkm_wr32(runl->fifo->engine.subdev.device, runl->chan + (chan->id * 4), 0xffffffff); +} + +static int +ga100_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + const u32 limit2 = ilog2(length / 8); + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x010, 0x0000face); + nvkm_wo32(chan->inst, 0x030, 0x7ffff902); + nvkm_wo32(chan->inst, 0x048, lower_32_bits(offset)); + nvkm_wo32(chan->inst, 0x04c, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->inst, 0x084, 0x20400000); + nvkm_wo32(chan->inst, 0x094, 0x30000000 | devm); + nvkm_wo32(chan->inst, 0x0e4, priv ? 0x00000020 : 0x00000000); + nvkm_wo32(chan->inst, 0x0e8, chan->id); + nvkm_wo32(chan->inst, 0x0f4, 0x00001000 | (priv ? 0x00000100 : 0x00000000)); + nvkm_wo32(chan->inst, 0x0f8, 0x80000000 | chan->cgrp->runl->nonstall.vector); + nvkm_mo32(chan->inst, 0x218, 0x00000000, 0x00000000); + nvkm_done(chan->inst); + return 0; +} + +static const struct nvkm_chan_func_ramfc +ga100_chan_ramfc = { + .write = ga100_chan_ramfc_write, + .devm = 0xfff, + .priv = true, +}; + +const struct nvkm_chan_func +ga100_chan = { + .inst = &gf100_chan_inst, + .userd = &gv100_chan_userd, + .ramfc = &ga100_chan_ramfc, + .unbind = ga100_chan_unbind, + .start = ga100_chan_start, + .stop = ga100_chan_stop, + .preempt = gk110_chan_preempt, + .doorbell_handle = ga100_chan_doorbell_handle, +}; + +static void +ga100_cgrp_preempt(struct nvkm_cgrp *cgrp) +{ + struct nvkm_runl *runl = cgrp->runl; + + nvkm_wr32(runl->fifo->engine.subdev.device, runl->addr + 0x098, 0x01000000 | cgrp->id); +} + +const struct nvkm_cgrp_func +ga100_cgrp = { + .preempt = ga100_cgrp_preempt, +}; + +static int +ga100_engn_cxid(struct nvkm_engn *engn, bool *cgid) +{ + struct nvkm_runl *runl = engn->runl; + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u32 stat = nvkm_rd32(device, runl->addr + 0x200 + engn->id * 0x40); + + ENGN_DEBUG(engn, "status %08x", stat); + *cgid = true; + + switch ((stat & 0x0000e000) >> 13) { + case 0 /* INVALID */: return -ENODEV; + case 1 /* VALID */: + case 5 /* SAVE */: return (stat & 0x00000fff); + case 6 /* LOAD */: return (stat & 0x0fff0000) >> 16; + case 7 /* SWITCH */: + if (nvkm_engine_chsw_load(engn->engine)) + return (stat & 0x0fff0000) >> 16; + return (stat & 0x00000fff); + default: + WARN_ON(1); + break; + } + + return -ENODEV; +} + +static int +ga100_engn_nonstall(struct nvkm_engn *engn) +{ + struct nvkm_engine *engine = engn->engine; + + if (WARN_ON(!engine->func->nonstall)) + return -EINVAL; + + return engine->func->nonstall(engine); +} + +const struct nvkm_engn_func +ga100_engn = { + .nonstall = ga100_engn_nonstall, + .cxid = ga100_engn_cxid, + .ctor = gk104_ectx_ctor, + .bind = gv100_ectx_bind, +}; + +const struct nvkm_engn_func +ga100_engn_ce = { + .nonstall = ga100_engn_nonstall, + .cxid = ga100_engn_cxid, + .ctor = gv100_ectx_ce_ctor, + .bind = gv100_ectx_ce_bind, +}; + +static bool +ga100_runq_idle(struct nvkm_runq *runq) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + + return !(nvkm_rd32(device, 0x04015c + (runq->id * 0x800)) & 0x0000e000); +} + +static bool +ga100_runq_intr_1(struct nvkm_runq *runq, struct nvkm_runl *runl) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + u32 inte = nvkm_rd32(device, 0x040180 + (runq->id * 0x800)); + u32 intr = nvkm_rd32(device, 0x040148 + (runq->id * 0x800)); + u32 stat = intr & inte; + + if (!stat) { + RUNQ_DEBUG(runq, "inte1 %08x %08x", intr, inte); + return false; + } + + if (stat & 0x80000000) { + u32 chid = nvkm_rd32(device, 0x040120 + (runq->id * 0x0800)) & runl->chid->mask; + struct nvkm_chan *chan; + unsigned long flags; + + RUNQ_ERROR(runq, "CTXNOTVALID chid:%d", chid); + chan = nvkm_runl_chan_get_chid(runl, chid, &flags); + if (chan) { + nvkm_chan_error(chan, true); + nvkm_chan_put(&chan, flags); + } + + nvkm_mask(device, 0x0400ac + (runq->id * 0x800), 0x00030000, 0x00030000); + stat &= ~0x80000000; + } + + if (stat) { + RUNQ_ERROR(runq, "intr1 %08x", stat); + nvkm_wr32(device, 0x0401a0 + (runq->id * 0x800), stat); + } + + nvkm_wr32(device, 0x040148 + (runq->id * 0x800), intr); + return true; +} + +static bool +ga100_runq_intr_0(struct nvkm_runq *runq, struct nvkm_runl *runl) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + u32 inte = nvkm_rd32(device, 0x040170 + (runq->id * 0x800)); + u32 intr = nvkm_rd32(device, 0x040108 + (runq->id * 0x800)); + u32 stat = intr & inte; + + if (!stat) { + RUNQ_DEBUG(runq, "inte0 %08x %08x", intr, inte); + return false; + } + + /*TODO: expand on this when fixing up gf100's version. */ + if (stat & 0xc6afe000) { + u32 chid = nvkm_rd32(device, 0x040120 + (runq->id * 0x0800)) & runl->chid->mask; + struct nvkm_chan *chan; + unsigned long flags; + + RUNQ_ERROR(runq, "intr0 %08x", stat); + chan = nvkm_runl_chan_get_chid(runl, chid, &flags); + if (chan) { + nvkm_chan_error(chan, true); + nvkm_chan_put(&chan, flags); + } + + stat &= ~0xc6afe000; + } + + if (stat) { + RUNQ_ERROR(runq, "intr0 %08x", stat); + nvkm_wr32(device, 0x040190 + (runq->id * 0x800), stat); + } + + nvkm_wr32(device, 0x040108 + (runq->id * 0x800), intr); + return true; +} + +static bool +ga100_runq_intr(struct nvkm_runq *runq, struct nvkm_runl *runl) +{ + bool intr0 = ga100_runq_intr_0(runq, runl); + bool intr1 = ga100_runq_intr_1(runq, runl); + + return intr0 || intr1; +} + +static void +ga100_runq_init(struct nvkm_runq *runq) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x040108 + (runq->id * 0x800), 0xffffffff); /* INTR_0 */ + nvkm_wr32(device, 0x040148 + (runq->id * 0x800), 0xffffffff); /* INTR_1 */ + nvkm_wr32(device, 0x040170 + (runq->id * 0x800), 0xffffffff); /* INTR_0_EN_SET_TREE */ + nvkm_wr32(device, 0x040180 + (runq->id * 0x800), 0xffffffff); /* INTR_1_EN_SET_TREE */ +} + +const struct nvkm_runq_func +ga100_runq = { + .init = ga100_runq_init, + .intr = ga100_runq_intr, + .idle = ga100_runq_idle, +}; + +static bool +ga100_runl_preempt_pending(struct nvkm_runl *runl) +{ + return nvkm_rd32(runl->fifo->engine.subdev.device, runl->addr + 0x098) & 0x00100000; +} + +static void +ga100_runl_preempt(struct nvkm_runl *runl) +{ + nvkm_wr32(runl->fifo->engine.subdev.device, runl->addr + 0x098, 0x00000000); +} + +static void +ga100_runl_allow(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, runl->addr + 0x094, 0x00000001, 0x00000000); +} + +static void +ga100_runl_block(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, runl->addr + 0x094, 0x00000001, 0x00000001); +} + +static bool +ga100_runl_pending(struct nvkm_runl *runl) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + + return nvkm_rd32(device, runl->addr + 0x08c) & 0x00008000; +} + +static void +ga100_runl_commit(struct nvkm_runl *runl, struct nvkm_memory *memory, u32 start, int count) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u64 addr = nvkm_memory_addr(memory) + start; + + nvkm_wr32(device, runl->addr + 0x080, lower_32_bits(addr)); + nvkm_wr32(device, runl->addr + 0x084, upper_32_bits(addr)); + nvkm_wr32(device, runl->addr + 0x088, count); +} + +static irqreturn_t +ga100_runl_intr(struct nvkm_inth *inth) +{ + struct nvkm_runl *runl = container_of(inth, typeof(*runl), inth); + struct nvkm_engn *engn; + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u32 inte = nvkm_rd32(device, runl->addr + 0x120); + u32 intr = nvkm_rd32(device, runl->addr + 0x100); + u32 stat = intr & inte; + u32 info; + + if (!stat) { + RUNL_DEBUG(runl, "inte %08x %08x", intr, inte); + return IRQ_NONE; + } + + if (stat & 0x00000007) { + nvkm_runl_foreach_engn_cond(engn, runl, stat & BIT(engn->id)) { + info = nvkm_rd32(device, runl->addr + 0x224 + (engn->id * 0x40)); + + tu102_fifo_intr_ctxsw_timeout_info(engn, info); + + nvkm_wr32(device, runl->addr + 0x100, BIT(engn->id)); + stat &= ~BIT(engn->id); + } + } + + if (stat & 0x00000300) { + nvkm_wr32(device, runl->addr + 0x100, stat & 0x00000300); + stat &= ~0x00000300; + } + + if (stat & 0x00010000) { + if (runl->runq[0]) { + if (runl->runq[0]->func->intr(runl->runq[0], runl)) + stat &= ~0x00010000; + } + } + + if (stat & 0x00020000) { + if (runl->runq[1]) { + if (runl->runq[1]->func->intr(runl->runq[1], runl)) + stat &= ~0x00020000; + } + } + + if (stat) { + RUNL_ERROR(runl, "intr %08x", stat); + nvkm_wr32(device, runl->addr + 0x140, stat); + } + + nvkm_wr32(device, runl->addr + 0x180, 0x00000001); + return IRQ_HANDLED; +} + +static void +ga100_runl_fini(struct nvkm_runl *runl) +{ + nvkm_mask(runl->fifo->engine.subdev.device, runl->addr + 0x300, 0x80000000, 0x00000000); + nvkm_inth_block(&runl->inth); +} + +static void +ga100_runl_init(struct nvkm_runl *runl) +{ + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_runq *runq; + struct nvkm_device *device = fifo->engine.subdev.device; + int i; + + /* Submit NULL runlist and preempt. */ + nvkm_wr32(device, runl->addr + 0x088, 0x00000000); + runl->func->preempt(runl); + + /* Enable doorbell. */ + nvkm_mask(device, runl->addr + 0x300, 0x80000000, 0x80000000); + + nvkm_wr32(device, runl->addr + 0x100, 0xffffffff); /* INTR_0 */ + nvkm_wr32(device, runl->addr + 0x140, 0xffffffff); /* INTR_0_EN_CLEAR_TREE(0) */ + nvkm_wr32(device, runl->addr + 0x120, 0x000f1307); /* INTR_0_EN_SET_TREE(0) */ + nvkm_wr32(device, runl->addr + 0x148, 0xffffffff); /* INTR_0_EN_CLEAR_TREE(1) */ + nvkm_wr32(device, runl->addr + 0x128, 0x00000000); /* INTR_0_EN_SET_TREE(1) */ + + /* Init PBDMA(s). */ + for (i = 0; i < runl->runq_nr; i++) { + runq = runl->runq[i]; + runq->func->init(runq); + } + + nvkm_inth_allow(&runl->inth); +} + +const struct nvkm_runl_func +ga100_runl = { + .init = ga100_runl_init, + .fini = ga100_runl_fini, + .size = 16, + .update = nv50_runl_update, + .insert_cgrp = gv100_runl_insert_cgrp, + .insert_chan = gv100_runl_insert_chan, + .commit = ga100_runl_commit, + .wait = nv50_runl_wait, + .pending = ga100_runl_pending, + .block = ga100_runl_block, + .allow = ga100_runl_allow, + .preempt = ga100_runl_preempt, + .preempt_pending = ga100_runl_preempt_pending, +}; + +static int +ga100_runl_new(struct nvkm_fifo *fifo, int id, u32 addr, struct nvkm_runl **prunl) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_top_device *tdev; + struct nvkm_runl *runl; + struct nvkm_engn *engn; + u32 chcfg = nvkm_rd32(device, addr + 0x004); + u32 chnum = 1 << (chcfg & 0x0000000f); + u32 chaddr = (chcfg & 0xfffffff0); + u32 dbcfg = nvkm_rd32(device, addr + 0x008); + u32 vector = nvkm_rd32(device, addr + 0x160); + int i, ret; + + runl = nvkm_runl_new(fifo, id, addr, chnum); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + *prunl = runl; + + for (i = 0; i < 2; i++) { + u32 pbcfg = nvkm_rd32(device, addr + 0x010 + (i * 0x04)); + if (pbcfg & 0x80000000) { + runl->runq[runl->runq_nr] = + nvkm_runq_new(fifo, ((pbcfg & 0x03fffc00) - 0x040000) / 0x800); + if (!runl->runq[runl->runq_nr]) { + RUNL_ERROR(runl, "runq %d", runl->runq_nr); + return -ENOMEM; + } + + runl->runq_nr++; + } + } + + nvkm_list_foreach(tdev, &device->top->device, head, tdev->runlist == runl->addr) { + if (tdev->engine < 0) { + RUNL_DEBUG(runl, "engn !top"); + return -EINVAL; + } + + engn = nvkm_runl_add(runl, tdev->engine, (tdev->type == NVKM_ENGINE_CE) ? + fifo->func->engn_ce : fifo->func->engn, + tdev->type, tdev->inst); + if (!engn) + return -EINVAL; + + if (!engn->engine->func->nonstall) { + RUNL_DEBUG(runl, "engn %s !nonstall", engn->engine->subdev.name); + return -EINVAL; + } + } + + if (list_empty(&runl->engns)) { + RUNL_DEBUG(runl, "!engns"); + return -EINVAL; + } + + ret = nvkm_inth_add(&device->vfn->intr, vector & 0x00000fff, NVKM_INTR_PRIO_NORMAL, + &fifo->engine.subdev, ga100_runl_intr, &runl->inth); + if (ret) { + RUNL_ERROR(runl, "inth %d", ret); + return ret; + } + + runl->chan = chaddr; + runl->doorbell = dbcfg >> 16; + return 0; +} + +static irqreturn_t +ga100_fifo_nonstall_intr(struct nvkm_inth *inth) +{ + struct nvkm_runl *runl = container_of(inth, typeof(*runl), nonstall.inth); + + nvkm_event_ntfy(&runl->fifo->nonstall.event, runl->id, NVKM_FIFO_NONSTALL_EVENT); + return IRQ_HANDLED; +} + +static void +ga100_fifo_nonstall_block(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + struct nvkm_runl *runl = nvkm_runl_get(fifo, index, 0); + + nvkm_inth_block(&runl->nonstall.inth); +} + +static void +ga100_fifo_nonstall_allow(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + struct nvkm_runl *runl = nvkm_runl_get(fifo, index, 0); + + nvkm_inth_allow(&runl->nonstall.inth); +} + +const struct nvkm_event_func +ga100_fifo_nonstall = { + .init = ga100_fifo_nonstall_allow, + .fini = ga100_fifo_nonstall_block, +}; + +int +ga100_fifo_nonstall_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_vfn *vfn = subdev->device->vfn; + struct nvkm_runl *runl; + int ret, nr = 0; + + nvkm_runl_foreach(runl, fifo) { + struct nvkm_engn *engn = list_first_entry(&runl->engns, typeof(*engn), head); + + runl->nonstall.vector = engn->func->nonstall(engn); + if (runl->nonstall.vector < 0) { + RUNL_ERROR(runl, "nonstall %d", runl->nonstall.vector); + return runl->nonstall.vector; + } + + ret = nvkm_inth_add(&vfn->intr, runl->nonstall.vector, NVKM_INTR_PRIO_NORMAL, + subdev, ga100_fifo_nonstall_intr, &runl->nonstall.inth); + if (ret) + return ret; + + nr = max(nr, runl->id + 1); + } + + return nr; +} + +int +ga100_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_top_device *tdev; + struct nvkm_runl *runl; + int id = 0, ret; + + nvkm_list_foreach(tdev, &device->top->device, head, tdev->runlist >= 0) { + runl = nvkm_runl_get(fifo, -1, tdev->runlist); + if (!runl) { + ret = ga100_runl_new(fifo, id++, tdev->runlist, &runl); + if (ret) { + if (runl) + nvkm_runl_del(runl); + + continue; + } + } + } + + return 0; +} + +static const struct nvkm_fifo_func +ga100_fifo = { + .runl_ctor = ga100_fifo_runl_ctor, + .mmu_fault = &tu102_fifo_mmu_fault, + .nonstall_ctor = ga100_fifo_nonstall_ctor, + .nonstall = &ga100_fifo_nonstall, + .runl = &ga100_runl, + .runq = &ga100_runq, + .engn = &ga100_engn, + .engn_ce = &ga100_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &ga100_cgrp, .force = true }, + .chan = {{ 0, 0, AMPERE_CHANNEL_GPFIFO_A }, &ga100_chan }, +}; + +int +ga100_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&ga100_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c new file mode 100644 index 0000000000..2cdf5da339 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c @@ -0,0 +1,45 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +static const struct nvkm_fifo_func +ga102_fifo = { + .runl_ctor = ga100_fifo_runl_ctor, + .mmu_fault = &tu102_fifo_mmu_fault, + .nonstall_ctor = ga100_fifo_nonstall_ctor, + .nonstall = &ga100_fifo_nonstall, + .runl = &ga100_runl, + .runq = &ga100_runq, + .engn = &ga100_engn, + .engn_ce = &ga100_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &ga100_cgrp, .force = true }, + .chan = {{ 0, 0, AMPERE_CHANNEL_GPFIFO_B }, &ga100_chan }, +}; + +int +ga102_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&ga102_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c new file mode 100644 index 0000000000..6c94451d0f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c @@ -0,0 +1,969 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" +#include "runq.h" + +#include +#include +#include +#include +#include +#include + +#include + +void +gf100_chan_preempt(struct nvkm_chan *chan) +{ + nvkm_wr32(chan->cgrp->runl->fifo->engine.subdev.device, 0x002634, chan->id); +} + +static void +gf100_chan_stop(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x003004 + (chan->id * 8), 0x00000001, 0x00000000); +} + +static void +gf100_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x003004 + (chan->id * 8), 0x001f0001); +} + +static void gf100_fifo_intr_engine(struct nvkm_fifo *); + +static void +gf100_chan_unbind(struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + + /*TODO: Is this cargo-culted, or necessary? RM does *something* here... Why? */ + gf100_fifo_intr_engine(fifo); + + nvkm_wr32(device, 0x003000 + (chan->id * 8), 0x00000000); +} + +static void +gf100_chan_bind(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x003000 + (chan->id * 8), 0xc0000000 | chan->inst->addr >> 12); +} + +static int +gf100_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + const u64 userd = nvkm_memory_addr(chan->userd.mem) + chan->userd.base; + const u32 limit2 = ilog2(length / 8); + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x08, lower_32_bits(userd)); + nvkm_wo32(chan->inst, 0x0c, upper_32_bits(userd)); + nvkm_wo32(chan->inst, 0x10, 0x0000face); + nvkm_wo32(chan->inst, 0x30, 0xfffff902); + nvkm_wo32(chan->inst, 0x48, lower_32_bits(offset)); + nvkm_wo32(chan->inst, 0x4c, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->inst, 0x54, 0x00000002); + nvkm_wo32(chan->inst, 0x84, 0x20400000); + nvkm_wo32(chan->inst, 0x94, 0x30000000 | devm); + nvkm_wo32(chan->inst, 0x9c, 0x00000100); + nvkm_wo32(chan->inst, 0xa4, 0x1f1f1f1f); + nvkm_wo32(chan->inst, 0xa8, 0x1f1f1f1f); + nvkm_wo32(chan->inst, 0xac, 0x0000001f); + nvkm_wo32(chan->inst, 0xb8, 0xf8000000); + nvkm_wo32(chan->inst, 0xf8, 0x10003080); /* 0x002310 */ + nvkm_wo32(chan->inst, 0xfc, 0x10000010); /* 0x002350 */ + nvkm_done(chan->inst); + return 0; +} + +static const struct nvkm_chan_func_ramfc +gf100_chan_ramfc = { + .write = gf100_chan_ramfc_write, + .devm = 0xfff, +}; + +void +gf100_chan_userd_clear(struct nvkm_chan *chan) +{ + nvkm_kmap(chan->userd.mem); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x040, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x044, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x048, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x04c, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x050, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x058, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x05c, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x060, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x088, 0x00000000); + nvkm_wo32(chan->userd.mem, chan->userd.base + 0x08c, 0x00000000); + nvkm_done(chan->userd.mem); +} + +static const struct nvkm_chan_func_userd +gf100_chan_userd = { + .bar = 1, + .size = 0x1000, + .clear = gf100_chan_userd_clear, +}; + +const struct nvkm_chan_func_inst +gf100_chan_inst = { + .size = 0x1000, + .zero = true, + .vmm = true, +}; + +static const struct nvkm_chan_func +gf100_chan = { + .inst = &gf100_chan_inst, + .userd = &gf100_chan_userd, + .ramfc = &gf100_chan_ramfc, + .bind = gf100_chan_bind, + .unbind = gf100_chan_unbind, + .start = gf100_chan_start, + .stop = gf100_chan_stop, + .preempt = gf100_chan_preempt, +}; + +static void +gf100_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + u64 addr = 0ULL; + u32 ptr0; + + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_SW : return; + case NVKM_ENGINE_GR : ptr0 = 0x0210; break; + case NVKM_ENGINE_CE : ptr0 = 0x0230 + (engn->engine->subdev.inst * 0x10); break; + case NVKM_ENGINE_MSPDEC: ptr0 = 0x0250; break; + case NVKM_ENGINE_MSPPP : ptr0 = 0x0260; break; + case NVKM_ENGINE_MSVLD : ptr0 = 0x0270; break; + default: + WARN_ON(1); + return; + } + + if (cctx) { + addr = cctx->vctx->vma->addr; + addr |= 4ULL; + } + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, ptr0 + 0, lower_32_bits(addr)); + nvkm_wo32(chan->inst, ptr0 + 4, upper_32_bits(addr)); + nvkm_done(chan->inst); +} + +static int +gf100_ectx_ctor(struct nvkm_engn *engn, struct nvkm_vctx *vctx) +{ + int ret; + + ret = nvkm_vmm_get(vctx->vmm, 12, vctx->inst->size, &vctx->vma); + if (ret) + return ret; + + return nvkm_memory_map(vctx->inst, 0, vctx->vmm, vctx->vma, NULL, 0); +} + +bool +gf100_engn_mmu_fault_triggered(struct nvkm_engn *engn) +{ + struct nvkm_runl *runl = engn->runl; + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + u32 data = nvkm_rd32(device, 0x002a30 + (engn->id * 4)); + + ENGN_DEBUG(engn, "%08x: mmu fault triggered", data); + if (!(data & 0x00000100)) + return false; + + spin_lock(&fifo->lock); + nvkm_mask(device, 0x002a30 + (engn->id * 4), 0x00000100, 0x00000000); + if (atomic_dec_and_test(&runl->rc_triggered)) + nvkm_mask(device, 0x002140, 0x00000100, 0x00000100); + spin_unlock(&fifo->lock); + return true; +} + +void +gf100_engn_mmu_fault_trigger(struct nvkm_engn *engn) +{ + struct nvkm_runl *runl = engn->runl; + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + + ENGN_DEBUG(engn, "triggering mmu fault on 0x%02x", engn->fault); + spin_lock(&fifo->lock); + if (atomic_inc_return(&runl->rc_triggered) == 1) + nvkm_mask(device, 0x002140, 0x00000100, 0x00000000); + nvkm_wr32(device, 0x002100, 0x00000100); + nvkm_wr32(device, 0x002a30 + (engn->id * 4), 0x00000100 | engn->fault); + spin_unlock(&fifo->lock); +} + +/*TODO: clean all this up. */ +struct gf100_engn_status { + bool busy; + bool save; + bool unk0; + bool unk1; + u8 chid; +}; + +static void +gf100_engn_status(struct nvkm_engn *engn, struct gf100_engn_status *status) +{ + u32 stat = nvkm_rd32(engn->engine->subdev.device, 0x002640 + (engn->id * 4)); + + status->busy = (stat & 0x10000000); + status->save = (stat & 0x00100000); + status->unk0 = (stat & 0x00004000); + status->unk1 = (stat & 0x00001000); + status->chid = (stat & 0x0000007f); + + ENGN_DEBUG(engn, "%08x: busy %d save %d unk0 %d unk1 %d chid %d", + stat, status->busy, status->save, status->unk0, status->unk1, status->chid); +} + +static int +gf100_engn_cxid(struct nvkm_engn *engn, bool *cgid) +{ + struct gf100_engn_status status; + + gf100_engn_status(engn, &status); + if (status.busy) { + *cgid = false; + return status.chid; + } + + return -ENODEV; +} + +static bool +gf100_engn_chsw(struct nvkm_engn *engn) +{ + struct gf100_engn_status status; + + gf100_engn_status(engn, &status); + if (status.busy && (status.unk0 || status.unk1)) + return true; + + return false; +} + +static const struct nvkm_engn_func +gf100_engn = { + .chsw = gf100_engn_chsw, + .cxid = gf100_engn_cxid, + .mmu_fault_trigger = gf100_engn_mmu_fault_trigger, + .mmu_fault_triggered = gf100_engn_mmu_fault_triggered, + .ctor = gf100_ectx_ctor, + .bind = gf100_ectx_bind, +}; + +const struct nvkm_engn_func +gf100_engn_sw = { +}; + +static const struct nvkm_bitfield +gf100_runq_intr_0_names[] = { +/* { 0x00008000, "" } seen with null ib push */ + { 0x00200000, "ILLEGAL_MTHD" }, + { 0x00800000, "EMPTY_SUBC" }, + {} +}; + +bool +gf100_runq_intr(struct nvkm_runq *runq, struct nvkm_runl *null) +{ + struct nvkm_subdev *subdev = &runq->fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, 0x04010c + (runq->id * 0x2000)); + u32 stat = nvkm_rd32(device, 0x040108 + (runq->id * 0x2000)) & mask; + u32 addr = nvkm_rd32(device, 0x0400c0 + (runq->id * 0x2000)); + u32 data = nvkm_rd32(device, 0x0400c4 + (runq->id * 0x2000)); + u32 chid = nvkm_rd32(device, 0x040120 + (runq->id * 0x2000)) & runq->fifo->chid->mask; + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00003ffc); + u32 show = stat; + struct nvkm_chan *chan; + unsigned long flags; + char msg[128]; + + if (stat & 0x00800000) { + if (device->sw) { + if (nvkm_sw_mthd(device->sw, chid, subc, mthd, data)) + show &= ~0x00800000; + } + } + + if (show) { + nvkm_snprintbf(msg, sizeof(msg), runq->func->intr_0_names, show); + chan = nvkm_chan_get_chid(&runq->fifo->engine, chid, &flags); + nvkm_error(subdev, "PBDMA%d: %08x [%s] ch %d [%010llx %s] " + "subc %d mthd %04x data %08x\n", + runq->id, show, msg, chid, chan ? chan->inst->addr : 0, + chan ? chan->name : "unknown", subc, mthd, data); + + /*TODO: use proper procedure for clearing each exception / debug output */ + if ((stat & 0xc67fe000) && chan) + nvkm_chan_error(chan, true); + nvkm_chan_put(&chan, flags); + } + + nvkm_wr32(device, 0x0400c0 + (runq->id * 0x2000), 0x80600008); + nvkm_wr32(device, 0x040108 + (runq->id * 0x2000), stat); + return true; +} + +void +gf100_runq_init(struct nvkm_runq *runq) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + + nvkm_mask(device, 0x04013c + (runq->id * 0x2000), 0x10000100, 0x00000000); + nvkm_wr32(device, 0x040108 + (runq->id * 0x2000), 0xffffffff); /* INTR */ + nvkm_wr32(device, 0x04010c + (runq->id * 0x2000), 0xfffffeff); /* INTREN */ +} + +static const struct nvkm_runq_func +gf100_runq = { + .init = gf100_runq_init, + .intr = gf100_runq_intr, + .intr_0_names = gf100_runq_intr_0_names, +}; + +bool +gf100_runl_preempt_pending(struct nvkm_runl *runl) +{ + return nvkm_rd32(runl->fifo->engine.subdev.device, 0x002634) & 0x00100000; +} + +static void +gf100_runl_fault_clear(struct nvkm_runl *runl) +{ + nvkm_mask(runl->fifo->engine.subdev.device, 0x00262c, 0x00000000, 0x00000000); +} + +static void +gf100_runl_allow(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, 0x002630, engm, 0x00000000); +} + +static void +gf100_runl_block(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, 0x002630, engm, engm); +} + +static bool +gf100_runl_pending(struct nvkm_runl *runl) +{ + return nvkm_rd32(runl->fifo->engine.subdev.device, 0x00227c) & 0x00100000; +} + +static void +gf100_runl_commit(struct nvkm_runl *runl, struct nvkm_memory *memory, u32 start, int count) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u64 addr = nvkm_memory_addr(memory) + start; + int target; + + switch (nvkm_memory_target(memory)) { + case NVKM_MEM_TARGET_VRAM: target = 0; break; + case NVKM_MEM_TARGET_NCOH: target = 3; break; + default: + WARN_ON(1); + return; + } + + nvkm_wr32(device, 0x002270, (target << 28) | (addr >> 12)); + nvkm_wr32(device, 0x002274, 0x01f00000 | count); +} + +static void +gf100_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0, chan->id); + nvkm_wo32(memory, offset + 4, 0x00000004); +} + +static const struct nvkm_runl_func +gf100_runl = { + .size = 8, + .update = nv50_runl_update, + .insert_chan = gf100_runl_insert_chan, + .commit = gf100_runl_commit, + .wait = nv50_runl_wait, + .pending = gf100_runl_pending, + .block = gf100_runl_block, + .allow = gf100_runl_allow, + .fault_clear = gf100_runl_fault_clear, + .preempt_pending = gf100_runl_preempt_pending, +}; + +static void +gf100_fifo_nonstall_allow(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + nvkm_mask(fifo->engine.subdev.device, 0x002140, 0x80000000, 0x80000000); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +static void +gf100_fifo_nonstall_block(struct nvkm_event *event, int type, int index) +{ + struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), nonstall.event); + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + nvkm_mask(fifo->engine.subdev.device, 0x002140, 0x80000000, 0x00000000); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +const struct nvkm_event_func +gf100_fifo_nonstall = { + .init = gf100_fifo_nonstall_allow, + .fini = gf100_fifo_nonstall_block, +}; + +static const struct nvkm_enum +gf100_fifo_mmu_fault_engine[] = { + { 0x00, "PGRAPH", NULL, NVKM_ENGINE_GR }, + { 0x03, "PEEPHOLE", NULL, NVKM_ENGINE_IFB }, + { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0x05, "BAR3", NULL, NVKM_SUBDEV_INSTMEM }, + { 0x07, "PFIFO" }, + { 0x10, "PMSVLD", NULL, NVKM_ENGINE_MSVLD }, + { 0x11, "PMSPPP", NULL, NVKM_ENGINE_MSPPP }, + { 0x13, "PCOUNTER" }, + { 0x14, "PMSPDEC", NULL, NVKM_ENGINE_MSPDEC }, + { 0x15, "PCE0", NULL, NVKM_ENGINE_CE, 0 }, + { 0x16, "PCE1", NULL, NVKM_ENGINE_CE, 1 }, + { 0x17, "PMU" }, + {} +}; + +static const struct nvkm_enum +gf100_fifo_mmu_fault_reason[] = { + { 0x00, "PT_NOT_PRESENT" }, + { 0x01, "PT_TOO_SHORT" }, + { 0x02, "PAGE_NOT_PRESENT" }, + { 0x03, "VM_LIMIT_EXCEEDED" }, + { 0x04, "NO_CHANNEL" }, + { 0x05, "PAGE_SYSTEM_ONLY" }, + { 0x06, "PAGE_READ_ONLY" }, + { 0x0a, "COMPRESSED_SYSRAM" }, + { 0x0c, "INVALID_STORAGE_TYPE" }, + {} +}; + +static const struct nvkm_enum +gf100_fifo_mmu_fault_hubclient[] = { + { 0x01, "PCOPY0" }, + { 0x02, "PCOPY1" }, + { 0x04, "DISPATCH" }, + { 0x05, "CTXCTL" }, + { 0x06, "PFIFO" }, + { 0x07, "BAR_READ" }, + { 0x08, "BAR_WRITE" }, + { 0x0b, "PVP" }, + { 0x0c, "PMSPPP" }, + { 0x0d, "PMSVLD" }, + { 0x11, "PCOUNTER" }, + { 0x12, "PMU" }, + { 0x14, "CCACHE" }, + { 0x15, "CCACHE_POST" }, + {} +}; + +static const struct nvkm_enum +gf100_fifo_mmu_fault_gpcclient[] = { + { 0x01, "TEX" }, + { 0x0c, "ESETUP" }, + { 0x0e, "CTXCTL" }, + { 0x0f, "PROP" }, + {} +}; + +const struct nvkm_enum +gf100_fifo_mmu_fault_access[] = { + { 0x00, "READ" }, + { 0x01, "WRITE" }, + {} +}; + +void +gf100_fifo_mmu_fault_recover(struct nvkm_fifo *fifo, struct nvkm_fault_data *info) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + const struct nvkm_enum *er, *ee, *ec, *ea; + struct nvkm_engine *engine = NULL; + struct nvkm_runl *runl; + struct nvkm_engn *engn; + struct nvkm_chan *chan; + unsigned long flags; + char ct[8] = "HUB/"; + + /* Lookup engine by MMU fault ID. */ + nvkm_runl_foreach(runl, fifo) { + engn = nvkm_runl_find_engn(engn, runl, engn->fault == info->engine); + if (engn) { + /* Fault triggered by CTXSW_TIMEOUT recovery procedure. */ + if (engn->func->mmu_fault_triggered && + engn->func->mmu_fault_triggered(engn)) { + nvkm_runl_rc_engn(runl, engn); + return; + } + + engine = engn->engine; + break; + } + } + + er = nvkm_enum_find(fifo->func->mmu_fault->reason, info->reason); + ee = nvkm_enum_find(fifo->func->mmu_fault->engine, info->engine); + if (info->hub) { + ec = nvkm_enum_find(fifo->func->mmu_fault->hubclient, info->client); + } else { + ec = nvkm_enum_find(fifo->func->mmu_fault->gpcclient, info->client); + snprintf(ct, sizeof(ct), "GPC%d/", info->gpc); + } + ea = nvkm_enum_find(fifo->func->mmu_fault->access, info->access); + + /* Handle BAR faults. */ + if (ee && ee->data2) { + switch (ee->data2) { + case NVKM_SUBDEV_BAR: + nvkm_bar_bar1_reset(device); + break; + case NVKM_SUBDEV_INSTMEM: + nvkm_bar_bar2_reset(device); + break; + case NVKM_ENGINE_IFB: + nvkm_mask(device, 0x001718, 0x00000000, 0x00000000); + break; + default: + break; + } + } + + chan = nvkm_chan_get_inst(&fifo->engine, info->inst, &flags); + + nvkm_error(subdev, + "fault %02x [%s] at %016llx engine %02x [%s] client %02x " + "[%s%s] reason %02x [%s] on channel %d [%010llx %s]\n", + info->access, ea ? ea->name : "", info->addr, + info->engine, ee ? ee->name : engine ? engine->subdev.name : "", + info->client, ct, ec ? ec->name : "", + info->reason, er ? er->name : "", + chan ? chan->id : -1, info->inst, chan ? chan->name : "unknown"); + + /* Handle host/engine faults. */ + if (chan) + nvkm_runl_rc_cgrp(chan->cgrp); + + nvkm_chan_put(&chan, flags); +} + +static const struct nvkm_fifo_func_mmu_fault +gf100_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gf100_fifo_mmu_fault_access, + .engine = gf100_fifo_mmu_fault_engine, + .reason = gf100_fifo_mmu_fault_reason, + .hubclient = gf100_fifo_mmu_fault_hubclient, + .gpcclient = gf100_fifo_mmu_fault_gpcclient, +}; + +void +gf100_fifo_intr_ctxsw_timeout(struct nvkm_fifo *fifo, u32 engm) +{ + struct nvkm_runl *runl; + struct nvkm_engn *engn, *engn2; + bool cgid, cgid2; + int id, id2; + + nvkm_runl_foreach(runl, fifo) { + /* Stop the runlist, and go through all engines serving it. */ + nvkm_runl_block(runl); + nvkm_runl_foreach_engn_cond(engn, runl, engm & BIT(engn->id)) { + /* Determine what channel (group) the engine is on. */ + id = engn->func->cxid(engn, &cgid); + if (id >= 0) { + /* Trigger MMU fault on any engine(s) on that channel (group). */ + nvkm_runl_foreach_engn_cond(engn2, runl, engn2->func->cxid) { + id2 = engn2->func->cxid(engn2, &cgid2); + if (cgid2 == cgid && id2 == id) + engn2->func->mmu_fault_trigger(engn2); + } + } + } + nvkm_runl_allow(runl); /* HW will keep runlist blocked via ERROR_SCHED_DISABLE. */ + } +} + +static void +gf100_fifo_intr_sched_ctxsw(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl; + struct nvkm_engn *engn; + u32 engm = 0; + + /* Look for any engines that are busy, and awaiting chsw ack. */ + nvkm_runl_foreach(runl, fifo) { + nvkm_runl_foreach_engn_cond(engn, runl, engn->func->chsw) { + if (WARN_ON(engn->fault < 0) || !engn->func->chsw(engn)) + continue; + + engm |= BIT(engn->id); + } + } + + if (!engm) + return; + + fifo->func->intr_ctxsw_timeout(fifo, engm); +} + +static const struct nvkm_enum +gf100_fifo_intr_sched_names[] = { + { 0x0a, "CTXSW_TIMEOUT" }, + {} +}; + +void +gf100_fifo_intr_sched(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 intr = nvkm_rd32(device, 0x00254c); + u32 code = intr & 0x000000ff; + const struct nvkm_enum *en; + + en = nvkm_enum_find(gf100_fifo_intr_sched_names, code); + + nvkm_error(subdev, "SCHED_ERROR %02x [%s]\n", code, en ? en->name : ""); + + switch (code) { + case 0x0a: + gf100_fifo_intr_sched_ctxsw(fifo); + break; + default: + break; + } +} + +void +gf100_fifo_intr_mmu_fault_unit(struct nvkm_fifo *fifo, int unit) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + u32 inst = nvkm_rd32(device, 0x002800 + (unit * 0x10)); + u32 valo = nvkm_rd32(device, 0x002804 + (unit * 0x10)); + u32 vahi = nvkm_rd32(device, 0x002808 + (unit * 0x10)); + u32 type = nvkm_rd32(device, 0x00280c + (unit * 0x10)); + struct nvkm_fault_data info; + + info.inst = (u64)inst << 12; + info.addr = ((u64)vahi << 32) | valo; + info.time = 0; + info.engine = unit; + info.valid = 1; + info.gpc = (type & 0x1f000000) >> 24; + info.client = (type & 0x00001f00) >> 8; + info.access = (type & 0x00000080) >> 7; + info.hub = (type & 0x00000040) >> 6; + info.reason = (type & 0x0000000f); + + nvkm_fifo_fault(fifo, &info); +} + +void +gf100_fifo_intr_mmu_fault(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + unsigned long mask = nvkm_rd32(device, 0x00259c); + int unit; + + for_each_set_bit(unit, &mask, 32) { + fifo->func->intr_mmu_fault_unit(fifo, unit); + nvkm_wr32(device, 0x00259c, BIT(unit)); + } +} + +bool +gf100_fifo_intr_pbdma(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_runq *runq; + u32 mask = nvkm_rd32(device, 0x0025a0); + bool handled = false; + + nvkm_runq_foreach_cond(runq, fifo, mask & BIT(runq->id)) { + if (runq->func->intr(runq, NULL)) + handled = true; + + nvkm_wr32(device, 0x0025a0, BIT(runq->id)); + } + + return handled; +} + +static void +gf100_fifo_intr_runlist(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 intr = nvkm_rd32(device, 0x002a00); + + if (intr & 0x10000000) { + nvkm_wr32(device, 0x002a00, 0x10000000); + intr &= ~0x10000000; + } + + if (intr) { + nvkm_error(subdev, "RUNLIST %08x\n", intr); + nvkm_wr32(device, 0x002a00, intr); + } +} + +static void +gf100_fifo_intr_engine_unit(struct nvkm_fifo *fifo, int engn) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 intr = nvkm_rd32(device, 0x0025a8 + (engn * 0x04)); + u32 inte = nvkm_rd32(device, 0x002628); + u32 unkn; + + nvkm_wr32(device, 0x0025a8 + (engn * 0x04), intr); + + for (unkn = 0; unkn < 8; unkn++) { + u32 ints = (intr >> (unkn * 0x04)) & inte; + if (ints & 0x1) { + nvkm_event_ntfy(&fifo->nonstall.event, 0, NVKM_FIFO_NONSTALL_EVENT); + ints &= ~1; + } + if (ints) { + nvkm_error(subdev, "ENGINE %d %d %01x", engn, unkn, ints); + nvkm_mask(device, 0x002628, ints, 0); + } + } +} + +static void +gf100_fifo_intr_engine(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + u32 mask = nvkm_rd32(device, 0x0025a4); + + while (mask) { + u32 unit = __ffs(mask); + gf100_fifo_intr_engine_unit(fifo, unit); + mask &= ~(1 << unit); + } +} + +static irqreturn_t +gf100_fifo_intr(struct nvkm_inth *inth) +{ + struct nvkm_fifo *fifo = container_of(inth, typeof(*fifo), engine.subdev.inth); + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, 0x002140); + u32 stat = nvkm_rd32(device, 0x002100) & mask; + + if (stat & 0x00000001) { + u32 intr = nvkm_rd32(device, 0x00252c); + nvkm_warn(subdev, "INTR 00000001: %08x\n", intr); + nvkm_wr32(device, 0x002100, 0x00000001); + stat &= ~0x00000001; + } + + if (stat & 0x00000100) { + gf100_fifo_intr_sched(fifo); + nvkm_wr32(device, 0x002100, 0x00000100); + stat &= ~0x00000100; + } + + if (stat & 0x00010000) { + u32 intr = nvkm_rd32(device, 0x00256c); + nvkm_warn(subdev, "INTR 00010000: %08x\n", intr); + nvkm_wr32(device, 0x002100, 0x00010000); + stat &= ~0x00010000; + } + + if (stat & 0x01000000) { + u32 intr = nvkm_rd32(device, 0x00258c); + nvkm_warn(subdev, "INTR 01000000: %08x\n", intr); + nvkm_wr32(device, 0x002100, 0x01000000); + stat &= ~0x01000000; + } + + if (stat & 0x10000000) { + gf100_fifo_intr_mmu_fault(fifo); + stat &= ~0x10000000; + } + + if (stat & 0x20000000) { + if (gf100_fifo_intr_pbdma(fifo)) + stat &= ~0x20000000; + } + + if (stat & 0x40000000) { + gf100_fifo_intr_runlist(fifo); + stat &= ~0x40000000; + } + + if (stat & 0x80000000) { + gf100_fifo_intr_engine(fifo); + stat &= ~0x80000000; + } + + if (stat) { + nvkm_error(subdev, "INTR %08x\n", stat); + spin_lock(&fifo->lock); + nvkm_mask(device, 0x002140, stat, 0x00000000); + spin_unlock(&fifo->lock); + nvkm_wr32(device, 0x002100, stat); + } + + return IRQ_HANDLED; +} + +static void +gf100_fifo_init_pbdmas(struct nvkm_fifo *fifo, u32 mask) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + + /* Enable PBDMAs. */ + nvkm_wr32(device, 0x000204, mask); + nvkm_wr32(device, 0x002204, mask); + + /* Assign engines to PBDMAs. */ + if ((mask & 7) == 7) { + nvkm_wr32(device, 0x002208, ~(1 << 0)); /* PGRAPH */ + nvkm_wr32(device, 0x00220c, ~(1 << 1)); /* PVP */ + nvkm_wr32(device, 0x002210, ~(1 << 1)); /* PMSPP */ + nvkm_wr32(device, 0x002214, ~(1 << 1)); /* PMSVLD */ + nvkm_wr32(device, 0x002218, ~(1 << 2)); /* PCE0 */ + nvkm_wr32(device, 0x00221c, ~(1 << 1)); /* PCE1 */ + } + + nvkm_mask(device, 0x002a04, 0xbfffffff, 0xbfffffff); +} + +static void +gf100_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + + nvkm_mask(device, 0x002200, 0x00000001, 0x00000001); + nvkm_wr32(device, 0x002254, 0x10000000 | fifo->userd.bar1->addr >> 12); + + nvkm_wr32(device, 0x002100, 0xffffffff); + nvkm_wr32(device, 0x002140, 0x7fffffff); + nvkm_wr32(device, 0x002628, 0x00000001); /* ENGINE_INTR_EN */ +} + +static int +gf100_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl; + + runl = nvkm_runl_new(fifo, 0, 0, 0); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + nvkm_runl_add(runl, 0, fifo->func->engn, NVKM_ENGINE_GR, 0); + nvkm_runl_add(runl, 1, fifo->func->engn, NVKM_ENGINE_MSPDEC, 0); + nvkm_runl_add(runl, 2, fifo->func->engn, NVKM_ENGINE_MSPPP, 0); + nvkm_runl_add(runl, 3, fifo->func->engn, NVKM_ENGINE_MSVLD, 0); + nvkm_runl_add(runl, 4, fifo->func->engn, NVKM_ENGINE_CE, 0); + nvkm_runl_add(runl, 5, fifo->func->engn, NVKM_ENGINE_CE, 1); + nvkm_runl_add(runl, 15, &gf100_engn_sw, NVKM_ENGINE_SW, 0); + return 0; +} + +int +gf100_fifo_runq_nr(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + u32 save; + + /* Determine number of PBDMAs by checking valid enable bits. */ + save = nvkm_mask(device, 0x000204, 0xffffffff, 0xffffffff); + save = nvkm_mask(device, 0x000204, 0xffffffff, save); + return hweight32(save); +} + +int +gf100_fifo_chid_ctor(struct nvkm_fifo *fifo, int nr) +{ + return nvkm_chid_new(&nvkm_chan_event, &fifo->engine.subdev, nr, 0, nr, &fifo->chid); +} + +static const struct nvkm_fifo_func +gf100_fifo = { + .chid_nr = nv50_fifo_chid_nr, + .chid_ctor = gf100_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gf100_fifo_runl_ctor, + .init = gf100_fifo_init, + .init_pbdmas = gf100_fifo_init_pbdmas, + .intr = gf100_fifo_intr, + .intr_mmu_fault_unit = gf100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gf100_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gf100_runl, + .runq = &gf100_runq, + .engn = &gf100_engn, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, FERMI_CHANNEL_GPFIFO }, &gf100_chan }, +}; + +int +gf100_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gf100_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c new file mode 100644 index 0000000000..d8a4d773a5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c @@ -0,0 +1,833 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" +#include "runq.h" + +#include +#include +#include +#include + +#include +#include + +void +gk104_chan_stop(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x800004 + (chan->id * 8), 0x00000800, 0x00000800); +} + +void +gk104_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x800004 + (chan->id * 8), 0x00000400, 0x00000400); +} + +void +gk104_chan_unbind(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x800000 + (chan->id * 8), 0x00000000); +} + +void +gk104_chan_bind_inst(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x800000 + (chan->id * 8), 0x80000000 | chan->inst->addr >> 12); +} + +void +gk104_chan_bind(struct nvkm_chan *chan) +{ + struct nvkm_runl *runl = chan->cgrp->runl; + struct nvkm_device *device = runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x800004 + (chan->id * 8), 0x000f0000, runl->id << 16); + gk104_chan_bind_inst(chan); +} + +static int +gk104_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + const u64 userd = nvkm_memory_addr(chan->userd.mem) + chan->userd.base; + const u32 limit2 = ilog2(length / 8); + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x08, lower_32_bits(userd)); + nvkm_wo32(chan->inst, 0x0c, upper_32_bits(userd)); + nvkm_wo32(chan->inst, 0x10, 0x0000face); + nvkm_wo32(chan->inst, 0x30, 0xfffff902); + nvkm_wo32(chan->inst, 0x48, lower_32_bits(offset)); + nvkm_wo32(chan->inst, 0x4c, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->inst, 0x84, 0x20400000); + nvkm_wo32(chan->inst, 0x94, 0x30000000 | devm); + nvkm_wo32(chan->inst, 0x9c, 0x00000100); + nvkm_wo32(chan->inst, 0xac, 0x0000001f); + nvkm_wo32(chan->inst, 0xe4, priv ? 0x00000020 : 0x00000000); + nvkm_wo32(chan->inst, 0xe8, chan->id); + nvkm_wo32(chan->inst, 0xb8, 0xf8000000); + nvkm_wo32(chan->inst, 0xf8, 0x10003080); /* 0x002310 */ + nvkm_wo32(chan->inst, 0xfc, 0x10000010); /* 0x002350 */ + nvkm_done(chan->inst); + return 0; +} + +const struct nvkm_chan_func_ramfc +gk104_chan_ramfc = { + .write = gk104_chan_ramfc_write, + .devm = 0xfff, + .priv = true, +}; + +const struct nvkm_chan_func_userd +gk104_chan_userd = { + .bar = 1, + .size = 0x200, + .clear = gf100_chan_userd_clear, +}; + +static const struct nvkm_chan_func +gk104_chan = { + .inst = &gf100_chan_inst, + .userd = &gk104_chan_userd, + .ramfc = &gk104_chan_ramfc, + .bind = gk104_chan_bind, + .unbind = gk104_chan_unbind, + .start = gk104_chan_start, + .stop = gk104_chan_stop, + .preempt = gf100_chan_preempt, +}; + +static void +gk104_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + u32 ptr0, ptr1 = 0; + u64 addr = 0ULL; + + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_SW : return; + case NVKM_ENGINE_GR : ptr0 = 0x0210; break; + case NVKM_ENGINE_SEC : ptr0 = 0x0220; break; + case NVKM_ENGINE_MSPDEC: ptr0 = 0x0250; break; + case NVKM_ENGINE_MSPPP : ptr0 = 0x0260; break; + case NVKM_ENGINE_MSVLD : ptr0 = 0x0270; break; + case NVKM_ENGINE_VIC : ptr0 = 0x0280; break; + case NVKM_ENGINE_MSENC : ptr0 = 0x0290; break; + case NVKM_ENGINE_NVDEC : + ptr1 = 0x0270; + ptr0 = 0x0210; + break; + case NVKM_ENGINE_NVENC : + if (!engn->engine->subdev.inst) + ptr1 = 0x0290; + ptr0 = 0x0210; + break; + default: + WARN_ON(1); + return; + } + + if (cctx) { + addr = cctx->vctx->vma->addr; + addr |= 4ULL; + } + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, ptr0 + 0, lower_32_bits(addr)); + nvkm_wo32(chan->inst, ptr0 + 4, upper_32_bits(addr)); + if (ptr1) { + nvkm_wo32(chan->inst, ptr1 + 0, lower_32_bits(addr)); + nvkm_wo32(chan->inst, ptr1 + 4, upper_32_bits(addr)); + } + nvkm_done(chan->inst); +} + +int +gk104_ectx_ctor(struct nvkm_engn *engn, struct nvkm_vctx *vctx) +{ + struct gf100_vmm_map_v0 args = { .priv = 1 }; + int ret; + + ret = nvkm_vmm_get(vctx->vmm, 12, vctx->inst->size, &vctx->vma); + if (ret) + return ret; + + return nvkm_memory_map(vctx->inst, 0, vctx->vmm, vctx->vma, &args, sizeof(args)); +} + +/*TODO: clean this up */ +struct gk104_engn_status { + bool busy; + bool faulted; + bool chsw; + bool save; + bool load; + struct { + bool tsg; + u32 id; + } prev, next, *chan; +}; + +static void +gk104_engn_status(struct nvkm_engn *engn, struct gk104_engn_status *status) +{ + u32 stat = nvkm_rd32(engn->runl->fifo->engine.subdev.device, 0x002640 + (engn->id * 0x08)); + + status->busy = !!(stat & 0x80000000); + status->faulted = !!(stat & 0x40000000); + status->next.tsg = !!(stat & 0x10000000); + status->next.id = (stat & 0x0fff0000) >> 16; + status->chsw = !!(stat & 0x00008000); + status->save = !!(stat & 0x00004000); + status->load = !!(stat & 0x00002000); + status->prev.tsg = !!(stat & 0x00001000); + status->prev.id = (stat & 0x00000fff); + status->chan = NULL; + + if (status->busy && status->chsw) { + if (status->load && status->save) { + if (nvkm_engine_chsw_load(engn->engine)) + status->chan = &status->next; + else + status->chan = &status->prev; + } else + if (status->load) { + status->chan = &status->next; + } else { + status->chan = &status->prev; + } + } else + if (status->load) { + status->chan = &status->prev; + } + + ENGN_DEBUG(engn, "%08x: busy %d faulted %d chsw %d save %d load %d %sid %d%s-> %sid %d%s", + stat, status->busy, status->faulted, status->chsw, status->save, status->load, + status->prev.tsg ? "tsg" : "ch", status->prev.id, + status->chan == &status->prev ? "*" : " ", + status->next.tsg ? "tsg" : "ch", status->next.id, + status->chan == &status->next ? "*" : " "); +} + +int +gk104_engn_cxid(struct nvkm_engn *engn, bool *cgid) +{ + struct gk104_engn_status status; + + gk104_engn_status(engn, &status); + if (status.chan) { + *cgid = status.chan->tsg; + return status.chan->id; + } + + return -ENODEV; +} + +bool +gk104_engn_chsw(struct nvkm_engn *engn) +{ + struct gk104_engn_status status; + + gk104_engn_status(engn, &status); + if (status.busy && status.chsw) + return true; + + return false; +} + +const struct nvkm_engn_func +gk104_engn = { + .chsw = gk104_engn_chsw, + .cxid = gk104_engn_cxid, + .mmu_fault_trigger = gf100_engn_mmu_fault_trigger, + .mmu_fault_triggered = gf100_engn_mmu_fault_triggered, + .ctor = gk104_ectx_ctor, + .bind = gk104_ectx_bind, +}; + +const struct nvkm_engn_func +gk104_engn_ce = { + .chsw = gk104_engn_chsw, + .cxid = gk104_engn_cxid, + .mmu_fault_trigger = gf100_engn_mmu_fault_trigger, + .mmu_fault_triggered = gf100_engn_mmu_fault_triggered, +}; + +bool +gk104_runq_idle(struct nvkm_runq *runq) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + + return !(nvkm_rd32(device, 0x003080 + (runq->id * 4)) & 0x0000e000); +} + +static const struct nvkm_bitfield +gk104_runq_intr_1_names[] = { + { 0x00000001, "HCE_RE_ILLEGAL_OP" }, + { 0x00000002, "HCE_RE_ALIGNB" }, + { 0x00000004, "HCE_PRIV" }, + { 0x00000008, "HCE_ILLEGAL_MTHD" }, + { 0x00000010, "HCE_ILLEGAL_CLASS" }, + {} +}; + +static bool +gk104_runq_intr_1(struct nvkm_runq *runq) +{ + struct nvkm_subdev *subdev = &runq->fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, 0x04014c + (runq->id * 0x2000)); + u32 stat = nvkm_rd32(device, 0x040148 + (runq->id * 0x2000)) & mask; + u32 chid = nvkm_rd32(device, 0x040120 + (runq->id * 0x2000)) & 0xfff; + char msg[128]; + + if (stat & 0x80000000) { + if (runq->func->intr_1_ctxnotvalid && + runq->func->intr_1_ctxnotvalid(runq, chid)) + stat &= ~0x80000000; + } + + if (stat) { + nvkm_snprintbf(msg, sizeof(msg), gk104_runq_intr_1_names, stat); + nvkm_error(subdev, "PBDMA%d: %08x [%s] ch %d %08x %08x\n", + runq->id, stat, msg, chid, + nvkm_rd32(device, 0x040150 + (runq->id * 0x2000)), + nvkm_rd32(device, 0x040154 + (runq->id * 0x2000))); + } + + nvkm_wr32(device, 0x040148 + (runq->id * 0x2000), stat); + return true; +} + +const struct nvkm_bitfield +gk104_runq_intr_0_names[] = { + { 0x00000001, "MEMREQ" }, + { 0x00000002, "MEMACK_TIMEOUT" }, + { 0x00000004, "MEMACK_EXTRA" }, + { 0x00000008, "MEMDAT_TIMEOUT" }, + { 0x00000010, "MEMDAT_EXTRA" }, + { 0x00000020, "MEMFLUSH" }, + { 0x00000040, "MEMOP" }, + { 0x00000080, "LBCONNECT" }, + { 0x00000100, "LBREQ" }, + { 0x00000200, "LBACK_TIMEOUT" }, + { 0x00000400, "LBACK_EXTRA" }, + { 0x00000800, "LBDAT_TIMEOUT" }, + { 0x00001000, "LBDAT_EXTRA" }, + { 0x00002000, "GPFIFO" }, + { 0x00004000, "GPPTR" }, + { 0x00008000, "GPENTRY" }, + { 0x00010000, "GPCRC" }, + { 0x00020000, "PBPTR" }, + { 0x00040000, "PBENTRY" }, + { 0x00080000, "PBCRC" }, + { 0x00100000, "XBARCONNECT" }, + { 0x00200000, "METHOD" }, + { 0x00400000, "METHODCRC" }, + { 0x00800000, "DEVICE" }, + { 0x02000000, "SEMAPHORE" }, + { 0x04000000, "ACQUIRE" }, + { 0x08000000, "PRI" }, + { 0x20000000, "NO_CTXSW_SEG" }, + { 0x40000000, "PBSEG" }, + { 0x80000000, "SIGNATURE" }, + {} +}; + +bool +gk104_runq_intr(struct nvkm_runq *runq, struct nvkm_runl *null) +{ + bool intr0 = gf100_runq_intr(runq, NULL); + bool intr1 = gk104_runq_intr_1(runq); + + return intr0 || intr1; +} + +void +gk104_runq_init(struct nvkm_runq *runq) +{ + struct nvkm_device *device = runq->fifo->engine.subdev.device; + + gf100_runq_init(runq); + + nvkm_wr32(device, 0x040148 + (runq->id * 0x2000), 0xffffffff); /* HCE.INTR */ + nvkm_wr32(device, 0x04014c + (runq->id * 0x2000), 0xffffffff); /* HCE.INTREN */ +} + +static u32 +gk104_runq_runm(struct nvkm_runq *runq) +{ + return nvkm_rd32(runq->fifo->engine.subdev.device, 0x002390 + (runq->id * 0x04)); +} + +const struct nvkm_runq_func +gk104_runq = { + .init = gk104_runq_init, + .intr = gk104_runq_intr, + .intr_0_names = gk104_runq_intr_0_names, + .idle = gk104_runq_idle, +}; + +void +gk104_runl_fault_clear(struct nvkm_runl *runl) +{ + nvkm_wr32(runl->fifo->engine.subdev.device, 0x00262c, BIT(runl->id)); +} + +void +gk104_runl_allow(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, 0x002630, BIT(runl->id), 0x00000000); +} + +void +gk104_runl_block(struct nvkm_runl *runl, u32 engm) +{ + nvkm_mask(runl->fifo->engine.subdev.device, 0x002630, BIT(runl->id), BIT(runl->id)); +} + +bool +gk104_runl_pending(struct nvkm_runl *runl) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + + return nvkm_rd32(device, 0x002284 + (runl->id * 0x08)) & 0x00100000; +} + +void +gk104_runl_commit(struct nvkm_runl *runl, struct nvkm_memory *memory, u32 start, int count) +{ + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + u64 addr = nvkm_memory_addr(memory) + start; + int target; + + switch (nvkm_memory_target(memory)) { + case NVKM_MEM_TARGET_VRAM: target = 0; break; + case NVKM_MEM_TARGET_NCOH: target = 3; break; + default: + WARN_ON(1); + return; + } + + spin_lock_irq(&fifo->lock); + nvkm_wr32(device, 0x002270, (target << 28) | (addr >> 12)); + nvkm_wr32(device, 0x002274, (runl->id << 20) | count); + spin_unlock_irq(&fifo->lock); +} + +void +gk104_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0, chan->id); + nvkm_wo32(memory, offset + 4, 0x00000000); +} + +static const struct nvkm_runl_func +gk104_runl = { + .size = 8, + .update = nv50_runl_update, + .insert_chan = gk104_runl_insert_chan, + .commit = gk104_runl_commit, + .wait = nv50_runl_wait, + .pending = gk104_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .fault_clear = gk104_runl_fault_clear, + .preempt_pending = gf100_runl_preempt_pending, +}; + +static const struct nvkm_enum +gk104_fifo_mmu_fault_engine[] = { + { 0x00, "GR", NULL, NVKM_ENGINE_GR }, + { 0x01, "DISPLAY" }, + { 0x02, "CAPTURE" }, + { 0x03, "IFB", NULL, NVKM_ENGINE_IFB }, + { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, + { 0x06, "SCHED" }, + { 0x07, "HOST0" }, + { 0x08, "HOST1" }, + { 0x09, "HOST2" }, + { 0x0a, "HOST3" }, + { 0x0b, "HOST4" }, + { 0x0c, "HOST5" }, + { 0x0d, "HOST6" }, + { 0x0e, "HOST7" }, + { 0x0f, "HOSTSR" }, + { 0x10, "MSVLD", NULL, NVKM_ENGINE_MSVLD }, + { 0x11, "MSPPP", NULL, NVKM_ENGINE_MSPPP }, + { 0x13, "PERF" }, + { 0x14, "MSPDEC", NULL, NVKM_ENGINE_MSPDEC }, + { 0x15, "CE0", NULL, NVKM_ENGINE_CE, 0 }, + { 0x16, "CE1", NULL, NVKM_ENGINE_CE, 1 }, + { 0x17, "PMU" }, + { 0x18, "PTP" }, + { 0x19, "MSENC", NULL, NVKM_ENGINE_MSENC }, + { 0x1b, "CE2", NULL, NVKM_ENGINE_CE, 2 }, + {} +}; + +const struct nvkm_enum +gk104_fifo_mmu_fault_reason[] = { + { 0x00, "PDE" }, + { 0x01, "PDE_SIZE" }, + { 0x02, "PTE" }, + { 0x03, "VA_LIMIT_VIOLATION" }, + { 0x04, "UNBOUND_INST_BLOCK" }, + { 0x05, "PRIV_VIOLATION" }, + { 0x06, "RO_VIOLATION" }, + { 0x07, "WO_VIOLATION" }, + { 0x08, "PITCH_MASK_VIOLATION" }, + { 0x09, "WORK_CREATION" }, + { 0x0a, "UNSUPPORTED_APERTURE" }, + { 0x0b, "COMPRESSION_FAILURE" }, + { 0x0c, "UNSUPPORTED_KIND" }, + { 0x0d, "REGION_VIOLATION" }, + { 0x0e, "BOTH_PTES_VALID" }, + { 0x0f, "INFO_TYPE_POISONED" }, + {} +}; + +const struct nvkm_enum +gk104_fifo_mmu_fault_hubclient[] = { + { 0x00, "VIP" }, + { 0x01, "CE0" }, + { 0x02, "CE1" }, + { 0x03, "DNISO" }, + { 0x04, "FE" }, + { 0x05, "FECS" }, + { 0x06, "HOST" }, + { 0x07, "HOST_CPU" }, + { 0x08, "HOST_CPU_NB" }, + { 0x09, "ISO" }, + { 0x0a, "MMU" }, + { 0x0b, "MSPDEC" }, + { 0x0c, "MSPPP" }, + { 0x0d, "MSVLD" }, + { 0x0e, "NISO" }, + { 0x0f, "P2P" }, + { 0x10, "PD" }, + { 0x11, "PERF" }, + { 0x12, "PMU" }, + { 0x13, "RASTERTWOD" }, + { 0x14, "SCC" }, + { 0x15, "SCC_NB" }, + { 0x16, "SEC" }, + { 0x17, "SSYNC" }, + { 0x18, "GR_CE" }, + { 0x19, "CE2" }, + { 0x1a, "XV" }, + { 0x1b, "MMU_NB" }, + { 0x1c, "MSENC" }, + { 0x1d, "DFALCON" }, + { 0x1e, "SKED" }, + { 0x1f, "AFALCON" }, + {} +}; + +const struct nvkm_enum +gk104_fifo_mmu_fault_gpcclient[] = { + { 0x00, "L1_0" }, { 0x01, "T1_0" }, { 0x02, "PE_0" }, + { 0x03, "L1_1" }, { 0x04, "T1_1" }, { 0x05, "PE_1" }, + { 0x06, "L1_2" }, { 0x07, "T1_2" }, { 0x08, "PE_2" }, + { 0x09, "L1_3" }, { 0x0a, "T1_3" }, { 0x0b, "PE_3" }, + { 0x0c, "RAST" }, + { 0x0d, "GCC" }, + { 0x0e, "GPCCS" }, + { 0x0f, "PROP_0" }, + { 0x10, "PROP_1" }, + { 0x11, "PROP_2" }, + { 0x12, "PROP_3" }, + { 0x13, "L1_4" }, { 0x14, "T1_4" }, { 0x15, "PE_4" }, + { 0x16, "L1_5" }, { 0x17, "T1_5" }, { 0x18, "PE_5" }, + { 0x19, "L1_6" }, { 0x1a, "T1_6" }, { 0x1b, "PE_6" }, + { 0x1c, "L1_7" }, { 0x1d, "T1_7" }, { 0x1e, "PE_7" }, + { 0x1f, "GPM" }, + { 0x20, "LTP_UTLB_0" }, + { 0x21, "LTP_UTLB_1" }, + { 0x22, "LTP_UTLB_2" }, + { 0x23, "LTP_UTLB_3" }, + { 0x24, "GPC_RGG_UTLB" }, + {} +}; + +const struct nvkm_fifo_func_mmu_fault +gk104_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gf100_fifo_mmu_fault_access, + .engine = gk104_fifo_mmu_fault_engine, + .reason = gk104_fifo_mmu_fault_reason, + .hubclient = gk104_fifo_mmu_fault_hubclient, + .gpcclient = gk104_fifo_mmu_fault_gpcclient, +}; + +static const struct nvkm_enum +gk104_fifo_intr_bind_reason[] = { + { 0x01, "BIND_NOT_UNBOUND" }, + { 0x02, "SNOOP_WITHOUT_BAR1" }, + { 0x03, "UNBIND_WHILE_RUNNING" }, + { 0x05, "INVALID_RUNLIST" }, + { 0x06, "INVALID_CTX_TGT" }, + { 0x0b, "UNBIND_WHILE_PARKED" }, + {} +}; + +void +gk104_fifo_intr_bind(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + u32 intr = nvkm_rd32(subdev->device, 0x00252c); + u32 code = intr & 0x000000ff; + const struct nvkm_enum *en = nvkm_enum_find(gk104_fifo_intr_bind_reason, code); + + nvkm_error(subdev, "BIND_ERROR %02x [%s]\n", code, en ? en->name : ""); +} + +void +gk104_fifo_intr_chsw(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x00256c); + + nvkm_error(subdev, "CHSW_ERROR %08x\n", stat); + nvkm_wr32(device, 0x00256c, stat); +} + +static void +gk104_fifo_intr_dropped_fault(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + u32 stat = nvkm_rd32(subdev->device, 0x00259c); + + nvkm_error(subdev, "DROPPED_MMU_FAULT %08x\n", stat); +} + +void +gk104_fifo_intr_runlist(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_runl *runl; + u32 mask = nvkm_rd32(device, 0x002a00); + + nvkm_runl_foreach_cond(runl, fifo, mask & BIT(runl->id)) { + nvkm_wr32(device, 0x002a00, BIT(runl->id)); + } +} + +irqreturn_t +gk104_fifo_intr(struct nvkm_inth *inth) +{ + struct nvkm_fifo *fifo = container_of(inth, typeof(*fifo), engine.subdev.inth); + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, 0x002140); + u32 stat = nvkm_rd32(device, 0x002100) & mask; + + if (stat & 0x00000001) { + gk104_fifo_intr_bind(fifo); + nvkm_wr32(device, 0x002100, 0x00000001); + stat &= ~0x00000001; + } + + if (stat & 0x00000010) { + nvkm_error(subdev, "PIO_ERROR\n"); + nvkm_wr32(device, 0x002100, 0x00000010); + stat &= ~0x00000010; + } + + if (stat & 0x00000100) { + gf100_fifo_intr_sched(fifo); + nvkm_wr32(device, 0x002100, 0x00000100); + stat &= ~0x00000100; + } + + if (stat & 0x00010000) { + gk104_fifo_intr_chsw(fifo); + nvkm_wr32(device, 0x002100, 0x00010000); + stat &= ~0x00010000; + } + + if (stat & 0x00800000) { + nvkm_error(subdev, "FB_FLUSH_TIMEOUT\n"); + nvkm_wr32(device, 0x002100, 0x00800000); + stat &= ~0x00800000; + } + + if (stat & 0x01000000) { + nvkm_error(subdev, "LB_ERROR\n"); + nvkm_wr32(device, 0x002100, 0x01000000); + stat &= ~0x01000000; + } + + if (stat & 0x08000000) { + gk104_fifo_intr_dropped_fault(fifo); + nvkm_wr32(device, 0x002100, 0x08000000); + stat &= ~0x08000000; + } + + if (stat & 0x10000000) { + gf100_fifo_intr_mmu_fault(fifo); + stat &= ~0x10000000; + } + + if (stat & 0x20000000) { + if (gf100_fifo_intr_pbdma(fifo)) + stat &= ~0x20000000; + } + + if (stat & 0x40000000) { + gk104_fifo_intr_runlist(fifo); + stat &= ~0x40000000; + } + + if (stat & 0x80000000) { + nvkm_wr32(device, 0x002100, 0x80000000); + nvkm_event_ntfy(&fifo->nonstall.event, 0, NVKM_FIFO_NONSTALL_EVENT); + stat &= ~0x80000000; + } + + if (stat) { + nvkm_error(subdev, "INTR %08x\n", stat); + spin_lock(&fifo->lock); + nvkm_mask(device, 0x002140, stat, 0x00000000); + spin_unlock(&fifo->lock); + nvkm_wr32(device, 0x002100, stat); + } + + return IRQ_HANDLED; +} + +void +gk104_fifo_init_pbdmas(struct nvkm_fifo *fifo, u32 mask) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + + nvkm_wr32(device, 0x000204, mask); + nvkm_mask(device, 0x002a04, 0xbfffffff, 0xbfffffff); +} + +void +gk104_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + + if (fifo->func->chan.func->userd->bar == 1) + nvkm_wr32(device, 0x002254, 0x10000000 | fifo->userd.bar1->addr >> 12); + + nvkm_wr32(device, 0x002100, 0xffffffff); + nvkm_wr32(device, 0x002140, 0x7fffffff); +} + +int +gk104_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_top_device *tdev; + struct nvkm_runl *runl; + struct nvkm_runq *runq; + const struct nvkm_engn_func *func; + + nvkm_list_foreach(tdev, &device->top->device, head, tdev->runlist >= 0) { + runl = nvkm_runl_get(fifo, tdev->runlist, tdev->runlist); + if (!runl) { + runl = nvkm_runl_new(fifo, tdev->runlist, tdev->runlist, 0); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + nvkm_runq_foreach_cond(runq, fifo, gk104_runq_runm(runq) & BIT(runl->id)) { + if (WARN_ON(runl->runq_nr == ARRAY_SIZE(runl->runq))) + return -ENOMEM; + + runl->runq[runl->runq_nr++] = runq; + } + + } + + if (tdev->engine < 0) + continue; + + switch (tdev->type) { + case NVKM_ENGINE_CE: + func = fifo->func->engn_ce; + break; + case NVKM_ENGINE_GR: + nvkm_runl_add(runl, 15, &gf100_engn_sw, NVKM_ENGINE_SW, 0); + fallthrough; + default: + func = fifo->func->engn; + break; + } + + nvkm_runl_add(runl, tdev->engine, func, tdev->type, tdev->inst); + } + + return 0; +} + +int +gk104_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 4096; +} + +static const struct nvkm_fifo_func +gk104_fifo = { + .chid_nr = gk104_fifo_chid_nr, + .chid_ctor = gf100_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gf100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gk104_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gk104_runl, + .runq = &gk104_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, KEPLER_CHANNEL_GPFIFO_A }, &gk104_chan }, +}; + +int +gk104_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gk104_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk110.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk110.c new file mode 100644 index 0000000000..a8ff21cf77 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk110.c @@ -0,0 +1,132 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include +#include + +#include + +void +gk110_chan_preempt(struct nvkm_chan *chan) +{ + struct nvkm_cgrp *cgrp = chan->cgrp; + + if (cgrp->hw) { + cgrp->func->preempt(cgrp); + return; + } + + gf100_chan_preempt(chan); +} + +const struct nvkm_chan_func +gk110_chan = { + .inst = &gf100_chan_inst, + .userd = &gk104_chan_userd, + .ramfc = &gk104_chan_ramfc, + .bind = gk104_chan_bind, + .unbind = gk104_chan_unbind, + .start = gk104_chan_start, + .stop = gk104_chan_stop, + .preempt = gk110_chan_preempt, +}; + +static void +gk110_cgrp_preempt(struct nvkm_cgrp *cgrp) +{ + nvkm_wr32(cgrp->runl->fifo->engine.subdev.device, 0x002634, 0x01000000 | cgrp->id); +} + +const struct nvkm_cgrp_func +gk110_cgrp = { + .preempt = gk110_cgrp_preempt, +}; + +void +gk110_runl_insert_cgrp(struct nvkm_cgrp *cgrp, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0, (cgrp->chan_nr << 26) | (128 << 18) | + (3 << 14) | 0x00002000 | cgrp->id); + nvkm_wo32(memory, offset + 4, 0x00000000); +} + +const struct nvkm_runl_func +gk110_runl = { + .size = 8, + .update = nv50_runl_update, + .insert_cgrp = gk110_runl_insert_cgrp, + .insert_chan = gk104_runl_insert_chan, + .commit = gk104_runl_commit, + .wait = nv50_runl_wait, + .pending = gk104_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .fault_clear = gk104_runl_fault_clear, + .preempt_pending = gf100_runl_preempt_pending, +}; + +int +gk110_fifo_chid_ctor(struct nvkm_fifo *fifo, int nr) +{ + int ret; + + ret = nvkm_chid_new(&nvkm_chan_event, &fifo->engine.subdev, nr, 0, nr, &fifo->cgid); + if (ret) + return ret; + + return gf100_fifo_chid_ctor(fifo, nr); +} + +static const struct nvkm_fifo_func +gk110_fifo = { + .chid_nr = gk104_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gf100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gk104_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gk110_runl, + .runq = &gk104_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp }, + .chan = {{ 0, 0, KEPLER_CHANNEL_GPFIFO_B }, &gk110_chan }, +}; + +int +gk110_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gk110_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk208.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk208.c new file mode 100644 index 0000000000..8fa2b0be14 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk208.c @@ -0,0 +1,77 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "runq.h" + +#include + +void +gk208_runq_init(struct nvkm_runq *runq) +{ + gk104_runq_init(runq); + + nvkm_wr32(runq->fifo->engine.subdev.device, 0x04012c + (runq->id * 0x2000), 0x000f4240); +} + +const struct nvkm_runq_func +gk208_runq = { + .init = gk208_runq_init, + .intr = gk104_runq_intr, + .intr_0_names = gk104_runq_intr_0_names, + .idle = gk104_runq_idle, +}; + +static int +gk208_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 1024; +} + +static const struct nvkm_fifo_func +gk208_fifo = { + .chid_nr = gk208_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gf100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gk104_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gk110_runl, + .runq = &gk208_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp }, + .chan = {{ 0, 0, KEPLER_CHANNEL_GPFIFO_A }, &gk110_chan }, +}; + +int +gk208_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gk208_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk20a.c new file mode 100644 index 0000000000..b63ca83613 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk20a.c @@ -0,0 +1,52 @@ +/* + * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +static const struct nvkm_fifo_func +gk20a_fifo = { + .chid_nr = nv50_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gf100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gk104_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gk110_runl, + .runq = &gk208_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ }, &gk110_cgrp }, + .chan = {{ 0, 0, KEPLER_CHANNEL_GPFIFO_A }, &gk110_chan }, +}; + +int +gk20a_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gk20a_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c new file mode 100644 index 0000000000..5ba60021b5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm107.c @@ -0,0 +1,156 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "runl.h" + +#include +#include + +#include + +const struct nvkm_chan_func +gm107_chan = { + .inst = &gf100_chan_inst, + .userd = &gk104_chan_userd, + .ramfc = &gk104_chan_ramfc, + .bind = gk104_chan_bind_inst, + .unbind = gk104_chan_unbind, + .start = gk104_chan_start, + .stop = gk104_chan_stop, + .preempt = gk110_chan_preempt, +}; + +static void +gm107_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0, chan->id); + nvkm_wo32(memory, offset + 4, chan->inst->addr >> 12); +} + +const struct nvkm_runl_func +gm107_runl = { + .size = 8, + .update = nv50_runl_update, + .insert_cgrp = gk110_runl_insert_cgrp, + .insert_chan = gm107_runl_insert_chan, + .commit = gk104_runl_commit, + .wait = nv50_runl_wait, + .pending = gk104_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .fault_clear = gk104_runl_fault_clear, + .preempt_pending = gf100_runl_preempt_pending, +}; + +static const struct nvkm_enum +gm107_fifo_mmu_fault_engine[] = { + { 0x01, "DISPLAY" }, + { 0x02, "CAPTURE" }, + { 0x03, "IFB", NULL, NVKM_ENGINE_IFB }, + { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, + { 0x06, "SCHED" }, + { 0x07, "HOST0" }, + { 0x08, "HOST1" }, + { 0x09, "HOST2" }, + { 0x0a, "HOST3" }, + { 0x0b, "HOST4" }, + { 0x0c, "HOST5" }, + { 0x0d, "HOST6" }, + { 0x0e, "HOST7" }, + { 0x0f, "HOSTSR" }, + { 0x13, "PERF" }, + { 0x17, "PMU" }, + { 0x18, "PTP" }, + {} +}; + +const struct nvkm_fifo_func_mmu_fault +gm107_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gf100_fifo_mmu_fault_access, + .engine = gm107_fifo_mmu_fault_engine, + .reason = gk104_fifo_mmu_fault_reason, + .hubclient = gk104_fifo_mmu_fault_hubclient, + .gpcclient = gk104_fifo_mmu_fault_gpcclient, +}; + +void +gm107_fifo_intr_mmu_fault_unit(struct nvkm_fifo *fifo, int unit) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + u32 inst = nvkm_rd32(device, 0x002800 + (unit * 0x10)); + u32 valo = nvkm_rd32(device, 0x002804 + (unit * 0x10)); + u32 vahi = nvkm_rd32(device, 0x002808 + (unit * 0x10)); + u32 type = nvkm_rd32(device, 0x00280c + (unit * 0x10)); + struct nvkm_fault_data info; + + info.inst = (u64)inst << 12; + info.addr = ((u64)vahi << 32) | valo; + info.time = 0; + info.engine = unit; + info.valid = 1; + info.gpc = (type & 0x1f000000) >> 24; + info.client = (type & 0x00003f00) >> 8; + info.access = (type & 0x00000080) >> 7; + info.hub = (type & 0x00000040) >> 6; + info.reason = (type & 0x0000000f); + + nvkm_fifo_fault(fifo, &info); +} + +static int +gm107_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 2048; +} + +static const struct nvkm_fifo_func +gm107_fifo = { + .chid_nr = gm107_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gf100_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gm107_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gm107_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gm107_runl, + .runq = &gk208_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp }, + .chan = {{ 0, 0, KEPLER_CHANNEL_GPFIFO_B }, &gm107_chan }, +}; + +int +gm107_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gm107_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm200.c new file mode 100644 index 0000000000..d92d1ac391 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gm200.c @@ -0,0 +1,66 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +int +gm200_fifo_runq_nr(struct nvkm_fifo *fifo) +{ + return nvkm_rd32(fifo->engine.subdev.device, 0x002004) & 0x000000ff; +} + +int +gm200_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return nvkm_rd32(fifo->engine.subdev.device, 0x002008); +} + +static const struct nvkm_fifo_func +gm200_fifo = { + .chid_nr = gm200_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gm200_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gm107_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gm107_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gm107_runl, + .runq = &gk208_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp }, + .chan = {{ 0, 0, MAXWELL_CHANNEL_GPFIFO_A }, &gm107_chan }, +}; + +int +gm200_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gm200_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c new file mode 100644 index 0000000000..65bdb6a7d5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gp100.c @@ -0,0 +1,139 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "runl.h" + +#include +#include + +#include + +static void +gp100_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0, chan->id | chan->runq << 14); + nvkm_wo32(memory, offset + 4, chan->inst->addr >> 12); +} + +static const struct nvkm_runl_func +gp100_runl = { + .runqs = 2, + .size = 8, + .update = nv50_runl_update, + .insert_cgrp = gk110_runl_insert_cgrp, + .insert_chan = gp100_runl_insert_chan, + .commit = gk104_runl_commit, + .wait = nv50_runl_wait, + .pending = gk104_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .fault_clear = gk104_runl_fault_clear, + .preempt_pending = gf100_runl_preempt_pending, +}; + +static const struct nvkm_enum +gp100_fifo_mmu_fault_engine[] = { + { 0x01, "DISPLAY" }, + { 0x03, "IFB", NULL, NVKM_ENGINE_IFB }, + { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, + { 0x06, "HOST0" }, + { 0x07, "HOST1" }, + { 0x08, "HOST2" }, + { 0x09, "HOST3" }, + { 0x0a, "HOST4" }, + { 0x0b, "HOST5" }, + { 0x0c, "HOST6" }, + { 0x0d, "HOST7" }, + { 0x0e, "HOST8" }, + { 0x0f, "HOST9" }, + { 0x10, "HOST10" }, + { 0x13, "PERF" }, + { 0x17, "PMU" }, + { 0x18, "PTP" }, + { 0x1f, "PHYSICAL" }, + {} +}; + +static const struct nvkm_fifo_func_mmu_fault +gp100_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gf100_fifo_mmu_fault_access, + .engine = gp100_fifo_mmu_fault_engine, + .reason = gk104_fifo_mmu_fault_reason, + .hubclient = gk104_fifo_mmu_fault_hubclient, + .gpcclient = gk104_fifo_mmu_fault_gpcclient, +}; + +static void +gp100_fifo_intr_mmu_fault_unit(struct nvkm_fifo *fifo, int unit) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + u32 inst = nvkm_rd32(device, 0x002800 + (unit * 0x10)); + u32 valo = nvkm_rd32(device, 0x002804 + (unit * 0x10)); + u32 vahi = nvkm_rd32(device, 0x002808 + (unit * 0x10)); + u32 type = nvkm_rd32(device, 0x00280c + (unit * 0x10)); + struct nvkm_fault_data info; + + info.inst = (u64)inst << 12; + info.addr = ((u64)vahi << 32) | valo; + info.time = 0; + info.engine = unit; + info.valid = 1; + info.gpc = (type & 0x1f000000) >> 24; + info.hub = (type & 0x00100000) >> 20; + info.access = (type & 0x00070000) >> 16; + info.client = (type & 0x00007f00) >> 8; + info.reason = (type & 0x0000001f); + + nvkm_fifo_fault(fifo, &info); +} + +static const struct nvkm_fifo_func +gp100_fifo = { + .chid_nr = gm200_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gm200_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_mmu_fault_unit = gp100_fifo_intr_mmu_fault_unit, + .intr_ctxsw_timeout = gf100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gp100_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gp100_runl, + .runq = &gk208_runq, + .engn = &gk104_engn, + .engn_ce = &gk104_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp, .force = true }, + .chan = {{ 0, 0, PASCAL_CHANNEL_GPFIFO_A }, &gm107_chan }, +}; + +int +gp100_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gp100_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gv100.c new file mode 100644 index 0000000000..33066c8cdc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gv100.c @@ -0,0 +1,492 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "chan.h" +#include "chid.h" +#include "cgrp.h" +#include "runl.h" +#include "runq.h" + +#include +#include + +#include + +static u32 +gv100_chan_doorbell_handle(struct nvkm_chan *chan) +{ + return chan->id; +} + +static int +gv100_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + const u64 userd = nvkm_memory_addr(chan->userd.mem) + chan->userd.base; + const u32 limit2 = ilog2(length / 8); + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x008, lower_32_bits(userd)); + nvkm_wo32(chan->inst, 0x00c, upper_32_bits(userd)); + nvkm_wo32(chan->inst, 0x010, 0x0000face); + nvkm_wo32(chan->inst, 0x030, 0x7ffff902); + nvkm_wo32(chan->inst, 0x048, lower_32_bits(offset)); + nvkm_wo32(chan->inst, 0x04c, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->inst, 0x084, 0x20400000); + nvkm_wo32(chan->inst, 0x094, 0x30000000 | devm); + nvkm_wo32(chan->inst, 0x0e4, priv ? 0x00000020 : 0x00000000); + nvkm_wo32(chan->inst, 0x0e8, chan->id); + nvkm_wo32(chan->inst, 0x0f4, 0x00001000 | (priv ? 0x00000100 : 0x00000000)); + nvkm_wo32(chan->inst, 0x0f8, 0x10003080); + nvkm_mo32(chan->inst, 0x218, 0x00000000, 0x00000000); + nvkm_done(chan->inst); + return 0; +} + +const struct nvkm_chan_func_ramfc +gv100_chan_ramfc = { + .write = gv100_chan_ramfc_write, + .devm = 0xfff, + .priv = true, +}; + +const struct nvkm_chan_func_userd +gv100_chan_userd = { + .bar = -1, + .size = 0x200, + .clear = gf100_chan_userd_clear, +}; + +static const struct nvkm_chan_func +gv100_chan = { + .inst = &gf100_chan_inst, + .userd = &gv100_chan_userd, + .ramfc = &gv100_chan_ramfc, + .bind = gk104_chan_bind_inst, + .unbind = gk104_chan_unbind, + .start = gk104_chan_start, + .stop = gk104_chan_stop, + .preempt = gk110_chan_preempt, + .doorbell_handle = gv100_chan_doorbell_handle, +}; + +void +gv100_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + u64 addr = 0ULL; + + if (cctx) { + addr = cctx->vctx->vma->addr; + addr |= 4ULL; + } + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x210, lower_32_bits(addr)); + nvkm_wo32(chan->inst, 0x214, upper_32_bits(addr)); + nvkm_mo32(chan->inst, 0x0ac, 0x00010000, cctx ? 0x00010000 : 0x00000000); + nvkm_done(chan->inst); +} + +const struct nvkm_engn_func +gv100_engn = { + .chsw = gk104_engn_chsw, + .cxid = gk104_engn_cxid, + .ctor = gk104_ectx_ctor, + .bind = gv100_ectx_bind, +}; + +void +gv100_ectx_ce_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + const u64 bar2 = cctx ? nvkm_memory_bar2(cctx->vctx->inst->memory) : 0ULL; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x220, lower_32_bits(bar2)); + nvkm_wo32(chan->inst, 0x224, upper_32_bits(bar2)); + nvkm_mo32(chan->inst, 0x0ac, 0x00020000, cctx ? 0x00020000 : 0x00000000); + nvkm_done(chan->inst); +} + +int +gv100_ectx_ce_ctor(struct nvkm_engn *engn, struct nvkm_vctx *vctx) +{ + if (nvkm_memory_bar2(vctx->inst->memory) == ~0ULL) + return -EFAULT; + + return 0; +} + +const struct nvkm_engn_func +gv100_engn_ce = { + .chsw = gk104_engn_chsw, + .cxid = gk104_engn_cxid, + .ctor = gv100_ectx_ce_ctor, + .bind = gv100_ectx_ce_bind, +}; + +static bool +gv100_runq_intr_1_ctxnotvalid(struct nvkm_runq *runq, int chid) +{ + struct nvkm_fifo *fifo = runq->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_chan *chan; + unsigned long flags; + + RUNQ_ERROR(runq, "CTXNOTVALID chid:%d", chid); + + chan = nvkm_chan_get_chid(&fifo->engine, chid, &flags); + if (WARN_ON_ONCE(!chan)) + return false; + + nvkm_chan_error(chan, true); + nvkm_chan_put(&chan, flags); + + nvkm_mask(device, 0x0400ac + (runq->id * 0x2000), 0x00030000, 0x00030000); + nvkm_wr32(device, 0x040148 + (runq->id * 0x2000), 0x80000000); + return true; +} + +const struct nvkm_runq_func +gv100_runq = { + .init = gk208_runq_init, + .intr = gk104_runq_intr, + .intr_0_names = gk104_runq_intr_0_names, + .intr_1_ctxnotvalid = gv100_runq_intr_1_ctxnotvalid, + .idle = gk104_runq_idle, +}; + +void +gv100_runl_preempt(struct nvkm_runl *runl) +{ + nvkm_wr32(runl->fifo->engine.subdev.device, 0x002638, BIT(runl->id)); +} + +void +gv100_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + const u64 user = nvkm_memory_addr(chan->userd.mem) + chan->userd.base; + const u64 inst = chan->inst->addr; + + nvkm_wo32(memory, offset + 0x0, lower_32_bits(user) | chan->runq << 1); + nvkm_wo32(memory, offset + 0x4, upper_32_bits(user)); + nvkm_wo32(memory, offset + 0x8, lower_32_bits(inst) | chan->id); + nvkm_wo32(memory, offset + 0xc, upper_32_bits(inst)); +} + +void +gv100_runl_insert_cgrp(struct nvkm_cgrp *cgrp, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset + 0x0, (128 << 24) | (3 << 16) | 0x00000001); + nvkm_wo32(memory, offset + 0x4, cgrp->chan_nr); + nvkm_wo32(memory, offset + 0x8, cgrp->id); + nvkm_wo32(memory, offset + 0xc, 0x00000000); +} + +static const struct nvkm_runl_func +gv100_runl = { + .runqs = 2, + .size = 16, + .update = nv50_runl_update, + .insert_cgrp = gv100_runl_insert_cgrp, + .insert_chan = gv100_runl_insert_chan, + .commit = gk104_runl_commit, + .wait = nv50_runl_wait, + .pending = gk104_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .preempt = gv100_runl_preempt, + .preempt_pending = gf100_runl_preempt_pending, +}; + +const struct nvkm_enum +gv100_fifo_mmu_fault_gpcclient[] = { + { 0x00, "T1_0" }, + { 0x01, "T1_1" }, + { 0x02, "T1_2" }, + { 0x03, "T1_3" }, + { 0x04, "T1_4" }, + { 0x05, "T1_5" }, + { 0x06, "T1_6" }, + { 0x07, "T1_7" }, + { 0x08, "PE_0" }, + { 0x09, "PE_1" }, + { 0x0a, "PE_2" }, + { 0x0b, "PE_3" }, + { 0x0c, "PE_4" }, + { 0x0d, "PE_5" }, + { 0x0e, "PE_6" }, + { 0x0f, "PE_7" }, + { 0x10, "RAST" }, + { 0x11, "GCC" }, + { 0x12, "GPCCS" }, + { 0x13, "PROP_0" }, + { 0x14, "PROP_1" }, + { 0x15, "PROP_2" }, + { 0x16, "PROP_3" }, + { 0x17, "GPM" }, + { 0x18, "LTP_UTLB_0" }, + { 0x19, "LTP_UTLB_1" }, + { 0x1a, "LTP_UTLB_2" }, + { 0x1b, "LTP_UTLB_3" }, + { 0x1c, "LTP_UTLB_4" }, + { 0x1d, "LTP_UTLB_5" }, + { 0x1e, "LTP_UTLB_6" }, + { 0x1f, "LTP_UTLB_7" }, + { 0x20, "RGG_UTLB" }, + { 0x21, "T1_8" }, + { 0x22, "T1_9" }, + { 0x23, "T1_10" }, + { 0x24, "T1_11" }, + { 0x25, "T1_12" }, + { 0x26, "T1_13" }, + { 0x27, "T1_14" }, + { 0x28, "T1_15" }, + { 0x29, "TPCCS_0" }, + { 0x2a, "TPCCS_1" }, + { 0x2b, "TPCCS_2" }, + { 0x2c, "TPCCS_3" }, + { 0x2d, "TPCCS_4" }, + { 0x2e, "TPCCS_5" }, + { 0x2f, "TPCCS_6" }, + { 0x30, "TPCCS_7" }, + { 0x31, "PE_8" }, + { 0x32, "PE_9" }, + { 0x33, "TPCCS_8" }, + { 0x34, "TPCCS_9" }, + { 0x35, "T1_16" }, + { 0x36, "T1_17" }, + { 0x37, "T1_18" }, + { 0x38, "T1_19" }, + { 0x39, "PE_10" }, + { 0x3a, "PE_11" }, + { 0x3b, "TPCCS_10" }, + { 0x3c, "TPCCS_11" }, + { 0x3d, "T1_20" }, + { 0x3e, "T1_21" }, + { 0x3f, "T1_22" }, + { 0x40, "T1_23" }, + { 0x41, "PE_12" }, + { 0x42, "PE_13" }, + { 0x43, "TPCCS_12" }, + { 0x44, "TPCCS_13" }, + { 0x45, "T1_24" }, + { 0x46, "T1_25" }, + { 0x47, "T1_26" }, + { 0x48, "T1_27" }, + { 0x49, "PE_14" }, + { 0x4a, "PE_15" }, + { 0x4b, "TPCCS_14" }, + { 0x4c, "TPCCS_15" }, + { 0x4d, "T1_28" }, + { 0x4e, "T1_29" }, + { 0x4f, "T1_30" }, + { 0x50, "T1_31" }, + { 0x51, "PE_16" }, + { 0x52, "PE_17" }, + { 0x53, "TPCCS_16" }, + { 0x54, "TPCCS_17" }, + { 0x55, "T1_32" }, + { 0x56, "T1_33" }, + { 0x57, "T1_34" }, + { 0x58, "T1_35" }, + { 0x59, "PE_18" }, + { 0x5a, "PE_19" }, + { 0x5b, "TPCCS_18" }, + { 0x5c, "TPCCS_19" }, + { 0x5d, "T1_36" }, + { 0x5e, "T1_37" }, + { 0x5f, "T1_38" }, + { 0x60, "T1_39" }, + {} +}; + +const struct nvkm_enum +gv100_fifo_mmu_fault_hubclient[] = { + { 0x00, "VIP" }, + { 0x01, "CE0" }, + { 0x02, "CE1" }, + { 0x03, "DNISO" }, + { 0x04, "FE" }, + { 0x05, "FECS" }, + { 0x06, "HOST" }, + { 0x07, "HOST_CPU" }, + { 0x08, "HOST_CPU_NB" }, + { 0x09, "ISO" }, + { 0x0a, "MMU" }, + { 0x0b, "NVDEC" }, + { 0x0d, "NVENC1" }, + { 0x0e, "NISO" }, + { 0x0f, "P2P" }, + { 0x10, "PD" }, + { 0x11, "PERF" }, + { 0x12, "PMU" }, + { 0x13, "RASTERTWOD" }, + { 0x14, "SCC" }, + { 0x15, "SCC_NB" }, + { 0x16, "SEC" }, + { 0x17, "SSYNC" }, + { 0x18, "CE2" }, + { 0x19, "XV" }, + { 0x1a, "MMU_NB" }, + { 0x1b, "NVENC0" }, + { 0x1c, "DFALCON" }, + { 0x1d, "SKED" }, + { 0x1e, "AFALCON" }, + { 0x1f, "DONT_CARE" }, + { 0x20, "HSCE0" }, + { 0x21, "HSCE1" }, + { 0x22, "HSCE2" }, + { 0x23, "HSCE3" }, + { 0x24, "HSCE4" }, + { 0x25, "HSCE5" }, + { 0x26, "HSCE6" }, + { 0x27, "HSCE7" }, + { 0x28, "HSCE8" }, + { 0x29, "HSCE9" }, + { 0x2a, "HSHUB" }, + { 0x2b, "PTP_X0" }, + { 0x2c, "PTP_X1" }, + { 0x2d, "PTP_X2" }, + { 0x2e, "PTP_X3" }, + { 0x2f, "PTP_X4" }, + { 0x30, "PTP_X5" }, + { 0x31, "PTP_X6" }, + { 0x32, "PTP_X7" }, + { 0x33, "NVENC2" }, + { 0x34, "VPR_SCRUBBER0" }, + { 0x35, "VPR_SCRUBBER1" }, + { 0x36, "DWBIF" }, + { 0x37, "FBFALCON" }, + { 0x38, "CE_SHIM" }, + { 0x39, "GSP" }, + {} +}; + +const struct nvkm_enum +gv100_fifo_mmu_fault_reason[] = { + { 0x00, "PDE" }, + { 0x01, "PDE_SIZE" }, + { 0x02, "PTE" }, + { 0x03, "VA_LIMIT_VIOLATION" }, + { 0x04, "UNBOUND_INST_BLOCK" }, + { 0x05, "PRIV_VIOLATION" }, + { 0x06, "RO_VIOLATION" }, + { 0x07, "WO_VIOLATION" }, + { 0x08, "PITCH_MASK_VIOLATION" }, + { 0x09, "WORK_CREATION" }, + { 0x0a, "UNSUPPORTED_APERTURE" }, + { 0x0b, "COMPRESSION_FAILURE" }, + { 0x0c, "UNSUPPORTED_KIND" }, + { 0x0d, "REGION_VIOLATION" }, + { 0x0e, "POISONED" }, + { 0x0f, "ATOMIC_VIOLATION" }, + {} +}; + +static const struct nvkm_enum +gv100_fifo_mmu_fault_engine[] = { + { 0x01, "DISPLAY" }, + { 0x03, "PTP" }, + { 0x04, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0x05, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, + { 0x06, "PWR_PMU" }, + { 0x08, "IFB", NULL, NVKM_ENGINE_IFB }, + { 0x09, "PERF" }, + { 0x1f, "PHYSICAL" }, + { 0x20, "HOST0" }, + { 0x21, "HOST1" }, + { 0x22, "HOST2" }, + { 0x23, "HOST3" }, + { 0x24, "HOST4" }, + { 0x25, "HOST5" }, + { 0x26, "HOST6" }, + { 0x27, "HOST7" }, + { 0x28, "HOST8" }, + { 0x29, "HOST9" }, + { 0x2a, "HOST10" }, + { 0x2b, "HOST11" }, + { 0x2c, "HOST12" }, + { 0x2d, "HOST13" }, + {} +}; + +const struct nvkm_enum +gv100_fifo_mmu_fault_access[] = { + { 0x0, "VIRT_READ" }, + { 0x1, "VIRT_WRITE" }, + { 0x2, "VIRT_ATOMIC" }, + { 0x3, "VIRT_PREFETCH" }, + { 0x4, "VIRT_ATOMIC_WEAK" }, + { 0x8, "PHYS_READ" }, + { 0x9, "PHYS_WRITE" }, + { 0xa, "PHYS_ATOMIC" }, + { 0xb, "PHYS_PREFETCH" }, + {} +}; + +static const struct nvkm_fifo_func_mmu_fault +gv100_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gv100_fifo_mmu_fault_access, + .engine = gv100_fifo_mmu_fault_engine, + .reason = gv100_fifo_mmu_fault_reason, + .hubclient = gv100_fifo_mmu_fault_hubclient, + .gpcclient = gv100_fifo_mmu_fault_gpcclient, +}; + +static void +gv100_fifo_intr_ctxsw_timeout(struct nvkm_fifo *fifo, u32 engm) +{ + struct nvkm_runl *runl; + struct nvkm_engn *engn; + + nvkm_runl_foreach(runl, fifo) { + nvkm_runl_foreach_engn_cond(engn, runl, engm & BIT(engn->id)) + nvkm_runl_rc_engn(runl, engn); + } +} + +static const struct nvkm_fifo_func +gv100_fifo = { + .chid_nr = gm200_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gm200_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = gk104_fifo_init_pbdmas, + .intr = gk104_fifo_intr, + .intr_ctxsw_timeout = gv100_fifo_intr_ctxsw_timeout, + .mmu_fault = &gv100_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &gv100_runl, + .runq = &gv100_runq, + .engn = &gv100_engn, + .engn_ce = &gv100_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp, .force = true }, + .chan = {{ 0, 0, VOLTA_CHANNEL_GPFIFO_A }, &gv100_chan }, +}; + +int +gv100_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&gv100_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c new file mode 100644 index 0000000000..674faf002b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c @@ -0,0 +1,543 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include "regsnv04.h" + +#include +#include +#include +#include +#include + +#include + +void +nv04_chan_stop(struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_memory *fctx = device->imem->ramfc; + const struct nvkm_ramfc_layout *c; + unsigned long flags; + u32 data = chan->ramfc_offset; + u32 chid; + + /* prevent fifo context switches */ + spin_lock_irqsave(&fifo->lock, flags); + nvkm_wr32(device, NV03_PFIFO_CACHES, 0); + + /* if this channel is active, replace it with a null context */ + chid = nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH1) & fifo->chid->mask; + if (chid == chan->id) { + nvkm_mask(device, NV04_PFIFO_CACHE1_DMA_PUSH, 0x00000001, 0); + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 0); + nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0); + + c = chan->func->ramfc->layout; + nvkm_kmap(fctx); + do { + u32 rm = ((1ULL << c->bits) - 1) << c->regs; + u32 cm = ((1ULL << c->bits) - 1) << c->ctxs; + u32 rv = (nvkm_rd32(device, c->regp) & rm) >> c->regs; + u32 cv = (nvkm_ro32(fctx, c->ctxp + data) & ~cm); + nvkm_wo32(fctx, c->ctxp + data, cv | (rv << c->ctxs)); + } while ((++c)->bits); + nvkm_done(fctx); + + c = chan->func->ramfc->layout; + do { + nvkm_wr32(device, c->regp, 0x00000000); + } while ((++c)->bits); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, 0); + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUT, 0); + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->chid->mask); + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); + } + + /* restore normal operation, after disabling dma mode */ + nvkm_mask(device, NV04_PFIFO_MODE, BIT(chan->id), 0); + nvkm_wr32(device, NV03_PFIFO_CACHES, 1); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +void +nv04_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + nvkm_mask(fifo->engine.subdev.device, NV04_PFIFO_MODE, BIT(chan->id), BIT(chan->id)); + spin_unlock_irqrestore(&fifo->lock, flags); +} + +void +nv04_chan_ramfc_clear(struct nvkm_chan *chan) +{ + struct nvkm_memory *ramfc = chan->cgrp->runl->fifo->engine.subdev.device->imem->ramfc; + const struct nvkm_ramfc_layout *c = chan->func->ramfc->layout; + + nvkm_kmap(ramfc); + do { + nvkm_wo32(ramfc, chan->ramfc_offset + c->ctxp, 0x00000000); + } while ((++c)->bits); + nvkm_done(ramfc); +} + +static int +nv04_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_memory *ramfc = chan->cgrp->runl->fifo->engine.subdev.device->imem->ramfc; + const u32 base = chan->id * 32; + + chan->ramfc_offset = base; + + nvkm_kmap(ramfc); + nvkm_wo32(ramfc, base + 0x00, offset); + nvkm_wo32(ramfc, base + 0x04, offset); + nvkm_wo32(ramfc, base + 0x08, chan->push->addr >> 4); + nvkm_wo32(ramfc, base + 0x10, NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES | + NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES | +#ifdef __BIG_ENDIAN + NV_PFIFO_CACHE1_BIG_ENDIAN | +#endif + NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8); + nvkm_done(ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +nv04_chan_ramfc = { + .layout = (const struct nvkm_ramfc_layout[]) { + { 32, 0, 0x00, 0, NV04_PFIFO_CACHE1_DMA_PUT }, + { 32, 0, 0x04, 0, NV04_PFIFO_CACHE1_DMA_GET }, + { 16, 0, 0x08, 0, NV04_PFIFO_CACHE1_DMA_INSTANCE }, + { 16, 16, 0x08, 0, NV04_PFIFO_CACHE1_DMA_DCOUNT }, + { 32, 0, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_STATE }, + { 32, 0, 0x10, 0, NV04_PFIFO_CACHE1_DMA_FETCH }, + { 32, 0, 0x14, 0, NV04_PFIFO_CACHE1_ENGINE }, + { 32, 0, 0x18, 0, NV04_PFIFO_CACHE1_PULL1 }, + {} + }, + .write = nv04_chan_ramfc_write, + .clear = nv04_chan_ramfc_clear, + .ctxdma = true, +}; + +const struct nvkm_chan_func_userd +nv04_chan_userd = { + .bar = 0, + .base = 0x800000, + .size = 0x010000, +}; + +const struct nvkm_chan_func_inst +nv04_chan_inst = { + .size = 0x1000, +}; + +static const struct nvkm_chan_func +nv04_chan = { + .inst = &nv04_chan_inst, + .userd = &nv04_chan_userd, + .ramfc = &nv04_chan_ramfc, + .start = nv04_chan_start, + .stop = nv04_chan_stop, +}; + +const struct nvkm_cgrp_func +nv04_cgrp = { +}; + +void +nv04_eobj_ramht_del(struct nvkm_chan *chan, int hash) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_instmem *imem = fifo->engine.subdev.device->imem; + + mutex_lock(&fifo->mutex); + nvkm_ramht_remove(imem->ramht, hash); + mutex_unlock(&fifo->mutex); +} + +static int +nv04_eobj_ramht_add(struct nvkm_engn *engn, struct nvkm_object *eobj, struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_instmem *imem = fifo->engine.subdev.device->imem; + u32 context = 0x80000000 | chan->id << 24 | engn->id << 16; + int hash; + + mutex_lock(&fifo->mutex); + hash = nvkm_ramht_insert(imem->ramht, eobj, chan->id, 4, eobj->handle, context); + mutex_unlock(&fifo->mutex); + return hash; +} + +const struct nvkm_engn_func +nv04_engn = { + .ramht_add = nv04_eobj_ramht_add, + .ramht_del = nv04_eobj_ramht_del, +}; + +void +nv04_fifo_pause(struct nvkm_fifo *fifo, unsigned long *pflags) +__acquires(fifo->lock) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + *pflags = flags; + + nvkm_wr32(device, NV03_PFIFO_CACHES, 0x00000000); + nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000000); + + /* in some cases the puller may be left in an inconsistent state + * if you try to stop it while it's busy translating handles. + * sometimes you get a CACHE_ERROR, sometimes it just fails + * silently; sending incorrect instance offsets to PGRAPH after + * it's started up again. + * + * to avoid this, we invalidate the most recently calculated + * instance. + */ + nvkm_msec(device, 2000, + u32 tmp = nvkm_rd32(device, NV04_PFIFO_CACHE1_PULL0); + if (!(tmp & NV04_PFIFO_CACHE1_PULL0_HASH_BUSY)) + break; + ); + + if (nvkm_rd32(device, NV04_PFIFO_CACHE1_PULL0) & + NV04_PFIFO_CACHE1_PULL0_HASH_FAILED) + nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR); + + nvkm_wr32(device, NV04_PFIFO_CACHE1_HASH, 0x00000000); +} + +void +nv04_fifo_start(struct nvkm_fifo *fifo, unsigned long *pflags) +__releases(fifo->lock) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + unsigned long flags = *pflags; + + nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000001); + nvkm_wr32(device, NV03_PFIFO_CACHES, 0x00000001); + + spin_unlock_irqrestore(&fifo->lock, flags); +} + +const struct nvkm_runl_func +nv04_runl = { +}; + +static const char * +nv_dma_state_err(u32 state) +{ + static const char * const desc[] = { + "NONE", "CALL_SUBR_ACTIVE", "INVALID_MTHD", "RET_SUBR_INACTIVE", + "INVALID_CMD", "IB_EMPTY"/* NV50+ */, "MEM_FAULT", "UNK" + }; + return desc[(state >> 29) & 0x7]; +} + +static bool +nv04_fifo_swmthd(struct nvkm_device *device, u32 chid, u32 addr, u32 data) +{ + struct nvkm_sw *sw = device->sw; + const int subc = (addr & 0x0000e000) >> 13; + const int mthd = (addr & 0x00001ffc); + const u32 mask = 0x0000000f << (subc * 4); + u32 engine = nvkm_rd32(device, 0x003280); + bool handled = false; + + switch (mthd) { + case 0x0000 ... 0x0000: /* subchannel's engine -> software */ + nvkm_wr32(device, 0x003280, (engine &= ~mask)); + fallthrough; + case 0x0180 ... 0x01fc: /* handle -> instance */ + data = nvkm_rd32(device, 0x003258) & 0x0000ffff; + fallthrough; + case 0x0100 ... 0x017c: + case 0x0200 ... 0x1ffc: /* pass method down to sw */ + if (!(engine & mask) && sw) + handled = nvkm_sw_mthd(sw, chid, subc, mthd, data); + break; + default: + break; + } + + return handled; +} + +static void +nv04_fifo_intr_cache_error(struct nvkm_fifo *fifo, u32 chid, u32 get) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_chan *chan; + unsigned long flags; + u32 pull0 = nvkm_rd32(device, 0x003250); + u32 mthd, data; + int ptr; + + /* NV_PFIFO_CACHE1_GET actually goes to 0xffc before wrapping on my + * G80 chips, but CACHE1 isn't big enough for this much data.. Tests + * show that it wraps around to the start at GET=0x800.. No clue as to + * why.. + */ + ptr = (get & 0x7ff) >> 2; + + if (device->card_type < NV_40) { + mthd = nvkm_rd32(device, NV04_PFIFO_CACHE1_METHOD(ptr)); + data = nvkm_rd32(device, NV04_PFIFO_CACHE1_DATA(ptr)); + } else { + mthd = nvkm_rd32(device, NV40_PFIFO_CACHE1_METHOD(ptr)); + data = nvkm_rd32(device, NV40_PFIFO_CACHE1_DATA(ptr)); + } + + if (!(pull0 & 0x00000100) || + !nv04_fifo_swmthd(device, chid, mthd, data)) { + chan = nvkm_chan_get_chid(&fifo->engine, chid, &flags); + nvkm_error(subdev, "CACHE_ERROR - " + "ch %d [%s] subc %d mthd %04x data %08x\n", + chid, chan ? chan->name : "unknown", + (mthd >> 13) & 7, mthd & 0x1ffc, data); + nvkm_chan_put(&chan, flags); + } + + nvkm_wr32(device, NV04_PFIFO_CACHE1_DMA_PUSH, 0); + nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, + nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH0) & ~1); + nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, get + 4); + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, + nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH0) | 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_HASH, 0); + + nvkm_wr32(device, NV04_PFIFO_CACHE1_DMA_PUSH, + nvkm_rd32(device, NV04_PFIFO_CACHE1_DMA_PUSH) | 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); +} + +static void +nv04_fifo_intr_dma_pusher(struct nvkm_fifo *fifo, u32 chid) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 dma_get = nvkm_rd32(device, 0x003244); + u32 dma_put = nvkm_rd32(device, 0x003240); + u32 push = nvkm_rd32(device, 0x003220); + u32 state = nvkm_rd32(device, 0x003228); + struct nvkm_chan *chan; + unsigned long flags; + const char *name; + + chan = nvkm_chan_get_chid(&fifo->engine, chid, &flags); + name = chan ? chan->name : "unknown"; + if (device->card_type == NV_50) { + u32 ho_get = nvkm_rd32(device, 0x003328); + u32 ho_put = nvkm_rd32(device, 0x003320); + u32 ib_get = nvkm_rd32(device, 0x003334); + u32 ib_put = nvkm_rd32(device, 0x003330); + + nvkm_error(subdev, "DMA_PUSHER - " + "ch %d [%s] get %02x%08x put %02x%08x ib_get %08x " + "ib_put %08x state %08x (err: %s) push %08x\n", + chid, name, ho_get, dma_get, ho_put, dma_put, + ib_get, ib_put, state, nv_dma_state_err(state), + push); + + /* METHOD_COUNT, in DMA_STATE on earlier chipsets */ + nvkm_wr32(device, 0x003364, 0x00000000); + if (dma_get != dma_put || ho_get != ho_put) { + nvkm_wr32(device, 0x003244, dma_put); + nvkm_wr32(device, 0x003328, ho_put); + } else + if (ib_get != ib_put) + nvkm_wr32(device, 0x003334, ib_put); + } else { + nvkm_error(subdev, "DMA_PUSHER - ch %d [%s] get %08x put %08x " + "state %08x (err: %s) push %08x\n", + chid, name, dma_get, dma_put, state, + nv_dma_state_err(state), push); + + if (dma_get != dma_put) + nvkm_wr32(device, 0x003244, dma_put); + } + nvkm_chan_put(&chan, flags); + + nvkm_wr32(device, 0x003228, 0x00000000); + nvkm_wr32(device, 0x003220, 0x00000001); + nvkm_wr32(device, 0x002100, NV_PFIFO_INTR_DMA_PUSHER); +} + +irqreturn_t +nv04_fifo_intr(struct nvkm_inth *inth) +{ + struct nvkm_fifo *fifo = container_of(inth, typeof(*fifo), engine.subdev.inth); + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, NV03_PFIFO_INTR_EN_0); + u32 stat = nvkm_rd32(device, NV03_PFIFO_INTR_0) & mask; + u32 reassign, chid, get, sem; + + reassign = nvkm_rd32(device, NV03_PFIFO_CACHES) & 1; + nvkm_wr32(device, NV03_PFIFO_CACHES, 0); + + chid = nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH1) & fifo->chid->mask; + get = nvkm_rd32(device, NV03_PFIFO_CACHE1_GET); + + if (stat & NV_PFIFO_INTR_CACHE_ERROR) { + nv04_fifo_intr_cache_error(fifo, chid, get); + stat &= ~NV_PFIFO_INTR_CACHE_ERROR; + } + + if (stat & NV_PFIFO_INTR_DMA_PUSHER) { + nv04_fifo_intr_dma_pusher(fifo, chid); + stat &= ~NV_PFIFO_INTR_DMA_PUSHER; + } + + if (stat & NV_PFIFO_INTR_SEMAPHORE) { + stat &= ~NV_PFIFO_INTR_SEMAPHORE; + nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_SEMAPHORE); + + sem = nvkm_rd32(device, NV10_PFIFO_CACHE1_SEMAPHORE); + nvkm_wr32(device, NV10_PFIFO_CACHE1_SEMAPHORE, sem | 0x1); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, get + 4); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); + } + + if (device->card_type == NV_50) { + if (stat & 0x00000010) { + stat &= ~0x00000010; + nvkm_wr32(device, 0x002100, 0x00000010); + } + + if (stat & 0x40000000) { + nvkm_wr32(device, 0x002100, 0x40000000); + nvkm_event_ntfy(&fifo->nonstall.event, 0, NVKM_FIFO_NONSTALL_EVENT); + stat &= ~0x40000000; + } + } + + if (stat) { + nvkm_warn(subdev, "intr %08x\n", stat); + nvkm_mask(device, NV03_PFIFO_INTR_EN_0, stat, 0x00000000); + nvkm_wr32(device, NV03_PFIFO_INTR_0, stat); + } + + nvkm_wr32(device, NV03_PFIFO_CACHES, reassign); + return IRQ_HANDLED; +} + +void +nv04_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_instmem *imem = device->imem; + struct nvkm_ramht *ramht = imem->ramht; + struct nvkm_memory *ramro = imem->ramro; + struct nvkm_memory *ramfc = imem->ramfc; + + nvkm_wr32(device, NV04_PFIFO_DELAY_0, 0x000000ff); + nvkm_wr32(device, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff); + + nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ | + ((ramht->bits - 9) << 16) | + (ramht->gpuobj->addr >> 8)); + nvkm_wr32(device, NV03_PFIFO_RAMRO, nvkm_memory_addr(ramro) >> 8); + nvkm_wr32(device, NV03_PFIFO_RAMFC, nvkm_memory_addr(ramfc) >> 8); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->chid->mask); + + nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff); + nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); + nvkm_wr32(device, NV03_PFIFO_CACHES, 1); +} + +int +nv04_fifo_runl_ctor(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl; + + runl = nvkm_runl_new(fifo, 0, 0, 0); + if (IS_ERR(runl)) + return PTR_ERR(runl); + + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_SW, 0); + nvkm_runl_add(runl, 0, fifo->func->engn_sw, NVKM_ENGINE_DMAOBJ, 0); + nvkm_runl_add(runl, 1, fifo->func->engn , NVKM_ENGINE_GR, 0); + nvkm_runl_add(runl, 2, fifo->func->engn , NVKM_ENGINE_MPEG, 0); /* NV31- */ + return 0; +} + +int +nv04_fifo_chid_ctor(struct nvkm_fifo *fifo, int nr) +{ + /* The last CHID is reserved by HW as a "channel invalid" marker. */ + return nvkm_chid_new(&nvkm_chan_event, &fifo->engine.subdev, nr, 0, nr - 1, &fifo->chid); +} + +static int +nv04_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 16; +} + +static const struct nvkm_fifo_func +nv04_fifo = { + .chid_nr = nv04_fifo_chid_nr, + .chid_ctor = nv04_fifo_chid_ctor, + .runl_ctor = nv04_fifo_runl_ctor, + .init = nv04_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .runl = &nv04_runl, + .engn = &nv04_engn, + .engn_sw = &nv04_engn, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, NV03_CHANNEL_DMA }, &nv04_chan }, +}; + +int +nv04_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&nv04_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv10.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv10.c new file mode 100644 index 0000000000..a4bcf6b0a7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv10.c @@ -0,0 +1,113 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "runl.h" + +#include +#include + +#include "regsnv04.h" + +#include + +static int +nv10_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_memory *ramfc = chan->cgrp->runl->fifo->engine.subdev.device->imem->ramfc; + const u32 base = chan->id * 32; + + chan->ramfc_offset = base; + + nvkm_kmap(ramfc); + nvkm_wo32(ramfc, base + 0x00, offset); + nvkm_wo32(ramfc, base + 0x04, offset); + nvkm_wo32(ramfc, base + 0x0c, chan->push->addr >> 4); + nvkm_wo32(ramfc, base + 0x14, NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES | + NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES | +#ifdef __BIG_ENDIAN + NV_PFIFO_CACHE1_BIG_ENDIAN | +#endif + NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8); + nvkm_done(ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +nv10_chan_ramfc = { + .layout = (const struct nvkm_ramfc_layout[]) { + { 32, 0, 0x00, 0, NV04_PFIFO_CACHE1_DMA_PUT }, + { 32, 0, 0x04, 0, NV04_PFIFO_CACHE1_DMA_GET }, + { 32, 0, 0x08, 0, NV10_PFIFO_CACHE1_REF_CNT }, + { 16, 0, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_INSTANCE }, + { 16, 16, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_DCOUNT }, + { 32, 0, 0x10, 0, NV04_PFIFO_CACHE1_DMA_STATE }, + { 32, 0, 0x14, 0, NV04_PFIFO_CACHE1_DMA_FETCH }, + { 32, 0, 0x18, 0, NV04_PFIFO_CACHE1_ENGINE }, + { 32, 0, 0x1c, 0, NV04_PFIFO_CACHE1_PULL1 }, + {} + }, + .write = nv10_chan_ramfc_write, + .clear = nv04_chan_ramfc_clear, + .ctxdma = true, +}; + +static const struct nvkm_chan_func +nv10_chan = { + .inst = &nv04_chan_inst, + .userd = &nv04_chan_userd, + .ramfc = &nv10_chan_ramfc, + .start = nv04_chan_start, + .stop = nv04_chan_stop, +}; + +int +nv10_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 32; +} + +static const struct nvkm_fifo_func +nv10_fifo = { + .chid_nr = nv10_fifo_chid_nr, + .chid_ctor = nv04_fifo_chid_ctor, + .runl_ctor = nv04_fifo_runl_ctor, + .init = nv04_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .runl = &nv04_runl, + .engn = &nv04_engn, + .engn_sw = &nv04_engn, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, NV10_CHANNEL_DMA }, &nv10_chan }, +}; + +int +nv10_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&nv10_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c new file mode 100644 index 0000000000..c70f44fd4f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c @@ -0,0 +1,142 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include "regsnv04.h" + +#include +#include + +#include + +static int +nv17_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_memory *ramfc = chan->cgrp->runl->fifo->engine.subdev.device->imem->ramfc; + const u32 base = chan->id * 64; + + chan->ramfc_offset = base; + + nvkm_kmap(ramfc); + nvkm_wo32(ramfc, base + 0x00, offset); + nvkm_wo32(ramfc, base + 0x04, offset); + nvkm_wo32(ramfc, base + 0x0c, chan->push->addr >> 4); + nvkm_wo32(ramfc, base + 0x14, NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES | + NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES | +#ifdef __BIG_ENDIAN + NV_PFIFO_CACHE1_BIG_ENDIAN | +#endif + NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8); + nvkm_done(ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +nv17_chan_ramfc = { + .layout = (const struct nvkm_ramfc_layout[]) { + { 32, 0, 0x00, 0, NV04_PFIFO_CACHE1_DMA_PUT }, + { 32, 0, 0x04, 0, NV04_PFIFO_CACHE1_DMA_GET }, + { 32, 0, 0x08, 0, NV10_PFIFO_CACHE1_REF_CNT }, + { 16, 0, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_INSTANCE }, + { 16, 16, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_DCOUNT }, + { 32, 0, 0x10, 0, NV04_PFIFO_CACHE1_DMA_STATE }, + { 32, 0, 0x14, 0, NV04_PFIFO_CACHE1_DMA_FETCH }, + { 32, 0, 0x18, 0, NV04_PFIFO_CACHE1_ENGINE }, + { 32, 0, 0x1c, 0, NV04_PFIFO_CACHE1_PULL1 }, + { 32, 0, 0x20, 0, NV10_PFIFO_CACHE1_ACQUIRE_VALUE }, + { 32, 0, 0x24, 0, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP }, + { 32, 0, 0x28, 0, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT }, + { 32, 0, 0x2c, 0, NV10_PFIFO_CACHE1_SEMAPHORE }, + { 32, 0, 0x30, 0, NV10_PFIFO_CACHE1_DMA_SUBROUTINE }, + {} + }, + .write = nv17_chan_ramfc_write, + .clear = nv04_chan_ramfc_clear, + .ctxdma = true, +}; + +static const struct nvkm_chan_func +nv17_chan = { + .inst = &nv04_chan_inst, + .userd = &nv04_chan_userd, + .ramfc = &nv17_chan_ramfc, + .start = nv04_chan_start, + .stop = nv04_chan_stop, +}; + +static void +nv17_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_instmem *imem = device->imem; + struct nvkm_ramht *ramht = imem->ramht; + struct nvkm_memory *ramro = imem->ramro; + struct nvkm_memory *ramfc = imem->ramfc; + + nvkm_wr32(device, NV04_PFIFO_DELAY_0, 0x000000ff); + nvkm_wr32(device, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff); + + nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ | + ((ramht->bits - 9) << 16) | + (ramht->gpuobj->addr >> 8)); + nvkm_wr32(device, NV03_PFIFO_RAMRO, nvkm_memory_addr(ramro) >> 8); + nvkm_wr32(device, NV03_PFIFO_RAMFC, nvkm_memory_addr(ramfc) >> 8 | + 0x00010000); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->chid->mask); + + nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff); + nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); + nvkm_wr32(device, NV03_PFIFO_CACHES, 1); +} + +static const struct nvkm_fifo_func +nv17_fifo = { + .chid_nr = nv10_fifo_chid_nr, + .chid_ctor = nv04_fifo_chid_ctor, + .runl_ctor = nv04_fifo_runl_ctor, + .init = nv17_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .runl = &nv04_runl, + .engn = &nv04_engn, + .engn_sw = &nv04_engn, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, NV17_CHANNEL_DMA }, &nv17_chan }, +}; + +int +nv17_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&nv17_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c new file mode 100644 index 0000000000..e50a94b6d7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c @@ -0,0 +1,252 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include "regsnv04.h" + +#include +#include +#include + +#include + +static int +nv40_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_memory *ramfc = chan->cgrp->runl->fifo->engine.subdev.device->imem->ramfc; + const u32 base = chan->id * 128; + + chan->ramfc_offset = base; + + nvkm_kmap(ramfc); + nvkm_wo32(ramfc, base + 0x00, offset); + nvkm_wo32(ramfc, base + 0x04, offset); + nvkm_wo32(ramfc, base + 0x0c, chan->push->addr >> 4); + nvkm_wo32(ramfc, base + 0x18, 0x30000000 | + NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES | + NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES | +#ifdef __BIG_ENDIAN + NV_PFIFO_CACHE1_BIG_ENDIAN | +#endif + NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8); + nvkm_wo32(ramfc, base + 0x3c, 0x0001ffff); + nvkm_done(ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +nv40_chan_ramfc = { + .layout = (const struct nvkm_ramfc_layout[]) { + { 32, 0, 0x00, 0, NV04_PFIFO_CACHE1_DMA_PUT }, + { 32, 0, 0x04, 0, NV04_PFIFO_CACHE1_DMA_GET }, + { 32, 0, 0x08, 0, NV10_PFIFO_CACHE1_REF_CNT }, + { 32, 0, 0x0c, 0, NV04_PFIFO_CACHE1_DMA_INSTANCE }, + { 32, 0, 0x10, 0, NV04_PFIFO_CACHE1_DMA_DCOUNT }, + { 32, 0, 0x14, 0, NV04_PFIFO_CACHE1_DMA_STATE }, + { 28, 0, 0x18, 0, NV04_PFIFO_CACHE1_DMA_FETCH }, + { 2, 28, 0x18, 28, 0x002058 }, + { 32, 0, 0x1c, 0, NV04_PFIFO_CACHE1_ENGINE }, + { 32, 0, 0x20, 0, NV04_PFIFO_CACHE1_PULL1 }, + { 32, 0, 0x24, 0, NV10_PFIFO_CACHE1_ACQUIRE_VALUE }, + { 32, 0, 0x28, 0, NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP }, + { 32, 0, 0x2c, 0, NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT }, + { 32, 0, 0x30, 0, NV10_PFIFO_CACHE1_SEMAPHORE }, + { 32, 0, 0x34, 0, NV10_PFIFO_CACHE1_DMA_SUBROUTINE }, + { 32, 0, 0x38, 0, NV40_PFIFO_GRCTX_INSTANCE }, + { 17, 0, 0x3c, 0, NV04_PFIFO_DMA_TIMESLICE }, + { 32, 0, 0x40, 0, 0x0032e4 }, + { 32, 0, 0x44, 0, 0x0032e8 }, + { 32, 0, 0x4c, 0, 0x002088 }, + { 32, 0, 0x50, 0, 0x003300 }, + { 32, 0, 0x54, 0, 0x00330c }, + {} + }, + .write = nv40_chan_ramfc_write, + .clear = nv04_chan_ramfc_clear, + .ctxdma = true, +}; + +static const struct nvkm_chan_func_userd +nv40_chan_userd = { + .bar = 0, + .base = 0xc00000, + .size = 0x001000, +}; + +static const struct nvkm_chan_func +nv40_chan = { + .inst = &nv04_chan_inst, + .userd = &nv40_chan_userd, + .ramfc = &nv40_chan_ramfc, + .start = nv04_chan_start, + .stop = nv04_chan_stop, +}; + +static int +nv40_eobj_ramht_add(struct nvkm_engn *engn, struct nvkm_object *eobj, struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_instmem *imem = fifo->engine.subdev.device->imem; + u32 context = chan->id << 23 | engn->id << 20; + int hash; + + mutex_lock(&fifo->mutex); + hash = nvkm_ramht_insert(imem->ramht, eobj, chan->id, 4, eobj->handle, context); + mutex_unlock(&fifo->mutex); + return hash; +} + +static void +nv40_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + struct nvkm_fifo *fifo = chan->cgrp->runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_memory *ramfc = device->imem->ramfc; + u32 inst = 0x00000000, reg, ctx; + int chid; + + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_GR: + reg = 0x0032e0; + ctx = 0x38; + break; + case NVKM_ENGINE_MPEG: + if (WARN_ON(device->chipset < 0x44)) + return; + reg = 0x00330c; + ctx = 0x54; + break; + default: + WARN_ON(1); + return; + } + + if (cctx) + inst = cctx->vctx->inst->addr >> 4; + + spin_lock_irq(&fifo->lock); + nvkm_mask(device, 0x002500, 0x00000001, 0x00000000); + + chid = nvkm_rd32(device, 0x003204) & (fifo->chid->nr - 1); + if (chid == chan->id) + nvkm_wr32(device, reg, inst); + + nvkm_kmap(ramfc); + nvkm_wo32(ramfc, chan->ramfc_offset + ctx, inst); + nvkm_done(ramfc); + + nvkm_mask(device, 0x002500, 0x00000001, 0x00000001); + spin_unlock_irq(&fifo->lock); +} + +static const struct nvkm_engn_func +nv40_engn = { + .bind = nv40_ectx_bind, + .ramht_add = nv40_eobj_ramht_add, + .ramht_del = nv04_eobj_ramht_del, +}; + +static const struct nvkm_engn_func +nv40_engn_sw = { + .ramht_add = nv40_eobj_ramht_add, + .ramht_del = nv04_eobj_ramht_del, +}; + +static void +nv40_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_fb *fb = device->fb; + struct nvkm_instmem *imem = device->imem; + struct nvkm_ramht *ramht = imem->ramht; + struct nvkm_memory *ramro = imem->ramro; + struct nvkm_memory *ramfc = imem->ramfc; + + nvkm_wr32(device, 0x002040, 0x000000ff); + nvkm_wr32(device, 0x002044, 0x2101ffff); + nvkm_wr32(device, 0x002058, 0x00000001); + + nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ | + ((ramht->bits - 9) << 16) | + (ramht->gpuobj->addr >> 8)); + nvkm_wr32(device, NV03_PFIFO_RAMRO, nvkm_memory_addr(ramro) >> 8); + + switch (device->chipset) { + case 0x47: + case 0x49: + case 0x4b: + nvkm_wr32(device, 0x002230, 0x00000001); + fallthrough; + case 0x40: + case 0x41: + case 0x42: + case 0x43: + case 0x45: + case 0x48: + nvkm_wr32(device, 0x002220, 0x00030002); + break; + default: + nvkm_wr32(device, 0x002230, 0x00000000); + nvkm_wr32(device, 0x002220, ((fb->ram->size - 512 * 1024 + + nvkm_memory_addr(ramfc)) >> 16) | + 0x00030000); + break; + } + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->chid->mask); + + nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff); + nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff); + + nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1); + nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1); + nvkm_wr32(device, NV03_PFIFO_CACHES, 1); +} + +static const struct nvkm_fifo_func +nv40_fifo = { + .chid_nr = nv10_fifo_chid_nr, + .chid_ctor = nv04_fifo_chid_ctor, + .runl_ctor = nv04_fifo_runl_ctor, + .init = nv40_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .runl = &nv04_runl, + .engn = &nv40_engn, + .engn_sw = &nv40_engn_sw, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, NV40_CHANNEL_DMA }, &nv40_chan }, +}; + +int +nv40_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&nv40_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c new file mode 100644 index 0000000000..954b5f3a7d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c @@ -0,0 +1,398 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include +#include + +#include + +void +nv50_eobj_ramht_del(struct nvkm_chan *chan, int hash) +{ + nvkm_ramht_remove(chan->ramht, hash); +} + +int +nv50_eobj_ramht_add(struct nvkm_engn *engn, struct nvkm_object *eobj, struct nvkm_chan *chan) +{ + return nvkm_ramht_insert(chan->ramht, eobj, 0, 4, eobj->handle, engn->id << 20); +} + +void +nv50_chan_stop(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x002600 + (chan->id * 4), 0x80000000, 0x00000000); +} + +void +nv50_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_mask(device, 0x002600 + (chan->id * 4), 0x80000000, 0x80000000); +} + +void +nv50_chan_unbind(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x002600 + (chan->id * 4), 0x00000000); +} + +static void +nv50_chan_bind(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + nvkm_wr32(device, 0x002600 + (chan->id * 4), chan->ramfc->addr >> 12); +} + +static int +nv50_chan_ramfc_write(struct nvkm_chan *chan, u64 offset, u64 length, u32 devm, bool priv) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + const u32 limit2 = ilog2(length / 8); + int ret; + + ret = nvkm_gpuobj_new(device, 0x0200, 0x1000, true, chan->inst, &chan->ramfc); + if (ret) + return ret; + + ret = nvkm_gpuobj_new(device, 0x1200, 0, true, chan->inst, &chan->eng); + if (ret) + return ret; + + ret = nvkm_gpuobj_new(device, 0x4000, 0, false, chan->inst, &chan->pgd); + if (ret) + return ret; + + ret = nvkm_ramht_new(device, 0x8000, 16, chan->inst, &chan->ramht); + if (ret) + return ret; + + nvkm_kmap(chan->ramfc); + nvkm_wo32(chan->ramfc, 0x3c, 0x403f6078); + nvkm_wo32(chan->ramfc, 0x44, 0x01003fff); + nvkm_wo32(chan->ramfc, 0x48, chan->push->node->offset >> 4); + nvkm_wo32(chan->ramfc, 0x50, lower_32_bits(offset)); + nvkm_wo32(chan->ramfc, 0x54, upper_32_bits(offset) | (limit2 << 16)); + nvkm_wo32(chan->ramfc, 0x60, 0x7fffffff); + nvkm_wo32(chan->ramfc, 0x78, 0x00000000); + nvkm_wo32(chan->ramfc, 0x7c, 0x30000000 | devm); + nvkm_wo32(chan->ramfc, 0x80, ((chan->ramht->bits - 9) << 27) | + (4 << 24) /* SEARCH_FULL */ | + (chan->ramht->gpuobj->node->offset >> 4)); + nvkm_done(chan->ramfc); + return 0; +} + +static const struct nvkm_chan_func_ramfc +nv50_chan_ramfc = { + .write = nv50_chan_ramfc_write, + .ctxdma = true, + .devm = 0xfff, +}; + +const struct nvkm_chan_func_userd +nv50_chan_userd = { + .bar = 0, + .base = 0xc00000, + .size = 0x002000, +}; + +const struct nvkm_chan_func_inst +nv50_chan_inst = { + .size = 0x10000, + .vmm = true, +}; + +static const struct nvkm_chan_func +nv50_chan = { + .inst = &nv50_chan_inst, + .userd = &nv50_chan_userd, + .ramfc = &nv50_chan_ramfc, + .bind = nv50_chan_bind, + .unbind = nv50_chan_unbind, + .start = nv50_chan_start, + .stop = nv50_chan_stop, +}; + +static void +nv50_ectx_bind(struct nvkm_engn *engn, struct nvkm_cctx *cctx, struct nvkm_chan *chan) +{ + struct nvkm_subdev *subdev = &chan->cgrp->runl->fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u64 start = 0, limit = 0; + u32 flags = 0, ptr0, save; + + switch (engn->engine->subdev.type) { + case NVKM_ENGINE_GR : ptr0 = 0x0000; break; + case NVKM_ENGINE_MPEG : ptr0 = 0x0060; break; + default: + WARN_ON(1); + return; + } + + if (!cctx) { + /* HW bug workaround: + * + * PFIFO will hang forever if the connected engines don't report + * that they've processed the context switch request. + * + * In order for the kickoff to work, we need to ensure all the + * connected engines are in a state where they can answer. + * + * Newer chipsets don't seem to suffer from this issue, and well, + * there's also a "ignore these engines" bitmask reg we can use + * if we hit the issue there.. + */ + save = nvkm_mask(device, 0x00b860, 0x00000001, 0x00000001); + + /* Tell engines to save out contexts. */ + nvkm_wr32(device, 0x0032fc, chan->inst->addr >> 12); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x0032fc) != 0xffffffff) + break; + ); + nvkm_wr32(device, 0x00b860, save); + } else { + flags = 0x00190000; + start = cctx->vctx->inst->addr; + limit = start + cctx->vctx->inst->size - 1; + } + + nvkm_kmap(chan->eng); + nvkm_wo32(chan->eng, ptr0 + 0x00, flags); + nvkm_wo32(chan->eng, ptr0 + 0x04, lower_32_bits(limit)); + nvkm_wo32(chan->eng, ptr0 + 0x08, lower_32_bits(start)); + nvkm_wo32(chan->eng, ptr0 + 0x0c, upper_32_bits(limit) << 24 | + lower_32_bits(start)); + nvkm_wo32(chan->eng, ptr0 + 0x10, 0x00000000); + nvkm_wo32(chan->eng, ptr0 + 0x14, 0x00000000); + nvkm_done(chan->eng); +} + +static const struct nvkm_engn_func +nv50_engn = { + .bind = nv50_ectx_bind, + .ramht_add = nv50_eobj_ramht_add, + .ramht_del = nv50_eobj_ramht_del, +}; + +const struct nvkm_engn_func +nv50_engn_sw = { + .ramht_add = nv50_eobj_ramht_add, + .ramht_del = nv50_eobj_ramht_del, +}; + +static bool +nv50_runl_pending(struct nvkm_runl *runl) +{ + return nvkm_rd32(runl->fifo->engine.subdev.device, 0x0032ec) & 0x00000100; +} + +int +nv50_runl_wait(struct nvkm_runl *runl) +{ + struct nvkm_fifo *fifo = runl->fifo; + + nvkm_msec(fifo->engine.subdev.device, fifo->timeout.chan_msec, + if (!nvkm_runl_update_pending(runl)) + return 0; + usleep_range(1, 2); + ); + + return -ETIMEDOUT; +} + +static void +nv50_runl_commit(struct nvkm_runl *runl, struct nvkm_memory *memory, u32 start, int count) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u64 addr = nvkm_memory_addr(memory) + start; + + nvkm_wr32(device, 0x0032f4, addr >> 12); + nvkm_wr32(device, 0x0032ec, count); +} + +static void +nv50_runl_insert_chan(struct nvkm_chan *chan, struct nvkm_memory *memory, u64 offset) +{ + nvkm_wo32(memory, offset, chan->id); +} + +static struct nvkm_memory * +nv50_runl_alloc(struct nvkm_runl *runl, u32 *offset) +{ + const u32 segment = ALIGN((runl->cgrp_nr + runl->chan_nr) * runl->func->size, 0x1000); + const u32 maxsize = (runl->cgid ? runl->cgid->nr : 0) + runl->chid->nr; + int ret; + + if (unlikely(!runl->mem)) { + ret = nvkm_memory_new(runl->fifo->engine.subdev.device, NVKM_MEM_TARGET_INST, + maxsize * 2 * runl->func->size, 0, false, &runl->mem); + if (ret) { + RUNL_ERROR(runl, "alloc %d\n", ret); + return ERR_PTR(ret); + } + } else { + if (runl->offset + segment >= nvkm_memory_size(runl->mem)) { + ret = runl->func->wait(runl); + if (ret) { + RUNL_DEBUG(runl, "rewind timeout"); + return ERR_PTR(ret); + } + + runl->offset = 0; + } + } + + *offset = runl->offset; + runl->offset += segment; + return runl->mem; +} + +int +nv50_runl_update(struct nvkm_runl *runl) +{ + struct nvkm_memory *memory; + struct nvkm_cgrp *cgrp; + struct nvkm_chan *chan; + u32 start, offset, count; + + /*TODO: prio, interleaving. */ + + RUNL_TRACE(runl, "RAMRL: update cgrps:%d chans:%d", runl->cgrp_nr, runl->chan_nr); + memory = nv50_runl_alloc(runl, &start); + if (IS_ERR(memory)) + return PTR_ERR(memory); + + RUNL_TRACE(runl, "RAMRL: update start:%08x", start); + offset = start; + + nvkm_kmap(memory); + nvkm_runl_foreach_cgrp(cgrp, runl) { + if (cgrp->hw) { + CGRP_TRACE(cgrp, " RAMRL+%08x: chans:%d", offset, cgrp->chan_nr); + runl->func->insert_cgrp(cgrp, memory, offset); + offset += runl->func->size; + } + + nvkm_cgrp_foreach_chan(chan, cgrp) { + CHAN_TRACE(chan, "RAMRL+%08x: [%s]", offset, chan->name); + runl->func->insert_chan(chan, memory, offset); + offset += runl->func->size; + } + } + nvkm_done(memory); + + /*TODO: look into using features on newer HW to guarantee forward progress. */ + list_rotate_left(&runl->cgrps); + + count = (offset - start) / runl->func->size; + RUNL_TRACE(runl, "RAMRL: commit start:%08x count:%d", start, count); + + runl->func->commit(runl, memory, start, count); + return 0; +} + +const struct nvkm_runl_func +nv50_runl = { + .size = 4, + .update = nv50_runl_update, + .insert_chan = nv50_runl_insert_chan, + .commit = nv50_runl_commit, + .wait = nv50_runl_wait, + .pending = nv50_runl_pending, +}; + +void +nv50_fifo_init(struct nvkm_fifo *fifo) +{ + struct nvkm_runl *runl = nvkm_runl_first(fifo); + struct nvkm_device *device = fifo->engine.subdev.device; + int i; + + nvkm_mask(device, 0x000200, 0x00000100, 0x00000000); + nvkm_mask(device, 0x000200, 0x00000100, 0x00000100); + nvkm_wr32(device, 0x00250c, 0x6f3cfc34); + nvkm_wr32(device, 0x002044, 0x01003fff); + + nvkm_wr32(device, 0x002100, 0xffffffff); + nvkm_wr32(device, 0x002140, 0xbfffffff); + + for (i = 0; i < 128; i++) + nvkm_wr32(device, 0x002600 + (i * 4), 0x00000000); + + atomic_set(&runl->changed, 1); + runl->func->update(runl); + + nvkm_wr32(device, 0x003200, 0x00000001); + nvkm_wr32(device, 0x003250, 0x00000001); + nvkm_wr32(device, 0x002500, 0x00000001); +} + +int +nv50_fifo_chid_ctor(struct nvkm_fifo *fifo, int nr) +{ + /* CHID 0 is unusable (some kind of PIO channel?), 127 is "channel invalid". */ + return nvkm_chid_new(&nvkm_chan_event, &fifo->engine.subdev, nr, 1, nr - 2, &fifo->chid); +} + +int +nv50_fifo_chid_nr(struct nvkm_fifo *fifo) +{ + return 128; +} + +static const struct nvkm_fifo_func +nv50_fifo = { + .chid_nr = nv50_fifo_chid_nr, + .chid_ctor = nv50_fifo_chid_ctor, + .runl_ctor = nv04_fifo_runl_ctor, + .init = nv50_fifo_init, + .intr = nv04_fifo_intr, + .pause = nv04_fifo_pause, + .start = nv04_fifo_start, + .runl = &nv50_runl, + .engn = &nv50_engn, + .engn_sw = &nv50_engn_sw, + .cgrp = {{ }, &nv04_cgrp }, + .chan = {{ 0, 0, NV50_CHANNEL_GPFIFO }, &nv50_chan }, +}; + +int +nv50_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&nv50_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h new file mode 100644 index 0000000000..4d448be192 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/priv.h @@ -0,0 +1,209 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_FIFO_PRIV_H__ +#define __NVKM_FIFO_PRIV_H__ +#define nvkm_fifo(p) container_of((p), struct nvkm_fifo, engine) +#include +#include +struct nvkm_cctx; +struct nvkm_cgrp; +struct nvkm_engn; +struct nvkm_memory; +struct nvkm_runl; +struct nvkm_runq; +struct nvkm_vctx; + +struct nvkm_fifo_func { + int (*chid_nr)(struct nvkm_fifo *); + int (*chid_ctor)(struct nvkm_fifo *, int nr); + int (*runq_nr)(struct nvkm_fifo *); + int (*runl_ctor)(struct nvkm_fifo *); + + void (*init)(struct nvkm_fifo *); + void (*init_pbdmas)(struct nvkm_fifo *, u32 mask); + + irqreturn_t (*intr)(struct nvkm_inth *); + void (*intr_mmu_fault_unit)(struct nvkm_fifo *, int unit); + void (*intr_ctxsw_timeout)(struct nvkm_fifo *, u32 engm); + + const struct nvkm_fifo_func_mmu_fault { + void (*recover)(struct nvkm_fifo *, struct nvkm_fault_data *); + const struct nvkm_enum *access; + const struct nvkm_enum *engine; + const struct nvkm_enum *reason; + const struct nvkm_enum *hubclient; + const struct nvkm_enum *gpcclient; + } *mmu_fault; + + void (*pause)(struct nvkm_fifo *, unsigned long *); + void (*start)(struct nvkm_fifo *, unsigned long *); + + int (*nonstall_ctor)(struct nvkm_fifo *); + const struct nvkm_event_func *nonstall; + + const struct nvkm_runl_func *runl; + const struct nvkm_runq_func *runq; + const struct nvkm_engn_func *engn; + const struct nvkm_engn_func *engn_sw; + const struct nvkm_engn_func *engn_ce; + + struct nvkm_fifo_func_cgrp { + struct nvkm_sclass user; + const struct nvkm_cgrp_func *func; + bool force; + } cgrp; + + struct nvkm_fifo_func_chan { + struct nvkm_sclass user; + const struct nvkm_chan_func *func; + } chan; +}; + +int nvkm_fifo_new_(const struct nvkm_fifo_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_fifo **); + +int nv04_fifo_chid_ctor(struct nvkm_fifo *, int); +int nv04_fifo_runl_ctor(struct nvkm_fifo *); +void nv04_fifo_init(struct nvkm_fifo *); +irqreturn_t nv04_fifo_intr(struct nvkm_inth *); +void nv04_fifo_pause(struct nvkm_fifo *, unsigned long *); +void nv04_fifo_start(struct nvkm_fifo *, unsigned long *); +extern const struct nvkm_runl_func nv04_runl; +extern const struct nvkm_engn_func nv04_engn; +extern const struct nvkm_cgrp_func nv04_cgrp; +extern const struct nvkm_chan_func_inst nv04_chan_inst; +extern const struct nvkm_chan_func_userd nv04_chan_userd; +void nv04_chan_ramfc_clear(struct nvkm_chan *); +void nv04_chan_start(struct nvkm_chan *); +void nv04_chan_stop(struct nvkm_chan *); +void nv04_eobj_ramht_del(struct nvkm_chan *, int); + +int nv10_fifo_chid_nr(struct nvkm_fifo *); + +int nv50_fifo_chid_nr(struct nvkm_fifo *); +int nv50_fifo_chid_ctor(struct nvkm_fifo *, int); +void nv50_fifo_init(struct nvkm_fifo *); +extern const struct nvkm_runl_func nv50_runl; +int nv50_runl_update(struct nvkm_runl *); +int nv50_runl_wait(struct nvkm_runl *); +extern const struct nvkm_engn_func nv50_engn_sw; +extern const struct nvkm_chan_func_inst nv50_chan_inst; +extern const struct nvkm_chan_func_userd nv50_chan_userd; +void nv50_chan_unbind(struct nvkm_chan *); +void nv50_chan_start(struct nvkm_chan *); +void nv50_chan_stop(struct nvkm_chan *); +void nv50_chan_preempt(struct nvkm_chan *); +int nv50_eobj_ramht_add(struct nvkm_engn *, struct nvkm_object *, struct nvkm_chan *); +void nv50_eobj_ramht_del(struct nvkm_chan *, int); + +extern const struct nvkm_event_func g84_fifo_nonstall; +extern const struct nvkm_engn_func g84_engn; +extern const struct nvkm_chan_func g84_chan; + +int gf100_fifo_chid_ctor(struct nvkm_fifo *, int); +int gf100_fifo_runq_nr(struct nvkm_fifo *); +bool gf100_fifo_intr_pbdma(struct nvkm_fifo *); +void gf100_fifo_intr_mmu_fault(struct nvkm_fifo *); +void gf100_fifo_intr_mmu_fault_unit(struct nvkm_fifo *, int); +void gf100_fifo_intr_sched(struct nvkm_fifo *); +void gf100_fifo_intr_ctxsw_timeout(struct nvkm_fifo *, u32); +void gf100_fifo_mmu_fault_recover(struct nvkm_fifo *, struct nvkm_fault_data *); +extern const struct nvkm_enum gf100_fifo_mmu_fault_access[]; +extern const struct nvkm_event_func gf100_fifo_nonstall; +bool gf100_runl_preempt_pending(struct nvkm_runl *); +void gf100_runq_init(struct nvkm_runq *); +bool gf100_runq_intr(struct nvkm_runq *, struct nvkm_runl *); +void gf100_engn_mmu_fault_trigger(struct nvkm_engn *); +bool gf100_engn_mmu_fault_triggered(struct nvkm_engn *); +extern const struct nvkm_engn_func gf100_engn_sw; +extern const struct nvkm_chan_func_inst gf100_chan_inst; +void gf100_chan_userd_clear(struct nvkm_chan *); +void gf100_chan_preempt(struct nvkm_chan *); + +int gk104_fifo_chid_nr(struct nvkm_fifo *); +int gk104_fifo_runl_ctor(struct nvkm_fifo *); +void gk104_fifo_init(struct nvkm_fifo *); +void gk104_fifo_init_pbdmas(struct nvkm_fifo *, u32); +irqreturn_t gk104_fifo_intr(struct nvkm_inth *); +void gk104_fifo_intr_runlist(struct nvkm_fifo *); +void gk104_fifo_intr_chsw(struct nvkm_fifo *); +void gk104_fifo_intr_bind(struct nvkm_fifo *); +extern const struct nvkm_fifo_func_mmu_fault gk104_fifo_mmu_fault; +extern const struct nvkm_enum gk104_fifo_mmu_fault_reason[]; +extern const struct nvkm_enum gk104_fifo_mmu_fault_hubclient[]; +extern const struct nvkm_enum gk104_fifo_mmu_fault_gpcclient[]; +void gk104_runl_insert_chan(struct nvkm_chan *, struct nvkm_memory *, u64); +void gk104_runl_commit(struct nvkm_runl *, struct nvkm_memory *, u32, int); +bool gk104_runl_pending(struct nvkm_runl *); +void gk104_runl_block(struct nvkm_runl *, u32); +void gk104_runl_allow(struct nvkm_runl *, u32); +void gk104_runl_fault_clear(struct nvkm_runl *); +extern const struct nvkm_runq_func gk104_runq; +void gk104_runq_init(struct nvkm_runq *); +bool gk104_runq_intr(struct nvkm_runq *, struct nvkm_runl *); +extern const struct nvkm_bitfield gk104_runq_intr_0_names[]; +bool gk104_runq_idle(struct nvkm_runq *); +extern const struct nvkm_engn_func gk104_engn; +bool gk104_engn_chsw(struct nvkm_engn *); +int gk104_engn_cxid(struct nvkm_engn *, bool *cgid); +int gk104_ectx_ctor(struct nvkm_engn *, struct nvkm_vctx *); +extern const struct nvkm_engn_func gk104_engn_ce; +extern const struct nvkm_chan_func_userd gk104_chan_userd; +extern const struct nvkm_chan_func_ramfc gk104_chan_ramfc; +void gk104_chan_bind(struct nvkm_chan *); +void gk104_chan_bind_inst(struct nvkm_chan *); +void gk104_chan_unbind(struct nvkm_chan *); +void gk104_chan_start(struct nvkm_chan *); +void gk104_chan_stop(struct nvkm_chan *); + +int gk110_fifo_chid_ctor(struct nvkm_fifo *, int); +extern const struct nvkm_runl_func gk110_runl; +extern const struct nvkm_cgrp_func gk110_cgrp; +void gk110_runl_insert_cgrp(struct nvkm_cgrp *, struct nvkm_memory *, u64); +extern const struct nvkm_chan_func gk110_chan; +void gk110_chan_preempt(struct nvkm_chan *); + +extern const struct nvkm_runq_func gk208_runq; +void gk208_runq_init(struct nvkm_runq *); + +void gm107_fifo_intr_mmu_fault_unit(struct nvkm_fifo *, int); +extern const struct nvkm_fifo_func_mmu_fault gm107_fifo_mmu_fault; +extern const struct nvkm_runl_func gm107_runl; +extern const struct nvkm_chan_func gm107_chan; + +int gm200_fifo_chid_nr(struct nvkm_fifo *); +int gm200_fifo_runq_nr(struct nvkm_fifo *); + +extern const struct nvkm_enum gv100_fifo_mmu_fault_access[]; +extern const struct nvkm_enum gv100_fifo_mmu_fault_reason[]; +extern const struct nvkm_enum gv100_fifo_mmu_fault_hubclient[]; +extern const struct nvkm_enum gv100_fifo_mmu_fault_gpcclient[]; +void gv100_runl_insert_cgrp(struct nvkm_cgrp *, struct nvkm_memory *, u64); +void gv100_runl_insert_chan(struct nvkm_chan *, struct nvkm_memory *, u64); +void gv100_runl_preempt(struct nvkm_runl *); +extern const struct nvkm_runq_func gv100_runq; +extern const struct nvkm_engn_func gv100_engn; +void gv100_ectx_bind(struct nvkm_engn *, struct nvkm_cctx *, struct nvkm_chan *); +extern const struct nvkm_engn_func gv100_engn_ce; +int gv100_ectx_ce_ctor(struct nvkm_engn *, struct nvkm_vctx *); +void gv100_ectx_ce_bind(struct nvkm_engn *, struct nvkm_cctx *, struct nvkm_chan *); +extern const struct nvkm_chan_func_userd gv100_chan_userd; +extern const struct nvkm_chan_func_ramfc gv100_chan_ramfc; + +void tu102_fifo_intr_ctxsw_timeout_info(struct nvkm_engn *, u32 info); +extern const struct nvkm_fifo_func_mmu_fault tu102_fifo_mmu_fault; + +int ga100_fifo_runl_ctor(struct nvkm_fifo *); +int ga100_fifo_nonstall_ctor(struct nvkm_fifo *); +extern const struct nvkm_event_func ga100_fifo_nonstall; +extern const struct nvkm_runl_func ga100_runl; +extern const struct nvkm_runq_func ga100_runq; +extern const struct nvkm_engn_func ga100_engn; +extern const struct nvkm_engn_func ga100_engn_ce; +extern const struct nvkm_cgrp_func ga100_cgrp; +extern const struct nvkm_chan_func ga100_chan; + +int nvkm_uchan_new(struct nvkm_fifo *, struct nvkm_cgrp *, const struct nvkm_oclass *, + void *argv, u32 argc, struct nvkm_object **); +int nvkm_ucgrp_new(struct nvkm_fifo *, const struct nvkm_oclass *, void *argv, u32 argc, + struct nvkm_object **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/regsnv04.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/regsnv04.h new file mode 100644 index 0000000000..4445a12b9a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/regsnv04.h @@ -0,0 +1,133 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV04_FIFO_REGS_H__ +#define __NV04_FIFO_REGS_H__ + +#define NV04_PFIFO_DELAY_0 0x00002040 +#define NV04_PFIFO_DMA_TIMESLICE 0x00002044 +#define NV04_PFIFO_NEXT_CHANNEL 0x00002050 +#define NV03_PFIFO_INTR_0 0x00002100 +#define NV03_PFIFO_INTR_EN_0 0x00002140 +# define NV_PFIFO_INTR_CACHE_ERROR (1<<0) +# define NV_PFIFO_INTR_RUNOUT (1<<4) +# define NV_PFIFO_INTR_RUNOUT_OVERFLOW (1<<8) +# define NV_PFIFO_INTR_DMA_PUSHER (1<<12) +# define NV_PFIFO_INTR_DMA_PT (1<<16) +# define NV_PFIFO_INTR_SEMAPHORE (1<<20) +# define NV_PFIFO_INTR_ACQUIRE_TIMEOUT (1<<24) +#define NV03_PFIFO_RAMHT 0x00002210 +#define NV03_PFIFO_RAMFC 0x00002214 +#define NV03_PFIFO_RAMRO 0x00002218 +#define NV40_PFIFO_RAMFC 0x00002220 +#define NV03_PFIFO_CACHES 0x00002500 +#define NV04_PFIFO_MODE 0x00002504 +#define NV04_PFIFO_DMA 0x00002508 +#define NV04_PFIFO_SIZE 0x0000250c +#define NV50_PFIFO_CTX_TABLE(c) (0x2600+(c)*4) +#define NV50_PFIFO_CTX_TABLE__SIZE 128 +#define NV50_PFIFO_CTX_TABLE_CHANNEL_ENABLED (1<<31) +#define NV50_PFIFO_CTX_TABLE_UNK30_BAD (1<<30) +#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G80 0x0FFFFFFF +#define NV50_PFIFO_CTX_TABLE_INSTANCE_MASK_G84 0x00FFFFFF +#define NV03_PFIFO_CACHE0_PUSH0 0x00003000 +#define NV03_PFIFO_CACHE0_PULL0 0x00003040 +#define NV04_PFIFO_CACHE0_PULL0 0x00003050 +#define NV04_PFIFO_CACHE0_PULL1 0x00003054 +#define NV03_PFIFO_CACHE1_PUSH0 0x00003200 +#define NV03_PFIFO_CACHE1_PUSH1 0x00003204 +#define NV03_PFIFO_CACHE1_PUSH1_DMA (1<<8) +#define NV40_PFIFO_CACHE1_PUSH1_DMA (1<<16) +#define NV03_PFIFO_CACHE1_PUSH1_CHID_MASK 0x0000000f +#define NV10_PFIFO_CACHE1_PUSH1_CHID_MASK 0x0000001f +#define NV50_PFIFO_CACHE1_PUSH1_CHID_MASK 0x0000007f +#define NV03_PFIFO_CACHE1_PUT 0x00003210 +#define NV04_PFIFO_CACHE1_DMA_PUSH 0x00003220 +#define NV04_PFIFO_CACHE1_DMA_FETCH 0x00003224 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_8_BYTES 0x00000000 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_16_BYTES 0x00000008 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_24_BYTES 0x00000010 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_32_BYTES 0x00000018 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_40_BYTES 0x00000020 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_48_BYTES 0x00000028 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_56_BYTES 0x00000030 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_64_BYTES 0x00000038 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_72_BYTES 0x00000040 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_80_BYTES 0x00000048 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_88_BYTES 0x00000050 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_96_BYTES 0x00000058 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_104_BYTES 0x00000060 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_112_BYTES 0x00000068 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_120_BYTES 0x00000070 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_128_BYTES 0x00000078 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_136_BYTES 0x00000080 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_144_BYTES 0x00000088 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_152_BYTES 0x00000090 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_160_BYTES 0x00000098 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_168_BYTES 0x000000A0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_176_BYTES 0x000000A8 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_184_BYTES 0x000000B0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_192_BYTES 0x000000B8 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_200_BYTES 0x000000C0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_208_BYTES 0x000000C8 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_216_BYTES 0x000000D0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_224_BYTES 0x000000D8 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_232_BYTES 0x000000E0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_240_BYTES 0x000000E8 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_248_BYTES 0x000000F0 +# define NV_PFIFO_CACHE1_DMA_FETCH_TRIG_256_BYTES 0x000000F8 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE 0x0000E000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_32_BYTES 0x00000000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_64_BYTES 0x00002000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_96_BYTES 0x00004000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_128_BYTES 0x00006000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_160_BYTES 0x00008000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_192_BYTES 0x0000A000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_224_BYTES 0x0000C000 +# define NV_PFIFO_CACHE1_DMA_FETCH_SIZE_256_BYTES 0x0000E000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS 0x001F0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_0 0x00000000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_1 0x00010000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_2 0x00020000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_3 0x00030000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_4 0x00040000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_5 0x00050000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_6 0x00060000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_7 0x00070000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_8 0x00080000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_9 0x00090000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_10 0x000A0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_11 0x000B0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_12 0x000C0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_13 0x000D0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_14 0x000E0000 +# define NV_PFIFO_CACHE1_DMA_FETCH_MAX_REQS_15 0x000F0000 +# define NV_PFIFO_CACHE1_ENDIAN 0x80000000 +# define NV_PFIFO_CACHE1_LITTLE_ENDIAN 0x7FFFFFFF +# define NV_PFIFO_CACHE1_BIG_ENDIAN 0x80000000 +#define NV04_PFIFO_CACHE1_DMA_STATE 0x00003228 +#define NV04_PFIFO_CACHE1_DMA_INSTANCE 0x0000322c +#define NV04_PFIFO_CACHE1_DMA_CTL 0x00003230 +#define NV04_PFIFO_CACHE1_DMA_PUT 0x00003240 +#define NV04_PFIFO_CACHE1_DMA_GET 0x00003244 +#define NV10_PFIFO_CACHE1_REF_CNT 0x00003248 +#define NV10_PFIFO_CACHE1_DMA_SUBROUTINE 0x0000324C +#define NV03_PFIFO_CACHE1_PULL0 0x00003240 +#define NV04_PFIFO_CACHE1_PULL0 0x00003250 +# define NV04_PFIFO_CACHE1_PULL0_HASH_FAILED 0x00000010 +# define NV04_PFIFO_CACHE1_PULL0_HASH_BUSY 0x00001000 +#define NV03_PFIFO_CACHE1_PULL1 0x00003250 +#define NV04_PFIFO_CACHE1_PULL1 0x00003254 +#define NV04_PFIFO_CACHE1_HASH 0x00003258 +#define NV10_PFIFO_CACHE1_ACQUIRE_TIMEOUT 0x00003260 +#define NV10_PFIFO_CACHE1_ACQUIRE_TIMESTAMP 0x00003264 +#define NV10_PFIFO_CACHE1_ACQUIRE_VALUE 0x00003268 +#define NV10_PFIFO_CACHE1_SEMAPHORE 0x0000326C +#define NV03_PFIFO_CACHE1_GET 0x00003270 +#define NV04_PFIFO_CACHE1_ENGINE 0x00003280 +#define NV04_PFIFO_CACHE1_DMA_DCOUNT 0x000032A0 +#define NV40_PFIFO_GRCTX_INSTANCE 0x000032E0 +#define NV40_PFIFO_UNK32E4 0x000032E4 +#define NV04_PFIFO_CACHE1_METHOD(i) (0x00003800+(i*8)) +#define NV04_PFIFO_CACHE1_DATA(i) (0x00003804+(i*8)) +#define NV40_PFIFO_CACHE1_METHOD(i) (0x00090000+(i*8)) +#define NV40_PFIFO_CACHE1_DATA(i) (0x00090004+(i*8)) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.c new file mode 100644 index 0000000000..454a481a0a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.c @@ -0,0 +1,430 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "runl.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "priv.h" +#include "runq.h" + +#include +#include +#include + +static struct nvkm_cgrp * +nvkm_engn_cgrp_get(struct nvkm_engn *engn, unsigned long *pirqflags) +{ + struct nvkm_cgrp *cgrp = NULL; + struct nvkm_chan *chan; + bool cgid; + int id; + + id = engn->func->cxid(engn, &cgid); + if (id < 0) + return NULL; + + if (!cgid) { + chan = nvkm_runl_chan_get_chid(engn->runl, id, pirqflags); + if (chan) + cgrp = chan->cgrp; + } else { + cgrp = nvkm_runl_cgrp_get_cgid(engn->runl, id, pirqflags); + } + + WARN_ON(!cgrp); + return cgrp; +} + +static void +nvkm_runl_rc(struct nvkm_runl *runl) +{ + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_cgrp *cgrp, *gtmp; + struct nvkm_chan *chan, *ctmp; + struct nvkm_engn *engn; + unsigned long flags; + int rc, state, i; + bool reset; + + /* Runlist is blocked before scheduling recovery - fetch count. */ + BUG_ON(!mutex_is_locked(&runl->mutex)); + rc = atomic_xchg(&runl->rc_pending, 0); + if (!rc) + return; + + /* Look for channel groups flagged for RC. */ + nvkm_runl_foreach_cgrp_safe(cgrp, gtmp, runl) { + state = atomic_cmpxchg(&cgrp->rc, NVKM_CGRP_RC_PENDING, NVKM_CGRP_RC_RUNNING); + if (state == NVKM_CGRP_RC_PENDING) { + /* Disable all channels in them, and remove from runlist. */ + nvkm_cgrp_foreach_chan_safe(chan, ctmp, cgrp) { + nvkm_chan_error(chan, false); + nvkm_chan_remove_locked(chan); + } + } + } + + /* On GPUs with runlist preempt, wait for PBDMA(s) servicing runlist to go idle. */ + if (runl->func->preempt) { + for (i = 0; i < runl->runq_nr; i++) { + struct nvkm_runq *runq = runl->runq[i]; + + if (runq) { + nvkm_msec(fifo->engine.subdev.device, 2000, + if (runq->func->idle(runq)) + break; + ); + } + } + } + + /* Look for engines that are still on flagged channel groups - reset them. */ + nvkm_runl_foreach_engn_cond(engn, runl, engn->func->cxid) { + cgrp = nvkm_engn_cgrp_get(engn, &flags); + if (!cgrp) { + ENGN_DEBUG(engn, "cxid not valid"); + continue; + } + + reset = atomic_read(&cgrp->rc) == NVKM_CGRP_RC_RUNNING; + nvkm_cgrp_put(&cgrp, flags); + if (!reset) { + ENGN_DEBUG(engn, "cxid not in recovery"); + continue; + } + + ENGN_DEBUG(engn, "resetting..."); + /*TODO: can we do something less of a potential catastrophe on failure? */ + WARN_ON(nvkm_engine_reset(engn->engine)); + } + + /* Submit runlist update, and clear any remaining exception state. */ + runl->func->update(runl); + if (runl->func->fault_clear) + runl->func->fault_clear(runl); + + /* Unblock runlist processing. */ + while (rc--) + nvkm_runl_allow(runl); + runl->func->wait(runl); +} + +static void +nvkm_runl_rc_runl(struct nvkm_runl *runl) +{ + RUNL_ERROR(runl, "rc scheduled"); + + nvkm_runl_block(runl); + if (runl->func->preempt) + runl->func->preempt(runl); + + atomic_inc(&runl->rc_pending); + schedule_work(&runl->work); +} + +void +nvkm_runl_rc_cgrp(struct nvkm_cgrp *cgrp) +{ + if (atomic_cmpxchg(&cgrp->rc, NVKM_CGRP_RC_NONE, NVKM_CGRP_RC_PENDING) != NVKM_CGRP_RC_NONE) + return; + + CGRP_ERROR(cgrp, "rc scheduled"); + nvkm_runl_rc_runl(cgrp->runl); +} + +void +nvkm_runl_rc_engn(struct nvkm_runl *runl, struct nvkm_engn *engn) +{ + struct nvkm_cgrp *cgrp; + unsigned long flags; + + /* Lookup channel group currently on engine. */ + cgrp = nvkm_engn_cgrp_get(engn, &flags); + if (!cgrp) { + ENGN_DEBUG(engn, "rc skipped, not on channel"); + return; + } + + nvkm_runl_rc_cgrp(cgrp); + nvkm_cgrp_put(&cgrp, flags); +} + +static void +nvkm_runl_work(struct work_struct *work) +{ + struct nvkm_runl *runl = container_of(work, typeof(*runl), work); + + mutex_lock(&runl->mutex); + nvkm_runl_rc(runl); + mutex_unlock(&runl->mutex); + +} + +struct nvkm_chan * +nvkm_runl_chan_get_inst(struct nvkm_runl *runl, u64 inst, unsigned long *pirqflags) +{ + struct nvkm_chid *chid = runl->chid; + struct nvkm_chan *chan; + unsigned long flags; + int id; + + spin_lock_irqsave(&chid->lock, flags); + for_each_set_bit(id, chid->used, chid->nr) { + chan = chid->data[id]; + if (likely(chan)) { + if (chan->inst->addr == inst) { + spin_lock(&chan->cgrp->lock); + *pirqflags = flags; + spin_unlock(&chid->lock); + return chan; + } + } + } + spin_unlock_irqrestore(&chid->lock, flags); + return NULL; +} + +struct nvkm_chan * +nvkm_runl_chan_get_chid(struct nvkm_runl *runl, int id, unsigned long *pirqflags) +{ + struct nvkm_chid *chid = runl->chid; + struct nvkm_chan *chan; + unsigned long flags; + + spin_lock_irqsave(&chid->lock, flags); + if (!WARN_ON(id >= chid->nr)) { + chan = chid->data[id]; + if (likely(chan)) { + spin_lock(&chan->cgrp->lock); + *pirqflags = flags; + spin_unlock(&chid->lock); + return chan; + } + } + spin_unlock_irqrestore(&chid->lock, flags); + return NULL; +} + +struct nvkm_cgrp * +nvkm_runl_cgrp_get_cgid(struct nvkm_runl *runl, int id, unsigned long *pirqflags) +{ + struct nvkm_chid *cgid = runl->cgid; + struct nvkm_cgrp *cgrp; + unsigned long flags; + + spin_lock_irqsave(&cgid->lock, flags); + if (!WARN_ON(id >= cgid->nr)) { + cgrp = cgid->data[id]; + if (likely(cgrp)) { + spin_lock(&cgrp->lock); + *pirqflags = flags; + spin_unlock(&cgid->lock); + return cgrp; + } + } + spin_unlock_irqrestore(&cgid->lock, flags); + return NULL; +} + +int +nvkm_runl_preempt_wait(struct nvkm_runl *runl) +{ + return nvkm_msec(runl->fifo->engine.subdev.device, runl->fifo->timeout.chan_msec, + if (!runl->func->preempt_pending(runl)) + break; + + nvkm_runl_rc(runl); + usleep_range(1, 2); + ) < 0 ? -ETIMEDOUT : 0; +} + +bool +nvkm_runl_update_pending(struct nvkm_runl *runl) +{ + if (!runl->func->pending(runl)) + return false; + + nvkm_runl_rc(runl); + return true; +} + +void +nvkm_runl_update_locked(struct nvkm_runl *runl, bool wait) +{ + if (atomic_xchg(&runl->changed, 0) && runl->func->update) { + runl->func->update(runl); + if (wait) + runl->func->wait(runl); + } +} + +void +nvkm_runl_allow(struct nvkm_runl *runl) +{ + struct nvkm_fifo *fifo = runl->fifo; + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + if (!--runl->blocked) { + RUNL_TRACE(runl, "running"); + runl->func->allow(runl, ~0); + } + spin_unlock_irqrestore(&fifo->lock, flags); +} + +void +nvkm_runl_block(struct nvkm_runl *runl) +{ + struct nvkm_fifo *fifo = runl->fifo; + unsigned long flags; + + spin_lock_irqsave(&fifo->lock, flags); + if (!runl->blocked++) { + RUNL_TRACE(runl, "stopped"); + runl->func->block(runl, ~0); + } + spin_unlock_irqrestore(&fifo->lock, flags); +} + +void +nvkm_runl_fini(struct nvkm_runl *runl) +{ + if (runl->func->fini) + runl->func->fini(runl); + + flush_work(&runl->work); +} + +void +nvkm_runl_del(struct nvkm_runl *runl) +{ + struct nvkm_engn *engn, *engt; + + nvkm_memory_unref(&runl->mem); + + list_for_each_entry_safe(engn, engt, &runl->engns, head) { + list_del(&engn->head); + kfree(engn); + } + + nvkm_chid_unref(&runl->chid); + nvkm_chid_unref(&runl->cgid); + + list_del(&runl->head); + mutex_destroy(&runl->mutex); + kfree(runl); +} + +struct nvkm_engn * +nvkm_runl_add(struct nvkm_runl *runl, int engi, const struct nvkm_engn_func *func, + enum nvkm_subdev_type type, int inst) +{ + struct nvkm_fifo *fifo = runl->fifo; + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_engine *engine; + struct nvkm_engn *engn; + + engine = nvkm_device_engine(device, type, inst); + if (!engine) { + RUNL_DEBUG(runl, "engn %d.%d[%s] not found", engi, inst, nvkm_subdev_type[type]); + return NULL; + } + + if (!(engn = kzalloc(sizeof(*engn), GFP_KERNEL))) + return NULL; + + engn->func = func; + engn->runl = runl; + engn->id = engi; + engn->engine = engine; + engn->fault = -1; + list_add_tail(&engn->head, &runl->engns); + + /* Lookup MMU engine ID for fault handling. */ + if (device->top) + engn->fault = nvkm_top_fault_id(device, engine->subdev.type, engine->subdev.inst); + + if (engn->fault < 0 && fifo->func->mmu_fault) { + const struct nvkm_enum *map = fifo->func->mmu_fault->engine; + + while (map->name) { + if (map->data2 == engine->subdev.type && map->inst == engine->subdev.inst) { + engn->fault = map->value; + break; + } + map++; + } + } + + return engn; +} + +struct nvkm_runl * +nvkm_runl_get(struct nvkm_fifo *fifo, int runi, u32 addr) +{ + struct nvkm_runl *runl; + + nvkm_runl_foreach(runl, fifo) { + if ((runi >= 0 && runl->id == runi) || (runi < 0 && runl->addr == addr)) + return runl; + } + + return NULL; +} + +struct nvkm_runl * +nvkm_runl_new(struct nvkm_fifo *fifo, int runi, u32 addr, int id_nr) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_runl *runl; + int ret; + + if (!(runl = kzalloc(sizeof(*runl), GFP_KERNEL))) + return ERR_PTR(-ENOMEM); + + runl->func = fifo->func->runl; + runl->fifo = fifo; + runl->id = runi; + runl->addr = addr; + INIT_LIST_HEAD(&runl->engns); + INIT_LIST_HEAD(&runl->cgrps); + atomic_set(&runl->changed, 0); + mutex_init(&runl->mutex); + INIT_WORK(&runl->work, nvkm_runl_work); + atomic_set(&runl->rc_triggered, 0); + atomic_set(&runl->rc_pending, 0); + list_add_tail(&runl->head, &fifo->runls); + + if (!fifo->chid) { + if ((ret = nvkm_chid_new(&nvkm_chan_event, subdev, id_nr, 0, id_nr, &runl->cgid)) || + (ret = nvkm_chid_new(&nvkm_chan_event, subdev, id_nr, 0, id_nr, &runl->chid))) { + RUNL_ERROR(runl, "cgid/chid: %d", ret); + nvkm_runl_del(runl); + return ERR_PTR(ret); + } + } else { + runl->cgid = nvkm_chid_ref(fifo->cgid); + runl->chid = nvkm_chid_ref(fifo->chid); + } + + return runl; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.h new file mode 100644 index 0000000000..5421321f8e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runl.h @@ -0,0 +1,131 @@ +#ifndef __NVKM_RUNL_H__ +#define __NVKM_RUNL_H__ +#include +struct nvkm_cctx; +struct nvkm_cgrp; +struct nvkm_chan; +struct nvkm_memory; +struct nvkm_object; +struct nvkm_vctx; +enum nvkm_subdev_type; + +struct nvkm_engn { + const struct nvkm_engn_func { + int (*nonstall)(struct nvkm_engn *); + bool (*chsw)(struct nvkm_engn *); + int (*cxid)(struct nvkm_engn *, bool *cgid); + void (*mmu_fault_trigger)(struct nvkm_engn *); + bool (*mmu_fault_triggered)(struct nvkm_engn *); + int (*ctor)(struct nvkm_engn *, struct nvkm_vctx *); + void (*bind)(struct nvkm_engn *, struct nvkm_cctx *, struct nvkm_chan *); + int (*ramht_add)(struct nvkm_engn *, struct nvkm_object *, struct nvkm_chan *); + void (*ramht_del)(struct nvkm_chan *, int hash); + } *func; + struct nvkm_runl *runl; + int id; + + struct nvkm_engine *engine; + + int fault; + + struct list_head head; +}; + +#define ENGN_PRINT(e,l,p,f,a...) \ + RUNL_PRINT((e)->runl, l, p, "%02d[%8s]:"f, (e)->id, (e)->engine->subdev.name, ##a) +#define ENGN_DEBUG(e,f,a...) ENGN_PRINT((e), DEBUG, info, " "f"\n", ##a) + +struct nvkm_runl { + const struct nvkm_runl_func { + void (*init)(struct nvkm_runl *); + void (*fini)(struct nvkm_runl *); + int runqs; + u8 size; + int (*update)(struct nvkm_runl *); + void (*insert_cgrp)(struct nvkm_cgrp *, struct nvkm_memory *, u64 offset); + void (*insert_chan)(struct nvkm_chan *, struct nvkm_memory *, u64 offset); + void (*commit)(struct nvkm_runl *, struct nvkm_memory *, u32 start, int count); + int (*wait)(struct nvkm_runl *); + bool (*pending)(struct nvkm_runl *); + void (*block)(struct nvkm_runl *, u32 engm); + void (*allow)(struct nvkm_runl *, u32 engm); + void (*fault_clear)(struct nvkm_runl *); + void (*preempt)(struct nvkm_runl *); + bool (*preempt_pending)(struct nvkm_runl *); + } *func; + struct nvkm_fifo *fifo; + int id; + u32 addr; + u32 chan; + u16 doorbell; + + struct nvkm_chid *cgid; +#define NVKM_CHAN_EVENT_ERRORED BIT(0) + struct nvkm_chid *chid; + + struct list_head engns; + + struct nvkm_runq *runq[2]; + int runq_nr; + + struct nvkm_inth inth; + + struct { + int vector; + struct nvkm_inth inth; + } nonstall; + + struct list_head cgrps; + int cgrp_nr; + int chan_nr; + atomic_t changed; + struct nvkm_memory *mem; + u32 offset; + struct mutex mutex; + + int blocked; + + struct work_struct work; + atomic_t rc_triggered; + atomic_t rc_pending; + + struct list_head head; +}; + +struct nvkm_runl *nvkm_runl_new(struct nvkm_fifo *, int runi, u32 addr, int id_nr); +struct nvkm_runl *nvkm_runl_get(struct nvkm_fifo *, int runi, u32 addr); +struct nvkm_engn *nvkm_runl_add(struct nvkm_runl *, int engi, const struct nvkm_engn_func *, + enum nvkm_subdev_type, int inst); +void nvkm_runl_del(struct nvkm_runl *); +void nvkm_runl_fini(struct nvkm_runl *); +void nvkm_runl_block(struct nvkm_runl *); +void nvkm_runl_allow(struct nvkm_runl *); +void nvkm_runl_update_locked(struct nvkm_runl *, bool wait); +bool nvkm_runl_update_pending(struct nvkm_runl *); +int nvkm_runl_preempt_wait(struct nvkm_runl *); + +void nvkm_runl_rc_engn(struct nvkm_runl *, struct nvkm_engn *); +void nvkm_runl_rc_cgrp(struct nvkm_cgrp *); + +struct nvkm_cgrp *nvkm_runl_cgrp_get_cgid(struct nvkm_runl *, int cgid, unsigned long *irqflags); +struct nvkm_chan *nvkm_runl_chan_get_chid(struct nvkm_runl *, int chid, unsigned long *irqflags); +struct nvkm_chan *nvkm_runl_chan_get_inst(struct nvkm_runl *, u64 inst, unsigned long *irqflags); + +#define nvkm_runl_find_engn(engn,runl,cond) nvkm_list_find(engn, &(runl)->engns, head, (cond)) + +#define nvkm_runl_first(fifo) list_first_entry(&(fifo)->runls, struct nvkm_runl, head) +#define nvkm_runl_foreach(runl,fifo) list_for_each_entry((runl), &(fifo)->runls, head) +#define nvkm_runl_foreach_cond(runl,fifo,cond) nvkm_list_foreach(runl, &(fifo)->runls, head, (cond)) +#define nvkm_runl_foreach_engn(engn,runl) list_for_each_entry((engn), &(runl)->engns, head) +#define nvkm_runl_foreach_engn_cond(engn,runl,cond) \ + nvkm_list_foreach(engn, &(runl)->engns, head, (cond)) +#define nvkm_runl_foreach_cgrp(cgrp,runl) list_for_each_entry((cgrp), &(runl)->cgrps, head) +#define nvkm_runl_foreach_cgrp_safe(cgrp,gtmp,runl) \ + list_for_each_entry_safe((cgrp), (gtmp), &(runl)->cgrps, head) + +#define RUNL_PRINT(r,l,p,f,a...) \ + nvkm_printk__(&(r)->fifo->engine.subdev, NV_DBG_##l, p, "%06x:"f, (r)->addr, ##a) +#define RUNL_ERROR(r,f,a...) RUNL_PRINT((r), ERROR, err, " "f"\n", ##a) +#define RUNL_DEBUG(r,f,a...) RUNL_PRINT((r), DEBUG, info, " "f"\n", ##a) +#define RUNL_TRACE(r,f,a...) RUNL_PRINT((r), TRACE, info, " "f"\n", ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.c new file mode 100644 index 0000000000..33bcf5fb3e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.c @@ -0,0 +1,45 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "runq.h" +#include "priv.h" + +void +nvkm_runq_del(struct nvkm_runq *runq) +{ + list_del(&runq->head); + kfree(runq); +} + +struct nvkm_runq * +nvkm_runq_new(struct nvkm_fifo *fifo, int pbid) +{ + struct nvkm_runq *runq; + + if (!(runq = kzalloc(sizeof(*runq), GFP_KERNEL))) + return NULL; + + runq->func = fifo->func->runq; + runq->fifo = fifo; + runq->id = pbid; + list_add_tail(&runq->head, &fifo->runqs); + return runq; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.h b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.h new file mode 100644 index 0000000000..2cb4836e8b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/runq.h @@ -0,0 +1,31 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_RUNQ_H__ +#define __NVKM_RUNQ_H__ +#include +struct nvkm_runl; + +struct nvkm_runq { + const struct nvkm_runq_func { + void (*init)(struct nvkm_runq *); + bool (*intr)(struct nvkm_runq *, struct nvkm_runl *); + const struct nvkm_bitfield *intr_0_names; + bool (*intr_1_ctxnotvalid)(struct nvkm_runq *, int chid); + bool (*idle)(struct nvkm_runq *); + } *func; + struct nvkm_fifo *fifo; + int id; + + struct list_head head; +}; + +struct nvkm_runq *nvkm_runq_new(struct nvkm_fifo *, int pbid); +void nvkm_runq_del(struct nvkm_runq *); + +#define nvkm_runq_foreach(runq,fifo) list_for_each_entry((runq), &(fifo)->runqs, head) +#define nvkm_runq_foreach_cond(runq,fifo,cond) nvkm_list_foreach(runq, &(fifo)->runqs, head, (cond)) + +#define RUNQ_PRINT(r,l,p,f,a...) \ + nvkm_printk__(&(r)->fifo->engine.subdev, NV_DBG_##l, p, "PBDMA%d:"f, (r)->id, ##a) +#define RUNQ_ERROR(r,f,a...) RUNQ_PRINT((r), ERROR, err, " "f"\n", ##a) +#define RUNQ_DEBUG(r,f,a...) RUNQ_PRINT((r), DEBUG, info, " "f"\n", ##a) +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/tu102.c new file mode 100644 index 0000000000..ea9e151dbb --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/tu102.c @@ -0,0 +1,286 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "runl.h" + +#include +#include +#include + +#include + +static u32 +tu102_chan_doorbell_handle(struct nvkm_chan *chan) +{ + return (chan->cgrp->runl->id << 16) | chan->id; +} + +static void +tu102_chan_start(struct nvkm_chan *chan) +{ + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + gk104_chan_start(chan); + nvkm_wr32(device, device->vfn->addr.user + 0x0090, chan->func->doorbell_handle(chan)); +} + +static const struct nvkm_chan_func +tu102_chan = { + .inst = &gf100_chan_inst, + .userd = &gv100_chan_userd, + .ramfc = &gv100_chan_ramfc, + .bind = gk104_chan_bind_inst, + .unbind = gk104_chan_unbind, + .start = tu102_chan_start, + .stop = gk104_chan_stop, + .preempt = gk110_chan_preempt, + .doorbell_handle = tu102_chan_doorbell_handle, +}; + +static bool +tu102_runl_pending(struct nvkm_runl *runl) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + + return nvkm_rd32(device, 0x002b0c + (runl->id * 0x10)) & 0x00008000; +} + +static void +tu102_runl_commit(struct nvkm_runl *runl, struct nvkm_memory *memory, u32 start, int count) +{ + struct nvkm_device *device = runl->fifo->engine.subdev.device; + u64 addr = nvkm_memory_addr(memory) + start; + /*XXX: target? */ + + nvkm_wr32(device, 0x002b00 + (runl->id * 0x10), lower_32_bits(addr)); + nvkm_wr32(device, 0x002b04 + (runl->id * 0x10), upper_32_bits(addr)); + nvkm_wr32(device, 0x002b08 + (runl->id * 0x10), count); +} + +static const struct nvkm_runl_func +tu102_runl = { + .runqs = 2, + .size = 16, + .update = nv50_runl_update, + .insert_cgrp = gv100_runl_insert_cgrp, + .insert_chan = gv100_runl_insert_chan, + .commit = tu102_runl_commit, + .wait = nv50_runl_wait, + .pending = tu102_runl_pending, + .block = gk104_runl_block, + .allow = gk104_runl_allow, + .preempt = gv100_runl_preempt, + .preempt_pending = gf100_runl_preempt_pending, +}; + +static const struct nvkm_enum +tu102_fifo_mmu_fault_engine[] = { + { 0x01, "DISPLAY" }, + { 0x03, "PTP" }, + { 0x06, "PWR_PMU" }, + { 0x08, "IFB", NULL, NVKM_ENGINE_IFB }, + { 0x09, "PERF" }, + { 0x1f, "PHYSICAL" }, + { 0x20, "HOST0" }, + { 0x21, "HOST1" }, + { 0x22, "HOST2" }, + { 0x23, "HOST3" }, + { 0x24, "HOST4" }, + { 0x25, "HOST5" }, + { 0x26, "HOST6" }, + { 0x27, "HOST7" }, + { 0x28, "HOST8" }, + { 0x29, "HOST9" }, + { 0x2a, "HOST10" }, + { 0x2b, "HOST11" }, + { 0x2c, "HOST12" }, + { 0x2d, "HOST13" }, + { 0x2e, "HOST14" }, + { 0x80, "BAR1", NULL, NVKM_SUBDEV_BAR }, + { 0xc0, "BAR2", NULL, NVKM_SUBDEV_INSTMEM }, + {} +}; + +const struct nvkm_fifo_func_mmu_fault +tu102_fifo_mmu_fault = { + .recover = gf100_fifo_mmu_fault_recover, + .access = gv100_fifo_mmu_fault_access, + .engine = tu102_fifo_mmu_fault_engine, + .reason = gv100_fifo_mmu_fault_reason, + .hubclient = gv100_fifo_mmu_fault_hubclient, + .gpcclient = gv100_fifo_mmu_fault_gpcclient, +}; + +void +tu102_fifo_intr_ctxsw_timeout_info(struct nvkm_engn *engn, u32 info) +{ + struct nvkm_runl *runl = engn->runl; + struct nvkm_cgrp *cgrp; + unsigned long flags; + + /* Check that engine hasn't become unstuck since timeout raised. */ + ENGN_DEBUG(engn, "CTXSW_TIMEOUT %08x", info); + if (info & 0xc0000000) + return; + + /* Determine channel group the engine is stuck on, and schedule recovery. */ + switch (info & 0x0000c000) { + case 0x00004000: /* LOAD */ + cgrp = nvkm_runl_cgrp_get_cgid(runl, info & 0x3fff0000, &flags); + break; + case 0x00008000: /* SAVE */ + case 0x0000c000: /* SWITCH */ + cgrp = nvkm_runl_cgrp_get_cgid(runl, info & 0x00003fff, &flags); + break; + default: + cgrp = NULL; + break; + } + + if (!WARN_ON(!cgrp)) { + nvkm_runl_rc_cgrp(cgrp); + nvkm_cgrp_put(&cgrp, flags); + } +} + +static void +tu102_fifo_intr_ctxsw_timeout(struct nvkm_fifo *fifo) +{ + struct nvkm_device *device = fifo->engine.subdev.device; + struct nvkm_runl *runl; + struct nvkm_engn *engn; + u32 engm = nvkm_rd32(device, 0x002a30); + u32 info; + + nvkm_runl_foreach(runl, fifo) { + nvkm_runl_foreach_engn_cond(engn, runl, engm & BIT(engn->id)) { + info = nvkm_rd32(device, 0x003200 + (engn->id * 4)); + tu102_fifo_intr_ctxsw_timeout_info(engn, info); + } + } + + nvkm_wr32(device, 0x002a30, engm); +} + +static void +tu102_fifo_intr_sched(struct nvkm_fifo *fifo) +{ + struct nvkm_subdev *subdev = &fifo->engine.subdev; + u32 intr = nvkm_rd32(subdev->device, 0x00254c); + u32 code = intr & 0x000000ff; + + nvkm_error(subdev, "SCHED_ERROR %02x\n", code); +} + +static irqreturn_t +tu102_fifo_intr(struct nvkm_inth *inth) +{ + struct nvkm_fifo *fifo = container_of(inth, typeof(*fifo), engine.subdev.inth); + struct nvkm_subdev *subdev = &fifo->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = nvkm_rd32(device, 0x002140); + u32 stat = nvkm_rd32(device, 0x002100) & mask; + + if (stat & 0x00000001) { + gk104_fifo_intr_bind(fifo); + nvkm_wr32(device, 0x002100, 0x00000001); + stat &= ~0x00000001; + } + + if (stat & 0x00000002) { + tu102_fifo_intr_ctxsw_timeout(fifo); + stat &= ~0x00000002; + } + + if (stat & 0x00000100) { + tu102_fifo_intr_sched(fifo); + nvkm_wr32(device, 0x002100, 0x00000100); + stat &= ~0x00000100; + } + + if (stat & 0x00010000) { + gk104_fifo_intr_chsw(fifo); + nvkm_wr32(device, 0x002100, 0x00010000); + stat &= ~0x00010000; + } + + if (stat & 0x20000000) { + if (gf100_fifo_intr_pbdma(fifo)) + stat &= ~0x20000000; + } + + if (stat & 0x40000000) { + gk104_fifo_intr_runlist(fifo); + stat &= ~0x40000000; + } + + if (stat & 0x80000000) { + nvkm_wr32(device, 0x002100, 0x80000000); + nvkm_event_ntfy(&fifo->nonstall.event, 0, NVKM_FIFO_NONSTALL_EVENT); + stat &= ~0x80000000; + } + + if (stat) { + nvkm_error(subdev, "INTR %08x\n", stat); + spin_lock(&fifo->lock); + nvkm_mask(device, 0x002140, stat, 0x00000000); + spin_unlock(&fifo->lock); + nvkm_wr32(device, 0x002100, stat); + } + + return IRQ_HANDLED; +} + +static void +tu102_fifo_init_pbdmas(struct nvkm_fifo *fifo, u32 mask) +{ + /* Not directly related to PBDMAs, but, enables doorbell to function. */ + nvkm_mask(fifo->engine.subdev.device, 0xb65000, 0x80000000, 0x80000000); +} + +static const struct nvkm_fifo_func +tu102_fifo = { + .chid_nr = gm200_fifo_chid_nr, + .chid_ctor = gk110_fifo_chid_ctor, + .runq_nr = gm200_fifo_runq_nr, + .runl_ctor = gk104_fifo_runl_ctor, + .init = gk104_fifo_init, + .init_pbdmas = tu102_fifo_init_pbdmas, + .intr = tu102_fifo_intr, + .mmu_fault = &tu102_fifo_mmu_fault, + .nonstall = &gf100_fifo_nonstall, + .runl = &tu102_runl, + .runq = &gv100_runq, + .engn = &gv100_engn, + .engn_ce = &gv100_engn_ce, + .cgrp = {{ 0, 0, KEPLER_CHANNEL_GROUP_A }, &gk110_cgrp, .force = true }, + .chan = {{ 0, 0, TURING_CHANNEL_GPFIFO_A }, &tu102_chan }, +}; + +int +tu102_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_fifo **pfifo) +{ + return nvkm_fifo_new_(&tu102_fifo, device, type, inst, pfifo); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ucgrp.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ucgrp.c new file mode 100644 index 0000000000..52c594dfb1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ucgrp.c @@ -0,0 +1,125 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define nvkm_ucgrp(p) container_of((p), struct nvkm_ucgrp, object) +#include "priv.h" +#include "cgrp.h" +#include "runl.h" + +#include + +#include + +struct nvkm_ucgrp { + struct nvkm_object object; + struct nvkm_cgrp *cgrp; +}; + +static int +nvkm_ucgrp_chan_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_cgrp *cgrp = nvkm_ucgrp(oclass->parent)->cgrp; + + return nvkm_uchan_new(cgrp->runl->fifo, cgrp, oclass, argv, argc, pobject); +} + +static int +nvkm_ucgrp_sclass(struct nvkm_object *object, int index, struct nvkm_oclass *oclass) +{ + struct nvkm_cgrp *cgrp = nvkm_ucgrp(object)->cgrp; + struct nvkm_fifo *fifo = cgrp->runl->fifo; + const struct nvkm_fifo_func_chan *chan = &fifo->func->chan; + int c = 0; + + /* *_CHANNEL_GPFIFO_* */ + if (chan->user.oclass) { + if (c++ == index) { + oclass->base = chan->user; + oclass->ctor = nvkm_ucgrp_chan_new; + return 0; + } + } + + return -EINVAL; +} + +static void * +nvkm_ucgrp_dtor(struct nvkm_object *object) +{ + struct nvkm_ucgrp *ucgrp = nvkm_ucgrp(object); + + nvkm_cgrp_unref(&ucgrp->cgrp); + return ucgrp; +} + +static const struct nvkm_object_func +nvkm_ucgrp = { + .dtor = nvkm_ucgrp_dtor, + .sclass = nvkm_ucgrp_sclass, +}; + +int +nvkm_ucgrp_new(struct nvkm_fifo *fifo, const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + union nvif_cgrp_args *args = argv; + struct nvkm_runl *runl; + struct nvkm_vmm *vmm; + struct nvkm_ucgrp *ucgrp; + int ret; + + if (argc < sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + argc -= sizeof(args->v0); + + if (args->v0.namelen != argc) + return -EINVAL; + + /* Lookup objects referenced in args. */ + runl = nvkm_runl_get(fifo, args->v0.runlist, 0); + if (!runl) + return -EINVAL; + + vmm = nvkm_uvmm_search(oclass->client, args->v0.vmm); + if (IS_ERR(vmm)) + return PTR_ERR(vmm); + + /* Allocate channel group. */ + if (!(ucgrp = kzalloc(sizeof(*ucgrp), GFP_KERNEL))) { + ret = -ENOMEM; + goto done; + } + + nvkm_object_ctor(&nvkm_ucgrp, oclass, &ucgrp->object); + *pobject = &ucgrp->object; + + ret = nvkm_cgrp_new(runl, args->v0.name, vmm, true, &ucgrp->cgrp); + if (ret) + goto done; + + /* Return channel group info to caller. */ + args->v0.cgid = ucgrp->cgrp->id; + +done: + nvkm_vmm_unref(&vmm); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/uchan.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/uchan.c new file mode 100644 index 0000000000..04140e0110 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/uchan.c @@ -0,0 +1,409 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#define nvkm_uchan(p) container_of((p), struct nvkm_uchan, object) +#include "priv.h" +#include "cgrp.h" +#include "chan.h" +#include "chid.h" +#include "runl.h" + +#include +#include +#include +#include + +#include + +struct nvkm_uchan { + struct nvkm_object object; + struct nvkm_chan *chan; +}; + +static int +nvkm_uchan_uevent(struct nvkm_object *object, void *argv, u32 argc, struct nvkm_uevent *uevent) +{ + struct nvkm_chan *chan = nvkm_uchan(object)->chan; + struct nvkm_runl *runl = chan->cgrp->runl; + union nvif_chan_event_args *args = argv; + + if (!uevent) + return 0; + if (argc != sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + + switch (args->v0.type) { + case NVIF_CHAN_EVENT_V0_NON_STALL_INTR: + return nvkm_uevent_add(uevent, &runl->fifo->nonstall.event, runl->id, + NVKM_FIFO_NONSTALL_EVENT, NULL); + case NVIF_CHAN_EVENT_V0_KILLED: + return nvkm_uevent_add(uevent, &runl->chid->event, chan->id, + NVKM_CHAN_EVENT_ERRORED, NULL); + default: + break; + } + + return -ENOSYS; +} + +struct nvkm_uobj { + struct nvkm_oproxy oproxy; + struct nvkm_chan *chan; + struct nvkm_cctx *cctx; + int hash; +}; + +static int +nvkm_uchan_object_fini_1(struct nvkm_oproxy *oproxy, bool suspend) +{ + struct nvkm_uobj *uobj = container_of(oproxy, typeof(*uobj), oproxy); + struct nvkm_chan *chan = uobj->chan; + struct nvkm_cctx *cctx = uobj->cctx; + struct nvkm_ectx *ectx = cctx->vctx->ectx; + + if (!ectx->object) + return 0; + + /* Unbind engine context from channel, if no longer required. */ + if (refcount_dec_and_mutex_lock(&cctx->uses, &chan->cgrp->mutex)) { + nvkm_chan_cctx_bind(chan, ectx->engn, NULL); + + if (refcount_dec_and_test(&ectx->uses)) + nvkm_object_fini(ectx->object, false); + mutex_unlock(&chan->cgrp->mutex); + } + + return 0; +} + +static int +nvkm_uchan_object_init_0(struct nvkm_oproxy *oproxy) +{ + struct nvkm_uobj *uobj = container_of(oproxy, typeof(*uobj), oproxy); + struct nvkm_chan *chan = uobj->chan; + struct nvkm_cctx *cctx = uobj->cctx; + struct nvkm_ectx *ectx = cctx->vctx->ectx; + int ret = 0; + + if (!ectx->object) + return 0; + + /* Bind engine context to channel, if it hasn't been already. */ + if (!refcount_inc_not_zero(&cctx->uses)) { + mutex_lock(&chan->cgrp->mutex); + if (!refcount_inc_not_zero(&cctx->uses)) { + if (!refcount_inc_not_zero(&ectx->uses)) { + ret = nvkm_object_init(ectx->object); + if (ret == 0) + refcount_set(&ectx->uses, 1); + } + + if (ret == 0) { + nvkm_chan_cctx_bind(chan, ectx->engn, cctx); + refcount_set(&cctx->uses, 1); + } + } + mutex_unlock(&chan->cgrp->mutex); + } + + return ret; +} + +static void +nvkm_uchan_object_dtor(struct nvkm_oproxy *oproxy) +{ + struct nvkm_uobj *uobj = container_of(oproxy, typeof(*uobj), oproxy); + struct nvkm_engn *engn; + + if (!uobj->cctx) + return; + + engn = uobj->cctx->vctx->ectx->engn; + if (engn->func->ramht_del) + engn->func->ramht_del(uobj->chan, uobj->hash); + + nvkm_chan_cctx_put(uobj->chan, &uobj->cctx); +} + +static const struct nvkm_oproxy_func +nvkm_uchan_object = { + .dtor[1] = nvkm_uchan_object_dtor, + .init[0] = nvkm_uchan_object_init_0, + .fini[1] = nvkm_uchan_object_fini_1, +}; + +static int +nvkm_uchan_object_new(const struct nvkm_oclass *oclass, void *argv, u32 argc, + struct nvkm_object **pobject) +{ + struct nvkm_chan *chan = nvkm_uchan(oclass->parent)->chan; + struct nvkm_cgrp *cgrp = chan->cgrp; + struct nvkm_engn *engn; + struct nvkm_uobj *uobj; + int ret; + + /* Lookup host engine state for target engine. */ + engn = nvkm_runl_find_engn(engn, cgrp->runl, engn->engine == oclass->engine); + if (WARN_ON(!engn)) + return -EINVAL; + + /* Allocate SW object. */ + if (!(uobj = kzalloc(sizeof(*uobj), GFP_KERNEL))) + return -ENOMEM; + + nvkm_oproxy_ctor(&nvkm_uchan_object, oclass, &uobj->oproxy); + uobj->chan = chan; + *pobject = &uobj->oproxy.base; + + /* Ref. channel context for target engine.*/ + ret = nvkm_chan_cctx_get(chan, engn, &uobj->cctx, oclass->client); + if (ret) + return ret; + + /* Allocate HW object. */ + ret = oclass->base.ctor(&(const struct nvkm_oclass) { + .base = oclass->base, + .engn = oclass->engn, + .handle = oclass->handle, + .object = oclass->object, + .client = oclass->client, + .parent = uobj->cctx->vctx->ectx->object ?: oclass->parent, + .engine = engn->engine, + }, argv, argc, &uobj->oproxy.object); + if (ret) + return ret; + + if (engn->func->ramht_add) { + uobj->hash = engn->func->ramht_add(engn, uobj->oproxy.object, uobj->chan); + if (uobj->hash < 0) + return uobj->hash; + } + + return 0; +} + +static int +nvkm_uchan_sclass(struct nvkm_object *object, int index, struct nvkm_oclass *oclass) +{ + struct nvkm_chan *chan = nvkm_uchan(object)->chan; + struct nvkm_engn *engn; + int ret, runq = 0; + + nvkm_runl_foreach_engn(engn, chan->cgrp->runl) { + struct nvkm_engine *engine = engn->engine; + int c = 0; + + /* Each runqueue, on runlists with multiple, has its own LCE. */ + if (engn->runl->func->runqs) { + if (engine->subdev.type == NVKM_ENGINE_CE) { + if (chan->runq != runq++) + continue; + } + } + + oclass->engine = engine; + oclass->base.oclass = 0; + + if (engine->func->fifo.sclass) { + ret = engine->func->fifo.sclass(oclass, index); + if (oclass->base.oclass) { + if (!oclass->base.ctor) + oclass->base.ctor = nvkm_object_new; + oclass->ctor = nvkm_uchan_object_new; + return 0; + } + + index -= ret; + continue; + } + + while (engine->func->sclass[c].oclass) { + if (c++ == index) { + oclass->base = engine->func->sclass[index]; + if (!oclass->base.ctor) + oclass->base.ctor = nvkm_object_new; + oclass->ctor = nvkm_uchan_object_new; + return 0; + } + } + + index -= c; + } + + return -EINVAL; +} + +static int +nvkm_uchan_map(struct nvkm_object *object, void *argv, u32 argc, + enum nvkm_object_map *type, u64 *addr, u64 *size) +{ + struct nvkm_chan *chan = nvkm_uchan(object)->chan; + struct nvkm_device *device = chan->cgrp->runl->fifo->engine.subdev.device; + + if (chan->func->userd->bar < 0) + return -ENOSYS; + + *type = NVKM_OBJECT_MAP_IO; + *addr = device->func->resource_addr(device, chan->func->userd->bar) + + chan->func->userd->base + chan->userd.base; + *size = chan->func->userd->size; + return 0; +} + +static int +nvkm_uchan_fini(struct nvkm_object *object, bool suspend) +{ + struct nvkm_chan *chan = nvkm_uchan(object)->chan; + + nvkm_chan_block(chan); + nvkm_chan_remove(chan, true); + + if (chan->func->unbind) + chan->func->unbind(chan); + + return 0; +} + +static int +nvkm_uchan_init(struct nvkm_object *object) +{ + struct nvkm_chan *chan = nvkm_uchan(object)->chan; + + if (atomic_read(&chan->errored)) + return 0; + + if (chan->func->bind) + chan->func->bind(chan); + + nvkm_chan_allow(chan); + nvkm_chan_insert(chan); + return 0; +} + +static void * +nvkm_uchan_dtor(struct nvkm_object *object) +{ + struct nvkm_uchan *uchan = nvkm_uchan(object); + + nvkm_chan_del(&uchan->chan); + return uchan; +} + +static const struct nvkm_object_func +nvkm_uchan = { + .dtor = nvkm_uchan_dtor, + .init = nvkm_uchan_init, + .fini = nvkm_uchan_fini, + .map = nvkm_uchan_map, + .sclass = nvkm_uchan_sclass, + .uevent = nvkm_uchan_uevent, +}; + +int +nvkm_uchan_new(struct nvkm_fifo *fifo, struct nvkm_cgrp *cgrp, const struct nvkm_oclass *oclass, + void *argv, u32 argc, struct nvkm_object **pobject) +{ + union nvif_chan_args *args = argv; + struct nvkm_runl *runl; + struct nvkm_vmm *vmm = NULL; + struct nvkm_dmaobj *ctxdma = NULL; + struct nvkm_memory *userd = NULL; + struct nvkm_uchan *uchan; + struct nvkm_chan *chan; + int ret; + + if (argc < sizeof(args->v0) || args->v0.version != 0) + return -ENOSYS; + argc -= sizeof(args->v0); + + if (args->v0.namelen != argc) + return -EINVAL; + + /* Lookup objects referenced in args. */ + runl = nvkm_runl_get(fifo, args->v0.runlist, 0); + if (!runl) + return -EINVAL; + + if (args->v0.vmm) { + vmm = nvkm_uvmm_search(oclass->client, args->v0.vmm); + if (IS_ERR(vmm)) + return PTR_ERR(vmm); + } + + if (args->v0.ctxdma) { + ctxdma = nvkm_dmaobj_search(oclass->client, args->v0.ctxdma); + if (IS_ERR(ctxdma)) { + ret = PTR_ERR(ctxdma); + goto done; + } + } + + if (args->v0.huserd) { + userd = nvkm_umem_search(oclass->client, args->v0.huserd); + if (IS_ERR(userd)) { + ret = PTR_ERR(userd); + userd = NULL; + goto done; + } + } + + /* Allocate channel. */ + if (!(uchan = kzalloc(sizeof(*uchan), GFP_KERNEL))) { + ret = -ENOMEM; + goto done; + } + + nvkm_object_ctor(&nvkm_uchan, oclass, &uchan->object); + *pobject = &uchan->object; + + ret = nvkm_chan_new_(fifo->func->chan.func, runl, args->v0.runq, cgrp, args->v0.name, + args->v0.priv != 0, args->v0.devm, vmm, ctxdma, args->v0.offset, + args->v0.length, userd, args->v0.ouserd, &uchan->chan); + if (ret) + goto done; + + chan = uchan->chan; + + /* Return channel info to caller. */ + if (chan->func->doorbell_handle) + args->v0.token = chan->func->doorbell_handle(chan); + else + args->v0.token = ~0; + + args->v0.chid = chan->id; + + switch (nvkm_memory_target(chan->inst->memory)) { + case NVKM_MEM_TARGET_INST: args->v0.aper = NVIF_CHAN_V0_INST_APER_INST; break; + case NVKM_MEM_TARGET_VRAM: args->v0.aper = NVIF_CHAN_V0_INST_APER_VRAM; break; + case NVKM_MEM_TARGET_HOST: args->v0.aper = NVIF_CHAN_V0_INST_APER_HOST; break; + case NVKM_MEM_TARGET_NCOH: args->v0.aper = NVIF_CHAN_V0_INST_APER_NCOH; break; + default: + WARN_ON(1); + ret = -EFAULT; + break; + } + + args->v0.inst = nvkm_memory_addr(chan->inst->memory); +done: + nvkm_memory_unref(&userd); + nvkm_vmm_unref(&vmm); + return ret; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild new file mode 100644 index 0000000000..b5418f05cc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild @@ -0,0 +1,67 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/gr/base.o +nvkm-y += nvkm/engine/gr/nv04.o +nvkm-y += nvkm/engine/gr/nv10.o +nvkm-y += nvkm/engine/gr/nv15.o +nvkm-y += nvkm/engine/gr/nv17.o +nvkm-y += nvkm/engine/gr/nv20.o +nvkm-y += nvkm/engine/gr/nv25.o +nvkm-y += nvkm/engine/gr/nv2a.o +nvkm-y += nvkm/engine/gr/nv30.o +nvkm-y += nvkm/engine/gr/nv34.o +nvkm-y += nvkm/engine/gr/nv35.o +nvkm-y += nvkm/engine/gr/nv40.o +nvkm-y += nvkm/engine/gr/nv44.o +nvkm-y += nvkm/engine/gr/nv50.o +nvkm-y += nvkm/engine/gr/g84.o +nvkm-y += nvkm/engine/gr/gt200.o +nvkm-y += nvkm/engine/gr/mcp79.o +nvkm-y += nvkm/engine/gr/gt215.o +nvkm-y += nvkm/engine/gr/mcp89.o +nvkm-y += nvkm/engine/gr/gf100.o +nvkm-y += nvkm/engine/gr/gf104.o +nvkm-y += nvkm/engine/gr/gf108.o +nvkm-y += nvkm/engine/gr/gf110.o +nvkm-y += nvkm/engine/gr/gf117.o +nvkm-y += nvkm/engine/gr/gf119.o +nvkm-y += nvkm/engine/gr/gk104.o +nvkm-y += nvkm/engine/gr/gk110.o +nvkm-y += nvkm/engine/gr/gk110b.o +nvkm-y += nvkm/engine/gr/gk208.o +nvkm-y += nvkm/engine/gr/gk20a.o +nvkm-y += nvkm/engine/gr/gm107.o +nvkm-y += nvkm/engine/gr/gm200.o +nvkm-y += nvkm/engine/gr/gm20b.o +nvkm-y += nvkm/engine/gr/gp100.o +nvkm-y += nvkm/engine/gr/gp102.o +nvkm-y += nvkm/engine/gr/gp104.o +nvkm-y += nvkm/engine/gr/gp107.o +nvkm-y += nvkm/engine/gr/gp108.o +nvkm-y += nvkm/engine/gr/gp10b.o +nvkm-y += nvkm/engine/gr/gv100.o +nvkm-y += nvkm/engine/gr/tu102.o +nvkm-y += nvkm/engine/gr/ga102.o + +nvkm-y += nvkm/engine/gr/ctxnv40.o +nvkm-y += nvkm/engine/gr/ctxnv50.o +nvkm-y += nvkm/engine/gr/ctxgf100.o +nvkm-y += nvkm/engine/gr/ctxgf104.o +nvkm-y += nvkm/engine/gr/ctxgf108.o +nvkm-y += nvkm/engine/gr/ctxgf110.o +nvkm-y += nvkm/engine/gr/ctxgf117.o +nvkm-y += nvkm/engine/gr/ctxgf119.o +nvkm-y += nvkm/engine/gr/ctxgk104.o +nvkm-y += nvkm/engine/gr/ctxgk110.o +nvkm-y += nvkm/engine/gr/ctxgk110b.o +nvkm-y += nvkm/engine/gr/ctxgk208.o +nvkm-y += nvkm/engine/gr/ctxgk20a.o +nvkm-y += nvkm/engine/gr/ctxgm107.o +nvkm-y += nvkm/engine/gr/ctxgm200.o +nvkm-y += nvkm/engine/gr/ctxgm20b.o +nvkm-y += nvkm/engine/gr/ctxgp100.o +nvkm-y += nvkm/engine/gr/ctxgp102.o +nvkm-y += nvkm/engine/gr/ctxgp104.o +nvkm-y += nvkm/engine/gr/ctxgp107.o +nvkm-y += nvkm/engine/gr/ctxgv100.o +nvkm-y += nvkm/engine/gr/ctxtu102.o +nvkm-y += nvkm/engine/gr/ctxga102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c new file mode 100644 index 0000000000..0096ad401b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c @@ -0,0 +1,205 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +u32 +nvkm_gr_ctxsw_inst(struct nvkm_device *device) +{ + struct nvkm_gr *gr = device->gr; + if (gr && gr->func->ctxsw.inst) + return gr->func->ctxsw.inst(gr); + return 0; +} + +int +nvkm_gr_ctxsw_resume(struct nvkm_device *device) +{ + struct nvkm_gr *gr = device->gr; + if (gr && gr->func->ctxsw.resume) + return gr->func->ctxsw.resume(gr); + return 0; +} + +int +nvkm_gr_ctxsw_pause(struct nvkm_device *device) +{ + struct nvkm_gr *gr = device->gr; + if (gr && gr->func->ctxsw.pause) + return gr->func->ctxsw.pause(gr); + return 0; +} + +static bool +nvkm_gr_chsw_load(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->chsw_load) + return gr->func->chsw_load(gr); + return false; +} + +static void +nvkm_gr_tile(struct nvkm_engine *engine, int region, struct nvkm_fb_tile *tile) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->tile) + gr->func->tile(gr, region, tile); +} + +u64 +nvkm_gr_units(struct nvkm_gr *gr) +{ + if (gr->func->units) + return gr->func->units(gr); + return 0; +} + +int +nvkm_gr_tlb_flush(struct nvkm_gr *gr) +{ + if (gr->func->tlb_flush) + return gr->func->tlb_flush(gr); + return -ENODEV; +} + +static int +nvkm_gr_oclass_get(struct nvkm_oclass *oclass, int index) +{ + struct nvkm_gr *gr = nvkm_gr(oclass->engine); + int c = 0; + + if (gr->func->object_get) { + int ret = gr->func->object_get(gr, index, &oclass->base); + if (oclass->base.oclass) + return index; + return ret; + } + + while (gr->func->sclass[c].oclass) { + if (c++ == index) { + oclass->base = gr->func->sclass[index]; + return index; + } + } + + return c; +} + +static int +nvkm_gr_cclass_new(struct nvkm_chan *chan, const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nvkm_gr *gr = nvkm_gr(oclass->engine); + if (gr->func->chan_new) + return gr->func->chan_new(gr, chan, oclass, pobject); + return 0; +} + +static void +nvkm_gr_intr(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + gr->func->intr(gr); +} + +static int +nvkm_gr_nonstall(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + + if (gr->func->nonstall) + return gr->func->nonstall(gr); + + return -EINVAL; +} + +static int +nvkm_gr_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->oneinit) + return gr->func->oneinit(gr); + return 0; +} + +static int +nvkm_gr_reset(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + + if (gr->func->reset) + return gr->func->reset(gr); + + return -ENOSYS; +} + +static int +nvkm_gr_init(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + return gr->func->init(gr); +} + +static int +nvkm_gr_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->fini) + return gr->func->fini(gr, suspend); + return 0; +} + +static void * +nvkm_gr_dtor(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->dtor) + return gr->func->dtor(gr); + return gr; +} + +static const struct nvkm_engine_func +nvkm_gr = { + .dtor = nvkm_gr_dtor, + .oneinit = nvkm_gr_oneinit, + .init = nvkm_gr_init, + .fini = nvkm_gr_fini, + .reset = nvkm_gr_reset, + .nonstall = nvkm_gr_nonstall, + .intr = nvkm_gr_intr, + .tile = nvkm_gr_tile, + .chsw_load = nvkm_gr_chsw_load, + .fifo.cclass = nvkm_gr_cclass_new, + .fifo.sclass = nvkm_gr_oclass_get, +}; + +int +nvkm_gr_ctor(const struct nvkm_gr_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, bool enable, struct nvkm_gr *gr) +{ + gr->func = func; + return nvkm_engine_ctor(&nvkm_gr, device, type, inst, enable, &gr->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxga102.c new file mode 100644 index 0000000000..11461adf50 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxga102.c @@ -0,0 +1,77 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +static void +ga102_grctx_generate_sm_id(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + tpc = gv100_gr_nonpes_aware_tpc(gr, gpc, tpc); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x608), sm); +} + +static void +ga102_grctx_generate_unkn(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x41be08, 0x00000004, 0x00000004); +} + +static void +ga102_grctx_generate_r419ea8(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x419ea8, nvkm_rd32(device, 0x504728) | 0x08000000); +} + +const struct gf100_grctx_func +ga102_grctx = { + .main = gf100_grctx_generate_main, + .unkn = ga102_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x1140, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gv100_grctx_generate_attrib_cb, + .attrib = gv100_grctx_generate_attrib, + .attrib_nr_max = 0x800, + .attrib_nr = 0x4a1, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .unknown_size = 0x80000, + .unknown = tu102_grctx_generate_unknown, + .gfxp_nr = 0xd28, + .sm_id = ga102_grctx_generate_sm_id, + .skip_pd_num_tpc_per_gpc = true, + .rop_mapping = gv100_grctx_generate_rop_mapping, + .r406500 = gm200_grctx_generate_r406500, + .r400088 = gv100_grctx_generate_r400088, + .r419ea8 = ga102_grctx_generate_r419ea8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c new file mode 100644 index 0000000000..cb390e0134 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c @@ -0,0 +1,1587 @@ +/* + * Copyright 2010 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include +#include +#include +#include + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf100_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000b2, 40, 0x01, 0x00000000 }, + { 0x000210, 8, 0x01, 0x00000040 }, + { 0x000218, 8, 0x01, 0x0000c080 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00001fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x003fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x0005e0, 5, 0x01, 0x00000022 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x00000c48 }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00300008 }, + { 0x000841, 1, 0x01, 0x04000080 }, + { 0x000842, 1, 0x01, 0x00300008 }, + { 0x000843, 1, 0x01, 0x04000080 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x000944, 1, 0x01, 0x00000022 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000014 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_icmd[] = { + { gf100_grctx_init_icmd_0 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_9097_0[] = { + { 0x000800, 8, 0x40, 0x00000000 }, + { 0x000804, 8, 0x40, 0x00000000 }, + { 0x000808, 8, 0x40, 0x00000400 }, + { 0x00080c, 8, 0x40, 0x00000300 }, + { 0x000810, 1, 0x04, 0x000000cf }, + { 0x000850, 7, 0x40, 0x00000000 }, + { 0x000814, 8, 0x40, 0x00000040 }, + { 0x000818, 8, 0x40, 0x00000001 }, + { 0x00081c, 8, 0x40, 0x00000000 }, + { 0x000820, 8, 0x40, 0x00000000 }, + { 0x002700, 8, 0x20, 0x00000000 }, + { 0x002704, 8, 0x20, 0x00000000 }, + { 0x002708, 8, 0x20, 0x00000000 }, + { 0x00270c, 8, 0x20, 0x00000000 }, + { 0x002710, 8, 0x20, 0x00014000 }, + { 0x002714, 8, 0x20, 0x00000040 }, + { 0x001c00, 16, 0x10, 0x00000000 }, + { 0x001c04, 16, 0x10, 0x00000000 }, + { 0x001c08, 16, 0x10, 0x00000000 }, + { 0x001c0c, 16, 0x10, 0x00000000 }, + { 0x001d00, 16, 0x10, 0x00000000 }, + { 0x001d04, 16, 0x10, 0x00000000 }, + { 0x001d08, 16, 0x10, 0x00000000 }, + { 0x001d0c, 16, 0x10, 0x00000000 }, + { 0x001f00, 16, 0x08, 0x00000000 }, + { 0x001f04, 16, 0x08, 0x00000000 }, + { 0x001f80, 16, 0x08, 0x00000000 }, + { 0x001f84, 16, 0x08, 0x00000000 }, + { 0x002200, 5, 0x10, 0x00000022 }, + { 0x002000, 1, 0x04, 0x00000000 }, + { 0x002040, 1, 0x04, 0x00000011 }, + { 0x002080, 1, 0x04, 0x00000020 }, + { 0x0020c0, 1, 0x04, 0x00000030 }, + { 0x002100, 1, 0x04, 0x00000040 }, + { 0x002140, 1, 0x04, 0x00000051 }, + { 0x00200c, 6, 0x40, 0x00000001 }, + { 0x002010, 1, 0x04, 0x00000000 }, + { 0x002050, 1, 0x04, 0x00000000 }, + { 0x002090, 1, 0x04, 0x00000001 }, + { 0x0020d0, 1, 0x04, 0x00000002 }, + { 0x002110, 1, 0x04, 0x00000003 }, + { 0x002150, 1, 0x04, 0x00000004 }, + { 0x000380, 4, 0x20, 0x00000000 }, + { 0x000384, 4, 0x20, 0x00000000 }, + { 0x000388, 4, 0x20, 0x00000000 }, + { 0x00038c, 4, 0x20, 0x00000000 }, + { 0x000700, 4, 0x10, 0x00000000 }, + { 0x000704, 4, 0x10, 0x00000000 }, + { 0x000708, 4, 0x10, 0x00000000 }, + { 0x002800, 128, 0x04, 0x00000000 }, + { 0x000a00, 16, 0x20, 0x00000000 }, + { 0x000a04, 16, 0x20, 0x00000000 }, + { 0x000a08, 16, 0x20, 0x00000000 }, + { 0x000a0c, 16, 0x20, 0x00000000 }, + { 0x000a10, 16, 0x20, 0x00000000 }, + { 0x000a14, 16, 0x20, 0x00000000 }, + { 0x000c00, 16, 0x10, 0x00000000 }, + { 0x000c04, 16, 0x10, 0x00000000 }, + { 0x000c08, 16, 0x10, 0x00000000 }, + { 0x000c0c, 16, 0x10, 0x3f800000 }, + { 0x000d00, 8, 0x08, 0xffff0000 }, + { 0x000d04, 8, 0x08, 0xffff0000 }, + { 0x000e00, 16, 0x10, 0x00000000 }, + { 0x000e04, 16, 0x10, 0xffff0000 }, + { 0x000e08, 16, 0x10, 0xffff0000 }, + { 0x000d40, 4, 0x08, 0x00000000 }, + { 0x000d44, 4, 0x08, 0x00000000 }, + { 0x001e00, 8, 0x20, 0x00000001 }, + { 0x001e04, 8, 0x20, 0x00000001 }, + { 0x001e08, 8, 0x20, 0x00000002 }, + { 0x001e0c, 8, 0x20, 0x00000001 }, + { 0x001e10, 8, 0x20, 0x00000001 }, + { 0x001e14, 8, 0x20, 0x00000002 }, + { 0x001e18, 8, 0x20, 0x00000001 }, + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x001514, 1, 0x04, 0x00000000 }, + { 0x000d68, 1, 0x04, 0x0000ffff }, + { 0x00121c, 1, 0x04, 0x0fac6881 }, + { 0x000fac, 1, 0x04, 0x00000001 }, + { 0x001538, 1, 0x04, 0x00000001 }, + { 0x000fe0, 2, 0x04, 0x00000000 }, + { 0x000fe8, 1, 0x04, 0x00000014 }, + { 0x000fec, 1, 0x04, 0x00000040 }, + { 0x000ff0, 1, 0x04, 0x00000000 }, + { 0x00179c, 1, 0x04, 0x00000000 }, + { 0x001228, 1, 0x04, 0x00000400 }, + { 0x00122c, 1, 0x04, 0x00000300 }, + { 0x001230, 1, 0x04, 0x00010001 }, + { 0x0007f8, 1, 0x04, 0x00000000 }, + { 0x0015b4, 1, 0x04, 0x00000001 }, + { 0x0015cc, 1, 0x04, 0x00000000 }, + { 0x001534, 1, 0x04, 0x00000000 }, + { 0x000fb0, 1, 0x04, 0x00000000 }, + { 0x0015d0, 1, 0x04, 0x00000000 }, + { 0x00153c, 1, 0x04, 0x00000000 }, + { 0x0016b4, 1, 0x04, 0x00000003 }, + { 0x000fbc, 4, 0x04, 0x0000ffff }, + { 0x000df8, 2, 0x04, 0x00000000 }, + { 0x001948, 1, 0x04, 0x00000000 }, + { 0x001970, 1, 0x04, 0x00000001 }, + { 0x00161c, 1, 0x04, 0x000009f0 }, + { 0x000dcc, 1, 0x04, 0x00000010 }, + { 0x00163c, 1, 0x04, 0x00000000 }, + { 0x0015e4, 1, 0x04, 0x00000000 }, + { 0x001160, 32, 0x04, 0x25e00040 }, + { 0x001880, 32, 0x04, 0x00000000 }, + { 0x000f84, 2, 0x04, 0x00000000 }, + { 0x0017c8, 2, 0x04, 0x00000000 }, + { 0x0017d0, 1, 0x04, 0x000000ff }, + { 0x0017d4, 1, 0x04, 0xffffffff }, + { 0x0017d8, 1, 0x04, 0x00000002 }, + { 0x0017dc, 1, 0x04, 0x00000000 }, + { 0x0015f4, 2, 0x04, 0x00000000 }, + { 0x001434, 2, 0x04, 0x00000000 }, + { 0x000d74, 1, 0x04, 0x00000000 }, + { 0x000dec, 1, 0x04, 0x00000001 }, + { 0x0013a4, 1, 0x04, 0x00000000 }, + { 0x001318, 1, 0x04, 0x00000001 }, + { 0x001644, 1, 0x04, 0x00000000 }, + { 0x000748, 1, 0x04, 0x00000000 }, + { 0x000de8, 1, 0x04, 0x00000000 }, + { 0x001648, 1, 0x04, 0x00000000 }, + { 0x0012a4, 1, 0x04, 0x00000000 }, + { 0x001120, 4, 0x04, 0x00000000 }, + { 0x001118, 1, 0x04, 0x00000000 }, + { 0x00164c, 1, 0x04, 0x00000000 }, + { 0x001658, 1, 0x04, 0x00000000 }, + { 0x001910, 1, 0x04, 0x00000290 }, + { 0x001518, 1, 0x04, 0x00000000 }, + { 0x00165c, 1, 0x04, 0x00000001 }, + { 0x001520, 1, 0x04, 0x00000000 }, + { 0x001604, 1, 0x04, 0x00000000 }, + { 0x001570, 1, 0x04, 0x00000000 }, + { 0x0013b0, 2, 0x04, 0x3f800000 }, + { 0x00020c, 1, 0x04, 0x00000000 }, + { 0x001670, 1, 0x04, 0x30201000 }, + { 0x001674, 1, 0x04, 0x70605040 }, + { 0x001678, 1, 0x04, 0xb8a89888 }, + { 0x00167c, 1, 0x04, 0xf8e8d8c8 }, + { 0x00166c, 1, 0x04, 0x00000000 }, + { 0x001680, 1, 0x04, 0x00ffff00 }, + { 0x0012d0, 1, 0x04, 0x00000003 }, + { 0x0012d4, 1, 0x04, 0x00000002 }, + { 0x001684, 2, 0x04, 0x00000000 }, + { 0x000dac, 2, 0x04, 0x00001b02 }, + { 0x000db4, 1, 0x04, 0x00000000 }, + { 0x00168c, 1, 0x04, 0x00000000 }, + { 0x0015bc, 1, 0x04, 0x00000000 }, + { 0x00156c, 1, 0x04, 0x00000000 }, + { 0x00187c, 1, 0x04, 0x00000000 }, + { 0x001110, 1, 0x04, 0x00000001 }, + { 0x000dc0, 3, 0x04, 0x00000000 }, + { 0x001234, 1, 0x04, 0x00000000 }, + { 0x001690, 1, 0x04, 0x00000000 }, + { 0x0012ac, 1, 0x04, 0x00000001 }, + { 0x0002c4, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x001000, 1, 0x04, 0x00000010 }, + { 0x0010fc, 1, 0x04, 0x00000000 }, + { 0x001290, 1, 0x04, 0x00000000 }, + { 0x000218, 1, 0x04, 0x00000010 }, + { 0x0012d8, 1, 0x04, 0x00000000 }, + { 0x0012dc, 1, 0x04, 0x00000010 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x00155c, 2, 0x04, 0x00000000 }, + { 0x001564, 1, 0x04, 0x00001fff }, + { 0x001574, 2, 0x04, 0x00000000 }, + { 0x00157c, 1, 0x04, 0x003fffff }, + { 0x001354, 1, 0x04, 0x00000000 }, + { 0x001664, 1, 0x04, 0x00000000 }, + { 0x001610, 1, 0x04, 0x00000012 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x00162c, 1, 0x04, 0x00000003 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000324, 6, 0x04, 0x3f800000 }, + { 0x000750, 1, 0x04, 0x00000000 }, + { 0x000760, 1, 0x04, 0x39291909 }, + { 0x000764, 1, 0x04, 0x79695949 }, + { 0x000768, 1, 0x04, 0xb9a99989 }, + { 0x00076c, 1, 0x04, 0xf9e9d9c9 }, + { 0x000770, 1, 0x04, 0x30201000 }, + { 0x000774, 1, 0x04, 0x70605040 }, + { 0x000778, 1, 0x04, 0x00009080 }, + { 0x000780, 1, 0x04, 0x39291909 }, + { 0x000784, 1, 0x04, 0x79695949 }, + { 0x000788, 1, 0x04, 0xb9a99989 }, + { 0x00078c, 1, 0x04, 0xf9e9d9c9 }, + { 0x0007d0, 1, 0x04, 0x30201000 }, + { 0x0007d4, 1, 0x04, 0x70605040 }, + { 0x0007d8, 1, 0x04, 0x00009080 }, + { 0x00037c, 1, 0x04, 0x00000001 }, + { 0x000740, 2, 0x04, 0x00000000 }, + { 0x002600, 1, 0x04, 0x00000000 }, + { 0x001918, 1, 0x04, 0x00000000 }, + { 0x00191c, 1, 0x04, 0x00000900 }, + { 0x001920, 1, 0x04, 0x00000405 }, + { 0x001308, 1, 0x04, 0x00000001 }, + { 0x001924, 1, 0x04, 0x00000000 }, + { 0x0013ac, 1, 0x04, 0x00000000 }, + { 0x00192c, 1, 0x04, 0x00000001 }, + { 0x00193c, 1, 0x04, 0x00002c1c }, + { 0x000d7c, 1, 0x04, 0x00000000 }, + { 0x000f8c, 1, 0x04, 0x00000000 }, + { 0x0002c0, 1, 0x04, 0x00000001 }, + { 0x001510, 1, 0x04, 0x00000000 }, + { 0x001940, 1, 0x04, 0x00000000 }, + { 0x000ff4, 2, 0x04, 0x00000000 }, + { 0x00194c, 2, 0x04, 0x00000000 }, + { 0x001968, 1, 0x04, 0x00000000 }, + { 0x001590, 1, 0x04, 0x0000003f }, + { 0x0007e8, 4, 0x04, 0x00000000 }, + { 0x00196c, 1, 0x04, 0x00000011 }, + { 0x00197c, 1, 0x04, 0x00000000 }, + { 0x000fcc, 2, 0x04, 0x00000000 }, + { 0x0002d8, 1, 0x04, 0x00000040 }, + { 0x001980, 1, 0x04, 0x00000080 }, + { 0x001504, 1, 0x04, 0x00000080 }, + { 0x001984, 1, 0x04, 0x00000000 }, + { 0x000300, 1, 0x04, 0x00000001 }, + { 0x0013a8, 1, 0x04, 0x00000000 }, + { 0x0012ec, 1, 0x04, 0x00000000 }, + { 0x001310, 1, 0x04, 0x00000000 }, + { 0x001314, 1, 0x04, 0x00000001 }, + { 0x001380, 1, 0x04, 0x00000000 }, + { 0x001384, 4, 0x04, 0x00000001 }, + { 0x001394, 1, 0x04, 0x00000000 }, + { 0x00139c, 1, 0x04, 0x00000000 }, + { 0x001398, 1, 0x04, 0x00000000 }, + { 0x001594, 1, 0x04, 0x00000000 }, + { 0x001598, 4, 0x04, 0x00000001 }, + { 0x000f54, 3, 0x04, 0x00000000 }, + { 0x0019bc, 1, 0x04, 0x00000000 }, + { 0x000f9c, 2, 0x04, 0x00000000 }, + { 0x0012cc, 1, 0x04, 0x00000000 }, + { 0x0012e8, 1, 0x04, 0x00000000 }, + { 0x00130c, 1, 0x04, 0x00000001 }, + { 0x001360, 8, 0x04, 0x00000000 }, + { 0x00133c, 2, 0x04, 0x00000001 }, + { 0x001344, 1, 0x04, 0x00000002 }, + { 0x001348, 2, 0x04, 0x00000001 }, + { 0x001350, 1, 0x04, 0x00000002 }, + { 0x001358, 1, 0x04, 0x00000001 }, + { 0x0012e4, 1, 0x04, 0x00000000 }, + { 0x00131c, 4, 0x04, 0x00000000 }, + { 0x0019c0, 1, 0x04, 0x00000000 }, + { 0x001140, 1, 0x04, 0x00000000 }, + { 0x0019c4, 1, 0x04, 0x00000000 }, + { 0x0019c8, 1, 0x04, 0x00001500 }, + { 0x00135c, 1, 0x04, 0x00000000 }, + { 0x000f90, 1, 0x04, 0x00000000 }, + { 0x0019e0, 8, 0x04, 0x00000001 }, + { 0x0019cc, 1, 0x04, 0x00000001 }, + { 0x0015b8, 1, 0x04, 0x00000000 }, + { 0x001a00, 1, 0x04, 0x00001111 }, + { 0x001a04, 7, 0x04, 0x00000000 }, + { 0x000d6c, 2, 0x04, 0xffff0000 }, + { 0x0010f8, 1, 0x04, 0x00001010 }, + { 0x000d80, 5, 0x04, 0x00000000 }, + { 0x000da0, 1, 0x04, 0x00000000 }, + { 0x001508, 1, 0x04, 0x80000000 }, + { 0x00150c, 1, 0x04, 0x40000000 }, + { 0x001668, 1, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000008 }, + { 0x000d9c, 1, 0x04, 0x00000001 }, + { 0x0007dc, 1, 0x04, 0x00000000 }, + { 0x00074c, 1, 0x04, 0x00000055 }, + { 0x001420, 1, 0x04, 0x00000003 }, + { 0x0017bc, 2, 0x04, 0x00000000 }, + { 0x0017c4, 1, 0x04, 0x00000001 }, + { 0x001008, 1, 0x04, 0x00000008 }, + { 0x00100c, 1, 0x04, 0x00000040 }, + { 0x001010, 1, 0x04, 0x0000012c }, + { 0x000d60, 1, 0x04, 0x00000040 }, + { 0x00075c, 1, 0x04, 0x00000003 }, + { 0x001018, 1, 0x04, 0x00000020 }, + { 0x00101c, 1, 0x04, 0x00000001 }, + { 0x001020, 1, 0x04, 0x00000020 }, + { 0x001024, 1, 0x04, 0x00000001 }, + { 0x001444, 3, 0x04, 0x00000000 }, + { 0x000360, 1, 0x04, 0x20164010 }, + { 0x000364, 1, 0x04, 0x00000020 }, + { 0x000368, 1, 0x04, 0x00000000 }, + { 0x000de4, 1, 0x04, 0x00000000 }, + { 0x000204, 1, 0x04, 0x00000006 }, + { 0x000208, 1, 0x04, 0x00000000 }, + { 0x0002cc, 1, 0x04, 0x003fffff }, + { 0x0002d0, 1, 0x04, 0x00000c48 }, + { 0x001220, 1, 0x04, 0x00000005 }, + { 0x000fdc, 1, 0x04, 0x00000000 }, + { 0x000f98, 1, 0x04, 0x00300008 }, + { 0x001284, 1, 0x04, 0x04000080 }, + { 0x001450, 1, 0x04, 0x00300008 }, + { 0x001454, 1, 0x04, 0x04000080 }, + { 0x000214, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_902d_0[] = { + { 0x000200, 1, 0x04, 0x000000cf }, + { 0x000204, 1, 0x04, 0x00000001 }, + { 0x000208, 1, 0x04, 0x00000020 }, + { 0x00020c, 1, 0x04, 0x00000001 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000214, 1, 0x04, 0x00000080 }, + { 0x000218, 2, 0x04, 0x00000100 }, + { 0x000220, 2, 0x04, 0x00000000 }, + { 0x000230, 1, 0x04, 0x000000cf }, + { 0x000234, 1, 0x04, 0x00000001 }, + { 0x000238, 1, 0x04, 0x00000020 }, + { 0x00023c, 1, 0x04, 0x00000001 }, + { 0x000244, 1, 0x04, 0x00000080 }, + { 0x000248, 2, 0x04, 0x00000100 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_9039_0[] = { + { 0x00030c, 3, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000238, 2, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_90c0_0[] = { + { 0x00270c, 8, 0x20, 0x00000000 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x000758, 1, 0x04, 0x00000100 }, + { 0x0002c4, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x000204, 3, 0x04, 0x00000000 }, + { 0x000214, 1, 0x04, 0x00000000 }, + { 0x00024c, 1, 0x04, 0x00000000 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x001664, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_mthd[] = { + { gf100_grctx_init_9097_0, 0x9097 }, + { gf100_grctx_init_902d_0, 0x902d }, + { gf100_grctx_init_9039_0, 0x9039 }, + { gf100_grctx_init_90c0_0, 0x90c0 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_main_0[] = { + { 0x400204, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_fe_0[] = { + { 0x404004, 11, 0x04, 0x00000000 }, + { 0x404044, 1, 0x04, 0x00000000 }, + { 0x404094, 13, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf0000087 }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040e8, 1, 0x04, 0x00001000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404138, 1, 0x04, 0x20000040 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000055 }, + { 0x404168, 1, 0x04, 0x00000000 }, + { 0x404174, 3, 0x04, 0x00000000 }, + { 0x404200, 8, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_pri_0[] = { + { 0x404404, 14, 0x04, 0x00000000 }, + { 0x404460, 2, 0x04, 0x00000000 }, + { 0x404468, 1, 0x04, 0x00ffffff }, + { 0x40446c, 1, 0x04, 0x00000000 }, + { 0x404480, 1, 0x04, 0x00000001 }, + { 0x404498, 1, 0x04, 0x00000001 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_memfmt_0[] = { + { 0x404604, 1, 0x04, 0x00000015 }, + { 0x404608, 1, 0x04, 0x00000000 }, + { 0x40460c, 1, 0x04, 0x00002e00 }, + { 0x404610, 1, 0x04, 0x00000100 }, + { 0x404618, 8, 0x04, 0x00000000 }, + { 0x404638, 1, 0x04, 0x00000004 }, + { 0x40463c, 8, 0x04, 0x00000000 }, + { 0x40465c, 1, 0x04, 0x007f0100 }, + { 0x404660, 7, 0x04, 0x00000000 }, + { 0x40467c, 1, 0x04, 0x00000002 }, + { 0x404680, 8, 0x04, 0x00000000 }, + { 0x4046a0, 1, 0x04, 0x007f0080 }, + { 0x4046a4, 18, 0x04, 0x00000000 }, + { 0x4046f0, 2, 0x04, 0x00000000 }, + { 0x404700, 13, 0x04, 0x00000000 }, + { 0x404734, 1, 0x04, 0x00000100 }, + { 0x404738, 8, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x078000bf }, + { 0x405830, 1, 0x04, 0x02180000 }, + { 0x405834, 2, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x000103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b4, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_rstr2d_0[] = { + { 0x407804, 1, 0x04, 0x00000023 }, + { 0x40780c, 1, 0x04, 0x0a418820 }, + { 0x407810, 1, 0x04, 0x062080e6 }, + { 0x407814, 1, 0x04, 0x020398a4 }, + { 0x407818, 1, 0x04, 0x0e629062 }, + { 0x40781c, 1, 0x04, 0x0a418820 }, + { 0x407820, 1, 0x04, 0x000000e6 }, + { 0x4078bc, 1, 0x04, 0x00000103 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_scc_0[] = { + { 0x408000, 2, 0x04, 0x00000000 }, + { 0x408008, 1, 0x04, 0x00000018 }, + { 0x40800c, 2, 0x04, 0x00000000 }, + { 0x408014, 1, 0x04, 0x00000069 }, + { 0x408018, 1, 0x04, 0xe100e100 }, + { 0x408064, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x02802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x0003e00d }, + { 0x408900, 1, 0x04, 0x3080b801 }, + { 0x408904, 1, 0x04, 0x02000001 }, + { 0x408908, 1, 0x04, 0x00c80929 }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gf100_grctx_init_fe_0 }, + { gf100_grctx_init_pri_0 }, + { gf100_grctx_init_memfmt_0 }, + { gf100_grctx_init_ds_0 }, + { gf100_grctx_init_pd_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gf100_grctx_init_scc_0 }, + { gf100_grctx_init_be_0 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_gpc_unk_0[] = { + { 0x418380, 1, 0x04, 0x00000016 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_prop_0[] = { + { 0x418400, 1, 0x04, 0x38004e00 }, + { 0x418404, 1, 0x04, 0x71e0ffff }, + { 0x418408, 1, 0x04, 0x00000000 }, + { 0x41840c, 1, 0x04, 0x00001008 }, + { 0x418410, 1, 0x04, 0x0fff0fff }, + { 0x418414, 1, 0x04, 0x00200fff }, + { 0x418450, 6, 0x04, 0x00000000 }, + { 0x418468, 1, 0x04, 0x00000001 }, + { 0x41846c, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_gpc_unk_1[] = { + { 0x418600, 1, 0x04, 0x0000001f }, + { 0x418684, 1, 0x04, 0x0000000f }, + { 0x418700, 1, 0x04, 0x00000002 }, + { 0x418704, 1, 0x04, 0x00000080 }, + { 0x418708, 1, 0x04, 0x00000000 }, + { 0x41870c, 1, 0x04, 0x07c80000 }, + { 0x418710, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x0006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00008442 }, + { 0x418830, 1, 0x04, 0x00000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x00100000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_zcull_0[] = { + { 0x41891c, 1, 0x04, 0x00ff00ff }, + { 0x418924, 1, 0x04, 0x00000000 }, + { 0x418928, 1, 0x04, 0x00ffff00 }, + { 0x41892c, 1, 0x04, 0x0000ff00 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_crstr_0[] = { + { 0x418b00, 1, 0x04, 0x00000000 }, + { 0x418b08, 1, 0x04, 0x0a418820 }, + { 0x418b0c, 1, 0x04, 0x062080e6 }, + { 0x418b10, 1, 0x04, 0x020398a4 }, + { 0x418b14, 1, 0x04, 0x0e629062 }, + { 0x418b18, 1, 0x04, 0x0a418820 }, + { 0x418b1c, 1, 0x04, 0x000000e6 }, + { 0x418bb8, 1, 0x04, 0x00000103 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_gpm_0[] = { + { 0x418c08, 1, 0x04, 0x00000001 }, + { 0x418c10, 8, 0x04, 0x00000000 }, + { 0x418c80, 1, 0x04, 0x20200004 }, + { 0x418c8c, 1, 0x04, 0x00000001 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_gcc_0[] = { + { 0x419000, 1, 0x04, 0x00000780 }, + { 0x419004, 2, 0x04, 0x00000000 }, + { 0x419014, 1, 0x04, 0x00000004 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf100_grctx_init_prop_0 }, + { gf100_grctx_init_gpc_unk_1 }, + { gf100_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_gpc_1[] = { + { gf100_grctx_init_crstr_0 }, + { gf100_grctx_init_gpm_0 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_zcullr_0[] = { + { 0x418a00, 3, 0x04, 0x00000000 }, + { 0x418a0c, 1, 0x04, 0x00010000 }, + { 0x418a10, 3, 0x04, 0x00000000 }, + { 0x418a20, 3, 0x04, 0x00000000 }, + { 0x418a2c, 1, 0x04, 0x00010000 }, + { 0x418a30, 3, 0x04, 0x00000000 }, + { 0x418a40, 3, 0x04, 0x00000000 }, + { 0x418a4c, 1, 0x04, 0x00010000 }, + { 0x418a50, 3, 0x04, 0x00000000 }, + { 0x418a60, 3, 0x04, 0x00000000 }, + { 0x418a6c, 1, 0x04, 0x00010000 }, + { 0x418a70, 3, 0x04, 0x00000000 }, + { 0x418a80, 3, 0x04, 0x00000000 }, + { 0x418a8c, 1, 0x04, 0x00010000 }, + { 0x418a90, 3, 0x04, 0x00000000 }, + { 0x418aa0, 3, 0x04, 0x00000000 }, + { 0x418aac, 1, 0x04, 0x00010000 }, + { 0x418ab0, 3, 0x04, 0x00000000 }, + { 0x418ac0, 3, 0x04, 0x00000000 }, + { 0x418acc, 1, 0x04, 0x00010000 }, + { 0x418ad0, 3, 0x04, 0x00000000 }, + { 0x418ae0, 3, 0x04, 0x00000000 }, + { 0x418aec, 1, 0x04, 0x00010000 }, + { 0x418af0, 3, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_zcull[] = { + { gf100_grctx_init_zcullr_0 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_pe_0[] = { + { 0x419818, 1, 0x04, 0x00000000 }, + { 0x41983c, 1, 0x04, 0x00038bc7 }, + { 0x419848, 1, 0x04, 0x00000000 }, + { 0x419864, 1, 0x04, 0x0000012a }, + { 0x419888, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000001f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000023 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_wwdx_0[] = { + { 0x419b00, 1, 0x04, 0x0a418820 }, + { 0x419b04, 1, 0x04, 0x062080e6 }, + { 0x419b08, 1, 0x04, 0x020398a4 }, + { 0x419b0c, 1, 0x04, 0x0e629062 }, + { 0x419b10, 1, 0x04, 0x0a418820 }, + { 0x419b14, 1, 0x04, 0x000000e6 }, + { 0x419bd0, 1, 0x04, 0x00900103 }, + { 0x419be0, 1, 0x04, 0x00000001 }, + { 0x419be4, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x00000002 }, + { 0x419c04, 1, 0x04, 0x00000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_l1c_0[] = { + { 0x419cb0, 1, 0x04, 0x00060048 }, + { 0x419ce8, 1, 0x04, 0x00000000 }, + { 0x419cf4, 1, 0x04, 0x00000183 }, + {} +}; + +const struct gf100_gr_init +gf100_grctx_init_tpccs_0[] = { + { 0x419d20, 1, 0x04, 0x02180000 }, + { 0x419d24, 1, 0x04, 0x00001fff }, + {} +}; + +static const struct gf100_gr_init +gf100_grctx_init_sm_0[] = { + { 0x419e04, 3, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00000002 }, + { 0x419e44, 1, 0x04, 0x001beff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000000f }, + { 0x419e50, 17, 0x04, 0x00000000 }, + { 0x419e98, 1, 0x04, 0x00000000 }, + { 0x419f50, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gf100_grctx_pack_tpc[] = { + { gf100_grctx_init_pe_0 }, + { gf100_grctx_init_tex_0 }, + { gf100_grctx_init_wwdx_0 }, + { gf100_grctx_init_mpc_0 }, + { gf100_grctx_init_l1c_0 }, + { gf100_grctx_init_tpccs_0 }, + { gf100_grctx_init_sm_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gf100_grctx_patch_wr32(struct gf100_gr_chan *chan, u32 addr, u32 data) +{ + if (unlikely(!chan->mmio)) { + nvkm_wr32(chan->gr->base.engine.subdev.device, addr, data); + return; + } + + nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, addr); + nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, data); +} + +void +gf100_grctx_generate_r419cb8(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419cb8, 0x00007c00, 0x00000000); +} + +void +gf100_grctx_generate_bundle(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gf100_grctx_patch_wr32(chan, 0x408004, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408008, 0x80000000 | (size >> 8)); + gf100_grctx_patch_wr32(chan, 0x418808, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x41880c, 0x80000000 | (size >> 8)); +} + +void +gf100_grctx_generate_pagepool(struct gf100_gr_chan *chan, u64 addr) +{ + gf100_grctx_patch_wr32(chan, 0x40800c, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408010, 0x80000000); + gf100_grctx_patch_wr32(chan, 0x419004, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x419008, 0x00000000); +} + +void +gf100_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 attrib = grctx->attrib_nr; + int gpc, tpc; + u32 bo = 0; + + gf100_grctx_patch_wr32(chan, 0x405830, (attrib << 16)); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + const u32 o = TPC_UNIT(gpc, tpc, 0x0520); + + gf100_grctx_patch_wr32(chan, o, (attrib << 16) | bo); + bo += grctx->attrib_nr_max; + } + } +} + +void +gf100_grctx_generate_attrib_cb(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gf100_grctx_patch_wr32(chan, 0x418810, 0x80000000 | addr >> 12); + gf100_grctx_patch_wr32(chan, 0x419848, 0x10000000 | addr >> 12); +} + +u32 +gf100_grctx_generate_attrib_cb_size(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + + return 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max) * gr->tpc_total; +} + +void +gf100_grctx_generate_unkn(struct gf100_gr *gr) +{ +} + +void +gf100_grctx_generate_r4060a8(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u8 gpcmax = nvkm_rd32(device, 0x022430); + const u8 tpcmax = nvkm_rd32(device, 0x022434) * gpcmax; + int i, j, sm = 0; + u32 data; + + for (i = 0; i < DIV_ROUND_UP(tpcmax, 4); i++) { + for (data = 0, j = 0; j < 4; j++) { + if (sm < gr->sm_nr) + data |= gr->sm[sm++].gpc << (j * 8); + else + data |= 0x1f << (j * 8); + } + nvkm_wr32(device, 0x4060a8 + (i * 4), data); + } +} + +void +gf100_grctx_generate_rop_mapping(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 data[6] = {}, data2[2] = {}; + u8 shift, ntpcv; + int i; + + /* Pack tile map into register format. */ + for (i = 0; i < 32; i++) + data[i / 6] |= (gr->tile[i] & 0x07) << ((i % 6) * 5); + + /* Magic. */ + shift = 0; + ntpcv = gr->tpc_total; + while (!(ntpcv & (1 << 4))) { + ntpcv <<= 1; + shift++; + } + + data2[0] = (ntpcv << 16); + data2[0] |= (shift << 21); + data2[0] |= (((1 << (0 + 5)) % ntpcv) << 24); + for (i = 1; i < 7; i++) + data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5); + + /* GPC_BROADCAST */ + nvkm_wr32(device, 0x418bb8, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x418b08 + (i * 4), data[i]); + + /* GPC_BROADCAST.TP_BROADCAST */ + nvkm_wr32(device, 0x419bd0, (gr->tpc_total << 8) | + gr->screen_tile_row_offset | data2[0]); + nvkm_wr32(device, 0x419be4, data2[1]); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x419b00 + (i * 4), data[i]); + + /* UNK78xx */ + nvkm_wr32(device, 0x4078bc, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x40780c + (i * 4), data[i]); +} + +void +gf100_grctx_generate_max_ways_evict(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 fbps = nvkm_rd32(device, 0x121c74); + if (fbps == 1) + nvkm_mask(device, 0x17e91c, 0x001f0000, 0x00090000); +} + +static const u32 +gf100_grctx_alpha_beta_map[17][32] = { + [1] = { + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + }, + [2] = { + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + }, + //XXX: 3 + [4] = { + 1, 1, 1, 1, 1, 1, 1, 1, + 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, + 3, 3, 3, 3, 3, 3, 3, 3, + }, + //XXX: 5 + //XXX: 6 + [7] = { + 1, 1, 1, 1, + 2, 2, 2, 2, 2, 2, + 3, 3, 3, 3, 3, 3, + 4, 4, 4, 4, 4, 4, + 5, 5, 5, 5, 5, 5, + 6, 6, 6, 6, + }, + [8] = { + 1, 1, 1, + 2, 2, 2, 2, 2, + 3, 3, 3, 3, 3, + 4, 4, 4, 4, 4, 4, + 5, 5, 5, 5, 5, + 6, 6, 6, 6, 6, + 7, 7, 7, + }, + //XXX: 9 + //XXX: 10 + [11] = { + 1, 1, + 2, 2, 2, 2, + 3, 3, 3, + 4, 4, 4, 4, + 5, 5, 5, + 6, 6, 6, + 7, 7, 7, 7, + 8, 8, 8, + 9, 9, 9, 9, + 10, 10, + }, + //XXX: 12 + //XXX: 13 + [14] = { + 1, 1, + 2, 2, + 3, 3, 3, + 4, 4, 4, + 5, 5, + 6, 6, 6, + 7, 7, + 8, 8, 8, + 9, 9, + 10, 10, 10, + 11, 11, 11, + 12, 12, + 13, 13, + }, + [15] = { + 1, 1, + 2, 2, + 3, 3, + 4, 4, 4, + 5, 5, + 6, 6, 6, + 7, 7, + 8, 8, + 9, 9, 9, + 10, 10, + 11, 11, 11, + 12, 12, + 13, 13, + 14, 14, + }, + [16] = { + 1, 1, + 2, 2, + 3, 3, + 4, 4, + 5, 5, + 6, 6, 6, + 7, 7, + 8, 8, + 9, 9, + 10, 10, 10, + 11, 11, + 12, 12, + 13, 13, + 14, 14, + 15, 15, + }, +}; + +void +gf100_grctx_generate_alpha_beta_tables(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + int i, gpc; + + for (i = 0; i < 32; i++) { + u32 atarget = gf100_grctx_alpha_beta_map[gr->tpc_total][i]; + u32 abits[GPC_MAX] = {}, amask = 0, bmask = 0; + + if (!atarget) { + nvkm_warn(subdev, "missing alpha/beta mapping table\n"); + atarget = max_t(u32, gr->tpc_total * i / 32, 1); + } + + while (atarget) { + for (gpc = 0; atarget && gpc < gr->gpc_nr; gpc++) { + if (abits[gpc] < gr->tpc_nr[gpc]) { + abits[gpc]++; + atarget--; + } + } + } + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + u32 bbits = gr->tpc_nr[gpc] - abits[gpc]; + amask |= ((1 << abits[gpc]) - 1) << (gpc * 8); + bmask |= ((1 << bbits) - 1) << abits[gpc] << (gpc * 8); + } + + nvkm_wr32(device, 0x406800 + (i * 0x20), amask); + nvkm_wr32(device, 0x406c00 + (i * 0x20), bmask); + } +} + +void +gf100_grctx_generate_tpc_nr(struct gf100_gr *gr, int gpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c08), gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c8c), gr->tpc_nr[gpc]); +} + +void +gf100_grctx_generate_sm_id(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x698), sm); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x4e8), sm); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), sm); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), sm); +} + +void +gf100_grctx_generate_floorsweep(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *func = gr->func->grctx; + int sm; + + for (sm = 0; sm < gr->sm_nr; sm++) { + func->sm_id(gr, gr->sm[sm].gpc, gr->sm[sm].tpc, sm); + if (func->tpc_nr) + func->tpc_nr(gr, gr->sm[sm].gpc); + } + + gf100_gr_init_num_tpc_per_gpc(gr, false, true); + if (!func->skip_pd_num_tpc_per_gpc) + gf100_gr_init_num_tpc_per_gpc(gr, true, false); + + if (func->r4060a8) + func->r4060a8(gr); + + func->rop_mapping(gr); + + if (func->alpha_beta_tables) + func->alpha_beta_tables(gr); + if (func->max_ways_evict) + func->max_ways_evict(gr); + if (func->dist_skip_table) + func->dist_skip_table(gr); + if (func->r406500) + func->r406500(gr); + if (func->gpc_tpc_nr) + func->gpc_tpc_nr(gr); + if (func->r419f78) + func->r419f78(gr); + if (func->tpc_mask) + func->tpc_mask(gr); + if (func->smid_config) + func->smid_config(gr); +} + +void +gf100_grctx_generate_main(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + u32 idle_timeout; + + nvkm_mc_unk260(device, 0); + + if (!gr->sw_ctx) { + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc_0); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->gpc_1); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); + } else { + gf100_gr_mmio(gr, gr->sw_ctx); + } + + if (gr->func->init_419bd8) + gr->func->init_419bd8(gr); + if (grctx->r419ea8) + grctx->r419ea8(gr); + + gf100_gr_wait_idle(gr); + + idle_timeout = nvkm_mask(device, 0x404154, 0xffffffff, 0x00000000); + + grctx->pagepool(chan, chan->pagepool->addr); + grctx->bundle(chan, chan->bundle_cb->addr, grctx->bundle_size); + grctx->attrib_cb(chan, chan->attrib_cb->addr, grctx->attrib_cb_size(gr)); + grctx->attrib(chan); + if (grctx->patch_ltc) + grctx->patch_ltc(chan); + if (grctx->unknown_size) + grctx->unknown(chan, chan->unknown->addr, grctx->unknown_size); + grctx->unkn(gr); + + gf100_grctx_generate_floorsweep(gr); + + gf100_gr_wait_idle(gr); + + if (grctx->r400088) grctx->r400088(gr, false); + + if (gr->bundle) + gf100_gr_icmd(gr, gr->bundle); + else + gf100_gr_icmd(gr, grctx->icmd); + + if (gr->bundle_veid) + gf100_gr_icmd(gr, gr->bundle_veid); + else + gf100_gr_icmd(gr, grctx->sw_veid_bundle_init); + + if (gr->bundle64) + gf100_gr_icmd(gr, gr->bundle64); + else + if (grctx->sw_bundle64_init) + gf100_gr_icmd(gr, grctx->sw_bundle64_init); + + if (grctx->r400088) grctx->r400088(gr, true); + + nvkm_wr32(device, 0x404154, idle_timeout); + + if (gr->method) + gf100_gr_mthd(gr, gr->method); + else + gf100_gr_mthd(gr, grctx->mthd); + nvkm_mc_unk260(device, 1); + + if (grctx->r419cb8) + grctx->r419cb8(gr); + if (grctx->r418800) + grctx->r418800(gr); + if (grctx->r419eb0) + grctx->r419eb0(gr); + if (grctx->r419e00) + grctx->r419e00(gr); + if (grctx->r418e94) + grctx->r418e94(gr); + if (grctx->r419a3c) + grctx->r419a3c(gr); + if (grctx->r408840) + grctx->r408840(gr); + if (grctx->r419c0c) + grctx->r419c0c(gr); + + gf100_gr_wait_idle(gr); +} + +#define CB_RESERVED 0x80000 + +int +gf100_grctx_generate(struct gf100_gr *gr, struct gf100_gr_chan *chan, struct nvkm_gpuobj *inst) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_memory *data = NULL; + struct nvkm_vma *ctx = NULL; + int ret, i; + u64 addr; + + /* NV_PGRAPH_FE_PWR_MODE_FORCE_ON. */ + nvkm_wr32(device, 0x404170, 0x00000012); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x404170) & 0x00000010)) + break; + ); + + if (grctx->unkn88c) + grctx->unkn88c(gr, true); + + /* Reset FECS. */ + gr->func->fecs.reset(gr); + + if (grctx->unkn88c) + grctx->unkn88c(gr, false); + + /* NV_PGRAPH_FE_PWR_MODE_AUTO. */ + nvkm_wr32(device, 0x404170, 0x00000010); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x404170) & 0x00000010)) + break; + ); + + /* Init SCC RAM. */ + nvkm_wr32(device, 0x40802c, 0x00000001); + + /* Allocate memory to store context, and dummy global context buffers. */ + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, + CB_RESERVED + gr->size, 0, true, &data); + if (ret) + goto done; + + ret = nvkm_vmm_get(chan->vmm, 0, nvkm_memory_size(data), &ctx); + if (ret) + goto done; + + ret = nvkm_memory_map(data, 0, chan->vmm, ctx, NULL, 0); + if (ret) + goto done; + + /* Setup context pointer. */ + nvkm_kmap(inst); + nvkm_wo32(inst, 0x0210, lower_32_bits(ctx->addr + CB_RESERVED) | 4); + nvkm_wo32(inst, 0x0214, upper_32_bits(ctx->addr + CB_RESERVED)); + nvkm_done(inst); + + /* Make channel current. */ + addr = inst->addr >> 12; + if (gr->firmware) { + ret = gf100_gr_fecs_bind_pointer(gr, 0x80000000 | addr); + if (ret) + goto done_inst; + + nvkm_kmap(data); + nvkm_wo32(data, 0x1c, 1); + nvkm_wo32(data, 0x20, 0); + nvkm_wo32(data, 0x28, 0); + nvkm_wo32(data, 0x2c, 0); + nvkm_done(data); + } else { + nvkm_wr32(device, 0x409840, 0x80000000); + nvkm_wr32(device, 0x409500, 0x80000000 | addr); + nvkm_wr32(device, 0x409504, 0x00000001); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x80000000) + break; + ); + } + + grctx->main(chan); + + if (!gr->firmware) { + /* Trigger a context unload by unsetting the "next channel valid" bit + * and faking a context switch interrupt. + */ + nvkm_mask(device, 0x409b04, 0x80000000, 0x00000000); + nvkm_wr32(device, 0x409000, 0x00000100); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x409b00) & 0x80000000)) + break; + ) < 0) { + ret = -EBUSY; + goto done_inst; + } + } else { + ret = gf100_gr_fecs_wfi_golden_save(gr, 0x80000000 | addr); + if (ret) + goto done_inst; + + nvkm_mask(device, 0x409b00, 0x80000000, 0x00000000); + } + + gr->data = kmalloc(gr->size, GFP_KERNEL); + if (gr->data) { + nvkm_kmap(data); + for (i = 0; i < gr->size; i += 4) + gr->data[i / 4] = nvkm_ro32(data, CB_RESERVED + i); + nvkm_done(data); + ret = 0; + } else { + ret = -ENOMEM; + } + +done_inst: + nvkm_kmap(inst); + nvkm_wo32(inst, 0x0210, 0); + nvkm_wo32(inst, 0x0214, 0); + nvkm_done(inst); +done: + nvkm_vmm_put(chan->vmm, &ctx); + nvkm_memory_unref(&data); + return ret; +} + +const struct gf100_grctx_func +gf100_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gf100_grctx_generate_unkn, + .hub = gf100_grctx_pack_hub, + .gpc_0 = gf100_grctx_pack_gpc_0, + .gpc_1 = gf100_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf100_grctx_pack_tpc, + .icmd = gf100_grctx_pack_icmd, + .mthd = gf100_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf100_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf100_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h new file mode 100644 index 0000000000..de161e7a04 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h @@ -0,0 +1,276 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_GRCTX_NVC0_H__ +#define __NVKM_GRCTX_NVC0_H__ +#include "gf100.h" + +void gf100_grctx_patch_wr32(struct gf100_gr_chan *, u32 addr, u32 data); + +struct gf100_grctx_func { + void (*unkn88c)(struct gf100_gr *, bool on); + /* main context generation function */ + void (*main)(struct gf100_gr_chan *); + /* context-specific modify-on-first-load list generation function */ + void (*unkn)(struct gf100_gr *); + /* mmio context data */ + const struct gf100_gr_pack *hub; + const struct gf100_gr_pack *gpc_0; + const struct gf100_gr_pack *gpc_1; + const struct gf100_gr_pack *zcull; + const struct gf100_gr_pack *tpc; + const struct gf100_gr_pack *ppc; + /* indirect context data, generated with icmds/mthds */ + const struct gf100_gr_pack *icmd; + const struct gf100_gr_pack *mthd; + const struct gf100_gr_pack *sw_veid_bundle_init; + const struct gf100_gr_pack *sw_bundle64_init; + /* bundle circular buffer */ + void (*bundle)(struct gf100_gr_chan *, u64 addr, u32 size); + u32 bundle_size; + u32 bundle_min_gpm_fifo_depth; + u32 bundle_token_limit; + /* pagepool */ + void (*pagepool)(struct gf100_gr_chan *, u64 addr); + u32 pagepool_size; + /* attribute(/alpha) circular buffer */ + u32 (*attrib_cb_size)(struct gf100_gr *); + void (*attrib_cb)(struct gf100_gr_chan *, u64 addr, u32 size); + void (*attrib)(struct gf100_gr_chan *); + u32 attrib_nr_max; + u32 attrib_nr; + u32 alpha_nr_max; + u32 alpha_nr; + u32 gfxp_nr; + /* some other context buffer */ + void (*unknown)(struct gf100_gr_chan *, u64 addr, u32 size); + u32 unknown_size; + /* other patch buffer stuff */ + void (*patch_ltc)(struct gf100_gr_chan *); + /* floorsweeping */ + void (*sm_id)(struct gf100_gr *, int gpc, int tpc, int sm); + void (*tpc_nr)(struct gf100_gr *, int gpc); + bool skip_pd_num_tpc_per_gpc; + void (*r4060a8)(struct gf100_gr *); + void (*rop_mapping)(struct gf100_gr *); + void (*alpha_beta_tables)(struct gf100_gr *); + void (*max_ways_evict)(struct gf100_gr *); + void (*dist_skip_table)(struct gf100_gr *); + void (*r406500)(struct gf100_gr *); + void (*gpc_tpc_nr)(struct gf100_gr *); + void (*r419f78)(struct gf100_gr *); + void (*tpc_mask)(struct gf100_gr *); + void (*smid_config)(struct gf100_gr *); + /* misc other things */ + void (*r400088)(struct gf100_gr *, bool); + void (*r419cb8)(struct gf100_gr *); + void (*r418800)(struct gf100_gr *); + void (*r419eb0)(struct gf100_gr *); + void (*r419e00)(struct gf100_gr *); + void (*r418e94)(struct gf100_gr *); + void (*r419a3c)(struct gf100_gr *); + void (*r408840)(struct gf100_gr *); + void (*r419c0c)(struct gf100_gr *); + void (*r419ea8)(struct gf100_gr *); +}; + +extern const struct gf100_grctx_func gf100_grctx; +int gf100_grctx_generate(struct gf100_gr *, struct gf100_gr_chan *, struct nvkm_gpuobj *inst); +void gf100_grctx_generate_main(struct gf100_gr_chan *); +void gf100_grctx_generate_pagepool(struct gf100_gr_chan *, u64); +void gf100_grctx_generate_bundle(struct gf100_gr_chan *, u64, u32); +u32 gf100_grctx_generate_attrib_cb_size(struct gf100_gr *); +void gf100_grctx_generate_attrib_cb(struct gf100_gr_chan *, u64, u32); +void gf100_grctx_generate_attrib(struct gf100_gr_chan *); +void gf100_grctx_generate_unkn(struct gf100_gr *); +void gf100_grctx_generate_floorsweep(struct gf100_gr *); +void gf100_grctx_generate_sm_id(struct gf100_gr *, int, int, int); +void gf100_grctx_generate_tpc_nr(struct gf100_gr *, int); +void gf100_grctx_generate_r4060a8(struct gf100_gr *); +void gf100_grctx_generate_rop_mapping(struct gf100_gr *); +void gf100_grctx_generate_alpha_beta_tables(struct gf100_gr *); +void gf100_grctx_generate_max_ways_evict(struct gf100_gr *); +void gf100_grctx_generate_r419cb8(struct gf100_gr *); + +extern const struct gf100_grctx_func gf108_grctx; +void gf108_grctx_generate_attrib(struct gf100_gr_chan *); +void gf108_grctx_generate_unkn(struct gf100_gr *); + +extern const struct gf100_grctx_func gf104_grctx; +extern const struct gf100_grctx_func gf110_grctx; + +extern const struct gf100_grctx_func gf117_grctx; +void gf117_grctx_generate_attrib(struct gf100_gr_chan *); +void gf117_grctx_generate_rop_mapping(struct gf100_gr *); +void gf117_grctx_generate_dist_skip_table(struct gf100_gr *); + +extern const struct gf100_grctx_func gf119_grctx; + +extern const struct gf100_grctx_func gk104_grctx; +void gk104_grctx_generate_alpha_beta_tables(struct gf100_gr *); +void gk104_grctx_generate_gpc_tpc_nr(struct gf100_gr *); + +extern const struct gf100_grctx_func gk20a_grctx; +void gk104_grctx_generate_pagepool(struct gf100_gr_chan *, u64); +void gk104_grctx_generate_bundle(struct gf100_gr_chan *, u64, u32); +void gk104_grctx_generate_patch_ltc(struct gf100_gr_chan *); +void gk104_grctx_generate_unkn(struct gf100_gr *); +void gk104_grctx_generate_r418800(struct gf100_gr *); + +extern const struct gf100_grctx_func gk110_grctx; +void gk110_grctx_generate_r419eb0(struct gf100_gr *); +void gk110_grctx_generate_r419f78(struct gf100_gr *); + +extern const struct gf100_grctx_func gk110b_grctx; +extern const struct gf100_grctx_func gk208_grctx; + +extern const struct gf100_grctx_func gm107_grctx; +void gm107_grctx_generate_pagepool(struct gf100_gr_chan *, u64); +void gm107_grctx_generate_bundle(struct gf100_gr_chan *, u64, u32); +void gm107_grctx_generate_attrib_cb(struct gf100_gr_chan *, u64, u32); +void gm107_grctx_generate_attrib(struct gf100_gr_chan *); +void gm107_grctx_generate_sm_id(struct gf100_gr *, int, int, int); + +extern const struct gf100_grctx_func gm200_grctx; +void gm200_grctx_generate_dist_skip_table(struct gf100_gr *); +void gm200_grctx_generate_r406500(struct gf100_gr *); +void gm200_grctx_generate_tpc_mask(struct gf100_gr *); +void gm200_grctx_generate_smid_config(struct gf100_gr *); +void gm200_grctx_generate_r419a3c(struct gf100_gr *); + +extern const struct gf100_grctx_func gm20b_grctx; + +extern const struct gf100_grctx_func gp100_grctx; +void gp100_grctx_generate_pagepool(struct gf100_gr_chan *, u64); +void gp100_grctx_generate_attrib_cb(struct gf100_gr_chan *, u64, u32); +void gp100_grctx_generate_smid_config(struct gf100_gr *); + +extern const struct gf100_grctx_func gp102_grctx; +u32 gp102_grctx_generate_attrib_cb_size(struct gf100_gr *); +void gp102_grctx_generate_attrib(struct gf100_gr_chan *); + +extern const struct gf100_grctx_func gp104_grctx; + +extern const struct gf100_grctx_func gp107_grctx; + +extern const struct gf100_grctx_func gv100_grctx; + +extern const struct gf100_grctx_func tu102_grctx; +void gv100_grctx_unkn88c(struct gf100_gr *, bool); +void gv100_grctx_generate_unkn(struct gf100_gr *); +void gv100_grctx_generate_attrib_cb(struct gf100_gr_chan *, u64, u32); +void gv100_grctx_generate_attrib(struct gf100_gr_chan *); +void gv100_grctx_generate_rop_mapping(struct gf100_gr *); +void gv100_grctx_generate_r400088(struct gf100_gr *, bool); + +void tu102_grctx_generate_unknown(struct gf100_gr_chan *, u64, u32); + +extern const struct gf100_grctx_func ga102_grctx; + +/* context init value lists */ + +extern const struct gf100_gr_pack gf100_grctx_pack_icmd[]; + +extern const struct gf100_gr_pack gf100_grctx_pack_mthd[]; +extern const struct gf100_gr_init gf100_grctx_init_902d_0[]; +extern const struct gf100_gr_init gf100_grctx_init_9039_0[]; +extern const struct gf100_gr_init gf100_grctx_init_90c0_0[]; + +extern const struct gf100_gr_pack gf100_grctx_pack_hub[]; +extern const struct gf100_gr_init gf100_grctx_init_main_0[]; +extern const struct gf100_gr_init gf100_grctx_init_fe_0[]; +extern const struct gf100_gr_init gf100_grctx_init_pri_0[]; +extern const struct gf100_gr_init gf100_grctx_init_memfmt_0[]; +extern const struct gf100_gr_init gf100_grctx_init_rstr2d_0[]; +extern const struct gf100_gr_init gf100_grctx_init_scc_0[]; + +extern const struct gf100_gr_pack gf100_grctx_pack_gpc_0[]; +extern const struct gf100_gr_pack gf100_grctx_pack_gpc_1[]; +extern const struct gf100_gr_init gf100_grctx_init_gpc_unk_0[]; +extern const struct gf100_gr_init gf100_grctx_init_prop_0[]; +extern const struct gf100_gr_init gf100_grctx_init_gpc_unk_1[]; +extern const struct gf100_gr_init gf100_grctx_init_zcull_0[]; +extern const struct gf100_gr_init gf100_grctx_init_crstr_0[]; +extern const struct gf100_gr_init gf100_grctx_init_gpm_0[]; +extern const struct gf100_gr_init gf100_grctx_init_gcc_0[]; + +extern const struct gf100_gr_pack gf100_grctx_pack_zcull[]; + +extern const struct gf100_gr_pack gf100_grctx_pack_tpc[]; +extern const struct gf100_gr_init gf100_grctx_init_pe_0[]; +extern const struct gf100_gr_init gf100_grctx_init_wwdx_0[]; +extern const struct gf100_gr_init gf100_grctx_init_mpc_0[]; +extern const struct gf100_gr_init gf100_grctx_init_tpccs_0[]; + +extern const struct gf100_gr_init gf104_grctx_init_tex_0[]; +extern const struct gf100_gr_init gf104_grctx_init_l1c_0[]; +extern const struct gf100_gr_init gf104_grctx_init_sm_0[]; + +extern const struct gf100_gr_init gf108_grctx_init_9097_0[]; + +extern const struct gf100_gr_init gf108_grctx_init_gpm_0[]; + +extern const struct gf100_gr_init gf108_grctx_init_pe_0[]; +extern const struct gf100_gr_init gf108_grctx_init_wwdx_0[]; +extern const struct gf100_gr_init gf108_grctx_init_tpccs_0[]; + +extern const struct gf100_gr_init gf110_grctx_init_9197_0[]; +extern const struct gf100_gr_init gf110_grctx_init_9297_0[]; + +extern const struct gf100_gr_pack gf119_grctx_pack_icmd[]; + +extern const struct gf100_gr_pack gf119_grctx_pack_mthd[]; + +extern const struct gf100_gr_init gf119_grctx_init_fe_0[]; +extern const struct gf100_gr_init gf119_grctx_init_be_0[]; + +extern const struct gf100_gr_init gf119_grctx_init_prop_0[]; +extern const struct gf100_gr_init gf119_grctx_init_gpc_unk_1[]; +extern const struct gf100_gr_init gf119_grctx_init_crstr_0[]; + +extern const struct gf100_gr_init gf119_grctx_init_sm_0[]; + +extern const struct gf100_gr_init gf117_grctx_init_pe_0[]; + +extern const struct gf100_gr_init gf117_grctx_init_wwdx_0[]; + +extern const struct gf100_gr_pack gf117_grctx_pack_gpc_1[]; + +extern const struct gf100_gr_init gk104_grctx_init_memfmt_0[]; +extern const struct gf100_gr_init gk104_grctx_init_ds_0[]; +extern const struct gf100_gr_init gk104_grctx_init_scc_0[]; + +extern const struct gf100_gr_init gk104_grctx_init_gpm_0[]; + +extern const struct gf100_gr_init gk104_grctx_init_pes_0[]; + +extern const struct gf100_gr_pack gk104_grctx_pack_hub[]; +extern const struct gf100_gr_pack gk104_grctx_pack_tpc[]; +extern const struct gf100_gr_pack gk104_grctx_pack_ppc[]; +extern const struct gf100_gr_pack gk104_grctx_pack_icmd[]; +extern const struct gf100_gr_init gk104_grctx_init_a097_0[]; + +extern const struct gf100_gr_pack gk110_grctx_pack_icmd[]; + +extern const struct gf100_gr_pack gk110_grctx_pack_mthd[]; + +extern const struct gf100_gr_pack gk110_grctx_pack_hub[]; +extern const struct gf100_gr_init gk110_grctx_init_pri_0[]; +extern const struct gf100_gr_init gk110_grctx_init_cwd_0[]; + +extern const struct gf100_gr_pack gk110_grctx_pack_gpc_0[]; +extern const struct gf100_gr_pack gk110_grctx_pack_gpc_1[]; +extern const struct gf100_gr_init gk110_grctx_init_gpc_unk_2[]; + +extern const struct gf100_gr_init gk110_grctx_init_tex_0[]; +extern const struct gf100_gr_init gk110_grctx_init_mpc_0[]; +extern const struct gf100_gr_init gk110_grctx_init_l1c_0[]; + +extern const struct gf100_gr_pack gk110_grctx_pack_ppc[]; + +extern const struct gf100_gr_init gk208_grctx_init_rstr2d_0[]; + +extern const struct gf100_gr_init gk208_grctx_init_prop_0[]; +extern const struct gf100_gr_init gk208_grctx_init_crstr_0[]; + +extern const struct gf100_gr_init gm107_grctx_init_gpc_unk_0[]; +extern const struct gf100_gr_init gm107_grctx_init_wwdx_0[]; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c new file mode 100644 index 0000000000..ba63a3b465 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c @@ -0,0 +1,109 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +const struct gf100_gr_init +gf104_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000001f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000023 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x00000000 }, + { 0x419a20, 1, 0x04, 0x00000800 }, + { 0x419ac4, 1, 0x04, 0x0007f440 }, + {} +}; + +const struct gf100_gr_init +gf104_grctx_init_l1c_0[] = { + { 0x419cb0, 1, 0x04, 0x00020048 }, + { 0x419ce8, 1, 0x04, 0x00000000 }, + { 0x419cf4, 1, 0x04, 0x00000183 }, + {} +}; + +const struct gf100_gr_init +gf104_grctx_init_sm_0[] = { + { 0x419e04, 3, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00000002 }, + { 0x419e44, 1, 0x04, 0x001beff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000000f }, + { 0x419e50, 17, 0x04, 0x00000000 }, + { 0x419e98, 1, 0x04, 0x00000000 }, + { 0x419ee0, 1, 0x04, 0x00011110 }, + { 0x419f30, 11, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf104_grctx_pack_tpc[] = { + { gf100_grctx_init_pe_0 }, + { gf104_grctx_init_tex_0 }, + { gf100_grctx_init_wwdx_0 }, + { gf100_grctx_init_mpc_0 }, + { gf104_grctx_init_l1c_0 }, + { gf100_grctx_init_tpccs_0 }, + { gf104_grctx_init_sm_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gf104_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gf100_grctx_generate_unkn, + .hub = gf100_grctx_pack_hub, + .gpc_0 = gf100_grctx_pack_gpc_0, + .gpc_1 = gf100_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf104_grctx_pack_tpc, + .icmd = gf100_grctx_pack_icmd, + .mthd = gf100_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf100_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf100_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c new file mode 100644 index 0000000000..0bc2eab6ad --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c @@ -0,0 +1,807 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf108_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000b2, 40, 0x01, 0x00000000 }, + { 0x000210, 8, 0x01, 0x00000040 }, + { 0x000218, 8, 0x01, 0x0000c080 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00001fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x003fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x0005e0, 5, 0x01, 0x00000022 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x00000c48 }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00300008 }, + { 0x000841, 1, 0x01, 0x04000080 }, + { 0x000842, 1, 0x01, 0x00300008 }, + { 0x000843, 1, 0x01, 0x04000080 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x000944, 1, 0x01, 0x00000022 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000014 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_icmd[] = { + { gf108_grctx_init_icmd_0 }, + {} +}; + +const struct gf100_gr_init +gf108_grctx_init_9097_0[] = { + { 0x000800, 8, 0x40, 0x00000000 }, + { 0x000804, 8, 0x40, 0x00000000 }, + { 0x000808, 8, 0x40, 0x00000400 }, + { 0x00080c, 8, 0x40, 0x00000300 }, + { 0x000810, 1, 0x04, 0x000000cf }, + { 0x000850, 7, 0x40, 0x00000000 }, + { 0x000814, 8, 0x40, 0x00000040 }, + { 0x000818, 8, 0x40, 0x00000001 }, + { 0x00081c, 8, 0x40, 0x00000000 }, + { 0x000820, 8, 0x40, 0x00000000 }, + { 0x002700, 8, 0x20, 0x00000000 }, + { 0x002704, 8, 0x20, 0x00000000 }, + { 0x002708, 8, 0x20, 0x00000000 }, + { 0x00270c, 8, 0x20, 0x00000000 }, + { 0x002710, 8, 0x20, 0x00014000 }, + { 0x002714, 8, 0x20, 0x00000040 }, + { 0x001c00, 16, 0x10, 0x00000000 }, + { 0x001c04, 16, 0x10, 0x00000000 }, + { 0x001c08, 16, 0x10, 0x00000000 }, + { 0x001c0c, 16, 0x10, 0x00000000 }, + { 0x001d00, 16, 0x10, 0x00000000 }, + { 0x001d04, 16, 0x10, 0x00000000 }, + { 0x001d08, 16, 0x10, 0x00000000 }, + { 0x001d0c, 16, 0x10, 0x00000000 }, + { 0x001f00, 16, 0x08, 0x00000000 }, + { 0x001f04, 16, 0x08, 0x00000000 }, + { 0x001f80, 16, 0x08, 0x00000000 }, + { 0x001f84, 16, 0x08, 0x00000000 }, + { 0x002200, 5, 0x10, 0x00000022 }, + { 0x002000, 1, 0x04, 0x00000000 }, + { 0x002040, 1, 0x04, 0x00000011 }, + { 0x002080, 1, 0x04, 0x00000020 }, + { 0x0020c0, 1, 0x04, 0x00000030 }, + { 0x002100, 1, 0x04, 0x00000040 }, + { 0x002140, 1, 0x04, 0x00000051 }, + { 0x00200c, 6, 0x40, 0x00000001 }, + { 0x002010, 1, 0x04, 0x00000000 }, + { 0x002050, 1, 0x04, 0x00000000 }, + { 0x002090, 1, 0x04, 0x00000001 }, + { 0x0020d0, 1, 0x04, 0x00000002 }, + { 0x002110, 1, 0x04, 0x00000003 }, + { 0x002150, 1, 0x04, 0x00000004 }, + { 0x000380, 4, 0x20, 0x00000000 }, + { 0x000384, 4, 0x20, 0x00000000 }, + { 0x000388, 4, 0x20, 0x00000000 }, + { 0x00038c, 4, 0x20, 0x00000000 }, + { 0x000700, 4, 0x10, 0x00000000 }, + { 0x000704, 4, 0x10, 0x00000000 }, + { 0x000708, 4, 0x10, 0x00000000 }, + { 0x002800, 128, 0x04, 0x00000000 }, + { 0x000a00, 16, 0x20, 0x00000000 }, + { 0x000a04, 16, 0x20, 0x00000000 }, + { 0x000a08, 16, 0x20, 0x00000000 }, + { 0x000a0c, 16, 0x20, 0x00000000 }, + { 0x000a10, 16, 0x20, 0x00000000 }, + { 0x000a14, 16, 0x20, 0x00000000 }, + { 0x000c00, 16, 0x10, 0x00000000 }, + { 0x000c04, 16, 0x10, 0x00000000 }, + { 0x000c08, 16, 0x10, 0x00000000 }, + { 0x000c0c, 16, 0x10, 0x3f800000 }, + { 0x000d00, 8, 0x08, 0xffff0000 }, + { 0x000d04, 8, 0x08, 0xffff0000 }, + { 0x000e00, 16, 0x10, 0x00000000 }, + { 0x000e04, 16, 0x10, 0xffff0000 }, + { 0x000e08, 16, 0x10, 0xffff0000 }, + { 0x000d40, 4, 0x08, 0x00000000 }, + { 0x000d44, 4, 0x08, 0x00000000 }, + { 0x001e00, 8, 0x20, 0x00000001 }, + { 0x001e04, 8, 0x20, 0x00000001 }, + { 0x001e08, 8, 0x20, 0x00000002 }, + { 0x001e0c, 8, 0x20, 0x00000001 }, + { 0x001e10, 8, 0x20, 0x00000001 }, + { 0x001e14, 8, 0x20, 0x00000002 }, + { 0x001e18, 8, 0x20, 0x00000001 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x001514, 1, 0x04, 0x00000000 }, + { 0x000d68, 1, 0x04, 0x0000ffff }, + { 0x00121c, 1, 0x04, 0x0fac6881 }, + { 0x000fac, 1, 0x04, 0x00000001 }, + { 0x001538, 1, 0x04, 0x00000001 }, + { 0x000fe0, 2, 0x04, 0x00000000 }, + { 0x000fe8, 1, 0x04, 0x00000014 }, + { 0x000fec, 1, 0x04, 0x00000040 }, + { 0x000ff0, 1, 0x04, 0x00000000 }, + { 0x00179c, 1, 0x04, 0x00000000 }, + { 0x001228, 1, 0x04, 0x00000400 }, + { 0x00122c, 1, 0x04, 0x00000300 }, + { 0x001230, 1, 0x04, 0x00010001 }, + { 0x0007f8, 1, 0x04, 0x00000000 }, + { 0x0015b4, 1, 0x04, 0x00000001 }, + { 0x0015cc, 1, 0x04, 0x00000000 }, + { 0x001534, 1, 0x04, 0x00000000 }, + { 0x000fb0, 1, 0x04, 0x00000000 }, + { 0x0015d0, 1, 0x04, 0x00000000 }, + { 0x00153c, 1, 0x04, 0x00000000 }, + { 0x0016b4, 1, 0x04, 0x00000003 }, + { 0x000fbc, 4, 0x04, 0x0000ffff }, + { 0x000df8, 2, 0x04, 0x00000000 }, + { 0x001948, 1, 0x04, 0x00000000 }, + { 0x001970, 1, 0x04, 0x00000001 }, + { 0x00161c, 1, 0x04, 0x000009f0 }, + { 0x000dcc, 1, 0x04, 0x00000010 }, + { 0x00163c, 1, 0x04, 0x00000000 }, + { 0x0015e4, 1, 0x04, 0x00000000 }, + { 0x001160, 32, 0x04, 0x25e00040 }, + { 0x001880, 32, 0x04, 0x00000000 }, + { 0x000f84, 2, 0x04, 0x00000000 }, + { 0x0017c8, 2, 0x04, 0x00000000 }, + { 0x0017d0, 1, 0x04, 0x000000ff }, + { 0x0017d4, 1, 0x04, 0xffffffff }, + { 0x0017d8, 1, 0x04, 0x00000002 }, + { 0x0017dc, 1, 0x04, 0x00000000 }, + { 0x0015f4, 2, 0x04, 0x00000000 }, + { 0x001434, 2, 0x04, 0x00000000 }, + { 0x000d74, 1, 0x04, 0x00000000 }, + { 0x000dec, 1, 0x04, 0x00000001 }, + { 0x0013a4, 1, 0x04, 0x00000000 }, + { 0x001318, 1, 0x04, 0x00000001 }, + { 0x001644, 1, 0x04, 0x00000000 }, + { 0x000748, 1, 0x04, 0x00000000 }, + { 0x000de8, 1, 0x04, 0x00000000 }, + { 0x001648, 1, 0x04, 0x00000000 }, + { 0x0012a4, 1, 0x04, 0x00000000 }, + { 0x001120, 4, 0x04, 0x00000000 }, + { 0x001118, 1, 0x04, 0x00000000 }, + { 0x00164c, 1, 0x04, 0x00000000 }, + { 0x001658, 1, 0x04, 0x00000000 }, + { 0x001910, 1, 0x04, 0x00000290 }, + { 0x001518, 1, 0x04, 0x00000000 }, + { 0x00165c, 1, 0x04, 0x00000001 }, + { 0x001520, 1, 0x04, 0x00000000 }, + { 0x001604, 1, 0x04, 0x00000000 }, + { 0x001570, 1, 0x04, 0x00000000 }, + { 0x0013b0, 2, 0x04, 0x3f800000 }, + { 0x00020c, 1, 0x04, 0x00000000 }, + { 0x001670, 1, 0x04, 0x30201000 }, + { 0x001674, 1, 0x04, 0x70605040 }, + { 0x001678, 1, 0x04, 0xb8a89888 }, + { 0x00167c, 1, 0x04, 0xf8e8d8c8 }, + { 0x00166c, 1, 0x04, 0x00000000 }, + { 0x001680, 1, 0x04, 0x00ffff00 }, + { 0x0012d0, 1, 0x04, 0x00000003 }, + { 0x0012d4, 1, 0x04, 0x00000002 }, + { 0x001684, 2, 0x04, 0x00000000 }, + { 0x000dac, 2, 0x04, 0x00001b02 }, + { 0x000db4, 1, 0x04, 0x00000000 }, + { 0x00168c, 1, 0x04, 0x00000000 }, + { 0x0015bc, 1, 0x04, 0x00000000 }, + { 0x00156c, 1, 0x04, 0x00000000 }, + { 0x00187c, 1, 0x04, 0x00000000 }, + { 0x001110, 1, 0x04, 0x00000001 }, + { 0x000dc0, 3, 0x04, 0x00000000 }, + { 0x001234, 1, 0x04, 0x00000000 }, + { 0x001690, 1, 0x04, 0x00000000 }, + { 0x0012ac, 1, 0x04, 0x00000001 }, + { 0x0002c4, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x001000, 1, 0x04, 0x00000010 }, + { 0x0010fc, 1, 0x04, 0x00000000 }, + { 0x001290, 1, 0x04, 0x00000000 }, + { 0x000218, 1, 0x04, 0x00000010 }, + { 0x0012d8, 1, 0x04, 0x00000000 }, + { 0x0012dc, 1, 0x04, 0x00000010 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x00155c, 2, 0x04, 0x00000000 }, + { 0x001564, 1, 0x04, 0x00001fff }, + { 0x001574, 2, 0x04, 0x00000000 }, + { 0x00157c, 1, 0x04, 0x003fffff }, + { 0x001354, 1, 0x04, 0x00000000 }, + { 0x001664, 1, 0x04, 0x00000000 }, + { 0x001610, 1, 0x04, 0x00000012 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x00162c, 1, 0x04, 0x00000003 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000324, 6, 0x04, 0x3f800000 }, + { 0x000750, 1, 0x04, 0x00000000 }, + { 0x000760, 1, 0x04, 0x39291909 }, + { 0x000764, 1, 0x04, 0x79695949 }, + { 0x000768, 1, 0x04, 0xb9a99989 }, + { 0x00076c, 1, 0x04, 0xf9e9d9c9 }, + { 0x000770, 1, 0x04, 0x30201000 }, + { 0x000774, 1, 0x04, 0x70605040 }, + { 0x000778, 1, 0x04, 0x00009080 }, + { 0x000780, 1, 0x04, 0x39291909 }, + { 0x000784, 1, 0x04, 0x79695949 }, + { 0x000788, 1, 0x04, 0xb9a99989 }, + { 0x00078c, 1, 0x04, 0xf9e9d9c9 }, + { 0x0007d0, 1, 0x04, 0x30201000 }, + { 0x0007d4, 1, 0x04, 0x70605040 }, + { 0x0007d8, 1, 0x04, 0x00009080 }, + { 0x00037c, 1, 0x04, 0x00000001 }, + { 0x000740, 2, 0x04, 0x00000000 }, + { 0x002600, 1, 0x04, 0x00000000 }, + { 0x001918, 1, 0x04, 0x00000000 }, + { 0x00191c, 1, 0x04, 0x00000900 }, + { 0x001920, 1, 0x04, 0x00000405 }, + { 0x001308, 1, 0x04, 0x00000001 }, + { 0x001924, 1, 0x04, 0x00000000 }, + { 0x0013ac, 1, 0x04, 0x00000000 }, + { 0x00192c, 1, 0x04, 0x00000001 }, + { 0x00193c, 1, 0x04, 0x00002c1c }, + { 0x000d7c, 1, 0x04, 0x00000000 }, + { 0x000f8c, 1, 0x04, 0x00000000 }, + { 0x0002c0, 1, 0x04, 0x00000001 }, + { 0x001510, 1, 0x04, 0x00000000 }, + { 0x001940, 1, 0x04, 0x00000000 }, + { 0x000ff4, 2, 0x04, 0x00000000 }, + { 0x00194c, 2, 0x04, 0x00000000 }, + { 0x001968, 1, 0x04, 0x00000000 }, + { 0x001590, 1, 0x04, 0x0000003f }, + { 0x0007e8, 4, 0x04, 0x00000000 }, + { 0x00196c, 1, 0x04, 0x00000011 }, + { 0x00197c, 1, 0x04, 0x00000000 }, + { 0x000fcc, 2, 0x04, 0x00000000 }, + { 0x0002d8, 1, 0x04, 0x00000040 }, + { 0x001980, 1, 0x04, 0x00000080 }, + { 0x001504, 1, 0x04, 0x00000080 }, + { 0x001984, 1, 0x04, 0x00000000 }, + { 0x000300, 1, 0x04, 0x00000001 }, + { 0x0013a8, 1, 0x04, 0x00000000 }, + { 0x0012ec, 1, 0x04, 0x00000000 }, + { 0x001310, 1, 0x04, 0x00000000 }, + { 0x001314, 1, 0x04, 0x00000001 }, + { 0x001380, 1, 0x04, 0x00000000 }, + { 0x001384, 4, 0x04, 0x00000001 }, + { 0x001394, 1, 0x04, 0x00000000 }, + { 0x00139c, 1, 0x04, 0x00000000 }, + { 0x001398, 1, 0x04, 0x00000000 }, + { 0x001594, 1, 0x04, 0x00000000 }, + { 0x001598, 4, 0x04, 0x00000001 }, + { 0x000f54, 3, 0x04, 0x00000000 }, + { 0x0019bc, 1, 0x04, 0x00000000 }, + { 0x000f9c, 2, 0x04, 0x00000000 }, + { 0x0012cc, 1, 0x04, 0x00000000 }, + { 0x0012e8, 1, 0x04, 0x00000000 }, + { 0x00130c, 1, 0x04, 0x00000001 }, + { 0x001360, 8, 0x04, 0x00000000 }, + { 0x00133c, 2, 0x04, 0x00000001 }, + { 0x001344, 1, 0x04, 0x00000002 }, + { 0x001348, 2, 0x04, 0x00000001 }, + { 0x001350, 1, 0x04, 0x00000002 }, + { 0x001358, 1, 0x04, 0x00000001 }, + { 0x0012e4, 1, 0x04, 0x00000000 }, + { 0x00131c, 4, 0x04, 0x00000000 }, + { 0x0019c0, 1, 0x04, 0x00000000 }, + { 0x001140, 1, 0x04, 0x00000000 }, + { 0x0019c4, 1, 0x04, 0x00000000 }, + { 0x0019c8, 1, 0x04, 0x00001500 }, + { 0x00135c, 1, 0x04, 0x00000000 }, + { 0x000f90, 1, 0x04, 0x00000000 }, + { 0x0019e0, 8, 0x04, 0x00000001 }, + { 0x0019cc, 1, 0x04, 0x00000001 }, + { 0x0015b8, 1, 0x04, 0x00000000 }, + { 0x001a00, 1, 0x04, 0x00001111 }, + { 0x001a04, 7, 0x04, 0x00000000 }, + { 0x000d6c, 2, 0x04, 0xffff0000 }, + { 0x0010f8, 1, 0x04, 0x00001010 }, + { 0x000d80, 5, 0x04, 0x00000000 }, + { 0x000da0, 1, 0x04, 0x00000000 }, + { 0x001508, 1, 0x04, 0x80000000 }, + { 0x00150c, 1, 0x04, 0x40000000 }, + { 0x001668, 1, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000008 }, + { 0x000d9c, 1, 0x04, 0x00000001 }, + { 0x0007dc, 1, 0x04, 0x00000000 }, + { 0x00074c, 1, 0x04, 0x00000055 }, + { 0x001420, 1, 0x04, 0x00000003 }, + { 0x0017bc, 2, 0x04, 0x00000000 }, + { 0x0017c4, 1, 0x04, 0x00000001 }, + { 0x001008, 1, 0x04, 0x00000008 }, + { 0x00100c, 1, 0x04, 0x00000040 }, + { 0x001010, 1, 0x04, 0x0000012c }, + { 0x000d60, 1, 0x04, 0x00000040 }, + { 0x00075c, 1, 0x04, 0x00000003 }, + { 0x001018, 1, 0x04, 0x00000020 }, + { 0x00101c, 1, 0x04, 0x00000001 }, + { 0x001020, 1, 0x04, 0x00000020 }, + { 0x001024, 1, 0x04, 0x00000001 }, + { 0x001444, 3, 0x04, 0x00000000 }, + { 0x000360, 1, 0x04, 0x20164010 }, + { 0x000364, 1, 0x04, 0x00000020 }, + { 0x000368, 1, 0x04, 0x00000000 }, + { 0x000de4, 1, 0x04, 0x00000000 }, + { 0x000204, 1, 0x04, 0x00000006 }, + { 0x000208, 1, 0x04, 0x00000000 }, + { 0x0002cc, 1, 0x04, 0x003fffff }, + { 0x0002d0, 1, 0x04, 0x00000c48 }, + { 0x001220, 1, 0x04, 0x00000005 }, + { 0x000fdc, 1, 0x04, 0x00000000 }, + { 0x000f98, 1, 0x04, 0x00300008 }, + { 0x001284, 1, 0x04, 0x04000080 }, + { 0x001450, 1, 0x04, 0x00300008 }, + { 0x001454, 1, 0x04, 0x04000080 }, + { 0x000214, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf108_grctx_init_9197_0[] = { + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x0002e4, 1, 0x04, 0x0000b001 }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_mthd[] = { + { gf108_grctx_init_9097_0, 0x9097 }, + { gf108_grctx_init_9197_0, 0x9197 }, + { gf100_grctx_init_902d_0, 0x902d }, + { gf100_grctx_init_9039_0, 0x9039 }, + { gf100_grctx_init_90c0_0, 0x90c0 }, + {} +}; + +static const struct gf100_gr_init +gf108_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8000bf }, + { 0x405830, 1, 0x04, 0x02180218 }, + { 0x405834, 2, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf108_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x000103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b4, 2, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x80140078 }, + { 0x4064c4, 1, 0x04, 0x0086ffff }, + {} +}; + +static const struct gf100_gr_init +gf108_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x02802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1003e005 }, + { 0x408900, 1, 0x04, 0x3080b801 }, + { 0x408904, 1, 0x04, 0x62000001 }, + { 0x408908, 1, 0x04, 0x00c80929 }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gf100_grctx_init_fe_0 }, + { gf100_grctx_init_pri_0 }, + { gf100_grctx_init_memfmt_0 }, + { gf108_grctx_init_ds_0 }, + { gf108_grctx_init_pd_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gf100_grctx_init_scc_0 }, + { gf108_grctx_init_be_0 }, + {} +}; + +static const struct gf100_gr_init +gf108_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x0006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00008442 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x00100018 }, + {} +}; + +const struct gf100_gr_init +gf108_grctx_init_gpm_0[] = { + { 0x418c08, 1, 0x04, 0x00000001 }, + { 0x418c10, 8, 0x04, 0x00000000 }, + { 0x418c6c, 1, 0x04, 0x00000001 }, + { 0x418c80, 1, 0x04, 0x20200004 }, + { 0x418c8c, 1, 0x04, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf100_grctx_init_prop_0 }, + { gf100_grctx_init_gpc_unk_1 }, + { gf108_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_gpc_1[] = { + { gf100_grctx_init_crstr_0 }, + { gf108_grctx_init_gpm_0 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +const struct gf100_gr_init +gf108_grctx_init_pe_0[] = { + { 0x419818, 1, 0x04, 0x00000000 }, + { 0x41983c, 1, 0x04, 0x00038bc7 }, + { 0x419848, 1, 0x04, 0x00000000 }, + { 0x419864, 1, 0x04, 0x00000129 }, + { 0x419888, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf108_grctx_init_wwdx_0[] = { + { 0x419b00, 1, 0x04, 0x0a418820 }, + { 0x419b04, 1, 0x04, 0x062080e6 }, + { 0x419b08, 1, 0x04, 0x020398a4 }, + { 0x419b0c, 1, 0x04, 0x0e629062 }, + { 0x419b10, 1, 0x04, 0x0a418820 }, + { 0x419b14, 1, 0x04, 0x000000e6 }, + { 0x419bd0, 1, 0x04, 0x00900103 }, + { 0x419be0, 1, 0x04, 0x00400001 }, + { 0x419be4, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf108_grctx_init_tpccs_0[] = { + { 0x419d20, 1, 0x04, 0x12180000 }, + { 0x419d24, 1, 0x04, 0x00001fff }, + { 0x419d44, 1, 0x04, 0x02180218 }, + {} +}; + +static const struct gf100_gr_pack +gf108_grctx_pack_tpc[] = { + { gf108_grctx_init_pe_0 }, + { gf104_grctx_init_tex_0 }, + { gf108_grctx_init_wwdx_0 }, + { gf100_grctx_init_mpc_0 }, + { gf104_grctx_init_l1c_0 }, + { gf108_grctx_init_tpccs_0 }, + { gf104_grctx_init_sm_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gf108_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 beta = grctx->attrib_nr; + const int timeslice_mode = 1; + const int max_batches = 0xffff; + u32 bo = 0; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; + int gpc, tpc; + + gf100_grctx_patch_wr32(chan, 0x405830, (beta << 16) | alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + const u32 a = alpha; + const u32 b = beta; + const u32 t = timeslice_mode; + const u32 o = TPC_UNIT(gpc, tpc, 0x500); + + gf100_grctx_patch_wr32(chan, o + 0x20, (t << 28) | (b << 16) | bo); + bo += grctx->attrib_nr_max; + gf100_grctx_patch_wr32(chan, o + 0x44, (a << 16) | ao); + ao += grctx->alpha_nr_max; + } + } +} + +void +gf108_grctx_generate_unkn(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x418c6c, 0x00000001, 0x00000001); + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x419814, 0x00000004, 0x00000004); + nvkm_mask(device, 0x4064c0, 0x80000000, 0x80000000); + nvkm_mask(device, 0x405800, 0x08000000, 0x08000000); + nvkm_mask(device, 0x419c00, 0x00000008, 0x00000008); +} + +const struct gf100_grctx_func +gf108_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gf108_grctx_generate_unkn, + .hub = gf108_grctx_pack_hub, + .gpc_0 = gf108_grctx_pack_gpc_0, + .gpc_1 = gf108_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf108_grctx_pack_tpc, + .icmd = gf108_grctx_pack_icmd, + .mthd = gf108_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf108_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x324, + .alpha_nr = 0x218, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf100_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c new file mode 100644 index 0000000000..64b723b0af --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c @@ -0,0 +1,357 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf110_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000b2, 40, 0x01, 0x00000000 }, + { 0x000210, 8, 0x01, 0x00000040 }, + { 0x000218, 8, 0x01, 0x0000c080 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00001fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x003fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x0005e0, 5, 0x01, 0x00000022 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x00097d, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x00000c48 }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00300008 }, + { 0x000841, 1, 0x01, 0x04000080 }, + { 0x000842, 1, 0x01, 0x00300008 }, + { 0x000843, 1, 0x01, 0x04000080 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x000944, 1, 0x01, 0x00000022 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000014 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gf110_grctx_pack_icmd[] = { + { gf110_grctx_init_icmd_0 }, + {} +}; + +const struct gf100_gr_init +gf110_grctx_init_9197_0[] = { + { 0x0002e4, 1, 0x04, 0x0000b001 }, + {} +}; + +const struct gf100_gr_init +gf110_grctx_init_9297_0[] = { + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x00036c, 2, 0x04, 0x00000000 }, + { 0x0007a4, 2, 0x04, 0x00000000 }, + { 0x000374, 1, 0x04, 0x00000000 }, + { 0x000378, 1, 0x04, 0x00000020 }, + {} +}; + +static const struct gf100_gr_pack +gf110_grctx_pack_mthd[] = { + { gf108_grctx_init_9097_0, 0x9097 }, + { gf110_grctx_init_9197_0, 0x9197 }, + { gf110_grctx_init_9297_0, 0x9297 }, + { gf100_grctx_init_902d_0, 0x902d }, + { gf100_grctx_init_9039_0, 0x9039 }, + { gf100_grctx_init_90c0_0, 0x90c0 }, + {} +}; + +static const struct gf100_gr_init +gf110_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x0006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00008442 }, + { 0x418830, 1, 0x04, 0x00000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100000 }, + {} +}; + +static const struct gf100_gr_pack +gf110_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf100_grctx_init_prop_0 }, + { gf100_grctx_init_gpc_unk_1 }, + { gf110_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gf110_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gf100_grctx_generate_unkn, + .hub = gf100_grctx_pack_hub, + .gpc_0 = gf110_grctx_pack_gpc_0, + .gpc_1 = gf100_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf100_grctx_pack_tpc, + .icmd = gf110_grctx_pack_icmd, + .mthd = gf110_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf100_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf100_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c new file mode 100644 index 0000000000..e34c5da2a9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c @@ -0,0 +1,308 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include +#include + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf117_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8000bf }, + { 0x405830, 1, 0x04, 0x02180324 }, + { 0x405834, 1, 0x04, 0x08000000 }, + { 0x405838, 1, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x000103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b4, 3, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x801a0078 }, + { 0x4064c4, 1, 0x04, 0x00c9ffff }, + { 0x4064d0, 8, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf117_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gf119_grctx_init_fe_0 }, + { gf100_grctx_init_pri_0 }, + { gf100_grctx_init_memfmt_0 }, + { gf117_grctx_init_ds_0 }, + { gf117_grctx_init_pd_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gf100_grctx_init_scc_0 }, + { gf119_grctx_init_be_0 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00008442 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100018 }, + {} +}; + +static const struct gf100_gr_pack +gf117_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf119_grctx_init_prop_0 }, + { gf119_grctx_init_gpc_unk_1 }, + { gf117_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +const struct gf100_gr_pack +gf117_grctx_pack_gpc_1[] = { + { gf119_grctx_init_crstr_0 }, + { gf108_grctx_init_gpm_0 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +const struct gf100_gr_init +gf117_grctx_init_pe_0[] = { + { 0x419848, 1, 0x04, 0x00000000 }, + { 0x419864, 1, 0x04, 0x00000129 }, + { 0x419888, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000001f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000023 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x00008000 }, + { 0x419a20, 1, 0x04, 0x00000800 }, + { 0x419ac4, 1, 0x04, 0x0017f440 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x0000000a }, + { 0x419c04, 1, 0x04, 0x00000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + { 0x419c24, 1, 0x04, 0x00084210 }, + { 0x419c28, 1, 0x04, 0x3efbefbe }, + {} +}; + +static const struct gf100_gr_pack +gf117_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gf117_grctx_init_tex_0 }, + { gf117_grctx_init_mpc_0 }, + { gf104_grctx_init_l1c_0 }, + { gf119_grctx_init_sm_0 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_pes_0[] = { + { 0x41be24, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_init +gf117_grctx_init_cbm_0[] = { + { 0x41bec0, 1, 0x04, 0x12180000 }, + { 0x41bec4, 1, 0x04, 0x00003fff }, + { 0x41bee4, 1, 0x04, 0x03240218 }, + {} +}; + +const struct gf100_gr_init +gf117_grctx_init_wwdx_0[] = { + { 0x41bf00, 1, 0x04, 0x0a418820 }, + { 0x41bf04, 1, 0x04, 0x062080e6 }, + { 0x41bf08, 1, 0x04, 0x020398a4 }, + { 0x41bf0c, 1, 0x04, 0x0e629062 }, + { 0x41bf10, 1, 0x04, 0x0a418820 }, + { 0x41bf14, 1, 0x04, 0x000000e6 }, + { 0x41bfd0, 1, 0x04, 0x00900103 }, + { 0x41bfe0, 1, 0x04, 0x00400001 }, + { 0x41bfe4, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf117_grctx_pack_ppc[] = { + { gf117_grctx_init_pes_0 }, + { gf117_grctx_init_cbm_0 }, + { gf117_grctx_init_wwdx_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gf117_grctx_generate_dist_skip_table(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int i; + + for (i = 0; i < 8; i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); +} + +void +gf117_grctx_generate_rop_mapping(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 data[6] = {}, data2[2] = {}; + u8 shift, ntpcv; + int i; + + /* Pack tile map into register format. */ + for (i = 0; i < 32; i++) + data[i / 6] |= (gr->tile[i] & 0x07) << ((i % 6) * 5); + + /* Magic. */ + shift = 0; + ntpcv = gr->tpc_total; + while (!(ntpcv & (1 << 4))) { + ntpcv <<= 1; + shift++; + } + + data2[0] = (ntpcv << 16); + data2[0] |= (shift << 21); + data2[0] |= (((1 << (0 + 5)) % ntpcv) << 24); + for (i = 1; i < 7; i++) + data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5); + + /* GPC_BROADCAST */ + nvkm_wr32(device, 0x418bb8, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x418b08 + (i * 4), data[i]); + + /* GPC_BROADCAST.TP_BROADCAST */ + nvkm_wr32(device, 0x41bfd0, (gr->tpc_total << 8) | + gr->screen_tile_row_offset | data2[0]); + nvkm_wr32(device, 0x41bfe4, data2[1]); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x41bf00 + (i * 4), data[i]); + + /* UNK78xx */ + nvkm_wr32(device, 0x4078bc, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0; i < 6; i++) + nvkm_wr32(device, 0x40780c + (i * 4), data[i]); +} + +void +gf117_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 beta = grctx->attrib_nr; + const int timeslice_mode = 1; + const int max_batches = 0xffff; + u32 bo = 0; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; + int gpc, ppc; + + gf100_grctx_patch_wr32(chan, 0x405830, (beta << 16) | alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++) { + const u32 a = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 b = beta * gr->ppc_tpc_nr[gpc][ppc]; + const u32 t = timeslice_mode; + const u32 o = PPC_UNIT(gpc, ppc, 0); + + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + + gf100_grctx_patch_wr32(chan, o + 0xc0, (t << 28) | (b << 16) | bo); + bo += grctx->attrib_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, o + 0xe4, (a << 16) | ao); + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + } + } +} + +const struct gf100_grctx_func +gf117_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gf117_grctx_pack_hub, + .gpc_0 = gf117_grctx_pack_gpc_0, + .gpc_1 = gf117_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf117_grctx_pack_tpc, + .ppc = gf117_grctx_pack_ppc, + .icmd = gf119_grctx_pack_icmd, + .mthd = gf119_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x7ff, + .alpha_nr = 0x324, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c new file mode 100644 index 0000000000..426ad1b8d4 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c @@ -0,0 +1,527 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf119_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000b2, 40, 0x01, 0x00000000 }, + { 0x000210, 8, 0x01, 0x00000040 }, + { 0x000400, 24, 0x01, 0x00000040 }, + { 0x000218, 8, 0x01, 0x0000c080 }, + { 0x000440, 24, 0x01, 0x0000c080 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00001fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x003fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x0005e0, 5, 0x01, 0x00000022 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x00097d, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x00000c48 }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00300008 }, + { 0x000841, 1, 0x01, 0x04000080 }, + { 0x000842, 1, 0x01, 0x00300008 }, + { 0x000843, 1, 0x01, 0x04000080 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x000944, 1, 0x01, 0x00000022 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000014 }, + { 0x000351, 1, 0x01, 0x00000100 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095d, 1, 0x01, 0x00000001 }, + { 0x00082b, 1, 0x01, 0x00000004 }, + { 0x000942, 1, 0x01, 0x00010001 }, + { 0x000943, 1, 0x01, 0x00000001 }, + { 0x0007c5, 1, 0x01, 0x00010001 }, + { 0x000834, 1, 0x01, 0x00000001 }, + { 0x0007c7, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x00080c, 1, 0x01, 0x00000002 }, + { 0x00080d, 2, 0x01, 0x00000100 }, + { 0x00080f, 1, 0x01, 0x00000001 }, + { 0x000823, 1, 0x01, 0x00000002 }, + { 0x000824, 2, 0x01, 0x00000100 }, + { 0x000826, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +const struct gf100_gr_pack +gf119_grctx_pack_icmd[] = { + { gf119_grctx_init_icmd_0 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_90c0_0[] = { + { 0x002700, 8, 0x20, 0x00000000 }, + { 0x002704, 8, 0x20, 0x00000000 }, + { 0x002708, 8, 0x20, 0x00000000 }, + { 0x00270c, 8, 0x20, 0x00000000 }, + { 0x002710, 8, 0x20, 0x00014000 }, + { 0x002714, 8, 0x20, 0x00000040 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x000758, 1, 0x04, 0x00000100 }, + { 0x0002c4, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x000204, 3, 0x04, 0x00000000 }, + { 0x000214, 1, 0x04, 0x00000000 }, + { 0x00024c, 1, 0x04, 0x00000000 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x001664, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gf119_grctx_pack_mthd[] = { + { gf108_grctx_init_9097_0, 0x9097 }, + { gf110_grctx_init_9197_0, 0x9197 }, + { gf110_grctx_init_9297_0, 0x9297 }, + { gf100_grctx_init_902d_0, 0x902d }, + { gf100_grctx_init_9039_0, 0x9039 }, + { gf119_grctx_init_90c0_0, 0x90c0 }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_fe_0[] = { + { 0x404004, 10, 0x04, 0x00000000 }, + { 0x404044, 1, 0x04, 0x00000000 }, + { 0x404094, 13, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf0000087 }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040e8, 1, 0x04, 0x00001000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404138, 1, 0x04, 0x20000040 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000055 }, + { 0x404168, 1, 0x04, 0x00000000 }, + { 0x404178, 2, 0x04, 0x00000000 }, + { 0x404200, 8, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8000bf }, + { 0x405830, 1, 0x04, 0x02180218 }, + { 0x405834, 1, 0x04, 0x08000000 }, + { 0x405838, 1, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x000103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b4, 3, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x80140078 }, + { 0x4064c4, 1, 0x04, 0x0086ffff }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x02802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1043e005 }, + { 0x408900, 1, 0x04, 0x3080b801 }, + { 0x408904, 1, 0x04, 0x62000001 }, + { 0x408908, 1, 0x04, 0x00c8102f }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +static const struct gf100_gr_pack +gf119_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gf119_grctx_init_fe_0 }, + { gf100_grctx_init_pri_0 }, + { gf100_grctx_init_memfmt_0 }, + { gf119_grctx_init_ds_0 }, + { gf119_grctx_init_pd_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gf100_grctx_init_scc_0 }, + { gf119_grctx_init_be_0 }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_prop_0[] = { + { 0x418400, 1, 0x04, 0x38004e00 }, + { 0x418404, 1, 0x04, 0x71e0ffff }, + { 0x41840c, 1, 0x04, 0x00001008 }, + { 0x418410, 1, 0x04, 0x0fff0fff }, + { 0x418414, 1, 0x04, 0x02200fff }, + { 0x418450, 6, 0x04, 0x00000000 }, + { 0x418468, 1, 0x04, 0x00000001 }, + { 0x41846c, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_gpc_unk_1[] = { + { 0x418600, 1, 0x04, 0x0000001f }, + { 0x418684, 1, 0x04, 0x0000000f }, + { 0x418700, 1, 0x04, 0x00000002 }, + { 0x418704, 1, 0x04, 0x00000080 }, + { 0x418708, 3, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00008442 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100008 }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_crstr_0[] = { + { 0x418b00, 1, 0x04, 0x00000006 }, + { 0x418b08, 1, 0x04, 0x0a418820 }, + { 0x418b0c, 1, 0x04, 0x062080e6 }, + { 0x418b10, 1, 0x04, 0x020398a4 }, + { 0x418b14, 1, 0x04, 0x0e629062 }, + { 0x418b18, 1, 0x04, 0x0a418820 }, + { 0x418b1c, 1, 0x04, 0x000000e6 }, + { 0x418bb8, 1, 0x04, 0x00000103 }, + {} +}; + +static const struct gf100_gr_pack +gf119_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf119_grctx_init_prop_0 }, + { gf119_grctx_init_gpc_unk_1 }, + { gf119_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000001f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000023 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x00000000 }, + { 0x419a20, 1, 0x04, 0x00000800 }, + { 0x419ac4, 1, 0x04, 0x0017f440 }, + {} +}; + +static const struct gf100_gr_init +gf119_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x0000000a }, + { 0x419c04, 1, 0x04, 0x00000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + { 0x419c24, 1, 0x04, 0x00084210 }, + { 0x419c28, 1, 0x04, 0x3cf3cf3c }, + {} +}; + +const struct gf100_gr_init +gf119_grctx_init_sm_0[] = { + { 0x419e04, 3, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00000002 }, + { 0x419e44, 1, 0x04, 0x001beff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000000f }, + { 0x419e50, 17, 0x04, 0x00000000 }, + { 0x419e98, 1, 0x04, 0x00000000 }, + { 0x419ee0, 1, 0x04, 0x00010110 }, + { 0x419f30, 11, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf119_grctx_pack_tpc[] = { + { gf108_grctx_init_pe_0 }, + { gf119_grctx_init_tex_0 }, + { gf108_grctx_init_wwdx_0 }, + { gf119_grctx_init_mpc_0 }, + { gf104_grctx_init_l1c_0 }, + { gf108_grctx_init_tpccs_0 }, + { gf119_grctx_init_sm_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gf119_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gf108_grctx_generate_unkn, + .hub = gf119_grctx_pack_hub, + .gpc_0 = gf119_grctx_pack_gpc_0, + .gpc_1 = gf117_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gf119_grctx_pack_tpc, + .icmd = gf119_grctx_pack_icmd, + .mthd = gf119_grctx_pack_mthd, + .bundle = gf100_grctx_generate_bundle, + .bundle_size = 0x1800, + .pagepool = gf100_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf108_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x324, + .alpha_nr = 0x218, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .r4060a8 = gf100_grctx_generate_r4060a8, + .rop_mapping = gf100_grctx_generate_rop_mapping, + .alpha_beta_tables = gf100_grctx_generate_alpha_beta_tables, + .max_ways_evict = gf100_grctx_generate_max_ways_evict, + .r419cb8 = gf100_grctx_generate_r419cb8, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c new file mode 100644 index 0000000000..52a234b1ef --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c @@ -0,0 +1,1002 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include +#include + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk104_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000383, 1, 0x01, 0x00000011 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x00097d, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x003fffff }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00400008 }, + { 0x000841, 1, 0x01, 0x08000080 }, + { 0x000842, 1, 0x01, 0x00400008 }, + { 0x000843, 1, 0x01, 0x08000080 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000008 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +const struct gf100_gr_pack +gk104_grctx_pack_icmd[] = { + { gk104_grctx_init_icmd_0 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_a097_0[] = { + { 0x000800, 8, 0x40, 0x00000000 }, + { 0x000804, 8, 0x40, 0x00000000 }, + { 0x000808, 8, 0x40, 0x00000400 }, + { 0x00080c, 8, 0x40, 0x00000300 }, + { 0x000810, 1, 0x04, 0x000000cf }, + { 0x000850, 7, 0x40, 0x00000000 }, + { 0x000814, 8, 0x40, 0x00000040 }, + { 0x000818, 8, 0x40, 0x00000001 }, + { 0x00081c, 8, 0x40, 0x00000000 }, + { 0x000820, 8, 0x40, 0x00000000 }, + { 0x001c00, 16, 0x10, 0x00000000 }, + { 0x001c04, 16, 0x10, 0x00000000 }, + { 0x001c08, 16, 0x10, 0x00000000 }, + { 0x001c0c, 16, 0x10, 0x00000000 }, + { 0x001d00, 16, 0x10, 0x00000000 }, + { 0x001d04, 16, 0x10, 0x00000000 }, + { 0x001d08, 16, 0x10, 0x00000000 }, + { 0x001d0c, 16, 0x10, 0x00000000 }, + { 0x001f00, 16, 0x08, 0x00000000 }, + { 0x001f04, 16, 0x08, 0x00000000 }, + { 0x001f80, 16, 0x08, 0x00000000 }, + { 0x001f84, 16, 0x08, 0x00000000 }, + { 0x002000, 1, 0x04, 0x00000000 }, + { 0x002040, 1, 0x04, 0x00000011 }, + { 0x002080, 1, 0x04, 0x00000020 }, + { 0x0020c0, 1, 0x04, 0x00000030 }, + { 0x002100, 1, 0x04, 0x00000040 }, + { 0x002140, 1, 0x04, 0x00000051 }, + { 0x00200c, 6, 0x40, 0x00000001 }, + { 0x002010, 1, 0x04, 0x00000000 }, + { 0x002050, 1, 0x04, 0x00000000 }, + { 0x002090, 1, 0x04, 0x00000001 }, + { 0x0020d0, 1, 0x04, 0x00000002 }, + { 0x002110, 1, 0x04, 0x00000003 }, + { 0x002150, 1, 0x04, 0x00000004 }, + { 0x000380, 4, 0x20, 0x00000000 }, + { 0x000384, 4, 0x20, 0x00000000 }, + { 0x000388, 4, 0x20, 0x00000000 }, + { 0x00038c, 4, 0x20, 0x00000000 }, + { 0x000700, 4, 0x10, 0x00000000 }, + { 0x000704, 4, 0x10, 0x00000000 }, + { 0x000708, 4, 0x10, 0x00000000 }, + { 0x002800, 128, 0x04, 0x00000000 }, + { 0x000a00, 16, 0x20, 0x00000000 }, + { 0x000a04, 16, 0x20, 0x00000000 }, + { 0x000a08, 16, 0x20, 0x00000000 }, + { 0x000a0c, 16, 0x20, 0x00000000 }, + { 0x000a10, 16, 0x20, 0x00000000 }, + { 0x000a14, 16, 0x20, 0x00000000 }, + { 0x000c00, 16, 0x10, 0x00000000 }, + { 0x000c04, 16, 0x10, 0x00000000 }, + { 0x000c08, 16, 0x10, 0x00000000 }, + { 0x000c0c, 16, 0x10, 0x3f800000 }, + { 0x000d00, 8, 0x08, 0xffff0000 }, + { 0x000d04, 8, 0x08, 0xffff0000 }, + { 0x000e00, 16, 0x10, 0x00000000 }, + { 0x000e04, 16, 0x10, 0xffff0000 }, + { 0x000e08, 16, 0x10, 0xffff0000 }, + { 0x000d40, 4, 0x08, 0x00000000 }, + { 0x000d44, 4, 0x08, 0x00000000 }, + { 0x001e00, 8, 0x20, 0x00000001 }, + { 0x001e04, 8, 0x20, 0x00000001 }, + { 0x001e08, 8, 0x20, 0x00000002 }, + { 0x001e0c, 8, 0x20, 0x00000001 }, + { 0x001e10, 8, 0x20, 0x00000001 }, + { 0x001e14, 8, 0x20, 0x00000002 }, + { 0x001e18, 8, 0x20, 0x00000001 }, + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x001514, 1, 0x04, 0x00000000 }, + { 0x000d68, 1, 0x04, 0x0000ffff }, + { 0x00121c, 1, 0x04, 0x0fac6881 }, + { 0x000fac, 1, 0x04, 0x00000001 }, + { 0x001538, 1, 0x04, 0x00000001 }, + { 0x000fe0, 2, 0x04, 0x00000000 }, + { 0x000fe8, 1, 0x04, 0x00000014 }, + { 0x000fec, 1, 0x04, 0x00000040 }, + { 0x000ff0, 1, 0x04, 0x00000000 }, + { 0x00179c, 1, 0x04, 0x00000000 }, + { 0x001228, 1, 0x04, 0x00000400 }, + { 0x00122c, 1, 0x04, 0x00000300 }, + { 0x001230, 1, 0x04, 0x00010001 }, + { 0x0007f8, 1, 0x04, 0x00000000 }, + { 0x0015b4, 1, 0x04, 0x00000001 }, + { 0x0015cc, 1, 0x04, 0x00000000 }, + { 0x001534, 1, 0x04, 0x00000000 }, + { 0x000fb0, 1, 0x04, 0x00000000 }, + { 0x0015d0, 1, 0x04, 0x00000000 }, + { 0x00153c, 1, 0x04, 0x00000000 }, + { 0x0016b4, 1, 0x04, 0x00000003 }, + { 0x000fbc, 4, 0x04, 0x0000ffff }, + { 0x000df8, 2, 0x04, 0x00000000 }, + { 0x001948, 1, 0x04, 0x00000000 }, + { 0x001970, 1, 0x04, 0x00000001 }, + { 0x00161c, 1, 0x04, 0x000009f0 }, + { 0x000dcc, 1, 0x04, 0x00000010 }, + { 0x00163c, 1, 0x04, 0x00000000 }, + { 0x0015e4, 1, 0x04, 0x00000000 }, + { 0x001160, 32, 0x04, 0x25e00040 }, + { 0x001880, 32, 0x04, 0x00000000 }, + { 0x000f84, 2, 0x04, 0x00000000 }, + { 0x0017c8, 2, 0x04, 0x00000000 }, + { 0x0017d0, 1, 0x04, 0x000000ff }, + { 0x0017d4, 1, 0x04, 0xffffffff }, + { 0x0017d8, 1, 0x04, 0x00000002 }, + { 0x0017dc, 1, 0x04, 0x00000000 }, + { 0x0015f4, 2, 0x04, 0x00000000 }, + { 0x001434, 2, 0x04, 0x00000000 }, + { 0x000d74, 1, 0x04, 0x00000000 }, + { 0x000dec, 1, 0x04, 0x00000001 }, + { 0x0013a4, 1, 0x04, 0x00000000 }, + { 0x001318, 1, 0x04, 0x00000001 }, + { 0x001644, 1, 0x04, 0x00000000 }, + { 0x000748, 1, 0x04, 0x00000000 }, + { 0x000de8, 1, 0x04, 0x00000000 }, + { 0x001648, 1, 0x04, 0x00000000 }, + { 0x0012a4, 1, 0x04, 0x00000000 }, + { 0x001120, 4, 0x04, 0x00000000 }, + { 0x001118, 1, 0x04, 0x00000000 }, + { 0x00164c, 1, 0x04, 0x00000000 }, + { 0x001658, 1, 0x04, 0x00000000 }, + { 0x001910, 1, 0x04, 0x00000290 }, + { 0x001518, 1, 0x04, 0x00000000 }, + { 0x00165c, 1, 0x04, 0x00000001 }, + { 0x001520, 1, 0x04, 0x00000000 }, + { 0x001604, 1, 0x04, 0x00000000 }, + { 0x001570, 1, 0x04, 0x00000000 }, + { 0x0013b0, 2, 0x04, 0x3f800000 }, + { 0x00020c, 1, 0x04, 0x00000000 }, + { 0x001670, 1, 0x04, 0x30201000 }, + { 0x001674, 1, 0x04, 0x70605040 }, + { 0x001678, 1, 0x04, 0xb8a89888 }, + { 0x00167c, 1, 0x04, 0xf8e8d8c8 }, + { 0x00166c, 1, 0x04, 0x00000000 }, + { 0x001680, 1, 0x04, 0x00ffff00 }, + { 0x0012d0, 1, 0x04, 0x00000003 }, + { 0x0012d4, 1, 0x04, 0x00000002 }, + { 0x001684, 2, 0x04, 0x00000000 }, + { 0x000dac, 2, 0x04, 0x00001b02 }, + { 0x000db4, 1, 0x04, 0x00000000 }, + { 0x00168c, 1, 0x04, 0x00000000 }, + { 0x0015bc, 1, 0x04, 0x00000000 }, + { 0x00156c, 1, 0x04, 0x00000000 }, + { 0x00187c, 1, 0x04, 0x00000000 }, + { 0x001110, 1, 0x04, 0x00000001 }, + { 0x000dc0, 3, 0x04, 0x00000000 }, + { 0x001234, 1, 0x04, 0x00000000 }, + { 0x001690, 1, 0x04, 0x00000000 }, + { 0x0012ac, 1, 0x04, 0x00000001 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x001000, 1, 0x04, 0x00000010 }, + { 0x0010fc, 1, 0x04, 0x00000000 }, + { 0x001290, 1, 0x04, 0x00000000 }, + { 0x000218, 1, 0x04, 0x00000010 }, + { 0x0012d8, 1, 0x04, 0x00000000 }, + { 0x0012dc, 1, 0x04, 0x00000010 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x00155c, 2, 0x04, 0x00000000 }, + { 0x001564, 1, 0x04, 0x00000fff }, + { 0x001574, 2, 0x04, 0x00000000 }, + { 0x00157c, 1, 0x04, 0x000fffff }, + { 0x001354, 1, 0x04, 0x00000000 }, + { 0x001610, 1, 0x04, 0x00000012 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x00260c, 1, 0x04, 0x00000000 }, + { 0x0007ac, 1, 0x04, 0x00000000 }, + { 0x00162c, 1, 0x04, 0x00000003 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000324, 6, 0x04, 0x3f800000 }, + { 0x000750, 1, 0x04, 0x00000000 }, + { 0x000760, 1, 0x04, 0x39291909 }, + { 0x000764, 1, 0x04, 0x79695949 }, + { 0x000768, 1, 0x04, 0xb9a99989 }, + { 0x00076c, 1, 0x04, 0xf9e9d9c9 }, + { 0x000770, 1, 0x04, 0x30201000 }, + { 0x000774, 1, 0x04, 0x70605040 }, + { 0x000778, 1, 0x04, 0x00009080 }, + { 0x000780, 1, 0x04, 0x39291909 }, + { 0x000784, 1, 0x04, 0x79695949 }, + { 0x000788, 1, 0x04, 0xb9a99989 }, + { 0x00078c, 1, 0x04, 0xf9e9d9c9 }, + { 0x0007d0, 1, 0x04, 0x30201000 }, + { 0x0007d4, 1, 0x04, 0x70605040 }, + { 0x0007d8, 1, 0x04, 0x00009080 }, + { 0x00037c, 1, 0x04, 0x00000001 }, + { 0x000740, 2, 0x04, 0x00000000 }, + { 0x002600, 1, 0x04, 0x00000000 }, + { 0x001918, 1, 0x04, 0x00000000 }, + { 0x00191c, 1, 0x04, 0x00000900 }, + { 0x001920, 1, 0x04, 0x00000405 }, + { 0x001308, 1, 0x04, 0x00000001 }, + { 0x001924, 1, 0x04, 0x00000000 }, + { 0x0013ac, 1, 0x04, 0x00000000 }, + { 0x00192c, 1, 0x04, 0x00000001 }, + { 0x00193c, 1, 0x04, 0x00002c1c }, + { 0x000d7c, 1, 0x04, 0x00000000 }, + { 0x000f8c, 1, 0x04, 0x00000000 }, + { 0x0002c0, 1, 0x04, 0x00000001 }, + { 0x001510, 1, 0x04, 0x00000000 }, + { 0x001940, 1, 0x04, 0x00000000 }, + { 0x000ff4, 2, 0x04, 0x00000000 }, + { 0x00194c, 2, 0x04, 0x00000000 }, + { 0x001968, 1, 0x04, 0x00000000 }, + { 0x001590, 1, 0x04, 0x0000003f }, + { 0x0007e8, 4, 0x04, 0x00000000 }, + { 0x00196c, 1, 0x04, 0x00000011 }, + { 0x0002e4, 1, 0x04, 0x0000b001 }, + { 0x00036c, 2, 0x04, 0x00000000 }, + { 0x00197c, 1, 0x04, 0x00000000 }, + { 0x000fcc, 2, 0x04, 0x00000000 }, + { 0x0002d8, 1, 0x04, 0x00000040 }, + { 0x001980, 1, 0x04, 0x00000080 }, + { 0x001504, 1, 0x04, 0x00000080 }, + { 0x001984, 1, 0x04, 0x00000000 }, + { 0x000300, 1, 0x04, 0x00000001 }, + { 0x0013a8, 1, 0x04, 0x00000000 }, + { 0x0012ec, 1, 0x04, 0x00000000 }, + { 0x001310, 1, 0x04, 0x00000000 }, + { 0x001314, 1, 0x04, 0x00000001 }, + { 0x001380, 1, 0x04, 0x00000000 }, + { 0x001384, 4, 0x04, 0x00000001 }, + { 0x001394, 1, 0x04, 0x00000000 }, + { 0x00139c, 1, 0x04, 0x00000000 }, + { 0x001398, 1, 0x04, 0x00000000 }, + { 0x001594, 1, 0x04, 0x00000000 }, + { 0x001598, 4, 0x04, 0x00000001 }, + { 0x000f54, 3, 0x04, 0x00000000 }, + { 0x0019bc, 1, 0x04, 0x00000000 }, + { 0x000f9c, 2, 0x04, 0x00000000 }, + { 0x0012cc, 1, 0x04, 0x00000000 }, + { 0x0012e8, 1, 0x04, 0x00000000 }, + { 0x00130c, 1, 0x04, 0x00000001 }, + { 0x001360, 8, 0x04, 0x00000000 }, + { 0x00133c, 2, 0x04, 0x00000001 }, + { 0x001344, 1, 0x04, 0x00000002 }, + { 0x001348, 2, 0x04, 0x00000001 }, + { 0x001350, 1, 0x04, 0x00000002 }, + { 0x001358, 1, 0x04, 0x00000001 }, + { 0x0012e4, 1, 0x04, 0x00000000 }, + { 0x00131c, 4, 0x04, 0x00000000 }, + { 0x0019c0, 1, 0x04, 0x00000000 }, + { 0x001140, 1, 0x04, 0x00000000 }, + { 0x0019c4, 1, 0x04, 0x00000000 }, + { 0x0019c8, 1, 0x04, 0x00001500 }, + { 0x00135c, 1, 0x04, 0x00000000 }, + { 0x000f90, 1, 0x04, 0x00000000 }, + { 0x0019e0, 8, 0x04, 0x00000001 }, + { 0x0019cc, 1, 0x04, 0x00000001 }, + { 0x0015b8, 1, 0x04, 0x00000000 }, + { 0x001a00, 1, 0x04, 0x00001111 }, + { 0x001a04, 7, 0x04, 0x00000000 }, + { 0x000d6c, 2, 0x04, 0xffff0000 }, + { 0x0010f8, 1, 0x04, 0x00001010 }, + { 0x000d80, 5, 0x04, 0x00000000 }, + { 0x000da0, 1, 0x04, 0x00000000 }, + { 0x0007a4, 2, 0x04, 0x00000000 }, + { 0x001508, 1, 0x04, 0x80000000 }, + { 0x00150c, 1, 0x04, 0x40000000 }, + { 0x001668, 1, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000008 }, + { 0x000d9c, 1, 0x04, 0x00000001 }, + { 0x000374, 1, 0x04, 0x00000000 }, + { 0x000378, 1, 0x04, 0x00000020 }, + { 0x0007dc, 1, 0x04, 0x00000000 }, + { 0x00074c, 1, 0x04, 0x00000055 }, + { 0x001420, 1, 0x04, 0x00000003 }, + { 0x0017bc, 2, 0x04, 0x00000000 }, + { 0x0017c4, 1, 0x04, 0x00000001 }, + { 0x001008, 1, 0x04, 0x00000008 }, + { 0x00100c, 1, 0x04, 0x00000040 }, + { 0x001010, 1, 0x04, 0x0000012c }, + { 0x000d60, 1, 0x04, 0x00000040 }, + { 0x00075c, 1, 0x04, 0x00000003 }, + { 0x001018, 1, 0x04, 0x00000020 }, + { 0x00101c, 1, 0x04, 0x00000001 }, + { 0x001020, 1, 0x04, 0x00000020 }, + { 0x001024, 1, 0x04, 0x00000001 }, + { 0x001444, 3, 0x04, 0x00000000 }, + { 0x000360, 1, 0x04, 0x20164010 }, + { 0x000364, 1, 0x04, 0x00000020 }, + { 0x000368, 1, 0x04, 0x00000000 }, + { 0x000de4, 1, 0x04, 0x00000000 }, + { 0x000204, 1, 0x04, 0x00000006 }, + { 0x000208, 1, 0x04, 0x00000000 }, + { 0x0002cc, 2, 0x04, 0x003fffff }, + { 0x001220, 1, 0x04, 0x00000005 }, + { 0x000fdc, 1, 0x04, 0x00000000 }, + { 0x000f98, 1, 0x04, 0x00400008 }, + { 0x001284, 1, 0x04, 0x08000080 }, + { 0x001450, 1, 0x04, 0x00400008 }, + { 0x001454, 1, 0x04, 0x08000080 }, + { 0x000214, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gk104_grctx_pack_mthd[] = { + { gk104_grctx_init_a097_0, 0xa097 }, + { gf100_grctx_init_902d_0, 0x902d }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_fe_0[] = { + { 0x404010, 5, 0x04, 0x00000000 }, + { 0x404024, 1, 0x04, 0x0000e000 }, + { 0x404028, 1, 0x04, 0x00000000 }, + { 0x4040a8, 8, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf800008f }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040e8, 1, 0x04, 0x00001000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404138, 1, 0x04, 0x20000040 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000055 }, + { 0x4041a0, 4, 0x04, 0x00000000 }, + { 0x404200, 4, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_memfmt_0[] = { + { 0x404604, 1, 0x04, 0x00000014 }, + { 0x404608, 1, 0x04, 0x00000000 }, + { 0x40460c, 1, 0x04, 0x00003fff }, + { 0x404610, 1, 0x04, 0x00000100 }, + { 0x404618, 4, 0x04, 0x00000000 }, + { 0x40462c, 2, 0x04, 0x00000000 }, + { 0x404640, 1, 0x04, 0x00000000 }, + { 0x404654, 1, 0x04, 0x00000000 }, + { 0x404660, 1, 0x04, 0x00000000 }, + { 0x404678, 1, 0x04, 0x00000000 }, + { 0x40467c, 1, 0x04, 0x00000002 }, + { 0x404680, 8, 0x04, 0x00000000 }, + { 0x4046a0, 1, 0x04, 0x007f0080 }, + { 0x4046a4, 8, 0x04, 0x00000000 }, + { 0x4046c8, 3, 0x04, 0x00000000 }, + { 0x404700, 3, 0x04, 0x00000000 }, + { 0x404718, 7, 0x04, 0x00000000 }, + { 0x404734, 1, 0x04, 0x00000100 }, + { 0x404738, 2, 0x04, 0x00000000 }, + { 0x404744, 2, 0x04, 0x00000000 }, + { 0x404754, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8000bf }, + { 0x405830, 1, 0x04, 0x02180648 }, + { 0x405834, 1, 0x04, 0x08000000 }, + { 0x405838, 1, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_cwd_0[] = { + { 0x405b00, 1, 0x04, 0x00000000 }, + { 0x405b10, 1, 0x04, 0x00001000 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x004103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b4, 2, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x801a00f0 }, + { 0x4064c4, 1, 0x04, 0x0192ffff }, + { 0x4064c8, 1, 0x04, 0x01800600 }, + { 0x4064cc, 9, 0x04, 0x00000000 }, + { 0x4064fc, 1, 0x04, 0x0000022a }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_sked_0[] = { + { 0x407040, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_scc_0[] = { + { 0x408000, 2, 0x04, 0x00000000 }, + { 0x408008, 1, 0x04, 0x00000030 }, + { 0x40800c, 2, 0x04, 0x00000000 }, + { 0x408014, 1, 0x04, 0x00000069 }, + { 0x408018, 1, 0x04, 0xe100e100 }, + { 0x408064, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x02802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1043e005 }, + { 0x408840, 1, 0x04, 0x0000000b }, + { 0x408900, 1, 0x04, 0x3080b801 }, + { 0x408904, 1, 0x04, 0x62000001 }, + { 0x408908, 1, 0x04, 0x00c8102f }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +const struct gf100_gr_pack +gk104_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gk104_grctx_init_fe_0 }, + { gf100_grctx_init_pri_0 }, + { gk104_grctx_init_memfmt_0 }, + { gk104_grctx_init_ds_0 }, + { gk104_grctx_init_cwd_0 }, + { gk104_grctx_init_pd_0 }, + { gk104_grctx_init_sked_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gk104_grctx_init_scc_0 }, + { gk104_grctx_init_be_0 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006860a }, + { 0x418808, 3, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00000044 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100018 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_gpm_0[] = { + { 0x418c08, 1, 0x04, 0x00000001 }, + { 0x418c10, 8, 0x04, 0x00000000 }, + { 0x418c40, 1, 0x04, 0xffffffff }, + { 0x418c6c, 1, 0x04, 0x00000001 }, + { 0x418c80, 1, 0x04, 0x20200004 }, + { 0x418c8c, 1, 0x04, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gk104_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf119_grctx_init_prop_0 }, + { gf119_grctx_init_gpc_unk_1 }, + { gk104_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +static const struct gf100_gr_pack +gk104_grctx_pack_gpc_1[] = { + { gf119_grctx_init_crstr_0 }, + { gk104_grctx_init_gpm_0 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000000f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000021 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x0000c000 }, + { 0x419a20, 1, 0x04, 0x00000800 }, + { 0x419a30, 1, 0x04, 0x00000001 }, + { 0x419ac4, 1, 0x04, 0x0037f440 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x0000000a }, + { 0x419c04, 1, 0x04, 0x80000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + { 0x419c24, 1, 0x04, 0x00084210 }, + { 0x419c28, 1, 0x04, 0x3efbefbe }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_l1c_0[] = { + { 0x419ce8, 1, 0x04, 0x00000000 }, + { 0x419cf4, 1, 0x04, 0x00003203 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_sm_0[] = { + { 0x419e04, 3, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00000402 }, + { 0x419e44, 1, 0x04, 0x0013eff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000007f }, + { 0x419e50, 19, 0x04, 0x00000000 }, + { 0x419eac, 1, 0x04, 0x00001f8f }, + { 0x419eb0, 1, 0x04, 0x00000d3f }, + { 0x419ec8, 1, 0x04, 0x0001304f }, + { 0x419f30, 8, 0x04, 0x00000000 }, + { 0x419f58, 1, 0x04, 0x00000000 }, + { 0x419f70, 1, 0x04, 0x00000000 }, + { 0x419f78, 1, 0x04, 0x0000000b }, + { 0x419f7c, 1, 0x04, 0x0000027c }, + {} +}; + +const struct gf100_gr_pack +gk104_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gk104_grctx_init_tex_0 }, + { gk104_grctx_init_mpc_0 }, + { gk104_grctx_init_l1c_0 }, + { gk104_grctx_init_sm_0 }, + {} +}; + +const struct gf100_gr_init +gk104_grctx_init_pes_0[] = { + { 0x41be24, 1, 0x04, 0x00000006 }, + {} +}; + +static const struct gf100_gr_init +gk104_grctx_init_cbm_0[] = { + { 0x41bec0, 1, 0x04, 0x12180000 }, + { 0x41bec4, 1, 0x04, 0x00037f7f }, + { 0x41bee4, 1, 0x04, 0x06480430 }, + {} +}; + +const struct gf100_gr_pack +gk104_grctx_pack_ppc[] = { + { gk104_grctx_init_pes_0 }, + { gk104_grctx_init_cbm_0 }, + { gf117_grctx_init_wwdx_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gk104_grctx_generate_r418800(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + /*XXX: Not real sure where to apply these, there doesn't seem + * to be any pattern to which chipsets it's done on. + * + * Perhaps a VBIOS tweak? + */ + if (0) { + nvkm_mask(device, 0x418800, 0x00200000, 0x00200000); + nvkm_mask(device, 0x41be10, 0x00800000, 0x00800000); + } +} + +void +gk104_grctx_generate_patch_ltc(struct gf100_gr_chan *chan) +{ + struct nvkm_device *device = chan->gr->base.engine.subdev.device; + u32 data0 = nvkm_rd32(device, 0x17e91c); + u32 data1 = nvkm_rd32(device, 0x17e920); + + /*XXX: Figure out how to modify this correctly! */ + gf100_grctx_patch_wr32(chan, 0x17e91c, data0); + gf100_grctx_patch_wr32(chan, 0x17e920, data1); +} + +void +gk104_grctx_generate_bundle(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + const struct gf100_grctx_func *grctx = chan->gr->func->grctx; + const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth, size / 0x20); + const u32 token_limit = grctx->bundle_token_limit; + + gf100_grctx_generate_bundle(chan, addr, size); + gf100_grctx_patch_wr32(chan, 0x4064c8, (state_limit << 16) | token_limit); +} + +void +gk104_grctx_generate_pagepool(struct gf100_gr_chan *chan, u64 addr) +{ + gf100_grctx_generate_pagepool(chan, addr); + gf100_grctx_patch_wr32(chan, 0x4064cc, 0x80000000); +} + +void +gk104_grctx_generate_unkn(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x418c6c, 0x00000001, 0x00000001); + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x41be08, 0x00000004, 0x00000004); + nvkm_mask(device, 0x4064c0, 0x80000000, 0x80000000); + nvkm_mask(device, 0x405800, 0x08000000, 0x08000000); + nvkm_mask(device, 0x419c00, 0x00000008, 0x00000008); +} + +static void +gk104_grctx_generate_r419f78(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + /* bit 3 set disables loads in fp helper invocations, we need it enabled */ + nvkm_mask(device, 0x419f78, 0x00000009, 0x00000000); +} + +void +gk104_grctx_generate_gpc_tpc_nr(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); +} + +void +gk104_grctx_generate_alpha_beta_tables(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int i, j, gpc, ppc; + + for (i = 0; i < 32; i++) { + u32 atarget = max_t(u32, gr->tpc_total * i / 32, 1); + u32 btarget = gr->tpc_total - atarget; + bool alpha = atarget < btarget; + u64 amask = 0, bmask = 0; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++) { + u32 ppc_tpcs = gr->ppc_tpc_nr[gpc][ppc]; + u32 abits, bbits, pmask; + + if (alpha) { + abits = atarget ? ppc_tpcs : 0; + bbits = ppc_tpcs - abits; + } else { + bbits = btarget ? ppc_tpcs : 0; + abits = ppc_tpcs - bbits; + } + + pmask = gr->ppc_tpc_mask[gpc][ppc]; + while (ppc_tpcs-- > abits) + pmask &= pmask - 1; + amask |= (u64)pmask << (gpc * 8); + + pmask ^= gr->ppc_tpc_mask[gpc][ppc]; + bmask |= (u64)pmask << (gpc * 8); + + atarget -= min(abits, atarget); + btarget -= min(bbits, btarget); + if ((abits > 0) || (bbits > 0)) + alpha = !alpha; + } + } + + for (j = 0; j < gr->gpc_nr; j += 4, amask >>= 32, bmask >>= 32) { + nvkm_wr32(device, 0x406800 + (i * 0x20) + j, amask); + nvkm_wr32(device, 0x406c00 + (i * 0x20) + j, bmask); + } + } +} + +const struct gf100_grctx_func +gk104_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gk104_grctx_pack_hub, + .gpc_0 = gk104_grctx_pack_gpc_0, + .gpc_1 = gk104_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gk104_grctx_pack_tpc, + .ppc = gk104_grctx_pack_ppc, + .icmd = gk104_grctx_pack_icmd, + .mthd = gk104_grctx_pack_mthd, + .bundle = gk104_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x600, + .pagepool = gk104_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x7ff, + .alpha_nr = 0x648, + .patch_ltc = gk104_grctx_generate_patch_ltc, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .r419f78 = gk104_grctx_generate_r419f78, + .r418800 = gk104_grctx_generate_r418800, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c new file mode 100644 index 0000000000..3acdd9eeb7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c @@ -0,0 +1,867 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk110_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x0002f2, 2, 0x01, 0x00000001 }, + { 0x0002f5, 1, 0x01, 0x00000001 }, + { 0x0002f7, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000383, 1, 0x01, 0x00000011 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000662, 1, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x00080b, 1, 0x01, 0x00000002 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x000a0d, 1, 0x01, 0x00000006 }, + { 0x00097d, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x003fffff }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00400008 }, + { 0x000841, 1, 0x01, 0x08000080 }, + { 0x000842, 1, 0x01, 0x00400008 }, + { 0x000843, 1, 0x01, 0x08000080 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000008 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_icmd[] = { + { gk110_grctx_init_icmd_0 }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_a197_0[] = { + { 0x000800, 8, 0x40, 0x00000000 }, + { 0x000804, 8, 0x40, 0x00000000 }, + { 0x000808, 8, 0x40, 0x00000400 }, + { 0x00080c, 8, 0x40, 0x00000300 }, + { 0x000810, 1, 0x04, 0x000000cf }, + { 0x000850, 7, 0x40, 0x00000000 }, + { 0x000814, 8, 0x40, 0x00000040 }, + { 0x000818, 8, 0x40, 0x00000001 }, + { 0x00081c, 8, 0x40, 0x00000000 }, + { 0x000820, 8, 0x40, 0x00000000 }, + { 0x001c00, 16, 0x10, 0x00000000 }, + { 0x001c04, 16, 0x10, 0x00000000 }, + { 0x001c08, 16, 0x10, 0x00000000 }, + { 0x001c0c, 16, 0x10, 0x00000000 }, + { 0x001d00, 16, 0x10, 0x00000000 }, + { 0x001d04, 16, 0x10, 0x00000000 }, + { 0x001d08, 16, 0x10, 0x00000000 }, + { 0x001d0c, 16, 0x10, 0x00000000 }, + { 0x001f00, 16, 0x08, 0x00000000 }, + { 0x001f04, 16, 0x08, 0x00000000 }, + { 0x001f80, 16, 0x08, 0x00000000 }, + { 0x001f84, 16, 0x08, 0x00000000 }, + { 0x002000, 1, 0x04, 0x00000000 }, + { 0x002040, 1, 0x04, 0x00000011 }, + { 0x002080, 1, 0x04, 0x00000020 }, + { 0x0020c0, 1, 0x04, 0x00000030 }, + { 0x002100, 1, 0x04, 0x00000040 }, + { 0x002140, 1, 0x04, 0x00000051 }, + { 0x00200c, 6, 0x40, 0x00000001 }, + { 0x002010, 1, 0x04, 0x00000000 }, + { 0x002050, 1, 0x04, 0x00000000 }, + { 0x002090, 1, 0x04, 0x00000001 }, + { 0x0020d0, 1, 0x04, 0x00000002 }, + { 0x002110, 1, 0x04, 0x00000003 }, + { 0x002150, 1, 0x04, 0x00000004 }, + { 0x000380, 4, 0x20, 0x00000000 }, + { 0x000384, 4, 0x20, 0x00000000 }, + { 0x000388, 4, 0x20, 0x00000000 }, + { 0x00038c, 4, 0x20, 0x00000000 }, + { 0x000700, 4, 0x10, 0x00000000 }, + { 0x000704, 4, 0x10, 0x00000000 }, + { 0x000708, 4, 0x10, 0x00000000 }, + { 0x002800, 128, 0x04, 0x00000000 }, + { 0x000a00, 16, 0x20, 0x00000000 }, + { 0x000a04, 16, 0x20, 0x00000000 }, + { 0x000a08, 16, 0x20, 0x00000000 }, + { 0x000a0c, 16, 0x20, 0x00000000 }, + { 0x000a10, 16, 0x20, 0x00000000 }, + { 0x000a14, 16, 0x20, 0x00000000 }, + { 0x000c00, 16, 0x10, 0x00000000 }, + { 0x000c04, 16, 0x10, 0x00000000 }, + { 0x000c08, 16, 0x10, 0x00000000 }, + { 0x000c0c, 16, 0x10, 0x3f800000 }, + { 0x000d00, 8, 0x08, 0xffff0000 }, + { 0x000d04, 8, 0x08, 0xffff0000 }, + { 0x000e00, 16, 0x10, 0x00000000 }, + { 0x000e04, 16, 0x10, 0xffff0000 }, + { 0x000e08, 16, 0x10, 0xffff0000 }, + { 0x000d40, 4, 0x08, 0x00000000 }, + { 0x000d44, 4, 0x08, 0x00000000 }, + { 0x001e00, 8, 0x20, 0x00000001 }, + { 0x001e04, 8, 0x20, 0x00000001 }, + { 0x001e08, 8, 0x20, 0x00000002 }, + { 0x001e0c, 8, 0x20, 0x00000001 }, + { 0x001e10, 8, 0x20, 0x00000001 }, + { 0x001e14, 8, 0x20, 0x00000002 }, + { 0x001e18, 8, 0x20, 0x00000001 }, + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x001514, 1, 0x04, 0x00000000 }, + { 0x000d68, 1, 0x04, 0x0000ffff }, + { 0x00121c, 1, 0x04, 0x0fac6881 }, + { 0x000fac, 1, 0x04, 0x00000001 }, + { 0x001538, 1, 0x04, 0x00000001 }, + { 0x000fe0, 2, 0x04, 0x00000000 }, + { 0x000fe8, 1, 0x04, 0x00000014 }, + { 0x000fec, 1, 0x04, 0x00000040 }, + { 0x000ff0, 1, 0x04, 0x00000000 }, + { 0x00179c, 1, 0x04, 0x00000000 }, + { 0x001228, 1, 0x04, 0x00000400 }, + { 0x00122c, 1, 0x04, 0x00000300 }, + { 0x001230, 1, 0x04, 0x00010001 }, + { 0x0007f8, 1, 0x04, 0x00000000 }, + { 0x0015b4, 1, 0x04, 0x00000001 }, + { 0x0015cc, 1, 0x04, 0x00000000 }, + { 0x001534, 1, 0x04, 0x00000000 }, + { 0x000fb0, 1, 0x04, 0x00000000 }, + { 0x0015d0, 1, 0x04, 0x00000000 }, + { 0x00153c, 1, 0x04, 0x00000000 }, + { 0x0016b4, 1, 0x04, 0x00000003 }, + { 0x000fbc, 4, 0x04, 0x0000ffff }, + { 0x000df8, 2, 0x04, 0x00000000 }, + { 0x001948, 1, 0x04, 0x00000000 }, + { 0x001970, 1, 0x04, 0x00000001 }, + { 0x00161c, 1, 0x04, 0x000009f0 }, + { 0x000dcc, 1, 0x04, 0x00000010 }, + { 0x00163c, 1, 0x04, 0x00000000 }, + { 0x0015e4, 1, 0x04, 0x00000000 }, + { 0x001160, 32, 0x04, 0x25e00040 }, + { 0x001880, 32, 0x04, 0x00000000 }, + { 0x000f84, 2, 0x04, 0x00000000 }, + { 0x0017c8, 2, 0x04, 0x00000000 }, + { 0x0017d0, 1, 0x04, 0x000000ff }, + { 0x0017d4, 1, 0x04, 0xffffffff }, + { 0x0017d8, 1, 0x04, 0x00000002 }, + { 0x0017dc, 1, 0x04, 0x00000000 }, + { 0x0015f4, 2, 0x04, 0x00000000 }, + { 0x001434, 2, 0x04, 0x00000000 }, + { 0x000d74, 1, 0x04, 0x00000000 }, + { 0x000dec, 1, 0x04, 0x00000001 }, + { 0x0013a4, 1, 0x04, 0x00000000 }, + { 0x001318, 1, 0x04, 0x00000001 }, + { 0x001644, 1, 0x04, 0x00000000 }, + { 0x000748, 1, 0x04, 0x00000000 }, + { 0x000de8, 1, 0x04, 0x00000000 }, + { 0x001648, 1, 0x04, 0x00000000 }, + { 0x0012a4, 1, 0x04, 0x00000000 }, + { 0x001120, 4, 0x04, 0x00000000 }, + { 0x001118, 1, 0x04, 0x00000000 }, + { 0x00164c, 1, 0x04, 0x00000000 }, + { 0x001658, 1, 0x04, 0x00000000 }, + { 0x001910, 1, 0x04, 0x00000290 }, + { 0x001518, 1, 0x04, 0x00000000 }, + { 0x00165c, 1, 0x04, 0x00000001 }, + { 0x001520, 1, 0x04, 0x00000000 }, + { 0x001604, 1, 0x04, 0x00000000 }, + { 0x001570, 1, 0x04, 0x00000000 }, + { 0x0013b0, 2, 0x04, 0x3f800000 }, + { 0x00020c, 1, 0x04, 0x00000000 }, + { 0x001670, 1, 0x04, 0x30201000 }, + { 0x001674, 1, 0x04, 0x70605040 }, + { 0x001678, 1, 0x04, 0xb8a89888 }, + { 0x00167c, 1, 0x04, 0xf8e8d8c8 }, + { 0x00166c, 1, 0x04, 0x00000000 }, + { 0x001680, 1, 0x04, 0x00ffff00 }, + { 0x0012d0, 1, 0x04, 0x00000003 }, + { 0x0012d4, 1, 0x04, 0x00000002 }, + { 0x001684, 2, 0x04, 0x00000000 }, + { 0x000dac, 2, 0x04, 0x00001b02 }, + { 0x000db4, 1, 0x04, 0x00000000 }, + { 0x00168c, 1, 0x04, 0x00000000 }, + { 0x0015bc, 1, 0x04, 0x00000000 }, + { 0x00156c, 1, 0x04, 0x00000000 }, + { 0x00187c, 1, 0x04, 0x00000000 }, + { 0x001110, 1, 0x04, 0x00000001 }, + { 0x000dc0, 3, 0x04, 0x00000000 }, + { 0x001234, 1, 0x04, 0x00000000 }, + { 0x001690, 1, 0x04, 0x00000000 }, + { 0x0012ac, 1, 0x04, 0x00000001 }, + { 0x0002c4, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x001000, 1, 0x04, 0x00000010 }, + { 0x0010fc, 1, 0x04, 0x00000000 }, + { 0x001290, 1, 0x04, 0x00000000 }, + { 0x000218, 1, 0x04, 0x00000010 }, + { 0x0012d8, 1, 0x04, 0x00000000 }, + { 0x0012dc, 1, 0x04, 0x00000010 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x00155c, 2, 0x04, 0x00000000 }, + { 0x001564, 1, 0x04, 0x00000fff }, + { 0x001574, 2, 0x04, 0x00000000 }, + { 0x00157c, 1, 0x04, 0x000fffff }, + { 0x001354, 1, 0x04, 0x00000000 }, + { 0x001610, 1, 0x04, 0x00000012 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x00260c, 1, 0x04, 0x00000000 }, + { 0x0007ac, 1, 0x04, 0x00000000 }, + { 0x00162c, 1, 0x04, 0x00000003 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000324, 6, 0x04, 0x3f800000 }, + { 0x000750, 1, 0x04, 0x00000000 }, + { 0x000760, 1, 0x04, 0x39291909 }, + { 0x000764, 1, 0x04, 0x79695949 }, + { 0x000768, 1, 0x04, 0xb9a99989 }, + { 0x00076c, 1, 0x04, 0xf9e9d9c9 }, + { 0x000770, 1, 0x04, 0x30201000 }, + { 0x000774, 1, 0x04, 0x70605040 }, + { 0x000778, 1, 0x04, 0x00009080 }, + { 0x000780, 1, 0x04, 0x39291909 }, + { 0x000784, 1, 0x04, 0x79695949 }, + { 0x000788, 1, 0x04, 0xb9a99989 }, + { 0x00078c, 1, 0x04, 0xf9e9d9c9 }, + { 0x0007d0, 1, 0x04, 0x30201000 }, + { 0x0007d4, 1, 0x04, 0x70605040 }, + { 0x0007d8, 1, 0x04, 0x00009080 }, + { 0x00037c, 1, 0x04, 0x00000001 }, + { 0x000740, 2, 0x04, 0x00000000 }, + { 0x002600, 1, 0x04, 0x00000000 }, + { 0x001918, 1, 0x04, 0x00000000 }, + { 0x00191c, 1, 0x04, 0x00000900 }, + { 0x001920, 1, 0x04, 0x00000405 }, + { 0x001308, 1, 0x04, 0x00000001 }, + { 0x001924, 1, 0x04, 0x00000000 }, + { 0x0013ac, 1, 0x04, 0x00000000 }, + { 0x00192c, 1, 0x04, 0x00000001 }, + { 0x00193c, 1, 0x04, 0x00002c1c }, + { 0x000d7c, 1, 0x04, 0x00000000 }, + { 0x000f8c, 1, 0x04, 0x00000000 }, + { 0x0002c0, 1, 0x04, 0x00000001 }, + { 0x001510, 1, 0x04, 0x00000000 }, + { 0x001940, 1, 0x04, 0x00000000 }, + { 0x000ff4, 2, 0x04, 0x00000000 }, + { 0x00194c, 2, 0x04, 0x00000000 }, + { 0x001968, 1, 0x04, 0x00000000 }, + { 0x001590, 1, 0x04, 0x0000003f }, + { 0x0007e8, 4, 0x04, 0x00000000 }, + { 0x00196c, 1, 0x04, 0x00000011 }, + { 0x0002e4, 1, 0x04, 0x0000b001 }, + { 0x00036c, 2, 0x04, 0x00000000 }, + { 0x00197c, 1, 0x04, 0x00000000 }, + { 0x000fcc, 2, 0x04, 0x00000000 }, + { 0x0002d8, 1, 0x04, 0x00000040 }, + { 0x001980, 1, 0x04, 0x00000080 }, + { 0x001504, 1, 0x04, 0x00000080 }, + { 0x001984, 1, 0x04, 0x00000000 }, + { 0x000300, 1, 0x04, 0x00000001 }, + { 0x0013a8, 1, 0x04, 0x00000000 }, + { 0x0012ec, 1, 0x04, 0x00000000 }, + { 0x001310, 1, 0x04, 0x00000000 }, + { 0x001314, 1, 0x04, 0x00000001 }, + { 0x001380, 1, 0x04, 0x00000000 }, + { 0x001384, 4, 0x04, 0x00000001 }, + { 0x001394, 1, 0x04, 0x00000000 }, + { 0x00139c, 1, 0x04, 0x00000000 }, + { 0x001398, 1, 0x04, 0x00000000 }, + { 0x001594, 1, 0x04, 0x00000000 }, + { 0x001598, 4, 0x04, 0x00000001 }, + { 0x000f54, 3, 0x04, 0x00000000 }, + { 0x0019bc, 1, 0x04, 0x00000000 }, + { 0x000f9c, 2, 0x04, 0x00000000 }, + { 0x0012cc, 1, 0x04, 0x00000000 }, + { 0x0012e8, 1, 0x04, 0x00000000 }, + { 0x00130c, 1, 0x04, 0x00000001 }, + { 0x001360, 8, 0x04, 0x00000000 }, + { 0x00133c, 2, 0x04, 0x00000001 }, + { 0x001344, 1, 0x04, 0x00000002 }, + { 0x001348, 2, 0x04, 0x00000001 }, + { 0x001350, 1, 0x04, 0x00000002 }, + { 0x001358, 1, 0x04, 0x00000001 }, + { 0x0012e4, 1, 0x04, 0x00000000 }, + { 0x00131c, 4, 0x04, 0x00000000 }, + { 0x0019c0, 1, 0x04, 0x00000000 }, + { 0x001140, 1, 0x04, 0x00000000 }, + { 0x0019c4, 1, 0x04, 0x00000000 }, + { 0x0019c8, 1, 0x04, 0x00001500 }, + { 0x00135c, 1, 0x04, 0x00000000 }, + { 0x000f90, 1, 0x04, 0x00000000 }, + { 0x0019e0, 8, 0x04, 0x00000001 }, + { 0x0019cc, 1, 0x04, 0x00000001 }, + { 0x0015b8, 1, 0x04, 0x00000000 }, + { 0x001a00, 1, 0x04, 0x00001111 }, + { 0x001a04, 7, 0x04, 0x00000000 }, + { 0x000d6c, 2, 0x04, 0xffff0000 }, + { 0x0010f8, 1, 0x04, 0x00001010 }, + { 0x000d80, 5, 0x04, 0x00000000 }, + { 0x000da0, 1, 0x04, 0x00000000 }, + { 0x0007a4, 2, 0x04, 0x00000000 }, + { 0x001508, 1, 0x04, 0x80000000 }, + { 0x00150c, 1, 0x04, 0x40000000 }, + { 0x001668, 1, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000008 }, + { 0x000d9c, 1, 0x04, 0x00000001 }, + { 0x000ddc, 1, 0x04, 0x00000002 }, + { 0x000374, 1, 0x04, 0x00000000 }, + { 0x000378, 1, 0x04, 0x00000020 }, + { 0x0007dc, 1, 0x04, 0x00000000 }, + { 0x00074c, 1, 0x04, 0x00000055 }, + { 0x001420, 1, 0x04, 0x00000003 }, + { 0x0017bc, 2, 0x04, 0x00000000 }, + { 0x0017c4, 1, 0x04, 0x00000001 }, + { 0x001008, 1, 0x04, 0x00000008 }, + { 0x00100c, 1, 0x04, 0x00000040 }, + { 0x001010, 1, 0x04, 0x0000012c }, + { 0x000d60, 1, 0x04, 0x00000040 }, + { 0x00075c, 1, 0x04, 0x00000003 }, + { 0x001018, 1, 0x04, 0x00000020 }, + { 0x00101c, 1, 0x04, 0x00000001 }, + { 0x001020, 1, 0x04, 0x00000020 }, + { 0x001024, 1, 0x04, 0x00000001 }, + { 0x001444, 3, 0x04, 0x00000000 }, + { 0x000360, 1, 0x04, 0x20164010 }, + { 0x000364, 1, 0x04, 0x00000020 }, + { 0x000368, 1, 0x04, 0x00000000 }, + { 0x000de4, 1, 0x04, 0x00000000 }, + { 0x000204, 1, 0x04, 0x00000006 }, + { 0x000208, 1, 0x04, 0x00000000 }, + { 0x0002cc, 2, 0x04, 0x003fffff }, + { 0x001220, 1, 0x04, 0x00000005 }, + { 0x000fdc, 1, 0x04, 0x00000000 }, + { 0x000f98, 1, 0x04, 0x00400008 }, + { 0x001284, 1, 0x04, 0x08000080 }, + { 0x001450, 1, 0x04, 0x00400008 }, + { 0x001454, 1, 0x04, 0x08000080 }, + { 0x000214, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_mthd[] = { + { gk110_grctx_init_a197_0, 0xa197 }, + { gf100_grctx_init_902d_0, 0x902d }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_fe_0[] = { + { 0x404004, 8, 0x04, 0x00000000 }, + { 0x404024, 1, 0x04, 0x0000e000 }, + { 0x404028, 8, 0x04, 0x00000000 }, + { 0x4040a8, 8, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf800008f }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040e8, 1, 0x04, 0x00001000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404100, 10, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404138, 1, 0x04, 0x20000040 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000055 }, + { 0x40417c, 2, 0x04, 0x00000000 }, + { 0x4041a0, 4, 0x04, 0x00000000 }, + { 0x404200, 1, 0x04, 0x0000a197 }, + { 0x404204, 1, 0x04, 0x0000a1c0 }, + { 0x404208, 1, 0x04, 0x0000a140 }, + { 0x40420c, 1, 0x04, 0x0000902d }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_pri_0[] = { + { 0x404404, 12, 0x04, 0x00000000 }, + { 0x404438, 1, 0x04, 0x00000000 }, + { 0x404460, 2, 0x04, 0x00000000 }, + { 0x404468, 1, 0x04, 0x00ffffff }, + { 0x40446c, 1, 0x04, 0x00000000 }, + { 0x404480, 1, 0x04, 0x00000001 }, + { 0x404498, 1, 0x04, 0x00000001 }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_cwd_0[] = { + { 0x405b00, 1, 0x04, 0x00000000 }, + { 0x405b10, 1, 0x04, 0x00001000 }, + { 0x405b20, 1, 0x04, 0x04000000 }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x034103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b0, 3, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x802000f0 }, + { 0x4064c4, 1, 0x04, 0x0192ffff }, + { 0x4064c8, 1, 0x04, 0x018007c0 }, + { 0x4064cc, 9, 0x04, 0x00000000 }, + { 0x4064fc, 1, 0x04, 0x0000022a }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x12802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1003e005 }, + { 0x408840, 1, 0x04, 0x0000000b }, + { 0x408900, 1, 0x04, 0x3080b801 }, + { 0x408904, 1, 0x04, 0x62000001 }, + { 0x408908, 1, 0x04, 0x00c8102f }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gk110_grctx_init_fe_0 }, + { gk110_grctx_init_pri_0 }, + { gk104_grctx_init_memfmt_0 }, + { gk104_grctx_init_ds_0 }, + { gk110_grctx_init_cwd_0 }, + { gk110_grctx_init_pd_0 }, + { gf100_grctx_init_rstr2d_0 }, + { gk104_grctx_init_scc_0 }, + { gk110_grctx_init_be_0 }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006860a }, + { 0x418808, 1, 0x04, 0x00000000 }, + { 0x41880c, 1, 0x04, 0x00000030 }, + { 0x418810, 1, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00000044 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100018 }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_gpc_unk_2[] = { + { 0x418d24, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gf119_grctx_init_prop_0 }, + { gf119_grctx_init_gpc_unk_1 }, + { gk110_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_gpc_1[] = { + { gf119_grctx_init_crstr_0 }, + { gk104_grctx_init_gpm_0 }, + { gk110_grctx_init_gpc_unk_2 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000000f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000021 }, + { 0x419a0c, 1, 0x04, 0x00020000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x0000c000 }, + { 0x419a20, 1, 0x04, 0x00020800 }, + { 0x419a30, 1, 0x04, 0x00000001 }, + { 0x419ac4, 1, 0x04, 0x0037f440 }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x0000001a }, + { 0x419c04, 1, 0x04, 0x80000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + { 0x419c24, 1, 0x04, 0x00084210 }, + { 0x419c28, 1, 0x04, 0x3efbefbe }, + {} +}; + +const struct gf100_gr_init +gk110_grctx_init_l1c_0[] = { + { 0x419ce8, 1, 0x04, 0x00000000 }, + { 0x419cf4, 1, 0x04, 0x00000203 }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_sm_0[] = { + { 0x419e04, 1, 0x04, 0x00000000 }, + { 0x419e08, 1, 0x04, 0x0000001d }, + { 0x419e0c, 1, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00001c02 }, + { 0x419e44, 1, 0x04, 0x0013eff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000007f }, + { 0x419e50, 2, 0x04, 0x00000000 }, + { 0x419e58, 1, 0x04, 0x00000001 }, + { 0x419e5c, 3, 0x04, 0x00000000 }, + { 0x419e68, 1, 0x04, 0x00000002 }, + { 0x419e6c, 12, 0x04, 0x00000000 }, + { 0x419eac, 1, 0x04, 0x00001f8f }, + { 0x419eb0, 1, 0x04, 0x0db00d2f }, + { 0x419eb8, 1, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x0001304f }, + { 0x419f30, 4, 0x04, 0x00000000 }, + { 0x419f40, 1, 0x04, 0x00000018 }, + { 0x419f44, 3, 0x04, 0x00000000 }, + { 0x419f58, 1, 0x04, 0x00000000 }, + { 0x419f70, 1, 0x04, 0x00007300 }, + { 0x419f78, 1, 0x04, 0x000000eb }, + { 0x419f7c, 1, 0x04, 0x00000404 }, + {} +}; + +static const struct gf100_gr_pack +gk110_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gk110_grctx_init_tex_0 }, + { gk110_grctx_init_mpc_0 }, + { gk110_grctx_init_l1c_0 }, + { gk110_grctx_init_sm_0 }, + {} +}; + +static const struct gf100_gr_init +gk110_grctx_init_cbm_0[] = { + { 0x41bec0, 1, 0x04, 0x10000000 }, + { 0x41bec4, 1, 0x04, 0x00037f7f }, + { 0x41bee4, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_pack +gk110_grctx_pack_ppc[] = { + { gk104_grctx_init_pes_0 }, + { gk110_grctx_init_cbm_0 }, + { gf117_grctx_init_wwdx_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gk110_grctx_generate_r419eb0(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419eb0, 0x00001000, 0x00001000); +} + +void +gk110_grctx_generate_r419f78(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + /* bit 3 set disables loads in fp helper invocations, we need it enabled */ + nvkm_mask(device, 0x419f78, 0x00000008, 0x00000000); +} + +const struct gf100_grctx_func +gk110_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gk110_grctx_pack_hub, + .gpc_0 = gk110_grctx_pack_gpc_0, + .gpc_1 = gk110_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gk110_grctx_pack_tpc, + .ppc = gk110_grctx_pack_ppc, + .icmd = gk110_grctx_pack_icmd, + .mthd = gk110_grctx_pack_mthd, + .bundle = gk104_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x7c0, + .pagepool = gk104_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x7ff, + .alpha_nr = 0x648, + .patch_ltc = gk104_grctx_generate_patch_ltc, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .r418800 = gk104_grctx_generate_r418800, + .r419eb0 = gk110_grctx_generate_r419eb0, + .r419f78 = gk110_grctx_generate_r419f78, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c new file mode 100644 index 0000000000..5597e87624 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c @@ -0,0 +1,107 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk110b_grctx_init_sm_0[] = { + { 0x419e04, 1, 0x04, 0x00000000 }, + { 0x419e08, 1, 0x04, 0x0000001d }, + { 0x419e0c, 1, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00001c02 }, + { 0x419e44, 1, 0x04, 0x0013eff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000007f }, + { 0x419e50, 2, 0x04, 0x00000000 }, + { 0x419e58, 1, 0x04, 0x00000001 }, + { 0x419e5c, 3, 0x04, 0x00000000 }, + { 0x419e68, 1, 0x04, 0x00000002 }, + { 0x419e6c, 12, 0x04, 0x00000000 }, + { 0x419eac, 1, 0x04, 0x00001f8f }, + { 0x419eb0, 1, 0x04, 0x0db00d2f }, + { 0x419eb8, 1, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x0001304f }, + { 0x419f30, 4, 0x04, 0x00000000 }, + { 0x419f40, 1, 0x04, 0x00000018 }, + { 0x419f44, 3, 0x04, 0x00000000 }, + { 0x419f58, 1, 0x04, 0x00000000 }, + { 0x419f70, 1, 0x04, 0x00006300 }, + { 0x419f78, 1, 0x04, 0x000000eb }, + { 0x419f7c, 1, 0x04, 0x00000404 }, + {} +}; + +static const struct gf100_gr_pack +gk110b_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gk110_grctx_init_tex_0 }, + { gk110_grctx_init_mpc_0 }, + { gk110_grctx_init_l1c_0 }, + { gk110b_grctx_init_sm_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gk110b_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gk110_grctx_pack_hub, + .gpc_0 = gk110_grctx_pack_gpc_0, + .gpc_1 = gk110_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gk110b_grctx_pack_tpc, + .ppc = gk110_grctx_pack_ppc, + .icmd = gk110_grctx_pack_icmd, + .mthd = gk110_grctx_pack_mthd, + .bundle = gk104_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x600, + .pagepool = gk104_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x7ff, + .alpha_nr = 0x648, + .patch_ltc = gk104_grctx_generate_patch_ltc, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .r418800 = gk104_grctx_generate_r418800, + .r419eb0 = gk110_grctx_generate_r419eb0, + .r419f78 = gk110_grctx_generate_r419f78, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c new file mode 100644 index 0000000000..6126564965 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c @@ -0,0 +1,572 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk208_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000b1, 1, 0x01, 0x00000001 }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002ec, 1, 0x01, 0x00000001 }, + { 0x0002f2, 2, 0x01, 0x00000001 }, + { 0x0002f5, 1, 0x01, 0x00000001 }, + { 0x0002f7, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x000375, 1, 0x01, 0x00000001 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000383, 1, 0x01, 0x00000011 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000662, 1, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x00080b, 1, 0x01, 0x00000002 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000937, 1, 0x01, 0x00000001 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000979, 1, 0x01, 0x00000003 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x000a0d, 1, 0x01, 0x00000006 }, + { 0x00097d, 1, 0x01, 0x00000020 }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000685, 1, 0x01, 0x003fffff }, + { 0x000687, 1, 0x01, 0x003fffff }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00400008 }, + { 0x000841, 1, 0x01, 0x08000080 }, + { 0x000842, 1, 0x01, 0x00400008 }, + { 0x000843, 1, 0x01, 0x08000080 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000008 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x00c500, 1, 0x01, 0x00000003 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_icmd[] = { + { gk208_grctx_init_icmd_0 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_fe_0[] = { + { 0x404004, 8, 0x04, 0x00000000 }, + { 0x404024, 1, 0x04, 0x0000e000 }, + { 0x404028, 8, 0x04, 0x00000000 }, + { 0x4040a8, 8, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf800008f }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040e8, 1, 0x04, 0x00001000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404100, 10, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404138, 1, 0x04, 0x20000040 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000055 }, + { 0x40417c, 2, 0x04, 0x00000000 }, + { 0x404194, 1, 0x04, 0x01000700 }, + { 0x4041a0, 4, 0x04, 0x00000000 }, + { 0x404200, 1, 0x04, 0x0000a197 }, + { 0x404204, 1, 0x04, 0x0000a1c0 }, + { 0x404208, 1, 0x04, 0x0000a140 }, + { 0x40420c, 1, 0x04, 0x0000902d }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8000bf }, + { 0x405830, 1, 0x04, 0x02180648 }, + { 0x405834, 1, 0x04, 0x08000000 }, + { 0x405838, 1, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + { 0x405a1c, 1, 0x04, 0x000000ff }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x034103c1 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b0, 3, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x802000f0 }, + { 0x4064c4, 1, 0x04, 0x0192ffff }, + { 0x4064c8, 1, 0x04, 0x00c20200 }, + { 0x4064cc, 9, 0x04, 0x00000000 }, + { 0x4064fc, 1, 0x04, 0x0000022a }, + {} +}; + +const struct gf100_gr_init +gk208_grctx_init_rstr2d_0[] = { + { 0x407804, 1, 0x04, 0x00000063 }, + { 0x40780c, 1, 0x04, 0x0a418820 }, + { 0x407810, 1, 0x04, 0x062080e6 }, + { 0x407814, 1, 0x04, 0x020398a4 }, + { 0x407818, 1, 0x04, 0x0e629062 }, + { 0x40781c, 1, 0x04, 0x0a418820 }, + { 0x407820, 1, 0x04, 0x000000e6 }, + { 0x4078bc, 1, 0x04, 0x00000103 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x32802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1003e005 }, + { 0x408840, 1, 0x04, 0x0000000b }, + { 0x408900, 1, 0x04, 0xb080b801 }, + { 0x408904, 1, 0x04, 0x62000001 }, + { 0x408908, 1, 0x04, 0x02c8102f }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gk208_grctx_init_fe_0 }, + { gk110_grctx_init_pri_0 }, + { gk104_grctx_init_memfmt_0 }, + { gk208_grctx_init_ds_0 }, + { gk110_grctx_init_cwd_0 }, + { gk208_grctx_init_pd_0 }, + { gk208_grctx_init_rstr2d_0 }, + { gk104_grctx_init_scc_0 }, + { gk208_grctx_init_be_0 }, + {} +}; + +const struct gf100_gr_init +gk208_grctx_init_prop_0[] = { + { 0x418400, 1, 0x04, 0x38005e00 }, + { 0x418404, 1, 0x04, 0x71e0ffff }, + { 0x41840c, 1, 0x04, 0x00001008 }, + { 0x418410, 1, 0x04, 0x0fff0fff }, + { 0x418414, 1, 0x04, 0x02200fff }, + { 0x418450, 6, 0x04, 0x00000000 }, + { 0x418468, 1, 0x04, 0x00000001 }, + { 0x41846c, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_gpc_unk_1[] = { + { 0x418600, 1, 0x04, 0x0000007f }, + { 0x418684, 1, 0x04, 0x0000001f }, + { 0x418700, 1, 0x04, 0x00000002 }, + { 0x418704, 2, 0x04, 0x00000080 }, + { 0x41870c, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006863a }, + { 0x418808, 1, 0x04, 0x00000000 }, + { 0x41880c, 1, 0x04, 0x00000030 }, + { 0x418810, 1, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00000044 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100058 }, + {} +}; + +const struct gf100_gr_init +gk208_grctx_init_crstr_0[] = { + { 0x418b00, 1, 0x04, 0x0000001e }, + { 0x418b08, 1, 0x04, 0x0a418820 }, + { 0x418b0c, 1, 0x04, 0x062080e6 }, + { 0x418b10, 1, 0x04, 0x020398a4 }, + { 0x418b14, 1, 0x04, 0x0e629062 }, + { 0x418b18, 1, 0x04, 0x0a418820 }, + { 0x418b1c, 1, 0x04, 0x000000e6 }, + { 0x418bb8, 1, 0x04, 0x00000103 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_gpm_0[] = { + { 0x418c08, 1, 0x04, 0x00000001 }, + { 0x418c10, 8, 0x04, 0x00000000 }, + { 0x418c40, 1, 0x04, 0xffffffff }, + { 0x418c6c, 1, 0x04, 0x00000001 }, + { 0x418c80, 1, 0x04, 0x2020000c }, + { 0x418c8c, 1, 0x04, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_gpc_0[] = { + { gf100_grctx_init_gpc_unk_0 }, + { gk208_grctx_init_prop_0 }, + { gk208_grctx_init_gpc_unk_1 }, + { gk208_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_gpc_1[] = { + { gk208_grctx_init_crstr_0 }, + { gk208_grctx_init_gpm_0 }, + { gk110_grctx_init_gpc_unk_2 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000100f0 }, + { 0x419a04, 1, 0x04, 0x00000001 }, + { 0x419a08, 1, 0x04, 0x00000421 }, + { 0x419a0c, 1, 0x04, 0x00120000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00000200 }, + { 0x419a1c, 1, 0x04, 0x0000c000 }, + { 0x419a20, 1, 0x04, 0x00000800 }, + { 0x419a30, 1, 0x04, 0x00000001 }, + { 0x419ac4, 1, 0x04, 0x0037f440 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_sm_0[] = { + { 0x419e04, 1, 0x04, 0x00000000 }, + { 0x419e08, 1, 0x04, 0x0000001d }, + { 0x419e0c, 1, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00001c02 }, + { 0x419e44, 1, 0x04, 0x0013eff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000007f }, + { 0x419e50, 2, 0x04, 0x00000000 }, + { 0x419e58, 1, 0x04, 0x00000001 }, + { 0x419e5c, 3, 0x04, 0x00000000 }, + { 0x419e68, 1, 0x04, 0x00000002 }, + { 0x419e6c, 12, 0x04, 0x00000000 }, + { 0x419eac, 1, 0x04, 0x00001f8f }, + { 0x419eb0, 1, 0x04, 0x0db00d2f }, + { 0x419eb8, 1, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x0001304f }, + { 0x419f30, 4, 0x04, 0x00000000 }, + { 0x419f40, 1, 0x04, 0x00000018 }, + { 0x419f44, 3, 0x04, 0x00000000 }, + { 0x419f58, 1, 0x04, 0x00000020 }, + { 0x419f70, 1, 0x04, 0x00000000 }, + { 0x419f78, 1, 0x04, 0x000001eb }, + { 0x419f7c, 1, 0x04, 0x00000404 }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gk208_grctx_init_tex_0 }, + { gk110_grctx_init_mpc_0 }, + { gk110_grctx_init_l1c_0 }, + { gk208_grctx_init_sm_0 }, + {} +}; + +static const struct gf100_gr_init +gk208_grctx_init_cbm_0[] = { + { 0x41bec0, 1, 0x04, 0x10000000 }, + { 0x41bec4, 1, 0x04, 0x00037f7f }, + { 0x41bee4, 1, 0x04, 0x00000000 }, + { 0x41bef0, 1, 0x04, 0x000003ff }, + {} +}; + +static const struct gf100_gr_pack +gk208_grctx_pack_ppc[] = { + { gk104_grctx_init_pes_0 }, + { gk208_grctx_init_cbm_0 }, + { gf117_grctx_init_wwdx_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gk208_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gk208_grctx_pack_hub, + .gpc_0 = gk208_grctx_pack_gpc_0, + .gpc_1 = gk208_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gk208_grctx_pack_tpc, + .ppc = gk208_grctx_pack_ppc, + .icmd = gk208_grctx_pack_icmd, + .mthd = gk110_grctx_pack_mthd, + .bundle = gk104_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0xc2, + .bundle_token_limit = 0x200, + .pagepool = gk104_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x324, + .attrib_nr = 0x218, + .alpha_nr_max = 0x7ff, + .alpha_nr = 0x648, + .patch_ltc = gk104_grctx_generate_patch_ltc, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .r418800 = gk104_grctx_generate_r418800, + .r419f78 = gk110_grctx_generate_r419f78, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c new file mode 100644 index 0000000000..ac5fdcb5cd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c @@ -0,0 +1,90 @@ +/* + * Copyright (c) 2014-2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" +#include "gf100.h" + +#include + +static void +gk20a_grctx_generate_main(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + u32 idle_timeout; + int i; + + gf100_gr_mmio(gr, gr->sw_ctx); + + gf100_gr_wait_idle(gr); + + idle_timeout = nvkm_mask(device, 0x404154, 0xffffffff, 0x00000000); + + grctx->attrib_cb(chan, chan->attrib_cb->addr, grctx->attrib_cb_size(gr)); + grctx->attrib(chan); + + grctx->unkn(gr); + + gf100_grctx_generate_floorsweep(gr); + + for (i = 0; i < 8; i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); + + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); + + nvkm_mask(device, 0x5044b0, 0x08000000, 0x08000000); + + gf100_gr_wait_idle(gr); + + nvkm_wr32(device, 0x404154, idle_timeout); + gf100_gr_wait_idle(gr); + + gf100_gr_mthd(gr, gr->method); + gf100_gr_wait_idle(gr); + + gf100_gr_icmd(gr, gr->bundle); + grctx->pagepool(chan, chan->pagepool->addr); + grctx->bundle(chan, chan->bundle_cb->addr, grctx->bundle_size); +} + +const struct gf100_grctx_func +gk20a_grctx = { + .main = gk20a_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gk104_grctx_generate_bundle, + .bundle_size = 0x1800, + .bundle_min_gpm_fifo_depth = 0x62, + .bundle_token_limit = 0x100, + .pagepool = gk104_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gf100_grctx_generate_attrib_cb, + .attrib = gf117_grctx_generate_attrib, + .attrib_nr_max = 0x240, + .attrib_nr = 0x240, + .alpha_nr_max = 0x648 + (0x648 / 2), + .alpha_nr = 0x648, + .sm_id = gf100_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c new file mode 100644 index 0000000000..9906974ac3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c @@ -0,0 +1,992 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include +#include + +/******************************************************************************* + * PGRAPH context register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gm107_grctx_init_icmd_0[] = { + { 0x001000, 1, 0x01, 0x00000004 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x0000a9, 1, 0x01, 0x0000ffff }, + { 0x000038, 1, 0x01, 0x0fac6881 }, + { 0x00003d, 1, 0x01, 0x00000001 }, + { 0x0000e8, 8, 0x01, 0x00000400 }, + { 0x000078, 8, 0x01, 0x00000300 }, + { 0x000050, 1, 0x01, 0x00000011 }, + { 0x000058, 8, 0x01, 0x00000008 }, + { 0x000208, 8, 0x01, 0x00000001 }, + { 0x000081, 1, 0x01, 0x00000001 }, + { 0x000085, 1, 0x01, 0x00000004 }, + { 0x000088, 1, 0x01, 0x00000400 }, + { 0x000090, 1, 0x01, 0x00000300 }, + { 0x000098, 1, 0x01, 0x00001001 }, + { 0x0000e3, 1, 0x01, 0x00000001 }, + { 0x0000da, 1, 0x01, 0x00000001 }, + { 0x0000f8, 1, 0x01, 0x00000003 }, + { 0x0000fa, 1, 0x01, 0x00000001 }, + { 0x0000b1, 2, 0x01, 0x00000001 }, + { 0x00009f, 4, 0x01, 0x0000ffff }, + { 0x0000a8, 1, 0x01, 0x0000ffff }, + { 0x0000ad, 1, 0x01, 0x0000013e }, + { 0x0000e1, 1, 0x01, 0x00000010 }, + { 0x000290, 16, 0x01, 0x00000000 }, + { 0x0003b0, 16, 0x01, 0x00000000 }, + { 0x0002a0, 16, 0x01, 0x00000000 }, + { 0x000420, 16, 0x01, 0x00000000 }, + { 0x0002b0, 16, 0x01, 0x00000000 }, + { 0x000430, 16, 0x01, 0x00000000 }, + { 0x0002c0, 16, 0x01, 0x00000000 }, + { 0x0004d0, 16, 0x01, 0x00000000 }, + { 0x000720, 16, 0x01, 0x00000000 }, + { 0x0008c0, 16, 0x01, 0x00000000 }, + { 0x000890, 16, 0x01, 0x00000000 }, + { 0x0008e0, 16, 0x01, 0x00000000 }, + { 0x0008a0, 16, 0x01, 0x00000000 }, + { 0x0008f0, 16, 0x01, 0x00000000 }, + { 0x00094c, 1, 0x01, 0x000000ff }, + { 0x00094d, 1, 0x01, 0xffffffff }, + { 0x00094e, 1, 0x01, 0x00000002 }, + { 0x0002f2, 2, 0x01, 0x00000001 }, + { 0x0002f5, 1, 0x01, 0x00000001 }, + { 0x0002f7, 1, 0x01, 0x00000001 }, + { 0x000303, 1, 0x01, 0x00000001 }, + { 0x0002e6, 1, 0x01, 0x00000001 }, + { 0x000466, 1, 0x01, 0x00000052 }, + { 0x000301, 1, 0x01, 0x3f800000 }, + { 0x000304, 1, 0x01, 0x30201000 }, + { 0x000305, 1, 0x01, 0x70605040 }, + { 0x000306, 1, 0x01, 0xb8a89888 }, + { 0x000307, 1, 0x01, 0xf8e8d8c8 }, + { 0x00030a, 1, 0x01, 0x00ffff00 }, + { 0x0000de, 1, 0x01, 0x00000001 }, + { 0x00030b, 1, 0x01, 0x0000001a }, + { 0x00030c, 1, 0x01, 0x00000001 }, + { 0x000318, 1, 0x01, 0x00000001 }, + { 0x000340, 1, 0x01, 0x00000000 }, + { 0x00037d, 1, 0x01, 0x00000006 }, + { 0x0003a0, 1, 0x01, 0x00000002 }, + { 0x0003aa, 1, 0x01, 0x00000001 }, + { 0x0003a9, 1, 0x01, 0x00000001 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000383, 1, 0x01, 0x00000011 }, + { 0x000360, 1, 0x01, 0x00000040 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x00037a, 1, 0x01, 0x00000012 }, + { 0x000619, 1, 0x01, 0x00000003 }, + { 0x000811, 1, 0x01, 0x00000003 }, + { 0x000812, 1, 0x01, 0x00000004 }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000815, 1, 0x01, 0x0000000b }, + { 0x000800, 6, 0x01, 0x00000001 }, + { 0x000632, 1, 0x01, 0x00000001 }, + { 0x000633, 1, 0x01, 0x00000002 }, + { 0x000634, 1, 0x01, 0x00000003 }, + { 0x000635, 1, 0x01, 0x00000004 }, + { 0x000654, 1, 0x01, 0x3f800000 }, + { 0x000657, 1, 0x01, 0x3f800000 }, + { 0x000655, 2, 0x01, 0x3f800000 }, + { 0x0006cd, 1, 0x01, 0x3f800000 }, + { 0x0007f5, 1, 0x01, 0x3f800000 }, + { 0x0007dc, 1, 0x01, 0x39291909 }, + { 0x0007dd, 1, 0x01, 0x79695949 }, + { 0x0007de, 1, 0x01, 0xb9a99989 }, + { 0x0007df, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007e8, 1, 0x01, 0x00003210 }, + { 0x0007e9, 1, 0x01, 0x00007654 }, + { 0x0007ea, 1, 0x01, 0x00000098 }, + { 0x0007ec, 1, 0x01, 0x39291909 }, + { 0x0007ed, 1, 0x01, 0x79695949 }, + { 0x0007ee, 1, 0x01, 0xb9a99989 }, + { 0x0007ef, 1, 0x01, 0xf9e9d9c9 }, + { 0x0007f0, 1, 0x01, 0x00003210 }, + { 0x0007f1, 1, 0x01, 0x00007654 }, + { 0x0007f2, 1, 0x01, 0x00000098 }, + { 0x0005a5, 1, 0x01, 0x00000001 }, + { 0x0005d0, 1, 0x01, 0x20181008 }, + { 0x0005d1, 1, 0x01, 0x40383028 }, + { 0x0005d2, 1, 0x01, 0x60585048 }, + { 0x0005d3, 1, 0x01, 0x80787068 }, + { 0x000980, 128, 0x01, 0x00000000 }, + { 0x000468, 1, 0x01, 0x00000004 }, + { 0x00046c, 1, 0x01, 0x00000001 }, + { 0x000470, 96, 0x01, 0x00000000 }, + { 0x000510, 16, 0x01, 0x3f800000 }, + { 0x000520, 1, 0x01, 0x000002b6 }, + { 0x000529, 1, 0x01, 0x00000001 }, + { 0x000530, 16, 0x01, 0xffff0000 }, + { 0x000550, 32, 0x01, 0xffff0000 }, + { 0x000585, 1, 0x01, 0x0000003f }, + { 0x000576, 1, 0x01, 0x00000003 }, + { 0x00057b, 1, 0x01, 0x00000059 }, + { 0x000586, 1, 0x01, 0x00000040 }, + { 0x000582, 2, 0x01, 0x00000080 }, + { 0x000595, 1, 0x01, 0x00400040 }, + { 0x000596, 1, 0x01, 0x00000492 }, + { 0x000597, 1, 0x01, 0x08080203 }, + { 0x0005ad, 1, 0x01, 0x00000008 }, + { 0x000598, 1, 0x01, 0x00020001 }, + { 0x0005c2, 1, 0x01, 0x00000001 }, + { 0x000638, 2, 0x01, 0x00000001 }, + { 0x00063a, 1, 0x01, 0x00000002 }, + { 0x00063b, 2, 0x01, 0x00000001 }, + { 0x00063d, 1, 0x01, 0x00000002 }, + { 0x00063e, 1, 0x01, 0x00000001 }, + { 0x0008b8, 8, 0x01, 0x00000001 }, + { 0x000900, 8, 0x01, 0x00000001 }, + { 0x000908, 8, 0x01, 0x00000002 }, + { 0x000910, 16, 0x01, 0x00000001 }, + { 0x000920, 8, 0x01, 0x00000002 }, + { 0x000928, 8, 0x01, 0x00000001 }, + { 0x000662, 1, 0x01, 0x00000001 }, + { 0x000648, 9, 0x01, 0x00000001 }, + { 0x000658, 1, 0x01, 0x0000000f }, + { 0x0007ff, 1, 0x01, 0x0000000a }, + { 0x00066a, 1, 0x01, 0x40000000 }, + { 0x00066b, 1, 0x01, 0x10000000 }, + { 0x00066c, 2, 0x01, 0xffff0000 }, + { 0x0007af, 2, 0x01, 0x00000008 }, + { 0x0007f6, 1, 0x01, 0x00000001 }, + { 0x0006b2, 1, 0x01, 0x00000055 }, + { 0x0007ad, 1, 0x01, 0x00000003 }, + { 0x000971, 1, 0x01, 0x00000008 }, + { 0x000972, 1, 0x01, 0x00000040 }, + { 0x000973, 1, 0x01, 0x0000012c }, + { 0x00097c, 1, 0x01, 0x00000040 }, + { 0x000975, 1, 0x01, 0x00000020 }, + { 0x000976, 1, 0x01, 0x00000001 }, + { 0x000977, 1, 0x01, 0x00000020 }, + { 0x000978, 1, 0x01, 0x00000001 }, + { 0x000957, 1, 0x01, 0x00000003 }, + { 0x00095e, 1, 0x01, 0x20164010 }, + { 0x00095f, 1, 0x01, 0x00000020 }, + { 0x000a0d, 1, 0x01, 0x00000006 }, + { 0x00097d, 1, 0x01, 0x0000000c }, + { 0x000683, 1, 0x01, 0x00000006 }, + { 0x000687, 1, 0x01, 0x003fffff }, + { 0x0006a0, 1, 0x01, 0x00000005 }, + { 0x000840, 1, 0x01, 0x00400008 }, + { 0x000841, 1, 0x01, 0x08000080 }, + { 0x000842, 1, 0x01, 0x00400008 }, + { 0x000843, 1, 0x01, 0x08000080 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ab, 1, 0x01, 0x00000002 }, + { 0x0006ac, 1, 0x01, 0x00000080 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x0006bb, 1, 0x01, 0x000000cf }, + { 0x0006ce, 1, 0x01, 0x2a712488 }, + { 0x000739, 1, 0x01, 0x4085c000 }, + { 0x00073a, 1, 0x01, 0x00000080 }, + { 0x000786, 1, 0x01, 0x80000100 }, + { 0x00073c, 1, 0x01, 0x00010100 }, + { 0x00073d, 1, 0x01, 0x02800000 }, + { 0x000787, 1, 0x01, 0x000000cf }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x000836, 1, 0x01, 0x00000001 }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x000833, 1, 0x01, 0x04444480 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c1b0, 8, 0x01, 0x0000000f }, + { 0x00c1b8, 1, 0x01, 0x0fac6881 }, + { 0x00c1b9, 1, 0x01, 0x00fac688 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000002 }, + { 0x0006aa, 1, 0x01, 0x00000001 }, + { 0x0006ad, 2, 0x01, 0x00000100 }, + { 0x0006b1, 1, 0x01, 0x00000011 }, + { 0x00078c, 1, 0x01, 0x00000008 }, + { 0x000792, 1, 0x01, 0x00000001 }, + { 0x000794, 3, 0x01, 0x00000001 }, + { 0x000797, 1, 0x01, 0x000000cf }, + { 0x00079a, 1, 0x01, 0x00000002 }, + { 0x0007a1, 1, 0x01, 0x00000001 }, + { 0x0007a3, 3, 0x01, 0x00000001 }, + { 0x000831, 1, 0x01, 0x00000004 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000008 }, + { 0x000039, 3, 0x01, 0x00000000 }, + { 0x000380, 1, 0x01, 0x00000001 }, + { 0x000366, 2, 0x01, 0x00000000 }, + { 0x000368, 1, 0x01, 0x00000fff }, + { 0x000370, 2, 0x01, 0x00000000 }, + { 0x000372, 1, 0x01, 0x000fffff }, + { 0x000813, 1, 0x01, 0x00000006 }, + { 0x000814, 1, 0x01, 0x00000008 }, + { 0x000818, 8, 0x01, 0x00000000 }, + { 0x000848, 16, 0x01, 0x00000000 }, + { 0x000738, 1, 0x01, 0x00000000 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x000a04, 1, 0x01, 0x000000ff }, + { 0x000a0b, 1, 0x01, 0x00000040 }, + { 0x00097f, 1, 0x01, 0x00000100 }, + { 0x000a02, 1, 0x01, 0x00000001 }, + { 0x000809, 1, 0x01, 0x00000007 }, + { 0x00c221, 1, 0x01, 0x00000040 }, + { 0x00c401, 1, 0x01, 0x00000001 }, + { 0x00c402, 1, 0x01, 0x00010001 }, + { 0x00c403, 2, 0x01, 0x00000001 }, + { 0x00c40e, 1, 0x01, 0x00000020 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + { 0x001000, 1, 0x01, 0x00000001 }, + { 0x000b07, 1, 0x01, 0x00000002 }, + { 0x000b08, 2, 0x01, 0x00000100 }, + { 0x000b0a, 1, 0x01, 0x00000001 }, + { 0x01e100, 1, 0x01, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_icmd[] = { + { gm107_grctx_init_icmd_0 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_b097_0[] = { + { 0x000800, 8, 0x40, 0x00000000 }, + { 0x000804, 8, 0x40, 0x00000000 }, + { 0x000808, 8, 0x40, 0x00000400 }, + { 0x00080c, 8, 0x40, 0x00000300 }, + { 0x000810, 1, 0x04, 0x000000cf }, + { 0x000850, 7, 0x40, 0x00000000 }, + { 0x000814, 8, 0x40, 0x00000040 }, + { 0x000818, 8, 0x40, 0x00000001 }, + { 0x00081c, 8, 0x40, 0x00000000 }, + { 0x000820, 8, 0x40, 0x00000000 }, + { 0x001c00, 16, 0x10, 0x00000000 }, + { 0x001c04, 16, 0x10, 0x00000000 }, + { 0x001c08, 16, 0x10, 0x00000000 }, + { 0x001c0c, 16, 0x10, 0x00000000 }, + { 0x001d00, 16, 0x10, 0x00000000 }, + { 0x001d04, 16, 0x10, 0x00000000 }, + { 0x001d08, 16, 0x10, 0x00000000 }, + { 0x001d0c, 16, 0x10, 0x00000000 }, + { 0x001f00, 16, 0x08, 0x00000000 }, + { 0x001f04, 16, 0x08, 0x00000000 }, + { 0x001f80, 16, 0x08, 0x00000000 }, + { 0x001f84, 16, 0x08, 0x00000000 }, + { 0x002000, 1, 0x04, 0x00000000 }, + { 0x002040, 1, 0x04, 0x00000011 }, + { 0x002080, 1, 0x04, 0x00000020 }, + { 0x0020c0, 1, 0x04, 0x00000030 }, + { 0x002100, 1, 0x04, 0x00000040 }, + { 0x002140, 1, 0x04, 0x00000051 }, + { 0x00200c, 6, 0x40, 0x00000001 }, + { 0x002010, 1, 0x04, 0x00000000 }, + { 0x002050, 1, 0x04, 0x00000000 }, + { 0x002090, 1, 0x04, 0x00000001 }, + { 0x0020d0, 1, 0x04, 0x00000002 }, + { 0x002110, 1, 0x04, 0x00000003 }, + { 0x002150, 1, 0x04, 0x00000004 }, + { 0x000380, 4, 0x20, 0x00000000 }, + { 0x000384, 4, 0x20, 0x00000000 }, + { 0x000388, 4, 0x20, 0x00000000 }, + { 0x00038c, 4, 0x20, 0x00000000 }, + { 0x000700, 4, 0x10, 0x00000000 }, + { 0x000704, 4, 0x10, 0x00000000 }, + { 0x000708, 4, 0x10, 0x00000000 }, + { 0x002800, 128, 0x04, 0x00000000 }, + { 0x000a00, 16, 0x20, 0x00000000 }, + { 0x000a04, 16, 0x20, 0x00000000 }, + { 0x000a08, 16, 0x20, 0x00000000 }, + { 0x000a0c, 16, 0x20, 0x00000000 }, + { 0x000a10, 16, 0x20, 0x00000000 }, + { 0x000a14, 16, 0x20, 0x00000000 }, + { 0x000c00, 16, 0x10, 0x00000000 }, + { 0x000c04, 16, 0x10, 0x00000000 }, + { 0x000c08, 16, 0x10, 0x00000000 }, + { 0x000c0c, 16, 0x10, 0x3f800000 }, + { 0x000d00, 8, 0x08, 0xffff0000 }, + { 0x000d04, 8, 0x08, 0xffff0000 }, + { 0x000e00, 16, 0x10, 0x00000000 }, + { 0x000e04, 16, 0x10, 0xffff0000 }, + { 0x000e08, 16, 0x10, 0xffff0000 }, + { 0x000d40, 4, 0x08, 0x00000000 }, + { 0x000d44, 4, 0x08, 0x00000000 }, + { 0x001e00, 8, 0x20, 0x00000001 }, + { 0x001e04, 8, 0x20, 0x00000001 }, + { 0x001e08, 8, 0x20, 0x00000002 }, + { 0x001e0c, 8, 0x20, 0x00000001 }, + { 0x001e10, 8, 0x20, 0x00000001 }, + { 0x001e14, 8, 0x20, 0x00000002 }, + { 0x001e18, 8, 0x20, 0x00000001 }, + { 0x001480, 8, 0x10, 0x00000000 }, + { 0x001484, 8, 0x10, 0x00000000 }, + { 0x001488, 8, 0x10, 0x00000000 }, + { 0x003400, 128, 0x04, 0x00000000 }, + { 0x00030c, 1, 0x04, 0x00000001 }, + { 0x001944, 1, 0x04, 0x00000000 }, + { 0x001514, 1, 0x04, 0x00000000 }, + { 0x000d68, 1, 0x04, 0x0000ffff }, + { 0x00121c, 1, 0x04, 0x0fac6881 }, + { 0x000fac, 1, 0x04, 0x00000001 }, + { 0x001538, 1, 0x04, 0x00000001 }, + { 0x000fe0, 2, 0x04, 0x00000000 }, + { 0x000fe8, 1, 0x04, 0x00000014 }, + { 0x000fec, 1, 0x04, 0x00000040 }, + { 0x000ff0, 1, 0x04, 0x00000000 }, + { 0x00179c, 1, 0x04, 0x00000000 }, + { 0x001228, 1, 0x04, 0x00000400 }, + { 0x00122c, 1, 0x04, 0x00000300 }, + { 0x001230, 1, 0x04, 0x00010001 }, + { 0x0007f8, 1, 0x04, 0x00000000 }, + { 0x0015b4, 1, 0x04, 0x00000001 }, + { 0x0015cc, 1, 0x04, 0x00000000 }, + { 0x001534, 1, 0x04, 0x00000000 }, + { 0x000754, 1, 0x04, 0x00000001 }, + { 0x000fb0, 1, 0x04, 0x00000000 }, + { 0x0015d0, 1, 0x04, 0x00000000 }, + { 0x00153c, 1, 0x04, 0x00000000 }, + { 0x0016b4, 1, 0x04, 0x00000003 }, + { 0x000fbc, 4, 0x04, 0x0000ffff }, + { 0x000df8, 2, 0x04, 0x00000000 }, + { 0x001948, 1, 0x04, 0x00000000 }, + { 0x001970, 1, 0x04, 0x00000001 }, + { 0x00161c, 1, 0x04, 0x000009f0 }, + { 0x000dcc, 1, 0x04, 0x00000010 }, + { 0x0015e4, 1, 0x04, 0x00000000 }, + { 0x001160, 32, 0x04, 0x25e00040 }, + { 0x001880, 32, 0x04, 0x00000000 }, + { 0x000f84, 2, 0x04, 0x00000000 }, + { 0x0017c8, 2, 0x04, 0x00000000 }, + { 0x0017d0, 1, 0x04, 0x000000ff }, + { 0x0017d4, 1, 0x04, 0xffffffff }, + { 0x0017d8, 1, 0x04, 0x00000002 }, + { 0x0017dc, 1, 0x04, 0x00000000 }, + { 0x0015f4, 2, 0x04, 0x00000000 }, + { 0x001434, 2, 0x04, 0x00000000 }, + { 0x000d74, 1, 0x04, 0x00000000 }, + { 0x0013a4, 1, 0x04, 0x00000000 }, + { 0x001318, 1, 0x04, 0x00000001 }, + { 0x001080, 2, 0x04, 0x00000000 }, + { 0x001088, 2, 0x04, 0x00000001 }, + { 0x001090, 1, 0x04, 0x00000000 }, + { 0x001094, 1, 0x04, 0x00000001 }, + { 0x001098, 1, 0x04, 0x00000000 }, + { 0x00109c, 1, 0x04, 0x00000001 }, + { 0x0010a0, 2, 0x04, 0x00000000 }, + { 0x001644, 1, 0x04, 0x00000000 }, + { 0x000748, 1, 0x04, 0x00000000 }, + { 0x000de8, 1, 0x04, 0x00000000 }, + { 0x001648, 1, 0x04, 0x00000000 }, + { 0x0012a4, 1, 0x04, 0x00000000 }, + { 0x001120, 4, 0x04, 0x00000000 }, + { 0x001118, 1, 0x04, 0x00000000 }, + { 0x00164c, 1, 0x04, 0x00000000 }, + { 0x001658, 1, 0x04, 0x00000000 }, + { 0x001910, 1, 0x04, 0x00000290 }, + { 0x001518, 1, 0x04, 0x00000000 }, + { 0x00165c, 1, 0x04, 0x00000001 }, + { 0x001520, 1, 0x04, 0x00000000 }, + { 0x001604, 1, 0x04, 0x00000000 }, + { 0x001570, 1, 0x04, 0x00000000 }, + { 0x0013b0, 2, 0x04, 0x3f800000 }, + { 0x00020c, 1, 0x04, 0x00000000 }, + { 0x001670, 1, 0x04, 0x30201000 }, + { 0x001674, 1, 0x04, 0x70605040 }, + { 0x001678, 1, 0x04, 0xb8a89888 }, + { 0x00167c, 1, 0x04, 0xf8e8d8c8 }, + { 0x00166c, 1, 0x04, 0x00000000 }, + { 0x001680, 1, 0x04, 0x00ffff00 }, + { 0x0012d0, 1, 0x04, 0x00000003 }, + { 0x0012d4, 1, 0x04, 0x00000002 }, + { 0x001684, 2, 0x04, 0x00000000 }, + { 0x000dac, 2, 0x04, 0x00001b02 }, + { 0x000db4, 1, 0x04, 0x00000000 }, + { 0x00168c, 1, 0x04, 0x00000000 }, + { 0x0015bc, 1, 0x04, 0x00000000 }, + { 0x00156c, 1, 0x04, 0x00000000 }, + { 0x00187c, 1, 0x04, 0x00000000 }, + { 0x001110, 1, 0x04, 0x00000001 }, + { 0x000dc0, 3, 0x04, 0x00000000 }, + { 0x000f40, 5, 0x04, 0x00000000 }, + { 0x001234, 1, 0x04, 0x00000000 }, + { 0x001690, 1, 0x04, 0x00000000 }, + { 0x000790, 5, 0x04, 0x00000000 }, + { 0x00077c, 1, 0x04, 0x00000000 }, + { 0x001000, 1, 0x04, 0x00000010 }, + { 0x0010fc, 1, 0x04, 0x00000000 }, + { 0x001290, 1, 0x04, 0x00000000 }, + { 0x000218, 1, 0x04, 0x00000010 }, + { 0x0012d8, 1, 0x04, 0x00000000 }, + { 0x0012dc, 1, 0x04, 0x00000010 }, + { 0x000d94, 1, 0x04, 0x00000001 }, + { 0x00155c, 2, 0x04, 0x00000000 }, + { 0x001564, 1, 0x04, 0x00000fff }, + { 0x001574, 2, 0x04, 0x00000000 }, + { 0x00157c, 1, 0x04, 0x000fffff }, + { 0x001354, 1, 0x04, 0x00000000 }, + { 0x001610, 1, 0x04, 0x00000012 }, + { 0x001608, 2, 0x04, 0x00000000 }, + { 0x00260c, 1, 0x04, 0x00000000 }, + { 0x0007ac, 1, 0x04, 0x00000000 }, + { 0x00162c, 1, 0x04, 0x00000003 }, + { 0x000210, 1, 0x04, 0x00000000 }, + { 0x000320, 1, 0x04, 0x00000000 }, + { 0x000324, 6, 0x04, 0x3f800000 }, + { 0x000750, 1, 0x04, 0x00000000 }, + { 0x000760, 1, 0x04, 0x39291909 }, + { 0x000764, 1, 0x04, 0x79695949 }, + { 0x000768, 1, 0x04, 0xb9a99989 }, + { 0x00076c, 1, 0x04, 0xf9e9d9c9 }, + { 0x000770, 1, 0x04, 0x30201000 }, + { 0x000774, 1, 0x04, 0x70605040 }, + { 0x000778, 1, 0x04, 0x00009080 }, + { 0x000780, 1, 0x04, 0x39291909 }, + { 0x000784, 1, 0x04, 0x79695949 }, + { 0x000788, 1, 0x04, 0xb9a99989 }, + { 0x00078c, 1, 0x04, 0xf9e9d9c9 }, + { 0x0007d0, 1, 0x04, 0x30201000 }, + { 0x0007d4, 1, 0x04, 0x70605040 }, + { 0x0007d8, 1, 0x04, 0x00009080 }, + { 0x00037c, 1, 0x04, 0x00000001 }, + { 0x000740, 2, 0x04, 0x00000000 }, + { 0x002600, 1, 0x04, 0x00000000 }, + { 0x001918, 1, 0x04, 0x00000000 }, + { 0x00191c, 1, 0x04, 0x00000900 }, + { 0x001920, 1, 0x04, 0x00000405 }, + { 0x001308, 1, 0x04, 0x00000001 }, + { 0x001924, 1, 0x04, 0x00000000 }, + { 0x0013ac, 1, 0x04, 0x00000000 }, + { 0x00192c, 1, 0x04, 0x00000001 }, + { 0x00193c, 1, 0x04, 0x00002c1c }, + { 0x000d7c, 1, 0x04, 0x00000000 }, + { 0x000f8c, 1, 0x04, 0x00000000 }, + { 0x0002c0, 1, 0x04, 0x00000001 }, + { 0x001510, 1, 0x04, 0x00000000 }, + { 0x001940, 1, 0x04, 0x00000000 }, + { 0x000ff4, 2, 0x04, 0x00000000 }, + { 0x00194c, 2, 0x04, 0x00000000 }, + { 0x001968, 1, 0x04, 0x00000000 }, + { 0x001590, 1, 0x04, 0x0000003f }, + { 0x0007e8, 4, 0x04, 0x00000000 }, + { 0x00196c, 1, 0x04, 0x00000011 }, + { 0x0002e4, 1, 0x04, 0x0000b001 }, + { 0x00036c, 2, 0x04, 0x00000000 }, + { 0x00197c, 1, 0x04, 0x00000000 }, + { 0x000fcc, 2, 0x04, 0x00000000 }, + { 0x0002d8, 1, 0x04, 0x00000040 }, + { 0x001980, 1, 0x04, 0x00000080 }, + { 0x001504, 1, 0x04, 0x00000080 }, + { 0x001984, 1, 0x04, 0x00000000 }, + { 0x000f60, 1, 0x04, 0x00000000 }, + { 0x000f64, 1, 0x04, 0x00400040 }, + { 0x000f68, 1, 0x04, 0x00002212 }, + { 0x000f6c, 1, 0x04, 0x08080203 }, + { 0x001108, 1, 0x04, 0x00000008 }, + { 0x000f70, 1, 0x04, 0x00080001 }, + { 0x000ffc, 1, 0x04, 0x00000000 }, + { 0x000300, 1, 0x04, 0x00000001 }, + { 0x0013a8, 1, 0x04, 0x00000000 }, + { 0x0012ec, 1, 0x04, 0x00000000 }, + { 0x001310, 1, 0x04, 0x00000000 }, + { 0x001314, 1, 0x04, 0x00000001 }, + { 0x001380, 1, 0x04, 0x00000000 }, + { 0x001384, 4, 0x04, 0x00000001 }, + { 0x001394, 1, 0x04, 0x00000000 }, + { 0x00139c, 1, 0x04, 0x00000000 }, + { 0x001398, 1, 0x04, 0x00000000 }, + { 0x001594, 1, 0x04, 0x00000000 }, + { 0x001598, 4, 0x04, 0x00000001 }, + { 0x000f54, 3, 0x04, 0x00000000 }, + { 0x0019bc, 1, 0x04, 0x00000000 }, + { 0x000f9c, 2, 0x04, 0x00000000 }, + { 0x0012cc, 1, 0x04, 0x00000000 }, + { 0x0012e8, 1, 0x04, 0x00000000 }, + { 0x00130c, 1, 0x04, 0x00000001 }, + { 0x001360, 8, 0x04, 0x00000000 }, + { 0x00133c, 2, 0x04, 0x00000001 }, + { 0x001344, 1, 0x04, 0x00000002 }, + { 0x001348, 2, 0x04, 0x00000001 }, + { 0x001350, 1, 0x04, 0x00000002 }, + { 0x001358, 1, 0x04, 0x00000001 }, + { 0x0012e4, 1, 0x04, 0x00000000 }, + { 0x00131c, 4, 0x04, 0x00000000 }, + { 0x0019c0, 1, 0x04, 0x00000000 }, + { 0x001140, 1, 0x04, 0x00000000 }, + { 0x000dd0, 1, 0x04, 0x00000000 }, + { 0x000dd4, 1, 0x04, 0x00000001 }, + { 0x0002f4, 1, 0x04, 0x00000000 }, + { 0x0019c4, 1, 0x04, 0x00000000 }, + { 0x0019c8, 1, 0x04, 0x00001500 }, + { 0x00135c, 1, 0x04, 0x00000000 }, + { 0x000f90, 1, 0x04, 0x00000000 }, + { 0x0019e0, 8, 0x04, 0x00000001 }, + { 0x0019cc, 1, 0x04, 0x00000001 }, + { 0x0015b8, 1, 0x04, 0x00000000 }, + { 0x001a00, 1, 0x04, 0x00001111 }, + { 0x001a04, 7, 0x04, 0x00000000 }, + { 0x000d6c, 2, 0x04, 0xffff0000 }, + { 0x0010f8, 1, 0x04, 0x00001010 }, + { 0x000d80, 5, 0x04, 0x00000000 }, + { 0x000da0, 1, 0x04, 0x00000000 }, + { 0x0007a4, 2, 0x04, 0x00000000 }, + { 0x001508, 1, 0x04, 0x80000000 }, + { 0x00150c, 1, 0x04, 0x40000000 }, + { 0x001668, 1, 0x04, 0x00000000 }, + { 0x000318, 2, 0x04, 0x00000008 }, + { 0x000d9c, 1, 0x04, 0x00000001 }, + { 0x000f14, 1, 0x04, 0x00000000 }, + { 0x000374, 1, 0x04, 0x00000000 }, + { 0x000378, 1, 0x04, 0x0000000c }, + { 0x0007dc, 1, 0x04, 0x00000000 }, + { 0x00074c, 1, 0x04, 0x00000055 }, + { 0x001420, 1, 0x04, 0x00000003 }, + { 0x001008, 1, 0x04, 0x00000008 }, + { 0x00100c, 1, 0x04, 0x00000040 }, + { 0x001010, 1, 0x04, 0x0000012c }, + { 0x000d60, 1, 0x04, 0x00000040 }, + { 0x001018, 1, 0x04, 0x00000020 }, + { 0x00101c, 1, 0x04, 0x00000001 }, + { 0x001020, 1, 0x04, 0x00000020 }, + { 0x001024, 1, 0x04, 0x00000001 }, + { 0x001444, 3, 0x04, 0x00000000 }, + { 0x000360, 1, 0x04, 0x20164010 }, + { 0x000364, 1, 0x04, 0x00000020 }, + { 0x000368, 1, 0x04, 0x00000000 }, + { 0x000da8, 1, 0x04, 0x00000030 }, + { 0x000de4, 1, 0x04, 0x00000000 }, + { 0x000204, 1, 0x04, 0x00000006 }, + { 0x0002d0, 1, 0x04, 0x003fffff }, + { 0x001220, 1, 0x04, 0x00000005 }, + { 0x000fdc, 1, 0x04, 0x00000000 }, + { 0x000f98, 1, 0x04, 0x00400008 }, + { 0x001284, 1, 0x04, 0x08000080 }, + { 0x001450, 1, 0x04, 0x00400008 }, + { 0x001454, 1, 0x04, 0x08000080 }, + { 0x000214, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_mthd[] = { + { gm107_grctx_init_b097_0, 0xb097 }, + { gf100_grctx_init_902d_0, 0x902d }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_fe_0[] = { + { 0x404004, 8, 0x04, 0x00000000 }, + { 0x404024, 1, 0x04, 0x0000e000 }, + { 0x404028, 8, 0x04, 0x00000000 }, + { 0x4040a8, 8, 0x04, 0x00000000 }, + { 0x4040c8, 1, 0x04, 0xf800008f }, + { 0x4040d0, 6, 0x04, 0x00000000 }, + { 0x4040f8, 1, 0x04, 0x00000000 }, + { 0x404100, 10, 0x04, 0x00000000 }, + { 0x404130, 2, 0x04, 0x00000000 }, + { 0x404150, 1, 0x04, 0x0000002e }, + { 0x404154, 1, 0x04, 0x00000400 }, + { 0x404158, 1, 0x04, 0x00000200 }, + { 0x404164, 1, 0x04, 0x00000045 }, + { 0x40417c, 2, 0x04, 0x00000000 }, + { 0x404194, 1, 0x04, 0x01000700 }, + { 0x4041a0, 4, 0x04, 0x00000000 }, + { 0x404200, 4, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_ds_0[] = { + { 0x405800, 1, 0x04, 0x0f8001bf }, + { 0x405830, 1, 0x04, 0x0aa01000 }, + { 0x405834, 1, 0x04, 0x08000000 }, + { 0x405838, 1, 0x04, 0x00000000 }, + { 0x405854, 1, 0x04, 0x00000000 }, + { 0x405870, 4, 0x04, 0x00000001 }, + { 0x405a00, 2, 0x04, 0x00000000 }, + { 0x405a18, 1, 0x04, 0x00000000 }, + { 0x405a1c, 1, 0x04, 0x000000ff }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_pd_0[] = { + { 0x406020, 1, 0x04, 0x07410001 }, + { 0x406028, 4, 0x04, 0x00000001 }, + { 0x4064a8, 1, 0x04, 0x00000000 }, + { 0x4064ac, 1, 0x04, 0x00003fff }, + { 0x4064b0, 3, 0x04, 0x00000000 }, + { 0x4064c0, 1, 0x04, 0x80400280 }, + { 0x4064c4, 1, 0x04, 0x0400ffff }, + { 0x4064c8, 1, 0x04, 0x018001ff }, + { 0x4064cc, 9, 0x04, 0x00000000 }, + { 0x4064fc, 1, 0x04, 0x0000022a }, + { 0x406500, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_be_0[] = { + { 0x408800, 1, 0x04, 0x32802a3c }, + { 0x408804, 1, 0x04, 0x00000040 }, + { 0x408808, 1, 0x04, 0x1003e005 }, + { 0x408840, 1, 0x04, 0x0000000b }, + { 0x408900, 1, 0x04, 0xb080b801 }, + { 0x408904, 1, 0x04, 0x63038001 }, + { 0x408908, 1, 0x04, 0x02c8102f }, + { 0x408980, 1, 0x04, 0x0000011d }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_hub[] = { + { gf100_grctx_init_main_0 }, + { gm107_grctx_init_fe_0 }, + { gk110_grctx_init_pri_0 }, + { gk104_grctx_init_memfmt_0 }, + { gm107_grctx_init_ds_0 }, + { gk110_grctx_init_cwd_0 }, + { gm107_grctx_init_pd_0 }, + { gk208_grctx_init_rstr2d_0 }, + { gk104_grctx_init_scc_0 }, + { gm107_grctx_init_be_0 }, + {} +}; + +const struct gf100_gr_init +gm107_grctx_init_gpc_unk_0[] = { + { 0x418380, 1, 0x04, 0x00000056 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_gpc_unk_1[] = { + { 0x418600, 1, 0x04, 0x0000007f }, + { 0x418684, 1, 0x04, 0x0000001f }, + { 0x418700, 1, 0x04, 0x00000002 }, + { 0x418704, 1, 0x04, 0x00000080 }, + { 0x418708, 1, 0x04, 0x40000000 }, + { 0x41870c, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_setup_0[] = { + { 0x418800, 1, 0x04, 0x7006863a }, + { 0x418810, 1, 0x04, 0x00000000 }, + { 0x418828, 1, 0x04, 0x00000044 }, + { 0x418830, 1, 0x04, 0x10000001 }, + { 0x4188d8, 1, 0x04, 0x00000008 }, + { 0x4188e0, 1, 0x04, 0x01000000 }, + { 0x4188e8, 5, 0x04, 0x00000000 }, + { 0x4188fc, 1, 0x04, 0x20100058 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_gpc_unk_2[] = { + { 0x418d24, 1, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x90000000 }, + { 0x418e24, 1, 0x04, 0x00000000 }, + { 0x418e28, 1, 0x04, 0x00000030 }, + { 0x418e30, 1, 0x04, 0x00000000 }, + { 0x418e34, 1, 0x04, 0x00010000 }, + { 0x418e38, 1, 0x04, 0x00000000 }, + { 0x418e40, 22, 0x04, 0x00000000 }, + { 0x418ea0, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_gpc_0[] = { + { gm107_grctx_init_gpc_unk_0 }, + { gk208_grctx_init_prop_0 }, + { gm107_grctx_init_gpc_unk_1 }, + { gm107_grctx_init_setup_0 }, + { gf100_grctx_init_zcull_0 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_gpc_1[] = { + { gk208_grctx_init_crstr_0 }, + { gk104_grctx_init_gpm_0 }, + { gm107_grctx_init_gpc_unk_2 }, + { gf100_grctx_init_gcc_0 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_tex_0[] = { + { 0x419a00, 1, 0x04, 0x000300f0 }, + { 0x419a04, 1, 0x04, 0x00000005 }, + { 0x419a08, 1, 0x04, 0x00000421 }, + { 0x419a0c, 1, 0x04, 0x00120000 }, + { 0x419a10, 1, 0x04, 0x00000000 }, + { 0x419a14, 1, 0x04, 0x00002200 }, + { 0x419a1c, 1, 0x04, 0x0000c000 }, + { 0x419a20, 1, 0x04, 0x20008a00 }, + { 0x419a30, 1, 0x04, 0x00000001 }, + { 0x419a3c, 1, 0x04, 0x00000002 }, + { 0x419ac4, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_mpc_0[] = { + { 0x419c00, 1, 0x04, 0x0000001a }, + { 0x419c04, 1, 0x04, 0x80000006 }, + { 0x419c08, 1, 0x04, 0x00000002 }, + { 0x419c20, 1, 0x04, 0x00000000 }, + { 0x419c24, 1, 0x04, 0x00084210 }, + { 0x419c28, 1, 0x04, 0x3efbefbe }, + { 0x419c2c, 1, 0x04, 0x00000000 }, + { 0x419c34, 1, 0x04, 0x01ff1ff3 }, + { 0x419c3c, 1, 0x04, 0x00001919 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_l1c_0[] = { + { 0x419c84, 1, 0x04, 0x00000020 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_sm_0[] = { + { 0x419e04, 3, 0x04, 0x00000000 }, + { 0x419e10, 1, 0x04, 0x00001c02 }, + { 0x419e44, 1, 0x04, 0x00d3eff2 }, + { 0x419e48, 1, 0x04, 0x00000000 }, + { 0x419e4c, 1, 0x04, 0x0000007f }, + { 0x419e50, 1, 0x04, 0x00000000 }, + { 0x419e60, 4, 0x04, 0x00000000 }, + { 0x419e74, 10, 0x04, 0x00000000 }, + { 0x419eac, 1, 0x04, 0x0001cf8b }, + { 0x419eb0, 1, 0x04, 0x00030300 }, + { 0x419eb8, 1, 0x04, 0x00000000 }, + { 0x419ef0, 24, 0x04, 0x00000000 }, + { 0x419f68, 2, 0x04, 0x00000000 }, + { 0x419f70, 1, 0x04, 0x00000020 }, + { 0x419f78, 1, 0x04, 0x000003eb }, + { 0x419f7c, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_tpc[] = { + { gf117_grctx_init_pe_0 }, + { gm107_grctx_init_tex_0 }, + { gm107_grctx_init_mpc_0 }, + { gm107_grctx_init_l1c_0 }, + { gm107_grctx_init_sm_0 }, + {} +}; + +static const struct gf100_gr_init +gm107_grctx_init_cbm_0[] = { + { 0x41bec0, 1, 0x04, 0x00000000 }, + { 0x41bec4, 1, 0x04, 0x01050000 }, + { 0x41bee4, 1, 0x04, 0x00000000 }, + { 0x41bef0, 1, 0x04, 0x000003ff }, + { 0x41bef4, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gm107_grctx_init_wwdx_0[] = { + { 0x41bf00, 1, 0x04, 0x0a418820 }, + { 0x41bf04, 1, 0x04, 0x062080e6 }, + { 0x41bf08, 1, 0x04, 0x020398a4 }, + { 0x41bf0c, 1, 0x04, 0x0e629062 }, + { 0x41bf10, 1, 0x04, 0x0a418820 }, + { 0x41bf14, 1, 0x04, 0x000000e6 }, + { 0x41bfd0, 1, 0x04, 0x00900103 }, + { 0x41bfe0, 1, 0x04, 0x80000000 }, + { 0x41bfe4, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gm107_grctx_pack_ppc[] = { + { gk104_grctx_init_pes_0 }, + { gm107_grctx_init_cbm_0 }, + { gm107_grctx_init_wwdx_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +static void +gm107_grctx_generate_r419e00(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419e00, 0x00808080, 0x00808080); + nvkm_mask(device, 0x419ccc, 0x80000000, 0x80000000); + nvkm_mask(device, 0x419f80, 0x80000000, 0x80000000); + nvkm_mask(device, 0x419f88, 0x80000000, 0x80000000); +} + +void +gm107_grctx_generate_bundle(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + const struct gf100_grctx_func *grctx = chan->gr->func->grctx; + const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth, size / 0x20); + const u32 token_limit = grctx->bundle_token_limit; + + gf100_grctx_patch_wr32(chan, 0x408004, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408008, 0x80000000 | (size >> 8)); + gf100_grctx_patch_wr32(chan, 0x418e24, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x418e28, 0x80000000 | (size >> 8)); + gf100_grctx_patch_wr32(chan, 0x4064c8, (state_limit << 16) | token_limit); +} + +void +gm107_grctx_generate_pagepool(struct gf100_gr_chan *chan, u64 addr) +{ + gk104_grctx_generate_pagepool(chan, addr); + gf100_grctx_patch_wr32(chan, 0x418e30, 0x80000000); +} + +void +gm107_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 attrib = grctx->attrib_nr; + const int max_batches = 0xffff; + u32 bo = 0; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; + int gpc, ppc, n = 0; + + gf100_grctx_patch_wr32(chan, 0x405830, (attrib << 16) | alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++, n++) { + const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 bs = attrib * gr->ppc_tpc_nr[gpc][ppc]; + const u32 u = 0x418ea0 + (n * 0x04); + const u32 o = PPC_UNIT(gpc, ppc, 0); + + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + + gf100_grctx_patch_wr32(chan, o + 0xc0, bs); + gf100_grctx_patch_wr32(chan, o + 0xf4, bo); + bo += grctx->attrib_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, o + 0xe4, as); + gf100_grctx_patch_wr32(chan, o + 0xf8, ao); + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, u, ((bs / 3) << 16) | bs); + } + } +} + +void +gm107_grctx_generate_attrib_cb(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gf100_grctx_generate_attrib_cb(chan, addr, size); + + gf100_grctx_patch_wr32(chan, 0x419c2c, 0x10000000 | addr >> 12); +} + +static void +gm107_grctx_generate_r406500(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x406500, 0x00000001); +} + +void +gm107_grctx_generate_sm_id(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x698), sm); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), sm); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), sm); +} + +const struct gf100_grctx_func +gm107_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .hub = gm107_grctx_pack_hub, + .gpc_0 = gm107_grctx_pack_gpc_0, + .gpc_1 = gm107_grctx_pack_gpc_1, + .zcull = gf100_grctx_pack_zcull, + .tpc = gm107_grctx_pack_tpc, + .ppc = gm107_grctx_pack_ppc, + .icmd = gm107_grctx_pack_icmd, + .mthd = gm107_grctx_pack_mthd, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x2c0, + .pagepool = gm107_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gm107_grctx_generate_attrib_cb, + .attrib = gm107_grctx_generate_attrib, + .attrib_nr_max = 0xff0, + .attrib_nr = 0xaa0, + .alpha_nr_max = 0x1800, + .alpha_nr = 0x1000, + .sm_id = gm107_grctx_generate_sm_id, + .tpc_nr = gf100_grctx_generate_tpc_nr, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .alpha_beta_tables = gk104_grctx_generate_alpha_beta_tables, + .dist_skip_table = gf117_grctx_generate_dist_skip_table, + .r406500 = gm107_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .r419e00 = gm107_grctx_generate_r419e00, + .r419f78 = gk110_grctx_generate_r419f78, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm200.c new file mode 100644 index 0000000000..175da8ac65 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm200.c @@ -0,0 +1,130 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gm200_grctx_generate_r419a3c(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419a3c, 0x00000014, 0x00000000); +} + +static void +gm200_grctx_generate_r418e94(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x418e94, 0xffffffff, 0xc4230000); + nvkm_mask(device, 0x418e4c, 0xffffffff, 0x70000000); +} + +void +gm200_grctx_generate_smid_config(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 dist_nr = DIV_ROUND_UP(gr->tpc_total, 4); + u32 dist[TPC_MAX / 4] = {}; + u32 gpcs[GPC_MAX] = {}; + u8 sm, i; + + for (sm = 0; sm < gr->sm_nr; sm++) { + const u8 gpc = gr->sm[sm].gpc; + const u8 tpc = gr->sm[sm].tpc; + dist[sm / 4] |= ((gpc << 4) | tpc) << ((sm % 4) * 8); + gpcs[gpc] |= sm << (tpc * 8); + } + + for (i = 0; i < dist_nr; i++) + nvkm_wr32(device, 0x405b60 + (i * 4), dist[i]); + for (i = 0; i < gr->gpc_nr; i++) + nvkm_wr32(device, 0x405ba0 + (i * 4), gpcs[i]); +} + +void +gm200_grctx_generate_tpc_mask(struct gf100_gr *gr) +{ + u32 tmp, i; + for (tmp = 0, i = 0; i < gr->gpc_nr; i++) + tmp |= ((1 << gr->tpc_nr[i]) - 1) << (i * gr->func->tpc_nr); + nvkm_wr32(gr->base.engine.subdev.device, 0x4041c4, tmp); +} + +void +gm200_grctx_generate_r406500(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x406500, 0x00000000); +} + +void +gm200_grctx_generate_dist_skip_table(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 data[8] = {}; + int gpc, ppc, i; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++) { + u8 ppc_tpcs = gr->ppc_tpc_nr[gpc][ppc]; + u8 ppc_tpcm = gr->ppc_tpc_mask[gpc][ppc]; + while (ppc_tpcs-- > gr->ppc_tpc_min) + ppc_tpcm &= ppc_tpcm - 1; + ppc_tpcm ^= gr->ppc_tpc_mask[gpc][ppc]; + ((u8 *)data)[gpc] |= ppc_tpcm; + } + } + + for (i = 0; i < ARRAY_SIZE(data); i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), data[i]); +} + +const struct gf100_grctx_func +gm200_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x780, + .pagepool = gm107_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gm107_grctx_generate_attrib_cb, + .attrib = gm107_grctx_generate_attrib, + .attrib_nr_max = 0x600, + .attrib_nr = 0x400, + .alpha_nr_max = 0x1800, + .alpha_nr = 0x1000, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .tpc_mask = gm200_grctx_generate_tpc_mask, + .smid_config = gm200_grctx_generate_smid_config, + .r418e94 = gm200_grctx_generate_r418e94, + .r419a3c = gm200_grctx_generate_r419a3c, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c new file mode 100644 index 0000000000..b8edccfada --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c @@ -0,0 +1,91 @@ +/* + * Copyright (c) 2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +static void +gm20b_grctx_generate_main(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + u32 idle_timeout; + int i, tmp; + + gf100_gr_mmio(gr, gr->sw_ctx); + + gf100_gr_wait_idle(gr); + + idle_timeout = nvkm_mask(device, 0x404154, 0xffffffff, 0x00000000); + + grctx->attrib_cb(chan, chan->attrib_cb->addr, grctx->attrib_cb_size(gr)); + grctx->attrib(chan); + + grctx->unkn(gr); + + gf100_grctx_generate_floorsweep(gr); + + for (i = 0; i < 8; i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); + + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); + + nvkm_wr32(device, 0x408908, nvkm_rd32(device, 0x410108) | 0x80000000); + + for (tmp = 0, i = 0; i < gr->gpc_nr; i++) + tmp |= ((1 << gr->tpc_nr[i]) - 1) << (i * 4); + nvkm_wr32(device, 0x4041c4, tmp); + + gm200_grctx_generate_smid_config(gr); + + gf100_gr_wait_idle(gr); + + nvkm_wr32(device, 0x404154, idle_timeout); + gf100_gr_wait_idle(gr); + + gf100_gr_mthd(gr, gr->method); + gf100_gr_wait_idle(gr); + + gf100_gr_icmd(gr, gr->bundle); + grctx->pagepool(chan, chan->pagepool->addr); + grctx->bundle(chan, chan->bundle_cb->addr, grctx->bundle_size); +} + +const struct gf100_grctx_func +gm20b_grctx = { + .main = gm20b_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x1800, + .bundle_min_gpm_fifo_depth = 0x182, + .bundle_token_limit = 0x1c0, + .pagepool = gm107_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib_cb_size = gf100_grctx_generate_attrib_cb_size, + .attrib_cb = gm107_grctx_generate_attrib_cb, + .attrib = gm107_grctx_generate_attrib, + .attrib_nr_max = 0x600, + .attrib_nr = 0x400, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c new file mode 100644 index 0000000000..8485aaeae7 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c @@ -0,0 +1,151 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +void +gp100_grctx_generate_pagepool(struct gf100_gr_chan *chan, u64 addr) +{ + gf100_grctx_patch_wr32(chan, 0x40800c, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408010, 0x8007d800); + gf100_grctx_patch_wr32(chan, 0x419004, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x419008, 0x00000000); +} + +static void +gp100_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 attrib = grctx->attrib_nr; + const int max_batches = 0xffff; + u32 size = grctx->alpha_nr_max * gr->tpc_total; + u32 ao = 0; + u32 bo = ao + size; + int gpc, ppc, n = 0; + + gf100_grctx_patch_wr32(chan, 0x405830, attrib); + gf100_grctx_patch_wr32(chan, 0x40585c, alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++, n++) { + const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 bs = attrib * gr->ppc_tpc_max; + const u32 u = 0x418ea0 + (n * 0x04); + const u32 o = PPC_UNIT(gpc, ppc, 0); + + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + + gf100_grctx_patch_wr32(chan, o + 0xc0, bs); + gf100_grctx_patch_wr32(chan, o + 0xf4, bo); + gf100_grctx_patch_wr32(chan, o + 0xf0, bs); + bo += grctx->attrib_nr_max * gr->ppc_tpc_max; + gf100_grctx_patch_wr32(chan, o + 0xe4, as); + gf100_grctx_patch_wr32(chan, o + 0xf8, ao); + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, u, bs); + } + } + + gf100_grctx_patch_wr32(chan, 0x418eec, 0x00000000); + gf100_grctx_patch_wr32(chan, 0x41befc, 0x00000000); +} + +void +gp100_grctx_generate_attrib_cb(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gm107_grctx_generate_attrib_cb(chan, addr, size); + + gf100_grctx_patch_wr32(chan, 0x419b00, 0x00000000 | addr >> 12); + gf100_grctx_patch_wr32(chan, 0x419b04, 0x80000000 | size >> 7); +} + +static u32 +gp100_grctx_generate_attrib_cb_size(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + u32 size = grctx->alpha_nr_max * gr->tpc_total; + int gpc; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) + size += grctx->attrib_nr_max * gr->func->ppc_nr * gr->ppc_tpc_max; + + return ((size * 0x20) + 128) & ~127; +} + +void +gp100_grctx_generate_smid_config(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 dist_nr = DIV_ROUND_UP(gr->tpc_total, 4); + u32 dist[TPC_MAX / 4] = {}, gpcs[16] = {}; + u8 sm, i; + + for (sm = 0; sm < gr->sm_nr; sm++) { + const u8 gpc = gr->sm[sm].gpc; + const u8 tpc = gr->sm[sm].tpc; + dist[sm / 4] |= ((gpc << 4) | tpc) << ((sm % 4) * 8); + gpcs[gpc + (gr->func->gpc_nr * (tpc / 4))] |= sm << ((tpc % 4) * 8); + } + + for (i = 0; i < dist_nr; i++) + nvkm_wr32(device, 0x405b60 + (i * 4), dist[i]); + for (i = 0; i < ARRAY_SIZE(gpcs); i++) + nvkm_wr32(device, 0x405ba0 + (i * 4), gpcs[i]); +} + +const struct gf100_grctx_func +gp100_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x1080, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp100_grctx_generate_attrib_cb_size, + .attrib_cb = gp100_grctx_generate_attrib_cb, + .attrib = gp100_grctx_generate_attrib, + .attrib_nr_max = 0x660, + .attrib_nr = 0x440, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .tpc_mask = gm200_grctx_generate_tpc_mask, + .smid_config = gp100_grctx_generate_smid_config, + .r419a3c = gm200_grctx_generate_r419a3c, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c new file mode 100644 index 0000000000..7537979a54 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c @@ -0,0 +1,125 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +static void +gp102_grctx_generate_r408840(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x408840, 0x00000003, 0x00000000); +} + +void +gp102_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 attrib = grctx->attrib_nr; + const u32 gfxp = grctx->gfxp_nr; + const int max_batches = 0xffff; + u32 size = grctx->alpha_nr_max * gr->tpc_total; + u32 ao = 0; + u32 bo = ao + size; + int gpc, ppc, n = 0; + + gf100_grctx_patch_wr32(chan, 0x405830, attrib); + gf100_grctx_patch_wr32(chan, 0x40585c, alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++, n++) { + const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 bs = attrib * gr->ppc_tpc_max; + const u32 gs = gfxp * gr->ppc_tpc_max; + const u32 u = 0x418ea0 + (n * 0x04); + const u32 o = PPC_UNIT(gpc, ppc, 0); + const u32 p = GPC_UNIT(gpc, 0xc44 + (ppc * 4)); + + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + + gf100_grctx_patch_wr32(chan, o + 0xc0, gs); + gf100_grctx_patch_wr32(chan, p, bs); + gf100_grctx_patch_wr32(chan, o + 0xf4, bo); + gf100_grctx_patch_wr32(chan, o + 0xf0, bs); + bo += gs; + gf100_grctx_patch_wr32(chan, o + 0xe4, as); + gf100_grctx_patch_wr32(chan, o + 0xf8, ao); + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, u, bs); + } + } + + gf100_grctx_patch_wr32(chan, 0x4181e4, 0x00000100); + gf100_grctx_patch_wr32(chan, 0x41befc, 0x00000100); +} + +u32 +gp102_grctx_generate_attrib_cb_size(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + u32 size = grctx->alpha_nr_max * gr->tpc_total; + int gpc; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) + size += grctx->gfxp_nr * gr->func->ppc_nr * gr->ppc_tpc_max; + + return ((size * 0x20) + 127) & ~127; +} + +const struct gf100_grctx_func +gp102_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x900, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gp100_grctx_generate_attrib_cb, + .attrib = gp102_grctx_generate_attrib, + .attrib_nr_max = 0x4b0, + .attrib_nr = 0x320, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .gfxp_nr = 0xba8, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .tpc_mask = gm200_grctx_generate_tpc_mask, + .smid_config = gp100_grctx_generate_smid_config, + .r419a3c = gm200_grctx_generate_r419a3c, + .r408840 = gp102_grctx_generate_r408840, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp104.c new file mode 100644 index 0000000000..90b5f793e5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp104.c @@ -0,0 +1,50 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +const struct gf100_grctx_func +gp104_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x900, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gp100_grctx_generate_attrib_cb, + .attrib = gp102_grctx_generate_attrib, + .attrib_nr_max = 0x4b0, + .attrib_nr = 0x320, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .gfxp_nr = 0xba8, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .tpc_mask = gm200_grctx_generate_tpc_mask, + .smid_config = gp100_grctx_generate_smid_config, + .r419a3c = gm200_grctx_generate_r419a3c, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp107.c new file mode 100644 index 0000000000..d191761a04 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp107.c @@ -0,0 +1,58 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +const struct gf100_grctx_func +gp107_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x300, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gp100_grctx_generate_attrib_cb, + .attrib = gp102_grctx_generate_attrib, + .attrib_nr_max = 0x15de, + .attrib_nr = 0x540, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .gfxp_nr = 0xe94, + .sm_id = gm107_grctx_generate_sm_id, + .rop_mapping = gf117_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .tpc_mask = gm200_grctx_generate_tpc_mask, + .smid_config = gp100_grctx_generate_smid_config, + .r419a3c = gm200_grctx_generate_r419a3c, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgv100.c new file mode 100644 index 0000000000..957ea9d6ba --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgv100.c @@ -0,0 +1,222 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +/******************************************************************************* + * PGRAPH context implementation + ******************************************************************************/ + +static const struct gf100_gr_init +gv100_grctx_init_sw_veid_bundle_init_0[] = { + { 0x00001000, 64, 0x00100000, 0x00000008 }, + { 0x00000941, 64, 0x00100000, 0x00000000 }, + { 0x0000097e, 64, 0x00100000, 0x00000000 }, + { 0x0000097f, 64, 0x00100000, 0x00000100 }, + { 0x0000035c, 64, 0x00100000, 0x00000000 }, + { 0x0000035d, 64, 0x00100000, 0x00000000 }, + { 0x00000a08, 64, 0x00100000, 0x00000000 }, + { 0x00000a09, 64, 0x00100000, 0x00000000 }, + { 0x00000a0a, 64, 0x00100000, 0x00000000 }, + { 0x00000352, 64, 0x00100000, 0x00000000 }, + { 0x00000353, 64, 0x00100000, 0x00000000 }, + { 0x00000358, 64, 0x00100000, 0x00000000 }, + { 0x00000359, 64, 0x00100000, 0x00000000 }, + { 0x00000370, 64, 0x00100000, 0x00000000 }, + { 0x00000371, 64, 0x00100000, 0x00000000 }, + { 0x00000372, 64, 0x00100000, 0x000fffff }, + { 0x00000366, 64, 0x00100000, 0x00000000 }, + { 0x00000367, 64, 0x00100000, 0x00000000 }, + { 0x00000368, 64, 0x00100000, 0x00000fff }, + { 0x00000623, 64, 0x00100000, 0x00000000 }, + { 0x00000624, 64, 0x00100000, 0x00000000 }, + { 0x0001e100, 1, 0x00000001, 0x02000001 }, + {} +}; + +static const struct gf100_gr_pack +gv100_grctx_pack_sw_veid_bundle_init[] = { + { gv100_grctx_init_sw_veid_bundle_init_0 }, + {} +}; + +void +gv100_grctx_generate_attrib(struct gf100_gr_chan *chan) +{ + struct gf100_gr *gr = chan->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 attrib = grctx->attrib_nr; + const u32 gfxp = grctx->gfxp_nr; + const int max_batches = 0xffff; + u32 size = grctx->alpha_nr_max * gr->tpc_total; + u32 ao = 0; + u32 bo = ao + size; + int gpc, ppc, n = 0; + + gf100_grctx_patch_wr32(chan, 0x405830, attrib); + gf100_grctx_patch_wr32(chan, 0x40585c, alpha); + gf100_grctx_patch_wr32(chan, 0x4064c4, ((alpha / 4) << 16) | max_batches); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++, n++) { + const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 bs = attrib * gr->ppc_tpc_max; + const u32 gs = gfxp * gr->ppc_tpc_max; + const u32 u = 0x418ea0 + (n * 0x04); + const u32 o = PPC_UNIT(gpc, ppc, 0); + + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + + gf100_grctx_patch_wr32(chan, o + 0xc0, gs); + gf100_grctx_patch_wr32(chan, o + 0xf4, bo); + gf100_grctx_patch_wr32(chan, o + 0xf0, bs); + bo += gs; + gf100_grctx_patch_wr32(chan, o + 0xe4, as); + gf100_grctx_patch_wr32(chan, o + 0xf8, ao); + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; + gf100_grctx_patch_wr32(chan, u, bs); + } + } + + gf100_grctx_patch_wr32(chan, 0x4181e4, 0x00000100); + gf100_grctx_patch_wr32(chan, 0x41befc, 0x00000100); +} + +void +gv100_grctx_generate_attrib_cb(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gm107_grctx_generate_attrib_cb(chan, addr, size); + + gf100_grctx_patch_wr32(chan, 0x419e00, 0x00000000 | addr >> 12); + gf100_grctx_patch_wr32(chan, 0x419e04, 0x80000000 | size >> 7); +} + +void +gv100_grctx_generate_rop_mapping(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 mapregs = DIV_ROUND_UP(gr->func->gpc_nr * gr->func->tpc_nr, 6); + u32 data; + int i, j; + + /* Pack tile map into register format. */ + nvkm_wr32(device, 0x418bb8, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0; i < mapregs; i++) { + for (data = 0, j = 0; j < 6; j++) + data |= (gr->tile[i * 6 + j] & 0x1f) << (j * 5); + nvkm_wr32(device, 0x418b08 + (i * 4), data); + nvkm_wr32(device, 0x41bf00 + (i * 4), data); + nvkm_wr32(device, 0x40780c + (i * 4), data); + } + + /* GPC_BROADCAST.TP_BROADCAST */ + nvkm_wr32(device, 0x41bfd0, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); + for (i = 0, j = 1; i < 5; i++, j += 4) { + u8 v19 = (1 << (j + 0)) % gr->tpc_total; + u8 v20 = (1 << (j + 1)) % gr->tpc_total; + u8 v21 = (1 << (j + 2)) % gr->tpc_total; + u8 v22 = (1 << (j + 3)) % gr->tpc_total; + nvkm_wr32(device, 0x41bfb0 + (i * 4), (v22 << 24) | + (v21 << 16) | + (v20 << 8) | + v19); + } + + /* UNK78xx */ + nvkm_wr32(device, 0x4078bc, (gr->tpc_total << 8) | + gr->screen_tile_row_offset); +} + +void +gv100_grctx_generate_r400088(struct gf100_gr *gr, bool on) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x400088, 0x00060000, on ? 0x00060000 : 0x00000000); +} + +static void +gv100_grctx_generate_sm_id(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + tpc = gv100_gr_nonpes_aware_tpc(gr, gpc, tpc); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x608), sm); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), sm); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), sm); +} + +void +gv100_grctx_generate_unkn(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x41be08, 0x00000004, 0x00000004); + nvkm_mask(device, 0x4064c0, 0x80000000, 0x80000000); + nvkm_mask(device, 0x405800, 0x08000000, 0x08000000); + nvkm_mask(device, 0x419c00, 0x00000008, 0x00000008); +} + +void +gv100_grctx_unkn88c(struct gf100_gr *gr, bool on) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 mask = 0x00000010, data = on ? mask : 0x00000000; + nvkm_mask(device, 0x40988c, mask, data); + nvkm_rd32(device, 0x40988c); + nvkm_mask(device, 0x41a88c, mask, data); + nvkm_rd32(device, 0x41a88c); + nvkm_mask(device, 0x408a14, mask, data); + nvkm_rd32(device, 0x408a14); +} + +const struct gf100_grctx_func +gv100_grctx = { + .unkn88c = gv100_grctx_unkn88c, + .main = gf100_grctx_generate_main, + .unkn = gv100_grctx_generate_unkn, + .sw_veid_bundle_init = gv100_grctx_pack_sw_veid_bundle_init, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0x1680, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gv100_grctx_generate_attrib_cb, + .attrib = gv100_grctx_generate_attrib, + .attrib_nr_max = 0x6c0, + .attrib_nr = 0x480, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .gfxp_nr = 0xd10, + .sm_id = gv100_grctx_generate_sm_id, + .rop_mapping = gv100_grctx_generate_rop_mapping, + .dist_skip_table = gm200_grctx_generate_dist_skip_table, + .r406500 = gm200_grctx_generate_r406500, + .gpc_tpc_nr = gk104_grctx_generate_gpc_tpc_nr, + .smid_config = gp100_grctx_generate_smid_config, + .r400088 = gv100_grctx_generate_r400088, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c new file mode 100644 index 0000000000..80a6b017af --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c @@ -0,0 +1,693 @@ +/* + * Copyright 2009 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +/* NVIDIA context programs handle a number of other conditions which are + * not implemented in our versions. It's not clear why NVIDIA context + * programs have this code, nor whether it's strictly necessary for + * correct operation. We'll implement additional handling if/when we + * discover it's necessary. + * + * - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state" + * flag is set, this gets saved into the context. + * - On context save, the context program for all cards load nsource + * into a flag register and check for ILLEGAL_MTHD. If it's set, + * opcode 0x60000d is called before resuming normal operation. + * - Some context programs check more conditions than the above. NV44 + * checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001)) + * and calls 0x60000d before resuming normal operation. + * - At the very beginning of NVIDIA's context programs, flag 9 is checked + * and if true 0x800001 is called with count=0, pos=0, the flag is cleared + * and then the ctxprog is aborted. It looks like a complicated NOP, + * its purpose is unknown. + * - In the section of code that loads the per-vs state, NVIDIA check + * flag 10. If it's set, they only transfer the small 0x300 byte block + * of state + the state for a single vs as opposed to the state for + * all vs units. It doesn't seem likely that it'll occur in normal + * operation, especially seeing as it appears NVIDIA may have screwed + * up the ctxprogs for some cards and have an invalid instruction + * rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction. + * - There's a number of places where context offset 0 (where we place + * the PRAMIN offset of the context) is loaded into either 0x408000, + * 0x408004 or 0x408008. Not sure what's up there either. + * - The ctxprogs for some cards save 0x400a00 again during the cleanup + * path for auto-loadctx. + */ + +#define CP_FLAG_CLEAR 0 +#define CP_FLAG_SET 1 +#define CP_FLAG_SWAP_DIRECTION ((0 * 32) + 0) +#define CP_FLAG_SWAP_DIRECTION_LOAD 0 +#define CP_FLAG_SWAP_DIRECTION_SAVE 1 +#define CP_FLAG_USER_SAVE ((0 * 32) + 5) +#define CP_FLAG_USER_SAVE_NOT_PENDING 0 +#define CP_FLAG_USER_SAVE_PENDING 1 +#define CP_FLAG_USER_LOAD ((0 * 32) + 6) +#define CP_FLAG_USER_LOAD_NOT_PENDING 0 +#define CP_FLAG_USER_LOAD_PENDING 1 +#define CP_FLAG_STATUS ((3 * 32) + 0) +#define CP_FLAG_STATUS_IDLE 0 +#define CP_FLAG_STATUS_BUSY 1 +#define CP_FLAG_AUTO_SAVE ((3 * 32) + 4) +#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0 +#define CP_FLAG_AUTO_SAVE_PENDING 1 +#define CP_FLAG_AUTO_LOAD ((3 * 32) + 5) +#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0 +#define CP_FLAG_AUTO_LOAD_PENDING 1 +#define CP_FLAG_UNK54 ((3 * 32) + 6) +#define CP_FLAG_UNK54_CLEAR 0 +#define CP_FLAG_UNK54_SET 1 +#define CP_FLAG_ALWAYS ((3 * 32) + 8) +#define CP_FLAG_ALWAYS_FALSE 0 +#define CP_FLAG_ALWAYS_TRUE 1 +#define CP_FLAG_UNK57 ((3 * 32) + 9) +#define CP_FLAG_UNK57_CLEAR 0 +#define CP_FLAG_UNK57_SET 1 + +#define CP_CTX 0x00100000 +#define CP_CTX_COUNT 0x000fc000 +#define CP_CTX_COUNT_SHIFT 14 +#define CP_CTX_REG 0x00003fff +#define CP_LOAD_SR 0x00200000 +#define CP_LOAD_SR_VALUE 0x000fffff +#define CP_BRA 0x00400000 +#define CP_BRA_IP 0x0000ff00 +#define CP_BRA_IP_SHIFT 8 +#define CP_BRA_IF_CLEAR 0x00000080 +#define CP_BRA_FLAG 0x0000007f +#define CP_WAIT 0x00500000 +#define CP_WAIT_SET 0x00000080 +#define CP_WAIT_FLAG 0x0000007f +#define CP_SET 0x00700000 +#define CP_SET_1 0x00000080 +#define CP_SET_FLAG 0x0000007f +#define CP_NEXT_TO_SWAP 0x00600007 +#define CP_NEXT_TO_CURRENT 0x00600009 +#define CP_SET_CONTEXT_POINTER 0x0060000a +#define CP_END 0x0060000e +#define CP_LOAD_MAGIC_UNK01 0x00800001 /* unknown */ +#define CP_LOAD_MAGIC_NV44TCL 0x00800029 /* per-vs state (0x4497) */ +#define CP_LOAD_MAGIC_NV40TCL 0x00800041 /* per-vs state (0x4097) */ + +#include "ctxnv40.h" +#include "nv40.h" + +/* TODO: + * - get vs count from 0x1540 + */ + +static int +nv40_gr_vs_count(struct nvkm_device *device) +{ + + switch (device->chipset) { + case 0x47: + case 0x49: + case 0x4b: + return 8; + case 0x40: + return 6; + case 0x41: + case 0x42: + return 5; + case 0x43: + case 0x44: + case 0x46: + case 0x4a: + return 3; + case 0x4c: + case 0x4e: + case 0x67: + default: + return 1; + } +} + + +enum cp_label { + cp_check_load = 1, + cp_setup_auto_load, + cp_setup_load, + cp_setup_save, + cp_swap_state, + cp_swap_state3d_3_is_save, + cp_prepare_exit, + cp_exit, +}; + +static void +nv40_gr_construct_general(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + + cp_ctx(ctx, 0x4000a4, 1); + gr_def(ctx, 0x4000a4, 0x00000008); + cp_ctx(ctx, 0x400144, 58); + gr_def(ctx, 0x400144, 0x00000001); + cp_ctx(ctx, 0x400314, 1); + gr_def(ctx, 0x400314, 0x00000000); + cp_ctx(ctx, 0x400400, 10); + cp_ctx(ctx, 0x400480, 10); + cp_ctx(ctx, 0x400500, 19); + gr_def(ctx, 0x400514, 0x00040000); + gr_def(ctx, 0x400524, 0x55555555); + gr_def(ctx, 0x400528, 0x55555555); + gr_def(ctx, 0x40052c, 0x55555555); + gr_def(ctx, 0x400530, 0x55555555); + cp_ctx(ctx, 0x400560, 6); + gr_def(ctx, 0x400568, 0x0000ffff); + gr_def(ctx, 0x40056c, 0x0000ffff); + cp_ctx(ctx, 0x40057c, 5); + cp_ctx(ctx, 0x400710, 3); + gr_def(ctx, 0x400710, 0x20010001); + gr_def(ctx, 0x400714, 0x0f73ef00); + cp_ctx(ctx, 0x400724, 1); + gr_def(ctx, 0x400724, 0x02008821); + cp_ctx(ctx, 0x400770, 3); + if (device->chipset == 0x40) { + cp_ctx(ctx, 0x400814, 4); + cp_ctx(ctx, 0x400828, 5); + cp_ctx(ctx, 0x400840, 5); + gr_def(ctx, 0x400850, 0x00000040); + cp_ctx(ctx, 0x400858, 4); + gr_def(ctx, 0x400858, 0x00000040); + gr_def(ctx, 0x40085c, 0x00000040); + gr_def(ctx, 0x400864, 0x80000000); + cp_ctx(ctx, 0x40086c, 9); + gr_def(ctx, 0x40086c, 0x80000000); + gr_def(ctx, 0x400870, 0x80000000); + gr_def(ctx, 0x400874, 0x80000000); + gr_def(ctx, 0x400878, 0x80000000); + gr_def(ctx, 0x400888, 0x00000040); + gr_def(ctx, 0x40088c, 0x80000000); + cp_ctx(ctx, 0x4009c0, 8); + gr_def(ctx, 0x4009cc, 0x80000000); + gr_def(ctx, 0x4009dc, 0x80000000); + } else { + cp_ctx(ctx, 0x400840, 20); + if (nv44_gr_class(ctx->device)) { + for (i = 0; i < 8; i++) + gr_def(ctx, 0x400860 + (i * 4), 0x00000001); + } + gr_def(ctx, 0x400880, 0x00000040); + gr_def(ctx, 0x400884, 0x00000040); + gr_def(ctx, 0x400888, 0x00000040); + cp_ctx(ctx, 0x400894, 11); + gr_def(ctx, 0x400894, 0x00000040); + if (!nv44_gr_class(ctx->device)) { + for (i = 0; i < 8; i++) + gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000); + } + cp_ctx(ctx, 0x4008e0, 2); + cp_ctx(ctx, 0x4008f8, 2); + if (device->chipset == 0x4c || + (device->chipset & 0xf0) == 0x60) + cp_ctx(ctx, 0x4009f8, 1); + } + cp_ctx(ctx, 0x400a00, 73); + gr_def(ctx, 0x400b0c, 0x0b0b0b0c); + cp_ctx(ctx, 0x401000, 4); + cp_ctx(ctx, 0x405004, 1); + switch (device->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x403448, 1); + gr_def(ctx, 0x403448, 0x00001010); + break; + default: + cp_ctx(ctx, 0x403440, 1); + switch (device->chipset) { + case 0x40: + gr_def(ctx, 0x403440, 0x00000010); + break; + case 0x44: + case 0x46: + case 0x4a: + gr_def(ctx, 0x403440, 0x00003010); + break; + case 0x41: + case 0x42: + case 0x43: + case 0x4c: + case 0x4e: + case 0x67: + default: + gr_def(ctx, 0x403440, 0x00001010); + break; + } + break; + } +} + +static void +nv40_gr_construct_state3d(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + + if (device->chipset == 0x40) { + cp_ctx(ctx, 0x401880, 51); + gr_def(ctx, 0x401940, 0x00000100); + } else + if (device->chipset == 0x46 || device->chipset == 0x47 || + device->chipset == 0x49 || device->chipset == 0x4b) { + cp_ctx(ctx, 0x401880, 32); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x401880 + (i * 4), 0x00000111); + if (device->chipset == 0x46) + cp_ctx(ctx, 0x401900, 16); + cp_ctx(ctx, 0x401940, 3); + } + cp_ctx(ctx, 0x40194c, 18); + gr_def(ctx, 0x401954, 0x00000111); + gr_def(ctx, 0x401958, 0x00080060); + gr_def(ctx, 0x401974, 0x00000080); + gr_def(ctx, 0x401978, 0xffff0000); + gr_def(ctx, 0x40197c, 0x00000001); + gr_def(ctx, 0x401990, 0x46400000); + if (device->chipset == 0x40) { + cp_ctx(ctx, 0x4019a0, 2); + cp_ctx(ctx, 0x4019ac, 5); + } else { + cp_ctx(ctx, 0x4019a0, 1); + cp_ctx(ctx, 0x4019b4, 3); + } + gr_def(ctx, 0x4019bc, 0xffff0000); + switch (device->chipset) { + case 0x46: + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x4019c0, 18); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888); + break; + } + cp_ctx(ctx, 0x401a08, 8); + gr_def(ctx, 0x401a10, 0x0fff0000); + gr_def(ctx, 0x401a14, 0x0fff0000); + gr_def(ctx, 0x401a1c, 0x00011100); + cp_ctx(ctx, 0x401a2c, 4); + cp_ctx(ctx, 0x401a44, 26); + for (i = 0; i < 16; i++) + gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000); + gr_def(ctx, 0x401a8c, 0x4b7fffff); + if (device->chipset == 0x40) { + cp_ctx(ctx, 0x401ab8, 3); + } else { + cp_ctx(ctx, 0x401ab8, 1); + cp_ctx(ctx, 0x401ac0, 1); + } + cp_ctx(ctx, 0x401ad0, 8); + gr_def(ctx, 0x401ad0, 0x30201000); + gr_def(ctx, 0x401ad4, 0x70605040); + gr_def(ctx, 0x401ad8, 0xb8a89888); + gr_def(ctx, 0x401adc, 0xf8e8d8c8); + cp_ctx(ctx, 0x401b10, device->chipset == 0x40 ? 2 : 1); + gr_def(ctx, 0x401b10, 0x40100000); + cp_ctx(ctx, 0x401b18, device->chipset == 0x40 ? 6 : 5); + gr_def(ctx, 0x401b28, device->chipset == 0x40 ? + 0x00000004 : 0x00000000); + cp_ctx(ctx, 0x401b30, 25); + gr_def(ctx, 0x401b34, 0x0000ffff); + gr_def(ctx, 0x401b68, 0x435185d6); + gr_def(ctx, 0x401b6c, 0x2155b699); + gr_def(ctx, 0x401b70, 0xfedcba98); + gr_def(ctx, 0x401b74, 0x00000098); + gr_def(ctx, 0x401b84, 0xffffffff); + gr_def(ctx, 0x401b88, 0x00ff7000); + gr_def(ctx, 0x401b8c, 0x0000ffff); + if (device->chipset != 0x44 && device->chipset != 0x4a && + device->chipset != 0x4e) + cp_ctx(ctx, 0x401b94, 1); + cp_ctx(ctx, 0x401b98, 8); + gr_def(ctx, 0x401b9c, 0x00ff0000); + cp_ctx(ctx, 0x401bc0, 9); + gr_def(ctx, 0x401be0, 0x00ffff00); + cp_ctx(ctx, 0x401c00, 192); + for (i = 0; i < 16; i++) { /* fragment texture units */ + gr_def(ctx, 0x401c40 + (i * 4), 0x00018488); + gr_def(ctx, 0x401c80 + (i * 4), 0x00028202); + gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4); + gr_def(ctx, 0x401d40 + (i * 4), 0x01012000); + gr_def(ctx, 0x401d80 + (i * 4), 0x00080008); + gr_def(ctx, 0x401e00 + (i * 4), 0x00100008); + } + for (i = 0; i < 4; i++) { /* vertex texture units */ + gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80); + gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202); + gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008); + gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008); + } + cp_ctx(ctx, 0x400f5c, 3); + gr_def(ctx, 0x400f5c, 0x00000002); + cp_ctx(ctx, 0x400f84, 1); +} + +static void +nv40_gr_construct_state3d_2(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + + cp_ctx(ctx, 0x402000, 1); + cp_ctx(ctx, 0x402404, device->chipset == 0x40 ? 1 : 2); + switch (device->chipset) { + case 0x40: + gr_def(ctx, 0x402404, 0x00000001); + break; + case 0x4c: + case 0x4e: + case 0x67: + gr_def(ctx, 0x402404, 0x00000020); + break; + case 0x46: + case 0x49: + case 0x4b: + gr_def(ctx, 0x402404, 0x00000421); + break; + default: + gr_def(ctx, 0x402404, 0x00000021); + } + if (device->chipset != 0x40) + gr_def(ctx, 0x402408, 0x030c30c3); + switch (device->chipset) { + case 0x44: + case 0x46: + case 0x4a: + case 0x4c: + case 0x4e: + case 0x67: + cp_ctx(ctx, 0x402440, 1); + gr_def(ctx, 0x402440, 0x00011001); + break; + default: + break; + } + cp_ctx(ctx, 0x402480, device->chipset == 0x40 ? 8 : 9); + gr_def(ctx, 0x402488, 0x3e020200); + gr_def(ctx, 0x40248c, 0x00ffffff); + switch (device->chipset) { + case 0x40: + gr_def(ctx, 0x402490, 0x60103f00); + break; + case 0x47: + gr_def(ctx, 0x402490, 0x40103f00); + break; + case 0x41: + case 0x42: + case 0x49: + case 0x4b: + gr_def(ctx, 0x402490, 0x20103f00); + break; + default: + gr_def(ctx, 0x402490, 0x0c103f00); + break; + } + gr_def(ctx, 0x40249c, device->chipset <= 0x43 ? + 0x00020000 : 0x00040000); + cp_ctx(ctx, 0x402500, 31); + gr_def(ctx, 0x402530, 0x00008100); + if (device->chipset == 0x40) + cp_ctx(ctx, 0x40257c, 6); + cp_ctx(ctx, 0x402594, 16); + cp_ctx(ctx, 0x402800, 17); + gr_def(ctx, 0x402800, 0x00000001); + switch (device->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x402864, 1); + gr_def(ctx, 0x402864, 0x00001001); + cp_ctx(ctx, 0x402870, 3); + gr_def(ctx, 0x402878, 0x00000003); + if (device->chipset != 0x47) { /* belong at end!! */ + cp_ctx(ctx, 0x402900, 1); + cp_ctx(ctx, 0x402940, 1); + cp_ctx(ctx, 0x402980, 1); + cp_ctx(ctx, 0x4029c0, 1); + cp_ctx(ctx, 0x402a00, 1); + cp_ctx(ctx, 0x402a40, 1); + cp_ctx(ctx, 0x402a80, 1); + cp_ctx(ctx, 0x402ac0, 1); + } + break; + case 0x40: + cp_ctx(ctx, 0x402844, 1); + gr_def(ctx, 0x402844, 0x00000001); + cp_ctx(ctx, 0x402850, 1); + break; + default: + cp_ctx(ctx, 0x402844, 1); + gr_def(ctx, 0x402844, 0x00001001); + cp_ctx(ctx, 0x402850, 2); + gr_def(ctx, 0x402854, 0x00000003); + break; + } + + cp_ctx(ctx, 0x402c00, 4); + gr_def(ctx, 0x402c00, device->chipset == 0x40 ? + 0x80800001 : 0x00888001); + switch (device->chipset) { + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x402c20, 40); + for (i = 0; i < 32; i++) + gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff); + cp_ctx(ctx, 0x4030b8, 13); + gr_def(ctx, 0x4030dc, 0x00000005); + gr_def(ctx, 0x4030e8, 0x0000ffff); + break; + default: + cp_ctx(ctx, 0x402c10, 4); + if (device->chipset == 0x40) + cp_ctx(ctx, 0x402c20, 36); + else + if (device->chipset <= 0x42) + cp_ctx(ctx, 0x402c20, 24); + else + if (device->chipset <= 0x4a) + cp_ctx(ctx, 0x402c20, 16); + else + cp_ctx(ctx, 0x402c20, 8); + cp_ctx(ctx, 0x402cb0, device->chipset == 0x40 ? 12 : 13); + gr_def(ctx, 0x402cd4, 0x00000005); + if (device->chipset != 0x40) + gr_def(ctx, 0x402ce0, 0x0000ffff); + break; + } + + cp_ctx(ctx, 0x403400, device->chipset == 0x40 ? 4 : 3); + cp_ctx(ctx, 0x403410, device->chipset == 0x40 ? 4 : 3); + cp_ctx(ctx, 0x403420, nv40_gr_vs_count(ctx->device)); + for (i = 0; i < nv40_gr_vs_count(ctx->device); i++) + gr_def(ctx, 0x403420 + (i * 4), 0x00005555); + + if (device->chipset != 0x40) { + cp_ctx(ctx, 0x403600, 1); + gr_def(ctx, 0x403600, 0x00000001); + } + cp_ctx(ctx, 0x403800, 1); + + cp_ctx(ctx, 0x403c18, 1); + gr_def(ctx, 0x403c18, 0x00000001); + switch (device->chipset) { + case 0x46: + case 0x47: + case 0x49: + case 0x4b: + cp_ctx(ctx, 0x405018, 1); + gr_def(ctx, 0x405018, 0x08e00001); + cp_ctx(ctx, 0x405c24, 1); + gr_def(ctx, 0x405c24, 0x000e3000); + break; + } + if (device->chipset != 0x4e) + cp_ctx(ctx, 0x405800, 11); + cp_ctx(ctx, 0x407000, 1); +} + +static void +nv40_gr_construct_state3d_3(struct nvkm_grctx *ctx) +{ + int len = nv44_gr_class(ctx->device) ? 0x0084 : 0x0684; + + cp_out (ctx, 0x300000); + cp_lsr (ctx, len - 4); + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save); + cp_lsr (ctx, len); + cp_name(ctx, cp_swap_state3d_3_is_save); + cp_out (ctx, 0x800001); + + ctx->ctxvals_pos += len; +} + +static void +nv40_gr_construct_shader(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + struct nvkm_gpuobj *obj = ctx->data; + int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset; + int offset, i; + + vs_nr = nv40_gr_vs_count(ctx->device); + vs_nr_b0 = 363; + vs_nr_b1 = device->chipset == 0x40 ? 128 : 64; + if (device->chipset == 0x40) { + b0_offset = 0x2200/4; /* 33a0 */ + b1_offset = 0x55a0/4; /* 1500 */ + vs_len = 0x6aa0/4; + } else + if (device->chipset == 0x41 || device->chipset == 0x42) { + b0_offset = 0x2200/4; /* 2200 */ + b1_offset = 0x4400/4; /* 0b00 */ + vs_len = 0x4f00/4; + } else { + b0_offset = 0x1d40/4; /* 2200 */ + b1_offset = 0x3f40/4; /* 0b00 : 0a40 */ + vs_len = nv44_gr_class(device) ? 0x4980/4 : 0x4a40/4; + } + + cp_lsr(ctx, vs_len * vs_nr + 0x300/4); + cp_out(ctx, nv44_gr_class(device) ? 0x800029 : 0x800041); + + offset = ctx->ctxvals_pos; + ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len)); + + if (ctx->mode != NVKM_GRCTX_VALS) + return; + + offset += 0x0280/4; + for (i = 0; i < 16; i++, offset += 2) + nvkm_wo32(obj, offset * 4, 0x3f800000); + + for (vs = 0; vs < vs_nr; vs++, offset += vs_len) { + for (i = 0; i < vs_nr_b0 * 6; i += 6) + nvkm_wo32(obj, (offset + b0_offset + i) * 4, 0x00000001); + for (i = 0; i < vs_nr_b1 * 4; i += 4) + nvkm_wo32(obj, (offset + b1_offset + i) * 4, 0x3f800000); + } +} + +static void +nv40_grctx_generate(struct nvkm_grctx *ctx) +{ + /* decide whether we're loading/unloading the context */ + cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save); + cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save); + + cp_name(ctx, cp_check_load); + cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load); + cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load); + cp_bra (ctx, ALWAYS, TRUE, cp_exit); + + /* setup for context load */ + cp_name(ctx, cp_setup_auto_load); + cp_wait(ctx, STATUS, IDLE); + cp_out (ctx, CP_NEXT_TO_SWAP); + cp_name(ctx, cp_setup_load); + cp_wait(ctx, STATUS, IDLE); + cp_set (ctx, SWAP_DIRECTION, LOAD); + cp_out (ctx, 0x00910880); /* ?? */ + cp_out (ctx, 0x00901ffe); /* ?? */ + cp_out (ctx, 0x01940000); /* ?? */ + cp_lsr (ctx, 0x20); + cp_out (ctx, 0x0060000b); /* ?? */ + cp_wait(ctx, UNK57, CLEAR); + cp_out (ctx, 0x0060000c); /* ?? */ + cp_bra (ctx, ALWAYS, TRUE, cp_swap_state); + + /* setup for context save */ + cp_name(ctx, cp_setup_save); + cp_set (ctx, SWAP_DIRECTION, SAVE); + + /* general PGRAPH state */ + cp_name(ctx, cp_swap_state); + cp_pos (ctx, 0x00020/4); + nv40_gr_construct_general(ctx); + cp_wait(ctx, STATUS, IDLE); + + /* 3D state, block 1 */ + cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit); + nv40_gr_construct_state3d(ctx); + cp_wait(ctx, STATUS, IDLE); + + /* 3D state, block 2 */ + nv40_gr_construct_state3d_2(ctx); + + /* Some other block of "random" state */ + nv40_gr_construct_state3d_3(ctx); + + /* Per-vertex shader state */ + cp_pos (ctx, ctx->ctxvals_pos); + nv40_gr_construct_shader(ctx); + + /* pre-exit state updates */ + cp_name(ctx, cp_prepare_exit); + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load); + cp_bra (ctx, USER_SAVE, PENDING, cp_exit); + cp_out (ctx, CP_NEXT_TO_CURRENT); + + cp_name(ctx, cp_exit); + cp_set (ctx, USER_SAVE, NOT_PENDING); + cp_set (ctx, USER_LOAD, NOT_PENDING); + cp_out (ctx, CP_END); +} + +void +nv40_grctx_fill(struct nvkm_device *device, struct nvkm_gpuobj *mem) +{ + nv40_grctx_generate(&(struct nvkm_grctx) { + .device = device, + .mode = NVKM_GRCTX_VALS, + .data = mem, + }); +} + +int +nv40_grctx_init(struct nvkm_device *device, u32 *size) +{ + u32 *ctxprog = kmalloc(256 * 4, GFP_KERNEL), i; + struct nvkm_grctx ctx = { + .device = device, + .mode = NVKM_GRCTX_PROG, + .ucode = ctxprog, + .ctxprog_max = 256, + }; + + if (!ctxprog) + return -ENOMEM; + + nv40_grctx_generate(&ctx); + + nvkm_wr32(device, 0x400324, 0); + for (i = 0; i < ctx.ctxprog_len; i++) + nvkm_wr32(device, 0x400328, ctxprog[i]); + *size = ctx.ctxvals_pos * 4; + + kfree(ctxprog); + return 0; +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h new file mode 100644 index 0000000000..7917567ade --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h @@ -0,0 +1,131 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_GRCTX_H__ +#define __NVKM_GRCTX_H__ +#include + +struct nvkm_grctx { + struct nvkm_device *device; + + enum { + NVKM_GRCTX_PROG, + NVKM_GRCTX_VALS + } mode; + u32 *ucode; + struct nvkm_gpuobj *data; + + u32 ctxprog_max; + u32 ctxprog_len; + u32 ctxprog_reg; + int ctxprog_label[32]; + u32 ctxvals_pos; + u32 ctxvals_base; +}; + +static inline void +cp_out(struct nvkm_grctx *ctx, u32 inst) +{ + u32 *ctxprog = ctx->ucode; + + if (ctx->mode != NVKM_GRCTX_PROG) + return; + + BUG_ON(ctx->ctxprog_len == ctx->ctxprog_max); + ctxprog[ctx->ctxprog_len++] = inst; +} + +static inline void +cp_lsr(struct nvkm_grctx *ctx, u32 val) +{ + cp_out(ctx, CP_LOAD_SR | val); +} + +static inline void +cp_ctx(struct nvkm_grctx *ctx, u32 reg, u32 length) +{ + ctx->ctxprog_reg = (reg - 0x00400000) >> 2; + + ctx->ctxvals_base = ctx->ctxvals_pos; + ctx->ctxvals_pos = ctx->ctxvals_base + length; + + if (length > (CP_CTX_COUNT >> CP_CTX_COUNT_SHIFT)) { + cp_lsr(ctx, length); + length = 0; + } + + cp_out(ctx, CP_CTX | (length << CP_CTX_COUNT_SHIFT) | ctx->ctxprog_reg); +} + +static inline void +cp_name(struct nvkm_grctx *ctx, int name) +{ + u32 *ctxprog = ctx->ucode; + int i; + + if (ctx->mode != NVKM_GRCTX_PROG) + return; + + ctx->ctxprog_label[name] = ctx->ctxprog_len; + for (i = 0; i < ctx->ctxprog_len; i++) { + if ((ctxprog[i] & 0xfff00000) != 0xff400000) + continue; + if ((ctxprog[i] & CP_BRA_IP) != ((name) << CP_BRA_IP_SHIFT)) + continue; + ctxprog[i] = (ctxprog[i] & 0x00ff00ff) | + (ctx->ctxprog_len << CP_BRA_IP_SHIFT); + } +} + +static inline void +_cp_bra(struct nvkm_grctx *ctx, u32 mod, int flag, int state, int name) +{ + int ip = 0; + + if (mod != 2) { + ip = ctx->ctxprog_label[name] << CP_BRA_IP_SHIFT; + if (ip == 0) + ip = 0xff000000 | (name << CP_BRA_IP_SHIFT); + } + + cp_out(ctx, CP_BRA | (mod << 18) | ip | flag | + (state ? 0 : CP_BRA_IF_CLEAR)); +} +#define cp_bra(c, f, s, n) _cp_bra((c), 0, CP_FLAG_##f, CP_FLAG_##f##_##s, n) +#define cp_cal(c, f, s, n) _cp_bra((c), 1, CP_FLAG_##f, CP_FLAG_##f##_##s, n) +#define cp_ret(c, f, s) _cp_bra((c), 2, CP_FLAG_##f, CP_FLAG_##f##_##s, 0) + +static inline void +_cp_wait(struct nvkm_grctx *ctx, int flag, int state) +{ + cp_out(ctx, CP_WAIT | flag | (state ? CP_WAIT_SET : 0)); +} +#define cp_wait(c, f, s) _cp_wait((c), CP_FLAG_##f, CP_FLAG_##f##_##s) + +static inline void +_cp_set(struct nvkm_grctx *ctx, int flag, int state) +{ + cp_out(ctx, CP_SET | flag | (state ? CP_SET_1 : 0)); +} +#define cp_set(c, f, s) _cp_set((c), CP_FLAG_##f, CP_FLAG_##f##_##s) + +static inline void +cp_pos(struct nvkm_grctx *ctx, int offset) +{ + ctx->ctxvals_pos = offset; + ctx->ctxvals_base = ctx->ctxvals_pos; + + cp_lsr(ctx, ctx->ctxvals_pos); + cp_out(ctx, CP_SET_CONTEXT_POINTER); +} + +static inline void +gr_def(struct nvkm_grctx *ctx, u32 reg, u32 val) +{ + if (ctx->mode != NVKM_GRCTX_VALS) + return; + + reg = (reg - 0x00400000) / 4; + reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base; + + nvkm_wo32(ctx->data, reg * 4, val); +} +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c new file mode 100644 index 0000000000..c8bb9191f9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c @@ -0,0 +1,3347 @@ +/* + * Copyright 2009 Marcin Kościelnicki + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#define CP_FLAG_CLEAR 0 +#define CP_FLAG_SET 1 +#define CP_FLAG_SWAP_DIRECTION ((0 * 32) + 0) +#define CP_FLAG_SWAP_DIRECTION_LOAD 0 +#define CP_FLAG_SWAP_DIRECTION_SAVE 1 +#define CP_FLAG_UNK01 ((0 * 32) + 1) +#define CP_FLAG_UNK01_CLEAR 0 +#define CP_FLAG_UNK01_SET 1 +#define CP_FLAG_UNK03 ((0 * 32) + 3) +#define CP_FLAG_UNK03_CLEAR 0 +#define CP_FLAG_UNK03_SET 1 +#define CP_FLAG_USER_SAVE ((0 * 32) + 5) +#define CP_FLAG_USER_SAVE_NOT_PENDING 0 +#define CP_FLAG_USER_SAVE_PENDING 1 +#define CP_FLAG_USER_LOAD ((0 * 32) + 6) +#define CP_FLAG_USER_LOAD_NOT_PENDING 0 +#define CP_FLAG_USER_LOAD_PENDING 1 +#define CP_FLAG_UNK0B ((0 * 32) + 0xb) +#define CP_FLAG_UNK0B_CLEAR 0 +#define CP_FLAG_UNK0B_SET 1 +#define CP_FLAG_XFER_SWITCH ((0 * 32) + 0xe) +#define CP_FLAG_XFER_SWITCH_DISABLE 0 +#define CP_FLAG_XFER_SWITCH_ENABLE 1 +#define CP_FLAG_STATE ((0 * 32) + 0x1c) +#define CP_FLAG_STATE_STOPPED 0 +#define CP_FLAG_STATE_RUNNING 1 +#define CP_FLAG_UNK1D ((0 * 32) + 0x1d) +#define CP_FLAG_UNK1D_CLEAR 0 +#define CP_FLAG_UNK1D_SET 1 +#define CP_FLAG_UNK20 ((1 * 32) + 0) +#define CP_FLAG_UNK20_CLEAR 0 +#define CP_FLAG_UNK20_SET 1 +#define CP_FLAG_STATUS ((2 * 32) + 0) +#define CP_FLAG_STATUS_BUSY 0 +#define CP_FLAG_STATUS_IDLE 1 +#define CP_FLAG_AUTO_SAVE ((2 * 32) + 4) +#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0 +#define CP_FLAG_AUTO_SAVE_PENDING 1 +#define CP_FLAG_AUTO_LOAD ((2 * 32) + 5) +#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0 +#define CP_FLAG_AUTO_LOAD_PENDING 1 +#define CP_FLAG_NEWCTX ((2 * 32) + 10) +#define CP_FLAG_NEWCTX_BUSY 0 +#define CP_FLAG_NEWCTX_DONE 1 +#define CP_FLAG_XFER ((2 * 32) + 11) +#define CP_FLAG_XFER_IDLE 0 +#define CP_FLAG_XFER_BUSY 1 +#define CP_FLAG_ALWAYS ((2 * 32) + 13) +#define CP_FLAG_ALWAYS_FALSE 0 +#define CP_FLAG_ALWAYS_TRUE 1 +#define CP_FLAG_INTR ((2 * 32) + 15) +#define CP_FLAG_INTR_NOT_PENDING 0 +#define CP_FLAG_INTR_PENDING 1 + +#define CP_CTX 0x00100000 +#define CP_CTX_COUNT 0x000f0000 +#define CP_CTX_COUNT_SHIFT 16 +#define CP_CTX_REG 0x00003fff +#define CP_LOAD_SR 0x00200000 +#define CP_LOAD_SR_VALUE 0x000fffff +#define CP_BRA 0x00400000 +#define CP_BRA_IP 0x0001ff00 +#define CP_BRA_IP_SHIFT 8 +#define CP_BRA_IF_CLEAR 0x00000080 +#define CP_BRA_FLAG 0x0000007f +#define CP_WAIT 0x00500000 +#define CP_WAIT_SET 0x00000080 +#define CP_WAIT_FLAG 0x0000007f +#define CP_SET 0x00700000 +#define CP_SET_1 0x00000080 +#define CP_SET_FLAG 0x0000007f +#define CP_NEWCTX 0x00600004 +#define CP_NEXT_TO_SWAP 0x00600005 +#define CP_SET_CONTEXT_POINTER 0x00600006 +#define CP_SET_XFER_POINTER 0x00600007 +#define CP_ENABLE 0x00600009 +#define CP_END 0x0060000c +#define CP_NEXT_TO_CURRENT 0x0060000d +#define CP_DISABLE1 0x0090ffff +#define CP_DISABLE2 0x0091ffff +#define CP_XFER_1 0x008000ff +#define CP_XFER_2 0x008800ff +#define CP_SEEK_1 0x00c000ff +#define CP_SEEK_2 0x00c800ff + +#include "ctxnv40.h" +#include "nv50.h" + +#include + +#define IS_NVA3F(x) (((x) > 0xa0 && (x) < 0xaa) || (x) == 0xaf) +#define IS_NVAAF(x) ((x) >= 0xaa && (x) <= 0xac) + +/* + * This code deals with PGRAPH contexts on NV50 family cards. Like NV40, it's + * the GPU itself that does context-switching, but it needs a special + * microcode to do it. And it's the driver's task to supply this microcode, + * further known as ctxprog, as well as the initial context values, known + * as ctxvals. + * + * Without ctxprog, you cannot switch contexts. Not even in software, since + * the majority of context [xfer strands] isn't accessible directly. You're + * stuck with a single channel, and you also suffer all the problems resulting + * from missing ctxvals, since you cannot load them. + * + * Without ctxvals, you're stuck with PGRAPH's default context. It's enough to + * run 2d operations, but trying to utilise 3d or CUDA will just lock you up, + * since you don't have... some sort of needed setup. + * + * Nouveau will just disable acceleration if not given ctxprog + ctxvals, since + * it's too much hassle to handle no-ctxprog as a special case. + */ + +/* + * How ctxprogs work. + * + * The ctxprog is written in its own kind of microcode, with very small and + * crappy set of available commands. You upload it to a small [512 insns] + * area of memory on PGRAPH, and it'll be run when PFIFO wants PGRAPH to + * switch channel. or when the driver explicitely requests it. Stuff visible + * to ctxprog consists of: PGRAPH MMIO registers, PGRAPH context strands, + * the per-channel context save area in VRAM [known as ctxvals or grctx], + * 4 flags registers, a scratch register, two grctx pointers, plus many + * random poorly-understood details. + * + * When ctxprog runs, it's supposed to check what operations are asked of it, + * save old context if requested, optionally reset PGRAPH and switch to the + * new channel, and load the new context. Context consists of three major + * parts: subset of MMIO registers and two "xfer areas". + */ + +/* TODO: + * - document unimplemented bits compared to nvidia + * - NVAx: make a TP subroutine, use it. + * - use 0x4008fc instead of 0x1540? + */ + +enum cp_label { + cp_check_load = 1, + cp_setup_auto_load, + cp_setup_load, + cp_setup_save, + cp_swap_state, + cp_prepare_exit, + cp_exit, +}; + +static void nv50_gr_construct_mmio(struct nvkm_grctx *ctx); +static void nv50_gr_construct_xfer1(struct nvkm_grctx *ctx); +static void nv50_gr_construct_xfer2(struct nvkm_grctx *ctx); + +/* Main function: construct the ctxprog skeleton, call the other functions. */ + +static int +nv50_grctx_generate(struct nvkm_grctx *ctx) +{ + cp_set (ctx, STATE, RUNNING); + cp_set (ctx, XFER_SWITCH, ENABLE); + /* decide whether we're loading/unloading the context */ + cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save); + cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save); + + cp_name(ctx, cp_check_load); + cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load); + cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load); + cp_bra (ctx, ALWAYS, TRUE, cp_prepare_exit); + + /* setup for context load */ + cp_name(ctx, cp_setup_auto_load); + cp_out (ctx, CP_DISABLE1); + cp_out (ctx, CP_DISABLE2); + cp_out (ctx, CP_ENABLE); + cp_out (ctx, CP_NEXT_TO_SWAP); + cp_set (ctx, UNK01, SET); + cp_name(ctx, cp_setup_load); + cp_out (ctx, CP_NEWCTX); + cp_wait(ctx, NEWCTX, BUSY); + cp_set (ctx, UNK1D, CLEAR); + cp_set (ctx, SWAP_DIRECTION, LOAD); + cp_bra (ctx, UNK0B, SET, cp_prepare_exit); + cp_bra (ctx, ALWAYS, TRUE, cp_swap_state); + + /* setup for context save */ + cp_name(ctx, cp_setup_save); + cp_set (ctx, UNK1D, SET); + cp_wait(ctx, STATUS, BUSY); + cp_wait(ctx, INTR, PENDING); + cp_bra (ctx, STATUS, BUSY, cp_setup_save); + cp_set (ctx, UNK01, SET); + cp_set (ctx, SWAP_DIRECTION, SAVE); + + /* general PGRAPH state */ + cp_name(ctx, cp_swap_state); + cp_set (ctx, UNK03, SET); + cp_pos (ctx, 0x00004/4); + cp_ctx (ctx, 0x400828, 1); /* needed. otherwise, flickering happens. */ + cp_pos (ctx, 0x00100/4); + nv50_gr_construct_mmio(ctx); + nv50_gr_construct_xfer1(ctx); + nv50_gr_construct_xfer2(ctx); + + cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load); + + cp_set (ctx, UNK20, SET); + cp_set (ctx, SWAP_DIRECTION, SAVE); /* no idea why this is needed, but fixes at least one lockup. */ + cp_lsr (ctx, ctx->ctxvals_base); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, 4); + cp_out (ctx, CP_SEEK_1); + cp_out (ctx, CP_XFER_1); + cp_wait(ctx, XFER, BUSY); + + /* pre-exit state updates */ + cp_name(ctx, cp_prepare_exit); + cp_set (ctx, UNK01, CLEAR); + cp_set (ctx, UNK03, CLEAR); + cp_set (ctx, UNK1D, CLEAR); + + cp_bra (ctx, USER_SAVE, PENDING, cp_exit); + cp_out (ctx, CP_NEXT_TO_CURRENT); + + cp_name(ctx, cp_exit); + cp_set (ctx, USER_SAVE, NOT_PENDING); + cp_set (ctx, USER_LOAD, NOT_PENDING); + cp_set (ctx, XFER_SWITCH, DISABLE); + cp_set (ctx, STATE, STOPPED); + cp_out (ctx, CP_END); + ctx->ctxvals_pos += 0x400; /* padding... no idea why you need it */ + + return 0; +} + +void +nv50_grctx_fill(struct nvkm_device *device, struct nvkm_gpuobj *mem) +{ + nv50_grctx_generate(&(struct nvkm_grctx) { + .device = device, + .mode = NVKM_GRCTX_VALS, + .data = mem, + }); +} + +int +nv50_grctx_init(struct nvkm_device *device, u32 *size) +{ + u32 *ctxprog = kmalloc(512 * 4, GFP_KERNEL), i; + struct nvkm_grctx ctx = { + .device = device, + .mode = NVKM_GRCTX_PROG, + .ucode = ctxprog, + .ctxprog_max = 512, + }; + + if (!ctxprog) + return -ENOMEM; + nv50_grctx_generate(&ctx); + + nvkm_wr32(device, 0x400324, 0); + for (i = 0; i < ctx.ctxprog_len; i++) + nvkm_wr32(device, 0x400328, ctxprog[i]); + *size = ctx.ctxvals_pos * 4; + kfree(ctxprog); + return 0; +} + +/* + * Constructs MMIO part of ctxprog and ctxvals. Just a matter of knowing which + * registers to save/restore and the default values for them. + */ + +static void +nv50_gr_construct_mmio_ddata(struct nvkm_grctx *ctx); + +static void +nv50_gr_construct_mmio(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i, j; + int offset, base; + u32 units = nvkm_rd32(device, 0x1540); + + /* 0800: DISPATCH */ + cp_ctx(ctx, 0x400808, 7); + gr_def(ctx, 0x400814, 0x00000030); + cp_ctx(ctx, 0x400834, 0x32); + if (device->chipset == 0x50) { + gr_def(ctx, 0x400834, 0xff400040); + gr_def(ctx, 0x400838, 0xfff00080); + gr_def(ctx, 0x40083c, 0xfff70090); + gr_def(ctx, 0x400840, 0xffe806a8); + } + gr_def(ctx, 0x400844, 0x00000002); + if (IS_NVA3F(device->chipset)) + gr_def(ctx, 0x400894, 0x00001000); + gr_def(ctx, 0x4008e8, 0x00000003); + gr_def(ctx, 0x4008ec, 0x00001000); + if (device->chipset == 0x50) + cp_ctx(ctx, 0x400908, 0xb); + else if (device->chipset < 0xa0) + cp_ctx(ctx, 0x400908, 0xc); + else + cp_ctx(ctx, 0x400908, 0xe); + + if (device->chipset >= 0xa0) + cp_ctx(ctx, 0x400b00, 0x1); + if (IS_NVA3F(device->chipset)) { + cp_ctx(ctx, 0x400b10, 0x1); + gr_def(ctx, 0x400b10, 0x0001629d); + cp_ctx(ctx, 0x400b20, 0x1); + gr_def(ctx, 0x400b20, 0x0001629d); + } + + nv50_gr_construct_mmio_ddata(ctx); + + /* 0C00: VFETCH */ + cp_ctx(ctx, 0x400c08, 0x2); + gr_def(ctx, 0x400c08, 0x0000fe0c); + + /* 1000 */ + if (device->chipset < 0xa0) { + cp_ctx(ctx, 0x401008, 0x4); + gr_def(ctx, 0x401014, 0x00001000); + } else if (!IS_NVA3F(device->chipset)) { + cp_ctx(ctx, 0x401008, 0x5); + gr_def(ctx, 0x401018, 0x00001000); + } else { + cp_ctx(ctx, 0x401008, 0x5); + gr_def(ctx, 0x401018, 0x00004000); + } + + /* 1400 */ + cp_ctx(ctx, 0x401400, 0x8); + cp_ctx(ctx, 0x401424, 0x3); + if (device->chipset == 0x50) + gr_def(ctx, 0x40142c, 0x0001fd87); + else + gr_def(ctx, 0x40142c, 0x00000187); + cp_ctx(ctx, 0x401540, 0x5); + gr_def(ctx, 0x401550, 0x00001018); + + /* 1800: STREAMOUT */ + cp_ctx(ctx, 0x401814, 0x1); + gr_def(ctx, 0x401814, 0x000000ff); + if (device->chipset == 0x50) { + cp_ctx(ctx, 0x40181c, 0xe); + gr_def(ctx, 0x401850, 0x00000004); + } else if (device->chipset < 0xa0) { + cp_ctx(ctx, 0x40181c, 0xf); + gr_def(ctx, 0x401854, 0x00000004); + } else { + cp_ctx(ctx, 0x40181c, 0x13); + gr_def(ctx, 0x401864, 0x00000004); + } + + /* 1C00 */ + cp_ctx(ctx, 0x401c00, 0x1); + switch (device->chipset) { + case 0x50: + gr_def(ctx, 0x401c00, 0x0001005f); + break; + case 0x84: + case 0x86: + case 0x94: + gr_def(ctx, 0x401c00, 0x044d00df); + break; + case 0x92: + case 0x96: + case 0x98: + case 0xa0: + case 0xaa: + case 0xac: + gr_def(ctx, 0x401c00, 0x042500df); + break; + case 0xa3: + case 0xa5: + case 0xa8: + case 0xaf: + gr_def(ctx, 0x401c00, 0x142500df); + break; + } + + /* 2000 */ + + /* 2400 */ + cp_ctx(ctx, 0x402400, 0x1); + if (device->chipset == 0x50) + cp_ctx(ctx, 0x402408, 0x1); + else + cp_ctx(ctx, 0x402408, 0x2); + gr_def(ctx, 0x402408, 0x00000600); + + /* 2800: CSCHED */ + cp_ctx(ctx, 0x402800, 0x1); + if (device->chipset == 0x50) + gr_def(ctx, 0x402800, 0x00000006); + + /* 2C00: ZCULL */ + cp_ctx(ctx, 0x402c08, 0x6); + if (device->chipset != 0x50) + gr_def(ctx, 0x402c14, 0x01000000); + gr_def(ctx, 0x402c18, 0x000000ff); + if (device->chipset == 0x50) + cp_ctx(ctx, 0x402ca0, 0x1); + else + cp_ctx(ctx, 0x402ca0, 0x2); + if (device->chipset < 0xa0) + gr_def(ctx, 0x402ca0, 0x00000400); + else if (!IS_NVA3F(device->chipset)) + gr_def(ctx, 0x402ca0, 0x00000800); + else + gr_def(ctx, 0x402ca0, 0x00000400); + cp_ctx(ctx, 0x402cac, 0x4); + + /* 3000: ENG2D */ + cp_ctx(ctx, 0x403004, 0x1); + gr_def(ctx, 0x403004, 0x00000001); + + /* 3400 */ + if (device->chipset >= 0xa0) { + cp_ctx(ctx, 0x403404, 0x1); + gr_def(ctx, 0x403404, 0x00000001); + } + + /* 5000: CCACHE */ + cp_ctx(ctx, 0x405000, 0x1); + switch (device->chipset) { + case 0x50: + gr_def(ctx, 0x405000, 0x00300080); + break; + case 0x84: + case 0xa0: + case 0xa3: + case 0xa5: + case 0xa8: + case 0xaa: + case 0xac: + case 0xaf: + gr_def(ctx, 0x405000, 0x000e0080); + break; + case 0x86: + case 0x92: + case 0x94: + case 0x96: + case 0x98: + gr_def(ctx, 0x405000, 0x00000080); + break; + } + cp_ctx(ctx, 0x405014, 0x1); + gr_def(ctx, 0x405014, 0x00000004); + cp_ctx(ctx, 0x40501c, 0x1); + cp_ctx(ctx, 0x405024, 0x1); + cp_ctx(ctx, 0x40502c, 0x1); + + /* 6000? */ + if (device->chipset == 0x50) + cp_ctx(ctx, 0x4063e0, 0x1); + + /* 6800: M2MF */ + if (device->chipset < 0x90) { + cp_ctx(ctx, 0x406814, 0x2b); + gr_def(ctx, 0x406818, 0x00000f80); + gr_def(ctx, 0x406860, 0x007f0080); + gr_def(ctx, 0x40689c, 0x007f0080); + } else { + cp_ctx(ctx, 0x406814, 0x4); + if (device->chipset == 0x98) + gr_def(ctx, 0x406818, 0x00000f80); + else + gr_def(ctx, 0x406818, 0x00001f80); + if (IS_NVA3F(device->chipset)) + gr_def(ctx, 0x40681c, 0x00000030); + cp_ctx(ctx, 0x406830, 0x3); + } + + /* 7000: per-ROP group state */ + for (i = 0; i < 8; i++) { + if (units & (1<<(i+16))) { + cp_ctx(ctx, 0x407000 + (i<<8), 3); + if (device->chipset == 0x50) + gr_def(ctx, 0x407000 + (i<<8), 0x1b74f820); + else if (device->chipset != 0xa5) + gr_def(ctx, 0x407000 + (i<<8), 0x3b74f821); + else + gr_def(ctx, 0x407000 + (i<<8), 0x7b74f821); + gr_def(ctx, 0x407004 + (i<<8), 0x89058001); + + if (device->chipset == 0x50) { + cp_ctx(ctx, 0x407010 + (i<<8), 1); + } else if (device->chipset < 0xa0) { + cp_ctx(ctx, 0x407010 + (i<<8), 2); + gr_def(ctx, 0x407010 + (i<<8), 0x00001000); + gr_def(ctx, 0x407014 + (i<<8), 0x0000001f); + } else { + cp_ctx(ctx, 0x407010 + (i<<8), 3); + gr_def(ctx, 0x407010 + (i<<8), 0x00001000); + if (device->chipset != 0xa5) + gr_def(ctx, 0x407014 + (i<<8), 0x000000ff); + else + gr_def(ctx, 0x407014 + (i<<8), 0x000001ff); + } + + cp_ctx(ctx, 0x407080 + (i<<8), 4); + if (device->chipset != 0xa5) + gr_def(ctx, 0x407080 + (i<<8), 0x027c10fa); + else + gr_def(ctx, 0x407080 + (i<<8), 0x827c10fa); + if (device->chipset == 0x50) + gr_def(ctx, 0x407084 + (i<<8), 0x000000c0); + else + gr_def(ctx, 0x407084 + (i<<8), 0x400000c0); + gr_def(ctx, 0x407088 + (i<<8), 0xb7892080); + + if (device->chipset < 0xa0) + cp_ctx(ctx, 0x407094 + (i<<8), 1); + else if (!IS_NVA3F(device->chipset)) + cp_ctx(ctx, 0x407094 + (i<<8), 3); + else { + cp_ctx(ctx, 0x407094 + (i<<8), 4); + gr_def(ctx, 0x4070a0 + (i<<8), 1); + } + } + } + + cp_ctx(ctx, 0x407c00, 0x3); + if (device->chipset < 0x90) + gr_def(ctx, 0x407c00, 0x00010040); + else if (device->chipset < 0xa0) + gr_def(ctx, 0x407c00, 0x00390040); + else + gr_def(ctx, 0x407c00, 0x003d0040); + gr_def(ctx, 0x407c08, 0x00000022); + if (device->chipset >= 0xa0) { + cp_ctx(ctx, 0x407c10, 0x3); + cp_ctx(ctx, 0x407c20, 0x1); + cp_ctx(ctx, 0x407c2c, 0x1); + } + + if (device->chipset < 0xa0) { + cp_ctx(ctx, 0x407d00, 0x9); + } else { + cp_ctx(ctx, 0x407d00, 0x15); + } + if (device->chipset == 0x98) + gr_def(ctx, 0x407d08, 0x00380040); + else { + if (device->chipset < 0x90) + gr_def(ctx, 0x407d08, 0x00010040); + else if (device->chipset < 0xa0) + gr_def(ctx, 0x407d08, 0x00390040); + else { + if (device->fb->ram->type != NVKM_RAM_TYPE_GDDR5) + gr_def(ctx, 0x407d08, 0x003d0040); + else + gr_def(ctx, 0x407d08, 0x003c0040); + } + gr_def(ctx, 0x407d0c, 0x00000022); + } + + /* 8000+: per-TP state */ + for (i = 0; i < 10; i++) { + if (units & (1<chipset < 0xa0) + base = 0x408000 + (i<<12); + else + base = 0x408000 + (i<<11); + if (device->chipset < 0xa0) + offset = base + 0xc00; + else + offset = base + 0x80; + cp_ctx(ctx, offset + 0x00, 1); + gr_def(ctx, offset + 0x00, 0x0000ff0a); + cp_ctx(ctx, offset + 0x08, 1); + + /* per-MP state */ + for (j = 0; j < (device->chipset < 0xa0 ? 2 : 4); j++) { + if (!(units & (1 << (j+24)))) continue; + if (device->chipset < 0xa0) + offset = base + 0x200 + (j<<7); + else + offset = base + 0x100 + (j<<7); + cp_ctx(ctx, offset, 0x20); + gr_def(ctx, offset + 0x00, 0x01800000); + gr_def(ctx, offset + 0x04, 0x00160000); + gr_def(ctx, offset + 0x08, 0x01800000); + gr_def(ctx, offset + 0x18, 0x0003ffff); + switch (device->chipset) { + case 0x50: + gr_def(ctx, offset + 0x1c, 0x00080000); + break; + case 0x84: + gr_def(ctx, offset + 0x1c, 0x00880000); + break; + case 0x86: + gr_def(ctx, offset + 0x1c, 0x018c0000); + break; + case 0x92: + case 0x96: + case 0x98: + gr_def(ctx, offset + 0x1c, 0x118c0000); + break; + case 0x94: + gr_def(ctx, offset + 0x1c, 0x10880000); + break; + case 0xa0: + case 0xa5: + gr_def(ctx, offset + 0x1c, 0x310c0000); + break; + case 0xa3: + case 0xa8: + case 0xaa: + case 0xac: + case 0xaf: + gr_def(ctx, offset + 0x1c, 0x300c0000); + break; + } + gr_def(ctx, offset + 0x40, 0x00010401); + if (device->chipset == 0x50) + gr_def(ctx, offset + 0x48, 0x00000040); + else + gr_def(ctx, offset + 0x48, 0x00000078); + gr_def(ctx, offset + 0x50, 0x000000bf); + gr_def(ctx, offset + 0x58, 0x00001210); + if (device->chipset == 0x50) + gr_def(ctx, offset + 0x5c, 0x00000080); + else + gr_def(ctx, offset + 0x5c, 0x08000080); + if (device->chipset >= 0xa0) + gr_def(ctx, offset + 0x68, 0x0000003e); + } + + if (device->chipset < 0xa0) + cp_ctx(ctx, base + 0x300, 0x4); + else + cp_ctx(ctx, base + 0x300, 0x5); + if (device->chipset == 0x50) + gr_def(ctx, base + 0x304, 0x00007070); + else if (device->chipset < 0xa0) + gr_def(ctx, base + 0x304, 0x00027070); + else if (!IS_NVA3F(device->chipset)) + gr_def(ctx, base + 0x304, 0x01127070); + else + gr_def(ctx, base + 0x304, 0x05127070); + + if (device->chipset < 0xa0) + cp_ctx(ctx, base + 0x318, 1); + else + cp_ctx(ctx, base + 0x320, 1); + if (device->chipset == 0x50) + gr_def(ctx, base + 0x318, 0x0003ffff); + else if (device->chipset < 0xa0) + gr_def(ctx, base + 0x318, 0x03ffffff); + else + gr_def(ctx, base + 0x320, 0x07ffffff); + + if (device->chipset < 0xa0) + cp_ctx(ctx, base + 0x324, 5); + else + cp_ctx(ctx, base + 0x328, 4); + + if (device->chipset < 0xa0) { + cp_ctx(ctx, base + 0x340, 9); + offset = base + 0x340; + } else if (!IS_NVA3F(device->chipset)) { + cp_ctx(ctx, base + 0x33c, 0xb); + offset = base + 0x344; + } else { + cp_ctx(ctx, base + 0x33c, 0xd); + offset = base + 0x344; + } + gr_def(ctx, offset + 0x0, 0x00120407); + gr_def(ctx, offset + 0x4, 0x05091507); + if (device->chipset == 0x84) + gr_def(ctx, offset + 0x8, 0x05100202); + else + gr_def(ctx, offset + 0x8, 0x05010202); + gr_def(ctx, offset + 0xc, 0x00030201); + if (device->chipset == 0xa3) + cp_ctx(ctx, base + 0x36c, 1); + + cp_ctx(ctx, base + 0x400, 2); + gr_def(ctx, base + 0x404, 0x00000040); + cp_ctx(ctx, base + 0x40c, 2); + gr_def(ctx, base + 0x40c, 0x0d0c0b0a); + gr_def(ctx, base + 0x410, 0x00141210); + + if (device->chipset < 0xa0) + offset = base + 0x800; + else + offset = base + 0x500; + cp_ctx(ctx, offset, 6); + gr_def(ctx, offset + 0x0, 0x000001f0); + gr_def(ctx, offset + 0x4, 0x00000001); + gr_def(ctx, offset + 0x8, 0x00000003); + if (device->chipset == 0x50 || IS_NVAAF(device->chipset)) + gr_def(ctx, offset + 0xc, 0x00008000); + gr_def(ctx, offset + 0x14, 0x00039e00); + cp_ctx(ctx, offset + 0x1c, 2); + if (device->chipset == 0x50) + gr_def(ctx, offset + 0x1c, 0x00000040); + else + gr_def(ctx, offset + 0x1c, 0x00000100); + gr_def(ctx, offset + 0x20, 0x00003800); + + if (device->chipset >= 0xa0) { + cp_ctx(ctx, base + 0x54c, 2); + if (!IS_NVA3F(device->chipset)) + gr_def(ctx, base + 0x54c, 0x003fe006); + else + gr_def(ctx, base + 0x54c, 0x003fe007); + gr_def(ctx, base + 0x550, 0x003fe000); + } + + if (device->chipset < 0xa0) + offset = base + 0xa00; + else + offset = base + 0x680; + cp_ctx(ctx, offset, 1); + gr_def(ctx, offset, 0x00404040); + + if (device->chipset < 0xa0) + offset = base + 0xe00; + else + offset = base + 0x700; + cp_ctx(ctx, offset, 2); + if (device->chipset < 0xa0) + gr_def(ctx, offset, 0x0077f005); + else if (device->chipset == 0xa5) + gr_def(ctx, offset, 0x6cf7f007); + else if (device->chipset == 0xa8) + gr_def(ctx, offset, 0x6cfff007); + else if (device->chipset == 0xac) + gr_def(ctx, offset, 0x0cfff007); + else + gr_def(ctx, offset, 0x0cf7f007); + if (device->chipset == 0x50) + gr_def(ctx, offset + 0x4, 0x00007fff); + else if (device->chipset < 0xa0) + gr_def(ctx, offset + 0x4, 0x003f7fff); + else + gr_def(ctx, offset + 0x4, 0x02bf7fff); + cp_ctx(ctx, offset + 0x2c, 1); + if (device->chipset == 0x50) { + cp_ctx(ctx, offset + 0x50, 9); + gr_def(ctx, offset + 0x54, 0x000003ff); + gr_def(ctx, offset + 0x58, 0x00000003); + gr_def(ctx, offset + 0x5c, 0x00000003); + gr_def(ctx, offset + 0x60, 0x000001ff); + gr_def(ctx, offset + 0x64, 0x0000001f); + gr_def(ctx, offset + 0x68, 0x0000000f); + gr_def(ctx, offset + 0x6c, 0x0000000f); + } else if (device->chipset < 0xa0) { + cp_ctx(ctx, offset + 0x50, 1); + cp_ctx(ctx, offset + 0x70, 1); + } else { + cp_ctx(ctx, offset + 0x50, 1); + cp_ctx(ctx, offset + 0x60, 5); + } + } + } +} + +static void +dd_emit(struct nvkm_grctx *ctx, int num, u32 val) { + int i; + if (val && ctx->mode == NVKM_GRCTX_VALS) { + for (i = 0; i < num; i++) + nvkm_wo32(ctx->data, 4 * (ctx->ctxvals_pos + i), val); + } + ctx->ctxvals_pos += num; +} + +static void +nv50_gr_construct_mmio_ddata(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int base, num; + base = ctx->ctxvals_pos; + + /* tesla state */ + dd_emit(ctx, 1, 0); /* 00000001 UNK0F90 */ + dd_emit(ctx, 1, 0); /* 00000001 UNK135C */ + + /* SRC_TIC state */ + dd_emit(ctx, 1, 0); /* 00000007 SRC_TILE_MODE_Z */ + dd_emit(ctx, 1, 2); /* 00000007 SRC_TILE_MODE_Y */ + dd_emit(ctx, 1, 1); /* 00000001 SRC_LINEAR #1 */ + dd_emit(ctx, 1, 0); /* 000000ff SRC_ADDRESS_HIGH */ + dd_emit(ctx, 1, 0); /* 00000001 SRC_SRGB */ + if (device->chipset >= 0x94) + dd_emit(ctx, 1, 0); /* 00000003 eng2d UNK0258 */ + dd_emit(ctx, 1, 1); /* 00000fff SRC_DEPTH */ + dd_emit(ctx, 1, 0x100); /* 0000ffff SRC_HEIGHT */ + + /* turing state */ + dd_emit(ctx, 1, 0); /* 0000000f TEXTURES_LOG2 */ + dd_emit(ctx, 1, 0); /* 0000000f SAMPLERS_LOG2 */ + dd_emit(ctx, 1, 0); /* 000000ff CB_DEF_ADDRESS_HIGH */ + dd_emit(ctx, 1, 0); /* ffffffff CB_DEF_ADDRESS_LOW */ + dd_emit(ctx, 1, 0); /* ffffffff SHARED_SIZE */ + dd_emit(ctx, 1, 2); /* ffffffff REG_MODE */ + dd_emit(ctx, 1, 1); /* 0000ffff BLOCK_ALLOC_THREADS */ + dd_emit(ctx, 1, 1); /* 00000001 LANES32 */ + dd_emit(ctx, 1, 0); /* 000000ff UNK370 */ + dd_emit(ctx, 1, 0); /* 000000ff USER_PARAM_UNK */ + dd_emit(ctx, 1, 0); /* 000000ff USER_PARAM_COUNT */ + dd_emit(ctx, 1, 1); /* 000000ff UNK384 bits 8-15 */ + dd_emit(ctx, 1, 0x3fffff); /* 003fffff TIC_LIMIT */ + dd_emit(ctx, 1, 0x1fff); /* 000fffff TSC_LIMIT */ + dd_emit(ctx, 1, 0); /* 0000ffff CB_ADDR_INDEX */ + dd_emit(ctx, 1, 1); /* 000007ff BLOCKDIM_X */ + dd_emit(ctx, 1, 1); /* 000007ff BLOCKDIM_XMY */ + dd_emit(ctx, 1, 0); /* 00000001 BLOCKDIM_XMY_OVERFLOW */ + dd_emit(ctx, 1, 1); /* 0003ffff BLOCKDIM_XMYMZ */ + dd_emit(ctx, 1, 1); /* 000007ff BLOCKDIM_Y */ + dd_emit(ctx, 1, 1); /* 0000007f BLOCKDIM_Z */ + dd_emit(ctx, 1, 4); /* 000000ff CP_REG_ALLOC_TEMP */ + dd_emit(ctx, 1, 1); /* 00000001 BLOCKDIM_DIRTY */ + if (IS_NVA3F(device->chipset)) + dd_emit(ctx, 1, 0); /* 00000003 UNK03E8 */ + dd_emit(ctx, 1, 1); /* 0000007f BLOCK_ALLOC_HALFWARPS */ + dd_emit(ctx, 1, 1); /* 00000007 LOCAL_WARPS_NO_CLAMP */ + dd_emit(ctx, 1, 7); /* 00000007 LOCAL_WARPS_LOG_ALLOC */ + dd_emit(ctx, 1, 1); /* 00000007 STACK_WARPS_NO_CLAMP */ + dd_emit(ctx, 1, 7); /* 00000007 STACK_WARPS_LOG_ALLOC */ + dd_emit(ctx, 1, 1); /* 00001fff BLOCK_ALLOC_REGSLOTS_PACKED */ + dd_emit(ctx, 1, 1); /* 00001fff BLOCK_ALLOC_REGSLOTS_STRIDED */ + dd_emit(ctx, 1, 1); /* 000007ff BLOCK_ALLOC_THREADS */ + + /* compat 2d state */ + if (device->chipset == 0x50) { + dd_emit(ctx, 4, 0); /* 0000ffff clip X, Y, W, H */ + + dd_emit(ctx, 1, 1); /* ffffffff chroma COLOR_FORMAT */ + + dd_emit(ctx, 1, 1); /* ffffffff pattern COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff pattern SHAPE */ + dd_emit(ctx, 1, 1); /* ffffffff pattern PATTERN_SELECT */ + + dd_emit(ctx, 1, 0xa); /* ffffffff surf2d SRC_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff surf2d DMA_SRC */ + dd_emit(ctx, 1, 0); /* 000000ff surf2d SRC_ADDRESS_HIGH */ + dd_emit(ctx, 1, 0); /* ffffffff surf2d SRC_ADDRESS_LOW */ + dd_emit(ctx, 1, 0x40); /* 0000ffff surf2d SRC_PITCH */ + dd_emit(ctx, 1, 0); /* 0000000f surf2d SRC_TILE_MODE_Z */ + dd_emit(ctx, 1, 2); /* 0000000f surf2d SRC_TILE_MODE_Y */ + dd_emit(ctx, 1, 0x100); /* ffffffff surf2d SRC_HEIGHT */ + dd_emit(ctx, 1, 1); /* 00000001 surf2d SRC_LINEAR */ + dd_emit(ctx, 1, 0x100); /* ffffffff surf2d SRC_WIDTH */ + + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_B_X */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_B_Y */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_C_X */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_C_Y */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_D_X */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect CLIP_D_Y */ + dd_emit(ctx, 1, 1); /* ffffffff gdirect COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff gdirect OPERATION */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect POINT_X */ + dd_emit(ctx, 1, 0); /* 0000ffff gdirect POINT_Y */ + + dd_emit(ctx, 1, 0); /* 0000ffff blit SRC_Y */ + dd_emit(ctx, 1, 0); /* ffffffff blit OPERATION */ + + dd_emit(ctx, 1, 0); /* ffffffff ifc OPERATION */ + + dd_emit(ctx, 1, 0); /* ffffffff iifc INDEX_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff iifc LUT_OFFSET */ + dd_emit(ctx, 1, 4); /* ffffffff iifc COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff iifc OPERATION */ + } + + /* m2mf state */ + dd_emit(ctx, 1, 0); /* ffffffff m2mf LINE_COUNT */ + dd_emit(ctx, 1, 0); /* ffffffff m2mf LINE_LENGTH_IN */ + dd_emit(ctx, 2, 0); /* ffffffff m2mf OFFSET_IN, OFFSET_OUT */ + dd_emit(ctx, 1, 1); /* ffffffff m2mf TILING_DEPTH_OUT */ + dd_emit(ctx, 1, 0x100); /* ffffffff m2mf TILING_HEIGHT_OUT */ + dd_emit(ctx, 1, 0); /* ffffffff m2mf TILING_POSITION_OUT_Z */ + dd_emit(ctx, 1, 1); /* 00000001 m2mf LINEAR_OUT */ + dd_emit(ctx, 2, 0); /* 0000ffff m2mf TILING_POSITION_OUT_X, Y */ + dd_emit(ctx, 1, 0x100); /* ffffffff m2mf TILING_PITCH_OUT */ + dd_emit(ctx, 1, 1); /* ffffffff m2mf TILING_DEPTH_IN */ + dd_emit(ctx, 1, 0x100); /* ffffffff m2mf TILING_HEIGHT_IN */ + dd_emit(ctx, 1, 0); /* ffffffff m2mf TILING_POSITION_IN_Z */ + dd_emit(ctx, 1, 1); /* 00000001 m2mf LINEAR_IN */ + dd_emit(ctx, 2, 0); /* 0000ffff m2mf TILING_POSITION_IN_X, Y */ + dd_emit(ctx, 1, 0x100); /* ffffffff m2mf TILING_PITCH_IN */ + + /* more compat 2d state */ + if (device->chipset == 0x50) { + dd_emit(ctx, 1, 1); /* ffffffff line COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff line OPERATION */ + + dd_emit(ctx, 1, 1); /* ffffffff triangle COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff triangle OPERATION */ + + dd_emit(ctx, 1, 0); /* 0000000f sifm TILE_MODE_Z */ + dd_emit(ctx, 1, 2); /* 0000000f sifm TILE_MODE_Y */ + dd_emit(ctx, 1, 0); /* 000000ff sifm FORMAT_FILTER */ + dd_emit(ctx, 1, 1); /* 000000ff sifm FORMAT_ORIGIN */ + dd_emit(ctx, 1, 0); /* 0000ffff sifm SRC_PITCH */ + dd_emit(ctx, 1, 1); /* 00000001 sifm SRC_LINEAR */ + dd_emit(ctx, 1, 0); /* 000000ff sifm SRC_OFFSET_HIGH */ + dd_emit(ctx, 1, 0); /* ffffffff sifm SRC_OFFSET */ + dd_emit(ctx, 1, 0); /* 0000ffff sifm SRC_HEIGHT */ + dd_emit(ctx, 1, 0); /* 0000ffff sifm SRC_WIDTH */ + dd_emit(ctx, 1, 3); /* ffffffff sifm COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff sifm OPERATION */ + + dd_emit(ctx, 1, 0); /* ffffffff sifc OPERATION */ + } + + /* tesla state */ + dd_emit(ctx, 1, 0); /* 0000000f GP_TEXTURES_LOG2 */ + dd_emit(ctx, 1, 0); /* 0000000f GP_SAMPLERS_LOG2 */ + dd_emit(ctx, 1, 0); /* 000000ff */ + dd_emit(ctx, 1, 0); /* ffffffff */ + dd_emit(ctx, 1, 4); /* 000000ff UNK12B0_0 */ + dd_emit(ctx, 1, 0x70); /* 000000ff UNK12B0_1 */ + dd_emit(ctx, 1, 0x80); /* 000000ff UNK12B0_3 */ + dd_emit(ctx, 1, 0); /* 000000ff UNK12B0_2 */ + dd_emit(ctx, 1, 0); /* 0000000f FP_TEXTURES_LOG2 */ + dd_emit(ctx, 1, 0); /* 0000000f FP_SAMPLERS_LOG2 */ + if (IS_NVA3F(device->chipset)) { + dd_emit(ctx, 1, 0); /* ffffffff */ + dd_emit(ctx, 1, 0); /* 0000007f MULTISAMPLE_SAMPLES_LOG2 */ + } else { + dd_emit(ctx, 1, 0); /* 0000000f MULTISAMPLE_SAMPLES_LOG2 */ + } + dd_emit(ctx, 1, 0xc); /* 000000ff SEMANTIC_COLOR.BFC0_ID */ + if (device->chipset != 0x50) + dd_emit(ctx, 1, 0); /* 00000001 SEMANTIC_COLOR.CLMP_EN */ + dd_emit(ctx, 1, 8); /* 000000ff SEMANTIC_COLOR.COLR_NR */ + dd_emit(ctx, 1, 0x14); /* 000000ff SEMANTIC_COLOR.FFC0_ID */ + if (device->chipset == 0x50) { + dd_emit(ctx, 1, 0); /* 000000ff SEMANTIC_LAYER */ + dd_emit(ctx, 1, 0); /* 00000001 */ + } else { + dd_emit(ctx, 1, 0); /* 00000001 SEMANTIC_PTSZ.ENABLE */ + dd_emit(ctx, 1, 0x29); /* 000000ff SEMANTIC_PTSZ.PTSZ_ID */ + dd_emit(ctx, 1, 0x27); /* 000000ff SEMANTIC_PRIM */ + dd_emit(ctx, 1, 0x26); /* 000000ff SEMANTIC_LAYER */ + dd_emit(ctx, 1, 8); /* 0000000f SMENATIC_CLIP.CLIP_HIGH */ + dd_emit(ctx, 1, 4); /* 000000ff SEMANTIC_CLIP.CLIP_LO */ + dd_emit(ctx, 1, 0x27); /* 000000ff UNK0FD4 */ + dd_emit(ctx, 1, 0); /* 00000001 UNK1900 */ + } + dd_emit(ctx, 1, 0); /* 00000007 RT_CONTROL_MAP0 */ + dd_emit(ctx, 1, 1); /* 00000007 RT_CONTROL_MAP1 */ + dd_emit(ctx, 1, 2); /* 00000007 RT_CONTROL_MAP2 */ + dd_emit(ctx, 1, 3); /* 00000007 RT_CONTROL_MAP3 */ + dd_emit(ctx, 1, 4); /* 00000007 RT_CONTROL_MAP4 */ + dd_emit(ctx, 1, 5); /* 00000007 RT_CONTROL_MAP5 */ + dd_emit(ctx, 1, 6); /* 00000007 RT_CONTROL_MAP6 */ + dd_emit(ctx, 1, 7); /* 00000007 RT_CONTROL_MAP7 */ + dd_emit(ctx, 1, 1); /* 0000000f RT_CONTROL_COUNT */ + dd_emit(ctx, 8, 0); /* 00000001 RT_HORIZ_UNK */ + dd_emit(ctx, 8, 0); /* ffffffff RT_ADDRESS_LOW */ + dd_emit(ctx, 1, 0xcf); /* 000000ff RT_FORMAT */ + dd_emit(ctx, 7, 0); /* 000000ff RT_FORMAT */ + if (device->chipset != 0x50) + dd_emit(ctx, 3, 0); /* 1, 1, 1 */ + else + dd_emit(ctx, 2, 0); /* 1, 1 */ + dd_emit(ctx, 1, 0); /* ffffffff GP_ENABLE */ + dd_emit(ctx, 1, 0x80); /* 0000ffff GP_VERTEX_OUTPUT_COUNT*/ + dd_emit(ctx, 1, 4); /* 000000ff GP_REG_ALLOC_RESULT */ + dd_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + if (IS_NVA3F(device->chipset)) { + dd_emit(ctx, 1, 3); /* 00000003 */ + dd_emit(ctx, 1, 0); /* 00000001 UNK1418. Alone. */ + } + if (device->chipset != 0x50) + dd_emit(ctx, 1, 3); /* 00000003 UNK15AC */ + dd_emit(ctx, 1, 1); /* ffffffff RASTERIZE_ENABLE */ + dd_emit(ctx, 1, 0); /* 00000001 FP_CONTROL.EXPORTS_Z */ + if (device->chipset != 0x50) + dd_emit(ctx, 1, 0); /* 00000001 FP_CONTROL.MULTIPLE_RESULTS */ + dd_emit(ctx, 1, 0x12); /* 000000ff FP_INTERPOLANT_CTRL.COUNT */ + dd_emit(ctx, 1, 0x10); /* 000000ff FP_INTERPOLANT_CTRL.COUNT_NONFLAT */ + dd_emit(ctx, 1, 0xc); /* 000000ff FP_INTERPOLANT_CTRL.OFFSET */ + dd_emit(ctx, 1, 1); /* 00000001 FP_INTERPOLANT_CTRL.UMASK.W */ + dd_emit(ctx, 1, 0); /* 00000001 FP_INTERPOLANT_CTRL.UMASK.X */ + dd_emit(ctx, 1, 0); /* 00000001 FP_INTERPOLANT_CTRL.UMASK.Y */ + dd_emit(ctx, 1, 0); /* 00000001 FP_INTERPOLANT_CTRL.UMASK.Z */ + dd_emit(ctx, 1, 4); /* 000000ff FP_RESULT_COUNT */ + dd_emit(ctx, 1, 2); /* ffffffff REG_MODE */ + dd_emit(ctx, 1, 4); /* 000000ff FP_REG_ALLOC_TEMP */ + if (device->chipset >= 0xa0) + dd_emit(ctx, 1, 0); /* ffffffff */ + dd_emit(ctx, 1, 0); /* 00000001 GP_BUILTIN_RESULT_EN.LAYER_IDX */ + dd_emit(ctx, 1, 0); /* ffffffff STRMOUT_ENABLE */ + dd_emit(ctx, 1, 0x3fffff); /* 003fffff TIC_LIMIT */ + dd_emit(ctx, 1, 0x1fff); /* 000fffff TSC_LIMIT */ + dd_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE*/ + if (device->chipset != 0x50) + dd_emit(ctx, 8, 0); /* 00000001 */ + if (device->chipset >= 0xa0) { + dd_emit(ctx, 1, 1); /* 00000007 VTX_ATTR_DEFINE.COMP */ + dd_emit(ctx, 1, 1); /* 00000007 VTX_ATTR_DEFINE.SIZE */ + dd_emit(ctx, 1, 2); /* 00000007 VTX_ATTR_DEFINE.TYPE */ + dd_emit(ctx, 1, 0); /* 000000ff VTX_ATTR_DEFINE.ATTR */ + } + dd_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + dd_emit(ctx, 1, 0x14); /* 0000001f ZETA_FORMAT */ + dd_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + dd_emit(ctx, 1, 0); /* 0000000f VP_TEXTURES_LOG2 */ + dd_emit(ctx, 1, 0); /* 0000000f VP_SAMPLERS_LOG2 */ + if (IS_NVA3F(device->chipset)) + dd_emit(ctx, 1, 0); /* 00000001 */ + dd_emit(ctx, 1, 2); /* 00000003 POLYGON_MODE_BACK */ + if (device->chipset >= 0xa0) + dd_emit(ctx, 1, 0); /* 00000003 VTX_ATTR_DEFINE.SIZE - 1 */ + dd_emit(ctx, 1, 0); /* 0000ffff CB_ADDR_INDEX */ + if (device->chipset >= 0xa0) + dd_emit(ctx, 1, 0); /* 00000003 */ + dd_emit(ctx, 1, 0); /* 00000001 CULL_FACE_ENABLE */ + dd_emit(ctx, 1, 1); /* 00000003 CULL_FACE */ + dd_emit(ctx, 1, 0); /* 00000001 FRONT_FACE */ + dd_emit(ctx, 1, 2); /* 00000003 POLYGON_MODE_FRONT */ + dd_emit(ctx, 1, 0x1000); /* 00007fff UNK141C */ + if (device->chipset != 0x50) { + dd_emit(ctx, 1, 0xe00); /* 7fff */ + dd_emit(ctx, 1, 0x1000); /* 7fff */ + dd_emit(ctx, 1, 0x1e00); /* 7fff */ + } + dd_emit(ctx, 1, 0); /* 00000001 BEGIN_END_ACTIVE */ + dd_emit(ctx, 1, 1); /* 00000001 POLYGON_MODE_??? */ + dd_emit(ctx, 1, 1); /* 000000ff GP_REG_ALLOC_TEMP / 4 rounded up */ + dd_emit(ctx, 1, 1); /* 000000ff FP_REG_ALLOC_TEMP... without /4? */ + dd_emit(ctx, 1, 1); /* 000000ff VP_REG_ALLOC_TEMP / 4 rounded up */ + dd_emit(ctx, 1, 1); /* 00000001 */ + dd_emit(ctx, 1, 0); /* 00000001 */ + dd_emit(ctx, 1, 0); /* 00000001 VTX_ATTR_MASK_UNK0 nonempty */ + dd_emit(ctx, 1, 0); /* 00000001 VTX_ATTR_MASK_UNK1 nonempty */ + dd_emit(ctx, 1, 0x200); /* 0003ffff GP_VERTEX_OUTPUT_COUNT*GP_REG_ALLOC_RESULT */ + if (IS_NVA3F(device->chipset)) + dd_emit(ctx, 1, 0x200); + dd_emit(ctx, 1, 0); /* 00000001 */ + if (device->chipset < 0xa0) { + dd_emit(ctx, 1, 1); /* 00000001 */ + dd_emit(ctx, 1, 0x70); /* 000000ff */ + dd_emit(ctx, 1, 0x80); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 00000001 */ + dd_emit(ctx, 1, 1); /* 00000001 */ + dd_emit(ctx, 1, 0x70); /* 000000ff */ + dd_emit(ctx, 1, 0x80); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 000000ff */ + } else { + dd_emit(ctx, 1, 1); /* 00000001 */ + dd_emit(ctx, 1, 0xf0); /* 000000ff */ + dd_emit(ctx, 1, 0xff); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 00000001 */ + dd_emit(ctx, 1, 1); /* 00000001 */ + dd_emit(ctx, 1, 0xf0); /* 000000ff */ + dd_emit(ctx, 1, 0xff); /* 000000ff */ + dd_emit(ctx, 1, 0); /* 000000ff */ + dd_emit(ctx, 1, 9); /* 0000003f UNK114C.COMP,SIZE */ + } + + /* eng2d state */ + dd_emit(ctx, 1, 0); /* 00000001 eng2d COLOR_KEY_ENABLE */ + dd_emit(ctx, 1, 0); /* 00000007 eng2d COLOR_KEY_FORMAT */ + dd_emit(ctx, 1, 1); /* ffffffff eng2d DST_DEPTH */ + dd_emit(ctx, 1, 0xcf); /* 000000ff eng2d DST_FORMAT */ + dd_emit(ctx, 1, 0); /* ffffffff eng2d DST_LAYER */ + dd_emit(ctx, 1, 1); /* 00000001 eng2d DST_LINEAR */ + dd_emit(ctx, 1, 0); /* 00000007 eng2d PATTERN_COLOR_FORMAT */ + dd_emit(ctx, 1, 0); /* 00000007 eng2d OPERATION */ + dd_emit(ctx, 1, 0); /* 00000003 eng2d PATTERN_SELECT */ + dd_emit(ctx, 1, 0xcf); /* 000000ff eng2d SIFC_FORMAT */ + dd_emit(ctx, 1, 0); /* 00000001 eng2d SIFC_BITMAP_ENABLE */ + dd_emit(ctx, 1, 2); /* 00000003 eng2d SIFC_BITMAP_UNK808 */ + dd_emit(ctx, 1, 0); /* ffffffff eng2d BLIT_DU_DX_FRACT */ + dd_emit(ctx, 1, 1); /* ffffffff eng2d BLIT_DU_DX_INT */ + dd_emit(ctx, 1, 0); /* ffffffff eng2d BLIT_DV_DY_FRACT */ + dd_emit(ctx, 1, 1); /* ffffffff eng2d BLIT_DV_DY_INT */ + dd_emit(ctx, 1, 0); /* 00000001 eng2d BLIT_CONTROL_FILTER */ + dd_emit(ctx, 1, 0xcf); /* 000000ff eng2d DRAW_COLOR_FORMAT */ + dd_emit(ctx, 1, 0xcf); /* 000000ff eng2d SRC_FORMAT */ + dd_emit(ctx, 1, 1); /* 00000001 eng2d SRC_LINEAR #2 */ + + num = ctx->ctxvals_pos - base; + ctx->ctxvals_pos = base; + if (IS_NVA3F(device->chipset)) + cp_ctx(ctx, 0x404800, num); + else + cp_ctx(ctx, 0x405400, num); +} + +/* + * xfer areas. These are a pain. + * + * There are 2 xfer areas: the first one is big and contains all sorts of + * stuff, the second is small and contains some per-TP context. + * + * Each area is split into 8 "strands". The areas, when saved to grctx, + * are made of 8-word blocks. Each block contains a single word from + * each strand. The strands are independent of each other, their + * addresses are unrelated to each other, and data in them is closely + * packed together. The strand layout varies a bit between cards: here + * and there, a single word is thrown out in the middle and the whole + * strand is offset by a bit from corresponding one on another chipset. + * For this reason, addresses of stuff in strands are almost useless. + * Knowing sequence of stuff and size of gaps between them is much more + * useful, and that's how we build the strands in our generator. + * + * NVA0 takes this mess to a whole new level by cutting the old strands + * into a few dozen pieces [known as genes], rearranging them randomly, + * and putting them back together to make new strands. Hopefully these + * genes correspond more or less directly to the same PGRAPH subunits + * as in 400040 register. + * + * The most common value in default context is 0, and when the genes + * are separated by 0's, gene bounduaries are quite speculative... + * some of them can be clearly deduced, others can be guessed, and yet + * others won't be resolved without figuring out the real meaning of + * given ctxval. For the same reason, ending point of each strand + * is unknown. Except for strand 0, which is the longest strand and + * its end corresponds to end of the whole xfer. + * + * An unsolved mystery is the seek instruction: it takes an argument + * in bits 8-18, and that argument is clearly the place in strands to + * seek to... but the offsets don't seem to correspond to offsets as + * seen in grctx. Perhaps there's another, real, not randomly-changing + * addressing in strands, and the xfer insn just happens to skip over + * the unused bits? NV10-NV30 PIPE comes to mind... + * + * As far as I know, there's no way to access the xfer areas directly + * without the help of ctxprog. + */ + +static void +xf_emit(struct nvkm_grctx *ctx, int num, u32 val) { + int i; + if (val && ctx->mode == NVKM_GRCTX_VALS) { + for (i = 0; i < num; i++) + nvkm_wo32(ctx->data, 4 * (ctx->ctxvals_pos + (i << 3)), val); + } + ctx->ctxvals_pos += num << 3; +} + +/* Gene declarations... */ + +static void nv50_gr_construct_gene_dispatch(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_m2mf(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_ccache(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_unk10xx(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_unk14xx(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_zcull(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_clipid(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_unk24xx(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_vfetch(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_eng2d(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_csched(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_unk1cxx(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_strmout(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_unk34xx(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_ropm1(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_ropm2(struct nvkm_grctx *ctx); +static void nv50_gr_construct_gene_ropc(struct nvkm_grctx *ctx); +static void nv50_gr_construct_xfer_tp(struct nvkm_grctx *ctx); + +static void +nv50_gr_construct_xfer1(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + int offset; + int size = 0; + u32 units = nvkm_rd32(device, 0x1540); + + offset = (ctx->ctxvals_pos+0x3f)&~0x3f; + ctx->ctxvals_base = offset; + + if (device->chipset < 0xa0) { + /* Strand 0 */ + ctx->ctxvals_pos = offset; + nv50_gr_construct_gene_dispatch(ctx); + nv50_gr_construct_gene_m2mf(ctx); + nv50_gr_construct_gene_unk24xx(ctx); + nv50_gr_construct_gene_clipid(ctx); + nv50_gr_construct_gene_zcull(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 1 */ + ctx->ctxvals_pos = offset + 0x1; + nv50_gr_construct_gene_vfetch(ctx); + nv50_gr_construct_gene_eng2d(ctx); + nv50_gr_construct_gene_csched(ctx); + nv50_gr_construct_gene_ropm1(ctx); + nv50_gr_construct_gene_ropm2(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 2 */ + ctx->ctxvals_pos = offset + 0x2; + nv50_gr_construct_gene_ccache(ctx); + nv50_gr_construct_gene_unk1cxx(ctx); + nv50_gr_construct_gene_strmout(ctx); + nv50_gr_construct_gene_unk14xx(ctx); + nv50_gr_construct_gene_unk10xx(ctx); + nv50_gr_construct_gene_unk34xx(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 3: per-ROP group state */ + ctx->ctxvals_pos = offset + 3; + for (i = 0; i < 6; i++) + if (units & (1 << (i + 16))) + nv50_gr_construct_gene_ropc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strands 4-7: per-TP state */ + for (i = 0; i < 4; i++) { + ctx->ctxvals_pos = offset + 4 + i; + if (units & (1 << (2 * i))) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << (2 * i + 1))) + nv50_gr_construct_xfer_tp(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + } else { + /* Strand 0 */ + ctx->ctxvals_pos = offset; + nv50_gr_construct_gene_dispatch(ctx); + nv50_gr_construct_gene_m2mf(ctx); + nv50_gr_construct_gene_unk34xx(ctx); + nv50_gr_construct_gene_csched(ctx); + nv50_gr_construct_gene_unk1cxx(ctx); + nv50_gr_construct_gene_strmout(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 1 */ + ctx->ctxvals_pos = offset + 1; + nv50_gr_construct_gene_unk10xx(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 2 */ + ctx->ctxvals_pos = offset + 2; + if (device->chipset == 0xa0) + nv50_gr_construct_gene_unk14xx(ctx); + nv50_gr_construct_gene_unk24xx(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 3 */ + ctx->ctxvals_pos = offset + 3; + nv50_gr_construct_gene_vfetch(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 4 */ + ctx->ctxvals_pos = offset + 4; + nv50_gr_construct_gene_ccache(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 5 */ + ctx->ctxvals_pos = offset + 5; + nv50_gr_construct_gene_ropm2(ctx); + nv50_gr_construct_gene_ropm1(ctx); + /* per-ROP context */ + for (i = 0; i < 8; i++) + if (units & (1<<(i+16))) + nv50_gr_construct_gene_ropc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 6 */ + ctx->ctxvals_pos = offset + 6; + nv50_gr_construct_gene_zcull(ctx); + nv50_gr_construct_gene_clipid(ctx); + nv50_gr_construct_gene_eng2d(ctx); + if (units & (1 << 0)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 1)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 2)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 3)) + nv50_gr_construct_xfer_tp(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 7 */ + ctx->ctxvals_pos = offset + 7; + if (device->chipset == 0xa0) { + if (units & (1 << 4)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 5)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 6)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 7)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 8)) + nv50_gr_construct_xfer_tp(ctx); + if (units & (1 << 9)) + nv50_gr_construct_xfer_tp(ctx); + } else { + nv50_gr_construct_gene_unk14xx(ctx); + } + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + + ctx->ctxvals_pos = offset + size * 8; + ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f; + cp_lsr (ctx, offset); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, size); + cp_out (ctx, CP_SEEK_1); + cp_out (ctx, CP_XFER_1); + cp_wait(ctx, XFER, BUSY); +} + +/* + * non-trivial demagiced parts of ctx init go here + */ + +static void +nv50_gr_construct_gene_dispatch(struct nvkm_grctx *ctx) +{ + /* start of strand 0 */ + struct nvkm_device *device = ctx->device; + /* SEEK */ + if (device->chipset == 0x50) + xf_emit(ctx, 5, 0); + else if (!IS_NVA3F(device->chipset)) + xf_emit(ctx, 6, 0); + else + xf_emit(ctx, 4, 0); + /* SEEK */ + /* the PGRAPH's internal FIFO */ + if (device->chipset == 0x50) + xf_emit(ctx, 8*3, 0); + else + xf_emit(ctx, 0x100*3, 0); + /* and another bonus slot?!? */ + xf_emit(ctx, 3, 0); + /* and YET ANOTHER bonus slot? */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 3, 0); + /* SEEK */ + /* CTX_SWITCH: caches of gr objects bound to subchannels. 8 values, last used index */ + xf_emit(ctx, 9, 0); + /* SEEK */ + xf_emit(ctx, 9, 0); + /* SEEK */ + xf_emit(ctx, 9, 0); + /* SEEK */ + xf_emit(ctx, 9, 0); + /* SEEK */ + if (device->chipset < 0x90) + xf_emit(ctx, 4, 0); + /* SEEK */ + xf_emit(ctx, 2, 0); + /* SEEK */ + xf_emit(ctx, 6*2, 0); + xf_emit(ctx, 2, 0); + /* SEEK */ + xf_emit(ctx, 2, 0); + /* SEEK */ + xf_emit(ctx, 6*2, 0); + xf_emit(ctx, 2, 0); + /* SEEK */ + if (device->chipset == 0x50) + xf_emit(ctx, 0x1c, 0); + else if (device->chipset < 0xa0) + xf_emit(ctx, 0x1e, 0); + else + xf_emit(ctx, 0x22, 0); + /* SEEK */ + xf_emit(ctx, 0x15, 0); +} + +static void +nv50_gr_construct_gene_m2mf(struct nvkm_grctx *ctx) +{ + /* Strand 0, right after dispatch */ + struct nvkm_device *device = ctx->device; + int smallm2mf = 0; + if (device->chipset < 0x92 || device->chipset == 0x98) + smallm2mf = 1; + /* SEEK */ + xf_emit (ctx, 1, 0); /* DMA_NOTIFY instance >> 4 */ + xf_emit (ctx, 1, 0); /* DMA_BUFFER_IN instance >> 4 */ + xf_emit (ctx, 1, 0); /* DMA_BUFFER_OUT instance >> 4 */ + xf_emit (ctx, 1, 0); /* OFFSET_IN */ + xf_emit (ctx, 1, 0); /* OFFSET_OUT */ + xf_emit (ctx, 1, 0); /* PITCH_IN */ + xf_emit (ctx, 1, 0); /* PITCH_OUT */ + xf_emit (ctx, 1, 0); /* LINE_LENGTH */ + xf_emit (ctx, 1, 0); /* LINE_COUNT */ + xf_emit (ctx, 1, 0x21); /* FORMAT: bits 0-4 INPUT_INC, bits 5-9 OUTPUT_INC */ + xf_emit (ctx, 1, 1); /* LINEAR_IN */ + xf_emit (ctx, 1, 0x2); /* TILING_MODE_IN: bits 0-2 y tiling, bits 3-5 z tiling */ + xf_emit (ctx, 1, 0x100); /* TILING_PITCH_IN */ + xf_emit (ctx, 1, 0x100); /* TILING_HEIGHT_IN */ + xf_emit (ctx, 1, 1); /* TILING_DEPTH_IN */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_IN_Z */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_IN */ + xf_emit (ctx, 1, 1); /* LINEAR_OUT */ + xf_emit (ctx, 1, 0x2); /* TILING_MODE_OUT: bits 0-2 y tiling, bits 3-5 z tiling */ + xf_emit (ctx, 1, 0x100); /* TILING_PITCH_OUT */ + xf_emit (ctx, 1, 0x100); /* TILING_HEIGHT_OUT */ + xf_emit (ctx, 1, 1); /* TILING_DEPTH_OUT */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_OUT_Z */ + xf_emit (ctx, 1, 0); /* TILING_POSITION_OUT */ + xf_emit (ctx, 1, 0); /* OFFSET_IN_HIGH */ + xf_emit (ctx, 1, 0); /* OFFSET_OUT_HIGH */ + /* SEEK */ + if (smallm2mf) + xf_emit(ctx, 0x40, 0); /* 20 * ffffffff, 3ffff */ + else + xf_emit(ctx, 0x100, 0); /* 80 * ffffffff, 3ffff */ + xf_emit(ctx, 4, 0); /* 1f/7f, 0, 1f/7f, 0 [1f for smallm2mf, 7f otherwise] */ + /* SEEK */ + if (smallm2mf) + xf_emit(ctx, 0x400, 0); /* ffffffff */ + else + xf_emit(ctx, 0x800, 0); /* ffffffff */ + xf_emit(ctx, 4, 0); /* ff/1ff, 0, 0, 0 [ff for smallm2mf, 1ff otherwise] */ + /* SEEK */ + xf_emit(ctx, 0x40, 0); /* 20 * bits ffffffff, 3ffff */ + xf_emit(ctx, 0x6, 0); /* 1f, 0, 1f, 0, 1f, 0 */ +} + +static void +nv50_gr_construct_gene_ccache(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 2, 0); /* RO */ + xf_emit(ctx, 0x800, 0); /* ffffffff */ + switch (device->chipset) { + case 0x50: + case 0x92: + case 0xa0: + xf_emit(ctx, 0x2b, 0); + break; + case 0x84: + xf_emit(ctx, 0x29, 0); + break; + case 0x94: + case 0x96: + case 0xa3: + xf_emit(ctx, 0x27, 0); + break; + case 0x86: + case 0x98: + case 0xa5: + case 0xa8: + case 0xaa: + case 0xac: + case 0xaf: + xf_emit(ctx, 0x25, 0); + break; + } + /* CB bindings, 0x80 of them. first word is address >> 8, second is + * size >> 4 | valid << 24 */ + xf_emit(ctx, 0x100, 0); /* ffffffff CB_DEF */ + xf_emit(ctx, 1, 0); /* 0000007f CB_ADDR_BUFFER */ + xf_emit(ctx, 1, 0); /* 0 */ + xf_emit(ctx, 0x30, 0); /* ff SET_PROGRAM_CB */ + xf_emit(ctx, 1, 0); /* 3f last SET_PROGRAM_CB */ + xf_emit(ctx, 4, 0); /* RO */ + xf_emit(ctx, 0x100, 0); /* ffffffff */ + xf_emit(ctx, 8, 0); /* 1f, 0, 0, ... */ + xf_emit(ctx, 8, 0); /* ffffffff */ + xf_emit(ctx, 4, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 3 */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_CODE_CB */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_TIC */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_TSC */ + xf_emit(ctx, 1, 0); /* 00000001 LINKED_TSC */ + xf_emit(ctx, 1, 0); /* 000000ff TIC_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff TIC_ADDRESS_LOW */ + xf_emit(ctx, 1, 0x3fffff); /* 003fffff TIC_LIMIT */ + xf_emit(ctx, 1, 0); /* 000000ff TSC_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff TSC_ADDRESS_LOW */ + xf_emit(ctx, 1, 0x1fff); /* 000fffff TSC_LIMIT */ + xf_emit(ctx, 1, 0); /* 000000ff VP_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff VP_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 00ffffff VP_START_ID */ + xf_emit(ctx, 1, 0); /* 000000ff CB_DEF_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff CB_DEF_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0); /* 000000ff GP_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff GP_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 00ffffff GP_START_ID */ + xf_emit(ctx, 1, 0); /* 000000ff FP_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff FP_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 00ffffff FP_START_ID */ +} + +static void +nv50_gr_construct_gene_unk10xx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + /* end of area 2 on pre-NVA0, area 1 on NVAx */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x80); /* 0000ffff GP_VERTEX_OUTPUT_COUNT */ + xf_emit(ctx, 1, 4); /* 000000ff GP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); + else + xf_emit(ctx, 1, 0x7ff); /* 000007ff */ + xf_emit(ctx, 1, 0); /* 111/113 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + for (i = 0; i < 8; i++) { + switch (device->chipset) { + case 0x50: + case 0x86: + case 0x98: + case 0xaa: + case 0xac: + xf_emit(ctx, 0xa0, 0); /* ffffffff */ + break; + case 0x84: + case 0x92: + case 0x94: + case 0x96: + xf_emit(ctx, 0x120, 0); + break; + case 0xa5: + case 0xa8: + xf_emit(ctx, 0x100, 0); /* ffffffff */ + break; + case 0xa0: + case 0xa3: + case 0xaf: + xf_emit(ctx, 0x400, 0); /* ffffffff */ + break; + } + xf_emit(ctx, 4, 0); /* 3f, 0, 0, 0 */ + xf_emit(ctx, 4, 0); /* ffffffff */ + } + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x80); /* 0000ffff GP_VERTEX_OUTPUT_COUNT */ + xf_emit(ctx, 1, 4); /* 000000ff GP_REG_ALLOC_TEMP */ + xf_emit(ctx, 1, 1); /* 00000001 RASTERIZE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0x27); /* 000000ff UNK0FD4 */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0x26); /* 000000ff SEMANTIC_LAYER */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ +} + +static void +nv50_gr_construct_gene_unk34xx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* end of area 2 on pre-NVA0, area 1 on NVAx */ + xf_emit(ctx, 1, 0); /* 00000001 VIEWPORT_CLIP_RECTS_EN */ + xf_emit(ctx, 1, 0); /* 00000003 VIEWPORT_CLIP_MODE */ + xf_emit(ctx, 0x10, 0x04000000); /* 07ffffff VIEWPORT_CLIP_HORIZ*8, VIEWPORT_CLIP_VERT*8 */ + xf_emit(ctx, 1, 0); /* 00000001 POLYGON_STIPPLE_ENABLE */ + xf_emit(ctx, 0x20, 0); /* ffffffff POLYGON_STIPPLE */ + xf_emit(ctx, 2, 0); /* 00007fff WINDOW_OFFSET_XY */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0x04e3bfdf); /* ffffffff UNK0D64 */ + xf_emit(ctx, 1, 0x04e3bfdf); /* ffffffff UNK0DF4 */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0x1fe21); /* 0001ffff tesla UNK0FAC */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0x0fac6881); + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 1); + xf_emit(ctx, 3, 0); + } +} + +static void +nv50_gr_construct_gene_unk14xx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* middle of area 2 on pre-NVA0, beginning of area 2 on NVA0, area 7 on >NVA0 */ + if (device->chipset != 0x50) { + xf_emit(ctx, 5, 0); /* ffffffff */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + xf_emit(ctx, 1, 0x804); /* 00000fff SEMANTIC_CLIP */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 2, 4); /* 7f, ff */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + } + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x10); /* 7f/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 000000ff VP_CLIP_DISTANCE_ENABLE */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0); /* 3ff */ + xf_emit(ctx, 1, 0); /* 000000ff tesla UNK1940 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0D7C */ + xf_emit(ctx, 1, 0x804); /* 00000fff SEMANTIC_CLIP */ + xf_emit(ctx, 1, 1); /* 00000001 VIEWPORT_TRANSFORM_EN */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0x7f); /* 000000ff tesla UNK0FFC */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 1); /* 00000001 SHADE_MODEL */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x10); /* 7f/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0D7C */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0F8C */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 1); /* 00000001 VIEWPORT_TRANSFORM_EN */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 4, 0); /* ffffffff NOPERSPECTIVE_BITMAP */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0); /* 0000000f */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); /* 000003ff tesla UNK0D68 */ + else + xf_emit(ctx, 1, 0x7ff); /* 000007ff tesla UNK0D68 */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE */ + xf_emit(ctx, 0x30, 0); /* ffffffff VIEWPORT_SCALE: X0, Y0, Z0, X1, Y1, ... */ + xf_emit(ctx, 3, 0); /* f, 0, 0 */ + xf_emit(ctx, 3, 0); /* ffffffff last VIEWPORT_SCALE? */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 1); /* 00000001 VIEWPORT_TRANSFORM_EN */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1924 */ + xf_emit(ctx, 1, 0x10); /* 000000ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 0x30, 0); /* ffffffff VIEWPORT_TRANSLATE */ + xf_emit(ctx, 3, 0); /* f, 0, 0 */ + xf_emit(ctx, 3, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 2, 0x88); /* 000001ff tesla UNK19D8 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1924 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 4); /* 0000000f CULL_MODE */ + xf_emit(ctx, 2, 0); /* 07ffffff SCREEN_SCISSOR */ + xf_emit(ctx, 2, 0); /* 00007fff WINDOW_OFFSET_XY */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + xf_emit(ctx, 0x10, 0); /* 00000001 SCISSOR_ENABLE */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0x26); /* 000000ff SEMANTIC_LAYER */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 0x3f800000); /* ffffffff LINE_WIDTH */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + xf_emit(ctx, 1, 0x10); /* 000000ff VIEW_VOLUME_CLIP_CTRL */ + if (device->chipset != 0x50) { + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + } + xf_emit(ctx, 0x20, 0); /* 10xbits ffffffff, 3fffff. SCISSOR_* */ + xf_emit(ctx, 1, 0); /* f */ + xf_emit(ctx, 1, 0); /* 0? */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 003fffff */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0x52); /* 000001ff SEMANTIC_PTSZ */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0x26); /* 000000ff SEMANTIC_LAYER */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0x00ffff00); /* 00ffffff LINE_STIPPLE_PATTERN */ + xf_emit(ctx, 1, 0); /* 0000000f */ +} + +static void +nv50_gr_construct_gene_zcull(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* end of strand 0 on pre-NVA0, beginning of strand 6 on NVAx */ + /* SEEK */ + xf_emit(ctx, 1, 0x3f); /* 0000003f UNK1590 */ + xf_emit(ctx, 1, 0); /* 00000001 ALPHA_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_BACK_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_REF */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 2); /* 00000003 tesla UNK143C */ + xf_emit(ctx, 2, 0x04000000); /* 07ffffff tesla UNK0D6C */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0); /* 00000001 CLIPID_ENABLE */ + xf_emit(ctx, 2, 0); /* ffffffff DEPTH_BOUNDS */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 00000007 DEPTH_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 4); /* 0000000f CULL_MODE */ + xf_emit(ctx, 1, 0); /* 0000ffff */ + xf_emit(ctx, 1, 0); /* 00000001 UNK0FB0 */ + xf_emit(ctx, 1, 0); /* 00000001 POLYGON_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0); /* 000000ff CLEAR_STENCIL */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_FRONT_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_REF */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff CLEAR_DEPTH */ + xf_emit(ctx, 1, 0); /* 00000007 */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1108 */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0x1001); /* 00001fff ZETA_ARRAY_MODE */ + /* SEEK */ + xf_emit(ctx, 4, 0xffff); /* 0000ffff MSAA_MASK */ + xf_emit(ctx, 0x10, 0); /* 00000001 SCISSOR_ENABLE */ + xf_emit(ctx, 0x10, 0); /* ffffffff DEPTH_RANGE_NEAR */ + xf_emit(ctx, 0x10, 0x3f800000); /* ffffffff DEPTH_RANGE_FAR */ + xf_emit(ctx, 1, 0x10); /* 7f/ff/3ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 00000001 VIEWPORT_CLIP_RECTS_EN */ + xf_emit(ctx, 1, 3); /* 00000003 FP_CTRL_UNK196C */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1968 */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0); /* 0fffffff tesla UNK1104 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK151C */ +} + +static void +nv50_gr_construct_gene_clipid(struct nvkm_grctx *ctx) +{ + /* middle of strand 0 on pre-NVA0 [after 24xx], middle of area 6 on NVAx */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000007 UNK0FB4 */ + /* SEEK */ + xf_emit(ctx, 4, 0); /* 07ffffff CLIPID_REGION_HORIZ */ + xf_emit(ctx, 4, 0); /* 07ffffff CLIPID_REGION_VERT */ + xf_emit(ctx, 2, 0); /* 07ffffff SCREEN_SCISSOR */ + xf_emit(ctx, 2, 0x04000000); /* 07ffffff UNK1508 */ + xf_emit(ctx, 1, 0); /* 00000001 CLIPID_ENABLE */ + xf_emit(ctx, 1, 0x80); /* 00003fff CLIPID_WIDTH */ + xf_emit(ctx, 1, 0); /* 000000ff CLIPID_ID */ + xf_emit(ctx, 1, 0); /* 000000ff CLIPID_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff CLIPID_ADDRESS_LOW */ + xf_emit(ctx, 1, 0x80); /* 00003fff CLIPID_HEIGHT */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_CLIPID */ +} + +static void +nv50_gr_construct_gene_unk24xx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + /* middle of strand 0 on pre-NVA0 [after m2mf], end of strand 2 on NVAx */ + /* SEEK */ + xf_emit(ctx, 0x33, 0); + /* SEEK */ + xf_emit(ctx, 2, 0); + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 4, 0); /* RO */ + xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */ + xf_emit(ctx, 1, 0); /* 1ff */ + xf_emit(ctx, 8, 0); /* 0? */ + xf_emit(ctx, 9, 0); /* ffffffff, 7ff */ + + xf_emit(ctx, 4, 0); /* RO */ + xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */ + xf_emit(ctx, 1, 0); /* 1ff */ + xf_emit(ctx, 8, 0); /* 0? */ + xf_emit(ctx, 9, 0); /* ffffffff, 7ff */ + } else { + xf_emit(ctx, 0xc, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */ + xf_emit(ctx, 1, 0); /* 1ff */ + xf_emit(ctx, 8, 0); /* 0? */ + + /* SEEK */ + xf_emit(ctx, 0xc, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 0xe10, 0); /* 190 * 9: 8*ffffffff, 7ff */ + xf_emit(ctx, 1, 0); /* 1ff */ + xf_emit(ctx, 8, 0); /* 0? */ + } + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK1100 */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 0); /* 0000000f VP_GP_BUILTIN_ATTR_EN */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 1); /* 00000001 */ + /* SEEK */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 2, 4); /* 000000ff */ + xf_emit(ctx, 1, 0x80c14); /* 01ffffff SEMANTIC_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 POINT_SPRITE_ENABLE */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 0x27); /* 000000ff SEMANTIC_PRIM_ID */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 1); /* 00000001 */ + for (i = 0; i < 10; i++) { + /* SEEK */ + xf_emit(ctx, 0x40, 0); /* ffffffff */ + xf_emit(ctx, 0x10, 0); /* 3, 0, 0.... */ + xf_emit(ctx, 0x10, 0); /* ffffffff */ + } + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000001 POINT_SPRITE_CTRL */ + xf_emit(ctx, 1, 1); /* 00000001 */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 4, 0); /* ffffffff NOPERSPECTIVE_BITMAP */ + xf_emit(ctx, 0x10, 0); /* 00ffffff POINT_COORD_REPLACE_MAP */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0); /* 000003ff */ +} + +static void +nv50_gr_construct_gene_vfetch(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int acnt = 0x10, rep, i; + /* beginning of strand 1 on pre-NVA0, strand 3 on NVAx */ + if (IS_NVA3F(device->chipset)) + acnt = 0x20; + /* SEEK */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK13A4 */ + xf_emit(ctx, 1, 1); /* 00000fff tesla UNK1318 */ + } + xf_emit(ctx, 1, 0); /* ffffffff VERTEX_BUFFER_FIRST */ + xf_emit(ctx, 1, 0); /* 00000001 PRIMITIVE_RESTART_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 UNK0DE8 */ + xf_emit(ctx, 1, 0); /* ffffffff PRIMITIVE_RESTART_INDEX */ + xf_emit(ctx, 1, 0xf); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, (acnt/8)-1, 0); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, acnt/8, 0); /* ffffffff VTX_ATR_MASK_UNK0DD0 */ + xf_emit(ctx, 1, 0); /* 0000000f VP_GP_BUILTIN_ATTR_EN */ + xf_emit(ctx, 1, 0x20); /* 0000ffff tesla UNK129C */ + xf_emit(ctx, 1, 0); /* 000000ff turing UNK370??? */ + xf_emit(ctx, 1, 0); /* 0000ffff turing USER_PARAM_COUNT */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0xb, 0); /* RO */ + else if (device->chipset >= 0xa0) + xf_emit(ctx, 0x9, 0); /* RO */ + else + xf_emit(ctx, 0x8, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 00000001 EDGE_FLAG */ + xf_emit(ctx, 1, 0); /* 00000001 PROVOKING_VERTEX_LAST */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + /* SEEK */ + xf_emit(ctx, 0xc, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 7f/ff */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 0000000f VP_GP_BUILTIN_ATTR_EN */ + xf_emit(ctx, 1, 4); /* 000001ff UNK1A28 */ + xf_emit(ctx, 1, 8); /* 000001ff UNK0DF0 */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); /* 3ff tesla UNK0D68 */ + else + xf_emit(ctx, 1, 0x7ff); /* 7ff tesla UNK0D68 */ + if (device->chipset == 0xa8) + xf_emit(ctx, 1, 0x1e00); /* 7fff */ + /* SEEK */ + xf_emit(ctx, 0xc, 0); /* RO or close */ + /* SEEK */ + xf_emit(ctx, 1, 0xf); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, (acnt/8)-1, 0); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, 1, 0); /* 0000000f VP_GP_BUILTIN_ATTR_EN */ + if (device->chipset > 0x50 && device->chipset < 0xa0) + xf_emit(ctx, 2, 0); /* ffffffff */ + else + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK0FD8 */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 0x10, 0); /* 0? */ + xf_emit(ctx, 2, 0); /* weird... */ + xf_emit(ctx, 2, 0); /* RO */ + } else { + xf_emit(ctx, 8, 0); /* 0? */ + xf_emit(ctx, 1, 0); /* weird... */ + xf_emit(ctx, 2, 0); /* RO */ + } + /* SEEK */ + xf_emit(ctx, 1, 0); /* ffffffff VB_ELEMENT_BASE */ + xf_emit(ctx, 1, 0); /* ffffffff UNK1438 */ + xf_emit(ctx, acnt, 0); /* 1 tesla UNK1000 */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1118? */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* ffffffff VERTEX_ARRAY_UNK90C */ + xf_emit(ctx, 1, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* ffffffff VERTEX_ARRAY_UNK90C */ + xf_emit(ctx, 1, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* RO */ + xf_emit(ctx, 2, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK111C? */ + xf_emit(ctx, 1, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 000000ff UNK15F4_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff UNK15F4_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 000000ff UNK0F84_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff UNK0F84_ADDRESS_LOW */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* 00003fff VERTEX_ARRAY_ATTRIB_OFFSET */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* 00000fff VERTEX_ARRAY_STRIDE */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* ffffffff VERTEX_ARRAY_LOW */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* 000000ff VERTEX_ARRAY_HIGH */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* ffffffff VERTEX_LIMIT_LOW */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + xf_emit(ctx, acnt, 0); /* 000000ff VERTEX_LIMIT_HIGH */ + xf_emit(ctx, 3, 0); /* f/1f */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, acnt, 0); /* f */ + xf_emit(ctx, 3, 0); /* f/1f */ + } + /* SEEK */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 2, 0); /* RO */ + else + xf_emit(ctx, 5, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* ffff DMA_VTXBUF */ + /* SEEK */ + if (device->chipset < 0xa0) { + xf_emit(ctx, 0x41, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 0x11, 0); /* RO */ + } else if (!IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x50, 0); /* RO */ + else + xf_emit(ctx, 0x58, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0xf); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, (acnt/8)-1, 0); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, 1, 1); /* 1 UNK0DEC */ + /* SEEK */ + xf_emit(ctx, acnt*4, 0); /* ffffffff VTX_ATTR */ + xf_emit(ctx, 4, 0); /* f/1f, 0, 0, 0 */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x1d, 0); /* RO */ + else + xf_emit(ctx, 0x16, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 1, 0xf); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, (acnt/8)-1, 0); /* ffffffff VP_ATTR_EN */ + /* SEEK */ + if (device->chipset < 0xa0) + xf_emit(ctx, 8, 0); /* RO */ + else if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0xc, 0); /* RO */ + else + xf_emit(ctx, 7, 0); /* RO */ + /* SEEK */ + xf_emit(ctx, 0xa, 0); /* RO */ + if (device->chipset == 0xa0) + rep = 0xc; + else + rep = 4; + for (i = 0; i < rep; i++) { + /* SEEK */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x20, 0); /* ffffffff */ + xf_emit(ctx, 0x200, 0); /* ffffffff */ + xf_emit(ctx, 4, 0); /* 7f/ff, 0, 0, 0 */ + xf_emit(ctx, 4, 0); /* ffffffff */ + } + /* SEEK */ + xf_emit(ctx, 1, 0); /* 113/111 */ + xf_emit(ctx, 1, 0xf); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, (acnt/8)-1, 0); /* ffffffff VP_ATTR_EN */ + xf_emit(ctx, acnt/8, 0); /* ffffffff VTX_ATTR_MASK_UNK0DD0 */ + xf_emit(ctx, 1, 0); /* 0000000f VP_GP_BUILTIN_ATTR_EN */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + /* SEEK */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 7, 0); /* weird... */ + else + xf_emit(ctx, 5, 0); /* weird... */ +} + +static void +nv50_gr_construct_gene_eng2d(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* middle of strand 1 on pre-NVA0 [after vfetch], middle of strand 6 on NVAx */ + /* SEEK */ + xf_emit(ctx, 2, 0); /* 0001ffff CLIP_X, CLIP_Y */ + xf_emit(ctx, 2, 0); /* 0000ffff CLIP_W, CLIP_H */ + xf_emit(ctx, 1, 0); /* 00000001 CLIP_ENABLE */ + if (device->chipset < 0xa0) { + /* this is useless on everything but the original NV50, + * guess they forgot to nuke it. Or just didn't bother. */ + xf_emit(ctx, 2, 0); /* 0000ffff IFC_CLIP_X, Y */ + xf_emit(ctx, 2, 1); /* 0000ffff IFC_CLIP_W, H */ + xf_emit(ctx, 1, 0); /* 00000001 IFC_CLIP_ENABLE */ + } + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + xf_emit(ctx, 1, 0x100); /* 0001ffff DST_WIDTH */ + xf_emit(ctx, 1, 0x100); /* 0001ffff DST_HEIGHT */ + xf_emit(ctx, 1, 0x11); /* 3f[NV50]/7f[NV84+] DST_FORMAT */ + xf_emit(ctx, 1, 0); /* 0001ffff DRAW_POINT_X */ + xf_emit(ctx, 1, 8); /* 0000000f DRAW_UNK58C */ + xf_emit(ctx, 1, 0); /* 000fffff SIFC_DST_X_FRACT */ + xf_emit(ctx, 1, 0); /* 0001ffff SIFC_DST_X_INT */ + xf_emit(ctx, 1, 0); /* 000fffff SIFC_DST_Y_FRACT */ + xf_emit(ctx, 1, 0); /* 0001ffff SIFC_DST_Y_INT */ + xf_emit(ctx, 1, 0); /* 000fffff SIFC_DX_DU_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff SIFC_DX_DU_INT */ + xf_emit(ctx, 1, 0); /* 000fffff SIFC_DY_DV_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff SIFC_DY_DV_INT */ + xf_emit(ctx, 1, 1); /* 0000ffff SIFC_WIDTH */ + xf_emit(ctx, 1, 1); /* 0000ffff SIFC_HEIGHT */ + xf_emit(ctx, 1, 0xcf); /* 000000ff SIFC_FORMAT */ + xf_emit(ctx, 1, 2); /* 00000003 SIFC_BITMAP_UNK808 */ + xf_emit(ctx, 1, 0); /* 00000003 SIFC_BITMAP_LINE_PACK_MODE */ + xf_emit(ctx, 1, 0); /* 00000001 SIFC_BITMAP_LSB_FIRST */ + xf_emit(ctx, 1, 0); /* 00000001 SIFC_BITMAP_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000ffff BLIT_DST_X */ + xf_emit(ctx, 1, 0); /* 0000ffff BLIT_DST_Y */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_DU_DX_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff BLIT_DU_DX_INT */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_DV_DY_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff BLIT_DV_DY_INT */ + xf_emit(ctx, 1, 1); /* 0000ffff BLIT_DST_W */ + xf_emit(ctx, 1, 1); /* 0000ffff BLIT_DST_H */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_SRC_X_FRACT */ + xf_emit(ctx, 1, 0); /* 0001ffff BLIT_SRC_X_INT */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_SRC_Y_FRACT */ + xf_emit(ctx, 1, 0); /* 00000001 UNK888 */ + xf_emit(ctx, 1, 4); /* 0000003f UNK884 */ + xf_emit(ctx, 1, 0); /* 00000007 UNK880 */ + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK0FB8 */ + xf_emit(ctx, 1, 0x15); /* 000000ff tesla UNK128C */ + xf_emit(ctx, 2, 0); /* 00000007, ffff0ff3 */ + xf_emit(ctx, 1, 0); /* 00000001 UNK260 */ + xf_emit(ctx, 1, 0x4444480); /* 1fffffff UNK870 */ + /* SEEK */ + xf_emit(ctx, 0x10, 0); + /* SEEK */ + xf_emit(ctx, 0x27, 0); +} + +static void +nv50_gr_construct_gene_csched(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* middle of strand 1 on pre-NVA0 [after eng2d], middle of strand 0 on NVAx */ + /* SEEK */ + xf_emit(ctx, 2, 0); /* 00007fff WINDOW_OFFSET_XY... what is it doing here??? */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1924 */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 0); /* 000003ff */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* ffffffff turing UNK364 */ + xf_emit(ctx, 1, 0); /* 0000000f turing UNK36C */ + xf_emit(ctx, 1, 0); /* 0000ffff USER_PARAM_COUNT */ + xf_emit(ctx, 1, 0x100); /* 00ffffff turing UNK384 */ + xf_emit(ctx, 1, 0); /* 0000000f turing UNK2A0 */ + xf_emit(ctx, 1, 0); /* 0000ffff GRIDID */ + xf_emit(ctx, 1, 0x10001); /* ffffffff GRIDDIM_XY */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0x10001); /* ffffffff BLOCKDIM_XY */ + xf_emit(ctx, 1, 1); /* 0000ffff BLOCKDIM_Z */ + xf_emit(ctx, 1, 0x10001); /* 00ffffff BLOCK_ALLOC */ + xf_emit(ctx, 1, 1); /* 00000001 LANES32 */ + xf_emit(ctx, 1, 4); /* 000000ff FP_REG_ALLOC_TEMP */ + xf_emit(ctx, 1, 2); /* 00000003 REG_MODE */ + /* SEEK */ + xf_emit(ctx, 0x40, 0); /* ffffffff USER_PARAM */ + switch (device->chipset) { + case 0x50: + case 0x92: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0x80, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 0x10*2, 0); /* ffffffff, 1f */ + break; + case 0x84: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0x60, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 0xc*2, 0); /* ffffffff, 1f */ + break; + case 0x94: + case 0x96: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0x40, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 8*2, 0); /* ffffffff, 1f */ + break; + case 0x86: + case 0x98: + xf_emit(ctx, 4, 0); /* f, 0, 0, 0 */ + xf_emit(ctx, 0x10, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 2*2, 0); /* ffffffff, 1f */ + break; + case 0xa0: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0xf0, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 0x1e*2, 0); /* ffffffff, 1f */ + break; + case 0xa3: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0x60, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 0xc*2, 0); /* ffffffff, 1f */ + break; + case 0xa5: + case 0xaf: + xf_emit(ctx, 8, 0); /* 7, 0, 0, 0, ... */ + xf_emit(ctx, 0x30, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 6*2, 0); /* ffffffff, 1f */ + break; + case 0xaa: + xf_emit(ctx, 0x12, 0); + break; + case 0xa8: + case 0xac: + xf_emit(ctx, 4, 0); /* f, 0, 0, 0 */ + xf_emit(ctx, 0x10, 0); /* fff */ + xf_emit(ctx, 2, 0); /* ff, fff */ + xf_emit(ctx, 2*2, 0); /* ffffffff, 1f */ + break; + } + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 0); /* 00000000 */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 0000001f */ + xf_emit(ctx, 4, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000003 turing UNK35C */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 4, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000003 turing UNK35C */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 000000ff */ +} + +static void +nv50_gr_construct_gene_unk1cxx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 2, 0); /* 00007fff WINDOW_OFFSET_XY */ + xf_emit(ctx, 1, 0x3f800000); /* ffffffff LINE_WIDTH */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1658 */ + xf_emit(ctx, 1, 0); /* 00000001 POLYGON_SMOOTH_ENABLE */ + xf_emit(ctx, 3, 0); /* 00000001 POLYGON_OFFSET_*_ENABLE */ + xf_emit(ctx, 1, 4); /* 0000000f CULL_MODE */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 0); /* 00000001 POINT_SPRITE_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK165C */ + xf_emit(ctx, 0x10, 0); /* 00000001 SCISSOR_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0x00ffff00); /* 00ffffff LINE_STIPPLE_PATTERN */ + xf_emit(ctx, 1, 0); /* ffffffff POLYGON_OFFSET_UNITS */ + xf_emit(ctx, 1, 0); /* ffffffff POLYGON_OFFSET_FACTOR */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1668 */ + xf_emit(ctx, 2, 0); /* 07ffffff SCREEN_SCISSOR */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0x11); /* 0000007f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 0000007f RT_FORMAT */ + xf_emit(ctx, 8, 0); /* 00000001 RT_HORIZ_LINEAR */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000001 ALPHA_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 ALPHA_TEST_FUNC */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 3); /* 00000003 UNK16B4 */ + else if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 1); /* 00000001 UNK16B4 */ + xf_emit(ctx, 1, 0); /* 00000003 MULTISAMPLE_CTRL */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK0F90 */ + xf_emit(ctx, 1, 2); /* 00000003 tesla UNK143C */ + xf_emit(ctx, 2, 0x04000000); /* 07ffffff tesla UNK0D6C */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 5); /* 0000000f UNK1408 */ + xf_emit(ctx, 1, 0x52); /* 000001ff SEMANTIC_PTSZ */ + xf_emit(ctx, 1, 0); /* ffffffff POINT_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 00000007 tesla UNK0FB4 */ + if (device->chipset != 0x50) { + xf_emit(ctx, 1, 0); /* 3ff */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK1110 */ + } + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1928 */ + xf_emit(ctx, 0x10, 0); /* ffffffff DEPTH_RANGE_NEAR */ + xf_emit(ctx, 0x10, 0x3f800000); /* ffffffff DEPTH_RANGE_FAR */ + xf_emit(ctx, 1, 0x10); /* 000000ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 0x20, 0); /* 07ffffff VIEWPORT_HORIZ, then VIEWPORT_VERT. (W&0x3fff)<<13 | (X&0x1fff). */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK187C */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_MASK */ + xf_emit(ctx, 1, 0x8100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 5); /* 0000000f tesla UNK1220 */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 000000ff tesla UNK1A20 */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE */ + xf_emit(ctx, 4, 0xffff); /* 0000ffff MSAA_MASK */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK1100 */ + if (device->chipset < 0xa0) + xf_emit(ctx, 0x1c, 0); /* RO */ + else if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x9, 0); + xf_emit(ctx, 1, 0); /* 00000001 UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0x00ffff00); /* 00ffffff LINE_STIPPLE_PATTERN */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ + xf_emit(ctx, 1, 0); /* 00000003 WINDOW_ORIGIN */ + if (device->chipset != 0x50) { + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK1100 */ + xf_emit(ctx, 1, 0); /* 3ff */ + } + /* XXX: the following block could belong either to unk1cxx, or + * to STRMOUT. Rather hard to tell. */ + if (device->chipset < 0xa0) + xf_emit(ctx, 0x25, 0); + else + xf_emit(ctx, 0x3b, 0); +} + +static void +nv50_gr_construct_gene_strmout(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 1, 0x102); /* 0000ffff STRMOUT_BUFFER_CTRL */ + xf_emit(ctx, 1, 0); /* ffffffff STRMOUT_PRIMITIVE_COUNT */ + xf_emit(ctx, 4, 4); /* 000000ff STRMOUT_NUM_ATTRIBS */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 4, 0); /* ffffffff UNK1A8C */ + xf_emit(ctx, 4, 0); /* ffffffff UNK1780 */ + } + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 4); /* 0000007f VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0x3ff); /* 000003ff tesla UNK0D68 */ + else + xf_emit(ctx, 1, 0x7ff); /* 000007ff tesla UNK0D68 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + /* SEEK */ + xf_emit(ctx, 1, 0x102); /* 0000ffff STRMOUT_BUFFER_CTRL */ + xf_emit(ctx, 1, 0); /* ffffffff STRMOUT_PRIMITIVE_COUNT */ + xf_emit(ctx, 4, 0); /* 000000ff STRMOUT_ADDRESS_HIGH */ + xf_emit(ctx, 4, 0); /* ffffffff STRMOUT_ADDRESS_LOW */ + xf_emit(ctx, 4, 4); /* 000000ff STRMOUT_NUM_ATTRIBS */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 4, 0); /* ffffffff UNK1A8C */ + xf_emit(ctx, 4, 0); /* ffffffff UNK1780 */ + } + xf_emit(ctx, 1, 0); /* 0000ffff DMA_STRMOUT */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_QUERY */ + xf_emit(ctx, 1, 0); /* 000000ff QUERY_ADDRESS_HIGH */ + xf_emit(ctx, 2, 0); /* ffffffff QUERY_ADDRESS_LOW QUERY_COUNTER */ + xf_emit(ctx, 2, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + /* SEEK */ + xf_emit(ctx, 0x20, 0); /* ffffffff STRMOUT_MAP */ + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 0); /* 00000000? */ + xf_emit(ctx, 2, 0); /* ffffffff */ +} + +static void +nv50_gr_construct_gene_ropm1(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 1, 0x4e3bfdf); /* ffffffff UNK0D64 */ + xf_emit(ctx, 1, 0x4e3bfdf); /* ffffffff UNK0DF4 */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0x11); /* 000000ff tesla UNK1968 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ +} + +static void +nv50_gr_construct_gene_ropm2(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + /* SEEK */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_QUERY */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 2, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 000000ff QUERY_ADDRESS_HIGH */ + xf_emit(ctx, 2, 0); /* ffffffff QUERY_ADDRESS_LOW, COUNTER */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 0); /* 7 */ + /* SEEK */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_QUERY */ + xf_emit(ctx, 1, 0); /* 000000ff QUERY_ADDRESS_HIGH */ + xf_emit(ctx, 2, 0); /* ffffffff QUERY_ADDRESS_LOW, COUNTER */ + xf_emit(ctx, 1, 0x4e3bfdf); /* ffffffff UNK0D64 */ + xf_emit(ctx, 1, 0x4e3bfdf); /* ffffffff UNK0DF4 */ + xf_emit(ctx, 1, 0); /* 00000001 eng2d UNK260 */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* 00000007 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0x11); /* 000000ff tesla UNK1968 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ +} + +static void +nv50_gr_construct_gene_ropc(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int magic2; + if (device->chipset == 0x50) { + magic2 = 0x00003e60; + } else if (!IS_NVA3F(device->chipset)) { + magic2 = 0x001ffe67; + } else { + magic2 = 0x00087e67; + } + xf_emit(ctx, 1, 0); /* f/7 MUTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_BACK_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 2); /* 00000003 tesla UNK143C */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, magic2); /* 001fffff tesla UNK0F78 */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000007 DEPTH_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_FRONT_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + if (device->chipset >= 0xa0 && !IS_NVAAF(device->chipset)) + xf_emit(ctx, 1, 0x15); /* 000000ff */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 0x10); /* 3ff/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* ffffffff CLEAR_DEPTH */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + if (device->chipset == 0x86 || device->chipset == 0x92 || device->chipset == 0x98 || device->chipset >= 0xa0) { + xf_emit(ctx, 3, 0); /* ff, ffffffff, ffffffff */ + xf_emit(ctx, 1, 4); /* 7 */ + xf_emit(ctx, 1, 0x400); /* fffffff */ + xf_emit(ctx, 1, 0x300); /* ffff */ + xf_emit(ctx, 1, 0x1001); /* 1fff */ + if (device->chipset != 0xa0) { + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0); /* 0000000f UNK15C8 */ + else + xf_emit(ctx, 1, 0x15); /* ff */ + } + } + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_BACK_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 2); /* 00000003 tesla UNK143C */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000007 DEPTH_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_FRONT_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 0x10); /* 7f/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1900 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_BACK_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_REF */ + xf_emit(ctx, 2, 0); /* ffffffff DEPTH_BOUNDS */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000007 DEPTH_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0FB0 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_FRONT_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_REF */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 0x10); /* 7f/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 0x10, 0); /* ffffffff DEPTH_RANGE_NEAR */ + xf_emit(ctx, 0x10, 0x3f800000); /* ffffffff DEPTH_RANGE_FAR */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_BACK_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_FUNC_REF */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 2, 0); /* ffffffff DEPTH_BOUNDS */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000007 DEPTH_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0); /* 000000ff CLEAR_STENCIL */ + xf_emit(ctx, 1, 0); /* 00000007 STENCIL_FRONT_FUNC_FUNC */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_FUNC_REF */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 1, 0x10); /* 7f/ff VIEW_VOLUME_CLIP_CTRL */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 0x3f); /* 0000003f UNK1590 */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 2, 0); /* ffff0ff3, ffff */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0FB0 */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff CLEAR_DEPTH */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK19CC */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 2, 0); + xf_emit(ctx, 1, 0x1001); + xf_emit(ctx, 0xb, 0); + } else { + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + } + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, 0x11); /* 3f/7f */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + if (device->chipset != 0x50) { + xf_emit(ctx, 1, 0); /* 0000000f LOGIC_OP */ + xf_emit(ctx, 1, 0); /* 000000ff */ + } + xf_emit(ctx, 1, 0); /* 00000007 OPERATION */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* 00000003 UNK0F90 */ + xf_emit(ctx, 2, 1); /* 00000007 BLEND_EQUATION_RGB, ALPHA */ + xf_emit(ctx, 1, 1); /* 00000001 UNK133C */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_RGB */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, magic2); /* 001fffff tesla UNK0F78 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK12E4 */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_RGB */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_ALPHA */ + xf_emit(ctx, 8, 1); /* 00000001 IBLEND_UNK00 */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_RGB */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1140 */ + xf_emit(ctx, 2, 0); /* 00000001 */ + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* 0000000f */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 2, 0); /* 00000001 */ + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + } else if (device->chipset >= 0xa0) { + xf_emit(ctx, 2, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 2, 0); /* 00000001 */ + } else { + xf_emit(ctx, 1, 0); /* 00000007 MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1430 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + } + xf_emit(ctx, 4, 0); /* ffffffff CLEAR_COLOR */ + xf_emit(ctx, 4, 0); /* ffffffff BLEND_COLOR A R G B */ + xf_emit(ctx, 1, 0); /* 00000fff eng2d UNK2B0 */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 2, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 UNK133C */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_RGB */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_RGB */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_ALPHA */ + xf_emit(ctx, 1, 0); /* 00000001 UNK19C0 */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f LOGIC_OP */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0); /* 00000001 UNK12E4? NVA3+ only? */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 8, 1); /* 00000001 IBLEND_UNK00 */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_RGB */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_RGB */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_ALPHA */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK15C4 */ + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1140 */ + } + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + xf_emit(ctx, 1, 0); /* 00000007 PATTERN_COLOR_FORMAT */ + xf_emit(ctx, 2, 0); /* ffffffff PATTERN_MONO_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 PATTERN_MONO_FORMAT */ + xf_emit(ctx, 2, 0); /* ffffffff PATTERN_MONO_BITMAP */ + xf_emit(ctx, 1, 0); /* 00000003 PATTERN_SELECT */ + xf_emit(ctx, 1, 0); /* 000000ff ROP */ + xf_emit(ctx, 1, 0); /* ffffffff BETA1 */ + xf_emit(ctx, 1, 0); /* ffffffff BETA4 */ + xf_emit(ctx, 1, 0); /* 00000007 OPERATION */ + xf_emit(ctx, 0x50, 0); /* 10x ffffff, ffffff, ffffff, ffffff, 3 PATTERN */ +} + +static void +nv50_gr_construct_xfer_unk84xx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int magic3; + switch (device->chipset) { + case 0x50: + magic3 = 0x1000; + break; + case 0x86: + case 0x98: + case 0xa8: + case 0xaa: + case 0xac: + case 0xaf: + magic3 = 0x1e00; + break; + default: + magic3 = 0; + } + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 7f/ff[NVA0+] VP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 111/113[NVA0+] */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x1f, 0); /* ffffffff */ + else if (device->chipset >= 0xa0) + xf_emit(ctx, 0x0f, 0); /* ffffffff */ + else + xf_emit(ctx, 0x10, 0); /* fffffff VP_RESULT_MAP_1 up */ + xf_emit(ctx, 2, 0); /* f/1f[NVA3], fffffff/ffffffff[NVA0+] */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_RESULT_MAP_SIZE */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0x03020100); /* ffffffff */ + else + xf_emit(ctx, 1, 0x00608080); /* fffffff VP_RESULT_MAP_0 */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 2, 0); /* 111/113, 7f/ff */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0x80); /* 0000ffff GP_VERTEX_OUTPUT_COUNT */ + if (magic3) + xf_emit(ctx, 1, magic3); /* 00007fff tesla UNK141C */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 111/113 */ + xf_emit(ctx, 0x1f, 0); /* ffffffff GP_RESULT_MAP_1 up */ + xf_emit(ctx, 1, 0); /* 0000001f */ + xf_emit(ctx, 1, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 0x80); /* 0000ffff GP_VERTEX_OUTPUT_COUNT */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0x03020100); /* ffffffff GP_RESULT_MAP_0 */ + xf_emit(ctx, 1, 3); /* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */ + if (magic3) + xf_emit(ctx, 1, magic3); /* 7fff tesla UNK141C */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 0); /* 00000001 PROVOKING_VERTEX_LAST */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 111/113 */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 3); /* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */ + xf_emit(ctx, 1, 0); /* 00000001 PROVOKING_VERTEX_LAST */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK13A0 */ + xf_emit(ctx, 1, 4); /* 7f/ff VP_REG_ALLOC_RESULT */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + xf_emit(ctx, 1, 0); /* 111/113 */ + if (device->chipset == 0x94 || device->chipset == 0x96) + xf_emit(ctx, 0x1020, 0); /* 4 x (0x400 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */ + else if (device->chipset < 0xa0) + xf_emit(ctx, 0xa20, 0); /* 4 x (0x280 x 0xffffffff, ff, 0, 0, 0, 4 x ffffffff) */ + else if (!IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x210, 0); /* ffffffff */ + else + xf_emit(ctx, 0x410, 0); /* ffffffff */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff GP_RESULT_MAP_SIZE */ + xf_emit(ctx, 1, 3); /* 00000003 GP_OUTPUT_PRIMITIVE_TYPE */ + xf_emit(ctx, 1, 0); /* 00000001 PROVOKING_VERTEX_LAST */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ +} + +static void +nv50_gr_construct_xfer_tprop(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int magic1, magic2; + if (device->chipset == 0x50) { + magic1 = 0x3ff; + magic2 = 0x00003e60; + } else if (!IS_NVA3F(device->chipset)) { + magic1 = 0x7ff; + magic2 = 0x001ffe67; + } else { + magic1 = 0x7ff; + magic2 = 0x00087e67; + } + xf_emit(ctx, 1, 0); /* 00000007 ALPHA_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* ffffffff ALPHA_TEST_REF */ + xf_emit(ctx, 1, 0); /* 00000001 ALPHA_TEST_ENABLE */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000000f UNK16A0 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_BACK_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_BACK_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 4, 0); /* ffffffff BLEND_COLOR */ + xf_emit(ctx, 1, 0); /* 00000001 UNK19C0 */ + xf_emit(ctx, 1, 0); /* 00000001 UNK0FDC */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + xf_emit(ctx, 1, 0); /* ff[NV50]/3ff[NV84+] */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 4, 0xffff); /* 0000ffff MSAA_MASK */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 3, 0); /* 00000007 STENCIL_FRONT_OP_FAIL, ZFAIL, ZPASS */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_BACK_ENABLE */ + xf_emit(ctx, 2, 0); /* 00007fff WINDOW_OFFSET_XY */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK19CC */ + xf_emit(ctx, 1, 0); /* 7 */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff COLOR_KEY */ + xf_emit(ctx, 1, 0); /* 00000001 COLOR_KEY_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 COLOR_KEY_FORMAT */ + xf_emit(ctx, 2, 0); /* ffffffff SIFC_BITMAP_COLOR */ + xf_emit(ctx, 1, 1); /* 00000001 SIFC_BITMAP_WRITE_BIT0_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 ALPHA_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 ALPHA_TEST_ENABLE */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK16B4 */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1298 */ + } else if (device->chipset >= 0xa0) { + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK16B4 */ + xf_emit(ctx, 1, 0); /* 00000003 */ + } else { + xf_emit(ctx, 1, 0); /* 00000003 MULTISAMPLE_CTRL */ + } + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_ALPHA */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_RGB */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_RGB */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_RGB */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 0); /* 00000001 UNK12E4 */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_RGB */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_ALPHA */ + xf_emit(ctx, 8, 1); /* 00000001 IBLEND_UNK00 */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_SRC_RGB */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_DST_RGB */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_SRC_ALPHA */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_DST_ALPHA */ + xf_emit(ctx, 1, 0); /* 00000001 UNK1140 */ + } + xf_emit(ctx, 1, 1); /* 00000001 UNK133C */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000003 UNK0F90 */ + xf_emit(ctx, 1, 0); /* 00000001 FRAMEBUFFER_SRGB */ + xf_emit(ctx, 1, 0); /* 7 */ + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + xf_emit(ctx, 1, 0); /* 00000007 OPERATION */ + xf_emit(ctx, 1, 0xcf); /* 000000ff SIFC_FORMAT */ + xf_emit(ctx, 1, 0xcf); /* 000000ff DRAW_COLOR_FORMAT */ + xf_emit(ctx, 1, 0xcf); /* 000000ff SRC_FORMAT */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + xf_emit(ctx, 1, 0); /* 7/f[NVA3] MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_ALPHA */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_RGB */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_RGB */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 1, 1); /* 00000001 UNK133C */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 8, 1); /* 00000001 UNK19E0 */ + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, magic2); /* 001fffff tesla UNK0F78 */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0); /* ff */ + else + xf_emit(ctx, 3, 0); /* 1, 7, 3ff */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000003 UNK0F90 */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_DU_DX_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff BLIT_DU_DX_INT */ + xf_emit(ctx, 1, 0); /* 000fffff BLIT_DV_DY_FRACT */ + xf_emit(ctx, 1, 1); /* 0001ffff BLIT_DV_DY_INT */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, magic1); /* 3ff/7ff tesla UNK0D68 */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 8, 0); /* 0000ffff DMA_COLOR */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_GLOBAL */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_LOCAL */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_STACK */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_DST */ + xf_emit(ctx, 1, 0); /* 7 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 8, 0); /* 000000ff RT_ADDRESS_HIGH */ + xf_emit(ctx, 8, 0); /* ffffffff RT_LAYER_STRIDE */ + xf_emit(ctx, 8, 0); /* ffffffff RT_ADDRESS_LOW */ + xf_emit(ctx, 8, 8); /* 0000007f RT_TILE_MODE */ + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 8, 0x400); /* 0fffffff RT_HORIZ */ + xf_emit(ctx, 8, 0x300); /* 0000ffff RT_VERT */ + xf_emit(ctx, 1, 1); /* 00001fff RT_ARRAY_MODE */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, 0x20); /* 00000fff DST_TILE_MODE */ + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 0x100); /* 0001ffff DST_HEIGHT */ + xf_emit(ctx, 1, 0); /* 000007ff DST_LAYER */ + xf_emit(ctx, 1, 1); /* 00000001 DST_LINEAR */ + xf_emit(ctx, 1, 0); /* ffffffff DST_ADDRESS_LOW */ + xf_emit(ctx, 1, 0); /* 000000ff DST_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0x40); /* 0007ffff DST_PITCH */ + xf_emit(ctx, 1, 0x100); /* 0001ffff DST_WIDTH */ + xf_emit(ctx, 1, 0); /* 0000ffff */ + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK15AC */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0); /* 00000003 UNK0F90 */ + xf_emit(ctx, 1, 0); /* 00000007 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, magic2); /* 001fffff tesla UNK0F78 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 2); /* 00000003 tesla UNK143C */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_ZETA */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 2, 0); /* ffff, ff/3ff */ + xf_emit(ctx, 1, 0); /* 0001ffff GP_BUILTIN_RESULT_EN */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 0); /* 000000ff STENCIL_FRONT_MASK */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* ffffffff ZETA_LAYER_STRIDE */ + xf_emit(ctx, 1, 0); /* 000000ff ZETA_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff ZETA_ADDRESS_LOW */ + xf_emit(ctx, 1, 4); /* 00000007 ZETA_TILE_MODE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + xf_emit(ctx, 1, 0x400); /* 0fffffff ZETA_HORIZ */ + xf_emit(ctx, 1, 0x300); /* 0000ffff ZETA_VERT */ + xf_emit(ctx, 1, 0x1001); /* 00001fff ZETA_ARRAY_MODE */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 0); /* 00000001 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0x11); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 3f/7f RT_FORMAT */ + xf_emit(ctx, 1, 0x0fac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0xf); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 7, 0); /* 0000000f COLOR_MASK */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000003 UNK0F90 */ + xf_emit(ctx, 1, 0); /* 00000001 FRAMEBUFFER_SRGB */ + xf_emit(ctx, 1, 0); /* 7 */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 0); /* 00000001 UNK1140 */ + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + } + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 UNK1534 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0x0fac6881); /* fffffff */ + xf_emit(ctx, 1, magic2); /* 001fffff tesla UNK0F78 */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_BOUNDS_EN */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE_ENABLE */ + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK0FB0 */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000001 STENCIL_FRONT_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK15B4 */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK19CC */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0); /* 00000001 SAMPLECNT_ENABLE */ + xf_emit(ctx, 1, 0); /* 0000000f ZETA_FORMAT */ + xf_emit(ctx, 1, 1); /* 00000001 ZETA_ENABLE */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* 0000000f tesla UNK15C8 */ + } + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A3C */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 3, 0); /* 7/f, 1, ffff0ff3 */ + xf_emit(ctx, 1, 0xfac6881); /* fffffff */ + xf_emit(ctx, 4, 0); /* 1, 1, 1, 3ff */ + xf_emit(ctx, 1, 4); /* 7 */ + xf_emit(ctx, 1, 0); /* 1 */ + xf_emit(ctx, 2, 1); /* 1 */ + xf_emit(ctx, 2, 0); /* 7, f */ + xf_emit(ctx, 1, 1); /* 1 */ + xf_emit(ctx, 1, 0); /* 7/f */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 0x9, 0); /* 1 */ + else + xf_emit(ctx, 0x8, 0); /* 1 */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 8, 1); /* 1 */ + xf_emit(ctx, 1, 0x11); /* 7f */ + xf_emit(ctx, 7, 0); /* 7f */ + xf_emit(ctx, 1, 0xfac6881); /* fffffff */ + xf_emit(ctx, 1, 0xf); /* f */ + xf_emit(ctx, 7, 0); /* f */ + xf_emit(ctx, 1, 0x11); /* 7f */ + xf_emit(ctx, 1, 1); /* 1 */ + xf_emit(ctx, 5, 0); /* 1, 7, 3ff, 3, 7 */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 0); /* 00000001 UNK1140 */ + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + } + } +} + +static void +nv50_gr_construct_xfer_tex(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 2, 0); /* 1 LINKED_TSC. yes, 2. */ + if (device->chipset != 0x50) + xf_emit(ctx, 1, 0); /* 3 */ + xf_emit(ctx, 1, 1); /* 1ffff BLIT_DU_DX_INT */ + xf_emit(ctx, 1, 0); /* fffff BLIT_DU_DX_FRACT */ + xf_emit(ctx, 1, 1); /* 1ffff BLIT_DV_DY_INT */ + xf_emit(ctx, 1, 0); /* fffff BLIT_DV_DY_FRACT */ + if (device->chipset == 0x50) + xf_emit(ctx, 1, 0); /* 3 BLIT_CONTROL */ + else + xf_emit(ctx, 2, 0); /* 3ff, 1 */ + xf_emit(ctx, 1, 0x2a712488); /* ffffffff SRC_TIC_0 */ + xf_emit(ctx, 1, 0); /* ffffffff SRC_TIC_1 */ + xf_emit(ctx, 1, 0x4085c000); /* ffffffff SRC_TIC_2 */ + xf_emit(ctx, 1, 0x40); /* ffffffff SRC_TIC_3 */ + xf_emit(ctx, 1, 0x100); /* ffffffff SRC_TIC_4 */ + xf_emit(ctx, 1, 0x10100); /* ffffffff SRC_TIC_5 */ + xf_emit(ctx, 1, 0x02800000); /* ffffffff SRC_TIC_6 */ + xf_emit(ctx, 1, 0); /* ffffffff SRC_TIC_7 */ + if (device->chipset == 0x50) { + xf_emit(ctx, 1, 0); /* 00000001 turing UNK358 */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A34? */ + xf_emit(ctx, 1, 0); /* 00000003 turing UNK37C tesla UNK1690 */ + xf_emit(ctx, 1, 0); /* 00000003 BLIT_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000001 turing UNK32C tesla UNK0F94 */ + } else if (!IS_NVAAF(device->chipset)) { + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A34? */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1664 / turing UNK03E8 */ + xf_emit(ctx, 1, 0); /* 00000003 */ + xf_emit(ctx, 1, 0); /* 000003ff */ + } else { + xf_emit(ctx, 0x6, 0); + } + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A34 */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_TEXTURE */ + xf_emit(ctx, 1, 0); /* 0000ffff DMA_SRC */ +} + +static void +nv50_gr_construct_xfer_unk8cxx(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + xf_emit(ctx, 1, 0); /* 00000001 UNK1534 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 2, 0); /* 7, ffff0ff3 */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE */ + xf_emit(ctx, 1, 0x04e3bfdf); /* ffffffff UNK0D64 */ + xf_emit(ctx, 1, 0x04e3bfdf); /* ffffffff UNK0DF4 */ + xf_emit(ctx, 1, 1); /* 00000001 UNK15B4 */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0x00ffff00); /* 00ffffff LINE_STIPPLE_PATTERN */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK0F98 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1668 */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_STIPPLE_ENABLE */ + xf_emit(ctx, 1, 0x00ffff00); /* 00ffffff LINE_STIPPLE_PATTERN */ + xf_emit(ctx, 1, 0); /* 00000001 POLYGON_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 UNK1534 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1658 */ + xf_emit(ctx, 1, 0); /* 00000001 LINE_SMOOTH_ENABLE */ + xf_emit(ctx, 1, 0); /* ffff0ff3 */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 DEPTH_WRITE */ + xf_emit(ctx, 1, 1); /* 00000001 UNK15B4 */ + xf_emit(ctx, 1, 0); /* 00000001 POINT_SPRITE_ENABLE */ + xf_emit(ctx, 1, 1); /* 00000001 tesla UNK165C */ + xf_emit(ctx, 1, 0x30201000); /* ffffffff tesla UNK1670 */ + xf_emit(ctx, 1, 0x70605040); /* ffffffff tesla UNK1670 */ + xf_emit(ctx, 1, 0xb8a89888); /* ffffffff tesla UNK1670 */ + xf_emit(ctx, 1, 0xf8e8d8c8); /* ffffffff tesla UNK1670 */ + xf_emit(ctx, 1, 0); /* 00000001 VERTEX_TWO_SIDE_ENABLE */ + xf_emit(ctx, 1, 0x1a); /* 0000001f POLYGON_MODE */ +} + +static void +nv50_gr_construct_xfer_tp(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + if (device->chipset < 0xa0) { + nv50_gr_construct_xfer_unk84xx(ctx); + nv50_gr_construct_xfer_tprop(ctx); + nv50_gr_construct_xfer_tex(ctx); + nv50_gr_construct_xfer_unk8cxx(ctx); + } else { + nv50_gr_construct_xfer_tex(ctx); + nv50_gr_construct_xfer_tprop(ctx); + nv50_gr_construct_xfer_unk8cxx(ctx); + nv50_gr_construct_xfer_unk84xx(ctx); + } +} + +static void +nv50_gr_construct_xfer_mpc(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i, mpcnt = 2; + switch (device->chipset) { + case 0x98: + case 0xaa: + mpcnt = 1; + break; + case 0x50: + case 0x84: + case 0x86: + case 0x92: + case 0x94: + case 0x96: + case 0xa8: + case 0xac: + mpcnt = 2; + break; + case 0xa0: + case 0xa3: + case 0xa5: + case 0xaf: + mpcnt = 3; + break; + } + for (i = 0; i < mpcnt; i++) { + xf_emit(ctx, 1, 0); /* ff */ + xf_emit(ctx, 1, 0x80); /* ffffffff tesla UNK1404 */ + xf_emit(ctx, 1, 0x80007004); /* ffffffff tesla UNK12B0 */ + xf_emit(ctx, 1, 0x04000400); /* ffffffff */ + if (device->chipset >= 0xa0) + xf_emit(ctx, 1, 0xc0); /* 00007fff tesla UNK152C */ + xf_emit(ctx, 1, 0x1000); /* 0000ffff tesla UNK0D60 */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A30 */ + if (device->chipset == 0x86 || device->chipset == 0x98 || device->chipset == 0xa8 || IS_NVAAF(device->chipset)) { + xf_emit(ctx, 1, 0xe00); /* 7fff */ + xf_emit(ctx, 1, 0x1e00); /* 7fff */ + } + xf_emit(ctx, 1, 1); /* 000000ff VP_REG_ALLOC_TEMP */ + xf_emit(ctx, 1, 0); /* 00000001 LINKED_TSC */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + if (device->chipset == 0x50) + xf_emit(ctx, 2, 0x1000); /* 7fff tesla UNK141C */ + xf_emit(ctx, 1, 1); /* 000000ff GP_REG_ALLOC_TEMP */ + xf_emit(ctx, 1, 0); /* 00000001 GP_ENABLE */ + xf_emit(ctx, 1, 4); /* 000000ff FP_REG_ALLOC_TEMP */ + xf_emit(ctx, 1, 2); /* 00000003 REG_MODE */ + if (IS_NVAAF(device->chipset)) + xf_emit(ctx, 0xb, 0); /* RO */ + else if (device->chipset >= 0xa0) + xf_emit(ctx, 0xc, 0); /* RO */ + else + xf_emit(ctx, 0xa, 0); /* RO */ + } + xf_emit(ctx, 1, 0x08100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + if (device->chipset >= 0xa0) { + xf_emit(ctx, 1, 0x1fe21); /* 0003ffff tesla UNK0FAC */ + } + xf_emit(ctx, 3, 0); /* 7fff, 0, 0 */ + xf_emit(ctx, 1, 0); /* 00000001 tesla UNK1534 */ + xf_emit(ctx, 1, 0); /* 7/f MULTISAMPLE_SAMPLES_LOG2 */ + xf_emit(ctx, 4, 0xffff); /* 0000ffff MSAA_MASK */ + xf_emit(ctx, 1, 1); /* 00000001 LANES32 */ + xf_emit(ctx, 1, 0x10001); /* 00ffffff BLOCK_ALLOC */ + xf_emit(ctx, 1, 0x10001); /* ffffffff BLOCKDIM_XY */ + xf_emit(ctx, 1, 1); /* 0000ffff BLOCKDIM_Z */ + xf_emit(ctx, 1, 0); /* ffffffff SHARED_SIZE */ + xf_emit(ctx, 1, 0x1fe21); /* 1ffff/3ffff[NVA0+] tesla UNk0FAC */ + xf_emit(ctx, 1, 0); /* ffffffff tesla UNK1A34 */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 1); /* 0000001f tesla UNK169C */ + xf_emit(ctx, 1, 0); /* ff/3ff */ + xf_emit(ctx, 1, 0); /* 1 LINKED_TSC */ + xf_emit(ctx, 1, 0); /* ff FP_ADDRESS_HIGH */ + xf_emit(ctx, 1, 0); /* ffffffff FP_ADDRESS_LOW */ + xf_emit(ctx, 1, 0x08100c12); /* 1fffffff FP_INTERPOLANT_CTRL */ + xf_emit(ctx, 1, 4); /* 00000007 FP_CONTROL */ + xf_emit(ctx, 1, 0); /* 000000ff FRAG_COLOR_CLAMP_EN */ + xf_emit(ctx, 1, 2); /* 00000003 REG_MODE */ + xf_emit(ctx, 1, 0x11); /* 0000007f RT_FORMAT */ + xf_emit(ctx, 7, 0); /* 0000007f RT_FORMAT */ + xf_emit(ctx, 1, 0); /* 00000007 */ + xf_emit(ctx, 1, 0xfac6881); /* 0fffffff RT_CONTROL */ + xf_emit(ctx, 1, 0); /* 00000003 MULTISAMPLE_CTRL */ + if (IS_NVA3F(device->chipset)) + xf_emit(ctx, 1, 3); /* 00000003 tesla UNK16B4 */ + xf_emit(ctx, 1, 0); /* 00000001 ALPHA_TEST_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000007 ALPHA_TEST_FUNC */ + xf_emit(ctx, 1, 0); /* 00000001 FRAMEBUFFER_SRGB */ + xf_emit(ctx, 1, 4); /* ffffffff tesla UNK1400 */ + xf_emit(ctx, 8, 0); /* 00000001 BLEND_ENABLE */ + xf_emit(ctx, 1, 0); /* 00000001 LOGIC_OP_ENABLE */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_RGB */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_RGB */ + xf_emit(ctx, 1, 2); /* 0000001f BLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 1, 1); /* 0000001f BLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 1, 1); /* 00000007 BLEND_EQUATION_ALPHA */ + xf_emit(ctx, 1, 1); /* 00000001 UNK133C */ + if (IS_NVA3F(device->chipset)) { + xf_emit(ctx, 1, 0); /* 00000001 UNK12E4 */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_RGB */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_RGB */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_RGB */ + xf_emit(ctx, 8, 2); /* 0000001f IBLEND_FUNC_SRC_ALPHA */ + xf_emit(ctx, 8, 1); /* 0000001f IBLEND_FUNC_DST_ALPHA */ + xf_emit(ctx, 8, 1); /* 00000007 IBLEND_EQUATION_ALPHA */ + xf_emit(ctx, 8, 1); /* 00000001 IBLEND_UNK00 */ + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK1928 */ + xf_emit(ctx, 1, 0); /* 00000001 UNK1140 */ + } + xf_emit(ctx, 1, 0); /* 00000003 tesla UNK0F90 */ + xf_emit(ctx, 1, 4); /* 000000ff FP_RESULT_COUNT */ + /* XXX: demagic this part some day */ + if (device->chipset == 0x50) + xf_emit(ctx, 0x3a0, 0); + else if (device->chipset < 0x94) + xf_emit(ctx, 0x3a2, 0); + else if (device->chipset == 0x98 || device->chipset == 0xaa) + xf_emit(ctx, 0x39f, 0); + else + xf_emit(ctx, 0x3a3, 0); + xf_emit(ctx, 1, 0x11); /* 3f/7f DST_FORMAT */ + xf_emit(ctx, 1, 0); /* 7 OPERATION */ + xf_emit(ctx, 1, 1); /* 1 DST_LINEAR */ + xf_emit(ctx, 0x2d, 0); +} + +static void +nv50_gr_construct_xfer2(struct nvkm_grctx *ctx) +{ + struct nvkm_device *device = ctx->device; + int i; + u32 offset; + u32 units = nvkm_rd32(device, 0x1540); + int size = 0; + + offset = (ctx->ctxvals_pos+0x3f)&~0x3f; + + if (device->chipset < 0xa0) { + for (i = 0; i < 8; i++) { + ctx->ctxvals_pos = offset + i; + /* that little bugger belongs to csched. No idea + * what it's doing here. */ + if (i == 0) + xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */ + if (units & (1 << i)) + nv50_gr_construct_xfer_mpc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + } else { + /* Strand 0: TPs 0, 1 */ + ctx->ctxvals_pos = offset; + /* that little bugger belongs to csched. No idea + * what it's doing here. */ + xf_emit(ctx, 1, 0x08100c12); /* FP_INTERPOLANT_CTRL */ + if (units & (1 << 0)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 1)) + nv50_gr_construct_xfer_mpc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 1: TPs 2, 3 */ + ctx->ctxvals_pos = offset + 1; + if (units & (1 << 2)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 3)) + nv50_gr_construct_xfer_mpc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 2: TPs 4, 5, 6 */ + ctx->ctxvals_pos = offset + 2; + if (units & (1 << 4)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 5)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 6)) + nv50_gr_construct_xfer_mpc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + + /* Strand 3: TPs 7, 8, 9 */ + ctx->ctxvals_pos = offset + 3; + if (units & (1 << 7)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 8)) + nv50_gr_construct_xfer_mpc(ctx); + if (units & (1 << 9)) + nv50_gr_construct_xfer_mpc(ctx); + if ((ctx->ctxvals_pos-offset)/8 > size) + size = (ctx->ctxvals_pos-offset)/8; + } + ctx->ctxvals_pos = offset + size * 8; + ctx->ctxvals_pos = (ctx->ctxvals_pos+0x3f)&~0x3f; + cp_lsr (ctx, offset); + cp_out (ctx, CP_SET_XFER_POINTER); + cp_lsr (ctx, size); + cp_out (ctx, CP_SEEK_2); + cp_out (ctx, CP_XFER_2); + cp_wait(ctx, XFER, BUSY); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxtu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxtu102.c new file mode 100644 index 0000000000..542ab0c78b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxtu102.c @@ -0,0 +1,94 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +static void +tu102_grctx_generate_r419c0c(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419c0c, 0x80000000, 0x80000000); + nvkm_mask(device, 0x40584c, 0x00000008, 0x00000000); + nvkm_mask(device, 0x400080, 0x00000000, 0x00000000); +} + +static void +tu102_grctx_generate_sm_id(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + tpc = gv100_gr_nonpes_aware_tpc(gr, gpc, tpc); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x608), sm); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), sm); +} + +static const struct gf100_gr_init +tu102_grctx_init_unknown_bundle_init_0[] = { + { 0x00001000, 1, 0x00000001, 0x00000004 }, + { 0x00002020, 64, 0x00000001, 0x00000000 }, + { 0x0001e100, 1, 0x00000001, 0x00000001 }, + {} +}; + +static const struct gf100_gr_pack +tu102_grctx_pack_sw_bundle64_init[] = { + { tu102_grctx_init_unknown_bundle_init_0, .type = 64 }, + {} +}; + +void +tu102_grctx_generate_unknown(struct gf100_gr_chan *chan, u64 addr, u32 size) +{ + gf100_grctx_patch_wr32(chan, 0x408070, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408074, size >> 8); /*XXX: guess */ + gf100_grctx_patch_wr32(chan, 0x419034, addr >> 8); + gf100_grctx_patch_wr32(chan, 0x408078, 0x00000000); +} + +const struct gf100_grctx_func +tu102_grctx = { + .main = gf100_grctx_generate_main, + .unkn = gv100_grctx_generate_unkn, + .sw_bundle64_init = tu102_grctx_pack_sw_bundle64_init, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x3000, + .bundle_min_gpm_fifo_depth = 0x180, + .bundle_token_limit = 0xa80, + .pagepool = gp100_grctx_generate_pagepool, + .pagepool_size = 0x20000, + .attrib_cb_size = gp102_grctx_generate_attrib_cb_size, + .attrib_cb = gv100_grctx_generate_attrib_cb, + .attrib = gv100_grctx_generate_attrib, + .attrib_nr_max = 0x800, + .attrib_nr = 0x700, + .unknown_size = 0x80000, + .unknown = tu102_grctx_generate_unknown, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, + .gfxp_nr = 0xfa8, + .sm_id = tu102_grctx_generate_sm_id, + .skip_pd_num_tpc_per_gpc = true, + .rop_mapping = gv100_grctx_generate_rop_mapping, + .r406500 = gm200_grctx_generate_r406500, + .r400088 = gv100_grctx_generate_r400088, + .r419c0c = tu102_grctx_generate_r419c0c, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/com.fuc b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/com.fuc new file mode 100644 index 0000000000..64208bf954 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/com.fuc @@ -0,0 +1,335 @@ +/* fuc microcode util functions for gf100 PGRAPH + * + * Copyright 2011 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#ifdef INCLUDE_CODE +// queue_put - add request to queue +// +// In : $r13 queue pointer +// $r14 command +// $r15 data +// +queue_put: + // make sure we have space.. + ld b32 $r8 D[$r13 + 0x0] // GET + ld b32 $r9 D[$r13 + 0x4] // PUT + xor $r8 8 + cmpu b32 $r8 $r9 + bra ne #queue_put_next + mov $r15 E_CMD_OVERFLOW + call(error) + ret + + // store cmd/data on queue + queue_put_next: + and $r8 $r9 7 + shl b32 $r8 3 + add b32 $r8 $r13 + add b32 $r8 8 + st b32 D[$r8 + 0x0] $r14 + st b32 D[$r8 + 0x4] $r15 + + // update PUT + add b32 $r9 1 + and $r9 0xf + st b32 D[$r13 + 0x4] $r9 + ret + +// queue_get - fetch request from queue +// +// In : $r13 queue pointer +// +// Out: $p1 clear on success (data available) +// $r14 command +// $r15 data +// +queue_get: + bset $flags $p1 + ld b32 $r8 D[$r13 + 0x0] // GET + ld b32 $r9 D[$r13 + 0x4] // PUT + cmpu b32 $r8 $r9 + bra e #queue_get_done + // fetch first cmd/data pair + and $r9 $r8 7 + shl b32 $r9 3 + add b32 $r9 $r13 + add b32 $r9 8 + ld b32 $r14 D[$r9 + 0x0] + ld b32 $r15 D[$r9 + 0x4] + + // update GET + add b32 $r8 1 + and $r8 0xf + st b32 D[$r13 + 0x0] $r8 + bclr $flags $p1 +queue_get_done: + ret + +// nv_rd32 - read 32-bit value from nv register +// +// In : $r14 register +// Out: $r15 value +// +nv_rd32: + mov b32 $r12 $r14 + bset $r12 31 // MMIO_CTRL_PENDING + nv_iowr(NV_PGRAPH_FECS_MMIO_CTRL, 0, $r12) + nv_rd32_wait: + nv_iord($r12, NV_PGRAPH_FECS_MMIO_CTRL, 0) + xbit $r12 $r12 31 + bra ne #nv_rd32_wait + mov $r10 6 // DONE_MMIO_RD + call(wait_doneo) + nv_iord($r15, NV_PGRAPH_FECS_MMIO_RDVAL, 0) + ret + +// nv_wr32 - write 32-bit value to nv register +// +// In : $r14 register +// $r15 value +// +nv_wr32: + nv_iowr(NV_PGRAPH_FECS_MMIO_WRVAL, 0, $r15) + mov b32 $r12 $r14 + bset $r12 31 // MMIO_CTRL_PENDING + bset $r12 30 // MMIO_CTRL_WRITE + nv_iowr(NV_PGRAPH_FECS_MMIO_CTRL, 0, $r12) + nv_wr32_wait: + nv_iord($r12, NV_PGRAPH_FECS_MMIO_CTRL, 0) + xbit $r12 $r12 31 + bra ne #nv_wr32_wait + ret + +// wait_donez - wait on FUC_DONE bit to become clear +// +// In : $r10 bit to wait on +// +wait_donez: + trace_set(T_WAIT); + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(6), 0, $r10) + wait_donez_ne: + nv_iord($r8, NV_PGRAPH_FECS_SIGNAL, 0) + xbit $r8 $r8 $r10 + bra ne #wait_donez_ne + trace_clr(T_WAIT) + ret + +// wait_doneo - wait on FUC_DONE bit to become set +// +// In : $r10 bit to wait on +// +wait_doneo: + trace_set(T_WAIT); + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(6), 0, $r10) + wait_doneo_e: + nv_iord($r8, NV_PGRAPH_FECS_SIGNAL, 0) + xbit $r8 $r8 $r10 + bra e #wait_doneo_e + trace_clr(T_WAIT) + ret + +// mmctx_size - determine size of a mmio list transfer +// +// In : $r14 mmio list head +// $r15 mmio list tail +// Out: $r15 transfer size (in bytes) +// +mmctx_size: + clear b32 $r9 + nv_mmctx_size_loop: + ld b32 $r8 D[$r14] + shr b32 $r8 26 + add b32 $r8 1 + shl b32 $r8 2 + add b32 $r9 $r8 + add b32 $r14 4 + cmpu b32 $r14 $r15 + bra ne #nv_mmctx_size_loop + mov b32 $r15 $r9 + ret + +// mmctx_xfer - execute a list of mmio transfers +// +// In : $r10 flags +// bit 0: direction (0 = save, 1 = load) +// bit 1: set if first transfer +// bit 2: set if last transfer +// $r11 base +// $r12 mmio list head +// $r13 mmio list tail +// $r14 multi_stride +// $r15 multi_mask +// +mmctx_xfer: + trace_set(T_MMCTX) + clear b32 $r9 + or $r11 $r11 + bra e #mmctx_base_disabled + nv_iowr(NV_PGRAPH_FECS_MMCTX_BASE, 0, $r11) + bset $r9 0 // BASE_EN + mmctx_base_disabled: + or $r14 $r14 + bra e #mmctx_multi_disabled + nv_iowr(NV_PGRAPH_FECS_MMCTX_MULTI_STRIDE, 0, $r14) + nv_iowr(NV_PGRAPH_FECS_MMCTX_MULTI_MASK, 0, $r15) + bset $r9 1 // MULTI_EN + mmctx_multi_disabled: + + xbit $r11 $r10 0 + shl b32 $r11 16 // DIR + bset $r11 12 // QLIMIT = 0x10 + xbit $r14 $r10 1 + shl b32 $r14 17 + or $r11 $r14 // START_TRIGGER + nv_iowr(NV_PGRAPH_FECS_MMCTX_CTRL, 0, $r11) + + // loop over the mmio list, and send requests to the hw + mmctx_exec_loop: + // wait for space in mmctx queue + mmctx_wait_free: + nv_iord($r14, NV_PGRAPH_FECS_MMCTX_CTRL, 0) + and $r14 0x1f + bra e #mmctx_wait_free + + // queue up an entry + ld b32 $r14 D[$r12] + or $r14 $r9 + nv_iowr(NV_PGRAPH_FECS_MMCTX_QUEUE, 0, $r14) + add b32 $r12 4 + cmpu b32 $r12 $r13 + bra ne #mmctx_exec_loop + + xbit $r11 $r10 2 + bra ne #mmctx_stop + // wait for queue to empty + mmctx_fini_wait: + nv_iord($r11, NV_PGRAPH_FECS_MMCTX_CTRL, 0) + and $r11 0x1f + cmpu b32 $r11 0x10 + bra ne #mmctx_fini_wait + mov $r10 5 // DONE_MMCTX + call(wait_donez) + bra #mmctx_done + mmctx_stop: + xbit $r11 $r10 0 + shl b32 $r11 16 // DIR + bset $r11 12 // QLIMIT = 0x10 + bset $r11 18 // STOP_TRIGGER + nv_iowr(NV_PGRAPH_FECS_MMCTX_CTRL, 0, $r11) + mmctx_stop_wait: + // wait for STOP_TRIGGER to clear + nv_iord($r11, NV_PGRAPH_FECS_MMCTX_CTRL, 0) + xbit $r11 $r11 18 + bra ne #mmctx_stop_wait + mmctx_done: + trace_clr(T_MMCTX) + ret + +// Wait for DONE_STRAND +// +strand_wait: + push $r10 + mov $r10 2 + call(wait_donez) + pop $r10 + ret + +// unknown - call before issuing strand commands +// +strand_pre: + mov $r9 NV_PGRAPH_FECS_STRAND_CMD_ENABLE + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r9) + call(strand_wait) + ret + +// unknown - call after issuing strand commands +// +strand_post: + mov $r9 NV_PGRAPH_FECS_STRAND_CMD_DISABLE + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r9) + call(strand_wait) + ret + +// Selects strand set?! +// +// In: $r14 id +// +strand_set: + mov $r12 0xf + nv_iowr(NV_PGRAPH_FECS_STRAND_FILTER, 0x3f, $r12) + mov $r12 NV_PGRAPH_FECS_STRAND_CMD_DEACTIVATE_FILTER + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r12) + nv_iowr(NV_PGRAPH_FECS_STRAND_FILTER, 0x3f, $r14) + mov $r12 NV_PGRAPH_FECS_STRAND_CMD_ACTIVATE_FILTER + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r12) + call(strand_wait) + ret + +// Initialise strand context data +// +// In : $r15 context base +// Out: $r15 context size (in bytes) +// +// Strandset(?) 3 hardcoded currently +// +strand_ctx_init: + trace_set(T_STRINIT) + call(strand_pre) + mov $r14 3 + call(strand_set) + + clear b32 $r12 + nv_iowr(NV_PGRAPH_FECS_STRAND_SELECT, 0x3f, $r12) + mov $r12 NV_PGRAPH_FECS_STRAND_CMD_SEEK + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r12) + call(strand_wait) + sub b32 $r12 $r0 1 + nv_iowr(NV_PGRAPH_FECS_STRAND_DATA, 0x3f, $r12) + mov $r12 NV_PGRAPH_FECS_STRAND_CMD_GET_INFO + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r12) + call(strand_wait) + call(strand_post) + + // read the size of each strand, poke the context offset of + // each into STRAND_{SAVE,LOAD}_SWBASE now, no need to worry + // about it later then. + nv_mkio($r8, NV_PGRAPH_FECS_STRAND_SAVE_SWBASE, 0x00) + nv_iord($r9, NV_PGRAPH_FECS_STRANDS_CNT, 0x00) + shr b32 $r14 $r15 8 + ctx_init_strand_loop: + iowr I[$r8 + 0x000] $r14 // STRAND_SAVE_SWBASE + iowr I[$r8 + 0x100] $r14 // STRAND_LOAD_SWBASE + iord $r10 I[$r8 + 0x200] // STRAND_SIZE + shr b32 $r10 6 + add b32 $r10 1 + add b32 $r14 $r10 + add b32 $r8 4 + sub b32 $r9 1 + bra ne #ctx_init_strand_loop + + shl b32 $r14 8 + sub b32 $r15 $r14 $r15 + trace_clr(T_STRINIT) + ret +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc new file mode 100644 index 0000000000..4984b0069d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc @@ -0,0 +1,492 @@ +/* fuc microcode for gf100 PGRAPH/GPC + * + * Copyright 2011 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +/* TODO + * - bracket certain functions with scratch writes, useful for debugging + * - watchdog timer around ctx operations + */ + +#ifdef INCLUDE_DATA +gpc_mmio_list_head: .b32 #mmio_list_base +gpc_mmio_list_tail: +tpc_mmio_list_head: .b32 #mmio_list_base +tpc_mmio_list_tail: +unk_mmio_list_head: .b32 #mmio_list_base +unk_mmio_list_tail: .b32 #mmio_list_base + +gpc_id: .b32 0 + +tpc_count: .b32 0 +tpc_mask: .b32 0 + +#if NV_PGRAPH_GPCX_UNK__SIZE > 0 +unk_count: .b32 0 +unk_mask: .b32 0 +#endif + +cmd_queue: queue_init + +mmio_list_base: +#endif + +#ifdef INCLUDE_CODE +#define gpc_addr(reg,addr) /* +*/ imm32(reg,addr) /* +*/ or reg NV_PGRAPH_GPCX_GPCCS_MMIO_CTRL_BASE_ENABLE +#define gpc_wr32(addr,reg) /* +*/ gpc_addr($r14,addr) /* +*/ mov b32 $r15 reg /* +*/ call(nv_wr32) + +// reports an exception to the host +// +// In: $r15 error code (see os.h) +// +error: + push $r14 + nv_wr32(NV_PGRAPH_FECS_CC_SCRATCH_VAL(5), $r15) + mov $r15 1 + nv_wr32(NV_PGRAPH_FECS_INTR_UP_SET, $r15) + pop $r14 + ret + +#if CHIPSET >= GM107 +tpc_strand_wait: + push $r9 + trace_set(T_STRTPC) + tpc_strand_busy: + nv_iord($r9, NV_PGRAPH_GPCX_GPCCS_TPC_STATUS, 0) + bra b32 $r9 0x0 ne #tpc_strand_busy + trace_clr(T_STRTPC) + pop $r9 + ret + +#define tpc_strand_wait() call(tpc_strand_wait) +#define tpc_strand_enable() /* +*/ mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_CMD_ENABLE /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_CMD, $r15) /* +*/ tpc_strand_wait() +#define tpc_strand_disable() /* +*/ mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_CMD_DISABLE /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_CMD, $r15) /* +*/ tpc_strand_wait() +#define tpc_strand_seek(p) /* +*/ mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_INDEX_ALL /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_INDEX, $r15) /* +*/ mov $r15 p /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_SELECT, $r15) /* +*/ mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_CMD_SEEK /* +*/ tpc_strand_wait() +#define tpc_strand_info(m) /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_CMD, $r15) /* +*/ mov $r15 m /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_DATA, $r15) /* +*/ mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_CMD_GET_INFO /* +*/ gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_CMD, $r15) /* +*/ tpc_strand_wait() +#endif + + +// GPC fuc initialisation, executed by triggering ucode start, will +// fall through to main loop after completion. +// +// Input: +// CC_SCRATCH[1]: context base +// +// Output: +// CC_SCRATCH[0]: +// 31:31: set to signal completion +// CC_SCRATCH[1]: +// 31:0: GPC context size +// +init: + clear b32 $r0 + + // setup stack + nv_iord($r1, NV_PGRAPH_GPCX_GPCCS_CAPS, 0) + extr $r1 $r1 9:17 + shl b32 $r1 8 + mov $sp $r1 + + // enable fifo access + mov $r2 NV_PGRAPH_GPCX_GPCCS_ACCESS_FIFO + nv_iowr(NV_PGRAPH_GPCX_GPCCS_ACCESS, 0, $r2) + + // setup i0 handler, and route all interrupts to it + mov $r1 #ih + mov $iv0 $r1 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_INTR_ROUTE, 0, $r0) + + // enable fifo interrupt + mov $r2 NV_PGRAPH_GPCX_GPCCS_INTR_EN_SET_FIFO + nv_iowr(NV_PGRAPH_GPCX_GPCCS_INTR_EN_SET, 0, $r2) + + // enable interrupts + bset $flags ie0 + + // how many TPCs do we have? + nv_iord($r2, NV_PGRAPH_GPCX_GPCCS_UNITS, 0) + mov $r3 1 + and $r2 0x1f + shl b32 $r3 $r2 + sub b32 $r3 1 + st b32 D[$r0 + #tpc_count] $r2 + st b32 D[$r0 + #tpc_mask] $r3 + + // determine which GPC we are, setup (optional) mmio access offset + nv_iord($r2, NV_PGRAPH_GPCX_GPCCS_MYINDEX, 0) + st b32 D[$r0 + #gpc_id] $r2 + shl b32 $r2 15 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_MMIO_BASE, 0, $r2) + +#if NV_PGRAPH_GPCX_UNK__SIZE > 0 + // figure out which, and how many, UNKs are actually present + gpc_addr($r14, 0x500c30) + clear b32 $r2 + clear b32 $r3 + clear b32 $r4 + init_unk_loop: + call(nv_rd32) + cmp b32 $r15 0 + bra z #init_unk_next + mov $r15 1 + shl b32 $r15 $r2 + or $r4 $r15 + add b32 $r3 1 + init_unk_next: + add b32 $r2 1 + add b32 $r14 4 + cmp b32 $r2 NV_PGRAPH_GPCX_UNK__SIZE + bra ne #init_unk_loop + init_unk_done: + st b32 D[$r0 + #unk_count] $r3 + st b32 D[$r0 + #unk_mask] $r4 +#endif + + // initialise context base, and size tracking + nv_iord($r2, NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_VAL(1), 0) + clear b32 $r3 // track GPC context size here + + // set mmctx base addresses now so we don't have to do it later, + // they don't currently ever change + shr b32 $r5 $r2 8 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_MMCTX_SAVE_SWBASE, 0, $r5) + nv_iowr(NV_PGRAPH_GPCX_GPCCS_MMCTX_LOAD_SWBASE, 0, $r5) + + // calculate GPC mmio context size + ld b32 $r14 D[$r0 + #gpc_mmio_list_head] + ld b32 $r15 D[$r0 + #gpc_mmio_list_tail] + call(mmctx_size) + add b32 $r2 $r15 + add b32 $r3 $r15 + + // calculate per-TPC mmio context size + ld b32 $r14 D[$r0 + #tpc_mmio_list_head] + ld b32 $r15 D[$r0 + #tpc_mmio_list_tail] + call(mmctx_size) + ld b32 $r14 D[$r0 + #tpc_count] + mulu $r14 $r15 + add b32 $r2 $r14 + add b32 $r3 $r14 + +#if NV_PGRAPH_GPCX_UNK__SIZE > 0 + // calculate per-UNK mmio context size + ld b32 $r14 D[$r0 + #unk_mmio_list_head] + ld b32 $r15 D[$r0 + #unk_mmio_list_tail] + call(mmctx_size) + ld b32 $r14 D[$r0 + #unk_count] + mulu $r14 $r15 + add b32 $r2 $r14 + add b32 $r3 $r14 +#endif + + // round up base/size to 256 byte boundary (for strand SWBASE) + shr b32 $r3 2 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_MMCTX_LOAD_COUNT, 0, $r3) // wtf for?! + shr b32 $r2 8 + shr b32 $r3 6 + add b32 $r2 1 + add b32 $r3 1 + shl b32 $r2 8 + shl b32 $r3 8 + + // calculate size of strand context data + mov b32 $r15 $r2 + call(strand_ctx_init) + add b32 $r2 $r15 + add b32 $r3 $r15 + +#if CHIPSET >= GM107 + // calculate size of tpc strand context data + mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_INDEX_ALL + gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_INDEX, $r15) + tpc_strand_enable(); + tpc_strand_seek(0); + tpc_strand_info(-1); + + ld b32 $r4 D[$r0 + #tpc_count] + gpc_addr($r5, NV_PGRAPH_GPC0_TPC0) + tpc_strand_init_tpc_loop: + add b32 $r14 $r5 NV_TPC_STRAND_CNT + call(nv_rd32) + mov b32 $r6 $r15 + clear b32 $r7 + tpc_strand_init_idx_loop: + add b32 $r14 $r5 NV_TPC_STRAND_INDEX + mov b32 $r15 $r7 + call(nv_wr32) + add b32 $r14 $r5 NV_TPC_STRAND_SAVE_SWBASE + shr b32 $r15 $r2 8 + call(nv_wr32) + add b32 $r14 $r5 NV_TPC_STRAND_LOAD_SWBASE + shr b32 $r15 $r2 8 + call(nv_wr32) + add b32 $r14 $r5 NV_TPC_STRAND_WORDS + call(nv_rd32) + shr b32 $r15 6 + add b32 $r15 1 + shl b32 $r15 8 + add b32 $r2 $r15 + add b32 $r3 $r15 + add b32 $r7 1 + sub b32 $r6 1 + bra nz #tpc_strand_init_idx_loop + add b32 $r5 NV_PGRAPH_GPC0_TPC0__SIZE + sub b32 $r4 1 + bra nz #tpc_strand_init_tpc_loop + + mov $r15 NV_PGRAPH_GPC0_TPCX_STRAND_INDEX_ALL + gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_INDEX, $r15) + tpc_strand_disable(); +#endif + + // save context size, and tell HUB we're done + nv_iowr(NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_VAL(1), 0, $r3) + clear b32 $r2 + bset $r2 31 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_SET(0), 0, $r2) + +// Main program loop, very simple, sleeps until woken up by the interrupt +// handler, pulls a command from the queue and executes its handler +// +wait: + sleep $p0 + bset $flags $p0 +main: + mov $r13 #cmd_queue + call(queue_get) + bra $p1 #wait + + // 0x0000-0x0003 are all context transfers + cmpu b32 $r14 0x04 + bra nc #main_not_ctx_xfer + // fetch $flags and mask off $p1/$p2 + mov $r1 $flags + mov $r2 0x0006 + not b32 $r2 + and $r1 $r2 + // set $p1/$p2 according to transfer type + shl b32 $r14 1 + or $r1 $r14 + mov $flags $r1 + // transfer context data + call(ctx_xfer) + bra #main + + main_not_ctx_xfer: + shl b32 $r15 $r14 16 + or $r15 E_BAD_COMMAND + call(error) + bra #main + +// interrupt handler +ih: + push $r0 + push $r8 + mov $r8 $flags + push $r8 + push $r9 + push $r10 + push $r11 + push $r13 + push $r14 + push $r15 + clear b32 $r0 + + // incoming fifo command? + nv_iord($r10, NV_PGRAPH_GPCX_GPCCS_INTR, 0) + and $r11 $r10 NV_PGRAPH_GPCX_GPCCS_INTR_FIFO + bra e #ih_no_fifo + // queue incoming fifo command for later processing + mov $r13 #cmd_queue + nv_iord($r14, NV_PGRAPH_GPCX_GPCCS_FIFO_CMD, 0) + nv_iord($r15, NV_PGRAPH_GPCX_GPCCS_FIFO_DATA, 0) + call(queue_put) + mov $r14 1 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_FIFO_ACK, 0, $r14) + + // ack, and wake up main() + ih_no_fifo: + nv_iowr(NV_PGRAPH_GPCX_GPCCS_INTR_ACK, 0, $r10) + + pop $r15 + pop $r14 + pop $r13 + pop $r11 + pop $r10 + pop $r9 + pop $r8 + mov $flags $r8 + pop $r8 + pop $r0 + bclr $flags $p0 + iret + +// Set this GPC's bit in HUB_BAR, used to signal completion of various +// activities to the HUB fuc +// +hub_barrier_done: + mov $r15 1 + ld b32 $r14 D[$r0 + #gpc_id] + shl b32 $r15 $r14 + nv_wr32(0x409418, $r15) // 0x409418 - HUB_BAR_SET + ret + +// Disables various things, waits a bit, and re-enables them.. +// +// Not sure how exactly this helps, perhaps "ENABLE" is not such a +// good description for the bits we turn off? Anyways, without this, +// funny things happen. +// +ctx_redswitch: + mov $r15 NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_POWER + nv_iowr(NV_PGRAPH_GPCX_GPCCS_RED_SWITCH, 0, $r15) + mov $r14 8 + ctx_redswitch_delay: + sub b32 $r14 1 + bra ne #ctx_redswitch_delay + or $r15 NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_UNK11 + or $r15 NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_ENABLE + nv_iowr(NV_PGRAPH_GPCX_GPCCS_RED_SWITCH, 0, $r15) + ret + +// Transfer GPC context data between GPU and storage area +// +// In: $r15 context base address +// $p1 clear on save, set on load +// $p2 set if opposite direction done/will be done, so: +// on save it means: "a load will follow this save" +// on load it means: "a save preceeded this load" +// +ctx_xfer: + // set context base address + nv_iowr(NV_PGRAPH_GPCX_GPCCS_MEM_BASE, 0, $r15) +#if CHIPSET >= GM107 + gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_MEM_BASE, $r15) +#endif + bra not $p1 #ctx_xfer_not_load + call(ctx_redswitch) + ctx_xfer_not_load: + + // strands + call(strand_pre) + clear b32 $r2 + nv_iowr(NV_PGRAPH_GPCX_GPCCS_STRAND_SELECT, 0x3f, $r2) + xbit $r2 $flags $p1 // SAVE/LOAD + add b32 $r2 NV_PGRAPH_GPCX_GPCCS_STRAND_CMD_SAVE + nv_iowr(NV_PGRAPH_GPCX_GPCCS_STRAND_CMD, 0x3f, $r2) + +#if CHIPSET >= GM107 + tpc_strand_enable(); + tpc_strand_seek(0); + xbit $r15 $flags $p1 // SAVE/LOAD + add b32 $r15 NV_PGRAPH_GPC0_TPCX_STRAND_CMD_SAVE + gpc_wr32(NV_PGRAPH_GPC0_TPCX_STRAND_CMD, $r15) +#endif + + // mmio context + xbit $r10 $flags $p1 // direction + or $r10 2 // first + imm32($r11,0x500000) + ld b32 $r12 D[$r0 + #gpc_id] + shl b32 $r12 15 + add b32 $r11 $r12 // base = NV_PGRAPH_GPCn + ld b32 $r12 D[$r0 + #gpc_mmio_list_head] + ld b32 $r13 D[$r0 + #gpc_mmio_list_tail] + mov $r14 0 // not multi + call(mmctx_xfer) + + // per-TPC mmio context + xbit $r10 $flags $p1 // direction +#if !NV_PGRAPH_GPCX_UNK__SIZE + or $r10 4 // last +#endif + imm32($r11, 0x504000) + ld b32 $r12 D[$r0 + #gpc_id] + shl b32 $r12 15 + add b32 $r11 $r12 // base = NV_PGRAPH_GPCn_TPC0 + ld b32 $r12 D[$r0 + #tpc_mmio_list_head] + ld b32 $r13 D[$r0 + #tpc_mmio_list_tail] + ld b32 $r15 D[$r0 + #tpc_mask] + mov $r14 0x800 // stride = 0x800 + call(mmctx_xfer) + +#if NV_PGRAPH_GPCX_UNK__SIZE > 0 + // per-UNK mmio context + xbit $r10 $flags $p1 // direction + or $r10 4 // last + imm32($r11, 0x503000) + ld b32 $r12 D[$r0 + #gpc_id] + shl b32 $r12 15 + add b32 $r11 $r12 // base = NV_PGRAPH_GPCn_UNK0 + ld b32 $r12 D[$r0 + #unk_mmio_list_head] + ld b32 $r13 D[$r0 + #unk_mmio_list_tail] + ld b32 $r15 D[$r0 + #unk_mask] + mov $r14 0x200 // stride = 0x200 + call(mmctx_xfer) +#endif + + // wait for strands to finish + call(strand_wait) +#if CHIPSET >= GM107 + tpc_strand_wait() +#endif + + // if load, or a save without a load following, do some + // unknown stuff that's done after finishing a block of + // strand commands + bra $p1 #ctx_xfer_post + bra not $p2 #ctx_xfer_done + ctx_xfer_post: + call(strand_post) +#if CHIPSET >= GM107 + tpc_strand_disable() +#endif + + // mark completion in HUB's barrier + ctx_xfer_done: + call(hub_barrier_done) + ret +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3 new file mode 100644 index 0000000000..7cf2bf9d95 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000000 + +#define CHIPSET GF100 +#include "macros.fuc" + +.section #gf100_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gf100_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h new file mode 100644 index 0000000000..54e14b4d31 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf100.fuc3.h @@ -0,0 +1,532 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gf100_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x00000064, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x00000064, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x00000064, +/* 0x000c: unk_mmio_list_tail */ + 0x00000064, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gf100_grgpc_code[] = { + 0x03a10ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0xe0f900f8, + 0xf102ffb9, + 0xf09814e7, + 0x21f440e3, + 0x01f7f09d, + 0xf102ffb9, + 0xf09c1ce7, + 0x21f440e3, + 0xf8e0fc9d, +/* 0x03a1: init */ + 0xf104bd00, + 0xf0420017, + 0x11cf0013, + 0x0911e700, + 0x0814b601, + 0xf00014fe, + 0x07f10227, + 0x03f01200, + 0x0002d000, + 0x17f104bd, + 0x10fe04f8, + 0x0007f100, + 0x0003f007, + 0xbd0000d0, + 0x0427f004, + 0x040007f1, + 0xd00003f0, + 0x04bd0002, + 0xf11031f4, + 0xf0820027, + 0x22cf0123, + 0x0137f000, + 0xbb1f24f0, + 0x32b60432, + 0x05028001, + 0xf1060380, + 0xf0860027, + 0x22cf0123, + 0x04028000, + 0xf10f24b6, + 0xf0c90007, + 0x02d00103, + 0xf104bd00, + 0xf0010027, + 0x22cf0223, + 0x9534bd00, + 0x07f10825, + 0x03f0c000, + 0x0005d001, + 0x07f104bd, + 0x03f0c100, + 0x0005d001, + 0x0e9804bd, + 0x010f9800, + 0x015021f5, + 0xbb002fbb, + 0x0e98003f, + 0x020f9801, + 0x015021f5, + 0xfd050e98, + 0x2ebb00ef, + 0x003ebb00, + 0xf10235b6, + 0xf0d30007, + 0x03d00103, + 0xb604bd00, + 0x35b60825, + 0x0120b606, + 0xb60130b6, + 0x34b60824, + 0x022fb908, + 0x02d321f5, + 0xbb002fbb, + 0x07f1003f, + 0x03f00100, + 0x0003d002, + 0x24bd04bd, + 0xf11f29f0, + 0xf0080007, + 0x02d00203, +/* 0x04bb: wait */ + 0xf404bd00, + 0x31f40028, +/* 0x04c1: main */ + 0x1cd7f000, + 0xf43921f4, + 0xe4b0f401, + 0x1e18f404, + 0xf00181fe, + 0x20bd0627, + 0xb60412fd, + 0x1efd01e4, + 0x0018fe05, + 0x05b421f5, +/* 0x04eb: main_not_ctx_xfer */ + 0x94d90ef4, + 0xf5f010ef, + 0x7e21f501, + 0xcc0ef403, +/* 0x04f8: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0xf104bdf0, + 0xf00200a7, + 0xaacf00a3, + 0x04abc400, + 0xf02c0bf4, + 0xe7f11cd7, + 0xe3f01a00, + 0x00eecf00, + 0x1900f7f1, + 0xcf00f3f0, + 0x21f400ff, + 0x01e7f004, + 0x1d0007f1, + 0xd00003f0, + 0x04bd000e, +/* 0x0548: ih_no_fifo */ + 0x010007f1, + 0xd00003f0, + 0x04bd000a, + 0xe0fcf0fc, + 0xb0fcd0fc, + 0x90fca0fc, + 0x88fe80fc, + 0xfc80fc00, + 0x0032f400, +/* 0x056e: hub_barrier_done */ + 0xf7f001f8, + 0x040e9801, + 0xb904febb, + 0xe7f102ff, + 0xe3f09418, + 0x9d21f440, +/* 0x0586: ctx_redswitch */ + 0xf7f000f8, + 0x0007f120, + 0x0103f085, + 0xbd000fd0, + 0x08e7f004, +/* 0x0598: ctx_redswitch_delay */ + 0xf401e2b6, + 0xf5f1fd1b, + 0xf5f10800, + 0x07f10200, + 0x03f08500, + 0x000fd001, + 0x00f804bd, +/* 0x05b4: ctx_xfer */ + 0x810007f1, + 0xd00203f0, + 0x04bd000f, + 0xf50711f4, +/* 0x05c7: ctx_xfer_not_load */ + 0xf5058621, + 0xbd026a21, + 0xfc07f124, + 0x0203f047, + 0xbd0002d0, + 0x012cf004, + 0xf10320b6, + 0xf04afc07, + 0x02d00203, + 0xf004bd00, + 0xa5f001ac, + 0x00b7f102, + 0x50b3f000, + 0xb6040c98, + 0xbcbb0fc4, + 0x000c9800, + 0xf0010d98, + 0x21f500e7, + 0xacf0016f, + 0x04a5f001, + 0x4000b7f1, + 0x9850b3f0, + 0xc4b6040c, + 0x00bcbb0f, + 0x98010c98, + 0x0f98020d, + 0x00e7f106, + 0x6f21f508, + 0x5e21f501, + 0x0601f402, +/* 0x063f: ctx_xfer_post */ + 0xf50712f4, +/* 0x0643: ctx_xfer_done */ + 0xf5027f21, + 0xf8056e21, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3 new file mode 100644 index 0000000000..c918f7d600 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000001 + +#define CHIPSET GF117 +#include "macros.fuc" + +.section #gf117_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gf117_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h new file mode 100644 index 0000000000..67524e615d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h @@ -0,0 +1,539 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gf117_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x0000006c, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x0000006c, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x0000006c, +/* 0x000c: unk_mmio_list_tail */ + 0x0000006c, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: unk_count */ + 0x00000000, +/* 0x0020: unk_mask */ + 0x00000000, +/* 0x0024: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gf117_grgpc_code[] = { + 0x03a10ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0xe0f900f8, + 0xf102ffb9, + 0xf09814e7, + 0x21f440e3, + 0x01f7f09d, + 0xf102ffb9, + 0xf09c1ce7, + 0x21f440e3, + 0xf8e0fc9d, +/* 0x03a1: init */ + 0xf104bd00, + 0xf0420017, + 0x11cf0013, + 0x0911e700, + 0x0814b601, + 0xf00014fe, + 0x07f10227, + 0x03f01200, + 0x0002d000, + 0x17f104bd, + 0x10fe0545, + 0x0007f100, + 0x0003f007, + 0xbd0000d0, + 0x0427f004, + 0x040007f1, + 0xd00003f0, + 0x04bd0002, + 0xf11031f4, + 0xf0820027, + 0x22cf0123, + 0x0137f000, + 0xbb1f24f0, + 0x32b60432, + 0x05028001, + 0xf1060380, + 0xf0860027, + 0x22cf0123, + 0x04028000, + 0xf10f24b6, + 0xf0c90007, + 0x02d00103, + 0xf104bd00, + 0xf00c30e7, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf401, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, + 0x04bd0003, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: wait */ + 0xf40028f4, +/* 0x050e: main */ + 0xd7f00031, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0x0121f500, + 0xd90ef406, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xf900f9cc, + 0x0188fe80, + 0x90f980f9, + 0xb0f9a0f9, + 0xe0f9d0f9, + 0x04bdf0f9, + 0x0200a7f1, + 0xcf00a3f0, + 0xabc400aa, + 0x2c0bf404, + 0xf124d7f0, + 0xf01a00e7, + 0xeecf00e3, + 0x00f7f100, + 0x00f3f019, + 0xf400ffcf, + 0xe7f00421, + 0x0007f101, + 0x0003f01d, + 0xbd000ed0, +/* 0x0595: ih_no_fifo */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x05bb: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05d3: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e5: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, + 0xf0850007, + 0x0fd00103, + 0xf804bd00, +/* 0x0601: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05d321f5, +/* 0x0614: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, + 0xf001acf0, + 0xb7f102a5, + 0xb3f00000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06b0: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b4: ctx_xfer_done */ + 0x05bb21f5, + 0x000000f8, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3 new file mode 100644 index 0000000000..b80cdfd337 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000001 + +#define CHIPSET GK100 +#include "macros.fuc" + +.section #gk104_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gk104_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h new file mode 100644 index 0000000000..60c8b7e899 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h @@ -0,0 +1,539 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk104_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x0000006c, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x0000006c, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x0000006c, +/* 0x000c: unk_mmio_list_tail */ + 0x0000006c, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: unk_count */ + 0x00000000, +/* 0x0020: unk_mask */ + 0x00000000, +/* 0x0024: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gk104_grgpc_code[] = { + 0x03a10ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0xe0f900f8, + 0xf102ffb9, + 0xf09814e7, + 0x21f440e3, + 0x01f7f09d, + 0xf102ffb9, + 0xf09c1ce7, + 0x21f440e3, + 0xf8e0fc9d, +/* 0x03a1: init */ + 0xf104bd00, + 0xf0420017, + 0x11cf0013, + 0x0911e700, + 0x0814b601, + 0xf00014fe, + 0x07f10227, + 0x03f01200, + 0x0002d000, + 0x17f104bd, + 0x10fe0545, + 0x0007f100, + 0x0003f007, + 0xbd0000d0, + 0x0427f004, + 0x040007f1, + 0xd00003f0, + 0x04bd0002, + 0xf11031f4, + 0xf0820027, + 0x22cf0123, + 0x0137f000, + 0xbb1f24f0, + 0x32b60432, + 0x05028001, + 0xf1060380, + 0xf0860027, + 0x22cf0123, + 0x04028000, + 0xf10f24b6, + 0xf0c90007, + 0x02d00103, + 0xf104bd00, + 0xf00c30e7, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf401, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, + 0x04bd0003, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: wait */ + 0xf40028f4, +/* 0x050e: main */ + 0xd7f00031, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0x0121f500, + 0xd90ef406, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xf900f9cc, + 0x0188fe80, + 0x90f980f9, + 0xb0f9a0f9, + 0xe0f9d0f9, + 0x04bdf0f9, + 0x0200a7f1, + 0xcf00a3f0, + 0xabc400aa, + 0x2c0bf404, + 0xf124d7f0, + 0xf01a00e7, + 0xeecf00e3, + 0x00f7f100, + 0x00f3f019, + 0xf400ffcf, + 0xe7f00421, + 0x0007f101, + 0x0003f01d, + 0xbd000ed0, +/* 0x0595: ih_no_fifo */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x05bb: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05d3: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e5: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, + 0xf0850007, + 0x0fd00103, + 0xf804bd00, +/* 0x0601: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05d321f5, +/* 0x0614: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, + 0xf001acf0, + 0xb7f102a5, + 0xb3f00000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06b0: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b4: ctx_xfer_done */ + 0x05bb21f5, + 0x000000f8, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3 new file mode 100644 index 0000000000..98d85fe210 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000002 + +#define CHIPSET GK110 +#include "macros.fuc" + +.section #gk110_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gk110_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h new file mode 100644 index 0000000000..c99d156655 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h @@ -0,0 +1,539 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk110_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x0000006c, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x0000006c, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x0000006c, +/* 0x000c: unk_mmio_list_tail */ + 0x0000006c, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: unk_count */ + 0x00000000, +/* 0x0020: unk_mask */ + 0x00000000, +/* 0x0024: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gk110_grgpc_code[] = { + 0x03a10ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f037, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f037, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x370007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x370007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0xe0f900f8, + 0xf102ffb9, + 0xf09814e7, + 0x21f440e3, + 0x01f7f09d, + 0xf102ffb9, + 0xf09c1ce7, + 0x21f440e3, + 0xf8e0fc9d, +/* 0x03a1: init */ + 0xf104bd00, + 0xf0420017, + 0x11cf0013, + 0x0911e700, + 0x0814b601, + 0xf00014fe, + 0x07f10227, + 0x03f01200, + 0x0002d000, + 0x17f104bd, + 0x10fe0545, + 0x0007f100, + 0x0003f007, + 0xbd0000d0, + 0x0427f004, + 0x040007f1, + 0xd00003f0, + 0x04bd0002, + 0xf11031f4, + 0xf0820027, + 0x22cf0123, + 0x0137f000, + 0xbb1f24f0, + 0x32b60432, + 0x05028001, + 0xf1060380, + 0xf0860027, + 0x22cf0123, + 0x04028000, + 0xf10f24b6, + 0xf0c90007, + 0x02d00103, + 0xf104bd00, + 0xf00c30e7, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf402, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, + 0x04bd0003, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x300007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: wait */ + 0xf40028f4, +/* 0x050e: main */ + 0xd7f00031, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0x0121f500, + 0xd90ef406, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xf900f9cc, + 0x0188fe80, + 0x90f980f9, + 0xb0f9a0f9, + 0xe0f9d0f9, + 0x04bdf0f9, + 0x0200a7f1, + 0xcf00a3f0, + 0xabc400aa, + 0x2c0bf404, + 0xf124d7f0, + 0xf01a00e7, + 0xeecf00e3, + 0x00f7f100, + 0x00f3f019, + 0xf400ffcf, + 0xe7f00421, + 0x0007f101, + 0x0003f01d, + 0xbd000ed0, +/* 0x0595: ih_no_fifo */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x05bb: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05d3: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e5: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, + 0xf0850007, + 0x0fd00103, + 0xf804bd00, +/* 0x0601: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05d321f5, +/* 0x0614: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, + 0xf001acf0, + 0xb7f102a5, + 0xb3f00000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06b0: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b4: ctx_xfer_done */ + 0x05bb21f5, + 0x000000f8, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5 new file mode 100644 index 0000000000..8f64299a3b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000001 + +#define CHIPSET GK208 +#include "macros.fuc" + +.section #gk208_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gk208_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h new file mode 100644 index 0000000000..753aa66729 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h @@ -0,0 +1,475 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk208_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x0000006c, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x0000006c, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x0000006c, +/* 0x000c: unk_mmio_list_tail */ + 0x0000006c, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: unk_count */ + 0x00000000, +/* 0x0020: unk_mask */ + 0x00000000, +/* 0x0024: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gk208_grgpc_code[] = { + 0x03140ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0xf489a408, + 0x020f0b1b, + 0x0002f87e, +/* 0x001a: queue_put_next */ + 0x98c400f8, + 0x0384b607, + 0xb6008dbb, + 0x8eb50880, + 0x018fb500, + 0xf00190b6, + 0xd9b50f94, +/* 0x0037: queue_get */ + 0xf400f801, + 0xd8980131, + 0x01d99800, + 0x0bf489a4, + 0x0789c421, + 0xbb0394b6, + 0x90b6009d, + 0x009e9808, + 0xb6019f98, + 0x84f00180, + 0x00d8b50f, +/* 0x0063: queue_get_done */ + 0xf80132f4, +/* 0x0065: nv_rd32 */ + 0xf0ecb200, + 0x00801fc9, + 0x0cf601ca, +/* 0x0073: nv_rd32_wait */ + 0x8c04bd00, + 0xcf01ca00, + 0xccc800cc, + 0xf61bf41f, + 0xec7e060a, + 0x008f0000, + 0xffcf01cb, +/* 0x008f: nv_wr32 */ + 0x8000f800, + 0xf601cc00, + 0x04bd000f, + 0xc9f0ecb2, + 0x1ec9f01f, + 0x01ca0080, + 0xbd000cf6, +/* 0x00a9: nv_wr32_wait */ + 0xca008c04, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f61b, +/* 0x00b8: wait_donez */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x00cf: wait_donez_ne */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf61bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x00ec: wait_doneo */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x0103: wait_doneo_e */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf60bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x0120: mmctx_size */ +/* 0x0122: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0x1bf4efa4, + 0xf89fb2ec, +/* 0x013d: mmctx_xfer */ + 0xf094bd00, + 0x00800199, + 0x09f60237, + 0xbd04bd00, + 0x05bbfd94, + 0x800f0bf4, + 0xf601c400, + 0x04bd000b, +/* 0x015f: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0xc6008018, + 0x000ef601, + 0x008004bd, + 0x0ff601c7, + 0xf004bd00, +/* 0x017a: mmctx_multi_disabled */ + 0xabc80199, + 0x10b4b600, + 0xc80cb9f0, + 0xe4b601ae, + 0x05befd11, + 0x01c50080, + 0xbd000bf6, +/* 0x0195: mmctx_exec_loop */ +/* 0x0195: mmctx_wait_free */ + 0xc5008e04, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f60b, + 0x05e9fd00, + 0x01c80080, + 0xbd000ef6, + 0x04c0b604, + 0x1bf4cda4, + 0x02abc8df, +/* 0x01bf: mmctx_fini_wait */ + 0x8b1c1bf4, + 0xcf01c500, + 0xb4f000bb, + 0x10b4b01f, + 0x0af31bf4, + 0x00b87e05, + 0x250ef400, +/* 0x01d8: mmctx_stop */ + 0xb600abc8, + 0xb9f010b4, + 0x12b9f00c, + 0x01c50080, + 0xbd000bf6, +/* 0x01ed: mmctx_stop_wait */ + 0xc5008b04, + 0x00bbcf01, + 0xf412bbc8, +/* 0x01fa: mmctx_done */ + 0x94bdf61b, + 0x800199f0, + 0xf6021700, + 0x04bd0009, +/* 0x020a: strand_wait */ + 0xa0f900f8, + 0xb87e020a, + 0xa0fc0000, +/* 0x0216: strand_pre */ + 0x0c0900f8, + 0x024afc80, + 0xbd0009f6, + 0x020a7e04, +/* 0x0227: strand_post */ + 0x0900f800, + 0x4afc800d, + 0x0009f602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0238: strand_set */ + 0xfc800f0c, + 0x0cf6024f, + 0x0c04bd00, + 0x4afc800b, + 0x000cf602, + 0xfc8004bd, + 0x0ef6024f, + 0x0c04bd00, + 0x4afc800a, + 0x000cf602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0268: strand_ctx_init */ + 0x99f094bd, + 0x37008003, + 0x0009f602, + 0x167e04bd, + 0x030e0002, + 0x0002387e, + 0xfc80c4bd, + 0x0cf60247, + 0x0c04bd00, + 0x4afc8001, + 0x000cf602, + 0x0a7e04bd, + 0x0c920002, + 0x46fc8001, + 0x000cf602, + 0x020c04bd, + 0x024afc80, + 0xbd000cf6, + 0x020a7e04, + 0x02277e00, + 0x42008800, + 0x20008902, + 0x0099cf02, +/* 0x02c7: ctx_init_strand_loop */ + 0xf608fe95, + 0x8ef6008e, + 0x808acf40, + 0xb606a5b6, + 0xeabb01a0, + 0x0480b600, + 0xf40192b6, + 0xe4b6e81b, + 0xf2efbc08, + 0x99f094bd, + 0x17008003, + 0x0009f602, + 0x00f804bd, +/* 0x02f8: error */ + 0xffb2e0f9, + 0x4098148e, + 0x00008f7e, + 0xffb2010f, + 0x409c1c8e, + 0x00008f7e, + 0x00f8e0fc, +/* 0x0314: init */ + 0x004104bd, + 0x0011cf42, + 0x010911e7, + 0xfe0814b6, + 0x02020014, + 0xf6120040, + 0x04bd0002, + 0xfe048441, + 0x00400010, + 0x0000f607, + 0x040204bd, + 0xf6040040, + 0x04bd0002, + 0x821031f4, + 0xcf018200, + 0x01030022, + 0xbb1f24f0, + 0x32b60432, + 0x0502b501, + 0x820603b5, + 0xcf018600, + 0x02b50022, + 0x0f24b604, + 0x01c90080, + 0xbd0002f6, + 0x0c308e04, + 0x01e5f050, + 0x34bd24bd, +/* 0x0386: init_unk_loop */ + 0x657e44bd, + 0xf6b00000, + 0x0e0bf400, + 0xf2bb010f, + 0x054ffd04, +/* 0x039b: init_unk_next */ + 0xb60130b6, + 0xe0b60120, + 0x0126b004, +/* 0x03a7: init_unk_done */ + 0xb5e21bf4, + 0x04b50703, + 0x01008208, + 0x0022cf02, + 0x259534bd, + 0xc0008008, + 0x0005f601, + 0x008004bd, + 0x05f601c1, + 0x9804bd00, + 0x0f98000e, + 0x01207e01, + 0x002fbb00, + 0x98003fbb, + 0x0f98010e, + 0x01207e02, + 0x050e9800, + 0xbb00effd, + 0x3ebb002e, + 0x020e9800, + 0x7e030f98, + 0x98000120, + 0xeffd070e, + 0x002ebb00, + 0xb6003ebb, + 0x00800235, + 0x03f601d3, + 0xb604bd00, + 0x35b60825, + 0x0120b606, + 0xb60130b6, + 0x34b60824, + 0x7e2fb208, + 0xbb000268, + 0x3fbb002f, + 0x01008000, + 0x0003f602, + 0x24bd04bd, + 0x801f29f0, + 0xf6023000, + 0x04bd0002, +/* 0x0448: wait */ + 0xf40028f4, +/* 0x044e: main */ + 0x240d0031, + 0x0000377e, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1d, + 0x20bd0602, + 0xb60412fd, + 0x1efd01e4, + 0x0018fe05, + 0x00051f7e, +/* 0x0477: main_not_ctx_xfer */ + 0x94da0ef4, + 0xf5f010ef, + 0x02f87e01, + 0xcd0ef400, +/* 0x0484: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0x4a04bdf0, + 0xaacf0200, + 0x04abc400, + 0x0d1f0bf4, + 0x1a004e24, + 0x4f00eecf, + 0xffcf1900, + 0x00047e00, + 0x40010e00, + 0x0ef61d00, +/* 0x04c3: ih_no_fifo */ + 0x4004bd00, + 0x0af60100, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x00fc80fc, + 0xf80032f4, +/* 0x04e5: hub_barrier_done */ + 0x98010f01, + 0xfebb040e, + 0x8effb204, + 0x7e409418, + 0xf800008f, +/* 0x04f9: ctx_redswitch */ + 0x80200f00, + 0xf6018500, + 0x04bd000f, +/* 0x0506: ctx_redswitch_delay */ + 0xe2b6080e, + 0xfd1bf401, + 0x0800f5f1, + 0x0200f5f1, + 0x01850080, + 0xbd000ff6, +/* 0x051f: ctx_xfer */ + 0x8000f804, + 0xf6028100, + 0x04bd000f, + 0x7e0711f4, +/* 0x052f: ctx_xfer_not_load */ + 0x7e0004f9, + 0xbd000216, + 0x47fc8024, + 0x0002f602, + 0x2cf004bd, + 0x0320b601, + 0x024afc80, + 0xbd0002f6, + 0x01acf004, + 0x8b02a5f0, + 0x98500000, + 0xc4b6040c, + 0x00bcbb0f, + 0x98000c98, + 0x000e010d, + 0x00013d7e, + 0x8b01acf0, + 0x98504000, + 0xc4b6040c, + 0x00bcbb0f, + 0x98010c98, + 0x0f98020d, + 0x08004e06, + 0x00013d7e, + 0xf001acf0, + 0x008b04a5, + 0x0c985030, + 0x0fc4b604, + 0x9800bcbb, + 0x0d98020c, + 0x080f9803, + 0x7e02004e, + 0x7e00013d, + 0xf400020a, + 0x12f40601, +/* 0x05b9: ctx_xfer_post */ + 0x02277e07, +/* 0x05bd: ctx_xfer_done */ + 0x04e57e00, + 0x0000f800, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5 new file mode 100644 index 0000000000..47802c7ecc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5 @@ -0,0 +1,42 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define NV_PGRAPH_GPCX_UNK__SIZE 0x00000002 + +#define CHIPSET GM107 +#include "macros.fuc" + +.section #gm107_grgpc_data +#define INCLUDE_DATA +#include "com.fuc" +#include "gpc.fuc" +#undef INCLUDE_DATA + +.section #gm107_grgpc_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "gpc.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h new file mode 100644 index 0000000000..db8b294cee --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h @@ -0,0 +1,607 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gm107_grgpc_data[] = { +/* 0x0000: gpc_mmio_list_head */ + 0x0000006c, +/* 0x0004: gpc_mmio_list_tail */ +/* 0x0004: tpc_mmio_list_head */ + 0x0000006c, +/* 0x0008: tpc_mmio_list_tail */ +/* 0x0008: unk_mmio_list_head */ + 0x0000006c, +/* 0x000c: unk_mmio_list_tail */ + 0x0000006c, +/* 0x0010: gpc_id */ + 0x00000000, +/* 0x0014: tpc_count */ + 0x00000000, +/* 0x0018: tpc_mask */ + 0x00000000, +/* 0x001c: unk_count */ + 0x00000000, +/* 0x0020: unk_mask */ + 0x00000000, +/* 0x0024: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t gm107_grgpc_code[] = { + 0x03410ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0xf489a408, + 0x020f0b1b, + 0x0002f87e, +/* 0x001a: queue_put_next */ + 0x98c400f8, + 0x0384b607, + 0xb6008dbb, + 0x8eb50880, + 0x018fb500, + 0xf00190b6, + 0xd9b50f94, +/* 0x0037: queue_get */ + 0xf400f801, + 0xd8980131, + 0x01d99800, + 0x0bf489a4, + 0x0789c421, + 0xbb0394b6, + 0x90b6009d, + 0x009e9808, + 0xb6019f98, + 0x84f00180, + 0x00d8b50f, +/* 0x0063: queue_get_done */ + 0xf80132f4, +/* 0x0065: nv_rd32 */ + 0xf0ecb200, + 0x00801fc9, + 0x0cf601ca, +/* 0x0073: nv_rd32_wait */ + 0x8c04bd00, + 0xcf01ca00, + 0xccc800cc, + 0xf61bf41f, + 0xec7e060a, + 0x008f0000, + 0xffcf01cb, +/* 0x008f: nv_wr32 */ + 0x8000f800, + 0xf601cc00, + 0x04bd000f, + 0xc9f0ecb2, + 0x1ec9f01f, + 0x01ca0080, + 0xbd000cf6, +/* 0x00a9: nv_wr32_wait */ + 0xca008c04, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f61b, +/* 0x00b8: wait_donez */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x00cf: wait_donez_ne */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf61bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x00ec: wait_doneo */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x0103: wait_doneo_e */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf60bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x0120: mmctx_size */ +/* 0x0122: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0x1bf4efa4, + 0xf89fb2ec, +/* 0x013d: mmctx_xfer */ + 0xf094bd00, + 0x00800199, + 0x09f60237, + 0xbd04bd00, + 0x05bbfd94, + 0x800f0bf4, + 0xf601c400, + 0x04bd000b, +/* 0x015f: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0xc6008018, + 0x000ef601, + 0x008004bd, + 0x0ff601c7, + 0xf004bd00, +/* 0x017a: mmctx_multi_disabled */ + 0xabc80199, + 0x10b4b600, + 0xc80cb9f0, + 0xe4b601ae, + 0x05befd11, + 0x01c50080, + 0xbd000bf6, +/* 0x0195: mmctx_exec_loop */ +/* 0x0195: mmctx_wait_free */ + 0xc5008e04, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f60b, + 0x05e9fd00, + 0x01c80080, + 0xbd000ef6, + 0x04c0b604, + 0x1bf4cda4, + 0x02abc8df, +/* 0x01bf: mmctx_fini_wait */ + 0x8b1c1bf4, + 0xcf01c500, + 0xb4f000bb, + 0x10b4b01f, + 0x0af31bf4, + 0x00b87e05, + 0x250ef400, +/* 0x01d8: mmctx_stop */ + 0xb600abc8, + 0xb9f010b4, + 0x12b9f00c, + 0x01c50080, + 0xbd000bf6, +/* 0x01ed: mmctx_stop_wait */ + 0xc5008b04, + 0x00bbcf01, + 0xf412bbc8, +/* 0x01fa: mmctx_done */ + 0x94bdf61b, + 0x800199f0, + 0xf6021700, + 0x04bd0009, +/* 0x020a: strand_wait */ + 0xa0f900f8, + 0xb87e020a, + 0xa0fc0000, +/* 0x0216: strand_pre */ + 0x0c0900f8, + 0x024afc80, + 0xbd0009f6, + 0x020a7e04, +/* 0x0227: strand_post */ + 0x0900f800, + 0x4afc800d, + 0x0009f602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0238: strand_set */ + 0xfc800f0c, + 0x0cf6024f, + 0x0c04bd00, + 0x4afc800b, + 0x000cf602, + 0xfc8004bd, + 0x0ef6024f, + 0x0c04bd00, + 0x4afc800a, + 0x000cf602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0268: strand_ctx_init */ + 0x99f094bd, + 0x37008003, + 0x0009f602, + 0x167e04bd, + 0x030e0002, + 0x0002387e, + 0xfc80c4bd, + 0x0cf60247, + 0x0c04bd00, + 0x4afc8001, + 0x000cf602, + 0x0a7e04bd, + 0x0c920002, + 0x46fc8001, + 0x000cf602, + 0x020c04bd, + 0x024afc80, + 0xbd000cf6, + 0x020a7e04, + 0x02277e00, + 0x42008800, + 0x20008902, + 0x0099cf02, +/* 0x02c7: ctx_init_strand_loop */ + 0xf608fe95, + 0x8ef6008e, + 0x808acf40, + 0xb606a5b6, + 0xeabb01a0, + 0x0480b600, + 0xf40192b6, + 0xe4b6e81b, + 0xf2efbc08, + 0x99f094bd, + 0x17008003, + 0x0009f602, + 0x00f804bd, +/* 0x02f8: error */ + 0xffb2e0f9, + 0x4098148e, + 0x00008f7e, + 0xffb2010f, + 0x409c1c8e, + 0x00008f7e, + 0x00f8e0fc, +/* 0x0314: tpc_strand_wait */ + 0x94bd90f9, + 0x800a99f0, + 0xf6023700, + 0x04bd0009, +/* 0x0324: tpc_strand_busy */ + 0x033f0089, + 0xb30099cf, + 0xbdf90094, + 0x0a99f094, + 0x02170080, + 0xbd0009f6, + 0xf890fc04, +/* 0x0341: init */ + 0x4104bd00, + 0x11cf4200, + 0x0911e700, + 0x0814b601, + 0x020014fe, + 0x12004002, + 0xbd0002f6, + 0x05ad4104, + 0x400010fe, + 0x00f60700, + 0x0204bd00, + 0x04004004, + 0xbd0002f6, + 0x1031f404, + 0x01820082, + 0x030022cf, + 0x1f24f001, + 0xb60432bb, + 0x02b50132, + 0x0603b505, + 0x01860082, + 0xb50022cf, + 0x24b60402, + 0xc900800f, + 0x0002f601, + 0x308e04bd, + 0xe5f0500c, + 0xbd24bd01, +/* 0x03b3: init_unk_loop */ + 0x7e44bd34, + 0xb0000065, + 0x0bf400f6, + 0xbb010f0e, + 0x4ffd04f2, + 0x0130b605, +/* 0x03c8: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf402, +/* 0x03d4: init_unk_done */ + 0xb50703b5, + 0x00820804, + 0x22cf0201, + 0x9534bd00, + 0x00800825, + 0x05f601c0, + 0x8004bd00, + 0xf601c100, + 0x04bd0005, + 0x98000e98, + 0x207e010f, + 0x2fbb0001, + 0x003fbb00, + 0x98010e98, + 0x207e020f, + 0x0e980001, + 0x00effd05, + 0xbb002ebb, + 0x0e98003e, + 0x030f9802, + 0x0001207e, + 0xfd070e98, + 0x2ebb00ef, + 0x003ebb00, + 0x800235b6, + 0xf601d300, + 0x04bd0003, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb20834, + 0x0002687e, + 0xbb002fbb, + 0x3f0f003f, + 0x501d608e, + 0xb201e5f0, + 0x008f7eff, + 0x8e0c0f00, + 0xf0501da8, + 0xffb201e5, + 0x00008f7e, + 0x0003147e, + 0x608e3f0f, + 0xe5f0501d, + 0x7effb201, + 0x0f00008f, + 0x1d9c8e00, + 0x01e5f050, + 0x8f7effb2, + 0x010f0000, + 0x0003147e, + 0x501da88e, + 0xb201e5f0, + 0x008f7eff, + 0x8eff0f00, + 0xf0501d98, + 0xffb201e5, + 0x00008f7e, + 0xa88e020f, + 0xe5f0501d, + 0x7effb201, + 0x7e00008f, + 0x98000314, + 0x00850504, + 0x55f05040, +/* 0x04dd: tpc_strand_init_tpc_loop */ + 0x705eb801, + 0x657e0005, + 0xf6b20000, +/* 0x04ea: tpc_strand_init_idx_loop */ + 0x5eb874bd, + 0xb2000560, + 0x008f7e7f, + 0x885eb800, + 0x2f950005, + 0x008f7e08, + 0x8c5eb800, + 0x2f950005, + 0x008f7e08, + 0x905eb800, + 0x657e0005, + 0xf5b60000, + 0x01f0b606, + 0xbb08f4b6, + 0x3fbb002f, + 0x0170b600, + 0xf40162b6, + 0x50b7bf1b, + 0x42b60800, + 0xa81bf401, + 0x608e3f0f, + 0xe5f0501d, + 0x7effb201, + 0x0f00008f, + 0x1da88e0d, + 0x01e5f050, + 0x8f7effb2, + 0x147e0000, + 0x00800003, + 0x03f60201, + 0xbd04bd00, + 0x1f29f024, + 0x02300080, + 0xbd0002f6, +/* 0x0571: wait */ + 0x0028f404, +/* 0x0577: main */ + 0x0d0031f4, + 0x00377e24, + 0xf401f400, + 0xf404e4b0, + 0x81fe1d18, + 0xbd060201, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0x06487e00, + 0xda0ef400, +/* 0x05a0: main_not_ctx_xfer */ + 0xf010ef94, + 0xf87e01f5, + 0x0ef40002, +/* 0x05ad: ih */ + 0xf900f9cd, + 0x0188fe80, + 0x90f980f9, + 0xb0f9a0f9, + 0xe0f9d0f9, + 0x04bdf0f9, + 0xcf02004a, + 0xabc400aa, + 0x1f0bf404, + 0x004e240d, + 0x00eecf1a, + 0xcf19004f, + 0x047e00ff, + 0x010e0000, + 0xf61d0040, + 0x04bd000e, +/* 0x05ec: ih_no_fifo */ + 0xf6010040, + 0x04bd000a, + 0xe0fcf0fc, + 0xb0fcd0fc, + 0x90fca0fc, + 0x88fe80fc, + 0xfc80fc00, + 0x0032f400, +/* 0x060e: hub_barrier_done */ + 0x010f01f8, + 0xbb040e98, + 0xffb204fe, + 0x4094188e, + 0x00008f7e, +/* 0x0622: ctx_redswitch */ + 0x200f00f8, + 0x01850080, + 0xbd000ff6, +/* 0x062f: ctx_redswitch_delay */ + 0xb6080e04, + 0x1bf401e2, + 0x00f5f1fd, + 0x00f5f108, + 0x85008002, + 0x000ff601, + 0x00f804bd, +/* 0x0648: ctx_xfer */ + 0x02810080, + 0xbd000ff6, + 0x1dc48e04, + 0x01e5f050, + 0x8f7effb2, + 0x11f40000, + 0x06227e07, +/* 0x0665: ctx_xfer_not_load */ + 0x02167e00, + 0x8024bd00, + 0xf60247fc, + 0x04bd0002, + 0xb6012cf0, + 0xfc800320, + 0x02f6024a, + 0x0f04bd00, + 0x1da88e0c, + 0x01e5f050, + 0x8f7effb2, + 0x147e0000, + 0x3f0f0003, + 0x501d608e, + 0xb201e5f0, + 0x008f7eff, + 0x8e000f00, + 0xf0501d9c, + 0xffb201e5, + 0x00008f7e, + 0x147e010f, + 0xfcf00003, + 0x03f0b601, + 0x501da88e, + 0xb201e5f0, + 0x008f7eff, + 0x01acf000, + 0x8b02a5f0, + 0x98500000, + 0xc4b6040c, + 0x00bcbb0f, + 0x98000c98, + 0x000e010d, + 0x00013d7e, + 0x8b01acf0, + 0x98504000, + 0xc4b6040c, + 0x00bcbb0f, + 0x98010c98, + 0x0f98020d, + 0x08004e06, + 0x00013d7e, + 0xf001acf0, + 0x008b04a5, + 0x0c985030, + 0x0fc4b604, + 0x9800bcbb, + 0x0d98020c, + 0x080f9803, + 0x7e02004e, + 0x7e00013d, + 0x7e00020a, + 0xf4000314, + 0x12f40601, +/* 0x073d: ctx_xfer_post */ + 0x02277e1a, + 0x8e0d0f00, + 0xf0501da8, + 0xffb201e5, + 0x00008f7e, + 0x0003147e, +/* 0x0754: ctx_xfer_done */ + 0x00060e7e, + 0x000000f8, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hub.fuc b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hub.fuc new file mode 100644 index 0000000000..4d416d4f82 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hub.fuc @@ -0,0 +1,699 @@ +/* fuc microcode for gf100 PGRAPH/HUB + * + * Copyright 2011 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#ifdef INCLUDE_DATA +hub_mmio_list_head: .b32 #hub_mmio_list_base +hub_mmio_list_tail: .b32 #hub_mmio_list_next + +gpc_count: .b32 0 +rop_count: .b32 0 +cmd_queue: queue_init + +ctx_current: .b32 0 + +.align 256 +chan_data: +chan_mmio_count: .b32 0 +chan_mmio_address: .b32 0 + +.align 256 +xfer_data: .skip 256 + +hub_mmio_list_base: +.b32 0x0417e91c // 0x17e91c, 2 +hub_mmio_list_next: +#endif + +#ifdef INCLUDE_CODE +// reports an exception to the host +// +// In: $r15 error code (see os.h) +// +error: + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(5), 0, $r15) + mov $r15 1 + nv_iowr(NV_PGRAPH_FECS_INTR_UP_SET, 0, $r15) + ret + +// HUB fuc initialisation, executed by triggering ucode start, will +// fall through to main loop after completion. +// +// Output: +// CC_SCRATCH[0]: +// 31:31: set to signal completion +// CC_SCRATCH[1]: +// 31:0: total PGRAPH context size +// +init: + clear b32 $r0 + mov $xdbase $r0 + + // setup stack + nv_iord($r1, NV_PGRAPH_FECS_CAPS, 0) + extr $r1 $r1 9:17 + shl b32 $r1 8 + mov $sp $r1 + + // enable fifo access + mov $r2 NV_PGRAPH_FECS_ACCESS_FIFO + nv_iowr(NV_PGRAPH_FECS_ACCESS, 0, $r2) + + // setup i0 handler, and route all interrupts to it + mov $r1 #ih + mov $iv0 $r1 + + clear b32 $r2 + nv_iowr(NV_PGRAPH_FECS_INTR_ROUTE, 0, $r2) + + // route HUB_CHSW_PULSE to fuc interrupt 8 + mov $r2 0x2003 // { HUB_CHSW_PULSE, ZERO } -> intr 8 + nv_iowr(NV_PGRAPH_FECS_IROUTE, 0, $r2) + + // not sure what these are, route them because NVIDIA does, and + // the IRQ handler will signal the host if we ever get one.. we + // may find out if/why we need to handle these if so.. + // + mov $r2 0x2004 // { 0x04, ZERO } -> intr 9 + nv_iowr(NV_PGRAPH_FECS_IROUTE, 1, $r2) + mov $r2 0x200b // { HUB_FIRMWARE_MTHD, ZERO } -> intr 10 + nv_iowr(NV_PGRAPH_FECS_IROUTE, 2, $r2) + mov $r2 0x200c // { 0x0c, ZERO } -> intr 15 + nv_iowr(NV_PGRAPH_FECS_IROUTE, 7, $r2) + + // enable all INTR_UP interrupts + sub b32 $r3 $r0 1 + nv_iowr(NV_PGRAPH_FECS_INTR_UP_EN, 0, $r3) + + // enable fifo, ctxsw, 9, fwmthd, 15 interrupts + imm32($r2, 0x8704) + nv_iowr(NV_PGRAPH_FECS_INTR_EN_SET, 0, $r2) + + // fifo level triggered, rest edge + mov $r2 NV_PGRAPH_FECS_INTR_MODE_FIFO_LEVEL + nv_iowr(NV_PGRAPH_FECS_INTR_MODE, 0, $r2) + + // enable interrupts + bset $flags ie0 + + // fetch enabled GPC/ROP counts + nv_rd32($r14, 0x409604) + extr $r1 $r15 16:20 + st b32 D[$r0 + #rop_count] $r1 + and $r15 0x1f + st b32 D[$r0 + #gpc_count] $r15 + + // set BAR_REQMASK to GPC mask + mov $r1 1 + shl b32 $r1 $r15 + sub b32 $r1 1 + nv_iowr(NV_PGRAPH_FECS_BAR_MASK0, 0, $r1) + nv_iowr(NV_PGRAPH_FECS_BAR_MASK1, 0, $r1) + + // context size calculation, reserve first 256 bytes for use by fuc + mov $r1 256 + + // + mov $r15 2 + call(ctx_4170s) + call(ctx_4170w) + mov $r15 0x10 + call(ctx_86c) + + // calculate size of mmio context data + ld b32 $r14 D[$r0 + #hub_mmio_list_head] + ld b32 $r15 D[$r0 + #hub_mmio_list_tail] + call(mmctx_size) + + // set mmctx base addresses now so we don't have to do it later, + // they don't (currently) ever change + shr b32 $r4 $r1 8 + nv_iowr(NV_PGRAPH_FECS_MMCTX_SAVE_SWBASE, 0, $r4) + nv_iowr(NV_PGRAPH_FECS_MMCTX_LOAD_SWBASE, 0, $r4) + add b32 $r3 0x1300 + add b32 $r1 $r15 + shr b32 $r15 2 + nv_iowr(NV_PGRAPH_FECS_MMCTX_LOAD_COUNT, 0, $r15) // wtf?? + + // strands, base offset needs to be aligned to 256 bytes + shr b32 $r1 8 + add b32 $r1 1 + shl b32 $r1 8 + mov b32 $r15 $r1 + call(strand_ctx_init) + add b32 $r1 $r15 + + // initialise each GPC in sequence by passing in the offset of its + // context data in GPCn_CC_SCRATCH[1], and starting its FUC (which + // has previously been uploaded by the host) running. + // + // the GPC fuc init sequence will set GPCn_CC_SCRATCH[0] bit 31 + // when it has completed, and return the size of its context data + // in GPCn_CC_SCRATCH[1] + // + ld b32 $r3 D[$r0 + #gpc_count] + imm32($r4, 0x502000) + init_gpc: + // setup, and start GPC ucode running + add b32 $r14 $r4 0x804 + mov b32 $r15 $r1 + call(nv_wr32) // CC_SCRATCH[1] = ctx offset + add b32 $r14 $r4 0x10c + clear b32 $r15 + call(nv_wr32) + add b32 $r14 $r4 0x104 + call(nv_wr32) // ENTRY + add b32 $r14 $r4 0x100 + mov $r15 2 // CTRL_START_TRIGGER + call(nv_wr32) // CTRL + + // wait for it to complete, and adjust context size + add b32 $r14 $r4 0x800 + init_gpc_wait: + call(nv_rd32) + xbit $r15 $r15 31 + bra e #init_gpc_wait + add b32 $r14 $r4 0x804 + call(nv_rd32) + add b32 $r1 $r15 + + // next! + add b32 $r4 0x8000 + sub b32 $r3 1 + bra ne #init_gpc + + // + mov $r15 0 + call(ctx_86c) + mov $r15 0 + call(ctx_4170s) + + // save context size, and tell host we're ready + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(1), 0, $r1) + clear b32 $r1 + bset $r1 31 + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_SET(0), 0, $r1) + +// Main program loop, very simple, sleeps until woken up by the interrupt +// handler, pulls a command from the queue and executes its handler +// +wait: + // sleep until we have something to do + sleep $p0 + bset $flags $p0 +main: + mov $r13 #cmd_queue + call(queue_get) + bra $p1 #wait + + // context switch, requested by GPU? + cmpu b32 $r14 0x4001 + bra ne #main_not_ctx_switch + trace_set(T_AUTO) + nv_iord($r1, NV_PGRAPH_FECS_CHAN_ADDR, 0) + nv_iord($r2, NV_PGRAPH_FECS_CHAN_NEXT, 0) + + xbit $r3 $r1 31 + bra e #chsw_no_prev + xbit $r3 $r2 31 + bra e #chsw_prev_no_next + push $r2 + mov b32 $r2 $r1 + trace_set(T_SAVE) + bclr $flags $p1 + bset $flags $p2 + call(ctx_xfer) + trace_clr(T_SAVE); + pop $r2 + trace_set(T_LOAD); + bset $flags $p1 + call(ctx_xfer) + trace_clr(T_LOAD); + bra #chsw_done + chsw_prev_no_next: + push $r2 + mov b32 $r2 $r1 + bclr $flags $p1 + bclr $flags $p2 + call(ctx_xfer) + pop $r2 + nv_iowr(NV_PGRAPH_FECS_CHAN_ADDR, 0, $r2) + bra #chsw_done + chsw_no_prev: + xbit $r3 $r2 31 + bra e #chsw_done + bset $flags $p1 + bclr $flags $p2 + call(ctx_xfer) + + // ack the context switch request + chsw_done: + mov $r2 NV_PGRAPH_FECS_CHSW_ACK + nv_iowr(NV_PGRAPH_FECS_CHSW, 0, $r2) + trace_clr(T_AUTO) + bra #main + + // request to set current channel? (*not* a context switch) + main_not_ctx_switch: + cmpu b32 $r14 0x0001 + bra ne #main_not_ctx_chan + mov b32 $r2 $r15 + call(ctx_chan) + bra #main_done + + // request to store current channel context? + main_not_ctx_chan: + cmpu b32 $r14 0x0002 + bra ne #main_not_ctx_save + trace_set(T_SAVE) + bclr $flags $p1 + bclr $flags $p2 + call(ctx_xfer) + trace_clr(T_SAVE) + bra #main_done + + main_not_ctx_save: + shl b32 $r15 $r14 16 + or $r15 E_BAD_COMMAND + call(error) + bra #main + + main_done: + clear b32 $r2 + bset $r2 31 + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_SET(0), 0, $r2) + bra #main + +// interrupt handler +ih: + push $r0 + push $r8 + mov $r8 $flags + push $r8 + push $r9 + push $r10 + push $r11 + push $r13 + push $r14 + push $r15 + clear b32 $r0 + + // incoming fifo command? + nv_iord($r10, NV_PGRAPH_FECS_INTR, 0) + and $r11 $r10 NV_PGRAPH_FECS_INTR_FIFO + bra e #ih_no_fifo + // queue incoming fifo command for later processing + mov $r13 #cmd_queue + nv_iord($r14, NV_PGRAPH_FECS_FIFO_CMD, 0) + nv_iord($r15, NV_PGRAPH_FECS_FIFO_DATA, 0) + call(queue_put) + add b32 $r11 0x400 + mov $r14 1 + nv_iowr(NV_PGRAPH_FECS_FIFO_ACK, 0, $r14) + + // context switch request? + ih_no_fifo: + and $r11 $r10 NV_PGRAPH_FECS_INTR_CHSW + bra e #ih_no_ctxsw + // enqueue a context switch for later processing + mov $r13 #cmd_queue + mov $r14 0x4001 + call(queue_put) + + // firmware method? + ih_no_ctxsw: + and $r11 $r10 NV_PGRAPH_FECS_INTR_FWMTHD + bra e #ih_no_fwmthd + // none we handle; report to host and ack + nv_rd32($r15, NV_PGRAPH_TRAPPED_DATA_LO) + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(4), 0, $r15) + nv_rd32($r15, NV_PGRAPH_TRAPPED_ADDR) + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(3), 0, $r15) + extr $r14 $r15 16:18 + shl b32 $r14 $r14 2 + imm32($r15, NV_PGRAPH_FE_OBJECT_TABLE(0)) + add b32 $r14 $r15 + call(nv_rd32) + nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_VAL(2), 0, $r15) + mov $r15 E_BAD_FWMTHD + call(error) + mov $r11 0x100 + nv_wr32(0x400144, $r11) + + // anything we didn't handle, bring it to the host's attention + ih_no_fwmthd: + mov $r11 0x504 // FIFO | CHSW | FWMTHD + not b32 $r11 + and $r11 $r10 $r11 + bra e #ih_no_other + nv_iowr(NV_PGRAPH_FECS_INTR_UP_SET, 0, $r11) + + // ack, and wake up main() + ih_no_other: + nv_iowr(NV_PGRAPH_FECS_INTR_ACK, 0, $r10) + + pop $r15 + pop $r14 + pop $r13 + pop $r11 + pop $r10 + pop $r9 + pop $r8 + mov $flags $r8 + pop $r8 + pop $r0 + bclr $flags $p0 + iret + +#if CHIPSET < GK100 +// Not real sure, but, MEM_CMD 7 will hang forever if this isn't done +ctx_4160s: + mov $r15 1 + nv_wr32(0x404160, $r15) + ctx_4160s_wait: + nv_rd32($r15, 0x404160) + xbit $r15 $r15 4 + bra e #ctx_4160s_wait + ret + +// Without clearing again at end of xfer, some things cause PGRAPH +// to hang with STATUS=0x00000007 until it's cleared.. fbcon can +// still function with it set however... +ctx_4160c: + clear b32 $r15 + nv_wr32(0x404160, $r15) + ret +#endif + +// Again, not real sure +// +// In: $r15 value to set 0x404170 to +// +ctx_4170s: + or $r15 0x10 + nv_wr32(0x404170, $r15) + ret + +// Waits for a ctx_4170s() call to complete +// +ctx_4170w: + nv_rd32($r15, 0x404170) + and $r15 0x10 + bra ne #ctx_4170w + ret + +// Disables various things, waits a bit, and re-enables them.. +// +// Not sure how exactly this helps, perhaps "ENABLE" is not such a +// good description for the bits we turn off? Anyways, without this, +// funny things happen. +// +ctx_redswitch: + mov $r14 NV_PGRAPH_FECS_RED_SWITCH_ENABLE_GPC + or $r14 NV_PGRAPH_FECS_RED_SWITCH_POWER_ROP + or $r14 NV_PGRAPH_FECS_RED_SWITCH_POWER_GPC + or $r14 NV_PGRAPH_FECS_RED_SWITCH_POWER_MAIN + nv_iowr(NV_PGRAPH_FECS_RED_SWITCH, 0, $r14) + mov $r15 8 + ctx_redswitch_delay: + sub b32 $r15 1 + bra ne #ctx_redswitch_delay + or $r14 NV_PGRAPH_FECS_RED_SWITCH_ENABLE_ROP + or $r14 NV_PGRAPH_FECS_RED_SWITCH_ENABLE_MAIN + nv_iowr(NV_PGRAPH_FECS_RED_SWITCH, 0, $r14) + ret + +// Not a clue what this is for, except that unless the value is 0x10, the +// strand context is saved (and presumably restored) incorrectly.. +// +// In: $r15 value to set to (0x00/0x10 are used) +// +ctx_86c: + nv_iowr(NV_PGRAPH_FECS_UNK86C, 0, $r15) + nv_wr32(0x408a14, $r15) + nv_wr32(NV_PGRAPH_GPCX_GPCCS_UNK86C, $r15) + ret + +// In: $r15 NV_PGRAPH_FECS_MEM_CMD_* +ctx_mem: + nv_iowr(NV_PGRAPH_FECS_MEM_CMD, 0, $r15) + ctx_mem_wait: + nv_iord($r15, NV_PGRAPH_FECS_MEM_CMD, 0) + or $r15 $r15 + bra ne #ctx_mem_wait + ret + +// ctx_load - load's a channel's ctxctl data, and selects its vm +// +// In: $r2 channel address +// +ctx_load: + trace_set(T_CHAN) + + // switch to channel, somewhat magic in parts.. + mov $r10 12 // DONE_UNK12 + call(wait_donez) + clear b32 $r15 + nv_iowr(0x409a24, 0, $r15) + nv_iowr(NV_PGRAPH_FECS_CHAN_NEXT, 0, $r2) + nv_iowr(NV_PGRAPH_FECS_MEM_CHAN, 0, $r2) + mov $r15 NV_PGRAPH_FECS_MEM_CMD_LOAD_CHAN + call(ctx_mem) + nv_iowr(NV_PGRAPH_FECS_CHAN_ADDR, 0, $r2) + + // load channel header, fetch PGRAPH context pointer + mov $xtargets $r0 + bclr $r2 31 + shl b32 $r2 4 + add b32 $r2 2 + + trace_set(T_LCHAN) + nv_iowr(NV_PGRAPH_FECS_MEM_BASE, 0, $r2) + imm32($r2, NV_PGRAPH_FECS_MEM_TARGET_UNK31) + or $r2 NV_PGRAPH_FECS_MEM_TARGET_AS_VRAM + nv_iowr(NV_PGRAPH_FECS_MEM_TARGET, 0, $r2) + mov $r1 0x10 // chan + 0x0210 + mov $r2 #xfer_data + sethi $r2 0x00020000 // 16 bytes + xdld $r1 $r2 + xdwait + trace_clr(T_LCHAN) + + // update current context + ld b32 $r1 D[$r0 + #xfer_data + 4] + shl b32 $r1 24 + ld b32 $r2 D[$r0 + #xfer_data + 0] + shr b32 $r2 8 + or $r1 $r2 + st b32 D[$r0 + #ctx_current] $r1 + + // set transfer base to start of context, and fetch context header + trace_set(T_LCTXH) + nv_iowr(NV_PGRAPH_FECS_MEM_BASE, 0, $r1) + mov $r2 NV_PGRAPH_FECS_MEM_TARGET_AS_VM + nv_iowr(NV_PGRAPH_FECS_MEM_TARGET, 0, $r2) + mov $r1 #chan_data + sethi $r1 0x00060000 // 256 bytes + xdld $r0 $r1 + xdwait + trace_clr(T_LCTXH) + + trace_clr(T_CHAN) + ret + +// ctx_chan - handler for HUB_SET_CHAN command, will set a channel as +// the active channel for ctxctl, but not actually transfer +// any context data. intended for use only during initial +// context construction. +// +// In: $r2 channel address +// +ctx_chan: +#if CHIPSET < GK100 + call(ctx_4160s) +#endif + call(ctx_load) + mov $r10 12 // DONE_UNK12 + call(wait_donez) + mov $r15 5 // MEM_CMD 5 ??? + call(ctx_mem) +#if CHIPSET < GK100 + call(ctx_4160c) +#endif + ret + +// Execute per-context state overrides list +// +// Only executed on the first load of a channel. Might want to look into +// removing this and having the host directly modify the channel's context +// to change this state... The nouveau DRM already builds this list as +// it's definitely needed for NVIDIA's, so we may as well use it for now +// +// Input: $r1 mmio list length +// +ctx_mmio_exec: + // set transfer base to be the mmio list + ld b32 $r3 D[$r0 + #chan_mmio_address] + nv_iowr(NV_PGRAPH_FECS_MEM_BASE, 0, $r3) + + clear b32 $r3 + ctx_mmio_loop: + // fetch next 256 bytes of mmio list if necessary + and $r4 $r3 0xff + bra ne #ctx_mmio_pull + mov $r5 #xfer_data + sethi $r5 0x00060000 // 256 bytes + xdld $r3 $r5 + xdwait + + // execute a single list entry + ctx_mmio_pull: + ld b32 $r14 D[$r4 + #xfer_data + 0x00] + ld b32 $r15 D[$r4 + #xfer_data + 0x04] + call(nv_wr32) + + // next! + add b32 $r3 8 + sub b32 $r1 1 + bra ne #ctx_mmio_loop + + // set transfer base back to the current context + ctx_mmio_done: + ld b32 $r3 D[$r0 + #ctx_current] + nv_iowr(NV_PGRAPH_FECS_MEM_BASE, 0, $r3) + + // disable the mmio list now, we don't need/want to execute it again + st b32 D[$r0 + #chan_mmio_count] $r0 + mov $r1 #chan_data + sethi $r1 0x00060000 // 256 bytes + xdst $r0 $r1 + xdwait + ret + +// Transfer HUB context data between GPU and storage area +// +// In: $r2 channel address +// $p1 clear on save, set on load +// $p2 set if opposite direction done/will be done, so: +// on save it means: "a load will follow this save" +// on load it means: "a save preceeded this load" +// +ctx_xfer: + // according to mwk, some kind of wait for idle + mov $r14 4 + nv_iowr(0x409c08, 0, $r14) + ctx_xfer_idle: + nv_iord($r14, 0x409c00, 0) + and $r14 0x2000 + bra ne #ctx_xfer_idle + + bra not $p1 #ctx_xfer_pre + bra $p2 #ctx_xfer_pre_load + ctx_xfer_pre: + mov $r15 0x10 + call(ctx_86c) +#if CHIPSET < GK100 + call(ctx_4160s) +#endif + bra not $p1 #ctx_xfer_exec + + ctx_xfer_pre_load: + mov $r15 2 + call(ctx_4170s) + call(ctx_4170w) + call(ctx_redswitch) + clear b32 $r15 + call(ctx_4170s) + call(ctx_load) + + // fetch context pointer, and initiate xfer on all GPCs + ctx_xfer_exec: + ld b32 $r1 D[$r0 + #ctx_current] + + clear b32 $r2 + nv_iowr(NV_PGRAPH_FECS_BAR, 0, $r2) + + nv_wr32(0x41a500, $r1) // GPC_BCAST_WRCMD_DATA = ctx pointer + xbit $r15 $flags $p1 + xbit $r2 $flags $p2 + shl b32 $r2 1 + or $r15 $r2 + nv_wr32(0x41a504, $r15) // GPC_BCAST_WRCMD_CMD = GPC_XFER(type) + + // strands + call(strand_pre) + clear b32 $r2 + nv_iowr(NV_PGRAPH_FECS_STRAND_SELECT, 0x3f, $r2) + xbit $r2 $flags $p1 // SAVE/LOAD + add b32 $r2 NV_PGRAPH_FECS_STRAND_CMD_SAVE + nv_iowr(NV_PGRAPH_FECS_STRAND_CMD, 0x3f, $r2) + + // mmio context + xbit $r10 $flags $p1 // direction + or $r10 6 // first, last + mov $r11 0 // base = 0 + ld b32 $r12 D[$r0 + #hub_mmio_list_head] + ld b32 $r13 D[$r0 + #hub_mmio_list_tail] + mov $r14 0 // not multi + call(mmctx_xfer) + + // wait for GPCs to all complete + mov $r10 8 // DONE_BAR + call(wait_doneo) + + // wait for strand xfer to complete + call(strand_wait) + + // post-op + bra $p1 #ctx_xfer_post + mov $r10 12 // DONE_UNK12 + call(wait_donez) + mov $r15 5 // MEM_CMD 5 ??? + call(ctx_mem) + + bra $p2 #ctx_xfer_done + ctx_xfer_post: + mov $r15 2 + call(ctx_4170s) + clear b32 $r15 + call(ctx_86c) + call(strand_post) + call(ctx_4170w) + clear b32 $r15 + call(ctx_4170s) + + bra not $p1 #ctx_xfer_no_post_mmio + ld b32 $r1 D[$r0 + #chan_mmio_count] + or $r1 $r1 + bra e #ctx_xfer_no_post_mmio + call(ctx_mmio_exec) + + ctx_xfer_no_post_mmio: +#if CHIPSET < GK100 + call(ctx_4160c) +#endif + + ctx_xfer_done: + ret +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3 new file mode 100644 index 0000000000..2c28e7199b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GF100 +#include "macros.fuc" + +.section #gf100_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gf100_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h new file mode 100644 index 0000000000..56162f6a6a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf100.fuc3.h @@ -0,0 +1,1049 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gf100_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gf100_grhub_code[] = { + 0x039b0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0x07f100f8, + 0x03f00500, + 0x000fd002, + 0xf7f004bd, + 0x0007f101, + 0x0303f007, + 0xbd000fd0, +/* 0x039b: init */ + 0xbd00f804, + 0x0007fe04, + 0x420017f1, + 0xcf0013f0, + 0x11e70011, + 0x14b60109, + 0x0014fe08, + 0xf10227f0, + 0xf0120007, + 0x02d00003, + 0xf104bd00, + 0xfe06c817, + 0x24bd0010, + 0x070007f1, + 0xd00003f0, + 0x04bd0002, + 0x200327f1, + 0x010007f1, + 0xd00103f0, + 0x04bd0002, + 0x200427f1, + 0x010407f1, + 0xd00103f0, + 0x04bd0002, + 0x200b27f1, + 0x010807f1, + 0xd00103f0, + 0x04bd0002, + 0x200c27f1, + 0x011c07f1, + 0xd00103f0, + 0x04bd0002, + 0xf1010392, + 0xf0090007, + 0x03d00303, + 0xf104bd00, + 0xf0870427, + 0x07f10023, + 0x03f00400, + 0x0002d000, + 0x27f004bd, + 0x0007f104, + 0x0003f003, + 0xbd0002d0, + 0x1031f404, + 0x9604e7f1, + 0xf440e3f0, + 0xfeb96821, + 0x90f1c702, + 0xf0030180, + 0x0f801ff4, + 0x0117f002, + 0xb6041fbb, + 0x07f10112, + 0x03f00300, + 0x0001d001, + 0x07f104bd, + 0x03f00400, + 0x0001d001, + 0x17f104bd, + 0xf7f00100, + 0x1121f502, + 0x2321f508, + 0x10f7f008, + 0x087021f5, + 0x98000e98, + 0x21f5010f, + 0x14950150, + 0x0007f108, + 0x0103f0c0, + 0xbd0004d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0004d0, + 0x0030b704, + 0x001fbb13, + 0xf102f5b6, + 0xf0d30007, + 0x0fd00103, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0xf5021fb9, + 0xbb02d321, + 0x0398001f, + 0x0047f102, + 0x5043f020, +/* 0x04f4: init_gpc */ + 0x08044ea0, + 0xf4021fb9, + 0x4ea09d21, + 0xf4bd010c, + 0xa09d21f4, + 0xf401044e, + 0x4ea09d21, + 0xf7f00100, + 0x9d21f402, + 0x08004ea0, +/* 0x051c: init_gpc_wait */ + 0xc86821f4, + 0x0bf41fff, + 0x044ea0fa, + 0x6821f408, + 0xb7001fbb, + 0xb6800040, + 0x1bf40132, + 0x00f7f0be, + 0x087021f5, + 0xf500f7f0, + 0xf1081121, + 0xf0010007, + 0x01d00203, + 0xbd04bd00, + 0x1f19f014, + 0x080007f1, + 0xd00203f0, + 0x04bd0001, +/* 0x0564: wait */ + 0xf40028f4, +/* 0x056a: main */ + 0xd7f00031, + 0x3921f410, + 0xb1f401f4, + 0xf54001e4, + 0xbd00e91b, + 0x0499f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xc00017f1, + 0xcf0213f0, + 0x27f10011, + 0x23f0c100, + 0x0022cf02, + 0xf51f13c8, + 0xc800890b, + 0x0bf41f23, + 0xb920f962, + 0x94bd0212, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x31f40132, + 0x4421f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x20fc04bd, + 0x99f094bd, + 0x0007f106, + 0x0203f00f, + 0xbd0009d0, + 0x0131f404, + 0x0a4421f5, + 0x99f094bd, + 0x0007f106, + 0x0203f017, + 0xbd0009d0, + 0x330ef404, +/* 0x060c: chsw_prev_no_next */ + 0x12b920f9, + 0x0132f402, + 0xf50232f4, + 0xfc0a4421, + 0x0007f120, + 0x0203f0c0, + 0xbd0002d0, + 0x130ef404, +/* 0x062c: chsw_no_prev */ + 0xf41f23c8, + 0x31f40d0b, + 0x0232f401, + 0x0a4421f5, +/* 0x063c: chsw_done */ + 0xf10127f0, + 0xf0c30007, + 0x02d00203, + 0xbd04bd00, + 0x0499f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xff0e0ef5, +/* 0x0660: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b90d1b, + 0xd421f502, + 0x460ef409, +/* 0x0670: main_not_ctx_chan */ + 0xf402e4b0, + 0x94bd321b, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x32f40132, + 0x4421f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x0ef404bd, +/* 0x06a5: main_not_ctx_save */ + 0x10ef9411, + 0xf501f5f0, + 0xf5037e21, +/* 0x06b3: main_done */ + 0xbdfebb0e, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, + 0xfea60ef5, +/* 0x06c8: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0xf104bdf0, + 0xf00200a7, + 0xaacf00a3, + 0x04abc400, + 0xf0300bf4, + 0xe7f110d7, + 0xe3f01a00, + 0x00eecf00, + 0x1900f7f1, + 0xcf00f3f0, + 0x21f400ff, + 0x00b0b704, + 0x01e7f004, + 0x1d0007f1, + 0xd00003f0, + 0x04bd000e, +/* 0x071c: ih_no_fifo */ + 0x0100abe4, + 0xf00d0bf4, + 0xe7f110d7, + 0x21f44001, +/* 0x072d: ih_no_ctxsw */ + 0x00abe404, + 0x6c0bf404, + 0x0708e7f1, + 0xf440e3f0, + 0xffb96821, + 0x0007f102, + 0x0203f004, + 0xbd000fd0, + 0x04e7f104, + 0x40e3f007, + 0xb96821f4, + 0x07f102ff, + 0x03f00300, + 0x000fd002, + 0xfec704bd, + 0x02ee9450, + 0x0700f7f1, + 0xbb40f3f0, + 0x21f400ef, + 0x0007f168, + 0x0203f002, + 0xbd000fd0, + 0x03f7f004, + 0x037e21f5, + 0x0100b7f1, + 0xf102bfb9, + 0xf00144e7, + 0x21f440e3, +/* 0x079d: ih_no_fwmthd */ + 0x04b7f19d, + 0xffb0bd05, + 0x0bf4b4ab, + 0x0007f10f, + 0x0303f007, + 0xbd000bd0, +/* 0x07b5: ih_no_other */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x07db: ctx_4160s */ + 0xf001f800, + 0xffb901f7, + 0x60e7f102, + 0x40e3f041, +/* 0x07eb: ctx_4160s_wait */ + 0xf19d21f4, + 0xf04160e7, + 0x21f440e3, + 0x02ffb968, + 0xf404ffc8, + 0x00f8f00b, +/* 0x0800: ctx_4160c */ + 0xffb9f4bd, + 0x60e7f102, + 0x40e3f041, + 0xf89d21f4, +/* 0x0811: ctx_4170s */ + 0x10f5f000, + 0xf102ffb9, + 0xf04170e7, + 0x21f440e3, +/* 0x0823: ctx_4170w */ + 0xf100f89d, + 0xf04170e7, + 0x21f440e3, + 0x02ffb968, + 0xf410f4f0, + 0x00f8f01b, +/* 0x0838: ctx_redswitch */ + 0x0200e7f1, + 0xf040e5f0, + 0xe5f020e5, + 0x0007f110, + 0x0103f085, + 0xbd000ed0, + 0x08f7f004, +/* 0x0854: ctx_redswitch_delay */ + 0xf401f2b6, + 0xe5f1fd1b, + 0xe5f10400, + 0x07f10100, + 0x03f08500, + 0x000ed001, + 0x00f804bd, +/* 0x0870: ctx_86c */ + 0x1b0007f1, + 0xd00203f0, + 0x04bd000f, + 0xf102ffb9, + 0xf08a14e7, + 0x21f440e3, + 0x02ffb99d, + 0xa86ce7f1, + 0xf441e3f0, + 0x00f89d21, +/* 0x0898: ctx_mem */ + 0x840007f1, + 0xd00203f0, + 0x04bd000f, +/* 0x08a4: ctx_mem_wait */ + 0x8400f7f1, + 0xcf02f3f0, + 0xfffd00ff, + 0xf31bf405, +/* 0x08b6: ctx_load */ + 0x94bd00f8, + 0xf10599f0, + 0xf00f0007, + 0x09d00203, + 0xf004bd00, + 0x21f40ca7, + 0xf1f4bdd0, + 0xf0890007, + 0x0fd00203, + 0xf104bd00, + 0xf0c10007, + 0x02d00203, + 0xf104bd00, + 0xf0830007, + 0x02d00203, + 0xf004bd00, + 0x21f507f7, + 0x07f10898, + 0x03f0c000, + 0x0002d002, + 0x0bfe04bd, + 0x1f2af000, + 0xb60424b6, + 0x94bd0220, + 0xf10899f0, + 0xf00f0007, + 0x09d00203, + 0xf104bd00, + 0xf0810007, + 0x02d00203, + 0xf104bd00, + 0xf1000027, + 0xf0800023, + 0x07f10225, + 0x03f08800, + 0x0002d002, + 0x17f004bd, + 0x0027f110, + 0x0223f002, + 0xf80512fa, + 0xf094bd03, + 0x07f10899, + 0x03f01700, + 0x0009d002, + 0x019804bd, + 0x1814b681, + 0xb6800298, + 0x12fd0825, + 0x16018005, + 0x99f094bd, + 0x0007f109, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f081, + 0xbd0001d0, + 0x0127f004, + 0x880007f1, + 0xd00203f0, + 0x04bd0002, + 0x010017f1, + 0xfa0613f0, + 0x03f80501, + 0x99f094bd, + 0x0007f109, + 0x0203f017, + 0xbd0009d0, + 0xf094bd04, + 0x07f10599, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x09d4: ctx_chan */ + 0x07db21f5, + 0x08b621f5, + 0xf40ca7f0, + 0xf7f0d021, + 0x9821f505, + 0x0021f508, +/* 0x09ef: ctx_mmio_exec */ + 0x9800f808, + 0x07f14103, + 0x03f08100, + 0x0003d002, + 0x34bd04bd, +/* 0x0a00: ctx_mmio_loop */ + 0xf4ff34c4, + 0x57f10f1b, + 0x53f00200, + 0x0535fa06, +/* 0x0a12: ctx_mmio_pull */ + 0x4e9803f8, + 0x814f9880, + 0xb69d21f4, + 0x12b60830, + 0xdf1bf401, +/* 0x0a24: ctx_mmio_done */ + 0xf1160398, + 0xf0810007, + 0x03d00203, + 0x8004bd00, + 0x17f14000, + 0x13f00100, + 0x0601fa06, + 0x00f803f8, +/* 0x0a44: ctx_xfer */ + 0xf104e7f0, + 0xf0020007, + 0x0ed00303, +/* 0x0a53: ctx_xfer_idle */ + 0xf104bd00, + 0xf00000e7, + 0xeecf03e3, + 0x00e4f100, + 0xf21bf420, + 0xf40611f4, +/* 0x0a6a: ctx_xfer_pre */ + 0xf7f01102, + 0x7021f510, + 0xdb21f508, + 0x1c11f407, +/* 0x0a78: ctx_xfer_pre_load */ + 0xf502f7f0, + 0xf5081121, + 0xf5082321, + 0xbd083821, + 0x1121f5f4, + 0xb621f508, +/* 0x0a91: ctx_xfer_exec */ + 0x16019808, + 0x07f124bd, + 0x03f00500, + 0x0002d001, + 0x1fb904bd, + 0x00e7f102, + 0x41e3f0a5, + 0xf09d21f4, + 0x2cf001fc, + 0x0124b602, + 0xb905f2fd, + 0xe7f102ff, + 0xe3f0a504, + 0x9d21f441, + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, + 0xf001acf0, + 0xb7f006a5, + 0x000c9800, + 0xf0010d98, + 0x21f500e7, + 0xa7f0016f, + 0x1021f508, + 0x5e21f501, + 0x1301f402, + 0xf40ca7f0, + 0xf7f0d021, + 0x9821f505, + 0x3202f408, +/* 0x0b20: ctx_xfer_post */ + 0xf502f7f0, + 0xbd081121, + 0x7021f5f4, + 0x7f21f508, + 0x2321f502, + 0xf5f4bd08, + 0xf4081121, + 0x01981011, + 0x0511fd40, + 0xf5070bf4, +/* 0x0b4b: ctx_xfer_no_post_mmio */ + 0xf509ef21, +/* 0x0b4f: ctx_xfer_done */ + 0xf8080021, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3 new file mode 100644 index 0000000000..581b2d53ab --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GF117 +#include "macros.fuc" + +.section #gf117_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gf117_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h new file mode 100644 index 0000000000..9b9f0d93f9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgf117.fuc3.h @@ -0,0 +1,1049 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gf117_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gf117_grhub_code[] = { + 0x039b0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0x07f100f8, + 0x03f00500, + 0x000fd002, + 0xf7f004bd, + 0x0007f101, + 0x0303f007, + 0xbd000fd0, +/* 0x039b: init */ + 0xbd00f804, + 0x0007fe04, + 0x420017f1, + 0xcf0013f0, + 0x11e70011, + 0x14b60109, + 0x0014fe08, + 0xf10227f0, + 0xf0120007, + 0x02d00003, + 0xf104bd00, + 0xfe06c817, + 0x24bd0010, + 0x070007f1, + 0xd00003f0, + 0x04bd0002, + 0x200327f1, + 0x010007f1, + 0xd00103f0, + 0x04bd0002, + 0x200427f1, + 0x010407f1, + 0xd00103f0, + 0x04bd0002, + 0x200b27f1, + 0x010807f1, + 0xd00103f0, + 0x04bd0002, + 0x200c27f1, + 0x011c07f1, + 0xd00103f0, + 0x04bd0002, + 0xf1010392, + 0xf0090007, + 0x03d00303, + 0xf104bd00, + 0xf0870427, + 0x07f10023, + 0x03f00400, + 0x0002d000, + 0x27f004bd, + 0x0007f104, + 0x0003f003, + 0xbd0002d0, + 0x1031f404, + 0x9604e7f1, + 0xf440e3f0, + 0xfeb96821, + 0x90f1c702, + 0xf0030180, + 0x0f801ff4, + 0x0117f002, + 0xb6041fbb, + 0x07f10112, + 0x03f00300, + 0x0001d001, + 0x07f104bd, + 0x03f00400, + 0x0001d001, + 0x17f104bd, + 0xf7f00100, + 0x1121f502, + 0x2321f508, + 0x10f7f008, + 0x087021f5, + 0x98000e98, + 0x21f5010f, + 0x14950150, + 0x0007f108, + 0x0103f0c0, + 0xbd0004d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0004d0, + 0x0030b704, + 0x001fbb13, + 0xf102f5b6, + 0xf0d30007, + 0x0fd00103, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0xf5021fb9, + 0xbb02d321, + 0x0398001f, + 0x0047f102, + 0x5043f020, +/* 0x04f4: init_gpc */ + 0x08044ea0, + 0xf4021fb9, + 0x4ea09d21, + 0xf4bd010c, + 0xa09d21f4, + 0xf401044e, + 0x4ea09d21, + 0xf7f00100, + 0x9d21f402, + 0x08004ea0, +/* 0x051c: init_gpc_wait */ + 0xc86821f4, + 0x0bf41fff, + 0x044ea0fa, + 0x6821f408, + 0xb7001fbb, + 0xb6800040, + 0x1bf40132, + 0x00f7f0be, + 0x087021f5, + 0xf500f7f0, + 0xf1081121, + 0xf0010007, + 0x01d00203, + 0xbd04bd00, + 0x1f19f014, + 0x080007f1, + 0xd00203f0, + 0x04bd0001, +/* 0x0564: wait */ + 0xf40028f4, +/* 0x056a: main */ + 0xd7f00031, + 0x3921f410, + 0xb1f401f4, + 0xf54001e4, + 0xbd00e91b, + 0x0499f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xc00017f1, + 0xcf0213f0, + 0x27f10011, + 0x23f0c100, + 0x0022cf02, + 0xf51f13c8, + 0xc800890b, + 0x0bf41f23, + 0xb920f962, + 0x94bd0212, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x31f40132, + 0x4421f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x20fc04bd, + 0x99f094bd, + 0x0007f106, + 0x0203f00f, + 0xbd0009d0, + 0x0131f404, + 0x0a4421f5, + 0x99f094bd, + 0x0007f106, + 0x0203f017, + 0xbd0009d0, + 0x330ef404, +/* 0x060c: chsw_prev_no_next */ + 0x12b920f9, + 0x0132f402, + 0xf50232f4, + 0xfc0a4421, + 0x0007f120, + 0x0203f0c0, + 0xbd0002d0, + 0x130ef404, +/* 0x062c: chsw_no_prev */ + 0xf41f23c8, + 0x31f40d0b, + 0x0232f401, + 0x0a4421f5, +/* 0x063c: chsw_done */ + 0xf10127f0, + 0xf0c30007, + 0x02d00203, + 0xbd04bd00, + 0x0499f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xff0e0ef5, +/* 0x0660: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b90d1b, + 0xd421f502, + 0x460ef409, +/* 0x0670: main_not_ctx_chan */ + 0xf402e4b0, + 0x94bd321b, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x32f40132, + 0x4421f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x0ef404bd, +/* 0x06a5: main_not_ctx_save */ + 0x10ef9411, + 0xf501f5f0, + 0xf5037e21, +/* 0x06b3: main_done */ + 0xbdfebb0e, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, + 0xfea60ef5, +/* 0x06c8: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0xf104bdf0, + 0xf00200a7, + 0xaacf00a3, + 0x04abc400, + 0xf0300bf4, + 0xe7f110d7, + 0xe3f01a00, + 0x00eecf00, + 0x1900f7f1, + 0xcf00f3f0, + 0x21f400ff, + 0x00b0b704, + 0x01e7f004, + 0x1d0007f1, + 0xd00003f0, + 0x04bd000e, +/* 0x071c: ih_no_fifo */ + 0x0100abe4, + 0xf00d0bf4, + 0xe7f110d7, + 0x21f44001, +/* 0x072d: ih_no_ctxsw */ + 0x00abe404, + 0x6c0bf404, + 0x0708e7f1, + 0xf440e3f0, + 0xffb96821, + 0x0007f102, + 0x0203f004, + 0xbd000fd0, + 0x04e7f104, + 0x40e3f007, + 0xb96821f4, + 0x07f102ff, + 0x03f00300, + 0x000fd002, + 0xfec704bd, + 0x02ee9450, + 0x0700f7f1, + 0xbb40f3f0, + 0x21f400ef, + 0x0007f168, + 0x0203f002, + 0xbd000fd0, + 0x03f7f004, + 0x037e21f5, + 0x0100b7f1, + 0xf102bfb9, + 0xf00144e7, + 0x21f440e3, +/* 0x079d: ih_no_fwmthd */ + 0x04b7f19d, + 0xffb0bd05, + 0x0bf4b4ab, + 0x0007f10f, + 0x0303f007, + 0xbd000bd0, +/* 0x07b5: ih_no_other */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x07db: ctx_4160s */ + 0xf001f800, + 0xffb901f7, + 0x60e7f102, + 0x40e3f041, +/* 0x07eb: ctx_4160s_wait */ + 0xf19d21f4, + 0xf04160e7, + 0x21f440e3, + 0x02ffb968, + 0xf404ffc8, + 0x00f8f00b, +/* 0x0800: ctx_4160c */ + 0xffb9f4bd, + 0x60e7f102, + 0x40e3f041, + 0xf89d21f4, +/* 0x0811: ctx_4170s */ + 0x10f5f000, + 0xf102ffb9, + 0xf04170e7, + 0x21f440e3, +/* 0x0823: ctx_4170w */ + 0xf100f89d, + 0xf04170e7, + 0x21f440e3, + 0x02ffb968, + 0xf410f4f0, + 0x00f8f01b, +/* 0x0838: ctx_redswitch */ + 0x0200e7f1, + 0xf040e5f0, + 0xe5f020e5, + 0x0007f110, + 0x0103f085, + 0xbd000ed0, + 0x08f7f004, +/* 0x0854: ctx_redswitch_delay */ + 0xf401f2b6, + 0xe5f1fd1b, + 0xe5f10400, + 0x07f10100, + 0x03f08500, + 0x000ed001, + 0x00f804bd, +/* 0x0870: ctx_86c */ + 0x1b0007f1, + 0xd00203f0, + 0x04bd000f, + 0xf102ffb9, + 0xf08a14e7, + 0x21f440e3, + 0x02ffb99d, + 0xa86ce7f1, + 0xf441e3f0, + 0x00f89d21, +/* 0x0898: ctx_mem */ + 0x840007f1, + 0xd00203f0, + 0x04bd000f, +/* 0x08a4: ctx_mem_wait */ + 0x8400f7f1, + 0xcf02f3f0, + 0xfffd00ff, + 0xf31bf405, +/* 0x08b6: ctx_load */ + 0x94bd00f8, + 0xf10599f0, + 0xf00f0007, + 0x09d00203, + 0xf004bd00, + 0x21f40ca7, + 0xf1f4bdd0, + 0xf0890007, + 0x0fd00203, + 0xf104bd00, + 0xf0c10007, + 0x02d00203, + 0xf104bd00, + 0xf0830007, + 0x02d00203, + 0xf004bd00, + 0x21f507f7, + 0x07f10898, + 0x03f0c000, + 0x0002d002, + 0x0bfe04bd, + 0x1f2af000, + 0xb60424b6, + 0x94bd0220, + 0xf10899f0, + 0xf00f0007, + 0x09d00203, + 0xf104bd00, + 0xf0810007, + 0x02d00203, + 0xf104bd00, + 0xf1000027, + 0xf0800023, + 0x07f10225, + 0x03f08800, + 0x0002d002, + 0x17f004bd, + 0x0027f110, + 0x0223f002, + 0xf80512fa, + 0xf094bd03, + 0x07f10899, + 0x03f01700, + 0x0009d002, + 0x019804bd, + 0x1814b681, + 0xb6800298, + 0x12fd0825, + 0x16018005, + 0x99f094bd, + 0x0007f109, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f081, + 0xbd0001d0, + 0x0127f004, + 0x880007f1, + 0xd00203f0, + 0x04bd0002, + 0x010017f1, + 0xfa0613f0, + 0x03f80501, + 0x99f094bd, + 0x0007f109, + 0x0203f017, + 0xbd0009d0, + 0xf094bd04, + 0x07f10599, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x09d4: ctx_chan */ + 0x07db21f5, + 0x08b621f5, + 0xf40ca7f0, + 0xf7f0d021, + 0x9821f505, + 0x0021f508, +/* 0x09ef: ctx_mmio_exec */ + 0x9800f808, + 0x07f14103, + 0x03f08100, + 0x0003d002, + 0x34bd04bd, +/* 0x0a00: ctx_mmio_loop */ + 0xf4ff34c4, + 0x57f10f1b, + 0x53f00200, + 0x0535fa06, +/* 0x0a12: ctx_mmio_pull */ + 0x4e9803f8, + 0x814f9880, + 0xb69d21f4, + 0x12b60830, + 0xdf1bf401, +/* 0x0a24: ctx_mmio_done */ + 0xf1160398, + 0xf0810007, + 0x03d00203, + 0x8004bd00, + 0x17f14000, + 0x13f00100, + 0x0601fa06, + 0x00f803f8, +/* 0x0a44: ctx_xfer */ + 0xf104e7f0, + 0xf0020007, + 0x0ed00303, +/* 0x0a53: ctx_xfer_idle */ + 0xf104bd00, + 0xf00000e7, + 0xeecf03e3, + 0x00e4f100, + 0xf21bf420, + 0xf40611f4, +/* 0x0a6a: ctx_xfer_pre */ + 0xf7f01102, + 0x7021f510, + 0xdb21f508, + 0x1c11f407, +/* 0x0a78: ctx_xfer_pre_load */ + 0xf502f7f0, + 0xf5081121, + 0xf5082321, + 0xbd083821, + 0x1121f5f4, + 0xb621f508, +/* 0x0a91: ctx_xfer_exec */ + 0x16019808, + 0x07f124bd, + 0x03f00500, + 0x0002d001, + 0x1fb904bd, + 0x00e7f102, + 0x41e3f0a5, + 0xf09d21f4, + 0x2cf001fc, + 0x0124b602, + 0xb905f2fd, + 0xe7f102ff, + 0xe3f0a504, + 0x9d21f441, + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, + 0xf001acf0, + 0xb7f006a5, + 0x000c9800, + 0xf0010d98, + 0x21f500e7, + 0xa7f0016f, + 0x1021f508, + 0x5e21f501, + 0x1301f402, + 0xf40ca7f0, + 0xf7f0d021, + 0x9821f505, + 0x3202f408, +/* 0x0b20: ctx_xfer_post */ + 0xf502f7f0, + 0xbd081121, + 0x7021f5f4, + 0x7f21f508, + 0x2321f502, + 0xf5f4bd08, + 0xf4081121, + 0x01981011, + 0x0511fd40, + 0xf5070bf4, +/* 0x0b4b: ctx_xfer_no_post_mmio */ + 0xf509ef21, +/* 0x0b4f: ctx_xfer_done */ + 0xf8080021, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3 new file mode 100644 index 0000000000..d977d393b6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GK100 +#include "macros.fuc" + +.section #gk104_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gk104_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h new file mode 100644 index 0000000000..fa11857b9d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk104.fuc3.h @@ -0,0 +1,1046 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk104_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gk104_grhub_code[] = { + 0x039b0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0x07f100f8, + 0x03f00500, + 0x000fd002, + 0xf7f004bd, + 0x0007f101, + 0x0303f007, + 0xbd000fd0, +/* 0x039b: init */ + 0xbd00f804, + 0x0007fe04, + 0x420017f1, + 0xcf0013f0, + 0x11e70011, + 0x14b60109, + 0x0014fe08, + 0xf10227f0, + 0xf0120007, + 0x02d00003, + 0xf104bd00, + 0xfe06c817, + 0x24bd0010, + 0x070007f1, + 0xd00003f0, + 0x04bd0002, + 0x200327f1, + 0x010007f1, + 0xd00103f0, + 0x04bd0002, + 0x200427f1, + 0x010407f1, + 0xd00103f0, + 0x04bd0002, + 0x200b27f1, + 0x010807f1, + 0xd00103f0, + 0x04bd0002, + 0x200c27f1, + 0x011c07f1, + 0xd00103f0, + 0x04bd0002, + 0xf1010392, + 0xf0090007, + 0x03d00303, + 0xf104bd00, + 0xf0870427, + 0x07f10023, + 0x03f00400, + 0x0002d000, + 0x27f004bd, + 0x0007f104, + 0x0003f003, + 0xbd0002d0, + 0x1031f404, + 0x9604e7f1, + 0xf440e3f0, + 0xfeb96821, + 0x90f1c702, + 0xf0030180, + 0x0f801ff4, + 0x0117f002, + 0xb6041fbb, + 0x07f10112, + 0x03f00300, + 0x0001d001, + 0x07f104bd, + 0x03f00400, + 0x0001d001, + 0x17f104bd, + 0xf7f00100, + 0xdb21f502, + 0xed21f507, + 0x10f7f007, + 0x083a21f5, + 0x98000e98, + 0x21f5010f, + 0x14950150, + 0x0007f108, + 0x0103f0c0, + 0xbd0004d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0004d0, + 0x0030b704, + 0x001fbb13, + 0xf102f5b6, + 0xf0d30007, + 0x0fd00103, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0xf5021fb9, + 0xbb02d321, + 0x0398001f, + 0x0047f102, + 0x5043f020, +/* 0x04f4: init_gpc */ + 0x08044ea0, + 0xf4021fb9, + 0x4ea09d21, + 0xf4bd010c, + 0xa09d21f4, + 0xf401044e, + 0x4ea09d21, + 0xf7f00100, + 0x9d21f402, + 0x08004ea0, +/* 0x051c: init_gpc_wait */ + 0xc86821f4, + 0x0bf41fff, + 0x044ea0fa, + 0x6821f408, + 0xb7001fbb, + 0xb6800040, + 0x1bf40132, + 0x00f7f0be, + 0x083a21f5, + 0xf500f7f0, + 0xf107db21, + 0xf0010007, + 0x01d00203, + 0xbd04bd00, + 0x1f19f014, + 0x080007f1, + 0xd00203f0, + 0x04bd0001, +/* 0x0564: wait */ + 0xf40028f4, +/* 0x056a: main */ + 0xd7f00031, + 0x3921f410, + 0xb1f401f4, + 0xf54001e4, + 0xbd00e91b, + 0x0499f094, + 0x0f0007f1, + 0xd00203f0, + 0x04bd0009, + 0xc00017f1, + 0xcf0213f0, + 0x27f10011, + 0x23f0c100, + 0x0022cf02, + 0xf51f13c8, + 0xc800890b, + 0x0bf41f23, + 0xb920f962, + 0x94bd0212, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x31f40132, + 0x0621f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x20fc04bd, + 0x99f094bd, + 0x0007f106, + 0x0203f00f, + 0xbd0009d0, + 0x0131f404, + 0x0a0621f5, + 0x99f094bd, + 0x0007f106, + 0x0203f017, + 0xbd0009d0, + 0x330ef404, +/* 0x060c: chsw_prev_no_next */ + 0x12b920f9, + 0x0132f402, + 0xf50232f4, + 0xfc0a0621, + 0x0007f120, + 0x0203f0c0, + 0xbd0002d0, + 0x130ef404, +/* 0x062c: chsw_no_prev */ + 0xf41f23c8, + 0x31f40d0b, + 0x0232f401, + 0x0a0621f5, +/* 0x063c: chsw_done */ + 0xf10127f0, + 0xf0c30007, + 0x02d00203, + 0xbd04bd00, + 0x0499f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xff0e0ef5, +/* 0x0660: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b90d1b, + 0x9e21f502, + 0x460ef409, +/* 0x0670: main_not_ctx_chan */ + 0xf402e4b0, + 0x94bd321b, + 0xf10799f0, + 0xf00f0007, + 0x09d00203, + 0xf404bd00, + 0x32f40132, + 0x0621f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x0ef404bd, +/* 0x06a5: main_not_ctx_save */ + 0x10ef9411, + 0xf501f5f0, + 0xf5037e21, +/* 0x06b3: main_done */ + 0xbdfebb0e, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, + 0xfea60ef5, +/* 0x06c8: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0xf104bdf0, + 0xf00200a7, + 0xaacf00a3, + 0x04abc400, + 0xf0300bf4, + 0xe7f110d7, + 0xe3f01a00, + 0x00eecf00, + 0x1900f7f1, + 0xcf00f3f0, + 0x21f400ff, + 0x00b0b704, + 0x01e7f004, + 0x1d0007f1, + 0xd00003f0, + 0x04bd000e, +/* 0x071c: ih_no_fifo */ + 0x0100abe4, + 0xf00d0bf4, + 0xe7f110d7, + 0x21f44001, +/* 0x072d: ih_no_ctxsw */ + 0x00abe404, + 0x6c0bf404, + 0x0708e7f1, + 0xf440e3f0, + 0xffb96821, + 0x0007f102, + 0x0203f004, + 0xbd000fd0, + 0x04e7f104, + 0x40e3f007, + 0xb96821f4, + 0x07f102ff, + 0x03f00300, + 0x000fd002, + 0xfec704bd, + 0x02ee9450, + 0x0700f7f1, + 0xbb40f3f0, + 0x21f400ef, + 0x0007f168, + 0x0203f002, + 0xbd000fd0, + 0x03f7f004, + 0x037e21f5, + 0x0100b7f1, + 0xf102bfb9, + 0xf00144e7, + 0x21f440e3, +/* 0x079d: ih_no_fwmthd */ + 0x04b7f19d, + 0xffb0bd05, + 0x0bf4b4ab, + 0x0007f10f, + 0x0303f007, + 0xbd000bd0, +/* 0x07b5: ih_no_other */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x07db: ctx_4170s */ + 0xf001f800, + 0xffb910f5, + 0x70e7f102, + 0x40e3f041, + 0xf89d21f4, +/* 0x07ed: ctx_4170w */ + 0x70e7f100, + 0x40e3f041, + 0xb96821f4, + 0xf4f002ff, + 0xf01bf410, +/* 0x0802: ctx_redswitch */ + 0xe7f100f8, + 0xe5f00200, + 0x20e5f040, + 0xf110e5f0, + 0xf0850007, + 0x0ed00103, + 0xf004bd00, +/* 0x081e: ctx_redswitch_delay */ + 0xf2b608f7, + 0xfd1bf401, + 0x0400e5f1, + 0x0100e5f1, + 0x850007f1, + 0xd00103f0, + 0x04bd000e, +/* 0x083a: ctx_86c */ + 0x07f100f8, + 0x03f01b00, + 0x000fd002, + 0xffb904bd, + 0x14e7f102, + 0x40e3f08a, + 0xb99d21f4, + 0xe7f102ff, + 0xe3f0a86c, + 0x9d21f441, +/* 0x0862: ctx_mem */ + 0x07f100f8, + 0x03f08400, + 0x000fd002, +/* 0x086e: ctx_mem_wait */ + 0xf7f104bd, + 0xf3f08400, + 0x00ffcf02, + 0xf405fffd, + 0x00f8f31b, +/* 0x0880: ctx_load */ + 0x99f094bd, + 0x0007f105, + 0x0203f00f, + 0xbd0009d0, + 0x0ca7f004, + 0xbdd021f4, + 0x0007f1f4, + 0x0203f089, + 0xbd000fd0, + 0x0007f104, + 0x0203f0c1, + 0xbd0002d0, + 0x0007f104, + 0x0203f083, + 0xbd0002d0, + 0x07f7f004, + 0x086221f5, + 0xc00007f1, + 0xd00203f0, + 0x04bd0002, + 0xf0000bfe, + 0x24b61f2a, + 0x0220b604, + 0x99f094bd, + 0x0007f108, + 0x0203f00f, + 0xbd0009d0, + 0x0007f104, + 0x0203f081, + 0xbd0002d0, + 0x0027f104, + 0x0023f100, + 0x0225f080, + 0x880007f1, + 0xd00203f0, + 0x04bd0002, + 0xf11017f0, + 0xf0020027, + 0x12fa0223, + 0xbd03f805, + 0x0899f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xb6810198, + 0x02981814, + 0x0825b680, + 0x800512fd, + 0x94bd1601, + 0xf10999f0, + 0xf00f0007, + 0x09d00203, + 0xf104bd00, + 0xf0810007, + 0x01d00203, + 0xf004bd00, + 0x07f10127, + 0x03f08800, + 0x0002d002, + 0x17f104bd, + 0x13f00100, + 0x0501fa06, + 0x94bd03f8, + 0xf10999f0, + 0xf0170007, + 0x09d00203, + 0xbd04bd00, + 0x0599f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x099e: ctx_chan */ + 0x21f500f8, + 0xa7f00880, + 0xd021f40c, + 0xf505f7f0, + 0xf8086221, +/* 0x09b1: ctx_mmio_exec */ + 0x41039800, + 0x810007f1, + 0xd00203f0, + 0x04bd0003, +/* 0x09c2: ctx_mmio_loop */ + 0x34c434bd, + 0x0f1bf4ff, + 0x020057f1, + 0xfa0653f0, + 0x03f80535, +/* 0x09d4: ctx_mmio_pull */ + 0x98804e98, + 0x21f4814f, + 0x0830b69d, + 0xf40112b6, +/* 0x09e6: ctx_mmio_done */ + 0x0398df1b, + 0x0007f116, + 0x0203f081, + 0xbd0003d0, + 0x40008004, + 0x010017f1, + 0xfa0613f0, + 0x03f80601, +/* 0x0a06: ctx_xfer */ + 0xe7f000f8, + 0x0007f104, + 0x0303f002, + 0xbd000ed0, +/* 0x0a15: ctx_xfer_idle */ + 0x00e7f104, + 0x03e3f000, + 0xf100eecf, + 0xf42000e4, + 0x11f4f21b, + 0x0d02f406, +/* 0x0a2c: ctx_xfer_pre */ + 0xf510f7f0, + 0xf4083a21, +/* 0x0a36: ctx_xfer_pre_load */ + 0xf7f01c11, + 0xdb21f502, + 0xed21f507, + 0x0221f507, + 0xf5f4bd08, + 0xf507db21, +/* 0x0a4f: ctx_xfer_exec */ + 0x98088021, + 0x24bd1601, + 0x050007f1, + 0xd00103f0, + 0x04bd0002, + 0xf1021fb9, + 0xf0a500e7, + 0x21f441e3, + 0x01fcf09d, + 0xb6022cf0, + 0xf2fd0124, + 0x02ffb905, + 0xa504e7f1, + 0xf441e3f0, + 0x21f59d21, + 0x24bd026a, + 0x47fc07f1, + 0xd00203f0, + 0x04bd0002, + 0xb6012cf0, + 0x07f10320, + 0x03f04afc, + 0x0002d002, + 0xacf004bd, + 0x06a5f001, + 0x9800b7f0, + 0x0d98000c, + 0x00e7f001, + 0x016f21f5, + 0xf508a7f0, + 0xf5011021, + 0xf4025e21, + 0xa7f01301, + 0xd021f40c, + 0xf505f7f0, + 0xf4086221, +/* 0x0ade: ctx_xfer_post */ + 0xf7f02e02, + 0xdb21f502, + 0xf5f4bd07, + 0xf5083a21, + 0xf5027f21, + 0xbd07ed21, + 0xdb21f5f4, + 0x1011f407, + 0xfd400198, + 0x0bf40511, + 0xb121f507, +/* 0x0b09: ctx_xfer_no_post_mmio */ +/* 0x0b09: ctx_xfer_done */ + 0x0000f809, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3 new file mode 100644 index 0000000000..760b4632f2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GK110 +#include "macros.fuc" + +.section #gk110_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gk110_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h new file mode 100644 index 0000000000..1d741b30a0 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk110.fuc3.h @@ -0,0 +1,1046 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk110_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gk110_grhub_code[] = { + 0x039b0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0x0489b808, + 0xf00c1bf4, + 0x21f502f7, + 0x00f8037e, +/* 0x001c: queue_put_next */ + 0xb60798c4, + 0x8dbb0384, + 0x0880b600, + 0x80008e80, + 0x90b6018f, + 0x0f94f001, + 0xf801d980, +/* 0x0039: queue_get */ + 0x0131f400, + 0x9800d898, + 0x89b801d9, + 0x210bf404, + 0xb60789c4, + 0x9dbb0394, + 0x0890b600, + 0x98009e98, + 0x80b6019f, + 0x0f84f001, + 0xf400d880, +/* 0x0066: queue_get_done */ + 0x00f80132, +/* 0x0068: nv_rd32 */ + 0xf002ecb9, + 0x07f11fc9, + 0x03f0ca00, + 0x000cd001, +/* 0x007a: nv_rd32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0xa7f0f31b, + 0x1021f506, + 0x00f7f101, + 0x01f3f0cb, + 0xf800ffcf, +/* 0x009d: nv_wr32 */ + 0x0007f100, + 0x0103f0cc, + 0xbd000fd0, + 0x02ecb904, + 0xf01fc9f0, + 0x07f11ec9, + 0x03f0ca00, + 0x000cd001, +/* 0x00be: nv_wr32_wait */ + 0xc7f104bd, + 0xc3f0ca00, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f31b, +/* 0x00d0: wait_donez */ + 0x99f094bd, + 0x0007f100, + 0x0203f037, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x00ed: wait_donez_ne */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x1bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0110: wait_doneo */ + 0x99f094bd, + 0x0007f100, + 0x0203f037, + 0xbd0009d0, + 0x0007f104, + 0x0203f006, + 0xbd000ad0, +/* 0x012d: wait_doneo_e */ + 0x0087f104, + 0x0183f000, + 0xff0088cf, + 0x0bf4888a, + 0xf094bdf3, + 0x07f10099, + 0x03f01700, + 0x0009d002, + 0x00f804bd, +/* 0x0150: mmctx_size */ +/* 0x0152: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0xf404efb8, + 0x9fb9eb1b, +/* 0x016f: mmctx_xfer */ + 0xbd00f802, + 0x0199f094, + 0x370007f1, + 0xd00203f0, + 0x04bd0009, + 0xbbfd94bd, + 0x120bf405, + 0xc40007f1, + 0xd00103f0, + 0x04bd000b, +/* 0x0197: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0x0007f11e, + 0x0103f0c6, + 0xbd000ed0, + 0x0007f104, + 0x0103f0c7, + 0xbd000fd0, + 0x0199f004, +/* 0x01b8: mmctx_multi_disabled */ + 0xb600abc8, + 0xb9f010b4, + 0x01aec80c, + 0xfd11e4b6, + 0x07f105be, + 0x03f0c500, + 0x000bd001, +/* 0x01d6: mmctx_exec_loop */ +/* 0x01d6: mmctx_wait_free */ + 0xe7f104bd, + 0xe3f0c500, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f30b, + 0x05e9fd00, + 0xc80007f1, + 0xd00103f0, + 0x04bd000e, + 0xb804c0b6, + 0x1bf404cd, + 0x02abc8d8, +/* 0x0207: mmctx_fini_wait */ + 0xf11f1bf4, + 0xf0c500b7, + 0xbbcf01b3, + 0x1fb4f000, + 0xf410b4b0, + 0xa7f0f01b, + 0xd021f405, +/* 0x0223: mmctx_stop */ + 0xc82b0ef4, + 0xb4b600ab, + 0x0cb9f010, + 0xf112b9f0, + 0xf0c50007, + 0x0bd00103, +/* 0x023b: mmctx_stop_wait */ + 0xf104bd00, + 0xf0c500b7, + 0xbbcf01b3, + 0x12bbc800, +/* 0x024b: mmctx_done */ + 0xbdf31bf4, + 0x0199f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x025e: strand_wait */ + 0xa0f900f8, + 0xf402a7f0, + 0xa0fcd021, +/* 0x026a: strand_pre */ + 0x97f000f8, + 0xfc07f10c, + 0x0203f04a, + 0xbd0009d0, + 0x5e21f504, +/* 0x027f: strand_post */ + 0xf000f802, + 0x07f10d97, + 0x03f04afc, + 0x0009d002, + 0x21f504bd, + 0x00f8025e, +/* 0x0294: strand_set */ + 0xf10fc7f0, + 0xf04ffc07, + 0x0cd00203, + 0xf004bd00, + 0x07f10bc7, + 0x03f04afc, + 0x000cd002, + 0x07f104bd, + 0x03f04ffc, + 0x000ed002, + 0xc7f004bd, + 0xfc07f10a, + 0x0203f04a, + 0xbd000cd0, + 0x5e21f504, +/* 0x02d3: strand_ctx_init */ + 0xbd00f802, + 0x0399f094, + 0x370007f1, + 0xd00203f0, + 0x04bd0009, + 0x026a21f5, + 0xf503e7f0, + 0xbd029421, + 0xfc07f1c4, + 0x0203f047, + 0xbd000cd0, + 0x01c7f004, + 0x4afc07f1, + 0xd00203f0, + 0x04bd000c, + 0x025e21f5, + 0xf1010c92, + 0xf046fc07, + 0x0cd00203, + 0xf004bd00, + 0x07f102c7, + 0x03f04afc, + 0x000cd002, + 0x21f504bd, + 0x21f5025e, + 0x87f1027f, + 0x83f04200, + 0x0097f102, + 0x0293f020, + 0x950099cf, +/* 0x034a: ctx_init_strand_loop */ + 0x8ed008fe, + 0x408ed000, + 0xb6808acf, + 0xa0b606a5, + 0x00eabb01, + 0xb60480b6, + 0x1bf40192, + 0x08e4b6e8, + 0xbdf2efbc, + 0x0399f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x037e: error */ + 0x07f100f8, + 0x03f00500, + 0x000fd002, + 0xf7f004bd, + 0x0007f101, + 0x0303f007, + 0xbd000fd0, +/* 0x039b: init */ + 0xbd00f804, + 0x0007fe04, + 0x420017f1, + 0xcf0013f0, + 0x11e70011, + 0x14b60109, + 0x0014fe08, + 0xf10227f0, + 0xf0120007, + 0x02d00003, + 0xf104bd00, + 0xfe06c817, + 0x24bd0010, + 0x070007f1, + 0xd00003f0, + 0x04bd0002, + 0x200327f1, + 0x010007f1, + 0xd00103f0, + 0x04bd0002, + 0x200427f1, + 0x010407f1, + 0xd00103f0, + 0x04bd0002, + 0x200b27f1, + 0x010807f1, + 0xd00103f0, + 0x04bd0002, + 0x200c27f1, + 0x011c07f1, + 0xd00103f0, + 0x04bd0002, + 0xf1010392, + 0xf0090007, + 0x03d00303, + 0xf104bd00, + 0xf0870427, + 0x07f10023, + 0x03f00400, + 0x0002d000, + 0x27f004bd, + 0x0007f104, + 0x0003f003, + 0xbd0002d0, + 0x1031f404, + 0x9604e7f1, + 0xf440e3f0, + 0xfeb96821, + 0x90f1c702, + 0xf0030180, + 0x0f801ff4, + 0x0117f002, + 0xb6041fbb, + 0x07f10112, + 0x03f00300, + 0x0001d001, + 0x07f104bd, + 0x03f00400, + 0x0001d001, + 0x17f104bd, + 0xf7f00100, + 0xdb21f502, + 0xed21f507, + 0x10f7f007, + 0x083a21f5, + 0x98000e98, + 0x21f5010f, + 0x14950150, + 0x0007f108, + 0x0103f0c0, + 0xbd0004d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0004d0, + 0x0030b704, + 0x001fbb13, + 0xf102f5b6, + 0xf0d30007, + 0x0fd00103, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0xf5021fb9, + 0xbb02d321, + 0x0398001f, + 0x0047f102, + 0x5043f020, +/* 0x04f4: init_gpc */ + 0x08044ea0, + 0xf4021fb9, + 0x4ea09d21, + 0xf4bd010c, + 0xa09d21f4, + 0xf401044e, + 0x4ea09d21, + 0xf7f00100, + 0x9d21f402, + 0x08004ea0, +/* 0x051c: init_gpc_wait */ + 0xc86821f4, + 0x0bf41fff, + 0x044ea0fa, + 0x6821f408, + 0xb7001fbb, + 0xb6800040, + 0x1bf40132, + 0x00f7f0be, + 0x083a21f5, + 0xf500f7f0, + 0xf107db21, + 0xf0010007, + 0x01d00203, + 0xbd04bd00, + 0x1f19f014, + 0x300007f1, + 0xd00203f0, + 0x04bd0001, +/* 0x0564: wait */ + 0xf40028f4, +/* 0x056a: main */ + 0xd7f00031, + 0x3921f410, + 0xb1f401f4, + 0xf54001e4, + 0xbd00e91b, + 0x0499f094, + 0x370007f1, + 0xd00203f0, + 0x04bd0009, + 0xc00017f1, + 0xcf0213f0, + 0x27f10011, + 0x23f0c100, + 0x0022cf02, + 0xf51f13c8, + 0xc800890b, + 0x0bf41f23, + 0xb920f962, + 0x94bd0212, + 0xf10799f0, + 0xf0370007, + 0x09d00203, + 0xf404bd00, + 0x31f40132, + 0x0621f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x20fc04bd, + 0x99f094bd, + 0x0007f106, + 0x0203f037, + 0xbd0009d0, + 0x0131f404, + 0x0a0621f5, + 0x99f094bd, + 0x0007f106, + 0x0203f017, + 0xbd0009d0, + 0x330ef404, +/* 0x060c: chsw_prev_no_next */ + 0x12b920f9, + 0x0132f402, + 0xf50232f4, + 0xfc0a0621, + 0x0007f120, + 0x0203f0c0, + 0xbd0002d0, + 0x130ef404, +/* 0x062c: chsw_no_prev */ + 0xf41f23c8, + 0x31f40d0b, + 0x0232f401, + 0x0a0621f5, +/* 0x063c: chsw_done */ + 0xf10127f0, + 0xf0c30007, + 0x02d00203, + 0xbd04bd00, + 0x0499f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xff0e0ef5, +/* 0x0660: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b90d1b, + 0x9e21f502, + 0x460ef409, +/* 0x0670: main_not_ctx_chan */ + 0xf402e4b0, + 0x94bd321b, + 0xf10799f0, + 0xf0370007, + 0x09d00203, + 0xf404bd00, + 0x32f40132, + 0x0621f502, + 0xf094bd0a, + 0x07f10799, + 0x03f01700, + 0x0009d002, + 0x0ef404bd, +/* 0x06a5: main_not_ctx_save */ + 0x10ef9411, + 0xf501f5f0, + 0xf5037e21, +/* 0x06b3: main_done */ + 0xbdfebb0e, + 0x1f29f024, + 0x300007f1, + 0xd00203f0, + 0x04bd0002, + 0xfea60ef5, +/* 0x06c8: ih */ + 0x80f900f9, + 0xf90188fe, + 0xf990f980, + 0xf9b0f9a0, + 0xf9e0f9d0, + 0xf104bdf0, + 0xf00200a7, + 0xaacf00a3, + 0x04abc400, + 0xf0300bf4, + 0xe7f110d7, + 0xe3f01a00, + 0x00eecf00, + 0x1900f7f1, + 0xcf00f3f0, + 0x21f400ff, + 0x00b0b704, + 0x01e7f004, + 0x1d0007f1, + 0xd00003f0, + 0x04bd000e, +/* 0x071c: ih_no_fifo */ + 0x0100abe4, + 0xf00d0bf4, + 0xe7f110d7, + 0x21f44001, +/* 0x072d: ih_no_ctxsw */ + 0x00abe404, + 0x6c0bf404, + 0x0708e7f1, + 0xf440e3f0, + 0xffb96821, + 0x0007f102, + 0x0203f004, + 0xbd000fd0, + 0x04e7f104, + 0x40e3f007, + 0xb96821f4, + 0x07f102ff, + 0x03f00300, + 0x000fd002, + 0xfec704bd, + 0x02ee9450, + 0x0700f7f1, + 0xbb40f3f0, + 0x21f400ef, + 0x0007f168, + 0x0203f002, + 0xbd000fd0, + 0x03f7f004, + 0x037e21f5, + 0x0100b7f1, + 0xf102bfb9, + 0xf00144e7, + 0x21f440e3, +/* 0x079d: ih_no_fwmthd */ + 0x04b7f19d, + 0xffb0bd05, + 0x0bf4b4ab, + 0x0007f10f, + 0x0303f007, + 0xbd000bd0, +/* 0x07b5: ih_no_other */ + 0x0007f104, + 0x0003f001, + 0xbd000ad0, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0x32f400fc, +/* 0x07db: ctx_4170s */ + 0xf001f800, + 0xffb910f5, + 0x70e7f102, + 0x40e3f041, + 0xf89d21f4, +/* 0x07ed: ctx_4170w */ + 0x70e7f100, + 0x40e3f041, + 0xb96821f4, + 0xf4f002ff, + 0xf01bf410, +/* 0x0802: ctx_redswitch */ + 0xe7f100f8, + 0xe5f00200, + 0x20e5f040, + 0xf110e5f0, + 0xf0850007, + 0x0ed00103, + 0xf004bd00, +/* 0x081e: ctx_redswitch_delay */ + 0xf2b608f7, + 0xfd1bf401, + 0x0400e5f1, + 0x0100e5f1, + 0x850007f1, + 0xd00103f0, + 0x04bd000e, +/* 0x083a: ctx_86c */ + 0x07f100f8, + 0x03f02300, + 0x000fd002, + 0xffb904bd, + 0x14e7f102, + 0x40e3f08a, + 0xb99d21f4, + 0xe7f102ff, + 0xe3f0a88c, + 0x9d21f441, +/* 0x0862: ctx_mem */ + 0x07f100f8, + 0x03f08400, + 0x000fd002, +/* 0x086e: ctx_mem_wait */ + 0xf7f104bd, + 0xf3f08400, + 0x00ffcf02, + 0xf405fffd, + 0x00f8f31b, +/* 0x0880: ctx_load */ + 0x99f094bd, + 0x0007f105, + 0x0203f037, + 0xbd0009d0, + 0x0ca7f004, + 0xbdd021f4, + 0x0007f1f4, + 0x0203f089, + 0xbd000fd0, + 0x0007f104, + 0x0203f0c1, + 0xbd0002d0, + 0x0007f104, + 0x0203f083, + 0xbd0002d0, + 0x07f7f004, + 0x086221f5, + 0xc00007f1, + 0xd00203f0, + 0x04bd0002, + 0xf0000bfe, + 0x24b61f2a, + 0x0220b604, + 0x99f094bd, + 0x0007f108, + 0x0203f037, + 0xbd0009d0, + 0x0007f104, + 0x0203f081, + 0xbd0002d0, + 0x0027f104, + 0x0023f100, + 0x0225f080, + 0x880007f1, + 0xd00203f0, + 0x04bd0002, + 0xf11017f0, + 0xf0020027, + 0x12fa0223, + 0xbd03f805, + 0x0899f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, + 0xb6810198, + 0x02981814, + 0x0825b680, + 0x800512fd, + 0x94bd1601, + 0xf10999f0, + 0xf0370007, + 0x09d00203, + 0xf104bd00, + 0xf0810007, + 0x01d00203, + 0xf004bd00, + 0x07f10127, + 0x03f08800, + 0x0002d002, + 0x17f104bd, + 0x13f00100, + 0x0501fa06, + 0x94bd03f8, + 0xf10999f0, + 0xf0170007, + 0x09d00203, + 0xbd04bd00, + 0x0599f094, + 0x170007f1, + 0xd00203f0, + 0x04bd0009, +/* 0x099e: ctx_chan */ + 0x21f500f8, + 0xa7f00880, + 0xd021f40c, + 0xf505f7f0, + 0xf8086221, +/* 0x09b1: ctx_mmio_exec */ + 0x41039800, + 0x810007f1, + 0xd00203f0, + 0x04bd0003, +/* 0x09c2: ctx_mmio_loop */ + 0x34c434bd, + 0x0f1bf4ff, + 0x020057f1, + 0xfa0653f0, + 0x03f80535, +/* 0x09d4: ctx_mmio_pull */ + 0x98804e98, + 0x21f4814f, + 0x0830b69d, + 0xf40112b6, +/* 0x09e6: ctx_mmio_done */ + 0x0398df1b, + 0x0007f116, + 0x0203f081, + 0xbd0003d0, + 0x40008004, + 0x010017f1, + 0xfa0613f0, + 0x03f80601, +/* 0x0a06: ctx_xfer */ + 0xe7f000f8, + 0x0007f104, + 0x0303f002, + 0xbd000ed0, +/* 0x0a15: ctx_xfer_idle */ + 0x00e7f104, + 0x03e3f000, + 0xf100eecf, + 0xf42000e4, + 0x11f4f21b, + 0x0d02f406, +/* 0x0a2c: ctx_xfer_pre */ + 0xf510f7f0, + 0xf4083a21, +/* 0x0a36: ctx_xfer_pre_load */ + 0xf7f01c11, + 0xdb21f502, + 0xed21f507, + 0x0221f507, + 0xf5f4bd08, + 0xf507db21, +/* 0x0a4f: ctx_xfer_exec */ + 0x98088021, + 0x24bd1601, + 0x050007f1, + 0xd00103f0, + 0x04bd0002, + 0xf1021fb9, + 0xf0a500e7, + 0x21f441e3, + 0x01fcf09d, + 0xb6022cf0, + 0xf2fd0124, + 0x02ffb905, + 0xa504e7f1, + 0xf441e3f0, + 0x21f59d21, + 0x24bd026a, + 0x47fc07f1, + 0xd00203f0, + 0x04bd0002, + 0xb6012cf0, + 0x07f10320, + 0x03f04afc, + 0x0002d002, + 0xacf004bd, + 0x06a5f001, + 0x9800b7f0, + 0x0d98000c, + 0x00e7f001, + 0x016f21f5, + 0xf508a7f0, + 0xf5011021, + 0xf4025e21, + 0xa7f01301, + 0xd021f40c, + 0xf505f7f0, + 0xf4086221, +/* 0x0ade: ctx_xfer_post */ + 0xf7f02e02, + 0xdb21f502, + 0xf5f4bd07, + 0xf5083a21, + 0xf5027f21, + 0xbd07ed21, + 0xdb21f5f4, + 0x1011f407, + 0xfd400198, + 0x0bf40511, + 0xb121f507, +/* 0x0b09: ctx_xfer_no_post_mmio */ +/* 0x0b09: ctx_xfer_done */ + 0x0000f809, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5 new file mode 100644 index 0000000000..43243a35f6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GK208 +#include "macros.fuc" + +.section #gk208_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gk208_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h new file mode 100644 index 0000000000..ae2d5b6891 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgk208.fuc5.h @@ -0,0 +1,918 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gk208_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gk208_grhub_code[] = { + 0x030e0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0xf489a408, + 0x020f0b1b, + 0x0002f87e, +/* 0x001a: queue_put_next */ + 0x98c400f8, + 0x0384b607, + 0xb6008dbb, + 0x8eb50880, + 0x018fb500, + 0xf00190b6, + 0xd9b50f94, +/* 0x0037: queue_get */ + 0xf400f801, + 0xd8980131, + 0x01d99800, + 0x0bf489a4, + 0x0789c421, + 0xbb0394b6, + 0x90b6009d, + 0x009e9808, + 0xb6019f98, + 0x84f00180, + 0x00d8b50f, +/* 0x0063: queue_get_done */ + 0xf80132f4, +/* 0x0065: nv_rd32 */ + 0xf0ecb200, + 0x00801fc9, + 0x0cf601ca, +/* 0x0073: nv_rd32_wait */ + 0x8c04bd00, + 0xcf01ca00, + 0xccc800cc, + 0xf61bf41f, + 0xec7e060a, + 0x008f0000, + 0xffcf01cb, +/* 0x008f: nv_wr32 */ + 0x8000f800, + 0xf601cc00, + 0x04bd000f, + 0xc9f0ecb2, + 0x1ec9f01f, + 0x01ca0080, + 0xbd000cf6, +/* 0x00a9: nv_wr32_wait */ + 0xca008c04, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f61b, +/* 0x00b8: wait_donez */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x00cf: wait_donez_ne */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf61bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x00ec: wait_doneo */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x0103: wait_doneo_e */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf60bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x0120: mmctx_size */ +/* 0x0122: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0x1bf4efa4, + 0xf89fb2ec, +/* 0x013d: mmctx_xfer */ + 0xf094bd00, + 0x00800199, + 0x09f60237, + 0xbd04bd00, + 0x05bbfd94, + 0x800f0bf4, + 0xf601c400, + 0x04bd000b, +/* 0x015f: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0xc6008018, + 0x000ef601, + 0x008004bd, + 0x0ff601c7, + 0xf004bd00, +/* 0x017a: mmctx_multi_disabled */ + 0xabc80199, + 0x10b4b600, + 0xc80cb9f0, + 0xe4b601ae, + 0x05befd11, + 0x01c50080, + 0xbd000bf6, +/* 0x0195: mmctx_exec_loop */ +/* 0x0195: mmctx_wait_free */ + 0xc5008e04, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f60b, + 0x05e9fd00, + 0x01c80080, + 0xbd000ef6, + 0x04c0b604, + 0x1bf4cda4, + 0x02abc8df, +/* 0x01bf: mmctx_fini_wait */ + 0x8b1c1bf4, + 0xcf01c500, + 0xb4f000bb, + 0x10b4b01f, + 0x0af31bf4, + 0x00b87e05, + 0x250ef400, +/* 0x01d8: mmctx_stop */ + 0xb600abc8, + 0xb9f010b4, + 0x12b9f00c, + 0x01c50080, + 0xbd000bf6, +/* 0x01ed: mmctx_stop_wait */ + 0xc5008b04, + 0x00bbcf01, + 0xf412bbc8, +/* 0x01fa: mmctx_done */ + 0x94bdf61b, + 0x800199f0, + 0xf6021700, + 0x04bd0009, +/* 0x020a: strand_wait */ + 0xa0f900f8, + 0xb87e020a, + 0xa0fc0000, +/* 0x0216: strand_pre */ + 0x0c0900f8, + 0x024afc80, + 0xbd0009f6, + 0x020a7e04, +/* 0x0227: strand_post */ + 0x0900f800, + 0x4afc800d, + 0x0009f602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0238: strand_set */ + 0xfc800f0c, + 0x0cf6024f, + 0x0c04bd00, + 0x4afc800b, + 0x000cf602, + 0xfc8004bd, + 0x0ef6024f, + 0x0c04bd00, + 0x4afc800a, + 0x000cf602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0268: strand_ctx_init */ + 0x99f094bd, + 0x37008003, + 0x0009f602, + 0x167e04bd, + 0x030e0002, + 0x0002387e, + 0xfc80c4bd, + 0x0cf60247, + 0x0c04bd00, + 0x4afc8001, + 0x000cf602, + 0x0a7e04bd, + 0x0c920002, + 0x46fc8001, + 0x000cf602, + 0x020c04bd, + 0x024afc80, + 0xbd000cf6, + 0x020a7e04, + 0x02277e00, + 0x42008800, + 0x20008902, + 0x0099cf02, +/* 0x02c7: ctx_init_strand_loop */ + 0xf608fe95, + 0x8ef6008e, + 0x808acf40, + 0xb606a5b6, + 0xeabb01a0, + 0x0480b600, + 0xf40192b6, + 0xe4b6e81b, + 0xf2efbc08, + 0x99f094bd, + 0x17008003, + 0x0009f602, + 0x00f804bd, +/* 0x02f8: error */ + 0x02050080, + 0xbd000ff6, + 0x80010f04, + 0xf6030700, + 0x04bd000f, +/* 0x030e: init */ + 0x04bd00f8, + 0x410007fe, + 0x11cf4200, + 0x0911e700, + 0x0814b601, + 0x020014fe, + 0x12004002, + 0xbd0002f6, + 0x05ca4104, + 0xbd0010fe, + 0x07004024, + 0xbd0002f6, + 0x20034204, + 0x01010080, + 0xbd0002f6, + 0x20044204, + 0x01010480, + 0xbd0002f6, + 0x200b4204, + 0x01010880, + 0xbd0002f6, + 0x200c4204, + 0x01011c80, + 0xbd0002f6, + 0x01039204, + 0x03090080, + 0xbd0003f6, + 0x87048204, + 0x04004000, + 0xbd0002f6, + 0x40040204, + 0x02f60300, + 0xf404bd00, + 0x048e1031, + 0x657e4096, + 0xfeb20000, + 0xb590f1c7, + 0xf4f00301, + 0x020fb51f, + 0x1fbb0101, + 0x0112b604, + 0x01030080, + 0xbd0001f6, + 0x04008004, + 0x0001f601, + 0x004104bd, + 0x7e020f01, + 0x7e0006ad, + 0x0f0006bc, + 0x06fe7e10, + 0x000e9800, + 0x7e010f98, + 0x95000120, + 0x00800814, + 0x04f601c0, + 0x8004bd00, + 0xf601c100, + 0x04bd0004, + 0x130030b7, + 0xb6001fbb, + 0x008002f5, + 0x0ff601d3, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0x687e1fb2, + 0x1fbb0002, + 0x02039800, + 0x50200084, +/* 0x0420: init_gpc */ + 0x08044eb8, + 0x7e1fb200, + 0xb800008f, + 0x00010c4e, + 0x8f7ef4bd, + 0x4eb80000, + 0x7e000104, + 0xb800008f, + 0x0001004e, + 0x8f7e020f, + 0x4eb80000, +/* 0x044f: init_gpc_wait */ + 0x7e000800, + 0xc8000065, + 0x0bf41fff, + 0x044eb8f9, + 0x657e0008, + 0x1fbb0000, + 0x0040b700, + 0x0132b680, + 0x0fb41bf4, + 0x06fe7e00, + 0x7e000f00, + 0x800006ad, + 0xf6020100, + 0x04bd0001, + 0x19f014bd, + 0x3000801f, + 0x0001f602, +/* 0x0492: wait */ + 0x28f404bd, + 0x0031f400, +/* 0x0498: main */ + 0x377e100d, + 0x01f40000, + 0x01e4b1f4, + 0xc71bf540, + 0xf094bd00, + 0x00800499, + 0x09f60237, + 0x8104bd00, + 0xcf02c000, + 0x00820011, + 0x22cf02c1, + 0x1f13c800, + 0xc8770bf4, + 0x0bf41f23, + 0xb220f955, + 0xf094bd12, + 0x00800799, + 0x09f60237, + 0xf404bd00, + 0x31f40132, + 0x08817e02, + 0xf094bd00, + 0x00800799, + 0x09f60217, + 0xfc04bd00, + 0xf094bd20, + 0x00800699, + 0x09f60237, + 0xf404bd00, + 0x817e0131, + 0x94bd0008, + 0x800699f0, + 0xf6021700, + 0x04bd0009, +/* 0x0523: chsw_prev_no_next */ + 0xf92f0ef4, + 0xf412b220, + 0x32f40132, + 0x08817e02, + 0x8020fc00, + 0xf602c000, + 0x04bd0002, +/* 0x053f: chsw_no_prev */ + 0xc8130ef4, + 0x0bf41f23, + 0x0131f40d, + 0x7e0232f4, +/* 0x054f: chsw_done */ + 0x02000881, + 0xc3008001, + 0x0002f602, + 0x94bd04bd, + 0x800499f0, + 0xf6021700, + 0x04bd0009, + 0xff300ef5, +/* 0x056c: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b20c1b, + 0x0008217e, +/* 0x057b: main_not_ctx_chan */ + 0xb0400ef4, + 0x1bf402e4, + 0xf094bd2c, + 0x00800799, + 0x09f60237, + 0xf404bd00, + 0x32f40132, + 0x08817e02, + 0xf094bd00, + 0x00800799, + 0x09f60217, + 0xf404bd00, +/* 0x05aa: main_not_ctx_save */ + 0xef94110e, + 0x01f5f010, + 0x0002f87e, + 0xfee40ef5, +/* 0x05b8: main_done */ + 0x29f024bd, + 0x3000801f, + 0x0002f602, + 0x0ef504bd, +/* 0x05ca: ih */ + 0x00f9fed2, + 0x88fe80f9, + 0xf980f901, + 0xf9a0f990, + 0xf9d0f9b0, + 0xbdf0f9e0, + 0x02004a04, + 0xc400aacf, + 0x0bf404ab, + 0x4e100d23, + 0xeecf1a00, + 0x19004f00, + 0x7e00ffcf, + 0xb7000004, + 0x0e0400b0, + 0x1d004001, + 0xbd000ef6, +/* 0x060d: ih_no_fifo */ + 0x00abe404, + 0x0c0bf401, + 0x014e100d, + 0x00047e40, +/* 0x061d: ih_no_ctxsw */ + 0x00abe400, + 0x560bf404, + 0x4007088e, + 0x0000657e, + 0x0080ffb2, + 0x0ff60204, + 0x8e04bd00, + 0x7e400704, + 0xb2000065, + 0x030080ff, + 0x000ff602, + 0xfec704bd, + 0x02ee9450, + 0x4007008f, + 0x7e00efbb, + 0x80000065, + 0xf6020200, + 0x04bd000f, + 0xf87e030f, + 0x004b0002, + 0x8ebfb201, + 0x7e400144, +/* 0x0677: ih_no_fwmthd */ + 0x4b00008f, + 0xb0bd0504, + 0xf4b4abff, + 0x00800c0b, + 0x0bf60307, +/* 0x068b: ih_no_other */ + 0x4004bd00, + 0x0af60100, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x00fc80fc, + 0xf80032f4, +/* 0x06ad: ctx_4170s */ + 0x10f5f001, + 0x708effb2, + 0x8f7e4041, + 0x00f80000, +/* 0x06bc: ctx_4170w */ + 0x4041708e, + 0x0000657e, + 0xf4f0ffb2, + 0xf31bf410, +/* 0x06ce: ctx_redswitch */ + 0x004e00f8, + 0x40e5f002, + 0xf020e5f0, + 0x008010e5, + 0x0ef60185, + 0x0f04bd00, +/* 0x06e5: ctx_redswitch_delay */ + 0x01f2b608, + 0xf1fd1bf4, + 0xf10400e5, + 0x800100e5, + 0xf6018500, + 0x04bd000e, +/* 0x06fe: ctx_86c */ + 0x008000f8, + 0x0ff60223, + 0xb204bd00, + 0x8a148eff, + 0x008f7e40, + 0x8effb200, + 0x7e41a88c, + 0xf800008f, +/* 0x071d: ctx_mem */ + 0x84008000, + 0x000ff602, +/* 0x0726: ctx_mem_wait */ + 0x008f04bd, + 0xffcf0284, + 0x05fffd00, + 0xf8f61bf4, +/* 0x0735: ctx_load */ + 0xf094bd00, + 0x00800599, + 0x09f60237, + 0x0a04bd00, + 0x00b87e0c, + 0x80f4bd00, + 0xf6028900, + 0x04bd000f, + 0x02c10080, + 0xbd0002f6, + 0x83008004, + 0x0002f602, + 0x070f04bd, + 0x00071d7e, + 0x02c00080, + 0xbd0002f6, + 0x000bfe04, + 0xb61f2af0, + 0x20b60424, + 0xf094bd02, + 0x00800899, + 0x09f60237, + 0x8004bd00, + 0xf6028100, + 0x04bd0002, + 0x000000d2, + 0x0225f080, + 0x02880080, + 0xbd0002f6, + 0x42100104, + 0x23f00200, + 0x0512fa02, + 0x94bd03f8, + 0x800899f0, + 0xf6021700, + 0x04bd0009, + 0xb6810198, + 0x02981814, + 0x0825b680, + 0xb50512fd, + 0x94bd1601, + 0x800999f0, + 0xf6023700, + 0x04bd0009, + 0x02810080, + 0xbd0001f6, + 0x80010204, + 0xf6028800, + 0x04bd0002, + 0xf0010041, + 0x01fa0613, + 0xbd03f805, + 0x0999f094, + 0x02170080, + 0xbd0009f6, + 0xf094bd04, + 0x00800599, + 0x09f60217, + 0xf804bd00, +/* 0x0821: ctx_chan */ + 0x07357e00, + 0x7e0c0a00, + 0x0f0000b8, + 0x071d7e05, +/* 0x0833: ctx_mmio_exec */ + 0x9800f800, + 0x00804103, + 0x03f60281, + 0xbd04bd00, +/* 0x0841: ctx_mmio_loop */ + 0xff34c434, + 0x450e1bf4, + 0x53f00200, + 0x0535fa06, +/* 0x0852: ctx_mmio_pull */ + 0x4e9803f8, + 0x814f9880, + 0x00008f7e, + 0xb60830b6, + 0x1bf40112, +/* 0x0865: ctx_mmio_done */ + 0x160398df, + 0x02810080, + 0xbd0003f6, + 0x4000b504, + 0xf0010041, + 0x01fa0613, + 0xf803f806, +/* 0x0881: ctx_xfer */ + 0x80040e00, + 0xf6030200, + 0x04bd000e, +/* 0x088c: ctx_xfer_idle */ + 0x0300008e, + 0xf100eecf, + 0xf42000e4, + 0x11f4f51b, + 0x0c02f406, +/* 0x08a0: ctx_xfer_pre */ + 0xfe7e100f, + 0x11f40006, +/* 0x08a9: ctx_xfer_pre_load */ + 0x7e020f1b, + 0x7e0006ad, + 0x7e0006bc, + 0xbd0006ce, + 0x06ad7ef4, + 0x07357e00, +/* 0x08c1: ctx_xfer_exec */ + 0x16019800, + 0x008024bd, + 0x02f60105, + 0xb204bd00, + 0xa5008e1f, + 0x008f7e41, + 0x01fcf000, + 0xb6022cf0, + 0xf2fd0124, + 0x8effb205, + 0x7e41a504, + 0x7e00008f, + 0xbd000216, + 0x47fc8024, + 0x0002f602, + 0x2cf004bd, + 0x0320b601, + 0x024afc80, + 0xbd0002f6, + 0x01acf004, + 0x0b06a5f0, + 0x000c9800, + 0x0e010d98, + 0x013d7e00, + 0x7e080a00, + 0x7e0000ec, + 0xf400020a, + 0x0c0a1201, + 0x0000b87e, + 0x1d7e050f, + 0x02f40007, +/* 0x093d: ctx_xfer_post */ + 0x7e020f2d, + 0xbd0006ad, + 0x06fe7ef4, + 0x02277e00, + 0x06bc7e00, + 0x7ef4bd00, + 0xf40006ad, + 0x01981011, + 0x0511fd40, + 0x7e070bf4, +/* 0x0967: ctx_xfer_no_post_mmio */ +/* 0x0967: ctx_xfer_done */ + 0xf8000833, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5 b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5 new file mode 100644 index 0000000000..27591b3086 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5 @@ -0,0 +1,40 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#define CHIPSET GK208 +#include "macros.fuc" + +.section #gm107_grhub_data +#define INCLUDE_DATA +#include "com.fuc" +#include "hub.fuc" +#undef INCLUDE_DATA + +.section #gm107_grhub_code +#define INCLUDE_CODE +bra #init +#include "com.fuc" +#include "hub.fuc" +.align 256 +#undef INCLUDE_CODE diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h new file mode 100644 index 0000000000..449dae7532 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/hubgm107.fuc5.h @@ -0,0 +1,918 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t gm107_grhub_data[] = { +/* 0x0000: hub_mmio_list_head */ + 0x00000300, +/* 0x0004: hub_mmio_list_tail */ + 0x00000304, +/* 0x0008: gpc_count */ + 0x00000000, +/* 0x000c: rop_count */ + 0x00000000, +/* 0x0010: cmd_queue */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0058: ctx_current */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0100: chan_data */ +/* 0x0100: chan_mmio_count */ + 0x00000000, +/* 0x0104: chan_mmio_address */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0200: xfer_data */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0300: hub_mmio_list_base */ + 0x0417e91c, +}; + +static uint32_t gm107_grhub_code[] = { + 0x030e0ef5, +/* 0x0004: queue_put */ + 0x9800d898, + 0x86f001d9, + 0xf489a408, + 0x020f0b1b, + 0x0002f87e, +/* 0x001a: queue_put_next */ + 0x98c400f8, + 0x0384b607, + 0xb6008dbb, + 0x8eb50880, + 0x018fb500, + 0xf00190b6, + 0xd9b50f94, +/* 0x0037: queue_get */ + 0xf400f801, + 0xd8980131, + 0x01d99800, + 0x0bf489a4, + 0x0789c421, + 0xbb0394b6, + 0x90b6009d, + 0x009e9808, + 0xb6019f98, + 0x84f00180, + 0x00d8b50f, +/* 0x0063: queue_get_done */ + 0xf80132f4, +/* 0x0065: nv_rd32 */ + 0xf0ecb200, + 0x00801fc9, + 0x0cf601ca, +/* 0x0073: nv_rd32_wait */ + 0x8c04bd00, + 0xcf01ca00, + 0xccc800cc, + 0xf61bf41f, + 0xec7e060a, + 0x008f0000, + 0xffcf01cb, +/* 0x008f: nv_wr32 */ + 0x8000f800, + 0xf601cc00, + 0x04bd000f, + 0xc9f0ecb2, + 0x1ec9f01f, + 0x01ca0080, + 0xbd000cf6, +/* 0x00a9: nv_wr32_wait */ + 0xca008c04, + 0x00cccf01, + 0xf41fccc8, + 0x00f8f61b, +/* 0x00b8: wait_donez */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x00cf: wait_donez_ne */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf61bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x00ec: wait_doneo */ + 0x99f094bd, + 0x37008000, + 0x0009f602, + 0x008004bd, + 0x0af60206, +/* 0x0103: wait_doneo_e */ + 0x8804bd00, + 0xcf010000, + 0x8aff0088, + 0xf60bf488, + 0x99f094bd, + 0x17008000, + 0x0009f602, + 0x00f804bd, +/* 0x0120: mmctx_size */ +/* 0x0122: nv_mmctx_size_loop */ + 0xe89894bd, + 0x1a85b600, + 0xb60180b6, + 0x98bb0284, + 0x04e0b600, + 0x1bf4efa4, + 0xf89fb2ec, +/* 0x013d: mmctx_xfer */ + 0xf094bd00, + 0x00800199, + 0x09f60237, + 0xbd04bd00, + 0x05bbfd94, + 0x800f0bf4, + 0xf601c400, + 0x04bd000b, +/* 0x015f: mmctx_base_disabled */ + 0xfd0099f0, + 0x0bf405ee, + 0xc6008018, + 0x000ef601, + 0x008004bd, + 0x0ff601c7, + 0xf004bd00, +/* 0x017a: mmctx_multi_disabled */ + 0xabc80199, + 0x10b4b600, + 0xc80cb9f0, + 0xe4b601ae, + 0x05befd11, + 0x01c50080, + 0xbd000bf6, +/* 0x0195: mmctx_exec_loop */ +/* 0x0195: mmctx_wait_free */ + 0xc5008e04, + 0x00eecf01, + 0xf41fe4f0, + 0xce98f60b, + 0x05e9fd00, + 0x01c80080, + 0xbd000ef6, + 0x04c0b604, + 0x1bf4cda4, + 0x02abc8df, +/* 0x01bf: mmctx_fini_wait */ + 0x8b1c1bf4, + 0xcf01c500, + 0xb4f000bb, + 0x10b4b01f, + 0x0af31bf4, + 0x00b87e05, + 0x250ef400, +/* 0x01d8: mmctx_stop */ + 0xb600abc8, + 0xb9f010b4, + 0x12b9f00c, + 0x01c50080, + 0xbd000bf6, +/* 0x01ed: mmctx_stop_wait */ + 0xc5008b04, + 0x00bbcf01, + 0xf412bbc8, +/* 0x01fa: mmctx_done */ + 0x94bdf61b, + 0x800199f0, + 0xf6021700, + 0x04bd0009, +/* 0x020a: strand_wait */ + 0xa0f900f8, + 0xb87e020a, + 0xa0fc0000, +/* 0x0216: strand_pre */ + 0x0c0900f8, + 0x024afc80, + 0xbd0009f6, + 0x020a7e04, +/* 0x0227: strand_post */ + 0x0900f800, + 0x4afc800d, + 0x0009f602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0238: strand_set */ + 0xfc800f0c, + 0x0cf6024f, + 0x0c04bd00, + 0x4afc800b, + 0x000cf602, + 0xfc8004bd, + 0x0ef6024f, + 0x0c04bd00, + 0x4afc800a, + 0x000cf602, + 0x0a7e04bd, + 0x00f80002, +/* 0x0268: strand_ctx_init */ + 0x99f094bd, + 0x37008003, + 0x0009f602, + 0x167e04bd, + 0x030e0002, + 0x0002387e, + 0xfc80c4bd, + 0x0cf60247, + 0x0c04bd00, + 0x4afc8001, + 0x000cf602, + 0x0a7e04bd, + 0x0c920002, + 0x46fc8001, + 0x000cf602, + 0x020c04bd, + 0x024afc80, + 0xbd000cf6, + 0x020a7e04, + 0x02277e00, + 0x42008800, + 0x20008902, + 0x0099cf02, +/* 0x02c7: ctx_init_strand_loop */ + 0xf608fe95, + 0x8ef6008e, + 0x808acf40, + 0xb606a5b6, + 0xeabb01a0, + 0x0480b600, + 0xf40192b6, + 0xe4b6e81b, + 0xf2efbc08, + 0x99f094bd, + 0x17008003, + 0x0009f602, + 0x00f804bd, +/* 0x02f8: error */ + 0x02050080, + 0xbd000ff6, + 0x80010f04, + 0xf6030700, + 0x04bd000f, +/* 0x030e: init */ + 0x04bd00f8, + 0x410007fe, + 0x11cf4200, + 0x0911e700, + 0x0814b601, + 0x020014fe, + 0x12004002, + 0xbd0002f6, + 0x05ca4104, + 0xbd0010fe, + 0x07004024, + 0xbd0002f6, + 0x20034204, + 0x01010080, + 0xbd0002f6, + 0x20044204, + 0x01010480, + 0xbd0002f6, + 0x200b4204, + 0x01010880, + 0xbd0002f6, + 0x200c4204, + 0x01011c80, + 0xbd0002f6, + 0x01039204, + 0x03090080, + 0xbd0003f6, + 0x87048204, + 0x04004000, + 0xbd0002f6, + 0x40040204, + 0x02f60300, + 0xf404bd00, + 0x048e1031, + 0x657e4096, + 0xfeb20000, + 0xb590f1c7, + 0xf4f00301, + 0x020fb51f, + 0x1fbb0101, + 0x0112b604, + 0x01030080, + 0xbd0001f6, + 0x04008004, + 0x0001f601, + 0x004104bd, + 0x7e020f01, + 0x7e0006ad, + 0x0f0006bc, + 0x06fe7e10, + 0x000e9800, + 0x7e010f98, + 0x95000120, + 0x00800814, + 0x04f601c0, + 0x8004bd00, + 0xf601c100, + 0x04bd0004, + 0x130030b7, + 0xb6001fbb, + 0x008002f5, + 0x0ff601d3, + 0xb604bd00, + 0x10b60815, + 0x0814b601, + 0x687e1fb2, + 0x1fbb0002, + 0x02039800, + 0x50200084, +/* 0x0420: init_gpc */ + 0x08044eb8, + 0x7e1fb200, + 0xb800008f, + 0x00010c4e, + 0x8f7ef4bd, + 0x4eb80000, + 0x7e000104, + 0xb800008f, + 0x0001004e, + 0x8f7e020f, + 0x4eb80000, +/* 0x044f: init_gpc_wait */ + 0x7e000800, + 0xc8000065, + 0x0bf41fff, + 0x044eb8f9, + 0x657e0008, + 0x1fbb0000, + 0x0040b700, + 0x0132b680, + 0x0fb41bf4, + 0x06fe7e00, + 0x7e000f00, + 0x800006ad, + 0xf6020100, + 0x04bd0001, + 0x19f014bd, + 0x3000801f, + 0x0001f602, +/* 0x0492: wait */ + 0x28f404bd, + 0x0031f400, +/* 0x0498: main */ + 0x377e100d, + 0x01f40000, + 0x01e4b1f4, + 0xc71bf540, + 0xf094bd00, + 0x00800499, + 0x09f60237, + 0x8104bd00, + 0xcf02c000, + 0x00820011, + 0x22cf02c1, + 0x1f13c800, + 0xc8770bf4, + 0x0bf41f23, + 0xb220f955, + 0xf094bd12, + 0x00800799, + 0x09f60237, + 0xf404bd00, + 0x31f40132, + 0x08817e02, + 0xf094bd00, + 0x00800799, + 0x09f60217, + 0xfc04bd00, + 0xf094bd20, + 0x00800699, + 0x09f60237, + 0xf404bd00, + 0x817e0131, + 0x94bd0008, + 0x800699f0, + 0xf6021700, + 0x04bd0009, +/* 0x0523: chsw_prev_no_next */ + 0xf92f0ef4, + 0xf412b220, + 0x32f40132, + 0x08817e02, + 0x8020fc00, + 0xf602c000, + 0x04bd0002, +/* 0x053f: chsw_no_prev */ + 0xc8130ef4, + 0x0bf41f23, + 0x0131f40d, + 0x7e0232f4, +/* 0x054f: chsw_done */ + 0x02000881, + 0xc3008001, + 0x0002f602, + 0x94bd04bd, + 0x800499f0, + 0xf6021700, + 0x04bd0009, + 0xff300ef5, +/* 0x056c: main_not_ctx_switch */ + 0xf401e4b0, + 0xf2b20c1b, + 0x0008217e, +/* 0x057b: main_not_ctx_chan */ + 0xb0400ef4, + 0x1bf402e4, + 0xf094bd2c, + 0x00800799, + 0x09f60237, + 0xf404bd00, + 0x32f40132, + 0x08817e02, + 0xf094bd00, + 0x00800799, + 0x09f60217, + 0xf404bd00, +/* 0x05aa: main_not_ctx_save */ + 0xef94110e, + 0x01f5f010, + 0x0002f87e, + 0xfee40ef5, +/* 0x05b8: main_done */ + 0x29f024bd, + 0x3000801f, + 0x0002f602, + 0x0ef504bd, +/* 0x05ca: ih */ + 0x00f9fed2, + 0x88fe80f9, + 0xf980f901, + 0xf9a0f990, + 0xf9d0f9b0, + 0xbdf0f9e0, + 0x02004a04, + 0xc400aacf, + 0x0bf404ab, + 0x4e100d23, + 0xeecf1a00, + 0x19004f00, + 0x7e00ffcf, + 0xb7000004, + 0x0e0400b0, + 0x1d004001, + 0xbd000ef6, +/* 0x060d: ih_no_fifo */ + 0x00abe404, + 0x0c0bf401, + 0x014e100d, + 0x00047e40, +/* 0x061d: ih_no_ctxsw */ + 0x00abe400, + 0x560bf404, + 0x4007088e, + 0x0000657e, + 0x0080ffb2, + 0x0ff60204, + 0x8e04bd00, + 0x7e400704, + 0xb2000065, + 0x030080ff, + 0x000ff602, + 0xfec704bd, + 0x02ee9450, + 0x4007008f, + 0x7e00efbb, + 0x80000065, + 0xf6020200, + 0x04bd000f, + 0xf87e030f, + 0x004b0002, + 0x8ebfb201, + 0x7e400144, +/* 0x0677: ih_no_fwmthd */ + 0x4b00008f, + 0xb0bd0504, + 0xf4b4abff, + 0x00800c0b, + 0x0bf60307, +/* 0x068b: ih_no_other */ + 0x4004bd00, + 0x0af60100, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x00fc80fc, + 0xf80032f4, +/* 0x06ad: ctx_4170s */ + 0x10f5f001, + 0x708effb2, + 0x8f7e4041, + 0x00f80000, +/* 0x06bc: ctx_4170w */ + 0x4041708e, + 0x0000657e, + 0xf4f0ffb2, + 0xf31bf410, +/* 0x06ce: ctx_redswitch */ + 0x004e00f8, + 0x40e5f002, + 0xf020e5f0, + 0x008010e5, + 0x0ef60185, + 0x0f04bd00, +/* 0x06e5: ctx_redswitch_delay */ + 0x01f2b608, + 0xf1fd1bf4, + 0xf10400e5, + 0x800100e5, + 0xf6018500, + 0x04bd000e, +/* 0x06fe: ctx_86c */ + 0x008000f8, + 0x0ff60223, + 0xb204bd00, + 0x8a148eff, + 0x008f7e40, + 0x8effb200, + 0x7e41a88c, + 0xf800008f, +/* 0x071d: ctx_mem */ + 0x84008000, + 0x000ff602, +/* 0x0726: ctx_mem_wait */ + 0x008f04bd, + 0xffcf0284, + 0x05fffd00, + 0xf8f61bf4, +/* 0x0735: ctx_load */ + 0xf094bd00, + 0x00800599, + 0x09f60237, + 0x0a04bd00, + 0x00b87e0c, + 0x80f4bd00, + 0xf6028900, + 0x04bd000f, + 0x02c10080, + 0xbd0002f6, + 0x83008004, + 0x0002f602, + 0x070f04bd, + 0x00071d7e, + 0x02c00080, + 0xbd0002f6, + 0x000bfe04, + 0xb61f2af0, + 0x20b60424, + 0xf094bd02, + 0x00800899, + 0x09f60237, + 0x8004bd00, + 0xf6028100, + 0x04bd0002, + 0x000000d2, + 0x0225f080, + 0x02880080, + 0xbd0002f6, + 0x42100104, + 0x23f00200, + 0x0512fa02, + 0x94bd03f8, + 0x800899f0, + 0xf6021700, + 0x04bd0009, + 0xb6810198, + 0x02981814, + 0x0825b680, + 0xb50512fd, + 0x94bd1601, + 0x800999f0, + 0xf6023700, + 0x04bd0009, + 0x02810080, + 0xbd0001f6, + 0x80010204, + 0xf6028800, + 0x04bd0002, + 0xf0010041, + 0x01fa0613, + 0xbd03f805, + 0x0999f094, + 0x02170080, + 0xbd0009f6, + 0xf094bd04, + 0x00800599, + 0x09f60217, + 0xf804bd00, +/* 0x0821: ctx_chan */ + 0x07357e00, + 0x7e0c0a00, + 0x0f0000b8, + 0x071d7e05, +/* 0x0833: ctx_mmio_exec */ + 0x9800f800, + 0x00804103, + 0x03f60281, + 0xbd04bd00, +/* 0x0841: ctx_mmio_loop */ + 0xff34c434, + 0x450e1bf4, + 0x53f00200, + 0x0535fa06, +/* 0x0852: ctx_mmio_pull */ + 0x4e9803f8, + 0x814f9880, + 0x00008f7e, + 0xb60830b6, + 0x1bf40112, +/* 0x0865: ctx_mmio_done */ + 0x160398df, + 0x02810080, + 0xbd0003f6, + 0x4000b504, + 0xf0010041, + 0x01fa0613, + 0xf803f806, +/* 0x0881: ctx_xfer */ + 0x80040e00, + 0xf6030200, + 0x04bd000e, +/* 0x088c: ctx_xfer_idle */ + 0x0300008e, + 0xf100eecf, + 0xf42000e4, + 0x11f4f51b, + 0x0c02f406, +/* 0x08a0: ctx_xfer_pre */ + 0xfe7e100f, + 0x11f40006, +/* 0x08a9: ctx_xfer_pre_load */ + 0x7e020f1b, + 0x7e0006ad, + 0x7e0006bc, + 0xbd0006ce, + 0x06ad7ef4, + 0x07357e00, +/* 0x08c1: ctx_xfer_exec */ + 0x16019800, + 0x008024bd, + 0x02f60105, + 0xb204bd00, + 0xa5008e1f, + 0x008f7e41, + 0x01fcf000, + 0xb6022cf0, + 0xf2fd0124, + 0x8effb205, + 0x7e41a504, + 0x7e00008f, + 0xbd000216, + 0x47fc8024, + 0x0002f602, + 0x2cf004bd, + 0x0320b601, + 0x024afc80, + 0xbd0002f6, + 0x01acf004, + 0x0b06a5f0, + 0x000c9800, + 0x0e010d98, + 0x013d7e00, + 0x7e080a00, + 0x7e0000ec, + 0xf400020a, + 0x0c0a1201, + 0x0000b87e, + 0x1d7e050f, + 0x02f40007, +/* 0x093d: ctx_xfer_post */ + 0x7e020f2d, + 0xbd0006ad, + 0x06fe7ef4, + 0x02277e00, + 0x06bc7e00, + 0x7ef4bd00, + 0xf40006ad, + 0x01981011, + 0x0511fd40, + 0x7e070bf4, +/* 0x0967: ctx_xfer_no_post_mmio */ +/* 0x0967: ctx_xfer_done */ + 0xf8000833, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/macros.fuc b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/macros.fuc new file mode 100644 index 0000000000..fa61806644 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/macros.fuc @@ -0,0 +1,261 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ + +#include "os.h" + +#define GF100 0xc0 +#define GF117 0xd7 +#define GK100 0xe0 +#define GK110 0xf0 +#define GK208 0x108 +#define GM107 0x117 + +#define NV_PGRAPH_TRAPPED_ADDR 0x400704 +#define NV_PGRAPH_TRAPPED_DATA_LO 0x400708 +#define NV_PGRAPH_TRAPPED_DATA_HI 0x40070c + +#define NV_PGRAPH_FE_OBJECT_TABLE(n) ((n) * 4 + 0x400700) + +#define NV_PGRAPH_FECS_INTR_ACK 0x409004 +#define NV_PGRAPH_FECS_INTR 0x409008 +#define NV_PGRAPH_FECS_INTR_FWMTHD 0x00000400 +#define NV_PGRAPH_FECS_INTR_CHSW 0x00000100 +#define NV_PGRAPH_FECS_INTR_FIFO 0x00000004 +#define NV_PGRAPH_FECS_INTR_MODE 0x40900c +#define NV_PGRAPH_FECS_INTR_MODE_FIFO 0x00000004 +#define NV_PGRAPH_FECS_INTR_MODE_FIFO_LEVEL 0x00000004 +#define NV_PGRAPH_FECS_INTR_MODE_FIFO_EDGE 0x00000000 +#define NV_PGRAPH_FECS_INTR_EN_SET 0x409010 +#define NV_PGRAPH_FECS_INTR_EN_SET_FIFO 0x00000004 +#define NV_PGRAPH_FECS_INTR_ROUTE 0x40901c +#define NV_PGRAPH_FECS_ACCESS 0x409048 +#define NV_PGRAPH_FECS_ACCESS_FIFO 0x00000002 +#define NV_PGRAPH_FECS_FIFO_DATA 0x409064 +#define NV_PGRAPH_FECS_FIFO_CMD 0x409068 +#define NV_PGRAPH_FECS_FIFO_ACK 0x409074 +#define NV_PGRAPH_FECS_CAPS 0x409108 +#define NV_PGRAPH_FECS_SIGNAL 0x409400 +#define NV_PGRAPH_FECS_IROUTE 0x409404 +#define NV_PGRAPH_FECS_BAR_MASK0 0x40940c +#define NV_PGRAPH_FECS_BAR_MASK1 0x409410 +#define NV_PGRAPH_FECS_BAR 0x409414 +#define NV_PGRAPH_FECS_BAR_SET 0x409418 +#define NV_PGRAPH_FECS_RED_SWITCH 0x409614 +#define NV_PGRAPH_FECS_RED_SWITCH_ENABLE_ROP 0x00000400 +#define NV_PGRAPH_FECS_RED_SWITCH_ENABLE_GPC 0x00000200 +#define NV_PGRAPH_FECS_RED_SWITCH_ENABLE_MAIN 0x00000100 +#define NV_PGRAPH_FECS_RED_SWITCH_POWER_ROP 0x00000040 +#define NV_PGRAPH_FECS_RED_SWITCH_POWER_GPC 0x00000020 +#define NV_PGRAPH_FECS_RED_SWITCH_POWER_MAIN 0x00000010 +#define NV_PGRAPH_FECS_RED_SWITCH_PAUSE_GPC 0x00000002 +#define NV_PGRAPH_FECS_RED_SWITCH_PAUSE_MAIN 0x00000001 +#define NV_PGRAPH_FECS_MMCTX_SAVE_SWBASE 0x409700 +#define NV_PGRAPH_FECS_MMCTX_LOAD_SWBASE 0x409704 +#define NV_PGRAPH_FECS_MMCTX_LOAD_COUNT 0x40974c +#define NV_PGRAPH_FECS_MMCTX_SAVE_SWBASE 0x409700 +#define NV_PGRAPH_FECS_MMCTX_LOAD_SWBASE 0x409704 +#define NV_PGRAPH_FECS_MMCTX_BASE 0x409710 +#define NV_PGRAPH_FECS_MMCTX_CTRL 0x409714 +#define NV_PGRAPH_FECS_MMCTX_MULTI_STRIDE 0x409718 +#define NV_PGRAPH_FECS_MMCTX_MULTI_MASK 0x40971c +#define NV_PGRAPH_FECS_MMCTX_QUEUE 0x409720 +#define NV_PGRAPH_FECS_MMIO_BASE 0x409724 +#define NV_PGRAPH_FECS_MMIO_CTRL 0x409728 +#define NV_PGRAPH_FECS_MMIO_CTRL_BASE_ENABLE 0x00000001 +#define NV_PGRAPH_FECS_MMIO_RDVAL 0x40972c +#define NV_PGRAPH_FECS_MMIO_WRVAL 0x409730 +#define NV_PGRAPH_FECS_MMCTX_LOAD_COUNT 0x40974c +#if CHIPSET < GK110 +#define NV_PGRAPH_FECS_CC_SCRATCH_VAL(n) ((n) * 4 + 0x409800) +#define NV_PGRAPH_FECS_CC_SCRATCH_SET(n) ((n) * 4 + 0x409820) +#define NV_PGRAPH_FECS_CC_SCRATCH_CLR(n) ((n) * 4 + 0x409840) +#define NV_PGRAPH_FECS_UNK86C 0x40986c +#else +#define NV_PGRAPH_FECS_CC_SCRATCH_VAL(n) ((n) * 4 + 0x409800) +#define NV_PGRAPH_FECS_CC_SCRATCH_CLR(n) ((n) * 4 + 0x409840) +#define NV_PGRAPH_FECS_UNK86C 0x40988c +#define NV_PGRAPH_FECS_CC_SCRATCH_SET(n) ((n) * 4 + 0x4098c0) +#endif +#define NV_PGRAPH_FECS_STRANDS_CNT 0x409880 +#define NV_PGRAPH_FECS_STRAND_SAVE_SWBASE 0x409908 +#define NV_PGRAPH_FECS_STRAND_LOAD_SWBASE 0x40990c +#define NV_PGRAPH_FECS_STRAND_WORDS 0x409910 +#define NV_PGRAPH_FECS_STRAND_DATA 0x409918 +#define NV_PGRAPH_FECS_STRAND_SELECT 0x40991c +#define NV_PGRAPH_FECS_STRAND_CMD 0x409928 +#define NV_PGRAPH_FECS_STRAND_CMD_SEEK 0x00000001 +#define NV_PGRAPH_FECS_STRAND_CMD_GET_INFO 0x00000002 +#define NV_PGRAPH_FECS_STRAND_CMD_SAVE 0x00000003 +#define NV_PGRAPH_FECS_STRAND_CMD_LOAD 0x00000004 +#define NV_PGRAPH_FECS_STRAND_CMD_ACTIVATE_FILTER 0x0000000a +#define NV_PGRAPH_FECS_STRAND_CMD_DEACTIVATE_FILTER 0x0000000b +#define NV_PGRAPH_FECS_STRAND_CMD_ENABLE 0x0000000c +#define NV_PGRAPH_FECS_STRAND_CMD_DISABLE 0x0000000d +#define NV_PGRAPH_FECS_STRAND_FILTER 0x40993c +#define NV_PGRAPH_FECS_MEM_BASE 0x409a04 +#define NV_PGRAPH_FECS_MEM_CHAN 0x409a0c +#define NV_PGRAPH_FECS_MEM_CMD 0x409a10 +#define NV_PGRAPH_FECS_MEM_CMD_LOAD_CHAN 0x00000007 +#define NV_PGRAPH_FECS_MEM_TARGET 0x409a20 +#define NV_PGRAPH_FECS_MEM_TARGET_UNK31 0x80000000 +#define NV_PGRAPH_FECS_MEM_TARGET_AS 0x0000001f +#define NV_PGRAPH_FECS_MEM_TARGET_AS_VM 0x00000001 +#define NV_PGRAPH_FECS_MEM_TARGET_AS_VRAM 0x00000002 +#define NV_PGRAPH_FECS_CHAN_ADDR 0x409b00 +#define NV_PGRAPH_FECS_CHAN_NEXT 0x409b04 +#define NV_PGRAPH_FECS_CHSW 0x409b0c +#define NV_PGRAPH_FECS_CHSW_ACK 0x00000001 +#define NV_PGRAPH_FECS_INTR_UP_SET 0x409c1c +#define NV_PGRAPH_FECS_INTR_UP_EN 0x409c24 + +#define NV_PGRAPH_GPCX_GPCCS_INTR_ACK 0x41a004 +#define NV_PGRAPH_GPCX_GPCCS_INTR 0x41a008 +#define NV_PGRAPH_GPCX_GPCCS_INTR_FIFO 0x00000004 +#define NV_PGRAPH_GPCX_GPCCS_INTR_EN_SET 0x41a010 +#define NV_PGRAPH_GPCX_GPCCS_INTR_EN_SET_FIFO 0x00000004 +#define NV_PGRAPH_GPCX_GPCCS_INTR_ROUTE 0x41a01c +#define NV_PGRAPH_GPCX_GPCCS_ACCESS 0x41a048 +#define NV_PGRAPH_GPCX_GPCCS_ACCESS_FIFO 0x00000002 +#define NV_PGRAPH_GPCX_GPCCS_FIFO_DATA 0x41a064 +#define NV_PGRAPH_GPCX_GPCCS_FIFO_CMD 0x41a068 +#define NV_PGRAPH_GPCX_GPCCS_FIFO_ACK 0x41a074 +#define NV_PGRAPH_GPCX_GPCCS_UNITS 0x41a608 +#define NV_PGRAPH_GPCX_GPCCS_CAPS 0x41a108 +#define NV_PGRAPH_GPCX_GPCCS_RED_SWITCH 0x41a614 +#define NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_UNK11 0x00000800 +#define NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_ENABLE 0x00000200 +#define NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_POWER 0x00000020 +#define NV_PGRAPH_GPCX_GPCCS_RED_SWITCH_PAUSE 0x00000002 +#define NV_PGRAPH_GPCX_GPCCS_MYINDEX 0x41a618 +#define NV_PGRAPH_GPCX_GPCCS_MMCTX_SAVE_SWBASE 0x41a700 +#define NV_PGRAPH_GPCX_GPCCS_MMCTX_LOAD_SWBASE 0x41a704 +#define NV_PGRAPH_GPCX_GPCCS_MMIO_BASE 0x41a724 +#define NV_PGRAPH_GPCX_GPCCS_MMIO_CTRL 0x41a728 +#define NV_PGRAPH_GPCX_GPCCS_MMIO_CTRL_BASE_ENABLE 0x00000001 +#define NV_PGRAPH_GPCX_GPCCS_MMIO_RDVAL 0x41a72c +#define NV_PGRAPH_GPCX_GPCCS_MMIO_WRVAL 0x41a730 +#define NV_PGRAPH_GPCX_GPCCS_MMCTX_LOAD_COUNT 0x41a74c +#if CHIPSET < GK110 +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_VAL(n) ((n) * 4 + 0x41a800) +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_SET(n) ((n) * 4 + 0x41a820) +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_CLR(n) ((n) * 4 + 0x41a840) +#define NV_PGRAPH_GPCX_GPCCS_UNK86C 0x41a86c +#else +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_VAL(n) ((n) * 4 + 0x41a800) +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_CLR(n) ((n) * 4 + 0x41a840) +#define NV_PGRAPH_GPCX_GPCCS_UNK86C 0x41a88c +#define NV_PGRAPH_GPCX_GPCCS_CC_SCRATCH_SET(n) ((n) * 4 + 0x41a8c0) +#endif +#define NV_PGRAPH_GPCX_GPCCS_STRAND_SELECT 0x41a91c +#define NV_PGRAPH_GPCX_GPCCS_STRAND_CMD 0x41a928 +#define NV_PGRAPH_GPCX_GPCCS_STRAND_CMD_SAVE 0x00000003 +#define NV_PGRAPH_GPCX_GPCCS_STRAND_CMD_LOAD 0x00000004 +#define NV_PGRAPH_GPCX_GPCCS_MEM_BASE 0x41aa04 +#define NV_PGRAPH_GPCX_GPCCS_TPC_STATUS 0x41acfc + +#define NV_PGRAPH_GPC0_TPC0 0x504000 +#define NV_PGRAPH_GPC0_TPC0__SIZE 0x000800 + +#define NV_PGRAPH_GPC0_TPCX_STRAND_INDEX 0x501d60 +#define NV_PGRAPH_GPC0_TPCX_STRAND_INDEX_ALL 0x0000003f +#define NV_PGRAPH_GPC0_TPCX_STRAND_DATA 0x501d98 +#define NV_PGRAPH_GPC0_TPCX_STRAND_SELECT 0x501d9c +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD 0x501da8 +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_SEEK 0x00000001 +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_GET_INFO 0x00000002 +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_SAVE 0x00000003 +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_LOAD 0x00000004 +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_ENABLE 0x0000000c +#define NV_PGRAPH_GPC0_TPCX_STRAND_CMD_DISABLE 0x0000000d +#define NV_PGRAPH_GPC0_TPCX_STRAND_MEM_BASE 0x501dc4 + +#define NV_TPC_STRAND_INDEX 0x560 +#define NV_TPC_STRAND_CNT 0x570 +#define NV_TPC_STRAND_SAVE_SWBASE 0x588 +#define NV_TPC_STRAND_LOAD_SWBASE 0x58c +#define NV_TPC_STRAND_WORDS 0x590 + +#define mmctx_data(r,c) .b32 (((c - 1) << 26) | r) +#define queue_init .skip 72 // (2 * 4) + ((8 * 4) * 2) + +#define T_WAIT 0 +#define T_MMCTX 1 +#define T_STRWAIT 2 +#define T_STRINIT 3 +#define T_AUTO 4 +#define T_CHAN 5 +#define T_LOAD 6 +#define T_SAVE 7 +#define T_LCHAN 8 +#define T_LCTXH 9 +#define T_STRTPC 10 + +#if CHIPSET < GK208 +#define imm32(reg,val) /* +*/ movw reg ((val) & 0x0000ffff) /* +*/ sethi reg ((val) & 0xffff0000) +#else +#define imm32(reg,val) /* +*/ mov reg (val) +#endif + +#define nv_mkio(rv,r,i) /* +*/ imm32(rv, (((r) & 0xffc) << 6) | ((i) << 2)) + +#define hash # +#define fn(a) a +#if CHIPSET < GK208 +#define call(a) call fn(hash)a +#else +#define call(a) lcall fn(hash)a +#endif + +#define nv_iord(rv,r,i) /* +*/ nv_mkio(rv,r,i) /* +*/ iord rv I[rv] + +#define nv_iowr(r,i,rv) /* +*/ nv_mkio($r0,r,i) /* +*/ iowr I[$r0] rv /* +*/ clear b32 $r0 + +#define nv_rd32(reg,addr) /* +*/ imm32($r14, addr) /* +*/ call(nv_rd32) /* +*/ mov b32 reg $r15 + +#define nv_wr32(addr,reg) /* +*/ mov b32 $r15 reg /* +*/ imm32($r14, addr) /* +*/ call(nv_wr32) + +#define trace_set(bit) /* +*/ clear b32 $r9 /* +*/ bset $r9 bit /* +*/ nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_SET(7), 0, $r9) + +#define trace_clr(bit) /* +*/ clear b32 $r9 /* +*/ bset $r9 bit /* +*/ nv_iowr(NV_PGRAPH_FECS_CC_SCRATCH_CLR(7), 0, $r9) diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/os.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/os.h new file mode 100644 index 0000000000..6ac155b966 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/os.h @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_GRAPH_OS_H__ +#define __NVKM_GRAPH_OS_H__ + +#define E_BAD_COMMAND 0x00000001 +#define E_CMD_OVERFLOW 0x00000002 +#define E_BAD_FWMTHD 0x00000003 + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c new file mode 100644 index 0000000000..65c332118f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c @@ -0,0 +1,198 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include + +#include + +static const struct nvkm_bitfield nv50_gr_status[] = { + { 0x00000001, "BUSY" }, /* set when any bit is set */ + { 0x00000002, "DISPATCH" }, + { 0x00000004, "UNK2" }, + { 0x00000008, "UNK3" }, + { 0x00000010, "UNK4" }, + { 0x00000020, "UNK5" }, + { 0x00000040, "M2MF" }, + { 0x00000080, "UNK7" }, + { 0x00000100, "CTXPROG" }, + { 0x00000200, "VFETCH" }, + { 0x00000400, "CCACHE_PREGEOM" }, + { 0x00000800, "STRMOUT_VATTR_POSTGEOM" }, + { 0x00001000, "VCLIP" }, + { 0x00002000, "RATTR_APLANE" }, + { 0x00004000, "TRAST" }, + { 0x00008000, "CLIPID" }, + { 0x00010000, "ZCULL" }, + { 0x00020000, "ENG2D" }, + { 0x00040000, "RMASK" }, + { 0x00080000, "TPC_RAST" }, + { 0x00100000, "TPC_PROP" }, + { 0x00200000, "TPC_TEX" }, + { 0x00400000, "TPC_GEOM" }, + { 0x00800000, "TPC_MP" }, + { 0x01000000, "ROP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_0[] = { + { 0x01, "VFETCH" }, + { 0x02, "CCACHE" }, + { 0x04, "PREGEOM" }, + { 0x08, "POSTGEOM" }, + { 0x10, "VATTR" }, + { 0x20, "STRMOUT" }, + { 0x40, "VCLIP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_1[] = { + { 0x01, "TPC_RAST" }, + { 0x02, "TPC_PROP" }, + { 0x04, "TPC_TEX" }, + { 0x08, "TPC_GEOM" }, + { 0x10, "TPC_MP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_2[] = { + { 0x01, "RATTR" }, + { 0x02, "APLANE" }, + { 0x04, "TRAST" }, + { 0x08, "CLIPID" }, + { 0x10, "ZCULL" }, + { 0x20, "ENG2D" }, + { 0x40, "RMASK" }, + { 0x80, "ROP" }, + {} +}; + +static void +nvkm_gr_vstatus_print(struct nv50_gr *gr, int r, + const struct nvkm_bitfield *units, u32 status) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + u32 stat = status; + u8 mask = 0x00; + char msg[64]; + int i; + + for (i = 0; units[i].name && status; i++) { + if ((status & 7) == 1) + mask |= (1 << i); + status >>= 3; + } + + nvkm_snprintbf(msg, sizeof(msg), units, mask); + nvkm_error(subdev, "PGRAPH_VSTATUS%d: %08x [%s]\n", r, stat, msg); +} + +int +g84_gr_tlb_flush(struct nvkm_gr *base) +{ + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_timer *tmr = device->timer; + bool idle, timeout = false; + unsigned long flags; + char status[128]; + u64 start; + u32 tmp; + + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, 0x400500, 0x00000001, 0x00000000); + + start = nvkm_timer_read(tmr); + do { + idle = true; + + for (tmp = nvkm_rd32(device, 0x400380); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + + for (tmp = nvkm_rd32(device, 0x400384); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + + for (tmp = nvkm_rd32(device, 0x400388); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + } while (!idle && + !(timeout = nvkm_timer_read(tmr) - start > 2000000000)); + + if (timeout) { + nvkm_error(subdev, "PGRAPH TLB flush idle timeout fail\n"); + + tmp = nvkm_rd32(device, 0x400700); + nvkm_snprintbf(status, sizeof(status), nv50_gr_status, tmp); + nvkm_error(subdev, "PGRAPH_STATUS %08x [%s]\n", tmp, status); + + nvkm_gr_vstatus_print(gr, 0, nv50_gr_vstatus_0, + nvkm_rd32(device, 0x400380)); + nvkm_gr_vstatus_print(gr, 1, nv50_gr_vstatus_1, + nvkm_rd32(device, 0x400384)); + nvkm_gr_vstatus_print(gr, 2, nv50_gr_vstatus_2, + nvkm_rd32(device, 0x400388)); + } + + + nvkm_wr32(device, 0x100c80, 0x00000001); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x100c80) & 0x00000001)) + break; + ); + nvkm_mask(device, 0x400500, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return timeout ? -EBUSY : 0; +} + +static const struct nvkm_gr_func +g84_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, G82_TESLA, &nv50_gr_object }, + {} + } +}; + +int +g84_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&g84_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ga102.c new file mode 100644 index 0000000000..00cd70abad --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ga102.c @@ -0,0 +1,354 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include +#include +#include +#include + +#include + +#include + +static void +ga102_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 invalid[] = { 0, 0, 0, 0 }, *color; + + if (gr->zbc_color[zbc].format) + color = gr->zbc_color[zbc].l2; + else + color = invalid; + + nvkm_mask(device, 0x41bcb4, 0x0000001f, zbc); + nvkm_wr32(device, 0x41bcec, color[0]); + nvkm_wr32(device, 0x41bcf0, color[1]); + nvkm_wr32(device, 0x41bcf4, color[2]); + nvkm_wr32(device, 0x41bcf8, color[3]); +} + +static const struct gf100_gr_func_zbc +ga102_gr_zbc = { + .clear_color = ga102_gr_zbc_clear_color, + .clear_depth = gp100_gr_zbc_clear_depth, + .stencil_get = gp102_gr_zbc_stencil_get, + .clear_stencil = gp102_gr_zbc_clear_stencil, +}; + +static void +ga102_gr_gpccs_reset(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x41a610, 0x00000000); + nvkm_msec(device, 1, NVKM_DELAY); + nvkm_wr32(device, 0x41a610, 0x00000001); +} + +static const struct nvkm_acr_lsf_func +ga102_gr_gpccs_acr = { + .flags = NVKM_ACR_LSF_FORCE_PRIV_LOAD, + .bl_entry = 0x3400, + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp108_gr_acr_bld_write, + .bld_patch = gp108_gr_acr_bld_patch, +}; + +static void +ga102_gr_fecs_reset(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409614, 0x00000010); + nvkm_wr32(device, 0x41a614, 0x00000020); + nvkm_usec(device, 10, NVKM_DELAY); + nvkm_wr32(device, 0x409614, 0x00000110); + nvkm_wr32(device, 0x41a614, 0x00000a20); + nvkm_usec(device, 10, NVKM_DELAY); + nvkm_rd32(device, 0x409614); + nvkm_rd32(device, 0x41a614); +} + +static const struct nvkm_acr_lsf_func +ga102_gr_fecs_acr = { + .bl_entry = 0x7e00, + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp108_gr_acr_bld_write, + .bld_patch = gp108_gr_acr_bld_patch, +}; + +static void +ga102_gr_init_rop_exceptions(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x41bcbc, 0x40000000); + nvkm_wr32(device, 0x41bc38, 0x40000000); + nvkm_wr32(device, 0x41ac94, nvkm_rd32(device, 0x502c94)); +} + +static void +ga102_gr_init_40a790(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x40a790, 0xc0000000); +} + +static void +ga102_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0xf8001fff); + nvkm_wr32(device, 0x418894, 0x00000000); + + nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); + nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); + nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); +} + +static struct nvkm_intr * +ga102_gr_oneinit_intr(struct gf100_gr *gr, enum nvkm_intr_type *pvector) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + *pvector = nvkm_rd32(device, 0x400154) & 0x00000fff; + return &device->vfn->intr; +} + +static int +ga102_gr_nonstall(struct gf100_gr *gr) +{ + return nvkm_rd32(gr->base.engine.subdev.device, 0x400160) & 0x00000fff; +} + +static const struct gf100_gr_func +ga102_gr = { + .nonstall = ga102_gr_nonstall, + .oneinit_intr = ga102_gr_oneinit_intr, + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gv100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_419bd8 = gv100_gr_init_419bd8, + .init_gpc_mmu = ga102_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = tu102_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fs = tu102_gr_init_fs, + .init_fecs_exceptions = tu102_gr_init_fecs_exceptions, + .init_40a790 = ga102_gr_init_40a790, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_504430 = gv100_gr_init_504430, + .init_shader_exceptions = gv100_gr_init_shader_exceptions, + .init_rop_exceptions = ga102_gr_init_rop_exceptions, + .init_4188a4 = gv100_gr_init_4188a4, + .trap_mp = gv100_gr_trap_mp, + .fecs.reset = ga102_gr_fecs_reset, + .gpccs.reset = ga102_gr_gpccs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 7, + .tpc_nr = 6, + .ppc_nr = 3, + .grctx = &ga102_grctx, + .zbc = &ga102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, AMPERE_B, &gf100_fermi }, + { -1, -1, AMPERE_COMPUTE_B }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/ga102/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga102/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga102/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga102/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga102/gr/NET_img.bin"); + +MODULE_FIRMWARE("nvidia/ga103/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga103/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga103/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga103/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga103/gr/NET_img.bin"); + +MODULE_FIRMWARE("nvidia/ga104/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga104/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga104/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga104/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga104/gr/NET_img.bin"); + +MODULE_FIRMWARE("nvidia/ga106/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga106/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga106/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga106/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga106/gr/NET_img.bin"); + +MODULE_FIRMWARE("nvidia/ga107/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga107/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga107/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/ga107/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/ga107/gr/NET_img.bin"); + +struct netlist_region { + u32 region_id; + u32 data_size; + u32 data_offset; +}; + +struct netlist_image_header { + u32 version; + u32 regions; +}; + +struct netlist_image { + struct netlist_image_header header; + struct netlist_region regions[]; +}; + +struct netlist_av64 { + u32 addr; + u32 data_hi; + u32 data_lo; +}; + +static int +ga102_gr_av64_to_init(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + int nent; + int i; + + nent = (blob->size / sizeof(struct netlist_av64)); + pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); + if (!pack) + return -ENOMEM; + + init = (void *)(pack + 2); + pack[0].init = init; + pack[0].type = 64; + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct netlist_av64 *av = &((struct netlist_av64 *)blob->data)[i]; + + ent->addr = av->addr; + ent->data = ((u64)av->data_hi << 32) | av->data_lo; + ent->count = 1; + ent->pitch = 1; + } + + *ppack = pack; + return 0; +} + +static int +ga102_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + const struct firmware *fw; + const struct netlist_image *net; + const struct netlist_region *fecs_inst = NULL; + const struct netlist_region *fecs_data = NULL; + const struct netlist_region *gpccs_inst = NULL; + const struct netlist_region *gpccs_data = NULL; + int ret, i; + + ret = nvkm_firmware_get(subdev, "gr/NET_img", 0, &fw); + if (ret) + return ret; + + net = (const void *)fw->data; + nvkm_debug(subdev, "netlist version %d, %d regions\n", + net->header.version, net->header.regions); + + for (i = 0; i < net->header.regions; i++) { + const struct netlist_region *reg = &net->regions[i]; + struct nvkm_blob blob = { + .data = (void *)fw->data + reg->data_offset, + .size = reg->data_size, + }; + + nvkm_debug(subdev, "\t%2d: %08x %08x\n", + reg->region_id, reg->data_offset, reg->data_size); + + switch (reg->region_id) { + case 0: fecs_data = reg; break; + case 1: fecs_inst = reg; break; + case 2: gpccs_data = reg; break; + case 3: gpccs_inst = reg; break; + case 4: gk20a_gr_av_to_init(&blob, &gr->bundle); break; + case 5: gk20a_gr_aiv_to_init(&blob, &gr->sw_ctx); break; + case 7: gk20a_gr_av_to_method(&blob, &gr->method); break; + case 28: tu102_gr_av_to_init_veid(&blob, &gr->bundle_veid); break; + case 34: ga102_gr_av64_to_init(&blob, &gr->bundle64); break; + case 48: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx1); break; + case 49: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx2); break; + case 50: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx3); break; + case 51: gk20a_gr_av_to_init(&blob, &gr->sw_nonctx4); break; + default: + break; + } + } + + ret = nvkm_acr_lsfw_load_bl_sig_net(subdev, &gr->fecs.falcon, NVKM_ACR_LSF_FECS, + "gr/fecs_", ver, fwif->fecs, + fw->data + fecs_inst->data_offset, + fecs_inst->data_size, + fw->data + fecs_data->data_offset, + fecs_data->data_size); + if (ret) + return ret; + + ret = nvkm_acr_lsfw_load_bl_sig_net(subdev, &gr->gpccs.falcon, NVKM_ACR_LSF_GPCCS, + "gr/gpccs_", ver, fwif->gpccs, + fw->data + gpccs_inst->data_offset, + gpccs_inst->data_size, + fw->data + gpccs_data->data_offset, + gpccs_data->data_size); + if (ret) + return ret; + + gr->firmware = true; + + nvkm_firmware_put(fw); + return 0; +} + +static const struct gf100_gr_fwif +ga102_gr_fwif[] = { + { 0, ga102_gr_load, &ga102_gr, &ga102_gr_fecs_acr, &ga102_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +ga102_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(ga102_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c new file mode 100644 index 0000000000..3648868bb9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c @@ -0,0 +1,2659 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" +#include "fuc/os.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +/******************************************************************************* + * Zero Bandwidth Clear + ******************************************************************************/ + +static void +gf100_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + if (gr->zbc_color[zbc].format) { + nvkm_wr32(device, 0x405804, gr->zbc_color[zbc].ds[0]); + nvkm_wr32(device, 0x405808, gr->zbc_color[zbc].ds[1]); + nvkm_wr32(device, 0x40580c, gr->zbc_color[zbc].ds[2]); + nvkm_wr32(device, 0x405810, gr->zbc_color[zbc].ds[3]); + } + nvkm_wr32(device, 0x405814, gr->zbc_color[zbc].format); + nvkm_wr32(device, 0x405820, zbc); + nvkm_wr32(device, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */ +} + +static int +gf100_gr_zbc_color_get(struct gf100_gr *gr, int format, + const u32 ds[4], const u32 l2[4]) +{ + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; + int zbc = -ENOSPC, i; + + for (i = ltc->zbc_color_min; i <= ltc->zbc_color_max; i++) { + if (gr->zbc_color[i].format) { + if (gr->zbc_color[i].format != format) + continue; + if (memcmp(gr->zbc_color[i].ds, ds, sizeof( + gr->zbc_color[i].ds))) + continue; + if (memcmp(gr->zbc_color[i].l2, l2, sizeof( + gr->zbc_color[i].l2))) { + WARN_ON(1); + return -EINVAL; + } + return i; + } else { + zbc = (zbc < 0) ? i : zbc; + } + } + + if (zbc < 0) + return zbc; + + memcpy(gr->zbc_color[zbc].ds, ds, sizeof(gr->zbc_color[zbc].ds)); + memcpy(gr->zbc_color[zbc].l2, l2, sizeof(gr->zbc_color[zbc].l2)); + gr->zbc_color[zbc].format = format; + nvkm_ltc_zbc_color_get(ltc, zbc, l2); + gr->func->zbc->clear_color(gr, zbc); + return zbc; +} + +static void +gf100_gr_zbc_clear_depth(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + if (gr->zbc_depth[zbc].format) + nvkm_wr32(device, 0x405818, gr->zbc_depth[zbc].ds); + nvkm_wr32(device, 0x40581c, gr->zbc_depth[zbc].format); + nvkm_wr32(device, 0x405820, zbc); + nvkm_wr32(device, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */ +} + +static int +gf100_gr_zbc_depth_get(struct gf100_gr *gr, int format, + const u32 ds, const u32 l2) +{ + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; + int zbc = -ENOSPC, i; + + for (i = ltc->zbc_depth_min; i <= ltc->zbc_depth_max; i++) { + if (gr->zbc_depth[i].format) { + if (gr->zbc_depth[i].format != format) + continue; + if (gr->zbc_depth[i].ds != ds) + continue; + if (gr->zbc_depth[i].l2 != l2) { + WARN_ON(1); + return -EINVAL; + } + return i; + } else { + zbc = (zbc < 0) ? i : zbc; + } + } + + if (zbc < 0) + return zbc; + + gr->zbc_depth[zbc].format = format; + gr->zbc_depth[zbc].ds = ds; + gr->zbc_depth[zbc].l2 = l2; + nvkm_ltc_zbc_depth_get(ltc, zbc, l2); + gr->func->zbc->clear_depth(gr, zbc); + return zbc; +} + +const struct gf100_gr_func_zbc +gf100_gr_zbc = { + .clear_color = gf100_gr_zbc_clear_color, + .clear_depth = gf100_gr_zbc_clear_depth, +}; + +/******************************************************************************* + * Graphics object classes + ******************************************************************************/ +#define gf100_gr_object(p) container_of((p), struct gf100_gr_object, object) + +struct gf100_gr_object { + struct nvkm_object object; + struct gf100_gr_chan *chan; +}; + +static int +gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) +{ + struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); + union { + struct fermi_a_zbc_color_v0 v0; + } *args = data; + int ret = -ENOSYS; + + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + switch (args->v0.format) { + case FERMI_A_ZBC_COLOR_V0_FMT_ZERO: + case FERMI_A_ZBC_COLOR_V0_FMT_UNORM_ONE: + case FERMI_A_ZBC_COLOR_V0_FMT_RF32_GF32_BF32_AF32: + case FERMI_A_ZBC_COLOR_V0_FMT_R16_G16_B16_A16: + case FERMI_A_ZBC_COLOR_V0_FMT_RN16_GN16_BN16_AN16: + case FERMI_A_ZBC_COLOR_V0_FMT_RS16_GS16_BS16_AS16: + case FERMI_A_ZBC_COLOR_V0_FMT_RU16_GU16_BU16_AU16: + case FERMI_A_ZBC_COLOR_V0_FMT_RF16_GF16_BF16_AF16: + case FERMI_A_ZBC_COLOR_V0_FMT_A8R8G8B8: + case FERMI_A_ZBC_COLOR_V0_FMT_A8RL8GL8BL8: + case FERMI_A_ZBC_COLOR_V0_FMT_A2B10G10R10: + case FERMI_A_ZBC_COLOR_V0_FMT_AU2BU10GU10RU10: + case FERMI_A_ZBC_COLOR_V0_FMT_A8B8G8R8: + case FERMI_A_ZBC_COLOR_V0_FMT_A8BL8GL8RL8: + case FERMI_A_ZBC_COLOR_V0_FMT_AN8BN8GN8RN8: + case FERMI_A_ZBC_COLOR_V0_FMT_AS8BS8GS8RS8: + case FERMI_A_ZBC_COLOR_V0_FMT_AU8BU8GU8RU8: + case FERMI_A_ZBC_COLOR_V0_FMT_A2R10G10B10: + case FERMI_A_ZBC_COLOR_V0_FMT_BF10GF11RF11: + ret = gf100_gr_zbc_color_get(gr, args->v0.format, + args->v0.ds, + args->v0.l2); + if (ret >= 0) { + args->v0.index = ret; + return 0; + } + break; + default: + return -EINVAL; + } + } + + return ret; +} + +static int +gf100_fermi_mthd_zbc_depth(struct nvkm_object *object, void *data, u32 size) +{ + struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); + union { + struct fermi_a_zbc_depth_v0 v0; + } *args = data; + int ret = -ENOSYS; + + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + switch (args->v0.format) { + case FERMI_A_ZBC_DEPTH_V0_FMT_FP32: + ret = gf100_gr_zbc_depth_get(gr, args->v0.format, + args->v0.ds, + args->v0.l2); + return (ret >= 0) ? 0 : -ENOSPC; + default: + return -EINVAL; + } + } + + return ret; +} + +static int +gf100_fermi_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + nvif_ioctl(object, "fermi mthd %08x\n", mthd); + switch (mthd) { + case FERMI_A_ZBC_COLOR: + return gf100_fermi_mthd_zbc_color(object, data, size); + case FERMI_A_ZBC_DEPTH: + return gf100_fermi_mthd_zbc_depth(object, data, size); + default: + break; + } + return -EINVAL; +} + +const struct nvkm_object_func +gf100_fermi = { + .mthd = gf100_fermi_mthd, +}; + +static void +gf100_gr_mthd_set_shader_exceptions(struct nvkm_device *device, u32 data) +{ + nvkm_wr32(device, 0x419e44, data ? 0xffffffff : 0x00000000); + nvkm_wr32(device, 0x419e4c, data ? 0xffffffff : 0x00000000); +} + +static bool +gf100_gr_mthd_sw(struct nvkm_device *device, u16 class, u32 mthd, u32 data) +{ + switch (class & 0x00ff) { + case 0x97: + case 0xc0: + switch (mthd) { + case 0x1528: + gf100_gr_mthd_set_shader_exceptions(device, data); + return true; + default: + break; + } + break; + default: + break; + } + return false; +} + +static const struct nvkm_object_func +gf100_gr_object_func = { +}; + +static int +gf100_gr_object_new(const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + struct gf100_gr_chan *chan = gf100_gr_chan(oclass->parent); + struct gf100_gr_object *object; + + if (!(object = kzalloc(sizeof(*object), GFP_KERNEL))) + return -ENOMEM; + *pobject = &object->object; + + nvkm_object_ctor(oclass->base.func ? oclass->base.func : + &gf100_gr_object_func, oclass, &object->object); + object->chan = chan; + return 0; +} + +static int +gf100_gr_object_get(struct nvkm_gr *base, int index, struct nvkm_sclass *sclass) +{ + struct gf100_gr *gr = gf100_gr(base); + int c = 0; + + while (gr->func->sclass[c].oclass) { + if (c++ == index) { + *sclass = gr->func->sclass[index]; + sclass->ctor = gf100_gr_object_new; + return index; + } + } + + return c; +} + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static int +gf100_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct gf100_gr_chan *chan = gf100_gr_chan(object); + struct gf100_gr *gr = chan->gr; + int ret, i; + + ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, false, parent, pgpuobj); + if (ret) + return ret; + + nvkm_kmap(*pgpuobj); + for (i = 0; i < gr->size; i += 4) + nvkm_wo32(*pgpuobj, i, gr->data[i / 4]); + + if (!gr->firmware) { + nvkm_wo32(*pgpuobj, 0x00, chan->mmio_nr / 2); + nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma->addr >> 8); + } else { + nvkm_wo32(*pgpuobj, 0xf4, 0); + nvkm_wo32(*pgpuobj, 0xf8, 0); + nvkm_wo32(*pgpuobj, 0x10, chan->mmio_nr / 2); + nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma->addr)); + nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma->addr)); + nvkm_wo32(*pgpuobj, 0x1c, 1); + nvkm_wo32(*pgpuobj, 0x20, 0); + nvkm_wo32(*pgpuobj, 0x28, 0); + nvkm_wo32(*pgpuobj, 0x2c, 0); + } + nvkm_done(*pgpuobj); + return 0; +} + +static void * +gf100_gr_chan_dtor(struct nvkm_object *object) +{ + struct gf100_gr_chan *chan = gf100_gr_chan(object); + + nvkm_vmm_put(chan->vmm, &chan->mmio_vma); + nvkm_memory_unref(&chan->mmio); + + nvkm_vmm_put(chan->vmm, &chan->attrib_cb); + nvkm_vmm_put(chan->vmm, &chan->unknown); + nvkm_vmm_put(chan->vmm, &chan->bundle_cb); + nvkm_vmm_put(chan->vmm, &chan->pagepool); + nvkm_vmm_unref(&chan->vmm); + return chan; +} + +static const struct nvkm_object_func +gf100_gr_chan = { + .dtor = gf100_gr_chan_dtor, + .bind = gf100_gr_chan_bind, +}; + +static int +gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct gf100_gr *gr = gf100_gr(base); + struct gf100_gr_chan *chan; + struct gf100_vmm_map_v0 args = { .priv = 1 }; + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&gf100_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->vmm = nvkm_vmm_ref(fifoch->vmm); + *pobject = &chan->object; + + /* Map pagepool. */ + ret = nvkm_vmm_get(chan->vmm, 12, nvkm_memory_size(gr->pagepool), &chan->pagepool); + if (ret) + return ret; + + ret = nvkm_memory_map(gr->pagepool, 0, chan->vmm, chan->pagepool, &args, sizeof(args)); + if (ret) + return ret; + + /* Map bundle circular buffer. */ + ret = nvkm_vmm_get(chan->vmm, 12, nvkm_memory_size(gr->bundle_cb), &chan->bundle_cb); + if (ret) + return ret; + + ret = nvkm_memory_map(gr->bundle_cb, 0, chan->vmm, chan->bundle_cb, &args, sizeof(args)); + if (ret) + return ret; + + /* Map attribute circular buffer. */ + ret = nvkm_vmm_get(chan->vmm, 12, nvkm_memory_size(gr->attrib_cb), &chan->attrib_cb); + if (ret) + return ret; + + if (device->card_type < GP100) { + ret = nvkm_memory_map(gr->attrib_cb, 0, chan->vmm, chan->attrib_cb, NULL, 0); + if (ret) + return ret; + } else { + ret = nvkm_memory_map(gr->attrib_cb, 0, chan->vmm, chan->attrib_cb, + &args, sizeof(args));; + if (ret) + return ret; + } + + /* Map some context buffer of unknown purpose. */ + if (gr->func->grctx->unknown_size) { + ret = nvkm_vmm_get(chan->vmm, 12, nvkm_memory_size(gr->unknown), &chan->unknown); + if (ret) + return ret; + + ret = nvkm_memory_map(gr->unknown, 0, chan->vmm, chan->unknown, + &args, sizeof(args)); + if (ret) + return ret; + } + + /* Generate golden context image. */ + mutex_lock(&gr->fecs.mutex); + if (gr->data == NULL) { + ret = gf100_grctx_generate(gr, chan, fifoch->inst); + if (ret) { + nvkm_error(&base->engine.subdev, "failed to construct context\n"); + return ret; + } + } + mutex_unlock(&gr->fecs.mutex); + + /* allocate memory for a "mmio list" buffer that's used by the HUB + * fuc to modify some per-context register settings on first load + * of the context. + */ + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x100, + false, &chan->mmio); + if (ret) + return ret; + + ret = nvkm_vmm_get(fifoch->vmm, 12, 0x1000, &chan->mmio_vma); + if (ret) + return ret; + + ret = nvkm_memory_map(chan->mmio, 0, fifoch->vmm, + chan->mmio_vma, &args, sizeof(args)); + if (ret) + return ret; + + /* finally, fill in the mmio list and point the context at it */ + nvkm_kmap(chan->mmio); + gr->func->grctx->pagepool(chan, chan->pagepool->addr); + gr->func->grctx->bundle(chan, chan->bundle_cb->addr, gr->func->grctx->bundle_size); + gr->func->grctx->attrib_cb(chan, chan->attrib_cb->addr, gr->func->grctx->attrib_cb_size(gr)); + gr->func->grctx->attrib(chan); + if (gr->func->grctx->patch_ltc) + gr->func->grctx->patch_ltc(chan); + if (gr->func->grctx->unknown_size) + gr->func->grctx->unknown(chan, chan->unknown->addr, gr->func->grctx->unknown_size); + nvkm_done(chan->mmio); + return 0; +} + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gf100_gr_init_main_0[] = { + { 0x400080, 1, 0x04, 0x003083c2 }, + { 0x400088, 1, 0x04, 0x00006fe7 }, + { 0x40008c, 1, 0x04, 0x00000000 }, + { 0x400090, 1, 0x04, 0x00000030 }, + { 0x40013c, 1, 0x04, 0x013901f7 }, + { 0x400140, 1, 0x04, 0x00000100 }, + { 0x400144, 1, 0x04, 0x00000000 }, + { 0x400148, 1, 0x04, 0x00000110 }, + { 0x400138, 1, 0x04, 0x00000000 }, + { 0x400130, 2, 0x04, 0x00000000 }, + { 0x400124, 1, 0x04, 0x00000002 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_fe_0[] = { + { 0x40415c, 1, 0x04, 0x00000000 }, + { 0x404170, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_pri_0[] = { + { 0x404488, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_rstr2d_0[] = { + { 0x407808, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_pd_0[] = { + { 0x406024, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405908, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_scc_0[] = { + { 0x40803c, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_prop_0[] = { + { 0x4184a0, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_gpc_unk_0[] = { + { 0x418604, 1, 0x04, 0x00000000 }, + { 0x418680, 1, 0x04, 0x00000000 }, + { 0x418714, 1, 0x04, 0x80000000 }, + { 0x418384, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_setup_0[] = { + { 0x418814, 3, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_crstr_0[] = { + { 0x418b04, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_setup_1[] = { + { 0x4188c8, 1, 0x04, 0x80000000 }, + { 0x4188cc, 1, 0x04, 0x00000000 }, + { 0x4188d0, 1, 0x04, 0x00010000 }, + { 0x4188d4, 1, 0x04, 0x00000001 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_zcull_0[] = { + { 0x418910, 1, 0x04, 0x00010001 }, + { 0x418914, 1, 0x04, 0x00000301 }, + { 0x418918, 1, 0x04, 0x00800000 }, + { 0x418980, 1, 0x04, 0x77777770 }, + { 0x418984, 3, 0x04, 0x77777777 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_gpm_0[] = { + { 0x418c04, 1, 0x04, 0x00000000 }, + { 0x418c88, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x00000050 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_gcc_0[] = { + { 0x41900c, 1, 0x04, 0x00000000 }, + { 0x419018, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_tpccs_0[] = { + { 0x419d08, 2, 0x04, 0x00000000 }, + { 0x419d10, 1, 0x04, 0x00000014 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419abc, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_pe_0[] = { + { 0x41980c, 3, 0x04, 0x00000000 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x41984c, 1, 0x04, 0x00005bc5 }, + { 0x419850, 4, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419ca8, 1, 0x04, 0x80000000 }, + { 0x419cb4, 1, 0x04, 0x00000000 }, + { 0x419cb8, 1, 0x04, 0x00008bf4 }, + { 0x419cbc, 1, 0x04, 0x28137606 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_wwdx_0[] = { + { 0x419bd4, 1, 0x04, 0x00800000 }, + { 0x419bdc, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_tpccs_1[] = { + { 0x419d2c, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_mpc_0[] = { + { 0x419c0c, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf100_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00001100 }, + { 0x419eac, 1, 0x04, 0x11100702 }, + { 0x419eb0, 1, 0x04, 0x00000003 }, + { 0x419eb4, 4, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x06060618 }, + { 0x419ed0, 1, 0x04, 0x0eff0e38 }, + { 0x419ed4, 1, 0x04, 0x011104f1 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419f2c, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_be_0[] = { + { 0x40880c, 1, 0x04, 0x00000000 }, + { 0x408910, 9, 0x04, 0x00000000 }, + { 0x408950, 1, 0x04, 0x00000000 }, + { 0x408954, 1, 0x04, 0x0000ffff }, + { 0x408984, 1, 0x04, 0x00000000 }, + { 0x408988, 1, 0x04, 0x08040201 }, + { 0x40898c, 1, 0x04, 0x80402010 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_fe_1[] = { + { 0x4040f0, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf100_gr_init_pe_1[] = { + { 0x419880, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_pack +gf100_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf100_gr_init_pd_0 }, + { gf100_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf100_gr_init_prop_0 }, + { gf100_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf100_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf100_gr_init_gpm_0 }, + { gf100_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf100_gr_init_tex_0 }, + { gf100_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf100_gr_init_wwdx_0 }, + { gf100_gr_init_tpccs_1 }, + { gf100_gr_init_mpc_0 }, + { gf100_gr_init_sm_0 }, + { gf100_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + { gf100_gr_init_pe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static u32 +gf100_gr_ctxsw_inst(struct nvkm_gr *gr) +{ + return nvkm_rd32(gr->engine.subdev.device, 0x409b00); +} + +static int +gf100_gr_fecs_ctrl_ctxsw(struct gf100_gr *gr, u32 mthd) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409804, 0xffffffff); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0xffffffff); + nvkm_wr32(device, 0x409504, mthd); + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x409804); + if (stat == 0x00000002) + return -EIO; + if (stat == 0x00000001) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_start_ctxsw(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + int ret = 0; + + mutex_lock(&gr->fecs.mutex); + if (!--gr->fecs.disable) { + if (WARN_ON(ret = gf100_gr_fecs_ctrl_ctxsw(gr, 0x39))) + gr->fecs.disable++; + } + mutex_unlock(&gr->fecs.mutex); + return ret; +} + +static int +gf100_gr_fecs_stop_ctxsw(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + int ret = 0; + + mutex_lock(&gr->fecs.mutex); + if (!gr->fecs.disable++) { + if (WARN_ON(ret = gf100_gr_fecs_ctrl_ctxsw(gr, 0x38))) + gr->fecs.disable--; + } + mutex_unlock(&gr->fecs.mutex); + return ret; +} + +static int +gf100_gr_fecs_halt_pipeline(struct gf100_gr *gr) +{ + int ret = 0; + + if (gr->firmware) { + mutex_lock(&gr->fecs.mutex); + ret = gf100_gr_fecs_ctrl_ctxsw(gr, 0x04); + mutex_unlock(&gr->fecs.mutex); + } + + return ret; +} + +int +gf100_gr_fecs_wfi_golden_save(struct gf100_gr *gr, u32 inst) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_mask(device, 0x409800, 0x00000003, 0x00000000); + nvkm_wr32(device, 0x409500, inst); + nvkm_wr32(device, 0x409504, 0x00000009); + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x409800); + if (stat & 0x00000002) + return -EIO; + if (stat & 0x00000001) + return 0; + ); + + return -ETIMEDOUT; +} + +int +gf100_gr_fecs_bind_pointer(struct gf100_gr *gr, u32 inst) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_mask(device, 0x409800, 0x00000030, 0x00000000); + nvkm_wr32(device, 0x409500, inst); + nvkm_wr32(device, 0x409504, 0x00000003); + nvkm_msec(device, 2000, + u32 stat = nvkm_rd32(device, 0x409800); + if (stat & 0x00000020) + return -EIO; + if (stat & 0x00000010) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_set_reglist_virtual_address(struct gf100_gr *gr, u64 addr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409810, addr >> 8); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000032); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) == 0x00000001) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_set_reglist_bind_instance(struct gf100_gr *gr, u32 inst) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409810, inst); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000031); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) == 0x00000001) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_discover_reglist_image_size(struct gf100_gr *gr, u32 *psize) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000030); + nvkm_msec(device, 2000, + if ((*psize = nvkm_rd32(device, 0x409800))) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_elpg_bind(struct gf100_gr *gr) +{ + u32 size; + int ret; + + ret = gf100_gr_fecs_discover_reglist_image_size(gr, &size); + if (ret) + return ret; + + /*XXX: We need to allocate + map the above into PMU's inst block, + * which which means we probably need a proper PMU before we + * even bother. + */ + + ret = gf100_gr_fecs_set_reglist_bind_instance(gr, 0); + if (ret) + return ret; + + return gf100_gr_fecs_set_reglist_virtual_address(gr, 0); +} + +static int +gf100_gr_fecs_discover_pm_image_size(struct gf100_gr *gr, u32 *psize) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000025); + nvkm_msec(device, 2000, + if ((*psize = nvkm_rd32(device, 0x409800))) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_discover_zcull_image_size(struct gf100_gr *gr, u32 *psize) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000016); + nvkm_msec(device, 2000, + if ((*psize = nvkm_rd32(device, 0x409800))) + return 0; + ); + + return -ETIMEDOUT; +} + +static int +gf100_gr_fecs_discover_image_size(struct gf100_gr *gr, u32 *psize) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000010); + nvkm_msec(device, 2000, + if ((*psize = nvkm_rd32(device, 0x409800))) + return 0; + ); + + return -ETIMEDOUT; +} + +static void +gf100_gr_fecs_set_watchdog_timeout(struct gf100_gr *gr, u32 timeout) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, timeout); + nvkm_wr32(device, 0x409504, 0x00000021); +} + +static bool +gf100_gr_chsw_load(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + if (!gr->firmware) { + u32 trace = nvkm_rd32(gr->base.engine.subdev.device, 0x40981c); + if (trace & 0x00000040) + return true; + } else { + u32 mthd = nvkm_rd32(gr->base.engine.subdev.device, 0x409808); + if (mthd & 0x00080000) + return true; + } + return false; +} + +int +gf100_gr_rops(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + return (nvkm_rd32(device, 0x409604) & 0x001f0000) >> 16; +} + +void +gf100_gr_zbc_init(struct gf100_gr *gr) +{ + const u32 zero[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; + const u32 one[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, + 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; + const u32 f32_0[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; + const u32 f32_1[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, + 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; + int index, c = ltc->zbc_color_min, d = ltc->zbc_depth_min, s = ltc->zbc_depth_min; + + if (!gr->zbc_color[0].format) { + gf100_gr_zbc_color_get(gr, 1, & zero[0], &zero[4]); c++; + gf100_gr_zbc_color_get(gr, 2, & one[0], &one[4]); c++; + gf100_gr_zbc_color_get(gr, 4, &f32_0[0], &f32_0[4]); c++; + gf100_gr_zbc_color_get(gr, 4, &f32_1[0], &f32_1[4]); c++; + gf100_gr_zbc_depth_get(gr, 1, 0x00000000, 0x00000000); d++; + gf100_gr_zbc_depth_get(gr, 1, 0x3f800000, 0x3f800000); d++; + if (gr->func->zbc->stencil_get) { + gr->func->zbc->stencil_get(gr, 1, 0x00, 0x00); s++; + gr->func->zbc->stencil_get(gr, 1, 0x01, 0x01); s++; + gr->func->zbc->stencil_get(gr, 1, 0xff, 0xff); s++; + } + } + + for (index = c; index <= ltc->zbc_color_max; index++) + gr->func->zbc->clear_color(gr, index); + for (index = d; index <= ltc->zbc_depth_max; index++) + gr->func->zbc->clear_depth(gr, index); + + if (gr->func->zbc->clear_stencil) { + for (index = s; index <= ltc->zbc_depth_max; index++) + gr->func->zbc->clear_stencil(gr, index); + } +} + +/** + * Wait until GR goes idle. GR is considered idle if it is disabled by the + * MC (0x200) register, or GR is not busy and a context switch is not in + * progress. + */ +int +gf100_gr_wait_idle(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + unsigned long end_jiffies = jiffies + msecs_to_jiffies(2000); + bool gr_enabled, ctxsw_active, gr_busy; + + do { + /* + * required to make sure FIFO_ENGINE_STATUS (0x2640) is + * up-to-date + */ + nvkm_rd32(device, 0x400700); + + gr_enabled = nvkm_rd32(device, 0x200) & 0x1000; + ctxsw_active = nvkm_fifo_ctxsw_in_progress(&gr->base.engine); + gr_busy = nvkm_rd32(device, 0x40060c) & 0x1; + + if (!gr_enabled || (!gr_busy && !ctxsw_active)) + return 0; + } while (time_before(jiffies, end_jiffies)); + + nvkm_error(subdev, + "wait for idle timeout (en: %d, ctxsw: %d, busy: %d)\n", + gr_enabled, ctxsw_active, gr_busy); + return -EAGAIN; +} + +void +gf100_gr_mmio(struct gf100_gr *gr, const struct gf100_gr_pack *p) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_gr_pack *pack; + const struct gf100_gr_init *init; + + pack_for_each_init(init, pack, p) { + u32 next = init->addr + init->count * init->pitch; + u32 addr = init->addr; + while (addr < next) { + nvkm_wr32(device, addr, init->data); + addr += init->pitch; + } + } +} + +void +gf100_gr_icmd(struct gf100_gr *gr, const struct gf100_gr_pack *p) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_gr_pack *pack; + const struct gf100_gr_init *init; + u64 data = 0; + + nvkm_wr32(device, 0x400208, 0x80000000); + + pack_for_each_init(init, pack, p) { + u32 next = init->addr + init->count * init->pitch; + u32 addr = init->addr; + + if ((pack == p && init == p->init) || data != init->data) { + nvkm_wr32(device, 0x400204, init->data); + if (pack->type == 64) + nvkm_wr32(device, 0x40020c, upper_32_bits(init->data)); + data = init->data; + } + + while (addr < next) { + nvkm_wr32(device, 0x400200, addr); + /** + * Wait for GR to go idle after submitting a + * GO_IDLE bundle + */ + if ((addr & 0xffff) == 0xe100) + gf100_gr_wait_idle(gr); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x400700) & 0x00000004)) + break; + ); + addr += init->pitch; + } + } + + nvkm_wr32(device, 0x400208, 0x00000000); +} + +void +gf100_gr_mthd(struct gf100_gr *gr, const struct gf100_gr_pack *p) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_gr_pack *pack; + const struct gf100_gr_init *init; + u32 data = 0; + + pack_for_each_init(init, pack, p) { + u32 ctrl = 0x80000000 | pack->type; + u32 next = init->addr + init->count * init->pitch; + u32 addr = init->addr; + + if ((pack == p && init == p->init) || data != init->data) { + nvkm_wr32(device, 0x40448c, init->data); + data = init->data; + } + + while (addr < next) { + nvkm_wr32(device, 0x404488, ctrl | (addr << 14)); + addr += init->pitch; + } + } +} + +u64 +gf100_gr_units(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + u64 cfg; + + cfg = (u32)gr->gpc_nr; + cfg |= (u32)gr->tpc_total << 8; + cfg |= (u64)gr->rop_nr << 32; + + return cfg; +} + +static const struct nvkm_bitfield gf100_dispatch_error[] = { + { 0x00000001, "INJECTED_BUNDLE_ERROR" }, + { 0x00000002, "CLASS_SUBCH_MISMATCH" }, + { 0x00000004, "SUBCHSW_DURING_NOTIFY" }, + {} +}; + +static const struct nvkm_bitfield gf100_m2mf_error[] = { + { 0x00000001, "PUSH_TOO_MUCH_DATA" }, + { 0x00000002, "PUSH_NOT_ENOUGH_DATA" }, + {} +}; + +static const struct nvkm_bitfield gf100_unk6_error[] = { + { 0x00000001, "TEMP_TOO_SMALL" }, + {} +}; + +static const struct nvkm_bitfield gf100_ccache_error[] = { + { 0x00000001, "INTR" }, + { 0x00000002, "LDCONST_OOB" }, + {} +}; + +static const struct nvkm_bitfield gf100_macro_error[] = { + { 0x00000001, "TOO_FEW_PARAMS" }, + { 0x00000002, "TOO_MANY_PARAMS" }, + { 0x00000004, "ILLEGAL_OPCODE" }, + { 0x00000008, "DOUBLE_BRANCH" }, + { 0x00000010, "WATCHDOG" }, + {} +}; + +static const struct nvkm_bitfield gk104_sked_error[] = { + { 0x00000040, "CTA_RESUME" }, + { 0x00000080, "CONSTANT_BUFFER_SIZE" }, + { 0x00000200, "LOCAL_MEMORY_SIZE_POS" }, + { 0x00000400, "LOCAL_MEMORY_SIZE_NEG" }, + { 0x00000800, "WARP_CSTACK_SIZE" }, + { 0x00001000, "TOTAL_TEMP_SIZE" }, + { 0x00002000, "REGISTER_COUNT" }, + { 0x00040000, "TOTAL_THREADS" }, + { 0x00100000, "PROGRAM_OFFSET" }, + { 0x00200000, "SHARED_MEMORY_SIZE" }, + { 0x00800000, "CTA_THREAD_DIMENSION_ZERO" }, + { 0x01000000, "MEMORY_WINDOW_OVERLAP" }, + { 0x02000000, "SHARED_CONFIG_TOO_SMALL" }, + { 0x04000000, "TOTAL_REGISTER_COUNT" }, + {} +}; + +static const struct nvkm_bitfield gf100_gpc_rop_error[] = { + { 0x00000002, "RT_PITCH_OVERRUN" }, + { 0x00000010, "RT_WIDTH_OVERRUN" }, + { 0x00000020, "RT_HEIGHT_OVERRUN" }, + { 0x00000080, "ZETA_STORAGE_TYPE_MISMATCH" }, + { 0x00000100, "RT_STORAGE_TYPE_MISMATCH" }, + { 0x00000400, "RT_LINEAR_MISMATCH" }, + {} +}; + +static void +gf100_gr_trap_gpc_rop(struct gf100_gr *gr, int gpc) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + char error[128]; + u32 trap[4]; + + trap[0] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0420)) & 0x3fffffff; + trap[1] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0434)); + trap[2] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0438)); + trap[3] = nvkm_rd32(device, GPC_UNIT(gpc, 0x043c)); + + nvkm_snprintbf(error, sizeof(error), gf100_gpc_rop_error, trap[0]); + + nvkm_error(subdev, "GPC%d/PROP trap: %08x [%s] x = %u, y = %u, " + "format = %x, storage type = %x\n", + gpc, trap[0], error, trap[1] & 0xffff, trap[1] >> 16, + (trap[2] >> 8) & 0x3f, trap[3] & 0xff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); +} + +const struct nvkm_enum gf100_mp_warp_error[] = { + { 0x01, "STACK_ERROR" }, + { 0x02, "API_STACK_ERROR" }, + { 0x03, "RET_EMPTY_STACK_ERROR" }, + { 0x04, "PC_WRAP" }, + { 0x05, "MISALIGNED_PC" }, + { 0x06, "PC_OVERFLOW" }, + { 0x07, "MISALIGNED_IMMC_ADDR" }, + { 0x08, "MISALIGNED_REG" }, + { 0x09, "ILLEGAL_INSTR_ENCODING" }, + { 0x0a, "ILLEGAL_SPH_INSTR_COMBO" }, + { 0x0b, "ILLEGAL_INSTR_PARAM" }, + { 0x0c, "INVALID_CONST_ADDR" }, + { 0x0d, "OOR_REG" }, + { 0x0e, "OOR_ADDR" }, + { 0x0f, "MISALIGNED_ADDR" }, + { 0x10, "INVALID_ADDR_SPACE" }, + { 0x11, "ILLEGAL_INSTR_PARAM2" }, + { 0x12, "INVALID_CONST_ADDR_LDC" }, + { 0x13, "GEOMETRY_SM_ERROR" }, + { 0x14, "DIVERGENT" }, + { 0x15, "WARP_EXIT" }, + {} +}; + +const struct nvkm_bitfield gf100_mp_global_error[] = { + { 0x00000001, "SM_TO_SM_FAULT" }, + { 0x00000002, "L1_ERROR" }, + { 0x00000004, "MULTIPLE_WARP_ERRORS" }, + { 0x00000008, "PHYSICAL_STACK_OVERFLOW" }, + { 0x00000010, "BPT_INT" }, + { 0x00000020, "BPT_PAUSE" }, + { 0x00000040, "SINGLE_STEP_COMPLETE" }, + { 0x20000000, "ECC_SEC_ERROR" }, + { 0x40000000, "ECC_DED_ERROR" }, + { 0x80000000, "TIMEOUT" }, + {} +}; + +void +gf100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x648)); + u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x650)); + const struct nvkm_enum *warp; + char glob[128]; + + nvkm_snprintbf(glob, sizeof(glob), gf100_mp_global_error, gerr); + warp = nvkm_enum_find(gf100_mp_warp_error, werr & 0xffff); + + nvkm_error(subdev, "GPC%i/TPC%i/MP trap: " + "global %08x [%s] warp %04x [%s]\n", + gpc, tpc, gerr, glob, werr, warp ? warp->name : ""); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x648), 0x00000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x650), gerr); +} + +static void +gf100_gr_trap_tpc(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0508)); + + if (stat & 0x00000001) { + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0224)); + nvkm_error(subdev, "GPC%d/TPC%d/TEX: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000); + stat &= ~0x00000001; + } + + if (stat & 0x00000002) { + gr->func->trap_mp(gr, gpc, tpc); + stat &= ~0x00000002; + } + + if (stat & 0x00000004) { + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0084)); + nvkm_error(subdev, "GPC%d/TPC%d/POLY: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000); + stat &= ~0x00000004; + } + + if (stat & 0x00000008) { + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x048c)); + nvkm_error(subdev, "GPC%d/TPC%d/L1C: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000); + stat &= ~0x00000008; + } + + if (stat & 0x00000010) { + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0430)); + nvkm_error(subdev, "GPC%d/TPC%d/MPC: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0430), 0xc0000000); + stat &= ~0x00000010; + } + + if (stat) { + nvkm_error(subdev, "GPC%d/TPC%d/%08x: unknown\n", gpc, tpc, stat); + } +} + +static void +gf100_gr_trap_gpc(struct gf100_gr *gr, int gpc) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, GPC_UNIT(gpc, 0x2c90)); + int tpc; + + if (stat & 0x00000001) { + gf100_gr_trap_gpc_rop(gr, gpc); + stat &= ~0x00000001; + } + + if (stat & 0x00000002) { + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0900)); + nvkm_error(subdev, "GPC%d/ZCULL: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + stat &= ~0x00000002; + } + + if (stat & 0x00000004) { + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x1028)); + nvkm_error(subdev, "GPC%d/CCACHE: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + stat &= ~0x00000004; + } + + if (stat & 0x00000008) { + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0824)); + nvkm_error(subdev, "GPC%d/ESETUP: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + stat &= ~0x00000009; + } + + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + u32 mask = 0x00010000 << tpc; + if (stat & mask) { + gf100_gr_trap_tpc(gr, gpc, tpc); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), mask); + stat &= ~mask; + } + } + + if (stat) { + nvkm_error(subdev, "GPC%d/%08x: unknown\n", gpc, stat); + } +} + +static void +gf100_gr_trap_intr(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + char error[128]; + u32 trap = nvkm_rd32(device, 0x400108); + int rop, gpc; + + if (trap & 0x00000001) { + u32 stat = nvkm_rd32(device, 0x404000); + + nvkm_snprintbf(error, sizeof(error), gf100_dispatch_error, + stat & 0x3fffffff); + nvkm_error(subdev, "DISPATCH %08x [%s]\n", stat, error); + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000001); + trap &= ~0x00000001; + } + + if (trap & 0x00000002) { + u32 stat = nvkm_rd32(device, 0x404600); + + nvkm_snprintbf(error, sizeof(error), gf100_m2mf_error, + stat & 0x3fffffff); + nvkm_error(subdev, "M2MF %08x [%s]\n", stat, error); + + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000002); + trap &= ~0x00000002; + } + + if (trap & 0x00000008) { + u32 stat = nvkm_rd32(device, 0x408030); + + nvkm_snprintbf(error, sizeof(error), gf100_ccache_error, + stat & 0x3fffffff); + nvkm_error(subdev, "CCACHE %08x [%s]\n", stat, error); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000008); + trap &= ~0x00000008; + } + + if (trap & 0x00000010) { + u32 stat = nvkm_rd32(device, 0x405840); + nvkm_error(subdev, "SHADER %08x, sph: 0x%06x, stage: 0x%02x\n", + stat, stat & 0xffffff, (stat >> 24) & 0x3f); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000010); + trap &= ~0x00000010; + } + + if (trap & 0x00000040) { + u32 stat = nvkm_rd32(device, 0x40601c); + + nvkm_snprintbf(error, sizeof(error), gf100_unk6_error, + stat & 0x3fffffff); + nvkm_error(subdev, "UNK6 %08x [%s]\n", stat, error); + + nvkm_wr32(device, 0x40601c, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000040); + trap &= ~0x00000040; + } + + if (trap & 0x00000080) { + u32 stat = nvkm_rd32(device, 0x404490); + u32 pc = nvkm_rd32(device, 0x404494); + u32 op = nvkm_rd32(device, 0x40449c); + + nvkm_snprintbf(error, sizeof(error), gf100_macro_error, + stat & 0x1fffffff); + nvkm_error(subdev, "MACRO %08x [%s], pc: 0x%03x%s, op: 0x%08x\n", + stat, error, pc & 0x7ff, + (pc & 0x10000000) ? "" : " (invalid)", + op); + + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000080); + trap &= ~0x00000080; + } + + if (trap & 0x00000100) { + u32 stat = nvkm_rd32(device, 0x407020) & 0x3fffffff; + + nvkm_snprintbf(error, sizeof(error), gk104_sked_error, stat); + nvkm_error(subdev, "SKED: %08x [%s]\n", stat, error); + + if (stat) + nvkm_wr32(device, 0x407020, 0x40000000); + nvkm_wr32(device, 0x400108, 0x00000100); + trap &= ~0x00000100; + } + + if (trap & 0x01000000) { + u32 stat = nvkm_rd32(device, 0x400118); + for (gpc = 0; stat && gpc < gr->gpc_nr; gpc++) { + u32 mask = 0x00000001 << gpc; + if (stat & mask) { + gf100_gr_trap_gpc(gr, gpc); + nvkm_wr32(device, 0x400118, mask); + stat &= ~mask; + } + } + nvkm_wr32(device, 0x400108, 0x01000000); + trap &= ~0x01000000; + } + + if (trap & 0x02000000) { + for (rop = 0; rop < gr->rop_nr; rop++) { + u32 statz = nvkm_rd32(device, ROP_UNIT(rop, 0x070)); + u32 statc = nvkm_rd32(device, ROP_UNIT(rop, 0x144)); + nvkm_error(subdev, "ROP%d %08x %08x\n", + rop, statz, statc); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); + } + nvkm_wr32(device, 0x400108, 0x02000000); + trap &= ~0x02000000; + } + + if (trap) { + nvkm_error(subdev, "TRAP UNHANDLED %08x\n", trap); + nvkm_wr32(device, 0x400108, trap); + } +} + +static void +gf100_gr_ctxctl_debug_unit(struct gf100_gr *gr, u32 base) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + nvkm_error(subdev, "%06x - done %08x\n", base, + nvkm_rd32(device, base + 0x400)); + nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, + nvkm_rd32(device, base + 0x800), + nvkm_rd32(device, base + 0x804), + nvkm_rd32(device, base + 0x808), + nvkm_rd32(device, base + 0x80c)); + nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, + nvkm_rd32(device, base + 0x810), + nvkm_rd32(device, base + 0x814), + nvkm_rd32(device, base + 0x818), + nvkm_rd32(device, base + 0x81c)); +} + +void +gf100_gr_ctxctl_debug(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 gpcnr = nvkm_rd32(device, 0x409604) & 0xffff; + u32 gpc; + + gf100_gr_ctxctl_debug_unit(gr, 0x409000); + for (gpc = 0; gpc < gpcnr; gpc++) + gf100_gr_ctxctl_debug_unit(gr, 0x502000 + (gpc * 0x8000)); +} + +static void +gf100_gr_ctxctl_isr(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x409c18); + + if (!gr->firmware && (stat & 0x00000001)) { + u32 code = nvkm_rd32(device, 0x409814); + if (code == E_BAD_FWMTHD) { + u32 class = nvkm_rd32(device, 0x409808); + u32 addr = nvkm_rd32(device, 0x40980c); + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00003ffc); + u32 data = nvkm_rd32(device, 0x409810); + + nvkm_error(subdev, "FECS MTHD subc %d class %04x " + "mthd %04x data %08x\n", + subc, class, mthd, data); + } else { + nvkm_error(subdev, "FECS ucode error %d\n", code); + } + nvkm_wr32(device, 0x409c20, 0x00000001); + stat &= ~0x00000001; + } + + if (!gr->firmware && (stat & 0x00080000)) { + nvkm_error(subdev, "FECS watchdog timeout\n"); + gf100_gr_ctxctl_debug(gr); + nvkm_wr32(device, 0x409c20, 0x00080000); + stat &= ~0x00080000; + } + + if (stat) { + nvkm_error(subdev, "FECS %08x\n", stat); + gf100_gr_ctxctl_debug(gr); + nvkm_wr32(device, 0x409c20, stat); + } +} + +static irqreturn_t +gf100_gr_intr(struct nvkm_inth *inth) +{ + struct gf100_gr *gr = container_of(inth, typeof(*gr), base.engine.subdev.inth); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_chan *chan; + unsigned long flags; + u64 inst = nvkm_rd32(device, 0x409b00) & 0x0fffffff; + u32 stat = nvkm_rd32(device, 0x400100); + u32 addr = nvkm_rd32(device, 0x400704); + u32 mthd = (addr & 0x00003ffc); + u32 subc = (addr & 0x00070000) >> 16; + u32 data = nvkm_rd32(device, 0x400708); + u32 code = nvkm_rd32(device, 0x400110); + u32 class; + const char *name = "unknown"; + int chid = -1; + + chan = nvkm_chan_get_inst(&gr->base.engine, (u64)inst << 12, &flags); + if (chan) { + name = chan->name; + chid = chan->id; + } + + if (device->card_type < NV_E0 || subc < 4) + class = nvkm_rd32(device, 0x404200 + (subc * 4)); + else + class = 0x0000; + + if (stat & 0x00000001) { + /* + * notifier interrupt, only needed for cyclestats + * can be safely ignored + */ + nvkm_wr32(device, 0x400100, 0x00000001); + stat &= ~0x00000001; + } + + if (stat & 0x00000010) { + if (!gf100_gr_mthd_sw(device, class, mthd, data)) { + nvkm_error(subdev, "ILLEGAL_MTHD ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + chid, inst << 12, name, subc, + class, mthd, data); + } + nvkm_wr32(device, 0x400100, 0x00000010); + stat &= ~0x00000010; + } + + if (stat & 0x00000020) { + nvkm_error(subdev, "ILLEGAL_CLASS ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + chid, inst << 12, name, subc, class, mthd, data); + nvkm_wr32(device, 0x400100, 0x00000020); + stat &= ~0x00000020; + } + + if (stat & 0x00100000) { + const struct nvkm_enum *en = + nvkm_enum_find(nv50_data_error_names, code); + nvkm_error(subdev, "DATA_ERROR %08x [%s] ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + code, en ? en->name : "", chid, inst << 12, + name, subc, class, mthd, data); + nvkm_wr32(device, 0x400100, 0x00100000); + stat &= ~0x00100000; + } + + if (stat & 0x00200000) { + nvkm_error(subdev, "TRAP ch %d [%010llx %s]\n", + chid, inst << 12, name); + gf100_gr_trap_intr(gr); + nvkm_wr32(device, 0x400100, 0x00200000); + stat &= ~0x00200000; + } + + if (stat & 0x00080000) { + gf100_gr_ctxctl_isr(gr); + nvkm_wr32(device, 0x400100, 0x00080000); + stat &= ~0x00080000; + } + + if (stat) { + nvkm_error(subdev, "intr %08x\n", stat); + nvkm_wr32(device, 0x400100, stat); + } + + nvkm_wr32(device, 0x400500, 0x00010001); + nvkm_chan_put(&chan, flags); + return IRQ_HANDLED; +} + +static void +gf100_gr_init_fw(struct nvkm_falcon *falcon, + struct nvkm_blob *code, struct nvkm_blob *data) +{ + nvkm_falcon_load_dmem(falcon, data->data, 0x0, data->size, 0); + nvkm_falcon_load_imem(falcon, code->data, 0x0, code->size, 0, 0, false); +} + +static void +gf100_gr_init_csdata(struct gf100_gr *gr, + const struct gf100_gr_pack *pack, + u32 falcon, u32 starstar, u32 base) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_gr_pack *iter; + const struct gf100_gr_init *init; + u32 addr = ~0, prev = ~0, xfer = 0; + u32 star, temp; + + nvkm_wr32(device, falcon + 0x01c0, 0x02000000 + starstar); + star = nvkm_rd32(device, falcon + 0x01c4); + temp = nvkm_rd32(device, falcon + 0x01c4); + if (temp > star) + star = temp; + nvkm_wr32(device, falcon + 0x01c0, 0x01000000 + star); + + pack_for_each_init(init, iter, pack) { + u32 head = init->addr - base; + u32 tail = head + init->count * init->pitch; + while (head < tail) { + if (head != prev + 4 || xfer >= 32) { + if (xfer) { + u32 data = ((--xfer << 26) | addr); + nvkm_wr32(device, falcon + 0x01c4, data); + star += 4; + } + addr = head; + xfer = 0; + } + prev = head; + xfer = xfer + 1; + head = head + init->pitch; + } + } + + nvkm_wr32(device, falcon + 0x01c4, (--xfer << 26) | addr); + nvkm_wr32(device, falcon + 0x01c0, 0x01000004 + starstar); + nvkm_wr32(device, falcon + 0x01c4, star + 4); +} + +/* Initialize context from an external (secure or not) firmware */ +static int +gf100_gr_init_ctxctl_ext(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 lsf_mask = 0; + int ret; + + /* load fuc microcode */ + nvkm_mc_unk260(device, 0); + + /* securely-managed falcons must be reset using secure boot */ + + if (!nvkm_acr_managed_falcon(device, NVKM_ACR_LSF_FECS)) { + gf100_gr_init_fw(&gr->fecs.falcon, &gr->fecs.inst, + &gr->fecs.data); + } else { + lsf_mask |= BIT(NVKM_ACR_LSF_FECS); + } + + if (!nvkm_acr_managed_falcon(device, NVKM_ACR_LSF_GPCCS)) { + gf100_gr_init_fw(&gr->gpccs.falcon, &gr->gpccs.inst, + &gr->gpccs.data); + } else { + lsf_mask |= BIT(NVKM_ACR_LSF_GPCCS); + } + + if (lsf_mask) { + ret = nvkm_acr_bootstrap_falcons(device, lsf_mask); + if (ret) + return ret; + } + + nvkm_mc_unk260(device, 1); + + /* start both of them running */ + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x41a10c, 0x00000000); + nvkm_wr32(device, 0x40910c, 0x00000000); + + nvkm_falcon_start(&gr->gpccs.falcon); + nvkm_falcon_start(&gr->fecs.falcon); + + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x00000001) + break; + ) < 0) + return -EBUSY; + + gf100_gr_fecs_set_watchdog_timeout(gr, 0x7fffffff); + + /* Determine how much memory is required to store main context image. */ + ret = gf100_gr_fecs_discover_image_size(gr, &gr->size); + if (ret) + return ret; + + /* Determine how much memory is required to store ZCULL image. */ + ret = gf100_gr_fecs_discover_zcull_image_size(gr, &gr->size_zcull); + if (ret) + return ret; + + /* Determine how much memory is required to store PerfMon image. */ + ret = gf100_gr_fecs_discover_pm_image_size(gr, &gr->size_pm); + if (ret) + return ret; + + /*XXX: We (likely) require PMU support to even bother with this. + * + * Also, it seems like not all GPUs support ELPG. Traces I + * have here show RM enabling it on Kepler/Turing, but none + * of the GPUs between those. NVGPU decides this by PCIID. + */ + if (0) { + ret = gf100_gr_fecs_elpg_bind(gr); + if (ret) + return ret; + } + + return 0; +} + +static int +gf100_gr_init_ctxctl_int(struct gf100_gr *gr) +{ + const struct gf100_grctx_func *grctx = gr->func->grctx; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + + if (!gr->func->fecs.ucode) { + return -ENOSYS; + } + + /* load HUB microcode */ + nvkm_mc_unk260(device, 0); + nvkm_falcon_load_dmem(&gr->fecs.falcon, + gr->func->fecs.ucode->data.data, 0x0, + gr->func->fecs.ucode->data.size, 0); + nvkm_falcon_load_imem(&gr->fecs.falcon, + gr->func->fecs.ucode->code.data, 0x0, + gr->func->fecs.ucode->code.size, 0, 0, false); + + /* load GPC microcode */ + nvkm_falcon_load_dmem(&gr->gpccs.falcon, + gr->func->gpccs.ucode->data.data, 0x0, + gr->func->gpccs.ucode->data.size, 0); + nvkm_falcon_load_imem(&gr->gpccs.falcon, + gr->func->gpccs.ucode->code.data, 0x0, + gr->func->gpccs.ucode->code.size, 0, 0, false); + nvkm_mc_unk260(device, 1); + + /* load register lists */ + gf100_gr_init_csdata(gr, grctx->hub, 0x409000, 0x000, 0x000000); + gf100_gr_init_csdata(gr, grctx->gpc_0, 0x41a000, 0x000, 0x418000); + gf100_gr_init_csdata(gr, grctx->gpc_1, 0x41a000, 0x000, 0x418000); + gf100_gr_init_csdata(gr, grctx->tpc, 0x41a000, 0x004, 0x419800); + gf100_gr_init_csdata(gr, grctx->ppc, 0x41a000, 0x008, 0x41be00); + + /* start HUB ucode running, it'll init the GPCs */ + nvkm_wr32(device, 0x40910c, 0x00000000); + nvkm_wr32(device, 0x409100, 0x00000002); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x80000000) + break; + ) < 0) { + gf100_gr_ctxctl_debug(gr); + return -EBUSY; + } + + gr->size = nvkm_rd32(device, 0x409804); + return 0; +} + +int +gf100_gr_init_ctxctl(struct gf100_gr *gr) +{ + int ret; + + if (gr->firmware) + ret = gf100_gr_init_ctxctl_ext(gr); + else + ret = gf100_gr_init_ctxctl_int(gr); + + return ret; +} + +int +gf100_gr_oneinit_sm_id(struct gf100_gr *gr) +{ + int tpc, gpc; + + for (tpc = 0; tpc < gr->tpc_max; tpc++) { + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + if (tpc < gr->tpc_nr[gpc]) { + gr->sm[gr->sm_nr].gpc = gpc; + gr->sm[gr->sm_nr].tpc = tpc; + gr->sm_nr++; + } + } + } + + return 0; +} + +void +gf100_gr_oneinit_tiles(struct gf100_gr *gr) +{ + static const u8 primes[] = { + 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61 + }; + int init_frac[GPC_MAX], init_err[GPC_MAX], run_err[GPC_MAX], i, j; + u32 mul_factor, comm_denom; + u8 gpc_map[GPC_MAX]; + bool sorted; + + switch (gr->tpc_total) { + case 15: gr->screen_tile_row_offset = 0x06; break; + case 14: gr->screen_tile_row_offset = 0x05; break; + case 13: gr->screen_tile_row_offset = 0x02; break; + case 11: gr->screen_tile_row_offset = 0x07; break; + case 10: gr->screen_tile_row_offset = 0x06; break; + case 7: + case 5: gr->screen_tile_row_offset = 0x01; break; + case 3: gr->screen_tile_row_offset = 0x02; break; + case 2: + case 1: gr->screen_tile_row_offset = 0x01; break; + default: gr->screen_tile_row_offset = 0x03; + for (i = 0; i < ARRAY_SIZE(primes); i++) { + if (gr->tpc_total % primes[i]) { + gr->screen_tile_row_offset = primes[i]; + break; + } + } + break; + } + + /* Sort GPCs by TPC count, highest-to-lowest. */ + for (i = 0; i < gr->gpc_nr; i++) + gpc_map[i] = i; + sorted = false; + + while (!sorted) { + for (sorted = true, i = 0; i < gr->gpc_nr - 1; i++) { + if (gr->tpc_nr[gpc_map[i + 1]] > + gr->tpc_nr[gpc_map[i + 0]]) { + u8 swap = gpc_map[i]; + gpc_map[i + 0] = gpc_map[i + 1]; + gpc_map[i + 1] = swap; + sorted = false; + } + } + } + + /* Determine tile->GPC mapping */ + mul_factor = gr->gpc_nr * gr->tpc_max; + if (mul_factor & 1) + mul_factor = 2; + else + mul_factor = 1; + + comm_denom = gr->gpc_nr * gr->tpc_max * mul_factor; + + for (i = 0; i < gr->gpc_nr; i++) { + init_frac[i] = gr->tpc_nr[gpc_map[i]] * gr->gpc_nr * mul_factor; + init_err[i] = i * gr->tpc_max * mul_factor - comm_denom/2; + run_err[i] = init_frac[i] + init_err[i]; + } + + for (i = 0; i < gr->tpc_total;) { + for (j = 0; j < gr->gpc_nr; j++) { + if ((run_err[j] * 2) >= comm_denom) { + gr->tile[i++] = gpc_map[j]; + run_err[j] += init_frac[j] - comm_denom; + } else { + run_err[j] += init_frac[j]; + } + } + } +} + +static int +gf100_gr_oneinit(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_intr *intr = &device->mc->intr; + enum nvkm_intr_type intr_type = NVKM_INTR_SUBDEV; + int ret, i, j; + + if (gr->func->oneinit_intr) + intr = gr->func->oneinit_intr(gr, &intr_type); + + ret = nvkm_inth_add(intr, intr_type, NVKM_INTR_PRIO_NORMAL, &gr->base.engine.subdev, + gf100_gr_intr, &gr->base.engine.subdev.inth); + if (ret) + return ret; + + nvkm_pmu_pgob(device->pmu, false); + + gr->rop_nr = gr->func->rops(gr); + gr->gpc_nr = nvkm_rd32(device, 0x409604) & 0x0000001f; + for (i = 0; i < gr->gpc_nr; i++) { + gr->tpc_nr[i] = nvkm_rd32(device, GPC_UNIT(i, 0x2608)); + gr->tpc_max = max(gr->tpc_max, gr->tpc_nr[i]); + gr->tpc_total += gr->tpc_nr[i]; + for (j = 0; j < gr->func->ppc_nr; j++) { + gr->ppc_tpc_mask[i][j] = + nvkm_rd32(device, GPC_UNIT(i, 0x0c30 + (j * 4))); + if (gr->ppc_tpc_mask[i][j] == 0) + continue; + + gr->ppc_nr[i]++; + + gr->ppc_mask[i] |= (1 << j); + gr->ppc_tpc_nr[i][j] = hweight8(gr->ppc_tpc_mask[i][j]); + if (gr->ppc_tpc_min == 0 || + gr->ppc_tpc_min > gr->ppc_tpc_nr[i][j]) + gr->ppc_tpc_min = gr->ppc_tpc_nr[i][j]; + if (gr->ppc_tpc_max < gr->ppc_tpc_nr[i][j]) + gr->ppc_tpc_max = gr->ppc_tpc_nr[i][j]; + } + + gr->ppc_total += gr->ppc_nr[i]; + } + + /* Allocate global context buffers. */ + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, gr->func->grctx->pagepool_size, + 0x100, false, &gr->pagepool); + if (ret) + return ret; + + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, gr->func->grctx->bundle_size, + 0x100, false, &gr->bundle_cb); + if (ret) + return ret; + + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, gr->func->grctx->attrib_cb_size(gr), + 0x1000, false, &gr->attrib_cb); + if (ret) + return ret; + + if (gr->func->grctx->unknown_size) { + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, gr->func->grctx->unknown_size, + 0x100, false, &gr->unknown); + if (ret) + return ret; + } + + memset(gr->tile, 0xff, sizeof(gr->tile)); + gr->func->oneinit_tiles(gr); + + return gr->func->oneinit_sm_id(gr); +} + +static int +gf100_gr_init_(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &base->engine.subdev; + struct nvkm_device *device = subdev->device; + bool reset = device->chipset == 0x137 || device->chipset == 0x138; + int ret; + + /* On certain GP107/GP108 boards, we trigger a weird issue where + * GR will stop responding to PRI accesses after we've asked the + * SEC2 RTOS to boot the GR falcons. This happens with far more + * frequency when cold-booting a board (ie. returning from D3). + * + * The root cause for this is not known and has proven difficult + * to isolate, with many avenues being dead-ends. + * + * A workaround was discovered by Karol, whereby putting GR into + * reset for an extended period right before initialisation + * prevents the problem from occuring. + * + * XXX: As RM does not require any such workaround, this is more + * of a hack than a true fix. + */ + reset = nvkm_boolopt(device->cfgopt, "NvGrResetWar", reset); + if (reset) { + nvkm_mask(device, 0x000200, 0x00001000, 0x00000000); + nvkm_rd32(device, 0x000200); + msleep(50); + nvkm_mask(device, 0x000200, 0x00001000, 0x00001000); + nvkm_rd32(device, 0x000200); + } + + nvkm_pmu_pgob(gr->base.engine.subdev.device->pmu, false); + + ret = nvkm_falcon_get(&gr->fecs.falcon, subdev); + if (ret) + return ret; + + ret = nvkm_falcon_get(&gr->gpccs.falcon, subdev); + if (ret) + return ret; + + ret = gr->func->init(gr); + if (ret) + return ret; + + nvkm_inth_allow(&subdev->inth); + return 0; +} + +static int +gf100_gr_fini(struct nvkm_gr *base, bool suspend) +{ + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + + nvkm_inth_block(&subdev->inth); + + nvkm_falcon_put(&gr->gpccs.falcon, subdev); + nvkm_falcon_put(&gr->fecs.falcon, subdev); + return 0; +} + +static void * +gf100_gr_dtor(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + + kfree(gr->data); + + nvkm_memory_unref(&gr->unknown); + nvkm_memory_unref(&gr->attrib_cb); + nvkm_memory_unref(&gr->bundle_cb); + nvkm_memory_unref(&gr->pagepool); + + nvkm_falcon_dtor(&gr->gpccs.falcon); + nvkm_falcon_dtor(&gr->fecs.falcon); + + nvkm_blob_dtor(&gr->fecs.inst); + nvkm_blob_dtor(&gr->fecs.data); + nvkm_blob_dtor(&gr->gpccs.inst); + nvkm_blob_dtor(&gr->gpccs.data); + + vfree(gr->bundle64); + vfree(gr->bundle_veid); + vfree(gr->bundle); + vfree(gr->method); + vfree(gr->sw_ctx); + vfree(gr->sw_nonctx); + vfree(gr->sw_nonctx1); + vfree(gr->sw_nonctx2); + vfree(gr->sw_nonctx3); + vfree(gr->sw_nonctx4); + + return gr; +} + +static const struct nvkm_falcon_func +gf100_gr_flcn = { + .load_imem = nvkm_falcon_v1_load_imem, + .load_dmem = nvkm_falcon_v1_load_dmem, + .start = nvkm_falcon_v1_start, +}; + +void +gf100_gr_init_num_tpc_per_gpc(struct gf100_gr *gr, bool pd, bool ds) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int gpc, i, j; + u32 data; + + for (gpc = 0, i = 0; i < 4; i++) { + for (data = 0, j = 0; j < 8 && gpc < gr->gpc_nr; j++, gpc++) + data |= gr->tpc_nr[gpc] << (j * 4); + if (pd) + nvkm_wr32(device, 0x406028 + (i * 4), data); + if (ds) + nvkm_wr32(device, 0x405870 + (i * 4), data); + } +} + +void +gf100_gr_init_400054(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x400054, 0x34ce3464); +} + +void +gf100_gr_init_exception2(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); +} + +void +gf100_gr_init_rop_exceptions(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int rop; + + for (rop = 0; rop < gr->rop_nr; rop++) { + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); + nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); + } +} + +void +gf100_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); +} + +void +gf100_gr_init_tex_hww_esr(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); +} + +void +gf100_gr_init_419eb4(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); +} + +void +gf100_gr_init_419cc0(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int gpc, tpc; + + nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); + } +} + +void +gf100_gr_init_40601c(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x40601c, 0xc0000000); +} + +void +gf100_gr_init_fecs_exceptions(struct gf100_gr *gr) +{ + const u32 data = gr->firmware ? 0x000e0000 : 0x000e0001; + nvkm_wr32(gr->base.engine.subdev.device, 0x409c24, data); +} + +void +gf100_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fb *fb = device->fb; + + nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0x00000001); + nvkm_wr32(device, 0x4188a4, 0x03000000); + nvkm_wr32(device, 0x418888, 0x00000000); + nvkm_wr32(device, 0x41888c, 0x00000000); + nvkm_wr32(device, 0x418890, 0x00000000); + nvkm_wr32(device, 0x418894, 0x00000000); + nvkm_wr32(device, 0x4188b4, nvkm_memory_addr(fb->mmu_wr) >> 8); + nvkm_wr32(device, 0x4188b8, nvkm_memory_addr(fb->mmu_rd) >> 8); +} + +void +gf100_gr_init_num_active_ltcs(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); +} + +void +gf100_gr_init_zcull(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + const u8 tile_nr = ALIGN(gr->tpc_total, 32); + u8 bank[GPC_MAX] = {}, gpc, i, j; + u32 data; + + for (i = 0; i < tile_nr; i += 8) { + for (data = 0, j = 0; j < 8 && i + j < gr->tpc_total; j++) { + data |= bank[gr->tile[i + j]] << (j * 4); + bank[gr->tile[i + j]]++; + } + nvkm_wr32(device, GPC_BCAST(0x0980 + ((i / 8) * 4)), data); + } + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->screen_tile_row_offset << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); + } + + nvkm_wr32(device, GPC_BCAST(0x1bd4), magicgpc918); +} + +void +gf100_gr_init_vsc_stream_master(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, TPC_UNIT(0, 0, 0x05c), 0x00000001, 0x00000001); +} + +static int +gf100_gr_reset(struct nvkm_gr *base) +{ + struct nvkm_subdev *subdev = &base->engine.subdev; + struct nvkm_device *device = subdev->device; + struct gf100_gr *gr = gf100_gr(base); + + nvkm_mask(device, 0x400500, 0x00000001, 0x00000000); + + WARN_ON(gf100_gr_fecs_halt_pipeline(gr)); + + subdev->func->fini(subdev, false); + nvkm_mc_disable(device, subdev->type, subdev->inst); + if (gr->func->gpccs.reset) + gr->func->gpccs.reset(gr); + + nvkm_mc_enable(device, subdev->type, subdev->inst); + return subdev->func->init(subdev); +} + +int +gf100_gr_init(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int gpc, tpc; + + nvkm_mask(device, 0x400500, 0x00010001, 0x00000000); + + gr->func->init_gpc_mmu(gr); + + if (gr->sw_nonctx1) { + gf100_gr_mmio(gr, gr->sw_nonctx1); + gf100_gr_mmio(gr, gr->sw_nonctx2); + gf100_gr_mmio(gr, gr->sw_nonctx3); + gf100_gr_mmio(gr, gr->sw_nonctx4); + } else + if (gr->sw_nonctx) { + gf100_gr_mmio(gr, gr->sw_nonctx); + } else { + gf100_gr_mmio(gr, gr->func->mmio); + } + + gf100_gr_wait_idle(gr); + + if (gr->func->init_r405a14) + gr->func->init_r405a14(gr); + + if (gr->func->clkgate_pack) + nvkm_therm_clkgate_init(device->therm, gr->func->clkgate_pack); + + if (gr->func->init_bios) + gr->func->init_bios(gr); + + gr->func->init_vsc_stream_master(gr); + gr->func->init_zcull(gr); + gr->func->init_num_active_ltcs(gr); + if (gr->func->init_rop_active_fbps) + gr->func->init_rop_active_fbps(gr); + if (gr->func->init_bios_2) + gr->func->init_bios_2(gr); + if (gr->func->init_swdx_pes_mask) + gr->func->init_swdx_pes_mask(gr); + if (gr->func->init_fs) + gr->func->init_fs(gr); + + nvkm_wr32(device, 0x400500, 0x00010001); + + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + nvkm_wr32(device, 0x400124, 0x00000002); + + gr->func->init_fecs_exceptions(gr); + + if (gr->func->init_40a790) + gr->func->init_40a790(gr); + + if (gr->func->init_ds_hww_esr_2) + gr->func->init_ds_hww_esr_2(gr); + + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x408030, 0xc0000000); + + if (gr->func->init_40601c) + gr->func->init_40601c(gr); + + nvkm_wr32(device, 0x406018, 0xc0000000); + nvkm_wr32(device, 0x404490, 0xc0000000); + + if (gr->func->init_sked_hww_esr) + gr->func->init_sked_hww_esr(gr); + + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x405844, 0x00ffffff); + + if (gr->func->init_419cc0) + gr->func->init_419cc0(gr); + if (gr->func->init_419eb4) + gr->func->init_419eb4(gr); + if (gr->func->init_419c9c) + gr->func->init_419c9c(gr); + + if (gr->func->init_ppc_exceptions) + gr->func->init_ppc_exceptions(gr); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); + if (gr->func->init_tex_hww_esr) + gr->func->init_tex_hww_esr(gr, gpc, tpc); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); + if (gr->func->init_504430) + gr->func->init_504430(gr, gpc, tpc); + gr->func->init_shader_exceptions(gr, gpc, tpc); + } + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); + } + + gr->func->init_rop_exceptions(gr); + + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + if (gr->func->init_exception2) + gr->func->init_exception2(gr); + + if (gr->func->init_400054) + gr->func->init_400054(gr); + + gf100_gr_zbc_init(gr); + + if (gr->func->init_4188a4) + gr->func->init_4188a4(gr); + + return gf100_gr_init_ctxctl(gr); +} + +void +gf100_gr_fecs_reset(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x409614, 0x00000070); + nvkm_usec(device, 10, NVKM_DELAY); + nvkm_mask(device, 0x409614, 0x00000700, 0x00000700); + nvkm_usec(device, 10, NVKM_DELAY); + nvkm_rd32(device, 0x409614); +} + +#include "fuc/hubgf100.fuc3.h" + +struct gf100_gr_ucode +gf100_gr_fecs_ucode = { + .code.data = gf100_grhub_code, + .code.size = sizeof(gf100_grhub_code), + .data.data = gf100_grhub_data, + .data.size = sizeof(gf100_grhub_data), +}; + +#include "fuc/gpcgf100.fuc3.h" + +struct gf100_gr_ucode +gf100_gr_gpccs_ucode = { + .code.data = gf100_grgpc_code, + .code.size = sizeof(gf100_grgpc_code), + .data.data = gf100_grgpc_data, + .data.size = sizeof(gf100_grgpc_data), +}; + +static int +gf100_gr_nonstall(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + + if (gr->func->nonstall) + return gr->func->nonstall(gr); + + return -EINVAL; +} + +static const struct nvkm_gr_func +gf100_gr_ = { + .dtor = gf100_gr_dtor, + .oneinit = gf100_gr_oneinit, + .init = gf100_gr_init_, + .fini = gf100_gr_fini, + .nonstall = gf100_gr_nonstall, + .reset = gf100_gr_reset, + .units = gf100_gr_units, + .chan_new = gf100_gr_chan_new, + .object_get = gf100_gr_object_get, + .chsw_load = gf100_gr_chsw_load, + .ctxsw.pause = gf100_gr_fecs_stop_ctxsw, + .ctxsw.resume = gf100_gr_fecs_start_ctxsw, + .ctxsw.inst = gf100_gr_ctxsw_inst, +}; + +static const struct gf100_gr_func +gf100_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf100_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf100_gr_pack_mmio, + .fecs.ucode = &gf100_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf100_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .grctx = &gf100_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +int +gf100_gr_nofw(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + gr->firmware = false; + return 0; +} + +static int +gf100_gr_load_fw(struct gf100_gr *gr, const char *name, + struct nvkm_blob *blob) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + const struct firmware *fw; + char f[32]; + int ret; + + snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, name); + ret = request_firmware(&fw, f, device->dev); + if (ret) { + snprintf(f, sizeof(f), "nouveau/%s", name); + ret = request_firmware(&fw, f, device->dev); + if (ret) { + nvkm_error(subdev, "failed to load %s\n", name); + return ret; + } + } + + blob->size = fw->size; + blob->data = kmemdup(fw->data, blob->size, GFP_KERNEL); + release_firmware(fw); + return (blob->data != NULL) ? 0 : -ENOMEM; +} + +int +gf100_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + if (!nvkm_boolopt(device->cfgopt, "NvGrUseFW", false)) + return -EINVAL; + + if (gf100_gr_load_fw(gr, "fuc409c", &gr->fecs.inst) || + gf100_gr_load_fw(gr, "fuc409d", &gr->fecs.data) || + gf100_gr_load_fw(gr, "fuc41ac", &gr->gpccs.inst) || + gf100_gr_load_fw(gr, "fuc41ad", &gr->gpccs.data)) + return -ENOENT; + + gr->firmware = true; + return 0; +} + +static const struct gf100_gr_fwif +gf100_gr_fwif[] = { + { -1, gf100_gr_load, &gf100_gr }, + { -1, gf100_gr_nofw, &gf100_gr }, + {} +}; + +int +gf100_gr_new_(const struct gf100_gr_fwif *fwif, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct gf100_gr *gr; + int ret; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + + ret = nvkm_gr_ctor(&gf100_gr_, device, type, inst, true, &gr->base); + if (ret) + return ret; + + fwif = nvkm_firmware_load(&gr->base.engine.subdev, fwif, "Gr", gr); + if (IS_ERR(fwif)) + return PTR_ERR(fwif); + + gr->func = fwif->func; + + ret = nvkm_falcon_ctor(&gf100_gr_flcn, &gr->base.engine.subdev, + "fecs", 0x409000, &gr->fecs.falcon); + if (ret) + return ret; + + mutex_init(&gr->fecs.mutex); + + ret = nvkm_falcon_ctor(&gf100_gr_flcn, &gr->base.engine.subdev, + "gpccs", 0x41a000, &gr->gpccs.falcon); + if (ret) + return ret; + + return 0; +} + +int +gf100_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf100_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h new file mode 100644 index 0000000000..54f686ba39 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h @@ -0,0 +1,448 @@ +/* + * Copyright 2010 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#ifndef __GF100_GR_H__ +#define __GF100_GR_H__ +#define gf100_gr(p) container_of((p), struct gf100_gr, base) +#include "priv.h" + +#include +#include +#include +#include + +struct nvkm_acr_lsfw; + +#define GPC_MAX 32 +#define TPC_MAX_PER_GPC 8 +#define TPC_MAX (GPC_MAX * TPC_MAX_PER_GPC) + +#define ROP_BCAST(r) (0x408800 + (r)) +#define ROP_UNIT(u, r) (0x410000 + (u) * 0x400 + (r)) +#define GPC_BCAST(r) (0x418000 + (r)) +#define GPC_UNIT(t, r) (0x500000 + (t) * 0x8000 + (r)) +#define PPC_UNIT(t, m, r) (0x503000 + (t) * 0x8000 + (m) * 0x200 + (r)) +#define TPC_UNIT(t, m, r) (0x504000 + (t) * 0x8000 + (m) * 0x800 + (r)) + +struct gf100_gr_zbc_color { + u32 format; + u32 ds[4]; + u32 l2[4]; +}; + +struct gf100_gr_zbc_depth { + u32 format; + u32 ds; + u32 l2; +}; + +struct gf100_gr_zbc_stencil { + u32 format; + u32 ds; + u32 l2; +}; + +struct gf100_gr { + const struct gf100_gr_func *func; + struct nvkm_gr base; + + struct { + struct nvkm_falcon falcon; + struct nvkm_blob inst; + struct nvkm_blob data; + + struct mutex mutex; + u32 disable; + } fecs; + + struct { + struct nvkm_falcon falcon; + struct nvkm_blob inst; + struct nvkm_blob data; + } gpccs; + + bool firmware; + + /* + * Used if the register packs are loaded from NVIDIA fw instead of + * using hardcoded arrays. To be allocated with vzalloc(). + */ + struct gf100_gr_pack *sw_nonctx; + struct gf100_gr_pack *sw_nonctx1; + struct gf100_gr_pack *sw_nonctx2; + struct gf100_gr_pack *sw_nonctx3; + struct gf100_gr_pack *sw_nonctx4; + struct gf100_gr_pack *sw_ctx; + struct gf100_gr_pack *bundle; + struct gf100_gr_pack *bundle_veid; + struct gf100_gr_pack *bundle64; + struct gf100_gr_pack *method; + + struct gf100_gr_zbc_color zbc_color[NVKM_LTC_MAX_ZBC_COLOR_CNT]; + struct gf100_gr_zbc_depth zbc_depth[NVKM_LTC_MAX_ZBC_DEPTH_CNT]; + struct gf100_gr_zbc_stencil zbc_stencil[NVKM_LTC_MAX_ZBC_DEPTH_CNT]; + + u8 rop_nr; + u8 gpc_nr; + u8 tpc_nr[GPC_MAX]; + u8 tpc_max; + u8 tpc_total; + u8 ppc_nr[GPC_MAX]; + u8 ppc_mask[GPC_MAX]; + u8 ppc_tpc_mask[GPC_MAX][4]; + u8 ppc_tpc_nr[GPC_MAX][4]; + u8 ppc_tpc_min; + u8 ppc_tpc_max; + u8 ppc_total; + + struct nvkm_memory *pagepool; + struct nvkm_memory *bundle_cb; + struct nvkm_memory *attrib_cb; + struct nvkm_memory *unknown; + + u8 screen_tile_row_offset; + u8 tile[TPC_MAX]; + + struct { + u8 gpc; + u8 tpc; + } sm[TPC_MAX]; + u8 sm_nr; + + u32 size; + u32 *data; + u32 size_zcull; + u32 size_pm; +}; + +int gf100_gr_fecs_bind_pointer(struct gf100_gr *, u32 inst); +int gf100_gr_fecs_wfi_golden_save(struct gf100_gr *, u32 inst); + +struct gf100_gr_func_zbc { + void (*clear_color)(struct gf100_gr *, int zbc); + void (*clear_depth)(struct gf100_gr *, int zbc); + int (*stencil_get)(struct gf100_gr *, int format, + const u32 ds, const u32 l2); + void (*clear_stencil)(struct gf100_gr *, int zbc); +}; + +struct gf100_gr_func { + int (*nonstall)(struct gf100_gr *); + struct nvkm_intr *(*oneinit_intr)(struct gf100_gr *, enum nvkm_intr_type *); + void (*oneinit_tiles)(struct gf100_gr *); + int (*oneinit_sm_id)(struct gf100_gr *); + int (*init)(struct gf100_gr *); + void (*init_419bd8)(struct gf100_gr *); + void (*init_gpc_mmu)(struct gf100_gr *); + void (*init_r405a14)(struct gf100_gr *); + void (*init_bios)(struct gf100_gr *); + void (*init_vsc_stream_master)(struct gf100_gr *); + void (*init_zcull)(struct gf100_gr *); + void (*init_num_active_ltcs)(struct gf100_gr *); + void (*init_rop_active_fbps)(struct gf100_gr *); + void (*init_bios_2)(struct gf100_gr *); + void (*init_swdx_pes_mask)(struct gf100_gr *); + void (*init_fs)(struct gf100_gr *); + void (*init_fecs_exceptions)(struct gf100_gr *); + void (*init_40a790)(struct gf100_gr *); + void (*init_ds_hww_esr_2)(struct gf100_gr *); + void (*init_40601c)(struct gf100_gr *); + void (*init_sked_hww_esr)(struct gf100_gr *); + void (*init_419cc0)(struct gf100_gr *); + void (*init_419eb4)(struct gf100_gr *); + void (*init_419c9c)(struct gf100_gr *); + void (*init_ppc_exceptions)(struct gf100_gr *); + void (*init_tex_hww_esr)(struct gf100_gr *, int gpc, int tpc); + void (*init_504430)(struct gf100_gr *, int gpc, int tpc); + void (*init_shader_exceptions)(struct gf100_gr *, int gpc, int tpc); + void (*init_rop_exceptions)(struct gf100_gr *); + void (*init_exception2)(struct gf100_gr *); + void (*init_400054)(struct gf100_gr *); + void (*init_4188a4)(struct gf100_gr *); + void (*trap_mp)(struct gf100_gr *, int gpc, int tpc); + void (*set_hww_esr_report_mask)(struct gf100_gr *); + const struct gf100_gr_pack *mmio; + struct { + struct gf100_gr_ucode *ucode; + void (*reset)(struct gf100_gr *); + } fecs; + struct { + struct gf100_gr_ucode *ucode; + void (*reset)(struct gf100_gr *); + } gpccs; + int (*rops)(struct gf100_gr *); + int gpc_nr; + int tpc_nr; + int ppc_nr; + const struct gf100_grctx_func *grctx; + const struct nvkm_therm_clkgate_pack *clkgate_pack; + const struct gf100_gr_func_zbc *zbc; + struct nvkm_sclass sclass[]; +}; + +int gf100_gr_rops(struct gf100_gr *); +void gf100_gr_oneinit_tiles(struct gf100_gr *); +int gf100_gr_oneinit_sm_id(struct gf100_gr *); +int gf100_gr_init(struct gf100_gr *); +void gf100_gr_init_vsc_stream_master(struct gf100_gr *); +void gf100_gr_init_zcull(struct gf100_gr *); +void gf100_gr_init_num_active_ltcs(struct gf100_gr *); +void gf100_gr_init_fecs_exceptions(struct gf100_gr *); +void gf100_gr_init_40601c(struct gf100_gr *); +void gf100_gr_init_419cc0(struct gf100_gr *); +void gf100_gr_init_419eb4(struct gf100_gr *); +void gf100_gr_init_tex_hww_esr(struct gf100_gr *, int, int); +void gf100_gr_init_shader_exceptions(struct gf100_gr *, int, int); +void gf100_gr_init_rop_exceptions(struct gf100_gr *); +void gf100_gr_init_exception2(struct gf100_gr *); +void gf100_gr_init_400054(struct gf100_gr *); +void gf100_gr_init_num_tpc_per_gpc(struct gf100_gr *, bool, bool); +extern const struct gf100_gr_func_zbc gf100_gr_zbc; +void gf100_gr_fecs_reset(struct gf100_gr *); + +void gf117_gr_init_zcull(struct gf100_gr *); + +void gk104_gr_init_vsc_stream_master(struct gf100_gr *); +void gk104_gr_init_rop_active_fbps(struct gf100_gr *); +void gk104_gr_init_ppc_exceptions(struct gf100_gr *); +void gk104_gr_init_sked_hww_esr(struct gf100_gr *); + +void gk110_gr_init_419eb4(struct gf100_gr *); + +void gm107_gr_init_504430(struct gf100_gr *, int, int); +void gm107_gr_init_shader_exceptions(struct gf100_gr *, int, int); +void gm107_gr_init_400054(struct gf100_gr *); + +int gk20a_gr_init(struct gf100_gr *); +int gk20a_gr_av_to_init_(struct nvkm_blob *, u8 count, u32 pitch, struct gf100_gr_pack **); +int gk20a_gr_av_to_init(struct nvkm_blob *, struct gf100_gr_pack **); +int gk20a_gr_aiv_to_init(struct nvkm_blob *, struct gf100_gr_pack **); +int gk20a_gr_av_to_method(struct nvkm_blob *, struct gf100_gr_pack **); + +void gm200_gr_oneinit_tiles(struct gf100_gr *); +int gm200_gr_oneinit_sm_id(struct gf100_gr *); +int gm200_gr_rops(struct gf100_gr *); +void gm200_gr_init_num_active_ltcs(struct gf100_gr *); +void gm200_gr_init_ds_hww_esr_2(struct gf100_gr *); + +void gp100_gr_init_rop_active_fbps(struct gf100_gr *); +void gp100_gr_init_fecs_exceptions(struct gf100_gr *); +void gp100_gr_init_shader_exceptions(struct gf100_gr *, int, int); +void gp100_gr_zbc_clear_color(struct gf100_gr *, int); +void gp100_gr_zbc_clear_depth(struct gf100_gr *, int); +extern const struct gf100_gr_func_zbc gp100_gr_zbc; + +void gp102_gr_init_swdx_pes_mask(struct gf100_gr *); +extern const struct gf100_gr_func_zbc gp102_gr_zbc; +int gp102_gr_zbc_stencil_get(struct gf100_gr *, int, const u32, const u32); +void gp102_gr_zbc_clear_stencil(struct gf100_gr *, int); + +extern const struct gf100_gr_func gp107_gr; + +int gv100_gr_oneinit_sm_id(struct gf100_gr *); +u32 gv100_gr_nonpes_aware_tpc(struct gf100_gr *gr, u32 gpc, u32 tpc); +void gv100_gr_init_419bd8(struct gf100_gr *); +void gv100_gr_init_504430(struct gf100_gr *, int, int); +void gv100_gr_init_shader_exceptions(struct gf100_gr *, int, int); +void gv100_gr_init_4188a4(struct gf100_gr *); +void gv100_gr_trap_mp(struct gf100_gr *, int, int); + +int tu102_gr_av_to_init_veid(struct nvkm_blob *, struct gf100_gr_pack **); +void tu102_gr_init_zcull(struct gf100_gr *); +void tu102_gr_init_fs(struct gf100_gr *); +void tu102_gr_init_fecs_exceptions(struct gf100_gr *); + +#define gf100_gr_chan(p) container_of((p), struct gf100_gr_chan, object) +#include + +struct gf100_gr_chan { + struct nvkm_object object; + struct gf100_gr *gr; + struct nvkm_vmm *vmm; + + struct nvkm_vma *pagepool; + struct nvkm_vma *bundle_cb; + struct nvkm_vma *attrib_cb; + struct nvkm_vma *unknown; + + struct nvkm_memory *mmio; + struct nvkm_vma *mmio_vma; + int mmio_nr; +}; + +void gf100_gr_ctxctl_debug(struct gf100_gr *); + +u64 gf100_gr_units(struct nvkm_gr *); +void gf100_gr_zbc_init(struct gf100_gr *); + +extern const struct nvkm_object_func gf100_fermi; + +struct gf100_gr_init { + u32 addr; + u8 count; + u32 pitch; + u64 data; +}; + +struct gf100_gr_pack { + const struct gf100_gr_init *init; + u32 type; +}; + +#define pack_for_each_init(init, pack, head) \ + for (pack = head; pack && pack->init; pack++) \ + for (init = pack->init; init && init->count; init++) + +struct gf100_gr_ucode { + struct nvkm_blob code; + struct nvkm_blob data; +}; + +extern struct gf100_gr_ucode gf100_gr_fecs_ucode; +extern struct gf100_gr_ucode gf100_gr_gpccs_ucode; + +extern struct gf100_gr_ucode gk110_gr_fecs_ucode; +extern struct gf100_gr_ucode gk110_gr_gpccs_ucode; + +int gf100_gr_wait_idle(struct gf100_gr *); +void gf100_gr_mmio(struct gf100_gr *, const struct gf100_gr_pack *); +void gf100_gr_icmd(struct gf100_gr *, const struct gf100_gr_pack *); +void gf100_gr_mthd(struct gf100_gr *, const struct gf100_gr_pack *); +int gf100_gr_init_ctxctl(struct gf100_gr *); + +/* register init value lists */ + +extern const struct gf100_gr_init gf100_gr_init_main_0[]; +extern const struct gf100_gr_init gf100_gr_init_fe_0[]; +extern const struct gf100_gr_init gf100_gr_init_pri_0[]; +extern const struct gf100_gr_init gf100_gr_init_rstr2d_0[]; +extern const struct gf100_gr_init gf100_gr_init_pd_0[]; +extern const struct gf100_gr_init gf100_gr_init_ds_0[]; +extern const struct gf100_gr_init gf100_gr_init_scc_0[]; +extern const struct gf100_gr_init gf100_gr_init_prop_0[]; +extern const struct gf100_gr_init gf100_gr_init_gpc_unk_0[]; +extern const struct gf100_gr_init gf100_gr_init_setup_0[]; +extern const struct gf100_gr_init gf100_gr_init_crstr_0[]; +extern const struct gf100_gr_init gf100_gr_init_setup_1[]; +extern const struct gf100_gr_init gf100_gr_init_zcull_0[]; +extern const struct gf100_gr_init gf100_gr_init_gpm_0[]; +extern const struct gf100_gr_init gf100_gr_init_gpc_unk_1[]; +extern const struct gf100_gr_init gf100_gr_init_gcc_0[]; +extern const struct gf100_gr_init gf100_gr_init_tpccs_0[]; +extern const struct gf100_gr_init gf100_gr_init_tex_0[]; +extern const struct gf100_gr_init gf100_gr_init_pe_0[]; +extern const struct gf100_gr_init gf100_gr_init_l1c_0[]; +extern const struct gf100_gr_init gf100_gr_init_wwdx_0[]; +extern const struct gf100_gr_init gf100_gr_init_tpccs_1[]; +extern const struct gf100_gr_init gf100_gr_init_mpc_0[]; +extern const struct gf100_gr_init gf100_gr_init_be_0[]; +extern const struct gf100_gr_init gf100_gr_init_fe_1[]; +extern const struct gf100_gr_init gf100_gr_init_pe_1[]; +void gf100_gr_init_gpc_mmu(struct gf100_gr *); +void gf100_gr_trap_mp(struct gf100_gr *, int, int); +extern const struct nvkm_bitfield gf100_mp_global_error[]; +extern const struct nvkm_enum gf100_mp_warp_error[]; + +extern const struct gf100_gr_init gf104_gr_init_ds_0[]; +extern const struct gf100_gr_init gf104_gr_init_tex_0[]; +extern const struct gf100_gr_init gf104_gr_init_sm_0[]; + +extern const struct gf100_gr_init gf108_gr_init_gpc_unk_0[]; +extern const struct gf100_gr_init gf108_gr_init_setup_1[]; + +extern const struct gf100_gr_init gf119_gr_init_pd_0[]; +extern const struct gf100_gr_init gf119_gr_init_ds_0[]; +extern const struct gf100_gr_init gf119_gr_init_prop_0[]; +extern const struct gf100_gr_init gf119_gr_init_gpm_0[]; +extern const struct gf100_gr_init gf119_gr_init_gpc_unk_1[]; +extern const struct gf100_gr_init gf119_gr_init_tex_0[]; +extern const struct gf100_gr_init gf119_gr_init_sm_0[]; +extern const struct gf100_gr_init gf119_gr_init_fe_1[]; + +extern const struct gf100_gr_init gf117_gr_init_pes_0[]; +extern const struct gf100_gr_init gf117_gr_init_wwdx_0[]; +extern const struct gf100_gr_init gf117_gr_init_cbm_0[]; + +extern const struct gf100_gr_init gk104_gr_init_main_0[]; +extern const struct gf100_gr_init gk104_gr_init_gpc_unk_2[]; +extern const struct gf100_gr_init gk104_gr_init_tpccs_0[]; +extern const struct gf100_gr_init gk104_gr_init_pe_0[]; +extern const struct gf100_gr_init gk104_gr_init_be_0[]; +extern const struct gf100_gr_pack gk104_gr_pack_mmio[]; + +extern const struct gf100_gr_init gk110_gr_init_fe_0[]; +extern const struct gf100_gr_init gk110_gr_init_ds_0[]; +extern const struct gf100_gr_init gk110_gr_init_sked_0[]; +extern const struct gf100_gr_init gk110_gr_init_cwd_0[]; +extern const struct gf100_gr_init gk110_gr_init_gpc_unk_1[]; +extern const struct gf100_gr_init gk110_gr_init_tex_0[]; +extern const struct gf100_gr_init gk110_gr_init_sm_0[]; + +extern const struct gf100_gr_init gk208_gr_init_gpc_unk_0[]; + +extern const struct gf100_gr_init gm107_gr_init_scc_0[]; +extern const struct gf100_gr_init gm107_gr_init_prop_0[]; +extern const struct gf100_gr_init gm107_gr_init_setup_1[]; +extern const struct gf100_gr_init gm107_gr_init_zcull_0[]; +extern const struct gf100_gr_init gm107_gr_init_gpc_unk_1[]; +extern const struct gf100_gr_init gm107_gr_init_tex_0[]; +extern const struct gf100_gr_init gm107_gr_init_l1c_0[]; +extern const struct gf100_gr_init gm107_gr_init_wwdx_0[]; +extern const struct gf100_gr_init gm107_gr_init_cbm_0[]; +void gm107_gr_init_bios(struct gf100_gr *); + +void gm200_gr_init_gpc_mmu(struct gf100_gr *); + +struct gf100_gr_fwif { + int version; + int (*load)(struct gf100_gr *, int ver, const struct gf100_gr_fwif *); + const struct gf100_gr_func *func; + const struct nvkm_acr_lsf_func *fecs; + const struct nvkm_acr_lsf_func *gpccs; +}; + +int gf100_gr_load(struct gf100_gr *, int, const struct gf100_gr_fwif *); +int gf100_gr_nofw(struct gf100_gr *, int, const struct gf100_gr_fwif *); + +int gk20a_gr_load_sw(struct gf100_gr *, const char *path, int ver); +int gk20a_gr_load_net(struct gf100_gr *, const char *, const char *, int, + int (*)(struct nvkm_blob *, struct gf100_gr_pack **), + struct gf100_gr_pack **); + +int gm200_gr_nofw(struct gf100_gr *, int, const struct gf100_gr_fwif *); +int gm200_gr_load(struct gf100_gr *, int, const struct gf100_gr_fwif *); +extern const struct nvkm_acr_lsf_func gm200_gr_gpccs_acr; +extern const struct nvkm_acr_lsf_func gm200_gr_fecs_acr; + +extern const struct nvkm_acr_lsf_func gm20b_gr_fecs_acr; +void gm20b_gr_acr_bld_write(struct nvkm_acr *, u32, struct nvkm_acr_lsfw *); +void gm20b_gr_acr_bld_patch(struct nvkm_acr *, u32, s64); + +extern const struct nvkm_acr_lsf_func gp108_gr_gpccs_acr; +extern const struct nvkm_acr_lsf_func gp108_gr_fecs_acr; +void gp108_gr_acr_bld_write(struct nvkm_acr *, u32, struct nvkm_acr_lsfw *); +void gp108_gr_acr_bld_patch(struct nvkm_acr *, u32, s64); + +int gf100_gr_new_(const struct gf100_gr_fwif *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_gr **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c new file mode 100644 index 0000000000..63bd29c22f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c @@ -0,0 +1,161 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gf104_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x00002834 }, + { 0x405908, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf104_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ac8, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419abc, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf104_gr_init_pe_0[] = { + { 0x41980c, 3, 0x04, 0x00000000 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x41984c, 1, 0x04, 0x00005bc5 }, + { 0x419850, 4, 0x04, 0x00000000 }, + { 0x419880, 1, 0x04, 0x00000002 }, + {} +}; + +const struct gf100_gr_init +gf104_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00001100 }, + { 0x419eac, 1, 0x04, 0x11100702 }, + { 0x419eb0, 1, 0x04, 0x00000003 }, + { 0x419eb4, 4, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x0e063818 }, + { 0x419ecc, 1, 0x04, 0x0e060e06 }, + { 0x419ed0, 1, 0x04, 0x00003818 }, + { 0x419ed4, 1, 0x04, 0x011104f1 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419f2c, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf104_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf100_gr_init_pd_0 }, + { gf104_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf100_gr_init_prop_0 }, + { gf100_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf100_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf100_gr_init_gpm_0 }, + { gf100_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf104_gr_init_tex_0 }, + { gf104_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf100_gr_init_wwdx_0 }, + { gf100_gr_init_tpccs_1 }, + { gf100_gr_init_mpc_0 }, + { gf104_gr_init_sm_0 }, + { gf100_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct gf100_gr_func +gf104_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf100_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf104_gr_pack_mmio, + .fecs.ucode = &gf100_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf100_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .grctx = &gf104_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +static const struct gf100_gr_fwif +gf104_gr_fwif[] = { + { -1, gf100_gr_load, &gf104_gr }, + { -1, gf100_gr_nofw, &gf104_gr }, + {} +}; + +int +gf104_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf104_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c new file mode 100644 index 0000000000..495a844f92 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c @@ -0,0 +1,160 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gf108_gr_init_gpc_unk_0[] = { + { 0x418604, 1, 0x04, 0x00000000 }, + { 0x418680, 1, 0x04, 0x00000000 }, + { 0x418714, 1, 0x04, 0x00000000 }, + { 0x418384, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf108_gr_init_setup_1[] = { + { 0x4188c8, 2, 0x04, 0x00000000 }, + { 0x4188d0, 1, 0x04, 0x00010000 }, + { 0x4188d4, 1, 0x04, 0x00000001 }, + {} +}; + +static const struct gf100_gr_init +gf108_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x00000003 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf108_gr_init_pe_0[] = { + { 0x41980c, 1, 0x04, 0x00000010 }, + { 0x419810, 1, 0x04, 0x00000000 }, + { 0x419814, 1, 0x04, 0x00000004 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x41984c, 1, 0x04, 0x00005bc5 }, + { 0x419850, 4, 0x04, 0x00000000 }, + { 0x419880, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_pack +gf108_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf100_gr_init_pd_0 }, + { gf104_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf100_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf100_gr_init_gpm_0 }, + { gf108_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf104_gr_init_tex_0 }, + { gf108_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf100_gr_init_wwdx_0 }, + { gf100_gr_init_tpccs_1 }, + { gf100_gr_init_mpc_0 }, + { gf104_gr_init_sm_0 }, + { gf100_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static void +gf108_gr_init_r405a14(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x405a14, 0x80000000); +} + +static const struct gf100_gr_func +gf108_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_r405a14 = gf108_gr_init_r405a14, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf100_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf108_gr_pack_mmio, + .fecs.ucode = &gf100_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf100_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .grctx = &gf108_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +static const struct gf100_gr_fwif +gf108_gr_fwif[] = { + { -1, gf100_gr_load, &gf108_gr }, + { -1, gf100_gr_nofw, &gf108_gr }, + {} +}; + +int +gf108_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf108_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c new file mode 100644 index 0000000000..70fad235d1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c @@ -0,0 +1,136 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf110_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00001100 }, + { 0x419eac, 1, 0x04, 0x11100f02 }, + { 0x419eb0, 1, 0x04, 0x00000003 }, + { 0x419eb4, 4, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x06060618 }, + { 0x419ed0, 1, 0x04, 0x0eff0e38 }, + { 0x419ed4, 1, 0x04, 0x011104f1 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419f2c, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf110_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf100_gr_init_pd_0 }, + { gf100_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf100_gr_init_prop_0 }, + { gf100_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf100_gr_init_gpm_0 }, + { gf100_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf100_gr_init_tex_0 }, + { gf100_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf100_gr_init_wwdx_0 }, + { gf100_gr_init_tpccs_1 }, + { gf100_gr_init_mpc_0 }, + { gf110_gr_init_sm_0 }, + { gf100_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + { gf100_gr_init_pe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct gf100_gr_func +gf110_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf100_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf110_gr_pack_mmio, + .fecs.ucode = &gf100_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf100_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .grctx = &gf110_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gf110_gr_fwif[] = { + { -1, gf100_gr_load, &gf110_gr }, + { -1, gf100_gr_nofw, &gf110_gr }, + {} +}; + +int +gf110_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf110_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c new file mode 100644 index 0000000000..f127282480 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c @@ -0,0 +1,203 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gf117_gr_init_pe_0[] = { + { 0x41980c, 1, 0x04, 0x00000010 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x41984c, 1, 0x04, 0x00005bc8 }, + { 0x419850, 3, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf117_gr_init_pes_0[] = { + { 0x41be04, 1, 0x04, 0x00000000 }, + { 0x41be08, 1, 0x04, 0x00000004 }, + { 0x41be0c, 1, 0x04, 0x00000000 }, + { 0x41be10, 1, 0x04, 0x003b8bc7 }, + { 0x41be14, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf117_gr_init_wwdx_0[] = { + { 0x41bfd4, 1, 0x04, 0x00800000 }, + { 0x41bfdc, 1, 0x04, 0x00000000 }, + { 0x41bff8, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf117_gr_init_cbm_0[] = { + { 0x41becc, 1, 0x04, 0x00000000 }, + { 0x41bee8, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf117_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gf119_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf119_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gf119_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf119_gr_init_tex_0 }, + { gf117_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gf119_gr_init_sm_0 }, + { gf117_gr_init_pes_0 }, + { gf117_gr_init_wwdx_0 }, + { gf117_gr_init_cbm_0 }, + { gf100_gr_init_be_0 }, + { gf119_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +#include "fuc/hubgf117.fuc3.h" + +static struct gf100_gr_ucode +gf117_gr_fecs_ucode = { + .code.data = gf117_grhub_code, + .code.size = sizeof(gf117_grhub_code), + .data.data = gf117_grhub_data, + .data.size = sizeof(gf117_grhub_data), +}; + +#include "fuc/gpcgf117.fuc3.h" + +static struct gf100_gr_ucode +gf117_gr_gpccs_ucode = { + .code.data = gf117_grgpc_code, + .code.size = sizeof(gf117_grgpc_code), + .data.data = gf117_grgpc_data, + .data.size = sizeof(gf117_grgpc_data), +}; + +void +gf117_gr_init_zcull(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + /*TODO: fill in litter vals for gf117-gm2xx */ + const u8 tile_nr = !gr->func->gpc_nr ? ALIGN(gr->tpc_total, 32) : + (gr->func->gpc_nr * gr->func->tpc_nr); + u8 bank[GPC_MAX] = {}, gpc, i, j; + u32 data; + + for (i = 0; i < tile_nr; i += 8) { + for (data = 0, j = 0; j < 8 && i + j < gr->tpc_total; j++) { + data |= bank[gr->tile[i + j]] << (j * 4); + bank[gr->tile[i + j]]++; + } + nvkm_wr32(device, GPC_BCAST(0x0980 + ((i / 8) * 4)), data); + } + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->screen_tile_row_offset << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); + } + + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); +} + +static const struct gf100_gr_func +gf117_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf117_gr_pack_mmio, + .fecs.ucode = &gf117_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf117_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 1, + .grctx = &gf117_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gf117_gr_fwif[] = { + { -1, gf100_gr_load, &gf117_gr }, + { -1, gf100_gr_nofw, &gf117_gr }, + {} +}; + +int +gf117_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf117_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c new file mode 100644 index 0000000000..75ceb514c0 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c @@ -0,0 +1,227 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gf119_gr_init_pd_0[] = { + { 0x406024, 1, 0x04, 0x00000000 }, + { 0x4064f0, 3, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x00002834 }, + { 0x405908, 1, 0x04, 0x00000000 }, + { 0x405928, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_prop_0[] = { + { 0x418408, 1, 0x04, 0x00000000 }, + { 0x4184a0, 3, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_gpm_0[] = { + { 0x418c04, 1, 0x04, 0x00000000 }, + { 0x418c64, 2, 0x04, 0x00000000 }, + { 0x418c88, 1, 0x04, 0x00000000 }, + { 0x418cb4, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418d28, 2, 0x04, 0x00000000 }, + { 0x418f00, 1, 0x04, 0x00000000 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418f20, 2, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x00000003 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + { 0x418e1c, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ac8, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419abc, 2, 0x04, 0x00000000 }, + { 0x419ab4, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf119_gr_init_pe_0[] = { + { 0x41980c, 1, 0x04, 0x00000010 }, + { 0x419810, 1, 0x04, 0x00000000 }, + { 0x419814, 1, 0x04, 0x00000004 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x41984c, 1, 0x04, 0x0000a918 }, + { 0x419850, 4, 0x04, 0x00000000 }, + { 0x419880, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_init +gf119_gr_init_wwdx_0[] = { + { 0x419bd4, 1, 0x04, 0x00800000 }, + { 0x419bdc, 1, 0x04, 0x00000000 }, + { 0x419bf8, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gf119_gr_init_tpccs_1[] = { + { 0x419d2c, 1, 0x04, 0x00000000 }, + { 0x419d48, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x02001100 }, + { 0x419eac, 1, 0x04, 0x11100702 }, + { 0x419eb0, 1, 0x04, 0x00000003 }, + { 0x419eb4, 4, 0x04, 0x00000000 }, + { 0x419ec8, 1, 0x04, 0x0e063818 }, + { 0x419ecc, 1, 0x04, 0x0e060e06 }, + { 0x419ed0, 1, 0x04, 0x00003818 }, + { 0x419ed4, 1, 0x04, 0x011104f1 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419f2c, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gf119_gr_init_fe_1[] = { + { 0x40402c, 1, 0x04, 0x00000000 }, + { 0x4040f0, 1, 0x04, 0x00000000 }, + { 0x404174, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gf119_gr_pack_mmio[] = { + { gf100_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gf119_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gf119_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gf119_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gf100_gr_init_tpccs_0 }, + { gf119_gr_init_tex_0 }, + { gf119_gr_init_pe_0 }, + { gf100_gr_init_l1c_0 }, + { gf119_gr_init_wwdx_0 }, + { gf119_gr_init_tpccs_1 }, + { gf100_gr_init_mpc_0 }, + { gf119_gr_init_sm_0 }, + { gf100_gr_init_be_0 }, + { gf119_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct gf100_gr_func +gf119_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, + .init_zcull = gf100_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_40601c = gf100_gr_init_40601c, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gf119_gr_pack_mmio, + .fecs.ucode = &gf100_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gf100_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .grctx = &gf119_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gf119_gr_fwif[] = { + { -1, gf100_gr_load, &gf119_gr }, + { -1, gf100_gr_nofw, &gf119_gr }, + {} +}; + +int +gf119_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gf119_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c new file mode 100644 index 0000000000..e53ade24ad --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c @@ -0,0 +1,506 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "gk104.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gk104_gr_init_main_0[] = { + { 0x400080, 1, 0x04, 0x003083c2 }, + { 0x400088, 1, 0x04, 0x0001ffe7 }, + { 0x40008c, 1, 0x04, 0x00000000 }, + { 0x400090, 1, 0x04, 0x00000030 }, + { 0x40013c, 1, 0x04, 0x003901f7 }, + { 0x400140, 1, 0x04, 0x00000100 }, + { 0x400144, 1, 0x04, 0x00000000 }, + { 0x400148, 1, 0x04, 0x00000110 }, + { 0x400138, 1, 0x04, 0x00000000 }, + { 0x400130, 2, 0x04, 0x00000000 }, + { 0x400124, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x0000ff34 }, + { 0x405908, 1, 0x04, 0x00000000 }, + { 0x405928, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_sked_0[] = { + { 0x407010, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_cwd_0[] = { + { 0x405b50, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418d28, 2, 0x04, 0x00000000 }, + { 0x418f00, 1, 0x04, 0x00000000 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418f20, 2, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x00000060 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + { 0x418e1c, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk104_gr_init_gpc_unk_2[] = { + { 0x418884, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk104_gr_init_tpccs_0[] = { + { 0x419d0c, 1, 0x04, 0x00000000 }, + { 0x419d10, 1, 0x04, 0x00000014 }, + {} +}; + +const struct gf100_gr_init +gk104_gr_init_pe_0[] = { + { 0x41980c, 1, 0x04, 0x00000010 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x419850, 1, 0x04, 0x00000004 }, + { 0x419854, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419ca8, 1, 0x04, 0x00000000 }, + { 0x419cb0, 1, 0x04, 0x01000000 }, + { 0x419cb4, 1, 0x04, 0x00000000 }, + { 0x419cb8, 1, 0x04, 0x00b08bea }, + { 0x419c84, 1, 0x04, 0x00010384 }, + { 0x419cbc, 1, 0x04, 0x28137646 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + { 0x419c80, 1, 0x04, 0x00020232 }, + {} +}; + +static const struct gf100_gr_init +gk104_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ee4, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00000000 }, + { 0x419eb4, 4, 0x04, 0x00000000 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419f74, 1, 0x04, 0x00000555 }, + {} +}; + +const struct gf100_gr_init +gk104_gr_init_be_0[] = { + { 0x40880c, 1, 0x04, 0x00000000 }, + { 0x408850, 1, 0x04, 0x00000004 }, + { 0x408910, 9, 0x04, 0x00000000 }, + { 0x408950, 1, 0x04, 0x00000000 }, + { 0x408954, 1, 0x04, 0x0000ffff }, + { 0x408958, 1, 0x04, 0x00000034 }, + { 0x408984, 1, 0x04, 0x00000000 }, + { 0x408988, 1, 0x04, 0x08040201 }, + { 0x40898c, 1, 0x04, 0x80402010 }, + {} +}; + +const struct gf100_gr_pack +gk104_gr_pack_mmio[] = { + { gk104_gr_init_main_0 }, + { gf100_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gk104_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gk104_gr_init_sked_0 }, + { gk104_gr_init_cwd_0 }, + { gf119_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gk104_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gk104_gr_init_gpc_unk_2 }, + { gk104_gr_init_tpccs_0 }, + { gf119_gr_init_tex_0 }, + { gk104_gr_init_pe_0 }, + { gk104_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gk104_gr_init_sm_0 }, + { gf117_gr_init_pes_0 }, + { gf117_gr_init_wwdx_0 }, + { gf117_gr_init_cbm_0 }, + { gk104_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_main_0[] = { + { 0x4041f0, 1, 0x00004046 }, + { 0x409890, 1, 0x00000045 }, + { 0x4098b0, 1, 0x0000007f }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_rstr2d_0[] = { + { 0x4078c0, 1, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_unk_0[] = { + { 0x406000, 1, 0x00004044 }, + { 0x405860, 1, 0x00004042 }, + { 0x40590c, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gcc_0[] = { + { 0x408040, 1, 0x00004044 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_sked_0[] = { + { 0x407000, 1, 0x00004044 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_unk_1[] = { + { 0x405bf0, 1, 0x00004044 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_ctxctl_0[] = { + { 0x41a890, 1, 0x00000042 }, + { 0x41a8b0, 1, 0x0000007f }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_unk_0[] = { + { 0x418500, 1, 0x00004042 }, + { 0x418608, 1, 0x00004042 }, + { 0x418688, 1, 0x00004042 }, + { 0x418718, 1, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_esetup_0[] = { + { 0x418828, 1, 0x00000044 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_tpbus_0[] = { + { 0x418bbc, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_zcull_0[] = { + { 0x418970, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_tpconf_0[] = { + { 0x418c70, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_unk_1[] = { + { 0x418cf0, 1, 0x00004042 }, + { 0x418d70, 1, 0x00004042 }, + { 0x418f0c, 1, 0x00004042 }, + { 0x418e0c, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_gcc_0[] = { + { 0x419020, 1, 0x00004042 }, + { 0x419038, 1, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_ffb_0[] = { + { 0x418898, 1, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_tex_0[] = { + { 0x419a40, 9, 0x00004042 }, + { 0x419acc, 1, 0x00004047 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_poly_0[] = { + { 0x419868, 1, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_l1c_0[] = { + { 0x419ccc, 3, 0x00000042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_unk_2[] = { + { 0x419c70, 1, 0x00004045 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_mp_0[] = { + { 0x419fd0, 1, 0x00004043 }, + { 0x419fd8, 1, 0x00004049 }, + { 0x419fe0, 2, 0x00004042 }, + { 0x419ff0, 1, 0x00004046 }, + { 0x419ff8, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_gpc_ppc_0[] = { + { 0x41be28, 1, 0x00000042 }, + { 0x41bfe8, 1, 0x00004042 }, + { 0x41bed0, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_rop_zrop_0[] = { + { 0x408810, 2, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_rop_0[] = { + { 0x408a80, 6, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_rop_crop_0[] = { + { 0x4089a8, 1, 0x00004042 }, + { 0x4089b0, 1, 0x00000042 }, + { 0x4089b8, 1, 0x00004042 }, + {} +}; + +const struct nvkm_therm_clkgate_init +gk104_clkgate_blcg_init_pxbar_0[] = { + { 0x13c820, 1, 0x0001007f }, + { 0x13cbe0, 1, 0x00000042 }, + {} +}; + +static const struct nvkm_therm_clkgate_pack +gk104_clkgate_pack[] = { + { gk104_clkgate_blcg_init_main_0 }, + { gk104_clkgate_blcg_init_rstr2d_0 }, + { gk104_clkgate_blcg_init_unk_0 }, + { gk104_clkgate_blcg_init_gcc_0 }, + { gk104_clkgate_blcg_init_sked_0 }, + { gk104_clkgate_blcg_init_unk_1 }, + { gk104_clkgate_blcg_init_gpc_ctxctl_0 }, + { gk104_clkgate_blcg_init_gpc_unk_0 }, + { gk104_clkgate_blcg_init_gpc_esetup_0 }, + { gk104_clkgate_blcg_init_gpc_tpbus_0 }, + { gk104_clkgate_blcg_init_gpc_zcull_0 }, + { gk104_clkgate_blcg_init_gpc_tpconf_0 }, + { gk104_clkgate_blcg_init_gpc_unk_1 }, + { gk104_clkgate_blcg_init_gpc_gcc_0 }, + { gk104_clkgate_blcg_init_gpc_ffb_0 }, + { gk104_clkgate_blcg_init_gpc_tex_0 }, + { gk104_clkgate_blcg_init_gpc_poly_0 }, + { gk104_clkgate_blcg_init_gpc_l1c_0 }, + { gk104_clkgate_blcg_init_gpc_unk_2 }, + { gk104_clkgate_blcg_init_gpc_mp_0 }, + { gk104_clkgate_blcg_init_gpc_ppc_0 }, + { gk104_clkgate_blcg_init_rop_zrop_0 }, + { gk104_clkgate_blcg_init_rop_0 }, + { gk104_clkgate_blcg_init_rop_crop_0 }, + { gk104_clkgate_blcg_init_pxbar_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +void +gk104_gr_init_sked_hww_esr(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x407020, 0x40000000); +} + +static void +gk104_gr_init_fecs_exceptions(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x409ffc, 0x00000000); + nvkm_wr32(device, 0x409c14, 0x00003e3e); + nvkm_wr32(device, 0x409c24, 0x000f0001); +} + +void +gk104_gr_init_rop_active_fbps(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 fbp_count = nvkm_rd32(device, 0x120074); + nvkm_mask(device, 0x408850, 0x0000000f, fbp_count); /* zrop */ + nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */ +} + +void +gk104_gr_init_ppc_exceptions(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int gpc, ppc; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->func->ppc_nr; ppc++) { + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; + nvkm_wr32(device, PPC_UNIT(gpc, ppc, 0x038), 0xc0000000); + } + } +} + +void +gk104_gr_init_vsc_stream_master(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, GPC_UNIT(0, 0x3018), 0x00000001); +} + +#include "fuc/hubgk104.fuc3.h" + +static struct gf100_gr_ucode +gk104_gr_fecs_ucode = { + .code.data = gk104_grhub_code, + .code.size = sizeof(gk104_grhub_code), + .data.data = gk104_grhub_data, + .data.size = sizeof(gk104_grhub_data), +}; + +#include "fuc/gpcgk104.fuc3.h" + +static struct gf100_gr_ucode +gk104_gr_gpccs_ucode = { + .code.data = gk104_grgpc_code, + .code.size = sizeof(gk104_grgpc_code), + .data.data = gk104_grgpc_data, + .data.size = sizeof(gk104_grgpc_data), +}; + +static const struct gf100_gr_func +gk104_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .init_fecs_exceptions = gk104_gr_init_fecs_exceptions, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gf100_gr_init_419eb4, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gk104_gr_pack_mmio, + .fecs.ucode = &gk104_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gk104_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 1, + .grctx = &gk104_grctx, + .clkgate_pack = gk104_clkgate_pack, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_A }, + { -1, -1, KEPLER_A, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_A }, + {} + } +}; + +static const struct gf100_gr_fwif +gk104_gr_fwif[] = { + { -1, gf100_gr_load, &gk104_gr }, + { -1, gf100_gr_nofw, &gk104_gr }, + {} +}; + +int +gk104_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gk104_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.h new file mode 100644 index 0000000000..a24c177365 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.h @@ -0,0 +1,55 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Lyude Paul + */ +#ifndef __GK104_GR_H__ +#define __GK104_GR_H__ + +#include + +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_main_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_rstr2d_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_unk_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gcc_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_sked_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_unk_1[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_ctxctl_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_unk_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_esetup_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_tpbus_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_zcull_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_tpconf_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_unk_1[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_gcc_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_ffb_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_tex_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_poly_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_l1c_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_unk_2[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_mp_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_gpc_ppc_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_rop_zrop_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_rop_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_rop_crop_0[]; +extern const struct nvkm_therm_clkgate_init gk104_clkgate_blcg_init_pxbar_0[]; + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c new file mode 100644 index 0000000000..c7e1c5dbc6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c @@ -0,0 +1,402 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "gk104.h" +#include "ctxgf100.h" + +#include + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +const struct gf100_gr_init +gk110_gr_init_fe_0[] = { + { 0x40415c, 1, 0x04, 0x00000000 }, + { 0x404170, 1, 0x04, 0x00000000 }, + { 0x4041b4, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x0000ff00 }, + { 0x405908, 1, 0x04, 0x00000000 }, + { 0x405928, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_sked_0[] = { + { 0x407010, 1, 0x04, 0x00000000 }, + { 0x407040, 1, 0x04, 0x80440424 }, + { 0x407048, 1, 0x04, 0x0000000a }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_cwd_0[] = { + { 0x405b44, 1, 0x04, 0x00000000 }, + { 0x405b50, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418d28, 2, 0x04, 0x00000000 }, + { 0x418f00, 1, 0x04, 0x00000400 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418f20, 2, 0x04, 0x00000000 }, + { 0x418e00, 1, 0x04, 0x00000000 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + { 0x418e1c, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ac8, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419aec, 1, 0x04, 0x00000000 }, + { 0x419abc, 2, 0x04, 0x00000000 }, + { 0x419ab4, 1, 0x04, 0x00000000 }, + { 0x419aa8, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk110_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419ca8, 1, 0x04, 0x00000000 }, + { 0x419cb0, 1, 0x04, 0x01000000 }, + { 0x419cb4, 1, 0x04, 0x00000000 }, + { 0x419cb8, 1, 0x04, 0x00b08bea }, + { 0x419c84, 1, 0x04, 0x00010384 }, + { 0x419cbc, 1, 0x04, 0x281b3646 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + { 0x419c80, 1, 0x04, 0x00020230 }, + { 0x419ccc, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk110_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000080 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ee4, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00000000 }, + { 0x419eb4, 1, 0x04, 0x00000000 }, + { 0x419ebc, 2, 0x04, 0x00000000 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419ed0, 1, 0x04, 0x00003234 }, + { 0x419f74, 1, 0x04, 0x00015555 }, + { 0x419f80, 4, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gk110_gr_pack_mmio[] = { + { gk104_gr_init_main_0 }, + { gk110_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gk110_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gk110_gr_init_sked_0 }, + { gk110_gr_init_cwd_0 }, + { gf119_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gk110_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gk104_gr_init_gpc_unk_2 }, + { gk104_gr_init_tpccs_0 }, + { gk110_gr_init_tex_0 }, + { gk104_gr_init_pe_0 }, + { gk110_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gk110_gr_init_sm_0 }, + { gf117_gr_init_pes_0 }, + { gf117_gr_init_wwdx_0 }, + { gf117_gr_init_cbm_0 }, + { gk104_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_blcg_init_sked_0[] = { + { 0x407000, 1, 0x00004041 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_blcg_init_gpc_gcc_0[] = { + { 0x419020, 1, 0x00000042 }, + { 0x419038, 1, 0x00000042 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_blcg_init_gpc_l1c_0[] = { + { 0x419cd4, 2, 0x00004042 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_blcg_init_gpc_mp_0[] = { + { 0x419fd0, 1, 0x00004043 }, + { 0x419fd8, 1, 0x00004049 }, + { 0x419fe0, 2, 0x00004042 }, + { 0x419ff0, 1, 0x00000046 }, + { 0x419ff8, 1, 0x00004042 }, + { 0x419f90, 1, 0x00004042 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_main_0[] = { + { 0x4041f4, 1, 0x00000000 }, + { 0x409894, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_unk_0[] = { + { 0x406004, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_sked_0[] = { + { 0x407004, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_ctxctl_0[] = { + { 0x41a894, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_unk_0[] = { + { 0x418504, 1, 0x00000000 }, + { 0x41860c, 1, 0x00000000 }, + { 0x41868c, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_esetup_0[] = { + { 0x41882c, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_zcull_0[] = { + { 0x418974, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_l1c_0[] = { + { 0x419cd8, 2, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_unk_1[] = { + { 0x419c74, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_mp_0[] = { + { 0x419fd4, 1, 0x00004a4a }, + { 0x419fdc, 1, 0x00000014 }, + { 0x419fe4, 1, 0x00000000 }, + { 0x419ff4, 1, 0x00001724 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_gpc_ppc_0[] = { + { 0x41be2c, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_init +gk110_clkgate_slcg_init_pcounter_0[] = { + { 0x1be018, 1, 0x000001ff }, + { 0x1bc018, 1, 0x000001ff }, + { 0x1b8018, 1, 0x000001ff }, + { 0x1b4124, 1, 0x00000000 }, + {} +}; + +static const struct nvkm_therm_clkgate_pack +gk110_clkgate_pack[] = { + { gk104_clkgate_blcg_init_main_0 }, + { gk104_clkgate_blcg_init_rstr2d_0 }, + { gk104_clkgate_blcg_init_unk_0 }, + { gk104_clkgate_blcg_init_gcc_0 }, + { gk110_clkgate_blcg_init_sked_0 }, + { gk104_clkgate_blcg_init_unk_1 }, + { gk104_clkgate_blcg_init_gpc_ctxctl_0 }, + { gk104_clkgate_blcg_init_gpc_unk_0 }, + { gk104_clkgate_blcg_init_gpc_esetup_0 }, + { gk104_clkgate_blcg_init_gpc_tpbus_0 }, + { gk104_clkgate_blcg_init_gpc_zcull_0 }, + { gk104_clkgate_blcg_init_gpc_tpconf_0 }, + { gk104_clkgate_blcg_init_gpc_unk_1 }, + { gk110_clkgate_blcg_init_gpc_gcc_0 }, + { gk104_clkgate_blcg_init_gpc_ffb_0 }, + { gk104_clkgate_blcg_init_gpc_tex_0 }, + { gk104_clkgate_blcg_init_gpc_poly_0 }, + { gk110_clkgate_blcg_init_gpc_l1c_0 }, + { gk104_clkgate_blcg_init_gpc_unk_2 }, + { gk110_clkgate_blcg_init_gpc_mp_0 }, + { gk104_clkgate_blcg_init_gpc_ppc_0 }, + { gk104_clkgate_blcg_init_rop_zrop_0 }, + { gk104_clkgate_blcg_init_rop_0 }, + { gk104_clkgate_blcg_init_rop_crop_0 }, + { gk104_clkgate_blcg_init_pxbar_0 }, + { gk110_clkgate_slcg_init_main_0 }, + { gk110_clkgate_slcg_init_unk_0 }, + { gk110_clkgate_slcg_init_sked_0 }, + { gk110_clkgate_slcg_init_gpc_ctxctl_0 }, + { gk110_clkgate_slcg_init_gpc_unk_0 }, + { gk110_clkgate_slcg_init_gpc_esetup_0 }, + { gk110_clkgate_slcg_init_gpc_zcull_0 }, + { gk110_clkgate_slcg_init_gpc_l1c_0 }, + { gk110_clkgate_slcg_init_gpc_unk_1 }, + { gk110_clkgate_slcg_init_gpc_mp_0 }, + { gk110_clkgate_slcg_init_gpc_ppc_0 }, + { gk110_clkgate_slcg_init_pcounter_0 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +#include "fuc/hubgk110.fuc3.h" + +struct gf100_gr_ucode +gk110_gr_fecs_ucode = { + .code.data = gk110_grhub_code, + .code.size = sizeof(gk110_grhub_code), + .data.data = gk110_grhub_data, + .data.size = sizeof(gk110_grhub_data), +}; + +#include "fuc/gpcgk110.fuc3.h" + +struct gf100_gr_ucode +gk110_gr_gpccs_ucode = { + .code.data = gk110_grgpc_code, + .code.size = sizeof(gk110_grgpc_code), + .data.data = gk110_grgpc_data, + .data.size = sizeof(gk110_grgpc_data), +}; + +void +gk110_gr_init_419eb4(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); + nvkm_mask(device, 0x419eb4, 0x00002000, 0x00002000); + nvkm_mask(device, 0x419eb4, 0x00004000, 0x00004000); + nvkm_mask(device, 0x419eb4, 0x00008000, 0x00008000); + nvkm_mask(device, 0x419eb4, 0x00001000, 0x00000000); + nvkm_mask(device, 0x419eb4, 0x00002000, 0x00000000); + nvkm_mask(device, 0x419eb4, 0x00004000, 0x00000000); + nvkm_mask(device, 0x419eb4, 0x00008000, 0x00000000); +} + +static const struct gf100_gr_func +gk110_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gk110_gr_init_419eb4, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gk110_gr_pack_mmio, + .fecs.ucode = &gk110_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gk110_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 2, + .grctx = &gk110_grctx, + .clkgate_pack = gk110_clkgate_pack, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gk110_gr_fwif[] = { + { -1, gf100_gr_load, &gk110_gr }, + { -1, gf100_gr_nofw, &gk110_gr }, + {} +}; + +int +gk110_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gk110_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c new file mode 100644 index 0000000000..458abae571 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c @@ -0,0 +1,154 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk110b_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419ca8, 1, 0x04, 0x00000000 }, + { 0x419cb0, 1, 0x04, 0x09000000 }, + { 0x419cb4, 1, 0x04, 0x00000000 }, + { 0x419cb8, 1, 0x04, 0x00b08bea }, + { 0x419c84, 1, 0x04, 0x00010384 }, + { 0x419cbc, 1, 0x04, 0x281b3646 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + { 0x419c80, 1, 0x04, 0x00020230 }, + { 0x419ccc, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk110b_gr_init_sm_0[] = { + { 0x419e00, 1, 0x04, 0x00000080 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ee4, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x00000000 }, + { 0x419eb4, 1, 0x04, 0x00000000 }, + { 0x419ebc, 2, 0x04, 0x00000000 }, + { 0x419edc, 1, 0x04, 0x00000000 }, + { 0x419f00, 1, 0x04, 0x00000000 }, + { 0x419ed0, 1, 0x04, 0x00002616 }, + { 0x419f74, 1, 0x04, 0x00015555 }, + { 0x419f80, 4, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gk110b_gr_pack_mmio[] = { + { gk104_gr_init_main_0 }, + { gk110_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gk110_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gk110_gr_init_sked_0 }, + { gk110_gr_init_cwd_0 }, + { gf119_gr_init_prop_0 }, + { gf108_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gf108_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gk110_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gk104_gr_init_gpc_unk_2 }, + { gk104_gr_init_tpccs_0 }, + { gk110_gr_init_tex_0 }, + { gk104_gr_init_pe_0 }, + { gk110b_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gk110b_gr_init_sm_0 }, + { gf117_gr_init_pes_0 }, + { gf117_gr_init_wwdx_0 }, + { gf117_gr_init_cbm_0 }, + { gk104_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct gf100_gr_func +gk110b_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419eb4 = gk110_gr_init_419eb4, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gk110b_gr_pack_mmio, + .fecs.ucode = &gk110_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gk110_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 2, + .grctx = &gk110b_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gk110b_gr_fwif[] = { + { -1, gf100_gr_load, &gk110b_gr }, + { -1, gf100_gr_nofw, &gk110b_gr }, + {} +}; + +int +gk110b_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gk110b_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c new file mode 100644 index 0000000000..d3f6b65c21 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c @@ -0,0 +1,211 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gk208_gr_init_main_0[] = { + { 0x400080, 1, 0x04, 0x003083c2 }, + { 0x400088, 1, 0x04, 0x0001bfe7 }, + { 0x40008c, 1, 0x04, 0x00000000 }, + { 0x400090, 1, 0x04, 0x00000030 }, + { 0x40013c, 1, 0x04, 0x003901f7 }, + { 0x400140, 1, 0x04, 0x00000100 }, + { 0x400144, 1, 0x04, 0x00000000 }, + { 0x400148, 1, 0x04, 0x00000110 }, + { 0x400138, 1, 0x04, 0x00000000 }, + { 0x400130, 2, 0x04, 0x00000000 }, + { 0x400124, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_init +gk208_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x00000000 }, + { 0x405908, 1, 0x04, 0x00000000 }, + { 0x405928, 2, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gk208_gr_init_gpc_unk_0[] = { + { 0x418604, 1, 0x04, 0x00000000 }, + { 0x418680, 1, 0x04, 0x00000000 }, + { 0x418714, 1, 0x04, 0x00000000 }, + { 0x418384, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk208_gr_init_setup_1[] = { + { 0x4188c8, 2, 0x04, 0x00000000 }, + { 0x4188d0, 1, 0x04, 0x00010000 }, + { 0x4188d4, 1, 0x04, 0x00000201 }, + {} +}; + +static const struct gf100_gr_init +gk208_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ac8, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419abc, 2, 0x04, 0x00000000 }, + { 0x419ab4, 1, 0x04, 0x00000000 }, + { 0x419aa8, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gk208_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419ca8, 1, 0x04, 0x00000000 }, + { 0x419cb0, 1, 0x04, 0x01000000 }, + { 0x419cb4, 1, 0x04, 0x00000000 }, + { 0x419cb8, 1, 0x04, 0x00b08bea }, + { 0x419c84, 1, 0x04, 0x00010384 }, + { 0x419cbc, 1, 0x04, 0x281b3646 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + { 0x419c80, 1, 0x04, 0x00000230 }, + { 0x419ccc, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gk208_gr_pack_mmio[] = { + { gk208_gr_init_main_0 }, + { gk110_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf119_gr_init_pd_0 }, + { gk208_gr_init_ds_0 }, + { gf100_gr_init_scc_0 }, + { gk110_gr_init_sked_0 }, + { gk110_gr_init_cwd_0 }, + { gf119_gr_init_prop_0 }, + { gk208_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gk208_gr_init_setup_1 }, + { gf100_gr_init_zcull_0 }, + { gf119_gr_init_gpm_0 }, + { gk110_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gk104_gr_init_gpc_unk_2 }, + { gk104_gr_init_tpccs_0 }, + { gk208_gr_init_tex_0 }, + { gk104_gr_init_pe_0 }, + { gk208_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gk110_gr_init_sm_0 }, + { gf117_gr_init_pes_0 }, + { gf117_gr_init_wwdx_0 }, + { gf117_gr_init_cbm_0 }, + { gk104_gr_init_be_0 }, + { gf100_gr_init_fe_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +#include "fuc/hubgk208.fuc5.h" + +static struct gf100_gr_ucode +gk208_gr_fecs_ucode = { + .code.data = gk208_grhub_code, + .code.size = sizeof(gk208_grhub_code), + .data.data = gk208_grhub_data, + .data.size = sizeof(gk208_grhub_data), +}; + +#include "fuc/gpcgk208.fuc5.h" + +static struct gf100_gr_ucode +gk208_gr_gpccs_ucode = { + .code.data = gk208_grgpc_code, + .code.size = sizeof(gk208_grgpc_code), + .data.data = gk208_grgpc_data, + .data.size = sizeof(gk208_grgpc_data), +}; + +static const struct gf100_gr_func +gk208_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gf100_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_shader_exceptions = gf100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gf100_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gk208_gr_pack_mmio, + .fecs.ucode = &gk208_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gk208_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 1, + .grctx = &gk208_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +static const struct gf100_gr_fwif +gk208_gr_fwif[] = { + { -1, gf100_gr_load, &gk208_gr }, + { -1, gf100_gr_nofw, &gk208_gr }, + {} +}; + +int +gk208_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gk208_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c new file mode 100644 index 0000000000..035ea213f5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c @@ -0,0 +1,346 @@ +/* + * Copyright (c) 2014-2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include +#include + +#include + +struct gk20a_fw_av +{ + u32 addr; + u32 data; +}; + +int +gk20a_gr_av_to_init_(struct nvkm_blob *blob, u8 count, u32 pitch, struct gf100_gr_pack **ppack) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + int nent; + int i; + + nent = (blob->size / sizeof(struct gk20a_fw_av)); + pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); + if (!pack) + return -ENOMEM; + + init = (void *)(pack + 2); + pack[0].init = init; + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct gk20a_fw_av *av = &((struct gk20a_fw_av *)blob->data)[i]; + + ent->addr = av->addr; + ent->data = av->data; + ent->count = ((ent->addr & 0xffff) != 0xe100) ? count : 1; + ent->pitch = pitch; + } + + *ppack = pack; + return 0; +} + +int +gk20a_gr_av_to_init(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) +{ + return gk20a_gr_av_to_init_(blob, 1, 1, ppack); +} + +struct gk20a_fw_aiv +{ + u32 addr; + u32 index; + u32 data; +}; + +int +gk20a_gr_aiv_to_init(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + int nent; + int i; + + nent = (blob->size / sizeof(struct gk20a_fw_aiv)); + pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); + if (!pack) + return -ENOMEM; + + init = (void *)(pack + 2); + pack[0].init = init; + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct gk20a_fw_aiv *av = &((struct gk20a_fw_aiv *)blob->data)[i]; + + ent->addr = av->addr; + ent->data = av->data; + ent->count = 1; + ent->pitch = 1; + } + + *ppack = pack; + return 0; +} + +int +gk20a_gr_av_to_method(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + /* We don't suppose we will initialize more than 16 classes here... */ + static const unsigned int max_classes = 16; + u32 classidx = 0, prevclass = 0; + int nent; + int i; + + nent = (blob->size / sizeof(struct gk20a_fw_av)); + pack = vzalloc((sizeof(*pack) * (max_classes + 1)) + + (sizeof(*init) * (nent + max_classes + 1))); + if (!pack) + return -ENOMEM; + + init = (void *)(pack + max_classes + 1); + + for (i = 0; i < nent; i++, init++) { + struct gk20a_fw_av *av = &((struct gk20a_fw_av *)blob->data)[i]; + u32 class = av->addr & 0xffff; + u32 addr = (av->addr & 0xffff0000) >> 14; + + if (prevclass != class) { + if (prevclass) /* Add terminator to the method list. */ + init++; + pack[classidx].init = init; + pack[classidx].type = class; + prevclass = class; + if (++classidx >= max_classes) { + vfree(pack); + return -ENOSPC; + } + } + + init->addr = addr; + init->data = av->data; + init->count = 1; + init->pitch = 1; + } + + *ppack = pack; + return 0; +} + +static int +gk20a_gr_wait_mem_scrubbing(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x40910c) & 0x00000006)) + break; + ) < 0) { + nvkm_error(subdev, "FECS mem scrubbing timeout\n"); + return -ETIMEDOUT; + } + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x41a10c) & 0x00000006)) + break; + ) < 0) { + nvkm_error(subdev, "GPCCS mem scrubbing timeout\n"); + return -ETIMEDOUT; + } + + return 0; +} + +static void +gk20a_gr_set_hww_esr_report_mask(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x419e44, 0x1ffffe); + nvkm_wr32(device, 0x419e4c, 0x7f); +} + +int +gk20a_gr_init(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret; + + /* Clear SCC RAM */ + nvkm_wr32(device, 0x40802c, 0x1); + + gf100_gr_mmio(gr, gr->sw_nonctx); + + ret = gk20a_gr_wait_mem_scrubbing(gr); + if (ret) + return ret; + + ret = gf100_gr_wait_idle(gr); + if (ret) + return ret; + + /* MMU debug buffer */ + if (gr->func->init_gpc_mmu) + gr->func->init_gpc_mmu(gr); + + /* Set the PE as stream master */ + nvkm_mask(device, 0x503018, 0x1, 0x1); + + /* Zcull init */ + gr->func->init_zcull(gr); + + gr->func->init_rop_active_fbps(gr); + + /* Enable FIFO access */ + nvkm_wr32(device, 0x400500, 0x00010001); + + /* Enable interrupts */ + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + + /* Enable FECS error interrupts */ + nvkm_wr32(device, 0x409c24, 0x000f0000); + + /* Enable hardware warning exceptions */ + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + + if (gr->func->set_hww_esr_report_mask) + gr->func->set_hww_esr_report_mask(gr); + + /* Enable TPC exceptions per GPC */ + nvkm_wr32(device, 0x419d0c, 0x2); + nvkm_wr32(device, 0x41ac94, (((1 << gr->tpc_total) - 1) & 0xff) << 16); + + /* Reset and enable all exceptions */ + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); + + gf100_gr_zbc_init(gr); + + return gf100_gr_init_ctxctl(gr); +} + +static const struct gf100_gr_func +gk20a_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gk20a_gr_init, + .init_zcull = gf117_gr_init_zcull, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .trap_mp = gf100_gr_trap_mp, + .set_hww_esr_report_mask = gk20a_gr_set_hww_esr_report_mask, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gf100_gr_rops, + .ppc_nr = 1, + .grctx = &gk20a_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_A }, + { -1, -1, KEPLER_C, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_A }, + {} + } +}; + +int +gk20a_gr_load_net(struct gf100_gr *gr, const char *path, const char *name, int ver, + int (*load)(struct nvkm_blob *, struct gf100_gr_pack **), + struct gf100_gr_pack **ppack) +{ + struct nvkm_blob blob; + int ret; + + ret = nvkm_firmware_load_blob(&gr->base.engine.subdev, path, name, ver, &blob); + if (ret) + return ret; + + ret = load(&blob, ppack); + nvkm_blob_dtor(&blob); + return 0; +} + +int +gk20a_gr_load_sw(struct gf100_gr *gr, const char *path, int ver) +{ + if (gk20a_gr_load_net(gr, path, "sw_nonctx", ver, gk20a_gr_av_to_init, &gr->sw_nonctx) || + gk20a_gr_load_net(gr, path, "sw_ctx", ver, gk20a_gr_aiv_to_init, &gr->sw_ctx) || + gk20a_gr_load_net(gr, path, "sw_bundle_init", ver, gk20a_gr_av_to_init, &gr->bundle) || + gk20a_gr_load_net(gr, path, "sw_method_init", ver, gk20a_gr_av_to_method, &gr->method)) + return -ENOENT; + + return 0; +} + +#if IS_ENABLED(CONFIG_ARCH_TEGRA_124_SOC) || IS_ENABLED(CONFIG_ARCH_TEGRA_132_SOC) +MODULE_FIRMWARE("nvidia/gk20a/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gk20a/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gk20a/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gk20a/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gk20a/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gk20a/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gk20a/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/gk20a/sw_nonctx.bin"); +#endif + +static int +gk20a_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + + if (nvkm_firmware_load_blob(subdev, "", "fecs_inst", ver, + &gr->fecs.inst) || + nvkm_firmware_load_blob(subdev, "", "fecs_data", ver, + &gr->fecs.data) || + nvkm_firmware_load_blob(subdev, "", "gpccs_inst", ver, + &gr->gpccs.inst) || + nvkm_firmware_load_blob(subdev, "", "gpccs_data", ver, + &gr->gpccs.data)) + return -ENOENT; + + gr->firmware = true; + + return gk20a_gr_load_sw(gr, "", ver); +} + +static const struct gf100_gr_fwif +gk20a_gr_fwif[] = { + { 0, gk20a_gr_load, &gk20a_gr }, + {} +}; + +int +gk20a_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gk20a_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c new file mode 100644 index 0000000000..797b828a94 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c @@ -0,0 +1,446 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include +#include +#include +#include +#include + +#include + +/******************************************************************************* + * PGRAPH register lists + ******************************************************************************/ + +static const struct gf100_gr_init +gm107_gr_init_main_0[] = { + { 0x40880c, 1, 0x04, 0x00000000 }, + { 0x408910, 1, 0x04, 0x00000000 }, + { 0x408984, 1, 0x04, 0x00000000 }, + { 0x41a8a0, 1, 0x04, 0x00000000 }, + { 0x400080, 1, 0x04, 0x003003c2 }, + { 0x400088, 1, 0x04, 0x0001bfe7 }, + { 0x40008c, 1, 0x04, 0x00060000 }, + { 0x400090, 1, 0x04, 0x00000030 }, + { 0x40013c, 1, 0x04, 0x003901f3 }, + { 0x400140, 1, 0x04, 0x00000100 }, + { 0x400144, 1, 0x04, 0x00000000 }, + { 0x400148, 1, 0x04, 0x00000110 }, + { 0x400138, 1, 0x04, 0x00000000 }, + { 0x400130, 2, 0x04, 0x00000000 }, + { 0x400124, 1, 0x04, 0x00000002 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_ds_0[] = { + { 0x405844, 1, 0x04, 0x00ffffff }, + { 0x405850, 1, 0x04, 0x00000000 }, + { 0x405900, 1, 0x04, 0x00000000 }, + { 0x405908, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_scc_0[] = { + { 0x40803c, 1, 0x04, 0x00000010 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_sked_0[] = { + { 0x407010, 1, 0x04, 0x00000000 }, + { 0x407040, 1, 0x04, 0x40440424 }, + { 0x407048, 1, 0x04, 0x0000000a }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_prop_0[] = { + { 0x418408, 1, 0x04, 0x00000000 }, + { 0x4184a0, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_setup_1[] = { + { 0x4188c8, 2, 0x04, 0x00000000 }, + { 0x4188d0, 1, 0x04, 0x00010000 }, + { 0x4188d4, 1, 0x04, 0x00010201 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_zcull_0[] = { + { 0x418910, 1, 0x04, 0x00010001 }, + { 0x418914, 1, 0x04, 0x00000301 }, + { 0x418918, 1, 0x04, 0x00800000 }, + { 0x418930, 2, 0x04, 0x00000000 }, + { 0x418980, 1, 0x04, 0x77777770 }, + { 0x418984, 3, 0x04, 0x77777777 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_gpc_unk_1[] = { + { 0x418d00, 1, 0x04, 0x00000000 }, + { 0x418f00, 1, 0x04, 0x00000400 }, + { 0x418f08, 1, 0x04, 0x00000000 }, + { 0x418e08, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_tpccs_0[] = { + { 0x419dc4, 1, 0x04, 0x00000000 }, + { 0x419dc8, 1, 0x04, 0x00000501 }, + { 0x419dd0, 1, 0x04, 0x00000000 }, + { 0x419dd4, 1, 0x04, 0x00000100 }, + { 0x419dd8, 1, 0x04, 0x00000001 }, + { 0x419ddc, 1, 0x04, 0x00000002 }, + { 0x419de0, 1, 0x04, 0x00000001 }, + { 0x419d0c, 1, 0x04, 0x00000000 }, + { 0x419d10, 1, 0x04, 0x00000014 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_tex_0[] = { + { 0x419ab0, 1, 0x04, 0x00000000 }, + { 0x419ab8, 1, 0x04, 0x000000e7 }, + { 0x419abc, 1, 0x04, 0x00000000 }, + { 0x419acc, 1, 0x04, 0x000000ff }, + { 0x419ac0, 1, 0x04, 0x00000000 }, + { 0x419aa8, 2, 0x04, 0x00000000 }, + { 0x419ad0, 2, 0x04, 0x00000000 }, + { 0x419ae0, 2, 0x04, 0x00000000 }, + { 0x419af0, 4, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_pe_0[] = { + { 0x419900, 1, 0x04, 0x000000ff }, + { 0x41980c, 1, 0x04, 0x00000010 }, + { 0x419844, 1, 0x04, 0x00000000 }, + { 0x419838, 1, 0x04, 0x000000ff }, + { 0x419850, 1, 0x04, 0x00000004 }, + { 0x419854, 2, 0x04, 0x00000000 }, + { 0x419894, 3, 0x04, 0x00100401 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_l1c_0[] = { + { 0x419c98, 1, 0x04, 0x00000000 }, + { 0x419cc0, 2, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_sm_0[] = { + { 0x419e30, 1, 0x04, 0x000000ff }, + { 0x419e00, 1, 0x04, 0x00000000 }, + { 0x419ea0, 1, 0x04, 0x00000000 }, + { 0x419ee4, 1, 0x04, 0x00000000 }, + { 0x419ea4, 1, 0x04, 0x00000100 }, + { 0x419ea8, 1, 0x04, 0x01000000 }, + { 0x419ee8, 1, 0x04, 0x00000091 }, + { 0x419eb4, 1, 0x04, 0x00000000 }, + { 0x419ebc, 2, 0x04, 0x00000000 }, + { 0x419edc, 1, 0x04, 0x000c1810 }, + { 0x419ed8, 1, 0x04, 0x00000000 }, + { 0x419ee0, 1, 0x04, 0x00000000 }, + { 0x419f74, 1, 0x04, 0x00005155 }, + { 0x419f80, 4, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_l1c_1[] = { + { 0x419ccc, 2, 0x04, 0x00000000 }, + { 0x419c80, 1, 0x04, 0x3f006022 }, + { 0x419c88, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_pes_0[] = { + { 0x41be50, 1, 0x04, 0x000000ff }, + { 0x41be04, 1, 0x04, 0x00000000 }, + { 0x41be08, 1, 0x04, 0x00000004 }, + { 0x41be0c, 1, 0x04, 0x00000008 }, + { 0x41be10, 1, 0x04, 0x0e3b8bc7 }, + { 0x41be14, 2, 0x04, 0x00000000 }, + { 0x41be3c, 5, 0x04, 0x00100401 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_wwdx_0[] = { + { 0x41bfd4, 1, 0x04, 0x00800000 }, + { 0x41bfdc, 1, 0x04, 0x00000000 }, + {} +}; + +const struct gf100_gr_init +gm107_gr_init_cbm_0[] = { + { 0x41becc, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_be_0[] = { + { 0x408890, 1, 0x04, 0x000000ff }, + { 0x408850, 1, 0x04, 0x00000004 }, + { 0x408878, 1, 0x04, 0x00c81603 }, + { 0x40887c, 1, 0x04, 0x80543432 }, + { 0x408880, 1, 0x04, 0x0010581e }, + { 0x408884, 1, 0x04, 0x00001205 }, + { 0x408974, 1, 0x04, 0x000000ff }, + { 0x408914, 8, 0x04, 0x00000000 }, + { 0x408950, 1, 0x04, 0x00000000 }, + { 0x408954, 1, 0x04, 0x0000ffff }, + { 0x408958, 1, 0x04, 0x00000034 }, + { 0x40895c, 1, 0x04, 0x8531a003 }, + { 0x408960, 1, 0x04, 0x0561985a }, + { 0x408964, 1, 0x04, 0x04e15c4f }, + { 0x408968, 1, 0x04, 0x02808833 }, + { 0x40896c, 1, 0x04, 0x01f02438 }, + { 0x408970, 1, 0x04, 0x00012c00 }, + { 0x408988, 1, 0x04, 0x08040201 }, + { 0x40898c, 1, 0x04, 0x80402010 }, + {} +}; + +static const struct gf100_gr_init +gm107_gr_init_sm_1[] = { + { 0x419e5c, 1, 0x04, 0x00000000 }, + { 0x419e58, 1, 0x04, 0x00000000 }, + {} +}; + +static const struct gf100_gr_pack +gm107_gr_pack_mmio[] = { + { gm107_gr_init_main_0 }, + { gk110_gr_init_fe_0 }, + { gf100_gr_init_pri_0 }, + { gf100_gr_init_rstr2d_0 }, + { gf100_gr_init_pd_0 }, + { gm107_gr_init_ds_0 }, + { gm107_gr_init_scc_0 }, + { gm107_gr_init_sked_0 }, + { gk110_gr_init_cwd_0 }, + { gm107_gr_init_prop_0 }, + { gk208_gr_init_gpc_unk_0 }, + { gf100_gr_init_setup_0 }, + { gf100_gr_init_crstr_0 }, + { gm107_gr_init_setup_1 }, + { gm107_gr_init_zcull_0 }, + { gf100_gr_init_gpm_0 }, + { gm107_gr_init_gpc_unk_1 }, + { gf100_gr_init_gcc_0 }, + { gk104_gr_init_gpc_unk_2 }, + { gm107_gr_init_tpccs_0 }, + { gm107_gr_init_tex_0 }, + { gm107_gr_init_pe_0 }, + { gm107_gr_init_l1c_0 }, + { gf100_gr_init_mpc_0 }, + { gm107_gr_init_sm_0 }, + { gm107_gr_init_l1c_1 }, + { gm107_gr_init_pes_0 }, + { gm107_gr_init_wwdx_0 }, + { gm107_gr_init_cbm_0 }, + { gm107_gr_init_be_0 }, + { gm107_gr_init_sm_1 }, + {} +}; + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +void +gm107_gr_init_400054(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x400054, 0x2c350f63); +} + +void +gm107_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x00000005); +} + +void +gm107_gr_init_504430(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x430), 0xc0000000); +} + +static void +gm107_gr_init_bios_2(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_bios *bios = device->bios; + struct bit_entry bit_P; + if (!bit_entry(bios, 'P', &bit_P) && + bit_P.version == 2 && bit_P.length >= 0x2c) { + u32 data = nvbios_rd32(bios, bit_P.offset + 0x28); + if (data) { + u8 ver = nvbios_rd08(bios, data + 0x00); + u8 hdr = nvbios_rd08(bios, data + 0x01); + if (ver == 0x20 && hdr >= 8) { + data = nvbios_rd32(bios, data + 0x04); + if (data) { + u32 save = nvkm_rd32(device, 0x619444); + nvbios_init(subdev, data); + nvkm_wr32(device, 0x619444, save); + } + } + } + } +} + +void +gm107_gr_init_bios(struct gf100_gr *gr) +{ + static const struct { + u32 ctrl; + u32 data; + } regs[] = { + { 0x419ed8, 0x419ee0 }, + { 0x419ad0, 0x419ad4 }, + { 0x419ae0, 0x419ae4 }, + { 0x419af0, 0x419af4 }, + { 0x419af8, 0x419afc }, + }; + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_bios *bios = device->bios; + struct nvbios_P0260E infoE; + struct nvbios_P0260X infoX; + int E = -1, X; + u8 ver, hdr; + + while (nvbios_P0260Ep(bios, ++E, &ver, &hdr, &infoE)) { + if (X = -1, E < ARRAY_SIZE(regs)) { + nvkm_wr32(device, regs[E].ctrl, infoE.data); + while (nvbios_P0260Xp(bios, ++X, &ver, &hdr, &infoX)) + nvkm_wr32(device, regs[E].data, infoX.data); + } + } +} + +static void +gm107_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fb *fb = device->fb; + + nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8); + nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8); +} + +#include "fuc/hubgm107.fuc5.h" + +static struct gf100_gr_ucode +gm107_gr_fecs_ucode = { + .code.data = gm107_grhub_code, + .code.size = sizeof(gm107_grhub_code), + .data.data = gm107_grhub_data, + .data.size = sizeof(gm107_grhub_data), +}; + +#include "fuc/gpcgm107.fuc5.h" + +static struct gf100_gr_ucode +gm107_gr_gpccs_ucode = { + .code.data = gm107_grgpc_code, + .code.size = sizeof(gm107_grgpc_code), + .data.data = gm107_grgpc_data, + .data.size = sizeof(gm107_grgpc_data), +}; + +static const struct gf100_gr_func +gm107_gr = { + .oneinit_tiles = gf100_gr_oneinit_tiles, + .oneinit_sm_id = gf100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm107_gr_init_gpc_mmu, + .init_bios = gm107_gr_init_bios, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .init_bios_2 = gm107_gr_init_bios_2, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gm107_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gm107_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .mmio = gm107_gr_pack_mmio, + .fecs.ucode = &gm107_gr_fecs_ucode, + .fecs.reset = gf100_gr_fecs_reset, + .gpccs.ucode = &gm107_gr_gpccs_ucode, + .rops = gf100_gr_rops, + .ppc_nr = 2, + .grctx = &gm107_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_A, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_A }, + {} + } +}; + +static const struct gf100_gr_fwif +gm107_gr_fwif[] = { + { -1, gf100_gr_load, &gm107_gr }, + { -1, gf100_gr_nofw, &gm107_gr }, + {} +}; + +int +gm107_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gm107_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c new file mode 100644 index 0000000000..b5210b31c1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c @@ -0,0 +1,296 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include +#include + +#include + +#include + +int +gm200_gr_nofw(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + nvkm_warn(&gr->base.engine.subdev, "firmware unavailable\n"); + return -ENODEV; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static void +gm200_gr_acr_bld_patch(struct nvkm_acr *acr, u32 bld, s64 adjust) +{ + struct flcn_bl_dmem_desc_v1 hdr; + nvkm_robj(acr->wpr, bld, &hdr, sizeof(hdr)); + hdr.code_dma_base = hdr.code_dma_base + adjust; + hdr.data_dma_base = hdr.data_dma_base + adjust; + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); + flcn_bl_dmem_desc_v1_dump(&acr->subdev, &hdr); +} + +static void +gm200_gr_acr_bld_write(struct nvkm_acr *acr, u32 bld, + struct nvkm_acr_lsfw *lsfw) +{ + const u64 base = lsfw->offset.img + lsfw->app_start_offset; + const u64 code = base + lsfw->app_resident_code_offset; + const u64 data = base + lsfw->app_resident_data_offset; + const struct flcn_bl_dmem_desc_v1 hdr = { + .ctx_dma = FALCON_DMAIDX_UCODE, + .code_dma_base = code, + .non_sec_code_off = lsfw->app_resident_code_offset, + .non_sec_code_size = lsfw->app_resident_code_size, + .code_entry_point = lsfw->app_imem_entry, + .data_dma_base = data, + .data_size = lsfw->app_resident_data_size, + }; + + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); +} + +const struct nvkm_acr_lsf_func +gm200_gr_gpccs_acr = { + .flags = NVKM_ACR_LSF_FORCE_PRIV_LOAD, + .bld_size = sizeof(struct flcn_bl_dmem_desc_v1), + .bld_write = gm200_gr_acr_bld_write, + .bld_patch = gm200_gr_acr_bld_patch, +}; + +const struct nvkm_acr_lsf_func +gm200_gr_fecs_acr = { + .bld_size = sizeof(struct flcn_bl_dmem_desc_v1), + .bld_write = gm200_gr_acr_bld_write, + .bld_patch = gm200_gr_acr_bld_patch, +}; + +int +gm200_gr_rops(struct gf100_gr *gr) +{ + return nvkm_rd32(gr->base.engine.subdev.device, 0x12006c); +} + +void +gm200_gr_init_ds_hww_esr_2(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x405848, 0xc0000000); + nvkm_mask(device, 0x40584c, 0x00000001, 0x00000001); +} + +void +gm200_gr_init_num_active_ltcs(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); + nvkm_wr32(device, GPC_BCAST(0x033c), nvkm_rd32(device, 0x100804)); +} + +void +gm200_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0xf0001fff); + nvkm_wr32(device, 0x418890, 0x00000000); + nvkm_wr32(device, 0x418894, 0x00000000); + + nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); + nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); + nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); +} + +static void +gm200_gr_init_rop_active_fbps(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 fbp_count = nvkm_rd32(device, 0x12006c); + nvkm_mask(device, 0x408850, 0x0000000f, fbp_count); /* zrop */ + nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */ +} + +static u8 +gm200_gr_tile_map_6_24[] = { + 0, 1, 2, 3, 4, 5, 3, 4, 5, 0, 1, 2, 0, 1, 2, 3, 4, 5, 3, 4, 5, 0, 1, 2, +}; + +static u8 +gm200_gr_tile_map_4_16[] = { + 0, 1, 2, 3, 2, 3, 0, 1, 3, 0, 1, 2, 1, 2, 3, 0, +}; + +static u8 +gm200_gr_tile_map_2_8[] = { + 0, 1, 1, 0, 0, 1, 1, 0, +}; + +int +gm200_gr_oneinit_sm_id(struct gf100_gr *gr) +{ + /*XXX: There's a different algorithm here I've not yet figured out. */ + return gf100_gr_oneinit_sm_id(gr); +} + +void +gm200_gr_oneinit_tiles(struct gf100_gr *gr) +{ + /*XXX: Not sure what this is about. The algorithm from NVGPU + * seems to work for all boards I tried from earlier (and + * later) GPUs except in these specific configurations. + * + * Let's just hardcode them for now. + */ + if (gr->gpc_nr == 2 && gr->tpc_total == 8) { + memcpy(gr->tile, gm200_gr_tile_map_2_8, gr->tpc_total); + gr->screen_tile_row_offset = 1; + } else + if (gr->gpc_nr == 4 && gr->tpc_total == 16) { + memcpy(gr->tile, gm200_gr_tile_map_4_16, gr->tpc_total); + gr->screen_tile_row_offset = 4; + } else + if (gr->gpc_nr == 6 && gr->tpc_total == 24) { + memcpy(gr->tile, gm200_gr_tile_map_6_24, gr->tpc_total); + gr->screen_tile_row_offset = 5; + } else { + gf100_gr_oneinit_tiles(gr); + } +} + +static const struct gf100_gr_func +gm200_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_bios = gm107_gr_init_bios, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gm200_gr_init_rop_active_fbps, + .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gm107_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_400054 = gm107_gr_init_400054, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .tpc_nr = 4, + .ppc_nr = 2, + .grctx = &gm200_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_B, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_B }, + {} + } +}; + +int +gm200_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + int ret; + + ret = nvkm_acr_lsfw_load_bl_inst_data_sig(&gr->base.engine.subdev, + &gr->fecs.falcon, + NVKM_ACR_LSF_FECS, + "gr/fecs_", ver, fwif->fecs); + if (ret) + return ret; + + ret = nvkm_acr_lsfw_load_bl_inst_data_sig(&gr->base.engine.subdev, + &gr->gpccs.falcon, + NVKM_ACR_LSF_GPCCS, + "gr/gpccs_", ver, + fwif->gpccs); + if (ret) + return ret; + + gr->firmware = true; + + return gk20a_gr_load_sw(gr, "gr/", ver); +} + +MODULE_FIRMWARE("nvidia/gm200/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gm200/gr/sw_method_init.bin"); + +MODULE_FIRMWARE("nvidia/gm204/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gm204/gr/sw_method_init.bin"); + +MODULE_FIRMWARE("nvidia/gm206/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gm206/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gm200_gr_fwif[] = { + { 0, gm200_gr_load, &gm200_gr, &gm200_gr_fecs_acr, &gm200_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gm200_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gm200_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c new file mode 100644 index 0000000000..458cd1a00d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c @@ -0,0 +1,188 @@ +/* + * Copyright (c) 2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include +#include +#include + +#include + +#include + +void +gm20b_gr_acr_bld_patch(struct nvkm_acr *acr, u32 bld, s64 adjust) +{ + struct flcn_bl_dmem_desc hdr; + u64 addr; + + nvkm_robj(acr->wpr, bld, &hdr, sizeof(hdr)); + addr = ((u64)hdr.code_dma_base1 << 40 | hdr.code_dma_base << 8); + hdr.code_dma_base = lower_32_bits((addr + adjust) >> 8); + hdr.code_dma_base1 = upper_32_bits((addr + adjust) >> 8); + addr = ((u64)hdr.data_dma_base1 << 40 | hdr.data_dma_base << 8); + hdr.data_dma_base = lower_32_bits((addr + adjust) >> 8); + hdr.data_dma_base1 = upper_32_bits((addr + adjust) >> 8); + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); + + flcn_bl_dmem_desc_dump(&acr->subdev, &hdr); +} + +void +gm20b_gr_acr_bld_write(struct nvkm_acr *acr, u32 bld, + struct nvkm_acr_lsfw *lsfw) +{ + const u64 base = lsfw->offset.img + lsfw->app_start_offset; + const u64 code = (base + lsfw->app_resident_code_offset) >> 8; + const u64 data = (base + lsfw->app_resident_data_offset) >> 8; + const struct flcn_bl_dmem_desc hdr = { + .ctx_dma = FALCON_DMAIDX_UCODE, + .code_dma_base = lower_32_bits(code), + .non_sec_code_off = lsfw->app_resident_code_offset, + .non_sec_code_size = lsfw->app_resident_code_size, + .code_entry_point = lsfw->app_imem_entry, + .data_dma_base = lower_32_bits(data), + .data_size = lsfw->app_resident_data_size, + .code_dma_base1 = upper_32_bits(code), + .data_dma_base1 = upper_32_bits(data), + }; + + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); +} + +const struct nvkm_acr_lsf_func +gm20b_gr_fecs_acr = { + .bld_size = sizeof(struct flcn_bl_dmem_desc), + .bld_write = gm20b_gr_acr_bld_write, + .bld_patch = gm20b_gr_acr_bld_patch, +}; + +static void +gm20b_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 val; + + /* Bypass MMU check for non-secure boot */ + if (!device->acr) { + nvkm_wr32(device, 0x100ce4, 0xffffffff); + + if (nvkm_rd32(device, 0x100ce4) != 0xffffffff) + nvdev_warn(device, + "cannot bypass secure boot - expect failure soon!\n"); + } + + val = nvkm_rd32(device, 0x100c80); + val &= 0xf000187f; + nvkm_wr32(device, 0x418880, val); + nvkm_wr32(device, 0x418890, 0); + nvkm_wr32(device, 0x418894, 0); + + nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); + nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); + nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); + + nvkm_wr32(device, 0x4188ac, nvkm_rd32(device, 0x100800)); +} + +static void +gm20b_gr_set_hww_esr_report_mask(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x419e44, 0xdffffe); + nvkm_wr32(device, 0x419e4c, 0x5); +} + +static const struct gf100_gr_func +gm20b_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gk20a_gr_init, + .init_zcull = gf117_gr_init_zcull, + .init_gpc_mmu = gm20b_gr_init_gpc_mmu, + .init_rop_active_fbps = gk104_gr_init_rop_active_fbps, + .trap_mp = gf100_gr_trap_mp, + .set_hww_esr_report_mask = gm20b_gr_set_hww_esr_report_mask, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .ppc_nr = 1, + .grctx = &gm20b_grctx, + .zbc = &gf100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_B, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_B }, + {} + } +}; + +static int +gm20b_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + int ret; + + ret = nvkm_acr_lsfw_load_bl_inst_data_sig(subdev, &gr->fecs.falcon, + NVKM_ACR_LSF_FECS, + "gr/fecs_", ver, fwif->fecs); + if (ret) + return ret; + + + if (nvkm_firmware_load_blob(subdev, "gr/", "gpccs_inst", ver, + &gr->gpccs.inst) || + nvkm_firmware_load_blob(subdev, "gr/", "gpccs_data", ver, + &gr->gpccs.data)) + return -ENOENT; + + gr->firmware = true; + + return gk20a_gr_load_sw(gr, "gr/", ver); +} + +#if IS_ENABLED(CONFIG_ARCH_TEGRA_210_SOC) +MODULE_FIRMWARE("nvidia/gm20b/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gm20b/gr/sw_method_init.bin"); +#endif + +static const struct gf100_gr_fwif +gm20b_gr_fwif[] = { + { 0, gm20b_gr_load, &gm20b_gr, &gm20b_gr_fecs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gm20b_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gm20b_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp100.c new file mode 100644 index 0000000000..851e743d2c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp100.c @@ -0,0 +1,165 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ +void +gp100_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const int znum = zbc - 1; + const u32 zoff = znum * 4; + + if (gr->zbc_color[zbc].format) { + nvkm_wr32(device, 0x418010 + zoff, gr->zbc_color[zbc].ds[0]); + nvkm_wr32(device, 0x41804c + zoff, gr->zbc_color[zbc].ds[1]); + nvkm_wr32(device, 0x418088 + zoff, gr->zbc_color[zbc].ds[2]); + nvkm_wr32(device, 0x4180c4 + zoff, gr->zbc_color[zbc].ds[3]); + } + + nvkm_mask(device, 0x418100 + ((znum / 4) * 4), + 0x0000007f << ((znum % 4) * 7), + gr->zbc_color[zbc].format << ((znum % 4) * 7)); +} + +void +gp100_gr_zbc_clear_depth(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const int znum = zbc - 1; + const u32 zoff = znum * 4; + + if (gr->zbc_depth[zbc].format) + nvkm_wr32(device, 0x418110 + zoff, gr->zbc_depth[zbc].ds); + nvkm_mask(device, 0x41814c + ((znum / 4) * 4), + 0x0000007f << ((znum % 4) * 7), + gr->zbc_depth[zbc].format << ((znum % 4) * 7)); +} + +const struct gf100_gr_func_zbc +gp100_gr_zbc = { + .clear_color = gp100_gr_zbc_clear_color, + .clear_depth = gp100_gr_zbc_clear_depth, +}; + +void +gp100_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x00000105); +} + +static void +gp100_gr_init_419c9c(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419c9c, 0x00010000, 0x00010000); + nvkm_mask(device, 0x419c9c, 0x00020000, 0x00020000); +} + +void +gp100_gr_init_fecs_exceptions(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x409c24, 0x000e0002); +} + +void +gp100_gr_init_rop_active_fbps(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + /*XXX: otherwise identical to gm200 aside from mask.. do everywhere? */ + const u32 fbp_count = nvkm_rd32(device, 0x12006c) & 0x0000000f; + nvkm_mask(device, 0x408850, 0x0000000f, fbp_count); /* zrop */ + nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */ +} + +static const struct gf100_gr_func +gp100_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_419c9c = gp100_gr_init_419c9c, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gp100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 6, + .tpc_nr = 5, + .ppc_nr = 2, + .grctx = &gp100_grctx, + .zbc = &gp100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, PASCAL_A, &gf100_fermi }, + { -1, -1, PASCAL_COMPUTE_A }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/gp100/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp100/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gp100_gr_fwif[] = { + { 0, gm200_gr_load, &gp100_gr, &gm200_gr_fecs_acr, &gm200_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp100_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp100_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp102.c new file mode 100644 index 0000000000..0e223b7b5f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp102.c @@ -0,0 +1,161 @@ +/* + * Copyright 2016 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +void +gp102_gr_zbc_clear_stencil(struct gf100_gr *gr, int zbc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const int znum = zbc - 1; + const u32 zoff = znum * 4; + + if (gr->zbc_stencil[zbc].format) + nvkm_wr32(device, 0x41815c + zoff, gr->zbc_stencil[zbc].ds); + nvkm_mask(device, 0x418198 + ((znum / 4) * 4), + 0x0000007f << ((znum % 4) * 7), + gr->zbc_stencil[zbc].format << ((znum % 4) * 7)); +} + +int +gp102_gr_zbc_stencil_get(struct gf100_gr *gr, int format, + const u32 ds, const u32 l2) +{ + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; + int zbc = -ENOSPC, i; + + for (i = ltc->zbc_depth_min; i <= ltc->zbc_depth_max; i++) { + if (gr->zbc_stencil[i].format) { + if (gr->zbc_stencil[i].format != format) + continue; + if (gr->zbc_stencil[i].ds != ds) + continue; + if (gr->zbc_stencil[i].l2 != l2) { + WARN_ON(1); + return -EINVAL; + } + return i; + } else { + zbc = (zbc < 0) ? i : zbc; + } + } + + if (zbc < 0) + return zbc; + + gr->zbc_stencil[zbc].format = format; + gr->zbc_stencil[zbc].ds = ds; + gr->zbc_stencil[zbc].l2 = l2; + nvkm_ltc_zbc_stencil_get(ltc, zbc, l2); + gr->func->zbc->clear_stencil(gr, zbc); + return zbc; +} + +const struct gf100_gr_func_zbc +gp102_gr_zbc = { + .clear_color = gp100_gr_zbc_clear_color, + .clear_depth = gp100_gr_zbc_clear_depth, + .stencil_get = gp102_gr_zbc_stencil_get, + .clear_stencil = gp102_gr_zbc_clear_stencil, +}; + +void +gp102_gr_init_swdx_pes_mask(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 mask = 0, data, gpc; + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + data = nvkm_rd32(device, GPC_UNIT(gpc, 0x0c50)) & 0x0000000f; + mask |= data << (gpc * 4); + } + + nvkm_wr32(device, 0x4181d0, mask); +} + +static const struct gf100_gr_func +gp102_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gp100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 6, + .tpc_nr = 5, + .ppc_nr = 3, + .grctx = &gp102_grctx, + .zbc = &gp102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, PASCAL_B, &gf100_fermi }, + { -1, -1, PASCAL_COMPUTE_B }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/gp102/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp102/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gp102_gr_fwif[] = { + { 0, gm200_gr_load, &gp102_gr, &gm200_gr_fecs_acr, &gm200_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp102_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp102_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp104.c new file mode 100644 index 0000000000..6802cb9b19 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp104.c @@ -0,0 +1,102 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +static const struct gf100_gr_func +gp104_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gp100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 6, + .tpc_nr = 5, + .ppc_nr = 3, + .grctx = &gp104_grctx, + .zbc = &gp102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, PASCAL_B, &gf100_fermi }, + { -1, -1, PASCAL_COMPUTE_B }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/gp104/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp104/gr/sw_method_init.bin"); + +MODULE_FIRMWARE("nvidia/gp106/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp106/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gp104_gr_fwif[] = { + { 0, gm200_gr_load, &gp104_gr, &gm200_gr_fecs_acr, &gm200_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp104_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp104_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp107.c new file mode 100644 index 0000000000..cc2bb0d0a9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp107.c @@ -0,0 +1,91 @@ +/* + * Copyright 2017 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +const struct gf100_gr_func +gp107_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gp100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 2, + .tpc_nr = 3, + .ppc_nr = 1, + .grctx = &gp107_grctx, + .zbc = &gp102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, PASCAL_B, &gf100_fermi }, + { -1, -1, PASCAL_COMPUTE_B }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/gp107/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp107/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gp107_gr_fwif[] = { + { 0, gm200_gr_load, &gp107_gr, &gm200_gr_fecs_acr, &gm200_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp107_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp107_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp108.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp108.c new file mode 100644 index 0000000000..311f703439 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp108.c @@ -0,0 +1,98 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" + +#include + +#include + +void +gp108_gr_acr_bld_patch(struct nvkm_acr *acr, u32 bld, s64 adjust) +{ + struct flcn_bl_dmem_desc_v2 hdr; + nvkm_robj(acr->wpr, bld, &hdr, sizeof(hdr)); + hdr.code_dma_base = hdr.code_dma_base + adjust; + hdr.data_dma_base = hdr.data_dma_base + adjust; + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); + flcn_bl_dmem_desc_v2_dump(&acr->subdev, &hdr); +} + +void +gp108_gr_acr_bld_write(struct nvkm_acr *acr, u32 bld, + struct nvkm_acr_lsfw *lsfw) +{ + const u64 base = lsfw->offset.img + lsfw->app_start_offset; + const u64 code = base + lsfw->app_resident_code_offset; + const u64 data = base + lsfw->app_resident_data_offset; + const struct flcn_bl_dmem_desc_v2 hdr = { + .ctx_dma = FALCON_DMAIDX_UCODE, + .code_dma_base = code, + .non_sec_code_off = lsfw->app_resident_code_offset, + .non_sec_code_size = lsfw->app_resident_code_size, + .code_entry_point = lsfw->app_imem_entry, + .data_dma_base = data, + .data_size = lsfw->app_resident_data_size, + }; + + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); +} + +const struct nvkm_acr_lsf_func +gp108_gr_gpccs_acr = { + .flags = NVKM_ACR_LSF_FORCE_PRIV_LOAD, + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp108_gr_acr_bld_write, + .bld_patch = gp108_gr_acr_bld_patch, +}; + +const struct nvkm_acr_lsf_func +gp108_gr_fecs_acr = { + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp108_gr_acr_bld_write, + .bld_patch = gp108_gr_acr_bld_patch, +}; + +MODULE_FIRMWARE("nvidia/gp108/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp108/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gp108_gr_fwif[] = { + { 0, gm200_gr_load, &gp107_gr, &gp108_gr_fecs_acr, &gp108_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp108_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp108_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp10b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp10b.c new file mode 100644 index 0000000000..5008881ca0 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gp10b.c @@ -0,0 +1,103 @@ +/* + * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "gf100.h" +#include "ctxgf100.h" + +#include + +#include + +#include + +static const struct nvkm_acr_lsf_func +gp10b_gr_gpccs_acr = { + .flags = NVKM_ACR_LSF_FORCE_PRIV_LOAD, + .bld_size = sizeof(struct flcn_bl_dmem_desc), + .bld_write = gm20b_gr_acr_bld_write, + .bld_patch = gm20b_gr_acr_bld_patch, +}; + +static const struct gf100_gr_func +gp10b_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gm200_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_419cc0 = gf100_gr_init_419cc0, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, + .init_504430 = gm107_gr_init_504430, + .init_shader_exceptions = gp100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .trap_mp = gf100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 1, + .tpc_nr = 2, + .ppc_nr = 1, + .grctx = &gp100_grctx, + .zbc = &gp100_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, PASCAL_A, &gf100_fermi }, + { -1, -1, PASCAL_COMPUTE_A }, + {} + } +}; + +#if IS_ENABLED(CONFIG_ARCH_TEGRA_186_SOC) +MODULE_FIRMWARE("nvidia/gp10b/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gp10b/gr/sw_method_init.bin"); +#endif + +static const struct gf100_gr_fwif +gp10b_gr_fwif[] = { + { 0, gm200_gr_load, &gp10b_gr, &gm20b_gr_fecs_acr, &gp10b_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gp10b_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gp10b_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c new file mode 100644 index 0000000000..1dfc65d45b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c @@ -0,0 +1,49 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include + +static const struct nvkm_gr_func +gt200_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT200_TESLA, &nv50_gr_object }, + {} + } +}; + +int +gt200_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(>200_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c new file mode 100644 index 0000000000..fcb5ead345 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include + +static const struct nvkm_gr_func +gt215_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT214_TESLA, &nv50_gr_object }, + { -1, -1, GT214_COMPUTE, &nv50_gr_object }, + {} + } +}; + +int +gt215_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(>215_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gv100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gv100.c new file mode 100644 index 0000000000..7f7404a761 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gv100.c @@ -0,0 +1,342 @@ +/* + * Copyright 2018 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +static void +gv100_gr_trap_sm(struct gf100_gr *gr, int gpc, int tpc, int sm) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x730 + (sm * 0x80))); + u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x734 + (sm * 0x80))); + const struct nvkm_enum *warp; + char glob[128]; + + nvkm_snprintbf(glob, sizeof(glob), gf100_mp_global_error, gerr); + warp = nvkm_enum_find(gf100_mp_warp_error, werr & 0xffff); + + nvkm_error(subdev, "GPC%i/TPC%i/SM%d trap: " + "global %08x [%s] warp %04x [%s]\n", + gpc, tpc, sm, gerr, glob, werr, warp ? warp->name : ""); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x730 + sm * 0x80), 0x00000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x734 + sm * 0x80), gerr); +} + +void +gv100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc) +{ + gv100_gr_trap_sm(gr, gpc, tpc, 0); + gv100_gr_trap_sm(gr, gpc, tpc, 1); +} + +void +gv100_gr_init_4188a4(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_mask(device, 0x4188a4, 0x03000000, 0x03000000); +} + +void +gv100_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int sm; + for (sm = 0; sm < 0x100; sm += 0x80) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x610), 0x00000001); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x72c + sm), 0x00000004); + } +} + +void +gv100_gr_init_504430(struct gf100_gr *gr, int gpc, int tpc) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x430), 0x403f0000); +} + +void +gv100_gr_init_419bd8(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x419bd8, 0x00000700, 0x00000000); +} + +u32 +gv100_gr_nonpes_aware_tpc(struct gf100_gr *gr, u32 gpc, u32 tpc) +{ + u32 pes, temp, tpc_new = 0; + + for (pes = 0; pes < gr->ppc_nr[gpc]; pes++) { + if (gr->ppc_tpc_mask[gpc][pes] & BIT(tpc)) + break; + + tpc_new += gr->ppc_tpc_nr[gpc][pes]; + } + + temp = (BIT(tpc) - 1) & gr->ppc_tpc_mask[gpc][pes]; + temp = hweight32(temp); + return tpc_new + temp; +} + +static int +gv100_gr_scg_estimate_perf(struct gf100_gr *gr, unsigned long *gpc_tpc_mask, + u32 disable_gpc, u32 disable_tpc, int *perf) +{ + const u32 scale_factor = 512UL; /* Use fx23.9 */ + const u32 pix_scale = 1024*1024UL; /* Pix perf in [29:20] */ + const u32 world_scale = 1024UL; /* World performance in [19:10] */ + const u32 tpc_scale = 1; /* TPC balancing in [9:0] */ + u32 scg_num_pes = 0; + u32 min_scg_gpc_pix_perf = scale_factor; /* Init perf as maximum */ + u32 average_tpcs = 0; /* Average of # of TPCs per GPC */ + u32 deviation; /* absolute diff between TPC# and average_tpcs, averaged across GPCs */ + u32 norm_tpc_deviation; /* deviation/max_tpc_per_gpc */ + u32 tpc_balance; + u32 scg_gpc_pix_perf; + u32 scg_world_perf; + u32 gpc; + u32 pes; + int diff; + bool tpc_removed_gpc = false; + bool tpc_removed_pes = false; + u32 max_tpc_gpc = 0; + u32 num_tpc_mask; + u32 *num_tpc_gpc; + int ret = -EINVAL; + + if (!(num_tpc_gpc = kcalloc(gr->gpc_nr, sizeof(*num_tpc_gpc), GFP_KERNEL))) + return -ENOMEM; + + /* Calculate pix-perf-reduction-rate per GPC and find bottleneck TPC */ + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + num_tpc_mask = gpc_tpc_mask[gpc]; + + if ((gpc == disable_gpc) && num_tpc_mask & BIT(disable_tpc)) { + /* Safety check if a TPC is removed twice */ + if (WARN_ON(tpc_removed_gpc)) + goto done; + + /* Remove logical TPC from set */ + num_tpc_mask &= ~BIT(disable_tpc); + tpc_removed_gpc = true; + } + + /* track balancing of tpcs across gpcs */ + num_tpc_gpc[gpc] = hweight32(num_tpc_mask); + average_tpcs += num_tpc_gpc[gpc]; + + /* save the maximum numer of gpcs */ + max_tpc_gpc = num_tpc_gpc[gpc] > max_tpc_gpc ? num_tpc_gpc[gpc] : max_tpc_gpc; + + /* + * Calculate ratio between TPC count and post-FS and post-SCG + * + * ratio represents relative throughput of the GPC + */ + scg_gpc_pix_perf = scale_factor * num_tpc_gpc[gpc] / gr->tpc_nr[gpc]; + if (min_scg_gpc_pix_perf > scg_gpc_pix_perf) + min_scg_gpc_pix_perf = scg_gpc_pix_perf; + + /* Calculate # of surviving PES */ + for (pes = 0; pes < gr->ppc_nr[gpc]; pes++) { + /* Count the number of TPC on the set */ + num_tpc_mask = gr->ppc_tpc_mask[gpc][pes] & gpc_tpc_mask[gpc]; + + if ((gpc == disable_gpc) && (num_tpc_mask & BIT(disable_tpc))) { + if (WARN_ON(tpc_removed_pes)) + goto done; + + num_tpc_mask &= ~BIT(disable_tpc); + tpc_removed_pes = true; + } + + if (hweight32(num_tpc_mask)) + scg_num_pes++; + } + } + + if (WARN_ON(!tpc_removed_gpc || !tpc_removed_pes)) + goto done; + + if (max_tpc_gpc == 0) { + *perf = 0; + goto done_ok; + } + + /* Now calculate perf */ + scg_world_perf = (scale_factor * scg_num_pes) / gr->ppc_total; + deviation = 0; + average_tpcs = scale_factor * average_tpcs / gr->gpc_nr; + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + diff = average_tpcs - scale_factor * num_tpc_gpc[gpc]; + if (diff < 0) + diff = -diff; + + deviation += diff; + } + + deviation /= gr->gpc_nr; + + norm_tpc_deviation = deviation / max_tpc_gpc; + + tpc_balance = scale_factor - norm_tpc_deviation; + + if ((tpc_balance > scale_factor) || + (scg_world_perf > scale_factor) || + (min_scg_gpc_pix_perf > scale_factor) || + (norm_tpc_deviation > scale_factor)) { + WARN_ON(1); + goto done; + } + + *perf = (pix_scale * min_scg_gpc_pix_perf) + + (world_scale * scg_world_perf) + + (tpc_scale * tpc_balance); +done_ok: + ret = 0; +done: + kfree(num_tpc_gpc); + return ret; +} + +int +gv100_gr_oneinit_sm_id(struct gf100_gr *gr) +{ + unsigned long *gpc_tpc_mask; + u32 *tpc_table, *gpc_table; + u32 gpc, tpc, pes, gtpc; + int perf, maxperf, ret = 0; + + gpc_tpc_mask = kcalloc(gr->gpc_nr, sizeof(*gpc_tpc_mask), GFP_KERNEL); + gpc_table = kcalloc(gr->tpc_total, sizeof(*gpc_table), GFP_KERNEL); + tpc_table = kcalloc(gr->tpc_total, sizeof(*tpc_table), GFP_KERNEL); + if (!gpc_table || !tpc_table || !gpc_tpc_mask) { + ret = -ENOMEM; + goto done; + } + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (pes = 0; pes < gr->ppc_nr[gpc]; pes++) + gpc_tpc_mask[gpc] |= gr->ppc_tpc_mask[gpc][pes]; + } + + for (gtpc = 0; gtpc < gr->tpc_total; gtpc++) { + for (maxperf = -1, gpc = 0; gpc < gr->gpc_nr; gpc++) { + for_each_set_bit(tpc, &gpc_tpc_mask[gpc], gr->tpc_nr[gpc]) { + ret = gv100_gr_scg_estimate_perf(gr, gpc_tpc_mask, gpc, tpc, &perf); + if (ret) + goto done; + + /* nvgpu does ">=" here, but this gets us RM's numbers. */ + if (perf > maxperf) { + maxperf = perf; + gpc_table[gtpc] = gpc; + tpc_table[gtpc] = tpc; + } + } + } + + gpc_tpc_mask[gpc_table[gtpc]] &= ~BIT(tpc_table[gtpc]); + } + + /*TODO: build table for sm_per_tpc != 1, don't use yet, but might need later? */ + for (gtpc = 0; gtpc < gr->tpc_total; gtpc++) { + gr->sm[gtpc].gpc = gpc_table[gtpc]; + gr->sm[gtpc].tpc = tpc_table[gtpc]; + gr->sm_nr++; + } + +done: + kfree(gpc_table); + kfree(tpc_table); + kfree(gpc_tpc_mask); + return ret; +} + +static const struct gf100_gr_func +gv100_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gv100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_419bd8 = gv100_gr_init_419bd8, + .init_gpc_mmu = gm200_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = gf117_gr_init_zcull, + .init_num_active_ltcs = gm200_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fecs_exceptions = gp100_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_504430 = gv100_gr_init_504430, + .init_shader_exceptions = gv100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_4188a4 = gv100_gr_init_4188a4, + .trap_mp = gv100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 6, + .tpc_nr = 7, + .ppc_nr = 3, + .grctx = &gv100_grctx, + .zbc = &gp102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, VOLTA_A, &gf100_fermi }, + { -1, -1, VOLTA_COMPUTE_A }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/gv100/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/gv100/gr/sw_method_init.bin"); + +static const struct gf100_gr_fwif +gv100_gr_fwif[] = { + { 0, gm200_gr_load, &gv100_gr, &gp108_gr_fecs_acr, &gp108_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +gv100_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(gv100_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c new file mode 100644 index 0000000000..cf782b64f6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c @@ -0,0 +1,48 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include + +static const struct nvkm_gr_func +mcp79_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT200_TESLA, &nv50_gr_object }, + {} + } +}; + +int +mcp79_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&mcp79_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c new file mode 100644 index 0000000000..6f90a63954 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include + +static const struct nvkm_gr_func +mcp89_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + { -1, -1, GT214_COMPUTE, &nv50_gr_object }, + { -1, -1, GT21A_TESLA, &nv50_gr_object }, + {} + } +}; + +int +mcp89_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&mcp89_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c new file mode 100644 index 0000000000..ca822f07b6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c @@ -0,0 +1,1426 @@ +/* + * Copyright 2007 Stephane Marchesin + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include "regs.h" + +#include +#include +#include +#include +#include +#include + +static u32 +nv04_gr_ctx_regs[] = { + 0x0040053c, + 0x00400544, + 0x00400540, + 0x00400548, + NV04_PGRAPH_CTX_SWITCH1, + NV04_PGRAPH_CTX_SWITCH2, + NV04_PGRAPH_CTX_SWITCH3, + NV04_PGRAPH_CTX_SWITCH4, + NV04_PGRAPH_CTX_CACHE1, + NV04_PGRAPH_CTX_CACHE2, + NV04_PGRAPH_CTX_CACHE3, + NV04_PGRAPH_CTX_CACHE4, + 0x00400184, + 0x004001a4, + 0x004001c4, + 0x004001e4, + 0x00400188, + 0x004001a8, + 0x004001c8, + 0x004001e8, + 0x0040018c, + 0x004001ac, + 0x004001cc, + 0x004001ec, + 0x00400190, + 0x004001b0, + 0x004001d0, + 0x004001f0, + 0x00400194, + 0x004001b4, + 0x004001d4, + 0x004001f4, + 0x00400198, + 0x004001b8, + 0x004001d8, + 0x004001f8, + 0x0040019c, + 0x004001bc, + 0x004001dc, + 0x004001fc, + 0x00400174, + NV04_PGRAPH_DMA_START_0, + NV04_PGRAPH_DMA_START_1, + NV04_PGRAPH_DMA_LENGTH, + NV04_PGRAPH_DMA_MISC, + NV04_PGRAPH_DMA_PITCH, + NV04_PGRAPH_BOFFSET0, + NV04_PGRAPH_BBASE0, + NV04_PGRAPH_BLIMIT0, + NV04_PGRAPH_BOFFSET1, + NV04_PGRAPH_BBASE1, + NV04_PGRAPH_BLIMIT1, + NV04_PGRAPH_BOFFSET2, + NV04_PGRAPH_BBASE2, + NV04_PGRAPH_BLIMIT2, + NV04_PGRAPH_BOFFSET3, + NV04_PGRAPH_BBASE3, + NV04_PGRAPH_BLIMIT3, + NV04_PGRAPH_BOFFSET4, + NV04_PGRAPH_BBASE4, + NV04_PGRAPH_BLIMIT4, + NV04_PGRAPH_BOFFSET5, + NV04_PGRAPH_BBASE5, + NV04_PGRAPH_BLIMIT5, + NV04_PGRAPH_BPITCH0, + NV04_PGRAPH_BPITCH1, + NV04_PGRAPH_BPITCH2, + NV04_PGRAPH_BPITCH3, + NV04_PGRAPH_BPITCH4, + NV04_PGRAPH_SURFACE, + NV04_PGRAPH_STATE, + NV04_PGRAPH_BSWIZZLE2, + NV04_PGRAPH_BSWIZZLE5, + NV04_PGRAPH_BPIXEL, + NV04_PGRAPH_NOTIFY, + NV04_PGRAPH_PATT_COLOR0, + NV04_PGRAPH_PATT_COLOR1, + NV04_PGRAPH_PATT_COLORRAM+0x00, + NV04_PGRAPH_PATT_COLORRAM+0x04, + NV04_PGRAPH_PATT_COLORRAM+0x08, + NV04_PGRAPH_PATT_COLORRAM+0x0c, + NV04_PGRAPH_PATT_COLORRAM+0x10, + NV04_PGRAPH_PATT_COLORRAM+0x14, + NV04_PGRAPH_PATT_COLORRAM+0x18, + NV04_PGRAPH_PATT_COLORRAM+0x1c, + NV04_PGRAPH_PATT_COLORRAM+0x20, + NV04_PGRAPH_PATT_COLORRAM+0x24, + NV04_PGRAPH_PATT_COLORRAM+0x28, + NV04_PGRAPH_PATT_COLORRAM+0x2c, + NV04_PGRAPH_PATT_COLORRAM+0x30, + NV04_PGRAPH_PATT_COLORRAM+0x34, + NV04_PGRAPH_PATT_COLORRAM+0x38, + NV04_PGRAPH_PATT_COLORRAM+0x3c, + NV04_PGRAPH_PATT_COLORRAM+0x40, + NV04_PGRAPH_PATT_COLORRAM+0x44, + NV04_PGRAPH_PATT_COLORRAM+0x48, + NV04_PGRAPH_PATT_COLORRAM+0x4c, + NV04_PGRAPH_PATT_COLORRAM+0x50, + NV04_PGRAPH_PATT_COLORRAM+0x54, + NV04_PGRAPH_PATT_COLORRAM+0x58, + NV04_PGRAPH_PATT_COLORRAM+0x5c, + NV04_PGRAPH_PATT_COLORRAM+0x60, + NV04_PGRAPH_PATT_COLORRAM+0x64, + NV04_PGRAPH_PATT_COLORRAM+0x68, + NV04_PGRAPH_PATT_COLORRAM+0x6c, + NV04_PGRAPH_PATT_COLORRAM+0x70, + NV04_PGRAPH_PATT_COLORRAM+0x74, + NV04_PGRAPH_PATT_COLORRAM+0x78, + NV04_PGRAPH_PATT_COLORRAM+0x7c, + NV04_PGRAPH_PATT_COLORRAM+0x80, + NV04_PGRAPH_PATT_COLORRAM+0x84, + NV04_PGRAPH_PATT_COLORRAM+0x88, + NV04_PGRAPH_PATT_COLORRAM+0x8c, + NV04_PGRAPH_PATT_COLORRAM+0x90, + NV04_PGRAPH_PATT_COLORRAM+0x94, + NV04_PGRAPH_PATT_COLORRAM+0x98, + NV04_PGRAPH_PATT_COLORRAM+0x9c, + NV04_PGRAPH_PATT_COLORRAM+0xa0, + NV04_PGRAPH_PATT_COLORRAM+0xa4, + NV04_PGRAPH_PATT_COLORRAM+0xa8, + NV04_PGRAPH_PATT_COLORRAM+0xac, + NV04_PGRAPH_PATT_COLORRAM+0xb0, + NV04_PGRAPH_PATT_COLORRAM+0xb4, + NV04_PGRAPH_PATT_COLORRAM+0xb8, + NV04_PGRAPH_PATT_COLORRAM+0xbc, + NV04_PGRAPH_PATT_COLORRAM+0xc0, + NV04_PGRAPH_PATT_COLORRAM+0xc4, + NV04_PGRAPH_PATT_COLORRAM+0xc8, + NV04_PGRAPH_PATT_COLORRAM+0xcc, + NV04_PGRAPH_PATT_COLORRAM+0xd0, + NV04_PGRAPH_PATT_COLORRAM+0xd4, + NV04_PGRAPH_PATT_COLORRAM+0xd8, + NV04_PGRAPH_PATT_COLORRAM+0xdc, + NV04_PGRAPH_PATT_COLORRAM+0xe0, + NV04_PGRAPH_PATT_COLORRAM+0xe4, + NV04_PGRAPH_PATT_COLORRAM+0xe8, + NV04_PGRAPH_PATT_COLORRAM+0xec, + NV04_PGRAPH_PATT_COLORRAM+0xf0, + NV04_PGRAPH_PATT_COLORRAM+0xf4, + NV04_PGRAPH_PATT_COLORRAM+0xf8, + NV04_PGRAPH_PATT_COLORRAM+0xfc, + NV04_PGRAPH_PATTERN, + 0x0040080c, + NV04_PGRAPH_PATTERN_SHAPE, + 0x00400600, + NV04_PGRAPH_ROP3, + NV04_PGRAPH_CHROMA, + NV04_PGRAPH_BETA_AND, + NV04_PGRAPH_BETA_PREMULT, + NV04_PGRAPH_CONTROL0, + NV04_PGRAPH_CONTROL1, + NV04_PGRAPH_CONTROL2, + NV04_PGRAPH_BLEND, + NV04_PGRAPH_STORED_FMT, + NV04_PGRAPH_SOURCE_COLOR, + 0x00400560, + 0x00400568, + 0x00400564, + 0x0040056c, + 0x00400400, + 0x00400480, + 0x00400404, + 0x00400484, + 0x00400408, + 0x00400488, + 0x0040040c, + 0x0040048c, + 0x00400410, + 0x00400490, + 0x00400414, + 0x00400494, + 0x00400418, + 0x00400498, + 0x0040041c, + 0x0040049c, + 0x00400420, + 0x004004a0, + 0x00400424, + 0x004004a4, + 0x00400428, + 0x004004a8, + 0x0040042c, + 0x004004ac, + 0x00400430, + 0x004004b0, + 0x00400434, + 0x004004b4, + 0x00400438, + 0x004004b8, + 0x0040043c, + 0x004004bc, + 0x00400440, + 0x004004c0, + 0x00400444, + 0x004004c4, + 0x00400448, + 0x004004c8, + 0x0040044c, + 0x004004cc, + 0x00400450, + 0x004004d0, + 0x00400454, + 0x004004d4, + 0x00400458, + 0x004004d8, + 0x0040045c, + 0x004004dc, + 0x00400460, + 0x004004e0, + 0x00400464, + 0x004004e4, + 0x00400468, + 0x004004e8, + 0x0040046c, + 0x004004ec, + 0x00400470, + 0x004004f0, + 0x00400474, + 0x004004f4, + 0x00400478, + 0x004004f8, + 0x0040047c, + 0x004004fc, + 0x00400534, + 0x00400538, + 0x00400514, + 0x00400518, + 0x0040051c, + 0x00400520, + 0x00400524, + 0x00400528, + 0x0040052c, + 0x00400530, + 0x00400d00, + 0x00400d40, + 0x00400d80, + 0x00400d04, + 0x00400d44, + 0x00400d84, + 0x00400d08, + 0x00400d48, + 0x00400d88, + 0x00400d0c, + 0x00400d4c, + 0x00400d8c, + 0x00400d10, + 0x00400d50, + 0x00400d90, + 0x00400d14, + 0x00400d54, + 0x00400d94, + 0x00400d18, + 0x00400d58, + 0x00400d98, + 0x00400d1c, + 0x00400d5c, + 0x00400d9c, + 0x00400d20, + 0x00400d60, + 0x00400da0, + 0x00400d24, + 0x00400d64, + 0x00400da4, + 0x00400d28, + 0x00400d68, + 0x00400da8, + 0x00400d2c, + 0x00400d6c, + 0x00400dac, + 0x00400d30, + 0x00400d70, + 0x00400db0, + 0x00400d34, + 0x00400d74, + 0x00400db4, + 0x00400d38, + 0x00400d78, + 0x00400db8, + 0x00400d3c, + 0x00400d7c, + 0x00400dbc, + 0x00400590, + 0x00400594, + 0x00400598, + 0x0040059c, + 0x004005a8, + 0x004005ac, + 0x004005b0, + 0x004005b4, + 0x004005c0, + 0x004005c4, + 0x004005c8, + 0x004005cc, + 0x004005d0, + 0x004005d4, + 0x004005d8, + 0x004005dc, + 0x004005e0, + NV04_PGRAPH_PASSTHRU_0, + NV04_PGRAPH_PASSTHRU_1, + NV04_PGRAPH_PASSTHRU_2, + NV04_PGRAPH_DVD_COLORFMT, + NV04_PGRAPH_SCALED_FORMAT, + NV04_PGRAPH_MISC24_0, + NV04_PGRAPH_MISC24_1, + NV04_PGRAPH_MISC24_2, + 0x00400500, + 0x00400504, + NV04_PGRAPH_VALID1, + NV04_PGRAPH_VALID2, + NV04_PGRAPH_DEBUG_3 +}; + +#define nv04_gr(p) container_of((p), struct nv04_gr, base) + +struct nv04_gr { + struct nvkm_gr base; + struct nv04_gr_chan *chan[16]; + spinlock_t lock; +}; + +#define nv04_gr_chan(p) container_of((p), struct nv04_gr_chan, object) + +struct nv04_gr_chan { + struct nvkm_object object; + struct nv04_gr *gr; + int chid; + u32 nv04[ARRAY_SIZE(nv04_gr_ctx_regs)]; +}; + +/******************************************************************************* + * Graphics object classes + ******************************************************************************/ + +/* + * Software methods, why they are needed, and how they all work: + * + * NV04 and NV05 keep most of the state in PGRAPH context itself, but some + * 2d engine settings are kept inside the grobjs themselves. The grobjs are + * 3 words long on both. grobj format on NV04 is: + * + * word 0: + * - bits 0-7: class + * - bit 12: color key active + * - bit 13: clip rect active + * - bit 14: if set, destination surface is swizzled and taken from buffer 5 + * [set by NV04_SWIZZLED_SURFACE], otherwise it's linear and taken + * from buffer 0 [set by NV04_CONTEXT_SURFACES_2D or + * NV03_CONTEXT_SURFACE_DST]. + * - bits 15-17: 2d operation [aka patch config] + * - bit 24: patch valid [enables rendering using this object] + * - bit 25: surf3d valid [for tex_tri and multitex_tri only] + * word 1: + * - bits 0-1: mono format + * - bits 8-13: color format + * - bits 16-31: DMA_NOTIFY instance + * word 2: + * - bits 0-15: DMA_A instance + * - bits 16-31: DMA_B instance + * + * On NV05 it's: + * + * word 0: + * - bits 0-7: class + * - bit 12: color key active + * - bit 13: clip rect active + * - bit 14: if set, destination surface is swizzled and taken from buffer 5 + * [set by NV04_SWIZZLED_SURFACE], otherwise it's linear and taken + * from buffer 0 [set by NV04_CONTEXT_SURFACES_2D or + * NV03_CONTEXT_SURFACE_DST]. + * - bits 15-17: 2d operation [aka patch config] + * - bits 20-22: dither mode + * - bit 24: patch valid [enables rendering using this object] + * - bit 25: surface_dst/surface_color/surf2d/surf3d valid + * - bit 26: surface_src/surface_zeta valid + * - bit 27: pattern valid + * - bit 28: rop valid + * - bit 29: beta1 valid + * - bit 30: beta4 valid + * word 1: + * - bits 0-1: mono format + * - bits 8-13: color format + * - bits 16-31: DMA_NOTIFY instance + * word 2: + * - bits 0-15: DMA_A instance + * - bits 16-31: DMA_B instance + * + * NV05 will set/unset the relevant valid bits when you poke the relevant + * object-binding methods with object of the proper type, or with the NULL + * type. It'll only allow rendering using the grobj if all needed objects + * are bound. The needed set of objects depends on selected operation: for + * example rop object is needed by ROP_AND, but not by SRCCOPY_AND. + * + * NV04 doesn't have these methods implemented at all, and doesn't have the + * relevant bits in grobj. Instead, it'll allow rendering whenever bit 24 + * is set. So we have to emulate them in software, internally keeping the + * same bits as NV05 does. Since grobjs are aligned to 16 bytes on nv04, + * but the last word isn't actually used for anything, we abuse it for this + * purpose. + * + * Actually, NV05 can optionally check bit 24 too, but we disable this since + * there's no use for it. + * + * For unknown reasons, NV04 implements surf3d binding in hardware as an + * exception. Also for unknown reasons, NV04 doesn't implement the clipping + * methods on the surf3d object, so we have to emulate them too. + */ + +static void +nv04_gr_set_ctx1(struct nvkm_device *device, u32 inst, u32 mask, u32 value) +{ + int subc = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7; + u32 tmp; + + tmp = nvkm_rd32(device, 0x700000 + inst); + tmp &= ~mask; + tmp |= value; + nvkm_wr32(device, 0x700000 + inst, tmp); + + nvkm_wr32(device, NV04_PGRAPH_CTX_SWITCH1, tmp); + nvkm_wr32(device, NV04_PGRAPH_CTX_CACHE1 + (subc << 2), tmp); +} + +static void +nv04_gr_set_ctx_val(struct nvkm_device *device, u32 inst, u32 mask, u32 value) +{ + int class, op, valid = 1; + u32 tmp, ctx1; + + ctx1 = nvkm_rd32(device, 0x700000 + inst); + class = ctx1 & 0xff; + op = (ctx1 >> 15) & 7; + + tmp = nvkm_rd32(device, 0x70000c + inst); + tmp &= ~mask; + tmp |= value; + nvkm_wr32(device, 0x70000c + inst, tmp); + + /* check for valid surf2d/surf_dst/surf_color */ + if (!(tmp & 0x02000000)) + valid = 0; + /* check for valid surf_src/surf_zeta */ + if ((class == 0x1f || class == 0x48) && !(tmp & 0x04000000)) + valid = 0; + + switch (op) { + /* SRCCOPY_AND, SRCCOPY: no extra objects required */ + case 0: + case 3: + break; + /* ROP_AND: requires pattern and rop */ + case 1: + if (!(tmp & 0x18000000)) + valid = 0; + break; + /* BLEND_AND: requires beta1 */ + case 2: + if (!(tmp & 0x20000000)) + valid = 0; + break; + /* SRCCOPY_PREMULT, BLEND_PREMULT: beta4 required */ + case 4: + case 5: + if (!(tmp & 0x40000000)) + valid = 0; + break; + } + + nv04_gr_set_ctx1(device, inst, 0x01000000, valid << 24); +} + +static bool +nv04_gr_mthd_set_operation(struct nvkm_device *device, u32 inst, u32 data) +{ + u8 class = nvkm_rd32(device, 0x700000) & 0x000000ff; + if (data > 5) + return false; + /* Old versions of the objects only accept first three operations. */ + if (data > 2 && class < 0x40) + return false; + nv04_gr_set_ctx1(device, inst, 0x00038000, data << 15); + /* changing operation changes set of objects needed for validation */ + nv04_gr_set_ctx_val(device, inst, 0, 0); + return true; +} + +static bool +nv04_gr_mthd_surf3d_clip_h(struct nvkm_device *device, u32 inst, u32 data) +{ + u32 min = data & 0xffff, max; + u32 w = data >> 16; + if (min & 0x8000) + /* too large */ + return false; + if (w & 0x8000) + /* yes, it accepts negative for some reason. */ + w |= 0xffff0000; + max = min + w; + max &= 0x3ffff; + nvkm_wr32(device, 0x40053c, min); + nvkm_wr32(device, 0x400544, max); + return true; +} + +static bool +nv04_gr_mthd_surf3d_clip_v(struct nvkm_device *device, u32 inst, u32 data) +{ + u32 min = data & 0xffff, max; + u32 w = data >> 16; + if (min & 0x8000) + /* too large */ + return false; + if (w & 0x8000) + /* yes, it accepts negative for some reason. */ + w |= 0xffff0000; + max = min + w; + max &= 0x3ffff; + nvkm_wr32(device, 0x400540, min); + nvkm_wr32(device, 0x400548, max); + return true; +} + +static u8 +nv04_gr_mthd_bind_class(struct nvkm_device *device, u32 inst) +{ + return nvkm_rd32(device, 0x700000 + (inst << 4)); +} + +static bool +nv04_gr_mthd_bind_surf2d(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; + case 0x42: + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_surf2d_swzsurf(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; + case 0x42: + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; + case 0x52: + nv04_gr_set_ctx1(device, inst, 0x00004000, 0x00004000); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; + } + return false; +} + +static bool +nv01_gr_mthd_bind_patt(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0); + return true; + case 0x18: + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0x08000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_patt(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0); + return true; + case 0x44: + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0x08000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_rop(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x10000000, 0); + return true; + case 0x43: + nv04_gr_set_ctx_val(device, inst, 0x10000000, 0x10000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_beta1(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x20000000, 0); + return true; + case 0x12: + nv04_gr_set_ctx_val(device, inst, 0x20000000, 0x20000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_beta4(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x40000000, 0); + return true; + case 0x72: + nv04_gr_set_ctx_val(device, inst, 0x40000000, 0x40000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_surf_dst(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; + case 0x58: + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_surf_src(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0); + return true; + case 0x59: + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0x04000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_surf_color(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; + case 0x5a: + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; + } + return false; +} + +static bool +nv04_gr_mthd_bind_surf_zeta(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0); + return true; + case 0x5b: + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0x04000000); + return true; + } + return false; +} + +static bool +nv01_gr_mthd_bind_clip(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx1(device, inst, 0x2000, 0); + return true; + case 0x19: + nv04_gr_set_ctx1(device, inst, 0x2000, 0x2000); + return true; + } + return false; +} + +static bool +nv01_gr_mthd_bind_chroma(struct nvkm_device *device, u32 inst, u32 data) +{ + switch (nv04_gr_mthd_bind_class(device, data)) { + case 0x30: + nv04_gr_set_ctx1(device, inst, 0x1000, 0); + return true; + /* Yes, for some reason even the old versions of objects + * accept 0x57 and not 0x17. Consistency be damned. + */ + case 0x57: + nv04_gr_set_ctx1(device, inst, 0x1000, 0x1000); + return true; + } + return false; +} + +static bool +nv03_gr_mthd_gdi(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_patt; break; + case 0x0188: func = nv04_gr_mthd_bind_rop; break; + case 0x018c: func = nv04_gr_mthd_bind_beta1; break; + case 0x0190: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_gdi(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv01_gr_mthd_blit(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv01_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x019c: func = nv04_gr_mthd_bind_surf_src; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_blit(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_beta4; break; + case 0x019c: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_iifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_chroma; break; + case 0x018c: func = nv01_gr_mthd_bind_clip; break; + case 0x0190: func = nv04_gr_mthd_bind_patt; break; + case 0x0194: func = nv04_gr_mthd_bind_rop; break; + case 0x0198: func = nv04_gr_mthd_bind_beta1; break; + case 0x019c: func = nv04_gr_mthd_bind_beta4; break; + case 0x01a0: func = nv04_gr_mthd_bind_surf2d_swzsurf; break; + case 0x03e4: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv01_gr_mthd_ifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv01_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_ifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_beta4; break; + case 0x019c: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv03_gr_mthd_sifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_sifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv03_gr_mthd_sifm(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x0304: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_sifm(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x0304: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_surf3d(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x02f8: func = nv04_gr_mthd_surf3d_clip_h; break; + case 0x02fc: func = nv04_gr_mthd_surf3d_clip_v; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv03_gr_mthd_ttri(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_surf_color; break; + case 0x0190: func = nv04_gr_mthd_bind_surf_zeta; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv01_gr_mthd_prim(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_clip; break; + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd_prim(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_clip; break; + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} + +static bool +nv04_gr_mthd(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32, u32); + switch (nvkm_rd32(device, 0x700000 + inst) & 0x000000ff) { + case 0x1c ... 0x1e: + func = nv01_gr_mthd_prim; break; + case 0x1f: func = nv01_gr_mthd_blit; break; + case 0x21: func = nv01_gr_mthd_ifc; break; + case 0x36: func = nv03_gr_mthd_sifc; break; + case 0x37: func = nv03_gr_mthd_sifm; break; + case 0x48: func = nv03_gr_mthd_ttri; break; + case 0x4a: func = nv04_gr_mthd_gdi; break; + case 0x4b: func = nv03_gr_mthd_gdi; break; + case 0x53: func = nv04_gr_mthd_surf3d; break; + case 0x5c ... 0x5e: + func = nv04_gr_mthd_prim; break; + case 0x5f: func = nv04_gr_mthd_blit; break; + case 0x60: func = nv04_gr_mthd_iifc; break; + case 0x61: func = nv04_gr_mthd_ifc; break; + case 0x76: func = nv04_gr_mthd_sifc; break; + case 0x77: func = nv04_gr_mthd_sifm; break; + default: + return false; + } + return func(device, inst, mthd, data); +} + +static int +nv04_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, align, + false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); +#ifdef __BIG_ENDIAN + nvkm_mo32(*pgpuobj, 0x00, 0x00080000, 0x00080000); +#endif + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; +} + +const struct nvkm_object_func +nv04_gr_object = { + .bind = nv04_gr_object_bind, +}; + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static struct nv04_gr_chan * +nv04_gr_channel(struct nv04_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nv04_gr_chan *chan = NULL; + if (nvkm_rd32(device, NV04_PGRAPH_CTX_CONTROL) & 0x00010000) { + int chid = nvkm_rd32(device, NV04_PGRAPH_CTX_USER) >> 24; + if (chid < ARRAY_SIZE(gr->chan)) + chan = gr->chan[chid]; + } + return chan; +} + +static int +nv04_gr_load_context(struct nv04_gr_chan *chan, int chid) +{ + struct nvkm_device *device = chan->gr->base.engine.subdev.device; + int i; + + for (i = 0; i < ARRAY_SIZE(nv04_gr_ctx_regs); i++) + nvkm_wr32(device, nv04_gr_ctx_regs[i], chan->nv04[i]); + + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, chid << 24); + nvkm_mask(device, NV04_PGRAPH_FFINTFC_ST2, 0xfff00000, 0x00000000); + return 0; +} + +static int +nv04_gr_unload_context(struct nv04_gr_chan *chan) +{ + struct nvkm_device *device = chan->gr->base.engine.subdev.device; + int i; + + for (i = 0; i < ARRAY_SIZE(nv04_gr_ctx_regs); i++) + chan->nv04[i] = nvkm_rd32(device, nv04_gr_ctx_regs[i]); + + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL, 0x10000000); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); + return 0; +} + +static void +nv04_gr_context_switch(struct nv04_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nv04_gr_chan *prev = NULL; + struct nv04_gr_chan *next = NULL; + int chid; + + nv04_gr_idle(&gr->base); + + /* If previous context is valid, we need to save it */ + prev = nv04_gr_channel(gr); + if (prev) + nv04_gr_unload_context(prev); + + /* load context for next channel */ + chid = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 24) & 0x0f; + next = gr->chan[chid]; + if (next) + nv04_gr_load_context(next, chid); +} + +static u32 *ctx_reg(struct nv04_gr_chan *chan, u32 reg) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(nv04_gr_ctx_regs); i++) { + if (nv04_gr_ctx_regs[i] == reg) + return &chan->nv04[i]; + } + + return NULL; +} + +static void * +nv04_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv04_gr_chan *chan = nv04_gr_chan(object); + struct nv04_gr *gr = chan->gr; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = NULL; + spin_unlock_irqrestore(&gr->lock, flags); + return chan; +} + +static int +nv04_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv04_gr_chan *chan = nv04_gr_chan(object); + struct nv04_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + if (nv04_gr_channel(gr) == chan) + nv04_gr_unload_context(chan); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} + +static const struct nvkm_object_func +nv04_gr_chan = { + .dtor = nv04_gr_chan_dtor, + .fini = nv04_gr_chan_fini, +}; + +static int +nv04_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv04_gr *gr = nv04_gr(base); + struct nv04_gr_chan *chan; + unsigned long flags; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv04_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + *ctx_reg(chan, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31; + + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = chan; + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +bool +nv04_gr_idle(struct nvkm_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 mask = 0xffffffff; + + if (device->card_type == NV_40) + mask &= ~NV40_PGRAPH_STATUS_SYNC_STALL; + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, NV04_PGRAPH_STATUS) & mask)) + break; + ) < 0) { + nvkm_error(subdev, "idle timed out with status %08x\n", + nvkm_rd32(device, NV04_PGRAPH_STATUS)); + return false; + } + + return true; +} + +static const struct nvkm_bitfield +nv04_gr_intr_name[] = { + { NV_PGRAPH_INTR_NOTIFY, "NOTIFY" }, + {} +}; + +static const struct nvkm_bitfield +nv04_gr_nstatus[] = { + { NV04_PGRAPH_NSTATUS_STATE_IN_USE, "STATE_IN_USE" }, + { NV04_PGRAPH_NSTATUS_INVALID_STATE, "INVALID_STATE" }, + { NV04_PGRAPH_NSTATUS_BAD_ARGUMENT, "BAD_ARGUMENT" }, + { NV04_PGRAPH_NSTATUS_PROTECTION_FAULT, "PROTECTION_FAULT" }, + {} +}; + +const struct nvkm_bitfield +nv04_gr_nsource[] = { + { NV03_PGRAPH_NSOURCE_NOTIFICATION, "NOTIFICATION" }, + { NV03_PGRAPH_NSOURCE_DATA_ERROR, "DATA_ERROR" }, + { NV03_PGRAPH_NSOURCE_PROTECTION_ERROR, "PROTECTION_ERROR" }, + { NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION, "RANGE_EXCEPTION" }, + { NV03_PGRAPH_NSOURCE_LIMIT_COLOR, "LIMIT_COLOR" }, + { NV03_PGRAPH_NSOURCE_LIMIT_ZETA, "LIMIT_ZETA" }, + { NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD, "ILLEGAL_MTHD" }, + { NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION, "DMA_R_PROTECTION" }, + { NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION, "DMA_W_PROTECTION" }, + { NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION, "FORMAT_EXCEPTION" }, + { NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION, "PATCH_EXCEPTION" }, + { NV03_PGRAPH_NSOURCE_STATE_INVALID, "STATE_INVALID" }, + { NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY, "DOUBLE_NOTIFY" }, + { NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE, "NOTIFY_IN_USE" }, + { NV03_PGRAPH_NSOURCE_METHOD_CNT, "METHOD_CNT" }, + { NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION, "BFR_NOTIFICATION" }, + { NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION, "DMA_VTX_PROTECTION" }, + { NV03_PGRAPH_NSOURCE_DMA_WIDTH_A, "DMA_WIDTH_A" }, + { NV03_PGRAPH_NSOURCE_DMA_WIDTH_B, "DMA_WIDTH_B" }, + {} +}; + +static void +nv04_gr_intr(struct nvkm_gr *base) +{ + struct nv04_gr *gr = nv04_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); + u32 chid = (addr & 0x0f000000) >> 24; + u32 subc = (addr & 0x0000e000) >> 13; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400180 + subc * 4) & 0xff; + u32 inst = (nvkm_rd32(device, 0x40016c) & 0xffff) << 4; + u32 show = stat; + char msg[128], src[128], sta[128]; + struct nv04_gr_chan *chan; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + chan = gr->chan[chid]; + + if (stat & NV_PGRAPH_INTR_NOTIFY) { + if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) { + if (!nv04_gr_mthd(device, inst, mthd, data)) + show &= ~NV_PGRAPH_INTR_NOTIFY; + } + } + + if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) { + nvkm_wr32(device, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); + stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; + show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; + nv04_gr_context_switch(gr); + } + + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); + + if (show) { + nvkm_snprintbf(msg, sizeof(msg), nv04_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv04_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->object.client->name : "unknown", + subc, class, mthd, data); + } + + spin_unlock_irqrestore(&gr->lock, flags); +} + +static int +nv04_gr_init(struct nvkm_gr *base) +{ + struct nv04_gr *gr = nv04_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + + /* Enable PGRAPH interrupts */ + nvkm_wr32(device, NV03_PGRAPH_INTR, 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_VALID1, 0); + nvkm_wr32(device, NV04_PGRAPH_VALID2, 0); + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x000001FF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x001FFFFF);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x1231c000); + /*1231C000 blob, 001 haiku*/ + /*V_WRITE(NV04_PGRAPH_DEBUG_1, 0xf2d91100);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x72111100); + /*0x72111100 blob , 01 haiku*/ + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x11d5f870);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x11d5f071); + /*haiku same*/ + + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xfad4ff31);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xf0d4ff31); + /*haiku and blob 10d4*/ + + nvkm_wr32(device, NV04_PGRAPH_STATE , 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL , 0x10000100); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); + + /* These don't belong here, they're part of a per-channel context */ + nvkm_wr32(device, NV04_PGRAPH_PATTERN_SHAPE, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_BETA_AND , 0xFFFFFFFF); + return 0; +} + +static const struct nvkm_gr_func +nv04_gr = { + .init = nv04_gr_init, + .intr = nv04_gr_intr, + .chan_new = nv04_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0017, &nv04_gr_object }, /* chroma */ + { -1, -1, 0x0018, &nv04_gr_object }, /* pattern (nv01) */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x001c, &nv04_gr_object }, /* line */ + { -1, -1, 0x001d, &nv04_gr_object }, /* tri */ + { -1, -1, 0x001e, &nv04_gr_object }, /* rect */ + { -1, -1, 0x001f, &nv04_gr_object }, + { -1, -1, 0x0021, &nv04_gr_object }, + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0036, &nv04_gr_object }, + { -1, -1, 0x0037, &nv04_gr_object }, + { -1, -1, 0x0038, &nv04_gr_object }, /* dvd subpicture */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0042, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x0048, &nv04_gr_object }, + { -1, -1, 0x004a, &nv04_gr_object }, + { -1, -1, 0x004b, &nv04_gr_object }, + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x0053, &nv04_gr_object }, + { -1, -1, 0x0054, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0055, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0057, &nv04_gr_object }, /* chroma */ + { -1, -1, 0x0058, &nv04_gr_object }, /* surf_dst */ + { -1, -1, 0x0059, &nv04_gr_object }, /* surf_src */ + { -1, -1, 0x005a, &nv04_gr_object }, /* surf_color */ + { -1, -1, 0x005b, &nv04_gr_object }, /* surf_zeta */ + { -1, -1, 0x005c, &nv04_gr_object }, /* line */ + { -1, -1, 0x005d, &nv04_gr_object }, /* tri */ + { -1, -1, 0x005e, &nv04_gr_object }, /* rect */ + { -1, -1, 0x005f, &nv04_gr_object }, + { -1, -1, 0x0060, &nv04_gr_object }, + { -1, -1, 0x0061, &nv04_gr_object }, + { -1, -1, 0x0064, &nv04_gr_object }, /* iifc (nv05) */ + { -1, -1, 0x0065, &nv04_gr_object }, /* ifc (nv05) */ + { -1, -1, 0x0066, &nv04_gr_object }, /* sifc (nv05) */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0076, &nv04_gr_object }, + { -1, -1, 0x0077, &nv04_gr_object }, + {} + } +}; + +int +nv04_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct nv04_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(&nv04_gr, device, type, inst, true, &gr->base); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c new file mode 100644 index 0000000000..92ef7c9b29 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c @@ -0,0 +1,1221 @@ +/* + * Copyright 2007 Matthieu CASTET + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "nv10.h" +#include "regs.h" + +#include +#include +#include +#include +#include + +struct pipe_state { + u32 pipe_0x0000[0x040/4]; + u32 pipe_0x0040[0x010/4]; + u32 pipe_0x0200[0x0c0/4]; + u32 pipe_0x4400[0x080/4]; + u32 pipe_0x6400[0x3b0/4]; + u32 pipe_0x6800[0x2f0/4]; + u32 pipe_0x6c00[0x030/4]; + u32 pipe_0x7000[0x130/4]; + u32 pipe_0x7400[0x0c0/4]; + u32 pipe_0x7800[0x0c0/4]; +}; + +static int nv10_gr_ctx_regs[] = { + NV10_PGRAPH_CTX_SWITCH(0), + NV10_PGRAPH_CTX_SWITCH(1), + NV10_PGRAPH_CTX_SWITCH(2), + NV10_PGRAPH_CTX_SWITCH(3), + NV10_PGRAPH_CTX_SWITCH(4), + NV10_PGRAPH_CTX_CACHE(0, 0), + NV10_PGRAPH_CTX_CACHE(0, 1), + NV10_PGRAPH_CTX_CACHE(0, 2), + NV10_PGRAPH_CTX_CACHE(0, 3), + NV10_PGRAPH_CTX_CACHE(0, 4), + NV10_PGRAPH_CTX_CACHE(1, 0), + NV10_PGRAPH_CTX_CACHE(1, 1), + NV10_PGRAPH_CTX_CACHE(1, 2), + NV10_PGRAPH_CTX_CACHE(1, 3), + NV10_PGRAPH_CTX_CACHE(1, 4), + NV10_PGRAPH_CTX_CACHE(2, 0), + NV10_PGRAPH_CTX_CACHE(2, 1), + NV10_PGRAPH_CTX_CACHE(2, 2), + NV10_PGRAPH_CTX_CACHE(2, 3), + NV10_PGRAPH_CTX_CACHE(2, 4), + NV10_PGRAPH_CTX_CACHE(3, 0), + NV10_PGRAPH_CTX_CACHE(3, 1), + NV10_PGRAPH_CTX_CACHE(3, 2), + NV10_PGRAPH_CTX_CACHE(3, 3), + NV10_PGRAPH_CTX_CACHE(3, 4), + NV10_PGRAPH_CTX_CACHE(4, 0), + NV10_PGRAPH_CTX_CACHE(4, 1), + NV10_PGRAPH_CTX_CACHE(4, 2), + NV10_PGRAPH_CTX_CACHE(4, 3), + NV10_PGRAPH_CTX_CACHE(4, 4), + NV10_PGRAPH_CTX_CACHE(5, 0), + NV10_PGRAPH_CTX_CACHE(5, 1), + NV10_PGRAPH_CTX_CACHE(5, 2), + NV10_PGRAPH_CTX_CACHE(5, 3), + NV10_PGRAPH_CTX_CACHE(5, 4), + NV10_PGRAPH_CTX_CACHE(6, 0), + NV10_PGRAPH_CTX_CACHE(6, 1), + NV10_PGRAPH_CTX_CACHE(6, 2), + NV10_PGRAPH_CTX_CACHE(6, 3), + NV10_PGRAPH_CTX_CACHE(6, 4), + NV10_PGRAPH_CTX_CACHE(7, 0), + NV10_PGRAPH_CTX_CACHE(7, 1), + NV10_PGRAPH_CTX_CACHE(7, 2), + NV10_PGRAPH_CTX_CACHE(7, 3), + NV10_PGRAPH_CTX_CACHE(7, 4), + NV10_PGRAPH_CTX_USER, + NV04_PGRAPH_DMA_START_0, + NV04_PGRAPH_DMA_START_1, + NV04_PGRAPH_DMA_LENGTH, + NV04_PGRAPH_DMA_MISC, + NV10_PGRAPH_DMA_PITCH, + NV04_PGRAPH_BOFFSET0, + NV04_PGRAPH_BBASE0, + NV04_PGRAPH_BLIMIT0, + NV04_PGRAPH_BOFFSET1, + NV04_PGRAPH_BBASE1, + NV04_PGRAPH_BLIMIT1, + NV04_PGRAPH_BOFFSET2, + NV04_PGRAPH_BBASE2, + NV04_PGRAPH_BLIMIT2, + NV04_PGRAPH_BOFFSET3, + NV04_PGRAPH_BBASE3, + NV04_PGRAPH_BLIMIT3, + NV04_PGRAPH_BOFFSET4, + NV04_PGRAPH_BBASE4, + NV04_PGRAPH_BLIMIT4, + NV04_PGRAPH_BOFFSET5, + NV04_PGRAPH_BBASE5, + NV04_PGRAPH_BLIMIT5, + NV04_PGRAPH_BPITCH0, + NV04_PGRAPH_BPITCH1, + NV04_PGRAPH_BPITCH2, + NV04_PGRAPH_BPITCH3, + NV04_PGRAPH_BPITCH4, + NV10_PGRAPH_SURFACE, + NV10_PGRAPH_STATE, + NV04_PGRAPH_BSWIZZLE2, + NV04_PGRAPH_BSWIZZLE5, + NV04_PGRAPH_BPIXEL, + NV10_PGRAPH_NOTIFY, + NV04_PGRAPH_PATT_COLOR0, + NV04_PGRAPH_PATT_COLOR1, + NV04_PGRAPH_PATT_COLORRAM, /* 64 values from 0x400900 to 0x4009fc */ + 0x00400904, + 0x00400908, + 0x0040090c, + 0x00400910, + 0x00400914, + 0x00400918, + 0x0040091c, + 0x00400920, + 0x00400924, + 0x00400928, + 0x0040092c, + 0x00400930, + 0x00400934, + 0x00400938, + 0x0040093c, + 0x00400940, + 0x00400944, + 0x00400948, + 0x0040094c, + 0x00400950, + 0x00400954, + 0x00400958, + 0x0040095c, + 0x00400960, + 0x00400964, + 0x00400968, + 0x0040096c, + 0x00400970, + 0x00400974, + 0x00400978, + 0x0040097c, + 0x00400980, + 0x00400984, + 0x00400988, + 0x0040098c, + 0x00400990, + 0x00400994, + 0x00400998, + 0x0040099c, + 0x004009a0, + 0x004009a4, + 0x004009a8, + 0x004009ac, + 0x004009b0, + 0x004009b4, + 0x004009b8, + 0x004009bc, + 0x004009c0, + 0x004009c4, + 0x004009c8, + 0x004009cc, + 0x004009d0, + 0x004009d4, + 0x004009d8, + 0x004009dc, + 0x004009e0, + 0x004009e4, + 0x004009e8, + 0x004009ec, + 0x004009f0, + 0x004009f4, + 0x004009f8, + 0x004009fc, + NV04_PGRAPH_PATTERN, /* 2 values from 0x400808 to 0x40080c */ + 0x0040080c, + NV04_PGRAPH_PATTERN_SHAPE, + NV03_PGRAPH_MONO_COLOR0, + NV04_PGRAPH_ROP3, + NV04_PGRAPH_CHROMA, + NV04_PGRAPH_BETA_AND, + NV04_PGRAPH_BETA_PREMULT, + 0x00400e70, + 0x00400e74, + 0x00400e78, + 0x00400e7c, + 0x00400e80, + 0x00400e84, + 0x00400e88, + 0x00400e8c, + 0x00400ea0, + 0x00400ea4, + 0x00400ea8, + 0x00400e90, + 0x00400e94, + 0x00400e98, + 0x00400e9c, + NV10_PGRAPH_WINDOWCLIP_HORIZONTAL, /* 8 values from 0x400f00-0x400f1c */ + NV10_PGRAPH_WINDOWCLIP_VERTICAL, /* 8 values from 0x400f20-0x400f3c */ + 0x00400f04, + 0x00400f24, + 0x00400f08, + 0x00400f28, + 0x00400f0c, + 0x00400f2c, + 0x00400f10, + 0x00400f30, + 0x00400f14, + 0x00400f34, + 0x00400f18, + 0x00400f38, + 0x00400f1c, + 0x00400f3c, + NV10_PGRAPH_XFMODE0, + NV10_PGRAPH_XFMODE1, + NV10_PGRAPH_GLOBALSTATE0, + NV10_PGRAPH_GLOBALSTATE1, + NV04_PGRAPH_STORED_FMT, + NV04_PGRAPH_SOURCE_COLOR, + NV03_PGRAPH_ABS_X_RAM, /* 32 values from 0x400400 to 0x40047c */ + NV03_PGRAPH_ABS_Y_RAM, /* 32 values from 0x400480 to 0x4004fc */ + 0x00400404, + 0x00400484, + 0x00400408, + 0x00400488, + 0x0040040c, + 0x0040048c, + 0x00400410, + 0x00400490, + 0x00400414, + 0x00400494, + 0x00400418, + 0x00400498, + 0x0040041c, + 0x0040049c, + 0x00400420, + 0x004004a0, + 0x00400424, + 0x004004a4, + 0x00400428, + 0x004004a8, + 0x0040042c, + 0x004004ac, + 0x00400430, + 0x004004b0, + 0x00400434, + 0x004004b4, + 0x00400438, + 0x004004b8, + 0x0040043c, + 0x004004bc, + 0x00400440, + 0x004004c0, + 0x00400444, + 0x004004c4, + 0x00400448, + 0x004004c8, + 0x0040044c, + 0x004004cc, + 0x00400450, + 0x004004d0, + 0x00400454, + 0x004004d4, + 0x00400458, + 0x004004d8, + 0x0040045c, + 0x004004dc, + 0x00400460, + 0x004004e0, + 0x00400464, + 0x004004e4, + 0x00400468, + 0x004004e8, + 0x0040046c, + 0x004004ec, + 0x00400470, + 0x004004f0, + 0x00400474, + 0x004004f4, + 0x00400478, + 0x004004f8, + 0x0040047c, + 0x004004fc, + NV03_PGRAPH_ABS_UCLIP_XMIN, + NV03_PGRAPH_ABS_UCLIP_XMAX, + NV03_PGRAPH_ABS_UCLIP_YMIN, + NV03_PGRAPH_ABS_UCLIP_YMAX, + 0x00400550, + 0x00400558, + 0x00400554, + 0x0040055c, + NV03_PGRAPH_ABS_UCLIPA_XMIN, + NV03_PGRAPH_ABS_UCLIPA_XMAX, + NV03_PGRAPH_ABS_UCLIPA_YMIN, + NV03_PGRAPH_ABS_UCLIPA_YMAX, + NV03_PGRAPH_ABS_ICLIP_XMAX, + NV03_PGRAPH_ABS_ICLIP_YMAX, + NV03_PGRAPH_XY_LOGIC_MISC0, + NV03_PGRAPH_XY_LOGIC_MISC1, + NV03_PGRAPH_XY_LOGIC_MISC2, + NV03_PGRAPH_XY_LOGIC_MISC3, + NV03_PGRAPH_CLIPX_0, + NV03_PGRAPH_CLIPX_1, + NV03_PGRAPH_CLIPY_0, + NV03_PGRAPH_CLIPY_1, + NV10_PGRAPH_COMBINER0_IN_ALPHA, + NV10_PGRAPH_COMBINER1_IN_ALPHA, + NV10_PGRAPH_COMBINER0_IN_RGB, + NV10_PGRAPH_COMBINER1_IN_RGB, + NV10_PGRAPH_COMBINER_COLOR0, + NV10_PGRAPH_COMBINER_COLOR1, + NV10_PGRAPH_COMBINER0_OUT_ALPHA, + NV10_PGRAPH_COMBINER1_OUT_ALPHA, + NV10_PGRAPH_COMBINER0_OUT_RGB, + NV10_PGRAPH_COMBINER1_OUT_RGB, + NV10_PGRAPH_COMBINER_FINAL0, + NV10_PGRAPH_COMBINER_FINAL1, + 0x00400e00, + 0x00400e04, + 0x00400e08, + 0x00400e0c, + 0x00400e10, + 0x00400e14, + 0x00400e18, + 0x00400e1c, + 0x00400e20, + 0x00400e24, + 0x00400e28, + 0x00400e2c, + 0x00400e30, + 0x00400e34, + 0x00400e38, + 0x00400e3c, + NV04_PGRAPH_PASSTHRU_0, + NV04_PGRAPH_PASSTHRU_1, + NV04_PGRAPH_PASSTHRU_2, + NV10_PGRAPH_DIMX_TEXTURE, + NV10_PGRAPH_WDIMX_TEXTURE, + NV10_PGRAPH_DVD_COLORFMT, + NV10_PGRAPH_SCALED_FORMAT, + NV04_PGRAPH_MISC24_0, + NV04_PGRAPH_MISC24_1, + NV04_PGRAPH_MISC24_2, + NV03_PGRAPH_X_MISC, + NV03_PGRAPH_Y_MISC, + NV04_PGRAPH_VALID1, + NV04_PGRAPH_VALID2, +}; + +static int nv17_gr_ctx_regs[] = { + NV10_PGRAPH_DEBUG_4, + 0x004006b0, + 0x00400eac, + 0x00400eb0, + 0x00400eb4, + 0x00400eb8, + 0x00400ebc, + 0x00400ec0, + 0x00400ec4, + 0x00400ec8, + 0x00400ecc, + 0x00400ed0, + 0x00400ed4, + 0x00400ed8, + 0x00400edc, + 0x00400ee0, + 0x00400a00, + 0x00400a04, +}; + +#define nv10_gr(p) container_of((p), struct nv10_gr, base) + +struct nv10_gr { + struct nvkm_gr base; + struct nv10_gr_chan *chan[32]; + spinlock_t lock; +}; + +#define nv10_gr_chan(p) container_of((p), struct nv10_gr_chan, object) + +struct nv10_gr_chan { + struct nvkm_object object; + struct nv10_gr *gr; + int chid; + int nv10[ARRAY_SIZE(nv10_gr_ctx_regs)]; + int nv17[ARRAY_SIZE(nv17_gr_ctx_regs)]; + struct pipe_state pipe_state; + u32 lma_window[4]; +}; + + +/******************************************************************************* + * Graphics object classes + ******************************************************************************/ + +#define PIPE_SAVE(gr, state, addr) \ + do { \ + int __i; \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ + state[__i] = nvkm_rd32(device, NV10_PGRAPH_PIPE_DATA); \ + } while (0) + +#define PIPE_RESTORE(gr, state, addr) \ + do { \ + int __i; \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, state[__i]); \ + } while (0) + +static void +nv17_gr_mthd_lma_window(struct nv10_gr_chan *chan, u32 mthd, u32 data) +{ + struct nvkm_device *device = chan->object.engine->subdev.device; + struct nvkm_gr *gr = &chan->gr->base; + struct pipe_state *pipe = &chan->pipe_state; + u32 pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3]; + u32 xfmode0, xfmode1; + int i; + + chan->lma_window[(mthd - 0x1638) / 4] = data; + + if (mthd != 0x1644) + return; + + nv04_gr_idle(gr); + + PIPE_SAVE(device, pipe_0x0040, 0x0040); + PIPE_SAVE(device, pipe->pipe_0x0200, 0x0200); + + PIPE_RESTORE(device, chan->lma_window, 0x6790); + + nv04_gr_idle(gr); + + xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0); + xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1); + + PIPE_SAVE(device, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(device, pipe_0x64c0, 0x64c0); + PIPE_SAVE(device, pipe_0x6ab0, 0x6ab0); + PIPE_SAVE(device, pipe_0x6a80, 0x6a80); + + nv04_gr_idle(gr); + + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); + for (i = 0; i < 4; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + for (i = 0; i < 4; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); + for (i = 0; i < 3; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); + for (i = 0; i < 3; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008); + + PIPE_RESTORE(device, pipe->pipe_0x0200, 0x0200); + + nv04_gr_idle(gr); + + PIPE_RESTORE(device, pipe_0x0040, 0x0040); + + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1); + + PIPE_RESTORE(device, pipe_0x64c0, 0x64c0); + PIPE_RESTORE(device, pipe_0x6ab0, 0x6ab0); + PIPE_RESTORE(device, pipe_0x6a80, 0x6a80); + PIPE_RESTORE(device, pipe->pipe_0x4400, 0x4400); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nv04_gr_idle(gr); +} + +static void +nv17_gr_mthd_lma_enable(struct nv10_gr_chan *chan, u32 mthd, u32 data) +{ + struct nvkm_device *device = chan->object.engine->subdev.device; + struct nvkm_gr *gr = &chan->gr->base; + + nv04_gr_idle(gr); + + nvkm_mask(device, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100); + nvkm_mask(device, 0x4006b0, 0x08000000, 0x08000000); +} + +static bool +nv17_gr_mthd_celcius(struct nv10_gr_chan *chan, u32 mthd, u32 data) +{ + void (*func)(struct nv10_gr_chan *, u32, u32); + switch (mthd) { + case 0x1638 ... 0x1644: + func = nv17_gr_mthd_lma_window; break; + case 0x1658: func = nv17_gr_mthd_lma_enable; break; + default: + return false; + } + func(chan, mthd, data); + return true; +} + +static bool +nv10_gr_mthd(struct nv10_gr_chan *chan, u8 class, u32 mthd, u32 data) +{ + bool (*func)(struct nv10_gr_chan *, u32, u32); + switch (class) { + case 0x99: func = nv17_gr_mthd_celcius; break; + default: + return false; + } + return func(chan, mthd, data); +} + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static struct nv10_gr_chan * +nv10_gr_channel(struct nv10_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nv10_gr_chan *chan = NULL; + if (nvkm_rd32(device, 0x400144) & 0x00010000) { + int chid = nvkm_rd32(device, 0x400148) >> 24; + if (chid < ARRAY_SIZE(gr->chan)) + chan = gr->chan[chid]; + } + return chan; +} + +static void +nv10_gr_save_pipe(struct nv10_gr_chan *chan) +{ + struct nv10_gr *gr = chan->gr; + struct pipe_state *pipe = &chan->pipe_state; + struct nvkm_device *device = gr->base.engine.subdev.device; + + PIPE_SAVE(gr, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(gr, pipe->pipe_0x0200, 0x0200); + PIPE_SAVE(gr, pipe->pipe_0x6400, 0x6400); + PIPE_SAVE(gr, pipe->pipe_0x6800, 0x6800); + PIPE_SAVE(gr, pipe->pipe_0x6c00, 0x6c00); + PIPE_SAVE(gr, pipe->pipe_0x7000, 0x7000); + PIPE_SAVE(gr, pipe->pipe_0x7400, 0x7400); + PIPE_SAVE(gr, pipe->pipe_0x7800, 0x7800); + PIPE_SAVE(gr, pipe->pipe_0x0040, 0x0040); + PIPE_SAVE(gr, pipe->pipe_0x0000, 0x0000); +} + +static void +nv10_gr_load_pipe(struct nv10_gr_chan *chan) +{ + struct nv10_gr *gr = chan->gr; + struct pipe_state *pipe = &chan->pipe_state; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 xfmode0, xfmode1; + int i; + + nv04_gr_idle(&gr->base); + /* XXX check haiku comments */ + xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0); + xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1); + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); + for (i = 0; i < 4; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + for (i = 0; i < 4; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); + for (i = 0; i < 3; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); + for (i = 0; i < 3; i++) + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); + + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008); + + + PIPE_RESTORE(gr, pipe->pipe_0x0200, 0x0200); + nv04_gr_idle(&gr->base); + + /* restore XFMODE */ + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1); + PIPE_RESTORE(gr, pipe->pipe_0x6400, 0x6400); + PIPE_RESTORE(gr, pipe->pipe_0x6800, 0x6800); + PIPE_RESTORE(gr, pipe->pipe_0x6c00, 0x6c00); + PIPE_RESTORE(gr, pipe->pipe_0x7000, 0x7000); + PIPE_RESTORE(gr, pipe->pipe_0x7400, 0x7400); + PIPE_RESTORE(gr, pipe->pipe_0x7800, 0x7800); + PIPE_RESTORE(gr, pipe->pipe_0x4400, 0x4400); + PIPE_RESTORE(gr, pipe->pipe_0x0000, 0x0000); + PIPE_RESTORE(gr, pipe->pipe_0x0040, 0x0040); + nv04_gr_idle(&gr->base); +} + +static void +nv10_gr_create_pipe(struct nv10_gr_chan *chan) +{ + struct nv10_gr *gr = chan->gr; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct pipe_state *pipe_state = &chan->pipe_state; + u32 *pipe_state_addr; + int i; +#define PIPE_INIT(addr) \ + do { \ + pipe_state_addr = pipe_state->pipe_##addr; \ + } while (0) +#define PIPE_INIT_END(addr) \ + do { \ + u32 *__end_addr = pipe_state->pipe_##addr + \ + ARRAY_SIZE(pipe_state->pipe_##addr); \ + if (pipe_state_addr != __end_addr) \ + nvkm_error(subdev, "incomplete pipe init for 0x%x : %p/%p\n", \ + addr, pipe_state_addr, __end_addr); \ + } while (0) +#define NV_WRITE_PIPE_INIT(value) *(pipe_state_addr++) = value + + PIPE_INIT(0x0200); + for (i = 0; i < 48; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x0200); + + PIPE_INIT(0x6400); + for (i = 0; i < 211; i++) + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x40000000); + NV_WRITE_PIPE_INIT(0x40000000); + NV_WRITE_PIPE_INIT(0x40000000); + NV_WRITE_PIPE_INIT(0x40000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f000000); + NV_WRITE_PIPE_INIT(0x3f000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x3f800000); + NV_WRITE_PIPE_INIT(0x3f800000); + PIPE_INIT_END(0x6400); + + PIPE_INIT(0x6800); + for (i = 0; i < 162; i++) + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x3f800000); + for (i = 0; i < 25; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x6800); + + PIPE_INIT(0x6c00); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0xbf800000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x6c00); + + PIPE_INIT(0x7000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x00000000); + NV_WRITE_PIPE_INIT(0x7149f2ca); + for (i = 0; i < 35; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x7000); + + PIPE_INIT(0x7400); + for (i = 0; i < 48; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x7400); + + PIPE_INIT(0x7800); + for (i = 0; i < 48; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x7800); + + PIPE_INIT(0x4400); + for (i = 0; i < 32; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x4400); + + PIPE_INIT(0x0000); + for (i = 0; i < 16; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x0000); + + PIPE_INIT(0x0040); + for (i = 0; i < 4; i++) + NV_WRITE_PIPE_INIT(0x00000000); + PIPE_INIT_END(0x0040); + +#undef PIPE_INIT +#undef PIPE_INIT_END +#undef NV_WRITE_PIPE_INIT +} + +static int +nv10_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + int i; + for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) { + if (nv10_gr_ctx_regs[i] == reg) + return i; + } + nvkm_error(subdev, "unknown offset nv10_ctx_regs %d\n", reg); + return -1; +} + +static int +nv17_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + int i; + for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) { + if (nv17_gr_ctx_regs[i] == reg) + return i; + } + nvkm_error(subdev, "unknown offset nv17_ctx_regs %d\n", reg); + return -1; +} + +static void +nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst) +{ + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 st2, st2_dl, st2_dh, fifo_ptr, fifo[0x60/4]; + u32 ctx_user, ctx_switch[5]; + int i, subchan = -1; + + /* NV10TCL_DMA_VTXBUF (method 0x18c) modifies hidden state + * that cannot be restored via MMIO. Do it through the FIFO + * instead. + */ + + /* Look for a celsius object */ + for (i = 0; i < 8; i++) { + int class = nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff; + + if (class == 0x56 || class == 0x96 || class == 0x99) { + subchan = i; + break; + } + } + + if (subchan < 0 || !inst) + return; + + /* Save the current ctx object */ + ctx_user = nvkm_rd32(device, NV10_PGRAPH_CTX_USER); + for (i = 0; i < 5; i++) + ctx_switch[i] = nvkm_rd32(device, NV10_PGRAPH_CTX_SWITCH(i)); + + /* Save the FIFO state */ + st2 = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2); + st2_dl = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DL); + st2_dh = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DH); + fifo_ptr = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR); + + for (i = 0; i < ARRAY_SIZE(fifo); i++) + fifo[i] = nvkm_rd32(device, 0x4007a0 + 4 * i); + + /* Switch to the celsius subchannel */ + for (i = 0; i < 5; i++) + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i), + nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(subchan, i))); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13); + + /* Inject NV10TCL_DMA_VTXBUF */ + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, + 0x2c000000 | chid << 20 | subchan << 16 | 0x18c); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, inst); + nvkm_mask(device, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + + /* Restore the FIFO state */ + for (i = 0; i < ARRAY_SIZE(fifo); i++) + nvkm_wr32(device, 0x4007a0 + 4 * i, fifo[i]); + + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, st2); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh); + + /* Restore the current ctx object */ + for (i = 0; i < 5; i++) + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]); + nvkm_wr32(device, NV10_PGRAPH_CTX_USER, ctx_user); +} + +static int +nv10_gr_load_context(struct nv10_gr_chan *chan, int chid) +{ + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 inst; + int i; + + for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) + nvkm_wr32(device, nv10_gr_ctx_regs[i], chan->nv10[i]); + + if (device->card_type >= NV_11 && device->chipset >= 0x17) { + for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) + nvkm_wr32(device, nv17_gr_ctx_regs[i], chan->nv17[i]); + } + + nv10_gr_load_pipe(chan); + + inst = nvkm_rd32(device, NV10_PGRAPH_GLOBALSTATE1) & 0xffff; + nv10_gr_load_dma_vtxbuf(chan, chid, inst); + + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24); + nvkm_mask(device, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000); + return 0; +} + +static int +nv10_gr_unload_context(struct nv10_gr_chan *chan) +{ + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + int i; + + for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) + chan->nv10[i] = nvkm_rd32(device, nv10_gr_ctx_regs[i]); + + if (device->card_type >= NV_11 && device->chipset >= 0x17) { + for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) + chan->nv17[i] = nvkm_rd32(device, nv17_gr_ctx_regs[i]); + } + + nv10_gr_save_pipe(chan); + + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000000); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); + return 0; +} + +static void +nv10_gr_context_switch(struct nv10_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nv10_gr_chan *prev = NULL; + struct nv10_gr_chan *next = NULL; + int chid; + + nv04_gr_idle(&gr->base); + + /* If previous context is valid, we need to save it */ + prev = nv10_gr_channel(gr); + if (prev) + nv10_gr_unload_context(prev); + + /* load context for next channel */ + chid = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f; + next = gr->chan[chid]; + if (next) + nv10_gr_load_context(next, chid); +} + +static int +nv10_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv10_gr_chan *chan = nv10_gr_chan(object); + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + if (nv10_gr_channel(gr) == chan) + nv10_gr_unload_context(chan); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} + +static void * +nv10_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv10_gr_chan *chan = nv10_gr_chan(object); + struct nv10_gr *gr = chan->gr; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = NULL; + spin_unlock_irqrestore(&gr->lock, flags); + return chan; +} + +static const struct nvkm_object_func +nv10_gr_chan = { + .dtor = nv10_gr_chan_dtor, + .fini = nv10_gr_chan_fini, +}; + +#define NV_WRITE_CTX(reg, val) do { \ + int offset = nv10_gr_ctx_regs_find_offset(gr, reg); \ + if (offset > 0) \ + chan->nv10[offset] = val; \ + } while (0) + +#define NV17_WRITE_CTX(reg, val) do { \ + int offset = nv17_gr_ctx_regs_find_offset(gr, reg); \ + if (offset > 0) \ + chan->nv17[offset] = val; \ + } while (0) + +int +nv10_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv10_gr *gr = nv10_gr(base); + struct nv10_gr_chan *chan; + struct nvkm_device *device = gr->base.engine.subdev.device; + unsigned long flags; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv10_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + NV_WRITE_CTX(0x00400e88, 0x08000000); + NV_WRITE_CTX(0x00400e9c, 0x4b7fffff); + NV_WRITE_CTX(NV03_PGRAPH_XY_LOGIC_MISC0, 0x0001ffff); + NV_WRITE_CTX(0x00400e10, 0x00001000); + NV_WRITE_CTX(0x00400e14, 0x00001000); + NV_WRITE_CTX(0x00400e30, 0x00080008); + NV_WRITE_CTX(0x00400e34, 0x00080008); + if (device->card_type >= NV_11 && device->chipset >= 0x17) { + /* is it really needed ??? */ + NV17_WRITE_CTX(NV10_PGRAPH_DEBUG_4, + nvkm_rd32(device, NV10_PGRAPH_DEBUG_4)); + NV17_WRITE_CTX(0x004006b0, nvkm_rd32(device, 0x004006b0)); + NV17_WRITE_CTX(0x00400eac, 0x0fff0000); + NV17_WRITE_CTX(0x00400eb0, 0x0fff0000); + NV17_WRITE_CTX(0x00400ec0, 0x00000080); + NV17_WRITE_CTX(0x00400ed0, 0x00000080); + } + NV_WRITE_CTX(NV10_PGRAPH_CTX_USER, chan->chid << 24); + + nv10_gr_create_pipe(chan); + + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = chan; + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +void +nv10_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) +{ + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; + unsigned long flags; + + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); + + nvkm_wr32(device, NV10_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV10_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV10_PGRAPH_TILE(i), tile->addr); + + nvkm_fifo_start(fifo, &flags); +} + +const struct nvkm_bitfield nv10_gr_intr_name[] = { + { NV_PGRAPH_INTR_NOTIFY, "NOTIFY" }, + { NV_PGRAPH_INTR_ERROR, "ERROR" }, + {} +}; + +const struct nvkm_bitfield nv10_gr_nstatus[] = { + { NV10_PGRAPH_NSTATUS_STATE_IN_USE, "STATE_IN_USE" }, + { NV10_PGRAPH_NSTATUS_INVALID_STATE, "INVALID_STATE" }, + { NV10_PGRAPH_NSTATUS_BAD_ARGUMENT, "BAD_ARGUMENT" }, + { NV10_PGRAPH_NSTATUS_PROTECTION_FAULT, "PROTECTION_FAULT" }, + {} +}; + +void +nv10_gr_intr(struct nvkm_gr *base) +{ + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); + u32 chid = (addr & 0x01f00000) >> 20; + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xfff; + u32 show = stat; + char msg[128], src[128], sta[128]; + struct nv10_gr_chan *chan; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + chan = gr->chan[chid]; + + if (stat & NV_PGRAPH_INTR_ERROR) { + if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) { + if (!nv10_gr_mthd(chan, class, mthd, data)) + show &= ~NV_PGRAPH_INTR_ERROR; + } + } + + if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) { + nvkm_wr32(device, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); + stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; + show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; + nv10_gr_context_switch(gr); + } + + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); + + if (show) { + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->object.client->name : "unknown", + subc, class, mthd, data); + } + + spin_unlock_irqrestore(&gr->lock, flags); +} + +int +nv10_gr_init(struct nvkm_gr *base) +{ + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x00118700); + /* nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x25f92ad9); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31)); + + if (device->card_type >= NV_11 && device->chipset >= 0x17) { + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x1f000000); + nvkm_wr32(device, 0x400a10, 0x03ff3fb6); + nvkm_wr32(device, 0x400838, 0x002f8684); + nvkm_wr32(device, 0x40083c, 0x00115f3f); + nvkm_wr32(device, 0x4006b0, 0x40000020); + } else { + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00000000); + } + + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_STATE, 0xFFFFFFFF); + + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, 0x08000000); + return 0; +} + +int +nv10_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct nv10_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, type, inst, true, &gr->base); +} + +static const struct nvkm_gr_func +nv10_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0056, &nv04_gr_object }, /* celcius */ + {} + } +}; + +int +nv10_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv10_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h new file mode 100644 index 0000000000..b86090c080 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h @@ -0,0 +1,14 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV10_GR_H__ +#define __NV10_GR_H__ +#include "priv.h" + +int nv10_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_gr **); +int nv10_gr_init(struct nvkm_gr *); +void nv10_gr_intr(struct nvkm_gr *); +void nv10_gr_tile(struct nvkm_gr *, int, struct nvkm_fb_tile *); + +int nv10_gr_chan_new(struct nvkm_gr *, struct nvkm_chan *, + const struct nvkm_oclass *, struct nvkm_object **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c new file mode 100644 index 0000000000..69ece259df --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c @@ -0,0 +1,59 @@ +/* + * Copyright 2007 Matthieu CASTET + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "nv10.h" + +static const struct nvkm_gr_func +nv15_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + {} + } +}; + +int +nv15_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv15_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c new file mode 100644 index 0000000000..e39dfc7d40 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c @@ -0,0 +1,59 @@ +/* + * Copyright 2007 Matthieu CASTET + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "nv10.h" + +static const struct nvkm_gr_func +nv17_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0099, &nv04_gr_object }, + {} + } +}; + +int +nv17_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv17_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c new file mode 100644 index 0000000000..02a8c62a0a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c @@ -0,0 +1,376 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +int +nv20_gr_chan_init(struct nvkm_object *object) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + struct nv20_gr *gr = chan->gr; + u32 inst = nvkm_memory_addr(chan->inst); + + nvkm_kmap(gr->ctxtab); + nvkm_wo32(gr->ctxtab, chan->chid * 4, inst >> 4); + nvkm_done(gr->ctxtab); + return 0; +} + +int +nv20_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + struct nv20_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 inst = nvkm_memory_addr(chan->inst); + int chid = -1; + + nvkm_mask(device, 0x400720, 0x00000001, 0x00000000); + if (nvkm_rd32(device, 0x400144) & 0x00010000) + chid = (nvkm_rd32(device, 0x400148) & 0x1f000000) >> 24; + if (chan->chid == chid) { + nvkm_wr32(device, 0x400784, inst >> 4); + nvkm_wr32(device, 0x400788, 0x00000002); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); + nvkm_wr32(device, 0x400144, 0x10000000); + nvkm_mask(device, 0x400148, 0xff000000, 0x1f000000); + } + nvkm_mask(device, 0x400720, 0x00000001, 0x00000001); + + nvkm_kmap(gr->ctxtab); + nvkm_wo32(gr->ctxtab, chan->chid * 4, 0x00000000); + nvkm_done(gr->ctxtab); + return 0; +} + +void * +nv20_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + nvkm_memory_unref(&chan->inst); + return chan; +} + +static const struct nvkm_object_func +nv20_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv20_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv20_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x37f0, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0000, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x033c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03a0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03a4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x047c, 0x00000101); + nvkm_wo32(chan->inst, 0x0490, 0x00000111); + nvkm_wo32(chan->inst, 0x04a8, 0x44400000); + for (i = 0x04d4; i <= 0x04e0; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x04f4; i <= 0x0500; i += 4) + nvkm_wo32(chan->inst, i, 0x00080000); + for (i = 0x050c; i <= 0x0518; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x051c; i <= 0x0528; i += 4) + nvkm_wo32(chan->inst, i, 0x000105b8); + for (i = 0x052c; i <= 0x0538; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + for (i = 0x055c; i <= 0x0598; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05a4, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x05fc, 0x00000001); + nvkm_wo32(chan->inst, 0x0604, 0x00004000); + nvkm_wo32(chan->inst, 0x0610, 0x00000001); + nvkm_wo32(chan->inst, 0x0618, 0x00040000); + nvkm_wo32(chan->inst, 0x061c, 0x00010000); + for (i = 0x1c1c; i <= 0x248c; i += 16) { + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); + } + nvkm_wo32(chan->inst, 0x281c, 0x3f800000); + nvkm_wo32(chan->inst, 0x2830, 0x3f800000); + nvkm_wo32(chan->inst, 0x285c, 0x40000000); + nvkm_wo32(chan->inst, 0x2860, 0x3f800000); + nvkm_wo32(chan->inst, 0x2864, 0x3f000000); + nvkm_wo32(chan->inst, 0x286c, 0x40000000); + nvkm_wo32(chan->inst, 0x2870, 0x3f800000); + nvkm_wo32(chan->inst, 0x2878, 0xbf800000); + nvkm_wo32(chan->inst, 0x2880, 0xbf800000); + nvkm_wo32(chan->inst, 0x34a4, 0x000fe000); + nvkm_wo32(chan->inst, 0x3530, 0x000003f8); + nvkm_wo32(chan->inst, 0x3540, 0x002fe000); + for (i = 0x355c; i <= 0x3578; i += 4) + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +void +nv20_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; + unsigned long flags; + + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); + + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->limit); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->pitch); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->addr); + + if (device->chipset != 0x34) { + nvkm_wr32(device, NV20_PGRAPH_ZCOMP(i), tile->zcomp); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00ea0090 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->zcomp); + } + + nvkm_fifo_start(fifo, &flags); +} + +void +nv20_gr_intr(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_chan *chan; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); + u32 chid = (addr & 0x01f00000) >> 20; + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xfff; + u32 show = stat; + char msg[128], src[128], sta[128]; + unsigned long flags; + + chan = nvkm_chan_get_chid(&gr->base.engine, chid, &flags); + + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); + + if (show) { + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->name : "unknown", + subc, class, mthd, data); + } + + nvkm_chan_put(&chan, flags); +} + +int +nv20_gr_oneinit(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + return nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 32 * 4, 16, + true, &gr->ctxtab); +} + +int +nv20_gr_init(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 tmp, vramsz; + int i; + + nvkm_wr32(device, NV20_PGRAPH_CHANNEL_CTX_TABLE, + nvkm_memory_addr(gr->ctxtab) >> 4); + + if (device->chipset == 0x20) { + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x003d0000); + for (i = 0; i < 15; i++) + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, 0x00000000); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); + } else { + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x02c80000); + for (i = 0; i < 32; i++) + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, 0x00000000); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); + } + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x00118700); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */ + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00000000); + nvkm_wr32(device, 0x40009C , 0x00000040); + + if (device->chipset >= 0x25) { + nvkm_wr32(device, 0x400890, 0x00a8cfff); + nvkm_wr32(device, 0x400610, 0x304B1FB6); + nvkm_wr32(device, 0x400B80, 0x1cbd3883); + nvkm_wr32(device, 0x400B84, 0x44000000); + nvkm_wr32(device, 0x400098, 0x40000080); + nvkm_wr32(device, 0x400B88, 0x000000ff); + + } else { + nvkm_wr32(device, 0x400880, 0x0008c7df); + nvkm_wr32(device, 0x400094, 0x00000005); + nvkm_wr32(device, 0x400B80, 0x45eae20e); + nvkm_wr32(device, 0x400B84, 0x24000000); + nvkm_wr32(device, 0x400098, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E00038); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000030); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E10038); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000030); + } + + nvkm_wr32(device, 0x4009a0, nvkm_rd32(device, 0x100324)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA000C); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, nvkm_rd32(device, 0x100324)); + + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); + + tmp = nvkm_rd32(device, NV10_PGRAPH_SURFACE) & 0x0007ff00; + nvkm_wr32(device, NV10_PGRAPH_SURFACE, tmp); + tmp = nvkm_rd32(device, NV10_PGRAPH_SURFACE) | 0x00020100; + nvkm_wr32(device, NV10_PGRAPH_SURFACE, tmp); + + /* begin RAM config */ + vramsz = device->func->resource_size(device, 1) - 1; + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400820, 0); + nvkm_wr32(device, 0x400824, 0); + nvkm_wr32(device, 0x400864, vramsz - 1); + nvkm_wr32(device, 0x400868, vramsz - 1); + + /* interesting.. the below overwrites some of the tile setup above.. */ + nvkm_wr32(device, 0x400B20, 0x00000000); + nvkm_wr32(device, 0x400B04, 0xFFFFFFFF); + + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_XMIN, 0); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_YMIN, 0); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff); + return 0; +} + +void * +nv20_gr_dtor(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + nvkm_memory_unref(&gr->ctxtab); + return gr; +} + +int +nv20_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct nv20_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, type, inst, true, &gr->base); +} + +static const struct nvkm_gr_func +nv20_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv20_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x0097, &nv04_gr_object }, /* kelvin */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + {} + } +}; + +int +nv20_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv20_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h new file mode 100644 index 0000000000..c0d2be5341 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h @@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV20_GR_H__ +#define __NV20_GR_H__ +#define nv20_gr(p) container_of((p), struct nv20_gr, base) +#include "priv.h" + +struct nv20_gr { + struct nvkm_gr base; + struct nvkm_memory *ctxtab; +}; + +int nv20_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_gr **); +void *nv20_gr_dtor(struct nvkm_gr *); +int nv20_gr_oneinit(struct nvkm_gr *); +int nv20_gr_init(struct nvkm_gr *); +void nv20_gr_intr(struct nvkm_gr *); +void nv20_gr_tile(struct nvkm_gr *, int, struct nvkm_fb_tile *); + +int nv30_gr_init(struct nvkm_gr *); + +#define nv20_gr_chan(p) container_of((p), struct nv20_gr_chan, object) +#include + +struct nv20_gr_chan { + struct nvkm_object object; + struct nv20_gr *gr; + int chid; + struct nvkm_memory *inst; +}; + +void *nv20_gr_chan_dtor(struct nvkm_object *); +int nv20_gr_chan_init(struct nvkm_object *); +int nv20_gr_chan_fini(struct nvkm_object *, bool); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c new file mode 100644 index 0000000000..d6bc6904dc --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c @@ -0,0 +1,135 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static const struct nvkm_object_func +nv25_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv25_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv25_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x3724, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x035c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03c0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03c4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x049c, 0x00000101); + nvkm_wo32(chan->inst, 0x04b0, 0x00000111); + nvkm_wo32(chan->inst, 0x04c8, 0x00000080); + nvkm_wo32(chan->inst, 0x04cc, 0xffff0000); + nvkm_wo32(chan->inst, 0x04d0, 0x00000001); + nvkm_wo32(chan->inst, 0x04e4, 0x44400000); + nvkm_wo32(chan->inst, 0x04fc, 0x4b800000); + for (i = 0x0510; i <= 0x051c; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x0530; i <= 0x053c; i += 4) + nvkm_wo32(chan->inst, i, 0x00080000); + for (i = 0x0548; i <= 0x0554; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x0558; i <= 0x0564; i += 4) + nvkm_wo32(chan->inst, i, 0x000105b8); + for (i = 0x0568; i <= 0x0574; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + for (i = 0x0598; i <= 0x05d4; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05e0, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x0620, 0x00000080); + nvkm_wo32(chan->inst, 0x0624, 0x30201000); + nvkm_wo32(chan->inst, 0x0628, 0x70605040); + nvkm_wo32(chan->inst, 0x062c, 0xb0a09080); + nvkm_wo32(chan->inst, 0x0630, 0xf0e0d0c0); + nvkm_wo32(chan->inst, 0x0664, 0x00000001); + nvkm_wo32(chan->inst, 0x066c, 0x00004000); + nvkm_wo32(chan->inst, 0x0678, 0x00000001); + nvkm_wo32(chan->inst, 0x0680, 0x00040000); + nvkm_wo32(chan->inst, 0x0684, 0x00010000); + for (i = 0x1b04; i <= 0x2374; i += 16) { + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); + } + nvkm_wo32(chan->inst, 0x2704, 0x3f800000); + nvkm_wo32(chan->inst, 0x2718, 0x3f800000); + nvkm_wo32(chan->inst, 0x2744, 0x40000000); + nvkm_wo32(chan->inst, 0x2748, 0x3f800000); + nvkm_wo32(chan->inst, 0x274c, 0x3f000000); + nvkm_wo32(chan->inst, 0x2754, 0x40000000); + nvkm_wo32(chan->inst, 0x2758, 0x3f800000); + nvkm_wo32(chan->inst, 0x2760, 0xbf800000); + nvkm_wo32(chan->inst, 0x2768, 0xbf800000); + nvkm_wo32(chan->inst, 0x308c, 0x000fe000); + nvkm_wo32(chan->inst, 0x3108, 0x000003f8); + nvkm_wo32(chan->inst, 0x3468, 0x002fe000); + for (i = 0x3484; i <= 0x34a0; i += 4) + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_gr_func +nv25_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv25_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; + +int +nv25_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv25_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c new file mode 100644 index 0000000000..e5a351b51e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c @@ -0,0 +1,126 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static const struct nvkm_object_func +nv2a_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv2a_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv2a_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x36b0, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0000, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x033c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03a0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03a4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x047c, 0x00000101); + nvkm_wo32(chan->inst, 0x0490, 0x00000111); + nvkm_wo32(chan->inst, 0x04a8, 0x44400000); + for (i = 0x04d4; i <= 0x04e0; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x04f4; i <= 0x0500; i += 4) + nvkm_wo32(chan->inst, i, 0x00080000); + for (i = 0x050c; i <= 0x0518; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x051c; i <= 0x0528; i += 4) + nvkm_wo32(chan->inst, i, 0x000105b8); + for (i = 0x052c; i <= 0x0538; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + for (i = 0x055c; i <= 0x0598; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05a4, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x05fc, 0x00000001); + nvkm_wo32(chan->inst, 0x0604, 0x00004000); + nvkm_wo32(chan->inst, 0x0610, 0x00000001); + nvkm_wo32(chan->inst, 0x0618, 0x00040000); + nvkm_wo32(chan->inst, 0x061c, 0x00010000); + for (i = 0x1a9c; i <= 0x22fc; i += 16) { /*XXX: check!! */ + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); + } + nvkm_wo32(chan->inst, 0x269c, 0x3f800000); + nvkm_wo32(chan->inst, 0x26b0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26dc, 0x40000000); + nvkm_wo32(chan->inst, 0x26e0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26e4, 0x3f000000); + nvkm_wo32(chan->inst, 0x26ec, 0x40000000); + nvkm_wo32(chan->inst, 0x26f0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26f8, 0xbf800000); + nvkm_wo32(chan->inst, 0x2700, 0xbf800000); + nvkm_wo32(chan->inst, 0x3024, 0x000fe000); + nvkm_wo32(chan->inst, 0x30a0, 0x000003f8); + nvkm_wo32(chan->inst, 0x33fc, 0x002fe000); + for (i = 0x341c; i <= 0x3438; i += 4) + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_gr_func +nv2a_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv2a_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; + +int +nv2a_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv2a_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c new file mode 100644 index 0000000000..8037032375 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c @@ -0,0 +1,200 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static const struct nvkm_object_func +nv30_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv30_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv30_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x5f48, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x0410, 0x00000101); + nvkm_wo32(chan->inst, 0x0424, 0x00000111); + nvkm_wo32(chan->inst, 0x0428, 0x00000060); + nvkm_wo32(chan->inst, 0x0444, 0x00000080); + nvkm_wo32(chan->inst, 0x0448, 0xffff0000); + nvkm_wo32(chan->inst, 0x044c, 0x00000001); + nvkm_wo32(chan->inst, 0x0460, 0x44400000); + nvkm_wo32(chan->inst, 0x048c, 0xffff0000); + for (i = 0x04e0; i < 0x04e8; i += 4) + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04ec, 0x00011100); + for (i = 0x0508; i < 0x0548; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x0550, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x058c, 0x00000080); + nvkm_wo32(chan->inst, 0x0590, 0x30201000); + nvkm_wo32(chan->inst, 0x0594, 0x70605040); + nvkm_wo32(chan->inst, 0x0598, 0xb8a89888); + nvkm_wo32(chan->inst, 0x059c, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05b0, 0xb0000000); + for (i = 0x0600; i < 0x0640; i += 4) + nvkm_wo32(chan->inst, i, 0x00010588); + for (i = 0x0640; i < 0x0680; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x06c0; i < 0x0700; i += 4) + nvkm_wo32(chan->inst, i, 0x0008aae4); + for (i = 0x0700; i < 0x0740; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x0740; i < 0x0780; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x085c, 0x00040000); + nvkm_wo32(chan->inst, 0x0860, 0x00010000); + for (i = 0x0864; i < 0x0874; i += 4) + nvkm_wo32(chan->inst, i, 0x00040004); + for (i = 0x1f18; i <= 0x3088 ; i += 16) { + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 4, 0x0436086c); + nvkm_wo32(chan->inst, i + 8, 0x000c001b); + } + for (i = 0x30b8; i < 0x30c8; i += 4) + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x344c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3808, 0x3f800000); + nvkm_wo32(chan->inst, 0x381c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3848, 0x40000000); + nvkm_wo32(chan->inst, 0x384c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3850, 0x3f000000); + nvkm_wo32(chan->inst, 0x3858, 0x40000000); + nvkm_wo32(chan->inst, 0x385c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3864, 0xbf800000); + nvkm_wo32(chan->inst, 0x386c, 0xbf800000); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +int +nv30_gr_init(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, NV20_PGRAPH_CHANNEL_CTX_TABLE, + nvkm_memory_addr(gr->ctxtab) >> 4); + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x401287c0); + nvkm_wr32(device, 0x400890, 0x01b463ff); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xf2de0475); + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00008000); + nvkm_wr32(device, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6); + nvkm_wr32(device, 0x400B80, 0x1003d888); + nvkm_wr32(device, 0x400B84, 0x0c000000); + nvkm_wr32(device, 0x400098, 0x00000000); + nvkm_wr32(device, 0x40009C, 0x0005ad00); + nvkm_wr32(device, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */ + nvkm_wr32(device, 0x4000a0, 0x00000000); + nvkm_wr32(device, 0x4000a4, 0x00000008); + nvkm_wr32(device, 0x4008a8, 0xb784a400); + nvkm_wr32(device, 0x400ba0, 0x002f8685); + nvkm_wr32(device, 0x400ba4, 0x00231f3f); + nvkm_wr32(device, 0x4008a4, 0x40000020); + + if (device->chipset == 0x34) { + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00200201); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0008); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000008); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000032); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E00004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000002); + } + + nvkm_wr32(device, 0x4000c0, 0x00000016); + + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); + nvkm_wr32(device, 0x0040075c , 0x00000001); + + /* begin RAM config */ + /* vramsz = pci_resource_len(gr->dev->pdev, 1) - 1; */ + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + if (device->chipset != 0x34) { + nvkm_wr32(device, 0x400750, 0x00EA0000); + nvkm_wr32(device, 0x400754, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x400750, 0x00EA0004); + nvkm_wr32(device, 0x400754, nvkm_rd32(device, 0x100204)); + } + + return 0; +} + +static const struct nvkm_gr_func +nv30_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv30_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0397, &nv04_gr_object }, /* rankine */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; + +int +nv30_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv30_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c new file mode 100644 index 0000000000..cdf043bbdd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c @@ -0,0 +1,137 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static const struct nvkm_object_func +nv34_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv34_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv34_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x46dc, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x040c, 0x01000101); + nvkm_wo32(chan->inst, 0x0420, 0x00000111); + nvkm_wo32(chan->inst, 0x0424, 0x00000060); + nvkm_wo32(chan->inst, 0x0440, 0x00000080); + nvkm_wo32(chan->inst, 0x0444, 0xffff0000); + nvkm_wo32(chan->inst, 0x0448, 0x00000001); + nvkm_wo32(chan->inst, 0x045c, 0x44400000); + nvkm_wo32(chan->inst, 0x0480, 0xffff0000); + for (i = 0x04d4; i < 0x04dc; i += 4) + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04e0, 0x00011100); + for (i = 0x04fc; i < 0x053c; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x0544, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x057c, 0x00000080); + nvkm_wo32(chan->inst, 0x0580, 0x30201000); + nvkm_wo32(chan->inst, 0x0584, 0x70605040); + nvkm_wo32(chan->inst, 0x0588, 0xb8a89888); + nvkm_wo32(chan->inst, 0x058c, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05a0, 0xb0000000); + for (i = 0x05f0; i < 0x0630; i += 4) + nvkm_wo32(chan->inst, i, 0x00010588); + for (i = 0x0630; i < 0x0670; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x06b0; i < 0x06f0; i += 4) + nvkm_wo32(chan->inst, i, 0x0008aae4); + for (i = 0x06f0; i < 0x0730; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x0730; i < 0x0770; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x0850, 0x00040000); + nvkm_wo32(chan->inst, 0x0854, 0x00010000); + for (i = 0x0858; i < 0x0868; i += 4) + nvkm_wo32(chan->inst, i, 0x00040004); + for (i = 0x15ac; i <= 0x271c ; i += 16) { + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 4, 0x0436086c); + nvkm_wo32(chan->inst, i + 8, 0x000c001b); + } + for (i = 0x274c; i < 0x275c; i += 4) + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x2ae0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2e9c, 0x3f800000); + nvkm_wo32(chan->inst, 0x2eb0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2edc, 0x40000000); + nvkm_wo32(chan->inst, 0x2ee0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2ee4, 0x3f000000); + nvkm_wo32(chan->inst, 0x2eec, 0x40000000); + nvkm_wo32(chan->inst, 0x2ef0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2ef8, 0xbf800000); + nvkm_wo32(chan->inst, 0x2f00, 0xbf800000); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_gr_func +nv34_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv34_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + { -1, -1, 0x0697, &nv04_gr_object }, /* rankine */ + {} + } +}; + +int +nv34_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv34_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c new file mode 100644 index 0000000000..fa5a6ccb87 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c @@ -0,0 +1,137 @@ +// SPDX-License-Identifier: MIT +#include "nv20.h" +#include "regs.h" + +#include +#include +#include + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static const struct nvkm_object_func +nv35_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + +static int +nv35_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv20_gr *gr = nv20_gr(base); + struct nv20_gr_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv35_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->id; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x577c, 16, true, + &chan->inst); + if (ret) + return ret; + + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x040c, 0x00000101); + nvkm_wo32(chan->inst, 0x0420, 0x00000111); + nvkm_wo32(chan->inst, 0x0424, 0x00000060); + nvkm_wo32(chan->inst, 0x0440, 0x00000080); + nvkm_wo32(chan->inst, 0x0444, 0xffff0000); + nvkm_wo32(chan->inst, 0x0448, 0x00000001); + nvkm_wo32(chan->inst, 0x045c, 0x44400000); + nvkm_wo32(chan->inst, 0x0488, 0xffff0000); + for (i = 0x04dc; i < 0x04e4; i += 4) + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04e8, 0x00011100); + for (i = 0x0504; i < 0x0544; i += 4) + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x054c, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x0588, 0x00000080); + nvkm_wo32(chan->inst, 0x058c, 0x30201000); + nvkm_wo32(chan->inst, 0x0590, 0x70605040); + nvkm_wo32(chan->inst, 0x0594, 0xb8a89888); + nvkm_wo32(chan->inst, 0x0598, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05ac, 0xb0000000); + for (i = 0x0604; i < 0x0644; i += 4) + nvkm_wo32(chan->inst, i, 0x00010588); + for (i = 0x0644; i < 0x0684; i += 4) + nvkm_wo32(chan->inst, i, 0x00030303); + for (i = 0x06c4; i < 0x0704; i += 4) + nvkm_wo32(chan->inst, i, 0x0008aae4); + for (i = 0x0704; i < 0x0744; i += 4) + nvkm_wo32(chan->inst, i, 0x01012000); + for (i = 0x0744; i < 0x0784; i += 4) + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x0860, 0x00040000); + nvkm_wo32(chan->inst, 0x0864, 0x00010000); + for (i = 0x0868; i < 0x0878; i += 4) + nvkm_wo32(chan->inst, i, 0x00040004); + for (i = 0x1f1c; i <= 0x308c ; i += 16) { + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 4, 0x0436086c); + nvkm_wo32(chan->inst, i + 8, 0x000c001b); + } + for (i = 0x30bc; i < 0x30cc; i += 4) + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x3450, 0x3f800000); + nvkm_wo32(chan->inst, 0x380c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3820, 0x3f800000); + nvkm_wo32(chan->inst, 0x384c, 0x40000000); + nvkm_wo32(chan->inst, 0x3850, 0x3f800000); + nvkm_wo32(chan->inst, 0x3854, 0x3f000000); + nvkm_wo32(chan->inst, 0x385c, 0x40000000); + nvkm_wo32(chan->inst, 0x3860, 0x3f800000); + nvkm_wo32(chan->inst, 0x3868, 0xbf800000); + nvkm_wo32(chan->inst, 0x3870, 0xbf800000); + nvkm_done(chan->inst); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_gr_func +nv35_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv35_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0497, &nv04_gr_object }, /* rankine */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; + +int +nv35_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv35_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c new file mode 100644 index 0000000000..a5e1f02791 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c @@ -0,0 +1,476 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" +#include "regs.h" + +#include +#include +#include +#include +#include + +u64 +nv40_gr_units(struct nvkm_gr *gr) +{ + return nvkm_rd32(gr->engine.subdev.device, 0x1540); +} + +/******************************************************************************* + * Graphics object classes + ******************************************************************************/ + +static int +nv40_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 20, align, + false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); +#ifdef __BIG_ENDIAN + nvkm_mo32(*pgpuobj, 0x08, 0x01000000, 0x01000000); +#endif + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_wo32(*pgpuobj, 0x10, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; +} + +const struct nvkm_object_func +nv40_gr_object = { + .bind = nv40_gr_object_bind, +}; + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static int +nv40_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nv40_gr_chan *chan = nv40_gr_chan(object); + struct nv40_gr *gr = chan->gr; + int ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, true, parent, pgpuobj); + if (ret == 0) { + chan->inst = (*pgpuobj)->addr; + nvkm_kmap(*pgpuobj); + nv40_grctx_fill(gr->base.engine.subdev.device, *pgpuobj); + nvkm_wo32(*pgpuobj, 0x00000, chan->inst >> 4); + nvkm_done(*pgpuobj); + } + return ret; +} + +static int +nv40_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv40_gr_chan *chan = nv40_gr_chan(object); + struct nv40_gr *gr = chan->gr; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 inst = 0x01000000 | chan->inst >> 4; + int ret = 0; + + nvkm_mask(device, 0x400720, 0x00000001, 0x00000000); + + if (nvkm_rd32(device, 0x40032c) == inst) { + if (suspend) { + nvkm_wr32(device, 0x400720, 0x00000000); + nvkm_wr32(device, 0x400784, inst); + nvkm_mask(device, 0x400310, 0x00000020, 0x00000020); + nvkm_mask(device, 0x400304, 0x00000001, 0x00000001); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x400300) & 0x00000001)) + break; + ) < 0) { + u32 insn = nvkm_rd32(device, 0x400308); + nvkm_warn(subdev, "ctxprog timeout %08x\n", insn); + ret = -EBUSY; + } + } + + nvkm_mask(device, 0x40032c, 0x01000000, 0x00000000); + } + + if (nvkm_rd32(device, 0x400330) == inst) + nvkm_mask(device, 0x400330, 0x01000000, 0x00000000); + + nvkm_mask(device, 0x400720, 0x00000001, 0x00000001); + return ret; +} + +static void * +nv40_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv40_gr_chan *chan = nv40_gr_chan(object); + unsigned long flags; + spin_lock_irqsave(&chan->gr->base.engine.lock, flags); + list_del(&chan->head); + spin_unlock_irqrestore(&chan->gr->base.engine.lock, flags); + return chan; +} + +static const struct nvkm_object_func +nv40_gr_chan = { + .dtor = nv40_gr_chan_dtor, + .fini = nv40_gr_chan_fini, + .bind = nv40_gr_chan_bind, +}; + +int +nv40_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nv40_gr_chan *chan; + unsigned long flags; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv40_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->fifo = fifoch; + *pobject = &chan->object; + + spin_lock_irqsave(&chan->gr->base.engine.lock, flags); + list_add(&chan->head, &gr->chan); + spin_unlock_irqrestore(&chan->gr->base.engine.lock, flags); + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static void +nv40_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; + unsigned long flags; + + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); + + switch (device->chipset) { + case 0x40: + case 0x41: + case 0x42: + case 0x43: + case 0x45: + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + switch (device->chipset) { + case 0x40: + case 0x45: + nvkm_wr32(device, NV20_PGRAPH_ZCOMP(i), tile->zcomp); + nvkm_wr32(device, NV40_PGRAPH_ZCOMP1(i), tile->zcomp); + break; + case 0x41: + case 0x42: + case 0x43: + nvkm_wr32(device, NV41_PGRAPH_ZCOMP0(i), tile->zcomp); + nvkm_wr32(device, NV41_PGRAPH_ZCOMP1(i), tile->zcomp); + break; + default: + break; + } + break; + case 0x47: + case 0x49: + case 0x4b: + nvkm_wr32(device, NV47_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV47_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV47_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + nvkm_wr32(device, NV47_PGRAPH_ZCOMP0(i), tile->zcomp); + nvkm_wr32(device, NV47_PGRAPH_ZCOMP1(i), tile->zcomp); + break; + default: + WARN_ON(1); + break; + } + + nvkm_fifo_start(fifo, &flags); +} + +void +nv40_gr_intr(struct nvkm_gr *base) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nv40_gr_chan *temp, *chan = NULL; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 inst = nvkm_rd32(device, 0x40032c) & 0x000fffff; + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xffff; + u32 show = stat; + char msg[128], src[128], sta[128]; + unsigned long flags; + + spin_lock_irqsave(&gr->base.engine.lock, flags); + list_for_each_entry(temp, &gr->chan, head) { + if (temp->inst >> 4 == inst) { + chan = temp; + list_del(&chan->head); + list_add(&chan->head, &gr->chan); + break; + } + } + + if (stat & NV_PGRAPH_INTR_ERROR) { + if (nsource & NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION) { + nvkm_mask(device, 0x402000, 0, 0); + } + } + + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); + + if (show) { + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%08x %s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, + chan ? chan->fifo->id : -1, inst << 4, + chan ? chan->fifo->name : "unknown", + subc, class, mthd, data); + } + + spin_unlock_irqrestore(&gr->base.engine.lock, flags); +} + +int +nv40_gr_init(struct nvkm_gr *base) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret, i, j; + u32 vramsz; + + /* generate and upload context program */ + ret = nv40_grctx_init(device, &gr->size); + if (ret) + return ret; + + /* No context present currently */ + nvkm_wr32(device, NV40_PGRAPH_CTXCTL_CUR, 0x00000000); + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV40_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x401287c0); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xe0de8055); + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00008000); + nvkm_wr32(device, NV04_PGRAPH_LIMIT_VIOL_PIX, 0x00be3c5f); + + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); + + j = nvkm_rd32(device, 0x1540) & 0xff; + if (j) { + for (i = 0; !(j & 1); j >>= 1, i++) + ; + nvkm_wr32(device, 0x405000, i); + } + + if (device->chipset == 0x40) { + nvkm_wr32(device, 0x4009b0, 0x83280fff); + nvkm_wr32(device, 0x4009b4, 0x000000a0); + } else { + nvkm_wr32(device, 0x400820, 0x83280eff); + nvkm_wr32(device, 0x400824, 0x000000a0); + } + + switch (device->chipset) { + case 0x40: + case 0x45: + nvkm_wr32(device, 0x4009b8, 0x0078e366); + nvkm_wr32(device, 0x4009bc, 0x0000014c); + break; + case 0x41: + case 0x42: /* pciid also 0x00Cx */ + /* case 0x0120: XXX (pciid) */ + nvkm_wr32(device, 0x400828, 0x007596ff); + nvkm_wr32(device, 0x40082c, 0x00000108); + break; + case 0x43: + nvkm_wr32(device, 0x400828, 0x0072cb77); + nvkm_wr32(device, 0x40082c, 0x00000108); + break; + case 0x44: + case 0x46: /* G72 */ + case 0x4a: + case 0x4c: /* G7x-based C51 */ + case 0x4e: + nvkm_wr32(device, 0x400860, 0); + nvkm_wr32(device, 0x400864, 0); + break; + case 0x47: /* G70 */ + case 0x49: /* G71 */ + case 0x4b: /* G73 */ + nvkm_wr32(device, 0x400828, 0x07830610); + nvkm_wr32(device, 0x40082c, 0x0000016A); + break; + default: + break; + } + + nvkm_wr32(device, 0x400b38, 0x2ffff800); + nvkm_wr32(device, 0x400b3c, 0x00006000); + + /* Tiling related stuff. */ + switch (device->chipset) { + case 0x44: + case 0x4a: + nvkm_wr32(device, 0x400bc4, 0x1003d888); + nvkm_wr32(device, 0x400bbc, 0xb7a7b500); + break; + case 0x46: + nvkm_wr32(device, 0x400bc4, 0x0000e024); + nvkm_wr32(device, 0x400bbc, 0xb7a7b520); + break; + case 0x4c: + case 0x4e: + case 0x67: + nvkm_wr32(device, 0x400bc4, 0x1003d888); + nvkm_wr32(device, 0x400bbc, 0xb7a7b540); + break; + default: + break; + } + + /* begin RAM config */ + vramsz = device->func->resource_size(device, 1) - 1; + switch (device->chipset) { + case 0x40: + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x4069A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4069A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400820, 0); + nvkm_wr32(device, 0x400824, 0); + nvkm_wr32(device, 0x400864, vramsz); + nvkm_wr32(device, 0x400868, vramsz); + break; + default: + switch (device->chipset) { + case 0x41: + case 0x42: + case 0x43: + case 0x45: + case 0x4e: + case 0x44: + case 0x4a: + nvkm_wr32(device, 0x4009F0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009F4, nvkm_rd32(device, 0x100204)); + break; + default: + nvkm_wr32(device, 0x400DF0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x400DF4, nvkm_rd32(device, 0x100204)); + break; + } + nvkm_wr32(device, 0x4069F0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4069F4, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400840, 0); + nvkm_wr32(device, 0x400844, 0); + nvkm_wr32(device, 0x4008A0, vramsz); + nvkm_wr32(device, 0x4008A4, vramsz); + break; + } + + return 0; +} + +int +nv40_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct nv40_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + INIT_LIST_HEAD(&gr->chan); + + return nvkm_gr_ctor(func, device, type, inst, true, &gr->base); +} + +static const struct nvkm_gr_func +nv40_gr = { + .init = nv40_gr_init, + .intr = nv40_gr_intr, + .tile = nv40_gr_tile, + .units = nv40_gr_units, + .chan_new = nv40_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv40_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv40_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv40_gr_object }, /* null */ + { -1, -1, 0x0039, &nv40_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv40_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv40_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv40_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv40_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv40_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv40_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv40_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv40_gr_object }, /* imageblit */ + { -1, -1, 0x3062, &nv40_gr_object }, /* surf2d (nv40) */ + { -1, -1, 0x3089, &nv40_gr_object }, /* sifm (nv40) */ + { -1, -1, 0x309e, &nv40_gr_object }, /* swzsurf (nv40) */ + { -1, -1, 0x4097, &nv40_gr_object }, /* curie */ + {} + } +}; + +int +nv40_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv40_gr_new_(&nv40_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h new file mode 100644 index 0000000000..84fbc99139 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h @@ -0,0 +1,49 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV40_GR_H__ +#define __NV40_GR_H__ +#define nv40_gr(p) container_of((p), struct nv40_gr, base) +#include "priv.h" + +struct nv40_gr { + struct nvkm_gr base; + u32 size; + struct list_head chan; +}; + +int nv40_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_gr **); +int nv40_gr_init(struct nvkm_gr *); +void nv40_gr_intr(struct nvkm_gr *); +u64 nv40_gr_units(struct nvkm_gr *); + +#define nv40_gr_chan(p) container_of((p), struct nv40_gr_chan, object) +#include + +struct nv40_gr_chan { + struct nvkm_object object; + struct nv40_gr *gr; + struct nvkm_chan *fifo; + u32 inst; + struct list_head head; +}; + +int nv40_gr_chan_new(struct nvkm_gr *, struct nvkm_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + +extern const struct nvkm_object_func nv40_gr_object; + +/* returns 1 if device is one of the nv4x using the 0x4497 object class, + * helpful to determine a number of other hardware features + */ +static inline int +nv44_gr_class(struct nvkm_device *device) +{ + if ((device->chipset & 0xf0) == 0x60) + return 1; + + return !(0x0aaf & (1 << (device->chipset & 0x0f))); +} + +int nv40_grctx_init(struct nvkm_device *, u32 *size); +void nv40_grctx_fill(struct nvkm_device *, struct nvkm_gpuobj *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c new file mode 100644 index 0000000000..22b6a38a70 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c @@ -0,0 +1,108 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" +#include "regs.h" + +#include +#include + +static void +nv44_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; + unsigned long flags; + + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); + + switch (device->chipset) { + case 0x44: + case 0x4a: + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + break; + case 0x46: + case 0x4c: + case 0x63: + case 0x67: + case 0x68: + nvkm_wr32(device, NV47_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV47_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV47_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + break; + case 0x4e: + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + break; + default: + WARN_ON(1); + break; + } + + nvkm_fifo_start(fifo, &flags); +} + +static const struct nvkm_gr_func +nv44_gr = { + .init = nv40_gr_init, + .intr = nv40_gr_intr, + .tile = nv44_gr_tile, + .units = nv40_gr_units, + .chan_new = nv40_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv40_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv40_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv40_gr_object }, /* null */ + { -1, -1, 0x0039, &nv40_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv40_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv40_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv40_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv40_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv40_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv40_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv40_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv40_gr_object }, /* imageblit */ + { -1, -1, 0x3062, &nv40_gr_object }, /* surf2d (nv40) */ + { -1, -1, 0x3089, &nv40_gr_object }, /* sifm (nv40) */ + { -1, -1, 0x309e, &nv40_gr_object }, /* swzsurf (nv40) */ + { -1, -1, 0x4497, &nv40_gr_object }, /* curie */ + {} + } +}; + +int +nv44_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv40_gr_new_(&nv44_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c new file mode 100644 index 0000000000..c8a0288c09 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c @@ -0,0 +1,796 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include +#include +#include + +#include + +u64 +nv50_gr_units(struct nvkm_gr *gr) +{ + return nvkm_rd32(gr->engine.subdev.device, 0x1540); +} + +/******************************************************************************* + * Graphics object classes + ******************************************************************************/ + +static int +nv50_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, + align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; +} + +const struct nvkm_object_func +nv50_gr_object = { + .bind = nv50_gr_object_bind, +}; + +/******************************************************************************* + * PGRAPH context + ******************************************************************************/ + +static int +nv50_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nv50_gr *gr = nv50_gr_chan(object)->gr; + int ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, true, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nv50_grctx_fill(gr->base.engine.subdev.device, *pgpuobj); + nvkm_done(*pgpuobj); + } + return ret; +} + +static const struct nvkm_object_func +nv50_gr_chan = { + .bind = nv50_gr_chan_bind, +}; + +int +nv50_gr_chan_new(struct nvkm_gr *base, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv50_gr *gr = nv50_gr(base); + struct nv50_gr_chan *chan; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv50_gr_chan, oclass, &chan->object); + chan->gr = gr; + *pobject = &chan->object; + return 0; +} + +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_bitfield nv50_mp_exec_errors[] = { + { 0x01, "STACK_UNDERFLOW" }, + { 0x02, "STACK_MISMATCH" }, + { 0x04, "QUADON_ACTIVE" }, + { 0x08, "TIMEOUT" }, + { 0x10, "INVALID_OPCODE" }, + { 0x20, "PM_OVERFLOW" }, + { 0x40, "BREAKPOINT" }, + {} +}; + +static const struct nvkm_bitfield nv50_mpc_traps[] = { + { 0x0000001, "LOCAL_LIMIT_READ" }, + { 0x0000010, "LOCAL_LIMIT_WRITE" }, + { 0x0000040, "STACK_LIMIT" }, + { 0x0000100, "GLOBAL_LIMIT_READ" }, + { 0x0001000, "GLOBAL_LIMIT_WRITE" }, + { 0x0010000, "MP0" }, + { 0x0020000, "MP1" }, + { 0x0040000, "GLOBAL_LIMIT_RED" }, + { 0x0400000, "GLOBAL_LIMIT_ATOM" }, + { 0x4000000, "MP2" }, + {} +}; + +static const struct nvkm_bitfield nv50_tex_traps[] = { + { 0x00000001, "" }, /* any bit set? */ + { 0x00000002, "FAULT" }, + { 0x00000004, "STORAGE_TYPE_MISMATCH" }, + { 0x00000008, "LINEAR_MISMATCH" }, + { 0x00000020, "WRONG_MEMTYPE" }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_trap_m2mf[] = { + { 0x00000001, "NOTIFY" }, + { 0x00000002, "IN" }, + { 0x00000004, "OUT" }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_trap_vfetch[] = { + { 0x00000001, "FAULT" }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_trap_strmout[] = { + { 0x00000001, "FAULT" }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_trap_ccache[] = { + { 0x00000001, "FAULT" }, + {} +}; + +/* There must be a *lot* of these. Will take some time to gather them up. */ +const struct nvkm_enum nv50_data_error_names[] = { + { 0x00000003, "INVALID_OPERATION", NULL }, + { 0x00000004, "INVALID_VALUE", NULL }, + { 0x00000005, "INVALID_ENUM", NULL }, + { 0x00000008, "INVALID_OBJECT", NULL }, + { 0x00000009, "READ_ONLY_OBJECT", NULL }, + { 0x0000000a, "SUPERVISOR_OBJECT", NULL }, + { 0x0000000b, "INVALID_ADDRESS_ALIGNMENT", NULL }, + { 0x0000000c, "INVALID_BITFIELD", NULL }, + { 0x0000000d, "BEGIN_END_ACTIVE", NULL }, + { 0x0000000e, "SEMANTIC_COLOR_BACK_OVER_LIMIT", NULL }, + { 0x0000000f, "VIEWPORT_ID_NEEDS_GP", NULL }, + { 0x00000010, "RT_DOUBLE_BIND", NULL }, + { 0x00000011, "RT_TYPES_MISMATCH", NULL }, + { 0x00000012, "RT_LINEAR_WITH_ZETA", NULL }, + { 0x00000015, "FP_TOO_FEW_REGS", NULL }, + { 0x00000016, "ZETA_FORMAT_CSAA_MISMATCH", NULL }, + { 0x00000017, "RT_LINEAR_WITH_MSAA", NULL }, + { 0x00000018, "FP_INTERPOLANT_START_OVER_LIMIT", NULL }, + { 0x00000019, "SEMANTIC_LAYER_OVER_LIMIT", NULL }, + { 0x0000001a, "RT_INVALID_ALIGNMENT", NULL }, + { 0x0000001b, "SAMPLER_OVER_LIMIT", NULL }, + { 0x0000001c, "TEXTURE_OVER_LIMIT", NULL }, + { 0x0000001e, "GP_TOO_MANY_OUTPUTS", NULL }, + { 0x0000001f, "RT_BPP128_WITH_MS8", NULL }, + { 0x00000021, "Z_OUT_OF_BOUNDS", NULL }, + { 0x00000023, "XY_OUT_OF_BOUNDS", NULL }, + { 0x00000024, "VP_ZERO_INPUTS", NULL }, + { 0x00000027, "CP_MORE_PARAMS_THAN_SHARED", NULL }, + { 0x00000028, "CP_NO_REG_SPACE_STRIPED", NULL }, + { 0x00000029, "CP_NO_REG_SPACE_PACKED", NULL }, + { 0x0000002a, "CP_NOT_ENOUGH_WARPS", NULL }, + { 0x0000002b, "CP_BLOCK_SIZE_MISMATCH", NULL }, + { 0x0000002c, "CP_NOT_ENOUGH_LOCAL_WARPS", NULL }, + { 0x0000002d, "CP_NOT_ENOUGH_STACK_WARPS", NULL }, + { 0x0000002e, "CP_NO_BLOCKDIM_LATCH", NULL }, + { 0x00000031, "ENG2D_FORMAT_MISMATCH", NULL }, + { 0x0000003f, "PRIMITIVE_ID_NEEDS_GP", NULL }, + { 0x00000044, "SEMANTIC_VIEWPORT_OVER_LIMIT", NULL }, + { 0x00000045, "SEMANTIC_COLOR_FRONT_OVER_LIMIT", NULL }, + { 0x00000046, "LAYER_ID_NEEDS_GP", NULL }, + { 0x00000047, "SEMANTIC_CLIP_OVER_LIMIT", NULL }, + { 0x00000048, "SEMANTIC_PTSZ_OVER_LIMIT", NULL }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_intr_name[] = { + { 0x00000001, "NOTIFY" }, + { 0x00000002, "COMPUTE_QUERY" }, + { 0x00000010, "ILLEGAL_MTHD" }, + { 0x00000020, "ILLEGAL_CLASS" }, + { 0x00000040, "DOUBLE_NOTIFY" }, + { 0x00001000, "CONTEXT_SWITCH" }, + { 0x00010000, "BUFFER_NOTIFY" }, + { 0x00100000, "DATA_ERROR" }, + { 0x00200000, "TRAP" }, + { 0x01000000, "SINGLE_STEP" }, + {} +}; + +static const struct nvkm_bitfield nv50_gr_trap_prop[] = { + { 0x00000004, "SURF_WIDTH_OVERRUN" }, + { 0x00000008, "SURF_HEIGHT_OVERRUN" }, + { 0x00000010, "DST2D_FAULT" }, + { 0x00000020, "ZETA_FAULT" }, + { 0x00000040, "RT_FAULT" }, + { 0x00000080, "CUDA_FAULT" }, + { 0x00000100, "DST2D_STORAGE_TYPE_MISMATCH" }, + { 0x00000200, "ZETA_STORAGE_TYPE_MISMATCH" }, + { 0x00000400, "RT_STORAGE_TYPE_MISMATCH" }, + { 0x00000800, "DST2D_LINEAR_MISMATCH" }, + { 0x00001000, "RT_LINEAR_MISMATCH" }, + {} +}; + +static void +nv50_gr_prop_trap(struct nv50_gr *gr, u32 ustatus_addr, u32 ustatus, u32 tp) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 e0c = nvkm_rd32(device, ustatus_addr + 0x04); + u32 e10 = nvkm_rd32(device, ustatus_addr + 0x08); + u32 e14 = nvkm_rd32(device, ustatus_addr + 0x0c); + u32 e18 = nvkm_rd32(device, ustatus_addr + 0x10); + u32 e1c = nvkm_rd32(device, ustatus_addr + 0x14); + u32 e20 = nvkm_rd32(device, ustatus_addr + 0x18); + u32 e24 = nvkm_rd32(device, ustatus_addr + 0x1c); + char msg[128]; + + /* CUDA memory: l[], g[] or stack. */ + if (ustatus & 0x00000080) { + if (e18 & 0x80000000) { + /* g[] read fault? */ + nvkm_error(subdev, "TRAP_PROP - TP %d - CUDA_FAULT - Global read fault at address %02x%08x\n", + tp, e14, e10 | ((e18 >> 24) & 0x1f)); + e18 &= ~0x1f000000; + } else if (e18 & 0xc) { + /* g[] write fault? */ + nvkm_error(subdev, "TRAP_PROP - TP %d - CUDA_FAULT - Global write fault at address %02x%08x\n", + tp, e14, e10 | ((e18 >> 7) & 0x1f)); + e18 &= ~0x00000f80; + } else { + nvkm_error(subdev, "TRAP_PROP - TP %d - Unknown CUDA fault at address %02x%08x\n", + tp, e14, e10); + } + ustatus &= ~0x00000080; + } + if (ustatus) { + nvkm_snprintbf(msg, sizeof(msg), nv50_gr_trap_prop, ustatus); + nvkm_error(subdev, "TRAP_PROP - TP %d - %08x [%s] - " + "Address %02x%08x\n", + tp, ustatus, msg, e14, e10); + } + nvkm_error(subdev, "TRAP_PROP - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", + tp, e0c, e18, e1c, e20, e24); +} + +static void +nv50_gr_mp_trap(struct nv50_gr *gr, int tpid, int display) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 units = nvkm_rd32(device, 0x1540); + u32 addr, mp10, status, pc, oplow, ophigh; + char msg[128]; + int i; + int mps = 0; + for (i = 0; i < 4; i++) { + if (!(units & 1 << (i+24))) + continue; + if (device->chipset < 0xa0) + addr = 0x408200 + (tpid << 12) + (i << 7); + else + addr = 0x408100 + (tpid << 11) + (i << 7); + mp10 = nvkm_rd32(device, addr + 0x10); + status = nvkm_rd32(device, addr + 0x14); + if (!status) + continue; + if (display) { + nvkm_rd32(device, addr + 0x20); + pc = nvkm_rd32(device, addr + 0x24); + oplow = nvkm_rd32(device, addr + 0x70); + ophigh = nvkm_rd32(device, addr + 0x74); + nvkm_snprintbf(msg, sizeof(msg), + nv50_mp_exec_errors, status); + nvkm_error(subdev, "TRAP_MP_EXEC - TP %d MP %d: " + "%08x [%s] at %06x warp %d, " + "opcode %08x %08x\n", + tpid, i, status, msg, pc & 0xffffff, + pc >> 24, oplow, ophigh); + } + nvkm_wr32(device, addr + 0x10, mp10); + nvkm_wr32(device, addr + 0x14, 0); + mps++; + } + if (!mps && display) + nvkm_error(subdev, "TRAP_MP_EXEC - TP %d: " + "No MPs claiming errors?\n", tpid); +} + +static void +nv50_gr_tp_trap(struct nv50_gr *gr, int type, u32 ustatus_old, + u32 ustatus_new, int display, const char *name) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 units = nvkm_rd32(device, 0x1540); + int tps = 0; + int i, r; + char msg[128]; + u32 ustatus_addr, ustatus; + for (i = 0; i < 16; i++) { + if (!(units & (1 << i))) + continue; + if (device->chipset < 0xa0) + ustatus_addr = ustatus_old + (i << 12); + else + ustatus_addr = ustatus_new + (i << 11); + ustatus = nvkm_rd32(device, ustatus_addr) & 0x7fffffff; + if (!ustatus) + continue; + tps++; + switch (type) { + case 6: /* texture error... unknown for now */ + if (display) { + nvkm_error(subdev, "magic set %d:\n", i); + for (r = ustatus_addr + 4; r <= ustatus_addr + 0x10; r += 4) + nvkm_error(subdev, "\t%08x: %08x\n", r, + nvkm_rd32(device, r)); + if (ustatus) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_tex_traps, ustatus); + nvkm_error(subdev, + "%s - TP%d: %08x [%s]\n", + name, i, ustatus, msg); + ustatus = 0; + } + } + break; + case 7: /* MP error */ + if (ustatus & 0x04030000) { + nv50_gr_mp_trap(gr, i, display); + ustatus &= ~0x04030000; + } + if (ustatus && display) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_mpc_traps, ustatus); + nvkm_error(subdev, "%s - TP%d: %08x [%s]\n", + name, i, ustatus, msg); + ustatus = 0; + } + break; + case 8: /* PROP error */ + if (display) + nv50_gr_prop_trap( + gr, ustatus_addr, ustatus, i); + ustatus = 0; + break; + } + if (ustatus) { + if (display) + nvkm_error(subdev, "%s - TP%d: Unhandled ustatus %08x\n", name, i, ustatus); + } + nvkm_wr32(device, ustatus_addr, 0xc0000000); + } + + if (!tps && display) + nvkm_warn(subdev, "%s - No TPs claiming errors?\n", name); +} + +static int +nv50_gr_trap_handler(struct nv50_gr *gr, u32 display, + int chid, u64 inst, const char *name) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 status = nvkm_rd32(device, 0x400108); + u32 ustatus; + char msg[128]; + + if (!status && display) { + nvkm_error(subdev, "TRAP: no units reporting traps?\n"); + return 1; + } + + /* DISPATCH: Relays commands to other units and handles NOTIFY, + * COND, QUERY. If you get a trap from it, the command is still stuck + * in DISPATCH and you need to do something about it. */ + if (status & 0x001) { + ustatus = nvkm_rd32(device, 0x400804) & 0x7fffffff; + if (!ustatus && display) { + nvkm_error(subdev, "TRAP_DISPATCH - no ustatus?\n"); + } + + nvkm_wr32(device, 0x400500, 0x00000000); + + /* Known to be triggered by screwed up NOTIFY and COND... */ + if (ustatus & 0x00000001) { + u32 addr = nvkm_rd32(device, 0x400808); + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 datal = nvkm_rd32(device, 0x40080c); + u32 datah = nvkm_rd32(device, 0x400810); + u32 class = nvkm_rd32(device, 0x400814); + u32 r848 = nvkm_rd32(device, 0x400848); + + nvkm_error(subdev, "TRAP DISPATCH_FAULT\n"); + if (display && (addr & 0x80000000)) { + nvkm_error(subdev, + "ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x%08x " + "400808 %08x 400848 %08x\n", + chid, inst, name, subc, class, mthd, + datah, datal, addr, r848); + } else + if (display) { + nvkm_error(subdev, "no stuck command?\n"); + } + + nvkm_wr32(device, 0x400808, 0); + nvkm_wr32(device, 0x4008e8, nvkm_rd32(device, 0x4008e8) & 3); + nvkm_wr32(device, 0x400848, 0); + ustatus &= ~0x00000001; + } + + if (ustatus & 0x00000002) { + u32 addr = nvkm_rd32(device, 0x40084c); + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, 0x40085c); + u32 class = nvkm_rd32(device, 0x400814); + + nvkm_error(subdev, "TRAP DISPATCH_QUERY\n"); + if (display && (addr & 0x80000000)) { + nvkm_error(subdev, + "ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x " + "40084c %08x\n", chid, inst, name, + subc, class, mthd, data, addr); + } else + if (display) { + nvkm_error(subdev, "no stuck command?\n"); + } + + nvkm_wr32(device, 0x40084c, 0); + ustatus &= ~0x00000002; + } + + if (ustatus && display) { + nvkm_error(subdev, "TRAP_DISPATCH " + "(unknown %08x)\n", ustatus); + } + + nvkm_wr32(device, 0x400804, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x001); + status &= ~0x001; + if (!status) + return 0; + } + + /* M2MF: Memory to memory copy engine. */ + if (status & 0x002) { + u32 ustatus = nvkm_rd32(device, 0x406800) & 0x7fffffff; + if (display) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_m2mf, ustatus); + nvkm_error(subdev, "TRAP_M2MF %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_M2MF %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x406804), + nvkm_rd32(device, 0x406808), + nvkm_rd32(device, 0x40680c), + nvkm_rd32(device, 0x406810)); + } + + /* No sane way found yet -- just reset the bugger. */ + nvkm_wr32(device, 0x400040, 2); + nvkm_wr32(device, 0x400040, 0); + nvkm_wr32(device, 0x406800, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x002); + status &= ~0x002; + } + + /* VFETCH: Fetches data from vertex buffers. */ + if (status & 0x004) { + u32 ustatus = nvkm_rd32(device, 0x400c04) & 0x7fffffff; + if (display) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_vfetch, ustatus); + nvkm_error(subdev, "TRAP_VFETCH %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_VFETCH %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x400c00), + nvkm_rd32(device, 0x400c08), + nvkm_rd32(device, 0x400c0c), + nvkm_rd32(device, 0x400c10)); + } + + nvkm_wr32(device, 0x400c04, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x004); + status &= ~0x004; + } + + /* STRMOUT: DirectX streamout / OpenGL transform feedback. */ + if (status & 0x008) { + ustatus = nvkm_rd32(device, 0x401800) & 0x7fffffff; + if (display) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_strmout, ustatus); + nvkm_error(subdev, "TRAP_STRMOUT %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_STRMOUT %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x401804), + nvkm_rd32(device, 0x401808), + nvkm_rd32(device, 0x40180c), + nvkm_rd32(device, 0x401810)); + } + + /* No sane way found yet -- just reset the bugger. */ + nvkm_wr32(device, 0x400040, 0x80); + nvkm_wr32(device, 0x400040, 0); + nvkm_wr32(device, 0x401800, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x008); + status &= ~0x008; + } + + /* CCACHE: Handles code and c[] caches and fills them. */ + if (status & 0x010) { + ustatus = nvkm_rd32(device, 0x405018) & 0x7fffffff; + if (display) { + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_ccache, ustatus); + nvkm_error(subdev, "TRAP_CCACHE %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_CCACHE %08x %08x %08x %08x " + "%08x %08x %08x\n", + nvkm_rd32(device, 0x405000), + nvkm_rd32(device, 0x405004), + nvkm_rd32(device, 0x405008), + nvkm_rd32(device, 0x40500c), + nvkm_rd32(device, 0x405010), + nvkm_rd32(device, 0x405014), + nvkm_rd32(device, 0x40501c)); + } + + nvkm_wr32(device, 0x405018, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x010); + status &= ~0x010; + } + + /* Unknown, not seen yet... 0x402000 is the only trap status reg + * remaining, so try to handle it anyway. Perhaps related to that + * unknown DMA slot on tesla? */ + if (status & 0x20) { + ustatus = nvkm_rd32(device, 0x402000) & 0x7fffffff; + if (display) + nvkm_error(subdev, "TRAP_UNKC04 %08x\n", ustatus); + nvkm_wr32(device, 0x402000, 0xc0000000); + /* no status modifiction on purpose */ + } + + /* TEXTURE: CUDA texturing units */ + if (status & 0x040) { + nv50_gr_tp_trap(gr, 6, 0x408900, 0x408600, display, + "TRAP_TEXTURE"); + nvkm_wr32(device, 0x400108, 0x040); + status &= ~0x040; + } + + /* MP: CUDA execution engines. */ + if (status & 0x080) { + nv50_gr_tp_trap(gr, 7, 0x408314, 0x40831c, display, + "TRAP_MP"); + nvkm_wr32(device, 0x400108, 0x080); + status &= ~0x080; + } + + /* PROP: Handles TP-initiated uncached memory accesses: + * l[], g[], stack, 2d surfaces, render targets. */ + if (status & 0x100) { + nv50_gr_tp_trap(gr, 8, 0x408e08, 0x408708, display, + "TRAP_PROP"); + nvkm_wr32(device, 0x400108, 0x100); + status &= ~0x100; + } + + if (status) { + if (display) + nvkm_error(subdev, "TRAP: unknown %08x\n", status); + nvkm_wr32(device, 0x400108, status); + } + + return 1; +} + +void +nv50_gr_intr(struct nvkm_gr *base) +{ + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_chan *chan; + u32 stat = nvkm_rd32(device, 0x400100); + u32 inst = nvkm_rd32(device, 0x40032c) & 0x0fffffff; + u32 addr = nvkm_rd32(device, 0x400704); + u32 subc = (addr & 0x00070000) >> 16; + u32 mthd = (addr & 0x00001ffc); + u32 data = nvkm_rd32(device, 0x400708); + u32 class = nvkm_rd32(device, 0x400814); + u32 show = stat, show_bitfield = stat; + const struct nvkm_enum *en; + unsigned long flags; + const char *name = "unknown"; + char msg[128]; + int chid = -1; + + chan = nvkm_chan_get_inst(&gr->base.engine, (u64)inst << 12, &flags); + if (chan) { + name = chan->name; + chid = chan->id; + } + + if (show & 0x00100000) { + u32 ecode = nvkm_rd32(device, 0x400110); + en = nvkm_enum_find(nv50_data_error_names, ecode); + nvkm_error(subdev, "DATA_ERROR %08x [%s]\n", + ecode, en ? en->name : ""); + show_bitfield &= ~0x00100000; + } + + if (stat & 0x00200000) { + if (!nv50_gr_trap_handler(gr, show, chid, (u64)inst << 12, name)) + show &= ~0x00200000; + show_bitfield &= ~0x00200000; + } + + nvkm_wr32(device, 0x400100, stat); + nvkm_wr32(device, 0x400500, 0x00010001); + + if (show) { + show &= show_bitfield; + nvkm_snprintbf(msg, sizeof(msg), nv50_gr_intr_name, show); + nvkm_error(subdev, "%08x [%s] ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x\n", + stat, msg, chid, (u64)inst << 12, name, + subc, class, mthd, data); + } + + if (nvkm_rd32(device, 0x400824) & (1 << 31)) + nvkm_wr32(device, 0x400824, nvkm_rd32(device, 0x400824) & ~(1 << 31)); + + nvkm_chan_put(&chan, flags); +} + +int +nv50_gr_init(struct nvkm_gr *base) +{ + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret, units, i; + + /* NV_PGRAPH_DEBUG_3_HW_CTX_SWITCH_ENABLED */ + nvkm_wr32(device, 0x40008c, 0x00000004); + + /* reset/enable traps and interrupts */ + nvkm_wr32(device, 0x400804, 0xc0000000); + nvkm_wr32(device, 0x406800, 0xc0000000); + nvkm_wr32(device, 0x400c04, 0xc0000000); + nvkm_wr32(device, 0x401800, 0xc0000000); + nvkm_wr32(device, 0x405018, 0xc0000000); + nvkm_wr32(device, 0x402000, 0xc0000000); + + units = nvkm_rd32(device, 0x001540); + for (i = 0; i < 16; i++) { + if (!(units & (1 << i))) + continue; + + if (device->chipset < 0xa0) { + nvkm_wr32(device, 0x408900 + (i << 12), 0xc0000000); + nvkm_wr32(device, 0x408e08 + (i << 12), 0xc0000000); + nvkm_wr32(device, 0x408314 + (i << 12), 0xc0000000); + } else { + nvkm_wr32(device, 0x408600 + (i << 11), 0xc0000000); + nvkm_wr32(device, 0x408708 + (i << 11), 0xc0000000); + nvkm_wr32(device, 0x40831c + (i << 11), 0xc0000000); + } + } + + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + nvkm_wr32(device, 0x400500, 0x00010001); + + /* upload context program, initialise ctxctl defaults */ + ret = nv50_grctx_init(device, &gr->size); + if (ret) + return ret; + + nvkm_wr32(device, 0x400824, 0x00000000); + nvkm_wr32(device, 0x400828, 0x00000000); + nvkm_wr32(device, 0x40082c, 0x00000000); + nvkm_wr32(device, 0x400830, 0x00000000); + nvkm_wr32(device, 0x40032c, 0x00000000); + nvkm_wr32(device, 0x400330, 0x00000000); + + /* some unknown zcull magic */ + switch (device->chipset & 0xf0) { + case 0x50: + case 0x80: + case 0x90: + nvkm_wr32(device, 0x402ca8, 0x00000800); + break; + case 0xa0: + default: + if (device->chipset == 0xa0 || + device->chipset == 0xaa || + device->chipset == 0xac) { + nvkm_wr32(device, 0x402ca8, 0x00000802); + } else { + nvkm_wr32(device, 0x402cc0, 0x00000000); + nvkm_wr32(device, 0x402ca8, 0x00000002); + } + + break; + } + + /* zero out zcull regions */ + for (i = 0; i < 8; i++) { + nvkm_wr32(device, 0x402c20 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c24 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c28 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c2c + (i * 0x10), 0x00000000); + } + + return 0; +} + +int +nv50_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + struct nv50_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, type, inst, true, &gr->base); +} + +static const struct nvkm_gr_func +nv50_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .units = nv50_gr_units, + .sclass = { + { -1, -1, NV_NULL_CLASS, &nv50_gr_object }, + { -1, -1, NV50_TWOD, &nv50_gr_object }, + { -1, -1, NV50_MEMORY_TO_MEMORY_FORMAT, &nv50_gr_object }, + { -1, -1, NV50_TESLA, &nv50_gr_object }, + { -1, -1, NV50_COMPUTE, &nv50_gr_object }, + {} + } +}; + +int +nv50_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&nv50_gr, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h new file mode 100644 index 0000000000..97ead00423 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h @@ -0,0 +1,37 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV50_GR_H__ +#define __NV50_GR_H__ +#define nv50_gr(p) container_of((p), struct nv50_gr, base) +#include "priv.h" + +struct nv50_gr { + struct nvkm_gr base; + const struct nv50_gr_func *func; + spinlock_t lock; + u32 size; +}; + +int nv50_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_gr **); +int nv50_gr_init(struct nvkm_gr *); +void nv50_gr_intr(struct nvkm_gr *); +u64 nv50_gr_units(struct nvkm_gr *); + +int g84_gr_tlb_flush(struct nvkm_gr *); + +#define nv50_gr_chan(p) container_of((p), struct nv50_gr_chan, object) +#include + +struct nv50_gr_chan { + struct nvkm_object object; + struct nv50_gr *gr; +}; + +int nv50_gr_chan_new(struct nvkm_gr *, struct nvkm_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + +extern const struct nvkm_object_func nv50_gr_object; + +int nv50_grctx_init(struct nvkm_device *, u32 *size); +void nv50_grctx_fill(struct nvkm_device *, struct nvkm_gpuobj *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h new file mode 100644 index 0000000000..0884abc73a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h @@ -0,0 +1,47 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_GR_PRIV_H__ +#define __NVKM_GR_PRIV_H__ +#define nvkm_gr(p) container_of((p), struct nvkm_gr, engine) +#include +#include +struct nvkm_fb_tile; +struct nvkm_chan; + +int nvkm_gr_ctor(const struct nvkm_gr_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + bool enable, struct nvkm_gr *); + +bool nv04_gr_idle(struct nvkm_gr *); + +struct nvkm_gr_func { + void *(*dtor)(struct nvkm_gr *); + int (*oneinit)(struct nvkm_gr *); + int (*init)(struct nvkm_gr *); + int (*fini)(struct nvkm_gr *, bool); + int (*reset)(struct nvkm_gr *); + int (*nonstall)(struct nvkm_gr *); + void (*intr)(struct nvkm_gr *); + void (*tile)(struct nvkm_gr *, int region, struct nvkm_fb_tile *); + int (*tlb_flush)(struct nvkm_gr *); + int (*chan_new)(struct nvkm_gr *, struct nvkm_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + int (*object_get)(struct nvkm_gr *, int, struct nvkm_sclass *); + /* Returns chipset-specific counts of units packed into an u64. + */ + u64 (*units)(struct nvkm_gr *); + bool (*chsw_load)(struct nvkm_gr *); + struct { + int (*pause)(struct nvkm_gr *); + int (*resume)(struct nvkm_gr *); + u32 (*inst)(struct nvkm_gr *); + } ctxsw; + struct nvkm_sclass sclass[]; +}; + +extern const struct nvkm_bitfield nv04_gr_nsource[]; +extern const struct nvkm_object_func nv04_gr_object; + +extern const struct nvkm_bitfield nv10_gr_intr_name[]; +extern const struct nvkm_bitfield nv10_gr_nstatus[]; + +extern const struct nvkm_enum nv50_data_error_names[]; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/regs.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/regs.h new file mode 100644 index 0000000000..fd1b7d35c6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/regs.h @@ -0,0 +1,275 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_GR_REGS_H__ +#define __NVKM_GR_REGS_H__ + +#define NV04_PGRAPH_DEBUG_0 0x00400080 +#define NV04_PGRAPH_DEBUG_1 0x00400084 +#define NV04_PGRAPH_DEBUG_2 0x00400088 +#define NV04_PGRAPH_DEBUG_3 0x0040008c +#define NV10_PGRAPH_DEBUG_4 0x00400090 +#define NV03_PGRAPH_INTR 0x00400100 +#define NV03_PGRAPH_NSTATUS 0x00400104 +# define NV04_PGRAPH_NSTATUS_STATE_IN_USE (1<<11) +# define NV04_PGRAPH_NSTATUS_INVALID_STATE (1<<12) +# define NV04_PGRAPH_NSTATUS_BAD_ARGUMENT (1<<13) +# define NV04_PGRAPH_NSTATUS_PROTECTION_FAULT (1<<14) +# define NV10_PGRAPH_NSTATUS_STATE_IN_USE (1<<23) +# define NV10_PGRAPH_NSTATUS_INVALID_STATE (1<<24) +# define NV10_PGRAPH_NSTATUS_BAD_ARGUMENT (1<<25) +# define NV10_PGRAPH_NSTATUS_PROTECTION_FAULT (1<<26) +#define NV03_PGRAPH_NSOURCE 0x00400108 +# define NV03_PGRAPH_NSOURCE_NOTIFICATION (1<<0) +# define NV03_PGRAPH_NSOURCE_DATA_ERROR (1<<1) +# define NV03_PGRAPH_NSOURCE_PROTECTION_ERROR (1<<2) +# define NV03_PGRAPH_NSOURCE_RANGE_EXCEPTION (1<<3) +# define NV03_PGRAPH_NSOURCE_LIMIT_COLOR (1<<4) +# define NV03_PGRAPH_NSOURCE_LIMIT_ZETA (1<<5) +# define NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD (1<<6) +# define NV03_PGRAPH_NSOURCE_DMA_R_PROTECTION (1<<7) +# define NV03_PGRAPH_NSOURCE_DMA_W_PROTECTION (1<<8) +# define NV03_PGRAPH_NSOURCE_FORMAT_EXCEPTION (1<<9) +# define NV03_PGRAPH_NSOURCE_PATCH_EXCEPTION (1<<10) +# define NV03_PGRAPH_NSOURCE_STATE_INVALID (1<<11) +# define NV03_PGRAPH_NSOURCE_DOUBLE_NOTIFY (1<<12) +# define NV03_PGRAPH_NSOURCE_NOTIFY_IN_USE (1<<13) +# define NV03_PGRAPH_NSOURCE_METHOD_CNT (1<<14) +# define NV03_PGRAPH_NSOURCE_BFR_NOTIFICATION (1<<15) +# define NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION (1<<16) +# define NV03_PGRAPH_NSOURCE_DMA_WIDTH_A (1<<17) +# define NV03_PGRAPH_NSOURCE_DMA_WIDTH_B (1<<18) +#define NV03_PGRAPH_INTR_EN 0x00400140 +#define NV40_PGRAPH_INTR_EN 0x0040013C +# define NV_PGRAPH_INTR_NOTIFY (1<<0) +# define NV_PGRAPH_INTR_MISSING_HW (1<<4) +# define NV_PGRAPH_INTR_CONTEXT_SWITCH (1<<12) +# define NV_PGRAPH_INTR_BUFFER_NOTIFY (1<<16) +# define NV_PGRAPH_INTR_ERROR (1<<20) +#define NV10_PGRAPH_CTX_CONTROL 0x00400144 +#define NV10_PGRAPH_CTX_USER 0x00400148 +#define NV10_PGRAPH_CTX_SWITCH(i) (0x0040014C + 0x4*(i)) +#define NV04_PGRAPH_CTX_SWITCH1 0x00400160 +#define NV10_PGRAPH_CTX_CACHE(i, j) (0x00400160 \ + + 0x4*(i) + 0x20*(j)) +#define NV04_PGRAPH_CTX_SWITCH2 0x00400164 +#define NV04_PGRAPH_CTX_SWITCH3 0x00400168 +#define NV04_PGRAPH_CTX_SWITCH4 0x0040016C +#define NV04_PGRAPH_CTX_CONTROL 0x00400170 +#define NV04_PGRAPH_CTX_USER 0x00400174 +#define NV04_PGRAPH_CTX_CACHE1 0x00400180 +#define NV03_PGRAPH_CTX_CONTROL 0x00400190 +#define NV03_PGRAPH_CTX_USER 0x00400194 +#define NV04_PGRAPH_CTX_CACHE2 0x004001A0 +#define NV04_PGRAPH_CTX_CACHE3 0x004001C0 +#define NV04_PGRAPH_CTX_CACHE4 0x004001E0 +#define NV40_PGRAPH_CTXCTL_0304 0x00400304 +#define NV40_PGRAPH_CTXCTL_0304_XFER_CTX 0x00000001 +#define NV40_PGRAPH_CTXCTL_UCODE_STAT 0x00400308 +#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_MASK 0xff000000 +#define NV40_PGRAPH_CTXCTL_UCODE_STAT_IP_SHIFT 24 +#define NV40_PGRAPH_CTXCTL_UCODE_STAT_OP_MASK 0x00ffffff +#define NV40_PGRAPH_CTXCTL_0310 0x00400310 +#define NV40_PGRAPH_CTXCTL_0310_XFER_SAVE 0x00000020 +#define NV40_PGRAPH_CTXCTL_0310_XFER_LOAD 0x00000040 +#define NV40_PGRAPH_CTXCTL_030C 0x0040030c +#define NV40_PGRAPH_CTXCTL_UCODE_INDEX 0x00400324 +#define NV40_PGRAPH_CTXCTL_UCODE_DATA 0x00400328 +#define NV40_PGRAPH_CTXCTL_CUR 0x0040032c +#define NV40_PGRAPH_CTXCTL_CUR_LOADED 0x01000000 +#define NV40_PGRAPH_CTXCTL_CUR_INSTANCE 0x000FFFFF +#define NV40_PGRAPH_CTXCTL_NEXT 0x00400330 +#define NV40_PGRAPH_CTXCTL_NEXT_INSTANCE 0x000fffff +#define NV50_PGRAPH_CTXCTL_CUR 0x0040032c +#define NV50_PGRAPH_CTXCTL_CUR_LOADED 0x80000000 +#define NV50_PGRAPH_CTXCTL_CUR_INSTANCE 0x00ffffff +#define NV50_PGRAPH_CTXCTL_NEXT 0x00400330 +#define NV50_PGRAPH_CTXCTL_NEXT_INSTANCE 0x00ffffff +#define NV03_PGRAPH_ABS_X_RAM 0x00400400 +#define NV03_PGRAPH_ABS_Y_RAM 0x00400480 +#define NV03_PGRAPH_X_MISC 0x00400500 +#define NV03_PGRAPH_Y_MISC 0x00400504 +#define NV04_PGRAPH_VALID1 0x00400508 +#define NV04_PGRAPH_SOURCE_COLOR 0x0040050C +#define NV04_PGRAPH_MISC24_0 0x00400510 +#define NV03_PGRAPH_XY_LOGIC_MISC0 0x00400514 +#define NV03_PGRAPH_XY_LOGIC_MISC1 0x00400518 +#define NV03_PGRAPH_XY_LOGIC_MISC2 0x0040051C +#define NV03_PGRAPH_XY_LOGIC_MISC3 0x00400520 +#define NV03_PGRAPH_CLIPX_0 0x00400524 +#define NV03_PGRAPH_CLIPX_1 0x00400528 +#define NV03_PGRAPH_CLIPY_0 0x0040052C +#define NV03_PGRAPH_CLIPY_1 0x00400530 +#define NV03_PGRAPH_ABS_ICLIP_XMAX 0x00400534 +#define NV03_PGRAPH_ABS_ICLIP_YMAX 0x00400538 +#define NV03_PGRAPH_ABS_UCLIP_XMIN 0x0040053C +#define NV03_PGRAPH_ABS_UCLIP_YMIN 0x00400540 +#define NV03_PGRAPH_ABS_UCLIP_XMAX 0x00400544 +#define NV03_PGRAPH_ABS_UCLIP_YMAX 0x00400548 +#define NV03_PGRAPH_ABS_UCLIPA_XMIN 0x00400560 +#define NV03_PGRAPH_ABS_UCLIPA_YMIN 0x00400564 +#define NV03_PGRAPH_ABS_UCLIPA_XMAX 0x00400568 +#define NV03_PGRAPH_ABS_UCLIPA_YMAX 0x0040056C +#define NV04_PGRAPH_MISC24_1 0x00400570 +#define NV04_PGRAPH_MISC24_2 0x00400574 +#define NV04_PGRAPH_VALID2 0x00400578 +#define NV04_PGRAPH_PASSTHRU_0 0x0040057C +#define NV04_PGRAPH_PASSTHRU_1 0x00400580 +#define NV04_PGRAPH_PASSTHRU_2 0x00400584 +#define NV10_PGRAPH_DIMX_TEXTURE 0x00400588 +#define NV10_PGRAPH_WDIMX_TEXTURE 0x0040058C +#define NV04_PGRAPH_COMBINE_0_ALPHA 0x00400590 +#define NV04_PGRAPH_COMBINE_0_COLOR 0x00400594 +#define NV04_PGRAPH_COMBINE_1_ALPHA 0x00400598 +#define NV04_PGRAPH_COMBINE_1_COLOR 0x0040059C +#define NV04_PGRAPH_FORMAT_0 0x004005A8 +#define NV04_PGRAPH_FORMAT_1 0x004005AC +#define NV04_PGRAPH_FILTER_0 0x004005B0 +#define NV04_PGRAPH_FILTER_1 0x004005B4 +#define NV03_PGRAPH_MONO_COLOR0 0x00400600 +#define NV04_PGRAPH_ROP3 0x00400604 +#define NV04_PGRAPH_BETA_AND 0x00400608 +#define NV04_PGRAPH_BETA_PREMULT 0x0040060C +#define NV04_PGRAPH_LIMIT_VIOL_PIX 0x00400610 +#define NV04_PGRAPH_FORMATS 0x00400618 +#define NV10_PGRAPH_DEBUG_2 0x00400620 +#define NV04_PGRAPH_BOFFSET0 0x00400640 +#define NV04_PGRAPH_BOFFSET1 0x00400644 +#define NV04_PGRAPH_BOFFSET2 0x00400648 +#define NV04_PGRAPH_BOFFSET3 0x0040064C +#define NV04_PGRAPH_BOFFSET4 0x00400650 +#define NV04_PGRAPH_BOFFSET5 0x00400654 +#define NV04_PGRAPH_BBASE0 0x00400658 +#define NV04_PGRAPH_BBASE1 0x0040065C +#define NV04_PGRAPH_BBASE2 0x00400660 +#define NV04_PGRAPH_BBASE3 0x00400664 +#define NV04_PGRAPH_BBASE4 0x00400668 +#define NV04_PGRAPH_BBASE5 0x0040066C +#define NV04_PGRAPH_BPITCH0 0x00400670 +#define NV04_PGRAPH_BPITCH1 0x00400674 +#define NV04_PGRAPH_BPITCH2 0x00400678 +#define NV04_PGRAPH_BPITCH3 0x0040067C +#define NV04_PGRAPH_BPITCH4 0x00400680 +#define NV04_PGRAPH_BLIMIT0 0x00400684 +#define NV04_PGRAPH_BLIMIT1 0x00400688 +#define NV04_PGRAPH_BLIMIT2 0x0040068C +#define NV04_PGRAPH_BLIMIT3 0x00400690 +#define NV04_PGRAPH_BLIMIT4 0x00400694 +#define NV04_PGRAPH_BLIMIT5 0x00400698 +#define NV04_PGRAPH_BSWIZZLE2 0x0040069C +#define NV04_PGRAPH_BSWIZZLE5 0x004006A0 +#define NV03_PGRAPH_STATUS 0x004006B0 +#define NV04_PGRAPH_STATUS 0x00400700 +# define NV40_PGRAPH_STATUS_SYNC_STALL 0x00004000 +#define NV04_PGRAPH_TRAPPED_ADDR 0x00400704 +#define NV04_PGRAPH_TRAPPED_DATA 0x00400708 +#define NV04_PGRAPH_SURFACE 0x0040070C +#define NV10_PGRAPH_TRAPPED_DATA_HIGH 0x0040070C +#define NV04_PGRAPH_STATE 0x00400710 +#define NV10_PGRAPH_SURFACE 0x00400710 +#define NV04_PGRAPH_NOTIFY 0x00400714 +#define NV10_PGRAPH_STATE 0x00400714 +#define NV10_PGRAPH_NOTIFY 0x00400718 + +#define NV04_PGRAPH_FIFO 0x00400720 + +#define NV04_PGRAPH_BPIXEL 0x00400724 +#define NV10_PGRAPH_RDI_INDEX 0x00400750 +#define NV04_PGRAPH_FFINTFC_ST2 0x00400754 +#define NV10_PGRAPH_RDI_DATA 0x00400754 +#define NV04_PGRAPH_DMA_PITCH 0x00400760 +#define NV10_PGRAPH_FFINTFC_FIFO_PTR 0x00400760 +#define NV04_PGRAPH_DVD_COLORFMT 0x00400764 +#define NV10_PGRAPH_FFINTFC_ST2 0x00400764 +#define NV04_PGRAPH_SCALED_FORMAT 0x00400768 +#define NV10_PGRAPH_FFINTFC_ST2_DL 0x00400768 +#define NV10_PGRAPH_FFINTFC_ST2_DH 0x0040076c +#define NV10_PGRAPH_DMA_PITCH 0x00400770 +#define NV10_PGRAPH_DVD_COLORFMT 0x00400774 +#define NV10_PGRAPH_SCALED_FORMAT 0x00400778 +#define NV20_PGRAPH_CHANNEL_CTX_TABLE 0x00400780 +#define NV20_PGRAPH_CHANNEL_CTX_POINTER 0x00400784 +#define NV20_PGRAPH_CHANNEL_CTX_XFER 0x00400788 +#define NV20_PGRAPH_CHANNEL_CTX_XFER_LOAD 0x00000001 +#define NV20_PGRAPH_CHANNEL_CTX_XFER_SAVE 0x00000002 +#define NV04_PGRAPH_PATT_COLOR0 0x00400800 +#define NV04_PGRAPH_PATT_COLOR1 0x00400804 +#define NV04_PGRAPH_PATTERN 0x00400808 +#define NV04_PGRAPH_PATTERN_SHAPE 0x00400810 +#define NV04_PGRAPH_CHROMA 0x00400814 +#define NV04_PGRAPH_CONTROL0 0x00400818 +#define NV04_PGRAPH_CONTROL1 0x0040081C +#define NV04_PGRAPH_CONTROL2 0x00400820 +#define NV04_PGRAPH_BLEND 0x00400824 +#define NV04_PGRAPH_STORED_FMT 0x00400830 +#define NV04_PGRAPH_PATT_COLORRAM 0x00400900 +#define NV20_PGRAPH_TILE(i) (0x00400900 + (i*16)) +#define NV20_PGRAPH_TLIMIT(i) (0x00400904 + (i*16)) +#define NV20_PGRAPH_TSIZE(i) (0x00400908 + (i*16)) +#define NV20_PGRAPH_TSTATUS(i) (0x0040090C + (i*16)) +#define NV20_PGRAPH_ZCOMP(i) (0x00400980 + 4*(i)) +#define NV41_PGRAPH_ZCOMP0(i) (0x004009c0 + 4*(i)) +#define NV10_PGRAPH_TILE(i) (0x00400B00 + (i*16)) +#define NV10_PGRAPH_TLIMIT(i) (0x00400B04 + (i*16)) +#define NV10_PGRAPH_TSIZE(i) (0x00400B08 + (i*16)) +#define NV10_PGRAPH_TSTATUS(i) (0x00400B0C + (i*16)) +#define NV04_PGRAPH_U_RAM 0x00400D00 +#define NV47_PGRAPH_TILE(i) (0x00400D00 + (i*16)) +#define NV47_PGRAPH_TLIMIT(i) (0x00400D04 + (i*16)) +#define NV47_PGRAPH_TSIZE(i) (0x00400D08 + (i*16)) +#define NV47_PGRAPH_TSTATUS(i) (0x00400D0C + (i*16)) +#define NV04_PGRAPH_V_RAM 0x00400D40 +#define NV04_PGRAPH_W_RAM 0x00400D80 +#define NV47_PGRAPH_ZCOMP0(i) (0x00400e00 + 4*(i)) +#define NV10_PGRAPH_COMBINER0_IN_ALPHA 0x00400E40 +#define NV10_PGRAPH_COMBINER1_IN_ALPHA 0x00400E44 +#define NV10_PGRAPH_COMBINER0_IN_RGB 0x00400E48 +#define NV10_PGRAPH_COMBINER1_IN_RGB 0x00400E4C +#define NV10_PGRAPH_COMBINER_COLOR0 0x00400E50 +#define NV10_PGRAPH_COMBINER_COLOR1 0x00400E54 +#define NV10_PGRAPH_COMBINER0_OUT_ALPHA 0x00400E58 +#define NV10_PGRAPH_COMBINER1_OUT_ALPHA 0x00400E5C +#define NV10_PGRAPH_COMBINER0_OUT_RGB 0x00400E60 +#define NV10_PGRAPH_COMBINER1_OUT_RGB 0x00400E64 +#define NV10_PGRAPH_COMBINER_FINAL0 0x00400E68 +#define NV10_PGRAPH_COMBINER_FINAL1 0x00400E6C +#define NV10_PGRAPH_WINDOWCLIP_HORIZONTAL 0x00400F00 +#define NV10_PGRAPH_WINDOWCLIP_VERTICAL 0x00400F20 +#define NV10_PGRAPH_XFMODE0 0x00400F40 +#define NV10_PGRAPH_XFMODE1 0x00400F44 +#define NV10_PGRAPH_GLOBALSTATE0 0x00400F48 +#define NV10_PGRAPH_GLOBALSTATE1 0x00400F4C +#define NV10_PGRAPH_PIPE_ADDRESS 0x00400F50 +#define NV10_PGRAPH_PIPE_DATA 0x00400F54 +#define NV04_PGRAPH_DMA_START_0 0x00401000 +#define NV04_PGRAPH_DMA_START_1 0x00401004 +#define NV04_PGRAPH_DMA_LENGTH 0x00401008 +#define NV04_PGRAPH_DMA_MISC 0x0040100C +#define NV04_PGRAPH_DMA_DATA_0 0x00401020 +#define NV04_PGRAPH_DMA_DATA_1 0x00401024 +#define NV04_PGRAPH_DMA_RM 0x00401030 +#define NV04_PGRAPH_DMA_A_XLATE_INST 0x00401040 +#define NV04_PGRAPH_DMA_A_CONTROL 0x00401044 +#define NV04_PGRAPH_DMA_A_LIMIT 0x00401048 +#define NV04_PGRAPH_DMA_A_TLB_PTE 0x0040104C +#define NV04_PGRAPH_DMA_A_TLB_TAG 0x00401050 +#define NV04_PGRAPH_DMA_A_ADJ_OFFSET 0x00401054 +#define NV04_PGRAPH_DMA_A_OFFSET 0x00401058 +#define NV04_PGRAPH_DMA_A_SIZE 0x0040105C +#define NV04_PGRAPH_DMA_A_Y_SIZE 0x00401060 +#define NV04_PGRAPH_DMA_B_XLATE_INST 0x00401080 +#define NV04_PGRAPH_DMA_B_CONTROL 0x00401084 +#define NV04_PGRAPH_DMA_B_LIMIT 0x00401088 +#define NV04_PGRAPH_DMA_B_TLB_PTE 0x0040108C +#define NV04_PGRAPH_DMA_B_TLB_TAG 0x00401090 +#define NV04_PGRAPH_DMA_B_ADJ_OFFSET 0x00401094 +#define NV04_PGRAPH_DMA_B_OFFSET 0x00401098 +#define NV04_PGRAPH_DMA_B_SIZE 0x0040109C +#define NV04_PGRAPH_DMA_B_Y_SIZE 0x004010A0 +#define NV47_PGRAPH_ZCOMP1(i) (0x004068c0 + 4*(i)) +#define NV40_PGRAPH_TILE1(i) (0x00406900 + (i*16)) +#define NV40_PGRAPH_TLIMIT1(i) (0x00406904 + (i*16)) +#define NV40_PGRAPH_TSIZE1(i) (0x00406908 + (i*16)) +#define NV40_PGRAPH_TSTATUS1(i) (0x0040690C + (i*16)) +#define NV40_PGRAPH_ZCOMP1(i) (0x00406980 + 4*(i)) +#define NV41_PGRAPH_ZCOMP1(i) (0x004069c0 + 4*(i)) + +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/tu102.c new file mode 100644 index 0000000000..a7775aa185 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/tu102.c @@ -0,0 +1,220 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include + +void +tu102_gr_init_fecs_exceptions(struct gf100_gr *gr) +{ + nvkm_wr32(gr->base.engine.subdev.device, 0x409c24, 0x006e0003); +} + +void +tu102_gr_init_fs(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + int sm; + + gp100_grctx_generate_smid_config(gr); + gk104_grctx_generate_gpc_tpc_nr(gr); + + for (sm = 0; sm < gr->sm_nr; sm++) { + int tpc = gv100_gr_nonpes_aware_tpc(gr, gr->sm[sm].gpc, gr->sm[sm].tpc); + + nvkm_wr32(device, GPC_UNIT(gr->sm[sm].gpc, 0x0c10 + tpc * 4), sm); + } + + gm200_grctx_generate_dist_skip_table(gr); + gf100_gr_init_num_tpc_per_gpc(gr, true, true); +} + +void +tu102_gr_init_zcull(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + const u8 tile_nr = gr->func->gpc_nr * gr->func->tpc_nr; + u8 bank[GPC_MAX] = {}, gpc, i, j; + u32 data; + + for (i = 0; i < tile_nr; i += 8) { + for (data = 0, j = 0; j < 8 && i + j < gr->tpc_total; j++) { + data |= bank[gr->tile[i + j]] << (j * 4); + bank[gr->tile[i + j]]++; + } + nvkm_wr32(device, GPC_BCAST(0x0980 + ((i / 8) * 4)), data); + } + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->screen_tile_row_offset << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); + } + + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); +} + +static void +tu102_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0xf8001fff); + nvkm_wr32(device, 0x418890, 0x00000000); + nvkm_wr32(device, 0x418894, 0x00000000); + + nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); + nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); + nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); +} + +static const struct gf100_gr_func +tu102_gr = { + .oneinit_tiles = gm200_gr_oneinit_tiles, + .oneinit_sm_id = gv100_gr_oneinit_sm_id, + .init = gf100_gr_init, + .init_419bd8 = gv100_gr_init_419bd8, + .init_gpc_mmu = tu102_gr_init_gpc_mmu, + .init_vsc_stream_master = gk104_gr_init_vsc_stream_master, + .init_zcull = tu102_gr_init_zcull, + .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, + .init_rop_active_fbps = gp100_gr_init_rop_active_fbps, + .init_swdx_pes_mask = gp102_gr_init_swdx_pes_mask, + .init_fs = tu102_gr_init_fs, + .init_fecs_exceptions = tu102_gr_init_fecs_exceptions, + .init_ds_hww_esr_2 = gm200_gr_init_ds_hww_esr_2, + .init_sked_hww_esr = gk104_gr_init_sked_hww_esr, + .init_ppc_exceptions = gk104_gr_init_ppc_exceptions, + .init_504430 = gv100_gr_init_504430, + .init_shader_exceptions = gv100_gr_init_shader_exceptions, + .init_rop_exceptions = gf100_gr_init_rop_exceptions, + .init_exception2 = gf100_gr_init_exception2, + .init_4188a4 = gv100_gr_init_4188a4, + .trap_mp = gv100_gr_trap_mp, + .fecs.reset = gf100_gr_fecs_reset, + .rops = gm200_gr_rops, + .gpc_nr = 6, + .tpc_nr = 6, + .ppc_nr = 3, + .grctx = &tu102_grctx, + .zbc = &gp102_gr_zbc, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, TURING_A, &gf100_fermi }, + { -1, -1, TURING_COMPUTE_A }, + {} + } +}; + +MODULE_FIRMWARE("nvidia/tu102/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/tu102/gr/sw_veid_bundle_init.bin"); + +MODULE_FIRMWARE("nvidia/tu104/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/tu104/gr/sw_veid_bundle_init.bin"); + +MODULE_FIRMWARE("nvidia/tu106/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/tu106/gr/sw_veid_bundle_init.bin"); + +MODULE_FIRMWARE("nvidia/tu117/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/tu117/gr/sw_veid_bundle_init.bin"); + +MODULE_FIRMWARE("nvidia/tu116/gr/fecs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/fecs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/fecs_data.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/fecs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/gpccs_bl.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/gpccs_inst.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/gpccs_data.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/gpccs_sig.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/sw_ctx.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/sw_nonctx.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/sw_bundle_init.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/sw_method_init.bin"); +MODULE_FIRMWARE("nvidia/tu116/gr/sw_veid_bundle_init.bin"); + +int +tu102_gr_av_to_init_veid(struct nvkm_blob *blob, struct gf100_gr_pack **ppack) +{ + return gk20a_gr_av_to_init_(blob, 64, 0x00100000, ppack); +} + +static const struct gf100_gr_fwif +tu102_gr_fwif[] = { + { 0, gm200_gr_load, &tu102_gr, &gp108_gr_fecs_acr, &gp108_gr_gpccs_acr }, + { -1, gm200_gr_nofw }, + {} +}; + +int +tu102_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(tu102_gr_fwif, device, type, inst, pgr); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/Kbuild new file mode 100644 index 0000000000..8d2d9eae56 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/Kbuild @@ -0,0 +1,6 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/mpeg/nv31.o +nvkm-y += nvkm/engine/mpeg/nv40.o +nvkm-y += nvkm/engine/mpeg/nv44.o +nvkm-y += nvkm/engine/mpeg/nv50.o +nvkm-y += nvkm/engine/mpeg/g84.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/g84.c new file mode 100644 index 0000000000..0fcc0ffa1e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/g84.c @@ -0,0 +1,44 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +static const struct nvkm_engine_func +g84_mpeg = { + .init = nv50_mpeg_init, + .intr = nv50_mpeg_intr, + .cclass = &nv50_mpeg_cclass, + .sclass = { + { -1, -1, G82_MPEG, &nv31_mpeg_object }, + {} + } +}; + +int +g84_mpeg_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pmpeg) +{ + return nvkm_engine_new_(&g84_mpeg, device, type, inst, true, pmpeg); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.c b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.c new file mode 100644 index 0000000000..db9fc1ecae --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.c @@ -0,0 +1,298 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv31.h" + +#include +#include +#include +#include +#include + +#include + +/******************************************************************************* + * MPEG object classes + ******************************************************************************/ + +static int +nv31_mpeg_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, align, + false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; +} + +const struct nvkm_object_func +nv31_mpeg_object = { + .bind = nv31_mpeg_object_bind, +}; + +/******************************************************************************* + * PMPEG context + ******************************************************************************/ + +static void * +nv31_mpeg_chan_dtor(struct nvkm_object *object) +{ + struct nv31_mpeg_chan *chan = nv31_mpeg_chan(object); + struct nv31_mpeg *mpeg = chan->mpeg; + unsigned long flags; + + spin_lock_irqsave(&mpeg->engine.lock, flags); + if (mpeg->chan == chan) + mpeg->chan = NULL; + spin_unlock_irqrestore(&mpeg->engine.lock, flags); + return chan; +} + +static const struct nvkm_object_func +nv31_mpeg_chan = { + .dtor = nv31_mpeg_chan_dtor, +}; + +int +nv31_mpeg_chan_new(struct nvkm_chan *fifoch, const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nv31_mpeg *mpeg = nv31_mpeg(oclass->engine); + struct nv31_mpeg_chan *chan; + unsigned long flags; + int ret = -EBUSY; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv31_mpeg_chan, oclass, &chan->object); + chan->mpeg = mpeg; + chan->fifo = fifoch; + *pobject = &chan->object; + + spin_lock_irqsave(&mpeg->engine.lock, flags); + if (!mpeg->chan) { + mpeg->chan = chan; + ret = 0; + } + spin_unlock_irqrestore(&mpeg->engine.lock, flags); + return ret; +} + +/******************************************************************************* + * PMPEG engine/subdev functions + ******************************************************************************/ + +void +nv31_mpeg_tile(struct nvkm_engine *engine, int i, struct nvkm_fb_tile *tile) +{ + struct nv31_mpeg *mpeg = nv31_mpeg(engine); + struct nvkm_device *device = mpeg->engine.subdev.device; + + nvkm_wr32(device, 0x00b008 + (i * 0x10), tile->pitch); + nvkm_wr32(device, 0x00b004 + (i * 0x10), tile->limit); + nvkm_wr32(device, 0x00b000 + (i * 0x10), tile->addr); +} + +static bool +nv31_mpeg_mthd_dma(struct nvkm_device *device, u32 mthd, u32 data) +{ + struct nv31_mpeg *mpeg = nv31_mpeg(device->mpeg); + struct nvkm_subdev *subdev = &mpeg->engine.subdev; + u32 inst = data << 4; + u32 dma0 = nvkm_rd32(device, 0x700000 + inst); + u32 dma1 = nvkm_rd32(device, 0x700004 + inst); + u32 dma2 = nvkm_rd32(device, 0x700008 + inst); + u32 base = (dma2 & 0xfffff000) | (dma0 >> 20); + u32 size = dma1 + 1; + + /* only allow linear DMA objects */ + if (!(dma0 & 0x00002000)) { + nvkm_error(subdev, "inst %08x dma0 %08x dma1 %08x dma2 %08x\n", + inst, dma0, dma1, dma2); + return false; + } + + if (mthd == 0x0190) { + /* DMA_CMD */ + nvkm_mask(device, 0x00b300, 0x00010000, + (dma0 & 0x00030000) ? 0x00010000 : 0); + nvkm_wr32(device, 0x00b334, base); + nvkm_wr32(device, 0x00b324, size); + } else + if (mthd == 0x01a0) { + /* DMA_DATA */ + nvkm_mask(device, 0x00b300, 0x00020000, + (dma0 & 0x00030000) ? 0x00020000 : 0); + nvkm_wr32(device, 0x00b360, base); + nvkm_wr32(device, 0x00b364, size); + } else { + /* DMA_IMAGE, VRAM only */ + if (dma0 & 0x00030000) + return false; + + nvkm_wr32(device, 0x00b370, base); + nvkm_wr32(device, 0x00b374, size); + } + + return true; +} + +static bool +nv31_mpeg_mthd(struct nv31_mpeg *mpeg, u32 mthd, u32 data) +{ + struct nvkm_device *device = mpeg->engine.subdev.device; + switch (mthd) { + case 0x190: + case 0x1a0: + case 0x1b0: + return mpeg->func->mthd_dma(device, mthd, data); + default: + break; + } + return false; +} + +static void +nv31_mpeg_intr(struct nvkm_engine *engine) +{ + struct nv31_mpeg *mpeg = nv31_mpeg(engine); + struct nvkm_subdev *subdev = &mpeg->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x00b100); + u32 type = nvkm_rd32(device, 0x00b230); + u32 mthd = nvkm_rd32(device, 0x00b234); + u32 data = nvkm_rd32(device, 0x00b238); + u32 show = stat; + unsigned long flags; + + spin_lock_irqsave(&mpeg->engine.lock, flags); + + if (stat & 0x01000000) { + /* happens on initial binding of the object */ + if (type == 0x00000020 && mthd == 0x0000) { + nvkm_mask(device, 0x00b308, 0x00000000, 0x00000000); + show &= ~0x01000000; + } + + if (type == 0x00000010) { + if (nv31_mpeg_mthd(mpeg, mthd, data)) + show &= ~0x01000000; + } + } + + nvkm_wr32(device, 0x00b100, stat); + nvkm_wr32(device, 0x00b230, 0x00000001); + + if (show) { + nvkm_error(subdev, "ch %d [%s] %08x %08x %08x %08x\n", + mpeg->chan ? mpeg->chan->fifo->id : -1, + mpeg->chan ? mpeg->chan->fifo->name : + "unknown", stat, type, mthd, data); + } + + spin_unlock_irqrestore(&mpeg->engine.lock, flags); +} + +int +nv31_mpeg_init(struct nvkm_engine *mpeg) +{ + struct nvkm_subdev *subdev = &mpeg->subdev; + struct nvkm_device *device = subdev->device; + + /* VPE init */ + nvkm_wr32(device, 0x00b0e0, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */ + nvkm_wr32(device, 0x00b0e8, 0x00000020); /* nvidia: rd 0x01, wr 0x20 */ + + /* PMPEG init */ + nvkm_wr32(device, 0x00b32c, 0x00000000); + nvkm_wr32(device, 0x00b314, 0x00000100); + nvkm_wr32(device, 0x00b220, 0x00000031); + nvkm_wr32(device, 0x00b300, 0x02001ec1); + nvkm_mask(device, 0x00b32c, 0x00000001, 0x00000001); + + nvkm_wr32(device, 0x00b100, 0xffffffff); + nvkm_wr32(device, 0x00b140, 0xffffffff); + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x00b200) & 0x00000001)) + break; + ) < 0) { + nvkm_error(subdev, "timeout %08x\n", + nvkm_rd32(device, 0x00b200)); + return -EBUSY; + } + + return 0; +} + +static void * +nv31_mpeg_dtor(struct nvkm_engine *engine) +{ + return nv31_mpeg(engine); +} + +static const struct nvkm_engine_func +nv31_mpeg_ = { + .dtor = nv31_mpeg_dtor, + .init = nv31_mpeg_init, + .intr = nv31_mpeg_intr, + .tile = nv31_mpeg_tile, + .fifo.cclass = nv31_mpeg_chan_new, + .sclass = { + { -1, -1, NV31_MPEG, &nv31_mpeg_object }, + {} + } +}; + +int +nv31_mpeg_new_(const struct nv31_mpeg_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_engine **pmpeg) +{ + struct nv31_mpeg *mpeg; + + if (!(mpeg = kzalloc(sizeof(*mpeg), GFP_KERNEL))) + return -ENOMEM; + mpeg->func = func; + *pmpeg = &mpeg->engine; + + return nvkm_engine_ctor(&nv31_mpeg_, device, type, inst, true, &mpeg->engine); +} + +static const struct nv31_mpeg_func +nv31_mpeg = { + .mthd_dma = nv31_mpeg_mthd_dma, +}; + +int +nv31_mpeg_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pmpeg) +{ + return nv31_mpeg_new_(&nv31_mpeg, device, type, inst, pmpeg); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.h b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.h new file mode 100644 index 0000000000..251d659565 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv31.h @@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NV31_MPEG_H__ +#define __NV31_MPEG_H__ +#define nv31_mpeg(p) container_of((p), struct nv31_mpeg, engine) +#include "priv.h" +#include + +struct nv31_mpeg { + const struct nv31_mpeg_func *func; + struct nvkm_engine engine; + struct nv31_mpeg_chan *chan; +}; + +int nv31_mpeg_new_(const struct nv31_mpeg_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_engine **); + +struct nv31_mpeg_func { + bool (*mthd_dma)(struct nvkm_device *, u32 mthd, u32 data); +}; + +#define nv31_mpeg_chan(p) container_of((p), struct nv31_mpeg_chan, object) +#include + +struct nv31_mpeg_chan { + struct nvkm_object object; + struct nv31_mpeg *mpeg; + struct nvkm_chan *fifo; +}; + +int nv31_mpeg_chan_new(struct nvkm_chan *, const struct nvkm_oclass *, + struct nvkm_object **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv40.c new file mode 100644 index 0000000000..179167484e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv40.c @@ -0,0 +1,83 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv31.h" + +#include + +#include + +bool +nv40_mpeg_mthd_dma(struct nvkm_device *device, u32 mthd, u32 data) +{ + struct nvkm_instmem *imem = device->imem; + struct nv31_mpeg *mpeg = nv31_mpeg(device->mpeg); + struct nvkm_subdev *subdev = &mpeg->engine.subdev; + u32 inst = data << 4; + u32 dma0 = nvkm_instmem_rd32(imem, inst + 0); + u32 dma1 = nvkm_instmem_rd32(imem, inst + 4); + u32 dma2 = nvkm_instmem_rd32(imem, inst + 8); + u32 base = (dma2 & 0xfffff000) | (dma0 >> 20); + u32 size = dma1 + 1; + + /* only allow linear DMA objects */ + if (!(dma0 & 0x00002000)) { + nvkm_error(subdev, "inst %08x dma0 %08x dma1 %08x dma2 %08x\n", + inst, dma0, dma1, dma2); + return false; + } + + if (mthd == 0x0190) { + /* DMA_CMD */ + nvkm_mask(device, 0x00b300, 0x00030000, (dma0 & 0x00030000)); + nvkm_wr32(device, 0x00b334, base); + nvkm_wr32(device, 0x00b324, size); + } else + if (mthd == 0x01a0) { + /* DMA_DATA */ + nvkm_mask(device, 0x00b300, 0x000c0000, (dma0 & 0x00030000) << 2); + nvkm_wr32(device, 0x00b360, base); + nvkm_wr32(device, 0x00b364, size); + } else { + /* DMA_IMAGE, VRAM only */ + if (dma0 & 0x00030000) + return false; + + nvkm_wr32(device, 0x00b370, base); + nvkm_wr32(device, 0x00b374, size); + } + + return true; +} + +static const struct nv31_mpeg_func +nv40_mpeg = { + .mthd_dma = nv40_mpeg_mthd_dma, +}; + +int +nv40_mpeg_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pmpeg) +{ + return nv31_mpeg_new_(&nv40_mpeg, device, type, inst, pmpeg); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv44.c b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv44.c new file mode 100644 index 0000000000..4b1374adbd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv44.c @@ -0,0 +1,216 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define nv44_mpeg(p) container_of((p), struct nv44_mpeg, engine) +#include "priv.h" + +#include +#include +#include + +#include + +struct nv44_mpeg { + struct nvkm_engine engine; + struct list_head chan; +}; + +/******************************************************************************* + * PMPEG context + ******************************************************************************/ +#define nv44_mpeg_chan(p) container_of((p), struct nv44_mpeg_chan, object) + +struct nv44_mpeg_chan { + struct nvkm_object object; + struct nv44_mpeg *mpeg; + struct nvkm_chan *fifo; + struct list_head head; + u32 inst; +}; + +static int +nv44_mpeg_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + struct nv44_mpeg_chan *chan = nv44_mpeg_chan(object); + int ret = nvkm_gpuobj_new(chan->object.engine->subdev.device, 264 * 4, + align, true, parent, pgpuobj); + if (ret == 0) { + chan->inst = (*pgpuobj)->addr; + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x78, 0x02001ec1); + nvkm_done(*pgpuobj); + } + return ret; +} + +static int +nv44_mpeg_chan_fini(struct nvkm_object *object, bool suspend) +{ + + struct nv44_mpeg_chan *chan = nv44_mpeg_chan(object); + struct nv44_mpeg *mpeg = chan->mpeg; + struct nvkm_device *device = mpeg->engine.subdev.device; + u32 inst = 0x80000000 | (chan->inst >> 4); + + nvkm_mask(device, 0x00b32c, 0x00000001, 0x00000000); + if (nvkm_rd32(device, 0x00b318) == inst) + nvkm_mask(device, 0x00b318, 0x80000000, 0x00000000); + nvkm_mask(device, 0x00b32c, 0x00000001, 0x00000001); + return 0; +} + +static void * +nv44_mpeg_chan_dtor(struct nvkm_object *object) +{ + struct nv44_mpeg_chan *chan = nv44_mpeg_chan(object); + struct nv44_mpeg *mpeg = chan->mpeg; + unsigned long flags; + spin_lock_irqsave(&mpeg->engine.lock, flags); + list_del(&chan->head); + spin_unlock_irqrestore(&mpeg->engine.lock, flags); + return chan; +} + +static const struct nvkm_object_func +nv44_mpeg_chan = { + .dtor = nv44_mpeg_chan_dtor, + .fini = nv44_mpeg_chan_fini, + .bind = nv44_mpeg_chan_bind, +}; + +static int +nv44_mpeg_chan_new(struct nvkm_chan *fifoch, const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nv44_mpeg *mpeg = nv44_mpeg(oclass->engine); + struct nv44_mpeg_chan *chan; + unsigned long flags; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv44_mpeg_chan, oclass, &chan->object); + chan->mpeg = mpeg; + chan->fifo = fifoch; + *pobject = &chan->object; + + spin_lock_irqsave(&mpeg->engine.lock, flags); + list_add(&chan->head, &mpeg->chan); + spin_unlock_irqrestore(&mpeg->engine.lock, flags); + return 0; +} + +/******************************************************************************* + * PMPEG engine/subdev functions + ******************************************************************************/ + +static bool +nv44_mpeg_mthd(struct nvkm_device *device, u32 mthd, u32 data) +{ + switch (mthd) { + case 0x190: + case 0x1a0: + case 0x1b0: + return nv40_mpeg_mthd_dma(device, mthd, data); + default: + break; + } + return false; +} + +static void +nv44_mpeg_intr(struct nvkm_engine *engine) +{ + struct nv44_mpeg *mpeg = nv44_mpeg(engine); + struct nvkm_subdev *subdev = &mpeg->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nv44_mpeg_chan *temp, *chan = NULL; + unsigned long flags; + u32 inst = nvkm_rd32(device, 0x00b318) & 0x000fffff; + u32 stat = nvkm_rd32(device, 0x00b100); + u32 type = nvkm_rd32(device, 0x00b230); + u32 mthd = nvkm_rd32(device, 0x00b234); + u32 data = nvkm_rd32(device, 0x00b238); + u32 show = stat; + + spin_lock_irqsave(&mpeg->engine.lock, flags); + list_for_each_entry(temp, &mpeg->chan, head) { + if (temp->inst >> 4 == inst) { + chan = temp; + list_del(&chan->head); + list_add(&chan->head, &mpeg->chan); + break; + } + } + + if (stat & 0x01000000) { + /* happens on initial binding of the object */ + if (type == 0x00000020 && mthd == 0x0000) { + nvkm_mask(device, 0x00b308, 0x00000000, 0x00000000); + show &= ~0x01000000; + } + + if (type == 0x00000010) { + if (nv44_mpeg_mthd(subdev->device, mthd, data)) + show &= ~0x01000000; + } + } + + nvkm_wr32(device, 0x00b100, stat); + nvkm_wr32(device, 0x00b230, 0x00000001); + + if (show) { + nvkm_error(subdev, "ch %d [%08x %s] %08x %08x %08x %08x\n", + chan ? chan->fifo->id : -1, inst << 4, + chan ? chan->fifo->name : "unknown", + stat, type, mthd, data); + } + + spin_unlock_irqrestore(&mpeg->engine.lock, flags); +} + +static const struct nvkm_engine_func +nv44_mpeg = { + .init = nv31_mpeg_init, + .intr = nv44_mpeg_intr, + .tile = nv31_mpeg_tile, + .fifo.cclass = nv44_mpeg_chan_new, + .sclass = { + { -1, -1, NV31_MPEG, &nv31_mpeg_object }, + {} + } +}; + +int +nv44_mpeg_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pmpeg) +{ + struct nv44_mpeg *mpeg; + + if (!(mpeg = kzalloc(sizeof(*mpeg), GFP_KERNEL))) + return -ENOMEM; + INIT_LIST_HEAD(&mpeg->chan); + *pmpeg = &mpeg->engine; + + return nvkm_engine_ctor(&nv44_mpeg, device, type, inst, true, &mpeg->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv50.c new file mode 100644 index 0000000000..e6374f3696 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/nv50.c @@ -0,0 +1,136 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include +#include +#include + +#include + +/******************************************************************************* + * PMPEG context + ******************************************************************************/ + +static int +nv50_mpeg_cclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 128 * 4, + align, true, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x70, 0x00801ec1); + nvkm_wo32(*pgpuobj, 0x7c, 0x0000037c); + nvkm_done(*pgpuobj); + } + return ret; +} + +const struct nvkm_object_func +nv50_mpeg_cclass = { + .bind = nv50_mpeg_cclass_bind, +}; + +/******************************************************************************* + * PMPEG engine/subdev functions + ******************************************************************************/ + +void +nv50_mpeg_intr(struct nvkm_engine *mpeg) +{ + struct nvkm_subdev *subdev = &mpeg->subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x00b100); + u32 type = nvkm_rd32(device, 0x00b230); + u32 mthd = nvkm_rd32(device, 0x00b234); + u32 data = nvkm_rd32(device, 0x00b238); + u32 show = stat; + + if (stat & 0x01000000) { + /* happens on initial binding of the object */ + if (type == 0x00000020 && mthd == 0x0000) { + nvkm_wr32(device, 0x00b308, 0x00000100); + show &= ~0x01000000; + } + } + + if (show) { + nvkm_info(subdev, "%08x %08x %08x %08x\n", + stat, type, mthd, data); + } + + nvkm_wr32(device, 0x00b100, stat); + nvkm_wr32(device, 0x00b230, 0x00000001); +} + +int +nv50_mpeg_init(struct nvkm_engine *mpeg) +{ + struct nvkm_subdev *subdev = &mpeg->subdev; + struct nvkm_device *device = subdev->device; + + nvkm_wr32(device, 0x00b32c, 0x00000000); + nvkm_wr32(device, 0x00b314, 0x00000100); + nvkm_wr32(device, 0x00b0e0, 0x0000001a); + + nvkm_wr32(device, 0x00b220, 0x00000044); + nvkm_wr32(device, 0x00b300, 0x00801ec1); + nvkm_wr32(device, 0x00b390, 0x00000000); + nvkm_wr32(device, 0x00b394, 0x00000000); + nvkm_wr32(device, 0x00b398, 0x00000000); + nvkm_mask(device, 0x00b32c, 0x00000001, 0x00000001); + + nvkm_wr32(device, 0x00b100, 0xffffffff); + nvkm_wr32(device, 0x00b140, 0xffffffff); + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x00b200) & 0x00000001)) + break; + ) < 0) { + nvkm_error(subdev, "timeout %08x\n", + nvkm_rd32(device, 0x00b200)); + return -EBUSY; + } + + return 0; +} + +static const struct nvkm_engine_func +nv50_mpeg = { + .init = nv50_mpeg_init, + .intr = nv50_mpeg_intr, + .cclass = &nv50_mpeg_cclass, + .sclass = { + { -1, -1, NV31_MPEG, &nv31_mpeg_object }, + {} + } +}; + +int +nv50_mpeg_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pmpeg) +{ + return nvkm_engine_new_(&nv50_mpeg, device, type, inst, true, pmpeg); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/priv.h new file mode 100644 index 0000000000..044ff41338 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mpeg/priv.h @@ -0,0 +1,17 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_MPEG_PRIV_H__ +#define __NVKM_MPEG_PRIV_H__ +#include +struct nvkm_chan; + +int nv31_mpeg_init(struct nvkm_engine *); +void nv31_mpeg_tile(struct nvkm_engine *, int, struct nvkm_fb_tile *); +extern const struct nvkm_object_func nv31_mpeg_object; + +bool nv40_mpeg_mthd_dma(struct nvkm_device *, u32, u32); + +int nv50_mpeg_init(struct nvkm_engine *); +void nv50_mpeg_intr(struct nvkm_engine *); + +extern const struct nvkm_object_func nv50_mpeg_cclass; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msenc/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/msenc/Kbuild new file mode 100644 index 0000000000..4bf033ff4d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msenc/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +#nvkm-y += nvkm/engine/msenc/base.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/Kbuild new file mode 100644 index 0000000000..0cd957d725 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/Kbuild @@ -0,0 +1,6 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/mspdec/base.o +nvkm-y += nvkm/engine/mspdec/g98.o +nvkm-y += nvkm/engine/mspdec/gt215.o +nvkm-y += nvkm/engine/mspdec/gf100.o +nvkm-y += nvkm/engine/mspdec/gk104.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/base.c new file mode 100644 index 0000000000..842fcfbd28 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/base.c @@ -0,0 +1,31 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +int +nvkm_mspdec_new_(const struct nvkm_falcon_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(func, device, type, inst, true, 0x085000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/g98.c b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/g98.c new file mode 100644 index 0000000000..ecb06d68f5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/g98.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +void +g98_mspdec_init(struct nvkm_falcon *mspdec) +{ + struct nvkm_device *device = mspdec->engine.subdev.device; + nvkm_wr32(device, 0x085010, 0x0000ffd2); + nvkm_wr32(device, 0x08501c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +g98_mspdec = { + .init = g98_mspdec_init, + .sclass = { + { -1, -1, G98_MSPDEC }, + {} + } +}; + +int +g98_mspdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_mspdec_new_(&g98_mspdec, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gf100.c new file mode 100644 index 0000000000..0a69bd767d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gf100.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Maarten Lankhorst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Maarten Lankhorst + */ +#include "priv.h" + +#include + +void +gf100_mspdec_init(struct nvkm_falcon *mspdec) +{ + struct nvkm_device *device = mspdec->engine.subdev.device; + nvkm_wr32(device, 0x085010, 0x0000fff2); + nvkm_wr32(device, 0x08501c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +gf100_mspdec = { + .init = gf100_mspdec_init, + .sclass = { + { -1, -1, GF100_MSPDEC }, + {} + } +}; + +int +gf100_mspdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_mspdec_new_(&gf100_mspdec, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gk104.c new file mode 100644 index 0000000000..a08991dca4 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gk104.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +gk104_mspdec = { + .init = gf100_mspdec_init, + .sclass = { + { -1, -1, GK104_MSPDEC }, + {} + } +}; + +int +gk104_mspdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_mspdec_new_(&gk104_mspdec, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gt215.c new file mode 100644 index 0000000000..791fb03a32 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/gt215.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +gt215_mspdec = { + .init = g98_mspdec_init, + .sclass = { + { -1, -1, GT212_MSPDEC }, + {} + } +}; + +int +gt215_mspdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_mspdec_new_(>215_mspdec, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/priv.h new file mode 100644 index 0000000000..2bc5537d40 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/mspdec/priv.h @@ -0,0 +1,12 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_MSPDEC_PRIV_H__ +#define __NVKM_MSPDEC_PRIV_H__ +#include + +int nvkm_mspdec_new_(const struct nvkm_falcon_func *, struct nvkm_device *, enum nvkm_subdev_type, + int, struct nvkm_engine **); + +void g98_mspdec_init(struct nvkm_falcon *); + +void gf100_mspdec_init(struct nvkm_falcon *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/Kbuild new file mode 100644 index 0000000000..788ce7255b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/Kbuild @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/msppp/base.o +nvkm-y += nvkm/engine/msppp/g98.o +nvkm-y += nvkm/engine/msppp/gt215.o +nvkm-y += nvkm/engine/msppp/gf100.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/base.c new file mode 100644 index 0000000000..45a9411ab2 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/base.c @@ -0,0 +1,31 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +int +nvkm_msppp_new_(const struct nvkm_falcon_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(func, device, type, inst, true, 0x086000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/g98.c b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/g98.c new file mode 100644 index 0000000000..160120b9bd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/g98.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +void +g98_msppp_init(struct nvkm_falcon *msppp) +{ + struct nvkm_device *device = msppp->engine.subdev.device; + nvkm_wr32(device, 0x086010, 0x0000ffd2); + nvkm_wr32(device, 0x08601c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +g98_msppp = { + .init = g98_msppp_init, + .sclass = { + { -1, -1, G98_MSPPP }, + {} + } +}; + +int +g98_msppp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msppp_new_(&g98_msppp, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gf100.c new file mode 100644 index 0000000000..debed9ae87 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gf100.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Maarten Lankhorst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Maarten Lankhorst + */ +#include "priv.h" + +#include + +static void +gf100_msppp_init(struct nvkm_falcon *msppp) +{ + struct nvkm_device *device = msppp->engine.subdev.device; + nvkm_wr32(device, 0x086010, 0x0000fff2); + nvkm_wr32(device, 0x08601c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +gf100_msppp = { + .init = gf100_msppp_init, + .sclass = { + { -1, -1, GF100_MSPPP }, + {} + } +}; + +int +gf100_msppp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msppp_new_(&gf100_msppp, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gt215.c new file mode 100644 index 0000000000..a2fd736fef --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/gt215.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +gt215_msppp = { + .init = g98_msppp_init, + .sclass = { + { -1, -1, GT212_MSPPP }, + {} + } +}; + +int +gt215_msppp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msppp_new_(>215_msppp, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msppp/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/priv.h new file mode 100644 index 0000000000..582ab8ce14 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msppp/priv.h @@ -0,0 +1,10 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_MSPPP_PRIV_H__ +#define __NVKM_MSPPP_PRIV_H__ +#include + +int nvkm_msppp_new_(const struct nvkm_falcon_func *, struct nvkm_device *, enum nvkm_subdev_type, + int, struct nvkm_engine **); + +void g98_msppp_init(struct nvkm_falcon *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/Kbuild new file mode 100644 index 0000000000..d68b4431cd --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/Kbuild @@ -0,0 +1,7 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/msvld/base.o +nvkm-y += nvkm/engine/msvld/g98.o +nvkm-y += nvkm/engine/msvld/gt215.o +nvkm-y += nvkm/engine/msvld/mcp89.o +nvkm-y += nvkm/engine/msvld/gf100.o +nvkm-y += nvkm/engine/msvld/gk104.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/base.c new file mode 100644 index 0000000000..7be42b980e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/base.c @@ -0,0 +1,31 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +int +nvkm_msvld_new_(const struct nvkm_falcon_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(func, device, type, inst, true, 0x084000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/g98.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/g98.c new file mode 100644 index 0000000000..cfa2065319 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/g98.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +void +g98_msvld_init(struct nvkm_falcon *msvld) +{ + struct nvkm_device *device = msvld->engine.subdev.device; + nvkm_wr32(device, 0x084010, 0x0000ffd2); + nvkm_wr32(device, 0x08401c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +g98_msvld = { + .init = g98_msvld_init, + .sclass = { + { -1, -1, G98_MSVLD }, + {} + } +}; + +int +g98_msvld_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msvld_new_(&g98_msvld, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gf100.c new file mode 100644 index 0000000000..8d58ad8e04 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gf100.c @@ -0,0 +1,50 @@ +/* + * Copyright 2012 Maarten Lankhorst + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Maarten Lankhorst + */ +#include "priv.h" + +#include + +void +gf100_msvld_init(struct nvkm_falcon *msvld) +{ + struct nvkm_device *device = msvld->engine.subdev.device; + nvkm_wr32(device, 0x084010, 0x0000fff2); + nvkm_wr32(device, 0x08401c, 0x0000fff2); +} + +static const struct nvkm_falcon_func +gf100_msvld = { + .init = gf100_msvld_init, + .sclass = { + { -1, -1, GF100_MSVLD }, + {} + } +}; + +int +gf100_msvld_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msvld_new_(&gf100_msvld, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gk104.c new file mode 100644 index 0000000000..b28be28046 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gk104.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +gk104_msvld = { + .init = gf100_msvld_init, + .sclass = { + { -1, -1, GK104_MSVLD }, + {} + } +}; + +int +gk104_msvld_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msvld_new_(&gk104_msvld, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gt215.c new file mode 100644 index 0000000000..d7489f972c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/gt215.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +gt215_msvld = { + .init = g98_msvld_init, + .sclass = { + { -1, -1, GT212_MSVLD }, + {} + } +}; + +int +gt215_msvld_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msvld_new_(>215_msvld, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/mcp89.c b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/mcp89.c new file mode 100644 index 0000000000..16c30b62ab --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/mcp89.c @@ -0,0 +1,42 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Maarten Lankhorst, Ilia Mirkin + */ +#include "priv.h" + +#include + +static const struct nvkm_falcon_func +mcp89_msvld = { + .init = g98_msvld_init, + .sclass = { + { -1, -1, IGT21A_MSVLD }, + {} + } +}; + +int +mcp89_msvld_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_msvld_new_(&mcp89_msvld, device, type, inst, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/msvld/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/priv.h new file mode 100644 index 0000000000..f729d919b0 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/msvld/priv.h @@ -0,0 +1,12 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_MSVLD_PRIV_H__ +#define __NVKM_MSVLD_PRIV_H__ +#include + +int nvkm_msvld_new_(const struct nvkm_falcon_func *, struct nvkm_device *, enum nvkm_subdev_type, + int, struct nvkm_engine **); + +void g98_msvld_init(struct nvkm_falcon *); + +void gf100_msvld_init(struct nvkm_falcon *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/Kbuild new file mode 100644 index 0000000000..f05e79670d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/Kbuild @@ -0,0 +1,4 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/nvdec/base.o +nvkm-y += nvkm/engine/nvdec/gm107.o +nvkm-y += nvkm/engine/nvdec/ga102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/base.c new file mode 100644 index 0000000000..1f6e3b32ba --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/base.c @@ -0,0 +1,61 @@ +/* + * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include + +static void * +nvkm_nvdec_dtor(struct nvkm_engine *engine) +{ + struct nvkm_nvdec *nvdec = nvkm_nvdec(engine); + nvkm_falcon_dtor(&nvdec->falcon); + return nvdec; +} + +static const struct nvkm_engine_func +nvkm_nvdec = { + .dtor = nvkm_nvdec_dtor, +}; + +int +nvkm_nvdec_new_(const struct nvkm_nvdec_fwif *fwif, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, u32 addr, struct nvkm_nvdec **pnvdec) +{ + struct nvkm_nvdec *nvdec; + int ret; + + if (!(nvdec = *pnvdec = kzalloc(sizeof(*nvdec), GFP_KERNEL))) + return -ENOMEM; + + ret = nvkm_engine_ctor(&nvkm_nvdec, device, type, inst, true, + &nvdec->engine); + if (ret) + return ret; + + fwif = nvkm_firmware_load(&nvdec->engine.subdev, fwif, "Nvdec", nvdec); + if (IS_ERR(fwif)) + return -ENODEV; + + nvdec->func = fwif->func; + + return nvkm_falcon_ctor(nvdec->func->flcn, &nvdec->engine.subdev, + nvdec->engine.subdev.name, addr, &nvdec->falcon); +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/ga102.c new file mode 100644 index 0000000000..37d8c3c0f3 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/ga102.c @@ -0,0 +1,61 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include +#include + +static const struct nvkm_falcon_func +ga102_nvdec_flcn = { + .disable = gm200_flcn_disable, + .enable = gm200_flcn_enable, + .addr2 = 0x1c00, + .reset_pmc = true, + .reset_prep = ga102_flcn_reset_prep, + .reset_wait_mem_scrubbing = ga102_flcn_reset_wait_mem_scrubbing, + .imem_dma = &ga102_flcn_dma, + .dmem_dma = &ga102_flcn_dma, +}; + +static const struct nvkm_nvdec_func +ga102_nvdec = { + .flcn = &ga102_nvdec_flcn, +}; + +static int +ga102_nvdec_nofw(struct nvkm_nvdec *nvdec, int ver, const struct nvkm_nvdec_fwif *fwif) +{ + return 0; +} + +static const struct nvkm_nvdec_fwif +ga102_nvdec_fwif[] = { + { -1, ga102_nvdec_nofw, &ga102_nvdec }, + {} +}; + +int +ga102_nvdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_nvdec **pnvdec) +{ + return nvkm_nvdec_new_(ga102_nvdec_fwif, device, type, inst, 0x848000, pnvdec); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/gm107.c new file mode 100644 index 0000000000..564f7e8960 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/gm107.c @@ -0,0 +1,58 @@ +/* + * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +static const struct nvkm_falcon_func +gm107_nvdec_flcn = { + .disable = gm200_flcn_disable, + .enable = gm200_flcn_enable, + .reset_pmc = true, + .reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing, + .debug = 0xd00, + .imem_pio = &gm200_flcn_imem_pio, + .dmem_pio = &gm200_flcn_dmem_pio, +}; + +static const struct nvkm_nvdec_func +gm107_nvdec = { + .flcn = &gm107_nvdec_flcn, +}; + +static int +gm107_nvdec_nofw(struct nvkm_nvdec *nvdec, int ver, + const struct nvkm_nvdec_fwif *fwif) +{ + return 0; +} + +static const struct nvkm_nvdec_fwif +gm107_nvdec_fwif[] = { + { -1, gm107_nvdec_nofw, &gm107_nvdec }, + {} +}; + +int +gm107_nvdec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_nvdec **pnvdec) +{ + return nvkm_nvdec_new_(gm107_nvdec_fwif, device, type, inst, 0, pnvdec); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/priv.h new file mode 100644 index 0000000000..61e1f7aaa5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvdec/priv.h @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_NVDEC_PRIV_H__ +#define __NVKM_NVDEC_PRIV_H__ +#include + +struct nvkm_nvdec_func { + const struct nvkm_falcon_func *flcn; +}; + +struct nvkm_nvdec_fwif { + int version; + int (*load)(struct nvkm_nvdec *, int ver, + const struct nvkm_nvdec_fwif *); + const struct nvkm_nvdec_func *func; +}; + +int nvkm_nvdec_new_(const struct nvkm_nvdec_fwif *fwif, struct nvkm_device *, + enum nvkm_subdev_type, int, u32 addr, struct nvkm_nvdec **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/Kbuild new file mode 100644 index 0000000000..75bf4436bf --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/Kbuild @@ -0,0 +1,3 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/nvenc/base.o +nvkm-y += nvkm/engine/nvenc/gm107.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/base.c new file mode 100644 index 0000000000..cf5dcfda7b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/base.c @@ -0,0 +1,62 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include + +static void * +nvkm_nvenc_dtor(struct nvkm_engine *engine) +{ + struct nvkm_nvenc *nvenc = nvkm_nvenc(engine); + nvkm_falcon_dtor(&nvenc->falcon); + return nvenc; +} + +static const struct nvkm_engine_func +nvkm_nvenc = { + .dtor = nvkm_nvenc_dtor, +}; + +int +nvkm_nvenc_new_(const struct nvkm_nvenc_fwif *fwif, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_nvenc **pnvenc) +{ + struct nvkm_nvenc *nvenc; + int ret; + + if (!(nvenc = *pnvenc = kzalloc(sizeof(*nvenc), GFP_KERNEL))) + return -ENOMEM; + + ret = nvkm_engine_ctor(&nvkm_nvenc, device, type, inst, true, + &nvenc->engine); + if (ret) + return ret; + + fwif = nvkm_firmware_load(&nvenc->engine.subdev, fwif, "Nvenc", nvenc); + if (IS_ERR(fwif)) + return -ENODEV; + + nvenc->func = fwif->func; + + return nvkm_falcon_ctor(nvenc->func->flcn, &nvenc->engine.subdev, + nvenc->engine.subdev.name, 0, &nvenc->falcon); +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/gm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/gm107.c new file mode 100644 index 0000000000..ad27d8b975 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/gm107.c @@ -0,0 +1,52 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "priv.h" + +static const struct nvkm_falcon_func +gm107_nvenc_flcn = { +}; + +static const struct nvkm_nvenc_func +gm107_nvenc = { + .flcn = &gm107_nvenc_flcn, +}; + +static int +gm107_nvenc_nofw(struct nvkm_nvenc *nvenc, int ver, + const struct nvkm_nvenc_fwif *fwif) +{ + return 0; +} + +static const struct nvkm_nvenc_fwif +gm107_nvenc_fwif[] = { + { -1, gm107_nvenc_nofw, &gm107_nvenc }, + {} +}; + +int +gm107_nvenc_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_nvenc **pnvenc) +{ + return nvkm_nvenc_new_(gm107_nvenc_fwif, device, type, inst, pnvenc); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/priv.h new file mode 100644 index 0000000000..4130a2bfbb --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/nvenc/priv.h @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_NVENC_PRIV_H__ +#define __NVKM_NVENC_PRIV_H__ +#include + +struct nvkm_nvenc_func { + const struct nvkm_falcon_func *flcn; +}; + +struct nvkm_nvenc_fwif { + int version; + int (*load)(struct nvkm_nvenc *, int ver, + const struct nvkm_nvenc_fwif *); + const struct nvkm_nvenc_func *func; +}; + +int nvkm_nvenc_new_(const struct nvkm_nvenc_fwif *, struct nvkm_device *, enum nvkm_subdev_type, + int, struct nvkm_nvenc **pnvenc); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/pm/Kbuild new file mode 100644 index 0000000000..2cc8a5f6fe --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/Kbuild @@ -0,0 +1,11 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/pm/base.o +nvkm-y += nvkm/engine/pm/nv40.o +nvkm-y += nvkm/engine/pm/nv50.o +nvkm-y += nvkm/engine/pm/g84.o +nvkm-y += nvkm/engine/pm/gt200.o +nvkm-y += nvkm/engine/pm/gt215.o +nvkm-y += nvkm/engine/pm/gf100.o +nvkm-y += nvkm/engine/pm/gf108.o +nvkm-y += nvkm/engine/pm/gf117.o +nvkm-y += nvkm/engine/pm/gk104.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c new file mode 100644 index 0000000000..8fe0444f76 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/base.c @@ -0,0 +1,868 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" + +#include +#include + +#include +#include +#include +#include +#include + +static u8 +nvkm_pm_count_perfdom(struct nvkm_pm *pm) +{ + struct nvkm_perfdom *dom; + u8 domain_nr = 0; + + list_for_each_entry(dom, &pm->domains, head) + domain_nr++; + return domain_nr; +} + +static u16 +nvkm_perfdom_count_perfsig(struct nvkm_perfdom *dom) +{ + u16 signal_nr = 0; + int i; + + if (dom) { + for (i = 0; i < dom->signal_nr; i++) { + if (dom->signal[i].name) + signal_nr++; + } + } + return signal_nr; +} + +static struct nvkm_perfdom * +nvkm_perfdom_find(struct nvkm_pm *pm, int di) +{ + struct nvkm_perfdom *dom; + int tmp = 0; + + list_for_each_entry(dom, &pm->domains, head) { + if (tmp++ == di) + return dom; + } + return NULL; +} + +static struct nvkm_perfsig * +nvkm_perfsig_find(struct nvkm_pm *pm, u8 di, u8 si, struct nvkm_perfdom **pdom) +{ + struct nvkm_perfdom *dom = *pdom; + + if (dom == NULL) { + dom = nvkm_perfdom_find(pm, di); + if (dom == NULL) + return NULL; + *pdom = dom; + } + + if (!dom->signal[si].name) + return NULL; + return &dom->signal[si]; +} + +static u8 +nvkm_perfsig_count_perfsrc(struct nvkm_perfsig *sig) +{ + u8 source_nr = 0, i; + + for (i = 0; i < ARRAY_SIZE(sig->source); i++) { + if (sig->source[i]) + source_nr++; + } + return source_nr; +} + +static struct nvkm_perfsrc * +nvkm_perfsrc_find(struct nvkm_pm *pm, struct nvkm_perfsig *sig, int si) +{ + struct nvkm_perfsrc *src; + bool found = false; + int tmp = 1; /* Sources ID start from 1 */ + u8 i; + + for (i = 0; i < ARRAY_SIZE(sig->source) && sig->source[i]; i++) { + if (sig->source[i] == si) { + found = true; + break; + } + } + + if (found) { + list_for_each_entry(src, &pm->sources, head) { + if (tmp++ == si) + return src; + } + } + + return NULL; +} + +static int +nvkm_perfsrc_enable(struct nvkm_pm *pm, struct nvkm_perfctr *ctr) +{ + struct nvkm_subdev *subdev = &pm->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_perfdom *dom = NULL; + struct nvkm_perfsig *sig; + struct nvkm_perfsrc *src; + u32 mask, value; + int i, j; + + for (i = 0; i < 4; i++) { + for (j = 0; j < 8 && ctr->source[i][j]; j++) { + sig = nvkm_perfsig_find(pm, ctr->domain, + ctr->signal[i], &dom); + if (!sig) + return -EINVAL; + + src = nvkm_perfsrc_find(pm, sig, ctr->source[i][j]); + if (!src) + return -EINVAL; + + /* set enable bit if needed */ + mask = value = 0x00000000; + if (src->enable) + mask = value = 0x80000000; + mask |= (src->mask << src->shift); + value |= ((ctr->source[i][j] >> 32) << src->shift); + + /* enable the source */ + nvkm_mask(device, src->addr, mask, value); + nvkm_debug(subdev, + "enabled source %08x %08x %08x\n", + src->addr, mask, value); + } + } + return 0; +} + +static int +nvkm_perfsrc_disable(struct nvkm_pm *pm, struct nvkm_perfctr *ctr) +{ + struct nvkm_subdev *subdev = &pm->engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_perfdom *dom = NULL; + struct nvkm_perfsig *sig; + struct nvkm_perfsrc *src; + u32 mask; + int i, j; + + for (i = 0; i < 4; i++) { + for (j = 0; j < 8 && ctr->source[i][j]; j++) { + sig = nvkm_perfsig_find(pm, ctr->domain, + ctr->signal[i], &dom); + if (!sig) + return -EINVAL; + + src = nvkm_perfsrc_find(pm, sig, ctr->source[i][j]); + if (!src) + return -EINVAL; + + /* unset enable bit if needed */ + mask = 0x00000000; + if (src->enable) + mask = 0x80000000; + mask |= (src->mask << src->shift); + + /* disable the source */ + nvkm_mask(device, src->addr, mask, 0); + nvkm_debug(subdev, "disabled source %08x %08x\n", + src->addr, mask); + } + } + return 0; +} + +/******************************************************************************* + * Perfdom object classes + ******************************************************************************/ +static int +nvkm_perfdom_init(struct nvkm_perfdom *dom, void *data, u32 size) +{ + union { + struct nvif_perfdom_init none; + } *args = data; + struct nvkm_object *object = &dom->object; + struct nvkm_pm *pm = dom->perfmon->pm; + int ret = -ENOSYS, i; + + nvif_ioctl(object, "perfdom init size %d\n", size); + if (!(ret = nvif_unvers(ret, &data, &size, args->none))) { + nvif_ioctl(object, "perfdom init\n"); + } else + return ret; + + for (i = 0; i < 4; i++) { + if (dom->ctr[i]) { + dom->func->init(pm, dom, dom->ctr[i]); + + /* enable sources */ + nvkm_perfsrc_enable(pm, dom->ctr[i]); + } + } + + /* start next batch of counters for sampling */ + dom->func->next(pm, dom); + return 0; +} + +static int +nvkm_perfdom_sample(struct nvkm_perfdom *dom, void *data, u32 size) +{ + union { + struct nvif_perfdom_sample none; + } *args = data; + struct nvkm_object *object = &dom->object; + struct nvkm_pm *pm = dom->perfmon->pm; + int ret = -ENOSYS; + + nvif_ioctl(object, "perfdom sample size %d\n", size); + if (!(ret = nvif_unvers(ret, &data, &size, args->none))) { + nvif_ioctl(object, "perfdom sample\n"); + } else + return ret; + pm->sequence++; + + /* sample previous batch of counters */ + list_for_each_entry(dom, &pm->domains, head) + dom->func->next(pm, dom); + + return 0; +} + +static int +nvkm_perfdom_read(struct nvkm_perfdom *dom, void *data, u32 size) +{ + union { + struct nvif_perfdom_read_v0 v0; + } *args = data; + struct nvkm_object *object = &dom->object; + struct nvkm_pm *pm = dom->perfmon->pm; + int ret = -ENOSYS, i; + + nvif_ioctl(object, "perfdom read size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "perfdom read vers %d\n", args->v0.version); + } else + return ret; + + for (i = 0; i < 4; i++) { + if (dom->ctr[i]) + dom->func->read(pm, dom, dom->ctr[i]); + } + + if (!dom->clk) + return -EAGAIN; + + for (i = 0; i < 4; i++) + if (dom->ctr[i]) + args->v0.ctr[i] = dom->ctr[i]->ctr; + args->v0.clk = dom->clk; + return 0; +} + +static int +nvkm_perfdom_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + struct nvkm_perfdom *dom = nvkm_perfdom(object); + switch (mthd) { + case NVIF_PERFDOM_V0_INIT: + return nvkm_perfdom_init(dom, data, size); + case NVIF_PERFDOM_V0_SAMPLE: + return nvkm_perfdom_sample(dom, data, size); + case NVIF_PERFDOM_V0_READ: + return nvkm_perfdom_read(dom, data, size); + default: + break; + } + return -EINVAL; +} + +static void * +nvkm_perfdom_dtor(struct nvkm_object *object) +{ + struct nvkm_perfdom *dom = nvkm_perfdom(object); + struct nvkm_pm *pm = dom->perfmon->pm; + int i; + + for (i = 0; i < 4; i++) { + struct nvkm_perfctr *ctr = dom->ctr[i]; + if (ctr) { + nvkm_perfsrc_disable(pm, ctr); + if (ctr->head.next) + list_del(&ctr->head); + } + kfree(ctr); + } + + return dom; +} + +static int +nvkm_perfctr_new(struct nvkm_perfdom *dom, int slot, u8 domain, + struct nvkm_perfsig *signal[4], u64 source[4][8], + u16 logic_op, struct nvkm_perfctr **pctr) +{ + struct nvkm_perfctr *ctr; + int i, j; + + if (!dom) + return -EINVAL; + + ctr = *pctr = kzalloc(sizeof(*ctr), GFP_KERNEL); + if (!ctr) + return -ENOMEM; + + ctr->domain = domain; + ctr->logic_op = logic_op; + ctr->slot = slot; + for (i = 0; i < 4; i++) { + if (signal[i]) { + ctr->signal[i] = signal[i] - dom->signal; + for (j = 0; j < 8; j++) + ctr->source[i][j] = source[i][j]; + } + } + list_add_tail(&ctr->head, &dom->list); + + return 0; +} + +static const struct nvkm_object_func +nvkm_perfdom = { + .dtor = nvkm_perfdom_dtor, + .mthd = nvkm_perfdom_mthd, +}; + +static int +nvkm_perfdom_new_(struct nvkm_perfmon *perfmon, + const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + union { + struct nvif_perfdom_v0 v0; + } *args = data; + struct nvkm_pm *pm = perfmon->pm; + struct nvkm_object *parent = oclass->parent; + struct nvkm_perfdom *sdom = NULL; + struct nvkm_perfctr *ctr[4] = {}; + struct nvkm_perfdom *dom; + int c, s, m; + int ret = -ENOSYS; + + nvif_ioctl(parent, "create perfdom size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(parent, "create perfdom vers %d dom %d mode %02x\n", + args->v0.version, args->v0.domain, args->v0.mode); + } else + return ret; + + for (c = 0; c < ARRAY_SIZE(args->v0.ctr); c++) { + struct nvkm_perfsig *sig[4] = {}; + u64 src[4][8] = {}; + + for (s = 0; s < ARRAY_SIZE(args->v0.ctr[c].signal); s++) { + sig[s] = nvkm_perfsig_find(pm, args->v0.domain, + args->v0.ctr[c].signal[s], + &sdom); + if (args->v0.ctr[c].signal[s] && !sig[s]) + return -EINVAL; + + for (m = 0; m < 8; m++) { + src[s][m] = args->v0.ctr[c].source[s][m]; + if (src[s][m] && !nvkm_perfsrc_find(pm, sig[s], + src[s][m])) + return -EINVAL; + } + } + + ret = nvkm_perfctr_new(sdom, c, args->v0.domain, sig, src, + args->v0.ctr[c].logic_op, &ctr[c]); + if (ret) + return ret; + } + + if (!sdom) + return -EINVAL; + + if (!(dom = kzalloc(sizeof(*dom), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nvkm_perfdom, oclass, &dom->object); + dom->perfmon = perfmon; + *pobject = &dom->object; + + dom->func = sdom->func; + dom->addr = sdom->addr; + dom->mode = args->v0.mode; + for (c = 0; c < ARRAY_SIZE(ctr); c++) + dom->ctr[c] = ctr[c]; + return 0; +} + +/******************************************************************************* + * Perfmon object classes + ******************************************************************************/ +static int +nvkm_perfmon_mthd_query_domain(struct nvkm_perfmon *perfmon, + void *data, u32 size) +{ + union { + struct nvif_perfmon_query_domain_v0 v0; + } *args = data; + struct nvkm_object *object = &perfmon->object; + struct nvkm_pm *pm = perfmon->pm; + struct nvkm_perfdom *dom; + u8 domain_nr; + int di, ret = -ENOSYS; + + nvif_ioctl(object, "perfmon query domain size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, "perfmon domain vers %d iter %02x\n", + args->v0.version, args->v0.iter); + di = (args->v0.iter & 0xff) - 1; + } else + return ret; + + domain_nr = nvkm_pm_count_perfdom(pm); + if (di >= (int)domain_nr) + return -EINVAL; + + if (di >= 0) { + dom = nvkm_perfdom_find(pm, di); + if (dom == NULL) + return -EINVAL; + + args->v0.id = di; + args->v0.signal_nr = nvkm_perfdom_count_perfsig(dom); + strncpy(args->v0.name, dom->name, sizeof(args->v0.name) - 1); + + /* Currently only global counters (PCOUNTER) are implemented + * but this will be different for local counters (MP). */ + args->v0.counter_nr = 4; + } + + if (++di < domain_nr) { + args->v0.iter = ++di; + return 0; + } + + args->v0.iter = 0xff; + return 0; +} + +static int +nvkm_perfmon_mthd_query_signal(struct nvkm_perfmon *perfmon, + void *data, u32 size) +{ + union { + struct nvif_perfmon_query_signal_v0 v0; + } *args = data; + struct nvkm_object *object = &perfmon->object; + struct nvkm_pm *pm = perfmon->pm; + struct nvkm_device *device = pm->engine.subdev.device; + struct nvkm_perfdom *dom; + struct nvkm_perfsig *sig; + const bool all = nvkm_boolopt(device->cfgopt, "NvPmShowAll", false); + const bool raw = nvkm_boolopt(device->cfgopt, "NvPmUnnamed", all); + int ret = -ENOSYS, si; + + nvif_ioctl(object, "perfmon query signal size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, + "perfmon query signal vers %d dom %d iter %04x\n", + args->v0.version, args->v0.domain, args->v0.iter); + si = (args->v0.iter & 0xffff) - 1; + } else + return ret; + + dom = nvkm_perfdom_find(pm, args->v0.domain); + if (dom == NULL || si >= (int)dom->signal_nr) + return -EINVAL; + + if (si >= 0) { + sig = &dom->signal[si]; + if (raw || !sig->name) { + snprintf(args->v0.name, sizeof(args->v0.name), + "/%s/%02x", dom->name, si); + } else { + strncpy(args->v0.name, sig->name, + sizeof(args->v0.name) - 1); + } + + args->v0.signal = si; + args->v0.source_nr = nvkm_perfsig_count_perfsrc(sig); + } + + while (++si < dom->signal_nr) { + if (all || dom->signal[si].name) { + args->v0.iter = ++si; + return 0; + } + } + + args->v0.iter = 0xffff; + return 0; +} + +static int +nvkm_perfmon_mthd_query_source(struct nvkm_perfmon *perfmon, + void *data, u32 size) +{ + union { + struct nvif_perfmon_query_source_v0 v0; + } *args = data; + struct nvkm_object *object = &perfmon->object; + struct nvkm_pm *pm = perfmon->pm; + struct nvkm_perfdom *dom = NULL; + struct nvkm_perfsig *sig; + struct nvkm_perfsrc *src; + u8 source_nr = 0; + int si, ret = -ENOSYS; + + nvif_ioctl(object, "perfmon query source size %d\n", size); + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + nvif_ioctl(object, + "perfmon source vers %d dom %d sig %02x iter %02x\n", + args->v0.version, args->v0.domain, args->v0.signal, + args->v0.iter); + si = (args->v0.iter & 0xff) - 1; + } else + return ret; + + sig = nvkm_perfsig_find(pm, args->v0.domain, args->v0.signal, &dom); + if (!sig) + return -EINVAL; + + source_nr = nvkm_perfsig_count_perfsrc(sig); + if (si >= (int)source_nr) + return -EINVAL; + + if (si >= 0) { + src = nvkm_perfsrc_find(pm, sig, sig->source[si]); + if (!src) + return -EINVAL; + + args->v0.source = sig->source[si]; + args->v0.mask = src->mask; + strncpy(args->v0.name, src->name, sizeof(args->v0.name) - 1); + } + + if (++si < source_nr) { + args->v0.iter = ++si; + return 0; + } + + args->v0.iter = 0xff; + return 0; +} + +static int +nvkm_perfmon_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + struct nvkm_perfmon *perfmon = nvkm_perfmon(object); + switch (mthd) { + case NVIF_PERFMON_V0_QUERY_DOMAIN: + return nvkm_perfmon_mthd_query_domain(perfmon, data, size); + case NVIF_PERFMON_V0_QUERY_SIGNAL: + return nvkm_perfmon_mthd_query_signal(perfmon, data, size); + case NVIF_PERFMON_V0_QUERY_SOURCE: + return nvkm_perfmon_mthd_query_source(perfmon, data, size); + default: + break; + } + return -EINVAL; +} + +static int +nvkm_perfmon_child_new(const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + struct nvkm_perfmon *perfmon = nvkm_perfmon(oclass->parent); + return nvkm_perfdom_new_(perfmon, oclass, data, size, pobject); +} + +static int +nvkm_perfmon_child_get(struct nvkm_object *object, int index, + struct nvkm_oclass *oclass) +{ + if (index == 0) { + oclass->base.oclass = NVIF_CLASS_PERFDOM; + oclass->base.minver = 0; + oclass->base.maxver = 0; + oclass->ctor = nvkm_perfmon_child_new; + return 0; + } + return -EINVAL; +} + +static void * +nvkm_perfmon_dtor(struct nvkm_object *object) +{ + struct nvkm_perfmon *perfmon = nvkm_perfmon(object); + struct nvkm_pm *pm = perfmon->pm; + spin_lock(&pm->client.lock); + if (pm->client.object == &perfmon->object) + pm->client.object = NULL; + spin_unlock(&pm->client.lock); + return perfmon; +} + +static const struct nvkm_object_func +nvkm_perfmon = { + .dtor = nvkm_perfmon_dtor, + .mthd = nvkm_perfmon_mthd, + .sclass = nvkm_perfmon_child_get, +}; + +static int +nvkm_perfmon_new(struct nvkm_pm *pm, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + struct nvkm_perfmon *perfmon; + + if (!(perfmon = kzalloc(sizeof(*perfmon), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nvkm_perfmon, oclass, &perfmon->object); + perfmon->pm = pm; + *pobject = &perfmon->object; + return 0; +} + +/******************************************************************************* + * PPM engine/subdev functions + ******************************************************************************/ + +static int +nvkm_pm_oclass_new(struct nvkm_device *device, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + struct nvkm_pm *pm = nvkm_pm(oclass->engine); + int ret; + + ret = nvkm_perfmon_new(pm, oclass, data, size, pobject); + if (ret) + return ret; + + spin_lock(&pm->client.lock); + if (pm->client.object == NULL) + pm->client.object = *pobject; + ret = (pm->client.object == *pobject) ? 0 : -EBUSY; + spin_unlock(&pm->client.lock); + return ret; +} + +static const struct nvkm_device_oclass +nvkm_pm_oclass = { + .base.oclass = NVIF_CLASS_PERFMON, + .base.minver = -1, + .base.maxver = -1, + .ctor = nvkm_pm_oclass_new, +}; + +static int +nvkm_pm_oclass_get(struct nvkm_oclass *oclass, int index, + const struct nvkm_device_oclass **class) +{ + if (index == 0) { + oclass->base = nvkm_pm_oclass.base; + *class = &nvkm_pm_oclass; + return index; + } + return 1; +} + +static int +nvkm_perfsrc_new(struct nvkm_pm *pm, struct nvkm_perfsig *sig, + const struct nvkm_specsrc *spec) +{ + const struct nvkm_specsrc *ssrc; + const struct nvkm_specmux *smux; + struct nvkm_perfsrc *src; + u8 source_nr = 0; + + if (!spec) { + /* No sources are defined for this signal. */ + return 0; + } + + ssrc = spec; + while (ssrc->name) { + smux = ssrc->mux; + while (smux->name) { + bool found = false; + u8 source_id = 0; + u32 len; + + list_for_each_entry(src, &pm->sources, head) { + if (src->addr == ssrc->addr && + src->shift == smux->shift) { + found = true; + break; + } + source_id++; + } + + if (!found) { + src = kzalloc(sizeof(*src), GFP_KERNEL); + if (!src) + return -ENOMEM; + + src->addr = ssrc->addr; + src->mask = smux->mask; + src->shift = smux->shift; + src->enable = smux->enable; + + len = strlen(ssrc->name) + + strlen(smux->name) + 2; + src->name = kzalloc(len, GFP_KERNEL); + if (!src->name) { + kfree(src); + return -ENOMEM; + } + snprintf(src->name, len, "%s_%s", ssrc->name, + smux->name); + + list_add_tail(&src->head, &pm->sources); + } + + sig->source[source_nr++] = source_id + 1; + smux++; + } + ssrc++; + } + + return 0; +} + +int +nvkm_perfdom_new(struct nvkm_pm *pm, const char *name, u32 mask, + u32 base, u32 size_unit, u32 size_domain, + const struct nvkm_specdom *spec) +{ + const struct nvkm_specdom *sdom; + const struct nvkm_specsig *ssig; + struct nvkm_perfdom *dom; + int ret, i; + + for (i = 0; i == 0 || mask; i++) { + u32 addr = base + (i * size_unit); + if (i && !(mask & (1 << i))) + continue; + + sdom = spec; + while (sdom->signal_nr) { + dom = kzalloc(struct_size(dom, signal, sdom->signal_nr), + GFP_KERNEL); + if (!dom) + return -ENOMEM; + + if (mask) { + snprintf(dom->name, sizeof(dom->name), + "%s/%02x/%02x", name, i, + (int)(sdom - spec)); + } else { + snprintf(dom->name, sizeof(dom->name), + "%s/%02x", name, (int)(sdom - spec)); + } + + list_add_tail(&dom->head, &pm->domains); + INIT_LIST_HEAD(&dom->list); + dom->func = sdom->func; + dom->addr = addr; + dom->signal_nr = sdom->signal_nr; + + ssig = (sdom++)->signal; + while (ssig->name) { + struct nvkm_perfsig *sig = + &dom->signal[ssig->signal]; + sig->name = ssig->name; + ret = nvkm_perfsrc_new(pm, sig, ssig->source); + if (ret) + return ret; + ssig++; + } + + addr += size_domain; + } + + mask &= ~(1 << i); + } + + return 0; +} + +static int +nvkm_pm_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_pm *pm = nvkm_pm(engine); + if (pm->func->fini) + pm->func->fini(pm); + return 0; +} + +static void * +nvkm_pm_dtor(struct nvkm_engine *engine) +{ + struct nvkm_pm *pm = nvkm_pm(engine); + struct nvkm_perfdom *dom, *next_dom; + struct nvkm_perfsrc *src, *next_src; + + list_for_each_entry_safe(dom, next_dom, &pm->domains, head) { + list_del(&dom->head); + kfree(dom); + } + + list_for_each_entry_safe(src, next_src, &pm->sources, head) { + list_del(&src->head); + kfree(src->name); + kfree(src); + } + + return pm; +} + +static const struct nvkm_engine_func +nvkm_pm = { + .dtor = nvkm_pm_dtor, + .fini = nvkm_pm_fini, + .base.sclass = nvkm_pm_oclass_get, +}; + +int +nvkm_pm_ctor(const struct nvkm_pm_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_pm *pm) +{ + pm->func = func; + INIT_LIST_HEAD(&pm->domains); + INIT_LIST_HEAD(&pm->sources); + spin_lock_init(&pm->client.lock); + return nvkm_engine_ctor(&nvkm_pm, device, type, inst, true, &pm->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/g84.c new file mode 100644 index 0000000000..0086d00eb1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/g84.c @@ -0,0 +1,165 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" + +const struct nvkm_specsrc +g84_vfetch_sources[] = { + { 0x400c0c, (const struct nvkm_specmux[]) { + { 0x3, 0, "unk0" }, + {} + }, "pgraph_vfetch_unk0c" }, + {} +}; + +static const struct nvkm_specsrc +g84_prop_sources[] = { + { 0x408e50, (const struct nvkm_specmux[]) { + { 0x1f, 0, "sel", true }, + {} + }, "pgraph_tpc0_prop_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +g84_crop_sources[] = { + { 0x407008, (const struct nvkm_specmux[]) { + { 0xf, 0, "sel0", true }, + { 0x7, 16, "sel1", true }, + {} + }, "pgraph_rop0_crop_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +g84_tex_sources[] = { + { 0x408808, (const struct nvkm_specmux[]) { + { 0xfffff, 0, "unk0" }, + {} + }, "pgraph_tpc0_tex_unk08" }, + {} +}; + +static const struct nvkm_specdom +g84_pm[] = { + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0xbd, "pc01_gr_idle" }, + { 0x5e, "pc01_strmout_00" }, + { 0x5f, "pc01_strmout_01" }, + { 0xd2, "pc01_trast_00" }, + { 0xd3, "pc01_trast_01" }, + { 0xd4, "pc01_trast_02" }, + { 0xd5, "pc01_trast_03" }, + { 0xd8, "pc01_trast_04" }, + { 0xd9, "pc01_trast_05" }, + { 0x5c, "pc01_vattr_00" }, + { 0x5d, "pc01_vattr_01" }, + { 0x66, "pc01_vfetch_00", g84_vfetch_sources }, + { 0x67, "pc01_vfetch_01", g84_vfetch_sources }, + { 0x68, "pc01_vfetch_02", g84_vfetch_sources }, + { 0x69, "pc01_vfetch_03", g84_vfetch_sources }, + { 0x6a, "pc01_vfetch_04", g84_vfetch_sources }, + { 0x6b, "pc01_vfetch_05", g84_vfetch_sources }, + { 0x6c, "pc01_vfetch_06", g84_vfetch_sources }, + { 0x6d, "pc01_vfetch_07", g84_vfetch_sources }, + { 0x6e, "pc01_vfetch_08", g84_vfetch_sources }, + { 0x6f, "pc01_vfetch_09", g84_vfetch_sources }, + { 0x70, "pc01_vfetch_0a", g84_vfetch_sources }, + { 0x71, "pc01_vfetch_0b", g84_vfetch_sources }, + { 0x72, "pc01_vfetch_0c", g84_vfetch_sources }, + { 0x73, "pc01_vfetch_0d", g84_vfetch_sources }, + { 0x74, "pc01_vfetch_0e", g84_vfetch_sources }, + { 0x75, "pc01_vfetch_0f", g84_vfetch_sources }, + { 0x76, "pc01_vfetch_10", g84_vfetch_sources }, + { 0x77, "pc01_vfetch_11", g84_vfetch_sources }, + { 0x78, "pc01_vfetch_12", g84_vfetch_sources }, + { 0x79, "pc01_vfetch_13", g84_vfetch_sources }, + { 0x7a, "pc01_vfetch_14", g84_vfetch_sources }, + { 0x7b, "pc01_vfetch_15", g84_vfetch_sources }, + { 0x7c, "pc01_vfetch_16", g84_vfetch_sources }, + { 0x7d, "pc01_vfetch_17", g84_vfetch_sources }, + { 0x7e, "pc01_vfetch_18", g84_vfetch_sources }, + { 0x7f, "pc01_vfetch_19", g84_vfetch_sources }, + { 0x07, "pc01_zcull_00", nv50_zcull_sources }, + { 0x08, "pc01_zcull_01", nv50_zcull_sources }, + { 0x09, "pc01_zcull_02", nv50_zcull_sources }, + { 0x0a, "pc01_zcull_03", nv50_zcull_sources }, + { 0x0b, "pc01_zcull_04", nv50_zcull_sources }, + { 0x0c, "pc01_zcull_05", nv50_zcull_sources }, + { 0xa4, "pc01_unk00" }, + { 0xec, "pc01_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0xa0, (const struct nvkm_specsig[]) { + { 0x30, "pc02_crop_00", g84_crop_sources }, + { 0x31, "pc02_crop_01", g84_crop_sources }, + { 0x32, "pc02_crop_02", g84_crop_sources }, + { 0x33, "pc02_crop_03", g84_crop_sources }, + { 0x00, "pc02_prop_00", g84_prop_sources }, + { 0x01, "pc02_prop_01", g84_prop_sources }, + { 0x02, "pc02_prop_02", g84_prop_sources }, + { 0x03, "pc02_prop_03", g84_prop_sources }, + { 0x04, "pc02_prop_04", g84_prop_sources }, + { 0x05, "pc02_prop_05", g84_prop_sources }, + { 0x06, "pc02_prop_06", g84_prop_sources }, + { 0x07, "pc02_prop_07", g84_prop_sources }, + { 0x48, "pc02_tex_00", g84_tex_sources }, + { 0x49, "pc02_tex_01", g84_tex_sources }, + { 0x4a, "pc02_tex_02", g84_tex_sources }, + { 0x4b, "pc02_tex_03", g84_tex_sources }, + { 0x1a, "pc02_tex_04", g84_tex_sources }, + { 0x1b, "pc02_tex_05", g84_tex_sources }, + { 0x1c, "pc02_tex_06", g84_tex_sources }, + { 0x44, "pc02_zrop_00", nv50_zrop_sources }, + { 0x45, "pc02_zrop_01", nv50_zrop_sources }, + { 0x46, "pc02_zrop_02", nv50_zrop_sources }, + { 0x47, "pc02_zrop_03", nv50_zrop_sources }, + { 0x8c, "pc02_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + {} +}; + +int +g84_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return nv40_pm_new_(g84_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c new file mode 100644 index 0000000000..8e02701def --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.c @@ -0,0 +1,243 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" + +const struct nvkm_specsrc +gf100_pbfb_sources[] = { + { 0x10f100, (const struct nvkm_specmux[]) { + { 0x1, 0, "unk0" }, + { 0x3f, 4, "unk4" }, + {} + }, "pbfb_broadcast_pm_unk100" }, + {} +}; + +const struct nvkm_specsrc +gf100_pmfb_sources[] = { + { 0x140028, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + { 0x7, 16, "unk16" }, + { 0x3, 24, "unk24" }, + { 0x2, 29, "unk29" }, + {} + }, "pmfb0_pm_unk28" }, + {} +}; + +static const struct nvkm_specsrc +gf100_l1_sources[] = { + { 0x5044a8, (const struct nvkm_specmux[]) { + { 0x3f, 0, "sel", true }, + {} + }, "pgraph_gpc0_tpc0_l1_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +gf100_tex_sources[] = { + { 0x5042c0, (const struct nvkm_specmux[]) { + { 0xf, 0, "sel0", true }, + { 0x7, 8, "sel1", true }, + {} + }, "pgraph_gpc0_tpc0_tex_pm_mux_c_d" }, + {} +}; + +static const struct nvkm_specsrc +gf100_unk400_sources[] = { + { 0x50440c, (const struct nvkm_specmux[]) { + { 0x3f, 0, "sel", true }, + {} + }, "pgraph_gpc0_tpc0_unk400_pm_mux" }, + {} +}; + +static const struct nvkm_specdom +gf100_pm_hub[] = { + {} +}; + +const struct nvkm_specdom +gf100_pm_gpc[] = { + { 0xe0, (const struct nvkm_specsig[]) { + { 0x00, "gpc00_l1_00", gf100_l1_sources }, + { 0x01, "gpc00_l1_01", gf100_l1_sources }, + { 0x02, "gpc00_l1_02", gf100_l1_sources }, + { 0x03, "gpc00_l1_03", gf100_l1_sources }, + { 0x05, "gpc00_l1_04", gf100_l1_sources }, + { 0x06, "gpc00_l1_05", gf100_l1_sources }, + { 0x0a, "gpc00_tex_00", gf100_tex_sources }, + { 0x0b, "gpc00_tex_01", gf100_tex_sources }, + { 0x0c, "gpc00_tex_02", gf100_tex_sources }, + { 0x0d, "gpc00_tex_03", gf100_tex_sources }, + { 0x0e, "gpc00_tex_04", gf100_tex_sources }, + { 0x0f, "gpc00_tex_05", gf100_tex_sources }, + { 0x10, "gpc00_tex_06", gf100_tex_sources }, + { 0x11, "gpc00_tex_07", gf100_tex_sources }, + { 0x12, "gpc00_tex_08", gf100_tex_sources }, + { 0x26, "gpc00_unk400_00", gf100_unk400_sources }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct nvkm_specdom +gf100_pm_part[] = { + { 0xe0, (const struct nvkm_specsig[]) { + { 0x0f, "part00_pbfb_00", gf100_pbfb_sources }, + { 0x10, "part00_pbfb_01", gf100_pbfb_sources }, + { 0x21, "part00_pmfb_00", gf100_pmfb_sources }, + { 0x04, "part00_pmfb_01", gf100_pmfb_sources }, + { 0x00, "part00_pmfb_02", gf100_pmfb_sources }, + { 0x02, "part00_pmfb_03", gf100_pmfb_sources }, + { 0x01, "part00_pmfb_04", gf100_pmfb_sources }, + { 0x2e, "part00_pmfb_05", gf100_pmfb_sources }, + { 0x2f, "part00_pmfb_06", gf100_pmfb_sources }, + { 0x1b, "part00_pmfb_07", gf100_pmfb_sources }, + { 0x1c, "part00_pmfb_08", gf100_pmfb_sources }, + { 0x1d, "part00_pmfb_09", gf100_pmfb_sources }, + { 0x1e, "part00_pmfb_0a", gf100_pmfb_sources }, + { 0x1f, "part00_pmfb_0b", gf100_pmfb_sources }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static void +gf100_perfctr_init(struct nvkm_pm *pm, struct nvkm_perfdom *dom, + struct nvkm_perfctr *ctr) +{ + struct nvkm_device *device = pm->engine.subdev.device; + u32 log = ctr->logic_op; + u32 src = 0x00000000; + int i; + + for (i = 0; i < 4; i++) + src |= ctr->signal[i] << (i * 8); + + nvkm_wr32(device, dom->addr + 0x09c, 0x00040002 | (dom->mode << 3)); + nvkm_wr32(device, dom->addr + 0x100, 0x00000000); + nvkm_wr32(device, dom->addr + 0x040 + (ctr->slot * 0x08), src); + nvkm_wr32(device, dom->addr + 0x044 + (ctr->slot * 0x08), log); +} + +static void +gf100_perfctr_read(struct nvkm_pm *pm, struct nvkm_perfdom *dom, + struct nvkm_perfctr *ctr) +{ + struct nvkm_device *device = pm->engine.subdev.device; + + switch (ctr->slot) { + case 0: ctr->ctr = nvkm_rd32(device, dom->addr + 0x08c); break; + case 1: ctr->ctr = nvkm_rd32(device, dom->addr + 0x088); break; + case 2: ctr->ctr = nvkm_rd32(device, dom->addr + 0x080); break; + case 3: ctr->ctr = nvkm_rd32(device, dom->addr + 0x090); break; + } + dom->clk = nvkm_rd32(device, dom->addr + 0x070); +} + +static void +gf100_perfctr_next(struct nvkm_pm *pm, struct nvkm_perfdom *dom) +{ + struct nvkm_device *device = pm->engine.subdev.device; + nvkm_wr32(device, dom->addr + 0x06c, dom->signal_nr - 0x40 + 0x27); + nvkm_wr32(device, dom->addr + 0x0ec, 0x00000011); +} + +const struct nvkm_funcdom +gf100_perfctr_func = { + .init = gf100_perfctr_init, + .read = gf100_perfctr_read, + .next = gf100_perfctr_next, +}; + +static void +gf100_pm_fini(struct nvkm_pm *pm) +{ + struct nvkm_device *device = pm->engine.subdev.device; + nvkm_mask(device, 0x000200, 0x10000000, 0x00000000); + nvkm_mask(device, 0x000200, 0x10000000, 0x10000000); +} + +static const struct nvkm_pm_func +gf100_pm_ = { + .fini = gf100_pm_fini, +}; + +int +gf100_pm_new_(const struct gf100_pm_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + struct nvkm_pm *pm; + u32 mask; + int ret; + + if (!(pm = *ppm = kzalloc(sizeof(*pm), GFP_KERNEL))) + return -ENOMEM; + + ret = nvkm_pm_ctor(&gf100_pm_, device, type, inst, pm); + if (ret) + return ret; + + /* HUB */ + ret = nvkm_perfdom_new(pm, "hub", 0, 0x1b0000, 0, 0x200, + func->doms_hub); + if (ret) + return ret; + + /* GPC */ + mask = (1 << nvkm_rd32(device, 0x022430)) - 1; + mask &= ~nvkm_rd32(device, 0x022504); + mask &= ~nvkm_rd32(device, 0x022584); + + ret = nvkm_perfdom_new(pm, "gpc", mask, 0x180000, + 0x1000, 0x200, func->doms_gpc); + if (ret) + return ret; + + /* PART */ + mask = (1 << nvkm_rd32(device, 0x022438)) - 1; + mask &= ~nvkm_rd32(device, 0x022548); + mask &= ~nvkm_rd32(device, 0x0225c8); + + ret = nvkm_perfdom_new(pm, "part", mask, 0x1a0000, + 0x1000, 0x200, func->doms_part); + if (ret) + return ret; + + return 0; +} + +static const struct gf100_pm_func +gf100_pm = { + .doms_gpc = gf100_pm_gpc, + .doms_hub = gf100_pm_hub, + .doms_part = gf100_pm_part, +}; + +int +gf100_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return gf100_pm_new_(&gf100_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.h new file mode 100644 index 0000000000..bc4b014c4e --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf100.h @@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_PM_NVC0_H__ +#define __NVKM_PM_NVC0_H__ +#include "priv.h" + +struct gf100_pm_func { + const struct nvkm_specdom *doms_hub; + const struct nvkm_specdom *doms_gpc; + const struct nvkm_specdom *doms_part; +}; + +int gf100_pm_new_(const struct gf100_pm_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_pm **); + +extern const struct nvkm_funcdom gf100_perfctr_func; +extern const struct nvkm_specdom gf100_pm_gpc[]; + +extern const struct nvkm_specsrc gf100_pbfb_sources[]; +extern const struct nvkm_specsrc gf100_pmfb_sources[]; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf108.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf108.c new file mode 100644 index 0000000000..505565866b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf108.c @@ -0,0 +1,66 @@ +/* + * Copyright 2015 Samuel Pitoiset + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Samuel Pitoiset + */ +#include "gf100.h" + +static const struct nvkm_specdom +gf108_pm_hub[] = { + {} +}; + +static const struct nvkm_specdom +gf108_pm_part[] = { + { 0xe0, (const struct nvkm_specsig[]) { + { 0x14, "part00_pbfb_00", gf100_pbfb_sources }, + { 0x15, "part00_pbfb_01", gf100_pbfb_sources }, + { 0x20, "part00_pbfb_02", gf100_pbfb_sources }, + { 0x21, "part00_pbfb_03", gf100_pbfb_sources }, + { 0x01, "part00_pmfb_00", gf100_pmfb_sources }, + { 0x04, "part00_pmfb_01", gf100_pmfb_sources }, + { 0x05, "part00_pmfb_02", gf100_pmfb_sources}, + { 0x07, "part00_pmfb_03", gf100_pmfb_sources }, + { 0x0d, "part00_pmfb_04", gf100_pmfb_sources }, + { 0x12, "part00_pmfb_05", gf100_pmfb_sources }, + { 0x13, "part00_pmfb_06", gf100_pmfb_sources }, + { 0x2c, "part00_pmfb_07", gf100_pmfb_sources }, + { 0x2d, "part00_pmfb_08", gf100_pmfb_sources }, + { 0x2e, "part00_pmfb_09", gf100_pmfb_sources }, + { 0x2f, "part00_pmfb_0a", gf100_pmfb_sources }, + { 0x30, "part00_pmfb_0b", gf100_pmfb_sources }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct gf100_pm_func +gf108_pm = { + .doms_gpc = gf100_pm_gpc, + .doms_hub = gf108_pm_hub, + .doms_part = gf108_pm_part, +}; + +int +gf108_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return gf100_pm_new_(&gf108_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf117.c new file mode 100644 index 0000000000..c61e8c010b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gf117.c @@ -0,0 +1,80 @@ +/* + * Copyright 2015 Samuel Pitoiset + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Samuel Pitoiset + */ +#include "gf100.h" + +static const struct nvkm_specsrc +gf117_pmfb_sources[] = { + { 0x140028, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + { 0x7, 16, "unk16" }, + { 0x3, 24, "unk24" }, + { 0x2, 28, "unk28" }, + {} + }, "pmfb0_pm_unk28" }, + { 0x14125c, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pmfb0_subp0_pm_unk25c" }, + {} +}; + +static const struct nvkm_specdom +gf117_pm_hub[] = { + {} +}; + +static const struct nvkm_specdom +gf117_pm_part[] = { + { 0xe0, (const struct nvkm_specsig[]) { + { 0x00, "part00_pbfb_00", gf100_pbfb_sources }, + { 0x01, "part00_pbfb_01", gf100_pbfb_sources }, + { 0x12, "part00_pmfb_00", gf117_pmfb_sources }, + { 0x15, "part00_pmfb_01", gf117_pmfb_sources }, + { 0x16, "part00_pmfb_02", gf117_pmfb_sources }, + { 0x18, "part00_pmfb_03", gf117_pmfb_sources }, + { 0x1e, "part00_pmfb_04", gf117_pmfb_sources }, + { 0x23, "part00_pmfb_05", gf117_pmfb_sources }, + { 0x24, "part00_pmfb_06", gf117_pmfb_sources }, + { 0x0c, "part00_pmfb_07", gf117_pmfb_sources }, + { 0x0d, "part00_pmfb_08", gf117_pmfb_sources }, + { 0x0e, "part00_pmfb_09", gf117_pmfb_sources }, + { 0x0f, "part00_pmfb_0a", gf117_pmfb_sources }, + { 0x10, "part00_pmfb_0b", gf117_pmfb_sources }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct gf100_pm_func +gf117_pm = { + .doms_gpc = gf100_pm_gpc, + .doms_hub = gf117_pm_hub, + .doms_part = gf117_pm_part, +}; + +int +gf117_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return gf100_pm_new_(&gf117_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gk104.c new file mode 100644 index 0000000000..75bf3df1cb --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gk104.c @@ -0,0 +1,184 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "gf100.h" + +static const struct nvkm_specsrc +gk104_pmfb_sources[] = { + { 0x140028, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + { 0x7, 16, "unk16" }, + { 0x3, 24, "unk24" }, + { 0x2, 28, "unk28" }, + {} + }, "pmfb0_pm_unk28" }, + { 0x14125c, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pmfb0_subp0_pm_unk25c" }, + { 0x14165c, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pmfb0_subp1_pm_unk25c" }, + { 0x141a5c, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pmfb0_subp2_pm_unk25c" }, + { 0x141e5c, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pmfb0_subp3_pm_unk25c" }, + {} +}; + +static const struct nvkm_specsrc +gk104_tex_sources[] = { + { 0x5042c0, (const struct nvkm_specmux[]) { + { 0xf, 0, "sel0", true }, + { 0x7, 8, "sel1", true }, + {} + }, "pgraph_gpc0_tpc0_tex_pm_mux_c_d" }, + { 0x5042c8, (const struct nvkm_specmux[]) { + { 0x1f, 0, "sel", true }, + {} + }, "pgraph_gpc0_tpc0_tex_pm_unkc8" }, + { 0x5042b8, (const struct nvkm_specmux[]) { + { 0xff, 0, "sel", true }, + {} + }, "pgraph_gpc0_tpc0_tex_pm_unkb8" }, + {} +}; + +static const struct nvkm_specdom +gk104_pm_hub[] = { + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "hub00_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x40, (const struct nvkm_specsig[]) { + { 0x27, "hub01_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "hub02_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "hub03_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x40, (const struct nvkm_specsig[]) { + { 0x03, "host_mmio_rd" }, + { 0x27, "hub04_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "hub05_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0xc0, (const struct nvkm_specsig[]) { + { 0x74, "host_fb_rd3x" }, + { 0x75, "host_fb_rd3x_2" }, + { 0xa7, "hub06_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "hub07_user_0" }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct nvkm_specdom +gk104_pm_gpc[] = { + { 0xe0, (const struct nvkm_specsig[]) { + { 0xc7, "gpc00_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &gf100_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + { 0x00, "gpc02_tex_00", gk104_tex_sources }, + { 0x01, "gpc02_tex_01", gk104_tex_sources }, + { 0x02, "gpc02_tex_02", gk104_tex_sources }, + { 0x03, "gpc02_tex_03", gk104_tex_sources }, + { 0x04, "gpc02_tex_04", gk104_tex_sources }, + { 0x05, "gpc02_tex_05", gk104_tex_sources }, + { 0x06, "gpc02_tex_06", gk104_tex_sources }, + { 0x07, "gpc02_tex_07", gk104_tex_sources }, + { 0x08, "gpc02_tex_08", gk104_tex_sources }, + { 0x0a, "gpc02_tex_0a", gk104_tex_sources }, + { 0x0b, "gpc02_tex_0b", gk104_tex_sources }, + { 0x0d, "gpc02_tex_0c", gk104_tex_sources }, + { 0x0c, "gpc02_tex_0d", gk104_tex_sources }, + { 0x0e, "gpc02_tex_0e", gk104_tex_sources }, + { 0x0f, "gpc02_tex_0f", gk104_tex_sources }, + { 0x10, "gpc02_tex_10", gk104_tex_sources }, + { 0x11, "gpc02_tex_11", gk104_tex_sources }, + { 0x12, "gpc02_tex_12", gk104_tex_sources }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct nvkm_specdom +gk104_pm_part[] = { + { 0x60, (const struct nvkm_specsig[]) { + { 0x00, "part00_pbfb_00", gf100_pbfb_sources }, + { 0x01, "part00_pbfb_01", gf100_pbfb_sources }, + { 0x0c, "part00_pmfb_00", gk104_pmfb_sources }, + { 0x0d, "part00_pmfb_01", gk104_pmfb_sources }, + { 0x0e, "part00_pmfb_02", gk104_pmfb_sources }, + { 0x0f, "part00_pmfb_03", gk104_pmfb_sources }, + { 0x10, "part00_pmfb_04", gk104_pmfb_sources }, + { 0x12, "part00_pmfb_05", gk104_pmfb_sources }, + { 0x15, "part00_pmfb_06", gk104_pmfb_sources }, + { 0x16, "part00_pmfb_07", gk104_pmfb_sources }, + { 0x18, "part00_pmfb_08", gk104_pmfb_sources }, + { 0x21, "part00_pmfb_09", gk104_pmfb_sources }, + { 0x25, "part00_pmfb_0a", gk104_pmfb_sources }, + { 0x26, "part00_pmfb_0b", gk104_pmfb_sources }, + { 0x27, "part00_pmfb_0c", gk104_pmfb_sources }, + { 0x47, "part00_user_0" }, + {} + }, &gf100_perfctr_func }, + { 0x60, (const struct nvkm_specsig[]) { + { 0x47, "part01_user_0" }, + {} + }, &gf100_perfctr_func }, + {} +}; + +static const struct gf100_pm_func +gk104_pm = { + .doms_gpc = gk104_pm_gpc, + .doms_hub = gk104_pm_hub, + .doms_part = gk104_pm_part, +}; + +int +gk104_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return gf100_pm_new_(&gk104_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt200.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt200.c new file mode 100644 index 0000000000..25874c5414 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt200.c @@ -0,0 +1,157 @@ +/* + * Copyright 2015 Nouveau project + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Samuel Pitoiset + */ +#include "nv40.h" + +const struct nvkm_specsrc +gt200_crop_sources[] = { + { 0x407008, (const struct nvkm_specmux[]) { + { 0xf, 0, "sel0", true }, + { 0x1f, 16, "sel1", true }, + {} + }, "pgraph_rop0_crop_pm_mux" }, + {} +}; + +const struct nvkm_specsrc +gt200_prop_sources[] = { + { 0x408750, (const struct nvkm_specmux[]) { + { 0x3f, 0, "sel", true }, + {} + }, "pgraph_tpc0_prop_pm_mux" }, + {} +}; + +const struct nvkm_specsrc +gt200_tex_sources[] = { + { 0x408508, (const struct nvkm_specmux[]) { + { 0xfffff, 0, "unk0" }, + {} + }, "pgraph_tpc0_tex_unk08" }, + {} +}; + +static const struct nvkm_specdom +gt200_pm[] = { + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0xc9, "pc01_gr_idle" }, + { 0x84, "pc01_strmout_00" }, + { 0x85, "pc01_strmout_01" }, + { 0xde, "pc01_trast_00" }, + { 0xdf, "pc01_trast_01" }, + { 0xe0, "pc01_trast_02" }, + { 0xe1, "pc01_trast_03" }, + { 0xe4, "pc01_trast_04" }, + { 0xe5, "pc01_trast_05" }, + { 0x82, "pc01_vattr_00" }, + { 0x83, "pc01_vattr_01" }, + { 0x46, "pc01_vfetch_00", g84_vfetch_sources }, + { 0x47, "pc01_vfetch_01", g84_vfetch_sources }, + { 0x48, "pc01_vfetch_02", g84_vfetch_sources }, + { 0x49, "pc01_vfetch_03", g84_vfetch_sources }, + { 0x4a, "pc01_vfetch_04", g84_vfetch_sources }, + { 0x4b, "pc01_vfetch_05", g84_vfetch_sources }, + { 0x4c, "pc01_vfetch_06", g84_vfetch_sources }, + { 0x4d, "pc01_vfetch_07", g84_vfetch_sources }, + { 0x4e, "pc01_vfetch_08", g84_vfetch_sources }, + { 0x4f, "pc01_vfetch_09", g84_vfetch_sources }, + { 0x50, "pc01_vfetch_0a", g84_vfetch_sources }, + { 0x51, "pc01_vfetch_0b", g84_vfetch_sources }, + { 0x52, "pc01_vfetch_0c", g84_vfetch_sources }, + { 0x53, "pc01_vfetch_0d", g84_vfetch_sources }, + { 0x54, "pc01_vfetch_0e", g84_vfetch_sources }, + { 0x55, "pc01_vfetch_0f", g84_vfetch_sources }, + { 0x56, "pc01_vfetch_10", g84_vfetch_sources }, + { 0x57, "pc01_vfetch_11", g84_vfetch_sources }, + { 0x58, "pc01_vfetch_12", g84_vfetch_sources }, + { 0x59, "pc01_vfetch_13", g84_vfetch_sources }, + { 0x5a, "pc01_vfetch_14", g84_vfetch_sources }, + { 0x5b, "pc01_vfetch_15", g84_vfetch_sources }, + { 0x5c, "pc01_vfetch_16", g84_vfetch_sources }, + { 0x5d, "pc01_vfetch_17", g84_vfetch_sources }, + { 0x5e, "pc01_vfetch_18", g84_vfetch_sources }, + { 0x5f, "pc01_vfetch_19", g84_vfetch_sources }, + { 0x07, "pc01_zcull_00", nv50_zcull_sources }, + { 0x08, "pc01_zcull_01", nv50_zcull_sources }, + { 0x09, "pc01_zcull_02", nv50_zcull_sources }, + { 0x0a, "pc01_zcull_03", nv50_zcull_sources }, + { 0x0b, "pc01_zcull_04", nv50_zcull_sources }, + { 0x0c, "pc01_zcull_05", nv50_zcull_sources }, + + { 0xb0, "pc01_unk00" }, + { 0xec, "pc01_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0x55, "pc02_crop_00", gt200_crop_sources }, + { 0x56, "pc02_crop_01", gt200_crop_sources }, + { 0x57, "pc02_crop_02", gt200_crop_sources }, + { 0x58, "pc02_crop_03", gt200_crop_sources }, + { 0x00, "pc02_prop_00", gt200_prop_sources }, + { 0x01, "pc02_prop_01", gt200_prop_sources }, + { 0x02, "pc02_prop_02", gt200_prop_sources }, + { 0x03, "pc02_prop_03", gt200_prop_sources }, + { 0x04, "pc02_prop_04", gt200_prop_sources }, + { 0x05, "pc02_prop_05", gt200_prop_sources }, + { 0x06, "pc02_prop_06", gt200_prop_sources }, + { 0x07, "pc02_prop_07", gt200_prop_sources }, + { 0x78, "pc02_tex_00", gt200_tex_sources }, + { 0x79, "pc02_tex_01", gt200_tex_sources }, + { 0x7a, "pc02_tex_02", gt200_tex_sources }, + { 0x7b, "pc02_tex_03", gt200_tex_sources }, + { 0x32, "pc02_tex_04", gt200_tex_sources }, + { 0x33, "pc02_tex_05", gt200_tex_sources }, + { 0x34, "pc02_tex_06", gt200_tex_sources }, + { 0x74, "pc02_zrop_00", nv50_zrop_sources }, + { 0x75, "pc02_zrop_01", nv50_zrop_sources }, + { 0x76, "pc02_zrop_02", nv50_zrop_sources }, + { 0x77, "pc02_zrop_03", nv50_zrop_sources }, + { 0xec, "pc02_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + {} +}; + +int +gt200_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return nv40_pm_new_(gt200_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt215.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt215.c new file mode 100644 index 0000000000..54c23e2b66 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/gt215.c @@ -0,0 +1,138 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" + +static const struct nvkm_specsrc +gt215_zcull_sources[] = { + { 0x402ca4, (const struct nvkm_specmux[]) { + { 0x7fff, 0, "unk0" }, + { 0xff, 24, "unk24" }, + {} + }, "pgraph_zcull_pm_unka4" }, + {} +}; + +static const struct nvkm_specdom +gt215_pm[] = { + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0xcb, "pc01_gr_idle" }, + { 0x86, "pc01_strmout_00" }, + { 0x87, "pc01_strmout_01" }, + { 0xe0, "pc01_trast_00" }, + { 0xe1, "pc01_trast_01" }, + { 0xe2, "pc01_trast_02" }, + { 0xe3, "pc01_trast_03" }, + { 0xe6, "pc01_trast_04" }, + { 0xe7, "pc01_trast_05" }, + { 0x84, "pc01_vattr_00" }, + { 0x85, "pc01_vattr_01" }, + { 0x46, "pc01_vfetch_00", g84_vfetch_sources }, + { 0x47, "pc01_vfetch_01", g84_vfetch_sources }, + { 0x48, "pc01_vfetch_02", g84_vfetch_sources }, + { 0x49, "pc01_vfetch_03", g84_vfetch_sources }, + { 0x4a, "pc01_vfetch_04", g84_vfetch_sources }, + { 0x4b, "pc01_vfetch_05", g84_vfetch_sources }, + { 0x4c, "pc01_vfetch_06", g84_vfetch_sources }, + { 0x4d, "pc01_vfetch_07", g84_vfetch_sources }, + { 0x4e, "pc01_vfetch_08", g84_vfetch_sources }, + { 0x4f, "pc01_vfetch_09", g84_vfetch_sources }, + { 0x50, "pc01_vfetch_0a", g84_vfetch_sources }, + { 0x51, "pc01_vfetch_0b", g84_vfetch_sources }, + { 0x52, "pc01_vfetch_0c", g84_vfetch_sources }, + { 0x53, "pc01_vfetch_0d", g84_vfetch_sources }, + { 0x54, "pc01_vfetch_0e", g84_vfetch_sources }, + { 0x55, "pc01_vfetch_0f", g84_vfetch_sources }, + { 0x56, "pc01_vfetch_10", g84_vfetch_sources }, + { 0x57, "pc01_vfetch_11", g84_vfetch_sources }, + { 0x58, "pc01_vfetch_12", g84_vfetch_sources }, + { 0x59, "pc01_vfetch_13", g84_vfetch_sources }, + { 0x5a, "pc01_vfetch_14", g84_vfetch_sources }, + { 0x5b, "pc01_vfetch_15", g84_vfetch_sources }, + { 0x5c, "pc01_vfetch_16", g84_vfetch_sources }, + { 0x5d, "pc01_vfetch_17", g84_vfetch_sources }, + { 0x5e, "pc01_vfetch_18", g84_vfetch_sources }, + { 0x5f, "pc01_vfetch_19", g84_vfetch_sources }, + { 0x07, "pc01_zcull_00", gt215_zcull_sources }, + { 0x08, "pc01_zcull_01", gt215_zcull_sources }, + { 0x09, "pc01_zcull_02", gt215_zcull_sources }, + { 0x0a, "pc01_zcull_03", gt215_zcull_sources }, + { 0x0b, "pc01_zcull_04", gt215_zcull_sources }, + { 0x0c, "pc01_zcull_05", gt215_zcull_sources }, + { 0xb2, "pc01_unk00" }, + { 0xec, "pc01_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0xe0, (const struct nvkm_specsig[]) { + { 0x64, "pc02_crop_00", gt200_crop_sources }, + { 0x65, "pc02_crop_01", gt200_crop_sources }, + { 0x66, "pc02_crop_02", gt200_crop_sources }, + { 0x67, "pc02_crop_03", gt200_crop_sources }, + { 0x00, "pc02_prop_00", gt200_prop_sources }, + { 0x01, "pc02_prop_01", gt200_prop_sources }, + { 0x02, "pc02_prop_02", gt200_prop_sources }, + { 0x03, "pc02_prop_03", gt200_prop_sources }, + { 0x04, "pc02_prop_04", gt200_prop_sources }, + { 0x05, "pc02_prop_05", gt200_prop_sources }, + { 0x06, "pc02_prop_06", gt200_prop_sources }, + { 0x07, "pc02_prop_07", gt200_prop_sources }, + { 0x80, "pc02_tex_00", gt200_tex_sources }, + { 0x81, "pc02_tex_01", gt200_tex_sources }, + { 0x82, "pc02_tex_02", gt200_tex_sources }, + { 0x83, "pc02_tex_03", gt200_tex_sources }, + { 0x3a, "pc02_tex_04", gt200_tex_sources }, + { 0x3b, "pc02_tex_05", gt200_tex_sources }, + { 0x3c, "pc02_tex_06", gt200_tex_sources }, + { 0x7c, "pc02_zrop_00", nv50_zrop_sources }, + { 0x7d, "pc02_zrop_01", nv50_zrop_sources }, + { 0x7e, "pc02_zrop_02", nv50_zrop_sources }, + { 0x7f, "pc02_zrop_03", nv50_zrop_sources }, + { 0xcc, "pc02_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + {} +}; + +int +gt215_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return nv40_pm_new_(gt215_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.c new file mode 100644 index 0000000000..eba5b3b793 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.c @@ -0,0 +1,123 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" + +static void +nv40_perfctr_init(struct nvkm_pm *pm, struct nvkm_perfdom *dom, + struct nvkm_perfctr *ctr) +{ + struct nvkm_device *device = pm->engine.subdev.device; + u32 log = ctr->logic_op; + u32 src = 0x00000000; + int i; + + for (i = 0; i < 4; i++) + src |= ctr->signal[i] << (i * 8); + + nvkm_wr32(device, 0x00a7c0 + dom->addr, 0x00000001 | (dom->mode << 4)); + nvkm_wr32(device, 0x00a400 + dom->addr + (ctr->slot * 0x40), src); + nvkm_wr32(device, 0x00a420 + dom->addr + (ctr->slot * 0x40), log); +} + +static void +nv40_perfctr_read(struct nvkm_pm *pm, struct nvkm_perfdom *dom, + struct nvkm_perfctr *ctr) +{ + struct nvkm_device *device = pm->engine.subdev.device; + + switch (ctr->slot) { + case 0: ctr->ctr = nvkm_rd32(device, 0x00a700 + dom->addr); break; + case 1: ctr->ctr = nvkm_rd32(device, 0x00a6c0 + dom->addr); break; + case 2: ctr->ctr = nvkm_rd32(device, 0x00a680 + dom->addr); break; + case 3: ctr->ctr = nvkm_rd32(device, 0x00a740 + dom->addr); break; + } + dom->clk = nvkm_rd32(device, 0x00a600 + dom->addr); +} + +static void +nv40_perfctr_next(struct nvkm_pm *pm, struct nvkm_perfdom *dom) +{ + struct nvkm_device *device = pm->engine.subdev.device; + struct nv40_pm *nv40pm = container_of(pm, struct nv40_pm, base); + + if (nv40pm->sequence != pm->sequence) { + nvkm_wr32(device, 0x400084, 0x00000020); + nv40pm->sequence = pm->sequence; + } +} + +const struct nvkm_funcdom +nv40_perfctr_func = { + .init = nv40_perfctr_init, + .read = nv40_perfctr_read, + .next = nv40_perfctr_next, +}; + +static const struct nvkm_pm_func +nv40_pm_ = { +}; + +int +nv40_pm_new_(const struct nvkm_specdom *doms, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + struct nv40_pm *pm; + int ret; + + if (!(pm = kzalloc(sizeof(*pm), GFP_KERNEL))) + return -ENOMEM; + *ppm = &pm->base; + + ret = nvkm_pm_ctor(&nv40_pm_, device, type, inst, &pm->base); + if (ret) + return ret; + + return nvkm_perfdom_new(&pm->base, "pc", 0, 0, 0, 4, doms); +} + +static const struct nvkm_specdom +nv40_pm[] = { + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + {} +}; + +int +nv40_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return nv40_pm_new_(nv40_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.h b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.h new file mode 100644 index 0000000000..afb7984372 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv40.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_PM_NV40_H__ +#define __NVKM_PM_NV40_H__ +#define nv40_pm(p) container_of((p), struct nv40_pm, base) +#include "priv.h" + +struct nv40_pm { + struct nvkm_pm base; + u32 sequence; +}; + +int nv40_pm_new_(const struct nvkm_specdom *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_pm **); +extern const struct nvkm_funcdom nv40_perfctr_func; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv50.c new file mode 100644 index 0000000000..bbd3404901 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/nv50.c @@ -0,0 +1,175 @@ +/* + * Copyright 2013 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" + +const struct nvkm_specsrc +nv50_zcull_sources[] = { + { 0x402ca4, (const struct nvkm_specmux[]) { + { 0x7fff, 0, "unk0" }, + {} + }, "pgraph_zcull_pm_unka4" }, + {} +}; + +const struct nvkm_specsrc +nv50_zrop_sources[] = { + { 0x40708c, (const struct nvkm_specmux[]) { + { 0xf, 0, "sel0", true }, + { 0xf, 16, "sel1", true }, + {} + }, "pgraph_rop0_zrop_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +nv50_prop_sources[] = { + { 0x40be50, (const struct nvkm_specmux[]) { + { 0x1f, 0, "sel", true }, + {} + }, "pgraph_tpc3_prop_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +nv50_crop_sources[] = { + { 0x407008, (const struct nvkm_specmux[]) { + { 0x7, 0, "sel0", true }, + { 0x7, 16, "sel1", true }, + {} + }, "pgraph_rop0_crop_pm_mux" }, + {} +}; + +static const struct nvkm_specsrc +nv50_tex_sources[] = { + { 0x40b808, (const struct nvkm_specmux[]) { + { 0x3fff, 0, "unk0" }, + {} + }, "pgraph_tpc3_tex_unk08" }, + {} +}; + +static const struct nvkm_specsrc +nv50_vfetch_sources[] = { + { 0x400c0c, (const struct nvkm_specmux[]) { + { 0x1, 0, "unk0" }, + {} + }, "pgraph_vfetch_unk0c" }, + {} +}; + +static const struct nvkm_specdom +nv50_pm[] = { + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0xc8, "pc01_gr_idle" }, + { 0x7f, "pc01_strmout_00" }, + { 0x80, "pc01_strmout_01" }, + { 0xdc, "pc01_trast_00" }, + { 0xdd, "pc01_trast_01" }, + { 0xde, "pc01_trast_02" }, + { 0xdf, "pc01_trast_03" }, + { 0xe2, "pc01_trast_04" }, + { 0xe3, "pc01_trast_05" }, + { 0x7c, "pc01_vattr_00" }, + { 0x7d, "pc01_vattr_01" }, + { 0x26, "pc01_vfetch_00", nv50_vfetch_sources }, + { 0x27, "pc01_vfetch_01", nv50_vfetch_sources }, + { 0x28, "pc01_vfetch_02", nv50_vfetch_sources }, + { 0x29, "pc01_vfetch_03", nv50_vfetch_sources }, + { 0x2a, "pc01_vfetch_04", nv50_vfetch_sources }, + { 0x2b, "pc01_vfetch_05", nv50_vfetch_sources }, + { 0x2c, "pc01_vfetch_06", nv50_vfetch_sources }, + { 0x2d, "pc01_vfetch_07", nv50_vfetch_sources }, + { 0x2e, "pc01_vfetch_08", nv50_vfetch_sources }, + { 0x2f, "pc01_vfetch_09", nv50_vfetch_sources }, + { 0x30, "pc01_vfetch_0a", nv50_vfetch_sources }, + { 0x31, "pc01_vfetch_0b", nv50_vfetch_sources }, + { 0x32, "pc01_vfetch_0c", nv50_vfetch_sources }, + { 0x33, "pc01_vfetch_0d", nv50_vfetch_sources }, + { 0x34, "pc01_vfetch_0e", nv50_vfetch_sources }, + { 0x35, "pc01_vfetch_0f", nv50_vfetch_sources }, + { 0x36, "pc01_vfetch_10", nv50_vfetch_sources }, + { 0x37, "pc01_vfetch_11", nv50_vfetch_sources }, + { 0x38, "pc01_vfetch_12", nv50_vfetch_sources }, + { 0x39, "pc01_vfetch_13", nv50_vfetch_sources }, + { 0x3a, "pc01_vfetch_14", nv50_vfetch_sources }, + { 0x3b, "pc01_vfetch_15", nv50_vfetch_sources }, + { 0x3c, "pc01_vfetch_16", nv50_vfetch_sources }, + { 0x3d, "pc01_vfetch_17", nv50_vfetch_sources }, + { 0x3e, "pc01_vfetch_18", nv50_vfetch_sources }, + { 0x3f, "pc01_vfetch_19", nv50_vfetch_sources }, + { 0x20, "pc01_zcull_00", nv50_zcull_sources }, + { 0x21, "pc01_zcull_01", nv50_zcull_sources }, + { 0x22, "pc01_zcull_02", nv50_zcull_sources }, + { 0x23, "pc01_zcull_03", nv50_zcull_sources }, + { 0x24, "pc01_zcull_04", nv50_zcull_sources }, + { 0x25, "pc01_zcull_05", nv50_zcull_sources }, + { 0xae, "pc01_unk00" }, + { 0xee, "pc01_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0xf0, (const struct nvkm_specsig[]) { + { 0x52, "pc02_crop_00", nv50_crop_sources }, + { 0x53, "pc02_crop_01", nv50_crop_sources }, + { 0x54, "pc02_crop_02", nv50_crop_sources }, + { 0x55, "pc02_crop_03", nv50_crop_sources }, + { 0x00, "pc02_prop_00", nv50_prop_sources }, + { 0x01, "pc02_prop_01", nv50_prop_sources }, + { 0x02, "pc02_prop_02", nv50_prop_sources }, + { 0x03, "pc02_prop_03", nv50_prop_sources }, + { 0x04, "pc02_prop_04", nv50_prop_sources }, + { 0x05, "pc02_prop_05", nv50_prop_sources }, + { 0x06, "pc02_prop_06", nv50_prop_sources }, + { 0x07, "pc02_prop_07", nv50_prop_sources }, + { 0x70, "pc02_tex_00", nv50_tex_sources }, + { 0x71, "pc02_tex_01", nv50_tex_sources }, + { 0x72, "pc02_tex_02", nv50_tex_sources }, + { 0x73, "pc02_tex_03", nv50_tex_sources }, + { 0x40, "pc02_tex_04", nv50_tex_sources }, + { 0x41, "pc02_tex_05", nv50_tex_sources }, + { 0x42, "pc02_tex_06", nv50_tex_sources }, + { 0x6c, "pc02_zrop_00", nv50_zrop_sources }, + { 0x6d, "pc02_zrop_01", nv50_zrop_sources }, + { 0x6e, "pc02_zrop_02", nv50_zrop_sources }, + { 0x6f, "pc02_zrop_03", nv50_zrop_sources }, + { 0xee, "pc02_trailer" }, + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + { 0x20, (const struct nvkm_specsig[]) { + {} + }, &nv40_perfctr_func }, + {} +}; + +int +nv50_pm_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_pm **ppm) +{ + return nv40_pm_new_(nv50_pm, device, type, inst, ppm); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/pm/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/pm/priv.h new file mode 100644 index 0000000000..6ae25d3e7f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/pm/priv.h @@ -0,0 +1,105 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_PM_PRIV_H__ +#define __NVKM_PM_PRIV_H__ +#define nvkm_pm(p) container_of((p), struct nvkm_pm, engine) +#include + +int nvkm_pm_ctor(const struct nvkm_pm_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_pm *); + +struct nvkm_pm_func { + void (*fini)(struct nvkm_pm *); +}; + +struct nvkm_perfctr { + struct list_head head; + u8 domain; + u8 signal[4]; + u64 source[4][8]; + int slot; + u32 logic_op; + u32 ctr; +}; + +struct nvkm_specmux { + u32 mask; + u8 shift; + const char *name; + bool enable; +}; + +struct nvkm_specsrc { + u32 addr; + const struct nvkm_specmux *mux; + const char *name; +}; + +struct nvkm_perfsrc { + struct list_head head; + char *name; + u32 addr; + u32 mask; + u8 shift; + bool enable; +}; + +extern const struct nvkm_specsrc nv50_zcull_sources[]; +extern const struct nvkm_specsrc nv50_zrop_sources[]; +extern const struct nvkm_specsrc g84_vfetch_sources[]; +extern const struct nvkm_specsrc gt200_crop_sources[]; +extern const struct nvkm_specsrc gt200_prop_sources[]; +extern const struct nvkm_specsrc gt200_tex_sources[]; + +struct nvkm_specsig { + u8 signal; + const char *name; + const struct nvkm_specsrc *source; +}; + +struct nvkm_perfsig { + const char *name; + u8 source[8]; +}; + +struct nvkm_specdom { + u16 signal_nr; + const struct nvkm_specsig *signal; + const struct nvkm_funcdom *func; +}; + +#define nvkm_perfdom(p) container_of((p), struct nvkm_perfdom, object) +#include + +struct nvkm_perfdom { + struct nvkm_object object; + struct nvkm_perfmon *perfmon; + struct list_head head; + struct list_head list; + const struct nvkm_funcdom *func; + struct nvkm_perfctr *ctr[4]; + char name[32]; + u32 addr; + u8 mode; + u32 clk; + u16 signal_nr; + struct nvkm_perfsig signal[]; +}; + +struct nvkm_funcdom { + void (*init)(struct nvkm_pm *, struct nvkm_perfdom *, + struct nvkm_perfctr *); + void (*read)(struct nvkm_pm *, struct nvkm_perfdom *, + struct nvkm_perfctr *); + void (*next)(struct nvkm_pm *, struct nvkm_perfdom *); +}; + +int nvkm_perfdom_new(struct nvkm_pm *, const char *, u32, u32, u32, u32, + const struct nvkm_specdom *); + +#define nvkm_perfmon(p) container_of((p), struct nvkm_perfmon, object) + +struct nvkm_perfmon { + struct nvkm_object object; + struct nvkm_pm *pm; +}; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/sec/Kbuild new file mode 100644 index 0000000000..b6e02ceb1c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/sec/g98.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s new file mode 100644 index 0000000000..66b147bd58 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s @@ -0,0 +1,698 @@ +/* + * fuc microcode for g98 sec engine + * Copyright (C) 2010 Marcin Kościelnicki + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +.section #g98_sec_data + +ctx_dma: +ctx_dma_query: .b32 0 +ctx_dma_src: .b32 0 +ctx_dma_dst: .b32 0 +.equ #dma_count 3 +ctx_query_address_high: .b32 0 +ctx_query_address_low: .b32 0 +ctx_query_counter: .b32 0 +ctx_cond_address_high: .b32 0 +ctx_cond_address_low: .b32 0 +ctx_cond_off: .b32 0 +ctx_src_address_high: .b32 0 +ctx_src_address_low: .b32 0 +ctx_dst_address_high: .b32 0 +ctx_dst_address_low: .b32 0 +ctx_mode: .b32 0 +.align 16 +ctx_key: .skip 16 +ctx_iv: .skip 16 + +.align 0x80 +swap: +.skip 32 + +.align 8 +common_cmd_dtable: +.b32 #ctx_query_address_high + 0x20000 ~0xff +.b32 #ctx_query_address_low + 0x20000 ~0xfffffff0 +.b32 #ctx_query_counter + 0x20000 ~0xffffffff +.b32 #cmd_query_get + 0x00000 ~1 +.b32 #ctx_cond_address_high + 0x20000 ~0xff +.b32 #ctx_cond_address_low + 0x20000 ~0xfffffff0 +.b32 #cmd_cond_mode + 0x00000 ~7 +.b32 #cmd_wrcache_flush + 0x00000 ~0 +.equ #common_cmd_max 0x88 + + +.align 8 +engine_cmd_dtable: +.b32 #ctx_key + 0x0 + 0x20000 ~0xffffffff +.b32 #ctx_key + 0x4 + 0x20000 ~0xffffffff +.b32 #ctx_key + 0x8 + 0x20000 ~0xffffffff +.b32 #ctx_key + 0xc + 0x20000 ~0xffffffff +.b32 #ctx_iv + 0x0 + 0x20000 ~0xffffffff +.b32 #ctx_iv + 0x4 + 0x20000 ~0xffffffff +.b32 #ctx_iv + 0x8 + 0x20000 ~0xffffffff +.b32 #ctx_iv + 0xc + 0x20000 ~0xffffffff +.b32 #ctx_src_address_high + 0x20000 ~0xff +.b32 #ctx_src_address_low + 0x20000 ~0xfffffff0 +.b32 #ctx_dst_address_high + 0x20000 ~0xff +.b32 #ctx_dst_address_low + 0x20000 ~0xfffffff0 +.b32 #sec_cmd_mode + 0x00000 ~0xf +.b32 #sec_cmd_length + 0x10000 ~0x0ffffff0 +.equ #engine_cmd_max 0xce + +.align 4 +sec_dtable: +.b16 #sec_copy_prep #sec_do_inout +.b16 #sec_store_prep #sec_do_out +.b16 #sec_ecb_e_prep #sec_do_inout +.b16 #sec_ecb_d_prep #sec_do_inout +.b16 #sec_cbc_e_prep #sec_do_inout +.b16 #sec_cbc_d_prep #sec_do_inout +.b16 #sec_pcbc_e_prep #sec_do_inout +.b16 #sec_pcbc_d_prep #sec_do_inout +.b16 #sec_cfb_e_prep #sec_do_inout +.b16 #sec_cfb_d_prep #sec_do_inout +.b16 #sec_ofb_prep #sec_do_inout +.b16 #sec_ctr_prep #sec_do_inout +.b16 #sec_cbc_mac_prep #sec_do_in +.b16 #sec_cmac_finish_complete_prep #sec_do_in +.b16 #sec_cmac_finish_partial_prep #sec_do_in + +.align 0x100 + +.section #g98_sec_code + + // $r0 is always set to 0 in our code - this allows some space savings. + clear b32 $r0 + + // set up the interrupt handler + mov $r1 #ih + mov $iv0 $r1 + + // init stack pointer + mov $sp $r0 + + // set interrupt dispatch - route timer, fifo, ctxswitch to i0, others to host + movw $r1 0xfff0 + sethi $r1 0 + mov $r2 0x400 + iowr I[$r2 + 0x300] $r1 + + // enable the interrupts + or $r1 0xc + iowr I[$r2] $r1 + + // enable fifo access and context switching + mov $r1 3 + mov $r2 0x1200 + iowr I[$r2] $r1 + + // enable i0 delivery + bset $flags ie0 + + // sleep forver, waking only for interrupts. + bset $flags $p0 + spin: + sleep $p0 + bra #spin + +// i0 handler +ih: + // see which interrupts we got + iord $r1 I[$r0 + 0x200] + + and $r2 $r1 0x8 + cmpu b32 $r2 0 + bra e #noctx + + // context switch... prepare the regs for xfer + mov $r2 0x7700 + mov $xtargets $r2 + mov $xdbase $r0 + // 128-byte context. + mov $r2 0 + sethi $r2 0x50000 + + // read current channel + mov $r3 0x1400 + iord $r4 I[$r3] + // if bit 30 set, it's active, so we have to unload it first. + shl b32 $r5 $r4 1 + cmps b32 $r5 0 + bra nc #ctxload + + // unload the current channel - save the context + xdst $r0 $r2 + xdwait + // and clear bit 30, then write back + bclr $r4 0x1e + iowr I[$r3] $r4 + // tell PFIFO we unloaded + mov $r4 1 + iowr I[$r3 + 0x200] $r4 + + bra #noctx + + ctxload: + // no channel loaded - perhaps we're requested to load one + iord $r4 I[$r3 + 0x100] + shl b32 $r15 $r4 1 + cmps b32 $r15 0 + // if bit 30 of next channel not set, probably PFIFO is just + // killing a context. do a faux load, without the active bit. + bra nc #dummyload + + // ok, do a real context load. + xdld $r0 $r2 + xdwait + mov $r5 #ctx_dma + mov $r6 #dma_count - 1 + ctxload_dma_loop: + ld b32 $r7 D[$r5 + $r6 * 4] + add b32 $r8 $r6 0x180 + shl b32 $r8 8 + iowr I[$r8] $r7 + sub b32 $r6 1 + bra nc #ctxload_dma_loop + + dummyload: + // tell PFIFO we're done + mov $r5 2 + iowr I[$r3 + 0x200] $r5 + + noctx: + and $r2 $r1 0x4 + cmpu b32 $r2 0 + bra e #nocmd + + // incoming fifo command. + mov $r3 0x1900 + iord $r2 I[$r3 + 0x100] + iord $r3 I[$r3] + // extract the method + and $r4 $r2 0x7ff + // shift the addr to proper position if we need to interrupt later + shl b32 $r2 0x10 + + // mthd 0 and 0x100 [NAME, NOP]: ignore + and $r5 $r4 0x7bf + cmpu b32 $r5 0 + bra e #cmddone + + mov $r5 #engine_cmd_dtable - 0xc0 * 8 + mov $r6 #engine_cmd_max + cmpu b32 $r4 0xc0 + bra nc #dtable_cmd + mov $r5 #common_cmd_dtable - 0x80 * 8 + mov $r6 #common_cmd_max + cmpu b32 $r4 0x80 + bra nc #dtable_cmd + cmpu b32 $r4 0x60 + bra nc #dma_cmd + cmpu b32 $r4 0x50 + bra ne #illegal_mthd + + // mthd 0x140: PM_TRIGGER + mov $r2 0x2200 + clear b32 $r3 + sethi $r3 0x20000 + iowr I[$r2] $r3 + bra #cmddone + + dma_cmd: + // mthd 0x180...: DMA_* + cmpu b32 $r4 0x60+#dma_count + bra nc #illegal_mthd + shl b32 $r5 $r4 2 + add b32 $r5 ((#ctx_dma - 0x60 * 4) & 0xffff) + bset $r3 0x1e + st b32 D[$r5] $r3 + add b32 $r4 0x180 - 0x60 + shl b32 $r4 8 + iowr I[$r4] $r3 + bra #cmddone + + dtable_cmd: + cmpu b32 $r4 $r6 + bra nc #illegal_mthd + shl b32 $r4 3 + add b32 $r4 $r5 + ld b32 $r5 D[$r4 + 4] + and $r5 $r3 + cmpu b32 $r5 0 + bra ne #invalid_bitfield + ld b16 $r5 D[$r4] + ld b16 $r6 D[$r4 + 2] + cmpu b32 $r6 2 + bra e #cmd_setctx + ld b32 $r7 D[$r0 + #ctx_cond_off] + and $r6 $r7 + cmpu b32 $r6 1 + bra e #cmddone + call $r5 + bra $p1 #dispatch_error + bra #cmddone + + cmd_setctx: + st b32 D[$r5] $r3 + bra #cmddone + + + invalid_bitfield: + or $r2 1 + dispatch_error: + illegal_mthd: + mov $r4 0x1000 + iowr I[$r4] $r2 + iowr I[$r4 + 0x100] $r3 + mov $r4 0x40 + iowr I[$r0] $r4 + + im_loop: + iord $r4 I[$r0 + 0x200] + and $r4 0x40 + cmpu b32 $r4 0 + bra ne #im_loop + + cmddone: + // remove the command from FIFO + mov $r3 0x1d00 + mov $r4 1 + iowr I[$r3] $r4 + + nocmd: + // ack the processed interrupts + and $r1 $r1 0xc + iowr I[$r0 + 0x100] $r1 +iret + +cmd_query_get: + // if bit 0 of param set, trigger interrupt afterwards. + setp $p1 $r3 + or $r2 3 + + // read PTIMER, beware of races... + mov $r4 0xb00 + ptimer_retry: + iord $r6 I[$r4 + 0x100] + iord $r5 I[$r4] + iord $r7 I[$r4 + 0x100] + cmpu b32 $r6 $r7 + bra ne #ptimer_retry + + // prepare the query structure + ld b32 $r4 D[$r0 + #ctx_query_counter] + st b32 D[$r0 + #swap + 0x0] $r4 + st b32 D[$r0 + #swap + 0x4] $r0 + st b32 D[$r0 + #swap + 0x8] $r5 + st b32 D[$r0 + #swap + 0xc] $r6 + + // will use target 0, DMA_QUERY. + mov $xtargets $r0 + + ld b32 $r4 D[$r0 + #ctx_query_address_high] + shl b32 $r4 0x18 + mov $xdbase $r4 + + ld b32 $r4 D[$r0 + #ctx_query_address_low] + mov $r5 #swap + sethi $r5 0x20000 + xdst $r4 $r5 + xdwait + + ret + +cmd_cond_mode: + // if >= 5, INVALID_ENUM + bset $flags $p1 + or $r2 2 + cmpu b32 $r3 5 + bra nc #return + + // otherwise, no error. + bclr $flags $p1 + + // if < 2, no QUERY object is involved + cmpu b32 $r3 2 + bra nc #cmd_cond_mode_queryful + + xor $r3 1 + st b32 D[$r0 + #ctx_cond_off] $r3 + return: + ret + + cmd_cond_mode_queryful: + // ok, will need to pull a QUERY object, prepare offsets + ld b32 $r4 D[$r0 + #ctx_cond_address_high] + ld b32 $r5 D[$r0 + #ctx_cond_address_low] + and $r6 $r5 0xff + shr b32 $r5 8 + shl b32 $r4 0x18 + or $r4 $r5 + mov $xdbase $r4 + mov $xtargets $r0 + + // pull the first one + mov $r5 #swap + sethi $r5 0x20000 + xdld $r6 $r5 + + // if == 2, only a single QUERY is involved... + cmpu b32 $r3 2 + bra ne #cmd_cond_mode_double + + xdwait + ld b32 $r4 D[$r0 + #swap + 4] + cmpu b32 $r4 0 + xbit $r4 $flags z + st b32 D[$r0 + #ctx_cond_off] $r4 + ret + + // ok, we'll need to pull second one too + cmd_cond_mode_double: + add b32 $r6 0x10 + add b32 $r5 0x10 + xdld $r6 $r5 + xdwait + + // compare COUNTERs + ld b32 $r5 D[$r0 + #swap + 0x00] + ld b32 $r6 D[$r0 + #swap + 0x10] + cmpu b32 $r5 $r6 + xbit $r4 $flags z + + // compare RESen + ld b32 $r5 D[$r0 + #swap + 0x04] + ld b32 $r6 D[$r0 + #swap + 0x14] + cmpu b32 $r5 $r6 + xbit $r5 $flags z + and $r4 $r5 + + // and negate or not, depending on mode + cmpu b32 $r3 3 + xbit $r5 $flags z + xor $r4 $r5 + st b32 D[$r0 + #ctx_cond_off] $r4 + ret + +cmd_wrcache_flush: + bclr $flags $p1 + mov $r2 0x2200 + clear b32 $r3 + sethi $r3 0x10000 + iowr I[$r2] $r3 + ret + +sec_cmd_mode: + // if >= 0xf, INVALID_ENUM + bset $flags $p1 + or $r2 2 + cmpu b32 $r3 0xf + bra nc #sec_cmd_mode_return + + bclr $flags $p1 + st b32 D[$r0 + #ctx_mode] $r3 + + sec_cmd_mode_return: + ret + +sec_cmd_length: + // nop if length == 0 + cmpu b32 $r3 0 + bra e #sec_cmd_mode_return + + // init key, IV + cxset 3 + mov $r4 #ctx_key + sethi $r4 0x70000 + xdst $r0 $r4 + mov $r4 #ctx_iv + sethi $r4 0x60000 + xdst $r0 $r4 + xdwait + ckeyreg $c7 + + // prepare the targets + mov $r4 0x2100 + mov $xtargets $r4 + + // prepare src address + ld b32 $r4 D[$r0 + #ctx_src_address_high] + ld b32 $r5 D[$r0 + #ctx_src_address_low] + shr b32 $r8 $r5 8 + shl b32 $r4 0x18 + or $r4 $r8 + and $r5 $r5 0xff + + // prepare dst address + ld b32 $r6 D[$r0 + #ctx_dst_address_high] + ld b32 $r7 D[$r0 + #ctx_dst_address_low] + shr b32 $r8 $r7 8 + shl b32 $r6 0x18 + or $r6 $r8 + and $r7 $r7 0xff + + // find the proper prep & do functions + ld b32 $r8 D[$r0 + #ctx_mode] + shl b32 $r8 2 + + // run prep + ld b16 $r9 D[$r8 + #sec_dtable] + call $r9 + + // do it + ld b16 $r9 D[$r8 + #sec_dtable + 2] + call $r9 + cxset 1 + xdwait + cxset 0x61 + xdwait + xdwait + + // update src address + shr b32 $r8 $r4 0x18 + shl b32 $r9 $r4 8 + add b32 $r9 $r5 + adc b32 $r8 0 + st b32 D[$r0 + #ctx_src_address_high] $r8 + st b32 D[$r0 + #ctx_src_address_low] $r9 + + // update dst address + shr b32 $r8 $r6 0x18 + shl b32 $r9 $r6 8 + add b32 $r9 $r7 + adc b32 $r8 0 + st b32 D[$r0 + #ctx_dst_address_high] $r8 + st b32 D[$r0 + #ctx_dst_address_low] $r9 + + // pull updated IV + cxset 2 + mov $r4 #ctx_iv + sethi $r4 0x60000 + xdld $r0 $r4 + xdwait + + ret + + +sec_copy_prep: + cs0begin 2 + cxsin $c0 + cxsout $c0 + ret + +sec_store_prep: + cs0begin 1 + cxsout $c6 + ret + +sec_ecb_e_prep: + cs0begin 3 + cxsin $c0 + cenc $c0 $c0 + cxsout $c0 + ret + +sec_ecb_d_prep: + ckexp $c7 $c7 + cs0begin 3 + cxsin $c0 + cdec $c0 $c0 + cxsout $c0 + ret + +sec_cbc_e_prep: + cs0begin 4 + cxsin $c0 + cxor $c6 $c0 + cenc $c6 $c6 + cxsout $c6 + ret + +sec_cbc_d_prep: + ckexp $c7 $c7 + cs0begin 5 + cmov $c2 $c6 + cxsin $c6 + cdec $c0 $c6 + cxor $c0 $c2 + cxsout $c0 + ret + +sec_pcbc_e_prep: + cs0begin 5 + cxsin $c0 + cxor $c6 $c0 + cenc $c6 $c6 + cxsout $c6 + cxor $c6 $c0 + ret + +sec_pcbc_d_prep: + ckexp $c7 $c7 + cs0begin 5 + cxsin $c0 + cdec $c1 $c0 + cxor $c6 $c1 + cxsout $c6 + cxor $c6 $c0 + ret + +sec_cfb_e_prep: + cs0begin 4 + cenc $c6 $c6 + cxsin $c0 + cxor $c6 $c0 + cxsout $c6 + ret + +sec_cfb_d_prep: + cs0begin 4 + cenc $c0 $c6 + cxsin $c6 + cxor $c0 $c6 + cxsout $c0 + ret + +sec_ofb_prep: + cs0begin 4 + cenc $c6 $c6 + cxsin $c0 + cxor $c0 $c6 + cxsout $c0 + ret + +sec_ctr_prep: + cs0begin 5 + cenc $c1 $c6 + cadd $c6 1 + cxsin $c0 + cxor $c0 $c1 + cxsout $c0 + ret + +sec_cbc_mac_prep: + cs0begin 3 + cxsin $c0 + cxor $c6 $c0 + cenc $c6 $c6 + ret + +sec_cmac_finish_complete_prep: + cs0begin 7 + cxsin $c0 + cxor $c6 $c0 + cxor $c0 $c0 + cenc $c0 $c0 + cprecmac $c0 $c0 + cxor $c6 $c0 + cenc $c6 $c6 + ret + +sec_cmac_finish_partial_prep: + cs0begin 8 + cxsin $c0 + cxor $c6 $c0 + cxor $c0 $c0 + cenc $c0 $c0 + cprecmac $c0 $c0 + cprecmac $c0 $c0 + cxor $c6 $c0 + cenc $c6 $c6 + ret + +// TODO +sec_do_in: + add b32 $r3 $r5 + mov $xdbase $r4 + mov $r9 #swap + sethi $r9 0x20000 + sec_do_in_loop: + xdld $r5 $r9 + xdwait + cxset 0x22 + xdst $r0 $r9 + cs0exec 1 + xdwait + add b32 $r5 0x10 + cmpu b32 $r5 $r3 + bra ne #sec_do_in_loop + cxset 1 + xdwait + ret + +sec_do_out: + add b32 $r3 $r7 + mov $xdbase $r6 + mov $r9 #swap + sethi $r9 0x20000 + sec_do_out_loop: + cs0exec 1 + cxset 0x61 + xdld $r7 $r9 + xdst $r7 $r9 + cxset 1 + xdwait + add b32 $r7 0x10 + cmpu b32 $r7 $r3 + bra ne #sec_do_out_loop + ret + +sec_do_inout: + add b32 $r3 $r5 + mov $r9 #swap + sethi $r9 0x20000 + sec_do_inout_loop: + mov $xdbase $r4 + xdld $r5 $r9 + xdwait + cxset 0x21 + xdst $r0 $r9 + cs0exec 1 + cxset 0x61 + mov $xdbase $r6 + xdld $r7 $r9 + xdst $r7 $r9 + cxset 1 + xdwait + add b32 $r5 0x10 + add b32 $r7 0x10 + cmpu b32 $r5 $r3 + bra ne #sec_do_inout_loop + ret + +.align 0x100 diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h new file mode 100644 index 0000000000..fe90f2e058 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec/fuc/g98.fuc0s.h @@ -0,0 +1,585 @@ +/* SPDX-License-Identifier: MIT */ +static uint32_t g98_sec_data[] = { +/* 0x0000: ctx_dma */ +/* 0x0000: ctx_dma_query */ + 0x00000000, +/* 0x0004: ctx_dma_src */ + 0x00000000, +/* 0x0008: ctx_dma_dst */ + 0x00000000, +/* 0x000c: ctx_query_address_high */ + 0x00000000, +/* 0x0010: ctx_query_address_low */ + 0x00000000, +/* 0x0014: ctx_query_counter */ + 0x00000000, +/* 0x0018: ctx_cond_address_high */ + 0x00000000, +/* 0x001c: ctx_cond_address_low */ + 0x00000000, +/* 0x0020: ctx_cond_off */ + 0x00000000, +/* 0x0024: ctx_src_address_high */ + 0x00000000, +/* 0x0028: ctx_src_address_low */ + 0x00000000, +/* 0x002c: ctx_dst_address_high */ + 0x00000000, +/* 0x0030: ctx_dst_address_low */ + 0x00000000, +/* 0x0034: ctx_mode */ + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0040: ctx_key */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0050: ctx_iv */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x0080: swap */ + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +/* 0x00a0: common_cmd_dtable */ + 0x0002000c, + 0xffffff00, + 0x00020010, + 0x0000000f, + 0x00020014, + 0x00000000, + 0x00000192, + 0xfffffffe, + 0x00020018, + 0xffffff00, + 0x0002001c, + 0x0000000f, + 0x000001d7, + 0xfffffff8, + 0x00000260, + 0xffffffff, +/* 0x00e0: engine_cmd_dtable */ + 0x00020040, + 0x00000000, + 0x00020044, + 0x00000000, + 0x00020048, + 0x00000000, + 0x0002004c, + 0x00000000, + 0x00020050, + 0x00000000, + 0x00020054, + 0x00000000, + 0x00020058, + 0x00000000, + 0x0002005c, + 0x00000000, + 0x00020024, + 0xffffff00, + 0x00020028, + 0x0000000f, + 0x0002002c, + 0xffffff00, + 0x00020030, + 0x0000000f, + 0x00000271, + 0xfffffff0, + 0x00010285, + 0xf000000f, +/* 0x0150: sec_dtable */ + 0x04db0321, + 0x04b1032f, + 0x04db0339, + 0x04db034b, + 0x04db0361, + 0x04db0377, + 0x04db0395, + 0x04db03af, + 0x04db03cd, + 0x04db03e3, + 0x04db03f9, + 0x04db040f, + 0x04830429, + 0x0483043b, + 0x0483045d, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; + +static uint32_t g98_sec_code[] = { + 0x17f004bd, + 0x0010fe35, + 0xf10004fe, + 0xf0fff017, + 0x27f10013, + 0x21d00400, + 0x0c15f0c0, + 0xf00021d0, + 0x27f10317, + 0x21d01200, + 0x1031f400, +/* 0x002f: spin */ + 0xf40031f4, + 0x0ef40028, +/* 0x0035: ih */ + 0x8001cffd, + 0xb00812c4, + 0x0bf40024, + 0x0027f167, + 0x002bfe77, + 0xf00007fe, + 0x23f00027, + 0x0037f105, + 0x0034cf14, + 0xb0014594, + 0x18f40055, + 0x0602fa17, + 0x4af003f8, + 0x0034d01e, + 0xd00147f0, + 0x0ef48034, +/* 0x0075: ctxload */ + 0x4034cf33, + 0xb0014f94, + 0x18f400f5, + 0x0502fa21, + 0x57f003f8, + 0x0267f000, +/* 0x008c: ctxload_dma_loop */ + 0xa07856bc, + 0xb6018068, + 0x87d00884, + 0x0162b600, +/* 0x009f: dummyload */ + 0xf0f018f4, + 0x35d00257, +/* 0x00a5: noctx */ + 0x0412c480, + 0xf50024b0, + 0xf100df0b, + 0xcf190037, + 0x33cf4032, + 0xff24e400, + 0x1024b607, + 0x07bf45e4, + 0xf50054b0, + 0xf100b90b, + 0xf1fae057, + 0xb000ce67, + 0x18f4c044, + 0xa057f14d, + 0x8867f1fc, + 0x8044b000, + 0xb03f18f4, + 0x18f46044, + 0x5044b019, + 0xf1741bf4, + 0xbd220027, + 0x0233f034, + 0xf50023d0, +/* 0x0103: dma_cmd */ + 0xb000810e, + 0x18f46344, + 0x0245945e, + 0xfe8050b7, + 0x801e39f0, + 0x40b70053, + 0x44b60120, + 0x0043d008, +/* 0x0123: dtable_cmd */ + 0xb8600ef4, + 0x18f40446, + 0x0344b63e, + 0x980045bb, + 0x53fd0145, + 0x0054b004, + 0x58291bf4, + 0x46580045, + 0x0264b001, + 0x98170bf4, + 0x67fd0807, + 0x0164b004, + 0xf9300bf4, + 0x0f01f455, +/* 0x015b: cmd_setctx */ + 0x80280ef4, + 0x0ef40053, +/* 0x0161: invalid_bitfield */ + 0x0125f022, +/* 0x0164: dispatch_error */ +/* 0x0164: illegal_mthd */ + 0x100047f1, + 0xd00042d0, + 0x47f04043, + 0x0004d040, +/* 0x0174: im_loop */ + 0xf08004cf, + 0x44b04044, + 0xf71bf400, +/* 0x0180: cmddone */ + 0x1d0037f1, + 0xd00147f0, +/* 0x018a: nocmd */ + 0x11c40034, + 0x4001d00c, +/* 0x0192: cmd_query_get */ + 0x38f201f8, + 0x0325f001, + 0x0b0047f1, +/* 0x019c: ptimer_retry */ + 0xcf4046cf, + 0x47cf0045, + 0x0467b840, + 0x98f41bf4, + 0x04800504, + 0x21008020, + 0x80220580, + 0x0bfe2306, + 0x03049800, + 0xfe1844b6, + 0x04980047, + 0x8057f104, + 0x0253f000, + 0xf80645fa, +/* 0x01d7: cmd_cond_mode */ + 0xf400f803, + 0x25f00131, + 0x0534b002, + 0xf41218f4, + 0x34b00132, + 0x0b18f402, + 0x800136f0, +/* 0x01f2: return */ + 0x00f80803, +/* 0x01f4: cmd_cond_mode_queryful */ + 0x98060498, + 0x56c40705, + 0x0855b6ff, + 0xfd1844b6, + 0x47fe0545, + 0x000bfe00, + 0x008057f1, + 0xfa0253f0, + 0x34b00565, + 0x131bf402, + 0x049803f8, + 0x0044b021, + 0x800b4cf0, + 0x00f80804, +/* 0x022c: cmd_cond_mode_double */ + 0xb61060b6, + 0x65fa1050, + 0x9803f805, + 0x06982005, + 0x0456b824, + 0x980b4cf0, + 0x06982105, + 0x0456b825, + 0xfd0b5cf0, + 0x34b00445, + 0x0b5cf003, + 0x800645fd, + 0x00f80804, +/* 0x0260: cmd_wrcache_flush */ + 0xf10132f4, + 0xbd220027, + 0x0133f034, + 0xf80023d0, +/* 0x0271: sec_cmd_mode */ + 0x0131f400, + 0xb00225f0, + 0x18f40f34, + 0x0132f409, +/* 0x0283: sec_cmd_mode_return */ + 0xf80d0380, +/* 0x0285: sec_cmd_length */ + 0x0034b000, + 0xf4fb0bf4, + 0x47f0033c, + 0x0743f040, + 0xf00604fa, + 0x43f05047, + 0x0604fa06, + 0x3cf503f8, + 0x47f1c407, + 0x4bfe2100, + 0x09049800, + 0x950a0598, + 0x44b60858, + 0x0548fd18, + 0x98ff55c4, + 0x07980b06, + 0x0878950c, + 0xfd1864b6, + 0x77c40568, + 0x0d0898ff, + 0x580284b6, + 0x95f9a889, + 0xf9a98958, + 0x013cf495, + 0x3cf403f8, + 0xf803f861, + 0x18489503, + 0xbb084994, + 0x81b60095, + 0x09088000, + 0x950a0980, + 0x69941868, + 0x0097bb08, + 0x800081b6, + 0x09800b08, + 0x023cf40c, + 0xf05047f0, + 0x04fa0643, + 0xf803f805, +/* 0x0321: sec_copy_prep */ + 0x203cf500, + 0x003cf594, + 0x003cf588, +/* 0x032f: sec_store_prep */ + 0xf500f88c, + 0xf594103c, + 0xf88c063c, +/* 0x0339: sec_ecb_e_prep */ + 0x303cf500, + 0x003cf594, + 0x003cf588, + 0x003cf5d0, +/* 0x034b: sec_ecb_d_prep */ + 0xf500f88c, + 0xf5c8773c, + 0xf594303c, + 0xf588003c, + 0xf5d4003c, + 0xf88c003c, +/* 0x0361: sec_cbc_e_prep */ + 0x403cf500, + 0x003cf594, + 0x063cf588, + 0x663cf5ac, + 0x063cf5d0, +/* 0x0377: sec_cbc_d_prep */ + 0xf500f88c, + 0xf5c8773c, + 0xf594503c, + 0xf584623c, + 0xf588063c, + 0xf5d4603c, + 0xf5ac203c, + 0xf88c003c, +/* 0x0395: sec_pcbc_e_prep */ + 0x503cf500, + 0x003cf594, + 0x063cf588, + 0x663cf5ac, + 0x063cf5d0, + 0x063cf58c, +/* 0x03af: sec_pcbc_d_prep */ + 0xf500f8ac, + 0xf5c8773c, + 0xf594503c, + 0xf588003c, + 0xf5d4013c, + 0xf5ac163c, + 0xf58c063c, + 0xf8ac063c, +/* 0x03cd: sec_cfb_e_prep */ + 0x403cf500, + 0x663cf594, + 0x003cf5d0, + 0x063cf588, + 0x063cf5ac, +/* 0x03e3: sec_cfb_d_prep */ + 0xf500f88c, + 0xf594403c, + 0xf5d0603c, + 0xf588063c, + 0xf5ac603c, + 0xf88c003c, +/* 0x03f9: sec_ofb_prep */ + 0x403cf500, + 0x663cf594, + 0x003cf5d0, + 0x603cf588, + 0x003cf5ac, +/* 0x040f: sec_ctr_prep */ + 0xf500f88c, + 0xf594503c, + 0xf5d0613c, + 0xf5b0163c, + 0xf588003c, + 0xf5ac103c, + 0xf88c003c, +/* 0x0429: sec_cbc_mac_prep */ + 0x303cf500, + 0x003cf594, + 0x063cf588, + 0x663cf5ac, +/* 0x043b: sec_cmac_finish_complete_prep */ + 0xf500f8d0, + 0xf594703c, + 0xf588003c, + 0xf5ac063c, + 0xf5ac003c, + 0xf5d0003c, + 0xf5bc003c, + 0xf5ac063c, + 0xf8d0663c, +/* 0x045d: sec_cmac_finish_partial_prep */ + 0x803cf500, + 0x003cf594, + 0x063cf588, + 0x003cf5ac, + 0x003cf5ac, + 0x003cf5d0, + 0x003cf5bc, + 0x063cf5bc, + 0x663cf5ac, +/* 0x0483: sec_do_in */ + 0xbb00f8d0, + 0x47fe0035, + 0x8097f100, + 0x0293f000, +/* 0x0490: sec_do_in_loop */ + 0xf80559fa, + 0x223cf403, + 0xf50609fa, + 0xf898103c, + 0x1050b603, + 0xf40453b8, + 0x3cf4e91b, + 0xf803f801, +/* 0x04b1: sec_do_out */ + 0x0037bb00, + 0xf10067fe, + 0xf0008097, +/* 0x04be: sec_do_out_loop */ + 0x3cf50293, + 0x3cf49810, + 0x0579fa61, + 0xf40679fa, + 0x03f8013c, + 0xb81070b6, + 0x1bf40473, +/* 0x04db: sec_do_inout */ + 0xbb00f8e8, + 0x97f10035, + 0x93f00080, +/* 0x04e5: sec_do_inout_loop */ + 0x0047fe02, + 0xf80559fa, + 0x213cf403, + 0xf50609fa, + 0xf498103c, + 0x67fe613c, + 0x0579fa00, + 0xf40679fa, + 0x03f8013c, + 0xb61050b6, + 0x53b81070, + 0xd41bf404, + 0x000000f8, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, + 0x00000000, +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec/g98.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec/g98.c new file mode 100644 index 0000000000..c15b2cbf50 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec/g98.c @@ -0,0 +1,81 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include +#include +#include "fuc/g98.fuc0s.h" + +#include +#include +#include + +#include + +static const struct nvkm_enum g98_sec_isr_error_name[] = { + { 0x0000, "ILLEGAL_MTHD" }, + { 0x0001, "INVALID_BITFIELD" }, + { 0x0002, "INVALID_ENUM" }, + { 0x0003, "QUERY" }, + {} +}; + +static void +g98_sec_intr(struct nvkm_falcon *sec, struct nvkm_chan *chan) +{ + struct nvkm_subdev *subdev = &sec->engine.subdev; + struct nvkm_device *device = subdev->device; + u32 ssta = nvkm_rd32(device, 0x087040) & 0x0000ffff; + u32 addr = nvkm_rd32(device, 0x087040) >> 16; + u32 mthd = (addr & 0x07ff) << 2; + u32 subc = (addr & 0x3800) >> 11; + u32 data = nvkm_rd32(device, 0x087044); + const struct nvkm_enum *en = + nvkm_enum_find(g98_sec_isr_error_name, ssta); + + nvkm_error(subdev, "DISPATCH_ERROR %04x [%s] ch %d [%010llx %s] " + "subc %d mthd %04x data %08x\n", ssta, + en ? en->name : "UNKNOWN", chan ? chan->id : -1, + chan ? chan->inst->addr : 0, + chan ? chan->name : "unknown", + subc, mthd, data); +} + +static const struct nvkm_falcon_func +g98_sec = { + .code.data = g98_sec_code, + .code.size = sizeof(g98_sec_code), + .data.data = g98_sec_data, + .data.size = sizeof(g98_sec_data), + .intr = g98_sec_intr, + .sclass = { + { -1, -1, G98_SEC }, + {} + } +}; + +int +g98_sec_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_falcon_new_(&g98_sec, device, type, inst, true, 0x087000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/Kbuild new file mode 100644 index 0000000000..19feadb1f6 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/Kbuild @@ -0,0 +1,6 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/sec2/base.o +nvkm-y += nvkm/engine/sec2/gp102.o +nvkm-y += nvkm/engine/sec2/gp108.o +nvkm-y += nvkm/engine/sec2/tu102.o +nvkm-y += nvkm/engine/sec2/ga102.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/base.c new file mode 100644 index 0000000000..f2c60da5d1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/base.c @@ -0,0 +1,164 @@ +/* + * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include +#include +#include + +#include + +static int +nvkm_sec2_finimsg(void *priv, struct nvfw_falcon_msg *hdr) +{ + struct nvkm_sec2 *sec2 = priv; + + atomic_set(&sec2->running, 0); + return 0; +} + +static int +nvkm_sec2_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_sec2 *sec2 = nvkm_sec2(engine); + struct nvkm_subdev *subdev = &sec2->engine.subdev; + struct nvkm_falcon *falcon = &sec2->falcon; + struct nvkm_falcon_cmdq *cmdq = sec2->cmdq; + struct nvfw_falcon_cmd cmd = { + .unit_id = sec2->func->unit_unload, + .size = sizeof(cmd), + }; + int ret; + + if (!subdev->use.enabled) + return 0; + + if (atomic_read(&sec2->initmsg) == 1) { + ret = nvkm_falcon_cmdq_send(cmdq, &cmd, nvkm_sec2_finimsg, sec2, + msecs_to_jiffies(1000)); + WARN_ON(ret); + + nvkm_msec(subdev->device, 2000, + if (nvkm_falcon_rd32(falcon, 0x100) & 0x00000010) + break; + ); + } + + nvkm_inth_block(&subdev->inth); + + nvkm_falcon_cmdq_fini(cmdq); + falcon->func->disable(falcon); + nvkm_falcon_put(falcon, subdev); + return 0; +} + +static int +nvkm_sec2_init(struct nvkm_engine *engine) +{ + struct nvkm_sec2 *sec2 = nvkm_sec2(engine); + struct nvkm_subdev *subdev = &sec2->engine.subdev; + struct nvkm_falcon *falcon = &sec2->falcon; + int ret; + + ret = nvkm_falcon_get(falcon, subdev); + if (ret) + return ret; + + nvkm_falcon_wr32(falcon, 0x014, 0xffffffff); + atomic_set(&sec2->initmsg, 0); + atomic_set(&sec2->running, 1); + nvkm_inth_allow(&subdev->inth); + + nvkm_falcon_start(falcon); + return 0; +} + +static int +nvkm_sec2_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_sec2 *sec2 = nvkm_sec2(engine); + struct nvkm_subdev *subdev = &sec2->engine.subdev; + struct nvkm_intr *intr = &sec2->engine.subdev.device->mc->intr; + enum nvkm_intr_type type = NVKM_INTR_SUBDEV; + + if (sec2->func->intr_vector) { + intr = sec2->func->intr_vector(sec2, &type); + if (IS_ERR(intr)) + return PTR_ERR(intr); + } + + return nvkm_inth_add(intr, type, NVKM_INTR_PRIO_NORMAL, subdev, sec2->func->intr, + &subdev->inth); +} + +static void * +nvkm_sec2_dtor(struct nvkm_engine *engine) +{ + struct nvkm_sec2 *sec2 = nvkm_sec2(engine); + + nvkm_falcon_msgq_del(&sec2->msgq); + nvkm_falcon_cmdq_del(&sec2->cmdq); + nvkm_falcon_qmgr_del(&sec2->qmgr); + nvkm_falcon_dtor(&sec2->falcon); + return sec2; +} + +static const struct nvkm_engine_func +nvkm_sec2 = { + .dtor = nvkm_sec2_dtor, + .oneinit = nvkm_sec2_oneinit, + .init = nvkm_sec2_init, + .fini = nvkm_sec2_fini, +}; + +int +nvkm_sec2_new_(const struct nvkm_sec2_fwif *fwif, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, u32 addr, struct nvkm_sec2 **psec2) +{ + struct nvkm_sec2 *sec2; + int ret; + + if (!(sec2 = *psec2 = kzalloc(sizeof(*sec2), GFP_KERNEL))) + return -ENOMEM; + + ret = nvkm_engine_ctor(&nvkm_sec2, device, type, inst, true, &sec2->engine); + if (ret) + return ret; + + fwif = nvkm_firmware_load(&sec2->engine.subdev, fwif, "Sec2", sec2); + if (IS_ERR(fwif)) + return PTR_ERR(fwif); + + sec2->func = fwif->func; + + ret = nvkm_falcon_ctor(sec2->func->flcn, &sec2->engine.subdev, + sec2->engine.subdev.name, addr, &sec2->falcon); + if (ret) + return ret; + + if ((ret = nvkm_falcon_qmgr_new(&sec2->falcon, &sec2->qmgr)) || + (ret = nvkm_falcon_cmdq_new(sec2->qmgr, "cmdq", &sec2->cmdq)) || + (ret = nvkm_falcon_msgq_new(sec2->qmgr, "msgq", &sec2->msgq))) + return ret; + + return 0; +}; diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/ga102.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/ga102.c new file mode 100644 index 0000000000..945abb8156 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/ga102.c @@ -0,0 +1,197 @@ +/* + * Copyright 2021 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include +#include + +#include +#include + +static int +ga102_sec2_initmsg(struct nvkm_sec2 *sec2) +{ + struct nv_sec2_init_msg_v1 msg; + int ret, i; + + ret = nvkm_falcon_msgq_recv_initmsg(sec2->msgq, &msg, sizeof(msg)); + if (ret) + return ret; + + if (msg.hdr.unit_id != NV_SEC2_UNIT_INIT || + msg.msg_type != NV_SEC2_INIT_MSG_INIT) + return -EINVAL; + + for (i = 0; i < ARRAY_SIZE(msg.queue_info); i++) { + if (msg.queue_info[i].id == NV_SEC2_INIT_MSG_QUEUE_ID_MSGQ) { + nvkm_falcon_msgq_init(sec2->msgq, msg.queue_info[i].index, + msg.queue_info[i].offset, + msg.queue_info[i].size); + } else { + nvkm_falcon_cmdq_init(sec2->cmdq, msg.queue_info[i].index, + msg.queue_info[i].offset, + msg.queue_info[i].size); + } + } + + return 0; +} + +static struct nvkm_intr * +ga102_sec2_intr_vector(struct nvkm_sec2 *sec2, enum nvkm_intr_type *pvector) +{ + struct nvkm_device *device = sec2->engine.subdev.device; + struct nvkm_falcon *falcon = &sec2->falcon; + int ret; + + ret = ga102_flcn_select(falcon); + if (ret) + return ERR_PTR(ret); + + *pvector = nvkm_rd32(device, 0x8403e0) & 0x000000ff; + return &device->vfn->intr; +} + +static int +ga102_sec2_acr_bootstrap_falcon_callback(void *priv, struct nvfw_falcon_msg *hdr) +{ + struct nv_sec2_acr_bootstrap_falcon_msg_v1 *msg = + container_of(hdr, typeof(*msg), msg.hdr); + struct nvkm_subdev *subdev = priv; + const char *name = nvkm_acr_lsf_id(msg->falcon_id); + + if (msg->error_code) { + nvkm_error(subdev, "ACR_BOOTSTRAP_FALCON failed for falcon %d [%s]: %08x %08x\n", + msg->falcon_id, name, msg->error_code, msg->unkn08); + return -EINVAL; + } + + nvkm_debug(subdev, "%s booted\n", name); + return 0; +} + +static int +ga102_sec2_acr_bootstrap_falcon(struct nvkm_falcon *falcon, enum nvkm_acr_lsf_id id) +{ + struct nvkm_sec2 *sec2 = container_of(falcon, typeof(*sec2), falcon); + struct nv_sec2_acr_bootstrap_falcon_cmd_v1 cmd = { + .cmd.hdr.unit_id = sec2->func->unit_acr, + .cmd.hdr.size = sizeof(cmd), + .cmd.cmd_type = NV_SEC2_ACR_CMD_BOOTSTRAP_FALCON, + .flags = NV_SEC2_ACR_BOOTSTRAP_FALCON_FLAGS_RESET_YES, + .falcon_id = id, + }; + + return nvkm_falcon_cmdq_send(sec2->cmdq, &cmd.cmd.hdr, + ga102_sec2_acr_bootstrap_falcon_callback, + &sec2->engine.subdev, + msecs_to_jiffies(1000)); +} + +static const struct nvkm_acr_lsf_func +ga102_sec2_acr_0 = { + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp102_sec2_acr_bld_write_1, + .bld_patch = gp102_sec2_acr_bld_patch_1, + .bootstrap_falcons = BIT_ULL(NVKM_ACR_LSF_FECS) | + BIT_ULL(NVKM_ACR_LSF_GPCCS) | + BIT_ULL(NVKM_ACR_LSF_SEC2), + .bootstrap_falcon = ga102_sec2_acr_bootstrap_falcon, +}; + +static const struct nvkm_falcon_func +ga102_sec2_flcn = { + .disable = gm200_flcn_disable, + .enable = gm200_flcn_enable, + .select = ga102_flcn_select, + .addr2 = 0x1000, + .reset_pmc = true, + .reset_eng = gp102_flcn_reset_eng, + .reset_prep = ga102_flcn_reset_prep, + .reset_wait_mem_scrubbing = ga102_flcn_reset_wait_mem_scrubbing, + .imem_dma = &ga102_flcn_dma, + .dmem_pio = &gm200_flcn_dmem_pio, + .dmem_dma = &ga102_flcn_dma, + .emem_addr = 0x01000000, + .emem_pio = &gp102_flcn_emem_pio, + .start = nvkm_falcon_v1_start, + .cmdq = { 0xc00, 0xc04, 8 }, + .msgq = { 0xc80, 0xc84, 8 }, +}; + +static const struct nvkm_sec2_func +ga102_sec2 = { + .flcn = &ga102_sec2_flcn, + .intr_vector = ga102_sec2_intr_vector, + .intr = gp102_sec2_intr, + .initmsg = ga102_sec2_initmsg, + .unit_acr = NV_SEC2_UNIT_V2_ACR, + .unit_unload = NV_SEC2_UNIT_V2_UNLOAD, +}; + +MODULE_FIRMWARE("nvidia/ga102/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/ga102/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/ga102/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/ga102/sec2/hs_bl_sig.bin"); + +MODULE_FIRMWARE("nvidia/ga103/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/ga103/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/ga103/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/ga103/sec2/hs_bl_sig.bin"); + +MODULE_FIRMWARE("nvidia/ga104/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/ga104/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/ga104/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/ga104/sec2/hs_bl_sig.bin"); + +MODULE_FIRMWARE("nvidia/ga106/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/ga106/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/ga106/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/ga106/sec2/hs_bl_sig.bin"); + +MODULE_FIRMWARE("nvidia/ga107/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/ga107/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/ga107/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/ga107/sec2/hs_bl_sig.bin"); + +static int +ga102_sec2_load(struct nvkm_sec2 *sec2, int ver, + const struct nvkm_sec2_fwif *fwif) +{ + return nvkm_acr_lsfw_load_sig_image_desc_v2(&sec2->engine.subdev, &sec2->falcon, + NVKM_ACR_LSF_SEC2, "sec2/", ver, fwif->acr); +} + +static const struct nvkm_sec2_fwif +ga102_sec2_fwif[] = { + { 0, ga102_sec2_load, &ga102_sec2, &ga102_sec2_acr_0 }, + { -1, gp102_sec2_nofw, &ga102_sec2 } +}; + +int +ga102_sec2_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_sec2 **psec2) +{ + /* TOP info wasn't updated on Turing to reflect the PRI + * address change for some reason. We override it here. + */ + return nvkm_sec2_new_(ga102_sec2_fwif, device, type, inst, 0x840000, psec2); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c new file mode 100644 index 0000000000..c64013d105 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c @@ -0,0 +1,318 @@ +/* + * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" + +#include +#include +#include + +#include +#include + +int +gp102_sec2_nofw(struct nvkm_sec2 *sec2, int ver, + const struct nvkm_sec2_fwif *fwif) +{ + nvkm_warn(&sec2->engine.subdev, "firmware unavailable\n"); + return 0; +} + +static int +gp102_sec2_acr_bootstrap_falcon_callback(void *priv, struct nvfw_falcon_msg *hdr) +{ + struct nv_sec2_acr_bootstrap_falcon_msg *msg = + container_of(hdr, typeof(*msg), msg.hdr); + struct nvkm_subdev *subdev = priv; + const char *name = nvkm_acr_lsf_id(msg->falcon_id); + + if (msg->error_code) { + nvkm_error(subdev, "ACR_BOOTSTRAP_FALCON failed for " + "falcon %d [%s]: %08x\n", + msg->falcon_id, name, msg->error_code); + return -EINVAL; + } + + nvkm_debug(subdev, "%s booted\n", name); + return 0; +} + +static int +gp102_sec2_acr_bootstrap_falcon(struct nvkm_falcon *falcon, + enum nvkm_acr_lsf_id id) +{ + struct nvkm_sec2 *sec2 = container_of(falcon, typeof(*sec2), falcon); + struct nv_sec2_acr_bootstrap_falcon_cmd cmd = { + .cmd.hdr.unit_id = sec2->func->unit_acr, + .cmd.hdr.size = sizeof(cmd), + .cmd.cmd_type = NV_SEC2_ACR_CMD_BOOTSTRAP_FALCON, + .flags = NV_SEC2_ACR_BOOTSTRAP_FALCON_FLAGS_RESET_YES, + .falcon_id = id, + }; + + return nvkm_falcon_cmdq_send(sec2->cmdq, &cmd.cmd.hdr, + gp102_sec2_acr_bootstrap_falcon_callback, + &sec2->engine.subdev, + msecs_to_jiffies(1000)); +} + +static void +gp102_sec2_acr_bld_patch(struct nvkm_acr *acr, u32 bld, s64 adjust) +{ + struct loader_config_v1 hdr; + nvkm_robj(acr->wpr, bld, &hdr, sizeof(hdr)); + hdr.code_dma_base = hdr.code_dma_base + adjust; + hdr.data_dma_base = hdr.data_dma_base + adjust; + hdr.overlay_dma_base = hdr.overlay_dma_base + adjust; + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); + loader_config_v1_dump(&acr->subdev, &hdr); +} + +static void +gp102_sec2_acr_bld_write(struct nvkm_acr *acr, u32 bld, + struct nvkm_acr_lsfw *lsfw) +{ + const struct loader_config_v1 hdr = { + .dma_idx = FALCON_SEC2_DMAIDX_UCODE, + .code_dma_base = lsfw->offset.img + lsfw->app_start_offset, + .code_size_total = lsfw->app_size, + .code_size_to_load = lsfw->app_resident_code_size, + .code_entry_point = lsfw->app_imem_entry, + .data_dma_base = lsfw->offset.img + lsfw->app_start_offset + + lsfw->app_resident_data_offset, + .data_size = lsfw->app_resident_data_size, + .overlay_dma_base = lsfw->offset.img + lsfw->app_start_offset, + .argc = 1, + .argv = lsfw->falcon->func->emem_addr, + }; + + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); +} + +static const struct nvkm_acr_lsf_func +gp102_sec2_acr_0 = { + .bld_size = sizeof(struct loader_config_v1), + .bld_write = gp102_sec2_acr_bld_write, + .bld_patch = gp102_sec2_acr_bld_patch, + .bootstrap_falcons = BIT_ULL(NVKM_ACR_LSF_FECS) | + BIT_ULL(NVKM_ACR_LSF_GPCCS) | + BIT_ULL(NVKM_ACR_LSF_SEC2), + .bootstrap_falcon = gp102_sec2_acr_bootstrap_falcon, +}; + +int +gp102_sec2_initmsg(struct nvkm_sec2 *sec2) +{ + struct nv_sec2_init_msg msg; + int ret, i; + + ret = nvkm_falcon_msgq_recv_initmsg(sec2->msgq, &msg, sizeof(msg)); + if (ret) + return ret; + + if (msg.hdr.unit_id != NV_SEC2_UNIT_INIT || + msg.msg_type != NV_SEC2_INIT_MSG_INIT) + return -EINVAL; + + for (i = 0; i < ARRAY_SIZE(msg.queue_info); i++) { + if (msg.queue_info[i].id == NV_SEC2_INIT_MSG_QUEUE_ID_MSGQ) { + nvkm_falcon_msgq_init(sec2->msgq, + msg.queue_info[i].index, + msg.queue_info[i].offset, + msg.queue_info[i].size); + } else { + nvkm_falcon_cmdq_init(sec2->cmdq, + msg.queue_info[i].index, + msg.queue_info[i].offset, + msg.queue_info[i].size); + } + } + + return 0; +} + +irqreturn_t +gp102_sec2_intr(struct nvkm_inth *inth) +{ + struct nvkm_sec2 *sec2 = container_of(inth, typeof(*sec2), engine.subdev.inth); + struct nvkm_subdev *subdev = &sec2->engine.subdev; + struct nvkm_falcon *falcon = &sec2->falcon; + u32 disp = nvkm_falcon_rd32(falcon, 0x01c); + u32 intr = nvkm_falcon_rd32(falcon, 0x008) & disp & ~(disp >> 16); + + if (intr & 0x00000040) { + if (unlikely(atomic_read(&sec2->initmsg) == 0)) { + int ret = sec2->func->initmsg(sec2); + + if (ret) + nvkm_error(subdev, "error parsing init message: %d\n", ret); + + atomic_set(&sec2->initmsg, ret ?: 1); + } + + if (atomic_read(&sec2->initmsg) > 0) { + if (!nvkm_falcon_msgq_empty(sec2->msgq)) + nvkm_falcon_msgq_recv(sec2->msgq); + } + + nvkm_falcon_wr32(falcon, 0x004, 0x00000040); + intr &= ~0x00000040; + } + + if (intr & 0x00000010) { + if (atomic_read(&sec2->running)) { + FLCN_ERR(falcon, "halted"); + gm200_flcn_tracepc(falcon); + } + + nvkm_falcon_wr32(falcon, 0x004, 0x00000010); + intr &= ~0x00000010; + } + + if (intr) { + nvkm_error(subdev, "unhandled intr %08x\n", intr); + nvkm_falcon_wr32(falcon, 0x004, intr); + } + + return IRQ_HANDLED; +} + +static const struct nvkm_falcon_func +gp102_sec2_flcn = { + .disable = gm200_flcn_disable, + .enable = gm200_flcn_enable, + .reset_pmc = true, + .reset_eng = gp102_flcn_reset_eng, + .reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing, + .debug = 0x408, + .bind_inst = gm200_flcn_bind_inst, + .bind_stat = gm200_flcn_bind_stat, + .bind_intr = true, + .imem_pio = &gm200_flcn_imem_pio, + .dmem_pio = &gm200_flcn_dmem_pio, + .emem_addr = 0x01000000, + .emem_pio = &gp102_flcn_emem_pio, + .start = nvkm_falcon_v1_start, + .cmdq = { 0xa00, 0xa04, 8 }, + .msgq = { 0xa30, 0xa34, 8 }, +}; + +const struct nvkm_sec2_func +gp102_sec2 = { + .flcn = &gp102_sec2_flcn, + .unit_unload = NV_SEC2_UNIT_UNLOAD, + .unit_acr = NV_SEC2_UNIT_ACR, + .intr = gp102_sec2_intr, + .initmsg = gp102_sec2_initmsg, +}; + +MODULE_FIRMWARE("nvidia/gp102/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gp102/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gp102/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/sig.bin"); + +void +gp102_sec2_acr_bld_patch_1(struct nvkm_acr *acr, u32 bld, s64 adjust) +{ + struct flcn_bl_dmem_desc_v2 hdr; + nvkm_robj(acr->wpr, bld, &hdr, sizeof(hdr)); + hdr.code_dma_base = hdr.code_dma_base + adjust; + hdr.data_dma_base = hdr.data_dma_base + adjust; + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); + flcn_bl_dmem_desc_v2_dump(&acr->subdev, &hdr); +} + +void +gp102_sec2_acr_bld_write_1(struct nvkm_acr *acr, u32 bld, + struct nvkm_acr_lsfw *lsfw) +{ + const struct flcn_bl_dmem_desc_v2 hdr = { + .ctx_dma = FALCON_SEC2_DMAIDX_UCODE, + .code_dma_base = lsfw->offset.img + lsfw->app_start_offset, + .non_sec_code_off = lsfw->app_resident_code_offset, + .non_sec_code_size = lsfw->app_resident_code_size, + .code_entry_point = lsfw->app_imem_entry, + .data_dma_base = lsfw->offset.img + lsfw->app_start_offset + + lsfw->app_resident_data_offset, + .data_size = lsfw->app_resident_data_size, + .argc = 1, + .argv = lsfw->falcon->func->emem_addr, + }; + + nvkm_wobj(acr->wpr, bld, &hdr, sizeof(hdr)); +} + +const struct nvkm_acr_lsf_func +gp102_sec2_acr_1 = { + .bld_size = sizeof(struct flcn_bl_dmem_desc_v2), + .bld_write = gp102_sec2_acr_bld_write_1, + .bld_patch = gp102_sec2_acr_bld_patch_1, + .bootstrap_falcons = BIT_ULL(NVKM_ACR_LSF_FECS) | + BIT_ULL(NVKM_ACR_LSF_GPCCS) | + BIT_ULL(NVKM_ACR_LSF_SEC2), + .bootstrap_falcon = gp102_sec2_acr_bootstrap_falcon, +}; + +int +gp102_sec2_load(struct nvkm_sec2 *sec2, int ver, + const struct nvkm_sec2_fwif *fwif) +{ + return nvkm_acr_lsfw_load_sig_image_desc_v1(&sec2->engine.subdev, + &sec2->falcon, + NVKM_ACR_LSF_SEC2, "sec2/", + ver, fwif->acr); +} + +MODULE_FIRMWARE("nvidia/gp102/sec2/desc-1.bin"); +MODULE_FIRMWARE("nvidia/gp102/sec2/image-1.bin"); +MODULE_FIRMWARE("nvidia/gp102/sec2/sig-1.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/desc-1.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/image-1.bin"); +MODULE_FIRMWARE("nvidia/gp104/sec2/sig-1.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/desc-1.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/image-1.bin"); +MODULE_FIRMWARE("nvidia/gp106/sec2/sig-1.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/desc-1.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/image-1.bin"); +MODULE_FIRMWARE("nvidia/gp107/sec2/sig-1.bin"); + +static const struct nvkm_sec2_fwif +gp102_sec2_fwif[] = { + { 1, gp102_sec2_load, &gp102_sec2, &gp102_sec2_acr_1 }, + { 0, gp102_sec2_load, &gp102_sec2, &gp102_sec2_acr_0 }, + { -1, gp102_sec2_nofw, &gp102_sec2 }, + {} +}; + +int +gp102_sec2_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_sec2 **psec2) +{ + return nvkm_sec2_new_(gp102_sec2_fwif, device, type, inst, 0, psec2); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp108.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp108.c new file mode 100644 index 0000000000..3e9f5c842f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp108.c @@ -0,0 +1,43 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include + +MODULE_FIRMWARE("nvidia/gp108/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gp108/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gp108/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/gv100/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/gv100/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/gv100/sec2/sig.bin"); + +static const struct nvkm_sec2_fwif +gp108_sec2_fwif[] = { + { 0, gp102_sec2_load, &gp102_sec2, &gp102_sec2_acr_1 }, + {} +}; + +int +gp108_sec2_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_sec2 **psec2) +{ + return nvkm_sec2_new_(gp108_sec2_fwif, device, type, inst, 0, psec2); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/priv.h new file mode 100644 index 0000000000..172d2705c1 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/priv.h @@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_SEC2_PRIV_H__ +#define __NVKM_SEC2_PRIV_H__ +#include +struct nvkm_acr_lsfw; + +struct nvkm_sec2_func { + const struct nvkm_falcon_func *flcn; + u8 unit_unload; + u8 unit_acr; + struct nvkm_intr *(*intr_vector)(struct nvkm_sec2 *, enum nvkm_intr_type *); + irqreturn_t (*intr)(struct nvkm_inth *); + int (*initmsg)(struct nvkm_sec2 *); +}; + +irqreturn_t gp102_sec2_intr(struct nvkm_inth *); +int gp102_sec2_initmsg(struct nvkm_sec2 *); + +struct nvkm_sec2_fwif { + int version; + int (*load)(struct nvkm_sec2 *, int ver, const struct nvkm_sec2_fwif *); + const struct nvkm_sec2_func *func; + const struct nvkm_acr_lsf_func *acr; +}; + +int gp102_sec2_nofw(struct nvkm_sec2 *, int, const struct nvkm_sec2_fwif *); +int gp102_sec2_load(struct nvkm_sec2 *, int, const struct nvkm_sec2_fwif *); +extern const struct nvkm_sec2_func gp102_sec2; +extern const struct nvkm_acr_lsf_func gp102_sec2_acr_1; +void gp102_sec2_acr_bld_write_1(struct nvkm_acr *, u32, struct nvkm_acr_lsfw *); +void gp102_sec2_acr_bld_patch_1(struct nvkm_acr *, u32, s64); + +int nvkm_sec2_new_(const struct nvkm_sec2_fwif *, struct nvkm_device *, enum nvkm_subdev_type, + int, u32 addr, struct nvkm_sec2 **); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c new file mode 100644 index 0000000000..0afc4b2fa5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c @@ -0,0 +1,86 @@ +/* + * Copyright 2019 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include "priv.h" +#include + +#include + +static const struct nvkm_falcon_func +tu102_sec2_flcn = { + .disable = gm200_flcn_disable, + .enable = gm200_flcn_enable, + .reset_pmc = true, + .reset_eng = gp102_flcn_reset_eng, + .reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing, + .debug = 0x408, + .bind_inst = gm200_flcn_bind_inst, + .bind_stat = gm200_flcn_bind_stat, + .bind_intr = true, + .imem_pio = &gm200_flcn_imem_pio, + .dmem_pio = &gm200_flcn_dmem_pio, + .emem_addr = 0x01000000, + .emem_pio = &gp102_flcn_emem_pio, + .start = nvkm_falcon_v1_start, + .cmdq = { 0xc00, 0xc04, 8 }, + .msgq = { 0xc80, 0xc84, 8 }, +}; + +static const struct nvkm_sec2_func +tu102_sec2 = { + .flcn = &tu102_sec2_flcn, + .unit_unload = NV_SEC2_UNIT_V2_UNLOAD, + .unit_acr = NV_SEC2_UNIT_V2_ACR, + .intr = gp102_sec2_intr, + .initmsg = gp102_sec2_initmsg, +}; + +MODULE_FIRMWARE("nvidia/tu102/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/tu102/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/tu102/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/tu104/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/tu104/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/tu104/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/tu106/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/tu106/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/tu106/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/tu116/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/tu116/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/tu116/sec2/sig.bin"); +MODULE_FIRMWARE("nvidia/tu117/sec2/desc.bin"); +MODULE_FIRMWARE("nvidia/tu117/sec2/image.bin"); +MODULE_FIRMWARE("nvidia/tu117/sec2/sig.bin"); + +static const struct nvkm_sec2_fwif +tu102_sec2_fwif[] = { + { 0, gp102_sec2_load, &tu102_sec2, &gp102_sec2_acr_1 }, + { -1, gp102_sec2_nofw, &tu102_sec2 } +}; + +int +tu102_sec2_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_sec2 **psec2) +{ + /* TOP info wasn't updated on Turing to reflect the PRI + * address change for some reason. We override it here. + */ + return nvkm_sec2_new_(tu102_sec2_fwif, device, type, inst, 0x840000, psec2); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/sw/Kbuild new file mode 100644 index 0000000000..94fe25964f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/Kbuild @@ -0,0 +1,10 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/sw/base.o +nvkm-y += nvkm/engine/sw/nv04.o +nvkm-y += nvkm/engine/sw/nv10.o +nvkm-y += nvkm/engine/sw/nv50.o +nvkm-y += nvkm/engine/sw/gf100.o + +nvkm-y += nvkm/engine/sw/chan.o + +nvkm-y += nvkm/engine/sw/nvsw.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/base.c new file mode 100644 index 0000000000..20220d6d4a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/base.c @@ -0,0 +1,109 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" + +#include + +bool +nvkm_sw_mthd(struct nvkm_sw *sw, int chid, int subc, u32 mthd, u32 data) +{ + struct nvkm_sw_chan *chan; + bool handled = false; + unsigned long flags; + + spin_lock_irqsave(&sw->engine.lock, flags); + list_for_each_entry(chan, &sw->chan, head) { + if (chan->fifo->id == chid) { + handled = nvkm_sw_chan_mthd(chan, subc, mthd, data); + list_del(&chan->head); + list_add(&chan->head, &sw->chan); + break; + } + } + spin_unlock_irqrestore(&sw->engine.lock, flags); + return handled; +} + +static int +nvkm_sw_oclass_new(const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + struct nvkm_sw_chan *chan = nvkm_sw_chan(oclass->parent); + const struct nvkm_sw_chan_sclass *sclass = oclass->engn; + return sclass->ctor(chan, oclass, data, size, pobject); +} + +static int +nvkm_sw_oclass_get(struct nvkm_oclass *oclass, int index) +{ + struct nvkm_sw *sw = nvkm_sw(oclass->engine); + int c = 0; + + while (sw->func->sclass[c].ctor) { + if (c++ == index) { + oclass->engn = &sw->func->sclass[index]; + oclass->base = sw->func->sclass[index].base; + oclass->base.ctor = nvkm_sw_oclass_new; + return index; + } + } + + return c; +} + +static int +nvkm_sw_cclass_get(struct nvkm_chan *fifoch, const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nvkm_sw *sw = nvkm_sw(oclass->engine); + return sw->func->chan_new(sw, fifoch, oclass, pobject); +} + +static void * +nvkm_sw_dtor(struct nvkm_engine *engine) +{ + return nvkm_sw(engine); +} + +static const struct nvkm_engine_func +nvkm_sw = { + .dtor = nvkm_sw_dtor, + .fifo.cclass = nvkm_sw_cclass_get, + .fifo.sclass = nvkm_sw_oclass_get, +}; + +int +nvkm_sw_new_(const struct nvkm_sw_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, struct nvkm_sw **psw) +{ + struct nvkm_sw *sw; + + if (!(sw = *psw = kzalloc(sizeof(*sw), GFP_KERNEL))) + return -ENOMEM; + INIT_LIST_HEAD(&sw->chan); + sw->func = func; + + return nvkm_engine_ctor(&nvkm_sw, device, type, inst, true, &sw->engine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.c new file mode 100644 index 0000000000..2bf45141de --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.c @@ -0,0 +1,91 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "chan.h" + +#include + +#include +#include + +bool +nvkm_sw_chan_mthd(struct nvkm_sw_chan *chan, int subc, u32 mthd, u32 data) +{ + switch (mthd) { + case 0x0000: + return true; + case 0x0500: + nvkm_event_ntfy(&chan->event, 0, NVKM_SW_CHAN_EVENT_PAGE_FLIP); + return true; + default: + if (chan->func->mthd) + return chan->func->mthd(chan, subc, mthd, data); + break; + } + return false; +} + +static const struct nvkm_event_func +nvkm_sw_chan_event = { +}; + +static void * +nvkm_sw_chan_dtor(struct nvkm_object *object) +{ + struct nvkm_sw_chan *chan = nvkm_sw_chan(object); + struct nvkm_sw *sw = chan->sw; + unsigned long flags; + void *data = chan; + + if (chan->func->dtor) + data = chan->func->dtor(chan); + nvkm_event_fini(&chan->event); + + spin_lock_irqsave(&sw->engine.lock, flags); + list_del(&chan->head); + spin_unlock_irqrestore(&sw->engine.lock, flags); + return data; +} + +static const struct nvkm_object_func +nvkm_sw_chan = { + .dtor = nvkm_sw_chan_dtor, +}; + +int +nvkm_sw_chan_ctor(const struct nvkm_sw_chan_func *func, struct nvkm_sw *sw, + struct nvkm_chan *fifo, const struct nvkm_oclass *oclass, + struct nvkm_sw_chan *chan) +{ + unsigned long flags; + + nvkm_object_ctor(&nvkm_sw_chan, oclass, &chan->object); + chan->func = func; + chan->sw = sw; + chan->fifo = fifo; + spin_lock_irqsave(&sw->engine.lock, flags); + list_add(&chan->head, &sw->chan); + spin_unlock_irqrestore(&sw->engine.lock, flags); + + return nvkm_event_init(&nvkm_sw_chan_event, &sw->engine.subdev, 1, 1, &chan->event); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.h b/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.h new file mode 100644 index 0000000000..c313aea16a --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/chan.h @@ -0,0 +1,30 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_SW_CHAN_H__ +#define __NVKM_SW_CHAN_H__ +#define nvkm_sw_chan(p) container_of((p), struct nvkm_sw_chan, object) +#include +#include + +#include "priv.h" + +struct nvkm_sw_chan { + const struct nvkm_sw_chan_func *func; + struct nvkm_object object; + struct nvkm_sw *sw; + struct nvkm_chan *fifo; + struct list_head head; + +#define NVKM_SW_CHAN_EVENT_PAGE_FLIP BIT(0) + struct nvkm_event event; +}; + +struct nvkm_sw_chan_func { + void *(*dtor)(struct nvkm_sw_chan *); + bool (*mthd)(struct nvkm_sw_chan *, int subc, u32 mthd, u32 data); +}; + +int nvkm_sw_chan_ctor(const struct nvkm_sw_chan_func *, struct nvkm_sw *, + struct nvkm_chan *, const struct nvkm_oclass *, + struct nvkm_sw_chan *); +bool nvkm_sw_chan_mthd(struct nvkm_sw_chan *, int subc, u32 mthd, u32 data); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/gf100.c new file mode 100644 index 0000000000..a0273baf4c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/gf100.c @@ -0,0 +1,147 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include +#include +#include +#include + +#include +#include + +/******************************************************************************* + * software context + ******************************************************************************/ + +static int +gf100_sw_chan_vblsem_release(struct nvkm_event_ntfy *notify, u32 bits) +{ + struct nv50_sw_chan *chan = + container_of(notify, typeof(*chan), vblank.notify[notify->id]); + struct nvkm_sw *sw = chan->base.sw; + struct nvkm_device *device = sw->engine.subdev.device; + u32 inst = chan->base.fifo->inst->addr >> 12; + + nvkm_wr32(device, 0x001718, 0x80000000 | inst); + nvkm_bar_flush(device->bar); + nvkm_wr32(device, 0x06000c, upper_32_bits(chan->vblank.offset)); + nvkm_wr32(device, 0x060010, lower_32_bits(chan->vblank.offset)); + nvkm_wr32(device, 0x060014, chan->vblank.value); + + return NVKM_EVENT_DROP; +} + +static bool +gf100_sw_chan_mthd(struct nvkm_sw_chan *base, int subc, u32 mthd, u32 data) +{ + struct nv50_sw_chan *chan = nv50_sw_chan(base); + struct nvkm_engine *engine = chan->base.object.engine; + struct nvkm_device *device = engine->subdev.device; + switch (mthd) { + case 0x0400: + chan->vblank.offset &= 0x00ffffffffULL; + chan->vblank.offset |= (u64)data << 32; + return true; + case 0x0404: + chan->vblank.offset &= 0xff00000000ULL; + chan->vblank.offset |= data; + return true; + case 0x0408: + chan->vblank.value = data; + return true; + case 0x040c: + if (data < device->disp->vblank.index_nr) { + nvkm_event_ntfy_allow(&chan->vblank.notify[data]); + return true; + } + break; + case 0x600: /* MP.PM_UNK000 */ + nvkm_wr32(device, 0x419e00, data); + return true; + case 0x644: /* MP.TRAP_WARP_ERROR_EN */ + if (!(data & ~0x001ffffe)) { + nvkm_wr32(device, 0x419e44, data); + return true; + } + break; + case 0x6ac: /* MP.PM_UNK0AC */ + nvkm_wr32(device, 0x419eac, data); + return true; + default: + break; + } + return false; +} + +static const struct nvkm_sw_chan_func +gf100_sw_chan = { + .dtor = nv50_sw_chan_dtor, + .mthd = gf100_sw_chan_mthd, +}; + +static int +gf100_sw_chan_new(struct nvkm_sw *sw, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = sw->engine.subdev.device->disp; + struct nv50_sw_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + *pobject = &chan->base.object; + + ret = nvkm_sw_chan_ctor(&gf100_sw_chan, sw, fifoch, oclass, + &chan->base); + if (ret) + return ret; + + for (i = 0; disp && i < disp->vblank.index_nr; i++) { + nvkm_event_ntfy_add(&disp->vblank, i, NVKM_DISP_HEAD_EVENT_VBLANK, true, + gf100_sw_chan_vblsem_release, &chan->vblank.notify[i]); + } + + return 0; +} + +/******************************************************************************* + * software engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_sw_func +gf100_sw = { + .chan_new = gf100_sw_chan_new, + .sclass = { + { nvkm_nvsw_new, { -1, -1, NVIF_CLASS_SW_GF100 } }, + {} + } +}; + +int +gf100_sw_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_sw **psw) +{ + return nvkm_sw_new_(&gf100_sw, device, type, inst, psw); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv04.c new file mode 100644 index 0000000000..8a1d112da8 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv04.c @@ -0,0 +1,139 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#define nv04_sw_chan(p) container_of((p), struct nv04_sw_chan, base) +#include "priv.h" +#include "chan.h" +#include "nvsw.h" + +#include +#include +#include +#include + +struct nv04_sw_chan { + struct nvkm_sw_chan base; + atomic_t ref; +}; + +/******************************************************************************* + * software object classes + ******************************************************************************/ + +static int +nv04_nvsw_mthd_get_ref(struct nvkm_nvsw *nvsw, void *data, u32 size) +{ + struct nv04_sw_chan *chan = nv04_sw_chan(nvsw->chan); + union { + struct nv04_nvsw_get_ref_v0 v0; + } *args = data; + int ret = -ENOSYS; + + if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { + args->v0.ref = atomic_read(&chan->ref); + } + + return ret; +} + +static int +nv04_nvsw_mthd(struct nvkm_nvsw *nvsw, u32 mthd, void *data, u32 size) +{ + switch (mthd) { + case NV04_NVSW_GET_REF: + return nv04_nvsw_mthd_get_ref(nvsw, data, size); + default: + break; + } + return -EINVAL; +} + +static const struct nvkm_nvsw_func +nv04_nvsw = { + .mthd = nv04_nvsw_mthd, +}; + +static int +nv04_nvsw_new(struct nvkm_sw_chan *chan, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + return nvkm_nvsw_new_(&nv04_nvsw, chan, oclass, data, size, pobject); +} + +/******************************************************************************* + * software context + ******************************************************************************/ + +static bool +nv04_sw_chan_mthd(struct nvkm_sw_chan *base, int subc, u32 mthd, u32 data) +{ + struct nv04_sw_chan *chan = nv04_sw_chan(base); + + switch (mthd) { + case 0x0150: + atomic_set(&chan->ref, data); + return true; + default: + break; + } + + return false; +} + +static const struct nvkm_sw_chan_func +nv04_sw_chan = { + .mthd = nv04_sw_chan_mthd, +}; + +static int +nv04_sw_chan_new(struct nvkm_sw *sw, struct nvkm_chan *fifo, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv04_sw_chan *chan; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + atomic_set(&chan->ref, 0); + *pobject = &chan->base.object; + + return nvkm_sw_chan_ctor(&nv04_sw_chan, sw, fifo, oclass, &chan->base); +} + +/******************************************************************************* + * software engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_sw_func +nv04_sw = { + .chan_new = nv04_sw_chan_new, + .sclass = { + { nv04_nvsw_new, { -1, -1, NVIF_CLASS_SW_NV04 } }, + {} + } +}; + +int +nv04_sw_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_sw **psw) +{ + return nvkm_sw_new_(&nv04_sw, device, type, inst, psw); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv10.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv10.c new file mode 100644 index 0000000000..742c758595 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv10.c @@ -0,0 +1,68 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "priv.h" +#include "chan.h" +#include "nvsw.h" + +#include + +/******************************************************************************* + * software context + ******************************************************************************/ + +static const struct nvkm_sw_chan_func +nv10_sw_chan = { +}; + +static int +nv10_sw_chan_new(struct nvkm_sw *sw, struct nvkm_chan *fifo, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nvkm_sw_chan *chan; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + *pobject = &chan->object; + + return nvkm_sw_chan_ctor(&nv10_sw_chan, sw, fifo, oclass, chan); +} + +/******************************************************************************* + * software engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_sw_func +nv10_sw = { + .chan_new = nv10_sw_chan_new, + .sclass = { + { nvkm_nvsw_new, { -1, -1, NVIF_CLASS_SW_NV10 } }, + {} + } +}; + +int +nv10_sw_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_sw **psw) +{ + return nvkm_sw_new_(&nv10_sw, device, type, inst, psw); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.c new file mode 100644 index 0000000000..99476d32c5 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.c @@ -0,0 +1,142 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include +#include +#include +#include + +#include +#include + +/******************************************************************************* + * software context + ******************************************************************************/ + +static int +nv50_sw_chan_vblsem_release(struct nvkm_event_ntfy *notify, u32 bits) +{ + struct nv50_sw_chan *chan = + container_of(notify, typeof(*chan), vblank.notify[notify->id]); + struct nvkm_sw *sw = chan->base.sw; + struct nvkm_device *device = sw->engine.subdev.device; + + nvkm_wr32(device, 0x001704, chan->base.fifo->inst->addr >> 12); + nvkm_wr32(device, 0x001710, 0x80000000 | chan->vblank.ctxdma); + nvkm_bar_flush(device->bar); + + if (device->chipset == 0x50) { + nvkm_wr32(device, 0x001570, chan->vblank.offset); + nvkm_wr32(device, 0x001574, chan->vblank.value); + } else { + nvkm_wr32(device, 0x060010, chan->vblank.offset); + nvkm_wr32(device, 0x060014, chan->vblank.value); + } + + return NVKM_EVENT_DROP; +} + +static bool +nv50_sw_chan_mthd(struct nvkm_sw_chan *base, int subc, u32 mthd, u32 data) +{ + struct nv50_sw_chan *chan = nv50_sw_chan(base); + struct nvkm_engine *engine = chan->base.object.engine; + struct nvkm_device *device = engine->subdev.device; + switch (mthd) { + case 0x018c: chan->vblank.ctxdma = data; return true; + case 0x0400: chan->vblank.offset = data; return true; + case 0x0404: chan->vblank.value = data; return true; + case 0x0408: + if (data < device->disp->vblank.index_nr) { + nvkm_event_ntfy_allow(&chan->vblank.notify[data]); + return true; + } + break; + default: + break; + } + return false; +} + +void * +nv50_sw_chan_dtor(struct nvkm_sw_chan *base) +{ + struct nv50_sw_chan *chan = nv50_sw_chan(base); + int i; + + for (i = 0; i < ARRAY_SIZE(chan->vblank.notify); i++) + nvkm_event_ntfy_del(&chan->vblank.notify[i]); + + return chan; +} + +static const struct nvkm_sw_chan_func +nv50_sw_chan = { + .dtor = nv50_sw_chan_dtor, + .mthd = nv50_sw_chan_mthd, +}; + +static int +nv50_sw_chan_new(struct nvkm_sw *sw, struct nvkm_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nvkm_disp *disp = sw->engine.subdev.device->disp; + struct nv50_sw_chan *chan; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + *pobject = &chan->base.object; + + ret = nvkm_sw_chan_ctor(&nv50_sw_chan, sw, fifoch, oclass, &chan->base); + if (ret) + return ret; + + for (i = 0; disp && i < disp->vblank.index_nr; i++) { + nvkm_event_ntfy_add(&disp->vblank, i, NVKM_DISP_HEAD_EVENT_VBLANK, true, + nv50_sw_chan_vblsem_release, &chan->vblank.notify[i]); + } + + return 0; +} + +/******************************************************************************* + * software engine/subdev functions + ******************************************************************************/ + +static const struct nvkm_sw_func +nv50_sw = { + .chan_new = nv50_sw_chan_new, + .sclass = { + { nvkm_nvsw_new, { -1, -1, NVIF_CLASS_SW_NV50 } }, + {} + } +}; + +int +nv50_sw_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_sw **psw) +{ + return nvkm_sw_new_(&nv50_sw, device, type, inst, psw); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.h new file mode 100644 index 0000000000..b42289ce88 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nv50.h @@ -0,0 +1,21 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_SW_NV50_H__ +#define __NVKM_SW_NV50_H__ +#define nv50_sw_chan(p) container_of((p), struct nv50_sw_chan, base) +#include "priv.h" +#include "chan.h" +#include "nvsw.h" +#include + +struct nv50_sw_chan { + struct nvkm_sw_chan base; + struct { + struct nvkm_event_ntfy notify[4]; + u32 ctxdma; + u64 offset; + u32 value; + } vblank; +}; + +void *nv50_sw_chan_dtor(struct nvkm_sw_chan *); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.c b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.c new file mode 100644 index 0000000000..f5affa1c8f --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.c @@ -0,0 +1,86 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nvsw.h" +#include "chan.h" + +#include + +static int +nvkm_nvsw_uevent(struct nvkm_object *object, void *argv, u32 argc, struct nvkm_uevent *uevent) +{ + union nv04_nvsw_event_args *args = argv; + + if (!uevent) + return 0; + if (argc != sizeof(args->vn)) + return -ENOSYS; + + return nvkm_uevent_add(uevent, &nvkm_nvsw(object)->chan->event, 0, + NVKM_SW_CHAN_EVENT_PAGE_FLIP, NULL); +} + +static int +nvkm_nvsw_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) +{ + struct nvkm_nvsw *nvsw = nvkm_nvsw(object); + + if (nvsw->func->mthd) + return nvsw->func->mthd(nvsw, mthd, data, size); + + return -ENODEV; +} + +static const struct nvkm_object_func +nvkm_nvsw_ = { + .mthd = nvkm_nvsw_mthd, + .uevent = nvkm_nvsw_uevent, +}; + +int +nvkm_nvsw_new_(const struct nvkm_nvsw_func *func, struct nvkm_sw_chan *chan, + const struct nvkm_oclass *oclass, void *data, u32 size, + struct nvkm_object **pobject) +{ + struct nvkm_nvsw *nvsw; + + if (!(nvsw = kzalloc(sizeof(*nvsw), GFP_KERNEL))) + return -ENOMEM; + *pobject = &nvsw->object; + + nvkm_object_ctor(&nvkm_nvsw_, oclass, &nvsw->object); + nvsw->func = func; + nvsw->chan = chan; + return 0; +} + +static const struct nvkm_nvsw_func +nvkm_nvsw = { +}; + +int +nvkm_nvsw_new(struct nvkm_sw_chan *chan, const struct nvkm_oclass *oclass, + void *data, u32 size, struct nvkm_object **pobject) +{ + return nvkm_nvsw_new_(&nvkm_nvsw, chan, oclass, data, size, pobject); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.h b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.h new file mode 100644 index 0000000000..d2f846499b --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/nvsw.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_NVSW_H__ +#define __NVKM_NVSW_H__ +#define nvkm_nvsw(p) container_of((p), struct nvkm_nvsw, object) +#include + +struct nvkm_nvsw { + struct nvkm_object object; + const struct nvkm_nvsw_func *func; + struct nvkm_sw_chan *chan; +}; + +struct nvkm_nvsw_func { + int (*mthd)(struct nvkm_nvsw *, u32 mthd, void *data, u32 size); +}; + +int nvkm_nvsw_new_(const struct nvkm_nvsw_func *, struct nvkm_sw_chan *, + const struct nvkm_oclass *, void *data, u32 size, + struct nvkm_object **pobject); +int nvkm_nvsw_new(struct nvkm_sw_chan *, const struct nvkm_oclass *, + void *data, u32 size, struct nvkm_object **pobject); +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sw/priv.h b/drivers/gpu/drm/nouveau/nvkm/engine/sw/priv.h new file mode 100644 index 0000000000..8015afaba9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/sw/priv.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: MIT */ +#ifndef __NVKM_SW_PRIV_H__ +#define __NVKM_SW_PRIV_H__ +#define nvkm_sw(p) container_of((p), struct nvkm_sw, engine) +#include +struct nvkm_sw_chan; + +int nvkm_sw_new_(const struct nvkm_sw_func *, struct nvkm_device *, enum nvkm_subdev_type, int, + struct nvkm_sw **); + +struct nvkm_sw_chan_sclass { + int (*ctor)(struct nvkm_sw_chan *, const struct nvkm_oclass *, + void *data, u32 size, struct nvkm_object **); + struct nvkm_sclass base; +}; + +struct nvkm_sw_func { + int (*chan_new)(struct nvkm_sw *, struct nvkm_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + const struct nvkm_sw_chan_sclass sclass[]; +}; +#endif diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/vic/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/vic/Kbuild new file mode 100644 index 0000000000..70164f482c --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/vic/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +#nvkm-y += nvkm/engine/vic/base.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/vp/Kbuild b/drivers/gpu/drm/nouveau/nvkm/engine/vp/Kbuild new file mode 100644 index 0000000000..d48ea0fd1d --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/vp/Kbuild @@ -0,0 +1,2 @@ +# SPDX-License-Identifier: MIT +nvkm-y += nvkm/engine/vp/g84.o diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/vp/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/vp/g84.c new file mode 100644 index 0000000000..b502266c76 --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/vp/g84.c @@ -0,0 +1,43 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs, Ilia Mirkin + */ +#include + +#include + +static const struct nvkm_xtensa_func +g84_vp = { + .fifo_val = 0x111, + .unkd28 = 0x9c544, + .sclass = { + { -1, -1, NV74_VP2 }, + {} + } +}; + +int +g84_vp_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, + struct nvkm_engine **pengine) +{ + return nvkm_xtensa_new_(&g84_vp, device, type, inst, true, 0x00f000, pengine); +} diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/xtensa.c b/drivers/gpu/drm/nouveau/nvkm/engine/xtensa.c new file mode 100644 index 0000000000..f7d3ba0afb --- /dev/null +++ b/drivers/gpu/drm/nouveau/nvkm/engine/xtensa.c @@ -0,0 +1,191 @@ +/* + * Copyright 2013 Ilia Mirkin + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ +#include + +#include +#include + +static int +nvkm_xtensa_oclass_get(struct nvkm_oclass *oclass, int index) +{ + struct nvkm_xtensa *xtensa = nvkm_xtensa(oclass->engine); + int c = 0; + + while (xtensa->func->sclass[c].oclass) { + if (c++ == index) { + oclass->base = xtensa->func->sclass[index]; + return index; + } + } + + return c; +} + +static int +nvkm_xtensa_cclass_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + return nvkm_gpuobj_new(object->engine->subdev.device, 0x10000, align, + true, parent, pgpuobj); +} + +static const struct nvkm_object_func +nvkm_xtensa_cclass = { + .bind = nvkm_xtensa_cclass_bind, +}; + +static void +nvkm_xtensa_intr(struct nvkm_engine *engine) +{ + struct nvkm_xtensa *xtensa = nvkm_xtensa(engine); + struct nvkm_subdev *subdev = &xtensa->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 base = xtensa->addr; + u32 unk104 = nvkm_rd32(device, base + 0xd04); + u32 intr = nvkm_rd32(device, base + 0xc20); + u32 chan = nvkm_rd32(device, base + 0xc28); + u32 unk10c = nvkm_rd32(device, base + 0xd0c); + + if (intr & 0x10) + nvkm_warn(subdev, "Watchdog interrupt, engine hung.\n"); + nvkm_wr32(device, base + 0xc20, intr); + intr = nvkm_rd32(device, base + 0xc20); + if (unk104 == 0x10001 && unk10c == 0x200 && chan && !intr) { + nvkm_debug(subdev, "Enabling FIFO_CTRL\n"); + nvkm_mask(device, xtensa->addr + 0xd94, 0, xtensa->func->fifo_val); + } +} + +static int +nvkm_xtensa_fini(struct nvkm_engine *engine, bool suspend) +{ + struct nvkm_xtensa *xtensa = nvkm_xtensa(engine); + struct nvkm_device *device = xtensa->engine.subdev.device; + const u32 base = xtensa->addr; + + nvkm_wr32(device, base + 0xd84, 0); /* INTR_EN */ + nvkm_wr32(device, base + 0xd94, 0); /* FIFO_CTRL */ + + if (!suspend) + nvkm_memory_unref(&xtensa->gpu_fw); + return 0; +} + +static int +nvkm_xtensa_init(struct nvkm_engine *engine) +{ + struct nvkm_xtensa *xtensa = nvkm_xtensa(engine); + struct nvkm_subdev *subdev = &xtensa->engine.subdev; + struct nvkm_device *device = subdev->device; + const u32 base = xtensa->addr; + const struct firmware *fw; + char name[32]; + int i, ret; + u64 addr, size; + u32 tmp; + + if (!xtensa->gpu_fw) { + snprintf(name, sizeof(name), "nouveau/nv84_xuc%03x", + xtensa->addr >> 12); + + ret = request_firmware(&fw, name, device->dev); + if (ret) { + nvkm_warn(subdev, "unable to load firmware %s\n", name); + return ret; + } + + if (fw->size > 0x40000) { + nvkm_warn(subdev, "firmware %s too large\n", name); + release_firmware(fw); + return -EINVAL; + } + + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, + 0x40000, 0x1000, false, + &xtensa->gpu_fw); + if (ret) { + release_firmware(fw); + return ret; + } + + nvkm_kmap(xtensa->gpu_fw); + for (i = 0; i < fw->size / 4; i++) + nvkm_wo32(xtensa->gpu_fw, i * 4, *((u32 *)fw->data + i)); + nvkm_done(xtensa->gpu_fw); + release_firmware(fw); + } + + addr = nvkm_memory_addr(xtensa->gpu_fw); + size = nvkm_memory_size(xtensa->gpu_fw); + + nvkm_wr32(device, base + 0xd10, 0x1fffffff); /* ?? */ + nvkm_wr32(device, base + 0xd08, 0x0fffffff); /* ?? */ + + nvkm_wr32(device, base + 0xd28, xtensa->func->unkd28); /* ?? */ + nvkm_wr32(device, base + 0xc20, 0x3f); /* INTR */ + nvkm_wr32(device, base + 0xd84, 0x3f); /* INTR_EN */ + + nvkm_wr32(device, base + 0xcc0, addr >> 8); /* XT_REGION_BASE */ + nvkm_wr32(device, base + 0xcc4, 0x1c); /* XT_REGION_SETUP */ + nvkm_wr32(device, base + 0xcc8, size >> 8); /* XT_REGION_LIMIT */ + + tmp = nvkm_rd32(device, 0x0); + nvkm_wr32(device, base + 0xde0, tmp); /* SCRATCH_H2X */ + + nvkm_wr32(device, base + 0xce8, 0xf); /* XT_REGION_SETUP */ + + nvkm_wr32(device, base + 0xc20, 0x3f); /* INTR */ + nvkm_wr32(device, base + 0xd84, 0x3f); /* INTR_EN */ + return 0; +} + +static void * +nvkm_xtensa_dtor(struct nvkm_engine *engine) +{ + return nvkm_xtensa(engine); +} + +static const struct nvkm_engine_func +nvkm_xtensa = { + .dtor = nvkm_xtensa_dtor, + .init = nvkm_xtensa_init, + .fini = nvkm_xtensa_fini, + .intr = nvkm_xtensa_intr, + .fifo.sclass = nvkm_xtensa_oclass_get, + .cclass = &nvkm_xtensa_cclass, +}; + +int +nvkm_xtensa_new_(const struct nvkm_xtensa_func *func, struct nvkm_device *device, + enum nvkm_subdev_type type, int inst, bool enable, u32 addr, + struct nvkm_engine **pengine) +{ + struct nvkm_xtensa *xtensa; + + if (!(xtensa = kzalloc(sizeof(*xtensa), GFP_KERNEL))) + return -ENOMEM; + xtensa->func = func; + xtensa->addr = addr; + *pengine = &xtensa->engine; + + return nvkm_engine_ctor(&nvkm_xtensa, device, type, inst, enable, &xtensa->engine); +} -- cgit v1.2.3