summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/i915/gvt
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/i915/gvt')
-rw-r--r--drivers/gpu/drm/i915/gvt/Makefile25
-rw-r--r--drivers/gpu/drm/i915/gvt/aperture_gm.c366
-rw-r--r--drivers/gpu/drm/i915/gvt/cfg_space.c400
-rw-r--r--drivers/gpu/drm/i915/gvt/cmd_parser.c3263
-rw-r--r--drivers/gpu/drm/i915/gvt/cmd_parser.h58
-rw-r--r--drivers/gpu/drm/i915/gvt/debug.h65
-rw-r--r--drivers/gpu/drm/i915/gvt/debugfs.c213
-rw-r--r--drivers/gpu/drm/i915/gvt/display.c808
-rw-r--r--drivers/gpu/drm/i915/gvt/display.h224
-rw-r--r--drivers/gpu/drm/i915/gvt/dmabuf.c589
-rw-r--r--drivers/gpu/drm/i915/gvt/dmabuf.h67
-rw-r--r--drivers/gpu/drm/i915/gvt/edid.c585
-rw-r--r--drivers/gpu/drm/i915/gvt/edid.h154
-rw-r--r--drivers/gpu/drm/i915/gvt/execlist.c558
-rw-r--r--drivers/gpu/drm/i915/gvt/execlist.h185
-rw-r--r--drivers/gpu/drm/i915/gvt/fb_decoder.c508
-rw-r--r--drivers/gpu/drm/i915/gvt/fb_decoder.h173
-rw-r--r--drivers/gpu/drm/i915/gvt/firmware.c266
-rw-r--r--drivers/gpu/drm/i915/gvt/gtt.c2926
-rw-r--r--drivers/gpu/drm/i915/gvt/gtt.h296
-rw-r--r--drivers/gpu/drm/i915/gvt/gvt.h767
-rw-r--r--drivers/gpu/drm/i915/gvt/handlers.c3231
-rw-r--r--drivers/gpu/drm/i915/gvt/interrupt.c717
-rw-r--r--drivers/gpu/drm/i915/gvt/interrupt.h232
-rw-r--r--drivers/gpu/drm/i915/gvt/kvmgt.c2017
-rw-r--r--drivers/gpu/drm/i915/gvt/mmio.c332
-rw-r--r--drivers/gpu/drm/i915/gvt/mmio.h113
-rw-r--r--drivers/gpu/drm/i915/gvt/mmio_context.c606
-rw-r--r--drivers/gpu/drm/i915/gvt/mmio_context.h73
-rw-r--r--drivers/gpu/drm/i915/gvt/opregion.c483
-rw-r--r--drivers/gpu/drm/i915/gvt/page_track.c185
-rw-r--r--drivers/gpu/drm/i915/gvt/page_track.h59
-rw-r--r--drivers/gpu/drm/i915/gvt/reg.h144
-rw-r--r--drivers/gpu/drm/i915/gvt/sched_policy.c480
-rw-r--r--drivers/gpu/drm/i915/gvt/sched_policy.h65
-rw-r--r--drivers/gpu/drm/i915/gvt/scheduler.c1786
-rw-r--r--drivers/gpu/drm/i915/gvt/scheduler.h171
-rw-r--r--drivers/gpu/drm/i915/gvt/trace.h383
-rw-r--r--drivers/gpu/drm/i915/gvt/trace_points.c34
-rw-r--r--drivers/gpu/drm/i915/gvt/vgpu.c512
40 files changed, 24119 insertions, 0 deletions
diff --git a/drivers/gpu/drm/i915/gvt/Makefile b/drivers/gpu/drm/i915/gvt/Makefile
new file mode 100644
index 000000000..1699f6442
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/Makefile
@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: GPL-2.0
+
+kvmgt-$(CONFIG_DRM_I915_GVT) += \
+ gvt/aperture_gm.o \
+ gvt/cfg_space.o \
+ gvt/cmd_parser.o \
+ gvt/debugfs.o \
+ gvt/display.o \
+ gvt/dmabuf.o \
+ gvt/edid.o \
+ gvt/execlist.o \
+ gvt/fb_decoder.o \
+ gvt/firmware.o \
+ gvt/gtt.o \
+ gvt/handlers.o \
+ gvt/interrupt.o \
+ gvt/kvmgt.o \
+ gvt/mmio.o \
+ gvt/mmio_context.o \
+ gvt/opregion.o \
+ gvt/page_track.o \
+ gvt/sched_policy.o \
+ gvt/scheduler.o \
+ gvt/trace_points.o \
+ gvt/vgpu.o
diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c
new file mode 100644
index 000000000..076c779f7
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c
@@ -0,0 +1,366 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Dexuan Cui
+ *
+ * Contributors:
+ * Pei Zhang <pei.zhang@intel.com>
+ * Min He <min.he@intel.com>
+ * Niu Bing <bing.niu@intel.com>
+ * Yulei Zhang <yulei.zhang@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gt/intel_ggtt_fencing.h"
+#include "gvt.h"
+
+static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gt *gt = gvt->gt;
+ unsigned int flags;
+ u64 start, end, size;
+ struct drm_mm_node *node;
+ int ret;
+
+ if (high_gm) {
+ node = &vgpu->gm.high_gm_node;
+ size = vgpu_hidden_sz(vgpu);
+ start = ALIGN(gvt_hidden_gmadr_base(gvt), I915_GTT_PAGE_SIZE);
+ end = ALIGN(gvt_hidden_gmadr_end(gvt), I915_GTT_PAGE_SIZE);
+ flags = PIN_HIGH;
+ } else {
+ node = &vgpu->gm.low_gm_node;
+ size = vgpu_aperture_sz(vgpu);
+ start = ALIGN(gvt_aperture_gmadr_base(gvt), I915_GTT_PAGE_SIZE);
+ end = ALIGN(gvt_aperture_gmadr_end(gvt), I915_GTT_PAGE_SIZE);
+ flags = PIN_MAPPABLE;
+ }
+
+ mutex_lock(&gt->ggtt->vm.mutex);
+ mmio_hw_access_pre(gt);
+ ret = i915_gem_gtt_insert(&gt->ggtt->vm, NULL, node,
+ size, I915_GTT_PAGE_SIZE,
+ I915_COLOR_UNEVICTABLE,
+ start, end, flags);
+ mmio_hw_access_post(gt);
+ mutex_unlock(&gt->ggtt->vm.mutex);
+ if (ret)
+ gvt_err("fail to alloc %s gm space from host\n",
+ high_gm ? "high" : "low");
+
+ return ret;
+}
+
+static int alloc_vgpu_gm(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gt *gt = gvt->gt;
+ int ret;
+
+ ret = alloc_gm(vgpu, false);
+ if (ret)
+ return ret;
+
+ ret = alloc_gm(vgpu, true);
+ if (ret)
+ goto out_free_aperture;
+
+ gvt_dbg_core("vgpu%d: alloc low GM start %llx size %llx\n", vgpu->id,
+ vgpu_aperture_offset(vgpu), vgpu_aperture_sz(vgpu));
+
+ gvt_dbg_core("vgpu%d: alloc high GM start %llx size %llx\n", vgpu->id,
+ vgpu_hidden_offset(vgpu), vgpu_hidden_sz(vgpu));
+
+ return 0;
+out_free_aperture:
+ mutex_lock(&gt->ggtt->vm.mutex);
+ drm_mm_remove_node(&vgpu->gm.low_gm_node);
+ mutex_unlock(&gt->ggtt->vm.mutex);
+ return ret;
+}
+
+static void free_vgpu_gm(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gt *gt = gvt->gt;
+
+ mutex_lock(&gt->ggtt->vm.mutex);
+ drm_mm_remove_node(&vgpu->gm.low_gm_node);
+ drm_mm_remove_node(&vgpu->gm.high_gm_node);
+ mutex_unlock(&gt->ggtt->vm.mutex);
+}
+
+/**
+ * intel_vgpu_write_fence - write fence registers owned by a vGPU
+ * @vgpu: vGPU instance
+ * @fence: vGPU fence register number
+ * @value: Fence register value to be written
+ *
+ * This function is used to write fence registers owned by a vGPU. The vGPU
+ * fence register number will be translated into HW fence register number.
+ *
+ */
+void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
+ u32 fence, u64 value)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ struct intel_uncore *uncore = gvt->gt->uncore;
+ struct i915_fence_reg *reg;
+ i915_reg_t fence_reg_lo, fence_reg_hi;
+
+ assert_rpm_wakelock_held(uncore->rpm);
+
+ if (drm_WARN_ON(&i915->drm, fence >= vgpu_fence_sz(vgpu)))
+ return;
+
+ reg = vgpu->fence.regs[fence];
+ if (drm_WARN_ON(&i915->drm, !reg))
+ return;
+
+ fence_reg_lo = FENCE_REG_GEN6_LO(reg->id);
+ fence_reg_hi = FENCE_REG_GEN6_HI(reg->id);
+
+ intel_uncore_write(uncore, fence_reg_lo, 0);
+ intel_uncore_posting_read(uncore, fence_reg_lo);
+
+ intel_uncore_write(uncore, fence_reg_hi, upper_32_bits(value));
+ intel_uncore_write(uncore, fence_reg_lo, lower_32_bits(value));
+ intel_uncore_posting_read(uncore, fence_reg_lo);
+}
+
+static void _clear_vgpu_fence(struct intel_vgpu *vgpu)
+{
+ int i;
+
+ for (i = 0; i < vgpu_fence_sz(vgpu); i++)
+ intel_vgpu_write_fence(vgpu, i, 0);
+}
+
+static void free_vgpu_fence(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_uncore *uncore = gvt->gt->uncore;
+ struct i915_fence_reg *reg;
+ intel_wakeref_t wakeref;
+ u32 i;
+
+ if (drm_WARN_ON(&gvt->gt->i915->drm, !vgpu_fence_sz(vgpu)))
+ return;
+
+ wakeref = intel_runtime_pm_get(uncore->rpm);
+
+ mutex_lock(&gvt->gt->ggtt->vm.mutex);
+ _clear_vgpu_fence(vgpu);
+ for (i = 0; i < vgpu_fence_sz(vgpu); i++) {
+ reg = vgpu->fence.regs[i];
+ i915_unreserve_fence(reg);
+ vgpu->fence.regs[i] = NULL;
+ }
+ mutex_unlock(&gvt->gt->ggtt->vm.mutex);
+
+ intel_runtime_pm_put(uncore->rpm, wakeref);
+}
+
+static int alloc_vgpu_fence(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_uncore *uncore = gvt->gt->uncore;
+ struct i915_fence_reg *reg;
+ intel_wakeref_t wakeref;
+ int i;
+
+ wakeref = intel_runtime_pm_get(uncore->rpm);
+
+ /* Request fences from host */
+ mutex_lock(&gvt->gt->ggtt->vm.mutex);
+
+ for (i = 0; i < vgpu_fence_sz(vgpu); i++) {
+ reg = i915_reserve_fence(gvt->gt->ggtt);
+ if (IS_ERR(reg))
+ goto out_free_fence;
+
+ vgpu->fence.regs[i] = reg;
+ }
+
+ _clear_vgpu_fence(vgpu);
+
+ mutex_unlock(&gvt->gt->ggtt->vm.mutex);
+ intel_runtime_pm_put(uncore->rpm, wakeref);
+ return 0;
+
+out_free_fence:
+ gvt_vgpu_err("Failed to alloc fences\n");
+ /* Return fences to host, if fail */
+ for (i = 0; i < vgpu_fence_sz(vgpu); i++) {
+ reg = vgpu->fence.regs[i];
+ if (!reg)
+ continue;
+ i915_unreserve_fence(reg);
+ vgpu->fence.regs[i] = NULL;
+ }
+ mutex_unlock(&gvt->gt->ggtt->vm.mutex);
+ intel_runtime_pm_put_unchecked(uncore->rpm);
+ return -ENOSPC;
+}
+
+static void free_resource(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+
+ gvt->gm.vgpu_allocated_low_gm_size -= vgpu_aperture_sz(vgpu);
+ gvt->gm.vgpu_allocated_high_gm_size -= vgpu_hidden_sz(vgpu);
+ gvt->fence.vgpu_allocated_fence_num -= vgpu_fence_sz(vgpu);
+}
+
+static int alloc_resource(struct intel_vgpu *vgpu,
+ const struct intel_vgpu_config *conf)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ unsigned long request, avail, max, taken;
+ const char *item;
+
+ if (!conf->low_mm || !conf->high_mm || !conf->fence) {
+ gvt_vgpu_err("Invalid vGPU creation params\n");
+ return -EINVAL;
+ }
+
+ item = "low GM space";
+ max = gvt_aperture_sz(gvt) - HOST_LOW_GM_SIZE;
+ taken = gvt->gm.vgpu_allocated_low_gm_size;
+ avail = max - taken;
+ request = conf->low_mm;
+
+ if (request > avail)
+ goto no_enough_resource;
+
+ vgpu_aperture_sz(vgpu) = ALIGN(request, I915_GTT_PAGE_SIZE);
+
+ item = "high GM space";
+ max = gvt_hidden_sz(gvt) - HOST_HIGH_GM_SIZE;
+ taken = gvt->gm.vgpu_allocated_high_gm_size;
+ avail = max - taken;
+ request = conf->high_mm;
+
+ if (request > avail)
+ goto no_enough_resource;
+
+ vgpu_hidden_sz(vgpu) = ALIGN(request, I915_GTT_PAGE_SIZE);
+
+ item = "fence";
+ max = gvt_fence_sz(gvt) - HOST_FENCE;
+ taken = gvt->fence.vgpu_allocated_fence_num;
+ avail = max - taken;
+ request = conf->fence;
+
+ if (request > avail)
+ goto no_enough_resource;
+
+ vgpu_fence_sz(vgpu) = request;
+
+ gvt->gm.vgpu_allocated_low_gm_size += conf->low_mm;
+ gvt->gm.vgpu_allocated_high_gm_size += conf->high_mm;
+ gvt->fence.vgpu_allocated_fence_num += conf->fence;
+ return 0;
+
+no_enough_resource:
+ gvt_err("fail to allocate resource %s\n", item);
+ gvt_err("request %luMB avail %luMB max %luMB taken %luMB\n",
+ BYTES_TO_MB(request), BYTES_TO_MB(avail),
+ BYTES_TO_MB(max), BYTES_TO_MB(taken));
+ return -ENOSPC;
+}
+
+/**
+ * intel_vgpu_free_resource() - free HW resource owned by a vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is used to free the HW resource owned by a vGPU.
+ *
+ */
+void intel_vgpu_free_resource(struct intel_vgpu *vgpu)
+{
+ free_vgpu_gm(vgpu);
+ free_vgpu_fence(vgpu);
+ free_resource(vgpu);
+}
+
+/**
+ * intel_vgpu_reset_resource - reset resource state owned by a vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is used to reset resource state owned by a vGPU.
+ *
+ */
+void intel_vgpu_reset_resource(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ intel_wakeref_t wakeref;
+
+ with_intel_runtime_pm(gvt->gt->uncore->rpm, wakeref)
+ _clear_vgpu_fence(vgpu);
+}
+
+/**
+ * intel_vgpu_alloc_resource() - allocate HW resource for a vGPU
+ * @vgpu: vGPU
+ * @param: vGPU creation params
+ *
+ * This function is used to allocate HW resource for a vGPU. User specifies
+ * the resource configuration through the creation params.
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_alloc_resource(struct intel_vgpu *vgpu,
+ const struct intel_vgpu_config *conf)
+{
+ int ret;
+
+ ret = alloc_resource(vgpu, conf);
+ if (ret)
+ return ret;
+
+ ret = alloc_vgpu_gm(vgpu);
+ if (ret)
+ goto out_free_resource;
+
+ ret = alloc_vgpu_fence(vgpu);
+ if (ret)
+ goto out_free_vgpu_gm;
+
+ return 0;
+
+out_free_vgpu_gm:
+ free_vgpu_gm(vgpu);
+out_free_resource:
+ free_resource(vgpu);
+ return ret;
+}
diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c
new file mode 100644
index 000000000..eef3bba8a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/cfg_space.c
@@ -0,0 +1,400 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Eddie Dong <eddie.dong@intel.com>
+ * Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+#include "intel_pci_config.h"
+
+enum {
+ INTEL_GVT_PCI_BAR_GTTMMIO = 0,
+ INTEL_GVT_PCI_BAR_APERTURE,
+ INTEL_GVT_PCI_BAR_PIO,
+ INTEL_GVT_PCI_BAR_MAX,
+};
+
+/* bitmap for writable bits (RW or RW1C bits, but cannot co-exist in one
+ * byte) byte by byte in standard pci configuration space. (not the full
+ * 256 bytes.)
+ */
+static const u8 pci_cfg_space_rw_bmp[PCI_INTERRUPT_LINE + 4] = {
+ [PCI_COMMAND] = 0xff, 0x07,
+ [PCI_STATUS] = 0x00, 0xf9, /* the only one RW1C byte */
+ [PCI_CACHE_LINE_SIZE] = 0xff,
+ [PCI_BASE_ADDRESS_0 ... PCI_CARDBUS_CIS - 1] = 0xff,
+ [PCI_ROM_ADDRESS] = 0x01, 0xf8, 0xff, 0xff,
+ [PCI_INTERRUPT_LINE] = 0xff,
+};
+
+/**
+ * vgpu_pci_cfg_mem_write - write virtual cfg space memory
+ * @vgpu: target vgpu
+ * @off: offset
+ * @src: src ptr to write
+ * @bytes: number of bytes
+ *
+ * Use this function to write virtual cfg space memory.
+ * For standard cfg space, only RW bits can be changed,
+ * and we emulates the RW1C behavior of PCI_STATUS register.
+ */
+static void vgpu_pci_cfg_mem_write(struct intel_vgpu *vgpu, unsigned int off,
+ u8 *src, unsigned int bytes)
+{
+ u8 *cfg_base = vgpu_cfg_space(vgpu);
+ u8 mask, new, old;
+ pci_power_t pwr;
+ int i = 0;
+
+ for (; i < bytes && (off + i < sizeof(pci_cfg_space_rw_bmp)); i++) {
+ mask = pci_cfg_space_rw_bmp[off + i];
+ old = cfg_base[off + i];
+ new = src[i] & mask;
+
+ /**
+ * The PCI_STATUS high byte has RW1C bits, here
+ * emulates clear by writing 1 for these bits.
+ * Writing a 0b to RW1C bits has no effect.
+ */
+ if (off + i == PCI_STATUS + 1)
+ new = (~new & old) & mask;
+
+ cfg_base[off + i] = (old & ~mask) | new;
+ }
+
+ /* For other configuration space directly copy as it is. */
+ if (i < bytes)
+ memcpy(cfg_base + off + i, src + i, bytes - i);
+
+ if (off == vgpu->cfg_space.pmcsr_off && vgpu->cfg_space.pmcsr_off) {
+ pwr = (pci_power_t __force)(*(u16*)(&vgpu_cfg_space(vgpu)[off])
+ & PCI_PM_CTRL_STATE_MASK);
+ if (pwr == PCI_D3hot)
+ vgpu->d3_entered = true;
+ gvt_dbg_core("vgpu-%d power status changed to %d\n",
+ vgpu->id, pwr);
+ }
+}
+
+/**
+ * intel_vgpu_emulate_cfg_read - emulate vGPU configuration space read
+ * @vgpu: target vgpu
+ * @offset: offset
+ * @p_data: return data ptr
+ * @bytes: number of bytes to read
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ if (drm_WARN_ON(&i915->drm, bytes > 4))
+ return -EINVAL;
+
+ if (drm_WARN_ON(&i915->drm,
+ offset + bytes > vgpu->gvt->device_info.cfg_space_size))
+ return -EINVAL;
+
+ memcpy(p_data, vgpu_cfg_space(vgpu) + offset, bytes);
+ return 0;
+}
+
+static void map_aperture(struct intel_vgpu *vgpu, bool map)
+{
+ if (map != vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked)
+ vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked = map;
+}
+
+static void trap_gttmmio(struct intel_vgpu *vgpu, bool trap)
+{
+ if (trap != vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].tracked)
+ vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].tracked = trap;
+}
+
+static int emulate_pci_command_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u8 old = vgpu_cfg_space(vgpu)[offset];
+ u8 new = *(u8 *)p_data;
+ u8 changed = old ^ new;
+
+ vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes);
+ if (!(changed & PCI_COMMAND_MEMORY))
+ return 0;
+
+ if (old & PCI_COMMAND_MEMORY) {
+ trap_gttmmio(vgpu, false);
+ map_aperture(vgpu, false);
+ } else {
+ trap_gttmmio(vgpu, true);
+ map_aperture(vgpu, true);
+ }
+
+ return 0;
+}
+
+static int emulate_pci_rom_bar_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 *pval = (u32 *)(vgpu_cfg_space(vgpu) + offset);
+ u32 new = *(u32 *)(p_data);
+
+ if ((new & PCI_ROM_ADDRESS_MASK) == PCI_ROM_ADDRESS_MASK)
+ /* We don't have rom, return size of 0. */
+ *pval = 0;
+ else
+ vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+static void emulate_pci_bar_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 new = *(u32 *)(p_data);
+ bool lo = IS_ALIGNED(offset, 8);
+ u64 size;
+ bool mmio_enabled =
+ vgpu_cfg_space(vgpu)[PCI_COMMAND] & PCI_COMMAND_MEMORY;
+ struct intel_vgpu_pci_bar *bars = vgpu->cfg_space.bar;
+
+ /*
+ * Power-up software can determine how much address
+ * space the device requires by writing a value of
+ * all 1's to the register and then reading the value
+ * back. The device will return 0's in all don't-care
+ * address bits.
+ */
+ if (new == 0xffffffff) {
+ switch (offset) {
+ case PCI_BASE_ADDRESS_0:
+ case PCI_BASE_ADDRESS_1:
+ size = ~(bars[INTEL_GVT_PCI_BAR_GTTMMIO].size -1);
+ intel_vgpu_write_pci_bar(vgpu, offset,
+ size >> (lo ? 0 : 32), lo);
+ /*
+ * Untrap the BAR, since guest hasn't configured a
+ * valid GPA
+ */
+ trap_gttmmio(vgpu, false);
+ break;
+ case PCI_BASE_ADDRESS_2:
+ case PCI_BASE_ADDRESS_3:
+ size = ~(bars[INTEL_GVT_PCI_BAR_APERTURE].size -1);
+ intel_vgpu_write_pci_bar(vgpu, offset,
+ size >> (lo ? 0 : 32), lo);
+ map_aperture(vgpu, false);
+ break;
+ default:
+ /* Unimplemented BARs */
+ intel_vgpu_write_pci_bar(vgpu, offset, 0x0, false);
+ }
+ } else {
+ switch (offset) {
+ case PCI_BASE_ADDRESS_0:
+ case PCI_BASE_ADDRESS_1:
+ /*
+ * Untrap the old BAR first, since guest has
+ * re-configured the BAR
+ */
+ trap_gttmmio(vgpu, false);
+ intel_vgpu_write_pci_bar(vgpu, offset, new, lo);
+ trap_gttmmio(vgpu, mmio_enabled);
+ break;
+ case PCI_BASE_ADDRESS_2:
+ case PCI_BASE_ADDRESS_3:
+ map_aperture(vgpu, false);
+ intel_vgpu_write_pci_bar(vgpu, offset, new, lo);
+ map_aperture(vgpu, mmio_enabled);
+ break;
+ default:
+ intel_vgpu_write_pci_bar(vgpu, offset, new, lo);
+ }
+ }
+}
+
+/**
+ * intel_vgpu_emulate_cfg_read - emulate vGPU configuration space write
+ * @vgpu: target vgpu
+ * @offset: offset
+ * @p_data: write data ptr
+ * @bytes: number of bytes to write
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ int ret;
+
+ if (drm_WARN_ON(&i915->drm, bytes > 4))
+ return -EINVAL;
+
+ if (drm_WARN_ON(&i915->drm,
+ offset + bytes > vgpu->gvt->device_info.cfg_space_size))
+ return -EINVAL;
+
+ /* First check if it's PCI_COMMAND */
+ if (IS_ALIGNED(offset, 2) && offset == PCI_COMMAND) {
+ if (drm_WARN_ON(&i915->drm, bytes > 2))
+ return -EINVAL;
+ return emulate_pci_command_write(vgpu, offset, p_data, bytes);
+ }
+
+ switch (rounddown(offset, 4)) {
+ case PCI_ROM_ADDRESS:
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+ return emulate_pci_rom_bar_write(vgpu, offset, p_data, bytes);
+
+ case PCI_BASE_ADDRESS_0 ... PCI_BASE_ADDRESS_5:
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+ emulate_pci_bar_write(vgpu, offset, p_data, bytes);
+ break;
+ case INTEL_GVT_PCI_SWSCI:
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+ ret = intel_vgpu_emulate_opregion_request(vgpu, *(u32 *)p_data);
+ if (ret)
+ return ret;
+ break;
+
+ case INTEL_GVT_PCI_OPREGION:
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+ ret = intel_vgpu_opregion_base_write_handler(vgpu,
+ *(u32 *)p_data);
+ if (ret)
+ return ret;
+
+ vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes);
+ break;
+ default:
+ vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes);
+ break;
+ }
+ return 0;
+}
+
+/**
+ * intel_vgpu_init_cfg_space - init vGPU configuration space when create vGPU
+ *
+ * @vgpu: a vGPU
+ * @primary: is the vGPU presented as primary
+ *
+ */
+void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu,
+ bool primary)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct pci_dev *pdev = to_pci_dev(gvt->gt->i915->drm.dev);
+ const struct intel_gvt_device_info *info = &gvt->device_info;
+ u16 *gmch_ctl;
+ u8 next;
+
+ memcpy(vgpu_cfg_space(vgpu), gvt->firmware.cfg_space,
+ info->cfg_space_size);
+
+ if (!primary) {
+ vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] =
+ INTEL_GVT_PCI_CLASS_VGA_OTHER;
+ vgpu_cfg_space(vgpu)[PCI_CLASS_PROG] =
+ INTEL_GVT_PCI_CLASS_VGA_OTHER;
+ }
+
+ /* Show guest that there isn't any stolen memory.*/
+ gmch_ctl = (u16 *)(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_GMCH_CONTROL);
+ *gmch_ctl &= ~(BDW_GMCH_GMS_MASK << BDW_GMCH_GMS_SHIFT);
+
+ intel_vgpu_write_pci_bar(vgpu, PCI_BASE_ADDRESS_2,
+ gvt_aperture_pa_base(gvt), true);
+
+ vgpu_cfg_space(vgpu)[PCI_COMMAND] &= ~(PCI_COMMAND_IO
+ | PCI_COMMAND_MEMORY
+ | PCI_COMMAND_MASTER);
+ /*
+ * Clear the bar upper 32bit and let guest to assign the new value
+ */
+ memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_1, 0, 4);
+ memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_3, 0, 4);
+ memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_4, 0, 8);
+ memset(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_OPREGION, 0, 4);
+
+ vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].size =
+ pci_resource_len(pdev, GTTMMADR_BAR);
+ vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].size =
+ pci_resource_len(pdev, GTT_APERTURE_BAR);
+
+ memset(vgpu_cfg_space(vgpu) + PCI_ROM_ADDRESS, 0, 4);
+
+ /* PM Support */
+ vgpu->cfg_space.pmcsr_off = 0;
+ if (vgpu_cfg_space(vgpu)[PCI_STATUS] & PCI_STATUS_CAP_LIST) {
+ next = vgpu_cfg_space(vgpu)[PCI_CAPABILITY_LIST];
+ do {
+ if (vgpu_cfg_space(vgpu)[next + PCI_CAP_LIST_ID] == PCI_CAP_ID_PM) {
+ vgpu->cfg_space.pmcsr_off = next + PCI_PM_CTRL;
+ break;
+ }
+ next = vgpu_cfg_space(vgpu)[next + PCI_CAP_LIST_NEXT];
+ } while (next);
+ }
+}
+
+/**
+ * intel_vgpu_reset_cfg_space - reset vGPU configuration space
+ *
+ * @vgpu: a vGPU
+ *
+ */
+void intel_vgpu_reset_cfg_space(struct intel_vgpu *vgpu)
+{
+ u8 cmd = vgpu_cfg_space(vgpu)[PCI_COMMAND];
+ bool primary = vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] !=
+ INTEL_GVT_PCI_CLASS_VGA_OTHER;
+
+ if (cmd & PCI_COMMAND_MEMORY) {
+ trap_gttmmio(vgpu, false);
+ map_aperture(vgpu, false);
+ }
+
+ /**
+ * Currently we only do such reset when vGPU is not
+ * owned by any VM, so we simply restore entire cfg
+ * space to default value.
+ */
+ intel_vgpu_init_cfg_space(vgpu, primary);
+}
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
new file mode 100644
index 000000000..de13f102d
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -0,0 +1,3263 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Kevin Tian <kevin.tian@intel.com>
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Yulei Zhang <yulei.zhang@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include <linux/slab.h>
+
+#include "i915_drv.h"
+#include "gt/intel_engine_regs.h"
+#include "gt/intel_gpu_commands.h"
+#include "gt/intel_gt_regs.h"
+#include "gt/intel_lrc.h"
+#include "gt/intel_ring.h"
+#include "gt/intel_gt_requests.h"
+#include "gt/shmem_utils.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+#include "trace.h"
+
+#include "gem/i915_gem_context.h"
+#include "gem/i915_gem_pm.h"
+#include "gt/intel_context.h"
+
+#define INVALID_OP (~0U)
+
+#define OP_LEN_MI 9
+#define OP_LEN_2D 10
+#define OP_LEN_3D_MEDIA 16
+#define OP_LEN_MFX_VC 16
+#define OP_LEN_VEBOX 16
+
+#define CMD_TYPE(cmd) (((cmd) >> 29) & 7)
+
+struct sub_op_bits {
+ int hi;
+ int low;
+};
+struct decode_info {
+ const char *name;
+ int op_len;
+ int nr_sub_op;
+ const struct sub_op_bits *sub_op;
+};
+
+#define MAX_CMD_BUDGET 0x7fffffff
+#define MI_WAIT_FOR_PLANE_C_FLIP_PENDING (1<<15)
+#define MI_WAIT_FOR_PLANE_B_FLIP_PENDING (1<<9)
+#define MI_WAIT_FOR_PLANE_A_FLIP_PENDING (1<<1)
+
+#define MI_WAIT_FOR_SPRITE_C_FLIP_PENDING (1<<20)
+#define MI_WAIT_FOR_SPRITE_B_FLIP_PENDING (1<<10)
+#define MI_WAIT_FOR_SPRITE_A_FLIP_PENDING (1<<2)
+
+/* Render Command Map */
+
+/* MI_* command Opcode (28:23) */
+#define OP_MI_NOOP 0x0
+#define OP_MI_SET_PREDICATE 0x1 /* HSW+ */
+#define OP_MI_USER_INTERRUPT 0x2
+#define OP_MI_WAIT_FOR_EVENT 0x3
+#define OP_MI_FLUSH 0x4
+#define OP_MI_ARB_CHECK 0x5
+#define OP_MI_RS_CONTROL 0x6 /* HSW+ */
+#define OP_MI_REPORT_HEAD 0x7
+#define OP_MI_ARB_ON_OFF 0x8
+#define OP_MI_URB_ATOMIC_ALLOC 0x9 /* HSW+ */
+#define OP_MI_BATCH_BUFFER_END 0xA
+#define OP_MI_SUSPEND_FLUSH 0xB
+#define OP_MI_PREDICATE 0xC /* IVB+ */
+#define OP_MI_TOPOLOGY_FILTER 0xD /* IVB+ */
+#define OP_MI_SET_APPID 0xE /* IVB+ */
+#define OP_MI_RS_CONTEXT 0xF /* HSW+ */
+#define OP_MI_LOAD_SCAN_LINES_INCL 0x12 /* HSW+ */
+#define OP_MI_DISPLAY_FLIP 0x14
+#define OP_MI_SEMAPHORE_MBOX 0x16
+#define OP_MI_SET_CONTEXT 0x18
+#define OP_MI_MATH 0x1A
+#define OP_MI_URB_CLEAR 0x19
+#define OP_MI_SEMAPHORE_SIGNAL 0x1B /* BDW+ */
+#define OP_MI_SEMAPHORE_WAIT 0x1C /* BDW+ */
+
+#define OP_MI_STORE_DATA_IMM 0x20
+#define OP_MI_STORE_DATA_INDEX 0x21
+#define OP_MI_LOAD_REGISTER_IMM 0x22
+#define OP_MI_UPDATE_GTT 0x23
+#define OP_MI_STORE_REGISTER_MEM 0x24
+#define OP_MI_FLUSH_DW 0x26
+#define OP_MI_CLFLUSH 0x27
+#define OP_MI_REPORT_PERF_COUNT 0x28
+#define OP_MI_LOAD_REGISTER_MEM 0x29 /* HSW+ */
+#define OP_MI_LOAD_REGISTER_REG 0x2A /* HSW+ */
+#define OP_MI_RS_STORE_DATA_IMM 0x2B /* HSW+ */
+#define OP_MI_LOAD_URB_MEM 0x2C /* HSW+ */
+#define OP_MI_STORE_URM_MEM 0x2D /* HSW+ */
+#define OP_MI_2E 0x2E /* BDW+ */
+#define OP_MI_2F 0x2F /* BDW+ */
+#define OP_MI_BATCH_BUFFER_START 0x31
+
+/* Bit definition for dword 0 */
+#define _CMDBIT_BB_START_IN_PPGTT (1UL << 8)
+
+#define OP_MI_CONDITIONAL_BATCH_BUFFER_END 0x36
+
+#define BATCH_BUFFER_ADDR_MASK ((1UL << 32) - (1U << 2))
+#define BATCH_BUFFER_ADDR_HIGH_MASK ((1UL << 16) - (1U))
+#define BATCH_BUFFER_ADR_SPACE_BIT(x) (((x) >> 8) & 1U)
+#define BATCH_BUFFER_2ND_LEVEL_BIT(x) ((x) >> 22 & 1U)
+
+/* 2D command: Opcode (28:22) */
+#define OP_2D(x) ((2<<7) | x)
+
+#define OP_XY_SETUP_BLT OP_2D(0x1)
+#define OP_XY_SETUP_CLIP_BLT OP_2D(0x3)
+#define OP_XY_SETUP_MONO_PATTERN_SL_BLT OP_2D(0x11)
+#define OP_XY_PIXEL_BLT OP_2D(0x24)
+#define OP_XY_SCANLINES_BLT OP_2D(0x25)
+#define OP_XY_TEXT_BLT OP_2D(0x26)
+#define OP_XY_TEXT_IMMEDIATE_BLT OP_2D(0x31)
+#define OP_XY_COLOR_BLT OP_2D(0x50)
+#define OP_XY_PAT_BLT OP_2D(0x51)
+#define OP_XY_MONO_PAT_BLT OP_2D(0x52)
+#define OP_XY_SRC_COPY_BLT OP_2D(0x53)
+#define OP_XY_MONO_SRC_COPY_BLT OP_2D(0x54)
+#define OP_XY_FULL_BLT OP_2D(0x55)
+#define OP_XY_FULL_MONO_SRC_BLT OP_2D(0x56)
+#define OP_XY_FULL_MONO_PATTERN_BLT OP_2D(0x57)
+#define OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT OP_2D(0x58)
+#define OP_XY_MONO_PAT_FIXED_BLT OP_2D(0x59)
+#define OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT OP_2D(0x71)
+#define OP_XY_PAT_BLT_IMMEDIATE OP_2D(0x72)
+#define OP_XY_SRC_COPY_CHROMA_BLT OP_2D(0x73)
+#define OP_XY_FULL_IMMEDIATE_PATTERN_BLT OP_2D(0x74)
+#define OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT OP_2D(0x75)
+#define OP_XY_PAT_CHROMA_BLT OP_2D(0x76)
+#define OP_XY_PAT_CHROMA_BLT_IMMEDIATE OP_2D(0x77)
+
+/* 3D/Media Command: Pipeline Type(28:27) Opcode(26:24) Sub Opcode(23:16) */
+#define OP_3D_MEDIA(sub_type, opcode, sub_opcode) \
+ ((3 << 13) | ((sub_type) << 11) | ((opcode) << 8) | (sub_opcode))
+
+#define OP_STATE_PREFETCH OP_3D_MEDIA(0x0, 0x0, 0x03)
+
+#define OP_STATE_BASE_ADDRESS OP_3D_MEDIA(0x0, 0x1, 0x01)
+#define OP_STATE_SIP OP_3D_MEDIA(0x0, 0x1, 0x02)
+#define OP_3D_MEDIA_0_1_4 OP_3D_MEDIA(0x0, 0x1, 0x04)
+#define OP_SWTESS_BASE_ADDRESS OP_3D_MEDIA(0x0, 0x1, 0x03)
+
+#define OP_3DSTATE_VF_STATISTICS_GM45 OP_3D_MEDIA(0x1, 0x0, 0x0B)
+
+#define OP_PIPELINE_SELECT OP_3D_MEDIA(0x1, 0x1, 0x04)
+
+#define OP_MEDIA_VFE_STATE OP_3D_MEDIA(0x2, 0x0, 0x0)
+#define OP_MEDIA_CURBE_LOAD OP_3D_MEDIA(0x2, 0x0, 0x1)
+#define OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD OP_3D_MEDIA(0x2, 0x0, 0x2)
+#define OP_MEDIA_GATEWAY_STATE OP_3D_MEDIA(0x2, 0x0, 0x3)
+#define OP_MEDIA_STATE_FLUSH OP_3D_MEDIA(0x2, 0x0, 0x4)
+#define OP_MEDIA_POOL_STATE OP_3D_MEDIA(0x2, 0x0, 0x5)
+
+#define OP_MEDIA_OBJECT OP_3D_MEDIA(0x2, 0x1, 0x0)
+#define OP_MEDIA_OBJECT_PRT OP_3D_MEDIA(0x2, 0x1, 0x2)
+#define OP_MEDIA_OBJECT_WALKER OP_3D_MEDIA(0x2, 0x1, 0x3)
+#define OP_GPGPU_WALKER OP_3D_MEDIA(0x2, 0x1, 0x5)
+
+#define OP_3DSTATE_CLEAR_PARAMS OP_3D_MEDIA(0x3, 0x0, 0x04) /* IVB+ */
+#define OP_3DSTATE_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x05) /* IVB+ */
+#define OP_3DSTATE_STENCIL_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x06) /* IVB+ */
+#define OP_3DSTATE_HIER_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x07) /* IVB+ */
+#define OP_3DSTATE_VERTEX_BUFFERS OP_3D_MEDIA(0x3, 0x0, 0x08)
+#define OP_3DSTATE_VERTEX_ELEMENTS OP_3D_MEDIA(0x3, 0x0, 0x09)
+#define OP_3DSTATE_INDEX_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x0A)
+#define OP_3DSTATE_VF_STATISTICS OP_3D_MEDIA(0x3, 0x0, 0x0B)
+#define OP_3DSTATE_VF OP_3D_MEDIA(0x3, 0x0, 0x0C) /* HSW+ */
+#define OP_3DSTATE_CC_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x0E)
+#define OP_3DSTATE_SCISSOR_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x0F)
+#define OP_3DSTATE_VS OP_3D_MEDIA(0x3, 0x0, 0x10)
+#define OP_3DSTATE_GS OP_3D_MEDIA(0x3, 0x0, 0x11)
+#define OP_3DSTATE_CLIP OP_3D_MEDIA(0x3, 0x0, 0x12)
+#define OP_3DSTATE_SF OP_3D_MEDIA(0x3, 0x0, 0x13)
+#define OP_3DSTATE_WM OP_3D_MEDIA(0x3, 0x0, 0x14)
+#define OP_3DSTATE_CONSTANT_VS OP_3D_MEDIA(0x3, 0x0, 0x15)
+#define OP_3DSTATE_CONSTANT_GS OP_3D_MEDIA(0x3, 0x0, 0x16)
+#define OP_3DSTATE_CONSTANT_PS OP_3D_MEDIA(0x3, 0x0, 0x17)
+#define OP_3DSTATE_SAMPLE_MASK OP_3D_MEDIA(0x3, 0x0, 0x18)
+#define OP_3DSTATE_CONSTANT_HS OP_3D_MEDIA(0x3, 0x0, 0x19) /* IVB+ */
+#define OP_3DSTATE_CONSTANT_DS OP_3D_MEDIA(0x3, 0x0, 0x1A) /* IVB+ */
+#define OP_3DSTATE_HS OP_3D_MEDIA(0x3, 0x0, 0x1B) /* IVB+ */
+#define OP_3DSTATE_TE OP_3D_MEDIA(0x3, 0x0, 0x1C) /* IVB+ */
+#define OP_3DSTATE_DS OP_3D_MEDIA(0x3, 0x0, 0x1D) /* IVB+ */
+#define OP_3DSTATE_STREAMOUT OP_3D_MEDIA(0x3, 0x0, 0x1E) /* IVB+ */
+#define OP_3DSTATE_SBE OP_3D_MEDIA(0x3, 0x0, 0x1F) /* IVB+ */
+#define OP_3DSTATE_PS OP_3D_MEDIA(0x3, 0x0, 0x20) /* IVB+ */
+#define OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP OP_3D_MEDIA(0x3, 0x0, 0x21) /* IVB+ */
+#define OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC OP_3D_MEDIA(0x3, 0x0, 0x23) /* IVB+ */
+#define OP_3DSTATE_BLEND_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x24) /* IVB+ */
+#define OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x25) /* IVB+ */
+#define OP_3DSTATE_BINDING_TABLE_POINTERS_VS OP_3D_MEDIA(0x3, 0x0, 0x26) /* IVB+ */
+#define OP_3DSTATE_BINDING_TABLE_POINTERS_HS OP_3D_MEDIA(0x3, 0x0, 0x27) /* IVB+ */
+#define OP_3DSTATE_BINDING_TABLE_POINTERS_DS OP_3D_MEDIA(0x3, 0x0, 0x28) /* IVB+ */
+#define OP_3DSTATE_BINDING_TABLE_POINTERS_GS OP_3D_MEDIA(0x3, 0x0, 0x29) /* IVB+ */
+#define OP_3DSTATE_BINDING_TABLE_POINTERS_PS OP_3D_MEDIA(0x3, 0x0, 0x2A) /* IVB+ */
+#define OP_3DSTATE_SAMPLER_STATE_POINTERS_VS OP_3D_MEDIA(0x3, 0x0, 0x2B) /* IVB+ */
+#define OP_3DSTATE_SAMPLER_STATE_POINTERS_HS OP_3D_MEDIA(0x3, 0x0, 0x2C) /* IVB+ */
+#define OP_3DSTATE_SAMPLER_STATE_POINTERS_DS OP_3D_MEDIA(0x3, 0x0, 0x2D) /* IVB+ */
+#define OP_3DSTATE_SAMPLER_STATE_POINTERS_GS OP_3D_MEDIA(0x3, 0x0, 0x2E) /* IVB+ */
+#define OP_3DSTATE_SAMPLER_STATE_POINTERS_PS OP_3D_MEDIA(0x3, 0x0, 0x2F) /* IVB+ */
+#define OP_3DSTATE_URB_VS OP_3D_MEDIA(0x3, 0x0, 0x30) /* IVB+ */
+#define OP_3DSTATE_URB_HS OP_3D_MEDIA(0x3, 0x0, 0x31) /* IVB+ */
+#define OP_3DSTATE_URB_DS OP_3D_MEDIA(0x3, 0x0, 0x32) /* IVB+ */
+#define OP_3DSTATE_URB_GS OP_3D_MEDIA(0x3, 0x0, 0x33) /* IVB+ */
+#define OP_3DSTATE_GATHER_CONSTANT_VS OP_3D_MEDIA(0x3, 0x0, 0x34) /* HSW+ */
+#define OP_3DSTATE_GATHER_CONSTANT_GS OP_3D_MEDIA(0x3, 0x0, 0x35) /* HSW+ */
+#define OP_3DSTATE_GATHER_CONSTANT_HS OP_3D_MEDIA(0x3, 0x0, 0x36) /* HSW+ */
+#define OP_3DSTATE_GATHER_CONSTANT_DS OP_3D_MEDIA(0x3, 0x0, 0x37) /* HSW+ */
+#define OP_3DSTATE_GATHER_CONSTANT_PS OP_3D_MEDIA(0x3, 0x0, 0x38) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTF_VS OP_3D_MEDIA(0x3, 0x0, 0x39) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTF_PS OP_3D_MEDIA(0x3, 0x0, 0x3A) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTI_VS OP_3D_MEDIA(0x3, 0x0, 0x3B) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTI_PS OP_3D_MEDIA(0x3, 0x0, 0x3C) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTB_VS OP_3D_MEDIA(0x3, 0x0, 0x3D) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANTB_PS OP_3D_MEDIA(0x3, 0x0, 0x3E) /* HSW+ */
+#define OP_3DSTATE_DX9_LOCAL_VALID_VS OP_3D_MEDIA(0x3, 0x0, 0x3F) /* HSW+ */
+#define OP_3DSTATE_DX9_LOCAL_VALID_PS OP_3D_MEDIA(0x3, 0x0, 0x40) /* HSW+ */
+#define OP_3DSTATE_DX9_GENERATE_ACTIVE_VS OP_3D_MEDIA(0x3, 0x0, 0x41) /* HSW+ */
+#define OP_3DSTATE_DX9_GENERATE_ACTIVE_PS OP_3D_MEDIA(0x3, 0x0, 0x42) /* HSW+ */
+#define OP_3DSTATE_BINDING_TABLE_EDIT_VS OP_3D_MEDIA(0x3, 0x0, 0x43) /* HSW+ */
+#define OP_3DSTATE_BINDING_TABLE_EDIT_GS OP_3D_MEDIA(0x3, 0x0, 0x44) /* HSW+ */
+#define OP_3DSTATE_BINDING_TABLE_EDIT_HS OP_3D_MEDIA(0x3, 0x0, 0x45) /* HSW+ */
+#define OP_3DSTATE_BINDING_TABLE_EDIT_DS OP_3D_MEDIA(0x3, 0x0, 0x46) /* HSW+ */
+#define OP_3DSTATE_BINDING_TABLE_EDIT_PS OP_3D_MEDIA(0x3, 0x0, 0x47) /* HSW+ */
+
+#define OP_3DSTATE_VF_INSTANCING OP_3D_MEDIA(0x3, 0x0, 0x49) /* BDW+ */
+#define OP_3DSTATE_VF_SGVS OP_3D_MEDIA(0x3, 0x0, 0x4A) /* BDW+ */
+#define OP_3DSTATE_VF_TOPOLOGY OP_3D_MEDIA(0x3, 0x0, 0x4B) /* BDW+ */
+#define OP_3DSTATE_WM_CHROMAKEY OP_3D_MEDIA(0x3, 0x0, 0x4C) /* BDW+ */
+#define OP_3DSTATE_PS_BLEND OP_3D_MEDIA(0x3, 0x0, 0x4D) /* BDW+ */
+#define OP_3DSTATE_WM_DEPTH_STENCIL OP_3D_MEDIA(0x3, 0x0, 0x4E) /* BDW+ */
+#define OP_3DSTATE_PS_EXTRA OP_3D_MEDIA(0x3, 0x0, 0x4F) /* BDW+ */
+#define OP_3DSTATE_RASTER OP_3D_MEDIA(0x3, 0x0, 0x50) /* BDW+ */
+#define OP_3DSTATE_SBE_SWIZ OP_3D_MEDIA(0x3, 0x0, 0x51) /* BDW+ */
+#define OP_3DSTATE_WM_HZ_OP OP_3D_MEDIA(0x3, 0x0, 0x52) /* BDW+ */
+#define OP_3DSTATE_COMPONENT_PACKING OP_3D_MEDIA(0x3, 0x0, 0x55) /* SKL+ */
+
+#define OP_3DSTATE_DRAWING_RECTANGLE OP_3D_MEDIA(0x3, 0x1, 0x00)
+#define OP_3DSTATE_SAMPLER_PALETTE_LOAD0 OP_3D_MEDIA(0x3, 0x1, 0x02)
+#define OP_3DSTATE_CHROMA_KEY OP_3D_MEDIA(0x3, 0x1, 0x04)
+#define OP_SNB_3DSTATE_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x05)
+#define OP_3DSTATE_POLY_STIPPLE_OFFSET OP_3D_MEDIA(0x3, 0x1, 0x06)
+#define OP_3DSTATE_POLY_STIPPLE_PATTERN OP_3D_MEDIA(0x3, 0x1, 0x07)
+#define OP_3DSTATE_LINE_STIPPLE OP_3D_MEDIA(0x3, 0x1, 0x08)
+#define OP_3DSTATE_AA_LINE_PARAMS OP_3D_MEDIA(0x3, 0x1, 0x0A)
+#define OP_3DSTATE_GS_SVB_INDEX OP_3D_MEDIA(0x3, 0x1, 0x0B)
+#define OP_3DSTATE_SAMPLER_PALETTE_LOAD1 OP_3D_MEDIA(0x3, 0x1, 0x0C)
+#define OP_3DSTATE_MULTISAMPLE_BDW OP_3D_MEDIA(0x3, 0x0, 0x0D)
+#define OP_SNB_3DSTATE_STENCIL_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x0E)
+#define OP_SNB_3DSTATE_HIER_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x0F)
+#define OP_SNB_3DSTATE_CLEAR_PARAMS OP_3D_MEDIA(0x3, 0x1, 0x10)
+#define OP_3DSTATE_MONOFILTER_SIZE OP_3D_MEDIA(0x3, 0x1, 0x11)
+#define OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS OP_3D_MEDIA(0x3, 0x1, 0x12) /* IVB+ */
+#define OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS OP_3D_MEDIA(0x3, 0x1, 0x13) /* IVB+ */
+#define OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS OP_3D_MEDIA(0x3, 0x1, 0x14) /* IVB+ */
+#define OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS OP_3D_MEDIA(0x3, 0x1, 0x15) /* IVB+ */
+#define OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS OP_3D_MEDIA(0x3, 0x1, 0x16) /* IVB+ */
+#define OP_3DSTATE_SO_DECL_LIST OP_3D_MEDIA(0x3, 0x1, 0x17)
+#define OP_3DSTATE_SO_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x18)
+#define OP_3DSTATE_BINDING_TABLE_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x19) /* HSW+ */
+#define OP_3DSTATE_GATHER_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x1A) /* HSW+ */
+#define OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x1B) /* HSW+ */
+#define OP_3DSTATE_SAMPLE_PATTERN OP_3D_MEDIA(0x3, 0x1, 0x1C)
+#define OP_PIPE_CONTROL OP_3D_MEDIA(0x3, 0x2, 0x00)
+#define OP_3DPRIMITIVE OP_3D_MEDIA(0x3, 0x3, 0x00)
+
+/* VCCP Command Parser */
+
+/*
+ * Below MFX and VBE cmd definition is from vaapi intel driver project (BSD License)
+ * git://anongit.freedesktop.org/vaapi/intel-driver
+ * src/i965_defines.h
+ *
+ */
+
+#define OP_MFX(pipeline, op, sub_opa, sub_opb) \
+ (3 << 13 | \
+ (pipeline) << 11 | \
+ (op) << 8 | \
+ (sub_opa) << 5 | \
+ (sub_opb))
+
+#define OP_MFX_PIPE_MODE_SELECT OP_MFX(2, 0, 0, 0) /* ALL */
+#define OP_MFX_SURFACE_STATE OP_MFX(2, 0, 0, 1) /* ALL */
+#define OP_MFX_PIPE_BUF_ADDR_STATE OP_MFX(2, 0, 0, 2) /* ALL */
+#define OP_MFX_IND_OBJ_BASE_ADDR_STATE OP_MFX(2, 0, 0, 3) /* ALL */
+#define OP_MFX_BSP_BUF_BASE_ADDR_STATE OP_MFX(2, 0, 0, 4) /* ALL */
+#define OP_2_0_0_5 OP_MFX(2, 0, 0, 5) /* ALL */
+#define OP_MFX_STATE_POINTER OP_MFX(2, 0, 0, 6) /* ALL */
+#define OP_MFX_QM_STATE OP_MFX(2, 0, 0, 7) /* IVB+ */
+#define OP_MFX_FQM_STATE OP_MFX(2, 0, 0, 8) /* IVB+ */
+#define OP_MFX_PAK_INSERT_OBJECT OP_MFX(2, 0, 2, 8) /* IVB+ */
+#define OP_MFX_STITCH_OBJECT OP_MFX(2, 0, 2, 0xA) /* IVB+ */
+
+#define OP_MFD_IT_OBJECT OP_MFX(2, 0, 1, 9) /* ALL */
+
+#define OP_MFX_WAIT OP_MFX(1, 0, 0, 0) /* IVB+ */
+#define OP_MFX_AVC_IMG_STATE OP_MFX(2, 1, 0, 0) /* ALL */
+#define OP_MFX_AVC_QM_STATE OP_MFX(2, 1, 0, 1) /* ALL */
+#define OP_MFX_AVC_DIRECTMODE_STATE OP_MFX(2, 1, 0, 2) /* ALL */
+#define OP_MFX_AVC_SLICE_STATE OP_MFX(2, 1, 0, 3) /* ALL */
+#define OP_MFX_AVC_REF_IDX_STATE OP_MFX(2, 1, 0, 4) /* ALL */
+#define OP_MFX_AVC_WEIGHTOFFSET_STATE OP_MFX(2, 1, 0, 5) /* ALL */
+#define OP_MFD_AVC_PICID_STATE OP_MFX(2, 1, 1, 5) /* HSW+ */
+#define OP_MFD_AVC_DPB_STATE OP_MFX(2, 1, 1, 6) /* IVB+ */
+#define OP_MFD_AVC_SLICEADDR OP_MFX(2, 1, 1, 7) /* IVB+ */
+#define OP_MFD_AVC_BSD_OBJECT OP_MFX(2, 1, 1, 8) /* ALL */
+#define OP_MFC_AVC_PAK_OBJECT OP_MFX(2, 1, 2, 9) /* ALL */
+
+#define OP_MFX_VC1_PRED_PIPE_STATE OP_MFX(2, 2, 0, 1) /* ALL */
+#define OP_MFX_VC1_DIRECTMODE_STATE OP_MFX(2, 2, 0, 2) /* ALL */
+#define OP_MFD_VC1_SHORT_PIC_STATE OP_MFX(2, 2, 1, 0) /* IVB+ */
+#define OP_MFD_VC1_LONG_PIC_STATE OP_MFX(2, 2, 1, 1) /* IVB+ */
+#define OP_MFD_VC1_BSD_OBJECT OP_MFX(2, 2, 1, 8) /* ALL */
+
+#define OP_MFX_MPEG2_PIC_STATE OP_MFX(2, 3, 0, 0) /* ALL */
+#define OP_MFX_MPEG2_QM_STATE OP_MFX(2, 3, 0, 1) /* ALL */
+#define OP_MFD_MPEG2_BSD_OBJECT OP_MFX(2, 3, 1, 8) /* ALL */
+#define OP_MFC_MPEG2_SLICEGROUP_STATE OP_MFX(2, 3, 2, 3) /* ALL */
+#define OP_MFC_MPEG2_PAK_OBJECT OP_MFX(2, 3, 2, 9) /* ALL */
+
+#define OP_MFX_2_6_0_0 OP_MFX(2, 6, 0, 0) /* IVB+ */
+#define OP_MFX_2_6_0_8 OP_MFX(2, 6, 0, 8) /* IVB+ */
+#define OP_MFX_2_6_0_9 OP_MFX(2, 6, 0, 9) /* IVB+ */
+
+#define OP_MFX_JPEG_PIC_STATE OP_MFX(2, 7, 0, 0)
+#define OP_MFX_JPEG_HUFF_TABLE_STATE OP_MFX(2, 7, 0, 2)
+#define OP_MFD_JPEG_BSD_OBJECT OP_MFX(2, 7, 1, 8)
+
+#define OP_VEB(pipeline, op, sub_opa, sub_opb) \
+ (3 << 13 | \
+ (pipeline) << 11 | \
+ (op) << 8 | \
+ (sub_opa) << 5 | \
+ (sub_opb))
+
+#define OP_VEB_SURFACE_STATE OP_VEB(2, 4, 0, 0)
+#define OP_VEB_STATE OP_VEB(2, 4, 0, 2)
+#define OP_VEB_DNDI_IECP_STATE OP_VEB(2, 4, 0, 3)
+
+struct parser_exec_state;
+
+typedef int (*parser_cmd_handler)(struct parser_exec_state *s);
+
+#define GVT_CMD_HASH_BITS 7
+
+/* which DWords need address fix */
+#define ADDR_FIX_1(x1) (1 << (x1))
+#define ADDR_FIX_2(x1, x2) (ADDR_FIX_1(x1) | ADDR_FIX_1(x2))
+#define ADDR_FIX_3(x1, x2, x3) (ADDR_FIX_1(x1) | ADDR_FIX_2(x2, x3))
+#define ADDR_FIX_4(x1, x2, x3, x4) (ADDR_FIX_1(x1) | ADDR_FIX_3(x2, x3, x4))
+#define ADDR_FIX_5(x1, x2, x3, x4, x5) (ADDR_FIX_1(x1) | ADDR_FIX_4(x2, x3, x4, x5))
+
+#define DWORD_FIELD(dword, end, start) \
+ FIELD_GET(GENMASK(end, start), cmd_val(s, dword))
+
+#define OP_LENGTH_BIAS 2
+#define CMD_LEN(value) (value + OP_LENGTH_BIAS)
+
+static int gvt_check_valid_cmd_length(int len, int valid_len)
+{
+ if (valid_len != len) {
+ gvt_err("len is not valid: len=%u valid_len=%u\n",
+ len, valid_len);
+ return -EFAULT;
+ }
+ return 0;
+}
+
+struct cmd_info {
+ const char *name;
+ u32 opcode;
+
+#define F_LEN_MASK 3U
+#define F_LEN_CONST 1U
+#define F_LEN_VAR 0U
+/* value is const although LEN maybe variable */
+#define F_LEN_VAR_FIXED (1<<1)
+
+/*
+ * command has its own ip advance logic
+ * e.g. MI_BATCH_START, MI_BATCH_END
+ */
+#define F_IP_ADVANCE_CUSTOM (1<<2)
+ u32 flag;
+
+#define R_RCS BIT(RCS0)
+#define R_VCS1 BIT(VCS0)
+#define R_VCS2 BIT(VCS1)
+#define R_VCS (R_VCS1 | R_VCS2)
+#define R_BCS BIT(BCS0)
+#define R_VECS BIT(VECS0)
+#define R_ALL (R_RCS | R_VCS | R_BCS | R_VECS)
+ /* rings that support this cmd: BLT/RCS/VCS/VECS */
+ intel_engine_mask_t rings;
+
+ /* devices that support this cmd: SNB/IVB/HSW/... */
+ u16 devices;
+
+ /* which DWords are address that need fix up.
+ * bit 0 means a 32-bit non address operand in command
+ * bit 1 means address operand, which could be 32-bit
+ * or 64-bit depending on different architectures.(
+ * defined by "gmadr_bytes_in_cmd" in intel_gvt.
+ * No matter the address length, each address only takes
+ * one bit in the bitmap.
+ */
+ u16 addr_bitmap;
+
+ /* flag == F_LEN_CONST : command length
+ * flag == F_LEN_VAR : length bias bits
+ * Note: length is in DWord
+ */
+ u32 len;
+
+ parser_cmd_handler handler;
+
+ /* valid length in DWord */
+ u32 valid_len;
+};
+
+struct cmd_entry {
+ struct hlist_node hlist;
+ const struct cmd_info *info;
+};
+
+enum {
+ RING_BUFFER_INSTRUCTION,
+ BATCH_BUFFER_INSTRUCTION,
+ BATCH_BUFFER_2ND_LEVEL,
+ RING_BUFFER_CTX,
+};
+
+enum {
+ GTT_BUFFER,
+ PPGTT_BUFFER
+};
+
+struct parser_exec_state {
+ struct intel_vgpu *vgpu;
+ const struct intel_engine_cs *engine;
+
+ int buf_type;
+
+ /* batch buffer address type */
+ int buf_addr_type;
+
+ /* graphics memory address of ring buffer start */
+ unsigned long ring_start;
+ unsigned long ring_size;
+ unsigned long ring_head;
+ unsigned long ring_tail;
+
+ /* instruction graphics memory address */
+ unsigned long ip_gma;
+
+ /* mapped va of the instr_gma */
+ void *ip_va;
+ void *rb_va;
+
+ void *ret_bb_va;
+ /* next instruction when return from batch buffer to ring buffer */
+ unsigned long ret_ip_gma_ring;
+
+ /* next instruction when return from 2nd batch buffer to batch buffer */
+ unsigned long ret_ip_gma_bb;
+
+ /* batch buffer address type (GTT or PPGTT)
+ * used when ret from 2nd level batch buffer
+ */
+ int saved_buf_addr_type;
+ bool is_ctx_wa;
+ bool is_init_ctx;
+
+ const struct cmd_info *info;
+
+ struct intel_vgpu_workload *workload;
+};
+
+#define gmadr_dw_number(s) \
+ (s->vgpu->gvt->device_info.gmadr_bytes_in_cmd >> 2)
+
+static unsigned long bypass_scan_mask = 0;
+
+/* ring ALL, type = 0 */
+static const struct sub_op_bits sub_op_mi[] = {
+ {31, 29},
+ {28, 23},
+};
+
+static const struct decode_info decode_info_mi = {
+ "MI",
+ OP_LEN_MI,
+ ARRAY_SIZE(sub_op_mi),
+ sub_op_mi,
+};
+
+/* ring RCS, command type 2 */
+static const struct sub_op_bits sub_op_2d[] = {
+ {31, 29},
+ {28, 22},
+};
+
+static const struct decode_info decode_info_2d = {
+ "2D",
+ OP_LEN_2D,
+ ARRAY_SIZE(sub_op_2d),
+ sub_op_2d,
+};
+
+/* ring RCS, command type 3 */
+static const struct sub_op_bits sub_op_3d_media[] = {
+ {31, 29},
+ {28, 27},
+ {26, 24},
+ {23, 16},
+};
+
+static const struct decode_info decode_info_3d_media = {
+ "3D_Media",
+ OP_LEN_3D_MEDIA,
+ ARRAY_SIZE(sub_op_3d_media),
+ sub_op_3d_media,
+};
+
+/* ring VCS, command type 3 */
+static const struct sub_op_bits sub_op_mfx_vc[] = {
+ {31, 29},
+ {28, 27},
+ {26, 24},
+ {23, 21},
+ {20, 16},
+};
+
+static const struct decode_info decode_info_mfx_vc = {
+ "MFX_VC",
+ OP_LEN_MFX_VC,
+ ARRAY_SIZE(sub_op_mfx_vc),
+ sub_op_mfx_vc,
+};
+
+/* ring VECS, command type 3 */
+static const struct sub_op_bits sub_op_vebox[] = {
+ {31, 29},
+ {28, 27},
+ {26, 24},
+ {23, 21},
+ {20, 16},
+};
+
+static const struct decode_info decode_info_vebox = {
+ "VEBOX",
+ OP_LEN_VEBOX,
+ ARRAY_SIZE(sub_op_vebox),
+ sub_op_vebox,
+};
+
+static const struct decode_info *ring_decode_info[I915_NUM_ENGINES][8] = {
+ [RCS0] = {
+ &decode_info_mi,
+ NULL,
+ NULL,
+ &decode_info_3d_media,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ },
+
+ [VCS0] = {
+ &decode_info_mi,
+ NULL,
+ NULL,
+ &decode_info_mfx_vc,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ },
+
+ [BCS0] = {
+ &decode_info_mi,
+ NULL,
+ &decode_info_2d,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ },
+
+ [VECS0] = {
+ &decode_info_mi,
+ NULL,
+ NULL,
+ &decode_info_vebox,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ },
+
+ [VCS1] = {
+ &decode_info_mi,
+ NULL,
+ NULL,
+ &decode_info_mfx_vc,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ },
+};
+
+static inline u32 get_opcode(u32 cmd, const struct intel_engine_cs *engine)
+{
+ const struct decode_info *d_info;
+
+ d_info = ring_decode_info[engine->id][CMD_TYPE(cmd)];
+ if (d_info == NULL)
+ return INVALID_OP;
+
+ return cmd >> (32 - d_info->op_len);
+}
+
+static inline const struct cmd_info *
+find_cmd_entry(struct intel_gvt *gvt, unsigned int opcode,
+ const struct intel_engine_cs *engine)
+{
+ struct cmd_entry *e;
+
+ hash_for_each_possible(gvt->cmd_table, e, hlist, opcode) {
+ if (opcode == e->info->opcode &&
+ e->info->rings & engine->mask)
+ return e->info;
+ }
+ return NULL;
+}
+
+static inline const struct cmd_info *
+get_cmd_info(struct intel_gvt *gvt, u32 cmd,
+ const struct intel_engine_cs *engine)
+{
+ u32 opcode;
+
+ opcode = get_opcode(cmd, engine);
+ if (opcode == INVALID_OP)
+ return NULL;
+
+ return find_cmd_entry(gvt, opcode, engine);
+}
+
+static inline u32 sub_op_val(u32 cmd, u32 hi, u32 low)
+{
+ return (cmd >> low) & ((1U << (hi - low + 1)) - 1);
+}
+
+static inline void print_opcode(u32 cmd, const struct intel_engine_cs *engine)
+{
+ const struct decode_info *d_info;
+ int i;
+
+ d_info = ring_decode_info[engine->id][CMD_TYPE(cmd)];
+ if (d_info == NULL)
+ return;
+
+ gvt_dbg_cmd("opcode=0x%x %s sub_ops:",
+ cmd >> (32 - d_info->op_len), d_info->name);
+
+ for (i = 0; i < d_info->nr_sub_op; i++)
+ pr_err("0x%x ", sub_op_val(cmd, d_info->sub_op[i].hi,
+ d_info->sub_op[i].low));
+
+ pr_err("\n");
+}
+
+static inline u32 *cmd_ptr(struct parser_exec_state *s, int index)
+{
+ return s->ip_va + (index << 2);
+}
+
+static inline u32 cmd_val(struct parser_exec_state *s, int index)
+{
+ return *cmd_ptr(s, index);
+}
+
+static inline bool is_init_ctx(struct parser_exec_state *s)
+{
+ return (s->buf_type == RING_BUFFER_CTX && s->is_init_ctx);
+}
+
+static void parser_exec_state_dump(struct parser_exec_state *s)
+{
+ int cnt = 0;
+ int i;
+
+ gvt_dbg_cmd(" vgpu%d RING%s: ring_start(%08lx) ring_end(%08lx)"
+ " ring_head(%08lx) ring_tail(%08lx)\n",
+ s->vgpu->id, s->engine->name,
+ s->ring_start, s->ring_start + s->ring_size,
+ s->ring_head, s->ring_tail);
+
+ gvt_dbg_cmd(" %s %s ip_gma(%08lx) ",
+ s->buf_type == RING_BUFFER_INSTRUCTION ?
+ "RING_BUFFER" : ((s->buf_type == RING_BUFFER_CTX) ?
+ "CTX_BUFFER" : "BATCH_BUFFER"),
+ s->buf_addr_type == GTT_BUFFER ?
+ "GTT" : "PPGTT", s->ip_gma);
+
+ if (s->ip_va == NULL) {
+ gvt_dbg_cmd(" ip_va(NULL)");
+ return;
+ }
+
+ gvt_dbg_cmd(" ip_va=%p: %08x %08x %08x %08x\n",
+ s->ip_va, cmd_val(s, 0), cmd_val(s, 1),
+ cmd_val(s, 2), cmd_val(s, 3));
+
+ print_opcode(cmd_val(s, 0), s->engine);
+
+ s->ip_va = (u32 *)((((u64)s->ip_va) >> 12) << 12);
+
+ while (cnt < 1024) {
+ gvt_dbg_cmd("ip_va=%p: ", s->ip_va);
+ for (i = 0; i < 8; i++)
+ gvt_dbg_cmd("%08x ", cmd_val(s, i));
+ gvt_dbg_cmd("\n");
+
+ s->ip_va += 8 * sizeof(u32);
+ cnt += 8;
+ }
+}
+
+static inline void update_ip_va(struct parser_exec_state *s)
+{
+ unsigned long len = 0;
+
+ if (WARN_ON(s->ring_head == s->ring_tail))
+ return;
+
+ if (s->buf_type == RING_BUFFER_INSTRUCTION ||
+ s->buf_type == RING_BUFFER_CTX) {
+ unsigned long ring_top = s->ring_start + s->ring_size;
+
+ if (s->ring_head > s->ring_tail) {
+ if (s->ip_gma >= s->ring_head && s->ip_gma < ring_top)
+ len = (s->ip_gma - s->ring_head);
+ else if (s->ip_gma >= s->ring_start &&
+ s->ip_gma <= s->ring_tail)
+ len = (ring_top - s->ring_head) +
+ (s->ip_gma - s->ring_start);
+ } else
+ len = (s->ip_gma - s->ring_head);
+
+ s->ip_va = s->rb_va + len;
+ } else {/* shadow batch buffer */
+ s->ip_va = s->ret_bb_va;
+ }
+}
+
+static inline int ip_gma_set(struct parser_exec_state *s,
+ unsigned long ip_gma)
+{
+ WARN_ON(!IS_ALIGNED(ip_gma, 4));
+
+ s->ip_gma = ip_gma;
+ update_ip_va(s);
+ return 0;
+}
+
+static inline int ip_gma_advance(struct parser_exec_state *s,
+ unsigned int dw_len)
+{
+ s->ip_gma += (dw_len << 2);
+
+ if (s->buf_type == RING_BUFFER_INSTRUCTION) {
+ if (s->ip_gma >= s->ring_start + s->ring_size)
+ s->ip_gma -= s->ring_size;
+ update_ip_va(s);
+ } else {
+ s->ip_va += (dw_len << 2);
+ }
+
+ return 0;
+}
+
+static inline int get_cmd_length(const struct cmd_info *info, u32 cmd)
+{
+ if ((info->flag & F_LEN_MASK) == F_LEN_CONST)
+ return info->len;
+ else
+ return (cmd & ((1U << info->len) - 1)) + 2;
+ return 0;
+}
+
+static inline int cmd_length(struct parser_exec_state *s)
+{
+ return get_cmd_length(s->info, cmd_val(s, 0));
+}
+
+/* do not remove this, some platform may need clflush here */
+#define patch_value(s, addr, val) do { \
+ *addr = val; \
+} while (0)
+
+static inline bool is_mocs_mmio(unsigned int offset)
+{
+ return ((offset >= 0xc800) && (offset <= 0xcff8)) ||
+ ((offset >= 0xb020) && (offset <= 0xb0a0));
+}
+
+static int is_cmd_update_pdps(unsigned int offset,
+ struct parser_exec_state *s)
+{
+ u32 base = s->workload->engine->mmio_base;
+ return i915_mmio_reg_equal(_MMIO(offset), GEN8_RING_PDP_UDW(base, 0));
+}
+
+static int cmd_pdp_mmio_update_handler(struct parser_exec_state *s,
+ unsigned int offset, unsigned int index)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ struct intel_vgpu_mm *shadow_mm = s->workload->shadow_mm;
+ struct intel_vgpu_mm *mm;
+ u64 pdps[GEN8_3LVL_PDPES];
+
+ if (shadow_mm->ppgtt_mm.root_entry_type ==
+ GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
+ pdps[0] = (u64)cmd_val(s, 2) << 32;
+ pdps[0] |= cmd_val(s, 4);
+
+ mm = intel_vgpu_find_ppgtt_mm(vgpu, pdps);
+ if (!mm) {
+ gvt_vgpu_err("failed to get the 4-level shadow vm\n");
+ return -EINVAL;
+ }
+ intel_vgpu_mm_get(mm);
+ list_add_tail(&mm->ppgtt_mm.link,
+ &s->workload->lri_shadow_mm);
+ *cmd_ptr(s, 2) = upper_32_bits(mm->ppgtt_mm.shadow_pdps[0]);
+ *cmd_ptr(s, 4) = lower_32_bits(mm->ppgtt_mm.shadow_pdps[0]);
+ } else {
+ /* Currently all guests use PML4 table and now can't
+ * have a guest with 3-level table but uses LRI for
+ * PPGTT update. So this is simply un-testable. */
+ GEM_BUG_ON(1);
+ gvt_vgpu_err("invalid shared shadow vm type\n");
+ return -EINVAL;
+ }
+ return 0;
+}
+
+static int cmd_reg_handler(struct parser_exec_state *s,
+ unsigned int offset, unsigned int index, char *cmd)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ u32 ctx_sr_ctl;
+ u32 *vreg, vreg_old;
+
+ if (offset + 4 > gvt->device_info.mmio_size) {
+ gvt_vgpu_err("%s access to (%x) outside of MMIO range\n",
+ cmd, offset);
+ return -EFAULT;
+ }
+
+ if (is_init_ctx(s)) {
+ struct intel_gvt_mmio_info *mmio_info;
+
+ intel_gvt_mmio_set_cmd_accessible(gvt, offset);
+ mmio_info = intel_gvt_find_mmio_info(gvt, offset);
+ if (mmio_info && mmio_info->write)
+ intel_gvt_mmio_set_cmd_write_patch(gvt, offset);
+ return 0;
+ }
+
+ if (!intel_gvt_mmio_is_cmd_accessible(gvt, offset)) {
+ gvt_vgpu_err("%s access to non-render register (%x)\n",
+ cmd, offset);
+ return -EBADRQC;
+ }
+
+ if (!strncmp(cmd, "srm", 3) ||
+ !strncmp(cmd, "lrm", 3)) {
+ if (offset == i915_mmio_reg_offset(GEN8_L3SQCREG4) ||
+ offset == 0x21f0 ||
+ (IS_BROADWELL(gvt->gt->i915) &&
+ offset == i915_mmio_reg_offset(INSTPM)))
+ return 0;
+ else {
+ gvt_vgpu_err("%s access to register (%x)\n",
+ cmd, offset);
+ return -EPERM;
+ }
+ }
+
+ if (!strncmp(cmd, "lrr-src", 7) ||
+ !strncmp(cmd, "lrr-dst", 7)) {
+ if (IS_BROADWELL(gvt->gt->i915) && offset == 0x215c)
+ return 0;
+ else {
+ gvt_vgpu_err("not allowed cmd %s reg (%x)\n", cmd, offset);
+ return -EPERM;
+ }
+ }
+
+ if (!strncmp(cmd, "pipe_ctrl", 9)) {
+ /* TODO: add LRI POST logic here */
+ return 0;
+ }
+
+ if (strncmp(cmd, "lri", 3))
+ return -EPERM;
+
+ /* below are all lri handlers */
+ vreg = &vgpu_vreg(s->vgpu, offset);
+
+ if (is_cmd_update_pdps(offset, s) &&
+ cmd_pdp_mmio_update_handler(s, offset, index))
+ return -EINVAL;
+
+ if (offset == i915_mmio_reg_offset(DERRMR) ||
+ offset == i915_mmio_reg_offset(FORCEWAKE_MT)) {
+ /* Writing to HW VGT_PVINFO_PAGE offset will be discarded */
+ patch_value(s, cmd_ptr(s, index), VGT_PVINFO_PAGE);
+ }
+
+ if (is_mocs_mmio(offset))
+ *vreg = cmd_val(s, index + 1);
+
+ vreg_old = *vreg;
+
+ if (intel_gvt_mmio_is_cmd_write_patch(gvt, offset)) {
+ u32 cmdval_new, cmdval;
+ struct intel_gvt_mmio_info *mmio_info;
+
+ cmdval = cmd_val(s, index + 1);
+
+ mmio_info = intel_gvt_find_mmio_info(gvt, offset);
+ if (!mmio_info) {
+ cmdval_new = cmdval;
+ } else {
+ u64 ro_mask = mmio_info->ro_mask;
+ int ret;
+
+ if (likely(!ro_mask))
+ ret = mmio_info->write(s->vgpu, offset,
+ &cmdval, 4);
+ else {
+ gvt_vgpu_err("try to write RO reg %x\n",
+ offset);
+ ret = -EBADRQC;
+ }
+ if (ret)
+ return ret;
+ cmdval_new = *vreg;
+ }
+ if (cmdval_new != cmdval)
+ patch_value(s, cmd_ptr(s, index+1), cmdval_new);
+ }
+
+ /* only patch cmd. restore vreg value if changed in mmio write handler*/
+ *vreg = vreg_old;
+
+ /* TODO
+ * In order to let workload with inhibit context to generate
+ * correct image data into memory, vregs values will be loaded to
+ * hw via LRIs in the workload with inhibit context. But as
+ * indirect context is loaded prior to LRIs in workload, we don't
+ * want reg values specified in indirect context overwritten by
+ * LRIs in workloads. So, when scanning an indirect context, we
+ * update reg values in it into vregs, so LRIs in workload with
+ * inhibit context will restore with correct values
+ */
+ if (GRAPHICS_VER(s->engine->i915) == 9 &&
+ intel_gvt_mmio_is_sr_in_ctx(gvt, offset) &&
+ !strncmp(cmd, "lri", 3)) {
+ intel_gvt_read_gpa(s->vgpu,
+ s->workload->ring_context_gpa + 12, &ctx_sr_ctl, 4);
+ /* check inhibit context */
+ if (ctx_sr_ctl & 1) {
+ u32 data = cmd_val(s, index + 1);
+
+ if (intel_gvt_mmio_has_mode_mask(s->vgpu->gvt, offset))
+ intel_vgpu_mask_mmio_write(vgpu,
+ offset, &data, 4);
+ else
+ vgpu_vreg(vgpu, offset) = data;
+ }
+ }
+
+ return 0;
+}
+
+#define cmd_reg(s, i) \
+ (cmd_val(s, i) & GENMASK(22, 2))
+
+#define cmd_reg_inhibit(s, i) \
+ (cmd_val(s, i) & GENMASK(22, 18))
+
+#define cmd_gma(s, i) \
+ (cmd_val(s, i) & GENMASK(31, 2))
+
+#define cmd_gma_hi(s, i) \
+ (cmd_val(s, i) & GENMASK(15, 0))
+
+static int cmd_handler_lri(struct parser_exec_state *s)
+{
+ int i, ret = 0;
+ int cmd_len = cmd_length(s);
+
+ for (i = 1; i < cmd_len; i += 2) {
+ if (IS_BROADWELL(s->engine->i915) && s->engine->id != RCS0) {
+ if (s->engine->id == BCS0 &&
+ cmd_reg(s, i) == i915_mmio_reg_offset(DERRMR))
+ ret |= 0;
+ else
+ ret |= cmd_reg_inhibit(s, i) ? -EBADRQC : 0;
+ }
+ if (ret)
+ break;
+ ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lri");
+ if (ret)
+ break;
+ }
+ return ret;
+}
+
+static int cmd_handler_lrr(struct parser_exec_state *s)
+{
+ int i, ret = 0;
+ int cmd_len = cmd_length(s);
+
+ for (i = 1; i < cmd_len; i += 2) {
+ if (IS_BROADWELL(s->engine->i915))
+ ret |= ((cmd_reg_inhibit(s, i) ||
+ (cmd_reg_inhibit(s, i + 1)))) ?
+ -EBADRQC : 0;
+ if (ret)
+ break;
+ ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrr-src");
+ if (ret)
+ break;
+ ret |= cmd_reg_handler(s, cmd_reg(s, i + 1), i, "lrr-dst");
+ if (ret)
+ break;
+ }
+ return ret;
+}
+
+static inline int cmd_address_audit(struct parser_exec_state *s,
+ unsigned long guest_gma, int op_size, bool index_mode);
+
+static int cmd_handler_lrm(struct parser_exec_state *s)
+{
+ struct intel_gvt *gvt = s->vgpu->gvt;
+ int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
+ unsigned long gma;
+ int i, ret = 0;
+ int cmd_len = cmd_length(s);
+
+ for (i = 1; i < cmd_len;) {
+ if (IS_BROADWELL(s->engine->i915))
+ ret |= (cmd_reg_inhibit(s, i)) ? -EBADRQC : 0;
+ if (ret)
+ break;
+ ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrm");
+ if (ret)
+ break;
+ if (cmd_val(s, 0) & (1 << 22)) {
+ gma = cmd_gma(s, i + 1);
+ if (gmadr_bytes == 8)
+ gma |= (cmd_gma_hi(s, i + 2)) << 32;
+ ret |= cmd_address_audit(s, gma, sizeof(u32), false);
+ if (ret)
+ break;
+ }
+ i += gmadr_dw_number(s) + 1;
+ }
+ return ret;
+}
+
+static int cmd_handler_srm(struct parser_exec_state *s)
+{
+ int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+ unsigned long gma;
+ int i, ret = 0;
+ int cmd_len = cmd_length(s);
+
+ for (i = 1; i < cmd_len;) {
+ ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "srm");
+ if (ret)
+ break;
+ if (cmd_val(s, 0) & (1 << 22)) {
+ gma = cmd_gma(s, i + 1);
+ if (gmadr_bytes == 8)
+ gma |= (cmd_gma_hi(s, i + 2)) << 32;
+ ret |= cmd_address_audit(s, gma, sizeof(u32), false);
+ if (ret)
+ break;
+ }
+ i += gmadr_dw_number(s) + 1;
+ }
+ return ret;
+}
+
+struct cmd_interrupt_event {
+ int pipe_control_notify;
+ int mi_flush_dw;
+ int mi_user_interrupt;
+};
+
+static const struct cmd_interrupt_event cmd_interrupt_events[] = {
+ [RCS0] = {
+ .pipe_control_notify = RCS_PIPE_CONTROL,
+ .mi_flush_dw = INTEL_GVT_EVENT_RESERVED,
+ .mi_user_interrupt = RCS_MI_USER_INTERRUPT,
+ },
+ [BCS0] = {
+ .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
+ .mi_flush_dw = BCS_MI_FLUSH_DW,
+ .mi_user_interrupt = BCS_MI_USER_INTERRUPT,
+ },
+ [VCS0] = {
+ .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
+ .mi_flush_dw = VCS_MI_FLUSH_DW,
+ .mi_user_interrupt = VCS_MI_USER_INTERRUPT,
+ },
+ [VCS1] = {
+ .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
+ .mi_flush_dw = VCS2_MI_FLUSH_DW,
+ .mi_user_interrupt = VCS2_MI_USER_INTERRUPT,
+ },
+ [VECS0] = {
+ .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
+ .mi_flush_dw = VECS_MI_FLUSH_DW,
+ .mi_user_interrupt = VECS_MI_USER_INTERRUPT,
+ },
+};
+
+static int cmd_handler_pipe_control(struct parser_exec_state *s)
+{
+ int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+ unsigned long gma;
+ bool index_mode = false;
+ unsigned int post_sync;
+ int ret = 0;
+ u32 hws_pga, val;
+
+ post_sync = (cmd_val(s, 1) & PIPE_CONTROL_POST_SYNC_OP_MASK) >> 14;
+
+ /* LRI post sync */
+ if (cmd_val(s, 1) & PIPE_CONTROL_MMIO_WRITE)
+ ret = cmd_reg_handler(s, cmd_reg(s, 2), 1, "pipe_ctrl");
+ /* post sync */
+ else if (post_sync) {
+ if (post_sync == 2)
+ ret = cmd_reg_handler(s, 0x2350, 1, "pipe_ctrl");
+ else if (post_sync == 3)
+ ret = cmd_reg_handler(s, 0x2358, 1, "pipe_ctrl");
+ else if (post_sync == 1) {
+ /* check ggtt*/
+ if ((cmd_val(s, 1) & PIPE_CONTROL_GLOBAL_GTT_IVB)) {
+ gma = cmd_val(s, 2) & GENMASK(31, 3);
+ if (gmadr_bytes == 8)
+ gma |= (cmd_gma_hi(s, 3)) << 32;
+ /* Store Data Index */
+ if (cmd_val(s, 1) & (1 << 21))
+ index_mode = true;
+ ret |= cmd_address_audit(s, gma, sizeof(u64),
+ index_mode);
+ if (ret)
+ return ret;
+ if (index_mode) {
+ hws_pga = s->vgpu->hws_pga[s->engine->id];
+ gma = hws_pga + gma;
+ patch_value(s, cmd_ptr(s, 2), gma);
+ val = cmd_val(s, 1) & (~(1 << 21));
+ patch_value(s, cmd_ptr(s, 1), val);
+ }
+ }
+ }
+ }
+
+ if (ret)
+ return ret;
+
+ if (cmd_val(s, 1) & PIPE_CONTROL_NOTIFY)
+ set_bit(cmd_interrupt_events[s->engine->id].pipe_control_notify,
+ s->workload->pending_events);
+ return 0;
+}
+
+static int cmd_handler_mi_user_interrupt(struct parser_exec_state *s)
+{
+ set_bit(cmd_interrupt_events[s->engine->id].mi_user_interrupt,
+ s->workload->pending_events);
+ patch_value(s, cmd_ptr(s, 0), MI_NOOP);
+ return 0;
+}
+
+static int cmd_advance_default(struct parser_exec_state *s)
+{
+ return ip_gma_advance(s, cmd_length(s));
+}
+
+static int cmd_handler_mi_batch_buffer_end(struct parser_exec_state *s)
+{
+ int ret;
+
+ if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
+ s->buf_type = BATCH_BUFFER_INSTRUCTION;
+ ret = ip_gma_set(s, s->ret_ip_gma_bb);
+ s->buf_addr_type = s->saved_buf_addr_type;
+ } else if (s->buf_type == RING_BUFFER_CTX) {
+ ret = ip_gma_set(s, s->ring_tail);
+ } else {
+ s->buf_type = RING_BUFFER_INSTRUCTION;
+ s->buf_addr_type = GTT_BUFFER;
+ if (s->ret_ip_gma_ring >= s->ring_start + s->ring_size)
+ s->ret_ip_gma_ring -= s->ring_size;
+ ret = ip_gma_set(s, s->ret_ip_gma_ring);
+ }
+ return ret;
+}
+
+struct mi_display_flip_command_info {
+ int pipe;
+ int plane;
+ int event;
+ i915_reg_t stride_reg;
+ i915_reg_t ctrl_reg;
+ i915_reg_t surf_reg;
+ u64 stride_val;
+ u64 tile_val;
+ u64 surf_val;
+ bool async_flip;
+};
+
+struct plane_code_mapping {
+ int pipe;
+ int plane;
+ int event;
+};
+
+static int gen8_decode_mi_display_flip(struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ struct drm_i915_private *dev_priv = s->engine->i915;
+ struct plane_code_mapping gen8_plane_code[] = {
+ [0] = {PIPE_A, PLANE_A, PRIMARY_A_FLIP_DONE},
+ [1] = {PIPE_B, PLANE_A, PRIMARY_B_FLIP_DONE},
+ [2] = {PIPE_A, PLANE_B, SPRITE_A_FLIP_DONE},
+ [3] = {PIPE_B, PLANE_B, SPRITE_B_FLIP_DONE},
+ [4] = {PIPE_C, PLANE_A, PRIMARY_C_FLIP_DONE},
+ [5] = {PIPE_C, PLANE_B, SPRITE_C_FLIP_DONE},
+ };
+ u32 dword0, dword1, dword2;
+ u32 v;
+
+ dword0 = cmd_val(s, 0);
+ dword1 = cmd_val(s, 1);
+ dword2 = cmd_val(s, 2);
+
+ v = (dword0 & GENMASK(21, 19)) >> 19;
+ if (drm_WARN_ON(&dev_priv->drm, v >= ARRAY_SIZE(gen8_plane_code)))
+ return -EBADRQC;
+
+ info->pipe = gen8_plane_code[v].pipe;
+ info->plane = gen8_plane_code[v].plane;
+ info->event = gen8_plane_code[v].event;
+ info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
+ info->tile_val = (dword1 & 0x1);
+ info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
+ info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
+
+ if (info->plane == PLANE_A) {
+ info->ctrl_reg = DSPCNTR(info->pipe);
+ info->stride_reg = DSPSTRIDE(info->pipe);
+ info->surf_reg = DSPSURF(info->pipe);
+ } else if (info->plane == PLANE_B) {
+ info->ctrl_reg = SPRCTL(info->pipe);
+ info->stride_reg = SPRSTRIDE(info->pipe);
+ info->surf_reg = SPRSURF(info->pipe);
+ } else {
+ drm_WARN_ON(&dev_priv->drm, 1);
+ return -EBADRQC;
+ }
+ return 0;
+}
+
+static int skl_decode_mi_display_flip(struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ struct drm_i915_private *dev_priv = s->engine->i915;
+ struct intel_vgpu *vgpu = s->vgpu;
+ u32 dword0 = cmd_val(s, 0);
+ u32 dword1 = cmd_val(s, 1);
+ u32 dword2 = cmd_val(s, 2);
+ u32 plane = (dword0 & GENMASK(12, 8)) >> 8;
+
+ info->plane = PRIMARY_PLANE;
+
+ switch (plane) {
+ case MI_DISPLAY_FLIP_SKL_PLANE_1_A:
+ info->pipe = PIPE_A;
+ info->event = PRIMARY_A_FLIP_DONE;
+ break;
+ case MI_DISPLAY_FLIP_SKL_PLANE_1_B:
+ info->pipe = PIPE_B;
+ info->event = PRIMARY_B_FLIP_DONE;
+ break;
+ case MI_DISPLAY_FLIP_SKL_PLANE_1_C:
+ info->pipe = PIPE_C;
+ info->event = PRIMARY_C_FLIP_DONE;
+ break;
+
+ case MI_DISPLAY_FLIP_SKL_PLANE_2_A:
+ info->pipe = PIPE_A;
+ info->event = SPRITE_A_FLIP_DONE;
+ info->plane = SPRITE_PLANE;
+ break;
+ case MI_DISPLAY_FLIP_SKL_PLANE_2_B:
+ info->pipe = PIPE_B;
+ info->event = SPRITE_B_FLIP_DONE;
+ info->plane = SPRITE_PLANE;
+ break;
+ case MI_DISPLAY_FLIP_SKL_PLANE_2_C:
+ info->pipe = PIPE_C;
+ info->event = SPRITE_C_FLIP_DONE;
+ info->plane = SPRITE_PLANE;
+ break;
+
+ default:
+ gvt_vgpu_err("unknown plane code %d\n", plane);
+ return -EBADRQC;
+ }
+
+ info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
+ info->tile_val = (dword1 & GENMASK(2, 0));
+ info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
+ info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
+
+ info->ctrl_reg = DSPCNTR(info->pipe);
+ info->stride_reg = DSPSTRIDE(info->pipe);
+ info->surf_reg = DSPSURF(info->pipe);
+
+ return 0;
+}
+
+static int gen8_check_mi_display_flip(struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ u32 stride, tile;
+
+ if (!info->async_flip)
+ return 0;
+
+ if (GRAPHICS_VER(s->engine->i915) >= 9) {
+ stride = vgpu_vreg_t(s->vgpu, info->stride_reg) & GENMASK(9, 0);
+ tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) &
+ GENMASK(12, 10)) >> 10;
+ } else {
+ stride = (vgpu_vreg_t(s->vgpu, info->stride_reg) &
+ GENMASK(15, 6)) >> 6;
+ tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) & (1 << 10)) >> 10;
+ }
+
+ if (stride != info->stride_val)
+ gvt_dbg_cmd("cannot change stride during async flip\n");
+
+ if (tile != info->tile_val)
+ gvt_dbg_cmd("cannot change tile during async flip\n");
+
+ return 0;
+}
+
+static int gen8_update_plane_mmio_from_mi_display_flip(
+ struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ struct drm_i915_private *dev_priv = s->engine->i915;
+ struct intel_vgpu *vgpu = s->vgpu;
+
+ set_mask_bits(&vgpu_vreg_t(vgpu, info->surf_reg), GENMASK(31, 12),
+ info->surf_val << 12);
+ if (GRAPHICS_VER(dev_priv) >= 9) {
+ set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(9, 0),
+ info->stride_val);
+ set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(12, 10),
+ info->tile_val << 10);
+ } else {
+ set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(15, 6),
+ info->stride_val << 6);
+ set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(10, 10),
+ info->tile_val << 10);
+ }
+
+ if (info->plane == PLANE_PRIMARY)
+ vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(info->pipe))++;
+
+ if (info->async_flip)
+ intel_vgpu_trigger_virtual_event(vgpu, info->event);
+ else
+ set_bit(info->event, vgpu->irq.flip_done_event[info->pipe]);
+
+ return 0;
+}
+
+static int decode_mi_display_flip(struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ if (IS_BROADWELL(s->engine->i915))
+ return gen8_decode_mi_display_flip(s, info);
+ if (GRAPHICS_VER(s->engine->i915) >= 9)
+ return skl_decode_mi_display_flip(s, info);
+
+ return -ENODEV;
+}
+
+static int check_mi_display_flip(struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ return gen8_check_mi_display_flip(s, info);
+}
+
+static int update_plane_mmio_from_mi_display_flip(
+ struct parser_exec_state *s,
+ struct mi_display_flip_command_info *info)
+{
+ return gen8_update_plane_mmio_from_mi_display_flip(s, info);
+}
+
+static int cmd_handler_mi_display_flip(struct parser_exec_state *s)
+{
+ struct mi_display_flip_command_info info;
+ struct intel_vgpu *vgpu = s->vgpu;
+ int ret;
+ int i;
+ int len = cmd_length(s);
+ u32 valid_len = CMD_LEN(1);
+
+ /* Flip Type == Stereo 3D Flip */
+ if (DWORD_FIELD(2, 1, 0) == 2)
+ valid_len++;
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ valid_len);
+ if (ret)
+ return ret;
+
+ ret = decode_mi_display_flip(s, &info);
+ if (ret) {
+ gvt_vgpu_err("fail to decode MI display flip command\n");
+ return ret;
+ }
+
+ ret = check_mi_display_flip(s, &info);
+ if (ret) {
+ gvt_vgpu_err("invalid MI display flip command\n");
+ return ret;
+ }
+
+ ret = update_plane_mmio_from_mi_display_flip(s, &info);
+ if (ret) {
+ gvt_vgpu_err("fail to update plane mmio\n");
+ return ret;
+ }
+
+ for (i = 0; i < len; i++)
+ patch_value(s, cmd_ptr(s, i), MI_NOOP);
+ return 0;
+}
+
+static bool is_wait_for_flip_pending(u32 cmd)
+{
+ return cmd & (MI_WAIT_FOR_PLANE_A_FLIP_PENDING |
+ MI_WAIT_FOR_PLANE_B_FLIP_PENDING |
+ MI_WAIT_FOR_PLANE_C_FLIP_PENDING |
+ MI_WAIT_FOR_SPRITE_A_FLIP_PENDING |
+ MI_WAIT_FOR_SPRITE_B_FLIP_PENDING |
+ MI_WAIT_FOR_SPRITE_C_FLIP_PENDING);
+}
+
+static int cmd_handler_mi_wait_for_event(struct parser_exec_state *s)
+{
+ u32 cmd = cmd_val(s, 0);
+
+ if (!is_wait_for_flip_pending(cmd))
+ return 0;
+
+ patch_value(s, cmd_ptr(s, 0), MI_NOOP);
+ return 0;
+}
+
+static unsigned long get_gma_bb_from_cmd(struct parser_exec_state *s, int index)
+{
+ unsigned long addr;
+ unsigned long gma_high, gma_low;
+ struct intel_vgpu *vgpu = s->vgpu;
+ int gmadr_bytes = vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+
+ if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) {
+ gvt_vgpu_err("invalid gma bytes %d\n", gmadr_bytes);
+ return INTEL_GVT_INVALID_ADDR;
+ }
+
+ gma_low = cmd_val(s, index) & BATCH_BUFFER_ADDR_MASK;
+ if (gmadr_bytes == 4) {
+ addr = gma_low;
+ } else {
+ gma_high = cmd_val(s, index + 1) & BATCH_BUFFER_ADDR_HIGH_MASK;
+ addr = (((unsigned long)gma_high) << 32) | gma_low;
+ }
+ return addr;
+}
+
+static inline int cmd_address_audit(struct parser_exec_state *s,
+ unsigned long guest_gma, int op_size, bool index_mode)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ u32 max_surface_size = vgpu->gvt->device_info.max_surface_size;
+ int i;
+ int ret;
+
+ if (op_size > max_surface_size) {
+ gvt_vgpu_err("command address audit fail name %s\n",
+ s->info->name);
+ return -EFAULT;
+ }
+
+ if (index_mode) {
+ if (guest_gma >= I915_GTT_PAGE_SIZE) {
+ ret = -EFAULT;
+ goto err;
+ }
+ } else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) {
+ ret = -EFAULT;
+ goto err;
+ }
+
+ return 0;
+
+err:
+ gvt_vgpu_err("cmd_parser: Malicious %s detected, addr=0x%lx, len=%d!\n",
+ s->info->name, guest_gma, op_size);
+
+ pr_err("cmd dump: ");
+ for (i = 0; i < cmd_length(s); i++) {
+ if (!(i % 4))
+ pr_err("\n%08x ", cmd_val(s, i));
+ else
+ pr_err("%08x ", cmd_val(s, i));
+ }
+ pr_err("\nvgpu%d: aperture 0x%llx - 0x%llx, hidden 0x%llx - 0x%llx\n",
+ vgpu->id,
+ vgpu_aperture_gmadr_base(vgpu),
+ vgpu_aperture_gmadr_end(vgpu),
+ vgpu_hidden_gmadr_base(vgpu),
+ vgpu_hidden_gmadr_end(vgpu));
+ return ret;
+}
+
+static int cmd_handler_mi_store_data_imm(struct parser_exec_state *s)
+{
+ int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+ int op_size = (cmd_length(s) - 3) * sizeof(u32);
+ int core_id = (cmd_val(s, 2) & (1 << 0)) ? 1 : 0;
+ unsigned long gma, gma_low, gma_high;
+ u32 valid_len = CMD_LEN(2);
+ int ret = 0;
+
+ /* check ppggt */
+ if (!(cmd_val(s, 0) & (1 << 22)))
+ return 0;
+
+ /* check if QWORD */
+ if (DWORD_FIELD(0, 21, 21))
+ valid_len++;
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ valid_len);
+ if (ret)
+ return ret;
+
+ gma = cmd_val(s, 2) & GENMASK(31, 2);
+
+ if (gmadr_bytes == 8) {
+ gma_low = cmd_val(s, 1) & GENMASK(31, 2);
+ gma_high = cmd_val(s, 2) & GENMASK(15, 0);
+ gma = (gma_high << 32) | gma_low;
+ core_id = (cmd_val(s, 1) & (1 << 0)) ? 1 : 0;
+ }
+ ret = cmd_address_audit(s, gma + op_size * core_id, op_size, false);
+ return ret;
+}
+
+static inline int unexpected_cmd(struct parser_exec_state *s)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+
+ gvt_vgpu_err("Unexpected %s in command buffer!\n", s->info->name);
+
+ return -EBADRQC;
+}
+
+static int cmd_handler_mi_semaphore_wait(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_report_perf_count(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_op_2e(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_op_2f(struct parser_exec_state *s)
+{
+ int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+ int op_size = (1 << ((cmd_val(s, 0) & GENMASK(20, 19)) >> 19)) *
+ sizeof(u32);
+ unsigned long gma, gma_high;
+ u32 valid_len = CMD_LEN(1);
+ int ret = 0;
+
+ if (!(cmd_val(s, 0) & (1 << 22)))
+ return ret;
+
+ /* check inline data */
+ if (cmd_val(s, 0) & BIT(18))
+ valid_len = CMD_LEN(9);
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ valid_len);
+ if (ret)
+ return ret;
+
+ gma = cmd_val(s, 1) & GENMASK(31, 2);
+ if (gmadr_bytes == 8) {
+ gma_high = cmd_val(s, 2) & GENMASK(15, 0);
+ gma = (gma_high << 32) | gma;
+ }
+ ret = cmd_address_audit(s, gma, op_size, false);
+ return ret;
+}
+
+static int cmd_handler_mi_store_data_index(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_clflush(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_conditional_batch_buffer_end(
+ struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_update_gtt(struct parser_exec_state *s)
+{
+ return unexpected_cmd(s);
+}
+
+static int cmd_handler_mi_flush_dw(struct parser_exec_state *s)
+{
+ int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+ unsigned long gma;
+ bool index_mode = false;
+ int ret = 0;
+ u32 hws_pga, val;
+ u32 valid_len = CMD_LEN(2);
+
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ valid_len);
+ if (ret) {
+ /* Check again for Qword */
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ ++valid_len);
+ return ret;
+ }
+
+ /* Check post-sync and ppgtt bit */
+ if (((cmd_val(s, 0) >> 14) & 0x3) && (cmd_val(s, 1) & (1 << 2))) {
+ gma = cmd_val(s, 1) & GENMASK(31, 3);
+ if (gmadr_bytes == 8)
+ gma |= (cmd_val(s, 2) & GENMASK(15, 0)) << 32;
+ /* Store Data Index */
+ if (cmd_val(s, 0) & (1 << 21))
+ index_mode = true;
+ ret = cmd_address_audit(s, gma, sizeof(u64), index_mode);
+ if (ret)
+ return ret;
+ if (index_mode) {
+ hws_pga = s->vgpu->hws_pga[s->engine->id];
+ gma = hws_pga + gma;
+ patch_value(s, cmd_ptr(s, 1), gma);
+ val = cmd_val(s, 0) & (~(1 << 21));
+ patch_value(s, cmd_ptr(s, 0), val);
+ }
+ }
+ /* Check notify bit */
+ if ((cmd_val(s, 0) & (1 << 8)))
+ set_bit(cmd_interrupt_events[s->engine->id].mi_flush_dw,
+ s->workload->pending_events);
+ return ret;
+}
+
+static void addr_type_update_snb(struct parser_exec_state *s)
+{
+ if ((s->buf_type == RING_BUFFER_INSTRUCTION) &&
+ (BATCH_BUFFER_ADR_SPACE_BIT(cmd_val(s, 0)) == 1)) {
+ s->buf_addr_type = PPGTT_BUFFER;
+ }
+}
+
+
+static int copy_gma_to_hva(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm,
+ unsigned long gma, unsigned long end_gma, void *va)
+{
+ unsigned long copy_len, offset;
+ unsigned long len = 0;
+ unsigned long gpa;
+
+ while (gma != end_gma) {
+ gpa = intel_vgpu_gma_to_gpa(mm, gma);
+ if (gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("invalid gma address: %lx\n", gma);
+ return -EFAULT;
+ }
+
+ offset = gma & (I915_GTT_PAGE_SIZE - 1);
+
+ copy_len = (end_gma - gma) >= (I915_GTT_PAGE_SIZE - offset) ?
+ I915_GTT_PAGE_SIZE - offset : end_gma - gma;
+
+ intel_gvt_read_gpa(vgpu, gpa, va + len, copy_len);
+
+ len += copy_len;
+ gma += copy_len;
+ }
+ return len;
+}
+
+
+/*
+ * Check whether a batch buffer needs to be scanned. Currently
+ * the only criteria is based on privilege.
+ */
+static int batch_buffer_needs_scan(struct parser_exec_state *s)
+{
+ /* Decide privilege based on address space */
+ if (cmd_val(s, 0) & BIT(8) &&
+ !(s->vgpu->scan_nonprivbb & s->engine->mask))
+ return 0;
+
+ return 1;
+}
+
+static const char *repr_addr_type(unsigned int type)
+{
+ return type == PPGTT_BUFFER ? "ppgtt" : "ggtt";
+}
+
+static int find_bb_size(struct parser_exec_state *s,
+ unsigned long *bb_size,
+ unsigned long *bb_end_cmd_offset)
+{
+ unsigned long gma = 0;
+ const struct cmd_info *info;
+ u32 cmd_len = 0;
+ bool bb_end = false;
+ struct intel_vgpu *vgpu = s->vgpu;
+ u32 cmd;
+ struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
+ s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
+
+ *bb_size = 0;
+ *bb_end_cmd_offset = 0;
+
+ /* get the start gm address of the batch buffer */
+ gma = get_gma_bb_from_cmd(s, 1);
+ if (gma == INTEL_GVT_INVALID_ADDR)
+ return -EFAULT;
+
+ cmd = cmd_val(s, 0);
+ info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
+ if (info == NULL) {
+ gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
+ cmd, get_opcode(cmd, s->engine),
+ repr_addr_type(s->buf_addr_type),
+ s->engine->name, s->workload);
+ return -EBADRQC;
+ }
+ do {
+ if (copy_gma_to_hva(s->vgpu, mm,
+ gma, gma + 4, &cmd) < 0)
+ return -EFAULT;
+ info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
+ if (info == NULL) {
+ gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
+ cmd, get_opcode(cmd, s->engine),
+ repr_addr_type(s->buf_addr_type),
+ s->engine->name, s->workload);
+ return -EBADRQC;
+ }
+
+ if (info->opcode == OP_MI_BATCH_BUFFER_END) {
+ bb_end = true;
+ } else if (info->opcode == OP_MI_BATCH_BUFFER_START) {
+ if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)
+ /* chained batch buffer */
+ bb_end = true;
+ }
+
+ if (bb_end)
+ *bb_end_cmd_offset = *bb_size;
+
+ cmd_len = get_cmd_length(info, cmd) << 2;
+ *bb_size += cmd_len;
+ gma += cmd_len;
+ } while (!bb_end);
+
+ return 0;
+}
+
+static int audit_bb_end(struct parser_exec_state *s, void *va)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ u32 cmd = *(u32 *)va;
+ const struct cmd_info *info;
+
+ info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
+ if (info == NULL) {
+ gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
+ cmd, get_opcode(cmd, s->engine),
+ repr_addr_type(s->buf_addr_type),
+ s->engine->name, s->workload);
+ return -EBADRQC;
+ }
+
+ if ((info->opcode == OP_MI_BATCH_BUFFER_END) ||
+ ((info->opcode == OP_MI_BATCH_BUFFER_START) &&
+ (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)))
+ return 0;
+
+ return -EBADRQC;
+}
+
+static int perform_bb_shadow(struct parser_exec_state *s)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ struct intel_vgpu_shadow_bb *bb;
+ unsigned long gma = 0;
+ unsigned long bb_size;
+ unsigned long bb_end_cmd_offset;
+ int ret = 0;
+ struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
+ s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
+ unsigned long start_offset = 0;
+
+ /* get the start gm address of the batch buffer */
+ gma = get_gma_bb_from_cmd(s, 1);
+ if (gma == INTEL_GVT_INVALID_ADDR)
+ return -EFAULT;
+
+ ret = find_bb_size(s, &bb_size, &bb_end_cmd_offset);
+ if (ret)
+ return ret;
+
+ bb = kzalloc(sizeof(*bb), GFP_KERNEL);
+ if (!bb)
+ return -ENOMEM;
+
+ bb->ppgtt = (s->buf_addr_type == GTT_BUFFER) ? false : true;
+
+ /* the start_offset stores the batch buffer's start gma's
+ * offset relative to page boundary. so for non-privileged batch
+ * buffer, the shadowed gem object holds exactly the same page
+ * layout as original gem object. This is for the convience of
+ * replacing the whole non-privilged batch buffer page to this
+ * shadowed one in PPGTT at the same gma address. (this replacing
+ * action is not implemented yet now, but may be necessary in
+ * future).
+ * for prileged batch buffer, we just change start gma address to
+ * that of shadowed page.
+ */
+ if (bb->ppgtt)
+ start_offset = gma & ~I915_GTT_PAGE_MASK;
+
+ bb->obj = i915_gem_object_create_shmem(s->engine->i915,
+ round_up(bb_size + start_offset,
+ PAGE_SIZE));
+ if (IS_ERR(bb->obj)) {
+ ret = PTR_ERR(bb->obj);
+ goto err_free_bb;
+ }
+
+ bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB);
+ if (IS_ERR(bb->va)) {
+ ret = PTR_ERR(bb->va);
+ goto err_free_obj;
+ }
+
+ ret = copy_gma_to_hva(s->vgpu, mm,
+ gma, gma + bb_size,
+ bb->va + start_offset);
+ if (ret < 0) {
+ gvt_vgpu_err("fail to copy guest ring buffer\n");
+ ret = -EFAULT;
+ goto err_unmap;
+ }
+
+ ret = audit_bb_end(s, bb->va + start_offset + bb_end_cmd_offset);
+ if (ret)
+ goto err_unmap;
+
+ i915_gem_object_unlock(bb->obj);
+ INIT_LIST_HEAD(&bb->list);
+ list_add(&bb->list, &s->workload->shadow_bb);
+
+ bb->bb_start_cmd_va = s->ip_va;
+
+ if ((s->buf_type == BATCH_BUFFER_INSTRUCTION) && (!s->is_ctx_wa))
+ bb->bb_offset = s->ip_va - s->rb_va;
+ else
+ bb->bb_offset = 0;
+
+ /*
+ * ip_va saves the virtual address of the shadow batch buffer, while
+ * ip_gma saves the graphics address of the original batch buffer.
+ * As the shadow batch buffer is just a copy from the originial one,
+ * it should be right to use shadow batch buffer'va and original batch
+ * buffer's gma in pair. After all, we don't want to pin the shadow
+ * buffer here (too early).
+ */
+ s->ip_va = bb->va + start_offset;
+ s->ip_gma = gma;
+ return 0;
+err_unmap:
+ i915_gem_object_unpin_map(bb->obj);
+err_free_obj:
+ i915_gem_object_put(bb->obj);
+err_free_bb:
+ kfree(bb);
+ return ret;
+}
+
+static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s)
+{
+ bool second_level;
+ int ret = 0;
+ struct intel_vgpu *vgpu = s->vgpu;
+
+ if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
+ gvt_vgpu_err("Found MI_BATCH_BUFFER_START in 2nd level BB\n");
+ return -EFAULT;
+ }
+
+ second_level = BATCH_BUFFER_2ND_LEVEL_BIT(cmd_val(s, 0)) == 1;
+ if (second_level && (s->buf_type != BATCH_BUFFER_INSTRUCTION)) {
+ gvt_vgpu_err("Jumping to 2nd level BB from RB is not allowed\n");
+ return -EFAULT;
+ }
+
+ s->saved_buf_addr_type = s->buf_addr_type;
+ addr_type_update_snb(s);
+ if (s->buf_type == RING_BUFFER_INSTRUCTION) {
+ s->ret_ip_gma_ring = s->ip_gma + cmd_length(s) * sizeof(u32);
+ s->buf_type = BATCH_BUFFER_INSTRUCTION;
+ } else if (second_level) {
+ s->buf_type = BATCH_BUFFER_2ND_LEVEL;
+ s->ret_ip_gma_bb = s->ip_gma + cmd_length(s) * sizeof(u32);
+ s->ret_bb_va = s->ip_va + cmd_length(s) * sizeof(u32);
+ }
+
+ if (batch_buffer_needs_scan(s)) {
+ ret = perform_bb_shadow(s);
+ if (ret < 0)
+ gvt_vgpu_err("invalid shadow batch buffer\n");
+ } else {
+ /* emulate a batch buffer end to do return right */
+ ret = cmd_handler_mi_batch_buffer_end(s);
+ if (ret < 0)
+ return ret;
+ }
+ return ret;
+}
+
+static int mi_noop_index;
+
+static const struct cmd_info cmd_info[] = {
+ {"MI_NOOP", OP_MI_NOOP, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
+
+ {"MI_SET_PREDICATE", OP_MI_SET_PREDICATE, F_LEN_CONST, R_ALL, D_ALL,
+ 0, 1, NULL},
+
+ {"MI_USER_INTERRUPT", OP_MI_USER_INTERRUPT, F_LEN_CONST, R_ALL, D_ALL,
+ 0, 1, cmd_handler_mi_user_interrupt},
+
+ {"MI_WAIT_FOR_EVENT", OP_MI_WAIT_FOR_EVENT, F_LEN_CONST, R_RCS | R_BCS,
+ D_ALL, 0, 1, cmd_handler_mi_wait_for_event},
+
+ {"MI_FLUSH", OP_MI_FLUSH, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
+
+ {"MI_ARB_CHECK", OP_MI_ARB_CHECK, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_RS_CONTROL", OP_MI_RS_CONTROL, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_REPORT_HEAD", OP_MI_REPORT_HEAD, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_ARB_ON_OFF", OP_MI_ARB_ON_OFF, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_URB_ATOMIC_ALLOC", OP_MI_URB_ATOMIC_ALLOC, F_LEN_CONST, R_RCS,
+ D_ALL, 0, 1, NULL},
+
+ {"MI_BATCH_BUFFER_END", OP_MI_BATCH_BUFFER_END,
+ F_IP_ADVANCE_CUSTOM | F_LEN_CONST, R_ALL, D_ALL, 0, 1,
+ cmd_handler_mi_batch_buffer_end},
+
+ {"MI_SUSPEND_FLUSH", OP_MI_SUSPEND_FLUSH, F_LEN_CONST, R_ALL, D_ALL,
+ 0, 1, NULL},
+
+ {"MI_PREDICATE", OP_MI_PREDICATE, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_TOPOLOGY_FILTER", OP_MI_TOPOLOGY_FILTER, F_LEN_CONST, R_ALL,
+ D_ALL, 0, 1, NULL},
+
+ {"MI_SET_APPID", OP_MI_SET_APPID, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_RS_CONTEXT", OP_MI_RS_CONTEXT, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
+ NULL},
+
+ {"MI_DISPLAY_FLIP", OP_MI_DISPLAY_FLIP, F_LEN_VAR,
+ R_RCS | R_BCS, D_ALL, 0, 8, cmd_handler_mi_display_flip},
+
+ {"MI_SEMAPHORE_MBOX", OP_MI_SEMAPHORE_MBOX, F_LEN_VAR | F_LEN_VAR_FIXED,
+ R_ALL, D_ALL, 0, 8, NULL, CMD_LEN(1)},
+
+ {"MI_MATH", OP_MI_MATH, F_LEN_VAR, R_ALL, D_ALL, 0, 8, NULL},
+
+ {"MI_URB_CLEAR", OP_MI_URB_CLEAR, F_LEN_VAR | F_LEN_VAR_FIXED, R_RCS,
+ D_ALL, 0, 8, NULL, CMD_LEN(0)},
+
+ {"MI_SEMAPHORE_SIGNAL", OP_MI_SEMAPHORE_SIGNAL,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS, 0, 8,
+ NULL, CMD_LEN(0)},
+
+ {"MI_SEMAPHORE_WAIT", OP_MI_SEMAPHORE_WAIT,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS, ADDR_FIX_1(2),
+ 8, cmd_handler_mi_semaphore_wait, CMD_LEN(2)},
+
+ {"MI_STORE_DATA_IMM", OP_MI_STORE_DATA_IMM, F_LEN_VAR, R_ALL, D_BDW_PLUS,
+ ADDR_FIX_1(1), 10, cmd_handler_mi_store_data_imm},
+
+ {"MI_STORE_DATA_INDEX", OP_MI_STORE_DATA_INDEX, F_LEN_VAR, R_ALL, D_ALL,
+ 0, 8, cmd_handler_mi_store_data_index},
+
+ {"MI_LOAD_REGISTER_IMM", OP_MI_LOAD_REGISTER_IMM, F_LEN_VAR, R_ALL,
+ D_ALL, 0, 8, cmd_handler_lri},
+
+ {"MI_UPDATE_GTT", OP_MI_UPDATE_GTT, F_LEN_VAR, R_ALL, D_BDW_PLUS, 0, 10,
+ cmd_handler_mi_update_gtt},
+
+ {"MI_STORE_REGISTER_MEM", OP_MI_STORE_REGISTER_MEM,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
+ cmd_handler_srm, CMD_LEN(2)},
+
+ {"MI_FLUSH_DW", OP_MI_FLUSH_DW, F_LEN_VAR, R_ALL, D_ALL, 0, 6,
+ cmd_handler_mi_flush_dw},
+
+ {"MI_CLFLUSH", OP_MI_CLFLUSH, F_LEN_VAR, R_ALL, D_ALL, ADDR_FIX_1(1),
+ 10, cmd_handler_mi_clflush},
+
+ {"MI_REPORT_PERF_COUNT", OP_MI_REPORT_PERF_COUNT,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(1), 6,
+ cmd_handler_mi_report_perf_count, CMD_LEN(2)},
+
+ {"MI_LOAD_REGISTER_MEM", OP_MI_LOAD_REGISTER_MEM,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
+ cmd_handler_lrm, CMD_LEN(2)},
+
+ {"MI_LOAD_REGISTER_REG", OP_MI_LOAD_REGISTER_REG,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, 0, 8,
+ cmd_handler_lrr, CMD_LEN(1)},
+
+ {"MI_RS_STORE_DATA_IMM", OP_MI_RS_STORE_DATA_IMM,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_RCS, D_ALL, 0,
+ 8, NULL, CMD_LEN(2)},
+
+ {"MI_LOAD_URB_MEM", OP_MI_LOAD_URB_MEM, F_LEN_VAR | F_LEN_VAR_FIXED,
+ R_RCS, D_ALL, ADDR_FIX_1(2), 8, NULL, CMD_LEN(2)},
+
+ {"MI_STORE_URM_MEM", OP_MI_STORE_URM_MEM, F_LEN_VAR, R_RCS, D_ALL,
+ ADDR_FIX_1(2), 8, NULL},
+
+ {"MI_OP_2E", OP_MI_2E, F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS,
+ ADDR_FIX_2(1, 2), 8, cmd_handler_mi_op_2e, CMD_LEN(3)},
+
+ {"MI_OP_2F", OP_MI_2F, F_LEN_VAR, R_ALL, D_BDW_PLUS, ADDR_FIX_1(1),
+ 8, cmd_handler_mi_op_2f},
+
+ {"MI_BATCH_BUFFER_START", OP_MI_BATCH_BUFFER_START,
+ F_IP_ADVANCE_CUSTOM, R_ALL, D_ALL, 0, 8,
+ cmd_handler_mi_batch_buffer_start},
+
+ {"MI_CONDITIONAL_BATCH_BUFFER_END", OP_MI_CONDITIONAL_BATCH_BUFFER_END,
+ F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
+ cmd_handler_mi_conditional_batch_buffer_end, CMD_LEN(2)},
+
+ {"MI_LOAD_SCAN_LINES_INCL", OP_MI_LOAD_SCAN_LINES_INCL, F_LEN_CONST,
+ R_RCS | R_BCS, D_ALL, 0, 2, NULL},
+
+ {"XY_SETUP_BLT", OP_XY_SETUP_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_2(4, 7), 8, NULL},
+
+ {"XY_SETUP_CLIP_BLT", OP_XY_SETUP_CLIP_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ 0, 8, NULL},
+
+ {"XY_SETUP_MONO_PATTERN_SL_BLT", OP_XY_SETUP_MONO_PATTERN_SL_BLT,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_PIXEL_BLT", OP_XY_PIXEL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
+
+ {"XY_SCANLINES_BLT", OP_XY_SCANLINES_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ 0, 8, NULL},
+
+ {"XY_TEXT_BLT", OP_XY_TEXT_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_1(3), 8, NULL},
+
+ {"XY_TEXT_IMMEDIATE_BLT", OP_XY_TEXT_IMMEDIATE_BLT, F_LEN_VAR, R_BCS,
+ D_ALL, 0, 8, NULL},
+
+ {"XY_COLOR_BLT", OP_XY_COLOR_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_PAT_BLT", OP_XY_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_2(4, 5), 8, NULL},
+
+ {"XY_MONO_PAT_BLT", OP_XY_MONO_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_SRC_COPY_BLT", OP_XY_SRC_COPY_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_2(4, 7), 8, NULL},
+
+ {"XY_MONO_SRC_COPY_BLT", OP_XY_MONO_SRC_COPY_BLT, F_LEN_VAR, R_BCS,
+ D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
+
+ {"XY_FULL_BLT", OP_XY_FULL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
+
+ {"XY_FULL_MONO_SRC_BLT", OP_XY_FULL_MONO_SRC_BLT, F_LEN_VAR, R_BCS,
+ D_ALL, ADDR_FIX_3(4, 5, 8), 8, NULL},
+
+ {"XY_FULL_MONO_PATTERN_BLT", OP_XY_FULL_MONO_PATTERN_BLT, F_LEN_VAR,
+ R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
+
+ {"XY_FULL_MONO_PATTERN_MONO_SRC_BLT",
+ OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
+
+ {"XY_MONO_PAT_FIXED_BLT", OP_XY_MONO_PAT_FIXED_BLT, F_LEN_VAR, R_BCS,
+ D_ALL, ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_MONO_SRC_COPY_IMMEDIATE_BLT", OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_PAT_BLT_IMMEDIATE", OP_XY_PAT_BLT_IMMEDIATE, F_LEN_VAR, R_BCS,
+ D_ALL, ADDR_FIX_1(4), 8, NULL},
+
+ {"XY_SRC_COPY_CHROMA_BLT", OP_XY_SRC_COPY_CHROMA_BLT, F_LEN_VAR, R_BCS,
+ D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
+
+ {"XY_FULL_IMMEDIATE_PATTERN_BLT", OP_XY_FULL_IMMEDIATE_PATTERN_BLT,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
+
+ {"XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT",
+ OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
+
+ {"XY_PAT_CHROMA_BLT", OP_XY_PAT_CHROMA_BLT, F_LEN_VAR, R_BCS, D_ALL,
+ ADDR_FIX_2(4, 5), 8, NULL},
+
+ {"XY_PAT_CHROMA_BLT_IMMEDIATE", OP_XY_PAT_CHROMA_BLT_IMMEDIATE,
+ F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
+
+ {"3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP",
+ OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_VIEWPORT_STATE_POINTERS_CC",
+ OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BLEND_STATE_POINTERS",
+ OP_3DSTATE_BLEND_STATE_POINTERS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DEPTH_STENCIL_STATE_POINTERS",
+ OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POINTERS_VS",
+ OP_3DSTATE_BINDING_TABLE_POINTERS_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POINTERS_HS",
+ OP_3DSTATE_BINDING_TABLE_POINTERS_HS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POINTERS_DS",
+ OP_3DSTATE_BINDING_TABLE_POINTERS_DS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POINTERS_GS",
+ OP_3DSTATE_BINDING_TABLE_POINTERS_GS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POINTERS_PS",
+ OP_3DSTATE_BINDING_TABLE_POINTERS_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_STATE_POINTERS_VS",
+ OP_3DSTATE_SAMPLER_STATE_POINTERS_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_STATE_POINTERS_HS",
+ OP_3DSTATE_SAMPLER_STATE_POINTERS_HS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_STATE_POINTERS_DS",
+ OP_3DSTATE_SAMPLER_STATE_POINTERS_DS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_STATE_POINTERS_GS",
+ OP_3DSTATE_SAMPLER_STATE_POINTERS_GS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_STATE_POINTERS_PS",
+ OP_3DSTATE_SAMPLER_STATE_POINTERS_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_URB_VS", OP_3DSTATE_URB_VS, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"3DSTATE_URB_HS", OP_3DSTATE_URB_HS, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"3DSTATE_URB_DS", OP_3DSTATE_URB_DS, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"3DSTATE_URB_GS", OP_3DSTATE_URB_GS, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"3DSTATE_GATHER_CONSTANT_VS", OP_3DSTATE_GATHER_CONSTANT_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GATHER_CONSTANT_GS", OP_3DSTATE_GATHER_CONSTANT_GS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GATHER_CONSTANT_HS", OP_3DSTATE_GATHER_CONSTANT_HS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GATHER_CONSTANT_DS", OP_3DSTATE_GATHER_CONSTANT_DS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GATHER_CONSTANT_PS", OP_3DSTATE_GATHER_CONSTANT_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_CONSTANTF_VS", OP_3DSTATE_DX9_CONSTANTF_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
+
+ {"3DSTATE_DX9_CONSTANTF_PS", OP_3DSTATE_DX9_CONSTANTF_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
+
+ {"3DSTATE_DX9_CONSTANTI_VS", OP_3DSTATE_DX9_CONSTANTI_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_CONSTANTI_PS", OP_3DSTATE_DX9_CONSTANTI_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_CONSTANTB_VS", OP_3DSTATE_DX9_CONSTANTB_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_CONSTANTB_PS", OP_3DSTATE_DX9_CONSTANTB_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_LOCAL_VALID_VS", OP_3DSTATE_DX9_LOCAL_VALID_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_LOCAL_VALID_PS", OP_3DSTATE_DX9_LOCAL_VALID_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_GENERATE_ACTIVE_VS", OP_3DSTATE_DX9_GENERATE_ACTIVE_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DX9_GENERATE_ACTIVE_PS", OP_3DSTATE_DX9_GENERATE_ACTIVE_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_EDIT_VS", OP_3DSTATE_BINDING_TABLE_EDIT_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_BINDING_TABLE_EDIT_GS", OP_3DSTATE_BINDING_TABLE_EDIT_GS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_BINDING_TABLE_EDIT_HS", OP_3DSTATE_BINDING_TABLE_EDIT_HS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_BINDING_TABLE_EDIT_DS", OP_3DSTATE_BINDING_TABLE_EDIT_DS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_BINDING_TABLE_EDIT_PS", OP_3DSTATE_BINDING_TABLE_EDIT_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_VF_INSTANCING", OP_3DSTATE_VF_INSTANCING, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_VF_SGVS", OP_3DSTATE_VF_SGVS, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
+ NULL},
+
+ {"3DSTATE_VF_TOPOLOGY", OP_3DSTATE_VF_TOPOLOGY, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_WM_CHROMAKEY", OP_3DSTATE_WM_CHROMAKEY, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_PS_BLEND", OP_3DSTATE_PS_BLEND, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
+ 8, NULL},
+
+ {"3DSTATE_WM_DEPTH_STENCIL", OP_3DSTATE_WM_DEPTH_STENCIL, F_LEN_VAR,
+ R_RCS, D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_PS_EXTRA", OP_3DSTATE_PS_EXTRA, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
+ 8, NULL},
+
+ {"3DSTATE_RASTER", OP_3DSTATE_RASTER, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
+ NULL},
+
+ {"3DSTATE_SBE_SWIZ", OP_3DSTATE_SBE_SWIZ, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
+ NULL},
+
+ {"3DSTATE_WM_HZ_OP", OP_3DSTATE_WM_HZ_OP, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
+ NULL},
+
+ {"3DSTATE_VERTEX_BUFFERS", OP_3DSTATE_VERTEX_BUFFERS, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_VERTEX_ELEMENTS", OP_3DSTATE_VERTEX_ELEMENTS, F_LEN_VAR,
+ R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_INDEX_BUFFER", OP_3DSTATE_INDEX_BUFFER, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, ADDR_FIX_1(2), 8, NULL},
+
+ {"3DSTATE_VF_STATISTICS", OP_3DSTATE_VF_STATISTICS, F_LEN_CONST,
+ R_RCS, D_ALL, 0, 1, NULL},
+
+ {"3DSTATE_VF", OP_3DSTATE_VF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CC_STATE_POINTERS", OP_3DSTATE_CC_STATE_POINTERS, F_LEN_VAR,
+ R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SCISSOR_STATE_POINTERS", OP_3DSTATE_SCISSOR_STATE_POINTERS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GS", OP_3DSTATE_GS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CLIP", OP_3DSTATE_CLIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_WM", OP_3DSTATE_WM, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CONSTANT_GS", OP_3DSTATE_CONSTANT_GS, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_CONSTANT_PS", OP_3DSTATE_CONSTANT_PS, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLE_MASK", OP_3DSTATE_SAMPLE_MASK, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CONSTANT_HS", OP_3DSTATE_CONSTANT_HS, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_CONSTANT_DS", OP_3DSTATE_CONSTANT_DS, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_HS", OP_3DSTATE_HS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_TE", OP_3DSTATE_TE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DS", OP_3DSTATE_DS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_STREAMOUT", OP_3DSTATE_STREAMOUT, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SBE", OP_3DSTATE_SBE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PS", OP_3DSTATE_PS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_DRAWING_RECTANGLE", OP_3DSTATE_DRAWING_RECTANGLE, F_LEN_VAR,
+ R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_PALETTE_LOAD0", OP_3DSTATE_SAMPLER_PALETTE_LOAD0,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CHROMA_KEY", OP_3DSTATE_CHROMA_KEY, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"3DSTATE_DEPTH_BUFFER", OP_3DSTATE_DEPTH_BUFFER, F_LEN_VAR, R_RCS,
+ D_ALL, ADDR_FIX_1(2), 8, NULL},
+
+ {"3DSTATE_POLY_STIPPLE_OFFSET", OP_3DSTATE_POLY_STIPPLE_OFFSET,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_POLY_STIPPLE_PATTERN", OP_3DSTATE_POLY_STIPPLE_PATTERN,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_LINE_STIPPLE", OP_3DSTATE_LINE_STIPPLE, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_AA_LINE_PARAMS", OP_3DSTATE_AA_LINE_PARAMS, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_GS_SVB_INDEX", OP_3DSTATE_GS_SVB_INDEX, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SAMPLER_PALETTE_LOAD1", OP_3DSTATE_SAMPLER_PALETTE_LOAD1,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_MULTISAMPLE", OP_3DSTATE_MULTISAMPLE_BDW, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"3DSTATE_STENCIL_BUFFER", OP_3DSTATE_STENCIL_BUFFER, F_LEN_VAR, R_RCS,
+ D_ALL, ADDR_FIX_1(2), 8, NULL},
+
+ {"3DSTATE_HIER_DEPTH_BUFFER", OP_3DSTATE_HIER_DEPTH_BUFFER, F_LEN_VAR,
+ R_RCS, D_ALL, ADDR_FIX_1(2), 8, NULL},
+
+ {"3DSTATE_CLEAR_PARAMS", OP_3DSTATE_CLEAR_PARAMS, F_LEN_VAR,
+ R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PUSH_CONSTANT_ALLOC_VS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PUSH_CONSTANT_ALLOC_HS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PUSH_CONSTANT_ALLOC_DS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PUSH_CONSTANT_ALLOC_GS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_PUSH_CONSTANT_ALLOC_PS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_MONOFILTER_SIZE", OP_3DSTATE_MONOFILTER_SIZE, F_LEN_VAR,
+ R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SO_DECL_LIST", OP_3DSTATE_SO_DECL_LIST, F_LEN_VAR, R_RCS,
+ D_ALL, 0, 9, NULL},
+
+ {"3DSTATE_SO_BUFFER", OP_3DSTATE_SO_BUFFER, F_LEN_VAR, R_RCS, D_BDW_PLUS,
+ ADDR_FIX_2(2, 4), 8, NULL},
+
+ {"3DSTATE_BINDING_TABLE_POOL_ALLOC",
+ OP_3DSTATE_BINDING_TABLE_POOL_ALLOC,
+ F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
+
+ {"3DSTATE_GATHER_POOL_ALLOC", OP_3DSTATE_GATHER_POOL_ALLOC,
+ F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
+
+ {"3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC",
+ OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC,
+ F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
+
+ {"3DSTATE_SAMPLE_PATTERN", OP_3DSTATE_SAMPLE_PATTERN, F_LEN_VAR, R_RCS,
+ D_BDW_PLUS, 0, 8, NULL},
+
+ {"PIPE_CONTROL", OP_PIPE_CONTROL, F_LEN_VAR, R_RCS, D_ALL,
+ ADDR_FIX_1(2), 8, cmd_handler_pipe_control},
+
+ {"3DPRIMITIVE", OP_3DPRIMITIVE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"PIPELINE_SELECT", OP_PIPELINE_SELECT, F_LEN_CONST, R_RCS, D_ALL, 0,
+ 1, NULL},
+
+ {"STATE_PREFETCH", OP_STATE_PREFETCH, F_LEN_VAR, R_RCS, D_ALL,
+ ADDR_FIX_1(1), 8, NULL},
+
+ {"STATE_SIP", OP_STATE_SIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"STATE_BASE_ADDRESS", OP_STATE_BASE_ADDRESS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
+ ADDR_FIX_5(1, 3, 4, 5, 6), 8, NULL},
+
+ {"OP_3D_MEDIA_0_1_4", OP_3D_MEDIA_0_1_4, F_LEN_VAR, R_RCS, D_ALL,
+ ADDR_FIX_1(1), 8, NULL},
+
+ {"OP_SWTESS_BASE_ADDRESS", OP_SWTESS_BASE_ADDRESS,
+ F_LEN_VAR, R_RCS, D_ALL, ADDR_FIX_2(1, 2), 3, NULL},
+
+ {"3DSTATE_VS", OP_3DSTATE_VS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_SF", OP_3DSTATE_SF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
+
+ {"3DSTATE_CONSTANT_VS", OP_3DSTATE_CONSTANT_VS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
+ 0, 8, NULL},
+
+ {"3DSTATE_COMPONENT_PACKING", OP_3DSTATE_COMPONENT_PACKING, F_LEN_VAR, R_RCS,
+ D_SKL_PLUS, 0, 8, NULL},
+
+ {"MEDIA_INTERFACE_DESCRIPTOR_LOAD", OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD,
+ F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
+
+ {"MEDIA_GATEWAY_STATE", OP_MEDIA_GATEWAY_STATE, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MEDIA_STATE_FLUSH", OP_MEDIA_STATE_FLUSH, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MEDIA_POOL_STATE", OP_MEDIA_POOL_STATE, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MEDIA_OBJECT", OP_MEDIA_OBJECT, F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
+
+ {"MEDIA_CURBE_LOAD", OP_MEDIA_CURBE_LOAD, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MEDIA_OBJECT_PRT", OP_MEDIA_OBJECT_PRT, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MEDIA_OBJECT_WALKER", OP_MEDIA_OBJECT_WALKER, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 16, NULL},
+
+ {"GPGPU_WALKER", OP_GPGPU_WALKER, F_LEN_VAR, R_RCS, D_ALL,
+ 0, 8, NULL},
+
+ {"MEDIA_VFE_STATE", OP_MEDIA_VFE_STATE, F_LEN_VAR, R_RCS, D_ALL, 0, 16,
+ NULL},
+
+ {"3DSTATE_VF_STATISTICS_GM45", OP_3DSTATE_VF_STATISTICS_GM45,
+ F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
+
+ {"MFX_PIPE_MODE_SELECT", OP_MFX_PIPE_MODE_SELECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_SURFACE_STATE", OP_MFX_SURFACE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_PIPE_BUF_ADDR_STATE", OP_MFX_PIPE_BUF_ADDR_STATE, F_LEN_VAR,
+ R_VCS, D_BDW_PLUS, 0, 12, NULL},
+
+ {"MFX_IND_OBJ_BASE_ADDR_STATE", OP_MFX_IND_OBJ_BASE_ADDR_STATE,
+ F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
+
+ {"MFX_BSP_BUF_BASE_ADDR_STATE", OP_MFX_BSP_BUF_BASE_ADDR_STATE,
+ F_LEN_VAR, R_VCS, D_BDW_PLUS, ADDR_FIX_3(1, 3, 5), 12, NULL},
+
+ {"OP_2_0_0_5", OP_2_0_0_5, F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
+
+ {"MFX_STATE_POINTER", OP_MFX_STATE_POINTER, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_QM_STATE", OP_MFX_QM_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_FQM_STATE", OP_MFX_FQM_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_PAK_INSERT_OBJECT", OP_MFX_PAK_INSERT_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_STITCH_OBJECT", OP_MFX_STITCH_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_IT_OBJECT", OP_MFD_IT_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_WAIT", OP_MFX_WAIT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 6, NULL},
+
+ {"MFX_AVC_IMG_STATE", OP_MFX_AVC_IMG_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_AVC_QM_STATE", OP_MFX_AVC_QM_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_AVC_DIRECTMODE_STATE", OP_MFX_AVC_DIRECTMODE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_AVC_SLICE_STATE", OP_MFX_AVC_SLICE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_AVC_REF_IDX_STATE", OP_MFX_AVC_REF_IDX_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_AVC_WEIGHTOFFSET_STATE", OP_MFX_AVC_WEIGHTOFFSET_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_AVC_PICID_STATE", OP_MFD_AVC_PICID_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+ {"MFD_AVC_DPB_STATE", OP_MFD_AVC_DPB_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_AVC_BSD_OBJECT", OP_MFD_AVC_BSD_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_AVC_SLICEADDR", OP_MFD_AVC_SLICEADDR, F_LEN_VAR,
+ R_VCS, D_ALL, ADDR_FIX_1(2), 12, NULL},
+
+ {"MFC_AVC_PAK_OBJECT", OP_MFC_AVC_PAK_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_VC1_PRED_PIPE_STATE", OP_MFX_VC1_PRED_PIPE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_VC1_DIRECTMODE_STATE", OP_MFX_VC1_DIRECTMODE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_VC1_SHORT_PIC_STATE", OP_MFD_VC1_SHORT_PIC_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_VC1_LONG_PIC_STATE", OP_MFD_VC1_LONG_PIC_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_VC1_BSD_OBJECT", OP_MFD_VC1_BSD_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFC_MPEG2_SLICEGROUP_STATE", OP_MFC_MPEG2_SLICEGROUP_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFC_MPEG2_PAK_OBJECT", OP_MFC_MPEG2_PAK_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_MPEG2_PIC_STATE", OP_MFX_MPEG2_PIC_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_MPEG2_QM_STATE", OP_MFX_MPEG2_QM_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_MPEG2_BSD_OBJECT", OP_MFD_MPEG2_BSD_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_2_6_0_0", OP_MFX_2_6_0_0, F_LEN_VAR, R_VCS, D_ALL,
+ 0, 16, NULL},
+
+ {"MFX_2_6_0_9", OP_MFX_2_6_0_9, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
+
+ {"MFX_2_6_0_8", OP_MFX_2_6_0_8, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
+
+ {"MFX_JPEG_PIC_STATE", OP_MFX_JPEG_PIC_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFX_JPEG_HUFF_TABLE_STATE", OP_MFX_JPEG_HUFF_TABLE_STATE, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"MFD_JPEG_BSD_OBJECT", OP_MFD_JPEG_BSD_OBJECT, F_LEN_VAR,
+ R_VCS, D_ALL, 0, 12, NULL},
+
+ {"VEBOX_STATE", OP_VEB_STATE, F_LEN_VAR, R_VECS, D_ALL, 0, 12, NULL},
+
+ {"VEBOX_SURFACE_STATE", OP_VEB_SURFACE_STATE, F_LEN_VAR, R_VECS, D_ALL,
+ 0, 12, NULL},
+
+ {"VEB_DI_IECP", OP_VEB_DNDI_IECP_STATE, F_LEN_VAR, R_VECS, D_BDW_PLUS,
+ 0, 12, NULL},
+};
+
+static void add_cmd_entry(struct intel_gvt *gvt, struct cmd_entry *e)
+{
+ hash_add(gvt->cmd_table, &e->hlist, e->info->opcode);
+}
+
+/* call the cmd handler, and advance ip */
+static int cmd_parser_exec(struct parser_exec_state *s)
+{
+ struct intel_vgpu *vgpu = s->vgpu;
+ const struct cmd_info *info;
+ u32 cmd;
+ int ret = 0;
+
+ cmd = cmd_val(s, 0);
+
+ /* fastpath for MI_NOOP */
+ if (cmd == MI_NOOP)
+ info = &cmd_info[mi_noop_index];
+ else
+ info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
+
+ if (info == NULL) {
+ gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
+ cmd, get_opcode(cmd, s->engine),
+ repr_addr_type(s->buf_addr_type),
+ s->engine->name, s->workload);
+ return -EBADRQC;
+ }
+
+ s->info = info;
+
+ trace_gvt_command(vgpu->id, s->engine->id, s->ip_gma, s->ip_va,
+ cmd_length(s), s->buf_type, s->buf_addr_type,
+ s->workload, info->name);
+
+ if ((info->flag & F_LEN_MASK) == F_LEN_VAR_FIXED) {
+ ret = gvt_check_valid_cmd_length(cmd_length(s),
+ info->valid_len);
+ if (ret)
+ return ret;
+ }
+
+ if (info->handler) {
+ ret = info->handler(s);
+ if (ret < 0) {
+ gvt_vgpu_err("%s handler error\n", info->name);
+ return ret;
+ }
+ }
+
+ if (!(info->flag & F_IP_ADVANCE_CUSTOM)) {
+ ret = cmd_advance_default(s);
+ if (ret) {
+ gvt_vgpu_err("%s IP advance error\n", info->name);
+ return ret;
+ }
+ }
+ return 0;
+}
+
+static inline bool gma_out_of_range(unsigned long gma,
+ unsigned long gma_head, unsigned int gma_tail)
+{
+ if (gma_tail >= gma_head)
+ return (gma < gma_head) || (gma > gma_tail);
+ else
+ return (gma > gma_tail) && (gma < gma_head);
+}
+
+/* Keep the consistent return type, e.g EBADRQC for unknown
+ * cmd, EFAULT for invalid address, EPERM for nonpriv. later
+ * works as the input of VM healthy status.
+ */
+static int command_scan(struct parser_exec_state *s,
+ unsigned long rb_head, unsigned long rb_tail,
+ unsigned long rb_start, unsigned long rb_len)
+{
+
+ unsigned long gma_head, gma_tail, gma_bottom;
+ int ret = 0;
+ struct intel_vgpu *vgpu = s->vgpu;
+
+ gma_head = rb_start + rb_head;
+ gma_tail = rb_start + rb_tail;
+ gma_bottom = rb_start + rb_len;
+
+ while (s->ip_gma != gma_tail) {
+ if (s->buf_type == RING_BUFFER_INSTRUCTION ||
+ s->buf_type == RING_BUFFER_CTX) {
+ if (!(s->ip_gma >= rb_start) ||
+ !(s->ip_gma < gma_bottom)) {
+ gvt_vgpu_err("ip_gma %lx out of ring scope."
+ "(base:0x%lx, bottom: 0x%lx)\n",
+ s->ip_gma, rb_start,
+ gma_bottom);
+ parser_exec_state_dump(s);
+ return -EFAULT;
+ }
+ if (gma_out_of_range(s->ip_gma, gma_head, gma_tail)) {
+ gvt_vgpu_err("ip_gma %lx out of range."
+ "base 0x%lx head 0x%lx tail 0x%lx\n",
+ s->ip_gma, rb_start,
+ rb_head, rb_tail);
+ parser_exec_state_dump(s);
+ break;
+ }
+ }
+ ret = cmd_parser_exec(s);
+ if (ret) {
+ gvt_vgpu_err("cmd parser error\n");
+ parser_exec_state_dump(s);
+ break;
+ }
+ }
+
+ return ret;
+}
+
+static int scan_workload(struct intel_vgpu_workload *workload)
+{
+ unsigned long gma_head, gma_tail, gma_bottom;
+ struct parser_exec_state s;
+ int ret = 0;
+
+ /* ring base is page aligned */
+ if (WARN_ON(!IS_ALIGNED(workload->rb_start, I915_GTT_PAGE_SIZE)))
+ return -EINVAL;
+
+ gma_head = workload->rb_start + workload->rb_head;
+ gma_tail = workload->rb_start + workload->rb_tail;
+ gma_bottom = workload->rb_start + _RING_CTL_BUF_SIZE(workload->rb_ctl);
+
+ s.buf_type = RING_BUFFER_INSTRUCTION;
+ s.buf_addr_type = GTT_BUFFER;
+ s.vgpu = workload->vgpu;
+ s.engine = workload->engine;
+ s.ring_start = workload->rb_start;
+ s.ring_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
+ s.ring_head = gma_head;
+ s.ring_tail = gma_tail;
+ s.rb_va = workload->shadow_ring_buffer_va;
+ s.workload = workload;
+ s.is_ctx_wa = false;
+
+ if (bypass_scan_mask & workload->engine->mask || gma_head == gma_tail)
+ return 0;
+
+ ret = ip_gma_set(&s, gma_head);
+ if (ret)
+ goto out;
+
+ ret = command_scan(&s, workload->rb_head, workload->rb_tail,
+ workload->rb_start, _RING_CTL_BUF_SIZE(workload->rb_ctl));
+
+out:
+ return ret;
+}
+
+static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+
+ unsigned long gma_head, gma_tail, gma_bottom, ring_size, ring_tail;
+ struct parser_exec_state s;
+ int ret = 0;
+ struct intel_vgpu_workload *workload = container_of(wa_ctx,
+ struct intel_vgpu_workload,
+ wa_ctx);
+
+ /* ring base is page aligned */
+ if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma,
+ I915_GTT_PAGE_SIZE)))
+ return -EINVAL;
+
+ ring_tail = wa_ctx->indirect_ctx.size + 3 * sizeof(u32);
+ ring_size = round_up(wa_ctx->indirect_ctx.size + CACHELINE_BYTES,
+ PAGE_SIZE);
+ gma_head = wa_ctx->indirect_ctx.guest_gma;
+ gma_tail = wa_ctx->indirect_ctx.guest_gma + ring_tail;
+ gma_bottom = wa_ctx->indirect_ctx.guest_gma + ring_size;
+
+ s.buf_type = RING_BUFFER_INSTRUCTION;
+ s.buf_addr_type = GTT_BUFFER;
+ s.vgpu = workload->vgpu;
+ s.engine = workload->engine;
+ s.ring_start = wa_ctx->indirect_ctx.guest_gma;
+ s.ring_size = ring_size;
+ s.ring_head = gma_head;
+ s.ring_tail = gma_tail;
+ s.rb_va = wa_ctx->indirect_ctx.shadow_va;
+ s.workload = workload;
+ s.is_ctx_wa = true;
+
+ ret = ip_gma_set(&s, gma_head);
+ if (ret)
+ goto out;
+
+ ret = command_scan(&s, 0, ring_tail,
+ wa_ctx->indirect_ctx.guest_gma, ring_size);
+out:
+ return ret;
+}
+
+static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ unsigned long gma_head, gma_tail, gma_top, guest_rb_size;
+ void *shadow_ring_buffer_va;
+ int ret;
+
+ guest_rb_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
+
+ /* calculate workload ring buffer size */
+ workload->rb_len = (workload->rb_tail + guest_rb_size -
+ workload->rb_head) % guest_rb_size;
+
+ gma_head = workload->rb_start + workload->rb_head;
+ gma_tail = workload->rb_start + workload->rb_tail;
+ gma_top = workload->rb_start + guest_rb_size;
+
+ if (workload->rb_len > s->ring_scan_buffer_size[workload->engine->id]) {
+ void *p;
+
+ /* realloc the new ring buffer if needed */
+ p = krealloc(s->ring_scan_buffer[workload->engine->id],
+ workload->rb_len, GFP_KERNEL);
+ if (!p) {
+ gvt_vgpu_err("fail to re-alloc ring scan buffer\n");
+ return -ENOMEM;
+ }
+ s->ring_scan_buffer[workload->engine->id] = p;
+ s->ring_scan_buffer_size[workload->engine->id] = workload->rb_len;
+ }
+
+ shadow_ring_buffer_va = s->ring_scan_buffer[workload->engine->id];
+
+ /* get shadow ring buffer va */
+ workload->shadow_ring_buffer_va = shadow_ring_buffer_va;
+
+ /* head > tail --> copy head <-> top */
+ if (gma_head > gma_tail) {
+ ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm,
+ gma_head, gma_top, shadow_ring_buffer_va);
+ if (ret < 0) {
+ gvt_vgpu_err("fail to copy guest ring buffer\n");
+ return ret;
+ }
+ shadow_ring_buffer_va += ret;
+ gma_head = workload->rb_start;
+ }
+
+ /* copy head or start <-> tail */
+ ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm, gma_head, gma_tail,
+ shadow_ring_buffer_va);
+ if (ret < 0) {
+ gvt_vgpu_err("fail to copy guest ring buffer\n");
+ return ret;
+ }
+ return 0;
+}
+
+int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload)
+{
+ int ret;
+ struct intel_vgpu *vgpu = workload->vgpu;
+
+ ret = shadow_workload_ring_buffer(workload);
+ if (ret) {
+ gvt_vgpu_err("fail to shadow workload ring_buffer\n");
+ return ret;
+ }
+
+ ret = scan_workload(workload);
+ if (ret) {
+ gvt_vgpu_err("scan workload error\n");
+ return ret;
+ }
+ return 0;
+}
+
+static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ int ctx_size = wa_ctx->indirect_ctx.size;
+ unsigned long guest_gma = wa_ctx->indirect_ctx.guest_gma;
+ struct intel_vgpu_workload *workload = container_of(wa_ctx,
+ struct intel_vgpu_workload,
+ wa_ctx);
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct drm_i915_gem_object *obj;
+ int ret = 0;
+ void *map;
+
+ obj = i915_gem_object_create_shmem(workload->engine->i915,
+ roundup(ctx_size + CACHELINE_BYTES,
+ PAGE_SIZE));
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
+
+ /* get the va of the shadow batch buffer */
+ map = i915_gem_object_pin_map(obj, I915_MAP_WB);
+ if (IS_ERR(map)) {
+ gvt_vgpu_err("failed to vmap shadow indirect ctx\n");
+ ret = PTR_ERR(map);
+ goto put_obj;
+ }
+
+ i915_gem_object_lock(obj, NULL);
+ ret = i915_gem_object_set_to_cpu_domain(obj, false);
+ i915_gem_object_unlock(obj);
+ if (ret) {
+ gvt_vgpu_err("failed to set shadow indirect ctx to CPU\n");
+ goto unmap_src;
+ }
+
+ ret = copy_gma_to_hva(workload->vgpu,
+ workload->vgpu->gtt.ggtt_mm,
+ guest_gma, guest_gma + ctx_size,
+ map);
+ if (ret < 0) {
+ gvt_vgpu_err("fail to copy guest indirect ctx\n");
+ goto unmap_src;
+ }
+
+ wa_ctx->indirect_ctx.obj = obj;
+ wa_ctx->indirect_ctx.shadow_va = map;
+ return 0;
+
+unmap_src:
+ i915_gem_object_unpin_map(obj);
+put_obj:
+ i915_gem_object_put(obj);
+ return ret;
+}
+
+static int combine_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ u32 per_ctx_start[CACHELINE_DWORDS] = {0};
+ unsigned char *bb_start_sva;
+
+ if (!wa_ctx->per_ctx.valid)
+ return 0;
+
+ per_ctx_start[0] = 0x18800001;
+ per_ctx_start[1] = wa_ctx->per_ctx.guest_gma;
+
+ bb_start_sva = (unsigned char *)wa_ctx->indirect_ctx.shadow_va +
+ wa_ctx->indirect_ctx.size;
+
+ memcpy(bb_start_sva, per_ctx_start, CACHELINE_BYTES);
+
+ return 0;
+}
+
+int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ int ret;
+ struct intel_vgpu_workload *workload = container_of(wa_ctx,
+ struct intel_vgpu_workload,
+ wa_ctx);
+ struct intel_vgpu *vgpu = workload->vgpu;
+
+ if (wa_ctx->indirect_ctx.size == 0)
+ return 0;
+
+ ret = shadow_indirect_ctx(wa_ctx);
+ if (ret) {
+ gvt_vgpu_err("fail to shadow indirect ctx\n");
+ return ret;
+ }
+
+ combine_wa_ctx(wa_ctx);
+
+ ret = scan_wa_ctx(wa_ctx);
+ if (ret) {
+ gvt_vgpu_err("scan wa ctx error\n");
+ return ret;
+ }
+
+ return 0;
+}
+
+/* generate dummy contexts by sending empty requests to HW, and let
+ * the HW to fill Engine Contexts. This dummy contexts are used for
+ * initialization purpose (update reg whitelist), so referred to as
+ * init context here
+ */
+void intel_gvt_update_reg_whitelist(struct intel_vgpu *vgpu)
+{
+ const unsigned long start = LRC_STATE_PN * PAGE_SIZE;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+
+ if (gvt->is_reg_whitelist_updated)
+ return;
+
+ /* scan init ctx to update cmd accessible list */
+ for_each_engine(engine, gvt->gt, id) {
+ struct parser_exec_state s;
+ void *vaddr;
+ int ret;
+
+ if (!engine->default_state)
+ continue;
+
+ vaddr = shmem_pin_map(engine->default_state);
+ if (!vaddr) {
+ gvt_err("failed to map %s->default state\n",
+ engine->name);
+ return;
+ }
+
+ s.buf_type = RING_BUFFER_CTX;
+ s.buf_addr_type = GTT_BUFFER;
+ s.vgpu = vgpu;
+ s.engine = engine;
+ s.ring_start = 0;
+ s.ring_size = engine->context_size - start;
+ s.ring_head = 0;
+ s.ring_tail = s.ring_size;
+ s.rb_va = vaddr + start;
+ s.workload = NULL;
+ s.is_ctx_wa = false;
+ s.is_init_ctx = true;
+
+ /* skipping the first RING_CTX_SIZE(0x50) dwords */
+ ret = ip_gma_set(&s, RING_CTX_SIZE);
+ if (ret == 0) {
+ ret = command_scan(&s, 0, s.ring_size, 0, s.ring_size);
+ if (ret)
+ gvt_err("Scan init ctx error\n");
+ }
+
+ shmem_unpin_map(engine->default_state, vaddr);
+ if (ret)
+ return;
+ }
+
+ gvt->is_reg_whitelist_updated = true;
+}
+
+int intel_gvt_scan_engine_context(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ unsigned long gma_head, gma_tail, gma_start, ctx_size;
+ struct parser_exec_state s;
+ int ring_id = workload->engine->id;
+ struct intel_context *ce = vgpu->submission.shadow[ring_id];
+ int ret;
+
+ GEM_BUG_ON(atomic_read(&ce->pin_count) < 0);
+
+ ctx_size = workload->engine->context_size - PAGE_SIZE;
+
+ /* Only ring contxt is loaded to HW for inhibit context, no need to
+ * scan engine context
+ */
+ if (is_inhibit_context(ce))
+ return 0;
+
+ gma_start = i915_ggtt_offset(ce->state) + LRC_STATE_PN*PAGE_SIZE;
+ gma_head = 0;
+ gma_tail = ctx_size;
+
+ s.buf_type = RING_BUFFER_CTX;
+ s.buf_addr_type = GTT_BUFFER;
+ s.vgpu = workload->vgpu;
+ s.engine = workload->engine;
+ s.ring_start = gma_start;
+ s.ring_size = ctx_size;
+ s.ring_head = gma_start + gma_head;
+ s.ring_tail = gma_start + gma_tail;
+ s.rb_va = ce->lrc_reg_state;
+ s.workload = workload;
+ s.is_ctx_wa = false;
+ s.is_init_ctx = false;
+
+ /* don't scan the first RING_CTX_SIZE(0x50) dwords, as it's ring
+ * context
+ */
+ ret = ip_gma_set(&s, gma_start + gma_head + RING_CTX_SIZE);
+ if (ret)
+ goto out;
+
+ ret = command_scan(&s, gma_head, gma_tail,
+ gma_start, ctx_size);
+out:
+ if (ret)
+ gvt_vgpu_err("scan shadow ctx error\n");
+
+ return ret;
+}
+
+static int init_cmd_table(struct intel_gvt *gvt)
+{
+ unsigned int gen_type = intel_gvt_get_device_type(gvt);
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(cmd_info); i++) {
+ struct cmd_entry *e;
+
+ if (!(cmd_info[i].devices & gen_type))
+ continue;
+
+ e = kzalloc(sizeof(*e), GFP_KERNEL);
+ if (!e)
+ return -ENOMEM;
+
+ e->info = &cmd_info[i];
+ if (cmd_info[i].opcode == OP_MI_NOOP)
+ mi_noop_index = i;
+
+ INIT_HLIST_NODE(&e->hlist);
+ add_cmd_entry(gvt, e);
+ gvt_dbg_cmd("add %-30s op %04x flag %x devs %02x rings %02x\n",
+ e->info->name, e->info->opcode, e->info->flag,
+ e->info->devices, e->info->rings);
+ }
+
+ return 0;
+}
+
+static void clean_cmd_table(struct intel_gvt *gvt)
+{
+ struct hlist_node *tmp;
+ struct cmd_entry *e;
+ int i;
+
+ hash_for_each_safe(gvt->cmd_table, i, tmp, e, hlist)
+ kfree(e);
+
+ hash_init(gvt->cmd_table);
+}
+
+void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt)
+{
+ clean_cmd_table(gvt);
+}
+
+int intel_gvt_init_cmd_parser(struct intel_gvt *gvt)
+{
+ int ret;
+
+ ret = init_cmd_table(gvt);
+ if (ret) {
+ intel_gvt_clean_cmd_parser(gvt);
+ return ret;
+ }
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.h b/drivers/gpu/drm/i915/gvt/cmd_parser.h
new file mode 100644
index 000000000..416d345e2
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.h
@@ -0,0 +1,58 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Kevin Tian <kevin.tian@intel.com>
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Yulei Zhang <yulei.zhang@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+#ifndef _GVT_CMD_PARSER_H_
+#define _GVT_CMD_PARSER_H_
+
+#define GVT_CMD_HASH_BITS 7
+
+struct intel_gvt;
+struct intel_shadow_wa_ctx;
+struct intel_vgpu;
+struct intel_vgpu_workload;
+
+void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt);
+
+int intel_gvt_init_cmd_parser(struct intel_gvt *gvt);
+
+int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload);
+
+int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx);
+
+void intel_gvt_update_reg_whitelist(struct intel_vgpu *vgpu);
+
+int intel_gvt_scan_engine_context(struct intel_vgpu_workload *workload);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/debug.h b/drivers/gpu/drm/i915/gvt/debug.h
new file mode 100644
index 000000000..c6027125c
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/debug.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ */
+
+#ifndef __GVT_DEBUG_H__
+#define __GVT_DEBUG_H__
+
+#define gvt_err(fmt, args...) \
+ pr_err("gvt: "fmt, ##args)
+
+#define gvt_vgpu_err(fmt, args...) \
+do { \
+ if (IS_ERR_OR_NULL(vgpu)) \
+ pr_err("gvt: "fmt, ##args); \
+ else \
+ pr_err("gvt: vgpu %d: "fmt, vgpu->id, ##args);\
+} while (0)
+
+#define gvt_dbg_core(fmt, args...) \
+ pr_debug("gvt: core: "fmt, ##args)
+
+#define gvt_dbg_irq(fmt, args...) \
+ pr_debug("gvt: irq: "fmt, ##args)
+
+#define gvt_dbg_mm(fmt, args...) \
+ pr_debug("gvt: mm: "fmt, ##args)
+
+#define gvt_dbg_mmio(fmt, args...) \
+ pr_debug("gvt: mmio: "fmt, ##args)
+
+#define gvt_dbg_dpy(fmt, args...) \
+ pr_debug("gvt: dpy: "fmt, ##args)
+
+#define gvt_dbg_el(fmt, args...) \
+ pr_debug("gvt: el: "fmt, ##args)
+
+#define gvt_dbg_sched(fmt, args...) \
+ pr_debug("gvt: sched: "fmt, ##args)
+
+#define gvt_dbg_render(fmt, args...) \
+ pr_debug("gvt: render: "fmt, ##args)
+
+#define gvt_dbg_cmd(fmt, args...) \
+ pr_debug("gvt: cmd: "fmt, ##args)
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c
new file mode 100644
index 000000000..e08ed0e9f
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/debugfs.c
@@ -0,0 +1,213 @@
+/*
+ * Copyright(c) 2011-2017 Intel 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 (including the next
+ * paragraph) 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 <linux/debugfs.h>
+#include <linux/list_sort.h>
+#include "i915_drv.h"
+#include "gvt.h"
+
+struct mmio_diff_param {
+ struct intel_vgpu *vgpu;
+ int total;
+ int diff;
+ struct list_head diff_mmio_list;
+};
+
+struct diff_mmio {
+ struct list_head node;
+ u32 offset;
+ u32 preg;
+ u32 vreg;
+};
+
+/* Compare two diff_mmio items. */
+static int mmio_offset_compare(void *priv,
+ const struct list_head *a, const struct list_head *b)
+{
+ struct diff_mmio *ma;
+ struct diff_mmio *mb;
+
+ ma = container_of(a, struct diff_mmio, node);
+ mb = container_of(b, struct diff_mmio, node);
+ if (ma->offset < mb->offset)
+ return -1;
+ else if (ma->offset > mb->offset)
+ return 1;
+ return 0;
+}
+
+static inline int mmio_diff_handler(struct intel_gvt *gvt,
+ u32 offset, void *data)
+{
+ struct mmio_diff_param *param = data;
+ struct diff_mmio *node;
+ u32 preg, vreg;
+
+ preg = intel_uncore_read_notrace(gvt->gt->uncore, _MMIO(offset));
+ vreg = vgpu_vreg(param->vgpu, offset);
+
+ if (preg != vreg) {
+ node = kmalloc(sizeof(*node), GFP_ATOMIC);
+ if (!node)
+ return -ENOMEM;
+
+ node->offset = offset;
+ node->preg = preg;
+ node->vreg = vreg;
+ list_add(&node->node, &param->diff_mmio_list);
+ param->diff++;
+ }
+ param->total++;
+ return 0;
+}
+
+/* Show the all the different values of tracked mmio. */
+static int vgpu_mmio_diff_show(struct seq_file *s, void *unused)
+{
+ struct intel_vgpu *vgpu = s->private;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct mmio_diff_param param = {
+ .vgpu = vgpu,
+ .total = 0,
+ .diff = 0,
+ };
+ struct diff_mmio *node, *next;
+
+ INIT_LIST_HEAD(&param.diff_mmio_list);
+
+ mutex_lock(&gvt->lock);
+ spin_lock_bh(&gvt->scheduler.mmio_context_lock);
+
+ mmio_hw_access_pre(gvt->gt);
+ /* Recognize all the diff mmios to list. */
+ intel_gvt_for_each_tracked_mmio(gvt, mmio_diff_handler, &param);
+ mmio_hw_access_post(gvt->gt);
+
+ spin_unlock_bh(&gvt->scheduler.mmio_context_lock);
+ mutex_unlock(&gvt->lock);
+
+ /* In an ascending order by mmio offset. */
+ list_sort(NULL, &param.diff_mmio_list, mmio_offset_compare);
+
+ seq_printf(s, "%-8s %-8s %-8s %-8s\n", "Offset", "HW", "vGPU", "Diff");
+ list_for_each_entry_safe(node, next, &param.diff_mmio_list, node) {
+ u32 diff = node->preg ^ node->vreg;
+
+ seq_printf(s, "%08x %08x %08x %*pbl\n",
+ node->offset, node->preg, node->vreg,
+ 32, &diff);
+ list_del(&node->node);
+ kfree(node);
+ }
+ seq_printf(s, "Total: %d, Diff: %d\n", param.total, param.diff);
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(vgpu_mmio_diff);
+
+static int
+vgpu_scan_nonprivbb_get(void *data, u64 *val)
+{
+ struct intel_vgpu *vgpu = (struct intel_vgpu *)data;
+
+ *val = vgpu->scan_nonprivbb;
+ return 0;
+}
+
+/*
+ * set/unset bit engine_id of vgpu->scan_nonprivbb to turn on/off scanning
+ * of non-privileged batch buffer. e.g.
+ * if vgpu->scan_nonprivbb=3, then it will scan non-privileged batch buffer
+ * on engine 0 and 1.
+ */
+static int
+vgpu_scan_nonprivbb_set(void *data, u64 val)
+{
+ struct intel_vgpu *vgpu = (struct intel_vgpu *)data;
+
+ vgpu->scan_nonprivbb = val;
+ return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(vgpu_scan_nonprivbb_fops,
+ vgpu_scan_nonprivbb_get, vgpu_scan_nonprivbb_set,
+ "0x%llx\n");
+
+/**
+ * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU
+ * @vgpu: a vGPU
+ */
+void intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu)
+{
+ char name[16] = "";
+
+ snprintf(name, 16, "vgpu%d", vgpu->id);
+ vgpu->debugfs = debugfs_create_dir(name, vgpu->gvt->debugfs_root);
+
+ debugfs_create_bool("active", 0444, vgpu->debugfs, &vgpu->active);
+ debugfs_create_file("mmio_diff", 0444, vgpu->debugfs, vgpu,
+ &vgpu_mmio_diff_fops);
+ debugfs_create_file("scan_nonprivbb", 0644, vgpu->debugfs, vgpu,
+ &vgpu_scan_nonprivbb_fops);
+}
+
+/**
+ * intel_gvt_debugfs_remove_vgpu - remove debugfs entries of a vGPU
+ * @vgpu: a vGPU
+ */
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_minor *minor = gvt->gt->i915->drm.primary;
+
+ if (minor->debugfs_root && gvt->debugfs_root) {
+ debugfs_remove_recursive(vgpu->debugfs);
+ vgpu->debugfs = NULL;
+ }
+}
+
+/**
+ * intel_gvt_debugfs_init - register gvt debugfs root entry
+ * @gvt: GVT device
+ */
+void intel_gvt_debugfs_init(struct intel_gvt *gvt)
+{
+ struct drm_minor *minor = gvt->gt->i915->drm.primary;
+
+ gvt->debugfs_root = debugfs_create_dir("gvt", minor->debugfs_root);
+
+ debugfs_create_ulong("num_tracked_mmio", 0444, gvt->debugfs_root,
+ &gvt->mmio.num_tracked_mmio);
+}
+
+/**
+ * intel_gvt_debugfs_clean - remove debugfs entries
+ * @gvt: GVT device
+ */
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt)
+{
+ struct drm_minor *minor = gvt->gt->i915->drm.primary;
+
+ if (minor->debugfs_root) {
+ debugfs_remove_recursive(gvt->debugfs_root);
+ gvt->debugfs_root = NULL;
+ }
+}
diff --git a/drivers/gpu/drm/i915/gvt/display.c b/drivers/gpu/drm/i915/gvt/display.c
new file mode 100644
index 000000000..c7722c818
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/display.c
@@ -0,0 +1,808 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Terrence Xu <terrence.xu@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gvt.h"
+
+static int get_edp_pipe(struct intel_vgpu *vgpu)
+{
+ u32 data = vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP);
+ int pipe = -1;
+
+ switch (data & TRANS_DDI_EDP_INPUT_MASK) {
+ case TRANS_DDI_EDP_INPUT_A_ON:
+ case TRANS_DDI_EDP_INPUT_A_ONOFF:
+ pipe = PIPE_A;
+ break;
+ case TRANS_DDI_EDP_INPUT_B_ONOFF:
+ pipe = PIPE_B;
+ break;
+ case TRANS_DDI_EDP_INPUT_C_ONOFF:
+ pipe = PIPE_C;
+ break;
+ }
+ return pipe;
+}
+
+static int edp_pipe_is_enabled(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ if (!(vgpu_vreg_t(vgpu, PIPECONF(_PIPE_EDP)) & PIPECONF_ENABLE))
+ return 0;
+
+ if (!(vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP) & TRANS_DDI_FUNC_ENABLE))
+ return 0;
+ return 1;
+}
+
+int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ if (drm_WARN_ON(&dev_priv->drm,
+ pipe < PIPE_A || pipe >= I915_MAX_PIPES))
+ return -EINVAL;
+
+ if (vgpu_vreg_t(vgpu, PIPECONF(pipe)) & PIPECONF_ENABLE)
+ return 1;
+
+ if (edp_pipe_is_enabled(vgpu) &&
+ get_edp_pipe(vgpu) == pipe)
+ return 1;
+ return 0;
+}
+
+static unsigned char virtual_dp_monitor_edid[GVT_EDID_NUM][EDID_SIZE] = {
+ {
+/* EDID with 1024x768 as its resolution */
+ /*Header*/
+ 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00,
+ /* Vendor & Product Identification */
+ 0x22, 0xf0, 0x54, 0x29, 0x00, 0x00, 0x00, 0x00, 0x04, 0x17,
+ /* Version & Revision */
+ 0x01, 0x04,
+ /* Basic Display Parameters & Features */
+ 0xa5, 0x34, 0x20, 0x78, 0x23,
+ /* Color Characteristics */
+ 0xfc, 0x81, 0xa4, 0x55, 0x4d, 0x9d, 0x25, 0x12, 0x50, 0x54,
+ /* Established Timings: maximum resolution is 1024x768 */
+ 0x21, 0x08, 0x00,
+ /* Standard Timings. All invalid */
+ 0x00, 0xc0, 0x00, 0xc0, 0x00, 0x40, 0x00, 0x80, 0x00, 0x00,
+ 0x00, 0x40, 0x00, 0x00, 0x00, 0x01,
+ /* 18 Byte Data Blocks 1: invalid */
+ 0x00, 0x00, 0x80, 0xa0, 0x70, 0xb0,
+ 0x23, 0x40, 0x30, 0x20, 0x36, 0x00, 0x06, 0x44, 0x21, 0x00, 0x00, 0x1a,
+ /* 18 Byte Data Blocks 2: invalid */
+ 0x00, 0x00, 0x00, 0xfd, 0x00, 0x18, 0x3c, 0x18, 0x50, 0x11, 0x00, 0x0a,
+ 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+ /* 18 Byte Data Blocks 3: invalid */
+ 0x00, 0x00, 0x00, 0xfc, 0x00, 0x48,
+ 0x50, 0x20, 0x5a, 0x52, 0x32, 0x34, 0x34, 0x30, 0x77, 0x0a, 0x20, 0x20,
+ /* 18 Byte Data Blocks 4: invalid */
+ 0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 0x4e, 0x34, 0x33, 0x30, 0x34, 0x30,
+ 0x44, 0x58, 0x51, 0x0a, 0x20, 0x20,
+ /* Extension Block Count */
+ 0x00,
+ /* Checksum */
+ 0xef,
+ },
+ {
+/* EDID with 1920x1200 as its resolution */
+ /*Header*/
+ 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00,
+ /* Vendor & Product Identification */
+ 0x22, 0xf0, 0x54, 0x29, 0x00, 0x00, 0x00, 0x00, 0x04, 0x17,
+ /* Version & Revision */
+ 0x01, 0x04,
+ /* Basic Display Parameters & Features */
+ 0xa5, 0x34, 0x20, 0x78, 0x23,
+ /* Color Characteristics */
+ 0xfc, 0x81, 0xa4, 0x55, 0x4d, 0x9d, 0x25, 0x12, 0x50, 0x54,
+ /* Established Timings: maximum resolution is 1024x768 */
+ 0x21, 0x08, 0x00,
+ /*
+ * Standard Timings.
+ * below new resolutions can be supported:
+ * 1920x1080, 1280x720, 1280x960, 1280x1024,
+ * 1440x900, 1600x1200, 1680x1050
+ */
+ 0xd1, 0xc0, 0x81, 0xc0, 0x81, 0x40, 0x81, 0x80, 0x95, 0x00,
+ 0xa9, 0x40, 0xb3, 0x00, 0x01, 0x01,
+ /* 18 Byte Data Blocks 1: max resolution is 1920x1200 */
+ 0x28, 0x3c, 0x80, 0xa0, 0x70, 0xb0,
+ 0x23, 0x40, 0x30, 0x20, 0x36, 0x00, 0x06, 0x44, 0x21, 0x00, 0x00, 0x1a,
+ /* 18 Byte Data Blocks 2: invalid */
+ 0x00, 0x00, 0x00, 0xfd, 0x00, 0x18, 0x3c, 0x18, 0x50, 0x11, 0x00, 0x0a,
+ 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+ /* 18 Byte Data Blocks 3: invalid */
+ 0x00, 0x00, 0x00, 0xfc, 0x00, 0x48,
+ 0x50, 0x20, 0x5a, 0x52, 0x32, 0x34, 0x34, 0x30, 0x77, 0x0a, 0x20, 0x20,
+ /* 18 Byte Data Blocks 4: invalid */
+ 0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 0x4e, 0x34, 0x33, 0x30, 0x34, 0x30,
+ 0x44, 0x58, 0x51, 0x0a, 0x20, 0x20,
+ /* Extension Block Count */
+ 0x00,
+ /* Checksum */
+ 0x45,
+ },
+};
+
+#define DPCD_HEADER_SIZE 0xb
+
+/* let the virtual display supports DP1.2 */
+static u8 dpcd_fix_data[DPCD_HEADER_SIZE] = {
+ 0x12, 0x014, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
+static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ int pipe;
+
+ if (IS_BROXTON(dev_priv)) {
+ enum transcoder trans;
+ enum port port;
+
+ /* Clear PIPE, DDI, PHY, HPD before setting new */
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &=
+ ~(GEN8_DE_PORT_HOTPLUG(HPD_PORT_A) |
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_B) |
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_C));
+
+ for_each_pipe(dev_priv, pipe) {
+ vgpu_vreg_t(vgpu, PIPECONF(pipe)) &=
+ ~(PIPECONF_ENABLE | PIPECONF_STATE_ENABLE);
+ vgpu_vreg_t(vgpu, DSPCNTR(pipe)) &= ~DISP_ENABLE;
+ vgpu_vreg_t(vgpu, SPRCTL(pipe)) &= ~SPRITE_ENABLE;
+ vgpu_vreg_t(vgpu, CURCNTR(pipe)) &= ~MCURSOR_MODE_MASK;
+ vgpu_vreg_t(vgpu, CURCNTR(pipe)) |= MCURSOR_MODE_DISABLE;
+ }
+
+ for (trans = TRANSCODER_A; trans <= TRANSCODER_EDP; trans++) {
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(trans)) &=
+ ~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
+ TRANS_DDI_PORT_MASK | TRANS_DDI_FUNC_ENABLE);
+ }
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+ ~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
+ TRANS_DDI_PORT_MASK);
+
+ for (port = PORT_A; port <= PORT_C; port++) {
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(port)) &=
+ ~BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(port)) |=
+ (BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK);
+
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_ENABLE(port)) &=
+ ~(PORT_PLL_POWER_STATE | PORT_PLL_POWER_ENABLE |
+ PORT_PLL_REF_SEL | PORT_PLL_LOCK |
+ PORT_PLL_ENABLE);
+
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(port)) &=
+ ~(DDI_INIT_DISPLAY_DETECTED |
+ DDI_BUF_CTL_ENABLE);
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(port)) |= DDI_BUF_IS_IDLE;
+ }
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~(PORTA_HOTPLUG_ENABLE | PORTA_HOTPLUG_STATUS_MASK);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~(PORTB_HOTPLUG_ENABLE | PORTB_HOTPLUG_STATUS_MASK);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~(PORTC_HOTPLUG_ENABLE | PORTC_HOTPLUG_STATUS_MASK);
+ /* No hpd_invert set in vgpu vbt, need to clear invert mask */
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &= ~BXT_DDI_HPD_INVERT_MASK;
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &= ~BXT_DE_PORT_HOTPLUG_MASK;
+
+ vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) &= ~(BIT(0) | BIT(1));
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) &=
+ ~PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY1)) &=
+ ~PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY0)) &= ~BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY1)) &= ~BIT(30);
+
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) &= ~SFUSE_STRAP_DDIB_DETECTED;
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) &= ~SFUSE_STRAP_DDIC_DETECTED;
+
+ /*
+ * Only 1 PIPE enabled in current vGPU display and PIPE_A is
+ * tied to TRANSCODER_A in HW, so it's safe to assume PIPE_A,
+ * TRANSCODER_A can be enabled. PORT_x depends on the input of
+ * setup_virtual_dp_monitor.
+ */
+ vgpu_vreg_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
+ vgpu_vreg_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_STATE_ENABLE;
+
+ /*
+ * Golden M/N are calculated based on:
+ * 24 bpp, 4 lanes, 154000 pixel clk (from virtual EDID),
+ * DP link clk 1620 MHz and non-constant_n.
+ * TODO: calculate DP link symbol clk and stream clk m/n.
+ */
+ vgpu_vreg_t(vgpu, PIPE_DATA_M1(TRANSCODER_A)) = TU_SIZE(64);
+ vgpu_vreg_t(vgpu, PIPE_DATA_M1(TRANSCODER_A)) |= 0x5b425e;
+ vgpu_vreg_t(vgpu, PIPE_DATA_N1(TRANSCODER_A)) = 0x800000;
+ vgpu_vreg_t(vgpu, PIPE_LINK_M1(TRANSCODER_A)) = 0x3cd6e;
+ vgpu_vreg_t(vgpu, PIPE_LINK_N1(TRANSCODER_A)) = 0x80000;
+
+ /* Enable per-DDI/PORT vreg */
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
+ vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) |= BIT(1);
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY1)) |=
+ PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY1)) |=
+ BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_A)) |=
+ BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_A)) &=
+ ~(BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK);
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_ENABLE(PORT_A)) |=
+ (PORT_PLL_POWER_STATE | PORT_PLL_POWER_ENABLE |
+ PORT_PLL_REF_SEL | PORT_PLL_LOCK |
+ PORT_PLL_ENABLE);
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_A)) |=
+ (DDI_BUF_CTL_ENABLE | DDI_INIT_DISPLAY_DETECTED);
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_A)) &=
+ ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_EDP)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ TRANS_DDI_FUNC_ENABLE);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTA_HOTPLUG_ENABLE;
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_A);
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIB_DETECTED;
+ vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) |= BIT(0);
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) |=
+ PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY0)) |=
+ BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_B)) |=
+ BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_B)) &=
+ ~(BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK);
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_ENABLE(PORT_B)) |=
+ (PORT_PLL_POWER_STATE | PORT_PLL_POWER_ENABLE |
+ PORT_PLL_REF_SEL | PORT_PLL_LOCK |
+ PORT_PLL_ENABLE);
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) |=
+ DDI_BUF_CTL_ENABLE;
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) &=
+ ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ (PORT_B << TRANS_DDI_PORT_SHIFT) |
+ TRANS_DDI_FUNC_ENABLE);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTB_HOTPLUG_ENABLE;
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_B);
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIC_DETECTED;
+ vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) |= BIT(0);
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) |=
+ PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY0)) |=
+ BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_C)) |=
+ BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_C)) &=
+ ~(BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK);
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_ENABLE(PORT_C)) |=
+ (PORT_PLL_POWER_STATE | PORT_PLL_POWER_ENABLE |
+ PORT_PLL_REF_SEL | PORT_PLL_LOCK |
+ PORT_PLL_ENABLE);
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) |=
+ DDI_BUF_CTL_ENABLE;
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) &=
+ ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ (PORT_B << TRANS_DDI_PORT_SHIFT) |
+ TRANS_DDI_FUNC_ENABLE);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTC_HOTPLUG_ENABLE;
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_C);
+ }
+
+ return;
+ }
+
+ vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTB_HOTPLUG_CPT |
+ SDE_PORTC_HOTPLUG_CPT |
+ SDE_PORTD_HOTPLUG_CPT);
+
+ if (IS_SKYLAKE(dev_priv) ||
+ IS_KABYLAKE(dev_priv) ||
+ IS_COFFEELAKE(dev_priv) ||
+ IS_COMETLAKE(dev_priv)) {
+ vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTA_HOTPLUG_SPT |
+ SDE_PORTE_HOTPLUG_SPT);
+ vgpu_vreg_t(vgpu, SKL_FUSE_STATUS) |=
+ SKL_FUSE_DOWNLOAD_STATUS |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG0) |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG1) |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG2);
+ /*
+ * Only 1 PIPE enabled in current vGPU display and PIPE_A is
+ * tied to TRANSCODER_A in HW, so it's safe to assume PIPE_A,
+ * TRANSCODER_A can be enabled. PORT_x depends on the input of
+ * setup_virtual_dp_monitor, we can bind DPLL0 to any PORT_x
+ * so we fixed to DPLL0 here.
+ * Setup DPLL0: DP link clk 1620 MHz, non SSC, DP Mode
+ */
+ vgpu_vreg_t(vgpu, DPLL_CTRL1) =
+ DPLL_CTRL1_OVERRIDE(DPLL_ID_SKL_DPLL0);
+ vgpu_vreg_t(vgpu, DPLL_CTRL1) |=
+ DPLL_CTRL1_LINK_RATE(DPLL_CTRL1_LINK_RATE_1620, DPLL_ID_SKL_DPLL0);
+ vgpu_vreg_t(vgpu, LCPLL1_CTL) =
+ LCPLL_PLL_ENABLE | LCPLL_PLL_LOCK;
+ vgpu_vreg_t(vgpu, DPLL_STATUS) = DPLL_LOCK(DPLL_ID_SKL_DPLL0);
+ /*
+ * Golden M/N are calculated based on:
+ * 24 bpp, 4 lanes, 154000 pixel clk (from virtual EDID),
+ * DP link clk 1620 MHz and non-constant_n.
+ * TODO: calculate DP link symbol clk and stream clk m/n.
+ */
+ vgpu_vreg_t(vgpu, PIPE_DATA_M1(TRANSCODER_A)) = TU_SIZE(64);
+ vgpu_vreg_t(vgpu, PIPE_DATA_M1(TRANSCODER_A)) |= 0x5b425e;
+ vgpu_vreg_t(vgpu, PIPE_DATA_N1(TRANSCODER_A)) = 0x800000;
+ vgpu_vreg_t(vgpu, PIPE_LINK_M1(TRANSCODER_A)) = 0x3cd6e;
+ vgpu_vreg_t(vgpu, PIPE_LINK_N1(TRANSCODER_A)) = 0x80000;
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) &=
+ ~DPLL_CTRL2_DDI_CLK_OFF(PORT_B);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_CLK_SEL(DPLL_ID_SKL_DPLL0, PORT_B);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_SEL_OVERRIDE(PORT_B);
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIB_DETECTED;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+ ~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
+ TRANS_DDI_PORT_MASK);
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ (PORT_B << TRANS_DDI_PORT_SHIFT) |
+ TRANS_DDI_FUNC_ENABLE);
+ if (IS_BROADWELL(dev_priv)) {
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) &=
+ ~PORT_CLK_SEL_MASK;
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) |=
+ PORT_CLK_SEL_LCPLL_810;
+ }
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) |= DDI_BUF_CTL_ENABLE;
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) &= ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTB_HOTPLUG_CPT;
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) &=
+ ~DPLL_CTRL2_DDI_CLK_OFF(PORT_C);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_CLK_SEL(DPLL_ID_SKL_DPLL0, PORT_C);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_SEL_OVERRIDE(PORT_C);
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTC_HOTPLUG_CPT;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+ ~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
+ TRANS_DDI_PORT_MASK);
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ (PORT_C << TRANS_DDI_PORT_SHIFT) |
+ TRANS_DDI_FUNC_ENABLE);
+ if (IS_BROADWELL(dev_priv)) {
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) &=
+ ~PORT_CLK_SEL_MASK;
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) |=
+ PORT_CLK_SEL_LCPLL_810;
+ }
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) |= DDI_BUF_CTL_ENABLE;
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) &= ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIC_DETECTED;
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_D)) {
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) &=
+ ~DPLL_CTRL2_DDI_CLK_OFF(PORT_D);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_CLK_SEL(DPLL_ID_SKL_DPLL0, PORT_D);
+ vgpu_vreg_t(vgpu, DPLL_CTRL2) |=
+ DPLL_CTRL2_DDI_SEL_OVERRIDE(PORT_D);
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+ ~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
+ TRANS_DDI_PORT_MASK);
+ vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+ (TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+ (PORT_D << TRANS_DDI_PORT_SHIFT) |
+ TRANS_DDI_FUNC_ENABLE);
+ if (IS_BROADWELL(dev_priv)) {
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) &=
+ ~PORT_CLK_SEL_MASK;
+ vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) |=
+ PORT_CLK_SEL_LCPLL_810;
+ }
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) |= DDI_BUF_CTL_ENABLE;
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) &= ~DDI_BUF_IS_IDLE;
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDID_DETECTED;
+ }
+
+ if ((IS_SKYLAKE(dev_priv) ||
+ IS_KABYLAKE(dev_priv) ||
+ IS_COFFEELAKE(dev_priv) ||
+ IS_COMETLAKE(dev_priv)) &&
+ intel_vgpu_has_monitor_on_port(vgpu, PORT_E)) {
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTE_HOTPLUG_SPT;
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
+ if (IS_BROADWELL(dev_priv))
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_A);
+ else
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTA_HOTPLUG_SPT;
+
+ vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_A)) |= DDI_INIT_DISPLAY_DETECTED;
+ }
+
+ /* Clear host CRT status, so guest couldn't detect this host CRT. */
+ if (IS_BROADWELL(dev_priv))
+ vgpu_vreg_t(vgpu, PCH_ADPA) &= ~ADPA_CRT_HOTPLUG_MONITOR_MASK;
+
+ /* Disable Primary/Sprite/Cursor plane */
+ for_each_pipe(dev_priv, pipe) {
+ vgpu_vreg_t(vgpu, DSPCNTR(pipe)) &= ~DISP_ENABLE;
+ vgpu_vreg_t(vgpu, SPRCTL(pipe)) &= ~SPRITE_ENABLE;
+ vgpu_vreg_t(vgpu, CURCNTR(pipe)) &= ~MCURSOR_MODE_MASK;
+ vgpu_vreg_t(vgpu, CURCNTR(pipe)) |= MCURSOR_MODE_DISABLE;
+ }
+
+ vgpu_vreg_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
+}
+
+static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
+{
+ struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
+
+ kfree(port->edid);
+ port->edid = NULL;
+
+ kfree(port->dpcd);
+ port->dpcd = NULL;
+}
+
+static enum hrtimer_restart vblank_timer_fn(struct hrtimer *data)
+{
+ struct intel_vgpu_vblank_timer *vblank_timer;
+ struct intel_vgpu *vgpu;
+
+ vblank_timer = container_of(data, struct intel_vgpu_vblank_timer, timer);
+ vgpu = container_of(vblank_timer, struct intel_vgpu, vblank_timer);
+
+ /* Set vblank emulation request per-vGPU bit */
+ intel_gvt_request_service(vgpu->gvt,
+ INTEL_GVT_REQUEST_EMULATE_VBLANK + vgpu->id);
+ hrtimer_add_expires_ns(&vblank_timer->timer, vblank_timer->period);
+ return HRTIMER_RESTART;
+}
+
+static int setup_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num,
+ int type, unsigned int resolution)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
+ struct intel_vgpu_vblank_timer *vblank_timer = &vgpu->vblank_timer;
+
+ if (drm_WARN_ON(&i915->drm, resolution >= GVT_EDID_NUM))
+ return -EINVAL;
+
+ port->edid = kzalloc(sizeof(*(port->edid)), GFP_KERNEL);
+ if (!port->edid)
+ return -ENOMEM;
+
+ port->dpcd = kzalloc(sizeof(*(port->dpcd)), GFP_KERNEL);
+ if (!port->dpcd) {
+ kfree(port->edid);
+ return -ENOMEM;
+ }
+
+ memcpy(port->edid->edid_block, virtual_dp_monitor_edid[resolution],
+ EDID_SIZE);
+ port->edid->data_valid = true;
+
+ memcpy(port->dpcd->data, dpcd_fix_data, DPCD_HEADER_SIZE);
+ port->dpcd->data_valid = true;
+ port->dpcd->data[DPCD_SINK_COUNT] = 0x1;
+ port->type = type;
+ port->id = resolution;
+ port->vrefresh_k = GVT_DEFAULT_REFRESH_RATE * MSEC_PER_SEC;
+ vgpu->display.port_num = port_num;
+
+ /* Init hrtimer based on default refresh rate */
+ hrtimer_init(&vblank_timer->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
+ vblank_timer->timer.function = vblank_timer_fn;
+ vblank_timer->vrefresh_k = port->vrefresh_k;
+ vblank_timer->period = DIV64_U64_ROUND_CLOSEST(NSEC_PER_SEC * MSEC_PER_SEC, vblank_timer->vrefresh_k);
+
+ emulate_monitor_status_change(vgpu);
+
+ return 0;
+}
+
+/**
+ * vgpu_update_vblank_emulation - Update per-vGPU vblank_timer
+ * @vgpu: vGPU operated
+ * @turnon: Turn ON/OFF vblank_timer
+ *
+ * This function is used to turn on/off or update the per-vGPU vblank_timer
+ * when PIPECONF is enabled or disabled. vblank_timer period is also updated
+ * if guest changed the refresh rate.
+ *
+ */
+void vgpu_update_vblank_emulation(struct intel_vgpu *vgpu, bool turnon)
+{
+ struct intel_vgpu_vblank_timer *vblank_timer = &vgpu->vblank_timer;
+ struct intel_vgpu_port *port =
+ intel_vgpu_port(vgpu, vgpu->display.port_num);
+
+ if (turnon) {
+ /*
+ * Skip the re-enable if already active and vrefresh unchanged.
+ * Otherwise, stop timer if already active and restart with new
+ * period.
+ */
+ if (vblank_timer->vrefresh_k != port->vrefresh_k ||
+ !hrtimer_active(&vblank_timer->timer)) {
+ /* Stop timer before start with new period if active */
+ if (hrtimer_active(&vblank_timer->timer))
+ hrtimer_cancel(&vblank_timer->timer);
+
+ /* Make sure new refresh rate updated to timer period */
+ vblank_timer->vrefresh_k = port->vrefresh_k;
+ vblank_timer->period = DIV64_U64_ROUND_CLOSEST(NSEC_PER_SEC * MSEC_PER_SEC, vblank_timer->vrefresh_k);
+ hrtimer_start(&vblank_timer->timer,
+ ktime_add_ns(ktime_get(), vblank_timer->period),
+ HRTIMER_MODE_ABS);
+ }
+ } else {
+ /* Caller request to stop vblank */
+ hrtimer_cancel(&vblank_timer->timer);
+ }
+}
+
+static void emulate_vblank_on_pipe(struct intel_vgpu *vgpu, int pipe)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ struct intel_vgpu_irq *irq = &vgpu->irq;
+ int vblank_event[] = {
+ [PIPE_A] = PIPE_A_VBLANK,
+ [PIPE_B] = PIPE_B_VBLANK,
+ [PIPE_C] = PIPE_C_VBLANK,
+ };
+ int event;
+
+ if (pipe < PIPE_A || pipe > PIPE_C)
+ return;
+
+ for_each_set_bit(event, irq->flip_done_event[pipe],
+ INTEL_GVT_EVENT_MAX) {
+ clear_bit(event, irq->flip_done_event[pipe]);
+ if (!pipe_is_enabled(vgpu, pipe))
+ continue;
+
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ }
+
+ if (pipe_is_enabled(vgpu, pipe)) {
+ vgpu_vreg_t(vgpu, PIPE_FRMCOUNT_G4X(pipe))++;
+ intel_vgpu_trigger_virtual_event(vgpu, vblank_event[pipe]);
+ }
+}
+
+void intel_vgpu_emulate_vblank(struct intel_vgpu *vgpu)
+{
+ int pipe;
+
+ mutex_lock(&vgpu->vgpu_lock);
+ for_each_pipe(vgpu->gvt->gt->i915, pipe)
+ emulate_vblank_on_pipe(vgpu, pipe);
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+/**
+ * intel_vgpu_emulate_hotplug - trigger hotplug event for vGPU
+ * @vgpu: a vGPU
+ * @connected: link state
+ *
+ * This function is used to trigger hotplug interrupt for vGPU
+ *
+ */
+void intel_vgpu_emulate_hotplug(struct intel_vgpu *vgpu, bool connected)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ /* TODO: add more platforms support */
+ if (IS_SKYLAKE(i915) ||
+ IS_KABYLAKE(i915) ||
+ IS_COFFEELAKE(i915) ||
+ IS_COMETLAKE(i915)) {
+ if (connected) {
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |=
+ SFUSE_STRAP_DDID_DETECTED;
+ vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
+ } else {
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) &=
+ ~SFUSE_STRAP_DDID_DETECTED;
+ vgpu_vreg_t(vgpu, SDEISR) &= ~SDE_PORTD_HOTPLUG_CPT;
+ }
+ vgpu_vreg_t(vgpu, SDEIIR) |= SDE_PORTD_HOTPLUG_CPT;
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTD_HOTPLUG_STATUS_MASK;
+ intel_vgpu_trigger_virtual_event(vgpu, DP_D_HOTPLUG);
+ } else if (IS_BROXTON(i915)) {
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
+ if (connected) {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_A);
+ } else {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &=
+ ~GEN8_DE_PORT_HOTPLUG(HPD_PORT_A);
+ }
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_IIR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_A);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~PORTA_HOTPLUG_STATUS_MASK;
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTA_HOTPLUG_LONG_DETECT;
+ intel_vgpu_trigger_virtual_event(vgpu, DP_A_HOTPLUG);
+ }
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
+ if (connected) {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_B);
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |=
+ SFUSE_STRAP_DDIB_DETECTED;
+ } else {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &=
+ ~GEN8_DE_PORT_HOTPLUG(HPD_PORT_B);
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) &=
+ ~SFUSE_STRAP_DDIB_DETECTED;
+ }
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_IIR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_B);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~PORTB_HOTPLUG_STATUS_MASK;
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTB_HOTPLUG_LONG_DETECT;
+ intel_vgpu_trigger_virtual_event(vgpu, DP_B_HOTPLUG);
+ }
+ if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
+ if (connected) {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_C);
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) |=
+ SFUSE_STRAP_DDIC_DETECTED;
+ } else {
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) &=
+ ~GEN8_DE_PORT_HOTPLUG(HPD_PORT_C);
+ vgpu_vreg_t(vgpu, SFUSE_STRAP) &=
+ ~SFUSE_STRAP_DDIC_DETECTED;
+ }
+ vgpu_vreg_t(vgpu, GEN8_DE_PORT_IIR) |=
+ GEN8_DE_PORT_HOTPLUG(HPD_PORT_C);
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) &=
+ ~PORTC_HOTPLUG_STATUS_MASK;
+ vgpu_vreg_t(vgpu, PCH_PORT_HOTPLUG) |=
+ PORTC_HOTPLUG_LONG_DETECT;
+ intel_vgpu_trigger_virtual_event(vgpu, DP_C_HOTPLUG);
+ }
+ }
+}
+
+/**
+ * intel_vgpu_clean_display - clean vGPU virtual display emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to clean vGPU virtual display emulation stuffs
+ *
+ */
+void intel_vgpu_clean_display(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ if (IS_SKYLAKE(dev_priv) ||
+ IS_KABYLAKE(dev_priv) ||
+ IS_COFFEELAKE(dev_priv) ||
+ IS_COMETLAKE(dev_priv))
+ clean_virtual_dp_monitor(vgpu, PORT_D);
+ else
+ clean_virtual_dp_monitor(vgpu, PORT_B);
+
+ vgpu_update_vblank_emulation(vgpu, false);
+}
+
+/**
+ * intel_vgpu_init_display- initialize vGPU virtual display emulation
+ * @vgpu: a vGPU
+ * @resolution: resolution index for intel_vgpu_edid
+ *
+ * This function is used to initialize vGPU virtual display emulation stuffs
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_init_display(struct intel_vgpu *vgpu, u64 resolution)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ intel_vgpu_init_i2c_edid(vgpu);
+
+ if (IS_SKYLAKE(dev_priv) ||
+ IS_KABYLAKE(dev_priv) ||
+ IS_COFFEELAKE(dev_priv) ||
+ IS_COMETLAKE(dev_priv))
+ return setup_virtual_dp_monitor(vgpu, PORT_D, GVT_DP_D,
+ resolution);
+ else
+ return setup_virtual_dp_monitor(vgpu, PORT_B, GVT_DP_B,
+ resolution);
+}
+
+/**
+ * intel_vgpu_reset_display- reset vGPU virtual display emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to reset vGPU virtual display emulation stuffs
+ *
+ */
+void intel_vgpu_reset_display(struct intel_vgpu *vgpu)
+{
+ emulate_monitor_status_change(vgpu);
+}
diff --git a/drivers/gpu/drm/i915/gvt/display.h b/drivers/gpu/drm/i915/gvt/display.h
new file mode 100644
index 000000000..f5616f99e
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/display.h
@@ -0,0 +1,224 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Terrence Xu <terrence.xu@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_DISPLAY_H_
+#define _GVT_DISPLAY_H_
+
+#include <linux/types.h>
+#include <linux/hrtimer.h>
+
+struct intel_gvt;
+struct intel_vgpu;
+
+#define SBI_REG_MAX 20
+#define DPCD_SIZE 0x700
+
+#define intel_vgpu_port(vgpu, port) \
+ (&(vgpu->display.ports[port]))
+
+#define intel_vgpu_has_monitor_on_port(vgpu, port) \
+ (intel_vgpu_port(vgpu, port)->edid && \
+ intel_vgpu_port(vgpu, port)->edid->data_valid)
+
+#define intel_vgpu_port_is_dp(vgpu, port) \
+ ((intel_vgpu_port(vgpu, port)->type == GVT_DP_A) || \
+ (intel_vgpu_port(vgpu, port)->type == GVT_DP_B) || \
+ (intel_vgpu_port(vgpu, port)->type == GVT_DP_C) || \
+ (intel_vgpu_port(vgpu, port)->type == GVT_DP_D))
+
+#define INTEL_GVT_MAX_UEVENT_VARS 3
+
+/* DPCD start */
+#define DPCD_SIZE 0x700
+
+/* DPCD */
+#define DP_SET_POWER 0x600
+#define DP_SET_POWER_D0 0x1
+#define AUX_NATIVE_WRITE 0x8
+#define AUX_NATIVE_READ 0x9
+
+#define AUX_NATIVE_REPLY_MASK (0x3 << 4)
+#define AUX_NATIVE_REPLY_ACK (0x0 << 4)
+#define AUX_NATIVE_REPLY_NAK (0x1 << 4)
+#define AUX_NATIVE_REPLY_DEFER (0x2 << 4)
+
+#define AUX_BURST_SIZE 20
+
+/* DPCD addresses */
+#define DPCD_REV 0x000
+#define DPCD_MAX_LINK_RATE 0x001
+#define DPCD_MAX_LANE_COUNT 0x002
+
+#define DPCD_TRAINING_PATTERN_SET 0x102
+#define DPCD_SINK_COUNT 0x200
+#define DPCD_LANE0_1_STATUS 0x202
+#define DPCD_LANE2_3_STATUS 0x203
+#define DPCD_LANE_ALIGN_STATUS_UPDATED 0x204
+#define DPCD_SINK_STATUS 0x205
+
+/* link training */
+#define DPCD_TRAINING_PATTERN_SET_MASK 0x03
+#define DPCD_LINK_TRAINING_DISABLED 0x00
+#define DPCD_TRAINING_PATTERN_1 0x01
+#define DPCD_TRAINING_PATTERN_2 0x02
+
+#define DPCD_CP_READY_MASK (1 << 6)
+
+/* lane status */
+#define DPCD_LANES_CR_DONE 0x11
+#define DPCD_LANES_EQ_DONE 0x22
+#define DPCD_SYMBOL_LOCKED 0x44
+
+#define DPCD_INTERLANE_ALIGN_DONE 0x01
+
+#define DPCD_SINK_IN_SYNC 0x03
+/* DPCD end */
+
+#define SBI_RESPONSE_MASK 0x3
+#define SBI_RESPONSE_SHIFT 0x1
+#define SBI_STAT_MASK 0x1
+#define SBI_STAT_SHIFT 0x0
+#define SBI_OPCODE_SHIFT 8
+#define SBI_OPCODE_MASK (0xff << SBI_OPCODE_SHIFT)
+#define SBI_CMD_IORD 2
+#define SBI_CMD_IOWR 3
+#define SBI_CMD_CRRD 6
+#define SBI_CMD_CRWR 7
+#define SBI_ADDR_OFFSET_SHIFT 16
+#define SBI_ADDR_OFFSET_MASK (0xffff << SBI_ADDR_OFFSET_SHIFT)
+
+struct intel_vgpu_sbi_register {
+ unsigned int offset;
+ u32 value;
+};
+
+struct intel_vgpu_sbi {
+ int number;
+ struct intel_vgpu_sbi_register registers[SBI_REG_MAX];
+};
+
+enum intel_gvt_plane_type {
+ PRIMARY_PLANE = 0,
+ CURSOR_PLANE,
+ SPRITE_PLANE,
+ MAX_PLANE
+};
+
+struct intel_vgpu_dpcd_data {
+ bool data_valid;
+ u8 data[DPCD_SIZE];
+};
+
+enum intel_vgpu_port_type {
+ GVT_CRT = 0,
+ GVT_DP_A,
+ GVT_DP_B,
+ GVT_DP_C,
+ GVT_DP_D,
+ GVT_HDMI_B,
+ GVT_HDMI_C,
+ GVT_HDMI_D,
+ GVT_PORT_MAX
+};
+
+enum intel_vgpu_edid {
+ GVT_EDID_1024_768,
+ GVT_EDID_1920_1200,
+ GVT_EDID_NUM,
+};
+
+#define GVT_DEFAULT_REFRESH_RATE 60
+struct intel_vgpu_port {
+ /* per display EDID information */
+ struct intel_vgpu_edid_data *edid;
+ /* per display DPCD information */
+ struct intel_vgpu_dpcd_data *dpcd;
+ int type;
+ enum intel_vgpu_edid id;
+ /* x1000 to get accurate 59.94, 24.976, 29.94, etc. in timing std. */
+ u32 vrefresh_k;
+};
+
+struct intel_vgpu_vblank_timer {
+ struct hrtimer timer;
+ u32 vrefresh_k;
+ u64 period;
+};
+
+static inline char *vgpu_edid_str(enum intel_vgpu_edid id)
+{
+ switch (id) {
+ case GVT_EDID_1024_768:
+ return "1024x768";
+ case GVT_EDID_1920_1200:
+ return "1920x1200";
+ default:
+ return "";
+ }
+}
+
+static inline unsigned int vgpu_edid_xres(enum intel_vgpu_edid id)
+{
+ switch (id) {
+ case GVT_EDID_1024_768:
+ return 1024;
+ case GVT_EDID_1920_1200:
+ return 1920;
+ default:
+ return 0;
+ }
+}
+
+static inline unsigned int vgpu_edid_yres(enum intel_vgpu_edid id)
+{
+ switch (id) {
+ case GVT_EDID_1024_768:
+ return 768;
+ case GVT_EDID_1920_1200:
+ return 1200;
+ default:
+ return 0;
+ }
+}
+
+void intel_vgpu_emulate_vblank(struct intel_vgpu *vgpu);
+void vgpu_update_vblank_emulation(struct intel_vgpu *vgpu, bool turnon);
+
+int intel_vgpu_init_display(struct intel_vgpu *vgpu, u64 resolution);
+void intel_vgpu_reset_display(struct intel_vgpu *vgpu);
+void intel_vgpu_clean_display(struct intel_vgpu *vgpu);
+
+int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.c b/drivers/gpu/drm/i915/gvt/dmabuf.c
new file mode 100644
index 000000000..01e54b45c
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.c
@@ -0,0 +1,589 @@
+/*
+ * Copyright 2017 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Xiaoguang Chen
+ * Tina Zhang <tina.zhang@intel.com>
+ */
+
+#include <linux/dma-buf.h>
+#include <linux/mdev.h>
+
+#include <drm/drm_fourcc.h>
+#include <drm/drm_plane.h>
+
+#include "gem/i915_gem_dmabuf.h"
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gvt.h"
+
+#define GEN8_DECODE_PTE(pte) (pte & GENMASK_ULL(63, 12))
+
+static int vgpu_gem_get_pages(
+ struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
+ struct intel_vgpu *vgpu;
+ struct sg_table *st;
+ struct scatterlist *sg;
+ int i, j, ret;
+ gen8_pte_t __iomem *gtt_entries;
+ struct intel_vgpu_fb_info *fb_info;
+ u32 page_num;
+
+ fb_info = (struct intel_vgpu_fb_info *)obj->gvt_info;
+ if (drm_WARN_ON(&dev_priv->drm, !fb_info))
+ return -ENODEV;
+
+ vgpu = fb_info->obj->vgpu;
+ if (drm_WARN_ON(&dev_priv->drm, !vgpu))
+ return -ENODEV;
+
+ st = kmalloc(sizeof(*st), GFP_KERNEL);
+ if (unlikely(!st))
+ return -ENOMEM;
+
+ page_num = obj->base.size >> PAGE_SHIFT;
+ ret = sg_alloc_table(st, page_num, GFP_KERNEL);
+ if (ret) {
+ kfree(st);
+ return ret;
+ }
+ gtt_entries = (gen8_pte_t __iomem *)to_gt(dev_priv)->ggtt->gsm +
+ (fb_info->start >> PAGE_SHIFT);
+ for_each_sg(st->sgl, sg, page_num, i) {
+ dma_addr_t dma_addr =
+ GEN8_DECODE_PTE(readq(&gtt_entries[i]));
+ if (intel_gvt_dma_pin_guest_page(vgpu, dma_addr)) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ sg->offset = 0;
+ sg->length = PAGE_SIZE;
+ sg_dma_len(sg) = PAGE_SIZE;
+ sg_dma_address(sg) = dma_addr;
+ }
+
+ __i915_gem_object_set_pages(obj, st, PAGE_SIZE);
+out:
+ if (ret) {
+ dma_addr_t dma_addr;
+
+ for_each_sg(st->sgl, sg, i, j) {
+ dma_addr = sg_dma_address(sg);
+ if (dma_addr)
+ intel_gvt_dma_unmap_guest_page(vgpu, dma_addr);
+ }
+ sg_free_table(st);
+ kfree(st);
+ }
+
+ return ret;
+
+}
+
+static void vgpu_gem_put_pages(struct drm_i915_gem_object *obj,
+ struct sg_table *pages)
+{
+ struct scatterlist *sg;
+
+ if (obj->base.dma_buf) {
+ struct intel_vgpu_fb_info *fb_info = obj->gvt_info;
+ struct intel_vgpu_dmabuf_obj *obj = fb_info->obj;
+ struct intel_vgpu *vgpu = obj->vgpu;
+ int i;
+
+ for_each_sg(pages->sgl, sg, fb_info->size, i)
+ intel_gvt_dma_unmap_guest_page(vgpu,
+ sg_dma_address(sg));
+ }
+
+ sg_free_table(pages);
+ kfree(pages);
+}
+
+static void dmabuf_gem_object_free(struct kref *kref)
+{
+ struct intel_vgpu_dmabuf_obj *obj =
+ container_of(kref, struct intel_vgpu_dmabuf_obj, kref);
+ struct intel_vgpu *vgpu = obj->vgpu;
+ struct list_head *pos;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+
+ if (vgpu && vgpu->active && !list_empty(&vgpu->dmabuf_obj_list_head)) {
+ list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+ dmabuf_obj = list_entry(pos, struct intel_vgpu_dmabuf_obj, list);
+ if (dmabuf_obj == obj) {
+ list_del(pos);
+ idr_remove(&vgpu->object_idr,
+ dmabuf_obj->dmabuf_id);
+ kfree(dmabuf_obj->info);
+ kfree(dmabuf_obj);
+ break;
+ }
+ }
+ } else {
+ /* Free the orphan dmabuf_objs here */
+ kfree(obj->info);
+ kfree(obj);
+ }
+}
+
+
+static inline void dmabuf_obj_get(struct intel_vgpu_dmabuf_obj *obj)
+{
+ kref_get(&obj->kref);
+}
+
+static inline void dmabuf_obj_put(struct intel_vgpu_dmabuf_obj *obj)
+{
+ kref_put(&obj->kref, dmabuf_gem_object_free);
+}
+
+static void vgpu_gem_release(struct drm_i915_gem_object *gem_obj)
+{
+
+ struct intel_vgpu_fb_info *fb_info = gem_obj->gvt_info;
+ struct intel_vgpu_dmabuf_obj *obj = fb_info->obj;
+ struct intel_vgpu *vgpu = obj->vgpu;
+
+ if (vgpu) {
+ mutex_lock(&vgpu->dmabuf_lock);
+ gem_obj->base.dma_buf = NULL;
+ dmabuf_obj_put(obj);
+ mutex_unlock(&vgpu->dmabuf_lock);
+ } else {
+ /* vgpu is NULL, as it has been removed already */
+ gem_obj->base.dma_buf = NULL;
+ dmabuf_obj_put(obj);
+ }
+}
+
+static const struct drm_i915_gem_object_ops intel_vgpu_gem_ops = {
+ .name = "i915_gem_object_vgpu",
+ .flags = I915_GEM_OBJECT_IS_PROXY,
+ .get_pages = vgpu_gem_get_pages,
+ .put_pages = vgpu_gem_put_pages,
+ .release = vgpu_gem_release,
+};
+
+static struct drm_i915_gem_object *vgpu_create_gem(struct drm_device *dev,
+ struct intel_vgpu_fb_info *info)
+{
+ static struct lock_class_key lock_class;
+ struct drm_i915_private *dev_priv = to_i915(dev);
+ struct drm_i915_gem_object *obj;
+
+ obj = i915_gem_object_alloc();
+ if (obj == NULL)
+ return NULL;
+
+ drm_gem_private_object_init(dev, &obj->base,
+ roundup(info->size, PAGE_SIZE));
+ i915_gem_object_init(obj, &intel_vgpu_gem_ops, &lock_class, 0);
+ i915_gem_object_set_readonly(obj);
+
+ obj->read_domains = I915_GEM_DOMAIN_GTT;
+ obj->write_domain = 0;
+ if (GRAPHICS_VER(dev_priv) >= 9) {
+ unsigned int tiling_mode = 0;
+ unsigned int stride = 0;
+
+ switch (info->drm_format_mod) {
+ case DRM_FORMAT_MOD_LINEAR:
+ tiling_mode = I915_TILING_NONE;
+ break;
+ case I915_FORMAT_MOD_X_TILED:
+ tiling_mode = I915_TILING_X;
+ stride = info->stride;
+ break;
+ case I915_FORMAT_MOD_Y_TILED:
+ case I915_FORMAT_MOD_Yf_TILED:
+ tiling_mode = I915_TILING_Y;
+ stride = info->stride;
+ break;
+ default:
+ gvt_dbg_core("invalid drm_format_mod %llx for tiling\n",
+ info->drm_format_mod);
+ }
+ obj->tiling_and_stride = tiling_mode | stride;
+ } else {
+ obj->tiling_and_stride = info->drm_format_mod ?
+ I915_TILING_X : 0;
+ }
+
+ return obj;
+}
+
+static bool validate_hotspot(struct intel_vgpu_cursor_plane_format *c)
+{
+ if (c && c->x_hot <= c->width && c->y_hot <= c->height)
+ return true;
+ else
+ return false;
+}
+
+static int vgpu_get_plane_info(struct drm_device *dev,
+ struct intel_vgpu *vgpu,
+ struct intel_vgpu_fb_info *info,
+ int plane_id)
+{
+ struct intel_vgpu_primary_plane_format p;
+ struct intel_vgpu_cursor_plane_format c;
+ int ret, tile_height = 1;
+
+ memset(info, 0, sizeof(*info));
+
+ if (plane_id == DRM_PLANE_TYPE_PRIMARY) {
+ ret = intel_vgpu_decode_primary_plane(vgpu, &p);
+ if (ret)
+ return ret;
+ info->start = p.base;
+ info->start_gpa = p.base_gpa;
+ info->width = p.width;
+ info->height = p.height;
+ info->stride = p.stride;
+ info->drm_format = p.drm_format;
+
+ switch (p.tiled) {
+ case PLANE_CTL_TILED_LINEAR:
+ info->drm_format_mod = DRM_FORMAT_MOD_LINEAR;
+ break;
+ case PLANE_CTL_TILED_X:
+ info->drm_format_mod = I915_FORMAT_MOD_X_TILED;
+ tile_height = 8;
+ break;
+ case PLANE_CTL_TILED_Y:
+ info->drm_format_mod = I915_FORMAT_MOD_Y_TILED;
+ tile_height = 32;
+ break;
+ case PLANE_CTL_TILED_YF:
+ info->drm_format_mod = I915_FORMAT_MOD_Yf_TILED;
+ tile_height = 32;
+ break;
+ default:
+ gvt_vgpu_err("invalid tiling mode: %x\n", p.tiled);
+ }
+ } else if (plane_id == DRM_PLANE_TYPE_CURSOR) {
+ ret = intel_vgpu_decode_cursor_plane(vgpu, &c);
+ if (ret)
+ return ret;
+ info->start = c.base;
+ info->start_gpa = c.base_gpa;
+ info->width = c.width;
+ info->height = c.height;
+ info->stride = c.width * (c.bpp / 8);
+ info->drm_format = c.drm_format;
+ info->drm_format_mod = 0;
+ info->x_pos = c.x_pos;
+ info->y_pos = c.y_pos;
+
+ if (validate_hotspot(&c)) {
+ info->x_hot = c.x_hot;
+ info->y_hot = c.y_hot;
+ } else {
+ info->x_hot = UINT_MAX;
+ info->y_hot = UINT_MAX;
+ }
+ } else {
+ gvt_vgpu_err("invalid plane id:%d\n", plane_id);
+ return -EINVAL;
+ }
+
+ info->size = info->stride * roundup(info->height, tile_height);
+ if (info->size == 0) {
+ gvt_vgpu_err("fb size is zero\n");
+ return -EINVAL;
+ }
+
+ if (info->start & (PAGE_SIZE - 1)) {
+ gvt_vgpu_err("Not aligned fb address:0x%llx\n", info->start);
+ return -EFAULT;
+ }
+
+ if (!intel_gvt_ggtt_validate_range(vgpu, info->start, info->size)) {
+ gvt_vgpu_err("invalid gma addr\n");
+ return -EFAULT;
+ }
+
+ return 0;
+}
+
+static struct intel_vgpu_dmabuf_obj *
+pick_dmabuf_by_info(struct intel_vgpu *vgpu,
+ struct intel_vgpu_fb_info *latest_info)
+{
+ struct list_head *pos;
+ struct intel_vgpu_fb_info *fb_info;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj = NULL;
+ struct intel_vgpu_dmabuf_obj *ret = NULL;
+
+ list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+ dmabuf_obj = list_entry(pos, struct intel_vgpu_dmabuf_obj, list);
+ if (!dmabuf_obj->info)
+ continue;
+
+ fb_info = (struct intel_vgpu_fb_info *)dmabuf_obj->info;
+ if ((fb_info->start == latest_info->start) &&
+ (fb_info->start_gpa == latest_info->start_gpa) &&
+ (fb_info->size == latest_info->size) &&
+ (fb_info->drm_format_mod == latest_info->drm_format_mod) &&
+ (fb_info->drm_format == latest_info->drm_format) &&
+ (fb_info->width == latest_info->width) &&
+ (fb_info->height == latest_info->height)) {
+ ret = dmabuf_obj;
+ break;
+ }
+ }
+
+ return ret;
+}
+
+static struct intel_vgpu_dmabuf_obj *
+pick_dmabuf_by_num(struct intel_vgpu *vgpu, u32 id)
+{
+ struct list_head *pos;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj = NULL;
+ struct intel_vgpu_dmabuf_obj *ret = NULL;
+
+ list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+ dmabuf_obj = list_entry(pos, struct intel_vgpu_dmabuf_obj, list);
+ if (dmabuf_obj->dmabuf_id == id) {
+ ret = dmabuf_obj;
+ break;
+ }
+ }
+
+ return ret;
+}
+
+static void update_fb_info(struct vfio_device_gfx_plane_info *gvt_dmabuf,
+ struct intel_vgpu_fb_info *fb_info)
+{
+ gvt_dmabuf->drm_format = fb_info->drm_format;
+ gvt_dmabuf->drm_format_mod = fb_info->drm_format_mod;
+ gvt_dmabuf->width = fb_info->width;
+ gvt_dmabuf->height = fb_info->height;
+ gvt_dmabuf->stride = fb_info->stride;
+ gvt_dmabuf->size = fb_info->size;
+ gvt_dmabuf->x_pos = fb_info->x_pos;
+ gvt_dmabuf->y_pos = fb_info->y_pos;
+ gvt_dmabuf->x_hot = fb_info->x_hot;
+ gvt_dmabuf->y_hot = fb_info->y_hot;
+}
+
+int intel_vgpu_query_plane(struct intel_vgpu *vgpu, void *args)
+{
+ struct drm_device *dev = &vgpu->gvt->gt->i915->drm;
+ struct vfio_device_gfx_plane_info *gfx_plane_info = args;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+ struct intel_vgpu_fb_info fb_info;
+ int ret = 0;
+
+ if (gfx_plane_info->flags == (VFIO_GFX_PLANE_TYPE_DMABUF |
+ VFIO_GFX_PLANE_TYPE_PROBE))
+ return ret;
+ else if ((gfx_plane_info->flags & ~VFIO_GFX_PLANE_TYPE_DMABUF) ||
+ (!gfx_plane_info->flags))
+ return -EINVAL;
+
+ ret = vgpu_get_plane_info(dev, vgpu, &fb_info,
+ gfx_plane_info->drm_plane_type);
+ if (ret != 0)
+ goto out;
+
+ mutex_lock(&vgpu->dmabuf_lock);
+ /* If exists, pick up the exposed dmabuf_obj */
+ dmabuf_obj = pick_dmabuf_by_info(vgpu, &fb_info);
+ if (dmabuf_obj) {
+ update_fb_info(gfx_plane_info, &fb_info);
+ gfx_plane_info->dmabuf_id = dmabuf_obj->dmabuf_id;
+
+ /* This buffer may be released between query_plane ioctl and
+ * get_dmabuf ioctl. Add the refcount to make sure it won't
+ * be released between the two ioctls.
+ */
+ if (!dmabuf_obj->initref) {
+ dmabuf_obj->initref = true;
+ dmabuf_obj_get(dmabuf_obj);
+ }
+ ret = 0;
+ gvt_dbg_dpy("vgpu%d: re-use dmabuf_obj ref %d, id %d\n",
+ vgpu->id, kref_read(&dmabuf_obj->kref),
+ gfx_plane_info->dmabuf_id);
+ mutex_unlock(&vgpu->dmabuf_lock);
+ goto out;
+ }
+
+ mutex_unlock(&vgpu->dmabuf_lock);
+
+ /* Need to allocate a new one*/
+ dmabuf_obj = kmalloc(sizeof(struct intel_vgpu_dmabuf_obj), GFP_KERNEL);
+ if (unlikely(!dmabuf_obj)) {
+ gvt_vgpu_err("alloc dmabuf_obj failed\n");
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ dmabuf_obj->info = kmalloc(sizeof(struct intel_vgpu_fb_info),
+ GFP_KERNEL);
+ if (unlikely(!dmabuf_obj->info)) {
+ gvt_vgpu_err("allocate intel vgpu fb info failed\n");
+ ret = -ENOMEM;
+ goto out_free_dmabuf;
+ }
+ memcpy(dmabuf_obj->info, &fb_info, sizeof(struct intel_vgpu_fb_info));
+
+ ((struct intel_vgpu_fb_info *)dmabuf_obj->info)->obj = dmabuf_obj;
+
+ dmabuf_obj->vgpu = vgpu;
+
+ ret = idr_alloc(&vgpu->object_idr, dmabuf_obj, 1, 0, GFP_NOWAIT);
+ if (ret < 0)
+ goto out_free_info;
+ gfx_plane_info->dmabuf_id = ret;
+ dmabuf_obj->dmabuf_id = ret;
+
+ dmabuf_obj->initref = true;
+
+ kref_init(&dmabuf_obj->kref);
+
+ update_fb_info(gfx_plane_info, &fb_info);
+
+ INIT_LIST_HEAD(&dmabuf_obj->list);
+ mutex_lock(&vgpu->dmabuf_lock);
+ list_add_tail(&dmabuf_obj->list, &vgpu->dmabuf_obj_list_head);
+ mutex_unlock(&vgpu->dmabuf_lock);
+
+ gvt_dbg_dpy("vgpu%d: %s new dmabuf_obj ref %d, id %d\n", vgpu->id,
+ __func__, kref_read(&dmabuf_obj->kref), ret);
+
+ return 0;
+
+out_free_info:
+ kfree(dmabuf_obj->info);
+out_free_dmabuf:
+ kfree(dmabuf_obj);
+out:
+ /* ENODEV means plane isn't ready, which might be a normal case. */
+ return (ret == -ENODEV) ? 0 : ret;
+}
+
+/* To associate an exposed dmabuf with the dmabuf_obj */
+int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id)
+{
+ struct drm_device *dev = &vgpu->gvt->gt->i915->drm;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+ struct drm_i915_gem_object *obj;
+ struct dma_buf *dmabuf;
+ int dmabuf_fd;
+ int ret = 0;
+
+ mutex_lock(&vgpu->dmabuf_lock);
+
+ dmabuf_obj = pick_dmabuf_by_num(vgpu, dmabuf_id);
+ if (dmabuf_obj == NULL) {
+ gvt_vgpu_err("invalid dmabuf id:%d\n", dmabuf_id);
+ ret = -EINVAL;
+ goto out;
+ }
+
+ obj = vgpu_create_gem(dev, dmabuf_obj->info);
+ if (obj == NULL) {
+ gvt_vgpu_err("create gvt gem obj failed\n");
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ obj->gvt_info = dmabuf_obj->info;
+
+ dmabuf = i915_gem_prime_export(&obj->base, DRM_CLOEXEC | DRM_RDWR);
+ if (IS_ERR(dmabuf)) {
+ gvt_vgpu_err("export dma-buf failed\n");
+ ret = PTR_ERR(dmabuf);
+ goto out_free_gem;
+ }
+
+ ret = dma_buf_fd(dmabuf, DRM_CLOEXEC | DRM_RDWR);
+ if (ret < 0) {
+ gvt_vgpu_err("create dma-buf fd failed ret:%d\n", ret);
+ goto out_free_dmabuf;
+ }
+ dmabuf_fd = ret;
+
+ dmabuf_obj_get(dmabuf_obj);
+
+ if (dmabuf_obj->initref) {
+ dmabuf_obj->initref = false;
+ dmabuf_obj_put(dmabuf_obj);
+ }
+
+ mutex_unlock(&vgpu->dmabuf_lock);
+
+ gvt_dbg_dpy("vgpu%d: dmabuf:%d, dmabuf ref %d, fd:%d\n"
+ " file count: %ld, GEM ref: %d\n",
+ vgpu->id, dmabuf_obj->dmabuf_id,
+ kref_read(&dmabuf_obj->kref),
+ dmabuf_fd,
+ file_count(dmabuf->file),
+ kref_read(&obj->base.refcount));
+
+ i915_gem_object_put(obj);
+
+ return dmabuf_fd;
+
+out_free_dmabuf:
+ dma_buf_put(dmabuf);
+out_free_gem:
+ i915_gem_object_put(obj);
+out:
+ mutex_unlock(&vgpu->dmabuf_lock);
+ return ret;
+}
+
+void intel_vgpu_dmabuf_cleanup(struct intel_vgpu *vgpu)
+{
+ struct list_head *pos, *n;
+ struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+
+ mutex_lock(&vgpu->dmabuf_lock);
+ list_for_each_safe(pos, n, &vgpu->dmabuf_obj_list_head) {
+ dmabuf_obj = list_entry(pos, struct intel_vgpu_dmabuf_obj, list);
+ dmabuf_obj->vgpu = NULL;
+
+ idr_remove(&vgpu->object_idr, dmabuf_obj->dmabuf_id);
+ list_del(pos);
+
+ /* dmabuf_obj might be freed in dmabuf_obj_put */
+ if (dmabuf_obj->initref) {
+ dmabuf_obj->initref = false;
+ dmabuf_obj_put(dmabuf_obj);
+ }
+
+ }
+ mutex_unlock(&vgpu->dmabuf_lock);
+}
diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.h b/drivers/gpu/drm/i915/gvt/dmabuf.h
new file mode 100644
index 000000000..5f8f03fb1
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright(c) 2017 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Xiaoguang Chen
+ * Tina Zhang <tina.zhang@intel.com>
+ */
+
+#ifndef _GVT_DMABUF_H_
+#define _GVT_DMABUF_H_
+#include <linux/vfio.h>
+
+struct intel_vgpu_fb_info {
+ __u64 start;
+ __u64 start_gpa;
+ __u64 drm_format_mod;
+ __u32 drm_format; /* drm format of plane */
+ __u32 width; /* width of plane */
+ __u32 height; /* height of plane */
+ __u32 stride; /* stride of plane */
+ __u32 size; /* size of plane in bytes, align on page */
+ __u32 x_pos; /* horizontal position of cursor plane */
+ __u32 y_pos; /* vertical position of cursor plane */
+ __u32 x_hot; /* horizontal position of cursor hotspot */
+ __u32 y_hot; /* vertical position of cursor hotspot */
+ struct intel_vgpu_dmabuf_obj *obj;
+};
+
+/**
+ * struct intel_vgpu_dmabuf_obj- Intel vGPU device buffer object
+ */
+struct intel_vgpu_dmabuf_obj {
+ struct intel_vgpu *vgpu;
+ struct intel_vgpu_fb_info *info;
+ __u32 dmabuf_id;
+ struct kref kref;
+ bool initref;
+ struct list_head list;
+};
+
+int intel_vgpu_query_plane(struct intel_vgpu *vgpu, void *args);
+int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id);
+void intel_vgpu_dmabuf_cleanup(struct intel_vgpu *vgpu);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/edid.c b/drivers/gpu/drm/i915/gvt/edid.c
new file mode 100644
index 000000000..1b509c1a1
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/edid.c
@@ -0,0 +1,585 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Terrence Xu <terrence.xu@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "display/intel_gmbus_regs.h"
+#include "gvt.h"
+#include "i915_drv.h"
+#include "i915_reg.h"
+
+#define GMBUS1_TOTAL_BYTES_SHIFT 16
+#define GMBUS1_TOTAL_BYTES_MASK 0x1ff
+#define gmbus1_total_byte_count(v) (((v) >> \
+ GMBUS1_TOTAL_BYTES_SHIFT) & GMBUS1_TOTAL_BYTES_MASK)
+#define gmbus1_slave_addr(v) (((v) & 0xff) >> 1)
+#define gmbus1_slave_index(v) (((v) >> 8) & 0xff)
+#define gmbus1_bus_cycle(v) (((v) >> 25) & 0x7)
+
+/* GMBUS0 bits definitions */
+#define _GMBUS_PIN_SEL_MASK (0x7)
+
+static unsigned char edid_get_byte(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_i2c_edid *edid = &vgpu->display.i2c_edid;
+ unsigned char chr = 0;
+
+ if (edid->state == I2C_NOT_SPECIFIED || !edid->slave_selected) {
+ gvt_vgpu_err("Driver tries to read EDID without proper sequence!\n");
+ return 0;
+ }
+ if (edid->current_edid_read >= EDID_SIZE) {
+ gvt_vgpu_err("edid_get_byte() exceeds the size of EDID!\n");
+ return 0;
+ }
+
+ if (!edid->edid_available) {
+ gvt_vgpu_err("Reading EDID but EDID is not available!\n");
+ return 0;
+ }
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, edid->port)) {
+ struct intel_vgpu_edid_data *edid_data =
+ intel_vgpu_port(vgpu, edid->port)->edid;
+
+ chr = edid_data->edid_block[edid->current_edid_read];
+ edid->current_edid_read++;
+ } else {
+ gvt_vgpu_err("No EDID available during the reading?\n");
+ }
+ return chr;
+}
+
+static inline int cnp_get_port_from_gmbus0(u32 gmbus0)
+{
+ int port_select = gmbus0 & _GMBUS_PIN_SEL_MASK;
+ int port = -EINVAL;
+
+ if (port_select == GMBUS_PIN_1_BXT)
+ port = PORT_B;
+ else if (port_select == GMBUS_PIN_2_BXT)
+ port = PORT_C;
+ else if (port_select == GMBUS_PIN_3_BXT)
+ port = PORT_D;
+ else if (port_select == GMBUS_PIN_4_CNP)
+ port = PORT_E;
+ return port;
+}
+
+static inline int bxt_get_port_from_gmbus0(u32 gmbus0)
+{
+ int port_select = gmbus0 & _GMBUS_PIN_SEL_MASK;
+ int port = -EINVAL;
+
+ if (port_select == GMBUS_PIN_1_BXT)
+ port = PORT_B;
+ else if (port_select == GMBUS_PIN_2_BXT)
+ port = PORT_C;
+ else if (port_select == GMBUS_PIN_3_BXT)
+ port = PORT_D;
+ return port;
+}
+
+static inline int get_port_from_gmbus0(u32 gmbus0)
+{
+ int port_select = gmbus0 & _GMBUS_PIN_SEL_MASK;
+ int port = -EINVAL;
+
+ if (port_select == GMBUS_PIN_VGADDC)
+ port = PORT_E;
+ else if (port_select == GMBUS_PIN_DPC)
+ port = PORT_C;
+ else if (port_select == GMBUS_PIN_DPB)
+ port = PORT_B;
+ else if (port_select == GMBUS_PIN_DPD)
+ port = PORT_D;
+ return port;
+}
+
+static void reset_gmbus_controller(struct intel_vgpu *vgpu)
+{
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) = GMBUS_HW_RDY;
+ if (!vgpu->display.i2c_edid.edid_available)
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+ vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
+}
+
+/* GMBUS0 */
+static int gmbus0_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ int port, pin_select;
+
+ memcpy(&vgpu_vreg(vgpu, offset), p_data, bytes);
+
+ pin_select = vgpu_vreg(vgpu, offset) & _GMBUS_PIN_SEL_MASK;
+
+ intel_vgpu_init_i2c_edid(vgpu);
+
+ if (pin_select == 0)
+ return 0;
+
+ if (IS_BROXTON(i915))
+ port = bxt_get_port_from_gmbus0(pin_select);
+ else if (IS_COFFEELAKE(i915) || IS_COMETLAKE(i915))
+ port = cnp_get_port_from_gmbus0(pin_select);
+ else
+ port = get_port_from_gmbus0(pin_select);
+ if (drm_WARN_ON(&i915->drm, port < 0))
+ return 0;
+
+ vgpu->display.i2c_edid.state = I2C_GMBUS;
+ vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
+
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY | GMBUS_HW_WAIT_PHASE;
+
+ if (intel_vgpu_has_monitor_on_port(vgpu, port) &&
+ !intel_vgpu_port_is_dp(vgpu, port)) {
+ vgpu->display.i2c_edid.port = port;
+ vgpu->display.i2c_edid.edid_available = true;
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_SATOER;
+ } else
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+ return 0;
+}
+
+static int gmbus1_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+ u32 slave_addr;
+ u32 wvalue = *(u32 *)p_data;
+
+ if (vgpu_vreg(vgpu, offset) & GMBUS_SW_CLR_INT) {
+ if (!(wvalue & GMBUS_SW_CLR_INT)) {
+ vgpu_vreg(vgpu, offset) &= ~GMBUS_SW_CLR_INT;
+ reset_gmbus_controller(vgpu);
+ }
+ /*
+ * TODO: "This bit is cleared to zero when an event
+ * causes the HW_RDY bit transition to occur "
+ */
+ } else {
+ /*
+ * per bspec setting this bit can cause:
+ * 1) INT status bit cleared
+ * 2) HW_RDY bit asserted
+ */
+ if (wvalue & GMBUS_SW_CLR_INT) {
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_INT;
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY;
+ }
+
+ /* For virtualization, we suppose that HW is always ready,
+ * so GMBUS_SW_RDY should always be cleared
+ */
+ if (wvalue & GMBUS_SW_RDY)
+ wvalue &= ~GMBUS_SW_RDY;
+
+ i2c_edid->gmbus.total_byte_count =
+ gmbus1_total_byte_count(wvalue);
+ slave_addr = gmbus1_slave_addr(wvalue);
+
+ /* vgpu gmbus only support EDID */
+ if (slave_addr == EDID_ADDR) {
+ i2c_edid->slave_selected = true;
+ } else if (slave_addr != 0) {
+ gvt_dbg_dpy(
+ "vgpu%d: unsupported gmbus slave addr(0x%x)\n"
+ " gmbus operations will be ignored.\n",
+ vgpu->id, slave_addr);
+ }
+
+ if (wvalue & GMBUS_CYCLE_INDEX)
+ i2c_edid->current_edid_read =
+ gmbus1_slave_index(wvalue);
+
+ i2c_edid->gmbus.cycle_type = gmbus1_bus_cycle(wvalue);
+ switch (gmbus1_bus_cycle(wvalue)) {
+ case GMBUS_NOCYCLE:
+ break;
+ case GMBUS_STOP:
+ /* From spec:
+ * This can only cause a STOP to be generated
+ * if a GMBUS cycle is generated, the GMBUS is
+ * currently in a data/wait/idle phase, or it is in a
+ * WAIT phase
+ */
+ if (gmbus1_bus_cycle(vgpu_vreg(vgpu, offset))
+ != GMBUS_NOCYCLE) {
+ intel_vgpu_init_i2c_edid(vgpu);
+ /* After the 'stop' cycle, hw state would become
+ * 'stop phase' and then 'idle phase' after a
+ * few milliseconds. In emulation, we just set
+ * it as 'idle phase' ('stop phase' is not
+ * visible in gmbus interface)
+ */
+ i2c_edid->gmbus.phase = GMBUS_IDLE_PHASE;
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+ }
+ break;
+ case NIDX_NS_W:
+ case IDX_NS_W:
+ case NIDX_STOP:
+ case IDX_STOP:
+ /* From hw spec the GMBUS phase
+ * transition like this:
+ * START (-->INDEX) -->DATA
+ */
+ i2c_edid->gmbus.phase = GMBUS_DATA_PHASE;
+ vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_ACTIVE;
+ break;
+ default:
+ gvt_vgpu_err("Unknown/reserved GMBUS cycle detected!\n");
+ break;
+ }
+ /*
+ * From hw spec the WAIT state will be
+ * cleared:
+ * (1) in a new GMBUS cycle
+ * (2) by generating a stop
+ */
+ vgpu_vreg(vgpu, offset) = wvalue;
+ }
+ return 0;
+}
+
+static int gmbus3_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ drm_WARN_ON(&i915->drm, 1);
+ return 0;
+}
+
+static int gmbus3_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ int i;
+ unsigned char byte_data;
+ struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+ int byte_left = i2c_edid->gmbus.total_byte_count -
+ i2c_edid->current_edid_read;
+ int byte_count = byte_left;
+ u32 reg_data = 0;
+
+ /* Data can only be recevied if previous settings correct */
+ if (vgpu_vreg_t(vgpu, PCH_GMBUS1) & GMBUS_SLAVE_READ) {
+ if (byte_left <= 0) {
+ memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+ return 0;
+ }
+
+ if (byte_count > 4)
+ byte_count = 4;
+ for (i = 0; i < byte_count; i++) {
+ byte_data = edid_get_byte(vgpu);
+ reg_data |= (byte_data << (i << 3));
+ }
+
+ memcpy(&vgpu_vreg(vgpu, offset), &reg_data, byte_count);
+ memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+
+ if (byte_left <= 4) {
+ switch (i2c_edid->gmbus.cycle_type) {
+ case NIDX_STOP:
+ case IDX_STOP:
+ i2c_edid->gmbus.phase = GMBUS_IDLE_PHASE;
+ break;
+ case NIDX_NS_W:
+ case IDX_NS_W:
+ default:
+ i2c_edid->gmbus.phase = GMBUS_WAIT_PHASE;
+ break;
+ }
+ intel_vgpu_init_i2c_edid(vgpu);
+ }
+ /*
+ * Read GMBUS3 during send operation,
+ * return the latest written value
+ */
+ } else {
+ memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+ gvt_vgpu_err("warning: gmbus3 read with nothing returned\n");
+ }
+ return 0;
+}
+
+static int gmbus2_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 value = vgpu_vreg(vgpu, offset);
+
+ if (!(vgpu_vreg(vgpu, offset) & GMBUS_INUSE))
+ vgpu_vreg(vgpu, offset) |= GMBUS_INUSE;
+ memcpy(p_data, (void *)&value, bytes);
+ return 0;
+}
+
+static int gmbus2_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 wvalue = *(u32 *)p_data;
+
+ if (wvalue & GMBUS_INUSE)
+ vgpu_vreg(vgpu, offset) &= ~GMBUS_INUSE;
+ /* All other bits are read-only */
+ return 0;
+}
+
+/**
+ * intel_gvt_i2c_handle_gmbus_read - emulate gmbus register mmio read
+ * @vgpu: a vGPU
+ * @offset: reg offset
+ * @p_data: data return buffer
+ * @bytes: access data length
+ *
+ * This function is used to emulate gmbus register mmio read
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_gvt_i2c_handle_gmbus_read(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ if (drm_WARN_ON(&i915->drm, bytes > 8 && (offset & (bytes - 1))))
+ return -EINVAL;
+
+ if (offset == i915_mmio_reg_offset(PCH_GMBUS2))
+ return gmbus2_mmio_read(vgpu, offset, p_data, bytes);
+ else if (offset == i915_mmio_reg_offset(PCH_GMBUS3))
+ return gmbus3_mmio_read(vgpu, offset, p_data, bytes);
+
+ memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+ return 0;
+}
+
+/**
+ * intel_gvt_i2c_handle_gmbus_write - emulate gmbus register mmio write
+ * @vgpu: a vGPU
+ * @offset: reg offset
+ * @p_data: data return buffer
+ * @bytes: access data length
+ *
+ * This function is used to emulate gmbus register mmio write
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_gvt_i2c_handle_gmbus_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ if (drm_WARN_ON(&i915->drm, bytes > 8 && (offset & (bytes - 1))))
+ return -EINVAL;
+
+ if (offset == i915_mmio_reg_offset(PCH_GMBUS0))
+ return gmbus0_mmio_write(vgpu, offset, p_data, bytes);
+ else if (offset == i915_mmio_reg_offset(PCH_GMBUS1))
+ return gmbus1_mmio_write(vgpu, offset, p_data, bytes);
+ else if (offset == i915_mmio_reg_offset(PCH_GMBUS2))
+ return gmbus2_mmio_write(vgpu, offset, p_data, bytes);
+ else if (offset == i915_mmio_reg_offset(PCH_GMBUS3))
+ return gmbus3_mmio_write(vgpu, offset, p_data, bytes);
+
+ memcpy(&vgpu_vreg(vgpu, offset), p_data, bytes);
+ return 0;
+}
+
+enum {
+ AUX_CH_CTL = 0,
+ AUX_CH_DATA1,
+ AUX_CH_DATA2,
+ AUX_CH_DATA3,
+ AUX_CH_DATA4,
+ AUX_CH_DATA5
+};
+
+static inline int get_aux_ch_reg(unsigned int offset)
+{
+ int reg;
+
+ switch (offset & 0xff) {
+ case 0x10:
+ reg = AUX_CH_CTL;
+ break;
+ case 0x14:
+ reg = AUX_CH_DATA1;
+ break;
+ case 0x18:
+ reg = AUX_CH_DATA2;
+ break;
+ case 0x1c:
+ reg = AUX_CH_DATA3;
+ break;
+ case 0x20:
+ reg = AUX_CH_DATA4;
+ break;
+ case 0x24:
+ reg = AUX_CH_DATA5;
+ break;
+ default:
+ reg = -1;
+ break;
+ }
+ return reg;
+}
+
+#define AUX_CTL_MSG_LENGTH(reg) \
+ ((reg & DP_AUX_CH_CTL_MESSAGE_SIZE_MASK) >> \
+ DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT)
+
+/**
+ * intel_gvt_i2c_handle_aux_ch_write - emulate AUX channel register write
+ * @vgpu: a vGPU
+ * @port_idx: port index
+ * @offset: reg offset
+ * @p_data: write ptr
+ *
+ * This function is used to emulate AUX channel register write
+ *
+ */
+void intel_gvt_i2c_handle_aux_ch_write(struct intel_vgpu *vgpu,
+ int port_idx,
+ unsigned int offset,
+ void *p_data)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+ int msg_length, ret_msg_size;
+ int msg, addr, ctrl, op;
+ u32 value = *(u32 *)p_data;
+ int aux_data_for_write = 0;
+ int reg = get_aux_ch_reg(offset);
+
+ if (reg != AUX_CH_CTL) {
+ vgpu_vreg(vgpu, offset) = value;
+ return;
+ }
+
+ msg_length = AUX_CTL_MSG_LENGTH(value);
+ // check the msg in DATA register.
+ msg = vgpu_vreg(vgpu, offset + 4);
+ addr = (msg >> 8) & 0xffff;
+ ctrl = (msg >> 24) & 0xff;
+ op = ctrl >> 4;
+ if (!(value & DP_AUX_CH_CTL_SEND_BUSY)) {
+ /* The ctl write to clear some states */
+ return;
+ }
+
+ /* Always set the wanted value for vms. */
+ ret_msg_size = (((op & 0x1) == GVT_AUX_I2C_READ) ? 2 : 1);
+ vgpu_vreg(vgpu, offset) =
+ DP_AUX_CH_CTL_DONE |
+ ((ret_msg_size << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) &
+ DP_AUX_CH_CTL_MESSAGE_SIZE_MASK);
+
+ if (msg_length == 3) {
+ if (!(op & GVT_AUX_I2C_MOT)) {
+ /* stop */
+ intel_vgpu_init_i2c_edid(vgpu);
+ } else {
+ /* start or restart */
+ i2c_edid->aux_ch.i2c_over_aux_ch = true;
+ i2c_edid->aux_ch.aux_ch_mot = true;
+ if (addr == 0) {
+ /* reset the address */
+ intel_vgpu_init_i2c_edid(vgpu);
+ } else if (addr == EDID_ADDR) {
+ i2c_edid->state = I2C_AUX_CH;
+ i2c_edid->port = port_idx;
+ i2c_edid->slave_selected = true;
+ if (intel_vgpu_has_monitor_on_port(vgpu,
+ port_idx) &&
+ intel_vgpu_port_is_dp(vgpu, port_idx))
+ i2c_edid->edid_available = true;
+ }
+ }
+ } else if ((op & 0x1) == GVT_AUX_I2C_WRITE) {
+ /* TODO
+ * We only support EDID reading from I2C_over_AUX. And
+ * we do not expect the index mode to be used. Right now
+ * the WRITE operation is ignored. It is good enough to
+ * support the gfx driver to do EDID access.
+ */
+ } else {
+ if (drm_WARN_ON(&i915->drm, (op & 0x1) != GVT_AUX_I2C_READ))
+ return;
+ if (drm_WARN_ON(&i915->drm, msg_length != 4))
+ return;
+ if (i2c_edid->edid_available && i2c_edid->slave_selected) {
+ unsigned char val = edid_get_byte(vgpu);
+
+ aux_data_for_write = (val << 16);
+ } else
+ aux_data_for_write = (0xff << 16);
+ }
+ /* write the return value in AUX_CH_DATA reg which includes:
+ * ACK of I2C_WRITE
+ * returned byte if it is READ
+ */
+ aux_data_for_write |= GVT_AUX_I2C_REPLY_ACK << 24;
+ vgpu_vreg(vgpu, offset + 4) = aux_data_for_write;
+}
+
+/**
+ * intel_vgpu_init_i2c_edid - initialize vGPU i2c edid emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to initialize vGPU i2c edid emulation stuffs
+ *
+ */
+void intel_vgpu_init_i2c_edid(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_i2c_edid *edid = &vgpu->display.i2c_edid;
+
+ edid->state = I2C_NOT_SPECIFIED;
+
+ edid->port = -1;
+ edid->slave_selected = false;
+ edid->edid_available = false;
+ edid->current_edid_read = 0;
+
+ memset(&edid->gmbus, 0, sizeof(struct intel_vgpu_i2c_gmbus));
+
+ edid->aux_ch.i2c_over_aux_ch = false;
+ edid->aux_ch.aux_ch_mot = false;
+}
diff --git a/drivers/gpu/drm/i915/gvt/edid.h b/drivers/gpu/drm/i915/gvt/edid.h
new file mode 100644
index 000000000..dfe0cbc6a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/edid.h
@@ -0,0 +1,154 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Terrence Xu <terrence.xu@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_EDID_H_
+#define _GVT_EDID_H_
+
+#include <linux/types.h>
+
+struct intel_vgpu;
+
+#define EDID_SIZE 128
+#define EDID_ADDR 0x50 /* Linux hvm EDID addr */
+
+#define GVT_AUX_NATIVE_WRITE 0x8
+#define GVT_AUX_NATIVE_READ 0x9
+#define GVT_AUX_I2C_WRITE 0x0
+#define GVT_AUX_I2C_READ 0x1
+#define GVT_AUX_I2C_STATUS 0x2
+#define GVT_AUX_I2C_MOT 0x4
+#define GVT_AUX_I2C_REPLY_ACK 0x0
+
+struct intel_vgpu_edid_data {
+ bool data_valid;
+ unsigned char edid_block[EDID_SIZE];
+};
+
+enum gmbus_cycle_type {
+ GMBUS_NOCYCLE = 0x0,
+ NIDX_NS_W = 0x1,
+ IDX_NS_W = 0x3,
+ GMBUS_STOP = 0x4,
+ NIDX_STOP = 0x5,
+ IDX_STOP = 0x7
+};
+
+/*
+ * States of GMBUS
+ *
+ * GMBUS0-3 could be related to the EDID virtualization. Another two GMBUS
+ * registers, GMBUS4 (interrupt mask) and GMBUS5 (2 byte indes register), are
+ * not considered here. Below describes the usage of GMBUS registers that are
+ * cared by the EDID virtualization
+ *
+ * GMBUS0:
+ * R/W
+ * port selection. value of bit0 - bit2 corresponds to the GPIO registers.
+ *
+ * GMBUS1:
+ * R/W Protect
+ * Command and Status.
+ * bit0 is the direction bit: 1 is read; 0 is write.
+ * bit1 - bit7 is slave 7-bit address.
+ * bit16 - bit24 total byte count (ignore?)
+ *
+ * GMBUS2:
+ * Most of bits are read only except bit 15 (IN_USE)
+ * Status register
+ * bit0 - bit8 current byte count
+ * bit 11: hardware ready;
+ *
+ * GMBUS3:
+ * Read/Write
+ * Data for transfer
+ */
+
+/* From hw specs, Other phases like START, ADDRESS, INDEX
+ * are invisible to GMBUS MMIO interface. So no definitions
+ * in below enum types
+ */
+enum gvt_gmbus_phase {
+ GMBUS_IDLE_PHASE = 0,
+ GMBUS_DATA_PHASE,
+ GMBUS_WAIT_PHASE,
+ //GMBUS_STOP_PHASE,
+ GMBUS_MAX_PHASE
+};
+
+struct intel_vgpu_i2c_gmbus {
+ unsigned int total_byte_count; /* from GMBUS1 */
+ enum gmbus_cycle_type cycle_type;
+ enum gvt_gmbus_phase phase;
+};
+
+struct intel_vgpu_i2c_aux_ch {
+ bool i2c_over_aux_ch;
+ bool aux_ch_mot;
+};
+
+enum i2c_state {
+ I2C_NOT_SPECIFIED = 0,
+ I2C_GMBUS = 1,
+ I2C_AUX_CH = 2
+};
+
+/* I2C sequences cannot interleave.
+ * GMBUS and AUX_CH sequences cannot interleave.
+ */
+struct intel_vgpu_i2c_edid {
+ enum i2c_state state;
+
+ unsigned int port;
+ bool slave_selected;
+ bool edid_available;
+ unsigned int current_edid_read;
+
+ struct intel_vgpu_i2c_gmbus gmbus;
+ struct intel_vgpu_i2c_aux_ch aux_ch;
+};
+
+void intel_vgpu_init_i2c_edid(struct intel_vgpu *vgpu);
+
+int intel_gvt_i2c_handle_gmbus_read(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes);
+
+int intel_gvt_i2c_handle_gmbus_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes);
+
+void intel_gvt_i2c_handle_aux_ch_write(struct intel_vgpu *vgpu,
+ int port_idx,
+ unsigned int offset,
+ void *p_data);
+
+#endif /*_GVT_EDID_H_*/
diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
new file mode 100644
index 000000000..274c6ef42
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/execlist.c
@@ -0,0 +1,558 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+
+#define _EL_OFFSET_STATUS 0x234
+#define _EL_OFFSET_STATUS_BUF 0x370
+#define _EL_OFFSET_STATUS_PTR 0x3A0
+
+#define execlist_ring_mmio(e, offset) ((e)->mmio_base + (offset))
+
+#define valid_context(ctx) ((ctx)->valid)
+#define same_context(a, b) (((a)->context_id == (b)->context_id) && \
+ ((a)->lrca == (b)->lrca))
+
+static int context_switch_events[] = {
+ [RCS0] = RCS_AS_CONTEXT_SWITCH,
+ [BCS0] = BCS_AS_CONTEXT_SWITCH,
+ [VCS0] = VCS_AS_CONTEXT_SWITCH,
+ [VCS1] = VCS2_AS_CONTEXT_SWITCH,
+ [VECS0] = VECS_AS_CONTEXT_SWITCH,
+};
+
+static int to_context_switch_event(const struct intel_engine_cs *engine)
+{
+ if (WARN_ON(engine->id >= ARRAY_SIZE(context_switch_events)))
+ return -EINVAL;
+
+ return context_switch_events[engine->id];
+}
+
+static void switch_virtual_execlist_slot(struct intel_vgpu_execlist *execlist)
+{
+ gvt_dbg_el("[before] running slot %d/context %x pending slot %d\n",
+ execlist->running_slot ?
+ execlist->running_slot->index : -1,
+ execlist->running_context ?
+ execlist->running_context->context_id : 0,
+ execlist->pending_slot ?
+ execlist->pending_slot->index : -1);
+
+ execlist->running_slot = execlist->pending_slot;
+ execlist->pending_slot = NULL;
+ execlist->running_context = execlist->running_context ?
+ &execlist->running_slot->ctx[0] : NULL;
+
+ gvt_dbg_el("[after] running slot %d/context %x pending slot %d\n",
+ execlist->running_slot ?
+ execlist->running_slot->index : -1,
+ execlist->running_context ?
+ execlist->running_context->context_id : 0,
+ execlist->pending_slot ?
+ execlist->pending_slot->index : -1);
+}
+
+static void emulate_execlist_status(struct intel_vgpu_execlist *execlist)
+{
+ struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+ struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
+ struct execlist_ctx_descriptor_format *desc = execlist->running_context;
+ struct intel_vgpu *vgpu = execlist->vgpu;
+ struct execlist_status_format status;
+ u32 status_reg =
+ execlist_ring_mmio(execlist->engine, _EL_OFFSET_STATUS);
+
+ status.ldw = vgpu_vreg(vgpu, status_reg);
+ status.udw = vgpu_vreg(vgpu, status_reg + 4);
+
+ if (running) {
+ status.current_execlist_pointer = !!running->index;
+ status.execlist_write_pointer = !!!running->index;
+ status.execlist_0_active = status.execlist_0_valid =
+ !!!(running->index);
+ status.execlist_1_active = status.execlist_1_valid =
+ !!(running->index);
+ } else {
+ status.context_id = 0;
+ status.execlist_0_active = status.execlist_0_valid = 0;
+ status.execlist_1_active = status.execlist_1_valid = 0;
+ }
+
+ status.context_id = desc ? desc->context_id : 0;
+ status.execlist_queue_full = !!(pending);
+
+ vgpu_vreg(vgpu, status_reg) = status.ldw;
+ vgpu_vreg(vgpu, status_reg + 4) = status.udw;
+
+ gvt_dbg_el("vgpu%d: status reg offset %x ldw %x udw %x\n",
+ vgpu->id, status_reg, status.ldw, status.udw);
+}
+
+static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
+ struct execlist_context_status_format *status,
+ bool trigger_interrupt_later)
+{
+ struct intel_vgpu *vgpu = execlist->vgpu;
+ struct execlist_context_status_pointer_format ctx_status_ptr;
+ u32 write_pointer;
+ u32 ctx_status_ptr_reg, ctx_status_buf_reg, offset;
+ unsigned long hwsp_gpa;
+
+ ctx_status_ptr_reg =
+ execlist_ring_mmio(execlist->engine, _EL_OFFSET_STATUS_PTR);
+ ctx_status_buf_reg =
+ execlist_ring_mmio(execlist->engine, _EL_OFFSET_STATUS_BUF);
+
+ ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
+
+ write_pointer = ctx_status_ptr.write_ptr;
+
+ if (write_pointer == 0x7)
+ write_pointer = 0;
+ else {
+ ++write_pointer;
+ write_pointer %= 0x6;
+ }
+
+ offset = ctx_status_buf_reg + write_pointer * 8;
+
+ vgpu_vreg(vgpu, offset) = status->ldw;
+ vgpu_vreg(vgpu, offset + 4) = status->udw;
+
+ ctx_status_ptr.write_ptr = write_pointer;
+ vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
+
+ /* Update the CSB and CSB write pointer in HWSP */
+ hwsp_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+ vgpu->hws_pga[execlist->engine->id]);
+ if (hwsp_gpa != INTEL_GVT_INVALID_ADDR) {
+ intel_gvt_write_gpa(vgpu,
+ hwsp_gpa + I915_HWS_CSB_BUF0_INDEX * 4 + write_pointer * 8,
+ status, 8);
+ intel_gvt_write_gpa(vgpu,
+ hwsp_gpa + INTEL_HWS_CSB_WRITE_INDEX(execlist->engine->i915) * 4,
+ &write_pointer, 4);
+ }
+
+ gvt_dbg_el("vgpu%d: w pointer %u reg %x csb l %x csb h %x\n",
+ vgpu->id, write_pointer, offset, status->ldw, status->udw);
+
+ if (trigger_interrupt_later)
+ return;
+
+ intel_vgpu_trigger_virtual_event(vgpu,
+ to_context_switch_event(execlist->engine));
+}
+
+static int emulate_execlist_ctx_schedule_out(
+ struct intel_vgpu_execlist *execlist,
+ struct execlist_ctx_descriptor_format *ctx)
+{
+ struct intel_vgpu *vgpu = execlist->vgpu;
+ struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+ struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
+ struct execlist_ctx_descriptor_format *ctx0 = &running->ctx[0];
+ struct execlist_ctx_descriptor_format *ctx1 = &running->ctx[1];
+ struct execlist_context_status_format status;
+
+ memset(&status, 0, sizeof(status));
+
+ gvt_dbg_el("schedule out context id %x\n", ctx->context_id);
+
+ if (WARN_ON(!same_context(ctx, execlist->running_context))) {
+ gvt_vgpu_err("schedule out context is not running context,"
+ "ctx id %x running ctx id %x\n",
+ ctx->context_id,
+ execlist->running_context->context_id);
+ return -EINVAL;
+ }
+
+ /* ctx1 is valid, ctx0/ctx is scheduled-out -> element switch */
+ if (valid_context(ctx1) && same_context(ctx0, ctx)) {
+ gvt_dbg_el("ctx 1 valid, ctx/ctx 0 is scheduled-out\n");
+
+ execlist->running_context = ctx1;
+
+ emulate_execlist_status(execlist);
+
+ status.context_complete = status.element_switch = 1;
+ status.context_id = ctx->context_id;
+
+ emulate_csb_update(execlist, &status, false);
+ /*
+ * ctx1 is not valid, ctx == ctx0
+ * ctx1 is valid, ctx1 == ctx
+ * --> last element is finished
+ * emulate:
+ * active-to-idle if there is *no* pending execlist
+ * context-complete if there *is* pending execlist
+ */
+ } else if ((!valid_context(ctx1) && same_context(ctx0, ctx))
+ || (valid_context(ctx1) && same_context(ctx1, ctx))) {
+ gvt_dbg_el("need to switch virtual execlist slot\n");
+
+ switch_virtual_execlist_slot(execlist);
+
+ emulate_execlist_status(execlist);
+
+ status.context_complete = status.active_to_idle = 1;
+ status.context_id = ctx->context_id;
+
+ if (!pending) {
+ emulate_csb_update(execlist, &status, false);
+ } else {
+ emulate_csb_update(execlist, &status, true);
+
+ memset(&status, 0, sizeof(status));
+
+ status.idle_to_active = 1;
+ status.context_id = 0;
+
+ emulate_csb_update(execlist, &status, false);
+ }
+ } else {
+ WARN_ON(1);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static struct intel_vgpu_execlist_slot *get_next_execlist_slot(
+ struct intel_vgpu_execlist *execlist)
+{
+ struct intel_vgpu *vgpu = execlist->vgpu;
+ u32 status_reg =
+ execlist_ring_mmio(execlist->engine, _EL_OFFSET_STATUS);
+ struct execlist_status_format status;
+
+ status.ldw = vgpu_vreg(vgpu, status_reg);
+ status.udw = vgpu_vreg(vgpu, status_reg + 4);
+
+ if (status.execlist_queue_full) {
+ gvt_vgpu_err("virtual execlist slots are full\n");
+ return NULL;
+ }
+
+ return &execlist->slot[status.execlist_write_pointer];
+}
+
+static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist,
+ struct execlist_ctx_descriptor_format ctx[2])
+{
+ struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+ struct intel_vgpu_execlist_slot *slot =
+ get_next_execlist_slot(execlist);
+
+ struct execlist_ctx_descriptor_format *ctx0, *ctx1;
+ struct execlist_context_status_format status;
+ struct intel_vgpu *vgpu = execlist->vgpu;
+
+ gvt_dbg_el("emulate schedule-in\n");
+
+ if (!slot) {
+ gvt_vgpu_err("no available execlist slot\n");
+ return -EINVAL;
+ }
+
+ memset(&status, 0, sizeof(status));
+ memset(slot->ctx, 0, sizeof(slot->ctx));
+
+ slot->ctx[0] = ctx[0];
+ slot->ctx[1] = ctx[1];
+
+ gvt_dbg_el("alloc slot index %d ctx 0 %x ctx 1 %x\n",
+ slot->index, ctx[0].context_id,
+ ctx[1].context_id);
+
+ /*
+ * no running execlist, make this write bundle as running execlist
+ * -> idle-to-active
+ */
+ if (!running) {
+ gvt_dbg_el("no current running execlist\n");
+
+ execlist->running_slot = slot;
+ execlist->pending_slot = NULL;
+ execlist->running_context = &slot->ctx[0];
+
+ gvt_dbg_el("running slot index %d running context %x\n",
+ execlist->running_slot->index,
+ execlist->running_context->context_id);
+
+ emulate_execlist_status(execlist);
+
+ status.idle_to_active = 1;
+ status.context_id = 0;
+
+ emulate_csb_update(execlist, &status, false);
+ return 0;
+ }
+
+ ctx0 = &running->ctx[0];
+ ctx1 = &running->ctx[1];
+
+ gvt_dbg_el("current running slot index %d ctx 0 %x ctx 1 %x\n",
+ running->index, ctx0->context_id, ctx1->context_id);
+
+ /*
+ * already has an running execlist
+ * a. running ctx1 is valid,
+ * ctx0 is finished, and running ctx1 == new execlist ctx[0]
+ * b. running ctx1 is not valid,
+ * ctx0 == new execlist ctx[0]
+ * ----> lite-restore + preempted
+ */
+ if ((valid_context(ctx1) && same_context(ctx1, &slot->ctx[0]) &&
+ /* condition a */
+ (!same_context(ctx0, execlist->running_context))) ||
+ (!valid_context(ctx1) &&
+ same_context(ctx0, &slot->ctx[0]))) { /* condition b */
+ gvt_dbg_el("need to switch virtual execlist slot\n");
+
+ execlist->pending_slot = slot;
+ switch_virtual_execlist_slot(execlist);
+
+ emulate_execlist_status(execlist);
+
+ status.lite_restore = status.preempted = 1;
+ status.context_id = ctx[0].context_id;
+
+ emulate_csb_update(execlist, &status, false);
+ } else {
+ gvt_dbg_el("emulate as pending slot\n");
+ /*
+ * otherwise
+ * --> emulate pending execlist exist + but no preemption case
+ */
+ execlist->pending_slot = slot;
+ emulate_execlist_status(execlist);
+ }
+ return 0;
+}
+
+#define get_desc_from_elsp_dwords(ed, i) \
+ ((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2]))
+
+static int prepare_execlist_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct execlist_ctx_descriptor_format ctx[2];
+ int ret;
+
+ if (!workload->emulate_schedule_in)
+ return 0;
+
+ ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
+ ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);
+
+ ret = emulate_execlist_schedule_in(&s->execlist[workload->engine->id],
+ ctx);
+ if (ret) {
+ gvt_vgpu_err("fail to emulate execlist schedule in\n");
+ return ret;
+ }
+ return 0;
+}
+
+static int complete_execlist_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist =
+ &s->execlist[workload->engine->id];
+ struct intel_vgpu_workload *next_workload;
+ struct list_head *next = workload_q_head(vgpu, workload->engine)->next;
+ bool lite_restore = false;
+ int ret = 0;
+
+ gvt_dbg_el("complete workload %p status %d\n",
+ workload, workload->status);
+
+ if (workload->status || vgpu->resetting_eng & workload->engine->mask)
+ goto out;
+
+ if (!list_empty(workload_q_head(vgpu, workload->engine))) {
+ struct execlist_ctx_descriptor_format *this_desc, *next_desc;
+
+ next_workload = container_of(next,
+ struct intel_vgpu_workload, list);
+ this_desc = &workload->ctx_desc;
+ next_desc = &next_workload->ctx_desc;
+
+ lite_restore = same_context(this_desc, next_desc);
+ }
+
+ if (lite_restore) {
+ gvt_dbg_el("next context == current - no schedule-out\n");
+ goto out;
+ }
+
+ ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc);
+out:
+ return ret;
+}
+
+static int submit_context(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine,
+ struct execlist_ctx_descriptor_format *desc,
+ bool emulate_schedule_in)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_workload *workload = NULL;
+
+ workload = intel_vgpu_create_workload(vgpu, engine, desc);
+ if (IS_ERR(workload))
+ return PTR_ERR(workload);
+
+ workload->prepare = prepare_execlist_workload;
+ workload->complete = complete_execlist_workload;
+ workload->emulate_schedule_in = emulate_schedule_in;
+
+ if (emulate_schedule_in)
+ workload->elsp_dwords = s->execlist[engine->id].elsp_dwords;
+
+ gvt_dbg_el("workload %p emulate schedule_in %d\n", workload,
+ emulate_schedule_in);
+
+ intel_vgpu_queue_workload(workload);
+ return 0;
+}
+
+int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist = &s->execlist[engine->id];
+ struct execlist_ctx_descriptor_format *desc[2];
+ int i, ret;
+
+ desc[0] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0);
+ desc[1] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1);
+
+ if (!desc[0]->valid) {
+ gvt_vgpu_err("invalid elsp submission, desc0 is invalid\n");
+ goto inv_desc;
+ }
+
+ for (i = 0; i < ARRAY_SIZE(desc); i++) {
+ if (!desc[i]->valid)
+ continue;
+ if (!desc[i]->privilege_access) {
+ gvt_vgpu_err("unexpected GGTT elsp submission\n");
+ goto inv_desc;
+ }
+ }
+
+ /* submit workload */
+ for (i = 0; i < ARRAY_SIZE(desc); i++) {
+ if (!desc[i]->valid)
+ continue;
+ ret = submit_context(vgpu, engine, desc[i], i == 0);
+ if (ret) {
+ gvt_vgpu_err("failed to submit desc %d\n", i);
+ return ret;
+ }
+ }
+
+ return 0;
+
+inv_desc:
+ gvt_vgpu_err("descriptors content: desc0 %08x %08x desc1 %08x %08x\n",
+ desc[0]->udw, desc[0]->ldw, desc[1]->udw, desc[1]->ldw);
+ return -EINVAL;
+}
+
+static void init_vgpu_execlist(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist = &s->execlist[engine->id];
+ struct execlist_context_status_pointer_format ctx_status_ptr;
+ u32 ctx_status_ptr_reg;
+
+ memset(execlist, 0, sizeof(*execlist));
+
+ execlist->vgpu = vgpu;
+ execlist->engine = engine;
+ execlist->slot[0].index = 0;
+ execlist->slot[1].index = 1;
+
+ ctx_status_ptr_reg = execlist_ring_mmio(engine, _EL_OFFSET_STATUS_PTR);
+ ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
+ ctx_status_ptr.read_ptr = 0;
+ ctx_status_ptr.write_ptr = 0x7;
+ vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
+}
+
+static void clean_execlist(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_engine_cs *engine;
+ intel_engine_mask_t tmp;
+
+ for_each_engine_masked(engine, vgpu->gvt->gt, engine_mask, tmp) {
+ kfree(s->ring_scan_buffer[engine->id]);
+ s->ring_scan_buffer[engine->id] = NULL;
+ s->ring_scan_buffer_size[engine->id] = 0;
+ }
+}
+
+static void reset_execlist(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask)
+{
+ struct intel_engine_cs *engine;
+ intel_engine_mask_t tmp;
+
+ for_each_engine_masked(engine, vgpu->gvt->gt, engine_mask, tmp)
+ init_vgpu_execlist(vgpu, engine);
+}
+
+static int init_execlist(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask)
+{
+ reset_execlist(vgpu, engine_mask);
+ return 0;
+}
+
+const struct intel_vgpu_submission_ops intel_vgpu_execlist_submission_ops = {
+ .name = "execlist",
+ .init = init_execlist,
+ .reset = reset_execlist,
+ .clean = clean_execlist,
+};
diff --git a/drivers/gpu/drm/i915/gvt/execlist.h b/drivers/gpu/drm/i915/gvt/execlist.h
new file mode 100644
index 000000000..84ad74b37
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/execlist.h
@@ -0,0 +1,185 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#ifndef _GVT_EXECLIST_H_
+#define _GVT_EXECLIST_H_
+
+#include <linux/types.h>
+
+struct execlist_ctx_descriptor_format {
+ union {
+ u32 ldw;
+ struct {
+ u32 valid : 1;
+ u32 force_pd_restore : 1;
+ u32 force_restore : 1;
+ u32 addressing_mode : 2;
+ u32 llc_coherency : 1;
+ u32 fault_handling : 2;
+ u32 privilege_access : 1;
+ u32 reserved : 3;
+ u32 lrca : 20;
+ };
+ };
+ union {
+ u32 udw;
+ u32 context_id;
+ };
+};
+
+struct execlist_status_format {
+ union {
+ u32 ldw;
+ struct {
+ u32 current_execlist_pointer :1;
+ u32 execlist_write_pointer :1;
+ u32 execlist_queue_full :1;
+ u32 execlist_1_valid :1;
+ u32 execlist_0_valid :1;
+ u32 last_ctx_switch_reason :9;
+ u32 current_active_elm_status :2;
+ u32 arbitration_enable :1;
+ u32 execlist_1_active :1;
+ u32 execlist_0_active :1;
+ u32 reserved :13;
+ };
+ };
+ union {
+ u32 udw;
+ u32 context_id;
+ };
+};
+
+struct execlist_context_status_pointer_format {
+ union {
+ u32 dw;
+ struct {
+ u32 write_ptr :3;
+ u32 reserved :5;
+ u32 read_ptr :3;
+ u32 reserved2 :5;
+ u32 mask :16;
+ };
+ };
+};
+
+struct execlist_context_status_format {
+ union {
+ u32 ldw;
+ struct {
+ u32 idle_to_active :1;
+ u32 preempted :1;
+ u32 element_switch :1;
+ u32 active_to_idle :1;
+ u32 context_complete :1;
+ u32 wait_on_sync_flip :1;
+ u32 wait_on_vblank :1;
+ u32 wait_on_semaphore :1;
+ u32 wait_on_scanline :1;
+ u32 reserved :2;
+ u32 semaphore_wait_mode :1;
+ u32 display_plane :3;
+ u32 lite_restore :1;
+ u32 reserved_2 :16;
+ };
+ };
+ union {
+ u32 udw;
+ u32 context_id;
+ };
+};
+
+struct execlist_mmio_pair {
+ u32 addr;
+ u32 val;
+};
+
+/* The first 52 dwords in register state context */
+struct execlist_ring_context {
+ u32 nop1;
+ u32 lri_cmd_1;
+ struct execlist_mmio_pair ctx_ctrl;
+ struct execlist_mmio_pair ring_header;
+ struct execlist_mmio_pair ring_tail;
+ struct execlist_mmio_pair rb_start;
+ struct execlist_mmio_pair rb_ctrl;
+ struct execlist_mmio_pair bb_cur_head_UDW;
+ struct execlist_mmio_pair bb_cur_head_LDW;
+ struct execlist_mmio_pair bb_state;
+ struct execlist_mmio_pair second_bb_addr_UDW;
+ struct execlist_mmio_pair second_bb_addr_LDW;
+ struct execlist_mmio_pair second_bb_state;
+ struct execlist_mmio_pair bb_per_ctx_ptr;
+ struct execlist_mmio_pair rcs_indirect_ctx;
+ struct execlist_mmio_pair rcs_indirect_ctx_offset;
+ u32 nop2;
+ u32 nop3;
+ u32 nop4;
+ u32 lri_cmd_2;
+ struct execlist_mmio_pair ctx_timestamp;
+ /*
+ * pdps[8]={ pdp3_UDW, pdp3_LDW, pdp2_UDW, pdp2_LDW,
+ * pdp1_UDW, pdp1_LDW, pdp0_UDW, pdp0_LDW}
+ */
+ struct execlist_mmio_pair pdps[8];
+};
+
+struct intel_vgpu_elsp_dwords {
+ u32 data[4];
+ u32 index;
+};
+
+struct intel_vgpu_execlist_slot {
+ struct execlist_ctx_descriptor_format ctx[2];
+ u32 index;
+};
+
+struct intel_vgpu_execlist {
+ struct intel_vgpu_execlist_slot slot[2];
+ struct intel_vgpu_execlist_slot *running_slot;
+ struct intel_vgpu_execlist_slot *pending_slot;
+ struct execlist_ctx_descriptor_format *running_context;
+ struct intel_vgpu *vgpu;
+ struct intel_vgpu_elsp_dwords elsp_dwords;
+ const struct intel_engine_cs *engine;
+};
+
+void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu);
+
+int intel_vgpu_init_execlist(struct intel_vgpu *vgpu);
+
+int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine);
+
+#endif /*_GVT_EXECLIST_H_*/
diff --git a/drivers/gpu/drm/i915/gvt/fb_decoder.c b/drivers/gpu/drm/i915/gvt/fb_decoder.c
new file mode 100644
index 000000000..835c3fde8
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/fb_decoder.c
@@ -0,0 +1,508 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Bing Niu <bing.niu@intel.com>
+ * Xu Han <xu.han@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Xiaoguang Chen <xiaoguang.chen@intel.com>
+ * Yang Liu <yang2.liu@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#include <uapi/drm/drm_fourcc.h>
+#include "i915_drv.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+#include "i915_reg.h"
+
+#define PRIMARY_FORMAT_NUM 16
+struct pixel_format {
+ int drm_format; /* Pixel format in DRM definition */
+ int bpp; /* Bits per pixel, 0 indicates invalid */
+ const char *desc; /* The description */
+};
+
+static const struct pixel_format bdw_pixel_formats[] = {
+ {DRM_FORMAT_C8, 8, "8-bit Indexed"},
+ {DRM_FORMAT_RGB565, 16, "16-bit BGRX (5:6:5 MSB-R:G:B)"},
+ {DRM_FORMAT_XRGB8888, 32, "32-bit BGRX (8:8:8:8 MSB-X:R:G:B)"},
+ {DRM_FORMAT_XBGR2101010, 32, "32-bit RGBX (2:10:10:10 MSB-X:B:G:R)"},
+
+ {DRM_FORMAT_XRGB2101010, 32, "32-bit BGRX (2:10:10:10 MSB-X:R:G:B)"},
+ {DRM_FORMAT_XBGR8888, 32, "32-bit RGBX (8:8:8:8 MSB-X:B:G:R)"},
+
+ /* non-supported format has bpp default to 0 */
+ {0, 0, NULL},
+};
+
+static const struct pixel_format skl_pixel_formats[] = {
+ {DRM_FORMAT_YUYV, 16, "16-bit packed YUYV (8:8:8:8 MSB-V:Y2:U:Y1)"},
+ {DRM_FORMAT_UYVY, 16, "16-bit packed UYVY (8:8:8:8 MSB-Y2:V:Y1:U)"},
+ {DRM_FORMAT_YVYU, 16, "16-bit packed YVYU (8:8:8:8 MSB-U:Y2:V:Y1)"},
+ {DRM_FORMAT_VYUY, 16, "16-bit packed VYUY (8:8:8:8 MSB-Y2:U:Y1:V)"},
+
+ {DRM_FORMAT_C8, 8, "8-bit Indexed"},
+ {DRM_FORMAT_RGB565, 16, "16-bit BGRX (5:6:5 MSB-R:G:B)"},
+ {DRM_FORMAT_ABGR8888, 32, "32-bit RGBA (8:8:8:8 MSB-A:B:G:R)"},
+ {DRM_FORMAT_XBGR8888, 32, "32-bit RGBX (8:8:8:8 MSB-X:B:G:R)"},
+
+ {DRM_FORMAT_ARGB8888, 32, "32-bit BGRA (8:8:8:8 MSB-A:R:G:B)"},
+ {DRM_FORMAT_XRGB8888, 32, "32-bit BGRX (8:8:8:8 MSB-X:R:G:B)"},
+ {DRM_FORMAT_XBGR2101010, 32, "32-bit RGBX (2:10:10:10 MSB-X:B:G:R)"},
+ {DRM_FORMAT_XRGB2101010, 32, "32-bit BGRX (2:10:10:10 MSB-X:R:G:B)"},
+
+ /* non-supported format has bpp default to 0 */
+ {0, 0, NULL},
+};
+
+static int bdw_format_to_drm(int format)
+{
+ int bdw_pixel_formats_index = 6;
+
+ switch (format) {
+ case DISP_FORMAT_8BPP:
+ bdw_pixel_formats_index = 0;
+ break;
+ case DISP_FORMAT_BGRX565:
+ bdw_pixel_formats_index = 1;
+ break;
+ case DISP_FORMAT_BGRX888:
+ bdw_pixel_formats_index = 2;
+ break;
+ case DISP_FORMAT_RGBX101010:
+ bdw_pixel_formats_index = 3;
+ break;
+ case DISP_FORMAT_BGRX101010:
+ bdw_pixel_formats_index = 4;
+ break;
+ case DISP_FORMAT_RGBX888:
+ bdw_pixel_formats_index = 5;
+ break;
+
+ default:
+ break;
+ }
+
+ return bdw_pixel_formats_index;
+}
+
+static int skl_format_to_drm(int format, bool rgb_order, bool alpha,
+ int yuv_order)
+{
+ int skl_pixel_formats_index = 12;
+
+ switch (format) {
+ case PLANE_CTL_FORMAT_INDEXED:
+ skl_pixel_formats_index = 4;
+ break;
+ case PLANE_CTL_FORMAT_RGB_565:
+ skl_pixel_formats_index = 5;
+ break;
+ case PLANE_CTL_FORMAT_XRGB_8888:
+ if (rgb_order)
+ skl_pixel_formats_index = alpha ? 6 : 7;
+ else
+ skl_pixel_formats_index = alpha ? 8 : 9;
+ break;
+ case PLANE_CTL_FORMAT_XRGB_2101010:
+ skl_pixel_formats_index = rgb_order ? 10 : 11;
+ break;
+ case PLANE_CTL_FORMAT_YUV422:
+ skl_pixel_formats_index = yuv_order >> 16;
+ if (skl_pixel_formats_index > 3)
+ return -EINVAL;
+ break;
+
+ default:
+ break;
+ }
+
+ return skl_pixel_formats_index;
+}
+
+static u32 intel_vgpu_get_stride(struct intel_vgpu *vgpu, int pipe,
+ u32 tiled, int stride_mask, int bpp)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ u32 stride_reg = vgpu_vreg_t(vgpu, DSPSTRIDE(pipe)) & stride_mask;
+ u32 stride = stride_reg;
+
+ if (GRAPHICS_VER(dev_priv) >= 9) {
+ switch (tiled) {
+ case PLANE_CTL_TILED_LINEAR:
+ stride = stride_reg * 64;
+ break;
+ case PLANE_CTL_TILED_X:
+ stride = stride_reg * 512;
+ break;
+ case PLANE_CTL_TILED_Y:
+ stride = stride_reg * 128;
+ break;
+ case PLANE_CTL_TILED_YF:
+ if (bpp == 8)
+ stride = stride_reg * 64;
+ else if (bpp == 16 || bpp == 32 || bpp == 64)
+ stride = stride_reg * 128;
+ else
+ gvt_dbg_core("skl: unsupported bpp:%d\n", bpp);
+ break;
+ default:
+ gvt_dbg_core("skl: unsupported tile format:%x\n",
+ tiled);
+ }
+ }
+
+ return stride;
+}
+
+static int get_active_pipe(struct intel_vgpu *vgpu)
+{
+ int i;
+
+ for (i = 0; i < I915_MAX_PIPES; i++)
+ if (pipe_is_enabled(vgpu, i))
+ break;
+
+ return i;
+}
+
+/**
+ * intel_vgpu_decode_primary_plane - Decode primary plane
+ * @vgpu: input vgpu
+ * @plane: primary plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_primary_plane_format *plane)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ u32 val, fmt;
+ int pipe;
+
+ pipe = get_active_pipe(vgpu);
+ if (pipe >= I915_MAX_PIPES)
+ return -ENODEV;
+
+ val = vgpu_vreg_t(vgpu, DSPCNTR(pipe));
+ plane->enabled = !!(val & DISP_ENABLE);
+ if (!plane->enabled)
+ return -ENODEV;
+
+ if (GRAPHICS_VER(dev_priv) >= 9) {
+ plane->tiled = val & PLANE_CTL_TILED_MASK;
+ fmt = skl_format_to_drm(
+ val & PLANE_CTL_FORMAT_MASK_SKL,
+ val & PLANE_CTL_ORDER_RGBX,
+ val & PLANE_CTL_ALPHA_MASK,
+ val & PLANE_CTL_YUV422_ORDER_MASK);
+
+ if (fmt >= ARRAY_SIZE(skl_pixel_formats)) {
+ gvt_vgpu_err("Out-of-bounds pixel format index\n");
+ return -EINVAL;
+ }
+
+ plane->bpp = skl_pixel_formats[fmt].bpp;
+ plane->drm_format = skl_pixel_formats[fmt].drm_format;
+ } else {
+ plane->tiled = val & DISP_TILED;
+ fmt = bdw_format_to_drm(val & DISP_FORMAT_MASK);
+ plane->bpp = bdw_pixel_formats[fmt].bpp;
+ plane->drm_format = bdw_pixel_formats[fmt].drm_format;
+ }
+
+ if (!plane->bpp) {
+ gvt_vgpu_err("Non-supported pixel format (0x%x)\n", fmt);
+ return -EINVAL;
+ }
+
+ plane->hw_format = fmt;
+
+ plane->base = vgpu_vreg_t(vgpu, DSPSURF(pipe)) & I915_GTT_PAGE_MASK;
+ if (!vgpu_gmadr_is_valid(vgpu, plane->base))
+ return -EINVAL;
+
+ plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+ if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("Translate primary plane gma 0x%x to gpa fail\n",
+ plane->base);
+ return -EINVAL;
+ }
+
+ plane->stride = intel_vgpu_get_stride(vgpu, pipe, plane->tiled,
+ (GRAPHICS_VER(dev_priv) >= 9) ?
+ (_PRI_PLANE_STRIDE_MASK >> 6) :
+ _PRI_PLANE_STRIDE_MASK, plane->bpp);
+
+ plane->width = (vgpu_vreg_t(vgpu, PIPESRC(pipe)) & _PIPE_H_SRCSZ_MASK) >>
+ _PIPE_H_SRCSZ_SHIFT;
+ plane->width += 1;
+ plane->height = (vgpu_vreg_t(vgpu, PIPESRC(pipe)) &
+ _PIPE_V_SRCSZ_MASK) >> _PIPE_V_SRCSZ_SHIFT;
+ plane->height += 1; /* raw height is one minus the real value */
+
+ val = vgpu_vreg_t(vgpu, DSPTILEOFF(pipe));
+ plane->x_offset = (val & _PRI_PLANE_X_OFF_MASK) >>
+ _PRI_PLANE_X_OFF_SHIFT;
+ plane->y_offset = (val & _PRI_PLANE_Y_OFF_MASK) >>
+ _PRI_PLANE_Y_OFF_SHIFT;
+
+ return 0;
+}
+
+#define CURSOR_FORMAT_NUM (1 << 6)
+struct cursor_mode_format {
+ int drm_format; /* Pixel format in DRM definition */
+ u8 bpp; /* Bits per pixel; 0 indicates invalid */
+ u32 width; /* In pixel */
+ u32 height; /* In lines */
+ const char *desc; /* The description */
+};
+
+static const struct cursor_mode_format cursor_pixel_formats[] = {
+ {DRM_FORMAT_ARGB8888, 32, 128, 128, "128x128 32bpp ARGB"},
+ {DRM_FORMAT_ARGB8888, 32, 256, 256, "256x256 32bpp ARGB"},
+ {DRM_FORMAT_ARGB8888, 32, 64, 64, "64x64 32bpp ARGB"},
+ {DRM_FORMAT_ARGB8888, 32, 64, 64, "64x64 32bpp ARGB"},
+
+ /* non-supported format has bpp default to 0 */
+ {0, 0, 0, 0, NULL},
+};
+
+static int cursor_mode_to_drm(int mode)
+{
+ int cursor_pixel_formats_index = 4;
+
+ switch (mode) {
+ case MCURSOR_MODE_128_ARGB_AX:
+ cursor_pixel_formats_index = 0;
+ break;
+ case MCURSOR_MODE_256_ARGB_AX:
+ cursor_pixel_formats_index = 1;
+ break;
+ case MCURSOR_MODE_64_ARGB_AX:
+ cursor_pixel_formats_index = 2;
+ break;
+ case MCURSOR_MODE_64_32B_AX:
+ cursor_pixel_formats_index = 3;
+ break;
+
+ default:
+ break;
+ }
+
+ return cursor_pixel_formats_index;
+}
+
+/**
+ * intel_vgpu_decode_cursor_plane - Decode sprite plane
+ * @vgpu: input vgpu
+ * @plane: cursor plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_cursor_plane_format *plane)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ u32 val, mode, index;
+ u32 alpha_plane, alpha_force;
+ int pipe;
+
+ pipe = get_active_pipe(vgpu);
+ if (pipe >= I915_MAX_PIPES)
+ return -ENODEV;
+
+ val = vgpu_vreg_t(vgpu, CURCNTR(pipe));
+ mode = val & MCURSOR_MODE_MASK;
+ plane->enabled = (mode != MCURSOR_MODE_DISABLE);
+ if (!plane->enabled)
+ return -ENODEV;
+
+ index = cursor_mode_to_drm(mode);
+
+ if (!cursor_pixel_formats[index].bpp) {
+ gvt_vgpu_err("Non-supported cursor mode (0x%x)\n", mode);
+ return -EINVAL;
+ }
+ plane->mode = mode;
+ plane->bpp = cursor_pixel_formats[index].bpp;
+ plane->drm_format = cursor_pixel_formats[index].drm_format;
+ plane->width = cursor_pixel_formats[index].width;
+ plane->height = cursor_pixel_formats[index].height;
+
+ alpha_plane = (val & _CURSOR_ALPHA_PLANE_MASK) >>
+ _CURSOR_ALPHA_PLANE_SHIFT;
+ alpha_force = (val & _CURSOR_ALPHA_FORCE_MASK) >>
+ _CURSOR_ALPHA_FORCE_SHIFT;
+ if (alpha_plane || alpha_force)
+ gvt_dbg_core("alpha_plane=0x%x, alpha_force=0x%x\n",
+ alpha_plane, alpha_force);
+
+ plane->base = vgpu_vreg_t(vgpu, CURBASE(pipe)) & I915_GTT_PAGE_MASK;
+ if (!vgpu_gmadr_is_valid(vgpu, plane->base))
+ return -EINVAL;
+
+ plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+ if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("Translate cursor plane gma 0x%x to gpa fail\n",
+ plane->base);
+ return -EINVAL;
+ }
+
+ val = vgpu_vreg_t(vgpu, CURPOS(pipe));
+ plane->x_pos = (val & _CURSOR_POS_X_MASK) >> _CURSOR_POS_X_SHIFT;
+ plane->x_sign = (val & _CURSOR_SIGN_X_MASK) >> _CURSOR_SIGN_X_SHIFT;
+ plane->y_pos = (val & _CURSOR_POS_Y_MASK) >> _CURSOR_POS_Y_SHIFT;
+ plane->y_sign = (val & _CURSOR_SIGN_Y_MASK) >> _CURSOR_SIGN_Y_SHIFT;
+
+ plane->x_hot = vgpu_vreg_t(vgpu, vgtif_reg(cursor_x_hot));
+ plane->y_hot = vgpu_vreg_t(vgpu, vgtif_reg(cursor_y_hot));
+ return 0;
+}
+
+#define SPRITE_FORMAT_NUM (1 << 3)
+
+static const struct pixel_format sprite_pixel_formats[SPRITE_FORMAT_NUM] = {
+ [0x0] = {DRM_FORMAT_YUV422, 16, "YUV 16-bit 4:2:2 packed"},
+ [0x1] = {DRM_FORMAT_XRGB2101010, 32, "RGB 32-bit 2:10:10:10"},
+ [0x2] = {DRM_FORMAT_XRGB8888, 32, "RGB 32-bit 8:8:8:8"},
+ [0x4] = {DRM_FORMAT_AYUV, 32,
+ "YUV 32-bit 4:4:4 packed (8:8:8:8 MSB-X:Y:U:V)"},
+};
+
+/**
+ * intel_vgpu_decode_sprite_plane - Decode sprite plane
+ * @vgpu: input vgpu
+ * @plane: sprite plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_sprite_plane_format *plane)
+{
+ u32 val, fmt;
+ u32 color_order, yuv_order;
+ int drm_format;
+ int pipe;
+
+ pipe = get_active_pipe(vgpu);
+ if (pipe >= I915_MAX_PIPES)
+ return -ENODEV;
+
+ val = vgpu_vreg_t(vgpu, SPRCTL(pipe));
+ plane->enabled = !!(val & SPRITE_ENABLE);
+ if (!plane->enabled)
+ return -ENODEV;
+
+ plane->tiled = !!(val & SPRITE_TILED);
+ color_order = !!(val & SPRITE_RGB_ORDER_RGBX);
+ yuv_order = (val & SPRITE_YUV_ORDER_MASK) >>
+ _SPRITE_YUV_ORDER_SHIFT;
+
+ fmt = (val & SPRITE_FORMAT_MASK) >> _SPRITE_FMT_SHIFT;
+ if (!sprite_pixel_formats[fmt].bpp) {
+ gvt_vgpu_err("Non-supported pixel format (0x%x)\n", fmt);
+ return -EINVAL;
+ }
+ plane->hw_format = fmt;
+ plane->bpp = sprite_pixel_formats[fmt].bpp;
+ drm_format = sprite_pixel_formats[fmt].drm_format;
+
+ /* Order of RGB values in an RGBxxx buffer may be ordered RGB or
+ * BGR depending on the state of the color_order field
+ */
+ if (!color_order) {
+ if (drm_format == DRM_FORMAT_XRGB2101010)
+ drm_format = DRM_FORMAT_XBGR2101010;
+ else if (drm_format == DRM_FORMAT_XRGB8888)
+ drm_format = DRM_FORMAT_XBGR8888;
+ }
+
+ if (drm_format == DRM_FORMAT_YUV422) {
+ switch (yuv_order) {
+ case 0:
+ drm_format = DRM_FORMAT_YUYV;
+ break;
+ case 1:
+ drm_format = DRM_FORMAT_UYVY;
+ break;
+ case 2:
+ drm_format = DRM_FORMAT_YVYU;
+ break;
+ case 3:
+ drm_format = DRM_FORMAT_VYUY;
+ break;
+ default:
+ /* yuv_order has only 2 bits */
+ break;
+ }
+ }
+
+ plane->drm_format = drm_format;
+
+ plane->base = vgpu_vreg_t(vgpu, SPRSURF(pipe)) & I915_GTT_PAGE_MASK;
+ if (!vgpu_gmadr_is_valid(vgpu, plane->base))
+ return -EINVAL;
+
+ plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+ if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("Translate sprite plane gma 0x%x to gpa fail\n",
+ plane->base);
+ return -EINVAL;
+ }
+
+ plane->stride = vgpu_vreg_t(vgpu, SPRSTRIDE(pipe)) &
+ _SPRITE_STRIDE_MASK;
+
+ val = vgpu_vreg_t(vgpu, SPRSIZE(pipe));
+ plane->height = (val & _SPRITE_SIZE_HEIGHT_MASK) >>
+ _SPRITE_SIZE_HEIGHT_SHIFT;
+ plane->width = (val & _SPRITE_SIZE_WIDTH_MASK) >>
+ _SPRITE_SIZE_WIDTH_SHIFT;
+ plane->height += 1; /* raw height is one minus the real value */
+ plane->width += 1; /* raw width is one minus the real value */
+
+ val = vgpu_vreg_t(vgpu, SPRPOS(pipe));
+ plane->x_pos = (val & _SPRITE_POS_X_MASK) >> _SPRITE_POS_X_SHIFT;
+ plane->y_pos = (val & _SPRITE_POS_Y_MASK) >> _SPRITE_POS_Y_SHIFT;
+
+ val = vgpu_vreg_t(vgpu, SPROFFSET(pipe));
+ plane->x_offset = (val & _SPRITE_OFFSET_START_X_MASK) >>
+ _SPRITE_OFFSET_START_X_SHIFT;
+ plane->y_offset = (val & _SPRITE_OFFSET_START_Y_MASK) >>
+ _SPRITE_OFFSET_START_Y_SHIFT;
+
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/fb_decoder.h b/drivers/gpu/drm/i915/gvt/fb_decoder.h
new file mode 100644
index 000000000..0daa3931a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/fb_decoder.h
@@ -0,0 +1,173 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Bing Niu <bing.niu@intel.com>
+ * Xu Han <xu.han@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Xiaoguang Chen <xiaoguang.chen@intel.com>
+ * Yang Liu <yang2.liu@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#ifndef _GVT_FB_DECODER_H_
+#define _GVT_FB_DECODER_H_
+
+#include <linux/types.h>
+
+#include "display/intel_display.h"
+
+struct intel_vgpu;
+
+#define _PLANE_CTL_FORMAT_SHIFT 24
+#define _PLANE_CTL_TILED_SHIFT 10
+#define _PIPE_V_SRCSZ_SHIFT 0
+#define _PIPE_V_SRCSZ_MASK (0xfff << _PIPE_V_SRCSZ_SHIFT)
+#define _PIPE_H_SRCSZ_SHIFT 16
+#define _PIPE_H_SRCSZ_MASK (0x1fff << _PIPE_H_SRCSZ_SHIFT)
+
+#define _PRI_PLANE_FMT_SHIFT 26
+#define _PRI_PLANE_STRIDE_MASK (0x3ff << 6)
+#define _PRI_PLANE_X_OFF_SHIFT 0
+#define _PRI_PLANE_X_OFF_MASK (0x1fff << _PRI_PLANE_X_OFF_SHIFT)
+#define _PRI_PLANE_Y_OFF_SHIFT 16
+#define _PRI_PLANE_Y_OFF_MASK (0xfff << _PRI_PLANE_Y_OFF_SHIFT)
+
+#define _CURSOR_MODE 0x3f
+#define _CURSOR_ALPHA_FORCE_SHIFT 8
+#define _CURSOR_ALPHA_FORCE_MASK (0x3 << _CURSOR_ALPHA_FORCE_SHIFT)
+#define _CURSOR_ALPHA_PLANE_SHIFT 10
+#define _CURSOR_ALPHA_PLANE_MASK (0x3 << _CURSOR_ALPHA_PLANE_SHIFT)
+#define _CURSOR_POS_X_SHIFT 0
+#define _CURSOR_POS_X_MASK (0x1fff << _CURSOR_POS_X_SHIFT)
+#define _CURSOR_SIGN_X_SHIFT 15
+#define _CURSOR_SIGN_X_MASK (1 << _CURSOR_SIGN_X_SHIFT)
+#define _CURSOR_POS_Y_SHIFT 16
+#define _CURSOR_POS_Y_MASK (0xfff << _CURSOR_POS_Y_SHIFT)
+#define _CURSOR_SIGN_Y_SHIFT 31
+#define _CURSOR_SIGN_Y_MASK (1 << _CURSOR_SIGN_Y_SHIFT)
+
+#define _SPRITE_FMT_SHIFT 25
+#define _SPRITE_COLOR_ORDER_SHIFT 20
+#define _SPRITE_YUV_ORDER_SHIFT 16
+#define _SPRITE_STRIDE_SHIFT 6
+#define _SPRITE_STRIDE_MASK (0x1ff << _SPRITE_STRIDE_SHIFT)
+#define _SPRITE_SIZE_WIDTH_SHIFT 0
+#define _SPRITE_SIZE_HEIGHT_SHIFT 16
+#define _SPRITE_SIZE_WIDTH_MASK (0x1fff << _SPRITE_SIZE_WIDTH_SHIFT)
+#define _SPRITE_SIZE_HEIGHT_MASK (0xfff << _SPRITE_SIZE_HEIGHT_SHIFT)
+#define _SPRITE_POS_X_SHIFT 0
+#define _SPRITE_POS_Y_SHIFT 16
+#define _SPRITE_POS_X_MASK (0x1fff << _SPRITE_POS_X_SHIFT)
+#define _SPRITE_POS_Y_MASK (0xfff << _SPRITE_POS_Y_SHIFT)
+#define _SPRITE_OFFSET_START_X_SHIFT 0
+#define _SPRITE_OFFSET_START_Y_SHIFT 16
+#define _SPRITE_OFFSET_START_X_MASK (0x1fff << _SPRITE_OFFSET_START_X_SHIFT)
+#define _SPRITE_OFFSET_START_Y_MASK (0xfff << _SPRITE_OFFSET_START_Y_SHIFT)
+
+enum GVT_FB_EVENT {
+ FB_MODE_SET_START = 1,
+ FB_MODE_SET_END,
+ FB_DISPLAY_FLIP,
+};
+
+enum DDI_PORT {
+ DDI_PORT_NONE = 0,
+ DDI_PORT_B = 1,
+ DDI_PORT_C = 2,
+ DDI_PORT_D = 3,
+ DDI_PORT_E = 4
+};
+
+/* color space conversion and gamma correction are not included */
+struct intel_vgpu_primary_plane_format {
+ u8 enabled; /* plane is enabled */
+ u32 tiled; /* tiling mode: linear, X-tiled, Y tiled, etc */
+ u8 bpp; /* bits per pixel */
+ u32 hw_format; /* format field in the PRI_CTL register */
+ u32 drm_format; /* format in DRM definition */
+ u32 base; /* framebuffer base in graphics memory */
+ u64 base_gpa;
+ u32 x_offset; /* in pixels */
+ u32 y_offset; /* in lines */
+ u32 width; /* in pixels */
+ u32 height; /* in lines */
+ u32 stride; /* in bytes */
+};
+
+struct intel_vgpu_sprite_plane_format {
+ u8 enabled; /* plane is enabled */
+ u8 tiled; /* X-tiled */
+ u8 bpp; /* bits per pixel */
+ u32 hw_format; /* format field in the SPR_CTL register */
+ u32 drm_format; /* format in DRM definition */
+ u32 base; /* sprite base in graphics memory */
+ u64 base_gpa;
+ u32 x_pos; /* in pixels */
+ u32 y_pos; /* in lines */
+ u32 x_offset; /* in pixels */
+ u32 y_offset; /* in lines */
+ u32 width; /* in pixels */
+ u32 height; /* in lines */
+ u32 stride; /* in bytes */
+};
+
+struct intel_vgpu_cursor_plane_format {
+ u8 enabled;
+ u8 mode; /* cursor mode select */
+ u8 bpp; /* bits per pixel */
+ u32 drm_format; /* format in DRM definition */
+ u32 base; /* cursor base in graphics memory */
+ u64 base_gpa;
+ u32 x_pos; /* in pixels */
+ u32 y_pos; /* in lines */
+ u8 x_sign; /* X Position Sign */
+ u8 y_sign; /* Y Position Sign */
+ u32 width; /* in pixels */
+ u32 height; /* in lines */
+ u32 x_hot; /* in pixels */
+ u32 y_hot; /* in pixels */
+};
+
+struct intel_vgpu_pipe_format {
+ struct intel_vgpu_primary_plane_format primary;
+ struct intel_vgpu_sprite_plane_format sprite;
+ struct intel_vgpu_cursor_plane_format cursor;
+ enum DDI_PORT ddi_port; /* the DDI port that pipe is connected to */
+};
+
+struct intel_vgpu_fb_format {
+ struct intel_vgpu_pipe_format pipes[I915_MAX_PIPES];
+};
+
+int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_primary_plane_format *plane);
+int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_cursor_plane_format *plane);
+int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
+ struct intel_vgpu_sprite_plane_format *plane);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/firmware.c b/drivers/gpu/drm/i915/gvt/firmware.c
new file mode 100644
index 000000000..54fe44223
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/firmware.c
@@ -0,0 +1,266 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Changbin Du <changbin.du@intel.com>
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/crc32.h>
+
+#include "i915_drv.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+
+#define FIRMWARE_VERSION (0x0)
+
+struct gvt_firmware_header {
+ u64 magic;
+ u32 crc32; /* protect the data after this field */
+ u32 version;
+ u64 cfg_space_size;
+ u64 cfg_space_offset; /* offset in the file */
+ u64 mmio_size;
+ u64 mmio_offset; /* offset in the file */
+ unsigned char data[1];
+};
+
+#define dev_to_drm_minor(d) dev_get_drvdata((d))
+
+static ssize_t
+gvt_firmware_read(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t offset, size_t count)
+{
+ memcpy(buf, attr->private + offset, count);
+ return count;
+}
+
+static struct bin_attribute firmware_attr = {
+ .attr = {.name = "gvt_firmware", .mode = (S_IRUSR)},
+ .read = gvt_firmware_read,
+ .write = NULL,
+ .mmap = NULL,
+};
+
+static int expose_firmware_sysfs(struct intel_gvt *gvt)
+{
+ struct intel_gvt_device_info *info = &gvt->device_info;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ struct pci_dev *pdev = to_pci_dev(i915->drm.dev);
+ struct gvt_firmware_header *h;
+ void *firmware;
+ void *p;
+ unsigned long size, crc32_start;
+ int ret;
+
+ size = sizeof(*h) + info->mmio_size + info->cfg_space_size;
+ firmware = vzalloc(size);
+ if (!firmware)
+ return -ENOMEM;
+
+ h = firmware;
+
+ h->magic = VGT_MAGIC;
+ h->version = FIRMWARE_VERSION;
+ h->cfg_space_size = info->cfg_space_size;
+ h->cfg_space_offset = offsetof(struct gvt_firmware_header, data);
+ h->mmio_size = info->mmio_size;
+ h->mmio_offset = h->cfg_space_offset + h->cfg_space_size;
+
+ p = firmware + h->cfg_space_offset;
+
+ memcpy(gvt->firmware.cfg_space, i915->vgpu.initial_cfg_space,
+ info->cfg_space_size);
+ memcpy(p, gvt->firmware.cfg_space, info->cfg_space_size);
+
+ p = firmware + h->mmio_offset;
+
+ memcpy(gvt->firmware.mmio, i915->vgpu.initial_mmio,
+ info->mmio_size);
+
+ memcpy(p, gvt->firmware.mmio, info->mmio_size);
+
+ crc32_start = offsetof(struct gvt_firmware_header, crc32) + 4;
+ h->crc32 = crc32_le(0, firmware + crc32_start, size - crc32_start);
+
+ firmware_attr.size = size;
+ firmware_attr.private = firmware;
+
+ ret = device_create_bin_file(&pdev->dev, &firmware_attr);
+ if (ret) {
+ vfree(firmware);
+ return ret;
+ }
+ return 0;
+}
+
+static void clean_firmware_sysfs(struct intel_gvt *gvt)
+{
+ struct pci_dev *pdev = to_pci_dev(gvt->gt->i915->drm.dev);
+
+ device_remove_bin_file(&pdev->dev, &firmware_attr);
+ vfree(firmware_attr.private);
+}
+
+/**
+ * intel_gvt_free_firmware - free GVT firmware
+ * @gvt: intel gvt device
+ *
+ */
+void intel_gvt_free_firmware(struct intel_gvt *gvt)
+{
+ if (!gvt->firmware.firmware_loaded)
+ clean_firmware_sysfs(gvt);
+
+ kfree(gvt->firmware.cfg_space);
+ vfree(gvt->firmware.mmio);
+}
+
+static int verify_firmware(struct intel_gvt *gvt,
+ const struct firmware *fw)
+{
+ struct intel_gvt_device_info *info = &gvt->device_info;
+ struct pci_dev *pdev = to_pci_dev(gvt->gt->i915->drm.dev);
+ struct gvt_firmware_header *h;
+ unsigned long id, crc32_start;
+ const void *mem;
+ const char *item;
+ u64 file, request;
+
+ h = (struct gvt_firmware_header *)fw->data;
+
+ crc32_start = offsetofend(struct gvt_firmware_header, crc32);
+ mem = fw->data + crc32_start;
+
+#define VERIFY(s, a, b) do { \
+ item = (s); file = (u64)(a); request = (u64)(b); \
+ if ((a) != (b)) \
+ goto invalid_firmware; \
+} while (0)
+
+ VERIFY("magic number", h->magic, VGT_MAGIC);
+ VERIFY("version", h->version, FIRMWARE_VERSION);
+ VERIFY("crc32", h->crc32, crc32_le(0, mem, fw->size - crc32_start));
+ VERIFY("cfg space size", h->cfg_space_size, info->cfg_space_size);
+ VERIFY("mmio size", h->mmio_size, info->mmio_size);
+
+ mem = (fw->data + h->cfg_space_offset);
+
+ id = *(u16 *)(mem + PCI_VENDOR_ID);
+ VERIFY("vender id", id, pdev->vendor);
+
+ id = *(u16 *)(mem + PCI_DEVICE_ID);
+ VERIFY("device id", id, pdev->device);
+
+ id = *(u8 *)(mem + PCI_REVISION_ID);
+ VERIFY("revision id", id, pdev->revision);
+
+#undef VERIFY
+ return 0;
+
+invalid_firmware:
+ gvt_dbg_core("Invalid firmware: %s [file] 0x%llx [request] 0x%llx\n",
+ item, file, request);
+ return -EINVAL;
+}
+
+#define GVT_FIRMWARE_PATH "i915/gvt"
+
+/**
+ * intel_gvt_load_firmware - load GVT firmware
+ * @gvt: intel gvt device
+ *
+ */
+int intel_gvt_load_firmware(struct intel_gvt *gvt)
+{
+ struct intel_gvt_device_info *info = &gvt->device_info;
+ struct pci_dev *pdev = to_pci_dev(gvt->gt->i915->drm.dev);
+ struct intel_gvt_firmware *firmware = &gvt->firmware;
+ struct gvt_firmware_header *h;
+ const struct firmware *fw;
+ char *path;
+ void *mem;
+ int ret;
+
+ path = kmalloc(PATH_MAX, GFP_KERNEL);
+ if (!path)
+ return -ENOMEM;
+
+ mem = kmalloc(info->cfg_space_size, GFP_KERNEL);
+ if (!mem) {
+ kfree(path);
+ return -ENOMEM;
+ }
+
+ firmware->cfg_space = mem;
+
+ mem = vmalloc(info->mmio_size);
+ if (!mem) {
+ kfree(path);
+ kfree(firmware->cfg_space);
+ return -ENOMEM;
+ }
+
+ firmware->mmio = mem;
+
+ sprintf(path, "%s/vid_0x%04x_did_0x%04x_rid_0x%02x.golden_hw_state",
+ GVT_FIRMWARE_PATH, pdev->vendor, pdev->device,
+ pdev->revision);
+
+ gvt_dbg_core("request hw state firmware %s...\n", path);
+
+ ret = request_firmware(&fw, path, gvt->gt->i915->drm.dev);
+ kfree(path);
+
+ if (ret)
+ goto expose_firmware;
+
+ gvt_dbg_core("success.\n");
+
+ ret = verify_firmware(gvt, fw);
+ if (ret)
+ goto out_free_fw;
+
+ gvt_dbg_core("verified.\n");
+
+ h = (struct gvt_firmware_header *)fw->data;
+
+ memcpy(firmware->cfg_space, fw->data + h->cfg_space_offset,
+ h->cfg_space_size);
+ memcpy(firmware->mmio, fw->data + h->mmio_offset,
+ h->mmio_size);
+
+ release_firmware(fw);
+ firmware->firmware_loaded = true;
+ return 0;
+
+out_free_fw:
+ release_firmware(fw);
+expose_firmware:
+ expose_firmware_sysfs(gvt);
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
new file mode 100644
index 000000000..79bf1be68
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -0,0 +1,2926 @@
+/*
+ * GTT virtualization
+ *
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ * Xiao Zheng <xiao.zheng@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+#include "trace.h"
+
+#include "gt/intel_gt_regs.h"
+
+#if defined(VERBOSE_DEBUG)
+#define gvt_vdbg_mm(fmt, args...) gvt_dbg_mm(fmt, ##args)
+#else
+#define gvt_vdbg_mm(fmt, args...)
+#endif
+
+static bool enable_out_of_sync = false;
+static int preallocated_oos_pages = 8192;
+
+static bool intel_gvt_is_valid_gfn(struct intel_vgpu *vgpu, unsigned long gfn)
+{
+ struct kvm *kvm = vgpu->vfio_device.kvm;
+ int idx;
+ bool ret;
+
+ if (!vgpu->attached)
+ return false;
+
+ idx = srcu_read_lock(&kvm->srcu);
+ ret = kvm_is_visible_gfn(kvm, gfn);
+ srcu_read_unlock(&kvm->srcu, idx);
+
+ return ret;
+}
+
+/*
+ * validate a gm address and related range size,
+ * translate it to host gm address
+ */
+bool intel_gvt_ggtt_validate_range(struct intel_vgpu *vgpu, u64 addr, u32 size)
+{
+ if (size == 0)
+ return vgpu_gmadr_is_valid(vgpu, addr);
+
+ if (vgpu_gmadr_is_aperture(vgpu, addr) &&
+ vgpu_gmadr_is_aperture(vgpu, addr + size - 1))
+ return true;
+ else if (vgpu_gmadr_is_hidden(vgpu, addr) &&
+ vgpu_gmadr_is_hidden(vgpu, addr + size - 1))
+ return true;
+
+ gvt_dbg_mm("Invalid ggtt range at 0x%llx, size: 0x%x\n",
+ addr, size);
+ return false;
+}
+
+/* translate a guest gmadr to host gmadr */
+int intel_gvt_ggtt_gmadr_g2h(struct intel_vgpu *vgpu, u64 g_addr, u64 *h_addr)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ if (drm_WARN(&i915->drm, !vgpu_gmadr_is_valid(vgpu, g_addr),
+ "invalid guest gmadr %llx\n", g_addr))
+ return -EACCES;
+
+ if (vgpu_gmadr_is_aperture(vgpu, g_addr))
+ *h_addr = vgpu_aperture_gmadr_base(vgpu)
+ + (g_addr - vgpu_aperture_offset(vgpu));
+ else
+ *h_addr = vgpu_hidden_gmadr_base(vgpu)
+ + (g_addr - vgpu_hidden_offset(vgpu));
+ return 0;
+}
+
+/* translate a host gmadr to guest gmadr */
+int intel_gvt_ggtt_gmadr_h2g(struct intel_vgpu *vgpu, u64 h_addr, u64 *g_addr)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+
+ if (drm_WARN(&i915->drm, !gvt_gmadr_is_valid(vgpu->gvt, h_addr),
+ "invalid host gmadr %llx\n", h_addr))
+ return -EACCES;
+
+ if (gvt_gmadr_is_aperture(vgpu->gvt, h_addr))
+ *g_addr = vgpu_aperture_gmadr_base(vgpu)
+ + (h_addr - gvt_aperture_gmadr_base(vgpu->gvt));
+ else
+ *g_addr = vgpu_hidden_gmadr_base(vgpu)
+ + (h_addr - gvt_hidden_gmadr_base(vgpu->gvt));
+ return 0;
+}
+
+int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index,
+ unsigned long *h_index)
+{
+ u64 h_addr;
+ int ret;
+
+ ret = intel_gvt_ggtt_gmadr_g2h(vgpu, g_index << I915_GTT_PAGE_SHIFT,
+ &h_addr);
+ if (ret)
+ return ret;
+
+ *h_index = h_addr >> I915_GTT_PAGE_SHIFT;
+ return 0;
+}
+
+int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index,
+ unsigned long *g_index)
+{
+ u64 g_addr;
+ int ret;
+
+ ret = intel_gvt_ggtt_gmadr_h2g(vgpu, h_index << I915_GTT_PAGE_SHIFT,
+ &g_addr);
+ if (ret)
+ return ret;
+
+ *g_index = g_addr >> I915_GTT_PAGE_SHIFT;
+ return 0;
+}
+
+#define gtt_type_is_entry(type) \
+ (type > GTT_TYPE_INVALID && type < GTT_TYPE_PPGTT_ENTRY \
+ && type != GTT_TYPE_PPGTT_PTE_ENTRY \
+ && type != GTT_TYPE_PPGTT_ROOT_ENTRY)
+
+#define gtt_type_is_pt(type) \
+ (type >= GTT_TYPE_PPGTT_PTE_PT && type < GTT_TYPE_MAX)
+
+#define gtt_type_is_pte_pt(type) \
+ (type == GTT_TYPE_PPGTT_PTE_PT)
+
+#define gtt_type_is_root_pointer(type) \
+ (gtt_type_is_entry(type) && type > GTT_TYPE_PPGTT_ROOT_ENTRY)
+
+#define gtt_init_entry(e, t, p, v) do { \
+ (e)->type = t; \
+ (e)->pdev = p; \
+ memcpy(&(e)->val64, &v, sizeof(v)); \
+} while (0)
+
+/*
+ * Mappings between GTT_TYPE* enumerations.
+ * Following information can be found according to the given type:
+ * - type of next level page table
+ * - type of entry inside this level page table
+ * - type of entry with PSE set
+ *
+ * If the given type doesn't have such a kind of information,
+ * e.g. give a l4 root entry type, then request to get its PSE type,
+ * give a PTE page table type, then request to get its next level page
+ * table type, as we know l4 root entry doesn't have a PSE bit,
+ * and a PTE page table doesn't have a next level page table type,
+ * GTT_TYPE_INVALID will be returned. This is useful when traversing a
+ * page table.
+ */
+
+struct gtt_type_table_entry {
+ int entry_type;
+ int pt_type;
+ int next_pt_type;
+ int pse_entry_type;
+};
+
+#define GTT_TYPE_TABLE_ENTRY(type, e_type, cpt_type, npt_type, pse_type) \
+ [type] = { \
+ .entry_type = e_type, \
+ .pt_type = cpt_type, \
+ .next_pt_type = npt_type, \
+ .pse_entry_type = pse_type, \
+ }
+
+static const struct gtt_type_table_entry gtt_type_table[] = {
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L4_ENTRY,
+ GTT_TYPE_PPGTT_ROOT_L4_ENTRY,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PML4_PT,
+ GTT_TYPE_INVALID),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_PT,
+ GTT_TYPE_PPGTT_PML4_ENTRY,
+ GTT_TYPE_PPGTT_PML4_PT,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_INVALID),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_ENTRY,
+ GTT_TYPE_PPGTT_PML4_ENTRY,
+ GTT_TYPE_PPGTT_PML4_PT,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_INVALID),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_PPGTT_PDP_ENTRY,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PTE_1G_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L3_ENTRY,
+ GTT_TYPE_PPGTT_ROOT_L3_ENTRY,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PTE_1G_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_ENTRY,
+ GTT_TYPE_PPGTT_PDP_ENTRY,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PTE_1G_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PDE_ENTRY,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_PPGTT_PTE_2M_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_ENTRY,
+ GTT_TYPE_PPGTT_PDE_ENTRY,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_PPGTT_PTE_2M_ENTRY),
+ /* We take IPS bit as 'PSE' for PTE level. */
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PTE_64K_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PTE_64K_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_64K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PTE_64K_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_2M_ENTRY,
+ GTT_TYPE_PPGTT_PDE_ENTRY,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PTE_2M_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_1G_ENTRY,
+ GTT_TYPE_PPGTT_PDP_ENTRY,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_PPGTT_PTE_1G_ENTRY),
+ GTT_TYPE_TABLE_ENTRY(GTT_TYPE_GGTT_PTE,
+ GTT_TYPE_GGTT_PTE,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_INVALID,
+ GTT_TYPE_INVALID),
+};
+
+static inline int get_next_pt_type(int type)
+{
+ return gtt_type_table[type].next_pt_type;
+}
+
+static inline int get_pt_type(int type)
+{
+ return gtt_type_table[type].pt_type;
+}
+
+static inline int get_entry_type(int type)
+{
+ return gtt_type_table[type].entry_type;
+}
+
+static inline int get_pse_type(int type)
+{
+ return gtt_type_table[type].pse_entry_type;
+}
+
+static u64 read_pte64(struct i915_ggtt *ggtt, unsigned long index)
+{
+ void __iomem *addr = (gen8_pte_t __iomem *)ggtt->gsm + index;
+
+ return readq(addr);
+}
+
+static void ggtt_invalidate(struct intel_gt *gt)
+{
+ mmio_hw_access_pre(gt);
+ intel_uncore_write(gt->uncore, GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN);
+ mmio_hw_access_post(gt);
+}
+
+static void write_pte64(struct i915_ggtt *ggtt, unsigned long index, u64 pte)
+{
+ void __iomem *addr = (gen8_pte_t __iomem *)ggtt->gsm + index;
+
+ writeq(pte, addr);
+}
+
+static inline int gtt_get_entry64(void *pt,
+ struct intel_gvt_gtt_entry *e,
+ unsigned long index, bool hypervisor_access, unsigned long gpa,
+ struct intel_vgpu *vgpu)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ int ret;
+
+ if (WARN_ON(info->gtt_entry_size != 8))
+ return -EINVAL;
+
+ if (hypervisor_access) {
+ ret = intel_gvt_read_gpa(vgpu, gpa +
+ (index << info->gtt_entry_size_shift),
+ &e->val64, 8);
+ if (WARN_ON(ret))
+ return ret;
+ } else if (!pt) {
+ e->val64 = read_pte64(vgpu->gvt->gt->ggtt, index);
+ } else {
+ e->val64 = *((u64 *)pt + index);
+ }
+ return 0;
+}
+
+static inline int gtt_set_entry64(void *pt,
+ struct intel_gvt_gtt_entry *e,
+ unsigned long index, bool hypervisor_access, unsigned long gpa,
+ struct intel_vgpu *vgpu)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ int ret;
+
+ if (WARN_ON(info->gtt_entry_size != 8))
+ return -EINVAL;
+
+ if (hypervisor_access) {
+ ret = intel_gvt_write_gpa(vgpu, gpa +
+ (index << info->gtt_entry_size_shift),
+ &e->val64, 8);
+ if (WARN_ON(ret))
+ return ret;
+ } else if (!pt) {
+ write_pte64(vgpu->gvt->gt->ggtt, index, e->val64);
+ } else {
+ *((u64 *)pt + index) = e->val64;
+ }
+ return 0;
+}
+
+#define GTT_HAW 46
+
+#define ADDR_1G_MASK GENMASK_ULL(GTT_HAW - 1, 30)
+#define ADDR_2M_MASK GENMASK_ULL(GTT_HAW - 1, 21)
+#define ADDR_64K_MASK GENMASK_ULL(GTT_HAW - 1, 16)
+#define ADDR_4K_MASK GENMASK_ULL(GTT_HAW - 1, 12)
+
+#define GTT_SPTE_FLAG_MASK GENMASK_ULL(62, 52)
+#define GTT_SPTE_FLAG_64K_SPLITED BIT(52) /* splited 64K gtt entry */
+
+#define GTT_64K_PTE_STRIDE 16
+
+static unsigned long gen8_gtt_get_pfn(struct intel_gvt_gtt_entry *e)
+{
+ unsigned long pfn;
+
+ if (e->type == GTT_TYPE_PPGTT_PTE_1G_ENTRY)
+ pfn = (e->val64 & ADDR_1G_MASK) >> PAGE_SHIFT;
+ else if (e->type == GTT_TYPE_PPGTT_PTE_2M_ENTRY)
+ pfn = (e->val64 & ADDR_2M_MASK) >> PAGE_SHIFT;
+ else if (e->type == GTT_TYPE_PPGTT_PTE_64K_ENTRY)
+ pfn = (e->val64 & ADDR_64K_MASK) >> PAGE_SHIFT;
+ else
+ pfn = (e->val64 & ADDR_4K_MASK) >> PAGE_SHIFT;
+ return pfn;
+}
+
+static void gen8_gtt_set_pfn(struct intel_gvt_gtt_entry *e, unsigned long pfn)
+{
+ if (e->type == GTT_TYPE_PPGTT_PTE_1G_ENTRY) {
+ e->val64 &= ~ADDR_1G_MASK;
+ pfn &= (ADDR_1G_MASK >> PAGE_SHIFT);
+ } else if (e->type == GTT_TYPE_PPGTT_PTE_2M_ENTRY) {
+ e->val64 &= ~ADDR_2M_MASK;
+ pfn &= (ADDR_2M_MASK >> PAGE_SHIFT);
+ } else if (e->type == GTT_TYPE_PPGTT_PTE_64K_ENTRY) {
+ e->val64 &= ~ADDR_64K_MASK;
+ pfn &= (ADDR_64K_MASK >> PAGE_SHIFT);
+ } else {
+ e->val64 &= ~ADDR_4K_MASK;
+ pfn &= (ADDR_4K_MASK >> PAGE_SHIFT);
+ }
+
+ e->val64 |= (pfn << PAGE_SHIFT);
+}
+
+static bool gen8_gtt_test_pse(struct intel_gvt_gtt_entry *e)
+{
+ return !!(e->val64 & _PAGE_PSE);
+}
+
+static void gen8_gtt_clear_pse(struct intel_gvt_gtt_entry *e)
+{
+ if (gen8_gtt_test_pse(e)) {
+ switch (e->type) {
+ case GTT_TYPE_PPGTT_PTE_2M_ENTRY:
+ e->val64 &= ~_PAGE_PSE;
+ e->type = GTT_TYPE_PPGTT_PDE_ENTRY;
+ break;
+ case GTT_TYPE_PPGTT_PTE_1G_ENTRY:
+ e->type = GTT_TYPE_PPGTT_PDP_ENTRY;
+ e->val64 &= ~_PAGE_PSE;
+ break;
+ default:
+ WARN_ON(1);
+ }
+ }
+}
+
+static bool gen8_gtt_test_ips(struct intel_gvt_gtt_entry *e)
+{
+ if (GEM_WARN_ON(e->type != GTT_TYPE_PPGTT_PDE_ENTRY))
+ return false;
+
+ return !!(e->val64 & GEN8_PDE_IPS_64K);
+}
+
+static void gen8_gtt_clear_ips(struct intel_gvt_gtt_entry *e)
+{
+ if (GEM_WARN_ON(e->type != GTT_TYPE_PPGTT_PDE_ENTRY))
+ return;
+
+ e->val64 &= ~GEN8_PDE_IPS_64K;
+}
+
+static bool gen8_gtt_test_present(struct intel_gvt_gtt_entry *e)
+{
+ /*
+ * i915 writes PDP root pointer registers without present bit,
+ * it also works, so we need to treat root pointer entry
+ * specifically.
+ */
+ if (e->type == GTT_TYPE_PPGTT_ROOT_L3_ENTRY
+ || e->type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY)
+ return (e->val64 != 0);
+ else
+ return (e->val64 & GEN8_PAGE_PRESENT);
+}
+
+static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e)
+{
+ e->val64 &= ~GEN8_PAGE_PRESENT;
+}
+
+static void gtt_entry_set_present(struct intel_gvt_gtt_entry *e)
+{
+ e->val64 |= GEN8_PAGE_PRESENT;
+}
+
+static bool gen8_gtt_test_64k_splited(struct intel_gvt_gtt_entry *e)
+{
+ return !!(e->val64 & GTT_SPTE_FLAG_64K_SPLITED);
+}
+
+static void gen8_gtt_set_64k_splited(struct intel_gvt_gtt_entry *e)
+{
+ e->val64 |= GTT_SPTE_FLAG_64K_SPLITED;
+}
+
+static void gen8_gtt_clear_64k_splited(struct intel_gvt_gtt_entry *e)
+{
+ e->val64 &= ~GTT_SPTE_FLAG_64K_SPLITED;
+}
+
+/*
+ * Per-platform GMA routines.
+ */
+static unsigned long gma_to_ggtt_pte_index(unsigned long gma)
+{
+ unsigned long x = (gma >> I915_GTT_PAGE_SHIFT);
+
+ trace_gma_index(__func__, gma, x);
+ return x;
+}
+
+#define DEFINE_PPGTT_GMA_TO_INDEX(prefix, ename, exp) \
+static unsigned long prefix##_gma_to_##ename##_index(unsigned long gma) \
+{ \
+ unsigned long x = (exp); \
+ trace_gma_index(__func__, gma, x); \
+ return x; \
+}
+
+DEFINE_PPGTT_GMA_TO_INDEX(gen8, pte, (gma >> 12 & 0x1ff));
+DEFINE_PPGTT_GMA_TO_INDEX(gen8, pde, (gma >> 21 & 0x1ff));
+DEFINE_PPGTT_GMA_TO_INDEX(gen8, l3_pdp, (gma >> 30 & 0x3));
+DEFINE_PPGTT_GMA_TO_INDEX(gen8, l4_pdp, (gma >> 30 & 0x1ff));
+DEFINE_PPGTT_GMA_TO_INDEX(gen8, pml4, (gma >> 39 & 0x1ff));
+
+static const struct intel_gvt_gtt_pte_ops gen8_gtt_pte_ops = {
+ .get_entry = gtt_get_entry64,
+ .set_entry = gtt_set_entry64,
+ .clear_present = gtt_entry_clear_present,
+ .set_present = gtt_entry_set_present,
+ .test_present = gen8_gtt_test_present,
+ .test_pse = gen8_gtt_test_pse,
+ .clear_pse = gen8_gtt_clear_pse,
+ .clear_ips = gen8_gtt_clear_ips,
+ .test_ips = gen8_gtt_test_ips,
+ .clear_64k_splited = gen8_gtt_clear_64k_splited,
+ .set_64k_splited = gen8_gtt_set_64k_splited,
+ .test_64k_splited = gen8_gtt_test_64k_splited,
+ .get_pfn = gen8_gtt_get_pfn,
+ .set_pfn = gen8_gtt_set_pfn,
+};
+
+static const struct intel_gvt_gtt_gma_ops gen8_gtt_gma_ops = {
+ .gma_to_ggtt_pte_index = gma_to_ggtt_pte_index,
+ .gma_to_pte_index = gen8_gma_to_pte_index,
+ .gma_to_pde_index = gen8_gma_to_pde_index,
+ .gma_to_l3_pdp_index = gen8_gma_to_l3_pdp_index,
+ .gma_to_l4_pdp_index = gen8_gma_to_l4_pdp_index,
+ .gma_to_pml4_index = gen8_gma_to_pml4_index,
+};
+
+/* Update entry type per pse and ips bit. */
+static void update_entry_type_for_real(const struct intel_gvt_gtt_pte_ops *pte_ops,
+ struct intel_gvt_gtt_entry *entry, bool ips)
+{
+ switch (entry->type) {
+ case GTT_TYPE_PPGTT_PDE_ENTRY:
+ case GTT_TYPE_PPGTT_PDP_ENTRY:
+ if (pte_ops->test_pse(entry))
+ entry->type = get_pse_type(entry->type);
+ break;
+ case GTT_TYPE_PPGTT_PTE_4K_ENTRY:
+ if (ips)
+ entry->type = get_pse_type(entry->type);
+ break;
+ default:
+ GEM_BUG_ON(!gtt_type_is_entry(entry->type));
+ }
+
+ GEM_BUG_ON(entry->type == GTT_TYPE_INVALID);
+}
+
+/*
+ * MM helpers.
+ */
+static void _ppgtt_get_root_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index,
+ bool guest)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_PPGTT);
+
+ entry->type = mm->ppgtt_mm.root_entry_type;
+ pte_ops->get_entry(guest ? mm->ppgtt_mm.guest_pdps :
+ mm->ppgtt_mm.shadow_pdps,
+ entry, index, false, 0, mm->vgpu);
+ update_entry_type_for_real(pte_ops, entry, false);
+}
+
+static inline void ppgtt_get_guest_root_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ _ppgtt_get_root_entry(mm, entry, index, true);
+}
+
+static inline void ppgtt_get_shadow_root_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ _ppgtt_get_root_entry(mm, entry, index, false);
+}
+
+static void _ppgtt_set_root_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index,
+ bool guest)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+
+ pte_ops->set_entry(guest ? mm->ppgtt_mm.guest_pdps :
+ mm->ppgtt_mm.shadow_pdps,
+ entry, index, false, 0, mm->vgpu);
+}
+
+static inline void ppgtt_set_shadow_root_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ _ppgtt_set_root_entry(mm, entry, index, false);
+}
+
+static void ggtt_get_guest_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_GGTT);
+
+ entry->type = GTT_TYPE_GGTT_PTE;
+ pte_ops->get_entry(mm->ggtt_mm.virtual_ggtt, entry, index,
+ false, 0, mm->vgpu);
+}
+
+static void ggtt_set_guest_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_GGTT);
+
+ pte_ops->set_entry(mm->ggtt_mm.virtual_ggtt, entry, index,
+ false, 0, mm->vgpu);
+}
+
+static void ggtt_get_host_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_GGTT);
+
+ pte_ops->get_entry(NULL, entry, index, false, 0, mm->vgpu);
+}
+
+static void ggtt_set_host_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *entry, unsigned long index)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = mm->vgpu->gvt->gtt.pte_ops;
+ unsigned long offset = index;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_GGTT);
+
+ if (vgpu_gmadr_is_aperture(mm->vgpu, index << I915_GTT_PAGE_SHIFT)) {
+ offset -= (vgpu_aperture_gmadr_base(mm->vgpu) >> PAGE_SHIFT);
+ mm->ggtt_mm.host_ggtt_aperture[offset] = entry->val64;
+ } else if (vgpu_gmadr_is_hidden(mm->vgpu, index << I915_GTT_PAGE_SHIFT)) {
+ offset -= (vgpu_hidden_gmadr_base(mm->vgpu) >> PAGE_SHIFT);
+ mm->ggtt_mm.host_ggtt_hidden[offset] = entry->val64;
+ }
+
+ pte_ops->set_entry(NULL, entry, index, false, 0, mm->vgpu);
+}
+
+/*
+ * PPGTT shadow page table helpers.
+ */
+static inline int ppgtt_spt_get_entry(
+ struct intel_vgpu_ppgtt_spt *spt,
+ void *page_table, int type,
+ struct intel_gvt_gtt_entry *e, unsigned long index,
+ bool guest)
+{
+ struct intel_gvt *gvt = spt->vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+ int ret;
+
+ e->type = get_entry_type(type);
+
+ if (WARN(!gtt_type_is_entry(e->type), "invalid entry type\n"))
+ return -EINVAL;
+
+ ret = ops->get_entry(page_table, e, index, guest,
+ spt->guest_page.gfn << I915_GTT_PAGE_SHIFT,
+ spt->vgpu);
+ if (ret)
+ return ret;
+
+ update_entry_type_for_real(ops, e, guest ?
+ spt->guest_page.pde_ips : false);
+
+ gvt_vdbg_mm("read ppgtt entry, spt type %d, entry type %d, index %lu, value %llx\n",
+ type, e->type, index, e->val64);
+ return 0;
+}
+
+static inline int ppgtt_spt_set_entry(
+ struct intel_vgpu_ppgtt_spt *spt,
+ void *page_table, int type,
+ struct intel_gvt_gtt_entry *e, unsigned long index,
+ bool guest)
+{
+ struct intel_gvt *gvt = spt->vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+
+ if (WARN(!gtt_type_is_entry(e->type), "invalid entry type\n"))
+ return -EINVAL;
+
+ gvt_vdbg_mm("set ppgtt entry, spt type %d, entry type %d, index %lu, value %llx\n",
+ type, e->type, index, e->val64);
+
+ return ops->set_entry(page_table, e, index, guest,
+ spt->guest_page.gfn << I915_GTT_PAGE_SHIFT,
+ spt->vgpu);
+}
+
+#define ppgtt_get_guest_entry(spt, e, index) \
+ ppgtt_spt_get_entry(spt, NULL, \
+ spt->guest_page.type, e, index, true)
+
+#define ppgtt_set_guest_entry(spt, e, index) \
+ ppgtt_spt_set_entry(spt, NULL, \
+ spt->guest_page.type, e, index, true)
+
+#define ppgtt_get_shadow_entry(spt, e, index) \
+ ppgtt_spt_get_entry(spt, spt->shadow_page.vaddr, \
+ spt->shadow_page.type, e, index, false)
+
+#define ppgtt_set_shadow_entry(spt, e, index) \
+ ppgtt_spt_set_entry(spt, spt->shadow_page.vaddr, \
+ spt->shadow_page.type, e, index, false)
+
+static void *alloc_spt(gfp_t gfp_mask)
+{
+ struct intel_vgpu_ppgtt_spt *spt;
+
+ spt = kzalloc(sizeof(*spt), gfp_mask);
+ if (!spt)
+ return NULL;
+
+ spt->shadow_page.page = alloc_page(gfp_mask);
+ if (!spt->shadow_page.page) {
+ kfree(spt);
+ return NULL;
+ }
+ return spt;
+}
+
+static void free_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ __free_page(spt->shadow_page.page);
+ kfree(spt);
+}
+
+static int detach_oos_page(struct intel_vgpu *vgpu,
+ struct intel_vgpu_oos_page *oos_page);
+
+static void ppgtt_free_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct device *kdev = spt->vgpu->gvt->gt->i915->drm.dev;
+
+ trace_spt_free(spt->vgpu->id, spt, spt->guest_page.type);
+
+ dma_unmap_page(kdev, spt->shadow_page.mfn << I915_GTT_PAGE_SHIFT, 4096,
+ DMA_BIDIRECTIONAL);
+
+ radix_tree_delete(&spt->vgpu->gtt.spt_tree, spt->shadow_page.mfn);
+
+ if (spt->guest_page.gfn) {
+ if (spt->guest_page.oos_page)
+ detach_oos_page(spt->vgpu, spt->guest_page.oos_page);
+
+ intel_vgpu_unregister_page_track(spt->vgpu, spt->guest_page.gfn);
+ }
+
+ list_del_init(&spt->post_shadow_list);
+ free_spt(spt);
+}
+
+static void ppgtt_free_all_spt(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_ppgtt_spt *spt, *spn;
+ struct radix_tree_iter iter;
+ LIST_HEAD(all_spt);
+ void __rcu **slot;
+
+ rcu_read_lock();
+ radix_tree_for_each_slot(slot, &vgpu->gtt.spt_tree, &iter, 0) {
+ spt = radix_tree_deref_slot(slot);
+ list_move(&spt->post_shadow_list, &all_spt);
+ }
+ rcu_read_unlock();
+
+ list_for_each_entry_safe(spt, spn, &all_spt, post_shadow_list)
+ ppgtt_free_spt(spt);
+}
+
+static int ppgtt_handle_guest_write_page_table_bytes(
+ struct intel_vgpu_ppgtt_spt *spt,
+ u64 pa, void *p_data, int bytes);
+
+static int ppgtt_write_protection_handler(
+ struct intel_vgpu_page_track *page_track,
+ u64 gpa, void *data, int bytes)
+{
+ struct intel_vgpu_ppgtt_spt *spt = page_track->priv_data;
+
+ int ret;
+
+ if (bytes != 4 && bytes != 8)
+ return -EINVAL;
+
+ ret = ppgtt_handle_guest_write_page_table_bytes(spt, gpa, data, bytes);
+ if (ret)
+ return ret;
+ return ret;
+}
+
+/* Find a spt by guest gfn. */
+static struct intel_vgpu_ppgtt_spt *intel_vgpu_find_spt_by_gfn(
+ struct intel_vgpu *vgpu, unsigned long gfn)
+{
+ struct intel_vgpu_page_track *track;
+
+ track = intel_vgpu_find_page_track(vgpu, gfn);
+ if (track && track->handler == ppgtt_write_protection_handler)
+ return track->priv_data;
+
+ return NULL;
+}
+
+/* Find the spt by shadow page mfn. */
+static inline struct intel_vgpu_ppgtt_spt *intel_vgpu_find_spt_by_mfn(
+ struct intel_vgpu *vgpu, unsigned long mfn)
+{
+ return radix_tree_lookup(&vgpu->gtt.spt_tree, mfn);
+}
+
+static int reclaim_one_ppgtt_mm(struct intel_gvt *gvt);
+
+/* Allocate shadow page table without guest page. */
+static struct intel_vgpu_ppgtt_spt *ppgtt_alloc_spt(
+ struct intel_vgpu *vgpu, enum intel_gvt_gtt_type type)
+{
+ struct device *kdev = vgpu->gvt->gt->i915->drm.dev;
+ struct intel_vgpu_ppgtt_spt *spt = NULL;
+ dma_addr_t daddr;
+ int ret;
+
+retry:
+ spt = alloc_spt(GFP_KERNEL | __GFP_ZERO);
+ if (!spt) {
+ if (reclaim_one_ppgtt_mm(vgpu->gvt))
+ goto retry;
+
+ gvt_vgpu_err("fail to allocate ppgtt shadow page\n");
+ return ERR_PTR(-ENOMEM);
+ }
+
+ spt->vgpu = vgpu;
+ atomic_set(&spt->refcount, 1);
+ INIT_LIST_HEAD(&spt->post_shadow_list);
+
+ /*
+ * Init shadow_page.
+ */
+ spt->shadow_page.type = type;
+ daddr = dma_map_page(kdev, spt->shadow_page.page,
+ 0, 4096, DMA_BIDIRECTIONAL);
+ if (dma_mapping_error(kdev, daddr)) {
+ gvt_vgpu_err("fail to map dma addr\n");
+ ret = -EINVAL;
+ goto err_free_spt;
+ }
+ spt->shadow_page.vaddr = page_address(spt->shadow_page.page);
+ spt->shadow_page.mfn = daddr >> I915_GTT_PAGE_SHIFT;
+
+ ret = radix_tree_insert(&vgpu->gtt.spt_tree, spt->shadow_page.mfn, spt);
+ if (ret)
+ goto err_unmap_dma;
+
+ return spt;
+
+err_unmap_dma:
+ dma_unmap_page(kdev, daddr, PAGE_SIZE, DMA_BIDIRECTIONAL);
+err_free_spt:
+ free_spt(spt);
+ return ERR_PTR(ret);
+}
+
+/* Allocate shadow page table associated with specific gfn. */
+static struct intel_vgpu_ppgtt_spt *ppgtt_alloc_spt_gfn(
+ struct intel_vgpu *vgpu, enum intel_gvt_gtt_type type,
+ unsigned long gfn, bool guest_pde_ips)
+{
+ struct intel_vgpu_ppgtt_spt *spt;
+ int ret;
+
+ spt = ppgtt_alloc_spt(vgpu, type);
+ if (IS_ERR(spt))
+ return spt;
+
+ /*
+ * Init guest_page.
+ */
+ ret = intel_vgpu_register_page_track(vgpu, gfn,
+ ppgtt_write_protection_handler, spt);
+ if (ret) {
+ ppgtt_free_spt(spt);
+ return ERR_PTR(ret);
+ }
+
+ spt->guest_page.type = type;
+ spt->guest_page.gfn = gfn;
+ spt->guest_page.pde_ips = guest_pde_ips;
+
+ trace_spt_alloc(vgpu->id, spt, type, spt->shadow_page.mfn, gfn);
+
+ return spt;
+}
+
+#define pt_entry_size_shift(spt) \
+ ((spt)->vgpu->gvt->device_info.gtt_entry_size_shift)
+
+#define pt_entries(spt) \
+ (I915_GTT_PAGE_SIZE >> pt_entry_size_shift(spt))
+
+#define for_each_present_guest_entry(spt, e, i) \
+ for (i = 0; i < pt_entries(spt); \
+ i += spt->guest_page.pde_ips ? GTT_64K_PTE_STRIDE : 1) \
+ if (!ppgtt_get_guest_entry(spt, e, i) && \
+ spt->vgpu->gvt->gtt.pte_ops->test_present(e))
+
+#define for_each_present_shadow_entry(spt, e, i) \
+ for (i = 0; i < pt_entries(spt); \
+ i += spt->shadow_page.pde_ips ? GTT_64K_PTE_STRIDE : 1) \
+ if (!ppgtt_get_shadow_entry(spt, e, i) && \
+ spt->vgpu->gvt->gtt.pte_ops->test_present(e))
+
+#define for_each_shadow_entry(spt, e, i) \
+ for (i = 0; i < pt_entries(spt); \
+ i += (spt->shadow_page.pde_ips ? GTT_64K_PTE_STRIDE : 1)) \
+ if (!ppgtt_get_shadow_entry(spt, e, i))
+
+static inline void ppgtt_get_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ int v = atomic_read(&spt->refcount);
+
+ trace_spt_refcount(spt->vgpu->id, "inc", spt, v, (v + 1));
+ atomic_inc(&spt->refcount);
+}
+
+static inline int ppgtt_put_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ int v = atomic_read(&spt->refcount);
+
+ trace_spt_refcount(spt->vgpu->id, "dec", spt, v, (v - 1));
+ return atomic_dec_return(&spt->refcount);
+}
+
+static int ppgtt_invalidate_spt(struct intel_vgpu_ppgtt_spt *spt);
+
+static int ppgtt_invalidate_spt_by_shadow_entry(struct intel_vgpu *vgpu,
+ struct intel_gvt_gtt_entry *e)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *s;
+ enum intel_gvt_gtt_type cur_pt_type;
+
+ GEM_BUG_ON(!gtt_type_is_pt(get_next_pt_type(e->type)));
+
+ if (e->type != GTT_TYPE_PPGTT_ROOT_L3_ENTRY
+ && e->type != GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
+ cur_pt_type = get_next_pt_type(e->type);
+
+ if (!gtt_type_is_pt(cur_pt_type) ||
+ !gtt_type_is_pt(cur_pt_type + 1)) {
+ drm_WARN(&i915->drm, 1,
+ "Invalid page table type, cur_pt_type is: %d\n",
+ cur_pt_type);
+ return -EINVAL;
+ }
+
+ cur_pt_type += 1;
+
+ if (ops->get_pfn(e) ==
+ vgpu->gtt.scratch_pt[cur_pt_type].page_mfn)
+ return 0;
+ }
+ s = intel_vgpu_find_spt_by_mfn(vgpu, ops->get_pfn(e));
+ if (!s) {
+ gvt_vgpu_err("fail to find shadow page: mfn: 0x%lx\n",
+ ops->get_pfn(e));
+ return -ENXIO;
+ }
+ return ppgtt_invalidate_spt(s);
+}
+
+static inline void ppgtt_invalidate_pte(struct intel_vgpu_ppgtt_spt *spt,
+ struct intel_gvt_gtt_entry *entry)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ unsigned long pfn;
+ int type;
+
+ pfn = ops->get_pfn(entry);
+ type = spt->shadow_page.type;
+
+ /* Uninitialized spte or unshadowed spte. */
+ if (!pfn || pfn == vgpu->gtt.scratch_pt[type].page_mfn)
+ return;
+
+ intel_gvt_dma_unmap_guest_page(vgpu, pfn << PAGE_SHIFT);
+}
+
+static int ppgtt_invalidate_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ struct intel_gvt_gtt_entry e;
+ unsigned long index;
+ int ret;
+
+ trace_spt_change(spt->vgpu->id, "die", spt,
+ spt->guest_page.gfn, spt->shadow_page.type);
+
+ if (ppgtt_put_spt(spt) > 0)
+ return 0;
+
+ for_each_present_shadow_entry(spt, &e, index) {
+ switch (e.type) {
+ case GTT_TYPE_PPGTT_PTE_4K_ENTRY:
+ gvt_vdbg_mm("invalidate 4K entry\n");
+ ppgtt_invalidate_pte(spt, &e);
+ break;
+ case GTT_TYPE_PPGTT_PTE_64K_ENTRY:
+ /* We don't setup 64K shadow entry so far. */
+ WARN(1, "suspicious 64K gtt entry\n");
+ continue;
+ case GTT_TYPE_PPGTT_PTE_2M_ENTRY:
+ gvt_vdbg_mm("invalidate 2M entry\n");
+ continue;
+ case GTT_TYPE_PPGTT_PTE_1G_ENTRY:
+ WARN(1, "GVT doesn't support 1GB page\n");
+ continue;
+ case GTT_TYPE_PPGTT_PML4_ENTRY:
+ case GTT_TYPE_PPGTT_PDP_ENTRY:
+ case GTT_TYPE_PPGTT_PDE_ENTRY:
+ gvt_vdbg_mm("invalidate PMUL4/PDP/PDE entry\n");
+ ret = ppgtt_invalidate_spt_by_shadow_entry(
+ spt->vgpu, &e);
+ if (ret)
+ goto fail;
+ break;
+ default:
+ GEM_BUG_ON(1);
+ }
+ }
+
+ trace_spt_change(spt->vgpu->id, "release", spt,
+ spt->guest_page.gfn, spt->shadow_page.type);
+ ppgtt_free_spt(spt);
+ return 0;
+fail:
+ gvt_vgpu_err("fail: shadow page %p shadow entry 0x%llx type %d\n",
+ spt, e.val64, e.type);
+ return ret;
+}
+
+static bool vgpu_ips_enabled(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+
+ if (GRAPHICS_VER(dev_priv) == 9) {
+ u32 ips = vgpu_vreg_t(vgpu, GEN8_GAMW_ECO_DEV_RW_IA) &
+ GAMW_ECO_ENABLE_64K_IPS_FIELD;
+
+ return ips == GAMW_ECO_ENABLE_64K_IPS_FIELD;
+ } else if (GRAPHICS_VER(dev_priv) >= 11) {
+ /* 64K paging only controlled by IPS bit in PTE now. */
+ return true;
+ } else
+ return false;
+}
+
+static int ppgtt_populate_spt(struct intel_vgpu_ppgtt_spt *spt);
+
+static struct intel_vgpu_ppgtt_spt *ppgtt_populate_spt_by_guest_entry(
+ struct intel_vgpu *vgpu, struct intel_gvt_gtt_entry *we)
+{
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *spt = NULL;
+ bool ips = false;
+ int ret;
+
+ GEM_BUG_ON(!gtt_type_is_pt(get_next_pt_type(we->type)));
+
+ if (we->type == GTT_TYPE_PPGTT_PDE_ENTRY)
+ ips = vgpu_ips_enabled(vgpu) && ops->test_ips(we);
+
+ spt = intel_vgpu_find_spt_by_gfn(vgpu, ops->get_pfn(we));
+ if (spt) {
+ ppgtt_get_spt(spt);
+
+ if (ips != spt->guest_page.pde_ips) {
+ spt->guest_page.pde_ips = ips;
+
+ gvt_dbg_mm("reshadow PDE since ips changed\n");
+ clear_page(spt->shadow_page.vaddr);
+ ret = ppgtt_populate_spt(spt);
+ if (ret) {
+ ppgtt_put_spt(spt);
+ goto err;
+ }
+ }
+ } else {
+ int type = get_next_pt_type(we->type);
+
+ if (!gtt_type_is_pt(type)) {
+ ret = -EINVAL;
+ goto err;
+ }
+
+ spt = ppgtt_alloc_spt_gfn(vgpu, type, ops->get_pfn(we), ips);
+ if (IS_ERR(spt)) {
+ ret = PTR_ERR(spt);
+ goto err;
+ }
+
+ ret = intel_vgpu_enable_page_track(vgpu, spt->guest_page.gfn);
+ if (ret)
+ goto err_free_spt;
+
+ ret = ppgtt_populate_spt(spt);
+ if (ret)
+ goto err_free_spt;
+
+ trace_spt_change(vgpu->id, "new", spt, spt->guest_page.gfn,
+ spt->shadow_page.type);
+ }
+ return spt;
+
+err_free_spt:
+ ppgtt_free_spt(spt);
+ spt = NULL;
+err:
+ gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d\n",
+ spt, we->val64, we->type);
+ return ERR_PTR(ret);
+}
+
+static inline void ppgtt_generate_shadow_entry(struct intel_gvt_gtt_entry *se,
+ struct intel_vgpu_ppgtt_spt *s, struct intel_gvt_gtt_entry *ge)
+{
+ const struct intel_gvt_gtt_pte_ops *ops = s->vgpu->gvt->gtt.pte_ops;
+
+ se->type = ge->type;
+ se->val64 = ge->val64;
+
+ /* Because we always split 64KB pages, so clear IPS in shadow PDE. */
+ if (se->type == GTT_TYPE_PPGTT_PDE_ENTRY)
+ ops->clear_ips(se);
+
+ ops->set_pfn(se, s->shadow_page.mfn);
+}
+
+/*
+ * Check if can do 2M page
+ * @vgpu: target vgpu
+ * @entry: target pfn's gtt entry
+ *
+ * Return 1 if 2MB huge gtt shadowing is possible, 0 if miscondition,
+ * negative if found err.
+ */
+static int is_2MB_gtt_possible(struct intel_vgpu *vgpu,
+ struct intel_gvt_gtt_entry *entry)
+{
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ kvm_pfn_t pfn;
+ int ret;
+
+ if (!HAS_PAGE_SIZES(vgpu->gvt->gt->i915, I915_GTT_PAGE_SIZE_2M))
+ return 0;
+
+ if (!vgpu->attached)
+ return -EINVAL;
+ pfn = gfn_to_pfn(vgpu->vfio_device.kvm, ops->get_pfn(entry));
+ if (is_error_noslot_pfn(pfn))
+ return -EINVAL;
+
+ if (!pfn_valid(pfn))
+ return -EINVAL;
+
+ ret = PageTransHuge(pfn_to_page(pfn));
+ kvm_release_pfn_clean(pfn);
+ return ret;
+}
+
+static int split_2MB_gtt_entry(struct intel_vgpu *vgpu,
+ struct intel_vgpu_ppgtt_spt *spt, unsigned long index,
+ struct intel_gvt_gtt_entry *se)
+{
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *sub_spt;
+ struct intel_gvt_gtt_entry sub_se;
+ unsigned long start_gfn;
+ dma_addr_t dma_addr;
+ unsigned long sub_index;
+ int ret;
+
+ gvt_dbg_mm("Split 2M gtt entry, index %lu\n", index);
+
+ start_gfn = ops->get_pfn(se);
+
+ sub_spt = ppgtt_alloc_spt(vgpu, GTT_TYPE_PPGTT_PTE_PT);
+ if (IS_ERR(sub_spt))
+ return PTR_ERR(sub_spt);
+
+ for_each_shadow_entry(sub_spt, &sub_se, sub_index) {
+ ret = intel_gvt_dma_map_guest_page(vgpu, start_gfn + sub_index,
+ PAGE_SIZE, &dma_addr);
+ if (ret)
+ goto err;
+ sub_se.val64 = se->val64;
+
+ /* Copy the PAT field from PDE. */
+ sub_se.val64 &= ~_PAGE_PAT;
+ sub_se.val64 |= (se->val64 & _PAGE_PAT_LARGE) >> 5;
+
+ ops->set_pfn(&sub_se, dma_addr >> PAGE_SHIFT);
+ ppgtt_set_shadow_entry(sub_spt, &sub_se, sub_index);
+ }
+
+ /* Clear dirty field. */
+ se->val64 &= ~_PAGE_DIRTY;
+
+ ops->clear_pse(se);
+ ops->clear_ips(se);
+ ops->set_pfn(se, sub_spt->shadow_page.mfn);
+ ppgtt_set_shadow_entry(spt, se, index);
+ return 0;
+err:
+ /* Cancel the existing addess mappings of DMA addr. */
+ for_each_present_shadow_entry(sub_spt, &sub_se, sub_index) {
+ gvt_vdbg_mm("invalidate 4K entry\n");
+ ppgtt_invalidate_pte(sub_spt, &sub_se);
+ }
+ /* Release the new allocated spt. */
+ trace_spt_change(sub_spt->vgpu->id, "release", sub_spt,
+ sub_spt->guest_page.gfn, sub_spt->shadow_page.type);
+ ppgtt_free_spt(sub_spt);
+ return ret;
+}
+
+static int split_64KB_gtt_entry(struct intel_vgpu *vgpu,
+ struct intel_vgpu_ppgtt_spt *spt, unsigned long index,
+ struct intel_gvt_gtt_entry *se)
+{
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_gvt_gtt_entry entry = *se;
+ unsigned long start_gfn;
+ dma_addr_t dma_addr;
+ int i, ret;
+
+ gvt_vdbg_mm("Split 64K gtt entry, index %lu\n", index);
+
+ GEM_BUG_ON(index % GTT_64K_PTE_STRIDE);
+
+ start_gfn = ops->get_pfn(se);
+
+ entry.type = GTT_TYPE_PPGTT_PTE_4K_ENTRY;
+ ops->set_64k_splited(&entry);
+
+ for (i = 0; i < GTT_64K_PTE_STRIDE; i++) {
+ ret = intel_gvt_dma_map_guest_page(vgpu, start_gfn + i,
+ PAGE_SIZE, &dma_addr);
+ if (ret)
+ return ret;
+
+ ops->set_pfn(&entry, dma_addr >> PAGE_SHIFT);
+ ppgtt_set_shadow_entry(spt, &entry, index + i);
+ }
+ return 0;
+}
+
+static int ppgtt_populate_shadow_entry(struct intel_vgpu *vgpu,
+ struct intel_vgpu_ppgtt_spt *spt, unsigned long index,
+ struct intel_gvt_gtt_entry *ge)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_gvt_gtt_entry se = *ge;
+ unsigned long gfn, page_size = PAGE_SIZE;
+ dma_addr_t dma_addr;
+ int ret;
+
+ if (!pte_ops->test_present(ge))
+ return 0;
+
+ gfn = pte_ops->get_pfn(ge);
+
+ switch (ge->type) {
+ case GTT_TYPE_PPGTT_PTE_4K_ENTRY:
+ gvt_vdbg_mm("shadow 4K gtt entry\n");
+ break;
+ case GTT_TYPE_PPGTT_PTE_64K_ENTRY:
+ gvt_vdbg_mm("shadow 64K gtt entry\n");
+ /*
+ * The layout of 64K page is special, the page size is
+ * controlled by uper PDE. To be simple, we always split
+ * 64K page to smaller 4K pages in shadow PT.
+ */
+ return split_64KB_gtt_entry(vgpu, spt, index, &se);
+ case GTT_TYPE_PPGTT_PTE_2M_ENTRY:
+ gvt_vdbg_mm("shadow 2M gtt entry\n");
+ ret = is_2MB_gtt_possible(vgpu, ge);
+ if (ret == 0)
+ return split_2MB_gtt_entry(vgpu, spt, index, &se);
+ else if (ret < 0)
+ return ret;
+ page_size = I915_GTT_PAGE_SIZE_2M;
+ break;
+ case GTT_TYPE_PPGTT_PTE_1G_ENTRY:
+ gvt_vgpu_err("GVT doesn't support 1GB entry\n");
+ return -EINVAL;
+ default:
+ GEM_BUG_ON(1);
+ }
+
+ /* direct shadow */
+ ret = intel_gvt_dma_map_guest_page(vgpu, gfn, page_size, &dma_addr);
+ if (ret)
+ return -ENXIO;
+
+ pte_ops->set_pfn(&se, dma_addr >> PAGE_SHIFT);
+ ppgtt_set_shadow_entry(spt, &se, index);
+ return 0;
+}
+
+static int ppgtt_populate_spt(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *s;
+ struct intel_gvt_gtt_entry se, ge;
+ unsigned long gfn, i;
+ int ret;
+
+ trace_spt_change(spt->vgpu->id, "born", spt,
+ spt->guest_page.gfn, spt->shadow_page.type);
+
+ for_each_present_guest_entry(spt, &ge, i) {
+ if (gtt_type_is_pt(get_next_pt_type(ge.type))) {
+ s = ppgtt_populate_spt_by_guest_entry(vgpu, &ge);
+ if (IS_ERR(s)) {
+ ret = PTR_ERR(s);
+ goto fail;
+ }
+ ppgtt_get_shadow_entry(spt, &se, i);
+ ppgtt_generate_shadow_entry(&se, s, &ge);
+ ppgtt_set_shadow_entry(spt, &se, i);
+ } else {
+ gfn = ops->get_pfn(&ge);
+ if (!intel_gvt_is_valid_gfn(vgpu, gfn)) {
+ ops->set_pfn(&se, gvt->gtt.scratch_mfn);
+ ppgtt_set_shadow_entry(spt, &se, i);
+ continue;
+ }
+
+ ret = ppgtt_populate_shadow_entry(vgpu, spt, i, &ge);
+ if (ret)
+ goto fail;
+ }
+ }
+ return 0;
+fail:
+ gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d\n",
+ spt, ge.val64, ge.type);
+ return ret;
+}
+
+static int ppgtt_handle_guest_entry_removal(struct intel_vgpu_ppgtt_spt *spt,
+ struct intel_gvt_gtt_entry *se, unsigned long index)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ int ret;
+
+ trace_spt_guest_change(spt->vgpu->id, "remove", spt,
+ spt->shadow_page.type, se->val64, index);
+
+ gvt_vdbg_mm("destroy old shadow entry, type %d, index %lu, value %llx\n",
+ se->type, index, se->val64);
+
+ if (!ops->test_present(se))
+ return 0;
+
+ if (ops->get_pfn(se) ==
+ vgpu->gtt.scratch_pt[spt->shadow_page.type].page_mfn)
+ return 0;
+
+ if (gtt_type_is_pt(get_next_pt_type(se->type))) {
+ struct intel_vgpu_ppgtt_spt *s =
+ intel_vgpu_find_spt_by_mfn(vgpu, ops->get_pfn(se));
+ if (!s) {
+ gvt_vgpu_err("fail to find guest page\n");
+ ret = -ENXIO;
+ goto fail;
+ }
+ ret = ppgtt_invalidate_spt(s);
+ if (ret)
+ goto fail;
+ } else {
+ /* We don't setup 64K shadow entry so far. */
+ WARN(se->type == GTT_TYPE_PPGTT_PTE_64K_ENTRY,
+ "suspicious 64K entry\n");
+ ppgtt_invalidate_pte(spt, se);
+ }
+
+ return 0;
+fail:
+ gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d\n",
+ spt, se->val64, se->type);
+ return ret;
+}
+
+static int ppgtt_handle_guest_entry_add(struct intel_vgpu_ppgtt_spt *spt,
+ struct intel_gvt_gtt_entry *we, unsigned long index)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ struct intel_gvt_gtt_entry m;
+ struct intel_vgpu_ppgtt_spt *s;
+ int ret;
+
+ trace_spt_guest_change(spt->vgpu->id, "add", spt, spt->shadow_page.type,
+ we->val64, index);
+
+ gvt_vdbg_mm("add shadow entry: type %d, index %lu, value %llx\n",
+ we->type, index, we->val64);
+
+ if (gtt_type_is_pt(get_next_pt_type(we->type))) {
+ s = ppgtt_populate_spt_by_guest_entry(vgpu, we);
+ if (IS_ERR(s)) {
+ ret = PTR_ERR(s);
+ goto fail;
+ }
+ ppgtt_get_shadow_entry(spt, &m, index);
+ ppgtt_generate_shadow_entry(&m, s, we);
+ ppgtt_set_shadow_entry(spt, &m, index);
+ } else {
+ ret = ppgtt_populate_shadow_entry(vgpu, spt, index, we);
+ if (ret)
+ goto fail;
+ }
+ return 0;
+fail:
+ gvt_vgpu_err("fail: spt %p guest entry 0x%llx type %d\n",
+ spt, we->val64, we->type);
+ return ret;
+}
+
+static int sync_oos_page(struct intel_vgpu *vgpu,
+ struct intel_vgpu_oos_page *oos_page)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *spt = oos_page->spt;
+ struct intel_gvt_gtt_entry old, new;
+ int index;
+ int ret;
+
+ trace_oos_change(vgpu->id, "sync", oos_page->id,
+ spt, spt->guest_page.type);
+
+ old.type = new.type = get_entry_type(spt->guest_page.type);
+ old.val64 = new.val64 = 0;
+
+ for (index = 0; index < (I915_GTT_PAGE_SIZE >>
+ info->gtt_entry_size_shift); index++) {
+ ops->get_entry(oos_page->mem, &old, index, false, 0, vgpu);
+ ops->get_entry(NULL, &new, index, true,
+ spt->guest_page.gfn << PAGE_SHIFT, vgpu);
+
+ if (old.val64 == new.val64
+ && !test_and_clear_bit(index, spt->post_shadow_bitmap))
+ continue;
+
+ trace_oos_sync(vgpu->id, oos_page->id,
+ spt, spt->guest_page.type,
+ new.val64, index);
+
+ ret = ppgtt_populate_shadow_entry(vgpu, spt, index, &new);
+ if (ret)
+ return ret;
+
+ ops->set_entry(oos_page->mem, &new, index, false, 0, vgpu);
+ }
+
+ spt->guest_page.write_cnt = 0;
+ list_del_init(&spt->post_shadow_list);
+ return 0;
+}
+
+static int detach_oos_page(struct intel_vgpu *vgpu,
+ struct intel_vgpu_oos_page *oos_page)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_vgpu_ppgtt_spt *spt = oos_page->spt;
+
+ trace_oos_change(vgpu->id, "detach", oos_page->id,
+ spt, spt->guest_page.type);
+
+ spt->guest_page.write_cnt = 0;
+ spt->guest_page.oos_page = NULL;
+ oos_page->spt = NULL;
+
+ list_del_init(&oos_page->vm_list);
+ list_move_tail(&oos_page->list, &gvt->gtt.oos_page_free_list_head);
+
+ return 0;
+}
+
+static int attach_oos_page(struct intel_vgpu_oos_page *oos_page,
+ struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_gvt *gvt = spt->vgpu->gvt;
+ int ret;
+
+ ret = intel_gvt_read_gpa(spt->vgpu,
+ spt->guest_page.gfn << I915_GTT_PAGE_SHIFT,
+ oos_page->mem, I915_GTT_PAGE_SIZE);
+ if (ret)
+ return ret;
+
+ oos_page->spt = spt;
+ spt->guest_page.oos_page = oos_page;
+
+ list_move_tail(&oos_page->list, &gvt->gtt.oos_page_use_list_head);
+
+ trace_oos_change(spt->vgpu->id, "attach", oos_page->id,
+ spt, spt->guest_page.type);
+ return 0;
+}
+
+static int ppgtt_set_guest_page_sync(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_vgpu_oos_page *oos_page = spt->guest_page.oos_page;
+ int ret;
+
+ ret = intel_vgpu_enable_page_track(spt->vgpu, spt->guest_page.gfn);
+ if (ret)
+ return ret;
+
+ trace_oos_change(spt->vgpu->id, "set page sync", oos_page->id,
+ spt, spt->guest_page.type);
+
+ list_del_init(&oos_page->vm_list);
+ return sync_oos_page(spt->vgpu, oos_page);
+}
+
+static int ppgtt_allocate_oos_page(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_gvt *gvt = spt->vgpu->gvt;
+ struct intel_gvt_gtt *gtt = &gvt->gtt;
+ struct intel_vgpu_oos_page *oos_page = spt->guest_page.oos_page;
+ int ret;
+
+ WARN(oos_page, "shadow PPGTT page has already has a oos page\n");
+
+ if (list_empty(&gtt->oos_page_free_list_head)) {
+ oos_page = container_of(gtt->oos_page_use_list_head.next,
+ struct intel_vgpu_oos_page, list);
+ ret = ppgtt_set_guest_page_sync(oos_page->spt);
+ if (ret)
+ return ret;
+ ret = detach_oos_page(spt->vgpu, oos_page);
+ if (ret)
+ return ret;
+ } else
+ oos_page = container_of(gtt->oos_page_free_list_head.next,
+ struct intel_vgpu_oos_page, list);
+ return attach_oos_page(oos_page, spt);
+}
+
+static int ppgtt_set_guest_page_oos(struct intel_vgpu_ppgtt_spt *spt)
+{
+ struct intel_vgpu_oos_page *oos_page = spt->guest_page.oos_page;
+
+ if (WARN(!oos_page, "shadow PPGTT page should have a oos page\n"))
+ return -EINVAL;
+
+ trace_oos_change(spt->vgpu->id, "set page out of sync", oos_page->id,
+ spt, spt->guest_page.type);
+
+ list_add_tail(&oos_page->vm_list, &spt->vgpu->gtt.oos_page_list_head);
+ return intel_vgpu_disable_page_track(spt->vgpu, spt->guest_page.gfn);
+}
+
+/**
+ * intel_vgpu_sync_oos_pages - sync all the out-of-synced shadow for vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is called before submitting a guest workload to host,
+ * to sync all the out-of-synced shadow for vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu)
+{
+ struct list_head *pos, *n;
+ struct intel_vgpu_oos_page *oos_page;
+ int ret;
+
+ if (!enable_out_of_sync)
+ return 0;
+
+ list_for_each_safe(pos, n, &vgpu->gtt.oos_page_list_head) {
+ oos_page = container_of(pos,
+ struct intel_vgpu_oos_page, vm_list);
+ ret = ppgtt_set_guest_page_sync(oos_page->spt);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+/*
+ * The heart of PPGTT shadow page table.
+ */
+static int ppgtt_handle_guest_write_page_table(
+ struct intel_vgpu_ppgtt_spt *spt,
+ struct intel_gvt_gtt_entry *we, unsigned long index)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ int type = spt->shadow_page.type;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_gvt_gtt_entry old_se;
+ int new_present;
+ int i, ret;
+
+ new_present = ops->test_present(we);
+
+ /*
+ * Adding the new entry first and then removing the old one, that can
+ * guarantee the ppgtt table is validated during the window between
+ * adding and removal.
+ */
+ ppgtt_get_shadow_entry(spt, &old_se, index);
+
+ if (new_present) {
+ ret = ppgtt_handle_guest_entry_add(spt, we, index);
+ if (ret)
+ goto fail;
+ }
+
+ ret = ppgtt_handle_guest_entry_removal(spt, &old_se, index);
+ if (ret)
+ goto fail;
+
+ if (!new_present) {
+ /* For 64KB splited entries, we need clear them all. */
+ if (ops->test_64k_splited(&old_se) &&
+ !(index % GTT_64K_PTE_STRIDE)) {
+ gvt_vdbg_mm("remove splited 64K shadow entries\n");
+ for (i = 0; i < GTT_64K_PTE_STRIDE; i++) {
+ ops->clear_64k_splited(&old_se);
+ ops->set_pfn(&old_se,
+ vgpu->gtt.scratch_pt[type].page_mfn);
+ ppgtt_set_shadow_entry(spt, &old_se, index + i);
+ }
+ } else if (old_se.type == GTT_TYPE_PPGTT_PTE_2M_ENTRY ||
+ old_se.type == GTT_TYPE_PPGTT_PTE_1G_ENTRY) {
+ ops->clear_pse(&old_se);
+ ops->set_pfn(&old_se,
+ vgpu->gtt.scratch_pt[type].page_mfn);
+ ppgtt_set_shadow_entry(spt, &old_se, index);
+ } else {
+ ops->set_pfn(&old_se,
+ vgpu->gtt.scratch_pt[type].page_mfn);
+ ppgtt_set_shadow_entry(spt, &old_se, index);
+ }
+ }
+
+ return 0;
+fail:
+ gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d.\n",
+ spt, we->val64, we->type);
+ return ret;
+}
+
+
+
+static inline bool can_do_out_of_sync(struct intel_vgpu_ppgtt_spt *spt)
+{
+ return enable_out_of_sync
+ && gtt_type_is_pte_pt(spt->guest_page.type)
+ && spt->guest_page.write_cnt >= 2;
+}
+
+static void ppgtt_set_post_shadow(struct intel_vgpu_ppgtt_spt *spt,
+ unsigned long index)
+{
+ set_bit(index, spt->post_shadow_bitmap);
+ if (!list_empty(&spt->post_shadow_list))
+ return;
+
+ list_add_tail(&spt->post_shadow_list,
+ &spt->vgpu->gtt.post_shadow_list_head);
+}
+
+/**
+ * intel_vgpu_flush_post_shadow - flush the post shadow transactions
+ * @vgpu: a vGPU
+ *
+ * This function is called before submitting a guest workload to host,
+ * to flush all the post shadows for a vGPU.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_flush_post_shadow(struct intel_vgpu *vgpu)
+{
+ struct list_head *pos, *n;
+ struct intel_vgpu_ppgtt_spt *spt;
+ struct intel_gvt_gtt_entry ge;
+ unsigned long index;
+ int ret;
+
+ list_for_each_safe(pos, n, &vgpu->gtt.post_shadow_list_head) {
+ spt = container_of(pos, struct intel_vgpu_ppgtt_spt,
+ post_shadow_list);
+
+ for_each_set_bit(index, spt->post_shadow_bitmap,
+ GTT_ENTRY_NUM_IN_ONE_PAGE) {
+ ppgtt_get_guest_entry(spt, &ge, index);
+
+ ret = ppgtt_handle_guest_write_page_table(spt,
+ &ge, index);
+ if (ret)
+ return ret;
+ clear_bit(index, spt->post_shadow_bitmap);
+ }
+ list_del_init(&spt->post_shadow_list);
+ }
+ return 0;
+}
+
+static int ppgtt_handle_guest_write_page_table_bytes(
+ struct intel_vgpu_ppgtt_spt *spt,
+ u64 pa, void *p_data, int bytes)
+{
+ struct intel_vgpu *vgpu = spt->vgpu;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ struct intel_gvt_gtt_entry we, se;
+ unsigned long index;
+ int ret;
+
+ index = (pa & (PAGE_SIZE - 1)) >> info->gtt_entry_size_shift;
+
+ ppgtt_get_guest_entry(spt, &we, index);
+
+ /*
+ * For page table which has 64K gtt entry, only PTE#0, PTE#16,
+ * PTE#32, ... PTE#496 are used. Unused PTEs update should be
+ * ignored.
+ */
+ if (we.type == GTT_TYPE_PPGTT_PTE_64K_ENTRY &&
+ (index % GTT_64K_PTE_STRIDE)) {
+ gvt_vdbg_mm("Ignore write to unused PTE entry, index %lu\n",
+ index);
+ return 0;
+ }
+
+ if (bytes == info->gtt_entry_size) {
+ ret = ppgtt_handle_guest_write_page_table(spt, &we, index);
+ if (ret)
+ return ret;
+ } else {
+ if (!test_bit(index, spt->post_shadow_bitmap)) {
+ int type = spt->shadow_page.type;
+
+ ppgtt_get_shadow_entry(spt, &se, index);
+ ret = ppgtt_handle_guest_entry_removal(spt, &se, index);
+ if (ret)
+ return ret;
+ ops->set_pfn(&se, vgpu->gtt.scratch_pt[type].page_mfn);
+ ppgtt_set_shadow_entry(spt, &se, index);
+ }
+ ppgtt_set_post_shadow(spt, index);
+ }
+
+ if (!enable_out_of_sync)
+ return 0;
+
+ spt->guest_page.write_cnt++;
+
+ if (spt->guest_page.oos_page)
+ ops->set_entry(spt->guest_page.oos_page->mem, &we, index,
+ false, 0, vgpu);
+
+ if (can_do_out_of_sync(spt)) {
+ if (!spt->guest_page.oos_page)
+ ppgtt_allocate_oos_page(spt);
+
+ ret = ppgtt_set_guest_page_oos(spt);
+ if (ret < 0)
+ return ret;
+ }
+ return 0;
+}
+
+static void invalidate_ppgtt_mm(struct intel_vgpu_mm *mm)
+{
+ struct intel_vgpu *vgpu = mm->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_gtt *gtt = &gvt->gtt;
+ const struct intel_gvt_gtt_pte_ops *ops = gtt->pte_ops;
+ struct intel_gvt_gtt_entry se;
+ int index;
+
+ if (!mm->ppgtt_mm.shadowed)
+ return;
+
+ for (index = 0; index < ARRAY_SIZE(mm->ppgtt_mm.shadow_pdps); index++) {
+ ppgtt_get_shadow_root_entry(mm, &se, index);
+
+ if (!ops->test_present(&se))
+ continue;
+
+ ppgtt_invalidate_spt_by_shadow_entry(vgpu, &se);
+ se.val64 = 0;
+ ppgtt_set_shadow_root_entry(mm, &se, index);
+
+ trace_spt_guest_change(vgpu->id, "destroy root pointer",
+ NULL, se.type, se.val64, index);
+ }
+
+ mm->ppgtt_mm.shadowed = false;
+}
+
+
+static int shadow_ppgtt_mm(struct intel_vgpu_mm *mm)
+{
+ struct intel_vgpu *vgpu = mm->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_gtt *gtt = &gvt->gtt;
+ const struct intel_gvt_gtt_pte_ops *ops = gtt->pte_ops;
+ struct intel_vgpu_ppgtt_spt *spt;
+ struct intel_gvt_gtt_entry ge, se;
+ int index, ret;
+
+ if (mm->ppgtt_mm.shadowed)
+ return 0;
+
+ mm->ppgtt_mm.shadowed = true;
+
+ for (index = 0; index < ARRAY_SIZE(mm->ppgtt_mm.guest_pdps); index++) {
+ ppgtt_get_guest_root_entry(mm, &ge, index);
+
+ if (!ops->test_present(&ge))
+ continue;
+
+ trace_spt_guest_change(vgpu->id, __func__, NULL,
+ ge.type, ge.val64, index);
+
+ spt = ppgtt_populate_spt_by_guest_entry(vgpu, &ge);
+ if (IS_ERR(spt)) {
+ gvt_vgpu_err("fail to populate guest root pointer\n");
+ ret = PTR_ERR(spt);
+ goto fail;
+ }
+ ppgtt_generate_shadow_entry(&se, spt, &ge);
+ ppgtt_set_shadow_root_entry(mm, &se, index);
+
+ trace_spt_guest_change(vgpu->id, "populate root pointer",
+ NULL, se.type, se.val64, index);
+ }
+
+ return 0;
+fail:
+ invalidate_ppgtt_mm(mm);
+ return ret;
+}
+
+static struct intel_vgpu_mm *vgpu_alloc_mm(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_mm *mm;
+
+ mm = kzalloc(sizeof(*mm), GFP_KERNEL);
+ if (!mm)
+ return NULL;
+
+ mm->vgpu = vgpu;
+ kref_init(&mm->ref);
+ atomic_set(&mm->pincount, 0);
+
+ return mm;
+}
+
+static void vgpu_free_mm(struct intel_vgpu_mm *mm)
+{
+ kfree(mm);
+}
+
+/**
+ * intel_vgpu_create_ppgtt_mm - create a ppgtt mm object for a vGPU
+ * @vgpu: a vGPU
+ * @root_entry_type: ppgtt root entry type
+ * @pdps: guest pdps.
+ *
+ * This function is used to create a ppgtt mm object for a vGPU.
+ *
+ * Returns:
+ * Zero on success, negative error code in pointer if failed.
+ */
+struct intel_vgpu_mm *intel_vgpu_create_ppgtt_mm(struct intel_vgpu *vgpu,
+ enum intel_gvt_gtt_type root_entry_type, u64 pdps[])
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_vgpu_mm *mm;
+ int ret;
+
+ mm = vgpu_alloc_mm(vgpu);
+ if (!mm)
+ return ERR_PTR(-ENOMEM);
+
+ mm->type = INTEL_GVT_MM_PPGTT;
+
+ GEM_BUG_ON(root_entry_type != GTT_TYPE_PPGTT_ROOT_L3_ENTRY &&
+ root_entry_type != GTT_TYPE_PPGTT_ROOT_L4_ENTRY);
+ mm->ppgtt_mm.root_entry_type = root_entry_type;
+
+ INIT_LIST_HEAD(&mm->ppgtt_mm.list);
+ INIT_LIST_HEAD(&mm->ppgtt_mm.lru_list);
+ INIT_LIST_HEAD(&mm->ppgtt_mm.link);
+
+ if (root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY)
+ mm->ppgtt_mm.guest_pdps[0] = pdps[0];
+ else
+ memcpy(mm->ppgtt_mm.guest_pdps, pdps,
+ sizeof(mm->ppgtt_mm.guest_pdps));
+
+ ret = shadow_ppgtt_mm(mm);
+ if (ret) {
+ gvt_vgpu_err("failed to shadow ppgtt mm\n");
+ vgpu_free_mm(mm);
+ return ERR_PTR(ret);
+ }
+
+ list_add_tail(&mm->ppgtt_mm.list, &vgpu->gtt.ppgtt_mm_list_head);
+
+ mutex_lock(&gvt->gtt.ppgtt_mm_lock);
+ list_add_tail(&mm->ppgtt_mm.lru_list, &gvt->gtt.ppgtt_mm_lru_list_head);
+ mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
+
+ return mm;
+}
+
+static struct intel_vgpu_mm *intel_vgpu_create_ggtt_mm(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_mm *mm;
+ unsigned long nr_entries;
+
+ mm = vgpu_alloc_mm(vgpu);
+ if (!mm)
+ return ERR_PTR(-ENOMEM);
+
+ mm->type = INTEL_GVT_MM_GGTT;
+
+ nr_entries = gvt_ggtt_gm_sz(vgpu->gvt) >> I915_GTT_PAGE_SHIFT;
+ mm->ggtt_mm.virtual_ggtt =
+ vzalloc(array_size(nr_entries,
+ vgpu->gvt->device_info.gtt_entry_size));
+ if (!mm->ggtt_mm.virtual_ggtt) {
+ vgpu_free_mm(mm);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ mm->ggtt_mm.host_ggtt_aperture = vzalloc((vgpu_aperture_sz(vgpu) >> PAGE_SHIFT) * sizeof(u64));
+ if (!mm->ggtt_mm.host_ggtt_aperture) {
+ vfree(mm->ggtt_mm.virtual_ggtt);
+ vgpu_free_mm(mm);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ mm->ggtt_mm.host_ggtt_hidden = vzalloc((vgpu_hidden_sz(vgpu) >> PAGE_SHIFT) * sizeof(u64));
+ if (!mm->ggtt_mm.host_ggtt_hidden) {
+ vfree(mm->ggtt_mm.host_ggtt_aperture);
+ vfree(mm->ggtt_mm.virtual_ggtt);
+ vgpu_free_mm(mm);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ return mm;
+}
+
+/**
+ * _intel_vgpu_mm_release - destroy a mm object
+ * @mm_ref: a kref object
+ *
+ * This function is used to destroy a mm object for vGPU
+ *
+ */
+void _intel_vgpu_mm_release(struct kref *mm_ref)
+{
+ struct intel_vgpu_mm *mm = container_of(mm_ref, typeof(*mm), ref);
+
+ if (GEM_WARN_ON(atomic_read(&mm->pincount)))
+ gvt_err("vgpu mm pin count bug detected\n");
+
+ if (mm->type == INTEL_GVT_MM_PPGTT) {
+ list_del(&mm->ppgtt_mm.list);
+
+ mutex_lock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
+ list_del(&mm->ppgtt_mm.lru_list);
+ mutex_unlock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
+
+ invalidate_ppgtt_mm(mm);
+ } else {
+ vfree(mm->ggtt_mm.virtual_ggtt);
+ vfree(mm->ggtt_mm.host_ggtt_aperture);
+ vfree(mm->ggtt_mm.host_ggtt_hidden);
+ }
+
+ vgpu_free_mm(mm);
+}
+
+/**
+ * intel_vgpu_unpin_mm - decrease the pin count of a vGPU mm object
+ * @mm: a vGPU mm object
+ *
+ * This function is called when user doesn't want to use a vGPU mm object
+ */
+void intel_vgpu_unpin_mm(struct intel_vgpu_mm *mm)
+{
+ atomic_dec_if_positive(&mm->pincount);
+}
+
+/**
+ * intel_vgpu_pin_mm - increase the pin count of a vGPU mm object
+ * @mm: target vgpu mm
+ *
+ * This function is called when user wants to use a vGPU mm object. If this
+ * mm object hasn't been shadowed yet, the shadow will be populated at this
+ * time.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_pin_mm(struct intel_vgpu_mm *mm)
+{
+ int ret;
+
+ atomic_inc(&mm->pincount);
+
+ if (mm->type == INTEL_GVT_MM_PPGTT) {
+ ret = shadow_ppgtt_mm(mm);
+ if (ret)
+ return ret;
+
+ mutex_lock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
+ list_move_tail(&mm->ppgtt_mm.lru_list,
+ &mm->vgpu->gvt->gtt.ppgtt_mm_lru_list_head);
+ mutex_unlock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
+ }
+
+ return 0;
+}
+
+static int reclaim_one_ppgtt_mm(struct intel_gvt *gvt)
+{
+ struct intel_vgpu_mm *mm;
+ struct list_head *pos, *n;
+
+ mutex_lock(&gvt->gtt.ppgtt_mm_lock);
+
+ list_for_each_safe(pos, n, &gvt->gtt.ppgtt_mm_lru_list_head) {
+ mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.lru_list);
+
+ if (atomic_read(&mm->pincount))
+ continue;
+
+ list_del_init(&mm->ppgtt_mm.lru_list);
+ mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
+ invalidate_ppgtt_mm(mm);
+ return 1;
+ }
+ mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
+ return 0;
+}
+
+/*
+ * GMA translation APIs.
+ */
+static inline int ppgtt_get_next_level_entry(struct intel_vgpu_mm *mm,
+ struct intel_gvt_gtt_entry *e, unsigned long index, bool guest)
+{
+ struct intel_vgpu *vgpu = mm->vgpu;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_vgpu_ppgtt_spt *s;
+
+ s = intel_vgpu_find_spt_by_mfn(vgpu, ops->get_pfn(e));
+ if (!s)
+ return -ENXIO;
+
+ if (!guest)
+ ppgtt_get_shadow_entry(s, e, index);
+ else
+ ppgtt_get_guest_entry(s, e, index);
+ return 0;
+}
+
+/**
+ * intel_vgpu_gma_to_gpa - translate a gma to GPA
+ * @mm: mm object. could be a PPGTT or GGTT mm object
+ * @gma: graphics memory address in this mm object
+ *
+ * This function is used to translate a graphics memory address in specific
+ * graphics memory space to guest physical address.
+ *
+ * Returns:
+ * Guest physical address on success, INTEL_GVT_INVALID_ADDR if failed.
+ */
+unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
+{
+ struct intel_vgpu *vgpu = mm->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *pte_ops = gvt->gtt.pte_ops;
+ const struct intel_gvt_gtt_gma_ops *gma_ops = gvt->gtt.gma_ops;
+ unsigned long gpa = INTEL_GVT_INVALID_ADDR;
+ unsigned long gma_index[4];
+ struct intel_gvt_gtt_entry e;
+ int i, levels = 0;
+ int ret;
+
+ GEM_BUG_ON(mm->type != INTEL_GVT_MM_GGTT &&
+ mm->type != INTEL_GVT_MM_PPGTT);
+
+ if (mm->type == INTEL_GVT_MM_GGTT) {
+ if (!vgpu_gmadr_is_valid(vgpu, gma))
+ goto err;
+
+ ggtt_get_guest_entry(mm, &e,
+ gma_ops->gma_to_ggtt_pte_index(gma));
+
+ gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT)
+ + (gma & ~I915_GTT_PAGE_MASK);
+
+ trace_gma_translate(vgpu->id, "ggtt", 0, 0, gma, gpa);
+ } else {
+ switch (mm->ppgtt_mm.root_entry_type) {
+ case GTT_TYPE_PPGTT_ROOT_L4_ENTRY:
+ ppgtt_get_shadow_root_entry(mm, &e, 0);
+
+ gma_index[0] = gma_ops->gma_to_pml4_index(gma);
+ gma_index[1] = gma_ops->gma_to_l4_pdp_index(gma);
+ gma_index[2] = gma_ops->gma_to_pde_index(gma);
+ gma_index[3] = gma_ops->gma_to_pte_index(gma);
+ levels = 4;
+ break;
+ case GTT_TYPE_PPGTT_ROOT_L3_ENTRY:
+ ppgtt_get_shadow_root_entry(mm, &e,
+ gma_ops->gma_to_l3_pdp_index(gma));
+
+ gma_index[0] = gma_ops->gma_to_pde_index(gma);
+ gma_index[1] = gma_ops->gma_to_pte_index(gma);
+ levels = 2;
+ break;
+ default:
+ GEM_BUG_ON(1);
+ }
+
+ /* walk the shadow page table and get gpa from guest entry */
+ for (i = 0; i < levels; i++) {
+ ret = ppgtt_get_next_level_entry(mm, &e, gma_index[i],
+ (i == levels - 1));
+ if (ret)
+ goto err;
+
+ if (!pte_ops->test_present(&e)) {
+ gvt_dbg_core("GMA 0x%lx is not present\n", gma);
+ goto err;
+ }
+ }
+
+ gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT) +
+ (gma & ~I915_GTT_PAGE_MASK);
+ trace_gma_translate(vgpu->id, "ppgtt", 0,
+ mm->ppgtt_mm.root_entry_type, gma, gpa);
+ }
+
+ return gpa;
+err:
+ gvt_vgpu_err("invalid mm type: %d gma %lx\n", mm->type, gma);
+ return INTEL_GVT_INVALID_ADDR;
+}
+
+static int emulate_ggtt_mmio_read(struct intel_vgpu *vgpu,
+ unsigned int off, void *p_data, unsigned int bytes)
+{
+ struct intel_vgpu_mm *ggtt_mm = vgpu->gtt.ggtt_mm;
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ unsigned long index = off >> info->gtt_entry_size_shift;
+ unsigned long gma;
+ struct intel_gvt_gtt_entry e;
+
+ if (bytes != 4 && bytes != 8)
+ return -EINVAL;
+
+ gma = index << I915_GTT_PAGE_SHIFT;
+ if (!intel_gvt_ggtt_validate_range(vgpu,
+ gma, 1 << I915_GTT_PAGE_SHIFT)) {
+ gvt_dbg_mm("read invalid ggtt at 0x%lx\n", gma);
+ memset(p_data, 0, bytes);
+ return 0;
+ }
+
+ ggtt_get_guest_entry(ggtt_mm, &e, index);
+ memcpy(p_data, (void *)&e.val64 + (off & (info->gtt_entry_size - 1)),
+ bytes);
+ return 0;
+}
+
+/**
+ * intel_vgpu_emulate_ggtt_mmio_read - emulate GTT MMIO register read
+ * @vgpu: a vGPU
+ * @off: register offset
+ * @p_data: data will be returned to guest
+ * @bytes: data length
+ *
+ * This function is used to emulate the GTT MMIO register read
+ *
+ * Returns:
+ * Zero on success, error code if failed.
+ */
+int intel_vgpu_emulate_ggtt_mmio_read(struct intel_vgpu *vgpu, unsigned int off,
+ void *p_data, unsigned int bytes)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ int ret;
+
+ if (bytes != 4 && bytes != 8)
+ return -EINVAL;
+
+ off -= info->gtt_start_offset;
+ ret = emulate_ggtt_mmio_read(vgpu, off, p_data, bytes);
+ return ret;
+}
+
+static void ggtt_invalidate_pte(struct intel_vgpu *vgpu,
+ struct intel_gvt_gtt_entry *entry)
+{
+ const struct intel_gvt_gtt_pte_ops *pte_ops = vgpu->gvt->gtt.pte_ops;
+ unsigned long pfn;
+
+ pfn = pte_ops->get_pfn(entry);
+ if (pfn != vgpu->gvt->gtt.scratch_mfn)
+ intel_gvt_dma_unmap_guest_page(vgpu, pfn << PAGE_SHIFT);
+}
+
+static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
+ void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_device_info *info = &gvt->device_info;
+ struct intel_vgpu_mm *ggtt_mm = vgpu->gtt.ggtt_mm;
+ const struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+ unsigned long g_gtt_index = off >> info->gtt_entry_size_shift;
+ unsigned long gma, gfn;
+ struct intel_gvt_gtt_entry e = {.val64 = 0, .type = GTT_TYPE_GGTT_PTE};
+ struct intel_gvt_gtt_entry m = {.val64 = 0, .type = GTT_TYPE_GGTT_PTE};
+ dma_addr_t dma_addr;
+ int ret;
+ struct intel_gvt_partial_pte *partial_pte, *pos, *n;
+ bool partial_update = false;
+
+ if (bytes != 4 && bytes != 8)
+ return -EINVAL;
+
+ gma = g_gtt_index << I915_GTT_PAGE_SHIFT;
+
+ /* the VM may configure the whole GM space when ballooning is used */
+ if (!vgpu_gmadr_is_valid(vgpu, gma))
+ return 0;
+
+ e.type = GTT_TYPE_GGTT_PTE;
+ memcpy((void *)&e.val64 + (off & (info->gtt_entry_size - 1)), p_data,
+ bytes);
+
+ /* If ggtt entry size is 8 bytes, and it's split into two 4 bytes
+ * write, save the first 4 bytes in a list and update virtual
+ * PTE. Only update shadow PTE when the second 4 bytes comes.
+ */
+ if (bytes < info->gtt_entry_size) {
+ bool found = false;
+
+ list_for_each_entry_safe(pos, n,
+ &ggtt_mm->ggtt_mm.partial_pte_list, list) {
+ if (g_gtt_index == pos->offset >>
+ info->gtt_entry_size_shift) {
+ if (off != pos->offset) {
+ /* the second partial part*/
+ int last_off = pos->offset &
+ (info->gtt_entry_size - 1);
+
+ memcpy((void *)&e.val64 + last_off,
+ (void *)&pos->data + last_off,
+ bytes);
+
+ list_del(&pos->list);
+ kfree(pos);
+ found = true;
+ break;
+ }
+
+ /* update of the first partial part */
+ pos->data = e.val64;
+ ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index);
+ return 0;
+ }
+ }
+
+ if (!found) {
+ /* the first partial part */
+ partial_pte = kzalloc(sizeof(*partial_pte), GFP_KERNEL);
+ if (!partial_pte)
+ return -ENOMEM;
+ partial_pte->offset = off;
+ partial_pte->data = e.val64;
+ list_add_tail(&partial_pte->list,
+ &ggtt_mm->ggtt_mm.partial_pte_list);
+ partial_update = true;
+ }
+ }
+
+ if (!partial_update && (ops->test_present(&e))) {
+ gfn = ops->get_pfn(&e);
+ m.val64 = e.val64;
+ m.type = e.type;
+
+ /* one PTE update may be issued in multiple writes and the
+ * first write may not construct a valid gfn
+ */
+ if (!intel_gvt_is_valid_gfn(vgpu, gfn)) {
+ ops->set_pfn(&m, gvt->gtt.scratch_mfn);
+ goto out;
+ }
+
+ ret = intel_gvt_dma_map_guest_page(vgpu, gfn, PAGE_SIZE,
+ &dma_addr);
+ if (ret) {
+ gvt_vgpu_err("fail to populate guest ggtt entry\n");
+ /* guest driver may read/write the entry when partial
+ * update the entry in this situation p2m will fail
+ * setting the shadow entry to point to a scratch page
+ */
+ ops->set_pfn(&m, gvt->gtt.scratch_mfn);
+ } else
+ ops->set_pfn(&m, dma_addr >> PAGE_SHIFT);
+ } else {
+ ops->set_pfn(&m, gvt->gtt.scratch_mfn);
+ ops->clear_present(&m);
+ }
+
+out:
+ ggtt_set_guest_entry(ggtt_mm, &e, g_gtt_index);
+
+ ggtt_get_host_entry(ggtt_mm, &e, g_gtt_index);
+ ggtt_invalidate_pte(vgpu, &e);
+
+ ggtt_set_host_entry(ggtt_mm, &m, g_gtt_index);
+ ggtt_invalidate(gvt->gt);
+ return 0;
+}
+
+/*
+ * intel_vgpu_emulate_ggtt_mmio_write - emulate GTT MMIO register write
+ * @vgpu: a vGPU
+ * @off: register offset
+ * @p_data: data from guest write
+ * @bytes: data length
+ *
+ * This function is used to emulate the GTT MMIO register write
+ *
+ * Returns:
+ * Zero on success, error code if failed.
+ */
+int intel_vgpu_emulate_ggtt_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int off, void *p_data, unsigned int bytes)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+ int ret;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_engine_cs *engine;
+ int i;
+
+ if (bytes != 4 && bytes != 8)
+ return -EINVAL;
+
+ off -= info->gtt_start_offset;
+ ret = emulate_ggtt_mmio_write(vgpu, off, p_data, bytes);
+
+ /* if ggtt of last submitted context is written,
+ * that context is probably got unpinned.
+ * Set last shadowed ctx to invalid.
+ */
+ for_each_engine(engine, vgpu->gvt->gt, i) {
+ if (!s->last_ctx[i].valid)
+ continue;
+
+ if (s->last_ctx[i].lrca == (off >> info->gtt_entry_size_shift))
+ s->last_ctx[i].valid = false;
+ }
+ return ret;
+}
+
+static int alloc_scratch_pages(struct intel_vgpu *vgpu,
+ enum intel_gvt_gtt_type type)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_vgpu_gtt *gtt = &vgpu->gtt;
+ const struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+ int page_entry_num = I915_GTT_PAGE_SIZE >>
+ vgpu->gvt->device_info.gtt_entry_size_shift;
+ void *scratch_pt;
+ int i;
+ struct device *dev = vgpu->gvt->gt->i915->drm.dev;
+ dma_addr_t daddr;
+
+ if (drm_WARN_ON(&i915->drm,
+ type < GTT_TYPE_PPGTT_PTE_PT || type >= GTT_TYPE_MAX))
+ return -EINVAL;
+
+ scratch_pt = (void *)get_zeroed_page(GFP_KERNEL);
+ if (!scratch_pt) {
+ gvt_vgpu_err("fail to allocate scratch page\n");
+ return -ENOMEM;
+ }
+
+ daddr = dma_map_page(dev, virt_to_page(scratch_pt), 0, 4096, DMA_BIDIRECTIONAL);
+ if (dma_mapping_error(dev, daddr)) {
+ gvt_vgpu_err("fail to dmamap scratch_pt\n");
+ __free_page(virt_to_page(scratch_pt));
+ return -ENOMEM;
+ }
+ gtt->scratch_pt[type].page_mfn =
+ (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT);
+ gtt->scratch_pt[type].page = virt_to_page(scratch_pt);
+ gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n",
+ vgpu->id, type, gtt->scratch_pt[type].page_mfn);
+
+ /* Build the tree by full filled the scratch pt with the entries which
+ * point to the next level scratch pt or scratch page. The
+ * scratch_pt[type] indicate the scratch pt/scratch page used by the
+ * 'type' pt.
+ * e.g. scratch_pt[GTT_TYPE_PPGTT_PDE_PT] is used by
+ * GTT_TYPE_PPGTT_PDE_PT level pt, that means this scratch_pt it self
+ * is GTT_TYPE_PPGTT_PTE_PT, and full filled by scratch page mfn.
+ */
+ if (type > GTT_TYPE_PPGTT_PTE_PT) {
+ struct intel_gvt_gtt_entry se;
+
+ memset(&se, 0, sizeof(struct intel_gvt_gtt_entry));
+ se.type = get_entry_type(type - 1);
+ ops->set_pfn(&se, gtt->scratch_pt[type - 1].page_mfn);
+
+ /* The entry parameters like present/writeable/cache type
+ * set to the same as i915's scratch page tree.
+ */
+ se.val64 |= GEN8_PAGE_PRESENT | GEN8_PAGE_RW;
+ if (type == GTT_TYPE_PPGTT_PDE_PT)
+ se.val64 |= PPAT_CACHED;
+
+ for (i = 0; i < page_entry_num; i++)
+ ops->set_entry(scratch_pt, &se, i, false, 0, vgpu);
+ }
+
+ return 0;
+}
+
+static int release_scratch_page_tree(struct intel_vgpu *vgpu)
+{
+ int i;
+ struct device *dev = vgpu->gvt->gt->i915->drm.dev;
+ dma_addr_t daddr;
+
+ for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) {
+ if (vgpu->gtt.scratch_pt[i].page != NULL) {
+ daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn <<
+ I915_GTT_PAGE_SHIFT);
+ dma_unmap_page(dev, daddr, 4096, DMA_BIDIRECTIONAL);
+ __free_page(vgpu->gtt.scratch_pt[i].page);
+ vgpu->gtt.scratch_pt[i].page = NULL;
+ vgpu->gtt.scratch_pt[i].page_mfn = 0;
+ }
+ }
+
+ return 0;
+}
+
+static int create_scratch_page_tree(struct intel_vgpu *vgpu)
+{
+ int i, ret;
+
+ for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) {
+ ret = alloc_scratch_pages(vgpu, i);
+ if (ret)
+ goto err;
+ }
+
+ return 0;
+
+err:
+ release_scratch_page_tree(vgpu);
+ return ret;
+}
+
+/**
+ * intel_vgpu_init_gtt - initialize per-vGPU graphics memory virulization
+ * @vgpu: a vGPU
+ *
+ * This function is used to initialize per-vGPU graphics memory virtualization
+ * components.
+ *
+ * Returns:
+ * Zero on success, error code if failed.
+ */
+int intel_vgpu_init_gtt(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_gtt *gtt = &vgpu->gtt;
+
+ INIT_RADIX_TREE(&gtt->spt_tree, GFP_KERNEL);
+
+ INIT_LIST_HEAD(&gtt->ppgtt_mm_list_head);
+ INIT_LIST_HEAD(&gtt->oos_page_list_head);
+ INIT_LIST_HEAD(&gtt->post_shadow_list_head);
+
+ gtt->ggtt_mm = intel_vgpu_create_ggtt_mm(vgpu);
+ if (IS_ERR(gtt->ggtt_mm)) {
+ gvt_vgpu_err("fail to create mm for ggtt.\n");
+ return PTR_ERR(gtt->ggtt_mm);
+ }
+
+ intel_vgpu_reset_ggtt(vgpu, false);
+
+ INIT_LIST_HEAD(&gtt->ggtt_mm->ggtt_mm.partial_pte_list);
+
+ return create_scratch_page_tree(vgpu);
+}
+
+void intel_vgpu_destroy_all_ppgtt_mm(struct intel_vgpu *vgpu)
+{
+ struct list_head *pos, *n;
+ struct intel_vgpu_mm *mm;
+
+ list_for_each_safe(pos, n, &vgpu->gtt.ppgtt_mm_list_head) {
+ mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.list);
+ intel_vgpu_destroy_mm(mm);
+ }
+
+ if (GEM_WARN_ON(!list_empty(&vgpu->gtt.ppgtt_mm_list_head)))
+ gvt_err("vgpu ppgtt mm is not fully destroyed\n");
+
+ if (GEM_WARN_ON(!radix_tree_empty(&vgpu->gtt.spt_tree))) {
+ gvt_err("Why we still has spt not freed?\n");
+ ppgtt_free_all_spt(vgpu);
+ }
+}
+
+static void intel_vgpu_destroy_ggtt_mm(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt_partial_pte *pos, *next;
+
+ list_for_each_entry_safe(pos, next,
+ &vgpu->gtt.ggtt_mm->ggtt_mm.partial_pte_list,
+ list) {
+ gvt_dbg_mm("partial PTE update on hold 0x%lx : 0x%llx\n",
+ pos->offset, pos->data);
+ kfree(pos);
+ }
+ intel_vgpu_destroy_mm(vgpu->gtt.ggtt_mm);
+ vgpu->gtt.ggtt_mm = NULL;
+}
+
+/**
+ * intel_vgpu_clean_gtt - clean up per-vGPU graphics memory virulization
+ * @vgpu: a vGPU
+ *
+ * This function is used to clean up per-vGPU graphics memory virtualization
+ * components.
+ *
+ * Returns:
+ * Zero on success, error code if failed.
+ */
+void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu)
+{
+ intel_vgpu_destroy_all_ppgtt_mm(vgpu);
+ intel_vgpu_destroy_ggtt_mm(vgpu);
+ release_scratch_page_tree(vgpu);
+}
+
+static void clean_spt_oos(struct intel_gvt *gvt)
+{
+ struct intel_gvt_gtt *gtt = &gvt->gtt;
+ struct list_head *pos, *n;
+ struct intel_vgpu_oos_page *oos_page;
+
+ WARN(!list_empty(&gtt->oos_page_use_list_head),
+ "someone is still using oos page\n");
+
+ list_for_each_safe(pos, n, &gtt->oos_page_free_list_head) {
+ oos_page = container_of(pos, struct intel_vgpu_oos_page, list);
+ list_del(&oos_page->list);
+ free_page((unsigned long)oos_page->mem);
+ kfree(oos_page);
+ }
+}
+
+static int setup_spt_oos(struct intel_gvt *gvt)
+{
+ struct intel_gvt_gtt *gtt = &gvt->gtt;
+ struct intel_vgpu_oos_page *oos_page;
+ int i;
+ int ret;
+
+ INIT_LIST_HEAD(&gtt->oos_page_free_list_head);
+ INIT_LIST_HEAD(&gtt->oos_page_use_list_head);
+
+ for (i = 0; i < preallocated_oos_pages; i++) {
+ oos_page = kzalloc(sizeof(*oos_page), GFP_KERNEL);
+ if (!oos_page) {
+ ret = -ENOMEM;
+ goto fail;
+ }
+ oos_page->mem = (void *)__get_free_pages(GFP_KERNEL, 0);
+ if (!oos_page->mem) {
+ ret = -ENOMEM;
+ kfree(oos_page);
+ goto fail;
+ }
+
+ INIT_LIST_HEAD(&oos_page->list);
+ INIT_LIST_HEAD(&oos_page->vm_list);
+ oos_page->id = i;
+ list_add_tail(&oos_page->list, &gtt->oos_page_free_list_head);
+ }
+
+ gvt_dbg_mm("%d oos pages preallocated\n", i);
+
+ return 0;
+fail:
+ clean_spt_oos(gvt);
+ return ret;
+}
+
+/**
+ * intel_vgpu_find_ppgtt_mm - find a PPGTT mm object
+ * @vgpu: a vGPU
+ * @pdps: pdp root array
+ *
+ * This function is used to find a PPGTT mm object from mm object pool
+ *
+ * Returns:
+ * pointer to mm object on success, NULL if failed.
+ */
+struct intel_vgpu_mm *intel_vgpu_find_ppgtt_mm(struct intel_vgpu *vgpu,
+ u64 pdps[])
+{
+ struct intel_vgpu_mm *mm;
+ struct list_head *pos;
+
+ list_for_each(pos, &vgpu->gtt.ppgtt_mm_list_head) {
+ mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.list);
+
+ switch (mm->ppgtt_mm.root_entry_type) {
+ case GTT_TYPE_PPGTT_ROOT_L4_ENTRY:
+ if (pdps[0] == mm->ppgtt_mm.guest_pdps[0])
+ return mm;
+ break;
+ case GTT_TYPE_PPGTT_ROOT_L3_ENTRY:
+ if (!memcmp(pdps, mm->ppgtt_mm.guest_pdps,
+ sizeof(mm->ppgtt_mm.guest_pdps)))
+ return mm;
+ break;
+ default:
+ GEM_BUG_ON(1);
+ }
+ }
+ return NULL;
+}
+
+/**
+ * intel_vgpu_get_ppgtt_mm - get or create a PPGTT mm object.
+ * @vgpu: a vGPU
+ * @root_entry_type: ppgtt root entry type
+ * @pdps: guest pdps
+ *
+ * This function is used to find or create a PPGTT mm object from a guest.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+struct intel_vgpu_mm *intel_vgpu_get_ppgtt_mm(struct intel_vgpu *vgpu,
+ enum intel_gvt_gtt_type root_entry_type, u64 pdps[])
+{
+ struct intel_vgpu_mm *mm;
+
+ mm = intel_vgpu_find_ppgtt_mm(vgpu, pdps);
+ if (mm) {
+ intel_vgpu_mm_get(mm);
+ } else {
+ mm = intel_vgpu_create_ppgtt_mm(vgpu, root_entry_type, pdps);
+ if (IS_ERR(mm))
+ gvt_vgpu_err("fail to create mm\n");
+ }
+ return mm;
+}
+
+/**
+ * intel_vgpu_put_ppgtt_mm - find and put a PPGTT mm object.
+ * @vgpu: a vGPU
+ * @pdps: guest pdps
+ *
+ * This function is used to find a PPGTT mm object from a guest and destroy it.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_put_ppgtt_mm(struct intel_vgpu *vgpu, u64 pdps[])
+{
+ struct intel_vgpu_mm *mm;
+
+ mm = intel_vgpu_find_ppgtt_mm(vgpu, pdps);
+ if (!mm) {
+ gvt_vgpu_err("fail to find ppgtt instance.\n");
+ return -EINVAL;
+ }
+ intel_vgpu_mm_put(mm);
+ return 0;
+}
+
+/**
+ * intel_gvt_init_gtt - initialize mm components of a GVT device
+ * @gvt: GVT device
+ *
+ * This function is called at the initialization stage, to initialize
+ * the mm components of a GVT device.
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ */
+int intel_gvt_init_gtt(struct intel_gvt *gvt)
+{
+ int ret;
+ void *page;
+ struct device *dev = gvt->gt->i915->drm.dev;
+ dma_addr_t daddr;
+
+ gvt_dbg_core("init gtt\n");
+
+ gvt->gtt.pte_ops = &gen8_gtt_pte_ops;
+ gvt->gtt.gma_ops = &gen8_gtt_gma_ops;
+
+ page = (void *)get_zeroed_page(GFP_KERNEL);
+ if (!page) {
+ gvt_err("fail to allocate scratch ggtt page\n");
+ return -ENOMEM;
+ }
+
+ daddr = dma_map_page(dev, virt_to_page(page), 0,
+ 4096, DMA_BIDIRECTIONAL);
+ if (dma_mapping_error(dev, daddr)) {
+ gvt_err("fail to dmamap scratch ggtt page\n");
+ __free_page(virt_to_page(page));
+ return -ENOMEM;
+ }
+
+ gvt->gtt.scratch_page = virt_to_page(page);
+ gvt->gtt.scratch_mfn = (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT);
+
+ if (enable_out_of_sync) {
+ ret = setup_spt_oos(gvt);
+ if (ret) {
+ gvt_err("fail to initialize SPT oos\n");
+ dma_unmap_page(dev, daddr, 4096, DMA_BIDIRECTIONAL);
+ __free_page(gvt->gtt.scratch_page);
+ return ret;
+ }
+ }
+ INIT_LIST_HEAD(&gvt->gtt.ppgtt_mm_lru_list_head);
+ mutex_init(&gvt->gtt.ppgtt_mm_lock);
+ return 0;
+}
+
+/**
+ * intel_gvt_clean_gtt - clean up mm components of a GVT device
+ * @gvt: GVT device
+ *
+ * This function is called at the driver unloading stage, to clean up the
+ * the mm components of a GVT device.
+ *
+ */
+void intel_gvt_clean_gtt(struct intel_gvt *gvt)
+{
+ struct device *dev = gvt->gt->i915->drm.dev;
+ dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_mfn <<
+ I915_GTT_PAGE_SHIFT);
+
+ dma_unmap_page(dev, daddr, 4096, DMA_BIDIRECTIONAL);
+
+ __free_page(gvt->gtt.scratch_page);
+
+ if (enable_out_of_sync)
+ clean_spt_oos(gvt);
+}
+
+/**
+ * intel_vgpu_invalidate_ppgtt - invalidate PPGTT instances
+ * @vgpu: a vGPU
+ *
+ * This function is called when invalidate all PPGTT instances of a vGPU.
+ *
+ */
+void intel_vgpu_invalidate_ppgtt(struct intel_vgpu *vgpu)
+{
+ struct list_head *pos, *n;
+ struct intel_vgpu_mm *mm;
+
+ list_for_each_safe(pos, n, &vgpu->gtt.ppgtt_mm_list_head) {
+ mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.list);
+ if (mm->type == INTEL_GVT_MM_PPGTT) {
+ mutex_lock(&vgpu->gvt->gtt.ppgtt_mm_lock);
+ list_del_init(&mm->ppgtt_mm.lru_list);
+ mutex_unlock(&vgpu->gvt->gtt.ppgtt_mm_lock);
+ if (mm->ppgtt_mm.shadowed)
+ invalidate_ppgtt_mm(mm);
+ }
+ }
+}
+
+/**
+ * intel_vgpu_reset_ggtt - reset the GGTT entry
+ * @vgpu: a vGPU
+ * @invalidate_old: invalidate old entries
+ *
+ * This function is called at the vGPU create stage
+ * to reset all the GGTT entries.
+ *
+ */
+void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu, bool invalidate_old)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_gtt_pte_ops *pte_ops = vgpu->gvt->gtt.pte_ops;
+ struct intel_gvt_gtt_entry entry = {.type = GTT_TYPE_GGTT_PTE};
+ struct intel_gvt_gtt_entry old_entry;
+ u32 index;
+ u32 num_entries;
+
+ pte_ops->set_pfn(&entry, gvt->gtt.scratch_mfn);
+ pte_ops->set_present(&entry);
+
+ index = vgpu_aperture_gmadr_base(vgpu) >> PAGE_SHIFT;
+ num_entries = vgpu_aperture_sz(vgpu) >> PAGE_SHIFT;
+ while (num_entries--) {
+ if (invalidate_old) {
+ ggtt_get_host_entry(vgpu->gtt.ggtt_mm, &old_entry, index);
+ ggtt_invalidate_pte(vgpu, &old_entry);
+ }
+ ggtt_set_host_entry(vgpu->gtt.ggtt_mm, &entry, index++);
+ }
+
+ index = vgpu_hidden_gmadr_base(vgpu) >> PAGE_SHIFT;
+ num_entries = vgpu_hidden_sz(vgpu) >> PAGE_SHIFT;
+ while (num_entries--) {
+ if (invalidate_old) {
+ ggtt_get_host_entry(vgpu->gtt.ggtt_mm, &old_entry, index);
+ ggtt_invalidate_pte(vgpu, &old_entry);
+ }
+ ggtt_set_host_entry(vgpu->gtt.ggtt_mm, &entry, index++);
+ }
+
+ ggtt_invalidate(gvt->gt);
+}
+
+/**
+ * intel_gvt_restore_ggtt - restore all vGPU's ggtt entries
+ * @gvt: intel gvt device
+ *
+ * This function is called at driver resume stage to restore
+ * GGTT entries of every vGPU.
+ *
+ */
+void intel_gvt_restore_ggtt(struct intel_gvt *gvt)
+{
+ struct intel_vgpu *vgpu;
+ struct intel_vgpu_mm *mm;
+ int id;
+ gen8_pte_t pte;
+ u32 idx, num_low, num_hi, offset;
+
+ /* Restore dirty host ggtt for all vGPUs */
+ idr_for_each_entry(&(gvt)->vgpu_idr, vgpu, id) {
+ mm = vgpu->gtt.ggtt_mm;
+
+ num_low = vgpu_aperture_sz(vgpu) >> PAGE_SHIFT;
+ offset = vgpu_aperture_gmadr_base(vgpu) >> PAGE_SHIFT;
+ for (idx = 0; idx < num_low; idx++) {
+ pte = mm->ggtt_mm.host_ggtt_aperture[idx];
+ if (pte & GEN8_PAGE_PRESENT)
+ write_pte64(vgpu->gvt->gt->ggtt, offset + idx, pte);
+ }
+
+ num_hi = vgpu_hidden_sz(vgpu) >> PAGE_SHIFT;
+ offset = vgpu_hidden_gmadr_base(vgpu) >> PAGE_SHIFT;
+ for (idx = 0; idx < num_hi; idx++) {
+ pte = mm->ggtt_mm.host_ggtt_hidden[idx];
+ if (pte & GEN8_PAGE_PRESENT)
+ write_pte64(vgpu->gvt->gt->ggtt, offset + idx, pte);
+ }
+ }
+}
diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
new file mode 100644
index 000000000..4cb183e06
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -0,0 +1,296 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ * Xiao Zheng <xiao.zheng@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#ifndef _GVT_GTT_H_
+#define _GVT_GTT_H_
+
+#include <linux/kernel.h>
+#include <linux/kref.h>
+#include <linux/mutex.h>
+#include <linux/radix-tree.h>
+
+#include "gt/intel_gtt.h"
+
+struct intel_gvt;
+struct intel_vgpu;
+struct intel_vgpu_mm;
+
+#define I915_GTT_PAGE_SHIFT 12
+
+#define INTEL_GVT_INVALID_ADDR (~0UL)
+
+struct intel_gvt_gtt_entry {
+ u64 val64;
+ int type;
+};
+
+struct intel_gvt_gtt_pte_ops {
+ int (*get_entry)(void *pt,
+ struct intel_gvt_gtt_entry *e,
+ unsigned long index,
+ bool hypervisor_access,
+ unsigned long gpa,
+ struct intel_vgpu *vgpu);
+ int (*set_entry)(void *pt,
+ struct intel_gvt_gtt_entry *e,
+ unsigned long index,
+ bool hypervisor_access,
+ unsigned long gpa,
+ struct intel_vgpu *vgpu);
+ bool (*test_present)(struct intel_gvt_gtt_entry *e);
+ void (*clear_present)(struct intel_gvt_gtt_entry *e);
+ void (*set_present)(struct intel_gvt_gtt_entry *e);
+ bool (*test_pse)(struct intel_gvt_gtt_entry *e);
+ void (*clear_pse)(struct intel_gvt_gtt_entry *e);
+ bool (*test_ips)(struct intel_gvt_gtt_entry *e);
+ void (*clear_ips)(struct intel_gvt_gtt_entry *e);
+ bool (*test_64k_splited)(struct intel_gvt_gtt_entry *e);
+ void (*clear_64k_splited)(struct intel_gvt_gtt_entry *e);
+ void (*set_64k_splited)(struct intel_gvt_gtt_entry *e);
+ void (*set_pfn)(struct intel_gvt_gtt_entry *e, unsigned long pfn);
+ unsigned long (*get_pfn)(struct intel_gvt_gtt_entry *e);
+};
+
+struct intel_gvt_gtt_gma_ops {
+ unsigned long (*gma_to_ggtt_pte_index)(unsigned long gma);
+ unsigned long (*gma_to_pte_index)(unsigned long gma);
+ unsigned long (*gma_to_pde_index)(unsigned long gma);
+ unsigned long (*gma_to_l3_pdp_index)(unsigned long gma);
+ unsigned long (*gma_to_l4_pdp_index)(unsigned long gma);
+ unsigned long (*gma_to_pml4_index)(unsigned long gma);
+};
+
+struct intel_gvt_gtt {
+ const struct intel_gvt_gtt_pte_ops *pte_ops;
+ const struct intel_gvt_gtt_gma_ops *gma_ops;
+ int (*mm_alloc_page_table)(struct intel_vgpu_mm *mm);
+ void (*mm_free_page_table)(struct intel_vgpu_mm *mm);
+ struct list_head oos_page_use_list_head;
+ struct list_head oos_page_free_list_head;
+ struct mutex ppgtt_mm_lock;
+ struct list_head ppgtt_mm_lru_list_head;
+
+ struct page *scratch_page;
+ unsigned long scratch_mfn;
+};
+
+enum intel_gvt_gtt_type {
+ GTT_TYPE_INVALID = 0,
+
+ GTT_TYPE_GGTT_PTE,
+
+ GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_64K_ENTRY,
+ GTT_TYPE_PPGTT_PTE_2M_ENTRY,
+ GTT_TYPE_PPGTT_PTE_1G_ENTRY,
+
+ GTT_TYPE_PPGTT_PTE_ENTRY,
+
+ GTT_TYPE_PPGTT_PDE_ENTRY,
+ GTT_TYPE_PPGTT_PDP_ENTRY,
+ GTT_TYPE_PPGTT_PML4_ENTRY,
+
+ GTT_TYPE_PPGTT_ROOT_ENTRY,
+
+ GTT_TYPE_PPGTT_ROOT_L3_ENTRY,
+ GTT_TYPE_PPGTT_ROOT_L4_ENTRY,
+
+ GTT_TYPE_PPGTT_ENTRY,
+
+ GTT_TYPE_PPGTT_PTE_PT,
+ GTT_TYPE_PPGTT_PDE_PT,
+ GTT_TYPE_PPGTT_PDP_PT,
+ GTT_TYPE_PPGTT_PML4_PT,
+
+ GTT_TYPE_MAX,
+};
+
+enum intel_gvt_mm_type {
+ INTEL_GVT_MM_GGTT,
+ INTEL_GVT_MM_PPGTT,
+};
+
+#define GVT_RING_CTX_NR_PDPS GEN8_3LVL_PDPES
+
+struct intel_gvt_partial_pte {
+ unsigned long offset;
+ u64 data;
+ struct list_head list;
+};
+
+struct intel_vgpu_mm {
+ enum intel_gvt_mm_type type;
+ struct intel_vgpu *vgpu;
+
+ struct kref ref;
+ atomic_t pincount;
+
+ union {
+ struct {
+ enum intel_gvt_gtt_type root_entry_type;
+ /*
+ * The 4 PDPs in ring context. For 48bit addressing,
+ * only PDP0 is valid and point to PML4. For 32it
+ * addressing, all 4 are used as true PDPs.
+ */
+ u64 guest_pdps[GVT_RING_CTX_NR_PDPS];
+ u64 shadow_pdps[GVT_RING_CTX_NR_PDPS];
+ bool shadowed;
+
+ struct list_head list;
+ struct list_head lru_list;
+ struct list_head link; /* possible LRI shadow mm list */
+ } ppgtt_mm;
+ struct {
+ void *virtual_ggtt;
+ /* Save/restore for PM */
+ u64 *host_ggtt_aperture;
+ u64 *host_ggtt_hidden;
+ struct list_head partial_pte_list;
+ } ggtt_mm;
+ };
+};
+
+struct intel_vgpu_mm *intel_vgpu_create_ppgtt_mm(struct intel_vgpu *vgpu,
+ enum intel_gvt_gtt_type root_entry_type, u64 pdps[]);
+
+static inline void intel_vgpu_mm_get(struct intel_vgpu_mm *mm)
+{
+ kref_get(&mm->ref);
+}
+
+void _intel_vgpu_mm_release(struct kref *mm_ref);
+
+static inline void intel_vgpu_mm_put(struct intel_vgpu_mm *mm)
+{
+ kref_put(&mm->ref, _intel_vgpu_mm_release);
+}
+
+static inline void intel_vgpu_destroy_mm(struct intel_vgpu_mm *mm)
+{
+ intel_vgpu_mm_put(mm);
+}
+
+struct intel_vgpu_guest_page;
+
+struct intel_vgpu_scratch_pt {
+ struct page *page;
+ unsigned long page_mfn;
+};
+
+struct intel_vgpu_gtt {
+ struct intel_vgpu_mm *ggtt_mm;
+ unsigned long active_ppgtt_mm_bitmap;
+ struct list_head ppgtt_mm_list_head;
+ struct radix_tree_root spt_tree;
+ struct list_head oos_page_list_head;
+ struct list_head post_shadow_list_head;
+ struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX];
+};
+
+int intel_vgpu_init_gtt(struct intel_vgpu *vgpu);
+void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu);
+void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu, bool invalidate_old);
+void intel_vgpu_invalidate_ppgtt(struct intel_vgpu *vgpu);
+
+int intel_gvt_init_gtt(struct intel_gvt *gvt);
+void intel_gvt_clean_gtt(struct intel_gvt *gvt);
+
+struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu,
+ int page_table_level,
+ void *root_entry);
+
+struct intel_vgpu_oos_page {
+ struct intel_vgpu_ppgtt_spt *spt;
+ struct list_head list;
+ struct list_head vm_list;
+ int id;
+ void *mem;
+};
+
+#define GTT_ENTRY_NUM_IN_ONE_PAGE 512
+
+/* Represent a vgpu shadow page table. */
+struct intel_vgpu_ppgtt_spt {
+ atomic_t refcount;
+ struct intel_vgpu *vgpu;
+
+ struct {
+ enum intel_gvt_gtt_type type;
+ bool pde_ips; /* for 64KB PTEs */
+ void *vaddr;
+ struct page *page;
+ unsigned long mfn;
+ } shadow_page;
+
+ struct {
+ enum intel_gvt_gtt_type type;
+ bool pde_ips; /* for 64KB PTEs */
+ unsigned long gfn;
+ unsigned long write_cnt;
+ struct intel_vgpu_oos_page *oos_page;
+ } guest_page;
+
+ DECLARE_BITMAP(post_shadow_bitmap, GTT_ENTRY_NUM_IN_ONE_PAGE);
+ struct list_head post_shadow_list;
+};
+
+int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu);
+
+int intel_vgpu_flush_post_shadow(struct intel_vgpu *vgpu);
+
+int intel_vgpu_pin_mm(struct intel_vgpu_mm *mm);
+
+void intel_vgpu_unpin_mm(struct intel_vgpu_mm *mm);
+
+unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm,
+ unsigned long gma);
+
+struct intel_vgpu_mm *intel_vgpu_find_ppgtt_mm(struct intel_vgpu *vgpu,
+ u64 pdps[]);
+
+struct intel_vgpu_mm *intel_vgpu_get_ppgtt_mm(struct intel_vgpu *vgpu,
+ enum intel_gvt_gtt_type root_entry_type, u64 pdps[]);
+
+int intel_vgpu_put_ppgtt_mm(struct intel_vgpu *vgpu, u64 pdps[]);
+
+int intel_vgpu_emulate_ggtt_mmio_read(struct intel_vgpu *vgpu,
+ unsigned int off, void *p_data, unsigned int bytes);
+
+int intel_vgpu_emulate_ggtt_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int off, void *p_data, unsigned int bytes);
+
+void intel_vgpu_destroy_all_ppgtt_mm(struct intel_vgpu *vgpu);
+void intel_gvt_restore_ggtt(struct intel_gvt *gvt);
+
+#endif /* _GVT_GTT_H_ */
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
new file mode 100644
index 000000000..dbf8d7470
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -0,0 +1,767 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Eddie Dong <eddie.dong@intel.com>
+ *
+ * Contributors:
+ * Niu Bing <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_H_
+#define _GVT_H_
+
+#include <uapi/linux/pci_regs.h>
+#include <linux/kvm_host.h>
+#include <linux/vfio.h>
+#include <linux/mdev.h>
+
+#include "i915_drv.h"
+#include "intel_gvt.h"
+
+#include "debug.h"
+#include "mmio.h"
+#include "reg.h"
+#include "interrupt.h"
+#include "gtt.h"
+#include "display.h"
+#include "edid.h"
+#include "execlist.h"
+#include "scheduler.h"
+#include "sched_policy.h"
+#include "mmio_context.h"
+#include "cmd_parser.h"
+#include "fb_decoder.h"
+#include "dmabuf.h"
+#include "page_track.h"
+
+#define GVT_MAX_VGPU 8
+
+/* Describe per-platform limitations. */
+struct intel_gvt_device_info {
+ u32 max_support_vgpus;
+ u32 cfg_space_size;
+ u32 mmio_size;
+ u32 mmio_bar;
+ unsigned long msi_cap_offset;
+ u32 gtt_start_offset;
+ u32 gtt_entry_size;
+ u32 gtt_entry_size_shift;
+ int gmadr_bytes_in_cmd;
+ u32 max_surface_size;
+};
+
+/* GM resources owned by a vGPU */
+struct intel_vgpu_gm {
+ u64 aperture_sz;
+ u64 hidden_sz;
+ struct drm_mm_node low_gm_node;
+ struct drm_mm_node high_gm_node;
+};
+
+#define INTEL_GVT_MAX_NUM_FENCES 32
+
+/* Fences owned by a vGPU */
+struct intel_vgpu_fence {
+ struct i915_fence_reg *regs[INTEL_GVT_MAX_NUM_FENCES];
+ u32 base;
+ u32 size;
+};
+
+struct intel_vgpu_mmio {
+ void *vreg;
+};
+
+#define INTEL_GVT_MAX_BAR_NUM 4
+
+struct intel_vgpu_pci_bar {
+ u64 size;
+ bool tracked;
+};
+
+struct intel_vgpu_cfg_space {
+ unsigned char virtual_cfg_space[PCI_CFG_SPACE_EXP_SIZE];
+ struct intel_vgpu_pci_bar bar[INTEL_GVT_MAX_BAR_NUM];
+ u32 pmcsr_off;
+};
+
+#define vgpu_cfg_space(vgpu) ((vgpu)->cfg_space.virtual_cfg_space)
+
+struct intel_vgpu_irq {
+ bool irq_warn_once[INTEL_GVT_EVENT_MAX];
+ DECLARE_BITMAP(flip_done_event[I915_MAX_PIPES],
+ INTEL_GVT_EVENT_MAX);
+};
+
+struct intel_vgpu_opregion {
+ bool mapped;
+ void *va;
+ u32 gfn[INTEL_GVT_OPREGION_PAGES];
+};
+
+#define vgpu_opregion(vgpu) (&(vgpu->opregion))
+
+struct intel_vgpu_display {
+ struct intel_vgpu_i2c_edid i2c_edid;
+ struct intel_vgpu_port ports[I915_MAX_PORTS];
+ struct intel_vgpu_sbi sbi;
+ enum port port_num;
+};
+
+struct vgpu_sched_ctl {
+ int weight;
+};
+
+enum {
+ INTEL_VGPU_EXECLIST_SUBMISSION = 1,
+ INTEL_VGPU_GUC_SUBMISSION,
+};
+
+struct intel_vgpu_submission_ops {
+ const char *name;
+ int (*init)(struct intel_vgpu *vgpu, intel_engine_mask_t engine_mask);
+ void (*clean)(struct intel_vgpu *vgpu, intel_engine_mask_t engine_mask);
+ void (*reset)(struct intel_vgpu *vgpu, intel_engine_mask_t engine_mask);
+};
+
+struct intel_vgpu_submission {
+ struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
+ struct list_head workload_q_head[I915_NUM_ENGINES];
+ struct intel_context *shadow[I915_NUM_ENGINES];
+ struct kmem_cache *workloads;
+ atomic_t running_workload_num;
+ union {
+ u64 i915_context_pml4;
+ u64 i915_context_pdps[GEN8_3LVL_PDPES];
+ };
+ DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+ DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
+ void *ring_scan_buffer[I915_NUM_ENGINES];
+ int ring_scan_buffer_size[I915_NUM_ENGINES];
+ const struct intel_vgpu_submission_ops *ops;
+ int virtual_submission_interface;
+ bool active;
+ struct {
+ u32 lrca;
+ bool valid;
+ u64 ring_context_gpa;
+ } last_ctx[I915_NUM_ENGINES];
+};
+
+#define KVMGT_DEBUGFS_FILENAME "kvmgt_nr_cache_entries"
+
+struct intel_vgpu {
+ struct vfio_device vfio_device;
+ struct intel_gvt *gvt;
+ struct mutex vgpu_lock;
+ int id;
+ bool active;
+ bool attached;
+ bool pv_notified;
+ bool failsafe;
+ unsigned int resetting_eng;
+
+ /* Both sched_data and sched_ctl can be seen a part of the global gvt
+ * scheduler structure. So below 2 vgpu data are protected
+ * by sched_lock, not vgpu_lock.
+ */
+ void *sched_data;
+ struct vgpu_sched_ctl sched_ctl;
+
+ struct intel_vgpu_fence fence;
+ struct intel_vgpu_gm gm;
+ struct intel_vgpu_cfg_space cfg_space;
+ struct intel_vgpu_mmio mmio;
+ struct intel_vgpu_irq irq;
+ struct intel_vgpu_gtt gtt;
+ struct intel_vgpu_opregion opregion;
+ struct intel_vgpu_display display;
+ struct intel_vgpu_submission submission;
+ struct radix_tree_root page_track_tree;
+ u32 hws_pga[I915_NUM_ENGINES];
+ /* Set on PCI_D3, reset on DMLR, not reflecting the actual PM state */
+ bool d3_entered;
+
+ struct dentry *debugfs;
+
+ struct list_head dmabuf_obj_list_head;
+ struct mutex dmabuf_lock;
+ struct idr object_idr;
+ struct intel_vgpu_vblank_timer vblank_timer;
+
+ u32 scan_nonprivbb;
+
+ struct vfio_region *region;
+ int num_regions;
+ struct eventfd_ctx *intx_trigger;
+ struct eventfd_ctx *msi_trigger;
+
+ /*
+ * Two caches are used to avoid mapping duplicated pages (eg.
+ * scratch pages). This help to reduce dma setup overhead.
+ */
+ struct rb_root gfn_cache;
+ struct rb_root dma_addr_cache;
+ unsigned long nr_cache_entries;
+ struct mutex cache_lock;
+
+ atomic_t released;
+
+ struct kvm_page_track_notifier_node track_node;
+#define NR_BKT (1 << 18)
+ struct hlist_head ptable[NR_BKT];
+#undef NR_BKT
+};
+
+/* validating GM healthy status*/
+#define vgpu_is_vm_unhealthy(ret_val) \
+ (((ret_val) == -EBADRQC) || ((ret_val) == -EFAULT))
+
+struct intel_gvt_gm {
+ unsigned long vgpu_allocated_low_gm_size;
+ unsigned long vgpu_allocated_high_gm_size;
+};
+
+struct intel_gvt_fence {
+ unsigned long vgpu_allocated_fence_num;
+};
+
+/* Special MMIO blocks. */
+struct gvt_mmio_block {
+ unsigned int device;
+ i915_reg_t offset;
+ unsigned int size;
+ gvt_mmio_func read;
+ gvt_mmio_func write;
+};
+
+#define INTEL_GVT_MMIO_HASH_BITS 11
+
+struct intel_gvt_mmio {
+ u16 *mmio_attribute;
+/* Register contains RO bits */
+#define F_RO (1 << 0)
+/* Register contains graphics address */
+#define F_GMADR (1 << 1)
+/* Mode mask registers with high 16 bits as the mask bits */
+#define F_MODE_MASK (1 << 2)
+/* This reg can be accessed by GPU commands */
+#define F_CMD_ACCESS (1 << 3)
+/* This reg has been accessed by a VM */
+#define F_ACCESSED (1 << 4)
+/* This reg requires save & restore during host PM suspend/resume */
+#define F_PM_SAVE (1 << 5)
+/* This reg could be accessed by unaligned address */
+#define F_UNALIGN (1 << 6)
+/* This reg is in GVT's mmio save-restor list and in hardware
+ * logical context image
+ */
+#define F_SR_IN_CTX (1 << 7)
+/* Value of command write of this reg needs to be patched */
+#define F_CMD_WRITE_PATCH (1 << 8)
+
+ struct gvt_mmio_block *mmio_block;
+ unsigned int num_mmio_block;
+
+ DECLARE_HASHTABLE(mmio_info_table, INTEL_GVT_MMIO_HASH_BITS);
+ unsigned long num_tracked_mmio;
+};
+
+struct intel_gvt_firmware {
+ void *cfg_space;
+ void *mmio;
+ bool firmware_loaded;
+};
+
+struct intel_vgpu_config {
+ unsigned int low_mm;
+ unsigned int high_mm;
+ unsigned int fence;
+
+ /*
+ * A vGPU with a weight of 8 will get twice as much GPU as a vGPU with
+ * a weight of 4 on a contended host, different vGPU type has different
+ * weight set. Legal weights range from 1 to 16.
+ */
+ unsigned int weight;
+ enum intel_vgpu_edid edid;
+ const char *name;
+};
+
+struct intel_vgpu_type {
+ struct mdev_type type;
+ char name[16];
+ const struct intel_vgpu_config *conf;
+};
+
+struct intel_gvt {
+ /* GVT scope lock, protect GVT itself, and all resource currently
+ * not yet protected by special locks(vgpu and scheduler lock).
+ */
+ struct mutex lock;
+ /* scheduler scope lock, protect gvt and vgpu schedule related data */
+ struct mutex sched_lock;
+
+ struct intel_gt *gt;
+ struct idr vgpu_idr; /* vGPU IDR pool */
+
+ struct intel_gvt_device_info device_info;
+ struct intel_gvt_gm gm;
+ struct intel_gvt_fence fence;
+ struct intel_gvt_mmio mmio;
+ struct intel_gvt_firmware firmware;
+ struct intel_gvt_irq irq;
+ struct intel_gvt_gtt gtt;
+ struct intel_gvt_workload_scheduler scheduler;
+ struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
+ DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
+ struct mdev_parent parent;
+ struct mdev_type **mdev_types;
+ struct intel_vgpu_type *types;
+ unsigned int num_types;
+ struct intel_vgpu *idle_vgpu;
+
+ struct task_struct *service_thread;
+ wait_queue_head_t service_thread_wq;
+
+ /* service_request is always used in bit operation, we should always
+ * use it with atomic bit ops so that no need to use gvt big lock.
+ */
+ unsigned long service_request;
+
+ struct {
+ struct engine_mmio *mmio;
+ int ctx_mmio_count[I915_NUM_ENGINES];
+ u32 *tlb_mmio_offset_list;
+ u32 tlb_mmio_offset_list_cnt;
+ u32 *mocs_mmio_offset_list;
+ u32 mocs_mmio_offset_list_cnt;
+ } engine_mmio_list;
+ bool is_reg_whitelist_updated;
+
+ struct dentry *debugfs_root;
+};
+
+static inline struct intel_gvt *to_gvt(struct drm_i915_private *i915)
+{
+ return i915->gvt;
+}
+
+enum {
+ /* Scheduling trigger by timer */
+ INTEL_GVT_REQUEST_SCHED = 0,
+
+ /* Scheduling trigger by event */
+ INTEL_GVT_REQUEST_EVENT_SCHED = 1,
+
+ /* per-vGPU vblank emulation request */
+ INTEL_GVT_REQUEST_EMULATE_VBLANK = 2,
+ INTEL_GVT_REQUEST_EMULATE_VBLANK_MAX = INTEL_GVT_REQUEST_EMULATE_VBLANK
+ + GVT_MAX_VGPU,
+};
+
+static inline void intel_gvt_request_service(struct intel_gvt *gvt,
+ int service)
+{
+ set_bit(service, (void *)&gvt->service_request);
+ wake_up(&gvt->service_thread_wq);
+}
+
+void intel_gvt_free_firmware(struct intel_gvt *gvt);
+int intel_gvt_load_firmware(struct intel_gvt *gvt);
+
+/* Aperture/GM space definitions for GVT device */
+#define MB_TO_BYTES(mb) ((mb) << 20ULL)
+#define BYTES_TO_MB(b) ((b) >> 20ULL)
+
+#define HOST_LOW_GM_SIZE MB_TO_BYTES(128)
+#define HOST_HIGH_GM_SIZE MB_TO_BYTES(384)
+#define HOST_FENCE 4
+
+#define gvt_to_ggtt(gvt) ((gvt)->gt->ggtt)
+
+/* Aperture/GM space definitions for GVT device */
+#define gvt_aperture_sz(gvt) gvt_to_ggtt(gvt)->mappable_end
+#define gvt_aperture_pa_base(gvt) gvt_to_ggtt(gvt)->gmadr.start
+
+#define gvt_ggtt_gm_sz(gvt) gvt_to_ggtt(gvt)->vm.total
+#define gvt_ggtt_sz(gvt) (gvt_to_ggtt(gvt)->vm.total >> PAGE_SHIFT << 3)
+#define gvt_hidden_sz(gvt) (gvt_ggtt_gm_sz(gvt) - gvt_aperture_sz(gvt))
+
+#define gvt_aperture_gmadr_base(gvt) (0)
+#define gvt_aperture_gmadr_end(gvt) (gvt_aperture_gmadr_base(gvt) \
+ + gvt_aperture_sz(gvt) - 1)
+
+#define gvt_hidden_gmadr_base(gvt) (gvt_aperture_gmadr_base(gvt) \
+ + gvt_aperture_sz(gvt))
+#define gvt_hidden_gmadr_end(gvt) (gvt_hidden_gmadr_base(gvt) \
+ + gvt_hidden_sz(gvt) - 1)
+
+#define gvt_fence_sz(gvt) (gvt_to_ggtt(gvt)->num_fences)
+
+/* Aperture/GM space definitions for vGPU */
+#define vgpu_aperture_offset(vgpu) ((vgpu)->gm.low_gm_node.start)
+#define vgpu_hidden_offset(vgpu) ((vgpu)->gm.high_gm_node.start)
+#define vgpu_aperture_sz(vgpu) ((vgpu)->gm.aperture_sz)
+#define vgpu_hidden_sz(vgpu) ((vgpu)->gm.hidden_sz)
+
+#define vgpu_aperture_pa_base(vgpu) \
+ (gvt_aperture_pa_base(vgpu->gvt) + vgpu_aperture_offset(vgpu))
+
+#define vgpu_ggtt_gm_sz(vgpu) ((vgpu)->gm.aperture_sz + (vgpu)->gm.hidden_sz)
+
+#define vgpu_aperture_pa_end(vgpu) \
+ (vgpu_aperture_pa_base(vgpu) + vgpu_aperture_sz(vgpu) - 1)
+
+#define vgpu_aperture_gmadr_base(vgpu) (vgpu_aperture_offset(vgpu))
+#define vgpu_aperture_gmadr_end(vgpu) \
+ (vgpu_aperture_gmadr_base(vgpu) + vgpu_aperture_sz(vgpu) - 1)
+
+#define vgpu_hidden_gmadr_base(vgpu) (vgpu_hidden_offset(vgpu))
+#define vgpu_hidden_gmadr_end(vgpu) \
+ (vgpu_hidden_gmadr_base(vgpu) + vgpu_hidden_sz(vgpu) - 1)
+
+#define vgpu_fence_base(vgpu) (vgpu->fence.base)
+#define vgpu_fence_sz(vgpu) (vgpu->fence.size)
+
+/* ring context size i.e. the first 0x50 dwords*/
+#define RING_CTX_SIZE 320
+
+int intel_vgpu_alloc_resource(struct intel_vgpu *vgpu,
+ const struct intel_vgpu_config *conf);
+void intel_vgpu_reset_resource(struct intel_vgpu *vgpu);
+void intel_vgpu_free_resource(struct intel_vgpu *vgpu);
+void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
+ u32 fence, u64 value);
+
+/* Macros for easily accessing vGPU virtual/shadow register.
+ Explicitly seperate use for typed MMIO reg or real offset.*/
+#define vgpu_vreg_t(vgpu, reg) \
+ (*(u32 *)(vgpu->mmio.vreg + i915_mmio_reg_offset(reg)))
+#define vgpu_vreg(vgpu, offset) \
+ (*(u32 *)(vgpu->mmio.vreg + (offset)))
+#define vgpu_vreg64_t(vgpu, reg) \
+ (*(u64 *)(vgpu->mmio.vreg + i915_mmio_reg_offset(reg)))
+#define vgpu_vreg64(vgpu, offset) \
+ (*(u64 *)(vgpu->mmio.vreg + (offset)))
+
+#define for_each_active_vgpu(gvt, vgpu, id) \
+ idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) \
+ for_each_if(vgpu->active)
+
+static inline void intel_vgpu_write_pci_bar(struct intel_vgpu *vgpu,
+ u32 offset, u32 val, bool low)
+{
+ u32 *pval;
+
+ /* BAR offset should be 32 bits algiend */
+ offset = rounddown(offset, 4);
+ pval = (u32 *)(vgpu_cfg_space(vgpu) + offset);
+
+ if (low) {
+ /*
+ * only update bit 31 - bit 4,
+ * leave the bit 3 - bit 0 unchanged.
+ */
+ *pval = (val & GENMASK(31, 4)) | (*pval & GENMASK(3, 0));
+ } else {
+ *pval = val;
+ }
+}
+
+int intel_gvt_init_vgpu_types(struct intel_gvt *gvt);
+void intel_gvt_clean_vgpu_types(struct intel_gvt *gvt);
+
+struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt);
+void intel_gvt_destroy_idle_vgpu(struct intel_vgpu *vgpu);
+int intel_gvt_create_vgpu(struct intel_vgpu *vgpu,
+ const struct intel_vgpu_config *conf);
+void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_release_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
+ intel_engine_mask_t engine_mask);
+void intel_gvt_reset_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_activate_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu);
+
+int intel_gvt_set_opregion(struct intel_vgpu *vgpu);
+int intel_gvt_set_edid(struct intel_vgpu *vgpu, int port_num);
+
+/* validating GM functions */
+#define vgpu_gmadr_is_aperture(vgpu, gmadr) \
+ ((gmadr >= vgpu_aperture_gmadr_base(vgpu)) && \
+ (gmadr <= vgpu_aperture_gmadr_end(vgpu)))
+
+#define vgpu_gmadr_is_hidden(vgpu, gmadr) \
+ ((gmadr >= vgpu_hidden_gmadr_base(vgpu)) && \
+ (gmadr <= vgpu_hidden_gmadr_end(vgpu)))
+
+#define vgpu_gmadr_is_valid(vgpu, gmadr) \
+ ((vgpu_gmadr_is_aperture(vgpu, gmadr) || \
+ (vgpu_gmadr_is_hidden(vgpu, gmadr))))
+
+#define gvt_gmadr_is_aperture(gvt, gmadr) \
+ ((gmadr >= gvt_aperture_gmadr_base(gvt)) && \
+ (gmadr <= gvt_aperture_gmadr_end(gvt)))
+
+#define gvt_gmadr_is_hidden(gvt, gmadr) \
+ ((gmadr >= gvt_hidden_gmadr_base(gvt)) && \
+ (gmadr <= gvt_hidden_gmadr_end(gvt)))
+
+#define gvt_gmadr_is_valid(gvt, gmadr) \
+ (gvt_gmadr_is_aperture(gvt, gmadr) || \
+ gvt_gmadr_is_hidden(gvt, gmadr))
+
+bool intel_gvt_ggtt_validate_range(struct intel_vgpu *vgpu, u64 addr, u32 size);
+int intel_gvt_ggtt_gmadr_g2h(struct intel_vgpu *vgpu, u64 g_addr, u64 *h_addr);
+int intel_gvt_ggtt_gmadr_h2g(struct intel_vgpu *vgpu, u64 h_addr, u64 *g_addr);
+int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index,
+ unsigned long *h_index);
+int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index,
+ unsigned long *g_index);
+
+void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu,
+ bool primary);
+void intel_vgpu_reset_cfg_space(struct intel_vgpu *vgpu);
+
+int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes);
+
+int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes);
+
+void intel_vgpu_emulate_hotplug(struct intel_vgpu *vgpu, bool connected);
+
+static inline u64 intel_vgpu_get_bar_gpa(struct intel_vgpu *vgpu, int bar)
+{
+ /* We are 64bit bar. */
+ return (*(u64 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
+ PCI_BASE_ADDRESS_MEM_MASK;
+}
+
+void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu);
+int intel_vgpu_init_opregion(struct intel_vgpu *vgpu);
+int intel_vgpu_opregion_base_write_handler(struct intel_vgpu *vgpu, u32 gpa);
+
+int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci);
+void populate_pvinfo_page(struct intel_vgpu *vgpu);
+
+int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload);
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason);
+void intel_vgpu_detach_regions(struct intel_vgpu *vgpu);
+
+enum {
+ GVT_FAILSAFE_UNSUPPORTED_GUEST,
+ GVT_FAILSAFE_INSUFFICIENT_RESOURCE,
+ GVT_FAILSAFE_GUEST_ERR,
+};
+
+static inline void mmio_hw_access_pre(struct intel_gt *gt)
+{
+ intel_runtime_pm_get(gt->uncore->rpm);
+}
+
+static inline void mmio_hw_access_post(struct intel_gt *gt)
+{
+ intel_runtime_pm_put_unchecked(gt->uncore->rpm);
+}
+
+/**
+ * intel_gvt_mmio_set_accessed - mark a MMIO has been accessed
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline void intel_gvt_mmio_set_accessed(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ gvt->mmio.mmio_attribute[offset >> 2] |= F_ACCESSED;
+}
+
+/**
+ * intel_gvt_mmio_is_cmd_accessible - if a MMIO could be accessed by command
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if an MMIO is able to be accessed by GPU commands
+ */
+static inline bool intel_gvt_mmio_is_cmd_accessible(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ return gvt->mmio.mmio_attribute[offset >> 2] & F_CMD_ACCESS;
+}
+
+/**
+ * intel_gvt_mmio_set_cmd_accessible -
+ * mark a MMIO could be accessible by command
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline void intel_gvt_mmio_set_cmd_accessible(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ gvt->mmio.mmio_attribute[offset >> 2] |= F_CMD_ACCESS;
+}
+
+/**
+ * intel_gvt_mmio_is_unalign - mark a MMIO could be accessed unaligned
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline bool intel_gvt_mmio_is_unalign(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ return gvt->mmio.mmio_attribute[offset >> 2] & F_UNALIGN;
+}
+
+/**
+ * intel_gvt_mmio_has_mode_mask - if a MMIO has a mode mask
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if a MMIO has a mode mask in its higher 16 bits, false if it isn't.
+ *
+ */
+static inline bool intel_gvt_mmio_has_mode_mask(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ return gvt->mmio.mmio_attribute[offset >> 2] & F_MODE_MASK;
+}
+
+/**
+ * intel_gvt_mmio_is_sr_in_ctx -
+ * check if an MMIO has F_SR_IN_CTX mask
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if an MMIO has an F_SR_IN_CTX mask, false if it isn't.
+ *
+ */
+static inline bool intel_gvt_mmio_is_sr_in_ctx(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ return gvt->mmio.mmio_attribute[offset >> 2] & F_SR_IN_CTX;
+}
+
+/**
+ * intel_gvt_mmio_set_sr_in_ctx -
+ * mask an MMIO in GVT's mmio save-restore list and also
+ * in hardware logical context image
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline void intel_gvt_mmio_set_sr_in_ctx(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ gvt->mmio.mmio_attribute[offset >> 2] |= F_SR_IN_CTX;
+}
+
+void intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu);
+/**
+ * intel_gvt_mmio_set_cmd_write_patch -
+ * mark an MMIO if its cmd write needs to be
+ * patched
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline void intel_gvt_mmio_set_cmd_write_patch(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ gvt->mmio.mmio_attribute[offset >> 2] |= F_CMD_WRITE_PATCH;
+}
+
+/**
+ * intel_gvt_mmio_is_cmd_write_patch - check if an mmio's cmd access needs to
+ * be patched
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if GPU commmand write to an MMIO should be patched
+ */
+static inline bool intel_gvt_mmio_is_cmd_write_patch(
+ struct intel_gvt *gvt, unsigned int offset)
+{
+ return gvt->mmio.mmio_attribute[offset >> 2] & F_CMD_WRITE_PATCH;
+}
+
+/**
+ * intel_gvt_read_gpa - copy data from GPA to host data buffer
+ * @vgpu: a vGPU
+ * @gpa: guest physical address
+ * @buf: host data buffer
+ * @len: data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_read_gpa(struct intel_vgpu *vgpu, unsigned long gpa,
+ void *buf, unsigned long len)
+{
+ if (!vgpu->attached)
+ return -ESRCH;
+ return vfio_dma_rw(&vgpu->vfio_device, gpa, buf, len, false);
+}
+
+/**
+ * intel_gvt_write_gpa - copy data from host data buffer to GPA
+ * @vgpu: a vGPU
+ * @gpa: guest physical address
+ * @buf: host data buffer
+ * @len: data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_write_gpa(struct intel_vgpu *vgpu,
+ unsigned long gpa, void *buf, unsigned long len)
+{
+ if (!vgpu->attached)
+ return -ESRCH;
+ return vfio_dma_rw(&vgpu->vfio_device, gpa, buf, len, true);
+}
+
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_debugfs_init(struct intel_gvt *gvt);
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt);
+
+int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn);
+int intel_gvt_page_track_remove(struct intel_vgpu *info, u64 gfn);
+int intel_gvt_dma_pin_guest_page(struct intel_vgpu *vgpu, dma_addr_t dma_addr);
+int intel_gvt_dma_map_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ unsigned long size, dma_addr_t *dma_addr);
+void intel_gvt_dma_unmap_guest_page(struct intel_vgpu *vgpu,
+ dma_addr_t dma_addr);
+
+#include "trace.h"
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
new file mode 100644
index 000000000..daac2050d
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -0,0 +1,3231 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Eddie Dong <eddie.dong@intel.com>
+ * Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Pei Zhang <pei.zhang@intel.com>
+ * Niu Bing <bing.niu@intel.com>
+ * Ping Gao <ping.a.gao@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+
+ */
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+#include "intel_mchbar_regs.h"
+#include "display/intel_display_types.h"
+#include "display/intel_dmc_regs.h"
+#include "display/intel_fbc.h"
+#include "display/vlv_dsi_pll_regs.h"
+#include "gt/intel_gt_regs.h"
+
+/* XXX FIXME i915 has changed PP_XXX definition */
+#define PCH_PP_STATUS _MMIO(0xc7200)
+#define PCH_PP_CONTROL _MMIO(0xc7204)
+#define PCH_PP_ON_DELAYS _MMIO(0xc7208)
+#define PCH_PP_OFF_DELAYS _MMIO(0xc720c)
+#define PCH_PP_DIVISOR _MMIO(0xc7210)
+
+unsigned long intel_gvt_get_device_type(struct intel_gvt *gvt)
+{
+ struct drm_i915_private *i915 = gvt->gt->i915;
+
+ if (IS_BROADWELL(i915))
+ return D_BDW;
+ else if (IS_SKYLAKE(i915))
+ return D_SKL;
+ else if (IS_KABYLAKE(i915))
+ return D_KBL;
+ else if (IS_BROXTON(i915))
+ return D_BXT;
+ else if (IS_COFFEELAKE(i915) || IS_COMETLAKE(i915))
+ return D_CFL;
+
+ return 0;
+}
+
+static bool intel_gvt_match_device(struct intel_gvt *gvt,
+ unsigned long device)
+{
+ return intel_gvt_get_device_type(gvt) & device;
+}
+
+static void read_vreg(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+}
+
+static void write_vreg(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ memcpy(&vgpu_vreg(vgpu, offset), p_data, bytes);
+}
+
+struct intel_gvt_mmio_info *intel_gvt_find_mmio_info(struct intel_gvt *gvt,
+ unsigned int offset)
+{
+ struct intel_gvt_mmio_info *e;
+
+ hash_for_each_possible(gvt->mmio.mmio_info_table, e, node, offset) {
+ if (e->offset == offset)
+ return e;
+ }
+ return NULL;
+}
+
+static int setup_mmio_info(struct intel_gvt *gvt, u32 offset, u32 size,
+ u16 flags, u32 addr_mask, u32 ro_mask, u32 device,
+ gvt_mmio_func read, gvt_mmio_func write)
+{
+ struct intel_gvt_mmio_info *p;
+ u32 start, end, i;
+
+ if (!intel_gvt_match_device(gvt, device))
+ return 0;
+
+ if (WARN_ON(!IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+
+ start = offset;
+ end = offset + size;
+
+ for (i = start; i < end; i += 4) {
+ p = intel_gvt_find_mmio_info(gvt, i);
+ if (!p) {
+ WARN(1, "assign a handler to a non-tracked mmio %x\n",
+ i);
+ return -ENODEV;
+ }
+ p->ro_mask = ro_mask;
+ gvt->mmio.mmio_attribute[i / 4] = flags;
+ if (read)
+ p->read = read;
+ if (write)
+ p->write = write;
+ }
+ return 0;
+}
+
+/**
+ * intel_gvt_render_mmio_to_engine - convert a mmio offset into the engine
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * The engine containing the offset within its mmio page.
+ */
+const struct intel_engine_cs *
+intel_gvt_render_mmio_to_engine(struct intel_gvt *gvt, unsigned int offset)
+{
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+
+ offset &= ~GENMASK(11, 0);
+ for_each_engine(engine, gvt->gt, id)
+ if (engine->mmio_base == offset)
+ return engine;
+
+ return NULL;
+}
+
+#define offset_to_fence_num(offset) \
+ ((offset - i915_mmio_reg_offset(FENCE_REG_GEN6_LO(0))) >> 3)
+
+#define fence_num_to_offset(num) \
+ (num * 8 + i915_mmio_reg_offset(FENCE_REG_GEN6_LO(0)))
+
+
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
+{
+ switch (reason) {
+ case GVT_FAILSAFE_UNSUPPORTED_GUEST:
+ pr_err("Detected your guest driver doesn't support GVT-g.\n");
+ break;
+ case GVT_FAILSAFE_INSUFFICIENT_RESOURCE:
+ pr_err("Graphics resource is not enough for the guest\n");
+ break;
+ case GVT_FAILSAFE_GUEST_ERR:
+ pr_err("GVT Internal error for the guest\n");
+ break;
+ default:
+ break;
+ }
+ pr_err("Now vgpu %d will enter failsafe mode.\n", vgpu->id);
+ vgpu->failsafe = true;
+}
+
+static int sanitize_fence_mmio_access(struct intel_vgpu *vgpu,
+ unsigned int fence_num, void *p_data, unsigned int bytes)
+{
+ unsigned int max_fence = vgpu_fence_sz(vgpu);
+
+ if (fence_num >= max_fence) {
+ gvt_vgpu_err("access oob fence reg %d/%d\n",
+ fence_num, max_fence);
+
+ /* When guest access oob fence regs without access
+ * pv_info first, we treat guest not supporting GVT,
+ * and we will let vgpu enter failsafe mode.
+ */
+ if (!vgpu->pv_notified)
+ enter_failsafe_mode(vgpu,
+ GVT_FAILSAFE_UNSUPPORTED_GUEST);
+
+ memset(p_data, 0, bytes);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+static int gamw_echo_dev_rw_ia_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 ips = (*(u32 *)p_data) & GAMW_ECO_ENABLE_64K_IPS_FIELD;
+
+ if (GRAPHICS_VER(vgpu->gvt->gt->i915) <= 10) {
+ if (ips == GAMW_ECO_ENABLE_64K_IPS_FIELD)
+ gvt_dbg_core("vgpu%d: ips enabled\n", vgpu->id);
+ else if (!ips)
+ gvt_dbg_core("vgpu%d: ips disabled\n", vgpu->id);
+ else {
+ /* All engines must be enabled together for vGPU,
+ * since we don't know which engine the ppgtt will
+ * bind to when shadowing.
+ */
+ gvt_vgpu_err("Unsupported IPS setting %x, cannot enable 64K gtt.\n",
+ ips);
+ return -EINVAL;
+ }
+ }
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+static int fence_mmio_read(struct intel_vgpu *vgpu, unsigned int off,
+ void *p_data, unsigned int bytes)
+{
+ int ret;
+
+ ret = sanitize_fence_mmio_access(vgpu, offset_to_fence_num(off),
+ p_data, bytes);
+ if (ret)
+ return ret;
+ read_vreg(vgpu, off, p_data, bytes);
+ return 0;
+}
+
+static int fence_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
+ void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ unsigned int fence_num = offset_to_fence_num(off);
+ int ret;
+
+ ret = sanitize_fence_mmio_access(vgpu, fence_num, p_data, bytes);
+ if (ret)
+ return ret;
+ write_vreg(vgpu, off, p_data, bytes);
+
+ mmio_hw_access_pre(gvt->gt);
+ intel_vgpu_write_fence(vgpu, fence_num,
+ vgpu_vreg64(vgpu, fence_num_to_offset(fence_num)));
+ mmio_hw_access_post(gvt->gt);
+ return 0;
+}
+
+#define CALC_MODE_MASK_REG(old, new) \
+ (((new) & GENMASK(31, 16)) \
+ | ((((old) & GENMASK(15, 0)) & ~((new) >> 16)) \
+ | ((new) & ((new) >> 16))))
+
+static int mul_force_wake_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 old, new;
+ u32 ack_reg_offset;
+
+ old = vgpu_vreg(vgpu, offset);
+ new = CALC_MODE_MASK_REG(old, *(u32 *)p_data);
+
+ if (GRAPHICS_VER(vgpu->gvt->gt->i915) >= 9) {
+ switch (offset) {
+ case FORCEWAKE_RENDER_GEN9_REG:
+ ack_reg_offset = FORCEWAKE_ACK_RENDER_GEN9_REG;
+ break;
+ case FORCEWAKE_GT_GEN9_REG:
+ ack_reg_offset = FORCEWAKE_ACK_GT_GEN9_REG;
+ break;
+ case FORCEWAKE_MEDIA_GEN9_REG:
+ ack_reg_offset = FORCEWAKE_ACK_MEDIA_GEN9_REG;
+ break;
+ default:
+ /*should not hit here*/
+ gvt_vgpu_err("invalid forcewake offset 0x%x\n", offset);
+ return -EINVAL;
+ }
+ } else {
+ ack_reg_offset = FORCEWAKE_ACK_HSW_REG;
+ }
+
+ vgpu_vreg(vgpu, offset) = new;
+ vgpu_vreg(vgpu, ack_reg_offset) = (new & GENMASK(15, 0));
+ return 0;
+}
+
+static int gdrst_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ intel_engine_mask_t engine_mask = 0;
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if (data & GEN6_GRDOM_FULL) {
+ gvt_dbg_mmio("vgpu%d: request full GPU reset\n", vgpu->id);
+ engine_mask = ALL_ENGINES;
+ } else {
+ if (data & GEN6_GRDOM_RENDER) {
+ gvt_dbg_mmio("vgpu%d: request RCS reset\n", vgpu->id);
+ engine_mask |= BIT(RCS0);
+ }
+ if (data & GEN6_GRDOM_MEDIA) {
+ gvt_dbg_mmio("vgpu%d: request VCS reset\n", vgpu->id);
+ engine_mask |= BIT(VCS0);
+ }
+ if (data & GEN6_GRDOM_BLT) {
+ gvt_dbg_mmio("vgpu%d: request BCS Reset\n", vgpu->id);
+ engine_mask |= BIT(BCS0);
+ }
+ if (data & GEN6_GRDOM_VECS) {
+ gvt_dbg_mmio("vgpu%d: request VECS Reset\n", vgpu->id);
+ engine_mask |= BIT(VECS0);
+ }
+ if (data & GEN8_GRDOM_MEDIA2) {
+ gvt_dbg_mmio("vgpu%d: request VCS2 Reset\n", vgpu->id);
+ engine_mask |= BIT(VCS1);
+ }
+ if (data & GEN9_GRDOM_GUC) {
+ gvt_dbg_mmio("vgpu%d: request GUC Reset\n", vgpu->id);
+ vgpu_vreg_t(vgpu, GUC_STATUS) |= GS_MIA_IN_RESET;
+ }
+ engine_mask &= vgpu->gvt->gt->info.engine_mask;
+ }
+
+ /* vgpu_lock already hold by emulate mmio r/w */
+ intel_gvt_reset_vgpu_locked(vgpu, false, engine_mask);
+
+ /* sw will wait for the device to ack the reset request */
+ vgpu_vreg(vgpu, offset) = 0;
+
+ return 0;
+}
+
+static int gmbus_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ return intel_gvt_i2c_handle_gmbus_read(vgpu, offset, p_data, bytes);
+}
+
+static int gmbus_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ return intel_gvt_i2c_handle_gmbus_write(vgpu, offset, p_data, bytes);
+}
+
+static int pch_pp_control_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & PANEL_POWER_ON) {
+ vgpu_vreg_t(vgpu, PCH_PP_STATUS) |= PP_ON;
+ vgpu_vreg_t(vgpu, PCH_PP_STATUS) |= PP_SEQUENCE_STATE_ON_IDLE;
+ vgpu_vreg_t(vgpu, PCH_PP_STATUS) &= ~PP_SEQUENCE_POWER_DOWN;
+ vgpu_vreg_t(vgpu, PCH_PP_STATUS) &= ~PP_CYCLE_DELAY_ACTIVE;
+
+ } else
+ vgpu_vreg_t(vgpu, PCH_PP_STATUS) &=
+ ~(PP_ON | PP_SEQUENCE_POWER_DOWN
+ | PP_CYCLE_DELAY_ACTIVE);
+ return 0;
+}
+
+static int transconf_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & TRANS_ENABLE)
+ vgpu_vreg(vgpu, offset) |= TRANS_STATE_ENABLE;
+ else
+ vgpu_vreg(vgpu, offset) &= ~TRANS_STATE_ENABLE;
+ return 0;
+}
+
+static int lcpll_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & LCPLL_PLL_DISABLE)
+ vgpu_vreg(vgpu, offset) &= ~LCPLL_PLL_LOCK;
+ else
+ vgpu_vreg(vgpu, offset) |= LCPLL_PLL_LOCK;
+
+ if (vgpu_vreg(vgpu, offset) & LCPLL_CD_SOURCE_FCLK)
+ vgpu_vreg(vgpu, offset) |= LCPLL_CD_SOURCE_FCLK_DONE;
+ else
+ vgpu_vreg(vgpu, offset) &= ~LCPLL_CD_SOURCE_FCLK_DONE;
+
+ return 0;
+}
+
+static int dpy_reg_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ switch (offset) {
+ case 0xe651c:
+ case 0xe661c:
+ case 0xe671c:
+ case 0xe681c:
+ vgpu_vreg(vgpu, offset) = 1 << 17;
+ break;
+ case 0xe6c04:
+ vgpu_vreg(vgpu, offset) = 0x3;
+ break;
+ case 0xe6e1c:
+ vgpu_vreg(vgpu, offset) = 0x2f << 16;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ read_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+/*
+ * Only PIPE_A is enabled in current vGPU display and PIPE_A is tied to
+ * TRANSCODER_A in HW. DDI/PORT could be PORT_x depends on
+ * setup_virtual_dp_monitor().
+ * emulate_monitor_status_change() set up PLL for PORT_x as the initial enabled
+ * DPLL. Later guest driver may setup a different DPLLx when setting mode.
+ * So the correct sequence to find DP stream clock is:
+ * Check TRANS_DDI_FUNC_CTL on TRANSCODER_A to get PORT_x.
+ * Check correct PLLx for PORT_x to get PLL frequency and DP bitrate.
+ * Then Refresh rate then can be calculated based on follow equations:
+ * Pixel clock = h_total * v_total * refresh_rate
+ * stream clock = Pixel clock
+ * ls_clk = DP bitrate
+ * Link M/N = strm_clk / ls_clk
+ */
+
+static u32 bdw_vgpu_get_dp_bitrate(struct intel_vgpu *vgpu, enum port port)
+{
+ u32 dp_br = 0;
+ u32 ddi_pll_sel = vgpu_vreg_t(vgpu, PORT_CLK_SEL(port));
+
+ switch (ddi_pll_sel) {
+ case PORT_CLK_SEL_LCPLL_2700:
+ dp_br = 270000 * 2;
+ break;
+ case PORT_CLK_SEL_LCPLL_1350:
+ dp_br = 135000 * 2;
+ break;
+ case PORT_CLK_SEL_LCPLL_810:
+ dp_br = 81000 * 2;
+ break;
+ case PORT_CLK_SEL_SPLL:
+ {
+ switch (vgpu_vreg_t(vgpu, SPLL_CTL) & SPLL_FREQ_MASK) {
+ case SPLL_FREQ_810MHz:
+ dp_br = 81000 * 2;
+ break;
+ case SPLL_FREQ_1350MHz:
+ dp_br = 135000 * 2;
+ break;
+ case SPLL_FREQ_2700MHz:
+ dp_br = 270000 * 2;
+ break;
+ default:
+ gvt_dbg_dpy("vgpu-%d PORT_%c can't get freq from SPLL 0x%08x\n",
+ vgpu->id, port_name(port), vgpu_vreg_t(vgpu, SPLL_CTL));
+ break;
+ }
+ break;
+ }
+ case PORT_CLK_SEL_WRPLL1:
+ case PORT_CLK_SEL_WRPLL2:
+ {
+ u32 wrpll_ctl;
+ int refclk, n, p, r;
+
+ if (ddi_pll_sel == PORT_CLK_SEL_WRPLL1)
+ wrpll_ctl = vgpu_vreg_t(vgpu, WRPLL_CTL(DPLL_ID_WRPLL1));
+ else
+ wrpll_ctl = vgpu_vreg_t(vgpu, WRPLL_CTL(DPLL_ID_WRPLL2));
+
+ switch (wrpll_ctl & WRPLL_REF_MASK) {
+ case WRPLL_REF_PCH_SSC:
+ refclk = vgpu->gvt->gt->i915->display.dpll.ref_clks.ssc;
+ break;
+ case WRPLL_REF_LCPLL:
+ refclk = 2700000;
+ break;
+ default:
+ gvt_dbg_dpy("vgpu-%d PORT_%c WRPLL can't get refclk 0x%08x\n",
+ vgpu->id, port_name(port), wrpll_ctl);
+ goto out;
+ }
+
+ r = wrpll_ctl & WRPLL_DIVIDER_REF_MASK;
+ p = (wrpll_ctl & WRPLL_DIVIDER_POST_MASK) >> WRPLL_DIVIDER_POST_SHIFT;
+ n = (wrpll_ctl & WRPLL_DIVIDER_FB_MASK) >> WRPLL_DIVIDER_FB_SHIFT;
+
+ dp_br = (refclk * n / 10) / (p * r) * 2;
+ break;
+ }
+ default:
+ gvt_dbg_dpy("vgpu-%d PORT_%c has invalid clock select 0x%08x\n",
+ vgpu->id, port_name(port), vgpu_vreg_t(vgpu, PORT_CLK_SEL(port)));
+ break;
+ }
+
+out:
+ return dp_br;
+}
+
+static u32 bxt_vgpu_get_dp_bitrate(struct intel_vgpu *vgpu, enum port port)
+{
+ u32 dp_br = 0;
+ int refclk = vgpu->gvt->gt->i915->display.dpll.ref_clks.nssc;
+ enum dpio_phy phy = DPIO_PHY0;
+ enum dpio_channel ch = DPIO_CH0;
+ struct dpll clock = {0};
+ u32 temp;
+
+ /* Port to PHY mapping is fixed, see bxt_ddi_phy_info{} */
+ switch (port) {
+ case PORT_A:
+ phy = DPIO_PHY1;
+ ch = DPIO_CH0;
+ break;
+ case PORT_B:
+ phy = DPIO_PHY0;
+ ch = DPIO_CH0;
+ break;
+ case PORT_C:
+ phy = DPIO_PHY0;
+ ch = DPIO_CH1;
+ break;
+ default:
+ gvt_dbg_dpy("vgpu-%d no PHY for PORT_%c\n", vgpu->id, port_name(port));
+ goto out;
+ }
+
+ temp = vgpu_vreg_t(vgpu, BXT_PORT_PLL_ENABLE(port));
+ if (!(temp & PORT_PLL_ENABLE) || !(temp & PORT_PLL_LOCK)) {
+ gvt_dbg_dpy("vgpu-%d PORT_%c PLL_ENABLE 0x%08x isn't enabled or locked\n",
+ vgpu->id, port_name(port), temp);
+ goto out;
+ }
+
+ clock.m1 = 2;
+ clock.m2 = REG_FIELD_GET(PORT_PLL_M2_INT_MASK,
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL(phy, ch, 0))) << 22;
+ if (vgpu_vreg_t(vgpu, BXT_PORT_PLL(phy, ch, 3)) & PORT_PLL_M2_FRAC_ENABLE)
+ clock.m2 |= REG_FIELD_GET(PORT_PLL_M2_FRAC_MASK,
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL(phy, ch, 2)));
+ clock.n = REG_FIELD_GET(PORT_PLL_N_MASK,
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL(phy, ch, 1)));
+ clock.p1 = REG_FIELD_GET(PORT_PLL_P1_MASK,
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_EBB_0(phy, ch)));
+ clock.p2 = REG_FIELD_GET(PORT_PLL_P2_MASK,
+ vgpu_vreg_t(vgpu, BXT_PORT_PLL_EBB_0(phy, ch)));
+ clock.m = clock.m1 * clock.m2;
+ clock.p = clock.p1 * clock.p2 * 5;
+
+ if (clock.n == 0 || clock.p == 0) {
+ gvt_dbg_dpy("vgpu-%d PORT_%c PLL has invalid divider\n", vgpu->id, port_name(port));
+ goto out;
+ }
+
+ clock.vco = DIV_ROUND_CLOSEST_ULL(mul_u32_u32(refclk, clock.m), clock.n << 22);
+ clock.dot = DIV_ROUND_CLOSEST(clock.vco, clock.p);
+
+ dp_br = clock.dot;
+
+out:
+ return dp_br;
+}
+
+static u32 skl_vgpu_get_dp_bitrate(struct intel_vgpu *vgpu, enum port port)
+{
+ u32 dp_br = 0;
+ enum intel_dpll_id dpll_id = DPLL_ID_SKL_DPLL0;
+
+ /* Find the enabled DPLL for the DDI/PORT */
+ if (!(vgpu_vreg_t(vgpu, DPLL_CTRL2) & DPLL_CTRL2_DDI_CLK_OFF(port)) &&
+ (vgpu_vreg_t(vgpu, DPLL_CTRL2) & DPLL_CTRL2_DDI_SEL_OVERRIDE(port))) {
+ dpll_id += (vgpu_vreg_t(vgpu, DPLL_CTRL2) &
+ DPLL_CTRL2_DDI_CLK_SEL_MASK(port)) >>
+ DPLL_CTRL2_DDI_CLK_SEL_SHIFT(port);
+ } else {
+ gvt_dbg_dpy("vgpu-%d DPLL for PORT_%c isn't turned on\n",
+ vgpu->id, port_name(port));
+ return dp_br;
+ }
+
+ /* Find PLL output frequency from correct DPLL, and get bir rate */
+ switch ((vgpu_vreg_t(vgpu, DPLL_CTRL1) &
+ DPLL_CTRL1_LINK_RATE_MASK(dpll_id)) >>
+ DPLL_CTRL1_LINK_RATE_SHIFT(dpll_id)) {
+ case DPLL_CTRL1_LINK_RATE_810:
+ dp_br = 81000 * 2;
+ break;
+ case DPLL_CTRL1_LINK_RATE_1080:
+ dp_br = 108000 * 2;
+ break;
+ case DPLL_CTRL1_LINK_RATE_1350:
+ dp_br = 135000 * 2;
+ break;
+ case DPLL_CTRL1_LINK_RATE_1620:
+ dp_br = 162000 * 2;
+ break;
+ case DPLL_CTRL1_LINK_RATE_2160:
+ dp_br = 216000 * 2;
+ break;
+ case DPLL_CTRL1_LINK_RATE_2700:
+ dp_br = 270000 * 2;
+ break;
+ default:
+ dp_br = 0;
+ gvt_dbg_dpy("vgpu-%d PORT_%c fail to get DPLL-%d freq\n",
+ vgpu->id, port_name(port), dpll_id);
+ }
+
+ return dp_br;
+}
+
+static void vgpu_update_refresh_rate(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ enum port port;
+ u32 dp_br, link_m, link_n, htotal, vtotal;
+
+ /* Find DDI/PORT assigned to TRANSCODER_A, expect B or D */
+ port = (vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &
+ TRANS_DDI_PORT_MASK) >> TRANS_DDI_PORT_SHIFT;
+ if (port != PORT_B && port != PORT_D) {
+ gvt_dbg_dpy("vgpu-%d unsupported PORT_%c\n", vgpu->id, port_name(port));
+ return;
+ }
+
+ /* Calculate DP bitrate from PLL */
+ if (IS_BROADWELL(dev_priv))
+ dp_br = bdw_vgpu_get_dp_bitrate(vgpu, port);
+ else if (IS_BROXTON(dev_priv))
+ dp_br = bxt_vgpu_get_dp_bitrate(vgpu, port);
+ else
+ dp_br = skl_vgpu_get_dp_bitrate(vgpu, port);
+
+ /* Get DP link symbol clock M/N */
+ link_m = vgpu_vreg_t(vgpu, PIPE_LINK_M1(TRANSCODER_A));
+ link_n = vgpu_vreg_t(vgpu, PIPE_LINK_N1(TRANSCODER_A));
+
+ /* Get H/V total from transcoder timing */
+ htotal = (vgpu_vreg_t(vgpu, HTOTAL(TRANSCODER_A)) >> TRANS_HTOTAL_SHIFT);
+ vtotal = (vgpu_vreg_t(vgpu, VTOTAL(TRANSCODER_A)) >> TRANS_VTOTAL_SHIFT);
+
+ if (dp_br && link_n && htotal && vtotal) {
+ u64 pixel_clk = 0;
+ u32 new_rate = 0;
+ u32 *old_rate = &(intel_vgpu_port(vgpu, vgpu->display.port_num)->vrefresh_k);
+
+ /* Calcuate pixel clock by (ls_clk * M / N) */
+ pixel_clk = div_u64(mul_u32_u32(link_m, dp_br), link_n);
+ pixel_clk *= MSEC_PER_SEC;
+
+ /* Calcuate refresh rate by (pixel_clk / (h_total * v_total)) */
+ new_rate = DIV64_U64_ROUND_CLOSEST(mul_u64_u32_shr(pixel_clk, MSEC_PER_SEC, 0), mul_u32_u32(htotal + 1, vtotal + 1));
+
+ if (*old_rate != new_rate)
+ *old_rate = new_rate;
+
+ gvt_dbg_dpy("vgpu-%d PIPE_%c refresh rate updated to %d\n",
+ vgpu->id, pipe_name(PIPE_A), new_rate);
+ }
+}
+
+static int pipeconf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if (data & PIPECONF_ENABLE) {
+ vgpu_vreg(vgpu, offset) |= PIPECONF_STATE_ENABLE;
+ vgpu_update_refresh_rate(vgpu);
+ vgpu_update_vblank_emulation(vgpu, true);
+ } else {
+ vgpu_vreg(vgpu, offset) &= ~PIPECONF_STATE_ENABLE;
+ vgpu_update_vblank_emulation(vgpu, false);
+ }
+ return 0;
+}
+
+/* sorted in ascending order */
+static i915_reg_t force_nonpriv_white_list[] = {
+ _MMIO(0xd80),
+ GEN9_CS_DEBUG_MODE1, //_MMIO(0x20ec)
+ GEN9_CTX_PREEMPT_REG,//_MMIO(0x2248)
+ CL_PRIMITIVES_COUNT, //_MMIO(0x2340)
+ PS_INVOCATION_COUNT, //_MMIO(0x2348)
+ PS_DEPTH_COUNT, //_MMIO(0x2350)
+ GEN8_CS_CHICKEN1,//_MMIO(0x2580)
+ _MMIO(0x2690),
+ _MMIO(0x2694),
+ _MMIO(0x2698),
+ _MMIO(0x2754),
+ _MMIO(0x28a0),
+ _MMIO(0x4de0),
+ _MMIO(0x4de4),
+ _MMIO(0x4dfc),
+ GEN7_COMMON_SLICE_CHICKEN1,//_MMIO(0x7010)
+ _MMIO(0x7014),
+ HDC_CHICKEN0,//_MMIO(0x7300)
+ GEN8_HDC_CHICKEN1,//_MMIO(0x7304)
+ _MMIO(0x7700),
+ _MMIO(0x7704),
+ _MMIO(0x7708),
+ _MMIO(0x770c),
+ _MMIO(0x83a8),
+ _MMIO(0xb110),
+ GEN8_L3SQCREG4,//_MMIO(0xb118)
+ _MMIO(0xe100),
+ _MMIO(0xe18c),
+ _MMIO(0xe48c),
+ _MMIO(0xe5f4),
+ _MMIO(0x64844),
+};
+
+/* a simple bsearch */
+static inline bool in_whitelist(u32 reg)
+{
+ int left = 0, right = ARRAY_SIZE(force_nonpriv_white_list);
+ i915_reg_t *array = force_nonpriv_white_list;
+
+ while (left < right) {
+ int mid = (left + right)/2;
+
+ if (reg > array[mid].reg)
+ left = mid + 1;
+ else if (reg < array[mid].reg)
+ right = mid;
+ else
+ return true;
+ }
+ return false;
+}
+
+static int force_nonpriv_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 reg_nonpriv = (*(u32 *)p_data) & REG_GENMASK(25, 2);
+ const struct intel_engine_cs *engine =
+ intel_gvt_render_mmio_to_engine(vgpu->gvt, offset);
+
+ if (bytes != 4 || !IS_ALIGNED(offset, bytes) || !engine) {
+ gvt_err("vgpu(%d) Invalid FORCE_NONPRIV offset %x(%dB)\n",
+ vgpu->id, offset, bytes);
+ return -EINVAL;
+ }
+
+ if (!in_whitelist(reg_nonpriv) &&
+ reg_nonpriv != i915_mmio_reg_offset(RING_NOPID(engine->mmio_base))) {
+ gvt_err("vgpu(%d) Invalid FORCE_NONPRIV write %x at offset %x\n",
+ vgpu->id, reg_nonpriv, offset);
+ } else
+ intel_vgpu_default_mmio_write(vgpu, offset, p_data, bytes);
+
+ return 0;
+}
+
+static int ddi_buf_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & DDI_BUF_CTL_ENABLE) {
+ vgpu_vreg(vgpu, offset) &= ~DDI_BUF_IS_IDLE;
+ } else {
+ vgpu_vreg(vgpu, offset) |= DDI_BUF_IS_IDLE;
+ if (offset == i915_mmio_reg_offset(DDI_BUF_CTL(PORT_E)))
+ vgpu_vreg_t(vgpu, DP_TP_STATUS(PORT_E))
+ &= ~DP_TP_STATUS_AUTOTRAIN_DONE;
+ }
+ return 0;
+}
+
+static int fdi_rx_iir_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ vgpu_vreg(vgpu, offset) &= ~*(u32 *)p_data;
+ return 0;
+}
+
+#define FDI_LINK_TRAIN_PATTERN1 0
+#define FDI_LINK_TRAIN_PATTERN2 1
+
+static int fdi_auto_training_started(struct intel_vgpu *vgpu)
+{
+ u32 ddi_buf_ctl = vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_E));
+ u32 rx_ctl = vgpu_vreg(vgpu, _FDI_RXA_CTL);
+ u32 tx_ctl = vgpu_vreg_t(vgpu, DP_TP_CTL(PORT_E));
+
+ if ((ddi_buf_ctl & DDI_BUF_CTL_ENABLE) &&
+ (rx_ctl & FDI_RX_ENABLE) &&
+ (rx_ctl & FDI_AUTO_TRAINING) &&
+ (tx_ctl & DP_TP_CTL_ENABLE) &&
+ (tx_ctl & DP_TP_CTL_FDI_AUTOTRAIN))
+ return 1;
+ else
+ return 0;
+}
+
+static int check_fdi_rx_train_status(struct intel_vgpu *vgpu,
+ enum pipe pipe, unsigned int train_pattern)
+{
+ i915_reg_t fdi_rx_imr, fdi_tx_ctl, fdi_rx_ctl;
+ unsigned int fdi_rx_check_bits, fdi_tx_check_bits;
+ unsigned int fdi_rx_train_bits, fdi_tx_train_bits;
+ unsigned int fdi_iir_check_bits;
+
+ fdi_rx_imr = FDI_RX_IMR(pipe);
+ fdi_tx_ctl = FDI_TX_CTL(pipe);
+ fdi_rx_ctl = FDI_RX_CTL(pipe);
+
+ if (train_pattern == FDI_LINK_TRAIN_PATTERN1) {
+ fdi_rx_train_bits = FDI_LINK_TRAIN_PATTERN_1_CPT;
+ fdi_tx_train_bits = FDI_LINK_TRAIN_PATTERN_1;
+ fdi_iir_check_bits = FDI_RX_BIT_LOCK;
+ } else if (train_pattern == FDI_LINK_TRAIN_PATTERN2) {
+ fdi_rx_train_bits = FDI_LINK_TRAIN_PATTERN_2_CPT;
+ fdi_tx_train_bits = FDI_LINK_TRAIN_PATTERN_2;
+ fdi_iir_check_bits = FDI_RX_SYMBOL_LOCK;
+ } else {
+ gvt_vgpu_err("Invalid train pattern %d\n", train_pattern);
+ return -EINVAL;
+ }
+
+ fdi_rx_check_bits = FDI_RX_ENABLE | fdi_rx_train_bits;
+ fdi_tx_check_bits = FDI_TX_ENABLE | fdi_tx_train_bits;
+
+ /* If imr bit has been masked */
+ if (vgpu_vreg_t(vgpu, fdi_rx_imr) & fdi_iir_check_bits)
+ return 0;
+
+ if (((vgpu_vreg_t(vgpu, fdi_tx_ctl) & fdi_tx_check_bits)
+ == fdi_tx_check_bits)
+ && ((vgpu_vreg_t(vgpu, fdi_rx_ctl) & fdi_rx_check_bits)
+ == fdi_rx_check_bits))
+ return 1;
+ else
+ return 0;
+}
+
+#define INVALID_INDEX (~0U)
+
+static unsigned int calc_index(unsigned int offset, unsigned int start,
+ unsigned int next, unsigned int end, i915_reg_t i915_end)
+{
+ unsigned int range = next - start;
+
+ if (!end)
+ end = i915_mmio_reg_offset(i915_end);
+ if (offset < start || offset > end)
+ return INVALID_INDEX;
+ offset -= start;
+ return offset / range;
+}
+
+#define FDI_RX_CTL_TO_PIPE(offset) \
+ calc_index(offset, _FDI_RXA_CTL, _FDI_RXB_CTL, 0, FDI_RX_CTL(PIPE_C))
+
+#define FDI_TX_CTL_TO_PIPE(offset) \
+ calc_index(offset, _FDI_TXA_CTL, _FDI_TXB_CTL, 0, FDI_TX_CTL(PIPE_C))
+
+#define FDI_RX_IMR_TO_PIPE(offset) \
+ calc_index(offset, _FDI_RXA_IMR, _FDI_RXB_IMR, 0, FDI_RX_IMR(PIPE_C))
+
+static int update_fdi_rx_iir_status(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ i915_reg_t fdi_rx_iir;
+ unsigned int index;
+ int ret;
+
+ if (FDI_RX_CTL_TO_PIPE(offset) != INVALID_INDEX)
+ index = FDI_RX_CTL_TO_PIPE(offset);
+ else if (FDI_TX_CTL_TO_PIPE(offset) != INVALID_INDEX)
+ index = FDI_TX_CTL_TO_PIPE(offset);
+ else if (FDI_RX_IMR_TO_PIPE(offset) != INVALID_INDEX)
+ index = FDI_RX_IMR_TO_PIPE(offset);
+ else {
+ gvt_vgpu_err("Unsupported registers %x\n", offset);
+ return -EINVAL;
+ }
+
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ fdi_rx_iir = FDI_RX_IIR(index);
+
+ ret = check_fdi_rx_train_status(vgpu, index, FDI_LINK_TRAIN_PATTERN1);
+ if (ret < 0)
+ return ret;
+ if (ret)
+ vgpu_vreg_t(vgpu, fdi_rx_iir) |= FDI_RX_BIT_LOCK;
+
+ ret = check_fdi_rx_train_status(vgpu, index, FDI_LINK_TRAIN_PATTERN2);
+ if (ret < 0)
+ return ret;
+ if (ret)
+ vgpu_vreg_t(vgpu, fdi_rx_iir) |= FDI_RX_SYMBOL_LOCK;
+
+ if (offset == _FDI_RXA_CTL)
+ if (fdi_auto_training_started(vgpu))
+ vgpu_vreg_t(vgpu, DP_TP_STATUS(PORT_E)) |=
+ DP_TP_STATUS_AUTOTRAIN_DONE;
+ return 0;
+}
+
+#define DP_TP_CTL_TO_PORT(offset) \
+ calc_index(offset, _DP_TP_CTL_A, _DP_TP_CTL_B, 0, DP_TP_CTL(PORT_E))
+
+static int dp_tp_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ i915_reg_t status_reg;
+ unsigned int index;
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ index = DP_TP_CTL_TO_PORT(offset);
+ data = (vgpu_vreg(vgpu, offset) & GENMASK(10, 8)) >> 8;
+ if (data == 0x2) {
+ status_reg = DP_TP_STATUS(index);
+ vgpu_vreg_t(vgpu, status_reg) |= (1 << 25);
+ }
+ return 0;
+}
+
+static int dp_tp_status_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 reg_val;
+ u32 sticky_mask;
+
+ reg_val = *((u32 *)p_data);
+ sticky_mask = GENMASK(27, 26) | (1 << 24);
+
+ vgpu_vreg(vgpu, offset) = (reg_val & ~sticky_mask) |
+ (vgpu_vreg(vgpu, offset) & sticky_mask);
+ vgpu_vreg(vgpu, offset) &= ~(reg_val & sticky_mask);
+ return 0;
+}
+
+static int pch_adpa_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if (data & ADPA_CRT_HOTPLUG_FORCE_TRIGGER)
+ vgpu_vreg(vgpu, offset) &= ~ADPA_CRT_HOTPLUG_FORCE_TRIGGER;
+ return 0;
+}
+
+static int south_chicken2_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if (data & FDI_MPHY_IOSFSB_RESET_CTL)
+ vgpu_vreg(vgpu, offset) |= FDI_MPHY_IOSFSB_RESET_STATUS;
+ else
+ vgpu_vreg(vgpu, offset) &= ~FDI_MPHY_IOSFSB_RESET_STATUS;
+ return 0;
+}
+
+#define DSPSURF_TO_PIPE(offset) \
+ calc_index(offset, _DSPASURF, _DSPBSURF, 0, DSPSURF(PIPE_C))
+
+static int pri_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ u32 pipe = DSPSURF_TO_PIPE(offset);
+ int event = SKL_FLIP_EVENT(pipe, PLANE_PRIMARY);
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ vgpu_vreg_t(vgpu, DSPSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+
+ vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+
+ if (vgpu_vreg_t(vgpu, DSPCNTR(pipe)) & PLANE_CTL_ASYNC_FLIP)
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ else
+ set_bit(event, vgpu->irq.flip_done_event[pipe]);
+
+ return 0;
+}
+
+#define SPRSURF_TO_PIPE(offset) \
+ calc_index(offset, _SPRA_SURF, _SPRB_SURF, 0, SPRSURF(PIPE_C))
+
+static int spr_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 pipe = SPRSURF_TO_PIPE(offset);
+ int event = SKL_FLIP_EVENT(pipe, PLANE_SPRITE0);
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ vgpu_vreg_t(vgpu, SPRSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+
+ if (vgpu_vreg_t(vgpu, SPRCTL(pipe)) & PLANE_CTL_ASYNC_FLIP)
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ else
+ set_bit(event, vgpu->irq.flip_done_event[pipe]);
+
+ return 0;
+}
+
+static int reg50080_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data,
+ unsigned int bytes)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ enum pipe pipe = REG_50080_TO_PIPE(offset);
+ enum plane_id plane = REG_50080_TO_PLANE(offset);
+ int event = SKL_FLIP_EVENT(pipe, plane);
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ if (plane == PLANE_PRIMARY) {
+ vgpu_vreg_t(vgpu, DSPSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+ vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+ } else {
+ vgpu_vreg_t(vgpu, SPRSURFLIVE(pipe)) = vgpu_vreg(vgpu, offset);
+ }
+
+ if ((vgpu_vreg(vgpu, offset) & REG50080_FLIP_TYPE_MASK) == REG50080_FLIP_TYPE_ASYNC)
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ else
+ set_bit(event, vgpu->irq.flip_done_event[pipe]);
+
+ return 0;
+}
+
+static int trigger_aux_channel_interrupt(struct intel_vgpu *vgpu,
+ unsigned int reg)
+{
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ enum intel_gvt_event_type event;
+
+ if (reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_A)))
+ event = AUX_CHANNEL_A;
+ else if (reg == _PCH_DPB_AUX_CH_CTL ||
+ reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_B)))
+ event = AUX_CHANNEL_B;
+ else if (reg == _PCH_DPC_AUX_CH_CTL ||
+ reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_C)))
+ event = AUX_CHANNEL_C;
+ else if (reg == _PCH_DPD_AUX_CH_CTL ||
+ reg == i915_mmio_reg_offset(DP_AUX_CH_CTL(AUX_CH_D)))
+ event = AUX_CHANNEL_D;
+ else {
+ drm_WARN_ON(&dev_priv->drm, true);
+ return -EINVAL;
+ }
+
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ return 0;
+}
+
+static int dp_aux_ch_ctl_trans_done(struct intel_vgpu *vgpu, u32 value,
+ unsigned int reg, int len, bool data_valid)
+{
+ /* mark transaction done */
+ value |= DP_AUX_CH_CTL_DONE;
+ value &= ~DP_AUX_CH_CTL_SEND_BUSY;
+ value &= ~DP_AUX_CH_CTL_RECEIVE_ERROR;
+
+ if (data_valid)
+ value &= ~DP_AUX_CH_CTL_TIME_OUT_ERROR;
+ else
+ value |= DP_AUX_CH_CTL_TIME_OUT_ERROR;
+
+ /* message size */
+ value &= ~(0xf << 20);
+ value |= (len << 20);
+ vgpu_vreg(vgpu, reg) = value;
+
+ if (value & DP_AUX_CH_CTL_INTERRUPT)
+ return trigger_aux_channel_interrupt(vgpu, reg);
+ return 0;
+}
+
+static void dp_aux_ch_ctl_link_training(struct intel_vgpu_dpcd_data *dpcd,
+ u8 t)
+{
+ if ((t & DPCD_TRAINING_PATTERN_SET_MASK) == DPCD_TRAINING_PATTERN_1) {
+ /* training pattern 1 for CR */
+ /* set LANE0_CR_DONE, LANE1_CR_DONE */
+ dpcd->data[DPCD_LANE0_1_STATUS] |= DPCD_LANES_CR_DONE;
+ /* set LANE2_CR_DONE, LANE3_CR_DONE */
+ dpcd->data[DPCD_LANE2_3_STATUS] |= DPCD_LANES_CR_DONE;
+ } else if ((t & DPCD_TRAINING_PATTERN_SET_MASK) ==
+ DPCD_TRAINING_PATTERN_2) {
+ /* training pattern 2 for EQ */
+ /* Set CHANNEL_EQ_DONE and SYMBOL_LOCKED for Lane0_1 */
+ dpcd->data[DPCD_LANE0_1_STATUS] |= DPCD_LANES_EQ_DONE;
+ dpcd->data[DPCD_LANE0_1_STATUS] |= DPCD_SYMBOL_LOCKED;
+ /* Set CHANNEL_EQ_DONE and SYMBOL_LOCKED for Lane2_3 */
+ dpcd->data[DPCD_LANE2_3_STATUS] |= DPCD_LANES_EQ_DONE;
+ dpcd->data[DPCD_LANE2_3_STATUS] |= DPCD_SYMBOL_LOCKED;
+ /* set INTERLANE_ALIGN_DONE */
+ dpcd->data[DPCD_LANE_ALIGN_STATUS_UPDATED] |=
+ DPCD_INTERLANE_ALIGN_DONE;
+ } else if ((t & DPCD_TRAINING_PATTERN_SET_MASK) ==
+ DPCD_LINK_TRAINING_DISABLED) {
+ /* finish link training */
+ /* set sink status as synchronized */
+ dpcd->data[DPCD_SINK_STATUS] = DPCD_SINK_IN_SYNC;
+ }
+}
+
+#define _REG_HSW_DP_AUX_CH_CTL(dp) \
+ ((dp) ? (_PCH_DPB_AUX_CH_CTL + ((dp)-1)*0x100) : 0x64010)
+
+#define _REG_SKL_DP_AUX_CH_CTL(dp) (0x64010 + (dp) * 0x100)
+
+#define OFFSET_TO_DP_AUX_PORT(offset) (((offset) & 0xF00) >> 8)
+
+#define dpy_is_valid_port(port) \
+ (((port) >= PORT_A) && ((port) < I915_MAX_PORTS))
+
+static int dp_aux_ch_ctl_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct intel_vgpu_display *display = &vgpu->display;
+ int msg, addr, ctrl, op, len;
+ int port_index = OFFSET_TO_DP_AUX_PORT(offset);
+ struct intel_vgpu_dpcd_data *dpcd = NULL;
+ struct intel_vgpu_port *port = NULL;
+ u32 data;
+
+ if (!dpy_is_valid_port(port_index)) {
+ gvt_vgpu_err("Unsupported DP port access!\n");
+ return 0;
+ }
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if ((GRAPHICS_VER(vgpu->gvt->gt->i915) >= 9)
+ && offset != _REG_SKL_DP_AUX_CH_CTL(port_index)) {
+ /* SKL DPB/C/D aux ctl register changed */
+ return 0;
+ } else if (IS_BROADWELL(vgpu->gvt->gt->i915) &&
+ offset != _REG_HSW_DP_AUX_CH_CTL(port_index)) {
+ /* write to the data registers */
+ return 0;
+ }
+
+ if (!(data & DP_AUX_CH_CTL_SEND_BUSY)) {
+ /* just want to clear the sticky bits */
+ vgpu_vreg(vgpu, offset) = 0;
+ return 0;
+ }
+
+ port = &display->ports[port_index];
+ dpcd = port->dpcd;
+
+ /* read out message from DATA1 register */
+ msg = vgpu_vreg(vgpu, offset + 4);
+ addr = (msg >> 8) & 0xffff;
+ ctrl = (msg >> 24) & 0xff;
+ len = msg & 0xff;
+ op = ctrl >> 4;
+
+ if (op == GVT_AUX_NATIVE_WRITE) {
+ int t;
+ u8 buf[16];
+
+ if ((addr + len + 1) >= DPCD_SIZE) {
+ /*
+ * Write request exceeds what we supported,
+ * DCPD spec: When a Source Device is writing a DPCD
+ * address not supported by the Sink Device, the Sink
+ * Device shall reply with AUX NACK and “M” equal to
+ * zero.
+ */
+
+ /* NAK the write */
+ vgpu_vreg(vgpu, offset + 4) = AUX_NATIVE_REPLY_NAK;
+ dp_aux_ch_ctl_trans_done(vgpu, data, offset, 2, true);
+ return 0;
+ }
+
+ /*
+ * Write request format: Headr (command + address + size) occupies
+ * 4 bytes, followed by (len + 1) bytes of data. See details at
+ * intel_dp_aux_transfer().
+ */
+ if ((len + 1 + 4) > AUX_BURST_SIZE) {
+ gvt_vgpu_err("dp_aux_header: len %d is too large\n", len);
+ return -EINVAL;
+ }
+
+ /* unpack data from vreg to buf */
+ for (t = 0; t < 4; t++) {
+ u32 r = vgpu_vreg(vgpu, offset + 8 + t * 4);
+
+ buf[t * 4] = (r >> 24) & 0xff;
+ buf[t * 4 + 1] = (r >> 16) & 0xff;
+ buf[t * 4 + 2] = (r >> 8) & 0xff;
+ buf[t * 4 + 3] = r & 0xff;
+ }
+
+ /* write to virtual DPCD */
+ if (dpcd && dpcd->data_valid) {
+ for (t = 0; t <= len; t++) {
+ int p = addr + t;
+
+ dpcd->data[p] = buf[t];
+ /* check for link training */
+ if (p == DPCD_TRAINING_PATTERN_SET)
+ dp_aux_ch_ctl_link_training(dpcd,
+ buf[t]);
+ }
+ }
+
+ /* ACK the write */
+ vgpu_vreg(vgpu, offset + 4) = 0;
+ dp_aux_ch_ctl_trans_done(vgpu, data, offset, 1,
+ dpcd && dpcd->data_valid);
+ return 0;
+ }
+
+ if (op == GVT_AUX_NATIVE_READ) {
+ int idx, i, ret = 0;
+
+ if ((addr + len + 1) >= DPCD_SIZE) {
+ /*
+ * read request exceeds what we supported
+ * DPCD spec: A Sink Device receiving a Native AUX CH
+ * read request for an unsupported DPCD address must
+ * reply with an AUX ACK and read data set equal to
+ * zero instead of replying with AUX NACK.
+ */
+
+ /* ACK the READ*/
+ vgpu_vreg(vgpu, offset + 4) = 0;
+ vgpu_vreg(vgpu, offset + 8) = 0;
+ vgpu_vreg(vgpu, offset + 12) = 0;
+ vgpu_vreg(vgpu, offset + 16) = 0;
+ vgpu_vreg(vgpu, offset + 20) = 0;
+
+ dp_aux_ch_ctl_trans_done(vgpu, data, offset, len + 2,
+ true);
+ return 0;
+ }
+
+ for (idx = 1; idx <= 5; idx++) {
+ /* clear the data registers */
+ vgpu_vreg(vgpu, offset + 4 * idx) = 0;
+ }
+
+ /*
+ * Read reply format: ACK (1 byte) plus (len + 1) bytes of data.
+ */
+ if ((len + 2) > AUX_BURST_SIZE) {
+ gvt_vgpu_err("dp_aux_header: len %d is too large\n", len);
+ return -EINVAL;
+ }
+
+ /* read from virtual DPCD to vreg */
+ /* first 4 bytes: [ACK][addr][addr+1][addr+2] */
+ if (dpcd && dpcd->data_valid) {
+ for (i = 1; i <= (len + 1); i++) {
+ int t;
+
+ t = dpcd->data[addr + i - 1];
+ t <<= (24 - 8 * (i % 4));
+ ret |= t;
+
+ if ((i % 4 == 3) || (i == (len + 1))) {
+ vgpu_vreg(vgpu, offset +
+ (i / 4 + 1) * 4) = ret;
+ ret = 0;
+ }
+ }
+ }
+ dp_aux_ch_ctl_trans_done(vgpu, data, offset, len + 2,
+ dpcd && dpcd->data_valid);
+ return 0;
+ }
+
+ /* i2c transaction starts */
+ intel_gvt_i2c_handle_aux_ch_write(vgpu, port_index, offset, p_data);
+
+ if (data & DP_AUX_CH_CTL_INTERRUPT)
+ trigger_aux_channel_interrupt(vgpu, offset);
+ return 0;
+}
+
+static int mbctl_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ *(u32 *)p_data &= (~GEN6_MBCTL_ENABLE_BOOT_FETCH);
+ write_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+static int vga_control_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ bool vga_disable;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ vga_disable = vgpu_vreg(vgpu, offset) & VGA_DISP_DISABLE;
+
+ gvt_dbg_core("vgpu%d: %s VGA mode\n", vgpu->id,
+ vga_disable ? "Disable" : "Enable");
+ return 0;
+}
+
+static u32 read_virtual_sbi_register(struct intel_vgpu *vgpu,
+ unsigned int sbi_offset)
+{
+ struct intel_vgpu_display *display = &vgpu->display;
+ int num = display->sbi.number;
+ int i;
+
+ for (i = 0; i < num; ++i)
+ if (display->sbi.registers[i].offset == sbi_offset)
+ break;
+
+ if (i == num)
+ return 0;
+
+ return display->sbi.registers[i].value;
+}
+
+static void write_virtual_sbi_register(struct intel_vgpu *vgpu,
+ unsigned int offset, u32 value)
+{
+ struct intel_vgpu_display *display = &vgpu->display;
+ int num = display->sbi.number;
+ int i;
+
+ for (i = 0; i < num; ++i) {
+ if (display->sbi.registers[i].offset == offset)
+ break;
+ }
+
+ if (i == num) {
+ if (num == SBI_REG_MAX) {
+ gvt_vgpu_err("SBI caching meets maximum limits\n");
+ return;
+ }
+ display->sbi.number++;
+ }
+
+ display->sbi.registers[i].offset = offset;
+ display->sbi.registers[i].value = value;
+}
+
+static int sbi_data_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ if (((vgpu_vreg_t(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
+ SBI_OPCODE_SHIFT) == SBI_CMD_CRRD) {
+ unsigned int sbi_offset = (vgpu_vreg_t(vgpu, SBI_ADDR) &
+ SBI_ADDR_OFFSET_MASK) >> SBI_ADDR_OFFSET_SHIFT;
+ vgpu_vreg(vgpu, offset) = read_virtual_sbi_register(vgpu,
+ sbi_offset);
+ }
+ read_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+static int sbi_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ data &= ~(SBI_STAT_MASK << SBI_STAT_SHIFT);
+ data |= SBI_READY;
+
+ data &= ~(SBI_RESPONSE_MASK << SBI_RESPONSE_SHIFT);
+ data |= SBI_RESPONSE_SUCCESS;
+
+ vgpu_vreg(vgpu, offset) = data;
+
+ if (((vgpu_vreg_t(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
+ SBI_OPCODE_SHIFT) == SBI_CMD_CRWR) {
+ unsigned int sbi_offset = (vgpu_vreg_t(vgpu, SBI_ADDR) &
+ SBI_ADDR_OFFSET_MASK) >> SBI_ADDR_OFFSET_SHIFT;
+
+ write_virtual_sbi_register(vgpu, sbi_offset,
+ vgpu_vreg_t(vgpu, SBI_DATA));
+ }
+ return 0;
+}
+
+#define _vgtif_reg(x) \
+ (VGT_PVINFO_PAGE + offsetof(struct vgt_if, x))
+
+static int pvinfo_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ bool invalid_read = false;
+
+ read_vreg(vgpu, offset, p_data, bytes);
+
+ switch (offset) {
+ case _vgtif_reg(magic) ... _vgtif_reg(vgt_id):
+ if (offset + bytes > _vgtif_reg(vgt_id) + 4)
+ invalid_read = true;
+ break;
+ case _vgtif_reg(avail_rs.mappable_gmadr.base) ...
+ _vgtif_reg(avail_rs.fence_num):
+ if (offset + bytes >
+ _vgtif_reg(avail_rs.fence_num) + 4)
+ invalid_read = true;
+ break;
+ case 0x78010: /* vgt_caps */
+ case 0x7881c:
+ break;
+ default:
+ invalid_read = true;
+ break;
+ }
+ if (invalid_read)
+ gvt_vgpu_err("invalid pvinfo read: [%x:%x] = %x\n",
+ offset, bytes, *(u32 *)p_data);
+ vgpu->pv_notified = true;
+ return 0;
+}
+
+static int handle_g2v_notification(struct intel_vgpu *vgpu, int notification)
+{
+ enum intel_gvt_gtt_type root_entry_type = GTT_TYPE_PPGTT_ROOT_L4_ENTRY;
+ struct intel_vgpu_mm *mm;
+ u64 *pdps;
+
+ pdps = (u64 *)&vgpu_vreg64_t(vgpu, vgtif_reg(pdp[0]));
+
+ switch (notification) {
+ case VGT_G2V_PPGTT_L3_PAGE_TABLE_CREATE:
+ root_entry_type = GTT_TYPE_PPGTT_ROOT_L3_ENTRY;
+ fallthrough;
+ case VGT_G2V_PPGTT_L4_PAGE_TABLE_CREATE:
+ mm = intel_vgpu_get_ppgtt_mm(vgpu, root_entry_type, pdps);
+ return PTR_ERR_OR_ZERO(mm);
+ case VGT_G2V_PPGTT_L3_PAGE_TABLE_DESTROY:
+ case VGT_G2V_PPGTT_L4_PAGE_TABLE_DESTROY:
+ return intel_vgpu_put_ppgtt_mm(vgpu, pdps);
+ case VGT_G2V_EXECLIST_CONTEXT_CREATE:
+ case VGT_G2V_EXECLIST_CONTEXT_DESTROY:
+ case 1: /* Remove this in guest driver. */
+ break;
+ default:
+ gvt_vgpu_err("Invalid PV notification %d\n", notification);
+ }
+ return 0;
+}
+
+static int send_display_ready_uevent(struct intel_vgpu *vgpu, int ready)
+{
+ struct kobject *kobj = &vgpu->gvt->gt->i915->drm.primary->kdev->kobj;
+ char *env[3] = {NULL, NULL, NULL};
+ char vmid_str[20];
+ char display_ready_str[20];
+
+ snprintf(display_ready_str, 20, "GVT_DISPLAY_READY=%d", ready);
+ env[0] = display_ready_str;
+
+ snprintf(vmid_str, 20, "VMID=%d", vgpu->id);
+ env[1] = vmid_str;
+
+ return kobject_uevent_env(kobj, KOBJ_ADD, env);
+}
+
+static int pvinfo_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 data = *(u32 *)p_data;
+ bool invalid_write = false;
+
+ switch (offset) {
+ case _vgtif_reg(display_ready):
+ send_display_ready_uevent(vgpu, data ? 1 : 0);
+ break;
+ case _vgtif_reg(g2v_notify):
+ handle_g2v_notification(vgpu, data);
+ break;
+ /* add xhot and yhot to handled list to avoid error log */
+ case _vgtif_reg(cursor_x_hot):
+ case _vgtif_reg(cursor_y_hot):
+ case _vgtif_reg(pdp[0].lo):
+ case _vgtif_reg(pdp[0].hi):
+ case _vgtif_reg(pdp[1].lo):
+ case _vgtif_reg(pdp[1].hi):
+ case _vgtif_reg(pdp[2].lo):
+ case _vgtif_reg(pdp[2].hi):
+ case _vgtif_reg(pdp[3].lo):
+ case _vgtif_reg(pdp[3].hi):
+ case _vgtif_reg(execlist_context_descriptor_lo):
+ case _vgtif_reg(execlist_context_descriptor_hi):
+ break;
+ case _vgtif_reg(rsv5[0])..._vgtif_reg(rsv5[3]):
+ invalid_write = true;
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_INSUFFICIENT_RESOURCE);
+ break;
+ default:
+ invalid_write = true;
+ gvt_vgpu_err("invalid pvinfo write offset %x bytes %x data %x\n",
+ offset, bytes, data);
+ break;
+ }
+
+ if (!invalid_write)
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ return 0;
+}
+
+static int pf_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ u32 val = *(u32 *)p_data;
+
+ if ((offset == _PS_1A_CTRL || offset == _PS_2A_CTRL ||
+ offset == _PS_1B_CTRL || offset == _PS_2B_CTRL ||
+ offset == _PS_1C_CTRL) && (val & PS_PLANE_SEL_MASK) != 0) {
+ drm_WARN_ONCE(&i915->drm, true,
+ "VM(%d): guest is trying to scaling a plane\n",
+ vgpu->id);
+ return 0;
+ }
+
+ return intel_vgpu_default_mmio_write(vgpu, offset, p_data, bytes);
+}
+
+static int power_well_ctl_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) &
+ HSW_PWR_WELL_CTL_REQ(HSW_PW_CTL_IDX_GLOBAL))
+ vgpu_vreg(vgpu, offset) |=
+ HSW_PWR_WELL_CTL_STATE(HSW_PW_CTL_IDX_GLOBAL);
+ else
+ vgpu_vreg(vgpu, offset) &=
+ ~HSW_PWR_WELL_CTL_STATE(HSW_PW_CTL_IDX_GLOBAL);
+ return 0;
+}
+
+static int gen9_dbuf_ctl_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & DBUF_POWER_REQUEST)
+ vgpu_vreg(vgpu, offset) |= DBUF_POWER_STATE;
+ else
+ vgpu_vreg(vgpu, offset) &= ~DBUF_POWER_STATE;
+
+ return 0;
+}
+
+static int fpga_dbg_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (vgpu_vreg(vgpu, offset) & FPGA_DBG_RM_NOCLAIM)
+ vgpu_vreg(vgpu, offset) &= ~FPGA_DBG_RM_NOCLAIM;
+ return 0;
+}
+
+static int dma_ctrl_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ u32 mode;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ mode = vgpu_vreg(vgpu, offset);
+
+ if (GFX_MODE_BIT_SET_IN_MASK(mode, START_DMA)) {
+ drm_WARN_ONCE(&i915->drm, 1,
+ "VM(%d): iGVT-g doesn't support GuC\n",
+ vgpu->id);
+ return 0;
+ }
+
+ return 0;
+}
+
+static int gen9_trtte_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ u32 trtte = *(u32 *)p_data;
+
+ if ((trtte & 1) && (trtte & (1 << 1)) == 0) {
+ drm_WARN(&i915->drm, 1,
+ "VM(%d): Use physical address for TRTT!\n",
+ vgpu->id);
+ return -EINVAL;
+ }
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ return 0;
+}
+
+static int gen9_trtt_chicken_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+static int dpll_status_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 v = 0;
+
+ if (vgpu_vreg(vgpu, 0x46010) & (1 << 31))
+ v |= (1 << 0);
+
+ if (vgpu_vreg(vgpu, 0x46014) & (1 << 31))
+ v |= (1 << 8);
+
+ if (vgpu_vreg(vgpu, 0x46040) & (1 << 31))
+ v |= (1 << 16);
+
+ if (vgpu_vreg(vgpu, 0x46060) & (1 << 31))
+ v |= (1 << 24);
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
+}
+
+static int mailbox_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 value = *(u32 *)p_data;
+ u32 cmd = value & 0xff;
+ u32 *data0 = &vgpu_vreg_t(vgpu, GEN6_PCODE_DATA);
+
+ switch (cmd) {
+ case GEN9_PCODE_READ_MEM_LATENCY:
+ if (IS_SKYLAKE(vgpu->gvt->gt->i915) ||
+ IS_KABYLAKE(vgpu->gvt->gt->i915) ||
+ IS_COFFEELAKE(vgpu->gvt->gt->i915) ||
+ IS_COMETLAKE(vgpu->gvt->gt->i915)) {
+ /**
+ * "Read memory latency" command on gen9.
+ * Below memory latency values are read
+ * from skylake platform.
+ */
+ if (!*data0)
+ *data0 = 0x1e1a1100;
+ else
+ *data0 = 0x61514b3d;
+ } else if (IS_BROXTON(vgpu->gvt->gt->i915)) {
+ /**
+ * "Read memory latency" command on gen9.
+ * Below memory latency values are read
+ * from Broxton MRB.
+ */
+ if (!*data0)
+ *data0 = 0x16080707;
+ else
+ *data0 = 0x16161616;
+ }
+ break;
+ case SKL_PCODE_CDCLK_CONTROL:
+ if (IS_SKYLAKE(vgpu->gvt->gt->i915) ||
+ IS_KABYLAKE(vgpu->gvt->gt->i915) ||
+ IS_COFFEELAKE(vgpu->gvt->gt->i915) ||
+ IS_COMETLAKE(vgpu->gvt->gt->i915))
+ *data0 = SKL_CDCLK_READY_FOR_CHANGE;
+ break;
+ case GEN6_PCODE_READ_RC6VIDS:
+ *data0 |= 0x1;
+ break;
+ }
+
+ gvt_dbg_core("VM(%d) write %x to mailbox, return data0 %x\n",
+ vgpu->id, value, *data0);
+ /**
+ * PCODE_READY clear means ready for pcode read/write,
+ * PCODE_ERROR_MASK clear means no error happened. In GVT-g we
+ * always emulate as pcode read/write success and ready for access
+ * anytime, since we don't touch real physical registers here.
+ */
+ value &= ~(GEN6_PCODE_READY | GEN6_PCODE_ERROR_MASK);
+ return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
+}
+
+static int hws_pga_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 value = *(u32 *)p_data;
+ const struct intel_engine_cs *engine =
+ intel_gvt_render_mmio_to_engine(vgpu->gvt, offset);
+
+ if (value != 0 &&
+ !intel_gvt_ggtt_validate_range(vgpu, value, I915_GTT_PAGE_SIZE)) {
+ gvt_vgpu_err("write invalid HWSP address, reg:0x%x, value:0x%x\n",
+ offset, value);
+ return -EINVAL;
+ }
+
+ /*
+ * Need to emulate all the HWSP register write to ensure host can
+ * update the VM CSB status correctly. Here listed registers can
+ * support BDW, SKL or other platforms with same HWSP registers.
+ */
+ if (unlikely(!engine)) {
+ gvt_vgpu_err("access unknown hardware status page register:0x%x\n",
+ offset);
+ return -EINVAL;
+ }
+ vgpu->hws_pga[engine->id] = value;
+ gvt_dbg_mmio("VM(%d) write: 0x%x to HWSP: 0x%x\n",
+ vgpu->id, value, offset);
+
+ return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
+}
+
+static int skl_power_well_ctl_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ if (IS_BROXTON(vgpu->gvt->gt->i915))
+ v &= (1 << 31) | (1 << 29);
+ else
+ v &= (1 << 31) | (1 << 29) | (1 << 9) |
+ (1 << 7) | (1 << 5) | (1 << 3) | (1 << 1);
+ v |= (v >> 1);
+
+ return intel_vgpu_default_mmio_write(vgpu, offset, &v, bytes);
+}
+
+static int skl_lcpll_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ /* other bits are MBZ. */
+ v &= (1 << 31) | (1 << 30);
+ v & (1 << 31) ? (v |= (1 << 30)) : (v &= ~(1 << 30));
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int bxt_de_pll_enable_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ if (v & BXT_DE_PLL_PLL_ENABLE)
+ v |= BXT_DE_PLL_LOCK;
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int bxt_port_pll_enable_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ if (v & PORT_PLL_ENABLE)
+ v |= PORT_PLL_LOCK;
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int bxt_phy_ctl_family_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+ u32 data = v & COMMON_RESET_DIS ? BXT_PHY_LANE_ENABLED : 0;
+
+ switch (offset) {
+ case _PHY_CTL_FAMILY_EDP:
+ vgpu_vreg(vgpu, _BXT_PHY_CTL_DDI_A) = data;
+ break;
+ case _PHY_CTL_FAMILY_DDI:
+ vgpu_vreg(vgpu, _BXT_PHY_CTL_DDI_B) = data;
+ vgpu_vreg(vgpu, _BXT_PHY_CTL_DDI_C) = data;
+ break;
+ }
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int bxt_port_tx_dw3_read(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = vgpu_vreg(vgpu, offset);
+
+ v &= ~UNIQUE_TRANGE_EN_METHOD;
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
+}
+
+static int bxt_pcs_dw12_grp_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ if (offset == _PORT_PCS_DW12_GRP_A || offset == _PORT_PCS_DW12_GRP_B) {
+ vgpu_vreg(vgpu, offset - 0x600) = v;
+ vgpu_vreg(vgpu, offset - 0x800) = v;
+ } else {
+ vgpu_vreg(vgpu, offset - 0x400) = v;
+ vgpu_vreg(vgpu, offset - 0x600) = v;
+ }
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int bxt_gt_disp_pwron_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 v = *(u32 *)p_data;
+
+ if (v & BIT(0)) {
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) &=
+ ~PHY_RESERVED;
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) |=
+ PHY_POWER_GOOD;
+ }
+
+ if (v & BIT(1)) {
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY1)) &=
+ ~PHY_RESERVED;
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY1)) |=
+ PHY_POWER_GOOD;
+ }
+
+
+ vgpu_vreg(vgpu, offset) = v;
+
+ return 0;
+}
+
+static int edp_psr_imr_iir_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ vgpu_vreg(vgpu, offset) = 0;
+ return 0;
+}
+
+/*
+ * FixMe:
+ * If guest fills non-priv batch buffer on ApolloLake/Broxton as Mesa i965 did:
+ * 717e7539124d (i965: Use a WC map and memcpy for the batch instead of pwrite.)
+ * Due to the missing flush of bb filled by VM vCPU, host GPU hangs on executing
+ * these MI_BATCH_BUFFER.
+ * Temporarily workaround this by setting SNOOP bit for PAT3 used by PPGTT
+ * PML4 PTE: PAT(0) PCD(1) PWT(1).
+ * The performance is still expected to be low, will need further improvement.
+ */
+static int bxt_ppat_low_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u64 pat =
+ GEN8_PPAT(0, CHV_PPAT_SNOOP) |
+ GEN8_PPAT(1, 0) |
+ GEN8_PPAT(2, 0) |
+ GEN8_PPAT(3, CHV_PPAT_SNOOP) |
+ GEN8_PPAT(4, CHV_PPAT_SNOOP) |
+ GEN8_PPAT(5, CHV_PPAT_SNOOP) |
+ GEN8_PPAT(6, CHV_PPAT_SNOOP) |
+ GEN8_PPAT(7, CHV_PPAT_SNOOP);
+
+ vgpu_vreg(vgpu, offset) = lower_32_bits(pat);
+
+ return 0;
+}
+
+static int guc_status_read(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data,
+ unsigned int bytes)
+{
+ /* keep MIA_IN_RESET before clearing */
+ read_vreg(vgpu, offset, p_data, bytes);
+ vgpu_vreg(vgpu, offset) &= ~GS_MIA_IN_RESET;
+ return 0;
+}
+
+static int mmio_read_from_hw(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_engine_cs *engine =
+ intel_gvt_render_mmio_to_engine(gvt, offset);
+
+ /**
+ * Read HW reg in following case
+ * a. the offset isn't a ring mmio
+ * b. the offset's ring is running on hw.
+ * c. the offset is ring time stamp mmio
+ */
+
+ if (!engine ||
+ vgpu == gvt->scheduler.engine_owner[engine->id] ||
+ offset == i915_mmio_reg_offset(RING_TIMESTAMP(engine->mmio_base)) ||
+ offset == i915_mmio_reg_offset(RING_TIMESTAMP_UDW(engine->mmio_base))) {
+ mmio_hw_access_pre(gvt->gt);
+ vgpu_vreg(vgpu, offset) =
+ intel_uncore_read(gvt->gt->uncore, _MMIO(offset));
+ mmio_hw_access_post(gvt->gt);
+ }
+
+ return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
+}
+
+static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ const struct intel_engine_cs *engine = intel_gvt_render_mmio_to_engine(vgpu->gvt, offset);
+ struct intel_vgpu_execlist *execlist;
+ u32 data = *(u32 *)p_data;
+ int ret = 0;
+
+ if (drm_WARN_ON(&i915->drm, !engine))
+ return -EINVAL;
+
+ /*
+ * Due to d3_entered is used to indicate skipping PPGTT invalidation on
+ * vGPU reset, it's set on D0->D3 on PCI config write, and cleared after
+ * vGPU reset if in resuming.
+ * In S0ix exit, the device power state also transite from D3 to D0 as
+ * S3 resume, but no vGPU reset (triggered by QEMU devic model). After
+ * S0ix exit, all engines continue to work. However the d3_entered
+ * remains set which will break next vGPU reset logic (miss the expected
+ * PPGTT invalidation).
+ * Engines can only work in D0. Thus the 1st elsp write gives GVT a
+ * chance to clear d3_entered.
+ */
+ if (vgpu->d3_entered)
+ vgpu->d3_entered = false;
+
+ execlist = &vgpu->submission.execlist[engine->id];
+
+ execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data;
+ if (execlist->elsp_dwords.index == 3) {
+ ret = intel_vgpu_submit_execlist(vgpu, engine);
+ if(ret)
+ gvt_vgpu_err("fail submit workload on ring %s\n",
+ engine->name);
+ }
+
+ ++execlist->elsp_dwords.index;
+ execlist->elsp_dwords.index &= 0x3;
+ return ret;
+}
+
+static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 data = *(u32 *)p_data;
+ const struct intel_engine_cs *engine =
+ intel_gvt_render_mmio_to_engine(vgpu->gvt, offset);
+ bool enable_execlist;
+ int ret;
+
+ (*(u32 *)p_data) &= ~_MASKED_BIT_ENABLE(1);
+ if (IS_COFFEELAKE(vgpu->gvt->gt->i915) ||
+ IS_COMETLAKE(vgpu->gvt->gt->i915))
+ (*(u32 *)p_data) &= ~_MASKED_BIT_ENABLE(2);
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (IS_MASKED_BITS_ENABLED(data, 1)) {
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_UNSUPPORTED_GUEST);
+ return 0;
+ }
+
+ if ((IS_COFFEELAKE(vgpu->gvt->gt->i915) ||
+ IS_COMETLAKE(vgpu->gvt->gt->i915)) &&
+ IS_MASKED_BITS_ENABLED(data, 2)) {
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_UNSUPPORTED_GUEST);
+ return 0;
+ }
+
+ /* when PPGTT mode enabled, we will check if guest has called
+ * pvinfo, if not, we will treat this guest as non-gvtg-aware
+ * guest, and stop emulating its cfg space, mmio, gtt, etc.
+ */
+ if ((IS_MASKED_BITS_ENABLED(data, GFX_PPGTT_ENABLE) ||
+ IS_MASKED_BITS_ENABLED(data, GFX_RUN_LIST_ENABLE)) &&
+ !vgpu->pv_notified) {
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_UNSUPPORTED_GUEST);
+ return 0;
+ }
+ if (IS_MASKED_BITS_ENABLED(data, GFX_RUN_LIST_ENABLE) ||
+ IS_MASKED_BITS_DISABLED(data, GFX_RUN_LIST_ENABLE)) {
+ enable_execlist = !!(data & GFX_RUN_LIST_ENABLE);
+
+ gvt_dbg_core("EXECLIST %s on ring %s\n",
+ (enable_execlist ? "enabling" : "disabling"),
+ engine->name);
+
+ if (!enable_execlist)
+ return 0;
+
+ ret = intel_vgpu_select_submission_ops(vgpu,
+ engine->mask,
+ INTEL_VGPU_EXECLIST_SUBMISSION);
+ if (ret)
+ return ret;
+
+ intel_vgpu_start_schedule(vgpu);
+ }
+ return 0;
+}
+
+static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ unsigned int id = 0;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ vgpu_vreg(vgpu, offset) = 0;
+
+ switch (offset) {
+ case 0x4260:
+ id = RCS0;
+ break;
+ case 0x4264:
+ id = VCS0;
+ break;
+ case 0x4268:
+ id = VCS1;
+ break;
+ case 0x426c:
+ id = BCS0;
+ break;
+ case 0x4270:
+ id = VECS0;
+ break;
+ default:
+ return -EINVAL;
+ }
+ set_bit(id, (void *)vgpu->submission.tlb_handle_pending);
+
+ return 0;
+}
+
+static int ring_reset_ctl_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data, unsigned int bytes)
+{
+ u32 data;
+
+ write_vreg(vgpu, offset, p_data, bytes);
+ data = vgpu_vreg(vgpu, offset);
+
+ if (IS_MASKED_BITS_ENABLED(data, RESET_CTL_REQUEST_RESET))
+ data |= RESET_CTL_READY_TO_RESET;
+ else if (data & _MASKED_BIT_DISABLE(RESET_CTL_REQUEST_RESET))
+ data &= ~RESET_CTL_READY_TO_RESET;
+
+ vgpu_vreg(vgpu, offset) = data;
+ return 0;
+}
+
+static int csfe_chicken1_mmio_write(struct intel_vgpu *vgpu,
+ unsigned int offset, void *p_data,
+ unsigned int bytes)
+{
+ u32 data = *(u32 *)p_data;
+
+ (*(u32 *)p_data) &= ~_MASKED_BIT_ENABLE(0x18);
+ write_vreg(vgpu, offset, p_data, bytes);
+
+ if (IS_MASKED_BITS_ENABLED(data, 0x10) ||
+ IS_MASKED_BITS_ENABLED(data, 0x8))
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_UNSUPPORTED_GUEST);
+
+ return 0;
+}
+
+#define MMIO_F(reg, s, f, am, rm, d, r, w) do { \
+ ret = setup_mmio_info(gvt, i915_mmio_reg_offset(reg), \
+ s, f, am, rm, d, r, w); \
+ if (ret) \
+ return ret; \
+} while (0)
+
+#define MMIO_DH(reg, d, r, w) \
+ MMIO_F(reg, 4, 0, 0, 0, d, r, w)
+
+#define MMIO_DFH(reg, d, f, r, w) \
+ MMIO_F(reg, 4, f, 0, 0, d, r, w)
+
+#define MMIO_GM(reg, d, r, w) \
+ MMIO_F(reg, 4, F_GMADR, 0xFFFFF000, 0, d, r, w)
+
+#define MMIO_GM_RDR(reg, d, r, w) \
+ MMIO_F(reg, 4, F_GMADR | F_CMD_ACCESS, 0xFFFFF000, 0, d, r, w)
+
+#define MMIO_RO(reg, d, f, rm, r, w) \
+ MMIO_F(reg, 4, F_RO | f, 0, rm, d, r, w)
+
+#define MMIO_RING_F(prefix, s, f, am, rm, d, r, w) do { \
+ MMIO_F(prefix(RENDER_RING_BASE), s, f, am, rm, d, r, w); \
+ MMIO_F(prefix(BLT_RING_BASE), s, f, am, rm, d, r, w); \
+ MMIO_F(prefix(GEN6_BSD_RING_BASE), s, f, am, rm, d, r, w); \
+ MMIO_F(prefix(VEBOX_RING_BASE), s, f, am, rm, d, r, w); \
+ if (HAS_ENGINE(gvt->gt, VCS1)) \
+ MMIO_F(prefix(GEN8_BSD2_RING_BASE), s, f, am, rm, d, r, w); \
+} while (0)
+
+#define MMIO_RING_DFH(prefix, d, f, r, w) \
+ MMIO_RING_F(prefix, 4, f, 0, 0, d, r, w)
+
+#define MMIO_RING_GM(prefix, d, r, w) \
+ MMIO_RING_F(prefix, 4, F_GMADR, 0xFFFF0000, 0, d, r, w)
+
+#define MMIO_RING_GM_RDR(prefix, d, r, w) \
+ MMIO_RING_F(prefix, 4, F_GMADR | F_CMD_ACCESS, 0xFFFF0000, 0, d, r, w)
+
+#define MMIO_RING_RO(prefix, d, f, rm, r, w) \
+ MMIO_RING_F(prefix, 4, F_RO | f, 0, rm, d, r, w)
+
+static int init_generic_mmio_info(struct intel_gvt *gvt)
+{
+ struct drm_i915_private *dev_priv = gvt->gt->i915;
+ int ret;
+
+ MMIO_RING_DFH(RING_IMR, D_ALL, 0, NULL,
+ intel_vgpu_reg_imr_handler);
+
+ MMIO_DFH(SDEIMR, D_ALL, 0, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DFH(SDEIER, D_ALL, 0, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DFH(SDEIIR, D_ALL, 0, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_RING_DFH(RING_HWSTAM, D_ALL, 0, NULL, NULL);
+
+
+ MMIO_DH(GEN8_GAMW_ECO_DEV_RW_IA, D_BDW_PLUS, NULL,
+ gamw_echo_dev_rw_ia_write);
+
+ MMIO_GM_RDR(BSD_HWS_PGA_GEN7, D_ALL, NULL, NULL);
+ MMIO_GM_RDR(BLT_HWS_PGA_GEN7, D_ALL, NULL, NULL);
+ MMIO_GM_RDR(VEBOX_HWS_PGA_GEN7, D_ALL, NULL, NULL);
+
+#define RING_REG(base) _MMIO((base) + 0x28)
+ MMIO_RING_DFH(RING_REG, D_ALL, F_CMD_ACCESS, NULL, NULL);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x134)
+ MMIO_RING_DFH(RING_REG, D_ALL, F_CMD_ACCESS, NULL, NULL);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x6c)
+ MMIO_RING_DFH(RING_REG, D_ALL, 0, mmio_read_from_hw, NULL);
+#undef RING_REG
+ MMIO_DH(GEN7_SC_INSTDONE, D_BDW_PLUS, mmio_read_from_hw, NULL);
+
+ MMIO_GM_RDR(_MMIO(0x2148), D_ALL, NULL, NULL);
+ MMIO_GM_RDR(CCID(RENDER_RING_BASE), D_ALL, NULL, NULL);
+ MMIO_GM_RDR(_MMIO(0x12198), D_ALL, NULL, NULL);
+
+ MMIO_RING_DFH(RING_TAIL, D_ALL, 0, NULL, NULL);
+ MMIO_RING_DFH(RING_HEAD, D_ALL, 0, NULL, NULL);
+ MMIO_RING_DFH(RING_CTL, D_ALL, 0, NULL, NULL);
+ MMIO_RING_DFH(RING_ACTHD, D_ALL, 0, mmio_read_from_hw, NULL);
+ MMIO_RING_GM(RING_START, D_ALL, NULL, NULL);
+
+ /* RING MODE */
+#define RING_REG(base) _MMIO((base) + 0x29c)
+ MMIO_RING_DFH(RING_REG, D_ALL,
+ F_MODE_MASK | F_CMD_ACCESS | F_CMD_WRITE_PATCH, NULL,
+ ring_mode_mmio_write);
+#undef RING_REG
+
+ MMIO_RING_DFH(RING_MI_MODE, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_RING_DFH(RING_INSTPM, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_RING_DFH(RING_TIMESTAMP, D_ALL, F_CMD_ACCESS,
+ mmio_read_from_hw, NULL);
+ MMIO_RING_DFH(RING_TIMESTAMP_UDW, D_ALL, F_CMD_ACCESS,
+ mmio_read_from_hw, NULL);
+
+ MMIO_DFH(GEN7_GT_MODE, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(CACHE_MODE_0_GEN7, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(CACHE_MODE_1, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(CACHE_MODE_0, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2124), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0x20dc), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_3D_CHICKEN3, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2088), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(FF_SLICE_CS_CHICKEN2, D_ALL,
+ F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2470), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GAM_ECOCHK, D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GEN7_COMMON_SLICE_CHICKEN1, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(COMMON_SLICE_CHICKEN2, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(_MMIO(0x9030), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x20a0), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2420), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2430), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2434), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2438), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x243c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x7018), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(HALF_SLICE_CHICKEN3, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GEN7_HALF_SLICE_CHICKEN1, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+
+ /* display */
+ MMIO_DH(PIPECONF(PIPE_A), D_ALL, NULL, pipeconf_mmio_write);
+ MMIO_DH(PIPECONF(PIPE_B), D_ALL, NULL, pipeconf_mmio_write);
+ MMIO_DH(PIPECONF(PIPE_C), D_ALL, NULL, pipeconf_mmio_write);
+ MMIO_DH(PIPECONF(_PIPE_EDP), D_ALL, NULL, pipeconf_mmio_write);
+ MMIO_DH(DSPSURF(PIPE_A), D_ALL, NULL, pri_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_A, PLANE_PRIMARY), D_ALL, NULL,
+ reg50080_mmio_write);
+ MMIO_DH(DSPSURF(PIPE_B), D_ALL, NULL, pri_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_B, PLANE_PRIMARY), D_ALL, NULL,
+ reg50080_mmio_write);
+ MMIO_DH(DSPSURF(PIPE_C), D_ALL, NULL, pri_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_C, PLANE_PRIMARY), D_ALL, NULL,
+ reg50080_mmio_write);
+ MMIO_DH(SPRSURF(PIPE_A), D_ALL, NULL, spr_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_A, PLANE_SPRITE0), D_ALL, NULL,
+ reg50080_mmio_write);
+ MMIO_DH(SPRSURF(PIPE_B), D_ALL, NULL, spr_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_B, PLANE_SPRITE0), D_ALL, NULL,
+ reg50080_mmio_write);
+ MMIO_DH(SPRSURF(PIPE_C), D_ALL, NULL, spr_surf_mmio_write);
+ MMIO_DH(REG_50080(PIPE_C, PLANE_SPRITE0), D_ALL, NULL,
+ reg50080_mmio_write);
+
+ MMIO_F(PCH_GMBUS0, 4 * 4, 0, 0, 0, D_ALL, gmbus_mmio_read,
+ gmbus_mmio_write);
+ MMIO_F(PCH_GPIO_BASE, 6 * 4, F_UNALIGN, 0, 0, D_ALL, NULL, NULL);
+
+ MMIO_F(_MMIO(_PCH_DPB_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+ dp_aux_ch_ctl_mmio_write);
+ MMIO_F(_MMIO(_PCH_DPC_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+ dp_aux_ch_ctl_mmio_write);
+ MMIO_F(_MMIO(_PCH_DPD_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+ dp_aux_ch_ctl_mmio_write);
+
+ MMIO_DH(PCH_ADPA, D_PRE_SKL, NULL, pch_adpa_mmio_write);
+
+ MMIO_DH(_MMIO(_PCH_TRANSACONF), D_ALL, NULL, transconf_mmio_write);
+ MMIO_DH(_MMIO(_PCH_TRANSBCONF), D_ALL, NULL, transconf_mmio_write);
+
+ MMIO_DH(FDI_RX_IIR(PIPE_A), D_ALL, NULL, fdi_rx_iir_mmio_write);
+ MMIO_DH(FDI_RX_IIR(PIPE_B), D_ALL, NULL, fdi_rx_iir_mmio_write);
+ MMIO_DH(FDI_RX_IIR(PIPE_C), D_ALL, NULL, fdi_rx_iir_mmio_write);
+ MMIO_DH(FDI_RX_IMR(PIPE_A), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(FDI_RX_IMR(PIPE_B), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(FDI_RX_IMR(PIPE_C), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(FDI_RX_CTL(PIPE_A), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(FDI_RX_CTL(PIPE_B), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(FDI_RX_CTL(PIPE_C), D_ALL, NULL, update_fdi_rx_iir_status);
+ MMIO_DH(PCH_PP_CONTROL, D_ALL, NULL, pch_pp_control_mmio_write);
+ MMIO_DH(_MMIO(0xe651c), D_ALL, dpy_reg_mmio_read, NULL);
+ MMIO_DH(_MMIO(0xe661c), D_ALL, dpy_reg_mmio_read, NULL);
+ MMIO_DH(_MMIO(0xe671c), D_ALL, dpy_reg_mmio_read, NULL);
+ MMIO_DH(_MMIO(0xe681c), D_ALL, dpy_reg_mmio_read, NULL);
+ MMIO_DH(_MMIO(0xe6c04), D_ALL, dpy_reg_mmio_read, NULL);
+ MMIO_DH(_MMIO(0xe6e1c), D_ALL, dpy_reg_mmio_read, NULL);
+
+ MMIO_RO(PCH_PORT_HOTPLUG, D_ALL, 0,
+ PORTA_HOTPLUG_STATUS_MASK
+ | PORTB_HOTPLUG_STATUS_MASK
+ | PORTC_HOTPLUG_STATUS_MASK
+ | PORTD_HOTPLUG_STATUS_MASK,
+ NULL, NULL);
+
+ MMIO_DH(LCPLL_CTL, D_ALL, NULL, lcpll_ctl_mmio_write);
+ MMIO_DH(SOUTH_CHICKEN2, D_ALL, NULL, south_chicken2_mmio_write);
+ MMIO_DH(SFUSE_STRAP, D_ALL, NULL, NULL);
+ MMIO_DH(SBI_DATA, D_ALL, sbi_data_mmio_read, NULL);
+ MMIO_DH(SBI_CTL_STAT, D_ALL, NULL, sbi_ctl_mmio_write);
+
+ MMIO_F(_MMIO(_DPA_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_ALL, NULL,
+ dp_aux_ch_ctl_mmio_write);
+
+ MMIO_DH(DDI_BUF_CTL(PORT_A), D_ALL, NULL, ddi_buf_ctl_mmio_write);
+ MMIO_DH(DDI_BUF_CTL(PORT_B), D_ALL, NULL, ddi_buf_ctl_mmio_write);
+ MMIO_DH(DDI_BUF_CTL(PORT_C), D_ALL, NULL, ddi_buf_ctl_mmio_write);
+ MMIO_DH(DDI_BUF_CTL(PORT_D), D_ALL, NULL, ddi_buf_ctl_mmio_write);
+ MMIO_DH(DDI_BUF_CTL(PORT_E), D_ALL, NULL, ddi_buf_ctl_mmio_write);
+
+ MMIO_DH(DP_TP_CTL(PORT_A), D_ALL, NULL, dp_tp_ctl_mmio_write);
+ MMIO_DH(DP_TP_CTL(PORT_B), D_ALL, NULL, dp_tp_ctl_mmio_write);
+ MMIO_DH(DP_TP_CTL(PORT_C), D_ALL, NULL, dp_tp_ctl_mmio_write);
+ MMIO_DH(DP_TP_CTL(PORT_D), D_ALL, NULL, dp_tp_ctl_mmio_write);
+ MMIO_DH(DP_TP_CTL(PORT_E), D_ALL, NULL, dp_tp_ctl_mmio_write);
+
+ MMIO_DH(DP_TP_STATUS(PORT_A), D_ALL, NULL, dp_tp_status_mmio_write);
+ MMIO_DH(DP_TP_STATUS(PORT_B), D_ALL, NULL, dp_tp_status_mmio_write);
+ MMIO_DH(DP_TP_STATUS(PORT_C), D_ALL, NULL, dp_tp_status_mmio_write);
+ MMIO_DH(DP_TP_STATUS(PORT_D), D_ALL, NULL, dp_tp_status_mmio_write);
+ MMIO_DH(DP_TP_STATUS(PORT_E), D_ALL, NULL, NULL);
+
+ MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_A), D_ALL, NULL, NULL);
+ MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_B), D_ALL, NULL, NULL);
+ MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_C), D_ALL, NULL, NULL);
+ MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_EDP), D_ALL, NULL, NULL);
+
+ MMIO_DH(FORCEWAKE, D_ALL, NULL, NULL);
+ MMIO_DFH(GTFIFODBG, D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GTFIFOCTL, D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DH(FORCEWAKE_MT, D_PRE_SKL, NULL, mul_force_wake_write);
+ MMIO_DH(FORCEWAKE_ACK_HSW, D_BDW, NULL, NULL);
+ MMIO_DH(GEN6_RC_CONTROL, D_ALL, NULL, NULL);
+ MMIO_DH(GEN6_RC_STATE, D_ALL, NULL, NULL);
+ MMIO_DH(HSW_PWR_WELL_CTL1, D_BDW, NULL, power_well_ctl_mmio_write);
+ MMIO_DH(HSW_PWR_WELL_CTL2, D_BDW, NULL, power_well_ctl_mmio_write);
+ MMIO_DH(HSW_PWR_WELL_CTL3, D_BDW, NULL, power_well_ctl_mmio_write);
+ MMIO_DH(HSW_PWR_WELL_CTL4, D_BDW, NULL, power_well_ctl_mmio_write);
+ MMIO_DH(HSW_PWR_WELL_CTL5, D_BDW, NULL, power_well_ctl_mmio_write);
+ MMIO_DH(HSW_PWR_WELL_CTL6, D_BDW, NULL, power_well_ctl_mmio_write);
+
+ MMIO_DH(GEN6_GDRST, D_ALL, NULL, gdrst_mmio_write);
+ MMIO_F(FENCE_REG_GEN6_LO(0), 0x80, 0, 0, 0, D_ALL, fence_mmio_read, fence_mmio_write);
+ MMIO_DH(CPU_VGACNTRL, D_ALL, NULL, vga_control_mmio_write);
+
+ MMIO_DH(GEN7_ERR_INT, D_ALL, NULL, NULL);
+ MMIO_DH(GFX_FLSH_CNTL_GEN6, D_ALL, NULL, NULL);
+
+ MMIO_DH(GEN6_MBCTL, D_ALL, NULL, mbctl_write);
+ MMIO_DFH(GEN7_UCGCTL4, D_ALL, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DH(FPGA_DBG, D_ALL, NULL, fpga_dbg_mmio_write);
+ MMIO_DFH(_MMIO(0x215c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2178), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x217c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x12178), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x1217c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_F(_MMIO(0x2290), 8, F_CMD_ACCESS, 0, 0, D_BDW_PLUS, NULL, NULL);
+ MMIO_F(_MMIO(0x5200), 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(_MMIO(0x5240), 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(_MMIO(0x5280), 16, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0x1c17c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x1c178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(BCS_SWCTRL, D_ALL, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_F(HS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(DS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(IA_VERTICES_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(IA_PRIMITIVES_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(VS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(GS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(GS_PRIMITIVES_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(CL_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(CL_PRIMITIVES_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(PS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_F(PS_DEPTH_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+ MMIO_DH(_MMIO(0x4260), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+ MMIO_DH(_MMIO(0x4264), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+ MMIO_DH(_MMIO(0x4268), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+ MMIO_DH(_MMIO(0x426c), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+ MMIO_DH(_MMIO(0x4270), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+ MMIO_DFH(_MMIO(0x4094), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(ARB_MODE, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_RING_GM(RING_BBADDR, D_ALL, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2220), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x12220), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x22220), D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_RING_DFH(RING_SYNC_1, D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_RING_DFH(RING_SYNC_0, D_ALL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x22178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x1a178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x1a17c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2217c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DH(EDP_PSR_IMR, D_BDW_PLUS, NULL, edp_psr_imr_iir_write);
+ MMIO_DH(EDP_PSR_IIR, D_BDW_PLUS, NULL, edp_psr_imr_iir_write);
+ MMIO_DH(GUC_STATUS, D_ALL, guc_status_read, NULL);
+
+ return 0;
+}
+
+static int init_bdw_mmio_info(struct intel_gvt *gvt)
+{
+ int ret;
+
+ MMIO_DH(GEN8_GT_IMR(0), D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_GT_IER(0), D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_GT_IIR(0), D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_GT_IMR(1), D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_GT_IER(1), D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_GT_IIR(1), D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_GT_IMR(2), D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_GT_IER(2), D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_GT_IIR(2), D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_GT_IMR(3), D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_GT_IER(3), D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_GT_IIR(3), D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_DE_PIPE_IMR(PIPE_A), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_DE_PIPE_IER(PIPE_A), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_DE_PIPE_IIR(PIPE_A), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_DE_PIPE_IMR(PIPE_B), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_DE_PIPE_IER(PIPE_B), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_DE_PIPE_IIR(PIPE_B), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_DE_PIPE_IMR(PIPE_C), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_DE_PIPE_IER(PIPE_C), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_DE_PIPE_IIR(PIPE_C), D_BDW_PLUS, NULL,
+ intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_DE_PORT_IMR, D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_DE_PORT_IER, D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_DE_PORT_IIR, D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_DE_MISC_IMR, D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_DE_MISC_IER, D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_DE_MISC_IIR, D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_PCU_IMR, D_BDW_PLUS, NULL, intel_vgpu_reg_imr_handler);
+ MMIO_DH(GEN8_PCU_IER, D_BDW_PLUS, NULL, intel_vgpu_reg_ier_handler);
+ MMIO_DH(GEN8_PCU_IIR, D_BDW_PLUS, NULL, intel_vgpu_reg_iir_handler);
+
+ MMIO_DH(GEN8_MASTER_IRQ, D_BDW_PLUS, NULL,
+ intel_vgpu_reg_master_irq_handler);
+
+ MMIO_RING_DFH(RING_ACTHD_UDW, D_BDW_PLUS, 0,
+ mmio_read_from_hw, NULL);
+
+#define RING_REG(base) _MMIO((base) + 0xd0)
+ MMIO_RING_F(RING_REG, 4, F_RO, 0,
+ ~_MASKED_BIT_ENABLE(RESET_CTL_REQUEST_RESET), D_BDW_PLUS, NULL,
+ ring_reset_ctl_write);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x230)
+ MMIO_RING_DFH(RING_REG, D_BDW_PLUS, 0, NULL, elsp_mmio_write);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x234)
+ MMIO_RING_F(RING_REG, 8, F_RO, 0, ~0, D_BDW_PLUS,
+ NULL, NULL);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x244)
+ MMIO_RING_DFH(RING_REG, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x370)
+ MMIO_RING_F(RING_REG, 48, F_RO, 0, ~0, D_BDW_PLUS, NULL, NULL);
+#undef RING_REG
+
+#define RING_REG(base) _MMIO((base) + 0x3a0)
+ MMIO_RING_DFH(RING_REG, D_BDW_PLUS, F_MODE_MASK, NULL, NULL);
+#undef RING_REG
+
+ MMIO_DH(GEN6_PCODE_MAILBOX, D_BDW_PLUS, NULL, mailbox_write);
+
+#define RING_REG(base) _MMIO((base) + 0x270)
+ MMIO_RING_F(RING_REG, 32, F_CMD_ACCESS, 0, 0, D_BDW_PLUS, NULL, NULL);
+#undef RING_REG
+
+ MMIO_RING_GM(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write);
+
+ MMIO_DFH(HDC_CHICKEN0, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(GEN8_ROW_CHICKEN, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(GEN7_ROW_CHICKEN2, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(GEN8_UCGCTL6, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0xb1f0), D_BDW, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xb1c0), D_BDW, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GEN8_L3SQCREG4, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xb100), D_BDW, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xb10c), D_BDW, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_F(_MMIO(0x24d0), 48, F_CMD_ACCESS | F_CMD_WRITE_PATCH, 0, 0,
+ D_BDW_PLUS, NULL, force_nonpriv_write);
+
+ MMIO_DFH(_MMIO(0x83a4), D_BDW, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0x8430), D_BDW, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0xe194), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe188), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(HALF_SLICE_CHICKEN2, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x2580), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0x2248), D_BDW, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DFH(_MMIO(0xe220), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe230), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe240), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe260), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe270), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe280), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe2a0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe2b0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe2c0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x21f0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+ return 0;
+}
+
+static int init_skl_mmio_info(struct intel_gvt *gvt)
+{
+ struct drm_i915_private *dev_priv = gvt->gt->i915;
+ int ret;
+
+ MMIO_DH(FORCEWAKE_RENDER_GEN9, D_SKL_PLUS, NULL, mul_force_wake_write);
+ MMIO_DH(FORCEWAKE_ACK_RENDER_GEN9, D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(FORCEWAKE_GT_GEN9, D_SKL_PLUS, NULL, mul_force_wake_write);
+ MMIO_DH(FORCEWAKE_ACK_GT_GEN9, D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(FORCEWAKE_MEDIA_GEN9, D_SKL_PLUS, NULL, mul_force_wake_write);
+ MMIO_DH(FORCEWAKE_ACK_MEDIA_GEN9, D_SKL_PLUS, NULL, NULL);
+
+ MMIO_F(DP_AUX_CH_CTL(AUX_CH_B), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+ dp_aux_ch_ctl_mmio_write);
+ MMIO_F(DP_AUX_CH_CTL(AUX_CH_C), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+ dp_aux_ch_ctl_mmio_write);
+ MMIO_F(DP_AUX_CH_CTL(AUX_CH_D), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+ dp_aux_ch_ctl_mmio_write);
+
+ MMIO_DH(HSW_PWR_WELL_CTL2, D_SKL_PLUS, NULL, skl_power_well_ctl_write);
+
+ MMIO_DH(DBUF_CTL_S(0), D_SKL_PLUS, NULL, gen9_dbuf_ctl_mmio_write);
+
+ MMIO_DFH(GEN9_GAMT_ECO_REG_RW_IA, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(MMCD_MISC_CTRL, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DH(CHICKEN_PAR1_1, D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(LCPLL1_CTL, D_SKL_PLUS, NULL, skl_lcpll_write);
+ MMIO_DH(LCPLL2_CTL, D_SKL_PLUS, NULL, skl_lcpll_write);
+ MMIO_DH(DPLL_STATUS, D_SKL_PLUS, dpll_status_read, NULL);
+
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_A, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_A, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_B, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_B, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_C, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_POS(PIPE_C, 1), D_SKL_PLUS, NULL, pf_write);
+
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_A, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_A, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_B, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_B, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_C, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_WIN_SZ(PIPE_C, 1), D_SKL_PLUS, NULL, pf_write);
+
+ MMIO_DH(SKL_PS_CTRL(PIPE_A, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_CTRL(PIPE_A, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_CTRL(PIPE_B, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_CTRL(PIPE_B, 1), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_CTRL(PIPE_C, 0), D_SKL_PLUS, NULL, pf_write);
+ MMIO_DH(SKL_PS_CTRL(PIPE_C, 1), D_SKL_PLUS, NULL, pf_write);
+
+ MMIO_DH(PLANE_BUF_CFG(PIPE_A, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_A, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_A, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_A, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_BUF_CFG(PIPE_B, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_B, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_B, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_B, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_BUF_CFG(PIPE_C, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_C, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_BUF_CFG(PIPE_C, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(CUR_BUF_CFG(PIPE_A), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(CUR_BUF_CFG(PIPE_B), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(CUR_BUF_CFG(PIPE_C), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_WM_TRANS(PIPE_A, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_A, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_A, 2), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_WM_TRANS(PIPE_B, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_B, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_B, 2), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_WM_TRANS(PIPE_C, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_C, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_WM_TRANS(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(CUR_WM_TRANS(PIPE_A), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(CUR_WM_TRANS(PIPE_B), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(CUR_WM_TRANS(PIPE_C), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_A, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_A, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_A, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_A, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_B, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_B, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_B, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_B, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 0), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 1), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 3), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 1)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 2)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 3)), D_SKL_PLUS, NULL, NULL);
+ MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 4)), D_SKL_PLUS, NULL, NULL);
+
+ MMIO_DFH(BDW_SCRATCH1, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_F(GEN9_GFX_MOCS(0), 0x7f8, F_CMD_ACCESS, 0, 0, D_SKL_PLUS,
+ NULL, NULL);
+ MMIO_F(GEN7_L3CNTLREG2, 0x80, F_CMD_ACCESS, 0, 0, D_SKL_PLUS,
+ NULL, NULL);
+
+ MMIO_DFH(GEN7_FF_SLICE_CS_CHICKEN1, D_SKL_PLUS,
+ F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GEN9_CS_DEBUG_MODE1, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+
+ /* TRTT */
+ MMIO_DFH(TRVATTL3PTRDW(0), D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(TRVATTL3PTRDW(1), D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(TRVATTL3PTRDW(2), D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(TRVATTL3PTRDW(3), D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(TRVADR, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(TRTTE, D_SKL_PLUS, F_CMD_ACCESS | F_PM_SAVE,
+ NULL, gen9_trtte_write);
+ MMIO_DFH(_MMIO(0x4dfc), D_SKL_PLUS, F_PM_SAVE,
+ NULL, gen9_trtt_chicken_write);
+
+ MMIO_DFH(GEN8_GARBCNTL, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DH(DMA_CTRL, D_SKL_PLUS, NULL, dma_ctrl_write);
+
+#define CSFE_CHICKEN1_REG(base) _MMIO((base) + 0xD4)
+ MMIO_RING_DFH(CSFE_CHICKEN1_REG, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, csfe_chicken1_mmio_write);
+#undef CSFE_CHICKEN1_REG
+ MMIO_DFH(GEN8_HDC_CHICKEN1, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+ MMIO_DFH(GEN9_WM_CHICKEN3, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS,
+ NULL, NULL);
+
+ MMIO_DFH(GAMT_CHKN_BIT_REG, D_KBL | D_CFL, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0xe4cc), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+
+ return 0;
+}
+
+static int init_bxt_mmio_info(struct intel_gvt *gvt)
+{
+ int ret;
+
+ MMIO_DH(BXT_P_CR_GT_DISP_PWRON, D_BXT, NULL, bxt_gt_disp_pwron_write);
+ MMIO_DH(BXT_PHY_CTL_FAMILY(DPIO_PHY0), D_BXT,
+ NULL, bxt_phy_ctl_family_write);
+ MMIO_DH(BXT_PHY_CTL_FAMILY(DPIO_PHY1), D_BXT,
+ NULL, bxt_phy_ctl_family_write);
+ MMIO_DH(BXT_PORT_PLL_ENABLE(PORT_A), D_BXT,
+ NULL, bxt_port_pll_enable_write);
+ MMIO_DH(BXT_PORT_PLL_ENABLE(PORT_B), D_BXT,
+ NULL, bxt_port_pll_enable_write);
+ MMIO_DH(BXT_PORT_PLL_ENABLE(PORT_C), D_BXT, NULL,
+ bxt_port_pll_enable_write);
+
+ MMIO_DH(BXT_PORT_PCS_DW12_GRP(DPIO_PHY0, DPIO_CH0), D_BXT,
+ NULL, bxt_pcs_dw12_grp_write);
+ MMIO_DH(BXT_PORT_TX_DW3_LN0(DPIO_PHY0, DPIO_CH0), D_BXT,
+ bxt_port_tx_dw3_read, NULL);
+ MMIO_DH(BXT_PORT_PCS_DW12_GRP(DPIO_PHY0, DPIO_CH1), D_BXT,
+ NULL, bxt_pcs_dw12_grp_write);
+ MMIO_DH(BXT_PORT_TX_DW3_LN0(DPIO_PHY0, DPIO_CH1), D_BXT,
+ bxt_port_tx_dw3_read, NULL);
+ MMIO_DH(BXT_PORT_PCS_DW12_GRP(DPIO_PHY1, DPIO_CH0), D_BXT,
+ NULL, bxt_pcs_dw12_grp_write);
+ MMIO_DH(BXT_PORT_TX_DW3_LN0(DPIO_PHY1, DPIO_CH0), D_BXT,
+ bxt_port_tx_dw3_read, NULL);
+ MMIO_DH(BXT_DE_PLL_ENABLE, D_BXT, NULL, bxt_de_pll_enable_write);
+ MMIO_DFH(GEN8_L3SQCREG1, D_BXT, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(GEN8_L3CNTLREG, D_BXT, F_CMD_ACCESS, NULL, NULL);
+ MMIO_DFH(_MMIO(0x20D8), D_BXT, F_CMD_ACCESS, NULL, NULL);
+ MMIO_F(GEN8_RING_CS_GPR(RENDER_RING_BASE, 0), 0x40, F_CMD_ACCESS,
+ 0, 0, D_BXT, NULL, NULL);
+ MMIO_F(GEN8_RING_CS_GPR(GEN6_BSD_RING_BASE, 0), 0x40, F_CMD_ACCESS,
+ 0, 0, D_BXT, NULL, NULL);
+ MMIO_F(GEN8_RING_CS_GPR(BLT_RING_BASE, 0), 0x40, F_CMD_ACCESS,
+ 0, 0, D_BXT, NULL, NULL);
+ MMIO_F(GEN8_RING_CS_GPR(VEBOX_RING_BASE, 0), 0x40, F_CMD_ACCESS,
+ 0, 0, D_BXT, NULL, NULL);
+
+ MMIO_DFH(GEN9_CTX_PREEMPT_REG, D_BXT, F_CMD_ACCESS, NULL, NULL);
+
+ MMIO_DH(GEN8_PRIVATE_PAT_LO, D_BXT, NULL, bxt_ppat_low_write);
+
+ return 0;
+}
+
+static struct gvt_mmio_block *find_mmio_block(struct intel_gvt *gvt,
+ unsigned int offset)
+{
+ struct gvt_mmio_block *block = gvt->mmio.mmio_block;
+ int num = gvt->mmio.num_mmio_block;
+ int i;
+
+ for (i = 0; i < num; i++, block++) {
+ if (offset >= i915_mmio_reg_offset(block->offset) &&
+ offset < i915_mmio_reg_offset(block->offset) + block->size)
+ return block;
+ }
+ return NULL;
+}
+
+/**
+ * intel_gvt_clean_mmio_info - clean up MMIO information table for GVT device
+ * @gvt: GVT device
+ *
+ * This function is called at the driver unloading stage, to clean up the MMIO
+ * information table of GVT device
+ *
+ */
+void intel_gvt_clean_mmio_info(struct intel_gvt *gvt)
+{
+ struct hlist_node *tmp;
+ struct intel_gvt_mmio_info *e;
+ int i;
+
+ hash_for_each_safe(gvt->mmio.mmio_info_table, i, tmp, e, node)
+ kfree(e);
+
+ kfree(gvt->mmio.mmio_block);
+ gvt->mmio.mmio_block = NULL;
+ gvt->mmio.num_mmio_block = 0;
+
+ vfree(gvt->mmio.mmio_attribute);
+ gvt->mmio.mmio_attribute = NULL;
+}
+
+static int handle_mmio(struct intel_gvt_mmio_table_iter *iter, u32 offset,
+ u32 size)
+{
+ struct intel_gvt *gvt = iter->data;
+ struct intel_gvt_mmio_info *info, *p;
+ u32 start, end, i;
+
+ if (WARN_ON(!IS_ALIGNED(offset, 4)))
+ return -EINVAL;
+
+ start = offset;
+ end = offset + size;
+
+ for (i = start; i < end; i += 4) {
+ p = intel_gvt_find_mmio_info(gvt, i);
+ if (p) {
+ WARN(1, "dup mmio definition offset %x\n",
+ info->offset);
+
+ /* We return -EEXIST here to make GVT-g load fail.
+ * So duplicated MMIO can be found as soon as
+ * possible.
+ */
+ return -EEXIST;
+ }
+
+ info = kzalloc(sizeof(*info), GFP_KERNEL);
+ if (!info)
+ return -ENOMEM;
+
+ info->offset = i;
+ info->read = intel_vgpu_default_mmio_read;
+ info->write = intel_vgpu_default_mmio_write;
+ INIT_HLIST_NODE(&info->node);
+ hash_add(gvt->mmio.mmio_info_table, &info->node, info->offset);
+ gvt->mmio.num_tracked_mmio++;
+ }
+ return 0;
+}
+
+static int handle_mmio_block(struct intel_gvt_mmio_table_iter *iter,
+ u32 offset, u32 size)
+{
+ struct intel_gvt *gvt = iter->data;
+ struct gvt_mmio_block *block = gvt->mmio.mmio_block;
+ void *ret;
+
+ ret = krealloc(block,
+ (gvt->mmio.num_mmio_block + 1) * sizeof(*block),
+ GFP_KERNEL);
+ if (!ret)
+ return -ENOMEM;
+
+ gvt->mmio.mmio_block = block = ret;
+
+ block += gvt->mmio.num_mmio_block;
+
+ memset(block, 0, sizeof(*block));
+
+ block->offset = _MMIO(offset);
+ block->size = size;
+
+ gvt->mmio.num_mmio_block++;
+
+ return 0;
+}
+
+static int handle_mmio_cb(struct intel_gvt_mmio_table_iter *iter, u32 offset,
+ u32 size)
+{
+ if (size < 1024 || offset == i915_mmio_reg_offset(GEN9_GFX_MOCS(0)))
+ return handle_mmio(iter, offset, size);
+ else
+ return handle_mmio_block(iter, offset, size);
+}
+
+static int init_mmio_info(struct intel_gvt *gvt)
+{
+ struct intel_gvt_mmio_table_iter iter = {
+ .i915 = gvt->gt->i915,
+ .data = gvt,
+ .handle_mmio_cb = handle_mmio_cb,
+ };
+
+ return intel_gvt_iterate_mmio_table(&iter);
+}
+
+static int init_mmio_block_handlers(struct intel_gvt *gvt)
+{
+ struct gvt_mmio_block *block;
+
+ block = find_mmio_block(gvt, VGT_PVINFO_PAGE);
+ if (!block) {
+ WARN(1, "fail to assign handlers to mmio block %x\n",
+ i915_mmio_reg_offset(gvt->mmio.mmio_block->offset));
+ return -ENODEV;
+ }
+
+ block->read = pvinfo_mmio_read;
+ block->write = pvinfo_mmio_write;
+
+ return 0;
+}
+
+/**
+ * intel_gvt_setup_mmio_info - setup MMIO information table for GVT device
+ * @gvt: GVT device
+ *
+ * This function is called at the initialization stage, to setup the MMIO
+ * information table for GVT device
+ *
+ * Returns:
+ * zero on success, negative if failed.
+ */
+int intel_gvt_setup_mmio_info(struct intel_gvt *gvt)
+{
+ struct intel_gvt_device_info *info = &gvt->device_info;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ int size = info->mmio_size / 4 * sizeof(*gvt->mmio.mmio_attribute);
+ int ret;
+
+ gvt->mmio.mmio_attribute = vzalloc(size);
+ if (!gvt->mmio.mmio_attribute)
+ return -ENOMEM;
+
+ ret = init_mmio_info(gvt);
+ if (ret)
+ goto err;
+
+ ret = init_mmio_block_handlers(gvt);
+ if (ret)
+ goto err;
+
+ ret = init_generic_mmio_info(gvt);
+ if (ret)
+ goto err;
+
+ if (IS_BROADWELL(i915)) {
+ ret = init_bdw_mmio_info(gvt);
+ if (ret)
+ goto err;
+ } else if (IS_SKYLAKE(i915) ||
+ IS_KABYLAKE(i915) ||
+ IS_COFFEELAKE(i915) ||
+ IS_COMETLAKE(i915)) {
+ ret = init_bdw_mmio_info(gvt);
+ if (ret)
+ goto err;
+ ret = init_skl_mmio_info(gvt);
+ if (ret)
+ goto err;
+ } else if (IS_BROXTON(i915)) {
+ ret = init_bdw_mmio_info(gvt);
+ if (ret)
+ goto err;
+ ret = init_skl_mmio_info(gvt);
+ if (ret)
+ goto err;
+ ret = init_bxt_mmio_info(gvt);
+ if (ret)
+ goto err;
+ }
+
+ return 0;
+err:
+ intel_gvt_clean_mmio_info(gvt);
+ return ret;
+}
+
+/**
+ * intel_gvt_for_each_tracked_mmio - iterate each tracked mmio
+ * @gvt: a GVT device
+ * @handler: the handler
+ * @data: private data given to handler
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+ int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+ void *data)
+{
+ struct gvt_mmio_block *block = gvt->mmio.mmio_block;
+ struct intel_gvt_mmio_info *e;
+ int i, j, ret;
+
+ hash_for_each(gvt->mmio.mmio_info_table, i, e, node) {
+ ret = handler(gvt, e->offset, data);
+ if (ret)
+ return ret;
+ }
+
+ for (i = 0; i < gvt->mmio.num_mmio_block; i++, block++) {
+ /* pvinfo data doesn't come from hw mmio */
+ if (i915_mmio_reg_offset(block->offset) == VGT_PVINFO_PAGE)
+ continue;
+
+ for (j = 0; j < block->size; j += 4) {
+ ret = handler(gvt, i915_mmio_reg_offset(block->offset) + j, data);
+ if (ret)
+ return ret;
+ }
+ }
+ return 0;
+}
+
+/**
+ * intel_vgpu_default_mmio_read - default MMIO read handler
+ * @vgpu: a vGPU
+ * @offset: access offset
+ * @p_data: data return buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_default_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ read_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+/**
+ * intel_vgpu_default_mmio_write() - default MMIO write handler
+ * @vgpu: a vGPU
+ * @offset: access offset
+ * @p_data: write data buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_default_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ write_vreg(vgpu, offset, p_data, bytes);
+ return 0;
+}
+
+/**
+ * intel_vgpu_mask_mmio_write - write mask register
+ * @vgpu: a vGPU
+ * @offset: access offset
+ * @p_data: write data buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_mask_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes)
+{
+ u32 mask, old_vreg;
+
+ old_vreg = vgpu_vreg(vgpu, offset);
+ write_vreg(vgpu, offset, p_data, bytes);
+ mask = vgpu_vreg(vgpu, offset) >> 16;
+ vgpu_vreg(vgpu, offset) = (old_vreg & ~mask) |
+ (vgpu_vreg(vgpu, offset) & mask);
+
+ return 0;
+}
+
+/**
+ * intel_gvt_in_force_nonpriv_whitelist - if a mmio is in whitelist to be
+ * force-nopriv register
+ *
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if the register is in force-nonpriv whitelist;
+ * False if outside;
+ */
+bool intel_gvt_in_force_nonpriv_whitelist(struct intel_gvt *gvt,
+ unsigned int offset)
+{
+ return in_whitelist(offset);
+}
+
+/**
+ * intel_vgpu_mmio_reg_rw - emulate tracked mmio registers
+ * @vgpu: a vGPU
+ * @offset: register offset
+ * @pdata: data buffer
+ * @bytes: data length
+ * @is_read: read or write
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_mmio_reg_rw(struct intel_vgpu *vgpu, unsigned int offset,
+ void *pdata, unsigned int bytes, bool is_read)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_mmio_info *mmio_info;
+ struct gvt_mmio_block *mmio_block;
+ gvt_mmio_func func;
+ int ret;
+
+ if (drm_WARN_ON(&i915->drm, bytes > 8))
+ return -EINVAL;
+
+ /*
+ * Handle special MMIO blocks.
+ */
+ mmio_block = find_mmio_block(gvt, offset);
+ if (mmio_block) {
+ func = is_read ? mmio_block->read : mmio_block->write;
+ if (func)
+ return func(vgpu, offset, pdata, bytes);
+ goto default_rw;
+ }
+
+ /*
+ * Normal tracked MMIOs.
+ */
+ mmio_info = intel_gvt_find_mmio_info(gvt, offset);
+ if (!mmio_info) {
+ gvt_dbg_mmio("untracked MMIO %08x len %d\n", offset, bytes);
+ goto default_rw;
+ }
+
+ if (is_read)
+ return mmio_info->read(vgpu, offset, pdata, bytes);
+ else {
+ u64 ro_mask = mmio_info->ro_mask;
+ u32 old_vreg = 0;
+ u64 data = 0;
+
+ if (intel_gvt_mmio_has_mode_mask(gvt, mmio_info->offset)) {
+ old_vreg = vgpu_vreg(vgpu, offset);
+ }
+
+ if (likely(!ro_mask))
+ ret = mmio_info->write(vgpu, offset, pdata, bytes);
+ else if (!~ro_mask) {
+ gvt_vgpu_err("try to write RO reg %x\n", offset);
+ return 0;
+ } else {
+ /* keep the RO bits in the virtual register */
+ memcpy(&data, pdata, bytes);
+ data &= ~ro_mask;
+ data |= vgpu_vreg(vgpu, offset) & ro_mask;
+ ret = mmio_info->write(vgpu, offset, &data, bytes);
+ }
+
+ /* higher 16bits of mode ctl regs are mask bits for change */
+ if (intel_gvt_mmio_has_mode_mask(gvt, mmio_info->offset)) {
+ u32 mask = vgpu_vreg(vgpu, offset) >> 16;
+
+ vgpu_vreg(vgpu, offset) = (old_vreg & ~mask)
+ | (vgpu_vreg(vgpu, offset) & mask);
+ }
+ }
+
+ return ret;
+
+default_rw:
+ return is_read ?
+ intel_vgpu_default_mmio_read(vgpu, offset, pdata, bytes) :
+ intel_vgpu_default_mmio_write(vgpu, offset, pdata, bytes);
+}
+
+void intel_gvt_restore_fence(struct intel_gvt *gvt)
+{
+ struct intel_vgpu *vgpu;
+ int i, id;
+
+ idr_for_each_entry(&(gvt)->vgpu_idr, vgpu, id) {
+ mmio_hw_access_pre(gvt->gt);
+ for (i = 0; i < vgpu_fence_sz(vgpu); i++)
+ intel_vgpu_write_fence(vgpu, i, vgpu_vreg64(vgpu, fence_num_to_offset(i)));
+ mmio_hw_access_post(gvt->gt);
+ }
+}
+
+static int mmio_pm_restore_handler(struct intel_gvt *gvt, u32 offset, void *data)
+{
+ struct intel_vgpu *vgpu = data;
+ struct drm_i915_private *dev_priv = gvt->gt->i915;
+
+ if (gvt->mmio.mmio_attribute[offset >> 2] & F_PM_SAVE)
+ intel_uncore_write(&dev_priv->uncore, _MMIO(offset), vgpu_vreg(vgpu, offset));
+
+ return 0;
+}
+
+void intel_gvt_restore_mmio(struct intel_gvt *gvt)
+{
+ struct intel_vgpu *vgpu;
+ int id;
+
+ idr_for_each_entry(&(gvt)->vgpu_idr, vgpu, id) {
+ mmio_hw_access_pre(gvt->gt);
+ intel_gvt_for_each_tracked_mmio(gvt, mmio_pm_restore_handler, vgpu);
+ mmio_hw_access_post(gvt->gt);
+ }
+}
diff --git a/drivers/gpu/drm/i915/gvt/interrupt.c b/drivers/gpu/drm/i915/gvt/interrupt.c
new file mode 100644
index 000000000..a6b2021b6
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/interrupt.c
@@ -0,0 +1,717 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Min he <min.he@intel.com>
+ *
+ */
+
+#include <linux/eventfd.h>
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gvt.h"
+#include "trace.h"
+
+/* common offset among interrupt control registers */
+#define regbase_to_isr(base) (base)
+#define regbase_to_imr(base) (base + 0x4)
+#define regbase_to_iir(base) (base + 0x8)
+#define regbase_to_ier(base) (base + 0xC)
+
+#define iir_to_regbase(iir) (iir - 0x8)
+#define ier_to_regbase(ier) (ier - 0xC)
+
+#define get_event_virt_handler(irq, e) (irq->events[e].v_handler)
+#define get_irq_info(irq, e) (irq->events[e].info)
+
+#define irq_to_gvt(irq) \
+ container_of(irq, struct intel_gvt, irq)
+
+static void update_upstream_irq(struct intel_vgpu *vgpu,
+ struct intel_gvt_irq_info *info);
+
+static const char * const irq_name[INTEL_GVT_EVENT_MAX] = {
+ [RCS_MI_USER_INTERRUPT] = "Render CS MI USER INTERRUPT",
+ [RCS_DEBUG] = "Render EU debug from SVG",
+ [RCS_MMIO_SYNC_FLUSH] = "Render MMIO sync flush status",
+ [RCS_CMD_STREAMER_ERR] = "Render CS error interrupt",
+ [RCS_PIPE_CONTROL] = "Render PIPE CONTROL notify",
+ [RCS_WATCHDOG_EXCEEDED] = "Render CS Watchdog counter exceeded",
+ [RCS_PAGE_DIRECTORY_FAULT] = "Render page directory faults",
+ [RCS_AS_CONTEXT_SWITCH] = "Render AS Context Switch Interrupt",
+
+ [VCS_MI_USER_INTERRUPT] = "Video CS MI USER INTERRUPT",
+ [VCS_MMIO_SYNC_FLUSH] = "Video MMIO sync flush status",
+ [VCS_CMD_STREAMER_ERR] = "Video CS error interrupt",
+ [VCS_MI_FLUSH_DW] = "Video MI FLUSH DW notify",
+ [VCS_WATCHDOG_EXCEEDED] = "Video CS Watchdog counter exceeded",
+ [VCS_PAGE_DIRECTORY_FAULT] = "Video page directory faults",
+ [VCS_AS_CONTEXT_SWITCH] = "Video AS Context Switch Interrupt",
+ [VCS2_MI_USER_INTERRUPT] = "VCS2 Video CS MI USER INTERRUPT",
+ [VCS2_MI_FLUSH_DW] = "VCS2 Video MI FLUSH DW notify",
+ [VCS2_AS_CONTEXT_SWITCH] = "VCS2 Context Switch Interrupt",
+
+ [BCS_MI_USER_INTERRUPT] = "Blitter CS MI USER INTERRUPT",
+ [BCS_MMIO_SYNC_FLUSH] = "Billter MMIO sync flush status",
+ [BCS_CMD_STREAMER_ERR] = "Blitter CS error interrupt",
+ [BCS_MI_FLUSH_DW] = "Blitter MI FLUSH DW notify",
+ [BCS_PAGE_DIRECTORY_FAULT] = "Blitter page directory faults",
+ [BCS_AS_CONTEXT_SWITCH] = "Blitter AS Context Switch Interrupt",
+
+ [VECS_MI_FLUSH_DW] = "Video Enhanced Streamer MI FLUSH DW notify",
+ [VECS_AS_CONTEXT_SWITCH] = "VECS Context Switch Interrupt",
+
+ [PIPE_A_FIFO_UNDERRUN] = "Pipe A FIFO underrun",
+ [PIPE_A_CRC_ERR] = "Pipe A CRC error",
+ [PIPE_A_CRC_DONE] = "Pipe A CRC done",
+ [PIPE_A_VSYNC] = "Pipe A vsync",
+ [PIPE_A_LINE_COMPARE] = "Pipe A line compare",
+ [PIPE_A_ODD_FIELD] = "Pipe A odd field",
+ [PIPE_A_EVEN_FIELD] = "Pipe A even field",
+ [PIPE_A_VBLANK] = "Pipe A vblank",
+ [PIPE_B_FIFO_UNDERRUN] = "Pipe B FIFO underrun",
+ [PIPE_B_CRC_ERR] = "Pipe B CRC error",
+ [PIPE_B_CRC_DONE] = "Pipe B CRC done",
+ [PIPE_B_VSYNC] = "Pipe B vsync",
+ [PIPE_B_LINE_COMPARE] = "Pipe B line compare",
+ [PIPE_B_ODD_FIELD] = "Pipe B odd field",
+ [PIPE_B_EVEN_FIELD] = "Pipe B even field",
+ [PIPE_B_VBLANK] = "Pipe B vblank",
+ [PIPE_C_VBLANK] = "Pipe C vblank",
+ [DPST_PHASE_IN] = "DPST phase in event",
+ [DPST_HISTOGRAM] = "DPST histogram event",
+ [GSE] = "GSE",
+ [DP_A_HOTPLUG] = "DP A Hotplug",
+ [AUX_CHANNEL_A] = "AUX Channel A",
+ [PERF_COUNTER] = "Performance counter",
+ [POISON] = "Poison",
+ [GTT_FAULT] = "GTT fault",
+ [PRIMARY_A_FLIP_DONE] = "Primary Plane A flip done",
+ [PRIMARY_B_FLIP_DONE] = "Primary Plane B flip done",
+ [PRIMARY_C_FLIP_DONE] = "Primary Plane C flip done",
+ [SPRITE_A_FLIP_DONE] = "Sprite Plane A flip done",
+ [SPRITE_B_FLIP_DONE] = "Sprite Plane B flip done",
+ [SPRITE_C_FLIP_DONE] = "Sprite Plane C flip done",
+
+ [PCU_THERMAL] = "PCU Thermal Event",
+ [PCU_PCODE2DRIVER_MAILBOX] = "PCU pcode2driver mailbox event",
+
+ [FDI_RX_INTERRUPTS_TRANSCODER_A] = "FDI RX Interrupts Combined A",
+ [AUDIO_CP_CHANGE_TRANSCODER_A] = "Audio CP Change Transcoder A",
+ [AUDIO_CP_REQUEST_TRANSCODER_A] = "Audio CP Request Transcoder A",
+ [FDI_RX_INTERRUPTS_TRANSCODER_B] = "FDI RX Interrupts Combined B",
+ [AUDIO_CP_CHANGE_TRANSCODER_B] = "Audio CP Change Transcoder B",
+ [AUDIO_CP_REQUEST_TRANSCODER_B] = "Audio CP Request Transcoder B",
+ [FDI_RX_INTERRUPTS_TRANSCODER_C] = "FDI RX Interrupts Combined C",
+ [AUDIO_CP_CHANGE_TRANSCODER_C] = "Audio CP Change Transcoder C",
+ [AUDIO_CP_REQUEST_TRANSCODER_C] = "Audio CP Request Transcoder C",
+ [ERR_AND_DBG] = "South Error and Debug Interrupts Combined",
+ [GMBUS] = "Gmbus",
+ [SDVO_B_HOTPLUG] = "SDVO B hotplug",
+ [CRT_HOTPLUG] = "CRT Hotplug",
+ [DP_B_HOTPLUG] = "DisplayPort/HDMI/DVI B Hotplug",
+ [DP_C_HOTPLUG] = "DisplayPort/HDMI/DVI C Hotplug",
+ [DP_D_HOTPLUG] = "DisplayPort/HDMI/DVI D Hotplug",
+ [AUX_CHANNEL_B] = "AUX Channel B",
+ [AUX_CHANNEL_C] = "AUX Channel C",
+ [AUX_CHANNEL_D] = "AUX Channel D",
+ [AUDIO_POWER_STATE_CHANGE_B] = "Audio Power State change Port B",
+ [AUDIO_POWER_STATE_CHANGE_C] = "Audio Power State change Port C",
+ [AUDIO_POWER_STATE_CHANGE_D] = "Audio Power State change Port D",
+
+ [INTEL_GVT_EVENT_RESERVED] = "RESERVED EVENTS!!!",
+};
+
+static inline struct intel_gvt_irq_info *regbase_to_irq_info(
+ struct intel_gvt *gvt,
+ unsigned int reg)
+{
+ struct intel_gvt_irq *irq = &gvt->irq;
+ int i;
+
+ for_each_set_bit(i, irq->irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX) {
+ if (i915_mmio_reg_offset(irq->info[i]->reg_base) == reg)
+ return irq->info[i];
+ }
+
+ return NULL;
+}
+
+/**
+ * intel_vgpu_reg_imr_handler - Generic IMR register emulation write handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IMR register bit change
+ * behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_imr_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+ u32 imr = *(u32 *)p_data;
+
+ trace_write_ir(vgpu->id, "IMR", reg, imr, vgpu_vreg(vgpu, reg),
+ (vgpu_vreg(vgpu, reg) ^ imr));
+
+ vgpu_vreg(vgpu, reg) = imr;
+
+ ops->check_pending_irq(vgpu);
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_reg_master_irq_handler - master IRQ write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the master IRQ register on gen8+.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_master_irq_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+ u32 ier = *(u32 *)p_data;
+ u32 virtual_ier = vgpu_vreg(vgpu, reg);
+
+ trace_write_ir(vgpu->id, "MASTER_IRQ", reg, ier, virtual_ier,
+ (virtual_ier ^ ier));
+
+ /*
+ * GEN8_MASTER_IRQ is a special irq register,
+ * only bit 31 is allowed to be modified
+ * and treated as an IER bit.
+ */
+ ier &= GEN8_MASTER_IRQ_CONTROL;
+ virtual_ier &= GEN8_MASTER_IRQ_CONTROL;
+ vgpu_vreg(vgpu, reg) &= ~GEN8_MASTER_IRQ_CONTROL;
+ vgpu_vreg(vgpu, reg) |= ier;
+
+ ops->check_pending_irq(vgpu);
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_reg_ier_handler - Generic IER write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IER register behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_ier_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ const struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+ struct intel_gvt_irq_info *info;
+ u32 ier = *(u32 *)p_data;
+
+ trace_write_ir(vgpu->id, "IER", reg, ier, vgpu_vreg(vgpu, reg),
+ (vgpu_vreg(vgpu, reg) ^ ier));
+
+ vgpu_vreg(vgpu, reg) = ier;
+
+ info = regbase_to_irq_info(gvt, ier_to_regbase(reg));
+ if (drm_WARN_ON(&i915->drm, !info))
+ return -EINVAL;
+
+ if (info->has_upstream_irq)
+ update_upstream_irq(vgpu, info);
+
+ ops->check_pending_irq(vgpu);
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_reg_iir_handler - Generic IIR write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IIR register behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_iir_handler(struct intel_vgpu *vgpu, unsigned int reg,
+ void *p_data, unsigned int bytes)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_gvt_irq_info *info = regbase_to_irq_info(vgpu->gvt,
+ iir_to_regbase(reg));
+ u32 iir = *(u32 *)p_data;
+
+ trace_write_ir(vgpu->id, "IIR", reg, iir, vgpu_vreg(vgpu, reg),
+ (vgpu_vreg(vgpu, reg) ^ iir));
+
+ if (drm_WARN_ON(&i915->drm, !info))
+ return -EINVAL;
+
+ vgpu_vreg(vgpu, reg) &= ~iir;
+
+ if (info->has_upstream_irq)
+ update_upstream_irq(vgpu, info);
+ return 0;
+}
+
+static struct intel_gvt_irq_map gen8_irq_map[] = {
+ { INTEL_GVT_IRQ_INFO_MASTER, 0, INTEL_GVT_IRQ_INFO_GT0, 0xffff },
+ { INTEL_GVT_IRQ_INFO_MASTER, 1, INTEL_GVT_IRQ_INFO_GT0, 0xffff0000 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 2, INTEL_GVT_IRQ_INFO_GT1, 0xffff },
+ { INTEL_GVT_IRQ_INFO_MASTER, 3, INTEL_GVT_IRQ_INFO_GT1, 0xffff0000 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 4, INTEL_GVT_IRQ_INFO_GT2, 0xffff },
+ { INTEL_GVT_IRQ_INFO_MASTER, 6, INTEL_GVT_IRQ_INFO_GT3, 0xffff },
+ { INTEL_GVT_IRQ_INFO_MASTER, 16, INTEL_GVT_IRQ_INFO_DE_PIPE_A, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 17, INTEL_GVT_IRQ_INFO_DE_PIPE_B, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 18, INTEL_GVT_IRQ_INFO_DE_PIPE_C, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 20, INTEL_GVT_IRQ_INFO_DE_PORT, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 22, INTEL_GVT_IRQ_INFO_DE_MISC, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 23, INTEL_GVT_IRQ_INFO_PCH, ~0 },
+ { INTEL_GVT_IRQ_INFO_MASTER, 30, INTEL_GVT_IRQ_INFO_PCU, ~0 },
+ { -1, -1, ~0 },
+};
+
+static void update_upstream_irq(struct intel_vgpu *vgpu,
+ struct intel_gvt_irq_info *info)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_gvt_irq *irq = &vgpu->gvt->irq;
+ struct intel_gvt_irq_map *map = irq->irq_map;
+ struct intel_gvt_irq_info *up_irq_info = NULL;
+ u32 set_bits = 0;
+ u32 clear_bits = 0;
+ int bit;
+ u32 val = vgpu_vreg(vgpu,
+ regbase_to_iir(i915_mmio_reg_offset(info->reg_base)))
+ & vgpu_vreg(vgpu,
+ regbase_to_ier(i915_mmio_reg_offset(info->reg_base)));
+
+ if (!info->has_upstream_irq)
+ return;
+
+ for (map = irq->irq_map; map->up_irq_bit != -1; map++) {
+ if (info->group != map->down_irq_group)
+ continue;
+
+ if (!up_irq_info)
+ up_irq_info = irq->info[map->up_irq_group];
+ else
+ drm_WARN_ON(&i915->drm, up_irq_info !=
+ irq->info[map->up_irq_group]);
+
+ bit = map->up_irq_bit;
+
+ if (val & map->down_irq_bitmask)
+ set_bits |= (1 << bit);
+ else
+ clear_bits |= (1 << bit);
+ }
+
+ if (drm_WARN_ON(&i915->drm, !up_irq_info))
+ return;
+
+ if (up_irq_info->group == INTEL_GVT_IRQ_INFO_MASTER) {
+ u32 isr = i915_mmio_reg_offset(up_irq_info->reg_base);
+
+ vgpu_vreg(vgpu, isr) &= ~clear_bits;
+ vgpu_vreg(vgpu, isr) |= set_bits;
+ } else {
+ u32 iir = regbase_to_iir(
+ i915_mmio_reg_offset(up_irq_info->reg_base));
+ u32 imr = regbase_to_imr(
+ i915_mmio_reg_offset(up_irq_info->reg_base));
+
+ vgpu_vreg(vgpu, iir) |= (set_bits & ~vgpu_vreg(vgpu, imr));
+ }
+
+ if (up_irq_info->has_upstream_irq)
+ update_upstream_irq(vgpu, up_irq_info);
+}
+
+static void init_irq_map(struct intel_gvt_irq *irq)
+{
+ struct intel_gvt_irq_map *map;
+ struct intel_gvt_irq_info *up_info, *down_info;
+ int up_bit;
+
+ for (map = irq->irq_map; map->up_irq_bit != -1; map++) {
+ up_info = irq->info[map->up_irq_group];
+ up_bit = map->up_irq_bit;
+ down_info = irq->info[map->down_irq_group];
+
+ set_bit(up_bit, up_info->downstream_irq_bitmap);
+ down_info->has_upstream_irq = true;
+
+ gvt_dbg_irq("[up] grp %d bit %d -> [down] grp %d bitmask %x\n",
+ up_info->group, up_bit,
+ down_info->group, map->down_irq_bitmask);
+ }
+}
+
+/* =======================vEvent injection===================== */
+
+#define MSI_CAP_CONTROL(offset) (offset + 2)
+#define MSI_CAP_ADDRESS(offset) (offset + 4)
+#define MSI_CAP_DATA(offset) (offset + 8)
+#define MSI_CAP_EN 0x1
+
+static int inject_virtual_interrupt(struct intel_vgpu *vgpu)
+{
+ unsigned long offset = vgpu->gvt->device_info.msi_cap_offset;
+ u16 control, data;
+ u32 addr;
+
+ control = *(u16 *)(vgpu_cfg_space(vgpu) + MSI_CAP_CONTROL(offset));
+ addr = *(u32 *)(vgpu_cfg_space(vgpu) + MSI_CAP_ADDRESS(offset));
+ data = *(u16 *)(vgpu_cfg_space(vgpu) + MSI_CAP_DATA(offset));
+
+ /* Do not generate MSI if MSIEN is disabled */
+ if (!(control & MSI_CAP_EN))
+ return 0;
+
+ if (WARN(control & GENMASK(15, 1), "only support one MSI format\n"))
+ return -EINVAL;
+
+ trace_inject_msi(vgpu->id, addr, data);
+
+ /*
+ * When guest is powered off, msi_trigger is set to NULL, but vgpu's
+ * config and mmio register isn't restored to default during guest
+ * poweroff. If this vgpu is still used in next vm, this vgpu's pipe
+ * may be enabled, then once this vgpu is active, it will get inject
+ * vblank interrupt request. But msi_trigger is null until msi is
+ * enabled by guest. so if msi_trigger is null, success is still
+ * returned and don't inject interrupt into guest.
+ */
+ if (!vgpu->attached)
+ return -ESRCH;
+ if (vgpu->msi_trigger && eventfd_signal(vgpu->msi_trigger, 1) != 1)
+ return -EFAULT;
+ return 0;
+}
+
+static void propagate_event(struct intel_gvt_irq *irq,
+ enum intel_gvt_event_type event, struct intel_vgpu *vgpu)
+{
+ struct intel_gvt_irq_info *info;
+ unsigned int reg_base;
+ int bit;
+
+ info = get_irq_info(irq, event);
+ if (WARN_ON(!info))
+ return;
+
+ reg_base = i915_mmio_reg_offset(info->reg_base);
+ bit = irq->events[event].bit;
+
+ if (!test_bit(bit, (void *)&vgpu_vreg(vgpu,
+ regbase_to_imr(reg_base)))) {
+ trace_propagate_event(vgpu->id, irq_name[event], bit);
+ set_bit(bit, (void *)&vgpu_vreg(vgpu,
+ regbase_to_iir(reg_base)));
+ }
+}
+
+/* =======================vEvent Handlers===================== */
+static void handle_default_event_virt(struct intel_gvt_irq *irq,
+ enum intel_gvt_event_type event, struct intel_vgpu *vgpu)
+{
+ if (!vgpu->irq.irq_warn_once[event]) {
+ gvt_dbg_core("vgpu%d: IRQ receive event %d (%s)\n",
+ vgpu->id, event, irq_name[event]);
+ vgpu->irq.irq_warn_once[event] = true;
+ }
+ propagate_event(irq, event, vgpu);
+}
+
+/* =====================GEN specific logic======================= */
+/* GEN8 interrupt routines. */
+
+#define DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(regname, regbase) \
+static struct intel_gvt_irq_info gen8_##regname##_info = { \
+ .name = #regname"-IRQ", \
+ .reg_base = (regbase), \
+ .bit_to_event = {[0 ... INTEL_GVT_IRQ_BITWIDTH-1] = \
+ INTEL_GVT_EVENT_RESERVED}, \
+}
+
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt0, GEN8_GT_ISR(0));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt1, GEN8_GT_ISR(1));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt2, GEN8_GT_ISR(2));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt3, GEN8_GT_ISR(3));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_a, GEN8_DE_PIPE_ISR(PIPE_A));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_b, GEN8_DE_PIPE_ISR(PIPE_B));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_c, GEN8_DE_PIPE_ISR(PIPE_C));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_port, GEN8_DE_PORT_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_misc, GEN8_DE_MISC_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(pcu, GEN8_PCU_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(master, GEN8_MASTER_IRQ);
+
+static struct intel_gvt_irq_info gvt_base_pch_info = {
+ .name = "PCH-IRQ",
+ .reg_base = SDEISR,
+ .bit_to_event = {[0 ... INTEL_GVT_IRQ_BITWIDTH-1] =
+ INTEL_GVT_EVENT_RESERVED},
+};
+
+static void gen8_check_pending_irq(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt_irq *irq = &vgpu->gvt->irq;
+ int i;
+
+ if (!(vgpu_vreg(vgpu, i915_mmio_reg_offset(GEN8_MASTER_IRQ)) &
+ GEN8_MASTER_IRQ_CONTROL))
+ return;
+
+ for_each_set_bit(i, irq->irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX) {
+ struct intel_gvt_irq_info *info = irq->info[i];
+ u32 reg_base;
+
+ if (!info->has_upstream_irq)
+ continue;
+
+ reg_base = i915_mmio_reg_offset(info->reg_base);
+ if ((vgpu_vreg(vgpu, regbase_to_iir(reg_base))
+ & vgpu_vreg(vgpu, regbase_to_ier(reg_base))))
+ update_upstream_irq(vgpu, info);
+ }
+
+ if (vgpu_vreg(vgpu, i915_mmio_reg_offset(GEN8_MASTER_IRQ))
+ & ~GEN8_MASTER_IRQ_CONTROL)
+ inject_virtual_interrupt(vgpu);
+}
+
+static void gen8_init_irq(
+ struct intel_gvt_irq *irq)
+{
+ struct intel_gvt *gvt = irq_to_gvt(irq);
+
+#define SET_BIT_INFO(s, b, e, i) \
+ do { \
+ s->events[e].bit = b; \
+ s->events[e].info = s->info[i]; \
+ s->info[i]->bit_to_event[b] = e;\
+ } while (0)
+
+#define SET_IRQ_GROUP(s, g, i) \
+ do { \
+ s->info[g] = i; \
+ (i)->group = g; \
+ set_bit(g, s->irq_info_bitmap); \
+ } while (0)
+
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_MASTER, &gen8_master_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT0, &gen8_gt0_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT1, &gen8_gt1_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT2, &gen8_gt2_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT3, &gen8_gt3_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_A, &gen8_de_pipe_a_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_B, &gen8_de_pipe_b_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_C, &gen8_de_pipe_c_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PORT, &gen8_de_port_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_MISC, &gen8_de_misc_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_PCU, &gen8_pcu_info);
+ SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_PCH, &gvt_base_pch_info);
+
+ /* GEN8 level 2 interrupts. */
+
+ /* GEN8 interrupt GT0 events */
+ SET_BIT_INFO(irq, 0, RCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 4, RCS_PIPE_CONTROL, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 8, RCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
+
+ SET_BIT_INFO(irq, 16, BCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 20, BCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 24, BCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
+
+ /* GEN8 interrupt GT1 events */
+ SET_BIT_INFO(irq, 0, VCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 4, VCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 8, VCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT1);
+
+ if (HAS_ENGINE(gvt->gt, VCS1)) {
+ SET_BIT_INFO(irq, 16, VCS2_MI_USER_INTERRUPT,
+ INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 20, VCS2_MI_FLUSH_DW,
+ INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 24, VCS2_AS_CONTEXT_SWITCH,
+ INTEL_GVT_IRQ_INFO_GT1);
+ }
+
+ /* GEN8 interrupt GT3 events */
+ SET_BIT_INFO(irq, 0, VECS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT3);
+ SET_BIT_INFO(irq, 4, VECS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT3);
+ SET_BIT_INFO(irq, 8, VECS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT3);
+
+ SET_BIT_INFO(irq, 0, PIPE_A_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+ SET_BIT_INFO(irq, 0, PIPE_B_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+ SET_BIT_INFO(irq, 0, PIPE_C_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+
+ /* GEN8 interrupt DE PORT events */
+ SET_BIT_INFO(irq, 0, AUX_CHANNEL_A, INTEL_GVT_IRQ_INFO_DE_PORT);
+ SET_BIT_INFO(irq, 3, DP_A_HOTPLUG, INTEL_GVT_IRQ_INFO_DE_PORT);
+
+ /* GEN8 interrupt DE MISC events */
+ SET_BIT_INFO(irq, 0, GSE, INTEL_GVT_IRQ_INFO_DE_MISC);
+
+ /* PCH events */
+ SET_BIT_INFO(irq, 17, GMBUS, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 19, CRT_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 21, DP_B_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 22, DP_C_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 23, DP_D_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+
+ if (IS_BROADWELL(gvt->gt->i915)) {
+ SET_BIT_INFO(irq, 25, AUX_CHANNEL_B, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 26, AUX_CHANNEL_C, INTEL_GVT_IRQ_INFO_PCH);
+ SET_BIT_INFO(irq, 27, AUX_CHANNEL_D, INTEL_GVT_IRQ_INFO_PCH);
+
+ SET_BIT_INFO(irq, 4, PRIMARY_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+ SET_BIT_INFO(irq, 5, SPRITE_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+
+ SET_BIT_INFO(irq, 4, PRIMARY_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+ SET_BIT_INFO(irq, 5, SPRITE_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+
+ SET_BIT_INFO(irq, 4, PRIMARY_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+ SET_BIT_INFO(irq, 5, SPRITE_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+ } else if (GRAPHICS_VER(gvt->gt->i915) >= 9) {
+ SET_BIT_INFO(irq, 25, AUX_CHANNEL_B, INTEL_GVT_IRQ_INFO_DE_PORT);
+ SET_BIT_INFO(irq, 26, AUX_CHANNEL_C, INTEL_GVT_IRQ_INFO_DE_PORT);
+ SET_BIT_INFO(irq, 27, AUX_CHANNEL_D, INTEL_GVT_IRQ_INFO_DE_PORT);
+
+ SET_BIT_INFO(irq, 3, PRIMARY_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+ SET_BIT_INFO(irq, 3, PRIMARY_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+ SET_BIT_INFO(irq, 3, PRIMARY_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+
+ SET_BIT_INFO(irq, 4, SPRITE_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+ SET_BIT_INFO(irq, 4, SPRITE_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+ SET_BIT_INFO(irq, 4, SPRITE_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+ }
+
+ /* GEN8 interrupt PCU events */
+ SET_BIT_INFO(irq, 24, PCU_THERMAL, INTEL_GVT_IRQ_INFO_PCU);
+ SET_BIT_INFO(irq, 25, PCU_PCODE2DRIVER_MAILBOX, INTEL_GVT_IRQ_INFO_PCU);
+}
+
+static const struct intel_gvt_irq_ops gen8_irq_ops = {
+ .init_irq = gen8_init_irq,
+ .check_pending_irq = gen8_check_pending_irq,
+};
+
+/**
+ * intel_vgpu_trigger_virtual_event - Trigger a virtual event for a vGPU
+ * @vgpu: a vGPU
+ * @event: interrupt event
+ *
+ * This function is used to trigger a virtual interrupt event for vGPU.
+ * The caller provides the event to be triggered, the framework itself
+ * will emulate the IRQ register bit change.
+ *
+ */
+void intel_vgpu_trigger_virtual_event(struct intel_vgpu *vgpu,
+ enum intel_gvt_event_type event)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_irq *irq = &gvt->irq;
+ gvt_event_virt_handler_t handler;
+ const struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+
+ handler = get_event_virt_handler(irq, event);
+ drm_WARN_ON(&i915->drm, !handler);
+
+ handler(irq, event, vgpu);
+
+ ops->check_pending_irq(vgpu);
+}
+
+static void init_events(
+ struct intel_gvt_irq *irq)
+{
+ int i;
+
+ for (i = 0; i < INTEL_GVT_EVENT_MAX; i++) {
+ irq->events[i].info = NULL;
+ irq->events[i].v_handler = handle_default_event_virt;
+ }
+}
+
+/**
+ * intel_gvt_init_irq - initialize GVT-g IRQ emulation subsystem
+ * @gvt: a GVT device
+ *
+ * This function is called at driver loading stage, to initialize the GVT-g IRQ
+ * emulation subsystem.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_init_irq(struct intel_gvt *gvt)
+{
+ struct intel_gvt_irq *irq = &gvt->irq;
+
+ gvt_dbg_core("init irq framework\n");
+
+ irq->ops = &gen8_irq_ops;
+ irq->irq_map = gen8_irq_map;
+
+ /* common event initialization */
+ init_events(irq);
+
+ /* gen specific initialization */
+ irq->ops->init_irq(irq);
+
+ init_irq_map(irq);
+
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/interrupt.h b/drivers/gpu/drm/i915/gvt/interrupt.h
new file mode 100644
index 000000000..b62f04ab4
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/interrupt.h
@@ -0,0 +1,232 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Min he <min.he@intel.com>
+ *
+ */
+
+#ifndef _GVT_INTERRUPT_H_
+#define _GVT_INTERRUPT_H_
+
+#include <linux/hrtimer.h>
+#include <linux/kernel.h>
+
+#include "i915_reg_defs.h"
+
+enum intel_gvt_event_type {
+ RCS_MI_USER_INTERRUPT = 0,
+ RCS_DEBUG,
+ RCS_MMIO_SYNC_FLUSH,
+ RCS_CMD_STREAMER_ERR,
+ RCS_PIPE_CONTROL,
+ RCS_L3_PARITY_ERR,
+ RCS_WATCHDOG_EXCEEDED,
+ RCS_PAGE_DIRECTORY_FAULT,
+ RCS_AS_CONTEXT_SWITCH,
+ RCS_MONITOR_BUFF_HALF_FULL,
+
+ VCS_MI_USER_INTERRUPT,
+ VCS_MMIO_SYNC_FLUSH,
+ VCS_CMD_STREAMER_ERR,
+ VCS_MI_FLUSH_DW,
+ VCS_WATCHDOG_EXCEEDED,
+ VCS_PAGE_DIRECTORY_FAULT,
+ VCS_AS_CONTEXT_SWITCH,
+
+ VCS2_MI_USER_INTERRUPT,
+ VCS2_MI_FLUSH_DW,
+ VCS2_AS_CONTEXT_SWITCH,
+
+ BCS_MI_USER_INTERRUPT,
+ BCS_MMIO_SYNC_FLUSH,
+ BCS_CMD_STREAMER_ERR,
+ BCS_MI_FLUSH_DW,
+ BCS_PAGE_DIRECTORY_FAULT,
+ BCS_AS_CONTEXT_SWITCH,
+
+ VECS_MI_USER_INTERRUPT,
+ VECS_MI_FLUSH_DW,
+ VECS_AS_CONTEXT_SWITCH,
+
+ PIPE_A_FIFO_UNDERRUN,
+ PIPE_B_FIFO_UNDERRUN,
+ PIPE_A_CRC_ERR,
+ PIPE_B_CRC_ERR,
+ PIPE_A_CRC_DONE,
+ PIPE_B_CRC_DONE,
+ PIPE_A_ODD_FIELD,
+ PIPE_B_ODD_FIELD,
+ PIPE_A_EVEN_FIELD,
+ PIPE_B_EVEN_FIELD,
+ PIPE_A_LINE_COMPARE,
+ PIPE_B_LINE_COMPARE,
+ PIPE_C_LINE_COMPARE,
+ PIPE_A_VBLANK,
+ PIPE_B_VBLANK,
+ PIPE_C_VBLANK,
+ PIPE_A_VSYNC,
+ PIPE_B_VSYNC,
+ PIPE_C_VSYNC,
+ PRIMARY_A_FLIP_DONE,
+ PRIMARY_B_FLIP_DONE,
+ PRIMARY_C_FLIP_DONE,
+ SPRITE_A_FLIP_DONE,
+ SPRITE_B_FLIP_DONE,
+ SPRITE_C_FLIP_DONE,
+
+ PCU_THERMAL,
+ PCU_PCODE2DRIVER_MAILBOX,
+
+ DPST_PHASE_IN,
+ DPST_HISTOGRAM,
+ GSE,
+ DP_A_HOTPLUG,
+ AUX_CHANNEL_A,
+ PERF_COUNTER,
+ POISON,
+ GTT_FAULT,
+ ERROR_INTERRUPT_COMBINED,
+
+ FDI_RX_INTERRUPTS_TRANSCODER_A,
+ AUDIO_CP_CHANGE_TRANSCODER_A,
+ AUDIO_CP_REQUEST_TRANSCODER_A,
+ FDI_RX_INTERRUPTS_TRANSCODER_B,
+ AUDIO_CP_CHANGE_TRANSCODER_B,
+ AUDIO_CP_REQUEST_TRANSCODER_B,
+ FDI_RX_INTERRUPTS_TRANSCODER_C,
+ AUDIO_CP_CHANGE_TRANSCODER_C,
+ AUDIO_CP_REQUEST_TRANSCODER_C,
+ ERR_AND_DBG,
+ GMBUS,
+ SDVO_B_HOTPLUG,
+ CRT_HOTPLUG,
+ DP_B_HOTPLUG,
+ DP_C_HOTPLUG,
+ DP_D_HOTPLUG,
+ AUX_CHANNEL_B,
+ AUX_CHANNEL_C,
+ AUX_CHANNEL_D,
+ AUDIO_POWER_STATE_CHANGE_B,
+ AUDIO_POWER_STATE_CHANGE_C,
+ AUDIO_POWER_STATE_CHANGE_D,
+
+ INTEL_GVT_EVENT_RESERVED,
+ INTEL_GVT_EVENT_MAX,
+};
+
+struct intel_gvt_irq;
+struct intel_gvt;
+struct intel_vgpu;
+
+typedef void (*gvt_event_virt_handler_t)(struct intel_gvt_irq *irq,
+ enum intel_gvt_event_type event, struct intel_vgpu *vgpu);
+
+struct intel_gvt_irq_ops {
+ void (*init_irq)(struct intel_gvt_irq *irq);
+ void (*check_pending_irq)(struct intel_vgpu *vgpu);
+};
+
+/* the list of physical interrupt control register groups */
+enum intel_gvt_irq_type {
+ INTEL_GVT_IRQ_INFO_GT,
+ INTEL_GVT_IRQ_INFO_DPY,
+ INTEL_GVT_IRQ_INFO_PCH,
+ INTEL_GVT_IRQ_INFO_PM,
+
+ INTEL_GVT_IRQ_INFO_MASTER,
+ INTEL_GVT_IRQ_INFO_GT0,
+ INTEL_GVT_IRQ_INFO_GT1,
+ INTEL_GVT_IRQ_INFO_GT2,
+ INTEL_GVT_IRQ_INFO_GT3,
+ INTEL_GVT_IRQ_INFO_DE_PIPE_A,
+ INTEL_GVT_IRQ_INFO_DE_PIPE_B,
+ INTEL_GVT_IRQ_INFO_DE_PIPE_C,
+ INTEL_GVT_IRQ_INFO_DE_PORT,
+ INTEL_GVT_IRQ_INFO_DE_MISC,
+ INTEL_GVT_IRQ_INFO_AUD,
+ INTEL_GVT_IRQ_INFO_PCU,
+
+ INTEL_GVT_IRQ_INFO_MAX,
+};
+
+#define INTEL_GVT_IRQ_BITWIDTH 32
+
+/* device specific interrupt bit definitions */
+struct intel_gvt_irq_info {
+ char *name;
+ i915_reg_t reg_base;
+ enum intel_gvt_event_type bit_to_event[INTEL_GVT_IRQ_BITWIDTH];
+ unsigned long warned;
+ int group;
+ DECLARE_BITMAP(downstream_irq_bitmap, INTEL_GVT_IRQ_BITWIDTH);
+ bool has_upstream_irq;
+};
+
+/* per-event information */
+struct intel_gvt_event_info {
+ int bit; /* map to register bit */
+ int policy; /* forwarding policy */
+ struct intel_gvt_irq_info *info; /* register info */
+ gvt_event_virt_handler_t v_handler; /* for v_event */
+};
+
+struct intel_gvt_irq_map {
+ int up_irq_group;
+ int up_irq_bit;
+ int down_irq_group;
+ u32 down_irq_bitmask;
+};
+
+/* structure containing device specific IRQ state */
+struct intel_gvt_irq {
+ const struct intel_gvt_irq_ops *ops;
+ struct intel_gvt_irq_info *info[INTEL_GVT_IRQ_INFO_MAX];
+ DECLARE_BITMAP(irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX);
+ struct intel_gvt_event_info events[INTEL_GVT_EVENT_MAX];
+ DECLARE_BITMAP(pending_events, INTEL_GVT_EVENT_MAX);
+ struct intel_gvt_irq_map *irq_map;
+};
+
+int intel_gvt_init_irq(struct intel_gvt *gvt);
+
+void intel_vgpu_trigger_virtual_event(struct intel_vgpu *vgpu,
+ enum intel_gvt_event_type event);
+
+int intel_vgpu_reg_iir_handler(struct intel_vgpu *vgpu, unsigned int reg,
+ void *p_data, unsigned int bytes);
+int intel_vgpu_reg_ier_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes);
+int intel_vgpu_reg_master_irq_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes);
+int intel_vgpu_reg_imr_handler(struct intel_vgpu *vgpu,
+ unsigned int reg, void *p_data, unsigned int bytes);
+
+int gvt_ring_id_to_pipe_control_notify_event(int ring_id);
+int gvt_ring_id_to_mi_flush_dw_event(int ring_id);
+int gvt_ring_id_to_mi_user_interrupt_event(int ring_id);
+
+#endif /* _GVT_INTERRUPT_H_ */
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
new file mode 100644
index 000000000..714221f9a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -0,0 +1,2017 @@
+/*
+ * KVMGT - the implementation of Intel mediated pass-through framework for KVM
+ *
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Kevin Tian <kevin.tian@intel.com>
+ * Jike Song <jike.song@intel.com>
+ * Xiaoguang Chen <xiaoguang.chen@intel.com>
+ * Eddie Dong <eddie.dong@intel.com>
+ *
+ * Contributors:
+ * Niu Bing <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ */
+
+#include <linux/init.h>
+#include <linux/mm.h>
+#include <linux/kthread.h>
+#include <linux/sched/mm.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/rbtree.h>
+#include <linux/spinlock.h>
+#include <linux/eventfd.h>
+#include <linux/mdev.h>
+#include <linux/debugfs.h>
+
+#include <linux/nospec.h>
+
+#include <drm/drm_edid.h>
+
+#include "i915_drv.h"
+#include "intel_gvt.h"
+#include "gvt.h"
+
+MODULE_IMPORT_NS(DMA_BUF);
+MODULE_IMPORT_NS(I915_GVT);
+
+/* helper macros copied from vfio-pci */
+#define VFIO_PCI_OFFSET_SHIFT 40
+#define VFIO_PCI_OFFSET_TO_INDEX(off) (off >> VFIO_PCI_OFFSET_SHIFT)
+#define VFIO_PCI_INDEX_TO_OFFSET(index) ((u64)(index) << VFIO_PCI_OFFSET_SHIFT)
+#define VFIO_PCI_OFFSET_MASK (((u64)(1) << VFIO_PCI_OFFSET_SHIFT) - 1)
+
+#define EDID_BLOB_OFFSET (PAGE_SIZE/2)
+
+#define OPREGION_SIGNATURE "IntelGraphicsMem"
+
+struct vfio_region;
+struct intel_vgpu_regops {
+ size_t (*rw)(struct intel_vgpu *vgpu, char *buf,
+ size_t count, loff_t *ppos, bool iswrite);
+ void (*release)(struct intel_vgpu *vgpu,
+ struct vfio_region *region);
+};
+
+struct vfio_region {
+ u32 type;
+ u32 subtype;
+ size_t size;
+ u32 flags;
+ const struct intel_vgpu_regops *ops;
+ void *data;
+};
+
+struct vfio_edid_region {
+ struct vfio_region_gfx_edid vfio_edid_regs;
+ void *edid_blob;
+};
+
+struct kvmgt_pgfn {
+ gfn_t gfn;
+ struct hlist_node hnode;
+};
+
+struct gvt_dma {
+ struct intel_vgpu *vgpu;
+ struct rb_node gfn_node;
+ struct rb_node dma_addr_node;
+ gfn_t gfn;
+ dma_addr_t dma_addr;
+ unsigned long size;
+ struct kref ref;
+};
+
+#define vfio_dev_to_vgpu(vfio_dev) \
+ container_of((vfio_dev), struct intel_vgpu, vfio_device)
+
+static void kvmgt_page_track_write(struct kvm_vcpu *vcpu, gpa_t gpa,
+ const u8 *val, int len,
+ struct kvm_page_track_notifier_node *node);
+static void kvmgt_page_track_flush_slot(struct kvm *kvm,
+ struct kvm_memory_slot *slot,
+ struct kvm_page_track_notifier_node *node);
+
+static ssize_t intel_vgpu_show_description(struct mdev_type *mtype, char *buf)
+{
+ struct intel_vgpu_type *type =
+ container_of(mtype, struct intel_vgpu_type, type);
+
+ return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n"
+ "fence: %d\nresolution: %s\n"
+ "weight: %d\n",
+ BYTES_TO_MB(type->conf->low_mm),
+ BYTES_TO_MB(type->conf->high_mm),
+ type->conf->fence, vgpu_edid_str(type->conf->edid),
+ type->conf->weight);
+}
+
+static void gvt_unpin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ unsigned long size)
+{
+ vfio_unpin_pages(&vgpu->vfio_device, gfn << PAGE_SHIFT,
+ DIV_ROUND_UP(size, PAGE_SIZE));
+}
+
+/* Pin a normal or compound guest page for dma. */
+static int gvt_pin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ unsigned long size, struct page **page)
+{
+ int total_pages = DIV_ROUND_UP(size, PAGE_SIZE);
+ struct page *base_page = NULL;
+ int npage;
+ int ret;
+
+ /*
+ * We pin the pages one-by-one to avoid allocating a big arrary
+ * on stack to hold pfns.
+ */
+ for (npage = 0; npage < total_pages; npage++) {
+ dma_addr_t cur_iova = (gfn + npage) << PAGE_SHIFT;
+ struct page *cur_page;
+
+ ret = vfio_pin_pages(&vgpu->vfio_device, cur_iova, 1,
+ IOMMU_READ | IOMMU_WRITE, &cur_page);
+ if (ret != 1) {
+ gvt_vgpu_err("vfio_pin_pages failed for iova %pad, ret %d\n",
+ &cur_iova, ret);
+ goto err;
+ }
+
+ if (npage == 0)
+ base_page = cur_page;
+ else if (base_page + npage != cur_page) {
+ gvt_vgpu_err("The pages are not continuous\n");
+ ret = -EINVAL;
+ npage++;
+ goto err;
+ }
+ }
+
+ *page = base_page;
+ return 0;
+err:
+ gvt_unpin_guest_page(vgpu, gfn, npage * PAGE_SIZE);
+ return ret;
+}
+
+static int gvt_dma_map_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ dma_addr_t *dma_addr, unsigned long size)
+{
+ struct device *dev = vgpu->gvt->gt->i915->drm.dev;
+ struct page *page = NULL;
+ int ret;
+
+ ret = gvt_pin_guest_page(vgpu, gfn, size, &page);
+ if (ret)
+ return ret;
+
+ /* Setup DMA mapping. */
+ *dma_addr = dma_map_page(dev, page, 0, size, DMA_BIDIRECTIONAL);
+ if (dma_mapping_error(dev, *dma_addr)) {
+ gvt_vgpu_err("DMA mapping failed for pfn 0x%lx, ret %d\n",
+ page_to_pfn(page), ret);
+ gvt_unpin_guest_page(vgpu, gfn, size);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static void gvt_dma_unmap_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ dma_addr_t dma_addr, unsigned long size)
+{
+ struct device *dev = vgpu->gvt->gt->i915->drm.dev;
+
+ dma_unmap_page(dev, dma_addr, size, DMA_BIDIRECTIONAL);
+ gvt_unpin_guest_page(vgpu, gfn, size);
+}
+
+static struct gvt_dma *__gvt_cache_find_dma_addr(struct intel_vgpu *vgpu,
+ dma_addr_t dma_addr)
+{
+ struct rb_node *node = vgpu->dma_addr_cache.rb_node;
+ struct gvt_dma *itr;
+
+ while (node) {
+ itr = rb_entry(node, struct gvt_dma, dma_addr_node);
+
+ if (dma_addr < itr->dma_addr)
+ node = node->rb_left;
+ else if (dma_addr > itr->dma_addr)
+ node = node->rb_right;
+ else
+ return itr;
+ }
+ return NULL;
+}
+
+static struct gvt_dma *__gvt_cache_find_gfn(struct intel_vgpu *vgpu, gfn_t gfn)
+{
+ struct rb_node *node = vgpu->gfn_cache.rb_node;
+ struct gvt_dma *itr;
+
+ while (node) {
+ itr = rb_entry(node, struct gvt_dma, gfn_node);
+
+ if (gfn < itr->gfn)
+ node = node->rb_left;
+ else if (gfn > itr->gfn)
+ node = node->rb_right;
+ else
+ return itr;
+ }
+ return NULL;
+}
+
+static int __gvt_cache_add(struct intel_vgpu *vgpu, gfn_t gfn,
+ dma_addr_t dma_addr, unsigned long size)
+{
+ struct gvt_dma *new, *itr;
+ struct rb_node **link, *parent = NULL;
+
+ new = kzalloc(sizeof(struct gvt_dma), GFP_KERNEL);
+ if (!new)
+ return -ENOMEM;
+
+ new->vgpu = vgpu;
+ new->gfn = gfn;
+ new->dma_addr = dma_addr;
+ new->size = size;
+ kref_init(&new->ref);
+
+ /* gfn_cache maps gfn to struct gvt_dma. */
+ link = &vgpu->gfn_cache.rb_node;
+ while (*link) {
+ parent = *link;
+ itr = rb_entry(parent, struct gvt_dma, gfn_node);
+
+ if (gfn < itr->gfn)
+ link = &parent->rb_left;
+ else
+ link = &parent->rb_right;
+ }
+ rb_link_node(&new->gfn_node, parent, link);
+ rb_insert_color(&new->gfn_node, &vgpu->gfn_cache);
+
+ /* dma_addr_cache maps dma addr to struct gvt_dma. */
+ parent = NULL;
+ link = &vgpu->dma_addr_cache.rb_node;
+ while (*link) {
+ parent = *link;
+ itr = rb_entry(parent, struct gvt_dma, dma_addr_node);
+
+ if (dma_addr < itr->dma_addr)
+ link = &parent->rb_left;
+ else
+ link = &parent->rb_right;
+ }
+ rb_link_node(&new->dma_addr_node, parent, link);
+ rb_insert_color(&new->dma_addr_node, &vgpu->dma_addr_cache);
+
+ vgpu->nr_cache_entries++;
+ return 0;
+}
+
+static void __gvt_cache_remove_entry(struct intel_vgpu *vgpu,
+ struct gvt_dma *entry)
+{
+ rb_erase(&entry->gfn_node, &vgpu->gfn_cache);
+ rb_erase(&entry->dma_addr_node, &vgpu->dma_addr_cache);
+ kfree(entry);
+ vgpu->nr_cache_entries--;
+}
+
+static void gvt_cache_destroy(struct intel_vgpu *vgpu)
+{
+ struct gvt_dma *dma;
+ struct rb_node *node = NULL;
+
+ for (;;) {
+ mutex_lock(&vgpu->cache_lock);
+ node = rb_first(&vgpu->gfn_cache);
+ if (!node) {
+ mutex_unlock(&vgpu->cache_lock);
+ break;
+ }
+ dma = rb_entry(node, struct gvt_dma, gfn_node);
+ gvt_dma_unmap_page(vgpu, dma->gfn, dma->dma_addr, dma->size);
+ __gvt_cache_remove_entry(vgpu, dma);
+ mutex_unlock(&vgpu->cache_lock);
+ }
+}
+
+static void gvt_cache_init(struct intel_vgpu *vgpu)
+{
+ vgpu->gfn_cache = RB_ROOT;
+ vgpu->dma_addr_cache = RB_ROOT;
+ vgpu->nr_cache_entries = 0;
+ mutex_init(&vgpu->cache_lock);
+}
+
+static void kvmgt_protect_table_init(struct intel_vgpu *info)
+{
+ hash_init(info->ptable);
+}
+
+static void kvmgt_protect_table_destroy(struct intel_vgpu *info)
+{
+ struct kvmgt_pgfn *p;
+ struct hlist_node *tmp;
+ int i;
+
+ hash_for_each_safe(info->ptable, i, tmp, p, hnode) {
+ hash_del(&p->hnode);
+ kfree(p);
+ }
+}
+
+static struct kvmgt_pgfn *
+__kvmgt_protect_table_find(struct intel_vgpu *info, gfn_t gfn)
+{
+ struct kvmgt_pgfn *p, *res = NULL;
+
+ hash_for_each_possible(info->ptable, p, hnode, gfn) {
+ if (gfn == p->gfn) {
+ res = p;
+ break;
+ }
+ }
+
+ return res;
+}
+
+static bool kvmgt_gfn_is_write_protected(struct intel_vgpu *info, gfn_t gfn)
+{
+ struct kvmgt_pgfn *p;
+
+ p = __kvmgt_protect_table_find(info, gfn);
+ return !!p;
+}
+
+static void kvmgt_protect_table_add(struct intel_vgpu *info, gfn_t gfn)
+{
+ struct kvmgt_pgfn *p;
+
+ if (kvmgt_gfn_is_write_protected(info, gfn))
+ return;
+
+ p = kzalloc(sizeof(struct kvmgt_pgfn), GFP_ATOMIC);
+ if (WARN(!p, "gfn: 0x%llx\n", gfn))
+ return;
+
+ p->gfn = gfn;
+ hash_add(info->ptable, &p->hnode, gfn);
+}
+
+static void kvmgt_protect_table_del(struct intel_vgpu *info, gfn_t gfn)
+{
+ struct kvmgt_pgfn *p;
+
+ p = __kvmgt_protect_table_find(info, gfn);
+ if (p) {
+ hash_del(&p->hnode);
+ kfree(p);
+ }
+}
+
+static size_t intel_vgpu_reg_rw_opregion(struct intel_vgpu *vgpu, char *buf,
+ size_t count, loff_t *ppos, bool iswrite)
+{
+ unsigned int i = VFIO_PCI_OFFSET_TO_INDEX(*ppos) -
+ VFIO_PCI_NUM_REGIONS;
+ void *base = vgpu->region[i].data;
+ loff_t pos = *ppos & VFIO_PCI_OFFSET_MASK;
+
+
+ if (pos >= vgpu->region[i].size || iswrite) {
+ gvt_vgpu_err("invalid op or offset for Intel vgpu OpRegion\n");
+ return -EINVAL;
+ }
+ count = min(count, (size_t)(vgpu->region[i].size - pos));
+ memcpy(buf, base + pos, count);
+
+ return count;
+}
+
+static void intel_vgpu_reg_release_opregion(struct intel_vgpu *vgpu,
+ struct vfio_region *region)
+{
+}
+
+static const struct intel_vgpu_regops intel_vgpu_regops_opregion = {
+ .rw = intel_vgpu_reg_rw_opregion,
+ .release = intel_vgpu_reg_release_opregion,
+};
+
+static int handle_edid_regs(struct intel_vgpu *vgpu,
+ struct vfio_edid_region *region, char *buf,
+ size_t count, u16 offset, bool is_write)
+{
+ struct vfio_region_gfx_edid *regs = &region->vfio_edid_regs;
+ unsigned int data;
+
+ if (offset + count > sizeof(*regs))
+ return -EINVAL;
+
+ if (count != 4)
+ return -EINVAL;
+
+ if (is_write) {
+ data = *((unsigned int *)buf);
+ switch (offset) {
+ case offsetof(struct vfio_region_gfx_edid, link_state):
+ if (data == VFIO_DEVICE_GFX_LINK_STATE_UP) {
+ if (!drm_edid_block_valid(
+ (u8 *)region->edid_blob,
+ 0,
+ true,
+ NULL)) {
+ gvt_vgpu_err("invalid EDID blob\n");
+ return -EINVAL;
+ }
+ intel_vgpu_emulate_hotplug(vgpu, true);
+ } else if (data == VFIO_DEVICE_GFX_LINK_STATE_DOWN)
+ intel_vgpu_emulate_hotplug(vgpu, false);
+ else {
+ gvt_vgpu_err("invalid EDID link state %d\n",
+ regs->link_state);
+ return -EINVAL;
+ }
+ regs->link_state = data;
+ break;
+ case offsetof(struct vfio_region_gfx_edid, edid_size):
+ if (data > regs->edid_max_size) {
+ gvt_vgpu_err("EDID size is bigger than %d!\n",
+ regs->edid_max_size);
+ return -EINVAL;
+ }
+ regs->edid_size = data;
+ break;
+ default:
+ /* read-only regs */
+ gvt_vgpu_err("write read-only EDID region at offset %d\n",
+ offset);
+ return -EPERM;
+ }
+ } else {
+ memcpy(buf, (char *)regs + offset, count);
+ }
+
+ return count;
+}
+
+static int handle_edid_blob(struct vfio_edid_region *region, char *buf,
+ size_t count, u16 offset, bool is_write)
+{
+ if (offset + count > region->vfio_edid_regs.edid_size)
+ return -EINVAL;
+
+ if (is_write)
+ memcpy(region->edid_blob + offset, buf, count);
+ else
+ memcpy(buf, region->edid_blob + offset, count);
+
+ return count;
+}
+
+static size_t intel_vgpu_reg_rw_edid(struct intel_vgpu *vgpu, char *buf,
+ size_t count, loff_t *ppos, bool iswrite)
+{
+ int ret;
+ unsigned int i = VFIO_PCI_OFFSET_TO_INDEX(*ppos) -
+ VFIO_PCI_NUM_REGIONS;
+ struct vfio_edid_region *region = vgpu->region[i].data;
+ loff_t pos = *ppos & VFIO_PCI_OFFSET_MASK;
+
+ if (pos < region->vfio_edid_regs.edid_offset) {
+ ret = handle_edid_regs(vgpu, region, buf, count, pos, iswrite);
+ } else {
+ pos -= EDID_BLOB_OFFSET;
+ ret = handle_edid_blob(region, buf, count, pos, iswrite);
+ }
+
+ if (ret < 0)
+ gvt_vgpu_err("failed to access EDID region\n");
+
+ return ret;
+}
+
+static void intel_vgpu_reg_release_edid(struct intel_vgpu *vgpu,
+ struct vfio_region *region)
+{
+ kfree(region->data);
+}
+
+static const struct intel_vgpu_regops intel_vgpu_regops_edid = {
+ .rw = intel_vgpu_reg_rw_edid,
+ .release = intel_vgpu_reg_release_edid,
+};
+
+static int intel_vgpu_register_reg(struct intel_vgpu *vgpu,
+ unsigned int type, unsigned int subtype,
+ const struct intel_vgpu_regops *ops,
+ size_t size, u32 flags, void *data)
+{
+ struct vfio_region *region;
+
+ region = krealloc(vgpu->region,
+ (vgpu->num_regions + 1) * sizeof(*region),
+ GFP_KERNEL);
+ if (!region)
+ return -ENOMEM;
+
+ vgpu->region = region;
+ vgpu->region[vgpu->num_regions].type = type;
+ vgpu->region[vgpu->num_regions].subtype = subtype;
+ vgpu->region[vgpu->num_regions].ops = ops;
+ vgpu->region[vgpu->num_regions].size = size;
+ vgpu->region[vgpu->num_regions].flags = flags;
+ vgpu->region[vgpu->num_regions].data = data;
+ vgpu->num_regions++;
+ return 0;
+}
+
+int intel_gvt_set_opregion(struct intel_vgpu *vgpu)
+{
+ void *base;
+ int ret;
+
+ /* Each vgpu has its own opregion, although VFIO would create another
+ * one later. This one is used to expose opregion to VFIO. And the
+ * other one created by VFIO later, is used by guest actually.
+ */
+ base = vgpu_opregion(vgpu)->va;
+ if (!base)
+ return -ENOMEM;
+
+ if (memcmp(base, OPREGION_SIGNATURE, 16)) {
+ memunmap(base);
+ return -EINVAL;
+ }
+
+ ret = intel_vgpu_register_reg(vgpu,
+ PCI_VENDOR_ID_INTEL | VFIO_REGION_TYPE_PCI_VENDOR_TYPE,
+ VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION,
+ &intel_vgpu_regops_opregion, OPREGION_SIZE,
+ VFIO_REGION_INFO_FLAG_READ, base);
+
+ return ret;
+}
+
+int intel_gvt_set_edid(struct intel_vgpu *vgpu, int port_num)
+{
+ struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
+ struct vfio_edid_region *base;
+ int ret;
+
+ base = kzalloc(sizeof(*base), GFP_KERNEL);
+ if (!base)
+ return -ENOMEM;
+
+ /* TODO: Add multi-port and EDID extension block support */
+ base->vfio_edid_regs.edid_offset = EDID_BLOB_OFFSET;
+ base->vfio_edid_regs.edid_max_size = EDID_SIZE;
+ base->vfio_edid_regs.edid_size = EDID_SIZE;
+ base->vfio_edid_regs.max_xres = vgpu_edid_xres(port->id);
+ base->vfio_edid_regs.max_yres = vgpu_edid_yres(port->id);
+ base->edid_blob = port->edid->edid_block;
+
+ ret = intel_vgpu_register_reg(vgpu,
+ VFIO_REGION_TYPE_GFX,
+ VFIO_REGION_SUBTYPE_GFX_EDID,
+ &intel_vgpu_regops_edid, EDID_SIZE,
+ VFIO_REGION_INFO_FLAG_READ |
+ VFIO_REGION_INFO_FLAG_WRITE |
+ VFIO_REGION_INFO_FLAG_CAPS, base);
+
+ return ret;
+}
+
+static void intel_vgpu_dma_unmap(struct vfio_device *vfio_dev, u64 iova,
+ u64 length)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ struct gvt_dma *entry;
+ u64 iov_pfn = iova >> PAGE_SHIFT;
+ u64 end_iov_pfn = iov_pfn + length / PAGE_SIZE;
+
+ mutex_lock(&vgpu->cache_lock);
+ for (; iov_pfn < end_iov_pfn; iov_pfn++) {
+ entry = __gvt_cache_find_gfn(vgpu, iov_pfn);
+ if (!entry)
+ continue;
+
+ gvt_dma_unmap_page(vgpu, entry->gfn, entry->dma_addr,
+ entry->size);
+ __gvt_cache_remove_entry(vgpu, entry);
+ }
+ mutex_unlock(&vgpu->cache_lock);
+}
+
+static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu *itr;
+ int id;
+ bool ret = false;
+
+ mutex_lock(&vgpu->gvt->lock);
+ for_each_active_vgpu(vgpu->gvt, itr, id) {
+ if (!itr->attached)
+ continue;
+
+ if (vgpu->vfio_device.kvm == itr->vfio_device.kvm) {
+ ret = true;
+ goto out;
+ }
+ }
+out:
+ mutex_unlock(&vgpu->gvt->lock);
+ return ret;
+}
+
+static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+
+ if (vgpu->attached)
+ return -EEXIST;
+
+ if (!vgpu->vfio_device.kvm ||
+ vgpu->vfio_device.kvm->mm != current->mm) {
+ gvt_vgpu_err("KVM is required to use Intel vGPU\n");
+ return -ESRCH;
+ }
+
+ if (__kvmgt_vgpu_exist(vgpu))
+ return -EEXIST;
+
+ vgpu->attached = true;
+
+ kvmgt_protect_table_init(vgpu);
+ gvt_cache_init(vgpu);
+
+ vgpu->track_node.track_write = kvmgt_page_track_write;
+ vgpu->track_node.track_flush_slot = kvmgt_page_track_flush_slot;
+ kvm_get_kvm(vgpu->vfio_device.kvm);
+ kvm_page_track_register_notifier(vgpu->vfio_device.kvm,
+ &vgpu->track_node);
+
+ debugfs_create_ulong(KVMGT_DEBUGFS_FILENAME, 0444, vgpu->debugfs,
+ &vgpu->nr_cache_entries);
+
+ intel_gvt_activate_vgpu(vgpu);
+
+ atomic_set(&vgpu->released, 0);
+ return 0;
+}
+
+static void intel_vgpu_release_msi_eventfd_ctx(struct intel_vgpu *vgpu)
+{
+ struct eventfd_ctx *trigger;
+
+ trigger = vgpu->msi_trigger;
+ if (trigger) {
+ eventfd_ctx_put(trigger);
+ vgpu->msi_trigger = NULL;
+ }
+}
+
+static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+
+ if (!vgpu->attached)
+ return;
+
+ if (atomic_cmpxchg(&vgpu->released, 0, 1))
+ return;
+
+ intel_gvt_release_vgpu(vgpu);
+
+ debugfs_remove(debugfs_lookup(KVMGT_DEBUGFS_FILENAME, vgpu->debugfs));
+
+ kvm_page_track_unregister_notifier(vgpu->vfio_device.kvm,
+ &vgpu->track_node);
+ kvm_put_kvm(vgpu->vfio_device.kvm);
+
+ kvmgt_protect_table_destroy(vgpu);
+ gvt_cache_destroy(vgpu);
+
+ intel_vgpu_release_msi_eventfd_ctx(vgpu);
+
+ vgpu->attached = false;
+}
+
+static u64 intel_vgpu_get_bar_addr(struct intel_vgpu *vgpu, int bar)
+{
+ u32 start_lo, start_hi;
+ u32 mem_type;
+
+ start_lo = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
+ PCI_BASE_ADDRESS_MEM_MASK;
+ mem_type = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
+ PCI_BASE_ADDRESS_MEM_TYPE_MASK;
+
+ switch (mem_type) {
+ case PCI_BASE_ADDRESS_MEM_TYPE_64:
+ start_hi = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space
+ + bar + 4));
+ break;
+ case PCI_BASE_ADDRESS_MEM_TYPE_32:
+ case PCI_BASE_ADDRESS_MEM_TYPE_1M:
+ /* 1M mem BAR treated as 32-bit BAR */
+ default:
+ /* mem unknown type treated as 32-bit BAR */
+ start_hi = 0;
+ break;
+ }
+
+ return ((u64)start_hi << 32) | start_lo;
+}
+
+static int intel_vgpu_bar_rw(struct intel_vgpu *vgpu, int bar, u64 off,
+ void *buf, unsigned int count, bool is_write)
+{
+ u64 bar_start = intel_vgpu_get_bar_addr(vgpu, bar);
+ int ret;
+
+ if (is_write)
+ ret = intel_vgpu_emulate_mmio_write(vgpu,
+ bar_start + off, buf, count);
+ else
+ ret = intel_vgpu_emulate_mmio_read(vgpu,
+ bar_start + off, buf, count);
+ return ret;
+}
+
+static inline bool intel_vgpu_in_aperture(struct intel_vgpu *vgpu, u64 off)
+{
+ return off >= vgpu_aperture_offset(vgpu) &&
+ off < vgpu_aperture_offset(vgpu) + vgpu_aperture_sz(vgpu);
+}
+
+static int intel_vgpu_aperture_rw(struct intel_vgpu *vgpu, u64 off,
+ void *buf, unsigned long count, bool is_write)
+{
+ void __iomem *aperture_va;
+
+ if (!intel_vgpu_in_aperture(vgpu, off) ||
+ !intel_vgpu_in_aperture(vgpu, off + count)) {
+ gvt_vgpu_err("Invalid aperture offset %llu\n", off);
+ return -EINVAL;
+ }
+
+ aperture_va = io_mapping_map_wc(&vgpu->gvt->gt->ggtt->iomap,
+ ALIGN_DOWN(off, PAGE_SIZE),
+ count + offset_in_page(off));
+ if (!aperture_va)
+ return -EIO;
+
+ if (is_write)
+ memcpy_toio(aperture_va + offset_in_page(off), buf, count);
+ else
+ memcpy_fromio(buf, aperture_va + offset_in_page(off), count);
+
+ io_mapping_unmap(aperture_va);
+
+ return 0;
+}
+
+static ssize_t intel_vgpu_rw(struct intel_vgpu *vgpu, char *buf,
+ size_t count, loff_t *ppos, bool is_write)
+{
+ unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos);
+ u64 pos = *ppos & VFIO_PCI_OFFSET_MASK;
+ int ret = -EINVAL;
+
+
+ if (index >= VFIO_PCI_NUM_REGIONS + vgpu->num_regions) {
+ gvt_vgpu_err("invalid index: %u\n", index);
+ return -EINVAL;
+ }
+
+ switch (index) {
+ case VFIO_PCI_CONFIG_REGION_INDEX:
+ if (is_write)
+ ret = intel_vgpu_emulate_cfg_write(vgpu, pos,
+ buf, count);
+ else
+ ret = intel_vgpu_emulate_cfg_read(vgpu, pos,
+ buf, count);
+ break;
+ case VFIO_PCI_BAR0_REGION_INDEX:
+ ret = intel_vgpu_bar_rw(vgpu, PCI_BASE_ADDRESS_0, pos,
+ buf, count, is_write);
+ break;
+ case VFIO_PCI_BAR2_REGION_INDEX:
+ ret = intel_vgpu_aperture_rw(vgpu, pos, buf, count, is_write);
+ break;
+ case VFIO_PCI_BAR1_REGION_INDEX:
+ case VFIO_PCI_BAR3_REGION_INDEX:
+ case VFIO_PCI_BAR4_REGION_INDEX:
+ case VFIO_PCI_BAR5_REGION_INDEX:
+ case VFIO_PCI_VGA_REGION_INDEX:
+ case VFIO_PCI_ROM_REGION_INDEX:
+ break;
+ default:
+ if (index >= VFIO_PCI_NUM_REGIONS + vgpu->num_regions)
+ return -EINVAL;
+
+ index -= VFIO_PCI_NUM_REGIONS;
+ return vgpu->region[index].ops->rw(vgpu, buf, count,
+ ppos, is_write);
+ }
+
+ return ret == 0 ? count : ret;
+}
+
+static bool gtt_entry(struct intel_vgpu *vgpu, loff_t *ppos)
+{
+ unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos);
+ struct intel_gvt *gvt = vgpu->gvt;
+ int offset;
+
+ /* Only allow MMIO GGTT entry access */
+ if (index != PCI_BASE_ADDRESS_0)
+ return false;
+
+ offset = (u64)(*ppos & VFIO_PCI_OFFSET_MASK) -
+ intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_0);
+
+ return (offset >= gvt->device_info.gtt_start_offset &&
+ offset < gvt->device_info.gtt_start_offset + gvt_ggtt_sz(gvt)) ?
+ true : false;
+}
+
+static ssize_t intel_vgpu_read(struct vfio_device *vfio_dev, char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ unsigned int done = 0;
+ int ret;
+
+ while (count) {
+ size_t filled;
+
+ /* Only support GGTT entry 8 bytes read */
+ if (count >= 8 && !(*ppos % 8) &&
+ gtt_entry(vgpu, ppos)) {
+ u64 val;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val, sizeof(val),
+ ppos, false);
+ if (ret <= 0)
+ goto read_err;
+
+ if (copy_to_user(buf, &val, sizeof(val)))
+ goto read_err;
+
+ filled = 8;
+ } else if (count >= 4 && !(*ppos % 4)) {
+ u32 val;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val, sizeof(val),
+ ppos, false);
+ if (ret <= 0)
+ goto read_err;
+
+ if (copy_to_user(buf, &val, sizeof(val)))
+ goto read_err;
+
+ filled = 4;
+ } else if (count >= 2 && !(*ppos % 2)) {
+ u16 val;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val, sizeof(val),
+ ppos, false);
+ if (ret <= 0)
+ goto read_err;
+
+ if (copy_to_user(buf, &val, sizeof(val)))
+ goto read_err;
+
+ filled = 2;
+ } else {
+ u8 val;
+
+ ret = intel_vgpu_rw(vgpu, &val, sizeof(val), ppos,
+ false);
+ if (ret <= 0)
+ goto read_err;
+
+ if (copy_to_user(buf, &val, sizeof(val)))
+ goto read_err;
+
+ filled = 1;
+ }
+
+ count -= filled;
+ done += filled;
+ *ppos += filled;
+ buf += filled;
+ }
+
+ return done;
+
+read_err:
+ return -EFAULT;
+}
+
+static ssize_t intel_vgpu_write(struct vfio_device *vfio_dev,
+ const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ unsigned int done = 0;
+ int ret;
+
+ while (count) {
+ size_t filled;
+
+ /* Only support GGTT entry 8 bytes write */
+ if (count >= 8 && !(*ppos % 8) &&
+ gtt_entry(vgpu, ppos)) {
+ u64 val;
+
+ if (copy_from_user(&val, buf, sizeof(val)))
+ goto write_err;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val, sizeof(val),
+ ppos, true);
+ if (ret <= 0)
+ goto write_err;
+
+ filled = 8;
+ } else if (count >= 4 && !(*ppos % 4)) {
+ u32 val;
+
+ if (copy_from_user(&val, buf, sizeof(val)))
+ goto write_err;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val, sizeof(val),
+ ppos, true);
+ if (ret <= 0)
+ goto write_err;
+
+ filled = 4;
+ } else if (count >= 2 && !(*ppos % 2)) {
+ u16 val;
+
+ if (copy_from_user(&val, buf, sizeof(val)))
+ goto write_err;
+
+ ret = intel_vgpu_rw(vgpu, (char *)&val,
+ sizeof(val), ppos, true);
+ if (ret <= 0)
+ goto write_err;
+
+ filled = 2;
+ } else {
+ u8 val;
+
+ if (copy_from_user(&val, buf, sizeof(val)))
+ goto write_err;
+
+ ret = intel_vgpu_rw(vgpu, &val, sizeof(val),
+ ppos, true);
+ if (ret <= 0)
+ goto write_err;
+
+ filled = 1;
+ }
+
+ count -= filled;
+ done += filled;
+ *ppos += filled;
+ buf += filled;
+ }
+
+ return done;
+write_err:
+ return -EFAULT;
+}
+
+static int intel_vgpu_mmap(struct vfio_device *vfio_dev,
+ struct vm_area_struct *vma)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ unsigned int index;
+ u64 virtaddr;
+ unsigned long req_size, pgoff, req_start;
+ pgprot_t pg_prot;
+
+ index = vma->vm_pgoff >> (VFIO_PCI_OFFSET_SHIFT - PAGE_SHIFT);
+ if (index >= VFIO_PCI_ROM_REGION_INDEX)
+ return -EINVAL;
+
+ if (vma->vm_end < vma->vm_start)
+ return -EINVAL;
+ if ((vma->vm_flags & VM_SHARED) == 0)
+ return -EINVAL;
+ if (index != VFIO_PCI_BAR2_REGION_INDEX)
+ return -EINVAL;
+
+ pg_prot = vma->vm_page_prot;
+ virtaddr = vma->vm_start;
+ req_size = vma->vm_end - vma->vm_start;
+ pgoff = vma->vm_pgoff &
+ ((1U << (VFIO_PCI_OFFSET_SHIFT - PAGE_SHIFT)) - 1);
+ req_start = pgoff << PAGE_SHIFT;
+
+ if (!intel_vgpu_in_aperture(vgpu, req_start))
+ return -EINVAL;
+ if (req_start + req_size >
+ vgpu_aperture_offset(vgpu) + vgpu_aperture_sz(vgpu))
+ return -EINVAL;
+
+ pgoff = (gvt_aperture_pa_base(vgpu->gvt) >> PAGE_SHIFT) + pgoff;
+
+ return remap_pfn_range(vma, virtaddr, pgoff, req_size, pg_prot);
+}
+
+static int intel_vgpu_get_irq_count(struct intel_vgpu *vgpu, int type)
+{
+ if (type == VFIO_PCI_INTX_IRQ_INDEX || type == VFIO_PCI_MSI_IRQ_INDEX)
+ return 1;
+
+ return 0;
+}
+
+static int intel_vgpu_set_intx_mask(struct intel_vgpu *vgpu,
+ unsigned int index, unsigned int start,
+ unsigned int count, u32 flags,
+ void *data)
+{
+ return 0;
+}
+
+static int intel_vgpu_set_intx_unmask(struct intel_vgpu *vgpu,
+ unsigned int index, unsigned int start,
+ unsigned int count, u32 flags, void *data)
+{
+ return 0;
+}
+
+static int intel_vgpu_set_intx_trigger(struct intel_vgpu *vgpu,
+ unsigned int index, unsigned int start, unsigned int count,
+ u32 flags, void *data)
+{
+ return 0;
+}
+
+static int intel_vgpu_set_msi_trigger(struct intel_vgpu *vgpu,
+ unsigned int index, unsigned int start, unsigned int count,
+ u32 flags, void *data)
+{
+ struct eventfd_ctx *trigger;
+
+ if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
+ int fd = *(int *)data;
+
+ trigger = eventfd_ctx_fdget(fd);
+ if (IS_ERR(trigger)) {
+ gvt_vgpu_err("eventfd_ctx_fdget failed\n");
+ return PTR_ERR(trigger);
+ }
+ vgpu->msi_trigger = trigger;
+ } else if ((flags & VFIO_IRQ_SET_DATA_NONE) && !count)
+ intel_vgpu_release_msi_eventfd_ctx(vgpu);
+
+ return 0;
+}
+
+static int intel_vgpu_set_irqs(struct intel_vgpu *vgpu, u32 flags,
+ unsigned int index, unsigned int start, unsigned int count,
+ void *data)
+{
+ int (*func)(struct intel_vgpu *vgpu, unsigned int index,
+ unsigned int start, unsigned int count, u32 flags,
+ void *data) = NULL;
+
+ switch (index) {
+ case VFIO_PCI_INTX_IRQ_INDEX:
+ switch (flags & VFIO_IRQ_SET_ACTION_TYPE_MASK) {
+ case VFIO_IRQ_SET_ACTION_MASK:
+ func = intel_vgpu_set_intx_mask;
+ break;
+ case VFIO_IRQ_SET_ACTION_UNMASK:
+ func = intel_vgpu_set_intx_unmask;
+ break;
+ case VFIO_IRQ_SET_ACTION_TRIGGER:
+ func = intel_vgpu_set_intx_trigger;
+ break;
+ }
+ break;
+ case VFIO_PCI_MSI_IRQ_INDEX:
+ switch (flags & VFIO_IRQ_SET_ACTION_TYPE_MASK) {
+ case VFIO_IRQ_SET_ACTION_MASK:
+ case VFIO_IRQ_SET_ACTION_UNMASK:
+ /* XXX Need masking support exported */
+ break;
+ case VFIO_IRQ_SET_ACTION_TRIGGER:
+ func = intel_vgpu_set_msi_trigger;
+ break;
+ }
+ break;
+ }
+
+ if (!func)
+ return -ENOTTY;
+
+ return func(vgpu, index, start, count, flags, data);
+}
+
+static long intel_vgpu_ioctl(struct vfio_device *vfio_dev, unsigned int cmd,
+ unsigned long arg)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ unsigned long minsz;
+
+ gvt_dbg_core("vgpu%d ioctl, cmd: %d\n", vgpu->id, cmd);
+
+ if (cmd == VFIO_DEVICE_GET_INFO) {
+ struct vfio_device_info info;
+
+ minsz = offsetofend(struct vfio_device_info, num_irqs);
+
+ if (copy_from_user(&info, (void __user *)arg, minsz))
+ return -EFAULT;
+
+ if (info.argsz < minsz)
+ return -EINVAL;
+
+ info.flags = VFIO_DEVICE_FLAGS_PCI;
+ info.flags |= VFIO_DEVICE_FLAGS_RESET;
+ info.num_regions = VFIO_PCI_NUM_REGIONS +
+ vgpu->num_regions;
+ info.num_irqs = VFIO_PCI_NUM_IRQS;
+
+ return copy_to_user((void __user *)arg, &info, minsz) ?
+ -EFAULT : 0;
+
+ } else if (cmd == VFIO_DEVICE_GET_REGION_INFO) {
+ struct vfio_region_info info;
+ struct vfio_info_cap caps = { .buf = NULL, .size = 0 };
+ unsigned int i;
+ int ret;
+ struct vfio_region_info_cap_sparse_mmap *sparse = NULL;
+ int nr_areas = 1;
+ int cap_type_id;
+
+ minsz = offsetofend(struct vfio_region_info, offset);
+
+ if (copy_from_user(&info, (void __user *)arg, minsz))
+ return -EFAULT;
+
+ if (info.argsz < minsz)
+ return -EINVAL;
+
+ switch (info.index) {
+ case VFIO_PCI_CONFIG_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = vgpu->gvt->device_info.cfg_space_size;
+ info.flags = VFIO_REGION_INFO_FLAG_READ |
+ VFIO_REGION_INFO_FLAG_WRITE;
+ break;
+ case VFIO_PCI_BAR0_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = vgpu->cfg_space.bar[info.index].size;
+ if (!info.size) {
+ info.flags = 0;
+ break;
+ }
+
+ info.flags = VFIO_REGION_INFO_FLAG_READ |
+ VFIO_REGION_INFO_FLAG_WRITE;
+ break;
+ case VFIO_PCI_BAR1_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = 0;
+ info.flags = 0;
+ break;
+ case VFIO_PCI_BAR2_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.flags = VFIO_REGION_INFO_FLAG_CAPS |
+ VFIO_REGION_INFO_FLAG_MMAP |
+ VFIO_REGION_INFO_FLAG_READ |
+ VFIO_REGION_INFO_FLAG_WRITE;
+ info.size = gvt_aperture_sz(vgpu->gvt);
+
+ sparse = kzalloc(struct_size(sparse, areas, nr_areas),
+ GFP_KERNEL);
+ if (!sparse)
+ return -ENOMEM;
+
+ sparse->header.id = VFIO_REGION_INFO_CAP_SPARSE_MMAP;
+ sparse->header.version = 1;
+ sparse->nr_areas = nr_areas;
+ cap_type_id = VFIO_REGION_INFO_CAP_SPARSE_MMAP;
+ sparse->areas[0].offset =
+ PAGE_ALIGN(vgpu_aperture_offset(vgpu));
+ sparse->areas[0].size = vgpu_aperture_sz(vgpu);
+ break;
+
+ case VFIO_PCI_BAR3_REGION_INDEX ... VFIO_PCI_BAR5_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = 0;
+ info.flags = 0;
+
+ gvt_dbg_core("get region info bar:%d\n", info.index);
+ break;
+
+ case VFIO_PCI_ROM_REGION_INDEX:
+ case VFIO_PCI_VGA_REGION_INDEX:
+ info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = 0;
+ info.flags = 0;
+
+ gvt_dbg_core("get region info index:%d\n", info.index);
+ break;
+ default:
+ {
+ struct vfio_region_info_cap_type cap_type = {
+ .header.id = VFIO_REGION_INFO_CAP_TYPE,
+ .header.version = 1 };
+
+ if (info.index >= VFIO_PCI_NUM_REGIONS +
+ vgpu->num_regions)
+ return -EINVAL;
+ info.index =
+ array_index_nospec(info.index,
+ VFIO_PCI_NUM_REGIONS +
+ vgpu->num_regions);
+
+ i = info.index - VFIO_PCI_NUM_REGIONS;
+
+ info.offset =
+ VFIO_PCI_INDEX_TO_OFFSET(info.index);
+ info.size = vgpu->region[i].size;
+ info.flags = vgpu->region[i].flags;
+
+ cap_type.type = vgpu->region[i].type;
+ cap_type.subtype = vgpu->region[i].subtype;
+
+ ret = vfio_info_add_capability(&caps,
+ &cap_type.header,
+ sizeof(cap_type));
+ if (ret)
+ return ret;
+ }
+ }
+
+ if ((info.flags & VFIO_REGION_INFO_FLAG_CAPS) && sparse) {
+ switch (cap_type_id) {
+ case VFIO_REGION_INFO_CAP_SPARSE_MMAP:
+ ret = vfio_info_add_capability(&caps,
+ &sparse->header,
+ struct_size(sparse, areas,
+ sparse->nr_areas));
+ if (ret) {
+ kfree(sparse);
+ return ret;
+ }
+ break;
+ default:
+ kfree(sparse);
+ return -EINVAL;
+ }
+ }
+
+ if (caps.size) {
+ info.flags |= VFIO_REGION_INFO_FLAG_CAPS;
+ if (info.argsz < sizeof(info) + caps.size) {
+ info.argsz = sizeof(info) + caps.size;
+ info.cap_offset = 0;
+ } else {
+ vfio_info_cap_shift(&caps, sizeof(info));
+ if (copy_to_user((void __user *)arg +
+ sizeof(info), caps.buf,
+ caps.size)) {
+ kfree(caps.buf);
+ kfree(sparse);
+ return -EFAULT;
+ }
+ info.cap_offset = sizeof(info);
+ }
+
+ kfree(caps.buf);
+ }
+
+ kfree(sparse);
+ return copy_to_user((void __user *)arg, &info, minsz) ?
+ -EFAULT : 0;
+ } else if (cmd == VFIO_DEVICE_GET_IRQ_INFO) {
+ struct vfio_irq_info info;
+
+ minsz = offsetofend(struct vfio_irq_info, count);
+
+ if (copy_from_user(&info, (void __user *)arg, minsz))
+ return -EFAULT;
+
+ if (info.argsz < minsz || info.index >= VFIO_PCI_NUM_IRQS)
+ return -EINVAL;
+
+ switch (info.index) {
+ case VFIO_PCI_INTX_IRQ_INDEX:
+ case VFIO_PCI_MSI_IRQ_INDEX:
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ info.flags = VFIO_IRQ_INFO_EVENTFD;
+
+ info.count = intel_vgpu_get_irq_count(vgpu, info.index);
+
+ if (info.index == VFIO_PCI_INTX_IRQ_INDEX)
+ info.flags |= (VFIO_IRQ_INFO_MASKABLE |
+ VFIO_IRQ_INFO_AUTOMASKED);
+ else
+ info.flags |= VFIO_IRQ_INFO_NORESIZE;
+
+ return copy_to_user((void __user *)arg, &info, minsz) ?
+ -EFAULT : 0;
+ } else if (cmd == VFIO_DEVICE_SET_IRQS) {
+ struct vfio_irq_set hdr;
+ u8 *data = NULL;
+ int ret = 0;
+ size_t data_size = 0;
+
+ minsz = offsetofend(struct vfio_irq_set, count);
+
+ if (copy_from_user(&hdr, (void __user *)arg, minsz))
+ return -EFAULT;
+
+ if (!(hdr.flags & VFIO_IRQ_SET_DATA_NONE)) {
+ int max = intel_vgpu_get_irq_count(vgpu, hdr.index);
+
+ ret = vfio_set_irqs_validate_and_prepare(&hdr, max,
+ VFIO_PCI_NUM_IRQS, &data_size);
+ if (ret) {
+ gvt_vgpu_err("intel:vfio_set_irqs_validate_and_prepare failed\n");
+ return -EINVAL;
+ }
+ if (data_size) {
+ data = memdup_user((void __user *)(arg + minsz),
+ data_size);
+ if (IS_ERR(data))
+ return PTR_ERR(data);
+ }
+ }
+
+ ret = intel_vgpu_set_irqs(vgpu, hdr.flags, hdr.index,
+ hdr.start, hdr.count, data);
+ kfree(data);
+
+ return ret;
+ } else if (cmd == VFIO_DEVICE_RESET) {
+ intel_gvt_reset_vgpu(vgpu);
+ return 0;
+ } else if (cmd == VFIO_DEVICE_QUERY_GFX_PLANE) {
+ struct vfio_device_gfx_plane_info dmabuf;
+ int ret = 0;
+
+ minsz = offsetofend(struct vfio_device_gfx_plane_info,
+ dmabuf_id);
+ if (copy_from_user(&dmabuf, (void __user *)arg, minsz))
+ return -EFAULT;
+ if (dmabuf.argsz < minsz)
+ return -EINVAL;
+
+ ret = intel_vgpu_query_plane(vgpu, &dmabuf);
+ if (ret != 0)
+ return ret;
+
+ return copy_to_user((void __user *)arg, &dmabuf, minsz) ?
+ -EFAULT : 0;
+ } else if (cmd == VFIO_DEVICE_GET_GFX_DMABUF) {
+ __u32 dmabuf_id;
+
+ if (get_user(dmabuf_id, (__u32 __user *)arg))
+ return -EFAULT;
+ return intel_vgpu_get_dmabuf(vgpu, dmabuf_id);
+ }
+
+ return -ENOTTY;
+}
+
+static ssize_t
+vgpu_id_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct intel_vgpu *vgpu = dev_get_drvdata(dev);
+
+ return sprintf(buf, "%d\n", vgpu->id);
+}
+
+static DEVICE_ATTR_RO(vgpu_id);
+
+static struct attribute *intel_vgpu_attrs[] = {
+ &dev_attr_vgpu_id.attr,
+ NULL
+};
+
+static const struct attribute_group intel_vgpu_group = {
+ .name = "intel_vgpu",
+ .attrs = intel_vgpu_attrs,
+};
+
+static const struct attribute_group *intel_vgpu_groups[] = {
+ &intel_vgpu_group,
+ NULL,
+};
+
+static int intel_vgpu_init_dev(struct vfio_device *vfio_dev)
+{
+ struct mdev_device *mdev = to_mdev_device(vfio_dev->dev);
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+ struct intel_vgpu_type *type =
+ container_of(mdev->type, struct intel_vgpu_type, type);
+
+ vgpu->gvt = kdev_to_i915(mdev->type->parent->dev)->gvt;
+ return intel_gvt_create_vgpu(vgpu, type->conf);
+}
+
+static void intel_vgpu_release_dev(struct vfio_device *vfio_dev)
+{
+ struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
+
+ intel_gvt_destroy_vgpu(vgpu);
+ vfio_free_device(vfio_dev);
+}
+
+static const struct vfio_device_ops intel_vgpu_dev_ops = {
+ .init = intel_vgpu_init_dev,
+ .release = intel_vgpu_release_dev,
+ .open_device = intel_vgpu_open_device,
+ .close_device = intel_vgpu_close_device,
+ .read = intel_vgpu_read,
+ .write = intel_vgpu_write,
+ .mmap = intel_vgpu_mmap,
+ .ioctl = intel_vgpu_ioctl,
+ .dma_unmap = intel_vgpu_dma_unmap,
+};
+
+static int intel_vgpu_probe(struct mdev_device *mdev)
+{
+ struct intel_vgpu *vgpu;
+ int ret;
+
+ vgpu = vfio_alloc_device(intel_vgpu, vfio_device, &mdev->dev,
+ &intel_vgpu_dev_ops);
+ if (IS_ERR(vgpu)) {
+ gvt_err("failed to create intel vgpu: %ld\n", PTR_ERR(vgpu));
+ return PTR_ERR(vgpu);
+ }
+
+ dev_set_drvdata(&mdev->dev, vgpu);
+ ret = vfio_register_emulated_iommu_dev(&vgpu->vfio_device);
+ if (ret)
+ goto out_put_vdev;
+
+ gvt_dbg_core("intel_vgpu_create succeeded for mdev: %s\n",
+ dev_name(mdev_dev(mdev)));
+ return 0;
+
+out_put_vdev:
+ vfio_put_device(&vgpu->vfio_device);
+ return ret;
+}
+
+static void intel_vgpu_remove(struct mdev_device *mdev)
+{
+ struct intel_vgpu *vgpu = dev_get_drvdata(&mdev->dev);
+
+ if (WARN_ON_ONCE(vgpu->attached))
+ return;
+
+ vfio_unregister_group_dev(&vgpu->vfio_device);
+ vfio_put_device(&vgpu->vfio_device);
+}
+
+static unsigned int intel_vgpu_get_available(struct mdev_type *mtype)
+{
+ struct intel_vgpu_type *type =
+ container_of(mtype, struct intel_vgpu_type, type);
+ struct intel_gvt *gvt = kdev_to_i915(mtype->parent->dev)->gvt;
+ unsigned int low_gm_avail, high_gm_avail, fence_avail;
+
+ mutex_lock(&gvt->lock);
+ low_gm_avail = gvt_aperture_sz(gvt) - HOST_LOW_GM_SIZE -
+ gvt->gm.vgpu_allocated_low_gm_size;
+ high_gm_avail = gvt_hidden_sz(gvt) - HOST_HIGH_GM_SIZE -
+ gvt->gm.vgpu_allocated_high_gm_size;
+ fence_avail = gvt_fence_sz(gvt) - HOST_FENCE -
+ gvt->fence.vgpu_allocated_fence_num;
+ mutex_unlock(&gvt->lock);
+
+ return min3(low_gm_avail / type->conf->low_mm,
+ high_gm_avail / type->conf->high_mm,
+ fence_avail / type->conf->fence);
+}
+
+static struct mdev_driver intel_vgpu_mdev_driver = {
+ .device_api = VFIO_DEVICE_API_PCI_STRING,
+ .driver = {
+ .name = "intel_vgpu_mdev",
+ .owner = THIS_MODULE,
+ .dev_groups = intel_vgpu_groups,
+ },
+ .probe = intel_vgpu_probe,
+ .remove = intel_vgpu_remove,
+ .get_available = intel_vgpu_get_available,
+ .show_description = intel_vgpu_show_description,
+};
+
+int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
+{
+ struct kvm *kvm = info->vfio_device.kvm;
+ struct kvm_memory_slot *slot;
+ int idx;
+
+ if (!info->attached)
+ return -ESRCH;
+
+ idx = srcu_read_lock(&kvm->srcu);
+ slot = gfn_to_memslot(kvm, gfn);
+ if (!slot) {
+ srcu_read_unlock(&kvm->srcu, idx);
+ return -EINVAL;
+ }
+
+ write_lock(&kvm->mmu_lock);
+
+ if (kvmgt_gfn_is_write_protected(info, gfn))
+ goto out;
+
+ kvm_slot_page_track_add_page(kvm, slot, gfn, KVM_PAGE_TRACK_WRITE);
+ kvmgt_protect_table_add(info, gfn);
+
+out:
+ write_unlock(&kvm->mmu_lock);
+ srcu_read_unlock(&kvm->srcu, idx);
+ return 0;
+}
+
+int intel_gvt_page_track_remove(struct intel_vgpu *info, u64 gfn)
+{
+ struct kvm *kvm = info->vfio_device.kvm;
+ struct kvm_memory_slot *slot;
+ int idx;
+
+ if (!info->attached)
+ return 0;
+
+ idx = srcu_read_lock(&kvm->srcu);
+ slot = gfn_to_memslot(kvm, gfn);
+ if (!slot) {
+ srcu_read_unlock(&kvm->srcu, idx);
+ return -EINVAL;
+ }
+
+ write_lock(&kvm->mmu_lock);
+
+ if (!kvmgt_gfn_is_write_protected(info, gfn))
+ goto out;
+
+ kvm_slot_page_track_remove_page(kvm, slot, gfn, KVM_PAGE_TRACK_WRITE);
+ kvmgt_protect_table_del(info, gfn);
+
+out:
+ write_unlock(&kvm->mmu_lock);
+ srcu_read_unlock(&kvm->srcu, idx);
+ return 0;
+}
+
+static void kvmgt_page_track_write(struct kvm_vcpu *vcpu, gpa_t gpa,
+ const u8 *val, int len,
+ struct kvm_page_track_notifier_node *node)
+{
+ struct intel_vgpu *info =
+ container_of(node, struct intel_vgpu, track_node);
+
+ if (kvmgt_gfn_is_write_protected(info, gpa_to_gfn(gpa)))
+ intel_vgpu_page_track_handler(info, gpa,
+ (void *)val, len);
+}
+
+static void kvmgt_page_track_flush_slot(struct kvm *kvm,
+ struct kvm_memory_slot *slot,
+ struct kvm_page_track_notifier_node *node)
+{
+ int i;
+ gfn_t gfn;
+ struct intel_vgpu *info =
+ container_of(node, struct intel_vgpu, track_node);
+
+ write_lock(&kvm->mmu_lock);
+ for (i = 0; i < slot->npages; i++) {
+ gfn = slot->base_gfn + i;
+ if (kvmgt_gfn_is_write_protected(info, gfn)) {
+ kvm_slot_page_track_remove_page(kvm, slot, gfn,
+ KVM_PAGE_TRACK_WRITE);
+ kvmgt_protect_table_del(info, gfn);
+ }
+ }
+ write_unlock(&kvm->mmu_lock);
+}
+
+void intel_vgpu_detach_regions(struct intel_vgpu *vgpu)
+{
+ int i;
+
+ if (!vgpu->region)
+ return;
+
+ for (i = 0; i < vgpu->num_regions; i++)
+ if (vgpu->region[i].ops->release)
+ vgpu->region[i].ops->release(vgpu,
+ &vgpu->region[i]);
+ vgpu->num_regions = 0;
+ kfree(vgpu->region);
+ vgpu->region = NULL;
+}
+
+int intel_gvt_dma_map_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+ unsigned long size, dma_addr_t *dma_addr)
+{
+ struct gvt_dma *entry;
+ int ret;
+
+ if (!vgpu->attached)
+ return -EINVAL;
+
+ mutex_lock(&vgpu->cache_lock);
+
+ entry = __gvt_cache_find_gfn(vgpu, gfn);
+ if (!entry) {
+ ret = gvt_dma_map_page(vgpu, gfn, dma_addr, size);
+ if (ret)
+ goto err_unlock;
+
+ ret = __gvt_cache_add(vgpu, gfn, *dma_addr, size);
+ if (ret)
+ goto err_unmap;
+ } else if (entry->size != size) {
+ /* the same gfn with different size: unmap and re-map */
+ gvt_dma_unmap_page(vgpu, gfn, entry->dma_addr, entry->size);
+ __gvt_cache_remove_entry(vgpu, entry);
+
+ ret = gvt_dma_map_page(vgpu, gfn, dma_addr, size);
+ if (ret)
+ goto err_unlock;
+
+ ret = __gvt_cache_add(vgpu, gfn, *dma_addr, size);
+ if (ret)
+ goto err_unmap;
+ } else {
+ kref_get(&entry->ref);
+ *dma_addr = entry->dma_addr;
+ }
+
+ mutex_unlock(&vgpu->cache_lock);
+ return 0;
+
+err_unmap:
+ gvt_dma_unmap_page(vgpu, gfn, *dma_addr, size);
+err_unlock:
+ mutex_unlock(&vgpu->cache_lock);
+ return ret;
+}
+
+int intel_gvt_dma_pin_guest_page(struct intel_vgpu *vgpu, dma_addr_t dma_addr)
+{
+ struct gvt_dma *entry;
+ int ret = 0;
+
+ if (!vgpu->attached)
+ return -ENODEV;
+
+ mutex_lock(&vgpu->cache_lock);
+ entry = __gvt_cache_find_dma_addr(vgpu, dma_addr);
+ if (entry)
+ kref_get(&entry->ref);
+ else
+ ret = -ENOMEM;
+ mutex_unlock(&vgpu->cache_lock);
+
+ return ret;
+}
+
+static void __gvt_dma_release(struct kref *ref)
+{
+ struct gvt_dma *entry = container_of(ref, typeof(*entry), ref);
+
+ gvt_dma_unmap_page(entry->vgpu, entry->gfn, entry->dma_addr,
+ entry->size);
+ __gvt_cache_remove_entry(entry->vgpu, entry);
+}
+
+void intel_gvt_dma_unmap_guest_page(struct intel_vgpu *vgpu,
+ dma_addr_t dma_addr)
+{
+ struct gvt_dma *entry;
+
+ if (!vgpu->attached)
+ return;
+
+ mutex_lock(&vgpu->cache_lock);
+ entry = __gvt_cache_find_dma_addr(vgpu, dma_addr);
+ if (entry)
+ kref_put(&entry->ref, __gvt_dma_release);
+ mutex_unlock(&vgpu->cache_lock);
+}
+
+static void init_device_info(struct intel_gvt *gvt)
+{
+ struct intel_gvt_device_info *info = &gvt->device_info;
+ struct pci_dev *pdev = to_pci_dev(gvt->gt->i915->drm.dev);
+
+ info->max_support_vgpus = 8;
+ info->cfg_space_size = PCI_CFG_SPACE_EXP_SIZE;
+ info->mmio_size = 2 * 1024 * 1024;
+ info->mmio_bar = 0;
+ info->gtt_start_offset = 8 * 1024 * 1024;
+ info->gtt_entry_size = 8;
+ info->gtt_entry_size_shift = 3;
+ info->gmadr_bytes_in_cmd = 8;
+ info->max_surface_size = 36 * 1024 * 1024;
+ info->msi_cap_offset = pdev->msi_cap;
+}
+
+static void intel_gvt_test_and_emulate_vblank(struct intel_gvt *gvt)
+{
+ struct intel_vgpu *vgpu;
+ int id;
+
+ mutex_lock(&gvt->lock);
+ idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) {
+ if (test_and_clear_bit(INTEL_GVT_REQUEST_EMULATE_VBLANK + id,
+ (void *)&gvt->service_request)) {
+ if (vgpu->active)
+ intel_vgpu_emulate_vblank(vgpu);
+ }
+ }
+ mutex_unlock(&gvt->lock);
+}
+
+static int gvt_service_thread(void *data)
+{
+ struct intel_gvt *gvt = (struct intel_gvt *)data;
+ int ret;
+
+ gvt_dbg_core("service thread start\n");
+
+ while (!kthread_should_stop()) {
+ ret = wait_event_interruptible(gvt->service_thread_wq,
+ kthread_should_stop() || gvt->service_request);
+
+ if (kthread_should_stop())
+ break;
+
+ if (WARN_ONCE(ret, "service thread is waken up by signal.\n"))
+ continue;
+
+ intel_gvt_test_and_emulate_vblank(gvt);
+
+ if (test_bit(INTEL_GVT_REQUEST_SCHED,
+ (void *)&gvt->service_request) ||
+ test_bit(INTEL_GVT_REQUEST_EVENT_SCHED,
+ (void *)&gvt->service_request)) {
+ intel_gvt_schedule(gvt);
+ }
+ }
+
+ return 0;
+}
+
+static void clean_service_thread(struct intel_gvt *gvt)
+{
+ kthread_stop(gvt->service_thread);
+}
+
+static int init_service_thread(struct intel_gvt *gvt)
+{
+ init_waitqueue_head(&gvt->service_thread_wq);
+
+ gvt->service_thread = kthread_run(gvt_service_thread,
+ gvt, "gvt_service_thread");
+ if (IS_ERR(gvt->service_thread)) {
+ gvt_err("fail to start service thread.\n");
+ return PTR_ERR(gvt->service_thread);
+ }
+ return 0;
+}
+
+/**
+ * intel_gvt_clean_device - clean a GVT device
+ * @i915: i915 private
+ *
+ * This function is called at the driver unloading stage, to free the
+ * resources owned by a GVT device.
+ *
+ */
+static void intel_gvt_clean_device(struct drm_i915_private *i915)
+{
+ struct intel_gvt *gvt = fetch_and_zero(&i915->gvt);
+
+ if (drm_WARN_ON(&i915->drm, !gvt))
+ return;
+
+ mdev_unregister_parent(&gvt->parent);
+ intel_gvt_destroy_idle_vgpu(gvt->idle_vgpu);
+ intel_gvt_clean_vgpu_types(gvt);
+
+ intel_gvt_debugfs_clean(gvt);
+ clean_service_thread(gvt);
+ intel_gvt_clean_cmd_parser(gvt);
+ intel_gvt_clean_sched_policy(gvt);
+ intel_gvt_clean_workload_scheduler(gvt);
+ intel_gvt_clean_gtt(gvt);
+ intel_gvt_free_firmware(gvt);
+ intel_gvt_clean_mmio_info(gvt);
+ idr_destroy(&gvt->vgpu_idr);
+
+ kfree(i915->gvt);
+}
+
+/**
+ * intel_gvt_init_device - initialize a GVT device
+ * @i915: drm i915 private data
+ *
+ * This function is called at the initialization stage, to initialize
+ * necessary GVT components.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+static int intel_gvt_init_device(struct drm_i915_private *i915)
+{
+ struct intel_gvt *gvt;
+ struct intel_vgpu *vgpu;
+ int ret;
+
+ if (drm_WARN_ON(&i915->drm, i915->gvt))
+ return -EEXIST;
+
+ gvt = kzalloc(sizeof(struct intel_gvt), GFP_KERNEL);
+ if (!gvt)
+ return -ENOMEM;
+
+ gvt_dbg_core("init gvt device\n");
+
+ idr_init_base(&gvt->vgpu_idr, 1);
+ spin_lock_init(&gvt->scheduler.mmio_context_lock);
+ mutex_init(&gvt->lock);
+ mutex_init(&gvt->sched_lock);
+ gvt->gt = to_gt(i915);
+ i915->gvt = gvt;
+
+ init_device_info(gvt);
+
+ ret = intel_gvt_setup_mmio_info(gvt);
+ if (ret)
+ goto out_clean_idr;
+
+ intel_gvt_init_engine_mmio_context(gvt);
+
+ ret = intel_gvt_load_firmware(gvt);
+ if (ret)
+ goto out_clean_mmio_info;
+
+ ret = intel_gvt_init_irq(gvt);
+ if (ret)
+ goto out_free_firmware;
+
+ ret = intel_gvt_init_gtt(gvt);
+ if (ret)
+ goto out_free_firmware;
+
+ ret = intel_gvt_init_workload_scheduler(gvt);
+ if (ret)
+ goto out_clean_gtt;
+
+ ret = intel_gvt_init_sched_policy(gvt);
+ if (ret)
+ goto out_clean_workload_scheduler;
+
+ ret = intel_gvt_init_cmd_parser(gvt);
+ if (ret)
+ goto out_clean_sched_policy;
+
+ ret = init_service_thread(gvt);
+ if (ret)
+ goto out_clean_cmd_parser;
+
+ ret = intel_gvt_init_vgpu_types(gvt);
+ if (ret)
+ goto out_clean_thread;
+
+ vgpu = intel_gvt_create_idle_vgpu(gvt);
+ if (IS_ERR(vgpu)) {
+ ret = PTR_ERR(vgpu);
+ gvt_err("failed to create idle vgpu\n");
+ goto out_clean_types;
+ }
+ gvt->idle_vgpu = vgpu;
+
+ intel_gvt_debugfs_init(gvt);
+
+ ret = mdev_register_parent(&gvt->parent, i915->drm.dev,
+ &intel_vgpu_mdev_driver,
+ gvt->mdev_types, gvt->num_types);
+ if (ret)
+ goto out_destroy_idle_vgpu;
+
+ gvt_dbg_core("gvt device initialization is done\n");
+ return 0;
+
+out_destroy_idle_vgpu:
+ intel_gvt_destroy_idle_vgpu(gvt->idle_vgpu);
+ intel_gvt_debugfs_clean(gvt);
+out_clean_types:
+ intel_gvt_clean_vgpu_types(gvt);
+out_clean_thread:
+ clean_service_thread(gvt);
+out_clean_cmd_parser:
+ intel_gvt_clean_cmd_parser(gvt);
+out_clean_sched_policy:
+ intel_gvt_clean_sched_policy(gvt);
+out_clean_workload_scheduler:
+ intel_gvt_clean_workload_scheduler(gvt);
+out_clean_gtt:
+ intel_gvt_clean_gtt(gvt);
+out_free_firmware:
+ intel_gvt_free_firmware(gvt);
+out_clean_mmio_info:
+ intel_gvt_clean_mmio_info(gvt);
+out_clean_idr:
+ idr_destroy(&gvt->vgpu_idr);
+ kfree(gvt);
+ i915->gvt = NULL;
+ return ret;
+}
+
+static void intel_gvt_pm_resume(struct drm_i915_private *i915)
+{
+ struct intel_gvt *gvt = i915->gvt;
+
+ intel_gvt_restore_fence(gvt);
+ intel_gvt_restore_mmio(gvt);
+ intel_gvt_restore_ggtt(gvt);
+}
+
+static const struct intel_vgpu_ops intel_gvt_vgpu_ops = {
+ .init_device = intel_gvt_init_device,
+ .clean_device = intel_gvt_clean_device,
+ .pm_resume = intel_gvt_pm_resume,
+};
+
+static int __init kvmgt_init(void)
+{
+ int ret;
+
+ ret = intel_gvt_set_ops(&intel_gvt_vgpu_ops);
+ if (ret)
+ return ret;
+
+ ret = mdev_register_driver(&intel_vgpu_mdev_driver);
+ if (ret)
+ intel_gvt_clear_ops(&intel_gvt_vgpu_ops);
+ return ret;
+}
+
+static void __exit kvmgt_exit(void)
+{
+ mdev_unregister_driver(&intel_vgpu_mdev_driver);
+ intel_gvt_clear_ops(&intel_gvt_vgpu_ops);
+}
+
+module_init(kvmgt_init);
+module_exit(kvmgt_exit);
+
+MODULE_LICENSE("GPL and additional rights");
+MODULE_AUTHOR("Intel Corporation");
diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c
new file mode 100644
index 000000000..9acc00505
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/mmio.c
@@ -0,0 +1,332 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Kevin Tian <kevin.tian@intel.com>
+ * Dexuan Cui
+ *
+ * Contributors:
+ * Tina Zhang <tina.zhang@intel.com>
+ * Min He <min.he@intel.com>
+ * Niu Bing <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "i915_reg.h"
+#include "gvt.h"
+
+#include "gt/intel_gt_regs.h"
+
+/**
+ * intel_vgpu_gpa_to_mmio_offset - translate a GPA to MMIO offset
+ * @vgpu: a vGPU
+ * @gpa: guest physical address
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa)
+{
+ u64 gttmmio_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_0);
+ return gpa - gttmmio_gpa;
+}
+
+#define reg_is_mmio(gvt, reg) \
+ (reg >= 0 && reg < gvt->device_info.mmio_size)
+
+#define reg_is_gtt(gvt, reg) \
+ (reg >= gvt->device_info.gtt_start_offset \
+ && reg < gvt->device_info.gtt_start_offset + gvt_ggtt_sz(gvt))
+
+static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, u64 pa,
+ void *p_data, unsigned int bytes, bool read)
+{
+ struct intel_gvt *gvt = NULL;
+ void *pt = NULL;
+ unsigned int offset = 0;
+
+ if (!vgpu || !p_data)
+ return;
+
+ gvt = vgpu->gvt;
+ mutex_lock(&vgpu->vgpu_lock);
+ offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
+ if (reg_is_mmio(gvt, offset)) {
+ if (read)
+ intel_vgpu_default_mmio_read(vgpu, offset, p_data,
+ bytes);
+ else
+ intel_vgpu_default_mmio_write(vgpu, offset, p_data,
+ bytes);
+ } else if (reg_is_gtt(gvt, offset)) {
+ offset -= gvt->device_info.gtt_start_offset;
+ pt = vgpu->gtt.ggtt_mm->ggtt_mm.virtual_ggtt + offset;
+ if (read)
+ memcpy(p_data, pt, bytes);
+ else
+ memcpy(pt, p_data, bytes);
+
+ }
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+/**
+ * intel_vgpu_emulate_mmio_read - emulate MMIO read
+ * @vgpu: a vGPU
+ * @pa: guest physical address
+ * @p_data: data return buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, u64 pa,
+ void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ unsigned int offset = 0;
+ int ret = -EINVAL;
+
+ if (vgpu->failsafe) {
+ failsafe_emulate_mmio_rw(vgpu, pa, p_data, bytes, true);
+ return 0;
+ }
+ mutex_lock(&vgpu->vgpu_lock);
+
+ offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
+
+ if (drm_WARN_ON(&i915->drm, bytes > 8))
+ goto err;
+
+ if (reg_is_gtt(gvt, offset)) {
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4) &&
+ !IS_ALIGNED(offset, 8)))
+ goto err;
+ if (drm_WARN_ON(&i915->drm, bytes != 4 && bytes != 8))
+ goto err;
+ if (drm_WARN_ON(&i915->drm,
+ !reg_is_gtt(gvt, offset + bytes - 1)))
+ goto err;
+
+ ret = intel_vgpu_emulate_ggtt_mmio_read(vgpu, offset,
+ p_data, bytes);
+ if (ret)
+ goto err;
+ goto out;
+ }
+
+ if (drm_WARN_ON_ONCE(&i915->drm, !reg_is_mmio(gvt, offset))) {
+ ret = intel_gvt_read_gpa(vgpu, pa, p_data, bytes);
+ goto out;
+ }
+
+ if (drm_WARN_ON(&i915->drm, !reg_is_mmio(gvt, offset + bytes - 1)))
+ goto err;
+
+ if (!intel_gvt_mmio_is_unalign(gvt, offset)) {
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, bytes)))
+ goto err;
+ }
+
+ ret = intel_vgpu_mmio_reg_rw(vgpu, offset, p_data, bytes, true);
+ if (ret < 0)
+ goto err;
+
+ intel_gvt_mmio_set_accessed(gvt, offset);
+ ret = 0;
+ goto out;
+
+err:
+ gvt_vgpu_err("fail to emulate MMIO read %08x len %d\n",
+ offset, bytes);
+out:
+ mutex_unlock(&vgpu->vgpu_lock);
+ return ret;
+}
+
+/**
+ * intel_vgpu_emulate_mmio_write - emulate MMIO write
+ * @vgpu: a vGPU
+ * @pa: guest physical address
+ * @p_data: write data buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, u64 pa,
+ void *p_data, unsigned int bytes)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+ unsigned int offset = 0;
+ int ret = -EINVAL;
+
+ if (vgpu->failsafe) {
+ failsafe_emulate_mmio_rw(vgpu, pa, p_data, bytes, false);
+ return 0;
+ }
+
+ mutex_lock(&vgpu->vgpu_lock);
+
+ offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
+
+ if (drm_WARN_ON(&i915->drm, bytes > 8))
+ goto err;
+
+ if (reg_is_gtt(gvt, offset)) {
+ if (drm_WARN_ON(&i915->drm, !IS_ALIGNED(offset, 4) &&
+ !IS_ALIGNED(offset, 8)))
+ goto err;
+ if (drm_WARN_ON(&i915->drm, bytes != 4 && bytes != 8))
+ goto err;
+ if (drm_WARN_ON(&i915->drm,
+ !reg_is_gtt(gvt, offset + bytes - 1)))
+ goto err;
+
+ ret = intel_vgpu_emulate_ggtt_mmio_write(vgpu, offset,
+ p_data, bytes);
+ if (ret)
+ goto err;
+ goto out;
+ }
+
+ if (drm_WARN_ON_ONCE(&i915->drm, !reg_is_mmio(gvt, offset))) {
+ ret = intel_gvt_write_gpa(vgpu, pa, p_data, bytes);
+ goto out;
+ }
+
+ ret = intel_vgpu_mmio_reg_rw(vgpu, offset, p_data, bytes, false);
+ if (ret < 0)
+ goto err;
+
+ intel_gvt_mmio_set_accessed(gvt, offset);
+ ret = 0;
+ goto out;
+err:
+ gvt_vgpu_err("fail to emulate MMIO write %08x len %d\n", offset,
+ bytes);
+out:
+ mutex_unlock(&vgpu->vgpu_lock);
+ return ret;
+}
+
+
+/**
+ * intel_vgpu_reset_mmio - reset virtual MMIO space
+ * @vgpu: a vGPU
+ * @dmlr: whether this is device model level reset
+ */
+void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ const struct intel_gvt_device_info *info = &gvt->device_info;
+ void *mmio = gvt->firmware.mmio;
+
+ if (dmlr) {
+ memcpy(vgpu->mmio.vreg, mmio, info->mmio_size);
+
+ vgpu_vreg_t(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0;
+
+ /* set the bit 0:2(Core C-State ) to C0 */
+ vgpu_vreg_t(vgpu, GEN6_GT_CORE_STATUS) = 0;
+
+ /* uc reset hw expect GS_MIA_IN_RESET */
+ vgpu_vreg_t(vgpu, GUC_STATUS) |= GS_MIA_IN_RESET;
+
+ if (IS_BROXTON(vgpu->gvt->gt->i915)) {
+ vgpu_vreg_t(vgpu, BXT_P_CR_GT_DISP_PWRON) &=
+ ~(BIT(0) | BIT(1));
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY0)) &=
+ ~PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PORT_CL1CM_DW0(DPIO_PHY1)) &=
+ ~PHY_POWER_GOOD;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY0)) &=
+ ~BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL_FAMILY(DPIO_PHY1)) &=
+ ~BIT(30);
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_A)) &=
+ ~BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_A)) |=
+ BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_B)) &=
+ ~BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_B)) |=
+ BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_C)) &=
+ ~BXT_PHY_LANE_ENABLED;
+ vgpu_vreg_t(vgpu, BXT_PHY_CTL(PORT_C)) |=
+ BXT_PHY_CMNLANE_POWERDOWN_ACK |
+ BXT_PHY_LANE_POWERDOWN_ACK;
+ vgpu_vreg_t(vgpu, SKL_FUSE_STATUS) |=
+ SKL_FUSE_DOWNLOAD_STATUS |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG0) |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG1) |
+ SKL_FUSE_PG_DIST_STATUS(SKL_PG2);
+ }
+ } else {
+#define GVT_GEN8_MMIO_RESET_OFFSET (0x44200)
+ /* only reset the engine related, so starting with 0x44200
+ * interrupt include DE,display mmio related will not be
+ * touched
+ */
+ memcpy(vgpu->mmio.vreg, mmio, GVT_GEN8_MMIO_RESET_OFFSET);
+ }
+
+}
+
+/**
+ * intel_vgpu_init_mmio - init MMIO space
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_init_mmio(struct intel_vgpu *vgpu)
+{
+ const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
+
+ vgpu->mmio.vreg = vzalloc(info->mmio_size);
+ if (!vgpu->mmio.vreg)
+ return -ENOMEM;
+
+ intel_vgpu_reset_mmio(vgpu, true);
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_clean_mmio - clean MMIO space
+ * @vgpu: a vGPU
+ *
+ */
+void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu)
+{
+ vfree(vgpu->mmio.vreg);
+ vgpu->mmio.vreg = NULL;
+}
diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h
new file mode 100644
index 000000000..bba154e38
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/mmio.h
@@ -0,0 +1,113 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Ke Yu
+ * Kevin Tian <kevin.tian@intel.com>
+ * Dexuan Cui
+ *
+ * Contributors:
+ * Tina Zhang <tina.zhang@intel.com>
+ * Min He <min.he@intel.com>
+ * Niu Bing <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_MMIO_H_
+#define _GVT_MMIO_H_
+
+#include <linux/types.h>
+
+struct intel_gvt;
+struct intel_vgpu;
+
+#define D_BDW (1 << 0)
+#define D_SKL (1 << 1)
+#define D_KBL (1 << 2)
+#define D_BXT (1 << 3)
+#define D_CFL (1 << 4)
+
+#define D_GEN9PLUS (D_SKL | D_KBL | D_BXT | D_CFL)
+#define D_GEN8PLUS (D_BDW | D_SKL | D_KBL | D_BXT | D_CFL)
+
+#define D_SKL_PLUS (D_SKL | D_KBL | D_BXT | D_CFL)
+#define D_BDW_PLUS (D_BDW | D_SKL | D_KBL | D_BXT | D_CFL)
+
+#define D_PRE_SKL (D_BDW)
+#define D_ALL (D_BDW | D_SKL | D_KBL | D_BXT | D_CFL)
+
+typedef int (*gvt_mmio_func)(struct intel_vgpu *, unsigned int, void *,
+ unsigned int);
+
+struct intel_gvt_mmio_info {
+ u32 offset;
+ u64 ro_mask;
+ u32 device;
+ gvt_mmio_func read;
+ gvt_mmio_func write;
+ u32 addr_range;
+ struct hlist_node node;
+};
+
+const struct intel_engine_cs *
+intel_gvt_render_mmio_to_engine(struct intel_gvt *gvt, unsigned int reg);
+unsigned long intel_gvt_get_device_type(struct intel_gvt *gvt);
+
+int intel_gvt_setup_mmio_info(struct intel_gvt *gvt);
+void intel_gvt_clean_mmio_info(struct intel_gvt *gvt);
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+ int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+ void *data);
+
+struct intel_gvt_mmio_info *intel_gvt_find_mmio_info(struct intel_gvt *gvt,
+ unsigned int offset);
+
+int intel_vgpu_init_mmio(struct intel_vgpu *vgpu);
+void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr);
+void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu);
+
+int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa);
+
+int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, u64 pa,
+ void *p_data, unsigned int bytes);
+int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, u64 pa,
+ void *p_data, unsigned int bytes);
+
+int intel_vgpu_default_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes);
+int intel_vgpu_default_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes);
+
+bool intel_gvt_in_force_nonpriv_whitelist(struct intel_gvt *gvt,
+ unsigned int offset);
+
+int intel_vgpu_mmio_reg_rw(struct intel_vgpu *vgpu, unsigned int offset,
+ void *pdata, unsigned int bytes, bool is_read);
+
+int intel_vgpu_mask_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+ void *p_data, unsigned int bytes);
+
+void intel_gvt_restore_fence(struct intel_gvt *gvt);
+void intel_gvt_restore_mmio(struct intel_gvt *gvt);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
new file mode 100644
index 000000000..1c6e941c9
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -0,0 +1,606 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Eddie Dong <eddie.dong@intel.com>
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gt/intel_context.h"
+#include "gt/intel_engine_regs.h"
+#include "gt/intel_gpu_commands.h"
+#include "gt/intel_gt_regs.h"
+#include "gt/intel_ring.h"
+#include "gvt.h"
+#include "trace.h"
+
+#define GEN9_MOCS_SIZE 64
+
+/* Raw offset is appened to each line for convenience. */
+static struct engine_mmio gen8_engine_mmio_list[] __cacheline_aligned = {
+ {RCS0, RING_MODE_GEN7(RENDER_RING_BASE), 0xffff, false}, /* 0x229c */
+ {RCS0, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+ {RCS0, HWSTAM, 0x0, false}, /* 0x2098 */
+ {RCS0, INSTPM, 0xffff, true}, /* 0x20c0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 0), 0, false}, /* 0x24d0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 1), 0, false}, /* 0x24d4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 2), 0, false}, /* 0x24d8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 3), 0, false}, /* 0x24dc */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 4), 0, false}, /* 0x24e0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 5), 0, false}, /* 0x24e4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 6), 0, false}, /* 0x24e8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 7), 0, false}, /* 0x24ec */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 8), 0, false}, /* 0x24f0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 9), 0, false}, /* 0x24f4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 10), 0, false}, /* 0x24f8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 11), 0, false}, /* 0x24fc */
+ {RCS0, CACHE_MODE_1, 0xffff, true}, /* 0x7004 */
+ {RCS0, GEN7_GT_MODE, 0xffff, true}, /* 0x7008 */
+ {RCS0, CACHE_MODE_0_GEN7, 0xffff, true}, /* 0x7000 */
+ {RCS0, GEN7_COMMON_SLICE_CHICKEN1, 0xffff, true}, /* 0x7010 */
+ {RCS0, HDC_CHICKEN0, 0xffff, true}, /* 0x7300 */
+ {RCS0, VF_GUARDBAND, 0xffff, true}, /* 0x83a4 */
+
+ {BCS0, RING_GFX_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2229c */
+ {BCS0, RING_MI_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2209c */
+ {BCS0, RING_INSTPM(BLT_RING_BASE), 0xffff, false}, /* 0x220c0 */
+ {BCS0, RING_HWSTAM(BLT_RING_BASE), 0x0, false}, /* 0x22098 */
+ {BCS0, RING_EXCC(BLT_RING_BASE), 0xffff, false}, /* 0x22028 */
+ {RCS0, INVALID_MMIO_REG, 0, false } /* Terminated */
+};
+
+static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
+ {RCS0, RING_MODE_GEN7(RENDER_RING_BASE), 0xffff, false}, /* 0x229c */
+ {RCS0, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+ {RCS0, HWSTAM, 0x0, false}, /* 0x2098 */
+ {RCS0, INSTPM, 0xffff, true}, /* 0x20c0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 0), 0, false}, /* 0x24d0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 1), 0, false}, /* 0x24d4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 2), 0, false}, /* 0x24d8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 3), 0, false}, /* 0x24dc */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 4), 0, false}, /* 0x24e0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 5), 0, false}, /* 0x24e4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 6), 0, false}, /* 0x24e8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 7), 0, false}, /* 0x24ec */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 8), 0, false}, /* 0x24f0 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 9), 0, false}, /* 0x24f4 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 10), 0, false}, /* 0x24f8 */
+ {RCS0, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 11), 0, false}, /* 0x24fc */
+ {RCS0, CACHE_MODE_1, 0xffff, true}, /* 0x7004 */
+ {RCS0, GEN7_GT_MODE, 0xffff, true}, /* 0x7008 */
+ {RCS0, CACHE_MODE_0_GEN7, 0xffff, true}, /* 0x7000 */
+ {RCS0, GEN7_COMMON_SLICE_CHICKEN1, 0xffff, true}, /* 0x7010 */
+ {RCS0, HDC_CHICKEN0, 0xffff, true}, /* 0x7300 */
+ {RCS0, VF_GUARDBAND, 0xffff, true}, /* 0x83a4 */
+
+ {RCS0, GEN8_PRIVATE_PAT_LO, 0, false}, /* 0x40e0 */
+ {RCS0, GEN8_PRIVATE_PAT_HI, 0, false}, /* 0x40e4 */
+ {RCS0, GEN8_CS_CHICKEN1, 0xffff, true}, /* 0x2580 */
+ {RCS0, COMMON_SLICE_CHICKEN2, 0xffff, true}, /* 0x7014 */
+ {RCS0, GEN9_CS_DEBUG_MODE1, 0xffff, false}, /* 0x20ec */
+ {RCS0, GEN8_L3SQCREG4, 0, false}, /* 0xb118 */
+ {RCS0, GEN9_SCRATCH1, 0, false}, /* 0xb11c */
+ {RCS0, GEN9_SCRATCH_LNCF1, 0, false}, /* 0xb008 */
+ {RCS0, GEN7_HALF_SLICE_CHICKEN1, 0xffff, true}, /* 0xe100 */
+ {RCS0, HALF_SLICE_CHICKEN2, 0xffff, true}, /* 0xe180 */
+ {RCS0, HALF_SLICE_CHICKEN3, 0xffff, true}, /* 0xe184 */
+ {RCS0, GEN9_HALF_SLICE_CHICKEN5, 0xffff, true}, /* 0xe188 */
+ {RCS0, GEN9_HALF_SLICE_CHICKEN7, 0xffff, true}, /* 0xe194 */
+ {RCS0, GEN8_ROW_CHICKEN, 0xffff, true}, /* 0xe4f0 */
+ {RCS0, TRVATTL3PTRDW(0), 0, true}, /* 0x4de0 */
+ {RCS0, TRVATTL3PTRDW(1), 0, true}, /* 0x4de4 */
+ {RCS0, TRNULLDETCT, 0, true}, /* 0x4de8 */
+ {RCS0, TRINVTILEDETCT, 0, true}, /* 0x4dec */
+ {RCS0, TRVADR, 0, true}, /* 0x4df0 */
+ {RCS0, TRTTE, 0, true}, /* 0x4df4 */
+ {RCS0, _MMIO(0x4dfc), 0, true},
+
+ {BCS0, RING_GFX_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2229c */
+ {BCS0, RING_MI_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2209c */
+ {BCS0, RING_INSTPM(BLT_RING_BASE), 0xffff, false}, /* 0x220c0 */
+ {BCS0, RING_HWSTAM(BLT_RING_BASE), 0x0, false}, /* 0x22098 */
+ {BCS0, RING_EXCC(BLT_RING_BASE), 0xffff, false}, /* 0x22028 */
+
+ {VCS1, RING_EXCC(GEN8_BSD2_RING_BASE), 0xffff, false}, /* 0x1c028 */
+
+ {VECS0, RING_EXCC(VEBOX_RING_BASE), 0xffff, false}, /* 0x1a028 */
+
+ {RCS0, GEN8_HDC_CHICKEN1, 0xffff, true}, /* 0x7304 */
+ {RCS0, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+ {RCS0, GEN7_UCGCTL4, 0x0, false}, /* 0x940c */
+ {RCS0, GAMT_CHKN_BIT_REG, 0x0, false}, /* 0x4ab8 */
+
+ {RCS0, GEN9_GAMT_ECO_REG_RW_IA, 0x0, false}, /* 0x4ab0 */
+ {RCS0, GEN9_CSFE_CHICKEN1_RCS, 0xffff, false}, /* 0x20d4 */
+ {RCS0, _MMIO(0x20D8), 0xffff, true}, /* 0x20d8 */
+
+ {RCS0, GEN8_GARBCNTL, 0x0, false}, /* 0xb004 */
+ {RCS0, GEN7_FF_THREAD_MODE, 0x0, false}, /* 0x20a0 */
+ {RCS0, FF_SLICE_CS_CHICKEN2, 0xffff, false}, /* 0x20e4 */
+ {RCS0, INVALID_MMIO_REG, 0, false } /* Terminated */
+};
+
+static struct {
+ bool initialized;
+ u32 control_table[I915_NUM_ENGINES][GEN9_MOCS_SIZE];
+ u32 l3cc_table[GEN9_MOCS_SIZE / 2];
+} gen9_render_mocs;
+
+static u32 gen9_mocs_mmio_offset_list[] = {
+ [RCS0] = 0xc800,
+ [VCS0] = 0xc900,
+ [VCS1] = 0xca00,
+ [BCS0] = 0xcc00,
+ [VECS0] = 0xcb00,
+};
+
+static void load_render_mocs(const struct intel_engine_cs *engine)
+{
+ struct intel_gvt *gvt = engine->i915->gvt;
+ struct intel_uncore *uncore = engine->uncore;
+ u32 cnt = gvt->engine_mmio_list.mocs_mmio_offset_list_cnt;
+ u32 *regs = gvt->engine_mmio_list.mocs_mmio_offset_list;
+ i915_reg_t offset;
+ int ring_id, i;
+
+ /* Platform doesn't have mocs mmios. */
+ if (!regs)
+ return;
+
+ for (ring_id = 0; ring_id < cnt; ring_id++) {
+ if (!HAS_ENGINE(engine->gt, ring_id))
+ continue;
+
+ offset.reg = regs[ring_id];
+ for (i = 0; i < GEN9_MOCS_SIZE; i++) {
+ gen9_render_mocs.control_table[ring_id][i] =
+ intel_uncore_read_fw(uncore, offset);
+ offset.reg += 4;
+ }
+ }
+
+ offset.reg = 0xb020;
+ for (i = 0; i < GEN9_MOCS_SIZE / 2; i++) {
+ gen9_render_mocs.l3cc_table[i] =
+ intel_uncore_read_fw(uncore, offset);
+ offset.reg += 4;
+ }
+ gen9_render_mocs.initialized = true;
+}
+
+static int
+restore_context_mmio_for_inhibit(struct intel_vgpu *vgpu,
+ struct i915_request *req)
+{
+ u32 *cs;
+ int ret;
+ struct engine_mmio *mmio;
+ struct intel_gvt *gvt = vgpu->gvt;
+ int ring_id = req->engine->id;
+ int count = gvt->engine_mmio_list.ctx_mmio_count[ring_id];
+
+ if (count == 0)
+ return 0;
+
+ ret = req->engine->emit_flush(req, EMIT_BARRIER);
+ if (ret)
+ return ret;
+
+ cs = intel_ring_begin(req, count * 2 + 2);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_LOAD_REGISTER_IMM(count);
+ for (mmio = gvt->engine_mmio_list.mmio;
+ i915_mmio_reg_valid(mmio->reg); mmio++) {
+ if (mmio->id != ring_id || !mmio->in_context)
+ continue;
+
+ *cs++ = i915_mmio_reg_offset(mmio->reg);
+ *cs++ = vgpu_vreg_t(vgpu, mmio->reg) | (mmio->mask << 16);
+ gvt_dbg_core("add lri reg pair 0x%x:0x%x in inhibit ctx, vgpu:%d, rind_id:%d\n",
+ *(cs-2), *(cs-1), vgpu->id, ring_id);
+ }
+
+ *cs++ = MI_NOOP;
+ intel_ring_advance(req, cs);
+
+ ret = req->engine->emit_flush(req, EMIT_BARRIER);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int
+restore_render_mocs_control_for_inhibit(struct intel_vgpu *vgpu,
+ struct i915_request *req)
+{
+ unsigned int index;
+ u32 *cs;
+
+ cs = intel_ring_begin(req, 2 * GEN9_MOCS_SIZE + 2);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_LOAD_REGISTER_IMM(GEN9_MOCS_SIZE);
+
+ for (index = 0; index < GEN9_MOCS_SIZE; index++) {
+ *cs++ = i915_mmio_reg_offset(GEN9_GFX_MOCS(index));
+ *cs++ = vgpu_vreg_t(vgpu, GEN9_GFX_MOCS(index));
+ gvt_dbg_core("add lri reg pair 0x%x:0x%x in inhibit ctx, vgpu:%d, rind_id:%d\n",
+ *(cs-2), *(cs-1), vgpu->id, req->engine->id);
+
+ }
+
+ *cs++ = MI_NOOP;
+ intel_ring_advance(req, cs);
+
+ return 0;
+}
+
+static int
+restore_render_mocs_l3cc_for_inhibit(struct intel_vgpu *vgpu,
+ struct i915_request *req)
+{
+ unsigned int index;
+ u32 *cs;
+
+ cs = intel_ring_begin(req, 2 * GEN9_MOCS_SIZE / 2 + 2);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_LOAD_REGISTER_IMM(GEN9_MOCS_SIZE / 2);
+
+ for (index = 0; index < GEN9_MOCS_SIZE / 2; index++) {
+ *cs++ = i915_mmio_reg_offset(GEN9_LNCFCMOCS(index));
+ *cs++ = vgpu_vreg_t(vgpu, GEN9_LNCFCMOCS(index));
+ gvt_dbg_core("add lri reg pair 0x%x:0x%x in inhibit ctx, vgpu:%d, rind_id:%d\n",
+ *(cs-2), *(cs-1), vgpu->id, req->engine->id);
+
+ }
+
+ *cs++ = MI_NOOP;
+ intel_ring_advance(req, cs);
+
+ return 0;
+}
+
+/*
+ * Use lri command to initialize the mmio which is in context state image for
+ * inhibit context, it contains tracked engine mmio, render_mocs and
+ * render_mocs_l3cc.
+ */
+int intel_vgpu_restore_inhibit_context(struct intel_vgpu *vgpu,
+ struct i915_request *req)
+{
+ int ret;
+ u32 *cs;
+
+ cs = intel_ring_begin(req, 2);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
+ *cs++ = MI_NOOP;
+ intel_ring_advance(req, cs);
+
+ ret = restore_context_mmio_for_inhibit(vgpu, req);
+ if (ret)
+ goto out;
+
+ /* no MOCS register in context except render engine */
+ if (req->engine->id != RCS0)
+ goto out;
+
+ ret = restore_render_mocs_control_for_inhibit(vgpu, req);
+ if (ret)
+ goto out;
+
+ ret = restore_render_mocs_l3cc_for_inhibit(vgpu, req);
+ if (ret)
+ goto out;
+
+out:
+ cs = intel_ring_begin(req, 2);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ *cs++ = MI_NOOP;
+ intel_ring_advance(req, cs);
+
+ return ret;
+}
+
+static u32 gen8_tlb_mmio_offset_list[] = {
+ [RCS0] = 0x4260,
+ [VCS0] = 0x4264,
+ [VCS1] = 0x4268,
+ [BCS0] = 0x426c,
+ [VECS0] = 0x4270,
+};
+
+static void handle_tlb_pending_event(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine)
+{
+ struct intel_uncore *uncore = engine->uncore;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ u32 *regs = vgpu->gvt->engine_mmio_list.tlb_mmio_offset_list;
+ u32 cnt = vgpu->gvt->engine_mmio_list.tlb_mmio_offset_list_cnt;
+ enum forcewake_domains fw;
+ i915_reg_t reg;
+
+ if (!regs)
+ return;
+
+ if (drm_WARN_ON(&engine->i915->drm, engine->id >= cnt))
+ return;
+
+ if (!test_and_clear_bit(engine->id, (void *)s->tlb_handle_pending))
+ return;
+
+ reg = _MMIO(regs[engine->id]);
+
+ /* WaForceWakeRenderDuringMmioTLBInvalidate:skl
+ * we need to put a forcewake when invalidating RCS TLB caches,
+ * otherwise device can go to RC6 state and interrupt invalidation
+ * process
+ */
+ fw = intel_uncore_forcewake_for_reg(uncore, reg,
+ FW_REG_READ | FW_REG_WRITE);
+ if (engine->id == RCS0 && GRAPHICS_VER(engine->i915) >= 9)
+ fw |= FORCEWAKE_RENDER;
+
+ intel_uncore_forcewake_get(uncore, fw);
+
+ intel_uncore_write_fw(uncore, reg, 0x1);
+
+ if (wait_for_atomic(intel_uncore_read_fw(uncore, reg) == 0, 50))
+ gvt_vgpu_err("timeout in invalidate ring %s tlb\n",
+ engine->name);
+ else
+ vgpu_vreg_t(vgpu, reg) = 0;
+
+ intel_uncore_forcewake_put(uncore, fw);
+
+ gvt_dbg_core("invalidate TLB for ring %s\n", engine->name);
+}
+
+static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
+ const struct intel_engine_cs *engine)
+{
+ u32 regs[] = {
+ [RCS0] = 0xc800,
+ [VCS0] = 0xc900,
+ [VCS1] = 0xca00,
+ [BCS0] = 0xcc00,
+ [VECS0] = 0xcb00,
+ };
+ struct intel_uncore *uncore = engine->uncore;
+ i915_reg_t offset, l3_offset;
+ u32 old_v, new_v;
+ int i;
+
+ if (drm_WARN_ON(&engine->i915->drm, engine->id >= ARRAY_SIZE(regs)))
+ return;
+
+ if (engine->id == RCS0 && GRAPHICS_VER(engine->i915) == 9)
+ return;
+
+ if (!pre && !gen9_render_mocs.initialized)
+ load_render_mocs(engine);
+
+ offset.reg = regs[engine->id];
+ for (i = 0; i < GEN9_MOCS_SIZE; i++) {
+ if (pre)
+ old_v = vgpu_vreg_t(pre, offset);
+ else
+ old_v = gen9_render_mocs.control_table[engine->id][i];
+ if (next)
+ new_v = vgpu_vreg_t(next, offset);
+ else
+ new_v = gen9_render_mocs.control_table[engine->id][i];
+
+ if (old_v != new_v)
+ intel_uncore_write_fw(uncore, offset, new_v);
+
+ offset.reg += 4;
+ }
+
+ if (engine->id == RCS0) {
+ l3_offset.reg = 0xb020;
+ for (i = 0; i < GEN9_MOCS_SIZE / 2; i++) {
+ if (pre)
+ old_v = vgpu_vreg_t(pre, l3_offset);
+ else
+ old_v = gen9_render_mocs.l3cc_table[i];
+ if (next)
+ new_v = vgpu_vreg_t(next, l3_offset);
+ else
+ new_v = gen9_render_mocs.l3cc_table[i];
+
+ if (old_v != new_v)
+ intel_uncore_write_fw(uncore, l3_offset, new_v);
+
+ l3_offset.reg += 4;
+ }
+ }
+}
+
+#define CTX_CONTEXT_CONTROL_VAL 0x03
+
+bool is_inhibit_context(struct intel_context *ce)
+{
+ const u32 *reg_state = ce->lrc_reg_state;
+ u32 inhibit_mask =
+ _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
+
+ return inhibit_mask ==
+ (reg_state[CTX_CONTEXT_CONTROL_VAL] & inhibit_mask);
+}
+
+/* Switch ring mmio values (context). */
+static void switch_mmio(struct intel_vgpu *pre,
+ struct intel_vgpu *next,
+ const struct intel_engine_cs *engine)
+{
+ struct intel_uncore *uncore = engine->uncore;
+ struct intel_vgpu_submission *s;
+ struct engine_mmio *mmio;
+ u32 old_v, new_v;
+
+ if (GRAPHICS_VER(engine->i915) >= 9)
+ switch_mocs(pre, next, engine);
+
+ for (mmio = engine->i915->gvt->engine_mmio_list.mmio;
+ i915_mmio_reg_valid(mmio->reg); mmio++) {
+ if (mmio->id != engine->id)
+ continue;
+ /*
+ * No need to do save or restore of the mmio which is in context
+ * state image on gen9, it's initialized by lri command and
+ * save or restore with context together.
+ */
+ if (GRAPHICS_VER(engine->i915) == 9 && mmio->in_context)
+ continue;
+
+ // save
+ if (pre) {
+ vgpu_vreg_t(pre, mmio->reg) =
+ intel_uncore_read_fw(uncore, mmio->reg);
+ if (mmio->mask)
+ vgpu_vreg_t(pre, mmio->reg) &=
+ ~(mmio->mask << 16);
+ old_v = vgpu_vreg_t(pre, mmio->reg);
+ } else {
+ old_v = mmio->value =
+ intel_uncore_read_fw(uncore, mmio->reg);
+ }
+
+ // restore
+ if (next) {
+ s = &next->submission;
+ /*
+ * No need to restore the mmio which is in context state
+ * image if it's not inhibit context, it will restore
+ * itself.
+ */
+ if (mmio->in_context &&
+ !is_inhibit_context(s->shadow[engine->id]))
+ continue;
+
+ if (mmio->mask)
+ new_v = vgpu_vreg_t(next, mmio->reg) |
+ (mmio->mask << 16);
+ else
+ new_v = vgpu_vreg_t(next, mmio->reg);
+ } else {
+ if (mmio->in_context)
+ continue;
+ if (mmio->mask)
+ new_v = mmio->value | (mmio->mask << 16);
+ else
+ new_v = mmio->value;
+ }
+
+ intel_uncore_write_fw(uncore, mmio->reg, new_v);
+
+ trace_render_mmio(pre ? pre->id : 0,
+ next ? next->id : 0,
+ "switch",
+ i915_mmio_reg_offset(mmio->reg),
+ old_v, new_v);
+ }
+
+ if (next)
+ handle_tlb_pending_event(next, engine);
+}
+
+/**
+ * intel_gvt_switch_mmio - switch mmio context of specific engine
+ * @pre: the last vGPU that own the engine
+ * @next: the vGPU to switch to
+ * @engine: the engine
+ *
+ * If pre is null indicates that host own the engine. If next is null
+ * indicates that we are switching to host workload.
+ */
+void intel_gvt_switch_mmio(struct intel_vgpu *pre,
+ struct intel_vgpu *next,
+ const struct intel_engine_cs *engine)
+{
+ if (WARN(!pre && !next, "switch ring %s from host to HOST\n",
+ engine->name))
+ return;
+
+ gvt_dbg_render("switch ring %s from %s to %s\n", engine->name,
+ pre ? "vGPU" : "host", next ? "vGPU" : "HOST");
+
+ /**
+ * We are using raw mmio access wrapper to improve the
+ * performace for batch mmio read/write, so we need
+ * handle forcewake mannually.
+ */
+ intel_uncore_forcewake_get(engine->uncore, FORCEWAKE_ALL);
+ switch_mmio(pre, next, engine);
+ intel_uncore_forcewake_put(engine->uncore, FORCEWAKE_ALL);
+}
+
+/**
+ * intel_gvt_init_engine_mmio_context - Initiate the engine mmio list
+ * @gvt: GVT device
+ *
+ */
+void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt)
+{
+ struct engine_mmio *mmio;
+
+ if (GRAPHICS_VER(gvt->gt->i915) >= 9) {
+ gvt->engine_mmio_list.mmio = gen9_engine_mmio_list;
+ gvt->engine_mmio_list.tlb_mmio_offset_list = gen8_tlb_mmio_offset_list;
+ gvt->engine_mmio_list.tlb_mmio_offset_list_cnt = ARRAY_SIZE(gen8_tlb_mmio_offset_list);
+ gvt->engine_mmio_list.mocs_mmio_offset_list = gen9_mocs_mmio_offset_list;
+ gvt->engine_mmio_list.mocs_mmio_offset_list_cnt = ARRAY_SIZE(gen9_mocs_mmio_offset_list);
+ } else {
+ gvt->engine_mmio_list.mmio = gen8_engine_mmio_list;
+ gvt->engine_mmio_list.tlb_mmio_offset_list = gen8_tlb_mmio_offset_list;
+ gvt->engine_mmio_list.tlb_mmio_offset_list_cnt = ARRAY_SIZE(gen8_tlb_mmio_offset_list);
+ }
+
+ for (mmio = gvt->engine_mmio_list.mmio;
+ i915_mmio_reg_valid(mmio->reg); mmio++) {
+ if (mmio->in_context) {
+ gvt->engine_mmio_list.ctx_mmio_count[mmio->id]++;
+ intel_gvt_mmio_set_sr_in_ctx(gvt, mmio->reg.reg);
+ }
+ }
+}
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.h b/drivers/gpu/drm/i915/gvt/mmio_context.h
new file mode 100644
index 000000000..9540813b8
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.h
@@ -0,0 +1,73 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Eddie Dong <eddie.dong@intel.com>
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Changbin Du <changbin.du@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#ifndef __GVT_RENDER_H__
+#define __GVT_RENDER_H__
+
+#include <linux/types.h>
+
+#include "gt/intel_engine_regs.h"
+#include "gt/intel_engine_types.h"
+#include "gt/intel_lrc_reg.h"
+
+struct i915_request;
+struct intel_context;
+struct intel_engine_cs;
+struct intel_gvt;
+struct intel_vgpu;
+
+struct engine_mmio {
+ enum intel_engine_id id;
+ i915_reg_t reg;
+ u32 mask;
+ bool in_context;
+ u32 value;
+};
+
+void intel_gvt_switch_mmio(struct intel_vgpu *pre,
+ struct intel_vgpu *next,
+ const struct intel_engine_cs *engine);
+
+void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt);
+
+bool is_inhibit_context(struct intel_context *ce);
+
+int intel_vgpu_restore_inhibit_context(struct intel_vgpu *vgpu,
+ struct i915_request *req);
+
+#define IS_RESTORE_INHIBIT(a) \
+ IS_MASKED_BITS_ENABLED(a, CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT)
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c
new file mode 100644
index 000000000..d2bed4665
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/opregion.c
@@ -0,0 +1,483 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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 <linux/acpi.h>
+#include "i915_drv.h"
+#include "gvt.h"
+
+/*
+ * Note: Only for GVT-g virtual VBT generation, other usage must
+ * not do like this.
+ */
+#define _INTEL_BIOS_PRIVATE
+#include "display/intel_vbt_defs.h"
+
+#define OPREGION_SIGNATURE "IntelGraphicsMem"
+#define MBOX_VBT (1<<3)
+
+/* device handle */
+#define DEVICE_TYPE_CRT 0x01
+#define DEVICE_TYPE_EFP1 0x04
+#define DEVICE_TYPE_EFP2 0x40
+#define DEVICE_TYPE_EFP3 0x20
+#define DEVICE_TYPE_EFP4 0x10
+
+struct opregion_header {
+ u8 signature[16];
+ u32 size;
+ u32 opregion_ver;
+ u8 bios_ver[32];
+ u8 vbios_ver[16];
+ u8 driver_ver[16];
+ u32 mboxes;
+ u32 driver_model;
+ u32 pcon;
+ u8 dver[32];
+ u8 rsvd[124];
+} __packed;
+
+struct bdb_data_header {
+ u8 id;
+ u16 size; /* data size */
+} __packed;
+
+/* For supporting windows guest with opregion, here hardcode the emulated
+ * bdb header version as '186', and the corresponding child_device_config
+ * length should be '33' but not '38'.
+ */
+struct efp_child_device_config {
+ u16 handle;
+ u16 device_type;
+ u16 device_class;
+ u8 i2c_speed;
+ u8 dp_onboard_redriver; /* 158 */
+ u8 dp_ondock_redriver; /* 158 */
+ u8 hdmi_level_shifter_value:4; /* 169 */
+ u8 hdmi_max_data_rate:4; /* 204 */
+ u16 dtd_buf_ptr; /* 161 */
+ u8 edidless_efp:1; /* 161 */
+ u8 compression_enable:1; /* 198 */
+ u8 compression_method:1; /* 198 */
+ u8 ganged_edp:1; /* 202 */
+ u8 skip0:4;
+ u8 compression_structure_index:4; /* 198 */
+ u8 skip1:4;
+ u8 slave_port; /* 202 */
+ u8 skip2;
+ u8 dvo_port;
+ u8 i2c_pin; /* for add-in card */
+ u8 slave_addr; /* for add-in card */
+ u8 ddc_pin;
+ u16 edid_ptr;
+ u8 dvo_config;
+ u8 efp_docked_port:1; /* 158 */
+ u8 lane_reversal:1; /* 184 */
+ u8 onboard_lspcon:1; /* 192 */
+ u8 iboost_enable:1; /* 196 */
+ u8 hpd_invert:1; /* BXT 196 */
+ u8 slip3:3;
+ u8 hdmi_compat:1;
+ u8 dp_compat:1;
+ u8 tmds_compat:1;
+ u8 skip4:5;
+ u8 aux_channel;
+ u8 dongle_detect;
+ u8 pipe_cap:2;
+ u8 sdvo_stall:1; /* 158 */
+ u8 hpd_status:2;
+ u8 integrated_encoder:1;
+ u8 skip5:2;
+ u8 dvo_wiring;
+ u8 mipi_bridge_type; /* 171 */
+ u16 device_class_ext;
+ u8 dvo_function;
+} __packed;
+
+struct vbt {
+ /* header->bdb_offset point to bdb_header offset */
+ struct vbt_header header;
+ struct bdb_header bdb_header;
+
+ struct bdb_data_header general_features_header;
+ struct bdb_general_features general_features;
+
+ struct bdb_data_header general_definitions_header;
+ struct bdb_general_definitions general_definitions;
+
+ struct efp_child_device_config child0;
+ struct efp_child_device_config child1;
+ struct efp_child_device_config child2;
+ struct efp_child_device_config child3;
+
+ struct bdb_data_header driver_features_header;
+ struct bdb_driver_features driver_features;
+};
+
+static void virt_vbt_generation(struct vbt *v)
+{
+ int num_child;
+
+ memset(v, 0, sizeof(struct vbt));
+
+ v->header.signature[0] = '$';
+ v->header.signature[1] = 'V';
+ v->header.signature[2] = 'B';
+ v->header.signature[3] = 'T';
+
+ /* there's features depending on version! */
+ v->header.version = 155;
+ v->header.header_size = sizeof(v->header);
+ v->header.vbt_size = sizeof(struct vbt);
+ v->header.bdb_offset = offsetof(struct vbt, bdb_header);
+
+ strcpy(&v->bdb_header.signature[0], "BIOS_DATA_BLOCK");
+ v->bdb_header.version = 186; /* child_dev_size = 33 */
+ v->bdb_header.header_size = sizeof(v->bdb_header);
+
+ v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header);
+
+ /* general features */
+ v->general_features_header.id = BDB_GENERAL_FEATURES;
+ v->general_features_header.size = sizeof(struct bdb_general_features);
+ v->general_features.int_crt_support = 0;
+ v->general_features.int_tv_support = 0;
+
+ /* child device */
+ num_child = 4; /* each port has one child */
+ v->general_definitions.child_dev_size =
+ sizeof(struct efp_child_device_config);
+ v->general_definitions_header.id = BDB_GENERAL_DEFINITIONS;
+ /* size will include child devices */
+ v->general_definitions_header.size =
+ sizeof(struct bdb_general_definitions) +
+ num_child * v->general_definitions.child_dev_size;
+
+ /* portA */
+ v->child0.handle = DEVICE_TYPE_EFP1;
+ v->child0.device_type = DEVICE_TYPE_DP;
+ v->child0.dvo_port = DVO_PORT_DPA;
+ v->child0.aux_channel = DP_AUX_A;
+ v->child0.dp_compat = true;
+ v->child0.integrated_encoder = true;
+
+ /* portB */
+ v->child1.handle = DEVICE_TYPE_EFP2;
+ v->child1.device_type = DEVICE_TYPE_DP;
+ v->child1.dvo_port = DVO_PORT_DPB;
+ v->child1.aux_channel = DP_AUX_B;
+ v->child1.dp_compat = true;
+ v->child1.integrated_encoder = true;
+
+ /* portC */
+ v->child2.handle = DEVICE_TYPE_EFP3;
+ v->child2.device_type = DEVICE_TYPE_DP;
+ v->child2.dvo_port = DVO_PORT_DPC;
+ v->child2.aux_channel = DP_AUX_C;
+ v->child2.dp_compat = true;
+ v->child2.integrated_encoder = true;
+
+ /* portD */
+ v->child3.handle = DEVICE_TYPE_EFP4;
+ v->child3.device_type = DEVICE_TYPE_DP;
+ v->child3.dvo_port = DVO_PORT_DPD;
+ v->child3.aux_channel = DP_AUX_D;
+ v->child3.dp_compat = true;
+ v->child3.integrated_encoder = true;
+
+ /* driver features */
+ v->driver_features_header.id = BDB_DRIVER_FEATURES;
+ v->driver_features_header.size = sizeof(struct bdb_driver_features);
+ v->driver_features.lvds_config = BDB_DRIVER_FEATURE_NO_LVDS;
+}
+
+/**
+ * intel_vgpu_init_opregion - initialize the stuff used to emulate opregion
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_init_opregion(struct intel_vgpu *vgpu)
+{
+ u8 *buf;
+ struct opregion_header *header;
+ struct vbt v;
+ const char opregion_signature[16] = OPREGION_SIGNATURE;
+
+ gvt_dbg_core("init vgpu%d opregion\n", vgpu->id);
+ vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL |
+ __GFP_ZERO,
+ get_order(INTEL_GVT_OPREGION_SIZE));
+ if (!vgpu_opregion(vgpu)->va) {
+ gvt_err("fail to get memory for vgpu virt opregion\n");
+ return -ENOMEM;
+ }
+
+ /* emulated opregion with VBT mailbox only */
+ buf = (u8 *)vgpu_opregion(vgpu)->va;
+ header = (struct opregion_header *)buf;
+ memcpy(header->signature, opregion_signature,
+ sizeof(opregion_signature));
+ header->size = 0x8;
+ header->opregion_ver = 0x02000000;
+ header->mboxes = MBOX_VBT;
+
+ /* for unknown reason, the value in LID field is incorrect
+ * which block the windows guest, so workaround it by force
+ * setting it to "OPEN"
+ */
+ buf[INTEL_GVT_OPREGION_CLID] = 0x3;
+
+ /* emulated vbt from virt vbt generation */
+ virt_vbt_generation(&v);
+ memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt));
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_opregion_base_write_handler - Opregion base register write handler
+ *
+ * @vgpu: a vGPU
+ * @gpa: guest physical address of opregion
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_opregion_base_write_handler(struct intel_vgpu *vgpu, u32 gpa)
+{
+
+ int i;
+
+ gvt_dbg_core("emulate opregion from kernel\n");
+
+ for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
+ vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
+ return 0;
+}
+
+/**
+ * intel_vgpu_clean_opregion - clean the stuff used to emulate opregion
+ * @vgpu: a vGPU
+ *
+ */
+void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu)
+{
+ gvt_dbg_core("vgpu%d: clean vgpu opregion\n", vgpu->id);
+
+ if (!vgpu_opregion(vgpu)->va)
+ return;
+
+ /* Guest opregion is released by VFIO */
+ free_pages((unsigned long)vgpu_opregion(vgpu)->va,
+ get_order(INTEL_GVT_OPREGION_SIZE));
+
+ vgpu_opregion(vgpu)->va = NULL;
+
+}
+
+
+#define GVT_OPREGION_FUNC(scic) \
+ ({ \
+ u32 __ret; \
+ __ret = (scic & OPREGION_SCIC_FUNC_MASK) >> \
+ OPREGION_SCIC_FUNC_SHIFT; \
+ __ret; \
+ })
+
+#define GVT_OPREGION_SUBFUNC(scic) \
+ ({ \
+ u32 __ret; \
+ __ret = (scic & OPREGION_SCIC_SUBFUNC_MASK) >> \
+ OPREGION_SCIC_SUBFUNC_SHIFT; \
+ __ret; \
+ })
+
+static const char *opregion_func_name(u32 func)
+{
+ const char *name = NULL;
+
+ switch (func) {
+ case 0 ... 3:
+ case 5:
+ case 7 ... 15:
+ name = "Reserved";
+ break;
+
+ case 4:
+ name = "Get BIOS Data";
+ break;
+
+ case 6:
+ name = "System BIOS Callbacks";
+ break;
+
+ default:
+ name = "Unknown";
+ break;
+ }
+ return name;
+}
+
+static const char *opregion_subfunc_name(u32 subfunc)
+{
+ const char *name = NULL;
+
+ switch (subfunc) {
+ case 0:
+ name = "Supported Calls";
+ break;
+
+ case 1:
+ name = "Requested Callbacks";
+ break;
+
+ case 2 ... 3:
+ case 8 ... 9:
+ name = "Reserved";
+ break;
+
+ case 5:
+ name = "Boot Display";
+ break;
+
+ case 6:
+ name = "TV-Standard/Video-Connector";
+ break;
+
+ case 7:
+ name = "Internal Graphics";
+ break;
+
+ case 10:
+ name = "Spread Spectrum Clocks";
+ break;
+
+ case 11:
+ name = "Get AKSV";
+ break;
+
+ default:
+ name = "Unknown";
+ break;
+ }
+ return name;
+};
+
+static bool querying_capabilities(u32 scic)
+{
+ u32 func, subfunc;
+
+ func = GVT_OPREGION_FUNC(scic);
+ subfunc = GVT_OPREGION_SUBFUNC(scic);
+
+ if ((func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA &&
+ subfunc == INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS)
+ || (func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA &&
+ subfunc == INTEL_GVT_OPREGION_SCIC_SF_REQEUSTEDCALLBACKS)
+ || (func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSCALLBACKS &&
+ subfunc == INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS)) {
+ return true;
+ }
+ return false;
+}
+
+/**
+ * intel_vgpu_emulate_opregion_request - emulating OpRegion request
+ * @vgpu: a vGPU
+ * @swsci: SWSCI request
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
+{
+ u32 scic, parm;
+ u32 func, subfunc;
+ u64 scic_pa = 0, parm_pa = 0;
+ int ret;
+
+ scic_pa = (vgpu_opregion(vgpu)->gfn[0] << PAGE_SHIFT) +
+ INTEL_GVT_OPREGION_SCIC;
+ parm_pa = (vgpu_opregion(vgpu)->gfn[0] << PAGE_SHIFT) +
+ INTEL_GVT_OPREGION_PARM;
+ ret = intel_gvt_read_gpa(vgpu, scic_pa, &scic, sizeof(scic));
+ if (ret) {
+ gvt_vgpu_err("guest opregion read error %d, gpa 0x%llx, len %lu\n",
+ ret, scic_pa, sizeof(scic));
+ return ret;
+ }
+
+ ret = intel_gvt_read_gpa(vgpu, parm_pa, &parm, sizeof(parm));
+ if (ret) {
+ gvt_vgpu_err("guest opregion read error %d, gpa 0x%llx, len %lu\n",
+ ret, scic_pa, sizeof(scic));
+ return ret;
+ }
+
+ if (!(swsci & SWSCI_SCI_SELECT)) {
+ gvt_vgpu_err("requesting SMI service\n");
+ return 0;
+ }
+ /* ignore non 0->1 trasitions */
+ if ((vgpu_cfg_space(vgpu)[INTEL_GVT_PCI_SWSCI]
+ & SWSCI_SCI_TRIGGER) ||
+ !(swsci & SWSCI_SCI_TRIGGER)) {
+ return 0;
+ }
+
+ func = GVT_OPREGION_FUNC(scic);
+ subfunc = GVT_OPREGION_SUBFUNC(scic);
+ if (!querying_capabilities(scic)) {
+ gvt_vgpu_err("requesting runtime service: func \"%s\","
+ " subfunc \"%s\"\n",
+ opregion_func_name(func),
+ opregion_subfunc_name(subfunc));
+ /*
+ * emulate exit status of function call, '0' means
+ * "failure, generic, unsupported or unknown cause"
+ */
+ scic &= ~OPREGION_SCIC_EXIT_MASK;
+ goto out;
+ }
+
+ scic = 0;
+ parm = 0;
+
+out:
+ ret = intel_gvt_write_gpa(vgpu, scic_pa, &scic, sizeof(scic));
+ if (ret) {
+ gvt_vgpu_err("guest opregion write error %d, gpa 0x%llx, len %lu\n",
+ ret, scic_pa, sizeof(scic));
+ return ret;
+ }
+
+ ret = intel_gvt_write_gpa(vgpu, parm_pa, &parm, sizeof(parm));
+ if (ret) {
+ gvt_vgpu_err("guest opregion write error %d, gpa 0x%llx, len %lu\n",
+ ret, scic_pa, sizeof(scic));
+ return ret;
+ }
+
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/page_track.c b/drivers/gpu/drm/i915/gvt/page_track.c
new file mode 100644
index 000000000..3375b51c7
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/page_track.c
@@ -0,0 +1,185 @@
+/*
+ * Copyright(c) 2011-2017 Intel 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 (including the next
+ * paragraph) 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 "i915_drv.h"
+#include "gvt.h"
+
+/**
+ * intel_vgpu_find_page_track - find page track rcord of guest page
+ * @vgpu: a vGPU
+ * @gfn: the gfn of guest page
+ *
+ * Returns:
+ * A pointer to struct intel_vgpu_page_track if found, else NULL returned.
+ */
+struct intel_vgpu_page_track *intel_vgpu_find_page_track(
+ struct intel_vgpu *vgpu, unsigned long gfn)
+{
+ return radix_tree_lookup(&vgpu->page_track_tree, gfn);
+}
+
+/**
+ * intel_vgpu_register_page_track - register a guest page to be tacked
+ * @vgpu: a vGPU
+ * @gfn: the gfn of guest page
+ * @handler: page track handler
+ * @priv: tracker private
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ */
+int intel_vgpu_register_page_track(struct intel_vgpu *vgpu, unsigned long gfn,
+ gvt_page_track_handler_t handler, void *priv)
+{
+ struct intel_vgpu_page_track *track;
+ int ret;
+
+ track = intel_vgpu_find_page_track(vgpu, gfn);
+ if (track)
+ return -EEXIST;
+
+ track = kzalloc(sizeof(*track), GFP_KERNEL);
+ if (!track)
+ return -ENOMEM;
+
+ track->handler = handler;
+ track->priv_data = priv;
+
+ ret = radix_tree_insert(&vgpu->page_track_tree, gfn, track);
+ if (ret) {
+ kfree(track);
+ return ret;
+ }
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_unregister_page_track - unregister the tracked guest page
+ * @vgpu: a vGPU
+ * @gfn: the gfn of guest page
+ *
+ */
+void intel_vgpu_unregister_page_track(struct intel_vgpu *vgpu,
+ unsigned long gfn)
+{
+ struct intel_vgpu_page_track *track;
+
+ track = radix_tree_delete(&vgpu->page_track_tree, gfn);
+ if (track) {
+ if (track->tracked)
+ intel_gvt_page_track_remove(vgpu, gfn);
+ kfree(track);
+ }
+}
+
+/**
+ * intel_vgpu_enable_page_track - set write-protection on guest page
+ * @vgpu: a vGPU
+ * @gfn: the gfn of guest page
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ */
+int intel_vgpu_enable_page_track(struct intel_vgpu *vgpu, unsigned long gfn)
+{
+ struct intel_vgpu_page_track *track;
+ int ret;
+
+ track = intel_vgpu_find_page_track(vgpu, gfn);
+ if (!track)
+ return -ENXIO;
+
+ if (track->tracked)
+ return 0;
+
+ ret = intel_gvt_page_track_add(vgpu, gfn);
+ if (ret)
+ return ret;
+ track->tracked = true;
+ return 0;
+}
+
+/**
+ * intel_vgpu_enable_page_track - cancel write-protection on guest page
+ * @vgpu: a vGPU
+ * @gfn: the gfn of guest page
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ */
+int intel_vgpu_disable_page_track(struct intel_vgpu *vgpu, unsigned long gfn)
+{
+ struct intel_vgpu_page_track *track;
+ int ret;
+
+ track = intel_vgpu_find_page_track(vgpu, gfn);
+ if (!track)
+ return -ENXIO;
+
+ if (!track->tracked)
+ return 0;
+
+ ret = intel_gvt_page_track_remove(vgpu, gfn);
+ if (ret)
+ return ret;
+ track->tracked = false;
+ return 0;
+}
+
+/**
+ * intel_vgpu_page_track_handler - called when write to write-protected page
+ * @vgpu: a vGPU
+ * @gpa: the gpa of this write
+ * @data: the writed data
+ * @bytes: the length of this write
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ */
+int intel_vgpu_page_track_handler(struct intel_vgpu *vgpu, u64 gpa,
+ void *data, unsigned int bytes)
+{
+ struct intel_vgpu_page_track *page_track;
+ int ret = 0;
+
+ mutex_lock(&vgpu->vgpu_lock);
+
+ page_track = intel_vgpu_find_page_track(vgpu, gpa >> PAGE_SHIFT);
+ if (!page_track) {
+ ret = -ENXIO;
+ goto out;
+ }
+
+ if (unlikely(vgpu->failsafe)) {
+ /* Remove write protection to prevent furture traps. */
+ intel_gvt_page_track_remove(vgpu, gpa >> PAGE_SHIFT);
+ } else {
+ ret = page_track->handler(page_track, gpa, data, bytes);
+ if (ret)
+ gvt_err("guest page write error, gpa %llx\n", gpa);
+ }
+
+out:
+ mutex_unlock(&vgpu->vgpu_lock);
+ return ret;
+}
diff --git a/drivers/gpu/drm/i915/gvt/page_track.h b/drivers/gpu/drm/i915/gvt/page_track.h
new file mode 100644
index 000000000..f6eb71355
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/page_track.h
@@ -0,0 +1,59 @@
+/*
+ * Copyright(c) 2011-2017 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ */
+
+#ifndef _GVT_PAGE_TRACK_H_
+#define _GVT_PAGE_TRACK_H_
+
+#include <linux/types.h>
+
+struct intel_vgpu;
+struct intel_vgpu_page_track;
+
+typedef int (*gvt_page_track_handler_t)(
+ struct intel_vgpu_page_track *page_track,
+ u64 gpa, void *data, int bytes);
+
+/* Track record for a write-protected guest page. */
+struct intel_vgpu_page_track {
+ gvt_page_track_handler_t handler;
+ bool tracked;
+ void *priv_data;
+};
+
+struct intel_vgpu_page_track *intel_vgpu_find_page_track(
+ struct intel_vgpu *vgpu, unsigned long gfn);
+
+int intel_vgpu_register_page_track(struct intel_vgpu *vgpu,
+ unsigned long gfn, gvt_page_track_handler_t handler,
+ void *priv);
+void intel_vgpu_unregister_page_track(struct intel_vgpu *vgpu,
+ unsigned long gfn);
+
+int intel_vgpu_enable_page_track(struct intel_vgpu *vgpu, unsigned long gfn);
+int intel_vgpu_disable_page_track(struct intel_vgpu *vgpu, unsigned long gfn);
+
+int intel_vgpu_page_track_handler(struct intel_vgpu *vgpu, u64 gpa,
+ void *data, unsigned int bytes);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/reg.h b/drivers/gpu/drm/i915/gvt/reg.h
new file mode 100644
index 000000000..d8216c63c
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/reg.h
@@ -0,0 +1,144 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ */
+
+#ifndef _GVT_REG_H
+#define _GVT_REG_H
+
+#define INTEL_GVT_PCI_CLASS_VGA_OTHER 0x80
+
+#define INTEL_GVT_PCI_GMCH_CONTROL 0x50
+#define BDW_GMCH_GMS_SHIFT 8
+#define BDW_GMCH_GMS_MASK 0xff
+
+#define INTEL_GVT_PCI_SWSCI 0xe8
+#define SWSCI_SCI_SELECT (1 << 15)
+#define SWSCI_SCI_TRIGGER 1
+
+#define INTEL_GVT_PCI_OPREGION 0xfc
+
+#define INTEL_GVT_OPREGION_CLID 0x1AC
+#define INTEL_GVT_OPREGION_SCIC 0x200
+#define OPREGION_SCIC_FUNC_MASK 0x1E
+#define OPREGION_SCIC_FUNC_SHIFT 1
+#define OPREGION_SCIC_SUBFUNC_MASK 0xFF00
+#define OPREGION_SCIC_SUBFUNC_SHIFT 8
+#define OPREGION_SCIC_EXIT_MASK 0xE0
+#define INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA 4
+#define INTEL_GVT_OPREGION_SCIC_F_GETBIOSCALLBACKS 6
+#define INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS 0
+#define INTEL_GVT_OPREGION_SCIC_SF_REQEUSTEDCALLBACKS 1
+#define INTEL_GVT_OPREGION_PARM 0x204
+
+#define INTEL_GVT_OPREGION_PAGES 2
+#define INTEL_GVT_OPREGION_SIZE (INTEL_GVT_OPREGION_PAGES * PAGE_SIZE)
+#define INTEL_GVT_OPREGION_VBT_OFFSET 0x400
+#define INTEL_GVT_OPREGION_VBT_SIZE \
+ (INTEL_GVT_OPREGION_SIZE - INTEL_GVT_OPREGION_VBT_OFFSET)
+
+#define VGT_SPRSTRIDE(pipe) _PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B)
+
+#define _REG_701C0(pipe, plane) (0x701c0 + pipe * 0x1000 + (plane - 1) * 0x100)
+#define _REG_701C4(pipe, plane) (0x701c4 + pipe * 0x1000 + (plane - 1) * 0x100)
+
+#define SKL_FLIP_EVENT(pipe, plane) (PRIMARY_A_FLIP_DONE + (plane) * 3 + (pipe))
+
+#define REG50080_FLIP_TYPE_MASK 0x3
+#define REG50080_FLIP_TYPE_ASYNC 0x1
+
+#define REG_50080(_pipe, _plane) ({ \
+ typeof(_pipe) (p) = (_pipe); \
+ typeof(_plane) (q) = (_plane); \
+ (((p) == PIPE_A) ? (((q) == PLANE_PRIMARY) ? (_MMIO(0x50080)) : \
+ (_MMIO(0x50090))) : \
+ (((p) == PIPE_B) ? (((q) == PLANE_PRIMARY) ? (_MMIO(0x50088)) : \
+ (_MMIO(0x50098))) : \
+ (((p) == PIPE_C) ? (((q) == PLANE_PRIMARY) ? (_MMIO(0x5008C)) : \
+ (_MMIO(0x5009C))) : \
+ (_MMIO(0x50080))))); })
+
+#define REG_50080_TO_PIPE(_reg) ({ \
+ typeof(_reg) (reg) = (_reg); \
+ (((reg) == 0x50080 || (reg) == 0x50090) ? (PIPE_A) : \
+ (((reg) == 0x50088 || (reg) == 0x50098) ? (PIPE_B) : \
+ (((reg) == 0x5008C || (reg) == 0x5009C) ? (PIPE_C) : \
+ (INVALID_PIPE)))); })
+
+#define REG_50080_TO_PLANE(_reg) ({ \
+ typeof(_reg) (reg) = (_reg); \
+ (((reg) == 0x50080 || (reg) == 0x50088 || (reg) == 0x5008C) ? \
+ (PLANE_PRIMARY) : \
+ (((reg) == 0x50090 || (reg) == 0x50098 || (reg) == 0x5009C) ? \
+ (PLANE_SPRITE0) : (I915_MAX_PLANES))); })
+
+#define GFX_MODE_BIT_SET_IN_MASK(val, bit) \
+ ((((bit) & 0xffff0000) == 0) && !!((val) & (((bit) << 16))))
+
+#define IS_MASKED_BITS_ENABLED(_val, _b) \
+ (((_val) & _MASKED_BIT_ENABLE(_b)) == _MASKED_BIT_ENABLE(_b))
+#define IS_MASKED_BITS_DISABLED(_val, _b) \
+ ((_val) & _MASKED_BIT_DISABLE(_b))
+
+#define FORCEWAKE_RENDER_GEN9_REG 0xa278
+#define FORCEWAKE_ACK_RENDER_GEN9_REG 0x0D84
+#define FORCEWAKE_GT_GEN9_REG 0xa188
+#define FORCEWAKE_ACK_GT_GEN9_REG 0x130044
+#define FORCEWAKE_MEDIA_GEN9_REG 0xa270
+#define FORCEWAKE_ACK_MEDIA_GEN9_REG 0x0D88
+#define FORCEWAKE_ACK_HSW_REG 0x130044
+
+#define RB_HEAD_WRAP_CNT_MAX ((1 << 11) - 1)
+#define RB_HEAD_WRAP_CNT_OFF 21
+#define RB_HEAD_OFF_MASK ((1U << 21) - (1U << 2))
+#define RB_TAIL_OFF_MASK ((1U << 21) - (1U << 3))
+#define RB_TAIL_SIZE_MASK ((1U << 21) - (1U << 12))
+#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + \
+ I915_GTT_PAGE_SIZE)
+
+#define PCH_GPIO_BASE _MMIO(0xc5010)
+
+#define PCH_GMBUS0 _MMIO(0xc5100)
+#define PCH_GMBUS1 _MMIO(0xc5104)
+#define PCH_GMBUS2 _MMIO(0xc5108)
+#define PCH_GMBUS3 _MMIO(0xc510c)
+#define PCH_GMBUS4 _MMIO(0xc5110)
+#define PCH_GMBUS5 _MMIO(0xc5120)
+
+#define TRVATTL3PTRDW(i) _MMIO(0x4de0 + (i) * 4)
+#define TRNULLDETCT _MMIO(0x4de8)
+#define TRINVTILEDETCT _MMIO(0x4dec)
+#define TRVADR _MMIO(0x4df0)
+#define TRTTE _MMIO(0x4df4)
+#define RING_EXCC(base) _MMIO((base) + 0x28)
+#define RING_GFX_MODE(base) _MMIO((base) + 0x29c)
+#define VF_GUARDBAND _MMIO(0x83a4)
+
+#define BCS_TILE_REGISTER_VAL_OFFSET (0x43*4)
+
+/* XXX FIXME i915 has changed PP_XXX definition */
+#define PCH_PP_STATUS _MMIO(0xc7200)
+#define PCH_PP_CONTROL _MMIO(0xc7204)
+#define PCH_PP_ON_DELAYS _MMIO(0xc7208)
+#define PCH_PP_OFF_DELAYS _MMIO(0xc720c)
+#define PCH_PP_DIVISOR _MMIO(0xc7210)
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/sched_policy.c b/drivers/gpu/drm/i915/gvt/sched_policy.c
new file mode 100644
index 000000000..c077fb467
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/sched_policy.c
@@ -0,0 +1,480 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Anhua Xu
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+
+static bool vgpu_has_pending_workload(struct intel_vgpu *vgpu)
+{
+ enum intel_engine_id i;
+ struct intel_engine_cs *engine;
+
+ for_each_engine(engine, vgpu->gvt->gt, i) {
+ if (!list_empty(workload_q_head(vgpu, engine)))
+ return true;
+ }
+
+ return false;
+}
+
+/* We give 2 seconds higher prio for vGPU during start */
+#define GVT_SCHED_VGPU_PRI_TIME 2
+
+struct vgpu_sched_data {
+ struct list_head lru_list;
+ struct intel_vgpu *vgpu;
+ bool active;
+ bool pri_sched;
+ ktime_t pri_time;
+ ktime_t sched_in_time;
+ ktime_t sched_time;
+ ktime_t left_ts;
+ ktime_t allocated_ts;
+
+ struct vgpu_sched_ctl sched_ctl;
+};
+
+struct gvt_sched_data {
+ struct intel_gvt *gvt;
+ struct hrtimer timer;
+ unsigned long period;
+ struct list_head lru_runq_head;
+ ktime_t expire_time;
+};
+
+static void vgpu_update_timeslice(struct intel_vgpu *vgpu, ktime_t cur_time)
+{
+ ktime_t delta_ts;
+ struct vgpu_sched_data *vgpu_data;
+
+ if (!vgpu || vgpu == vgpu->gvt->idle_vgpu)
+ return;
+
+ vgpu_data = vgpu->sched_data;
+ delta_ts = ktime_sub(cur_time, vgpu_data->sched_in_time);
+ vgpu_data->sched_time = ktime_add(vgpu_data->sched_time, delta_ts);
+ vgpu_data->left_ts = ktime_sub(vgpu_data->left_ts, delta_ts);
+ vgpu_data->sched_in_time = cur_time;
+}
+
+#define GVT_TS_BALANCE_PERIOD_MS 100
+#define GVT_TS_BALANCE_STAGE_NUM 10
+
+static void gvt_balance_timeslice(struct gvt_sched_data *sched_data)
+{
+ struct vgpu_sched_data *vgpu_data;
+ struct list_head *pos;
+ static u64 stage_check;
+ int stage = stage_check++ % GVT_TS_BALANCE_STAGE_NUM;
+
+ /* The timeslice accumulation reset at stage 0, which is
+ * allocated again without adding previous debt.
+ */
+ if (stage == 0) {
+ int total_weight = 0;
+ ktime_t fair_timeslice;
+
+ list_for_each(pos, &sched_data->lru_runq_head) {
+ vgpu_data = container_of(pos, struct vgpu_sched_data, lru_list);
+ total_weight += vgpu_data->sched_ctl.weight;
+ }
+
+ list_for_each(pos, &sched_data->lru_runq_head) {
+ vgpu_data = container_of(pos, struct vgpu_sched_data, lru_list);
+ fair_timeslice = ktime_divns(ms_to_ktime(GVT_TS_BALANCE_PERIOD_MS),
+ total_weight) * vgpu_data->sched_ctl.weight;
+
+ vgpu_data->allocated_ts = fair_timeslice;
+ vgpu_data->left_ts = vgpu_data->allocated_ts;
+ }
+ } else {
+ list_for_each(pos, &sched_data->lru_runq_head) {
+ vgpu_data = container_of(pos, struct vgpu_sched_data, lru_list);
+
+ /* timeslice for next 100ms should add the left/debt
+ * slice of previous stages.
+ */
+ vgpu_data->left_ts += vgpu_data->allocated_ts;
+ }
+ }
+}
+
+static void try_to_schedule_next_vgpu(struct intel_gvt *gvt)
+{
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ enum intel_engine_id i;
+ struct intel_engine_cs *engine;
+ struct vgpu_sched_data *vgpu_data;
+ ktime_t cur_time;
+
+ /* no need to schedule if next_vgpu is the same with current_vgpu,
+ * let scheduler chose next_vgpu again by setting it to NULL.
+ */
+ if (scheduler->next_vgpu == scheduler->current_vgpu) {
+ scheduler->next_vgpu = NULL;
+ return;
+ }
+
+ /*
+ * after the flag is set, workload dispatch thread will
+ * stop dispatching workload for current vgpu
+ */
+ scheduler->need_reschedule = true;
+
+ /* still have uncompleted workload? */
+ for_each_engine(engine, gvt->gt, i) {
+ if (scheduler->current_workload[engine->id])
+ return;
+ }
+
+ cur_time = ktime_get();
+ vgpu_update_timeslice(scheduler->current_vgpu, cur_time);
+ vgpu_data = scheduler->next_vgpu->sched_data;
+ vgpu_data->sched_in_time = cur_time;
+
+ /* switch current vgpu */
+ scheduler->current_vgpu = scheduler->next_vgpu;
+ scheduler->next_vgpu = NULL;
+
+ scheduler->need_reschedule = false;
+
+ /* wake up workload dispatch thread */
+ for_each_engine(engine, gvt->gt, i)
+ wake_up(&scheduler->waitq[engine->id]);
+}
+
+static struct intel_vgpu *find_busy_vgpu(struct gvt_sched_data *sched_data)
+{
+ struct vgpu_sched_data *vgpu_data;
+ struct intel_vgpu *vgpu = NULL;
+ struct list_head *head = &sched_data->lru_runq_head;
+ struct list_head *pos;
+
+ /* search a vgpu with pending workload */
+ list_for_each(pos, head) {
+
+ vgpu_data = container_of(pos, struct vgpu_sched_data, lru_list);
+ if (!vgpu_has_pending_workload(vgpu_data->vgpu))
+ continue;
+
+ if (vgpu_data->pri_sched) {
+ if (ktime_before(ktime_get(), vgpu_data->pri_time)) {
+ vgpu = vgpu_data->vgpu;
+ break;
+ } else
+ vgpu_data->pri_sched = false;
+ }
+
+ /* Return the vGPU only if it has time slice left */
+ if (vgpu_data->left_ts > 0) {
+ vgpu = vgpu_data->vgpu;
+ break;
+ }
+ }
+
+ return vgpu;
+}
+
+/* in nanosecond */
+#define GVT_DEFAULT_TIME_SLICE 1000000
+
+static void tbs_sched_func(struct gvt_sched_data *sched_data)
+{
+ struct intel_gvt *gvt = sched_data->gvt;
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct vgpu_sched_data *vgpu_data;
+ struct intel_vgpu *vgpu = NULL;
+
+ /* no active vgpu or has already had a target */
+ if (list_empty(&sched_data->lru_runq_head) || scheduler->next_vgpu)
+ goto out;
+
+ vgpu = find_busy_vgpu(sched_data);
+ if (vgpu) {
+ scheduler->next_vgpu = vgpu;
+ vgpu_data = vgpu->sched_data;
+ if (!vgpu_data->pri_sched) {
+ /* Move the last used vGPU to the tail of lru_list */
+ list_del_init(&vgpu_data->lru_list);
+ list_add_tail(&vgpu_data->lru_list,
+ &sched_data->lru_runq_head);
+ }
+ } else {
+ scheduler->next_vgpu = gvt->idle_vgpu;
+ }
+out:
+ if (scheduler->next_vgpu)
+ try_to_schedule_next_vgpu(gvt);
+}
+
+void intel_gvt_schedule(struct intel_gvt *gvt)
+{
+ struct gvt_sched_data *sched_data = gvt->scheduler.sched_data;
+ ktime_t cur_time;
+
+ mutex_lock(&gvt->sched_lock);
+ cur_time = ktime_get();
+
+ if (test_and_clear_bit(INTEL_GVT_REQUEST_SCHED,
+ (void *)&gvt->service_request)) {
+ if (cur_time >= sched_data->expire_time) {
+ gvt_balance_timeslice(sched_data);
+ sched_data->expire_time = ktime_add_ms(
+ cur_time, GVT_TS_BALANCE_PERIOD_MS);
+ }
+ }
+ clear_bit(INTEL_GVT_REQUEST_EVENT_SCHED, (void *)&gvt->service_request);
+
+ vgpu_update_timeslice(gvt->scheduler.current_vgpu, cur_time);
+ tbs_sched_func(sched_data);
+
+ mutex_unlock(&gvt->sched_lock);
+}
+
+static enum hrtimer_restart tbs_timer_fn(struct hrtimer *timer_data)
+{
+ struct gvt_sched_data *data;
+
+ data = container_of(timer_data, struct gvt_sched_data, timer);
+
+ intel_gvt_request_service(data->gvt, INTEL_GVT_REQUEST_SCHED);
+
+ hrtimer_add_expires_ns(&data->timer, data->period);
+
+ return HRTIMER_RESTART;
+}
+
+static int tbs_sched_init(struct intel_gvt *gvt)
+{
+ struct intel_gvt_workload_scheduler *scheduler =
+ &gvt->scheduler;
+
+ struct gvt_sched_data *data;
+
+ data = kzalloc(sizeof(*data), GFP_KERNEL);
+ if (!data)
+ return -ENOMEM;
+
+ INIT_LIST_HEAD(&data->lru_runq_head);
+ hrtimer_init(&data->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
+ data->timer.function = tbs_timer_fn;
+ data->period = GVT_DEFAULT_TIME_SLICE;
+ data->gvt = gvt;
+
+ scheduler->sched_data = data;
+
+ return 0;
+}
+
+static void tbs_sched_clean(struct intel_gvt *gvt)
+{
+ struct intel_gvt_workload_scheduler *scheduler =
+ &gvt->scheduler;
+ struct gvt_sched_data *data = scheduler->sched_data;
+
+ hrtimer_cancel(&data->timer);
+
+ kfree(data);
+ scheduler->sched_data = NULL;
+}
+
+static int tbs_sched_init_vgpu(struct intel_vgpu *vgpu)
+{
+ struct vgpu_sched_data *data;
+
+ data = kzalloc(sizeof(*data), GFP_KERNEL);
+ if (!data)
+ return -ENOMEM;
+
+ data->sched_ctl.weight = vgpu->sched_ctl.weight;
+ data->vgpu = vgpu;
+ INIT_LIST_HEAD(&data->lru_list);
+
+ vgpu->sched_data = data;
+
+ return 0;
+}
+
+static void tbs_sched_clean_vgpu(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct gvt_sched_data *sched_data = gvt->scheduler.sched_data;
+
+ kfree(vgpu->sched_data);
+ vgpu->sched_data = NULL;
+
+ /* this vgpu id has been removed */
+ if (idr_is_empty(&gvt->vgpu_idr))
+ hrtimer_cancel(&sched_data->timer);
+}
+
+static void tbs_sched_start_schedule(struct intel_vgpu *vgpu)
+{
+ struct gvt_sched_data *sched_data = vgpu->gvt->scheduler.sched_data;
+ struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
+ ktime_t now;
+
+ if (!list_empty(&vgpu_data->lru_list))
+ return;
+
+ now = ktime_get();
+ vgpu_data->pri_time = ktime_add(now,
+ ktime_set(GVT_SCHED_VGPU_PRI_TIME, 0));
+ vgpu_data->pri_sched = true;
+
+ list_add(&vgpu_data->lru_list, &sched_data->lru_runq_head);
+
+ if (!hrtimer_active(&sched_data->timer))
+ hrtimer_start(&sched_data->timer, ktime_add_ns(ktime_get(),
+ sched_data->period), HRTIMER_MODE_ABS);
+ vgpu_data->active = true;
+}
+
+static void tbs_sched_stop_schedule(struct intel_vgpu *vgpu)
+{
+ struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
+
+ list_del_init(&vgpu_data->lru_list);
+ vgpu_data->active = false;
+}
+
+static const struct intel_gvt_sched_policy_ops tbs_schedule_ops = {
+ .init = tbs_sched_init,
+ .clean = tbs_sched_clean,
+ .init_vgpu = tbs_sched_init_vgpu,
+ .clean_vgpu = tbs_sched_clean_vgpu,
+ .start_schedule = tbs_sched_start_schedule,
+ .stop_schedule = tbs_sched_stop_schedule,
+};
+
+int intel_gvt_init_sched_policy(struct intel_gvt *gvt)
+{
+ int ret;
+
+ mutex_lock(&gvt->sched_lock);
+ gvt->scheduler.sched_ops = &tbs_schedule_ops;
+ ret = gvt->scheduler.sched_ops->init(gvt);
+ mutex_unlock(&gvt->sched_lock);
+
+ return ret;
+}
+
+void intel_gvt_clean_sched_policy(struct intel_gvt *gvt)
+{
+ mutex_lock(&gvt->sched_lock);
+ gvt->scheduler.sched_ops->clean(gvt);
+ mutex_unlock(&gvt->sched_lock);
+}
+
+/* for per-vgpu scheduler policy, there are 2 per-vgpu data:
+ * sched_data, and sched_ctl. We see these 2 data as part of
+ * the global scheduler which are proteced by gvt->sched_lock.
+ * Caller should make their decision if the vgpu_lock should
+ * be hold outside.
+ */
+
+int intel_vgpu_init_sched_policy(struct intel_vgpu *vgpu)
+{
+ int ret;
+
+ mutex_lock(&vgpu->gvt->sched_lock);
+ ret = vgpu->gvt->scheduler.sched_ops->init_vgpu(vgpu);
+ mutex_unlock(&vgpu->gvt->sched_lock);
+
+ return ret;
+}
+
+void intel_vgpu_clean_sched_policy(struct intel_vgpu *vgpu)
+{
+ mutex_lock(&vgpu->gvt->sched_lock);
+ vgpu->gvt->scheduler.sched_ops->clean_vgpu(vgpu);
+ mutex_unlock(&vgpu->gvt->sched_lock);
+}
+
+void intel_vgpu_start_schedule(struct intel_vgpu *vgpu)
+{
+ struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
+
+ mutex_lock(&vgpu->gvt->sched_lock);
+ if (!vgpu_data->active) {
+ gvt_dbg_core("vgpu%d: start schedule\n", vgpu->id);
+ vgpu->gvt->scheduler.sched_ops->start_schedule(vgpu);
+ }
+ mutex_unlock(&vgpu->gvt->sched_lock);
+}
+
+void intel_gvt_kick_schedule(struct intel_gvt *gvt)
+{
+ mutex_lock(&gvt->sched_lock);
+ intel_gvt_request_service(gvt, INTEL_GVT_REQUEST_EVENT_SCHED);
+ mutex_unlock(&gvt->sched_lock);
+}
+
+void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt_workload_scheduler *scheduler =
+ &vgpu->gvt->scheduler;
+ struct vgpu_sched_data *vgpu_data = vgpu->sched_data;
+ struct drm_i915_private *dev_priv = vgpu->gvt->gt->i915;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+
+ if (!vgpu_data->active)
+ return;
+
+ gvt_dbg_core("vgpu%d: stop schedule\n", vgpu->id);
+
+ mutex_lock(&vgpu->gvt->sched_lock);
+ scheduler->sched_ops->stop_schedule(vgpu);
+
+ if (scheduler->next_vgpu == vgpu)
+ scheduler->next_vgpu = NULL;
+
+ if (scheduler->current_vgpu == vgpu) {
+ /* stop workload dispatching */
+ scheduler->need_reschedule = true;
+ scheduler->current_vgpu = NULL;
+ }
+
+ intel_runtime_pm_get(&dev_priv->runtime_pm);
+ spin_lock_bh(&scheduler->mmio_context_lock);
+ for_each_engine(engine, vgpu->gvt->gt, id) {
+ if (scheduler->engine_owner[engine->id] == vgpu) {
+ intel_gvt_switch_mmio(vgpu, NULL, engine);
+ scheduler->engine_owner[engine->id] = NULL;
+ }
+ }
+ spin_unlock_bh(&scheduler->mmio_context_lock);
+ intel_runtime_pm_put_unchecked(&dev_priv->runtime_pm);
+ mutex_unlock(&vgpu->gvt->sched_lock);
+}
diff --git a/drivers/gpu/drm/i915/gvt/sched_policy.h b/drivers/gpu/drm/i915/gvt/sched_policy.h
new file mode 100644
index 000000000..3dacdad5f
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/sched_policy.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Anhua Xu
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef __GVT_SCHED_POLICY__
+#define __GVT_SCHED_POLICY__
+
+struct intel_gvt;
+struct intel_vgpu;
+
+struct intel_gvt_sched_policy_ops {
+ int (*init)(struct intel_gvt *gvt);
+ void (*clean)(struct intel_gvt *gvt);
+ int (*init_vgpu)(struct intel_vgpu *vgpu);
+ void (*clean_vgpu)(struct intel_vgpu *vgpu);
+ void (*start_schedule)(struct intel_vgpu *vgpu);
+ void (*stop_schedule)(struct intel_vgpu *vgpu);
+};
+
+void intel_gvt_schedule(struct intel_gvt *gvt);
+
+int intel_gvt_init_sched_policy(struct intel_gvt *gvt);
+
+void intel_gvt_clean_sched_policy(struct intel_gvt *gvt);
+
+int intel_vgpu_init_sched_policy(struct intel_vgpu *vgpu);
+
+void intel_vgpu_clean_sched_policy(struct intel_vgpu *vgpu);
+
+void intel_vgpu_start_schedule(struct intel_vgpu *vgpu);
+
+void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu);
+
+void intel_gvt_kick_schedule(struct intel_gvt *gvt);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
new file mode 100644
index 000000000..8342d95f5
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -0,0 +1,1786 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Chanbin Du <changbin.du@intel.com>
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ *
+ */
+
+#include <linux/kthread.h>
+
+#include "gem/i915_gem_pm.h"
+#include "gt/intel_context.h"
+#include "gt/intel_execlists_submission.h"
+#include "gt/intel_gt_regs.h"
+#include "gt/intel_lrc.h"
+#include "gt/intel_ring.h"
+
+#include "i915_drv.h"
+#include "i915_gem_gtt.h"
+#include "i915_perf_oa_regs.h"
+#include "gvt.h"
+
+#define RING_CTX_OFF(x) \
+ offsetof(struct execlist_ring_context, x)
+
+static void set_context_pdp_root_pointer(
+ struct execlist_ring_context *ring_context,
+ u32 pdp[8])
+{
+ int i;
+
+ for (i = 0; i < 8; i++)
+ ring_context->pdps[i].val = pdp[7 - i];
+}
+
+static void update_shadow_pdps(struct intel_vgpu_workload *workload)
+{
+ struct execlist_ring_context *shadow_ring_context;
+ struct intel_context *ctx = workload->req->context;
+
+ if (WARN_ON(!workload->shadow_mm))
+ return;
+
+ if (WARN_ON(!atomic_read(&workload->shadow_mm->pincount)))
+ return;
+
+ shadow_ring_context = (struct execlist_ring_context *)ctx->lrc_reg_state;
+ set_context_pdp_root_pointer(shadow_ring_context,
+ (void *)workload->shadow_mm->ppgtt_mm.shadow_pdps);
+}
+
+/*
+ * when populating shadow ctx from guest, we should not overrride oa related
+ * registers, so that they will not be overlapped by guest oa configs. Thus
+ * made it possible to capture oa data from host for both host and guests.
+ */
+static void sr_oa_regs(struct intel_vgpu_workload *workload,
+ u32 *reg_state, bool save)
+{
+ struct drm_i915_private *dev_priv = workload->vgpu->gvt->gt->i915;
+ u32 ctx_oactxctrl = dev_priv->perf.ctx_oactxctrl_offset;
+ u32 ctx_flexeu0 = dev_priv->perf.ctx_flexeu0_offset;
+ int i = 0;
+ u32 flex_mmio[] = {
+ i915_mmio_reg_offset(EU_PERF_CNTL0),
+ i915_mmio_reg_offset(EU_PERF_CNTL1),
+ i915_mmio_reg_offset(EU_PERF_CNTL2),
+ i915_mmio_reg_offset(EU_PERF_CNTL3),
+ i915_mmio_reg_offset(EU_PERF_CNTL4),
+ i915_mmio_reg_offset(EU_PERF_CNTL5),
+ i915_mmio_reg_offset(EU_PERF_CNTL6),
+ };
+
+ if (workload->engine->id != RCS0)
+ return;
+
+ if (save) {
+ workload->oactxctrl = reg_state[ctx_oactxctrl + 1];
+
+ for (i = 0; i < ARRAY_SIZE(workload->flex_mmio); i++) {
+ u32 state_offset = ctx_flexeu0 + i * 2;
+
+ workload->flex_mmio[i] = reg_state[state_offset + 1];
+ }
+ } else {
+ reg_state[ctx_oactxctrl] =
+ i915_mmio_reg_offset(GEN8_OACTXCONTROL);
+ reg_state[ctx_oactxctrl + 1] = workload->oactxctrl;
+
+ for (i = 0; i < ARRAY_SIZE(workload->flex_mmio); i++) {
+ u32 state_offset = ctx_flexeu0 + i * 2;
+ u32 mmio = flex_mmio[i];
+
+ reg_state[state_offset] = mmio;
+ reg_state[state_offset + 1] = workload->flex_mmio[i];
+ }
+ }
+}
+
+static int populate_shadow_context(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_context *ctx = workload->req->context;
+ struct execlist_ring_context *shadow_ring_context;
+ void *dst;
+ void *context_base;
+ unsigned long context_gpa, context_page_num;
+ unsigned long gpa_base; /* first gpa of consecutive GPAs */
+ unsigned long gpa_size; /* size of consecutive GPAs */
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ int i;
+ bool skip = false;
+ int ring_id = workload->engine->id;
+ int ret;
+
+ GEM_BUG_ON(!intel_context_is_pinned(ctx));
+
+ context_base = (void *) ctx->lrc_reg_state -
+ (LRC_STATE_PN << I915_GTT_PAGE_SHIFT);
+
+ shadow_ring_context = (void *) ctx->lrc_reg_state;
+
+ sr_oa_regs(workload, (u32 *)shadow_ring_context, true);
+#define COPY_REG(name) \
+ intel_gvt_read_gpa(vgpu, workload->ring_context_gpa \
+ + RING_CTX_OFF(name.val), &shadow_ring_context->name.val, 4)
+#define COPY_REG_MASKED(name) {\
+ intel_gvt_read_gpa(vgpu, workload->ring_context_gpa \
+ + RING_CTX_OFF(name.val),\
+ &shadow_ring_context->name.val, 4);\
+ shadow_ring_context->name.val |= 0xffff << 16;\
+ }
+
+ COPY_REG_MASKED(ctx_ctrl);
+ COPY_REG(ctx_timestamp);
+
+ if (workload->engine->id == RCS0) {
+ COPY_REG(bb_per_ctx_ptr);
+ COPY_REG(rcs_indirect_ctx);
+ COPY_REG(rcs_indirect_ctx_offset);
+ } else if (workload->engine->id == BCS0)
+ intel_gvt_read_gpa(vgpu,
+ workload->ring_context_gpa +
+ BCS_TILE_REGISTER_VAL_OFFSET,
+ (void *)shadow_ring_context +
+ BCS_TILE_REGISTER_VAL_OFFSET, 4);
+#undef COPY_REG
+#undef COPY_REG_MASKED
+
+ /* don't copy Ring Context (the first 0x50 dwords),
+ * only copy the Engine Context part from guest
+ */
+ intel_gvt_read_gpa(vgpu,
+ workload->ring_context_gpa +
+ RING_CTX_SIZE,
+ (void *)shadow_ring_context +
+ RING_CTX_SIZE,
+ I915_GTT_PAGE_SIZE - RING_CTX_SIZE);
+
+ sr_oa_regs(workload, (u32 *)shadow_ring_context, false);
+
+ gvt_dbg_sched("ring %s workload lrca %x, ctx_id %x, ctx gpa %llx",
+ workload->engine->name, workload->ctx_desc.lrca,
+ workload->ctx_desc.context_id,
+ workload->ring_context_gpa);
+
+ /* only need to ensure this context is not pinned/unpinned during the
+ * period from last submission to this this submission.
+ * Upon reaching this function, the currently submitted context is not
+ * supposed to get unpinned. If a misbehaving guest driver ever does
+ * this, it would corrupt itself.
+ */
+ if (s->last_ctx[ring_id].valid &&
+ (s->last_ctx[ring_id].lrca ==
+ workload->ctx_desc.lrca) &&
+ (s->last_ctx[ring_id].ring_context_gpa ==
+ workload->ring_context_gpa))
+ skip = true;
+
+ s->last_ctx[ring_id].lrca = workload->ctx_desc.lrca;
+ s->last_ctx[ring_id].ring_context_gpa = workload->ring_context_gpa;
+
+ if (IS_RESTORE_INHIBIT(shadow_ring_context->ctx_ctrl.val) || skip)
+ return 0;
+
+ s->last_ctx[ring_id].valid = false;
+ context_page_num = workload->engine->context_size;
+ context_page_num = context_page_num >> PAGE_SHIFT;
+
+ if (IS_BROADWELL(gvt->gt->i915) && workload->engine->id == RCS0)
+ context_page_num = 19;
+
+ /* find consecutive GPAs from gma until the first inconsecutive GPA.
+ * read from the continuous GPAs into dst virtual address
+ */
+ gpa_size = 0;
+ for (i = 2; i < context_page_num; i++) {
+ context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+ (u32)((workload->ctx_desc.lrca + i) <<
+ I915_GTT_PAGE_SHIFT));
+ if (context_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("Invalid guest context descriptor\n");
+ return -EFAULT;
+ }
+
+ if (gpa_size == 0) {
+ gpa_base = context_gpa;
+ dst = context_base + (i << I915_GTT_PAGE_SHIFT);
+ } else if (context_gpa != gpa_base + gpa_size)
+ goto read;
+
+ gpa_size += I915_GTT_PAGE_SIZE;
+
+ if (i == context_page_num - 1)
+ goto read;
+
+ continue;
+
+read:
+ intel_gvt_read_gpa(vgpu, gpa_base, dst, gpa_size);
+ gpa_base = context_gpa;
+ gpa_size = I915_GTT_PAGE_SIZE;
+ dst = context_base + (i << I915_GTT_PAGE_SHIFT);
+ }
+ ret = intel_gvt_scan_engine_context(workload);
+ if (ret) {
+ gvt_vgpu_err("invalid cmd found in guest context pages\n");
+ return ret;
+ }
+ s->last_ctx[ring_id].valid = true;
+ return 0;
+}
+
+static inline bool is_gvt_request(struct i915_request *rq)
+{
+ return intel_context_force_single_submission(rq->context);
+}
+
+static void save_ring_hw_state(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine)
+{
+ struct intel_uncore *uncore = engine->uncore;
+ i915_reg_t reg;
+
+ reg = RING_INSTDONE(engine->mmio_base);
+ vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) =
+ intel_uncore_read(uncore, reg);
+
+ reg = RING_ACTHD(engine->mmio_base);
+ vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) =
+ intel_uncore_read(uncore, reg);
+
+ reg = RING_ACTHD_UDW(engine->mmio_base);
+ vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) =
+ intel_uncore_read(uncore, reg);
+}
+
+static int shadow_context_status_change(struct notifier_block *nb,
+ unsigned long action, void *data)
+{
+ struct i915_request *rq = data;
+ struct intel_gvt *gvt = container_of(nb, struct intel_gvt,
+ shadow_ctx_notifier_block[rq->engine->id]);
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ enum intel_engine_id ring_id = rq->engine->id;
+ struct intel_vgpu_workload *workload;
+ unsigned long flags;
+
+ if (!is_gvt_request(rq)) {
+ spin_lock_irqsave(&scheduler->mmio_context_lock, flags);
+ if (action == INTEL_CONTEXT_SCHEDULE_IN &&
+ scheduler->engine_owner[ring_id]) {
+ /* Switch ring from vGPU to host. */
+ intel_gvt_switch_mmio(scheduler->engine_owner[ring_id],
+ NULL, rq->engine);
+ scheduler->engine_owner[ring_id] = NULL;
+ }
+ spin_unlock_irqrestore(&scheduler->mmio_context_lock, flags);
+
+ return NOTIFY_OK;
+ }
+
+ workload = scheduler->current_workload[ring_id];
+ if (unlikely(!workload))
+ return NOTIFY_OK;
+
+ switch (action) {
+ case INTEL_CONTEXT_SCHEDULE_IN:
+ spin_lock_irqsave(&scheduler->mmio_context_lock, flags);
+ if (workload->vgpu != scheduler->engine_owner[ring_id]) {
+ /* Switch ring from host to vGPU or vGPU to vGPU. */
+ intel_gvt_switch_mmio(scheduler->engine_owner[ring_id],
+ workload->vgpu, rq->engine);
+ scheduler->engine_owner[ring_id] = workload->vgpu;
+ } else
+ gvt_dbg_sched("skip ring %d mmio switch for vgpu%d\n",
+ ring_id, workload->vgpu->id);
+ spin_unlock_irqrestore(&scheduler->mmio_context_lock, flags);
+ atomic_set(&workload->shadow_ctx_active, 1);
+ break;
+ case INTEL_CONTEXT_SCHEDULE_OUT:
+ save_ring_hw_state(workload->vgpu, rq->engine);
+ atomic_set(&workload->shadow_ctx_active, 0);
+ break;
+ case INTEL_CONTEXT_SCHEDULE_PREEMPTED:
+ save_ring_hw_state(workload->vgpu, rq->engine);
+ break;
+ default:
+ WARN_ON(1);
+ return NOTIFY_OK;
+ }
+ wake_up(&workload->shadow_ctx_status_wq);
+ return NOTIFY_OK;
+}
+
+static void
+shadow_context_descriptor_update(struct intel_context *ce,
+ struct intel_vgpu_workload *workload)
+{
+ u64 desc = ce->lrc.desc;
+
+ /*
+ * Update bits 0-11 of the context descriptor which includes flags
+ * like GEN8_CTX_* cached in desc_template
+ */
+ desc &= ~(0x3ull << GEN8_CTX_ADDRESSING_MODE_SHIFT);
+ desc |= (u64)workload->ctx_desc.addressing_mode <<
+ GEN8_CTX_ADDRESSING_MODE_SHIFT;
+
+ ce->lrc.desc = desc;
+}
+
+static int copy_workload_to_ring_buffer(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct i915_request *req = workload->req;
+ void *shadow_ring_buffer_va;
+ u32 *cs;
+ int err;
+
+ if (GRAPHICS_VER(req->engine->i915) == 9 && is_inhibit_context(req->context))
+ intel_vgpu_restore_inhibit_context(vgpu, req);
+
+ /*
+ * To track whether a request has started on HW, we can emit a
+ * breadcrumb at the beginning of the request and check its
+ * timeline's HWSP to see if the breadcrumb has advanced past the
+ * start of this request. Actually, the request must have the
+ * init_breadcrumb if its timeline set has_init_bread_crumb, or the
+ * scheduler might get a wrong state of it during reset. Since the
+ * requests from gvt always set the has_init_breadcrumb flag, here
+ * need to do the emit_init_breadcrumb for all the requests.
+ */
+ if (req->engine->emit_init_breadcrumb) {
+ err = req->engine->emit_init_breadcrumb(req);
+ if (err) {
+ gvt_vgpu_err("fail to emit init breadcrumb\n");
+ return err;
+ }
+ }
+
+ /* allocate shadow ring buffer */
+ cs = intel_ring_begin(workload->req, workload->rb_len / sizeof(u32));
+ if (IS_ERR(cs)) {
+ gvt_vgpu_err("fail to alloc size =%ld shadow ring buffer\n",
+ workload->rb_len);
+ return PTR_ERR(cs);
+ }
+
+ shadow_ring_buffer_va = workload->shadow_ring_buffer_va;
+
+ /* get shadow ring buffer va */
+ workload->shadow_ring_buffer_va = cs;
+
+ memcpy(cs, shadow_ring_buffer_va,
+ workload->rb_len);
+
+ cs += workload->rb_len / sizeof(u32);
+ intel_ring_advance(workload->req, cs);
+
+ return 0;
+}
+
+static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ if (!wa_ctx->indirect_ctx.obj)
+ return;
+
+ i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
+ i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+ i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+ i915_gem_object_put(wa_ctx->indirect_ctx.obj);
+
+ wa_ctx->indirect_ctx.obj = NULL;
+ wa_ctx->indirect_ctx.shadow_va = NULL;
+}
+
+static void set_dma_address(struct i915_page_directory *pd, dma_addr_t addr)
+{
+ struct scatterlist *sg = pd->pt.base->mm.pages->sgl;
+
+ /* This is not a good idea */
+ sg->dma_address = addr;
+}
+
+static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
+ struct intel_context *ce)
+{
+ struct intel_vgpu_mm *mm = workload->shadow_mm;
+ struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ce->vm);
+ int i = 0;
+
+ if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
+ set_dma_address(ppgtt->pd, mm->ppgtt_mm.shadow_pdps[0]);
+ } else {
+ for (i = 0; i < GVT_RING_CTX_NR_PDPS; i++) {
+ struct i915_page_directory * const pd =
+ i915_pd_entry(ppgtt->pd, i);
+ /* skip now as current i915 ppgtt alloc won't allocate
+ top level pdp for non 4-level table, won't impact
+ shadow ppgtt. */
+ if (!pd)
+ break;
+
+ set_dma_address(pd, mm->ppgtt_mm.shadow_pdps[i]);
+ }
+ }
+}
+
+static int
+intel_gvt_workload_req_alloc(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_request *rq;
+
+ if (workload->req)
+ return 0;
+
+ rq = i915_request_create(s->shadow[workload->engine->id]);
+ if (IS_ERR(rq)) {
+ gvt_vgpu_err("fail to allocate gem request\n");
+ return PTR_ERR(rq);
+ }
+
+ workload->req = i915_request_get(rq);
+ return 0;
+}
+
+/**
+ * intel_gvt_scan_and_shadow_workload - audit the workload by scanning and
+ * shadow it as well, include ringbuffer,wa_ctx and ctx.
+ * @workload: an abstract entity for each execlist submission.
+ *
+ * This function is called before the workload submitting to i915, to make
+ * sure the content of the workload is valid.
+ */
+int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ int ret;
+
+ lockdep_assert_held(&vgpu->vgpu_lock);
+
+ if (workload->shadow)
+ return 0;
+
+ if (!test_and_set_bit(workload->engine->id, s->shadow_ctx_desc_updated))
+ shadow_context_descriptor_update(s->shadow[workload->engine->id],
+ workload);
+
+ ret = intel_gvt_scan_and_shadow_ringbuffer(workload);
+ if (ret)
+ return ret;
+
+ if (workload->engine->id == RCS0 &&
+ workload->wa_ctx.indirect_ctx.size) {
+ ret = intel_gvt_scan_and_shadow_wa_ctx(&workload->wa_ctx);
+ if (ret)
+ goto err_shadow;
+ }
+
+ workload->shadow = true;
+ return 0;
+
+err_shadow:
+ release_shadow_wa_ctx(&workload->wa_ctx);
+ return ret;
+}
+
+static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload);
+
+static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+ struct intel_gvt *gvt = workload->vgpu->gvt;
+ const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
+ struct intel_vgpu_shadow_bb *bb;
+ struct i915_gem_ww_ctx ww;
+ int ret;
+
+ list_for_each_entry(bb, &workload->shadow_bb, list) {
+ /* For privilge batch buffer and not wa_ctx, the bb_start_cmd_va
+ * is only updated into ring_scan_buffer, not real ring address
+ * allocated in later copy_workload_to_ring_buffer. pls be noted
+ * shadow_ring_buffer_va is now pointed to real ring buffer va
+ * in copy_workload_to_ring_buffer.
+ */
+
+ if (bb->bb_offset)
+ bb->bb_start_cmd_va = workload->shadow_ring_buffer_va
+ + bb->bb_offset;
+
+ /*
+ * For non-priv bb, scan&shadow is only for
+ * debugging purpose, so the content of shadow bb
+ * is the same as original bb. Therefore,
+ * here, rather than switch to shadow bb's gma
+ * address, we directly use original batch buffer's
+ * gma address, and send original bb to hardware
+ * directly
+ */
+ if (!bb->ppgtt) {
+ i915_gem_ww_ctx_init(&ww, false);
+retry:
+ i915_gem_object_lock(bb->obj, &ww);
+
+ bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+ NULL, 0, 0, 0);
+ if (IS_ERR(bb->vma)) {
+ ret = PTR_ERR(bb->vma);
+ if (ret == -EDEADLK) {
+ ret = i915_gem_ww_ctx_backoff(&ww);
+ if (!ret)
+ goto retry;
+ }
+ goto err;
+ }
+
+ /* relocate shadow batch buffer */
+ bb->bb_start_cmd_va[1] = i915_ggtt_offset(bb->vma);
+ if (gmadr_bytes == 8)
+ bb->bb_start_cmd_va[2] = 0;
+
+ ret = i915_vma_move_to_active(bb->vma,
+ workload->req,
+ 0);
+ if (ret)
+ goto err;
+
+ /* No one is going to touch shadow bb from now on. */
+ i915_gem_object_flush_map(bb->obj);
+ i915_gem_ww_ctx_fini(&ww);
+ }
+ }
+ return 0;
+err:
+ i915_gem_ww_ctx_fini(&ww);
+ release_shadow_batch_buffer(workload);
+ return ret;
+}
+
+static void update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ struct intel_vgpu_workload *workload =
+ container_of(wa_ctx, struct intel_vgpu_workload, wa_ctx);
+ struct i915_request *rq = workload->req;
+ struct execlist_ring_context *shadow_ring_context =
+ (struct execlist_ring_context *)rq->context->lrc_reg_state;
+
+ shadow_ring_context->bb_per_ctx_ptr.val =
+ (shadow_ring_context->bb_per_ctx_ptr.val &
+ (~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma;
+ shadow_ring_context->rcs_indirect_ctx.val =
+ (shadow_ring_context->rcs_indirect_ctx.val &
+ (~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma;
+}
+
+static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+ struct i915_vma *vma;
+ unsigned char *per_ctx_va =
+ (unsigned char *)wa_ctx->indirect_ctx.shadow_va +
+ wa_ctx->indirect_ctx.size;
+ struct i915_gem_ww_ctx ww;
+ int ret;
+
+ if (wa_ctx->indirect_ctx.size == 0)
+ return 0;
+
+ i915_gem_ww_ctx_init(&ww, false);
+retry:
+ i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+ vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+ 0, CACHELINE_BYTES, 0);
+ if (IS_ERR(vma)) {
+ ret = PTR_ERR(vma);
+ if (ret == -EDEADLK) {
+ ret = i915_gem_ww_ctx_backoff(&ww);
+ if (!ret)
+ goto retry;
+ }
+ return ret;
+ }
+
+ i915_gem_ww_ctx_fini(&ww);
+
+ /* FIXME: we are not tracking our pinned VMA leaving it
+ * up to the core to fix up the stray pin_count upon
+ * free.
+ */
+
+ wa_ctx->indirect_ctx.shadow_gma = i915_ggtt_offset(vma);
+
+ wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1);
+ memset(per_ctx_va, 0, CACHELINE_BYTES);
+
+ update_wa_ctx_2_shadow_ctx(wa_ctx);
+ return 0;
+}
+
+static void update_vreg_in_ctx(struct intel_vgpu_workload *workload)
+{
+ vgpu_vreg_t(workload->vgpu, RING_START(workload->engine->mmio_base)) =
+ workload->rb_start;
+}
+
+static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu_shadow_bb *bb, *pos;
+
+ if (list_empty(&workload->shadow_bb))
+ return;
+
+ bb = list_first_entry(&workload->shadow_bb,
+ struct intel_vgpu_shadow_bb, list);
+
+ list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
+ if (bb->obj) {
+ i915_gem_object_lock(bb->obj, NULL);
+ if (bb->va && !IS_ERR(bb->va))
+ i915_gem_object_unpin_map(bb->obj);
+
+ if (bb->vma && !IS_ERR(bb->vma))
+ i915_vma_unpin(bb->vma);
+
+ i915_gem_object_unlock(bb->obj);
+ i915_gem_object_put(bb->obj);
+ }
+ list_del(&bb->list);
+ kfree(bb);
+ }
+}
+
+static int
+intel_vgpu_shadow_mm_pin(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_mm *m;
+ int ret = 0;
+
+ ret = intel_vgpu_pin_mm(workload->shadow_mm);
+ if (ret) {
+ gvt_vgpu_err("fail to vgpu pin mm\n");
+ return ret;
+ }
+
+ if (workload->shadow_mm->type != INTEL_GVT_MM_PPGTT ||
+ !workload->shadow_mm->ppgtt_mm.shadowed) {
+ intel_vgpu_unpin_mm(workload->shadow_mm);
+ gvt_vgpu_err("workload shadow ppgtt isn't ready\n");
+ return -EINVAL;
+ }
+
+ if (!list_empty(&workload->lri_shadow_mm)) {
+ list_for_each_entry(m, &workload->lri_shadow_mm,
+ ppgtt_mm.link) {
+ ret = intel_vgpu_pin_mm(m);
+ if (ret) {
+ list_for_each_entry_from_reverse(m,
+ &workload->lri_shadow_mm,
+ ppgtt_mm.link)
+ intel_vgpu_unpin_mm(m);
+ gvt_vgpu_err("LRI shadow ppgtt fail to pin\n");
+ break;
+ }
+ }
+ }
+
+ if (ret)
+ intel_vgpu_unpin_mm(workload->shadow_mm);
+
+ return ret;
+}
+
+static void
+intel_vgpu_shadow_mm_unpin(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu_mm *m;
+
+ if (!list_empty(&workload->lri_shadow_mm)) {
+ list_for_each_entry(m, &workload->lri_shadow_mm,
+ ppgtt_mm.link)
+ intel_vgpu_unpin_mm(m);
+ }
+ intel_vgpu_unpin_mm(workload->shadow_mm);
+}
+
+static int prepare_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ int ret = 0;
+
+ ret = intel_vgpu_shadow_mm_pin(workload);
+ if (ret) {
+ gvt_vgpu_err("fail to pin shadow mm\n");
+ return ret;
+ }
+
+ update_shadow_pdps(workload);
+
+ set_context_ppgtt_from_shadow(workload, s->shadow[workload->engine->id]);
+
+ ret = intel_vgpu_sync_oos_pages(workload->vgpu);
+ if (ret) {
+ gvt_vgpu_err("fail to vgpu sync oos pages\n");
+ goto err_unpin_mm;
+ }
+
+ ret = intel_vgpu_flush_post_shadow(workload->vgpu);
+ if (ret) {
+ gvt_vgpu_err("fail to flush post shadow\n");
+ goto err_unpin_mm;
+ }
+
+ ret = copy_workload_to_ring_buffer(workload);
+ if (ret) {
+ gvt_vgpu_err("fail to generate request\n");
+ goto err_unpin_mm;
+ }
+
+ ret = prepare_shadow_batch_buffer(workload);
+ if (ret) {
+ gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n");
+ goto err_unpin_mm;
+ }
+
+ ret = prepare_shadow_wa_ctx(&workload->wa_ctx);
+ if (ret) {
+ gvt_vgpu_err("fail to prepare_shadow_wa_ctx\n");
+ goto err_shadow_batch;
+ }
+
+ if (workload->prepare) {
+ ret = workload->prepare(workload);
+ if (ret)
+ goto err_shadow_wa_ctx;
+ }
+
+ return 0;
+err_shadow_wa_ctx:
+ release_shadow_wa_ctx(&workload->wa_ctx);
+err_shadow_batch:
+ release_shadow_batch_buffer(workload);
+err_unpin_mm:
+ intel_vgpu_shadow_mm_unpin(workload);
+ return ret;
+}
+
+static int dispatch_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct i915_request *rq;
+ int ret;
+
+ gvt_dbg_sched("ring id %s prepare to dispatch workload %p\n",
+ workload->engine->name, workload);
+
+ mutex_lock(&vgpu->vgpu_lock);
+
+ ret = intel_gvt_workload_req_alloc(workload);
+ if (ret)
+ goto err_req;
+
+ ret = intel_gvt_scan_and_shadow_workload(workload);
+ if (ret)
+ goto out;
+
+ ret = populate_shadow_context(workload);
+ if (ret) {
+ release_shadow_wa_ctx(&workload->wa_ctx);
+ goto out;
+ }
+
+ ret = prepare_workload(workload);
+out:
+ if (ret) {
+ /* We might still need to add request with
+ * clean ctx to retire it properly..
+ */
+ rq = fetch_and_zero(&workload->req);
+ i915_request_put(rq);
+ }
+
+ if (!IS_ERR_OR_NULL(workload->req)) {
+ gvt_dbg_sched("ring id %s submit workload to i915 %p\n",
+ workload->engine->name, workload->req);
+ i915_request_add(workload->req);
+ workload->dispatched = true;
+ }
+err_req:
+ if (ret)
+ workload->status = ret;
+ mutex_unlock(&vgpu->vgpu_lock);
+ return ret;
+}
+
+static struct intel_vgpu_workload *
+pick_next_workload(struct intel_gvt *gvt, struct intel_engine_cs *engine)
+{
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct intel_vgpu_workload *workload = NULL;
+
+ mutex_lock(&gvt->sched_lock);
+
+ /*
+ * no current vgpu / will be scheduled out / no workload
+ * bail out
+ */
+ if (!scheduler->current_vgpu) {
+ gvt_dbg_sched("ring %s stop - no current vgpu\n", engine->name);
+ goto out;
+ }
+
+ if (scheduler->need_reschedule) {
+ gvt_dbg_sched("ring %s stop - will reschedule\n", engine->name);
+ goto out;
+ }
+
+ if (!scheduler->current_vgpu->active ||
+ list_empty(workload_q_head(scheduler->current_vgpu, engine)))
+ goto out;
+
+ /*
+ * still have current workload, maybe the workload disptacher
+ * fail to submit it for some reason, resubmit it.
+ */
+ if (scheduler->current_workload[engine->id]) {
+ workload = scheduler->current_workload[engine->id];
+ gvt_dbg_sched("ring %s still have current workload %p\n",
+ engine->name, workload);
+ goto out;
+ }
+
+ /*
+ * pick a workload as current workload
+ * once current workload is set, schedule policy routines
+ * will wait the current workload is finished when trying to
+ * schedule out a vgpu.
+ */
+ scheduler->current_workload[engine->id] =
+ list_first_entry(workload_q_head(scheduler->current_vgpu,
+ engine),
+ struct intel_vgpu_workload, list);
+
+ workload = scheduler->current_workload[engine->id];
+
+ gvt_dbg_sched("ring %s pick new workload %p\n", engine->name, workload);
+
+ atomic_inc(&workload->vgpu->submission.running_workload_num);
+out:
+ mutex_unlock(&gvt->sched_lock);
+ return workload;
+}
+
+static void update_guest_pdps(struct intel_vgpu *vgpu,
+ u64 ring_context_gpa, u32 pdp[8])
+{
+ u64 gpa;
+ int i;
+
+ gpa = ring_context_gpa + RING_CTX_OFF(pdps[0].val);
+
+ for (i = 0; i < 8; i++)
+ intel_gvt_write_gpa(vgpu, gpa + i * 8, &pdp[7 - i], 4);
+}
+
+static __maybe_unused bool
+check_shadow_context_ppgtt(struct execlist_ring_context *c, struct intel_vgpu_mm *m)
+{
+ if (m->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
+ u64 shadow_pdp = c->pdps[7].val | (u64) c->pdps[6].val << 32;
+
+ if (shadow_pdp != m->ppgtt_mm.shadow_pdps[0]) {
+ gvt_dbg_mm("4-level context ppgtt not match LRI command\n");
+ return false;
+ }
+ return true;
+ } else {
+ /* see comment in LRI handler in cmd_parser.c */
+ gvt_dbg_mm("invalid shadow mm type\n");
+ return false;
+ }
+}
+
+static void update_guest_context(struct intel_vgpu_workload *workload)
+{
+ struct i915_request *rq = workload->req;
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct execlist_ring_context *shadow_ring_context;
+ struct intel_context *ctx = workload->req->context;
+ void *context_base;
+ void *src;
+ unsigned long context_gpa, context_page_num;
+ unsigned long gpa_base; /* first gpa of consecutive GPAs */
+ unsigned long gpa_size; /* size of consecutive GPAs*/
+ int i;
+ u32 ring_base;
+ u32 head, tail;
+ u16 wrap_count;
+
+ gvt_dbg_sched("ring id %d workload lrca %x\n", rq->engine->id,
+ workload->ctx_desc.lrca);
+
+ GEM_BUG_ON(!intel_context_is_pinned(ctx));
+
+ head = workload->rb_head;
+ tail = workload->rb_tail;
+ wrap_count = workload->guest_rb_head >> RB_HEAD_WRAP_CNT_OFF;
+
+ if (tail < head) {
+ if (wrap_count == RB_HEAD_WRAP_CNT_MAX)
+ wrap_count = 0;
+ else
+ wrap_count += 1;
+ }
+
+ head = (wrap_count << RB_HEAD_WRAP_CNT_OFF) | tail;
+
+ ring_base = rq->engine->mmio_base;
+ vgpu_vreg_t(vgpu, RING_TAIL(ring_base)) = tail;
+ vgpu_vreg_t(vgpu, RING_HEAD(ring_base)) = head;
+
+ context_page_num = rq->engine->context_size;
+ context_page_num = context_page_num >> PAGE_SHIFT;
+
+ if (IS_BROADWELL(rq->engine->i915) && rq->engine->id == RCS0)
+ context_page_num = 19;
+
+ context_base = (void *) ctx->lrc_reg_state -
+ (LRC_STATE_PN << I915_GTT_PAGE_SHIFT);
+
+ /* find consecutive GPAs from gma until the first inconsecutive GPA.
+ * write to the consecutive GPAs from src virtual address
+ */
+ gpa_size = 0;
+ for (i = 2; i < context_page_num; i++) {
+ context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+ (u32)((workload->ctx_desc.lrca + i) <<
+ I915_GTT_PAGE_SHIFT));
+ if (context_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("invalid guest context descriptor\n");
+ return;
+ }
+
+ if (gpa_size == 0) {
+ gpa_base = context_gpa;
+ src = context_base + (i << I915_GTT_PAGE_SHIFT);
+ } else if (context_gpa != gpa_base + gpa_size)
+ goto write;
+
+ gpa_size += I915_GTT_PAGE_SIZE;
+
+ if (i == context_page_num - 1)
+ goto write;
+
+ continue;
+
+write:
+ intel_gvt_write_gpa(vgpu, gpa_base, src, gpa_size);
+ gpa_base = context_gpa;
+ gpa_size = I915_GTT_PAGE_SIZE;
+ src = context_base + (i << I915_GTT_PAGE_SHIFT);
+ }
+
+ intel_gvt_write_gpa(vgpu, workload->ring_context_gpa +
+ RING_CTX_OFF(ring_header.val), &workload->rb_tail, 4);
+
+ shadow_ring_context = (void *) ctx->lrc_reg_state;
+
+ if (!list_empty(&workload->lri_shadow_mm)) {
+ struct intel_vgpu_mm *m = list_last_entry(&workload->lri_shadow_mm,
+ struct intel_vgpu_mm,
+ ppgtt_mm.link);
+ GEM_BUG_ON(!check_shadow_context_ppgtt(shadow_ring_context, m));
+ update_guest_pdps(vgpu, workload->ring_context_gpa,
+ (void *)m->ppgtt_mm.guest_pdps);
+ }
+
+#define COPY_REG(name) \
+ intel_gvt_write_gpa(vgpu, workload->ring_context_gpa + \
+ RING_CTX_OFF(name.val), &shadow_ring_context->name.val, 4)
+
+ COPY_REG(ctx_ctrl);
+ COPY_REG(ctx_timestamp);
+
+#undef COPY_REG
+
+ intel_gvt_write_gpa(vgpu,
+ workload->ring_context_gpa +
+ sizeof(*shadow_ring_context),
+ (void *)shadow_ring_context +
+ sizeof(*shadow_ring_context),
+ I915_GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
+}
+
+void intel_vgpu_clean_workloads(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_engine_cs *engine;
+ struct intel_vgpu_workload *pos, *n;
+ intel_engine_mask_t tmp;
+
+ /* free the unsubmited workloads in the queues. */
+ for_each_engine_masked(engine, vgpu->gvt->gt, engine_mask, tmp) {
+ list_for_each_entry_safe(pos, n,
+ &s->workload_q_head[engine->id], list) {
+ list_del_init(&pos->list);
+ intel_vgpu_destroy_workload(pos);
+ }
+ clear_bit(engine->id, s->shadow_ctx_desc_updated);
+ }
+}
+
+static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
+{
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct intel_vgpu_workload *workload =
+ scheduler->current_workload[ring_id];
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_request *rq = workload->req;
+ int event;
+
+ mutex_lock(&vgpu->vgpu_lock);
+ mutex_lock(&gvt->sched_lock);
+
+ /* For the workload w/ request, needs to wait for the context
+ * switch to make sure request is completed.
+ * For the workload w/o request, directly complete the workload.
+ */
+ if (rq) {
+ wait_event(workload->shadow_ctx_status_wq,
+ !atomic_read(&workload->shadow_ctx_active));
+
+ /* If this request caused GPU hang, req->fence.error will
+ * be set to -EIO. Use -EIO to set workload status so
+ * that when this request caused GPU hang, didn't trigger
+ * context switch interrupt to guest.
+ */
+ if (likely(workload->status == -EINPROGRESS)) {
+ if (workload->req->fence.error == -EIO)
+ workload->status = -EIO;
+ else
+ workload->status = 0;
+ }
+
+ if (!workload->status &&
+ !(vgpu->resetting_eng & BIT(ring_id))) {
+ update_guest_context(workload);
+
+ for_each_set_bit(event, workload->pending_events,
+ INTEL_GVT_EVENT_MAX)
+ intel_vgpu_trigger_virtual_event(vgpu, event);
+ }
+
+ i915_request_put(fetch_and_zero(&workload->req));
+ }
+
+ gvt_dbg_sched("ring id %d complete workload %p status %d\n",
+ ring_id, workload, workload->status);
+
+ scheduler->current_workload[ring_id] = NULL;
+
+ list_del_init(&workload->list);
+
+ if (workload->status || vgpu->resetting_eng & BIT(ring_id)) {
+ /* if workload->status is not successful means HW GPU
+ * has occurred GPU hang or something wrong with i915/GVT,
+ * and GVT won't inject context switch interrupt to guest.
+ * So this error is a vGPU hang actually to the guest.
+ * According to this we should emunlate a vGPU hang. If
+ * there are pending workloads which are already submitted
+ * from guest, we should clean them up like HW GPU does.
+ *
+ * if it is in middle of engine resetting, the pending
+ * workloads won't be submitted to HW GPU and will be
+ * cleaned up during the resetting process later, so doing
+ * the workload clean up here doesn't have any impact.
+ **/
+ intel_vgpu_clean_workloads(vgpu, BIT(ring_id));
+ }
+
+ workload->complete(workload);
+
+ intel_vgpu_shadow_mm_unpin(workload);
+ intel_vgpu_destroy_workload(workload);
+
+ atomic_dec(&s->running_workload_num);
+ wake_up(&scheduler->workload_complete_wq);
+
+ if (gvt->scheduler.need_reschedule)
+ intel_gvt_request_service(gvt, INTEL_GVT_REQUEST_EVENT_SCHED);
+
+ mutex_unlock(&gvt->sched_lock);
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+static int workload_thread(void *arg)
+{
+ struct intel_engine_cs *engine = arg;
+ const bool need_force_wake = GRAPHICS_VER(engine->i915) >= 9;
+ struct intel_gvt *gvt = engine->i915->gvt;
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct intel_vgpu_workload *workload = NULL;
+ struct intel_vgpu *vgpu = NULL;
+ int ret;
+ DEFINE_WAIT_FUNC(wait, woken_wake_function);
+
+ gvt_dbg_core("workload thread for ring %s started\n", engine->name);
+
+ while (!kthread_should_stop()) {
+ intel_wakeref_t wakeref;
+
+ add_wait_queue(&scheduler->waitq[engine->id], &wait);
+ do {
+ workload = pick_next_workload(gvt, engine);
+ if (workload)
+ break;
+ wait_woken(&wait, TASK_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT);
+ } while (!kthread_should_stop());
+ remove_wait_queue(&scheduler->waitq[engine->id], &wait);
+
+ if (!workload)
+ break;
+
+ gvt_dbg_sched("ring %s next workload %p vgpu %d\n",
+ engine->name, workload,
+ workload->vgpu->id);
+
+ wakeref = intel_runtime_pm_get(engine->uncore->rpm);
+
+ gvt_dbg_sched("ring %s will dispatch workload %p\n",
+ engine->name, workload);
+
+ if (need_force_wake)
+ intel_uncore_forcewake_get(engine->uncore,
+ FORCEWAKE_ALL);
+ /*
+ * Update the vReg of the vGPU which submitted this
+ * workload. The vGPU may use these registers for checking
+ * the context state. The value comes from GPU commands
+ * in this workload.
+ */
+ update_vreg_in_ctx(workload);
+
+ ret = dispatch_workload(workload);
+
+ if (ret) {
+ vgpu = workload->vgpu;
+ gvt_vgpu_err("fail to dispatch workload, skip\n");
+ goto complete;
+ }
+
+ gvt_dbg_sched("ring %s wait workload %p\n",
+ engine->name, workload);
+ i915_request_wait(workload->req, 0, MAX_SCHEDULE_TIMEOUT);
+
+complete:
+ gvt_dbg_sched("will complete workload %p, status: %d\n",
+ workload, workload->status);
+
+ complete_current_workload(gvt, engine->id);
+
+ if (need_force_wake)
+ intel_uncore_forcewake_put(engine->uncore,
+ FORCEWAKE_ALL);
+
+ intel_runtime_pm_put(engine->uncore->rpm, wakeref);
+ if (ret && (vgpu_is_vm_unhealthy(ret)))
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
+ }
+ return 0;
+}
+
+void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+
+ if (atomic_read(&s->running_workload_num)) {
+ gvt_dbg_sched("wait vgpu idle\n");
+
+ wait_event(scheduler->workload_complete_wq,
+ !atomic_read(&s->running_workload_num));
+ }
+}
+
+void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt)
+{
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id i;
+
+ gvt_dbg_core("clean workload scheduler\n");
+
+ for_each_engine(engine, gvt->gt, i) {
+ atomic_notifier_chain_unregister(
+ &engine->context_status_notifier,
+ &gvt->shadow_ctx_notifier_block[i]);
+ kthread_stop(scheduler->thread[i]);
+ }
+}
+
+int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt)
+{
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id i;
+ int ret;
+
+ gvt_dbg_core("init workload scheduler\n");
+
+ init_waitqueue_head(&scheduler->workload_complete_wq);
+
+ for_each_engine(engine, gvt->gt, i) {
+ init_waitqueue_head(&scheduler->waitq[i]);
+
+ scheduler->thread[i] = kthread_run(workload_thread, engine,
+ "gvt:%s", engine->name);
+ if (IS_ERR(scheduler->thread[i])) {
+ gvt_err("fail to create workload thread\n");
+ ret = PTR_ERR(scheduler->thread[i]);
+ goto err;
+ }
+
+ gvt->shadow_ctx_notifier_block[i].notifier_call =
+ shadow_context_status_change;
+ atomic_notifier_chain_register(&engine->context_status_notifier,
+ &gvt->shadow_ctx_notifier_block[i]);
+ }
+
+ return 0;
+
+err:
+ intel_gvt_clean_workload_scheduler(gvt);
+ return ret;
+}
+
+static void
+i915_context_ppgtt_root_restore(struct intel_vgpu_submission *s,
+ struct i915_ppgtt *ppgtt)
+{
+ int i;
+
+ if (i915_vm_is_4lvl(&ppgtt->vm)) {
+ set_dma_address(ppgtt->pd, s->i915_context_pml4);
+ } else {
+ for (i = 0; i < GEN8_3LVL_PDPES; i++) {
+ struct i915_page_directory * const pd =
+ i915_pd_entry(ppgtt->pd, i);
+
+ set_dma_address(pd, s->i915_context_pdps[i]);
+ }
+ }
+}
+
+/**
+ * intel_vgpu_clean_submission - free submission-related resource for vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is called when a vGPU is being destroyed.
+ *
+ */
+void intel_vgpu_clean_submission(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+
+ intel_vgpu_select_submission_ops(vgpu, ALL_ENGINES, 0);
+
+ i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(s->shadow[0]->vm));
+ for_each_engine(engine, vgpu->gvt->gt, id)
+ intel_context_put(s->shadow[id]);
+
+ kmem_cache_destroy(s->workloads);
+}
+
+
+/**
+ * intel_vgpu_reset_submission - reset submission-related resource for vGPU
+ * @vgpu: a vGPU
+ * @engine_mask: engines expected to be reset
+ *
+ * This function is called when a vGPU is being destroyed.
+ *
+ */
+void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+
+ if (!s->active)
+ return;
+
+ intel_vgpu_clean_workloads(vgpu, engine_mask);
+ s->ops->reset(vgpu, engine_mask);
+}
+
+static void
+i915_context_ppgtt_root_save(struct intel_vgpu_submission *s,
+ struct i915_ppgtt *ppgtt)
+{
+ int i;
+
+ if (i915_vm_is_4lvl(&ppgtt->vm)) {
+ s->i915_context_pml4 = px_dma(ppgtt->pd);
+ } else {
+ for (i = 0; i < GEN8_3LVL_PDPES; i++) {
+ struct i915_page_directory * const pd =
+ i915_pd_entry(ppgtt->pd, i);
+
+ s->i915_context_pdps[i] = px_dma(pd);
+ }
+ }
+}
+
+/**
+ * intel_vgpu_setup_submission - setup submission-related resource for vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is called when a vGPU is being created.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_engine_cs *engine;
+ struct i915_ppgtt *ppgtt;
+ enum intel_engine_id i;
+ int ret;
+
+ ppgtt = i915_ppgtt_create(to_gt(i915), I915_BO_ALLOC_PM_EARLY);
+ if (IS_ERR(ppgtt))
+ return PTR_ERR(ppgtt);
+
+ i915_context_ppgtt_root_save(s, ppgtt);
+
+ for_each_engine(engine, vgpu->gvt->gt, i) {
+ struct intel_context *ce;
+
+ INIT_LIST_HEAD(&s->workload_q_head[i]);
+ s->shadow[i] = ERR_PTR(-EINVAL);
+
+ ce = intel_context_create(engine);
+ if (IS_ERR(ce)) {
+ ret = PTR_ERR(ce);
+ goto out_shadow_ctx;
+ }
+
+ i915_vm_put(ce->vm);
+ ce->vm = i915_vm_get(&ppgtt->vm);
+ intel_context_set_single_submission(ce);
+
+ /* Max ring buffer size */
+ if (!intel_uc_wants_guc_submission(&engine->gt->uc))
+ ce->ring_size = SZ_2M;
+
+ s->shadow[i] = ce;
+ }
+
+ bitmap_zero(s->shadow_ctx_desc_updated, I915_NUM_ENGINES);
+
+ s->workloads = kmem_cache_create_usercopy("gvt-g_vgpu_workload",
+ sizeof(struct intel_vgpu_workload), 0,
+ SLAB_HWCACHE_ALIGN,
+ offsetof(struct intel_vgpu_workload, rb_tail),
+ sizeof_field(struct intel_vgpu_workload, rb_tail),
+ NULL);
+
+ if (!s->workloads) {
+ ret = -ENOMEM;
+ goto out_shadow_ctx;
+ }
+
+ atomic_set(&s->running_workload_num, 0);
+ bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
+
+ memset(s->last_ctx, 0, sizeof(s->last_ctx));
+
+ i915_vm_put(&ppgtt->vm);
+ return 0;
+
+out_shadow_ctx:
+ i915_context_ppgtt_root_restore(s, ppgtt);
+ for_each_engine(engine, vgpu->gvt->gt, i) {
+ if (IS_ERR(s->shadow[i]))
+ break;
+
+ intel_context_put(s->shadow[i]);
+ }
+ i915_vm_put(&ppgtt->vm);
+ return ret;
+}
+
+/**
+ * intel_vgpu_select_submission_ops - select virtual submission interface
+ * @vgpu: a vGPU
+ * @engine_mask: either ALL_ENGINES or target engine mask
+ * @interface: expected vGPU virtual submission interface
+ *
+ * This function is called when guest configures submission interface.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask,
+ unsigned int interface)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ const struct intel_vgpu_submission_ops *ops[] = {
+ [INTEL_VGPU_EXECLIST_SUBMISSION] =
+ &intel_vgpu_execlist_submission_ops,
+ };
+ int ret;
+
+ if (drm_WARN_ON(&i915->drm, interface >= ARRAY_SIZE(ops)))
+ return -EINVAL;
+
+ if (drm_WARN_ON(&i915->drm,
+ interface == 0 && engine_mask != ALL_ENGINES))
+ return -EINVAL;
+
+ if (s->active)
+ s->ops->clean(vgpu, engine_mask);
+
+ if (interface == 0) {
+ s->ops = NULL;
+ s->virtual_submission_interface = 0;
+ s->active = false;
+ gvt_dbg_core("vgpu%d: remove submission ops\n", vgpu->id);
+ return 0;
+ }
+
+ ret = ops[interface]->init(vgpu, engine_mask);
+ if (ret)
+ return ret;
+
+ s->ops = ops[interface];
+ s->virtual_submission_interface = interface;
+ s->active = true;
+
+ gvt_dbg_core("vgpu%d: activate ops [ %s ]\n",
+ vgpu->id, s->ops->name);
+
+ return 0;
+}
+
+/**
+ * intel_vgpu_destroy_workload - destroy a vGPU workload
+ * @workload: workload to destroy
+ *
+ * This function is called when destroy a vGPU workload.
+ *
+ */
+void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload)
+{
+ struct intel_vgpu_submission *s = &workload->vgpu->submission;
+
+ intel_context_unpin(s->shadow[workload->engine->id]);
+ release_shadow_batch_buffer(workload);
+ release_shadow_wa_ctx(&workload->wa_ctx);
+
+ if (!list_empty(&workload->lri_shadow_mm)) {
+ struct intel_vgpu_mm *m, *mm;
+ list_for_each_entry_safe(m, mm, &workload->lri_shadow_mm,
+ ppgtt_mm.link) {
+ list_del(&m->ppgtt_mm.link);
+ intel_vgpu_mm_put(m);
+ }
+ }
+
+ GEM_BUG_ON(!list_empty(&workload->lri_shadow_mm));
+ if (workload->shadow_mm)
+ intel_vgpu_mm_put(workload->shadow_mm);
+
+ kmem_cache_free(s->workloads, workload);
+}
+
+static struct intel_vgpu_workload *
+alloc_workload(struct intel_vgpu *vgpu)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_workload *workload;
+
+ workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL);
+ if (!workload)
+ return ERR_PTR(-ENOMEM);
+
+ INIT_LIST_HEAD(&workload->list);
+ INIT_LIST_HEAD(&workload->shadow_bb);
+ INIT_LIST_HEAD(&workload->lri_shadow_mm);
+
+ init_waitqueue_head(&workload->shadow_ctx_status_wq);
+ atomic_set(&workload->shadow_ctx_active, 0);
+
+ workload->status = -EINPROGRESS;
+ workload->vgpu = vgpu;
+
+ return workload;
+}
+
+#define RING_CTX_OFF(x) \
+ offsetof(struct execlist_ring_context, x)
+
+static void read_guest_pdps(struct intel_vgpu *vgpu,
+ u64 ring_context_gpa, u32 pdp[8])
+{
+ u64 gpa;
+ int i;
+
+ gpa = ring_context_gpa + RING_CTX_OFF(pdps[0].val);
+
+ for (i = 0; i < 8; i++)
+ intel_gvt_read_gpa(vgpu,
+ gpa + i * 8, &pdp[7 - i], 4);
+}
+
+static int prepare_mm(struct intel_vgpu_workload *workload)
+{
+ struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc;
+ struct intel_vgpu_mm *mm;
+ struct intel_vgpu *vgpu = workload->vgpu;
+ enum intel_gvt_gtt_type root_entry_type;
+ u64 pdps[GVT_RING_CTX_NR_PDPS];
+
+ switch (desc->addressing_mode) {
+ case 1: /* legacy 32-bit */
+ root_entry_type = GTT_TYPE_PPGTT_ROOT_L3_ENTRY;
+ break;
+ case 3: /* legacy 64-bit */
+ root_entry_type = GTT_TYPE_PPGTT_ROOT_L4_ENTRY;
+ break;
+ default:
+ gvt_vgpu_err("Advanced Context mode(SVM) is not supported!\n");
+ return -EINVAL;
+ }
+
+ read_guest_pdps(workload->vgpu, workload->ring_context_gpa, (void *)pdps);
+
+ mm = intel_vgpu_get_ppgtt_mm(workload->vgpu, root_entry_type, pdps);
+ if (IS_ERR(mm))
+ return PTR_ERR(mm);
+
+ workload->shadow_mm = mm;
+ return 0;
+}
+
+#define same_context(a, b) (((a)->context_id == (b)->context_id) && \
+ ((a)->lrca == (b)->lrca))
+
+/**
+ * intel_vgpu_create_workload - create a vGPU workload
+ * @vgpu: a vGPU
+ * @engine: the engine
+ * @desc: a guest context descriptor
+ *
+ * This function is called when creating a vGPU workload.
+ *
+ * Returns:
+ * struct intel_vgpu_workload * on success, negative error code in
+ * pointer if failed.
+ *
+ */
+struct intel_vgpu_workload *
+intel_vgpu_create_workload(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine,
+ struct execlist_ctx_descriptor_format *desc)
+{
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct list_head *q = workload_q_head(vgpu, engine);
+ struct intel_vgpu_workload *last_workload = NULL;
+ struct intel_vgpu_workload *workload = NULL;
+ u64 ring_context_gpa;
+ u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx;
+ u32 guest_head;
+ int ret;
+
+ ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+ (u32)((desc->lrca + 1) << I915_GTT_PAGE_SHIFT));
+ if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) {
+ gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca);
+ return ERR_PTR(-EINVAL);
+ }
+
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(ring_header.val), &head, 4);
+
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(ring_tail.val), &tail, 4);
+
+ guest_head = head;
+
+ head &= RB_HEAD_OFF_MASK;
+ tail &= RB_TAIL_OFF_MASK;
+
+ list_for_each_entry_reverse(last_workload, q, list) {
+
+ if (same_context(&last_workload->ctx_desc, desc)) {
+ gvt_dbg_el("ring %s cur workload == last\n",
+ engine->name);
+ gvt_dbg_el("ctx head %x real head %lx\n", head,
+ last_workload->rb_tail);
+ /*
+ * cannot use guest context head pointer here,
+ * as it might not be updated at this time
+ */
+ head = last_workload->rb_tail;
+ break;
+ }
+ }
+
+ gvt_dbg_el("ring %s begin a new workload\n", engine->name);
+
+ /* record some ring buffer register values for scan and shadow */
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(rb_start.val), &start, 4);
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(rb_ctrl.val), &ctl, 4);
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4);
+
+ if (!intel_gvt_ggtt_validate_range(vgpu, start,
+ _RING_CTL_BUF_SIZE(ctl))) {
+ gvt_vgpu_err("context contain invalid rb at: 0x%x\n", start);
+ return ERR_PTR(-EINVAL);
+ }
+
+ workload = alloc_workload(vgpu);
+ if (IS_ERR(workload))
+ return workload;
+
+ workload->engine = engine;
+ workload->ctx_desc = *desc;
+ workload->ring_context_gpa = ring_context_gpa;
+ workload->rb_head = head;
+ workload->guest_rb_head = guest_head;
+ workload->rb_tail = tail;
+ workload->rb_start = start;
+ workload->rb_ctl = ctl;
+
+ if (engine->id == RCS0) {
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4);
+ intel_gvt_read_gpa(vgpu, ring_context_gpa +
+ RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4);
+
+ workload->wa_ctx.indirect_ctx.guest_gma =
+ indirect_ctx & INDIRECT_CTX_ADDR_MASK;
+ workload->wa_ctx.indirect_ctx.size =
+ (indirect_ctx & INDIRECT_CTX_SIZE_MASK) *
+ CACHELINE_BYTES;
+
+ if (workload->wa_ctx.indirect_ctx.size != 0) {
+ if (!intel_gvt_ggtt_validate_range(vgpu,
+ workload->wa_ctx.indirect_ctx.guest_gma,
+ workload->wa_ctx.indirect_ctx.size)) {
+ gvt_vgpu_err("invalid wa_ctx at: 0x%lx\n",
+ workload->wa_ctx.indirect_ctx.guest_gma);
+ kmem_cache_free(s->workloads, workload);
+ return ERR_PTR(-EINVAL);
+ }
+ }
+
+ workload->wa_ctx.per_ctx.guest_gma =
+ per_ctx & PER_CTX_ADDR_MASK;
+ workload->wa_ctx.per_ctx.valid = per_ctx & 1;
+ if (workload->wa_ctx.per_ctx.valid) {
+ if (!intel_gvt_ggtt_validate_range(vgpu,
+ workload->wa_ctx.per_ctx.guest_gma,
+ CACHELINE_BYTES)) {
+ gvt_vgpu_err("invalid per_ctx at: 0x%lx\n",
+ workload->wa_ctx.per_ctx.guest_gma);
+ kmem_cache_free(s->workloads, workload);
+ return ERR_PTR(-EINVAL);
+ }
+ }
+ }
+
+ gvt_dbg_el("workload %p ring %s head %x tail %x start %x ctl %x\n",
+ workload, engine->name, head, tail, start, ctl);
+
+ ret = prepare_mm(workload);
+ if (ret) {
+ kmem_cache_free(s->workloads, workload);
+ return ERR_PTR(ret);
+ }
+
+ /* Only scan and shadow the first workload in the queue
+ * as there is only one pre-allocated buf-obj for shadow.
+ */
+ if (list_empty(q)) {
+ intel_wakeref_t wakeref;
+
+ with_intel_runtime_pm(engine->gt->uncore->rpm, wakeref)
+ ret = intel_gvt_scan_and_shadow_workload(workload);
+ }
+
+ if (ret) {
+ if (vgpu_is_vm_unhealthy(ret))
+ enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
+ intel_vgpu_destroy_workload(workload);
+ return ERR_PTR(ret);
+ }
+
+ ret = intel_context_pin(s->shadow[engine->id]);
+ if (ret) {
+ intel_vgpu_destroy_workload(workload);
+ return ERR_PTR(ret);
+ }
+
+ return workload;
+}
+
+/**
+ * intel_vgpu_queue_workload - Qeue a vGPU workload
+ * @workload: the workload to queue in
+ */
+void intel_vgpu_queue_workload(struct intel_vgpu_workload *workload)
+{
+ list_add_tail(&workload->list,
+ workload_q_head(workload->vgpu, workload->engine));
+ intel_gvt_kick_schedule(workload->vgpu->gvt);
+ wake_up(&workload->vgpu->gvt->scheduler.waitq[workload->engine->id]);
+}
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h
new file mode 100644
index 000000000..1f391b3da
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -0,0 +1,171 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ * Ping Gao <ping.a.gao@intel.com>
+ * Tina Zhang <tina.zhang@intel.com>
+ * Chanbin Du <changbin.du@intel.com>
+ * Min He <min.he@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ * Zhenyu Wang <zhenyuw@linux.intel.com>
+ *
+ */
+
+#ifndef _GVT_SCHEDULER_H_
+#define _GVT_SCHEDULER_H_
+
+#include "gt/intel_engine_types.h"
+
+#include "execlist.h"
+#include "interrupt.h"
+
+struct intel_gvt_workload_scheduler {
+ struct intel_vgpu *current_vgpu;
+ struct intel_vgpu *next_vgpu;
+ struct intel_vgpu_workload *current_workload[I915_NUM_ENGINES];
+ bool need_reschedule;
+
+ spinlock_t mmio_context_lock;
+ /* can be null when owner is host */
+ struct intel_vgpu *engine_owner[I915_NUM_ENGINES];
+
+ wait_queue_head_t workload_complete_wq;
+ struct task_struct *thread[I915_NUM_ENGINES];
+ wait_queue_head_t waitq[I915_NUM_ENGINES];
+
+ void *sched_data;
+ const struct intel_gvt_sched_policy_ops *sched_ops;
+};
+
+#define INDIRECT_CTX_ADDR_MASK 0xffffffc0
+#define INDIRECT_CTX_SIZE_MASK 0x3f
+struct shadow_indirect_ctx {
+ struct drm_i915_gem_object *obj;
+ unsigned long guest_gma;
+ unsigned long shadow_gma;
+ void *shadow_va;
+ u32 size;
+};
+
+#define PER_CTX_ADDR_MASK 0xfffff000
+struct shadow_per_ctx {
+ unsigned long guest_gma;
+ unsigned long shadow_gma;
+ unsigned valid;
+};
+
+struct intel_shadow_wa_ctx {
+ struct shadow_indirect_ctx indirect_ctx;
+ struct shadow_per_ctx per_ctx;
+
+};
+
+struct intel_vgpu_workload {
+ struct intel_vgpu *vgpu;
+ const struct intel_engine_cs *engine;
+ struct i915_request *req;
+ /* if this workload has been dispatched to i915? */
+ bool dispatched;
+ bool shadow; /* if workload has done shadow of guest request */
+ int status;
+
+ struct intel_vgpu_mm *shadow_mm;
+ struct list_head lri_shadow_mm; /* For PPGTT load cmd */
+
+ /* different submission model may need different handler */
+ int (*prepare)(struct intel_vgpu_workload *);
+ int (*complete)(struct intel_vgpu_workload *);
+ struct list_head list;
+
+ DECLARE_BITMAP(pending_events, INTEL_GVT_EVENT_MAX);
+ void *shadow_ring_buffer_va;
+
+ /* execlist context information */
+ struct execlist_ctx_descriptor_format ctx_desc;
+ struct execlist_ring_context *ring_context;
+ unsigned long rb_head, rb_tail, rb_ctl, rb_start, rb_len;
+ unsigned long guest_rb_head;
+ bool restore_inhibit;
+ struct intel_vgpu_elsp_dwords elsp_dwords;
+ bool emulate_schedule_in;
+ atomic_t shadow_ctx_active;
+ wait_queue_head_t shadow_ctx_status_wq;
+ u64 ring_context_gpa;
+
+ /* shadow batch buffer */
+ struct list_head shadow_bb;
+ struct intel_shadow_wa_ctx wa_ctx;
+
+ /* oa registers */
+ u32 oactxctrl;
+ u32 flex_mmio[7];
+};
+
+struct intel_vgpu_shadow_bb {
+ struct list_head list;
+ struct drm_i915_gem_object *obj;
+ struct i915_vma *vma;
+ void *va;
+ u32 *bb_start_cmd_va;
+ unsigned long bb_offset;
+ bool ppgtt;
+};
+
+#define workload_q_head(vgpu, e) \
+ (&(vgpu)->submission.workload_q_head[(e)->id])
+
+void intel_vgpu_queue_workload(struct intel_vgpu_workload *workload);
+
+int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt);
+
+void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt);
+
+void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu);
+
+int intel_vgpu_setup_submission(struct intel_vgpu *vgpu);
+
+void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask);
+
+void intel_vgpu_clean_submission(struct intel_vgpu *vgpu);
+
+int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask,
+ unsigned int interface);
+
+extern const struct intel_vgpu_submission_ops
+intel_vgpu_execlist_submission_ops;
+
+struct intel_vgpu_workload *
+intel_vgpu_create_workload(struct intel_vgpu *vgpu,
+ const struct intel_engine_cs *engine,
+ struct execlist_ctx_descriptor_format *desc);
+
+void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload);
+
+void intel_vgpu_clean_workloads(struct intel_vgpu *vgpu,
+ intel_engine_mask_t engine_mask);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/trace.h b/drivers/gpu/drm/i915/gvt/trace.h
new file mode 100644
index 000000000..020f1aa28
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/trace.h
@@ -0,0 +1,383 @@
+/*
+ * Copyright © 2011-2016 Intel Corporation
+ *
+ * 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
+ * paragraph) 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.
+ *
+ * Authors:
+ * Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#if !defined(_GVT_TRACE_H_) || defined(TRACE_HEADER_MULTI_READ)
+#define _GVT_TRACE_H_
+
+#include <linux/types.h>
+#include <linux/stringify.h>
+#include <linux/tracepoint.h>
+#include <asm/tsc.h>
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM gvt
+
+TRACE_EVENT(spt_alloc,
+ TP_PROTO(int id, void *spt, int type, unsigned long mfn,
+ unsigned long gpt_gfn),
+
+ TP_ARGS(id, spt, type, mfn, gpt_gfn),
+
+ TP_STRUCT__entry(
+ __field(int, id)
+ __field(void *, spt)
+ __field(int, type)
+ __field(unsigned long, mfn)
+ __field(unsigned long, gpt_gfn)
+ ),
+
+ TP_fast_assign(
+ __entry->id = id;
+ __entry->spt = spt;
+ __entry->type = type;
+ __entry->mfn = mfn;
+ __entry->gpt_gfn = gpt_gfn;
+ ),
+
+ TP_printk("VM%d [alloc] spt %p type %d mfn 0x%lx gfn 0x%lx\n",
+ __entry->id,
+ __entry->spt,
+ __entry->type,
+ __entry->mfn,
+ __entry->gpt_gfn)
+);
+
+TRACE_EVENT(spt_free,
+ TP_PROTO(int id, void *spt, int type),
+
+ TP_ARGS(id, spt, type),
+
+ TP_STRUCT__entry(
+ __field(int, id)
+ __field(void *, spt)
+ __field(int, type)
+ ),
+
+ TP_fast_assign(
+ __entry->id = id;
+ __entry->spt = spt;
+ __entry->type = type;
+ ),
+
+ TP_printk("VM%u [free] spt %p type %d\n",
+ __entry->id,
+ __entry->spt,
+ __entry->type)
+);
+
+#define MAX_BUF_LEN 256
+
+TRACE_EVENT(gma_index,
+ TP_PROTO(const char *prefix, unsigned long gma,
+ unsigned long index),
+
+ TP_ARGS(prefix, gma, index),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "%s gma 0x%lx index 0x%lx\n", prefix, gma, index);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(gma_translate,
+ TP_PROTO(int id, char *type, int ring_id, int root_entry_type,
+ unsigned long gma, unsigned long gpa),
+
+ TP_ARGS(id, type, ring_id, root_entry_type, gma, gpa),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d %s ring %d root_entry_type %d gma 0x%lx -> gpa 0x%lx\n",
+ id, type, ring_id, root_entry_type, gma, gpa);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(spt_refcount,
+ TP_PROTO(int id, char *action, void *spt, int before, int after),
+
+ TP_ARGS(id, action, spt, before, after),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d [%s] spt %p before %d -> after %d\n",
+ id, action, spt, before, after);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(spt_change,
+ TP_PROTO(int id, char *action, void *spt, unsigned long gfn,
+ int type),
+
+ TP_ARGS(id, action, spt, gfn, type),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d [%s] spt %p gfn 0x%lx type %d\n",
+ id, action, spt, gfn, type);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(spt_guest_change,
+ TP_PROTO(int id, const char *tag, void *spt, int type, u64 v,
+ unsigned long index),
+
+ TP_ARGS(id, tag, spt, type, v, index),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d [%s] spt %p type %d entry 0x%llx index 0x%lx\n",
+ id, tag, spt, type, v, index);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(oos_change,
+ TP_PROTO(int id, const char *tag, int page_id, void *gpt, int type),
+
+ TP_ARGS(id, tag, page_id, gpt, type),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d [oos %s] page id %d gpt %p type %d\n",
+ id, tag, page_id, gpt, type);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(oos_sync,
+ TP_PROTO(int id, int page_id, void *gpt, int type, u64 v,
+ unsigned long index),
+
+ TP_ARGS(id, page_id, gpt, type, v, index),
+
+ TP_STRUCT__entry(
+ __array(char, buf, MAX_BUF_LEN)
+ ),
+
+ TP_fast_assign(
+ snprintf(__entry->buf, MAX_BUF_LEN,
+ "VM%d [oos sync] page id %d gpt %p type %d entry 0x%llx index 0x%lx\n",
+ id, page_id, gpt, type, v, index);
+ ),
+
+ TP_printk("%s", __entry->buf)
+);
+
+#define GVT_CMD_STR_LEN 40
+TRACE_EVENT(gvt_command,
+ TP_PROTO(u8 vgpu_id, u8 ring_id, u32 ip_gma, u32 *cmd_va,
+ u32 cmd_len, u32 buf_type, u32 buf_addr_type,
+ void *workload, const char *cmd_name),
+
+ TP_ARGS(vgpu_id, ring_id, ip_gma, cmd_va, cmd_len, buf_type,
+ buf_addr_type, workload, cmd_name),
+
+ TP_STRUCT__entry(
+ __field(u8, vgpu_id)
+ __field(u8, ring_id)
+ __field(u32, ip_gma)
+ __field(u32, buf_type)
+ __field(u32, buf_addr_type)
+ __field(u32, cmd_len)
+ __field(void*, workload)
+ __dynamic_array(u32, raw_cmd, cmd_len)
+ __array(char, cmd_name, GVT_CMD_STR_LEN)
+ ),
+
+ TP_fast_assign(
+ __entry->vgpu_id = vgpu_id;
+ __entry->ring_id = ring_id;
+ __entry->ip_gma = ip_gma;
+ __entry->buf_type = buf_type;
+ __entry->buf_addr_type = buf_addr_type;
+ __entry->cmd_len = cmd_len;
+ __entry->workload = workload;
+ snprintf(__entry->cmd_name, GVT_CMD_STR_LEN, "%s", cmd_name);
+ memcpy(__get_dynamic_array(raw_cmd), cmd_va, cmd_len * sizeof(*cmd_va));
+ ),
+
+
+ TP_printk("vgpu%d ring %d: address_type %u, buf_type %u, ip_gma %08x,cmd (name=%s,len=%u,raw cmd=%s), workload=%p\n",
+ __entry->vgpu_id,
+ __entry->ring_id,
+ __entry->buf_addr_type,
+ __entry->buf_type,
+ __entry->ip_gma,
+ __entry->cmd_name,
+ __entry->cmd_len,
+ __print_array(__get_dynamic_array(raw_cmd),
+ __entry->cmd_len, 4),
+ __entry->workload)
+);
+
+#define GVT_TEMP_STR_LEN 10
+TRACE_EVENT(write_ir,
+ TP_PROTO(int id, char *reg_name, unsigned int reg, unsigned int new_val,
+ unsigned int old_val, bool changed),
+
+ TP_ARGS(id, reg_name, reg, new_val, old_val, changed),
+
+ TP_STRUCT__entry(
+ __field(int, id)
+ __array(char, buf, GVT_TEMP_STR_LEN)
+ __field(unsigned int, reg)
+ __field(unsigned int, new_val)
+ __field(unsigned int, old_val)
+ __field(bool, changed)
+ ),
+
+ TP_fast_assign(
+ __entry->id = id;
+ snprintf(__entry->buf, GVT_TEMP_STR_LEN, "%s", reg_name);
+ __entry->reg = reg;
+ __entry->new_val = new_val;
+ __entry->old_val = old_val;
+ __entry->changed = changed;
+ ),
+
+ TP_printk("VM%u write [%s] %x, new %08x, old %08x, changed %08x\n",
+ __entry->id, __entry->buf, __entry->reg, __entry->new_val,
+ __entry->old_val, __entry->changed)
+);
+
+TRACE_EVENT(propagate_event,
+ TP_PROTO(int id, const char *irq_name, int bit),
+
+ TP_ARGS(id, irq_name, bit),
+
+ TP_STRUCT__entry(
+ __field(int, id)
+ __array(char, buf, GVT_TEMP_STR_LEN)
+ __field(int, bit)
+ ),
+
+ TP_fast_assign(
+ __entry->id = id;
+ snprintf(__entry->buf, GVT_TEMP_STR_LEN, "%s", irq_name);
+ __entry->bit = bit;
+ ),
+
+ TP_printk("Set bit (%d) for (%s) for vgpu (%d)\n",
+ __entry->bit, __entry->buf, __entry->id)
+);
+
+TRACE_EVENT(inject_msi,
+ TP_PROTO(int id, unsigned int address, unsigned int data),
+
+ TP_ARGS(id, address, data),
+
+ TP_STRUCT__entry(
+ __field(int, id)
+ __field(unsigned int, address)
+ __field(unsigned int, data)
+ ),
+
+ TP_fast_assign(
+ __entry->id = id;
+ __entry->address = address;
+ __entry->data = data;
+ ),
+
+ TP_printk("vgpu%d:inject msi address %x data %x\n",
+ __entry->id, __entry->address, __entry->data)
+);
+
+TRACE_EVENT(render_mmio,
+ TP_PROTO(int old_id, int new_id, char *action, unsigned int reg,
+ unsigned int old_val, unsigned int new_val),
+
+ TP_ARGS(old_id, new_id, action, reg, old_val, new_val),
+
+ TP_STRUCT__entry(
+ __field(int, old_id)
+ __field(int, new_id)
+ __array(char, buf, GVT_TEMP_STR_LEN)
+ __field(unsigned int, reg)
+ __field(unsigned int, old_val)
+ __field(unsigned int, new_val)
+ ),
+
+ TP_fast_assign(
+ __entry->old_id = old_id;
+ __entry->new_id = new_id;
+ snprintf(__entry->buf, GVT_TEMP_STR_LEN, "%s", action);
+ __entry->reg = reg;
+ __entry->old_val = old_val;
+ __entry->new_val = new_val;
+ ),
+
+ TP_printk("VM%u -> VM%u %s reg %x, old %08x new %08x\n",
+ __entry->old_id, __entry->new_id,
+ __entry->buf, __entry->reg,
+ __entry->old_val, __entry->new_val)
+);
+
+#endif /* _GVT_TRACE_H_ */
+
+/* This part must be out of protection */
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/i915/gvt
+#define TRACE_INCLUDE_FILE trace
+#include <trace/define_trace.h>
diff --git a/drivers/gpu/drm/i915/gvt/trace_points.c b/drivers/gpu/drm/i915/gvt/trace_points.c
new file mode 100644
index 000000000..fe552e877
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/trace_points.c
@@ -0,0 +1,34 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ * Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef __CHECKER__
+#define CREATE_TRACE_POINTS
+#include "trace.h"
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c
new file mode 100644
index 000000000..56c714740
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -0,0 +1,512 @@
+/*
+ * Copyright(c) 2011-2016 Intel 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 (including the next
+ * paragraph) 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.
+ *
+ * Authors:
+ * Eddie Dong <eddie.dong@intel.com>
+ * Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ * Ping Gao <ping.a.gao@intel.com>
+ * Zhi Wang <zhi.a.wang@intel.com>
+ * Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+#include "i915_pvinfo.h"
+
+void populate_pvinfo_page(struct intel_vgpu *vgpu)
+{
+ struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
+ /* setup the ballooning information */
+ vgpu_vreg64_t(vgpu, vgtif_reg(magic)) = VGT_MAGIC;
+ vgpu_vreg_t(vgpu, vgtif_reg(version_major)) = 1;
+ vgpu_vreg_t(vgpu, vgtif_reg(version_minor)) = 0;
+ vgpu_vreg_t(vgpu, vgtif_reg(display_ready)) = 0;
+ vgpu_vreg_t(vgpu, vgtif_reg(vgt_id)) = vgpu->id;
+
+ vgpu_vreg_t(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_PPGTT;
+ vgpu_vreg_t(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HWSP_EMULATION;
+ vgpu_vreg_t(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HUGE_GTT;
+
+ vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) =
+ vgpu_aperture_gmadr_base(vgpu);
+ vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
+ vgpu_aperture_sz(vgpu);
+ vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.base)) =
+ vgpu_hidden_gmadr_base(vgpu);
+ vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.size)) =
+ vgpu_hidden_sz(vgpu);
+
+ vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.fence_num)) = vgpu_fence_sz(vgpu);
+
+ vgpu_vreg_t(vgpu, vgtif_reg(cursor_x_hot)) = UINT_MAX;
+ vgpu_vreg_t(vgpu, vgtif_reg(cursor_y_hot)) = UINT_MAX;
+
+ gvt_dbg_core("Populate PVINFO PAGE for vGPU %d\n", vgpu->id);
+ gvt_dbg_core("aperture base [GMADR] 0x%llx size 0x%llx\n",
+ vgpu_aperture_gmadr_base(vgpu), vgpu_aperture_sz(vgpu));
+ gvt_dbg_core("hidden base [GMADR] 0x%llx size=0x%llx\n",
+ vgpu_hidden_gmadr_base(vgpu), vgpu_hidden_sz(vgpu));
+ gvt_dbg_core("fence size %d\n", vgpu_fence_sz(vgpu));
+
+ drm_WARN_ON(&i915->drm, sizeof(struct vgt_if) != VGT_PVINFO_SIZE);
+}
+
+/*
+ * vGPU type name is defined as GVTg_Vx_y which contains the physical GPU
+ * generation type (e.g V4 as BDW server, V5 as SKL server).
+ *
+ * Depening on the physical SKU resource, we might see vGPU types like
+ * GVTg_V4_8, GVTg_V4_4, GVTg_V4_2, etc. We can create different types of
+ * vGPU on same physical GPU depending on available resource. Each vGPU
+ * type will have a different number of avail_instance to indicate how
+ * many vGPU instance can be created for this type.
+ */
+#define VGPU_MAX_WEIGHT 16
+#define VGPU_WEIGHT(vgpu_num) \
+ (VGPU_MAX_WEIGHT / (vgpu_num))
+
+static const struct intel_vgpu_config intel_vgpu_configs[] = {
+ { MB_TO_BYTES(64), MB_TO_BYTES(384), 4, VGPU_WEIGHT(8), GVT_EDID_1024_768, "8" },
+ { MB_TO_BYTES(128), MB_TO_BYTES(512), 4, VGPU_WEIGHT(4), GVT_EDID_1920_1200, "4" },
+ { MB_TO_BYTES(256), MB_TO_BYTES(1024), 4, VGPU_WEIGHT(2), GVT_EDID_1920_1200, "2" },
+ { MB_TO_BYTES(512), MB_TO_BYTES(2048), 4, VGPU_WEIGHT(1), GVT_EDID_1920_1200, "1" },
+};
+
+/**
+ * intel_gvt_init_vgpu_types - initialize vGPU type list
+ * @gvt : GVT device
+ *
+ * Initialize vGPU type list based on available resource.
+ *
+ */
+int intel_gvt_init_vgpu_types(struct intel_gvt *gvt)
+{
+ unsigned int low_avail = gvt_aperture_sz(gvt) - HOST_LOW_GM_SIZE;
+ unsigned int high_avail = gvt_hidden_sz(gvt) - HOST_HIGH_GM_SIZE;
+ unsigned int num_types = ARRAY_SIZE(intel_vgpu_configs);
+ unsigned int i;
+
+ gvt->types = kcalloc(num_types, sizeof(struct intel_vgpu_type),
+ GFP_KERNEL);
+ if (!gvt->types)
+ return -ENOMEM;
+
+ gvt->mdev_types = kcalloc(num_types, sizeof(*gvt->mdev_types),
+ GFP_KERNEL);
+ if (!gvt->mdev_types)
+ goto out_free_types;
+
+ for (i = 0; i < num_types; ++i) {
+ const struct intel_vgpu_config *conf = &intel_vgpu_configs[i];
+
+ if (low_avail / conf->low_mm == 0)
+ break;
+ if (conf->weight < 1 || conf->weight > VGPU_MAX_WEIGHT)
+ goto out_free_mdev_types;
+
+ sprintf(gvt->types[i].name, "GVTg_V%u_%s",
+ GRAPHICS_VER(gvt->gt->i915) == 8 ? 4 : 5, conf->name);
+ gvt->types[i].conf = conf;
+
+ gvt_dbg_core("type[%d]: %s avail %u low %u high %u fence %u weight %u res %s\n",
+ i, gvt->types[i].name,
+ min(low_avail / conf->low_mm,
+ high_avail / conf->high_mm),
+ conf->low_mm, conf->high_mm, conf->fence,
+ conf->weight, vgpu_edid_str(conf->edid));
+
+ gvt->mdev_types[i] = &gvt->types[i].type;
+ gvt->mdev_types[i]->sysfs_name = gvt->types[i].name;
+ }
+
+ gvt->num_types = i;
+ return 0;
+
+out_free_mdev_types:
+ kfree(gvt->mdev_types);
+out_free_types:
+ kfree(gvt->types);
+ return -EINVAL;
+}
+
+void intel_gvt_clean_vgpu_types(struct intel_gvt *gvt)
+{
+ kfree(gvt->mdev_types);
+ kfree(gvt->types);
+}
+
+/**
+ * intel_gvt_active_vgpu - activate a virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to activate a virtual GPU.
+ *
+ */
+void intel_gvt_activate_vgpu(struct intel_vgpu *vgpu)
+{
+ mutex_lock(&vgpu->vgpu_lock);
+ vgpu->active = true;
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+/**
+ * intel_gvt_deactive_vgpu - deactivate a virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to deactivate a virtual GPU.
+ * The virtual GPU will be stopped.
+ *
+ */
+void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu)
+{
+ mutex_lock(&vgpu->vgpu_lock);
+
+ vgpu->active = false;
+
+ if (atomic_read(&vgpu->submission.running_workload_num)) {
+ mutex_unlock(&vgpu->vgpu_lock);
+ intel_gvt_wait_vgpu_idle(vgpu);
+ mutex_lock(&vgpu->vgpu_lock);
+ }
+
+ intel_vgpu_stop_schedule(vgpu);
+
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+/**
+ * intel_gvt_release_vgpu - release a virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to release a virtual GPU.
+ * The virtual GPU will be stopped and all runtime information will be
+ * destroyed.
+ *
+ */
+void intel_gvt_release_vgpu(struct intel_vgpu *vgpu)
+{
+ intel_gvt_deactivate_vgpu(vgpu);
+
+ mutex_lock(&vgpu->vgpu_lock);
+ vgpu->d3_entered = false;
+ intel_vgpu_clean_workloads(vgpu, ALL_ENGINES);
+ intel_vgpu_dmabuf_cleanup(vgpu);
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+/**
+ * intel_gvt_destroy_vgpu - destroy a virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to destroy a virtual GPU.
+ *
+ */
+void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *i915 = gvt->gt->i915;
+
+ drm_WARN(&i915->drm, vgpu->active, "vGPU is still active!\n");
+
+ /*
+ * remove idr first so later clean can judge if need to stop
+ * service if no active vgpu.
+ */
+ mutex_lock(&gvt->lock);
+ idr_remove(&gvt->vgpu_idr, vgpu->id);
+ mutex_unlock(&gvt->lock);
+
+ mutex_lock(&vgpu->vgpu_lock);
+ intel_gvt_debugfs_remove_vgpu(vgpu);
+ intel_vgpu_clean_sched_policy(vgpu);
+ intel_vgpu_clean_submission(vgpu);
+ intel_vgpu_clean_display(vgpu);
+ intel_vgpu_clean_opregion(vgpu);
+ intel_vgpu_reset_ggtt(vgpu, true);
+ intel_vgpu_clean_gtt(vgpu);
+ intel_vgpu_detach_regions(vgpu);
+ intel_vgpu_free_resource(vgpu);
+ intel_vgpu_clean_mmio(vgpu);
+ intel_vgpu_dmabuf_cleanup(vgpu);
+ mutex_unlock(&vgpu->vgpu_lock);
+}
+
+#define IDLE_VGPU_IDR 0
+
+/**
+ * intel_gvt_create_idle_vgpu - create an idle virtual GPU
+ * @gvt: GVT device
+ *
+ * This function is called when user wants to create an idle virtual GPU.
+ *
+ * Returns:
+ * pointer to intel_vgpu, error pointer if failed.
+ */
+struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt)
+{
+ struct intel_vgpu *vgpu;
+ enum intel_engine_id i;
+ int ret;
+
+ vgpu = vzalloc(sizeof(*vgpu));
+ if (!vgpu)
+ return ERR_PTR(-ENOMEM);
+
+ vgpu->id = IDLE_VGPU_IDR;
+ vgpu->gvt = gvt;
+ mutex_init(&vgpu->vgpu_lock);
+
+ for (i = 0; i < I915_NUM_ENGINES; i++)
+ INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]);
+
+ ret = intel_vgpu_init_sched_policy(vgpu);
+ if (ret)
+ goto out_free_vgpu;
+
+ vgpu->active = false;
+
+ return vgpu;
+
+out_free_vgpu:
+ vfree(vgpu);
+ return ERR_PTR(ret);
+}
+
+/**
+ * intel_gvt_destroy_vgpu - destroy an idle virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to destroy an idle virtual GPU.
+ *
+ */
+void intel_gvt_destroy_idle_vgpu(struct intel_vgpu *vgpu)
+{
+ mutex_lock(&vgpu->vgpu_lock);
+ intel_vgpu_clean_sched_policy(vgpu);
+ mutex_unlock(&vgpu->vgpu_lock);
+
+ vfree(vgpu);
+}
+
+int intel_gvt_create_vgpu(struct intel_vgpu *vgpu,
+ const struct intel_vgpu_config *conf)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct drm_i915_private *dev_priv = gvt->gt->i915;
+ int ret;
+
+ gvt_dbg_core("low %u MB high %u MB fence %u\n",
+ BYTES_TO_MB(conf->low_mm), BYTES_TO_MB(conf->high_mm),
+ conf->fence);
+
+ mutex_lock(&gvt->lock);
+ ret = idr_alloc(&gvt->vgpu_idr, vgpu, IDLE_VGPU_IDR + 1, GVT_MAX_VGPU,
+ GFP_KERNEL);
+ if (ret < 0)
+ goto out_unlock;;
+
+ vgpu->id = ret;
+ vgpu->sched_ctl.weight = conf->weight;
+ mutex_init(&vgpu->vgpu_lock);
+ mutex_init(&vgpu->dmabuf_lock);
+ INIT_LIST_HEAD(&vgpu->dmabuf_obj_list_head);
+ INIT_RADIX_TREE(&vgpu->page_track_tree, GFP_KERNEL);
+ idr_init_base(&vgpu->object_idr, 1);
+ intel_vgpu_init_cfg_space(vgpu, 1);
+ vgpu->d3_entered = false;
+
+ ret = intel_vgpu_init_mmio(vgpu);
+ if (ret)
+ goto out_clean_idr;
+
+ ret = intel_vgpu_alloc_resource(vgpu, conf);
+ if (ret)
+ goto out_clean_vgpu_mmio;
+
+ populate_pvinfo_page(vgpu);
+
+ ret = intel_vgpu_init_gtt(vgpu);
+ if (ret)
+ goto out_clean_vgpu_resource;
+
+ ret = intel_vgpu_init_opregion(vgpu);
+ if (ret)
+ goto out_clean_gtt;
+
+ ret = intel_vgpu_init_display(vgpu, conf->edid);
+ if (ret)
+ goto out_clean_opregion;
+
+ ret = intel_vgpu_setup_submission(vgpu);
+ if (ret)
+ goto out_clean_display;
+
+ ret = intel_vgpu_init_sched_policy(vgpu);
+ if (ret)
+ goto out_clean_submission;
+
+ intel_gvt_debugfs_add_vgpu(vgpu);
+
+ ret = intel_gvt_set_opregion(vgpu);
+ if (ret)
+ goto out_clean_sched_policy;
+
+ if (IS_BROADWELL(dev_priv) || IS_BROXTON(dev_priv))
+ ret = intel_gvt_set_edid(vgpu, PORT_B);
+ else
+ ret = intel_gvt_set_edid(vgpu, PORT_D);
+ if (ret)
+ goto out_clean_sched_policy;
+
+ intel_gvt_update_reg_whitelist(vgpu);
+ mutex_unlock(&gvt->lock);
+ return 0;
+
+out_clean_sched_policy:
+ intel_vgpu_clean_sched_policy(vgpu);
+out_clean_submission:
+ intel_vgpu_clean_submission(vgpu);
+out_clean_display:
+ intel_vgpu_clean_display(vgpu);
+out_clean_opregion:
+ intel_vgpu_clean_opregion(vgpu);
+out_clean_gtt:
+ intel_vgpu_clean_gtt(vgpu);
+out_clean_vgpu_resource:
+ intel_vgpu_free_resource(vgpu);
+out_clean_vgpu_mmio:
+ intel_vgpu_clean_mmio(vgpu);
+out_clean_idr:
+ idr_remove(&gvt->vgpu_idr, vgpu->id);
+out_unlock:
+ mutex_unlock(&gvt->lock);
+ return ret;
+}
+
+/**
+ * intel_gvt_reset_vgpu_locked - reset a virtual GPU by DMLR or GT reset
+ * @vgpu: virtual GPU
+ * @dmlr: vGPU Device Model Level Reset or GT Reset
+ * @engine_mask: engines to reset for GT reset
+ *
+ * This function is called when user wants to reset a virtual GPU through
+ * device model reset or GT reset. The caller should hold the vgpu lock.
+ *
+ * vGPU Device Model Level Reset (DMLR) simulates the PCI level reset to reset
+ * the whole vGPU to default state as when it is created. This vGPU function
+ * is required both for functionary and security concerns.The ultimate goal
+ * of vGPU FLR is that reuse a vGPU instance by virtual machines. When we
+ * assign a vGPU to a virtual machine we must isse such reset first.
+ *
+ * Full GT Reset and Per-Engine GT Reset are soft reset flow for GPU engines
+ * (Render, Blitter, Video, Video Enhancement). It is defined by GPU Spec.
+ * Unlike the FLR, GT reset only reset particular resource of a vGPU per
+ * the reset request. Guest driver can issue a GT reset by programming the
+ * virtual GDRST register to reset specific virtual GPU engine or all
+ * engines.
+ *
+ * The parameter dev_level is to identify if we will do DMLR or GT reset.
+ * The parameter engine_mask is to specific the engines that need to be
+ * resetted. If value ALL_ENGINES is given for engine_mask, it means
+ * the caller requests a full GT reset that we will reset all virtual
+ * GPU engines. For FLR, engine_mask is ignored.
+ */
+void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
+ intel_engine_mask_t engine_mask)
+{
+ struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+ intel_engine_mask_t resetting_eng = dmlr ? ALL_ENGINES : engine_mask;
+
+ gvt_dbg_core("------------------------------------------\n");
+ gvt_dbg_core("resseting vgpu%d, dmlr %d, engine_mask %08x\n",
+ vgpu->id, dmlr, engine_mask);
+
+ vgpu->resetting_eng = resetting_eng;
+
+ intel_vgpu_stop_schedule(vgpu);
+ /*
+ * The current_vgpu will set to NULL after stopping the
+ * scheduler when the reset is triggered by current vgpu.
+ */
+ if (scheduler->current_vgpu == NULL) {
+ mutex_unlock(&vgpu->vgpu_lock);
+ intel_gvt_wait_vgpu_idle(vgpu);
+ mutex_lock(&vgpu->vgpu_lock);
+ }
+
+ intel_vgpu_reset_submission(vgpu, resetting_eng);
+ /* full GPU reset or device model level reset */
+ if (engine_mask == ALL_ENGINES || dmlr) {
+ intel_vgpu_select_submission_ops(vgpu, ALL_ENGINES, 0);
+ if (engine_mask == ALL_ENGINES)
+ intel_vgpu_invalidate_ppgtt(vgpu);
+ /*fence will not be reset during virtual reset */
+ if (dmlr) {
+ if(!vgpu->d3_entered) {
+ intel_vgpu_invalidate_ppgtt(vgpu);
+ intel_vgpu_destroy_all_ppgtt_mm(vgpu);
+ }
+ intel_vgpu_reset_ggtt(vgpu, true);
+ intel_vgpu_reset_resource(vgpu);
+ }
+
+ intel_vgpu_reset_mmio(vgpu, dmlr);
+ populate_pvinfo_page(vgpu);
+
+ if (dmlr) {
+ intel_vgpu_reset_display(vgpu);
+ intel_vgpu_reset_cfg_space(vgpu);
+ /* only reset the failsafe mode when dmlr reset */
+ vgpu->failsafe = false;
+ /*
+ * PCI_D0 is set before dmlr, so reset d3_entered here
+ * after done using.
+ */
+ if(vgpu->d3_entered)
+ vgpu->d3_entered = false;
+ else
+ vgpu->pv_notified = false;
+ }
+ }
+
+ vgpu->resetting_eng = 0;
+ gvt_dbg_core("reset vgpu%d done\n", vgpu->id);
+ gvt_dbg_core("------------------------------------------\n");
+}
+
+/**
+ * intel_gvt_reset_vgpu - reset a virtual GPU (Function Level)
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to reset a virtual GPU.
+ *
+ */
+void intel_gvt_reset_vgpu(struct intel_vgpu *vgpu)
+{
+ mutex_lock(&vgpu->vgpu_lock);
+ intel_gvt_reset_vgpu_locked(vgpu, true, 0);
+ mutex_unlock(&vgpu->vgpu_lock);
+}