summaryrefslogtreecommitdiffstats
path: root/drivers/edac
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-11 08:27:49 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-11 08:27:49 +0000
commitace9429bb58fd418f0c81d4c2835699bddf6bde6 (patch)
treeb2d64bc10158fdd5497876388cd68142ca374ed3 /drivers/edac
parentInitial commit. (diff)
downloadlinux-ace9429bb58fd418f0c81d4c2835699bddf6bde6.tar.xz
linux-ace9429bb58fd418f0c81d4c2835699bddf6bde6.zip
Adding upstream version 6.6.15.upstream/6.6.15
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r--drivers/edac/Kconfig564
-rw-r--r--drivers/edac/Makefile88
-rw-r--r--drivers/edac/al_mc_edac.c350
-rw-r--r--drivers/edac/altera_edac.c2227
-rw-r--r--drivers/edac/altera_edac.h398
-rw-r--r--drivers/edac/amd64_edac.c4472
-rw-r--r--drivers/edac/amd64_edac.h529
-rw-r--r--drivers/edac/amd76x_edac.c375
-rw-r--r--drivers/edac/amd8111_edac.c597
-rw-r--r--drivers/edac/amd8111_edac.h118
-rw-r--r--drivers/edac/amd8131_edac.c358
-rw-r--r--drivers/edac/amd8131_edac.h107
-rw-r--r--drivers/edac/armada_xp_edac.c624
-rw-r--r--drivers/edac/aspeed_edac.c399
-rw-r--r--drivers/edac/bluefield_edac.c356
-rw-r--r--drivers/edac/cell_edac.c282
-rw-r--r--drivers/edac/cpc925_edac.c1067
-rw-r--r--drivers/edac/debugfs.c151
-rw-r--r--drivers/edac/dmc520_edac.c656
-rw-r--r--drivers/edac/e752x_edac.c1481
-rw-r--r--drivers/edac/e7xxx_edac.c602
-rw-r--r--drivers/edac/edac_device.c621
-rw-r--r--drivers/edac/edac_device.h365
-rw-r--r--drivers/edac/edac_device_sysfs.c873
-rw-r--r--drivers/edac/edac_mc.c1091
-rw-r--r--drivers/edac/edac_mc.h258
-rw-r--r--drivers/edac/edac_mc_sysfs.c1055
-rw-r--r--drivers/edac/edac_module.c168
-rw-r--r--drivers/edac/edac_module.h125
-rw-r--r--drivers/edac/edac_pci.c337
-rw-r--r--drivers/edac/edac_pci.h271
-rw-r--r--drivers/edac/edac_pci_sysfs.c743
-rw-r--r--drivers/edac/fsl_ddr_edac.c633
-rw-r--r--drivers/edac/fsl_ddr_edac.h76
-rw-r--r--drivers/edac/ghes_edac.c573
-rw-r--r--drivers/edac/highbank_l2_edac.c143
-rw-r--r--drivers/edac/highbank_mc_edac.c276
-rw-r--r--drivers/edac/i10nm_base.c1230
-rw-r--r--drivers/edac/i3000_edac.c565
-rw-r--r--drivers/edac/i3200_edac.c545
-rw-r--r--drivers/edac/i5000_edac.c1582
-rw-r--r--drivers/edac/i5100_edac.c1224
-rw-r--r--drivers/edac/i5400_edac.c1469
-rw-r--r--drivers/edac/i7300_edac.c1214
-rw-r--r--drivers/edac/i7core_edac.c2399
-rw-r--r--drivers/edac/i82443bxgx_edac.c462
-rw-r--r--drivers/edac/i82860_edac.c362
-rw-r--r--drivers/edac/i82875p_edac.c599
-rw-r--r--drivers/edac/i82975x_edac.c706
-rw-r--r--drivers/edac/ie31200_edac.c667
-rw-r--r--drivers/edac/igen6_edac.c1304
-rw-r--r--drivers/edac/layerscape_edac.c75
-rw-r--r--drivers/edac/mce_amd.c1447
-rw-r--r--drivers/edac/mce_amd.h82
-rw-r--r--drivers/edac/mpc85xx_edac.c713
-rw-r--r--drivers/edac/mpc85xx_edac.h105
-rw-r--r--drivers/edac/npcm_edac.c544
-rw-r--r--drivers/edac/octeon_edac-l2c.c207
-rw-r--r--drivers/edac/octeon_edac-lmc.c324
-rw-r--r--drivers/edac/octeon_edac-pc.c142
-rw-r--r--drivers/edac/octeon_edac-pci.c110
-rw-r--r--drivers/edac/pasemi_edac.c303
-rw-r--r--drivers/edac/pnd2_edac.c1587
-rw-r--r--drivers/edac/pnd2_edac.h293
-rw-r--r--drivers/edac/ppc4xx_edac.c1428
-rw-r--r--drivers/edac/ppc4xx_edac.h167
-rw-r--r--drivers/edac/qcom_edac.c420
-rw-r--r--drivers/edac/r82600_edac.c426
-rw-r--r--drivers/edac/sb_edac.c3688
-rw-r--r--drivers/edac/sifive_edac.c119
-rw-r--r--drivers/edac/skx_base.c759
-rw-r--r--drivers/edac/skx_common.c723
-rw-r--r--drivers/edac/skx_common.h263
-rw-r--r--drivers/edac/synopsys_edac.c1445
-rw-r--r--drivers/edac/thunderx_edac.c2154
-rw-r--r--drivers/edac/ti_edac.c338
-rw-r--r--drivers/edac/wq.c42
-rw-r--r--drivers/edac/x38_edac.c523
-rw-r--r--drivers/edac/xgene_edac.c2045
-rw-r--r--drivers/edac/zynqmp_edac.c467
80 files changed, 58676 insertions, 0 deletions
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
new file mode 100644
index 0000000000..110e99b86a
--- /dev/null
+++ b/drivers/edac/Kconfig
@@ -0,0 +1,564 @@
+#
+# EDAC Kconfig
+# Copyright (c) 2008 Doug Thompson www.softwarebitmaker.com
+# Licensed and distributed under the GPL
+
+config EDAC_ATOMIC_SCRUB
+ bool
+
+config EDAC_SUPPORT
+ bool
+
+menuconfig EDAC
+ tristate "EDAC (Error Detection And Correction) reporting"
+ depends on HAS_IOMEM && EDAC_SUPPORT && RAS
+ help
+ EDAC is a subsystem along with hardware-specific drivers designed to
+ report hardware errors. These are low-level errors that are reported
+ in the CPU or supporting chipset or other subsystems:
+ memory errors, cache errors, PCI errors, thermal throttling, etc..
+ If unsure, select 'Y'.
+
+ The mailing list for the EDAC project is linux-edac@vger.kernel.org.
+
+if EDAC
+
+config EDAC_LEGACY_SYSFS
+ bool "EDAC legacy sysfs"
+ default y
+ help
+ Enable the compatibility sysfs nodes.
+ Use 'Y' if your edac utilities aren't ported to work with the newer
+ structures.
+
+config EDAC_DEBUG
+ bool "Debugging"
+ select DEBUG_FS
+ help
+ This turns on debugging information for the entire EDAC subsystem.
+ You do so by inserting edac_module with "edac_debug_level=x." Valid
+ levels are 0-4 (from low to high) and by default it is set to 2.
+ Usually you should select 'N' here.
+
+config EDAC_DECODE_MCE
+ tristate "Decode MCEs in human-readable form (only on AMD for now)"
+ depends on CPU_SUP_AMD && X86_MCE_AMD
+ default y
+ help
+ Enable this option if you want to decode Machine Check Exceptions
+ occurring on your machine in human-readable form.
+
+ You should definitely say Y here in case you want to decode MCEs
+ which occur really early upon boot, before the module infrastructure
+ has been initialized.
+
+config EDAC_GHES
+ tristate "Output ACPI APEI/GHES BIOS detected errors via EDAC"
+ depends on ACPI_APEI_GHES
+ select UEFI_CPER
+ help
+ Not all machines support hardware-driven error report. Some of those
+ provide a BIOS-driven error report mechanism via ACPI, using the
+ APEI/GHES driver. By enabling this option, the error reports provided
+ by GHES are sent to userspace via the EDAC API.
+
+ When this option is enabled, it will disable the hardware-driven
+ mechanisms, if a GHES BIOS is detected, entering into the
+ "Firmware First" mode.
+
+ It should be noticed that keeping both GHES and a hardware-driven
+ error mechanism won't work well, as BIOS will race with OS, while
+ reading the error registers. So, if you want to not use "Firmware
+ first" GHES error mechanism, you should disable GHES either at
+ compilation time or by passing "ghes.disable=1" Kernel parameter
+ at boot time.
+
+ In doubt, say 'Y'.
+
+config EDAC_AMD64
+ tristate "AMD64 (Opteron, Athlon64)"
+ depends on AMD_NB && EDAC_DECODE_MCE
+ help
+ Support for error detection and correction of DRAM ECC errors on
+ the AMD64 families (>= K8) of memory controllers.
+
+ When EDAC_DEBUG is enabled, hardware error injection facilities
+ through sysfs are available:
+
+ AMD CPUs up to and excluding family 0x17 provide for Memory
+ Error Injection into the ECC detection circuits. The amd64_edac
+ module allows the operator/user to inject Uncorrectable and
+ Correctable errors into DRAM.
+
+ When enabled, in each of the respective memory controller directories
+ (/sys/devices/system/edac/mc/mcX), there are 3 input files:
+
+ - inject_section (0..3, 16-byte section of 64-byte cacheline),
+ - inject_word (0..8, 16-bit word of 16-byte section),
+ - inject_ecc_vector (hex ecc vector: select bits of inject word)
+
+ In addition, there are two control files, inject_read and inject_write,
+ which trigger the DRAM ECC Read and Write respectively.
+
+config EDAC_AL_MC
+ tristate "Amazon's Annapurna Lab Memory Controller"
+ depends on (ARCH_ALPINE || COMPILE_TEST)
+ help
+ Support for error detection and correction for Amazon's Annapurna
+ Labs Alpine chips which allow 1 bit correction and 2 bits detection.
+
+config EDAC_AMD76X
+ tristate "AMD 76x (760, 762, 768)"
+ depends on PCI && X86_32
+ help
+ Support for error detection and correction on the AMD 76x
+ series of chipsets used with the Athlon processor.
+
+config EDAC_E7XXX
+ tristate "Intel e7xxx (e7205, e7500, e7501, e7505)"
+ depends on PCI && X86_32
+ help
+ Support for error detection and correction on the Intel
+ E7205, E7500, E7501 and E7505 server chipsets.
+
+config EDAC_E752X
+ tristate "Intel e752x (e7520, e7525, e7320) and 3100"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ E7520, E7525, E7320 server chipsets.
+
+config EDAC_I82443BXGX
+ tristate "Intel 82443BX/GX (440BX/GX)"
+ depends on PCI && X86_32
+ depends on BROKEN
+ help
+ Support for error detection and correction on the Intel
+ 82443BX/GX memory controllers (440BX/GX chipsets).
+
+config EDAC_I82875P
+ tristate "Intel 82875p (D82875P, E7210)"
+ depends on PCI && X86_32
+ help
+ Support for error detection and correction on the Intel
+ DP82785P and E7210 server chipsets.
+
+config EDAC_I82975X
+ tristate "Intel 82975x (D82975x)"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ DP82975x server chipsets.
+
+config EDAC_I3000
+ tristate "Intel 3000/3010"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ 3000 and 3010 server chipsets.
+
+config EDAC_I3200
+ tristate "Intel 3200"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ 3200 and 3210 server chipsets.
+
+config EDAC_IE31200
+ tristate "Intel e312xx"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ E3-1200 based DRAM controllers.
+
+config EDAC_X38
+ tristate "Intel X38"
+ depends on PCI && X86
+ help
+ Support for error detection and correction on the Intel
+ X38 server chipsets.
+
+config EDAC_I5400
+ tristate "Intel 5400 (Seaburg) chipsets"
+ depends on PCI && X86
+ help
+ Support for error detection and correction the Intel
+ i5400 MCH chipset (Seaburg).
+
+config EDAC_I7CORE
+ tristate "Intel i7 Core (Nehalem) processors"
+ depends on PCI && X86 && X86_MCE_INTEL
+ help
+ Support for error detection and correction the Intel
+ i7 Core (Nehalem) Integrated Memory Controller that exists on
+ newer processors like i7 Core, i7 Core Extreme, Xeon 35xx
+ and Xeon 55xx processors.
+
+config EDAC_I82860
+ tristate "Intel 82860"
+ depends on PCI && X86_32
+ help
+ Support for error detection and correction on the Intel
+ 82860 chipset.
+
+config EDAC_R82600
+ tristate "Radisys 82600 embedded chipset"
+ depends on PCI && X86_32
+ help
+ Support for error detection and correction on the Radisys
+ 82600 embedded chipset.
+
+config EDAC_I5000
+ tristate "Intel Greencreek/Blackford chipset"
+ depends on X86 && PCI
+ depends on BROKEN
+ help
+ Support for error detection and correction the Intel
+ Greekcreek/Blackford chipsets.
+
+config EDAC_I5100
+ tristate "Intel San Clemente MCH"
+ depends on X86 && PCI
+ help
+ Support for error detection and correction the Intel
+ San Clemente MCH.
+
+config EDAC_I7300
+ tristate "Intel Clarksboro MCH"
+ depends on X86 && PCI
+ help
+ Support for error detection and correction the Intel
+ Clarksboro MCH (Intel 7300 chipset).
+
+config EDAC_SBRIDGE
+ tristate "Intel Sandy-Bridge/Ivy-Bridge/Haswell Integrated MC"
+ depends on PCI && X86_64 && X86_MCE_INTEL && PCI_MMCONFIG
+ help
+ Support for error detection and correction the Intel
+ Sandy Bridge, Ivy Bridge and Haswell Integrated Memory Controllers.
+
+config EDAC_SKX
+ tristate "Intel Skylake server Integrated MC"
+ depends on PCI && X86_64 && X86_MCE_INTEL && PCI_MMCONFIG && ACPI
+ depends on ACPI_NFIT || !ACPI_NFIT # if ACPI_NFIT=m, EDAC_SKX can't be y
+ select DMI
+ select ACPI_ADXL
+ help
+ Support for error detection and correction the Intel
+ Skylake server Integrated Memory Controllers. If your
+ system has non-volatile DIMMs you should also manually
+ select CONFIG_ACPI_NFIT.
+
+config EDAC_I10NM
+ tristate "Intel 10nm server Integrated MC"
+ depends on PCI && X86_64 && X86_MCE_INTEL && PCI_MMCONFIG && ACPI
+ depends on ACPI_NFIT || !ACPI_NFIT # if ACPI_NFIT=m, EDAC_I10NM can't be y
+ select DMI
+ select ACPI_ADXL
+ help
+ Support for error detection and correction the Intel
+ 10nm server Integrated Memory Controllers. If your
+ system has non-volatile DIMMs you should also manually
+ select CONFIG_ACPI_NFIT.
+
+config EDAC_PND2
+ tristate "Intel Pondicherry2"
+ depends on PCI && X86_64 && X86_MCE_INTEL
+ select P2SB if X86
+ help
+ Support for error detection and correction on the Intel
+ Pondicherry2 Integrated Memory Controller. This SoC IP is
+ first used on the Apollo Lake platform and Denverton
+ micro-server but may appear on others in the future.
+
+config EDAC_IGEN6
+ tristate "Intel client SoC Integrated MC"
+ depends on PCI && PCI_MMCONFIG && ARCH_HAVE_NMI_SAFE_CMPXCHG
+ depends on X86_64 && X86_MCE_INTEL
+ help
+ Support for error detection and correction on the Intel
+ client SoC Integrated Memory Controller using In-Band ECC IP.
+ This In-Band ECC is first used on the Elkhart Lake SoC but
+ may appear on others in the future.
+
+config EDAC_MPC85XX
+ bool "Freescale MPC83xx / MPC85xx"
+ depends on FSL_SOC && EDAC=y
+ help
+ Support for error detection and correction on the Freescale
+ MPC8349, MPC8560, MPC8540, MPC8548, T4240
+
+config EDAC_LAYERSCAPE
+ tristate "Freescale Layerscape DDR"
+ depends on ARCH_LAYERSCAPE || SOC_LS1021A
+ help
+ Support for error detection and correction on Freescale memory
+ controllers on Layerscape SoCs.
+
+config EDAC_PASEMI
+ tristate "PA Semi PWRficient"
+ depends on PPC_PASEMI && PCI
+ help
+ Support for error detection and correction on PA Semi
+ PWRficient.
+
+config EDAC_CELL
+ tristate "Cell Broadband Engine memory controller"
+ depends on PPC_CELL_COMMON
+ help
+ Support for error detection and correction on the
+ Cell Broadband Engine internal memory controller
+ on platform without a hypervisor
+
+config EDAC_PPC4XX
+ tristate "PPC4xx IBM DDR2 Memory Controller"
+ depends on 4xx
+ help
+ This enables support for EDAC on the ECC memory used
+ with the IBM DDR2 memory controller found in various
+ PowerPC 4xx embedded processors such as the 405EX[r],
+ 440SP, 440SPe, 460EX, 460GT and 460SX.
+
+config EDAC_AMD8131
+ tristate "AMD8131 HyperTransport PCI-X Tunnel"
+ depends on PCI && PPC_MAPLE
+ help
+ Support for error detection and correction on the
+ AMD8131 HyperTransport PCI-X Tunnel chip.
+ Note, add more Kconfig dependency if it's adopted
+ on some machine other than Maple.
+
+config EDAC_AMD8111
+ tristate "AMD8111 HyperTransport I/O Hub"
+ depends on PCI && PPC_MAPLE
+ help
+ Support for error detection and correction on the
+ AMD8111 HyperTransport I/O Hub chip.
+ Note, add more Kconfig dependency if it's adopted
+ on some machine other than Maple.
+
+config EDAC_CPC925
+ tristate "IBM CPC925 Memory Controller (PPC970FX)"
+ depends on PPC64
+ help
+ Support for error detection and correction on the
+ IBM CPC925 Bridge and Memory Controller, which is
+ a companion chip to the PowerPC 970 family of
+ processors.
+
+config EDAC_HIGHBANK_MC
+ tristate "Highbank Memory Controller"
+ depends on ARCH_HIGHBANK
+ help
+ Support for error detection and correction on the
+ Calxeda Highbank memory controller.
+
+config EDAC_HIGHBANK_L2
+ tristate "Highbank L2 Cache"
+ depends on ARCH_HIGHBANK
+ help
+ Support for error detection and correction on the
+ Calxeda Highbank memory controller.
+
+config EDAC_OCTEON_PC
+ tristate "Cavium Octeon Primary Caches"
+ depends on CPU_CAVIUM_OCTEON
+ help
+ Support for error detection and correction on the primary caches of
+ the cnMIPS cores of Cavium Octeon family SOCs.
+
+config EDAC_OCTEON_L2C
+ tristate "Cavium Octeon Secondary Caches (L2C)"
+ depends on CAVIUM_OCTEON_SOC
+ help
+ Support for error detection and correction on the
+ Cavium Octeon family of SOCs.
+
+config EDAC_OCTEON_LMC
+ tristate "Cavium Octeon DRAM Memory Controller (LMC)"
+ depends on CAVIUM_OCTEON_SOC
+ help
+ Support for error detection and correction on the
+ Cavium Octeon family of SOCs.
+
+config EDAC_OCTEON_PCI
+ tristate "Cavium Octeon PCI Controller"
+ depends on PCI && CAVIUM_OCTEON_SOC
+ help
+ Support for error detection and correction on the
+ Cavium Octeon family of SOCs.
+
+config EDAC_THUNDERX
+ tristate "Cavium ThunderX EDAC"
+ depends on ARM64
+ depends on PCI
+ help
+ Support for error detection and correction on the
+ Cavium ThunderX memory controllers (LMC), Cache
+ Coherent Processor Interconnect (CCPI) and L2 cache
+ blocks (TAD, CBC, MCI).
+
+config EDAC_ALTERA
+ bool "Altera SOCFPGA ECC"
+ depends on EDAC=y && ARCH_INTEL_SOCFPGA
+ help
+ Support for error detection and correction on the
+ Altera SOCs. This is the global enable for the
+ various Altera peripherals.
+
+config EDAC_ALTERA_SDRAM
+ bool "Altera SDRAM ECC"
+ depends on EDAC_ALTERA=y
+ help
+ Support for error detection and correction on the
+ Altera SDRAM Memory for Altera SoCs. Note that the
+ preloader must initialize the SDRAM before loading
+ the kernel.
+
+config EDAC_ALTERA_L2C
+ bool "Altera L2 Cache ECC"
+ depends on EDAC_ALTERA=y && CACHE_L2X0
+ help
+ Support for error detection and correction on the
+ Altera L2 cache Memory for Altera SoCs. This option
+ requires L2 cache.
+
+config EDAC_ALTERA_OCRAM
+ bool "Altera On-Chip RAM ECC"
+ depends on EDAC_ALTERA=y && SRAM && GENERIC_ALLOCATOR
+ help
+ Support for error detection and correction on the
+ Altera On-Chip RAM Memory for Altera SoCs.
+
+config EDAC_ALTERA_ETHERNET
+ bool "Altera Ethernet FIFO ECC"
+ depends on EDAC_ALTERA=y
+ help
+ Support for error detection and correction on the
+ Altera Ethernet FIFO Memory for Altera SoCs.
+
+config EDAC_ALTERA_NAND
+ bool "Altera NAND FIFO ECC"
+ depends on EDAC_ALTERA=y && MTD_NAND_DENALI
+ help
+ Support for error detection and correction on the
+ Altera NAND FIFO Memory for Altera SoCs.
+
+config EDAC_ALTERA_DMA
+ bool "Altera DMA FIFO ECC"
+ depends on EDAC_ALTERA=y && PL330_DMA=y
+ help
+ Support for error detection and correction on the
+ Altera DMA FIFO Memory for Altera SoCs.
+
+config EDAC_ALTERA_USB
+ bool "Altera USB FIFO ECC"
+ depends on EDAC_ALTERA=y && USB_DWC2
+ help
+ Support for error detection and correction on the
+ Altera USB FIFO Memory for Altera SoCs.
+
+config EDAC_ALTERA_QSPI
+ bool "Altera QSPI FIFO ECC"
+ depends on EDAC_ALTERA=y && SPI_CADENCE_QUADSPI
+ help
+ Support for error detection and correction on the
+ Altera QSPI FIFO Memory for Altera SoCs.
+
+config EDAC_ALTERA_SDMMC
+ bool "Altera SDMMC FIFO ECC"
+ depends on EDAC_ALTERA=y && MMC_DW
+ help
+ Support for error detection and correction on the
+ Altera SDMMC FIFO Memory for Altera SoCs.
+
+config EDAC_SIFIVE
+ bool "Sifive platform EDAC driver"
+ depends on EDAC=y && SIFIVE_CCACHE
+ help
+ Support for error detection and correction on the SiFive SoCs.
+
+config EDAC_ARMADA_XP
+ bool "Marvell Armada XP DDR and L2 Cache ECC"
+ depends on MACH_MVEBU_V7
+ help
+ Support for error correction and detection on the Marvell Aramada XP
+ DDR RAM and L2 cache controllers.
+
+config EDAC_SYNOPSYS
+ tristate "Synopsys DDR Memory Controller"
+ depends on ARCH_ZYNQ || ARCH_ZYNQMP || ARCH_INTEL_SOCFPGA || ARCH_MXC
+ help
+ Support for error detection and correction on the Synopsys DDR
+ memory controller.
+
+config EDAC_XGENE
+ tristate "APM X-Gene SoC"
+ depends on (ARM64 || COMPILE_TEST)
+ help
+ Support for error detection and correction on the
+ APM X-Gene family of SOCs.
+
+config EDAC_TI
+ tristate "Texas Instruments DDR3 ECC Controller"
+ depends on ARCH_KEYSTONE || SOC_DRA7XX
+ help
+ Support for error detection and correction on the TI SoCs.
+
+config EDAC_QCOM
+ tristate "QCOM EDAC Controller"
+ depends on ARCH_QCOM && QCOM_LLCC
+ help
+ Support for error detection and correction on the
+ Qualcomm Technologies, Inc. SoCs.
+
+ This driver reports Single Bit Errors (SBEs) and Double Bit Errors (DBEs).
+ As of now, it supports error reporting for Last Level Cache Controller (LLCC)
+ of Tag RAM and Data RAM.
+
+ For debugging issues having to do with stability and overall system
+ health, you should probably say 'Y' here.
+
+config EDAC_ASPEED
+ tristate "Aspeed AST BMC SoC"
+ depends on ARCH_ASPEED
+ help
+ Support for error detection and correction on the Aspeed AST BMC SoC.
+
+ First, ECC must be configured in the bootloader. Then, this driver
+ will expose error counters via the EDAC kernel framework.
+
+config EDAC_BLUEFIELD
+ tristate "Mellanox BlueField Memory ECC"
+ depends on ARM64 && ((MELLANOX_PLATFORM && ACPI) || COMPILE_TEST)
+ help
+ Support for error detection and correction on the
+ Mellanox BlueField SoCs.
+
+config EDAC_DMC520
+ tristate "ARM DMC-520 ECC"
+ depends on ARM64
+ help
+ Support for error detection and correction on the
+ SoCs with ARM DMC-520 DRAM controller.
+
+config EDAC_ZYNQMP
+ tristate "Xilinx ZynqMP OCM Controller"
+ depends on ARCH_ZYNQMP || COMPILE_TEST
+ help
+ This driver supports error detection and correction for the
+ Xilinx ZynqMP OCM (On Chip Memory) controller. It can also be
+ built as a module. In that case it will be called zynqmp_edac.
+
+config EDAC_NPCM
+ tristate "Nuvoton NPCM DDR Memory Controller"
+ depends on (ARCH_NPCM || COMPILE_TEST)
+ help
+ Support for error detection and correction on the Nuvoton NPCM DDR
+ memory controller.
+
+ The memory controller supports single bit error correction, double bit
+ error detection (in-line ECC in which a section 1/8th of the memory
+ device used to store data is used for ECC storage).
+
+endif # EDAC
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
new file mode 100644
index 0000000000..61945d3113
--- /dev/null
+++ b/drivers/edac/Makefile
@@ -0,0 +1,88 @@
+#
+# Makefile for the Linux kernel EDAC drivers.
+#
+# Copyright 02 Jul 2003, Linux Networx (http://lnxi.com)
+# This file may be distributed under the terms of the
+# GNU General Public License.
+#
+
+obj-$(CONFIG_EDAC) := edac_core.o
+
+edac_core-y := edac_mc.o edac_device.o edac_mc_sysfs.o
+edac_core-y += edac_module.o edac_device_sysfs.o wq.o
+
+edac_core-$(CONFIG_EDAC_DEBUG) += debugfs.o
+
+ifdef CONFIG_PCI
+edac_core-y += edac_pci.o edac_pci_sysfs.o
+endif
+
+obj-$(CONFIG_EDAC_GHES) += ghes_edac.o
+
+edac_mce_amd-y := mce_amd.o
+obj-$(CONFIG_EDAC_DECODE_MCE) += edac_mce_amd.o
+
+obj-$(CONFIG_EDAC_AL_MC) += al_mc_edac.o
+obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
+obj-$(CONFIG_EDAC_CPC925) += cpc925_edac.o
+obj-$(CONFIG_EDAC_I5000) += i5000_edac.o
+obj-$(CONFIG_EDAC_I5100) += i5100_edac.o
+obj-$(CONFIG_EDAC_I5400) += i5400_edac.o
+obj-$(CONFIG_EDAC_I7300) += i7300_edac.o
+obj-$(CONFIG_EDAC_I7CORE) += i7core_edac.o
+obj-$(CONFIG_EDAC_SBRIDGE) += sb_edac.o
+obj-$(CONFIG_EDAC_PND2) += pnd2_edac.o
+obj-$(CONFIG_EDAC_IGEN6) += igen6_edac.o
+obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o
+obj-$(CONFIG_EDAC_E752X) += e752x_edac.o
+obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o
+obj-$(CONFIG_EDAC_I82875P) += i82875p_edac.o
+obj-$(CONFIG_EDAC_I82975X) += i82975x_edac.o
+obj-$(CONFIG_EDAC_I3000) += i3000_edac.o
+obj-$(CONFIG_EDAC_I3200) += i3200_edac.o
+obj-$(CONFIG_EDAC_IE31200) += ie31200_edac.o
+obj-$(CONFIG_EDAC_X38) += x38_edac.o
+obj-$(CONFIG_EDAC_I82860) += i82860_edac.o
+obj-$(CONFIG_EDAC_R82600) += r82600_edac.o
+obj-$(CONFIG_EDAC_AMD64) += amd64_edac.o
+
+obj-$(CONFIG_EDAC_PASEMI) += pasemi_edac.o
+
+mpc85xx_edac_mod-y := fsl_ddr_edac.o mpc85xx_edac.o
+obj-$(CONFIG_EDAC_MPC85XX) += mpc85xx_edac_mod.o
+
+layerscape_edac_mod-y := fsl_ddr_edac.o layerscape_edac.o
+obj-$(CONFIG_EDAC_LAYERSCAPE) += layerscape_edac_mod.o
+
+skx_edac-y := skx_common.o skx_base.o
+obj-$(CONFIG_EDAC_SKX) += skx_edac.o
+
+i10nm_edac-y := skx_common.o i10nm_base.o
+obj-$(CONFIG_EDAC_I10NM) += i10nm_edac.o
+
+obj-$(CONFIG_EDAC_CELL) += cell_edac.o
+obj-$(CONFIG_EDAC_PPC4XX) += ppc4xx_edac.o
+obj-$(CONFIG_EDAC_AMD8111) += amd8111_edac.o
+obj-$(CONFIG_EDAC_AMD8131) += amd8131_edac.o
+
+obj-$(CONFIG_EDAC_HIGHBANK_MC) += highbank_mc_edac.o
+obj-$(CONFIG_EDAC_HIGHBANK_L2) += highbank_l2_edac.o
+
+obj-$(CONFIG_EDAC_OCTEON_PC) += octeon_edac-pc.o
+obj-$(CONFIG_EDAC_OCTEON_L2C) += octeon_edac-l2c.o
+obj-$(CONFIG_EDAC_OCTEON_LMC) += octeon_edac-lmc.o
+obj-$(CONFIG_EDAC_OCTEON_PCI) += octeon_edac-pci.o
+obj-$(CONFIG_EDAC_THUNDERX) += thunderx_edac.o
+
+obj-$(CONFIG_EDAC_ALTERA) += altera_edac.o
+obj-$(CONFIG_EDAC_SIFIVE) += sifive_edac.o
+obj-$(CONFIG_EDAC_ARMADA_XP) += armada_xp_edac.o
+obj-$(CONFIG_EDAC_SYNOPSYS) += synopsys_edac.o
+obj-$(CONFIG_EDAC_XGENE) += xgene_edac.o
+obj-$(CONFIG_EDAC_TI) += ti_edac.o
+obj-$(CONFIG_EDAC_QCOM) += qcom_edac.o
+obj-$(CONFIG_EDAC_ASPEED) += aspeed_edac.o
+obj-$(CONFIG_EDAC_BLUEFIELD) += bluefield_edac.o
+obj-$(CONFIG_EDAC_DMC520) += dmc520_edac.o
+obj-$(CONFIG_EDAC_NPCM) += npcm_edac.o
+obj-$(CONFIG_EDAC_ZYNQMP) += zynqmp_edac.o
diff --git a/drivers/edac/al_mc_edac.c b/drivers/edac/al_mc_edac.c
new file mode 100644
index 0000000000..178b9e581a
--- /dev/null
+++ b/drivers/edac/al_mc_edac.c
@@ -0,0 +1,350 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright 2019 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+ */
+#include <linux/bitfield.h>
+#include <linux/bitops.h>
+#include <linux/edac.h>
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>
+#include <linux/spinlock.h>
+#include "edac_module.h"
+
+/* Registers Offset */
+#define AL_MC_ECC_CFG 0x70
+#define AL_MC_ECC_CLEAR 0x7c
+#define AL_MC_ECC_ERR_COUNT 0x80
+#define AL_MC_ECC_CE_ADDR0 0x84
+#define AL_MC_ECC_CE_ADDR1 0x88
+#define AL_MC_ECC_UE_ADDR0 0xa4
+#define AL_MC_ECC_UE_ADDR1 0xa8
+#define AL_MC_ECC_CE_SYND0 0x8c
+#define AL_MC_ECC_CE_SYND1 0x90
+#define AL_MC_ECC_CE_SYND2 0x94
+#define AL_MC_ECC_UE_SYND0 0xac
+#define AL_MC_ECC_UE_SYND1 0xb0
+#define AL_MC_ECC_UE_SYND2 0xb4
+
+/* Registers Fields */
+#define AL_MC_ECC_CFG_SCRUB_DISABLED BIT(4)
+
+#define AL_MC_ECC_CLEAR_UE_COUNT BIT(3)
+#define AL_MC_ECC_CLEAR_CE_COUNT BIT(2)
+#define AL_MC_ECC_CLEAR_UE_ERR BIT(1)
+#define AL_MC_ECC_CLEAR_CE_ERR BIT(0)
+
+#define AL_MC_ECC_ERR_COUNT_UE GENMASK(31, 16)
+#define AL_MC_ECC_ERR_COUNT_CE GENMASK(15, 0)
+
+#define AL_MC_ECC_CE_ADDR0_RANK GENMASK(25, 24)
+#define AL_MC_ECC_CE_ADDR0_ROW GENMASK(17, 0)
+
+#define AL_MC_ECC_CE_ADDR1_BG GENMASK(25, 24)
+#define AL_MC_ECC_CE_ADDR1_BANK GENMASK(18, 16)
+#define AL_MC_ECC_CE_ADDR1_COLUMN GENMASK(11, 0)
+
+#define AL_MC_ECC_UE_ADDR0_RANK GENMASK(25, 24)
+#define AL_MC_ECC_UE_ADDR0_ROW GENMASK(17, 0)
+
+#define AL_MC_ECC_UE_ADDR1_BG GENMASK(25, 24)
+#define AL_MC_ECC_UE_ADDR1_BANK GENMASK(18, 16)
+#define AL_MC_ECC_UE_ADDR1_COLUMN GENMASK(11, 0)
+
+#define DRV_NAME "al_mc_edac"
+#define AL_MC_EDAC_MSG_MAX 256
+
+struct al_mc_edac {
+ void __iomem *mmio_base;
+ spinlock_t lock;
+ int irq_ce;
+ int irq_ue;
+};
+
+static void prepare_msg(char *message, size_t buffer_size,
+ enum hw_event_mc_err_type type,
+ u8 rank, u32 row, u8 bg, u8 bank, u16 column,
+ u32 syn0, u32 syn1, u32 syn2)
+{
+ snprintf(message, buffer_size,
+ "%s rank=0x%x row=0x%x bg=0x%x bank=0x%x col=0x%x syn0: 0x%x syn1: 0x%x syn2: 0x%x",
+ type == HW_EVENT_ERR_UNCORRECTED ? "UE" : "CE",
+ rank, row, bg, bank, column, syn0, syn1, syn2);
+}
+
+static int handle_ce(struct mem_ctl_info *mci)
+{
+ u32 eccerrcnt, ecccaddr0, ecccaddr1, ecccsyn0, ecccsyn1, ecccsyn2, row;
+ struct al_mc_edac *al_mc = mci->pvt_info;
+ char msg[AL_MC_EDAC_MSG_MAX];
+ u16 ce_count, column;
+ unsigned long flags;
+ u8 rank, bg, bank;
+
+ eccerrcnt = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_ERR_COUNT);
+ ce_count = FIELD_GET(AL_MC_ECC_ERR_COUNT_CE, eccerrcnt);
+ if (!ce_count)
+ return 0;
+
+ ecccaddr0 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_CE_ADDR0);
+ ecccaddr1 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_CE_ADDR1);
+ ecccsyn0 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_CE_SYND0);
+ ecccsyn1 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_CE_SYND1);
+ ecccsyn2 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_CE_SYND2);
+
+ writel_relaxed(AL_MC_ECC_CLEAR_CE_COUNT | AL_MC_ECC_CLEAR_CE_ERR,
+ al_mc->mmio_base + AL_MC_ECC_CLEAR);
+
+ dev_dbg(mci->pdev, "eccuaddr0=0x%08x eccuaddr1=0x%08x\n",
+ ecccaddr0, ecccaddr1);
+
+ rank = FIELD_GET(AL_MC_ECC_CE_ADDR0_RANK, ecccaddr0);
+ row = FIELD_GET(AL_MC_ECC_CE_ADDR0_ROW, ecccaddr0);
+
+ bg = FIELD_GET(AL_MC_ECC_CE_ADDR1_BG, ecccaddr1);
+ bank = FIELD_GET(AL_MC_ECC_CE_ADDR1_BANK, ecccaddr1);
+ column = FIELD_GET(AL_MC_ECC_CE_ADDR1_COLUMN, ecccaddr1);
+
+ prepare_msg(msg, sizeof(msg), HW_EVENT_ERR_CORRECTED,
+ rank, row, bg, bank, column,
+ ecccsyn0, ecccsyn1, ecccsyn2);
+
+ spin_lock_irqsave(&al_mc->lock, flags);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
+ ce_count, 0, 0, 0, 0, 0, -1, mci->ctl_name, msg);
+ spin_unlock_irqrestore(&al_mc->lock, flags);
+
+ return ce_count;
+}
+
+static int handle_ue(struct mem_ctl_info *mci)
+{
+ u32 eccerrcnt, eccuaddr0, eccuaddr1, eccusyn0, eccusyn1, eccusyn2, row;
+ struct al_mc_edac *al_mc = mci->pvt_info;
+ char msg[AL_MC_EDAC_MSG_MAX];
+ u16 ue_count, column;
+ unsigned long flags;
+ u8 rank, bg, bank;
+
+ eccerrcnt = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_ERR_COUNT);
+ ue_count = FIELD_GET(AL_MC_ECC_ERR_COUNT_UE, eccerrcnt);
+ if (!ue_count)
+ return 0;
+
+ eccuaddr0 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_UE_ADDR0);
+ eccuaddr1 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_UE_ADDR1);
+ eccusyn0 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_UE_SYND0);
+ eccusyn1 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_UE_SYND1);
+ eccusyn2 = readl_relaxed(al_mc->mmio_base + AL_MC_ECC_UE_SYND2);
+
+ writel_relaxed(AL_MC_ECC_CLEAR_UE_COUNT | AL_MC_ECC_CLEAR_UE_ERR,
+ al_mc->mmio_base + AL_MC_ECC_CLEAR);
+
+ dev_dbg(mci->pdev, "eccuaddr0=0x%08x eccuaddr1=0x%08x\n",
+ eccuaddr0, eccuaddr1);
+
+ rank = FIELD_GET(AL_MC_ECC_UE_ADDR0_RANK, eccuaddr0);
+ row = FIELD_GET(AL_MC_ECC_UE_ADDR0_ROW, eccuaddr0);
+
+ bg = FIELD_GET(AL_MC_ECC_UE_ADDR1_BG, eccuaddr1);
+ bank = FIELD_GET(AL_MC_ECC_UE_ADDR1_BANK, eccuaddr1);
+ column = FIELD_GET(AL_MC_ECC_UE_ADDR1_COLUMN, eccuaddr1);
+
+ prepare_msg(msg, sizeof(msg), HW_EVENT_ERR_UNCORRECTED,
+ rank, row, bg, bank, column,
+ eccusyn0, eccusyn1, eccusyn2);
+
+ spin_lock_irqsave(&al_mc->lock, flags);
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
+ ue_count, 0, 0, 0, 0, 0, -1, mci->ctl_name, msg);
+ spin_unlock_irqrestore(&al_mc->lock, flags);
+
+ return ue_count;
+}
+
+static void al_mc_edac_check(struct mem_ctl_info *mci)
+{
+ struct al_mc_edac *al_mc = mci->pvt_info;
+
+ if (al_mc->irq_ue <= 0)
+ handle_ue(mci);
+
+ if (al_mc->irq_ce <= 0)
+ handle_ce(mci);
+}
+
+static irqreturn_t al_mc_edac_irq_handler_ue(int irq, void *info)
+{
+ struct platform_device *pdev = info;
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ if (handle_ue(mci))
+ return IRQ_HANDLED;
+ return IRQ_NONE;
+}
+
+static irqreturn_t al_mc_edac_irq_handler_ce(int irq, void *info)
+{
+ struct platform_device *pdev = info;
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ if (handle_ce(mci))
+ return IRQ_HANDLED;
+ return IRQ_NONE;
+}
+
+static enum scrub_type get_scrub_mode(void __iomem *mmio_base)
+{
+ u32 ecccfg0;
+
+ ecccfg0 = readl(mmio_base + AL_MC_ECC_CFG);
+
+ if (FIELD_GET(AL_MC_ECC_CFG_SCRUB_DISABLED, ecccfg0))
+ return SCRUB_NONE;
+ else
+ return SCRUB_HW_SRC;
+}
+
+static void devm_al_mc_edac_free(void *data)
+{
+ edac_mc_free(data);
+}
+
+static void devm_al_mc_edac_del(void *data)
+{
+ edac_mc_del_mc(data);
+}
+
+static int al_mc_edac_probe(struct platform_device *pdev)
+{
+ struct edac_mc_layer layers[1];
+ struct mem_ctl_info *mci;
+ struct al_mc_edac *al_mc;
+ void __iomem *mmio_base;
+ struct dimm_info *dimm;
+ int ret;
+
+ mmio_base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(mmio_base)) {
+ dev_err(&pdev->dev, "failed to ioremap memory (%ld)\n",
+ PTR_ERR(mmio_base));
+ return PTR_ERR(mmio_base);
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct al_mc_edac));
+ if (!mci)
+ return -ENOMEM;
+
+ ret = devm_add_action_or_reset(&pdev->dev, devm_al_mc_edac_free, mci);
+ if (ret)
+ return ret;
+
+ platform_set_drvdata(pdev, mci);
+ al_mc = mci->pvt_info;
+
+ al_mc->mmio_base = mmio_base;
+
+ al_mc->irq_ue = of_irq_get_byname(pdev->dev.of_node, "ue");
+ if (al_mc->irq_ue <= 0)
+ dev_dbg(&pdev->dev,
+ "no IRQ defined for UE - falling back to polling\n");
+
+ al_mc->irq_ce = of_irq_get_byname(pdev->dev.of_node, "ce");
+ if (al_mc->irq_ce <= 0)
+ dev_dbg(&pdev->dev,
+ "no IRQ defined for CE - falling back to polling\n");
+
+ /*
+ * In case both interrupts (ue/ce) are to be found, use interrupt mode.
+ * In case none of the interrupt are foud, use polling mode.
+ * In case only one interrupt is found, use interrupt mode for it but
+ * keep polling mode enable for the other.
+ */
+ if (al_mc->irq_ue <= 0 || al_mc->irq_ce <= 0) {
+ edac_op_state = EDAC_OPSTATE_POLL;
+ mci->edac_check = al_mc_edac_check;
+ } else {
+ edac_op_state = EDAC_OPSTATE_INT;
+ }
+
+ spin_lock_init(&al_mc->lock);
+
+ mci->mtype_cap = MEM_FLAG_DDR3 | MEM_FLAG_DDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = DRV_NAME;
+ mci->ctl_name = "al_mc";
+ mci->pdev = &pdev->dev;
+ mci->scrub_mode = get_scrub_mode(mmio_base);
+
+ dimm = *mci->dimms;
+ dimm->grain = 1;
+
+ ret = edac_mc_add_mc(mci);
+ if (ret < 0) {
+ dev_err(&pdev->dev,
+ "fail to add memory controller device (%d)\n",
+ ret);
+ return ret;
+ }
+
+ ret = devm_add_action_or_reset(&pdev->dev, devm_al_mc_edac_del, &pdev->dev);
+ if (ret)
+ return ret;
+
+ if (al_mc->irq_ue > 0) {
+ ret = devm_request_irq(&pdev->dev,
+ al_mc->irq_ue,
+ al_mc_edac_irq_handler_ue,
+ IRQF_SHARED,
+ pdev->name,
+ pdev);
+ if (ret != 0) {
+ dev_err(&pdev->dev,
+ "failed to request UE IRQ %d (%d)\n",
+ al_mc->irq_ue, ret);
+ return ret;
+ }
+ }
+
+ if (al_mc->irq_ce > 0) {
+ ret = devm_request_irq(&pdev->dev,
+ al_mc->irq_ce,
+ al_mc_edac_irq_handler_ce,
+ IRQF_SHARED,
+ pdev->name,
+ pdev);
+ if (ret != 0) {
+ dev_err(&pdev->dev,
+ "failed to request CE IRQ %d (%d)\n",
+ al_mc->irq_ce, ret);
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+static const struct of_device_id al_mc_edac_of_match[] = {
+ { .compatible = "amazon,al-mc-edac", },
+ {},
+};
+
+MODULE_DEVICE_TABLE(of, al_mc_edac_of_match);
+
+static struct platform_driver al_mc_edac_driver = {
+ .probe = al_mc_edac_probe,
+ .driver = {
+ .name = DRV_NAME,
+ .of_match_table = al_mc_edac_of_match,
+ },
+};
+
+module_platform_driver(al_mc_edac_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Talel Shenhar");
+MODULE_DESCRIPTION("Amazon's Annapurna Lab's Memory Controller EDAC Driver");
diff --git a/drivers/edac/altera_edac.c b/drivers/edac/altera_edac.c
new file mode 100644
index 0000000000..8b31cd54bd
--- /dev/null
+++ b/drivers/edac/altera_edac.c
@@ -0,0 +1,2227 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2017-2018, Intel Corporation. All rights reserved
+ * Copyright Altera Corporation (C) 2014-2016. All rights reserved.
+ * Copyright 2011-2012 Calxeda, Inc.
+ */
+
+#include <asm/cacheflush.h>
+#include <linux/ctype.h>
+#include <linux/delay.h>
+#include <linux/edac.h>
+#include <linux/firmware/intel/stratix10-smc.h>
+#include <linux/genalloc.h>
+#include <linux/interrupt.h>
+#include <linux/irqchip/chained_irq.h>
+#include <linux/kernel.h>
+#include <linux/mfd/altera-sysmgr.h>
+#include <linux/mfd/syscon.h>
+#include <linux/notifier.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/of_platform.h>
+#include <linux/panic_notifier.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+#include <linux/types.h>
+#include <linux/uaccess.h>
+
+#include "altera_edac.h"
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "altera_edac"
+#define EDAC_DEVICE "Altera"
+
+#ifdef CONFIG_EDAC_ALTERA_SDRAM
+static const struct altr_sdram_prv_data c5_data = {
+ .ecc_ctrl_offset = CV_CTLCFG_OFST,
+ .ecc_ctl_en_mask = CV_CTLCFG_ECC_AUTO_EN,
+ .ecc_stat_offset = CV_DRAMSTS_OFST,
+ .ecc_stat_ce_mask = CV_DRAMSTS_SBEERR,
+ .ecc_stat_ue_mask = CV_DRAMSTS_DBEERR,
+ .ecc_saddr_offset = CV_ERRADDR_OFST,
+ .ecc_daddr_offset = CV_ERRADDR_OFST,
+ .ecc_cecnt_offset = CV_SBECOUNT_OFST,
+ .ecc_uecnt_offset = CV_DBECOUNT_OFST,
+ .ecc_irq_en_offset = CV_DRAMINTR_OFST,
+ .ecc_irq_en_mask = CV_DRAMINTR_INTREN,
+ .ecc_irq_clr_offset = CV_DRAMINTR_OFST,
+ .ecc_irq_clr_mask = (CV_DRAMINTR_INTRCLR | CV_DRAMINTR_INTREN),
+ .ecc_cnt_rst_offset = CV_DRAMINTR_OFST,
+ .ecc_cnt_rst_mask = CV_DRAMINTR_INTRCLR,
+ .ce_ue_trgr_offset = CV_CTLCFG_OFST,
+ .ce_set_mask = CV_CTLCFG_GEN_SB_ERR,
+ .ue_set_mask = CV_CTLCFG_GEN_DB_ERR,
+};
+
+static const struct altr_sdram_prv_data a10_data = {
+ .ecc_ctrl_offset = A10_ECCCTRL1_OFST,
+ .ecc_ctl_en_mask = A10_ECCCTRL1_ECC_EN,
+ .ecc_stat_offset = A10_INTSTAT_OFST,
+ .ecc_stat_ce_mask = A10_INTSTAT_SBEERR,
+ .ecc_stat_ue_mask = A10_INTSTAT_DBEERR,
+ .ecc_saddr_offset = A10_SERRADDR_OFST,
+ .ecc_daddr_offset = A10_DERRADDR_OFST,
+ .ecc_irq_en_offset = A10_ERRINTEN_OFST,
+ .ecc_irq_en_mask = A10_ECC_IRQ_EN_MASK,
+ .ecc_irq_clr_offset = A10_INTSTAT_OFST,
+ .ecc_irq_clr_mask = (A10_INTSTAT_SBEERR | A10_INTSTAT_DBEERR),
+ .ecc_cnt_rst_offset = A10_ECCCTRL1_OFST,
+ .ecc_cnt_rst_mask = A10_ECC_CNT_RESET_MASK,
+ .ce_ue_trgr_offset = A10_DIAGINTTEST_OFST,
+ .ce_set_mask = A10_DIAGINT_TSERRA_MASK,
+ .ue_set_mask = A10_DIAGINT_TDERRA_MASK,
+};
+
+/*********************** EDAC Memory Controller Functions ****************/
+
+/* The SDRAM controller uses the EDAC Memory Controller framework. */
+
+static irqreturn_t altr_sdram_mc_err_handler(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct altr_sdram_mc_data *drvdata = mci->pvt_info;
+ const struct altr_sdram_prv_data *priv = drvdata->data;
+ u32 status, err_count = 1, err_addr;
+
+ regmap_read(drvdata->mc_vbase, priv->ecc_stat_offset, &status);
+
+ if (status & priv->ecc_stat_ue_mask) {
+ regmap_read(drvdata->mc_vbase, priv->ecc_daddr_offset,
+ &err_addr);
+ if (priv->ecc_uecnt_offset)
+ regmap_read(drvdata->mc_vbase, priv->ecc_uecnt_offset,
+ &err_count);
+ panic("\nEDAC: [%d Uncorrectable errors @ 0x%08X]\n",
+ err_count, err_addr);
+ }
+ if (status & priv->ecc_stat_ce_mask) {
+ regmap_read(drvdata->mc_vbase, priv->ecc_saddr_offset,
+ &err_addr);
+ if (priv->ecc_uecnt_offset)
+ regmap_read(drvdata->mc_vbase, priv->ecc_cecnt_offset,
+ &err_count);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, err_count,
+ err_addr >> PAGE_SHIFT,
+ err_addr & ~PAGE_MASK, 0,
+ 0, 0, -1, mci->ctl_name, "");
+ /* Clear IRQ to resume */
+ regmap_write(drvdata->mc_vbase, priv->ecc_irq_clr_offset,
+ priv->ecc_irq_clr_mask);
+
+ return IRQ_HANDLED;
+ }
+ return IRQ_NONE;
+}
+
+static ssize_t altr_sdr_mc_err_inject_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct mem_ctl_info *mci = file->private_data;
+ struct altr_sdram_mc_data *drvdata = mci->pvt_info;
+ const struct altr_sdram_prv_data *priv = drvdata->data;
+ u32 *ptemp;
+ dma_addr_t dma_handle;
+ u32 reg, read_reg;
+
+ ptemp = dma_alloc_coherent(mci->pdev, 16, &dma_handle, GFP_KERNEL);
+ if (!ptemp) {
+ dma_free_coherent(mci->pdev, 16, ptemp, dma_handle);
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Inject: Buffer Allocation error\n");
+ return -ENOMEM;
+ }
+
+ regmap_read(drvdata->mc_vbase, priv->ce_ue_trgr_offset,
+ &read_reg);
+ read_reg &= ~(priv->ce_set_mask | priv->ue_set_mask);
+
+ /* Error are injected by writing a word while the SBE or DBE
+ * bit in the CTLCFG register is set. Reading the word will
+ * trigger the SBE or DBE error and the corresponding IRQ.
+ */
+ if (count == 3) {
+ edac_printk(KERN_ALERT, EDAC_MC,
+ "Inject Double bit error\n");
+ local_irq_disable();
+ regmap_write(drvdata->mc_vbase, priv->ce_ue_trgr_offset,
+ (read_reg | priv->ue_set_mask));
+ local_irq_enable();
+ } else {
+ edac_printk(KERN_ALERT, EDAC_MC,
+ "Inject Single bit error\n");
+ local_irq_disable();
+ regmap_write(drvdata->mc_vbase, priv->ce_ue_trgr_offset,
+ (read_reg | priv->ce_set_mask));
+ local_irq_enable();
+ }
+
+ ptemp[0] = 0x5A5A5A5A;
+ ptemp[1] = 0xA5A5A5A5;
+
+ /* Clear the error injection bits */
+ regmap_write(drvdata->mc_vbase, priv->ce_ue_trgr_offset, read_reg);
+ /* Ensure it has been written out */
+ wmb();
+
+ /*
+ * To trigger the error, we need to read the data back
+ * (the data was written with errors above).
+ * The READ_ONCE macros and printk are used to prevent the
+ * the compiler optimizing these reads out.
+ */
+ reg = READ_ONCE(ptemp[0]);
+ read_reg = READ_ONCE(ptemp[1]);
+ /* Force Read */
+ rmb();
+
+ edac_printk(KERN_ALERT, EDAC_MC, "Read Data [0x%X, 0x%X]\n",
+ reg, read_reg);
+
+ dma_free_coherent(mci->pdev, 16, ptemp, dma_handle);
+
+ return count;
+}
+
+static const struct file_operations altr_sdr_mc_debug_inject_fops = {
+ .open = simple_open,
+ .write = altr_sdr_mc_err_inject_write,
+ .llseek = generic_file_llseek,
+};
+
+static void altr_sdr_mc_create_debugfs_nodes(struct mem_ctl_info *mci)
+{
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+ return;
+
+ if (!mci->debugfs)
+ return;
+
+ edac_debugfs_create_file("altr_trigger", S_IWUSR, mci->debugfs, mci,
+ &altr_sdr_mc_debug_inject_fops);
+}
+
+/* Get total memory size from Open Firmware DTB */
+static unsigned long get_total_mem(void)
+{
+ struct device_node *np = NULL;
+ struct resource res;
+ int ret;
+ unsigned long total_mem = 0;
+
+ for_each_node_by_type(np, "memory") {
+ ret = of_address_to_resource(np, 0, &res);
+ if (ret)
+ continue;
+
+ total_mem += resource_size(&res);
+ }
+ edac_dbg(0, "total_mem 0x%lx\n", total_mem);
+ return total_mem;
+}
+
+static const struct of_device_id altr_sdram_ctrl_of_match[] = {
+ { .compatible = "altr,sdram-edac", .data = &c5_data},
+ { .compatible = "altr,sdram-edac-a10", .data = &a10_data},
+ {},
+};
+MODULE_DEVICE_TABLE(of, altr_sdram_ctrl_of_match);
+
+static int a10_init(struct regmap *mc_vbase)
+{
+ if (regmap_update_bits(mc_vbase, A10_INTMODE_OFST,
+ A10_INTMODE_SB_INT, A10_INTMODE_SB_INT)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Error setting SB IRQ mode\n");
+ return -ENODEV;
+ }
+
+ if (regmap_write(mc_vbase, A10_SERRCNTREG_OFST, 1)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Error setting trigger count\n");
+ return -ENODEV;
+ }
+
+ return 0;
+}
+
+static int a10_unmask_irq(struct platform_device *pdev, u32 mask)
+{
+ void __iomem *sm_base;
+ int ret = 0;
+
+ if (!request_mem_region(A10_SYMAN_INTMASK_CLR, sizeof(u32),
+ dev_name(&pdev->dev))) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Unable to request mem region\n");
+ return -EBUSY;
+ }
+
+ sm_base = ioremap(A10_SYMAN_INTMASK_CLR, sizeof(u32));
+ if (!sm_base) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Unable to ioremap device\n");
+
+ ret = -ENOMEM;
+ goto release;
+ }
+
+ iowrite32(mask, sm_base);
+
+ iounmap(sm_base);
+
+release:
+ release_mem_region(A10_SYMAN_INTMASK_CLR, sizeof(u32));
+
+ return ret;
+}
+
+static int altr_sdram_probe(struct platform_device *pdev)
+{
+ const struct of_device_id *id;
+ struct edac_mc_layer layers[2];
+ struct mem_ctl_info *mci;
+ struct altr_sdram_mc_data *drvdata;
+ const struct altr_sdram_prv_data *priv;
+ struct regmap *mc_vbase;
+ struct dimm_info *dimm;
+ u32 read_reg;
+ int irq, irq2, res = 0;
+ unsigned long mem_size, irqflags = 0;
+
+ id = of_match_device(altr_sdram_ctrl_of_match, &pdev->dev);
+ if (!id)
+ return -ENODEV;
+
+ /* Grab the register range from the sdr controller in device tree */
+ mc_vbase = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "altr,sdr-syscon");
+ if (IS_ERR(mc_vbase)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "regmap for altr,sdr-syscon lookup failed.\n");
+ return -ENODEV;
+ }
+
+ /* Check specific dependencies for the module */
+ priv = of_match_node(altr_sdram_ctrl_of_match,
+ pdev->dev.of_node)->data;
+
+ /* Validate the SDRAM controller has ECC enabled */
+ if (regmap_read(mc_vbase, priv->ecc_ctrl_offset, &read_reg) ||
+ ((read_reg & priv->ecc_ctl_en_mask) != priv->ecc_ctl_en_mask)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "No ECC/ECC disabled [0x%08X]\n", read_reg);
+ return -ENODEV;
+ }
+
+ /* Grab memory size from device tree. */
+ mem_size = get_total_mem();
+ if (!mem_size) {
+ edac_printk(KERN_ERR, EDAC_MC, "Unable to calculate memory size\n");
+ return -ENODEV;
+ }
+
+ /* Ensure the SDRAM Interrupt is disabled */
+ if (regmap_update_bits(mc_vbase, priv->ecc_irq_en_offset,
+ priv->ecc_irq_en_mask, 0)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Error disabling SDRAM ECC IRQ\n");
+ return -ENODEV;
+ }
+
+ /* Toggle to clear the SDRAM Error count */
+ if (regmap_update_bits(mc_vbase, priv->ecc_cnt_rst_offset,
+ priv->ecc_cnt_rst_mask,
+ priv->ecc_cnt_rst_mask)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Error clearing SDRAM ECC count\n");
+ return -ENODEV;
+ }
+
+ if (regmap_update_bits(mc_vbase, priv->ecc_cnt_rst_offset,
+ priv->ecc_cnt_rst_mask, 0)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Error clearing SDRAM ECC count\n");
+ return -ENODEV;
+ }
+
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "No irq %d in DT\n", irq);
+ return irq;
+ }
+
+ /* Arria10 has a 2nd IRQ */
+ irq2 = platform_get_irq(pdev, 1);
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct altr_sdram_mc_data));
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ drvdata = mci->pvt_info;
+ drvdata->mc_vbase = mc_vbase;
+ drvdata->data = priv;
+ platform_set_drvdata(pdev, mci);
+
+ if (!devres_open_group(&pdev->dev, NULL, GFP_KERNEL)) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Unable to get managed device resource\n");
+ res = -ENOMEM;
+ goto free;
+ }
+
+ mci->mtype_cap = MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = dev_name(&pdev->dev);
+ mci->scrub_mode = SCRUB_SW_SRC;
+ mci->dev_name = dev_name(&pdev->dev);
+
+ dimm = *mci->dimms;
+ dimm->nr_pages = ((mem_size - 1) >> PAGE_SHIFT) + 1;
+ dimm->grain = 8;
+ dimm->dtype = DEV_X8;
+ dimm->mtype = MEM_DDR3;
+ dimm->edac_mode = EDAC_SECDED;
+
+ res = edac_mc_add_mc(mci);
+ if (res < 0)
+ goto err;
+
+ /* Only the Arria10 has separate IRQs */
+ if (of_machine_is_compatible("altr,socfpga-arria10")) {
+ /* Arria10 specific initialization */
+ res = a10_init(mc_vbase);
+ if (res < 0)
+ goto err2;
+
+ res = devm_request_irq(&pdev->dev, irq2,
+ altr_sdram_mc_err_handler,
+ IRQF_SHARED, dev_name(&pdev->dev), mci);
+ if (res < 0) {
+ edac_mc_printk(mci, KERN_ERR,
+ "Unable to request irq %d\n", irq2);
+ res = -ENODEV;
+ goto err2;
+ }
+
+ res = a10_unmask_irq(pdev, A10_DDR0_IRQ_MASK);
+ if (res < 0)
+ goto err2;
+
+ irqflags = IRQF_SHARED;
+ }
+
+ res = devm_request_irq(&pdev->dev, irq, altr_sdram_mc_err_handler,
+ irqflags, dev_name(&pdev->dev), mci);
+ if (res < 0) {
+ edac_mc_printk(mci, KERN_ERR,
+ "Unable to request irq %d\n", irq);
+ res = -ENODEV;
+ goto err2;
+ }
+
+ /* Infrastructure ready - enable the IRQ */
+ if (regmap_update_bits(drvdata->mc_vbase, priv->ecc_irq_en_offset,
+ priv->ecc_irq_en_mask, priv->ecc_irq_en_mask)) {
+ edac_mc_printk(mci, KERN_ERR,
+ "Error enabling SDRAM ECC IRQ\n");
+ res = -ENODEV;
+ goto err2;
+ }
+
+ altr_sdr_mc_create_debugfs_nodes(mci);
+
+ devres_close_group(&pdev->dev, NULL);
+
+ return 0;
+
+err2:
+ edac_mc_del_mc(&pdev->dev);
+err:
+ devres_release_group(&pdev->dev, NULL);
+free:
+ edac_mc_free(mci);
+ edac_printk(KERN_ERR, EDAC_MC,
+ "EDAC Probe Failed; Error %d\n", res);
+
+ return res;
+}
+
+static int altr_sdram_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+ platform_set_drvdata(pdev, NULL);
+
+ return 0;
+}
+
+/*
+ * If you want to suspend, need to disable EDAC by removing it
+ * from the device tree or defconfig.
+ */
+#ifdef CONFIG_PM
+static int altr_sdram_prepare(struct device *dev)
+{
+ pr_err("Suspend not allowed when EDAC is enabled.\n");
+
+ return -EPERM;
+}
+
+static const struct dev_pm_ops altr_sdram_pm_ops = {
+ .prepare = altr_sdram_prepare,
+};
+#endif
+
+static struct platform_driver altr_sdram_edac_driver = {
+ .probe = altr_sdram_probe,
+ .remove = altr_sdram_remove,
+ .driver = {
+ .name = "altr_sdram_edac",
+#ifdef CONFIG_PM
+ .pm = &altr_sdram_pm_ops,
+#endif
+ .of_match_table = altr_sdram_ctrl_of_match,
+ },
+};
+
+module_platform_driver(altr_sdram_edac_driver);
+
+#endif /* CONFIG_EDAC_ALTERA_SDRAM */
+
+/************************* EDAC Parent Probe *************************/
+
+static const struct of_device_id altr_edac_device_of_match[];
+
+static const struct of_device_id altr_edac_of_match[] = {
+ { .compatible = "altr,socfpga-ecc-manager" },
+ {},
+};
+MODULE_DEVICE_TABLE(of, altr_edac_of_match);
+
+static int altr_edac_probe(struct platform_device *pdev)
+{
+ of_platform_populate(pdev->dev.of_node, altr_edac_device_of_match,
+ NULL, &pdev->dev);
+ return 0;
+}
+
+static struct platform_driver altr_edac_driver = {
+ .probe = altr_edac_probe,
+ .driver = {
+ .name = "socfpga_ecc_manager",
+ .of_match_table = altr_edac_of_match,
+ },
+};
+module_platform_driver(altr_edac_driver);
+
+/************************* EDAC Device Functions *************************/
+
+/*
+ * EDAC Device Functions (shared between various IPs).
+ * The discrete memories use the EDAC Device framework. The probe
+ * and error handling functions are very similar between memories
+ * so they are shared. The memory allocation and freeing for EDAC
+ * trigger testing are different for each memory.
+ */
+
+#ifdef CONFIG_EDAC_ALTERA_OCRAM
+static const struct edac_device_prv_data ocramecc_data;
+#endif
+#ifdef CONFIG_EDAC_ALTERA_L2C
+static const struct edac_device_prv_data l2ecc_data;
+#endif
+#ifdef CONFIG_EDAC_ALTERA_OCRAM
+static const struct edac_device_prv_data a10_ocramecc_data;
+#endif
+#ifdef CONFIG_EDAC_ALTERA_L2C
+static const struct edac_device_prv_data a10_l2ecc_data;
+#endif
+
+static irqreturn_t altr_edac_device_handler(int irq, void *dev_id)
+{
+ irqreturn_t ret_value = IRQ_NONE;
+ struct edac_device_ctl_info *dci = dev_id;
+ struct altr_edac_device_dev *drvdata = dci->pvt_info;
+ const struct edac_device_prv_data *priv = drvdata->data;
+
+ if (irq == drvdata->sb_irq) {
+ if (priv->ce_clear_mask)
+ writel(priv->ce_clear_mask, drvdata->base);
+ edac_device_handle_ce(dci, 0, 0, drvdata->edac_dev_name);
+ ret_value = IRQ_HANDLED;
+ } else if (irq == drvdata->db_irq) {
+ if (priv->ue_clear_mask)
+ writel(priv->ue_clear_mask, drvdata->base);
+ edac_device_handle_ue(dci, 0, 0, drvdata->edac_dev_name);
+ panic("\nEDAC:ECC_DEVICE[Uncorrectable errors]\n");
+ ret_value = IRQ_HANDLED;
+ } else {
+ WARN_ON(1);
+ }
+
+ return ret_value;
+}
+
+static ssize_t __maybe_unused
+altr_edac_device_trig(struct file *file, const char __user *user_buf,
+ size_t count, loff_t *ppos)
+
+{
+ u32 *ptemp, i, error_mask;
+ int result = 0;
+ u8 trig_type;
+ unsigned long flags;
+ struct edac_device_ctl_info *edac_dci = file->private_data;
+ struct altr_edac_device_dev *drvdata = edac_dci->pvt_info;
+ const struct edac_device_prv_data *priv = drvdata->data;
+ void *generic_ptr = edac_dci->dev;
+
+ if (!user_buf || get_user(trig_type, user_buf))
+ return -EFAULT;
+
+ if (!priv->alloc_mem)
+ return -ENOMEM;
+
+ /*
+ * Note that generic_ptr is initialized to the device * but in
+ * some alloc_functions, this is overridden and returns data.
+ */
+ ptemp = priv->alloc_mem(priv->trig_alloc_sz, &generic_ptr);
+ if (!ptemp) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Inject: Buffer Allocation error\n");
+ return -ENOMEM;
+ }
+
+ if (trig_type == ALTR_UE_TRIGGER_CHAR)
+ error_mask = priv->ue_set_mask;
+ else
+ error_mask = priv->ce_set_mask;
+
+ edac_printk(KERN_ALERT, EDAC_DEVICE,
+ "Trigger Error Mask (0x%X)\n", error_mask);
+
+ local_irq_save(flags);
+ /* write ECC corrupted data out. */
+ for (i = 0; i < (priv->trig_alloc_sz / sizeof(*ptemp)); i++) {
+ /* Read data so we're in the correct state */
+ rmb();
+ if (READ_ONCE(ptemp[i]))
+ result = -1;
+ /* Toggle Error bit (it is latched), leave ECC enabled */
+ writel(error_mask, (drvdata->base + priv->set_err_ofst));
+ writel(priv->ecc_enable_mask, (drvdata->base +
+ priv->set_err_ofst));
+ ptemp[i] = i;
+ }
+ /* Ensure it has been written out */
+ wmb();
+ local_irq_restore(flags);
+
+ if (result)
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Mem Not Cleared\n");
+
+ /* Read out written data. ECC error caused here */
+ for (i = 0; i < ALTR_TRIGGER_READ_WRD_CNT; i++)
+ if (READ_ONCE(ptemp[i]) != i)
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Read doesn't match written data\n");
+
+ if (priv->free_mem)
+ priv->free_mem(ptemp, priv->trig_alloc_sz, generic_ptr);
+
+ return count;
+}
+
+static const struct file_operations altr_edac_device_inject_fops __maybe_unused = {
+ .open = simple_open,
+ .write = altr_edac_device_trig,
+ .llseek = generic_file_llseek,
+};
+
+static ssize_t __maybe_unused
+altr_edac_a10_device_trig(struct file *file, const char __user *user_buf,
+ size_t count, loff_t *ppos);
+
+static const struct file_operations altr_edac_a10_device_inject_fops __maybe_unused = {
+ .open = simple_open,
+ .write = altr_edac_a10_device_trig,
+ .llseek = generic_file_llseek,
+};
+
+static ssize_t __maybe_unused
+altr_edac_a10_device_trig2(struct file *file, const char __user *user_buf,
+ size_t count, loff_t *ppos);
+
+static const struct file_operations altr_edac_a10_device_inject2_fops __maybe_unused = {
+ .open = simple_open,
+ .write = altr_edac_a10_device_trig2,
+ .llseek = generic_file_llseek,
+};
+
+static void altr_create_edacdev_dbgfs(struct edac_device_ctl_info *edac_dci,
+ const struct edac_device_prv_data *priv)
+{
+ struct altr_edac_device_dev *drvdata = edac_dci->pvt_info;
+
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+ return;
+
+ drvdata->debugfs_dir = edac_debugfs_create_dir(drvdata->edac_dev_name);
+ if (!drvdata->debugfs_dir)
+ return;
+
+ if (!edac_debugfs_create_file("altr_trigger", S_IWUSR,
+ drvdata->debugfs_dir, edac_dci,
+ priv->inject_fops))
+ debugfs_remove_recursive(drvdata->debugfs_dir);
+}
+
+static const struct of_device_id altr_edac_device_of_match[] = {
+#ifdef CONFIG_EDAC_ALTERA_L2C
+ { .compatible = "altr,socfpga-l2-ecc", .data = &l2ecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_OCRAM
+ { .compatible = "altr,socfpga-ocram-ecc", .data = &ocramecc_data },
+#endif
+ {},
+};
+MODULE_DEVICE_TABLE(of, altr_edac_device_of_match);
+
+/*
+ * altr_edac_device_probe()
+ * This is a generic EDAC device driver that will support
+ * various Altera memory devices such as the L2 cache ECC and
+ * OCRAM ECC as well as the memories for other peripherals.
+ * Module specific initialization is done by passing the
+ * function index in the device tree.
+ */
+static int altr_edac_device_probe(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci;
+ struct altr_edac_device_dev *drvdata;
+ struct resource *r;
+ int res = 0;
+ struct device_node *np = pdev->dev.of_node;
+ char *ecc_name = (char *)np->name;
+ static int dev_instance;
+
+ if (!devres_open_group(&pdev->dev, NULL, GFP_KERNEL)) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to open devm\n");
+ return -ENOMEM;
+ }
+
+ r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!r) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to get mem resource\n");
+ res = -ENODEV;
+ goto fail;
+ }
+
+ if (!devm_request_mem_region(&pdev->dev, r->start, resource_size(r),
+ dev_name(&pdev->dev))) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s:Error requesting mem region\n", ecc_name);
+ res = -EBUSY;
+ goto fail;
+ }
+
+ dci = edac_device_alloc_ctl_info(sizeof(*drvdata), ecc_name,
+ 1, ecc_name, 1, 0, NULL, 0,
+ dev_instance++);
+
+ if (!dci) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: Unable to allocate EDAC device\n", ecc_name);
+ res = -ENOMEM;
+ goto fail;
+ }
+
+ drvdata = dci->pvt_info;
+ dci->dev = &pdev->dev;
+ platform_set_drvdata(pdev, dci);
+ drvdata->edac_dev_name = ecc_name;
+
+ drvdata->base = devm_ioremap(&pdev->dev, r->start, resource_size(r));
+ if (!drvdata->base) {
+ res = -ENOMEM;
+ goto fail1;
+ }
+
+ /* Get driver specific data for this EDAC device */
+ drvdata->data = of_match_node(altr_edac_device_of_match, np)->data;
+
+ /* Check specific dependencies for the module */
+ if (drvdata->data->setup) {
+ res = drvdata->data->setup(drvdata);
+ if (res)
+ goto fail1;
+ }
+
+ drvdata->sb_irq = platform_get_irq(pdev, 0);
+ res = devm_request_irq(&pdev->dev, drvdata->sb_irq,
+ altr_edac_device_handler,
+ 0, dev_name(&pdev->dev), dci);
+ if (res)
+ goto fail1;
+
+ drvdata->db_irq = platform_get_irq(pdev, 1);
+ res = devm_request_irq(&pdev->dev, drvdata->db_irq,
+ altr_edac_device_handler,
+ 0, dev_name(&pdev->dev), dci);
+ if (res)
+ goto fail1;
+
+ dci->mod_name = "Altera ECC Manager";
+ dci->dev_name = drvdata->edac_dev_name;
+
+ res = edac_device_add_device(dci);
+ if (res)
+ goto fail1;
+
+ altr_create_edacdev_dbgfs(dci, drvdata->data);
+
+ devres_close_group(&pdev->dev, NULL);
+
+ return 0;
+
+fail1:
+ edac_device_free_ctl_info(dci);
+fail:
+ devres_release_group(&pdev->dev, NULL);
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s:Error setting up EDAC device: %d\n", ecc_name, res);
+
+ return res;
+}
+
+static int altr_edac_device_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
+ struct altr_edac_device_dev *drvdata = dci->pvt_info;
+
+ debugfs_remove_recursive(drvdata->debugfs_dir);
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);
+
+ return 0;
+}
+
+static struct platform_driver altr_edac_device_driver = {
+ .probe = altr_edac_device_probe,
+ .remove = altr_edac_device_remove,
+ .driver = {
+ .name = "altr_edac_device",
+ .of_match_table = altr_edac_device_of_match,
+ },
+};
+module_platform_driver(altr_edac_device_driver);
+
+/******************* Arria10 Device ECC Shared Functions *****************/
+
+/*
+ * Test for memory's ECC dependencies upon entry because platform specific
+ * startup should have initialized the memory and enabled the ECC.
+ * Can't turn on ECC here because accessing un-initialized memory will
+ * cause CE/UE errors possibly causing an ABORT.
+ */
+static int __maybe_unused
+altr_check_ecc_deps(struct altr_edac_device_dev *device)
+{
+ void __iomem *base = device->base;
+ const struct edac_device_prv_data *prv = device->data;
+
+ if (readl(base + prv->ecc_en_ofst) & prv->ecc_enable_mask)
+ return 0;
+
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: No ECC present or ECC disabled.\n",
+ device->edac_dev_name);
+ return -ENODEV;
+}
+
+static irqreturn_t __maybe_unused altr_edac_a10_ecc_irq(int irq, void *dev_id)
+{
+ struct altr_edac_device_dev *dci = dev_id;
+ void __iomem *base = dci->base;
+
+ if (irq == dci->sb_irq) {
+ writel(ALTR_A10_ECC_SERRPENA,
+ base + ALTR_A10_ECC_INTSTAT_OFST);
+ edac_device_handle_ce(dci->edac_dev, 0, 0, dci->edac_dev_name);
+
+ return IRQ_HANDLED;
+ } else if (irq == dci->db_irq) {
+ writel(ALTR_A10_ECC_DERRPENA,
+ base + ALTR_A10_ECC_INTSTAT_OFST);
+ edac_device_handle_ue(dci->edac_dev, 0, 0, dci->edac_dev_name);
+ if (dci->data->panic)
+ panic("\nEDAC:ECC_DEVICE[Uncorrectable errors]\n");
+
+ return IRQ_HANDLED;
+ }
+
+ WARN_ON(1);
+
+ return IRQ_NONE;
+}
+
+/******************* Arria10 Memory Buffer Functions *********************/
+
+static inline int a10_get_irq_mask(struct device_node *np)
+{
+ int irq;
+ const u32 *handle = of_get_property(np, "interrupts", NULL);
+
+ if (!handle)
+ return -ENODEV;
+ irq = be32_to_cpup(handle);
+ return irq;
+}
+
+static inline void ecc_set_bits(u32 bit_mask, void __iomem *ioaddr)
+{
+ u32 value = readl(ioaddr);
+
+ value |= bit_mask;
+ writel(value, ioaddr);
+}
+
+static inline void ecc_clear_bits(u32 bit_mask, void __iomem *ioaddr)
+{
+ u32 value = readl(ioaddr);
+
+ value &= ~bit_mask;
+ writel(value, ioaddr);
+}
+
+static inline int ecc_test_bits(u32 bit_mask, void __iomem *ioaddr)
+{
+ u32 value = readl(ioaddr);
+
+ return (value & bit_mask) ? 1 : 0;
+}
+
+/*
+ * This function uses the memory initialization block in the Arria10 ECC
+ * controller to initialize/clear the entire memory data and ECC data.
+ */
+static int __maybe_unused altr_init_memory_port(void __iomem *ioaddr, int port)
+{
+ int limit = ALTR_A10_ECC_INIT_WATCHDOG_10US;
+ u32 init_mask, stat_mask, clear_mask;
+ int ret = 0;
+
+ if (port) {
+ init_mask = ALTR_A10_ECC_INITB;
+ stat_mask = ALTR_A10_ECC_INITCOMPLETEB;
+ clear_mask = ALTR_A10_ECC_ERRPENB_MASK;
+ } else {
+ init_mask = ALTR_A10_ECC_INITA;
+ stat_mask = ALTR_A10_ECC_INITCOMPLETEA;
+ clear_mask = ALTR_A10_ECC_ERRPENA_MASK;
+ }
+
+ ecc_set_bits(init_mask, (ioaddr + ALTR_A10_ECC_CTRL_OFST));
+ while (limit--) {
+ if (ecc_test_bits(stat_mask,
+ (ioaddr + ALTR_A10_ECC_INITSTAT_OFST)))
+ break;
+ udelay(1);
+ }
+ if (limit < 0)
+ ret = -EBUSY;
+
+ /* Clear any pending ECC interrupts */
+ writel(clear_mask, (ioaddr + ALTR_A10_ECC_INTSTAT_OFST));
+
+ return ret;
+}
+
+static __init int __maybe_unused
+altr_init_a10_ecc_block(struct device_node *np, u32 irq_mask,
+ u32 ecc_ctrl_en_mask, bool dual_port)
+{
+ int ret = 0;
+ void __iomem *ecc_block_base;
+ struct regmap *ecc_mgr_map;
+ char *ecc_name;
+ struct device_node *np_eccmgr;
+
+ ecc_name = (char *)np->name;
+
+ /* Get the ECC Manager - parent of the device EDACs */
+ np_eccmgr = of_get_parent(np);
+
+ ecc_mgr_map =
+ altr_sysmgr_regmap_lookup_by_phandle(np_eccmgr,
+ "altr,sysmgr-syscon");
+
+ of_node_put(np_eccmgr);
+ if (IS_ERR(ecc_mgr_map)) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to get syscon altr,sysmgr-syscon\n");
+ return -ENODEV;
+ }
+
+ /* Map the ECC Block */
+ ecc_block_base = of_iomap(np, 0);
+ if (!ecc_block_base) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to map %s ECC block\n", ecc_name);
+ return -ENODEV;
+ }
+
+ /* Disable ECC */
+ regmap_write(ecc_mgr_map, A10_SYSMGR_ECC_INTMASK_SET_OFST, irq_mask);
+ writel(ALTR_A10_ECC_SERRINTEN,
+ (ecc_block_base + ALTR_A10_ECC_ERRINTENR_OFST));
+ ecc_clear_bits(ecc_ctrl_en_mask,
+ (ecc_block_base + ALTR_A10_ECC_CTRL_OFST));
+ /* Ensure all writes complete */
+ wmb();
+ /* Use HW initialization block to initialize memory for ECC */
+ ret = altr_init_memory_port(ecc_block_base, 0);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "ECC: cannot init %s PORTA memory\n", ecc_name);
+ goto out;
+ }
+
+ if (dual_port) {
+ ret = altr_init_memory_port(ecc_block_base, 1);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "ECC: cannot init %s PORTB memory\n",
+ ecc_name);
+ goto out;
+ }
+ }
+
+ /* Interrupt mode set to every SBERR */
+ regmap_write(ecc_mgr_map, ALTR_A10_ECC_INTMODE_OFST,
+ ALTR_A10_ECC_INTMODE);
+ /* Enable ECC */
+ ecc_set_bits(ecc_ctrl_en_mask, (ecc_block_base +
+ ALTR_A10_ECC_CTRL_OFST));
+ writel(ALTR_A10_ECC_SERRINTEN,
+ (ecc_block_base + ALTR_A10_ECC_ERRINTENS_OFST));
+ regmap_write(ecc_mgr_map, A10_SYSMGR_ECC_INTMASK_CLR_OFST, irq_mask);
+ /* Ensure all writes complete */
+ wmb();
+out:
+ iounmap(ecc_block_base);
+ return ret;
+}
+
+static int validate_parent_available(struct device_node *np);
+static const struct of_device_id altr_edac_a10_device_of_match[];
+static int __init __maybe_unused altr_init_a10_ecc_device_type(char *compat)
+{
+ int irq;
+ struct device_node *child, *np;
+
+ np = of_find_compatible_node(NULL, NULL,
+ "altr,socfpga-a10-ecc-manager");
+ if (!np) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "ECC Manager not found\n");
+ return -ENODEV;
+ }
+
+ for_each_child_of_node(np, child) {
+ const struct of_device_id *pdev_id;
+ const struct edac_device_prv_data *prv;
+
+ if (!of_device_is_available(child))
+ continue;
+ if (!of_device_is_compatible(child, compat))
+ continue;
+
+ if (validate_parent_available(child))
+ continue;
+
+ irq = a10_get_irq_mask(child);
+ if (irq < 0)
+ continue;
+
+ /* Get matching node and check for valid result */
+ pdev_id = of_match_node(altr_edac_a10_device_of_match, child);
+ if (IS_ERR_OR_NULL(pdev_id))
+ continue;
+
+ /* Validate private data pointer before dereferencing */
+ prv = pdev_id->data;
+ if (!prv)
+ continue;
+
+ altr_init_a10_ecc_block(child, BIT(irq),
+ prv->ecc_enable_mask, 0);
+ }
+
+ of_node_put(np);
+ return 0;
+}
+
+/*********************** SDRAM EDAC Device Functions *********************/
+
+#ifdef CONFIG_EDAC_ALTERA_SDRAM
+
+/*
+ * A legacy U-Boot bug only enabled memory mapped access to the ECC Enable
+ * register if ECC is enabled. Linux checks the ECC Enable register to
+ * determine ECC status.
+ * Use an SMC call (which always works) to determine ECC enablement.
+ */
+static int altr_s10_sdram_check_ecc_deps(struct altr_edac_device_dev *device)
+{
+ const struct edac_device_prv_data *prv = device->data;
+ unsigned long sdram_ecc_addr;
+ struct arm_smccc_res result;
+ struct device_node *np;
+ phys_addr_t sdram_addr;
+ u32 read_reg;
+ int ret;
+
+ np = of_find_compatible_node(NULL, NULL, "altr,sdr-ctl");
+ if (!np)
+ goto sdram_err;
+
+ sdram_addr = of_translate_address(np, of_get_address(np, 0,
+ NULL, NULL));
+ of_node_put(np);
+ sdram_ecc_addr = (unsigned long)sdram_addr + prv->ecc_en_ofst;
+ arm_smccc_smc(INTEL_SIP_SMC_REG_READ, sdram_ecc_addr,
+ 0, 0, 0, 0, 0, 0, &result);
+ read_reg = (unsigned int)result.a1;
+ ret = (int)result.a0;
+ if (!ret && (read_reg & prv->ecc_enable_mask))
+ return 0;
+
+sdram_err:
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: No ECC present or ECC disabled.\n",
+ device->edac_dev_name);
+ return -ENODEV;
+}
+
+static const struct edac_device_prv_data s10_sdramecc_data = {
+ .setup = altr_s10_sdram_check_ecc_deps,
+ .ce_clear_mask = ALTR_S10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_S10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_S10_ECC_EN,
+ .ecc_en_ofst = ALTR_S10_ECC_CTRL_SDRAM_OFST,
+ .ce_set_mask = ALTR_S10_ECC_TSERRA,
+ .ue_set_mask = ALTR_S10_ECC_TDERRA,
+ .set_err_ofst = ALTR_S10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+#endif /* CONFIG_EDAC_ALTERA_SDRAM */
+
+/*********************** OCRAM EDAC Device Functions *********************/
+
+#ifdef CONFIG_EDAC_ALTERA_OCRAM
+
+static void *ocram_alloc_mem(size_t size, void **other)
+{
+ struct device_node *np;
+ struct gen_pool *gp;
+ void *sram_addr;
+
+ np = of_find_compatible_node(NULL, NULL, "altr,socfpga-ocram-ecc");
+ if (!np)
+ return NULL;
+
+ gp = of_gen_pool_get(np, "iram", 0);
+ of_node_put(np);
+ if (!gp)
+ return NULL;
+
+ sram_addr = (void *)gen_pool_alloc(gp, size);
+ if (!sram_addr)
+ return NULL;
+
+ memset(sram_addr, 0, size);
+ /* Ensure data is written out */
+ wmb();
+
+ /* Remember this handle for freeing later */
+ *other = gp;
+
+ return sram_addr;
+}
+
+static void ocram_free_mem(void *p, size_t size, void *other)
+{
+ gen_pool_free((struct gen_pool *)other, (unsigned long)p, size);
+}
+
+static const struct edac_device_prv_data ocramecc_data = {
+ .setup = altr_check_ecc_deps,
+ .ce_clear_mask = (ALTR_OCR_ECC_EN | ALTR_OCR_ECC_SERR),
+ .ue_clear_mask = (ALTR_OCR_ECC_EN | ALTR_OCR_ECC_DERR),
+ .alloc_mem = ocram_alloc_mem,
+ .free_mem = ocram_free_mem,
+ .ecc_enable_mask = ALTR_OCR_ECC_EN,
+ .ecc_en_ofst = ALTR_OCR_ECC_REG_OFFSET,
+ .ce_set_mask = (ALTR_OCR_ECC_EN | ALTR_OCR_ECC_INJS),
+ .ue_set_mask = (ALTR_OCR_ECC_EN | ALTR_OCR_ECC_INJD),
+ .set_err_ofst = ALTR_OCR_ECC_REG_OFFSET,
+ .trig_alloc_sz = ALTR_TRIG_OCRAM_BYTE_SIZE,
+ .inject_fops = &altr_edac_device_inject_fops,
+};
+
+static int __maybe_unused
+altr_check_ocram_deps_init(struct altr_edac_device_dev *device)
+{
+ void __iomem *base = device->base;
+ int ret;
+
+ ret = altr_check_ecc_deps(device);
+ if (ret)
+ return ret;
+
+ /* Verify OCRAM has been initialized */
+ if (!ecc_test_bits(ALTR_A10_ECC_INITCOMPLETEA,
+ (base + ALTR_A10_ECC_INITSTAT_OFST)))
+ return -ENODEV;
+
+ /* Enable IRQ on Single Bit Error */
+ writel(ALTR_A10_ECC_SERRINTEN, (base + ALTR_A10_ECC_ERRINTENS_OFST));
+ /* Ensure all writes complete */
+ wmb();
+
+ return 0;
+}
+
+static const struct edac_device_prv_data a10_ocramecc_data = {
+ .setup = altr_check_ocram_deps_init,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .irq_status_mask = A10_SYSMGR_ECC_INTSTAT_OCRAM,
+ .ecc_enable_mask = ALTR_A10_OCRAM_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject2_fops,
+ /*
+ * OCRAM panic on uncorrectable error because sleep/resume
+ * functions and FPGA contents are stored in OCRAM. Prefer
+ * a kernel panic over executing/loading corrupted data.
+ */
+ .panic = true,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_OCRAM */
+
+/********************* L2 Cache EDAC Device Functions ********************/
+
+#ifdef CONFIG_EDAC_ALTERA_L2C
+
+static void *l2_alloc_mem(size_t size, void **other)
+{
+ struct device *dev = *other;
+ void *ptemp = devm_kzalloc(dev, size, GFP_KERNEL);
+
+ if (!ptemp)
+ return NULL;
+
+ /* Make sure everything is written out */
+ wmb();
+
+ /*
+ * Clean all cache levels up to LoC (includes L2)
+ * This ensures the corrupted data is written into
+ * L2 cache for readback test (which causes ECC error).
+ */
+ flush_cache_all();
+
+ return ptemp;
+}
+
+static void l2_free_mem(void *p, size_t size, void *other)
+{
+ struct device *dev = other;
+
+ if (dev && p)
+ devm_kfree(dev, p);
+}
+
+/*
+ * altr_l2_check_deps()
+ * Test for L2 cache ECC dependencies upon entry because
+ * platform specific startup should have initialized the L2
+ * memory and enabled the ECC.
+ * Bail if ECC is not enabled.
+ * Note that L2 Cache Enable is forced at build time.
+ */
+static int altr_l2_check_deps(struct altr_edac_device_dev *device)
+{
+ void __iomem *base = device->base;
+ const struct edac_device_prv_data *prv = device->data;
+
+ if ((readl(base) & prv->ecc_enable_mask) ==
+ prv->ecc_enable_mask)
+ return 0;
+
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "L2: No ECC present, or ECC disabled\n");
+ return -ENODEV;
+}
+
+static irqreturn_t altr_edac_a10_l2_irq(int irq, void *dev_id)
+{
+ struct altr_edac_device_dev *dci = dev_id;
+
+ if (irq == dci->sb_irq) {
+ regmap_write(dci->edac->ecc_mgr_map,
+ A10_SYSGMR_MPU_CLEAR_L2_ECC_OFST,
+ A10_SYSGMR_MPU_CLEAR_L2_ECC_SB);
+ edac_device_handle_ce(dci->edac_dev, 0, 0, dci->edac_dev_name);
+
+ return IRQ_HANDLED;
+ } else if (irq == dci->db_irq) {
+ regmap_write(dci->edac->ecc_mgr_map,
+ A10_SYSGMR_MPU_CLEAR_L2_ECC_OFST,
+ A10_SYSGMR_MPU_CLEAR_L2_ECC_MB);
+ edac_device_handle_ue(dci->edac_dev, 0, 0, dci->edac_dev_name);
+ panic("\nEDAC:ECC_DEVICE[Uncorrectable errors]\n");
+
+ return IRQ_HANDLED;
+ }
+
+ WARN_ON(1);
+
+ return IRQ_NONE;
+}
+
+static const struct edac_device_prv_data l2ecc_data = {
+ .setup = altr_l2_check_deps,
+ .ce_clear_mask = 0,
+ .ue_clear_mask = 0,
+ .alloc_mem = l2_alloc_mem,
+ .free_mem = l2_free_mem,
+ .ecc_enable_mask = ALTR_L2_ECC_EN,
+ .ce_set_mask = (ALTR_L2_ECC_EN | ALTR_L2_ECC_INJS),
+ .ue_set_mask = (ALTR_L2_ECC_EN | ALTR_L2_ECC_INJD),
+ .set_err_ofst = ALTR_L2_ECC_REG_OFFSET,
+ .trig_alloc_sz = ALTR_TRIG_L2C_BYTE_SIZE,
+ .inject_fops = &altr_edac_device_inject_fops,
+};
+
+static const struct edac_device_prv_data a10_l2ecc_data = {
+ .setup = altr_l2_check_deps,
+ .ce_clear_mask = ALTR_A10_L2_ECC_SERR_CLR,
+ .ue_clear_mask = ALTR_A10_L2_ECC_MERR_CLR,
+ .irq_status_mask = A10_SYSMGR_ECC_INTSTAT_L2,
+ .alloc_mem = l2_alloc_mem,
+ .free_mem = l2_free_mem,
+ .ecc_enable_mask = ALTR_A10_L2_ECC_EN_CTL,
+ .ce_set_mask = ALTR_A10_L2_ECC_CE_INJ_MASK,
+ .ue_set_mask = ALTR_A10_L2_ECC_UE_INJ_MASK,
+ .set_err_ofst = ALTR_A10_L2_ECC_INJ_OFST,
+ .ecc_irq_handler = altr_edac_a10_l2_irq,
+ .trig_alloc_sz = ALTR_TRIG_L2C_BYTE_SIZE,
+ .inject_fops = &altr_edac_device_inject_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_L2C */
+
+/********************* Ethernet Device Functions ********************/
+
+#ifdef CONFIG_EDAC_ALTERA_ETHERNET
+
+static int __init socfpga_init_ethernet_ecc(struct altr_edac_device_dev *dev)
+{
+ int ret;
+
+ ret = altr_init_a10_ecc_device_type("altr,socfpga-eth-mac-ecc");
+ if (ret)
+ return ret;
+
+ return altr_check_ecc_deps(dev);
+}
+
+static const struct edac_device_prv_data a10_enetecc_data = {
+ .setup = socfpga_init_ethernet_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject2_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_ETHERNET */
+
+/********************** NAND Device Functions **********************/
+
+#ifdef CONFIG_EDAC_ALTERA_NAND
+
+static int __init socfpga_init_nand_ecc(struct altr_edac_device_dev *device)
+{
+ int ret;
+
+ ret = altr_init_a10_ecc_device_type("altr,socfpga-nand-ecc");
+ if (ret)
+ return ret;
+
+ return altr_check_ecc_deps(device);
+}
+
+static const struct edac_device_prv_data a10_nandecc_data = {
+ .setup = socfpga_init_nand_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_NAND */
+
+/********************** DMA Device Functions **********************/
+
+#ifdef CONFIG_EDAC_ALTERA_DMA
+
+static int __init socfpga_init_dma_ecc(struct altr_edac_device_dev *device)
+{
+ int ret;
+
+ ret = altr_init_a10_ecc_device_type("altr,socfpga-dma-ecc");
+ if (ret)
+ return ret;
+
+ return altr_check_ecc_deps(device);
+}
+
+static const struct edac_device_prv_data a10_dmaecc_data = {
+ .setup = socfpga_init_dma_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_DMA */
+
+/********************** USB Device Functions **********************/
+
+#ifdef CONFIG_EDAC_ALTERA_USB
+
+static int __init socfpga_init_usb_ecc(struct altr_edac_device_dev *device)
+{
+ int ret;
+
+ ret = altr_init_a10_ecc_device_type("altr,socfpga-usb-ecc");
+ if (ret)
+ return ret;
+
+ return altr_check_ecc_deps(device);
+}
+
+static const struct edac_device_prv_data a10_usbecc_data = {
+ .setup = socfpga_init_usb_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject2_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_USB */
+
+/********************** QSPI Device Functions **********************/
+
+#ifdef CONFIG_EDAC_ALTERA_QSPI
+
+static int __init socfpga_init_qspi_ecc(struct altr_edac_device_dev *device)
+{
+ int ret;
+
+ ret = altr_init_a10_ecc_device_type("altr,socfpga-qspi-ecc");
+ if (ret)
+ return ret;
+
+ return altr_check_ecc_deps(device);
+}
+
+static const struct edac_device_prv_data a10_qspiecc_data = {
+ .setup = socfpga_init_qspi_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRA,
+ .ue_set_mask = ALTR_A10_ECC_TDERRA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_QSPI */
+
+/********************* SDMMC Device Functions **********************/
+
+#ifdef CONFIG_EDAC_ALTERA_SDMMC
+
+static const struct edac_device_prv_data a10_sdmmceccb_data;
+static int altr_portb_setup(struct altr_edac_device_dev *device)
+{
+ struct edac_device_ctl_info *dci;
+ struct altr_edac_device_dev *altdev;
+ char *ecc_name = "sdmmcb-ecc";
+ int edac_idx, rc;
+ struct device_node *np;
+ const struct edac_device_prv_data *prv = &a10_sdmmceccb_data;
+
+ rc = altr_check_ecc_deps(device);
+ if (rc)
+ return rc;
+
+ np = of_find_compatible_node(NULL, NULL, "altr,socfpga-sdmmc-ecc");
+ if (!np) {
+ edac_printk(KERN_WARNING, EDAC_DEVICE, "SDMMC node not found\n");
+ return -ENODEV;
+ }
+
+ /* Create the PortB EDAC device */
+ edac_idx = edac_device_alloc_index();
+ dci = edac_device_alloc_ctl_info(sizeof(*altdev), ecc_name, 1,
+ ecc_name, 1, 0, NULL, 0, edac_idx);
+ if (!dci) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: Unable to allocate PortB EDAC device\n",
+ ecc_name);
+ return -ENOMEM;
+ }
+
+ /* Initialize the PortB EDAC device structure from PortA structure */
+ altdev = dci->pvt_info;
+ *altdev = *device;
+
+ if (!devres_open_group(&altdev->ddev, altr_portb_setup, GFP_KERNEL))
+ return -ENOMEM;
+
+ /* Update PortB specific values */
+ altdev->edac_dev_name = ecc_name;
+ altdev->edac_idx = edac_idx;
+ altdev->edac_dev = dci;
+ altdev->data = prv;
+ dci->dev = &altdev->ddev;
+ dci->ctl_name = "Altera ECC Manager";
+ dci->mod_name = ecc_name;
+ dci->dev_name = ecc_name;
+
+ /*
+ * Update the PortB IRQs - A10 has 4, S10 has 2, Index accordingly
+ *
+ * FIXME: Instead of ifdefs with different architectures the driver
+ * should properly use compatibles.
+ */
+#ifdef CONFIG_64BIT
+ altdev->sb_irq = irq_of_parse_and_map(np, 1);
+#else
+ altdev->sb_irq = irq_of_parse_and_map(np, 2);
+#endif
+ if (!altdev->sb_irq) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Error PortB SBIRQ alloc\n");
+ rc = -ENODEV;
+ goto err_release_group_1;
+ }
+ rc = devm_request_irq(&altdev->ddev, altdev->sb_irq,
+ prv->ecc_irq_handler,
+ IRQF_ONESHOT | IRQF_TRIGGER_HIGH,
+ ecc_name, altdev);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "PortB SBERR IRQ error\n");
+ goto err_release_group_1;
+ }
+
+#ifdef CONFIG_64BIT
+ /* Use IRQ to determine SError origin instead of assigning IRQ */
+ rc = of_property_read_u32_index(np, "interrupts", 1, &altdev->db_irq);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Error PortB DBIRQ alloc\n");
+ goto err_release_group_1;
+ }
+#else
+ altdev->db_irq = irq_of_parse_and_map(np, 3);
+ if (!altdev->db_irq) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Error PortB DBIRQ alloc\n");
+ rc = -ENODEV;
+ goto err_release_group_1;
+ }
+ rc = devm_request_irq(&altdev->ddev, altdev->db_irq,
+ prv->ecc_irq_handler,
+ IRQF_ONESHOT | IRQF_TRIGGER_HIGH,
+ ecc_name, altdev);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "PortB DBERR IRQ error\n");
+ goto err_release_group_1;
+ }
+#endif
+
+ rc = edac_device_add_device(dci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "edac_device_add_device portB failed\n");
+ rc = -ENOMEM;
+ goto err_release_group_1;
+ }
+ altr_create_edacdev_dbgfs(dci, prv);
+
+ list_add(&altdev->next, &altdev->edac->a10_ecc_devices);
+
+ devres_remove_group(&altdev->ddev, altr_portb_setup);
+
+ return 0;
+
+err_release_group_1:
+ edac_device_free_ctl_info(dci);
+ devres_release_group(&altdev->ddev, altr_portb_setup);
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s:Error setting up EDAC device: %d\n", ecc_name, rc);
+ return rc;
+}
+
+static int __init socfpga_init_sdmmc_ecc(struct altr_edac_device_dev *device)
+{
+ int rc = -ENODEV;
+ struct device_node *child;
+
+ child = of_find_compatible_node(NULL, NULL, "altr,socfpga-sdmmc-ecc");
+ if (!child)
+ return -ENODEV;
+
+ if (!of_device_is_available(child))
+ goto exit;
+
+ if (validate_parent_available(child))
+ goto exit;
+
+ /* Init portB */
+ rc = altr_init_a10_ecc_block(child, ALTR_A10_SDMMC_IRQ_MASK,
+ a10_sdmmceccb_data.ecc_enable_mask, 1);
+ if (rc)
+ goto exit;
+
+ /* Setup portB */
+ return altr_portb_setup(device);
+
+exit:
+ of_node_put(child);
+ return rc;
+}
+
+static irqreturn_t altr_edac_a10_ecc_irq_portb(int irq, void *dev_id)
+{
+ struct altr_edac_device_dev *ad = dev_id;
+ void __iomem *base = ad->base;
+ const struct edac_device_prv_data *priv = ad->data;
+
+ if (irq == ad->sb_irq) {
+ writel(priv->ce_clear_mask,
+ base + ALTR_A10_ECC_INTSTAT_OFST);
+ edac_device_handle_ce(ad->edac_dev, 0, 0, ad->edac_dev_name);
+ return IRQ_HANDLED;
+ } else if (irq == ad->db_irq) {
+ writel(priv->ue_clear_mask,
+ base + ALTR_A10_ECC_INTSTAT_OFST);
+ edac_device_handle_ue(ad->edac_dev, 0, 0, ad->edac_dev_name);
+ return IRQ_HANDLED;
+ }
+
+ WARN_ONCE(1, "Unhandled IRQ%d on Port B.", irq);
+
+ return IRQ_NONE;
+}
+
+static const struct edac_device_prv_data a10_sdmmcecca_data = {
+ .setup = socfpga_init_sdmmc_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENA,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_SERRPENA,
+ .ue_set_mask = ALTR_A10_ECC_DERRPENA,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+
+static const struct edac_device_prv_data a10_sdmmceccb_data = {
+ .setup = socfpga_init_sdmmc_ecc,
+ .ce_clear_mask = ALTR_A10_ECC_SERRPENB,
+ .ue_clear_mask = ALTR_A10_ECC_DERRPENB,
+ .ecc_enable_mask = ALTR_A10_COMMON_ECC_EN_CTL,
+ .ecc_en_ofst = ALTR_A10_ECC_CTRL_OFST,
+ .ce_set_mask = ALTR_A10_ECC_TSERRB,
+ .ue_set_mask = ALTR_A10_ECC_TDERRB,
+ .set_err_ofst = ALTR_A10_ECC_INTTEST_OFST,
+ .ecc_irq_handler = altr_edac_a10_ecc_irq_portb,
+ .inject_fops = &altr_edac_a10_device_inject_fops,
+};
+
+#endif /* CONFIG_EDAC_ALTERA_SDMMC */
+
+/********************* Arria10 EDAC Device Functions *************************/
+static const struct of_device_id altr_edac_a10_device_of_match[] = {
+#ifdef CONFIG_EDAC_ALTERA_L2C
+ { .compatible = "altr,socfpga-a10-l2-ecc", .data = &a10_l2ecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_OCRAM
+ { .compatible = "altr,socfpga-a10-ocram-ecc",
+ .data = &a10_ocramecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_ETHERNET
+ { .compatible = "altr,socfpga-eth-mac-ecc",
+ .data = &a10_enetecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_NAND
+ { .compatible = "altr,socfpga-nand-ecc", .data = &a10_nandecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_DMA
+ { .compatible = "altr,socfpga-dma-ecc", .data = &a10_dmaecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_USB
+ { .compatible = "altr,socfpga-usb-ecc", .data = &a10_usbecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_QSPI
+ { .compatible = "altr,socfpga-qspi-ecc", .data = &a10_qspiecc_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_SDMMC
+ { .compatible = "altr,socfpga-sdmmc-ecc", .data = &a10_sdmmcecca_data },
+#endif
+#ifdef CONFIG_EDAC_ALTERA_SDRAM
+ { .compatible = "altr,sdram-edac-s10", .data = &s10_sdramecc_data },
+#endif
+ {},
+};
+MODULE_DEVICE_TABLE(of, altr_edac_a10_device_of_match);
+
+/*
+ * The Arria10 EDAC Device Functions differ from the Cyclone5/Arria5
+ * because 2 IRQs are shared among the all ECC peripherals. The ECC
+ * manager manages the IRQs and the children.
+ * Based on xgene_edac.c peripheral code.
+ */
+
+static ssize_t __maybe_unused
+altr_edac_a10_device_trig(struct file *file, const char __user *user_buf,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dci = file->private_data;
+ struct altr_edac_device_dev *drvdata = edac_dci->pvt_info;
+ const struct edac_device_prv_data *priv = drvdata->data;
+ void __iomem *set_addr = (drvdata->base + priv->set_err_ofst);
+ unsigned long flags;
+ u8 trig_type;
+
+ if (!user_buf || get_user(trig_type, user_buf))
+ return -EFAULT;
+
+ local_irq_save(flags);
+ if (trig_type == ALTR_UE_TRIGGER_CHAR)
+ writel(priv->ue_set_mask, set_addr);
+ else
+ writel(priv->ce_set_mask, set_addr);
+
+ /* Ensure the interrupt test bits are set */
+ wmb();
+ local_irq_restore(flags);
+
+ return count;
+}
+
+/*
+ * The Stratix10 EDAC Error Injection Functions differ from Arria10
+ * slightly. A few Arria10 peripherals can use this injection function.
+ * Inject the error into the memory and then readback to trigger the IRQ.
+ */
+static ssize_t __maybe_unused
+altr_edac_a10_device_trig2(struct file *file, const char __user *user_buf,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dci = file->private_data;
+ struct altr_edac_device_dev *drvdata = edac_dci->pvt_info;
+ const struct edac_device_prv_data *priv = drvdata->data;
+ void __iomem *set_addr = (drvdata->base + priv->set_err_ofst);
+ unsigned long flags;
+ u8 trig_type;
+
+ if (!user_buf || get_user(trig_type, user_buf))
+ return -EFAULT;
+
+ local_irq_save(flags);
+ if (trig_type == ALTR_UE_TRIGGER_CHAR) {
+ writel(priv->ue_set_mask, set_addr);
+ } else {
+ /* Setup read/write of 4 bytes */
+ writel(ECC_WORD_WRITE, drvdata->base + ECC_BLK_DBYTECTRL_OFST);
+ /* Setup Address to 0 */
+ writel(0, drvdata->base + ECC_BLK_ADDRESS_OFST);
+ /* Setup accctrl to read & ecc & data override */
+ writel(ECC_READ_EDOVR, drvdata->base + ECC_BLK_ACCCTRL_OFST);
+ /* Kick it. */
+ writel(ECC_XACT_KICK, drvdata->base + ECC_BLK_STARTACC_OFST);
+ /* Setup write for single bit change */
+ writel(readl(drvdata->base + ECC_BLK_RDATA0_OFST) ^ 0x1,
+ drvdata->base + ECC_BLK_WDATA0_OFST);
+ writel(readl(drvdata->base + ECC_BLK_RDATA1_OFST),
+ drvdata->base + ECC_BLK_WDATA1_OFST);
+ writel(readl(drvdata->base + ECC_BLK_RDATA2_OFST),
+ drvdata->base + ECC_BLK_WDATA2_OFST);
+ writel(readl(drvdata->base + ECC_BLK_RDATA3_OFST),
+ drvdata->base + ECC_BLK_WDATA3_OFST);
+
+ /* Copy Read ECC to Write ECC */
+ writel(readl(drvdata->base + ECC_BLK_RECC0_OFST),
+ drvdata->base + ECC_BLK_WECC0_OFST);
+ writel(readl(drvdata->base + ECC_BLK_RECC1_OFST),
+ drvdata->base + ECC_BLK_WECC1_OFST);
+ /* Setup accctrl to write & ecc override & data override */
+ writel(ECC_WRITE_EDOVR, drvdata->base + ECC_BLK_ACCCTRL_OFST);
+ /* Kick it. */
+ writel(ECC_XACT_KICK, drvdata->base + ECC_BLK_STARTACC_OFST);
+ /* Setup accctrl to read & ecc overwrite & data overwrite */
+ writel(ECC_READ_EDOVR, drvdata->base + ECC_BLK_ACCCTRL_OFST);
+ /* Kick it. */
+ writel(ECC_XACT_KICK, drvdata->base + ECC_BLK_STARTACC_OFST);
+ }
+
+ /* Ensure the interrupt test bits are set */
+ wmb();
+ local_irq_restore(flags);
+
+ return count;
+}
+
+static void altr_edac_a10_irq_handler(struct irq_desc *desc)
+{
+ int dberr, bit, sm_offset, irq_status;
+ struct altr_arria10_edac *edac = irq_desc_get_handler_data(desc);
+ struct irq_chip *chip = irq_desc_get_chip(desc);
+ int irq = irq_desc_get_irq(desc);
+ unsigned long bits;
+
+ dberr = (irq == edac->db_irq) ? 1 : 0;
+ sm_offset = dberr ? A10_SYSMGR_ECC_INTSTAT_DERR_OFST :
+ A10_SYSMGR_ECC_INTSTAT_SERR_OFST;
+
+ chained_irq_enter(chip, desc);
+
+ regmap_read(edac->ecc_mgr_map, sm_offset, &irq_status);
+
+ bits = irq_status;
+ for_each_set_bit(bit, &bits, 32)
+ generic_handle_domain_irq(edac->domain, dberr * 32 + bit);
+
+ chained_irq_exit(chip, desc);
+}
+
+static int validate_parent_available(struct device_node *np)
+{
+ struct device_node *parent;
+ int ret = 0;
+
+ /* SDRAM must be present for Linux (implied parent) */
+ if (of_device_is_compatible(np, "altr,sdram-edac-s10"))
+ return 0;
+
+ /* Ensure parent device is enabled if parent node exists */
+ parent = of_parse_phandle(np, "altr,ecc-parent", 0);
+ if (parent && !of_device_is_available(parent))
+ ret = -ENODEV;
+
+ of_node_put(parent);
+ return ret;
+}
+
+static int get_s10_sdram_edac_resource(struct device_node *np,
+ struct resource *res)
+{
+ struct device_node *parent;
+ int ret;
+
+ parent = of_parse_phandle(np, "altr,sdr-syscon", 0);
+ if (!parent)
+ return -ENODEV;
+
+ ret = of_address_to_resource(parent, 0, res);
+ of_node_put(parent);
+
+ return ret;
+}
+
+static int altr_edac_a10_device_add(struct altr_arria10_edac *edac,
+ struct device_node *np)
+{
+ struct edac_device_ctl_info *dci;
+ struct altr_edac_device_dev *altdev;
+ char *ecc_name = (char *)np->name;
+ struct resource res;
+ int edac_idx;
+ int rc = 0;
+ const struct edac_device_prv_data *prv;
+ /* Get matching node and check for valid result */
+ const struct of_device_id *pdev_id =
+ of_match_node(altr_edac_a10_device_of_match, np);
+ if (IS_ERR_OR_NULL(pdev_id))
+ return -ENODEV;
+
+ /* Get driver specific data for this EDAC device */
+ prv = pdev_id->data;
+ if (IS_ERR_OR_NULL(prv))
+ return -ENODEV;
+
+ if (validate_parent_available(np))
+ return -ENODEV;
+
+ if (!devres_open_group(edac->dev, altr_edac_a10_device_add, GFP_KERNEL))
+ return -ENOMEM;
+
+ if (of_device_is_compatible(np, "altr,sdram-edac-s10"))
+ rc = get_s10_sdram_edac_resource(np, &res);
+ else
+ rc = of_address_to_resource(np, 0, &res);
+
+ if (rc < 0) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: no resource address\n", ecc_name);
+ goto err_release_group;
+ }
+
+ edac_idx = edac_device_alloc_index();
+ dci = edac_device_alloc_ctl_info(sizeof(*altdev), ecc_name,
+ 1, ecc_name, 1, 0, NULL, 0,
+ edac_idx);
+
+ if (!dci) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s: Unable to allocate EDAC device\n", ecc_name);
+ rc = -ENOMEM;
+ goto err_release_group;
+ }
+
+ altdev = dci->pvt_info;
+ dci->dev = edac->dev;
+ altdev->edac_dev_name = ecc_name;
+ altdev->edac_idx = edac_idx;
+ altdev->edac = edac;
+ altdev->edac_dev = dci;
+ altdev->data = prv;
+ altdev->ddev = *edac->dev;
+ dci->dev = &altdev->ddev;
+ dci->ctl_name = "Altera ECC Manager";
+ dci->mod_name = ecc_name;
+ dci->dev_name = ecc_name;
+
+ altdev->base = devm_ioremap_resource(edac->dev, &res);
+ if (IS_ERR(altdev->base)) {
+ rc = PTR_ERR(altdev->base);
+ goto err_release_group1;
+ }
+
+ /* Check specific dependencies for the module */
+ if (altdev->data->setup) {
+ rc = altdev->data->setup(altdev);
+ if (rc)
+ goto err_release_group1;
+ }
+
+ altdev->sb_irq = irq_of_parse_and_map(np, 0);
+ if (!altdev->sb_irq) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Error allocating SBIRQ\n");
+ rc = -ENODEV;
+ goto err_release_group1;
+ }
+ rc = devm_request_irq(edac->dev, altdev->sb_irq, prv->ecc_irq_handler,
+ IRQF_ONESHOT | IRQF_TRIGGER_HIGH,
+ ecc_name, altdev);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "No SBERR IRQ resource\n");
+ goto err_release_group1;
+ }
+
+#ifdef CONFIG_64BIT
+ /* Use IRQ to determine SError origin instead of assigning IRQ */
+ rc = of_property_read_u32_index(np, "interrupts", 0, &altdev->db_irq);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to parse DB IRQ index\n");
+ goto err_release_group1;
+ }
+#else
+ altdev->db_irq = irq_of_parse_and_map(np, 1);
+ if (!altdev->db_irq) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Error allocating DBIRQ\n");
+ rc = -ENODEV;
+ goto err_release_group1;
+ }
+ rc = devm_request_irq(edac->dev, altdev->db_irq, prv->ecc_irq_handler,
+ IRQF_ONESHOT | IRQF_TRIGGER_HIGH,
+ ecc_name, altdev);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "No DBERR IRQ resource\n");
+ goto err_release_group1;
+ }
+#endif
+
+ rc = edac_device_add_device(dci);
+ if (rc) {
+ dev_err(edac->dev, "edac_device_add_device failed\n");
+ rc = -ENOMEM;
+ goto err_release_group1;
+ }
+
+ altr_create_edacdev_dbgfs(dci, prv);
+
+ list_add(&altdev->next, &edac->a10_ecc_devices);
+
+ devres_remove_group(edac->dev, altr_edac_a10_device_add);
+
+ return 0;
+
+err_release_group1:
+ edac_device_free_ctl_info(dci);
+err_release_group:
+ devres_release_group(edac->dev, NULL);
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "%s:Error setting up EDAC device: %d\n", ecc_name, rc);
+
+ return rc;
+}
+
+static void a10_eccmgr_irq_mask(struct irq_data *d)
+{
+ struct altr_arria10_edac *edac = irq_data_get_irq_chip_data(d);
+
+ regmap_write(edac->ecc_mgr_map, A10_SYSMGR_ECC_INTMASK_SET_OFST,
+ BIT(d->hwirq));
+}
+
+static void a10_eccmgr_irq_unmask(struct irq_data *d)
+{
+ struct altr_arria10_edac *edac = irq_data_get_irq_chip_data(d);
+
+ regmap_write(edac->ecc_mgr_map, A10_SYSMGR_ECC_INTMASK_CLR_OFST,
+ BIT(d->hwirq));
+}
+
+static int a10_eccmgr_irqdomain_map(struct irq_domain *d, unsigned int irq,
+ irq_hw_number_t hwirq)
+{
+ struct altr_arria10_edac *edac = d->host_data;
+
+ irq_set_chip_and_handler(irq, &edac->irq_chip, handle_simple_irq);
+ irq_set_chip_data(irq, edac);
+ irq_set_noprobe(irq);
+
+ return 0;
+}
+
+static const struct irq_domain_ops a10_eccmgr_ic_ops = {
+ .map = a10_eccmgr_irqdomain_map,
+ .xlate = irq_domain_xlate_twocell,
+};
+
+/************** Stratix 10 EDAC Double Bit Error Handler ************/
+#define to_a10edac(p, m) container_of(p, struct altr_arria10_edac, m)
+
+#ifdef CONFIG_64BIT
+/* panic routine issues reboot on non-zero panic_timeout */
+extern int panic_timeout;
+
+/*
+ * The double bit error is handled through SError which is fatal. This is
+ * called as a panic notifier to printout ECC error info as part of the panic.
+ */
+static int s10_edac_dberr_handler(struct notifier_block *this,
+ unsigned long event, void *ptr)
+{
+ struct altr_arria10_edac *edac = to_a10edac(this, panic_notifier);
+ int err_addr, dberror;
+
+ regmap_read(edac->ecc_mgr_map, S10_SYSMGR_ECC_INTSTAT_DERR_OFST,
+ &dberror);
+ regmap_write(edac->ecc_mgr_map, S10_SYSMGR_UE_VAL_OFST, dberror);
+ if (dberror & S10_DBE_IRQ_MASK) {
+ struct list_head *position;
+ struct altr_edac_device_dev *ed;
+ struct arm_smccc_res result;
+
+ /* Find the matching DBE in the list of devices */
+ list_for_each(position, &edac->a10_ecc_devices) {
+ ed = list_entry(position, struct altr_edac_device_dev,
+ next);
+ if (!(BIT(ed->db_irq) & dberror))
+ continue;
+
+ writel(ALTR_A10_ECC_DERRPENA,
+ ed->base + ALTR_A10_ECC_INTSTAT_OFST);
+ err_addr = readl(ed->base + ALTR_S10_DERR_ADDRA_OFST);
+ regmap_write(edac->ecc_mgr_map,
+ S10_SYSMGR_UE_ADDR_OFST, err_addr);
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "EDAC: [Fatal DBE on %s @ 0x%08X]\n",
+ ed->edac_dev_name, err_addr);
+ break;
+ }
+ /* Notify the System through SMC. Reboot delay = 1 second */
+ panic_timeout = 1;
+ arm_smccc_smc(INTEL_SIP_SMC_ECC_DBE, dberror, 0, 0, 0, 0,
+ 0, 0, &result);
+ }
+
+ return NOTIFY_DONE;
+}
+#endif
+
+/****************** Arria 10 EDAC Probe Function *********************/
+static int altr_edac_a10_probe(struct platform_device *pdev)
+{
+ struct altr_arria10_edac *edac;
+ struct device_node *child;
+
+ edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL);
+ if (!edac)
+ return -ENOMEM;
+
+ edac->dev = &pdev->dev;
+ platform_set_drvdata(pdev, edac);
+ INIT_LIST_HEAD(&edac->a10_ecc_devices);
+
+ edac->ecc_mgr_map =
+ altr_sysmgr_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "altr,sysmgr-syscon");
+
+ if (IS_ERR(edac->ecc_mgr_map)) {
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Unable to get syscon altr,sysmgr-syscon\n");
+ return PTR_ERR(edac->ecc_mgr_map);
+ }
+
+ edac->irq_chip.name = pdev->dev.of_node->name;
+ edac->irq_chip.irq_mask = a10_eccmgr_irq_mask;
+ edac->irq_chip.irq_unmask = a10_eccmgr_irq_unmask;
+ edac->domain = irq_domain_add_linear(pdev->dev.of_node, 64,
+ &a10_eccmgr_ic_ops, edac);
+ if (!edac->domain) {
+ dev_err(&pdev->dev, "Error adding IRQ domain\n");
+ return -ENOMEM;
+ }
+
+ edac->sb_irq = platform_get_irq(pdev, 0);
+ if (edac->sb_irq < 0)
+ return edac->sb_irq;
+
+ irq_set_chained_handler_and_data(edac->sb_irq,
+ altr_edac_a10_irq_handler,
+ edac);
+
+#ifdef CONFIG_64BIT
+ {
+ int dberror, err_addr;
+
+ edac->panic_notifier.notifier_call = s10_edac_dberr_handler;
+ atomic_notifier_chain_register(&panic_notifier_list,
+ &edac->panic_notifier);
+
+ /* Printout a message if uncorrectable error previously. */
+ regmap_read(edac->ecc_mgr_map, S10_SYSMGR_UE_VAL_OFST,
+ &dberror);
+ if (dberror) {
+ regmap_read(edac->ecc_mgr_map, S10_SYSMGR_UE_ADDR_OFST,
+ &err_addr);
+ edac_printk(KERN_ERR, EDAC_DEVICE,
+ "Previous Boot UE detected[0x%X] @ 0x%X\n",
+ dberror, err_addr);
+ /* Reset the sticky registers */
+ regmap_write(edac->ecc_mgr_map,
+ S10_SYSMGR_UE_VAL_OFST, 0);
+ regmap_write(edac->ecc_mgr_map,
+ S10_SYSMGR_UE_ADDR_OFST, 0);
+ }
+ }
+#else
+ edac->db_irq = platform_get_irq(pdev, 1);
+ if (edac->db_irq < 0)
+ return edac->db_irq;
+
+ irq_set_chained_handler_and_data(edac->db_irq,
+ altr_edac_a10_irq_handler, edac);
+#endif
+
+ for_each_child_of_node(pdev->dev.of_node, child) {
+ if (!of_device_is_available(child))
+ continue;
+
+ if (of_match_node(altr_edac_a10_device_of_match, child))
+ altr_edac_a10_device_add(edac, child);
+
+#ifdef CONFIG_EDAC_ALTERA_SDRAM
+ else if (of_device_is_compatible(child, "altr,sdram-edac-a10"))
+ of_platform_populate(pdev->dev.of_node,
+ altr_sdram_ctrl_of_match,
+ NULL, &pdev->dev);
+#endif
+ }
+
+ return 0;
+}
+
+static const struct of_device_id altr_edac_a10_of_match[] = {
+ { .compatible = "altr,socfpga-a10-ecc-manager" },
+ { .compatible = "altr,socfpga-s10-ecc-manager" },
+ {},
+};
+MODULE_DEVICE_TABLE(of, altr_edac_a10_of_match);
+
+static struct platform_driver altr_edac_a10_driver = {
+ .probe = altr_edac_a10_probe,
+ .driver = {
+ .name = "socfpga_a10_ecc_manager",
+ .of_match_table = altr_edac_a10_of_match,
+ },
+};
+module_platform_driver(altr_edac_a10_driver);
+
+MODULE_AUTHOR("Thor Thayer");
+MODULE_DESCRIPTION("EDAC Driver for Altera Memories");
diff --git a/drivers/edac/altera_edac.h b/drivers/edac/altera_edac.h
new file mode 100644
index 0000000000..3727e72c8c
--- /dev/null
+++ b/drivers/edac/altera_edac.h
@@ -0,0 +1,398 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2017-2018, Intel Corporation
+ * Copyright (C) 2015 Altera Corporation
+ */
+
+#ifndef _ALTERA_EDAC_H
+#define _ALTERA_EDAC_H
+
+#include <linux/arm-smccc.h>
+#include <linux/edac.h>
+#include <linux/types.h>
+
+/* SDRAM Controller CtrlCfg Register */
+#define CV_CTLCFG_OFST 0x00
+
+/* SDRAM Controller CtrlCfg Register Bit Masks */
+#define CV_CTLCFG_ECC_EN 0x400
+#define CV_CTLCFG_ECC_CORR_EN 0x800
+#define CV_CTLCFG_GEN_SB_ERR 0x2000
+#define CV_CTLCFG_GEN_DB_ERR 0x4000
+
+#define CV_CTLCFG_ECC_AUTO_EN (CV_CTLCFG_ECC_EN)
+
+/* SDRAM Controller Address Width Register */
+#define CV_DRAMADDRW_OFST 0x2C
+
+/* SDRAM Controller Address Widths Field Register */
+#define DRAMADDRW_COLBIT_MASK 0x001F
+#define DRAMADDRW_COLBIT_SHIFT 0
+#define DRAMADDRW_ROWBIT_MASK 0x03E0
+#define DRAMADDRW_ROWBIT_SHIFT 5
+#define CV_DRAMADDRW_BANKBIT_MASK 0x1C00
+#define CV_DRAMADDRW_BANKBIT_SHIFT 10
+#define CV_DRAMADDRW_CSBIT_MASK 0xE000
+#define CV_DRAMADDRW_CSBIT_SHIFT 13
+
+/* SDRAM Controller Interface Data Width Register */
+#define CV_DRAMIFWIDTH_OFST 0x30
+
+/* SDRAM Controller Interface Data Width Defines */
+#define CV_DRAMIFWIDTH_16B_ECC 24
+#define CV_DRAMIFWIDTH_32B_ECC 40
+
+/* SDRAM Controller DRAM Status Register */
+#define CV_DRAMSTS_OFST 0x38
+
+/* SDRAM Controller DRAM Status Register Bit Masks */
+#define CV_DRAMSTS_SBEERR 0x04
+#define CV_DRAMSTS_DBEERR 0x08
+#define CV_DRAMSTS_CORR_DROP 0x10
+
+/* SDRAM Controller DRAM IRQ Register */
+#define CV_DRAMINTR_OFST 0x3C
+
+/* SDRAM Controller DRAM IRQ Register Bit Masks */
+#define CV_DRAMINTR_INTREN 0x01
+#define CV_DRAMINTR_SBEMASK 0x02
+#define CV_DRAMINTR_DBEMASK 0x04
+#define CV_DRAMINTR_CORRDROPMASK 0x08
+#define CV_DRAMINTR_INTRCLR 0x10
+
+/* SDRAM Controller Single Bit Error Count Register */
+#define CV_SBECOUNT_OFST 0x40
+
+/* SDRAM Controller Double Bit Error Count Register */
+#define CV_DBECOUNT_OFST 0x44
+
+/* SDRAM Controller ECC Error Address Register */
+#define CV_ERRADDR_OFST 0x48
+
+/*-----------------------------------------*/
+
+/* SDRAM Controller EccCtrl Register */
+#define A10_ECCCTRL1_OFST 0x00
+
+/* SDRAM Controller EccCtrl Register Bit Masks */
+#define A10_ECCCTRL1_ECC_EN 0x001
+#define A10_ECCCTRL1_CNT_RST 0x010
+#define A10_ECCCTRL1_AWB_CNT_RST 0x100
+#define A10_ECC_CNT_RESET_MASK (A10_ECCCTRL1_CNT_RST | \
+ A10_ECCCTRL1_AWB_CNT_RST)
+
+/* SDRAM Controller Address Width Register */
+#define CV_DRAMADDRW 0xFFC2502C
+#define A10_DRAMADDRW 0xFFCFA0A8
+#define S10_DRAMADDRW 0xF80110E0
+
+/* SDRAM Controller Address Widths Field Register */
+#define DRAMADDRW_COLBIT_MASK 0x001F
+#define DRAMADDRW_COLBIT_SHIFT 0
+#define DRAMADDRW_ROWBIT_MASK 0x03E0
+#define DRAMADDRW_ROWBIT_SHIFT 5
+#define CV_DRAMADDRW_BANKBIT_MASK 0x1C00
+#define CV_DRAMADDRW_BANKBIT_SHIFT 10
+#define CV_DRAMADDRW_CSBIT_MASK 0xE000
+#define CV_DRAMADDRW_CSBIT_SHIFT 13
+
+#define A10_DRAMADDRW_BANKBIT_MASK 0x3C00
+#define A10_DRAMADDRW_BANKBIT_SHIFT 10
+#define A10_DRAMADDRW_GRPBIT_MASK 0xC000
+#define A10_DRAMADDRW_GRPBIT_SHIFT 14
+#define A10_DRAMADDRW_CSBIT_MASK 0x70000
+#define A10_DRAMADDRW_CSBIT_SHIFT 16
+
+/* SDRAM Controller Interface Data Width Register */
+#define CV_DRAMIFWIDTH 0xFFC25030
+#define A10_DRAMIFWIDTH 0xFFCFB008
+#define S10_DRAMIFWIDTH 0xF8011008
+
+/* SDRAM Controller Interface Data Width Defines */
+#define CV_DRAMIFWIDTH_16B_ECC 24
+#define CV_DRAMIFWIDTH_32B_ECC 40
+
+#define A10_DRAMIFWIDTH_16B 0x0
+#define A10_DRAMIFWIDTH_32B 0x1
+#define A10_DRAMIFWIDTH_64B 0x2
+
+/* SDRAM Controller DRAM IRQ Register */
+#define A10_ERRINTEN_OFST 0x10
+
+/* SDRAM Controller DRAM IRQ Register Bit Masks */
+#define A10_ERRINTEN_SERRINTEN 0x01
+#define A10_ERRINTEN_DERRINTEN 0x02
+#define A10_ECC_IRQ_EN_MASK (A10_ERRINTEN_SERRINTEN | \
+ A10_ERRINTEN_DERRINTEN)
+
+/* SDRAM Interrupt Mode Register */
+#define A10_INTMODE_OFST 0x1C
+#define A10_INTMODE_SB_INT 1
+
+/* SDRAM Controller Error Status Register */
+#define A10_INTSTAT_OFST 0x20
+
+/* SDRAM Controller Error Status Register Bit Masks */
+#define A10_INTSTAT_SBEERR 0x01
+#define A10_INTSTAT_DBEERR 0x02
+
+/* SDRAM Controller ECC Error Address Register */
+#define A10_DERRADDR_OFST 0x2C
+#define A10_SERRADDR_OFST 0x30
+
+/* SDRAM Controller ECC Diagnostic Register */
+#define A10_DIAGINTTEST_OFST 0x24
+
+#define A10_DIAGINT_TSERRA_MASK 0x0001
+#define A10_DIAGINT_TDERRA_MASK 0x0100
+
+#define A10_SBERR_IRQ 34
+#define A10_DBERR_IRQ 32
+
+/* SDRAM Single Bit Error Count Compare Set Register */
+#define A10_SERRCNTREG_OFST 0x3C
+
+#define A10_SYMAN_INTMASK_CLR 0xFFD06098
+#define A10_INTMASK_CLR_OFST 0x10
+#define A10_DDR0_IRQ_MASK BIT(17)
+
+struct altr_sdram_prv_data {
+ int ecc_ctrl_offset;
+ int ecc_ctl_en_mask;
+ int ecc_cecnt_offset;
+ int ecc_uecnt_offset;
+ int ecc_stat_offset;
+ int ecc_stat_ce_mask;
+ int ecc_stat_ue_mask;
+ int ecc_saddr_offset;
+ int ecc_daddr_offset;
+ int ecc_irq_en_offset;
+ int ecc_irq_en_mask;
+ int ecc_irq_clr_offset;
+ int ecc_irq_clr_mask;
+ int ecc_cnt_rst_offset;
+ int ecc_cnt_rst_mask;
+ struct edac_dev_sysfs_attribute *eccmgr_sysfs_attr;
+ int ecc_enable_mask;
+ int ce_set_mask;
+ int ue_set_mask;
+ int ce_ue_trgr_offset;
+};
+
+/* Altera SDRAM Memory Controller data */
+struct altr_sdram_mc_data {
+ struct regmap *mc_vbase;
+ int sb_irq;
+ int db_irq;
+ const struct altr_sdram_prv_data *data;
+};
+
+/************************** EDAC Device Defines **************************/
+/***** General Device Trigger Defines *****/
+#define ALTR_UE_TRIGGER_CHAR 'U' /* Trigger for UE */
+#define ALTR_TRIGGER_READ_WRD_CNT 32 /* Line size x 4 */
+#define ALTR_TRIG_OCRAM_BYTE_SIZE 128 /* Line size x 4 */
+#define ALTR_TRIG_L2C_BYTE_SIZE 4096 /* Full Page */
+
+/******* Cyclone5 and Arria5 Defines *******/
+/* OCRAM ECC Management Group Defines */
+#define ALTR_MAN_GRP_OCRAM_ECC_OFFSET 0x04
+#define ALTR_OCR_ECC_REG_OFFSET 0x00
+#define ALTR_OCR_ECC_EN BIT(0)
+#define ALTR_OCR_ECC_INJS BIT(1)
+#define ALTR_OCR_ECC_INJD BIT(2)
+#define ALTR_OCR_ECC_SERR BIT(3)
+#define ALTR_OCR_ECC_DERR BIT(4)
+
+/* L2 ECC Management Group Defines */
+#define ALTR_MAN_GRP_L2_ECC_OFFSET 0x00
+#define ALTR_L2_ECC_REG_OFFSET 0x00
+#define ALTR_L2_ECC_EN BIT(0)
+#define ALTR_L2_ECC_INJS BIT(1)
+#define ALTR_L2_ECC_INJD BIT(2)
+
+/* Arria10 General ECC Block Module Defines */
+#define ALTR_A10_ECC_CTRL_OFST 0x08
+#define ALTR_A10_ECC_EN BIT(0)
+#define ALTR_A10_ECC_INITA BIT(16)
+#define ALTR_A10_ECC_INITB BIT(24)
+
+#define ALTR_A10_ECC_INITSTAT_OFST 0x0C
+#define ALTR_A10_ECC_INITCOMPLETEA BIT(0)
+#define ALTR_A10_ECC_INITCOMPLETEB BIT(8)
+
+#define ALTR_A10_ECC_ERRINTEN_OFST 0x10
+#define ALTR_A10_ECC_ERRINTENS_OFST 0x14
+#define ALTR_A10_ECC_ERRINTENR_OFST 0x18
+#define ALTR_A10_ECC_SERRINTEN BIT(0)
+
+#define ALTR_A10_ECC_INTMODE_OFST 0x1C
+#define ALTR_A10_ECC_INTMODE BIT(0)
+
+#define ALTR_A10_ECC_INTSTAT_OFST 0x20
+#define ALTR_A10_ECC_SERRPENA BIT(0)
+#define ALTR_A10_ECC_DERRPENA BIT(8)
+#define ALTR_A10_ECC_ERRPENA_MASK (ALTR_A10_ECC_SERRPENA | \
+ ALTR_A10_ECC_DERRPENA)
+#define ALTR_A10_ECC_SERRPENB BIT(16)
+#define ALTR_A10_ECC_DERRPENB BIT(24)
+#define ALTR_A10_ECC_ERRPENB_MASK (ALTR_A10_ECC_SERRPENB | \
+ ALTR_A10_ECC_DERRPENB)
+
+#define ALTR_A10_ECC_INTTEST_OFST 0x24
+#define ALTR_A10_ECC_TSERRA BIT(0)
+#define ALTR_A10_ECC_TDERRA BIT(8)
+#define ALTR_A10_ECC_TSERRB BIT(16)
+#define ALTR_A10_ECC_TDERRB BIT(24)
+
+/* ECC Manager Defines */
+#define A10_SYSMGR_ECC_INTMASK_SET_OFST 0x94
+#define A10_SYSMGR_ECC_INTMASK_CLR_OFST 0x98
+#define A10_SYSMGR_ECC_INTMASK_OCRAM BIT(1)
+
+#define A10_SYSMGR_ECC_INTSTAT_SERR_OFST 0x9C
+#define A10_SYSMGR_ECC_INTSTAT_DERR_OFST 0xA0
+#define A10_SYSMGR_ECC_INTSTAT_L2 BIT(0)
+#define A10_SYSMGR_ECC_INTSTAT_OCRAM BIT(1)
+
+#define A10_SYSGMR_MPU_CLEAR_L2_ECC_OFST 0xA8
+#define A10_SYSGMR_MPU_CLEAR_L2_ECC_SB BIT(15)
+#define A10_SYSGMR_MPU_CLEAR_L2_ECC_MB BIT(31)
+
+/* Arria 10 L2 ECC Management Group Defines */
+#define ALTR_A10_L2_ECC_CTL_OFST 0x0
+#define ALTR_A10_L2_ECC_EN_CTL BIT(0)
+
+#define ALTR_A10_L2_ECC_STATUS 0xFFD060A4
+#define ALTR_A10_L2_ECC_STAT_OFST 0xA4
+#define ALTR_A10_L2_ECC_SERR_PEND BIT(0)
+#define ALTR_A10_L2_ECC_MERR_PEND BIT(0)
+
+#define ALTR_A10_L2_ECC_CLR_OFST 0x4
+#define ALTR_A10_L2_ECC_SERR_CLR BIT(15)
+#define ALTR_A10_L2_ECC_MERR_CLR BIT(31)
+
+#define ALTR_A10_L2_ECC_INJ_OFST ALTR_A10_L2_ECC_CTL_OFST
+#define ALTR_A10_L2_ECC_CE_INJ_MASK 0x00000101
+#define ALTR_A10_L2_ECC_UE_INJ_MASK 0x00010101
+
+/* Arria 10 OCRAM ECC Management Group Defines */
+#define ALTR_A10_OCRAM_ECC_EN_CTL (BIT(1) | BIT(0))
+
+/* Arria 10 Ethernet ECC Management Group Defines */
+#define ALTR_A10_COMMON_ECC_EN_CTL BIT(0)
+
+/* Arria 10 SDMMC ECC Management Group Defines */
+#define ALTR_A10_SDMMC_IRQ_MASK (BIT(16) | BIT(15))
+
+/* A10 ECC Controller memory initialization timeout */
+#define ALTR_A10_ECC_INIT_WATCHDOG_10US 10000
+
+/************* Stratix10 Defines **************/
+#define ALTR_S10_ECC_CTRL_SDRAM_OFST 0x00
+#define ALTR_S10_ECC_EN BIT(0)
+
+#define ALTR_S10_ECC_ERRINTEN_OFST 0x10
+#define ALTR_S10_ECC_ERRINTENS_OFST 0x14
+#define ALTR_S10_ECC_ERRINTENR_OFST 0x18
+#define ALTR_S10_ECC_SERRINTEN BIT(0)
+
+#define ALTR_S10_ECC_INTMODE_OFST 0x1C
+#define ALTR_S10_ECC_INTMODE BIT(0)
+
+#define ALTR_S10_ECC_INTSTAT_OFST 0x20
+#define ALTR_S10_ECC_SERRPENA BIT(0)
+#define ALTR_S10_ECC_DERRPENA BIT(8)
+#define ALTR_S10_ECC_ERRPENA_MASK (ALTR_S10_ECC_SERRPENA | \
+ ALTR_S10_ECC_DERRPENA)
+
+#define ALTR_S10_ECC_INTTEST_OFST 0x24
+#define ALTR_S10_ECC_TSERRA BIT(0)
+#define ALTR_S10_ECC_TDERRA BIT(8)
+#define ALTR_S10_ECC_TSERRB BIT(16)
+#define ALTR_S10_ECC_TDERRB BIT(24)
+
+#define ALTR_S10_DERR_ADDRA_OFST 0x2C
+
+/* Stratix10 ECC Manager Defines */
+#define S10_SYSMGR_ECC_INTMASK_CLR_OFST 0x98
+#define S10_SYSMGR_ECC_INTSTAT_DERR_OFST 0xA0
+
+/* Sticky registers for Uncorrected Errors */
+#define S10_SYSMGR_UE_VAL_OFST 0x220
+#define S10_SYSMGR_UE_ADDR_OFST 0x224
+
+#define S10_DDR0_IRQ_MASK BIT(16)
+#define S10_DBE_IRQ_MASK 0x3FFFE
+
+/* Define ECC Block Offsets for peripherals */
+#define ECC_BLK_ADDRESS_OFST 0x40
+#define ECC_BLK_RDATA0_OFST 0x44
+#define ECC_BLK_RDATA1_OFST 0x48
+#define ECC_BLK_RDATA2_OFST 0x4C
+#define ECC_BLK_RDATA3_OFST 0x50
+#define ECC_BLK_WDATA0_OFST 0x54
+#define ECC_BLK_WDATA1_OFST 0x58
+#define ECC_BLK_WDATA2_OFST 0x5C
+#define ECC_BLK_WDATA3_OFST 0x60
+#define ECC_BLK_RECC0_OFST 0x64
+#define ECC_BLK_RECC1_OFST 0x68
+#define ECC_BLK_WECC0_OFST 0x6C
+#define ECC_BLK_WECC1_OFST 0x70
+#define ECC_BLK_DBYTECTRL_OFST 0x74
+#define ECC_BLK_ACCCTRL_OFST 0x78
+#define ECC_BLK_STARTACC_OFST 0x7C
+
+#define ECC_XACT_KICK 0x10000
+#define ECC_WORD_WRITE 0xFF
+#define ECC_WRITE_DOVR 0x101
+#define ECC_WRITE_EDOVR 0x103
+#define ECC_READ_EOVR 0x2
+#define ECC_READ_EDOVR 0x3
+
+struct altr_edac_device_dev;
+
+struct edac_device_prv_data {
+ int (*setup)(struct altr_edac_device_dev *device);
+ int ce_clear_mask;
+ int ue_clear_mask;
+ int irq_status_mask;
+ void * (*alloc_mem)(size_t size, void **other);
+ void (*free_mem)(void *p, size_t size, void *other);
+ int ecc_enable_mask;
+ int ecc_en_ofst;
+ int ce_set_mask;
+ int ue_set_mask;
+ int set_err_ofst;
+ irqreturn_t (*ecc_irq_handler)(int irq, void *dev_id);
+ int trig_alloc_sz;
+ const struct file_operations *inject_fops;
+ bool panic;
+};
+
+struct altr_edac_device_dev {
+ struct list_head next;
+ void __iomem *base;
+ int sb_irq;
+ int db_irq;
+ const struct edac_device_prv_data *data;
+ struct dentry *debugfs_dir;
+ char *edac_dev_name;
+ struct altr_arria10_edac *edac;
+ struct edac_device_ctl_info *edac_dev;
+ struct device ddev;
+ int edac_idx;
+};
+
+struct altr_arria10_edac {
+ struct device *dev;
+ struct regmap *ecc_mgr_map;
+ int sb_irq;
+ int db_irq;
+ struct irq_domain *domain;
+ struct irq_chip irq_chip;
+ struct list_head a10_ecc_devices;
+ struct notifier_block panic_notifier;
+};
+
+#endif /* #ifndef _ALTERA_EDAC_H */
diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
new file mode 100644
index 0000000000..9b6642d008
--- /dev/null
+++ b/drivers/edac/amd64_edac.c
@@ -0,0 +1,4472 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include "amd64_edac.h"
+#include <asm/amd_nb.h>
+
+static struct edac_pci_ctl_info *pci_ctl;
+
+/*
+ * Set by command line parameter. If BIOS has enabled the ECC, this override is
+ * cleared to prevent re-enabling the hardware by this driver.
+ */
+static int ecc_enable_override;
+module_param(ecc_enable_override, int, 0644);
+
+static struct msr __percpu *msrs;
+
+static inline u32 get_umc_reg(struct amd64_pvt *pvt, u32 reg)
+{
+ if (!pvt->flags.zn_regs_v2)
+ return reg;
+
+ switch (reg) {
+ case UMCCH_ADDR_CFG: return UMCCH_ADDR_CFG_DDR5;
+ case UMCCH_ADDR_MASK_SEC: return UMCCH_ADDR_MASK_SEC_DDR5;
+ case UMCCH_DIMM_CFG: return UMCCH_DIMM_CFG_DDR5;
+ }
+
+ WARN_ONCE(1, "%s: unknown register 0x%x", __func__, reg);
+ return 0;
+}
+
+/* Per-node stuff */
+static struct ecc_settings **ecc_stngs;
+
+/* Device for the PCI component */
+static struct device *pci_ctl_dev;
+
+/*
+ * Valid scrub rates for the K8 hardware memory scrubber. We map the scrubbing
+ * bandwidth to a valid bit pattern. The 'set' operation finds the 'matching-
+ * or higher value'.
+ *
+ *FIXME: Produce a better mapping/linearisation.
+ */
+static const struct scrubrate {
+ u32 scrubval; /* bit pattern for scrub rate */
+ u32 bandwidth; /* bandwidth consumed (bytes/sec) */
+} scrubrates[] = {
+ { 0x01, 1600000000UL},
+ { 0x02, 800000000UL},
+ { 0x03, 400000000UL},
+ { 0x04, 200000000UL},
+ { 0x05, 100000000UL},
+ { 0x06, 50000000UL},
+ { 0x07, 25000000UL},
+ { 0x08, 12284069UL},
+ { 0x09, 6274509UL},
+ { 0x0A, 3121951UL},
+ { 0x0B, 1560975UL},
+ { 0x0C, 781440UL},
+ { 0x0D, 390720UL},
+ { 0x0E, 195300UL},
+ { 0x0F, 97650UL},
+ { 0x10, 48854UL},
+ { 0x11, 24427UL},
+ { 0x12, 12213UL},
+ { 0x13, 6101UL},
+ { 0x14, 3051UL},
+ { 0x15, 1523UL},
+ { 0x16, 761UL},
+ { 0x00, 0UL}, /* scrubbing off */
+};
+
+int __amd64_read_pci_cfg_dword(struct pci_dev *pdev, int offset,
+ u32 *val, const char *func)
+{
+ int err = 0;
+
+ err = pci_read_config_dword(pdev, offset, val);
+ if (err)
+ amd64_warn("%s: error reading F%dx%03x.\n",
+ func, PCI_FUNC(pdev->devfn), offset);
+
+ return err;
+}
+
+int __amd64_write_pci_cfg_dword(struct pci_dev *pdev, int offset,
+ u32 val, const char *func)
+{
+ int err = 0;
+
+ err = pci_write_config_dword(pdev, offset, val);
+ if (err)
+ amd64_warn("%s: error writing to F%dx%03x.\n",
+ func, PCI_FUNC(pdev->devfn), offset);
+
+ return err;
+}
+
+/*
+ * Select DCT to which PCI cfg accesses are routed
+ */
+static void f15h_select_dct(struct amd64_pvt *pvt, u8 dct)
+{
+ u32 reg = 0;
+
+ amd64_read_pci_cfg(pvt->F1, DCT_CFG_SEL, &reg);
+ reg &= (pvt->model == 0x30) ? ~3 : ~1;
+ reg |= dct;
+ amd64_write_pci_cfg(pvt->F1, DCT_CFG_SEL, reg);
+}
+
+/*
+ *
+ * Depending on the family, F2 DCT reads need special handling:
+ *
+ * K8: has a single DCT only and no address offsets >= 0x100
+ *
+ * F10h: each DCT has its own set of regs
+ * DCT0 -> F2x040..
+ * DCT1 -> F2x140..
+ *
+ * F16h: has only 1 DCT
+ *
+ * F15h: we select which DCT we access using F1x10C[DctCfgSel]
+ */
+static inline int amd64_read_dct_pci_cfg(struct amd64_pvt *pvt, u8 dct,
+ int offset, u32 *val)
+{
+ switch (pvt->fam) {
+ case 0xf:
+ if (dct || offset >= 0x100)
+ return -EINVAL;
+ break;
+
+ case 0x10:
+ if (dct) {
+ /*
+ * Note: If ganging is enabled, barring the regs
+ * F2x[1,0]98 and F2x[1,0]9C; reads reads to F2x1xx
+ * return 0. (cf. Section 2.8.1 F10h BKDG)
+ */
+ if (dct_ganging_enabled(pvt))
+ return 0;
+
+ offset += 0x100;
+ }
+ break;
+
+ case 0x15:
+ /*
+ * F15h: F2x1xx addresses do not map explicitly to DCT1.
+ * We should select which DCT we access using F1x10C[DctCfgSel]
+ */
+ dct = (dct && pvt->model == 0x30) ? 3 : dct;
+ f15h_select_dct(pvt, dct);
+ break;
+
+ case 0x16:
+ if (dct)
+ return -EINVAL;
+ break;
+
+ default:
+ break;
+ }
+ return amd64_read_pci_cfg(pvt->F2, offset, val);
+}
+
+/*
+ * Memory scrubber control interface. For K8, memory scrubbing is handled by
+ * hardware and can involve L2 cache, dcache as well as the main memory. With
+ * F10, this is extended to L3 cache scrubbing on CPU models sporting that
+ * functionality.
+ *
+ * This causes the "units" for the scrubbing speed to vary from 64 byte blocks
+ * (dram) over to cache lines. This is nasty, so we will use bandwidth in
+ * bytes/sec for the setting.
+ *
+ * Currently, we only do dram scrubbing. If the scrubbing is done in software on
+ * other archs, we might not have access to the caches directly.
+ */
+
+/*
+ * Scan the scrub rate mapping table for a close or matching bandwidth value to
+ * issue. If requested is too big, then use last maximum value found.
+ */
+static int __set_scrub_rate(struct amd64_pvt *pvt, u32 new_bw, u32 min_rate)
+{
+ u32 scrubval;
+ int i;
+
+ /*
+ * map the configured rate (new_bw) to a value specific to the AMD64
+ * memory controller and apply to register. Search for the first
+ * bandwidth entry that is greater or equal than the setting requested
+ * and program that. If at last entry, turn off DRAM scrubbing.
+ *
+ * If no suitable bandwidth is found, turn off DRAM scrubbing entirely
+ * by falling back to the last element in scrubrates[].
+ */
+ for (i = 0; i < ARRAY_SIZE(scrubrates) - 1; i++) {
+ /*
+ * skip scrub rates which aren't recommended
+ * (see F10 BKDG, F3x58)
+ */
+ if (scrubrates[i].scrubval < min_rate)
+ continue;
+
+ if (scrubrates[i].bandwidth <= new_bw)
+ break;
+ }
+
+ scrubval = scrubrates[i].scrubval;
+
+ if (pvt->fam == 0x15 && pvt->model == 0x60) {
+ f15h_select_dct(pvt, 0);
+ pci_write_bits32(pvt->F2, F15H_M60H_SCRCTRL, scrubval, 0x001F);
+ f15h_select_dct(pvt, 1);
+ pci_write_bits32(pvt->F2, F15H_M60H_SCRCTRL, scrubval, 0x001F);
+ } else {
+ pci_write_bits32(pvt->F3, SCRCTRL, scrubval, 0x001F);
+ }
+
+ if (scrubval)
+ return scrubrates[i].bandwidth;
+
+ return 0;
+}
+
+static int set_scrub_rate(struct mem_ctl_info *mci, u32 bw)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ u32 min_scrubrate = 0x5;
+
+ if (pvt->fam == 0xf)
+ min_scrubrate = 0x0;
+
+ if (pvt->fam == 0x15) {
+ /* Erratum #505 */
+ if (pvt->model < 0x10)
+ f15h_select_dct(pvt, 0);
+
+ if (pvt->model == 0x60)
+ min_scrubrate = 0x6;
+ }
+ return __set_scrub_rate(pvt, bw, min_scrubrate);
+}
+
+static int get_scrub_rate(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ int i, retval = -EINVAL;
+ u32 scrubval = 0;
+
+ if (pvt->fam == 0x15) {
+ /* Erratum #505 */
+ if (pvt->model < 0x10)
+ f15h_select_dct(pvt, 0);
+
+ if (pvt->model == 0x60)
+ amd64_read_pci_cfg(pvt->F2, F15H_M60H_SCRCTRL, &scrubval);
+ else
+ amd64_read_pci_cfg(pvt->F3, SCRCTRL, &scrubval);
+ } else {
+ amd64_read_pci_cfg(pvt->F3, SCRCTRL, &scrubval);
+ }
+
+ scrubval = scrubval & 0x001F;
+
+ for (i = 0; i < ARRAY_SIZE(scrubrates); i++) {
+ if (scrubrates[i].scrubval == scrubval) {
+ retval = scrubrates[i].bandwidth;
+ break;
+ }
+ }
+ return retval;
+}
+
+/*
+ * returns true if the SysAddr given by sys_addr matches the
+ * DRAM base/limit associated with node_id
+ */
+static bool base_limit_match(struct amd64_pvt *pvt, u64 sys_addr, u8 nid)
+{
+ u64 addr;
+
+ /* The K8 treats this as a 40-bit value. However, bits 63-40 will be
+ * all ones if the most significant implemented address bit is 1.
+ * Here we discard bits 63-40. See section 3.4.2 of AMD publication
+ * 24592: AMD x86-64 Architecture Programmer's Manual Volume 1
+ * Application Programming.
+ */
+ addr = sys_addr & 0x000000ffffffffffull;
+
+ return ((addr >= get_dram_base(pvt, nid)) &&
+ (addr <= get_dram_limit(pvt, nid)));
+}
+
+/*
+ * Attempt to map a SysAddr to a node. On success, return a pointer to the
+ * mem_ctl_info structure for the node that the SysAddr maps to.
+ *
+ * On failure, return NULL.
+ */
+static struct mem_ctl_info *find_mc_by_sys_addr(struct mem_ctl_info *mci,
+ u64 sys_addr)
+{
+ struct amd64_pvt *pvt;
+ u8 node_id;
+ u32 intlv_en, bits;
+
+ /*
+ * Here we use the DRAM Base (section 3.4.4.1) and DRAM Limit (section
+ * 3.4.4.2) registers to map the SysAddr to a node ID.
+ */
+ pvt = mci->pvt_info;
+
+ /*
+ * The value of this field should be the same for all DRAM Base
+ * registers. Therefore we arbitrarily choose to read it from the
+ * register for node 0.
+ */
+ intlv_en = dram_intlv_en(pvt, 0);
+
+ if (intlv_en == 0) {
+ for (node_id = 0; node_id < DRAM_RANGES; node_id++) {
+ if (base_limit_match(pvt, sys_addr, node_id))
+ goto found;
+ }
+ goto err_no_match;
+ }
+
+ if (unlikely((intlv_en != 0x01) &&
+ (intlv_en != 0x03) &&
+ (intlv_en != 0x07))) {
+ amd64_warn("DRAM Base[IntlvEn] junk value: 0x%x, BIOS bug?\n", intlv_en);
+ return NULL;
+ }
+
+ bits = (((u32) sys_addr) >> 12) & intlv_en;
+
+ for (node_id = 0; ; ) {
+ if ((dram_intlv_sel(pvt, node_id) & intlv_en) == bits)
+ break; /* intlv_sel field matches */
+
+ if (++node_id >= DRAM_RANGES)
+ goto err_no_match;
+ }
+
+ /* sanity test for sys_addr */
+ if (unlikely(!base_limit_match(pvt, sys_addr, node_id))) {
+ amd64_warn("%s: sys_addr 0x%llx falls outside base/limit address"
+ "range for node %d with node interleaving enabled.\n",
+ __func__, sys_addr, node_id);
+ return NULL;
+ }
+
+found:
+ return edac_mc_find((int)node_id);
+
+err_no_match:
+ edac_dbg(2, "sys_addr 0x%lx doesn't match any node\n",
+ (unsigned long)sys_addr);
+
+ return NULL;
+}
+
+/*
+ * compute the CS base address of the @csrow on the DRAM controller @dct.
+ * For details see F2x[5C:40] in the processor's BKDG
+ */
+static void get_cs_base_and_mask(struct amd64_pvt *pvt, int csrow, u8 dct,
+ u64 *base, u64 *mask)
+{
+ u64 csbase, csmask, base_bits, mask_bits;
+ u8 addr_shift;
+
+ if (pvt->fam == 0xf && pvt->ext_model < K8_REV_F) {
+ csbase = pvt->csels[dct].csbases[csrow];
+ csmask = pvt->csels[dct].csmasks[csrow];
+ base_bits = GENMASK_ULL(31, 21) | GENMASK_ULL(15, 9);
+ mask_bits = GENMASK_ULL(29, 21) | GENMASK_ULL(15, 9);
+ addr_shift = 4;
+
+ /*
+ * F16h and F15h, models 30h and later need two addr_shift values:
+ * 8 for high and 6 for low (cf. F16h BKDG).
+ */
+ } else if (pvt->fam == 0x16 ||
+ (pvt->fam == 0x15 && pvt->model >= 0x30)) {
+ csbase = pvt->csels[dct].csbases[csrow];
+ csmask = pvt->csels[dct].csmasks[csrow >> 1];
+
+ *base = (csbase & GENMASK_ULL(15, 5)) << 6;
+ *base |= (csbase & GENMASK_ULL(30, 19)) << 8;
+
+ *mask = ~0ULL;
+ /* poke holes for the csmask */
+ *mask &= ~((GENMASK_ULL(15, 5) << 6) |
+ (GENMASK_ULL(30, 19) << 8));
+
+ *mask |= (csmask & GENMASK_ULL(15, 5)) << 6;
+ *mask |= (csmask & GENMASK_ULL(30, 19)) << 8;
+
+ return;
+ } else {
+ csbase = pvt->csels[dct].csbases[csrow];
+ csmask = pvt->csels[dct].csmasks[csrow >> 1];
+ addr_shift = 8;
+
+ if (pvt->fam == 0x15)
+ base_bits = mask_bits =
+ GENMASK_ULL(30,19) | GENMASK_ULL(13,5);
+ else
+ base_bits = mask_bits =
+ GENMASK_ULL(28,19) | GENMASK_ULL(13,5);
+ }
+
+ *base = (csbase & base_bits) << addr_shift;
+
+ *mask = ~0ULL;
+ /* poke holes for the csmask */
+ *mask &= ~(mask_bits << addr_shift);
+ /* OR them in */
+ *mask |= (csmask & mask_bits) << addr_shift;
+}
+
+#define for_each_chip_select(i, dct, pvt) \
+ for (i = 0; i < pvt->csels[dct].b_cnt; i++)
+
+#define chip_select_base(i, dct, pvt) \
+ pvt->csels[dct].csbases[i]
+
+#define for_each_chip_select_mask(i, dct, pvt) \
+ for (i = 0; i < pvt->csels[dct].m_cnt; i++)
+
+#define for_each_umc(i) \
+ for (i = 0; i < pvt->max_mcs; i++)
+
+/*
+ * @input_addr is an InputAddr associated with the node given by mci. Return the
+ * csrow that input_addr maps to, or -1 on failure (no csrow claims input_addr).
+ */
+static int input_addr_to_csrow(struct mem_ctl_info *mci, u64 input_addr)
+{
+ struct amd64_pvt *pvt;
+ int csrow;
+ u64 base, mask;
+
+ pvt = mci->pvt_info;
+
+ for_each_chip_select(csrow, 0, pvt) {
+ if (!csrow_enabled(csrow, 0, pvt))
+ continue;
+
+ get_cs_base_and_mask(pvt, csrow, 0, &base, &mask);
+
+ mask = ~mask;
+
+ if ((input_addr & mask) == (base & mask)) {
+ edac_dbg(2, "InputAddr 0x%lx matches csrow %d (node %d)\n",
+ (unsigned long)input_addr, csrow,
+ pvt->mc_node_id);
+
+ return csrow;
+ }
+ }
+ edac_dbg(2, "no matching csrow for InputAddr 0x%lx (MC node %d)\n",
+ (unsigned long)input_addr, pvt->mc_node_id);
+
+ return -1;
+}
+
+/*
+ * Obtain info from the DRAM Hole Address Register (section 3.4.8, pub #26094)
+ * for the node represented by mci. Info is passed back in *hole_base,
+ * *hole_offset, and *hole_size. Function returns 0 if info is valid or 1 if
+ * info is invalid. Info may be invalid for either of the following reasons:
+ *
+ * - The revision of the node is not E or greater. In this case, the DRAM Hole
+ * Address Register does not exist.
+ *
+ * - The DramHoleValid bit is cleared in the DRAM Hole Address Register,
+ * indicating that its contents are not valid.
+ *
+ * The values passed back in *hole_base, *hole_offset, and *hole_size are
+ * complete 32-bit values despite the fact that the bitfields in the DHAR
+ * only represent bits 31-24 of the base and offset values.
+ */
+static int get_dram_hole_info(struct mem_ctl_info *mci, u64 *hole_base,
+ u64 *hole_offset, u64 *hole_size)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ /* only revE and later have the DRAM Hole Address Register */
+ if (pvt->fam == 0xf && pvt->ext_model < K8_REV_E) {
+ edac_dbg(1, " revision %d for node %d does not support DHAR\n",
+ pvt->ext_model, pvt->mc_node_id);
+ return 1;
+ }
+
+ /* valid for Fam10h and above */
+ if (pvt->fam >= 0x10 && !dhar_mem_hoist_valid(pvt)) {
+ edac_dbg(1, " Dram Memory Hoisting is DISABLED on this system\n");
+ return 1;
+ }
+
+ if (!dhar_valid(pvt)) {
+ edac_dbg(1, " Dram Memory Hoisting is DISABLED on this node %d\n",
+ pvt->mc_node_id);
+ return 1;
+ }
+
+ /* This node has Memory Hoisting */
+
+ /* +------------------+--------------------+--------------------+-----
+ * | memory | DRAM hole | relocated |
+ * | [0, (x - 1)] | [x, 0xffffffff] | addresses from |
+ * | | | DRAM hole |
+ * | | | [0x100000000, |
+ * | | | (0x100000000+ |
+ * | | | (0xffffffff-x))] |
+ * +------------------+--------------------+--------------------+-----
+ *
+ * Above is a diagram of physical memory showing the DRAM hole and the
+ * relocated addresses from the DRAM hole. As shown, the DRAM hole
+ * starts at address x (the base address) and extends through address
+ * 0xffffffff. The DRAM Hole Address Register (DHAR) relocates the
+ * addresses in the hole so that they start at 0x100000000.
+ */
+
+ *hole_base = dhar_base(pvt);
+ *hole_size = (1ULL << 32) - *hole_base;
+
+ *hole_offset = (pvt->fam > 0xf) ? f10_dhar_offset(pvt)
+ : k8_dhar_offset(pvt);
+
+ edac_dbg(1, " DHAR info for node %d base 0x%lx offset 0x%lx size 0x%lx\n",
+ pvt->mc_node_id, (unsigned long)*hole_base,
+ (unsigned long)*hole_offset, (unsigned long)*hole_size);
+
+ return 0;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+#define EDAC_DCT_ATTR_SHOW(reg) \
+static ssize_t reg##_show(struct device *dev, \
+ struct device_attribute *mattr, char *data) \
+{ \
+ struct mem_ctl_info *mci = to_mci(dev); \
+ struct amd64_pvt *pvt = mci->pvt_info; \
+ \
+ return sprintf(data, "0x%016llx\n", (u64)pvt->reg); \
+}
+
+EDAC_DCT_ATTR_SHOW(dhar);
+EDAC_DCT_ATTR_SHOW(dbam0);
+EDAC_DCT_ATTR_SHOW(top_mem);
+EDAC_DCT_ATTR_SHOW(top_mem2);
+
+static ssize_t dram_hole_show(struct device *dev, struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ u64 hole_base = 0;
+ u64 hole_offset = 0;
+ u64 hole_size = 0;
+
+ get_dram_hole_info(mci, &hole_base, &hole_offset, &hole_size);
+
+ return sprintf(data, "%llx %llx %llx\n", hole_base, hole_offset,
+ hole_size);
+}
+
+/*
+ * update NUM_DBG_ATTRS in case you add new members
+ */
+static DEVICE_ATTR(dhar, S_IRUGO, dhar_show, NULL);
+static DEVICE_ATTR(dbam, S_IRUGO, dbam0_show, NULL);
+static DEVICE_ATTR(topmem, S_IRUGO, top_mem_show, NULL);
+static DEVICE_ATTR(topmem2, S_IRUGO, top_mem2_show, NULL);
+static DEVICE_ATTR_RO(dram_hole);
+
+static struct attribute *dbg_attrs[] = {
+ &dev_attr_dhar.attr,
+ &dev_attr_dbam.attr,
+ &dev_attr_topmem.attr,
+ &dev_attr_topmem2.attr,
+ &dev_attr_dram_hole.attr,
+ NULL
+};
+
+static const struct attribute_group dbg_group = {
+ .attrs = dbg_attrs,
+};
+
+static ssize_t inject_section_show(struct device *dev,
+ struct device_attribute *mattr, char *buf)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ return sprintf(buf, "0x%x\n", pvt->injection.section);
+}
+
+/*
+ * store error injection section value which refers to one of 4 16-byte sections
+ * within a 64-byte cacheline
+ *
+ * range: 0..3
+ */
+static ssize_t inject_section_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int ret;
+
+ ret = kstrtoul(data, 10, &value);
+ if (ret < 0)
+ return ret;
+
+ if (value > 3) {
+ amd64_warn("%s: invalid section 0x%lx\n", __func__, value);
+ return -EINVAL;
+ }
+
+ pvt->injection.section = (u32) value;
+ return count;
+}
+
+static ssize_t inject_word_show(struct device *dev,
+ struct device_attribute *mattr, char *buf)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ return sprintf(buf, "0x%x\n", pvt->injection.word);
+}
+
+/*
+ * store error injection word value which refers to one of 9 16-bit word of the
+ * 16-byte (128-bit + ECC bits) section
+ *
+ * range: 0..8
+ */
+static ssize_t inject_word_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int ret;
+
+ ret = kstrtoul(data, 10, &value);
+ if (ret < 0)
+ return ret;
+
+ if (value > 8) {
+ amd64_warn("%s: invalid word 0x%lx\n", __func__, value);
+ return -EINVAL;
+ }
+
+ pvt->injection.word = (u32) value;
+ return count;
+}
+
+static ssize_t inject_ecc_vector_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *buf)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ return sprintf(buf, "0x%x\n", pvt->injection.bit_map);
+}
+
+/*
+ * store 16 bit error injection vector which enables injecting errors to the
+ * corresponding bit within the error injection word above. When used during a
+ * DRAM ECC read, it holds the contents of the of the DRAM ECC bits.
+ */
+static ssize_t inject_ecc_vector_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int ret;
+
+ ret = kstrtoul(data, 16, &value);
+ if (ret < 0)
+ return ret;
+
+ if (value & 0xFFFF0000) {
+ amd64_warn("%s: invalid EccVector: 0x%lx\n", __func__, value);
+ return -EINVAL;
+ }
+
+ pvt->injection.bit_map = (u32) value;
+ return count;
+}
+
+/*
+ * Do a DRAM ECC read. Assemble staged values in the pvt area, format into
+ * fields needed by the injection registers and read the NB Array Data Port.
+ */
+static ssize_t inject_read_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ u32 section, word_bits;
+ int ret;
+
+ ret = kstrtoul(data, 10, &value);
+ if (ret < 0)
+ return ret;
+
+ /* Form value to choose 16-byte section of cacheline */
+ section = F10_NB_ARRAY_DRAM | SET_NB_ARRAY_ADDR(pvt->injection.section);
+
+ amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_ADDR, section);
+
+ word_bits = SET_NB_DRAM_INJECTION_READ(pvt->injection);
+
+ /* Issue 'word' and 'bit' along with the READ request */
+ amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, word_bits);
+
+ edac_dbg(0, "section=0x%x word_bits=0x%x\n", section, word_bits);
+
+ return count;
+}
+
+/*
+ * Do a DRAM ECC write. Assemble staged values in the pvt area and format into
+ * fields needed by the injection registers.
+ */
+static ssize_t inject_write_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+ u32 section, word_bits, tmp;
+ unsigned long value;
+ int ret;
+
+ ret = kstrtoul(data, 10, &value);
+ if (ret < 0)
+ return ret;
+
+ /* Form value to choose 16-byte section of cacheline */
+ section = F10_NB_ARRAY_DRAM | SET_NB_ARRAY_ADDR(pvt->injection.section);
+
+ amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_ADDR, section);
+
+ word_bits = SET_NB_DRAM_INJECTION_WRITE(pvt->injection);
+
+ pr_notice_once("Don't forget to decrease MCE polling interval in\n"
+ "/sys/bus/machinecheck/devices/machinecheck<CPUNUM>/check_interval\n"
+ "so that you can get the error report faster.\n");
+
+ on_each_cpu(disable_caches, NULL, 1);
+
+ /* Issue 'word' and 'bit' along with the READ request */
+ amd64_write_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, word_bits);
+
+ retry:
+ /* wait until injection happens */
+ amd64_read_pci_cfg(pvt->F3, F10_NB_ARRAY_DATA, &tmp);
+ if (tmp & F10_NB_ARR_ECC_WR_REQ) {
+ cpu_relax();
+ goto retry;
+ }
+
+ on_each_cpu(enable_caches, NULL, 1);
+
+ edac_dbg(0, "section=0x%x word_bits=0x%x\n", section, word_bits);
+
+ return count;
+}
+
+/*
+ * update NUM_INJ_ATTRS in case you add new members
+ */
+
+static DEVICE_ATTR_RW(inject_section);
+static DEVICE_ATTR_RW(inject_word);
+static DEVICE_ATTR_RW(inject_ecc_vector);
+static DEVICE_ATTR_WO(inject_write);
+static DEVICE_ATTR_WO(inject_read);
+
+static struct attribute *inj_attrs[] = {
+ &dev_attr_inject_section.attr,
+ &dev_attr_inject_word.attr,
+ &dev_attr_inject_ecc_vector.attr,
+ &dev_attr_inject_write.attr,
+ &dev_attr_inject_read.attr,
+ NULL
+};
+
+static umode_t inj_is_visible(struct kobject *kobj, struct attribute *attr, int idx)
+{
+ struct device *dev = kobj_to_dev(kobj);
+ struct mem_ctl_info *mci = container_of(dev, struct mem_ctl_info, dev);
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ /* Families which have that injection hw */
+ if (pvt->fam >= 0x10 && pvt->fam <= 0x16)
+ return attr->mode;
+
+ return 0;
+}
+
+static const struct attribute_group inj_group = {
+ .attrs = inj_attrs,
+ .is_visible = inj_is_visible,
+};
+#endif /* CONFIG_EDAC_DEBUG */
+
+/*
+ * Return the DramAddr that the SysAddr given by @sys_addr maps to. It is
+ * assumed that sys_addr maps to the node given by mci.
+ *
+ * The first part of section 3.4.4 (p. 70) shows how the DRAM Base (section
+ * 3.4.4.1) and DRAM Limit (section 3.4.4.2) registers are used to translate a
+ * SysAddr to a DramAddr. If the DRAM Hole Address Register (DHAR) is enabled,
+ * then it is also involved in translating a SysAddr to a DramAddr. Sections
+ * 3.4.8 and 3.5.8.2 describe the DHAR and how it is used for memory hoisting.
+ * These parts of the documentation are unclear. I interpret them as follows:
+ *
+ * When node n receives a SysAddr, it processes the SysAddr as follows:
+ *
+ * 1. It extracts the DRAMBase and DRAMLimit values from the DRAM Base and DRAM
+ * Limit registers for node n. If the SysAddr is not within the range
+ * specified by the base and limit values, then node n ignores the Sysaddr
+ * (since it does not map to node n). Otherwise continue to step 2 below.
+ *
+ * 2. If the DramHoleValid bit of the DHAR for node n is clear, the DHAR is
+ * disabled so skip to step 3 below. Otherwise see if the SysAddr is within
+ * the range of relocated addresses (starting at 0x100000000) from the DRAM
+ * hole. If not, skip to step 3 below. Else get the value of the
+ * DramHoleOffset field from the DHAR. To obtain the DramAddr, subtract the
+ * offset defined by this value from the SysAddr.
+ *
+ * 3. Obtain the base address for node n from the DRAMBase field of the DRAM
+ * Base register for node n. To obtain the DramAddr, subtract the base
+ * address from the SysAddr, as shown near the start of section 3.4.4 (p.70).
+ */
+static u64 sys_addr_to_dram_addr(struct mem_ctl_info *mci, u64 sys_addr)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ u64 dram_base, hole_base, hole_offset, hole_size, dram_addr;
+ int ret;
+
+ dram_base = get_dram_base(pvt, pvt->mc_node_id);
+
+ ret = get_dram_hole_info(mci, &hole_base, &hole_offset, &hole_size);
+ if (!ret) {
+ if ((sys_addr >= (1ULL << 32)) &&
+ (sys_addr < ((1ULL << 32) + hole_size))) {
+ /* use DHAR to translate SysAddr to DramAddr */
+ dram_addr = sys_addr - hole_offset;
+
+ edac_dbg(2, "using DHAR to translate SysAddr 0x%lx to DramAddr 0x%lx\n",
+ (unsigned long)sys_addr,
+ (unsigned long)dram_addr);
+
+ return dram_addr;
+ }
+ }
+
+ /*
+ * Translate the SysAddr to a DramAddr as shown near the start of
+ * section 3.4.4 (p. 70). Although sys_addr is a 64-bit value, the k8
+ * only deals with 40-bit values. Therefore we discard bits 63-40 of
+ * sys_addr below. If bit 39 of sys_addr is 1 then the bits we
+ * discard are all 1s. Otherwise the bits we discard are all 0s. See
+ * section 3.4.2 of AMD publication 24592: AMD x86-64 Architecture
+ * Programmer's Manual Volume 1 Application Programming.
+ */
+ dram_addr = (sys_addr & GENMASK_ULL(39, 0)) - dram_base;
+
+ edac_dbg(2, "using DRAM Base register to translate SysAddr 0x%lx to DramAddr 0x%lx\n",
+ (unsigned long)sys_addr, (unsigned long)dram_addr);
+ return dram_addr;
+}
+
+/*
+ * @intlv_en is the value of the IntlvEn field from a DRAM Base register
+ * (section 3.4.4.1). Return the number of bits from a SysAddr that are used
+ * for node interleaving.
+ */
+static int num_node_interleave_bits(unsigned intlv_en)
+{
+ static const int intlv_shift_table[] = { 0, 1, 0, 2, 0, 0, 0, 3 };
+ int n;
+
+ BUG_ON(intlv_en > 7);
+ n = intlv_shift_table[intlv_en];
+ return n;
+}
+
+/* Translate the DramAddr given by @dram_addr to an InputAddr. */
+static u64 dram_addr_to_input_addr(struct mem_ctl_info *mci, u64 dram_addr)
+{
+ struct amd64_pvt *pvt;
+ int intlv_shift;
+ u64 input_addr;
+
+ pvt = mci->pvt_info;
+
+ /*
+ * See the start of section 3.4.4 (p. 70, BKDG #26094, K8, revA-E)
+ * concerning translating a DramAddr to an InputAddr.
+ */
+ intlv_shift = num_node_interleave_bits(dram_intlv_en(pvt, 0));
+ input_addr = ((dram_addr >> intlv_shift) & GENMASK_ULL(35, 12)) +
+ (dram_addr & 0xfff);
+
+ edac_dbg(2, " Intlv Shift=%d DramAddr=0x%lx maps to InputAddr=0x%lx\n",
+ intlv_shift, (unsigned long)dram_addr,
+ (unsigned long)input_addr);
+
+ return input_addr;
+}
+
+/*
+ * Translate the SysAddr represented by @sys_addr to an InputAddr. It is
+ * assumed that @sys_addr maps to the node given by mci.
+ */
+static u64 sys_addr_to_input_addr(struct mem_ctl_info *mci, u64 sys_addr)
+{
+ u64 input_addr;
+
+ input_addr =
+ dram_addr_to_input_addr(mci, sys_addr_to_dram_addr(mci, sys_addr));
+
+ edac_dbg(2, "SysAddr 0x%lx translates to InputAddr 0x%lx\n",
+ (unsigned long)sys_addr, (unsigned long)input_addr);
+
+ return input_addr;
+}
+
+/* Map the Error address to a PAGE and PAGE OFFSET. */
+static inline void error_address_to_page_and_offset(u64 error_address,
+ struct err_info *err)
+{
+ err->page = (u32) (error_address >> PAGE_SHIFT);
+ err->offset = ((u32) error_address) & ~PAGE_MASK;
+}
+
+/*
+ * @sys_addr is an error address (a SysAddr) extracted from the MCA NB Address
+ * Low (section 3.6.4.5) and MCA NB Address High (section 3.6.4.6) registers
+ * of a node that detected an ECC memory error. mci represents the node that
+ * the error address maps to (possibly different from the node that detected
+ * the error). Return the number of the csrow that sys_addr maps to, or -1 on
+ * error.
+ */
+static int sys_addr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr)
+{
+ int csrow;
+
+ csrow = input_addr_to_csrow(mci, sys_addr_to_input_addr(mci, sys_addr));
+
+ if (csrow == -1)
+ amd64_mc_err(mci, "Failed to translate InputAddr to csrow for "
+ "address 0x%lx\n", (unsigned long)sys_addr);
+ return csrow;
+}
+
+/*
+ * See AMD PPR DF::LclNodeTypeMap
+ *
+ * This register gives information for nodes of the same type within a system.
+ *
+ * Reading this register from a GPU node will tell how many GPU nodes are in the
+ * system and what the lowest AMD Node ID value is for the GPU nodes. Use this
+ * info to fixup the Linux logical "Node ID" value set in the AMD NB code and EDAC.
+ */
+static struct local_node_map {
+ u16 node_count;
+ u16 base_node_id;
+} gpu_node_map;
+
+#define PCI_DEVICE_ID_AMD_MI200_DF_F1 0x14d1
+#define REG_LOCAL_NODE_TYPE_MAP 0x144
+
+/* Local Node Type Map (LNTM) fields */
+#define LNTM_NODE_COUNT GENMASK(27, 16)
+#define LNTM_BASE_NODE_ID GENMASK(11, 0)
+
+static int gpu_get_node_map(void)
+{
+ struct pci_dev *pdev;
+ int ret;
+ u32 tmp;
+
+ /*
+ * Node ID 0 is reserved for CPUs.
+ * Therefore, a non-zero Node ID means we've already cached the values.
+ */
+ if (gpu_node_map.base_node_id)
+ return 0;
+
+ pdev = pci_get_device(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_MI200_DF_F1, NULL);
+ if (!pdev) {
+ ret = -ENODEV;
+ goto out;
+ }
+
+ ret = pci_read_config_dword(pdev, REG_LOCAL_NODE_TYPE_MAP, &tmp);
+ if (ret)
+ goto out;
+
+ gpu_node_map.node_count = FIELD_GET(LNTM_NODE_COUNT, tmp);
+ gpu_node_map.base_node_id = FIELD_GET(LNTM_BASE_NODE_ID, tmp);
+
+out:
+ pci_dev_put(pdev);
+ return ret;
+}
+
+static int fixup_node_id(int node_id, struct mce *m)
+{
+ /* MCA_IPID[InstanceIdHi] give the AMD Node ID for the bank. */
+ u8 nid = (m->ipid >> 44) & 0xF;
+
+ if (smca_get_bank_type(m->extcpu, m->bank) != SMCA_UMC_V2)
+ return node_id;
+
+ /* Nodes below the GPU base node are CPU nodes and don't need a fixup. */
+ if (nid < gpu_node_map.base_node_id)
+ return node_id;
+
+ /* Convert the hardware-provided AMD Node ID to a Linux logical one. */
+ return nid - gpu_node_map.base_node_id + 1;
+}
+
+/* Protect the PCI config register pairs used for DF indirect access. */
+static DEFINE_MUTEX(df_indirect_mutex);
+
+/*
+ * Data Fabric Indirect Access uses FICAA/FICAD.
+ *
+ * Fabric Indirect Configuration Access Address (FICAA): Constructed based
+ * on the device's Instance Id and the PCI function and register offset of
+ * the desired register.
+ *
+ * Fabric Indirect Configuration Access Data (FICAD): There are FICAD LO
+ * and FICAD HI registers but so far we only need the LO register.
+ *
+ * Use Instance Id 0xFF to indicate a broadcast read.
+ */
+#define DF_BROADCAST 0xFF
+static int __df_indirect_read(u16 node, u8 func, u16 reg, u8 instance_id, u32 *lo)
+{
+ struct pci_dev *F4;
+ u32 ficaa;
+ int err = -ENODEV;
+
+ if (node >= amd_nb_num())
+ goto out;
+
+ F4 = node_to_amd_nb(node)->link;
+ if (!F4)
+ goto out;
+
+ ficaa = (instance_id == DF_BROADCAST) ? 0 : 1;
+ ficaa |= reg & 0x3FC;
+ ficaa |= (func & 0x7) << 11;
+ ficaa |= instance_id << 16;
+
+ mutex_lock(&df_indirect_mutex);
+
+ err = pci_write_config_dword(F4, 0x5C, ficaa);
+ if (err) {
+ pr_warn("Error writing DF Indirect FICAA, FICAA=0x%x\n", ficaa);
+ goto out_unlock;
+ }
+
+ err = pci_read_config_dword(F4, 0x98, lo);
+ if (err)
+ pr_warn("Error reading DF Indirect FICAD LO, FICAA=0x%x.\n", ficaa);
+
+out_unlock:
+ mutex_unlock(&df_indirect_mutex);
+
+out:
+ return err;
+}
+
+static int df_indirect_read_instance(u16 node, u8 func, u16 reg, u8 instance_id, u32 *lo)
+{
+ return __df_indirect_read(node, func, reg, instance_id, lo);
+}
+
+static int df_indirect_read_broadcast(u16 node, u8 func, u16 reg, u32 *lo)
+{
+ return __df_indirect_read(node, func, reg, DF_BROADCAST, lo);
+}
+
+struct addr_ctx {
+ u64 ret_addr;
+ u32 tmp;
+ u16 nid;
+ u8 inst_id;
+};
+
+static int umc_normaddr_to_sysaddr(u64 norm_addr, u16 nid, u8 umc, u64 *sys_addr)
+{
+ u64 dram_base_addr, dram_limit_addr, dram_hole_base;
+
+ u8 die_id_shift, die_id_mask, socket_id_shift, socket_id_mask;
+ u8 intlv_num_dies, intlv_num_chan, intlv_num_sockets;
+ u8 intlv_addr_sel, intlv_addr_bit;
+ u8 num_intlv_bits, hashed_bit;
+ u8 lgcy_mmio_hole_en, base = 0;
+ u8 cs_mask, cs_id = 0;
+ bool hash_enabled = false;
+
+ struct addr_ctx ctx;
+
+ memset(&ctx, 0, sizeof(ctx));
+
+ /* Start from the normalized address */
+ ctx.ret_addr = norm_addr;
+
+ ctx.nid = nid;
+ ctx.inst_id = umc;
+
+ /* Read D18F0x1B4 (DramOffset), check if base 1 is used. */
+ if (df_indirect_read_instance(nid, 0, 0x1B4, umc, &ctx.tmp))
+ goto out_err;
+
+ /* Remove HiAddrOffset from normalized address, if enabled: */
+ if (ctx.tmp & BIT(0)) {
+ u64 hi_addr_offset = (ctx.tmp & GENMASK_ULL(31, 20)) << 8;
+
+ if (norm_addr >= hi_addr_offset) {
+ ctx.ret_addr -= hi_addr_offset;
+ base = 1;
+ }
+ }
+
+ /* Read D18F0x110 (DramBaseAddress). */
+ if (df_indirect_read_instance(nid, 0, 0x110 + (8 * base), umc, &ctx.tmp))
+ goto out_err;
+
+ /* Check if address range is valid. */
+ if (!(ctx.tmp & BIT(0))) {
+ pr_err("%s: Invalid DramBaseAddress range: 0x%x.\n",
+ __func__, ctx.tmp);
+ goto out_err;
+ }
+
+ lgcy_mmio_hole_en = ctx.tmp & BIT(1);
+ intlv_num_chan = (ctx.tmp >> 4) & 0xF;
+ intlv_addr_sel = (ctx.tmp >> 8) & 0x7;
+ dram_base_addr = (ctx.tmp & GENMASK_ULL(31, 12)) << 16;
+
+ /* {0, 1, 2, 3} map to address bits {8, 9, 10, 11} respectively */
+ if (intlv_addr_sel > 3) {
+ pr_err("%s: Invalid interleave address select %d.\n",
+ __func__, intlv_addr_sel);
+ goto out_err;
+ }
+
+ /* Read D18F0x114 (DramLimitAddress). */
+ if (df_indirect_read_instance(nid, 0, 0x114 + (8 * base), umc, &ctx.tmp))
+ goto out_err;
+
+ intlv_num_sockets = (ctx.tmp >> 8) & 0x1;
+ intlv_num_dies = (ctx.tmp >> 10) & 0x3;
+ dram_limit_addr = ((ctx.tmp & GENMASK_ULL(31, 12)) << 16) | GENMASK_ULL(27, 0);
+
+ intlv_addr_bit = intlv_addr_sel + 8;
+
+ /* Re-use intlv_num_chan by setting it equal to log2(#channels) */
+ switch (intlv_num_chan) {
+ case 0: intlv_num_chan = 0; break;
+ case 1: intlv_num_chan = 1; break;
+ case 3: intlv_num_chan = 2; break;
+ case 5: intlv_num_chan = 3; break;
+ case 7: intlv_num_chan = 4; break;
+
+ case 8: intlv_num_chan = 1;
+ hash_enabled = true;
+ break;
+ default:
+ pr_err("%s: Invalid number of interleaved channels %d.\n",
+ __func__, intlv_num_chan);
+ goto out_err;
+ }
+
+ num_intlv_bits = intlv_num_chan;
+
+ if (intlv_num_dies > 2) {
+ pr_err("%s: Invalid number of interleaved nodes/dies %d.\n",
+ __func__, intlv_num_dies);
+ goto out_err;
+ }
+
+ num_intlv_bits += intlv_num_dies;
+
+ /* Add a bit if sockets are interleaved. */
+ num_intlv_bits += intlv_num_sockets;
+
+ /* Assert num_intlv_bits <= 4 */
+ if (num_intlv_bits > 4) {
+ pr_err("%s: Invalid interleave bits %d.\n",
+ __func__, num_intlv_bits);
+ goto out_err;
+ }
+
+ if (num_intlv_bits > 0) {
+ u64 temp_addr_x, temp_addr_i, temp_addr_y;
+ u8 die_id_bit, sock_id_bit, cs_fabric_id;
+
+ /*
+ * Read FabricBlockInstanceInformation3_CS[BlockFabricID].
+ * This is the fabric id for this coherent slave. Use
+ * umc/channel# as instance id of the coherent slave
+ * for FICAA.
+ */
+ if (df_indirect_read_instance(nid, 0, 0x50, umc, &ctx.tmp))
+ goto out_err;
+
+ cs_fabric_id = (ctx.tmp >> 8) & 0xFF;
+ die_id_bit = 0;
+
+ /* If interleaved over more than 1 channel: */
+ if (intlv_num_chan) {
+ die_id_bit = intlv_num_chan;
+ cs_mask = (1 << die_id_bit) - 1;
+ cs_id = cs_fabric_id & cs_mask;
+ }
+
+ sock_id_bit = die_id_bit;
+
+ /* Read D18F1x208 (SystemFabricIdMask). */
+ if (intlv_num_dies || intlv_num_sockets)
+ if (df_indirect_read_broadcast(nid, 1, 0x208, &ctx.tmp))
+ goto out_err;
+
+ /* If interleaved over more than 1 die. */
+ if (intlv_num_dies) {
+ sock_id_bit = die_id_bit + intlv_num_dies;
+ die_id_shift = (ctx.tmp >> 24) & 0xF;
+ die_id_mask = (ctx.tmp >> 8) & 0xFF;
+
+ cs_id |= ((cs_fabric_id & die_id_mask) >> die_id_shift) << die_id_bit;
+ }
+
+ /* If interleaved over more than 1 socket. */
+ if (intlv_num_sockets) {
+ socket_id_shift = (ctx.tmp >> 28) & 0xF;
+ socket_id_mask = (ctx.tmp >> 16) & 0xFF;
+
+ cs_id |= ((cs_fabric_id & socket_id_mask) >> socket_id_shift) << sock_id_bit;
+ }
+
+ /*
+ * The pre-interleaved address consists of XXXXXXIIIYYYYY
+ * where III is the ID for this CS, and XXXXXXYYYYY are the
+ * address bits from the post-interleaved address.
+ * "num_intlv_bits" has been calculated to tell us how many "I"
+ * bits there are. "intlv_addr_bit" tells us how many "Y" bits
+ * there are (where "I" starts).
+ */
+ temp_addr_y = ctx.ret_addr & GENMASK_ULL(intlv_addr_bit - 1, 0);
+ temp_addr_i = (cs_id << intlv_addr_bit);
+ temp_addr_x = (ctx.ret_addr & GENMASK_ULL(63, intlv_addr_bit)) << num_intlv_bits;
+ ctx.ret_addr = temp_addr_x | temp_addr_i | temp_addr_y;
+ }
+
+ /* Add dram base address */
+ ctx.ret_addr += dram_base_addr;
+
+ /* If legacy MMIO hole enabled */
+ if (lgcy_mmio_hole_en) {
+ if (df_indirect_read_broadcast(nid, 0, 0x104, &ctx.tmp))
+ goto out_err;
+
+ dram_hole_base = ctx.tmp & GENMASK(31, 24);
+ if (ctx.ret_addr >= dram_hole_base)
+ ctx.ret_addr += (BIT_ULL(32) - dram_hole_base);
+ }
+
+ if (hash_enabled) {
+ /* Save some parentheses and grab ls-bit at the end. */
+ hashed_bit = (ctx.ret_addr >> 12) ^
+ (ctx.ret_addr >> 18) ^
+ (ctx.ret_addr >> 21) ^
+ (ctx.ret_addr >> 30) ^
+ cs_id;
+
+ hashed_bit &= BIT(0);
+
+ if (hashed_bit != ((ctx.ret_addr >> intlv_addr_bit) & BIT(0)))
+ ctx.ret_addr ^= BIT(intlv_addr_bit);
+ }
+
+ /* Is calculated system address is above DRAM limit address? */
+ if (ctx.ret_addr > dram_limit_addr)
+ goto out_err;
+
+ *sys_addr = ctx.ret_addr;
+ return 0;
+
+out_err:
+ return -EINVAL;
+}
+
+static int get_channel_from_ecc_syndrome(struct mem_ctl_info *, u16);
+
+/*
+ * Determine if the DIMMs have ECC enabled. ECC is enabled ONLY if all the DIMMs
+ * are ECC capable.
+ */
+static unsigned long dct_determine_edac_cap(struct amd64_pvt *pvt)
+{
+ unsigned long edac_cap = EDAC_FLAG_NONE;
+ u8 bit;
+
+ bit = (pvt->fam > 0xf || pvt->ext_model >= K8_REV_F)
+ ? 19
+ : 17;
+
+ if (pvt->dclr0 & BIT(bit))
+ edac_cap = EDAC_FLAG_SECDED;
+
+ return edac_cap;
+}
+
+static unsigned long umc_determine_edac_cap(struct amd64_pvt *pvt)
+{
+ u8 i, umc_en_mask = 0, dimm_ecc_en_mask = 0;
+ unsigned long edac_cap = EDAC_FLAG_NONE;
+
+ for_each_umc(i) {
+ if (!(pvt->umc[i].sdp_ctrl & UMC_SDP_INIT))
+ continue;
+
+ umc_en_mask |= BIT(i);
+
+ /* UMC Configuration bit 12 (DimmEccEn) */
+ if (pvt->umc[i].umc_cfg & BIT(12))
+ dimm_ecc_en_mask |= BIT(i);
+ }
+
+ if (umc_en_mask == dimm_ecc_en_mask)
+ edac_cap = EDAC_FLAG_SECDED;
+
+ return edac_cap;
+}
+
+/*
+ * debug routine to display the memory sizes of all logical DIMMs and its
+ * CSROWs
+ */
+static void dct_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
+{
+ u32 *dcsb = ctrl ? pvt->csels[1].csbases : pvt->csels[0].csbases;
+ u32 dbam = ctrl ? pvt->dbam1 : pvt->dbam0;
+ int dimm, size0, size1;
+
+ if (pvt->fam == 0xf) {
+ /* K8 families < revF not supported yet */
+ if (pvt->ext_model < K8_REV_F)
+ return;
+
+ WARN_ON(ctrl != 0);
+ }
+
+ if (pvt->fam == 0x10) {
+ dbam = (ctrl && !dct_ganging_enabled(pvt)) ? pvt->dbam1
+ : pvt->dbam0;
+ dcsb = (ctrl && !dct_ganging_enabled(pvt)) ?
+ pvt->csels[1].csbases :
+ pvt->csels[0].csbases;
+ } else if (ctrl) {
+ dbam = pvt->dbam0;
+ dcsb = pvt->csels[1].csbases;
+ }
+ edac_dbg(1, "F2x%d80 (DRAM Bank Address Mapping): 0x%08x\n",
+ ctrl, dbam);
+
+ edac_printk(KERN_DEBUG, EDAC_MC, "DCT%d chip selects:\n", ctrl);
+
+ /* Dump memory sizes for DIMM and its CSROWs */
+ for (dimm = 0; dimm < 4; dimm++) {
+ size0 = 0;
+ if (dcsb[dimm * 2] & DCSB_CS_ENABLE)
+ /*
+ * For F15m60h, we need multiplier for LRDIMM cs_size
+ * calculation. We pass dimm value to the dbam_to_cs
+ * mapper so we can find the multiplier from the
+ * corresponding DCSM.
+ */
+ size0 = pvt->ops->dbam_to_cs(pvt, ctrl,
+ DBAM_DIMM(dimm, dbam),
+ dimm);
+
+ size1 = 0;
+ if (dcsb[dimm * 2 + 1] & DCSB_CS_ENABLE)
+ size1 = pvt->ops->dbam_to_cs(pvt, ctrl,
+ DBAM_DIMM(dimm, dbam),
+ dimm);
+
+ amd64_info(EDAC_MC ": %d: %5dMB %d: %5dMB\n",
+ dimm * 2, size0,
+ dimm * 2 + 1, size1);
+ }
+}
+
+
+static void debug_dump_dramcfg_low(struct amd64_pvt *pvt, u32 dclr, int chan)
+{
+ edac_dbg(1, "F2x%d90 (DRAM Cfg Low): 0x%08x\n", chan, dclr);
+
+ if (pvt->dram_type == MEM_LRDDR3) {
+ u32 dcsm = pvt->csels[chan].csmasks[0];
+ /*
+ * It's assumed all LRDIMMs in a DCT are going to be of
+ * same 'type' until proven otherwise. So, use a cs
+ * value of '0' here to get dcsm value.
+ */
+ edac_dbg(1, " LRDIMM %dx rank multiply\n", (dcsm & 0x3));
+ }
+
+ edac_dbg(1, "All DIMMs support ECC:%s\n",
+ (dclr & BIT(19)) ? "yes" : "no");
+
+
+ edac_dbg(1, " PAR/ERR parity: %s\n",
+ (dclr & BIT(8)) ? "enabled" : "disabled");
+
+ if (pvt->fam == 0x10)
+ edac_dbg(1, " DCT 128bit mode width: %s\n",
+ (dclr & BIT(11)) ? "128b" : "64b");
+
+ edac_dbg(1, " x4 logical DIMMs present: L0: %s L1: %s L2: %s L3: %s\n",
+ (dclr & BIT(12)) ? "yes" : "no",
+ (dclr & BIT(13)) ? "yes" : "no",
+ (dclr & BIT(14)) ? "yes" : "no",
+ (dclr & BIT(15)) ? "yes" : "no");
+}
+
+#define CS_EVEN_PRIMARY BIT(0)
+#define CS_ODD_PRIMARY BIT(1)
+#define CS_EVEN_SECONDARY BIT(2)
+#define CS_ODD_SECONDARY BIT(3)
+#define CS_3R_INTERLEAVE BIT(4)
+
+#define CS_EVEN (CS_EVEN_PRIMARY | CS_EVEN_SECONDARY)
+#define CS_ODD (CS_ODD_PRIMARY | CS_ODD_SECONDARY)
+
+static int umc_get_cs_mode(int dimm, u8 ctrl, struct amd64_pvt *pvt)
+{
+ u8 base, count = 0;
+ int cs_mode = 0;
+
+ if (csrow_enabled(2 * dimm, ctrl, pvt))
+ cs_mode |= CS_EVEN_PRIMARY;
+
+ if (csrow_enabled(2 * dimm + 1, ctrl, pvt))
+ cs_mode |= CS_ODD_PRIMARY;
+
+ /* Asymmetric dual-rank DIMM support. */
+ if (csrow_sec_enabled(2 * dimm + 1, ctrl, pvt))
+ cs_mode |= CS_ODD_SECONDARY;
+
+ /*
+ * 3 Rank inteleaving support.
+ * There should be only three bases enabled and their two masks should
+ * be equal.
+ */
+ for_each_chip_select(base, ctrl, pvt)
+ count += csrow_enabled(base, ctrl, pvt);
+
+ if (count == 3 &&
+ pvt->csels[ctrl].csmasks[0] == pvt->csels[ctrl].csmasks[1]) {
+ edac_dbg(1, "3R interleaving in use.\n");
+ cs_mode |= CS_3R_INTERLEAVE;
+ }
+
+ return cs_mode;
+}
+
+static int __addr_mask_to_cs_size(u32 addr_mask_orig, unsigned int cs_mode,
+ int csrow_nr, int dimm)
+{
+ u32 msb, weight, num_zero_bits;
+ u32 addr_mask_deinterleaved;
+ int size = 0;
+
+ /*
+ * The number of zero bits in the mask is equal to the number of bits
+ * in a full mask minus the number of bits in the current mask.
+ *
+ * The MSB is the number of bits in the full mask because BIT[0] is
+ * always 0.
+ *
+ * In the special 3 Rank interleaving case, a single bit is flipped
+ * without swapping with the most significant bit. This can be handled
+ * by keeping the MSB where it is and ignoring the single zero bit.
+ */
+ msb = fls(addr_mask_orig) - 1;
+ weight = hweight_long(addr_mask_orig);
+ num_zero_bits = msb - weight - !!(cs_mode & CS_3R_INTERLEAVE);
+
+ /* Take the number of zero bits off from the top of the mask. */
+ addr_mask_deinterleaved = GENMASK_ULL(msb - num_zero_bits, 1);
+
+ edac_dbg(1, "CS%d DIMM%d AddrMasks:\n", csrow_nr, dimm);
+ edac_dbg(1, " Original AddrMask: 0x%x\n", addr_mask_orig);
+ edac_dbg(1, " Deinterleaved AddrMask: 0x%x\n", addr_mask_deinterleaved);
+
+ /* Register [31:1] = Address [39:9]. Size is in kBs here. */
+ size = (addr_mask_deinterleaved >> 2) + 1;
+
+ /* Return size in MBs. */
+ return size >> 10;
+}
+
+static int umc_addr_mask_to_cs_size(struct amd64_pvt *pvt, u8 umc,
+ unsigned int cs_mode, int csrow_nr)
+{
+ int cs_mask_nr = csrow_nr;
+ u32 addr_mask_orig;
+ int dimm, size = 0;
+
+ /* No Chip Selects are enabled. */
+ if (!cs_mode)
+ return size;
+
+ /* Requested size of an even CS but none are enabled. */
+ if (!(cs_mode & CS_EVEN) && !(csrow_nr & 1))
+ return size;
+
+ /* Requested size of an odd CS but none are enabled. */
+ if (!(cs_mode & CS_ODD) && (csrow_nr & 1))
+ return size;
+
+ /*
+ * Family 17h introduced systems with one mask per DIMM,
+ * and two Chip Selects per DIMM.
+ *
+ * CS0 and CS1 -> MASK0 / DIMM0
+ * CS2 and CS3 -> MASK1 / DIMM1
+ *
+ * Family 19h Model 10h introduced systems with one mask per Chip Select,
+ * and two Chip Selects per DIMM.
+ *
+ * CS0 -> MASK0 -> DIMM0
+ * CS1 -> MASK1 -> DIMM0
+ * CS2 -> MASK2 -> DIMM1
+ * CS3 -> MASK3 -> DIMM1
+ *
+ * Keep the mask number equal to the Chip Select number for newer systems,
+ * and shift the mask number for older systems.
+ */
+ dimm = csrow_nr >> 1;
+
+ if (!pvt->flags.zn_regs_v2)
+ cs_mask_nr >>= 1;
+
+ /* Asymmetric dual-rank DIMM support. */
+ if ((csrow_nr & 1) && (cs_mode & CS_ODD_SECONDARY))
+ addr_mask_orig = pvt->csels[umc].csmasks_sec[cs_mask_nr];
+ else
+ addr_mask_orig = pvt->csels[umc].csmasks[cs_mask_nr];
+
+ return __addr_mask_to_cs_size(addr_mask_orig, cs_mode, csrow_nr, dimm);
+}
+
+static void umc_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
+{
+ int dimm, size0, size1, cs0, cs1, cs_mode;
+
+ edac_printk(KERN_DEBUG, EDAC_MC, "UMC%d chip selects:\n", ctrl);
+
+ for (dimm = 0; dimm < 2; dimm++) {
+ cs0 = dimm * 2;
+ cs1 = dimm * 2 + 1;
+
+ cs_mode = umc_get_cs_mode(dimm, ctrl, pvt);
+
+ size0 = umc_addr_mask_to_cs_size(pvt, ctrl, cs_mode, cs0);
+ size1 = umc_addr_mask_to_cs_size(pvt, ctrl, cs_mode, cs1);
+
+ amd64_info(EDAC_MC ": %d: %5dMB %d: %5dMB\n",
+ cs0, size0,
+ cs1, size1);
+ }
+}
+
+static void umc_dump_misc_regs(struct amd64_pvt *pvt)
+{
+ struct amd64_umc *umc;
+ u32 i, tmp, umc_base;
+
+ for_each_umc(i) {
+ umc_base = get_umc_base(i);
+ umc = &pvt->umc[i];
+
+ edac_dbg(1, "UMC%d DIMM cfg: 0x%x\n", i, umc->dimm_cfg);
+ edac_dbg(1, "UMC%d UMC cfg: 0x%x\n", i, umc->umc_cfg);
+ edac_dbg(1, "UMC%d SDP ctrl: 0x%x\n", i, umc->sdp_ctrl);
+ edac_dbg(1, "UMC%d ECC ctrl: 0x%x\n", i, umc->ecc_ctrl);
+
+ amd_smn_read(pvt->mc_node_id, umc_base + UMCCH_ECC_BAD_SYMBOL, &tmp);
+ edac_dbg(1, "UMC%d ECC bad symbol: 0x%x\n", i, tmp);
+
+ amd_smn_read(pvt->mc_node_id, umc_base + UMCCH_UMC_CAP, &tmp);
+ edac_dbg(1, "UMC%d UMC cap: 0x%x\n", i, tmp);
+ edac_dbg(1, "UMC%d UMC cap high: 0x%x\n", i, umc->umc_cap_hi);
+
+ edac_dbg(1, "UMC%d ECC capable: %s, ChipKill ECC capable: %s\n",
+ i, (umc->umc_cap_hi & BIT(30)) ? "yes" : "no",
+ (umc->umc_cap_hi & BIT(31)) ? "yes" : "no");
+ edac_dbg(1, "UMC%d All DIMMs support ECC: %s\n",
+ i, (umc->umc_cfg & BIT(12)) ? "yes" : "no");
+ edac_dbg(1, "UMC%d x4 DIMMs present: %s\n",
+ i, (umc->dimm_cfg & BIT(6)) ? "yes" : "no");
+ edac_dbg(1, "UMC%d x16 DIMMs present: %s\n",
+ i, (umc->dimm_cfg & BIT(7)) ? "yes" : "no");
+
+ if (umc->dram_type == MEM_LRDDR4 || umc->dram_type == MEM_LRDDR5) {
+ amd_smn_read(pvt->mc_node_id,
+ umc_base + get_umc_reg(pvt, UMCCH_ADDR_CFG),
+ &tmp);
+ edac_dbg(1, "UMC%d LRDIMM %dx rank multiply\n",
+ i, 1 << ((tmp >> 4) & 0x3));
+ }
+
+ umc_debug_display_dimm_sizes(pvt, i);
+ }
+}
+
+static void dct_dump_misc_regs(struct amd64_pvt *pvt)
+{
+ edac_dbg(1, "F3xE8 (NB Cap): 0x%08x\n", pvt->nbcap);
+
+ edac_dbg(1, " NB two channel DRAM capable: %s\n",
+ (pvt->nbcap & NBCAP_DCT_DUAL) ? "yes" : "no");
+
+ edac_dbg(1, " ECC capable: %s, ChipKill ECC capable: %s\n",
+ (pvt->nbcap & NBCAP_SECDED) ? "yes" : "no",
+ (pvt->nbcap & NBCAP_CHIPKILL) ? "yes" : "no");
+
+ debug_dump_dramcfg_low(pvt, pvt->dclr0, 0);
+
+ edac_dbg(1, "F3xB0 (Online Spare): 0x%08x\n", pvt->online_spare);
+
+ edac_dbg(1, "F1xF0 (DRAM Hole Address): 0x%08x, base: 0x%08x, offset: 0x%08x\n",
+ pvt->dhar, dhar_base(pvt),
+ (pvt->fam == 0xf) ? k8_dhar_offset(pvt)
+ : f10_dhar_offset(pvt));
+
+ dct_debug_display_dimm_sizes(pvt, 0);
+
+ /* everything below this point is Fam10h and above */
+ if (pvt->fam == 0xf)
+ return;
+
+ dct_debug_display_dimm_sizes(pvt, 1);
+
+ /* Only if NOT ganged does dclr1 have valid info */
+ if (!dct_ganging_enabled(pvt))
+ debug_dump_dramcfg_low(pvt, pvt->dclr1, 1);
+
+ edac_dbg(1, " DramHoleValid: %s\n", dhar_valid(pvt) ? "yes" : "no");
+
+ amd64_info("using x%u syndromes.\n", pvt->ecc_sym_sz);
+}
+
+/*
+ * See BKDG, F2x[1,0][5C:40], F2[1,0][6C:60]
+ */
+static void dct_prep_chip_selects(struct amd64_pvt *pvt)
+{
+ if (pvt->fam == 0xf && pvt->ext_model < K8_REV_F) {
+ pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 8;
+ pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 8;
+ } else if (pvt->fam == 0x15 && pvt->model == 0x30) {
+ pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 4;
+ pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 2;
+ } else {
+ pvt->csels[0].b_cnt = pvt->csels[1].b_cnt = 8;
+ pvt->csels[0].m_cnt = pvt->csels[1].m_cnt = 4;
+ }
+}
+
+static void umc_prep_chip_selects(struct amd64_pvt *pvt)
+{
+ int umc;
+
+ for_each_umc(umc) {
+ pvt->csels[umc].b_cnt = 4;
+ pvt->csels[umc].m_cnt = pvt->flags.zn_regs_v2 ? 4 : 2;
+ }
+}
+
+static void umc_read_base_mask(struct amd64_pvt *pvt)
+{
+ u32 umc_base_reg, umc_base_reg_sec;
+ u32 umc_mask_reg, umc_mask_reg_sec;
+ u32 base_reg, base_reg_sec;
+ u32 mask_reg, mask_reg_sec;
+ u32 *base, *base_sec;
+ u32 *mask, *mask_sec;
+ int cs, umc;
+
+ for_each_umc(umc) {
+ umc_base_reg = get_umc_base(umc) + UMCCH_BASE_ADDR;
+ umc_base_reg_sec = get_umc_base(umc) + UMCCH_BASE_ADDR_SEC;
+
+ for_each_chip_select(cs, umc, pvt) {
+ base = &pvt->csels[umc].csbases[cs];
+ base_sec = &pvt->csels[umc].csbases_sec[cs];
+
+ base_reg = umc_base_reg + (cs * 4);
+ base_reg_sec = umc_base_reg_sec + (cs * 4);
+
+ if (!amd_smn_read(pvt->mc_node_id, base_reg, base))
+ edac_dbg(0, " DCSB%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *base, base_reg);
+
+ if (!amd_smn_read(pvt->mc_node_id, base_reg_sec, base_sec))
+ edac_dbg(0, " DCSB_SEC%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *base_sec, base_reg_sec);
+ }
+
+ umc_mask_reg = get_umc_base(umc) + UMCCH_ADDR_MASK;
+ umc_mask_reg_sec = get_umc_base(umc) + get_umc_reg(pvt, UMCCH_ADDR_MASK_SEC);
+
+ for_each_chip_select_mask(cs, umc, pvt) {
+ mask = &pvt->csels[umc].csmasks[cs];
+ mask_sec = &pvt->csels[umc].csmasks_sec[cs];
+
+ mask_reg = umc_mask_reg + (cs * 4);
+ mask_reg_sec = umc_mask_reg_sec + (cs * 4);
+
+ if (!amd_smn_read(pvt->mc_node_id, mask_reg, mask))
+ edac_dbg(0, " DCSM%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *mask, mask_reg);
+
+ if (!amd_smn_read(pvt->mc_node_id, mask_reg_sec, mask_sec))
+ edac_dbg(0, " DCSM_SEC%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *mask_sec, mask_reg_sec);
+ }
+ }
+}
+
+/*
+ * Function 2 Offset F10_DCSB0; read in the DCS Base and DCS Mask registers
+ */
+static void dct_read_base_mask(struct amd64_pvt *pvt)
+{
+ int cs;
+
+ for_each_chip_select(cs, 0, pvt) {
+ int reg0 = DCSB0 + (cs * 4);
+ int reg1 = DCSB1 + (cs * 4);
+ u32 *base0 = &pvt->csels[0].csbases[cs];
+ u32 *base1 = &pvt->csels[1].csbases[cs];
+
+ if (!amd64_read_dct_pci_cfg(pvt, 0, reg0, base0))
+ edac_dbg(0, " DCSB0[%d]=0x%08x reg: F2x%x\n",
+ cs, *base0, reg0);
+
+ if (pvt->fam == 0xf)
+ continue;
+
+ if (!amd64_read_dct_pci_cfg(pvt, 1, reg0, base1))
+ edac_dbg(0, " DCSB1[%d]=0x%08x reg: F2x%x\n",
+ cs, *base1, (pvt->fam == 0x10) ? reg1
+ : reg0);
+ }
+
+ for_each_chip_select_mask(cs, 0, pvt) {
+ int reg0 = DCSM0 + (cs * 4);
+ int reg1 = DCSM1 + (cs * 4);
+ u32 *mask0 = &pvt->csels[0].csmasks[cs];
+ u32 *mask1 = &pvt->csels[1].csmasks[cs];
+
+ if (!amd64_read_dct_pci_cfg(pvt, 0, reg0, mask0))
+ edac_dbg(0, " DCSM0[%d]=0x%08x reg: F2x%x\n",
+ cs, *mask0, reg0);
+
+ if (pvt->fam == 0xf)
+ continue;
+
+ if (!amd64_read_dct_pci_cfg(pvt, 1, reg0, mask1))
+ edac_dbg(0, " DCSM1[%d]=0x%08x reg: F2x%x\n",
+ cs, *mask1, (pvt->fam == 0x10) ? reg1
+ : reg0);
+ }
+}
+
+static void umc_determine_memory_type(struct amd64_pvt *pvt)
+{
+ struct amd64_umc *umc;
+ u32 i;
+
+ for_each_umc(i) {
+ umc = &pvt->umc[i];
+
+ if (!(umc->sdp_ctrl & UMC_SDP_INIT)) {
+ umc->dram_type = MEM_EMPTY;
+ continue;
+ }
+
+ /*
+ * Check if the system supports the "DDR Type" field in UMC Config
+ * and has DDR5 DIMMs in use.
+ */
+ if (pvt->flags.zn_regs_v2 && ((umc->umc_cfg & GENMASK(2, 0)) == 0x1)) {
+ if (umc->dimm_cfg & BIT(5))
+ umc->dram_type = MEM_LRDDR5;
+ else if (umc->dimm_cfg & BIT(4))
+ umc->dram_type = MEM_RDDR5;
+ else
+ umc->dram_type = MEM_DDR5;
+ } else {
+ if (umc->dimm_cfg & BIT(5))
+ umc->dram_type = MEM_LRDDR4;
+ else if (umc->dimm_cfg & BIT(4))
+ umc->dram_type = MEM_RDDR4;
+ else
+ umc->dram_type = MEM_DDR4;
+ }
+
+ edac_dbg(1, " UMC%d DIMM type: %s\n", i, edac_mem_types[umc->dram_type]);
+ }
+}
+
+static void dct_determine_memory_type(struct amd64_pvt *pvt)
+{
+ u32 dram_ctrl, dcsm;
+
+ switch (pvt->fam) {
+ case 0xf:
+ if (pvt->ext_model >= K8_REV_F)
+ goto ddr3;
+
+ pvt->dram_type = (pvt->dclr0 & BIT(18)) ? MEM_DDR : MEM_RDDR;
+ return;
+
+ case 0x10:
+ if (pvt->dchr0 & DDR3_MODE)
+ goto ddr3;
+
+ pvt->dram_type = (pvt->dclr0 & BIT(16)) ? MEM_DDR2 : MEM_RDDR2;
+ return;
+
+ case 0x15:
+ if (pvt->model < 0x60)
+ goto ddr3;
+
+ /*
+ * Model 0x60h needs special handling:
+ *
+ * We use a Chip Select value of '0' to obtain dcsm.
+ * Theoretically, it is possible to populate LRDIMMs of different
+ * 'Rank' value on a DCT. But this is not the common case. So,
+ * it's reasonable to assume all DIMMs are going to be of same
+ * 'type' until proven otherwise.
+ */
+ amd64_read_dct_pci_cfg(pvt, 0, DRAM_CONTROL, &dram_ctrl);
+ dcsm = pvt->csels[0].csmasks[0];
+
+ if (((dram_ctrl >> 8) & 0x7) == 0x2)
+ pvt->dram_type = MEM_DDR4;
+ else if (pvt->dclr0 & BIT(16))
+ pvt->dram_type = MEM_DDR3;
+ else if (dcsm & 0x3)
+ pvt->dram_type = MEM_LRDDR3;
+ else
+ pvt->dram_type = MEM_RDDR3;
+
+ return;
+
+ case 0x16:
+ goto ddr3;
+
+ default:
+ WARN(1, KERN_ERR "%s: Family??? 0x%x\n", __func__, pvt->fam);
+ pvt->dram_type = MEM_EMPTY;
+ }
+
+ edac_dbg(1, " DIMM type: %s\n", edac_mem_types[pvt->dram_type]);
+ return;
+
+ddr3:
+ pvt->dram_type = (pvt->dclr0 & BIT(16)) ? MEM_DDR3 : MEM_RDDR3;
+}
+
+/* On F10h and later ErrAddr is MC4_ADDR[47:1] */
+static u64 get_error_address(struct amd64_pvt *pvt, struct mce *m)
+{
+ u16 mce_nid = topology_die_id(m->extcpu);
+ struct mem_ctl_info *mci;
+ u8 start_bit = 1;
+ u8 end_bit = 47;
+ u64 addr;
+
+ mci = edac_mc_find(mce_nid);
+ if (!mci)
+ return 0;
+
+ pvt = mci->pvt_info;
+
+ if (pvt->fam == 0xf) {
+ start_bit = 3;
+ end_bit = 39;
+ }
+
+ addr = m->addr & GENMASK_ULL(end_bit, start_bit);
+
+ /*
+ * Erratum 637 workaround
+ */
+ if (pvt->fam == 0x15) {
+ u64 cc6_base, tmp_addr;
+ u32 tmp;
+ u8 intlv_en;
+
+ if ((addr & GENMASK_ULL(47, 24)) >> 24 != 0x00fdf7)
+ return addr;
+
+
+ amd64_read_pci_cfg(pvt->F1, DRAM_LOCAL_NODE_LIM, &tmp);
+ intlv_en = tmp >> 21 & 0x7;
+
+ /* add [47:27] + 3 trailing bits */
+ cc6_base = (tmp & GENMASK_ULL(20, 0)) << 3;
+
+ /* reverse and add DramIntlvEn */
+ cc6_base |= intlv_en ^ 0x7;
+
+ /* pin at [47:24] */
+ cc6_base <<= 24;
+
+ if (!intlv_en)
+ return cc6_base | (addr & GENMASK_ULL(23, 0));
+
+ amd64_read_pci_cfg(pvt->F1, DRAM_LOCAL_NODE_BASE, &tmp);
+
+ /* faster log2 */
+ tmp_addr = (addr & GENMASK_ULL(23, 12)) << __fls(intlv_en + 1);
+
+ /* OR DramIntlvSel into bits [14:12] */
+ tmp_addr |= (tmp & GENMASK_ULL(23, 21)) >> 9;
+
+ /* add remaining [11:0] bits from original MC4_ADDR */
+ tmp_addr |= addr & GENMASK_ULL(11, 0);
+
+ return cc6_base | tmp_addr;
+ }
+
+ return addr;
+}
+
+static struct pci_dev *pci_get_related_function(unsigned int vendor,
+ unsigned int device,
+ struct pci_dev *related)
+{
+ struct pci_dev *dev = NULL;
+
+ while ((dev = pci_get_device(vendor, device, dev))) {
+ if (pci_domain_nr(dev->bus) == pci_domain_nr(related->bus) &&
+ (dev->bus->number == related->bus->number) &&
+ (PCI_SLOT(dev->devfn) == PCI_SLOT(related->devfn)))
+ break;
+ }
+
+ return dev;
+}
+
+static void read_dram_base_limit_regs(struct amd64_pvt *pvt, unsigned range)
+{
+ struct amd_northbridge *nb;
+ struct pci_dev *f1 = NULL;
+ unsigned int pci_func;
+ int off = range << 3;
+ u32 llim;
+
+ amd64_read_pci_cfg(pvt->F1, DRAM_BASE_LO + off, &pvt->ranges[range].base.lo);
+ amd64_read_pci_cfg(pvt->F1, DRAM_LIMIT_LO + off, &pvt->ranges[range].lim.lo);
+
+ if (pvt->fam == 0xf)
+ return;
+
+ if (!dram_rw(pvt, range))
+ return;
+
+ amd64_read_pci_cfg(pvt->F1, DRAM_BASE_HI + off, &pvt->ranges[range].base.hi);
+ amd64_read_pci_cfg(pvt->F1, DRAM_LIMIT_HI + off, &pvt->ranges[range].lim.hi);
+
+ /* F15h: factor in CC6 save area by reading dst node's limit reg */
+ if (pvt->fam != 0x15)
+ return;
+
+ nb = node_to_amd_nb(dram_dst_node(pvt, range));
+ if (WARN_ON(!nb))
+ return;
+
+ if (pvt->model == 0x60)
+ pci_func = PCI_DEVICE_ID_AMD_15H_M60H_NB_F1;
+ else if (pvt->model == 0x30)
+ pci_func = PCI_DEVICE_ID_AMD_15H_M30H_NB_F1;
+ else
+ pci_func = PCI_DEVICE_ID_AMD_15H_NB_F1;
+
+ f1 = pci_get_related_function(nb->misc->vendor, pci_func, nb->misc);
+ if (WARN_ON(!f1))
+ return;
+
+ amd64_read_pci_cfg(f1, DRAM_LOCAL_NODE_LIM, &llim);
+
+ pvt->ranges[range].lim.lo &= GENMASK_ULL(15, 0);
+
+ /* {[39:27],111b} */
+ pvt->ranges[range].lim.lo |= ((llim & 0x1fff) << 3 | 0x7) << 16;
+
+ pvt->ranges[range].lim.hi &= GENMASK_ULL(7, 0);
+
+ /* [47:40] */
+ pvt->ranges[range].lim.hi |= llim >> 13;
+
+ pci_dev_put(f1);
+}
+
+static void k8_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
+ struct err_info *err)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ error_address_to_page_and_offset(sys_addr, err);
+
+ /*
+ * Find out which node the error address belongs to. This may be
+ * different from the node that detected the error.
+ */
+ err->src_mci = find_mc_by_sys_addr(mci, sys_addr);
+ if (!err->src_mci) {
+ amd64_mc_err(mci, "failed to map error addr 0x%lx to a node\n",
+ (unsigned long)sys_addr);
+ err->err_code = ERR_NODE;
+ return;
+ }
+
+ /* Now map the sys_addr to a CSROW */
+ err->csrow = sys_addr_to_csrow(err->src_mci, sys_addr);
+ if (err->csrow < 0) {
+ err->err_code = ERR_CSROW;
+ return;
+ }
+
+ /* CHIPKILL enabled */
+ if (pvt->nbcfg & NBCFG_CHIPKILL) {
+ err->channel = get_channel_from_ecc_syndrome(mci, err->syndrome);
+ if (err->channel < 0) {
+ /*
+ * Syndrome didn't map, so we don't know which of the
+ * 2 DIMMs is in error. So we need to ID 'both' of them
+ * as suspect.
+ */
+ amd64_mc_warn(err->src_mci, "unknown syndrome 0x%04x - "
+ "possible error reporting race\n",
+ err->syndrome);
+ err->err_code = ERR_CHANNEL;
+ return;
+ }
+ } else {
+ /*
+ * non-chipkill ecc mode
+ *
+ * The k8 documentation is unclear about how to determine the
+ * channel number when using non-chipkill memory. This method
+ * was obtained from email communication with someone at AMD.
+ * (Wish the email was placed in this comment - norsk)
+ */
+ err->channel = ((sys_addr & BIT(3)) != 0);
+ }
+}
+
+static int ddr2_cs_size(unsigned i, bool dct_width)
+{
+ unsigned shift = 0;
+
+ if (i <= 2)
+ shift = i;
+ else if (!(i & 0x1))
+ shift = i >> 1;
+ else
+ shift = (i + 1) >> 1;
+
+ return 128 << (shift + !!dct_width);
+}
+
+static int k8_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
+ unsigned cs_mode, int cs_mask_nr)
+{
+ u32 dclr = dct ? pvt->dclr1 : pvt->dclr0;
+
+ if (pvt->ext_model >= K8_REV_F) {
+ WARN_ON(cs_mode > 11);
+ return ddr2_cs_size(cs_mode, dclr & WIDTH_128);
+ }
+ else if (pvt->ext_model >= K8_REV_D) {
+ unsigned diff;
+ WARN_ON(cs_mode > 10);
+
+ /*
+ * the below calculation, besides trying to win an obfuscated C
+ * contest, maps cs_mode values to DIMM chip select sizes. The
+ * mappings are:
+ *
+ * cs_mode CS size (mb)
+ * ======= ============
+ * 0 32
+ * 1 64
+ * 2 128
+ * 3 128
+ * 4 256
+ * 5 512
+ * 6 256
+ * 7 512
+ * 8 1024
+ * 9 1024
+ * 10 2048
+ *
+ * Basically, it calculates a value with which to shift the
+ * smallest CS size of 32MB.
+ *
+ * ddr[23]_cs_size have a similar purpose.
+ */
+ diff = cs_mode/3 + (unsigned)(cs_mode > 5);
+
+ return 32 << (cs_mode - diff);
+ }
+ else {
+ WARN_ON(cs_mode > 6);
+ return 32 << cs_mode;
+ }
+}
+
+static int ddr3_cs_size(unsigned i, bool dct_width)
+{
+ unsigned shift = 0;
+ int cs_size = 0;
+
+ if (i == 0 || i == 3 || i == 4)
+ cs_size = -1;
+ else if (i <= 2)
+ shift = i;
+ else if (i == 12)
+ shift = 7;
+ else if (!(i & 0x1))
+ shift = i >> 1;
+ else
+ shift = (i + 1) >> 1;
+
+ if (cs_size != -1)
+ cs_size = (128 * (1 << !!dct_width)) << shift;
+
+ return cs_size;
+}
+
+static int ddr3_lrdimm_cs_size(unsigned i, unsigned rank_multiply)
+{
+ unsigned shift = 0;
+ int cs_size = 0;
+
+ if (i < 4 || i == 6)
+ cs_size = -1;
+ else if (i == 12)
+ shift = 7;
+ else if (!(i & 0x1))
+ shift = i >> 1;
+ else
+ shift = (i + 1) >> 1;
+
+ if (cs_size != -1)
+ cs_size = rank_multiply * (128 << shift);
+
+ return cs_size;
+}
+
+static int ddr4_cs_size(unsigned i)
+{
+ int cs_size = 0;
+
+ if (i == 0)
+ cs_size = -1;
+ else if (i == 1)
+ cs_size = 1024;
+ else
+ /* Min cs_size = 1G */
+ cs_size = 1024 * (1 << (i >> 1));
+
+ return cs_size;
+}
+
+static int f10_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
+ unsigned cs_mode, int cs_mask_nr)
+{
+ u32 dclr = dct ? pvt->dclr1 : pvt->dclr0;
+
+ WARN_ON(cs_mode > 11);
+
+ if (pvt->dchr0 & DDR3_MODE || pvt->dchr1 & DDR3_MODE)
+ return ddr3_cs_size(cs_mode, dclr & WIDTH_128);
+ else
+ return ddr2_cs_size(cs_mode, dclr & WIDTH_128);
+}
+
+/*
+ * F15h supports only 64bit DCT interfaces
+ */
+static int f15_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
+ unsigned cs_mode, int cs_mask_nr)
+{
+ WARN_ON(cs_mode > 12);
+
+ return ddr3_cs_size(cs_mode, false);
+}
+
+/* F15h M60h supports DDR4 mapping as well.. */
+static int f15_m60h_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
+ unsigned cs_mode, int cs_mask_nr)
+{
+ int cs_size;
+ u32 dcsm = pvt->csels[dct].csmasks[cs_mask_nr];
+
+ WARN_ON(cs_mode > 12);
+
+ if (pvt->dram_type == MEM_DDR4) {
+ if (cs_mode > 9)
+ return -1;
+
+ cs_size = ddr4_cs_size(cs_mode);
+ } else if (pvt->dram_type == MEM_LRDDR3) {
+ unsigned rank_multiply = dcsm & 0xf;
+
+ if (rank_multiply == 3)
+ rank_multiply = 4;
+ cs_size = ddr3_lrdimm_cs_size(cs_mode, rank_multiply);
+ } else {
+ /* Minimum cs size is 512mb for F15hM60h*/
+ if (cs_mode == 0x1)
+ return -1;
+
+ cs_size = ddr3_cs_size(cs_mode, false);
+ }
+
+ return cs_size;
+}
+
+/*
+ * F16h and F15h model 30h have only limited cs_modes.
+ */
+static int f16_dbam_to_chip_select(struct amd64_pvt *pvt, u8 dct,
+ unsigned cs_mode, int cs_mask_nr)
+{
+ WARN_ON(cs_mode > 12);
+
+ if (cs_mode == 6 || cs_mode == 8 ||
+ cs_mode == 9 || cs_mode == 12)
+ return -1;
+ else
+ return ddr3_cs_size(cs_mode, false);
+}
+
+static void read_dram_ctl_register(struct amd64_pvt *pvt)
+{
+
+ if (pvt->fam == 0xf)
+ return;
+
+ if (!amd64_read_pci_cfg(pvt->F2, DCT_SEL_LO, &pvt->dct_sel_lo)) {
+ edac_dbg(0, "F2x110 (DCTSelLow): 0x%08x, High range addrs at: 0x%x\n",
+ pvt->dct_sel_lo, dct_sel_baseaddr(pvt));
+
+ edac_dbg(0, " DCTs operate in %s mode\n",
+ (dct_ganging_enabled(pvt) ? "ganged" : "unganged"));
+
+ if (!dct_ganging_enabled(pvt))
+ edac_dbg(0, " Address range split per DCT: %s\n",
+ (dct_high_range_enabled(pvt) ? "yes" : "no"));
+
+ edac_dbg(0, " data interleave for ECC: %s, DRAM cleared since last warm reset: %s\n",
+ (dct_data_intlv_enabled(pvt) ? "enabled" : "disabled"),
+ (dct_memory_cleared(pvt) ? "yes" : "no"));
+
+ edac_dbg(0, " channel interleave: %s, "
+ "interleave bits selector: 0x%x\n",
+ (dct_interleave_enabled(pvt) ? "enabled" : "disabled"),
+ dct_sel_interleave_addr(pvt));
+ }
+
+ amd64_read_pci_cfg(pvt->F2, DCT_SEL_HI, &pvt->dct_sel_hi);
+}
+
+/*
+ * Determine channel (DCT) based on the interleaving mode (see F15h M30h BKDG,
+ * 2.10.12 Memory Interleaving Modes).
+ */
+static u8 f15_m30h_determine_channel(struct amd64_pvt *pvt, u64 sys_addr,
+ u8 intlv_en, int num_dcts_intlv,
+ u32 dct_sel)
+{
+ u8 channel = 0;
+ u8 select;
+
+ if (!(intlv_en))
+ return (u8)(dct_sel);
+
+ if (num_dcts_intlv == 2) {
+ select = (sys_addr >> 8) & 0x3;
+ channel = select ? 0x3 : 0;
+ } else if (num_dcts_intlv == 4) {
+ u8 intlv_addr = dct_sel_interleave_addr(pvt);
+ switch (intlv_addr) {
+ case 0x4:
+ channel = (sys_addr >> 8) & 0x3;
+ break;
+ case 0x5:
+ channel = (sys_addr >> 9) & 0x3;
+ break;
+ }
+ }
+ return channel;
+}
+
+/*
+ * Determine channel (DCT) based on the interleaving mode: F10h BKDG, 2.8.9 Memory
+ * Interleaving Modes.
+ */
+static u8 f1x_determine_channel(struct amd64_pvt *pvt, u64 sys_addr,
+ bool hi_range_sel, u8 intlv_en)
+{
+ u8 dct_sel_high = (pvt->dct_sel_lo >> 1) & 1;
+
+ if (dct_ganging_enabled(pvt))
+ return 0;
+
+ if (hi_range_sel)
+ return dct_sel_high;
+
+ /*
+ * see F2x110[DctSelIntLvAddr] - channel interleave mode
+ */
+ if (dct_interleave_enabled(pvt)) {
+ u8 intlv_addr = dct_sel_interleave_addr(pvt);
+
+ /* return DCT select function: 0=DCT0, 1=DCT1 */
+ if (!intlv_addr)
+ return sys_addr >> 6 & 1;
+
+ if (intlv_addr & 0x2) {
+ u8 shift = intlv_addr & 0x1 ? 9 : 6;
+ u32 temp = hweight_long((u32) ((sys_addr >> 16) & 0x1F)) & 1;
+
+ return ((sys_addr >> shift) & 1) ^ temp;
+ }
+
+ if (intlv_addr & 0x4) {
+ u8 shift = intlv_addr & 0x1 ? 9 : 8;
+
+ return (sys_addr >> shift) & 1;
+ }
+
+ return (sys_addr >> (12 + hweight8(intlv_en))) & 1;
+ }
+
+ if (dct_high_range_enabled(pvt))
+ return ~dct_sel_high & 1;
+
+ return 0;
+}
+
+/* Convert the sys_addr to the normalized DCT address */
+static u64 f1x_get_norm_dct_addr(struct amd64_pvt *pvt, u8 range,
+ u64 sys_addr, bool hi_rng,
+ u32 dct_sel_base_addr)
+{
+ u64 chan_off;
+ u64 dram_base = get_dram_base(pvt, range);
+ u64 hole_off = f10_dhar_offset(pvt);
+ u64 dct_sel_base_off = (u64)(pvt->dct_sel_hi & 0xFFFFFC00) << 16;
+
+ if (hi_rng) {
+ /*
+ * if
+ * base address of high range is below 4Gb
+ * (bits [47:27] at [31:11])
+ * DRAM address space on this DCT is hoisted above 4Gb &&
+ * sys_addr > 4Gb
+ *
+ * remove hole offset from sys_addr
+ * else
+ * remove high range offset from sys_addr
+ */
+ if ((!(dct_sel_base_addr >> 16) ||
+ dct_sel_base_addr < dhar_base(pvt)) &&
+ dhar_valid(pvt) &&
+ (sys_addr >= BIT_64(32)))
+ chan_off = hole_off;
+ else
+ chan_off = dct_sel_base_off;
+ } else {
+ /*
+ * if
+ * we have a valid hole &&
+ * sys_addr > 4Gb
+ *
+ * remove hole
+ * else
+ * remove dram base to normalize to DCT address
+ */
+ if (dhar_valid(pvt) && (sys_addr >= BIT_64(32)))
+ chan_off = hole_off;
+ else
+ chan_off = dram_base;
+ }
+
+ return (sys_addr & GENMASK_ULL(47,6)) - (chan_off & GENMASK_ULL(47,23));
+}
+
+/*
+ * checks if the csrow passed in is marked as SPARED, if so returns the new
+ * spare row
+ */
+static int f10_process_possible_spare(struct amd64_pvt *pvt, u8 dct, int csrow)
+{
+ int tmp_cs;
+
+ if (online_spare_swap_done(pvt, dct) &&
+ csrow == online_spare_bad_dramcs(pvt, dct)) {
+
+ for_each_chip_select(tmp_cs, dct, pvt) {
+ if (chip_select_base(tmp_cs, dct, pvt) & 0x2) {
+ csrow = tmp_cs;
+ break;
+ }
+ }
+ }
+ return csrow;
+}
+
+/*
+ * Iterate over the DRAM DCT "base" and "mask" registers looking for a
+ * SystemAddr match on the specified 'ChannelSelect' and 'NodeID'
+ *
+ * Return:
+ * -EINVAL: NOT FOUND
+ * 0..csrow = Chip-Select Row
+ */
+static int f1x_lookup_addr_in_dct(u64 in_addr, u8 nid, u8 dct)
+{
+ struct mem_ctl_info *mci;
+ struct amd64_pvt *pvt;
+ u64 cs_base, cs_mask;
+ int cs_found = -EINVAL;
+ int csrow;
+
+ mci = edac_mc_find(nid);
+ if (!mci)
+ return cs_found;
+
+ pvt = mci->pvt_info;
+
+ edac_dbg(1, "input addr: 0x%llx, DCT: %d\n", in_addr, dct);
+
+ for_each_chip_select(csrow, dct, pvt) {
+ if (!csrow_enabled(csrow, dct, pvt))
+ continue;
+
+ get_cs_base_and_mask(pvt, csrow, dct, &cs_base, &cs_mask);
+
+ edac_dbg(1, " CSROW=%d CSBase=0x%llx CSMask=0x%llx\n",
+ csrow, cs_base, cs_mask);
+
+ cs_mask = ~cs_mask;
+
+ edac_dbg(1, " (InputAddr & ~CSMask)=0x%llx (CSBase & ~CSMask)=0x%llx\n",
+ (in_addr & cs_mask), (cs_base & cs_mask));
+
+ if ((in_addr & cs_mask) == (cs_base & cs_mask)) {
+ if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+ cs_found = csrow;
+ break;
+ }
+ cs_found = f10_process_possible_spare(pvt, dct, csrow);
+
+ edac_dbg(1, " MATCH csrow=%d\n", cs_found);
+ break;
+ }
+ }
+ return cs_found;
+}
+
+/*
+ * See F2x10C. Non-interleaved graphics framebuffer memory under the 16G is
+ * swapped with a region located at the bottom of memory so that the GPU can use
+ * the interleaved region and thus two channels.
+ */
+static u64 f1x_swap_interleaved_region(struct amd64_pvt *pvt, u64 sys_addr)
+{
+ u32 swap_reg, swap_base, swap_limit, rgn_size, tmp_addr;
+
+ if (pvt->fam == 0x10) {
+ /* only revC3 and revE have that feature */
+ if (pvt->model < 4 || (pvt->model < 0xa && pvt->stepping < 3))
+ return sys_addr;
+ }
+
+ amd64_read_pci_cfg(pvt->F2, SWAP_INTLV_REG, &swap_reg);
+
+ if (!(swap_reg & 0x1))
+ return sys_addr;
+
+ swap_base = (swap_reg >> 3) & 0x7f;
+ swap_limit = (swap_reg >> 11) & 0x7f;
+ rgn_size = (swap_reg >> 20) & 0x7f;
+ tmp_addr = sys_addr >> 27;
+
+ if (!(sys_addr >> 34) &&
+ (((tmp_addr >= swap_base) &&
+ (tmp_addr <= swap_limit)) ||
+ (tmp_addr < rgn_size)))
+ return sys_addr ^ (u64)swap_base << 27;
+
+ return sys_addr;
+}
+
+/* For a given @dram_range, check if @sys_addr falls within it. */
+static int f1x_match_to_this_node(struct amd64_pvt *pvt, unsigned range,
+ u64 sys_addr, int *chan_sel)
+{
+ int cs_found = -EINVAL;
+ u64 chan_addr;
+ u32 dct_sel_base;
+ u8 channel;
+ bool high_range = false;
+
+ u8 node_id = dram_dst_node(pvt, range);
+ u8 intlv_en = dram_intlv_en(pvt, range);
+ u32 intlv_sel = dram_intlv_sel(pvt, range);
+
+ edac_dbg(1, "(range %d) SystemAddr= 0x%llx Limit=0x%llx\n",
+ range, sys_addr, get_dram_limit(pvt, range));
+
+ if (dhar_valid(pvt) &&
+ dhar_base(pvt) <= sys_addr &&
+ sys_addr < BIT_64(32)) {
+ amd64_warn("Huh? Address is in the MMIO hole: 0x%016llx\n",
+ sys_addr);
+ return -EINVAL;
+ }
+
+ if (intlv_en && (intlv_sel != ((sys_addr >> 12) & intlv_en)))
+ return -EINVAL;
+
+ sys_addr = f1x_swap_interleaved_region(pvt, sys_addr);
+
+ dct_sel_base = dct_sel_baseaddr(pvt);
+
+ /*
+ * check whether addresses >= DctSelBaseAddr[47:27] are to be used to
+ * select between DCT0 and DCT1.
+ */
+ if (dct_high_range_enabled(pvt) &&
+ !dct_ganging_enabled(pvt) &&
+ ((sys_addr >> 27) >= (dct_sel_base >> 11)))
+ high_range = true;
+
+ channel = f1x_determine_channel(pvt, sys_addr, high_range, intlv_en);
+
+ chan_addr = f1x_get_norm_dct_addr(pvt, range, sys_addr,
+ high_range, dct_sel_base);
+
+ /* Remove node interleaving, see F1x120 */
+ if (intlv_en)
+ chan_addr = ((chan_addr >> (12 + hweight8(intlv_en))) << 12) |
+ (chan_addr & 0xfff);
+
+ /* remove channel interleave */
+ if (dct_interleave_enabled(pvt) &&
+ !dct_high_range_enabled(pvt) &&
+ !dct_ganging_enabled(pvt)) {
+
+ if (dct_sel_interleave_addr(pvt) != 1) {
+ if (dct_sel_interleave_addr(pvt) == 0x3)
+ /* hash 9 */
+ chan_addr = ((chan_addr >> 10) << 9) |
+ (chan_addr & 0x1ff);
+ else
+ /* A[6] or hash 6 */
+ chan_addr = ((chan_addr >> 7) << 6) |
+ (chan_addr & 0x3f);
+ } else
+ /* A[12] */
+ chan_addr = ((chan_addr >> 13) << 12) |
+ (chan_addr & 0xfff);
+ }
+
+ edac_dbg(1, " Normalized DCT addr: 0x%llx\n", chan_addr);
+
+ cs_found = f1x_lookup_addr_in_dct(chan_addr, node_id, channel);
+
+ if (cs_found >= 0)
+ *chan_sel = channel;
+
+ return cs_found;
+}
+
+static int f15_m30h_match_to_this_node(struct amd64_pvt *pvt, unsigned range,
+ u64 sys_addr, int *chan_sel)
+{
+ int cs_found = -EINVAL;
+ int num_dcts_intlv = 0;
+ u64 chan_addr, chan_offset;
+ u64 dct_base, dct_limit;
+ u32 dct_cont_base_reg, dct_cont_limit_reg, tmp;
+ u8 channel, alias_channel, leg_mmio_hole, dct_sel, dct_offset_en;
+
+ u64 dhar_offset = f10_dhar_offset(pvt);
+ u8 intlv_addr = dct_sel_interleave_addr(pvt);
+ u8 node_id = dram_dst_node(pvt, range);
+ u8 intlv_en = dram_intlv_en(pvt, range);
+
+ amd64_read_pci_cfg(pvt->F1, DRAM_CONT_BASE, &dct_cont_base_reg);
+ amd64_read_pci_cfg(pvt->F1, DRAM_CONT_LIMIT, &dct_cont_limit_reg);
+
+ dct_offset_en = (u8) ((dct_cont_base_reg >> 3) & BIT(0));
+ dct_sel = (u8) ((dct_cont_base_reg >> 4) & 0x7);
+
+ edac_dbg(1, "(range %d) SystemAddr= 0x%llx Limit=0x%llx\n",
+ range, sys_addr, get_dram_limit(pvt, range));
+
+ if (!(get_dram_base(pvt, range) <= sys_addr) &&
+ !(get_dram_limit(pvt, range) >= sys_addr))
+ return -EINVAL;
+
+ if (dhar_valid(pvt) &&
+ dhar_base(pvt) <= sys_addr &&
+ sys_addr < BIT_64(32)) {
+ amd64_warn("Huh? Address is in the MMIO hole: 0x%016llx\n",
+ sys_addr);
+ return -EINVAL;
+ }
+
+ /* Verify sys_addr is within DCT Range. */
+ dct_base = (u64) dct_sel_baseaddr(pvt);
+ dct_limit = (dct_cont_limit_reg >> 11) & 0x1FFF;
+
+ if (!(dct_cont_base_reg & BIT(0)) &&
+ !(dct_base <= (sys_addr >> 27) &&
+ dct_limit >= (sys_addr >> 27)))
+ return -EINVAL;
+
+ /* Verify number of dct's that participate in channel interleaving. */
+ num_dcts_intlv = (int) hweight8(intlv_en);
+
+ if (!(num_dcts_intlv % 2 == 0) || (num_dcts_intlv > 4))
+ return -EINVAL;
+
+ if (pvt->model >= 0x60)
+ channel = f1x_determine_channel(pvt, sys_addr, false, intlv_en);
+ else
+ channel = f15_m30h_determine_channel(pvt, sys_addr, intlv_en,
+ num_dcts_intlv, dct_sel);
+
+ /* Verify we stay within the MAX number of channels allowed */
+ if (channel > 3)
+ return -EINVAL;
+
+ leg_mmio_hole = (u8) (dct_cont_base_reg >> 1 & BIT(0));
+
+ /* Get normalized DCT addr */
+ if (leg_mmio_hole && (sys_addr >= BIT_64(32)))
+ chan_offset = dhar_offset;
+ else
+ chan_offset = dct_base << 27;
+
+ chan_addr = sys_addr - chan_offset;
+
+ /* remove channel interleave */
+ if (num_dcts_intlv == 2) {
+ if (intlv_addr == 0x4)
+ chan_addr = ((chan_addr >> 9) << 8) |
+ (chan_addr & 0xff);
+ else if (intlv_addr == 0x5)
+ chan_addr = ((chan_addr >> 10) << 9) |
+ (chan_addr & 0x1ff);
+ else
+ return -EINVAL;
+
+ } else if (num_dcts_intlv == 4) {
+ if (intlv_addr == 0x4)
+ chan_addr = ((chan_addr >> 10) << 8) |
+ (chan_addr & 0xff);
+ else if (intlv_addr == 0x5)
+ chan_addr = ((chan_addr >> 11) << 9) |
+ (chan_addr & 0x1ff);
+ else
+ return -EINVAL;
+ }
+
+ if (dct_offset_en) {
+ amd64_read_pci_cfg(pvt->F1,
+ DRAM_CONT_HIGH_OFF + (int) channel * 4,
+ &tmp);
+ chan_addr += (u64) ((tmp >> 11) & 0xfff) << 27;
+ }
+
+ f15h_select_dct(pvt, channel);
+
+ edac_dbg(1, " Normalized DCT addr: 0x%llx\n", chan_addr);
+
+ /*
+ * Find Chip select:
+ * if channel = 3, then alias it to 1. This is because, in F15 M30h,
+ * there is support for 4 DCT's, but only 2 are currently functional.
+ * They are DCT0 and DCT3. But we have read all registers of DCT3 into
+ * pvt->csels[1]. So we need to use '1' here to get correct info.
+ * Refer F15 M30h BKDG Section 2.10 and 2.10.3 for clarifications.
+ */
+ alias_channel = (channel == 3) ? 1 : channel;
+
+ cs_found = f1x_lookup_addr_in_dct(chan_addr, node_id, alias_channel);
+
+ if (cs_found >= 0)
+ *chan_sel = alias_channel;
+
+ return cs_found;
+}
+
+static int f1x_translate_sysaddr_to_cs(struct amd64_pvt *pvt,
+ u64 sys_addr,
+ int *chan_sel)
+{
+ int cs_found = -EINVAL;
+ unsigned range;
+
+ for (range = 0; range < DRAM_RANGES; range++) {
+ if (!dram_rw(pvt, range))
+ continue;
+
+ if (pvt->fam == 0x15 && pvt->model >= 0x30)
+ cs_found = f15_m30h_match_to_this_node(pvt, range,
+ sys_addr,
+ chan_sel);
+
+ else if ((get_dram_base(pvt, range) <= sys_addr) &&
+ (get_dram_limit(pvt, range) >= sys_addr)) {
+ cs_found = f1x_match_to_this_node(pvt, range,
+ sys_addr, chan_sel);
+ if (cs_found >= 0)
+ break;
+ }
+ }
+ return cs_found;
+}
+
+/*
+ * For reference see "2.8.5 Routing DRAM Requests" in F10 BKDG. This code maps
+ * a @sys_addr to NodeID, DCT (channel) and chip select (CSROW).
+ *
+ * The @sys_addr is usually an error address received from the hardware
+ * (MCX_ADDR).
+ */
+static void f1x_map_sysaddr_to_csrow(struct mem_ctl_info *mci, u64 sys_addr,
+ struct err_info *err)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ error_address_to_page_and_offset(sys_addr, err);
+
+ err->csrow = f1x_translate_sysaddr_to_cs(pvt, sys_addr, &err->channel);
+ if (err->csrow < 0) {
+ err->err_code = ERR_CSROW;
+ return;
+ }
+
+ /*
+ * We need the syndromes for channel detection only when we're
+ * ganged. Otherwise @chan should already contain the channel at
+ * this point.
+ */
+ if (dct_ganging_enabled(pvt))
+ err->channel = get_channel_from_ecc_syndrome(mci, err->syndrome);
+}
+
+/*
+ * These are tables of eigenvectors (one per line) which can be used for the
+ * construction of the syndrome tables. The modified syndrome search algorithm
+ * uses those to find the symbol in error and thus the DIMM.
+ *
+ * Algorithm courtesy of Ross LaFetra from AMD.
+ */
+static const u16 x4_vectors[] = {
+ 0x2f57, 0x1afe, 0x66cc, 0xdd88,
+ 0x11eb, 0x3396, 0x7f4c, 0xeac8,
+ 0x0001, 0x0002, 0x0004, 0x0008,
+ 0x1013, 0x3032, 0x4044, 0x8088,
+ 0x106b, 0x30d6, 0x70fc, 0xe0a8,
+ 0x4857, 0xc4fe, 0x13cc, 0x3288,
+ 0x1ac5, 0x2f4a, 0x5394, 0xa1e8,
+ 0x1f39, 0x251e, 0xbd6c, 0x6bd8,
+ 0x15c1, 0x2a42, 0x89ac, 0x4758,
+ 0x2b03, 0x1602, 0x4f0c, 0xca08,
+ 0x1f07, 0x3a0e, 0x6b04, 0xbd08,
+ 0x8ba7, 0x465e, 0x244c, 0x1cc8,
+ 0x2b87, 0x164e, 0x642c, 0xdc18,
+ 0x40b9, 0x80de, 0x1094, 0x20e8,
+ 0x27db, 0x1eb6, 0x9dac, 0x7b58,
+ 0x11c1, 0x2242, 0x84ac, 0x4c58,
+ 0x1be5, 0x2d7a, 0x5e34, 0xa718,
+ 0x4b39, 0x8d1e, 0x14b4, 0x28d8,
+ 0x4c97, 0xc87e, 0x11fc, 0x33a8,
+ 0x8e97, 0x497e, 0x2ffc, 0x1aa8,
+ 0x16b3, 0x3d62, 0x4f34, 0x8518,
+ 0x1e2f, 0x391a, 0x5cac, 0xf858,
+ 0x1d9f, 0x3b7a, 0x572c, 0xfe18,
+ 0x15f5, 0x2a5a, 0x5264, 0xa3b8,
+ 0x1dbb, 0x3b66, 0x715c, 0xe3f8,
+ 0x4397, 0xc27e, 0x17fc, 0x3ea8,
+ 0x1617, 0x3d3e, 0x6464, 0xb8b8,
+ 0x23ff, 0x12aa, 0xab6c, 0x56d8,
+ 0x2dfb, 0x1ba6, 0x913c, 0x7328,
+ 0x185d, 0x2ca6, 0x7914, 0x9e28,
+ 0x171b, 0x3e36, 0x7d7c, 0xebe8,
+ 0x4199, 0x82ee, 0x19f4, 0x2e58,
+ 0x4807, 0xc40e, 0x130c, 0x3208,
+ 0x1905, 0x2e0a, 0x5804, 0xac08,
+ 0x213f, 0x132a, 0xadfc, 0x5ba8,
+ 0x19a9, 0x2efe, 0xb5cc, 0x6f88,
+};
+
+static const u16 x8_vectors[] = {
+ 0x0145, 0x028a, 0x2374, 0x43c8, 0xa1f0, 0x0520, 0x0a40, 0x1480,
+ 0x0211, 0x0422, 0x0844, 0x1088, 0x01b0, 0x44e0, 0x23c0, 0xed80,
+ 0x1011, 0x0116, 0x022c, 0x0458, 0x08b0, 0x8c60, 0x2740, 0x4e80,
+ 0x0411, 0x0822, 0x1044, 0x0158, 0x02b0, 0x2360, 0x46c0, 0xab80,
+ 0x0811, 0x1022, 0x012c, 0x0258, 0x04b0, 0x4660, 0x8cc0, 0x2780,
+ 0x2071, 0x40e2, 0xa0c4, 0x0108, 0x0210, 0x0420, 0x0840, 0x1080,
+ 0x4071, 0x80e2, 0x0104, 0x0208, 0x0410, 0x0820, 0x1040, 0x2080,
+ 0x8071, 0x0102, 0x0204, 0x0408, 0x0810, 0x1020, 0x2040, 0x4080,
+ 0x019d, 0x03d6, 0x136c, 0x2198, 0x50b0, 0xb2e0, 0x0740, 0x0e80,
+ 0x0189, 0x03ea, 0x072c, 0x0e58, 0x1cb0, 0x56e0, 0x37c0, 0xf580,
+ 0x01fd, 0x0376, 0x06ec, 0x0bb8, 0x1110, 0x2220, 0x4440, 0x8880,
+ 0x0163, 0x02c6, 0x1104, 0x0758, 0x0eb0, 0x2be0, 0x6140, 0xc280,
+ 0x02fd, 0x01c6, 0x0b5c, 0x1108, 0x07b0, 0x25a0, 0x8840, 0x6180,
+ 0x0801, 0x012e, 0x025c, 0x04b8, 0x1370, 0x26e0, 0x57c0, 0xb580,
+ 0x0401, 0x0802, 0x015c, 0x02b8, 0x22b0, 0x13e0, 0x7140, 0xe280,
+ 0x0201, 0x0402, 0x0804, 0x01b8, 0x11b0, 0x31a0, 0x8040, 0x7180,
+ 0x0101, 0x0202, 0x0404, 0x0808, 0x1010, 0x2020, 0x4040, 0x8080,
+ 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+ 0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000, 0x8000,
+};
+
+static int decode_syndrome(u16 syndrome, const u16 *vectors, unsigned num_vecs,
+ unsigned v_dim)
+{
+ unsigned int i, err_sym;
+
+ for (err_sym = 0; err_sym < num_vecs / v_dim; err_sym++) {
+ u16 s = syndrome;
+ unsigned v_idx = err_sym * v_dim;
+ unsigned v_end = (err_sym + 1) * v_dim;
+
+ /* walk over all 16 bits of the syndrome */
+ for (i = 1; i < (1U << 16); i <<= 1) {
+
+ /* if bit is set in that eigenvector... */
+ if (v_idx < v_end && vectors[v_idx] & i) {
+ u16 ev_comp = vectors[v_idx++];
+
+ /* ... and bit set in the modified syndrome, */
+ if (s & i) {
+ /* remove it. */
+ s ^= ev_comp;
+
+ if (!s)
+ return err_sym;
+ }
+
+ } else if (s & i)
+ /* can't get to zero, move to next symbol */
+ break;
+ }
+ }
+
+ edac_dbg(0, "syndrome(%x) not found\n", syndrome);
+ return -1;
+}
+
+static int map_err_sym_to_channel(int err_sym, int sym_size)
+{
+ if (sym_size == 4)
+ switch (err_sym) {
+ case 0x20:
+ case 0x21:
+ return 0;
+ case 0x22:
+ case 0x23:
+ return 1;
+ default:
+ return err_sym >> 4;
+ }
+ /* x8 symbols */
+ else
+ switch (err_sym) {
+ /* imaginary bits not in a DIMM */
+ case 0x10:
+ WARN(1, KERN_ERR "Invalid error symbol: 0x%x\n",
+ err_sym);
+ return -1;
+ case 0x11:
+ return 0;
+ case 0x12:
+ return 1;
+ default:
+ return err_sym >> 3;
+ }
+ return -1;
+}
+
+static int get_channel_from_ecc_syndrome(struct mem_ctl_info *mci, u16 syndrome)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ int err_sym = -1;
+
+ if (pvt->ecc_sym_sz == 8)
+ err_sym = decode_syndrome(syndrome, x8_vectors,
+ ARRAY_SIZE(x8_vectors),
+ pvt->ecc_sym_sz);
+ else if (pvt->ecc_sym_sz == 4)
+ err_sym = decode_syndrome(syndrome, x4_vectors,
+ ARRAY_SIZE(x4_vectors),
+ pvt->ecc_sym_sz);
+ else {
+ amd64_warn("Illegal syndrome type: %u\n", pvt->ecc_sym_sz);
+ return err_sym;
+ }
+
+ return map_err_sym_to_channel(err_sym, pvt->ecc_sym_sz);
+}
+
+static void __log_ecc_error(struct mem_ctl_info *mci, struct err_info *err,
+ u8 ecc_type)
+{
+ enum hw_event_mc_err_type err_type;
+ const char *string;
+
+ if (ecc_type == 2)
+ err_type = HW_EVENT_ERR_CORRECTED;
+ else if (ecc_type == 1)
+ err_type = HW_EVENT_ERR_UNCORRECTED;
+ else if (ecc_type == 3)
+ err_type = HW_EVENT_ERR_DEFERRED;
+ else {
+ WARN(1, "Something is rotten in the state of Denmark.\n");
+ return;
+ }
+
+ switch (err->err_code) {
+ case DECODE_OK:
+ string = "";
+ break;
+ case ERR_NODE:
+ string = "Failed to map error addr to a node";
+ break;
+ case ERR_CSROW:
+ string = "Failed to map error addr to a csrow";
+ break;
+ case ERR_CHANNEL:
+ string = "Unknown syndrome - possible error reporting race";
+ break;
+ case ERR_SYND:
+ string = "MCA_SYND not valid - unknown syndrome and csrow";
+ break;
+ case ERR_NORM_ADDR:
+ string = "Cannot decode normalized address";
+ break;
+ default:
+ string = "WTF error";
+ break;
+ }
+
+ edac_mc_handle_error(err_type, mci, 1,
+ err->page, err->offset, err->syndrome,
+ err->csrow, err->channel, -1,
+ string, "");
+}
+
+static inline void decode_bus_error(int node_id, struct mce *m)
+{
+ struct mem_ctl_info *mci;
+ struct amd64_pvt *pvt;
+ u8 ecc_type = (m->status >> 45) & 0x3;
+ u8 xec = XEC(m->status, 0x1f);
+ u16 ec = EC(m->status);
+ u64 sys_addr;
+ struct err_info err;
+
+ mci = edac_mc_find(node_id);
+ if (!mci)
+ return;
+
+ pvt = mci->pvt_info;
+
+ /* Bail out early if this was an 'observed' error */
+ if (PP(ec) == NBSL_PP_OBS)
+ return;
+
+ /* Do only ECC errors */
+ if (xec && xec != F10_NBSL_EXT_ERR_ECC)
+ return;
+
+ memset(&err, 0, sizeof(err));
+
+ sys_addr = get_error_address(pvt, m);
+
+ if (ecc_type == 2)
+ err.syndrome = extract_syndrome(m->status);
+
+ pvt->ops->map_sysaddr_to_csrow(mci, sys_addr, &err);
+
+ __log_ecc_error(mci, &err, ecc_type);
+}
+
+/*
+ * To find the UMC channel represented by this bank we need to match on its
+ * instance_id. The instance_id of a bank is held in the lower 32 bits of its
+ * IPID.
+ *
+ * Currently, we can derive the channel number by looking at the 6th nibble in
+ * the instance_id. For example, instance_id=0xYXXXXX where Y is the channel
+ * number.
+ *
+ * For DRAM ECC errors, the Chip Select number is given in bits [2:0] of
+ * the MCA_SYND[ErrorInformation] field.
+ */
+static void umc_get_err_info(struct mce *m, struct err_info *err)
+{
+ err->channel = (m->ipid & GENMASK(31, 0)) >> 20;
+ err->csrow = m->synd & 0x7;
+}
+
+static void decode_umc_error(int node_id, struct mce *m)
+{
+ u8 ecc_type = (m->status >> 45) & 0x3;
+ struct mem_ctl_info *mci;
+ struct amd64_pvt *pvt;
+ struct err_info err;
+ u64 sys_addr;
+
+ node_id = fixup_node_id(node_id, m);
+
+ mci = edac_mc_find(node_id);
+ if (!mci)
+ return;
+
+ pvt = mci->pvt_info;
+
+ memset(&err, 0, sizeof(err));
+
+ if (m->status & MCI_STATUS_DEFERRED)
+ ecc_type = 3;
+
+ if (!(m->status & MCI_STATUS_SYNDV)) {
+ err.err_code = ERR_SYND;
+ goto log_error;
+ }
+
+ if (ecc_type == 2) {
+ u8 length = (m->synd >> 18) & 0x3f;
+
+ if (length)
+ err.syndrome = (m->synd >> 32) & GENMASK(length - 1, 0);
+ else
+ err.err_code = ERR_CHANNEL;
+ }
+
+ pvt->ops->get_err_info(m, &err);
+
+ if (umc_normaddr_to_sysaddr(m->addr, pvt->mc_node_id, err.channel, &sys_addr)) {
+ err.err_code = ERR_NORM_ADDR;
+ goto log_error;
+ }
+
+ error_address_to_page_and_offset(sys_addr, &err);
+
+log_error:
+ __log_ecc_error(mci, &err, ecc_type);
+}
+
+/*
+ * Use pvt->F3 which contains the F3 CPU PCI device to get the related
+ * F1 (AddrMap) and F2 (Dct) devices. Return negative value on error.
+ */
+static int
+reserve_mc_sibling_devs(struct amd64_pvt *pvt, u16 pci_id1, u16 pci_id2)
+{
+ /* Reserve the ADDRESS MAP Device */
+ pvt->F1 = pci_get_related_function(pvt->F3->vendor, pci_id1, pvt->F3);
+ if (!pvt->F1) {
+ edac_dbg(1, "F1 not found: device 0x%x\n", pci_id1);
+ return -ENODEV;
+ }
+
+ /* Reserve the DCT Device */
+ pvt->F2 = pci_get_related_function(pvt->F3->vendor, pci_id2, pvt->F3);
+ if (!pvt->F2) {
+ pci_dev_put(pvt->F1);
+ pvt->F1 = NULL;
+
+ edac_dbg(1, "F2 not found: device 0x%x\n", pci_id2);
+ return -ENODEV;
+ }
+
+ if (!pci_ctl_dev)
+ pci_ctl_dev = &pvt->F2->dev;
+
+ edac_dbg(1, "F1: %s\n", pci_name(pvt->F1));
+ edac_dbg(1, "F2: %s\n", pci_name(pvt->F2));
+ edac_dbg(1, "F3: %s\n", pci_name(pvt->F3));
+
+ return 0;
+}
+
+static void determine_ecc_sym_sz(struct amd64_pvt *pvt)
+{
+ pvt->ecc_sym_sz = 4;
+
+ if (pvt->fam >= 0x10) {
+ u32 tmp;
+
+ amd64_read_pci_cfg(pvt->F3, EXT_NB_MCA_CFG, &tmp);
+ /* F16h has only DCT0, so no need to read dbam1. */
+ if (pvt->fam != 0x16)
+ amd64_read_dct_pci_cfg(pvt, 1, DBAM0, &pvt->dbam1);
+
+ /* F10h, revD and later can do x8 ECC too. */
+ if ((pvt->fam > 0x10 || pvt->model > 7) && tmp & BIT(25))
+ pvt->ecc_sym_sz = 8;
+ }
+}
+
+/*
+ * Retrieve the hardware registers of the memory controller.
+ */
+static void umc_read_mc_regs(struct amd64_pvt *pvt)
+{
+ u8 nid = pvt->mc_node_id;
+ struct amd64_umc *umc;
+ u32 i, umc_base;
+
+ /* Read registers from each UMC */
+ for_each_umc(i) {
+
+ umc_base = get_umc_base(i);
+ umc = &pvt->umc[i];
+
+ amd_smn_read(nid, umc_base + get_umc_reg(pvt, UMCCH_DIMM_CFG), &umc->dimm_cfg);
+ amd_smn_read(nid, umc_base + UMCCH_UMC_CFG, &umc->umc_cfg);
+ amd_smn_read(nid, umc_base + UMCCH_SDP_CTRL, &umc->sdp_ctrl);
+ amd_smn_read(nid, umc_base + UMCCH_ECC_CTRL, &umc->ecc_ctrl);
+ amd_smn_read(nid, umc_base + UMCCH_UMC_CAP_HI, &umc->umc_cap_hi);
+ }
+}
+
+/*
+ * Retrieve the hardware registers of the memory controller (this includes the
+ * 'Address Map' and 'Misc' device regs)
+ */
+static void dct_read_mc_regs(struct amd64_pvt *pvt)
+{
+ unsigned int range;
+ u64 msr_val;
+
+ /*
+ * Retrieve TOP_MEM and TOP_MEM2; no masking off of reserved bits since
+ * those are Read-As-Zero.
+ */
+ rdmsrl(MSR_K8_TOP_MEM1, pvt->top_mem);
+ edac_dbg(0, " TOP_MEM: 0x%016llx\n", pvt->top_mem);
+
+ /* Check first whether TOP_MEM2 is enabled: */
+ rdmsrl(MSR_AMD64_SYSCFG, msr_val);
+ if (msr_val & BIT(21)) {
+ rdmsrl(MSR_K8_TOP_MEM2, pvt->top_mem2);
+ edac_dbg(0, " TOP_MEM2: 0x%016llx\n", pvt->top_mem2);
+ } else {
+ edac_dbg(0, " TOP_MEM2 disabled\n");
+ }
+
+ amd64_read_pci_cfg(pvt->F3, NBCAP, &pvt->nbcap);
+
+ read_dram_ctl_register(pvt);
+
+ for (range = 0; range < DRAM_RANGES; range++) {
+ u8 rw;
+
+ /* read settings for this DRAM range */
+ read_dram_base_limit_regs(pvt, range);
+
+ rw = dram_rw(pvt, range);
+ if (!rw)
+ continue;
+
+ edac_dbg(1, " DRAM range[%d], base: 0x%016llx; limit: 0x%016llx\n",
+ range,
+ get_dram_base(pvt, range),
+ get_dram_limit(pvt, range));
+
+ edac_dbg(1, " IntlvEn=%s; Range access: %s%s IntlvSel=%d DstNode=%d\n",
+ dram_intlv_en(pvt, range) ? "Enabled" : "Disabled",
+ (rw & 0x1) ? "R" : "-",
+ (rw & 0x2) ? "W" : "-",
+ dram_intlv_sel(pvt, range),
+ dram_dst_node(pvt, range));
+ }
+
+ amd64_read_pci_cfg(pvt->F1, DHAR, &pvt->dhar);
+ amd64_read_dct_pci_cfg(pvt, 0, DBAM0, &pvt->dbam0);
+
+ amd64_read_pci_cfg(pvt->F3, F10_ONLINE_SPARE, &pvt->online_spare);
+
+ amd64_read_dct_pci_cfg(pvt, 0, DCLR0, &pvt->dclr0);
+ amd64_read_dct_pci_cfg(pvt, 0, DCHR0, &pvt->dchr0);
+
+ if (!dct_ganging_enabled(pvt)) {
+ amd64_read_dct_pci_cfg(pvt, 1, DCLR0, &pvt->dclr1);
+ amd64_read_dct_pci_cfg(pvt, 1, DCHR0, &pvt->dchr1);
+ }
+
+ determine_ecc_sym_sz(pvt);
+}
+
+/*
+ * NOTE: CPU Revision Dependent code
+ *
+ * Input:
+ * @csrow_nr ChipSelect Row Number (0..NUM_CHIPSELECTS-1)
+ * k8 private pointer to -->
+ * DRAM Bank Address mapping register
+ * node_id
+ * DCL register where dual_channel_active is
+ *
+ * The DBAM register consists of 4 sets of 4 bits each definitions:
+ *
+ * Bits: CSROWs
+ * 0-3 CSROWs 0 and 1
+ * 4-7 CSROWs 2 and 3
+ * 8-11 CSROWs 4 and 5
+ * 12-15 CSROWs 6 and 7
+ *
+ * Values range from: 0 to 15
+ * The meaning of the values depends on CPU revision and dual-channel state,
+ * see relevant BKDG more info.
+ *
+ * The memory controller provides for total of only 8 CSROWs in its current
+ * architecture. Each "pair" of CSROWs normally represents just one DIMM in
+ * single channel or two (2) DIMMs in dual channel mode.
+ *
+ * The following code logic collapses the various tables for CSROW based on CPU
+ * revision.
+ *
+ * Returns:
+ * The number of PAGE_SIZE pages on the specified CSROW number it
+ * encompasses
+ *
+ */
+static u32 dct_get_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
+{
+ u32 dbam = dct ? pvt->dbam1 : pvt->dbam0;
+ u32 cs_mode, nr_pages;
+
+ csrow_nr >>= 1;
+ cs_mode = DBAM_DIMM(csrow_nr, dbam);
+
+ nr_pages = pvt->ops->dbam_to_cs(pvt, dct, cs_mode, csrow_nr);
+ nr_pages <<= 20 - PAGE_SHIFT;
+
+ edac_dbg(0, "csrow: %d, channel: %d, DBAM idx: %d\n",
+ csrow_nr, dct, cs_mode);
+ edac_dbg(0, "nr_pages/channel: %u\n", nr_pages);
+
+ return nr_pages;
+}
+
+static u32 umc_get_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr_orig)
+{
+ int csrow_nr = csrow_nr_orig;
+ u32 cs_mode, nr_pages;
+
+ cs_mode = umc_get_cs_mode(csrow_nr >> 1, dct, pvt);
+
+ nr_pages = umc_addr_mask_to_cs_size(pvt, dct, cs_mode, csrow_nr);
+ nr_pages <<= 20 - PAGE_SHIFT;
+
+ edac_dbg(0, "csrow: %d, channel: %d, cs_mode %d\n",
+ csrow_nr_orig, dct, cs_mode);
+ edac_dbg(0, "nr_pages/channel: %u\n", nr_pages);
+
+ return nr_pages;
+}
+
+static void umc_init_csrows(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ enum edac_type edac_mode = EDAC_NONE;
+ enum dev_type dev_type = DEV_UNKNOWN;
+ struct dimm_info *dimm;
+ u8 umc, cs;
+
+ if (mci->edac_ctl_cap & EDAC_FLAG_S16ECD16ED) {
+ edac_mode = EDAC_S16ECD16ED;
+ dev_type = DEV_X16;
+ } else if (mci->edac_ctl_cap & EDAC_FLAG_S8ECD8ED) {
+ edac_mode = EDAC_S8ECD8ED;
+ dev_type = DEV_X8;
+ } else if (mci->edac_ctl_cap & EDAC_FLAG_S4ECD4ED) {
+ edac_mode = EDAC_S4ECD4ED;
+ dev_type = DEV_X4;
+ } else if (mci->edac_ctl_cap & EDAC_FLAG_SECDED) {
+ edac_mode = EDAC_SECDED;
+ }
+
+ for_each_umc(umc) {
+ for_each_chip_select(cs, umc, pvt) {
+ if (!csrow_enabled(cs, umc, pvt))
+ continue;
+
+ dimm = mci->csrows[cs]->channels[umc]->dimm;
+
+ edac_dbg(1, "MC node: %d, csrow: %d\n",
+ pvt->mc_node_id, cs);
+
+ dimm->nr_pages = umc_get_csrow_nr_pages(pvt, umc, cs);
+ dimm->mtype = pvt->umc[umc].dram_type;
+ dimm->edac_mode = edac_mode;
+ dimm->dtype = dev_type;
+ dimm->grain = 64;
+ }
+ }
+}
+
+/*
+ * Initialize the array of csrow attribute instances, based on the values
+ * from pci config hardware registers.
+ */
+static void dct_init_csrows(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ enum edac_type edac_mode = EDAC_NONE;
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ int nr_pages = 0;
+ int i, j;
+ u32 val;
+
+ amd64_read_pci_cfg(pvt->F3, NBCFG, &val);
+
+ pvt->nbcfg = val;
+
+ edac_dbg(0, "node %d, NBCFG=0x%08x[ChipKillEccCap: %d|DramEccEn: %d]\n",
+ pvt->mc_node_id, val,
+ !!(val & NBCFG_CHIPKILL), !!(val & NBCFG_ECC_ENABLE));
+
+ /*
+ * We iterate over DCT0 here but we look at DCT1 in parallel, if needed.
+ */
+ for_each_chip_select(i, 0, pvt) {
+ bool row_dct0 = !!csrow_enabled(i, 0, pvt);
+ bool row_dct1 = false;
+
+ if (pvt->fam != 0xf)
+ row_dct1 = !!csrow_enabled(i, 1, pvt);
+
+ if (!row_dct0 && !row_dct1)
+ continue;
+
+ csrow = mci->csrows[i];
+
+ edac_dbg(1, "MC node: %d, csrow: %d\n",
+ pvt->mc_node_id, i);
+
+ if (row_dct0) {
+ nr_pages = dct_get_csrow_nr_pages(pvt, 0, i);
+ csrow->channels[0]->dimm->nr_pages = nr_pages;
+ }
+
+ /* K8 has only one DCT */
+ if (pvt->fam != 0xf && row_dct1) {
+ int row_dct1_pages = dct_get_csrow_nr_pages(pvt, 1, i);
+
+ csrow->channels[1]->dimm->nr_pages = row_dct1_pages;
+ nr_pages += row_dct1_pages;
+ }
+
+ edac_dbg(1, "Total csrow%d pages: %u\n", i, nr_pages);
+
+ /* Determine DIMM ECC mode: */
+ if (pvt->nbcfg & NBCFG_ECC_ENABLE) {
+ edac_mode = (pvt->nbcfg & NBCFG_CHIPKILL)
+ ? EDAC_S4ECD4ED
+ : EDAC_SECDED;
+ }
+
+ for (j = 0; j < pvt->max_mcs; j++) {
+ dimm = csrow->channels[j]->dimm;
+ dimm->mtype = pvt->dram_type;
+ dimm->edac_mode = edac_mode;
+ dimm->grain = 64;
+ }
+ }
+}
+
+/* get all cores on this DCT */
+static void get_cpus_on_this_dct_cpumask(struct cpumask *mask, u16 nid)
+{
+ int cpu;
+
+ for_each_online_cpu(cpu)
+ if (topology_die_id(cpu) == nid)
+ cpumask_set_cpu(cpu, mask);
+}
+
+/* check MCG_CTL on all the cpus on this node */
+static bool nb_mce_bank_enabled_on_node(u16 nid)
+{
+ cpumask_var_t mask;
+ int cpu, nbe;
+ bool ret = false;
+
+ if (!zalloc_cpumask_var(&mask, GFP_KERNEL)) {
+ amd64_warn("%s: Error allocating mask\n", __func__);
+ return false;
+ }
+
+ get_cpus_on_this_dct_cpumask(mask, nid);
+
+ rdmsr_on_cpus(mask, MSR_IA32_MCG_CTL, msrs);
+
+ for_each_cpu(cpu, mask) {
+ struct msr *reg = per_cpu_ptr(msrs, cpu);
+ nbe = reg->l & MSR_MCGCTL_NBE;
+
+ edac_dbg(0, "core: %u, MCG_CTL: 0x%llx, NB MSR is %s\n",
+ cpu, reg->q,
+ (nbe ? "enabled" : "disabled"));
+
+ if (!nbe)
+ goto out;
+ }
+ ret = true;
+
+out:
+ free_cpumask_var(mask);
+ return ret;
+}
+
+static int toggle_ecc_err_reporting(struct ecc_settings *s, u16 nid, bool on)
+{
+ cpumask_var_t cmask;
+ int cpu;
+
+ if (!zalloc_cpumask_var(&cmask, GFP_KERNEL)) {
+ amd64_warn("%s: error allocating mask\n", __func__);
+ return -ENOMEM;
+ }
+
+ get_cpus_on_this_dct_cpumask(cmask, nid);
+
+ rdmsr_on_cpus(cmask, MSR_IA32_MCG_CTL, msrs);
+
+ for_each_cpu(cpu, cmask) {
+
+ struct msr *reg = per_cpu_ptr(msrs, cpu);
+
+ if (on) {
+ if (reg->l & MSR_MCGCTL_NBE)
+ s->flags.nb_mce_enable = 1;
+
+ reg->l |= MSR_MCGCTL_NBE;
+ } else {
+ /*
+ * Turn off NB MCE reporting only when it was off before
+ */
+ if (!s->flags.nb_mce_enable)
+ reg->l &= ~MSR_MCGCTL_NBE;
+ }
+ }
+ wrmsr_on_cpus(cmask, MSR_IA32_MCG_CTL, msrs);
+
+ free_cpumask_var(cmask);
+
+ return 0;
+}
+
+static bool enable_ecc_error_reporting(struct ecc_settings *s, u16 nid,
+ struct pci_dev *F3)
+{
+ bool ret = true;
+ u32 value, mask = 0x3; /* UECC/CECC enable */
+
+ if (toggle_ecc_err_reporting(s, nid, ON)) {
+ amd64_warn("Error enabling ECC reporting over MCGCTL!\n");
+ return false;
+ }
+
+ amd64_read_pci_cfg(F3, NBCTL, &value);
+
+ s->old_nbctl = value & mask;
+ s->nbctl_valid = true;
+
+ value |= mask;
+ amd64_write_pci_cfg(F3, NBCTL, value);
+
+ amd64_read_pci_cfg(F3, NBCFG, &value);
+
+ edac_dbg(0, "1: node %d, NBCFG=0x%08x[DramEccEn: %d]\n",
+ nid, value, !!(value & NBCFG_ECC_ENABLE));
+
+ if (!(value & NBCFG_ECC_ENABLE)) {
+ amd64_warn("DRAM ECC disabled on this node, enabling...\n");
+
+ s->flags.nb_ecc_prev = 0;
+
+ /* Attempt to turn on DRAM ECC Enable */
+ value |= NBCFG_ECC_ENABLE;
+ amd64_write_pci_cfg(F3, NBCFG, value);
+
+ amd64_read_pci_cfg(F3, NBCFG, &value);
+
+ if (!(value & NBCFG_ECC_ENABLE)) {
+ amd64_warn("Hardware rejected DRAM ECC enable,"
+ "check memory DIMM configuration.\n");
+ ret = false;
+ } else {
+ amd64_info("Hardware accepted DRAM ECC Enable\n");
+ }
+ } else {
+ s->flags.nb_ecc_prev = 1;
+ }
+
+ edac_dbg(0, "2: node %d, NBCFG=0x%08x[DramEccEn: %d]\n",
+ nid, value, !!(value & NBCFG_ECC_ENABLE));
+
+ return ret;
+}
+
+static void restore_ecc_error_reporting(struct ecc_settings *s, u16 nid,
+ struct pci_dev *F3)
+{
+ u32 value, mask = 0x3; /* UECC/CECC enable */
+
+ if (!s->nbctl_valid)
+ return;
+
+ amd64_read_pci_cfg(F3, NBCTL, &value);
+ value &= ~mask;
+ value |= s->old_nbctl;
+
+ amd64_write_pci_cfg(F3, NBCTL, value);
+
+ /* restore previous BIOS DRAM ECC "off" setting we force-enabled */
+ if (!s->flags.nb_ecc_prev) {
+ amd64_read_pci_cfg(F3, NBCFG, &value);
+ value &= ~NBCFG_ECC_ENABLE;
+ amd64_write_pci_cfg(F3, NBCFG, value);
+ }
+
+ /* restore the NB Enable MCGCTL bit */
+ if (toggle_ecc_err_reporting(s, nid, OFF))
+ amd64_warn("Error restoring NB MCGCTL settings!\n");
+}
+
+static bool dct_ecc_enabled(struct amd64_pvt *pvt)
+{
+ u16 nid = pvt->mc_node_id;
+ bool nb_mce_en = false;
+ u8 ecc_en = 0;
+ u32 value;
+
+ amd64_read_pci_cfg(pvt->F3, NBCFG, &value);
+
+ ecc_en = !!(value & NBCFG_ECC_ENABLE);
+
+ nb_mce_en = nb_mce_bank_enabled_on_node(nid);
+ if (!nb_mce_en)
+ edac_dbg(0, "NB MCE bank disabled, set MSR 0x%08x[4] on node %d to enable.\n",
+ MSR_IA32_MCG_CTL, nid);
+
+ edac_dbg(3, "Node %d: DRAM ECC %s.\n", nid, (ecc_en ? "enabled" : "disabled"));
+
+ if (!ecc_en || !nb_mce_en)
+ return false;
+ else
+ return true;
+}
+
+static bool umc_ecc_enabled(struct amd64_pvt *pvt)
+{
+ u8 umc_en_mask = 0, ecc_en_mask = 0;
+ u16 nid = pvt->mc_node_id;
+ struct amd64_umc *umc;
+ u8 ecc_en = 0, i;
+
+ for_each_umc(i) {
+ umc = &pvt->umc[i];
+
+ /* Only check enabled UMCs. */
+ if (!(umc->sdp_ctrl & UMC_SDP_INIT))
+ continue;
+
+ umc_en_mask |= BIT(i);
+
+ if (umc->umc_cap_hi & UMC_ECC_ENABLED)
+ ecc_en_mask |= BIT(i);
+ }
+
+ /* Check whether at least one UMC is enabled: */
+ if (umc_en_mask)
+ ecc_en = umc_en_mask == ecc_en_mask;
+ else
+ edac_dbg(0, "Node %d: No enabled UMCs.\n", nid);
+
+ edac_dbg(3, "Node %d: DRAM ECC %s.\n", nid, (ecc_en ? "enabled" : "disabled"));
+
+ if (!ecc_en)
+ return false;
+ else
+ return true;
+}
+
+static inline void
+umc_determine_edac_ctl_cap(struct mem_ctl_info *mci, struct amd64_pvt *pvt)
+{
+ u8 i, ecc_en = 1, cpk_en = 1, dev_x4 = 1, dev_x16 = 1;
+
+ for_each_umc(i) {
+ if (pvt->umc[i].sdp_ctrl & UMC_SDP_INIT) {
+ ecc_en &= !!(pvt->umc[i].umc_cap_hi & UMC_ECC_ENABLED);
+ cpk_en &= !!(pvt->umc[i].umc_cap_hi & UMC_ECC_CHIPKILL_CAP);
+
+ dev_x4 &= !!(pvt->umc[i].dimm_cfg & BIT(6));
+ dev_x16 &= !!(pvt->umc[i].dimm_cfg & BIT(7));
+ }
+ }
+
+ /* Set chipkill only if ECC is enabled: */
+ if (ecc_en) {
+ mci->edac_ctl_cap |= EDAC_FLAG_SECDED;
+
+ if (!cpk_en)
+ return;
+
+ if (dev_x4)
+ mci->edac_ctl_cap |= EDAC_FLAG_S4ECD4ED;
+ else if (dev_x16)
+ mci->edac_ctl_cap |= EDAC_FLAG_S16ECD16ED;
+ else
+ mci->edac_ctl_cap |= EDAC_FLAG_S8ECD8ED;
+ }
+}
+
+static void dct_setup_mci_misc_attrs(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ mci->mtype_cap = MEM_FLAG_DDR2 | MEM_FLAG_RDDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+
+ if (pvt->nbcap & NBCAP_SECDED)
+ mci->edac_ctl_cap |= EDAC_FLAG_SECDED;
+
+ if (pvt->nbcap & NBCAP_CHIPKILL)
+ mci->edac_ctl_cap |= EDAC_FLAG_S4ECD4ED;
+
+ mci->edac_cap = dct_determine_edac_cap(pvt);
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = pvt->ctl_name;
+ mci->dev_name = pci_name(pvt->F3);
+ mci->ctl_page_to_phys = NULL;
+
+ /* memory scrubber interface */
+ mci->set_sdram_scrub_rate = set_scrub_rate;
+ mci->get_sdram_scrub_rate = get_scrub_rate;
+
+ dct_init_csrows(mci);
+}
+
+static void umc_setup_mci_misc_attrs(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ mci->mtype_cap = MEM_FLAG_DDR4 | MEM_FLAG_RDDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+
+ umc_determine_edac_ctl_cap(mci, pvt);
+
+ mci->edac_cap = umc_determine_edac_cap(pvt);
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = pvt->ctl_name;
+ mci->dev_name = pci_name(pvt->F3);
+ mci->ctl_page_to_phys = NULL;
+
+ umc_init_csrows(mci);
+}
+
+static int dct_hw_info_get(struct amd64_pvt *pvt)
+{
+ int ret = reserve_mc_sibling_devs(pvt, pvt->f1_id, pvt->f2_id);
+
+ if (ret)
+ return ret;
+
+ dct_prep_chip_selects(pvt);
+ dct_read_base_mask(pvt);
+ dct_read_mc_regs(pvt);
+ dct_determine_memory_type(pvt);
+
+ return 0;
+}
+
+static int umc_hw_info_get(struct amd64_pvt *pvt)
+{
+ pvt->umc = kcalloc(pvt->max_mcs, sizeof(struct amd64_umc), GFP_KERNEL);
+ if (!pvt->umc)
+ return -ENOMEM;
+
+ umc_prep_chip_selects(pvt);
+ umc_read_base_mask(pvt);
+ umc_read_mc_regs(pvt);
+ umc_determine_memory_type(pvt);
+
+ return 0;
+}
+
+/*
+ * The CPUs have one channel per UMC, so UMC number is equivalent to a
+ * channel number. The GPUs have 8 channels per UMC, so the UMC number no
+ * longer works as a channel number.
+ *
+ * The channel number within a GPU UMC is given in MCA_IPID[15:12].
+ * However, the IDs are split such that two UMC values go to one UMC, and
+ * the channel numbers are split in two groups of four.
+ *
+ * Refer to comment on gpu_get_umc_base().
+ *
+ * For example,
+ * UMC0 CH[3:0] = 0x0005[3:0]000
+ * UMC0 CH[7:4] = 0x0015[3:0]000
+ * UMC1 CH[3:0] = 0x0025[3:0]000
+ * UMC1 CH[7:4] = 0x0035[3:0]000
+ */
+static void gpu_get_err_info(struct mce *m, struct err_info *err)
+{
+ u8 ch = (m->ipid & GENMASK(31, 0)) >> 20;
+ u8 phy = ((m->ipid >> 12) & 0xf);
+
+ err->channel = ch % 2 ? phy + 4 : phy;
+ err->csrow = phy;
+}
+
+static int gpu_addr_mask_to_cs_size(struct amd64_pvt *pvt, u8 umc,
+ unsigned int cs_mode, int csrow_nr)
+{
+ u32 addr_mask_orig = pvt->csels[umc].csmasks[csrow_nr];
+
+ return __addr_mask_to_cs_size(addr_mask_orig, cs_mode, csrow_nr, csrow_nr >> 1);
+}
+
+static void gpu_debug_display_dimm_sizes(struct amd64_pvt *pvt, u8 ctrl)
+{
+ int size, cs_mode, cs = 0;
+
+ edac_printk(KERN_DEBUG, EDAC_MC, "UMC%d chip selects:\n", ctrl);
+
+ cs_mode = CS_EVEN_PRIMARY | CS_ODD_PRIMARY;
+
+ for_each_chip_select(cs, ctrl, pvt) {
+ size = gpu_addr_mask_to_cs_size(pvt, ctrl, cs_mode, cs);
+ amd64_info(EDAC_MC ": %d: %5dMB\n", cs, size);
+ }
+}
+
+static void gpu_dump_misc_regs(struct amd64_pvt *pvt)
+{
+ struct amd64_umc *umc;
+ u32 i;
+
+ for_each_umc(i) {
+ umc = &pvt->umc[i];
+
+ edac_dbg(1, "UMC%d UMC cfg: 0x%x\n", i, umc->umc_cfg);
+ edac_dbg(1, "UMC%d SDP ctrl: 0x%x\n", i, umc->sdp_ctrl);
+ edac_dbg(1, "UMC%d ECC ctrl: 0x%x\n", i, umc->ecc_ctrl);
+ edac_dbg(1, "UMC%d All HBMs support ECC: yes\n", i);
+
+ gpu_debug_display_dimm_sizes(pvt, i);
+ }
+}
+
+static u32 gpu_get_csrow_nr_pages(struct amd64_pvt *pvt, u8 dct, int csrow_nr)
+{
+ u32 nr_pages;
+ int cs_mode = CS_EVEN_PRIMARY | CS_ODD_PRIMARY;
+
+ nr_pages = gpu_addr_mask_to_cs_size(pvt, dct, cs_mode, csrow_nr);
+ nr_pages <<= 20 - PAGE_SHIFT;
+
+ edac_dbg(0, "csrow: %d, channel: %d\n", csrow_nr, dct);
+ edac_dbg(0, "nr_pages/channel: %u\n", nr_pages);
+
+ return nr_pages;
+}
+
+static void gpu_init_csrows(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+ struct dimm_info *dimm;
+ u8 umc, cs;
+
+ for_each_umc(umc) {
+ for_each_chip_select(cs, umc, pvt) {
+ if (!csrow_enabled(cs, umc, pvt))
+ continue;
+
+ dimm = mci->csrows[umc]->channels[cs]->dimm;
+
+ edac_dbg(1, "MC node: %d, csrow: %d\n",
+ pvt->mc_node_id, cs);
+
+ dimm->nr_pages = gpu_get_csrow_nr_pages(pvt, umc, cs);
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->mtype = MEM_HBM2;
+ dimm->dtype = DEV_X16;
+ dimm->grain = 64;
+ }
+ }
+}
+
+static void gpu_setup_mci_misc_attrs(struct mem_ctl_info *mci)
+{
+ struct amd64_pvt *pvt = mci->pvt_info;
+
+ mci->mtype_cap = MEM_FLAG_HBM2;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+
+ mci->edac_cap = EDAC_FLAG_EC;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = pvt->ctl_name;
+ mci->dev_name = pci_name(pvt->F3);
+ mci->ctl_page_to_phys = NULL;
+
+ gpu_init_csrows(mci);
+}
+
+/* ECC is enabled by default on GPU nodes */
+static bool gpu_ecc_enabled(struct amd64_pvt *pvt)
+{
+ return true;
+}
+
+static inline u32 gpu_get_umc_base(u8 umc, u8 channel)
+{
+ /*
+ * On CPUs, there is one channel per UMC, so UMC numbering equals
+ * channel numbering. On GPUs, there are eight channels per UMC,
+ * so the channel numbering is different from UMC numbering.
+ *
+ * On CPU nodes channels are selected in 6th nibble
+ * UMC chY[3:0]= [(chY*2 + 1) : (chY*2)]50000;
+ *
+ * On GPU nodes channels are selected in 3rd nibble
+ * HBM chX[3:0]= [Y ]5X[3:0]000;
+ * HBM chX[7:4]= [Y+1]5X[3:0]000
+ */
+ umc *= 2;
+
+ if (channel >= 4)
+ umc++;
+
+ return 0x50000 + (umc << 20) + ((channel % 4) << 12);
+}
+
+static void gpu_read_mc_regs(struct amd64_pvt *pvt)
+{
+ u8 nid = pvt->mc_node_id;
+ struct amd64_umc *umc;
+ u32 i, umc_base;
+
+ /* Read registers from each UMC */
+ for_each_umc(i) {
+ umc_base = gpu_get_umc_base(i, 0);
+ umc = &pvt->umc[i];
+
+ amd_smn_read(nid, umc_base + UMCCH_UMC_CFG, &umc->umc_cfg);
+ amd_smn_read(nid, umc_base + UMCCH_SDP_CTRL, &umc->sdp_ctrl);
+ amd_smn_read(nid, umc_base + UMCCH_ECC_CTRL, &umc->ecc_ctrl);
+ }
+}
+
+static void gpu_read_base_mask(struct amd64_pvt *pvt)
+{
+ u32 base_reg, mask_reg;
+ u32 *base, *mask;
+ int umc, cs;
+
+ for_each_umc(umc) {
+ for_each_chip_select(cs, umc, pvt) {
+ base_reg = gpu_get_umc_base(umc, cs) + UMCCH_BASE_ADDR;
+ base = &pvt->csels[umc].csbases[cs];
+
+ if (!amd_smn_read(pvt->mc_node_id, base_reg, base)) {
+ edac_dbg(0, " DCSB%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *base, base_reg);
+ }
+
+ mask_reg = gpu_get_umc_base(umc, cs) + UMCCH_ADDR_MASK;
+ mask = &pvt->csels[umc].csmasks[cs];
+
+ if (!amd_smn_read(pvt->mc_node_id, mask_reg, mask)) {
+ edac_dbg(0, " DCSM%d[%d]=0x%08x reg: 0x%x\n",
+ umc, cs, *mask, mask_reg);
+ }
+ }
+ }
+}
+
+static void gpu_prep_chip_selects(struct amd64_pvt *pvt)
+{
+ int umc;
+
+ for_each_umc(umc) {
+ pvt->csels[umc].b_cnt = 8;
+ pvt->csels[umc].m_cnt = 8;
+ }
+}
+
+static int gpu_hw_info_get(struct amd64_pvt *pvt)
+{
+ int ret;
+
+ ret = gpu_get_node_map();
+ if (ret)
+ return ret;
+
+ pvt->umc = kcalloc(pvt->max_mcs, sizeof(struct amd64_umc), GFP_KERNEL);
+ if (!pvt->umc)
+ return -ENOMEM;
+
+ gpu_prep_chip_selects(pvt);
+ gpu_read_base_mask(pvt);
+ gpu_read_mc_regs(pvt);
+
+ return 0;
+}
+
+static void hw_info_put(struct amd64_pvt *pvt)
+{
+ pci_dev_put(pvt->F1);
+ pci_dev_put(pvt->F2);
+ kfree(pvt->umc);
+}
+
+static struct low_ops umc_ops = {
+ .hw_info_get = umc_hw_info_get,
+ .ecc_enabled = umc_ecc_enabled,
+ .setup_mci_misc_attrs = umc_setup_mci_misc_attrs,
+ .dump_misc_regs = umc_dump_misc_regs,
+ .get_err_info = umc_get_err_info,
+};
+
+static struct low_ops gpu_ops = {
+ .hw_info_get = gpu_hw_info_get,
+ .ecc_enabled = gpu_ecc_enabled,
+ .setup_mci_misc_attrs = gpu_setup_mci_misc_attrs,
+ .dump_misc_regs = gpu_dump_misc_regs,
+ .get_err_info = gpu_get_err_info,
+};
+
+/* Use Family 16h versions for defaults and adjust as needed below. */
+static struct low_ops dct_ops = {
+ .map_sysaddr_to_csrow = f1x_map_sysaddr_to_csrow,
+ .dbam_to_cs = f16_dbam_to_chip_select,
+ .hw_info_get = dct_hw_info_get,
+ .ecc_enabled = dct_ecc_enabled,
+ .setup_mci_misc_attrs = dct_setup_mci_misc_attrs,
+ .dump_misc_regs = dct_dump_misc_regs,
+};
+
+static int per_family_init(struct amd64_pvt *pvt)
+{
+ pvt->ext_model = boot_cpu_data.x86_model >> 4;
+ pvt->stepping = boot_cpu_data.x86_stepping;
+ pvt->model = boot_cpu_data.x86_model;
+ pvt->fam = boot_cpu_data.x86;
+ pvt->max_mcs = 2;
+
+ /*
+ * Decide on which ops group to use here and do any family/model
+ * overrides below.
+ */
+ if (pvt->fam >= 0x17)
+ pvt->ops = &umc_ops;
+ else
+ pvt->ops = &dct_ops;
+
+ switch (pvt->fam) {
+ case 0xf:
+ pvt->ctl_name = (pvt->ext_model >= K8_REV_F) ?
+ "K8 revF or later" : "K8 revE or earlier";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_K8_NB_ADDRMAP;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_K8_NB_MEMCTL;
+ pvt->ops->map_sysaddr_to_csrow = k8_map_sysaddr_to_csrow;
+ pvt->ops->dbam_to_cs = k8_dbam_to_chip_select;
+ break;
+
+ case 0x10:
+ pvt->ctl_name = "F10h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_10H_NB_MAP;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_10H_NB_DRAM;
+ pvt->ops->dbam_to_cs = f10_dbam_to_chip_select;
+ break;
+
+ case 0x15:
+ switch (pvt->model) {
+ case 0x30:
+ pvt->ctl_name = "F15h_M30h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_15H_M30H_NB_F1;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_15H_M30H_NB_F2;
+ break;
+ case 0x60:
+ pvt->ctl_name = "F15h_M60h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_15H_M60H_NB_F1;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_15H_M60H_NB_F2;
+ pvt->ops->dbam_to_cs = f15_m60h_dbam_to_chip_select;
+ break;
+ case 0x13:
+ /* Richland is only client */
+ return -ENODEV;
+ default:
+ pvt->ctl_name = "F15h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_15H_NB_F1;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_15H_NB_F2;
+ pvt->ops->dbam_to_cs = f15_dbam_to_chip_select;
+ break;
+ }
+ break;
+
+ case 0x16:
+ switch (pvt->model) {
+ case 0x30:
+ pvt->ctl_name = "F16h_M30h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_16H_M30H_NB_F1;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_16H_M30H_NB_F2;
+ break;
+ default:
+ pvt->ctl_name = "F16h";
+ pvt->f1_id = PCI_DEVICE_ID_AMD_16H_NB_F1;
+ pvt->f2_id = PCI_DEVICE_ID_AMD_16H_NB_F2;
+ break;
+ }
+ break;
+
+ case 0x17:
+ switch (pvt->model) {
+ case 0x10 ... 0x2f:
+ pvt->ctl_name = "F17h_M10h";
+ break;
+ case 0x30 ... 0x3f:
+ pvt->ctl_name = "F17h_M30h";
+ pvt->max_mcs = 8;
+ break;
+ case 0x60 ... 0x6f:
+ pvt->ctl_name = "F17h_M60h";
+ break;
+ case 0x70 ... 0x7f:
+ pvt->ctl_name = "F17h_M70h";
+ break;
+ default:
+ pvt->ctl_name = "F17h";
+ break;
+ }
+ break;
+
+ case 0x18:
+ pvt->ctl_name = "F18h";
+ break;
+
+ case 0x19:
+ switch (pvt->model) {
+ case 0x00 ... 0x0f:
+ pvt->ctl_name = "F19h";
+ pvt->max_mcs = 8;
+ break;
+ case 0x10 ... 0x1f:
+ pvt->ctl_name = "F19h_M10h";
+ pvt->max_mcs = 12;
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ case 0x20 ... 0x2f:
+ pvt->ctl_name = "F19h_M20h";
+ break;
+ case 0x30 ... 0x3f:
+ if (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) {
+ pvt->ctl_name = "MI200";
+ pvt->max_mcs = 4;
+ pvt->ops = &gpu_ops;
+ } else {
+ pvt->ctl_name = "F19h_M30h";
+ pvt->max_mcs = 8;
+ }
+ break;
+ case 0x50 ... 0x5f:
+ pvt->ctl_name = "F19h_M50h";
+ break;
+ case 0x60 ... 0x6f:
+ pvt->ctl_name = "F19h_M60h";
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ case 0x70 ... 0x7f:
+ pvt->ctl_name = "F19h_M70h";
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ case 0xa0 ... 0xaf:
+ pvt->ctl_name = "F19h_MA0h";
+ pvt->max_mcs = 12;
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ }
+ break;
+
+ case 0x1A:
+ switch (pvt->model) {
+ case 0x00 ... 0x1f:
+ pvt->ctl_name = "F1Ah";
+ pvt->max_mcs = 12;
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ case 0x40 ... 0x4f:
+ pvt->ctl_name = "F1Ah_M40h";
+ pvt->flags.zn_regs_v2 = 1;
+ break;
+ }
+ break;
+
+ default:
+ amd64_err("Unsupported family!\n");
+ return -ENODEV;
+ }
+
+ return 0;
+}
+
+static const struct attribute_group *amd64_edac_attr_groups[] = {
+#ifdef CONFIG_EDAC_DEBUG
+ &dbg_group,
+ &inj_group,
+#endif
+ NULL
+};
+
+static int init_one_instance(struct amd64_pvt *pvt)
+{
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ int ret = -ENOMEM;
+
+ /*
+ * For Heterogeneous family EDAC CHIP_SELECT and CHANNEL layers should
+ * be swapped to fit into the layers.
+ */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
+ pvt->max_mcs : pvt->csels[0].b_cnt;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
+ pvt->csels[0].b_cnt : pvt->max_mcs;
+ layers[1].is_virt_csrow = false;
+
+ mci = edac_mc_alloc(pvt->mc_node_id, ARRAY_SIZE(layers), layers, 0);
+ if (!mci)
+ return ret;
+
+ mci->pvt_info = pvt;
+ mci->pdev = &pvt->F3->dev;
+
+ pvt->ops->setup_mci_misc_attrs(mci);
+
+ ret = -ENODEV;
+ if (edac_mc_add_mc_with_groups(mci, amd64_edac_attr_groups)) {
+ edac_dbg(1, "failed edac_mc_add_mc()\n");
+ edac_mc_free(mci);
+ return ret;
+ }
+
+ return 0;
+}
+
+static bool instance_has_memory(struct amd64_pvt *pvt)
+{
+ bool cs_enabled = false;
+ int cs = 0, dct = 0;
+
+ for (dct = 0; dct < pvt->max_mcs; dct++) {
+ for_each_chip_select(cs, dct, pvt)
+ cs_enabled |= csrow_enabled(cs, dct, pvt);
+ }
+
+ return cs_enabled;
+}
+
+static int probe_one_instance(unsigned int nid)
+{
+ struct pci_dev *F3 = node_to_amd_nb(nid)->misc;
+ struct amd64_pvt *pvt = NULL;
+ struct ecc_settings *s;
+ int ret;
+
+ ret = -ENOMEM;
+ s = kzalloc(sizeof(struct ecc_settings), GFP_KERNEL);
+ if (!s)
+ goto err_out;
+
+ ecc_stngs[nid] = s;
+
+ pvt = kzalloc(sizeof(struct amd64_pvt), GFP_KERNEL);
+ if (!pvt)
+ goto err_settings;
+
+ pvt->mc_node_id = nid;
+ pvt->F3 = F3;
+
+ ret = per_family_init(pvt);
+ if (ret < 0)
+ goto err_enable;
+
+ ret = pvt->ops->hw_info_get(pvt);
+ if (ret < 0)
+ goto err_enable;
+
+ ret = 0;
+ if (!instance_has_memory(pvt)) {
+ amd64_info("Node %d: No DIMMs detected.\n", nid);
+ goto err_enable;
+ }
+
+ if (!pvt->ops->ecc_enabled(pvt)) {
+ ret = -ENODEV;
+
+ if (!ecc_enable_override)
+ goto err_enable;
+
+ if (boot_cpu_data.x86 >= 0x17) {
+ amd64_warn("Forcing ECC on is not recommended on newer systems. Please enable ECC in BIOS.");
+ goto err_enable;
+ } else
+ amd64_warn("Forcing ECC on!\n");
+
+ if (!enable_ecc_error_reporting(s, nid, F3))
+ goto err_enable;
+ }
+
+ ret = init_one_instance(pvt);
+ if (ret < 0) {
+ amd64_err("Error probing instance: %d\n", nid);
+
+ if (boot_cpu_data.x86 < 0x17)
+ restore_ecc_error_reporting(s, nid, F3);
+
+ goto err_enable;
+ }
+
+ amd64_info("%s detected (node %d).\n", pvt->ctl_name, pvt->mc_node_id);
+
+ /* Display and decode various registers for debug purposes. */
+ pvt->ops->dump_misc_regs(pvt);
+
+ return ret;
+
+err_enable:
+ hw_info_put(pvt);
+ kfree(pvt);
+
+err_settings:
+ kfree(s);
+ ecc_stngs[nid] = NULL;
+
+err_out:
+ return ret;
+}
+
+static void remove_one_instance(unsigned int nid)
+{
+ struct pci_dev *F3 = node_to_amd_nb(nid)->misc;
+ struct ecc_settings *s = ecc_stngs[nid];
+ struct mem_ctl_info *mci;
+ struct amd64_pvt *pvt;
+
+ /* Remove from EDAC CORE tracking list */
+ mci = edac_mc_del_mc(&F3->dev);
+ if (!mci)
+ return;
+
+ pvt = mci->pvt_info;
+
+ restore_ecc_error_reporting(s, nid, F3);
+
+ kfree(ecc_stngs[nid]);
+ ecc_stngs[nid] = NULL;
+
+ /* Free the EDAC CORE resources */
+ mci->pvt_info = NULL;
+
+ hw_info_put(pvt);
+ kfree(pvt);
+ edac_mc_free(mci);
+}
+
+static void setup_pci_device(void)
+{
+ if (pci_ctl)
+ return;
+
+ pci_ctl = edac_pci_create_generic_ctl(pci_ctl_dev, EDAC_MOD_STR);
+ if (!pci_ctl) {
+ pr_warn("%s(): Unable to create PCI control\n", __func__);
+ pr_warn("%s(): PCI error report via EDAC not set\n", __func__);
+ }
+}
+
+static const struct x86_cpu_id amd64_cpuids[] = {
+ X86_MATCH_VENDOR_FAM(AMD, 0x0F, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x10, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x15, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x16, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x17, NULL),
+ X86_MATCH_VENDOR_FAM(HYGON, 0x18, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x19, NULL),
+ X86_MATCH_VENDOR_FAM(AMD, 0x1A, NULL),
+ { }
+};
+MODULE_DEVICE_TABLE(x86cpu, amd64_cpuids);
+
+static int __init amd64_edac_init(void)
+{
+ const char *owner;
+ int err = -ENODEV;
+ int i;
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ if (!x86_match_cpu(amd64_cpuids))
+ return -ENODEV;
+
+ if (!amd_nb_num())
+ return -ENODEV;
+
+ opstate_init();
+
+ err = -ENOMEM;
+ ecc_stngs = kcalloc(amd_nb_num(), sizeof(ecc_stngs[0]), GFP_KERNEL);
+ if (!ecc_stngs)
+ goto err_free;
+
+ msrs = msrs_alloc();
+ if (!msrs)
+ goto err_free;
+
+ for (i = 0; i < amd_nb_num(); i++) {
+ err = probe_one_instance(i);
+ if (err) {
+ /* unwind properly */
+ while (--i >= 0)
+ remove_one_instance(i);
+
+ goto err_pci;
+ }
+ }
+
+ if (!edac_has_mcs()) {
+ err = -ENODEV;
+ goto err_pci;
+ }
+
+ /* register stuff with EDAC MCE */
+ if (boot_cpu_data.x86 >= 0x17) {
+ amd_register_ecc_decoder(decode_umc_error);
+ } else {
+ amd_register_ecc_decoder(decode_bus_error);
+ setup_pci_device();
+ }
+
+#ifdef CONFIG_X86_32
+ amd64_err("%s on 32-bit is unsupported. USE AT YOUR OWN RISK!\n", EDAC_MOD_STR);
+#endif
+
+ return 0;
+
+err_pci:
+ pci_ctl_dev = NULL;
+
+ msrs_free(msrs);
+ msrs = NULL;
+
+err_free:
+ kfree(ecc_stngs);
+ ecc_stngs = NULL;
+
+ return err;
+}
+
+static void __exit amd64_edac_exit(void)
+{
+ int i;
+
+ if (pci_ctl)
+ edac_pci_release_generic_ctl(pci_ctl);
+
+ /* unregister from EDAC MCE */
+ if (boot_cpu_data.x86 >= 0x17)
+ amd_unregister_ecc_decoder(decode_umc_error);
+ else
+ amd_unregister_ecc_decoder(decode_bus_error);
+
+ for (i = 0; i < amd_nb_num(); i++)
+ remove_one_instance(i);
+
+ kfree(ecc_stngs);
+ ecc_stngs = NULL;
+
+ pci_ctl_dev = NULL;
+
+ msrs_free(msrs);
+ msrs = NULL;
+}
+
+module_init(amd64_edac_init);
+module_exit(amd64_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("SoftwareBitMaker: Doug Thompson, Dave Peterson, Thayne Harbaugh; AMD");
+MODULE_DESCRIPTION("MC support for AMD64 memory controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h
new file mode 100644
index 0000000000..5a4e4a5968
--- /dev/null
+++ b/drivers/edac/amd64_edac.h
@@ -0,0 +1,529 @@
+/*
+ * AMD64 class Memory Controller kernel module
+ *
+ * Copyright (c) 2009 SoftwareBitMaker.
+ * Copyright (c) 2009-15 Advanced Micro Devices, Inc.
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ */
+
+#include <linux/module.h>
+#include <linux/ctype.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/mmzone.h>
+#include <linux/edac.h>
+#include <linux/bitfield.h>
+#include <asm/cpu_device_id.h>
+#include <asm/msr.h>
+#include "edac_module.h"
+#include "mce_amd.h"
+
+#define amd64_info(fmt, arg...) \
+ edac_printk(KERN_INFO, "amd64", fmt, ##arg)
+
+#define amd64_warn(fmt, arg...) \
+ edac_printk(KERN_WARNING, "amd64", "Warning: " fmt, ##arg)
+
+#define amd64_err(fmt, arg...) \
+ edac_printk(KERN_ERR, "amd64", "Error: " fmt, ##arg)
+
+#define amd64_mc_warn(mci, fmt, arg...) \
+ edac_mc_chipset_printk(mci, KERN_WARNING, "amd64", fmt, ##arg)
+
+#define amd64_mc_err(mci, fmt, arg...) \
+ edac_mc_chipset_printk(mci, KERN_ERR, "amd64", fmt, ##arg)
+
+/*
+ * Throughout the comments in this code, the following terms are used:
+ *
+ * SysAddr, DramAddr, and InputAddr
+ *
+ * These terms come directly from the amd64 documentation
+ * (AMD publication #26094). They are defined as follows:
+ *
+ * SysAddr:
+ * This is a physical address generated by a CPU core or a device
+ * doing DMA. If generated by a CPU core, a SysAddr is the result of
+ * a virtual to physical address translation by the CPU core's address
+ * translation mechanism (MMU).
+ *
+ * DramAddr:
+ * A DramAddr is derived from a SysAddr by subtracting an offset that
+ * depends on which node the SysAddr maps to and whether the SysAddr
+ * is within a range affected by memory hoisting. The DRAM Base
+ * (section 3.4.4.1) and DRAM Limit (section 3.4.4.2) registers
+ * determine which node a SysAddr maps to.
+ *
+ * If the DRAM Hole Address Register (DHAR) is enabled and the SysAddr
+ * is within the range of addresses specified by this register, then
+ * a value x from the DHAR is subtracted from the SysAddr to produce a
+ * DramAddr. Here, x represents the base address for the node that
+ * the SysAddr maps to plus an offset due to memory hoisting. See
+ * section 3.4.8 and the comments in amd64_get_dram_hole_info() and
+ * sys_addr_to_dram_addr() below for more information.
+ *
+ * If the SysAddr is not affected by the DHAR then a value y is
+ * subtracted from the SysAddr to produce a DramAddr. Here, y is the
+ * base address for the node that the SysAddr maps to. See section
+ * 3.4.4 and the comments in sys_addr_to_dram_addr() below for more
+ * information.
+ *
+ * InputAddr:
+ * A DramAddr is translated to an InputAddr before being passed to the
+ * memory controller for the node that the DramAddr is associated
+ * with. The memory controller then maps the InputAddr to a csrow.
+ * If node interleaving is not in use, then the InputAddr has the same
+ * value as the DramAddr. Otherwise, the InputAddr is produced by
+ * discarding the bits used for node interleaving from the DramAddr.
+ * See section 3.4.4 for more information.
+ *
+ * The memory controller for a given node uses its DRAM CS Base and
+ * DRAM CS Mask registers to map an InputAddr to a csrow. See
+ * sections 3.5.4 and 3.5.5 for more information.
+ */
+
+#define EDAC_MOD_STR "amd64_edac"
+
+/* Extended Model from CPUID, for CPU Revision numbers */
+#define K8_REV_D 1
+#define K8_REV_E 2
+#define K8_REV_F 4
+
+/* Hardware limit on ChipSelect rows per MC and processors per system */
+#define NUM_CHIPSELECTS 8
+#define DRAM_RANGES 8
+#define NUM_CONTROLLERS 12
+
+#define ON true
+#define OFF false
+
+/*
+ * PCI-defined configuration space registers
+ */
+#define PCI_DEVICE_ID_AMD_15H_NB_F1 0x1601
+#define PCI_DEVICE_ID_AMD_15H_NB_F2 0x1602
+#define PCI_DEVICE_ID_AMD_15H_M30H_NB_F1 0x141b
+#define PCI_DEVICE_ID_AMD_15H_M30H_NB_F2 0x141c
+#define PCI_DEVICE_ID_AMD_15H_M60H_NB_F1 0x1571
+#define PCI_DEVICE_ID_AMD_15H_M60H_NB_F2 0x1572
+#define PCI_DEVICE_ID_AMD_16H_NB_F1 0x1531
+#define PCI_DEVICE_ID_AMD_16H_NB_F2 0x1532
+#define PCI_DEVICE_ID_AMD_16H_M30H_NB_F1 0x1581
+#define PCI_DEVICE_ID_AMD_16H_M30H_NB_F2 0x1582
+
+/*
+ * Function 1 - Address Map
+ */
+#define DRAM_BASE_LO 0x40
+#define DRAM_LIMIT_LO 0x44
+
+/*
+ * F15 M30h D18F1x2[1C:00]
+ */
+#define DRAM_CONT_BASE 0x200
+#define DRAM_CONT_LIMIT 0x204
+
+/*
+ * F15 M30h D18F1x2[4C:40]
+ */
+#define DRAM_CONT_HIGH_OFF 0x240
+
+#define dram_rw(pvt, i) ((u8)(pvt->ranges[i].base.lo & 0x3))
+#define dram_intlv_sel(pvt, i) ((u8)((pvt->ranges[i].lim.lo >> 8) & 0x7))
+#define dram_dst_node(pvt, i) ((u8)(pvt->ranges[i].lim.lo & 0x7))
+
+#define DHAR 0xf0
+#define dhar_mem_hoist_valid(pvt) ((pvt)->dhar & BIT(1))
+#define dhar_base(pvt) ((pvt)->dhar & 0xff000000)
+#define k8_dhar_offset(pvt) (((pvt)->dhar & 0x0000ff00) << 16)
+
+ /* NOTE: Extra mask bit vs K8 */
+#define f10_dhar_offset(pvt) (((pvt)->dhar & 0x0000ff80) << 16)
+
+#define DCT_CFG_SEL 0x10C
+
+#define DRAM_LOCAL_NODE_BASE 0x120
+#define DRAM_LOCAL_NODE_LIM 0x124
+
+#define DRAM_BASE_HI 0x140
+#define DRAM_LIMIT_HI 0x144
+
+
+/*
+ * Function 2 - DRAM controller
+ */
+#define DCSB0 0x40
+#define DCSB1 0x140
+#define DCSB_CS_ENABLE BIT(0)
+
+#define DCSM0 0x60
+#define DCSM1 0x160
+
+#define csrow_enabled(i, dct, pvt) ((pvt)->csels[(dct)].csbases[(i)] & DCSB_CS_ENABLE)
+#define csrow_sec_enabled(i, dct, pvt) ((pvt)->csels[(dct)].csbases_sec[(i)] & DCSB_CS_ENABLE)
+
+#define DRAM_CONTROL 0x78
+
+#define DBAM0 0x80
+#define DBAM1 0x180
+
+/* Extract the DIMM 'type' on the i'th DIMM from the DBAM reg value passed */
+#define DBAM_DIMM(i, reg) ((((reg) >> (4*(i)))) & 0xF)
+
+#define DBAM_MAX_VALUE 11
+
+#define DCLR0 0x90
+#define DCLR1 0x190
+#define REVE_WIDTH_128 BIT(16)
+#define WIDTH_128 BIT(11)
+
+#define DCHR0 0x94
+#define DCHR1 0x194
+#define DDR3_MODE BIT(8)
+
+#define DCT_SEL_LO 0x110
+#define dct_high_range_enabled(pvt) ((pvt)->dct_sel_lo & BIT(0))
+#define dct_interleave_enabled(pvt) ((pvt)->dct_sel_lo & BIT(2))
+
+#define dct_ganging_enabled(pvt) ((boot_cpu_data.x86 == 0x10) && ((pvt)->dct_sel_lo & BIT(4)))
+
+#define dct_data_intlv_enabled(pvt) ((pvt)->dct_sel_lo & BIT(5))
+#define dct_memory_cleared(pvt) ((pvt)->dct_sel_lo & BIT(10))
+
+#define SWAP_INTLV_REG 0x10c
+
+#define DCT_SEL_HI 0x114
+
+#define F15H_M60H_SCRCTRL 0x1C8
+
+/*
+ * Function 3 - Misc Control
+ */
+#define NBCTL 0x40
+
+#define NBCFG 0x44
+#define NBCFG_CHIPKILL BIT(23)
+#define NBCFG_ECC_ENABLE BIT(22)
+
+/* F3x48: NBSL */
+#define F10_NBSL_EXT_ERR_ECC 0x8
+#define NBSL_PP_OBS 0x2
+
+#define SCRCTRL 0x58
+
+#define F10_ONLINE_SPARE 0xB0
+#define online_spare_swap_done(pvt, c) (((pvt)->online_spare >> (1 + 2 * (c))) & 0x1)
+#define online_spare_bad_dramcs(pvt, c) (((pvt)->online_spare >> (4 + 4 * (c))) & 0x7)
+
+#define F10_NB_ARRAY_ADDR 0xB8
+#define F10_NB_ARRAY_DRAM BIT(31)
+
+/* Bits [2:1] are used to select 16-byte section within a 64-byte cacheline */
+#define SET_NB_ARRAY_ADDR(section) (((section) & 0x3) << 1)
+
+#define F10_NB_ARRAY_DATA 0xBC
+#define F10_NB_ARR_ECC_WR_REQ BIT(17)
+#define SET_NB_DRAM_INJECTION_WRITE(inj) \
+ (BIT(((inj.word) & 0xF) + 20) | \
+ F10_NB_ARR_ECC_WR_REQ | inj.bit_map)
+#define SET_NB_DRAM_INJECTION_READ(inj) \
+ (BIT(((inj.word) & 0xF) + 20) | \
+ BIT(16) | inj.bit_map)
+
+
+#define NBCAP 0xE8
+#define NBCAP_CHIPKILL BIT(4)
+#define NBCAP_SECDED BIT(3)
+#define NBCAP_DCT_DUAL BIT(0)
+
+#define EXT_NB_MCA_CFG 0x180
+
+/* MSRs */
+#define MSR_MCGCTL_NBE BIT(4)
+
+/* F17h */
+
+/* F0: */
+#define DF_DHAR 0x104
+
+/* UMC CH register offsets */
+#define UMCCH_BASE_ADDR 0x0
+#define UMCCH_BASE_ADDR_SEC 0x10
+#define UMCCH_ADDR_MASK 0x20
+#define UMCCH_ADDR_MASK_SEC 0x28
+#define UMCCH_ADDR_MASK_SEC_DDR5 0x30
+#define UMCCH_ADDR_CFG 0x30
+#define UMCCH_ADDR_CFG_DDR5 0x40
+#define UMCCH_DIMM_CFG 0x80
+#define UMCCH_DIMM_CFG_DDR5 0x90
+#define UMCCH_UMC_CFG 0x100
+#define UMCCH_SDP_CTRL 0x104
+#define UMCCH_ECC_CTRL 0x14C
+#define UMCCH_ECC_BAD_SYMBOL 0xD90
+#define UMCCH_UMC_CAP 0xDF0
+#define UMCCH_UMC_CAP_HI 0xDF4
+
+/* UMC CH bitfields */
+#define UMC_ECC_CHIPKILL_CAP BIT(31)
+#define UMC_ECC_ENABLED BIT(30)
+
+#define UMC_SDP_INIT BIT(31)
+
+/* Error injection control structure */
+struct error_injection {
+ u32 section;
+ u32 word;
+ u32 bit_map;
+};
+
+/* low and high part of PCI config space regs */
+struct reg_pair {
+ u32 lo, hi;
+};
+
+/*
+ * See F1x[1, 0][7C:40] DRAM Base/Limit Registers
+ */
+struct dram_range {
+ struct reg_pair base;
+ struct reg_pair lim;
+};
+
+/* A DCT chip selects collection */
+struct chip_select {
+ u32 csbases[NUM_CHIPSELECTS];
+ u32 csbases_sec[NUM_CHIPSELECTS];
+ u8 b_cnt;
+
+ u32 csmasks[NUM_CHIPSELECTS];
+ u32 csmasks_sec[NUM_CHIPSELECTS];
+ u8 m_cnt;
+};
+
+struct amd64_umc {
+ u32 dimm_cfg; /* DIMM Configuration reg */
+ u32 umc_cfg; /* Configuration reg */
+ u32 sdp_ctrl; /* SDP Control reg */
+ u32 ecc_ctrl; /* DRAM ECC Control reg */
+ u32 umc_cap_hi; /* Capabilities High reg */
+
+ /* cache the dram_type */
+ enum mem_type dram_type;
+};
+
+struct amd64_family_flags {
+ /*
+ * Indicates that the system supports the new register offsets, etc.
+ * first introduced with Family 19h Model 10h.
+ */
+ __u64 zn_regs_v2 : 1,
+
+ __reserved : 63;
+};
+
+struct amd64_pvt {
+ struct low_ops *ops;
+
+ /* pci_device handles which we utilize */
+ struct pci_dev *F1, *F2, *F3;
+
+ u16 mc_node_id; /* MC index of this MC node */
+ u8 fam; /* CPU family */
+ u8 model; /* ... model */
+ u8 stepping; /* ... stepping */
+
+ int ext_model; /* extended model value of this node */
+
+ /* Raw registers */
+ u32 dclr0; /* DRAM Configuration Low DCT0 reg */
+ u32 dclr1; /* DRAM Configuration Low DCT1 reg */
+ u32 dchr0; /* DRAM Configuration High DCT0 reg */
+ u32 dchr1; /* DRAM Configuration High DCT1 reg */
+ u32 nbcap; /* North Bridge Capabilities */
+ u32 nbcfg; /* F10 North Bridge Configuration */
+ u32 ext_nbcfg; /* Extended F10 North Bridge Configuration */
+ u32 dhar; /* DRAM Hoist reg */
+ u32 dbam0; /* DRAM Base Address Mapping reg for DCT0 */
+ u32 dbam1; /* DRAM Base Address Mapping reg for DCT1 */
+
+ /* one for each DCT/UMC */
+ struct chip_select csels[NUM_CONTROLLERS];
+
+ /* DRAM base and limit pairs F1x[78,70,68,60,58,50,48,40] */
+ struct dram_range ranges[DRAM_RANGES];
+
+ u64 top_mem; /* top of memory below 4GB */
+ u64 top_mem2; /* top of memory above 4GB */
+
+ u32 dct_sel_lo; /* DRAM Controller Select Low */
+ u32 dct_sel_hi; /* DRAM Controller Select High */
+ u32 online_spare; /* On-Line spare Reg */
+
+ /* x4, x8, or x16 syndromes in use */
+ u8 ecc_sym_sz;
+
+ const char *ctl_name;
+ u16 f1_id, f2_id;
+ /* Maximum number of memory controllers per die/node. */
+ u8 max_mcs;
+
+ struct amd64_family_flags flags;
+ /* place to store error injection parameters prior to issue */
+ struct error_injection injection;
+
+ /*
+ * cache the dram_type
+ *
+ * NOTE: Don't use this for Family 17h and later.
+ * Use dram_type in struct amd64_umc instead.
+ */
+ enum mem_type dram_type;
+
+ struct amd64_umc *umc; /* UMC registers */
+};
+
+enum err_codes {
+ DECODE_OK = 0,
+ ERR_NODE = -1,
+ ERR_CSROW = -2,
+ ERR_CHANNEL = -3,
+ ERR_SYND = -4,
+ ERR_NORM_ADDR = -5,
+};
+
+struct err_info {
+ int err_code;
+ struct mem_ctl_info *src_mci;
+ int csrow;
+ int channel;
+ u16 syndrome;
+ u32 page;
+ u32 offset;
+};
+
+static inline u32 get_umc_base(u8 channel)
+{
+ /* chY: 0xY50000 */
+ return 0x50000 + (channel << 20);
+}
+
+static inline u64 get_dram_base(struct amd64_pvt *pvt, u8 i)
+{
+ u64 addr = ((u64)pvt->ranges[i].base.lo & 0xffff0000) << 8;
+
+ if (boot_cpu_data.x86 == 0xf)
+ return addr;
+
+ return (((u64)pvt->ranges[i].base.hi & 0x000000ff) << 40) | addr;
+}
+
+static inline u64 get_dram_limit(struct amd64_pvt *pvt, u8 i)
+{
+ u64 lim = (((u64)pvt->ranges[i].lim.lo & 0xffff0000) << 8) | 0x00ffffff;
+
+ if (boot_cpu_data.x86 == 0xf)
+ return lim;
+
+ return (((u64)pvt->ranges[i].lim.hi & 0x000000ff) << 40) | lim;
+}
+
+static inline u16 extract_syndrome(u64 status)
+{
+ return ((status >> 47) & 0xff) | ((status >> 16) & 0xff00);
+}
+
+static inline u8 dct_sel_interleave_addr(struct amd64_pvt *pvt)
+{
+ if (pvt->fam == 0x15 && pvt->model >= 0x30)
+ return (((pvt->dct_sel_hi >> 9) & 0x1) << 2) |
+ ((pvt->dct_sel_lo >> 6) & 0x3);
+
+ return ((pvt)->dct_sel_lo >> 6) & 0x3;
+}
+/*
+ * per-node ECC settings descriptor
+ */
+struct ecc_settings {
+ u32 old_nbctl;
+ bool nbctl_valid;
+
+ struct flags {
+ unsigned long nb_mce_enable:1;
+ unsigned long nb_ecc_prev:1;
+ } flags;
+};
+
+/*
+ * Each of the PCI Device IDs types have their own set of hardware accessor
+ * functions and per device encoding/decoding logic.
+ */
+struct low_ops {
+ void (*map_sysaddr_to_csrow)(struct mem_ctl_info *mci, u64 sys_addr,
+ struct err_info *err);
+ int (*dbam_to_cs)(struct amd64_pvt *pvt, u8 dct,
+ unsigned int cs_mode, int cs_mask_nr);
+ int (*hw_info_get)(struct amd64_pvt *pvt);
+ bool (*ecc_enabled)(struct amd64_pvt *pvt);
+ void (*setup_mci_misc_attrs)(struct mem_ctl_info *mci);
+ void (*dump_misc_regs)(struct amd64_pvt *pvt);
+ void (*get_err_info)(struct mce *m, struct err_info *err);
+};
+
+int __amd64_read_pci_cfg_dword(struct pci_dev *pdev, int offset,
+ u32 *val, const char *func);
+int __amd64_write_pci_cfg_dword(struct pci_dev *pdev, int offset,
+ u32 val, const char *func);
+
+#define amd64_read_pci_cfg(pdev, offset, val) \
+ __amd64_read_pci_cfg_dword(pdev, offset, val, __func__)
+
+#define amd64_write_pci_cfg(pdev, offset, val) \
+ __amd64_write_pci_cfg_dword(pdev, offset, val, __func__)
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+/* Injection helpers */
+static inline void disable_caches(void *dummy)
+{
+ write_cr0(read_cr0() | X86_CR0_CD);
+ wbinvd();
+}
+
+static inline void enable_caches(void *dummy)
+{
+ write_cr0(read_cr0() & ~X86_CR0_CD);
+}
+
+static inline u8 dram_intlv_en(struct amd64_pvt *pvt, unsigned int i)
+{
+ if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+ u32 tmp;
+ amd64_read_pci_cfg(pvt->F1, DRAM_CONT_LIMIT, &tmp);
+ return (u8) tmp & 0xF;
+ }
+ return (u8) (pvt->ranges[i].base.lo >> 8) & 0x7;
+}
+
+static inline u8 dhar_valid(struct amd64_pvt *pvt)
+{
+ if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+ u32 tmp;
+ amd64_read_pci_cfg(pvt->F1, DRAM_CONT_BASE, &tmp);
+ return (tmp >> 1) & BIT(0);
+ }
+ return (pvt)->dhar & BIT(0);
+}
+
+static inline u32 dct_sel_baseaddr(struct amd64_pvt *pvt)
+{
+ if (pvt->fam == 0x15 && pvt->model >= 0x30) {
+ u32 tmp;
+ amd64_read_pci_cfg(pvt->F1, DRAM_CONT_BASE, &tmp);
+ return (tmp >> 11) & 0x1FFF;
+ }
+ return (pvt)->dct_sel_lo & 0xFFFFF800;
+}
diff --git a/drivers/edac/amd76x_edac.c b/drivers/edac/amd76x_edac.c
new file mode 100644
index 0000000000..2a49f68a7c
--- /dev/null
+++ b/drivers/edac/amd76x_edac.c
@@ -0,0 +1,375 @@
+/*
+ * AMD 76x Memory Controller kernel module
+ * (C) 2003 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * $Id: edac_amd76x.c,v 1.4.2.5 2005/10/05 00:43:44 dsp_llnl Exp $
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "amd76x_edac"
+
+#define amd76x_printk(level, fmt, arg...) \
+ edac_printk(level, "amd76x", fmt, ##arg)
+
+#define amd76x_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "amd76x", fmt, ##arg)
+
+#define AMD76X_NR_CSROWS 8
+#define AMD76X_NR_DIMMS 4
+
+/* AMD 76x register addresses - device 0 function 0 - PCI bridge */
+
+#define AMD76X_ECC_MODE_STATUS 0x48 /* Mode and status of ECC (32b)
+ *
+ * 31:16 reserved
+ * 15:14 SERR enabled: x1=ue 1x=ce
+ * 13 reserved
+ * 12 diag: disabled, enabled
+ * 11:10 mode: dis, EC, ECC, ECC+scrub
+ * 9:8 status: x1=ue 1x=ce
+ * 7:4 UE cs row
+ * 3:0 CE cs row
+ */
+
+#define AMD76X_DRAM_MODE_STATUS 0x58 /* DRAM Mode and status (32b)
+ *
+ * 31:26 clock disable 5 - 0
+ * 25 SDRAM init
+ * 24 reserved
+ * 23 mode register service
+ * 22:21 suspend to RAM
+ * 20 burst refresh enable
+ * 19 refresh disable
+ * 18 reserved
+ * 17:16 cycles-per-refresh
+ * 15:8 reserved
+ * 7:0 x4 mode enable 7 - 0
+ */
+
+#define AMD76X_MEM_BASE_ADDR 0xC0 /* Memory base address (8 x 32b)
+ *
+ * 31:23 chip-select base
+ * 22:16 reserved
+ * 15:7 chip-select mask
+ * 6:3 reserved
+ * 2:1 address mode
+ * 0 chip-select enable
+ */
+
+struct amd76x_error_info {
+ u32 ecc_mode_status;
+};
+
+enum amd76x_chips {
+ AMD761 = 0,
+ AMD762
+};
+
+struct amd76x_dev_info {
+ const char *ctl_name;
+};
+
+static const struct amd76x_dev_info amd76x_devs[] = {
+ [AMD761] = {
+ .ctl_name = "AMD761"},
+ [AMD762] = {
+ .ctl_name = "AMD762"},
+};
+
+static struct edac_pci_ctl_info *amd76x_pci;
+
+/**
+ * amd76x_get_error_info - fetch error information
+ * @mci: Memory controller
+ * @info: Info to fill in
+ *
+ * Fetch and store the AMD76x ECC status. Clear pending status
+ * on the chip so that further errors will be reported
+ */
+static void amd76x_get_error_info(struct mem_ctl_info *mci,
+ struct amd76x_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+ pci_read_config_dword(pdev, AMD76X_ECC_MODE_STATUS,
+ &info->ecc_mode_status);
+
+ if (info->ecc_mode_status & BIT(8))
+ pci_write_bits32(pdev, AMD76X_ECC_MODE_STATUS,
+ (u32) BIT(8), (u32) BIT(8));
+
+ if (info->ecc_mode_status & BIT(9))
+ pci_write_bits32(pdev, AMD76X_ECC_MODE_STATUS,
+ (u32) BIT(9), (u32) BIT(9));
+}
+
+/**
+ * amd76x_process_error_info - Error check
+ * @mci: Memory controller
+ * @info: Previously fetched information from chip
+ * @handle_errors: 1 if we should do recovery
+ *
+ * Process the chip state and decide if an error has occurred.
+ * A return of 1 indicates an error. Also if handle_errors is true
+ * then attempt to handle and clean up after the error
+ */
+static int amd76x_process_error_info(struct mem_ctl_info *mci,
+ struct amd76x_error_info *info,
+ int handle_errors)
+{
+ int error_found;
+ u32 row;
+
+ error_found = 0;
+
+ /*
+ * Check for an uncorrectable error
+ */
+ if (info->ecc_mode_status & BIT(8)) {
+ error_found = 1;
+
+ if (handle_errors) {
+ row = (info->ecc_mode_status >> 4) & 0xf;
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ mci->csrows[row]->first_page, 0, 0,
+ row, 0, -1,
+ mci->ctl_name, "");
+ }
+ }
+
+ /*
+ * Check for a correctable error
+ */
+ if (info->ecc_mode_status & BIT(9)) {
+ error_found = 1;
+
+ if (handle_errors) {
+ row = info->ecc_mode_status & 0xf;
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ mci->csrows[row]->first_page, 0, 0,
+ row, 0, -1,
+ mci->ctl_name, "");
+ }
+ }
+
+ return error_found;
+}
+
+/**
+ * amd76x_check - Poll the controller
+ * @mci: Memory controller
+ *
+ * Called by the poll handlers this function reads the status
+ * from the controller and checks for errors.
+ */
+static void amd76x_check(struct mem_ctl_info *mci)
+{
+ struct amd76x_error_info info;
+ amd76x_get_error_info(mci, &info);
+ amd76x_process_error_info(mci, &info, 1);
+}
+
+static void amd76x_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
+ enum edac_type edac_mode)
+{
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ u32 mba, mba_base, mba_mask, dms;
+ int index;
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ /* find the DRAM Chip Select Base address and mask */
+ pci_read_config_dword(pdev,
+ AMD76X_MEM_BASE_ADDR + (index * 4), &mba);
+
+ if (!(mba & BIT(0)))
+ continue;
+
+ mba_base = mba & 0xff800000UL;
+ mba_mask = ((mba & 0xff80) << 16) | 0x7fffffUL;
+ pci_read_config_dword(pdev, AMD76X_DRAM_MODE_STATUS, &dms);
+ csrow->first_page = mba_base >> PAGE_SHIFT;
+ dimm->nr_pages = (mba_mask + 1) >> PAGE_SHIFT;
+ csrow->last_page = csrow->first_page + dimm->nr_pages - 1;
+ csrow->page_mask = mba_mask >> PAGE_SHIFT;
+ dimm->grain = dimm->nr_pages << PAGE_SHIFT;
+ dimm->mtype = MEM_RDDR;
+ dimm->dtype = ((dms >> index) & 0x1) ? DEV_X4 : DEV_UNKNOWN;
+ dimm->edac_mode = edac_mode;
+ }
+}
+
+/**
+ * amd76x_probe1 - Perform set up for detected device
+ * @pdev; PCI device detected
+ * @dev_idx: Device type index
+ *
+ * We have found an AMD76x and now need to set up the memory
+ * controller status reporting. We configure and set up the
+ * memory controller reporting and claim the device.
+ */
+static int amd76x_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ static const enum edac_type ems_modes[] = {
+ EDAC_NONE,
+ EDAC_EC,
+ EDAC_SECDED,
+ EDAC_SECDED
+ };
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ u32 ems;
+ u32 ems_mode;
+ struct amd76x_error_info discard;
+
+ edac_dbg(0, "\n");
+ pci_read_config_dword(pdev, AMD76X_ECC_MODE_STATUS, &ems);
+ ems_mode = (ems >> 10) & 0x3;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = AMD76X_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "mci = %p\n", mci);
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_RDDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ mci->edac_cap = ems_mode ?
+ (EDAC_FLAG_EC | EDAC_FLAG_SECDED) : EDAC_FLAG_NONE;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = amd76x_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = amd76x_check;
+ mci->ctl_page_to_phys = NULL;
+
+ amd76x_init_csrows(mci, pdev, ems_modes[ems_mode]);
+ amd76x_get_error_info(mci, &discard); /* clear counters */
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* allocating generic PCI control info */
+ amd76x_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!amd76x_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int amd76x_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ edac_dbg(0, "\n");
+
+ /* don't need to call pci_enable_device() */
+ return amd76x_probe1(pdev, ent->driver_data);
+}
+
+/**
+ * amd76x_remove_one - driver shutdown
+ * @pdev: PCI device being handed back
+ *
+ * Called when the driver is unloaded. Find the matching mci
+ * structure for the device then delete the mci and free the
+ * resources.
+ */
+static void amd76x_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (amd76x_pci)
+ edac_pci_release_generic_ctl(amd76x_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id amd76x_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(AMD, FE_GATE_700C), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ AMD762},
+ {
+ PCI_VEND_DEV(AMD, FE_GATE_700E), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ AMD761},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, amd76x_pci_tbl);
+
+static struct pci_driver amd76x_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = amd76x_init_one,
+ .remove = amd76x_remove_one,
+ .id_table = amd76x_pci_tbl,
+};
+
+static int __init amd76x_init(void)
+{
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ return pci_register_driver(&amd76x_driver);
+}
+
+static void __exit amd76x_exit(void)
+{
+ pci_unregister_driver(&amd76x_driver);
+}
+
+module_init(amd76x_init);
+module_exit(amd76x_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh");
+MODULE_DESCRIPTION("MC support for AMD 76x memory controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/amd8111_edac.c b/drivers/edac/amd8111_edac.c
new file mode 100644
index 0000000000..ca718f63fc
--- /dev/null
+++ b/drivers/edac/amd8111_edac.c
@@ -0,0 +1,597 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * amd8111_edac.c, AMD8111 Hyper Transport chip EDAC kernel module
+ *
+ * Copyright (c) 2008 Wind River Systems, Inc.
+ *
+ * Authors: Cao Qingtao <qingtao.cao@windriver.com>
+ * Benjamin Walsh <benjamin.walsh@windriver.com>
+ * Hu Yongqi <yongqi.hu@windriver.com>
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/bitops.h>
+#include <linux/edac.h>
+#include <linux/pci_ids.h>
+#include <asm/io.h>
+
+#include "edac_module.h"
+#include "amd8111_edac.h"
+
+#define AMD8111_EDAC_REVISION " Ver: 1.0.0"
+#define AMD8111_EDAC_MOD_STR "amd8111_edac"
+
+#define PCI_DEVICE_ID_AMD_8111_PCI 0x7460
+
+enum amd8111_edac_devs {
+ LPC_BRIDGE = 0,
+};
+
+enum amd8111_edac_pcis {
+ PCI_BRIDGE = 0,
+};
+
+/* Wrapper functions for accessing PCI configuration space */
+static int edac_pci_read_dword(struct pci_dev *dev, int reg, u32 *val32)
+{
+ int ret;
+
+ ret = pci_read_config_dword(dev, reg, val32);
+ if (ret != 0)
+ printk(KERN_ERR AMD8111_EDAC_MOD_STR
+ " PCI Access Read Error at 0x%x\n", reg);
+
+ return ret;
+}
+
+static void edac_pci_read_byte(struct pci_dev *dev, int reg, u8 *val8)
+{
+ int ret;
+
+ ret = pci_read_config_byte(dev, reg, val8);
+ if (ret != 0)
+ printk(KERN_ERR AMD8111_EDAC_MOD_STR
+ " PCI Access Read Error at 0x%x\n", reg);
+}
+
+static void edac_pci_write_dword(struct pci_dev *dev, int reg, u32 val32)
+{
+ int ret;
+
+ ret = pci_write_config_dword(dev, reg, val32);
+ if (ret != 0)
+ printk(KERN_ERR AMD8111_EDAC_MOD_STR
+ " PCI Access Write Error at 0x%x\n", reg);
+}
+
+static void edac_pci_write_byte(struct pci_dev *dev, int reg, u8 val8)
+{
+ int ret;
+
+ ret = pci_write_config_byte(dev, reg, val8);
+ if (ret != 0)
+ printk(KERN_ERR AMD8111_EDAC_MOD_STR
+ " PCI Access Write Error at 0x%x\n", reg);
+}
+
+/*
+ * device-specific methods for amd8111 PCI Bridge Controller
+ *
+ * Error Reporting and Handling for amd8111 chipset could be found
+ * in its datasheet 3.1.2 section, P37
+ */
+static void amd8111_pci_bridge_init(struct amd8111_pci_info *pci_info)
+{
+ u32 val32;
+ struct pci_dev *dev = pci_info->dev;
+
+ /* First clear error detection flags on the host interface */
+
+ /* Clear SSE/SMA/STA flags in the global status register*/
+ edac_pci_read_dword(dev, REG_PCI_STSCMD, &val32);
+ if (val32 & PCI_STSCMD_CLEAR_MASK)
+ edac_pci_write_dword(dev, REG_PCI_STSCMD, val32);
+
+ /* Clear CRC and Link Fail flags in HT Link Control reg */
+ edac_pci_read_dword(dev, REG_HT_LINK, &val32);
+ if (val32 & HT_LINK_CLEAR_MASK)
+ edac_pci_write_dword(dev, REG_HT_LINK, val32);
+
+ /* Second clear all fault on the secondary interface */
+
+ /* Clear error flags in the memory-base limit reg. */
+ edac_pci_read_dword(dev, REG_MEM_LIM, &val32);
+ if (val32 & MEM_LIMIT_CLEAR_MASK)
+ edac_pci_write_dword(dev, REG_MEM_LIM, val32);
+
+ /* Clear Discard Timer Expired flag in Interrupt/Bridge Control reg */
+ edac_pci_read_dword(dev, REG_PCI_INTBRG_CTRL, &val32);
+ if (val32 & PCI_INTBRG_CTRL_CLEAR_MASK)
+ edac_pci_write_dword(dev, REG_PCI_INTBRG_CTRL, val32);
+
+ /* Last enable error detections */
+ if (edac_op_state == EDAC_OPSTATE_POLL) {
+ /* Enable System Error reporting in global status register */
+ edac_pci_read_dword(dev, REG_PCI_STSCMD, &val32);
+ val32 |= PCI_STSCMD_SERREN;
+ edac_pci_write_dword(dev, REG_PCI_STSCMD, val32);
+
+ /* Enable CRC Sync flood packets to HyperTransport Link */
+ edac_pci_read_dword(dev, REG_HT_LINK, &val32);
+ val32 |= HT_LINK_CRCFEN;
+ edac_pci_write_dword(dev, REG_HT_LINK, val32);
+
+ /* Enable SSE reporting etc in Interrupt control reg */
+ edac_pci_read_dword(dev, REG_PCI_INTBRG_CTRL, &val32);
+ val32 |= PCI_INTBRG_CTRL_POLL_MASK;
+ edac_pci_write_dword(dev, REG_PCI_INTBRG_CTRL, val32);
+ }
+}
+
+static void amd8111_pci_bridge_exit(struct amd8111_pci_info *pci_info)
+{
+ u32 val32;
+ struct pci_dev *dev = pci_info->dev;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL) {
+ /* Disable System Error reporting */
+ edac_pci_read_dword(dev, REG_PCI_STSCMD, &val32);
+ val32 &= ~PCI_STSCMD_SERREN;
+ edac_pci_write_dword(dev, REG_PCI_STSCMD, val32);
+
+ /* Disable CRC flood packets */
+ edac_pci_read_dword(dev, REG_HT_LINK, &val32);
+ val32 &= ~HT_LINK_CRCFEN;
+ edac_pci_write_dword(dev, REG_HT_LINK, val32);
+
+ /* Disable DTSERREN/MARSP/SERREN in Interrupt Control reg */
+ edac_pci_read_dword(dev, REG_PCI_INTBRG_CTRL, &val32);
+ val32 &= ~PCI_INTBRG_CTRL_POLL_MASK;
+ edac_pci_write_dword(dev, REG_PCI_INTBRG_CTRL, val32);
+ }
+}
+
+static void amd8111_pci_bridge_check(struct edac_pci_ctl_info *edac_dev)
+{
+ struct amd8111_pci_info *pci_info = edac_dev->pvt_info;
+ struct pci_dev *dev = pci_info->dev;
+ u32 val32;
+
+ /* Check out PCI Bridge Status and Command Register */
+ edac_pci_read_dword(dev, REG_PCI_STSCMD, &val32);
+ if (val32 & PCI_STSCMD_CLEAR_MASK) {
+ printk(KERN_INFO "Error(s) in PCI bridge status and command"
+ "register on device %s\n", pci_info->ctl_name);
+ printk(KERN_INFO "SSE: %d, RMA: %d, RTA: %d\n",
+ (val32 & PCI_STSCMD_SSE) != 0,
+ (val32 & PCI_STSCMD_RMA) != 0,
+ (val32 & PCI_STSCMD_RTA) != 0);
+
+ val32 |= PCI_STSCMD_CLEAR_MASK;
+ edac_pci_write_dword(dev, REG_PCI_STSCMD, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check out HyperTransport Link Control Register */
+ edac_pci_read_dword(dev, REG_HT_LINK, &val32);
+ if (val32 & HT_LINK_LKFAIL) {
+ printk(KERN_INFO "Error(s) in hypertransport link control"
+ "register on device %s\n", pci_info->ctl_name);
+ printk(KERN_INFO "LKFAIL: %d\n",
+ (val32 & HT_LINK_LKFAIL) != 0);
+
+ val32 |= HT_LINK_LKFAIL;
+ edac_pci_write_dword(dev, REG_HT_LINK, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check out PCI Interrupt and Bridge Control Register */
+ edac_pci_read_dword(dev, REG_PCI_INTBRG_CTRL, &val32);
+ if (val32 & PCI_INTBRG_CTRL_DTSTAT) {
+ printk(KERN_INFO "Error(s) in PCI interrupt and bridge control"
+ "register on device %s\n", pci_info->ctl_name);
+ printk(KERN_INFO "DTSTAT: %d\n",
+ (val32 & PCI_INTBRG_CTRL_DTSTAT) != 0);
+
+ val32 |= PCI_INTBRG_CTRL_DTSTAT;
+ edac_pci_write_dword(dev, REG_PCI_INTBRG_CTRL, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check out PCI Bridge Memory Base-Limit Register */
+ edac_pci_read_dword(dev, REG_MEM_LIM, &val32);
+ if (val32 & MEM_LIMIT_CLEAR_MASK) {
+ printk(KERN_INFO
+ "Error(s) in mem limit register on %s device\n",
+ pci_info->ctl_name);
+ printk(KERN_INFO "DPE: %d, RSE: %d, RMA: %d\n"
+ "RTA: %d, STA: %d, MDPE: %d\n",
+ (val32 & MEM_LIMIT_DPE) != 0,
+ (val32 & MEM_LIMIT_RSE) != 0,
+ (val32 & MEM_LIMIT_RMA) != 0,
+ (val32 & MEM_LIMIT_RTA) != 0,
+ (val32 & MEM_LIMIT_STA) != 0,
+ (val32 & MEM_LIMIT_MDPE) != 0);
+
+ val32 |= MEM_LIMIT_CLEAR_MASK;
+ edac_pci_write_dword(dev, REG_MEM_LIM, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+}
+
+static struct resource *legacy_io_res;
+static int at_compat_reg_broken;
+#define LEGACY_NR_PORTS 1
+
+/* device-specific methods for amd8111 LPC Bridge device */
+static void amd8111_lpc_bridge_init(struct amd8111_dev_info *dev_info)
+{
+ u8 val8;
+ struct pci_dev *dev = dev_info->dev;
+
+ /* First clear REG_AT_COMPAT[SERR, IOCHK] if necessary */
+ legacy_io_res = request_region(REG_AT_COMPAT, LEGACY_NR_PORTS,
+ AMD8111_EDAC_MOD_STR);
+ if (!legacy_io_res)
+ printk(KERN_INFO "%s: failed to request legacy I/O region "
+ "start %d, len %d\n", __func__,
+ REG_AT_COMPAT, LEGACY_NR_PORTS);
+ else {
+ val8 = __do_inb(REG_AT_COMPAT);
+ if (val8 == 0xff) { /* buggy port */
+ printk(KERN_INFO "%s: port %d is buggy, not supported"
+ " by hardware?\n", __func__, REG_AT_COMPAT);
+ at_compat_reg_broken = 1;
+ release_region(REG_AT_COMPAT, LEGACY_NR_PORTS);
+ legacy_io_res = NULL;
+ } else {
+ u8 out8 = 0;
+ if (val8 & AT_COMPAT_SERR)
+ out8 = AT_COMPAT_CLRSERR;
+ if (val8 & AT_COMPAT_IOCHK)
+ out8 |= AT_COMPAT_CLRIOCHK;
+ if (out8 > 0)
+ __do_outb(out8, REG_AT_COMPAT);
+ }
+ }
+
+ /* Second clear error flags on LPC bridge */
+ edac_pci_read_byte(dev, REG_IO_CTRL_1, &val8);
+ if (val8 & IO_CTRL_1_CLEAR_MASK)
+ edac_pci_write_byte(dev, REG_IO_CTRL_1, val8);
+}
+
+static void amd8111_lpc_bridge_exit(struct amd8111_dev_info *dev_info)
+{
+ if (legacy_io_res)
+ release_region(REG_AT_COMPAT, LEGACY_NR_PORTS);
+}
+
+static void amd8111_lpc_bridge_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct amd8111_dev_info *dev_info = edac_dev->pvt_info;
+ struct pci_dev *dev = dev_info->dev;
+ u8 val8;
+
+ edac_pci_read_byte(dev, REG_IO_CTRL_1, &val8);
+ if (val8 & IO_CTRL_1_CLEAR_MASK) {
+ printk(KERN_INFO
+ "Error(s) in IO control register on %s device\n",
+ dev_info->ctl_name);
+ printk(KERN_INFO "LPC ERR: %d, PW2LPC: %d\n",
+ (val8 & IO_CTRL_1_LPC_ERR) != 0,
+ (val8 & IO_CTRL_1_PW2LPC) != 0);
+
+ val8 |= IO_CTRL_1_CLEAR_MASK;
+ edac_pci_write_byte(dev, REG_IO_CTRL_1, val8);
+
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+ }
+
+ if (at_compat_reg_broken == 0) {
+ u8 out8 = 0;
+ val8 = __do_inb(REG_AT_COMPAT);
+ if (val8 & AT_COMPAT_SERR)
+ out8 = AT_COMPAT_CLRSERR;
+ if (val8 & AT_COMPAT_IOCHK)
+ out8 |= AT_COMPAT_CLRIOCHK;
+ if (out8 > 0) {
+ __do_outb(out8, REG_AT_COMPAT);
+ edac_device_handle_ue(edac_dev, 0, 0,
+ edac_dev->ctl_name);
+ }
+ }
+}
+
+/* General devices represented by edac_device_ctl_info */
+static struct amd8111_dev_info amd8111_devices[] = {
+ [LPC_BRIDGE] = {
+ .err_dev = PCI_DEVICE_ID_AMD_8111_LPC,
+ .ctl_name = "lpc",
+ .init = amd8111_lpc_bridge_init,
+ .exit = amd8111_lpc_bridge_exit,
+ .check = amd8111_lpc_bridge_check,
+ },
+ {0},
+};
+
+/* PCI controllers represented by edac_pci_ctl_info */
+static struct amd8111_pci_info amd8111_pcis[] = {
+ [PCI_BRIDGE] = {
+ .err_dev = PCI_DEVICE_ID_AMD_8111_PCI,
+ .ctl_name = "AMD8111_PCI_Controller",
+ .init = amd8111_pci_bridge_init,
+ .exit = amd8111_pci_bridge_exit,
+ .check = amd8111_pci_bridge_check,
+ },
+ {0},
+};
+
+static int amd8111_dev_probe(struct pci_dev *dev,
+ const struct pci_device_id *id)
+{
+ struct amd8111_dev_info *dev_info = &amd8111_devices[id->driver_data];
+ int ret = -ENODEV;
+
+ dev_info->dev = pci_get_device(PCI_VENDOR_ID_AMD,
+ dev_info->err_dev, NULL);
+
+ if (!dev_info->dev) {
+ printk(KERN_ERR "EDAC device not found:"
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, dev_info->err_dev,
+ dev_info->ctl_name);
+ goto err;
+ }
+
+ if (pci_enable_device(dev_info->dev)) {
+ printk(KERN_ERR "failed to enable:"
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, dev_info->err_dev,
+ dev_info->ctl_name);
+ goto err_dev_put;
+ }
+
+ /*
+ * we do not allocate extra private structure for
+ * edac_device_ctl_info, but make use of existing
+ * one instead.
+ */
+ dev_info->edac_idx = edac_device_alloc_index();
+ dev_info->edac_dev =
+ edac_device_alloc_ctl_info(0, dev_info->ctl_name, 1,
+ NULL, 0, 0,
+ NULL, 0, dev_info->edac_idx);
+ if (!dev_info->edac_dev) {
+ ret = -ENOMEM;
+ goto err_dev_put;
+ }
+
+ dev_info->edac_dev->pvt_info = dev_info;
+ dev_info->edac_dev->dev = &dev_info->dev->dev;
+ dev_info->edac_dev->mod_name = AMD8111_EDAC_MOD_STR;
+ dev_info->edac_dev->ctl_name = dev_info->ctl_name;
+ dev_info->edac_dev->dev_name = dev_name(&dev_info->dev->dev);
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ dev_info->edac_dev->edac_check = dev_info->check;
+
+ if (dev_info->init)
+ dev_info->init(dev_info);
+
+ if (edac_device_add_device(dev_info->edac_dev) > 0) {
+ printk(KERN_ERR "failed to add edac_dev for %s\n",
+ dev_info->ctl_name);
+ goto err_edac_free_ctl;
+ }
+
+ printk(KERN_INFO "added one edac_dev on AMD8111 "
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, dev_info->err_dev,
+ dev_info->ctl_name);
+
+ return 0;
+
+err_edac_free_ctl:
+ edac_device_free_ctl_info(dev_info->edac_dev);
+err_dev_put:
+ pci_dev_put(dev_info->dev);
+err:
+ return ret;
+}
+
+static void amd8111_dev_remove(struct pci_dev *dev)
+{
+ struct amd8111_dev_info *dev_info;
+
+ for (dev_info = amd8111_devices; dev_info->err_dev; dev_info++)
+ if (dev_info->dev->device == dev->device)
+ break;
+
+ if (!dev_info->err_dev) /* should never happen */
+ return;
+
+ if (dev_info->edac_dev) {
+ edac_device_del_device(dev_info->edac_dev->dev);
+ edac_device_free_ctl_info(dev_info->edac_dev);
+ }
+
+ if (dev_info->exit)
+ dev_info->exit(dev_info);
+
+ pci_dev_put(dev_info->dev);
+}
+
+static int amd8111_pci_probe(struct pci_dev *dev,
+ const struct pci_device_id *id)
+{
+ struct amd8111_pci_info *pci_info = &amd8111_pcis[id->driver_data];
+ int ret = -ENODEV;
+
+ pci_info->dev = pci_get_device(PCI_VENDOR_ID_AMD,
+ pci_info->err_dev, NULL);
+
+ if (!pci_info->dev) {
+ printk(KERN_ERR "EDAC device not found:"
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, pci_info->err_dev,
+ pci_info->ctl_name);
+ goto err;
+ }
+
+ if (pci_enable_device(pci_info->dev)) {
+ printk(KERN_ERR "failed to enable:"
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, pci_info->err_dev,
+ pci_info->ctl_name);
+ goto err_dev_put;
+ }
+
+ /*
+ * we do not allocate extra private structure for
+ * edac_pci_ctl_info, but make use of existing
+ * one instead.
+ */
+ pci_info->edac_idx = edac_pci_alloc_index();
+ pci_info->edac_dev = edac_pci_alloc_ctl_info(0, pci_info->ctl_name);
+ if (!pci_info->edac_dev) {
+ ret = -ENOMEM;
+ goto err_dev_put;
+ }
+
+ pci_info->edac_dev->pvt_info = pci_info;
+ pci_info->edac_dev->dev = &pci_info->dev->dev;
+ pci_info->edac_dev->mod_name = AMD8111_EDAC_MOD_STR;
+ pci_info->edac_dev->ctl_name = pci_info->ctl_name;
+ pci_info->edac_dev->dev_name = dev_name(&pci_info->dev->dev);
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ pci_info->edac_dev->edac_check = pci_info->check;
+
+ if (pci_info->init)
+ pci_info->init(pci_info);
+
+ if (edac_pci_add_device(pci_info->edac_dev, pci_info->edac_idx) > 0) {
+ printk(KERN_ERR "failed to add edac_pci for %s\n",
+ pci_info->ctl_name);
+ goto err_edac_free_ctl;
+ }
+
+ printk(KERN_INFO "added one edac_pci on AMD8111 "
+ "vendor %x, device %x, name %s\n",
+ PCI_VENDOR_ID_AMD, pci_info->err_dev,
+ pci_info->ctl_name);
+
+ return 0;
+
+err_edac_free_ctl:
+ edac_pci_free_ctl_info(pci_info->edac_dev);
+err_dev_put:
+ pci_dev_put(pci_info->dev);
+err:
+ return ret;
+}
+
+static void amd8111_pci_remove(struct pci_dev *dev)
+{
+ struct amd8111_pci_info *pci_info;
+
+ for (pci_info = amd8111_pcis; pci_info->err_dev; pci_info++)
+ if (pci_info->dev->device == dev->device)
+ break;
+
+ if (!pci_info->err_dev) /* should never happen */
+ return;
+
+ if (pci_info->edac_dev) {
+ edac_pci_del_device(pci_info->edac_dev->dev);
+ edac_pci_free_ctl_info(pci_info->edac_dev);
+ }
+
+ if (pci_info->exit)
+ pci_info->exit(pci_info);
+
+ pci_dev_put(pci_info->dev);
+}
+
+/* PCI Device ID talbe for general EDAC device */
+static const struct pci_device_id amd8111_edac_dev_tbl[] = {
+ {
+ PCI_VEND_DEV(AMD, 8111_LPC),
+ .subvendor = PCI_ANY_ID,
+ .subdevice = PCI_ANY_ID,
+ .class = 0,
+ .class_mask = 0,
+ .driver_data = LPC_BRIDGE,
+ },
+ {
+ 0,
+ } /* table is NULL-terminated */
+};
+MODULE_DEVICE_TABLE(pci, amd8111_edac_dev_tbl);
+
+static struct pci_driver amd8111_edac_dev_driver = {
+ .name = "AMD8111_EDAC_DEV",
+ .probe = amd8111_dev_probe,
+ .remove = amd8111_dev_remove,
+ .id_table = amd8111_edac_dev_tbl,
+};
+
+/* PCI Device ID table for EDAC PCI controller */
+static const struct pci_device_id amd8111_edac_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(AMD, 8111_PCI),
+ .subvendor = PCI_ANY_ID,
+ .subdevice = PCI_ANY_ID,
+ .class = 0,
+ .class_mask = 0,
+ .driver_data = PCI_BRIDGE,
+ },
+ {
+ 0,
+ } /* table is NULL-terminated */
+};
+MODULE_DEVICE_TABLE(pci, amd8111_edac_pci_tbl);
+
+static struct pci_driver amd8111_edac_pci_driver = {
+ .name = "AMD8111_EDAC_PCI",
+ .probe = amd8111_pci_probe,
+ .remove = amd8111_pci_remove,
+ .id_table = amd8111_edac_pci_tbl,
+};
+
+static int __init amd8111_edac_init(void)
+{
+ int val;
+
+ printk(KERN_INFO "AMD8111 EDAC driver " AMD8111_EDAC_REVISION "\n");
+ printk(KERN_INFO "\t(c) 2008 Wind River Systems, Inc.\n");
+
+ /* Only POLL mode supported so far */
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ val = pci_register_driver(&amd8111_edac_dev_driver);
+ val |= pci_register_driver(&amd8111_edac_pci_driver);
+
+ return val;
+}
+
+static void __exit amd8111_edac_exit(void)
+{
+ pci_unregister_driver(&amd8111_edac_pci_driver);
+ pci_unregister_driver(&amd8111_edac_dev_driver);
+}
+
+
+module_init(amd8111_edac_init);
+module_exit(amd8111_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cao Qingtao <qingtao.cao@windriver.com>");
+MODULE_DESCRIPTION("AMD8111 HyperTransport I/O Hub EDAC kernel module");
diff --git a/drivers/edac/amd8111_edac.h b/drivers/edac/amd8111_edac.h
new file mode 100644
index 0000000000..200cab1b3e
--- /dev/null
+++ b/drivers/edac/amd8111_edac.h
@@ -0,0 +1,118 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * amd8111_edac.h, EDAC defs for AMD8111 hypertransport chip
+ *
+ * Copyright (c) 2008 Wind River Systems, Inc.
+ *
+ * Authors: Cao Qingtao <qingtao.cao@windriver.com>
+ * Benjamin Walsh <benjamin.walsh@windriver.com>
+ * Hu Yongqi <yongqi.hu@windriver.com>
+ */
+
+#ifndef _AMD8111_EDAC_H_
+#define _AMD8111_EDAC_H_
+
+/************************************************************
+ * PCI Bridge Status and Command Register, DevA:0x04
+ ************************************************************/
+#define REG_PCI_STSCMD 0x04
+enum pci_stscmd_bits {
+ PCI_STSCMD_SSE = BIT(30),
+ PCI_STSCMD_RMA = BIT(29),
+ PCI_STSCMD_RTA = BIT(28),
+ PCI_STSCMD_SERREN = BIT(8),
+ PCI_STSCMD_CLEAR_MASK = (PCI_STSCMD_SSE |
+ PCI_STSCMD_RMA |
+ PCI_STSCMD_RTA)
+};
+
+/************************************************************
+ * PCI Bridge Memory Base-Limit Register, DevA:0x1c
+ ************************************************************/
+#define REG_MEM_LIM 0x1c
+enum mem_limit_bits {
+ MEM_LIMIT_DPE = BIT(31),
+ MEM_LIMIT_RSE = BIT(30),
+ MEM_LIMIT_RMA = BIT(29),
+ MEM_LIMIT_RTA = BIT(28),
+ MEM_LIMIT_STA = BIT(27),
+ MEM_LIMIT_MDPE = BIT(24),
+ MEM_LIMIT_CLEAR_MASK = (MEM_LIMIT_DPE |
+ MEM_LIMIT_RSE |
+ MEM_LIMIT_RMA |
+ MEM_LIMIT_RTA |
+ MEM_LIMIT_STA |
+ MEM_LIMIT_MDPE)
+};
+
+/************************************************************
+ * HyperTransport Link Control Register, DevA:0xc4
+ ************************************************************/
+#define REG_HT_LINK 0xc4
+enum ht_link_bits {
+ HT_LINK_LKFAIL = BIT(4),
+ HT_LINK_CRCFEN = BIT(1),
+ HT_LINK_CLEAR_MASK = (HT_LINK_LKFAIL)
+};
+
+/************************************************************
+ * PCI Bridge Interrupt and Bridge Control, DevA:0x3c
+ ************************************************************/
+#define REG_PCI_INTBRG_CTRL 0x3c
+enum pci_intbrg_ctrl_bits {
+ PCI_INTBRG_CTRL_DTSERREN = BIT(27),
+ PCI_INTBRG_CTRL_DTSTAT = BIT(26),
+ PCI_INTBRG_CTRL_MARSP = BIT(21),
+ PCI_INTBRG_CTRL_SERREN = BIT(17),
+ PCI_INTBRG_CTRL_PEREN = BIT(16),
+ PCI_INTBRG_CTRL_CLEAR_MASK = (PCI_INTBRG_CTRL_DTSTAT),
+ PCI_INTBRG_CTRL_POLL_MASK = (PCI_INTBRG_CTRL_DTSERREN |
+ PCI_INTBRG_CTRL_MARSP |
+ PCI_INTBRG_CTRL_SERREN)
+};
+
+/************************************************************
+ * I/O Control 1 Register, DevB:0x40
+ ************************************************************/
+#define REG_IO_CTRL_1 0x40
+enum io_ctrl_1_bits {
+ IO_CTRL_1_NMIONERR = BIT(7),
+ IO_CTRL_1_LPC_ERR = BIT(6),
+ IO_CTRL_1_PW2LPC = BIT(1),
+ IO_CTRL_1_CLEAR_MASK = (IO_CTRL_1_LPC_ERR | IO_CTRL_1_PW2LPC)
+};
+
+/************************************************************
+ * Legacy I/O Space Registers
+ ************************************************************/
+#define REG_AT_COMPAT 0x61
+enum at_compat_bits {
+ AT_COMPAT_SERR = BIT(7),
+ AT_COMPAT_IOCHK = BIT(6),
+ AT_COMPAT_CLRIOCHK = BIT(3),
+ AT_COMPAT_CLRSERR = BIT(2),
+};
+
+struct amd8111_dev_info {
+ u16 err_dev; /* PCI Device ID */
+ struct pci_dev *dev;
+ int edac_idx; /* device index */
+ char *ctl_name;
+ struct edac_device_ctl_info *edac_dev;
+ void (*init)(struct amd8111_dev_info *dev_info);
+ void (*exit)(struct amd8111_dev_info *dev_info);
+ void (*check)(struct edac_device_ctl_info *edac_dev);
+};
+
+struct amd8111_pci_info {
+ u16 err_dev; /* PCI Device ID */
+ struct pci_dev *dev;
+ int edac_idx; /* pci index */
+ const char *ctl_name;
+ struct edac_pci_ctl_info *edac_dev;
+ void (*init)(struct amd8111_pci_info *dev_info);
+ void (*exit)(struct amd8111_pci_info *dev_info);
+ void (*check)(struct edac_pci_ctl_info *edac_dev);
+};
+
+#endif /* _AMD8111_EDAC_H_ */
diff --git a/drivers/edac/amd8131_edac.c b/drivers/edac/amd8131_edac.c
new file mode 100644
index 0000000000..28610ba514
--- /dev/null
+++ b/drivers/edac/amd8131_edac.c
@@ -0,0 +1,358 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * amd8131_edac.c, AMD8131 hypertransport chip EDAC kernel module
+ *
+ * Copyright (c) 2008 Wind River Systems, Inc.
+ *
+ * Authors: Cao Qingtao <qingtao.cao@windriver.com>
+ * Benjamin Walsh <benjamin.walsh@windriver.com>
+ * Hu Yongqi <yongqi.hu@windriver.com>
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/bitops.h>
+#include <linux/edac.h>
+#include <linux/pci_ids.h>
+
+#include "edac_module.h"
+#include "amd8131_edac.h"
+
+#define AMD8131_EDAC_REVISION " Ver: 1.0.0"
+#define AMD8131_EDAC_MOD_STR "amd8131_edac"
+
+/* Wrapper functions for accessing PCI configuration space */
+static void edac_pci_read_dword(struct pci_dev *dev, int reg, u32 *val32)
+{
+ int ret;
+
+ ret = pci_read_config_dword(dev, reg, val32);
+ if (ret != 0)
+ printk(KERN_ERR AMD8131_EDAC_MOD_STR
+ " PCI Access Read Error at 0x%x\n", reg);
+}
+
+static void edac_pci_write_dword(struct pci_dev *dev, int reg, u32 val32)
+{
+ int ret;
+
+ ret = pci_write_config_dword(dev, reg, val32);
+ if (ret != 0)
+ printk(KERN_ERR AMD8131_EDAC_MOD_STR
+ " PCI Access Write Error at 0x%x\n", reg);
+}
+
+/* Support up to two AMD8131 chipsets on a platform */
+static struct amd8131_dev_info amd8131_devices[] = {
+ {
+ .inst = NORTH_A,
+ .devfn = DEVFN_PCIX_BRIDGE_NORTH_A,
+ .ctl_name = "AMD8131_PCIX_NORTH_A",
+ },
+ {
+ .inst = NORTH_B,
+ .devfn = DEVFN_PCIX_BRIDGE_NORTH_B,
+ .ctl_name = "AMD8131_PCIX_NORTH_B",
+ },
+ {
+ .inst = SOUTH_A,
+ .devfn = DEVFN_PCIX_BRIDGE_SOUTH_A,
+ .ctl_name = "AMD8131_PCIX_SOUTH_A",
+ },
+ {
+ .inst = SOUTH_B,
+ .devfn = DEVFN_PCIX_BRIDGE_SOUTH_B,
+ .ctl_name = "AMD8131_PCIX_SOUTH_B",
+ },
+ {.inst = NO_BRIDGE,},
+};
+
+static void amd8131_pcix_init(struct amd8131_dev_info *dev_info)
+{
+ u32 val32;
+ struct pci_dev *dev = dev_info->dev;
+
+ /* First clear error detection flags */
+ edac_pci_read_dword(dev, REG_MEM_LIM, &val32);
+ if (val32 & MEM_LIMIT_MASK)
+ edac_pci_write_dword(dev, REG_MEM_LIM, val32);
+
+ /* Clear Discard Timer Timedout flag */
+ edac_pci_read_dword(dev, REG_INT_CTLR, &val32);
+ if (val32 & INT_CTLR_DTS)
+ edac_pci_write_dword(dev, REG_INT_CTLR, val32);
+
+ /* Clear CRC Error flag on link side A */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_A, &val32);
+ if (val32 & LNK_CTRL_CRCERR_A)
+ edac_pci_write_dword(dev, REG_LNK_CTRL_A, val32);
+
+ /* Clear CRC Error flag on link side B */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_B, &val32);
+ if (val32 & LNK_CTRL_CRCERR_B)
+ edac_pci_write_dword(dev, REG_LNK_CTRL_B, val32);
+
+ /*
+ * Then enable all error detections.
+ *
+ * Setup Discard Timer Sync Flood Enable,
+ * System Error Enable and Parity Error Enable.
+ */
+ edac_pci_read_dword(dev, REG_INT_CTLR, &val32);
+ val32 |= INT_CTLR_PERR | INT_CTLR_SERR | INT_CTLR_DTSE;
+ edac_pci_write_dword(dev, REG_INT_CTLR, val32);
+
+ /* Enable overall SERR Error detection */
+ edac_pci_read_dword(dev, REG_STS_CMD, &val32);
+ val32 |= STS_CMD_SERREN;
+ edac_pci_write_dword(dev, REG_STS_CMD, val32);
+
+ /* Setup CRC Flood Enable for link side A */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_A, &val32);
+ val32 |= LNK_CTRL_CRCFEN;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_A, val32);
+
+ /* Setup CRC Flood Enable for link side B */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_B, &val32);
+ val32 |= LNK_CTRL_CRCFEN;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_B, val32);
+}
+
+static void amd8131_pcix_exit(struct amd8131_dev_info *dev_info)
+{
+ u32 val32;
+ struct pci_dev *dev = dev_info->dev;
+
+ /* Disable SERR, PERR and DTSE Error detection */
+ edac_pci_read_dword(dev, REG_INT_CTLR, &val32);
+ val32 &= ~(INT_CTLR_PERR | INT_CTLR_SERR | INT_CTLR_DTSE);
+ edac_pci_write_dword(dev, REG_INT_CTLR, val32);
+
+ /* Disable overall System Error detection */
+ edac_pci_read_dword(dev, REG_STS_CMD, &val32);
+ val32 &= ~STS_CMD_SERREN;
+ edac_pci_write_dword(dev, REG_STS_CMD, val32);
+
+ /* Disable CRC Sync Flood on link side A */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_A, &val32);
+ val32 &= ~LNK_CTRL_CRCFEN;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_A, val32);
+
+ /* Disable CRC Sync Flood on link side B */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_B, &val32);
+ val32 &= ~LNK_CTRL_CRCFEN;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_B, val32);
+}
+
+static void amd8131_pcix_check(struct edac_pci_ctl_info *edac_dev)
+{
+ struct amd8131_dev_info *dev_info = edac_dev->pvt_info;
+ struct pci_dev *dev = dev_info->dev;
+ u32 val32;
+
+ /* Check PCI-X Bridge Memory Base-Limit Register for errors */
+ edac_pci_read_dword(dev, REG_MEM_LIM, &val32);
+ if (val32 & MEM_LIMIT_MASK) {
+ printk(KERN_INFO "Error(s) in mem limit register "
+ "on %s bridge\n", dev_info->ctl_name);
+ printk(KERN_INFO "DPE: %d, RSE: %d, RMA: %d\n"
+ "RTA: %d, STA: %d, MDPE: %d\n",
+ val32 & MEM_LIMIT_DPE,
+ val32 & MEM_LIMIT_RSE,
+ val32 & MEM_LIMIT_RMA,
+ val32 & MEM_LIMIT_RTA,
+ val32 & MEM_LIMIT_STA,
+ val32 & MEM_LIMIT_MDPE);
+
+ val32 |= MEM_LIMIT_MASK;
+ edac_pci_write_dword(dev, REG_MEM_LIM, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check if Discard Timer timed out */
+ edac_pci_read_dword(dev, REG_INT_CTLR, &val32);
+ if (val32 & INT_CTLR_DTS) {
+ printk(KERN_INFO "Error(s) in interrupt and control register "
+ "on %s bridge\n", dev_info->ctl_name);
+ printk(KERN_INFO "DTS: %d\n", val32 & INT_CTLR_DTS);
+
+ val32 |= INT_CTLR_DTS;
+ edac_pci_write_dword(dev, REG_INT_CTLR, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check if CRC error happens on link side A */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_A, &val32);
+ if (val32 & LNK_CTRL_CRCERR_A) {
+ printk(KERN_INFO "Error(s) in link conf and control register "
+ "on %s bridge\n", dev_info->ctl_name);
+ printk(KERN_INFO "CRCERR: %d\n", val32 & LNK_CTRL_CRCERR_A);
+
+ val32 |= LNK_CTRL_CRCERR_A;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_A, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+
+ /* Check if CRC error happens on link side B */
+ edac_pci_read_dword(dev, REG_LNK_CTRL_B, &val32);
+ if (val32 & LNK_CTRL_CRCERR_B) {
+ printk(KERN_INFO "Error(s) in link conf and control register "
+ "on %s bridge\n", dev_info->ctl_name);
+ printk(KERN_INFO "CRCERR: %d\n", val32 & LNK_CTRL_CRCERR_B);
+
+ val32 |= LNK_CTRL_CRCERR_B;
+ edac_pci_write_dword(dev, REG_LNK_CTRL_B, val32);
+
+ edac_pci_handle_npe(edac_dev, edac_dev->ctl_name);
+ }
+}
+
+static struct amd8131_info amd8131_chipset = {
+ .err_dev = PCI_DEVICE_ID_AMD_8131_APIC,
+ .devices = amd8131_devices,
+ .init = amd8131_pcix_init,
+ .exit = amd8131_pcix_exit,
+ .check = amd8131_pcix_check,
+};
+
+/*
+ * There are 4 PCIX Bridges on ATCA-6101 that share the same PCI Device ID,
+ * so amd8131_probe() would be called by kernel 4 times, with different
+ * address of pci_dev for each of them each time.
+ */
+static int amd8131_probe(struct pci_dev *dev, const struct pci_device_id *id)
+{
+ struct amd8131_dev_info *dev_info;
+
+ for (dev_info = amd8131_chipset.devices; dev_info->inst != NO_BRIDGE;
+ dev_info++)
+ if (dev_info->devfn == dev->devfn)
+ break;
+
+ if (dev_info->inst == NO_BRIDGE) /* should never happen */
+ return -ENODEV;
+
+ /*
+ * We can't call pci_get_device() as we are used to do because
+ * there are 4 of them but pci_dev_get() instead.
+ */
+ dev_info->dev = pci_dev_get(dev);
+
+ if (pci_enable_device(dev_info->dev)) {
+ pci_dev_put(dev_info->dev);
+ printk(KERN_ERR "failed to enable:"
+ "vendor %x, device %x, devfn %x, name %s\n",
+ PCI_VENDOR_ID_AMD, amd8131_chipset.err_dev,
+ dev_info->devfn, dev_info->ctl_name);
+ return -ENODEV;
+ }
+
+ /*
+ * we do not allocate extra private structure for
+ * edac_pci_ctl_info, but make use of existing
+ * one instead.
+ */
+ dev_info->edac_idx = edac_pci_alloc_index();
+ dev_info->edac_dev = edac_pci_alloc_ctl_info(0, dev_info->ctl_name);
+ if (!dev_info->edac_dev)
+ return -ENOMEM;
+
+ dev_info->edac_dev->pvt_info = dev_info;
+ dev_info->edac_dev->dev = &dev_info->dev->dev;
+ dev_info->edac_dev->mod_name = AMD8131_EDAC_MOD_STR;
+ dev_info->edac_dev->ctl_name = dev_info->ctl_name;
+ dev_info->edac_dev->dev_name = dev_name(&dev_info->dev->dev);
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ dev_info->edac_dev->edac_check = amd8131_chipset.check;
+
+ if (amd8131_chipset.init)
+ amd8131_chipset.init(dev_info);
+
+ if (edac_pci_add_device(dev_info->edac_dev, dev_info->edac_idx) > 0) {
+ printk(KERN_ERR "failed edac_pci_add_device() for %s\n",
+ dev_info->ctl_name);
+ edac_pci_free_ctl_info(dev_info->edac_dev);
+ return -ENODEV;
+ }
+
+ printk(KERN_INFO "added one device on AMD8131 "
+ "vendor %x, device %x, devfn %x, name %s\n",
+ PCI_VENDOR_ID_AMD, amd8131_chipset.err_dev,
+ dev_info->devfn, dev_info->ctl_name);
+
+ return 0;
+}
+
+static void amd8131_remove(struct pci_dev *dev)
+{
+ struct amd8131_dev_info *dev_info;
+
+ for (dev_info = amd8131_chipset.devices; dev_info->inst != NO_BRIDGE;
+ dev_info++)
+ if (dev_info->devfn == dev->devfn)
+ break;
+
+ if (dev_info->inst == NO_BRIDGE) /* should never happen */
+ return;
+
+ if (dev_info->edac_dev) {
+ edac_pci_del_device(dev_info->edac_dev->dev);
+ edac_pci_free_ctl_info(dev_info->edac_dev);
+ }
+
+ if (amd8131_chipset.exit)
+ amd8131_chipset.exit(dev_info);
+
+ pci_dev_put(dev_info->dev);
+}
+
+static const struct pci_device_id amd8131_edac_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(AMD, 8131_BRIDGE),
+ .subvendor = PCI_ANY_ID,
+ .subdevice = PCI_ANY_ID,
+ .class = 0,
+ .class_mask = 0,
+ .driver_data = 0,
+ },
+ {
+ 0,
+ } /* table is NULL-terminated */
+};
+MODULE_DEVICE_TABLE(pci, amd8131_edac_pci_tbl);
+
+static struct pci_driver amd8131_edac_driver = {
+ .name = AMD8131_EDAC_MOD_STR,
+ .probe = amd8131_probe,
+ .remove = amd8131_remove,
+ .id_table = amd8131_edac_pci_tbl,
+};
+
+static int __init amd8131_edac_init(void)
+{
+ printk(KERN_INFO "AMD8131 EDAC driver " AMD8131_EDAC_REVISION "\n");
+ printk(KERN_INFO "\t(c) 2008 Wind River Systems, Inc.\n");
+
+ /* Only POLL mode supported so far */
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ return pci_register_driver(&amd8131_edac_driver);
+}
+
+static void __exit amd8131_edac_exit(void)
+{
+ pci_unregister_driver(&amd8131_edac_driver);
+}
+
+module_init(amd8131_edac_init);
+module_exit(amd8131_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cao Qingtao <qingtao.cao@windriver.com>");
+MODULE_DESCRIPTION("AMD8131 HyperTransport PCI-X Tunnel EDAC kernel module");
diff --git a/drivers/edac/amd8131_edac.h b/drivers/edac/amd8131_edac.h
new file mode 100644
index 0000000000..5f362abdaf
--- /dev/null
+++ b/drivers/edac/amd8131_edac.h
@@ -0,0 +1,107 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * amd8131_edac.h, EDAC defs for AMD8131 hypertransport chip
+ *
+ * Copyright (c) 2008 Wind River Systems, Inc.
+ *
+ * Authors: Cao Qingtao <qingtao.cao@windriver.com>
+ * Benjamin Walsh <benjamin.walsh@windriver.com>
+ * Hu Yongqi <yongqi.hu@windriver.com>
+ */
+
+#ifndef _AMD8131_EDAC_H_
+#define _AMD8131_EDAC_H_
+
+#define DEVFN_PCIX_BRIDGE_NORTH_A 8
+#define DEVFN_PCIX_BRIDGE_NORTH_B 16
+#define DEVFN_PCIX_BRIDGE_SOUTH_A 24
+#define DEVFN_PCIX_BRIDGE_SOUTH_B 32
+
+/************************************************************
+ * PCI-X Bridge Status and Command Register, DevA:0x04
+ ************************************************************/
+#define REG_STS_CMD 0x04
+enum sts_cmd_bits {
+ STS_CMD_SSE = BIT(30),
+ STS_CMD_SERREN = BIT(8)
+};
+
+/************************************************************
+ * PCI-X Bridge Interrupt and Bridge Control Register,
+ ************************************************************/
+#define REG_INT_CTLR 0x3c
+enum int_ctlr_bits {
+ INT_CTLR_DTSE = BIT(27),
+ INT_CTLR_DTS = BIT(26),
+ INT_CTLR_SERR = BIT(17),
+ INT_CTLR_PERR = BIT(16)
+};
+
+/************************************************************
+ * PCI-X Bridge Memory Base-Limit Register, DevA:0x1C
+ ************************************************************/
+#define REG_MEM_LIM 0x1c
+enum mem_limit_bits {
+ MEM_LIMIT_DPE = BIT(31),
+ MEM_LIMIT_RSE = BIT(30),
+ MEM_LIMIT_RMA = BIT(29),
+ MEM_LIMIT_RTA = BIT(28),
+ MEM_LIMIT_STA = BIT(27),
+ MEM_LIMIT_MDPE = BIT(24),
+ MEM_LIMIT_MASK = MEM_LIMIT_DPE|MEM_LIMIT_RSE|MEM_LIMIT_RMA|
+ MEM_LIMIT_RTA|MEM_LIMIT_STA|MEM_LIMIT_MDPE
+};
+
+/************************************************************
+ * Link Configuration And Control Register, side A
+ ************************************************************/
+#define REG_LNK_CTRL_A 0xc4
+
+/************************************************************
+ * Link Configuration And Control Register, side B
+ ************************************************************/
+#define REG_LNK_CTRL_B 0xc8
+
+enum lnk_ctrl_bits {
+ LNK_CTRL_CRCERR_A = BIT(9),
+ LNK_CTRL_CRCERR_B = BIT(8),
+ LNK_CTRL_CRCFEN = BIT(1)
+};
+
+enum pcix_bridge_inst {
+ NORTH_A = 0,
+ NORTH_B = 1,
+ SOUTH_A = 2,
+ SOUTH_B = 3,
+ NO_BRIDGE = 4
+};
+
+struct amd8131_dev_info {
+ int devfn;
+ enum pcix_bridge_inst inst;
+ struct pci_dev *dev;
+ int edac_idx; /* pci device index */
+ char *ctl_name;
+ struct edac_pci_ctl_info *edac_dev;
+};
+
+/*
+ * AMD8131 chipset has two pairs of PCIX Bridge and related IOAPIC
+ * Controller, and ATCA-6101 has two AMD8131 chipsets, so there are
+ * four PCIX Bridges on ATCA-6101 altogether.
+ *
+ * These PCIX Bridges share the same PCI Device ID and are all of
+ * Function Zero, they could be discrimated by their pci_dev->devfn.
+ * They share the same set of init/check/exit methods, and their
+ * private structures are collected in the devices[] array.
+ */
+struct amd8131_info {
+ u16 err_dev; /* PCI Device ID for AMD8131 APIC*/
+ struct amd8131_dev_info *devices;
+ void (*init)(struct amd8131_dev_info *dev_info);
+ void (*exit)(struct amd8131_dev_info *dev_info);
+ void (*check)(struct edac_pci_ctl_info *edac_dev);
+};
+
+#endif /* _AMD8131_EDAC_H_ */
+
diff --git a/drivers/edac/armada_xp_edac.c b/drivers/edac/armada_xp_edac.c
new file mode 100644
index 0000000000..c4bd2fb9c4
--- /dev/null
+++ b/drivers/edac/armada_xp_edac.c
@@ -0,0 +1,624 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2017 Pengutronix, Jan Luebbe <kernel@pengutronix.de>
+ */
+
+#include <linux/kernel.h>
+#include <linux/edac.h>
+#include <linux/of_platform.h>
+
+#include <asm/hardware/cache-l2x0.h>
+#include <asm/hardware/cache-aurora-l2.h>
+
+#include "edac_mc.h"
+#include "edac_device.h"
+#include "edac_module.h"
+
+/************************ EDAC MC (DDR RAM) ********************************/
+
+#define SDRAM_NUM_CS 4
+
+#define SDRAM_CONFIG_REG 0x0
+#define SDRAM_CONFIG_ECC_MASK BIT(18)
+#define SDRAM_CONFIG_REGISTERED_MASK BIT(17)
+#define SDRAM_CONFIG_BUS_WIDTH_MASK BIT(15)
+
+#define SDRAM_ADDR_CTRL_REG 0x10
+#define SDRAM_ADDR_CTRL_SIZE_HIGH_OFFSET(cs) (20+cs)
+#define SDRAM_ADDR_CTRL_SIZE_HIGH_MASK(cs) (0x1 << SDRAM_ADDR_CTRL_SIZE_HIGH_OFFSET(cs))
+#define SDRAM_ADDR_CTRL_ADDR_SEL_MASK(cs) BIT(16+cs)
+#define SDRAM_ADDR_CTRL_SIZE_LOW_OFFSET(cs) (cs*4+2)
+#define SDRAM_ADDR_CTRL_SIZE_LOW_MASK(cs) (0x3 << SDRAM_ADDR_CTRL_SIZE_LOW_OFFSET(cs))
+#define SDRAM_ADDR_CTRL_STRUCT_OFFSET(cs) (cs*4)
+#define SDRAM_ADDR_CTRL_STRUCT_MASK(cs) (0x3 << SDRAM_ADDR_CTRL_STRUCT_OFFSET(cs))
+
+#define SDRAM_ERR_DATA_H_REG 0x40
+#define SDRAM_ERR_DATA_L_REG 0x44
+
+#define SDRAM_ERR_RECV_ECC_REG 0x48
+#define SDRAM_ERR_RECV_ECC_VALUE_MASK 0xff
+
+#define SDRAM_ERR_CALC_ECC_REG 0x4c
+#define SDRAM_ERR_CALC_ECC_ROW_OFFSET 8
+#define SDRAM_ERR_CALC_ECC_ROW_MASK (0xffff << SDRAM_ERR_CALC_ECC_ROW_OFFSET)
+#define SDRAM_ERR_CALC_ECC_VALUE_MASK 0xff
+
+#define SDRAM_ERR_ADDR_REG 0x50
+#define SDRAM_ERR_ADDR_BANK_OFFSET 23
+#define SDRAM_ERR_ADDR_BANK_MASK (0x7 << SDRAM_ERR_ADDR_BANK_OFFSET)
+#define SDRAM_ERR_ADDR_COL_OFFSET 8
+#define SDRAM_ERR_ADDR_COL_MASK (0x7fff << SDRAM_ERR_ADDR_COL_OFFSET)
+#define SDRAM_ERR_ADDR_CS_OFFSET 1
+#define SDRAM_ERR_ADDR_CS_MASK (0x3 << SDRAM_ERR_ADDR_CS_OFFSET)
+#define SDRAM_ERR_ADDR_TYPE_MASK BIT(0)
+
+#define SDRAM_ERR_CTRL_REG 0x54
+#define SDRAM_ERR_CTRL_THR_OFFSET 16
+#define SDRAM_ERR_CTRL_THR_MASK (0xff << SDRAM_ERR_CTRL_THR_OFFSET)
+#define SDRAM_ERR_CTRL_PROP_MASK BIT(9)
+
+#define SDRAM_ERR_SBE_COUNT_REG 0x58
+#define SDRAM_ERR_DBE_COUNT_REG 0x5c
+
+#define SDRAM_ERR_CAUSE_ERR_REG 0xd0
+#define SDRAM_ERR_CAUSE_MSG_REG 0xd8
+#define SDRAM_ERR_CAUSE_DBE_MASK BIT(1)
+#define SDRAM_ERR_CAUSE_SBE_MASK BIT(0)
+
+#define SDRAM_RANK_CTRL_REG 0x1e0
+#define SDRAM_RANK_CTRL_EXIST_MASK(cs) BIT(cs)
+
+struct axp_mc_drvdata {
+ void __iomem *base;
+ /* width in bytes */
+ unsigned int width;
+ /* bank interleaving */
+ bool cs_addr_sel[SDRAM_NUM_CS];
+
+ char msg[128];
+};
+
+/* derived from "DRAM Address Multiplexing" in the ARMADA XP Functional Spec */
+static uint32_t axp_mc_calc_address(struct axp_mc_drvdata *drvdata,
+ uint8_t cs, uint8_t bank, uint16_t row,
+ uint16_t col)
+{
+ if (drvdata->width == 8) {
+ /* 64 bit */
+ if (drvdata->cs_addr_sel[cs])
+ /* bank interleaved */
+ return (((row & 0xfff8) << 16) |
+ ((bank & 0x7) << 16) |
+ ((row & 0x7) << 13) |
+ ((col & 0x3ff) << 3));
+ else
+ return (((row & 0xffff << 16) |
+ ((bank & 0x7) << 13) |
+ ((col & 0x3ff)) << 3));
+ } else if (drvdata->width == 4) {
+ /* 32 bit */
+ if (drvdata->cs_addr_sel[cs])
+ /* bank interleaved */
+ return (((row & 0xfff0) << 15) |
+ ((bank & 0x7) << 16) |
+ ((row & 0xf) << 12) |
+ ((col & 0x3ff) << 2));
+ else
+ return (((row & 0xffff << 15) |
+ ((bank & 0x7) << 12) |
+ ((col & 0x3ff)) << 2));
+ } else {
+ /* 16 bit */
+ if (drvdata->cs_addr_sel[cs])
+ /* bank interleaved */
+ return (((row & 0xffe0) << 14) |
+ ((bank & 0x7) << 16) |
+ ((row & 0x1f) << 11) |
+ ((col & 0x3ff) << 1));
+ else
+ return (((row & 0xffff << 14) |
+ ((bank & 0x7) << 11) |
+ ((col & 0x3ff)) << 1));
+ }
+}
+
+static void axp_mc_check(struct mem_ctl_info *mci)
+{
+ struct axp_mc_drvdata *drvdata = mci->pvt_info;
+ uint32_t data_h, data_l, recv_ecc, calc_ecc, addr;
+ uint32_t cnt_sbe, cnt_dbe, cause_err, cause_msg;
+ uint32_t row_val, col_val, bank_val, addr_val;
+ uint8_t syndrome_val, cs_val;
+ char *msg = drvdata->msg;
+
+ data_h = readl(drvdata->base + SDRAM_ERR_DATA_H_REG);
+ data_l = readl(drvdata->base + SDRAM_ERR_DATA_L_REG);
+ recv_ecc = readl(drvdata->base + SDRAM_ERR_RECV_ECC_REG);
+ calc_ecc = readl(drvdata->base + SDRAM_ERR_CALC_ECC_REG);
+ addr = readl(drvdata->base + SDRAM_ERR_ADDR_REG);
+ cnt_sbe = readl(drvdata->base + SDRAM_ERR_SBE_COUNT_REG);
+ cnt_dbe = readl(drvdata->base + SDRAM_ERR_DBE_COUNT_REG);
+ cause_err = readl(drvdata->base + SDRAM_ERR_CAUSE_ERR_REG);
+ cause_msg = readl(drvdata->base + SDRAM_ERR_CAUSE_MSG_REG);
+
+ /* clear cause registers */
+ writel(~(SDRAM_ERR_CAUSE_DBE_MASK | SDRAM_ERR_CAUSE_SBE_MASK),
+ drvdata->base + SDRAM_ERR_CAUSE_ERR_REG);
+ writel(~(SDRAM_ERR_CAUSE_DBE_MASK | SDRAM_ERR_CAUSE_SBE_MASK),
+ drvdata->base + SDRAM_ERR_CAUSE_MSG_REG);
+
+ /* clear error counter registers */
+ if (cnt_sbe)
+ writel(0, drvdata->base + SDRAM_ERR_SBE_COUNT_REG);
+ if (cnt_dbe)
+ writel(0, drvdata->base + SDRAM_ERR_DBE_COUNT_REG);
+
+ if (!cnt_sbe && !cnt_dbe)
+ return;
+
+ if (!(addr & SDRAM_ERR_ADDR_TYPE_MASK)) {
+ if (cnt_sbe)
+ cnt_sbe--;
+ else
+ dev_warn(mci->pdev, "inconsistent SBE count detected\n");
+ } else {
+ if (cnt_dbe)
+ cnt_dbe--;
+ else
+ dev_warn(mci->pdev, "inconsistent DBE count detected\n");
+ }
+
+ /* report earlier errors */
+ if (cnt_sbe)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
+ cnt_sbe, /* error count */
+ 0, 0, 0, /* pfn, offset, syndrome */
+ -1, -1, -1, /* top, mid, low layer */
+ mci->ctl_name,
+ "details unavailable (multiple errors)");
+ if (cnt_dbe)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
+ cnt_dbe, /* error count */
+ 0, 0, 0, /* pfn, offset, syndrome */
+ -1, -1, -1, /* top, mid, low layer */
+ mci->ctl_name,
+ "details unavailable (multiple errors)");
+
+ /* report details for most recent error */
+ cs_val = (addr & SDRAM_ERR_ADDR_CS_MASK) >> SDRAM_ERR_ADDR_CS_OFFSET;
+ bank_val = (addr & SDRAM_ERR_ADDR_BANK_MASK) >> SDRAM_ERR_ADDR_BANK_OFFSET;
+ row_val = (calc_ecc & SDRAM_ERR_CALC_ECC_ROW_MASK) >> SDRAM_ERR_CALC_ECC_ROW_OFFSET;
+ col_val = (addr & SDRAM_ERR_ADDR_COL_MASK) >> SDRAM_ERR_ADDR_COL_OFFSET;
+ syndrome_val = (recv_ecc ^ calc_ecc) & 0xff;
+ addr_val = axp_mc_calc_address(drvdata, cs_val, bank_val, row_val,
+ col_val);
+ msg += sprintf(msg, "row=0x%04x ", row_val); /* 11 chars */
+ msg += sprintf(msg, "bank=0x%x ", bank_val); /* 9 chars */
+ msg += sprintf(msg, "col=0x%04x ", col_val); /* 11 chars */
+ msg += sprintf(msg, "cs=%d", cs_val); /* 4 chars */
+
+ if (!(addr & SDRAM_ERR_ADDR_TYPE_MASK)) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
+ 1, /* error count */
+ addr_val >> PAGE_SHIFT,
+ addr_val & ~PAGE_MASK,
+ syndrome_val,
+ cs_val, -1, -1, /* top, mid, low layer */
+ mci->ctl_name, drvdata->msg);
+ } else {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
+ 1, /* error count */
+ addr_val >> PAGE_SHIFT,
+ addr_val & ~PAGE_MASK,
+ syndrome_val,
+ cs_val, -1, -1, /* top, mid, low layer */
+ mci->ctl_name, drvdata->msg);
+ }
+}
+
+static void axp_mc_read_config(struct mem_ctl_info *mci)
+{
+ struct axp_mc_drvdata *drvdata = mci->pvt_info;
+ uint32_t config, addr_ctrl, rank_ctrl;
+ unsigned int i, cs_struct, cs_size;
+ struct dimm_info *dimm;
+
+ config = readl(drvdata->base + SDRAM_CONFIG_REG);
+ if (config & SDRAM_CONFIG_BUS_WIDTH_MASK)
+ /* 64 bit */
+ drvdata->width = 8;
+ else
+ /* 32 bit */
+ drvdata->width = 4;
+
+ addr_ctrl = readl(drvdata->base + SDRAM_ADDR_CTRL_REG);
+ rank_ctrl = readl(drvdata->base + SDRAM_RANK_CTRL_REG);
+ for (i = 0; i < SDRAM_NUM_CS; i++) {
+ dimm = mci->dimms[i];
+
+ if (!(rank_ctrl & SDRAM_RANK_CTRL_EXIST_MASK(i)))
+ continue;
+
+ drvdata->cs_addr_sel[i] =
+ !!(addr_ctrl & SDRAM_ADDR_CTRL_ADDR_SEL_MASK(i));
+
+ cs_struct = (addr_ctrl & SDRAM_ADDR_CTRL_STRUCT_MASK(i)) >> SDRAM_ADDR_CTRL_STRUCT_OFFSET(i);
+ cs_size = ((addr_ctrl & SDRAM_ADDR_CTRL_SIZE_HIGH_MASK(i)) >> (SDRAM_ADDR_CTRL_SIZE_HIGH_OFFSET(i) - 2) |
+ ((addr_ctrl & SDRAM_ADDR_CTRL_SIZE_LOW_MASK(i)) >> SDRAM_ADDR_CTRL_SIZE_LOW_OFFSET(i)));
+
+ switch (cs_size) {
+ case 0: /* 2GBit */
+ dimm->nr_pages = 524288;
+ break;
+ case 1: /* 256MBit */
+ dimm->nr_pages = 65536;
+ break;
+ case 2: /* 512MBit */
+ dimm->nr_pages = 131072;
+ break;
+ case 3: /* 1GBit */
+ dimm->nr_pages = 262144;
+ break;
+ case 4: /* 4GBit */
+ dimm->nr_pages = 1048576;
+ break;
+ case 5: /* 8GBit */
+ dimm->nr_pages = 2097152;
+ break;
+ }
+ dimm->grain = 8;
+ dimm->dtype = cs_struct ? DEV_X16 : DEV_X8;
+ dimm->mtype = (config & SDRAM_CONFIG_REGISTERED_MASK) ?
+ MEM_RDDR3 : MEM_DDR3;
+ dimm->edac_mode = EDAC_SECDED;
+ }
+}
+
+static const struct of_device_id axp_mc_of_match[] = {
+ {.compatible = "marvell,armada-xp-sdram-controller",},
+ {},
+};
+MODULE_DEVICE_TABLE(of, axp_mc_of_match);
+
+static int axp_mc_probe(struct platform_device *pdev)
+{
+ struct axp_mc_drvdata *drvdata;
+ struct edac_mc_layer layers[1];
+ const struct of_device_id *id;
+ struct mem_ctl_info *mci;
+ void __iomem *base;
+ uint32_t config;
+
+ base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(base)) {
+ dev_err(&pdev->dev, "Unable to map regs\n");
+ return PTR_ERR(base);
+ }
+
+ config = readl(base + SDRAM_CONFIG_REG);
+ if (!(config & SDRAM_CONFIG_ECC_MASK)) {
+ dev_warn(&pdev->dev, "SDRAM ECC is not enabled\n");
+ return -EINVAL;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = SDRAM_NUM_CS;
+ layers[0].is_virt_csrow = true;
+
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*drvdata));
+ if (!mci)
+ return -ENOMEM;
+
+ drvdata = mci->pvt_info;
+ drvdata->base = base;
+ mci->pdev = &pdev->dev;
+ platform_set_drvdata(pdev, mci);
+
+ id = of_match_device(axp_mc_of_match, &pdev->dev);
+ mci->edac_check = axp_mc_check;
+ mci->mtype_cap = MEM_FLAG_DDR3;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = pdev->dev.driver->name;
+ mci->ctl_name = id ? id->compatible : "unknown";
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->scrub_mode = SCRUB_NONE;
+
+ axp_mc_read_config(mci);
+
+ /* These SoCs have a reduced width bus */
+ if (of_machine_is_compatible("marvell,armada380") ||
+ of_machine_is_compatible("marvell,armadaxp-98dx3236"))
+ drvdata->width /= 2;
+
+ /* configure SBE threshold */
+ /* it seems that SBEs are not captured otherwise */
+ writel(1 << SDRAM_ERR_CTRL_THR_OFFSET, drvdata->base + SDRAM_ERR_CTRL_REG);
+
+ /* clear cause registers */
+ writel(~(SDRAM_ERR_CAUSE_DBE_MASK | SDRAM_ERR_CAUSE_SBE_MASK), drvdata->base + SDRAM_ERR_CAUSE_ERR_REG);
+ writel(~(SDRAM_ERR_CAUSE_DBE_MASK | SDRAM_ERR_CAUSE_SBE_MASK), drvdata->base + SDRAM_ERR_CAUSE_MSG_REG);
+
+ /* clear counter registers */
+ writel(0, drvdata->base + SDRAM_ERR_SBE_COUNT_REG);
+ writel(0, drvdata->base + SDRAM_ERR_DBE_COUNT_REG);
+
+ if (edac_mc_add_mc(mci)) {
+ edac_mc_free(mci);
+ return -EINVAL;
+ }
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ return 0;
+}
+
+static int axp_mc_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+ platform_set_drvdata(pdev, NULL);
+
+ return 0;
+}
+
+static struct platform_driver axp_mc_driver = {
+ .probe = axp_mc_probe,
+ .remove = axp_mc_remove,
+ .driver = {
+ .name = "armada_xp_mc_edac",
+ .of_match_table = of_match_ptr(axp_mc_of_match),
+ },
+};
+
+/************************ EDAC Device (L2 Cache) ***************************/
+
+struct aurora_l2_drvdata {
+ void __iomem *base;
+
+ char msg[128];
+
+ /* error injection via debugfs */
+ uint32_t inject_addr;
+ uint32_t inject_mask;
+ uint8_t inject_ctl;
+
+ struct dentry *debugfs;
+};
+
+#ifdef CONFIG_EDAC_DEBUG
+static void aurora_l2_inject(struct aurora_l2_drvdata *drvdata)
+{
+ drvdata->inject_addr &= AURORA_ERR_INJECT_CTL_ADDR_MASK;
+ drvdata->inject_ctl &= AURORA_ERR_INJECT_CTL_EN_MASK;
+ writel(0, drvdata->base + AURORA_ERR_INJECT_CTL_REG);
+ writel(drvdata->inject_mask, drvdata->base + AURORA_ERR_INJECT_MASK_REG);
+ writel(drvdata->inject_addr | drvdata->inject_ctl, drvdata->base + AURORA_ERR_INJECT_CTL_REG);
+}
+#endif
+
+static void aurora_l2_check(struct edac_device_ctl_info *dci)
+{
+ struct aurora_l2_drvdata *drvdata = dci->pvt_info;
+ uint32_t cnt, src, txn, err, attr_cap, addr_cap, way_cap;
+ unsigned int cnt_ce, cnt_ue;
+ char *msg = drvdata->msg;
+ size_t size = sizeof(drvdata->msg);
+ size_t len = 0;
+
+ cnt = readl(drvdata->base + AURORA_ERR_CNT_REG);
+ attr_cap = readl(drvdata->base + AURORA_ERR_ATTR_CAP_REG);
+ addr_cap = readl(drvdata->base + AURORA_ERR_ADDR_CAP_REG);
+ way_cap = readl(drvdata->base + AURORA_ERR_WAY_CAP_REG);
+
+ cnt_ce = (cnt & AURORA_ERR_CNT_CE_MASK) >> AURORA_ERR_CNT_CE_OFFSET;
+ cnt_ue = (cnt & AURORA_ERR_CNT_UE_MASK) >> AURORA_ERR_CNT_UE_OFFSET;
+ /* clear error counter registers */
+ if (cnt_ce || cnt_ue)
+ writel(AURORA_ERR_CNT_CLR, drvdata->base + AURORA_ERR_CNT_REG);
+
+ if (!(attr_cap & AURORA_ERR_ATTR_CAP_VALID))
+ goto clear_remaining;
+
+ src = (attr_cap & AURORA_ERR_ATTR_SRC_MSK) >> AURORA_ERR_ATTR_SRC_OFF;
+ if (src <= 3)
+ len += scnprintf(msg+len, size-len, "src=CPU%d ", src);
+ else
+ len += scnprintf(msg+len, size-len, "src=IO ");
+
+ txn = (attr_cap & AURORA_ERR_ATTR_TXN_MSK) >> AURORA_ERR_ATTR_TXN_OFF;
+ switch (txn) {
+ case 0:
+ len += scnprintf(msg+len, size-len, "txn=Data-Read ");
+ break;
+ case 1:
+ len += scnprintf(msg+len, size-len, "txn=Isn-Read ");
+ break;
+ case 2:
+ len += scnprintf(msg+len, size-len, "txn=Clean-Flush ");
+ break;
+ case 3:
+ len += scnprintf(msg+len, size-len, "txn=Eviction ");
+ break;
+ case 4:
+ len += scnprintf(msg+len, size-len,
+ "txn=Read-Modify-Write ");
+ break;
+ }
+
+ err = (attr_cap & AURORA_ERR_ATTR_ERR_MSK) >> AURORA_ERR_ATTR_ERR_OFF;
+ switch (err) {
+ case 0:
+ len += scnprintf(msg+len, size-len, "err=CorrECC ");
+ break;
+ case 1:
+ len += scnprintf(msg+len, size-len, "err=UnCorrECC ");
+ break;
+ case 2:
+ len += scnprintf(msg+len, size-len, "err=TagParity ");
+ break;
+ }
+
+ len += scnprintf(msg+len, size-len, "addr=0x%x ", addr_cap & AURORA_ERR_ADDR_CAP_ADDR_MASK);
+ len += scnprintf(msg+len, size-len, "index=0x%x ", (way_cap & AURORA_ERR_WAY_IDX_MSK) >> AURORA_ERR_WAY_IDX_OFF);
+ len += scnprintf(msg+len, size-len, "way=0x%x", (way_cap & AURORA_ERR_WAY_CAP_WAY_MASK) >> AURORA_ERR_WAY_CAP_WAY_OFFSET);
+
+ /* clear error capture registers */
+ writel(AURORA_ERR_ATTR_CAP_VALID, drvdata->base + AURORA_ERR_ATTR_CAP_REG);
+ if (err) {
+ /* UnCorrECC or TagParity */
+ if (cnt_ue)
+ cnt_ue--;
+ edac_device_handle_ue(dci, 0, 0, drvdata->msg);
+ } else {
+ if (cnt_ce)
+ cnt_ce--;
+ edac_device_handle_ce(dci, 0, 0, drvdata->msg);
+ }
+
+clear_remaining:
+ /* report remaining errors */
+ while (cnt_ue--)
+ edac_device_handle_ue(dci, 0, 0, "details unavailable (multiple errors)");
+ while (cnt_ce--)
+ edac_device_handle_ue(dci, 0, 0, "details unavailable (multiple errors)");
+}
+
+static void aurora_l2_poll(struct edac_device_ctl_info *dci)
+{
+#ifdef CONFIG_EDAC_DEBUG
+ struct aurora_l2_drvdata *drvdata = dci->pvt_info;
+#endif
+
+ aurora_l2_check(dci);
+#ifdef CONFIG_EDAC_DEBUG
+ aurora_l2_inject(drvdata);
+#endif
+}
+
+static const struct of_device_id aurora_l2_of_match[] = {
+ {.compatible = "marvell,aurora-system-cache",},
+ {},
+};
+MODULE_DEVICE_TABLE(of, aurora_l2_of_match);
+
+static int aurora_l2_probe(struct platform_device *pdev)
+{
+ struct aurora_l2_drvdata *drvdata;
+ struct edac_device_ctl_info *dci;
+ const struct of_device_id *id;
+ uint32_t l2x0_aux_ctrl;
+ void __iomem *base;
+
+ base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(base)) {
+ dev_err(&pdev->dev, "Unable to map regs\n");
+ return PTR_ERR(base);
+ }
+
+ l2x0_aux_ctrl = readl(base + L2X0_AUX_CTRL);
+ if (!(l2x0_aux_ctrl & AURORA_ACR_PARITY_EN))
+ dev_warn(&pdev->dev, "tag parity is not enabled\n");
+ if (!(l2x0_aux_ctrl & AURORA_ACR_ECC_EN))
+ dev_warn(&pdev->dev, "data ECC is not enabled\n");
+
+ dci = edac_device_alloc_ctl_info(sizeof(*drvdata),
+ "cpu", 1, "L", 1, 2, NULL, 0, 0);
+ if (!dci)
+ return -ENOMEM;
+
+ drvdata = dci->pvt_info;
+ drvdata->base = base;
+ dci->dev = &pdev->dev;
+ platform_set_drvdata(pdev, dci);
+
+ id = of_match_device(aurora_l2_of_match, &pdev->dev);
+ dci->edac_check = aurora_l2_poll;
+ dci->mod_name = pdev->dev.driver->name;
+ dci->ctl_name = id ? id->compatible : "unknown";
+ dci->dev_name = dev_name(&pdev->dev);
+
+ /* clear registers */
+ writel(AURORA_ERR_CNT_CLR, drvdata->base + AURORA_ERR_CNT_REG);
+ writel(AURORA_ERR_ATTR_CAP_VALID, drvdata->base + AURORA_ERR_ATTR_CAP_REG);
+
+ if (edac_device_add_device(dci)) {
+ edac_device_free_ctl_info(dci);
+ return -EINVAL;
+ }
+
+#ifdef CONFIG_EDAC_DEBUG
+ drvdata->debugfs = edac_debugfs_create_dir(dev_name(&pdev->dev));
+ if (drvdata->debugfs) {
+ edac_debugfs_create_x32("inject_addr", 0644,
+ drvdata->debugfs,
+ &drvdata->inject_addr);
+ edac_debugfs_create_x32("inject_mask", 0644,
+ drvdata->debugfs,
+ &drvdata->inject_mask);
+ edac_debugfs_create_x8("inject_ctl", 0644,
+ drvdata->debugfs, &drvdata->inject_ctl);
+ }
+#endif
+
+ return 0;
+}
+
+static int aurora_l2_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
+#ifdef CONFIG_EDAC_DEBUG
+ struct aurora_l2_drvdata *drvdata = dci->pvt_info;
+
+ edac_debugfs_remove_recursive(drvdata->debugfs);
+#endif
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);
+ platform_set_drvdata(pdev, NULL);
+
+ return 0;
+}
+
+static struct platform_driver aurora_l2_driver = {
+ .probe = aurora_l2_probe,
+ .remove = aurora_l2_remove,
+ .driver = {
+ .name = "aurora_l2_edac",
+ .of_match_table = of_match_ptr(aurora_l2_of_match),
+ },
+};
+
+/************************ Driver registration ******************************/
+
+static struct platform_driver * const drivers[] = {
+ &axp_mc_driver,
+ &aurora_l2_driver,
+};
+
+static int __init armada_xp_edac_init(void)
+{
+ int res;
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ /* only polling is supported */
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ res = platform_register_drivers(drivers, ARRAY_SIZE(drivers));
+ if (res)
+ pr_warn("Armada XP EDAC drivers fail to register\n");
+
+ return 0;
+}
+module_init(armada_xp_edac_init);
+
+static void __exit armada_xp_edac_exit(void)
+{
+ platform_unregister_drivers(drivers, ARRAY_SIZE(drivers));
+}
+module_exit(armada_xp_edac_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Pengutronix");
+MODULE_DESCRIPTION("EDAC Drivers for Marvell Armada XP SDRAM and L2 Cache Controller");
diff --git a/drivers/edac/aspeed_edac.c b/drivers/edac/aspeed_edac.c
new file mode 100644
index 0000000000..6bd5f88159
--- /dev/null
+++ b/drivers/edac/aspeed_edac.c
@@ -0,0 +1,399 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright 2018, 2019 Cisco Systems
+ */
+
+#include <linux/edac.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/platform_device.h>
+#include <linux/stop_machine.h>
+#include <linux/io.h>
+#include <linux/of_address.h>
+#include <linux/regmap.h>
+#include "edac_module.h"
+
+
+#define DRV_NAME "aspeed-edac"
+
+
+#define ASPEED_MCR_PROT 0x00 /* protection key register */
+#define ASPEED_MCR_CONF 0x04 /* configuration register */
+#define ASPEED_MCR_INTR_CTRL 0x50 /* interrupt control/status register */
+#define ASPEED_MCR_ADDR_UNREC 0x58 /* address of first un-recoverable error */
+#define ASPEED_MCR_ADDR_REC 0x5c /* address of last recoverable error */
+#define ASPEED_MCR_LAST ASPEED_MCR_ADDR_REC
+
+
+#define ASPEED_MCR_PROT_PASSWD 0xfc600309
+#define ASPEED_MCR_CONF_DRAM_TYPE BIT(4)
+#define ASPEED_MCR_CONF_ECC BIT(7)
+#define ASPEED_MCR_INTR_CTRL_CLEAR BIT(31)
+#define ASPEED_MCR_INTR_CTRL_CNT_REC GENMASK(23, 16)
+#define ASPEED_MCR_INTR_CTRL_CNT_UNREC GENMASK(15, 12)
+#define ASPEED_MCR_INTR_CTRL_ENABLE (BIT(0) | BIT(1))
+
+
+static struct regmap *aspeed_regmap;
+
+
+static int regmap_reg_write(void *context, unsigned int reg, unsigned int val)
+{
+ void __iomem *regs = (void __iomem *)context;
+
+ /* enable write to MCR register set */
+ writel(ASPEED_MCR_PROT_PASSWD, regs + ASPEED_MCR_PROT);
+
+ writel(val, regs + reg);
+
+ /* disable write to MCR register set */
+ writel(~ASPEED_MCR_PROT_PASSWD, regs + ASPEED_MCR_PROT);
+
+ return 0;
+}
+
+
+static int regmap_reg_read(void *context, unsigned int reg, unsigned int *val)
+{
+ void __iomem *regs = (void __iomem *)context;
+
+ *val = readl(regs + reg);
+
+ return 0;
+}
+
+static bool regmap_is_volatile(struct device *dev, unsigned int reg)
+{
+ switch (reg) {
+ case ASPEED_MCR_PROT:
+ case ASPEED_MCR_INTR_CTRL:
+ case ASPEED_MCR_ADDR_UNREC:
+ case ASPEED_MCR_ADDR_REC:
+ return true;
+ default:
+ return false;
+ }
+}
+
+
+static const struct regmap_config aspeed_regmap_config = {
+ .reg_bits = 32,
+ .val_bits = 32,
+ .reg_stride = 4,
+ .max_register = ASPEED_MCR_LAST,
+ .reg_write = regmap_reg_write,
+ .reg_read = regmap_reg_read,
+ .volatile_reg = regmap_is_volatile,
+ .fast_io = true,
+};
+
+
+static void count_rec(struct mem_ctl_info *mci, u8 rec_cnt, u32 rec_addr)
+{
+ struct csrow_info *csrow = mci->csrows[0];
+ u32 page, offset, syndrome;
+
+ if (!rec_cnt)
+ return;
+
+ /* report first few errors (if there are) */
+ /* note: no addresses are recorded */
+ if (rec_cnt > 1) {
+ /* page, offset and syndrome are not available */
+ page = 0;
+ offset = 0;
+ syndrome = 0;
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, rec_cnt-1,
+ page, offset, syndrome, 0, 0, -1,
+ "address(es) not available", "");
+ }
+
+ /* report last error */
+ /* note: rec_addr is the last recoverable error addr */
+ page = rec_addr >> PAGE_SHIFT;
+ offset = rec_addr & ~PAGE_MASK;
+ /* syndrome is not available */
+ syndrome = 0;
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ csrow->first_page + page, offset, syndrome,
+ 0, 0, -1, "", "");
+}
+
+
+static void count_un_rec(struct mem_ctl_info *mci, u8 un_rec_cnt,
+ u32 un_rec_addr)
+{
+ struct csrow_info *csrow = mci->csrows[0];
+ u32 page, offset, syndrome;
+
+ if (!un_rec_cnt)
+ return;
+
+ /* report 1. error */
+ /* note: un_rec_addr is the first unrecoverable error addr */
+ page = un_rec_addr >> PAGE_SHIFT;
+ offset = un_rec_addr & ~PAGE_MASK;
+ /* syndrome is not available */
+ syndrome = 0;
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ csrow->first_page + page, offset, syndrome,
+ 0, 0, -1, "", "");
+
+ /* report further errors (if there are) */
+ /* note: no addresses are recorded */
+ if (un_rec_cnt > 1) {
+ /* page, offset and syndrome are not available */
+ page = 0;
+ offset = 0;
+ syndrome = 0;
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, un_rec_cnt-1,
+ page, offset, syndrome, 0, 0, -1,
+ "address(es) not available", "");
+ }
+}
+
+
+static irqreturn_t mcr_isr(int irq, void *arg)
+{
+ struct mem_ctl_info *mci = arg;
+ u32 rec_addr, un_rec_addr;
+ u32 reg50, reg5c, reg58;
+ u8 rec_cnt, un_rec_cnt;
+
+ regmap_read(aspeed_regmap, ASPEED_MCR_INTR_CTRL, &reg50);
+ dev_dbg(mci->pdev, "received edac interrupt w/ mcr register 50: 0x%x\n",
+ reg50);
+
+ /* collect data about recoverable and unrecoverable errors */
+ rec_cnt = (reg50 & ASPEED_MCR_INTR_CTRL_CNT_REC) >> 16;
+ un_rec_cnt = (reg50 & ASPEED_MCR_INTR_CTRL_CNT_UNREC) >> 12;
+
+ dev_dbg(mci->pdev, "%d recoverable interrupts and %d unrecoverable interrupts\n",
+ rec_cnt, un_rec_cnt);
+
+ regmap_read(aspeed_regmap, ASPEED_MCR_ADDR_UNREC, &reg58);
+ un_rec_addr = reg58;
+
+ regmap_read(aspeed_regmap, ASPEED_MCR_ADDR_REC, &reg5c);
+ rec_addr = reg5c;
+
+ /* clear interrupt flags and error counters: */
+ regmap_update_bits(aspeed_regmap, ASPEED_MCR_INTR_CTRL,
+ ASPEED_MCR_INTR_CTRL_CLEAR,
+ ASPEED_MCR_INTR_CTRL_CLEAR);
+
+ regmap_update_bits(aspeed_regmap, ASPEED_MCR_INTR_CTRL,
+ ASPEED_MCR_INTR_CTRL_CLEAR, 0);
+
+ /* process recoverable and unrecoverable errors */
+ count_rec(mci, rec_cnt, rec_addr);
+ count_un_rec(mci, un_rec_cnt, un_rec_addr);
+
+ if (!rec_cnt && !un_rec_cnt)
+ dev_dbg(mci->pdev, "received edac interrupt, but did not find any ECC counters\n");
+
+ regmap_read(aspeed_regmap, ASPEED_MCR_INTR_CTRL, &reg50);
+ dev_dbg(mci->pdev, "edac interrupt handled. mcr reg 50 is now: 0x%x\n",
+ reg50);
+
+ return IRQ_HANDLED;
+}
+
+
+static int config_irq(void *ctx, struct platform_device *pdev)
+{
+ int irq;
+ int rc;
+
+ /* register interrupt handler */
+ irq = platform_get_irq(pdev, 0);
+ dev_dbg(&pdev->dev, "got irq %d\n", irq);
+ if (irq < 0)
+ return irq;
+
+ rc = devm_request_irq(&pdev->dev, irq, mcr_isr, IRQF_TRIGGER_HIGH,
+ DRV_NAME, ctx);
+ if (rc) {
+ dev_err(&pdev->dev, "unable to request irq %d\n", irq);
+ return rc;
+ }
+
+ /* enable interrupts */
+ regmap_update_bits(aspeed_regmap, ASPEED_MCR_INTR_CTRL,
+ ASPEED_MCR_INTR_CTRL_ENABLE,
+ ASPEED_MCR_INTR_CTRL_ENABLE);
+
+ return 0;
+}
+
+
+static int init_csrows(struct mem_ctl_info *mci)
+{
+ struct csrow_info *csrow = mci->csrows[0];
+ u32 nr_pages, dram_type;
+ struct dimm_info *dimm;
+ struct device_node *np;
+ struct resource r;
+ u32 reg04;
+ int rc;
+
+ /* retrieve info about physical memory from device tree */
+ np = of_find_node_by_name(NULL, "memory");
+ if (!np) {
+ dev_err(mci->pdev, "dt: missing /memory node\n");
+ return -ENODEV;
+ }
+
+ rc = of_address_to_resource(np, 0, &r);
+
+ of_node_put(np);
+
+ if (rc) {
+ dev_err(mci->pdev, "dt: failed requesting resource for /memory node\n");
+ return rc;
+ }
+
+ dev_dbg(mci->pdev, "dt: /memory node resources: first page %pR, PAGE_SHIFT macro=0x%x\n",
+ &r, PAGE_SHIFT);
+
+ csrow->first_page = r.start >> PAGE_SHIFT;
+ nr_pages = resource_size(&r) >> PAGE_SHIFT;
+ csrow->last_page = csrow->first_page + nr_pages - 1;
+
+ regmap_read(aspeed_regmap, ASPEED_MCR_CONF, &reg04);
+ dram_type = (reg04 & ASPEED_MCR_CONF_DRAM_TYPE) ? MEM_DDR4 : MEM_DDR3;
+
+ dimm = csrow->channels[0]->dimm;
+ dimm->mtype = dram_type;
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->nr_pages = nr_pages / csrow->nr_channels;
+
+ dev_dbg(mci->pdev, "initialized dimm with first_page=0x%lx and nr_pages=0x%x\n",
+ csrow->first_page, nr_pages);
+
+ return 0;
+}
+
+
+static int aspeed_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct edac_mc_layer layers[2];
+ struct mem_ctl_info *mci;
+ void __iomem *regs;
+ u32 reg04;
+ int rc;
+
+ regs = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(regs))
+ return PTR_ERR(regs);
+
+ aspeed_regmap = devm_regmap_init(dev, NULL, (__force void *)regs,
+ &aspeed_regmap_config);
+ if (IS_ERR(aspeed_regmap))
+ return PTR_ERR(aspeed_regmap);
+
+ /* bail out if ECC mode is not configured */
+ regmap_read(aspeed_regmap, ASPEED_MCR_CONF, &reg04);
+ if (!(reg04 & ASPEED_MCR_CONF_ECC)) {
+ dev_err(&pdev->dev, "ECC mode is not configured in u-boot\n");
+ return -EPERM;
+ }
+
+ edac_op_state = EDAC_OPSTATE_INT;
+
+ /* allocate & init EDAC MC data structure */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 1;
+ layers[1].is_virt_csrow = false;
+
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR3 | MEM_FLAG_DDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ mci->scrub_mode = SCRUB_HW_SRC;
+ mci->mod_name = DRV_NAME;
+ mci->ctl_name = "MIC";
+ mci->dev_name = dev_name(&pdev->dev);
+
+ rc = init_csrows(mci);
+ if (rc) {
+ dev_err(&pdev->dev, "failed to init csrows\n");
+ goto probe_exit02;
+ }
+
+ platform_set_drvdata(pdev, mci);
+
+ /* register with edac core */
+ rc = edac_mc_add_mc(mci);
+ if (rc) {
+ dev_err(&pdev->dev, "failed to register with EDAC core\n");
+ goto probe_exit02;
+ }
+
+ /* register interrupt handler and enable interrupts */
+ rc = config_irq(mci, pdev);
+ if (rc) {
+ dev_err(&pdev->dev, "failed setting up irq\n");
+ goto probe_exit01;
+ }
+
+ return 0;
+
+probe_exit01:
+ edac_mc_del_mc(&pdev->dev);
+probe_exit02:
+ edac_mc_free(mci);
+ return rc;
+}
+
+
+static int aspeed_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ /* disable interrupts */
+ regmap_update_bits(aspeed_regmap, ASPEED_MCR_INTR_CTRL,
+ ASPEED_MCR_INTR_CTRL_ENABLE, 0);
+
+ /* free resources */
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (mci)
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+
+static const struct of_device_id aspeed_of_match[] = {
+ { .compatible = "aspeed,ast2400-sdram-edac" },
+ { .compatible = "aspeed,ast2500-sdram-edac" },
+ { .compatible = "aspeed,ast2600-sdram-edac" },
+ {},
+};
+
+MODULE_DEVICE_TABLE(of, aspeed_of_match);
+
+static struct platform_driver aspeed_driver = {
+ .driver = {
+ .name = DRV_NAME,
+ .of_match_table = aspeed_of_match
+ },
+ .probe = aspeed_probe,
+ .remove = aspeed_remove
+};
+module_platform_driver(aspeed_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Stefan Schaeckeler <sschaeck@cisco.com>");
+MODULE_DESCRIPTION("Aspeed BMC SoC EDAC driver");
+MODULE_VERSION("1.0");
diff --git a/drivers/edac/bluefield_edac.c b/drivers/edac/bluefield_edac.c
new file mode 100644
index 0000000000..e4736eb37b
--- /dev/null
+++ b/drivers/edac/bluefield_edac.c
@@ -0,0 +1,356 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Bluefield-specific EDAC driver.
+ *
+ * Copyright (c) 2019 Mellanox Technologies.
+ */
+
+#include <linux/acpi.h>
+#include <linux/arm-smccc.h>
+#include <linux/bitfield.h>
+#include <linux/edac.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#include "edac_module.h"
+
+#define DRIVER_NAME "bluefield-edac"
+
+/*
+ * Mellanox BlueField EMI (External Memory Interface) register definitions.
+ */
+
+#define MLXBF_ECC_CNT 0x340
+#define MLXBF_ECC_CNT__SERR_CNT GENMASK(15, 0)
+#define MLXBF_ECC_CNT__DERR_CNT GENMASK(31, 16)
+
+#define MLXBF_ECC_ERR 0x348
+#define MLXBF_ECC_ERR__SECC BIT(0)
+#define MLXBF_ECC_ERR__DECC BIT(16)
+
+#define MLXBF_ECC_LATCH_SEL 0x354
+#define MLXBF_ECC_LATCH_SEL__START BIT(24)
+
+#define MLXBF_ERR_ADDR_0 0x358
+
+#define MLXBF_ERR_ADDR_1 0x37c
+
+#define MLXBF_SYNDROM 0x35c
+#define MLXBF_SYNDROM__DERR BIT(0)
+#define MLXBF_SYNDROM__SERR BIT(1)
+#define MLXBF_SYNDROM__SYN GENMASK(25, 16)
+
+#define MLXBF_ADD_INFO 0x364
+#define MLXBF_ADD_INFO__ERR_PRANK GENMASK(9, 8)
+
+#define MLXBF_EDAC_MAX_DIMM_PER_MC 2
+#define MLXBF_EDAC_ERROR_GRAIN 8
+
+/*
+ * Request MLNX_SIP_GET_DIMM_INFO
+ *
+ * Retrieve information about DIMM on a certain slot.
+ *
+ * Call register usage:
+ * a0: MLNX_SIP_GET_DIMM_INFO
+ * a1: (Memory controller index) << 16 | (Dimm index in memory controller)
+ * a2-7: not used.
+ *
+ * Return status:
+ * a0: MLXBF_DIMM_INFO defined below describing the DIMM.
+ * a1-3: not used.
+ */
+#define MLNX_SIP_GET_DIMM_INFO 0x82000008
+
+/* Format for the SMC response about the memory information */
+#define MLXBF_DIMM_INFO__SIZE_GB GENMASK_ULL(15, 0)
+#define MLXBF_DIMM_INFO__IS_RDIMM BIT(16)
+#define MLXBF_DIMM_INFO__IS_LRDIMM BIT(17)
+#define MLXBF_DIMM_INFO__IS_NVDIMM BIT(18)
+#define MLXBF_DIMM_INFO__RANKS GENMASK_ULL(23, 21)
+#define MLXBF_DIMM_INFO__PACKAGE_X GENMASK_ULL(31, 24)
+
+struct bluefield_edac_priv {
+ int dimm_ranks[MLXBF_EDAC_MAX_DIMM_PER_MC];
+ void __iomem *emi_base;
+ int dimm_per_mc;
+};
+
+static u64 smc_call1(u64 smc_op, u64 smc_arg)
+{
+ struct arm_smccc_res res;
+
+ arm_smccc_smc(smc_op, smc_arg, 0, 0, 0, 0, 0, 0, &res);
+
+ return res.a0;
+}
+
+/*
+ * Gather the ECC information from the External Memory Interface registers
+ * and report it to the edac handler.
+ */
+static void bluefield_gather_report_ecc(struct mem_ctl_info *mci,
+ int error_cnt,
+ int is_single_ecc)
+{
+ struct bluefield_edac_priv *priv = mci->pvt_info;
+ u32 dram_additional_info, err_prank, edea0, edea1;
+ u32 ecc_latch_select, dram_syndrom, serr, derr, syndrom;
+ enum hw_event_mc_err_type ecc_type;
+ u64 ecc_dimm_addr;
+ int ecc_dimm;
+
+ ecc_type = is_single_ecc ? HW_EVENT_ERR_CORRECTED :
+ HW_EVENT_ERR_UNCORRECTED;
+
+ /*
+ * Tell the External Memory Interface to populate the relevant
+ * registers with information about the last ECC error occurrence.
+ */
+ ecc_latch_select = MLXBF_ECC_LATCH_SEL__START;
+ writel(ecc_latch_select, priv->emi_base + MLXBF_ECC_LATCH_SEL);
+
+ /*
+ * Verify that the ECC reported info in the registers is of the
+ * same type as the one asked to report. If not, just report the
+ * error without the detailed information.
+ */
+ dram_syndrom = readl(priv->emi_base + MLXBF_SYNDROM);
+ serr = FIELD_GET(MLXBF_SYNDROM__SERR, dram_syndrom);
+ derr = FIELD_GET(MLXBF_SYNDROM__DERR, dram_syndrom);
+ syndrom = FIELD_GET(MLXBF_SYNDROM__SYN, dram_syndrom);
+
+ if ((is_single_ecc && !serr) || (!is_single_ecc && !derr)) {
+ edac_mc_handle_error(ecc_type, mci, error_cnt, 0, 0, 0,
+ 0, 0, -1, mci->ctl_name, "");
+ return;
+ }
+
+ dram_additional_info = readl(priv->emi_base + MLXBF_ADD_INFO);
+ err_prank = FIELD_GET(MLXBF_ADD_INFO__ERR_PRANK, dram_additional_info);
+
+ ecc_dimm = (err_prank >= 2 && priv->dimm_ranks[0] <= 2) ? 1 : 0;
+
+ edea0 = readl(priv->emi_base + MLXBF_ERR_ADDR_0);
+ edea1 = readl(priv->emi_base + MLXBF_ERR_ADDR_1);
+
+ ecc_dimm_addr = ((u64)edea1 << 32) | edea0;
+
+ edac_mc_handle_error(ecc_type, mci, error_cnt,
+ PFN_DOWN(ecc_dimm_addr),
+ offset_in_page(ecc_dimm_addr),
+ syndrom, ecc_dimm, 0, 0, mci->ctl_name, "");
+}
+
+static void bluefield_edac_check(struct mem_ctl_info *mci)
+{
+ struct bluefield_edac_priv *priv = mci->pvt_info;
+ u32 ecc_count, single_error_count, double_error_count, ecc_error = 0;
+
+ /*
+ * The memory controller might not be initialized by the firmware
+ * when there isn't memory, which may lead to bad register readings.
+ */
+ if (mci->edac_cap == EDAC_FLAG_NONE)
+ return;
+
+ ecc_count = readl(priv->emi_base + MLXBF_ECC_CNT);
+ single_error_count = FIELD_GET(MLXBF_ECC_CNT__SERR_CNT, ecc_count);
+ double_error_count = FIELD_GET(MLXBF_ECC_CNT__DERR_CNT, ecc_count);
+
+ if (single_error_count) {
+ ecc_error |= MLXBF_ECC_ERR__SECC;
+
+ bluefield_gather_report_ecc(mci, single_error_count, 1);
+ }
+
+ if (double_error_count) {
+ ecc_error |= MLXBF_ECC_ERR__DECC;
+
+ bluefield_gather_report_ecc(mci, double_error_count, 0);
+ }
+
+ /* Write to clear reported errors. */
+ if (ecc_count)
+ writel(ecc_error, priv->emi_base + MLXBF_ECC_ERR);
+}
+
+/* Initialize the DIMMs information for the given memory controller. */
+static void bluefield_edac_init_dimms(struct mem_ctl_info *mci)
+{
+ struct bluefield_edac_priv *priv = mci->pvt_info;
+ int mem_ctrl_idx = mci->mc_idx;
+ struct dimm_info *dimm;
+ u64 smc_info, smc_arg;
+ int is_empty = 1, i;
+
+ for (i = 0; i < priv->dimm_per_mc; i++) {
+ dimm = mci->dimms[i];
+
+ smc_arg = mem_ctrl_idx << 16 | i;
+ smc_info = smc_call1(MLNX_SIP_GET_DIMM_INFO, smc_arg);
+
+ if (!FIELD_GET(MLXBF_DIMM_INFO__SIZE_GB, smc_info)) {
+ dimm->mtype = MEM_EMPTY;
+ continue;
+ }
+
+ is_empty = 0;
+
+ dimm->edac_mode = EDAC_SECDED;
+
+ if (FIELD_GET(MLXBF_DIMM_INFO__IS_NVDIMM, smc_info))
+ dimm->mtype = MEM_NVDIMM;
+ else if (FIELD_GET(MLXBF_DIMM_INFO__IS_LRDIMM, smc_info))
+ dimm->mtype = MEM_LRDDR4;
+ else if (FIELD_GET(MLXBF_DIMM_INFO__IS_RDIMM, smc_info))
+ dimm->mtype = MEM_RDDR4;
+ else
+ dimm->mtype = MEM_DDR4;
+
+ dimm->nr_pages =
+ FIELD_GET(MLXBF_DIMM_INFO__SIZE_GB, smc_info) *
+ (SZ_1G / PAGE_SIZE);
+ dimm->grain = MLXBF_EDAC_ERROR_GRAIN;
+
+ /* Mem controller for BlueField only supports x4, x8 and x16 */
+ switch (FIELD_GET(MLXBF_DIMM_INFO__PACKAGE_X, smc_info)) {
+ case 4:
+ dimm->dtype = DEV_X4;
+ break;
+ case 8:
+ dimm->dtype = DEV_X8;
+ break;
+ case 16:
+ dimm->dtype = DEV_X16;
+ break;
+ default:
+ dimm->dtype = DEV_UNKNOWN;
+ }
+
+ priv->dimm_ranks[i] =
+ FIELD_GET(MLXBF_DIMM_INFO__RANKS, smc_info);
+ }
+
+ if (is_empty)
+ mci->edac_cap = EDAC_FLAG_NONE;
+ else
+ mci->edac_cap = EDAC_FLAG_SECDED;
+}
+
+static int bluefield_edac_mc_probe(struct platform_device *pdev)
+{
+ struct bluefield_edac_priv *priv;
+ struct device *dev = &pdev->dev;
+ struct edac_mc_layer layers[1];
+ struct mem_ctl_info *mci;
+ struct resource *emi_res;
+ unsigned int mc_idx, dimm_count;
+ int rc, ret;
+
+ /* Read the MSS (Memory SubSystem) index from ACPI table. */
+ if (device_property_read_u32(dev, "mss_number", &mc_idx)) {
+ dev_warn(dev, "bf_edac: MSS number unknown\n");
+ return -EINVAL;
+ }
+
+ /* Read the DIMMs per MC from ACPI table. */
+ if (device_property_read_u32(dev, "dimm_per_mc", &dimm_count)) {
+ dev_warn(dev, "bf_edac: DIMMs per MC unknown\n");
+ return -EINVAL;
+ }
+
+ if (dimm_count > MLXBF_EDAC_MAX_DIMM_PER_MC) {
+ dev_warn(dev, "bf_edac: DIMMs per MC not valid\n");
+ return -EINVAL;
+ }
+
+ emi_res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!emi_res)
+ return -EINVAL;
+
+ layers[0].type = EDAC_MC_LAYER_SLOT;
+ layers[0].size = dimm_count;
+ layers[0].is_virt_csrow = true;
+
+ mci = edac_mc_alloc(mc_idx, ARRAY_SIZE(layers), layers, sizeof(*priv));
+ if (!mci)
+ return -ENOMEM;
+
+ priv = mci->pvt_info;
+
+ priv->dimm_per_mc = dimm_count;
+ priv->emi_base = devm_ioremap_resource(dev, emi_res);
+ if (IS_ERR(priv->emi_base)) {
+ dev_err(dev, "failed to map EMI IO resource\n");
+ ret = PTR_ERR(priv->emi_base);
+ goto err;
+ }
+
+ mci->pdev = dev;
+ mci->mtype_cap = MEM_FLAG_DDR4 | MEM_FLAG_RDDR4 |
+ MEM_FLAG_LRDDR4 | MEM_FLAG_NVDIMM;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+
+ mci->mod_name = DRIVER_NAME;
+ mci->ctl_name = "BlueField_Memory_Controller";
+ mci->dev_name = dev_name(dev);
+ mci->edac_check = bluefield_edac_check;
+
+ /* Initialize mci with the actual populated DIMM information. */
+ bluefield_edac_init_dimms(mci);
+
+ platform_set_drvdata(pdev, mci);
+
+ /* Register with EDAC core */
+ rc = edac_mc_add_mc(mci);
+ if (rc) {
+ dev_err(dev, "failed to register with EDAC core\n");
+ ret = rc;
+ goto err;
+ }
+
+ /* Only POLL mode supported so far. */
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ return 0;
+
+err:
+ edac_mc_free(mci);
+
+ return ret;
+
+}
+
+static int bluefield_edac_mc_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+static const struct acpi_device_id bluefield_mc_acpi_ids[] = {
+ {"MLNXBF08", 0},
+ {}
+};
+
+MODULE_DEVICE_TABLE(acpi, bluefield_mc_acpi_ids);
+
+static struct platform_driver bluefield_edac_mc_driver = {
+ .driver = {
+ .name = DRIVER_NAME,
+ .acpi_match_table = bluefield_mc_acpi_ids,
+ },
+ .probe = bluefield_edac_mc_probe,
+ .remove = bluefield_edac_mc_remove,
+};
+
+module_platform_driver(bluefield_edac_mc_driver);
+
+MODULE_DESCRIPTION("Mellanox BlueField memory edac driver");
+MODULE_AUTHOR("Mellanox Technologies");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/cell_edac.c b/drivers/edac/cell_edac.c
new file mode 100644
index 0000000000..bc1f341640
--- /dev/null
+++ b/drivers/edac/cell_edac.c
@@ -0,0 +1,282 @@
+/*
+ * Cell MIC driver for ECC counting
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ * <benh@kernel.crashing.org>
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ */
+#undef DEBUG
+
+#include <linux/edac.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+#include <linux/stop_machine.h>
+#include <linux/io.h>
+#include <linux/of_address.h>
+#include <asm/machdep.h>
+#include <asm/cell-regs.h>
+
+#include "edac_module.h"
+
+struct cell_edac_priv
+{
+ struct cbe_mic_tm_regs __iomem *regs;
+ int node;
+ int chanmask;
+#ifdef DEBUG
+ u64 prev_fir;
+#endif
+};
+
+static void cell_edac_count_ce(struct mem_ctl_info *mci, int chan, u64 ar)
+{
+ struct cell_edac_priv *priv = mci->pvt_info;
+ struct csrow_info *csrow = mci->csrows[0];
+ unsigned long address, pfn, offset, syndrome;
+
+ dev_dbg(mci->pdev, "ECC CE err on node %d, channel %d, ar = 0x%016llx\n",
+ priv->node, chan, ar);
+
+ /* Address decoding is likely a bit bogus, to dbl check */
+ address = (ar & 0xffffffffe0000000ul) >> 29;
+ if (priv->chanmask == 0x3)
+ address = (address << 1) | chan;
+ pfn = address >> PAGE_SHIFT;
+ offset = address & ~PAGE_MASK;
+ syndrome = (ar & 0x000000001fe00000ul) >> 21;
+
+ /* TODO: Decoding of the error address */
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ csrow->first_page + pfn, offset, syndrome,
+ 0, chan, -1, "", "");
+}
+
+static void cell_edac_count_ue(struct mem_ctl_info *mci, int chan, u64 ar)
+{
+ struct cell_edac_priv *priv = mci->pvt_info;
+ struct csrow_info *csrow = mci->csrows[0];
+ unsigned long address, pfn, offset;
+
+ dev_dbg(mci->pdev, "ECC UE err on node %d, channel %d, ar = 0x%016llx\n",
+ priv->node, chan, ar);
+
+ /* Address decoding is likely a bit bogus, to dbl check */
+ address = (ar & 0xffffffffe0000000ul) >> 29;
+ if (priv->chanmask == 0x3)
+ address = (address << 1) | chan;
+ pfn = address >> PAGE_SHIFT;
+ offset = address & ~PAGE_MASK;
+
+ /* TODO: Decoding of the error address */
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ csrow->first_page + pfn, offset, 0,
+ 0, chan, -1, "", "");
+}
+
+static void cell_edac_check(struct mem_ctl_info *mci)
+{
+ struct cell_edac_priv *priv = mci->pvt_info;
+ u64 fir, addreg, clear = 0;
+
+ fir = in_be64(&priv->regs->mic_fir);
+#ifdef DEBUG
+ if (fir != priv->prev_fir) {
+ dev_dbg(mci->pdev, "fir change : 0x%016lx\n", fir);
+ priv->prev_fir = fir;
+ }
+#endif
+ if ((priv->chanmask & 0x1) && (fir & CBE_MIC_FIR_ECC_SINGLE_0_ERR)) {
+ addreg = in_be64(&priv->regs->mic_df_ecc_address_0);
+ clear |= CBE_MIC_FIR_ECC_SINGLE_0_RESET;
+ cell_edac_count_ce(mci, 0, addreg);
+ }
+ if ((priv->chanmask & 0x2) && (fir & CBE_MIC_FIR_ECC_SINGLE_1_ERR)) {
+ addreg = in_be64(&priv->regs->mic_df_ecc_address_1);
+ clear |= CBE_MIC_FIR_ECC_SINGLE_1_RESET;
+ cell_edac_count_ce(mci, 1, addreg);
+ }
+ if ((priv->chanmask & 0x1) && (fir & CBE_MIC_FIR_ECC_MULTI_0_ERR)) {
+ addreg = in_be64(&priv->regs->mic_df_ecc_address_0);
+ clear |= CBE_MIC_FIR_ECC_MULTI_0_RESET;
+ cell_edac_count_ue(mci, 0, addreg);
+ }
+ if ((priv->chanmask & 0x2) && (fir & CBE_MIC_FIR_ECC_MULTI_1_ERR)) {
+ addreg = in_be64(&priv->regs->mic_df_ecc_address_1);
+ clear |= CBE_MIC_FIR_ECC_MULTI_1_RESET;
+ cell_edac_count_ue(mci, 1, addreg);
+ }
+
+ /* The procedure for clearing FIR bits is a bit ... weird */
+ if (clear) {
+ fir &= ~(CBE_MIC_FIR_ECC_ERR_MASK | CBE_MIC_FIR_ECC_SET_MASK);
+ fir |= CBE_MIC_FIR_ECC_RESET_MASK;
+ fir &= ~clear;
+ out_be64(&priv->regs->mic_fir, fir);
+ (void)in_be64(&priv->regs->mic_fir);
+
+ mb(); /* sync up */
+#ifdef DEBUG
+ fir = in_be64(&priv->regs->mic_fir);
+ dev_dbg(mci->pdev, "fir clear : 0x%016lx\n", fir);
+#endif
+ }
+}
+
+static void cell_edac_init_csrows(struct mem_ctl_info *mci)
+{
+ struct csrow_info *csrow = mci->csrows[0];
+ struct dimm_info *dimm;
+ struct cell_edac_priv *priv = mci->pvt_info;
+ struct device_node *np;
+ int j;
+ u32 nr_pages;
+
+ for_each_node_by_name(np, "memory") {
+ struct resource r;
+
+ /* We "know" that the Cell firmware only creates one entry
+ * in the "memory" nodes. If that changes, this code will
+ * need to be adapted.
+ */
+ if (of_address_to_resource(np, 0, &r))
+ continue;
+ if (of_node_to_nid(np) != priv->node)
+ continue;
+ csrow->first_page = r.start >> PAGE_SHIFT;
+ nr_pages = resource_size(&r) >> PAGE_SHIFT;
+ csrow->last_page = csrow->first_page + nr_pages - 1;
+
+ for (j = 0; j < csrow->nr_channels; j++) {
+ dimm = csrow->channels[j]->dimm;
+ dimm->mtype = MEM_XDR;
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->nr_pages = nr_pages / csrow->nr_channels;
+ }
+ dev_dbg(mci->pdev,
+ "Initialized on node %d, chanmask=0x%x,"
+ " first_page=0x%lx, nr_pages=0x%x\n",
+ priv->node, priv->chanmask,
+ csrow->first_page, nr_pages);
+ break;
+ }
+ of_node_put(np);
+}
+
+static int cell_edac_probe(struct platform_device *pdev)
+{
+ struct cbe_mic_tm_regs __iomem *regs;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct cell_edac_priv *priv;
+ u64 reg;
+ int rc, chanmask, num_chans;
+
+ regs = cbe_get_cpu_mic_tm_regs(cbe_node_to_cpu(pdev->id));
+ if (regs == NULL)
+ return -ENODEV;
+
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ /* Get channel population */
+ reg = in_be64(&regs->mic_mnt_cfg);
+ dev_dbg(&pdev->dev, "MIC_MNT_CFG = 0x%016llx\n", reg);
+ chanmask = 0;
+ if (reg & CBE_MIC_MNT_CFG_CHAN_0_POP)
+ chanmask |= 0x1;
+ if (reg & CBE_MIC_MNT_CFG_CHAN_1_POP)
+ chanmask |= 0x2;
+ if (chanmask == 0) {
+ dev_warn(&pdev->dev,
+ "Yuck ! No channel populated ? Aborting !\n");
+ return -ENODEV;
+ }
+ dev_dbg(&pdev->dev, "Initial FIR = 0x%016llx\n",
+ in_be64(&regs->mic_fir));
+
+ /* Allocate & init EDAC MC data structure */
+ num_chans = chanmask == 3 ? 2 : 1;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = num_chans;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(pdev->id, ARRAY_SIZE(layers), layers,
+ sizeof(struct cell_edac_priv));
+ if (mci == NULL)
+ return -ENOMEM;
+ priv = mci->pvt_info;
+ priv->regs = regs;
+ priv->node = pdev->id;
+ priv->chanmask = chanmask;
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_XDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ mci->mod_name = "cell_edac";
+ mci->ctl_name = "MIC";
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->edac_check = cell_edac_check;
+ cell_edac_init_csrows(mci);
+
+ /* Register with EDAC core */
+ rc = edac_mc_add_mc(mci);
+ if (rc) {
+ dev_err(&pdev->dev, "failed to register with EDAC core\n");
+ edac_mc_free(mci);
+ return rc;
+ }
+
+ return 0;
+}
+
+static int cell_edac_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = edac_mc_del_mc(&pdev->dev);
+ if (mci)
+ edac_mc_free(mci);
+ return 0;
+}
+
+static struct platform_driver cell_edac_driver = {
+ .driver = {
+ .name = "cbe-mic",
+ },
+ .probe = cell_edac_probe,
+ .remove = cell_edac_remove,
+};
+
+static int __init cell_edac_init(void)
+{
+ /* Sanity check registers data structure */
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_df_ecc_address_0) != 0xf8);
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_df_ecc_address_1) != 0x1b8);
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_df_config) != 0x218);
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_fir) != 0x230);
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_mnt_cfg) != 0x210);
+ BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs,
+ mic_exc) != 0x208);
+
+ return platform_driver_register(&cell_edac_driver);
+}
+
+static void __exit cell_edac_exit(void)
+{
+ platform_driver_unregister(&cell_edac_driver);
+}
+
+module_init(cell_edac_init);
+module_exit(cell_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Benjamin Herrenschmidt <benh@kernel.crashing.org>");
+MODULE_DESCRIPTION("ECC counting for Cell MIC");
diff --git a/drivers/edac/cpc925_edac.c b/drivers/edac/cpc925_edac.c
new file mode 100644
index 0000000000..9797e6d60d
--- /dev/null
+++ b/drivers/edac/cpc925_edac.c
@@ -0,0 +1,1067 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * cpc925_edac.c, EDAC driver for IBM CPC925 Bridge and Memory Controller.
+ *
+ * Copyright (c) 2008 Wind River Systems, Inc.
+ *
+ * Authors: Cao Qingtao <qingtao.cao@windriver.com>
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/edac.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/gfp.h>
+
+#include "edac_module.h"
+
+#define CPC925_EDAC_REVISION " Ver: 1.0.0"
+#define CPC925_EDAC_MOD_STR "cpc925_edac"
+
+#define cpc925_printk(level, fmt, arg...) \
+ edac_printk(level, "CPC925", fmt, ##arg)
+
+#define cpc925_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "CPC925", fmt, ##arg)
+
+/*
+ * CPC925 registers are of 32 bits with bit0 defined at the
+ * most significant bit and bit31 at that of least significant.
+ */
+#define CPC925_BITS_PER_REG 32
+#define CPC925_BIT(nr) (1UL << (CPC925_BITS_PER_REG - 1 - nr))
+
+/*
+ * EDAC device names for the error detections of
+ * CPU Interface and Hypertransport Link.
+ */
+#define CPC925_CPU_ERR_DEV "cpu"
+#define CPC925_HT_LINK_DEV "htlink"
+
+/* Suppose DDR Refresh cycle is 15.6 microsecond */
+#define CPC925_REF_FREQ 0xFA69
+#define CPC925_SCRUB_BLOCK_SIZE 64 /* bytes */
+#define CPC925_NR_CSROWS 8
+
+/*
+ * All registers and bits definitions are taken from
+ * "CPC925 Bridge and Memory Controller User Manual, SA14-2761-02".
+ */
+
+/*
+ * CPU and Memory Controller Registers
+ */
+/************************************************************
+ * Processor Interface Exception Mask Register (APIMASK)
+ ************************************************************/
+#define REG_APIMASK_OFFSET 0x30070
+enum apimask_bits {
+ APIMASK_DART = CPC925_BIT(0), /* DART Exception */
+ APIMASK_ADI0 = CPC925_BIT(1), /* Handshake Error on PI0_ADI */
+ APIMASK_ADI1 = CPC925_BIT(2), /* Handshake Error on PI1_ADI */
+ APIMASK_STAT = CPC925_BIT(3), /* Status Exception */
+ APIMASK_DERR = CPC925_BIT(4), /* Data Error Exception */
+ APIMASK_ADRS0 = CPC925_BIT(5), /* Addressing Exception on PI0 */
+ APIMASK_ADRS1 = CPC925_BIT(6), /* Addressing Exception on PI1 */
+ /* BIT(7) Reserved */
+ APIMASK_ECC_UE_H = CPC925_BIT(8), /* UECC upper */
+ APIMASK_ECC_CE_H = CPC925_BIT(9), /* CECC upper */
+ APIMASK_ECC_UE_L = CPC925_BIT(10), /* UECC lower */
+ APIMASK_ECC_CE_L = CPC925_BIT(11), /* CECC lower */
+
+ CPU_MASK_ENABLE = (APIMASK_DART | APIMASK_ADI0 | APIMASK_ADI1 |
+ APIMASK_STAT | APIMASK_DERR | APIMASK_ADRS0 |
+ APIMASK_ADRS1),
+ ECC_MASK_ENABLE = (APIMASK_ECC_UE_H | APIMASK_ECC_CE_H |
+ APIMASK_ECC_UE_L | APIMASK_ECC_CE_L),
+};
+#define APIMASK_ADI(n) CPC925_BIT(((n)+1))
+
+/************************************************************
+ * Processor Interface Exception Register (APIEXCP)
+ ************************************************************/
+#define REG_APIEXCP_OFFSET 0x30060
+enum apiexcp_bits {
+ APIEXCP_DART = CPC925_BIT(0), /* DART Exception */
+ APIEXCP_ADI0 = CPC925_BIT(1), /* Handshake Error on PI0_ADI */
+ APIEXCP_ADI1 = CPC925_BIT(2), /* Handshake Error on PI1_ADI */
+ APIEXCP_STAT = CPC925_BIT(3), /* Status Exception */
+ APIEXCP_DERR = CPC925_BIT(4), /* Data Error Exception */
+ APIEXCP_ADRS0 = CPC925_BIT(5), /* Addressing Exception on PI0 */
+ APIEXCP_ADRS1 = CPC925_BIT(6), /* Addressing Exception on PI1 */
+ /* BIT(7) Reserved */
+ APIEXCP_ECC_UE_H = CPC925_BIT(8), /* UECC upper */
+ APIEXCP_ECC_CE_H = CPC925_BIT(9), /* CECC upper */
+ APIEXCP_ECC_UE_L = CPC925_BIT(10), /* UECC lower */
+ APIEXCP_ECC_CE_L = CPC925_BIT(11), /* CECC lower */
+
+ CPU_EXCP_DETECTED = (APIEXCP_DART | APIEXCP_ADI0 | APIEXCP_ADI1 |
+ APIEXCP_STAT | APIEXCP_DERR | APIEXCP_ADRS0 |
+ APIEXCP_ADRS1),
+ UECC_EXCP_DETECTED = (APIEXCP_ECC_UE_H | APIEXCP_ECC_UE_L),
+ CECC_EXCP_DETECTED = (APIEXCP_ECC_CE_H | APIEXCP_ECC_CE_L),
+ ECC_EXCP_DETECTED = (UECC_EXCP_DETECTED | CECC_EXCP_DETECTED),
+};
+
+/************************************************************
+ * Memory Bus Configuration Register (MBCR)
+************************************************************/
+#define REG_MBCR_OFFSET 0x2190
+#define MBCR_64BITCFG_SHIFT 23
+#define MBCR_64BITCFG_MASK (1UL << MBCR_64BITCFG_SHIFT)
+#define MBCR_64BITBUS_SHIFT 22
+#define MBCR_64BITBUS_MASK (1UL << MBCR_64BITBUS_SHIFT)
+
+/************************************************************
+ * Memory Bank Mode Register (MBMR)
+************************************************************/
+#define REG_MBMR_OFFSET 0x21C0
+#define MBMR_MODE_MAX_VALUE 0xF
+#define MBMR_MODE_SHIFT 25
+#define MBMR_MODE_MASK (MBMR_MODE_MAX_VALUE << MBMR_MODE_SHIFT)
+#define MBMR_BBA_SHIFT 24
+#define MBMR_BBA_MASK (1UL << MBMR_BBA_SHIFT)
+
+/************************************************************
+ * Memory Bank Boundary Address Register (MBBAR)
+ ************************************************************/
+#define REG_MBBAR_OFFSET 0x21D0
+#define MBBAR_BBA_MAX_VALUE 0xFF
+#define MBBAR_BBA_SHIFT 24
+#define MBBAR_BBA_MASK (MBBAR_BBA_MAX_VALUE << MBBAR_BBA_SHIFT)
+
+/************************************************************
+ * Memory Scrub Control Register (MSCR)
+ ************************************************************/
+#define REG_MSCR_OFFSET 0x2400
+#define MSCR_SCRUB_MOD_MASK 0xC0000000 /* scrub_mod - bit0:1*/
+#define MSCR_BACKGR_SCRUB 0x40000000 /* 01 */
+#define MSCR_SI_SHIFT 16 /* si - bit8:15*/
+#define MSCR_SI_MAX_VALUE 0xFF
+#define MSCR_SI_MASK (MSCR_SI_MAX_VALUE << MSCR_SI_SHIFT)
+
+/************************************************************
+ * Memory Scrub Range Start Register (MSRSR)
+ ************************************************************/
+#define REG_MSRSR_OFFSET 0x2410
+
+/************************************************************
+ * Memory Scrub Range End Register (MSRER)
+ ************************************************************/
+#define REG_MSRER_OFFSET 0x2420
+
+/************************************************************
+ * Memory Scrub Pattern Register (MSPR)
+ ************************************************************/
+#define REG_MSPR_OFFSET 0x2430
+
+/************************************************************
+ * Memory Check Control Register (MCCR)
+ ************************************************************/
+#define REG_MCCR_OFFSET 0x2440
+enum mccr_bits {
+ MCCR_ECC_EN = CPC925_BIT(0), /* ECC high and low check */
+};
+
+/************************************************************
+ * Memory Check Range End Register (MCRER)
+ ************************************************************/
+#define REG_MCRER_OFFSET 0x2450
+
+/************************************************************
+ * Memory Error Address Register (MEAR)
+ ************************************************************/
+#define REG_MEAR_OFFSET 0x2460
+#define MEAR_BCNT_MAX_VALUE 0x3
+#define MEAR_BCNT_SHIFT 30
+#define MEAR_BCNT_MASK (MEAR_BCNT_MAX_VALUE << MEAR_BCNT_SHIFT)
+#define MEAR_RANK_MAX_VALUE 0x7
+#define MEAR_RANK_SHIFT 27
+#define MEAR_RANK_MASK (MEAR_RANK_MAX_VALUE << MEAR_RANK_SHIFT)
+#define MEAR_COL_MAX_VALUE 0x7FF
+#define MEAR_COL_SHIFT 16
+#define MEAR_COL_MASK (MEAR_COL_MAX_VALUE << MEAR_COL_SHIFT)
+#define MEAR_BANK_MAX_VALUE 0x3
+#define MEAR_BANK_SHIFT 14
+#define MEAR_BANK_MASK (MEAR_BANK_MAX_VALUE << MEAR_BANK_SHIFT)
+#define MEAR_ROW_MASK 0x00003FFF
+
+/************************************************************
+ * Memory Error Syndrome Register (MESR)
+ ************************************************************/
+#define REG_MESR_OFFSET 0x2470
+#define MESR_ECC_SYN_H_MASK 0xFF00
+#define MESR_ECC_SYN_L_MASK 0x00FF
+
+/************************************************************
+ * Memory Mode Control Register (MMCR)
+ ************************************************************/
+#define REG_MMCR_OFFSET 0x2500
+enum mmcr_bits {
+ MMCR_REG_DIMM_MODE = CPC925_BIT(3),
+};
+
+/*
+ * HyperTransport Link Registers
+ */
+/************************************************************
+ * Error Handling/Enumeration Scratch Pad Register (ERRCTRL)
+ ************************************************************/
+#define REG_ERRCTRL_OFFSET 0x70140
+enum errctrl_bits { /* nonfatal interrupts for */
+ ERRCTRL_SERR_NF = CPC925_BIT(0), /* system error */
+ ERRCTRL_CRC_NF = CPC925_BIT(1), /* CRC error */
+ ERRCTRL_RSP_NF = CPC925_BIT(2), /* Response error */
+ ERRCTRL_EOC_NF = CPC925_BIT(3), /* End-Of-Chain error */
+ ERRCTRL_OVF_NF = CPC925_BIT(4), /* Overflow error */
+ ERRCTRL_PROT_NF = CPC925_BIT(5), /* Protocol error */
+
+ ERRCTRL_RSP_ERR = CPC925_BIT(6), /* Response error received */
+ ERRCTRL_CHN_FAL = CPC925_BIT(7), /* Sync flooding detected */
+
+ HT_ERRCTRL_ENABLE = (ERRCTRL_SERR_NF | ERRCTRL_CRC_NF |
+ ERRCTRL_RSP_NF | ERRCTRL_EOC_NF |
+ ERRCTRL_OVF_NF | ERRCTRL_PROT_NF),
+ HT_ERRCTRL_DETECTED = (ERRCTRL_RSP_ERR | ERRCTRL_CHN_FAL),
+};
+
+/************************************************************
+ * Link Configuration and Link Control Register (LINKCTRL)
+ ************************************************************/
+#define REG_LINKCTRL_OFFSET 0x70110
+enum linkctrl_bits {
+ LINKCTRL_CRC_ERR = (CPC925_BIT(22) | CPC925_BIT(23)),
+ LINKCTRL_LINK_FAIL = CPC925_BIT(27),
+
+ HT_LINKCTRL_DETECTED = (LINKCTRL_CRC_ERR | LINKCTRL_LINK_FAIL),
+};
+
+/************************************************************
+ * Link FreqCap/Error/Freq/Revision ID Register (LINKERR)
+ ************************************************************/
+#define REG_LINKERR_OFFSET 0x70120
+enum linkerr_bits {
+ LINKERR_EOC_ERR = CPC925_BIT(17), /* End-Of-Chain error */
+ LINKERR_OVF_ERR = CPC925_BIT(18), /* Receive Buffer Overflow */
+ LINKERR_PROT_ERR = CPC925_BIT(19), /* Protocol error */
+
+ HT_LINKERR_DETECTED = (LINKERR_EOC_ERR | LINKERR_OVF_ERR |
+ LINKERR_PROT_ERR),
+};
+
+/************************************************************
+ * Bridge Control Register (BRGCTRL)
+ ************************************************************/
+#define REG_BRGCTRL_OFFSET 0x70300
+enum brgctrl_bits {
+ BRGCTRL_DETSERR = CPC925_BIT(0), /* SERR on Secondary Bus */
+ BRGCTRL_SECBUSRESET = CPC925_BIT(9), /* Secondary Bus Reset */
+};
+
+/* Private structure for edac memory controller */
+struct cpc925_mc_pdata {
+ void __iomem *vbase;
+ unsigned long total_mem;
+ const char *name;
+ int edac_idx;
+};
+
+/* Private structure for common edac device */
+struct cpc925_dev_info {
+ void __iomem *vbase;
+ struct platform_device *pdev;
+ char *ctl_name;
+ int edac_idx;
+ struct edac_device_ctl_info *edac_dev;
+ void (*init)(struct cpc925_dev_info *dev_info);
+ void (*exit)(struct cpc925_dev_info *dev_info);
+ void (*check)(struct edac_device_ctl_info *edac_dev);
+};
+
+/* Get total memory size from Open Firmware DTB */
+static void get_total_mem(struct cpc925_mc_pdata *pdata)
+{
+ struct device_node *np = NULL;
+ const unsigned int *reg, *reg_end;
+ int len, sw, aw;
+ unsigned long start, size;
+
+ np = of_find_node_by_type(NULL, "memory");
+ if (!np)
+ return;
+
+ aw = of_n_addr_cells(np);
+ sw = of_n_size_cells(np);
+ reg = (const unsigned int *)of_get_property(np, "reg", &len);
+ reg_end = reg + len/4;
+
+ pdata->total_mem = 0;
+ do {
+ start = of_read_number(reg, aw);
+ reg += aw;
+ size = of_read_number(reg, sw);
+ reg += sw;
+ edac_dbg(1, "start 0x%lx, size 0x%lx\n", start, size);
+ pdata->total_mem += size;
+ } while (reg < reg_end);
+
+ of_node_put(np);
+ edac_dbg(0, "total_mem 0x%lx\n", pdata->total_mem);
+}
+
+static void cpc925_init_csrows(struct mem_ctl_info *mci)
+{
+ struct cpc925_mc_pdata *pdata = mci->pvt_info;
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ enum dev_type dtype;
+ int index, j;
+ u32 mbmr, mbbar, bba, grain;
+ unsigned long row_size, nr_pages, last_nr_pages = 0;
+
+ get_total_mem(pdata);
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ mbmr = __raw_readl(pdata->vbase + REG_MBMR_OFFSET +
+ 0x20 * index);
+ mbbar = __raw_readl(pdata->vbase + REG_MBBAR_OFFSET +
+ 0x20 + index);
+ bba = (((mbmr & MBMR_BBA_MASK) >> MBMR_BBA_SHIFT) << 8) |
+ ((mbbar & MBBAR_BBA_MASK) >> MBBAR_BBA_SHIFT);
+
+ if (bba == 0)
+ continue; /* not populated */
+
+ csrow = mci->csrows[index];
+
+ row_size = bba * (1UL << 28); /* 256M */
+ csrow->first_page = last_nr_pages;
+ nr_pages = row_size >> PAGE_SHIFT;
+ csrow->last_page = csrow->first_page + nr_pages - 1;
+ last_nr_pages = csrow->last_page + 1;
+
+ switch (csrow->nr_channels) {
+ case 1: /* Single channel */
+ grain = 32; /* four-beat burst of 32 bytes */
+ break;
+ case 2: /* Dual channel */
+ default:
+ grain = 64; /* four-beat burst of 64 bytes */
+ break;
+ }
+ switch ((mbmr & MBMR_MODE_MASK) >> MBMR_MODE_SHIFT) {
+ case 6: /* 0110, no way to differentiate X8 VS X16 */
+ case 5: /* 0101 */
+ case 8: /* 1000 */
+ dtype = DEV_X16;
+ break;
+ case 7: /* 0111 */
+ case 9: /* 1001 */
+ dtype = DEV_X8;
+ break;
+ default:
+ dtype = DEV_UNKNOWN;
+ break;
+ }
+ for (j = 0; j < csrow->nr_channels; j++) {
+ dimm = csrow->channels[j]->dimm;
+ dimm->nr_pages = nr_pages / csrow->nr_channels;
+ dimm->mtype = MEM_RDDR;
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->grain = grain;
+ dimm->dtype = dtype;
+ }
+ }
+}
+
+/* Enable memory controller ECC detection */
+static void cpc925_mc_init(struct mem_ctl_info *mci)
+{
+ struct cpc925_mc_pdata *pdata = mci->pvt_info;
+ u32 apimask;
+ u32 mccr;
+
+ /* Enable various ECC error exceptions */
+ apimask = __raw_readl(pdata->vbase + REG_APIMASK_OFFSET);
+ if ((apimask & ECC_MASK_ENABLE) == 0) {
+ apimask |= ECC_MASK_ENABLE;
+ __raw_writel(apimask, pdata->vbase + REG_APIMASK_OFFSET);
+ }
+
+ /* Enable ECC detection */
+ mccr = __raw_readl(pdata->vbase + REG_MCCR_OFFSET);
+ if ((mccr & MCCR_ECC_EN) == 0) {
+ mccr |= MCCR_ECC_EN;
+ __raw_writel(mccr, pdata->vbase + REG_MCCR_OFFSET);
+ }
+}
+
+/* Disable memory controller ECC detection */
+static void cpc925_mc_exit(struct mem_ctl_info *mci)
+{
+ /*
+ * WARNING:
+ * We are supposed to clear the ECC error detection bits,
+ * and it will be no problem to do so. However, once they
+ * are cleared here if we want to re-install CPC925 EDAC
+ * module later, setting them up in cpc925_mc_init() will
+ * trigger machine check exception.
+ * Also, it's ok to leave ECC error detection bits enabled,
+ * since they are reset to 1 by default or by boot loader.
+ */
+
+ return;
+}
+
+/*
+ * Revert DDR column/row/bank addresses into page frame number and
+ * offset in page.
+ *
+ * Suppose memory mode is 0x0111(128-bit mode, identical DIMM pairs),
+ * physical address(PA) bits to column address(CA) bits mappings are:
+ * CA 0 1 2 3 4 5 6 7 8 9 10
+ * PA 59 58 57 56 55 54 53 52 51 50 49
+ *
+ * physical address(PA) bits to bank address(BA) bits mappings are:
+ * BA 0 1
+ * PA 43 44
+ *
+ * physical address(PA) bits to row address(RA) bits mappings are:
+ * RA 0 1 2 3 4 5 6 7 8 9 10 11 12
+ * PA 36 35 34 48 47 46 45 40 41 42 39 38 37
+ */
+static void cpc925_mc_get_pfn(struct mem_ctl_info *mci, u32 mear,
+ unsigned long *pfn, unsigned long *offset, int *csrow)
+{
+ u32 bcnt, rank, col, bank, row;
+ u32 c;
+ unsigned long pa;
+ int i;
+
+ bcnt = (mear & MEAR_BCNT_MASK) >> MEAR_BCNT_SHIFT;
+ rank = (mear & MEAR_RANK_MASK) >> MEAR_RANK_SHIFT;
+ col = (mear & MEAR_COL_MASK) >> MEAR_COL_SHIFT;
+ bank = (mear & MEAR_BANK_MASK) >> MEAR_BANK_SHIFT;
+ row = mear & MEAR_ROW_MASK;
+
+ *csrow = rank;
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (mci->csrows[rank]->first_page == 0) {
+ cpc925_mc_printk(mci, KERN_ERR, "ECC occurs in a "
+ "non-populated csrow, broken hardware?\n");
+ return;
+ }
+#endif
+
+ /* Revert csrow number */
+ pa = mci->csrows[rank]->first_page << PAGE_SHIFT;
+
+ /* Revert column address */
+ col += bcnt;
+ for (i = 0; i < 11; i++) {
+ c = col & 0x1;
+ col >>= 1;
+ pa |= c << (14 - i);
+ }
+
+ /* Revert bank address */
+ pa |= bank << 19;
+
+ /* Revert row address, in 4 steps */
+ for (i = 0; i < 3; i++) {
+ c = row & 0x1;
+ row >>= 1;
+ pa |= c << (26 - i);
+ }
+
+ for (i = 0; i < 3; i++) {
+ c = row & 0x1;
+ row >>= 1;
+ pa |= c << (21 + i);
+ }
+
+ for (i = 0; i < 4; i++) {
+ c = row & 0x1;
+ row >>= 1;
+ pa |= c << (18 - i);
+ }
+
+ for (i = 0; i < 3; i++) {
+ c = row & 0x1;
+ row >>= 1;
+ pa |= c << (29 - i);
+ }
+
+ *offset = pa & (PAGE_SIZE - 1);
+ *pfn = pa >> PAGE_SHIFT;
+
+ edac_dbg(0, "ECC physical address 0x%lx\n", pa);
+}
+
+static int cpc925_mc_find_channel(struct mem_ctl_info *mci, u16 syndrome)
+{
+ if ((syndrome & MESR_ECC_SYN_H_MASK) == 0)
+ return 0;
+
+ if ((syndrome & MESR_ECC_SYN_L_MASK) == 0)
+ return 1;
+
+ cpc925_mc_printk(mci, KERN_INFO, "Unexpected syndrome value: 0x%x\n",
+ syndrome);
+ return 1;
+}
+
+/* Check memory controller registers for ECC errors */
+static void cpc925_mc_check(struct mem_ctl_info *mci)
+{
+ struct cpc925_mc_pdata *pdata = mci->pvt_info;
+ u32 apiexcp;
+ u32 mear;
+ u32 mesr;
+ u16 syndrome;
+ unsigned long pfn = 0, offset = 0;
+ int csrow = 0, channel = 0;
+
+ /* APIEXCP is cleared when read */
+ apiexcp = __raw_readl(pdata->vbase + REG_APIEXCP_OFFSET);
+ if ((apiexcp & ECC_EXCP_DETECTED) == 0)
+ return;
+
+ mesr = __raw_readl(pdata->vbase + REG_MESR_OFFSET);
+ syndrome = mesr | (MESR_ECC_SYN_H_MASK | MESR_ECC_SYN_L_MASK);
+
+ mear = __raw_readl(pdata->vbase + REG_MEAR_OFFSET);
+
+ /* Revert column/row addresses into page frame number, etc */
+ cpc925_mc_get_pfn(mci, mear, &pfn, &offset, &csrow);
+
+ if (apiexcp & CECC_EXCP_DETECTED) {
+ cpc925_mc_printk(mci, KERN_INFO, "DRAM CECC Fault\n");
+ channel = cpc925_mc_find_channel(mci, syndrome);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ pfn, offset, syndrome,
+ csrow, channel, -1,
+ mci->ctl_name, "");
+ }
+
+ if (apiexcp & UECC_EXCP_DETECTED) {
+ cpc925_mc_printk(mci, KERN_INFO, "DRAM UECC Fault\n");
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ pfn, offset, 0,
+ csrow, -1, -1,
+ mci->ctl_name, "");
+ }
+
+ cpc925_mc_printk(mci, KERN_INFO, "Dump registers:\n");
+ cpc925_mc_printk(mci, KERN_INFO, "APIMASK 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_APIMASK_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "APIEXCP 0x%08x\n",
+ apiexcp);
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Ctrl 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MSCR_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge Start 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MSRSR_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Rge End 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MSRER_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Scrub Pattern 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MSPR_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Ctrl 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MCCR_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Chk Rge End 0x%08x\n",
+ __raw_readl(pdata->vbase + REG_MCRER_OFFSET));
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Err Address 0x%08x\n",
+ mesr);
+ cpc925_mc_printk(mci, KERN_INFO, "Mem Err Syndrome 0x%08x\n",
+ syndrome);
+}
+
+/******************** CPU err device********************************/
+static u32 cpc925_cpu_mask_disabled(void)
+{
+ struct device_node *cpunode;
+ static u32 mask = 0;
+
+ /* use cached value if available */
+ if (mask != 0)
+ return mask;
+
+ mask = APIMASK_ADI0 | APIMASK_ADI1;
+
+ for_each_of_cpu_node(cpunode) {
+ const u32 *reg = of_get_property(cpunode, "reg", NULL);
+ if (reg == NULL || *reg > 2) {
+ cpc925_printk(KERN_ERR, "Bad reg value at %pOF\n", cpunode);
+ continue;
+ }
+
+ mask &= ~APIMASK_ADI(*reg);
+ }
+
+ if (mask != (APIMASK_ADI0 | APIMASK_ADI1)) {
+ /* We assume that each CPU sits on it's own PI and that
+ * for present CPUs the reg property equals to the PI
+ * interface id */
+ cpc925_printk(KERN_WARNING,
+ "Assuming PI id is equal to CPU MPIC id!\n");
+ }
+
+ return mask;
+}
+
+/* Enable CPU Errors detection */
+static void cpc925_cpu_init(struct cpc925_dev_info *dev_info)
+{
+ u32 apimask;
+ u32 cpumask;
+
+ apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
+
+ cpumask = cpc925_cpu_mask_disabled();
+ if (apimask & cpumask) {
+ cpc925_printk(KERN_WARNING, "CPU(s) not present, "
+ "but enabled in APIMASK, disabling\n");
+ apimask &= ~cpumask;
+ }
+
+ if ((apimask & CPU_MASK_ENABLE) == 0)
+ apimask |= CPU_MASK_ENABLE;
+
+ __raw_writel(apimask, dev_info->vbase + REG_APIMASK_OFFSET);
+}
+
+/* Disable CPU Errors detection */
+static void cpc925_cpu_exit(struct cpc925_dev_info *dev_info)
+{
+ /*
+ * WARNING:
+ * We are supposed to clear the CPU error detection bits,
+ * and it will be no problem to do so. However, once they
+ * are cleared here if we want to re-install CPC925 EDAC
+ * module later, setting them up in cpc925_cpu_init() will
+ * trigger machine check exception.
+ * Also, it's ok to leave CPU error detection bits enabled,
+ * since they are reset to 1 by default.
+ */
+
+ return;
+}
+
+/* Check for CPU Errors */
+static void cpc925_cpu_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
+ u32 apiexcp;
+ u32 apimask;
+
+ /* APIEXCP is cleared when read */
+ apiexcp = __raw_readl(dev_info->vbase + REG_APIEXCP_OFFSET);
+ if ((apiexcp & CPU_EXCP_DETECTED) == 0)
+ return;
+
+ if ((apiexcp & ~cpc925_cpu_mask_disabled()) == 0)
+ return;
+
+ apimask = __raw_readl(dev_info->vbase + REG_APIMASK_OFFSET);
+ cpc925_printk(KERN_INFO, "Processor Interface Fault\n"
+ "Processor Interface register dump:\n");
+ cpc925_printk(KERN_INFO, "APIMASK 0x%08x\n", apimask);
+ cpc925_printk(KERN_INFO, "APIEXCP 0x%08x\n", apiexcp);
+
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+/******************** HT Link err device****************************/
+/* Enable HyperTransport Link Error detection */
+static void cpc925_htlink_init(struct cpc925_dev_info *dev_info)
+{
+ u32 ht_errctrl;
+
+ ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
+ if ((ht_errctrl & HT_ERRCTRL_ENABLE) == 0) {
+ ht_errctrl |= HT_ERRCTRL_ENABLE;
+ __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
+ }
+}
+
+/* Disable HyperTransport Link Error detection */
+static void cpc925_htlink_exit(struct cpc925_dev_info *dev_info)
+{
+ u32 ht_errctrl;
+
+ ht_errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
+ ht_errctrl &= ~HT_ERRCTRL_ENABLE;
+ __raw_writel(ht_errctrl, dev_info->vbase + REG_ERRCTRL_OFFSET);
+}
+
+/* Check for HyperTransport Link errors */
+static void cpc925_htlink_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct cpc925_dev_info *dev_info = edac_dev->pvt_info;
+ u32 brgctrl = __raw_readl(dev_info->vbase + REG_BRGCTRL_OFFSET);
+ u32 linkctrl = __raw_readl(dev_info->vbase + REG_LINKCTRL_OFFSET);
+ u32 errctrl = __raw_readl(dev_info->vbase + REG_ERRCTRL_OFFSET);
+ u32 linkerr = __raw_readl(dev_info->vbase + REG_LINKERR_OFFSET);
+
+ if (!((brgctrl & BRGCTRL_DETSERR) ||
+ (linkctrl & HT_LINKCTRL_DETECTED) ||
+ (errctrl & HT_ERRCTRL_DETECTED) ||
+ (linkerr & HT_LINKERR_DETECTED)))
+ return;
+
+ cpc925_printk(KERN_INFO, "HT Link Fault\n"
+ "HT register dump:\n");
+ cpc925_printk(KERN_INFO, "Bridge Ctrl 0x%08x\n",
+ brgctrl);
+ cpc925_printk(KERN_INFO, "Link Config Ctrl 0x%08x\n",
+ linkctrl);
+ cpc925_printk(KERN_INFO, "Error Enum and Ctrl 0x%08x\n",
+ errctrl);
+ cpc925_printk(KERN_INFO, "Link Error 0x%08x\n",
+ linkerr);
+
+ /* Clear by write 1 */
+ if (brgctrl & BRGCTRL_DETSERR)
+ __raw_writel(BRGCTRL_DETSERR,
+ dev_info->vbase + REG_BRGCTRL_OFFSET);
+
+ if (linkctrl & HT_LINKCTRL_DETECTED)
+ __raw_writel(HT_LINKCTRL_DETECTED,
+ dev_info->vbase + REG_LINKCTRL_OFFSET);
+
+ /* Initiate Secondary Bus Reset to clear the chain failure */
+ if (errctrl & ERRCTRL_CHN_FAL)
+ __raw_writel(BRGCTRL_SECBUSRESET,
+ dev_info->vbase + REG_BRGCTRL_OFFSET);
+
+ if (errctrl & ERRCTRL_RSP_ERR)
+ __raw_writel(ERRCTRL_RSP_ERR,
+ dev_info->vbase + REG_ERRCTRL_OFFSET);
+
+ if (linkerr & HT_LINKERR_DETECTED)
+ __raw_writel(HT_LINKERR_DETECTED,
+ dev_info->vbase + REG_LINKERR_OFFSET);
+
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static struct cpc925_dev_info cpc925_devs[] = {
+ {
+ .ctl_name = CPC925_CPU_ERR_DEV,
+ .init = cpc925_cpu_init,
+ .exit = cpc925_cpu_exit,
+ .check = cpc925_cpu_check,
+ },
+ {
+ .ctl_name = CPC925_HT_LINK_DEV,
+ .init = cpc925_htlink_init,
+ .exit = cpc925_htlink_exit,
+ .check = cpc925_htlink_check,
+ },
+ { }
+};
+
+/*
+ * Add CPU Err detection and HyperTransport Link Err detection
+ * as common "edac_device", they have no corresponding device
+ * nodes in the Open Firmware DTB and we have to add platform
+ * devices for them. Also, they will share the MMIO with that
+ * of memory controller.
+ */
+static void cpc925_add_edac_devices(void __iomem *vbase)
+{
+ struct cpc925_dev_info *dev_info;
+
+ if (!vbase) {
+ cpc925_printk(KERN_ERR, "MMIO not established yet\n");
+ return;
+ }
+
+ for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
+ dev_info->vbase = vbase;
+ dev_info->pdev = platform_device_register_simple(
+ dev_info->ctl_name, 0, NULL, 0);
+ if (IS_ERR(dev_info->pdev)) {
+ cpc925_printk(KERN_ERR,
+ "Can't register platform device for %s\n",
+ dev_info->ctl_name);
+ continue;
+ }
+
+ /*
+ * Don't have to allocate private structure but
+ * make use of cpc925_devs[] instead.
+ */
+ dev_info->edac_idx = edac_device_alloc_index();
+ dev_info->edac_dev =
+ edac_device_alloc_ctl_info(0, dev_info->ctl_name,
+ 1, NULL, 0, 0, NULL, 0, dev_info->edac_idx);
+ if (!dev_info->edac_dev) {
+ cpc925_printk(KERN_ERR, "No memory for edac device\n");
+ goto err1;
+ }
+
+ dev_info->edac_dev->pvt_info = dev_info;
+ dev_info->edac_dev->dev = &dev_info->pdev->dev;
+ dev_info->edac_dev->ctl_name = dev_info->ctl_name;
+ dev_info->edac_dev->mod_name = CPC925_EDAC_MOD_STR;
+ dev_info->edac_dev->dev_name = dev_name(&dev_info->pdev->dev);
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ dev_info->edac_dev->edac_check = dev_info->check;
+
+ if (dev_info->init)
+ dev_info->init(dev_info);
+
+ if (edac_device_add_device(dev_info->edac_dev) > 0) {
+ cpc925_printk(KERN_ERR,
+ "Unable to add edac device for %s\n",
+ dev_info->ctl_name);
+ goto err2;
+ }
+
+ edac_dbg(0, "Successfully added edac device for %s\n",
+ dev_info->ctl_name);
+
+ continue;
+
+err2:
+ if (dev_info->exit)
+ dev_info->exit(dev_info);
+ edac_device_free_ctl_info(dev_info->edac_dev);
+err1:
+ platform_device_unregister(dev_info->pdev);
+ }
+}
+
+/*
+ * Delete the common "edac_device" for CPU Err Detection
+ * and HyperTransport Link Err Detection
+ */
+static void cpc925_del_edac_devices(void)
+{
+ struct cpc925_dev_info *dev_info;
+
+ for (dev_info = &cpc925_devs[0]; dev_info->init; dev_info++) {
+ if (dev_info->edac_dev) {
+ edac_device_del_device(dev_info->edac_dev->dev);
+ edac_device_free_ctl_info(dev_info->edac_dev);
+ platform_device_unregister(dev_info->pdev);
+ }
+
+ if (dev_info->exit)
+ dev_info->exit(dev_info);
+
+ edac_dbg(0, "Successfully deleted edac device for %s\n",
+ dev_info->ctl_name);
+ }
+}
+
+/* Convert current back-ground scrub rate into byte/sec bandwidth */
+static int cpc925_get_sdram_scrub_rate(struct mem_ctl_info *mci)
+{
+ struct cpc925_mc_pdata *pdata = mci->pvt_info;
+ int bw;
+ u32 mscr;
+ u8 si;
+
+ mscr = __raw_readl(pdata->vbase + REG_MSCR_OFFSET);
+ si = (mscr & MSCR_SI_MASK) >> MSCR_SI_SHIFT;
+
+ edac_dbg(0, "Mem Scrub Ctrl Register 0x%x\n", mscr);
+
+ if (((mscr & MSCR_SCRUB_MOD_MASK) != MSCR_BACKGR_SCRUB) ||
+ (si == 0)) {
+ cpc925_mc_printk(mci, KERN_INFO, "Scrub mode not enabled\n");
+ bw = 0;
+ } else
+ bw = CPC925_SCRUB_BLOCK_SIZE * 0xFA67 / si;
+
+ return bw;
+}
+
+/* Return 0 for single channel; 1 for dual channel */
+static int cpc925_mc_get_channels(void __iomem *vbase)
+{
+ int dual = 0;
+ u32 mbcr;
+
+ mbcr = __raw_readl(vbase + REG_MBCR_OFFSET);
+
+ /*
+ * Dual channel only when 128-bit wide physical bus
+ * and 128-bit configuration.
+ */
+ if (((mbcr & MBCR_64BITCFG_MASK) == 0) &&
+ ((mbcr & MBCR_64BITBUS_MASK) == 0))
+ dual = 1;
+
+ edac_dbg(0, "%s channel\n", (dual > 0) ? "Dual" : "Single");
+
+ return dual;
+}
+
+static int cpc925_probe(struct platform_device *pdev)
+{
+ static int edac_mc_idx;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ void __iomem *vbase;
+ struct cpc925_mc_pdata *pdata;
+ struct resource *r;
+ int res = 0, nr_channels;
+
+ edac_dbg(0, "%s platform device found!\n", pdev->name);
+
+ if (!devres_open_group(&pdev->dev, cpc925_probe, GFP_KERNEL)) {
+ res = -ENOMEM;
+ goto out;
+ }
+
+ r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!r) {
+ cpc925_printk(KERN_ERR, "Unable to get resource\n");
+ res = -ENOENT;
+ goto err1;
+ }
+
+ if (!devm_request_mem_region(&pdev->dev,
+ r->start,
+ resource_size(r),
+ pdev->name)) {
+ cpc925_printk(KERN_ERR, "Unable to request mem region\n");
+ res = -EBUSY;
+ goto err1;
+ }
+
+ vbase = devm_ioremap(&pdev->dev, r->start, resource_size(r));
+ if (!vbase) {
+ cpc925_printk(KERN_ERR, "Unable to ioremap device\n");
+ res = -ENOMEM;
+ goto err2;
+ }
+
+ nr_channels = cpc925_mc_get_channels(vbase) + 1;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = CPC925_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = nr_channels;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(edac_mc_idx, ARRAY_SIZE(layers), layers,
+ sizeof(struct cpc925_mc_pdata));
+ if (!mci) {
+ cpc925_printk(KERN_ERR, "No memory for mem_ctl_info\n");
+ res = -ENOMEM;
+ goto err2;
+ }
+
+ pdata = mci->pvt_info;
+ pdata->vbase = vbase;
+ pdata->edac_idx = edac_mc_idx++;
+ pdata->name = pdev->name;
+
+ mci->pdev = &pdev->dev;
+ platform_set_drvdata(pdev, mci);
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_DDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = CPC925_EDAC_MOD_STR;
+ mci->ctl_name = pdev->name;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ mci->edac_check = cpc925_mc_check;
+
+ mci->ctl_page_to_phys = NULL;
+ mci->scrub_mode = SCRUB_SW_SRC;
+ mci->set_sdram_scrub_rate = NULL;
+ mci->get_sdram_scrub_rate = cpc925_get_sdram_scrub_rate;
+
+ cpc925_init_csrows(mci);
+
+ /* Setup memory controller registers */
+ cpc925_mc_init(mci);
+
+ if (edac_mc_add_mc(mci) > 0) {
+ cpc925_mc_printk(mci, KERN_ERR, "Failed edac_mc_add_mc()\n");
+ goto err3;
+ }
+
+ cpc925_add_edac_devices(vbase);
+
+ /* get this far and it's successful */
+ edac_dbg(0, "success\n");
+
+ res = 0;
+ goto out;
+
+err3:
+ cpc925_mc_exit(mci);
+ edac_mc_free(mci);
+err2:
+ devm_release_mem_region(&pdev->dev, r->start, resource_size(r));
+err1:
+ devres_release_group(&pdev->dev, cpc925_probe);
+out:
+ return res;
+}
+
+static int cpc925_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ /*
+ * Delete common edac devices before edac mc, because
+ * the former share the MMIO of the latter.
+ */
+ cpc925_del_edac_devices();
+ cpc925_mc_exit(mci);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+static struct platform_driver cpc925_edac_driver = {
+ .probe = cpc925_probe,
+ .remove = cpc925_remove,
+ .driver = {
+ .name = "cpc925_edac",
+ }
+};
+
+static int __init cpc925_edac_init(void)
+{
+ int ret = 0;
+
+ printk(KERN_INFO "IBM CPC925 EDAC driver " CPC925_EDAC_REVISION "\n");
+ printk(KERN_INFO "\t(c) 2008 Wind River Systems, Inc\n");
+
+ /* Only support POLL mode so far */
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+ ret = platform_driver_register(&cpc925_edac_driver);
+ if (ret) {
+ printk(KERN_WARNING "Failed to register %s\n",
+ CPC925_EDAC_MOD_STR);
+ }
+
+ return ret;
+}
+
+static void __exit cpc925_edac_exit(void)
+{
+ platform_driver_unregister(&cpc925_edac_driver);
+}
+
+module_init(cpc925_edac_init);
+module_exit(cpc925_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cao Qingtao <qingtao.cao@windriver.com>");
+MODULE_DESCRIPTION("IBM CPC925 Bridge and MC EDAC kernel module");
diff --git a/drivers/edac/debugfs.c b/drivers/edac/debugfs.c
new file mode 100644
index 0000000000..4804332d99
--- /dev/null
+++ b/drivers/edac/debugfs.c
@@ -0,0 +1,151 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include "edac_module.h"
+
+static struct dentry *edac_debugfs;
+
+static ssize_t edac_fake_inject_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct device *dev = file->private_data;
+ struct mem_ctl_info *mci = to_mci(dev);
+ static enum hw_event_mc_err_type type;
+ u16 errcount = mci->fake_inject_count;
+
+ if (!errcount)
+ errcount = 1;
+
+ type = mci->fake_inject_ue ? HW_EVENT_ERR_UNCORRECTED
+ : HW_EVENT_ERR_CORRECTED;
+
+ printk(KERN_DEBUG
+ "Generating %d %s fake error%s to %d.%d.%d to test core handling. NOTE: this won't test the driver-specific decoding logic.\n",
+ errcount,
+ (type == HW_EVENT_ERR_UNCORRECTED) ? "UE" : "CE",
+ errcount > 1 ? "s" : "",
+ mci->fake_inject_layer[0],
+ mci->fake_inject_layer[1],
+ mci->fake_inject_layer[2]
+ );
+ edac_mc_handle_error(type, mci, errcount, 0, 0, 0,
+ mci->fake_inject_layer[0],
+ mci->fake_inject_layer[1],
+ mci->fake_inject_layer[2],
+ "FAKE ERROR", "for EDAC testing only");
+
+ return count;
+}
+
+static const struct file_operations debug_fake_inject_fops = {
+ .open = simple_open,
+ .write = edac_fake_inject_write,
+ .llseek = generic_file_llseek,
+};
+
+void __init edac_debugfs_init(void)
+{
+ edac_debugfs = debugfs_create_dir("edac", NULL);
+}
+
+void edac_debugfs_exit(void)
+{
+ debugfs_remove_recursive(edac_debugfs);
+}
+
+void edac_create_debugfs_nodes(struct mem_ctl_info *mci)
+{
+ struct dentry *parent;
+ char name[80];
+ int i;
+
+ parent = debugfs_create_dir(mci->dev.kobj.name, edac_debugfs);
+
+ for (i = 0; i < mci->n_layers; i++) {
+ sprintf(name, "fake_inject_%s",
+ edac_layer_name[mci->layers[i].type]);
+ debugfs_create_u8(name, S_IRUGO | S_IWUSR, parent,
+ &mci->fake_inject_layer[i]);
+ }
+
+ debugfs_create_bool("fake_inject_ue", S_IRUGO | S_IWUSR, parent,
+ &mci->fake_inject_ue);
+
+ debugfs_create_u16("fake_inject_count", S_IRUGO | S_IWUSR, parent,
+ &mci->fake_inject_count);
+
+ debugfs_create_file("fake_inject", S_IWUSR, parent, &mci->dev,
+ &debug_fake_inject_fops);
+
+ mci->debugfs = parent;
+}
+
+/* Create a toplevel dir under EDAC's debugfs hierarchy */
+struct dentry *edac_debugfs_create_dir(const char *dirname)
+{
+ if (!edac_debugfs)
+ return NULL;
+
+ return debugfs_create_dir(dirname, edac_debugfs);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_dir);
+
+/* Create a toplevel dir under EDAC's debugfs hierarchy with parent @parent */
+struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent)
+{
+ return debugfs_create_dir(dirname, parent);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_dir_at);
+
+/*
+ * Create a file under EDAC's hierarchy or a sub-hierarchy:
+ *
+ * @name: file name
+ * @mode: file permissions
+ * @parent: parent dentry. If NULL, it becomes the toplevel EDAC dir
+ * @data: private data of caller
+ * @fops: file operations of this file
+ */
+struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+ void *data, const struct file_operations *fops)
+{
+ if (!parent)
+ parent = edac_debugfs;
+
+ return debugfs_create_file(name, mode, parent, data, fops);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_file);
+
+/* Wrapper for debugfs_create_x8() */
+void edac_debugfs_create_x8(const char *name, umode_t mode,
+ struct dentry *parent, u8 *value)
+{
+ if (!parent)
+ parent = edac_debugfs;
+
+ debugfs_create_x8(name, mode, parent, value);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_x8);
+
+/* Wrapper for debugfs_create_x16() */
+void edac_debugfs_create_x16(const char *name, umode_t mode,
+ struct dentry *parent, u16 *value)
+{
+ if (!parent)
+ parent = edac_debugfs;
+
+ debugfs_create_x16(name, mode, parent, value);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_x16);
+
+/* Wrapper for debugfs_create_x32() */
+void edac_debugfs_create_x32(const char *name, umode_t mode,
+ struct dentry *parent, u32 *value)
+{
+ if (!parent)
+ parent = edac_debugfs;
+
+ debugfs_create_x32(name, mode, parent, value);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_x32);
diff --git a/drivers/edac/dmc520_edac.c b/drivers/edac/dmc520_edac.c
new file mode 100644
index 0000000000..1fa5ca57e9
--- /dev/null
+++ b/drivers/edac/dmc520_edac.c
@@ -0,0 +1,656 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * EDAC driver for DMC-520 memory controller.
+ *
+ * The driver supports 10 interrupt lines,
+ * though only dram_ecc_errc and dram_ecc_errd are currently handled.
+ *
+ * Authors: Rui Zhao <ruizhao@microsoft.com>
+ * Lei Wang <lewan@microsoft.com>
+ * Shiping Ji <shji@microsoft.com>
+ */
+
+#include <linux/bitfield.h>
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include "edac_mc.h"
+
+/* DMC-520 registers */
+#define REG_OFFSET_FEATURE_CONFIG 0x130
+#define REG_OFFSET_ECC_ERRC_COUNT_31_00 0x158
+#define REG_OFFSET_ECC_ERRC_COUNT_63_32 0x15C
+#define REG_OFFSET_ECC_ERRD_COUNT_31_00 0x160
+#define REG_OFFSET_ECC_ERRD_COUNT_63_32 0x164
+#define REG_OFFSET_INTERRUPT_CONTROL 0x500
+#define REG_OFFSET_INTERRUPT_CLR 0x508
+#define REG_OFFSET_INTERRUPT_STATUS 0x510
+#define REG_OFFSET_DRAM_ECC_ERRC_INT_INFO_31_00 0x528
+#define REG_OFFSET_DRAM_ECC_ERRC_INT_INFO_63_32 0x52C
+#define REG_OFFSET_DRAM_ECC_ERRD_INT_INFO_31_00 0x530
+#define REG_OFFSET_DRAM_ECC_ERRD_INT_INFO_63_32 0x534
+#define REG_OFFSET_ADDRESS_CONTROL_NOW 0x1010
+#define REG_OFFSET_MEMORY_TYPE_NOW 0x1128
+#define REG_OFFSET_SCRUB_CONTROL0_NOW 0x1170
+#define REG_OFFSET_FORMAT_CONTROL 0x18
+
+/* DMC-520 types, masks and bitfields */
+#define RAM_ECC_INT_CE_BIT BIT(0)
+#define RAM_ECC_INT_UE_BIT BIT(1)
+#define DRAM_ECC_INT_CE_BIT BIT(2)
+#define DRAM_ECC_INT_UE_BIT BIT(3)
+#define FAILED_ACCESS_INT_BIT BIT(4)
+#define FAILED_PROG_INT_BIT BIT(5)
+#define LINK_ERR_INT_BIT BIT(6)
+#define TEMPERATURE_EVENT_INT_BIT BIT(7)
+#define ARCH_FSM_INT_BIT BIT(8)
+#define PHY_REQUEST_INT_BIT BIT(9)
+#define MEMORY_WIDTH_MASK GENMASK(1, 0)
+#define SCRUB_TRIGGER0_NEXT_MASK GENMASK(1, 0)
+#define REG_FIELD_DRAM_ECC_ENABLED GENMASK(1, 0)
+#define REG_FIELD_MEMORY_TYPE GENMASK(2, 0)
+#define REG_FIELD_DEVICE_WIDTH GENMASK(9, 8)
+#define REG_FIELD_ADDRESS_CONTROL_COL GENMASK(2, 0)
+#define REG_FIELD_ADDRESS_CONTROL_ROW GENMASK(10, 8)
+#define REG_FIELD_ADDRESS_CONTROL_BANK GENMASK(18, 16)
+#define REG_FIELD_ADDRESS_CONTROL_RANK GENMASK(25, 24)
+#define REG_FIELD_ERR_INFO_LOW_VALID BIT(0)
+#define REG_FIELD_ERR_INFO_LOW_COL GENMASK(10, 1)
+#define REG_FIELD_ERR_INFO_LOW_ROW GENMASK(28, 11)
+#define REG_FIELD_ERR_INFO_LOW_RANK GENMASK(31, 29)
+#define REG_FIELD_ERR_INFO_HIGH_BANK GENMASK(3, 0)
+#define REG_FIELD_ERR_INFO_HIGH_VALID BIT(31)
+
+#define DRAM_ADDRESS_CONTROL_MIN_COL_BITS 8
+#define DRAM_ADDRESS_CONTROL_MIN_ROW_BITS 11
+
+#define DMC520_SCRUB_TRIGGER_ERR_DETECT 2
+#define DMC520_SCRUB_TRIGGER_IDLE 3
+
+/* Driver settings */
+/*
+ * The max-length message would be: "rank:7 bank:15 row:262143 col:1023".
+ * Max length is 34. Using a 40-size buffer is enough.
+ */
+#define DMC520_MSG_BUF_SIZE 40
+#define EDAC_MOD_NAME "dmc520-edac"
+#define EDAC_CTL_NAME "dmc520"
+
+/* the data bus width for the attached memory chips. */
+enum dmc520_mem_width {
+ MEM_WIDTH_X32 = 2,
+ MEM_WIDTH_X64 = 3
+};
+
+/* memory type */
+enum dmc520_mem_type {
+ MEM_TYPE_DDR3 = 1,
+ MEM_TYPE_DDR4 = 2
+};
+
+/* memory device width */
+enum dmc520_dev_width {
+ DEV_WIDTH_X4 = 0,
+ DEV_WIDTH_X8 = 1,
+ DEV_WIDTH_X16 = 2
+};
+
+struct ecc_error_info {
+ u32 col;
+ u32 row;
+ u32 bank;
+ u32 rank;
+};
+
+/* The interrupt config */
+struct dmc520_irq_config {
+ char *name;
+ int mask;
+};
+
+/* The interrupt mappings */
+static struct dmc520_irq_config dmc520_irq_configs[] = {
+ {
+ .name = "ram_ecc_errc",
+ .mask = RAM_ECC_INT_CE_BIT
+ },
+ {
+ .name = "ram_ecc_errd",
+ .mask = RAM_ECC_INT_UE_BIT
+ },
+ {
+ .name = "dram_ecc_errc",
+ .mask = DRAM_ECC_INT_CE_BIT
+ },
+ {
+ .name = "dram_ecc_errd",
+ .mask = DRAM_ECC_INT_UE_BIT
+ },
+ {
+ .name = "failed_access",
+ .mask = FAILED_ACCESS_INT_BIT
+ },
+ {
+ .name = "failed_prog",
+ .mask = FAILED_PROG_INT_BIT
+ },
+ {
+ .name = "link_err",
+ .mask = LINK_ERR_INT_BIT
+ },
+ {
+ .name = "temperature_event",
+ .mask = TEMPERATURE_EVENT_INT_BIT
+ },
+ {
+ .name = "arch_fsm",
+ .mask = ARCH_FSM_INT_BIT
+ },
+ {
+ .name = "phy_request",
+ .mask = PHY_REQUEST_INT_BIT
+ }
+};
+
+#define NUMBER_OF_IRQS ARRAY_SIZE(dmc520_irq_configs)
+
+/*
+ * The EDAC driver private data.
+ * error_lock is to protect concurrent writes to the mci->error_desc through
+ * edac_mc_handle_error().
+ */
+struct dmc520_edac {
+ void __iomem *reg_base;
+ spinlock_t error_lock;
+ u32 mem_width_in_bytes;
+ int irqs[NUMBER_OF_IRQS];
+ int masks[NUMBER_OF_IRQS];
+};
+
+static int dmc520_mc_idx;
+
+static u32 dmc520_read_reg(struct dmc520_edac *pvt, u32 offset)
+{
+ return readl(pvt->reg_base + offset);
+}
+
+static void dmc520_write_reg(struct dmc520_edac *pvt, u32 val, u32 offset)
+{
+ writel(val, pvt->reg_base + offset);
+}
+
+static u32 dmc520_calc_dram_ecc_error(u32 value)
+{
+ u32 total = 0;
+
+ /* Each rank's error counter takes one byte. */
+ while (value > 0) {
+ total += (value & 0xFF);
+ value >>= 8;
+ }
+ return total;
+}
+
+static u32 dmc520_get_dram_ecc_error_count(struct dmc520_edac *pvt,
+ bool is_ce)
+{
+ u32 reg_offset_low, reg_offset_high;
+ u32 err_low, err_high;
+ u32 err_count;
+
+ reg_offset_low = is_ce ? REG_OFFSET_ECC_ERRC_COUNT_31_00 :
+ REG_OFFSET_ECC_ERRD_COUNT_31_00;
+ reg_offset_high = is_ce ? REG_OFFSET_ECC_ERRC_COUNT_63_32 :
+ REG_OFFSET_ECC_ERRD_COUNT_63_32;
+
+ err_low = dmc520_read_reg(pvt, reg_offset_low);
+ err_high = dmc520_read_reg(pvt, reg_offset_high);
+ /* Reset error counters */
+ dmc520_write_reg(pvt, 0, reg_offset_low);
+ dmc520_write_reg(pvt, 0, reg_offset_high);
+
+ err_count = dmc520_calc_dram_ecc_error(err_low) +
+ dmc520_calc_dram_ecc_error(err_high);
+
+ return err_count;
+}
+
+static void dmc520_get_dram_ecc_error_info(struct dmc520_edac *pvt,
+ bool is_ce,
+ struct ecc_error_info *info)
+{
+ u32 reg_offset_low, reg_offset_high;
+ u32 reg_val_low, reg_val_high;
+ bool valid;
+
+ reg_offset_low = is_ce ? REG_OFFSET_DRAM_ECC_ERRC_INT_INFO_31_00 :
+ REG_OFFSET_DRAM_ECC_ERRD_INT_INFO_31_00;
+ reg_offset_high = is_ce ? REG_OFFSET_DRAM_ECC_ERRC_INT_INFO_63_32 :
+ REG_OFFSET_DRAM_ECC_ERRD_INT_INFO_63_32;
+
+ reg_val_low = dmc520_read_reg(pvt, reg_offset_low);
+ reg_val_high = dmc520_read_reg(pvt, reg_offset_high);
+
+ valid = (FIELD_GET(REG_FIELD_ERR_INFO_LOW_VALID, reg_val_low) != 0) &&
+ (FIELD_GET(REG_FIELD_ERR_INFO_HIGH_VALID, reg_val_high) != 0);
+
+ if (valid) {
+ info->col = FIELD_GET(REG_FIELD_ERR_INFO_LOW_COL, reg_val_low);
+ info->row = FIELD_GET(REG_FIELD_ERR_INFO_LOW_ROW, reg_val_low);
+ info->rank = FIELD_GET(REG_FIELD_ERR_INFO_LOW_RANK, reg_val_low);
+ info->bank = FIELD_GET(REG_FIELD_ERR_INFO_HIGH_BANK, reg_val_high);
+ } else {
+ memset(info, 0, sizeof(*info));
+ }
+}
+
+static bool dmc520_is_ecc_enabled(void __iomem *reg_base)
+{
+ u32 reg_val = readl(reg_base + REG_OFFSET_FEATURE_CONFIG);
+
+ return FIELD_GET(REG_FIELD_DRAM_ECC_ENABLED, reg_val);
+}
+
+static enum scrub_type dmc520_get_scrub_type(struct dmc520_edac *pvt)
+{
+ enum scrub_type type = SCRUB_NONE;
+ u32 reg_val, scrub_cfg;
+
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_SCRUB_CONTROL0_NOW);
+ scrub_cfg = FIELD_GET(SCRUB_TRIGGER0_NEXT_MASK, reg_val);
+
+ if (scrub_cfg == DMC520_SCRUB_TRIGGER_ERR_DETECT ||
+ scrub_cfg == DMC520_SCRUB_TRIGGER_IDLE)
+ type = SCRUB_HW_PROG;
+
+ return type;
+}
+
+/* Get the memory data bus width, in number of bytes. */
+static u32 dmc520_get_memory_width(struct dmc520_edac *pvt)
+{
+ enum dmc520_mem_width mem_width_field;
+ u32 mem_width_in_bytes = 0;
+ u32 reg_val;
+
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_FORMAT_CONTROL);
+ mem_width_field = FIELD_GET(MEMORY_WIDTH_MASK, reg_val);
+
+ if (mem_width_field == MEM_WIDTH_X32)
+ mem_width_in_bytes = 4;
+ else if (mem_width_field == MEM_WIDTH_X64)
+ mem_width_in_bytes = 8;
+ return mem_width_in_bytes;
+}
+
+static enum mem_type dmc520_get_mtype(struct dmc520_edac *pvt)
+{
+ enum mem_type mt = MEM_UNKNOWN;
+ enum dmc520_mem_type type;
+ u32 reg_val;
+
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_MEMORY_TYPE_NOW);
+ type = FIELD_GET(REG_FIELD_MEMORY_TYPE, reg_val);
+
+ switch (type) {
+ case MEM_TYPE_DDR3:
+ mt = MEM_DDR3;
+ break;
+
+ case MEM_TYPE_DDR4:
+ mt = MEM_DDR4;
+ break;
+ }
+
+ return mt;
+}
+
+static enum dev_type dmc520_get_dtype(struct dmc520_edac *pvt)
+{
+ enum dmc520_dev_width device_width;
+ enum dev_type dt = DEV_UNKNOWN;
+ u32 reg_val;
+
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_MEMORY_TYPE_NOW);
+ device_width = FIELD_GET(REG_FIELD_DEVICE_WIDTH, reg_val);
+
+ switch (device_width) {
+ case DEV_WIDTH_X4:
+ dt = DEV_X4;
+ break;
+
+ case DEV_WIDTH_X8:
+ dt = DEV_X8;
+ break;
+
+ case DEV_WIDTH_X16:
+ dt = DEV_X16;
+ break;
+ }
+
+ return dt;
+}
+
+static u32 dmc520_get_rank_count(void __iomem *reg_base)
+{
+ u32 reg_val, rank_bits;
+
+ reg_val = readl(reg_base + REG_OFFSET_ADDRESS_CONTROL_NOW);
+ rank_bits = FIELD_GET(REG_FIELD_ADDRESS_CONTROL_RANK, reg_val);
+
+ return BIT(rank_bits);
+}
+
+static u64 dmc520_get_rank_size(struct dmc520_edac *pvt)
+{
+ u32 reg_val, col_bits, row_bits, bank_bits;
+
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_ADDRESS_CONTROL_NOW);
+
+ col_bits = FIELD_GET(REG_FIELD_ADDRESS_CONTROL_COL, reg_val) +
+ DRAM_ADDRESS_CONTROL_MIN_COL_BITS;
+ row_bits = FIELD_GET(REG_FIELD_ADDRESS_CONTROL_ROW, reg_val) +
+ DRAM_ADDRESS_CONTROL_MIN_ROW_BITS;
+ bank_bits = FIELD_GET(REG_FIELD_ADDRESS_CONTROL_BANK, reg_val);
+
+ return (u64)pvt->mem_width_in_bytes << (col_bits + row_bits + bank_bits);
+}
+
+static void dmc520_handle_dram_ecc_errors(struct mem_ctl_info *mci,
+ bool is_ce)
+{
+ struct dmc520_edac *pvt = mci->pvt_info;
+ char message[DMC520_MSG_BUF_SIZE];
+ struct ecc_error_info info;
+ u32 cnt;
+
+ dmc520_get_dram_ecc_error_info(pvt, is_ce, &info);
+
+ cnt = dmc520_get_dram_ecc_error_count(pvt, is_ce);
+ if (!cnt)
+ return;
+
+ snprintf(message, ARRAY_SIZE(message),
+ "rank:%d bank:%d row:%d col:%d",
+ info.rank, info.bank,
+ info.row, info.col);
+
+ spin_lock(&pvt->error_lock);
+ edac_mc_handle_error((is_ce ? HW_EVENT_ERR_CORRECTED :
+ HW_EVENT_ERR_UNCORRECTED),
+ mci, cnt, 0, 0, 0, info.rank, -1, -1,
+ message, "");
+ spin_unlock(&pvt->error_lock);
+}
+
+static irqreturn_t dmc520_edac_dram_ecc_isr(int irq, struct mem_ctl_info *mci,
+ bool is_ce)
+{
+ struct dmc520_edac *pvt = mci->pvt_info;
+ u32 i_mask;
+
+ i_mask = is_ce ? DRAM_ECC_INT_CE_BIT : DRAM_ECC_INT_UE_BIT;
+
+ dmc520_handle_dram_ecc_errors(mci, is_ce);
+
+ dmc520_write_reg(pvt, i_mask, REG_OFFSET_INTERRUPT_CLR);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t dmc520_edac_dram_all_isr(int irq, struct mem_ctl_info *mci,
+ u32 irq_mask)
+{
+ struct dmc520_edac *pvt = mci->pvt_info;
+ irqreturn_t irq_ret = IRQ_NONE;
+ u32 status;
+
+ status = dmc520_read_reg(pvt, REG_OFFSET_INTERRUPT_STATUS);
+
+ if ((irq_mask & DRAM_ECC_INT_CE_BIT) &&
+ (status & DRAM_ECC_INT_CE_BIT))
+ irq_ret = dmc520_edac_dram_ecc_isr(irq, mci, true);
+
+ if ((irq_mask & DRAM_ECC_INT_UE_BIT) &&
+ (status & DRAM_ECC_INT_UE_BIT))
+ irq_ret = dmc520_edac_dram_ecc_isr(irq, mci, false);
+
+ return irq_ret;
+}
+
+static irqreturn_t dmc520_isr(int irq, void *data)
+{
+ struct mem_ctl_info *mci = data;
+ struct dmc520_edac *pvt = mci->pvt_info;
+ u32 mask = 0;
+ int idx;
+
+ for (idx = 0; idx < NUMBER_OF_IRQS; idx++) {
+ if (pvt->irqs[idx] == irq) {
+ mask = pvt->masks[idx];
+ break;
+ }
+ }
+ return dmc520_edac_dram_all_isr(irq, mci, mask);
+}
+
+static void dmc520_init_csrow(struct mem_ctl_info *mci)
+{
+ struct dmc520_edac *pvt = mci->pvt_info;
+ struct csrow_info *csi;
+ struct dimm_info *dimm;
+ u32 pages_per_rank;
+ enum dev_type dt;
+ enum mem_type mt;
+ int row, ch;
+ u64 rs;
+
+ dt = dmc520_get_dtype(pvt);
+ mt = dmc520_get_mtype(pvt);
+ rs = dmc520_get_rank_size(pvt);
+ pages_per_rank = rs >> PAGE_SHIFT;
+
+ for (row = 0; row < mci->nr_csrows; row++) {
+ csi = mci->csrows[row];
+
+ for (ch = 0; ch < csi->nr_channels; ch++) {
+ dimm = csi->channels[ch]->dimm;
+ dimm->grain = pvt->mem_width_in_bytes;
+ dimm->dtype = dt;
+ dimm->mtype = mt;
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->nr_pages = pages_per_rank / csi->nr_channels;
+ }
+ }
+}
+
+static int dmc520_edac_probe(struct platform_device *pdev)
+{
+ bool registered[NUMBER_OF_IRQS] = { false };
+ int irqs[NUMBER_OF_IRQS] = { -ENXIO };
+ int masks[NUMBER_OF_IRQS] = { 0 };
+ struct edac_mc_layer layers[1];
+ struct dmc520_edac *pvt = NULL;
+ struct mem_ctl_info *mci;
+ void __iomem *reg_base;
+ u32 irq_mask_all = 0;
+ struct resource *res;
+ struct device *dev;
+ int ret, idx, irq;
+ u32 reg_val;
+
+ /* Parse the device node */
+ dev = &pdev->dev;
+
+ for (idx = 0; idx < NUMBER_OF_IRQS; idx++) {
+ irq = platform_get_irq_byname_optional(pdev, dmc520_irq_configs[idx].name);
+ irqs[idx] = irq;
+ masks[idx] = dmc520_irq_configs[idx].mask;
+ if (irq >= 0) {
+ irq_mask_all |= dmc520_irq_configs[idx].mask;
+ edac_dbg(0, "Discovered %s, irq: %d.\n", dmc520_irq_configs[idx].name, irq);
+ }
+ }
+
+ if (!irq_mask_all) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME,
+ "At least one valid interrupt line is expected.\n");
+ return -EINVAL;
+ }
+
+ /* Initialize dmc520 edac */
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ reg_base = devm_ioremap_resource(dev, res);
+ if (IS_ERR(reg_base))
+ return PTR_ERR(reg_base);
+
+ if (!dmc520_is_ecc_enabled(reg_base))
+ return -ENXIO;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = dmc520_get_rank_count(reg_base);
+ layers[0].is_virt_csrow = true;
+
+ mci = edac_mc_alloc(dmc520_mc_idx++, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (!mci) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME,
+ "Failed to allocate memory for mc instance\n");
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ pvt = mci->pvt_info;
+
+ pvt->reg_base = reg_base;
+ spin_lock_init(&pvt->error_lock);
+ memcpy(pvt->irqs, irqs, sizeof(irqs));
+ memcpy(pvt->masks, masks, sizeof(masks));
+
+ platform_set_drvdata(pdev, mci);
+
+ mci->pdev = dev;
+ mci->mtype_cap = MEM_FLAG_DDR3 | MEM_FLAG_DDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ mci->scrub_mode = dmc520_get_scrub_type(pvt);
+ mci->ctl_name = EDAC_CTL_NAME;
+ mci->dev_name = dev_name(mci->pdev);
+ mci->mod_name = EDAC_MOD_NAME;
+
+ edac_op_state = EDAC_OPSTATE_INT;
+
+ pvt->mem_width_in_bytes = dmc520_get_memory_width(pvt);
+
+ dmc520_init_csrow(mci);
+
+ /* Clear interrupts, not affecting other unrelated interrupts */
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_INTERRUPT_CONTROL);
+ dmc520_write_reg(pvt, reg_val & (~irq_mask_all),
+ REG_OFFSET_INTERRUPT_CONTROL);
+ dmc520_write_reg(pvt, irq_mask_all, REG_OFFSET_INTERRUPT_CLR);
+
+ for (idx = 0; idx < NUMBER_OF_IRQS; idx++) {
+ irq = irqs[idx];
+ if (irq >= 0) {
+ ret = devm_request_irq(&pdev->dev, irq,
+ dmc520_isr, IRQF_SHARED,
+ dev_name(&pdev->dev), mci);
+ if (ret < 0) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Failed to request irq %d\n", irq);
+ goto err;
+ }
+ registered[idx] = true;
+ }
+ }
+
+ /* Reset DRAM CE/UE counters */
+ if (irq_mask_all & DRAM_ECC_INT_CE_BIT)
+ dmc520_get_dram_ecc_error_count(pvt, true);
+
+ if (irq_mask_all & DRAM_ECC_INT_UE_BIT)
+ dmc520_get_dram_ecc_error_count(pvt, false);
+
+ ret = edac_mc_add_mc(mci);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME,
+ "Failed to register with EDAC core\n");
+ goto err;
+ }
+
+ /* Enable interrupts, not affecting other unrelated interrupts */
+ dmc520_write_reg(pvt, reg_val | irq_mask_all,
+ REG_OFFSET_INTERRUPT_CONTROL);
+
+ return 0;
+
+err:
+ for (idx = 0; idx < NUMBER_OF_IRQS; idx++) {
+ if (registered[idx])
+ devm_free_irq(&pdev->dev, pvt->irqs[idx], mci);
+ }
+ if (mci)
+ edac_mc_free(mci);
+
+ return ret;
+}
+
+static int dmc520_edac_remove(struct platform_device *pdev)
+{
+ u32 reg_val, idx, irq_mask_all = 0;
+ struct mem_ctl_info *mci;
+ struct dmc520_edac *pvt;
+
+ mci = platform_get_drvdata(pdev);
+ pvt = mci->pvt_info;
+
+ /* Disable interrupts */
+ reg_val = dmc520_read_reg(pvt, REG_OFFSET_INTERRUPT_CONTROL);
+ dmc520_write_reg(pvt, reg_val & (~irq_mask_all),
+ REG_OFFSET_INTERRUPT_CONTROL);
+
+ /* free irq's */
+ for (idx = 0; idx < NUMBER_OF_IRQS; idx++) {
+ if (pvt->irqs[idx] >= 0) {
+ irq_mask_all |= pvt->masks[idx];
+ devm_free_irq(&pdev->dev, pvt->irqs[idx], mci);
+ }
+ }
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+static const struct of_device_id dmc520_edac_driver_id[] = {
+ { .compatible = "arm,dmc-520", },
+ { /* end of table */ }
+};
+
+MODULE_DEVICE_TABLE(of, dmc520_edac_driver_id);
+
+static struct platform_driver dmc520_edac_driver = {
+ .driver = {
+ .name = "dmc520",
+ .of_match_table = dmc520_edac_driver_id,
+ },
+
+ .probe = dmc520_edac_probe,
+ .remove = dmc520_edac_remove
+};
+
+module_platform_driver(dmc520_edac_driver);
+
+MODULE_AUTHOR("Rui Zhao <ruizhao@microsoft.com>");
+MODULE_AUTHOR("Lei Wang <lewan@microsoft.com>");
+MODULE_AUTHOR("Shiping Ji <shji@microsoft.com>");
+MODULE_DESCRIPTION("DMC-520 ECC driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/e752x_edac.c b/drivers/edac/e752x_edac.c
new file mode 100644
index 0000000000..7221b4bb6d
--- /dev/null
+++ b/drivers/edac/e752x_edac.c
@@ -0,0 +1,1481 @@
+/*
+ * Intel e752x Memory Controller kernel module
+ * (C) 2004 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Implement support for the e7520, E7525, e7320 and i3100 memory controllers.
+ *
+ * Datasheets:
+ * https://www.intel.in/content/www/in/en/chipsets/e7525-memory-controller-hub-datasheet.html
+ * ftp://download.intel.com/design/intarch/datashts/31345803.pdf
+ *
+ * Written by Tom Zimmerman
+ *
+ * Contributors:
+ * Thayne Harbaugh at realmsys.com (?)
+ * Wang Zhenyu at intel.com
+ * Dave Jiang at mvista.com
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "e752x_edac"
+
+static int report_non_memory_errors;
+static int force_function_unhide;
+static int sysbus_parity = -1;
+
+static struct edac_pci_ctl_info *e752x_pci;
+
+#define e752x_printk(level, fmt, arg...) \
+ edac_printk(level, "e752x", fmt, ##arg)
+
+#define e752x_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "e752x", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_7520_0
+#define PCI_DEVICE_ID_INTEL_7520_0 0x3590
+#endif /* PCI_DEVICE_ID_INTEL_7520_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7520_1_ERR
+#define PCI_DEVICE_ID_INTEL_7520_1_ERR 0x3591
+#endif /* PCI_DEVICE_ID_INTEL_7520_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_7525_0
+#define PCI_DEVICE_ID_INTEL_7525_0 0x359E
+#endif /* PCI_DEVICE_ID_INTEL_7525_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7525_1_ERR
+#define PCI_DEVICE_ID_INTEL_7525_1_ERR 0x3593
+#endif /* PCI_DEVICE_ID_INTEL_7525_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_7320_0
+#define PCI_DEVICE_ID_INTEL_7320_0 0x3592
+#endif /* PCI_DEVICE_ID_INTEL_7320_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7320_1_ERR
+#define PCI_DEVICE_ID_INTEL_7320_1_ERR 0x3593
+#endif /* PCI_DEVICE_ID_INTEL_7320_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_3100_0
+#define PCI_DEVICE_ID_INTEL_3100_0 0x35B0
+#endif /* PCI_DEVICE_ID_INTEL_3100_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_3100_1_ERR
+#define PCI_DEVICE_ID_INTEL_3100_1_ERR 0x35B1
+#endif /* PCI_DEVICE_ID_INTEL_3100_1_ERR */
+
+#define E752X_NR_CSROWS 8 /* number of csrows */
+
+/* E752X register addresses - device 0 function 0 */
+#define E752X_MCHSCRB 0x52 /* Memory Scrub register (16b) */
+ /*
+ * 6:5 Scrub Completion Count
+ * 3:2 Scrub Rate (i3100 only)
+ * 01=fast 10=normal
+ * 1:0 Scrub Mode enable
+ * 00=off 10=on
+ */
+#define E752X_DRB 0x60 /* DRAM row boundary register (8b) */
+#define E752X_DRA 0x70 /* DRAM row attribute register (8b) */
+ /*
+ * 31:30 Device width row 7
+ * 01=x8 10=x4 11=x8 DDR2
+ * 27:26 Device width row 6
+ * 23:22 Device width row 5
+ * 19:20 Device width row 4
+ * 15:14 Device width row 3
+ * 11:10 Device width row 2
+ * 7:6 Device width row 1
+ * 3:2 Device width row 0
+ */
+#define E752X_DRC 0x7C /* DRAM controller mode reg (32b) */
+ /* FIXME:IS THIS RIGHT? */
+ /*
+ * 22 Number channels 0=1,1=2
+ * 19:18 DRB Granularity 32/64MB
+ */
+#define E752X_DRM 0x80 /* Dimm mapping register */
+#define E752X_DDRCSR 0x9A /* DDR control and status reg (16b) */
+ /*
+ * 14:12 1 single A, 2 single B, 3 dual
+ */
+#define E752X_TOLM 0xC4 /* DRAM top of low memory reg (16b) */
+#define E752X_REMAPBASE 0xC6 /* DRAM remap base address reg (16b) */
+#define E752X_REMAPLIMIT 0xC8 /* DRAM remap limit address reg (16b) */
+#define E752X_REMAPOFFSET 0xCA /* DRAM remap limit offset reg (16b) */
+
+/* E752X register addresses - device 0 function 1 */
+#define E752X_FERR_GLOBAL 0x40 /* Global first error register (32b) */
+#define E752X_NERR_GLOBAL 0x44 /* Global next error register (32b) */
+#define E752X_HI_FERR 0x50 /* Hub interface first error reg (8b) */
+#define E752X_HI_NERR 0x52 /* Hub interface next error reg (8b) */
+#define E752X_HI_ERRMASK 0x54 /* Hub interface error mask reg (8b) */
+#define E752X_HI_SMICMD 0x5A /* Hub interface SMI command reg (8b) */
+#define E752X_SYSBUS_FERR 0x60 /* System buss first error reg (16b) */
+#define E752X_SYSBUS_NERR 0x62 /* System buss next error reg (16b) */
+#define E752X_SYSBUS_ERRMASK 0x64 /* System buss error mask reg (16b) */
+#define E752X_SYSBUS_SMICMD 0x6A /* System buss SMI command reg (16b) */
+#define E752X_BUF_FERR 0x70 /* Memory buffer first error reg (8b) */
+#define E752X_BUF_NERR 0x72 /* Memory buffer next error reg (8b) */
+#define E752X_BUF_ERRMASK 0x74 /* Memory buffer error mask reg (8b) */
+#define E752X_BUF_SMICMD 0x7A /* Memory buffer SMI cmd reg (8b) */
+#define E752X_DRAM_FERR 0x80 /* DRAM first error register (16b) */
+#define E752X_DRAM_NERR 0x82 /* DRAM next error register (16b) */
+#define E752X_DRAM_ERRMASK 0x84 /* DRAM error mask register (8b) */
+#define E752X_DRAM_SMICMD 0x8A /* DRAM SMI command register (8b) */
+#define E752X_DRAM_RETR_ADD 0xAC /* DRAM Retry address register (32b) */
+#define E752X_DRAM_SEC1_ADD 0xA0 /* DRAM first correctable memory */
+ /* error address register (32b) */
+ /*
+ * 31 Reserved
+ * 30:2 CE address (64 byte block 34:6
+ * 1 Reserved
+ * 0 HiLoCS
+ */
+#define E752X_DRAM_SEC2_ADD 0xC8 /* DRAM first correctable memory */
+ /* error address register (32b) */
+ /*
+ * 31 Reserved
+ * 30:2 CE address (64 byte block 34:6)
+ * 1 Reserved
+ * 0 HiLoCS
+ */
+#define E752X_DRAM_DED_ADD 0xA4 /* DRAM first uncorrectable memory */
+ /* error address register (32b) */
+ /*
+ * 31 Reserved
+ * 30:2 CE address (64 byte block 34:6)
+ * 1 Reserved
+ * 0 HiLoCS
+ */
+#define E752X_DRAM_SCRB_ADD 0xA8 /* DRAM 1st uncorrectable scrub mem */
+ /* error address register (32b) */
+ /*
+ * 31 Reserved
+ * 30:2 CE address (64 byte block 34:6
+ * 1 Reserved
+ * 0 HiLoCS
+ */
+#define E752X_DRAM_SEC1_SYNDROME 0xC4 /* DRAM first correctable memory */
+ /* error syndrome register (16b) */
+#define E752X_DRAM_SEC2_SYNDROME 0xC6 /* DRAM second correctable memory */
+ /* error syndrome register (16b) */
+#define E752X_DEVPRES1 0xF4 /* Device Present 1 register (8b) */
+
+/* 3100 IMCH specific register addresses - device 0 function 1 */
+#define I3100_NSI_FERR 0x48 /* NSI first error reg (32b) */
+#define I3100_NSI_NERR 0x4C /* NSI next error reg (32b) */
+#define I3100_NSI_SMICMD 0x54 /* NSI SMI command register (32b) */
+#define I3100_NSI_EMASK 0x90 /* NSI error mask register (32b) */
+
+/* ICH5R register addresses - device 30 function 0 */
+#define ICH5R_PCI_STAT 0x06 /* PCI status register (16b) */
+#define ICH5R_PCI_2ND_STAT 0x1E /* PCI status secondary reg (16b) */
+#define ICH5R_PCI_BRIDGE_CTL 0x3E /* PCI bridge control register (16b) */
+
+enum e752x_chips {
+ E7520 = 0,
+ E7525 = 1,
+ E7320 = 2,
+ I3100 = 3
+};
+
+/*
+ * Those chips Support single-rank and dual-rank memories only.
+ *
+ * On e752x chips, the odd rows are present only on dual-rank memories.
+ * Dividing the rank by two will provide the dimm#
+ *
+ * i3100 MC has a different mapping: it supports only 4 ranks.
+ *
+ * The mapping is (from 1 to n):
+ * slot single-ranked double-ranked
+ * dimm #1 -> rank #4 NA
+ * dimm #2 -> rank #3 NA
+ * dimm #3 -> rank #2 Ranks 2 and 3
+ * dimm #4 -> rank $1 Ranks 1 and 4
+ *
+ * FIXME: The current mapping for i3100 considers that it supports up to 8
+ * ranks/chanel, but datasheet says that the MC supports only 4 ranks.
+ */
+
+struct e752x_pvt {
+ struct pci_dev *dev_d0f0;
+ struct pci_dev *dev_d0f1;
+ u32 tolm;
+ u32 remapbase;
+ u32 remaplimit;
+ int mc_symmetric;
+ u8 map[8];
+ int map_type;
+ const struct e752x_dev_info *dev_info;
+};
+
+struct e752x_dev_info {
+ u16 err_dev;
+ u16 ctl_dev;
+ const char *ctl_name;
+};
+
+struct e752x_error_info {
+ u32 ferr_global;
+ u32 nerr_global;
+ u32 nsi_ferr; /* 3100 only */
+ u32 nsi_nerr; /* 3100 only */
+ u8 hi_ferr; /* all but 3100 */
+ u8 hi_nerr; /* all but 3100 */
+ u16 sysbus_ferr;
+ u16 sysbus_nerr;
+ u8 buf_ferr;
+ u8 buf_nerr;
+ u16 dram_ferr;
+ u16 dram_nerr;
+ u32 dram_sec1_add;
+ u32 dram_sec2_add;
+ u16 dram_sec1_syndrome;
+ u16 dram_sec2_syndrome;
+ u32 dram_ded_add;
+ u32 dram_scrb_add;
+ u32 dram_retr_add;
+};
+
+static const struct e752x_dev_info e752x_devs[] = {
+ [E7520] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7520_1_ERR,
+ .ctl_dev = PCI_DEVICE_ID_INTEL_7520_0,
+ .ctl_name = "E7520"},
+ [E7525] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7525_1_ERR,
+ .ctl_dev = PCI_DEVICE_ID_INTEL_7525_0,
+ .ctl_name = "E7525"},
+ [E7320] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7320_1_ERR,
+ .ctl_dev = PCI_DEVICE_ID_INTEL_7320_0,
+ .ctl_name = "E7320"},
+ [I3100] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_3100_1_ERR,
+ .ctl_dev = PCI_DEVICE_ID_INTEL_3100_0,
+ .ctl_name = "3100"},
+};
+
+/* Valid scrub rates for the e752x/3100 hardware memory scrubber. We
+ * map the scrubbing bandwidth to a hardware register value. The 'set'
+ * operation finds the 'matching or higher value'. Note that scrubbing
+ * on the e752x can only be enabled/disabled. The 3100 supports
+ * a normal and fast mode.
+ */
+
+#define SDRATE_EOT 0xFFFFFFFF
+
+struct scrubrate {
+ u32 bandwidth; /* bandwidth consumed by scrubbing in bytes/sec */
+ u16 scrubval; /* register value for scrub rate */
+};
+
+/* Rate below assumes same performance as i3100 using PC3200 DDR2 in
+ * normal mode. e752x bridges don't support choosing normal or fast mode,
+ * so the scrubbing bandwidth value isn't all that important - scrubbing is
+ * either on or off.
+ */
+static const struct scrubrate scrubrates_e752x[] = {
+ {0, 0x00}, /* Scrubbing Off */
+ {500000, 0x02}, /* Scrubbing On */
+ {SDRATE_EOT, 0x00} /* End of Table */
+};
+
+/* Fast mode: 2 GByte PC3200 DDR2 scrubbed in 33s = 63161283 bytes/s
+ * Normal mode: 125 (32000 / 256) times slower than fast mode.
+ */
+static const struct scrubrate scrubrates_i3100[] = {
+ {0, 0x00}, /* Scrubbing Off */
+ {500000, 0x0a}, /* Normal mode - 32k clocks */
+ {62500000, 0x06}, /* Fast mode - 256 clocks */
+ {SDRATE_EOT, 0x00} /* End of Table */
+};
+
+static unsigned long ctl_page_to_phys(struct mem_ctl_info *mci,
+ unsigned long page)
+{
+ u32 remap;
+ struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
+
+ edac_dbg(3, "\n");
+
+ if (page < pvt->tolm)
+ return page;
+
+ if ((page >= 0x100000) && (page < pvt->remapbase))
+ return page;
+
+ remap = (page - pvt->tolm) + pvt->remapbase;
+
+ if (remap < pvt->remaplimit)
+ return remap;
+
+ e752x_printk(KERN_ERR, "Invalid page %lx - out of range\n", page);
+ return pvt->tolm - 1;
+}
+
+static void do_process_ce(struct mem_ctl_info *mci, u16 error_one,
+ u32 sec1_add, u16 sec1_syndrome)
+{
+ u32 page;
+ int row;
+ int channel;
+ int i;
+ struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
+
+ edac_dbg(3, "\n");
+
+ /* convert the addr to 4k page */
+ page = sec1_add >> (PAGE_SHIFT - 4);
+
+ /* FIXME - check for -1 */
+ if (pvt->mc_symmetric) {
+ /* chip select are bits 14 & 13 */
+ row = ((page >> 1) & 3);
+ e752x_printk(KERN_WARNING,
+ "Test row %d Table %d %d %d %d %d %d %d %d\n", row,
+ pvt->map[0], pvt->map[1], pvt->map[2], pvt->map[3],
+ pvt->map[4], pvt->map[5], pvt->map[6],
+ pvt->map[7]);
+
+ /* test for channel remapping */
+ for (i = 0; i < 8; i++) {
+ if (pvt->map[i] == row)
+ break;
+ }
+
+ e752x_printk(KERN_WARNING, "Test computed row %d\n", i);
+
+ if (i < 8)
+ row = i;
+ else
+ e752x_mc_printk(mci, KERN_WARNING,
+ "row %d not found in remap table\n",
+ row);
+ } else
+ row = edac_mc_find_csrow_by_page(mci, page);
+
+ /* 0 = channel A, 1 = channel B */
+ channel = !(error_one & 1);
+
+ /* e752x mc reads 34:6 of the DRAM linear address */
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ page, offset_in_page(sec1_add << 4), sec1_syndrome,
+ row, channel, -1,
+ "e752x CE", "");
+}
+
+static inline void process_ce(struct mem_ctl_info *mci, u16 error_one,
+ u32 sec1_add, u16 sec1_syndrome, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_process_ce(mci, error_one, sec1_add, sec1_syndrome);
+}
+
+static void do_process_ue(struct mem_ctl_info *mci, u16 error_one,
+ u32 ded_add, u32 scrb_add)
+{
+ u32 error_2b, block_page;
+ int row;
+ struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
+
+ edac_dbg(3, "\n");
+
+ if (error_one & 0x0202) {
+ error_2b = ded_add;
+
+ /* convert to 4k address */
+ block_page = error_2b >> (PAGE_SHIFT - 4);
+
+ row = pvt->mc_symmetric ?
+ /* chip select are bits 14 & 13 */
+ ((block_page >> 1) & 3) :
+ edac_mc_find_csrow_by_page(mci, block_page);
+
+ /* e752x mc reads 34:6 of the DRAM linear address */
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ block_page,
+ offset_in_page(error_2b << 4), 0,
+ row, -1, -1,
+ "e752x UE from Read", "");
+
+ }
+ if (error_one & 0x0404) {
+ error_2b = scrb_add;
+
+ /* convert to 4k address */
+ block_page = error_2b >> (PAGE_SHIFT - 4);
+
+ row = pvt->mc_symmetric ?
+ /* chip select are bits 14 & 13 */
+ ((block_page >> 1) & 3) :
+ edac_mc_find_csrow_by_page(mci, block_page);
+
+ /* e752x mc reads 34:6 of the DRAM linear address */
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ block_page,
+ offset_in_page(error_2b << 4), 0,
+ row, -1, -1,
+ "e752x UE from Scruber", "");
+ }
+}
+
+static inline void process_ue(struct mem_ctl_info *mci, u16 error_one,
+ u32 ded_add, u32 scrb_add, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_process_ue(mci, error_one, ded_add, scrb_add);
+}
+
+static inline void process_ue_no_info_wr(struct mem_ctl_info *mci,
+ int *error_found, int handle_error)
+{
+ *error_found = 1;
+
+ if (!handle_error)
+ return;
+
+ edac_dbg(3, "\n");
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1,
+ "e752x UE log memory write", "");
+}
+
+static void do_process_ded_retry(struct mem_ctl_info *mci, u16 error,
+ u32 retry_add)
+{
+ u32 error_1b, page;
+ int row;
+ struct e752x_pvt *pvt = (struct e752x_pvt *)mci->pvt_info;
+
+ error_1b = retry_add;
+ page = error_1b >> (PAGE_SHIFT - 4); /* convert the addr to 4k page */
+
+ /* chip select are bits 14 & 13 */
+ row = pvt->mc_symmetric ? ((page >> 1) & 3) :
+ edac_mc_find_csrow_by_page(mci, page);
+
+ e752x_mc_printk(mci, KERN_WARNING,
+ "CE page 0x%lx, row %d : Memory read retry\n",
+ (long unsigned int)page, row);
+}
+
+static inline void process_ded_retry(struct mem_ctl_info *mci, u16 error,
+ u32 retry_add, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_process_ded_retry(mci, error, retry_add);
+}
+
+static inline void process_threshold_ce(struct mem_ctl_info *mci, u16 error,
+ int *error_found, int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ e752x_mc_printk(mci, KERN_WARNING, "Memory threshold CE\n");
+}
+
+static char *global_message[11] = {
+ "PCI Express C1",
+ "PCI Express C",
+ "PCI Express B1",
+ "PCI Express B",
+ "PCI Express A1",
+ "PCI Express A",
+ "DMA Controller",
+ "HUB or NS Interface",
+ "System Bus",
+ "DRAM Controller", /* 9th entry */
+ "Internal Buffer"
+};
+
+#define DRAM_ENTRY 9
+
+static char *fatal_message[2] = { "Non-Fatal ", "Fatal " };
+
+static void do_global_error(int fatal, u32 errors)
+{
+ int i;
+
+ for (i = 0; i < 11; i++) {
+ if (errors & (1 << i)) {
+ /* If the error is from DRAM Controller OR
+ * we are to report ALL errors, then
+ * report the error
+ */
+ if ((i == DRAM_ENTRY) || report_non_memory_errors)
+ e752x_printk(KERN_WARNING, "%sError %s\n",
+ fatal_message[fatal],
+ global_message[i]);
+ }
+ }
+}
+
+static inline void global_error(int fatal, u32 errors, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_global_error(fatal, errors);
+}
+
+static char *hub_message[7] = {
+ "HI Address or Command Parity", "HI Illegal Access",
+ "HI Internal Parity", "Out of Range Access",
+ "HI Data Parity", "Enhanced Config Access",
+ "Hub Interface Target Abort"
+};
+
+static void do_hub_error(int fatal, u8 errors)
+{
+ int i;
+
+ for (i = 0; i < 7; i++) {
+ if (errors & (1 << i))
+ e752x_printk(KERN_WARNING, "%sError %s\n",
+ fatal_message[fatal], hub_message[i]);
+ }
+}
+
+static inline void hub_error(int fatal, u8 errors, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_hub_error(fatal, errors);
+}
+
+#define NSI_FATAL_MASK 0x0c080081
+#define NSI_NON_FATAL_MASK 0x23a0ba64
+#define NSI_ERR_MASK (NSI_FATAL_MASK | NSI_NON_FATAL_MASK)
+
+static char *nsi_message[30] = {
+ "NSI Link Down", /* NSI_FERR/NSI_NERR bit 0, fatal error */
+ "", /* reserved */
+ "NSI Parity Error", /* bit 2, non-fatal */
+ "", /* reserved */
+ "", /* reserved */
+ "Correctable Error Message", /* bit 5, non-fatal */
+ "Non-Fatal Error Message", /* bit 6, non-fatal */
+ "Fatal Error Message", /* bit 7, fatal */
+ "", /* reserved */
+ "Receiver Error", /* bit 9, non-fatal */
+ "", /* reserved */
+ "Bad TLP", /* bit 11, non-fatal */
+ "Bad DLLP", /* bit 12, non-fatal */
+ "REPLAY_NUM Rollover", /* bit 13, non-fatal */
+ "", /* reserved */
+ "Replay Timer Timeout", /* bit 15, non-fatal */
+ "", /* reserved */
+ "", /* reserved */
+ "", /* reserved */
+ "Data Link Protocol Error", /* bit 19, fatal */
+ "", /* reserved */
+ "Poisoned TLP", /* bit 21, non-fatal */
+ "", /* reserved */
+ "Completion Timeout", /* bit 23, non-fatal */
+ "Completer Abort", /* bit 24, non-fatal */
+ "Unexpected Completion", /* bit 25, non-fatal */
+ "Receiver Overflow", /* bit 26, fatal */
+ "Malformed TLP", /* bit 27, fatal */
+ "", /* reserved */
+ "Unsupported Request" /* bit 29, non-fatal */
+};
+
+static void do_nsi_error(int fatal, u32 errors)
+{
+ int i;
+
+ for (i = 0; i < 30; i++) {
+ if (errors & (1 << i))
+ printk(KERN_WARNING "%sError %s\n",
+ fatal_message[fatal], nsi_message[i]);
+ }
+}
+
+static inline void nsi_error(int fatal, u32 errors, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_nsi_error(fatal, errors);
+}
+
+static char *membuf_message[4] = {
+ "Internal PMWB to DRAM parity",
+ "Internal PMWB to System Bus Parity",
+ "Internal System Bus or IO to PMWB Parity",
+ "Internal DRAM to PMWB Parity"
+};
+
+static void do_membuf_error(u8 errors)
+{
+ int i;
+
+ for (i = 0; i < 4; i++) {
+ if (errors & (1 << i))
+ e752x_printk(KERN_WARNING, "Non-Fatal Error %s\n",
+ membuf_message[i]);
+ }
+}
+
+static inline void membuf_error(u8 errors, int *error_found, int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_membuf_error(errors);
+}
+
+static char *sysbus_message[10] = {
+ "Addr or Request Parity",
+ "Data Strobe Glitch",
+ "Addr Strobe Glitch",
+ "Data Parity",
+ "Addr Above TOM",
+ "Non DRAM Lock Error",
+ "MCERR", "BINIT",
+ "Memory Parity",
+ "IO Subsystem Parity"
+};
+
+static void do_sysbus_error(int fatal, u32 errors)
+{
+ int i;
+
+ for (i = 0; i < 10; i++) {
+ if (errors & (1 << i))
+ e752x_printk(KERN_WARNING, "%sError System Bus %s\n",
+ fatal_message[fatal], sysbus_message[i]);
+ }
+}
+
+static inline void sysbus_error(int fatal, u32 errors, int *error_found,
+ int handle_error)
+{
+ *error_found = 1;
+
+ if (handle_error)
+ do_sysbus_error(fatal, errors);
+}
+
+static void e752x_check_hub_interface(struct e752x_error_info *info,
+ int *error_found, int handle_error)
+{
+ u8 stat8;
+
+ //pci_read_config_byte(dev,E752X_HI_FERR,&stat8);
+
+ stat8 = info->hi_ferr;
+
+ if (stat8 & 0x7f) { /* Error, so process */
+ stat8 &= 0x7f;
+
+ if (stat8 & 0x2b)
+ hub_error(1, stat8 & 0x2b, error_found, handle_error);
+
+ if (stat8 & 0x54)
+ hub_error(0, stat8 & 0x54, error_found, handle_error);
+ }
+ //pci_read_config_byte(dev,E752X_HI_NERR,&stat8);
+
+ stat8 = info->hi_nerr;
+
+ if (stat8 & 0x7f) { /* Error, so process */
+ stat8 &= 0x7f;
+
+ if (stat8 & 0x2b)
+ hub_error(1, stat8 & 0x2b, error_found, handle_error);
+
+ if (stat8 & 0x54)
+ hub_error(0, stat8 & 0x54, error_found, handle_error);
+ }
+}
+
+static void e752x_check_ns_interface(struct e752x_error_info *info,
+ int *error_found, int handle_error)
+{
+ u32 stat32;
+
+ stat32 = info->nsi_ferr;
+ if (stat32 & NSI_ERR_MASK) { /* Error, so process */
+ if (stat32 & NSI_FATAL_MASK) /* check for fatal errors */
+ nsi_error(1, stat32 & NSI_FATAL_MASK, error_found,
+ handle_error);
+ if (stat32 & NSI_NON_FATAL_MASK) /* check for non-fatal ones */
+ nsi_error(0, stat32 & NSI_NON_FATAL_MASK, error_found,
+ handle_error);
+ }
+ stat32 = info->nsi_nerr;
+ if (stat32 & NSI_ERR_MASK) {
+ if (stat32 & NSI_FATAL_MASK)
+ nsi_error(1, stat32 & NSI_FATAL_MASK, error_found,
+ handle_error);
+ if (stat32 & NSI_NON_FATAL_MASK)
+ nsi_error(0, stat32 & NSI_NON_FATAL_MASK, error_found,
+ handle_error);
+ }
+}
+
+static void e752x_check_sysbus(struct e752x_error_info *info,
+ int *error_found, int handle_error)
+{
+ u32 stat32, error32;
+
+ //pci_read_config_dword(dev,E752X_SYSBUS_FERR,&stat32);
+ stat32 = info->sysbus_ferr + (info->sysbus_nerr << 16);
+
+ if (stat32 == 0)
+ return; /* no errors */
+
+ error32 = (stat32 >> 16) & 0x3ff;
+ stat32 = stat32 & 0x3ff;
+
+ if (stat32 & 0x087)
+ sysbus_error(1, stat32 & 0x087, error_found, handle_error);
+
+ if (stat32 & 0x378)
+ sysbus_error(0, stat32 & 0x378, error_found, handle_error);
+
+ if (error32 & 0x087)
+ sysbus_error(1, error32 & 0x087, error_found, handle_error);
+
+ if (error32 & 0x378)
+ sysbus_error(0, error32 & 0x378, error_found, handle_error);
+}
+
+static void e752x_check_membuf(struct e752x_error_info *info,
+ int *error_found, int handle_error)
+{
+ u8 stat8;
+
+ stat8 = info->buf_ferr;
+
+ if (stat8 & 0x0f) { /* Error, so process */
+ stat8 &= 0x0f;
+ membuf_error(stat8, error_found, handle_error);
+ }
+
+ stat8 = info->buf_nerr;
+
+ if (stat8 & 0x0f) { /* Error, so process */
+ stat8 &= 0x0f;
+ membuf_error(stat8, error_found, handle_error);
+ }
+}
+
+static void e752x_check_dram(struct mem_ctl_info *mci,
+ struct e752x_error_info *info, int *error_found,
+ int handle_error)
+{
+ u16 error_one, error_next;
+
+ error_one = info->dram_ferr;
+ error_next = info->dram_nerr;
+
+ /* decode and report errors */
+ if (error_one & 0x0101) /* check first error correctable */
+ process_ce(mci, error_one, info->dram_sec1_add,
+ info->dram_sec1_syndrome, error_found, handle_error);
+
+ if (error_next & 0x0101) /* check next error correctable */
+ process_ce(mci, error_next, info->dram_sec2_add,
+ info->dram_sec2_syndrome, error_found, handle_error);
+
+ if (error_one & 0x4040)
+ process_ue_no_info_wr(mci, error_found, handle_error);
+
+ if (error_next & 0x4040)
+ process_ue_no_info_wr(mci, error_found, handle_error);
+
+ if (error_one & 0x2020)
+ process_ded_retry(mci, error_one, info->dram_retr_add,
+ error_found, handle_error);
+
+ if (error_next & 0x2020)
+ process_ded_retry(mci, error_next, info->dram_retr_add,
+ error_found, handle_error);
+
+ if (error_one & 0x0808)
+ process_threshold_ce(mci, error_one, error_found, handle_error);
+
+ if (error_next & 0x0808)
+ process_threshold_ce(mci, error_next, error_found,
+ handle_error);
+
+ if (error_one & 0x0606)
+ process_ue(mci, error_one, info->dram_ded_add,
+ info->dram_scrb_add, error_found, handle_error);
+
+ if (error_next & 0x0606)
+ process_ue(mci, error_next, info->dram_ded_add,
+ info->dram_scrb_add, error_found, handle_error);
+}
+
+static void e752x_get_error_info(struct mem_ctl_info *mci,
+ struct e752x_error_info *info)
+{
+ struct pci_dev *dev;
+ struct e752x_pvt *pvt;
+
+ memset(info, 0, sizeof(*info));
+ pvt = (struct e752x_pvt *)mci->pvt_info;
+ dev = pvt->dev_d0f1;
+ pci_read_config_dword(dev, E752X_FERR_GLOBAL, &info->ferr_global);
+
+ if (info->ferr_global) {
+ if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
+ pci_read_config_dword(dev, I3100_NSI_FERR,
+ &info->nsi_ferr);
+ info->hi_ferr = 0;
+ } else {
+ pci_read_config_byte(dev, E752X_HI_FERR,
+ &info->hi_ferr);
+ info->nsi_ferr = 0;
+ }
+ pci_read_config_word(dev, E752X_SYSBUS_FERR,
+ &info->sysbus_ferr);
+ pci_read_config_byte(dev, E752X_BUF_FERR, &info->buf_ferr);
+ pci_read_config_word(dev, E752X_DRAM_FERR, &info->dram_ferr);
+ pci_read_config_dword(dev, E752X_DRAM_SEC1_ADD,
+ &info->dram_sec1_add);
+ pci_read_config_word(dev, E752X_DRAM_SEC1_SYNDROME,
+ &info->dram_sec1_syndrome);
+ pci_read_config_dword(dev, E752X_DRAM_DED_ADD,
+ &info->dram_ded_add);
+ pci_read_config_dword(dev, E752X_DRAM_SCRB_ADD,
+ &info->dram_scrb_add);
+ pci_read_config_dword(dev, E752X_DRAM_RETR_ADD,
+ &info->dram_retr_add);
+
+ /* ignore the reserved bits just in case */
+ if (info->hi_ferr & 0x7f)
+ pci_write_config_byte(dev, E752X_HI_FERR,
+ info->hi_ferr);
+
+ if (info->nsi_ferr & NSI_ERR_MASK)
+ pci_write_config_dword(dev, I3100_NSI_FERR,
+ info->nsi_ferr);
+
+ if (info->sysbus_ferr)
+ pci_write_config_word(dev, E752X_SYSBUS_FERR,
+ info->sysbus_ferr);
+
+ if (info->buf_ferr & 0x0f)
+ pci_write_config_byte(dev, E752X_BUF_FERR,
+ info->buf_ferr);
+
+ if (info->dram_ferr)
+ pci_write_bits16(pvt->dev_d0f1, E752X_DRAM_FERR,
+ info->dram_ferr, info->dram_ferr);
+
+ pci_write_config_dword(dev, E752X_FERR_GLOBAL,
+ info->ferr_global);
+ }
+
+ pci_read_config_dword(dev, E752X_NERR_GLOBAL, &info->nerr_global);
+
+ if (info->nerr_global) {
+ if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
+ pci_read_config_dword(dev, I3100_NSI_NERR,
+ &info->nsi_nerr);
+ info->hi_nerr = 0;
+ } else {
+ pci_read_config_byte(dev, E752X_HI_NERR,
+ &info->hi_nerr);
+ info->nsi_nerr = 0;
+ }
+ pci_read_config_word(dev, E752X_SYSBUS_NERR,
+ &info->sysbus_nerr);
+ pci_read_config_byte(dev, E752X_BUF_NERR, &info->buf_nerr);
+ pci_read_config_word(dev, E752X_DRAM_NERR, &info->dram_nerr);
+ pci_read_config_dword(dev, E752X_DRAM_SEC2_ADD,
+ &info->dram_sec2_add);
+ pci_read_config_word(dev, E752X_DRAM_SEC2_SYNDROME,
+ &info->dram_sec2_syndrome);
+
+ if (info->hi_nerr & 0x7f)
+ pci_write_config_byte(dev, E752X_HI_NERR,
+ info->hi_nerr);
+
+ if (info->nsi_nerr & NSI_ERR_MASK)
+ pci_write_config_dword(dev, I3100_NSI_NERR,
+ info->nsi_nerr);
+
+ if (info->sysbus_nerr)
+ pci_write_config_word(dev, E752X_SYSBUS_NERR,
+ info->sysbus_nerr);
+
+ if (info->buf_nerr & 0x0f)
+ pci_write_config_byte(dev, E752X_BUF_NERR,
+ info->buf_nerr);
+
+ if (info->dram_nerr)
+ pci_write_bits16(pvt->dev_d0f1, E752X_DRAM_NERR,
+ info->dram_nerr, info->dram_nerr);
+
+ pci_write_config_dword(dev, E752X_NERR_GLOBAL,
+ info->nerr_global);
+ }
+}
+
+static int e752x_process_error_info(struct mem_ctl_info *mci,
+ struct e752x_error_info *info,
+ int handle_errors)
+{
+ u32 error32, stat32;
+ int error_found;
+
+ error_found = 0;
+ error32 = (info->ferr_global >> 18) & 0x3ff;
+ stat32 = (info->ferr_global >> 4) & 0x7ff;
+
+ if (error32)
+ global_error(1, error32, &error_found, handle_errors);
+
+ if (stat32)
+ global_error(0, stat32, &error_found, handle_errors);
+
+ error32 = (info->nerr_global >> 18) & 0x3ff;
+ stat32 = (info->nerr_global >> 4) & 0x7ff;
+
+ if (error32)
+ global_error(1, error32, &error_found, handle_errors);
+
+ if (stat32)
+ global_error(0, stat32, &error_found, handle_errors);
+
+ e752x_check_hub_interface(info, &error_found, handle_errors);
+ e752x_check_ns_interface(info, &error_found, handle_errors);
+ e752x_check_sysbus(info, &error_found, handle_errors);
+ e752x_check_membuf(info, &error_found, handle_errors);
+ e752x_check_dram(mci, info, &error_found, handle_errors);
+ return error_found;
+}
+
+static void e752x_check(struct mem_ctl_info *mci)
+{
+ struct e752x_error_info info;
+
+ e752x_get_error_info(mci, &info);
+ e752x_process_error_info(mci, &info, 1);
+}
+
+/* Program byte/sec bandwidth scrub rate to hardware */
+static int set_sdram_scrub_rate(struct mem_ctl_info *mci, u32 new_bw)
+{
+ const struct scrubrate *scrubrates;
+ struct e752x_pvt *pvt = (struct e752x_pvt *) mci->pvt_info;
+ struct pci_dev *pdev = pvt->dev_d0f0;
+ int i;
+
+ if (pvt->dev_info->ctl_dev == PCI_DEVICE_ID_INTEL_3100_0)
+ scrubrates = scrubrates_i3100;
+ else
+ scrubrates = scrubrates_e752x;
+
+ /* Translate the desired scrub rate to a e752x/3100 register value.
+ * Search for the bandwidth that is equal or greater than the
+ * desired rate and program the cooresponding register value.
+ */
+ for (i = 0; scrubrates[i].bandwidth != SDRATE_EOT; i++)
+ if (scrubrates[i].bandwidth >= new_bw)
+ break;
+
+ if (scrubrates[i].bandwidth == SDRATE_EOT)
+ return -1;
+
+ pci_write_config_word(pdev, E752X_MCHSCRB, scrubrates[i].scrubval);
+
+ return scrubrates[i].bandwidth;
+}
+
+/* Convert current scrub rate value into byte/sec bandwidth */
+static int get_sdram_scrub_rate(struct mem_ctl_info *mci)
+{
+ const struct scrubrate *scrubrates;
+ struct e752x_pvt *pvt = (struct e752x_pvt *) mci->pvt_info;
+ struct pci_dev *pdev = pvt->dev_d0f0;
+ u16 scrubval;
+ int i;
+
+ if (pvt->dev_info->ctl_dev == PCI_DEVICE_ID_INTEL_3100_0)
+ scrubrates = scrubrates_i3100;
+ else
+ scrubrates = scrubrates_e752x;
+
+ /* Find the bandwidth matching the memory scrubber configuration */
+ pci_read_config_word(pdev, E752X_MCHSCRB, &scrubval);
+ scrubval = scrubval & 0x0f;
+
+ for (i = 0; scrubrates[i].bandwidth != SDRATE_EOT; i++)
+ if (scrubrates[i].scrubval == scrubval)
+ break;
+
+ if (scrubrates[i].bandwidth == SDRATE_EOT) {
+ e752x_printk(KERN_WARNING,
+ "Invalid sdram scrub control value: 0x%x\n", scrubval);
+ return -1;
+ }
+ return scrubrates[i].bandwidth;
+
+}
+
+/* Return 1 if dual channel mode is active. Else return 0. */
+static inline int dual_channel_active(u16 ddrcsr)
+{
+ return (((ddrcsr >> 12) & 3) == 3);
+}
+
+/* Remap csrow index numbers if map_type is "reverse"
+ */
+static inline int remap_csrow_index(struct mem_ctl_info *mci, int index)
+{
+ struct e752x_pvt *pvt = mci->pvt_info;
+
+ if (!pvt->map_type)
+ return (7 - index);
+
+ return (index);
+}
+
+static void e752x_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
+ u16 ddrcsr)
+{
+ struct csrow_info *csrow;
+ enum edac_type edac_mode;
+ unsigned long last_cumul_size;
+ int index, mem_dev, drc_chan;
+ int drc_drbg; /* DRB granularity 0=64mb, 1=128mb */
+ int drc_ddim; /* DRAM Data Integrity Mode 0=none, 2=edac */
+ u8 value;
+ u32 dra, drc, cumul_size, i, nr_pages;
+
+ dra = 0;
+ for (index = 0; index < 4; index++) {
+ u8 dra_reg;
+ pci_read_config_byte(pdev, E752X_DRA + index, &dra_reg);
+ dra |= dra_reg << (index * 8);
+ }
+ pci_read_config_dword(pdev, E752X_DRC, &drc);
+ drc_chan = dual_channel_active(ddrcsr) ? 1 : 0;
+ drc_drbg = drc_chan + 1; /* 128 in dual mode, 64 in single */
+ drc_ddim = (drc >> 20) & 0x3;
+
+ /* The dram row boundary (DRB) reg values are boundary address for
+ * each DRAM row with a granularity of 64 or 128MB (single/dual
+ * channel operation). DRB regs are cumulative; therefore DRB7 will
+ * contain the total memory contained in all eight rows.
+ */
+ for (last_cumul_size = index = 0; index < mci->nr_csrows; index++) {
+ /* mem_dev 0=x8, 1=x4 */
+ mem_dev = (dra >> (index * 4 + 2)) & 0x3;
+ csrow = mci->csrows[remap_csrow_index(mci, index)];
+
+ mem_dev = (mem_dev == 2);
+ pci_read_config_byte(pdev, E752X_DRB + index, &value);
+ /* convert a 128 or 64 MiB DRB to a page size. */
+ cumul_size = value << (25 + drc_drbg - PAGE_SHIFT);
+ edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
+ if (cumul_size == last_cumul_size)
+ continue; /* not populated */
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ nr_pages = cumul_size - last_cumul_size;
+ last_cumul_size = cumul_size;
+
+ /*
+ * if single channel or x8 devices then SECDED
+ * if dual channel and x4 then S4ECD4ED
+ */
+ if (drc_ddim) {
+ if (drc_chan && mem_dev) {
+ edac_mode = EDAC_S4ECD4ED;
+ mci->edac_cap |= EDAC_FLAG_S4ECD4ED;
+ } else {
+ edac_mode = EDAC_SECDED;
+ mci->edac_cap |= EDAC_FLAG_SECDED;
+ }
+ } else
+ edac_mode = EDAC_NONE;
+ for (i = 0; i < csrow->nr_channels; i++) {
+ struct dimm_info *dimm = csrow->channels[i]->dimm;
+
+ edac_dbg(3, "Initializing rank at (%i,%i)\n", index, i);
+ dimm->nr_pages = nr_pages / csrow->nr_channels;
+ dimm->grain = 1 << 12; /* 4KiB - resolution of CELOG */
+ dimm->mtype = MEM_RDDR; /* only one type supported */
+ dimm->dtype = mem_dev ? DEV_X4 : DEV_X8;
+ dimm->edac_mode = edac_mode;
+ }
+ }
+}
+
+static void e752x_init_mem_map_table(struct pci_dev *pdev,
+ struct e752x_pvt *pvt)
+{
+ int index;
+ u8 value, last, row;
+
+ last = 0;
+ row = 0;
+
+ for (index = 0; index < 8; index += 2) {
+ pci_read_config_byte(pdev, E752X_DRB + index, &value);
+ /* test if there is a dimm in this slot */
+ if (value == last) {
+ /* no dimm in the slot, so flag it as empty */
+ pvt->map[index] = 0xff;
+ pvt->map[index + 1] = 0xff;
+ } else { /* there is a dimm in the slot */
+ pvt->map[index] = row;
+ row++;
+ last = value;
+ /* test the next value to see if the dimm is double
+ * sided
+ */
+ pci_read_config_byte(pdev, E752X_DRB + index + 1,
+ &value);
+
+ /* the dimm is single sided, so flag as empty */
+ /* this is a double sided dimm to save the next row #*/
+ pvt->map[index + 1] = (value == last) ? 0xff : row;
+ row++;
+ last = value;
+ }
+ }
+}
+
+/* Return 0 on success or 1 on failure. */
+static int e752x_get_devs(struct pci_dev *pdev, int dev_idx,
+ struct e752x_pvt *pvt)
+{
+ pvt->dev_d0f1 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ pvt->dev_info->err_dev, NULL);
+
+ if (pvt->dev_d0f1 == NULL) {
+ pvt->dev_d0f1 = pci_scan_single_device(pdev->bus,
+ PCI_DEVFN(0, 1));
+ pci_dev_get(pvt->dev_d0f1);
+ }
+
+ if (pvt->dev_d0f1 == NULL) {
+ e752x_printk(KERN_ERR, "error reporting device not found:"
+ "vendor %x device 0x%x (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, e752x_devs[dev_idx].err_dev);
+ return 1;
+ }
+
+ pvt->dev_d0f0 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ e752x_devs[dev_idx].ctl_dev,
+ NULL);
+
+ if (pvt->dev_d0f0 == NULL)
+ goto fail;
+
+ return 0;
+
+fail:
+ pci_dev_put(pvt->dev_d0f1);
+ return 1;
+}
+
+/* Setup system bus parity mask register.
+ * Sysbus parity supported on:
+ * e7320/e7520/e7525 + Xeon
+ */
+static void e752x_init_sysbus_parity_mask(struct e752x_pvt *pvt)
+{
+ char *cpu_id = cpu_data(0).x86_model_id;
+ struct pci_dev *dev = pvt->dev_d0f1;
+ int enable = 1;
+
+ /* Allow module parameter override, else see if CPU supports parity */
+ if (sysbus_parity != -1) {
+ enable = sysbus_parity;
+ } else if (cpu_id[0] && !strstr(cpu_id, "Xeon")) {
+ e752x_printk(KERN_INFO, "System Bus Parity not "
+ "supported by CPU, disabling\n");
+ enable = 0;
+ }
+
+ if (enable)
+ pci_write_config_word(dev, E752X_SYSBUS_ERRMASK, 0x0000);
+ else
+ pci_write_config_word(dev, E752X_SYSBUS_ERRMASK, 0x0309);
+}
+
+static void e752x_init_error_reporting_regs(struct e752x_pvt *pvt)
+{
+ struct pci_dev *dev;
+
+ dev = pvt->dev_d0f1;
+ /* Turn off error disable & SMI in case the BIOS turned it on */
+ if (pvt->dev_info->err_dev == PCI_DEVICE_ID_INTEL_3100_1_ERR) {
+ pci_write_config_dword(dev, I3100_NSI_EMASK, 0);
+ pci_write_config_dword(dev, I3100_NSI_SMICMD, 0);
+ } else {
+ pci_write_config_byte(dev, E752X_HI_ERRMASK, 0x00);
+ pci_write_config_byte(dev, E752X_HI_SMICMD, 0x00);
+ }
+
+ e752x_init_sysbus_parity_mask(pvt);
+
+ pci_write_config_word(dev, E752X_SYSBUS_SMICMD, 0x00);
+ pci_write_config_byte(dev, E752X_BUF_ERRMASK, 0x00);
+ pci_write_config_byte(dev, E752X_BUF_SMICMD, 0x00);
+ pci_write_config_byte(dev, E752X_DRAM_ERRMASK, 0x00);
+ pci_write_config_byte(dev, E752X_DRAM_SMICMD, 0x00);
+}
+
+static int e752x_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ u16 pci_data;
+ u8 stat8;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct e752x_pvt *pvt;
+ u16 ddrcsr;
+ int drc_chan; /* Number of channels 0=1chan,1=2chan */
+ struct e752x_error_info discard;
+
+ edac_dbg(0, "mci\n");
+ edac_dbg(0, "Starting Probe1\n");
+
+ /* check to see if device 0 function 1 is enabled; if it isn't, we
+ * assume the BIOS has reserved it for a reason and is expecting
+ * exclusive access, we take care not to violate that assumption and
+ * fail the probe. */
+ pci_read_config_byte(pdev, E752X_DEVPRES1, &stat8);
+ if (!force_function_unhide && !(stat8 & (1 << 5))) {
+ printk(KERN_INFO "Contact your BIOS vendor to see if the "
+ "E752x error registers can be safely un-hidden\n");
+ return -ENODEV;
+ }
+ stat8 |= (1 << 5);
+ pci_write_config_byte(pdev, E752X_DEVPRES1, stat8);
+
+ pci_read_config_word(pdev, E752X_DDRCSR, &ddrcsr);
+ /* FIXME: should check >>12 or 0xf, true for all? */
+ /* Dual channel = 1, Single channel = 0 */
+ drc_chan = dual_channel_active(ddrcsr);
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = E752X_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = drc_chan + 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(3, "init mci\n");
+ mci->mtype_cap = MEM_FLAG_RDDR;
+ /* 3100 IMCH supports SECDEC only */
+ mci->edac_ctl_cap = (dev_idx == I3100) ? EDAC_FLAG_SECDED :
+ (EDAC_FLAG_NONE | EDAC_FLAG_SECDED | EDAC_FLAG_S4ECD4ED);
+ /* FIXME - what if different memory types are in different csrows? */
+ mci->mod_name = EDAC_MOD_STR;
+ mci->pdev = &pdev->dev;
+
+ edac_dbg(3, "init pvt\n");
+ pvt = (struct e752x_pvt *)mci->pvt_info;
+ pvt->dev_info = &e752x_devs[dev_idx];
+ pvt->mc_symmetric = ((ddrcsr & 0x10) != 0);
+
+ if (e752x_get_devs(pdev, dev_idx, pvt)) {
+ edac_mc_free(mci);
+ return -ENODEV;
+ }
+
+ edac_dbg(3, "more mci init\n");
+ mci->ctl_name = pvt->dev_info->ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = e752x_check;
+ mci->ctl_page_to_phys = ctl_page_to_phys;
+ mci->set_sdram_scrub_rate = set_sdram_scrub_rate;
+ mci->get_sdram_scrub_rate = get_sdram_scrub_rate;
+
+ /* set the map type. 1 = normal, 0 = reversed
+ * Must be set before e752x_init_csrows in case csrow mapping
+ * is reversed.
+ */
+ pci_read_config_byte(pdev, E752X_DRM, &stat8);
+ pvt->map_type = ((stat8 & 0x0f) > ((stat8 >> 4) & 0x0f));
+
+ e752x_init_csrows(mci, pdev, ddrcsr);
+ e752x_init_mem_map_table(pdev, pvt);
+
+ if (dev_idx == I3100)
+ mci->edac_cap = EDAC_FLAG_SECDED; /* the only mode supported */
+ else
+ mci->edac_cap |= EDAC_FLAG_NONE;
+ edac_dbg(3, "tolm, remapbase, remaplimit\n");
+
+ /* load the top of low memory, remap base, and remap limit vars */
+ pci_read_config_word(pdev, E752X_TOLM, &pci_data);
+ pvt->tolm = ((u32) pci_data) << 4;
+ pci_read_config_word(pdev, E752X_REMAPBASE, &pci_data);
+ pvt->remapbase = ((u32) pci_data) << 14;
+ pci_read_config_word(pdev, E752X_REMAPLIMIT, &pci_data);
+ pvt->remaplimit = ((u32) pci_data) << 14;
+ e752x_printk(KERN_INFO,
+ "tolm = %x, remapbase = %x, remaplimit = %x\n",
+ pvt->tolm, pvt->remapbase, pvt->remaplimit);
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ e752x_init_error_reporting_regs(pvt);
+ e752x_get_error_info(mci, &discard); /* clear other MCH errors */
+
+ /* allocating generic PCI control info */
+ e752x_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!e752x_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n", __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail:
+ pci_dev_put(pvt->dev_d0f0);
+ pci_dev_put(pvt->dev_d0f1);
+ edac_mc_free(mci);
+
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int e752x_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ edac_dbg(0, "\n");
+
+ /* wake up and enable device */
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ return e752x_probe1(pdev, ent->driver_data);
+}
+
+static void e752x_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct e752x_pvt *pvt;
+
+ edac_dbg(0, "\n");
+
+ if (e752x_pci)
+ edac_pci_release_generic_ctl(e752x_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ pvt = (struct e752x_pvt *)mci->pvt_info;
+ pci_dev_put(pvt->dev_d0f0);
+ pci_dev_put(pvt->dev_d0f1);
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id e752x_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 7520_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7520},
+ {
+ PCI_VEND_DEV(INTEL, 7525_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7525},
+ {
+ PCI_VEND_DEV(INTEL, 7320_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7320},
+ {
+ PCI_VEND_DEV(INTEL, 3100_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I3100},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, e752x_pci_tbl);
+
+static struct pci_driver e752x_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = e752x_init_one,
+ .remove = e752x_remove_one,
+ .id_table = e752x_pci_tbl,
+};
+
+static int __init e752x_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&e752x_driver);
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+static void __exit e752x_exit(void)
+{
+ edac_dbg(3, "\n");
+ pci_unregister_driver(&e752x_driver);
+}
+
+module_init(e752x_init);
+module_exit(e752x_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Linux Networx (http://lnxi.com) Tom Zimmerman");
+MODULE_DESCRIPTION("MC support for Intel e752x/3100 memory controllers");
+
+module_param(force_function_unhide, int, 0444);
+MODULE_PARM_DESC(force_function_unhide, "if BIOS sets Dev0:Fun1 up as hidden:"
+ " 1=force unhide and hope BIOS doesn't fight driver for "
+ "Dev0:Fun1 access");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
+module_param(sysbus_parity, int, 0444);
+MODULE_PARM_DESC(sysbus_parity, "0=disable system bus parity checking,"
+ " 1=enable system bus parity checking, default=auto-detect");
+module_param(report_non_memory_errors, int, 0644);
+MODULE_PARM_DESC(report_non_memory_errors, "0=disable non-memory error "
+ "reporting, 1=enable non-memory error reporting");
diff --git a/drivers/edac/e7xxx_edac.c b/drivers/edac/e7xxx_edac.c
new file mode 100644
index 0000000000..5852b95fa4
--- /dev/null
+++ b/drivers/edac/e7xxx_edac.c
@@ -0,0 +1,602 @@
+/*
+ * Intel e7xxx Memory Controller kernel module
+ * (C) 2003 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * See "enum e7xxx_chips" below for supported chipsets
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * Datasheet:
+ * http://www.intel.com/content/www/us/en/chipsets/e7501-chipset-memory-controller-hub-datasheet.html
+ *
+ * Contributors:
+ * Eric Biederman (Linux Networx)
+ * Tom Zimmerman (Linux Networx)
+ * Jim Garlick (Lawrence Livermore National Labs)
+ * Dave Peterson (Lawrence Livermore National Labs)
+ * That One Guy (Some other place)
+ * Wang Zhenyu (intel.com)
+ *
+ * $Id: edac_e7xxx.c,v 1.5.2.9 2005/10/05 00:43:44 dsp_llnl Exp $
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "e7xxx_edac"
+
+#define e7xxx_printk(level, fmt, arg...) \
+ edac_printk(level, "e7xxx", fmt, ##arg)
+
+#define e7xxx_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "e7xxx", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_7205_0
+#define PCI_DEVICE_ID_INTEL_7205_0 0x255d
+#endif /* PCI_DEVICE_ID_INTEL_7205_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7205_1_ERR
+#define PCI_DEVICE_ID_INTEL_7205_1_ERR 0x2551
+#endif /* PCI_DEVICE_ID_INTEL_7205_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_7500_0
+#define PCI_DEVICE_ID_INTEL_7500_0 0x2540
+#endif /* PCI_DEVICE_ID_INTEL_7500_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7500_1_ERR
+#define PCI_DEVICE_ID_INTEL_7500_1_ERR 0x2541
+#endif /* PCI_DEVICE_ID_INTEL_7500_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_7501_0
+#define PCI_DEVICE_ID_INTEL_7501_0 0x254c
+#endif /* PCI_DEVICE_ID_INTEL_7501_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7501_1_ERR
+#define PCI_DEVICE_ID_INTEL_7501_1_ERR 0x2541
+#endif /* PCI_DEVICE_ID_INTEL_7501_1_ERR */
+
+#ifndef PCI_DEVICE_ID_INTEL_7505_0
+#define PCI_DEVICE_ID_INTEL_7505_0 0x2550
+#endif /* PCI_DEVICE_ID_INTEL_7505_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_7505_1_ERR
+#define PCI_DEVICE_ID_INTEL_7505_1_ERR 0x2551
+#endif /* PCI_DEVICE_ID_INTEL_7505_1_ERR */
+
+#define E7XXX_NR_CSROWS 8 /* number of csrows */
+#define E7XXX_NR_DIMMS 8 /* 2 channels, 4 dimms/channel */
+
+/* E7XXX register addresses - device 0 function 0 */
+#define E7XXX_DRB 0x60 /* DRAM row boundary register (8b) */
+#define E7XXX_DRA 0x70 /* DRAM row attribute register (8b) */
+ /*
+ * 31 Device width row 7 0=x8 1=x4
+ * 27 Device width row 6
+ * 23 Device width row 5
+ * 19 Device width row 4
+ * 15 Device width row 3
+ * 11 Device width row 2
+ * 7 Device width row 1
+ * 3 Device width row 0
+ */
+#define E7XXX_DRC 0x7C /* DRAM controller mode reg (32b) */
+ /*
+ * 22 Number channels 0=1,1=2
+ * 19:18 DRB Granularity 32/64MB
+ */
+#define E7XXX_TOLM 0xC4 /* DRAM top of low memory reg (16b) */
+#define E7XXX_REMAPBASE 0xC6 /* DRAM remap base address reg (16b) */
+#define E7XXX_REMAPLIMIT 0xC8 /* DRAM remap limit address reg (16b) */
+
+/* E7XXX register addresses - device 0 function 1 */
+#define E7XXX_DRAM_FERR 0x80 /* DRAM first error register (8b) */
+#define E7XXX_DRAM_NERR 0x82 /* DRAM next error register (8b) */
+#define E7XXX_DRAM_CELOG_ADD 0xA0 /* DRAM first correctable memory */
+ /* error address register (32b) */
+ /*
+ * 31:28 Reserved
+ * 27:6 CE address (4k block 33:12)
+ * 5:0 Reserved
+ */
+#define E7XXX_DRAM_UELOG_ADD 0xB0 /* DRAM first uncorrectable memory */
+ /* error address register (32b) */
+ /*
+ * 31:28 Reserved
+ * 27:6 CE address (4k block 33:12)
+ * 5:0 Reserved
+ */
+#define E7XXX_DRAM_CELOG_SYNDROME 0xD0 /* DRAM first correctable memory */
+ /* error syndrome register (16b) */
+
+enum e7xxx_chips {
+ E7500 = 0,
+ E7501,
+ E7505,
+ E7205,
+};
+
+struct e7xxx_pvt {
+ struct pci_dev *bridge_ck;
+ u32 tolm;
+ u32 remapbase;
+ u32 remaplimit;
+ const struct e7xxx_dev_info *dev_info;
+};
+
+struct e7xxx_dev_info {
+ u16 err_dev;
+ const char *ctl_name;
+};
+
+struct e7xxx_error_info {
+ u8 dram_ferr;
+ u8 dram_nerr;
+ u32 dram_celog_add;
+ u16 dram_celog_syndrome;
+ u32 dram_uelog_add;
+};
+
+static struct edac_pci_ctl_info *e7xxx_pci;
+
+static const struct e7xxx_dev_info e7xxx_devs[] = {
+ [E7500] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7500_1_ERR,
+ .ctl_name = "E7500"},
+ [E7501] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7501_1_ERR,
+ .ctl_name = "E7501"},
+ [E7505] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7505_1_ERR,
+ .ctl_name = "E7505"},
+ [E7205] = {
+ .err_dev = PCI_DEVICE_ID_INTEL_7205_1_ERR,
+ .ctl_name = "E7205"},
+};
+
+/* FIXME - is this valid for both SECDED and S4ECD4ED? */
+static inline int e7xxx_find_channel(u16 syndrome)
+{
+ edac_dbg(3, "\n");
+
+ if ((syndrome & 0xff00) == 0)
+ return 0;
+
+ if ((syndrome & 0x00ff) == 0)
+ return 1;
+
+ if ((syndrome & 0xf000) == 0 || (syndrome & 0x0f00) == 0)
+ return 0;
+
+ return 1;
+}
+
+static unsigned long ctl_page_to_phys(struct mem_ctl_info *mci,
+ unsigned long page)
+{
+ u32 remap;
+ struct e7xxx_pvt *pvt = (struct e7xxx_pvt *)mci->pvt_info;
+
+ edac_dbg(3, "\n");
+
+ if ((page < pvt->tolm) ||
+ ((page >= 0x100000) && (page < pvt->remapbase)))
+ return page;
+
+ remap = (page - pvt->tolm) + pvt->remapbase;
+
+ if (remap < pvt->remaplimit)
+ return remap;
+
+ e7xxx_printk(KERN_ERR, "Invalid page %lx - out of range\n", page);
+ return pvt->tolm - 1;
+}
+
+static void process_ce(struct mem_ctl_info *mci, struct e7xxx_error_info *info)
+{
+ u32 error_1b, page;
+ u16 syndrome;
+ int row;
+ int channel;
+
+ edac_dbg(3, "\n");
+ /* read the error address */
+ error_1b = info->dram_celog_add;
+ /* FIXME - should use PAGE_SHIFT */
+ page = error_1b >> 6; /* convert the address to 4k page */
+ /* read the syndrome */
+ syndrome = info->dram_celog_syndrome;
+ /* FIXME - check for -1 */
+ row = edac_mc_find_csrow_by_page(mci, page);
+ /* convert syndrome to channel */
+ channel = e7xxx_find_channel(syndrome);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, page, 0, syndrome,
+ row, channel, -1, "e7xxx CE", "");
+}
+
+static void process_ce_no_info(struct mem_ctl_info *mci)
+{
+ edac_dbg(3, "\n");
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0, -1, -1, -1,
+ "e7xxx CE log register overflow", "");
+}
+
+static void process_ue(struct mem_ctl_info *mci, struct e7xxx_error_info *info)
+{
+ u32 error_2b, block_page;
+ int row;
+
+ edac_dbg(3, "\n");
+ /* read the error address */
+ error_2b = info->dram_uelog_add;
+ /* FIXME - should use PAGE_SHIFT */
+ block_page = error_2b >> 6; /* convert to 4k address */
+ row = edac_mc_find_csrow_by_page(mci, block_page);
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, block_page, 0, 0,
+ row, -1, -1, "e7xxx UE", "");
+}
+
+static void process_ue_no_info(struct mem_ctl_info *mci)
+{
+ edac_dbg(3, "\n");
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0, -1, -1, -1,
+ "e7xxx UE log register overflow", "");
+}
+
+static void e7xxx_get_error_info(struct mem_ctl_info *mci,
+ struct e7xxx_error_info *info)
+{
+ struct e7xxx_pvt *pvt;
+
+ pvt = (struct e7xxx_pvt *)mci->pvt_info;
+ pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_FERR, &info->dram_ferr);
+ pci_read_config_byte(pvt->bridge_ck, E7XXX_DRAM_NERR, &info->dram_nerr);
+
+ if ((info->dram_ferr & 1) || (info->dram_nerr & 1)) {
+ pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_CELOG_ADD,
+ &info->dram_celog_add);
+ pci_read_config_word(pvt->bridge_ck,
+ E7XXX_DRAM_CELOG_SYNDROME,
+ &info->dram_celog_syndrome);
+ }
+
+ if ((info->dram_ferr & 2) || (info->dram_nerr & 2))
+ pci_read_config_dword(pvt->bridge_ck, E7XXX_DRAM_UELOG_ADD,
+ &info->dram_uelog_add);
+
+ if (info->dram_ferr & 3)
+ pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_FERR, 0x03, 0x03);
+
+ if (info->dram_nerr & 3)
+ pci_write_bits8(pvt->bridge_ck, E7XXX_DRAM_NERR, 0x03, 0x03);
+}
+
+static int e7xxx_process_error_info(struct mem_ctl_info *mci,
+ struct e7xxx_error_info *info,
+ int handle_errors)
+{
+ int error_found;
+
+ error_found = 0;
+
+ /* decode and report errors */
+ if (info->dram_ferr & 1) { /* check first error correctable */
+ error_found = 1;
+
+ if (handle_errors)
+ process_ce(mci, info);
+ }
+
+ if (info->dram_ferr & 2) { /* check first error uncorrectable */
+ error_found = 1;
+
+ if (handle_errors)
+ process_ue(mci, info);
+ }
+
+ if (info->dram_nerr & 1) { /* check next error correctable */
+ error_found = 1;
+
+ if (handle_errors) {
+ if (info->dram_ferr & 1)
+ process_ce_no_info(mci);
+ else
+ process_ce(mci, info);
+ }
+ }
+
+ if (info->dram_nerr & 2) { /* check next error uncorrectable */
+ error_found = 1;
+
+ if (handle_errors) {
+ if (info->dram_ferr & 2)
+ process_ue_no_info(mci);
+ else
+ process_ue(mci, info);
+ }
+ }
+
+ return error_found;
+}
+
+static void e7xxx_check(struct mem_ctl_info *mci)
+{
+ struct e7xxx_error_info info;
+
+ e7xxx_get_error_info(mci, &info);
+ e7xxx_process_error_info(mci, &info, 1);
+}
+
+/* Return 1 if dual channel mode is active. Else return 0. */
+static inline int dual_channel_active(u32 drc, int dev_idx)
+{
+ return (dev_idx == E7501) ? ((drc >> 22) & 0x1) : 1;
+}
+
+/* Return DRB granularity (0=32mb, 1=64mb). */
+static inline int drb_granularity(u32 drc, int dev_idx)
+{
+ /* only e7501 can be single channel */
+ return (dev_idx == E7501) ? ((drc >> 18) & 0x3) : 1;
+}
+
+static void e7xxx_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
+ int dev_idx, u32 drc)
+{
+ unsigned long last_cumul_size;
+ int index, j;
+ u8 value;
+ u32 dra, cumul_size, nr_pages;
+ int drc_chan, drc_drbg, drc_ddim, mem_dev;
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ enum edac_type edac_mode;
+
+ pci_read_config_dword(pdev, E7XXX_DRA, &dra);
+ drc_chan = dual_channel_active(drc, dev_idx);
+ drc_drbg = drb_granularity(drc, dev_idx);
+ drc_ddim = (drc >> 20) & 0x3;
+ last_cumul_size = 0;
+
+ /* The dram row boundary (DRB) reg values are boundary address
+ * for each DRAM row with a granularity of 32 or 64MB (single/dual
+ * channel operation). DRB regs are cumulative; therefore DRB7 will
+ * contain the total memory contained in all eight rows.
+ */
+ for (index = 0; index < mci->nr_csrows; index++) {
+ /* mem_dev 0=x8, 1=x4 */
+ mem_dev = (dra >> (index * 4 + 3)) & 0x1;
+ csrow = mci->csrows[index];
+
+ pci_read_config_byte(pdev, E7XXX_DRB + index, &value);
+ /* convert a 64 or 32 MiB DRB to a page size. */
+ cumul_size = value << (25 + drc_drbg - PAGE_SHIFT);
+ edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
+ if (cumul_size == last_cumul_size)
+ continue; /* not populated */
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ nr_pages = cumul_size - last_cumul_size;
+ last_cumul_size = cumul_size;
+
+ /*
+ * if single channel or x8 devices then SECDED
+ * if dual channel and x4 then S4ECD4ED
+ */
+ if (drc_ddim) {
+ if (drc_chan && mem_dev) {
+ edac_mode = EDAC_S4ECD4ED;
+ mci->edac_cap |= EDAC_FLAG_S4ECD4ED;
+ } else {
+ edac_mode = EDAC_SECDED;
+ mci->edac_cap |= EDAC_FLAG_SECDED;
+ }
+ } else
+ edac_mode = EDAC_NONE;
+
+ for (j = 0; j < drc_chan + 1; j++) {
+ dimm = csrow->channels[j]->dimm;
+
+ dimm->nr_pages = nr_pages / (drc_chan + 1);
+ dimm->grain = 1 << 12; /* 4KiB - resolution of CELOG */
+ dimm->mtype = MEM_RDDR; /* only one type supported */
+ dimm->dtype = mem_dev ? DEV_X4 : DEV_X8;
+ dimm->edac_mode = edac_mode;
+ }
+ }
+}
+
+static int e7xxx_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ u16 pci_data;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ struct e7xxx_pvt *pvt = NULL;
+ u32 drc;
+ int drc_chan;
+ struct e7xxx_error_info discard;
+
+ edac_dbg(0, "mci\n");
+
+ pci_read_config_dword(pdev, E7XXX_DRC, &drc);
+
+ drc_chan = dual_channel_active(drc, dev_idx);
+ /*
+ * According with the datasheet, this device has a maximum of
+ * 4 DIMMS per channel, either single-rank or dual-rank. So, the
+ * total amount of dimms is 8 (E7XXX_NR_DIMMS).
+ * That means that the DIMM is mapped as CSROWs, and the channel
+ * will map the rank. So, an error to either channel should be
+ * attributed to the same dimm.
+ */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = E7XXX_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = drc_chan + 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(3, "init mci\n");
+ mci->mtype_cap = MEM_FLAG_RDDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED |
+ EDAC_FLAG_S4ECD4ED;
+ /* FIXME - what if different memory types are in different csrows? */
+ mci->mod_name = EDAC_MOD_STR;
+ mci->pdev = &pdev->dev;
+ edac_dbg(3, "init pvt\n");
+ pvt = (struct e7xxx_pvt *)mci->pvt_info;
+ pvt->dev_info = &e7xxx_devs[dev_idx];
+ pvt->bridge_ck = pci_get_device(PCI_VENDOR_ID_INTEL,
+ pvt->dev_info->err_dev, pvt->bridge_ck);
+
+ if (!pvt->bridge_ck) {
+ e7xxx_printk(KERN_ERR, "error reporting device not found:"
+ "vendor %x device 0x%x (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, e7xxx_devs[dev_idx].err_dev);
+ goto fail0;
+ }
+
+ edac_dbg(3, "more mci init\n");
+ mci->ctl_name = pvt->dev_info->ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = e7xxx_check;
+ mci->ctl_page_to_phys = ctl_page_to_phys;
+ e7xxx_init_csrows(mci, pdev, dev_idx, drc);
+ mci->edac_cap |= EDAC_FLAG_NONE;
+ edac_dbg(3, "tolm, remapbase, remaplimit\n");
+ /* load the top of low memory, remap base, and remap limit vars */
+ pci_read_config_word(pdev, E7XXX_TOLM, &pci_data);
+ pvt->tolm = ((u32) pci_data) << 4;
+ pci_read_config_word(pdev, E7XXX_REMAPBASE, &pci_data);
+ pvt->remapbase = ((u32) pci_data) << 14;
+ pci_read_config_word(pdev, E7XXX_REMAPLIMIT, &pci_data);
+ pvt->remaplimit = ((u32) pci_data) << 14;
+ e7xxx_printk(KERN_INFO,
+ "tolm = %x, remapbase = %x, remaplimit = %x\n", pvt->tolm,
+ pvt->remapbase, pvt->remaplimit);
+
+ /* clear any pending errors, or initial state bits */
+ e7xxx_get_error_info(mci, &discard);
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail1;
+ }
+
+ /* allocating generic PCI control info */
+ e7xxx_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!e7xxx_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail1:
+ pci_dev_put(pvt->bridge_ck);
+
+fail0:
+ edac_mc_free(mci);
+
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int e7xxx_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ edac_dbg(0, "\n");
+
+ /* wake up and enable device */
+ return pci_enable_device(pdev) ?
+ -EIO : e7xxx_probe1(pdev, ent->driver_data);
+}
+
+static void e7xxx_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct e7xxx_pvt *pvt;
+
+ edac_dbg(0, "\n");
+
+ if (e7xxx_pci)
+ edac_pci_release_generic_ctl(e7xxx_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ pvt = (struct e7xxx_pvt *)mci->pvt_info;
+ pci_dev_put(pvt->bridge_ck);
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id e7xxx_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 7205_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7205},
+ {
+ PCI_VEND_DEV(INTEL, 7500_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7500},
+ {
+ PCI_VEND_DEV(INTEL, 7501_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7501},
+ {
+ PCI_VEND_DEV(INTEL, 7505_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ E7505},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, e7xxx_pci_tbl);
+
+static struct pci_driver e7xxx_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = e7xxx_init_one,
+ .remove = e7xxx_remove_one,
+ .id_table = e7xxx_pci_tbl,
+};
+
+static int __init e7xxx_init(void)
+{
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ return pci_register_driver(&e7xxx_driver);
+}
+
+static void __exit e7xxx_exit(void)
+{
+ pci_unregister_driver(&e7xxx_driver);
+}
+
+module_init(e7xxx_init);
+module_exit(e7xxx_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al");
+MODULE_DESCRIPTION("MC support for Intel e7xxx memory controllers");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/edac_device.c b/drivers/edac/edac_device.c
new file mode 100644
index 0000000000..0689e15107
--- /dev/null
+++ b/drivers/edac/edac_device.c
@@ -0,0 +1,621 @@
+
+/*
+ * edac_device.c
+ * (C) 2007 www.douglaskthompson.com
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Doug Thompson <norsk5@xmission.com>
+ *
+ * edac_device API implementation
+ * 19 Jan 2007
+ */
+
+#include <asm/page.h>
+#include <linux/uaccess.h>
+#include <linux/ctype.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/jiffies.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/spinlock.h>
+#include <linux/sysctl.h>
+#include <linux/timer.h>
+
+#include "edac_device.h"
+#include "edac_module.h"
+
+/* lock for the list: 'edac_device_list', manipulation of this list
+ * is protected by the 'device_ctls_mutex' lock
+ */
+static DEFINE_MUTEX(device_ctls_mutex);
+static LIST_HEAD(edac_device_list);
+
+/* Default workqueue processing interval on this instance, in msecs */
+#define DEFAULT_POLL_INTERVAL 1000
+
+#ifdef CONFIG_EDAC_DEBUG
+static void edac_device_dump_device(struct edac_device_ctl_info *edac_dev)
+{
+ edac_dbg(3, "\tedac_dev = %p dev_idx=%d\n",
+ edac_dev, edac_dev->dev_idx);
+ edac_dbg(4, "\tedac_dev->edac_check = %p\n", edac_dev->edac_check);
+ edac_dbg(3, "\tdev = %p\n", edac_dev->dev);
+ edac_dbg(3, "\tmod_name:ctl_name = %s:%s\n",
+ edac_dev->mod_name, edac_dev->ctl_name);
+ edac_dbg(3, "\tpvt_info = %p\n\n", edac_dev->pvt_info);
+}
+#endif /* CONFIG_EDAC_DEBUG */
+
+/*
+ * @off_val: zero, 1, or other based offset
+ */
+struct edac_device_ctl_info *
+edac_device_alloc_ctl_info(unsigned pvt_sz, char *dev_name, unsigned nr_instances,
+ char *blk_name, unsigned nr_blocks, unsigned off_val,
+ struct edac_dev_sysfs_block_attribute *attrib_spec,
+ unsigned nr_attrib, int device_index)
+{
+ struct edac_dev_sysfs_block_attribute *dev_attrib, *attrib_p, *attrib;
+ struct edac_device_block *dev_blk, *blk_p, *blk;
+ struct edac_device_instance *dev_inst, *inst;
+ struct edac_device_ctl_info *dev_ctl;
+ unsigned instance, block, attr;
+ void *pvt;
+ int err;
+
+ edac_dbg(4, "instances=%d blocks=%d\n", nr_instances, nr_blocks);
+
+ dev_ctl = kzalloc(sizeof(struct edac_device_ctl_info), GFP_KERNEL);
+ if (!dev_ctl)
+ return NULL;
+
+ dev_inst = kcalloc(nr_instances, sizeof(struct edac_device_instance), GFP_KERNEL);
+ if (!dev_inst)
+ goto free;
+
+ dev_ctl->instances = dev_inst;
+
+ dev_blk = kcalloc(nr_instances * nr_blocks, sizeof(struct edac_device_block), GFP_KERNEL);
+ if (!dev_blk)
+ goto free;
+
+ dev_ctl->blocks = dev_blk;
+
+ if (nr_attrib) {
+ dev_attrib = kcalloc(nr_attrib, sizeof(struct edac_dev_sysfs_block_attribute),
+ GFP_KERNEL);
+ if (!dev_attrib)
+ goto free;
+
+ dev_ctl->attribs = dev_attrib;
+ }
+
+ if (pvt_sz) {
+ pvt = kzalloc(pvt_sz, GFP_KERNEL);
+ if (!pvt)
+ goto free;
+
+ dev_ctl->pvt_info = pvt;
+ }
+
+ dev_ctl->dev_idx = device_index;
+ dev_ctl->nr_instances = nr_instances;
+
+ /* Default logging of CEs and UEs */
+ dev_ctl->log_ce = 1;
+ dev_ctl->log_ue = 1;
+
+ /* Name of this edac device */
+ snprintf(dev_ctl->name, sizeof(dev_ctl->name),"%s", dev_name);
+
+ /* Initialize every Instance */
+ for (instance = 0; instance < nr_instances; instance++) {
+ inst = &dev_inst[instance];
+ inst->ctl = dev_ctl;
+ inst->nr_blocks = nr_blocks;
+ blk_p = &dev_blk[instance * nr_blocks];
+ inst->blocks = blk_p;
+
+ /* name of this instance */
+ snprintf(inst->name, sizeof(inst->name), "%s%u", dev_name, instance);
+
+ /* Initialize every block in each instance */
+ for (block = 0; block < nr_blocks; block++) {
+ blk = &blk_p[block];
+ blk->instance = inst;
+ snprintf(blk->name, sizeof(blk->name),
+ "%s%d", blk_name, block + off_val);
+
+ edac_dbg(4, "instance=%d inst_p=%p block=#%d block_p=%p name='%s'\n",
+ instance, inst, block, blk, blk->name);
+
+ /* if there are NO attributes OR no attribute pointer
+ * then continue on to next block iteration
+ */
+ if ((nr_attrib == 0) || (attrib_spec == NULL))
+ continue;
+
+ /* setup the attribute array for this block */
+ blk->nr_attribs = nr_attrib;
+ attrib_p = &dev_attrib[block*nr_instances*nr_attrib];
+ blk->block_attributes = attrib_p;
+
+ edac_dbg(4, "THIS BLOCK_ATTRIB=%p\n",
+ blk->block_attributes);
+
+ /* Initialize every user specified attribute in this
+ * block with the data the caller passed in
+ * Each block gets its own copy of pointers,
+ * and its unique 'value'
+ */
+ for (attr = 0; attr < nr_attrib; attr++) {
+ attrib = &attrib_p[attr];
+
+ /* populate the unique per attrib
+ * with the code pointers and info
+ */
+ attrib->attr = attrib_spec[attr].attr;
+ attrib->show = attrib_spec[attr].show;
+ attrib->store = attrib_spec[attr].store;
+
+ attrib->block = blk; /* up link */
+
+ edac_dbg(4, "alloc-attrib=%p attrib_name='%s' attrib-spec=%p spec-name=%s\n",
+ attrib, attrib->attr.name,
+ &attrib_spec[attr],
+ attrib_spec[attr].attr.name
+ );
+ }
+ }
+ }
+
+ /* Mark this instance as merely ALLOCATED */
+ dev_ctl->op_state = OP_ALLOC;
+
+ /*
+ * Initialize the 'root' kobj for the edac_device controller
+ */
+ err = edac_device_register_sysfs_main_kobj(dev_ctl);
+ if (err)
+ goto free;
+
+ /* at this point, the root kobj is valid, and in order to
+ * 'free' the object, then the function:
+ * edac_device_unregister_sysfs_main_kobj() must be called
+ * which will perform kobj unregistration and the actual free
+ * will occur during the kobject callback operation
+ */
+
+ return dev_ctl;
+
+free:
+ __edac_device_free_ctl_info(dev_ctl);
+
+ return NULL;
+}
+EXPORT_SYMBOL_GPL(edac_device_alloc_ctl_info);
+
+void edac_device_free_ctl_info(struct edac_device_ctl_info *ctl_info)
+{
+ edac_device_unregister_sysfs_main_kobj(ctl_info);
+}
+EXPORT_SYMBOL_GPL(edac_device_free_ctl_info);
+
+/*
+ * find_edac_device_by_dev
+ * scans the edac_device list for a specific 'struct device *'
+ *
+ * lock to be held prior to call: device_ctls_mutex
+ *
+ * Return:
+ * pointer to control structure managing 'dev'
+ * NULL if not found on list
+ */
+static struct edac_device_ctl_info *find_edac_device_by_dev(struct device *dev)
+{
+ struct edac_device_ctl_info *edac_dev;
+ struct list_head *item;
+
+ edac_dbg(0, "\n");
+
+ list_for_each(item, &edac_device_list) {
+ edac_dev = list_entry(item, struct edac_device_ctl_info, link);
+
+ if (edac_dev->dev == dev)
+ return edac_dev;
+ }
+
+ return NULL;
+}
+
+/*
+ * add_edac_dev_to_global_list
+ * Before calling this function, caller must
+ * assign a unique value to edac_dev->dev_idx.
+ *
+ * lock to be held prior to call: device_ctls_mutex
+ *
+ * Return:
+ * 0 on success
+ * 1 on failure.
+ */
+static int add_edac_dev_to_global_list(struct edac_device_ctl_info *edac_dev)
+{
+ struct list_head *item, *insert_before;
+ struct edac_device_ctl_info *rover;
+
+ insert_before = &edac_device_list;
+
+ /* Determine if already on the list */
+ rover = find_edac_device_by_dev(edac_dev->dev);
+ if (unlikely(rover != NULL))
+ goto fail0;
+
+ /* Insert in ascending order by 'dev_idx', so find position */
+ list_for_each(item, &edac_device_list) {
+ rover = list_entry(item, struct edac_device_ctl_info, link);
+
+ if (rover->dev_idx >= edac_dev->dev_idx) {
+ if (unlikely(rover->dev_idx == edac_dev->dev_idx))
+ goto fail1;
+
+ insert_before = item;
+ break;
+ }
+ }
+
+ list_add_tail_rcu(&edac_dev->link, insert_before);
+ return 0;
+
+fail0:
+ edac_printk(KERN_WARNING, EDAC_MC,
+ "%s (%s) %s %s already assigned %d\n",
+ dev_name(rover->dev), edac_dev_name(rover),
+ rover->mod_name, rover->ctl_name, rover->dev_idx);
+ return 1;
+
+fail1:
+ edac_printk(KERN_WARNING, EDAC_MC,
+ "bug in low-level driver: attempt to assign\n"
+ " duplicate dev_idx %d in %s()\n", rover->dev_idx,
+ __func__);
+ return 1;
+}
+
+/*
+ * del_edac_device_from_global_list
+ */
+static void del_edac_device_from_global_list(struct edac_device_ctl_info
+ *edac_device)
+{
+ list_del_rcu(&edac_device->link);
+
+ /* these are for safe removal of devices from global list while
+ * NMI handlers may be traversing list
+ */
+ synchronize_rcu();
+ INIT_LIST_HEAD(&edac_device->link);
+}
+
+/*
+ * edac_device_workq_function
+ * performs the operation scheduled by a workq request
+ *
+ * this workq is embedded within an edac_device_ctl_info
+ * structure, that needs to be polled for possible error events.
+ *
+ * This operation is to acquire the list mutex lock
+ * (thus preventing insertation or deletion)
+ * and then call the device's poll function IFF this device is
+ * running polled and there is a poll function defined.
+ */
+static void edac_device_workq_function(struct work_struct *work_req)
+{
+ struct delayed_work *d_work = to_delayed_work(work_req);
+ struct edac_device_ctl_info *edac_dev = to_edac_device_ctl_work(d_work);
+
+ mutex_lock(&device_ctls_mutex);
+
+ /* If we are being removed, bail out immediately */
+ if (edac_dev->op_state == OP_OFFLINE) {
+ mutex_unlock(&device_ctls_mutex);
+ return;
+ }
+
+ /* Only poll controllers that are running polled and have a check */
+ if ((edac_dev->op_state == OP_RUNNING_POLL) &&
+ (edac_dev->edac_check != NULL)) {
+ edac_dev->edac_check(edac_dev);
+ }
+
+ mutex_unlock(&device_ctls_mutex);
+
+ /* Reschedule the workq for the next time period to start again
+ * if the number of msec is for 1 sec, then adjust to the next
+ * whole one second to save timers firing all over the period
+ * between integral seconds
+ */
+ if (edac_dev->poll_msec == DEFAULT_POLL_INTERVAL)
+ edac_queue_work(&edac_dev->work, round_jiffies_relative(edac_dev->delay));
+ else
+ edac_queue_work(&edac_dev->work, edac_dev->delay);
+}
+
+/*
+ * edac_device_workq_setup
+ * initialize a workq item for this edac_device instance
+ * passing in the new delay period in msec
+ */
+static void edac_device_workq_setup(struct edac_device_ctl_info *edac_dev,
+ unsigned msec)
+{
+ edac_dbg(0, "\n");
+
+ /* take the arg 'msec' and set it into the control structure
+ * to used in the time period calculation
+ * then calc the number of jiffies that represents
+ */
+ edac_dev->poll_msec = msec;
+ edac_dev->delay = msecs_to_jiffies(msec);
+
+ INIT_DELAYED_WORK(&edac_dev->work, edac_device_workq_function);
+
+ /* optimize here for the 1 second case, which will be normal value, to
+ * fire ON the 1 second time event. This helps reduce all sorts of
+ * timers firing on sub-second basis, while they are happy
+ * to fire together on the 1 second exactly
+ */
+ if (edac_dev->poll_msec == DEFAULT_POLL_INTERVAL)
+ edac_queue_work(&edac_dev->work, round_jiffies_relative(edac_dev->delay));
+ else
+ edac_queue_work(&edac_dev->work, edac_dev->delay);
+}
+
+/*
+ * edac_device_workq_teardown
+ * stop the workq processing on this edac_dev
+ */
+static void edac_device_workq_teardown(struct edac_device_ctl_info *edac_dev)
+{
+ if (!edac_dev->edac_check)
+ return;
+
+ edac_dev->op_state = OP_OFFLINE;
+
+ edac_stop_work(&edac_dev->work);
+}
+
+/*
+ * edac_device_reset_delay_period
+ *
+ * need to stop any outstanding workq queued up at this time
+ * because we will be resetting the sleep time.
+ * Then restart the workq on the new delay
+ */
+void edac_device_reset_delay_period(struct edac_device_ctl_info *edac_dev,
+ unsigned long msec)
+{
+ edac_dev->poll_msec = msec;
+ edac_dev->delay = msecs_to_jiffies(msec);
+
+ /* See comment in edac_device_workq_setup() above */
+ if (edac_dev->poll_msec == DEFAULT_POLL_INTERVAL)
+ edac_mod_work(&edac_dev->work, round_jiffies_relative(edac_dev->delay));
+ else
+ edac_mod_work(&edac_dev->work, edac_dev->delay);
+}
+
+int edac_device_alloc_index(void)
+{
+ static atomic_t device_indexes = ATOMIC_INIT(0);
+
+ return atomic_inc_return(&device_indexes) - 1;
+}
+EXPORT_SYMBOL_GPL(edac_device_alloc_index);
+
+int edac_device_add_device(struct edac_device_ctl_info *edac_dev)
+{
+ edac_dbg(0, "\n");
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (edac_debug_level >= 3)
+ edac_device_dump_device(edac_dev);
+#endif
+ mutex_lock(&device_ctls_mutex);
+
+ if (add_edac_dev_to_global_list(edac_dev))
+ goto fail0;
+
+ /* set load time so that error rate can be tracked */
+ edac_dev->start_time = jiffies;
+
+ /* create this instance's sysfs entries */
+ if (edac_device_create_sysfs(edac_dev)) {
+ edac_device_printk(edac_dev, KERN_WARNING,
+ "failed to create sysfs device\n");
+ goto fail1;
+ }
+
+ /* If there IS a check routine, then we are running POLLED */
+ if (edac_dev->edac_check != NULL) {
+ /* This instance is NOW RUNNING */
+ edac_dev->op_state = OP_RUNNING_POLL;
+
+ edac_device_workq_setup(edac_dev, edac_dev->poll_msec ?: DEFAULT_POLL_INTERVAL);
+ } else {
+ edac_dev->op_state = OP_RUNNING_INTERRUPT;
+ }
+
+ /* Report action taken */
+ edac_device_printk(edac_dev, KERN_INFO,
+ "Giving out device to module %s controller %s: DEV %s (%s)\n",
+ edac_dev->mod_name, edac_dev->ctl_name, edac_dev->dev_name,
+ edac_op_state_to_string(edac_dev->op_state));
+
+ mutex_unlock(&device_ctls_mutex);
+ return 0;
+
+fail1:
+ /* Some error, so remove the entry from the lsit */
+ del_edac_device_from_global_list(edac_dev);
+
+fail0:
+ mutex_unlock(&device_ctls_mutex);
+ return 1;
+}
+EXPORT_SYMBOL_GPL(edac_device_add_device);
+
+struct edac_device_ctl_info *edac_device_del_device(struct device *dev)
+{
+ struct edac_device_ctl_info *edac_dev;
+
+ edac_dbg(0, "\n");
+
+ mutex_lock(&device_ctls_mutex);
+
+ /* Find the structure on the list, if not there, then leave */
+ edac_dev = find_edac_device_by_dev(dev);
+ if (edac_dev == NULL) {
+ mutex_unlock(&device_ctls_mutex);
+ return NULL;
+ }
+
+ /* mark this instance as OFFLINE */
+ edac_dev->op_state = OP_OFFLINE;
+
+ /* deregister from global list */
+ del_edac_device_from_global_list(edac_dev);
+
+ mutex_unlock(&device_ctls_mutex);
+
+ /* clear workq processing on this instance */
+ edac_device_workq_teardown(edac_dev);
+
+ /* Tear down the sysfs entries for this instance */
+ edac_device_remove_sysfs(edac_dev);
+
+ edac_printk(KERN_INFO, EDAC_MC,
+ "Removed device %d for %s %s: DEV %s\n",
+ edac_dev->dev_idx,
+ edac_dev->mod_name, edac_dev->ctl_name, edac_dev_name(edac_dev));
+
+ return edac_dev;
+}
+EXPORT_SYMBOL_GPL(edac_device_del_device);
+
+static inline int edac_device_get_log_ce(struct edac_device_ctl_info *edac_dev)
+{
+ return edac_dev->log_ce;
+}
+
+static inline int edac_device_get_log_ue(struct edac_device_ctl_info *edac_dev)
+{
+ return edac_dev->log_ue;
+}
+
+static inline int edac_device_get_panic_on_ue(struct edac_device_ctl_info
+ *edac_dev)
+{
+ return edac_dev->panic_on_ue;
+}
+
+void edac_device_handle_ce_count(struct edac_device_ctl_info *edac_dev,
+ unsigned int count, int inst_nr, int block_nr,
+ const char *msg)
+{
+ struct edac_device_instance *instance;
+ struct edac_device_block *block = NULL;
+
+ if (!count)
+ return;
+
+ if ((inst_nr >= edac_dev->nr_instances) || (inst_nr < 0)) {
+ edac_device_printk(edac_dev, KERN_ERR,
+ "INTERNAL ERROR: 'instance' out of range "
+ "(%d >= %d)\n", inst_nr,
+ edac_dev->nr_instances);
+ return;
+ }
+
+ instance = edac_dev->instances + inst_nr;
+
+ if ((block_nr >= instance->nr_blocks) || (block_nr < 0)) {
+ edac_device_printk(edac_dev, KERN_ERR,
+ "INTERNAL ERROR: instance %d 'block' "
+ "out of range (%d >= %d)\n",
+ inst_nr, block_nr,
+ instance->nr_blocks);
+ return;
+ }
+
+ if (instance->nr_blocks > 0) {
+ block = instance->blocks + block_nr;
+ block->counters.ce_count += count;
+ }
+
+ /* Propagate the count up the 'totals' tree */
+ instance->counters.ce_count += count;
+ edac_dev->counters.ce_count += count;
+
+ if (edac_device_get_log_ce(edac_dev))
+ edac_device_printk(edac_dev, KERN_WARNING,
+ "CE: %s instance: %s block: %s count: %d '%s'\n",
+ edac_dev->ctl_name, instance->name,
+ block ? block->name : "N/A", count, msg);
+}
+EXPORT_SYMBOL_GPL(edac_device_handle_ce_count);
+
+void edac_device_handle_ue_count(struct edac_device_ctl_info *edac_dev,
+ unsigned int count, int inst_nr, int block_nr,
+ const char *msg)
+{
+ struct edac_device_instance *instance;
+ struct edac_device_block *block = NULL;
+
+ if (!count)
+ return;
+
+ if ((inst_nr >= edac_dev->nr_instances) || (inst_nr < 0)) {
+ edac_device_printk(edac_dev, KERN_ERR,
+ "INTERNAL ERROR: 'instance' out of range "
+ "(%d >= %d)\n", inst_nr,
+ edac_dev->nr_instances);
+ return;
+ }
+
+ instance = edac_dev->instances + inst_nr;
+
+ if ((block_nr >= instance->nr_blocks) || (block_nr < 0)) {
+ edac_device_printk(edac_dev, KERN_ERR,
+ "INTERNAL ERROR: instance %d 'block' "
+ "out of range (%d >= %d)\n",
+ inst_nr, block_nr,
+ instance->nr_blocks);
+ return;
+ }
+
+ if (instance->nr_blocks > 0) {
+ block = instance->blocks + block_nr;
+ block->counters.ue_count += count;
+ }
+
+ /* Propagate the count up the 'totals' tree */
+ instance->counters.ue_count += count;
+ edac_dev->counters.ue_count += count;
+
+ if (edac_device_get_log_ue(edac_dev))
+ edac_device_printk(edac_dev, KERN_EMERG,
+ "UE: %s instance: %s block: %s count: %d '%s'\n",
+ edac_dev->ctl_name, instance->name,
+ block ? block->name : "N/A", count, msg);
+
+ if (edac_device_get_panic_on_ue(edac_dev))
+ panic("EDAC %s: UE instance: %s block %s count: %d '%s'\n",
+ edac_dev->ctl_name, instance->name,
+ block ? block->name : "N/A", count, msg);
+}
+EXPORT_SYMBOL_GPL(edac_device_handle_ue_count);
diff --git a/drivers/edac/edac_device.h b/drivers/edac/edac_device.h
new file mode 100644
index 0000000000..3f44e6b9d3
--- /dev/null
+++ b/drivers/edac/edac_device.h
@@ -0,0 +1,365 @@
+/*
+ * Defines, structures, APIs for edac_device
+ *
+ * (C) 2007 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * NMI handling support added by
+ * Dave Peterson <dsp@llnl.gov> <dave_peterson@pobox.com>
+ *
+ * Refactored for multi-source files:
+ * Doug Thompson <norsk5@xmission.com>
+ *
+ * Please look at Documentation/driver-api/edac.rst for more info about
+ * EDAC core structs and functions.
+ */
+
+#ifndef _EDAC_DEVICE_H_
+#define _EDAC_DEVICE_H_
+
+#include <linux/completion.h>
+#include <linux/device.h>
+#include <linux/edac.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/types.h>
+#include <linux/sysfs.h>
+#include <linux/workqueue.h>
+
+
+/*
+ * The following are the structures to provide for a generic
+ * or abstract 'edac_device'. This set of structures and the
+ * code that implements the APIs for the same, provide for
+ * registering EDAC type devices which are NOT standard memory.
+ *
+ * CPU caches (L1 and L2)
+ * DMA engines
+ * Core CPU switches
+ * Fabric switch units
+ * PCIe interface controllers
+ * other EDAC/ECC type devices that can be monitored for
+ * errors, etc.
+ *
+ * It allows for a 2 level set of hierarchy. For example:
+ *
+ * cache could be composed of L1, L2 and L3 levels of cache.
+ * Each CPU core would have its own L1 cache, while sharing
+ * L2 and maybe L3 caches.
+ *
+ * View them arranged, via the sysfs presentation:
+ * /sys/devices/system/edac/..
+ *
+ * mc/ <existing memory device directory>
+ * cpu/cpu0/.. <L1 and L2 block directory>
+ * /L1-cache/ce_count
+ * /ue_count
+ * /L2-cache/ce_count
+ * /ue_count
+ * cpu/cpu1/.. <L1 and L2 block directory>
+ * /L1-cache/ce_count
+ * /ue_count
+ * /L2-cache/ce_count
+ * /ue_count
+ * ...
+ *
+ * the L1 and L2 directories would be "edac_device_block's"
+ */
+
+struct edac_device_counter {
+ u32 ue_count;
+ u32 ce_count;
+};
+
+/* forward reference */
+struct edac_device_ctl_info;
+struct edac_device_block;
+
+/* edac_dev_sysfs_attribute structure
+ * used for driver sysfs attributes in mem_ctl_info
+ * for extra controls and attributes:
+ * like high level error Injection controls
+ */
+struct edac_dev_sysfs_attribute {
+ struct attribute attr;
+ ssize_t (*show)(struct edac_device_ctl_info *, char *);
+ ssize_t (*store)(struct edac_device_ctl_info *, const char *, size_t);
+};
+
+/* edac_dev_sysfs_block_attribute structure
+ *
+ * used in leaf 'block' nodes for adding controls/attributes
+ *
+ * each block in each instance of the containing control structure
+ * can have an array of the following. The show and store functions
+ * will be filled in with the show/store function in the
+ * low level driver.
+ *
+ * The 'value' field will be the actual value field used for
+ * counting
+ */
+struct edac_dev_sysfs_block_attribute {
+ struct attribute attr;
+ ssize_t (*show)(struct kobject *, struct attribute *, char *);
+ ssize_t (*store)(struct kobject *, struct attribute *,
+ const char *, size_t);
+ struct edac_device_block *block;
+
+ unsigned int value;
+};
+
+/* device block control structure */
+struct edac_device_block {
+ struct edac_device_instance *instance; /* Up Pointer */
+ char name[EDAC_DEVICE_NAME_LEN + 1];
+
+ struct edac_device_counter counters; /* basic UE and CE counters */
+
+ int nr_attribs; /* how many attributes */
+
+ /* this block's attributes, could be NULL */
+ struct edac_dev_sysfs_block_attribute *block_attributes;
+
+ /* edac sysfs device control */
+ struct kobject kobj;
+};
+
+/* device instance control structure */
+struct edac_device_instance {
+ struct edac_device_ctl_info *ctl; /* Up pointer */
+ char name[EDAC_DEVICE_NAME_LEN + 4];
+
+ struct edac_device_counter counters; /* instance counters */
+
+ u32 nr_blocks; /* how many blocks */
+ struct edac_device_block *blocks; /* block array */
+
+ /* edac sysfs device control */
+ struct kobject kobj;
+};
+
+
+/*
+ * Abstract edac_device control info structure
+ *
+ */
+struct edac_device_ctl_info {
+ /* for global list of edac_device_ctl_info structs */
+ struct list_head link;
+
+ struct module *owner; /* Module owner of this control struct */
+
+ int dev_idx;
+
+ /* Per instance controls for this edac_device */
+ int log_ue; /* boolean for logging UEs */
+ int log_ce; /* boolean for logging CEs */
+ int panic_on_ue; /* boolean for panic'ing on an UE */
+ unsigned poll_msec; /* number of milliseconds to poll interval */
+ unsigned long delay; /* number of jiffies for poll_msec */
+
+ /* Additional top controller level attributes, but specified
+ * by the low level driver.
+ *
+ * Set by the low level driver to provide attributes at the
+ * controller level, same level as 'ue_count' and 'ce_count' above.
+ * An array of structures, NULL terminated
+ *
+ * If attributes are desired, then set to array of attributes
+ * If no attributes are desired, leave NULL
+ */
+ struct edac_dev_sysfs_attribute *sysfs_attributes;
+
+ /* pointer to main 'edac' subsys in sysfs */
+ struct bus_type *edac_subsys;
+
+ /* the internal state of this controller instance */
+ int op_state;
+ /* work struct for this instance */
+ struct delayed_work work;
+
+ /* pointer to edac polling checking routine:
+ * If NOT NULL: points to polling check routine
+ * If NULL: Then assumes INTERRUPT operation, where
+ * MC driver will receive events
+ */
+ void (*edac_check) (struct edac_device_ctl_info * edac_dev);
+
+ struct device *dev; /* pointer to device structure */
+
+ const char *mod_name; /* module name */
+ const char *ctl_name; /* edac controller name */
+ const char *dev_name; /* pci/platform/etc... name */
+
+ void *pvt_info; /* pointer to 'private driver' info */
+
+ unsigned long start_time; /* edac_device load start time (jiffies) */
+
+ struct completion removal_complete;
+
+ /* sysfs top name under 'edac' directory
+ * and instance name:
+ * cpu/cpu0/...
+ * cpu/cpu1/...
+ * cpu/cpu2/...
+ * ...
+ */
+ char name[EDAC_DEVICE_NAME_LEN + 1];
+
+ /* Number of instances supported on this control structure
+ * and the array of those instances
+ */
+ u32 nr_instances;
+ struct edac_device_instance *instances;
+ struct edac_device_block *blocks;
+ struct edac_dev_sysfs_block_attribute *attribs;
+
+ /* Event counters for the this whole EDAC Device */
+ struct edac_device_counter counters;
+
+ /* edac sysfs device control for the 'name'
+ * device this structure controls
+ */
+ struct kobject kobj;
+};
+
+/* To get from the instance's wq to the beginning of the ctl structure */
+#define to_edac_mem_ctl_work(w) \
+ container_of(w, struct mem_ctl_info, work)
+
+#define to_edac_device_ctl_work(w) \
+ container_of(w,struct edac_device_ctl_info,work)
+
+/*
+ * The alloc() and free() functions for the 'edac_device' control info
+ * structure. A MC driver will allocate one of these for each edac_device
+ * it is going to control/register with the EDAC CORE.
+ */
+extern struct edac_device_ctl_info *edac_device_alloc_ctl_info(
+ unsigned sizeof_private,
+ char *edac_device_name, unsigned nr_instances,
+ char *edac_block_name, unsigned nr_blocks,
+ unsigned offset_value,
+ struct edac_dev_sysfs_block_attribute *block_attributes,
+ unsigned nr_attribs,
+ int device_index);
+
+/* The offset value can be:
+ * -1 indicating no offset value
+ * 0 for zero-based block numbers
+ * 1 for 1-based block number
+ * other for other-based block number
+ */
+#define BLOCK_OFFSET_VALUE_OFF ((unsigned) -1)
+
+extern void edac_device_free_ctl_info(struct edac_device_ctl_info *ctl_info);
+
+/**
+ * edac_device_add_device - Insert the 'edac_dev' structure into the
+ * edac_device global list and create sysfs entries associated with
+ * edac_device structure.
+ *
+ * @edac_dev: pointer to edac_device structure to be added to the list
+ * 'edac_device' structure.
+ *
+ * Returns:
+ * 0 on Success, or an error code on failure
+ */
+extern int edac_device_add_device(struct edac_device_ctl_info *edac_dev);
+
+/**
+ * edac_device_del_device - Remove sysfs entries for specified edac_device
+ * structure and then remove edac_device structure from global list
+ *
+ * @dev:
+ * Pointer to struct &device representing the edac device
+ * structure to remove.
+ *
+ * Returns:
+ * Pointer to removed edac_device structure,
+ * or %NULL if device not found.
+ */
+extern struct edac_device_ctl_info *edac_device_del_device(struct device *dev);
+
+/**
+ * edac_device_handle_ce_count - Log correctable errors.
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the CE error happened
+ * @count: Number of errors to log.
+ * @block_nr: number of the block where the CE error happened
+ * @msg: message to be printed
+ */
+void edac_device_handle_ce_count(struct edac_device_ctl_info *edac_dev,
+ unsigned int count, int inst_nr, int block_nr,
+ const char *msg);
+
+/**
+ * edac_device_handle_ue_count - Log uncorrectable errors.
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the CE error happened
+ * @count: Number of errors to log.
+ * @block_nr: number of the block where the CE error happened
+ * @msg: message to be printed
+ */
+void edac_device_handle_ue_count(struct edac_device_ctl_info *edac_dev,
+ unsigned int count, int inst_nr, int block_nr,
+ const char *msg);
+
+/**
+ * edac_device_handle_ce(): Log a single correctable error
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the CE error happened
+ * @block_nr: number of the block where the CE error happened
+ * @msg: message to be printed
+ */
+static inline void
+edac_device_handle_ce(struct edac_device_ctl_info *edac_dev, int inst_nr,
+ int block_nr, const char *msg)
+{
+ edac_device_handle_ce_count(edac_dev, 1, inst_nr, block_nr, msg);
+}
+
+/**
+ * edac_device_handle_ue(): Log a single uncorrectable error
+ *
+ * @edac_dev: pointer to struct &edac_device_ctl_info
+ * @inst_nr: number of the instance where the UE error happened
+ * @block_nr: number of the block where the UE error happened
+ * @msg: message to be printed
+ */
+static inline void
+edac_device_handle_ue(struct edac_device_ctl_info *edac_dev, int inst_nr,
+ int block_nr, const char *msg)
+{
+ edac_device_handle_ue_count(edac_dev, 1, inst_nr, block_nr, msg);
+}
+
+/**
+ * edac_device_alloc_index: Allocate a unique device index number
+ *
+ * Returns:
+ * allocated index number
+ */
+extern int edac_device_alloc_index(void);
+extern const char *edac_layer_name[];
+
+/* Free the actual struct */
+static inline void __edac_device_free_ctl_info(struct edac_device_ctl_info *ci)
+{
+ if (ci) {
+ kfree(ci->pvt_info);
+ kfree(ci->attribs);
+ kfree(ci->blocks);
+ kfree(ci->instances);
+ kfree(ci);
+ }
+}
+#endif
diff --git a/drivers/edac/edac_device_sysfs.c b/drivers/edac/edac_device_sysfs.c
new file mode 100644
index 0000000000..010c26be58
--- /dev/null
+++ b/drivers/edac/edac_device_sysfs.c
@@ -0,0 +1,873 @@
+/*
+ * file for managing the edac_device subsystem of devices for EDAC
+ *
+ * (C) 2007 SoftwareBitMaker
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written Doug Thompson <norsk5@xmission.com>
+ *
+ */
+
+#include <linux/ctype.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+
+#include "edac_device.h"
+#include "edac_module.h"
+
+#define EDAC_DEVICE_SYMLINK "device"
+
+#define to_edacdev(k) container_of(k, struct edac_device_ctl_info, kobj)
+#define to_edacdev_attr(a) container_of(a, struct edacdev_attribute, attr)
+
+
+/*
+ * Set of edac_device_ctl_info attribute store/show functions
+ */
+
+/* 'log_ue' */
+static ssize_t edac_device_ctl_log_ue_show(struct edac_device_ctl_info
+ *ctl_info, char *data)
+{
+ return sprintf(data, "%u\n", ctl_info->log_ue);
+}
+
+static ssize_t edac_device_ctl_log_ue_store(struct edac_device_ctl_info
+ *ctl_info, const char *data,
+ size_t count)
+{
+ /* if parameter is zero, turn off flag, if non-zero turn on flag */
+ ctl_info->log_ue = (simple_strtoul(data, NULL, 0) != 0);
+
+ return count;
+}
+
+/* 'log_ce' */
+static ssize_t edac_device_ctl_log_ce_show(struct edac_device_ctl_info
+ *ctl_info, char *data)
+{
+ return sprintf(data, "%u\n", ctl_info->log_ce);
+}
+
+static ssize_t edac_device_ctl_log_ce_store(struct edac_device_ctl_info
+ *ctl_info, const char *data,
+ size_t count)
+{
+ /* if parameter is zero, turn off flag, if non-zero turn on flag */
+ ctl_info->log_ce = (simple_strtoul(data, NULL, 0) != 0);
+
+ return count;
+}
+
+/* 'panic_on_ue' */
+static ssize_t edac_device_ctl_panic_on_ue_show(struct edac_device_ctl_info
+ *ctl_info, char *data)
+{
+ return sprintf(data, "%u\n", ctl_info->panic_on_ue);
+}
+
+static ssize_t edac_device_ctl_panic_on_ue_store(struct edac_device_ctl_info
+ *ctl_info, const char *data,
+ size_t count)
+{
+ /* if parameter is zero, turn off flag, if non-zero turn on flag */
+ ctl_info->panic_on_ue = (simple_strtoul(data, NULL, 0) != 0);
+
+ return count;
+}
+
+/* 'poll_msec' show and store functions*/
+static ssize_t edac_device_ctl_poll_msec_show(struct edac_device_ctl_info
+ *ctl_info, char *data)
+{
+ return sprintf(data, "%u\n", ctl_info->poll_msec);
+}
+
+static ssize_t edac_device_ctl_poll_msec_store(struct edac_device_ctl_info
+ *ctl_info, const char *data,
+ size_t count)
+{
+ unsigned long value;
+
+ /* get the value and enforce that it is non-zero, must be at least
+ * one millisecond for the delay period, between scans
+ * Then cancel last outstanding delay for the work request
+ * and set a new one.
+ */
+ value = simple_strtoul(data, NULL, 0);
+ edac_device_reset_delay_period(ctl_info, value);
+
+ return count;
+}
+
+/* edac_device_ctl_info specific attribute structure */
+struct ctl_info_attribute {
+ struct attribute attr;
+ ssize_t(*show) (struct edac_device_ctl_info *, char *);
+ ssize_t(*store) (struct edac_device_ctl_info *, const char *, size_t);
+};
+
+#define to_ctl_info(k) container_of(k, struct edac_device_ctl_info, kobj)
+#define to_ctl_info_attr(a) container_of(a,struct ctl_info_attribute,attr)
+
+/* Function to 'show' fields from the edac_dev 'ctl_info' structure */
+static ssize_t edac_dev_ctl_info_show(struct kobject *kobj,
+ struct attribute *attr, char *buffer)
+{
+ struct edac_device_ctl_info *edac_dev = to_ctl_info(kobj);
+ struct ctl_info_attribute *ctl_info_attr = to_ctl_info_attr(attr);
+
+ if (ctl_info_attr->show)
+ return ctl_info_attr->show(edac_dev, buffer);
+ return -EIO;
+}
+
+/* Function to 'store' fields into the edac_dev 'ctl_info' structure */
+static ssize_t edac_dev_ctl_info_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
+{
+ struct edac_device_ctl_info *edac_dev = to_ctl_info(kobj);
+ struct ctl_info_attribute *ctl_info_attr = to_ctl_info_attr(attr);
+
+ if (ctl_info_attr->store)
+ return ctl_info_attr->store(edac_dev, buffer, count);
+ return -EIO;
+}
+
+/* edac_dev file operations for an 'ctl_info' */
+static const struct sysfs_ops device_ctl_info_ops = {
+ .show = edac_dev_ctl_info_show,
+ .store = edac_dev_ctl_info_store
+};
+
+#define CTL_INFO_ATTR(_name,_mode,_show,_store) \
+static struct ctl_info_attribute attr_ctl_info_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .show = _show, \
+ .store = _store, \
+};
+
+/* Declare the various ctl_info attributes here and their respective ops */
+CTL_INFO_ATTR(log_ue, S_IRUGO | S_IWUSR,
+ edac_device_ctl_log_ue_show, edac_device_ctl_log_ue_store);
+CTL_INFO_ATTR(log_ce, S_IRUGO | S_IWUSR,
+ edac_device_ctl_log_ce_show, edac_device_ctl_log_ce_store);
+CTL_INFO_ATTR(panic_on_ue, S_IRUGO | S_IWUSR,
+ edac_device_ctl_panic_on_ue_show,
+ edac_device_ctl_panic_on_ue_store);
+CTL_INFO_ATTR(poll_msec, S_IRUGO | S_IWUSR,
+ edac_device_ctl_poll_msec_show, edac_device_ctl_poll_msec_store);
+
+/* Base Attributes of the EDAC_DEVICE ECC object */
+static struct attribute *device_ctrl_attrs[] = {
+ &attr_ctl_info_panic_on_ue.attr,
+ &attr_ctl_info_log_ue.attr,
+ &attr_ctl_info_log_ce.attr,
+ &attr_ctl_info_poll_msec.attr,
+ NULL,
+};
+ATTRIBUTE_GROUPS(device_ctrl);
+
+/*
+ * edac_device_ctrl_master_release
+ *
+ * called when the reference count for the 'main' kobj
+ * for a edac_device control struct reaches zero
+ *
+ * Reference count model:
+ * One 'main' kobject for each control structure allocated.
+ * That main kobj is initially set to one AND
+ * the reference count for the EDAC 'core' module is
+ * bumped by one, thus added 'keep in memory' dependency.
+ *
+ * Each new internal kobj (in instances and blocks) then
+ * bumps the 'main' kobject.
+ *
+ * When they are released their release functions decrement
+ * the 'main' kobj.
+ *
+ * When the main kobj reaches zero (0) then THIS function
+ * is called which then decrements the EDAC 'core' module.
+ * When the module reference count reaches zero then the
+ * module no longer has dependency on keeping the release
+ * function code in memory and module can be unloaded.
+ *
+ * This will support several control objects as well, each
+ * with its own 'main' kobj.
+ */
+static void edac_device_ctrl_master_release(struct kobject *kobj)
+{
+ struct edac_device_ctl_info *edac_dev = to_edacdev(kobj);
+
+ edac_dbg(4, "control index=%d\n", edac_dev->dev_idx);
+
+ /* decrement the EDAC CORE module ref count */
+ module_put(edac_dev->owner);
+
+ __edac_device_free_ctl_info(edac_dev);
+}
+
+/* ktype for the main (master) kobject */
+static struct kobj_type ktype_device_ctrl = {
+ .release = edac_device_ctrl_master_release,
+ .sysfs_ops = &device_ctl_info_ops,
+ .default_groups = device_ctrl_groups,
+};
+
+/*
+ * edac_device_register_sysfs_main_kobj
+ *
+ * perform the high level setup for the new edac_device instance
+ *
+ * Return: 0 SUCCESS
+ * !0 FAILURE
+ */
+int edac_device_register_sysfs_main_kobj(struct edac_device_ctl_info *edac_dev)
+{
+ struct device *dev_root;
+ struct bus_type *edac_subsys;
+ int err = -ENODEV;
+
+ edac_dbg(1, "\n");
+
+ /* get the /sys/devices/system/edac reference */
+ edac_subsys = edac_get_sysfs_subsys();
+
+ /* Point to the 'edac_subsys' this instance 'reports' to */
+ edac_dev->edac_subsys = edac_subsys;
+
+ /* Init the devices's kobject */
+ memset(&edac_dev->kobj, 0, sizeof(struct kobject));
+
+ /* Record which module 'owns' this control structure
+ * and bump the ref count of the module
+ */
+ edac_dev->owner = THIS_MODULE;
+
+ if (!try_module_get(edac_dev->owner))
+ goto err_out;
+
+ /* register */
+ dev_root = bus_get_dev_root(edac_subsys);
+ if (dev_root) {
+ err = kobject_init_and_add(&edac_dev->kobj, &ktype_device_ctrl,
+ &dev_root->kobj, "%s", edac_dev->name);
+ put_device(dev_root);
+ }
+ if (err) {
+ edac_dbg(1, "Failed to register '.../edac/%s'\n",
+ edac_dev->name);
+ goto err_kobj_reg;
+ }
+ kobject_uevent(&edac_dev->kobj, KOBJ_ADD);
+
+ /* At this point, to 'free' the control struct,
+ * edac_device_unregister_sysfs_main_kobj() must be used
+ */
+
+ edac_dbg(4, "Registered '.../edac/%s' kobject\n", edac_dev->name);
+
+ return 0;
+
+ /* Error exit stack */
+err_kobj_reg:
+ kobject_put(&edac_dev->kobj);
+ module_put(edac_dev->owner);
+
+err_out:
+ return err;
+}
+
+/*
+ * edac_device_unregister_sysfs_main_kobj:
+ * the '..../edac/<name>' kobject
+ */
+void edac_device_unregister_sysfs_main_kobj(struct edac_device_ctl_info *dev)
+{
+ edac_dbg(0, "\n");
+ edac_dbg(4, "name of kobject is: %s\n", kobject_name(&dev->kobj));
+
+ /*
+ * Unregister the edac device's kobject and
+ * allow for reference count to reach 0 at which point
+ * the callback will be called to:
+ * a) module_put() this module
+ * b) 'kfree' the memory
+ */
+ kobject_put(&dev->kobj);
+}
+
+/* edac_dev -> instance information */
+
+/*
+ * Set of low-level instance attribute show functions
+ */
+static ssize_t instance_ue_count_show(struct edac_device_instance *instance,
+ char *data)
+{
+ return sprintf(data, "%u\n", instance->counters.ue_count);
+}
+
+static ssize_t instance_ce_count_show(struct edac_device_instance *instance,
+ char *data)
+{
+ return sprintf(data, "%u\n", instance->counters.ce_count);
+}
+
+#define to_instance(k) container_of(k, struct edac_device_instance, kobj)
+#define to_instance_attr(a) container_of(a,struct instance_attribute,attr)
+
+/* DEVICE instance kobject release() function */
+static void edac_device_ctrl_instance_release(struct kobject *kobj)
+{
+ struct edac_device_instance *instance;
+
+ edac_dbg(1, "\n");
+
+ /* map from this kobj to the main control struct
+ * and then dec the main kobj count
+ */
+ instance = to_instance(kobj);
+ kobject_put(&instance->ctl->kobj);
+}
+
+/* instance specific attribute structure */
+struct instance_attribute {
+ struct attribute attr;
+ ssize_t(*show) (struct edac_device_instance *, char *);
+ ssize_t(*store) (struct edac_device_instance *, const char *, size_t);
+};
+
+/* Function to 'show' fields from the edac_dev 'instance' structure */
+static ssize_t edac_dev_instance_show(struct kobject *kobj,
+ struct attribute *attr, char *buffer)
+{
+ struct edac_device_instance *instance = to_instance(kobj);
+ struct instance_attribute *instance_attr = to_instance_attr(attr);
+
+ if (instance_attr->show)
+ return instance_attr->show(instance, buffer);
+ return -EIO;
+}
+
+/* Function to 'store' fields into the edac_dev 'instance' structure */
+static ssize_t edac_dev_instance_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
+{
+ struct edac_device_instance *instance = to_instance(kobj);
+ struct instance_attribute *instance_attr = to_instance_attr(attr);
+
+ if (instance_attr->store)
+ return instance_attr->store(instance, buffer, count);
+ return -EIO;
+}
+
+/* edac_dev file operations for an 'instance' */
+static const struct sysfs_ops device_instance_ops = {
+ .show = edac_dev_instance_show,
+ .store = edac_dev_instance_store
+};
+
+#define INSTANCE_ATTR(_name,_mode,_show,_store) \
+static struct instance_attribute attr_instance_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .show = _show, \
+ .store = _store, \
+};
+
+/*
+ * Define attributes visible for the edac_device instance object
+ * Each contains a pointer to a show and an optional set
+ * function pointer that does the low level output/input
+ */
+INSTANCE_ATTR(ce_count, S_IRUGO, instance_ce_count_show, NULL);
+INSTANCE_ATTR(ue_count, S_IRUGO, instance_ue_count_show, NULL);
+
+/* list of edac_dev 'instance' attributes */
+static struct attribute *device_instance_attrs[] = {
+ &attr_instance_ce_count.attr,
+ &attr_instance_ue_count.attr,
+ NULL,
+};
+ATTRIBUTE_GROUPS(device_instance);
+
+/* The 'ktype' for each edac_dev 'instance' */
+static struct kobj_type ktype_instance_ctrl = {
+ .release = edac_device_ctrl_instance_release,
+ .sysfs_ops = &device_instance_ops,
+ .default_groups = device_instance_groups,
+};
+
+/* edac_dev -> instance -> block information */
+
+#define to_block(k) container_of(k, struct edac_device_block, kobj)
+#define to_block_attr(a) \
+ container_of(a, struct edac_dev_sysfs_block_attribute, attr)
+
+/*
+ * Set of low-level block attribute show functions
+ */
+static ssize_t block_ue_count_show(struct kobject *kobj,
+ struct attribute *attr, char *data)
+{
+ struct edac_device_block *block = to_block(kobj);
+
+ return sprintf(data, "%u\n", block->counters.ue_count);
+}
+
+static ssize_t block_ce_count_show(struct kobject *kobj,
+ struct attribute *attr, char *data)
+{
+ struct edac_device_block *block = to_block(kobj);
+
+ return sprintf(data, "%u\n", block->counters.ce_count);
+}
+
+/* DEVICE block kobject release() function */
+static void edac_device_ctrl_block_release(struct kobject *kobj)
+{
+ struct edac_device_block *block;
+
+ edac_dbg(1, "\n");
+
+ /* get the container of the kobj */
+ block = to_block(kobj);
+
+ /* map from 'block kobj' to 'block->instance->controller->main_kobj'
+ * now 'release' the block kobject
+ */
+ kobject_put(&block->instance->ctl->kobj);
+}
+
+
+/* Function to 'show' fields from the edac_dev 'block' structure */
+static ssize_t edac_dev_block_show(struct kobject *kobj,
+ struct attribute *attr, char *buffer)
+{
+ struct edac_dev_sysfs_block_attribute *block_attr =
+ to_block_attr(attr);
+
+ if (block_attr->show)
+ return block_attr->show(kobj, attr, buffer);
+ return -EIO;
+}
+
+/* Function to 'store' fields into the edac_dev 'block' structure */
+static ssize_t edac_dev_block_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
+{
+ struct edac_dev_sysfs_block_attribute *block_attr;
+
+ block_attr = to_block_attr(attr);
+
+ if (block_attr->store)
+ return block_attr->store(kobj, attr, buffer, count);
+ return -EIO;
+}
+
+/* edac_dev file operations for a 'block' */
+static const struct sysfs_ops device_block_ops = {
+ .show = edac_dev_block_show,
+ .store = edac_dev_block_store
+};
+
+#define BLOCK_ATTR(_name,_mode,_show,_store) \
+static struct edac_dev_sysfs_block_attribute attr_block_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .show = _show, \
+ .store = _store, \
+};
+
+BLOCK_ATTR(ce_count, S_IRUGO, block_ce_count_show, NULL);
+BLOCK_ATTR(ue_count, S_IRUGO, block_ue_count_show, NULL);
+
+/* list of edac_dev 'block' attributes */
+static struct attribute *device_block_attrs[] = {
+ &attr_block_ce_count.attr,
+ &attr_block_ue_count.attr,
+ NULL,
+};
+ATTRIBUTE_GROUPS(device_block);
+
+/* The 'ktype' for each edac_dev 'block' */
+static struct kobj_type ktype_block_ctrl = {
+ .release = edac_device_ctrl_block_release,
+ .sysfs_ops = &device_block_ops,
+ .default_groups = device_block_groups,
+};
+
+/* block ctor/dtor code */
+
+/*
+ * edac_device_create_block
+ */
+static int edac_device_create_block(struct edac_device_ctl_info *edac_dev,
+ struct edac_device_instance *instance,
+ struct edac_device_block *block)
+{
+ int i;
+ int err;
+ struct edac_dev_sysfs_block_attribute *sysfs_attrib;
+ struct kobject *main_kobj;
+
+ edac_dbg(4, "Instance '%s' inst_p=%p block '%s' block_p=%p\n",
+ instance->name, instance, block->name, block);
+ edac_dbg(4, "block kobj=%p block kobj->parent=%p\n",
+ &block->kobj, &block->kobj.parent);
+
+ /* init this block's kobject */
+ memset(&block->kobj, 0, sizeof(struct kobject));
+
+ /* bump the main kobject's reference count for this controller
+ * and this instance is dependent on the main
+ */
+ main_kobj = kobject_get(&edac_dev->kobj);
+ if (!main_kobj) {
+ err = -ENODEV;
+ goto err_out;
+ }
+
+ /* Add this block's kobject */
+ err = kobject_init_and_add(&block->kobj, &ktype_block_ctrl,
+ &instance->kobj,
+ "%s", block->name);
+ if (err) {
+ edac_dbg(1, "Failed to register instance '%s'\n", block->name);
+ kobject_put(main_kobj);
+ err = -ENODEV;
+ goto err_out;
+ }
+
+ /* If there are driver level block attributes, then added them
+ * to the block kobject
+ */
+ sysfs_attrib = block->block_attributes;
+ if (sysfs_attrib && block->nr_attribs) {
+ for (i = 0; i < block->nr_attribs; i++, sysfs_attrib++) {
+
+ edac_dbg(4, "creating block attrib='%s' attrib->%p to kobj=%p\n",
+ sysfs_attrib->attr.name,
+ sysfs_attrib, &block->kobj);
+
+ /* Create each block_attribute file */
+ err = sysfs_create_file(&block->kobj,
+ &sysfs_attrib->attr);
+ if (err)
+ goto err_on_attrib;
+ }
+ }
+ kobject_uevent(&block->kobj, KOBJ_ADD);
+
+ return 0;
+
+ /* Error unwind stack */
+err_on_attrib:
+ kobject_put(&block->kobj);
+
+err_out:
+ return err;
+}
+
+/*
+ * edac_device_delete_block(edac_dev,block);
+ */
+static void edac_device_delete_block(struct edac_device_ctl_info *edac_dev,
+ struct edac_device_block *block)
+{
+ struct edac_dev_sysfs_block_attribute *sysfs_attrib;
+ int i;
+
+ /* if this block has 'attributes' then we need to iterate over the list
+ * and 'remove' the attributes on this block
+ */
+ sysfs_attrib = block->block_attributes;
+ if (sysfs_attrib && block->nr_attribs) {
+ for (i = 0; i < block->nr_attribs; i++, sysfs_attrib++) {
+
+ /* remove each block_attrib file */
+ sysfs_remove_file(&block->kobj,
+ (struct attribute *) sysfs_attrib);
+ }
+ }
+
+ /* unregister this block's kobject, SEE:
+ * edac_device_ctrl_block_release() callback operation
+ */
+ kobject_put(&block->kobj);
+}
+
+/* instance ctor/dtor code */
+
+/*
+ * edac_device_create_instance
+ * create just one instance of an edac_device 'instance'
+ */
+static int edac_device_create_instance(struct edac_device_ctl_info *edac_dev,
+ int idx)
+{
+ int i, j;
+ int err;
+ struct edac_device_instance *instance;
+ struct kobject *main_kobj;
+
+ instance = &edac_dev->instances[idx];
+
+ /* Init the instance's kobject */
+ memset(&instance->kobj, 0, sizeof(struct kobject));
+
+ instance->ctl = edac_dev;
+
+ /* bump the main kobject's reference count for this controller
+ * and this instance is dependent on the main
+ */
+ main_kobj = kobject_get(&edac_dev->kobj);
+ if (!main_kobj) {
+ err = -ENODEV;
+ goto err_out;
+ }
+
+ /* Formally register this instance's kobject under the edac_device */
+ err = kobject_init_and_add(&instance->kobj, &ktype_instance_ctrl,
+ &edac_dev->kobj, "%s", instance->name);
+ if (err != 0) {
+ edac_dbg(2, "Failed to register instance '%s'\n",
+ instance->name);
+ kobject_put(main_kobj);
+ goto err_out;
+ }
+
+ edac_dbg(4, "now register '%d' blocks for instance %d\n",
+ instance->nr_blocks, idx);
+
+ /* register all blocks of this instance */
+ for (i = 0; i < instance->nr_blocks; i++) {
+ err = edac_device_create_block(edac_dev, instance,
+ &instance->blocks[i]);
+ if (err) {
+ /* If any fail, remove all previous ones */
+ for (j = 0; j < i; j++)
+ edac_device_delete_block(edac_dev,
+ &instance->blocks[j]);
+ goto err_release_instance_kobj;
+ }
+ }
+ kobject_uevent(&instance->kobj, KOBJ_ADD);
+
+ edac_dbg(4, "Registered instance %d '%s' kobject\n",
+ idx, instance->name);
+
+ return 0;
+
+ /* error unwind stack */
+err_release_instance_kobj:
+ kobject_put(&instance->kobj);
+
+err_out:
+ return err;
+}
+
+/*
+ * edac_device_remove_instance
+ * remove an edac_device instance
+ */
+static void edac_device_delete_instance(struct edac_device_ctl_info *edac_dev,
+ int idx)
+{
+ struct edac_device_instance *instance;
+ int i;
+
+ instance = &edac_dev->instances[idx];
+
+ /* unregister all blocks in this instance */
+ for (i = 0; i < instance->nr_blocks; i++)
+ edac_device_delete_block(edac_dev, &instance->blocks[i]);
+
+ /* unregister this instance's kobject, SEE:
+ * edac_device_ctrl_instance_release() for callback operation
+ */
+ kobject_put(&instance->kobj);
+}
+
+/*
+ * edac_device_create_instances
+ * create the first level of 'instances' for this device
+ * (ie 'cache' might have 'cache0', 'cache1', 'cache2', etc
+ */
+static int edac_device_create_instances(struct edac_device_ctl_info *edac_dev)
+{
+ int i, j;
+ int err;
+
+ edac_dbg(0, "\n");
+
+ /* iterate over creation of the instances */
+ for (i = 0; i < edac_dev->nr_instances; i++) {
+ err = edac_device_create_instance(edac_dev, i);
+ if (err) {
+ /* unwind previous instances on error */
+ for (j = 0; j < i; j++)
+ edac_device_delete_instance(edac_dev, j);
+ return err;
+ }
+ }
+
+ return 0;
+}
+
+/*
+ * edac_device_delete_instances(edac_dev);
+ * unregister all the kobjects of the instances
+ */
+static void edac_device_delete_instances(struct edac_device_ctl_info *edac_dev)
+{
+ int i;
+
+ /* iterate over creation of the instances */
+ for (i = 0; i < edac_dev->nr_instances; i++)
+ edac_device_delete_instance(edac_dev, i);
+}
+
+/* edac_dev sysfs ctor/dtor code */
+
+/*
+ * edac_device_add_main_sysfs_attributes
+ * add some attributes to this instance's main kobject
+ */
+static int edac_device_add_main_sysfs_attributes(
+ struct edac_device_ctl_info *edac_dev)
+{
+ struct edac_dev_sysfs_attribute *sysfs_attrib;
+ int err = 0;
+
+ sysfs_attrib = edac_dev->sysfs_attributes;
+ if (sysfs_attrib) {
+ /* iterate over the array and create an attribute for each
+ * entry in the list
+ */
+ while (sysfs_attrib->attr.name != NULL) {
+ err = sysfs_create_file(&edac_dev->kobj,
+ (struct attribute*) sysfs_attrib);
+ if (err)
+ goto err_out;
+
+ sysfs_attrib++;
+ }
+ }
+
+err_out:
+ return err;
+}
+
+/*
+ * edac_device_remove_main_sysfs_attributes
+ * remove any attributes to this instance's main kobject
+ */
+static void edac_device_remove_main_sysfs_attributes(
+ struct edac_device_ctl_info *edac_dev)
+{
+ struct edac_dev_sysfs_attribute *sysfs_attrib;
+
+ /* if there are main attributes, defined, remove them. First,
+ * point to the start of the array and iterate over it
+ * removing each attribute listed from this device's instance's kobject
+ */
+ sysfs_attrib = edac_dev->sysfs_attributes;
+ if (sysfs_attrib) {
+ while (sysfs_attrib->attr.name != NULL) {
+ sysfs_remove_file(&edac_dev->kobj,
+ (struct attribute *) sysfs_attrib);
+ sysfs_attrib++;
+ }
+ }
+}
+
+/*
+ * edac_device_create_sysfs() Constructor
+ *
+ * accept a created edac_device control structure
+ * and 'export' it to sysfs. The 'main' kobj should already have been
+ * created. 'instance' and 'block' kobjects should be registered
+ * along with any 'block' attributes from the low driver. In addition,
+ * the main attributes (if any) are connected to the main kobject of
+ * the control structure.
+ *
+ * Return:
+ * 0 Success
+ * !0 Failure
+ */
+int edac_device_create_sysfs(struct edac_device_ctl_info *edac_dev)
+{
+ int err;
+ struct kobject *edac_kobj = &edac_dev->kobj;
+
+ edac_dbg(0, "idx=%d\n", edac_dev->dev_idx);
+
+ /* go create any main attributes callers wants */
+ err = edac_device_add_main_sysfs_attributes(edac_dev);
+ if (err) {
+ edac_dbg(0, "failed to add sysfs attribs\n");
+ goto err_out;
+ }
+
+ /* create a symlink from the edac device
+ * to the platform 'device' being used for this
+ */
+ err = sysfs_create_link(edac_kobj,
+ &edac_dev->dev->kobj, EDAC_DEVICE_SYMLINK);
+ if (err) {
+ edac_dbg(0, "sysfs_create_link() returned err= %d\n", err);
+ goto err_remove_main_attribs;
+ }
+
+ /* Create the first level instance directories
+ * In turn, the nested blocks beneath the instances will
+ * be registered as well
+ */
+ err = edac_device_create_instances(edac_dev);
+ if (err) {
+ edac_dbg(0, "edac_device_create_instances() returned err= %d\n",
+ err);
+ goto err_remove_link;
+ }
+
+
+ edac_dbg(4, "create-instances done, idx=%d\n", edac_dev->dev_idx);
+
+ return 0;
+
+ /* Error unwind stack */
+err_remove_link:
+ /* remove the sym link */
+ sysfs_remove_link(&edac_dev->kobj, EDAC_DEVICE_SYMLINK);
+
+err_remove_main_attribs:
+ edac_device_remove_main_sysfs_attributes(edac_dev);
+
+err_out:
+ return err;
+}
+
+/*
+ * edac_device_remove_sysfs() destructor
+ *
+ * given an edac_device struct, tear down the kobject resources
+ */
+void edac_device_remove_sysfs(struct edac_device_ctl_info *edac_dev)
+{
+ edac_dbg(0, "\n");
+
+ /* remove any main attributes for this device */
+ edac_device_remove_main_sysfs_attributes(edac_dev);
+
+ /* remove the device sym link */
+ sysfs_remove_link(&edac_dev->kobj, EDAC_DEVICE_SYMLINK);
+
+ /* walk the instance/block kobject tree, deconstructing it */
+ edac_device_delete_instances(edac_dev);
+}
diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c
new file mode 100644
index 0000000000..6faeb2ab39
--- /dev/null
+++ b/drivers/edac/edac_mc.c
@@ -0,0 +1,1091 @@
+/*
+ * edac_mc kernel module
+ * (C) 2005, 2006 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * Modified by Dave Peterson and Doug Thompson
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/proc_fs.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/smp.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/highmem.h>
+#include <linux/timer.h>
+#include <linux/slab.h>
+#include <linux/jiffies.h>
+#include <linux/spinlock.h>
+#include <linux/list.h>
+#include <linux/ctype.h>
+#include <linux/edac.h>
+#include <linux/bitops.h>
+#include <linux/uaccess.h>
+#include <asm/page.h>
+#include "edac_mc.h"
+#include "edac_module.h"
+#include <ras/ras_event.h>
+
+#ifdef CONFIG_EDAC_ATOMIC_SCRUB
+#include <asm/edac.h>
+#else
+#define edac_atomic_scrub(va, size) do { } while (0)
+#endif
+
+int edac_op_state = EDAC_OPSTATE_INVAL;
+EXPORT_SYMBOL_GPL(edac_op_state);
+
+/* lock to memory controller's control array */
+static DEFINE_MUTEX(mem_ctls_mutex);
+static LIST_HEAD(mc_devices);
+
+/*
+ * Used to lock EDAC MC to just one module, avoiding two drivers e. g.
+ * apei/ghes and i7core_edac to be used at the same time.
+ */
+static const char *edac_mc_owner;
+
+static struct mem_ctl_info *error_desc_to_mci(struct edac_raw_error_desc *e)
+{
+ return container_of(e, struct mem_ctl_info, error_desc);
+}
+
+unsigned int edac_dimm_info_location(struct dimm_info *dimm, char *buf,
+ unsigned int len)
+{
+ struct mem_ctl_info *mci = dimm->mci;
+ int i, n, count = 0;
+ char *p = buf;
+
+ for (i = 0; i < mci->n_layers; i++) {
+ n = scnprintf(p, len, "%s %d ",
+ edac_layer_name[mci->layers[i].type],
+ dimm->location[i]);
+ p += n;
+ len -= n;
+ count += n;
+ }
+
+ return count;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+
+static void edac_mc_dump_channel(struct rank_info *chan)
+{
+ edac_dbg(4, " channel->chan_idx = %d\n", chan->chan_idx);
+ edac_dbg(4, " channel = %p\n", chan);
+ edac_dbg(4, " channel->csrow = %p\n", chan->csrow);
+ edac_dbg(4, " channel->dimm = %p\n", chan->dimm);
+}
+
+static void edac_mc_dump_dimm(struct dimm_info *dimm)
+{
+ char location[80];
+
+ if (!dimm->nr_pages)
+ return;
+
+ edac_dimm_info_location(dimm, location, sizeof(location));
+
+ edac_dbg(4, "%s%i: %smapped as virtual row %d, chan %d\n",
+ dimm->mci->csbased ? "rank" : "dimm",
+ dimm->idx, location, dimm->csrow, dimm->cschannel);
+ edac_dbg(4, " dimm = %p\n", dimm);
+ edac_dbg(4, " dimm->label = '%s'\n", dimm->label);
+ edac_dbg(4, " dimm->nr_pages = 0x%x\n", dimm->nr_pages);
+ edac_dbg(4, " dimm->grain = %d\n", dimm->grain);
+}
+
+static void edac_mc_dump_csrow(struct csrow_info *csrow)
+{
+ edac_dbg(4, "csrow->csrow_idx = %d\n", csrow->csrow_idx);
+ edac_dbg(4, " csrow = %p\n", csrow);
+ edac_dbg(4, " csrow->first_page = 0x%lx\n", csrow->first_page);
+ edac_dbg(4, " csrow->last_page = 0x%lx\n", csrow->last_page);
+ edac_dbg(4, " csrow->page_mask = 0x%lx\n", csrow->page_mask);
+ edac_dbg(4, " csrow->nr_channels = %d\n", csrow->nr_channels);
+ edac_dbg(4, " csrow->channels = %p\n", csrow->channels);
+ edac_dbg(4, " csrow->mci = %p\n", csrow->mci);
+}
+
+static void edac_mc_dump_mci(struct mem_ctl_info *mci)
+{
+ edac_dbg(3, "\tmci = %p\n", mci);
+ edac_dbg(3, "\tmci->mtype_cap = %lx\n", mci->mtype_cap);
+ edac_dbg(3, "\tmci->edac_ctl_cap = %lx\n", mci->edac_ctl_cap);
+ edac_dbg(3, "\tmci->edac_cap = %lx\n", mci->edac_cap);
+ edac_dbg(4, "\tmci->edac_check = %p\n", mci->edac_check);
+ edac_dbg(3, "\tmci->nr_csrows = %d, csrows = %p\n",
+ mci->nr_csrows, mci->csrows);
+ edac_dbg(3, "\tmci->nr_dimms = %d, dimms = %p\n",
+ mci->tot_dimms, mci->dimms);
+ edac_dbg(3, "\tdev = %p\n", mci->pdev);
+ edac_dbg(3, "\tmod_name:ctl_name = %s:%s\n",
+ mci->mod_name, mci->ctl_name);
+ edac_dbg(3, "\tpvt_info = %p\n\n", mci->pvt_info);
+}
+
+#endif /* CONFIG_EDAC_DEBUG */
+
+const char * const edac_mem_types[] = {
+ [MEM_EMPTY] = "Empty",
+ [MEM_RESERVED] = "Reserved",
+ [MEM_UNKNOWN] = "Unknown",
+ [MEM_FPM] = "FPM",
+ [MEM_EDO] = "EDO",
+ [MEM_BEDO] = "BEDO",
+ [MEM_SDR] = "Unbuffered-SDR",
+ [MEM_RDR] = "Registered-SDR",
+ [MEM_DDR] = "Unbuffered-DDR",
+ [MEM_RDDR] = "Registered-DDR",
+ [MEM_RMBS] = "RMBS",
+ [MEM_DDR2] = "Unbuffered-DDR2",
+ [MEM_FB_DDR2] = "FullyBuffered-DDR2",
+ [MEM_RDDR2] = "Registered-DDR2",
+ [MEM_XDR] = "XDR",
+ [MEM_DDR3] = "Unbuffered-DDR3",
+ [MEM_RDDR3] = "Registered-DDR3",
+ [MEM_LRDDR3] = "Load-Reduced-DDR3-RAM",
+ [MEM_LPDDR3] = "Low-Power-DDR3-RAM",
+ [MEM_DDR4] = "Unbuffered-DDR4",
+ [MEM_RDDR4] = "Registered-DDR4",
+ [MEM_LPDDR4] = "Low-Power-DDR4-RAM",
+ [MEM_LRDDR4] = "Load-Reduced-DDR4-RAM",
+ [MEM_DDR5] = "Unbuffered-DDR5",
+ [MEM_RDDR5] = "Registered-DDR5",
+ [MEM_LRDDR5] = "Load-Reduced-DDR5-RAM",
+ [MEM_NVDIMM] = "Non-volatile-RAM",
+ [MEM_WIO2] = "Wide-IO-2",
+ [MEM_HBM2] = "High-bandwidth-memory-Gen2",
+};
+EXPORT_SYMBOL_GPL(edac_mem_types);
+
+static void _edac_mc_free(struct mem_ctl_info *mci)
+{
+ put_device(&mci->dev);
+}
+
+static void mci_release(struct device *dev)
+{
+ struct mem_ctl_info *mci = container_of(dev, struct mem_ctl_info, dev);
+ struct csrow_info *csr;
+ int i, chn, row;
+
+ if (mci->dimms) {
+ for (i = 0; i < mci->tot_dimms; i++)
+ kfree(mci->dimms[i]);
+ kfree(mci->dimms);
+ }
+
+ if (mci->csrows) {
+ for (row = 0; row < mci->nr_csrows; row++) {
+ csr = mci->csrows[row];
+ if (!csr)
+ continue;
+
+ if (csr->channels) {
+ for (chn = 0; chn < mci->num_cschannel; chn++)
+ kfree(csr->channels[chn]);
+ kfree(csr->channels);
+ }
+ kfree(csr);
+ }
+ kfree(mci->csrows);
+ }
+ kfree(mci->pvt_info);
+ kfree(mci->layers);
+ kfree(mci);
+}
+
+static int edac_mc_alloc_csrows(struct mem_ctl_info *mci)
+{
+ unsigned int tot_channels = mci->num_cschannel;
+ unsigned int tot_csrows = mci->nr_csrows;
+ unsigned int row, chn;
+
+ /*
+ * Alocate and fill the csrow/channels structs
+ */
+ mci->csrows = kcalloc(tot_csrows, sizeof(*mci->csrows), GFP_KERNEL);
+ if (!mci->csrows)
+ return -ENOMEM;
+
+ for (row = 0; row < tot_csrows; row++) {
+ struct csrow_info *csr;
+
+ csr = kzalloc(sizeof(**mci->csrows), GFP_KERNEL);
+ if (!csr)
+ return -ENOMEM;
+
+ mci->csrows[row] = csr;
+ csr->csrow_idx = row;
+ csr->mci = mci;
+ csr->nr_channels = tot_channels;
+ csr->channels = kcalloc(tot_channels, sizeof(*csr->channels),
+ GFP_KERNEL);
+ if (!csr->channels)
+ return -ENOMEM;
+
+ for (chn = 0; chn < tot_channels; chn++) {
+ struct rank_info *chan;
+
+ chan = kzalloc(sizeof(**csr->channels), GFP_KERNEL);
+ if (!chan)
+ return -ENOMEM;
+
+ csr->channels[chn] = chan;
+ chan->chan_idx = chn;
+ chan->csrow = csr;
+ }
+ }
+
+ return 0;
+}
+
+static int edac_mc_alloc_dimms(struct mem_ctl_info *mci)
+{
+ unsigned int pos[EDAC_MAX_LAYERS];
+ unsigned int row, chn, idx;
+ int layer;
+ void *p;
+
+ /*
+ * Allocate and fill the dimm structs
+ */
+ mci->dimms = kcalloc(mci->tot_dimms, sizeof(*mci->dimms), GFP_KERNEL);
+ if (!mci->dimms)
+ return -ENOMEM;
+
+ memset(&pos, 0, sizeof(pos));
+ row = 0;
+ chn = 0;
+ for (idx = 0; idx < mci->tot_dimms; idx++) {
+ struct dimm_info *dimm;
+ struct rank_info *chan;
+ int n, len;
+
+ chan = mci->csrows[row]->channels[chn];
+
+ dimm = kzalloc(sizeof(**mci->dimms), GFP_KERNEL);
+ if (!dimm)
+ return -ENOMEM;
+ mci->dimms[idx] = dimm;
+ dimm->mci = mci;
+ dimm->idx = idx;
+
+ /*
+ * Copy DIMM location and initialize it.
+ */
+ len = sizeof(dimm->label);
+ p = dimm->label;
+ n = scnprintf(p, len, "mc#%u", mci->mc_idx);
+ p += n;
+ len -= n;
+ for (layer = 0; layer < mci->n_layers; layer++) {
+ n = scnprintf(p, len, "%s#%u",
+ edac_layer_name[mci->layers[layer].type],
+ pos[layer]);
+ p += n;
+ len -= n;
+ dimm->location[layer] = pos[layer];
+ }
+
+ /* Link it to the csrows old API data */
+ chan->dimm = dimm;
+ dimm->csrow = row;
+ dimm->cschannel = chn;
+
+ /* Increment csrow location */
+ if (mci->layers[0].is_virt_csrow) {
+ chn++;
+ if (chn == mci->num_cschannel) {
+ chn = 0;
+ row++;
+ }
+ } else {
+ row++;
+ if (row == mci->nr_csrows) {
+ row = 0;
+ chn++;
+ }
+ }
+
+ /* Increment dimm location */
+ for (layer = mci->n_layers - 1; layer >= 0; layer--) {
+ pos[layer]++;
+ if (pos[layer] < mci->layers[layer].size)
+ break;
+ pos[layer] = 0;
+ }
+ }
+
+ return 0;
+}
+
+struct mem_ctl_info *edac_mc_alloc(unsigned int mc_num,
+ unsigned int n_layers,
+ struct edac_mc_layer *layers,
+ unsigned int sz_pvt)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer *layer;
+ unsigned int idx, tot_dimms = 1;
+ unsigned int tot_csrows = 1, tot_channels = 1;
+ bool per_rank = false;
+
+ if (WARN_ON(n_layers > EDAC_MAX_LAYERS || n_layers == 0))
+ return NULL;
+
+ /*
+ * Calculate the total amount of dimms and csrows/cschannels while
+ * in the old API emulation mode
+ */
+ for (idx = 0; idx < n_layers; idx++) {
+ tot_dimms *= layers[idx].size;
+
+ if (layers[idx].is_virt_csrow)
+ tot_csrows *= layers[idx].size;
+ else
+ tot_channels *= layers[idx].size;
+
+ if (layers[idx].type == EDAC_MC_LAYER_CHIP_SELECT)
+ per_rank = true;
+ }
+
+ mci = kzalloc(sizeof(struct mem_ctl_info), GFP_KERNEL);
+ if (!mci)
+ return NULL;
+
+ mci->layers = kcalloc(n_layers, sizeof(struct edac_mc_layer), GFP_KERNEL);
+ if (!mci->layers)
+ goto error;
+
+ mci->pvt_info = kzalloc(sz_pvt, GFP_KERNEL);
+ if (!mci->pvt_info)
+ goto error;
+
+ mci->dev.release = mci_release;
+ device_initialize(&mci->dev);
+
+ /* setup index and various internal pointers */
+ mci->mc_idx = mc_num;
+ mci->tot_dimms = tot_dimms;
+ mci->n_layers = n_layers;
+ memcpy(mci->layers, layers, sizeof(*layer) * n_layers);
+ mci->nr_csrows = tot_csrows;
+ mci->num_cschannel = tot_channels;
+ mci->csbased = per_rank;
+
+ if (edac_mc_alloc_csrows(mci))
+ goto error;
+
+ if (edac_mc_alloc_dimms(mci))
+ goto error;
+
+ mci->op_state = OP_ALLOC;
+
+ return mci;
+
+error:
+ _edac_mc_free(mci);
+
+ return NULL;
+}
+EXPORT_SYMBOL_GPL(edac_mc_alloc);
+
+void edac_mc_free(struct mem_ctl_info *mci)
+{
+ edac_dbg(1, "\n");
+
+ _edac_mc_free(mci);
+}
+EXPORT_SYMBOL_GPL(edac_mc_free);
+
+bool edac_has_mcs(void)
+{
+ bool ret;
+
+ mutex_lock(&mem_ctls_mutex);
+
+ ret = list_empty(&mc_devices);
+
+ mutex_unlock(&mem_ctls_mutex);
+
+ return !ret;
+}
+EXPORT_SYMBOL_GPL(edac_has_mcs);
+
+/* Caller must hold mem_ctls_mutex */
+static struct mem_ctl_info *__find_mci_by_dev(struct device *dev)
+{
+ struct mem_ctl_info *mci;
+ struct list_head *item;
+
+ edac_dbg(3, "\n");
+
+ list_for_each(item, &mc_devices) {
+ mci = list_entry(item, struct mem_ctl_info, link);
+
+ if (mci->pdev == dev)
+ return mci;
+ }
+
+ return NULL;
+}
+
+/**
+ * find_mci_by_dev
+ *
+ * scan list of controllers looking for the one that manages
+ * the 'dev' device
+ * @dev: pointer to a struct device related with the MCI
+ */
+struct mem_ctl_info *find_mci_by_dev(struct device *dev)
+{
+ struct mem_ctl_info *ret;
+
+ mutex_lock(&mem_ctls_mutex);
+ ret = __find_mci_by_dev(dev);
+ mutex_unlock(&mem_ctls_mutex);
+
+ return ret;
+}
+EXPORT_SYMBOL_GPL(find_mci_by_dev);
+
+/*
+ * edac_mc_workq_function
+ * performs the operation scheduled by a workq request
+ */
+static void edac_mc_workq_function(struct work_struct *work_req)
+{
+ struct delayed_work *d_work = to_delayed_work(work_req);
+ struct mem_ctl_info *mci = to_edac_mem_ctl_work(d_work);
+
+ mutex_lock(&mem_ctls_mutex);
+
+ if (mci->op_state != OP_RUNNING_POLL) {
+ mutex_unlock(&mem_ctls_mutex);
+ return;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ mci->edac_check(mci);
+
+ mutex_unlock(&mem_ctls_mutex);
+
+ /* Queue ourselves again. */
+ edac_queue_work(&mci->work, msecs_to_jiffies(edac_mc_get_poll_msec()));
+}
+
+/*
+ * edac_mc_reset_delay_period(unsigned long value)
+ *
+ * user space has updated our poll period value, need to
+ * reset our workq delays
+ */
+void edac_mc_reset_delay_period(unsigned long value)
+{
+ struct mem_ctl_info *mci;
+ struct list_head *item;
+
+ mutex_lock(&mem_ctls_mutex);
+
+ list_for_each(item, &mc_devices) {
+ mci = list_entry(item, struct mem_ctl_info, link);
+
+ if (mci->op_state == OP_RUNNING_POLL)
+ edac_mod_work(&mci->work, value);
+ }
+ mutex_unlock(&mem_ctls_mutex);
+}
+
+
+
+/* Return 0 on success, 1 on failure.
+ * Before calling this function, caller must
+ * assign a unique value to mci->mc_idx.
+ *
+ * locking model:
+ *
+ * called with the mem_ctls_mutex lock held
+ */
+static int add_mc_to_global_list(struct mem_ctl_info *mci)
+{
+ struct list_head *item, *insert_before;
+ struct mem_ctl_info *p;
+
+ insert_before = &mc_devices;
+
+ p = __find_mci_by_dev(mci->pdev);
+ if (unlikely(p != NULL))
+ goto fail0;
+
+ list_for_each(item, &mc_devices) {
+ p = list_entry(item, struct mem_ctl_info, link);
+
+ if (p->mc_idx >= mci->mc_idx) {
+ if (unlikely(p->mc_idx == mci->mc_idx))
+ goto fail1;
+
+ insert_before = item;
+ break;
+ }
+ }
+
+ list_add_tail_rcu(&mci->link, insert_before);
+ return 0;
+
+fail0:
+ edac_printk(KERN_WARNING, EDAC_MC,
+ "%s (%s) %s %s already assigned %d\n", dev_name(p->pdev),
+ edac_dev_name(mci), p->mod_name, p->ctl_name, p->mc_idx);
+ return 1;
+
+fail1:
+ edac_printk(KERN_WARNING, EDAC_MC,
+ "bug in low-level driver: attempt to assign\n"
+ " duplicate mc_idx %d in %s()\n", p->mc_idx, __func__);
+ return 1;
+}
+
+static int del_mc_from_global_list(struct mem_ctl_info *mci)
+{
+ list_del_rcu(&mci->link);
+
+ /* these are for safe removal of devices from global list while
+ * NMI handlers may be traversing list
+ */
+ synchronize_rcu();
+ INIT_LIST_HEAD(&mci->link);
+
+ return list_empty(&mc_devices);
+}
+
+struct mem_ctl_info *edac_mc_find(int idx)
+{
+ struct mem_ctl_info *mci;
+ struct list_head *item;
+
+ mutex_lock(&mem_ctls_mutex);
+
+ list_for_each(item, &mc_devices) {
+ mci = list_entry(item, struct mem_ctl_info, link);
+ if (mci->mc_idx == idx)
+ goto unlock;
+ }
+
+ mci = NULL;
+unlock:
+ mutex_unlock(&mem_ctls_mutex);
+ return mci;
+}
+EXPORT_SYMBOL(edac_mc_find);
+
+const char *edac_get_owner(void)
+{
+ return edac_mc_owner;
+}
+EXPORT_SYMBOL_GPL(edac_get_owner);
+
+/* FIXME - should a warning be printed if no error detection? correction? */
+int edac_mc_add_mc_with_groups(struct mem_ctl_info *mci,
+ const struct attribute_group **groups)
+{
+ int ret = -EINVAL;
+ edac_dbg(0, "\n");
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (edac_debug_level >= 3)
+ edac_mc_dump_mci(mci);
+
+ if (edac_debug_level >= 4) {
+ struct dimm_info *dimm;
+ int i;
+
+ for (i = 0; i < mci->nr_csrows; i++) {
+ struct csrow_info *csrow = mci->csrows[i];
+ u32 nr_pages = 0;
+ int j;
+
+ for (j = 0; j < csrow->nr_channels; j++)
+ nr_pages += csrow->channels[j]->dimm->nr_pages;
+ if (!nr_pages)
+ continue;
+ edac_mc_dump_csrow(csrow);
+ for (j = 0; j < csrow->nr_channels; j++)
+ if (csrow->channels[j]->dimm->nr_pages)
+ edac_mc_dump_channel(csrow->channels[j]);
+ }
+
+ mci_for_each_dimm(mci, dimm)
+ edac_mc_dump_dimm(dimm);
+ }
+#endif
+ mutex_lock(&mem_ctls_mutex);
+
+ if (edac_mc_owner && edac_mc_owner != mci->mod_name) {
+ ret = -EPERM;
+ goto fail0;
+ }
+
+ if (add_mc_to_global_list(mci))
+ goto fail0;
+
+ /* set load time so that error rate can be tracked */
+ mci->start_time = jiffies;
+
+ mci->bus = edac_get_sysfs_subsys();
+
+ if (edac_create_sysfs_mci_device(mci, groups)) {
+ edac_mc_printk(mci, KERN_WARNING,
+ "failed to create sysfs device\n");
+ goto fail1;
+ }
+
+ if (mci->edac_check) {
+ mci->op_state = OP_RUNNING_POLL;
+
+ INIT_DELAYED_WORK(&mci->work, edac_mc_workq_function);
+ edac_queue_work(&mci->work, msecs_to_jiffies(edac_mc_get_poll_msec()));
+
+ } else {
+ mci->op_state = OP_RUNNING_INTERRUPT;
+ }
+
+ /* Report action taken */
+ edac_mc_printk(mci, KERN_INFO,
+ "Giving out device to module %s controller %s: DEV %s (%s)\n",
+ mci->mod_name, mci->ctl_name, mci->dev_name,
+ edac_op_state_to_string(mci->op_state));
+
+ edac_mc_owner = mci->mod_name;
+
+ mutex_unlock(&mem_ctls_mutex);
+ return 0;
+
+fail1:
+ del_mc_from_global_list(mci);
+
+fail0:
+ mutex_unlock(&mem_ctls_mutex);
+ return ret;
+}
+EXPORT_SYMBOL_GPL(edac_mc_add_mc_with_groups);
+
+struct mem_ctl_info *edac_mc_del_mc(struct device *dev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ mutex_lock(&mem_ctls_mutex);
+
+ /* find the requested mci struct in the global list */
+ mci = __find_mci_by_dev(dev);
+ if (mci == NULL) {
+ mutex_unlock(&mem_ctls_mutex);
+ return NULL;
+ }
+
+ /* mark MCI offline: */
+ mci->op_state = OP_OFFLINE;
+
+ if (del_mc_from_global_list(mci))
+ edac_mc_owner = NULL;
+
+ mutex_unlock(&mem_ctls_mutex);
+
+ if (mci->edac_check)
+ edac_stop_work(&mci->work);
+
+ /* remove from sysfs */
+ edac_remove_sysfs_mci_device(mci);
+
+ edac_printk(KERN_INFO, EDAC_MC,
+ "Removed device %d for %s %s: DEV %s\n", mci->mc_idx,
+ mci->mod_name, mci->ctl_name, edac_dev_name(mci));
+
+ return mci;
+}
+EXPORT_SYMBOL_GPL(edac_mc_del_mc);
+
+static void edac_mc_scrub_block(unsigned long page, unsigned long offset,
+ u32 size)
+{
+ struct page *pg;
+ void *virt_addr;
+ unsigned long flags = 0;
+
+ edac_dbg(3, "\n");
+
+ /* ECC error page was not in our memory. Ignore it. */
+ if (!pfn_valid(page))
+ return;
+
+ /* Find the actual page structure then map it and fix */
+ pg = pfn_to_page(page);
+
+ if (PageHighMem(pg))
+ local_irq_save(flags);
+
+ virt_addr = kmap_atomic(pg);
+
+ /* Perform architecture specific atomic scrub operation */
+ edac_atomic_scrub(virt_addr + offset, size);
+
+ /* Unmap and complete */
+ kunmap_atomic(virt_addr);
+
+ if (PageHighMem(pg))
+ local_irq_restore(flags);
+}
+
+/* FIXME - should return -1 */
+int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci, unsigned long page)
+{
+ struct csrow_info **csrows = mci->csrows;
+ int row, i, j, n;
+
+ edac_dbg(1, "MC%d: 0x%lx\n", mci->mc_idx, page);
+ row = -1;
+
+ for (i = 0; i < mci->nr_csrows; i++) {
+ struct csrow_info *csrow = csrows[i];
+ n = 0;
+ for (j = 0; j < csrow->nr_channels; j++) {
+ struct dimm_info *dimm = csrow->channels[j]->dimm;
+ n += dimm->nr_pages;
+ }
+ if (n == 0)
+ continue;
+
+ edac_dbg(3, "MC%d: first(0x%lx) page(0x%lx) last(0x%lx) mask(0x%lx)\n",
+ mci->mc_idx,
+ csrow->first_page, page, csrow->last_page,
+ csrow->page_mask);
+
+ if ((page >= csrow->first_page) &&
+ (page <= csrow->last_page) &&
+ ((page & csrow->page_mask) ==
+ (csrow->first_page & csrow->page_mask))) {
+ row = i;
+ break;
+ }
+ }
+
+ if (row == -1)
+ edac_mc_printk(mci, KERN_ERR,
+ "could not look up page error address %lx\n",
+ (unsigned long)page);
+
+ return row;
+}
+EXPORT_SYMBOL_GPL(edac_mc_find_csrow_by_page);
+
+const char *edac_layer_name[] = {
+ [EDAC_MC_LAYER_BRANCH] = "branch",
+ [EDAC_MC_LAYER_CHANNEL] = "channel",
+ [EDAC_MC_LAYER_SLOT] = "slot",
+ [EDAC_MC_LAYER_CHIP_SELECT] = "csrow",
+ [EDAC_MC_LAYER_ALL_MEM] = "memory",
+};
+EXPORT_SYMBOL_GPL(edac_layer_name);
+
+static void edac_inc_ce_error(struct edac_raw_error_desc *e)
+{
+ int pos[EDAC_MAX_LAYERS] = { e->top_layer, e->mid_layer, e->low_layer };
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+ struct dimm_info *dimm = edac_get_dimm(mci, pos[0], pos[1], pos[2]);
+
+ mci->ce_mc += e->error_count;
+
+ if (dimm)
+ dimm->ce_count += e->error_count;
+ else
+ mci->ce_noinfo_count += e->error_count;
+}
+
+static void edac_inc_ue_error(struct edac_raw_error_desc *e)
+{
+ int pos[EDAC_MAX_LAYERS] = { e->top_layer, e->mid_layer, e->low_layer };
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+ struct dimm_info *dimm = edac_get_dimm(mci, pos[0], pos[1], pos[2]);
+
+ mci->ue_mc += e->error_count;
+
+ if (dimm)
+ dimm->ue_count += e->error_count;
+ else
+ mci->ue_noinfo_count += e->error_count;
+}
+
+static void edac_ce_error(struct edac_raw_error_desc *e)
+{
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+ unsigned long remapped_page;
+
+ if (edac_mc_get_log_ce()) {
+ edac_mc_printk(mci, KERN_WARNING,
+ "%d CE %s%son %s (%s page:0x%lx offset:0x%lx grain:%ld syndrome:0x%lx%s%s)\n",
+ e->error_count, e->msg,
+ *e->msg ? " " : "",
+ e->label, e->location, e->page_frame_number, e->offset_in_page,
+ e->grain, e->syndrome,
+ *e->other_detail ? " - " : "",
+ e->other_detail);
+ }
+
+ edac_inc_ce_error(e);
+
+ if (mci->scrub_mode == SCRUB_SW_SRC) {
+ /*
+ * Some memory controllers (called MCs below) can remap
+ * memory so that it is still available at a different
+ * address when PCI devices map into memory.
+ * MC's that can't do this, lose the memory where PCI
+ * devices are mapped. This mapping is MC-dependent
+ * and so we call back into the MC driver for it to
+ * map the MC page to a physical (CPU) page which can
+ * then be mapped to a virtual page - which can then
+ * be scrubbed.
+ */
+ remapped_page = mci->ctl_page_to_phys ?
+ mci->ctl_page_to_phys(mci, e->page_frame_number) :
+ e->page_frame_number;
+
+ edac_mc_scrub_block(remapped_page, e->offset_in_page, e->grain);
+ }
+}
+
+static void edac_ue_error(struct edac_raw_error_desc *e)
+{
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+
+ if (edac_mc_get_log_ue()) {
+ edac_mc_printk(mci, KERN_WARNING,
+ "%d UE %s%son %s (%s page:0x%lx offset:0x%lx grain:%ld%s%s)\n",
+ e->error_count, e->msg,
+ *e->msg ? " " : "",
+ e->label, e->location, e->page_frame_number, e->offset_in_page,
+ e->grain,
+ *e->other_detail ? " - " : "",
+ e->other_detail);
+ }
+
+ edac_inc_ue_error(e);
+
+ if (edac_mc_get_panic_on_ue()) {
+ panic("UE %s%son %s (%s page:0x%lx offset:0x%lx grain:%ld%s%s)\n",
+ e->msg,
+ *e->msg ? " " : "",
+ e->label, e->location, e->page_frame_number, e->offset_in_page,
+ e->grain,
+ *e->other_detail ? " - " : "",
+ e->other_detail);
+ }
+}
+
+static void edac_inc_csrow(struct edac_raw_error_desc *e, int row, int chan)
+{
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+ enum hw_event_mc_err_type type = e->type;
+ u16 count = e->error_count;
+
+ if (row < 0)
+ return;
+
+ edac_dbg(4, "csrow/channel to increment: (%d,%d)\n", row, chan);
+
+ if (type == HW_EVENT_ERR_CORRECTED) {
+ mci->csrows[row]->ce_count += count;
+ if (chan >= 0)
+ mci->csrows[row]->channels[chan]->ce_count += count;
+ } else {
+ mci->csrows[row]->ue_count += count;
+ }
+}
+
+void edac_raw_mc_handle_error(struct edac_raw_error_desc *e)
+{
+ struct mem_ctl_info *mci = error_desc_to_mci(e);
+ u8 grain_bits;
+
+ /* Sanity-check driver-supplied grain value. */
+ if (WARN_ON_ONCE(!e->grain))
+ e->grain = 1;
+
+ grain_bits = fls_long(e->grain - 1);
+
+ /* Report the error via the trace interface */
+ if (IS_ENABLED(CONFIG_RAS))
+ trace_mc_event(e->type, e->msg, e->label, e->error_count,
+ mci->mc_idx, e->top_layer, e->mid_layer,
+ e->low_layer,
+ (e->page_frame_number << PAGE_SHIFT) | e->offset_in_page,
+ grain_bits, e->syndrome, e->other_detail);
+
+ if (e->type == HW_EVENT_ERR_CORRECTED)
+ edac_ce_error(e);
+ else
+ edac_ue_error(e);
+}
+EXPORT_SYMBOL_GPL(edac_raw_mc_handle_error);
+
+void edac_mc_handle_error(const enum hw_event_mc_err_type type,
+ struct mem_ctl_info *mci,
+ const u16 error_count,
+ const unsigned long page_frame_number,
+ const unsigned long offset_in_page,
+ const unsigned long syndrome,
+ const int top_layer,
+ const int mid_layer,
+ const int low_layer,
+ const char *msg,
+ const char *other_detail)
+{
+ struct dimm_info *dimm;
+ char *p, *end;
+ int row = -1, chan = -1;
+ int pos[EDAC_MAX_LAYERS] = { top_layer, mid_layer, low_layer };
+ int i, n_labels = 0;
+ struct edac_raw_error_desc *e = &mci->error_desc;
+ bool any_memory = true;
+ const char *prefix;
+
+ edac_dbg(3, "MC%d\n", mci->mc_idx);
+
+ /* Fills the error report buffer */
+ memset(e, 0, sizeof (*e));
+ e->error_count = error_count;
+ e->type = type;
+ e->top_layer = top_layer;
+ e->mid_layer = mid_layer;
+ e->low_layer = low_layer;
+ e->page_frame_number = page_frame_number;
+ e->offset_in_page = offset_in_page;
+ e->syndrome = syndrome;
+ /* need valid strings here for both: */
+ e->msg = msg ?: "";
+ e->other_detail = other_detail ?: "";
+
+ /*
+ * Check if the event report is consistent and if the memory location is
+ * known. If it is, the DIMM(s) label info will be filled and the DIMM's
+ * error counters will be incremented.
+ */
+ for (i = 0; i < mci->n_layers; i++) {
+ if (pos[i] >= (int)mci->layers[i].size) {
+
+ edac_mc_printk(mci, KERN_ERR,
+ "INTERNAL ERROR: %s value is out of range (%d >= %d)\n",
+ edac_layer_name[mci->layers[i].type],
+ pos[i], mci->layers[i].size);
+ /*
+ * Instead of just returning it, let's use what's
+ * known about the error. The increment routines and
+ * the DIMM filter logic will do the right thing by
+ * pointing the likely damaged DIMMs.
+ */
+ pos[i] = -1;
+ }
+ if (pos[i] >= 0)
+ any_memory = false;
+ }
+
+ /*
+ * Get the dimm label/grain that applies to the match criteria.
+ * As the error algorithm may not be able to point to just one memory
+ * stick, the logic here will get all possible labels that could
+ * pottentially be affected by the error.
+ * On FB-DIMM memory controllers, for uncorrected errors, it is common
+ * to have only the MC channel and the MC dimm (also called "branch")
+ * but the channel is not known, as the memory is arranged in pairs,
+ * where each memory belongs to a separate channel within the same
+ * branch.
+ */
+ p = e->label;
+ *p = '\0';
+ end = p + sizeof(e->label);
+ prefix = "";
+
+ mci_for_each_dimm(mci, dimm) {
+ if (top_layer >= 0 && top_layer != dimm->location[0])
+ continue;
+ if (mid_layer >= 0 && mid_layer != dimm->location[1])
+ continue;
+ if (low_layer >= 0 && low_layer != dimm->location[2])
+ continue;
+
+ /* get the max grain, over the error match range */
+ if (dimm->grain > e->grain)
+ e->grain = dimm->grain;
+
+ /*
+ * If the error is memory-controller wide, there's no need to
+ * seek for the affected DIMMs because the whole channel/memory
+ * controller/... may be affected. Also, don't show errors for
+ * empty DIMM slots.
+ */
+ if (!dimm->nr_pages)
+ continue;
+
+ n_labels++;
+ if (n_labels > EDAC_MAX_LABELS) {
+ p = e->label;
+ *p = '\0';
+ } else {
+ p += scnprintf(p, end - p, "%s%s", prefix, dimm->label);
+ prefix = OTHER_LABEL;
+ }
+
+ /*
+ * get csrow/channel of the DIMM, in order to allow
+ * incrementing the compat API counters
+ */
+ edac_dbg(4, "%s csrows map: (%d,%d)\n",
+ mci->csbased ? "rank" : "dimm",
+ dimm->csrow, dimm->cschannel);
+ if (row == -1)
+ row = dimm->csrow;
+ else if (row >= 0 && row != dimm->csrow)
+ row = -2;
+
+ if (chan == -1)
+ chan = dimm->cschannel;
+ else if (chan >= 0 && chan != dimm->cschannel)
+ chan = -2;
+ }
+
+ if (any_memory)
+ strscpy(e->label, "any memory", sizeof(e->label));
+ else if (!*e->label)
+ strscpy(e->label, "unknown memory", sizeof(e->label));
+
+ edac_inc_csrow(e, row, chan);
+
+ /* Fill the RAM location data */
+ p = e->location;
+ end = p + sizeof(e->location);
+ prefix = "";
+
+ for (i = 0; i < mci->n_layers; i++) {
+ if (pos[i] < 0)
+ continue;
+
+ p += scnprintf(p, end - p, "%s%s:%d", prefix,
+ edac_layer_name[mci->layers[i].type], pos[i]);
+ prefix = " ";
+ }
+
+ edac_raw_mc_handle_error(e);
+}
+EXPORT_SYMBOL_GPL(edac_mc_handle_error);
diff --git a/drivers/edac/edac_mc.h b/drivers/edac/edac_mc.h
new file mode 100644
index 0000000000..881b00eadf
--- /dev/null
+++ b/drivers/edac/edac_mc.h
@@ -0,0 +1,258 @@
+/*
+ * Defines, structures, APIs for edac_mc module
+ *
+ * (C) 2007 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * NMI handling support added by
+ * Dave Peterson <dsp@llnl.gov> <dave_peterson@pobox.com>
+ *
+ * Refactored for multi-source files:
+ * Doug Thompson <norsk5@xmission.com>
+ *
+ * Please look at Documentation/driver-api/edac.rst for more info about
+ * EDAC core structs and functions.
+ */
+
+#ifndef _EDAC_MC_H_
+#define _EDAC_MC_H_
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/smp.h>
+#include <linux/pci.h>
+#include <linux/time.h>
+#include <linux/nmi.h>
+#include <linux/rcupdate.h>
+#include <linux/completion.h>
+#include <linux/kobject.h>
+#include <linux/platform_device.h>
+#include <linux/workqueue.h>
+#include <linux/edac.h>
+
+#if PAGE_SHIFT < 20
+#define PAGES_TO_MiB(pages) ((pages) >> (20 - PAGE_SHIFT))
+#define MiB_TO_PAGES(mb) ((mb) << (20 - PAGE_SHIFT))
+#else /* PAGE_SHIFT > 20 */
+#define PAGES_TO_MiB(pages) ((pages) << (PAGE_SHIFT - 20))
+#define MiB_TO_PAGES(mb) ((mb) >> (PAGE_SHIFT - 20))
+#endif
+
+#define edac_printk(level, prefix, fmt, arg...) \
+ printk(level "EDAC " prefix ": " fmt, ##arg)
+
+#define edac_mc_printk(mci, level, fmt, arg...) \
+ printk(level "EDAC MC%d: " fmt, mci->mc_idx, ##arg)
+
+#define edac_mc_chipset_printk(mci, level, prefix, fmt, arg...) \
+ printk(level "EDAC " prefix " MC%d: " fmt, mci->mc_idx, ##arg)
+
+#define edac_device_printk(ctl, level, fmt, arg...) \
+ printk(level "EDAC DEVICE%d: " fmt, ctl->dev_idx, ##arg)
+
+#define edac_pci_printk(ctl, level, fmt, arg...) \
+ printk(level "EDAC PCI%d: " fmt, ctl->pci_idx, ##arg)
+
+/* prefixes for edac_printk() and edac_mc_printk() */
+#define EDAC_MC "MC"
+#define EDAC_PCI "PCI"
+#define EDAC_DEBUG "DEBUG"
+
+extern const char * const edac_mem_types[];
+
+#ifdef CONFIG_EDAC_DEBUG
+extern int edac_debug_level;
+
+#define edac_dbg(level, fmt, ...) \
+do { \
+ if (level <= edac_debug_level) \
+ edac_printk(KERN_DEBUG, EDAC_DEBUG, \
+ "%s: " fmt, __func__, ##__VA_ARGS__); \
+} while (0)
+
+#else /* !CONFIG_EDAC_DEBUG */
+
+#define edac_dbg(level, fmt, ...) \
+do { \
+ if (0) \
+ edac_printk(KERN_DEBUG, EDAC_DEBUG, \
+ "%s: " fmt, __func__, ##__VA_ARGS__); \
+} while (0)
+
+#endif /* !CONFIG_EDAC_DEBUG */
+
+#define PCI_VEND_DEV(vend, dev) PCI_VENDOR_ID_ ## vend, \
+ PCI_DEVICE_ID_ ## vend ## _ ## dev
+
+#define edac_dev_name(dev) (dev)->dev_name
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+/**
+ * edac_mc_alloc() - Allocate and partially fill a struct &mem_ctl_info.
+ *
+ * @mc_num: Memory controller number
+ * @n_layers: Number of MC hierarchy layers
+ * @layers: Describes each layer as seen by the Memory Controller
+ * @sz_pvt: size of private storage needed
+ *
+ *
+ * Everything is kmalloc'ed as one big chunk - more efficient.
+ * Only can be used if all structures have the same lifetime - otherwise
+ * you have to allocate and initialize your own structures.
+ *
+ * Use edac_mc_free() to free mc structures allocated by this function.
+ *
+ * .. note::
+ *
+ * drivers handle multi-rank memories in different ways: in some
+ * drivers, one multi-rank memory stick is mapped as one entry, while, in
+ * others, a single multi-rank memory stick would be mapped into several
+ * entries. Currently, this function will allocate multiple struct dimm_info
+ * on such scenarios, as grouping the multiple ranks require drivers change.
+ *
+ * Returns:
+ * On success, return a pointer to struct mem_ctl_info pointer;
+ * %NULL otherwise
+ */
+struct mem_ctl_info *edac_mc_alloc(unsigned int mc_num,
+ unsigned int n_layers,
+ struct edac_mc_layer *layers,
+ unsigned int sz_pvt);
+
+/**
+ * edac_get_owner - Return the owner's mod_name of EDAC MC
+ *
+ * Returns:
+ * Pointer to mod_name string when EDAC MC is owned. NULL otherwise.
+ */
+extern const char *edac_get_owner(void);
+
+/*
+ * edac_mc_add_mc_with_groups() - Insert the @mci structure into the mci
+ * global list and create sysfs entries associated with @mci structure.
+ *
+ * @mci: pointer to the mci structure to be added to the list
+ * @groups: optional attribute groups for the driver-specific sysfs entries
+ *
+ * Returns:
+ * 0 on Success, or an error code on failure
+ */
+extern int edac_mc_add_mc_with_groups(struct mem_ctl_info *mci,
+ const struct attribute_group **groups);
+#define edac_mc_add_mc(mci) edac_mc_add_mc_with_groups(mci, NULL)
+
+/**
+ * edac_mc_free() - Frees a previously allocated @mci structure
+ *
+ * @mci: pointer to a struct mem_ctl_info structure
+ */
+extern void edac_mc_free(struct mem_ctl_info *mci);
+
+/**
+ * edac_has_mcs() - Check if any MCs have been allocated.
+ *
+ * Returns:
+ * True if MC instances have been registered successfully.
+ * False otherwise.
+ */
+extern bool edac_has_mcs(void);
+
+/**
+ * edac_mc_find() - Search for a mem_ctl_info structure whose index is @idx.
+ *
+ * @idx: index to be seek
+ *
+ * If found, return a pointer to the structure.
+ * Else return NULL.
+ */
+extern struct mem_ctl_info *edac_mc_find(int idx);
+
+/**
+ * find_mci_by_dev() - Scan list of controllers looking for the one that
+ * manages the @dev device.
+ *
+ * @dev: pointer to a struct device related with the MCI
+ *
+ * Returns: on success, returns a pointer to struct &mem_ctl_info;
+ * %NULL otherwise.
+ */
+extern struct mem_ctl_info *find_mci_by_dev(struct device *dev);
+
+/**
+ * edac_mc_del_mc() - Remove sysfs entries for mci structure associated with
+ * @dev and remove mci structure from global list.
+ *
+ * @dev: Pointer to struct &device representing mci structure to remove.
+ *
+ * Returns: pointer to removed mci structure, or %NULL if device not found.
+ */
+extern struct mem_ctl_info *edac_mc_del_mc(struct device *dev);
+
+/**
+ * edac_mc_find_csrow_by_page() - Ancillary routine to identify what csrow
+ * contains a memory page.
+ *
+ * @mci: pointer to a struct mem_ctl_info structure
+ * @page: memory page to find
+ *
+ * Returns: on success, returns the csrow. -1 if not found.
+ */
+extern int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci,
+ unsigned long page);
+
+/**
+ * edac_raw_mc_handle_error() - Reports a memory event to userspace without
+ * doing anything to discover the error location.
+ *
+ * @e: error description
+ *
+ * This raw function is used internally by edac_mc_handle_error(). It should
+ * only be called directly when the hardware error come directly from BIOS,
+ * like in the case of APEI GHES driver.
+ */
+void edac_raw_mc_handle_error(struct edac_raw_error_desc *e);
+
+/**
+ * edac_mc_handle_error() - Reports a memory event to userspace.
+ *
+ * @type: severity of the error (CE/UE/Fatal)
+ * @mci: a struct mem_ctl_info pointer
+ * @error_count: Number of errors of the same type
+ * @page_frame_number: mem page where the error occurred
+ * @offset_in_page: offset of the error inside the page
+ * @syndrome: ECC syndrome
+ * @top_layer: Memory layer[0] position
+ * @mid_layer: Memory layer[1] position
+ * @low_layer: Memory layer[2] position
+ * @msg: Message meaningful to the end users that
+ * explains the event
+ * @other_detail: Technical details about the event that
+ * may help hardware manufacturers and
+ * EDAC developers to analyse the event
+ */
+void edac_mc_handle_error(const enum hw_event_mc_err_type type,
+ struct mem_ctl_info *mci,
+ const u16 error_count,
+ const unsigned long page_frame_number,
+ const unsigned long offset_in_page,
+ const unsigned long syndrome,
+ const int top_layer,
+ const int mid_layer,
+ const int low_layer,
+ const char *msg,
+ const char *other_detail);
+
+/*
+ * edac misc APIs
+ */
+extern char *edac_op_state_to_string(int op_state);
+
+#endif /* _EDAC_MC_H_ */
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
new file mode 100644
index 0000000000..15f63452a9
--- /dev/null
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -0,0 +1,1055 @@
+/*
+ * edac_mc kernel module
+ * (C) 2005-2007 Linux Networx (http://lnxi.com)
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written Doug Thompson <norsk5@xmission.com> www.softwarebitmaker.com
+ *
+ * (c) 2012-2013 - Mauro Carvalho Chehab
+ * The entire API were re-written, and ported to use struct device
+ *
+ */
+
+#include <linux/ctype.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <linux/bug.h>
+#include <linux/pm_runtime.h>
+#include <linux/uaccess.h>
+
+#include "edac_mc.h"
+#include "edac_module.h"
+
+/* MC EDAC Controls, setable by module parameter, and sysfs */
+static int edac_mc_log_ue = 1;
+static int edac_mc_log_ce = 1;
+static int edac_mc_panic_on_ue;
+static unsigned int edac_mc_poll_msec = 1000;
+
+/* Getter functions for above */
+int edac_mc_get_log_ue(void)
+{
+ return edac_mc_log_ue;
+}
+
+int edac_mc_get_log_ce(void)
+{
+ return edac_mc_log_ce;
+}
+
+int edac_mc_get_panic_on_ue(void)
+{
+ return edac_mc_panic_on_ue;
+}
+
+/* this is temporary */
+unsigned int edac_mc_get_poll_msec(void)
+{
+ return edac_mc_poll_msec;
+}
+
+static int edac_set_poll_msec(const char *val, const struct kernel_param *kp)
+{
+ unsigned int i;
+ int ret;
+
+ if (!val)
+ return -EINVAL;
+
+ ret = kstrtouint(val, 0, &i);
+ if (ret)
+ return ret;
+
+ if (i < 1000)
+ return -EINVAL;
+
+ *((unsigned int *)kp->arg) = i;
+
+ /* notify edac_mc engine to reset the poll period */
+ edac_mc_reset_delay_period(i);
+
+ return 0;
+}
+
+/* Parameter declarations for above */
+module_param(edac_mc_panic_on_ue, int, 0644);
+MODULE_PARM_DESC(edac_mc_panic_on_ue, "Panic on uncorrected error: 0=off 1=on");
+module_param(edac_mc_log_ue, int, 0644);
+MODULE_PARM_DESC(edac_mc_log_ue,
+ "Log uncorrectable error to console: 0=off 1=on");
+module_param(edac_mc_log_ce, int, 0644);
+MODULE_PARM_DESC(edac_mc_log_ce,
+ "Log correctable error to console: 0=off 1=on");
+module_param_call(edac_mc_poll_msec, edac_set_poll_msec, param_get_uint,
+ &edac_mc_poll_msec, 0644);
+MODULE_PARM_DESC(edac_mc_poll_msec, "Polling period in milliseconds");
+
+static struct device *mci_pdev;
+
+/*
+ * various constants for Memory Controllers
+ */
+static const char * const dev_types[] = {
+ [DEV_UNKNOWN] = "Unknown",
+ [DEV_X1] = "x1",
+ [DEV_X2] = "x2",
+ [DEV_X4] = "x4",
+ [DEV_X8] = "x8",
+ [DEV_X16] = "x16",
+ [DEV_X32] = "x32",
+ [DEV_X64] = "x64"
+};
+
+static const char * const edac_caps[] = {
+ [EDAC_UNKNOWN] = "Unknown",
+ [EDAC_NONE] = "None",
+ [EDAC_RESERVED] = "Reserved",
+ [EDAC_PARITY] = "PARITY",
+ [EDAC_EC] = "EC",
+ [EDAC_SECDED] = "SECDED",
+ [EDAC_S2ECD2ED] = "S2ECD2ED",
+ [EDAC_S4ECD4ED] = "S4ECD4ED",
+ [EDAC_S8ECD8ED] = "S8ECD8ED",
+ [EDAC_S16ECD16ED] = "S16ECD16ED"
+};
+
+#ifdef CONFIG_EDAC_LEGACY_SYSFS
+/*
+ * EDAC sysfs CSROW data structures and methods
+ */
+
+#define to_csrow(k) container_of(k, struct csrow_info, dev)
+
+/*
+ * We need it to avoid namespace conflicts between the legacy API
+ * and the per-dimm/per-rank one
+ */
+#define DEVICE_ATTR_LEGACY(_name, _mode, _show, _store) \
+ static struct device_attribute dev_attr_legacy_##_name = __ATTR(_name, _mode, _show, _store)
+
+struct dev_ch_attribute {
+ struct device_attribute attr;
+ unsigned int channel;
+};
+
+#define DEVICE_CHANNEL(_name, _mode, _show, _store, _var) \
+ static struct dev_ch_attribute dev_attr_legacy_##_name = \
+ { __ATTR(_name, _mode, _show, _store), (_var) }
+
+#define to_channel(k) (container_of(k, struct dev_ch_attribute, attr)->channel)
+
+/* Set of more default csrow<id> attribute show/store functions */
+static ssize_t csrow_ue_count_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+
+ return sprintf(data, "%u\n", csrow->ue_count);
+}
+
+static ssize_t csrow_ce_count_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+
+ return sprintf(data, "%u\n", csrow->ce_count);
+}
+
+static ssize_t csrow_size_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+ int i;
+ u32 nr_pages = 0;
+
+ for (i = 0; i < csrow->nr_channels; i++)
+ nr_pages += csrow->channels[i]->dimm->nr_pages;
+ return sprintf(data, "%u\n", PAGES_TO_MiB(nr_pages));
+}
+
+static ssize_t csrow_mem_type_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+
+ return sprintf(data, "%s\n", edac_mem_types[csrow->channels[0]->dimm->mtype]);
+}
+
+static ssize_t csrow_dev_type_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+
+ return sprintf(data, "%s\n", dev_types[csrow->channels[0]->dimm->dtype]);
+}
+
+static ssize_t csrow_edac_mode_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+
+ return sprintf(data, "%s\n", edac_caps[csrow->channels[0]->dimm->edac_mode]);
+}
+
+/* show/store functions for DIMM Label attributes */
+static ssize_t channel_dimm_label_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+ unsigned int chan = to_channel(mattr);
+ struct rank_info *rank = csrow->channels[chan];
+
+ /* if field has not been initialized, there is nothing to send */
+ if (!rank->dimm->label[0])
+ return 0;
+
+ return snprintf(data, sizeof(rank->dimm->label) + 1, "%s\n",
+ rank->dimm->label);
+}
+
+static ssize_t channel_dimm_label_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+ unsigned int chan = to_channel(mattr);
+ struct rank_info *rank = csrow->channels[chan];
+ size_t copy_count = count;
+
+ if (count == 0)
+ return -EINVAL;
+
+ if (data[count - 1] == '\0' || data[count - 1] == '\n')
+ copy_count -= 1;
+
+ if (copy_count == 0 || copy_count >= sizeof(rank->dimm->label))
+ return -EINVAL;
+
+ strncpy(rank->dimm->label, data, copy_count);
+ rank->dimm->label[copy_count] = '\0';
+
+ return count;
+}
+
+/* show function for dynamic chX_ce_count attribute */
+static ssize_t channel_ce_count_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct csrow_info *csrow = to_csrow(dev);
+ unsigned int chan = to_channel(mattr);
+ struct rank_info *rank = csrow->channels[chan];
+
+ return sprintf(data, "%u\n", rank->ce_count);
+}
+
+/* cwrow<id>/attribute files */
+DEVICE_ATTR_LEGACY(size_mb, S_IRUGO, csrow_size_show, NULL);
+DEVICE_ATTR_LEGACY(dev_type, S_IRUGO, csrow_dev_type_show, NULL);
+DEVICE_ATTR_LEGACY(mem_type, S_IRUGO, csrow_mem_type_show, NULL);
+DEVICE_ATTR_LEGACY(edac_mode, S_IRUGO, csrow_edac_mode_show, NULL);
+DEVICE_ATTR_LEGACY(ue_count, S_IRUGO, csrow_ue_count_show, NULL);
+DEVICE_ATTR_LEGACY(ce_count, S_IRUGO, csrow_ce_count_show, NULL);
+
+/* default attributes of the CSROW<id> object */
+static struct attribute *csrow_attrs[] = {
+ &dev_attr_legacy_dev_type.attr,
+ &dev_attr_legacy_mem_type.attr,
+ &dev_attr_legacy_edac_mode.attr,
+ &dev_attr_legacy_size_mb.attr,
+ &dev_attr_legacy_ue_count.attr,
+ &dev_attr_legacy_ce_count.attr,
+ NULL,
+};
+
+static const struct attribute_group csrow_attr_grp = {
+ .attrs = csrow_attrs,
+};
+
+static const struct attribute_group *csrow_attr_groups[] = {
+ &csrow_attr_grp,
+ NULL
+};
+
+static const struct device_type csrow_attr_type = {
+ .groups = csrow_attr_groups,
+};
+
+/*
+ * possible dynamic channel DIMM Label attribute files
+ *
+ */
+DEVICE_CHANNEL(ch0_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 0);
+DEVICE_CHANNEL(ch1_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 1);
+DEVICE_CHANNEL(ch2_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 2);
+DEVICE_CHANNEL(ch3_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 3);
+DEVICE_CHANNEL(ch4_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 4);
+DEVICE_CHANNEL(ch5_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 5);
+DEVICE_CHANNEL(ch6_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 6);
+DEVICE_CHANNEL(ch7_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 7);
+DEVICE_CHANNEL(ch8_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 8);
+DEVICE_CHANNEL(ch9_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 9);
+DEVICE_CHANNEL(ch10_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 10);
+DEVICE_CHANNEL(ch11_dimm_label, S_IRUGO | S_IWUSR,
+ channel_dimm_label_show, channel_dimm_label_store, 11);
+
+/* Total possible dynamic DIMM Label attribute file table */
+static struct attribute *dynamic_csrow_dimm_attr[] = {
+ &dev_attr_legacy_ch0_dimm_label.attr.attr,
+ &dev_attr_legacy_ch1_dimm_label.attr.attr,
+ &dev_attr_legacy_ch2_dimm_label.attr.attr,
+ &dev_attr_legacy_ch3_dimm_label.attr.attr,
+ &dev_attr_legacy_ch4_dimm_label.attr.attr,
+ &dev_attr_legacy_ch5_dimm_label.attr.attr,
+ &dev_attr_legacy_ch6_dimm_label.attr.attr,
+ &dev_attr_legacy_ch7_dimm_label.attr.attr,
+ &dev_attr_legacy_ch8_dimm_label.attr.attr,
+ &dev_attr_legacy_ch9_dimm_label.attr.attr,
+ &dev_attr_legacy_ch10_dimm_label.attr.attr,
+ &dev_attr_legacy_ch11_dimm_label.attr.attr,
+ NULL
+};
+
+/* possible dynamic channel ce_count attribute files */
+DEVICE_CHANNEL(ch0_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 0);
+DEVICE_CHANNEL(ch1_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 1);
+DEVICE_CHANNEL(ch2_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 2);
+DEVICE_CHANNEL(ch3_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 3);
+DEVICE_CHANNEL(ch4_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 4);
+DEVICE_CHANNEL(ch5_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 5);
+DEVICE_CHANNEL(ch6_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 6);
+DEVICE_CHANNEL(ch7_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 7);
+DEVICE_CHANNEL(ch8_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 8);
+DEVICE_CHANNEL(ch9_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 9);
+DEVICE_CHANNEL(ch10_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 10);
+DEVICE_CHANNEL(ch11_ce_count, S_IRUGO,
+ channel_ce_count_show, NULL, 11);
+
+/* Total possible dynamic ce_count attribute file table */
+static struct attribute *dynamic_csrow_ce_count_attr[] = {
+ &dev_attr_legacy_ch0_ce_count.attr.attr,
+ &dev_attr_legacy_ch1_ce_count.attr.attr,
+ &dev_attr_legacy_ch2_ce_count.attr.attr,
+ &dev_attr_legacy_ch3_ce_count.attr.attr,
+ &dev_attr_legacy_ch4_ce_count.attr.attr,
+ &dev_attr_legacy_ch5_ce_count.attr.attr,
+ &dev_attr_legacy_ch6_ce_count.attr.attr,
+ &dev_attr_legacy_ch7_ce_count.attr.attr,
+ &dev_attr_legacy_ch8_ce_count.attr.attr,
+ &dev_attr_legacy_ch9_ce_count.attr.attr,
+ &dev_attr_legacy_ch10_ce_count.attr.attr,
+ &dev_attr_legacy_ch11_ce_count.attr.attr,
+ NULL
+};
+
+static umode_t csrow_dev_is_visible(struct kobject *kobj,
+ struct attribute *attr, int idx)
+{
+ struct device *dev = kobj_to_dev(kobj);
+ struct csrow_info *csrow = container_of(dev, struct csrow_info, dev);
+
+ if (idx >= csrow->nr_channels)
+ return 0;
+
+ if (idx >= ARRAY_SIZE(dynamic_csrow_ce_count_attr) - 1) {
+ WARN_ONCE(1, "idx: %d\n", idx);
+ return 0;
+ }
+
+ /* Only expose populated DIMMs */
+ if (!csrow->channels[idx]->dimm->nr_pages)
+ return 0;
+
+ return attr->mode;
+}
+
+
+static const struct attribute_group csrow_dev_dimm_group = {
+ .attrs = dynamic_csrow_dimm_attr,
+ .is_visible = csrow_dev_is_visible,
+};
+
+static const struct attribute_group csrow_dev_ce_count_group = {
+ .attrs = dynamic_csrow_ce_count_attr,
+ .is_visible = csrow_dev_is_visible,
+};
+
+static const struct attribute_group *csrow_dev_groups[] = {
+ &csrow_dev_dimm_group,
+ &csrow_dev_ce_count_group,
+ NULL
+};
+
+static void csrow_release(struct device *dev)
+{
+ /*
+ * Nothing to do, just unregister sysfs here. The mci
+ * device owns the data and will also release it.
+ */
+}
+
+static inline int nr_pages_per_csrow(struct csrow_info *csrow)
+{
+ int chan, nr_pages = 0;
+
+ for (chan = 0; chan < csrow->nr_channels; chan++)
+ nr_pages += csrow->channels[chan]->dimm->nr_pages;
+
+ return nr_pages;
+}
+
+/* Create a CSROW object under specifed edac_mc_device */
+static int edac_create_csrow_object(struct mem_ctl_info *mci,
+ struct csrow_info *csrow, int index)
+{
+ int err;
+
+ csrow->dev.type = &csrow_attr_type;
+ csrow->dev.groups = csrow_dev_groups;
+ csrow->dev.release = csrow_release;
+ device_initialize(&csrow->dev);
+ csrow->dev.parent = &mci->dev;
+ csrow->mci = mci;
+ dev_set_name(&csrow->dev, "csrow%d", index);
+ dev_set_drvdata(&csrow->dev, csrow);
+
+ err = device_add(&csrow->dev);
+ if (err) {
+ edac_dbg(1, "failure: create device %s\n", dev_name(&csrow->dev));
+ put_device(&csrow->dev);
+ return err;
+ }
+
+ edac_dbg(0, "device %s created\n", dev_name(&csrow->dev));
+
+ return 0;
+}
+
+/* Create a CSROW object under specifed edac_mc_device */
+static int edac_create_csrow_objects(struct mem_ctl_info *mci)
+{
+ int err, i;
+ struct csrow_info *csrow;
+
+ for (i = 0; i < mci->nr_csrows; i++) {
+ csrow = mci->csrows[i];
+ if (!nr_pages_per_csrow(csrow))
+ continue;
+ err = edac_create_csrow_object(mci, mci->csrows[i], i);
+ if (err < 0)
+ goto error;
+ }
+ return 0;
+
+error:
+ for (--i; i >= 0; i--) {
+ if (device_is_registered(&mci->csrows[i]->dev))
+ device_unregister(&mci->csrows[i]->dev);
+ }
+
+ return err;
+}
+
+static void edac_delete_csrow_objects(struct mem_ctl_info *mci)
+{
+ int i;
+
+ for (i = 0; i < mci->nr_csrows; i++) {
+ if (device_is_registered(&mci->csrows[i]->dev))
+ device_unregister(&mci->csrows[i]->dev);
+ }
+}
+
+#endif
+
+/*
+ * Per-dimm (or per-rank) devices
+ */
+
+#define to_dimm(k) container_of(k, struct dimm_info, dev)
+
+/* show/store functions for DIMM Label attributes */
+static ssize_t dimmdev_location_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+ ssize_t count;
+
+ count = edac_dimm_info_location(dimm, data, PAGE_SIZE);
+ count += scnprintf(data + count, PAGE_SIZE - count, "\n");
+
+ return count;
+}
+
+static ssize_t dimmdev_label_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ /* if field has not been initialized, there is nothing to send */
+ if (!dimm->label[0])
+ return 0;
+
+ return snprintf(data, sizeof(dimm->label) + 1, "%s\n", dimm->label);
+}
+
+static ssize_t dimmdev_label_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data,
+ size_t count)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+ size_t copy_count = count;
+
+ if (count == 0)
+ return -EINVAL;
+
+ if (data[count - 1] == '\0' || data[count - 1] == '\n')
+ copy_count -= 1;
+
+ if (copy_count == 0 || copy_count >= sizeof(dimm->label))
+ return -EINVAL;
+
+ strncpy(dimm->label, data, copy_count);
+ dimm->label[copy_count] = '\0';
+
+ return count;
+}
+
+static ssize_t dimmdev_size_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%u\n", PAGES_TO_MiB(dimm->nr_pages));
+}
+
+static ssize_t dimmdev_mem_type_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%s\n", edac_mem_types[dimm->mtype]);
+}
+
+static ssize_t dimmdev_dev_type_show(struct device *dev,
+ struct device_attribute *mattr, char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%s\n", dev_types[dimm->dtype]);
+}
+
+static ssize_t dimmdev_edac_mode_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%s\n", edac_caps[dimm->edac_mode]);
+}
+
+static ssize_t dimmdev_ce_count_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%u\n", dimm->ce_count);
+}
+
+static ssize_t dimmdev_ue_count_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct dimm_info *dimm = to_dimm(dev);
+
+ return sprintf(data, "%u\n", dimm->ue_count);
+}
+
+/* dimm/rank attribute files */
+static DEVICE_ATTR(dimm_label, S_IRUGO | S_IWUSR,
+ dimmdev_label_show, dimmdev_label_store);
+static DEVICE_ATTR(dimm_location, S_IRUGO, dimmdev_location_show, NULL);
+static DEVICE_ATTR(size, S_IRUGO, dimmdev_size_show, NULL);
+static DEVICE_ATTR(dimm_mem_type, S_IRUGO, dimmdev_mem_type_show, NULL);
+static DEVICE_ATTR(dimm_dev_type, S_IRUGO, dimmdev_dev_type_show, NULL);
+static DEVICE_ATTR(dimm_edac_mode, S_IRUGO, dimmdev_edac_mode_show, NULL);
+static DEVICE_ATTR(dimm_ce_count, S_IRUGO, dimmdev_ce_count_show, NULL);
+static DEVICE_ATTR(dimm_ue_count, S_IRUGO, dimmdev_ue_count_show, NULL);
+
+/* attributes of the dimm<id>/rank<id> object */
+static struct attribute *dimm_attrs[] = {
+ &dev_attr_dimm_label.attr,
+ &dev_attr_dimm_location.attr,
+ &dev_attr_size.attr,
+ &dev_attr_dimm_mem_type.attr,
+ &dev_attr_dimm_dev_type.attr,
+ &dev_attr_dimm_edac_mode.attr,
+ &dev_attr_dimm_ce_count.attr,
+ &dev_attr_dimm_ue_count.attr,
+ NULL,
+};
+
+static const struct attribute_group dimm_attr_grp = {
+ .attrs = dimm_attrs,
+};
+
+static const struct attribute_group *dimm_attr_groups[] = {
+ &dimm_attr_grp,
+ NULL
+};
+
+static const struct device_type dimm_attr_type = {
+ .groups = dimm_attr_groups,
+};
+
+static void dimm_release(struct device *dev)
+{
+ /*
+ * Nothing to do, just unregister sysfs here. The mci
+ * device owns the data and will also release it.
+ */
+}
+
+/* Create a DIMM object under specifed memory controller device */
+static int edac_create_dimm_object(struct mem_ctl_info *mci,
+ struct dimm_info *dimm)
+{
+ int err;
+ dimm->mci = mci;
+
+ dimm->dev.type = &dimm_attr_type;
+ dimm->dev.release = dimm_release;
+ device_initialize(&dimm->dev);
+
+ dimm->dev.parent = &mci->dev;
+ if (mci->csbased)
+ dev_set_name(&dimm->dev, "rank%d", dimm->idx);
+ else
+ dev_set_name(&dimm->dev, "dimm%d", dimm->idx);
+ dev_set_drvdata(&dimm->dev, dimm);
+ pm_runtime_forbid(&mci->dev);
+
+ err = device_add(&dimm->dev);
+ if (err) {
+ edac_dbg(1, "failure: create device %s\n", dev_name(&dimm->dev));
+ put_device(&dimm->dev);
+ return err;
+ }
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG)) {
+ char location[80];
+
+ edac_dimm_info_location(dimm, location, sizeof(location));
+ edac_dbg(0, "device %s created at location %s\n",
+ dev_name(&dimm->dev), location);
+ }
+
+ return 0;
+}
+
+/*
+ * Memory controller device
+ */
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+static ssize_t mci_reset_counters_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct dimm_info *dimm;
+ int row, chan;
+
+ mci->ue_mc = 0;
+ mci->ce_mc = 0;
+ mci->ue_noinfo_count = 0;
+ mci->ce_noinfo_count = 0;
+
+ for (row = 0; row < mci->nr_csrows; row++) {
+ struct csrow_info *ri = mci->csrows[row];
+
+ ri->ue_count = 0;
+ ri->ce_count = 0;
+
+ for (chan = 0; chan < ri->nr_channels; chan++)
+ ri->channels[chan]->ce_count = 0;
+ }
+
+ mci_for_each_dimm(mci, dimm) {
+ dimm->ue_count = 0;
+ dimm->ce_count = 0;
+ }
+
+ mci->start_time = jiffies;
+ return count;
+}
+
+/* Memory scrubbing interface:
+ *
+ * A MC driver can limit the scrubbing bandwidth based on the CPU type.
+ * Therefore, ->set_sdram_scrub_rate should be made to return the actual
+ * bandwidth that is accepted or 0 when scrubbing is to be disabled.
+ *
+ * Negative value still means that an error has occurred while setting
+ * the scrub rate.
+ */
+static ssize_t mci_sdram_scrub_rate_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ unsigned long bandwidth = 0;
+ int new_bw = 0;
+
+ if (kstrtoul(data, 10, &bandwidth) < 0)
+ return -EINVAL;
+
+ new_bw = mci->set_sdram_scrub_rate(mci, bandwidth);
+ if (new_bw < 0) {
+ edac_printk(KERN_WARNING, EDAC_MC,
+ "Error setting scrub rate to: %lu\n", bandwidth);
+ return -EINVAL;
+ }
+
+ return count;
+}
+
+/*
+ * ->get_sdram_scrub_rate() return value semantics same as above.
+ */
+static ssize_t mci_sdram_scrub_rate_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ int bandwidth = 0;
+
+ bandwidth = mci->get_sdram_scrub_rate(mci);
+ if (bandwidth < 0) {
+ edac_printk(KERN_DEBUG, EDAC_MC, "Error reading scrub rate\n");
+ return bandwidth;
+ }
+
+ return sprintf(data, "%d\n", bandwidth);
+}
+
+/* default attribute files for the MCI object */
+static ssize_t mci_ue_count_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%u\n", mci->ue_mc);
+}
+
+static ssize_t mci_ce_count_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%u\n", mci->ce_mc);
+}
+
+static ssize_t mci_ce_noinfo_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%u\n", mci->ce_noinfo_count);
+}
+
+static ssize_t mci_ue_noinfo_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%u\n", mci->ue_noinfo_count);
+}
+
+static ssize_t mci_seconds_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%ld\n", (jiffies - mci->start_time) / HZ);
+}
+
+static ssize_t mci_ctl_name_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ return sprintf(data, "%s\n", mci->ctl_name);
+}
+
+static ssize_t mci_size_mb_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ int total_pages = 0, csrow_idx, j;
+
+ for (csrow_idx = 0; csrow_idx < mci->nr_csrows; csrow_idx++) {
+ struct csrow_info *csrow = mci->csrows[csrow_idx];
+
+ for (j = 0; j < csrow->nr_channels; j++) {
+ struct dimm_info *dimm = csrow->channels[j]->dimm;
+
+ total_pages += dimm->nr_pages;
+ }
+ }
+
+ return sprintf(data, "%u\n", PAGES_TO_MiB(total_pages));
+}
+
+static ssize_t mci_max_location_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ int len = PAGE_SIZE;
+ char *p = data;
+ int i, n;
+
+ for (i = 0; i < mci->n_layers; i++) {
+ n = scnprintf(p, len, "%s %d ",
+ edac_layer_name[mci->layers[i].type],
+ mci->layers[i].size - 1);
+ len -= n;
+ if (len <= 0)
+ goto out;
+
+ p += n;
+ }
+
+ p += scnprintf(p, len, "\n");
+out:
+ return p - data;
+}
+
+/* default Control file */
+static DEVICE_ATTR(reset_counters, S_IWUSR, NULL, mci_reset_counters_store);
+
+/* default Attribute files */
+static DEVICE_ATTR(mc_name, S_IRUGO, mci_ctl_name_show, NULL);
+static DEVICE_ATTR(size_mb, S_IRUGO, mci_size_mb_show, NULL);
+static DEVICE_ATTR(seconds_since_reset, S_IRUGO, mci_seconds_show, NULL);
+static DEVICE_ATTR(ue_noinfo_count, S_IRUGO, mci_ue_noinfo_show, NULL);
+static DEVICE_ATTR(ce_noinfo_count, S_IRUGO, mci_ce_noinfo_show, NULL);
+static DEVICE_ATTR(ue_count, S_IRUGO, mci_ue_count_show, NULL);
+static DEVICE_ATTR(ce_count, S_IRUGO, mci_ce_count_show, NULL);
+static DEVICE_ATTR(max_location, S_IRUGO, mci_max_location_show, NULL);
+
+/* memory scrubber attribute file */
+static DEVICE_ATTR(sdram_scrub_rate, 0, mci_sdram_scrub_rate_show,
+ mci_sdram_scrub_rate_store); /* umode set later in is_visible */
+
+static struct attribute *mci_attrs[] = {
+ &dev_attr_reset_counters.attr,
+ &dev_attr_mc_name.attr,
+ &dev_attr_size_mb.attr,
+ &dev_attr_seconds_since_reset.attr,
+ &dev_attr_ue_noinfo_count.attr,
+ &dev_attr_ce_noinfo_count.attr,
+ &dev_attr_ue_count.attr,
+ &dev_attr_ce_count.attr,
+ &dev_attr_max_location.attr,
+ &dev_attr_sdram_scrub_rate.attr,
+ NULL
+};
+
+static umode_t mci_attr_is_visible(struct kobject *kobj,
+ struct attribute *attr, int idx)
+{
+ struct device *dev = kobj_to_dev(kobj);
+ struct mem_ctl_info *mci = to_mci(dev);
+ umode_t mode = 0;
+
+ if (attr != &dev_attr_sdram_scrub_rate.attr)
+ return attr->mode;
+ if (mci->get_sdram_scrub_rate)
+ mode |= S_IRUGO;
+ if (mci->set_sdram_scrub_rate)
+ mode |= S_IWUSR;
+ return mode;
+}
+
+static const struct attribute_group mci_attr_grp = {
+ .attrs = mci_attrs,
+ .is_visible = mci_attr_is_visible,
+};
+
+static const struct attribute_group *mci_attr_groups[] = {
+ &mci_attr_grp,
+ NULL
+};
+
+static const struct device_type mci_attr_type = {
+ .groups = mci_attr_groups,
+};
+
+/*
+ * Create a new Memory Controller kobject instance,
+ * mc<id> under the 'mc' directory
+ *
+ * Return:
+ * 0 Success
+ * !0 Failure
+ */
+int edac_create_sysfs_mci_device(struct mem_ctl_info *mci,
+ const struct attribute_group **groups)
+{
+ struct dimm_info *dimm;
+ int err;
+
+ /* get the /sys/devices/system/edac subsys reference */
+ mci->dev.type = &mci_attr_type;
+ mci->dev.parent = mci_pdev;
+ mci->dev.groups = groups;
+ dev_set_name(&mci->dev, "mc%d", mci->mc_idx);
+ dev_set_drvdata(&mci->dev, mci);
+ pm_runtime_forbid(&mci->dev);
+
+ err = device_add(&mci->dev);
+ if (err < 0) {
+ edac_dbg(1, "failure: create device %s\n", dev_name(&mci->dev));
+ /* no put_device() here, free mci with _edac_mc_free() */
+ return err;
+ }
+
+ edac_dbg(0, "device %s created\n", dev_name(&mci->dev));
+
+ /*
+ * Create the dimm/rank devices
+ */
+ mci_for_each_dimm(mci, dimm) {
+ /* Only expose populated DIMMs */
+ if (!dimm->nr_pages)
+ continue;
+
+ err = edac_create_dimm_object(mci, dimm);
+ if (err)
+ goto fail;
+ }
+
+#ifdef CONFIG_EDAC_LEGACY_SYSFS
+ err = edac_create_csrow_objects(mci);
+ if (err < 0)
+ goto fail;
+#endif
+
+ edac_create_debugfs_nodes(mci);
+ return 0;
+
+fail:
+ edac_remove_sysfs_mci_device(mci);
+
+ return err;
+}
+
+/*
+ * remove a Memory Controller instance
+ */
+void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci)
+{
+ struct dimm_info *dimm;
+
+ if (!device_is_registered(&mci->dev))
+ return;
+
+ edac_dbg(0, "\n");
+
+#ifdef CONFIG_EDAC_DEBUG
+ edac_debugfs_remove_recursive(mci->debugfs);
+#endif
+#ifdef CONFIG_EDAC_LEGACY_SYSFS
+ edac_delete_csrow_objects(mci);
+#endif
+
+ mci_for_each_dimm(mci, dimm) {
+ if (!device_is_registered(&dimm->dev))
+ continue;
+ edac_dbg(1, "unregistering device %s\n", dev_name(&dimm->dev));
+ device_unregister(&dimm->dev);
+ }
+
+ /* only remove the device, but keep mci */
+ device_del(&mci->dev);
+}
+
+static void mc_attr_release(struct device *dev)
+{
+ /*
+ * There's no container structure here, as this is just the mci
+ * parent device, used to create the /sys/devices/mc sysfs node.
+ * So, there are no attributes on it.
+ */
+ edac_dbg(1, "device %s released\n", dev_name(dev));
+ kfree(dev);
+}
+
+/*
+ * Init/exit code for the module. Basically, creates/removes /sys/class/rc
+ */
+int __init edac_mc_sysfs_init(void)
+{
+ int err;
+
+ mci_pdev = kzalloc(sizeof(*mci_pdev), GFP_KERNEL);
+ if (!mci_pdev)
+ return -ENOMEM;
+
+ mci_pdev->bus = edac_get_sysfs_subsys();
+ mci_pdev->release = mc_attr_release;
+ mci_pdev->init_name = "mc";
+
+ err = device_register(mci_pdev);
+ if (err < 0) {
+ edac_dbg(1, "failure: create device %s\n", dev_name(mci_pdev));
+ put_device(mci_pdev);
+ return err;
+ }
+
+ edac_dbg(0, "device %s created\n", dev_name(mci_pdev));
+
+ return 0;
+}
+
+void edac_mc_sysfs_exit(void)
+{
+ device_unregister(mci_pdev);
+}
diff --git a/drivers/edac/edac_module.c b/drivers/edac/edac_module.c
new file mode 100644
index 0000000000..32a931d0cb
--- /dev/null
+++ b/drivers/edac/edac_module.c
@@ -0,0 +1,168 @@
+/*
+ * edac_module.c
+ *
+ * (C) 2007 www.softwarebitmaker.com
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2. This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ *
+ * Author: Doug Thompson <dougthompson@xmission.com>
+ *
+ */
+#include <linux/edac.h>
+
+#include "edac_mc.h"
+#include "edac_module.h"
+
+#define EDAC_VERSION "Ver: 3.0.0"
+
+#ifdef CONFIG_EDAC_DEBUG
+
+static int edac_set_debug_level(const char *buf,
+ const struct kernel_param *kp)
+{
+ unsigned long val;
+ int ret;
+
+ ret = kstrtoul(buf, 0, &val);
+ if (ret)
+ return ret;
+
+ if (val > 4)
+ return -EINVAL;
+
+ return param_set_int(buf, kp);
+}
+
+/* Values of 0 to 4 will generate output */
+int edac_debug_level = 2;
+EXPORT_SYMBOL_GPL(edac_debug_level);
+
+module_param_call(edac_debug_level, edac_set_debug_level, param_get_int,
+ &edac_debug_level, 0644);
+MODULE_PARM_DESC(edac_debug_level, "EDAC debug level: [0-4], default: 2");
+#endif
+
+/*
+ * edac_op_state_to_string()
+ */
+char *edac_op_state_to_string(int opstate)
+{
+ if (opstate == OP_RUNNING_POLL)
+ return "POLLED";
+ else if (opstate == OP_RUNNING_INTERRUPT)
+ return "INTERRUPT";
+ else if (opstate == OP_RUNNING_POLL_INTR)
+ return "POLL-INTR";
+ else if (opstate == OP_ALLOC)
+ return "ALLOC";
+ else if (opstate == OP_OFFLINE)
+ return "OFFLINE";
+
+ return "UNKNOWN";
+}
+
+/*
+ * sysfs object: /sys/devices/system/edac
+ * need to export to other files
+ */
+static struct bus_type edac_subsys = {
+ .name = "edac",
+ .dev_name = "edac",
+};
+
+static int edac_subsys_init(void)
+{
+ int err;
+
+ /* create the /sys/devices/system/edac directory */
+ err = subsys_system_register(&edac_subsys, NULL);
+ if (err)
+ printk(KERN_ERR "Error registering toplevel EDAC sysfs dir\n");
+
+ return err;
+}
+
+static void edac_subsys_exit(void)
+{
+ bus_unregister(&edac_subsys);
+}
+
+/* return pointer to the 'edac' node in sysfs */
+struct bus_type *edac_get_sysfs_subsys(void)
+{
+ return &edac_subsys;
+}
+EXPORT_SYMBOL_GPL(edac_get_sysfs_subsys);
+/*
+ * edac_init
+ * module initialization entry point
+ */
+static int __init edac_init(void)
+{
+ int err = 0;
+
+ edac_printk(KERN_INFO, EDAC_MC, EDAC_VERSION "\n");
+
+ err = edac_subsys_init();
+ if (err)
+ return err;
+
+ /*
+ * Harvest and clear any boot/initialization PCI parity errors
+ *
+ * FIXME: This only clears errors logged by devices present at time of
+ * module initialization. We should also do an initial clear
+ * of each newly hotplugged device.
+ */
+ edac_pci_clear_parity_errors();
+
+ err = edac_mc_sysfs_init();
+ if (err)
+ goto err_sysfs;
+
+ edac_debugfs_init();
+
+ err = edac_workqueue_setup();
+ if (err) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failure initializing workqueue\n");
+ goto err_wq;
+ }
+
+ return 0;
+
+err_wq:
+ edac_debugfs_exit();
+ edac_mc_sysfs_exit();
+
+err_sysfs:
+ edac_subsys_exit();
+
+ return err;
+}
+
+/*
+ * edac_exit()
+ * module exit/termination function
+ */
+static void __exit edac_exit(void)
+{
+ edac_dbg(0, "\n");
+
+ /* tear down the various subsystems */
+ edac_workqueue_teardown();
+ edac_mc_sysfs_exit();
+ edac_debugfs_exit();
+ edac_subsys_exit();
+}
+
+/*
+ * Inform the kernel of our entry and exit points
+ */
+subsys_initcall(edac_init);
+module_exit(edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Doug Thompson www.softwarebitmaker.com, et al");
+MODULE_DESCRIPTION("Core library routines for EDAC reporting");
diff --git a/drivers/edac/edac_module.h b/drivers/edac/edac_module.h
new file mode 100644
index 0000000000..47593afdc2
--- /dev/null
+++ b/drivers/edac/edac_module.h
@@ -0,0 +1,125 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * edac_module.h
+ *
+ * For defining functions/data for within the EDAC_CORE module only
+ *
+ * written by doug thompson <norsk5@xmission.h>
+ */
+
+#ifndef __EDAC_MODULE_H__
+#define __EDAC_MODULE_H__
+
+#include <acpi/ghes.h>
+#include "edac_mc.h"
+#include "edac_pci.h"
+#include "edac_device.h"
+
+/*
+ * INTERNAL EDAC MODULE:
+ * EDAC memory controller sysfs create/remove functions
+ * and setup/teardown functions
+ *
+ * edac_mc objects
+ */
+ /* on edac_mc_sysfs.c */
+int edac_mc_sysfs_init(void);
+void edac_mc_sysfs_exit(void);
+extern int edac_create_sysfs_mci_device(struct mem_ctl_info *mci,
+ const struct attribute_group **groups);
+extern void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci);
+extern int edac_mc_get_log_ue(void);
+extern int edac_mc_get_log_ce(void);
+extern int edac_mc_get_panic_on_ue(void);
+extern unsigned int edac_mc_get_poll_msec(void);
+
+unsigned edac_dimm_info_location(struct dimm_info *dimm, char *buf,
+ unsigned len);
+
+ /* on edac_device.c */
+extern int edac_device_register_sysfs_main_kobj(
+ struct edac_device_ctl_info *edac_dev);
+extern void edac_device_unregister_sysfs_main_kobj(
+ struct edac_device_ctl_info *edac_dev);
+extern int edac_device_create_sysfs(struct edac_device_ctl_info *edac_dev);
+extern void edac_device_remove_sysfs(struct edac_device_ctl_info *edac_dev);
+
+/* edac core workqueue: single CPU mode */
+int edac_workqueue_setup(void);
+void edac_workqueue_teardown(void);
+bool edac_queue_work(struct delayed_work *work, unsigned long delay);
+bool edac_stop_work(struct delayed_work *work);
+bool edac_mod_work(struct delayed_work *work, unsigned long delay);
+
+extern void edac_device_reset_delay_period(struct edac_device_ctl_info
+ *edac_dev, unsigned long msec);
+extern void edac_mc_reset_delay_period(unsigned long value);
+
+/*
+ * EDAC debugfs functions
+ */
+
+#define edac_debugfs_remove_recursive debugfs_remove_recursive
+#define edac_debugfs_remove debugfs_remove
+#ifdef CONFIG_EDAC_DEBUG
+void edac_debugfs_init(void);
+void edac_debugfs_exit(void);
+void edac_create_debugfs_nodes(struct mem_ctl_info *mci);
+struct dentry *edac_debugfs_create_dir(const char *dirname);
+struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent);
+struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+ void *data, const struct file_operations *fops);
+void edac_debugfs_create_x8(const char *name, umode_t mode,
+ struct dentry *parent, u8 *value);
+void edac_debugfs_create_x16(const char *name, umode_t mode,
+ struct dentry *parent, u16 *value);
+void edac_debugfs_create_x32(const char *name, umode_t mode,
+ struct dentry *parent, u32 *value);
+#else
+static inline void edac_debugfs_init(void) { }
+static inline void edac_debugfs_exit(void) { }
+static inline void edac_create_debugfs_nodes(struct mem_ctl_info *mci) { }
+static inline struct dentry *edac_debugfs_create_dir(const char *dirname) { return NULL; }
+static inline struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent) { return NULL; }
+static inline struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+ void *data, const struct file_operations *fops) { return NULL; }
+static inline void edac_debugfs_create_x8(const char *name, umode_t mode,
+ struct dentry *parent, u8 *value) { }
+static inline void edac_debugfs_create_x16(const char *name, umode_t mode,
+ struct dentry *parent, u16 *value) { }
+static inline void edac_debugfs_create_x32(const char *name, umode_t mode,
+ struct dentry *parent, u32 *value) { }
+#endif
+
+/*
+ * EDAC PCI functions
+ */
+#ifdef CONFIG_PCI
+extern void edac_pci_do_parity_check(void);
+extern void edac_pci_clear_parity_errors(void);
+extern int edac_sysfs_pci_setup(void);
+extern void edac_sysfs_pci_teardown(void);
+extern int edac_pci_get_check_errors(void);
+extern int edac_pci_get_poll_msec(void);
+extern void edac_pci_remove_sysfs(struct edac_pci_ctl_info *pci);
+extern void edac_pci_handle_pe(struct edac_pci_ctl_info *pci, const char *msg);
+extern void edac_pci_handle_npe(struct edac_pci_ctl_info *pci,
+ const char *msg);
+#else /* CONFIG_PCI */
+/* pre-process these away */
+#define edac_pci_do_parity_check()
+#define edac_pci_clear_parity_errors()
+#define edac_sysfs_pci_setup() (0)
+#define edac_sysfs_pci_teardown()
+#define edac_pci_get_check_errors()
+#define edac_pci_get_poll_msec()
+#define edac_pci_handle_pe()
+#define edac_pci_handle_npe()
+#endif /* CONFIG_PCI */
+
+#endif /* __EDAC_MODULE_H__ */
diff --git a/drivers/edac/edac_pci.c b/drivers/edac/edac_pci.c
new file mode 100644
index 0000000000..64c142aecc
--- /dev/null
+++ b/drivers/edac/edac_pci.c
@@ -0,0 +1,337 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * EDAC PCI component
+ *
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2007 (c) MontaVista Software, Inc.
+ */
+#include <asm/page.h>
+#include <linux/uaccess.h>
+#include <linux/ctype.h>
+#include <linux/highmem.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/smp.h>
+#include <linux/spinlock.h>
+#include <linux/sysctl.h>
+#include <linux/timer.h>
+
+#include "edac_pci.h"
+#include "edac_module.h"
+
+static DEFINE_MUTEX(edac_pci_ctls_mutex);
+static LIST_HEAD(edac_pci_list);
+static atomic_t pci_indexes = ATOMIC_INIT(0);
+
+struct edac_pci_ctl_info *edac_pci_alloc_ctl_info(unsigned int sz_pvt,
+ const char *edac_pci_name)
+{
+ struct edac_pci_ctl_info *pci;
+
+ edac_dbg(1, "\n");
+
+ pci = kzalloc(sizeof(struct edac_pci_ctl_info), GFP_KERNEL);
+ if (!pci)
+ return NULL;
+
+ if (sz_pvt) {
+ pci->pvt_info = kzalloc(sz_pvt, GFP_KERNEL);
+ if (!pci->pvt_info)
+ goto free;
+ }
+
+ pci->op_state = OP_ALLOC;
+
+ snprintf(pci->name, strlen(edac_pci_name) + 1, "%s", edac_pci_name);
+
+ return pci;
+
+free:
+ kfree(pci);
+ return NULL;
+}
+EXPORT_SYMBOL_GPL(edac_pci_alloc_ctl_info);
+
+void edac_pci_free_ctl_info(struct edac_pci_ctl_info *pci)
+{
+ edac_dbg(1, "\n");
+
+ edac_pci_remove_sysfs(pci);
+}
+EXPORT_SYMBOL_GPL(edac_pci_free_ctl_info);
+
+/*
+ * find_edac_pci_by_dev()
+ * scans the edac_pci list for a specific 'struct device *'
+ *
+ * return NULL if not found, or return control struct pointer
+ */
+static struct edac_pci_ctl_info *find_edac_pci_by_dev(struct device *dev)
+{
+ struct edac_pci_ctl_info *pci;
+ struct list_head *item;
+
+ edac_dbg(1, "\n");
+
+ list_for_each(item, &edac_pci_list) {
+ pci = list_entry(item, struct edac_pci_ctl_info, link);
+
+ if (pci->dev == dev)
+ return pci;
+ }
+
+ return NULL;
+}
+
+/*
+ * add_edac_pci_to_global_list
+ * Before calling this function, caller must assign a unique value to
+ * edac_dev->pci_idx.
+ * Return:
+ * 0 on success
+ * 1 on failure
+ */
+static int add_edac_pci_to_global_list(struct edac_pci_ctl_info *pci)
+{
+ struct list_head *item, *insert_before;
+ struct edac_pci_ctl_info *rover;
+
+ edac_dbg(1, "\n");
+
+ insert_before = &edac_pci_list;
+
+ /* Determine if already on the list */
+ rover = find_edac_pci_by_dev(pci->dev);
+ if (unlikely(rover != NULL))
+ goto fail0;
+
+ /* Insert in ascending order by 'pci_idx', so find position */
+ list_for_each(item, &edac_pci_list) {
+ rover = list_entry(item, struct edac_pci_ctl_info, link);
+
+ if (rover->pci_idx >= pci->pci_idx) {
+ if (unlikely(rover->pci_idx == pci->pci_idx))
+ goto fail1;
+
+ insert_before = item;
+ break;
+ }
+ }
+
+ list_add_tail_rcu(&pci->link, insert_before);
+ return 0;
+
+fail0:
+ edac_printk(KERN_WARNING, EDAC_PCI,
+ "%s (%s) %s %s already assigned %d\n",
+ dev_name(rover->dev), edac_dev_name(rover),
+ rover->mod_name, rover->ctl_name, rover->pci_idx);
+ return 1;
+
+fail1:
+ edac_printk(KERN_WARNING, EDAC_PCI,
+ "but in low-level driver: attempt to assign\n"
+ "\tduplicate pci_idx %d in %s()\n", rover->pci_idx,
+ __func__);
+ return 1;
+}
+
+/*
+ * del_edac_pci_from_global_list
+ *
+ * remove the PCI control struct from the global list
+ */
+static void del_edac_pci_from_global_list(struct edac_pci_ctl_info *pci)
+{
+ list_del_rcu(&pci->link);
+
+ /* these are for safe removal of devices from global list while
+ * NMI handlers may be traversing list
+ */
+ synchronize_rcu();
+ INIT_LIST_HEAD(&pci->link);
+}
+
+/*
+ * edac_pci_workq_function()
+ *
+ * periodic function that performs the operation
+ * scheduled by a workq request, for a given PCI control struct
+ */
+static void edac_pci_workq_function(struct work_struct *work_req)
+{
+ struct delayed_work *d_work = to_delayed_work(work_req);
+ struct edac_pci_ctl_info *pci = to_edac_pci_ctl_work(d_work);
+ int msec;
+ unsigned long delay;
+
+ edac_dbg(3, "checking\n");
+
+ mutex_lock(&edac_pci_ctls_mutex);
+
+ if (pci->op_state != OP_RUNNING_POLL) {
+ mutex_unlock(&edac_pci_ctls_mutex);
+ return;
+ }
+
+ if (edac_pci_get_check_errors())
+ pci->edac_check(pci);
+
+ /* if we are on a one second period, then use round */
+ msec = edac_pci_get_poll_msec();
+ if (msec == 1000)
+ delay = round_jiffies_relative(msecs_to_jiffies(msec));
+ else
+ delay = msecs_to_jiffies(msec);
+
+ edac_queue_work(&pci->work, delay);
+
+ mutex_unlock(&edac_pci_ctls_mutex);
+}
+
+int edac_pci_alloc_index(void)
+{
+ return atomic_inc_return(&pci_indexes) - 1;
+}
+EXPORT_SYMBOL_GPL(edac_pci_alloc_index);
+
+int edac_pci_add_device(struct edac_pci_ctl_info *pci, int edac_idx)
+{
+ edac_dbg(0, "\n");
+
+ pci->pci_idx = edac_idx;
+ pci->start_time = jiffies;
+
+ mutex_lock(&edac_pci_ctls_mutex);
+
+ if (add_edac_pci_to_global_list(pci))
+ goto fail0;
+
+ if (edac_pci_create_sysfs(pci)) {
+ edac_pci_printk(pci, KERN_WARNING,
+ "failed to create sysfs pci\n");
+ goto fail1;
+ }
+
+ if (pci->edac_check) {
+ pci->op_state = OP_RUNNING_POLL;
+
+ INIT_DELAYED_WORK(&pci->work, edac_pci_workq_function);
+ edac_queue_work(&pci->work, msecs_to_jiffies(edac_pci_get_poll_msec()));
+
+ } else {
+ pci->op_state = OP_RUNNING_INTERRUPT;
+ }
+
+ edac_pci_printk(pci, KERN_INFO,
+ "Giving out device to module %s controller %s: DEV %s (%s)\n",
+ pci->mod_name, pci->ctl_name, pci->dev_name,
+ edac_op_state_to_string(pci->op_state));
+
+ mutex_unlock(&edac_pci_ctls_mutex);
+ return 0;
+
+ /* error unwind stack */
+fail1:
+ del_edac_pci_from_global_list(pci);
+fail0:
+ mutex_unlock(&edac_pci_ctls_mutex);
+ return 1;
+}
+EXPORT_SYMBOL_GPL(edac_pci_add_device);
+
+struct edac_pci_ctl_info *edac_pci_del_device(struct device *dev)
+{
+ struct edac_pci_ctl_info *pci;
+
+ edac_dbg(0, "\n");
+
+ mutex_lock(&edac_pci_ctls_mutex);
+
+ /* ensure the control struct is on the global list
+ * if not, then leave
+ */
+ pci = find_edac_pci_by_dev(dev);
+ if (pci == NULL) {
+ mutex_unlock(&edac_pci_ctls_mutex);
+ return NULL;
+ }
+
+ pci->op_state = OP_OFFLINE;
+
+ del_edac_pci_from_global_list(pci);
+
+ mutex_unlock(&edac_pci_ctls_mutex);
+
+ if (pci->edac_check)
+ edac_stop_work(&pci->work);
+
+ edac_printk(KERN_INFO, EDAC_PCI,
+ "Removed device %d for %s %s: DEV %s\n",
+ pci->pci_idx, pci->mod_name, pci->ctl_name, edac_dev_name(pci));
+
+ return pci;
+}
+EXPORT_SYMBOL_GPL(edac_pci_del_device);
+
+/*
+ * edac_pci_generic_check
+ *
+ * a Generic parity check API
+ */
+static void edac_pci_generic_check(struct edac_pci_ctl_info *pci)
+{
+ edac_dbg(4, "\n");
+ edac_pci_do_parity_check();
+}
+
+/* free running instance index counter */
+static int edac_pci_idx;
+#define EDAC_PCI_GENCTL_NAME "EDAC PCI controller"
+
+struct edac_pci_gen_data {
+ int edac_idx;
+};
+
+struct edac_pci_ctl_info *edac_pci_create_generic_ctl(struct device *dev,
+ const char *mod_name)
+{
+ struct edac_pci_ctl_info *pci;
+ struct edac_pci_gen_data *pdata;
+
+ pci = edac_pci_alloc_ctl_info(sizeof(*pdata), EDAC_PCI_GENCTL_NAME);
+ if (!pci)
+ return NULL;
+
+ pdata = pci->pvt_info;
+ pci->dev = dev;
+ dev_set_drvdata(pci->dev, pci);
+ pci->dev_name = pci_name(to_pci_dev(dev));
+
+ pci->mod_name = mod_name;
+ pci->ctl_name = EDAC_PCI_GENCTL_NAME;
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ pci->edac_check = edac_pci_generic_check;
+
+ pdata->edac_idx = edac_pci_idx++;
+
+ if (edac_pci_add_device(pci, pdata->edac_idx) > 0) {
+ edac_dbg(3, "failed edac_pci_add_device()\n");
+ edac_pci_free_ctl_info(pci);
+ return NULL;
+ }
+
+ return pci;
+}
+EXPORT_SYMBOL_GPL(edac_pci_create_generic_ctl);
+
+void edac_pci_release_generic_ctl(struct edac_pci_ctl_info *pci)
+{
+ edac_dbg(0, "pci mod=%s\n", pci->mod_name);
+
+ edac_pci_del_device(pci->dev);
+ edac_pci_free_ctl_info(pci);
+}
+EXPORT_SYMBOL_GPL(edac_pci_release_generic_ctl);
diff --git a/drivers/edac/edac_pci.h b/drivers/edac/edac_pci.h
new file mode 100644
index 0000000000..5175f5724c
--- /dev/null
+++ b/drivers/edac/edac_pci.h
@@ -0,0 +1,271 @@
+/*
+ * Defines, structures, APIs for edac_pci and edac_pci_sysfs
+ *
+ * (C) 2007 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Based on work by Dan Hollis <goemon at anime dot net> and others.
+ * http://www.anime.net/~goemon/linux-ecc/
+ *
+ * NMI handling support added by
+ * Dave Peterson <dsp@llnl.gov> <dave_peterson@pobox.com>
+ *
+ * Refactored for multi-source files:
+ * Doug Thompson <norsk5@xmission.com>
+ *
+ * Please look at Documentation/driver-api/edac.rst for more info about
+ * EDAC core structs and functions.
+ */
+
+#ifndef _EDAC_PCI_H_
+#define _EDAC_PCI_H_
+
+#include <linux/completion.h>
+#include <linux/device.h>
+#include <linux/edac.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/pci.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#ifdef CONFIG_PCI
+
+struct edac_pci_counter {
+ atomic_t pe_count;
+ atomic_t npe_count;
+};
+
+/*
+ * Abstract edac_pci control info structure
+ *
+ */
+struct edac_pci_ctl_info {
+ /* for global list of edac_pci_ctl_info structs */
+ struct list_head link;
+
+ int pci_idx;
+
+ struct bus_type *edac_subsys; /* pointer to subsystem */
+
+ /* the internal state of this controller instance */
+ int op_state;
+ /* work struct for this instance */
+ struct delayed_work work;
+
+ /* pointer to edac polling checking routine:
+ * If NOT NULL: points to polling check routine
+ * If NULL: Then assumes INTERRUPT operation, where
+ * MC driver will receive events
+ */
+ void (*edac_check) (struct edac_pci_ctl_info * edac_dev);
+
+ struct device *dev; /* pointer to device structure */
+
+ const char *mod_name; /* module name */
+ const char *ctl_name; /* edac controller name */
+ const char *dev_name; /* pci/platform/etc... name */
+
+ void *pvt_info; /* pointer to 'private driver' info */
+
+ unsigned long start_time; /* edac_pci load start time (jiffies) */
+
+ struct completion complete;
+
+ /* sysfs top name under 'edac' directory
+ * and instance name:
+ * cpu/cpu0/...
+ * cpu/cpu1/...
+ * cpu/cpu2/...
+ * ...
+ */
+ char name[EDAC_DEVICE_NAME_LEN + 1];
+
+ /* Event counters for the this whole EDAC Device */
+ struct edac_pci_counter counters;
+
+ /* edac sysfs device control for the 'name'
+ * device this structure controls
+ */
+ struct kobject kobj;
+};
+
+#define to_edac_pci_ctl_work(w) \
+ container_of(w, struct edac_pci_ctl_info,work)
+
+/* write all or some bits in a byte-register*/
+static inline void pci_write_bits8(struct pci_dev *pdev, int offset, u8 value,
+ u8 mask)
+{
+ if (mask != 0xff) {
+ u8 buf;
+
+ pci_read_config_byte(pdev, offset, &buf);
+ value &= mask;
+ buf &= ~mask;
+ value |= buf;
+ }
+
+ pci_write_config_byte(pdev, offset, value);
+}
+
+/* write all or some bits in a word-register*/
+static inline void pci_write_bits16(struct pci_dev *pdev, int offset,
+ u16 value, u16 mask)
+{
+ if (mask != 0xffff) {
+ u16 buf;
+
+ pci_read_config_word(pdev, offset, &buf);
+ value &= mask;
+ buf &= ~mask;
+ value |= buf;
+ }
+
+ pci_write_config_word(pdev, offset, value);
+}
+
+/*
+ * pci_write_bits32
+ *
+ * edac local routine to do pci_write_config_dword, but adds
+ * a mask parameter. If mask is all ones, ignore the mask.
+ * Otherwise utilize the mask to isolate specified bits
+ *
+ * write all or some bits in a dword-register
+ */
+static inline void pci_write_bits32(struct pci_dev *pdev, int offset,
+ u32 value, u32 mask)
+{
+ if (mask != 0xffffffff) {
+ u32 buf;
+
+ pci_read_config_dword(pdev, offset, &buf);
+ value &= mask;
+ buf &= ~mask;
+ value |= buf;
+ }
+
+ pci_write_config_dword(pdev, offset, value);
+}
+
+#endif /* CONFIG_PCI */
+
+/*
+ * edac_pci APIs
+ */
+
+/**
+ * edac_pci_alloc_ctl_info:
+ * The alloc() function for the 'edac_pci' control info
+ * structure.
+ *
+ * @sz_pvt: size of the private info at struct &edac_pci_ctl_info
+ * @edac_pci_name: name of the PCI device
+ *
+ * The chip driver will allocate one of these for each
+ * edac_pci it is going to control/register with the EDAC CORE.
+ *
+ * Returns: a pointer to struct &edac_pci_ctl_info on success; %NULL otherwise.
+ */
+extern struct edac_pci_ctl_info *edac_pci_alloc_ctl_info(unsigned int sz_pvt,
+ const char *edac_pci_name);
+
+/**
+ * edac_pci_free_ctl_info():
+ * Last action on the pci control structure.
+ *
+ * @pci: pointer to struct &edac_pci_ctl_info
+ *
+ * Calls the remove sysfs information, which will unregister
+ * this control struct's kobj. When that kobj's ref count
+ * goes to zero, its release function will be call and then
+ * kfree() the memory.
+ */
+extern void edac_pci_free_ctl_info(struct edac_pci_ctl_info *pci);
+
+/**
+ * edac_pci_alloc_index: Allocate a unique PCI index number
+ *
+ * Returns:
+ * allocated index number
+ *
+ */
+extern int edac_pci_alloc_index(void);
+
+/**
+ * edac_pci_add_device(): Insert the 'edac_dev' structure into the
+ * edac_pci global list and create sysfs entries associated with
+ * edac_pci structure.
+ *
+ * @pci: pointer to the edac_device structure to be added to the list
+ * @edac_idx: A unique numeric identifier to be assigned to the
+ * 'edac_pci' structure.
+ *
+ * Returns:
+ * 0 on Success, or an error code on failure
+ */
+extern int edac_pci_add_device(struct edac_pci_ctl_info *pci, int edac_idx);
+
+/**
+ * edac_pci_del_device()
+ * Remove sysfs entries for specified edac_pci structure and
+ * then remove edac_pci structure from global list
+ *
+ * @dev:
+ * Pointer to 'struct device' representing edac_pci structure
+ * to remove
+ *
+ * Returns:
+ * Pointer to removed edac_pci structure,
+ * or %NULL if device not found
+ */
+extern struct edac_pci_ctl_info *edac_pci_del_device(struct device *dev);
+
+/**
+ * edac_pci_create_generic_ctl()
+ * A generic constructor for a PCI parity polling device
+ * Some systems have more than one domain of PCI busses.
+ * For systems with one domain, then this API will
+ * provide for a generic poller.
+ *
+ * @dev: pointer to struct &device;
+ * @mod_name: name of the PCI device
+ *
+ * This routine calls the edac_pci_alloc_ctl_info() for
+ * the generic device, with default values
+ *
+ * Returns: Pointer to struct &edac_pci_ctl_info on success, %NULL on
+ * failure.
+ */
+extern struct edac_pci_ctl_info *edac_pci_create_generic_ctl(
+ struct device *dev,
+ const char *mod_name);
+
+/**
+ * edac_pci_release_generic_ctl
+ * The release function of a generic EDAC PCI polling device
+ *
+ * @pci: pointer to struct &edac_pci_ctl_info
+ */
+extern void edac_pci_release_generic_ctl(struct edac_pci_ctl_info *pci);
+
+/**
+ * edac_pci_create_sysfs
+ * Create the controls/attributes for the specified EDAC PCI device
+ *
+ * @pci: pointer to struct &edac_pci_ctl_info
+ */
+extern int edac_pci_create_sysfs(struct edac_pci_ctl_info *pci);
+
+/**
+ * edac_pci_remove_sysfs()
+ * remove the controls and attributes for this EDAC PCI device
+ *
+ * @pci: pointer to struct &edac_pci_ctl_info
+ */
+extern void edac_pci_remove_sysfs(struct edac_pci_ctl_info *pci);
+
+#endif
diff --git a/drivers/edac/edac_pci_sysfs.c b/drivers/edac/edac_pci_sysfs.c
new file mode 100644
index 0000000000..287cc51dbc
--- /dev/null
+++ b/drivers/edac/edac_pci_sysfs.c
@@ -0,0 +1,743 @@
+/*
+ * (C) 2005, 2006 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written Doug Thompson <norsk5@xmission.com>
+ *
+ */
+#include <linux/module.h>
+#include <linux/edac.h>
+#include <linux/slab.h>
+#include <linux/ctype.h>
+
+#include "edac_pci.h"
+#include "edac_module.h"
+
+#define EDAC_PCI_SYMLINK "device"
+
+/* data variables exported via sysfs */
+static int check_pci_errors; /* default NO check PCI parity */
+static int edac_pci_panic_on_pe; /* default NO panic on PCI Parity */
+static int edac_pci_log_pe = 1; /* log PCI parity errors */
+static int edac_pci_log_npe = 1; /* log PCI non-parity error errors */
+static int edac_pci_poll_msec = 1000; /* one second workq period */
+
+static atomic_t pci_parity_count = ATOMIC_INIT(0);
+static atomic_t pci_nonparity_count = ATOMIC_INIT(0);
+
+static struct kobject *edac_pci_top_main_kobj;
+static atomic_t edac_pci_sysfs_refcount = ATOMIC_INIT(0);
+
+/* getter functions for the data variables */
+int edac_pci_get_check_errors(void)
+{
+ return check_pci_errors;
+}
+
+static int edac_pci_get_log_pe(void)
+{
+ return edac_pci_log_pe;
+}
+
+static int edac_pci_get_log_npe(void)
+{
+ return edac_pci_log_npe;
+}
+
+static int edac_pci_get_panic_on_pe(void)
+{
+ return edac_pci_panic_on_pe;
+}
+
+int edac_pci_get_poll_msec(void)
+{
+ return edac_pci_poll_msec;
+}
+
+/**************************** EDAC PCI sysfs instance *******************/
+static ssize_t instance_pe_count_show(struct edac_pci_ctl_info *pci, char *data)
+{
+ return sprintf(data, "%u\n", atomic_read(&pci->counters.pe_count));
+}
+
+static ssize_t instance_npe_count_show(struct edac_pci_ctl_info *pci,
+ char *data)
+{
+ return sprintf(data, "%u\n", atomic_read(&pci->counters.npe_count));
+}
+
+#define to_instance(k) container_of(k, struct edac_pci_ctl_info, kobj)
+#define to_instance_attr(a) container_of(a, struct instance_attribute, attr)
+
+/* DEVICE instance kobject release() function */
+static void edac_pci_instance_release(struct kobject *kobj)
+{
+ struct edac_pci_ctl_info *pci;
+
+ edac_dbg(0, "\n");
+
+ /* Form pointer to containing struct, the pci control struct */
+ pci = to_instance(kobj);
+
+ /* decrement reference count on top main kobj */
+ kobject_put(edac_pci_top_main_kobj);
+
+ kfree(pci); /* Free the control struct */
+}
+
+/* instance specific attribute structure */
+struct instance_attribute {
+ struct attribute attr;
+ ssize_t(*show) (struct edac_pci_ctl_info *, char *);
+ ssize_t(*store) (struct edac_pci_ctl_info *, const char *, size_t);
+};
+
+/* Function to 'show' fields from the edac_pci 'instance' structure */
+static ssize_t edac_pci_instance_show(struct kobject *kobj,
+ struct attribute *attr, char *buffer)
+{
+ struct edac_pci_ctl_info *pci = to_instance(kobj);
+ struct instance_attribute *instance_attr = to_instance_attr(attr);
+
+ if (instance_attr->show)
+ return instance_attr->show(pci, buffer);
+ return -EIO;
+}
+
+/* Function to 'store' fields into the edac_pci 'instance' structure */
+static ssize_t edac_pci_instance_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
+{
+ struct edac_pci_ctl_info *pci = to_instance(kobj);
+ struct instance_attribute *instance_attr = to_instance_attr(attr);
+
+ if (instance_attr->store)
+ return instance_attr->store(pci, buffer, count);
+ return -EIO;
+}
+
+/* fs_ops table */
+static const struct sysfs_ops pci_instance_ops = {
+ .show = edac_pci_instance_show,
+ .store = edac_pci_instance_store
+};
+
+#define INSTANCE_ATTR(_name, _mode, _show, _store) \
+static struct instance_attribute attr_instance_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .show = _show, \
+ .store = _store, \
+};
+
+INSTANCE_ATTR(pe_count, S_IRUGO, instance_pe_count_show, NULL);
+INSTANCE_ATTR(npe_count, S_IRUGO, instance_npe_count_show, NULL);
+
+/* pci instance attributes */
+static struct attribute *pci_instance_attrs[] = {
+ &attr_instance_pe_count.attr,
+ &attr_instance_npe_count.attr,
+ NULL
+};
+ATTRIBUTE_GROUPS(pci_instance);
+
+/* the ktype for a pci instance */
+static struct kobj_type ktype_pci_instance = {
+ .release = edac_pci_instance_release,
+ .sysfs_ops = &pci_instance_ops,
+ .default_groups = pci_instance_groups,
+};
+
+/*
+ * edac_pci_create_instance_kobj
+ *
+ * construct one EDAC PCI instance's kobject for use
+ */
+static int edac_pci_create_instance_kobj(struct edac_pci_ctl_info *pci, int idx)
+{
+ struct kobject *main_kobj;
+ int err;
+
+ edac_dbg(0, "\n");
+
+ /* First bump the ref count on the top main kobj, which will
+ * track the number of PCI instances we have, and thus nest
+ * properly on keeping the module loaded
+ */
+ main_kobj = kobject_get(edac_pci_top_main_kobj);
+ if (!main_kobj) {
+ err = -ENODEV;
+ goto error_out;
+ }
+
+ /* And now register this new kobject under the main kobj */
+ err = kobject_init_and_add(&pci->kobj, &ktype_pci_instance,
+ edac_pci_top_main_kobj, "pci%d", idx);
+ if (err != 0) {
+ edac_dbg(2, "failed to register instance pci%d\n", idx);
+ kobject_put(edac_pci_top_main_kobj);
+ goto error_out;
+ }
+
+ kobject_uevent(&pci->kobj, KOBJ_ADD);
+ edac_dbg(1, "Register instance 'pci%d' kobject\n", idx);
+
+ return 0;
+
+ /* Error unwind statck */
+error_out:
+ return err;
+}
+
+/*
+ * edac_pci_unregister_sysfs_instance_kobj
+ *
+ * unregister the kobj for the EDAC PCI instance
+ */
+static void edac_pci_unregister_sysfs_instance_kobj(
+ struct edac_pci_ctl_info *pci)
+{
+ edac_dbg(0, "\n");
+
+ /* Unregister the instance kobject and allow its release
+ * function release the main reference count and then
+ * kfree the memory
+ */
+ kobject_put(&pci->kobj);
+}
+
+/***************************** EDAC PCI sysfs root **********************/
+#define to_edacpci(k) container_of(k, struct edac_pci_ctl_info, kobj)
+#define to_edacpci_attr(a) container_of(a, struct edac_pci_attr, attr)
+
+/* simple show/store functions for attributes */
+static ssize_t edac_pci_int_show(void *ptr, char *buffer)
+{
+ int *value = ptr;
+ return sprintf(buffer, "%d\n", *value);
+}
+
+static ssize_t edac_pci_int_store(void *ptr, const char *buffer, size_t count)
+{
+ int *value = ptr;
+
+ if (isdigit(*buffer))
+ *value = simple_strtoul(buffer, NULL, 0);
+
+ return count;
+}
+
+struct edac_pci_dev_attribute {
+ struct attribute attr;
+ void *value;
+ ssize_t(*show) (void *, char *);
+ ssize_t(*store) (void *, const char *, size_t);
+};
+
+/* Set of show/store abstract level functions for PCI Parity object */
+static ssize_t edac_pci_dev_show(struct kobject *kobj, struct attribute *attr,
+ char *buffer)
+{
+ struct edac_pci_dev_attribute *edac_pci_dev;
+ edac_pci_dev = (struct edac_pci_dev_attribute *)attr;
+
+ if (edac_pci_dev->show)
+ return edac_pci_dev->show(edac_pci_dev->value, buffer);
+ return -EIO;
+}
+
+static ssize_t edac_pci_dev_store(struct kobject *kobj,
+ struct attribute *attr, const char *buffer,
+ size_t count)
+{
+ struct edac_pci_dev_attribute *edac_pci_dev;
+ edac_pci_dev = (struct edac_pci_dev_attribute *)attr;
+
+ if (edac_pci_dev->store)
+ return edac_pci_dev->store(edac_pci_dev->value, buffer, count);
+ return -EIO;
+}
+
+static const struct sysfs_ops edac_pci_sysfs_ops = {
+ .show = edac_pci_dev_show,
+ .store = edac_pci_dev_store
+};
+
+#define EDAC_PCI_ATTR(_name,_mode,_show,_store) \
+static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .value = &_name, \
+ .show = _show, \
+ .store = _store, \
+};
+
+#define EDAC_PCI_STRING_ATTR(_name,_data,_mode,_show,_store) \
+static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
+ .attr = {.name = __stringify(_name), .mode = _mode }, \
+ .value = _data, \
+ .show = _show, \
+ .store = _store, \
+};
+
+/* PCI Parity control files */
+EDAC_PCI_ATTR(check_pci_errors, S_IRUGO | S_IWUSR, edac_pci_int_show,
+ edac_pci_int_store);
+EDAC_PCI_ATTR(edac_pci_log_pe, S_IRUGO | S_IWUSR, edac_pci_int_show,
+ edac_pci_int_store);
+EDAC_PCI_ATTR(edac_pci_log_npe, S_IRUGO | S_IWUSR, edac_pci_int_show,
+ edac_pci_int_store);
+EDAC_PCI_ATTR(edac_pci_panic_on_pe, S_IRUGO | S_IWUSR, edac_pci_int_show,
+ edac_pci_int_store);
+EDAC_PCI_ATTR(pci_parity_count, S_IRUGO, edac_pci_int_show, NULL);
+EDAC_PCI_ATTR(pci_nonparity_count, S_IRUGO, edac_pci_int_show, NULL);
+
+/* Base Attributes of the memory ECC object */
+static struct attribute *edac_pci_attrs[] = {
+ &edac_pci_attr_check_pci_errors.attr,
+ &edac_pci_attr_edac_pci_log_pe.attr,
+ &edac_pci_attr_edac_pci_log_npe.attr,
+ &edac_pci_attr_edac_pci_panic_on_pe.attr,
+ &edac_pci_attr_pci_parity_count.attr,
+ &edac_pci_attr_pci_nonparity_count.attr,
+ NULL,
+};
+ATTRIBUTE_GROUPS(edac_pci);
+
+/*
+ * edac_pci_release_main_kobj
+ *
+ * This release function is called when the reference count to the
+ * passed kobj goes to zero.
+ *
+ * This kobj is the 'main' kobject that EDAC PCI instances
+ * link to, and thus provide for proper nesting counts
+ */
+static void edac_pci_release_main_kobj(struct kobject *kobj)
+{
+ edac_dbg(0, "here to module_put(THIS_MODULE)\n");
+
+ kfree(kobj);
+
+ /* last reference to top EDAC PCI kobject has been removed,
+ * NOW release our ref count on the core module
+ */
+ module_put(THIS_MODULE);
+}
+
+/* ktype struct for the EDAC PCI main kobj */
+static struct kobj_type ktype_edac_pci_main_kobj = {
+ .release = edac_pci_release_main_kobj,
+ .sysfs_ops = &edac_pci_sysfs_ops,
+ .default_groups = edac_pci_groups,
+};
+
+/**
+ * edac_pci_main_kobj_setup: Setup the sysfs for EDAC PCI attributes.
+ */
+static int edac_pci_main_kobj_setup(void)
+{
+ int err = -ENODEV;
+ struct bus_type *edac_subsys;
+ struct device *dev_root;
+
+ edac_dbg(0, "\n");
+
+ /* check and count if we have already created the main kobject */
+ if (atomic_inc_return(&edac_pci_sysfs_refcount) != 1)
+ return 0;
+
+ /* First time, so create the main kobject and its
+ * controls and attributes
+ */
+ edac_subsys = edac_get_sysfs_subsys();
+
+ /* Bump the reference count on this module to ensure the
+ * modules isn't unloaded until we deconstruct the top
+ * level main kobj for EDAC PCI
+ */
+ if (!try_module_get(THIS_MODULE)) {
+ edac_dbg(1, "try_module_get() failed\n");
+ goto decrement_count_fail;
+ }
+
+ edac_pci_top_main_kobj = kzalloc(sizeof(struct kobject), GFP_KERNEL);
+ if (!edac_pci_top_main_kobj) {
+ edac_dbg(1, "Failed to allocate\n");
+ err = -ENOMEM;
+ goto kzalloc_fail;
+ }
+
+ /* Instanstiate the pci object */
+ dev_root = bus_get_dev_root(edac_subsys);
+ if (dev_root) {
+ err = kobject_init_and_add(edac_pci_top_main_kobj,
+ &ktype_edac_pci_main_kobj,
+ &dev_root->kobj, "pci");
+ put_device(dev_root);
+ }
+ if (err) {
+ edac_dbg(1, "Failed to register '.../edac/pci'\n");
+ goto kobject_init_and_add_fail;
+ }
+
+ /* At this point, to 'release' the top level kobject
+ * for EDAC PCI, then edac_pci_main_kobj_teardown()
+ * must be used, for resources to be cleaned up properly
+ */
+ kobject_uevent(edac_pci_top_main_kobj, KOBJ_ADD);
+ edac_dbg(1, "Registered '.../edac/pci' kobject\n");
+
+ return 0;
+
+ /* Error unwind statck */
+kobject_init_and_add_fail:
+ kobject_put(edac_pci_top_main_kobj);
+
+kzalloc_fail:
+ module_put(THIS_MODULE);
+
+decrement_count_fail:
+ /* if are on this error exit, nothing to tear down */
+ atomic_dec(&edac_pci_sysfs_refcount);
+
+ return err;
+}
+
+/*
+ * edac_pci_main_kobj_teardown()
+ *
+ * if no longer linked (needed) remove the top level EDAC PCI
+ * kobject with its controls and attributes
+ */
+static void edac_pci_main_kobj_teardown(void)
+{
+ edac_dbg(0, "\n");
+
+ /* Decrement the count and only if no more controller instances
+ * are connected perform the unregisteration of the top level
+ * main kobj
+ */
+ if (atomic_dec_return(&edac_pci_sysfs_refcount) == 0) {
+ edac_dbg(0, "called kobject_put on main kobj\n");
+ kobject_put(edac_pci_top_main_kobj);
+ }
+}
+
+int edac_pci_create_sysfs(struct edac_pci_ctl_info *pci)
+{
+ int err;
+ struct kobject *edac_kobj = &pci->kobj;
+
+ edac_dbg(0, "idx=%d\n", pci->pci_idx);
+
+ /* create the top main EDAC PCI kobject, IF needed */
+ err = edac_pci_main_kobj_setup();
+ if (err)
+ return err;
+
+ /* Create this instance's kobject under the MAIN kobject */
+ err = edac_pci_create_instance_kobj(pci, pci->pci_idx);
+ if (err)
+ goto unregister_cleanup;
+
+ err = sysfs_create_link(edac_kobj, &pci->dev->kobj, EDAC_PCI_SYMLINK);
+ if (err) {
+ edac_dbg(0, "sysfs_create_link() returned err= %d\n", err);
+ goto symlink_fail;
+ }
+
+ return 0;
+
+ /* Error unwind stack */
+symlink_fail:
+ edac_pci_unregister_sysfs_instance_kobj(pci);
+
+unregister_cleanup:
+ edac_pci_main_kobj_teardown();
+
+ return err;
+}
+
+void edac_pci_remove_sysfs(struct edac_pci_ctl_info *pci)
+{
+ edac_dbg(0, "index=%d\n", pci->pci_idx);
+
+ /* Remove the symlink */
+ sysfs_remove_link(&pci->kobj, EDAC_PCI_SYMLINK);
+
+ /* remove this PCI instance's sysfs entries */
+ edac_pci_unregister_sysfs_instance_kobj(pci);
+
+ /* Call the main unregister function, which will determine
+ * if this 'pci' is the last instance.
+ * If it is, the main kobject will be unregistered as a result
+ */
+ edac_dbg(0, "calling edac_pci_main_kobj_teardown()\n");
+ edac_pci_main_kobj_teardown();
+}
+
+/************************ PCI error handling *************************/
+static u16 get_pci_parity_status(struct pci_dev *dev, int secondary)
+{
+ int where;
+ u16 status;
+
+ where = secondary ? PCI_SEC_STATUS : PCI_STATUS;
+ pci_read_config_word(dev, where, &status);
+
+ /* If we get back 0xFFFF then we must suspect that the card has been
+ * pulled but the Linux PCI layer has not yet finished cleaning up.
+ * We don't want to report on such devices
+ */
+
+ if (status == 0xFFFF) {
+ u32 sanity;
+
+ pci_read_config_dword(dev, 0, &sanity);
+
+ if (sanity == 0xFFFFFFFF)
+ return 0;
+ }
+
+ status &= PCI_STATUS_DETECTED_PARITY | PCI_STATUS_SIG_SYSTEM_ERROR |
+ PCI_STATUS_PARITY;
+
+ if (status)
+ /* reset only the bits we are interested in */
+ pci_write_config_word(dev, where, status);
+
+ return status;
+}
+
+
+/* Clear any PCI parity errors logged by this device. */
+static void edac_pci_dev_parity_clear(struct pci_dev *dev)
+{
+ u8 header_type;
+
+ get_pci_parity_status(dev, 0);
+
+ /* read the device TYPE, looking for bridges */
+ pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
+
+ if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE)
+ get_pci_parity_status(dev, 1);
+}
+
+/*
+ * PCI Parity polling
+ *
+ * Function to retrieve the current parity status
+ * and decode it
+ *
+ */
+static void edac_pci_dev_parity_test(struct pci_dev *dev)
+{
+ unsigned long flags;
+ u16 status;
+ u8 header_type;
+
+ /* stop any interrupts until we can acquire the status */
+ local_irq_save(flags);
+
+ /* read the STATUS register on this device */
+ status = get_pci_parity_status(dev, 0);
+
+ /* read the device TYPE, looking for bridges */
+ pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
+
+ local_irq_restore(flags);
+
+ edac_dbg(4, "PCI STATUS= 0x%04x %s\n", status, dev_name(&dev->dev));
+
+ /* check the status reg for errors on boards NOT marked as broken
+ * if broken, we cannot trust any of the status bits
+ */
+ if (status && !dev->broken_parity_status) {
+ if (status & (PCI_STATUS_SIG_SYSTEM_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_PCI,
+ "Signaled System Error on %s\n",
+ pci_name(dev));
+ atomic_inc(&pci_nonparity_count);
+ }
+
+ if (status & (PCI_STATUS_PARITY)) {
+ edac_printk(KERN_CRIT, EDAC_PCI,
+ "Master Data Parity Error on %s\n",
+ pci_name(dev));
+
+ atomic_inc(&pci_parity_count);
+ }
+
+ if (status & (PCI_STATUS_DETECTED_PARITY)) {
+ edac_printk(KERN_CRIT, EDAC_PCI,
+ "Detected Parity Error on %s\n",
+ pci_name(dev));
+
+ atomic_inc(&pci_parity_count);
+ }
+ }
+
+
+ edac_dbg(4, "PCI HEADER TYPE= 0x%02x %s\n",
+ header_type, dev_name(&dev->dev));
+
+ if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE) {
+ /* On bridges, need to examine secondary status register */
+ status = get_pci_parity_status(dev, 1);
+
+ edac_dbg(4, "PCI SEC_STATUS= 0x%04x %s\n",
+ status, dev_name(&dev->dev));
+
+ /* check the secondary status reg for errors,
+ * on NOT broken boards
+ */
+ if (status && !dev->broken_parity_status) {
+ if (status & (PCI_STATUS_SIG_SYSTEM_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
+ "Signaled System Error on %s\n",
+ pci_name(dev));
+ atomic_inc(&pci_nonparity_count);
+ }
+
+ if (status & (PCI_STATUS_PARITY)) {
+ edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
+ "Master Data Parity Error on "
+ "%s\n", pci_name(dev));
+
+ atomic_inc(&pci_parity_count);
+ }
+
+ if (status & (PCI_STATUS_DETECTED_PARITY)) {
+ edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
+ "Detected Parity Error on %s\n",
+ pci_name(dev));
+
+ atomic_inc(&pci_parity_count);
+ }
+ }
+ }
+}
+
+/* reduce some complexity in definition of the iterator */
+typedef void (*pci_parity_check_fn_t) (struct pci_dev *dev);
+
+/*
+ * pci_dev parity list iterator
+ *
+ * Scan the PCI device list looking for SERRORs, Master Parity ERRORS or
+ * Parity ERRORs on primary or secondary devices.
+ */
+static inline void edac_pci_dev_parity_iterator(pci_parity_check_fn_t fn)
+{
+ struct pci_dev *dev = NULL;
+
+ for_each_pci_dev(dev)
+ fn(dev);
+}
+
+/*
+ * edac_pci_do_parity_check
+ *
+ * performs the actual PCI parity check operation
+ */
+void edac_pci_do_parity_check(void)
+{
+ int before_count;
+
+ edac_dbg(3, "\n");
+
+ /* if policy has PCI check off, leave now */
+ if (!check_pci_errors)
+ return;
+
+ before_count = atomic_read(&pci_parity_count);
+
+ /* scan all PCI devices looking for a Parity Error on devices and
+ * bridges.
+ * The iterator calls pci_get_device() which might sleep, thus
+ * we cannot disable interrupts in this scan.
+ */
+ edac_pci_dev_parity_iterator(edac_pci_dev_parity_test);
+
+ /* Only if operator has selected panic on PCI Error */
+ if (edac_pci_get_panic_on_pe()) {
+ /* If the count is different 'after' from 'before' */
+ if (before_count != atomic_read(&pci_parity_count))
+ panic("EDAC: PCI Parity Error");
+ }
+}
+
+/*
+ * edac_pci_clear_parity_errors
+ *
+ * function to perform an iteration over the PCI devices
+ * and clearn their current status
+ */
+void edac_pci_clear_parity_errors(void)
+{
+ /* Clear any PCI bus parity errors that devices initially have logged
+ * in their registers.
+ */
+ edac_pci_dev_parity_iterator(edac_pci_dev_parity_clear);
+}
+
+/*
+ * edac_pci_handle_pe
+ *
+ * Called to handle a PARITY ERROR event
+ */
+void edac_pci_handle_pe(struct edac_pci_ctl_info *pci, const char *msg)
+{
+
+ /* global PE counter incremented by edac_pci_do_parity_check() */
+ atomic_inc(&pci->counters.pe_count);
+
+ if (edac_pci_get_log_pe())
+ edac_pci_printk(pci, KERN_WARNING,
+ "Parity Error ctl: %s %d: %s\n",
+ pci->ctl_name, pci->pci_idx, msg);
+
+ /*
+ * poke all PCI devices and see which one is the troublemaker
+ * panic() is called if set
+ */
+ edac_pci_do_parity_check();
+}
+EXPORT_SYMBOL_GPL(edac_pci_handle_pe);
+
+
+/*
+ * edac_pci_handle_npe
+ *
+ * Called to handle a NON-PARITY ERROR event
+ */
+void edac_pci_handle_npe(struct edac_pci_ctl_info *pci, const char *msg)
+{
+
+ /* global NPE counter incremented by edac_pci_do_parity_check() */
+ atomic_inc(&pci->counters.npe_count);
+
+ if (edac_pci_get_log_npe())
+ edac_pci_printk(pci, KERN_WARNING,
+ "Non-Parity Error ctl: %s %d: %s\n",
+ pci->ctl_name, pci->pci_idx, msg);
+
+ /*
+ * poke all PCI devices and see which one is the troublemaker
+ * panic() is called if set
+ */
+ edac_pci_do_parity_check();
+}
+EXPORT_SYMBOL_GPL(edac_pci_handle_npe);
+
+/*
+ * Define the PCI parameter to the module
+ */
+module_param(check_pci_errors, int, 0644);
+MODULE_PARM_DESC(check_pci_errors,
+ "Check for PCI bus parity errors: 0=off 1=on");
+module_param(edac_pci_panic_on_pe, int, 0644);
+MODULE_PARM_DESC(edac_pci_panic_on_pe,
+ "Panic on PCI Bus Parity error: 0=off 1=on");
diff --git a/drivers/edac/fsl_ddr_edac.c b/drivers/edac/fsl_ddr_edac.c
new file mode 100644
index 0000000000..b81757555a
--- /dev/null
+++ b/drivers/edac/fsl_ddr_edac.c
@@ -0,0 +1,633 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Freescale Memory Controller kernel module
+ *
+ * Support Power-based SoCs including MPC85xx, MPC86xx, MPC83xx and
+ * ARM-based Layerscape SoCs including LS2xxx and LS1021A. Originally
+ * split out from mpc85xx_edac EDAC driver.
+ *
+ * Parts Copyrighted (c) 2013 by Freescale Semiconductor, Inc.
+ *
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2006-2007 (c) MontaVista Software, Inc.
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/ctype.h>
+#include <linux/io.h>
+#include <linux/mod_devicetable.h>
+#include <linux/edac.h>
+#include <linux/smp.h>
+#include <linux/gfp.h>
+
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include "edac_module.h"
+#include "fsl_ddr_edac.h"
+
+#define EDAC_MOD_STR "fsl_ddr_edac"
+
+static int edac_mc_idx;
+
+static u32 orig_ddr_err_disable;
+static u32 orig_ddr_err_sbe;
+static bool little_endian;
+
+static inline u32 ddr_in32(void __iomem *addr)
+{
+ return little_endian ? ioread32(addr) : ioread32be(addr);
+}
+
+static inline void ddr_out32(void __iomem *addr, u32 value)
+{
+ if (little_endian)
+ iowrite32(value, addr);
+ else
+ iowrite32be(value, addr);
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/************************ MC SYSFS parts ***********************************/
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+static ssize_t fsl_mc_inject_data_hi_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ return sprintf(data, "0x%08x",
+ ddr_in32(pdata->mc_vbase + FSL_MC_DATA_ERR_INJECT_HI));
+}
+
+static ssize_t fsl_mc_inject_data_lo_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ return sprintf(data, "0x%08x",
+ ddr_in32(pdata->mc_vbase + FSL_MC_DATA_ERR_INJECT_LO));
+}
+
+static ssize_t fsl_mc_inject_ctrl_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ return sprintf(data, "0x%08x",
+ ddr_in32(pdata->mc_vbase + FSL_MC_ECC_ERR_INJECT));
+}
+
+static ssize_t fsl_mc_inject_data_hi_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ unsigned long val;
+ int rc;
+
+ if (isdigit(*data)) {
+ rc = kstrtoul(data, 0, &val);
+ if (rc)
+ return rc;
+
+ ddr_out32(pdata->mc_vbase + FSL_MC_DATA_ERR_INJECT_HI, val);
+ return count;
+ }
+ return 0;
+}
+
+static ssize_t fsl_mc_inject_data_lo_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ unsigned long val;
+ int rc;
+
+ if (isdigit(*data)) {
+ rc = kstrtoul(data, 0, &val);
+ if (rc)
+ return rc;
+
+ ddr_out32(pdata->mc_vbase + FSL_MC_DATA_ERR_INJECT_LO, val);
+ return count;
+ }
+ return 0;
+}
+
+static ssize_t fsl_mc_inject_ctrl_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ unsigned long val;
+ int rc;
+
+ if (isdigit(*data)) {
+ rc = kstrtoul(data, 0, &val);
+ if (rc)
+ return rc;
+
+ ddr_out32(pdata->mc_vbase + FSL_MC_ECC_ERR_INJECT, val);
+ return count;
+ }
+ return 0;
+}
+
+static DEVICE_ATTR(inject_data_hi, S_IRUGO | S_IWUSR,
+ fsl_mc_inject_data_hi_show, fsl_mc_inject_data_hi_store);
+static DEVICE_ATTR(inject_data_lo, S_IRUGO | S_IWUSR,
+ fsl_mc_inject_data_lo_show, fsl_mc_inject_data_lo_store);
+static DEVICE_ATTR(inject_ctrl, S_IRUGO | S_IWUSR,
+ fsl_mc_inject_ctrl_show, fsl_mc_inject_ctrl_store);
+#endif /* CONFIG_EDAC_DEBUG */
+
+static struct attribute *fsl_ddr_dev_attrs[] = {
+#ifdef CONFIG_EDAC_DEBUG
+ &dev_attr_inject_data_hi.attr,
+ &dev_attr_inject_data_lo.attr,
+ &dev_attr_inject_ctrl.attr,
+#endif
+ NULL
+};
+
+ATTRIBUTE_GROUPS(fsl_ddr_dev);
+
+/**************************** MC Err device ***************************/
+
+/*
+ * Taken from table 8-55 in the MPC8641 User's Manual and/or 9-61 in the
+ * MPC8572 User's Manual. Each line represents a syndrome bit column as a
+ * 64-bit value, but split into an upper and lower 32-bit chunk. The labels
+ * below correspond to Freescale's manuals.
+ */
+static unsigned int ecc_table[16] = {
+ /* MSB LSB */
+ /* [0:31] [32:63] */
+ 0xf00fe11e, 0xc33c0ff7, /* Syndrome bit 7 */
+ 0x00ff00ff, 0x00fff0ff,
+ 0x0f0f0f0f, 0x0f0fff00,
+ 0x11113333, 0x7777000f,
+ 0x22224444, 0x8888222f,
+ 0x44448888, 0xffff4441,
+ 0x8888ffff, 0x11118882,
+ 0xffff1111, 0x22221114, /* Syndrome bit 0 */
+};
+
+/*
+ * Calculate the correct ECC value for a 64-bit value specified by high:low
+ */
+static u8 calculate_ecc(u32 high, u32 low)
+{
+ u32 mask_low;
+ u32 mask_high;
+ int bit_cnt;
+ u8 ecc = 0;
+ int i;
+ int j;
+
+ for (i = 0; i < 8; i++) {
+ mask_high = ecc_table[i * 2];
+ mask_low = ecc_table[i * 2 + 1];
+ bit_cnt = 0;
+
+ for (j = 0; j < 32; j++) {
+ if ((mask_high >> j) & 1)
+ bit_cnt ^= (high >> j) & 1;
+ if ((mask_low >> j) & 1)
+ bit_cnt ^= (low >> j) & 1;
+ }
+
+ ecc |= bit_cnt << i;
+ }
+
+ return ecc;
+}
+
+/*
+ * Create the syndrome code which is generated if the data line specified by
+ * 'bit' failed. Eg generate an 8-bit codes seen in Table 8-55 in the MPC8641
+ * User's Manual and 9-61 in the MPC8572 User's Manual.
+ */
+static u8 syndrome_from_bit(unsigned int bit) {
+ int i;
+ u8 syndrome = 0;
+
+ /*
+ * Cycle through the upper or lower 32-bit portion of each value in
+ * ecc_table depending on if 'bit' is in the upper or lower half of
+ * 64-bit data.
+ */
+ for (i = bit < 32; i < 16; i += 2)
+ syndrome |= ((ecc_table[i] >> (bit % 32)) & 1) << (i / 2);
+
+ return syndrome;
+}
+
+/*
+ * Decode data and ecc syndrome to determine what went wrong
+ * Note: This can only decode single-bit errors
+ */
+static void sbe_ecc_decode(u32 cap_high, u32 cap_low, u32 cap_ecc,
+ int *bad_data_bit, int *bad_ecc_bit)
+{
+ int i;
+ u8 syndrome;
+
+ *bad_data_bit = -1;
+ *bad_ecc_bit = -1;
+
+ /*
+ * Calculate the ECC of the captured data and XOR it with the captured
+ * ECC to find an ECC syndrome value we can search for
+ */
+ syndrome = calculate_ecc(cap_high, cap_low) ^ cap_ecc;
+
+ /* Check if a data line is stuck... */
+ for (i = 0; i < 64; i++) {
+ if (syndrome == syndrome_from_bit(i)) {
+ *bad_data_bit = i;
+ return;
+ }
+ }
+
+ /* If data is correct, check ECC bits for errors... */
+ for (i = 0; i < 8; i++) {
+ if ((syndrome >> i) & 0x1) {
+ *bad_ecc_bit = i;
+ return;
+ }
+ }
+}
+
+#define make64(high, low) (((u64)(high) << 32) | (low))
+
+static void fsl_mc_check(struct mem_ctl_info *mci)
+{
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ struct csrow_info *csrow;
+ u32 bus_width;
+ u32 err_detect;
+ u32 syndrome;
+ u64 err_addr;
+ u32 pfn;
+ int row_index;
+ u32 cap_high;
+ u32 cap_low;
+ int bad_data_bit;
+ int bad_ecc_bit;
+
+ err_detect = ddr_in32(pdata->mc_vbase + FSL_MC_ERR_DETECT);
+ if (!err_detect)
+ return;
+
+ fsl_mc_printk(mci, KERN_ERR, "Err Detect Register: %#8.8x\n",
+ err_detect);
+
+ /* no more processing if not ECC bit errors */
+ if (!(err_detect & (DDR_EDE_SBE | DDR_EDE_MBE))) {
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_DETECT, err_detect);
+ return;
+ }
+
+ syndrome = ddr_in32(pdata->mc_vbase + FSL_MC_CAPTURE_ECC);
+
+ /* Mask off appropriate bits of syndrome based on bus width */
+ bus_width = (ddr_in32(pdata->mc_vbase + FSL_MC_DDR_SDRAM_CFG) &
+ DSC_DBW_MASK) ? 32 : 64;
+ if (bus_width == 64)
+ syndrome &= 0xff;
+ else
+ syndrome &= 0xffff;
+
+ err_addr = make64(
+ ddr_in32(pdata->mc_vbase + FSL_MC_CAPTURE_EXT_ADDRESS),
+ ddr_in32(pdata->mc_vbase + FSL_MC_CAPTURE_ADDRESS));
+ pfn = err_addr >> PAGE_SHIFT;
+
+ for (row_index = 0; row_index < mci->nr_csrows; row_index++) {
+ csrow = mci->csrows[row_index];
+ if ((pfn >= csrow->first_page) && (pfn <= csrow->last_page))
+ break;
+ }
+
+ cap_high = ddr_in32(pdata->mc_vbase + FSL_MC_CAPTURE_DATA_HI);
+ cap_low = ddr_in32(pdata->mc_vbase + FSL_MC_CAPTURE_DATA_LO);
+
+ /*
+ * Analyze single-bit errors on 64-bit wide buses
+ * TODO: Add support for 32-bit wide buses
+ */
+ if ((err_detect & DDR_EDE_SBE) && (bus_width == 64)) {
+ sbe_ecc_decode(cap_high, cap_low, syndrome,
+ &bad_data_bit, &bad_ecc_bit);
+
+ if (bad_data_bit != -1)
+ fsl_mc_printk(mci, KERN_ERR,
+ "Faulty Data bit: %d\n", bad_data_bit);
+ if (bad_ecc_bit != -1)
+ fsl_mc_printk(mci, KERN_ERR,
+ "Faulty ECC bit: %d\n", bad_ecc_bit);
+
+ fsl_mc_printk(mci, KERN_ERR,
+ "Expected Data / ECC:\t%#8.8x_%08x / %#2.2x\n",
+ cap_high ^ (1 << (bad_data_bit - 32)),
+ cap_low ^ (1 << bad_data_bit),
+ syndrome ^ (1 << bad_ecc_bit));
+ }
+
+ fsl_mc_printk(mci, KERN_ERR,
+ "Captured Data / ECC:\t%#8.8x_%08x / %#2.2x\n",
+ cap_high, cap_low, syndrome);
+ fsl_mc_printk(mci, KERN_ERR, "Err addr: %#8.8llx\n", err_addr);
+ fsl_mc_printk(mci, KERN_ERR, "PFN: %#8.8x\n", pfn);
+
+ /* we are out of range */
+ if (row_index == mci->nr_csrows)
+ fsl_mc_printk(mci, KERN_ERR, "PFN out of range!\n");
+
+ if (err_detect & DDR_EDE_SBE)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ pfn, err_addr & ~PAGE_MASK, syndrome,
+ row_index, 0, -1,
+ mci->ctl_name, "");
+
+ if (err_detect & DDR_EDE_MBE)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ pfn, err_addr & ~PAGE_MASK, syndrome,
+ row_index, 0, -1,
+ mci->ctl_name, "");
+
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_DETECT, err_detect);
+}
+
+static irqreturn_t fsl_mc_isr(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ u32 err_detect;
+
+ err_detect = ddr_in32(pdata->mc_vbase + FSL_MC_ERR_DETECT);
+ if (!err_detect)
+ return IRQ_NONE;
+
+ fsl_mc_check(mci);
+
+ return IRQ_HANDLED;
+}
+
+static void fsl_ddr_init_csrows(struct mem_ctl_info *mci)
+{
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ u32 sdram_ctl;
+ u32 sdtype;
+ enum mem_type mtype;
+ u32 cs_bnds;
+ int index;
+
+ sdram_ctl = ddr_in32(pdata->mc_vbase + FSL_MC_DDR_SDRAM_CFG);
+
+ sdtype = sdram_ctl & DSC_SDTYPE_MASK;
+ if (sdram_ctl & DSC_RD_EN) {
+ switch (sdtype) {
+ case 0x02000000:
+ mtype = MEM_RDDR;
+ break;
+ case 0x03000000:
+ mtype = MEM_RDDR2;
+ break;
+ case 0x07000000:
+ mtype = MEM_RDDR3;
+ break;
+ case 0x05000000:
+ mtype = MEM_RDDR4;
+ break;
+ default:
+ mtype = MEM_UNKNOWN;
+ break;
+ }
+ } else {
+ switch (sdtype) {
+ case 0x02000000:
+ mtype = MEM_DDR;
+ break;
+ case 0x03000000:
+ mtype = MEM_DDR2;
+ break;
+ case 0x07000000:
+ mtype = MEM_DDR3;
+ break;
+ case 0x05000000:
+ mtype = MEM_DDR4;
+ break;
+ default:
+ mtype = MEM_UNKNOWN;
+ break;
+ }
+ }
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ u32 start;
+ u32 end;
+
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ cs_bnds = ddr_in32(pdata->mc_vbase + FSL_MC_CS_BNDS_0 +
+ (index * FSL_MC_CS_BNDS_OFS));
+
+ start = (cs_bnds & 0xffff0000) >> 16;
+ end = (cs_bnds & 0x0000ffff);
+
+ if (start == end)
+ continue; /* not populated */
+
+ start <<= (24 - PAGE_SHIFT);
+ end <<= (24 - PAGE_SHIFT);
+ end |= (1 << (24 - PAGE_SHIFT)) - 1;
+
+ csrow->first_page = start;
+ csrow->last_page = end;
+
+ dimm->nr_pages = end + 1 - start;
+ dimm->grain = 8;
+ dimm->mtype = mtype;
+ dimm->dtype = DEV_UNKNOWN;
+ if (sdram_ctl & DSC_X32_EN)
+ dimm->dtype = DEV_X32;
+ dimm->edac_mode = EDAC_SECDED;
+ }
+}
+
+int fsl_mc_err_probe(struct platform_device *op)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct fsl_mc_pdata *pdata;
+ struct resource r;
+ u32 sdram_ctl;
+ int res;
+
+ if (!devres_open_group(&op->dev, fsl_mc_err_probe, GFP_KERNEL))
+ return -ENOMEM;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 4;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(edac_mc_idx, ARRAY_SIZE(layers), layers,
+ sizeof(*pdata));
+ if (!mci) {
+ devres_release_group(&op->dev, fsl_mc_err_probe);
+ return -ENOMEM;
+ }
+
+ pdata = mci->pvt_info;
+ pdata->name = "fsl_mc_err";
+ mci->pdev = &op->dev;
+ pdata->edac_idx = edac_mc_idx++;
+ dev_set_drvdata(mci->pdev, mci);
+ mci->ctl_name = pdata->name;
+ mci->dev_name = pdata->name;
+
+ /*
+ * Get the endianness of DDR controller registers.
+ * Default is big endian.
+ */
+ little_endian = of_property_read_bool(op->dev.of_node, "little-endian");
+
+ res = of_address_to_resource(op->dev.of_node, 0, &r);
+ if (res) {
+ pr_err("%s: Unable to get resource for MC err regs\n",
+ __func__);
+ goto err;
+ }
+
+ if (!devm_request_mem_region(&op->dev, r.start, resource_size(&r),
+ pdata->name)) {
+ pr_err("%s: Error while requesting mem region\n",
+ __func__);
+ res = -EBUSY;
+ goto err;
+ }
+
+ pdata->mc_vbase = devm_ioremap(&op->dev, r.start, resource_size(&r));
+ if (!pdata->mc_vbase) {
+ pr_err("%s: Unable to setup MC err regs\n", __func__);
+ res = -ENOMEM;
+ goto err;
+ }
+
+ sdram_ctl = ddr_in32(pdata->mc_vbase + FSL_MC_DDR_SDRAM_CFG);
+ if (!(sdram_ctl & DSC_ECC_EN)) {
+ /* no ECC */
+ pr_warn("%s: No ECC DIMMs discovered\n", __func__);
+ res = -ENODEV;
+ goto err;
+ }
+
+ edac_dbg(3, "init mci\n");
+ mci->mtype_cap = MEM_FLAG_DDR | MEM_FLAG_RDDR |
+ MEM_FLAG_DDR2 | MEM_FLAG_RDDR2 |
+ MEM_FLAG_DDR3 | MEM_FLAG_RDDR3 |
+ MEM_FLAG_DDR4 | MEM_FLAG_RDDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ mci->edac_check = fsl_mc_check;
+
+ mci->ctl_page_to_phys = NULL;
+
+ mci->scrub_mode = SCRUB_SW_SRC;
+
+ fsl_ddr_init_csrows(mci);
+
+ /* store the original error disable bits */
+ orig_ddr_err_disable = ddr_in32(pdata->mc_vbase + FSL_MC_ERR_DISABLE);
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_DISABLE, 0);
+
+ /* clear all error bits */
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_DETECT, ~0);
+
+ res = edac_mc_add_mc_with_groups(mci, fsl_ddr_dev_groups);
+ if (res) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto err;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_INT_EN,
+ DDR_EIE_MBEE | DDR_EIE_SBEE);
+
+ /* store the original error management threshold */
+ orig_ddr_err_sbe = ddr_in32(pdata->mc_vbase +
+ FSL_MC_ERR_SBE) & 0xff0000;
+
+ /* set threshold to 1 error per interrupt */
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_SBE, 0x10000);
+
+ /* register interrupts */
+ pdata->irq = platform_get_irq(op, 0);
+ res = devm_request_irq(&op->dev, pdata->irq,
+ fsl_mc_isr,
+ IRQF_SHARED,
+ "[EDAC] MC err", mci);
+ if (res < 0) {
+ pr_err("%s: Unable to request irq %d for FSL DDR DRAM ERR\n",
+ __func__, pdata->irq);
+ res = -ENODEV;
+ goto err2;
+ }
+
+ pr_info(EDAC_MOD_STR " acquired irq %d for MC\n",
+ pdata->irq);
+ }
+
+ devres_remove_group(&op->dev, fsl_mc_err_probe);
+ edac_dbg(3, "success\n");
+ pr_info(EDAC_MOD_STR " MC err registered\n");
+
+ return 0;
+
+err2:
+ edac_mc_del_mc(&op->dev);
+err:
+ devres_release_group(&op->dev, fsl_mc_err_probe);
+ edac_mc_free(mci);
+ return res;
+}
+
+int fsl_mc_err_remove(struct platform_device *op)
+{
+ struct mem_ctl_info *mci = dev_get_drvdata(&op->dev);
+ struct fsl_mc_pdata *pdata = mci->pvt_info;
+
+ edac_dbg(0, "\n");
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_INT_EN, 0);
+ }
+
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_DISABLE,
+ orig_ddr_err_disable);
+ ddr_out32(pdata->mc_vbase + FSL_MC_ERR_SBE, orig_ddr_err_sbe);
+
+ edac_mc_del_mc(&op->dev);
+ edac_mc_free(mci);
+ return 0;
+}
diff --git a/drivers/edac/fsl_ddr_edac.h b/drivers/edac/fsl_ddr_edac.h
new file mode 100644
index 0000000000..332439d7b2
--- /dev/null
+++ b/drivers/edac/fsl_ddr_edac.h
@@ -0,0 +1,76 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Freescale Memory Controller kernel module
+ *
+ * Support Power-based SoCs including MPC85xx, MPC86xx, MPC83xx and
+ * ARM-based Layerscape SoCs including LS2xxx and LS1021A. Originally
+ * split out from mpc85xx_edac EDAC driver.
+ *
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2006-2007 (c) MontaVista Software, Inc.
+ */
+#ifndef _FSL_DDR_EDAC_H_
+#define _FSL_DDR_EDAC_H_
+
+#define fsl_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "FSL_DDR", fmt, ##arg)
+
+/*
+ * DRAM error defines
+ */
+
+/* DDR_SDRAM_CFG */
+#define FSL_MC_DDR_SDRAM_CFG 0x0110
+#define FSL_MC_CS_BNDS_0 0x0000
+#define FSL_MC_CS_BNDS_OFS 0x0008
+
+#define FSL_MC_DATA_ERR_INJECT_HI 0x0e00
+#define FSL_MC_DATA_ERR_INJECT_LO 0x0e04
+#define FSL_MC_ECC_ERR_INJECT 0x0e08
+#define FSL_MC_CAPTURE_DATA_HI 0x0e20
+#define FSL_MC_CAPTURE_DATA_LO 0x0e24
+#define FSL_MC_CAPTURE_ECC 0x0e28
+#define FSL_MC_ERR_DETECT 0x0e40
+#define FSL_MC_ERR_DISABLE 0x0e44
+#define FSL_MC_ERR_INT_EN 0x0e48
+#define FSL_MC_CAPTURE_ATRIBUTES 0x0e4c
+#define FSL_MC_CAPTURE_ADDRESS 0x0e50
+#define FSL_MC_CAPTURE_EXT_ADDRESS 0x0e54
+#define FSL_MC_ERR_SBE 0x0e58
+
+#define DSC_MEM_EN 0x80000000
+#define DSC_ECC_EN 0x20000000
+#define DSC_RD_EN 0x10000000
+#define DSC_DBW_MASK 0x00180000
+#define DSC_DBW_32 0x00080000
+#define DSC_DBW_64 0x00000000
+
+#define DSC_SDTYPE_MASK 0x07000000
+#define DSC_X32_EN 0x00000020
+
+/* Err_Int_En */
+#define DDR_EIE_MSEE 0x1 /* memory select */
+#define DDR_EIE_SBEE 0x4 /* single-bit ECC error */
+#define DDR_EIE_MBEE 0x8 /* multi-bit ECC error */
+
+/* Err_Detect */
+#define DDR_EDE_MSE 0x1 /* memory select */
+#define DDR_EDE_SBE 0x4 /* single-bit ECC error */
+#define DDR_EDE_MBE 0x8 /* multi-bit ECC error */
+#define DDR_EDE_MME 0x80000000 /* multiple memory errors */
+
+/* Err_Disable */
+#define DDR_EDI_MSED 0x1 /* memory select disable */
+#define DDR_EDI_SBED 0x4 /* single-bit ECC error disable */
+#define DDR_EDI_MBED 0x8 /* multi-bit ECC error disable */
+
+struct fsl_mc_pdata {
+ char *name;
+ int edac_idx;
+ void __iomem *mc_vbase;
+ int irq;
+};
+int fsl_mc_err_probe(struct platform_device *op);
+int fsl_mc_err_remove(struct platform_device *op);
+#endif
diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c
new file mode 100644
index 0000000000..cf2b618c1a
--- /dev/null
+++ b/drivers/edac/ghes_edac.c
@@ -0,0 +1,573 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * GHES/EDAC Linux driver
+ *
+ * Copyright (c) 2013 by Mauro Carvalho Chehab
+ *
+ * Red Hat Inc. https://www.redhat.com
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <acpi/ghes.h>
+#include <linux/edac.h>
+#include <linux/dmi.h>
+#include "edac_module.h"
+#include <ras/ras_event.h>
+#include <linux/notifier.h>
+
+#define OTHER_DETAIL_LEN 400
+
+struct ghes_pvt {
+ struct mem_ctl_info *mci;
+
+ /* Buffers for the error handling routine */
+ char other_detail[OTHER_DETAIL_LEN];
+ char msg[80];
+};
+
+static refcount_t ghes_refcount = REFCOUNT_INIT(0);
+
+/*
+ * Access to ghes_pvt must be protected by ghes_lock. The spinlock
+ * also provides the necessary (implicit) memory barrier for the SMP
+ * case to make the pointer visible on another CPU.
+ */
+static struct ghes_pvt *ghes_pvt;
+
+/*
+ * This driver's representation of the system hardware, as collected
+ * from DMI.
+ */
+static struct ghes_hw_desc {
+ int num_dimms;
+ struct dimm_info *dimms;
+} ghes_hw;
+
+/* GHES registration mutex */
+static DEFINE_MUTEX(ghes_reg_mutex);
+
+/*
+ * Sync with other, potentially concurrent callers of
+ * ghes_edac_report_mem_error(). We don't know what the
+ * "inventive" firmware would do.
+ */
+static DEFINE_SPINLOCK(ghes_lock);
+
+static bool system_scanned;
+
+static struct list_head *ghes_devs;
+
+/* Memory Device - Type 17 of SMBIOS spec */
+struct memdev_dmi_entry {
+ u8 type;
+ u8 length;
+ u16 handle;
+ u16 phys_mem_array_handle;
+ u16 mem_err_info_handle;
+ u16 total_width;
+ u16 data_width;
+ u16 size;
+ u8 form_factor;
+ u8 device_set;
+ u8 device_locator;
+ u8 bank_locator;
+ u8 memory_type;
+ u16 type_detail;
+ u16 speed;
+ u8 manufacturer;
+ u8 serial_number;
+ u8 asset_tag;
+ u8 part_number;
+ u8 attributes;
+ u32 extended_size;
+ u16 conf_mem_clk_speed;
+} __attribute__((__packed__));
+
+static struct dimm_info *find_dimm_by_handle(struct mem_ctl_info *mci, u16 handle)
+{
+ struct dimm_info *dimm;
+
+ mci_for_each_dimm(mci, dimm) {
+ if (dimm->smbios_handle == handle)
+ return dimm;
+ }
+
+ return NULL;
+}
+
+static void dimm_setup_label(struct dimm_info *dimm, u16 handle)
+{
+ const char *bank = NULL, *device = NULL;
+
+ dmi_memdev_name(handle, &bank, &device);
+
+ /*
+ * Set to a NULL string when both bank and device are zero. In this case,
+ * the label assigned by default will be preserved.
+ */
+ snprintf(dimm->label, sizeof(dimm->label), "%s%s%s",
+ (bank && *bank) ? bank : "",
+ (bank && *bank && device && *device) ? " " : "",
+ (device && *device) ? device : "");
+}
+
+static void assign_dmi_dimm_info(struct dimm_info *dimm, struct memdev_dmi_entry *entry)
+{
+ u16 rdr_mask = BIT(7) | BIT(13);
+
+ if (entry->size == 0xffff) {
+ pr_info("Can't get DIMM%i size\n", dimm->idx);
+ dimm->nr_pages = MiB_TO_PAGES(32);/* Unknown */
+ } else if (entry->size == 0x7fff) {
+ dimm->nr_pages = MiB_TO_PAGES(entry->extended_size);
+ } else {
+ if (entry->size & BIT(15))
+ dimm->nr_pages = MiB_TO_PAGES((entry->size & 0x7fff) << 10);
+ else
+ dimm->nr_pages = MiB_TO_PAGES(entry->size);
+ }
+
+ switch (entry->memory_type) {
+ case 0x12:
+ if (entry->type_detail & BIT(13))
+ dimm->mtype = MEM_RDDR;
+ else
+ dimm->mtype = MEM_DDR;
+ break;
+ case 0x13:
+ if (entry->type_detail & BIT(13))
+ dimm->mtype = MEM_RDDR2;
+ else
+ dimm->mtype = MEM_DDR2;
+ break;
+ case 0x14:
+ dimm->mtype = MEM_FB_DDR2;
+ break;
+ case 0x18:
+ if (entry->type_detail & BIT(12))
+ dimm->mtype = MEM_NVDIMM;
+ else if (entry->type_detail & BIT(13))
+ dimm->mtype = MEM_RDDR3;
+ else
+ dimm->mtype = MEM_DDR3;
+ break;
+ case 0x1a:
+ if (entry->type_detail & BIT(12))
+ dimm->mtype = MEM_NVDIMM;
+ else if (entry->type_detail & BIT(13))
+ dimm->mtype = MEM_RDDR4;
+ else
+ dimm->mtype = MEM_DDR4;
+ break;
+ default:
+ if (entry->type_detail & BIT(6))
+ dimm->mtype = MEM_RMBS;
+ else if ((entry->type_detail & rdr_mask) == rdr_mask)
+ dimm->mtype = MEM_RDR;
+ else if (entry->type_detail & BIT(7))
+ dimm->mtype = MEM_SDR;
+ else if (entry->type_detail & BIT(9))
+ dimm->mtype = MEM_EDO;
+ else
+ dimm->mtype = MEM_UNKNOWN;
+ }
+
+ /*
+ * Actually, we can only detect if the memory has bits for
+ * checksum or not
+ */
+ if (entry->total_width == entry->data_width)
+ dimm->edac_mode = EDAC_NONE;
+ else
+ dimm->edac_mode = EDAC_SECDED;
+
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->grain = 128; /* Likely, worse case */
+
+ dimm_setup_label(dimm, entry->handle);
+
+ if (dimm->nr_pages) {
+ edac_dbg(1, "DIMM%i: %s size = %d MB%s\n",
+ dimm->idx, edac_mem_types[dimm->mtype],
+ PAGES_TO_MiB(dimm->nr_pages),
+ (dimm->edac_mode != EDAC_NONE) ? "(ECC)" : "");
+ edac_dbg(2, "\ttype %d, detail 0x%02x, width %d(total %d)\n",
+ entry->memory_type, entry->type_detail,
+ entry->total_width, entry->data_width);
+ }
+
+ dimm->smbios_handle = entry->handle;
+}
+
+static void enumerate_dimms(const struct dmi_header *dh, void *arg)
+{
+ struct memdev_dmi_entry *entry = (struct memdev_dmi_entry *)dh;
+ struct ghes_hw_desc *hw = (struct ghes_hw_desc *)arg;
+ struct dimm_info *d;
+
+ if (dh->type != DMI_ENTRY_MEM_DEVICE)
+ return;
+
+ /* Enlarge the array with additional 16 */
+ if (!hw->num_dimms || !(hw->num_dimms % 16)) {
+ struct dimm_info *new;
+
+ new = krealloc_array(hw->dimms, hw->num_dimms + 16,
+ sizeof(struct dimm_info), GFP_KERNEL);
+ if (!new) {
+ WARN_ON_ONCE(1);
+ return;
+ }
+
+ hw->dimms = new;
+ }
+
+ d = &hw->dimms[hw->num_dimms];
+ d->idx = hw->num_dimms;
+
+ assign_dmi_dimm_info(d, entry);
+
+ hw->num_dimms++;
+}
+
+static void ghes_scan_system(void)
+{
+ if (system_scanned)
+ return;
+
+ dmi_walk(enumerate_dimms, &ghes_hw);
+
+ system_scanned = true;
+}
+
+static int print_mem_error_other_detail(const struct cper_sec_mem_err *mem, char *msg,
+ const char *location, unsigned int len)
+{
+ u32 n;
+
+ if (!msg)
+ return 0;
+
+ n = 0;
+ len -= 1;
+
+ n += scnprintf(msg + n, len - n, "APEI location: %s ", location);
+
+ if (!(mem->validation_bits & CPER_MEM_VALID_ERROR_STATUS))
+ goto out;
+
+ n += scnprintf(msg + n, len - n, "status(0x%016llx): ", mem->error_status);
+ n += scnprintf(msg + n, len - n, "%s ", cper_mem_err_status_str(mem->error_status));
+
+out:
+ msg[n] = '\0';
+
+ return n;
+}
+
+static int ghes_edac_report_mem_error(struct notifier_block *nb,
+ unsigned long val, void *data)
+{
+ struct cper_sec_mem_err *mem_err = (struct cper_sec_mem_err *)data;
+ struct cper_mem_err_compact cmem;
+ struct edac_raw_error_desc *e;
+ struct mem_ctl_info *mci;
+ unsigned long sev = val;
+ struct ghes_pvt *pvt;
+ unsigned long flags;
+ char *p;
+
+ /*
+ * We can do the locking below because GHES defers error processing
+ * from NMI to IRQ context. Whenever that changes, we'd at least
+ * know.
+ */
+ if (WARN_ON_ONCE(in_nmi()))
+ return NOTIFY_OK;
+
+ spin_lock_irqsave(&ghes_lock, flags);
+
+ pvt = ghes_pvt;
+ if (!pvt)
+ goto unlock;
+
+ mci = pvt->mci;
+ e = &mci->error_desc;
+
+ /* Cleans the error report buffer */
+ memset(e, 0, sizeof (*e));
+ e->error_count = 1;
+ e->grain = 1;
+ e->msg = pvt->msg;
+ e->other_detail = pvt->other_detail;
+ e->top_layer = -1;
+ e->mid_layer = -1;
+ e->low_layer = -1;
+ *pvt->other_detail = '\0';
+ *pvt->msg = '\0';
+
+ switch (sev) {
+ case GHES_SEV_CORRECTED:
+ e->type = HW_EVENT_ERR_CORRECTED;
+ break;
+ case GHES_SEV_RECOVERABLE:
+ e->type = HW_EVENT_ERR_UNCORRECTED;
+ break;
+ case GHES_SEV_PANIC:
+ e->type = HW_EVENT_ERR_FATAL;
+ break;
+ default:
+ case GHES_SEV_NO:
+ e->type = HW_EVENT_ERR_INFO;
+ }
+
+ edac_dbg(1, "error validation_bits: 0x%08llx\n",
+ (long long)mem_err->validation_bits);
+
+ /* Error type, mapped on e->msg */
+ if (mem_err->validation_bits & CPER_MEM_VALID_ERROR_TYPE) {
+ u8 etype = mem_err->error_type;
+
+ p = pvt->msg;
+ p += snprintf(p, sizeof(pvt->msg), "%s", cper_mem_err_type_str(etype));
+ } else {
+ strcpy(pvt->msg, "unknown error");
+ }
+
+ /* Error address */
+ if (mem_err->validation_bits & CPER_MEM_VALID_PA) {
+ e->page_frame_number = PHYS_PFN(mem_err->physical_addr);
+ e->offset_in_page = offset_in_page(mem_err->physical_addr);
+ }
+
+ /* Error grain */
+ if (mem_err->validation_bits & CPER_MEM_VALID_PA_MASK)
+ e->grain = ~mem_err->physical_addr_mask + 1;
+
+ /* Memory error location, mapped on e->location */
+ p = e->location;
+ cper_mem_err_pack(mem_err, &cmem);
+ p += cper_mem_err_location(&cmem, p);
+
+ if (mem_err->validation_bits & CPER_MEM_VALID_MODULE_HANDLE) {
+ struct dimm_info *dimm;
+
+ p += cper_dimm_err_location(&cmem, p);
+ dimm = find_dimm_by_handle(mci, mem_err->mem_dev_handle);
+ if (dimm) {
+ e->top_layer = dimm->idx;
+ strcpy(e->label, dimm->label);
+ }
+ }
+ if (p > e->location)
+ *(p - 1) = '\0';
+
+ if (!*e->label)
+ strcpy(e->label, "unknown memory");
+
+ /* All other fields are mapped on e->other_detail */
+ p = pvt->other_detail;
+ p += print_mem_error_other_detail(mem_err, p, e->location, OTHER_DETAIL_LEN);
+ if (p > pvt->other_detail)
+ *(p - 1) = '\0';
+
+ edac_raw_mc_handle_error(e);
+
+unlock:
+ spin_unlock_irqrestore(&ghes_lock, flags);
+
+ return NOTIFY_OK;
+}
+
+static struct notifier_block ghes_edac_mem_err_nb = {
+ .notifier_call = ghes_edac_report_mem_error,
+ .priority = 0,
+};
+
+static int ghes_edac_register(struct device *dev)
+{
+ bool fake = false;
+ struct mem_ctl_info *mci;
+ struct ghes_pvt *pvt;
+ struct edac_mc_layer layers[1];
+ unsigned long flags;
+ int rc = 0;
+
+ /* finish another registration/unregistration instance first */
+ mutex_lock(&ghes_reg_mutex);
+
+ /*
+ * We have only one logical memory controller to which all DIMMs belong.
+ */
+ if (refcount_inc_not_zero(&ghes_refcount))
+ goto unlock;
+
+ ghes_scan_system();
+
+ /* Check if we've got a bogus BIOS */
+ if (!ghes_hw.num_dimms) {
+ fake = true;
+ ghes_hw.num_dimms = 1;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_ALL_MEM;
+ layers[0].size = ghes_hw.num_dimms;
+ layers[0].is_virt_csrow = true;
+
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(struct ghes_pvt));
+ if (!mci) {
+ pr_info("Can't allocate memory for EDAC data\n");
+ rc = -ENOMEM;
+ goto unlock;
+ }
+
+ pvt = mci->pvt_info;
+ pvt->mci = mci;
+
+ mci->pdev = dev;
+ mci->mtype_cap = MEM_FLAG_EMPTY;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "ghes_edac.c";
+ mci->ctl_name = "ghes_edac";
+ mci->dev_name = "ghes";
+
+ if (fake) {
+ pr_info("This system has a very crappy BIOS: It doesn't even list the DIMMS.\n");
+ pr_info("Its SMBIOS info is wrong. It is doubtful that the error report would\n");
+ pr_info("work on such system. Use this driver with caution\n");
+ }
+
+ pr_info("This system has %d DIMM sockets.\n", ghes_hw.num_dimms);
+
+ if (!fake) {
+ struct dimm_info *src, *dst;
+ int i = 0;
+
+ mci_for_each_dimm(mci, dst) {
+ src = &ghes_hw.dimms[i];
+
+ dst->idx = src->idx;
+ dst->smbios_handle = src->smbios_handle;
+ dst->nr_pages = src->nr_pages;
+ dst->mtype = src->mtype;
+ dst->edac_mode = src->edac_mode;
+ dst->dtype = src->dtype;
+ dst->grain = src->grain;
+
+ /*
+ * If no src->label, preserve default label assigned
+ * from EDAC core.
+ */
+ if (strlen(src->label))
+ memcpy(dst->label, src->label, sizeof(src->label));
+
+ i++;
+ }
+
+ } else {
+ struct dimm_info *dimm = edac_get_dimm(mci, 0, 0, 0);
+
+ dimm->nr_pages = 1;
+ dimm->grain = 128;
+ dimm->mtype = MEM_UNKNOWN;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_SECDED;
+ }
+
+ rc = edac_mc_add_mc(mci);
+ if (rc < 0) {
+ pr_info("Can't register with the EDAC core\n");
+ edac_mc_free(mci);
+ rc = -ENODEV;
+ goto unlock;
+ }
+
+ spin_lock_irqsave(&ghes_lock, flags);
+ ghes_pvt = pvt;
+ spin_unlock_irqrestore(&ghes_lock, flags);
+
+ ghes_register_report_chain(&ghes_edac_mem_err_nb);
+
+ /* only set on success */
+ refcount_set(&ghes_refcount, 1);
+
+unlock:
+
+ /* Not needed anymore */
+ kfree(ghes_hw.dimms);
+ ghes_hw.dimms = NULL;
+
+ mutex_unlock(&ghes_reg_mutex);
+
+ return rc;
+}
+
+static void ghes_edac_unregister(struct ghes *ghes)
+{
+ struct mem_ctl_info *mci;
+ unsigned long flags;
+
+ mutex_lock(&ghes_reg_mutex);
+
+ system_scanned = false;
+ memset(&ghes_hw, 0, sizeof(struct ghes_hw_desc));
+
+ if (!refcount_dec_and_test(&ghes_refcount))
+ goto unlock;
+
+ /*
+ * Wait for the irq handler being finished.
+ */
+ spin_lock_irqsave(&ghes_lock, flags);
+ mci = ghes_pvt ? ghes_pvt->mci : NULL;
+ ghes_pvt = NULL;
+ spin_unlock_irqrestore(&ghes_lock, flags);
+
+ if (!mci)
+ goto unlock;
+
+ mci = edac_mc_del_mc(mci->pdev);
+ if (mci)
+ edac_mc_free(mci);
+
+ ghes_unregister_report_chain(&ghes_edac_mem_err_nb);
+
+unlock:
+ mutex_unlock(&ghes_reg_mutex);
+}
+
+static int __init ghes_edac_init(void)
+{
+ struct ghes *g, *g_tmp;
+
+ ghes_devs = ghes_get_devices();
+ if (!ghes_devs)
+ return -ENODEV;
+
+ if (list_empty(ghes_devs)) {
+ pr_info("GHES probing device list is empty");
+ return -ENODEV;
+ }
+
+ list_for_each_entry_safe(g, g_tmp, ghes_devs, elist) {
+ ghes_edac_register(g->dev);
+ }
+
+ return 0;
+}
+module_init(ghes_edac_init);
+
+static void __exit ghes_edac_exit(void)
+{
+ struct ghes *g, *g_tmp;
+
+ list_for_each_entry_safe(g, g_tmp, ghes_devs, elist) {
+ ghes_edac_unregister(g);
+ }
+}
+module_exit(ghes_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Output ACPI APEI/GHES BIOS detected errors via EDAC");
diff --git a/drivers/edac/highbank_l2_edac.c b/drivers/edac/highbank_l2_edac.c
new file mode 100644
index 0000000000..140d4431bd
--- /dev/null
+++ b/drivers/edac/highbank_l2_edac.c
@@ -0,0 +1,143 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2011-2012 Calxeda, Inc.
+ */
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/ctype.h>
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+
+#include "edac_module.h"
+
+#define SR_CLR_SB_ECC_INTR 0x0
+#define SR_CLR_DB_ECC_INTR 0x4
+
+struct hb_l2_drvdata {
+ void __iomem *base;
+ int sb_irq;
+ int db_irq;
+};
+
+static irqreturn_t highbank_l2_err_handler(int irq, void *dev_id)
+{
+ struct edac_device_ctl_info *dci = dev_id;
+ struct hb_l2_drvdata *drvdata = dci->pvt_info;
+
+ if (irq == drvdata->sb_irq) {
+ writel(1, drvdata->base + SR_CLR_SB_ECC_INTR);
+ edac_device_handle_ce(dci, 0, 0, dci->ctl_name);
+ }
+ if (irq == drvdata->db_irq) {
+ writel(1, drvdata->base + SR_CLR_DB_ECC_INTR);
+ edac_device_handle_ue(dci, 0, 0, dci->ctl_name);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static const struct of_device_id hb_l2_err_of_match[] = {
+ { .compatible = "calxeda,hb-sregs-l2-ecc", },
+ {},
+};
+MODULE_DEVICE_TABLE(of, hb_l2_err_of_match);
+
+static int highbank_l2_err_probe(struct platform_device *pdev)
+{
+ const struct of_device_id *id;
+ struct edac_device_ctl_info *dci;
+ struct hb_l2_drvdata *drvdata;
+ struct resource *r;
+ int res = 0;
+
+ dci = edac_device_alloc_ctl_info(sizeof(*drvdata), "cpu",
+ 1, "L", 1, 2, NULL, 0, 0);
+ if (!dci)
+ return -ENOMEM;
+
+ drvdata = dci->pvt_info;
+ dci->dev = &pdev->dev;
+ platform_set_drvdata(pdev, dci);
+
+ if (!devres_open_group(&pdev->dev, NULL, GFP_KERNEL))
+ return -ENOMEM;
+
+ r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!r) {
+ dev_err(&pdev->dev, "Unable to get mem resource\n");
+ res = -ENODEV;
+ goto err;
+ }
+
+ if (!devm_request_mem_region(&pdev->dev, r->start,
+ resource_size(r), dev_name(&pdev->dev))) {
+ dev_err(&pdev->dev, "Error while requesting mem region\n");
+ res = -EBUSY;
+ goto err;
+ }
+
+ drvdata->base = devm_ioremap(&pdev->dev, r->start, resource_size(r));
+ if (!drvdata->base) {
+ dev_err(&pdev->dev, "Unable to map regs\n");
+ res = -ENOMEM;
+ goto err;
+ }
+
+ id = of_match_device(hb_l2_err_of_match, &pdev->dev);
+ dci->mod_name = pdev->dev.driver->name;
+ dci->ctl_name = id ? id->compatible : "unknown";
+ dci->dev_name = dev_name(&pdev->dev);
+
+ if (edac_device_add_device(dci))
+ goto err;
+
+ drvdata->db_irq = platform_get_irq(pdev, 0);
+ res = devm_request_irq(&pdev->dev, drvdata->db_irq,
+ highbank_l2_err_handler,
+ 0, dev_name(&pdev->dev), dci);
+ if (res < 0)
+ goto err2;
+
+ drvdata->sb_irq = platform_get_irq(pdev, 1);
+ res = devm_request_irq(&pdev->dev, drvdata->sb_irq,
+ highbank_l2_err_handler,
+ 0, dev_name(&pdev->dev), dci);
+ if (res < 0)
+ goto err2;
+
+ devres_close_group(&pdev->dev, NULL);
+ return 0;
+err2:
+ edac_device_del_device(&pdev->dev);
+err:
+ devres_release_group(&pdev->dev, NULL);
+ edac_device_free_ctl_info(dci);
+ return res;
+}
+
+static int highbank_l2_err_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);
+ return 0;
+}
+
+static struct platform_driver highbank_l2_edac_driver = {
+ .probe = highbank_l2_err_probe,
+ .remove = highbank_l2_err_remove,
+ .driver = {
+ .name = "hb_l2_edac",
+ .of_match_table = hb_l2_err_of_match,
+ },
+};
+
+module_platform_driver(highbank_l2_edac_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Calxeda, Inc.");
+MODULE_DESCRIPTION("EDAC Driver for Calxeda Highbank L2 Cache");
diff --git a/drivers/edac/highbank_mc_edac.c b/drivers/edac/highbank_mc_edac.c
new file mode 100644
index 0000000000..a0c04a7f95
--- /dev/null
+++ b/drivers/edac/highbank_mc_edac.c
@@ -0,0 +1,276 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2011-2012 Calxeda, Inc.
+ */
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/ctype.h>
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/uaccess.h>
+
+#include "edac_module.h"
+
+/* DDR Ctrlr Error Registers */
+
+#define HB_DDR_ECC_ERR_BASE 0x128
+#define MW_DDR_ECC_ERR_BASE 0x1b4
+
+#define HB_DDR_ECC_OPT 0x00
+#define HB_DDR_ECC_U_ERR_ADDR 0x08
+#define HB_DDR_ECC_U_ERR_STAT 0x0c
+#define HB_DDR_ECC_U_ERR_DATAL 0x10
+#define HB_DDR_ECC_U_ERR_DATAH 0x14
+#define HB_DDR_ECC_C_ERR_ADDR 0x18
+#define HB_DDR_ECC_C_ERR_STAT 0x1c
+#define HB_DDR_ECC_C_ERR_DATAL 0x20
+#define HB_DDR_ECC_C_ERR_DATAH 0x24
+
+#define HB_DDR_ECC_OPT_MODE_MASK 0x3
+#define HB_DDR_ECC_OPT_FWC 0x100
+#define HB_DDR_ECC_OPT_XOR_SHIFT 16
+
+/* DDR Ctrlr Interrupt Registers */
+
+#define HB_DDR_ECC_INT_BASE 0x180
+#define MW_DDR_ECC_INT_BASE 0x218
+
+#define HB_DDR_ECC_INT_STATUS 0x00
+#define HB_DDR_ECC_INT_ACK 0x04
+
+#define HB_DDR_ECC_INT_STAT_CE 0x8
+#define HB_DDR_ECC_INT_STAT_DOUBLE_CE 0x10
+#define HB_DDR_ECC_INT_STAT_UE 0x20
+#define HB_DDR_ECC_INT_STAT_DOUBLE_UE 0x40
+
+struct hb_mc_drvdata {
+ void __iomem *mc_err_base;
+ void __iomem *mc_int_base;
+};
+
+static irqreturn_t highbank_mc_err_handler(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct hb_mc_drvdata *drvdata = mci->pvt_info;
+ u32 status, err_addr;
+
+ /* Read the interrupt status register */
+ status = readl(drvdata->mc_int_base + HB_DDR_ECC_INT_STATUS);
+
+ if (status & HB_DDR_ECC_INT_STAT_UE) {
+ err_addr = readl(drvdata->mc_err_base + HB_DDR_ECC_U_ERR_ADDR);
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ err_addr >> PAGE_SHIFT,
+ err_addr & ~PAGE_MASK, 0,
+ 0, 0, -1,
+ mci->ctl_name, "");
+ }
+ if (status & HB_DDR_ECC_INT_STAT_CE) {
+ u32 syndrome = readl(drvdata->mc_err_base + HB_DDR_ECC_C_ERR_STAT);
+ syndrome = (syndrome >> 8) & 0xff;
+ err_addr = readl(drvdata->mc_err_base + HB_DDR_ECC_C_ERR_ADDR);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ err_addr >> PAGE_SHIFT,
+ err_addr & ~PAGE_MASK, syndrome,
+ 0, 0, -1,
+ mci->ctl_name, "");
+ }
+
+ /* clear the error, clears the interrupt */
+ writel(status, drvdata->mc_int_base + HB_DDR_ECC_INT_ACK);
+ return IRQ_HANDLED;
+}
+
+static void highbank_mc_err_inject(struct mem_ctl_info *mci, u8 synd)
+{
+ struct hb_mc_drvdata *pdata = mci->pvt_info;
+ u32 reg;
+
+ reg = readl(pdata->mc_err_base + HB_DDR_ECC_OPT);
+ reg &= HB_DDR_ECC_OPT_MODE_MASK;
+ reg |= (synd << HB_DDR_ECC_OPT_XOR_SHIFT) | HB_DDR_ECC_OPT_FWC;
+ writel(reg, pdata->mc_err_base + HB_DDR_ECC_OPT);
+}
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+static ssize_t highbank_mc_inject_ctrl(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ u8 synd;
+
+ if (kstrtou8(buf, 16, &synd))
+ return -EINVAL;
+
+ highbank_mc_err_inject(mci, synd);
+
+ return count;
+}
+
+static DEVICE_ATTR(inject_ctrl, S_IWUSR, NULL, highbank_mc_inject_ctrl);
+
+static struct attribute *highbank_dev_attrs[] = {
+ &dev_attr_inject_ctrl.attr,
+ NULL
+};
+
+ATTRIBUTE_GROUPS(highbank_dev);
+
+struct hb_mc_settings {
+ int err_offset;
+ int int_offset;
+};
+
+static struct hb_mc_settings hb_settings = {
+ .err_offset = HB_DDR_ECC_ERR_BASE,
+ .int_offset = HB_DDR_ECC_INT_BASE,
+};
+
+static struct hb_mc_settings mw_settings = {
+ .err_offset = MW_DDR_ECC_ERR_BASE,
+ .int_offset = MW_DDR_ECC_INT_BASE,
+};
+
+static const struct of_device_id hb_ddr_ctrl_of_match[] = {
+ { .compatible = "calxeda,hb-ddr-ctrl", .data = &hb_settings },
+ { .compatible = "calxeda,ecx-2000-ddr-ctrl", .data = &mw_settings },
+ {},
+};
+MODULE_DEVICE_TABLE(of, hb_ddr_ctrl_of_match);
+
+static int highbank_mc_probe(struct platform_device *pdev)
+{
+ const struct of_device_id *id;
+ const struct hb_mc_settings *settings;
+ struct edac_mc_layer layers[2];
+ struct mem_ctl_info *mci;
+ struct hb_mc_drvdata *drvdata;
+ struct dimm_info *dimm;
+ struct resource *r;
+ void __iomem *base;
+ u32 control;
+ int irq;
+ int res = 0;
+
+ id = of_match_device(hb_ddr_ctrl_of_match, &pdev->dev);
+ if (!id)
+ return -ENODEV;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 1;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct hb_mc_drvdata));
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ drvdata = mci->pvt_info;
+ platform_set_drvdata(pdev, mci);
+
+ if (!devres_open_group(&pdev->dev, NULL, GFP_KERNEL)) {
+ res = -ENOMEM;
+ goto free;
+ }
+
+ r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (!r) {
+ dev_err(&pdev->dev, "Unable to get mem resource\n");
+ res = -ENODEV;
+ goto err;
+ }
+
+ if (!devm_request_mem_region(&pdev->dev, r->start,
+ resource_size(r), dev_name(&pdev->dev))) {
+ dev_err(&pdev->dev, "Error while requesting mem region\n");
+ res = -EBUSY;
+ goto err;
+ }
+
+ base = devm_ioremap(&pdev->dev, r->start, resource_size(r));
+ if (!base) {
+ dev_err(&pdev->dev, "Unable to map regs\n");
+ res = -ENOMEM;
+ goto err;
+ }
+
+ settings = id->data;
+ drvdata->mc_err_base = base + settings->err_offset;
+ drvdata->mc_int_base = base + settings->int_offset;
+
+ control = readl(drvdata->mc_err_base + HB_DDR_ECC_OPT) & 0x3;
+ if (!control || (control == 0x2)) {
+ dev_err(&pdev->dev, "No ECC present, or ECC disabled\n");
+ res = -ENODEV;
+ goto err;
+ }
+
+ mci->mtype_cap = MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = pdev->dev.driver->name;
+ mci->ctl_name = id->compatible;
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->scrub_mode = SCRUB_SW_SRC;
+
+ /* Only a single 4GB DIMM is supported */
+ dimm = *mci->dimms;
+ dimm->nr_pages = (~0UL >> PAGE_SHIFT) + 1;
+ dimm->grain = 8;
+ dimm->dtype = DEV_X8;
+ dimm->mtype = MEM_DDR3;
+ dimm->edac_mode = EDAC_SECDED;
+
+ res = edac_mc_add_mc_with_groups(mci, highbank_dev_groups);
+ if (res < 0)
+ goto err;
+
+ irq = platform_get_irq(pdev, 0);
+ res = devm_request_irq(&pdev->dev, irq, highbank_mc_err_handler,
+ 0, dev_name(&pdev->dev), mci);
+ if (res < 0) {
+ dev_err(&pdev->dev, "Unable to request irq %d\n", irq);
+ goto err2;
+ }
+
+ devres_close_group(&pdev->dev, NULL);
+ return 0;
+err2:
+ edac_mc_del_mc(&pdev->dev);
+err:
+ devres_release_group(&pdev->dev, NULL);
+free:
+ edac_mc_free(mci);
+ return res;
+}
+
+static int highbank_mc_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+ return 0;
+}
+
+static struct platform_driver highbank_mc_edac_driver = {
+ .probe = highbank_mc_probe,
+ .remove = highbank_mc_remove,
+ .driver = {
+ .name = "hb_mc_edac",
+ .of_match_table = hb_ddr_ctrl_of_match,
+ },
+};
+
+module_platform_driver(highbank_mc_edac_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Calxeda, Inc.");
+MODULE_DESCRIPTION("EDAC Driver for Calxeda Highbank");
diff --git a/drivers/edac/i10nm_base.c b/drivers/edac/i10nm_base.c
new file mode 100644
index 0000000000..2b83d6de93
--- /dev/null
+++ b/drivers/edac/i10nm_base.c
@@ -0,0 +1,1230 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Driver for Intel(R) 10nm server memory controller.
+ * Copyright (c) 2019, Intel Corporation.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <asm/cpu_device_id.h>
+#include <asm/intel-family.h>
+#include <asm/mce.h>
+#include "edac_module.h"
+#include "skx_common.h"
+
+#define I10NM_REVISION "v0.0.6"
+#define EDAC_MOD_STR "i10nm_edac"
+
+/* Debug macros */
+#define i10nm_printk(level, fmt, arg...) \
+ edac_printk(level, "i10nm", fmt, ##arg)
+
+#define I10NM_GET_SCK_BAR(d, reg) \
+ pci_read_config_dword((d)->uracu, 0xd0, &(reg))
+#define I10NM_GET_IMC_BAR(d, i, reg) \
+ pci_read_config_dword((d)->uracu, \
+ (res_cfg->type == GNR ? 0xd4 : 0xd8) + (i) * 4, &(reg))
+#define I10NM_GET_SAD(d, offset, i, reg)\
+ pci_read_config_dword((d)->sad_all, (offset) + (i) * \
+ (res_cfg->type == GNR ? 12 : 8), &(reg))
+#define I10NM_GET_HBM_IMC_BAR(d, reg) \
+ pci_read_config_dword((d)->uracu, 0xd4, &(reg))
+#define I10NM_GET_CAPID3_CFG(d, reg) \
+ pci_read_config_dword((d)->pcu_cr3, \
+ res_cfg->type == GNR ? 0x290 : 0x90, &(reg))
+#define I10NM_GET_CAPID5_CFG(d, reg) \
+ pci_read_config_dword((d)->pcu_cr3, \
+ res_cfg->type == GNR ? 0x298 : 0x98, &(reg))
+#define I10NM_GET_DIMMMTR(m, i, j) \
+ readl((m)->mbase + ((m)->hbm_mc ? 0x80c : \
+ (res_cfg->type == GNR ? 0xc0c : 0x2080c)) + \
+ (i) * (m)->chan_mmio_sz + (j) * 4)
+#define I10NM_GET_MCDDRTCFG(m, i) \
+ readl((m)->mbase + ((m)->hbm_mc ? 0x970 : 0x20970) + \
+ (i) * (m)->chan_mmio_sz)
+#define I10NM_GET_MCMTR(m, i) \
+ readl((m)->mbase + ((m)->hbm_mc ? 0xef8 : \
+ (res_cfg->type == GNR ? 0xaf8 : 0x20ef8)) + \
+ (i) * (m)->chan_mmio_sz)
+#define I10NM_GET_AMAP(m, i) \
+ readl((m)->mbase + ((m)->hbm_mc ? 0x814 : \
+ (res_cfg->type == GNR ? 0xc14 : 0x20814)) + \
+ (i) * (m)->chan_mmio_sz)
+#define I10NM_GET_REG32(m, i, offset) \
+ readl((m)->mbase + (i) * (m)->chan_mmio_sz + (offset))
+#define I10NM_GET_REG64(m, i, offset) \
+ readq((m)->mbase + (i) * (m)->chan_mmio_sz + (offset))
+#define I10NM_SET_REG32(m, i, offset, v) \
+ writel(v, (m)->mbase + (i) * (m)->chan_mmio_sz + (offset))
+
+#define I10NM_GET_SCK_MMIO_BASE(reg) (GET_BITFIELD(reg, 0, 28) << 23)
+#define I10NM_GET_IMC_MMIO_OFFSET(reg) (GET_BITFIELD(reg, 0, 10) << 12)
+#define I10NM_GET_IMC_MMIO_SIZE(reg) ((GET_BITFIELD(reg, 13, 23) - \
+ GET_BITFIELD(reg, 0, 10) + 1) << 12)
+#define I10NM_GET_HBM_IMC_MMIO_OFFSET(reg) \
+ ((GET_BITFIELD(reg, 0, 10) << 12) + 0x140000)
+
+#define I10NM_GNR_IMC_MMIO_OFFSET 0x24c000
+#define I10NM_GNR_IMC_MMIO_SIZE 0x4000
+#define I10NM_HBM_IMC_MMIO_SIZE 0x9000
+#define I10NM_DDR_IMC_CH_CNT(reg) GET_BITFIELD(reg, 21, 24)
+#define I10NM_IS_HBM_PRESENT(reg) GET_BITFIELD(reg, 27, 30)
+#define I10NM_IS_HBM_IMC(reg) GET_BITFIELD(reg, 29, 29)
+
+#define I10NM_MAX_SAD 16
+#define I10NM_SAD_ENABLE(reg) GET_BITFIELD(reg, 0, 0)
+#define I10NM_SAD_NM_CACHEABLE(reg) GET_BITFIELD(reg, 5, 5)
+
+#define RETRY_RD_ERR_LOG_UC BIT(1)
+#define RETRY_RD_ERR_LOG_NOOVER BIT(14)
+#define RETRY_RD_ERR_LOG_EN BIT(15)
+#define RETRY_RD_ERR_LOG_NOOVER_UC (BIT(14) | BIT(1))
+#define RETRY_RD_ERR_LOG_OVER_UC_V (BIT(2) | BIT(1) | BIT(0))
+
+static struct list_head *i10nm_edac_list;
+
+static struct res_config *res_cfg;
+static int retry_rd_err_log;
+static int decoding_via_mca;
+static bool mem_cfg_2lm;
+
+static u32 offsets_scrub_icx[] = {0x22c60, 0x22c54, 0x22c5c, 0x22c58, 0x22c28, 0x20ed8};
+static u32 offsets_scrub_spr[] = {0x22c60, 0x22c54, 0x22f08, 0x22c58, 0x22c28, 0x20ed8};
+static u32 offsets_scrub_spr_hbm0[] = {0x2860, 0x2854, 0x2b08, 0x2858, 0x2828, 0x0ed8};
+static u32 offsets_scrub_spr_hbm1[] = {0x2c60, 0x2c54, 0x2f08, 0x2c58, 0x2c28, 0x0fa8};
+static u32 offsets_demand_icx[] = {0x22e54, 0x22e60, 0x22e64, 0x22e58, 0x22e5c, 0x20ee0};
+static u32 offsets_demand_spr[] = {0x22e54, 0x22e60, 0x22f10, 0x22e58, 0x22e5c, 0x20ee0};
+static u32 offsets_demand2_spr[] = {0x22c70, 0x22d80, 0x22f18, 0x22d58, 0x22c64, 0x20f10};
+static u32 offsets_demand_spr_hbm0[] = {0x2a54, 0x2a60, 0x2b10, 0x2a58, 0x2a5c, 0x0ee0};
+static u32 offsets_demand_spr_hbm1[] = {0x2e54, 0x2e60, 0x2f10, 0x2e58, 0x2e5c, 0x0fb0};
+
+static void __enable_retry_rd_err_log(struct skx_imc *imc, int chan, bool enable,
+ u32 *offsets_scrub, u32 *offsets_demand,
+ u32 *offsets_demand2)
+{
+ u32 s, d, d2;
+
+ s = I10NM_GET_REG32(imc, chan, offsets_scrub[0]);
+ d = I10NM_GET_REG32(imc, chan, offsets_demand[0]);
+ if (offsets_demand2)
+ d2 = I10NM_GET_REG32(imc, chan, offsets_demand2[0]);
+
+ if (enable) {
+ /* Save default configurations */
+ imc->chan[chan].retry_rd_err_log_s = s;
+ imc->chan[chan].retry_rd_err_log_d = d;
+ if (offsets_demand2)
+ imc->chan[chan].retry_rd_err_log_d2 = d2;
+
+ s &= ~RETRY_RD_ERR_LOG_NOOVER_UC;
+ s |= RETRY_RD_ERR_LOG_EN;
+ d &= ~RETRY_RD_ERR_LOG_NOOVER_UC;
+ d |= RETRY_RD_ERR_LOG_EN;
+
+ if (offsets_demand2) {
+ d2 &= ~RETRY_RD_ERR_LOG_UC;
+ d2 |= RETRY_RD_ERR_LOG_NOOVER;
+ d2 |= RETRY_RD_ERR_LOG_EN;
+ }
+ } else {
+ /* Restore default configurations */
+ if (imc->chan[chan].retry_rd_err_log_s & RETRY_RD_ERR_LOG_UC)
+ s |= RETRY_RD_ERR_LOG_UC;
+ if (imc->chan[chan].retry_rd_err_log_s & RETRY_RD_ERR_LOG_NOOVER)
+ s |= RETRY_RD_ERR_LOG_NOOVER;
+ if (!(imc->chan[chan].retry_rd_err_log_s & RETRY_RD_ERR_LOG_EN))
+ s &= ~RETRY_RD_ERR_LOG_EN;
+ if (imc->chan[chan].retry_rd_err_log_d & RETRY_RD_ERR_LOG_UC)
+ d |= RETRY_RD_ERR_LOG_UC;
+ if (imc->chan[chan].retry_rd_err_log_d & RETRY_RD_ERR_LOG_NOOVER)
+ d |= RETRY_RD_ERR_LOG_NOOVER;
+ if (!(imc->chan[chan].retry_rd_err_log_d & RETRY_RD_ERR_LOG_EN))
+ d &= ~RETRY_RD_ERR_LOG_EN;
+
+ if (offsets_demand2) {
+ if (imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_UC)
+ d2 |= RETRY_RD_ERR_LOG_UC;
+ if (!(imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_NOOVER))
+ d2 &= ~RETRY_RD_ERR_LOG_NOOVER;
+ if (!(imc->chan[chan].retry_rd_err_log_d2 & RETRY_RD_ERR_LOG_EN))
+ d2 &= ~RETRY_RD_ERR_LOG_EN;
+ }
+ }
+
+ I10NM_SET_REG32(imc, chan, offsets_scrub[0], s);
+ I10NM_SET_REG32(imc, chan, offsets_demand[0], d);
+ if (offsets_demand2)
+ I10NM_SET_REG32(imc, chan, offsets_demand2[0], d2);
+}
+
+static void enable_retry_rd_err_log(bool enable)
+{
+ int i, j, imc_num, chan_num;
+ struct skx_imc *imc;
+ struct skx_dev *d;
+
+ edac_dbg(2, "\n");
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ imc_num = res_cfg->ddr_imc_num;
+ chan_num = res_cfg->ddr_chan_num;
+
+ for (i = 0; i < imc_num; i++) {
+ imc = &d->imc[i];
+ if (!imc->mbase)
+ continue;
+
+ for (j = 0; j < chan_num; j++)
+ __enable_retry_rd_err_log(imc, j, enable,
+ res_cfg->offsets_scrub,
+ res_cfg->offsets_demand,
+ res_cfg->offsets_demand2);
+ }
+
+ imc_num += res_cfg->hbm_imc_num;
+ chan_num = res_cfg->hbm_chan_num;
+
+ for (; i < imc_num; i++) {
+ imc = &d->imc[i];
+ if (!imc->mbase || !imc->hbm_mc)
+ continue;
+
+ for (j = 0; j < chan_num; j++) {
+ __enable_retry_rd_err_log(imc, j, enable,
+ res_cfg->offsets_scrub_hbm0,
+ res_cfg->offsets_demand_hbm0,
+ NULL);
+ __enable_retry_rd_err_log(imc, j, enable,
+ res_cfg->offsets_scrub_hbm1,
+ res_cfg->offsets_demand_hbm1,
+ NULL);
+ }
+ }
+ }
+}
+
+static void show_retry_rd_err_log(struct decoded_addr *res, char *msg,
+ int len, bool scrub_err)
+{
+ struct skx_imc *imc = &res->dev->imc[res->imc];
+ u32 log0, log1, log2, log3, log4;
+ u32 corr0, corr1, corr2, corr3;
+ u32 lxg0, lxg1, lxg3, lxg4;
+ u32 *xffsets = NULL;
+ u64 log2a, log5;
+ u64 lxg2a, lxg5;
+ u32 *offsets;
+ int n, pch;
+
+ if (!imc->mbase)
+ return;
+
+ if (imc->hbm_mc) {
+ pch = res->cs & 1;
+
+ if (pch)
+ offsets = scrub_err ? res_cfg->offsets_scrub_hbm1 :
+ res_cfg->offsets_demand_hbm1;
+ else
+ offsets = scrub_err ? res_cfg->offsets_scrub_hbm0 :
+ res_cfg->offsets_demand_hbm0;
+ } else {
+ if (scrub_err) {
+ offsets = res_cfg->offsets_scrub;
+ } else {
+ offsets = res_cfg->offsets_demand;
+ xffsets = res_cfg->offsets_demand2;
+ }
+ }
+
+ log0 = I10NM_GET_REG32(imc, res->channel, offsets[0]);
+ log1 = I10NM_GET_REG32(imc, res->channel, offsets[1]);
+ log3 = I10NM_GET_REG32(imc, res->channel, offsets[3]);
+ log4 = I10NM_GET_REG32(imc, res->channel, offsets[4]);
+ log5 = I10NM_GET_REG64(imc, res->channel, offsets[5]);
+
+ if (xffsets) {
+ lxg0 = I10NM_GET_REG32(imc, res->channel, xffsets[0]);
+ lxg1 = I10NM_GET_REG32(imc, res->channel, xffsets[1]);
+ lxg3 = I10NM_GET_REG32(imc, res->channel, xffsets[3]);
+ lxg4 = I10NM_GET_REG32(imc, res->channel, xffsets[4]);
+ lxg5 = I10NM_GET_REG64(imc, res->channel, xffsets[5]);
+ }
+
+ if (res_cfg->type == SPR) {
+ log2a = I10NM_GET_REG64(imc, res->channel, offsets[2]);
+ n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.16llx %.8x %.8x %.16llx",
+ log0, log1, log2a, log3, log4, log5);
+
+ if (len - n > 0) {
+ if (xffsets) {
+ lxg2a = I10NM_GET_REG64(imc, res->channel, xffsets[2]);
+ n += snprintf(msg + n, len - n, " %.8x %.8x %.16llx %.8x %.8x %.16llx]",
+ lxg0, lxg1, lxg2a, lxg3, lxg4, lxg5);
+ } else {
+ n += snprintf(msg + n, len - n, "]");
+ }
+ }
+ } else {
+ log2 = I10NM_GET_REG32(imc, res->channel, offsets[2]);
+ n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.8x %.8x %.8x %.16llx]",
+ log0, log1, log2, log3, log4, log5);
+ }
+
+ if (imc->hbm_mc) {
+ if (pch) {
+ corr0 = I10NM_GET_REG32(imc, res->channel, 0x2c18);
+ corr1 = I10NM_GET_REG32(imc, res->channel, 0x2c1c);
+ corr2 = I10NM_GET_REG32(imc, res->channel, 0x2c20);
+ corr3 = I10NM_GET_REG32(imc, res->channel, 0x2c24);
+ } else {
+ corr0 = I10NM_GET_REG32(imc, res->channel, 0x2818);
+ corr1 = I10NM_GET_REG32(imc, res->channel, 0x281c);
+ corr2 = I10NM_GET_REG32(imc, res->channel, 0x2820);
+ corr3 = I10NM_GET_REG32(imc, res->channel, 0x2824);
+ }
+ } else {
+ corr0 = I10NM_GET_REG32(imc, res->channel, 0x22c18);
+ corr1 = I10NM_GET_REG32(imc, res->channel, 0x22c1c);
+ corr2 = I10NM_GET_REG32(imc, res->channel, 0x22c20);
+ corr3 = I10NM_GET_REG32(imc, res->channel, 0x22c24);
+ }
+
+ if (len - n > 0)
+ snprintf(msg + n, len - n,
+ " correrrcnt[%.4x %.4x %.4x %.4x %.4x %.4x %.4x %.4x]",
+ corr0 & 0xffff, corr0 >> 16,
+ corr1 & 0xffff, corr1 >> 16,
+ corr2 & 0xffff, corr2 >> 16,
+ corr3 & 0xffff, corr3 >> 16);
+
+ /* Clear status bits */
+ if (retry_rd_err_log == 2) {
+ if (log0 & RETRY_RD_ERR_LOG_OVER_UC_V) {
+ log0 &= ~RETRY_RD_ERR_LOG_OVER_UC_V;
+ I10NM_SET_REG32(imc, res->channel, offsets[0], log0);
+ }
+
+ if (xffsets && (lxg0 & RETRY_RD_ERR_LOG_OVER_UC_V)) {
+ lxg0 &= ~RETRY_RD_ERR_LOG_OVER_UC_V;
+ I10NM_SET_REG32(imc, res->channel, xffsets[0], lxg0);
+ }
+ }
+}
+
+static struct pci_dev *pci_get_dev_wrapper(int dom, unsigned int bus,
+ unsigned int dev, unsigned int fun)
+{
+ struct pci_dev *pdev;
+
+ pdev = pci_get_domain_bus_and_slot(dom, bus, PCI_DEVFN(dev, fun));
+ if (!pdev) {
+ edac_dbg(2, "No device %02x:%02x.%x\n",
+ bus, dev, fun);
+ return NULL;
+ }
+
+ if (unlikely(pci_enable_device(pdev) < 0)) {
+ edac_dbg(2, "Failed to enable device %02x:%02x.%x\n",
+ bus, dev, fun);
+ pci_dev_put(pdev);
+ return NULL;
+ }
+
+ return pdev;
+}
+
+/**
+ * i10nm_get_imc_num() - Get the number of present DDR memory controllers.
+ *
+ * @cfg : The pointer to the structure of EDAC resource configurations.
+ *
+ * For Granite Rapids CPUs, the number of present DDR memory controllers read
+ * at runtime overwrites the value statically configured in @cfg->ddr_imc_num.
+ * For other CPUs, the number of present DDR memory controllers is statically
+ * configured in @cfg->ddr_imc_num.
+ *
+ * RETURNS : 0 on success, < 0 on failure.
+ */
+static int i10nm_get_imc_num(struct res_config *cfg)
+{
+ int n, imc_num, chan_num = 0;
+ struct skx_dev *d;
+ u32 reg;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ d->pcu_cr3 = pci_get_dev_wrapper(d->seg, d->bus[res_cfg->pcu_cr3_bdf.bus],
+ res_cfg->pcu_cr3_bdf.dev,
+ res_cfg->pcu_cr3_bdf.fun);
+ if (!d->pcu_cr3)
+ continue;
+
+ if (I10NM_GET_CAPID5_CFG(d, reg))
+ continue;
+
+ n = I10NM_DDR_IMC_CH_CNT(reg);
+
+ if (!chan_num) {
+ chan_num = n;
+ edac_dbg(2, "Get DDR CH number: %d\n", chan_num);
+ } else if (chan_num != n) {
+ i10nm_printk(KERN_NOTICE, "Get DDR CH numbers: %d, %d\n", chan_num, n);
+ }
+ }
+
+ switch (cfg->type) {
+ case GNR:
+ /*
+ * One channel per DDR memory controller for Granite Rapids CPUs.
+ */
+ imc_num = chan_num;
+
+ if (!imc_num) {
+ i10nm_printk(KERN_ERR, "Invalid DDR MC number\n");
+ return -ENODEV;
+ }
+
+ if (imc_num > I10NM_NUM_DDR_IMC) {
+ i10nm_printk(KERN_ERR, "Need to make I10NM_NUM_DDR_IMC >= %d\n", imc_num);
+ return -EINVAL;
+ }
+
+ if (cfg->ddr_imc_num != imc_num) {
+ /*
+ * Store the number of present DDR memory controllers.
+ */
+ cfg->ddr_imc_num = imc_num;
+ edac_dbg(2, "Set DDR MC number: %d", imc_num);
+ }
+
+ return 0;
+ default:
+ /*
+ * For other CPUs, the number of present DDR memory controllers
+ * is statically pre-configured in cfg->ddr_imc_num.
+ */
+ return 0;
+ }
+}
+
+static bool i10nm_check_2lm(struct res_config *cfg)
+{
+ struct skx_dev *d;
+ u32 reg;
+ int i;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ d->sad_all = pci_get_dev_wrapper(d->seg, d->bus[res_cfg->sad_all_bdf.bus],
+ res_cfg->sad_all_bdf.dev,
+ res_cfg->sad_all_bdf.fun);
+ if (!d->sad_all)
+ continue;
+
+ for (i = 0; i < I10NM_MAX_SAD; i++) {
+ I10NM_GET_SAD(d, cfg->sad_all_offset, i, reg);
+ if (I10NM_SAD_ENABLE(reg) && I10NM_SAD_NM_CACHEABLE(reg)) {
+ edac_dbg(2, "2-level memory configuration.\n");
+ return true;
+ }
+ }
+ }
+
+ return false;
+}
+
+/*
+ * Check whether the error comes from DDRT by ICX/Tremont/SPR model specific error code.
+ * Refer to SDM vol3B 17.11.3/17.13.2 Intel IMC MC error codes for IA32_MCi_STATUS.
+ */
+static bool i10nm_mscod_is_ddrt(u32 mscod)
+{
+ switch (res_cfg->type) {
+ case I10NM:
+ switch (mscod) {
+ case 0x0106: case 0x0107:
+ case 0x0800: case 0x0804:
+ case 0x0806 ... 0x0808:
+ case 0x080a ... 0x080e:
+ case 0x0810: case 0x0811:
+ case 0x0816: case 0x081e:
+ case 0x081f:
+ return true;
+ }
+
+ break;
+ case SPR:
+ switch (mscod) {
+ case 0x0800: case 0x0804:
+ case 0x0806 ... 0x0808:
+ case 0x080a ... 0x080e:
+ case 0x0810: case 0x0811:
+ case 0x0816: case 0x081e:
+ case 0x081f:
+ return true;
+ }
+
+ break;
+ default:
+ return false;
+ }
+
+ return false;
+}
+
+static bool i10nm_mc_decode_available(struct mce *mce)
+{
+#define ICX_IMCx_CHy 0x06666000
+ u8 bank;
+
+ if (!decoding_via_mca || mem_cfg_2lm)
+ return false;
+
+ if ((mce->status & (MCI_STATUS_MISCV | MCI_STATUS_ADDRV))
+ != (MCI_STATUS_MISCV | MCI_STATUS_ADDRV))
+ return false;
+
+ bank = mce->bank;
+
+ switch (res_cfg->type) {
+ case I10NM:
+ /* Check whether the bank is one of {13,14,17,18,21,22,25,26} */
+ if (!(ICX_IMCx_CHy & (1 << bank)))
+ return false;
+ break;
+ case SPR:
+ if (bank < 13 || bank > 20)
+ return false;
+ break;
+ default:
+ return false;
+ }
+
+ /* DDRT errors can't be decoded from MCA bank registers */
+ if (MCI_MISC_ECC_MODE(mce->misc) == MCI_MISC_ECC_DDRT)
+ return false;
+
+ if (i10nm_mscod_is_ddrt(MCI_STATUS_MSCOD(mce->status)))
+ return false;
+
+ return true;
+}
+
+static bool i10nm_mc_decode(struct decoded_addr *res)
+{
+ struct mce *m = res->mce;
+ struct skx_dev *d;
+ u8 bank;
+
+ if (!i10nm_mc_decode_available(m))
+ return false;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ if (d->imc[0].src_id == m->socketid) {
+ res->socket = m->socketid;
+ res->dev = d;
+ break;
+ }
+ }
+
+ switch (res_cfg->type) {
+ case I10NM:
+ bank = m->bank - 13;
+ res->imc = bank / 4;
+ res->channel = bank % 2;
+ res->column = GET_BITFIELD(m->misc, 9, 18) << 2;
+ res->row = GET_BITFIELD(m->misc, 19, 39);
+ res->bank_group = GET_BITFIELD(m->misc, 40, 41);
+ res->bank_address = GET_BITFIELD(m->misc, 42, 43);
+ res->bank_group |= GET_BITFIELD(m->misc, 44, 44) << 2;
+ res->rank = GET_BITFIELD(m->misc, 56, 58);
+ res->dimm = res->rank >> 2;
+ res->rank = res->rank % 4;
+ break;
+ case SPR:
+ bank = m->bank - 13;
+ res->imc = bank / 2;
+ res->channel = bank % 2;
+ res->column = GET_BITFIELD(m->misc, 9, 18) << 2;
+ res->row = GET_BITFIELD(m->misc, 19, 36);
+ res->bank_group = GET_BITFIELD(m->misc, 37, 38);
+ res->bank_address = GET_BITFIELD(m->misc, 39, 40);
+ res->bank_group |= GET_BITFIELD(m->misc, 41, 41) << 2;
+ res->rank = GET_BITFIELD(m->misc, 57, 57);
+ res->dimm = GET_BITFIELD(m->misc, 58, 58);
+ break;
+ default:
+ return false;
+ }
+
+ if (!res->dev) {
+ skx_printk(KERN_ERR, "No device for src_id %d imc %d\n",
+ m->socketid, res->imc);
+ return false;
+ }
+
+ return true;
+}
+
+/**
+ * get_gnr_mdev() - Get the PCI device of the @logical_idx-th DDR memory controller.
+ *
+ * @d : The pointer to the structure of CPU socket EDAC device.
+ * @logical_idx : The logical index of the present memory controller (0 ~ max present MC# - 1).
+ * @physical_idx : To store the corresponding physical index of @logical_idx.
+ *
+ * RETURNS : The PCI device of the @logical_idx-th DDR memory controller, NULL on failure.
+ */
+static struct pci_dev *get_gnr_mdev(struct skx_dev *d, int logical_idx, int *physical_idx)
+{
+#define GNR_MAX_IMC_PCI_CNT 28
+
+ struct pci_dev *mdev;
+ int i, logical = 0;
+
+ /*
+ * Detect present memory controllers from { PCI device: 8-5, function 7-1 }
+ */
+ for (i = 0; i < GNR_MAX_IMC_PCI_CNT; i++) {
+ mdev = pci_get_dev_wrapper(d->seg,
+ d->bus[res_cfg->ddr_mdev_bdf.bus],
+ res_cfg->ddr_mdev_bdf.dev + i / 7,
+ res_cfg->ddr_mdev_bdf.fun + i % 7);
+
+ if (mdev) {
+ if (logical == logical_idx) {
+ *physical_idx = i;
+ return mdev;
+ }
+
+ pci_dev_put(mdev);
+ logical++;
+ }
+ }
+
+ return NULL;
+}
+
+/**
+ * get_ddr_munit() - Get the resource of the i-th DDR memory controller.
+ *
+ * @d : The pointer to the structure of CPU socket EDAC device.
+ * @i : The index of the CPU socket relative DDR memory controller.
+ * @offset : To store the MMIO offset of the i-th DDR memory controller.
+ * @size : To store the MMIO size of the i-th DDR memory controller.
+ *
+ * RETURNS : The PCI device of the i-th DDR memory controller, NULL on failure.
+ */
+static struct pci_dev *get_ddr_munit(struct skx_dev *d, int i, u32 *offset, unsigned long *size)
+{
+ struct pci_dev *mdev;
+ int physical_idx;
+ u32 reg;
+
+ switch (res_cfg->type) {
+ case GNR:
+ if (I10NM_GET_IMC_BAR(d, 0, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to get mc0 bar\n");
+ return NULL;
+ }
+
+ mdev = get_gnr_mdev(d, i, &physical_idx);
+ if (!mdev)
+ return NULL;
+
+ *offset = I10NM_GET_IMC_MMIO_OFFSET(reg) +
+ I10NM_GNR_IMC_MMIO_OFFSET +
+ physical_idx * I10NM_GNR_IMC_MMIO_SIZE;
+ *size = I10NM_GNR_IMC_MMIO_SIZE;
+
+ break;
+ default:
+ if (I10NM_GET_IMC_BAR(d, i, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to get mc%d bar\n", i);
+ return NULL;
+ }
+
+ mdev = pci_get_dev_wrapper(d->seg,
+ d->bus[res_cfg->ddr_mdev_bdf.bus],
+ res_cfg->ddr_mdev_bdf.dev + i,
+ res_cfg->ddr_mdev_bdf.fun);
+ if (!mdev)
+ return NULL;
+
+ *offset = I10NM_GET_IMC_MMIO_OFFSET(reg);
+ *size = I10NM_GET_IMC_MMIO_SIZE(reg);
+ }
+
+ return mdev;
+}
+
+/**
+ * i10nm_imc_absent() - Check whether the memory controller @imc is absent
+ *
+ * @imc : The pointer to the structure of memory controller EDAC device.
+ *
+ * RETURNS : true if the memory controller EDAC device is absent, false otherwise.
+ */
+static bool i10nm_imc_absent(struct skx_imc *imc)
+{
+ u32 mcmtr;
+ int i;
+
+ switch (res_cfg->type) {
+ case SPR:
+ for (i = 0; i < res_cfg->ddr_chan_num; i++) {
+ mcmtr = I10NM_GET_MCMTR(imc, i);
+ edac_dbg(1, "ch%d mcmtr reg %x\n", i, mcmtr);
+ if (mcmtr != ~0)
+ return false;
+ }
+
+ /*
+ * Some workstations' absent memory controllers still
+ * appear as PCIe devices, misleading the EDAC driver.
+ * By observing that the MMIO registers of these absent
+ * memory controllers consistently hold the value of ~0.
+ *
+ * We identify a memory controller as absent by checking
+ * if its MMIO register "mcmtr" == ~0 in all its channels.
+ */
+ return true;
+ default:
+ return false;
+ }
+}
+
+static int i10nm_get_ddr_munits(void)
+{
+ struct pci_dev *mdev;
+ void __iomem *mbase;
+ unsigned long size;
+ struct skx_dev *d;
+ int i, lmc, j = 0;
+ u32 reg, off;
+ u64 base;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ d->util_all = pci_get_dev_wrapper(d->seg, d->bus[res_cfg->util_all_bdf.bus],
+ res_cfg->util_all_bdf.dev,
+ res_cfg->util_all_bdf.fun);
+ if (!d->util_all)
+ return -ENODEV;
+
+ d->uracu = pci_get_dev_wrapper(d->seg, d->bus[res_cfg->uracu_bdf.bus],
+ res_cfg->uracu_bdf.dev,
+ res_cfg->uracu_bdf.fun);
+ if (!d->uracu)
+ return -ENODEV;
+
+ if (I10NM_GET_SCK_BAR(d, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to socket bar\n");
+ return -ENODEV;
+ }
+
+ base = I10NM_GET_SCK_MMIO_BASE(reg);
+ edac_dbg(2, "socket%d mmio base 0x%llx (reg 0x%x)\n",
+ j++, base, reg);
+
+ for (lmc = 0, i = 0; i < res_cfg->ddr_imc_num; i++) {
+ mdev = get_ddr_munit(d, i, &off, &size);
+
+ if (i == 0 && !mdev) {
+ i10nm_printk(KERN_ERR, "No IMC found\n");
+ return -ENODEV;
+ }
+ if (!mdev)
+ continue;
+
+ edac_dbg(2, "mc%d mmio base 0x%llx size 0x%lx (reg 0x%x)\n",
+ i, base + off, size, reg);
+
+ mbase = ioremap(base + off, size);
+ if (!mbase) {
+ i10nm_printk(KERN_ERR, "Failed to ioremap 0x%llx\n",
+ base + off);
+ return -ENODEV;
+ }
+
+ d->imc[lmc].mbase = mbase;
+ if (i10nm_imc_absent(&d->imc[lmc])) {
+ pci_dev_put(mdev);
+ iounmap(mbase);
+ d->imc[lmc].mbase = NULL;
+ edac_dbg(2, "Skip absent mc%d\n", i);
+ continue;
+ } else {
+ d->imc[lmc].mdev = mdev;
+ lmc++;
+ }
+ }
+ }
+
+ return 0;
+}
+
+static bool i10nm_check_hbm_imc(struct skx_dev *d)
+{
+ u32 reg;
+
+ if (I10NM_GET_CAPID3_CFG(d, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to get capid3_cfg\n");
+ return false;
+ }
+
+ return I10NM_IS_HBM_PRESENT(reg) != 0;
+}
+
+static int i10nm_get_hbm_munits(void)
+{
+ struct pci_dev *mdev;
+ void __iomem *mbase;
+ u32 reg, off, mcmtr;
+ struct skx_dev *d;
+ int i, lmc;
+ u64 base;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ if (!d->pcu_cr3)
+ return -ENODEV;
+
+ if (!i10nm_check_hbm_imc(d)) {
+ i10nm_printk(KERN_DEBUG, "No hbm memory\n");
+ return -ENODEV;
+ }
+
+ if (I10NM_GET_SCK_BAR(d, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to get socket bar\n");
+ return -ENODEV;
+ }
+ base = I10NM_GET_SCK_MMIO_BASE(reg);
+
+ if (I10NM_GET_HBM_IMC_BAR(d, reg)) {
+ i10nm_printk(KERN_ERR, "Failed to get hbm mc bar\n");
+ return -ENODEV;
+ }
+ base += I10NM_GET_HBM_IMC_MMIO_OFFSET(reg);
+
+ lmc = res_cfg->ddr_imc_num;
+
+ for (i = 0; i < res_cfg->hbm_imc_num; i++) {
+ mdev = pci_get_dev_wrapper(d->seg, d->bus[res_cfg->hbm_mdev_bdf.bus],
+ res_cfg->hbm_mdev_bdf.dev + i / 4,
+ res_cfg->hbm_mdev_bdf.fun + i % 4);
+
+ if (i == 0 && !mdev) {
+ i10nm_printk(KERN_ERR, "No hbm mc found\n");
+ return -ENODEV;
+ }
+ if (!mdev)
+ continue;
+
+ d->imc[lmc].mdev = mdev;
+ off = i * I10NM_HBM_IMC_MMIO_SIZE;
+
+ edac_dbg(2, "hbm mc%d mmio base 0x%llx size 0x%x\n",
+ lmc, base + off, I10NM_HBM_IMC_MMIO_SIZE);
+
+ mbase = ioremap(base + off, I10NM_HBM_IMC_MMIO_SIZE);
+ if (!mbase) {
+ pci_dev_put(d->imc[lmc].mdev);
+ d->imc[lmc].mdev = NULL;
+
+ i10nm_printk(KERN_ERR, "Failed to ioremap for hbm mc 0x%llx\n",
+ base + off);
+ return -ENOMEM;
+ }
+
+ d->imc[lmc].mbase = mbase;
+ d->imc[lmc].hbm_mc = true;
+
+ mcmtr = I10NM_GET_MCMTR(&d->imc[lmc], 0);
+ if (!I10NM_IS_HBM_IMC(mcmtr)) {
+ iounmap(d->imc[lmc].mbase);
+ d->imc[lmc].mbase = NULL;
+ d->imc[lmc].hbm_mc = false;
+ pci_dev_put(d->imc[lmc].mdev);
+ d->imc[lmc].mdev = NULL;
+
+ i10nm_printk(KERN_ERR, "This isn't an hbm mc!\n");
+ return -ENODEV;
+ }
+
+ lmc++;
+ }
+ }
+
+ return 0;
+}
+
+static struct res_config i10nm_cfg0 = {
+ .type = I10NM,
+ .decs_did = 0x3452,
+ .busno_cfg_offset = 0xcc,
+ .ddr_imc_num = 4,
+ .ddr_chan_num = 2,
+ .ddr_dimm_num = 2,
+ .ddr_chan_mmio_sz = 0x4000,
+ .sad_all_bdf = {1, 29, 0},
+ .pcu_cr3_bdf = {1, 30, 3},
+ .util_all_bdf = {1, 29, 1},
+ .uracu_bdf = {0, 0, 1},
+ .ddr_mdev_bdf = {0, 12, 0},
+ .hbm_mdev_bdf = {0, 12, 1},
+ .sad_all_offset = 0x108,
+ .offsets_scrub = offsets_scrub_icx,
+ .offsets_demand = offsets_demand_icx,
+};
+
+static struct res_config i10nm_cfg1 = {
+ .type = I10NM,
+ .decs_did = 0x3452,
+ .busno_cfg_offset = 0xd0,
+ .ddr_imc_num = 4,
+ .ddr_chan_num = 2,
+ .ddr_dimm_num = 2,
+ .ddr_chan_mmio_sz = 0x4000,
+ .sad_all_bdf = {1, 29, 0},
+ .pcu_cr3_bdf = {1, 30, 3},
+ .util_all_bdf = {1, 29, 1},
+ .uracu_bdf = {0, 0, 1},
+ .ddr_mdev_bdf = {0, 12, 0},
+ .hbm_mdev_bdf = {0, 12, 1},
+ .sad_all_offset = 0x108,
+ .offsets_scrub = offsets_scrub_icx,
+ .offsets_demand = offsets_demand_icx,
+};
+
+static struct res_config spr_cfg = {
+ .type = SPR,
+ .decs_did = 0x3252,
+ .busno_cfg_offset = 0xd0,
+ .ddr_imc_num = 4,
+ .ddr_chan_num = 2,
+ .ddr_dimm_num = 2,
+ .hbm_imc_num = 16,
+ .hbm_chan_num = 2,
+ .hbm_dimm_num = 1,
+ .ddr_chan_mmio_sz = 0x8000,
+ .hbm_chan_mmio_sz = 0x4000,
+ .support_ddr5 = true,
+ .sad_all_bdf = {1, 10, 0},
+ .pcu_cr3_bdf = {1, 30, 3},
+ .util_all_bdf = {1, 29, 1},
+ .uracu_bdf = {0, 0, 1},
+ .ddr_mdev_bdf = {0, 12, 0},
+ .hbm_mdev_bdf = {0, 12, 1},
+ .sad_all_offset = 0x300,
+ .offsets_scrub = offsets_scrub_spr,
+ .offsets_scrub_hbm0 = offsets_scrub_spr_hbm0,
+ .offsets_scrub_hbm1 = offsets_scrub_spr_hbm1,
+ .offsets_demand = offsets_demand_spr,
+ .offsets_demand2 = offsets_demand2_spr,
+ .offsets_demand_hbm0 = offsets_demand_spr_hbm0,
+ .offsets_demand_hbm1 = offsets_demand_spr_hbm1,
+};
+
+static struct res_config gnr_cfg = {
+ .type = GNR,
+ .decs_did = 0x3252,
+ .busno_cfg_offset = 0xd0,
+ .ddr_imc_num = 12,
+ .ddr_chan_num = 1,
+ .ddr_dimm_num = 2,
+ .ddr_chan_mmio_sz = 0x4000,
+ .support_ddr5 = true,
+ .sad_all_bdf = {0, 13, 0},
+ .pcu_cr3_bdf = {0, 5, 0},
+ .util_all_bdf = {0, 13, 1},
+ .uracu_bdf = {0, 0, 1},
+ .ddr_mdev_bdf = {0, 5, 1},
+ .sad_all_offset = 0x300,
+};
+
+static const struct x86_cpu_id i10nm_cpuids[] = {
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ATOM_TREMONT_D, X86_STEPPINGS(0x0, 0x3), &i10nm_cfg0),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ATOM_TREMONT_D, X86_STEPPINGS(0x4, 0xf), &i10nm_cfg1),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ICELAKE_X, X86_STEPPINGS(0x0, 0x3), &i10nm_cfg0),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ICELAKE_X, X86_STEPPINGS(0x4, 0xf), &i10nm_cfg1),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ICELAKE_D, X86_STEPPINGS(0x0, 0xf), &i10nm_cfg1),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(SAPPHIRERAPIDS_X, X86_STEPPINGS(0x0, 0xf), &spr_cfg),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(EMERALDRAPIDS_X, X86_STEPPINGS(0x0, 0xf), &spr_cfg),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(GRANITERAPIDS_X, X86_STEPPINGS(0x0, 0xf), &gnr_cfg),
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(ATOM_CRESTMONT_X, X86_STEPPINGS(0x0, 0xf), &gnr_cfg),
+ {}
+};
+MODULE_DEVICE_TABLE(x86cpu, i10nm_cpuids);
+
+static bool i10nm_check_ecc(struct skx_imc *imc, int chan)
+{
+ u32 mcmtr;
+
+ mcmtr = I10NM_GET_MCMTR(imc, chan);
+ edac_dbg(1, "ch%d mcmtr reg %x\n", chan, mcmtr);
+
+ return !!GET_BITFIELD(mcmtr, 2, 2);
+}
+
+static int i10nm_get_dimm_config(struct mem_ctl_info *mci,
+ struct res_config *cfg)
+{
+ struct skx_pvt *pvt = mci->pvt_info;
+ struct skx_imc *imc = pvt->imc;
+ u32 mtr, amap, mcddrtcfg = 0;
+ struct dimm_info *dimm;
+ int i, j, ndimms;
+
+ for (i = 0; i < imc->num_channels; i++) {
+ if (!imc->mbase)
+ continue;
+
+ ndimms = 0;
+ amap = I10NM_GET_AMAP(imc, i);
+
+ if (res_cfg->type != GNR)
+ mcddrtcfg = I10NM_GET_MCDDRTCFG(imc, i);
+
+ for (j = 0; j < imc->num_dimms; j++) {
+ dimm = edac_get_dimm(mci, i, j, 0);
+ mtr = I10NM_GET_DIMMMTR(imc, i, j);
+ edac_dbg(1, "dimmmtr 0x%x mcddrtcfg 0x%x (mc%d ch%d dimm%d)\n",
+ mtr, mcddrtcfg, imc->mc, i, j);
+
+ if (IS_DIMM_PRESENT(mtr))
+ ndimms += skx_get_dimm_info(mtr, 0, amap, dimm,
+ imc, i, j, cfg);
+ else if (IS_NVDIMM_PRESENT(mcddrtcfg, j))
+ ndimms += skx_get_nvdimm_info(dimm, imc, i, j,
+ EDAC_MOD_STR);
+ }
+ if (ndimms && !i10nm_check_ecc(imc, i)) {
+ i10nm_printk(KERN_ERR, "ECC is disabled on imc %d channel %d\n",
+ imc->mc, i);
+ return -ENODEV;
+ }
+ }
+
+ return 0;
+}
+
+static struct notifier_block i10nm_mce_dec = {
+ .notifier_call = skx_mce_check_error,
+ .priority = MCE_PRIO_EDAC,
+};
+
+#ifdef CONFIG_EDAC_DEBUG
+/*
+ * Debug feature.
+ * Exercise the address decode logic by writing an address to
+ * /sys/kernel/debug/edac/i10nm_test/addr.
+ */
+static struct dentry *i10nm_test;
+
+static int debugfs_u64_set(void *data, u64 val)
+{
+ struct mce m;
+
+ pr_warn_once("Fake error to 0x%llx injected via debugfs\n", val);
+
+ memset(&m, 0, sizeof(m));
+ /* ADDRV + MemRd + Unknown channel */
+ m.status = MCI_STATUS_ADDRV + 0x90;
+ /* One corrected error */
+ m.status |= BIT_ULL(MCI_STATUS_CEC_SHIFT);
+ m.addr = val;
+ skx_mce_check_error(NULL, 0, &m);
+
+ return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n");
+
+static void setup_i10nm_debug(void)
+{
+ i10nm_test = edac_debugfs_create_dir("i10nm_test");
+ if (!i10nm_test)
+ return;
+
+ if (!edac_debugfs_create_file("addr", 0200, i10nm_test,
+ NULL, &fops_u64_wo)) {
+ debugfs_remove(i10nm_test);
+ i10nm_test = NULL;
+ }
+}
+
+static void teardown_i10nm_debug(void)
+{
+ debugfs_remove_recursive(i10nm_test);
+}
+#else
+static inline void setup_i10nm_debug(void) {}
+static inline void teardown_i10nm_debug(void) {}
+#endif /*CONFIG_EDAC_DEBUG*/
+
+static int __init i10nm_init(void)
+{
+ u8 mc = 0, src_id = 0, node_id = 0;
+ const struct x86_cpu_id *id;
+ struct res_config *cfg;
+ const char *owner;
+ struct skx_dev *d;
+ int rc, i, off[3] = {0xd0, 0xc8, 0xcc};
+ u64 tolm, tohm;
+ int imc_num;
+
+ edac_dbg(2, "\n");
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
+ return -ENODEV;
+
+ id = x86_match_cpu(i10nm_cpuids);
+ if (!id)
+ return -ENODEV;
+
+ cfg = (struct res_config *)id->driver_data;
+ res_cfg = cfg;
+
+ rc = skx_get_hi_lo(0x09a2, off, &tolm, &tohm);
+ if (rc)
+ return rc;
+
+ rc = skx_get_all_bus_mappings(cfg, &i10nm_edac_list);
+ if (rc < 0)
+ goto fail;
+ if (rc == 0) {
+ i10nm_printk(KERN_ERR, "No memory controllers found\n");
+ return -ENODEV;
+ }
+
+ rc = i10nm_get_imc_num(cfg);
+ if (rc < 0)
+ goto fail;
+
+ mem_cfg_2lm = i10nm_check_2lm(cfg);
+ skx_set_mem_cfg(mem_cfg_2lm);
+
+ rc = i10nm_get_ddr_munits();
+
+ if (i10nm_get_hbm_munits() && rc)
+ goto fail;
+
+ imc_num = res_cfg->ddr_imc_num + res_cfg->hbm_imc_num;
+
+ list_for_each_entry(d, i10nm_edac_list, list) {
+ rc = skx_get_src_id(d, 0xf8, &src_id);
+ if (rc < 0)
+ goto fail;
+
+ rc = skx_get_node_id(d, &node_id);
+ if (rc < 0)
+ goto fail;
+
+ edac_dbg(2, "src_id = %d node_id = %d\n", src_id, node_id);
+ for (i = 0; i < imc_num; i++) {
+ if (!d->imc[i].mdev)
+ continue;
+
+ d->imc[i].mc = mc++;
+ d->imc[i].lmc = i;
+ d->imc[i].src_id = src_id;
+ d->imc[i].node_id = node_id;
+ if (d->imc[i].hbm_mc) {
+ d->imc[i].chan_mmio_sz = cfg->hbm_chan_mmio_sz;
+ d->imc[i].num_channels = cfg->hbm_chan_num;
+ d->imc[i].num_dimms = cfg->hbm_dimm_num;
+ } else {
+ d->imc[i].chan_mmio_sz = cfg->ddr_chan_mmio_sz;
+ d->imc[i].num_channels = cfg->ddr_chan_num;
+ d->imc[i].num_dimms = cfg->ddr_dimm_num;
+ }
+
+ rc = skx_register_mci(&d->imc[i], d->imc[i].mdev,
+ "Intel_10nm Socket", EDAC_MOD_STR,
+ i10nm_get_dimm_config, cfg);
+ if (rc < 0)
+ goto fail;
+ }
+ }
+
+ rc = skx_adxl_get();
+ if (rc)
+ goto fail;
+
+ opstate_init();
+ mce_register_decode_chain(&i10nm_mce_dec);
+ setup_i10nm_debug();
+
+ if (retry_rd_err_log && res_cfg->offsets_scrub && res_cfg->offsets_demand) {
+ skx_set_decode(i10nm_mc_decode, show_retry_rd_err_log);
+ if (retry_rd_err_log == 2)
+ enable_retry_rd_err_log(true);
+ } else {
+ skx_set_decode(i10nm_mc_decode, NULL);
+ }
+
+ i10nm_printk(KERN_INFO, "%s\n", I10NM_REVISION);
+
+ return 0;
+fail:
+ skx_remove();
+ return rc;
+}
+
+static void __exit i10nm_exit(void)
+{
+ edac_dbg(2, "\n");
+
+ if (retry_rd_err_log && res_cfg->offsets_scrub && res_cfg->offsets_demand) {
+ skx_set_decode(NULL, NULL);
+ if (retry_rd_err_log == 2)
+ enable_retry_rd_err_log(false);
+ }
+
+ teardown_i10nm_debug();
+ mce_unregister_decode_chain(&i10nm_mce_dec);
+ skx_adxl_put();
+ skx_remove();
+}
+
+module_init(i10nm_init);
+module_exit(i10nm_exit);
+
+static int set_decoding_via_mca(const char *buf, const struct kernel_param *kp)
+{
+ unsigned long val;
+ int ret;
+
+ ret = kstrtoul(buf, 0, &val);
+
+ if (ret || val > 1)
+ return -EINVAL;
+
+ if (val && mem_cfg_2lm) {
+ i10nm_printk(KERN_NOTICE, "Decoding errors via MCA banks for 2LM isn't supported yet\n");
+ return -EIO;
+ }
+
+ ret = param_set_int(buf, kp);
+
+ return ret;
+}
+
+static const struct kernel_param_ops decoding_via_mca_param_ops = {
+ .set = set_decoding_via_mca,
+ .get = param_get_int,
+};
+
+module_param_cb(decoding_via_mca, &decoding_via_mca_param_ops, &decoding_via_mca, 0644);
+MODULE_PARM_DESC(decoding_via_mca, "decoding_via_mca: 0=off(default), 1=enable");
+
+module_param(retry_rd_err_log, int, 0444);
+MODULE_PARM_DESC(retry_rd_err_log, "retry_rd_err_log: 0=off(default), 1=bios(Linux doesn't reset any control bits, but just reports values.), 2=linux(Linux tries to take control and resets mode bits, clear valid/UC bits after reading.)");
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("MC Driver for Intel 10nm server processors");
diff --git a/drivers/edac/i3000_edac.c b/drivers/edac/i3000_edac.c
new file mode 100644
index 0000000000..9065bc4386
--- /dev/null
+++ b/drivers/edac/i3000_edac.c
@@ -0,0 +1,565 @@
+/*
+ * Intel 3000/3010 Memory Controller kernel module
+ * Copyright (C) 2007 Akamai Technologies, Inc.
+ * Shamelessly copied from:
+ * Intel D82875P Memory Controller kernel module
+ * (C) 2003 Linux Networx (http://lnxi.com)
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "i3000_edac"
+
+#define I3000_RANKS 8
+#define I3000_RANKS_PER_CHANNEL 4
+#define I3000_CHANNELS 2
+
+/* Intel 3000 register addresses - device 0 function 0 - DRAM Controller */
+
+#define I3000_MCHBAR 0x44 /* MCH Memory Mapped Register BAR */
+#define I3000_MCHBAR_MASK 0xffffc000
+#define I3000_MMR_WINDOW_SIZE 16384
+
+#define I3000_EDEAP 0x70 /* Extended DRAM Error Address Pointer (8b)
+ *
+ * 7:1 reserved
+ * 0 bit 32 of address
+ */
+#define I3000_DEAP 0x58 /* DRAM Error Address Pointer (32b)
+ *
+ * 31:7 address
+ * 6:1 reserved
+ * 0 Error channel 0/1
+ */
+#define I3000_DEAP_GRAIN (1 << 7)
+
+/*
+ * Helper functions to decode the DEAP/EDEAP hardware registers.
+ *
+ * The type promotion here is deliberate; we're deriving an
+ * unsigned long pfn and offset from hardware regs which are u8/u32.
+ */
+
+static inline unsigned long deap_pfn(u8 edeap, u32 deap)
+{
+ deap >>= PAGE_SHIFT;
+ deap |= (edeap & 1) << (32 - PAGE_SHIFT);
+ return deap;
+}
+
+static inline unsigned long deap_offset(u32 deap)
+{
+ return deap & ~(I3000_DEAP_GRAIN - 1) & ~PAGE_MASK;
+}
+
+static inline int deap_channel(u32 deap)
+{
+ return deap & 1;
+}
+
+#define I3000_DERRSYN 0x5c /* DRAM Error Syndrome (8b)
+ *
+ * 7:0 DRAM ECC Syndrome
+ */
+
+#define I3000_ERRSTS 0xc8 /* Error Status Register (16b)
+ *
+ * 15:12 reserved
+ * 11 MCH Thermal Sensor Event
+ * for SMI/SCI/SERR
+ * 10 reserved
+ * 9 LOCK to non-DRAM Memory Flag (LCKF)
+ * 8 Received Refresh Timeout Flag (RRTOF)
+ * 7:2 reserved
+ * 1 Multi-bit DRAM ECC Error Flag (DMERR)
+ * 0 Single-bit DRAM ECC Error Flag (DSERR)
+ */
+#define I3000_ERRSTS_BITS 0x0b03 /* bits which indicate errors */
+#define I3000_ERRSTS_UE 0x0002
+#define I3000_ERRSTS_CE 0x0001
+
+#define I3000_ERRCMD 0xca /* Error Command (16b)
+ *
+ * 15:12 reserved
+ * 11 SERR on MCH Thermal Sensor Event
+ * (TSESERR)
+ * 10 reserved
+ * 9 SERR on LOCK to non-DRAM Memory
+ * (LCKERR)
+ * 8 SERR on DRAM Refresh Timeout
+ * (DRTOERR)
+ * 7:2 reserved
+ * 1 SERR Multi-Bit DRAM ECC Error
+ * (DMERR)
+ * 0 SERR on Single-Bit ECC Error
+ * (DSERR)
+ */
+
+/* Intel MMIO register space - device 0 function 0 - MMR space */
+
+#define I3000_DRB_SHIFT 25 /* 32MiB grain */
+
+#define I3000_C0DRB 0x100 /* Channel 0 DRAM Rank Boundary (8b x 4)
+ *
+ * 7:0 Channel 0 DRAM Rank Boundary Address
+ */
+#define I3000_C1DRB 0x180 /* Channel 1 DRAM Rank Boundary (8b x 4)
+ *
+ * 7:0 Channel 1 DRAM Rank Boundary Address
+ */
+
+#define I3000_C0DRA 0x108 /* Channel 0 DRAM Rank Attribute (8b x 2)
+ *
+ * 7 reserved
+ * 6:4 DRAM odd Rank Attribute
+ * 3 reserved
+ * 2:0 DRAM even Rank Attribute
+ *
+ * Each attribute defines the page
+ * size of the corresponding rank:
+ * 000: unpopulated
+ * 001: reserved
+ * 010: 4 KB
+ * 011: 8 KB
+ * 100: 16 KB
+ * Others: reserved
+ */
+#define I3000_C1DRA 0x188 /* Channel 1 DRAM Rank Attribute (8b x 2) */
+
+static inline unsigned char odd_rank_attrib(unsigned char dra)
+{
+ return (dra & 0x70) >> 4;
+}
+
+static inline unsigned char even_rank_attrib(unsigned char dra)
+{
+ return dra & 0x07;
+}
+
+#define I3000_C0DRC0 0x120 /* DRAM Controller Mode 0 (32b)
+ *
+ * 31:30 reserved
+ * 29 Initialization Complete (IC)
+ * 28:11 reserved
+ * 10:8 Refresh Mode Select (RMS)
+ * 7 reserved
+ * 6:4 Mode Select (SMS)
+ * 3:2 reserved
+ * 1:0 DRAM Type (DT)
+ */
+
+#define I3000_C0DRC1 0x124 /* DRAM Controller Mode 1 (32b)
+ *
+ * 31 Enhanced Addressing Enable (ENHADE)
+ * 30:0 reserved
+ */
+
+enum i3000p_chips {
+ I3000 = 0,
+};
+
+struct i3000_dev_info {
+ const char *ctl_name;
+};
+
+struct i3000_error_info {
+ u16 errsts;
+ u8 derrsyn;
+ u8 edeap;
+ u32 deap;
+ u16 errsts2;
+};
+
+static const struct i3000_dev_info i3000_devs[] = {
+ [I3000] = {
+ .ctl_name = "i3000"},
+};
+
+static struct pci_dev *mci_pdev;
+static int i3000_registered = 1;
+static struct edac_pci_ctl_info *i3000_pci;
+
+static void i3000_get_error_info(struct mem_ctl_info *mci,
+ struct i3000_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, I3000_ERRSTS, &info->errsts);
+ if (!(info->errsts & I3000_ERRSTS_BITS))
+ return;
+ pci_read_config_byte(pdev, I3000_EDEAP, &info->edeap);
+ pci_read_config_dword(pdev, I3000_DEAP, &info->deap);
+ pci_read_config_byte(pdev, I3000_DERRSYN, &info->derrsyn);
+ pci_read_config_word(pdev, I3000_ERRSTS, &info->errsts2);
+
+ /*
+ * If the error is the same for both reads then the first set
+ * of reads is valid. If there is a change then there is a CE
+ * with no info and the second set of reads is valid and
+ * should be UE info.
+ */
+ if ((info->errsts ^ info->errsts2) & I3000_ERRSTS_BITS) {
+ pci_read_config_byte(pdev, I3000_EDEAP, &info->edeap);
+ pci_read_config_dword(pdev, I3000_DEAP, &info->deap);
+ pci_read_config_byte(pdev, I3000_DERRSYN, &info->derrsyn);
+ }
+
+ /*
+ * Clear any error bits.
+ * (Yes, we really clear bits by writing 1 to them.)
+ */
+ pci_write_bits16(pdev, I3000_ERRSTS, I3000_ERRSTS_BITS,
+ I3000_ERRSTS_BITS);
+}
+
+static int i3000_process_error_info(struct mem_ctl_info *mci,
+ struct i3000_error_info *info,
+ int handle_errors)
+{
+ int row, multi_chan, channel;
+ unsigned long pfn, offset;
+
+ multi_chan = mci->csrows[0]->nr_channels - 1;
+
+ if (!(info->errsts & I3000_ERRSTS_BITS))
+ return 0;
+
+ if (!handle_errors)
+ return 1;
+
+ if ((info->errsts ^ info->errsts2) & I3000_ERRSTS_BITS) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1,
+ "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ pfn = deap_pfn(info->edeap, info->deap);
+ offset = deap_offset(info->deap);
+ channel = deap_channel(info->deap);
+
+ row = edac_mc_find_csrow_by_page(mci, pfn);
+
+ if (info->errsts & I3000_ERRSTS_UE)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ pfn, offset, 0,
+ row, -1, -1,
+ "i3000 UE", "");
+ else
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ pfn, offset, info->derrsyn,
+ row, multi_chan ? channel : 0, -1,
+ "i3000 CE", "");
+
+ return 1;
+}
+
+static void i3000_check(struct mem_ctl_info *mci)
+{
+ struct i3000_error_info info;
+
+ i3000_get_error_info(mci, &info);
+ i3000_process_error_info(mci, &info, 1);
+}
+
+static int i3000_is_interleaved(const unsigned char *c0dra,
+ const unsigned char *c1dra,
+ const unsigned char *c0drb,
+ const unsigned char *c1drb)
+{
+ int i;
+
+ /*
+ * If the channels aren't populated identically then
+ * we're not interleaved.
+ */
+ for (i = 0; i < I3000_RANKS_PER_CHANNEL / 2; i++)
+ if (odd_rank_attrib(c0dra[i]) != odd_rank_attrib(c1dra[i]) ||
+ even_rank_attrib(c0dra[i]) !=
+ even_rank_attrib(c1dra[i]))
+ return 0;
+
+ /*
+ * If the rank boundaries for the two channels are different
+ * then we're not interleaved.
+ */
+ for (i = 0; i < I3000_RANKS_PER_CHANNEL; i++)
+ if (c0drb[i] != c1drb[i])
+ return 0;
+
+ return 1;
+}
+
+static int i3000_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int rc;
+ int i, j;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ unsigned long last_cumul_size, nr_pages;
+ int interleaved, nr_channels;
+ unsigned char dra[I3000_RANKS / 2], drb[I3000_RANKS];
+ unsigned char *c0dra = dra, *c1dra = &dra[I3000_RANKS_PER_CHANNEL / 2];
+ unsigned char *c0drb = drb, *c1drb = &drb[I3000_RANKS_PER_CHANNEL];
+ unsigned long mchbar;
+ void __iomem *window;
+
+ edac_dbg(0, "MC:\n");
+
+ pci_read_config_dword(pdev, I3000_MCHBAR, (u32 *) & mchbar);
+ mchbar &= I3000_MCHBAR_MASK;
+ window = ioremap(mchbar, I3000_MMR_WINDOW_SIZE);
+ if (!window) {
+ printk(KERN_ERR "i3000: cannot map mmio space at 0x%lx\n",
+ mchbar);
+ return -ENODEV;
+ }
+
+ c0dra[0] = readb(window + I3000_C0DRA + 0); /* ranks 0,1 */
+ c0dra[1] = readb(window + I3000_C0DRA + 1); /* ranks 2,3 */
+ c1dra[0] = readb(window + I3000_C1DRA + 0); /* ranks 0,1 */
+ c1dra[1] = readb(window + I3000_C1DRA + 1); /* ranks 2,3 */
+
+ for (i = 0; i < I3000_RANKS_PER_CHANNEL; i++) {
+ c0drb[i] = readb(window + I3000_C0DRB + i);
+ c1drb[i] = readb(window + I3000_C1DRB + i);
+ }
+
+ iounmap(window);
+
+ /*
+ * Figure out how many channels we have.
+ *
+ * If we have what the datasheet calls "asymmetric channels"
+ * (essentially the same as what was called "virtual single
+ * channel mode" in the i82875) then it's a single channel as
+ * far as EDAC is concerned.
+ */
+ interleaved = i3000_is_interleaved(c0dra, c1dra, c0drb, c1drb);
+ nr_channels = interleaved ? 2 : 1;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = I3000_RANKS / nr_channels;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = nr_channels;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (!mci)
+ return -ENOMEM;
+
+ edac_dbg(3, "MC: init mci\n");
+
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR2;
+
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = i3000_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i3000_check;
+ mci->ctl_page_to_phys = NULL;
+
+ /*
+ * The dram rank boundary (DRB) reg values are boundary addresses
+ * for each DRAM rank with a granularity of 32MB. DRB regs are
+ * cumulative; the last one will contain the total memory
+ * contained in all ranks.
+ *
+ * If we're in interleaved mode then we're only walking through
+ * the ranks of controller 0, so we double all the values we see.
+ */
+ for (last_cumul_size = i = 0; i < mci->nr_csrows; i++) {
+ u8 value;
+ u32 cumul_size;
+ struct csrow_info *csrow = mci->csrows[i];
+
+ value = drb[i];
+ cumul_size = value << (I3000_DRB_SHIFT - PAGE_SHIFT);
+ if (interleaved)
+ cumul_size <<= 1;
+ edac_dbg(3, "MC: (%d) cumul_size 0x%x\n", i, cumul_size);
+ if (cumul_size == last_cumul_size)
+ continue;
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ nr_pages = cumul_size - last_cumul_size;
+ last_cumul_size = cumul_size;
+
+ for (j = 0; j < nr_channels; j++) {
+ struct dimm_info *dimm = csrow->channels[j]->dimm;
+
+ dimm->nr_pages = nr_pages / nr_channels;
+ dimm->grain = I3000_DEAP_GRAIN;
+ dimm->mtype = MEM_DDR2;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_UNKNOWN;
+ }
+ }
+
+ /*
+ * Clear any error bits.
+ * (Yes, we really clear bits by writing 1 to them.)
+ */
+ pci_write_bits16(pdev, I3000_ERRSTS, I3000_ERRSTS_BITS,
+ I3000_ERRSTS_BITS);
+
+ rc = -ENODEV;
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "MC: failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* allocating generic PCI control info */
+ i3000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i3000_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "MC: success\n");
+ return 0;
+
+fail:
+ if (mci)
+ edac_mc_free(mci);
+
+ return rc;
+}
+
+/* returns count (>= 0), or negative on error */
+static int i3000_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = i3000_probe1(pdev, ent->driver_data);
+ if (!mci_pdev)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i3000_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (i3000_pci)
+ edac_pci_release_generic_ctl(i3000_pci);
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i3000_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 3000_HB), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I3000},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i3000_pci_tbl);
+
+static struct pci_driver i3000_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i3000_init_one,
+ .remove = i3000_remove_one,
+ .id_table = i3000_pci_tbl,
+};
+
+static int __init i3000_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "MC:\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i3000_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (!mci_pdev) {
+ i3000_registered = 0;
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_3000_HB, NULL);
+ if (!mci_pdev) {
+ edac_dbg(0, "i3000 pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i3000_init_one(mci_pdev, i3000_pci_tbl);
+ if (pci_rc < 0) {
+ edac_dbg(0, "i3000 init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i3000_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+
+ return pci_rc;
+}
+
+static void __exit i3000_exit(void)
+{
+ edac_dbg(3, "MC:\n");
+
+ pci_unregister_driver(&i3000_driver);
+ if (!i3000_registered) {
+ i3000_remove_one(mci_pdev);
+ pci_dev_put(mci_pdev);
+ }
+}
+
+module_init(i3000_init);
+module_exit(i3000_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Akamai Technologies Arthur Ulfeldt/Jason Uhlenkott");
+MODULE_DESCRIPTION("MC support for Intel 3000 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i3200_edac.c b/drivers/edac/i3200_edac.c
new file mode 100644
index 0000000000..afccdebf5a
--- /dev/null
+++ b/drivers/edac/i3200_edac.c
@@ -0,0 +1,545 @@
+/*
+ * Intel 3200/3210 Memory Controller kernel module
+ * Copyright (C) 2008-2009 Akamai Technologies, Inc.
+ * Portions by Hitoshi Mitake <h.mitake@gmail.com>.
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include <linux/io.h>
+#include "edac_module.h"
+
+#include <linux/io-64-nonatomic-lo-hi.h>
+
+#define EDAC_MOD_STR "i3200_edac"
+
+#define PCI_DEVICE_ID_INTEL_3200_HB 0x29f0
+
+#define I3200_DIMMS 4
+#define I3200_RANKS 8
+#define I3200_RANKS_PER_CHANNEL 4
+#define I3200_CHANNELS 2
+
+/* Intel 3200 register addresses - device 0 function 0 - DRAM Controller */
+
+#define I3200_MCHBAR_LOW 0x48 /* MCH Memory Mapped Register BAR */
+#define I3200_MCHBAR_HIGH 0x4c
+#define I3200_MCHBAR_MASK 0xfffffc000ULL /* bits 35:14 */
+#define I3200_MMR_WINDOW_SIZE 16384
+
+#define I3200_TOM 0xa0 /* Top of Memory (16b)
+ *
+ * 15:10 reserved
+ * 9:0 total populated physical memory
+ */
+#define I3200_TOM_MASK 0x3ff /* bits 9:0 */
+#define I3200_TOM_SHIFT 26 /* 64MiB grain */
+
+#define I3200_ERRSTS 0xc8 /* Error Status Register (16b)
+ *
+ * 15 reserved
+ * 14 Isochronous TBWRR Run Behind FIFO Full
+ * (ITCV)
+ * 13 Isochronous TBWRR Run Behind FIFO Put
+ * (ITSTV)
+ * 12 reserved
+ * 11 MCH Thermal Sensor Event
+ * for SMI/SCI/SERR (GTSE)
+ * 10 reserved
+ * 9 LOCK to non-DRAM Memory Flag (LCKF)
+ * 8 reserved
+ * 7 DRAM Throttle Flag (DTF)
+ * 6:2 reserved
+ * 1 Multi-bit DRAM ECC Error Flag (DMERR)
+ * 0 Single-bit DRAM ECC Error Flag (DSERR)
+ */
+#define I3200_ERRSTS_UE 0x0002
+#define I3200_ERRSTS_CE 0x0001
+#define I3200_ERRSTS_BITS (I3200_ERRSTS_UE | I3200_ERRSTS_CE)
+
+
+/* Intel MMIO register space - device 0 function 0 - MMR space */
+
+#define I3200_C0DRB 0x200 /* Channel 0 DRAM Rank Boundary (16b x 4)
+ *
+ * 15:10 reserved
+ * 9:0 Channel 0 DRAM Rank Boundary Address
+ */
+#define I3200_C1DRB 0x600 /* Channel 1 DRAM Rank Boundary (16b x 4) */
+#define I3200_DRB_MASK 0x3ff /* bits 9:0 */
+#define I3200_DRB_SHIFT 26 /* 64MiB grain */
+
+#define I3200_C0ECCERRLOG 0x280 /* Channel 0 ECC Error Log (64b)
+ *
+ * 63:48 Error Column Address (ERRCOL)
+ * 47:32 Error Row Address (ERRROW)
+ * 31:29 Error Bank Address (ERRBANK)
+ * 28:27 Error Rank Address (ERRRANK)
+ * 26:24 reserved
+ * 23:16 Error Syndrome (ERRSYND)
+ * 15: 2 reserved
+ * 1 Multiple Bit Error Status (MERRSTS)
+ * 0 Correctable Error Status (CERRSTS)
+ */
+#define I3200_C1ECCERRLOG 0x680 /* Chan 1 ECC Error Log (64b) */
+#define I3200_ECCERRLOG_CE 0x1
+#define I3200_ECCERRLOG_UE 0x2
+#define I3200_ECCERRLOG_RANK_BITS 0x18000000
+#define I3200_ECCERRLOG_RANK_SHIFT 27
+#define I3200_ECCERRLOG_SYNDROME_BITS 0xff0000
+#define I3200_ECCERRLOG_SYNDROME_SHIFT 16
+#define I3200_CAPID0 0xe0 /* P.95 of spec for details */
+
+struct i3200_priv {
+ void __iomem *window;
+};
+
+static int nr_channels;
+
+static int how_many_channels(struct pci_dev *pdev)
+{
+ int n_channels;
+
+ unsigned char capid0_8b; /* 8th byte of CAPID0 */
+
+ pci_read_config_byte(pdev, I3200_CAPID0 + 8, &capid0_8b);
+
+ if (capid0_8b & 0x20) { /* check DCD: Dual Channel Disable */
+ edac_dbg(0, "In single channel mode\n");
+ n_channels = 1;
+ } else {
+ edac_dbg(0, "In dual channel mode\n");
+ n_channels = 2;
+ }
+
+ if (capid0_8b & 0x10) /* check if both channels are filled */
+ edac_dbg(0, "2 DIMMS per channel disabled\n");
+ else
+ edac_dbg(0, "2 DIMMS per channel enabled\n");
+
+ return n_channels;
+}
+
+static unsigned long eccerrlog_syndrome(u64 log)
+{
+ return (log & I3200_ECCERRLOG_SYNDROME_BITS) >>
+ I3200_ECCERRLOG_SYNDROME_SHIFT;
+}
+
+static int eccerrlog_row(int channel, u64 log)
+{
+ u64 rank = ((log & I3200_ECCERRLOG_RANK_BITS) >>
+ I3200_ECCERRLOG_RANK_SHIFT);
+ return rank | (channel * I3200_RANKS_PER_CHANNEL);
+}
+
+enum i3200_chips {
+ I3200 = 0,
+};
+
+struct i3200_dev_info {
+ const char *ctl_name;
+};
+
+struct i3200_error_info {
+ u16 errsts;
+ u16 errsts2;
+ u64 eccerrlog[I3200_CHANNELS];
+};
+
+static const struct i3200_dev_info i3200_devs[] = {
+ [I3200] = {
+ .ctl_name = "i3200"
+ },
+};
+
+static struct pci_dev *mci_pdev;
+static int i3200_registered = 1;
+
+
+static void i3200_clear_error_info(struct mem_ctl_info *mci)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * Clear any error bits.
+ * (Yes, we really clear bits by writing 1 to them.)
+ */
+ pci_write_bits16(pdev, I3200_ERRSTS, I3200_ERRSTS_BITS,
+ I3200_ERRSTS_BITS);
+}
+
+static void i3200_get_and_clear_error_info(struct mem_ctl_info *mci,
+ struct i3200_error_info *info)
+{
+ struct pci_dev *pdev;
+ struct i3200_priv *priv = mci->pvt_info;
+ void __iomem *window = priv->window;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, I3200_ERRSTS, &info->errsts);
+ if (!(info->errsts & I3200_ERRSTS_BITS))
+ return;
+
+ info->eccerrlog[0] = readq(window + I3200_C0ECCERRLOG);
+ if (nr_channels == 2)
+ info->eccerrlog[1] = readq(window + I3200_C1ECCERRLOG);
+
+ pci_read_config_word(pdev, I3200_ERRSTS, &info->errsts2);
+
+ /*
+ * If the error is the same for both reads then the first set
+ * of reads is valid. If there is a change then there is a CE
+ * with no info and the second set of reads is valid and
+ * should be UE info.
+ */
+ if ((info->errsts ^ info->errsts2) & I3200_ERRSTS_BITS) {
+ info->eccerrlog[0] = readq(window + I3200_C0ECCERRLOG);
+ if (nr_channels == 2)
+ info->eccerrlog[1] = readq(window + I3200_C1ECCERRLOG);
+ }
+
+ i3200_clear_error_info(mci);
+}
+
+static void i3200_process_error_info(struct mem_ctl_info *mci,
+ struct i3200_error_info *info)
+{
+ int channel;
+ u64 log;
+
+ if (!(info->errsts & I3200_ERRSTS_BITS))
+ return;
+
+ if ((info->errsts ^ info->errsts2) & I3200_ERRSTS_BITS) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ for (channel = 0; channel < nr_channels; channel++) {
+ log = info->eccerrlog[channel];
+ if (log & I3200_ECCERRLOG_UE) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ 0, 0, 0,
+ eccerrlog_row(channel, log),
+ -1, -1,
+ "i3000 UE", "");
+ } else if (log & I3200_ECCERRLOG_CE) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ 0, 0, eccerrlog_syndrome(log),
+ eccerrlog_row(channel, log),
+ -1, -1,
+ "i3000 CE", "");
+ }
+ }
+}
+
+static void i3200_check(struct mem_ctl_info *mci)
+{
+ struct i3200_error_info info;
+
+ i3200_get_and_clear_error_info(mci, &info);
+ i3200_process_error_info(mci, &info);
+}
+
+static void __iomem *i3200_map_mchbar(struct pci_dev *pdev)
+{
+ union {
+ u64 mchbar;
+ struct {
+ u32 mchbar_low;
+ u32 mchbar_high;
+ };
+ } u;
+ void __iomem *window;
+
+ pci_read_config_dword(pdev, I3200_MCHBAR_LOW, &u.mchbar_low);
+ pci_read_config_dword(pdev, I3200_MCHBAR_HIGH, &u.mchbar_high);
+ u.mchbar &= I3200_MCHBAR_MASK;
+
+ if (u.mchbar != (resource_size_t)u.mchbar) {
+ printk(KERN_ERR
+ "i3200: mmio space beyond accessible range (0x%llx)\n",
+ (unsigned long long)u.mchbar);
+ return NULL;
+ }
+
+ window = ioremap(u.mchbar, I3200_MMR_WINDOW_SIZE);
+ if (!window)
+ printk(KERN_ERR "i3200: cannot map mmio space at 0x%llx\n",
+ (unsigned long long)u.mchbar);
+
+ return window;
+}
+
+
+static void i3200_get_drbs(void __iomem *window,
+ u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL])
+{
+ int i;
+
+ for (i = 0; i < I3200_RANKS_PER_CHANNEL; i++) {
+ drbs[0][i] = readw(window + I3200_C0DRB + 2*i) & I3200_DRB_MASK;
+ drbs[1][i] = readw(window + I3200_C1DRB + 2*i) & I3200_DRB_MASK;
+
+ edac_dbg(0, "drb[0][%d] = %d, drb[1][%d] = %d\n", i, drbs[0][i], i, drbs[1][i]);
+ }
+}
+
+static bool i3200_is_stacked(struct pci_dev *pdev,
+ u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL])
+{
+ u16 tom;
+
+ pci_read_config_word(pdev, I3200_TOM, &tom);
+ tom &= I3200_TOM_MASK;
+
+ return drbs[I3200_CHANNELS - 1][I3200_RANKS_PER_CHANNEL - 1] == tom;
+}
+
+static unsigned long drb_to_nr_pages(
+ u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL], bool stacked,
+ int channel, int rank)
+{
+ int n;
+
+ n = drbs[channel][rank];
+ if (!n)
+ return 0;
+
+ if (rank > 0)
+ n -= drbs[channel][rank - 1];
+ if (stacked && (channel == 1) &&
+ drbs[channel][rank] == drbs[channel][I3200_RANKS_PER_CHANNEL - 1])
+ n -= drbs[0][I3200_RANKS_PER_CHANNEL - 1];
+
+ n <<= (I3200_DRB_SHIFT - PAGE_SHIFT);
+ return n;
+}
+
+static int i3200_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int rc;
+ int i, j;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ u16 drbs[I3200_CHANNELS][I3200_RANKS_PER_CHANNEL];
+ bool stacked;
+ void __iomem *window;
+ struct i3200_priv *priv;
+
+ edac_dbg(0, "MC:\n");
+
+ window = i3200_map_mchbar(pdev);
+ if (!window)
+ return -ENODEV;
+
+ i3200_get_drbs(window, drbs);
+ nr_channels = how_many_channels(pdev);
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = I3200_DIMMS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = nr_channels;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct i3200_priv));
+ if (!mci)
+ return -ENOMEM;
+
+ edac_dbg(3, "MC: init mci\n");
+
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR2;
+
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = i3200_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i3200_check;
+ mci->ctl_page_to_phys = NULL;
+ priv = mci->pvt_info;
+ priv->window = window;
+
+ stacked = i3200_is_stacked(pdev, drbs);
+
+ /*
+ * The dram rank boundary (DRB) reg values are boundary addresses
+ * for each DRAM rank with a granularity of 64MB. DRB regs are
+ * cumulative; the last one will contain the total memory
+ * contained in all ranks.
+ */
+ for (i = 0; i < I3200_DIMMS; i++) {
+ unsigned long nr_pages;
+
+ for (j = 0; j < nr_channels; j++) {
+ struct dimm_info *dimm = edac_get_dimm(mci, i, j, 0);
+
+ nr_pages = drb_to_nr_pages(drbs, stacked, j, i);
+ if (nr_pages == 0)
+ continue;
+
+ edac_dbg(0, "csrow %d, channel %d%s, size = %ld MiB\n", i, j,
+ stacked ? " (stacked)" : "", PAGES_TO_MiB(nr_pages));
+
+ dimm->nr_pages = nr_pages;
+ dimm->grain = nr_pages << PAGE_SHIFT;
+ dimm->mtype = MEM_DDR2;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_UNKNOWN;
+ }
+ }
+
+ i3200_clear_error_info(mci);
+
+ rc = -ENODEV;
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "MC: failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "MC: success\n");
+ return 0;
+
+fail:
+ iounmap(window);
+ if (mci)
+ edac_mc_free(mci);
+
+ return rc;
+}
+
+static int i3200_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = i3200_probe1(pdev, ent->driver_data);
+ if (!mci_pdev)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i3200_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct i3200_priv *priv;
+
+ edac_dbg(0, "\n");
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ priv = mci->pvt_info;
+ iounmap(priv->window);
+
+ edac_mc_free(mci);
+
+ pci_disable_device(pdev);
+}
+
+static const struct pci_device_id i3200_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 3200_HB), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I3200},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i3200_pci_tbl);
+
+static struct pci_driver i3200_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i3200_init_one,
+ .remove = i3200_remove_one,
+ .id_table = i3200_pci_tbl,
+};
+
+static int __init i3200_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "MC:\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i3200_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (!mci_pdev) {
+ i3200_registered = 0;
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_3200_HB, NULL);
+ if (!mci_pdev) {
+ edac_dbg(0, "i3200 pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i3200_init_one(mci_pdev, i3200_pci_tbl);
+ if (pci_rc < 0) {
+ edac_dbg(0, "i3200 init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i3200_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+
+ return pci_rc;
+}
+
+static void __exit i3200_exit(void)
+{
+ edac_dbg(3, "MC:\n");
+
+ pci_unregister_driver(&i3200_driver);
+ if (!i3200_registered) {
+ i3200_remove_one(mci_pdev);
+ pci_dev_put(mci_pdev);
+ }
+}
+
+module_init(i3200_init);
+module_exit(i3200_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Akamai Technologies, Inc.");
+MODULE_DESCRIPTION("MC support for Intel 3200 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i5000_edac.c b/drivers/edac/i5000_edac.c
new file mode 100644
index 0000000000..4b5a71f873
--- /dev/null
+++ b/drivers/edac/i5000_edac.c
@@ -0,0 +1,1582 @@
+/*
+ * Intel 5000(P/V/X) class Memory Controllers kernel module
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Douglas Thompson Linux Networx (http://lnxi.com)
+ * norsk5@xmission.com
+ *
+ * This module is based on the following document:
+ *
+ * Intel 5000X Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://developer.intel.com/design/chipsets/datashts/313070.htm
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <asm/mmzone.h>
+
+#include "edac_module.h"
+
+/*
+ * Alter this version for the I5000 module when modifications are made
+ */
+#define I5000_REVISION " Ver: 2.0.12"
+#define EDAC_MOD_STR "i5000_edac"
+
+#define i5000_printk(level, fmt, arg...) \
+ edac_printk(level, "i5000", fmt, ##arg)
+
+#define i5000_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i5000", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_FBD_0
+#define PCI_DEVICE_ID_INTEL_FBD_0 0x25F5
+#endif
+#ifndef PCI_DEVICE_ID_INTEL_FBD_1
+#define PCI_DEVICE_ID_INTEL_FBD_1 0x25F6
+#endif
+
+/* Device 16,
+ * Function 0: System Address
+ * Function 1: Memory Branch Map, Control, Errors Register
+ * Function 2: FSB Error Registers
+ *
+ * All 3 functions of Device 16 (0,1,2) share the SAME DID
+ */
+#define PCI_DEVICE_ID_INTEL_I5000_DEV16 0x25F0
+
+/* OFFSETS for Function 0 */
+
+/* OFFSETS for Function 1 */
+#define AMBASE 0x48
+#define MAXCH 0x56
+#define MAXDIMMPERCH 0x57
+#define TOLM 0x6C
+#define REDMEMB 0x7C
+#define RED_ECC_LOCATOR(x) ((x) & 0x3FFFF)
+#define REC_ECC_LOCATOR_EVEN(x) ((x) & 0x001FF)
+#define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3FE00)
+#define MIR0 0x80
+#define MIR1 0x84
+#define MIR2 0x88
+#define AMIR0 0x8C
+#define AMIR1 0x90
+#define AMIR2 0x94
+
+#define FERR_FAT_FBD 0x98
+#define NERR_FAT_FBD 0x9C
+#define EXTRACT_FBDCHAN_INDX(x) (((x)>>28) & 0x3)
+#define FERR_FAT_FBDCHAN 0x30000000
+#define FERR_FAT_M3ERR 0x00000004
+#define FERR_FAT_M2ERR 0x00000002
+#define FERR_FAT_M1ERR 0x00000001
+#define FERR_FAT_MASK (FERR_FAT_M1ERR | \
+ FERR_FAT_M2ERR | \
+ FERR_FAT_M3ERR)
+
+#define FERR_NF_FBD 0xA0
+
+/* Thermal and SPD or BFD errors */
+#define FERR_NF_M28ERR 0x01000000
+#define FERR_NF_M27ERR 0x00800000
+#define FERR_NF_M26ERR 0x00400000
+#define FERR_NF_M25ERR 0x00200000
+#define FERR_NF_M24ERR 0x00100000
+#define FERR_NF_M23ERR 0x00080000
+#define FERR_NF_M22ERR 0x00040000
+#define FERR_NF_M21ERR 0x00020000
+
+/* Correctable errors */
+#define FERR_NF_M20ERR 0x00010000
+#define FERR_NF_M19ERR 0x00008000
+#define FERR_NF_M18ERR 0x00004000
+#define FERR_NF_M17ERR 0x00002000
+
+/* Non-Retry or redundant Retry errors */
+#define FERR_NF_M16ERR 0x00001000
+#define FERR_NF_M15ERR 0x00000800
+#define FERR_NF_M14ERR 0x00000400
+#define FERR_NF_M13ERR 0x00000200
+
+/* Uncorrectable errors */
+#define FERR_NF_M12ERR 0x00000100
+#define FERR_NF_M11ERR 0x00000080
+#define FERR_NF_M10ERR 0x00000040
+#define FERR_NF_M9ERR 0x00000020
+#define FERR_NF_M8ERR 0x00000010
+#define FERR_NF_M7ERR 0x00000008
+#define FERR_NF_M6ERR 0x00000004
+#define FERR_NF_M5ERR 0x00000002
+#define FERR_NF_M4ERR 0x00000001
+
+#define FERR_NF_UNCORRECTABLE (FERR_NF_M12ERR | \
+ FERR_NF_M11ERR | \
+ FERR_NF_M10ERR | \
+ FERR_NF_M9ERR | \
+ FERR_NF_M8ERR | \
+ FERR_NF_M7ERR | \
+ FERR_NF_M6ERR | \
+ FERR_NF_M5ERR | \
+ FERR_NF_M4ERR)
+#define FERR_NF_CORRECTABLE (FERR_NF_M20ERR | \
+ FERR_NF_M19ERR | \
+ FERR_NF_M18ERR | \
+ FERR_NF_M17ERR)
+#define FERR_NF_DIMM_SPARE (FERR_NF_M27ERR | \
+ FERR_NF_M28ERR)
+#define FERR_NF_THERMAL (FERR_NF_M26ERR | \
+ FERR_NF_M25ERR | \
+ FERR_NF_M24ERR | \
+ FERR_NF_M23ERR)
+#define FERR_NF_SPD_PROTOCOL (FERR_NF_M22ERR)
+#define FERR_NF_NORTH_CRC (FERR_NF_M21ERR)
+#define FERR_NF_NON_RETRY (FERR_NF_M13ERR | \
+ FERR_NF_M14ERR | \
+ FERR_NF_M15ERR)
+
+#define NERR_NF_FBD 0xA4
+#define FERR_NF_MASK (FERR_NF_UNCORRECTABLE | \
+ FERR_NF_CORRECTABLE | \
+ FERR_NF_DIMM_SPARE | \
+ FERR_NF_THERMAL | \
+ FERR_NF_SPD_PROTOCOL | \
+ FERR_NF_NORTH_CRC | \
+ FERR_NF_NON_RETRY)
+
+#define EMASK_FBD 0xA8
+#define EMASK_FBD_M28ERR 0x08000000
+#define EMASK_FBD_M27ERR 0x04000000
+#define EMASK_FBD_M26ERR 0x02000000
+#define EMASK_FBD_M25ERR 0x01000000
+#define EMASK_FBD_M24ERR 0x00800000
+#define EMASK_FBD_M23ERR 0x00400000
+#define EMASK_FBD_M22ERR 0x00200000
+#define EMASK_FBD_M21ERR 0x00100000
+#define EMASK_FBD_M20ERR 0x00080000
+#define EMASK_FBD_M19ERR 0x00040000
+#define EMASK_FBD_M18ERR 0x00020000
+#define EMASK_FBD_M17ERR 0x00010000
+
+#define EMASK_FBD_M15ERR 0x00004000
+#define EMASK_FBD_M14ERR 0x00002000
+#define EMASK_FBD_M13ERR 0x00001000
+#define EMASK_FBD_M12ERR 0x00000800
+#define EMASK_FBD_M11ERR 0x00000400
+#define EMASK_FBD_M10ERR 0x00000200
+#define EMASK_FBD_M9ERR 0x00000100
+#define EMASK_FBD_M8ERR 0x00000080
+#define EMASK_FBD_M7ERR 0x00000040
+#define EMASK_FBD_M6ERR 0x00000020
+#define EMASK_FBD_M5ERR 0x00000010
+#define EMASK_FBD_M4ERR 0x00000008
+#define EMASK_FBD_M3ERR 0x00000004
+#define EMASK_FBD_M2ERR 0x00000002
+#define EMASK_FBD_M1ERR 0x00000001
+
+#define ENABLE_EMASK_FBD_FATAL_ERRORS (EMASK_FBD_M1ERR | \
+ EMASK_FBD_M2ERR | \
+ EMASK_FBD_M3ERR)
+
+#define ENABLE_EMASK_FBD_UNCORRECTABLE (EMASK_FBD_M4ERR | \
+ EMASK_FBD_M5ERR | \
+ EMASK_FBD_M6ERR | \
+ EMASK_FBD_M7ERR | \
+ EMASK_FBD_M8ERR | \
+ EMASK_FBD_M9ERR | \
+ EMASK_FBD_M10ERR | \
+ EMASK_FBD_M11ERR | \
+ EMASK_FBD_M12ERR)
+#define ENABLE_EMASK_FBD_CORRECTABLE (EMASK_FBD_M17ERR | \
+ EMASK_FBD_M18ERR | \
+ EMASK_FBD_M19ERR | \
+ EMASK_FBD_M20ERR)
+#define ENABLE_EMASK_FBD_DIMM_SPARE (EMASK_FBD_M27ERR | \
+ EMASK_FBD_M28ERR)
+#define ENABLE_EMASK_FBD_THERMALS (EMASK_FBD_M26ERR | \
+ EMASK_FBD_M25ERR | \
+ EMASK_FBD_M24ERR | \
+ EMASK_FBD_M23ERR)
+#define ENABLE_EMASK_FBD_SPD_PROTOCOL (EMASK_FBD_M22ERR)
+#define ENABLE_EMASK_FBD_NORTH_CRC (EMASK_FBD_M21ERR)
+#define ENABLE_EMASK_FBD_NON_RETRY (EMASK_FBD_M15ERR | \
+ EMASK_FBD_M14ERR | \
+ EMASK_FBD_M13ERR)
+
+#define ENABLE_EMASK_ALL (ENABLE_EMASK_FBD_NON_RETRY | \
+ ENABLE_EMASK_FBD_NORTH_CRC | \
+ ENABLE_EMASK_FBD_SPD_PROTOCOL | \
+ ENABLE_EMASK_FBD_THERMALS | \
+ ENABLE_EMASK_FBD_DIMM_SPARE | \
+ ENABLE_EMASK_FBD_FATAL_ERRORS | \
+ ENABLE_EMASK_FBD_CORRECTABLE | \
+ ENABLE_EMASK_FBD_UNCORRECTABLE)
+
+#define ERR0_FBD 0xAC
+#define ERR1_FBD 0xB0
+#define ERR2_FBD 0xB4
+#define MCERR_FBD 0xB8
+#define NRECMEMA 0xBE
+#define NREC_BANK(x) (((x)>>12) & 0x7)
+#define NREC_RDWR(x) (((x)>>11) & 1)
+#define NREC_RANK(x) (((x)>>8) & 0x7)
+#define NRECMEMB 0xC0
+#define NREC_CAS(x) (((x)>>16) & 0xFFF)
+#define NREC_RAS(x) ((x) & 0x7FFF)
+#define NRECFGLOG 0xC4
+#define NREEECFBDA 0xC8
+#define NREEECFBDB 0xCC
+#define NREEECFBDC 0xD0
+#define NREEECFBDD 0xD4
+#define NREEECFBDE 0xD8
+#define REDMEMA 0xDC
+#define RECMEMA 0xE2
+#define REC_BANK(x) (((x)>>12) & 0x7)
+#define REC_RDWR(x) (((x)>>11) & 1)
+#define REC_RANK(x) (((x)>>8) & 0x7)
+#define RECMEMB 0xE4
+#define REC_CAS(x) (((x)>>16) & 0xFFFFFF)
+#define REC_RAS(x) ((x) & 0x7FFF)
+#define RECFGLOG 0xE8
+#define RECFBDA 0xEC
+#define RECFBDB 0xF0
+#define RECFBDC 0xF4
+#define RECFBDD 0xF8
+#define RECFBDE 0xFC
+
+/* OFFSETS for Function 2 */
+
+/*
+ * Device 21,
+ * Function 0: Memory Map Branch 0
+ *
+ * Device 22,
+ * Function 0: Memory Map Branch 1
+ */
+#define PCI_DEVICE_ID_I5000_BRANCH_0 0x25F5
+#define PCI_DEVICE_ID_I5000_BRANCH_1 0x25F6
+
+#define AMB_PRESENT_0 0x64
+#define AMB_PRESENT_1 0x66
+#define MTR0 0x80
+#define MTR1 0x84
+#define MTR2 0x88
+#define MTR3 0x8C
+
+#define NUM_MTRS 4
+#define CHANNELS_PER_BRANCH 2
+#define MAX_BRANCHES 2
+
+/* Defines to extract the various fields from the
+ * MTRx - Memory Technology Registers
+ */
+#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (0x1 << 8))
+#define MTR_DRAM_WIDTH(mtr) ((((mtr) >> 6) & 0x1) ? 8 : 4)
+#define MTR_DRAM_BANKS(mtr) ((((mtr) >> 5) & 0x1) ? 8 : 4)
+#define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
+#define MTR_DIMM_RANK(mtr) (((mtr) >> 4) & 0x1)
+#define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
+#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
+#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
+#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
+#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
+
+/* enables the report of miscellaneous messages as CE errors - default off */
+static int misc_messages;
+
+/* Enumeration of supported devices */
+enum i5000_chips {
+ I5000P = 0,
+ I5000V = 1, /* future */
+ I5000X = 2 /* future */
+};
+
+/* Device name and register DID (Device ID) */
+struct i5000_dev_info {
+ const char *ctl_name; /* name for this device */
+ u16 fsb_mapping_errors; /* DID for the branchmap,control */
+};
+
+/* Table of devices attributes supported by this driver */
+static const struct i5000_dev_info i5000_devs[] = {
+ [I5000P] = {
+ .ctl_name = "I5000",
+ .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I5000_DEV16,
+ },
+};
+
+struct i5000_dimm_info {
+ int megabytes; /* size, 0 means not present */
+ int dual_rank;
+};
+
+#define MAX_CHANNELS 6 /* max possible channels */
+#define MAX_CSROWS (8*2) /* max possible csrows per channel */
+
+/* driver private data structure */
+struct i5000_pvt {
+ struct pci_dev *system_address; /* 16.0 */
+ struct pci_dev *branchmap_werrors; /* 16.1 */
+ struct pci_dev *fsb_error_regs; /* 16.2 */
+ struct pci_dev *branch_0; /* 21.0 */
+ struct pci_dev *branch_1; /* 22.0 */
+
+ u16 tolm; /* top of low memory */
+ union {
+ u64 ambase; /* AMB BAR */
+ struct {
+ u32 ambase_bottom;
+ u32 ambase_top;
+ } u __packed;
+ };
+
+ u16 mir0, mir1, mir2;
+
+ u16 b0_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
+ u16 b0_ambpresent0; /* Branch 0, Channel 0 */
+ u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
+
+ u16 b1_mtr[NUM_MTRS]; /* Memory Technlogy Reg */
+ u16 b1_ambpresent0; /* Branch 1, Channel 8 */
+ u16 b1_ambpresent1; /* Branch 1, Channel 1 */
+
+ /* DIMM information matrix, allocating architecture maximums */
+ struct i5000_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
+
+ /* Actual values for this controller */
+ int maxch; /* Max channels */
+ int maxdimmperch; /* Max DIMMs per channel */
+};
+
+/* I5000 MCH error information retrieved from Hardware */
+struct i5000_error_info {
+
+ /* These registers are always read from the MC */
+ u32 ferr_fat_fbd; /* First Errors Fatal */
+ u32 nerr_fat_fbd; /* Next Errors Fatal */
+ u32 ferr_nf_fbd; /* First Errors Non-Fatal */
+ u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
+
+ /* These registers are input ONLY if there was a Recoverable Error */
+ u32 redmemb; /* Recoverable Mem Data Error log B */
+ u16 recmema; /* Recoverable Mem Error log A */
+ u32 recmemb; /* Recoverable Mem Error log B */
+
+ /* These registers are input ONLY if there was a
+ * Non-Recoverable Error */
+ u16 nrecmema; /* Non-Recoverable Mem log A */
+ u32 nrecmemb; /* Non-Recoverable Mem log B */
+
+};
+
+static struct edac_pci_ctl_info *i5000_pci;
+
+/*
+ * i5000_get_error_info Retrieve the hardware error information from
+ * the hardware and cache it in the 'info'
+ * structure
+ */
+static void i5000_get_error_info(struct mem_ctl_info *mci,
+ struct i5000_error_info *info)
+{
+ struct i5000_pvt *pvt;
+ u32 value;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
+
+ /* Mask only the bits that the doc says are valid
+ */
+ value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
+
+ /* If there is an error, then read in the */
+ /* NEXT FATAL error register and the Memory Error Log Register A */
+ if (value & FERR_FAT_MASK) {
+ info->ferr_fat_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_FAT_FBD, &info->nerr_fat_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMA, &info->nrecmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NRECMEMB, &info->nrecmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_FAT_FBD, value);
+ } else {
+ info->ferr_fat_fbd = 0;
+ info->nerr_fat_fbd = 0;
+ info->nrecmema = 0;
+ info->nrecmemb = 0;
+ }
+
+ /* read in the 1st NON-FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
+
+ /* If there is an error, then read in the 1st NON-FATAL error
+ * register as well */
+ if (value & FERR_NF_MASK) {
+ info->ferr_nf_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_NF_FBD, &info->nerr_nf_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ RECMEMA, &info->recmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ RECMEMB, &info->recmemb);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ REDMEMB, &info->redmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_NF_FBD, value);
+ } else {
+ info->ferr_nf_fbd = 0;
+ info->nerr_nf_fbd = 0;
+ info->recmema = 0;
+ info->recmemb = 0;
+ info->redmemb = 0;
+ }
+}
+
+/*
+ * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
+ * struct i5000_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel FATAL errors, if any
+ */
+static void i5000_process_fatal_error_info(struct mem_ctl_info *mci,
+ struct i5000_error_info *info,
+ int handle_errors)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 160];
+ char *specific = NULL;
+ u32 allErrors;
+ int channel;
+ int bank;
+ int rank;
+ int rdwr;
+ int ras, cas;
+
+ /* mask off the Error bits that are possible */
+ allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ channel = EXTRACT_FBDCHAN_INDX(info->ferr_fat_fbd);
+
+ /* Use the NON-Recoverable macros to extract data */
+ bank = NREC_BANK(info->nrecmema);
+ rank = NREC_RANK(info->nrecmema);
+ rdwr = NREC_RDWR(info->nrecmema);
+ ras = NREC_RAS(info->nrecmemb);
+ cas = NREC_CAS(info->nrecmemb);
+
+ edac_dbg(0, "\t\tCSROW= %d Channel= %d (DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, bank,
+ rdwr ? "Write" : "Read", ras, cas);
+
+ /* Only 1 bit will be on */
+ switch (allErrors) {
+ case FERR_FAT_M1ERR:
+ specific = "Alert on non-redundant retry or fast "
+ "reset timeout";
+ break;
+ case FERR_FAT_M2ERR:
+ specific = "Northbound CRC error on non-redundant "
+ "retry";
+ break;
+ case FERR_FAT_M3ERR:
+ {
+ static int done;
+
+ /*
+ * This error is generated to inform that the intelligent
+ * throttling is disabled and the temperature passed the
+ * specified middle point. Since this is something the BIOS
+ * should take care of, we'll warn only once to avoid
+ * worthlessly flooding the log.
+ */
+ if (done)
+ return;
+ done++;
+
+ specific = ">Tmid Thermal event with intelligent "
+ "throttling disabled";
+ }
+ break;
+ }
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Bank=%d RAS=%d CAS=%d FATAL Err=0x%x (%s)",
+ bank, ras, cas, allErrors, specific);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, 1, 0, 0, 0,
+ channel >> 1, channel & 1, rank,
+ rdwr ? "Write error" : "Read error",
+ msg);
+}
+
+/*
+ * i5000_process_fatal_error_info(struct mem_ctl_info *mci,
+ * struct i5000_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel NON-FATAL errors, if any
+ */
+static void i5000_process_nonfatal_error_info(struct mem_ctl_info *mci,
+ struct i5000_error_info *info,
+ int handle_errors)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 170];
+ char *specific = NULL;
+ u32 allErrors;
+ u32 ue_errors;
+ u32 ce_errors;
+ u32 misc_errors;
+ int branch;
+ int channel;
+ int bank;
+ int rank;
+ int rdwr;
+ int ras, cas;
+
+ /* mask off the Error bits that are possible */
+ allErrors = (info->ferr_nf_fbd & FERR_NF_MASK);
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+ ue_errors = allErrors & FERR_NF_UNCORRECTABLE;
+ if (ue_errors) {
+ edac_dbg(0, "\tUncorrected bits= 0x%x\n", ue_errors);
+
+ branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
+
+ /*
+ * According with i5000 datasheet, bit 28 has no significance
+ * for errors M4Err-M12Err and M17Err-M21Err, on FERR_NF_FBD
+ */
+ channel = branch & 2;
+
+ bank = NREC_BANK(info->nrecmema);
+ rank = NREC_RANK(info->nrecmema);
+ rdwr = NREC_RDWR(info->nrecmema);
+ ras = NREC_RAS(info->nrecmemb);
+ cas = NREC_CAS(info->nrecmemb);
+
+ edac_dbg(0, "\t\tCSROW= %d Channels= %d,%d (Branch= %d DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, channel + 1, branch >> 1, bank,
+ rdwr ? "Write" : "Read", ras, cas);
+
+ switch (ue_errors) {
+ case FERR_NF_M12ERR:
+ specific = "Non-Aliased Uncorrectable Patrol Data ECC";
+ break;
+ case FERR_NF_M11ERR:
+ specific = "Non-Aliased Uncorrectable Spare-Copy "
+ "Data ECC";
+ break;
+ case FERR_NF_M10ERR:
+ specific = "Non-Aliased Uncorrectable Mirrored Demand "
+ "Data ECC";
+ break;
+ case FERR_NF_M9ERR:
+ specific = "Non-Aliased Uncorrectable Non-Mirrored "
+ "Demand Data ECC";
+ break;
+ case FERR_NF_M8ERR:
+ specific = "Aliased Uncorrectable Patrol Data ECC";
+ break;
+ case FERR_NF_M7ERR:
+ specific = "Aliased Uncorrectable Spare-Copy Data ECC";
+ break;
+ case FERR_NF_M6ERR:
+ specific = "Aliased Uncorrectable Mirrored Demand "
+ "Data ECC";
+ break;
+ case FERR_NF_M5ERR:
+ specific = "Aliased Uncorrectable Non-Mirrored Demand "
+ "Data ECC";
+ break;
+ case FERR_NF_M4ERR:
+ specific = "Uncorrectable Data ECC on Replay";
+ break;
+ }
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Rank=%d Bank=%d RAS=%d CAS=%d, UE Err=0x%x (%s)",
+ rank, bank, ras, cas, ue_errors, specific);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ channel >> 1, -1, rank,
+ rdwr ? "Write error" : "Read error",
+ msg);
+ }
+
+ /* Check correctable errors */
+ ce_errors = allErrors & FERR_NF_CORRECTABLE;
+ if (ce_errors) {
+ edac_dbg(0, "\tCorrected bits= 0x%x\n", ce_errors);
+
+ branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
+
+ channel = 0;
+ if (REC_ECC_LOCATOR_ODD(info->redmemb))
+ channel = 1;
+
+ /* Convert channel to be based from zero, instead of
+ * from branch base of 0 */
+ channel += branch;
+
+ bank = REC_BANK(info->recmema);
+ rank = REC_RANK(info->recmema);
+ rdwr = REC_RDWR(info->recmema);
+ ras = REC_RAS(info->recmemb);
+ cas = REC_CAS(info->recmemb);
+
+ edac_dbg(0, "\t\tCSROW= %d Channel= %d (Branch %d DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, branch >> 1, bank,
+ rdwr ? "Write" : "Read", ras, cas);
+
+ switch (ce_errors) {
+ case FERR_NF_M17ERR:
+ specific = "Correctable Non-Mirrored Demand Data ECC";
+ break;
+ case FERR_NF_M18ERR:
+ specific = "Correctable Mirrored Demand Data ECC";
+ break;
+ case FERR_NF_M19ERR:
+ specific = "Correctable Spare-Copy Data ECC";
+ break;
+ case FERR_NF_M20ERR:
+ specific = "Correctable Patrol Data ECC";
+ break;
+ }
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Rank=%d Bank=%d RDWR=%s RAS=%d "
+ "CAS=%d, CE Err=0x%x (%s))", branch >> 1, bank,
+ rdwr ? "Write" : "Read", ras, cas, ce_errors,
+ specific);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0,
+ channel >> 1, channel % 2, rank,
+ rdwr ? "Write error" : "Read error",
+ msg);
+ }
+
+ if (!misc_messages)
+ return;
+
+ misc_errors = allErrors & (FERR_NF_NON_RETRY | FERR_NF_NORTH_CRC |
+ FERR_NF_SPD_PROTOCOL | FERR_NF_DIMM_SPARE);
+ if (misc_errors) {
+ switch (misc_errors) {
+ case FERR_NF_M13ERR:
+ specific = "Non-Retry or Redundant Retry FBD Memory "
+ "Alert or Redundant Fast Reset Timeout";
+ break;
+ case FERR_NF_M14ERR:
+ specific = "Non-Retry or Redundant Retry FBD "
+ "Configuration Alert";
+ break;
+ case FERR_NF_M15ERR:
+ specific = "Non-Retry or Redundant Retry FBD "
+ "Northbound CRC error on read data";
+ break;
+ case FERR_NF_M21ERR:
+ specific = "FBD Northbound CRC error on "
+ "FBD Sync Status";
+ break;
+ case FERR_NF_M22ERR:
+ specific = "SPD protocol error";
+ break;
+ case FERR_NF_M27ERR:
+ specific = "DIMM-spare copy started";
+ break;
+ case FERR_NF_M28ERR:
+ specific = "DIMM-spare copy completed";
+ break;
+ }
+ branch = EXTRACT_FBDCHAN_INDX(info->ferr_nf_fbd);
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Err=%#x (%s)", misc_errors, specific);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0,
+ branch >> 1, -1, -1,
+ "Misc error", msg);
+ }
+}
+
+/*
+ * i5000_process_error_info Process the error info that is
+ * in the 'info' structure, previously retrieved from hardware
+ */
+static void i5000_process_error_info(struct mem_ctl_info *mci,
+ struct i5000_error_info *info,
+ int handle_errors)
+{
+ /* First handle any fatal errors that occurred */
+ i5000_process_fatal_error_info(mci, info, handle_errors);
+
+ /* now handle any non-fatal errors that occurred */
+ i5000_process_nonfatal_error_info(mci, info, handle_errors);
+}
+
+/*
+ * i5000_clear_error Retrieve any error from the hardware
+ * but do NOT process that error.
+ * Used for 'clearing' out of previous errors
+ * Called by the Core module.
+ */
+static void i5000_clear_error(struct mem_ctl_info *mci)
+{
+ struct i5000_error_info info;
+
+ i5000_get_error_info(mci, &info);
+}
+
+/*
+ * i5000_check_error Retrieve and process errors reported by the
+ * hardware. Called by the Core module.
+ */
+static void i5000_check_error(struct mem_ctl_info *mci)
+{
+ struct i5000_error_info info;
+
+ i5000_get_error_info(mci, &info);
+ i5000_process_error_info(mci, &info, 1);
+}
+
+/*
+ * i5000_get_devices Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ *
+ * Need to 'get' device 16 func 1 and func 2
+ */
+static int i5000_get_devices(struct mem_ctl_info *mci, int dev_idx)
+{
+ //const struct i5000_dev_info *i5000_dev = &i5000_devs[dev_idx];
+ struct i5000_pvt *pvt;
+ struct pci_dev *pdev;
+
+ pvt = mci->pvt_info;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while (1) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
+
+ /* End of list, leave */
+ if (pdev == NULL) {
+ i5000_printk(KERN_ERR,
+ "'system address,Process Bus' "
+ "device not found:"
+ "vendor 0x%x device 0x%x FUNC 1 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I5000_DEV16);
+
+ return 1;
+ }
+
+ /* Scan for device 16 func 1 */
+ if (PCI_FUNC(pdev->devfn) == 1)
+ break;
+ }
+
+ pvt->branchmap_werrors = pdev;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while (1) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I5000_DEV16, pdev);
+
+ if (pdev == NULL) {
+ i5000_printk(KERN_ERR,
+ "MC: 'branchmap,control,errors' "
+ "device not found:"
+ "vendor 0x%x device 0x%x Func 2 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I5000_DEV16);
+
+ pci_dev_put(pvt->branchmap_werrors);
+ return 1;
+ }
+
+ /* Scan for device 16 func 1 */
+ if (PCI_FUNC(pdev->devfn) == 2)
+ break;
+ }
+
+ pvt->fsb_error_regs = pdev;
+
+ edac_dbg(1, "System Address, processor bus- PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->system_address),
+ pvt->system_address->vendor, pvt->system_address->device);
+ edac_dbg(1, "Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->branchmap_werrors),
+ pvt->branchmap_werrors->vendor,
+ pvt->branchmap_werrors->device);
+ edac_dbg(1, "FSB Error Regs - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->fsb_error_regs),
+ pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
+
+ pdev = NULL;
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_I5000_BRANCH_0, pdev);
+
+ if (pdev == NULL) {
+ i5000_printk(KERN_ERR,
+ "MC: 'BRANCH 0' device not found:"
+ "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_I5000_BRANCH_0);
+
+ pci_dev_put(pvt->branchmap_werrors);
+ pci_dev_put(pvt->fsb_error_regs);
+ return 1;
+ }
+
+ pvt->branch_0 = pdev;
+
+ /* If this device claims to have more than 2 channels then
+ * fetch Branch 1's information
+ */
+ if (pvt->maxch >= CHANNELS_PER_BRANCH) {
+ pdev = NULL;
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_I5000_BRANCH_1, pdev);
+
+ if (pdev == NULL) {
+ i5000_printk(KERN_ERR,
+ "MC: 'BRANCH 1' device not found:"
+ "vendor 0x%x device 0x%x Func 0 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_I5000_BRANCH_1);
+
+ pci_dev_put(pvt->branchmap_werrors);
+ pci_dev_put(pvt->fsb_error_regs);
+ pci_dev_put(pvt->branch_0);
+ return 1;
+ }
+
+ pvt->branch_1 = pdev;
+ }
+
+ return 0;
+}
+
+/*
+ * i5000_put_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void i5000_put_devices(struct mem_ctl_info *mci)
+{
+ struct i5000_pvt *pvt;
+
+ pvt = mci->pvt_info;
+
+ pci_dev_put(pvt->branchmap_werrors); /* FUNC 1 */
+ pci_dev_put(pvt->fsb_error_regs); /* FUNC 2 */
+ pci_dev_put(pvt->branch_0); /* DEV 21 */
+
+ /* Only if more than 2 channels do we release the second branch */
+ if (pvt->maxch >= CHANNELS_PER_BRANCH)
+ pci_dev_put(pvt->branch_1); /* DEV 22 */
+}
+
+/*
+ * determine_amb_resent
+ *
+ * the information is contained in NUM_MTRS different registers
+ * determineing which of the NUM_MTRS requires knowing
+ * which channel is in question
+ *
+ * 2 branches, each with 2 channels
+ * b0_ambpresent0 for channel '0'
+ * b0_ambpresent1 for channel '1'
+ * b1_ambpresent0 for channel '2'
+ * b1_ambpresent1 for channel '3'
+ */
+static int determine_amb_present_reg(struct i5000_pvt *pvt, int channel)
+{
+ int amb_present;
+
+ if (channel < CHANNELS_PER_BRANCH) {
+ if (channel & 0x1)
+ amb_present = pvt->b0_ambpresent1;
+ else
+ amb_present = pvt->b0_ambpresent0;
+ } else {
+ if (channel & 0x1)
+ amb_present = pvt->b1_ambpresent1;
+ else
+ amb_present = pvt->b1_ambpresent0;
+ }
+
+ return amb_present;
+}
+
+/*
+ * determine_mtr(pvt, csrow, channel)
+ *
+ * return the proper MTR register as determine by the csrow and channel desired
+ */
+static int determine_mtr(struct i5000_pvt *pvt, int slot, int channel)
+{
+ int mtr;
+
+ if (channel < CHANNELS_PER_BRANCH)
+ mtr = pvt->b0_mtr[slot];
+ else
+ mtr = pvt->b1_mtr[slot];
+
+ return mtr;
+}
+
+/*
+ */
+static void decode_mtr(int slot_row, u16 mtr)
+{
+ int ans;
+
+ ans = MTR_DIMMS_PRESENT(mtr);
+
+ edac_dbg(2, "\tMTR%d=0x%x: DIMMs are %sPresent\n",
+ slot_row, mtr, ans ? "" : "NOT ");
+ if (!ans)
+ return;
+
+ edac_dbg(2, "\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
+ edac_dbg(2, "\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
+ edac_dbg(2, "\t\tNUMRANK: %s\n",
+ MTR_DIMM_RANK(mtr) ? "double" : "single");
+ edac_dbg(2, "\t\tNUMROW: %s\n",
+ MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" :
+ MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" :
+ MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" :
+ "reserved");
+ edac_dbg(2, "\t\tNUMCOL: %s\n",
+ MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" :
+ MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" :
+ MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" :
+ "reserved");
+}
+
+static void handle_channel(struct i5000_pvt *pvt, int slot, int channel,
+ struct i5000_dimm_info *dinfo)
+{
+ int mtr;
+ int amb_present_reg;
+ int addrBits;
+
+ mtr = determine_mtr(pvt, slot, channel);
+ if (MTR_DIMMS_PRESENT(mtr)) {
+ amb_present_reg = determine_amb_present_reg(pvt, channel);
+
+ /* Determine if there is a DIMM present in this DIMM slot */
+ if (amb_present_reg) {
+ dinfo->dual_rank = MTR_DIMM_RANK(mtr);
+
+ /* Start with the number of bits for a Bank
+ * on the DRAM */
+ addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
+ /* Add the number of ROW bits */
+ addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
+ /* add the number of COLUMN bits */
+ addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
+
+ /* Dual-rank memories have twice the size */
+ if (dinfo->dual_rank)
+ addrBits++;
+
+ addrBits += 6; /* add 64 bits per DIMM */
+ addrBits -= 20; /* divide by 2^^20 */
+ addrBits -= 3; /* 8 bits per bytes */
+
+ dinfo->megabytes = 1 << addrBits;
+ }
+ }
+}
+
+/*
+ * calculate_dimm_size
+ *
+ * also will output a DIMM matrix map, if debug is enabled, for viewing
+ * how the DIMMs are populated
+ */
+static void calculate_dimm_size(struct i5000_pvt *pvt)
+{
+ struct i5000_dimm_info *dinfo;
+ int slot, channel, branch;
+ char *p, *mem_buffer;
+ int space, n;
+
+ /* ================= Generate some debug output ================= */
+ space = PAGE_SIZE;
+ mem_buffer = p = kmalloc(space, GFP_KERNEL);
+ if (p == NULL) {
+ i5000_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
+ __FILE__, __func__);
+ return;
+ }
+
+ /* Scan all the actual slots
+ * and calculate the information for each DIMM
+ * Start with the highest slot first, to display it first
+ * and work toward the 0th slot
+ */
+ for (slot = pvt->maxdimmperch - 1; slot >= 0; slot--) {
+
+ /* on an odd slot, first output a 'boundary' marker,
+ * then reset the message buffer */
+ if (slot & 0x1) {
+ n = snprintf(p, space, "--------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+ n = snprintf(p, space, "slot %2d ", slot);
+ p += n;
+ space -= n;
+
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ dinfo = &pvt->dimm_info[slot][channel];
+ handle_channel(pvt, slot, channel, dinfo);
+ if (dinfo->megabytes)
+ n = snprintf(p, space, "%4d MB %dR| ",
+ dinfo->megabytes, dinfo->dual_rank + 1);
+ else
+ n = snprintf(p, space, "%4d MB | ", 0);
+ p += n;
+ space -= n;
+ }
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+
+ /* Output the last bottom 'boundary' marker */
+ n = snprintf(p, space, "--------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ /* now output the 'channel' labels */
+ n = snprintf(p, space, " ");
+ p += n;
+ space -= n;
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ n = snprintf(p, space, "channel %d | ", channel);
+ p += n;
+ space -= n;
+ }
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ n = snprintf(p, space, " ");
+ p += n;
+ for (branch = 0; branch < MAX_BRANCHES; branch++) {
+ n = snprintf(p, space, " branch %d | ", branch);
+ p += n;
+ space -= n;
+ }
+
+ /* output the last message and free buffer */
+ edac_dbg(2, "%s\n", mem_buffer);
+ kfree(mem_buffer);
+}
+
+/*
+ * i5000_get_mc_regs read in the necessary registers and
+ * cache locally
+ *
+ * Fills in the private data members
+ */
+static void i5000_get_mc_regs(struct mem_ctl_info *mci)
+{
+ struct i5000_pvt *pvt;
+ u32 actual_tolm;
+ u16 limit;
+ int slot_row;
+ int way0, way1;
+
+ pvt = mci->pvt_info;
+
+ pci_read_config_dword(pvt->system_address, AMBASE,
+ &pvt->u.ambase_bottom);
+ pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
+ &pvt->u.ambase_top);
+
+ edac_dbg(2, "AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
+ (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
+
+ /* Get the Branch Map regs */
+ pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
+ pvt->tolm >>= 12;
+ edac_dbg(2, "TOLM (number of 256M regions) =%u (0x%x)\n",
+ pvt->tolm, pvt->tolm);
+
+ actual_tolm = pvt->tolm << 28;
+ edac_dbg(2, "Actual TOLM byte addr=%u (0x%x)\n",
+ actual_tolm, actual_tolm);
+
+ pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
+ pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
+ pci_read_config_word(pvt->branchmap_werrors, MIR2, &pvt->mir2);
+
+ /* Get the MIR[0-2] regs */
+ limit = (pvt->mir0 >> 4) & 0x0FFF;
+ way0 = pvt->mir0 & 0x1;
+ way1 = pvt->mir0 & 0x2;
+ edac_dbg(2, "MIR0: limit= 0x%x WAY1= %u WAY0= %x\n",
+ limit, way1, way0);
+ limit = (pvt->mir1 >> 4) & 0x0FFF;
+ way0 = pvt->mir1 & 0x1;
+ way1 = pvt->mir1 & 0x2;
+ edac_dbg(2, "MIR1: limit= 0x%x WAY1= %u WAY0= %x\n",
+ limit, way1, way0);
+ limit = (pvt->mir2 >> 4) & 0x0FFF;
+ way0 = pvt->mir2 & 0x1;
+ way1 = pvt->mir2 & 0x2;
+ edac_dbg(2, "MIR2: limit= 0x%x WAY1= %u WAY0= %x\n",
+ limit, way1, way0);
+
+ /* Get the MTR[0-3] regs */
+ for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
+ int where = MTR0 + (slot_row * sizeof(u32));
+
+ pci_read_config_word(pvt->branch_0, where,
+ &pvt->b0_mtr[slot_row]);
+
+ edac_dbg(2, "MTR%d where=0x%x B0 value=0x%x\n",
+ slot_row, where, pvt->b0_mtr[slot_row]);
+
+ if (pvt->maxch >= CHANNELS_PER_BRANCH) {
+ pci_read_config_word(pvt->branch_1, where,
+ &pvt->b1_mtr[slot_row]);
+ edac_dbg(2, "MTR%d where=0x%x B1 value=0x%x\n",
+ slot_row, where, pvt->b1_mtr[slot_row]);
+ } else {
+ pvt->b1_mtr[slot_row] = 0;
+ }
+ }
+
+ /* Read and dump branch 0's MTRs */
+ edac_dbg(2, "Memory Technology Registers:\n");
+ edac_dbg(2, " Branch 0:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
+ decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
+ }
+ pci_read_config_word(pvt->branch_0, AMB_PRESENT_0,
+ &pvt->b0_ambpresent0);
+ edac_dbg(2, "\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
+ pci_read_config_word(pvt->branch_0, AMB_PRESENT_1,
+ &pvt->b0_ambpresent1);
+ edac_dbg(2, "\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
+
+ /* Only if we have 2 branchs (4 channels) */
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_ambpresent0 = 0;
+ pvt->b1_ambpresent1 = 0;
+ } else {
+ /* Read and dump branch 1's MTRs */
+ edac_dbg(2, " Branch 1:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS; slot_row++) {
+ decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
+ }
+ pci_read_config_word(pvt->branch_1, AMB_PRESENT_0,
+ &pvt->b1_ambpresent0);
+ edac_dbg(2, "\t\tAMB-Branch 1-present0 0x%x:\n",
+ pvt->b1_ambpresent0);
+ pci_read_config_word(pvt->branch_1, AMB_PRESENT_1,
+ &pvt->b1_ambpresent1);
+ edac_dbg(2, "\t\tAMB-Branch 1-present1 0x%x:\n",
+ pvt->b1_ambpresent1);
+ }
+
+ /* Go and determine the size of each DIMM and place in an
+ * orderly matrix */
+ calculate_dimm_size(pvt);
+}
+
+/*
+ * i5000_init_csrows Initialize the 'csrows' table within
+ * the mci control structure with the
+ * addressing of memory.
+ *
+ * return:
+ * 0 success
+ * 1 no actual memory found on this MC
+ */
+static int i5000_init_csrows(struct mem_ctl_info *mci)
+{
+ struct i5000_pvt *pvt;
+ struct dimm_info *dimm;
+ int empty;
+ int max_csrows;
+ int mtr;
+ int csrow_megs;
+ int channel;
+ int slot;
+
+ pvt = mci->pvt_info;
+ max_csrows = pvt->maxdimmperch * 2;
+
+ empty = 1; /* Assume NO memory */
+
+ /*
+ * FIXME: The memory layout used to map slot/channel into the
+ * real memory architecture is weird: branch+slot are "csrows"
+ * and channel is channel. That required an extra array (dimm_info)
+ * to map the dimms. A good cleanup would be to remove this array,
+ * and do a loop here with branch, channel, slot
+ */
+ for (slot = 0; slot < max_csrows; slot++) {
+ for (channel = 0; channel < pvt->maxch; channel++) {
+
+ mtr = determine_mtr(pvt, slot, channel);
+
+ if (!MTR_DIMMS_PRESENT(mtr))
+ continue;
+
+ dimm = edac_get_dimm(mci, channel / MAX_BRANCHES,
+ channel % MAX_BRANCHES, slot);
+
+ csrow_megs = pvt->dimm_info[slot][channel].megabytes;
+ dimm->grain = 8;
+
+ /* Assume DDR2 for now */
+ dimm->mtype = MEM_FB_DDR2;
+
+ /* ask what device type on this row */
+ if (MTR_DRAM_WIDTH(mtr) == 8)
+ dimm->dtype = DEV_X8;
+ else
+ dimm->dtype = DEV_X4;
+
+ dimm->edac_mode = EDAC_S8ECD8ED;
+ dimm->nr_pages = csrow_megs << 8;
+ }
+
+ empty = 0;
+ }
+
+ return empty;
+}
+
+/*
+ * i5000_enable_error_reporting
+ * Turn on the memory reporting features of the hardware
+ */
+static void i5000_enable_error_reporting(struct mem_ctl_info *mci)
+{
+ struct i5000_pvt *pvt;
+ u32 fbd_error_mask;
+
+ pvt = mci->pvt_info;
+
+ /* Read the FBD Error Mask Register */
+ pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ &fbd_error_mask);
+
+ /* Enable with a '0' */
+ fbd_error_mask &= ~(ENABLE_EMASK_ALL);
+
+ pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ fbd_error_mask);
+}
+
+/*
+ * i5000_get_dimm_and_channel_counts(pdev, &nr_csrows, &num_channels)
+ *
+ * ask the device how many channels are present and how many CSROWS
+ * as well
+ */
+static void i5000_get_dimm_and_channel_counts(struct pci_dev *pdev,
+ int *num_dimms_per_channel,
+ int *num_channels)
+{
+ u8 value;
+
+ /* Need to retrieve just how many channels and dimms per channel are
+ * supported on this memory controller
+ */
+ pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
+ *num_dimms_per_channel = (int)value;
+
+ pci_read_config_byte(pdev, MAXCH, &value);
+ *num_channels = (int)value;
+}
+
+/*
+ * i5000_probe1 Probe for ONE instance of device to see if it is
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[3];
+ struct i5000_pvt *pvt;
+ int num_channels;
+ int num_dimms_per_channel;
+
+ edac_dbg(0, "MC: pdev bus %u dev=0x%x fn=0x%x\n",
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+
+ /* We only are looking for func 0 of the set */
+ if (PCI_FUNC(pdev->devfn) != 0)
+ return -ENODEV;
+
+ /* Ask the devices for the number of CSROWS and CHANNELS so
+ * that we can calculate the memory resources, etc
+ *
+ * The Chipset will report what it can handle which will be greater
+ * or equal to what the motherboard manufacturer will implement.
+ *
+ * As we don't have a motherboard identification routine to determine
+ * actual number of slots/dimms per channel, we thus utilize the
+ * resource as specified by the chipset. Thus, we might have
+ * have more DIMMs per channel than actually on the mobo, but this
+ * allows the driver to support up to the chipset max, without
+ * some fancy mobo determination.
+ */
+ i5000_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
+ &num_channels);
+
+ edac_dbg(0, "MC: Number of Branches=2 Channels= %d DIMMS= %d\n",
+ num_channels, num_dimms_per_channel);
+
+ /* allocate a new MC control structure */
+
+ layers[0].type = EDAC_MC_LAYER_BRANCH;
+ layers[0].size = MAX_BRANCHES;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = num_channels / MAX_BRANCHES;
+ layers[1].is_virt_csrow = false;
+ layers[2].type = EDAC_MC_LAYER_SLOT;
+ layers[2].size = num_dimms_per_channel;
+ layers[2].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p\n", mci);
+
+ mci->pdev = &pdev->dev; /* record ptr to the generic device */
+
+ pvt = mci->pvt_info;
+ pvt->system_address = pdev; /* Record this device in our private */
+ pvt->maxch = num_channels;
+ pvt->maxdimmperch = num_dimms_per_channel;
+
+ /* 'get' the pci devices we want to reserve for our use */
+ if (i5000_get_devices(mci, dev_idx))
+ goto fail0;
+
+ /* Time to get serious */
+ i5000_get_mc_regs(mci); /* retrieve the hardware registers */
+
+ mci->mc_idx = 0;
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i5000_edac.c";
+ mci->ctl_name = i5000_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Set the function pointer to an actual operation function */
+ mci->edac_check = i5000_check_error;
+
+ /* initialize the MC control structure 'csrows' table
+ * with the mapping and control information */
+ if (i5000_init_csrows(mci)) {
+ edac_dbg(0, "MC: Setting mci->edac_cap to EDAC_FLAG_NONE because i5000_init_csrows() returned nonzero value\n");
+ mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
+ } else {
+ edac_dbg(1, "MC: Enable error reporting now\n");
+ i5000_enable_error_reporting(mci);
+ }
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+ goto fail1;
+ }
+
+ i5000_clear_error(mci);
+
+ /* allocating generic PCI control info */
+ i5000_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i5000_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ return 0;
+
+ /* Error exit unwinding stack */
+fail1:
+
+ i5000_put_devices(mci);
+
+fail0:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/*
+ * i5000_init_one constructor for one instance of device
+ *
+ * returns:
+ * negative on error
+ * count (>= 0)
+ */
+static int i5000_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ /* wake up device */
+ rc = pci_enable_device(pdev);
+ if (rc)
+ return rc;
+
+ /* now probe and enable the device */
+ return i5000_probe1(pdev, id->driver_data);
+}
+
+/*
+ * i5000_remove_one destructor for one instance of device
+ *
+ */
+static void i5000_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (i5000_pci)
+ edac_pci_release_generic_ctl(i5000_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ /* retrieve references to resources, and free those resources */
+ i5000_put_devices(mci);
+ edac_mc_free(mci);
+}
+
+/*
+ * pci_device_id table for which devices we are looking for
+ *
+ * The "E500P" device is the first device supported.
+ */
+static const struct pci_device_id i5000_pci_tbl[] = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I5000_DEV16),
+ .driver_data = I5000P},
+
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i5000_pci_tbl);
+
+/*
+ * i5000_driver pci_driver structure for this module
+ *
+ */
+static struct pci_driver i5000_driver = {
+ .name = KBUILD_BASENAME,
+ .probe = i5000_init_one,
+ .remove = i5000_remove_one,
+ .id_table = i5000_pci_tbl,
+};
+
+/*
+ * i5000_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init i5000_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(2, "MC:\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i5000_driver);
+
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+/*
+ * i5000_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit i5000_exit(void)
+{
+ edac_dbg(2, "MC:\n");
+ pci_unregister_driver(&i5000_driver);
+}
+
+module_init(i5000_init);
+module_exit(i5000_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Linux Networx (http://lnxi.com) Doug Thompson <norsk5@xmission.com>");
+MODULE_DESCRIPTION("MC Driver for Intel I5000 memory controllers - " I5000_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+module_param(misc_messages, int, 0444);
+MODULE_PARM_DESC(misc_messages, "Log miscellaneous non fatal messages");
diff --git a/drivers/edac/i5100_edac.c b/drivers/edac/i5100_edac.c
new file mode 100644
index 0000000000..d470afe650
--- /dev/null
+++ b/drivers/edac/i5100_edac.c
@@ -0,0 +1,1224 @@
+/*
+ * Intel 5100 Memory Controllers kernel module
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * This module is based on the following document:
+ *
+ * Intel 5100X Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://download.intel.com/design/chipsets/datashts/318378.pdf
+ *
+ * The intel 5100 has two independent channels. EDAC core currently
+ * can not reflect this configuration so instead the chip-select
+ * rows for each respective channel are laid out one after another,
+ * the first half belonging to channel 0, the second half belonging
+ * to channel 1.
+ *
+ * This driver is for DDR2 DIMMs, and it uses chip select to select among the
+ * several ranks. However, instead of showing memories as ranks, it outputs
+ * them as DIMM's. An internal table creates the association between ranks
+ * and DIMM's.
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include <linux/delay.h>
+#include <linux/mmzone.h>
+#include <linux/debugfs.h>
+
+#include "edac_module.h"
+
+/* register addresses */
+
+/* device 16, func 1 */
+#define I5100_MC 0x40 /* Memory Control Register */
+#define I5100_MC_SCRBEN_MASK (1 << 7)
+#define I5100_MC_SCRBDONE_MASK (1 << 4)
+#define I5100_MS 0x44 /* Memory Status Register */
+#define I5100_SPDDATA 0x48 /* Serial Presence Detect Status Reg */
+#define I5100_SPDCMD 0x4c /* Serial Presence Detect Command Reg */
+#define I5100_TOLM 0x6c /* Top of Low Memory */
+#define I5100_MIR0 0x80 /* Memory Interleave Range 0 */
+#define I5100_MIR1 0x84 /* Memory Interleave Range 1 */
+#define I5100_AMIR_0 0x8c /* Adjusted Memory Interleave Range 0 */
+#define I5100_AMIR_1 0x90 /* Adjusted Memory Interleave Range 1 */
+#define I5100_FERR_NF_MEM 0xa0 /* MC First Non Fatal Errors */
+#define I5100_FERR_NF_MEM_M16ERR_MASK (1 << 16)
+#define I5100_FERR_NF_MEM_M15ERR_MASK (1 << 15)
+#define I5100_FERR_NF_MEM_M14ERR_MASK (1 << 14)
+#define I5100_FERR_NF_MEM_M12ERR_MASK (1 << 12)
+#define I5100_FERR_NF_MEM_M11ERR_MASK (1 << 11)
+#define I5100_FERR_NF_MEM_M10ERR_MASK (1 << 10)
+#define I5100_FERR_NF_MEM_M6ERR_MASK (1 << 6)
+#define I5100_FERR_NF_MEM_M5ERR_MASK (1 << 5)
+#define I5100_FERR_NF_MEM_M4ERR_MASK (1 << 4)
+#define I5100_FERR_NF_MEM_M1ERR_MASK (1 << 1)
+#define I5100_FERR_NF_MEM_ANY_MASK \
+ (I5100_FERR_NF_MEM_M16ERR_MASK | \
+ I5100_FERR_NF_MEM_M15ERR_MASK | \
+ I5100_FERR_NF_MEM_M14ERR_MASK | \
+ I5100_FERR_NF_MEM_M12ERR_MASK | \
+ I5100_FERR_NF_MEM_M11ERR_MASK | \
+ I5100_FERR_NF_MEM_M10ERR_MASK | \
+ I5100_FERR_NF_MEM_M6ERR_MASK | \
+ I5100_FERR_NF_MEM_M5ERR_MASK | \
+ I5100_FERR_NF_MEM_M4ERR_MASK | \
+ I5100_FERR_NF_MEM_M1ERR_MASK)
+#define I5100_NERR_NF_MEM 0xa4 /* MC Next Non-Fatal Errors */
+#define I5100_EMASK_MEM 0xa8 /* MC Error Mask Register */
+#define I5100_MEM0EINJMSK0 0x200 /* Injection Mask0 Register Channel 0 */
+#define I5100_MEM1EINJMSK0 0x208 /* Injection Mask0 Register Channel 1 */
+#define I5100_MEMXEINJMSK0_EINJEN (1 << 27)
+#define I5100_MEM0EINJMSK1 0x204 /* Injection Mask1 Register Channel 0 */
+#define I5100_MEM1EINJMSK1 0x206 /* Injection Mask1 Register Channel 1 */
+
+/* Device 19, Function 0 */
+#define I5100_DINJ0 0x9a
+
+/* device 21 and 22, func 0 */
+#define I5100_MTR_0 0x154 /* Memory Technology Registers 0-3 */
+#define I5100_DMIR 0x15c /* DIMM Interleave Range */
+#define I5100_VALIDLOG 0x18c /* Valid Log Markers */
+#define I5100_NRECMEMA 0x190 /* Non-Recoverable Memory Error Log Reg A */
+#define I5100_NRECMEMB 0x194 /* Non-Recoverable Memory Error Log Reg B */
+#define I5100_REDMEMA 0x198 /* Recoverable Memory Data Error Log Reg A */
+#define I5100_REDMEMB 0x19c /* Recoverable Memory Data Error Log Reg B */
+#define I5100_RECMEMA 0x1a0 /* Recoverable Memory Error Log Reg A */
+#define I5100_RECMEMB 0x1a4 /* Recoverable Memory Error Log Reg B */
+#define I5100_MTR_4 0x1b0 /* Memory Technology Registers 4,5 */
+
+/* bit field accessors */
+
+static inline u32 i5100_mc_scrben(u32 mc)
+{
+ return mc >> 7 & 1;
+}
+
+static inline u32 i5100_mc_errdeten(u32 mc)
+{
+ return mc >> 5 & 1;
+}
+
+static inline u32 i5100_mc_scrbdone(u32 mc)
+{
+ return mc >> 4 & 1;
+}
+
+static inline u16 i5100_spddata_rdo(u16 a)
+{
+ return a >> 15 & 1;
+}
+
+static inline u16 i5100_spddata_sbe(u16 a)
+{
+ return a >> 13 & 1;
+}
+
+static inline u16 i5100_spddata_busy(u16 a)
+{
+ return a >> 12 & 1;
+}
+
+static inline u16 i5100_spddata_data(u16 a)
+{
+ return a & ((1 << 8) - 1);
+}
+
+static inline u32 i5100_spdcmd_create(u32 dti, u32 ckovrd, u32 sa, u32 ba,
+ u32 data, u32 cmd)
+{
+ return ((dti & ((1 << 4) - 1)) << 28) |
+ ((ckovrd & 1) << 27) |
+ ((sa & ((1 << 3) - 1)) << 24) |
+ ((ba & ((1 << 8) - 1)) << 16) |
+ ((data & ((1 << 8) - 1)) << 8) |
+ (cmd & 1);
+}
+
+static inline u16 i5100_tolm_tolm(u16 a)
+{
+ return a >> 12 & ((1 << 4) - 1);
+}
+
+static inline u16 i5100_mir_limit(u16 a)
+{
+ return a >> 4 & ((1 << 12) - 1);
+}
+
+static inline u16 i5100_mir_way1(u16 a)
+{
+ return a >> 1 & 1;
+}
+
+static inline u16 i5100_mir_way0(u16 a)
+{
+ return a & 1;
+}
+
+static inline u32 i5100_ferr_nf_mem_chan_indx(u32 a)
+{
+ return a >> 28 & 1;
+}
+
+static inline u32 i5100_ferr_nf_mem_any(u32 a)
+{
+ return a & I5100_FERR_NF_MEM_ANY_MASK;
+}
+
+static inline u32 i5100_nerr_nf_mem_any(u32 a)
+{
+ return i5100_ferr_nf_mem_any(a);
+}
+
+static inline u32 i5100_dmir_limit(u32 a)
+{
+ return a >> 16 & ((1 << 11) - 1);
+}
+
+static inline u32 i5100_dmir_rank(u32 a, u32 i)
+{
+ return a >> (4 * i) & ((1 << 2) - 1);
+}
+
+static inline u16 i5100_mtr_present(u16 a)
+{
+ return a >> 10 & 1;
+}
+
+static inline u16 i5100_mtr_ethrottle(u16 a)
+{
+ return a >> 9 & 1;
+}
+
+static inline u16 i5100_mtr_width(u16 a)
+{
+ return a >> 8 & 1;
+}
+
+static inline u16 i5100_mtr_numbank(u16 a)
+{
+ return a >> 6 & 1;
+}
+
+static inline u16 i5100_mtr_numrow(u16 a)
+{
+ return a >> 2 & ((1 << 2) - 1);
+}
+
+static inline u16 i5100_mtr_numcol(u16 a)
+{
+ return a & ((1 << 2) - 1);
+}
+
+
+static inline u32 i5100_validlog_redmemvalid(u32 a)
+{
+ return a >> 2 & 1;
+}
+
+static inline u32 i5100_validlog_recmemvalid(u32 a)
+{
+ return a >> 1 & 1;
+}
+
+static inline u32 i5100_validlog_nrecmemvalid(u32 a)
+{
+ return a & 1;
+}
+
+static inline u32 i5100_nrecmema_merr(u32 a)
+{
+ return a >> 15 & ((1 << 5) - 1);
+}
+
+static inline u32 i5100_nrecmema_bank(u32 a)
+{
+ return a >> 12 & ((1 << 3) - 1);
+}
+
+static inline u32 i5100_nrecmema_rank(u32 a)
+{
+ return a >> 8 & ((1 << 3) - 1);
+}
+
+static inline u32 i5100_nrecmemb_cas(u32 a)
+{
+ return a >> 16 & ((1 << 13) - 1);
+}
+
+static inline u32 i5100_nrecmemb_ras(u32 a)
+{
+ return a & ((1 << 16) - 1);
+}
+
+static inline u32 i5100_recmema_merr(u32 a)
+{
+ return i5100_nrecmema_merr(a);
+}
+
+static inline u32 i5100_recmema_bank(u32 a)
+{
+ return i5100_nrecmema_bank(a);
+}
+
+static inline u32 i5100_recmema_rank(u32 a)
+{
+ return i5100_nrecmema_rank(a);
+}
+
+static inline u32 i5100_recmemb_cas(u32 a)
+{
+ return i5100_nrecmemb_cas(a);
+}
+
+static inline u32 i5100_recmemb_ras(u32 a)
+{
+ return i5100_nrecmemb_ras(a);
+}
+
+/* some generic limits */
+#define I5100_MAX_RANKS_PER_CHAN 6
+#define I5100_CHANNELS 2
+#define I5100_MAX_RANKS_PER_DIMM 4
+#define I5100_DIMM_ADDR_LINES (6 - 3) /* 64 bits / 8 bits per byte */
+#define I5100_MAX_DIMM_SLOTS_PER_CHAN 4
+#define I5100_MAX_RANK_INTERLEAVE 4
+#define I5100_MAX_DMIRS 5
+#define I5100_SCRUB_REFRESH_RATE (5 * 60 * HZ)
+
+struct i5100_priv {
+ /* ranks on each dimm -- 0 maps to not present -- obtained via SPD */
+ int dimm_numrank[I5100_CHANNELS][I5100_MAX_DIMM_SLOTS_PER_CHAN];
+
+ /*
+ * mainboard chip select map -- maps i5100 chip selects to
+ * DIMM slot chip selects. In the case of only 4 ranks per
+ * channel, the mapping is fairly obvious but not unique.
+ * we map -1 -> NC and assume both channels use the same
+ * map...
+ *
+ */
+ int dimm_csmap[I5100_MAX_DIMM_SLOTS_PER_CHAN][I5100_MAX_RANKS_PER_DIMM];
+
+ /* memory interleave range */
+ struct {
+ u64 limit;
+ unsigned way[2];
+ } mir[I5100_CHANNELS];
+
+ /* adjusted memory interleave range register */
+ unsigned amir[I5100_CHANNELS];
+
+ /* dimm interleave range */
+ struct {
+ unsigned rank[I5100_MAX_RANK_INTERLEAVE];
+ u64 limit;
+ } dmir[I5100_CHANNELS][I5100_MAX_DMIRS];
+
+ /* memory technology registers... */
+ struct {
+ unsigned present; /* 0 or 1 */
+ unsigned ethrottle; /* 0 or 1 */
+ unsigned width; /* 4 or 8 bits */
+ unsigned numbank; /* 2 or 3 lines */
+ unsigned numrow; /* 13 .. 16 lines */
+ unsigned numcol; /* 11 .. 12 lines */
+ } mtr[I5100_CHANNELS][I5100_MAX_RANKS_PER_CHAN];
+
+ u64 tolm; /* top of low memory in bytes */
+ unsigned ranksperchan; /* number of ranks per channel */
+
+ struct pci_dev *mc; /* device 16 func 1 */
+ struct pci_dev *einj; /* device 19 func 0 */
+ struct pci_dev *ch0mm; /* device 21 func 0 */
+ struct pci_dev *ch1mm; /* device 22 func 0 */
+
+ struct delayed_work i5100_scrubbing;
+ int scrub_enable;
+
+ /* Error injection */
+ u8 inject_channel;
+ u8 inject_hlinesel;
+ u8 inject_deviceptr1;
+ u8 inject_deviceptr2;
+ u16 inject_eccmask1;
+ u16 inject_eccmask2;
+
+ struct dentry *debugfs;
+};
+
+static struct dentry *i5100_debugfs;
+
+/* map a rank/chan to a slot number on the mainboard */
+static int i5100_rank_to_slot(const struct mem_ctl_info *mci,
+ int chan, int rank)
+{
+ const struct i5100_priv *priv = mci->pvt_info;
+ int i;
+
+ for (i = 0; i < I5100_MAX_DIMM_SLOTS_PER_CHAN; i++) {
+ int j;
+ const int numrank = priv->dimm_numrank[chan][i];
+
+ for (j = 0; j < numrank; j++)
+ if (priv->dimm_csmap[i][j] == rank)
+ return i * 2 + chan;
+ }
+
+ return -1;
+}
+
+static const char *i5100_err_msg(unsigned err)
+{
+ static const char *merrs[] = {
+ "unknown", /* 0 */
+ "uncorrectable data ECC on replay", /* 1 */
+ "unknown", /* 2 */
+ "unknown", /* 3 */
+ "aliased uncorrectable demand data ECC", /* 4 */
+ "aliased uncorrectable spare-copy data ECC", /* 5 */
+ "aliased uncorrectable patrol data ECC", /* 6 */
+ "unknown", /* 7 */
+ "unknown", /* 8 */
+ "unknown", /* 9 */
+ "non-aliased uncorrectable demand data ECC", /* 10 */
+ "non-aliased uncorrectable spare-copy data ECC", /* 11 */
+ "non-aliased uncorrectable patrol data ECC", /* 12 */
+ "unknown", /* 13 */
+ "correctable demand data ECC", /* 14 */
+ "correctable spare-copy data ECC", /* 15 */
+ "correctable patrol data ECC", /* 16 */
+ "unknown", /* 17 */
+ "SPD protocol error", /* 18 */
+ "unknown", /* 19 */
+ "spare copy initiated", /* 20 */
+ "spare copy completed", /* 21 */
+ };
+ unsigned i;
+
+ for (i = 0; i < ARRAY_SIZE(merrs); i++)
+ if (1 << i & err)
+ return merrs[i];
+
+ return "none";
+}
+
+/* convert csrow index into a rank (per channel -- 0..5) */
+static unsigned int i5100_csrow_to_rank(const struct mem_ctl_info *mci,
+ unsigned int csrow)
+{
+ const struct i5100_priv *priv = mci->pvt_info;
+
+ return csrow % priv->ranksperchan;
+}
+
+/* convert csrow index into a channel (0..1) */
+static unsigned int i5100_csrow_to_chan(const struct mem_ctl_info *mci,
+ unsigned int csrow)
+{
+ const struct i5100_priv *priv = mci->pvt_info;
+
+ return csrow / priv->ranksperchan;
+}
+
+static void i5100_handle_ce(struct mem_ctl_info *mci,
+ int chan,
+ unsigned bank,
+ unsigned rank,
+ unsigned long syndrome,
+ unsigned cas,
+ unsigned ras,
+ const char *msg)
+{
+ char detail[80];
+
+ /* Form out message */
+ snprintf(detail, sizeof(detail),
+ "bank %u, cas %u, ras %u\n",
+ bank, cas, ras);
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ 0, 0, syndrome,
+ chan, rank, -1,
+ msg, detail);
+}
+
+static void i5100_handle_ue(struct mem_ctl_info *mci,
+ int chan,
+ unsigned bank,
+ unsigned rank,
+ unsigned long syndrome,
+ unsigned cas,
+ unsigned ras,
+ const char *msg)
+{
+ char detail[80];
+
+ /* Form out message */
+ snprintf(detail, sizeof(detail),
+ "bank %u, cas %u, ras %u\n",
+ bank, cas, ras);
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ 0, 0, syndrome,
+ chan, rank, -1,
+ msg, detail);
+}
+
+static void i5100_read_log(struct mem_ctl_info *mci, int chan,
+ u32 ferr, u32 nerr)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ struct pci_dev *pdev = (chan) ? priv->ch1mm : priv->ch0mm;
+ u32 dw;
+ u32 dw2;
+ unsigned syndrome = 0;
+ unsigned merr;
+ unsigned bank;
+ unsigned rank;
+ unsigned cas;
+ unsigned ras;
+
+ pci_read_config_dword(pdev, I5100_VALIDLOG, &dw);
+
+ if (i5100_validlog_redmemvalid(dw)) {
+ pci_read_config_dword(pdev, I5100_REDMEMA, &dw2);
+ syndrome = dw2;
+ pci_read_config_dword(pdev, I5100_REDMEMB, &dw2);
+ }
+
+ if (i5100_validlog_recmemvalid(dw)) {
+ const char *msg;
+
+ pci_read_config_dword(pdev, I5100_RECMEMA, &dw2);
+ merr = i5100_recmema_merr(dw2);
+ bank = i5100_recmema_bank(dw2);
+ rank = i5100_recmema_rank(dw2);
+
+ pci_read_config_dword(pdev, I5100_RECMEMB, &dw2);
+ cas = i5100_recmemb_cas(dw2);
+ ras = i5100_recmemb_ras(dw2);
+
+ /* FIXME: not really sure if this is what merr is...
+ */
+ if (!merr)
+ msg = i5100_err_msg(ferr);
+ else
+ msg = i5100_err_msg(nerr);
+
+ i5100_handle_ce(mci, chan, bank, rank, syndrome, cas, ras, msg);
+ }
+
+ if (i5100_validlog_nrecmemvalid(dw)) {
+ const char *msg;
+
+ pci_read_config_dword(pdev, I5100_NRECMEMA, &dw2);
+ merr = i5100_nrecmema_merr(dw2);
+ bank = i5100_nrecmema_bank(dw2);
+ rank = i5100_nrecmema_rank(dw2);
+
+ pci_read_config_dword(pdev, I5100_NRECMEMB, &dw2);
+ cas = i5100_nrecmemb_cas(dw2);
+ ras = i5100_nrecmemb_ras(dw2);
+
+ /* FIXME: not really sure if this is what merr is...
+ */
+ if (!merr)
+ msg = i5100_err_msg(ferr);
+ else
+ msg = i5100_err_msg(nerr);
+
+ i5100_handle_ue(mci, chan, bank, rank, syndrome, cas, ras, msg);
+ }
+
+ pci_write_config_dword(pdev, I5100_VALIDLOG, dw);
+}
+
+static void i5100_check_error(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ u32 dw, dw2;
+
+ pci_read_config_dword(priv->mc, I5100_FERR_NF_MEM, &dw);
+ if (i5100_ferr_nf_mem_any(dw)) {
+
+ pci_read_config_dword(priv->mc, I5100_NERR_NF_MEM, &dw2);
+
+ i5100_read_log(mci, i5100_ferr_nf_mem_chan_indx(dw),
+ i5100_ferr_nf_mem_any(dw),
+ i5100_nerr_nf_mem_any(dw2));
+
+ pci_write_config_dword(priv->mc, I5100_NERR_NF_MEM, dw2);
+ }
+ pci_write_config_dword(priv->mc, I5100_FERR_NF_MEM, dw);
+}
+
+/* The i5100 chipset will scrub the entire memory once, then
+ * set a done bit. Continuous scrubbing is achieved by enqueing
+ * delayed work to a workqueue, checking every few minutes if
+ * the scrubbing has completed and if so reinitiating it.
+ */
+
+static void i5100_refresh_scrubbing(struct work_struct *work)
+{
+ struct delayed_work *i5100_scrubbing = to_delayed_work(work);
+ struct i5100_priv *priv = container_of(i5100_scrubbing,
+ struct i5100_priv,
+ i5100_scrubbing);
+ u32 dw;
+
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+
+ if (priv->scrub_enable) {
+
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+
+ if (i5100_mc_scrbdone(dw)) {
+ dw |= I5100_MC_SCRBEN_MASK;
+ pci_write_config_dword(priv->mc, I5100_MC, dw);
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+ }
+
+ schedule_delayed_work(&(priv->i5100_scrubbing),
+ I5100_SCRUB_REFRESH_RATE);
+ }
+}
+/*
+ * The bandwidth is based on experimentation, feel free to refine it.
+ */
+static int i5100_set_scrub_rate(struct mem_ctl_info *mci, u32 bandwidth)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ u32 dw;
+
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+ if (bandwidth) {
+ priv->scrub_enable = 1;
+ dw |= I5100_MC_SCRBEN_MASK;
+ schedule_delayed_work(&(priv->i5100_scrubbing),
+ I5100_SCRUB_REFRESH_RATE);
+ } else {
+ priv->scrub_enable = 0;
+ dw &= ~I5100_MC_SCRBEN_MASK;
+ cancel_delayed_work(&(priv->i5100_scrubbing));
+ }
+ pci_write_config_dword(priv->mc, I5100_MC, dw);
+
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+
+ bandwidth = 5900000 * i5100_mc_scrben(dw);
+
+ return bandwidth;
+}
+
+static int i5100_get_scrub_rate(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ u32 dw;
+
+ pci_read_config_dword(priv->mc, I5100_MC, &dw);
+
+ return 5900000 * i5100_mc_scrben(dw);
+}
+
+static struct pci_dev *pci_get_device_func(unsigned vendor,
+ unsigned device,
+ unsigned func)
+{
+ struct pci_dev *ret = NULL;
+
+ while (1) {
+ ret = pci_get_device(vendor, device, ret);
+
+ if (!ret)
+ break;
+
+ if (PCI_FUNC(ret->devfn) == func)
+ break;
+ }
+
+ return ret;
+}
+
+static unsigned long i5100_npages(struct mem_ctl_info *mci, unsigned int csrow)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ const unsigned int chan_rank = i5100_csrow_to_rank(mci, csrow);
+ const unsigned int chan = i5100_csrow_to_chan(mci, csrow);
+ unsigned addr_lines;
+
+ /* dimm present? */
+ if (!priv->mtr[chan][chan_rank].present)
+ return 0ULL;
+
+ addr_lines =
+ I5100_DIMM_ADDR_LINES +
+ priv->mtr[chan][chan_rank].numcol +
+ priv->mtr[chan][chan_rank].numrow +
+ priv->mtr[chan][chan_rank].numbank;
+
+ return (unsigned long)
+ ((unsigned long long) (1ULL << addr_lines) / PAGE_SIZE);
+}
+
+static void i5100_init_mtr(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ struct pci_dev *mms[2] = { priv->ch0mm, priv->ch1mm };
+ int i;
+
+ for (i = 0; i < I5100_CHANNELS; i++) {
+ int j;
+ struct pci_dev *pdev = mms[i];
+
+ for (j = 0; j < I5100_MAX_RANKS_PER_CHAN; j++) {
+ const unsigned addr =
+ (j < 4) ? I5100_MTR_0 + j * 2 :
+ I5100_MTR_4 + (j - 4) * 2;
+ u16 w;
+
+ pci_read_config_word(pdev, addr, &w);
+
+ priv->mtr[i][j].present = i5100_mtr_present(w);
+ priv->mtr[i][j].ethrottle = i5100_mtr_ethrottle(w);
+ priv->mtr[i][j].width = 4 + 4 * i5100_mtr_width(w);
+ priv->mtr[i][j].numbank = 2 + i5100_mtr_numbank(w);
+ priv->mtr[i][j].numrow = 13 + i5100_mtr_numrow(w);
+ priv->mtr[i][j].numcol = 10 + i5100_mtr_numcol(w);
+ }
+ }
+}
+
+/*
+ * FIXME: make this into a real i2c adapter (so that dimm-decode
+ * will work)?
+ */
+static int i5100_read_spd_byte(const struct mem_ctl_info *mci,
+ u8 ch, u8 slot, u8 addr, u8 *byte)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ u16 w;
+
+ pci_read_config_word(priv->mc, I5100_SPDDATA, &w);
+ if (i5100_spddata_busy(w))
+ return -1;
+
+ pci_write_config_dword(priv->mc, I5100_SPDCMD,
+ i5100_spdcmd_create(0xa, 1, ch * 4 + slot, addr,
+ 0, 0));
+
+ /* wait up to 100ms */
+ udelay(100);
+ while (1) {
+ pci_read_config_word(priv->mc, I5100_SPDDATA, &w);
+ if (!i5100_spddata_busy(w))
+ break;
+ udelay(100);
+ }
+
+ if (!i5100_spddata_rdo(w) || i5100_spddata_sbe(w))
+ return -1;
+
+ *byte = i5100_spddata_data(w);
+
+ return 0;
+}
+
+/*
+ * fill dimm chip select map
+ *
+ * FIXME:
+ * o not the only way to may chip selects to dimm slots
+ * o investigate if there is some way to obtain this map from the bios
+ */
+static void i5100_init_dimm_csmap(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ int i;
+
+ for (i = 0; i < I5100_MAX_DIMM_SLOTS_PER_CHAN; i++) {
+ int j;
+
+ for (j = 0; j < I5100_MAX_RANKS_PER_DIMM; j++)
+ priv->dimm_csmap[i][j] = -1; /* default NC */
+ }
+
+ /* only 2 chip selects per slot... */
+ if (priv->ranksperchan == 4) {
+ priv->dimm_csmap[0][0] = 0;
+ priv->dimm_csmap[0][1] = 3;
+ priv->dimm_csmap[1][0] = 1;
+ priv->dimm_csmap[1][1] = 2;
+ priv->dimm_csmap[2][0] = 2;
+ priv->dimm_csmap[3][0] = 3;
+ } else {
+ priv->dimm_csmap[0][0] = 0;
+ priv->dimm_csmap[0][1] = 1;
+ priv->dimm_csmap[1][0] = 2;
+ priv->dimm_csmap[1][1] = 3;
+ priv->dimm_csmap[2][0] = 4;
+ priv->dimm_csmap[2][1] = 5;
+ }
+}
+
+static void i5100_init_dimm_layout(struct pci_dev *pdev,
+ struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ int i;
+
+ for (i = 0; i < I5100_CHANNELS; i++) {
+ int j;
+
+ for (j = 0; j < I5100_MAX_DIMM_SLOTS_PER_CHAN; j++) {
+ u8 rank;
+
+ if (i5100_read_spd_byte(mci, i, j, 5, &rank) < 0)
+ priv->dimm_numrank[i][j] = 0;
+ else
+ priv->dimm_numrank[i][j] = (rank & 3) + 1;
+ }
+ }
+
+ i5100_init_dimm_csmap(mci);
+}
+
+static void i5100_init_interleaving(struct pci_dev *pdev,
+ struct mem_ctl_info *mci)
+{
+ u16 w;
+ u32 dw;
+ struct i5100_priv *priv = mci->pvt_info;
+ struct pci_dev *mms[2] = { priv->ch0mm, priv->ch1mm };
+ int i;
+
+ pci_read_config_word(pdev, I5100_TOLM, &w);
+ priv->tolm = (u64) i5100_tolm_tolm(w) * 256 * 1024 * 1024;
+
+ pci_read_config_word(pdev, I5100_MIR0, &w);
+ priv->mir[0].limit = (u64) i5100_mir_limit(w) << 28;
+ priv->mir[0].way[1] = i5100_mir_way1(w);
+ priv->mir[0].way[0] = i5100_mir_way0(w);
+
+ pci_read_config_word(pdev, I5100_MIR1, &w);
+ priv->mir[1].limit = (u64) i5100_mir_limit(w) << 28;
+ priv->mir[1].way[1] = i5100_mir_way1(w);
+ priv->mir[1].way[0] = i5100_mir_way0(w);
+
+ pci_read_config_word(pdev, I5100_AMIR_0, &w);
+ priv->amir[0] = w;
+ pci_read_config_word(pdev, I5100_AMIR_1, &w);
+ priv->amir[1] = w;
+
+ for (i = 0; i < I5100_CHANNELS; i++) {
+ int j;
+
+ for (j = 0; j < 5; j++) {
+ int k;
+
+ pci_read_config_dword(mms[i], I5100_DMIR + j * 4, &dw);
+
+ priv->dmir[i][j].limit =
+ (u64) i5100_dmir_limit(dw) << 28;
+ for (k = 0; k < I5100_MAX_RANKS_PER_DIMM; k++)
+ priv->dmir[i][j].rank[k] =
+ i5100_dmir_rank(dw, k);
+ }
+ }
+
+ i5100_init_mtr(mci);
+}
+
+static void i5100_init_csrows(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ struct dimm_info *dimm;
+
+ mci_for_each_dimm(mci, dimm) {
+ const unsigned long npages = i5100_npages(mci, dimm->idx);
+ const unsigned int chan = i5100_csrow_to_chan(mci, dimm->idx);
+ const unsigned int rank = i5100_csrow_to_rank(mci, dimm->idx);
+
+ if (!npages)
+ continue;
+
+ dimm->nr_pages = npages;
+ dimm->grain = 32;
+ dimm->dtype = (priv->mtr[chan][rank].width == 4) ?
+ DEV_X4 : DEV_X8;
+ dimm->mtype = MEM_RDDR2;
+ dimm->edac_mode = EDAC_SECDED;
+ snprintf(dimm->label, sizeof(dimm->label), "DIMM%u",
+ i5100_rank_to_slot(mci, chan, rank));
+
+ edac_dbg(2, "dimm channel %d, rank %d, size %ld\n",
+ chan, rank, (long)PAGES_TO_MiB(npages));
+ }
+}
+
+/****************************************************************************
+ * Error injection routines
+ ****************************************************************************/
+
+static void i5100_do_inject(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+ u32 mask0;
+ u16 mask1;
+
+ /* MEM[1:0]EINJMSK0
+ * 31 - ADDRMATCHEN
+ * 29:28 - HLINESEL
+ * 00 Reserved
+ * 01 Lower half of cache line
+ * 10 Upper half of cache line
+ * 11 Both upper and lower parts of cache line
+ * 27 - EINJEN
+ * 25:19 - XORMASK1 for deviceptr1
+ * 9:5 - SEC2RAM for deviceptr2
+ * 4:0 - FIR2RAM for deviceptr1
+ */
+ mask0 = ((priv->inject_hlinesel & 0x3) << 28) |
+ I5100_MEMXEINJMSK0_EINJEN |
+ ((priv->inject_eccmask1 & 0xffff) << 10) |
+ ((priv->inject_deviceptr2 & 0x1f) << 5) |
+ (priv->inject_deviceptr1 & 0x1f);
+
+ /* MEM[1:0]EINJMSK1
+ * 15:0 - XORMASK2 for deviceptr2
+ */
+ mask1 = priv->inject_eccmask2;
+
+ if (priv->inject_channel == 0) {
+ pci_write_config_dword(priv->mc, I5100_MEM0EINJMSK0, mask0);
+ pci_write_config_word(priv->mc, I5100_MEM0EINJMSK1, mask1);
+ } else {
+ pci_write_config_dword(priv->mc, I5100_MEM1EINJMSK0, mask0);
+ pci_write_config_word(priv->mc, I5100_MEM1EINJMSK1, mask1);
+ }
+
+ /* Error Injection Response Function
+ * Intel 5100 Memory Controller Hub Chipset (318378) datasheet
+ * hints about this register but carry no data about them. All
+ * data regarding device 19 is based on experimentation and the
+ * Intel 7300 Chipset Memory Controller Hub (318082) datasheet
+ * which appears to be accurate for the i5100 in this area.
+ *
+ * The injection code don't work without setting this register.
+ * The register needs to be flipped off then on else the hardware
+ * will only perform the first injection.
+ *
+ * Stop condition bits 7:4
+ * 1010 - Stop after one injection
+ * 1011 - Never stop injecting faults
+ *
+ * Start condition bits 3:0
+ * 1010 - Never start
+ * 1011 - Start immediately
+ */
+ pci_write_config_byte(priv->einj, I5100_DINJ0, 0xaa);
+ pci_write_config_byte(priv->einj, I5100_DINJ0, 0xab);
+}
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+static ssize_t inject_enable_write(struct file *file, const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct device *dev = file->private_data;
+ struct mem_ctl_info *mci = to_mci(dev);
+
+ i5100_do_inject(mci);
+
+ return count;
+}
+
+static const struct file_operations i5100_inject_enable_fops = {
+ .open = simple_open,
+ .write = inject_enable_write,
+ .llseek = generic_file_llseek,
+};
+
+static int i5100_setup_debugfs(struct mem_ctl_info *mci)
+{
+ struct i5100_priv *priv = mci->pvt_info;
+
+ if (!i5100_debugfs)
+ return -ENODEV;
+
+ priv->debugfs = edac_debugfs_create_dir_at(mci->bus->name, i5100_debugfs);
+
+ if (!priv->debugfs)
+ return -ENOMEM;
+
+ edac_debugfs_create_x8("inject_channel", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_channel);
+ edac_debugfs_create_x8("inject_hlinesel", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_hlinesel);
+ edac_debugfs_create_x8("inject_deviceptr1", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_deviceptr1);
+ edac_debugfs_create_x8("inject_deviceptr2", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_deviceptr2);
+ edac_debugfs_create_x16("inject_eccmask1", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_eccmask1);
+ edac_debugfs_create_x16("inject_eccmask2", S_IRUGO | S_IWUSR, priv->debugfs,
+ &priv->inject_eccmask2);
+ edac_debugfs_create_file("inject_enable", S_IWUSR, priv->debugfs,
+ &mci->dev, &i5100_inject_enable_fops);
+
+ return 0;
+
+}
+
+static int i5100_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rc;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct i5100_priv *priv;
+ struct pci_dev *ch0mm, *ch1mm, *einj;
+ int ret = 0;
+ u32 dw;
+ int ranksperch;
+
+ if (PCI_FUNC(pdev->devfn) != 1)
+ return -ENODEV;
+
+ rc = pci_enable_device(pdev);
+ if (rc < 0) {
+ ret = rc;
+ goto bail;
+ }
+
+ /* ECC enabled? */
+ pci_read_config_dword(pdev, I5100_MC, &dw);
+ if (!i5100_mc_errdeten(dw)) {
+ printk(KERN_INFO "i5100_edac: ECC not enabled.\n");
+ ret = -ENODEV;
+ goto bail_pdev;
+ }
+
+ /* figure out how many ranks, from strapped state of 48GB_Mode input */
+ pci_read_config_dword(pdev, I5100_MS, &dw);
+ ranksperch = !!(dw & (1 << 8)) * 2 + 4;
+
+ /* enable error reporting... */
+ pci_read_config_dword(pdev, I5100_EMASK_MEM, &dw);
+ dw &= ~I5100_FERR_NF_MEM_ANY_MASK;
+ pci_write_config_dword(pdev, I5100_EMASK_MEM, dw);
+
+ /* device 21, func 0, Channel 0 Memory Map, Error Flag/Mask, etc... */
+ ch0mm = pci_get_device_func(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5100_21, 0);
+ if (!ch0mm) {
+ ret = -ENODEV;
+ goto bail_pdev;
+ }
+
+ rc = pci_enable_device(ch0mm);
+ if (rc < 0) {
+ ret = rc;
+ goto bail_ch0;
+ }
+
+ /* device 22, func 0, Channel 1 Memory Map, Error Flag/Mask, etc... */
+ ch1mm = pci_get_device_func(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5100_22, 0);
+ if (!ch1mm) {
+ ret = -ENODEV;
+ goto bail_disable_ch0;
+ }
+
+ rc = pci_enable_device(ch1mm);
+ if (rc < 0) {
+ ret = rc;
+ goto bail_ch1;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = 2;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = ranksperch;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(*priv));
+ if (!mci) {
+ ret = -ENOMEM;
+ goto bail_disable_ch1;
+ }
+
+
+ /* device 19, func 0, Error injection */
+ einj = pci_get_device_func(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5100_19, 0);
+ if (!einj) {
+ ret = -ENODEV;
+ goto bail_mc_free;
+ }
+
+ rc = pci_enable_device(einj);
+ if (rc < 0) {
+ ret = rc;
+ goto bail_einj;
+ }
+
+ mci->pdev = &pdev->dev;
+
+ priv = mci->pvt_info;
+ priv->ranksperchan = ranksperch;
+ priv->mc = pdev;
+ priv->ch0mm = ch0mm;
+ priv->ch1mm = ch1mm;
+ priv->einj = einj;
+
+ INIT_DELAYED_WORK(&(priv->i5100_scrubbing), i5100_refresh_scrubbing);
+
+ /* If scrubbing was already enabled by the bios, start maintaining it */
+ pci_read_config_dword(pdev, I5100_MC, &dw);
+ if (i5100_mc_scrben(dw)) {
+ priv->scrub_enable = 1;
+ schedule_delayed_work(&(priv->i5100_scrubbing),
+ I5100_SCRUB_REFRESH_RATE);
+ }
+
+ i5100_init_dimm_layout(pdev, mci);
+ i5100_init_interleaving(pdev, mci);
+
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = "i5100_edac.c";
+ mci->ctl_name = "i5100";
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ mci->edac_check = i5100_check_error;
+ mci->set_sdram_scrub_rate = i5100_set_scrub_rate;
+ mci->get_sdram_scrub_rate = i5100_get_scrub_rate;
+
+ priv->inject_channel = 0;
+ priv->inject_hlinesel = 0;
+ priv->inject_deviceptr1 = 0;
+ priv->inject_deviceptr2 = 0;
+ priv->inject_eccmask1 = 0;
+ priv->inject_eccmask2 = 0;
+
+ i5100_init_csrows(mci);
+
+ /* this strange construction seems to be in every driver, dunno why */
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_NMI:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_POLL;
+ break;
+ }
+
+ if (edac_mc_add_mc(mci)) {
+ ret = -ENODEV;
+ goto bail_scrub;
+ }
+
+ i5100_setup_debugfs(mci);
+
+ return ret;
+
+bail_scrub:
+ priv->scrub_enable = 0;
+ cancel_delayed_work_sync(&(priv->i5100_scrubbing));
+ pci_disable_device(einj);
+
+bail_einj:
+ pci_dev_put(einj);
+
+bail_mc_free:
+ edac_mc_free(mci);
+
+bail_disable_ch1:
+ pci_disable_device(ch1mm);
+
+bail_ch1:
+ pci_dev_put(ch1mm);
+
+bail_disable_ch0:
+ pci_disable_device(ch0mm);
+
+bail_ch0:
+ pci_dev_put(ch0mm);
+
+bail_pdev:
+ pci_disable_device(pdev);
+
+bail:
+ return ret;
+}
+
+static void i5100_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct i5100_priv *priv;
+
+ mci = edac_mc_del_mc(&pdev->dev);
+
+ if (!mci)
+ return;
+
+ priv = mci->pvt_info;
+
+ edac_debugfs_remove_recursive(priv->debugfs);
+
+ priv->scrub_enable = 0;
+ cancel_delayed_work_sync(&(priv->i5100_scrubbing));
+
+ pci_disable_device(pdev);
+ pci_disable_device(priv->ch0mm);
+ pci_disable_device(priv->ch1mm);
+ pci_disable_device(priv->einj);
+ pci_dev_put(priv->ch0mm);
+ pci_dev_put(priv->ch1mm);
+ pci_dev_put(priv->einj);
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i5100_pci_tbl[] = {
+ /* Device 16, Function 0, Channel 0 Memory Map, Error Flag/Mask, ... */
+ { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5100_16) },
+ { 0, }
+};
+MODULE_DEVICE_TABLE(pci, i5100_pci_tbl);
+
+static struct pci_driver i5100_driver = {
+ .name = KBUILD_BASENAME,
+ .probe = i5100_init_one,
+ .remove = i5100_remove_one,
+ .id_table = i5100_pci_tbl,
+};
+
+static int __init i5100_init(void)
+{
+ int pci_rc;
+
+ i5100_debugfs = edac_debugfs_create_dir_at("i5100_edac", NULL);
+
+ pci_rc = pci_register_driver(&i5100_driver);
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+static void __exit i5100_exit(void)
+{
+ edac_debugfs_remove(i5100_debugfs);
+
+ pci_unregister_driver(&i5100_driver);
+}
+
+module_init(i5100_init);
+module_exit(i5100_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Arthur Jones <ajones@riverbed.com>");
+MODULE_DESCRIPTION("MC Driver for Intel I5100 memory controllers");
diff --git a/drivers/edac/i5400_edac.c b/drivers/edac/i5400_edac.c
new file mode 100644
index 0000000000..49b4499269
--- /dev/null
+++ b/drivers/edac/i5400_edac.c
@@ -0,0 +1,1469 @@
+/*
+ * Intel 5400 class Memory Controllers kernel module (Seaburg)
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Copyright (c) 2008 by:
+ * Ben Woodard <woodard@redhat.com>
+ * Mauro Carvalho Chehab
+ *
+ * Red Hat Inc. https://www.redhat.com
+ *
+ * Forked and adapted from the i5000_edac driver which was
+ * written by Douglas Thompson Linux Networx <norsk5@xmission.com>
+ *
+ * This module is based on the following document:
+ *
+ * Intel 5400 Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://developer.intel.com/design/chipsets/datashts/313070.htm
+ *
+ * This Memory Controller manages DDR2 FB-DIMMs. It has 2 branches, each with
+ * 2 channels operating in lockstep no-mirror mode. Each channel can have up to
+ * 4 dimm's, each with up to 8GB.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+
+#include "edac_module.h"
+
+/*
+ * Alter this version for the I5400 module when modifications are made
+ */
+#define I5400_REVISION " Ver: 1.0.0"
+
+#define EDAC_MOD_STR "i5400_edac"
+
+#define i5400_printk(level, fmt, arg...) \
+ edac_printk(level, "i5400", fmt, ##arg)
+
+#define i5400_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i5400", fmt, ##arg)
+
+/* Limits for i5400 */
+#define MAX_BRANCHES 2
+#define CHANNELS_PER_BRANCH 2
+#define DIMMS_PER_CHANNEL 4
+#define MAX_CHANNELS (MAX_BRANCHES * CHANNELS_PER_BRANCH)
+
+/* Device 16,
+ * Function 0: System Address
+ * Function 1: Memory Branch Map, Control, Errors Register
+ * Function 2: FSB Error Registers
+ *
+ * All 3 functions of Device 16 (0,1,2) share the SAME DID and
+ * uses PCI_DEVICE_ID_INTEL_5400_ERR for device 16 (0,1,2),
+ * PCI_DEVICE_ID_INTEL_5400_FBD0 and PCI_DEVICE_ID_INTEL_5400_FBD1
+ * for device 21 (0,1).
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBASE 0x48 /* AMB Mem Mapped Reg Region Base */
+#define MAXCH 0x56 /* Max Channel Number */
+#define MAXDIMMPERCH 0x57 /* Max DIMM PER Channel Number */
+
+ /* OFFSETS for Function 1 */
+#define TOLM 0x6C
+#define REDMEMB 0x7C
+#define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3fe00) /* bits [17:9] indicate ODD, [8:0] indicate EVEN */
+#define MIR0 0x80
+#define MIR1 0x84
+#define AMIR0 0x8c
+#define AMIR1 0x90
+
+ /* Fatal error registers */
+#define FERR_FAT_FBD 0x98 /* also called as FERR_FAT_FB_DIMM at datasheet */
+#define FERR_FAT_FBDCHAN (3<<28) /* channel index where the highest-order error occurred */
+
+#define NERR_FAT_FBD 0x9c
+#define FERR_NF_FBD 0xa0 /* also called as FERR_NFAT_FB_DIMM at datasheet */
+
+ /* Non-fatal error register */
+#define NERR_NF_FBD 0xa4
+
+ /* Enable error mask */
+#define EMASK_FBD 0xa8
+
+#define ERR0_FBD 0xac
+#define ERR1_FBD 0xb0
+#define ERR2_FBD 0xb4
+#define MCERR_FBD 0xb8
+
+ /* No OFFSETS for Device 16 Function 2 */
+
+/*
+ * Device 21,
+ * Function 0: Memory Map Branch 0
+ *
+ * Device 22,
+ * Function 0: Memory Map Branch 1
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBPRESENT_0 0x64
+#define AMBPRESENT_1 0x66
+#define MTR0 0x80
+#define MTR1 0x82
+#define MTR2 0x84
+#define MTR3 0x86
+
+ /* OFFSETS for Function 1 */
+#define NRECFGLOG 0x74
+#define RECFGLOG 0x78
+#define NRECMEMA 0xbe
+#define NRECMEMB 0xc0
+#define NRECFB_DIMMA 0xc4
+#define NRECFB_DIMMB 0xc8
+#define NRECFB_DIMMC 0xcc
+#define NRECFB_DIMMD 0xd0
+#define NRECFB_DIMME 0xd4
+#define NRECFB_DIMMF 0xd8
+#define REDMEMA 0xdC
+#define RECMEMA 0xf0
+#define RECMEMB 0xf4
+#define RECFB_DIMMA 0xf8
+#define RECFB_DIMMB 0xec
+#define RECFB_DIMMC 0xf0
+#define RECFB_DIMMD 0xf4
+#define RECFB_DIMME 0xf8
+#define RECFB_DIMMF 0xfC
+
+/*
+ * Error indicator bits and masks
+ * Error masks are according with Table 5-17 of i5400 datasheet
+ */
+
+enum error_mask {
+ EMASK_M1 = 1<<0, /* Memory Write error on non-redundant retry */
+ EMASK_M2 = 1<<1, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M3 = 1<<2, /* Reserved */
+ EMASK_M4 = 1<<3, /* Uncorrectable Data ECC on Replay */
+ EMASK_M5 = 1<<4, /* Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M6 = 1<<5, /* Unsupported on i5400 */
+ EMASK_M7 = 1<<6, /* Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M8 = 1<<7, /* Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M9 = 1<<8, /* Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M10 = 1<<9, /* Unsupported on i5400 */
+ EMASK_M11 = 1<<10, /* Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M12 = 1<<11, /* Non-Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M13 = 1<<12, /* Memory Write error on first attempt */
+ EMASK_M14 = 1<<13, /* FB-DIMM Configuration Write error on first attempt */
+ EMASK_M15 = 1<<14, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M16 = 1<<15, /* Channel Failed-Over Occurred */
+ EMASK_M17 = 1<<16, /* Correctable Non-Mirrored Demand Data ECC */
+ EMASK_M18 = 1<<17, /* Unsupported on i5400 */
+ EMASK_M19 = 1<<18, /* Correctable Resilver- or Spare-Copy Data ECC */
+ EMASK_M20 = 1<<19, /* Correctable Patrol Data ECC */
+ EMASK_M21 = 1<<20, /* FB-DIMM Northbound parity error on FB-DIMM Sync Status */
+ EMASK_M22 = 1<<21, /* SPD protocol Error */
+ EMASK_M23 = 1<<22, /* Non-Redundant Fast Reset Timeout */
+ EMASK_M24 = 1<<23, /* Refresh error */
+ EMASK_M25 = 1<<24, /* Memory Write error on redundant retry */
+ EMASK_M26 = 1<<25, /* Redundant Fast Reset Timeout */
+ EMASK_M27 = 1<<26, /* Correctable Counter Threshold Exceeded */
+ EMASK_M28 = 1<<27, /* DIMM-Spare Copy Completed */
+ EMASK_M29 = 1<<28, /* DIMM-Isolation Completed */
+};
+
+/*
+ * Names to translate bit error into something useful
+ */
+static const char *error_name[] = {
+ [0] = "Memory Write error on non-redundant retry",
+ [1] = "Memory or FB-DIMM configuration CRC read error",
+ /* Reserved */
+ [3] = "Uncorrectable Data ECC on Replay",
+ [4] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* M6 Unsupported on i5400 */
+ [6] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [7] = "Aliased Uncorrectable Patrol Data ECC",
+ [8] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* M10 Unsupported on i5400 */
+ [10] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [11] = "Non-Aliased Uncorrectable Patrol Data ECC",
+ [12] = "Memory Write error on first attempt",
+ [13] = "FB-DIMM Configuration Write error on first attempt",
+ [14] = "Memory or FB-DIMM configuration CRC read error",
+ [15] = "Channel Failed-Over Occurred",
+ [16] = "Correctable Non-Mirrored Demand Data ECC",
+ /* M18 Unsupported on i5400 */
+ [18] = "Correctable Resilver- or Spare-Copy Data ECC",
+ [19] = "Correctable Patrol Data ECC",
+ [20] = "FB-DIMM Northbound parity error on FB-DIMM Sync Status",
+ [21] = "SPD protocol Error",
+ [22] = "Non-Redundant Fast Reset Timeout",
+ [23] = "Refresh error",
+ [24] = "Memory Write error on redundant retry",
+ [25] = "Redundant Fast Reset Timeout",
+ [26] = "Correctable Counter Threshold Exceeded",
+ [27] = "DIMM-Spare Copy Completed",
+ [28] = "DIMM-Isolation Completed",
+};
+
+/* Fatal errors */
+#define ERROR_FAT_MASK (EMASK_M1 | \
+ EMASK_M2 | \
+ EMASK_M23)
+
+/* Correctable errors */
+#define ERROR_NF_CORRECTABLE (EMASK_M27 | \
+ EMASK_M20 | \
+ EMASK_M19 | \
+ EMASK_M18 | \
+ EMASK_M17 | \
+ EMASK_M16)
+#define ERROR_NF_DIMM_SPARE (EMASK_M29 | \
+ EMASK_M28)
+#define ERROR_NF_SPD_PROTOCOL (EMASK_M22)
+#define ERROR_NF_NORTH_CRC (EMASK_M21)
+
+/* Recoverable errors */
+#define ERROR_NF_RECOVERABLE (EMASK_M26 | \
+ EMASK_M25 | \
+ EMASK_M24 | \
+ EMASK_M15 | \
+ EMASK_M14 | \
+ EMASK_M13 | \
+ EMASK_M12 | \
+ EMASK_M11 | \
+ EMASK_M9 | \
+ EMASK_M8 | \
+ EMASK_M7 | \
+ EMASK_M5)
+
+/* uncorrectable errors */
+#define ERROR_NF_UNCORRECTABLE (EMASK_M4)
+
+/* mask to all non-fatal errors */
+#define ERROR_NF_MASK (ERROR_NF_CORRECTABLE | \
+ ERROR_NF_UNCORRECTABLE | \
+ ERROR_NF_RECOVERABLE | \
+ ERROR_NF_DIMM_SPARE | \
+ ERROR_NF_SPD_PROTOCOL | \
+ ERROR_NF_NORTH_CRC)
+
+/*
+ * Define error masks for the several registers
+ */
+
+/* Enable all fatal and non fatal errors */
+#define ENABLE_EMASK_ALL (ERROR_FAT_MASK | ERROR_NF_MASK)
+
+/* mask for fatal error registers */
+#define FERR_FAT_MASK ERROR_FAT_MASK
+
+/* masks for non-fatal error register */
+static inline int to_nf_mask(unsigned int mask)
+{
+ return (mask & EMASK_M29) | (mask >> 3);
+};
+
+static inline int from_nf_ferr(unsigned int mask)
+{
+ return (mask & EMASK_M29) | /* Bit 28 */
+ (mask & ((1 << 28) - 1) << 3); /* Bits 0 to 27 */
+};
+
+#define FERR_NF_MASK to_nf_mask(ERROR_NF_MASK)
+#define FERR_NF_CORRECTABLE to_nf_mask(ERROR_NF_CORRECTABLE)
+#define FERR_NF_DIMM_SPARE to_nf_mask(ERROR_NF_DIMM_SPARE)
+#define FERR_NF_SPD_PROTOCOL to_nf_mask(ERROR_NF_SPD_PROTOCOL)
+#define FERR_NF_NORTH_CRC to_nf_mask(ERROR_NF_NORTH_CRC)
+#define FERR_NF_RECOVERABLE to_nf_mask(ERROR_NF_RECOVERABLE)
+#define FERR_NF_UNCORRECTABLE to_nf_mask(ERROR_NF_UNCORRECTABLE)
+
+/*
+ * Defines to extract the various fields from the
+ * MTRx - Memory Technology Registers
+ */
+#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 10))
+#define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 9))
+#define MTR_DRAM_WIDTH(mtr) (((mtr) & (1 << 8)) ? 8 : 4)
+#define MTR_DRAM_BANKS(mtr) (((mtr) & (1 << 6)) ? 8 : 4)
+#define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
+#define MTR_DIMM_RANK(mtr) (((mtr) >> 5) & 0x1)
+#define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
+#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
+#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
+#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
+#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
+
+/* This applies to FERR_NF_FB-DIMM as well as FERR_FAT_FB-DIMM */
+static inline int extract_fbdchan_indx(u32 x)
+{
+ return (x>>28) & 0x3;
+}
+
+/* Device name and register DID (Device ID) */
+struct i5400_dev_info {
+ const char *ctl_name; /* name for this device */
+ u16 fsb_mapping_errors; /* DID for the branchmap,control */
+};
+
+/* Table of devices attributes supported by this driver */
+static const struct i5400_dev_info i5400_devs[] = {
+ {
+ .ctl_name = "I5400",
+ .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_5400_ERR,
+ },
+};
+
+struct i5400_dimm_info {
+ int megabytes; /* size, 0 means not present */
+};
+
+/* driver private data structure */
+struct i5400_pvt {
+ struct pci_dev *system_address; /* 16.0 */
+ struct pci_dev *branchmap_werrors; /* 16.1 */
+ struct pci_dev *fsb_error_regs; /* 16.2 */
+ struct pci_dev *branch_0; /* 21.0 */
+ struct pci_dev *branch_1; /* 22.0 */
+
+ u16 tolm; /* top of low memory */
+ union {
+ u64 ambase; /* AMB BAR */
+ struct {
+ u32 ambase_bottom;
+ u32 ambase_top;
+ } u __packed;
+ };
+
+ u16 mir0, mir1;
+
+ u16 b0_mtr[DIMMS_PER_CHANNEL]; /* Memory Technlogy Reg */
+ u16 b0_ambpresent0; /* Branch 0, Channel 0 */
+ u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
+
+ u16 b1_mtr[DIMMS_PER_CHANNEL]; /* Memory Technlogy Reg */
+ u16 b1_ambpresent0; /* Branch 1, Channel 8 */
+ u16 b1_ambpresent1; /* Branch 1, Channel 1 */
+
+ /* DIMM information matrix, allocating architecture maximums */
+ struct i5400_dimm_info dimm_info[DIMMS_PER_CHANNEL][MAX_CHANNELS];
+
+ /* Actual values for this controller */
+ int maxch; /* Max channels */
+ int maxdimmperch; /* Max DIMMs per channel */
+};
+
+/* I5400 MCH error information retrieved from Hardware */
+struct i5400_error_info {
+ /* These registers are always read from the MC */
+ u32 ferr_fat_fbd; /* First Errors Fatal */
+ u32 nerr_fat_fbd; /* Next Errors Fatal */
+ u32 ferr_nf_fbd; /* First Errors Non-Fatal */
+ u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
+
+ /* These registers are input ONLY if there was a Recoverable Error */
+ u32 redmemb; /* Recoverable Mem Data Error log B */
+ u16 recmema; /* Recoverable Mem Error log A */
+ u32 recmemb; /* Recoverable Mem Error log B */
+
+ /* These registers are input ONLY if there was a Non-Rec Error */
+ u16 nrecmema; /* Non-Recoverable Mem log A */
+ u32 nrecmemb; /* Non-Recoverable Mem log B */
+
+};
+
+/* note that nrec_rdwr changed from NRECMEMA to NRECMEMB between the 5000 and
+ 5400 better to use an inline function than a macro in this case */
+static inline int nrec_bank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 12) & 0x7;
+}
+static inline int nrec_rank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 8) & 0xf;
+}
+static inline int nrec_buf_id(struct i5400_error_info *info)
+{
+ return ((info->nrecmema)) & 0xff;
+}
+static inline int nrec_rdwr(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) >> 31;
+}
+/* This applies to both NREC and REC string so it can be used with nrec_rdwr
+ and rec_rdwr */
+static inline const char *rdwr_str(int rdwr)
+{
+ return rdwr ? "Write" : "Read";
+}
+static inline int nrec_cas(struct i5400_error_info *info)
+{
+ return ((info->nrecmemb) >> 16) & 0x1fff;
+}
+static inline int nrec_ras(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) & 0xffff;
+}
+static inline int rec_bank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 12) & 0x7;
+}
+static inline int rec_rank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 8) & 0xf;
+}
+static inline int rec_rdwr(struct i5400_error_info *info)
+{
+ return (info->recmemb) >> 31;
+}
+static inline int rec_cas(struct i5400_error_info *info)
+{
+ return ((info->recmemb) >> 16) & 0x1fff;
+}
+static inline int rec_ras(struct i5400_error_info *info)
+{
+ return (info->recmemb) & 0xffff;
+}
+
+static struct edac_pci_ctl_info *i5400_pci;
+
+/*
+ * i5400_get_error_info Retrieve the hardware error information from
+ * the hardware and cache it in the 'info'
+ * structure
+ */
+static void i5400_get_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ struct i5400_pvt *pvt;
+ u32 value;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
+
+ /* Mask only the bits that the doc says are valid
+ */
+ value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
+
+ /* If there is an error, then read in the
+ NEXT FATAL error register and the Memory Error Log Register A
+ */
+ if (value & FERR_FAT_MASK) {
+ info->ferr_fat_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_FAT_FBD, &info->nerr_fat_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMA, &info->nrecmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NRECMEMB, &info->nrecmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_FAT_FBD, value);
+ } else {
+ info->ferr_fat_fbd = 0;
+ info->nerr_fat_fbd = 0;
+ info->nrecmema = 0;
+ info->nrecmemb = 0;
+ }
+
+ /* read in the 1st NON-FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
+
+ /* If there is an error, then read in the 1st NON-FATAL error
+ * register as well */
+ if (value & FERR_NF_MASK) {
+ info->ferr_nf_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_NF_FBD, &info->nerr_nf_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ RECMEMA, &info->recmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ RECMEMB, &info->recmemb);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ REDMEMB, &info->redmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_NF_FBD, value);
+ } else {
+ info->ferr_nf_fbd = 0;
+ info->nerr_nf_fbd = 0;
+ info->recmema = 0;
+ info->recmemb = 0;
+ info->redmemb = 0;
+ }
+}
+
+/*
+ * i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel FATAL and unrecoverable errors, if any
+ */
+static void i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info,
+ unsigned long allErrors)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ int branch;
+ int channel;
+ int bank;
+ int buf_id;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+ char *type = NULL;
+ enum hw_event_mc_err_type tp_event = HW_EVENT_ERR_UNCORRECTED;
+
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ if (allErrors & ERROR_FAT_MASK) {
+ type = "FATAL";
+ tp_event = HW_EVENT_ERR_FATAL;
+ } else if (allErrors & FERR_NF_UNCORRECTABLE)
+ type = "NON-FATAL uncorrected";
+ else
+ type = "NON-FATAL recoverable";
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ branch = extract_fbdchan_indx(info->ferr_fat_fbd);
+ channel = branch;
+
+ /* Use the NON-Recoverable macros to extract data */
+ bank = nrec_bank(info);
+ rank = nrec_rank(info);
+ buf_id = nrec_buf_id(info);
+ rdwr = nrec_rdwr(info);
+ ras = nrec_ras(info);
+ cas = nrec_cas(info);
+
+ edac_dbg(0, "\t\t%s DIMM= %d Channels= %d,%d (Branch= %d DRAM Bank= %d Buffer ID = %d rdwr= %s ras= %d cas= %d)\n",
+ type, rank, channel, channel + 1, branch >> 1, bank,
+ buf_id, rdwr_str(rdwr), ras, cas);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Bank=%d Buffer ID = %d RAS=%d CAS=%d Err=0x%lx (%s)",
+ bank, buf_id, ras, cas, allErrors, error_name[errnum]);
+
+ edac_mc_handle_error(tp_event, mci, 1, 0, 0, 0,
+ branch >> 1, -1, rank,
+ rdwr ? "Write error" : "Read error",
+ msg);
+}
+
+/*
+ * i5400_process_fatal_error_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel NON-FATAL errors, if any
+ */
+static void i5400_process_nonfatal_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ unsigned long allErrors;
+ int branch;
+ int channel;
+ int bank;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+
+ /* mask off the Error bits that are possible */
+ allErrors = from_nf_ferr(info->ferr_nf_fbd & FERR_NF_MASK);
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ if (allErrors & (ERROR_NF_UNCORRECTABLE | ERROR_NF_RECOVERABLE)) {
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+ return;
+ }
+
+ /* Correctable errors */
+ if (allErrors & ERROR_NF_CORRECTABLE) {
+ edac_dbg(0, "\tCorrected bits= 0x%lx\n", allErrors);
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ channel = 0;
+ if (REC_ECC_LOCATOR_ODD(info->redmemb))
+ channel = 1;
+
+ /* Convert channel to be based from zero, instead of
+ * from branch base of 0 */
+ channel += branch;
+
+ bank = rec_bank(info);
+ rank = rec_rank(info);
+ rdwr = rec_rdwr(info);
+ ras = rec_ras(info);
+ cas = rec_cas(info);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ edac_dbg(0, "\t\tDIMM= %d Channel= %d (Branch %d DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, branch >> 1, bank,
+ rdwr_str(rdwr), ras, cas);
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Corrected error (Branch=%d DRAM-Bank=%d RDWR=%s "
+ "RAS=%d CAS=%d, CE Err=0x%lx (%s))",
+ branch >> 1, bank, rdwr_str(rdwr), ras, cas,
+ allErrors, error_name[errnum]);
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0,
+ branch >> 1, channel % 2, rank,
+ rdwr ? "Write error" : "Read error",
+ msg);
+
+ return;
+ }
+
+ /* Miscellaneous errors */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ i5400_mc_printk(mci, KERN_EMERG,
+ "Non-Fatal misc error (Branch=%d Err=%#lx (%s))",
+ branch >> 1, allErrors, error_name[errnum]);
+}
+
+/*
+ * i5400_process_error_info Process the error info that is
+ * in the 'info' structure, previously retrieved from hardware
+ */
+static void i5400_process_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{ u32 allErrors;
+
+ /* First handle any fatal errors that occurred */
+ allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+
+ /* now handle any non-fatal errors that occurred */
+ i5400_process_nonfatal_error_info(mci, info);
+}
+
+/*
+ * i5400_clear_error Retrieve any error from the hardware
+ * but do NOT process that error.
+ * Used for 'clearing' out of previous errors
+ * Called by the Core module.
+ */
+static void i5400_clear_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+
+ i5400_get_error_info(mci, &info);
+}
+
+/*
+ * i5400_check_error Retrieve and process errors reported by the
+ * hardware. Called by the Core module.
+ */
+static void i5400_check_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+
+ i5400_get_error_info(mci, &info);
+ i5400_process_error_info(mci, &info);
+}
+
+/*
+ * i5400_put_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void i5400_put_devices(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+
+ pvt = mci->pvt_info;
+
+ /* Decrement usage count for devices */
+ pci_dev_put(pvt->branch_1);
+ pci_dev_put(pvt->branch_0);
+ pci_dev_put(pvt->fsb_error_regs);
+ pci_dev_put(pvt->branchmap_werrors);
+}
+
+/*
+ * i5400_get_devices Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ *
+ * Need to 'get' device 16 func 1 and func 2
+ */
+static int i5400_get_devices(struct mem_ctl_info *mci, int dev_idx)
+{
+ struct i5400_pvt *pvt;
+ struct pci_dev *pdev;
+
+ pvt = mci->pvt_info;
+ pvt->branchmap_werrors = NULL;
+ pvt->fsb_error_regs = NULL;
+ pvt->branch_0 = NULL;
+ pvt->branch_1 = NULL;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while (1) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
+ if (!pdev) {
+ /* End of list, leave */
+ i5400_printk(KERN_ERR,
+ "'system address,Process Bus' "
+ "device not found:"
+ "vendor 0x%x device 0x%x ERR func 1 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR);
+ return -ENODEV;
+ }
+
+ /* Store device 16 func 1 */
+ if (PCI_FUNC(pdev->devfn) == 1)
+ break;
+ }
+ pvt->branchmap_werrors = pdev;
+
+ pdev = NULL;
+ while (1) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
+ if (!pdev) {
+ /* End of list, leave */
+ i5400_printk(KERN_ERR,
+ "'system address,Process Bus' "
+ "device not found:"
+ "vendor 0x%x device 0x%x ERR func 2 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR);
+
+ pci_dev_put(pvt->branchmap_werrors);
+ return -ENODEV;
+ }
+
+ /* Store device 16 func 2 */
+ if (PCI_FUNC(pdev->devfn) == 2)
+ break;
+ }
+ pvt->fsb_error_regs = pdev;
+
+ edac_dbg(1, "System Address, processor bus- PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->system_address),
+ pvt->system_address->vendor, pvt->system_address->device);
+ edac_dbg(1, "Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->branchmap_werrors),
+ pvt->branchmap_werrors->vendor,
+ pvt->branchmap_werrors->device);
+ edac_dbg(1, "FSB Error Regs - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->fsb_error_regs),
+ pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
+
+ pvt->branch_0 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD0, NULL);
+ if (!pvt->branch_0) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 0' device not found:"
+ "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_FBD0);
+
+ pci_dev_put(pvt->fsb_error_regs);
+ pci_dev_put(pvt->branchmap_werrors);
+ return -ENODEV;
+ }
+
+ /* If this device claims to have more than 2 channels then
+ * fetch Branch 1's information
+ */
+ if (pvt->maxch < CHANNELS_PER_BRANCH)
+ return 0;
+
+ pvt->branch_1 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1, NULL);
+ if (!pvt->branch_1) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 1' device not found:"
+ "vendor 0x%x device 0x%x Func 0 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1);
+
+ pci_dev_put(pvt->branch_0);
+ pci_dev_put(pvt->fsb_error_regs);
+ pci_dev_put(pvt->branchmap_werrors);
+ return -ENODEV;
+ }
+
+ return 0;
+}
+
+/*
+ * determine_amb_present
+ *
+ * the information is contained in DIMMS_PER_CHANNEL different
+ * registers determining which of the DIMMS_PER_CHANNEL requires
+ * knowing which channel is in question
+ *
+ * 2 branches, each with 2 channels
+ * b0_ambpresent0 for channel '0'
+ * b0_ambpresent1 for channel '1'
+ * b1_ambpresent0 for channel '2'
+ * b1_ambpresent1 for channel '3'
+ */
+static int determine_amb_present_reg(struct i5400_pvt *pvt, int channel)
+{
+ int amb_present;
+
+ if (channel < CHANNELS_PER_BRANCH) {
+ if (channel & 0x1)
+ amb_present = pvt->b0_ambpresent1;
+ else
+ amb_present = pvt->b0_ambpresent0;
+ } else {
+ if (channel & 0x1)
+ amb_present = pvt->b1_ambpresent1;
+ else
+ amb_present = pvt->b1_ambpresent0;
+ }
+
+ return amb_present;
+}
+
+/*
+ * determine_mtr(pvt, dimm, channel)
+ *
+ * return the proper MTR register as determine by the dimm and desired channel
+ */
+static int determine_mtr(struct i5400_pvt *pvt, int dimm, int channel)
+{
+ int mtr;
+ int n;
+
+ /* There is one MTR for each slot pair of FB-DIMMs,
+ Each slot pair may be at branch 0 or branch 1.
+ */
+ n = dimm;
+
+ if (n >= DIMMS_PER_CHANNEL) {
+ edac_dbg(0, "ERROR: trying to access an invalid dimm: %d\n",
+ dimm);
+ return 0;
+ }
+
+ if (channel < CHANNELS_PER_BRANCH)
+ mtr = pvt->b0_mtr[n];
+ else
+ mtr = pvt->b1_mtr[n];
+
+ return mtr;
+}
+
+/*
+ */
+static void decode_mtr(int slot_row, u16 mtr)
+{
+ int ans;
+
+ ans = MTR_DIMMS_PRESENT(mtr);
+
+ edac_dbg(2, "\tMTR%d=0x%x: DIMMs are %sPresent\n",
+ slot_row, mtr, ans ? "" : "NOT ");
+ if (!ans)
+ return;
+
+ edac_dbg(2, "\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
+
+ edac_dbg(2, "\t\tELECTRICAL THROTTLING is %s\n",
+ MTR_DIMMS_ETHROTTLE(mtr) ? "enabled" : "disabled");
+
+ edac_dbg(2, "\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
+ edac_dbg(2, "\t\tNUMRANK: %s\n",
+ MTR_DIMM_RANK(mtr) ? "double" : "single");
+ edac_dbg(2, "\t\tNUMROW: %s\n",
+ MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" :
+ MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" :
+ MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" :
+ "65,536 - 16 rows");
+ edac_dbg(2, "\t\tNUMCOL: %s\n",
+ MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" :
+ MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" :
+ MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" :
+ "reserved");
+}
+
+static void handle_channel(struct i5400_pvt *pvt, int dimm, int channel,
+ struct i5400_dimm_info *dinfo)
+{
+ int mtr;
+ int amb_present_reg;
+ int addrBits;
+
+ mtr = determine_mtr(pvt, dimm, channel);
+ if (MTR_DIMMS_PRESENT(mtr)) {
+ amb_present_reg = determine_amb_present_reg(pvt, channel);
+
+ /* Determine if there is a DIMM present in this DIMM slot */
+ if (amb_present_reg & (1 << dimm)) {
+ /* Start with the number of bits for a Bank
+ * on the DRAM */
+ addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
+ /* Add thenumber of ROW bits */
+ addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
+ /* add the number of COLUMN bits */
+ addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
+ /* add the number of RANK bits */
+ addrBits += MTR_DIMM_RANK(mtr);
+
+ addrBits += 6; /* add 64 bits per DIMM */
+ addrBits -= 20; /* divide by 2^^20 */
+ addrBits -= 3; /* 8 bits per bytes */
+
+ dinfo->megabytes = 1 << addrBits;
+ }
+ }
+}
+
+/*
+ * calculate_dimm_size
+ *
+ * also will output a DIMM matrix map, if debug is enabled, for viewing
+ * how the DIMMs are populated
+ */
+static void calculate_dimm_size(struct i5400_pvt *pvt)
+{
+ struct i5400_dimm_info *dinfo;
+ int dimm, max_dimms;
+ char *p, *mem_buffer;
+ int space, n;
+ int channel, branch;
+
+ /* ================= Generate some debug output ================= */
+ space = PAGE_SIZE;
+ mem_buffer = p = kmalloc(space, GFP_KERNEL);
+ if (p == NULL) {
+ i5400_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
+ __FILE__, __func__);
+ return;
+ }
+
+ /* Scan all the actual DIMMS
+ * and calculate the information for each DIMM
+ * Start with the highest dimm first, to display it first
+ * and work toward the 0th dimm
+ */
+ max_dimms = pvt->maxdimmperch;
+ for (dimm = max_dimms - 1; dimm >= 0; dimm--) {
+
+ /* on an odd dimm, first output a 'boundary' marker,
+ * then reset the message buffer */
+ if (dimm & 0x1) {
+ n = snprintf(p, space, "---------------------------"
+ "-------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+ n = snprintf(p, space, "dimm %2d ", dimm);
+ p += n;
+ space -= n;
+
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ dinfo = &pvt->dimm_info[dimm][channel];
+ handle_channel(pvt, dimm, channel, dinfo);
+ n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
+ p += n;
+ space -= n;
+ }
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+
+ /* Output the last bottom 'boundary' marker */
+ n = snprintf(p, space, "---------------------------"
+ "-------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ /* now output the 'channel' labels */
+ n = snprintf(p, space, " ");
+ p += n;
+ space -= n;
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ n = snprintf(p, space, "channel %d | ", channel);
+ p += n;
+ space -= n;
+ }
+
+ space -= n;
+ edac_dbg(2, "%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ n = snprintf(p, space, " ");
+ p += n;
+ for (branch = 0; branch < MAX_BRANCHES; branch++) {
+ n = snprintf(p, space, " branch %d | ", branch);
+ p += n;
+ space -= n;
+ }
+
+ /* output the last message and free buffer */
+ edac_dbg(2, "%s\n", mem_buffer);
+ kfree(mem_buffer);
+}
+
+/*
+ * i5400_get_mc_regs read in the necessary registers and
+ * cache locally
+ *
+ * Fills in the private data members
+ */
+static void i5400_get_mc_regs(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 actual_tolm;
+ u16 limit;
+ int slot_row;
+ int way0, way1;
+
+ pvt = mci->pvt_info;
+
+ pci_read_config_dword(pvt->system_address, AMBASE,
+ &pvt->u.ambase_bottom);
+ pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
+ &pvt->u.ambase_top);
+
+ edac_dbg(2, "AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
+ (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
+
+ /* Get the Branch Map regs */
+ pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
+ pvt->tolm >>= 12;
+ edac_dbg(2, "\nTOLM (number of 256M regions) =%u (0x%x)\n",
+ pvt->tolm, pvt->tolm);
+
+ actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28));
+ edac_dbg(2, "Actual TOLM byte addr=%u.%03u GB (0x%x)\n",
+ actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28);
+
+ pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
+ pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
+
+ /* Get the MIR[0-1] regs */
+ limit = (pvt->mir0 >> 4) & 0x0fff;
+ way0 = pvt->mir0 & 0x1;
+ way1 = pvt->mir0 & 0x2;
+ edac_dbg(2, "MIR0: limit= 0x%x WAY1= %u WAY0= %x\n",
+ limit, way1, way0);
+ limit = (pvt->mir1 >> 4) & 0xfff;
+ way0 = pvt->mir1 & 0x1;
+ way1 = pvt->mir1 & 0x2;
+ edac_dbg(2, "MIR1: limit= 0x%x WAY1= %u WAY0= %x\n",
+ limit, way1, way0);
+
+ /* Get the set of MTR[0-3] regs by each branch */
+ for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++) {
+ int where = MTR0 + (slot_row * sizeof(u16));
+
+ /* Branch 0 set of MTR registers */
+ pci_read_config_word(pvt->branch_0, where,
+ &pvt->b0_mtr[slot_row]);
+
+ edac_dbg(2, "MTR%d where=0x%x B0 value=0x%x\n",
+ slot_row, where, pvt->b0_mtr[slot_row]);
+
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_mtr[slot_row] = 0;
+ continue;
+ }
+
+ /* Branch 1 set of MTR registers */
+ pci_read_config_word(pvt->branch_1, where,
+ &pvt->b1_mtr[slot_row]);
+ edac_dbg(2, "MTR%d where=0x%x B1 value=0x%x\n",
+ slot_row, where, pvt->b1_mtr[slot_row]);
+ }
+
+ /* Read and dump branch 0's MTRs */
+ edac_dbg(2, "Memory Technology Registers:\n");
+ edac_dbg(2, " Branch 0:\n");
+ for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++)
+ decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_0,
+ &pvt->b0_ambpresent0);
+ edac_dbg(2, "\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_1,
+ &pvt->b0_ambpresent1);
+ edac_dbg(2, "\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
+
+ /* Only if we have 2 branchs (4 channels) */
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_ambpresent0 = 0;
+ pvt->b1_ambpresent1 = 0;
+ } else {
+ /* Read and dump branch 1's MTRs */
+ edac_dbg(2, " Branch 1:\n");
+ for (slot_row = 0; slot_row < DIMMS_PER_CHANNEL; slot_row++)
+ decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_0,
+ &pvt->b1_ambpresent0);
+ edac_dbg(2, "\t\tAMB-Branch 1-present0 0x%x:\n",
+ pvt->b1_ambpresent0);
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_1,
+ &pvt->b1_ambpresent1);
+ edac_dbg(2, "\t\tAMB-Branch 1-present1 0x%x:\n",
+ pvt->b1_ambpresent1);
+ }
+
+ /* Go and determine the size of each DIMM and place in an
+ * orderly matrix */
+ calculate_dimm_size(pvt);
+}
+
+/*
+ * i5400_init_dimms Initialize the 'dimms' table within
+ * the mci control structure with the
+ * addressing of memory.
+ *
+ * return:
+ * 0 success
+ * 1 no actual memory found on this MC
+ */
+static int i5400_init_dimms(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ struct dimm_info *dimm;
+ int ndimms;
+ int mtr;
+ int size_mb;
+ int channel, slot;
+
+ pvt = mci->pvt_info;
+
+ ndimms = 0;
+
+ /*
+ * FIXME: remove pvt->dimm_info[slot][channel] and use the 3
+ * layers here.
+ */
+ for (channel = 0; channel < mci->layers[0].size * mci->layers[1].size;
+ channel++) {
+ for (slot = 0; slot < mci->layers[2].size; slot++) {
+ mtr = determine_mtr(pvt, slot, channel);
+
+ /* if no DIMMS on this slot, continue */
+ if (!MTR_DIMMS_PRESENT(mtr))
+ continue;
+
+ dimm = edac_get_dimm(mci, channel / 2, channel % 2, slot);
+
+ size_mb = pvt->dimm_info[slot][channel].megabytes;
+
+ edac_dbg(2, "dimm (branch %d channel %d slot %d): %d.%03d GB\n",
+ channel / 2, channel % 2, slot,
+ size_mb / 1000, size_mb % 1000);
+
+ dimm->nr_pages = size_mb << 8;
+ dimm->grain = 8;
+ dimm->dtype = MTR_DRAM_WIDTH(mtr) == 8 ?
+ DEV_X8 : DEV_X4;
+ dimm->mtype = MEM_FB_DDR2;
+ /*
+ * The eccc mechanism is SDDC (aka SECC), with
+ * is similar to Chipkill.
+ */
+ dimm->edac_mode = MTR_DRAM_WIDTH(mtr) == 8 ?
+ EDAC_S8ECD8ED : EDAC_S4ECD4ED;
+ ndimms++;
+ }
+ }
+
+ /*
+ * When just one memory is provided, it should be at location (0,0,0).
+ * With such single-DIMM mode, the SDCC algorithm degrades to SECDEC+.
+ */
+ if (ndimms == 1)
+ mci->dimms[0]->edac_mode = EDAC_SECDED;
+
+ return (ndimms == 0);
+}
+
+/*
+ * i5400_enable_error_reporting
+ * Turn on the memory reporting features of the hardware
+ */
+static void i5400_enable_error_reporting(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 fbd_error_mask;
+
+ pvt = mci->pvt_info;
+
+ /* Read the FBD Error Mask Register */
+ pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ &fbd_error_mask);
+
+ /* Enable with a '0' */
+ fbd_error_mask &= ~(ENABLE_EMASK_ALL);
+
+ pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ fbd_error_mask);
+}
+
+/*
+ * i5400_probe1 Probe for ONE instance of device to see if it is
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct i5400_pvt *pvt;
+ struct edac_mc_layer layers[3];
+
+ if (dev_idx >= ARRAY_SIZE(i5400_devs))
+ return -EINVAL;
+
+ edac_dbg(0, "MC: pdev bus %u dev=0x%x fn=0x%x\n",
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+
+ /* We only are looking for func 0 of the set */
+ if (PCI_FUNC(pdev->devfn) != 0)
+ return -ENODEV;
+
+ /*
+ * allocate a new MC control structure
+ *
+ * This drivers uses the DIMM slot as "csrow" and the rest as "channel".
+ */
+ layers[0].type = EDAC_MC_LAYER_BRANCH;
+ layers[0].size = MAX_BRANCHES;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = CHANNELS_PER_BRANCH;
+ layers[1].is_virt_csrow = false;
+ layers[2].type = EDAC_MC_LAYER_SLOT;
+ layers[2].size = DIMMS_PER_CHANNEL;
+ layers[2].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p\n", mci);
+
+ mci->pdev = &pdev->dev; /* record ptr to the generic device */
+
+ pvt = mci->pvt_info;
+ pvt->system_address = pdev; /* Record this device in our private */
+ pvt->maxch = MAX_CHANNELS;
+ pvt->maxdimmperch = DIMMS_PER_CHANNEL;
+
+ /* 'get' the pci devices we want to reserve for our use */
+ if (i5400_get_devices(mci, dev_idx))
+ goto fail0;
+
+ /* Time to get serious */
+ i5400_get_mc_regs(mci); /* retrieve the hardware registers */
+
+ mci->mc_idx = 0;
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i5400_edac.c";
+ mci->ctl_name = i5400_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Set the function pointer to an actual operation function */
+ mci->edac_check = i5400_check_error;
+
+ /* initialize the MC control structure 'dimms' table
+ * with the mapping and control information */
+ if (i5400_init_dimms(mci)) {
+ edac_dbg(0, "MC: Setting mci->edac_cap to EDAC_FLAG_NONE because i5400_init_dimms() returned nonzero value\n");
+ mci->edac_cap = EDAC_FLAG_NONE; /* no dimms found */
+ } else {
+ edac_dbg(1, "MC: Enable error reporting now\n");
+ i5400_enable_error_reporting(mci);
+ }
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+ goto fail1;
+ }
+
+ i5400_clear_error(mci);
+
+ /* allocating generic PCI control info */
+ i5400_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i5400_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ return 0;
+
+ /* Error exit unwinding stack */
+fail1:
+
+ i5400_put_devices(mci);
+
+fail0:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/*
+ * i5400_init_one constructor for one instance of device
+ *
+ * returns:
+ * negative on error
+ * count (>= 0)
+ */
+static int i5400_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ /* wake up device */
+ rc = pci_enable_device(pdev);
+ if (rc)
+ return rc;
+
+ /* now probe and enable the device */
+ return i5400_probe1(pdev, id->driver_data);
+}
+
+/*
+ * i5400_remove_one destructor for one instance of device
+ *
+ */
+static void i5400_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (i5400_pci)
+ edac_pci_release_generic_ctl(i5400_pci);
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ /* retrieve references to resources, and free those resources */
+ i5400_put_devices(mci);
+
+ pci_disable_device(pdev);
+
+ edac_mc_free(mci);
+}
+
+/*
+ * pci_device_id table for which devices we are looking for
+ *
+ * The "E500P" device is the first device supported.
+ */
+static const struct pci_device_id i5400_pci_tbl[] = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_ERR)},
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i5400_pci_tbl);
+
+/*
+ * i5400_driver pci_driver structure for this module
+ *
+ */
+static struct pci_driver i5400_driver = {
+ .name = "i5400_edac",
+ .probe = i5400_init_one,
+ .remove = i5400_remove_one,
+ .id_table = i5400_pci_tbl,
+};
+
+/*
+ * i5400_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init i5400_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(2, "MC:\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i5400_driver);
+
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+/*
+ * i5400_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit i5400_exit(void)
+{
+ edac_dbg(2, "MC:\n");
+ pci_unregister_driver(&i5400_driver);
+}
+
+module_init(i5400_init);
+module_exit(i5400_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ben Woodard <woodard@redhat.com>");
+MODULE_AUTHOR("Mauro Carvalho Chehab");
+MODULE_AUTHOR("Red Hat Inc. (https://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel I5400 memory controllers - "
+ I5400_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i7300_edac.c b/drivers/edac/i7300_edac.c
new file mode 100644
index 0000000000..61adaa872b
--- /dev/null
+++ b/drivers/edac/i7300_edac.c
@@ -0,0 +1,1214 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Intel 7300 class Memory Controllers kernel module (Clarksboro)
+ *
+ * Copyright (c) 2010 by:
+ * Mauro Carvalho Chehab
+ *
+ * Red Hat Inc. https://www.redhat.com
+ *
+ * Intel 7300 Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://www.intel.com/Assets/PDF/datasheet/318082.pdf
+ *
+ * TODO: The chipset allow checking for PCI Express errors also. Currently,
+ * the driver covers only memory error errors
+ *
+ * This driver uses "csrows" EDAC attribute to represent DIMM slot#
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+
+#include "edac_module.h"
+
+/*
+ * Alter this version for the I7300 module when modifications are made
+ */
+#define I7300_REVISION " Ver: 1.0.0"
+
+#define EDAC_MOD_STR "i7300_edac"
+
+#define i7300_printk(level, fmt, arg...) \
+ edac_printk(level, "i7300", fmt, ##arg)
+
+#define i7300_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i7300", fmt, ##arg)
+
+/***********************************************
+ * i7300 Limit constants Structs and static vars
+ ***********************************************/
+
+/*
+ * Memory topology is organized as:
+ * Branch 0 - 2 channels: channels 0 and 1 (FDB0 PCI dev 21.0)
+ * Branch 1 - 2 channels: channels 2 and 3 (FDB1 PCI dev 22.0)
+ * Each channel can have to 8 DIMM sets (called as SLOTS)
+ * Slots should generally be filled in pairs
+ * Except on Single Channel mode of operation
+ * just slot 0/channel0 filled on this mode
+ * On normal operation mode, the two channels on a branch should be
+ * filled together for the same SLOT#
+ * When in mirrored mode, Branch 1 replicate memory at Branch 0, so, the four
+ * channels on both branches should be filled
+ */
+
+/* Limits for i7300 */
+#define MAX_SLOTS 8
+#define MAX_BRANCHES 2
+#define MAX_CH_PER_BRANCH 2
+#define MAX_CHANNELS (MAX_CH_PER_BRANCH * MAX_BRANCHES)
+#define MAX_MIR 3
+
+#define to_channel(ch, branch) ((((branch)) << 1) | (ch))
+
+#define to_csrow(slot, ch, branch) \
+ (to_channel(ch, branch) | ((slot) << 2))
+
+/* Device name and register DID (Device ID) */
+struct i7300_dev_info {
+ const char *ctl_name; /* name for this device */
+ u16 fsb_mapping_errors; /* DID for the branchmap,control */
+};
+
+/* Table of devices attributes supported by this driver */
+static const struct i7300_dev_info i7300_devs[] = {
+ {
+ .ctl_name = "I7300",
+ .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I7300_MCH_ERR,
+ },
+};
+
+struct i7300_dimm_info {
+ int megabytes; /* size, 0 means not present */
+};
+
+/* driver private data structure */
+struct i7300_pvt {
+ struct pci_dev *pci_dev_16_0_fsb_ctlr; /* 16.0 */
+ struct pci_dev *pci_dev_16_1_fsb_addr_map; /* 16.1 */
+ struct pci_dev *pci_dev_16_2_fsb_err_regs; /* 16.2 */
+ struct pci_dev *pci_dev_2x_0_fbd_branch[MAX_BRANCHES]; /* 21.0 and 22.0 */
+
+ u16 tolm; /* top of low memory */
+ u64 ambase; /* AMB BAR */
+
+ u32 mc_settings; /* Report several settings */
+ u32 mc_settings_a;
+
+ u16 mir[MAX_MIR]; /* Memory Interleave Reg*/
+
+ u16 mtr[MAX_SLOTS][MAX_BRANCHES]; /* Memory Technlogy Reg */
+ u16 ambpresent[MAX_CHANNELS]; /* AMB present regs */
+
+ /* DIMM information matrix, allocating architecture maximums */
+ struct i7300_dimm_info dimm_info[MAX_SLOTS][MAX_CHANNELS];
+
+ /* Temporary buffer for use when preparing error messages */
+ char *tmp_prt_buffer;
+};
+
+/* FIXME: Why do we need to have this static? */
+static struct edac_pci_ctl_info *i7300_pci;
+
+/***************************************************
+ * i7300 Register definitions for memory enumeration
+ ***************************************************/
+
+/*
+ * Device 16,
+ * Function 0: System Address (not documented)
+ * Function 1: Memory Branch Map, Control, Errors Register
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBASE 0x48 /* AMB Mem Mapped Reg Region Base */
+#define MAXCH 0x56 /* Max Channel Number */
+#define MAXDIMMPERCH 0x57 /* Max DIMM PER Channel Number */
+
+ /* OFFSETS for Function 1 */
+#define MC_SETTINGS 0x40
+ #define IS_MIRRORED(mc) ((mc) & (1 << 16))
+ #define IS_ECC_ENABLED(mc) ((mc) & (1 << 5))
+ #define IS_RETRY_ENABLED(mc) ((mc) & (1 << 31))
+ #define IS_SCRBALGO_ENHANCED(mc) ((mc) & (1 << 8))
+
+#define MC_SETTINGS_A 0x58
+ #define IS_SINGLE_MODE(mca) ((mca) & (1 << 14))
+
+#define TOLM 0x6C
+
+#define MIR0 0x80
+#define MIR1 0x84
+#define MIR2 0x88
+
+/*
+ * Note: Other Intel EDAC drivers use AMBPRESENT to identify if the available
+ * memory. From datasheet item 7.3.1 (FB-DIMM technology & organization), it
+ * seems that we cannot use this information directly for the same usage.
+ * Each memory slot may have up to 2 AMB interfaces, one for income and another
+ * for outcome interface to the next slot.
+ * For now, the driver just stores the AMB present registers, but rely only at
+ * the MTR info to detect memory.
+ * Datasheet is also not clear about how to map each AMBPRESENT registers to
+ * one of the 4 available channels.
+ */
+#define AMBPRESENT_0 0x64
+#define AMBPRESENT_1 0x66
+
+static const u16 mtr_regs[MAX_SLOTS] = {
+ 0x80, 0x84, 0x88, 0x8c,
+ 0x82, 0x86, 0x8a, 0x8e
+};
+
+/*
+ * Defines to extract the vaious fields from the
+ * MTRx - Memory Technology Registers
+ */
+#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 8))
+#define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 7))
+#define MTR_DRAM_WIDTH(mtr) (((mtr) & (1 << 6)) ? 8 : 4)
+#define MTR_DRAM_BANKS(mtr) (((mtr) & (1 << 5)) ? 8 : 4)
+#define MTR_DIMM_RANKS(mtr) (((mtr) & (1 << 4)) ? 1 : 0)
+#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
+#define MTR_DRAM_BANKS_ADDR_BITS 2
+#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
+#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
+#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
+
+/************************************************
+ * i7300 Register definitions for error detection
+ ************************************************/
+
+/*
+ * Device 16.1: FBD Error Registers
+ */
+#define FERR_FAT_FBD 0x98
+static const char *ferr_fat_fbd_name[] = {
+ [22] = "Non-Redundant Fast Reset Timeout",
+ [2] = ">Tmid Thermal event with intelligent throttling disabled",
+ [1] = "Memory or FBD configuration CRC read error",
+ [0] = "Memory Write error on non-redundant retry or "
+ "FBD configuration Write error on retry",
+};
+#define GET_FBD_FAT_IDX(fbderr) (((fbderr) >> 28) & 3)
+#define FERR_FAT_FBD_ERR_MASK ((1 << 0) | (1 << 1) | (1 << 2) | (1 << 22))
+
+#define FERR_NF_FBD 0xa0
+static const char *ferr_nf_fbd_name[] = {
+ [24] = "DIMM-Spare Copy Completed",
+ [23] = "DIMM-Spare Copy Initiated",
+ [22] = "Redundant Fast Reset Timeout",
+ [21] = "Memory Write error on redundant retry",
+ [18] = "SPD protocol Error",
+ [17] = "FBD Northbound parity error on FBD Sync Status",
+ [16] = "Correctable Patrol Data ECC",
+ [15] = "Correctable Resilver- or Spare-Copy Data ECC",
+ [14] = "Correctable Mirrored Demand Data ECC",
+ [13] = "Correctable Non-Mirrored Demand Data ECC",
+ [11] = "Memory or FBD configuration CRC read error",
+ [10] = "FBD Configuration Write error on first attempt",
+ [9] = "Memory Write error on first attempt",
+ [8] = "Non-Aliased Uncorrectable Patrol Data ECC",
+ [7] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [6] = "Non-Aliased Uncorrectable Mirrored Demand Data ECC",
+ [5] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ [4] = "Aliased Uncorrectable Patrol Data ECC",
+ [3] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [2] = "Aliased Uncorrectable Mirrored Demand Data ECC",
+ [1] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ [0] = "Uncorrectable Data ECC on Replay",
+};
+#define GET_FBD_NF_IDX(fbderr) (((fbderr) >> 28) & 3)
+#define FERR_NF_FBD_ERR_MASK ((1 << 24) | (1 << 23) | (1 << 22) | (1 << 21) |\
+ (1 << 18) | (1 << 17) | (1 << 16) | (1 << 15) |\
+ (1 << 14) | (1 << 13) | (1 << 11) | (1 << 10) |\
+ (1 << 9) | (1 << 8) | (1 << 7) | (1 << 6) |\
+ (1 << 5) | (1 << 4) | (1 << 3) | (1 << 2) |\
+ (1 << 1) | (1 << 0))
+
+#define EMASK_FBD 0xa8
+#define EMASK_FBD_ERR_MASK ((1 << 27) | (1 << 26) | (1 << 25) | (1 << 24) |\
+ (1 << 22) | (1 << 21) | (1 << 20) | (1 << 19) |\
+ (1 << 18) | (1 << 17) | (1 << 16) | (1 << 14) |\
+ (1 << 13) | (1 << 12) | (1 << 11) | (1 << 10) |\
+ (1 << 9) | (1 << 8) | (1 << 7) | (1 << 6) |\
+ (1 << 5) | (1 << 4) | (1 << 3) | (1 << 2) |\
+ (1 << 1) | (1 << 0))
+
+/*
+ * Device 16.2: Global Error Registers
+ */
+
+#define FERR_GLOBAL_HI 0x48
+static const char *ferr_global_hi_name[] = {
+ [3] = "FSB 3 Fatal Error",
+ [2] = "FSB 2 Fatal Error",
+ [1] = "FSB 1 Fatal Error",
+ [0] = "FSB 0 Fatal Error",
+};
+#define ferr_global_hi_is_fatal(errno) 1
+
+#define FERR_GLOBAL_LO 0x40
+static const char *ferr_global_lo_name[] = {
+ [31] = "Internal MCH Fatal Error",
+ [30] = "Intel QuickData Technology Device Fatal Error",
+ [29] = "FSB1 Fatal Error",
+ [28] = "FSB0 Fatal Error",
+ [27] = "FBD Channel 3 Fatal Error",
+ [26] = "FBD Channel 2 Fatal Error",
+ [25] = "FBD Channel 1 Fatal Error",
+ [24] = "FBD Channel 0 Fatal Error",
+ [23] = "PCI Express Device 7Fatal Error",
+ [22] = "PCI Express Device 6 Fatal Error",
+ [21] = "PCI Express Device 5 Fatal Error",
+ [20] = "PCI Express Device 4 Fatal Error",
+ [19] = "PCI Express Device 3 Fatal Error",
+ [18] = "PCI Express Device 2 Fatal Error",
+ [17] = "PCI Express Device 1 Fatal Error",
+ [16] = "ESI Fatal Error",
+ [15] = "Internal MCH Non-Fatal Error",
+ [14] = "Intel QuickData Technology Device Non Fatal Error",
+ [13] = "FSB1 Non-Fatal Error",
+ [12] = "FSB 0 Non-Fatal Error",
+ [11] = "FBD Channel 3 Non-Fatal Error",
+ [10] = "FBD Channel 2 Non-Fatal Error",
+ [9] = "FBD Channel 1 Non-Fatal Error",
+ [8] = "FBD Channel 0 Non-Fatal Error",
+ [7] = "PCI Express Device 7 Non-Fatal Error",
+ [6] = "PCI Express Device 6 Non-Fatal Error",
+ [5] = "PCI Express Device 5 Non-Fatal Error",
+ [4] = "PCI Express Device 4 Non-Fatal Error",
+ [3] = "PCI Express Device 3 Non-Fatal Error",
+ [2] = "PCI Express Device 2 Non-Fatal Error",
+ [1] = "PCI Express Device 1 Non-Fatal Error",
+ [0] = "ESI Non-Fatal Error",
+};
+#define ferr_global_lo_is_fatal(errno) ((errno < 16) ? 0 : 1)
+
+#define NRECMEMA 0xbe
+ #define NRECMEMA_BANK(v) (((v) >> 12) & 7)
+ #define NRECMEMA_RANK(v) (((v) >> 8) & 15)
+
+#define NRECMEMB 0xc0
+ #define NRECMEMB_IS_WR(v) ((v) & (1 << 31))
+ #define NRECMEMB_CAS(v) (((v) >> 16) & 0x1fff)
+ #define NRECMEMB_RAS(v) ((v) & 0xffff)
+
+#define REDMEMA 0xdc
+
+#define REDMEMB 0x7c
+
+#define RECMEMA 0xe0
+ #define RECMEMA_BANK(v) (((v) >> 12) & 7)
+ #define RECMEMA_RANK(v) (((v) >> 8) & 15)
+
+#define RECMEMB 0xe4
+ #define RECMEMB_IS_WR(v) ((v) & (1 << 31))
+ #define RECMEMB_CAS(v) (((v) >> 16) & 0x1fff)
+ #define RECMEMB_RAS(v) ((v) & 0xffff)
+
+/********************************************
+ * i7300 Functions related to error detection
+ ********************************************/
+
+/**
+ * get_err_from_table() - Gets the error message from a table
+ * @table: table name (array of char *)
+ * @size: number of elements at the table
+ * @pos: position of the element to be returned
+ *
+ * This is a small routine that gets the pos-th element of a table. If the
+ * element doesn't exist (or it is empty), it returns "reserved".
+ * Instead of calling it directly, the better is to call via the macro
+ * GET_ERR_FROM_TABLE(), that automatically checks the table size via
+ * ARRAY_SIZE() macro
+ */
+static const char *get_err_from_table(const char *table[], int size, int pos)
+{
+ if (unlikely(pos >= size))
+ return "Reserved";
+
+ if (unlikely(!table[pos]))
+ return "Reserved";
+
+ return table[pos];
+}
+
+#define GET_ERR_FROM_TABLE(table, pos) \
+ get_err_from_table(table, ARRAY_SIZE(table), pos)
+
+/**
+ * i7300_process_error_global() - Retrieve the hardware error information from
+ * the hardware global error registers and
+ * sends it to dmesg
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_process_error_global(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ u32 errnum, error_reg;
+ unsigned long errors;
+ const char *specific;
+ bool is_fatal;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_HI, &error_reg);
+ if (unlikely(error_reg)) {
+ errors = error_reg;
+ errnum = find_first_bit(&errors,
+ ARRAY_SIZE(ferr_global_hi_name));
+ specific = GET_ERR_FROM_TABLE(ferr_global_hi_name, errnum);
+ is_fatal = ferr_global_hi_is_fatal(errnum);
+
+ /* Clear the error bit */
+ pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_HI, error_reg);
+
+ goto error_global;
+ }
+
+ pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_LO, &error_reg);
+ if (unlikely(error_reg)) {
+ errors = error_reg;
+ errnum = find_first_bit(&errors,
+ ARRAY_SIZE(ferr_global_lo_name));
+ specific = GET_ERR_FROM_TABLE(ferr_global_lo_name, errnum);
+ is_fatal = ferr_global_lo_is_fatal(errnum);
+
+ /* Clear the error bit */
+ pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_LO, error_reg);
+
+ goto error_global;
+ }
+ return;
+
+error_global:
+ i7300_mc_printk(mci, KERN_EMERG, "%s misc error: %s\n",
+ is_fatal ? "Fatal" : "NOT fatal", specific);
+}
+
+/**
+ * i7300_process_fbd_error() - Retrieve the hardware error information from
+ * the FBD error registers and sends it via
+ * EDAC error API calls
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_process_fbd_error(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ u32 errnum, value, error_reg;
+ u16 val16;
+ unsigned branch, channel, bank, rank, cas, ras;
+ u32 syndrome;
+
+ unsigned long errors;
+ const char *specific;
+ bool is_wr;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_FAT_FBD, &error_reg);
+ if (unlikely(error_reg & FERR_FAT_FBD_ERR_MASK)) {
+ errors = error_reg & FERR_FAT_FBD_ERR_MASK ;
+ errnum = find_first_bit(&errors,
+ ARRAY_SIZE(ferr_fat_fbd_name));
+ specific = GET_ERR_FROM_TABLE(ferr_fat_fbd_name, errnum);
+ branch = (GET_FBD_FAT_IDX(error_reg) == 2) ? 1 : 0;
+
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map,
+ NRECMEMA, &val16);
+ bank = NRECMEMA_BANK(val16);
+ rank = NRECMEMA_RANK(val16);
+
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ NRECMEMB, &value);
+ is_wr = NRECMEMB_IS_WR(value);
+ cas = NRECMEMB_CAS(value);
+ ras = NRECMEMB_RAS(value);
+
+ /* Clean the error register */
+ pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_FAT_FBD, error_reg);
+
+ snprintf(pvt->tmp_prt_buffer, PAGE_SIZE,
+ "Bank=%d RAS=%d CAS=%d Err=0x%lx (%s))",
+ bank, ras, cas, errors, specific);
+
+ edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, 1, 0, 0, 0,
+ branch, -1, rank,
+ is_wr ? "Write error" : "Read error",
+ pvt->tmp_prt_buffer);
+
+ }
+
+ /* read in the 1st NON-FATAL error register */
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_NF_FBD, &error_reg);
+ if (unlikely(error_reg & FERR_NF_FBD_ERR_MASK)) {
+ errors = error_reg & FERR_NF_FBD_ERR_MASK;
+ errnum = find_first_bit(&errors,
+ ARRAY_SIZE(ferr_nf_fbd_name));
+ specific = GET_ERR_FROM_TABLE(ferr_nf_fbd_name, errnum);
+ branch = (GET_FBD_NF_IDX(error_reg) == 2) ? 1 : 0;
+
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ REDMEMA, &syndrome);
+
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map,
+ RECMEMA, &val16);
+ bank = RECMEMA_BANK(val16);
+ rank = RECMEMA_RANK(val16);
+
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ RECMEMB, &value);
+ is_wr = RECMEMB_IS_WR(value);
+ cas = RECMEMB_CAS(value);
+ ras = RECMEMB_RAS(value);
+
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ REDMEMB, &value);
+ channel = (branch << 1);
+
+ /* Second channel ? */
+ channel += !!(value & BIT(17));
+
+ /* Clear the error bit */
+ pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_NF_FBD, error_reg);
+
+ /* Form out message */
+ snprintf(pvt->tmp_prt_buffer, PAGE_SIZE,
+ "DRAM-Bank=%d RAS=%d CAS=%d, Err=0x%lx (%s))",
+ bank, ras, cas, errors, specific);
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0,
+ syndrome,
+ branch >> 1, channel % 2, rank,
+ is_wr ? "Write error" : "Read error",
+ pvt->tmp_prt_buffer);
+ }
+ return;
+}
+
+/**
+ * i7300_check_error() - Calls the error checking subroutines
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_check_error(struct mem_ctl_info *mci)
+{
+ i7300_process_error_global(mci);
+ i7300_process_fbd_error(mci);
+};
+
+/**
+ * i7300_clear_error() - Clears the error registers
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_clear_error(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt = mci->pvt_info;
+ u32 value;
+ /*
+ * All error values are RWC - we need to read and write 1 to the
+ * bit that we want to cleanup
+ */
+
+ /* Clear global error registers */
+ pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_HI, &value);
+ pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_HI, value);
+
+ pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_LO, &value);
+ pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs,
+ FERR_GLOBAL_LO, value);
+
+ /* Clear FBD error registers */
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_FAT_FBD, &value);
+ pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_FAT_FBD, value);
+
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_NF_FBD, &value);
+ pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ FERR_NF_FBD, value);
+}
+
+/**
+ * i7300_enable_error_reporting() - Enable the memory reporting logic at the
+ * hardware
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_enable_error_reporting(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt = mci->pvt_info;
+ u32 fbd_error_mask;
+
+ /* Read the FBD Error Mask Register */
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ EMASK_FBD, &fbd_error_mask);
+
+ /* Enable with a '0' */
+ fbd_error_mask &= ~(EMASK_FBD_ERR_MASK);
+
+ pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map,
+ EMASK_FBD, fbd_error_mask);
+}
+
+/************************************************
+ * i7300 Functions related to memory enumberation
+ ************************************************/
+
+/**
+ * decode_mtr() - Decodes the MTR descriptor, filling the edac structs
+ * @pvt: pointer to the private data struct used by i7300 driver
+ * @slot: DIMM slot (0 to 7)
+ * @ch: Channel number within the branch (0 or 1)
+ * @branch: Branch number (0 or 1)
+ * @dinfo: Pointer to DIMM info where dimm size is stored
+ * @dimm: Pointer to the struct dimm_info that corresponds to that element
+ */
+static int decode_mtr(struct i7300_pvt *pvt,
+ int slot, int ch, int branch,
+ struct i7300_dimm_info *dinfo,
+ struct dimm_info *dimm)
+{
+ int mtr, ans, addrBits, channel;
+
+ channel = to_channel(ch, branch);
+
+ mtr = pvt->mtr[slot][branch];
+ ans = MTR_DIMMS_PRESENT(mtr) ? 1 : 0;
+
+ edac_dbg(2, "\tMTR%d CH%d: DIMMs are %sPresent (mtr)\n",
+ slot, channel, ans ? "" : "NOT ");
+
+ /* Determine if there is a DIMM present in this DIMM slot */
+ if (!ans)
+ return 0;
+
+ /* Start with the number of bits for a Bank
+ * on the DRAM */
+ addrBits = MTR_DRAM_BANKS_ADDR_BITS;
+ /* Add thenumber of ROW bits */
+ addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
+ /* add the number of COLUMN bits */
+ addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
+ /* add the number of RANK bits */
+ addrBits += MTR_DIMM_RANKS(mtr);
+
+ addrBits += 6; /* add 64 bits per DIMM */
+ addrBits -= 20; /* divide by 2^^20 */
+ addrBits -= 3; /* 8 bits per bytes */
+
+ dinfo->megabytes = 1 << addrBits;
+
+ edac_dbg(2, "\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
+
+ edac_dbg(2, "\t\tELECTRICAL THROTTLING is %s\n",
+ MTR_DIMMS_ETHROTTLE(mtr) ? "enabled" : "disabled");
+
+ edac_dbg(2, "\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
+ edac_dbg(2, "\t\tNUMRANK: %s\n",
+ MTR_DIMM_RANKS(mtr) ? "double" : "single");
+ edac_dbg(2, "\t\tNUMROW: %s\n",
+ MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" :
+ MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" :
+ MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" :
+ "65,536 - 16 rows");
+ edac_dbg(2, "\t\tNUMCOL: %s\n",
+ MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" :
+ MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" :
+ MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" :
+ "reserved");
+ edac_dbg(2, "\t\tSIZE: %d MB\n", dinfo->megabytes);
+
+ /*
+ * The type of error detection actually depends of the
+ * mode of operation. When it is just one single memory chip, at
+ * socket 0, channel 0, it uses 8-byte-over-32-byte SECDED+ code.
+ * In normal or mirrored mode, it uses Lockstep mode,
+ * with the possibility of using an extended algorithm for x8 memories
+ * See datasheet Sections 7.3.6 to 7.3.8
+ */
+
+ dimm->nr_pages = MiB_TO_PAGES(dinfo->megabytes);
+ dimm->grain = 8;
+ dimm->mtype = MEM_FB_DDR2;
+ if (IS_SINGLE_MODE(pvt->mc_settings_a)) {
+ dimm->edac_mode = EDAC_SECDED;
+ edac_dbg(2, "\t\tECC code is 8-byte-over-32-byte SECDED+ code\n");
+ } else {
+ edac_dbg(2, "\t\tECC code is on Lockstep mode\n");
+ if (MTR_DRAM_WIDTH(mtr) == 8)
+ dimm->edac_mode = EDAC_S8ECD8ED;
+ else
+ dimm->edac_mode = EDAC_S4ECD4ED;
+ }
+
+ /* ask what device type on this row */
+ if (MTR_DRAM_WIDTH(mtr) == 8) {
+ edac_dbg(2, "\t\tScrub algorithm for x8 is on %s mode\n",
+ IS_SCRBALGO_ENHANCED(pvt->mc_settings) ?
+ "enhanced" : "normal");
+
+ dimm->dtype = DEV_X8;
+ } else
+ dimm->dtype = DEV_X4;
+
+ return mtr;
+}
+
+/**
+ * print_dimm_size() - Prints dump of the memory organization
+ * @pvt: pointer to the private data struct used by i7300 driver
+ *
+ * Useful for debug. If debug is disabled, this routine do nothing
+ */
+static void print_dimm_size(struct i7300_pvt *pvt)
+{
+#ifdef CONFIG_EDAC_DEBUG
+ struct i7300_dimm_info *dinfo;
+ char *p;
+ int space, n;
+ int channel, slot;
+
+ space = PAGE_SIZE;
+ p = pvt->tmp_prt_buffer;
+
+ n = snprintf(p, space, " ");
+ p += n;
+ space -= n;
+ for (channel = 0; channel < MAX_CHANNELS; channel++) {
+ n = snprintf(p, space, "channel %d | ", channel);
+ p += n;
+ space -= n;
+ }
+ edac_dbg(2, "%s\n", pvt->tmp_prt_buffer);
+ p = pvt->tmp_prt_buffer;
+ space = PAGE_SIZE;
+ n = snprintf(p, space, "-------------------------------"
+ "------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", pvt->tmp_prt_buffer);
+ p = pvt->tmp_prt_buffer;
+ space = PAGE_SIZE;
+
+ for (slot = 0; slot < MAX_SLOTS; slot++) {
+ n = snprintf(p, space, "csrow/SLOT %d ", slot);
+ p += n;
+ space -= n;
+
+ for (channel = 0; channel < MAX_CHANNELS; channel++) {
+ dinfo = &pvt->dimm_info[slot][channel];
+ n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
+ p += n;
+ space -= n;
+ }
+
+ edac_dbg(2, "%s\n", pvt->tmp_prt_buffer);
+ p = pvt->tmp_prt_buffer;
+ space = PAGE_SIZE;
+ }
+
+ n = snprintf(p, space, "-------------------------------"
+ "------------------------------");
+ p += n;
+ space -= n;
+ edac_dbg(2, "%s\n", pvt->tmp_prt_buffer);
+ p = pvt->tmp_prt_buffer;
+ space = PAGE_SIZE;
+#endif
+}
+
+/**
+ * i7300_init_csrows() - Initialize the 'csrows' table within
+ * the mci control structure with the
+ * addressing of memory.
+ * @mci: struct mem_ctl_info pointer
+ */
+static int i7300_init_csrows(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ struct i7300_dimm_info *dinfo;
+ int rc = -ENODEV;
+ int mtr;
+ int ch, branch, slot, channel, max_channel, max_branch;
+ struct dimm_info *dimm;
+
+ pvt = mci->pvt_info;
+
+ edac_dbg(2, "Memory Technology Registers:\n");
+
+ if (IS_SINGLE_MODE(pvt->mc_settings_a)) {
+ max_branch = 1;
+ max_channel = 1;
+ } else {
+ max_branch = MAX_BRANCHES;
+ max_channel = MAX_CH_PER_BRANCH;
+ }
+
+ /* Get the AMB present registers for the four channels */
+ for (branch = 0; branch < max_branch; branch++) {
+ /* Read and dump branch 0's MTRs */
+ channel = to_channel(0, branch);
+ pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch],
+ AMBPRESENT_0,
+ &pvt->ambpresent[channel]);
+ edac_dbg(2, "\t\tAMB-present CH%d = 0x%x:\n",
+ channel, pvt->ambpresent[channel]);
+
+ if (max_channel == 1)
+ continue;
+
+ channel = to_channel(1, branch);
+ pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch],
+ AMBPRESENT_1,
+ &pvt->ambpresent[channel]);
+ edac_dbg(2, "\t\tAMB-present CH%d = 0x%x:\n",
+ channel, pvt->ambpresent[channel]);
+ }
+
+ /* Get the set of MTR[0-7] regs by each branch */
+ for (slot = 0; slot < MAX_SLOTS; slot++) {
+ int where = mtr_regs[slot];
+ for (branch = 0; branch < max_branch; branch++) {
+ pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch],
+ where,
+ &pvt->mtr[slot][branch]);
+ for (ch = 0; ch < max_channel; ch++) {
+ int channel = to_channel(ch, branch);
+
+ dimm = edac_get_dimm(mci, branch, ch, slot);
+
+ dinfo = &pvt->dimm_info[slot][channel];
+
+ mtr = decode_mtr(pvt, slot, ch, branch,
+ dinfo, dimm);
+
+ /* if no DIMMS on this row, continue */
+ if (!MTR_DIMMS_PRESENT(mtr))
+ continue;
+
+ rc = 0;
+
+ }
+ }
+ }
+
+ return rc;
+}
+
+/**
+ * decode_mir() - Decodes Memory Interleave Register (MIR) info
+ * @mir_no: number of the MIR register to decode
+ * @mir: array with the MIR data cached on the driver
+ */
+static void decode_mir(int mir_no, u16 mir[MAX_MIR])
+{
+ if (mir[mir_no] & 3)
+ edac_dbg(2, "MIR%d: limit= 0x%x Branch(es) that participate: %s %s\n",
+ mir_no,
+ (mir[mir_no] >> 4) & 0xfff,
+ (mir[mir_no] & 1) ? "B0" : "",
+ (mir[mir_no] & 2) ? "B1" : "");
+}
+
+/**
+ * i7300_get_mc_regs() - Get the contents of the MC enumeration registers
+ * @mci: struct mem_ctl_info pointer
+ *
+ * Data read is cached internally for its usage when needed
+ */
+static int i7300_get_mc_regs(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ u32 actual_tolm;
+ int i, rc;
+
+ pvt = mci->pvt_info;
+
+ pci_read_config_dword(pvt->pci_dev_16_0_fsb_ctlr, AMBASE,
+ (u32 *) &pvt->ambase);
+
+ edac_dbg(2, "AMBASE= 0x%lx\n", (long unsigned int)pvt->ambase);
+
+ /* Get the Branch Map regs */
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, TOLM, &pvt->tolm);
+ pvt->tolm >>= 12;
+ edac_dbg(2, "TOLM (number of 256M regions) =%u (0x%x)\n",
+ pvt->tolm, pvt->tolm);
+
+ actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28));
+ edac_dbg(2, "Actual TOLM byte addr=%u.%03u GB (0x%x)\n",
+ actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28);
+
+ /* Get memory controller settings */
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, MC_SETTINGS,
+ &pvt->mc_settings);
+ pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, MC_SETTINGS_A,
+ &pvt->mc_settings_a);
+
+ if (IS_SINGLE_MODE(pvt->mc_settings_a))
+ edac_dbg(0, "Memory controller operating on single mode\n");
+ else
+ edac_dbg(0, "Memory controller operating on %smirrored mode\n",
+ IS_MIRRORED(pvt->mc_settings) ? "" : "non-");
+
+ edac_dbg(0, "Error detection is %s\n",
+ IS_ECC_ENABLED(pvt->mc_settings) ? "enabled" : "disabled");
+ edac_dbg(0, "Retry is %s\n",
+ IS_RETRY_ENABLED(pvt->mc_settings) ? "enabled" : "disabled");
+
+ /* Get Memory Interleave Range registers */
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR0,
+ &pvt->mir[0]);
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR1,
+ &pvt->mir[1]);
+ pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR2,
+ &pvt->mir[2]);
+
+ /* Decode the MIR regs */
+ for (i = 0; i < MAX_MIR; i++)
+ decode_mir(i, pvt->mir);
+
+ rc = i7300_init_csrows(mci);
+ if (rc < 0)
+ return rc;
+
+ /* Go and determine the size of each DIMM and place in an
+ * orderly matrix */
+ print_dimm_size(pvt);
+
+ return 0;
+}
+
+/*************************************************
+ * i7300 Functions related to device probe/release
+ *************************************************/
+
+/**
+ * i7300_put_devices() - Release the PCI devices
+ * @mci: struct mem_ctl_info pointer
+ */
+static void i7300_put_devices(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ int branch;
+
+ pvt = mci->pvt_info;
+
+ /* Decrement usage count for devices */
+ for (branch = 0; branch < MAX_CH_PER_BRANCH; branch++)
+ pci_dev_put(pvt->pci_dev_2x_0_fbd_branch[branch]);
+ pci_dev_put(pvt->pci_dev_16_2_fsb_err_regs);
+ pci_dev_put(pvt->pci_dev_16_1_fsb_addr_map);
+}
+
+/**
+ * i7300_get_devices() - Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ * @mci: struct mem_ctl_info pointer
+ *
+ * Access and prepare the several devices for usage:
+ * I7300 devices used by this driver:
+ * Device 16, functions 0,1 and 2: PCI_DEVICE_ID_INTEL_I7300_MCH_ERR
+ * Device 21 function 0: PCI_DEVICE_ID_INTEL_I7300_MCH_FB0
+ * Device 22 function 0: PCI_DEVICE_ID_INTEL_I7300_MCH_FB1
+ */
+static int i7300_get_devices(struct mem_ctl_info *mci)
+{
+ struct i7300_pvt *pvt;
+ struct pci_dev *pdev;
+
+ pvt = mci->pvt_info;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while ((pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7300_MCH_ERR,
+ pdev))) {
+ /* Store device 16 funcs 1 and 2 */
+ switch (PCI_FUNC(pdev->devfn)) {
+ case 1:
+ if (!pvt->pci_dev_16_1_fsb_addr_map)
+ pvt->pci_dev_16_1_fsb_addr_map =
+ pci_dev_get(pdev);
+ break;
+ case 2:
+ if (!pvt->pci_dev_16_2_fsb_err_regs)
+ pvt->pci_dev_16_2_fsb_err_regs =
+ pci_dev_get(pdev);
+ break;
+ }
+ }
+
+ if (!pvt->pci_dev_16_1_fsb_addr_map ||
+ !pvt->pci_dev_16_2_fsb_err_regs) {
+ /* At least one device was not found */
+ i7300_printk(KERN_ERR,
+ "'system address,Process Bus' device not found:"
+ "vendor 0x%x device 0x%x ERR funcs (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7300_MCH_ERR);
+ goto error;
+ }
+
+ edac_dbg(1, "System Address, processor bus- PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->pci_dev_16_0_fsb_ctlr),
+ pvt->pci_dev_16_0_fsb_ctlr->vendor,
+ pvt->pci_dev_16_0_fsb_ctlr->device);
+ edac_dbg(1, "Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->pci_dev_16_1_fsb_addr_map),
+ pvt->pci_dev_16_1_fsb_addr_map->vendor,
+ pvt->pci_dev_16_1_fsb_addr_map->device);
+ edac_dbg(1, "FSB Error Regs - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->pci_dev_16_2_fsb_err_regs),
+ pvt->pci_dev_16_2_fsb_err_regs->vendor,
+ pvt->pci_dev_16_2_fsb_err_regs->device);
+
+ pvt->pci_dev_2x_0_fbd_branch[0] = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7300_MCH_FB0,
+ NULL);
+ if (!pvt->pci_dev_2x_0_fbd_branch[0]) {
+ i7300_printk(KERN_ERR,
+ "MC: 'BRANCH 0' device not found:"
+ "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I7300_MCH_FB0);
+ goto error;
+ }
+
+ pvt->pci_dev_2x_0_fbd_branch[1] = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7300_MCH_FB1,
+ NULL);
+ if (!pvt->pci_dev_2x_0_fbd_branch[1]) {
+ i7300_printk(KERN_ERR,
+ "MC: 'BRANCH 1' device not found:"
+ "vendor 0x%x device 0x%x Func 0 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7300_MCH_FB1);
+ goto error;
+ }
+
+ return 0;
+
+error:
+ i7300_put_devices(mci);
+ return -ENODEV;
+}
+
+/**
+ * i7300_init_one() - Probe for one instance of the device
+ * @pdev: struct pci_dev pointer
+ * @id: struct pci_device_id pointer - currently unused
+ */
+static int i7300_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[3];
+ struct i7300_pvt *pvt;
+ int rc;
+
+ /* wake up device */
+ rc = pci_enable_device(pdev);
+ if (rc == -EIO)
+ return rc;
+
+ edac_dbg(0, "MC: pdev bus %u dev=0x%x fn=0x%x\n",
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+
+ /* We only are looking for func 0 of the set */
+ if (PCI_FUNC(pdev->devfn) != 0)
+ return -ENODEV;
+
+ /* allocate a new MC control structure */
+ layers[0].type = EDAC_MC_LAYER_BRANCH;
+ layers[0].size = MAX_BRANCHES;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = MAX_CH_PER_BRANCH;
+ layers[1].is_virt_csrow = true;
+ layers[2].type = EDAC_MC_LAYER_SLOT;
+ layers[2].size = MAX_SLOTS;
+ layers[2].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p\n", mci);
+
+ mci->pdev = &pdev->dev; /* record ptr to the generic device */
+
+ pvt = mci->pvt_info;
+ pvt->pci_dev_16_0_fsb_ctlr = pdev; /* Record this device in our private */
+
+ pvt->tmp_prt_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
+ if (!pvt->tmp_prt_buffer) {
+ edac_mc_free(mci);
+ return -ENOMEM;
+ }
+
+ /* 'get' the pci devices we want to reserve for our use */
+ if (i7300_get_devices(mci))
+ goto fail0;
+
+ mci->mc_idx = 0;
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i7300_edac.c";
+ mci->ctl_name = i7300_devs[0].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Set the function pointer to an actual operation function */
+ mci->edac_check = i7300_check_error;
+
+ /* initialize the MC control structure 'csrows' table
+ * with the mapping and control information */
+ if (i7300_get_mc_regs(mci)) {
+ edac_dbg(0, "MC: Setting mci->edac_cap to EDAC_FLAG_NONE because i7300_init_csrows() returned nonzero value\n");
+ mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
+ } else {
+ edac_dbg(1, "MC: Enable error reporting now\n");
+ i7300_enable_error_reporting(mci);
+ }
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+ goto fail1;
+ }
+
+ i7300_clear_error(mci);
+
+ /* allocating generic PCI control info */
+ i7300_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i7300_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ return 0;
+
+ /* Error exit unwinding stack */
+fail1:
+
+ i7300_put_devices(mci);
+
+fail0:
+ kfree(pvt->tmp_prt_buffer);
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/**
+ * i7300_remove_one() - Remove the driver
+ * @pdev: struct pci_dev pointer
+ */
+static void i7300_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ char *tmp;
+
+ edac_dbg(0, "\n");
+
+ if (i7300_pci)
+ edac_pci_release_generic_ctl(i7300_pci);
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ tmp = ((struct i7300_pvt *)mci->pvt_info)->tmp_prt_buffer;
+
+ /* retrieve references to resources, and free those resources */
+ i7300_put_devices(mci);
+
+ kfree(tmp);
+ edac_mc_free(mci);
+}
+
+/*
+ * pci_device_id: table for which devices we are looking for
+ *
+ * Has only 8086:360c PCI ID
+ */
+static const struct pci_device_id i7300_pci_tbl[] = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I7300_MCH_ERR)},
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i7300_pci_tbl);
+
+/*
+ * i7300_driver: pci_driver structure for this module
+ */
+static struct pci_driver i7300_driver = {
+ .name = "i7300_edac",
+ .probe = i7300_init_one,
+ .remove = i7300_remove_one,
+ .id_table = i7300_pci_tbl,
+};
+
+/**
+ * i7300_init() - Registers the driver
+ */
+static int __init i7300_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(2, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i7300_driver);
+
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+/**
+ * i7300_exit() - Unregisters the driver
+ */
+static void __exit i7300_exit(void)
+{
+ edac_dbg(2, "\n");
+ pci_unregister_driver(&i7300_driver);
+}
+
+module_init(i7300_init);
+module_exit(i7300_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab");
+MODULE_AUTHOR("Red Hat Inc. (https://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel I7300 memory controllers - "
+ I7300_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i7core_edac.c b/drivers/edac/i7core_edac.c
new file mode 100644
index 0000000000..23d25724ba
--- /dev/null
+++ b/drivers/edac/i7core_edac.c
@@ -0,0 +1,2399 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Intel i7 core/Nehalem Memory Controller kernel module
+ *
+ * This driver supports the memory controllers found on the Intel
+ * processor families i7core, i7core 7xx/8xx, i5core, Xeon 35xx,
+ * Xeon 55xx and Xeon 56xx also known as Nehalem, Nehalem-EP, Lynnfield
+ * and Westmere-EP.
+ *
+ * Copyright (c) 2009-2010 by:
+ * Mauro Carvalho Chehab
+ *
+ * Red Hat Inc. https://www.redhat.com
+ *
+ * Forked and adapted from the i5400_edac driver
+ *
+ * Based on the following public Intel datasheets:
+ * Intel Core i7 Processor Extreme Edition and Intel Core i7 Processor
+ * Datasheet, Volume 2:
+ * http://download.intel.com/design/processor/datashts/320835.pdf
+ * Intel Xeon Processor 5500 Series Datasheet Volume 2
+ * http://www.intel.com/Assets/PDF/datasheet/321322.pdf
+ * also available at:
+ * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/dmi.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+#include <linux/smp.h>
+#include <asm/mce.h>
+#include <asm/processor.h>
+#include <asm/div64.h>
+
+#include "edac_module.h"
+
+/* Static vars */
+static LIST_HEAD(i7core_edac_list);
+static DEFINE_MUTEX(i7core_edac_lock);
+static int probed;
+
+static int use_pci_fixup;
+module_param(use_pci_fixup, int, 0444);
+MODULE_PARM_DESC(use_pci_fixup, "Enable PCI fixup to seek for hidden devices");
+/*
+ * This is used for Nehalem-EP and Nehalem-EX devices, where the non-core
+ * registers start at bus 255, and are not reported by BIOS.
+ * We currently find devices with only 2 sockets. In order to support more QPI
+ * Quick Path Interconnect, just increment this number.
+ */
+#define MAX_SOCKET_BUSES 2
+
+
+/*
+ * Alter this version for the module when modifications are made
+ */
+#define I7CORE_REVISION " Ver: 1.0.0"
+#define EDAC_MOD_STR "i7core_edac"
+
+/*
+ * Debug macros
+ */
+#define i7core_printk(level, fmt, arg...) \
+ edac_printk(level, "i7core", fmt, ##arg)
+
+#define i7core_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i7core", fmt, ##arg)
+
+/*
+ * i7core Memory Controller Registers
+ */
+
+ /* OFFSETS for Device 0 Function 0 */
+
+#define MC_CFG_CONTROL 0x90
+ #define MC_CFG_UNLOCK 0x02
+ #define MC_CFG_LOCK 0x00
+
+ /* OFFSETS for Device 3 Function 0 */
+
+#define MC_CONTROL 0x48
+#define MC_STATUS 0x4c
+#define MC_MAX_DOD 0x64
+
+/*
+ * OFFSETS for Device 3 Function 4, as indicated on Xeon 5500 datasheet:
+ * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+
+#define MC_TEST_ERR_RCV1 0x60
+ #define DIMM2_COR_ERR(r) ((r) & 0x7fff)
+
+#define MC_TEST_ERR_RCV0 0x64
+ #define DIMM1_COR_ERR(r) (((r) >> 16) & 0x7fff)
+ #define DIMM0_COR_ERR(r) ((r) & 0x7fff)
+
+/* OFFSETS for Device 3 Function 2, as indicated on Xeon 5500 datasheet */
+#define MC_SSRCONTROL 0x48
+ #define SSR_MODE_DISABLE 0x00
+ #define SSR_MODE_ENABLE 0x01
+ #define SSR_MODE_MASK 0x03
+
+#define MC_SCRUB_CONTROL 0x4c
+ #define STARTSCRUB (1 << 24)
+ #define SCRUBINTERVAL_MASK 0xffffff
+
+#define MC_COR_ECC_CNT_0 0x80
+#define MC_COR_ECC_CNT_1 0x84
+#define MC_COR_ECC_CNT_2 0x88
+#define MC_COR_ECC_CNT_3 0x8c
+#define MC_COR_ECC_CNT_4 0x90
+#define MC_COR_ECC_CNT_5 0x94
+
+#define DIMM_TOP_COR_ERR(r) (((r) >> 16) & 0x7fff)
+#define DIMM_BOT_COR_ERR(r) ((r) & 0x7fff)
+
+
+ /* OFFSETS for Devices 4,5 and 6 Function 0 */
+
+#define MC_CHANNEL_DIMM_INIT_PARAMS 0x58
+ #define THREE_DIMMS_PRESENT (1 << 24)
+ #define SINGLE_QUAD_RANK_PRESENT (1 << 23)
+ #define QUAD_RANK_PRESENT (1 << 22)
+ #define REGISTERED_DIMM (1 << 15)
+
+#define MC_CHANNEL_MAPPER 0x60
+ #define RDLCH(r, ch) ((((r) >> (3 + (ch * 6))) & 0x07) - 1)
+ #define WRLCH(r, ch) ((((r) >> (ch * 6)) & 0x07) - 1)
+
+#define MC_CHANNEL_RANK_PRESENT 0x7c
+ #define RANK_PRESENT_MASK 0xffff
+
+#define MC_CHANNEL_ADDR_MATCH 0xf0
+#define MC_CHANNEL_ERROR_MASK 0xf8
+#define MC_CHANNEL_ERROR_INJECT 0xfc
+ #define INJECT_ADDR_PARITY 0x10
+ #define INJECT_ECC 0x08
+ #define MASK_CACHELINE 0x06
+ #define MASK_FULL_CACHELINE 0x06
+ #define MASK_MSB32_CACHELINE 0x04
+ #define MASK_LSB32_CACHELINE 0x02
+ #define NO_MASK_CACHELINE 0x00
+ #define REPEAT_EN 0x01
+
+ /* OFFSETS for Devices 4,5 and 6 Function 1 */
+
+#define MC_DOD_CH_DIMM0 0x48
+#define MC_DOD_CH_DIMM1 0x4c
+#define MC_DOD_CH_DIMM2 0x50
+ #define RANKOFFSET_MASK ((1 << 12) | (1 << 11) | (1 << 10))
+ #define RANKOFFSET(x) ((x & RANKOFFSET_MASK) >> 10)
+ #define DIMM_PRESENT_MASK (1 << 9)
+ #define DIMM_PRESENT(x) (((x) & DIMM_PRESENT_MASK) >> 9)
+ #define MC_DOD_NUMBANK_MASK ((1 << 8) | (1 << 7))
+ #define MC_DOD_NUMBANK(x) (((x) & MC_DOD_NUMBANK_MASK) >> 7)
+ #define MC_DOD_NUMRANK_MASK ((1 << 6) | (1 << 5))
+ #define MC_DOD_NUMRANK(x) (((x) & MC_DOD_NUMRANK_MASK) >> 5)
+ #define MC_DOD_NUMROW_MASK ((1 << 4) | (1 << 3) | (1 << 2))
+ #define MC_DOD_NUMROW(x) (((x) & MC_DOD_NUMROW_MASK) >> 2)
+ #define MC_DOD_NUMCOL_MASK 3
+ #define MC_DOD_NUMCOL(x) ((x) & MC_DOD_NUMCOL_MASK)
+
+#define MC_RANK_PRESENT 0x7c
+
+#define MC_SAG_CH_0 0x80
+#define MC_SAG_CH_1 0x84
+#define MC_SAG_CH_2 0x88
+#define MC_SAG_CH_3 0x8c
+#define MC_SAG_CH_4 0x90
+#define MC_SAG_CH_5 0x94
+#define MC_SAG_CH_6 0x98
+#define MC_SAG_CH_7 0x9c
+
+#define MC_RIR_LIMIT_CH_0 0x40
+#define MC_RIR_LIMIT_CH_1 0x44
+#define MC_RIR_LIMIT_CH_2 0x48
+#define MC_RIR_LIMIT_CH_3 0x4C
+#define MC_RIR_LIMIT_CH_4 0x50
+#define MC_RIR_LIMIT_CH_5 0x54
+#define MC_RIR_LIMIT_CH_6 0x58
+#define MC_RIR_LIMIT_CH_7 0x5C
+#define MC_RIR_LIMIT_MASK ((1 << 10) - 1)
+
+#define MC_RIR_WAY_CH 0x80
+ #define MC_RIR_WAY_OFFSET_MASK (((1 << 14) - 1) & ~0x7)
+ #define MC_RIR_WAY_RANK_MASK 0x7
+
+/*
+ * i7core structs
+ */
+
+#define NUM_CHANS 3
+#define MAX_DIMMS 3 /* Max DIMMS per channel */
+#define MAX_MCR_FUNC 4
+#define MAX_CHAN_FUNC 3
+
+struct i7core_info {
+ u32 mc_control;
+ u32 mc_status;
+ u32 max_dod;
+ u32 ch_map;
+};
+
+
+struct i7core_inject {
+ int enable;
+
+ u32 section;
+ u32 type;
+ u32 eccmask;
+
+ /* Error address mask */
+ int channel, dimm, rank, bank, page, col;
+};
+
+struct i7core_channel {
+ bool is_3dimms_present;
+ bool is_single_4rank;
+ bool has_4rank;
+ u32 dimms;
+};
+
+struct pci_id_descr {
+ int dev;
+ int func;
+ int dev_id;
+ int optional;
+};
+
+struct pci_id_table {
+ const struct pci_id_descr *descr;
+ int n_devs;
+};
+
+struct i7core_dev {
+ struct list_head list;
+ u8 socket;
+ struct pci_dev **pdev;
+ int n_devs;
+ struct mem_ctl_info *mci;
+};
+
+struct i7core_pvt {
+ struct device *addrmatch_dev, *chancounts_dev;
+
+ struct pci_dev *pci_noncore;
+ struct pci_dev *pci_mcr[MAX_MCR_FUNC + 1];
+ struct pci_dev *pci_ch[NUM_CHANS][MAX_CHAN_FUNC + 1];
+
+ struct i7core_dev *i7core_dev;
+
+ struct i7core_info info;
+ struct i7core_inject inject;
+ struct i7core_channel channel[NUM_CHANS];
+
+ int ce_count_available;
+
+ /* ECC corrected errors counts per udimm */
+ unsigned long udimm_ce_count[MAX_DIMMS];
+ int udimm_last_ce_count[MAX_DIMMS];
+ /* ECC corrected errors counts per rdimm */
+ unsigned long rdimm_ce_count[NUM_CHANS][MAX_DIMMS];
+ int rdimm_last_ce_count[NUM_CHANS][MAX_DIMMS];
+
+ bool is_registered, enable_scrub;
+
+ /* DCLK Frequency used for computing scrub rate */
+ int dclk_freq;
+
+ /* Struct to control EDAC polling */
+ struct edac_pci_ctl_info *i7core_pci;
+};
+
+#define PCI_DESCR(device, function, device_id) \
+ .dev = (device), \
+ .func = (function), \
+ .dev_id = (device_id)
+
+static const struct pci_id_descr pci_dev_descr_i7core_nehalem[] = {
+ /* Memory controller */
+ { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_I7_MCR) },
+ { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_I7_MC_TAD) },
+ /* Exists only for RDIMM */
+ { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_I7_MC_RAS), .optional = 1 },
+ { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_I7_MC_TEST) },
+
+ /* Channel 0 */
+ { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH0_CTRL) },
+ { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH0_ADDR) },
+ { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH0_RANK) },
+ { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH0_TC) },
+
+ /* Channel 1 */
+ { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH1_CTRL) },
+ { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH1_ADDR) },
+ { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH1_RANK) },
+ { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH1_TC) },
+
+ /* Channel 2 */
+ { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH2_CTRL) },
+ { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH2_ADDR) },
+ { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH2_RANK) },
+ { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH2_TC) },
+
+ /* Generic Non-core registers */
+ /*
+ * This is the PCI device on i7core and on Xeon 35xx (8086:2c41)
+ * On Xeon 55xx, however, it has a different id (8086:2c40). So,
+ * the probing code needs to test for the other address in case of
+ * failure of this one
+ */
+ { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_I7_NONCORE) },
+
+};
+
+static const struct pci_id_descr pci_dev_descr_lynnfield[] = {
+ { PCI_DESCR( 3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR) },
+ { PCI_DESCR( 3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD) },
+ { PCI_DESCR( 3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST) },
+
+ { PCI_DESCR( 4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL) },
+ { PCI_DESCR( 4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR) },
+ { PCI_DESCR( 4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK) },
+ { PCI_DESCR( 4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC) },
+
+ { PCI_DESCR( 5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL) },
+ { PCI_DESCR( 5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR) },
+ { PCI_DESCR( 5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK) },
+ { PCI_DESCR( 5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC) },
+
+ /*
+ * This is the PCI device has an alternate address on some
+ * processors like Core i7 860
+ */
+ { PCI_DESCR( 0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE) },
+};
+
+static const struct pci_id_descr pci_dev_descr_i7core_westmere[] = {
+ /* Memory controller */
+ { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR_REV2) },
+ { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD_REV2) },
+ /* Exists only for RDIMM */
+ { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_RAS_REV2), .optional = 1 },
+ { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST_REV2) },
+
+ /* Channel 0 */
+ { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL_REV2) },
+ { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR_REV2) },
+ { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK_REV2) },
+ { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC_REV2) },
+
+ /* Channel 1 */
+ { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL_REV2) },
+ { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR_REV2) },
+ { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK_REV2) },
+ { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC_REV2) },
+
+ /* Channel 2 */
+ { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_CTRL_REV2) },
+ { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_ADDR_REV2) },
+ { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_RANK_REV2) },
+ { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_TC_REV2) },
+
+ /* Generic Non-core registers */
+ { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2) },
+
+};
+
+#define PCI_ID_TABLE_ENTRY(A) { .descr=A, .n_devs = ARRAY_SIZE(A) }
+static const struct pci_id_table pci_dev_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_nehalem),
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_lynnfield),
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_westmere),
+ {0,} /* 0 terminated list. */
+};
+
+/*
+ * pci_device_id table for which devices we are looking for
+ */
+static const struct pci_device_id i7core_pci_tbl[] = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_X58_HUB_MGMT)},
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0)},
+ {0,} /* 0 terminated list. */
+};
+
+/****************************************************************************
+ Ancillary status routines
+ ****************************************************************************/
+
+ /* MC_CONTROL bits */
+#define CH_ACTIVE(pvt, ch) ((pvt)->info.mc_control & (1 << (8 + ch)))
+#define ECCx8(pvt) ((pvt)->info.mc_control & (1 << 1))
+
+ /* MC_STATUS bits */
+#define ECC_ENABLED(pvt) ((pvt)->info.mc_status & (1 << 4))
+#define CH_DISABLED(pvt, ch) ((pvt)->info.mc_status & (1 << ch))
+
+ /* MC_MAX_DOD read functions */
+static inline int numdimms(u32 dimms)
+{
+ return (dimms & 0x3) + 1;
+}
+
+static inline int numrank(u32 rank)
+{
+ static const int ranks[] = { 1, 2, 4, -EINVAL };
+
+ return ranks[rank & 0x3];
+}
+
+static inline int numbank(u32 bank)
+{
+ static const int banks[] = { 4, 8, 16, -EINVAL };
+
+ return banks[bank & 0x3];
+}
+
+static inline int numrow(u32 row)
+{
+ static const int rows[] = {
+ 1 << 12, 1 << 13, 1 << 14, 1 << 15,
+ 1 << 16, -EINVAL, -EINVAL, -EINVAL,
+ };
+
+ return rows[row & 0x7];
+}
+
+static inline int numcol(u32 col)
+{
+ static const int cols[] = {
+ 1 << 10, 1 << 11, 1 << 12, -EINVAL,
+ };
+ return cols[col & 0x3];
+}
+
+static struct i7core_dev *get_i7core_dev(u8 socket)
+{
+ struct i7core_dev *i7core_dev;
+
+ list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
+ if (i7core_dev->socket == socket)
+ return i7core_dev;
+ }
+
+ return NULL;
+}
+
+static struct i7core_dev *alloc_i7core_dev(u8 socket,
+ const struct pci_id_table *table)
+{
+ struct i7core_dev *i7core_dev;
+
+ i7core_dev = kzalloc(sizeof(*i7core_dev), GFP_KERNEL);
+ if (!i7core_dev)
+ return NULL;
+
+ i7core_dev->pdev = kcalloc(table->n_devs, sizeof(*i7core_dev->pdev),
+ GFP_KERNEL);
+ if (!i7core_dev->pdev) {
+ kfree(i7core_dev);
+ return NULL;
+ }
+
+ i7core_dev->socket = socket;
+ i7core_dev->n_devs = table->n_devs;
+ list_add_tail(&i7core_dev->list, &i7core_edac_list);
+
+ return i7core_dev;
+}
+
+static void free_i7core_dev(struct i7core_dev *i7core_dev)
+{
+ list_del(&i7core_dev->list);
+ kfree(i7core_dev->pdev);
+ kfree(i7core_dev);
+}
+
+/****************************************************************************
+ Memory check routines
+ ****************************************************************************/
+
+static int get_dimm_config(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ int i, j;
+ enum edac_type mode;
+ enum mem_type mtype;
+ struct dimm_info *dimm;
+
+ /* Get data from the MC register, function 0 */
+ pdev = pvt->pci_mcr[0];
+ if (!pdev)
+ return -ENODEV;
+
+ /* Device 3 function 0 reads */
+ pci_read_config_dword(pdev, MC_CONTROL, &pvt->info.mc_control);
+ pci_read_config_dword(pdev, MC_STATUS, &pvt->info.mc_status);
+ pci_read_config_dword(pdev, MC_MAX_DOD, &pvt->info.max_dod);
+ pci_read_config_dword(pdev, MC_CHANNEL_MAPPER, &pvt->info.ch_map);
+
+ edac_dbg(0, "QPI %d control=0x%08x status=0x%08x dod=0x%08x map=0x%08x\n",
+ pvt->i7core_dev->socket, pvt->info.mc_control,
+ pvt->info.mc_status, pvt->info.max_dod, pvt->info.ch_map);
+
+ if (ECC_ENABLED(pvt)) {
+ edac_dbg(0, "ECC enabled with x%d SDCC\n", ECCx8(pvt) ? 8 : 4);
+ if (ECCx8(pvt))
+ mode = EDAC_S8ECD8ED;
+ else
+ mode = EDAC_S4ECD4ED;
+ } else {
+ edac_dbg(0, "ECC disabled\n");
+ mode = EDAC_NONE;
+ }
+
+ /* FIXME: need to handle the error codes */
+ edac_dbg(0, "DOD Max limits: DIMMS: %d, %d-ranked, %d-banked x%x x 0x%x\n",
+ numdimms(pvt->info.max_dod),
+ numrank(pvt->info.max_dod >> 2),
+ numbank(pvt->info.max_dod >> 4),
+ numrow(pvt->info.max_dod >> 6),
+ numcol(pvt->info.max_dod >> 9));
+
+ for (i = 0; i < NUM_CHANS; i++) {
+ u32 data, dimm_dod[3], value[8];
+
+ if (!pvt->pci_ch[i][0])
+ continue;
+
+ if (!CH_ACTIVE(pvt, i)) {
+ edac_dbg(0, "Channel %i is not active\n", i);
+ continue;
+ }
+ if (CH_DISABLED(pvt, i)) {
+ edac_dbg(0, "Channel %i is disabled\n", i);
+ continue;
+ }
+
+ /* Devices 4-6 function 0 */
+ pci_read_config_dword(pvt->pci_ch[i][0],
+ MC_CHANNEL_DIMM_INIT_PARAMS, &data);
+
+
+ if (data & THREE_DIMMS_PRESENT)
+ pvt->channel[i].is_3dimms_present = true;
+
+ if (data & SINGLE_QUAD_RANK_PRESENT)
+ pvt->channel[i].is_single_4rank = true;
+
+ if (data & QUAD_RANK_PRESENT)
+ pvt->channel[i].has_4rank = true;
+
+ if (data & REGISTERED_DIMM)
+ mtype = MEM_RDDR3;
+ else
+ mtype = MEM_DDR3;
+
+ /* Devices 4-6 function 1 */
+ pci_read_config_dword(pvt->pci_ch[i][1],
+ MC_DOD_CH_DIMM0, &dimm_dod[0]);
+ pci_read_config_dword(pvt->pci_ch[i][1],
+ MC_DOD_CH_DIMM1, &dimm_dod[1]);
+ pci_read_config_dword(pvt->pci_ch[i][1],
+ MC_DOD_CH_DIMM2, &dimm_dod[2]);
+
+ edac_dbg(0, "Ch%d phy rd%d, wr%d (0x%08x): %s%s%s%cDIMMs\n",
+ i,
+ RDLCH(pvt->info.ch_map, i), WRLCH(pvt->info.ch_map, i),
+ data,
+ pvt->channel[i].is_3dimms_present ? "3DIMMS " : "",
+ pvt->channel[i].is_3dimms_present ? "SINGLE_4R " : "",
+ pvt->channel[i].has_4rank ? "HAS_4R " : "",
+ (data & REGISTERED_DIMM) ? 'R' : 'U');
+
+ for (j = 0; j < 3; j++) {
+ u32 banks, ranks, rows, cols;
+ u32 size, npages;
+
+ if (!DIMM_PRESENT(dimm_dod[j]))
+ continue;
+
+ dimm = edac_get_dimm(mci, i, j, 0);
+ banks = numbank(MC_DOD_NUMBANK(dimm_dod[j]));
+ ranks = numrank(MC_DOD_NUMRANK(dimm_dod[j]));
+ rows = numrow(MC_DOD_NUMROW(dimm_dod[j]));
+ cols = numcol(MC_DOD_NUMCOL(dimm_dod[j]));
+
+ /* DDR3 has 8 I/O banks */
+ size = (rows * cols * banks * ranks) >> (20 - 3);
+
+ edac_dbg(0, "\tdimm %d %d MiB offset: %x, bank: %d, rank: %d, row: %#x, col: %#x\n",
+ j, size,
+ RANKOFFSET(dimm_dod[j]),
+ banks, ranks, rows, cols);
+
+ npages = MiB_TO_PAGES(size);
+
+ dimm->nr_pages = npages;
+
+ switch (banks) {
+ case 4:
+ dimm->dtype = DEV_X4;
+ break;
+ case 8:
+ dimm->dtype = DEV_X8;
+ break;
+ case 16:
+ dimm->dtype = DEV_X16;
+ break;
+ default:
+ dimm->dtype = DEV_UNKNOWN;
+ }
+
+ snprintf(dimm->label, sizeof(dimm->label),
+ "CPU#%uChannel#%u_DIMM#%u",
+ pvt->i7core_dev->socket, i, j);
+ dimm->grain = 8;
+ dimm->edac_mode = mode;
+ dimm->mtype = mtype;
+ }
+
+ pci_read_config_dword(pdev, MC_SAG_CH_0, &value[0]);
+ pci_read_config_dword(pdev, MC_SAG_CH_1, &value[1]);
+ pci_read_config_dword(pdev, MC_SAG_CH_2, &value[2]);
+ pci_read_config_dword(pdev, MC_SAG_CH_3, &value[3]);
+ pci_read_config_dword(pdev, MC_SAG_CH_4, &value[4]);
+ pci_read_config_dword(pdev, MC_SAG_CH_5, &value[5]);
+ pci_read_config_dword(pdev, MC_SAG_CH_6, &value[6]);
+ pci_read_config_dword(pdev, MC_SAG_CH_7, &value[7]);
+ edac_dbg(1, "\t[%i] DIVBY3\tREMOVED\tOFFSET\n", i);
+ for (j = 0; j < 8; j++)
+ edac_dbg(1, "\t\t%#x\t%#x\t%#x\n",
+ (value[j] >> 27) & 0x1,
+ (value[j] >> 24) & 0x7,
+ (value[j] & ((1 << 24) - 1)));
+ }
+
+ return 0;
+}
+
+/****************************************************************************
+ Error insertion routines
+ ****************************************************************************/
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+/* The i7core has independent error injection features per channel.
+ However, to have a simpler code, we don't allow enabling error injection
+ on more than one channel.
+ Also, since a change at an inject parameter will be applied only at enable,
+ we're disabling error injection on all write calls to the sysfs nodes that
+ controls the error code injection.
+ */
+static int disable_inject(const struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+
+ pvt->inject.enable = 0;
+
+ if (!pvt->pci_ch[pvt->inject.channel][0])
+ return -ENODEV;
+
+ pci_write_config_dword(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ERROR_INJECT, 0);
+
+ return 0;
+}
+
+/*
+ * i7core inject inject.section
+ *
+ * accept and store error injection inject.section value
+ * bit 0 - refers to the lower 32-byte half cacheline
+ * bit 1 - refers to the upper 32-byte half cacheline
+ */
+static ssize_t i7core_inject_section_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int rc;
+
+ if (pvt->inject.enable)
+ disable_inject(mci);
+
+ rc = kstrtoul(data, 10, &value);
+ if ((rc < 0) || (value > 3))
+ return -EIO;
+
+ pvt->inject.section = (u32) value;
+ return count;
+}
+
+static ssize_t i7core_inject_section_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ return sprintf(data, "0x%08x\n", pvt->inject.section);
+}
+
+/*
+ * i7core inject.type
+ *
+ * accept and store error injection inject.section value
+ * bit 0 - repeat enable - Enable error repetition
+ * bit 1 - inject ECC error
+ * bit 2 - inject parity error
+ */
+static ssize_t i7core_inject_type_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int rc;
+
+ if (pvt->inject.enable)
+ disable_inject(mci);
+
+ rc = kstrtoul(data, 10, &value);
+ if ((rc < 0) || (value > 7))
+ return -EIO;
+
+ pvt->inject.type = (u32) value;
+ return count;
+}
+
+static ssize_t i7core_inject_type_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+
+ return sprintf(data, "0x%08x\n", pvt->inject.type);
+}
+
+/*
+ * i7core_inject_inject.eccmask_store
+ *
+ * The type of error (UE/CE) will depend on the inject.eccmask value:
+ * Any bits set to a 1 will flip the corresponding ECC bit
+ * Correctable errors can be injected by flipping 1 bit or the bits within
+ * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
+ * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an
+ * uncorrectable error to be injected.
+ */
+static ssize_t i7core_inject_eccmask_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ unsigned long value;
+ int rc;
+
+ if (pvt->inject.enable)
+ disable_inject(mci);
+
+ rc = kstrtoul(data, 10, &value);
+ if (rc < 0)
+ return -EIO;
+
+ pvt->inject.eccmask = (u32) value;
+ return count;
+}
+
+static ssize_t i7core_inject_eccmask_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+
+ return sprintf(data, "0x%08x\n", pvt->inject.eccmask);
+}
+
+/*
+ * i7core_addrmatch
+ *
+ * The type of error (UE/CE) will depend on the inject.eccmask value:
+ * Any bits set to a 1 will flip the corresponding ECC bit
+ * Correctable errors can be injected by flipping 1 bit or the bits within
+ * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
+ * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an
+ * uncorrectable error to be injected.
+ */
+
+#define DECLARE_ADDR_MATCH(param, limit) \
+static ssize_t i7core_inject_store_##param( \
+ struct device *dev, \
+ struct device_attribute *mattr, \
+ const char *data, size_t count) \
+{ \
+ struct mem_ctl_info *mci = dev_get_drvdata(dev); \
+ struct i7core_pvt *pvt; \
+ long value; \
+ int rc; \
+ \
+ edac_dbg(1, "\n"); \
+ pvt = mci->pvt_info; \
+ \
+ if (pvt->inject.enable) \
+ disable_inject(mci); \
+ \
+ if (!strcasecmp(data, "any") || !strcasecmp(data, "any\n"))\
+ value = -1; \
+ else { \
+ rc = kstrtoul(data, 10, &value); \
+ if ((rc < 0) || (value >= limit)) \
+ return -EIO; \
+ } \
+ \
+ pvt->inject.param = value; \
+ \
+ return count; \
+} \
+ \
+static ssize_t i7core_inject_show_##param( \
+ struct device *dev, \
+ struct device_attribute *mattr, \
+ char *data) \
+{ \
+ struct mem_ctl_info *mci = dev_get_drvdata(dev); \
+ struct i7core_pvt *pvt; \
+ \
+ pvt = mci->pvt_info; \
+ edac_dbg(1, "pvt=%p\n", pvt); \
+ if (pvt->inject.param < 0) \
+ return sprintf(data, "any\n"); \
+ else \
+ return sprintf(data, "%d\n", pvt->inject.param);\
+}
+
+#define ATTR_ADDR_MATCH(param) \
+ static DEVICE_ATTR(param, S_IRUGO | S_IWUSR, \
+ i7core_inject_show_##param, \
+ i7core_inject_store_##param)
+
+DECLARE_ADDR_MATCH(channel, 3);
+DECLARE_ADDR_MATCH(dimm, 3);
+DECLARE_ADDR_MATCH(rank, 4);
+DECLARE_ADDR_MATCH(bank, 32);
+DECLARE_ADDR_MATCH(page, 0x10000);
+DECLARE_ADDR_MATCH(col, 0x4000);
+
+ATTR_ADDR_MATCH(channel);
+ATTR_ADDR_MATCH(dimm);
+ATTR_ADDR_MATCH(rank);
+ATTR_ADDR_MATCH(bank);
+ATTR_ADDR_MATCH(page);
+ATTR_ADDR_MATCH(col);
+
+static int write_and_test(struct pci_dev *dev, const int where, const u32 val)
+{
+ u32 read;
+ int count;
+
+ edac_dbg(0, "setting pci %02x:%02x.%x reg=%02x value=%08x\n",
+ dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
+ where, val);
+
+ for (count = 0; count < 10; count++) {
+ if (count)
+ msleep(100);
+ pci_write_config_dword(dev, where, val);
+ pci_read_config_dword(dev, where, &read);
+
+ if (read == val)
+ return 0;
+ }
+
+ i7core_printk(KERN_ERR, "Error during set pci %02x:%02x.%x reg=%02x "
+ "write=%08x. Read=%08x\n",
+ dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
+ where, val, read);
+
+ return -EINVAL;
+}
+
+/*
+ * This routine prepares the Memory Controller for error injection.
+ * The error will be injected when some process tries to write to the
+ * memory that matches the given criteria.
+ * The criteria can be set in terms of a mask where dimm, rank, bank, page
+ * and col can be specified.
+ * A -1 value for any of the mask items will make the MCU to ignore
+ * that matching criteria for error injection.
+ *
+ * It should be noticed that the error will only happen after a write operation
+ * on a memory that matches the condition. if REPEAT_EN is not enabled at
+ * inject mask, then it will produce just one error. Otherwise, it will repeat
+ * until the injectmask would be cleaned.
+ *
+ * FIXME: This routine assumes that MAXNUMDIMMS value of MC_MAX_DOD
+ * is reliable enough to check if the MC is using the
+ * three channels. However, this is not clear at the datasheet.
+ */
+static ssize_t i7core_inject_enable_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 injectmask;
+ u64 mask = 0;
+ int rc;
+ long enable;
+
+ if (!pvt->pci_ch[pvt->inject.channel][0])
+ return 0;
+
+ rc = kstrtoul(data, 10, &enable);
+ if ((rc < 0))
+ return 0;
+
+ if (enable) {
+ pvt->inject.enable = 1;
+ } else {
+ disable_inject(mci);
+ return count;
+ }
+
+ /* Sets pvt->inject.dimm mask */
+ if (pvt->inject.dimm < 0)
+ mask |= 1LL << 41;
+ else {
+ if (pvt->channel[pvt->inject.channel].dimms > 2)
+ mask |= (pvt->inject.dimm & 0x3LL) << 35;
+ else
+ mask |= (pvt->inject.dimm & 0x1LL) << 36;
+ }
+
+ /* Sets pvt->inject.rank mask */
+ if (pvt->inject.rank < 0)
+ mask |= 1LL << 40;
+ else {
+ if (pvt->channel[pvt->inject.channel].dimms > 2)
+ mask |= (pvt->inject.rank & 0x1LL) << 34;
+ else
+ mask |= (pvt->inject.rank & 0x3LL) << 34;
+ }
+
+ /* Sets pvt->inject.bank mask */
+ if (pvt->inject.bank < 0)
+ mask |= 1LL << 39;
+ else
+ mask |= (pvt->inject.bank & 0x15LL) << 30;
+
+ /* Sets pvt->inject.page mask */
+ if (pvt->inject.page < 0)
+ mask |= 1LL << 38;
+ else
+ mask |= (pvt->inject.page & 0xffff) << 14;
+
+ /* Sets pvt->inject.column mask */
+ if (pvt->inject.col < 0)
+ mask |= 1LL << 37;
+ else
+ mask |= (pvt->inject.col & 0x3fff);
+
+ /*
+ * bit 0: REPEAT_EN
+ * bits 1-2: MASK_HALF_CACHELINE
+ * bit 3: INJECT_ECC
+ * bit 4: INJECT_ADDR_PARITY
+ */
+
+ injectmask = (pvt->inject.type & 1) |
+ (pvt->inject.section & 0x3) << 1 |
+ (pvt->inject.type & 0x6) << (3 - 1);
+
+ /* Unlock writes to registers - this register is write only */
+ pci_write_config_dword(pvt->pci_noncore,
+ MC_CFG_CONTROL, 0x2);
+
+ write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ADDR_MATCH, mask);
+ write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ADDR_MATCH + 4, mask >> 32L);
+
+ write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ERROR_MASK, pvt->inject.eccmask);
+
+ write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ERROR_INJECT, injectmask);
+
+ /*
+ * This is something undocumented, based on my tests
+ * Without writing 8 to this register, errors aren't injected. Not sure
+ * why.
+ */
+ pci_write_config_dword(pvt->pci_noncore,
+ MC_CFG_CONTROL, 8);
+
+ edac_dbg(0, "Error inject addr match 0x%016llx, ecc 0x%08x, inject 0x%08x\n",
+ mask, pvt->inject.eccmask, injectmask);
+
+
+ return count;
+}
+
+static ssize_t i7core_inject_enable_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 injectmask;
+
+ if (!pvt->pci_ch[pvt->inject.channel][0])
+ return 0;
+
+ pci_read_config_dword(pvt->pci_ch[pvt->inject.channel][0],
+ MC_CHANNEL_ERROR_INJECT, &injectmask);
+
+ edac_dbg(0, "Inject error read: 0x%018x\n", injectmask);
+
+ if (injectmask & 0x0c)
+ pvt->inject.enable = 1;
+
+ return sprintf(data, "%d\n", pvt->inject.enable);
+}
+
+#define DECLARE_COUNTER(param) \
+static ssize_t i7core_show_counter_##param( \
+ struct device *dev, \
+ struct device_attribute *mattr, \
+ char *data) \
+{ \
+ struct mem_ctl_info *mci = dev_get_drvdata(dev); \
+ struct i7core_pvt *pvt = mci->pvt_info; \
+ \
+ edac_dbg(1, "\n"); \
+ if (!pvt->ce_count_available || (pvt->is_registered)) \
+ return sprintf(data, "data unavailable\n"); \
+ return sprintf(data, "%lu\n", \
+ pvt->udimm_ce_count[param]); \
+}
+
+#define ATTR_COUNTER(param) \
+ static DEVICE_ATTR(udimm##param, S_IRUGO | S_IWUSR, \
+ i7core_show_counter_##param, \
+ NULL)
+
+DECLARE_COUNTER(0);
+DECLARE_COUNTER(1);
+DECLARE_COUNTER(2);
+
+ATTR_COUNTER(0);
+ATTR_COUNTER(1);
+ATTR_COUNTER(2);
+
+/*
+ * inject_addrmatch device sysfs struct
+ */
+
+static struct attribute *i7core_addrmatch_attrs[] = {
+ &dev_attr_channel.attr,
+ &dev_attr_dimm.attr,
+ &dev_attr_rank.attr,
+ &dev_attr_bank.attr,
+ &dev_attr_page.attr,
+ &dev_attr_col.attr,
+ NULL
+};
+
+static const struct attribute_group addrmatch_grp = {
+ .attrs = i7core_addrmatch_attrs,
+};
+
+static const struct attribute_group *addrmatch_groups[] = {
+ &addrmatch_grp,
+ NULL
+};
+
+static void addrmatch_release(struct device *device)
+{
+ edac_dbg(1, "Releasing device %s\n", dev_name(device));
+ kfree(device);
+}
+
+static const struct device_type addrmatch_type = {
+ .groups = addrmatch_groups,
+ .release = addrmatch_release,
+};
+
+/*
+ * all_channel_counts sysfs struct
+ */
+
+static struct attribute *i7core_udimm_counters_attrs[] = {
+ &dev_attr_udimm0.attr,
+ &dev_attr_udimm1.attr,
+ &dev_attr_udimm2.attr,
+ NULL
+};
+
+static const struct attribute_group all_channel_counts_grp = {
+ .attrs = i7core_udimm_counters_attrs,
+};
+
+static const struct attribute_group *all_channel_counts_groups[] = {
+ &all_channel_counts_grp,
+ NULL
+};
+
+static void all_channel_counts_release(struct device *device)
+{
+ edac_dbg(1, "Releasing device %s\n", dev_name(device));
+ kfree(device);
+}
+
+static const struct device_type all_channel_counts_type = {
+ .groups = all_channel_counts_groups,
+ .release = all_channel_counts_release,
+};
+
+/*
+ * inject sysfs attributes
+ */
+
+static DEVICE_ATTR(inject_section, S_IRUGO | S_IWUSR,
+ i7core_inject_section_show, i7core_inject_section_store);
+
+static DEVICE_ATTR(inject_type, S_IRUGO | S_IWUSR,
+ i7core_inject_type_show, i7core_inject_type_store);
+
+
+static DEVICE_ATTR(inject_eccmask, S_IRUGO | S_IWUSR,
+ i7core_inject_eccmask_show, i7core_inject_eccmask_store);
+
+static DEVICE_ATTR(inject_enable, S_IRUGO | S_IWUSR,
+ i7core_inject_enable_show, i7core_inject_enable_store);
+
+static struct attribute *i7core_dev_attrs[] = {
+ &dev_attr_inject_section.attr,
+ &dev_attr_inject_type.attr,
+ &dev_attr_inject_eccmask.attr,
+ &dev_attr_inject_enable.attr,
+ NULL
+};
+
+ATTRIBUTE_GROUPS(i7core_dev);
+
+static int i7core_create_sysfs_devices(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ int rc;
+
+ pvt->addrmatch_dev = kzalloc(sizeof(*pvt->addrmatch_dev), GFP_KERNEL);
+ if (!pvt->addrmatch_dev)
+ return -ENOMEM;
+
+ pvt->addrmatch_dev->type = &addrmatch_type;
+ pvt->addrmatch_dev->bus = mci->dev.bus;
+ device_initialize(pvt->addrmatch_dev);
+ pvt->addrmatch_dev->parent = &mci->dev;
+ dev_set_name(pvt->addrmatch_dev, "inject_addrmatch");
+ dev_set_drvdata(pvt->addrmatch_dev, mci);
+
+ edac_dbg(1, "creating %s\n", dev_name(pvt->addrmatch_dev));
+
+ rc = device_add(pvt->addrmatch_dev);
+ if (rc < 0)
+ goto err_put_addrmatch;
+
+ if (!pvt->is_registered) {
+ pvt->chancounts_dev = kzalloc(sizeof(*pvt->chancounts_dev),
+ GFP_KERNEL);
+ if (!pvt->chancounts_dev) {
+ rc = -ENOMEM;
+ goto err_del_addrmatch;
+ }
+
+ pvt->chancounts_dev->type = &all_channel_counts_type;
+ pvt->chancounts_dev->bus = mci->dev.bus;
+ device_initialize(pvt->chancounts_dev);
+ pvt->chancounts_dev->parent = &mci->dev;
+ dev_set_name(pvt->chancounts_dev, "all_channel_counts");
+ dev_set_drvdata(pvt->chancounts_dev, mci);
+
+ edac_dbg(1, "creating %s\n", dev_name(pvt->chancounts_dev));
+
+ rc = device_add(pvt->chancounts_dev);
+ if (rc < 0)
+ goto err_put_chancounts;
+ }
+ return 0;
+
+err_put_chancounts:
+ put_device(pvt->chancounts_dev);
+err_del_addrmatch:
+ device_del(pvt->addrmatch_dev);
+err_put_addrmatch:
+ put_device(pvt->addrmatch_dev);
+
+ return rc;
+}
+
+static void i7core_delete_sysfs_devices(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+
+ edac_dbg(1, "\n");
+
+ if (!pvt->is_registered) {
+ device_del(pvt->chancounts_dev);
+ put_device(pvt->chancounts_dev);
+ }
+ device_del(pvt->addrmatch_dev);
+ put_device(pvt->addrmatch_dev);
+}
+
+/****************************************************************************
+ Device initialization routines: put/get, init/exit
+ ****************************************************************************/
+
+/*
+ * i7core_put_all_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void i7core_put_devices(struct i7core_dev *i7core_dev)
+{
+ int i;
+
+ edac_dbg(0, "\n");
+ for (i = 0; i < i7core_dev->n_devs; i++) {
+ struct pci_dev *pdev = i7core_dev->pdev[i];
+ if (!pdev)
+ continue;
+ edac_dbg(0, "Removing dev %02x:%02x.%d\n",
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+ pci_dev_put(pdev);
+ }
+}
+
+static void i7core_put_all_devices(void)
+{
+ struct i7core_dev *i7core_dev, *tmp;
+
+ list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) {
+ i7core_put_devices(i7core_dev);
+ free_i7core_dev(i7core_dev);
+ }
+}
+
+static void __init i7core_xeon_pci_fixup(const struct pci_id_table *table)
+{
+ struct pci_dev *pdev = NULL;
+ int i;
+
+ /*
+ * On Xeon 55xx, the Intel Quick Path Arch Generic Non-core pci buses
+ * aren't announced by acpi. So, we need to use a legacy scan probing
+ * to detect them
+ */
+ while (table && table->descr) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, table->descr[0].dev_id, NULL);
+ if (unlikely(!pdev)) {
+ for (i = 0; i < MAX_SOCKET_BUSES; i++)
+ pcibios_scan_specific_bus(255-i);
+ }
+ pci_dev_put(pdev);
+ table++;
+ }
+}
+
+static unsigned i7core_pci_lastbus(void)
+{
+ int last_bus = 0, bus;
+ struct pci_bus *b = NULL;
+
+ while ((b = pci_find_next_bus(b)) != NULL) {
+ bus = b->number;
+ edac_dbg(0, "Found bus %d\n", bus);
+ if (bus > last_bus)
+ last_bus = bus;
+ }
+
+ edac_dbg(0, "Last bus %d\n", last_bus);
+
+ return last_bus;
+}
+
+/*
+ * i7core_get_all_devices Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ *
+ * Need to 'get' device 16 func 1 and func 2
+ */
+static int i7core_get_onedevice(struct pci_dev **prev,
+ const struct pci_id_table *table,
+ const unsigned devno,
+ const unsigned last_bus)
+{
+ struct i7core_dev *i7core_dev;
+ const struct pci_id_descr *dev_descr = &table->descr[devno];
+
+ struct pci_dev *pdev = NULL;
+ u8 bus = 0;
+ u8 socket = 0;
+
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ dev_descr->dev_id, *prev);
+
+ /*
+ * On Xeon 55xx, the Intel QuickPath Arch Generic Non-core regs
+ * is at addr 8086:2c40, instead of 8086:2c41. So, we need
+ * to probe for the alternate address in case of failure
+ */
+ if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_I7_NONCORE && !pdev) {
+ pci_dev_get(*prev); /* pci_get_device will put it */
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT, *prev);
+ }
+
+ if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE &&
+ !pdev) {
+ pci_dev_get(*prev); /* pci_get_device will put it */
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT,
+ *prev);
+ }
+
+ if (!pdev) {
+ if (*prev) {
+ *prev = pdev;
+ return 0;
+ }
+
+ if (dev_descr->optional)
+ return 0;
+
+ if (devno == 0)
+ return -ENODEV;
+
+ i7core_printk(KERN_INFO,
+ "Device not found: dev %02x.%d PCI ID %04x:%04x\n",
+ dev_descr->dev, dev_descr->func,
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+ /* End of list, leave */
+ return -ENODEV;
+ }
+ bus = pdev->bus->number;
+
+ socket = last_bus - bus;
+
+ i7core_dev = get_i7core_dev(socket);
+ if (!i7core_dev) {
+ i7core_dev = alloc_i7core_dev(socket, table);
+ if (!i7core_dev) {
+ pci_dev_put(pdev);
+ return -ENOMEM;
+ }
+ }
+
+ if (i7core_dev->pdev[devno]) {
+ i7core_printk(KERN_ERR,
+ "Duplicated device for "
+ "dev %02x:%02x.%d PCI ID %04x:%04x\n",
+ bus, dev_descr->dev, dev_descr->func,
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+ pci_dev_put(pdev);
+ return -ENODEV;
+ }
+
+ i7core_dev->pdev[devno] = pdev;
+
+ /* Sanity check */
+ if (unlikely(PCI_SLOT(pdev->devfn) != dev_descr->dev ||
+ PCI_FUNC(pdev->devfn) != dev_descr->func)) {
+ i7core_printk(KERN_ERR,
+ "Device PCI ID %04x:%04x "
+ "has dev %02x:%02x.%d instead of dev %02x:%02x.%d\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id,
+ bus, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+ bus, dev_descr->dev, dev_descr->func);
+ return -ENODEV;
+ }
+
+ /* Be sure that the device is enabled */
+ if (unlikely(pci_enable_device(pdev) < 0)) {
+ i7core_printk(KERN_ERR,
+ "Couldn't enable "
+ "dev %02x:%02x.%d PCI ID %04x:%04x\n",
+ bus, dev_descr->dev, dev_descr->func,
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+ return -ENODEV;
+ }
+
+ edac_dbg(0, "Detected socket %d dev %02x:%02x.%d PCI ID %04x:%04x\n",
+ socket, bus, dev_descr->dev,
+ dev_descr->func,
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+ /*
+ * As stated on drivers/pci/search.c, the reference count for
+ * @from is always decremented if it is not %NULL. So, as we need
+ * to get all devices up to null, we need to do a get for the device
+ */
+ pci_dev_get(pdev);
+
+ *prev = pdev;
+
+ return 0;
+}
+
+static int i7core_get_all_devices(void)
+{
+ int i, rc, last_bus;
+ struct pci_dev *pdev = NULL;
+ const struct pci_id_table *table = pci_dev_table;
+
+ last_bus = i7core_pci_lastbus();
+
+ while (table && table->descr) {
+ for (i = 0; i < table->n_devs; i++) {
+ pdev = NULL;
+ do {
+ rc = i7core_get_onedevice(&pdev, table, i,
+ last_bus);
+ if (rc < 0) {
+ if (i == 0) {
+ i = table->n_devs;
+ break;
+ }
+ i7core_put_all_devices();
+ return -ENODEV;
+ }
+ } while (pdev);
+ }
+ table++;
+ }
+
+ return 0;
+}
+
+static int mci_bind_devs(struct mem_ctl_info *mci,
+ struct i7core_dev *i7core_dev)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ int i, func, slot;
+ char *family;
+
+ pvt->is_registered = false;
+ pvt->enable_scrub = false;
+ for (i = 0; i < i7core_dev->n_devs; i++) {
+ pdev = i7core_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ func = PCI_FUNC(pdev->devfn);
+ slot = PCI_SLOT(pdev->devfn);
+ if (slot == 3) {
+ if (unlikely(func > MAX_MCR_FUNC))
+ goto error;
+ pvt->pci_mcr[func] = pdev;
+ } else if (likely(slot >= 4 && slot < 4 + NUM_CHANS)) {
+ if (unlikely(func > MAX_CHAN_FUNC))
+ goto error;
+ pvt->pci_ch[slot - 4][func] = pdev;
+ } else if (!slot && !func) {
+ pvt->pci_noncore = pdev;
+
+ /* Detect the processor family */
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_I7_NONCORE:
+ family = "Xeon 35xx/ i7core";
+ pvt->enable_scrub = false;
+ break;
+ case PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT:
+ family = "i7-800/i5-700";
+ pvt->enable_scrub = false;
+ break;
+ case PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE:
+ family = "Xeon 34xx";
+ pvt->enable_scrub = false;
+ break;
+ case PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT:
+ family = "Xeon 55xx";
+ pvt->enable_scrub = true;
+ break;
+ case PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2:
+ family = "Xeon 56xx / i7-900";
+ pvt->enable_scrub = true;
+ break;
+ default:
+ family = "unknown";
+ pvt->enable_scrub = false;
+ }
+ edac_dbg(0, "Detected a processor type %s\n", family);
+ } else
+ goto error;
+
+ edac_dbg(0, "Associated fn %d.%d, dev = %p, socket %d\n",
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+ pdev, i7core_dev->socket);
+
+ if (PCI_SLOT(pdev->devfn) == 3 &&
+ PCI_FUNC(pdev->devfn) == 2)
+ pvt->is_registered = true;
+ }
+
+ return 0;
+
+error:
+ i7core_printk(KERN_ERR, "Device %d, function %d "
+ "is out of the expected range\n",
+ slot, func);
+ return -EINVAL;
+}
+
+/****************************************************************************
+ Error check routines
+ ****************************************************************************/
+
+static void i7core_rdimm_update_ce_count(struct mem_ctl_info *mci,
+ const int chan,
+ const int new0,
+ const int new1,
+ const int new2)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ int add0 = 0, add1 = 0, add2 = 0;
+ /* Updates CE counters if it is not the first time here */
+ if (pvt->ce_count_available) {
+ /* Updates CE counters */
+
+ add2 = new2 - pvt->rdimm_last_ce_count[chan][2];
+ add1 = new1 - pvt->rdimm_last_ce_count[chan][1];
+ add0 = new0 - pvt->rdimm_last_ce_count[chan][0];
+
+ if (add2 < 0)
+ add2 += 0x7fff;
+ pvt->rdimm_ce_count[chan][2] += add2;
+
+ if (add1 < 0)
+ add1 += 0x7fff;
+ pvt->rdimm_ce_count[chan][1] += add1;
+
+ if (add0 < 0)
+ add0 += 0x7fff;
+ pvt->rdimm_ce_count[chan][0] += add0;
+ } else
+ pvt->ce_count_available = 1;
+
+ /* Store the new values */
+ pvt->rdimm_last_ce_count[chan][2] = new2;
+ pvt->rdimm_last_ce_count[chan][1] = new1;
+ pvt->rdimm_last_ce_count[chan][0] = new0;
+
+ /*updated the edac core */
+ if (add0 != 0)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, add0,
+ 0, 0, 0,
+ chan, 0, -1, "error", "");
+ if (add1 != 0)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, add1,
+ 0, 0, 0,
+ chan, 1, -1, "error", "");
+ if (add2 != 0)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, add2,
+ 0, 0, 0,
+ chan, 2, -1, "error", "");
+}
+
+static void i7core_rdimm_check_mc_ecc_err(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 rcv[3][2];
+ int i, new0, new1, new2;
+
+ /*Read DEV 3: FUN 2: MC_COR_ECC_CNT regs directly*/
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_0,
+ &rcv[0][0]);
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_1,
+ &rcv[0][1]);
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_2,
+ &rcv[1][0]);
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_3,
+ &rcv[1][1]);
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_4,
+ &rcv[2][0]);
+ pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_5,
+ &rcv[2][1]);
+ for (i = 0 ; i < 3; i++) {
+ edac_dbg(3, "MC_COR_ECC_CNT%d = 0x%x; MC_COR_ECC_CNT%d = 0x%x\n",
+ (i * 2), rcv[i][0], (i * 2) + 1, rcv[i][1]);
+ /*if the channel has 3 dimms*/
+ if (pvt->channel[i].dimms > 2) {
+ new0 = DIMM_BOT_COR_ERR(rcv[i][0]);
+ new1 = DIMM_TOP_COR_ERR(rcv[i][0]);
+ new2 = DIMM_BOT_COR_ERR(rcv[i][1]);
+ } else {
+ new0 = DIMM_TOP_COR_ERR(rcv[i][0]) +
+ DIMM_BOT_COR_ERR(rcv[i][0]);
+ new1 = DIMM_TOP_COR_ERR(rcv[i][1]) +
+ DIMM_BOT_COR_ERR(rcv[i][1]);
+ new2 = 0;
+ }
+
+ i7core_rdimm_update_ce_count(mci, i, new0, new1, new2);
+ }
+}
+
+/* This function is based on the device 3 function 4 registers as described on:
+ * Intel Xeon Processor 5500 Series Datasheet Volume 2
+ * http://www.intel.com/Assets/PDF/datasheet/321322.pdf
+ * also available at:
+ * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+static void i7core_udimm_check_mc_ecc_err(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 rcv1, rcv0;
+ int new0, new1, new2;
+
+ if (!pvt->pci_mcr[4]) {
+ edac_dbg(0, "MCR registers not found\n");
+ return;
+ }
+
+ /* Corrected test errors */
+ pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV1, &rcv1);
+ pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV0, &rcv0);
+
+ /* Store the new values */
+ new2 = DIMM2_COR_ERR(rcv1);
+ new1 = DIMM1_COR_ERR(rcv0);
+ new0 = DIMM0_COR_ERR(rcv0);
+
+ /* Updates CE counters if it is not the first time here */
+ if (pvt->ce_count_available) {
+ /* Updates CE counters */
+ int add0, add1, add2;
+
+ add2 = new2 - pvt->udimm_last_ce_count[2];
+ add1 = new1 - pvt->udimm_last_ce_count[1];
+ add0 = new0 - pvt->udimm_last_ce_count[0];
+
+ if (add2 < 0)
+ add2 += 0x7fff;
+ pvt->udimm_ce_count[2] += add2;
+
+ if (add1 < 0)
+ add1 += 0x7fff;
+ pvt->udimm_ce_count[1] += add1;
+
+ if (add0 < 0)
+ add0 += 0x7fff;
+ pvt->udimm_ce_count[0] += add0;
+
+ if (add0 | add1 | add2)
+ i7core_printk(KERN_ERR, "New Corrected error(s): "
+ "dimm0: +%d, dimm1: +%d, dimm2 +%d\n",
+ add0, add1, add2);
+ } else
+ pvt->ce_count_available = 1;
+
+ /* Store the new values */
+ pvt->udimm_last_ce_count[2] = new2;
+ pvt->udimm_last_ce_count[1] = new1;
+ pvt->udimm_last_ce_count[0] = new0;
+}
+
+/*
+ * According with tables E-11 and E-12 of chapter E.3.3 of Intel 64 and IA-32
+ * Architectures Software Developer’s Manual Volume 3B.
+ * Nehalem are defined as family 0x06, model 0x1a
+ *
+ * The MCA registers used here are the following ones:
+ * struct mce field MCA Register
+ * m->status MSR_IA32_MC8_STATUS
+ * m->addr MSR_IA32_MC8_ADDR
+ * m->misc MSR_IA32_MC8_MISC
+ * In the case of Nehalem, the error information is masked at .status and .misc
+ * fields
+ */
+static void i7core_mce_output_error(struct mem_ctl_info *mci,
+ const struct mce *m)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ char *optype, *err;
+ enum hw_event_mc_err_type tp_event;
+ unsigned long error = m->status & 0x1ff0000l;
+ bool uncorrected_error = m->mcgstatus & 1ll << 61;
+ bool ripv = m->mcgstatus & 1;
+ u32 optypenum = (m->status >> 4) & 0x07;
+ u32 core_err_cnt = (m->status >> 38) & 0x7fff;
+ u32 dimm = (m->misc >> 16) & 0x3;
+ u32 channel = (m->misc >> 18) & 0x3;
+ u32 syndrome = m->misc >> 32;
+ u32 errnum = find_first_bit(&error, 32);
+
+ if (uncorrected_error) {
+ core_err_cnt = 1;
+ if (ripv)
+ tp_event = HW_EVENT_ERR_UNCORRECTED;
+ else
+ tp_event = HW_EVENT_ERR_FATAL;
+ } else {
+ tp_event = HW_EVENT_ERR_CORRECTED;
+ }
+
+ switch (optypenum) {
+ case 0:
+ optype = "generic undef request";
+ break;
+ case 1:
+ optype = "read error";
+ break;
+ case 2:
+ optype = "write error";
+ break;
+ case 3:
+ optype = "addr/cmd error";
+ break;
+ case 4:
+ optype = "scrubbing error";
+ break;
+ default:
+ optype = "reserved";
+ break;
+ }
+
+ switch (errnum) {
+ case 16:
+ err = "read ECC error";
+ break;
+ case 17:
+ err = "RAS ECC error";
+ break;
+ case 18:
+ err = "write parity error";
+ break;
+ case 19:
+ err = "redundancy loss";
+ break;
+ case 20:
+ err = "reserved";
+ break;
+ case 21:
+ err = "memory range error";
+ break;
+ case 22:
+ err = "RTID out of range";
+ break;
+ case 23:
+ err = "address parity error";
+ break;
+ case 24:
+ err = "byte enable parity error";
+ break;
+ default:
+ err = "unknown";
+ }
+
+ /*
+ * Call the helper to output message
+ * FIXME: what to do if core_err_cnt > 1? Currently, it generates
+ * only one event
+ */
+ if (uncorrected_error || !pvt->is_registered)
+ edac_mc_handle_error(tp_event, mci, core_err_cnt,
+ m->addr >> PAGE_SHIFT,
+ m->addr & ~PAGE_MASK,
+ syndrome,
+ channel, dimm, -1,
+ err, optype);
+}
+
+/*
+ * i7core_check_error Retrieve and process errors reported by the
+ * hardware. Called by the Core module.
+ */
+static void i7core_check_error(struct mem_ctl_info *mci, struct mce *m)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+
+ i7core_mce_output_error(mci, m);
+
+ /*
+ * Now, let's increment CE error counts
+ */
+ if (!pvt->is_registered)
+ i7core_udimm_check_mc_ecc_err(mci);
+ else
+ i7core_rdimm_check_mc_ecc_err(mci);
+}
+
+/*
+ * Check that logging is enabled and that this is the right type
+ * of error for us to handle.
+ */
+static int i7core_mce_check_error(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct mce *mce = (struct mce *)data;
+ struct i7core_dev *i7_dev;
+ struct mem_ctl_info *mci;
+
+ i7_dev = get_i7core_dev(mce->socketid);
+ if (!i7_dev || (mce->kflags & MCE_HANDLED_CEC))
+ return NOTIFY_DONE;
+
+ mci = i7_dev->mci;
+
+ /*
+ * Just let mcelog handle it if the error is
+ * outside the memory controller
+ */
+ if (((mce->status & 0xffff) >> 7) != 1)
+ return NOTIFY_DONE;
+
+ /* Bank 8 registers are the only ones that we know how to handle */
+ if (mce->bank != 8)
+ return NOTIFY_DONE;
+
+ i7core_check_error(mci, mce);
+
+ /* Advise mcelog that the errors were handled */
+ mce->kflags |= MCE_HANDLED_EDAC;
+ return NOTIFY_OK;
+}
+
+static struct notifier_block i7_mce_dec = {
+ .notifier_call = i7core_mce_check_error,
+ .priority = MCE_PRIO_EDAC,
+};
+
+struct memdev_dmi_entry {
+ u8 type;
+ u8 length;
+ u16 handle;
+ u16 phys_mem_array_handle;
+ u16 mem_err_info_handle;
+ u16 total_width;
+ u16 data_width;
+ u16 size;
+ u8 form;
+ u8 device_set;
+ u8 device_locator;
+ u8 bank_locator;
+ u8 memory_type;
+ u16 type_detail;
+ u16 speed;
+ u8 manufacturer;
+ u8 serial_number;
+ u8 asset_tag;
+ u8 part_number;
+ u8 attributes;
+ u32 extended_size;
+ u16 conf_mem_clk_speed;
+} __attribute__((__packed__));
+
+
+/*
+ * Decode the DRAM Clock Frequency, be paranoid, make sure that all
+ * memory devices show the same speed, and if they don't then consider
+ * all speeds to be invalid.
+ */
+static void decode_dclk(const struct dmi_header *dh, void *_dclk_freq)
+{
+ int *dclk_freq = _dclk_freq;
+ u16 dmi_mem_clk_speed;
+
+ if (*dclk_freq == -1)
+ return;
+
+ if (dh->type == DMI_ENTRY_MEM_DEVICE) {
+ struct memdev_dmi_entry *memdev_dmi_entry =
+ (struct memdev_dmi_entry *)dh;
+ unsigned long conf_mem_clk_speed_offset =
+ (unsigned long)&memdev_dmi_entry->conf_mem_clk_speed -
+ (unsigned long)&memdev_dmi_entry->type;
+ unsigned long speed_offset =
+ (unsigned long)&memdev_dmi_entry->speed -
+ (unsigned long)&memdev_dmi_entry->type;
+
+ /* Check that a DIMM is present */
+ if (memdev_dmi_entry->size == 0)
+ return;
+
+ /*
+ * Pick the configured speed if it's available, otherwise
+ * pick the DIMM speed, or we don't have a speed.
+ */
+ if (memdev_dmi_entry->length > conf_mem_clk_speed_offset) {
+ dmi_mem_clk_speed =
+ memdev_dmi_entry->conf_mem_clk_speed;
+ } else if (memdev_dmi_entry->length > speed_offset) {
+ dmi_mem_clk_speed = memdev_dmi_entry->speed;
+ } else {
+ *dclk_freq = -1;
+ return;
+ }
+
+ if (*dclk_freq == 0) {
+ /* First pass, speed was 0 */
+ if (dmi_mem_clk_speed > 0) {
+ /* Set speed if a valid speed is read */
+ *dclk_freq = dmi_mem_clk_speed;
+ } else {
+ /* Otherwise we don't have a valid speed */
+ *dclk_freq = -1;
+ }
+ } else if (*dclk_freq > 0 &&
+ *dclk_freq != dmi_mem_clk_speed) {
+ /*
+ * If we have a speed, check that all DIMMS are the same
+ * speed, otherwise set the speed as invalid.
+ */
+ *dclk_freq = -1;
+ }
+ }
+}
+
+/*
+ * The default DCLK frequency is used as a fallback if we
+ * fail to find anything reliable in the DMI. The value
+ * is taken straight from the datasheet.
+ */
+#define DEFAULT_DCLK_FREQ 800
+
+static int get_dclk_freq(void)
+{
+ int dclk_freq = 0;
+
+ dmi_walk(decode_dclk, (void *)&dclk_freq);
+
+ if (dclk_freq < 1)
+ return DEFAULT_DCLK_FREQ;
+
+ return dclk_freq;
+}
+
+/*
+ * set_sdram_scrub_rate This routine sets byte/sec bandwidth scrub rate
+ * to hardware according to SCRUBINTERVAL formula
+ * found in datasheet.
+ */
+static int set_sdram_scrub_rate(struct mem_ctl_info *mci, u32 new_bw)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ u32 dw_scrub;
+ u32 dw_ssr;
+
+ /* Get data from the MC register, function 2 */
+ pdev = pvt->pci_mcr[2];
+ if (!pdev)
+ return -ENODEV;
+
+ pci_read_config_dword(pdev, MC_SCRUB_CONTROL, &dw_scrub);
+
+ if (new_bw == 0) {
+ /* Prepare to disable petrol scrub */
+ dw_scrub &= ~STARTSCRUB;
+ /* Stop the patrol scrub engine */
+ write_and_test(pdev, MC_SCRUB_CONTROL,
+ dw_scrub & ~SCRUBINTERVAL_MASK);
+
+ /* Get current status of scrub rate and set bit to disable */
+ pci_read_config_dword(pdev, MC_SSRCONTROL, &dw_ssr);
+ dw_ssr &= ~SSR_MODE_MASK;
+ dw_ssr |= SSR_MODE_DISABLE;
+ } else {
+ const int cache_line_size = 64;
+ const u32 freq_dclk_mhz = pvt->dclk_freq;
+ unsigned long long scrub_interval;
+ /*
+ * Translate the desired scrub rate to a register value and
+ * program the corresponding register value.
+ */
+ scrub_interval = (unsigned long long)freq_dclk_mhz *
+ cache_line_size * 1000000;
+ do_div(scrub_interval, new_bw);
+
+ if (!scrub_interval || scrub_interval > SCRUBINTERVAL_MASK)
+ return -EINVAL;
+
+ dw_scrub = SCRUBINTERVAL_MASK & scrub_interval;
+
+ /* Start the patrol scrub engine */
+ pci_write_config_dword(pdev, MC_SCRUB_CONTROL,
+ STARTSCRUB | dw_scrub);
+
+ /* Get current status of scrub rate and set bit to enable */
+ pci_read_config_dword(pdev, MC_SSRCONTROL, &dw_ssr);
+ dw_ssr &= ~SSR_MODE_MASK;
+ dw_ssr |= SSR_MODE_ENABLE;
+ }
+ /* Disable or enable scrubbing */
+ pci_write_config_dword(pdev, MC_SSRCONTROL, dw_ssr);
+
+ return new_bw;
+}
+
+/*
+ * get_sdram_scrub_rate This routine convert current scrub rate value
+ * into byte/sec bandwidth according to
+ * SCRUBINTERVAL formula found in datasheet.
+ */
+static int get_sdram_scrub_rate(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ const u32 cache_line_size = 64;
+ const u32 freq_dclk_mhz = pvt->dclk_freq;
+ unsigned long long scrub_rate;
+ u32 scrubval;
+
+ /* Get data from the MC register, function 2 */
+ pdev = pvt->pci_mcr[2];
+ if (!pdev)
+ return -ENODEV;
+
+ /* Get current scrub control data */
+ pci_read_config_dword(pdev, MC_SCRUB_CONTROL, &scrubval);
+
+ /* Mask highest 8-bits to 0 */
+ scrubval &= SCRUBINTERVAL_MASK;
+ if (!scrubval)
+ return 0;
+
+ /* Calculate scrub rate value into byte/sec bandwidth */
+ scrub_rate = (unsigned long long)freq_dclk_mhz *
+ 1000000 * cache_line_size;
+ do_div(scrub_rate, scrubval);
+ return (int)scrub_rate;
+}
+
+static void enable_sdram_scrub_setting(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 pci_lock;
+
+ /* Unlock writes to pci registers */
+ pci_read_config_dword(pvt->pci_noncore, MC_CFG_CONTROL, &pci_lock);
+ pci_lock &= ~0x3;
+ pci_write_config_dword(pvt->pci_noncore, MC_CFG_CONTROL,
+ pci_lock | MC_CFG_UNLOCK);
+
+ mci->set_sdram_scrub_rate = set_sdram_scrub_rate;
+ mci->get_sdram_scrub_rate = get_sdram_scrub_rate;
+}
+
+static void disable_sdram_scrub_setting(struct mem_ctl_info *mci)
+{
+ struct i7core_pvt *pvt = mci->pvt_info;
+ u32 pci_lock;
+
+ /* Lock writes to pci registers */
+ pci_read_config_dword(pvt->pci_noncore, MC_CFG_CONTROL, &pci_lock);
+ pci_lock &= ~0x3;
+ pci_write_config_dword(pvt->pci_noncore, MC_CFG_CONTROL,
+ pci_lock | MC_CFG_LOCK);
+}
+
+static void i7core_pci_ctl_create(struct i7core_pvt *pvt)
+{
+ pvt->i7core_pci = edac_pci_create_generic_ctl(
+ &pvt->i7core_dev->pdev[0]->dev,
+ EDAC_MOD_STR);
+ if (unlikely(!pvt->i7core_pci))
+ i7core_printk(KERN_WARNING,
+ "Unable to setup PCI error report via EDAC\n");
+}
+
+static void i7core_pci_ctl_release(struct i7core_pvt *pvt)
+{
+ if (likely(pvt->i7core_pci))
+ edac_pci_release_generic_ctl(pvt->i7core_pci);
+ else
+ i7core_printk(KERN_ERR,
+ "Couldn't find mem_ctl_info for socket %d\n",
+ pvt->i7core_dev->socket);
+ pvt->i7core_pci = NULL;
+}
+
+static void i7core_unregister_mci(struct i7core_dev *i7core_dev)
+{
+ struct mem_ctl_info *mci = i7core_dev->mci;
+ struct i7core_pvt *pvt;
+
+ if (unlikely(!mci || !mci->pvt_info)) {
+ edac_dbg(0, "MC: dev = %p\n", &i7core_dev->pdev[0]->dev);
+
+ i7core_printk(KERN_ERR, "Couldn't find mci handler\n");
+ return;
+ }
+
+ pvt = mci->pvt_info;
+
+ edac_dbg(0, "MC: mci = %p, dev = %p\n", mci, &i7core_dev->pdev[0]->dev);
+
+ /* Disable scrubrate setting */
+ if (pvt->enable_scrub)
+ disable_sdram_scrub_setting(mci);
+
+ /* Disable EDAC polling */
+ i7core_pci_ctl_release(pvt);
+
+ /* Remove MC sysfs nodes */
+ i7core_delete_sysfs_devices(mci);
+ edac_mc_del_mc(mci->pdev);
+
+ edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
+ kfree(mci->ctl_name);
+ edac_mc_free(mci);
+ i7core_dev->mci = NULL;
+}
+
+static int i7core_register_mci(struct i7core_dev *i7core_dev)
+{
+ struct mem_ctl_info *mci;
+ struct i7core_pvt *pvt;
+ int rc;
+ struct edac_mc_layer layers[2];
+
+ /* allocate a new MC control structure */
+
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = NUM_CHANS;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = MAX_DIMMS;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(i7core_dev->socket, ARRAY_SIZE(layers), layers,
+ sizeof(*pvt));
+ if (unlikely(!mci))
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p, dev = %p\n", mci, &i7core_dev->pdev[0]->dev);
+
+ pvt = mci->pvt_info;
+ memset(pvt, 0, sizeof(*pvt));
+
+ /* Associates i7core_dev and mci for future usage */
+ pvt->i7core_dev = i7core_dev;
+ i7core_dev->mci = mci;
+
+ /*
+ * FIXME: how to handle RDDR3 at MCI level? It is possible to have
+ * Mixed RDDR3/UDDR3 with Nehalem, provided that they are on different
+ * memory channels
+ */
+ mci->mtype_cap = MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i7core_edac.c";
+
+ mci->ctl_name = kasprintf(GFP_KERNEL, "i7 core #%d", i7core_dev->socket);
+ if (!mci->ctl_name) {
+ rc = -ENOMEM;
+ goto fail1;
+ }
+
+ mci->dev_name = pci_name(i7core_dev->pdev[0]);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Store pci devices at mci for faster access */
+ rc = mci_bind_devs(mci, i7core_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+
+
+ /* Get dimm basic config */
+ get_dimm_config(mci);
+ /* record ptr to the generic device */
+ mci->pdev = &i7core_dev->pdev[0]->dev;
+
+ /* Enable scrubrate setting */
+ if (pvt->enable_scrub)
+ enable_sdram_scrub_setting(mci);
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (unlikely(edac_mc_add_mc_with_groups(mci, i7core_dev_groups))) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+
+ rc = -EINVAL;
+ goto fail0;
+ }
+ if (i7core_create_sysfs_devices(mci)) {
+ edac_dbg(0, "MC: failed to create sysfs nodes\n");
+ edac_mc_del_mc(mci->pdev);
+ rc = -EINVAL;
+ goto fail0;
+ }
+
+ /* Default error mask is any memory */
+ pvt->inject.channel = 0;
+ pvt->inject.dimm = -1;
+ pvt->inject.rank = -1;
+ pvt->inject.bank = -1;
+ pvt->inject.page = -1;
+ pvt->inject.col = -1;
+
+ /* allocating generic PCI control info */
+ i7core_pci_ctl_create(pvt);
+
+ /* DCLK for scrub rate setting */
+ pvt->dclk_freq = get_dclk_freq();
+
+ return 0;
+
+fail0:
+ kfree(mci->ctl_name);
+
+fail1:
+ edac_mc_free(mci);
+ i7core_dev->mci = NULL;
+ return rc;
+}
+
+/*
+ * i7core_probe Probe for ONE instance of device to see if it is
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+
+static int i7core_probe(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rc, count = 0;
+ struct i7core_dev *i7core_dev;
+
+ /* get the pci devices we want to reserve for our use */
+ mutex_lock(&i7core_edac_lock);
+
+ /*
+ * All memory controllers are allocated at the first pass.
+ */
+ if (unlikely(probed >= 1)) {
+ mutex_unlock(&i7core_edac_lock);
+ return -ENODEV;
+ }
+ probed++;
+
+ rc = i7core_get_all_devices();
+ if (unlikely(rc < 0))
+ goto fail0;
+
+ list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
+ count++;
+ rc = i7core_register_mci(i7core_dev);
+ if (unlikely(rc < 0))
+ goto fail1;
+ }
+
+ /*
+ * Nehalem-EX uses a different memory controller. However, as the
+ * memory controller is not visible on some Nehalem/Nehalem-EP, we
+ * need to indirectly probe via a X58 PCI device. The same devices
+ * are found on (some) Nehalem-EX. So, on those machines, the
+ * probe routine needs to return -ENODEV, as the actual Memory
+ * Controller registers won't be detected.
+ */
+ if (!count) {
+ rc = -ENODEV;
+ goto fail1;
+ }
+
+ i7core_printk(KERN_INFO,
+ "Driver loaded, %d memory controller(s) found.\n",
+ count);
+
+ mutex_unlock(&i7core_edac_lock);
+ return 0;
+
+fail1:
+ list_for_each_entry(i7core_dev, &i7core_edac_list, list)
+ i7core_unregister_mci(i7core_dev);
+
+ i7core_put_all_devices();
+fail0:
+ mutex_unlock(&i7core_edac_lock);
+ return rc;
+}
+
+/*
+ * i7core_remove destructor for one instance of device
+ *
+ */
+static void i7core_remove(struct pci_dev *pdev)
+{
+ struct i7core_dev *i7core_dev;
+
+ edac_dbg(0, "\n");
+
+ /*
+ * we have a trouble here: pdev value for removal will be wrong, since
+ * it will point to the X58 register used to detect that the machine
+ * is a Nehalem or upper design. However, due to the way several PCI
+ * devices are grouped together to provide MC functionality, we need
+ * to use a different method for releasing the devices
+ */
+
+ mutex_lock(&i7core_edac_lock);
+
+ if (unlikely(!probed)) {
+ mutex_unlock(&i7core_edac_lock);
+ return;
+ }
+
+ list_for_each_entry(i7core_dev, &i7core_edac_list, list)
+ i7core_unregister_mci(i7core_dev);
+
+ /* Release PCI resources */
+ i7core_put_all_devices();
+
+ probed--;
+
+ mutex_unlock(&i7core_edac_lock);
+}
+
+MODULE_DEVICE_TABLE(pci, i7core_pci_tbl);
+
+/*
+ * i7core_driver pci_driver structure for this module
+ *
+ */
+static struct pci_driver i7core_driver = {
+ .name = "i7core_edac",
+ .probe = i7core_probe,
+ .remove = i7core_remove,
+ .id_table = i7core_pci_tbl,
+};
+
+/*
+ * i7core_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init i7core_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(2, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ if (use_pci_fixup)
+ i7core_xeon_pci_fixup(pci_dev_table);
+
+ pci_rc = pci_register_driver(&i7core_driver);
+
+ if (pci_rc >= 0) {
+ mce_register_decode_chain(&i7_mce_dec);
+ return 0;
+ }
+
+ i7core_printk(KERN_ERR, "Failed to register device with error %d.\n",
+ pci_rc);
+
+ return pci_rc;
+}
+
+/*
+ * i7core_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit i7core_exit(void)
+{
+ edac_dbg(2, "\n");
+ pci_unregister_driver(&i7core_driver);
+ mce_unregister_decode_chain(&i7_mce_dec);
+}
+
+module_init(i7core_init);
+module_exit(i7core_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab");
+MODULE_AUTHOR("Red Hat Inc. (https://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel i7 Core memory controllers - "
+ I7CORE_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i82443bxgx_edac.c b/drivers/edac/i82443bxgx_edac.c
new file mode 100644
index 0000000000..933dcf3cfd
--- /dev/null
+++ b/drivers/edac/i82443bxgx_edac.c
@@ -0,0 +1,462 @@
+/*
+ * Intel 82443BX/GX (440BX/GX chipset) Memory Controller EDAC kernel
+ * module (C) 2006 Tim Small
+ *
+ * This file may be distributed under the terms of the GNU General
+ * Public License.
+ *
+ * Written by Tim Small <tim@buttersideup.com>, based on work by Linux
+ * Networx, Thayne Harbaugh, Dan Hollis <goemon at anime dot net> and
+ * others.
+ *
+ * 440GX fix by Jason Uhlenkott <juhlenko@akamai.com>.
+ *
+ * Written with reference to 82443BX Host Bridge Datasheet:
+ * http://download.intel.com/design/chipsets/datashts/29063301.pdf
+ * references to this document given in [].
+ *
+ * This module doesn't support the 440LX, but it may be possible to
+ * make it do so (the 440LX's register definitions are different, but
+ * not completely so - I haven't studied them in enough detail to know
+ * how easy this would be).
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+
+
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "i82443bxgx_edac"
+
+/* The 82443BX supports SDRAM, or EDO (EDO for mobile only), "Memory
+ * Size: 8 MB to 512 MB (1GB with Registered DIMMs) with eight memory
+ * rows" "The 82443BX supports multiple-bit error detection and
+ * single-bit error correction when ECC mode is enabled and
+ * single/multi-bit error detection when correction is disabled.
+ * During writes to the DRAM, the 82443BX generates ECC for the data
+ * on a QWord basis. Partial QWord writes require a read-modify-write
+ * cycle when ECC is enabled."
+*/
+
+/* "Additionally, the 82443BX ensures that the data is corrected in
+ * main memory so that accumulation of errors is prevented. Another
+ * error within the same QWord would result in a double-bit error
+ * which is unrecoverable. This is known as hardware scrubbing since
+ * it requires no software intervention to correct the data in memory."
+ */
+
+/* [Also see page 100 (section 4.3), "DRAM Interface"]
+ * [Also see page 112 (section 4.6.1.4), ECC]
+ */
+
+#define I82443BXGX_NR_CSROWS 8
+#define I82443BXGX_NR_CHANS 1
+#define I82443BXGX_NR_DIMMS 4
+
+/* 82443 PCI Device 0 */
+#define I82443BXGX_NBXCFG 0x50 /* 32bit register starting at this PCI
+ * config space offset */
+#define I82443BXGX_NBXCFG_OFFSET_NON_ECCROW 24 /* Array of bits, zero if
+ * row is non-ECC */
+#define I82443BXGX_NBXCFG_OFFSET_DRAM_FREQ 12 /* 2 bits,00=100MHz,10=66 MHz */
+
+#define I82443BXGX_NBXCFG_OFFSET_DRAM_INTEGRITY 7 /* 2 bits: */
+#define I82443BXGX_NBXCFG_INTEGRITY_NONE 0x0 /* 00 = Non-ECC */
+#define I82443BXGX_NBXCFG_INTEGRITY_EC 0x1 /* 01 = EC (only) */
+#define I82443BXGX_NBXCFG_INTEGRITY_ECC 0x2 /* 10 = ECC */
+#define I82443BXGX_NBXCFG_INTEGRITY_SCRUB 0x3 /* 11 = ECC + HW Scrub */
+
+#define I82443BXGX_NBXCFG_OFFSET_ECC_DIAG_ENABLE 6
+
+/* 82443 PCI Device 0 */
+#define I82443BXGX_EAP 0x80 /* 32bit register starting at this PCI
+ * config space offset, Error Address
+ * Pointer Register */
+#define I82443BXGX_EAP_OFFSET_EAP 12 /* High 20 bits of error address */
+#define I82443BXGX_EAP_OFFSET_MBE BIT(1) /* Err at EAP was multi-bit (W1TC) */
+#define I82443BXGX_EAP_OFFSET_SBE BIT(0) /* Err at EAP was single-bit (W1TC) */
+
+#define I82443BXGX_ERRCMD 0x90 /* 8bit register starting at this PCI
+ * config space offset. */
+#define I82443BXGX_ERRCMD_OFFSET_SERR_ON_MBE BIT(1) /* 1 = enable */
+#define I82443BXGX_ERRCMD_OFFSET_SERR_ON_SBE BIT(0) /* 1 = enable */
+
+#define I82443BXGX_ERRSTS 0x91 /* 16bit register starting at this PCI
+ * config space offset. */
+#define I82443BXGX_ERRSTS_OFFSET_MBFRE 5 /* 3 bits - first err row multibit */
+#define I82443BXGX_ERRSTS_OFFSET_MEF BIT(4) /* 1 = MBE occurred */
+#define I82443BXGX_ERRSTS_OFFSET_SBFRE 1 /* 3 bits - first err row singlebit */
+#define I82443BXGX_ERRSTS_OFFSET_SEF BIT(0) /* 1 = SBE occurred */
+
+#define I82443BXGX_DRAMC 0x57 /* 8bit register starting at this PCI
+ * config space offset. */
+#define I82443BXGX_DRAMC_OFFSET_DT 3 /* 2 bits, DRAM Type */
+#define I82443BXGX_DRAMC_DRAM_IS_EDO 0 /* 00 = EDO */
+#define I82443BXGX_DRAMC_DRAM_IS_SDRAM 1 /* 01 = SDRAM */
+#define I82443BXGX_DRAMC_DRAM_IS_RSDRAM 2 /* 10 = Registered SDRAM */
+
+#define I82443BXGX_DRB 0x60 /* 8x 8bit registers starting at this PCI
+ * config space offset. */
+
+/* FIXME - don't poll when ECC disabled? */
+
+struct i82443bxgx_edacmc_error_info {
+ u32 eap;
+};
+
+static struct edac_pci_ctl_info *i82443bxgx_pci;
+
+static struct pci_dev *mci_pdev; /* init dev: in case that AGP code has
+ * already registered driver
+ */
+
+static int i82443bxgx_registered = 1;
+
+static void i82443bxgx_edacmc_get_error_info(struct mem_ctl_info *mci,
+ struct i82443bxgx_edacmc_error_info
+ *info)
+{
+ struct pci_dev *pdev;
+ pdev = to_pci_dev(mci->pdev);
+ pci_read_config_dword(pdev, I82443BXGX_EAP, &info->eap);
+ if (info->eap & I82443BXGX_EAP_OFFSET_SBE)
+ /* Clear error to allow next error to be reported [p.61] */
+ pci_write_bits32(pdev, I82443BXGX_EAP,
+ I82443BXGX_EAP_OFFSET_SBE,
+ I82443BXGX_EAP_OFFSET_SBE);
+
+ if (info->eap & I82443BXGX_EAP_OFFSET_MBE)
+ /* Clear error to allow next error to be reported [p.61] */
+ pci_write_bits32(pdev, I82443BXGX_EAP,
+ I82443BXGX_EAP_OFFSET_MBE,
+ I82443BXGX_EAP_OFFSET_MBE);
+}
+
+static int i82443bxgx_edacmc_process_error_info(struct mem_ctl_info *mci,
+ struct
+ i82443bxgx_edacmc_error_info
+ *info, int handle_errors)
+{
+ int error_found = 0;
+ u32 eapaddr, page, pageoffset;
+
+ /* bits 30:12 hold the 4kb block in which the error occurred
+ * [p.61] */
+ eapaddr = (info->eap & 0xfffff000);
+ page = eapaddr >> PAGE_SHIFT;
+ pageoffset = eapaddr - (page << PAGE_SHIFT);
+
+ if (info->eap & I82443BXGX_EAP_OFFSET_SBE) {
+ error_found = 1;
+ if (handle_errors)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ page, pageoffset, 0,
+ edac_mc_find_csrow_by_page(mci, page),
+ 0, -1, mci->ctl_name, "");
+ }
+
+ if (info->eap & I82443BXGX_EAP_OFFSET_MBE) {
+ error_found = 1;
+ if (handle_errors)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ page, pageoffset, 0,
+ edac_mc_find_csrow_by_page(mci, page),
+ 0, -1, mci->ctl_name, "");
+ }
+
+ return error_found;
+}
+
+static void i82443bxgx_edacmc_check(struct mem_ctl_info *mci)
+{
+ struct i82443bxgx_edacmc_error_info info;
+
+ i82443bxgx_edacmc_get_error_info(mci, &info);
+ i82443bxgx_edacmc_process_error_info(mci, &info, 1);
+}
+
+static void i82443bxgx_init_csrows(struct mem_ctl_info *mci,
+ struct pci_dev *pdev,
+ enum edac_type edac_mode,
+ enum mem_type mtype)
+{
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ int index;
+ u8 drbar, dramc;
+ u32 row_base, row_high_limit, row_high_limit_last;
+
+ pci_read_config_byte(pdev, I82443BXGX_DRAMC, &dramc);
+ row_high_limit_last = 0;
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ pci_read_config_byte(pdev, I82443BXGX_DRB + index, &drbar);
+ edac_dbg(1, "MC%d: Row=%d DRB = %#0x\n",
+ mci->mc_idx, index, drbar);
+ row_high_limit = ((u32) drbar << 23);
+ /* find the DRAM Chip Select Base address and mask */
+ edac_dbg(1, "MC%d: Row=%d, Boundary Address=%#0x, Last = %#0x\n",
+ mci->mc_idx, index, row_high_limit,
+ row_high_limit_last);
+
+ /* 440GX goes to 2GB, represented with a DRB of 0. */
+ if (row_high_limit_last && !row_high_limit)
+ row_high_limit = 1UL << 31;
+
+ /* This row is empty [p.49] */
+ if (row_high_limit == row_high_limit_last)
+ continue;
+ row_base = row_high_limit_last;
+ csrow->first_page = row_base >> PAGE_SHIFT;
+ csrow->last_page = (row_high_limit >> PAGE_SHIFT) - 1;
+ dimm->nr_pages = csrow->last_page - csrow->first_page + 1;
+ /* EAP reports in 4kilobyte granularity [61] */
+ dimm->grain = 1 << 12;
+ dimm->mtype = mtype;
+ /* I don't think 440BX can tell you device type? FIXME? */
+ dimm->dtype = DEV_UNKNOWN;
+ /* Mode is global to all rows on 440BX */
+ dimm->edac_mode = edac_mode;
+ row_high_limit_last = row_high_limit;
+ }
+}
+
+static int i82443bxgx_edacmc_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ u8 dramc;
+ u32 nbxcfg, ecc_mode;
+ enum mem_type mtype;
+ enum edac_type edac_mode;
+
+ edac_dbg(0, "MC:\n");
+
+ /* Something is really hosed if PCI config space reads from
+ * the MC aren't working.
+ */
+ if (pci_read_config_dword(pdev, I82443BXGX_NBXCFG, &nbxcfg))
+ return -EIO;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = I82443BXGX_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = I82443BXGX_NR_CHANS;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p\n", mci);
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_EDO | MEM_FLAG_SDR | MEM_FLAG_RDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ pci_read_config_byte(pdev, I82443BXGX_DRAMC, &dramc);
+ switch ((dramc >> I82443BXGX_DRAMC_OFFSET_DT) & (BIT(0) | BIT(1))) {
+ case I82443BXGX_DRAMC_DRAM_IS_EDO:
+ mtype = MEM_EDO;
+ break;
+ case I82443BXGX_DRAMC_DRAM_IS_SDRAM:
+ mtype = MEM_SDR;
+ break;
+ case I82443BXGX_DRAMC_DRAM_IS_RSDRAM:
+ mtype = MEM_RDR;
+ break;
+ default:
+ edac_dbg(0, "Unknown/reserved DRAM type value in DRAMC register!\n");
+ mtype = -MEM_UNKNOWN;
+ }
+
+ if ((mtype == MEM_SDR) || (mtype == MEM_RDR))
+ mci->edac_cap = mci->edac_ctl_cap;
+ else
+ mci->edac_cap = EDAC_FLAG_NONE;
+
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ pci_read_config_dword(pdev, I82443BXGX_NBXCFG, &nbxcfg);
+ ecc_mode = ((nbxcfg >> I82443BXGX_NBXCFG_OFFSET_DRAM_INTEGRITY) &
+ (BIT(0) | BIT(1)));
+
+ mci->scrub_mode = (ecc_mode == I82443BXGX_NBXCFG_INTEGRITY_SCRUB)
+ ? SCRUB_HW_SRC : SCRUB_NONE;
+
+ switch (ecc_mode) {
+ case I82443BXGX_NBXCFG_INTEGRITY_NONE:
+ edac_mode = EDAC_NONE;
+ break;
+ case I82443BXGX_NBXCFG_INTEGRITY_EC:
+ edac_mode = EDAC_EC;
+ break;
+ case I82443BXGX_NBXCFG_INTEGRITY_ECC:
+ case I82443BXGX_NBXCFG_INTEGRITY_SCRUB:
+ edac_mode = EDAC_SECDED;
+ break;
+ default:
+ edac_dbg(0, "Unknown/reserved ECC state in NBXCFG register!\n");
+ edac_mode = EDAC_UNKNOWN;
+ break;
+ }
+
+ i82443bxgx_init_csrows(mci, pdev, edac_mode, mtype);
+
+ /* Many BIOSes don't clear error flags on boot, so do this
+ * here, or we get "phantom" errors occurring at module-load
+ * time. */
+ pci_write_bits32(pdev, I82443BXGX_EAP,
+ (I82443BXGX_EAP_OFFSET_SBE |
+ I82443BXGX_EAP_OFFSET_MBE),
+ (I82443BXGX_EAP_OFFSET_SBE |
+ I82443BXGX_EAP_OFFSET_MBE));
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = "I82443BXGX";
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i82443bxgx_edacmc_check;
+ mci->ctl_page_to_phys = NULL;
+
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* allocating generic PCI control info */
+ i82443bxgx_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i82443bxgx_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ edac_dbg(3, "MC: success\n");
+ return 0;
+
+fail:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int i82443bxgx_edacmc_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ /* don't need to call pci_enable_device() */
+ rc = i82443bxgx_edacmc_probe1(pdev, ent->driver_data);
+
+ if (mci_pdev == NULL)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i82443bxgx_edacmc_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (i82443bxgx_pci)
+ edac_pci_release_generic_ctl(i82443bxgx_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i82443bxgx_pci_tbl[] = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82443BX_0)},
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82443BX_2)},
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82443GX_0)},
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82443GX_2)},
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i82443bxgx_pci_tbl);
+
+static struct pci_driver i82443bxgx_edacmc_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i82443bxgx_edacmc_init_one,
+ .remove = i82443bxgx_edacmc_remove_one,
+ .id_table = i82443bxgx_pci_tbl,
+};
+
+static int __init i82443bxgx_edacmc_init(void)
+{
+ int pci_rc;
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i82443bxgx_edacmc_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (mci_pdev == NULL) {
+ const struct pci_device_id *id = &i82443bxgx_pci_tbl[0];
+ int i = 0;
+ i82443bxgx_registered = 0;
+
+ while (mci_pdev == NULL && id->vendor != 0) {
+ mci_pdev = pci_get_device(id->vendor,
+ id->device, NULL);
+ i++;
+ id = &i82443bxgx_pci_tbl[i];
+ }
+ if (!mci_pdev) {
+ edac_dbg(0, "i82443bxgx pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i82443bxgx_edacmc_init_one(mci_pdev, i82443bxgx_pci_tbl);
+
+ if (pci_rc < 0) {
+ edac_dbg(0, "i82443bxgx init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i82443bxgx_edacmc_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+ return pci_rc;
+}
+
+static void __exit i82443bxgx_edacmc_exit(void)
+{
+ pci_unregister_driver(&i82443bxgx_edacmc_driver);
+
+ if (!i82443bxgx_registered)
+ i82443bxgx_edacmc_remove_one(mci_pdev);
+
+ pci_dev_put(mci_pdev);
+}
+
+module_init(i82443bxgx_edacmc_init);
+module_exit(i82443bxgx_edacmc_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Tim Small <tim@buttersideup.com> - WPAD");
+MODULE_DESCRIPTION("EDAC MC support for Intel 82443BX/GX memory controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i82860_edac.c b/drivers/edac/i82860_edac.c
new file mode 100644
index 0000000000..b8a497f0de
--- /dev/null
+++ b/drivers/edac/i82860_edac.c
@@ -0,0 +1,362 @@
+/*
+ * Intel 82860 Memory Controller kernel module
+ * (C) 2005 Red Hat (http://www.redhat.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Ben Woodard <woodard@redhat.com>
+ * shamelessly copied from and based upon the edac_i82875 driver
+ * by Thayne Harbaugh of Linux Networx. (http://lnxi.com)
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "i82860_edac"
+
+#define i82860_printk(level, fmt, arg...) \
+ edac_printk(level, "i82860", fmt, ##arg)
+
+#define i82860_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i82860", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_82860_0
+#define PCI_DEVICE_ID_INTEL_82860_0 0x2531
+#endif /* PCI_DEVICE_ID_INTEL_82860_0 */
+
+#define I82860_MCHCFG 0x50
+#define I82860_GBA 0x60
+#define I82860_GBA_MASK 0x7FF
+#define I82860_GBA_SHIFT 24
+#define I82860_ERRSTS 0xC8
+#define I82860_EAP 0xE4
+#define I82860_DERRCTL_STS 0xE2
+
+enum i82860_chips {
+ I82860 = 0,
+};
+
+struct i82860_dev_info {
+ const char *ctl_name;
+};
+
+struct i82860_error_info {
+ u16 errsts;
+ u32 eap;
+ u16 derrsyn;
+ u16 errsts2;
+};
+
+static const struct i82860_dev_info i82860_devs[] = {
+ [I82860] = {
+ .ctl_name = "i82860"},
+};
+
+static struct pci_dev *mci_pdev; /* init dev: in case that AGP code
+ * has already registered driver
+ */
+static struct edac_pci_ctl_info *i82860_pci;
+
+static void i82860_get_error_info(struct mem_ctl_info *mci,
+ struct i82860_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, I82860_ERRSTS, &info->errsts);
+ pci_read_config_dword(pdev, I82860_EAP, &info->eap);
+ pci_read_config_word(pdev, I82860_DERRCTL_STS, &info->derrsyn);
+ pci_read_config_word(pdev, I82860_ERRSTS, &info->errsts2);
+
+ pci_write_bits16(pdev, I82860_ERRSTS, 0x0003, 0x0003);
+
+ /*
+ * If the error is the same for both reads then the first set of reads
+ * is valid. If there is a change then there is a CE no info and the
+ * second set of reads is valid and should be UE info.
+ */
+ if (!(info->errsts2 & 0x0003))
+ return;
+
+ if ((info->errsts ^ info->errsts2) & 0x0003) {
+ pci_read_config_dword(pdev, I82860_EAP, &info->eap);
+ pci_read_config_word(pdev, I82860_DERRCTL_STS, &info->derrsyn);
+ }
+}
+
+static int i82860_process_error_info(struct mem_ctl_info *mci,
+ struct i82860_error_info *info,
+ int handle_errors)
+{
+ struct dimm_info *dimm;
+ int row;
+
+ if (!(info->errsts2 & 0x0003))
+ return 0;
+
+ if (!handle_errors)
+ return 1;
+
+ if ((info->errsts ^ info->errsts2) & 0x0003) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ info->eap >>= PAGE_SHIFT;
+ row = edac_mc_find_csrow_by_page(mci, info->eap);
+ dimm = mci->csrows[row]->channels[0]->dimm;
+
+ if (info->errsts & 0x0002)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ info->eap, 0, 0,
+ dimm->location[0], dimm->location[1], -1,
+ "i82860 UE", "");
+ else
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ info->eap, 0, info->derrsyn,
+ dimm->location[0], dimm->location[1], -1,
+ "i82860 CE", "");
+
+ return 1;
+}
+
+static void i82860_check(struct mem_ctl_info *mci)
+{
+ struct i82860_error_info info;
+
+ i82860_get_error_info(mci, &info);
+ i82860_process_error_info(mci, &info, 1);
+}
+
+static void i82860_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev)
+{
+ unsigned long last_cumul_size;
+ u16 mchcfg_ddim; /* DRAM Data Integrity Mode 0=none, 2=edac */
+ u16 value;
+ u32 cumul_size;
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ int index;
+
+ pci_read_config_word(pdev, I82860_MCHCFG, &mchcfg_ddim);
+ mchcfg_ddim = mchcfg_ddim & 0x180;
+ last_cumul_size = 0;
+
+ /* The group row boundary (GRA) reg values are boundary address
+ * for each DRAM row with a granularity of 16MB. GRA regs are
+ * cumulative; therefore GRA15 will contain the total memory contained
+ * in all eight rows.
+ */
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ pci_read_config_word(pdev, I82860_GBA + index * 2, &value);
+ cumul_size = (value & I82860_GBA_MASK) <<
+ (I82860_GBA_SHIFT - PAGE_SHIFT);
+ edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
+
+ if (cumul_size == last_cumul_size)
+ continue; /* not populated */
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ dimm->nr_pages = cumul_size - last_cumul_size;
+ last_cumul_size = cumul_size;
+ dimm->grain = 1 << 12; /* I82860_EAP has 4KiB reolution */
+ dimm->mtype = MEM_RMBS;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = mchcfg_ddim ? EDAC_SECDED : EDAC_NONE;
+ }
+}
+
+static int i82860_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct i82860_error_info discard;
+
+ /*
+ * RDRAM has channels but these don't map onto the csrow abstraction.
+ * According with the datasheet, there are 2 Rambus channels, supporting
+ * up to 16 direct RDRAM devices.
+ * The device groups from the GRA registers seem to map reasonably
+ * well onto the notion of a chip select row.
+ * There are 16 GRA registers and since the name is associated with
+ * the channel and the GRA registers map to physical devices so we are
+ * going to make 1 channel for group.
+ */
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = 2;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = 8;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (!mci)
+ return -ENOMEM;
+
+ edac_dbg(3, "init mci\n");
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ /* I"m not sure about this but I think that all RDRAM is SECDED */
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = i82860_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i82860_check;
+ mci->ctl_page_to_phys = NULL;
+ i82860_init_csrows(mci, pdev);
+ i82860_get_error_info(mci, &discard); /* clear counters */
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* allocating generic PCI control info */
+ i82860_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i82860_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+
+ return 0;
+
+fail:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int i82860_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "\n");
+ i82860_printk(KERN_INFO, "i82860 init one\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = i82860_probe1(pdev, ent->driver_data);
+
+ if (rc == 0)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i82860_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (i82860_pci)
+ edac_pci_release_generic_ctl(i82860_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i82860_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 82860_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I82860},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i82860_pci_tbl);
+
+static struct pci_driver i82860_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i82860_init_one,
+ .remove = i82860_remove_one,
+ .id_table = i82860_pci_tbl,
+};
+
+static int __init i82860_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ if ((pci_rc = pci_register_driver(&i82860_driver)) < 0)
+ goto fail0;
+
+ if (!mci_pdev) {
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_82860_0, NULL);
+
+ if (mci_pdev == NULL) {
+ edac_dbg(0, "860 pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i82860_init_one(mci_pdev, i82860_pci_tbl);
+
+ if (pci_rc < 0) {
+ edac_dbg(0, "860 init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i82860_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+ return pci_rc;
+}
+
+static void __exit i82860_exit(void)
+{
+ edac_dbg(3, "\n");
+ pci_unregister_driver(&i82860_driver);
+ pci_dev_put(mci_pdev);
+}
+
+module_init(i82860_init);
+module_exit(i82860_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com) Ben Woodard <woodard@redhat.com>");
+MODULE_DESCRIPTION("ECC support for Intel 82860 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i82875p_edac.c b/drivers/edac/i82875p_edac.c
new file mode 100644
index 0000000000..553880b9fc
--- /dev/null
+++ b/drivers/edac/i82875p_edac.c
@@ -0,0 +1,599 @@
+/*
+ * Intel D82875P Memory Controller kernel module
+ * (C) 2003 Linux Networx (http://lnxi.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Thayne Harbaugh
+ * Contributors:
+ * Wang Zhenyu at intel.com
+ *
+ * $Id: edac_i82875p.c,v 1.5.2.11 2005/10/05 00:43:44 dsp_llnl Exp $
+ *
+ * Note: E7210 appears same as D82875P - zhenyu.z.wang at intel.com
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "i82875p_edac"
+
+#define i82875p_printk(level, fmt, arg...) \
+ edac_printk(level, "i82875p", fmt, ##arg)
+
+#define i82875p_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i82875p", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_82875_0
+#define PCI_DEVICE_ID_INTEL_82875_0 0x2578
+#endif /* PCI_DEVICE_ID_INTEL_82875_0 */
+
+#ifndef PCI_DEVICE_ID_INTEL_82875_6
+#define PCI_DEVICE_ID_INTEL_82875_6 0x257e
+#endif /* PCI_DEVICE_ID_INTEL_82875_6 */
+
+/* four csrows in dual channel, eight in single channel */
+#define I82875P_NR_DIMMS 8
+#define I82875P_NR_CSROWS(nr_chans) (I82875P_NR_DIMMS / (nr_chans))
+
+/* Intel 82875p register addresses - device 0 function 0 - DRAM Controller */
+#define I82875P_EAP 0x58 /* Error Address Pointer (32b)
+ *
+ * 31:12 block address
+ * 11:0 reserved
+ */
+
+#define I82875P_DERRSYN 0x5c /* DRAM Error Syndrome (8b)
+ *
+ * 7:0 DRAM ECC Syndrome
+ */
+
+#define I82875P_DES 0x5d /* DRAM Error Status (8b)
+ *
+ * 7:1 reserved
+ * 0 Error channel 0/1
+ */
+
+#define I82875P_ERRSTS 0xc8 /* Error Status Register (16b)
+ *
+ * 15:10 reserved
+ * 9 non-DRAM lock error (ndlock)
+ * 8 Sftwr Generated SMI
+ * 7 ECC UE
+ * 6 reserved
+ * 5 MCH detects unimplemented cycle
+ * 4 AGP access outside GA
+ * 3 Invalid AGP access
+ * 2 Invalid GA translation table
+ * 1 Unsupported AGP command
+ * 0 ECC CE
+ */
+
+#define I82875P_ERRCMD 0xca /* Error Command (16b)
+ *
+ * 15:10 reserved
+ * 9 SERR on non-DRAM lock
+ * 8 SERR on ECC UE
+ * 7 SERR on ECC CE
+ * 6 target abort on high exception
+ * 5 detect unimplemented cyc
+ * 4 AGP access outside of GA
+ * 3 SERR on invalid AGP access
+ * 2 invalid translation table
+ * 1 SERR on unsupported AGP command
+ * 0 reserved
+ */
+
+/* Intel 82875p register addresses - device 6 function 0 - DRAM Controller */
+#define I82875P_PCICMD6 0x04 /* PCI Command Register (16b)
+ *
+ * 15:10 reserved
+ * 9 fast back-to-back - ro 0
+ * 8 SERR enable - ro 0
+ * 7 addr/data stepping - ro 0
+ * 6 parity err enable - ro 0
+ * 5 VGA palette snoop - ro 0
+ * 4 mem wr & invalidate - ro 0
+ * 3 special cycle - ro 0
+ * 2 bus master - ro 0
+ * 1 mem access dev6 - 0(dis),1(en)
+ * 0 IO access dev3 - 0(dis),1(en)
+ */
+
+#define I82875P_BAR6 0x10 /* Mem Delays Base ADDR Reg (32b)
+ *
+ * 31:12 mem base addr [31:12]
+ * 11:4 address mask - ro 0
+ * 3 prefetchable - ro 0(non),1(pre)
+ * 2:1 mem type - ro 0
+ * 0 mem space - ro 0
+ */
+
+/* Intel 82875p MMIO register space - device 0 function 0 - MMR space */
+
+#define I82875P_DRB_SHIFT 26 /* 64MiB grain */
+#define I82875P_DRB 0x00 /* DRAM Row Boundary (8b x 8)
+ *
+ * 7 reserved
+ * 6:0 64MiB row boundary addr
+ */
+
+#define I82875P_DRA 0x10 /* DRAM Row Attribute (4b x 8)
+ *
+ * 7 reserved
+ * 6:4 row attr row 1
+ * 3 reserved
+ * 2:0 row attr row 0
+ *
+ * 000 = 4KiB
+ * 001 = 8KiB
+ * 010 = 16KiB
+ * 011 = 32KiB
+ */
+
+#define I82875P_DRC 0x68 /* DRAM Controller Mode (32b)
+ *
+ * 31:30 reserved
+ * 29 init complete
+ * 28:23 reserved
+ * 22:21 nr chan 00=1,01=2
+ * 20 reserved
+ * 19:18 Data Integ Mode 00=none,01=ecc
+ * 17:11 reserved
+ * 10:8 refresh mode
+ * 7 reserved
+ * 6:4 mode select
+ * 3:2 reserved
+ * 1:0 DRAM type 01=DDR
+ */
+
+enum i82875p_chips {
+ I82875P = 0,
+};
+
+struct i82875p_pvt {
+ struct pci_dev *ovrfl_pdev;
+ void __iomem *ovrfl_window;
+};
+
+struct i82875p_dev_info {
+ const char *ctl_name;
+};
+
+struct i82875p_error_info {
+ u16 errsts;
+ u32 eap;
+ u8 des;
+ u8 derrsyn;
+ u16 errsts2;
+};
+
+static const struct i82875p_dev_info i82875p_devs[] = {
+ [I82875P] = {
+ .ctl_name = "i82875p"},
+};
+
+static struct pci_dev *mci_pdev; /* init dev: in case that AGP code has
+ * already registered driver
+ */
+
+static struct edac_pci_ctl_info *i82875p_pci;
+
+static void i82875p_get_error_info(struct mem_ctl_info *mci,
+ struct i82875p_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, I82875P_ERRSTS, &info->errsts);
+
+ if (!(info->errsts & 0x0081))
+ return;
+
+ pci_read_config_dword(pdev, I82875P_EAP, &info->eap);
+ pci_read_config_byte(pdev, I82875P_DES, &info->des);
+ pci_read_config_byte(pdev, I82875P_DERRSYN, &info->derrsyn);
+ pci_read_config_word(pdev, I82875P_ERRSTS, &info->errsts2);
+
+ /*
+ * If the error is the same then we can for both reads then
+ * the first set of reads is valid. If there is a change then
+ * there is a CE no info and the second set of reads is valid
+ * and should be UE info.
+ */
+ if ((info->errsts ^ info->errsts2) & 0x0081) {
+ pci_read_config_dword(pdev, I82875P_EAP, &info->eap);
+ pci_read_config_byte(pdev, I82875P_DES, &info->des);
+ pci_read_config_byte(pdev, I82875P_DERRSYN, &info->derrsyn);
+ }
+
+ pci_write_bits16(pdev, I82875P_ERRSTS, 0x0081, 0x0081);
+}
+
+static int i82875p_process_error_info(struct mem_ctl_info *mci,
+ struct i82875p_error_info *info,
+ int handle_errors)
+{
+ int row, multi_chan;
+
+ multi_chan = mci->csrows[0]->nr_channels - 1;
+
+ if (!(info->errsts & 0x0081))
+ return 0;
+
+ if (!handle_errors)
+ return 1;
+
+ if ((info->errsts ^ info->errsts2) & 0x0081) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1,
+ "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ info->eap >>= PAGE_SHIFT;
+ row = edac_mc_find_csrow_by_page(mci, info->eap);
+
+ if (info->errsts & 0x0080)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ info->eap, 0, 0,
+ row, -1, -1,
+ "i82875p UE", "");
+ else
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ info->eap, 0, info->derrsyn,
+ row, multi_chan ? (info->des & 0x1) : 0,
+ -1, "i82875p CE", "");
+
+ return 1;
+}
+
+static void i82875p_check(struct mem_ctl_info *mci)
+{
+ struct i82875p_error_info info;
+
+ i82875p_get_error_info(mci, &info);
+ i82875p_process_error_info(mci, &info, 1);
+}
+
+/* Return 0 on success or 1 on failure. */
+static int i82875p_setup_overfl_dev(struct pci_dev *pdev,
+ struct pci_dev **ovrfl_pdev,
+ void __iomem **ovrfl_window)
+{
+ struct pci_dev *dev;
+ void __iomem *window;
+
+ *ovrfl_pdev = NULL;
+ *ovrfl_window = NULL;
+ dev = pci_get_device(PCI_VEND_DEV(INTEL, 82875_6), NULL);
+
+ if (dev == NULL) {
+ /* Intel tells BIOS developers to hide device 6 which
+ * configures the overflow device access containing
+ * the DRBs - this is where we expose device 6.
+ * http://www.x86-secret.com/articles/tweak/pat/patsecrets-2.htm
+ */
+ pci_write_bits8(pdev, 0xf4, 0x2, 0x2);
+ dev = pci_scan_single_device(pdev->bus, PCI_DEVFN(6, 0));
+
+ if (dev == NULL)
+ return 1;
+
+ pci_bus_assign_resources(dev->bus);
+ pci_bus_add_device(dev);
+ }
+
+ *ovrfl_pdev = dev;
+
+ if (pci_enable_device(dev)) {
+ i82875p_printk(KERN_ERR, "%s(): Failed to enable overflow "
+ "device\n", __func__);
+ return 1;
+ }
+
+ if (pci_request_regions(dev, pci_name(dev))) {
+#ifdef CORRECT_BIOS
+ goto fail0;
+#endif
+ }
+
+ /* cache is irrelevant for PCI bus reads/writes */
+ window = pci_ioremap_bar(dev, 0);
+ if (window == NULL) {
+ i82875p_printk(KERN_ERR, "%s(): Failed to ioremap bar6\n",
+ __func__);
+ goto fail1;
+ }
+
+ *ovrfl_window = window;
+ return 0;
+
+fail1:
+ pci_release_regions(dev);
+
+#ifdef CORRECT_BIOS
+fail0:
+ pci_disable_device(dev);
+#endif
+ /* NOTE: the ovrfl proc entry and pci_dev are intentionally left */
+ return 1;
+}
+
+/* Return 1 if dual channel mode is active. Else return 0. */
+static inline int dual_channel_active(u32 drc)
+{
+ return (drc >> 21) & 0x1;
+}
+
+static void i82875p_init_csrows(struct mem_ctl_info *mci,
+ struct pci_dev *pdev,
+ void __iomem * ovrfl_window, u32 drc)
+{
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ unsigned nr_chans = dual_channel_active(drc) + 1;
+ unsigned long last_cumul_size;
+ u8 value;
+ u32 drc_ddim; /* DRAM Data Integrity Mode 0=none,2=edac */
+ u32 cumul_size, nr_pages;
+ int index, j;
+
+ drc_ddim = (drc >> 18) & 0x1;
+ last_cumul_size = 0;
+
+ /* The dram row boundary (DRB) reg values are boundary address
+ * for each DRAM row with a granularity of 32 or 64MB (single/dual
+ * channel operation). DRB regs are cumulative; therefore DRB7 will
+ * contain the total memory contained in all eight rows.
+ */
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+
+ value = readb(ovrfl_window + I82875P_DRB + index);
+ cumul_size = value << (I82875P_DRB_SHIFT - PAGE_SHIFT);
+ edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
+ if (cumul_size == last_cumul_size)
+ continue; /* not populated */
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ nr_pages = cumul_size - last_cumul_size;
+ last_cumul_size = cumul_size;
+
+ for (j = 0; j < nr_chans; j++) {
+ dimm = csrow->channels[j]->dimm;
+
+ dimm->nr_pages = nr_pages / nr_chans;
+ dimm->grain = 1 << 12; /* I82875P_EAP has 4KiB reolution */
+ dimm->mtype = MEM_DDR;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = drc_ddim ? EDAC_SECDED : EDAC_NONE;
+ }
+ }
+}
+
+static int i82875p_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int rc = -ENODEV;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct i82875p_pvt *pvt;
+ struct pci_dev *ovrfl_pdev;
+ void __iomem *ovrfl_window;
+ u32 drc;
+ u32 nr_chans;
+ struct i82875p_error_info discard;
+
+ edac_dbg(0, "\n");
+
+ if (i82875p_setup_overfl_dev(pdev, &ovrfl_pdev, &ovrfl_window))
+ return -ENODEV;
+ drc = readl(ovrfl_window + I82875P_DRC);
+ nr_chans = dual_channel_active(drc) + 1;
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = I82875P_NR_CSROWS(nr_chans);
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = nr_chans;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (!mci) {
+ rc = -ENOMEM;
+ goto fail0;
+ }
+
+ edac_dbg(3, "init mci\n");
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_UNKNOWN;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = i82875p_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i82875p_check;
+ mci->ctl_page_to_phys = NULL;
+ edac_dbg(3, "init pvt\n");
+ pvt = (struct i82875p_pvt *)mci->pvt_info;
+ pvt->ovrfl_pdev = ovrfl_pdev;
+ pvt->ovrfl_window = ovrfl_window;
+ i82875p_init_csrows(mci, pdev, ovrfl_window, drc);
+ i82875p_get_error_info(mci, &discard); /* clear counters */
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail1;
+ }
+
+ /* allocating generic PCI control info */
+ i82875p_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i82875p_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail1:
+ edac_mc_free(mci);
+
+fail0:
+ iounmap(ovrfl_window);
+ pci_release_regions(ovrfl_pdev);
+
+ pci_disable_device(ovrfl_pdev);
+ /* NOTE: the ovrfl proc entry and pci_dev are intentionally left */
+ return rc;
+}
+
+/* returns count (>= 0), or negative on error */
+static int i82875p_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "\n");
+ i82875p_printk(KERN_INFO, "i82875p init one\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = i82875p_probe1(pdev, ent->driver_data);
+
+ if (mci_pdev == NULL)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i82875p_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct i82875p_pvt *pvt = NULL;
+
+ edac_dbg(0, "\n");
+
+ if (i82875p_pci)
+ edac_pci_release_generic_ctl(i82875p_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ pvt = (struct i82875p_pvt *)mci->pvt_info;
+
+ if (pvt->ovrfl_window)
+ iounmap(pvt->ovrfl_window);
+
+ if (pvt->ovrfl_pdev) {
+#ifdef CORRECT_BIOS
+ pci_release_regions(pvt->ovrfl_pdev);
+#endif /*CORRECT_BIOS */
+ pci_disable_device(pvt->ovrfl_pdev);
+ pci_dev_put(pvt->ovrfl_pdev);
+ }
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i82875p_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 82875_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I82875P},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i82875p_pci_tbl);
+
+static struct pci_driver i82875p_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i82875p_init_one,
+ .remove = i82875p_remove_one,
+ .id_table = i82875p_pci_tbl,
+};
+
+static int __init i82875p_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i82875p_driver);
+
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (mci_pdev == NULL) {
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_82875_0, NULL);
+
+ if (!mci_pdev) {
+ edac_dbg(0, "875p pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i82875p_init_one(mci_pdev, i82875p_pci_tbl);
+
+ if (pci_rc < 0) {
+ edac_dbg(0, "875p init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i82875p_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+ return pci_rc;
+}
+
+static void __exit i82875p_exit(void)
+{
+ edac_dbg(3, "\n");
+
+ i82875p_remove_one(mci_pdev);
+ pci_dev_put(mci_pdev);
+
+ pci_unregister_driver(&i82875p_driver);
+
+}
+
+module_init(i82875p_init);
+module_exit(i82875p_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh");
+MODULE_DESCRIPTION("MC support for Intel 82875 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i82975x_edac.c b/drivers/edac/i82975x_edac.c
new file mode 100644
index 0000000000..d99f005832
--- /dev/null
+++ b/drivers/edac/i82975x_edac.c
@@ -0,0 +1,706 @@
+/*
+ * Intel 82975X Memory Controller kernel module
+ * (C) 2007 aCarLab (India) Pvt. Ltd. (http://acarlab.com)
+ * (C) 2007 jetzbroadband (http://jetzbroadband.com)
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Arvind R.
+ * Copied from i82875p_edac.c source:
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "i82975x_edac"
+
+#define i82975x_printk(level, fmt, arg...) \
+ edac_printk(level, "i82975x", fmt, ##arg)
+
+#define i82975x_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i82975x", fmt, ##arg)
+
+#ifndef PCI_DEVICE_ID_INTEL_82975_0
+#define PCI_DEVICE_ID_INTEL_82975_0 0x277c
+#endif /* PCI_DEVICE_ID_INTEL_82975_0 */
+
+#define I82975X_NR_DIMMS 8
+#define I82975X_NR_CSROWS(nr_chans) (I82975X_NR_DIMMS / (nr_chans))
+
+/* Intel 82975X register addresses - device 0 function 0 - DRAM Controller */
+#define I82975X_EAP 0x58 /* Dram Error Address Pointer (32b)
+ *
+ * 31:7 128 byte cache-line address
+ * 6:1 reserved
+ * 0 0: CH0; 1: CH1
+ */
+
+#define I82975X_DERRSYN 0x5c /* Dram Error SYNdrome (8b)
+ *
+ * 7:0 DRAM ECC Syndrome
+ */
+
+#define I82975X_DES 0x5d /* Dram ERRor DeSTination (8b)
+ * 0h: Processor Memory Reads
+ * 1h:7h reserved
+ * More - See Page 65 of Intel DocSheet.
+ */
+
+#define I82975X_ERRSTS 0xc8 /* Error Status Register (16b)
+ *
+ * 15:12 reserved
+ * 11 Thermal Sensor Event
+ * 10 reserved
+ * 9 non-DRAM lock error (ndlock)
+ * 8 Refresh Timeout
+ * 7:2 reserved
+ * 1 ECC UE (multibit DRAM error)
+ * 0 ECC CE (singlebit DRAM error)
+ */
+
+/* Error Reporting is supported by 3 mechanisms:
+ 1. DMI SERR generation ( ERRCMD )
+ 2. SMI DMI generation ( SMICMD )
+ 3. SCI DMI generation ( SCICMD )
+NOTE: Only ONE of the three must be enabled
+*/
+#define I82975X_ERRCMD 0xca /* Error Command (16b)
+ *
+ * 15:12 reserved
+ * 11 Thermal Sensor Event
+ * 10 reserved
+ * 9 non-DRAM lock error (ndlock)
+ * 8 Refresh Timeout
+ * 7:2 reserved
+ * 1 ECC UE (multibit DRAM error)
+ * 0 ECC CE (singlebit DRAM error)
+ */
+
+#define I82975X_SMICMD 0xcc /* Error Command (16b)
+ *
+ * 15:2 reserved
+ * 1 ECC UE (multibit DRAM error)
+ * 0 ECC CE (singlebit DRAM error)
+ */
+
+#define I82975X_SCICMD 0xce /* Error Command (16b)
+ *
+ * 15:2 reserved
+ * 1 ECC UE (multibit DRAM error)
+ * 0 ECC CE (singlebit DRAM error)
+ */
+
+#define I82975X_XEAP 0xfc /* Extended Dram Error Address Pointer (8b)
+ *
+ * 7:1 reserved
+ * 0 Bit32 of the Dram Error Address
+ */
+
+#define I82975X_MCHBAR 0x44 /*
+ *
+ * 31:14 Base Addr of 16K memory-mapped
+ * configuration space
+ * 13:1 reserved
+ * 0 mem-mapped config space enable
+ */
+
+/* NOTE: Following addresses have to indexed using MCHBAR offset (44h, 32b) */
+/* Intel 82975x memory mapped register space */
+
+#define I82975X_DRB_SHIFT 25 /* fixed 32MiB grain */
+
+#define I82975X_DRB 0x100 /* DRAM Row Boundary (8b x 8)
+ *
+ * 7 set to 1 in highest DRB of
+ * channel if 4GB in ch.
+ * 6:2 upper boundary of rank in
+ * 32MB grains
+ * 1:0 set to 0
+ */
+#define I82975X_DRB_CH0R0 0x100
+#define I82975X_DRB_CH0R1 0x101
+#define I82975X_DRB_CH0R2 0x102
+#define I82975X_DRB_CH0R3 0x103
+#define I82975X_DRB_CH1R0 0x180
+#define I82975X_DRB_CH1R1 0x181
+#define I82975X_DRB_CH1R2 0x182
+#define I82975X_DRB_CH1R3 0x183
+
+
+#define I82975X_DRA 0x108 /* DRAM Row Attribute (4b x 8)
+ * defines the PAGE SIZE to be used
+ * for the rank
+ * 7 reserved
+ * 6:4 row attr of odd rank, i.e. 1
+ * 3 reserved
+ * 2:0 row attr of even rank, i.e. 0
+ *
+ * 000 = unpopulated
+ * 001 = reserved
+ * 010 = 4KiB
+ * 011 = 8KiB
+ * 100 = 16KiB
+ * others = reserved
+ */
+#define I82975X_DRA_CH0R01 0x108
+#define I82975X_DRA_CH0R23 0x109
+#define I82975X_DRA_CH1R01 0x188
+#define I82975X_DRA_CH1R23 0x189
+
+
+#define I82975X_BNKARC 0x10e /* Type of device in each rank - Bank Arch (16b)
+ *
+ * 15:8 reserved
+ * 7:6 Rank 3 architecture
+ * 5:4 Rank 2 architecture
+ * 3:2 Rank 1 architecture
+ * 1:0 Rank 0 architecture
+ *
+ * 00 => 4 banks
+ * 01 => 8 banks
+ */
+#define I82975X_C0BNKARC 0x10e
+#define I82975X_C1BNKARC 0x18e
+
+
+
+#define I82975X_DRC 0x120 /* DRAM Controller Mode0 (32b)
+ *
+ * 31:30 reserved
+ * 29 init complete
+ * 28:11 reserved, according to Intel
+ * 22:21 number of channels
+ * 00=1 01=2 in 82875
+ * seems to be ECC mode
+ * bits in 82975 in Asus
+ * P5W
+ * 19:18 Data Integ Mode
+ * 00=none 01=ECC in 82875
+ * 10:8 refresh mode
+ * 7 reserved
+ * 6:4 mode select
+ * 3:2 reserved
+ * 1:0 DRAM type 10=Second Revision
+ * DDR2 SDRAM
+ * 00, 01, 11 reserved
+ */
+#define I82975X_DRC_CH0M0 0x120
+#define I82975X_DRC_CH1M0 0x1A0
+
+
+#define I82975X_DRC_M1 0x124 /* DRAM Controller Mode1 (32b)
+ * 31 0=Standard Address Map
+ * 1=Enhanced Address Map
+ * 30:0 reserved
+ */
+
+#define I82975X_DRC_CH0M1 0x124
+#define I82975X_DRC_CH1M1 0x1A4
+
+enum i82975x_chips {
+ I82975X = 0,
+};
+
+struct i82975x_pvt {
+ void __iomem *mch_window;
+};
+
+struct i82975x_dev_info {
+ const char *ctl_name;
+};
+
+struct i82975x_error_info {
+ u16 errsts;
+ u32 eap;
+ u8 des;
+ u8 derrsyn;
+ u16 errsts2;
+ u8 chan; /* the channel is bit 0 of EAP */
+ u8 xeap; /* extended eap bit */
+};
+
+static const struct i82975x_dev_info i82975x_devs[] = {
+ [I82975X] = {
+ .ctl_name = "i82975x"
+ },
+};
+
+static struct pci_dev *mci_pdev; /* init dev: in case that AGP code has
+ * already registered driver
+ */
+
+static int i82975x_registered = 1;
+
+static void i82975x_get_error_info(struct mem_ctl_info *mci,
+ struct i82975x_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, I82975X_ERRSTS, &info->errsts);
+ pci_read_config_dword(pdev, I82975X_EAP, &info->eap);
+ pci_read_config_byte(pdev, I82975X_XEAP, &info->xeap);
+ pci_read_config_byte(pdev, I82975X_DES, &info->des);
+ pci_read_config_byte(pdev, I82975X_DERRSYN, &info->derrsyn);
+ pci_read_config_word(pdev, I82975X_ERRSTS, &info->errsts2);
+
+ pci_write_bits16(pdev, I82975X_ERRSTS, 0x0003, 0x0003);
+
+ /*
+ * If the error is the same then we can for both reads then
+ * the first set of reads is valid. If there is a change then
+ * there is a CE no info and the second set of reads is valid
+ * and should be UE info.
+ */
+ if (!(info->errsts2 & 0x0003))
+ return;
+
+ if ((info->errsts ^ info->errsts2) & 0x0003) {
+ pci_read_config_dword(pdev, I82975X_EAP, &info->eap);
+ pci_read_config_byte(pdev, I82975X_XEAP, &info->xeap);
+ pci_read_config_byte(pdev, I82975X_DES, &info->des);
+ pci_read_config_byte(pdev, I82975X_DERRSYN,
+ &info->derrsyn);
+ }
+}
+
+static int i82975x_process_error_info(struct mem_ctl_info *mci,
+ struct i82975x_error_info *info, int handle_errors)
+{
+ int row, chan;
+ unsigned long offst, page;
+
+ if (!(info->errsts2 & 0x0003))
+ return 0;
+
+ if (!handle_errors)
+ return 1;
+
+ if ((info->errsts ^ info->errsts2) & 0x0003) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ page = (unsigned long) info->eap;
+ page >>= 1;
+ if (info->xeap & 1)
+ page |= 0x80000000;
+ page >>= (PAGE_SHIFT - 1);
+ row = edac_mc_find_csrow_by_page(mci, page);
+
+ if (row == -1) {
+ i82975x_mc_printk(mci, KERN_ERR, "error processing EAP:\n"
+ "\tXEAP=%u\n"
+ "\t EAP=0x%08x\n"
+ "\tPAGE=0x%08x\n",
+ (info->xeap & 1) ? 1 : 0, info->eap, (unsigned int) page);
+ return 0;
+ }
+ chan = (mci->csrows[row]->nr_channels == 1) ? 0 : info->eap & 1;
+ offst = info->eap
+ & ((1 << PAGE_SHIFT) -
+ (1 << mci->csrows[row]->channels[chan]->dimm->grain));
+
+ if (info->errsts & 0x0002)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ page, offst, 0,
+ row, -1, -1,
+ "i82975x UE", "");
+ else
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ page, offst, info->derrsyn,
+ row, chan ? chan : 0, -1,
+ "i82975x CE", "");
+
+ return 1;
+}
+
+static void i82975x_check(struct mem_ctl_info *mci)
+{
+ struct i82975x_error_info info;
+
+ i82975x_get_error_info(mci, &info);
+ i82975x_process_error_info(mci, &info, 1);
+}
+
+/* Return 1 if dual channel mode is active. Else return 0. */
+static int dual_channel_active(void __iomem *mch_window)
+{
+ /*
+ * We treat interleaved-symmetric configuration as dual-channel - EAP's
+ * bit-0 giving the channel of the error location.
+ *
+ * All other configurations are treated as single channel - the EAP's
+ * bit-0 will resolve ok in symmetric area of mixed
+ * (symmetric/asymmetric) configurations
+ */
+ u8 drb[4][2];
+ int row;
+ int dualch;
+
+ for (dualch = 1, row = 0; dualch && (row < 4); row++) {
+ drb[row][0] = readb(mch_window + I82975X_DRB + row);
+ drb[row][1] = readb(mch_window + I82975X_DRB + row + 0x80);
+ dualch = dualch && (drb[row][0] == drb[row][1]);
+ }
+ return dualch;
+}
+
+static void i82975x_init_csrows(struct mem_ctl_info *mci,
+ struct pci_dev *pdev, void __iomem *mch_window)
+{
+ struct csrow_info *csrow;
+ unsigned long last_cumul_size;
+ u8 value;
+ u32 cumul_size, nr_pages;
+ int index, chan;
+ struct dimm_info *dimm;
+
+ last_cumul_size = 0;
+
+ /*
+ * 82875 comment:
+ * The dram row boundary (DRB) reg values are boundary address
+ * for each DRAM row with a granularity of 32 or 64MB (single/dual
+ * channel operation). DRB regs are cumulative; therefore DRB7 will
+ * contain the total memory contained in all rows.
+ *
+ */
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+
+ value = readb(mch_window + I82975X_DRB + index +
+ ((index >= 4) ? 0x80 : 0));
+ cumul_size = value;
+ cumul_size <<= (I82975X_DRB_SHIFT - PAGE_SHIFT);
+ /*
+ * Adjust cumul_size w.r.t number of channels
+ *
+ */
+ if (csrow->nr_channels > 1)
+ cumul_size <<= 1;
+ edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
+
+ nr_pages = cumul_size - last_cumul_size;
+ if (!nr_pages)
+ continue;
+
+ /*
+ * Initialise dram labels
+ * index values:
+ * [0-7] for single-channel; i.e. csrow->nr_channels = 1
+ * [0-3] for dual-channel; i.e. csrow->nr_channels = 2
+ */
+ for (chan = 0; chan < csrow->nr_channels; chan++) {
+ dimm = mci->csrows[index]->channels[chan]->dimm;
+
+ dimm->nr_pages = nr_pages / csrow->nr_channels;
+
+ snprintf(csrow->channels[chan]->dimm->label, EDAC_MC_LABEL_LEN, "DIMM %c%d",
+ (chan == 0) ? 'A' : 'B',
+ index);
+ dimm->grain = 1 << 7; /* 128Byte cache-line resolution */
+
+ /* ECC is possible on i92975x ONLY with DEV_X8. */
+ dimm->dtype = DEV_X8;
+
+ dimm->mtype = MEM_DDR2; /* I82975x supports only DDR2 */
+ dimm->edac_mode = EDAC_SECDED; /* only supported */
+ }
+
+ csrow->first_page = last_cumul_size;
+ csrow->last_page = cumul_size - 1;
+ last_cumul_size = cumul_size;
+ }
+}
+
+/* #define i82975x_DEBUG_IOMEM */
+
+#ifdef i82975x_DEBUG_IOMEM
+static void i82975x_print_dram_timings(void __iomem *mch_window)
+{
+ /*
+ * The register meanings are from Intel specs;
+ * (shows 13-5-5-5 for 800-DDR2)
+ * Asus P5W Bios reports 15-5-4-4
+ * What's your religion?
+ */
+ static const int caslats[4] = { 5, 4, 3, 6 };
+ u32 dtreg[2];
+
+ dtreg[0] = readl(mch_window + 0x114);
+ dtreg[1] = readl(mch_window + 0x194);
+ i82975x_printk(KERN_INFO, "DRAM Timings : Ch0 Ch1\n"
+ " RAS Active Min = %d %d\n"
+ " CAS latency = %d %d\n"
+ " RAS to CAS = %d %d\n"
+ " RAS precharge = %d %d\n",
+ (dtreg[0] >> 19 ) & 0x0f,
+ (dtreg[1] >> 19) & 0x0f,
+ caslats[(dtreg[0] >> 8) & 0x03],
+ caslats[(dtreg[1] >> 8) & 0x03],
+ ((dtreg[0] >> 4) & 0x07) + 2,
+ ((dtreg[1] >> 4) & 0x07) + 2,
+ (dtreg[0] & 0x07) + 2,
+ (dtreg[1] & 0x07) + 2
+ );
+
+}
+#endif
+
+static int i82975x_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int rc = -ENODEV;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct i82975x_pvt *pvt;
+ void __iomem *mch_window;
+ u32 mchbar;
+ u32 drc[2];
+ struct i82975x_error_info discard;
+ int chans;
+#ifdef i82975x_DEBUG_IOMEM
+ u8 c0drb[4];
+ u8 c1drb[4];
+#endif
+
+ edac_dbg(0, "\n");
+
+ pci_read_config_dword(pdev, I82975X_MCHBAR, &mchbar);
+ if (!(mchbar & 1)) {
+ edac_dbg(3, "failed, MCHBAR disabled!\n");
+ goto fail0;
+ }
+ mchbar &= 0xffffc000; /* bits 31:14 used for 16K window */
+ mch_window = ioremap(mchbar, 0x1000);
+ if (!mch_window) {
+ edac_dbg(3, "error ioremapping MCHBAR!\n");
+ goto fail0;
+ }
+
+#ifdef i82975x_DEBUG_IOMEM
+ i82975x_printk(KERN_INFO, "MCHBAR real = %0x, remapped = %p\n",
+ mchbar, mch_window);
+
+ c0drb[0] = readb(mch_window + I82975X_DRB_CH0R0);
+ c0drb[1] = readb(mch_window + I82975X_DRB_CH0R1);
+ c0drb[2] = readb(mch_window + I82975X_DRB_CH0R2);
+ c0drb[3] = readb(mch_window + I82975X_DRB_CH0R3);
+ c1drb[0] = readb(mch_window + I82975X_DRB_CH1R0);
+ c1drb[1] = readb(mch_window + I82975X_DRB_CH1R1);
+ c1drb[2] = readb(mch_window + I82975X_DRB_CH1R2);
+ c1drb[3] = readb(mch_window + I82975X_DRB_CH1R3);
+ i82975x_printk(KERN_INFO, "DRBCH0R0 = 0x%02x\n", c0drb[0]);
+ i82975x_printk(KERN_INFO, "DRBCH0R1 = 0x%02x\n", c0drb[1]);
+ i82975x_printk(KERN_INFO, "DRBCH0R2 = 0x%02x\n", c0drb[2]);
+ i82975x_printk(KERN_INFO, "DRBCH0R3 = 0x%02x\n", c0drb[3]);
+ i82975x_printk(KERN_INFO, "DRBCH1R0 = 0x%02x\n", c1drb[0]);
+ i82975x_printk(KERN_INFO, "DRBCH1R1 = 0x%02x\n", c1drb[1]);
+ i82975x_printk(KERN_INFO, "DRBCH1R2 = 0x%02x\n", c1drb[2]);
+ i82975x_printk(KERN_INFO, "DRBCH1R3 = 0x%02x\n", c1drb[3]);
+#endif
+
+ drc[0] = readl(mch_window + I82975X_DRC_CH0M0);
+ drc[1] = readl(mch_window + I82975X_DRC_CH1M0);
+#ifdef i82975x_DEBUG_IOMEM
+ i82975x_printk(KERN_INFO, "DRC_CH0 = %0x, %s\n", drc[0],
+ ((drc[0] >> 21) & 3) == 1 ?
+ "ECC enabled" : "ECC disabled");
+ i82975x_printk(KERN_INFO, "DRC_CH1 = %0x, %s\n", drc[1],
+ ((drc[1] >> 21) & 3) == 1 ?
+ "ECC enabled" : "ECC disabled");
+
+ i82975x_printk(KERN_INFO, "C0 BNKARC = %0x\n",
+ readw(mch_window + I82975X_C0BNKARC));
+ i82975x_printk(KERN_INFO, "C1 BNKARC = %0x\n",
+ readw(mch_window + I82975X_C1BNKARC));
+ i82975x_print_dram_timings(mch_window);
+ goto fail1;
+#endif
+ if (!(((drc[0] >> 21) & 3) == 1 || ((drc[1] >> 21) & 3) == 1)) {
+ i82975x_printk(KERN_INFO, "ECC disabled on both channels.\n");
+ goto fail1;
+ }
+
+ chans = dual_channel_active(mch_window) + 1;
+
+ /* assuming only one controller, index thus is 0 */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = I82975X_NR_DIMMS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = I82975X_NR_CSROWS(chans);
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (!mci) {
+ rc = -ENOMEM;
+ goto fail1;
+ }
+
+ edac_dbg(3, "init mci\n");
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = i82975x_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = i82975x_check;
+ mci->ctl_page_to_phys = NULL;
+ edac_dbg(3, "init pvt\n");
+ pvt = (struct i82975x_pvt *) mci->pvt_info;
+ pvt->mch_window = mch_window;
+ i82975x_init_csrows(mci, pdev, mch_window);
+ mci->scrub_mode = SCRUB_HW_SRC;
+ i82975x_get_error_info(mci, &discard); /* clear counters */
+
+ /* finalize this instance of memory controller with edac core */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail2;
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail2:
+ edac_mc_free(mci);
+
+fail1:
+ iounmap(mch_window);
+fail0:
+ return rc;
+}
+
+/* returns count (>= 0), or negative on error */
+static int i82975x_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = i82975x_probe1(pdev, ent->driver_data);
+
+ if (mci_pdev == NULL)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void i82975x_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct i82975x_pvt *pvt;
+
+ edac_dbg(0, "\n");
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (mci == NULL)
+ return;
+
+ pvt = mci->pvt_info;
+ if (pvt->mch_window)
+ iounmap( pvt->mch_window );
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id i82975x_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, 82975_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ I82975X
+ },
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i82975x_pci_tbl);
+
+static struct pci_driver i82975x_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = i82975x_init_one,
+ .remove = i82975x_remove_one,
+ .id_table = i82975x_pci_tbl,
+};
+
+static int __init i82975x_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i82975x_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (mci_pdev == NULL) {
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_82975_0, NULL);
+
+ if (!mci_pdev) {
+ edac_dbg(0, "i82975x pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = i82975x_init_one(mci_pdev, i82975x_pci_tbl);
+
+ if (pci_rc < 0) {
+ edac_dbg(0, "i82975x init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&i82975x_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+ return pci_rc;
+}
+
+static void __exit i82975x_exit(void)
+{
+ edac_dbg(3, "\n");
+
+ pci_unregister_driver(&i82975x_driver);
+
+ if (!i82975x_registered) {
+ i82975x_remove_one(mci_pdev);
+ pci_dev_put(mci_pdev);
+ }
+}
+
+module_init(i82975x_init);
+module_exit(i82975x_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Arvind R. <arvino55@gmail.com>");
+MODULE_DESCRIPTION("MC support for Intel 82975 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/ie31200_edac.c b/drivers/edac/ie31200_edac.c
new file mode 100644
index 0000000000..9ef13570f2
--- /dev/null
+++ b/drivers/edac/ie31200_edac.c
@@ -0,0 +1,667 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Intel E3-1200
+ * Copyright (C) 2014 Jason Baron <jbaron@akamai.com>
+ *
+ * Support for the E3-1200 processor family. Heavily based on previous
+ * Intel EDAC drivers.
+ *
+ * Since the DRAM controller is on the cpu chip, we can use its PCI device
+ * id to identify these processors.
+ *
+ * PCI DRAM controller device ids (Taken from The PCI ID Repository - https://pci-ids.ucw.cz/)
+ *
+ * 0108: Xeon E3-1200 Processor Family DRAM Controller
+ * 010c: Xeon E3-1200/2nd Generation Core Processor Family DRAM Controller
+ * 0150: Xeon E3-1200 v2/3rd Gen Core processor DRAM Controller
+ * 0158: Xeon E3-1200 v2/Ivy Bridge DRAM Controller
+ * 015c: Xeon E3-1200 v2/3rd Gen Core processor DRAM Controller
+ * 0c04: Xeon E3-1200 v3/4th Gen Core Processor DRAM Controller
+ * 0c08: Xeon E3-1200 v3 Processor DRAM Controller
+ * 1918: Xeon E3-1200 v5 Skylake Host Bridge/DRAM Registers
+ * 5918: Xeon E3-1200 Xeon E3-1200 v6/7th Gen Core Processor Host Bridge/DRAM Registers
+ * 190f: 6th Gen Core Dual-Core Processor Host Bridge/DRAM Registers
+ * 191f: 6th Gen Core Quad-Core Processor Host Bridge/DRAM Registers
+ * 3e..: 8th/9th Gen Core Processor Host Bridge/DRAM Registers
+ *
+ * Based on Intel specification:
+ * https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e3-1200v3-vol-2-datasheet.pdf
+ * http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200-family-vol-2-datasheet.html
+ * https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/desktop-6th-gen-core-family-datasheet-vol-2.pdf
+ * https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e3-1200v6-vol-2-datasheet.pdf
+ * https://www.intel.com/content/www/us/en/processors/core/7th-gen-core-family-mobile-h-processor-lines-datasheet-vol-2.html
+ * https://www.intel.com/content/www/us/en/products/docs/processors/core/8th-gen-core-family-datasheet-vol-2.html
+ *
+ * According to the above datasheet (p.16):
+ * "
+ * 6. Software must not access B0/D0/F0 32-bit memory-mapped registers with
+ * requests that cross a DW boundary.
+ * "
+ *
+ * Thus, we make use of the explicit: lo_hi_readq(), which breaks the readq into
+ * 2 readl() calls. This restriction may be lifted in subsequent chip releases,
+ * but lo_hi_readq() ensures that we are safe across all e3-1200 processors.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+
+#include <linux/io-64-nonatomic-lo-hi.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "ie31200_edac"
+
+#define ie31200_printk(level, fmt, arg...) \
+ edac_printk(level, "ie31200", fmt, ##arg)
+
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_1 0x0108
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_2 0x010c
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_3 0x0150
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_4 0x0158
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_5 0x015c
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_6 0x0c04
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_7 0x0c08
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_8 0x190F
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_9 0x1918
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_10 0x191F
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_11 0x5918
+
+/* Coffee Lake-S */
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_MASK 0x3e00
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_1 0x3e0f
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_2 0x3e18
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_3 0x3e1f
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_4 0x3e30
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_5 0x3e31
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_6 0x3e32
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_7 0x3e33
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_8 0x3ec2
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_9 0x3ec6
+#define PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_10 0x3eca
+
+/* Test if HB is for Skylake or later. */
+#define DEVICE_ID_SKYLAKE_OR_LATER(did) \
+ (((did) == PCI_DEVICE_ID_INTEL_IE31200_HB_8) || \
+ ((did) == PCI_DEVICE_ID_INTEL_IE31200_HB_9) || \
+ ((did) == PCI_DEVICE_ID_INTEL_IE31200_HB_10) || \
+ ((did) == PCI_DEVICE_ID_INTEL_IE31200_HB_11) || \
+ (((did) & PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_MASK) == \
+ PCI_DEVICE_ID_INTEL_IE31200_HB_CFL_MASK))
+
+#define IE31200_DIMMS 4
+#define IE31200_RANKS 8
+#define IE31200_RANKS_PER_CHANNEL 4
+#define IE31200_DIMMS_PER_CHANNEL 2
+#define IE31200_CHANNELS 2
+
+/* Intel IE31200 register addresses - device 0 function 0 - DRAM Controller */
+#define IE31200_MCHBAR_LOW 0x48
+#define IE31200_MCHBAR_HIGH 0x4c
+#define IE31200_MCHBAR_MASK GENMASK_ULL(38, 15)
+#define IE31200_MMR_WINDOW_SIZE BIT(15)
+
+/*
+ * Error Status Register (16b)
+ *
+ * 15 reserved
+ * 14 Isochronous TBWRR Run Behind FIFO Full
+ * (ITCV)
+ * 13 Isochronous TBWRR Run Behind FIFO Put
+ * (ITSTV)
+ * 12 reserved
+ * 11 MCH Thermal Sensor Event
+ * for SMI/SCI/SERR (GTSE)
+ * 10 reserved
+ * 9 LOCK to non-DRAM Memory Flag (LCKF)
+ * 8 reserved
+ * 7 DRAM Throttle Flag (DTF)
+ * 6:2 reserved
+ * 1 Multi-bit DRAM ECC Error Flag (DMERR)
+ * 0 Single-bit DRAM ECC Error Flag (DSERR)
+ */
+#define IE31200_ERRSTS 0xc8
+#define IE31200_ERRSTS_UE BIT(1)
+#define IE31200_ERRSTS_CE BIT(0)
+#define IE31200_ERRSTS_BITS (IE31200_ERRSTS_UE | IE31200_ERRSTS_CE)
+
+/*
+ * Channel 0 ECC Error Log (64b)
+ *
+ * 63:48 Error Column Address (ERRCOL)
+ * 47:32 Error Row Address (ERRROW)
+ * 31:29 Error Bank Address (ERRBANK)
+ * 28:27 Error Rank Address (ERRRANK)
+ * 26:24 reserved
+ * 23:16 Error Syndrome (ERRSYND)
+ * 15: 2 reserved
+ * 1 Multiple Bit Error Status (MERRSTS)
+ * 0 Correctable Error Status (CERRSTS)
+ */
+
+#define IE31200_C0ECCERRLOG 0x40c8
+#define IE31200_C1ECCERRLOG 0x44c8
+#define IE31200_C0ECCERRLOG_SKL 0x4048
+#define IE31200_C1ECCERRLOG_SKL 0x4448
+#define IE31200_ECCERRLOG_CE BIT(0)
+#define IE31200_ECCERRLOG_UE BIT(1)
+#define IE31200_ECCERRLOG_RANK_BITS GENMASK_ULL(28, 27)
+#define IE31200_ECCERRLOG_RANK_SHIFT 27
+#define IE31200_ECCERRLOG_SYNDROME_BITS GENMASK_ULL(23, 16)
+#define IE31200_ECCERRLOG_SYNDROME_SHIFT 16
+
+#define IE31200_ECCERRLOG_SYNDROME(log) \
+ ((log & IE31200_ECCERRLOG_SYNDROME_BITS) >> \
+ IE31200_ECCERRLOG_SYNDROME_SHIFT)
+
+#define IE31200_CAPID0 0xe4
+#define IE31200_CAPID0_PDCD BIT(4)
+#define IE31200_CAPID0_DDPCD BIT(6)
+#define IE31200_CAPID0_ECC BIT(1)
+
+#define IE31200_MAD_DIMM_0_OFFSET 0x5004
+#define IE31200_MAD_DIMM_0_OFFSET_SKL 0x500C
+#define IE31200_MAD_DIMM_SIZE GENMASK_ULL(7, 0)
+#define IE31200_MAD_DIMM_A_RANK BIT(17)
+#define IE31200_MAD_DIMM_A_RANK_SHIFT 17
+#define IE31200_MAD_DIMM_A_RANK_SKL BIT(10)
+#define IE31200_MAD_DIMM_A_RANK_SKL_SHIFT 10
+#define IE31200_MAD_DIMM_A_WIDTH BIT(19)
+#define IE31200_MAD_DIMM_A_WIDTH_SHIFT 19
+#define IE31200_MAD_DIMM_A_WIDTH_SKL GENMASK_ULL(9, 8)
+#define IE31200_MAD_DIMM_A_WIDTH_SKL_SHIFT 8
+
+/* Skylake reports 1GB increments, everything else is 256MB */
+#define IE31200_PAGES(n, skl) \
+ (n << (28 + (2 * skl) - PAGE_SHIFT))
+
+static int nr_channels;
+static struct pci_dev *mci_pdev;
+static int ie31200_registered = 1;
+
+struct ie31200_priv {
+ void __iomem *window;
+ void __iomem *c0errlog;
+ void __iomem *c1errlog;
+};
+
+enum ie31200_chips {
+ IE31200 = 0,
+};
+
+struct ie31200_dev_info {
+ const char *ctl_name;
+};
+
+struct ie31200_error_info {
+ u16 errsts;
+ u16 errsts2;
+ u64 eccerrlog[IE31200_CHANNELS];
+};
+
+static const struct ie31200_dev_info ie31200_devs[] = {
+ [IE31200] = {
+ .ctl_name = "IE31200"
+ },
+};
+
+struct dimm_data {
+ u8 size; /* in multiples of 256MB, except Skylake is 1GB */
+ u8 dual_rank : 1,
+ x16_width : 2; /* 0 means x8 width */
+};
+
+static int how_many_channels(struct pci_dev *pdev)
+{
+ int n_channels;
+ unsigned char capid0_2b; /* 2nd byte of CAPID0 */
+
+ pci_read_config_byte(pdev, IE31200_CAPID0 + 1, &capid0_2b);
+
+ /* check PDCD: Dual Channel Disable */
+ if (capid0_2b & IE31200_CAPID0_PDCD) {
+ edac_dbg(0, "In single channel mode\n");
+ n_channels = 1;
+ } else {
+ edac_dbg(0, "In dual channel mode\n");
+ n_channels = 2;
+ }
+
+ /* check DDPCD - check if both channels are filled */
+ if (capid0_2b & IE31200_CAPID0_DDPCD)
+ edac_dbg(0, "2 DIMMS per channel disabled\n");
+ else
+ edac_dbg(0, "2 DIMMS per channel enabled\n");
+
+ return n_channels;
+}
+
+static bool ecc_capable(struct pci_dev *pdev)
+{
+ unsigned char capid0_4b; /* 4th byte of CAPID0 */
+
+ pci_read_config_byte(pdev, IE31200_CAPID0 + 3, &capid0_4b);
+ if (capid0_4b & IE31200_CAPID0_ECC)
+ return false;
+ return true;
+}
+
+static int eccerrlog_row(u64 log)
+{
+ return ((log & IE31200_ECCERRLOG_RANK_BITS) >>
+ IE31200_ECCERRLOG_RANK_SHIFT);
+}
+
+static void ie31200_clear_error_info(struct mem_ctl_info *mci)
+{
+ /*
+ * Clear any error bits.
+ * (Yes, we really clear bits by writing 1 to them.)
+ */
+ pci_write_bits16(to_pci_dev(mci->pdev), IE31200_ERRSTS,
+ IE31200_ERRSTS_BITS, IE31200_ERRSTS_BITS);
+}
+
+static void ie31200_get_and_clear_error_info(struct mem_ctl_info *mci,
+ struct ie31200_error_info *info)
+{
+ struct pci_dev *pdev;
+ struct ie31200_priv *priv = mci->pvt_info;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, IE31200_ERRSTS, &info->errsts);
+ if (!(info->errsts & IE31200_ERRSTS_BITS))
+ return;
+
+ info->eccerrlog[0] = lo_hi_readq(priv->c0errlog);
+ if (nr_channels == 2)
+ info->eccerrlog[1] = lo_hi_readq(priv->c1errlog);
+
+ pci_read_config_word(pdev, IE31200_ERRSTS, &info->errsts2);
+
+ /*
+ * If the error is the same for both reads then the first set
+ * of reads is valid. If there is a change then there is a CE
+ * with no info and the second set of reads is valid and
+ * should be UE info.
+ */
+ if ((info->errsts ^ info->errsts2) & IE31200_ERRSTS_BITS) {
+ info->eccerrlog[0] = lo_hi_readq(priv->c0errlog);
+ if (nr_channels == 2)
+ info->eccerrlog[1] =
+ lo_hi_readq(priv->c1errlog);
+ }
+
+ ie31200_clear_error_info(mci);
+}
+
+static void ie31200_process_error_info(struct mem_ctl_info *mci,
+ struct ie31200_error_info *info)
+{
+ int channel;
+ u64 log;
+
+ if (!(info->errsts & IE31200_ERRSTS_BITS))
+ return;
+
+ if ((info->errsts ^ info->errsts2) & IE31200_ERRSTS_BITS) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ for (channel = 0; channel < nr_channels; channel++) {
+ log = info->eccerrlog[channel];
+ if (log & IE31200_ECCERRLOG_UE) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ 0, 0, 0,
+ eccerrlog_row(log),
+ channel, -1,
+ "ie31200 UE", "");
+ } else if (log & IE31200_ECCERRLOG_CE) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ 0, 0,
+ IE31200_ECCERRLOG_SYNDROME(log),
+ eccerrlog_row(log),
+ channel, -1,
+ "ie31200 CE", "");
+ }
+ }
+}
+
+static void ie31200_check(struct mem_ctl_info *mci)
+{
+ struct ie31200_error_info info;
+
+ ie31200_get_and_clear_error_info(mci, &info);
+ ie31200_process_error_info(mci, &info);
+}
+
+static void __iomem *ie31200_map_mchbar(struct pci_dev *pdev)
+{
+ union {
+ u64 mchbar;
+ struct {
+ u32 mchbar_low;
+ u32 mchbar_high;
+ };
+ } u;
+ void __iomem *window;
+
+ pci_read_config_dword(pdev, IE31200_MCHBAR_LOW, &u.mchbar_low);
+ pci_read_config_dword(pdev, IE31200_MCHBAR_HIGH, &u.mchbar_high);
+ u.mchbar &= IE31200_MCHBAR_MASK;
+
+ if (u.mchbar != (resource_size_t)u.mchbar) {
+ ie31200_printk(KERN_ERR, "mmio space beyond accessible range (0x%llx)\n",
+ (unsigned long long)u.mchbar);
+ return NULL;
+ }
+
+ window = ioremap(u.mchbar, IE31200_MMR_WINDOW_SIZE);
+ if (!window)
+ ie31200_printk(KERN_ERR, "Cannot map mmio space at 0x%llx\n",
+ (unsigned long long)u.mchbar);
+
+ return window;
+}
+
+static void __skl_populate_dimm_info(struct dimm_data *dd, u32 addr_decode,
+ int chan)
+{
+ dd->size = (addr_decode >> (chan << 4)) & IE31200_MAD_DIMM_SIZE;
+ dd->dual_rank = (addr_decode & (IE31200_MAD_DIMM_A_RANK_SKL << (chan << 4))) ? 1 : 0;
+ dd->x16_width = ((addr_decode & (IE31200_MAD_DIMM_A_WIDTH_SKL << (chan << 4))) >>
+ (IE31200_MAD_DIMM_A_WIDTH_SKL_SHIFT + (chan << 4)));
+}
+
+static void __populate_dimm_info(struct dimm_data *dd, u32 addr_decode,
+ int chan)
+{
+ dd->size = (addr_decode >> (chan << 3)) & IE31200_MAD_DIMM_SIZE;
+ dd->dual_rank = (addr_decode & (IE31200_MAD_DIMM_A_RANK << chan)) ? 1 : 0;
+ dd->x16_width = (addr_decode & (IE31200_MAD_DIMM_A_WIDTH << chan)) ? 1 : 0;
+}
+
+static void populate_dimm_info(struct dimm_data *dd, u32 addr_decode, int chan,
+ bool skl)
+{
+ if (skl)
+ __skl_populate_dimm_info(dd, addr_decode, chan);
+ else
+ __populate_dimm_info(dd, addr_decode, chan);
+}
+
+
+static int ie31200_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int i, j, ret;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ struct dimm_data dimm_info[IE31200_CHANNELS][IE31200_DIMMS_PER_CHANNEL];
+ void __iomem *window;
+ struct ie31200_priv *priv;
+ u32 addr_decode, mad_offset;
+
+ /*
+ * Kaby Lake, Coffee Lake seem to work like Skylake. Please re-visit
+ * this logic when adding new CPU support.
+ */
+ bool skl = DEVICE_ID_SKYLAKE_OR_LATER(pdev->device);
+
+ edac_dbg(0, "MC:\n");
+
+ if (!ecc_capable(pdev)) {
+ ie31200_printk(KERN_INFO, "No ECC support\n");
+ return -ENODEV;
+ }
+
+ nr_channels = how_many_channels(pdev);
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = IE31200_DIMMS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = nr_channels;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct ie31200_priv));
+ if (!mci)
+ return -ENOMEM;
+
+ window = ie31200_map_mchbar(pdev);
+ if (!window) {
+ ret = -ENODEV;
+ goto fail_free;
+ }
+
+ edac_dbg(3, "MC: init mci\n");
+ mci->pdev = &pdev->dev;
+ if (skl)
+ mci->mtype_cap = MEM_FLAG_DDR4;
+ else
+ mci->mtype_cap = MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = ie31200_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = ie31200_check;
+ mci->ctl_page_to_phys = NULL;
+ priv = mci->pvt_info;
+ priv->window = window;
+ if (skl) {
+ priv->c0errlog = window + IE31200_C0ECCERRLOG_SKL;
+ priv->c1errlog = window + IE31200_C1ECCERRLOG_SKL;
+ mad_offset = IE31200_MAD_DIMM_0_OFFSET_SKL;
+ } else {
+ priv->c0errlog = window + IE31200_C0ECCERRLOG;
+ priv->c1errlog = window + IE31200_C1ECCERRLOG;
+ mad_offset = IE31200_MAD_DIMM_0_OFFSET;
+ }
+
+ /* populate DIMM info */
+ for (i = 0; i < IE31200_CHANNELS; i++) {
+ addr_decode = readl(window + mad_offset +
+ (i * 4));
+ edac_dbg(0, "addr_decode: 0x%x\n", addr_decode);
+ for (j = 0; j < IE31200_DIMMS_PER_CHANNEL; j++) {
+ populate_dimm_info(&dimm_info[i][j], addr_decode, j,
+ skl);
+ edac_dbg(0, "size: 0x%x, rank: %d, width: %d\n",
+ dimm_info[i][j].size,
+ dimm_info[i][j].dual_rank,
+ dimm_info[i][j].x16_width);
+ }
+ }
+
+ /*
+ * The dram rank boundary (DRB) reg values are boundary addresses
+ * for each DRAM rank with a granularity of 64MB. DRB regs are
+ * cumulative; the last one will contain the total memory
+ * contained in all ranks.
+ */
+ for (i = 0; i < IE31200_DIMMS_PER_CHANNEL; i++) {
+ for (j = 0; j < IE31200_CHANNELS; j++) {
+ struct dimm_info *dimm;
+ unsigned long nr_pages;
+
+ nr_pages = IE31200_PAGES(dimm_info[j][i].size, skl);
+ if (nr_pages == 0)
+ continue;
+
+ if (dimm_info[j][i].dual_rank) {
+ nr_pages = nr_pages / 2;
+ dimm = edac_get_dimm(mci, (i * 2) + 1, j, 0);
+ dimm->nr_pages = nr_pages;
+ edac_dbg(0, "set nr pages: 0x%lx\n", nr_pages);
+ dimm->grain = 8; /* just a guess */
+ if (skl)
+ dimm->mtype = MEM_DDR4;
+ else
+ dimm->mtype = MEM_DDR3;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_UNKNOWN;
+ }
+ dimm = edac_get_dimm(mci, i * 2, j, 0);
+ dimm->nr_pages = nr_pages;
+ edac_dbg(0, "set nr pages: 0x%lx\n", nr_pages);
+ dimm->grain = 8; /* same guess */
+ if (skl)
+ dimm->mtype = MEM_DDR4;
+ else
+ dimm->mtype = MEM_DDR3;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_UNKNOWN;
+ }
+ }
+
+ ie31200_clear_error_info(mci);
+
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "MC: failed edac_mc_add_mc()\n");
+ ret = -ENODEV;
+ goto fail_unmap;
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "MC: success\n");
+ return 0;
+
+fail_unmap:
+ iounmap(window);
+
+fail_free:
+ edac_mc_free(mci);
+
+ return ret;
+}
+
+static int ie31200_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+ rc = ie31200_probe1(pdev, ent->driver_data);
+ if (rc == 0 && !mci_pdev)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void ie31200_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct ie31200_priv *priv;
+
+ edac_dbg(0, "\n");
+ pci_dev_put(mci_pdev);
+ mci_pdev = NULL;
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+ priv = mci->pvt_info;
+ iounmap(priv->window);
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id ie31200_pci_tbl[] = {
+ { PCI_VEND_DEV(INTEL, IE31200_HB_1), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_2), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_3), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_4), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_5), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_6), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_7), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_8), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_9), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_10), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_11), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_1), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_2), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_3), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_4), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_5), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_6), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_7), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_8), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_9), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { PCI_VEND_DEV(INTEL, IE31200_HB_CFL_10), PCI_ANY_ID, PCI_ANY_ID, 0, 0, IE31200 },
+ { 0, } /* 0 terminated list. */
+};
+MODULE_DEVICE_TABLE(pci, ie31200_pci_tbl);
+
+static struct pci_driver ie31200_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = ie31200_init_one,
+ .remove = ie31200_remove_one,
+ .id_table = ie31200_pci_tbl,
+};
+
+static int __init ie31200_init(void)
+{
+ int pci_rc, i;
+
+ edac_dbg(3, "MC:\n");
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&ie31200_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (!mci_pdev) {
+ ie31200_registered = 0;
+ for (i = 0; ie31200_pci_tbl[i].vendor != 0; i++) {
+ mci_pdev = pci_get_device(ie31200_pci_tbl[i].vendor,
+ ie31200_pci_tbl[i].device,
+ NULL);
+ if (mci_pdev)
+ break;
+ }
+ if (!mci_pdev) {
+ edac_dbg(0, "ie31200 pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ pci_rc = ie31200_init_one(mci_pdev, &ie31200_pci_tbl[i]);
+ if (pci_rc < 0) {
+ edac_dbg(0, "ie31200 init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+ return 0;
+
+fail1:
+ pci_unregister_driver(&ie31200_driver);
+fail0:
+ pci_dev_put(mci_pdev);
+
+ return pci_rc;
+}
+
+static void __exit ie31200_exit(void)
+{
+ edac_dbg(3, "MC:\n");
+ pci_unregister_driver(&ie31200_driver);
+ if (!ie31200_registered)
+ ie31200_remove_one(mci_pdev);
+}
+
+module_init(ie31200_init);
+module_exit(ie31200_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jason Baron <jbaron@akamai.com>");
+MODULE_DESCRIPTION("MC support for Intel Processor E31200 memory hub controllers");
diff --git a/drivers/edac/igen6_edac.c b/drivers/edac/igen6_edac.c
new file mode 100644
index 0000000000..1a18693294
--- /dev/null
+++ b/drivers/edac/igen6_edac.c
@@ -0,0 +1,1304 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Driver for Intel client SoC with integrated memory controller using IBECC
+ *
+ * Copyright (C) 2020 Intel Corporation
+ *
+ * The In-Band ECC (IBECC) IP provides ECC protection to all or specific
+ * regions of the physical memory space. It's used for memory controllers
+ * that don't support the out-of-band ECC which often needs an additional
+ * storage device to each channel for storing ECC data.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/irq_work.h>
+#include <linux/llist.h>
+#include <linux/genalloc.h>
+#include <linux/edac.h>
+#include <linux/bits.h>
+#include <linux/io.h>
+#include <asm/mach_traps.h>
+#include <asm/nmi.h>
+#include <asm/mce.h>
+
+#include "edac_mc.h"
+#include "edac_module.h"
+
+#define IGEN6_REVISION "v2.5.1"
+
+#define EDAC_MOD_STR "igen6_edac"
+#define IGEN6_NMI_NAME "igen6_ibecc"
+
+/* Debug macros */
+#define igen6_printk(level, fmt, arg...) \
+ edac_printk(level, "igen6", fmt, ##arg)
+
+#define igen6_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "igen6", fmt, ##arg)
+
+#define GET_BITFIELD(v, lo, hi) (((v) & GENMASK_ULL(hi, lo)) >> (lo))
+
+#define NUM_IMC 2 /* Max memory controllers */
+#define NUM_CHANNELS 2 /* Max channels */
+#define NUM_DIMMS 2 /* Max DIMMs per channel */
+
+#define _4GB BIT_ULL(32)
+
+/* Size of physical memory */
+#define TOM_OFFSET 0xa0
+/* Top of low usable DRAM */
+#define TOLUD_OFFSET 0xbc
+/* Capability register C */
+#define CAPID_C_OFFSET 0xec
+#define CAPID_C_IBECC BIT(15)
+
+/* Capability register E */
+#define CAPID_E_OFFSET 0xf0
+#define CAPID_E_IBECC BIT(12)
+
+/* Error Status */
+#define ERRSTS_OFFSET 0xc8
+#define ERRSTS_CE BIT_ULL(6)
+#define ERRSTS_UE BIT_ULL(7)
+
+/* Error Command */
+#define ERRCMD_OFFSET 0xca
+#define ERRCMD_CE BIT_ULL(6)
+#define ERRCMD_UE BIT_ULL(7)
+
+/* IBECC MMIO base address */
+#define IBECC_BASE (res_cfg->ibecc_base)
+#define IBECC_ACTIVATE_OFFSET IBECC_BASE
+#define IBECC_ACTIVATE_EN BIT(0)
+
+/* IBECC error log */
+#define ECC_ERROR_LOG_OFFSET (IBECC_BASE + res_cfg->ibecc_error_log_offset)
+#define ECC_ERROR_LOG_CE BIT_ULL(62)
+#define ECC_ERROR_LOG_UE BIT_ULL(63)
+#define ECC_ERROR_LOG_ADDR_SHIFT 5
+#define ECC_ERROR_LOG_ADDR(v) GET_BITFIELD(v, 5, 38)
+#define ECC_ERROR_LOG_SYND(v) GET_BITFIELD(v, 46, 61)
+
+/* Host MMIO base address */
+#define MCHBAR_OFFSET 0x48
+#define MCHBAR_EN BIT_ULL(0)
+#define MCHBAR_BASE(v) (GET_BITFIELD(v, 16, 38) << 16)
+#define MCHBAR_SIZE 0x10000
+
+/* Parameters for the channel decode stage */
+#define IMC_BASE (res_cfg->imc_base)
+#define MAD_INTER_CHANNEL_OFFSET IMC_BASE
+#define MAD_INTER_CHANNEL_DDR_TYPE(v) GET_BITFIELD(v, 0, 2)
+#define MAD_INTER_CHANNEL_ECHM(v) GET_BITFIELD(v, 3, 3)
+#define MAD_INTER_CHANNEL_CH_L_MAP(v) GET_BITFIELD(v, 4, 4)
+#define MAD_INTER_CHANNEL_CH_S_SIZE(v) ((u64)GET_BITFIELD(v, 12, 19) << 29)
+
+/* Parameters for DRAM decode stage */
+#define MAD_INTRA_CH0_OFFSET (IMC_BASE + 4)
+#define MAD_INTRA_CH_DIMM_L_MAP(v) GET_BITFIELD(v, 0, 0)
+
+/* DIMM characteristics */
+#define MAD_DIMM_CH0_OFFSET (IMC_BASE + 0xc)
+#define MAD_DIMM_CH_DIMM_L_SIZE(v) ((u64)GET_BITFIELD(v, 0, 6) << 29)
+#define MAD_DIMM_CH_DLW(v) GET_BITFIELD(v, 7, 8)
+#define MAD_DIMM_CH_DIMM_S_SIZE(v) ((u64)GET_BITFIELD(v, 16, 22) << 29)
+#define MAD_DIMM_CH_DSW(v) GET_BITFIELD(v, 24, 25)
+
+/* Hash for memory controller selection */
+#define MAD_MC_HASH_OFFSET (IMC_BASE + 0x1b8)
+#define MAC_MC_HASH_LSB(v) GET_BITFIELD(v, 1, 3)
+
+/* Hash for channel selection */
+#define CHANNEL_HASH_OFFSET (IMC_BASE + 0x24)
+/* Hash for enhanced channel selection */
+#define CHANNEL_EHASH_OFFSET (IMC_BASE + 0x28)
+#define CHANNEL_HASH_MASK(v) (GET_BITFIELD(v, 6, 19) << 6)
+#define CHANNEL_HASH_LSB_MASK_BIT(v) GET_BITFIELD(v, 24, 26)
+#define CHANNEL_HASH_MODE(v) GET_BITFIELD(v, 28, 28)
+
+/* Parameters for memory slice decode stage */
+#define MEM_SLICE_HASH_MASK(v) (GET_BITFIELD(v, 6, 19) << 6)
+#define MEM_SLICE_HASH_LSB_MASK_BIT(v) GET_BITFIELD(v, 24, 26)
+
+static struct res_config {
+ bool machine_check;
+ int num_imc;
+ u32 imc_base;
+ u32 cmf_base;
+ u32 cmf_size;
+ u32 ms_hash_offset;
+ u32 ibecc_base;
+ u32 ibecc_error_log_offset;
+ bool (*ibecc_available)(struct pci_dev *pdev);
+ /* Convert error address logged in IBECC to system physical address */
+ u64 (*err_addr_to_sys_addr)(u64 eaddr, int mc);
+ /* Convert error address logged in IBECC to integrated memory controller address */
+ u64 (*err_addr_to_imc_addr)(u64 eaddr, int mc);
+} *res_cfg;
+
+struct igen6_imc {
+ int mc;
+ struct mem_ctl_info *mci;
+ struct pci_dev *pdev;
+ struct device dev;
+ void __iomem *window;
+ u64 size;
+ u64 ch_s_size;
+ int ch_l_map;
+ u64 dimm_s_size[NUM_CHANNELS];
+ u64 dimm_l_size[NUM_CHANNELS];
+ int dimm_l_map[NUM_CHANNELS];
+};
+
+static struct igen6_pvt {
+ struct igen6_imc imc[NUM_IMC];
+ u64 ms_hash;
+ u64 ms_s_size;
+ int ms_l_map;
+} *igen6_pvt;
+
+/* The top of low usable DRAM */
+static u32 igen6_tolud;
+/* The size of physical memory */
+static u64 igen6_tom;
+
+struct decoded_addr {
+ int mc;
+ u64 imc_addr;
+ u64 sys_addr;
+ int channel_idx;
+ u64 channel_addr;
+ int sub_channel_idx;
+ u64 sub_channel_addr;
+};
+
+struct ecclog_node {
+ struct llist_node llnode;
+ int mc;
+ u64 ecclog;
+};
+
+/*
+ * In the NMI handler, the driver uses the lock-less memory allocator
+ * to allocate memory to store the IBECC error logs and links the logs
+ * to the lock-less list. Delay printk() and the work of error reporting
+ * to EDAC core in a worker.
+ */
+#define ECCLOG_POOL_SIZE PAGE_SIZE
+static LLIST_HEAD(ecclog_llist);
+static struct gen_pool *ecclog_pool;
+static char ecclog_buf[ECCLOG_POOL_SIZE];
+static struct irq_work ecclog_irq_work;
+static struct work_struct ecclog_work;
+
+/* Compute die IDs for Elkhart Lake with IBECC */
+#define DID_EHL_SKU5 0x4514
+#define DID_EHL_SKU6 0x4528
+#define DID_EHL_SKU7 0x452a
+#define DID_EHL_SKU8 0x4516
+#define DID_EHL_SKU9 0x452c
+#define DID_EHL_SKU10 0x452e
+#define DID_EHL_SKU11 0x4532
+#define DID_EHL_SKU12 0x4518
+#define DID_EHL_SKU13 0x451a
+#define DID_EHL_SKU14 0x4534
+#define DID_EHL_SKU15 0x4536
+
+/* Compute die IDs for ICL-NNPI with IBECC */
+#define DID_ICL_SKU8 0x4581
+#define DID_ICL_SKU10 0x4585
+#define DID_ICL_SKU11 0x4589
+#define DID_ICL_SKU12 0x458d
+
+/* Compute die IDs for Tiger Lake with IBECC */
+#define DID_TGL_SKU 0x9a14
+
+/* Compute die IDs for Alder Lake with IBECC */
+#define DID_ADL_SKU1 0x4601
+#define DID_ADL_SKU2 0x4602
+#define DID_ADL_SKU3 0x4621
+#define DID_ADL_SKU4 0x4641
+
+static bool ehl_ibecc_available(struct pci_dev *pdev)
+{
+ u32 v;
+
+ if (pci_read_config_dword(pdev, CAPID_C_OFFSET, &v))
+ return false;
+
+ return !!(CAPID_C_IBECC & v);
+}
+
+static u64 ehl_err_addr_to_sys_addr(u64 eaddr, int mc)
+{
+ return eaddr;
+}
+
+static u64 ehl_err_addr_to_imc_addr(u64 eaddr, int mc)
+{
+ if (eaddr < igen6_tolud)
+ return eaddr;
+
+ if (igen6_tom <= _4GB)
+ return eaddr + igen6_tolud - _4GB;
+
+ if (eaddr < _4GB)
+ return eaddr + igen6_tolud - igen6_tom;
+
+ return eaddr;
+}
+
+static bool icl_ibecc_available(struct pci_dev *pdev)
+{
+ u32 v;
+
+ if (pci_read_config_dword(pdev, CAPID_C_OFFSET, &v))
+ return false;
+
+ return !(CAPID_C_IBECC & v) &&
+ (boot_cpu_data.x86_stepping >= 1);
+}
+
+static bool tgl_ibecc_available(struct pci_dev *pdev)
+{
+ u32 v;
+
+ if (pci_read_config_dword(pdev, CAPID_E_OFFSET, &v))
+ return false;
+
+ return !(CAPID_E_IBECC & v);
+}
+
+static u64 mem_addr_to_sys_addr(u64 maddr)
+{
+ if (maddr < igen6_tolud)
+ return maddr;
+
+ if (igen6_tom <= _4GB)
+ return maddr - igen6_tolud + _4GB;
+
+ if (maddr < _4GB)
+ return maddr - igen6_tolud + igen6_tom;
+
+ return maddr;
+}
+
+static u64 mem_slice_hash(u64 addr, u64 mask, u64 hash_init, int intlv_bit)
+{
+ u64 hash_addr = addr & mask, hash = hash_init;
+ u64 intlv = (addr >> intlv_bit) & 1;
+ int i;
+
+ for (i = 6; i < 20; i++)
+ hash ^= (hash_addr >> i) & 1;
+
+ return hash ^ intlv;
+}
+
+static u64 tgl_err_addr_to_mem_addr(u64 eaddr, int mc)
+{
+ u64 maddr, hash, mask, ms_s_size;
+ int intlv_bit;
+ u32 ms_hash;
+
+ ms_s_size = igen6_pvt->ms_s_size;
+ if (eaddr >= ms_s_size)
+ return eaddr + ms_s_size;
+
+ ms_hash = igen6_pvt->ms_hash;
+
+ mask = MEM_SLICE_HASH_MASK(ms_hash);
+ intlv_bit = MEM_SLICE_HASH_LSB_MASK_BIT(ms_hash) + 6;
+
+ maddr = GET_BITFIELD(eaddr, intlv_bit, 63) << (intlv_bit + 1) |
+ GET_BITFIELD(eaddr, 0, intlv_bit - 1);
+
+ hash = mem_slice_hash(maddr, mask, mc, intlv_bit);
+
+ return maddr | (hash << intlv_bit);
+}
+
+static u64 tgl_err_addr_to_sys_addr(u64 eaddr, int mc)
+{
+ u64 maddr = tgl_err_addr_to_mem_addr(eaddr, mc);
+
+ return mem_addr_to_sys_addr(maddr);
+}
+
+static u64 tgl_err_addr_to_imc_addr(u64 eaddr, int mc)
+{
+ return eaddr;
+}
+
+static u64 adl_err_addr_to_sys_addr(u64 eaddr, int mc)
+{
+ return mem_addr_to_sys_addr(eaddr);
+}
+
+static u64 adl_err_addr_to_imc_addr(u64 eaddr, int mc)
+{
+ u64 imc_addr, ms_s_size = igen6_pvt->ms_s_size;
+ struct igen6_imc *imc = &igen6_pvt->imc[mc];
+ int intlv_bit;
+ u32 mc_hash;
+
+ if (eaddr >= 2 * ms_s_size)
+ return eaddr - ms_s_size;
+
+ mc_hash = readl(imc->window + MAD_MC_HASH_OFFSET);
+
+ intlv_bit = MAC_MC_HASH_LSB(mc_hash) + 6;
+
+ imc_addr = GET_BITFIELD(eaddr, intlv_bit + 1, 63) << intlv_bit |
+ GET_BITFIELD(eaddr, 0, intlv_bit - 1);
+
+ return imc_addr;
+}
+
+static struct res_config ehl_cfg = {
+ .num_imc = 1,
+ .imc_base = 0x5000,
+ .ibecc_base = 0xdc00,
+ .ibecc_available = ehl_ibecc_available,
+ .ibecc_error_log_offset = 0x170,
+ .err_addr_to_sys_addr = ehl_err_addr_to_sys_addr,
+ .err_addr_to_imc_addr = ehl_err_addr_to_imc_addr,
+};
+
+static struct res_config icl_cfg = {
+ .num_imc = 1,
+ .imc_base = 0x5000,
+ .ibecc_base = 0xd800,
+ .ibecc_error_log_offset = 0x170,
+ .ibecc_available = icl_ibecc_available,
+ .err_addr_to_sys_addr = ehl_err_addr_to_sys_addr,
+ .err_addr_to_imc_addr = ehl_err_addr_to_imc_addr,
+};
+
+static struct res_config tgl_cfg = {
+ .machine_check = true,
+ .num_imc = 2,
+ .imc_base = 0x5000,
+ .cmf_base = 0x11000,
+ .cmf_size = 0x800,
+ .ms_hash_offset = 0xac,
+ .ibecc_base = 0xd400,
+ .ibecc_error_log_offset = 0x170,
+ .ibecc_available = tgl_ibecc_available,
+ .err_addr_to_sys_addr = tgl_err_addr_to_sys_addr,
+ .err_addr_to_imc_addr = tgl_err_addr_to_imc_addr,
+};
+
+static struct res_config adl_cfg = {
+ .machine_check = true,
+ .num_imc = 2,
+ .imc_base = 0xd800,
+ .ibecc_base = 0xd400,
+ .ibecc_error_log_offset = 0x68,
+ .ibecc_available = tgl_ibecc_available,
+ .err_addr_to_sys_addr = adl_err_addr_to_sys_addr,
+ .err_addr_to_imc_addr = adl_err_addr_to_imc_addr,
+};
+
+static const struct pci_device_id igen6_pci_tbl[] = {
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU5), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU6), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU7), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU8), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU9), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU10), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU11), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU12), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU13), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU14), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_EHL_SKU15), (kernel_ulong_t)&ehl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ICL_SKU8), (kernel_ulong_t)&icl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ICL_SKU10), (kernel_ulong_t)&icl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ICL_SKU11), (kernel_ulong_t)&icl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ICL_SKU12), (kernel_ulong_t)&icl_cfg },
+ { PCI_VDEVICE(INTEL, DID_TGL_SKU), (kernel_ulong_t)&tgl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ADL_SKU1), (kernel_ulong_t)&adl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ADL_SKU2), (kernel_ulong_t)&adl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ADL_SKU3), (kernel_ulong_t)&adl_cfg },
+ { PCI_VDEVICE(INTEL, DID_ADL_SKU4), (kernel_ulong_t)&adl_cfg },
+ { },
+};
+MODULE_DEVICE_TABLE(pci, igen6_pci_tbl);
+
+static enum dev_type get_width(int dimm_l, u32 mad_dimm)
+{
+ u32 w = dimm_l ? MAD_DIMM_CH_DLW(mad_dimm) :
+ MAD_DIMM_CH_DSW(mad_dimm);
+
+ switch (w) {
+ case 0:
+ return DEV_X8;
+ case 1:
+ return DEV_X16;
+ case 2:
+ return DEV_X32;
+ default:
+ return DEV_UNKNOWN;
+ }
+}
+
+static enum mem_type get_memory_type(u32 mad_inter)
+{
+ u32 t = MAD_INTER_CHANNEL_DDR_TYPE(mad_inter);
+
+ switch (t) {
+ case 0:
+ return MEM_DDR4;
+ case 1:
+ return MEM_DDR3;
+ case 2:
+ return MEM_LPDDR3;
+ case 3:
+ return MEM_LPDDR4;
+ case 4:
+ return MEM_WIO2;
+ default:
+ return MEM_UNKNOWN;
+ }
+}
+
+static int decode_chan_idx(u64 addr, u64 mask, int intlv_bit)
+{
+ u64 hash_addr = addr & mask, hash = 0;
+ u64 intlv = (addr >> intlv_bit) & 1;
+ int i;
+
+ for (i = 6; i < 20; i++)
+ hash ^= (hash_addr >> i) & 1;
+
+ return (int)hash ^ intlv;
+}
+
+static u64 decode_channel_addr(u64 addr, int intlv_bit)
+{
+ u64 channel_addr;
+
+ /* Remove the interleave bit and shift upper part down to fill gap */
+ channel_addr = GET_BITFIELD(addr, intlv_bit + 1, 63) << intlv_bit;
+ channel_addr |= GET_BITFIELD(addr, 0, intlv_bit - 1);
+
+ return channel_addr;
+}
+
+static void decode_addr(u64 addr, u32 hash, u64 s_size, int l_map,
+ int *idx, u64 *sub_addr)
+{
+ int intlv_bit = CHANNEL_HASH_LSB_MASK_BIT(hash) + 6;
+
+ if (addr > 2 * s_size) {
+ *sub_addr = addr - s_size;
+ *idx = l_map;
+ return;
+ }
+
+ if (CHANNEL_HASH_MODE(hash)) {
+ *sub_addr = decode_channel_addr(addr, intlv_bit);
+ *idx = decode_chan_idx(addr, CHANNEL_HASH_MASK(hash), intlv_bit);
+ } else {
+ *sub_addr = decode_channel_addr(addr, 6);
+ *idx = GET_BITFIELD(addr, 6, 6);
+ }
+}
+
+static int igen6_decode(struct decoded_addr *res)
+{
+ struct igen6_imc *imc = &igen6_pvt->imc[res->mc];
+ u64 addr = res->imc_addr, sub_addr, s_size;
+ int idx, l_map;
+ u32 hash;
+
+ if (addr >= igen6_tom) {
+ edac_dbg(0, "Address 0x%llx out of range\n", addr);
+ return -EINVAL;
+ }
+
+ /* Decode channel */
+ hash = readl(imc->window + CHANNEL_HASH_OFFSET);
+ s_size = imc->ch_s_size;
+ l_map = imc->ch_l_map;
+ decode_addr(addr, hash, s_size, l_map, &idx, &sub_addr);
+ res->channel_idx = idx;
+ res->channel_addr = sub_addr;
+
+ /* Decode sub-channel/DIMM */
+ hash = readl(imc->window + CHANNEL_EHASH_OFFSET);
+ s_size = imc->dimm_s_size[idx];
+ l_map = imc->dimm_l_map[idx];
+ decode_addr(res->channel_addr, hash, s_size, l_map, &idx, &sub_addr);
+ res->sub_channel_idx = idx;
+ res->sub_channel_addr = sub_addr;
+
+ return 0;
+}
+
+static void igen6_output_error(struct decoded_addr *res,
+ struct mem_ctl_info *mci, u64 ecclog)
+{
+ enum hw_event_mc_err_type type = ecclog & ECC_ERROR_LOG_UE ?
+ HW_EVENT_ERR_UNCORRECTED :
+ HW_EVENT_ERR_CORRECTED;
+
+ edac_mc_handle_error(type, mci, 1,
+ res->sys_addr >> PAGE_SHIFT,
+ res->sys_addr & ~PAGE_MASK,
+ ECC_ERROR_LOG_SYND(ecclog),
+ res->channel_idx, res->sub_channel_idx,
+ -1, "", "");
+}
+
+static struct gen_pool *ecclog_gen_pool_create(void)
+{
+ struct gen_pool *pool;
+
+ pool = gen_pool_create(ilog2(sizeof(struct ecclog_node)), -1);
+ if (!pool)
+ return NULL;
+
+ if (gen_pool_add(pool, (unsigned long)ecclog_buf, ECCLOG_POOL_SIZE, -1)) {
+ gen_pool_destroy(pool);
+ return NULL;
+ }
+
+ return pool;
+}
+
+static int ecclog_gen_pool_add(int mc, u64 ecclog)
+{
+ struct ecclog_node *node;
+
+ node = (void *)gen_pool_alloc(ecclog_pool, sizeof(*node));
+ if (!node)
+ return -ENOMEM;
+
+ node->mc = mc;
+ node->ecclog = ecclog;
+ llist_add(&node->llnode, &ecclog_llist);
+
+ return 0;
+}
+
+/*
+ * Either the memory-mapped I/O status register ECC_ERROR_LOG or the PCI
+ * configuration space status register ERRSTS can indicate whether a
+ * correctable error or an uncorrectable error occurred. We only use the
+ * ECC_ERROR_LOG register to check error type, but need to clear both
+ * registers to enable future error events.
+ */
+static u64 ecclog_read_and_clear(struct igen6_imc *imc)
+{
+ u64 ecclog = readq(imc->window + ECC_ERROR_LOG_OFFSET);
+
+ if (ecclog & (ECC_ERROR_LOG_CE | ECC_ERROR_LOG_UE)) {
+ /* Clear CE/UE bits by writing 1s */
+ writeq(ecclog, imc->window + ECC_ERROR_LOG_OFFSET);
+ return ecclog;
+ }
+
+ return 0;
+}
+
+static void errsts_clear(struct igen6_imc *imc)
+{
+ u16 errsts;
+
+ if (pci_read_config_word(imc->pdev, ERRSTS_OFFSET, &errsts)) {
+ igen6_printk(KERN_ERR, "Failed to read ERRSTS\n");
+ return;
+ }
+
+ /* Clear CE/UE bits by writing 1s */
+ if (errsts & (ERRSTS_CE | ERRSTS_UE))
+ pci_write_config_word(imc->pdev, ERRSTS_OFFSET, errsts);
+}
+
+static int errcmd_enable_error_reporting(bool enable)
+{
+ struct igen6_imc *imc = &igen6_pvt->imc[0];
+ u16 errcmd;
+ int rc;
+
+ rc = pci_read_config_word(imc->pdev, ERRCMD_OFFSET, &errcmd);
+ if (rc)
+ return rc;
+
+ if (enable)
+ errcmd |= ERRCMD_CE | ERRSTS_UE;
+ else
+ errcmd &= ~(ERRCMD_CE | ERRSTS_UE);
+
+ rc = pci_write_config_word(imc->pdev, ERRCMD_OFFSET, errcmd);
+ if (rc)
+ return rc;
+
+ return 0;
+}
+
+static int ecclog_handler(void)
+{
+ struct igen6_imc *imc;
+ int i, n = 0;
+ u64 ecclog;
+
+ for (i = 0; i < res_cfg->num_imc; i++) {
+ imc = &igen6_pvt->imc[i];
+
+ /* errsts_clear() isn't NMI-safe. Delay it in the IRQ context */
+
+ ecclog = ecclog_read_and_clear(imc);
+ if (!ecclog)
+ continue;
+
+ if (!ecclog_gen_pool_add(i, ecclog))
+ irq_work_queue(&ecclog_irq_work);
+
+ n++;
+ }
+
+ return n;
+}
+
+static void ecclog_work_cb(struct work_struct *work)
+{
+ struct ecclog_node *node, *tmp;
+ struct mem_ctl_info *mci;
+ struct llist_node *head;
+ struct decoded_addr res;
+ u64 eaddr;
+
+ head = llist_del_all(&ecclog_llist);
+ if (!head)
+ return;
+
+ llist_for_each_entry_safe(node, tmp, head, llnode) {
+ memset(&res, 0, sizeof(res));
+ eaddr = ECC_ERROR_LOG_ADDR(node->ecclog) <<
+ ECC_ERROR_LOG_ADDR_SHIFT;
+ res.mc = node->mc;
+ res.sys_addr = res_cfg->err_addr_to_sys_addr(eaddr, res.mc);
+ res.imc_addr = res_cfg->err_addr_to_imc_addr(eaddr, res.mc);
+
+ mci = igen6_pvt->imc[res.mc].mci;
+
+ edac_dbg(2, "MC %d, ecclog = 0x%llx\n", node->mc, node->ecclog);
+ igen6_mc_printk(mci, KERN_DEBUG, "HANDLING IBECC MEMORY ERROR\n");
+ igen6_mc_printk(mci, KERN_DEBUG, "ADDR 0x%llx ", res.sys_addr);
+
+ if (!igen6_decode(&res))
+ igen6_output_error(&res, mci, node->ecclog);
+
+ gen_pool_free(ecclog_pool, (unsigned long)node, sizeof(*node));
+ }
+}
+
+static void ecclog_irq_work_cb(struct irq_work *irq_work)
+{
+ int i;
+
+ for (i = 0; i < res_cfg->num_imc; i++)
+ errsts_clear(&igen6_pvt->imc[i]);
+
+ if (!llist_empty(&ecclog_llist))
+ schedule_work(&ecclog_work);
+}
+
+static int ecclog_nmi_handler(unsigned int cmd, struct pt_regs *regs)
+{
+ unsigned char reason;
+
+ if (!ecclog_handler())
+ return NMI_DONE;
+
+ /*
+ * Both In-Band ECC correctable error and uncorrectable error are
+ * reported by SERR# NMI. The NMI generic code (see pci_serr_error())
+ * doesn't clear the bit NMI_REASON_CLEAR_SERR (in port 0x61) to
+ * re-enable the SERR# NMI after NMI handling. So clear this bit here
+ * to re-enable SERR# NMI for receiving future In-Band ECC errors.
+ */
+ reason = x86_platform.get_nmi_reason() & NMI_REASON_CLEAR_MASK;
+ reason |= NMI_REASON_CLEAR_SERR;
+ outb(reason, NMI_REASON_PORT);
+ reason &= ~NMI_REASON_CLEAR_SERR;
+ outb(reason, NMI_REASON_PORT);
+
+ return NMI_HANDLED;
+}
+
+static int ecclog_mce_handler(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct mce *mce = (struct mce *)data;
+ char *type;
+
+ if (mce->kflags & MCE_HANDLED_CEC)
+ return NOTIFY_DONE;
+
+ /*
+ * Ignore unless this is a memory related error.
+ * We don't check the bit MCI_STATUS_ADDRV of MCi_STATUS here,
+ * since this bit isn't set on some CPU (e.g., Tiger Lake UP3).
+ */
+ if ((mce->status & 0xefff) >> 7 != 1)
+ return NOTIFY_DONE;
+
+ if (mce->mcgstatus & MCG_STATUS_MCIP)
+ type = "Exception";
+ else
+ type = "Event";
+
+ edac_dbg(0, "CPU %d: Machine Check %s: 0x%llx Bank %d: 0x%llx\n",
+ mce->extcpu, type, mce->mcgstatus,
+ mce->bank, mce->status);
+ edac_dbg(0, "TSC 0x%llx\n", mce->tsc);
+ edac_dbg(0, "ADDR 0x%llx\n", mce->addr);
+ edac_dbg(0, "MISC 0x%llx\n", mce->misc);
+ edac_dbg(0, "PROCESSOR %u:0x%x TIME %llu SOCKET %u APIC 0x%x\n",
+ mce->cpuvendor, mce->cpuid, mce->time,
+ mce->socketid, mce->apicid);
+ /*
+ * We just use the Machine Check for the memory error notification.
+ * Each memory controller is associated with an IBECC instance.
+ * Directly read and clear the error information(error address and
+ * error type) on all the IBECC instances so that we know on which
+ * memory controller the memory error(s) occurred.
+ */
+ if (!ecclog_handler())
+ return NOTIFY_DONE;
+
+ mce->kflags |= MCE_HANDLED_EDAC;
+
+ return NOTIFY_DONE;
+}
+
+static struct notifier_block ecclog_mce_dec = {
+ .notifier_call = ecclog_mce_handler,
+ .priority = MCE_PRIO_EDAC,
+};
+
+static bool igen6_check_ecc(struct igen6_imc *imc)
+{
+ u32 activate = readl(imc->window + IBECC_ACTIVATE_OFFSET);
+
+ return !!(activate & IBECC_ACTIVATE_EN);
+}
+
+static int igen6_get_dimm_config(struct mem_ctl_info *mci)
+{
+ struct igen6_imc *imc = mci->pvt_info;
+ u32 mad_inter, mad_intra, mad_dimm;
+ int i, j, ndimms, mc = imc->mc;
+ struct dimm_info *dimm;
+ enum mem_type mtype;
+ enum dev_type dtype;
+ u64 dsize;
+ bool ecc;
+
+ edac_dbg(2, "\n");
+
+ mad_inter = readl(imc->window + MAD_INTER_CHANNEL_OFFSET);
+ mtype = get_memory_type(mad_inter);
+ ecc = igen6_check_ecc(imc);
+ imc->ch_s_size = MAD_INTER_CHANNEL_CH_S_SIZE(mad_inter);
+ imc->ch_l_map = MAD_INTER_CHANNEL_CH_L_MAP(mad_inter);
+
+ for (i = 0; i < NUM_CHANNELS; i++) {
+ mad_intra = readl(imc->window + MAD_INTRA_CH0_OFFSET + i * 4);
+ mad_dimm = readl(imc->window + MAD_DIMM_CH0_OFFSET + i * 4);
+
+ imc->dimm_l_size[i] = MAD_DIMM_CH_DIMM_L_SIZE(mad_dimm);
+ imc->dimm_s_size[i] = MAD_DIMM_CH_DIMM_S_SIZE(mad_dimm);
+ imc->dimm_l_map[i] = MAD_INTRA_CH_DIMM_L_MAP(mad_intra);
+ imc->size += imc->dimm_s_size[i];
+ imc->size += imc->dimm_l_size[i];
+ ndimms = 0;
+
+ for (j = 0; j < NUM_DIMMS; j++) {
+ dimm = edac_get_dimm(mci, i, j, 0);
+
+ if (j ^ imc->dimm_l_map[i]) {
+ dtype = get_width(0, mad_dimm);
+ dsize = imc->dimm_s_size[i];
+ } else {
+ dtype = get_width(1, mad_dimm);
+ dsize = imc->dimm_l_size[i];
+ }
+
+ if (!dsize)
+ continue;
+
+ dimm->grain = 64;
+ dimm->mtype = mtype;
+ dimm->dtype = dtype;
+ dimm->nr_pages = MiB_TO_PAGES(dsize >> 20);
+ dimm->edac_mode = EDAC_SECDED;
+ snprintf(dimm->label, sizeof(dimm->label),
+ "MC#%d_Chan#%d_DIMM#%d", mc, i, j);
+ edac_dbg(0, "MC %d, Channel %d, DIMM %d, Size %llu MiB (%u pages)\n",
+ mc, i, j, dsize >> 20, dimm->nr_pages);
+
+ ndimms++;
+ }
+
+ if (ndimms && !ecc) {
+ igen6_printk(KERN_ERR, "MC%d In-Band ECC is disabled\n", mc);
+ return -ENODEV;
+ }
+ }
+
+ edac_dbg(0, "MC %d, total size %llu MiB\n", mc, imc->size >> 20);
+
+ return 0;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/* Top of upper usable DRAM */
+static u64 igen6_touud;
+#define TOUUD_OFFSET 0xa8
+
+static void igen6_reg_dump(struct igen6_imc *imc)
+{
+ int i;
+
+ edac_dbg(2, "CHANNEL_HASH : 0x%x\n",
+ readl(imc->window + CHANNEL_HASH_OFFSET));
+ edac_dbg(2, "CHANNEL_EHASH : 0x%x\n",
+ readl(imc->window + CHANNEL_EHASH_OFFSET));
+ edac_dbg(2, "MAD_INTER_CHANNEL: 0x%x\n",
+ readl(imc->window + MAD_INTER_CHANNEL_OFFSET));
+ edac_dbg(2, "ECC_ERROR_LOG : 0x%llx\n",
+ readq(imc->window + ECC_ERROR_LOG_OFFSET));
+
+ for (i = 0; i < NUM_CHANNELS; i++) {
+ edac_dbg(2, "MAD_INTRA_CH%d : 0x%x\n", i,
+ readl(imc->window + MAD_INTRA_CH0_OFFSET + i * 4));
+ edac_dbg(2, "MAD_DIMM_CH%d : 0x%x\n", i,
+ readl(imc->window + MAD_DIMM_CH0_OFFSET + i * 4));
+ }
+ edac_dbg(2, "TOLUD : 0x%x", igen6_tolud);
+ edac_dbg(2, "TOUUD : 0x%llx", igen6_touud);
+ edac_dbg(2, "TOM : 0x%llx", igen6_tom);
+}
+
+static struct dentry *igen6_test;
+
+static int debugfs_u64_set(void *data, u64 val)
+{
+ u64 ecclog;
+
+ if ((val >= igen6_tolud && val < _4GB) || val >= igen6_touud) {
+ edac_dbg(0, "Address 0x%llx out of range\n", val);
+ return 0;
+ }
+
+ pr_warn_once("Fake error to 0x%llx injected via debugfs\n", val);
+
+ val >>= ECC_ERROR_LOG_ADDR_SHIFT;
+ ecclog = (val << ECC_ERROR_LOG_ADDR_SHIFT) | ECC_ERROR_LOG_CE;
+
+ if (!ecclog_gen_pool_add(0, ecclog))
+ irq_work_queue(&ecclog_irq_work);
+
+ return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n");
+
+static void igen6_debug_setup(void)
+{
+ igen6_test = edac_debugfs_create_dir("igen6_test");
+ if (!igen6_test)
+ return;
+
+ if (!edac_debugfs_create_file("addr", 0200, igen6_test,
+ NULL, &fops_u64_wo)) {
+ debugfs_remove(igen6_test);
+ igen6_test = NULL;
+ }
+}
+
+static void igen6_debug_teardown(void)
+{
+ debugfs_remove_recursive(igen6_test);
+}
+#else
+static void igen6_reg_dump(struct igen6_imc *imc) {}
+static void igen6_debug_setup(void) {}
+static void igen6_debug_teardown(void) {}
+#endif
+
+static int igen6_pci_setup(struct pci_dev *pdev, u64 *mchbar)
+{
+ union {
+ u64 v;
+ struct {
+ u32 v_lo;
+ u32 v_hi;
+ };
+ } u;
+
+ edac_dbg(2, "\n");
+
+ if (!res_cfg->ibecc_available(pdev)) {
+ edac_dbg(2, "No In-Band ECC IP\n");
+ goto fail;
+ }
+
+ if (pci_read_config_dword(pdev, TOLUD_OFFSET, &igen6_tolud)) {
+ igen6_printk(KERN_ERR, "Failed to read TOLUD\n");
+ goto fail;
+ }
+
+ igen6_tolud &= GENMASK(31, 20);
+
+ if (pci_read_config_dword(pdev, TOM_OFFSET, &u.v_lo)) {
+ igen6_printk(KERN_ERR, "Failed to read lower TOM\n");
+ goto fail;
+ }
+
+ if (pci_read_config_dword(pdev, TOM_OFFSET + 4, &u.v_hi)) {
+ igen6_printk(KERN_ERR, "Failed to read upper TOM\n");
+ goto fail;
+ }
+
+ igen6_tom = u.v & GENMASK_ULL(38, 20);
+
+ if (pci_read_config_dword(pdev, MCHBAR_OFFSET, &u.v_lo)) {
+ igen6_printk(KERN_ERR, "Failed to read lower MCHBAR\n");
+ goto fail;
+ }
+
+ if (pci_read_config_dword(pdev, MCHBAR_OFFSET + 4, &u.v_hi)) {
+ igen6_printk(KERN_ERR, "Failed to read upper MCHBAR\n");
+ goto fail;
+ }
+
+ if (!(u.v & MCHBAR_EN)) {
+ igen6_printk(KERN_ERR, "MCHBAR is disabled\n");
+ goto fail;
+ }
+
+ *mchbar = MCHBAR_BASE(u.v);
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (pci_read_config_dword(pdev, TOUUD_OFFSET, &u.v_lo))
+ edac_dbg(2, "Failed to read lower TOUUD\n");
+ else if (pci_read_config_dword(pdev, TOUUD_OFFSET + 4, &u.v_hi))
+ edac_dbg(2, "Failed to read upper TOUUD\n");
+ else
+ igen6_touud = u.v & GENMASK_ULL(38, 20);
+#endif
+
+ return 0;
+fail:
+ return -ENODEV;
+}
+
+static int igen6_register_mci(int mc, u64 mchbar, struct pci_dev *pdev)
+{
+ struct edac_mc_layer layers[2];
+ struct mem_ctl_info *mci;
+ struct igen6_imc *imc;
+ void __iomem *window;
+ int rc;
+
+ edac_dbg(2, "\n");
+
+ mchbar += mc * MCHBAR_SIZE;
+ window = ioremap(mchbar, MCHBAR_SIZE);
+ if (!window) {
+ igen6_printk(KERN_ERR, "Failed to ioremap 0x%llx\n", mchbar);
+ return -ENODEV;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = NUM_CHANNELS;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = NUM_DIMMS;
+ layers[1].is_virt_csrow = true;
+
+ mci = edac_mc_alloc(mc, ARRAY_SIZE(layers), layers, 0);
+ if (!mci) {
+ rc = -ENOMEM;
+ goto fail;
+ }
+
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Intel_client_SoC MC#%d", mc);
+ if (!mci->ctl_name) {
+ rc = -ENOMEM;
+ goto fail2;
+ }
+
+ mci->mtype_cap = MEM_FLAG_LPDDR4 | MEM_FLAG_DDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->dev_name = pci_name(pdev);
+ mci->pvt_info = &igen6_pvt->imc[mc];
+
+ imc = mci->pvt_info;
+ device_initialize(&imc->dev);
+ /*
+ * EDAC core uses mci->pdev(pointer of structure device) as
+ * memory controller ID. The client SoCs attach one or more
+ * memory controllers to single pci_dev (single pci_dev->dev
+ * can be for multiple memory controllers).
+ *
+ * To make mci->pdev unique, assign pci_dev->dev to mci->pdev
+ * for the first memory controller and assign a unique imc->dev
+ * to mci->pdev for each non-first memory controller.
+ */
+ mci->pdev = mc ? &imc->dev : &pdev->dev;
+ imc->mc = mc;
+ imc->pdev = pdev;
+ imc->window = window;
+
+ igen6_reg_dump(imc);
+
+ rc = igen6_get_dimm_config(mci);
+ if (rc)
+ goto fail3;
+
+ rc = edac_mc_add_mc(mci);
+ if (rc) {
+ igen6_printk(KERN_ERR, "Failed to register mci#%d\n", mc);
+ goto fail3;
+ }
+
+ imc->mci = mci;
+ return 0;
+fail3:
+ kfree(mci->ctl_name);
+fail2:
+ edac_mc_free(mci);
+fail:
+ iounmap(window);
+ return rc;
+}
+
+static void igen6_unregister_mcis(void)
+{
+ struct mem_ctl_info *mci;
+ struct igen6_imc *imc;
+ int i;
+
+ edac_dbg(2, "\n");
+
+ for (i = 0; i < res_cfg->num_imc; i++) {
+ imc = &igen6_pvt->imc[i];
+ mci = imc->mci;
+ if (!mci)
+ continue;
+
+ edac_mc_del_mc(mci->pdev);
+ kfree(mci->ctl_name);
+ edac_mc_free(mci);
+ iounmap(imc->window);
+ }
+}
+
+static int igen6_mem_slice_setup(u64 mchbar)
+{
+ struct igen6_imc *imc = &igen6_pvt->imc[0];
+ u64 base = mchbar + res_cfg->cmf_base;
+ u32 offset = res_cfg->ms_hash_offset;
+ u32 size = res_cfg->cmf_size;
+ u64 ms_s_size, ms_hash;
+ void __iomem *cmf;
+ int ms_l_map;
+
+ edac_dbg(2, "\n");
+
+ if (imc[0].size < imc[1].size) {
+ ms_s_size = imc[0].size;
+ ms_l_map = 1;
+ } else {
+ ms_s_size = imc[1].size;
+ ms_l_map = 0;
+ }
+
+ igen6_pvt->ms_s_size = ms_s_size;
+ igen6_pvt->ms_l_map = ms_l_map;
+
+ edac_dbg(0, "ms_s_size: %llu MiB, ms_l_map %d\n",
+ ms_s_size >> 20, ms_l_map);
+
+ if (!size)
+ return 0;
+
+ cmf = ioremap(base, size);
+ if (!cmf) {
+ igen6_printk(KERN_ERR, "Failed to ioremap cmf 0x%llx\n", base);
+ return -ENODEV;
+ }
+
+ ms_hash = readq(cmf + offset);
+ igen6_pvt->ms_hash = ms_hash;
+
+ edac_dbg(0, "MEM_SLICE_HASH: 0x%llx\n", ms_hash);
+
+ iounmap(cmf);
+
+ return 0;
+}
+
+static int register_err_handler(void)
+{
+ int rc;
+
+ if (res_cfg->machine_check) {
+ mce_register_decode_chain(&ecclog_mce_dec);
+ return 0;
+ }
+
+ rc = register_nmi_handler(NMI_SERR, ecclog_nmi_handler,
+ 0, IGEN6_NMI_NAME);
+ if (rc) {
+ igen6_printk(KERN_ERR, "Failed to register NMI handler\n");
+ return rc;
+ }
+
+ return 0;
+}
+
+static void unregister_err_handler(void)
+{
+ if (res_cfg->machine_check) {
+ mce_unregister_decode_chain(&ecclog_mce_dec);
+ return;
+ }
+
+ unregister_nmi_handler(NMI_SERR, IGEN6_NMI_NAME);
+}
+
+static int igen6_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ u64 mchbar;
+ int i, rc;
+
+ edac_dbg(2, "\n");
+
+ igen6_pvt = kzalloc(sizeof(*igen6_pvt), GFP_KERNEL);
+ if (!igen6_pvt)
+ return -ENOMEM;
+
+ res_cfg = (struct res_config *)ent->driver_data;
+
+ rc = igen6_pci_setup(pdev, &mchbar);
+ if (rc)
+ goto fail;
+
+ for (i = 0; i < res_cfg->num_imc; i++) {
+ rc = igen6_register_mci(i, mchbar, pdev);
+ if (rc)
+ goto fail2;
+ }
+
+ if (res_cfg->num_imc > 1) {
+ rc = igen6_mem_slice_setup(mchbar);
+ if (rc)
+ goto fail2;
+ }
+
+ ecclog_pool = ecclog_gen_pool_create();
+ if (!ecclog_pool) {
+ rc = -ENOMEM;
+ goto fail2;
+ }
+
+ INIT_WORK(&ecclog_work, ecclog_work_cb);
+ init_irq_work(&ecclog_irq_work, ecclog_irq_work_cb);
+
+ rc = register_err_handler();
+ if (rc)
+ goto fail3;
+
+ /* Enable error reporting */
+ rc = errcmd_enable_error_reporting(true);
+ if (rc) {
+ igen6_printk(KERN_ERR, "Failed to enable error reporting\n");
+ goto fail4;
+ }
+
+ /* Check if any pending errors before/during the registration of the error handler */
+ ecclog_handler();
+
+ igen6_debug_setup();
+ return 0;
+fail4:
+ unregister_nmi_handler(NMI_SERR, IGEN6_NMI_NAME);
+fail3:
+ gen_pool_destroy(ecclog_pool);
+fail2:
+ igen6_unregister_mcis();
+fail:
+ kfree(igen6_pvt);
+ return rc;
+}
+
+static void igen6_remove(struct pci_dev *pdev)
+{
+ edac_dbg(2, "\n");
+
+ igen6_debug_teardown();
+ errcmd_enable_error_reporting(false);
+ unregister_err_handler();
+ irq_work_sync(&ecclog_irq_work);
+ flush_work(&ecclog_work);
+ gen_pool_destroy(ecclog_pool);
+ igen6_unregister_mcis();
+ kfree(igen6_pvt);
+}
+
+static struct pci_driver igen6_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = igen6_probe,
+ .remove = igen6_remove,
+ .id_table = igen6_pci_tbl,
+};
+
+static int __init igen6_init(void)
+{
+ const char *owner;
+ int rc;
+
+ edac_dbg(2, "\n");
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ edac_op_state = EDAC_OPSTATE_NMI;
+
+ rc = pci_register_driver(&igen6_driver);
+ if (rc)
+ return rc;
+
+ igen6_printk(KERN_INFO, "%s\n", IGEN6_REVISION);
+
+ return 0;
+}
+
+static void __exit igen6_exit(void)
+{
+ edac_dbg(2, "\n");
+
+ pci_unregister_driver(&igen6_driver);
+}
+
+module_init(igen6_init);
+module_exit(igen6_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Qiuxu Zhuo");
+MODULE_DESCRIPTION("MC Driver for Intel client SoC using In-Band ECC");
diff --git a/drivers/edac/layerscape_edac.c b/drivers/edac/layerscape_edac.c
new file mode 100644
index 0000000000..7c5e2b3c0d
--- /dev/null
+++ b/drivers/edac/layerscape_edac.c
@@ -0,0 +1,75 @@
+/*
+ * Freescale Memory Controller kernel module
+ *
+ * Author: York Sun <york.sun@nxp.com>
+ *
+ * Copyright 2016 NXP Semiconductor
+ *
+ * Derived from mpc85xx_edac.c
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2006-2007 (c) MontaVista Software, Inc. This file is licensed under
+ * the terms of the GNU General Public License version 2. This program
+ * is licensed "as is" without any warranty of any kind, whether express
+ * or implied.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include "edac_module.h"
+#include "fsl_ddr_edac.h"
+
+static const struct of_device_id fsl_ddr_mc_err_of_match[] = {
+ { .compatible = "fsl,qoriq-memory-controller", },
+ {},
+};
+MODULE_DEVICE_TABLE(of, fsl_ddr_mc_err_of_match);
+
+static struct platform_driver fsl_ddr_mc_err_driver = {
+ .probe = fsl_mc_err_probe,
+ .remove = fsl_mc_err_remove,
+ .driver = {
+ .name = "fsl_ddr_mc_err",
+ .of_match_table = fsl_ddr_mc_err_of_match,
+ },
+};
+
+static int __init fsl_ddr_mc_init(void)
+{
+ int res;
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ /* make sure error reporting method is sane */
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_INT:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_INT;
+ break;
+ }
+
+ res = platform_driver_register(&fsl_ddr_mc_err_driver);
+ if (res) {
+ pr_err("MC fails to register\n");
+ return res;
+ }
+
+ return 0;
+}
+
+module_init(fsl_ddr_mc_init);
+
+static void __exit fsl_ddr_mc_exit(void)
+{
+ platform_driver_unregister(&fsl_ddr_mc_err_driver);
+}
+
+module_exit(fsl_ddr_mc_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("NXP Semiconductor");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll, 2=Interrupt");
diff --git a/drivers/edac/mce_amd.c b/drivers/edac/mce_amd.c
new file mode 100644
index 0000000000..9215c06783
--- /dev/null
+++ b/drivers/edac/mce_amd.c
@@ -0,0 +1,1447 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#include <asm/cpu.h>
+
+#include "mce_amd.h"
+
+static struct amd_decoder_ops fam_ops;
+
+static u8 xec_mask = 0xf;
+
+static void (*decode_dram_ecc)(int node_id, struct mce *m);
+
+void amd_register_ecc_decoder(void (*f)(int, struct mce *))
+{
+ decode_dram_ecc = f;
+}
+EXPORT_SYMBOL_GPL(amd_register_ecc_decoder);
+
+void amd_unregister_ecc_decoder(void (*f)(int, struct mce *))
+{
+ if (decode_dram_ecc) {
+ WARN_ON(decode_dram_ecc != f);
+
+ decode_dram_ecc = NULL;
+ }
+}
+EXPORT_SYMBOL_GPL(amd_unregister_ecc_decoder);
+
+/*
+ * string representation for the different MCA reported error types, see F3x48
+ * or MSR0000_0411.
+ */
+
+/* transaction type */
+static const char * const tt_msgs[] = { "INSN", "DATA", "GEN", "RESV" };
+
+/* cache level */
+static const char * const ll_msgs[] = { "RESV", "L1", "L2", "L3/GEN" };
+
+/* memory transaction type */
+static const char * const rrrr_msgs[] = {
+ "GEN", "RD", "WR", "DRD", "DWR", "IRD", "PRF", "EV", "SNP"
+};
+
+/* participating processor */
+const char * const pp_msgs[] = { "SRC", "RES", "OBS", "GEN" };
+EXPORT_SYMBOL_GPL(pp_msgs);
+
+/* request timeout */
+static const char * const to_msgs[] = { "no timeout", "timed out" };
+
+/* memory or i/o */
+static const char * const ii_msgs[] = { "MEM", "RESV", "IO", "GEN" };
+
+/* internal error type */
+static const char * const uu_msgs[] = { "RESV", "RESV", "HWA", "RESV" };
+
+static const char * const f15h_mc1_mce_desc[] = {
+ "UC during a demand linefill from L2",
+ "Parity error during data load from IC",
+ "Parity error for IC valid bit",
+ "Main tag parity error",
+ "Parity error in prediction queue",
+ "PFB data/address parity error",
+ "Parity error in the branch status reg",
+ "PFB promotion address error",
+ "Tag error during probe/victimization",
+ "Parity error for IC probe tag valid bit",
+ "PFB non-cacheable bit parity error",
+ "PFB valid bit parity error", /* xec = 0xd */
+ "Microcode Patch Buffer", /* xec = 010 */
+ "uop queue",
+ "insn buffer",
+ "predecode buffer",
+ "fetch address FIFO",
+ "dispatch uop queue"
+};
+
+static const char * const f15h_mc2_mce_desc[] = {
+ "Fill ECC error on data fills", /* xec = 0x4 */
+ "Fill parity error on insn fills",
+ "Prefetcher request FIFO parity error",
+ "PRQ address parity error",
+ "PRQ data parity error",
+ "WCC Tag ECC error",
+ "WCC Data ECC error",
+ "WCB Data parity error",
+ "VB Data ECC or parity error",
+ "L2 Tag ECC error", /* xec = 0x10 */
+ "Hard L2 Tag ECC error",
+ "Multiple hits on L2 tag",
+ "XAB parity error",
+ "PRB address parity error"
+};
+
+static const char * const mc4_mce_desc[] = {
+ "DRAM ECC error detected on the NB",
+ "CRC error detected on HT link",
+ "Link-defined sync error packets detected on HT link",
+ "HT Master abort",
+ "HT Target abort",
+ "Invalid GART PTE entry during GART table walk",
+ "Unsupported atomic RMW received from an IO link",
+ "Watchdog timeout due to lack of progress",
+ "DRAM ECC error detected on the NB",
+ "SVM DMA Exclusion Vector error",
+ "HT data error detected on link",
+ "Protocol error (link, L3, probe filter)",
+ "NB internal arrays parity error",
+ "DRAM addr/ctl signals parity error",
+ "IO link transmission error",
+ "L3 data cache ECC error", /* xec = 0x1c */
+ "L3 cache tag error",
+ "L3 LRU parity bits error",
+ "ECC Error in the Probe Filter directory"
+};
+
+static const char * const mc5_mce_desc[] = {
+ "CPU Watchdog timer expire",
+ "Wakeup array dest tag",
+ "AG payload array",
+ "EX payload array",
+ "IDRF array",
+ "Retire dispatch queue",
+ "Mapper checkpoint array",
+ "Physical register file EX0 port",
+ "Physical register file EX1 port",
+ "Physical register file AG0 port",
+ "Physical register file AG1 port",
+ "Flag register file",
+ "DE error occurred",
+ "Retire status queue"
+};
+
+static const char * const mc6_mce_desc[] = {
+ "Hardware Assertion",
+ "Free List",
+ "Physical Register File",
+ "Retire Queue",
+ "Scheduler table",
+ "Status Register File",
+};
+
+/* Scalable MCA error strings */
+static const char * const smca_ls_mce_desc[] = {
+ "Load queue parity error",
+ "Store queue parity error",
+ "Miss address buffer payload parity error",
+ "Level 1 TLB parity error",
+ "DC Tag error type 5",
+ "DC Tag error type 6",
+ "DC Tag error type 1",
+ "Internal error type 1",
+ "Internal error type 2",
+ "System Read Data Error Thread 0",
+ "System Read Data Error Thread 1",
+ "DC Tag error type 2",
+ "DC Data error type 1 and poison consumption",
+ "DC Data error type 2",
+ "DC Data error type 3",
+ "DC Tag error type 4",
+ "Level 2 TLB parity error",
+ "PDC parity error",
+ "DC Tag error type 3",
+ "DC Tag error type 5",
+ "L2 Fill Data error",
+};
+
+static const char * const smca_ls2_mce_desc[] = {
+ "An ECC error was detected on a data cache read by a probe or victimization",
+ "An ECC error or L2 poison was detected on a data cache read by a load",
+ "An ECC error was detected on a data cache read-modify-write by a store",
+ "An ECC error or poison bit mismatch was detected on a tag read by a probe or victimization",
+ "An ECC error or poison bit mismatch was detected on a tag read by a load",
+ "An ECC error or poison bit mismatch was detected on a tag read by a store",
+ "An ECC error was detected on an EMEM read by a load",
+ "An ECC error was detected on an EMEM read-modify-write by a store",
+ "A parity error was detected in an L1 TLB entry by any access",
+ "A parity error was detected in an L2 TLB entry by any access",
+ "A parity error was detected in a PWC entry by any access",
+ "A parity error was detected in an STQ entry by any access",
+ "A parity error was detected in an LDQ entry by any access",
+ "A parity error was detected in a MAB entry by any access",
+ "A parity error was detected in an SCB entry state field by any access",
+ "A parity error was detected in an SCB entry address field by any access",
+ "A parity error was detected in an SCB entry data field by any access",
+ "A parity error was detected in a WCB entry by any access",
+ "A poisoned line was detected in an SCB entry by any access",
+ "A SystemReadDataError error was reported on read data returned from L2 for a load",
+ "A SystemReadDataError error was reported on read data returned from L2 for an SCB store",
+ "A SystemReadDataError error was reported on read data returned from L2 for a WCB store",
+ "A hardware assertion error was reported",
+ "A parity error was detected in an STLF, SCB EMEM entry or SRB store data by any access",
+};
+
+static const char * const smca_if_mce_desc[] = {
+ "Op Cache Microtag Probe Port Parity Error",
+ "IC Microtag or Full Tag Multi-hit Error",
+ "IC Full Tag Parity Error",
+ "IC Data Array Parity Error",
+ "Decoupling Queue PhysAddr Parity Error",
+ "L0 ITLB Parity Error",
+ "L1 ITLB Parity Error",
+ "L2 ITLB Parity Error",
+ "BPQ Thread 0 Snoop Parity Error",
+ "BPQ Thread 1 Snoop Parity Error",
+ "L1 BTB Multi-Match Error",
+ "L2 BTB Multi-Match Error",
+ "L2 Cache Response Poison Error",
+ "System Read Data Error",
+ "Hardware Assertion Error",
+ "L1-TLB Multi-Hit",
+ "L2-TLB Multi-Hit",
+ "BSR Parity Error",
+ "CT MCE",
+};
+
+static const char * const smca_l2_mce_desc[] = {
+ "L2M Tag Multiple-Way-Hit error",
+ "L2M Tag or State Array ECC Error",
+ "L2M Data Array ECC Error",
+ "Hardware Assert Error",
+};
+
+static const char * const smca_de_mce_desc[] = {
+ "Micro-op cache tag parity error",
+ "Micro-op cache data parity error",
+ "Instruction buffer parity error",
+ "Micro-op queue parity error",
+ "Instruction dispatch queue parity error",
+ "Fetch address FIFO parity error",
+ "Patch RAM data parity error",
+ "Patch RAM sequencer parity error",
+ "Micro-op buffer parity error",
+ "Hardware Assertion MCA Error",
+};
+
+static const char * const smca_ex_mce_desc[] = {
+ "Watchdog Timeout error",
+ "Physical register file parity error",
+ "Flag register file parity error",
+ "Immediate displacement register file parity error",
+ "Address generator payload parity error",
+ "EX payload parity error",
+ "Checkpoint queue parity error",
+ "Retire dispatch queue parity error",
+ "Retire status queue parity error",
+ "Scheduling queue parity error",
+ "Branch buffer queue parity error",
+ "Hardware Assertion error",
+ "Spec Map parity error",
+ "Retire Map parity error",
+};
+
+static const char * const smca_fp_mce_desc[] = {
+ "Physical register file (PRF) parity error",
+ "Freelist (FL) parity error",
+ "Schedule queue parity error",
+ "NSQ parity error",
+ "Retire queue (RQ) parity error",
+ "Status register file (SRF) parity error",
+ "Hardware assertion",
+};
+
+static const char * const smca_l3_mce_desc[] = {
+ "Shadow Tag Macro ECC Error",
+ "Shadow Tag Macro Multi-way-hit Error",
+ "L3M Tag ECC Error",
+ "L3M Tag Multi-way-hit Error",
+ "L3M Data ECC Error",
+ "SDP Parity Error or SystemReadDataError from XI",
+ "L3 Victim Queue Parity Error",
+ "L3 Hardware Assertion",
+};
+
+static const char * const smca_cs_mce_desc[] = {
+ "Illegal Request",
+ "Address Violation",
+ "Security Violation",
+ "Illegal Response",
+ "Unexpected Response",
+ "Request or Probe Parity Error",
+ "Read Response Parity Error",
+ "Atomic Request Parity Error",
+ "Probe Filter ECC Error",
+};
+
+static const char * const smca_cs2_mce_desc[] = {
+ "Illegal Request",
+ "Address Violation",
+ "Security Violation",
+ "Illegal Response",
+ "Unexpected Response",
+ "Request or Probe Parity Error",
+ "Read Response Parity Error",
+ "Atomic Request Parity Error",
+ "SDP read response had no match in the CS queue",
+ "Probe Filter Protocol Error",
+ "Probe Filter ECC Error",
+ "SDP read response had an unexpected RETRY error",
+ "Counter overflow error",
+ "Counter underflow error",
+};
+
+static const char * const smca_pie_mce_desc[] = {
+ "Hardware Assert",
+ "Register security violation",
+ "Link Error",
+ "Poison data consumption",
+ "A deferred error was detected in the DF"
+};
+
+static const char * const smca_umc_mce_desc[] = {
+ "DRAM ECC error",
+ "Data poison error",
+ "SDP parity error",
+ "Advanced peripheral bus error",
+ "Address/Command parity error",
+ "Write data CRC error",
+ "DCQ SRAM ECC error",
+ "AES SRAM ECC error",
+};
+
+static const char * const smca_umc2_mce_desc[] = {
+ "DRAM ECC error",
+ "Data poison error",
+ "SDP parity error",
+ "Reserved",
+ "Address/Command parity error",
+ "Write data parity error",
+ "DCQ SRAM ECC error",
+ "Reserved",
+ "Read data parity error",
+ "Rdb SRAM ECC error",
+ "RdRsp SRAM ECC error",
+ "LM32 MP errors",
+};
+
+static const char * const smca_pb_mce_desc[] = {
+ "An ECC error in the Parameter Block RAM array",
+};
+
+static const char * const smca_psp_mce_desc[] = {
+ "An ECC or parity error in a PSP RAM instance",
+};
+
+static const char * const smca_psp2_mce_desc[] = {
+ "High SRAM ECC or parity error",
+ "Low SRAM ECC or parity error",
+ "Instruction Cache Bank 0 ECC or parity error",
+ "Instruction Cache Bank 1 ECC or parity error",
+ "Instruction Tag Ram 0 parity error",
+ "Instruction Tag Ram 1 parity error",
+ "Data Cache Bank 0 ECC or parity error",
+ "Data Cache Bank 1 ECC or parity error",
+ "Data Cache Bank 2 ECC or parity error",
+ "Data Cache Bank 3 ECC or parity error",
+ "Data Tag Bank 0 parity error",
+ "Data Tag Bank 1 parity error",
+ "Data Tag Bank 2 parity error",
+ "Data Tag Bank 3 parity error",
+ "Dirty Data Ram parity error",
+ "TLB Bank 0 parity error",
+ "TLB Bank 1 parity error",
+ "System Hub Read Buffer ECC or parity error",
+};
+
+static const char * const smca_smu_mce_desc[] = {
+ "An ECC or parity error in an SMU RAM instance",
+};
+
+static const char * const smca_smu2_mce_desc[] = {
+ "High SRAM ECC or parity error",
+ "Low SRAM ECC or parity error",
+ "Data Cache Bank A ECC or parity error",
+ "Data Cache Bank B ECC or parity error",
+ "Data Tag Cache Bank A ECC or parity error",
+ "Data Tag Cache Bank B ECC or parity error",
+ "Instruction Cache Bank A ECC or parity error",
+ "Instruction Cache Bank B ECC or parity error",
+ "Instruction Tag Cache Bank A ECC or parity error",
+ "Instruction Tag Cache Bank B ECC or parity error",
+ "System Hub Read Buffer ECC or parity error",
+ "PHY RAM ECC error",
+};
+
+static const char * const smca_mp5_mce_desc[] = {
+ "High SRAM ECC or parity error",
+ "Low SRAM ECC or parity error",
+ "Data Cache Bank A ECC or parity error",
+ "Data Cache Bank B ECC or parity error",
+ "Data Tag Cache Bank A ECC or parity error",
+ "Data Tag Cache Bank B ECC or parity error",
+ "Instruction Cache Bank A ECC or parity error",
+ "Instruction Cache Bank B ECC or parity error",
+ "Instruction Tag Cache Bank A ECC or parity error",
+ "Instruction Tag Cache Bank B ECC or parity error",
+};
+
+static const char * const smca_mpdma_mce_desc[] = {
+ "Main SRAM [31:0] bank ECC or parity error",
+ "Main SRAM [63:32] bank ECC or parity error",
+ "Main SRAM [95:64] bank ECC or parity error",
+ "Main SRAM [127:96] bank ECC or parity error",
+ "Data Cache Bank A ECC or parity error",
+ "Data Cache Bank B ECC or parity error",
+ "Data Tag Cache Bank A ECC or parity error",
+ "Data Tag Cache Bank B ECC or parity error",
+ "Instruction Cache Bank A ECC or parity error",
+ "Instruction Cache Bank B ECC or parity error",
+ "Instruction Tag Cache Bank A ECC or parity error",
+ "Instruction Tag Cache Bank B ECC or parity error",
+ "Data Cache Bank A ECC or parity error",
+ "Data Cache Bank B ECC or parity error",
+ "Data Tag Cache Bank A ECC or parity error",
+ "Data Tag Cache Bank B ECC or parity error",
+ "Instruction Cache Bank A ECC or parity error",
+ "Instruction Cache Bank B ECC or parity error",
+ "Instruction Tag Cache Bank A ECC or parity error",
+ "Instruction Tag Cache Bank B ECC or parity error",
+ "Data Cache Bank A ECC or parity error",
+ "Data Cache Bank B ECC or parity error",
+ "Data Tag Cache Bank A ECC or parity error",
+ "Data Tag Cache Bank B ECC or parity error",
+ "Instruction Cache Bank A ECC or parity error",
+ "Instruction Cache Bank B ECC or parity error",
+ "Instruction Tag Cache Bank A ECC or parity error",
+ "Instruction Tag Cache Bank B ECC or parity error",
+ "System Hub Read Buffer ECC or parity error",
+ "MPDMA TVF DVSEC Memory ECC or parity error",
+ "MPDMA TVF MMIO Mailbox0 ECC or parity error",
+ "MPDMA TVF MMIO Mailbox1 ECC or parity error",
+ "MPDMA TVF Doorbell Memory ECC or parity error",
+ "MPDMA TVF SDP Slave Memory 0 ECC or parity error",
+ "MPDMA TVF SDP Slave Memory 1 ECC or parity error",
+ "MPDMA TVF SDP Slave Memory 2 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 0 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 1 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 2 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 3 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 4 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 5 ECC or parity error",
+ "MPDMA TVF SDP Master Memory 6 ECC or parity error",
+ "MPDMA PTE Command FIFO ECC or parity error",
+ "MPDMA PTE Hub Data FIFO ECC or parity error",
+ "MPDMA PTE Internal Data FIFO ECC or parity error",
+ "MPDMA PTE Command Memory DMA ECC or parity error",
+ "MPDMA PTE Command Memory Internal ECC or parity error",
+ "MPDMA PTE DMA Completion FIFO ECC or parity error",
+ "MPDMA PTE Tablewalk Completion FIFO ECC or parity error",
+ "MPDMA PTE Descriptor Completion FIFO ECC or parity error",
+ "MPDMA PTE ReadOnly Completion FIFO ECC or parity error",
+ "MPDMA PTE DirectWrite Completion FIFO ECC or parity error",
+ "SDP Watchdog Timer expired",
+};
+
+static const char * const smca_nbio_mce_desc[] = {
+ "ECC or Parity error",
+ "PCIE error",
+ "SDP ErrEvent error",
+ "SDP Egress Poison Error",
+ "IOHC Internal Poison Error",
+};
+
+static const char * const smca_pcie_mce_desc[] = {
+ "CCIX PER Message logging",
+ "CCIX Read Response with Status: Non-Data Error",
+ "CCIX Write Response with Status: Non-Data Error",
+ "CCIX Read Response with Status: Data Error",
+ "CCIX Non-okay write response with data error",
+};
+
+static const char * const smca_pcie2_mce_desc[] = {
+ "SDP Parity Error logging",
+};
+
+static const char * const smca_xgmipcs_mce_desc[] = {
+ "Data Loss Error",
+ "Training Error",
+ "Flow Control Acknowledge Error",
+ "Rx Fifo Underflow Error",
+ "Rx Fifo Overflow Error",
+ "CRC Error",
+ "BER Exceeded Error",
+ "Tx Vcid Data Error",
+ "Replay Buffer Parity Error",
+ "Data Parity Error",
+ "Replay Fifo Overflow Error",
+ "Replay Fifo Underflow Error",
+ "Elastic Fifo Overflow Error",
+ "Deskew Error",
+ "Flow Control CRC Error",
+ "Data Startup Limit Error",
+ "FC Init Timeout Error",
+ "Recovery Timeout Error",
+ "Ready Serial Timeout Error",
+ "Ready Serial Attempt Error",
+ "Recovery Attempt Error",
+ "Recovery Relock Attempt Error",
+ "Replay Attempt Error",
+ "Sync Header Error",
+ "Tx Replay Timeout Error",
+ "Rx Replay Timeout Error",
+ "LinkSub Tx Timeout Error",
+ "LinkSub Rx Timeout Error",
+ "Rx CMD Packet Error",
+};
+
+static const char * const smca_xgmiphy_mce_desc[] = {
+ "RAM ECC Error",
+ "ARC instruction buffer parity error",
+ "ARC data buffer parity error",
+ "PHY APB error",
+};
+
+static const char * const smca_nbif_mce_desc[] = {
+ "Timeout error from GMI",
+ "SRAM ECC error",
+ "NTB Error Event",
+ "SDP Parity error",
+};
+
+static const char * const smca_sata_mce_desc[] = {
+ "Parity error for port 0",
+ "Parity error for port 1",
+ "Parity error for port 2",
+ "Parity error for port 3",
+ "Parity error for port 4",
+ "Parity error for port 5",
+ "Parity error for port 6",
+ "Parity error for port 7",
+};
+
+static const char * const smca_usb_mce_desc[] = {
+ "Parity error or ECC error for S0 RAM0",
+ "Parity error or ECC error for S0 RAM1",
+ "Parity error or ECC error for S0 RAM2",
+ "Parity error for PHY RAM0",
+ "Parity error for PHY RAM1",
+ "AXI Slave Response error",
+};
+
+static const char * const smca_gmipcs_mce_desc[] = {
+ "Data Loss Error",
+ "Training Error",
+ "Replay Parity Error",
+ "Rx Fifo Underflow Error",
+ "Rx Fifo Overflow Error",
+ "CRC Error",
+ "BER Exceeded Error",
+ "Tx Fifo Underflow Error",
+ "Replay Buffer Parity Error",
+ "Tx Overflow Error",
+ "Replay Fifo Overflow Error",
+ "Replay Fifo Underflow Error",
+ "Elastic Fifo Overflow Error",
+ "Deskew Error",
+ "Offline Error",
+ "Data Startup Limit Error",
+ "FC Init Timeout Error",
+ "Recovery Timeout Error",
+ "Ready Serial Timeout Error",
+ "Ready Serial Attempt Error",
+ "Recovery Attempt Error",
+ "Recovery Relock Attempt Error",
+ "Deskew Abort Error",
+ "Rx Buffer Error",
+ "Rx LFDS Fifo Overflow Error",
+ "Rx LFDS Fifo Underflow Error",
+ "LinkSub Tx Timeout Error",
+ "LinkSub Rx Timeout Error",
+ "Rx CMD Packet Error",
+ "LFDS Training Timeout Error",
+ "LFDS FC Init Timeout Error",
+ "Data Loss Error",
+};
+
+struct smca_mce_desc {
+ const char * const *descs;
+ unsigned int num_descs;
+};
+
+static struct smca_mce_desc smca_mce_descs[] = {
+ [SMCA_LS] = { smca_ls_mce_desc, ARRAY_SIZE(smca_ls_mce_desc) },
+ [SMCA_LS_V2] = { smca_ls2_mce_desc, ARRAY_SIZE(smca_ls2_mce_desc) },
+ [SMCA_IF] = { smca_if_mce_desc, ARRAY_SIZE(smca_if_mce_desc) },
+ [SMCA_L2_CACHE] = { smca_l2_mce_desc, ARRAY_SIZE(smca_l2_mce_desc) },
+ [SMCA_DE] = { smca_de_mce_desc, ARRAY_SIZE(smca_de_mce_desc) },
+ [SMCA_EX] = { smca_ex_mce_desc, ARRAY_SIZE(smca_ex_mce_desc) },
+ [SMCA_FP] = { smca_fp_mce_desc, ARRAY_SIZE(smca_fp_mce_desc) },
+ [SMCA_L3_CACHE] = { smca_l3_mce_desc, ARRAY_SIZE(smca_l3_mce_desc) },
+ [SMCA_CS] = { smca_cs_mce_desc, ARRAY_SIZE(smca_cs_mce_desc) },
+ [SMCA_CS_V2] = { smca_cs2_mce_desc, ARRAY_SIZE(smca_cs2_mce_desc) },
+ [SMCA_PIE] = { smca_pie_mce_desc, ARRAY_SIZE(smca_pie_mce_desc) },
+ [SMCA_UMC] = { smca_umc_mce_desc, ARRAY_SIZE(smca_umc_mce_desc) },
+ [SMCA_UMC_V2] = { smca_umc2_mce_desc, ARRAY_SIZE(smca_umc2_mce_desc) },
+ [SMCA_PB] = { smca_pb_mce_desc, ARRAY_SIZE(smca_pb_mce_desc) },
+ [SMCA_PSP] = { smca_psp_mce_desc, ARRAY_SIZE(smca_psp_mce_desc) },
+ [SMCA_PSP_V2] = { smca_psp2_mce_desc, ARRAY_SIZE(smca_psp2_mce_desc) },
+ [SMCA_SMU] = { smca_smu_mce_desc, ARRAY_SIZE(smca_smu_mce_desc) },
+ [SMCA_SMU_V2] = { smca_smu2_mce_desc, ARRAY_SIZE(smca_smu2_mce_desc) },
+ [SMCA_MP5] = { smca_mp5_mce_desc, ARRAY_SIZE(smca_mp5_mce_desc) },
+ [SMCA_MPDMA] = { smca_mpdma_mce_desc, ARRAY_SIZE(smca_mpdma_mce_desc) },
+ [SMCA_NBIO] = { smca_nbio_mce_desc, ARRAY_SIZE(smca_nbio_mce_desc) },
+ [SMCA_PCIE] = { smca_pcie_mce_desc, ARRAY_SIZE(smca_pcie_mce_desc) },
+ [SMCA_PCIE_V2] = { smca_pcie2_mce_desc, ARRAY_SIZE(smca_pcie2_mce_desc) },
+ [SMCA_XGMI_PCS] = { smca_xgmipcs_mce_desc, ARRAY_SIZE(smca_xgmipcs_mce_desc) },
+ /* NBIF and SHUB have the same error descriptions, for now. */
+ [SMCA_NBIF] = { smca_nbif_mce_desc, ARRAY_SIZE(smca_nbif_mce_desc) },
+ [SMCA_SHUB] = { smca_nbif_mce_desc, ARRAY_SIZE(smca_nbif_mce_desc) },
+ [SMCA_SATA] = { smca_sata_mce_desc, ARRAY_SIZE(smca_sata_mce_desc) },
+ [SMCA_USB] = { smca_usb_mce_desc, ARRAY_SIZE(smca_usb_mce_desc) },
+ [SMCA_GMI_PCS] = { smca_gmipcs_mce_desc, ARRAY_SIZE(smca_gmipcs_mce_desc) },
+ /* All the PHY bank types have the same error descriptions, for now. */
+ [SMCA_XGMI_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
+ [SMCA_WAFL_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
+ [SMCA_GMI_PHY] = { smca_xgmiphy_mce_desc, ARRAY_SIZE(smca_xgmiphy_mce_desc) },
+};
+
+static bool f12h_mc0_mce(u16 ec, u8 xec)
+{
+ bool ret = false;
+
+ if (MEM_ERROR(ec)) {
+ u8 ll = LL(ec);
+ ret = true;
+
+ if (ll == LL_L2)
+ pr_cont("during L1 linefill from L2.\n");
+ else if (ll == LL_L1)
+ pr_cont("Data/Tag %s error.\n", R4_MSG(ec));
+ else
+ ret = false;
+ }
+ return ret;
+}
+
+static bool f10h_mc0_mce(u16 ec, u8 xec)
+{
+ if (R4(ec) == R4_GEN && LL(ec) == LL_L1) {
+ pr_cont("during data scrub.\n");
+ return true;
+ }
+ return f12h_mc0_mce(ec, xec);
+}
+
+static bool k8_mc0_mce(u16 ec, u8 xec)
+{
+ if (BUS_ERROR(ec)) {
+ pr_cont("during system linefill.\n");
+ return true;
+ }
+
+ return f10h_mc0_mce(ec, xec);
+}
+
+static bool cat_mc0_mce(u16 ec, u8 xec)
+{
+ u8 r4 = R4(ec);
+ bool ret = true;
+
+ if (MEM_ERROR(ec)) {
+
+ if (TT(ec) != TT_DATA || LL(ec) != LL_L1)
+ return false;
+
+ switch (r4) {
+ case R4_DRD:
+ case R4_DWR:
+ pr_cont("Data/Tag parity error due to %s.\n",
+ (r4 == R4_DRD ? "load/hw prf" : "store"));
+ break;
+ case R4_EVICT:
+ pr_cont("Copyback parity error on a tag miss.\n");
+ break;
+ case R4_SNOOP:
+ pr_cont("Tag parity error during snoop.\n");
+ break;
+ default:
+ ret = false;
+ }
+ } else if (BUS_ERROR(ec)) {
+
+ if ((II(ec) != II_MEM && II(ec) != II_IO) || LL(ec) != LL_LG)
+ return false;
+
+ pr_cont("System read data error on a ");
+
+ switch (r4) {
+ case R4_RD:
+ pr_cont("TLB reload.\n");
+ break;
+ case R4_DWR:
+ pr_cont("store.\n");
+ break;
+ case R4_DRD:
+ pr_cont("load.\n");
+ break;
+ default:
+ ret = false;
+ }
+ } else {
+ ret = false;
+ }
+
+ return ret;
+}
+
+static bool f15h_mc0_mce(u16 ec, u8 xec)
+{
+ bool ret = true;
+
+ if (MEM_ERROR(ec)) {
+
+ switch (xec) {
+ case 0x0:
+ pr_cont("Data Array access error.\n");
+ break;
+
+ case 0x1:
+ pr_cont("UC error during a linefill from L2/NB.\n");
+ break;
+
+ case 0x2:
+ case 0x11:
+ pr_cont("STQ access error.\n");
+ break;
+
+ case 0x3:
+ pr_cont("SCB access error.\n");
+ break;
+
+ case 0x10:
+ pr_cont("Tag error.\n");
+ break;
+
+ case 0x12:
+ pr_cont("LDQ access error.\n");
+ break;
+
+ default:
+ ret = false;
+ }
+ } else if (BUS_ERROR(ec)) {
+
+ if (!xec)
+ pr_cont("System Read Data Error.\n");
+ else
+ pr_cont(" Internal error condition type %d.\n", xec);
+ } else if (INT_ERROR(ec)) {
+ if (xec <= 0x1f)
+ pr_cont("Hardware Assert.\n");
+ else
+ ret = false;
+
+ } else
+ ret = false;
+
+ return ret;
+}
+
+static void decode_mc0_mce(struct mce *m)
+{
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, xec_mask);
+
+ pr_emerg(HW_ERR "MC0 Error: ");
+
+ /* TLB error signatures are the same across families */
+ if (TLB_ERROR(ec)) {
+ if (TT(ec) == TT_DATA) {
+ pr_cont("%s TLB %s.\n", LL_MSG(ec),
+ ((xec == 2) ? "locked miss"
+ : (xec ? "multimatch" : "parity")));
+ return;
+ }
+ } else if (fam_ops.mc0_mce(ec, xec))
+ ;
+ else
+ pr_emerg(HW_ERR "Corrupted MC0 MCE info?\n");
+}
+
+static bool k8_mc1_mce(u16 ec, u8 xec)
+{
+ u8 ll = LL(ec);
+ bool ret = true;
+
+ if (!MEM_ERROR(ec))
+ return false;
+
+ if (ll == 0x2)
+ pr_cont("during a linefill from L2.\n");
+ else if (ll == 0x1) {
+ switch (R4(ec)) {
+ case R4_IRD:
+ pr_cont("Parity error during data load.\n");
+ break;
+
+ case R4_EVICT:
+ pr_cont("Copyback Parity/Victim error.\n");
+ break;
+
+ case R4_SNOOP:
+ pr_cont("Tag Snoop error.\n");
+ break;
+
+ default:
+ ret = false;
+ break;
+ }
+ } else
+ ret = false;
+
+ return ret;
+}
+
+static bool cat_mc1_mce(u16 ec, u8 xec)
+{
+ u8 r4 = R4(ec);
+ bool ret = true;
+
+ if (!MEM_ERROR(ec))
+ return false;
+
+ if (TT(ec) != TT_INSTR)
+ return false;
+
+ if (r4 == R4_IRD)
+ pr_cont("Data/tag array parity error for a tag hit.\n");
+ else if (r4 == R4_SNOOP)
+ pr_cont("Tag error during snoop/victimization.\n");
+ else if (xec == 0x0)
+ pr_cont("Tag parity error from victim castout.\n");
+ else if (xec == 0x2)
+ pr_cont("Microcode patch RAM parity error.\n");
+ else
+ ret = false;
+
+ return ret;
+}
+
+static bool f15h_mc1_mce(u16 ec, u8 xec)
+{
+ bool ret = true;
+
+ if (!MEM_ERROR(ec))
+ return false;
+
+ switch (xec) {
+ case 0x0 ... 0xa:
+ pr_cont("%s.\n", f15h_mc1_mce_desc[xec]);
+ break;
+
+ case 0xd:
+ pr_cont("%s.\n", f15h_mc1_mce_desc[xec-2]);
+ break;
+
+ case 0x10:
+ pr_cont("%s.\n", f15h_mc1_mce_desc[xec-4]);
+ break;
+
+ case 0x11 ... 0x15:
+ pr_cont("Decoder %s parity error.\n", f15h_mc1_mce_desc[xec-4]);
+ break;
+
+ default:
+ ret = false;
+ }
+ return ret;
+}
+
+static void decode_mc1_mce(struct mce *m)
+{
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, xec_mask);
+
+ pr_emerg(HW_ERR "MC1 Error: ");
+
+ if (TLB_ERROR(ec))
+ pr_cont("%s TLB %s.\n", LL_MSG(ec),
+ (xec ? "multimatch" : "parity error"));
+ else if (BUS_ERROR(ec)) {
+ bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58)));
+
+ pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read"));
+ } else if (INT_ERROR(ec)) {
+ if (xec <= 0x3f)
+ pr_cont("Hardware Assert.\n");
+ else
+ goto wrong_mc1_mce;
+ } else if (fam_ops.mc1_mce(ec, xec))
+ ;
+ else
+ goto wrong_mc1_mce;
+
+ return;
+
+wrong_mc1_mce:
+ pr_emerg(HW_ERR "Corrupted MC1 MCE info?\n");
+}
+
+static bool k8_mc2_mce(u16 ec, u8 xec)
+{
+ bool ret = true;
+
+ if (xec == 0x1)
+ pr_cont(" in the write data buffers.\n");
+ else if (xec == 0x3)
+ pr_cont(" in the victim data buffers.\n");
+ else if (xec == 0x2 && MEM_ERROR(ec))
+ pr_cont(": %s error in the L2 cache tags.\n", R4_MSG(ec));
+ else if (xec == 0x0) {
+ if (TLB_ERROR(ec))
+ pr_cont("%s error in a Page Descriptor Cache or Guest TLB.\n",
+ TT_MSG(ec));
+ else if (BUS_ERROR(ec))
+ pr_cont(": %s/ECC error in data read from NB: %s.\n",
+ R4_MSG(ec), PP_MSG(ec));
+ else if (MEM_ERROR(ec)) {
+ u8 r4 = R4(ec);
+
+ if (r4 >= 0x7)
+ pr_cont(": %s error during data copyback.\n",
+ R4_MSG(ec));
+ else if (r4 <= 0x1)
+ pr_cont(": %s parity/ECC error during data "
+ "access from L2.\n", R4_MSG(ec));
+ else
+ ret = false;
+ } else
+ ret = false;
+ } else
+ ret = false;
+
+ return ret;
+}
+
+static bool f15h_mc2_mce(u16 ec, u8 xec)
+{
+ bool ret = true;
+
+ if (TLB_ERROR(ec)) {
+ if (xec == 0x0)
+ pr_cont("Data parity TLB read error.\n");
+ else if (xec == 0x1)
+ pr_cont("Poison data provided for TLB fill.\n");
+ else
+ ret = false;
+ } else if (BUS_ERROR(ec)) {
+ if (xec > 2)
+ ret = false;
+
+ pr_cont("Error during attempted NB data read.\n");
+ } else if (MEM_ERROR(ec)) {
+ switch (xec) {
+ case 0x4 ... 0xc:
+ pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x4]);
+ break;
+
+ case 0x10 ... 0x14:
+ pr_cont("%s.\n", f15h_mc2_mce_desc[xec - 0x7]);
+ break;
+
+ default:
+ ret = false;
+ }
+ } else if (INT_ERROR(ec)) {
+ if (xec <= 0x3f)
+ pr_cont("Hardware Assert.\n");
+ else
+ ret = false;
+ }
+
+ return ret;
+}
+
+static bool f16h_mc2_mce(u16 ec, u8 xec)
+{
+ u8 r4 = R4(ec);
+
+ if (!MEM_ERROR(ec))
+ return false;
+
+ switch (xec) {
+ case 0x04 ... 0x05:
+ pr_cont("%cBUFF parity error.\n", (r4 == R4_RD) ? 'I' : 'O');
+ break;
+
+ case 0x09 ... 0x0b:
+ case 0x0d ... 0x0f:
+ pr_cont("ECC error in L2 tag (%s).\n",
+ ((r4 == R4_GEN) ? "BankReq" :
+ ((r4 == R4_SNOOP) ? "Prb" : "Fill")));
+ break;
+
+ case 0x10 ... 0x19:
+ case 0x1b:
+ pr_cont("ECC error in L2 data array (%s).\n",
+ (((r4 == R4_RD) && !(xec & 0x3)) ? "Hit" :
+ ((r4 == R4_GEN) ? "Attr" :
+ ((r4 == R4_EVICT) ? "Vict" : "Fill"))));
+ break;
+
+ case 0x1c ... 0x1d:
+ case 0x1f:
+ pr_cont("Parity error in L2 attribute bits (%s).\n",
+ ((r4 == R4_RD) ? "Hit" :
+ ((r4 == R4_GEN) ? "Attr" : "Fill")));
+ break;
+
+ default:
+ return false;
+ }
+
+ return true;
+}
+
+static void decode_mc2_mce(struct mce *m)
+{
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, xec_mask);
+
+ pr_emerg(HW_ERR "MC2 Error: ");
+
+ if (!fam_ops.mc2_mce(ec, xec))
+ pr_cont(HW_ERR "Corrupted MC2 MCE info?\n");
+}
+
+static void decode_mc3_mce(struct mce *m)
+{
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, xec_mask);
+
+ if (boot_cpu_data.x86 >= 0x14) {
+ pr_emerg("You shouldn't be seeing MC3 MCE on this cpu family,"
+ " please report on LKML.\n");
+ return;
+ }
+
+ pr_emerg(HW_ERR "MC3 Error");
+
+ if (xec == 0x0) {
+ u8 r4 = R4(ec);
+
+ if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR))
+ goto wrong_mc3_mce;
+
+ pr_cont(" during %s.\n", R4_MSG(ec));
+ } else
+ goto wrong_mc3_mce;
+
+ return;
+
+ wrong_mc3_mce:
+ pr_emerg(HW_ERR "Corrupted MC3 MCE info?\n");
+}
+
+static void decode_mc4_mce(struct mce *m)
+{
+ unsigned int fam = x86_family(m->cpuid);
+ int node_id = topology_die_id(m->extcpu);
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, 0x1f);
+ u8 offset = 0;
+
+ pr_emerg(HW_ERR "MC4 Error (node %d): ", node_id);
+
+ switch (xec) {
+ case 0x0 ... 0xe:
+
+ /* special handling for DRAM ECCs */
+ if (xec == 0x0 || xec == 0x8) {
+ /* no ECCs on F11h */
+ if (fam == 0x11)
+ goto wrong_mc4_mce;
+
+ pr_cont("%s.\n", mc4_mce_desc[xec]);
+
+ if (decode_dram_ecc)
+ decode_dram_ecc(node_id, m);
+ return;
+ }
+ break;
+
+ case 0xf:
+ if (TLB_ERROR(ec))
+ pr_cont("GART Table Walk data error.\n");
+ else if (BUS_ERROR(ec))
+ pr_cont("DMA Exclusion Vector Table Walk error.\n");
+ else
+ goto wrong_mc4_mce;
+ return;
+
+ case 0x19:
+ if (fam == 0x15 || fam == 0x16)
+ pr_cont("Compute Unit Data Error.\n");
+ else
+ goto wrong_mc4_mce;
+ return;
+
+ case 0x1c ... 0x1f:
+ offset = 13;
+ break;
+
+ default:
+ goto wrong_mc4_mce;
+ }
+
+ pr_cont("%s.\n", mc4_mce_desc[xec - offset]);
+ return;
+
+ wrong_mc4_mce:
+ pr_emerg(HW_ERR "Corrupted MC4 MCE info?\n");
+}
+
+static void decode_mc5_mce(struct mce *m)
+{
+ unsigned int fam = x86_family(m->cpuid);
+ u16 ec = EC(m->status);
+ u8 xec = XEC(m->status, xec_mask);
+
+ if (fam == 0xf || fam == 0x11)
+ goto wrong_mc5_mce;
+
+ pr_emerg(HW_ERR "MC5 Error: ");
+
+ if (INT_ERROR(ec)) {
+ if (xec <= 0x1f) {
+ pr_cont("Hardware Assert.\n");
+ return;
+ } else
+ goto wrong_mc5_mce;
+ }
+
+ if (xec == 0x0 || xec == 0xc)
+ pr_cont("%s.\n", mc5_mce_desc[xec]);
+ else if (xec <= 0xd)
+ pr_cont("%s parity error.\n", mc5_mce_desc[xec]);
+ else
+ goto wrong_mc5_mce;
+
+ return;
+
+ wrong_mc5_mce:
+ pr_emerg(HW_ERR "Corrupted MC5 MCE info?\n");
+}
+
+static void decode_mc6_mce(struct mce *m)
+{
+ u8 xec = XEC(m->status, xec_mask);
+
+ pr_emerg(HW_ERR "MC6 Error: ");
+
+ if (xec > 0x5)
+ goto wrong_mc6_mce;
+
+ pr_cont("%s parity error.\n", mc6_mce_desc[xec]);
+ return;
+
+ wrong_mc6_mce:
+ pr_emerg(HW_ERR "Corrupted MC6 MCE info?\n");
+}
+
+/* Decode errors according to Scalable MCA specification */
+static void decode_smca_error(struct mce *m)
+{
+ enum smca_bank_types bank_type = smca_get_bank_type(m->extcpu, m->bank);
+ const char *ip_name;
+ u8 xec = XEC(m->status, xec_mask);
+
+ if (bank_type >= N_SMCA_BANK_TYPES)
+ return;
+
+ if (bank_type == SMCA_RESERVED) {
+ pr_emerg(HW_ERR "Bank %d is reserved.\n", m->bank);
+ return;
+ }
+
+ ip_name = smca_get_long_name(bank_type);
+
+ pr_emerg(HW_ERR "%s Ext. Error Code: %d", ip_name, xec);
+
+ /* Only print the decode of valid error codes */
+ if (xec < smca_mce_descs[bank_type].num_descs)
+ pr_cont(", %s.\n", smca_mce_descs[bank_type].descs[xec]);
+
+ if ((bank_type == SMCA_UMC || bank_type == SMCA_UMC_V2) &&
+ xec == 0 && decode_dram_ecc)
+ decode_dram_ecc(topology_die_id(m->extcpu), m);
+}
+
+static inline void amd_decode_err_code(u16 ec)
+{
+ if (INT_ERROR(ec)) {
+ pr_emerg(HW_ERR "internal: %s\n", UU_MSG(ec));
+ return;
+ }
+
+ pr_emerg(HW_ERR "cache level: %s", LL_MSG(ec));
+
+ if (BUS_ERROR(ec))
+ pr_cont(", mem/io: %s", II_MSG(ec));
+ else
+ pr_cont(", tx: %s", TT_MSG(ec));
+
+ if (MEM_ERROR(ec) || BUS_ERROR(ec)) {
+ pr_cont(", mem-tx: %s", R4_MSG(ec));
+
+ if (BUS_ERROR(ec))
+ pr_cont(", part-proc: %s (%s)", PP_MSG(ec), TO_MSG(ec));
+ }
+
+ pr_cont("\n");
+}
+
+static const char *decode_error_status(struct mce *m)
+{
+ if (m->status & MCI_STATUS_UC) {
+ if (m->status & MCI_STATUS_PCC)
+ return "System Fatal error.";
+ if (m->mcgstatus & MCG_STATUS_RIPV)
+ return "Uncorrected, software restartable error.";
+ return "Uncorrected, software containable error.";
+ }
+
+ if (m->status & MCI_STATUS_DEFERRED)
+ return "Deferred error, no action required.";
+
+ return "Corrected error, no action required.";
+}
+
+static int
+amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data)
+{
+ struct mce *m = (struct mce *)data;
+ unsigned int fam = x86_family(m->cpuid);
+ int ecc;
+
+ if (m->kflags & MCE_HANDLED_CEC)
+ return NOTIFY_DONE;
+
+ pr_emerg(HW_ERR "%s\n", decode_error_status(m));
+
+ pr_emerg(HW_ERR "CPU:%d (%x:%x:%x) MC%d_STATUS[%s|%s|%s|%s|%s",
+ m->extcpu,
+ fam, x86_model(m->cpuid), x86_stepping(m->cpuid),
+ m->bank,
+ ((m->status & MCI_STATUS_OVER) ? "Over" : "-"),
+ ((m->status & MCI_STATUS_UC) ? "UE" :
+ (m->status & MCI_STATUS_DEFERRED) ? "-" : "CE"),
+ ((m->status & MCI_STATUS_MISCV) ? "MiscV" : "-"),
+ ((m->status & MCI_STATUS_ADDRV) ? "AddrV" : "-"),
+ ((m->status & MCI_STATUS_PCC) ? "PCC" : "-"));
+
+ if (boot_cpu_has(X86_FEATURE_SMCA)) {
+ u32 low, high;
+ u32 addr = MSR_AMD64_SMCA_MCx_CONFIG(m->bank);
+
+ if (!rdmsr_safe(addr, &low, &high) &&
+ (low & MCI_CONFIG_MCAX))
+ pr_cont("|%s", ((m->status & MCI_STATUS_TCC) ? "TCC" : "-"));
+
+ pr_cont("|%s", ((m->status & MCI_STATUS_SYNDV) ? "SyndV" : "-"));
+ }
+
+ /* do the two bits[14:13] together */
+ ecc = (m->status >> 45) & 0x3;
+ if (ecc)
+ pr_cont("|%sECC", ((ecc == 2) ? "C" : "U"));
+
+ if (fam >= 0x15) {
+ pr_cont("|%s", (m->status & MCI_STATUS_DEFERRED ? "Deferred" : "-"));
+
+ /* F15h, bank4, bit 43 is part of McaStatSubCache. */
+ if (fam != 0x15 || m->bank != 4)
+ pr_cont("|%s", (m->status & MCI_STATUS_POISON ? "Poison" : "-"));
+ }
+
+ if (fam >= 0x17)
+ pr_cont("|%s", (m->status & MCI_STATUS_SCRUB ? "Scrub" : "-"));
+
+ pr_cont("]: 0x%016llx\n", m->status);
+
+ if (m->status & MCI_STATUS_ADDRV)
+ pr_emerg(HW_ERR "Error Addr: 0x%016llx\n", m->addr);
+
+ if (m->ppin)
+ pr_emerg(HW_ERR "PPIN: 0x%016llx\n", m->ppin);
+
+ if (boot_cpu_has(X86_FEATURE_SMCA)) {
+ pr_emerg(HW_ERR "IPID: 0x%016llx", m->ipid);
+
+ if (m->status & MCI_STATUS_SYNDV)
+ pr_cont(", Syndrome: 0x%016llx", m->synd);
+
+ pr_cont("\n");
+
+ decode_smca_error(m);
+ goto err_code;
+ }
+
+ if (m->tsc)
+ pr_emerg(HW_ERR "TSC: %llu\n", m->tsc);
+
+ /* Doesn't matter which member to test. */
+ if (!fam_ops.mc0_mce)
+ goto err_code;
+
+ switch (m->bank) {
+ case 0:
+ decode_mc0_mce(m);
+ break;
+
+ case 1:
+ decode_mc1_mce(m);
+ break;
+
+ case 2:
+ decode_mc2_mce(m);
+ break;
+
+ case 3:
+ decode_mc3_mce(m);
+ break;
+
+ case 4:
+ decode_mc4_mce(m);
+ break;
+
+ case 5:
+ decode_mc5_mce(m);
+ break;
+
+ case 6:
+ decode_mc6_mce(m);
+ break;
+
+ default:
+ break;
+ }
+
+ err_code:
+ amd_decode_err_code(m->status & 0xffff);
+
+ m->kflags |= MCE_HANDLED_EDAC;
+ return NOTIFY_OK;
+}
+
+static struct notifier_block amd_mce_dec_nb = {
+ .notifier_call = amd_decode_mce,
+ .priority = MCE_PRIO_EDAC,
+};
+
+static int __init mce_amd_init(void)
+{
+ struct cpuinfo_x86 *c = &boot_cpu_data;
+
+ if (c->x86_vendor != X86_VENDOR_AMD &&
+ c->x86_vendor != X86_VENDOR_HYGON)
+ return -ENODEV;
+
+ if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
+ return -ENODEV;
+
+ if (boot_cpu_has(X86_FEATURE_SMCA)) {
+ xec_mask = 0x3f;
+ goto out;
+ }
+
+ switch (c->x86) {
+ case 0xf:
+ fam_ops.mc0_mce = k8_mc0_mce;
+ fam_ops.mc1_mce = k8_mc1_mce;
+ fam_ops.mc2_mce = k8_mc2_mce;
+ break;
+
+ case 0x10:
+ fam_ops.mc0_mce = f10h_mc0_mce;
+ fam_ops.mc1_mce = k8_mc1_mce;
+ fam_ops.mc2_mce = k8_mc2_mce;
+ break;
+
+ case 0x11:
+ fam_ops.mc0_mce = k8_mc0_mce;
+ fam_ops.mc1_mce = k8_mc1_mce;
+ fam_ops.mc2_mce = k8_mc2_mce;
+ break;
+
+ case 0x12:
+ fam_ops.mc0_mce = f12h_mc0_mce;
+ fam_ops.mc1_mce = k8_mc1_mce;
+ fam_ops.mc2_mce = k8_mc2_mce;
+ break;
+
+ case 0x14:
+ fam_ops.mc0_mce = cat_mc0_mce;
+ fam_ops.mc1_mce = cat_mc1_mce;
+ fam_ops.mc2_mce = k8_mc2_mce;
+ break;
+
+ case 0x15:
+ xec_mask = c->x86_model == 0x60 ? 0x3f : 0x1f;
+
+ fam_ops.mc0_mce = f15h_mc0_mce;
+ fam_ops.mc1_mce = f15h_mc1_mce;
+ fam_ops.mc2_mce = f15h_mc2_mce;
+ break;
+
+ case 0x16:
+ xec_mask = 0x1f;
+ fam_ops.mc0_mce = cat_mc0_mce;
+ fam_ops.mc1_mce = cat_mc1_mce;
+ fam_ops.mc2_mce = f16h_mc2_mce;
+ break;
+
+ case 0x17:
+ case 0x18:
+ pr_warn_once("Decoding supported only on Scalable MCA processors.\n");
+ return -EINVAL;
+
+ default:
+ printk(KERN_WARNING "Huh? What family is it: 0x%x?!\n", c->x86);
+ return -EINVAL;
+ }
+
+out:
+ pr_info("MCE: In-kernel MCE decoding enabled.\n");
+
+ mce_register_decode_chain(&amd_mce_dec_nb);
+
+ return 0;
+}
+early_initcall(mce_amd_init);
+
+#ifdef MODULE
+static void __exit mce_amd_exit(void)
+{
+ mce_unregister_decode_chain(&amd_mce_dec_nb);
+}
+
+MODULE_DESCRIPTION("AMD MCE decoder");
+MODULE_ALIAS("edac-mce-amd");
+MODULE_LICENSE("GPL");
+module_exit(mce_amd_exit);
+#endif
diff --git a/drivers/edac/mce_amd.h b/drivers/edac/mce_amd.h
new file mode 100644
index 0000000000..4811b18d96
--- /dev/null
+++ b/drivers/edac/mce_amd.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _EDAC_MCE_AMD_H
+#define _EDAC_MCE_AMD_H
+
+#include <linux/notifier.h>
+
+#include <asm/mce.h>
+
+#define EC(x) ((x) & 0xffff)
+
+#define LOW_SYNDROME(x) (((x) >> 15) & 0xff)
+#define HIGH_SYNDROME(x) (((x) >> 24) & 0xff)
+
+#define TLB_ERROR(x) (((x) & 0xFFF0) == 0x0010)
+#define MEM_ERROR(x) (((x) & 0xFF00) == 0x0100)
+#define BUS_ERROR(x) (((x) & 0xF800) == 0x0800)
+#define INT_ERROR(x) (((x) & 0xF4FF) == 0x0400)
+
+#define TT(x) (((x) >> 2) & 0x3)
+#define TT_MSG(x) tt_msgs[TT(x)]
+#define II(x) (((x) >> 2) & 0x3)
+#define II_MSG(x) ii_msgs[II(x)]
+#define LL(x) ((x) & 0x3)
+#define LL_MSG(x) ll_msgs[LL(x)]
+#define TO(x) (((x) >> 8) & 0x1)
+#define TO_MSG(x) to_msgs[TO(x)]
+#define PP(x) (((x) >> 9) & 0x3)
+#define PP_MSG(x) pp_msgs[PP(x)]
+#define UU(x) (((x) >> 8) & 0x3)
+#define UU_MSG(x) uu_msgs[UU(x)]
+
+#define R4(x) (((x) >> 4) & 0xf)
+#define R4_MSG(x) ((R4(x) < 9) ? rrrr_msgs[R4(x)] : "Wrong R4!")
+
+extern const char * const pp_msgs[];
+
+enum tt_ids {
+ TT_INSTR = 0,
+ TT_DATA,
+ TT_GEN,
+ TT_RESV,
+};
+
+enum ll_ids {
+ LL_RESV = 0,
+ LL_L1,
+ LL_L2,
+ LL_LG,
+};
+
+enum ii_ids {
+ II_MEM = 0,
+ II_RESV,
+ II_IO,
+ II_GEN,
+};
+
+enum rrrr_ids {
+ R4_GEN = 0,
+ R4_RD,
+ R4_WR,
+ R4_DRD,
+ R4_DWR,
+ R4_IRD,
+ R4_PREF,
+ R4_EVICT,
+ R4_SNOOP,
+};
+
+/*
+ * per-family decoder ops
+ */
+struct amd_decoder_ops {
+ bool (*mc0_mce)(u16, u8);
+ bool (*mc1_mce)(u16, u8);
+ bool (*mc2_mce)(u16, u8);
+};
+
+void amd_register_ecc_decoder(void (*f)(int, struct mce *));
+void amd_unregister_ecc_decoder(void (*f)(int, struct mce *));
+
+#endif /* _EDAC_MCE_AMD_H */
diff --git a/drivers/edac/mpc85xx_edac.c b/drivers/edac/mpc85xx_edac.c
new file mode 100644
index 0000000000..2b5703e506
--- /dev/null
+++ b/drivers/edac/mpc85xx_edac.c
@@ -0,0 +1,713 @@
+/*
+ * Freescale MPC85xx Memory Controller kernel module
+ *
+ * Parts Copyrighted (c) 2013 by Freescale Semiconductor, Inc.
+ *
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2006-2007 (c) MontaVista Software, Inc. This file is licensed under
+ * the terms of the GNU General Public License version 2. This program
+ * is licensed "as is" without any warranty of any kind, whether express
+ * or implied.
+ *
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/ctype.h>
+#include <linux/io.h>
+#include <linux/mod_devicetable.h>
+#include <linux/edac.h>
+#include <linux/smp.h>
+#include <linux/gfp.h>
+#include <linux/fsl/edac.h>
+
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include "edac_module.h"
+#include "mpc85xx_edac.h"
+#include "fsl_ddr_edac.h"
+
+static int edac_dev_idx;
+#ifdef CONFIG_PCI
+static int edac_pci_idx;
+#endif
+
+/*
+ * PCI Err defines
+ */
+#ifdef CONFIG_PCI
+static u32 orig_pci_err_cap_dr;
+static u32 orig_pci_err_en;
+#endif
+
+static u32 orig_l2_err_disable;
+
+/**************************** PCI Err device ***************************/
+#ifdef CONFIG_PCI
+
+static void mpc85xx_pci_check(struct edac_pci_ctl_info *pci)
+{
+ struct mpc85xx_pci_pdata *pdata = pci->pvt_info;
+ u32 err_detect;
+
+ err_detect = in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR);
+
+ /* master aborts can happen during PCI config cycles */
+ if (!(err_detect & ~(PCI_EDE_MULTI_ERR | PCI_EDE_MST_ABRT))) {
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR, err_detect);
+ return;
+ }
+
+ pr_err("PCI error(s) detected\n");
+ pr_err("PCI/X ERR_DR register: %#08x\n", err_detect);
+
+ pr_err("PCI/X ERR_ATTRIB register: %#08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ATTRIB));
+ pr_err("PCI/X ERR_ADDR register: %#08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ADDR));
+ pr_err("PCI/X ERR_EXT_ADDR register: %#08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EXT_ADDR));
+ pr_err("PCI/X ERR_DL register: %#08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DL));
+ pr_err("PCI/X ERR_DH register: %#08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DH));
+
+ /* clear error bits */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR, err_detect);
+
+ if (err_detect & PCI_EDE_PERR_MASK)
+ edac_pci_handle_pe(pci, pci->ctl_name);
+
+ if ((err_detect & ~PCI_EDE_MULTI_ERR) & ~PCI_EDE_PERR_MASK)
+ edac_pci_handle_npe(pci, pci->ctl_name);
+}
+
+static void mpc85xx_pcie_check(struct edac_pci_ctl_info *pci)
+{
+ struct mpc85xx_pci_pdata *pdata = pci->pvt_info;
+ u32 err_detect, err_cap_stat;
+
+ err_detect = in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR);
+ err_cap_stat = in_be32(pdata->pci_vbase + MPC85XX_PCI_GAS_TIMR);
+
+ pr_err("PCIe error(s) detected\n");
+ pr_err("PCIe ERR_DR register: 0x%08x\n", err_detect);
+ pr_err("PCIe ERR_CAP_STAT register: 0x%08x\n", err_cap_stat);
+ pr_err("PCIe ERR_CAP_R0 register: 0x%08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCIE_ERR_CAP_R0));
+ pr_err("PCIe ERR_CAP_R1 register: 0x%08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCIE_ERR_CAP_R1));
+ pr_err("PCIe ERR_CAP_R2 register: 0x%08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCIE_ERR_CAP_R2));
+ pr_err("PCIe ERR_CAP_R3 register: 0x%08x\n",
+ in_be32(pdata->pci_vbase + MPC85XX_PCIE_ERR_CAP_R3));
+
+ /* clear error bits */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR, err_detect);
+
+ /* reset error capture */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_GAS_TIMR, err_cap_stat | 0x1);
+}
+
+static int mpc85xx_pcie_find_capability(struct device_node *np)
+{
+ struct pci_controller *hose;
+
+ if (!np)
+ return -EINVAL;
+
+ hose = pci_find_hose_for_OF_device(np);
+
+ return early_find_capability(hose, 0, 0, PCI_CAP_ID_EXP);
+}
+
+static irqreturn_t mpc85xx_pci_isr(int irq, void *dev_id)
+{
+ struct edac_pci_ctl_info *pci = dev_id;
+ struct mpc85xx_pci_pdata *pdata = pci->pvt_info;
+ u32 err_detect;
+
+ err_detect = in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR);
+
+ if (!err_detect)
+ return IRQ_NONE;
+
+ if (pdata->is_pcie)
+ mpc85xx_pcie_check(pci);
+ else
+ mpc85xx_pci_check(pci);
+
+ return IRQ_HANDLED;
+}
+
+static int mpc85xx_pci_err_probe(struct platform_device *op)
+{
+ struct edac_pci_ctl_info *pci;
+ struct mpc85xx_pci_pdata *pdata;
+ struct mpc85xx_edac_pci_plat_data *plat_data;
+ struct device_node *of_node;
+ struct resource r;
+ int res = 0;
+
+ if (!devres_open_group(&op->dev, mpc85xx_pci_err_probe, GFP_KERNEL))
+ return -ENOMEM;
+
+ pci = edac_pci_alloc_ctl_info(sizeof(*pdata), "mpc85xx_pci_err");
+ if (!pci)
+ return -ENOMEM;
+
+ /* make sure error reporting method is sane */
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_INT:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_INT;
+ break;
+ }
+
+ pdata = pci->pvt_info;
+ pdata->name = "mpc85xx_pci_err";
+
+ plat_data = op->dev.platform_data;
+ if (!plat_data) {
+ dev_err(&op->dev, "no platform data");
+ res = -ENXIO;
+ goto err;
+ }
+ of_node = plat_data->of_node;
+
+ if (mpc85xx_pcie_find_capability(of_node) > 0)
+ pdata->is_pcie = true;
+
+ dev_set_drvdata(&op->dev, pci);
+ pci->dev = &op->dev;
+ pci->mod_name = EDAC_MOD_STR;
+ pci->ctl_name = pdata->name;
+ pci->dev_name = dev_name(&op->dev);
+
+ if (edac_op_state == EDAC_OPSTATE_POLL) {
+ if (pdata->is_pcie)
+ pci->edac_check = mpc85xx_pcie_check;
+ else
+ pci->edac_check = mpc85xx_pci_check;
+ }
+
+ pdata->edac_idx = edac_pci_idx++;
+
+ res = of_address_to_resource(of_node, 0, &r);
+ if (res) {
+ pr_err("%s: Unable to get resource for PCI err regs\n", __func__);
+ goto err;
+ }
+
+ /* we only need the error registers */
+ r.start += 0xe00;
+
+ if (!devm_request_mem_region(&op->dev, r.start, resource_size(&r),
+ pdata->name)) {
+ pr_err("%s: Error while requesting mem region\n", __func__);
+ res = -EBUSY;
+ goto err;
+ }
+
+ pdata->pci_vbase = devm_ioremap(&op->dev, r.start, resource_size(&r));
+ if (!pdata->pci_vbase) {
+ pr_err("%s: Unable to setup PCI err regs\n", __func__);
+ res = -ENOMEM;
+ goto err;
+ }
+
+ if (pdata->is_pcie) {
+ orig_pci_err_cap_dr =
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ADDR);
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ADDR, ~0);
+ orig_pci_err_en =
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN);
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN, 0);
+ } else {
+ orig_pci_err_cap_dr =
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_CAP_DR);
+
+ /* PCI master abort is expected during config cycles */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_CAP_DR, 0x40);
+
+ orig_pci_err_en =
+ in_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN);
+
+ /* disable master abort reporting */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN, ~0x40);
+ }
+
+ /* clear error bits */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_DR, ~0);
+
+ /* reset error capture */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_GAS_TIMR, 0x1);
+
+ if (edac_pci_add_device(pci, pdata->edac_idx) > 0) {
+ edac_dbg(3, "failed edac_pci_add_device()\n");
+ goto err;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ pdata->irq = irq_of_parse_and_map(of_node, 0);
+ res = devm_request_irq(&op->dev, pdata->irq,
+ mpc85xx_pci_isr,
+ IRQF_SHARED,
+ "[EDAC] PCI err", pci);
+ if (res < 0) {
+ pr_err("%s: Unable to request irq %d for MPC85xx PCI err\n",
+ __func__, pdata->irq);
+ irq_dispose_mapping(pdata->irq);
+ res = -ENODEV;
+ goto err2;
+ }
+
+ pr_info(EDAC_MOD_STR " acquired irq %d for PCI Err\n",
+ pdata->irq);
+ }
+
+ if (pdata->is_pcie) {
+ /*
+ * Enable all PCIe error interrupt & error detect except invalid
+ * PEX_CONFIG_ADDR/PEX_CONFIG_DATA access interrupt generation
+ * enable bit and invalid PEX_CONFIG_ADDR/PEX_CONFIG_DATA access
+ * detection enable bit. Because PCIe bus code to initialize and
+ * configure these PCIe devices on booting will use some invalid
+ * PEX_CONFIG_ADDR/PEX_CONFIG_DATA, edac driver prints the much
+ * notice information. So disable this detect to fix ugly print.
+ */
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN, ~0
+ & ~PEX_ERR_ICCAIE_EN_BIT);
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ADDR, 0
+ | PEX_ERR_ICCAD_DISR_BIT);
+ }
+
+ devres_remove_group(&op->dev, mpc85xx_pci_err_probe);
+ edac_dbg(3, "success\n");
+ pr_info(EDAC_MOD_STR " PCI err registered\n");
+
+ return 0;
+
+err2:
+ edac_pci_del_device(&op->dev);
+err:
+ edac_pci_free_ctl_info(pci);
+ devres_release_group(&op->dev, mpc85xx_pci_err_probe);
+ return res;
+}
+
+static int mpc85xx_pci_err_remove(struct platform_device *op)
+{
+ struct edac_pci_ctl_info *pci = dev_get_drvdata(&op->dev);
+ struct mpc85xx_pci_pdata *pdata = pci->pvt_info;
+
+ edac_dbg(0, "\n");
+
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_ADDR, orig_pci_err_cap_dr);
+ out_be32(pdata->pci_vbase + MPC85XX_PCI_ERR_EN, orig_pci_err_en);
+
+ edac_pci_del_device(&op->dev);
+ edac_pci_free_ctl_info(pci);
+
+ return 0;
+}
+
+static const struct platform_device_id mpc85xx_pci_err_match[] = {
+ {
+ .name = "mpc85xx-pci-edac"
+ },
+ {}
+};
+
+static struct platform_driver mpc85xx_pci_err_driver = {
+ .probe = mpc85xx_pci_err_probe,
+ .remove = mpc85xx_pci_err_remove,
+ .id_table = mpc85xx_pci_err_match,
+ .driver = {
+ .name = "mpc85xx_pci_err",
+ .suppress_bind_attrs = true,
+ },
+};
+#endif /* CONFIG_PCI */
+
+/**************************** L2 Err device ***************************/
+
+/************************ L2 SYSFS parts ***********************************/
+
+static ssize_t mpc85xx_l2_inject_data_hi_show(struct edac_device_ctl_info
+ *edac_dev, char *data)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ return sprintf(data, "0x%08x",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJHI));
+}
+
+static ssize_t mpc85xx_l2_inject_data_lo_show(struct edac_device_ctl_info
+ *edac_dev, char *data)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ return sprintf(data, "0x%08x",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJLO));
+}
+
+static ssize_t mpc85xx_l2_inject_ctrl_show(struct edac_device_ctl_info
+ *edac_dev, char *data)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ return sprintf(data, "0x%08x",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJCTL));
+}
+
+static ssize_t mpc85xx_l2_inject_data_hi_store(struct edac_device_ctl_info
+ *edac_dev, const char *data,
+ size_t count)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ if (isdigit(*data)) {
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJHI,
+ simple_strtoul(data, NULL, 0));
+ return count;
+ }
+ return 0;
+}
+
+static ssize_t mpc85xx_l2_inject_data_lo_store(struct edac_device_ctl_info
+ *edac_dev, const char *data,
+ size_t count)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ if (isdigit(*data)) {
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJLO,
+ simple_strtoul(data, NULL, 0));
+ return count;
+ }
+ return 0;
+}
+
+static ssize_t mpc85xx_l2_inject_ctrl_store(struct edac_device_ctl_info
+ *edac_dev, const char *data,
+ size_t count)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ if (isdigit(*data)) {
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRINJCTL,
+ simple_strtoul(data, NULL, 0));
+ return count;
+ }
+ return 0;
+}
+
+static struct edac_dev_sysfs_attribute mpc85xx_l2_sysfs_attributes[] = {
+ {
+ .attr = {
+ .name = "inject_data_hi",
+ .mode = (S_IRUGO | S_IWUSR)
+ },
+ .show = mpc85xx_l2_inject_data_hi_show,
+ .store = mpc85xx_l2_inject_data_hi_store},
+ {
+ .attr = {
+ .name = "inject_data_lo",
+ .mode = (S_IRUGO | S_IWUSR)
+ },
+ .show = mpc85xx_l2_inject_data_lo_show,
+ .store = mpc85xx_l2_inject_data_lo_store},
+ {
+ .attr = {
+ .name = "inject_ctrl",
+ .mode = (S_IRUGO | S_IWUSR)
+ },
+ .show = mpc85xx_l2_inject_ctrl_show,
+ .store = mpc85xx_l2_inject_ctrl_store},
+
+ /* End of list */
+ {
+ .attr = {.name = NULL}
+ }
+};
+
+static void mpc85xx_set_l2_sysfs_attributes(struct edac_device_ctl_info
+ *edac_dev)
+{
+ edac_dev->sysfs_attributes = mpc85xx_l2_sysfs_attributes;
+}
+
+/***************************** L2 ops ***********************************/
+
+static void mpc85xx_l2_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ u32 err_detect;
+
+ err_detect = in_be32(pdata->l2_vbase + MPC85XX_L2_ERRDET);
+
+ if (!(err_detect & L2_EDE_MASK))
+ return;
+
+ pr_err("ECC Error in CPU L2 cache\n");
+ pr_err("L2 Error Detect Register: 0x%08x\n", err_detect);
+ pr_err("L2 Error Capture Data High Register: 0x%08x\n",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_CAPTDATAHI));
+ pr_err("L2 Error Capture Data Lo Register: 0x%08x\n",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_CAPTDATALO));
+ pr_err("L2 Error Syndrome Register: 0x%08x\n",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_CAPTECC));
+ pr_err("L2 Error Attributes Capture Register: 0x%08x\n",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_ERRATTR));
+ pr_err("L2 Error Address Capture Register: 0x%08x\n",
+ in_be32(pdata->l2_vbase + MPC85XX_L2_ERRADDR));
+
+ /* clear error detect register */
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRDET, err_detect);
+
+ if (err_detect & L2_EDE_CE_MASK)
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+
+ if (err_detect & L2_EDE_UE_MASK)
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static irqreturn_t mpc85xx_l2_isr(int irq, void *dev_id)
+{
+ struct edac_device_ctl_info *edac_dev = dev_id;
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+ u32 err_detect;
+
+ err_detect = in_be32(pdata->l2_vbase + MPC85XX_L2_ERRDET);
+
+ if (!(err_detect & L2_EDE_MASK))
+ return IRQ_NONE;
+
+ mpc85xx_l2_check(edac_dev);
+
+ return IRQ_HANDLED;
+}
+
+static int mpc85xx_l2_err_probe(struct platform_device *op)
+{
+ struct edac_device_ctl_info *edac_dev;
+ struct mpc85xx_l2_pdata *pdata;
+ struct resource r;
+ int res;
+
+ if (!devres_open_group(&op->dev, mpc85xx_l2_err_probe, GFP_KERNEL))
+ return -ENOMEM;
+
+ edac_dev = edac_device_alloc_ctl_info(sizeof(*pdata),
+ "cpu", 1, "L", 1, 2, NULL, 0,
+ edac_dev_idx);
+ if (!edac_dev) {
+ devres_release_group(&op->dev, mpc85xx_l2_err_probe);
+ return -ENOMEM;
+ }
+
+ pdata = edac_dev->pvt_info;
+ pdata->name = "mpc85xx_l2_err";
+ edac_dev->dev = &op->dev;
+ dev_set_drvdata(edac_dev->dev, edac_dev);
+ edac_dev->ctl_name = pdata->name;
+ edac_dev->dev_name = pdata->name;
+
+ res = of_address_to_resource(op->dev.of_node, 0, &r);
+ if (res) {
+ pr_err("%s: Unable to get resource for L2 err regs\n", __func__);
+ goto err;
+ }
+
+ /* we only need the error registers */
+ r.start += 0xe00;
+
+ if (!devm_request_mem_region(&op->dev, r.start, resource_size(&r),
+ pdata->name)) {
+ pr_err("%s: Error while requesting mem region\n", __func__);
+ res = -EBUSY;
+ goto err;
+ }
+
+ pdata->l2_vbase = devm_ioremap(&op->dev, r.start, resource_size(&r));
+ if (!pdata->l2_vbase) {
+ pr_err("%s: Unable to setup L2 err regs\n", __func__);
+ res = -ENOMEM;
+ goto err;
+ }
+
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRDET, ~0);
+
+ orig_l2_err_disable = in_be32(pdata->l2_vbase + MPC85XX_L2_ERRDIS);
+
+ /* clear the err_dis */
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRDIS, 0);
+
+ edac_dev->mod_name = EDAC_MOD_STR;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ edac_dev->edac_check = mpc85xx_l2_check;
+
+ mpc85xx_set_l2_sysfs_attributes(edac_dev);
+
+ pdata->edac_idx = edac_dev_idx++;
+
+ if (edac_device_add_device(edac_dev) > 0) {
+ edac_dbg(3, "failed edac_device_add_device()\n");
+ goto err;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ pdata->irq = irq_of_parse_and_map(op->dev.of_node, 0);
+ res = devm_request_irq(&op->dev, pdata->irq,
+ mpc85xx_l2_isr, IRQF_SHARED,
+ "[EDAC] L2 err", edac_dev);
+ if (res < 0) {
+ pr_err("%s: Unable to request irq %d for MPC85xx L2 err\n",
+ __func__, pdata->irq);
+ irq_dispose_mapping(pdata->irq);
+ res = -ENODEV;
+ goto err2;
+ }
+
+ pr_info(EDAC_MOD_STR " acquired irq %d for L2 Err\n", pdata->irq);
+
+ edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRINTEN, L2_EIE_MASK);
+ }
+
+ devres_remove_group(&op->dev, mpc85xx_l2_err_probe);
+
+ edac_dbg(3, "success\n");
+ pr_info(EDAC_MOD_STR " L2 err registered\n");
+
+ return 0;
+
+err2:
+ edac_device_del_device(&op->dev);
+err:
+ devres_release_group(&op->dev, mpc85xx_l2_err_probe);
+ edac_device_free_ctl_info(edac_dev);
+ return res;
+}
+
+static int mpc85xx_l2_err_remove(struct platform_device *op)
+{
+ struct edac_device_ctl_info *edac_dev = dev_get_drvdata(&op->dev);
+ struct mpc85xx_l2_pdata *pdata = edac_dev->pvt_info;
+
+ edac_dbg(0, "\n");
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRINTEN, 0);
+ irq_dispose_mapping(pdata->irq);
+ }
+
+ out_be32(pdata->l2_vbase + MPC85XX_L2_ERRDIS, orig_l2_err_disable);
+ edac_device_del_device(&op->dev);
+ edac_device_free_ctl_info(edac_dev);
+ return 0;
+}
+
+static const struct of_device_id mpc85xx_l2_err_of_match[] = {
+ { .compatible = "fsl,mpc8536-l2-cache-controller", },
+ { .compatible = "fsl,mpc8540-l2-cache-controller", },
+ { .compatible = "fsl,mpc8541-l2-cache-controller", },
+ { .compatible = "fsl,mpc8544-l2-cache-controller", },
+ { .compatible = "fsl,mpc8548-l2-cache-controller", },
+ { .compatible = "fsl,mpc8555-l2-cache-controller", },
+ { .compatible = "fsl,mpc8560-l2-cache-controller", },
+ { .compatible = "fsl,mpc8568-l2-cache-controller", },
+ { .compatible = "fsl,mpc8569-l2-cache-controller", },
+ { .compatible = "fsl,mpc8572-l2-cache-controller", },
+ { .compatible = "fsl,p1020-l2-cache-controller", },
+ { .compatible = "fsl,p1021-l2-cache-controller", },
+ { .compatible = "fsl,p2020-l2-cache-controller", },
+ { .compatible = "fsl,t2080-l2-cache-controller", },
+ {},
+};
+MODULE_DEVICE_TABLE(of, mpc85xx_l2_err_of_match);
+
+static struct platform_driver mpc85xx_l2_err_driver = {
+ .probe = mpc85xx_l2_err_probe,
+ .remove = mpc85xx_l2_err_remove,
+ .driver = {
+ .name = "mpc85xx_l2_err",
+ .of_match_table = mpc85xx_l2_err_of_match,
+ },
+};
+
+static const struct of_device_id mpc85xx_mc_err_of_match[] = {
+ { .compatible = "fsl,mpc8536-memory-controller", },
+ { .compatible = "fsl,mpc8540-memory-controller", },
+ { .compatible = "fsl,mpc8541-memory-controller", },
+ { .compatible = "fsl,mpc8544-memory-controller", },
+ { .compatible = "fsl,mpc8548-memory-controller", },
+ { .compatible = "fsl,mpc8555-memory-controller", },
+ { .compatible = "fsl,mpc8560-memory-controller", },
+ { .compatible = "fsl,mpc8568-memory-controller", },
+ { .compatible = "fsl,mpc8569-memory-controller", },
+ { .compatible = "fsl,mpc8572-memory-controller", },
+ { .compatible = "fsl,mpc8349-memory-controller", },
+ { .compatible = "fsl,p1020-memory-controller", },
+ { .compatible = "fsl,p1021-memory-controller", },
+ { .compatible = "fsl,p2020-memory-controller", },
+ { .compatible = "fsl,qoriq-memory-controller", },
+ {},
+};
+MODULE_DEVICE_TABLE(of, mpc85xx_mc_err_of_match);
+
+static struct platform_driver mpc85xx_mc_err_driver = {
+ .probe = fsl_mc_err_probe,
+ .remove = fsl_mc_err_remove,
+ .driver = {
+ .name = "mpc85xx_mc_err",
+ .of_match_table = mpc85xx_mc_err_of_match,
+ },
+};
+
+static struct platform_driver * const drivers[] = {
+ &mpc85xx_mc_err_driver,
+ &mpc85xx_l2_err_driver,
+#ifdef CONFIG_PCI
+ &mpc85xx_pci_err_driver,
+#endif
+};
+
+static int __init mpc85xx_mc_init(void)
+{
+ int res = 0;
+ u32 __maybe_unused pvr = 0;
+
+ pr_info("Freescale(R) MPC85xx EDAC driver, (C) 2006 Montavista Software\n");
+
+ /* make sure error reporting method is sane */
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_INT:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_INT;
+ break;
+ }
+
+ res = platform_register_drivers(drivers, ARRAY_SIZE(drivers));
+ if (res)
+ pr_warn(EDAC_MOD_STR "drivers fail to register\n");
+
+ return 0;
+}
+
+module_init(mpc85xx_mc_init);
+
+static void __exit mpc85xx_mc_exit(void)
+{
+ platform_unregister_drivers(drivers, ARRAY_SIZE(drivers));
+}
+
+module_exit(mpc85xx_mc_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Montavista Software, Inc.");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll, 2=Interrupt");
diff --git a/drivers/edac/mpc85xx_edac.h b/drivers/edac/mpc85xx_edac.h
new file mode 100644
index 0000000000..66a046ae33
--- /dev/null
+++ b/drivers/edac/mpc85xx_edac.h
@@ -0,0 +1,105 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Freescale MPC85xx Memory Controller kernel module
+ * Author: Dave Jiang <djiang@mvista.com>
+ *
+ * 2006-2007 (c) MontaVista Software, Inc.
+ */
+#ifndef _MPC85XX_EDAC_H_
+#define _MPC85XX_EDAC_H_
+
+#define MPC85XX_REVISION " Ver: 2.0.0"
+#define EDAC_MOD_STR "MPC85xx_edac"
+
+#define mpc85xx_printk(level, fmt, arg...) \
+ edac_printk(level, "MPC85xx", fmt, ##arg)
+
+/*
+ * L2 Err defines
+ */
+#define MPC85XX_L2_ERRINJHI 0x0000
+#define MPC85XX_L2_ERRINJLO 0x0004
+#define MPC85XX_L2_ERRINJCTL 0x0008
+#define MPC85XX_L2_CAPTDATAHI 0x0020
+#define MPC85XX_L2_CAPTDATALO 0x0024
+#define MPC85XX_L2_CAPTECC 0x0028
+#define MPC85XX_L2_ERRDET 0x0040
+#define MPC85XX_L2_ERRDIS 0x0044
+#define MPC85XX_L2_ERRINTEN 0x0048
+#define MPC85XX_L2_ERRATTR 0x004c
+#define MPC85XX_L2_ERRADDR 0x0050
+#define MPC85XX_L2_ERRCTL 0x0058
+
+/* Error Interrupt Enable */
+#define L2_EIE_L2CFGINTEN 0x1
+#define L2_EIE_SBECCINTEN 0x4
+#define L2_EIE_MBECCINTEN 0x8
+#define L2_EIE_TPARINTEN 0x10
+#define L2_EIE_MASK (L2_EIE_L2CFGINTEN | L2_EIE_SBECCINTEN | \
+ L2_EIE_MBECCINTEN | L2_EIE_TPARINTEN)
+
+/* Error Detect */
+#define L2_EDE_L2CFGERR 0x1
+#define L2_EDE_SBECCERR 0x4
+#define L2_EDE_MBECCERR 0x8
+#define L2_EDE_TPARERR 0x10
+#define L2_EDE_MULL2ERR 0x80000000
+
+#define L2_EDE_CE_MASK L2_EDE_SBECCERR
+#define L2_EDE_UE_MASK (L2_EDE_L2CFGERR | L2_EDE_MBECCERR | \
+ L2_EDE_TPARERR)
+#define L2_EDE_MASK (L2_EDE_L2CFGERR | L2_EDE_SBECCERR | \
+ L2_EDE_MBECCERR | L2_EDE_TPARERR | L2_EDE_MULL2ERR)
+
+/*
+ * PCI Err defines
+ */
+#define PCI_EDE_TOE 0x00000001
+#define PCI_EDE_SCM 0x00000002
+#define PCI_EDE_IRMSV 0x00000004
+#define PCI_EDE_ORMSV 0x00000008
+#define PCI_EDE_OWMSV 0x00000010
+#define PCI_EDE_TGT_ABRT 0x00000020
+#define PCI_EDE_MST_ABRT 0x00000040
+#define PCI_EDE_TGT_PERR 0x00000080
+#define PCI_EDE_MST_PERR 0x00000100
+#define PCI_EDE_RCVD_SERR 0x00000200
+#define PCI_EDE_ADDR_PERR 0x00000400
+#define PCI_EDE_MULTI_ERR 0x80000000
+
+#define PCI_EDE_PERR_MASK (PCI_EDE_TGT_PERR | PCI_EDE_MST_PERR | \
+ PCI_EDE_ADDR_PERR)
+
+#define MPC85XX_PCI_ERR_DR 0x0000
+#define MPC85XX_PCI_ERR_CAP_DR 0x0004
+#define MPC85XX_PCI_ERR_EN 0x0008
+#define PEX_ERR_ICCAIE_EN_BIT 0x00020000
+#define MPC85XX_PCI_ERR_ATTRIB 0x000c
+#define MPC85XX_PCI_ERR_ADDR 0x0010
+#define PEX_ERR_ICCAD_DISR_BIT 0x00020000
+#define MPC85XX_PCI_ERR_EXT_ADDR 0x0014
+#define MPC85XX_PCI_ERR_DL 0x0018
+#define MPC85XX_PCI_ERR_DH 0x001c
+#define MPC85XX_PCI_GAS_TIMR 0x0020
+#define MPC85XX_PCI_PCIX_TIMR 0x0024
+#define MPC85XX_PCIE_ERR_CAP_R0 0x0028
+#define MPC85XX_PCIE_ERR_CAP_R1 0x002c
+#define MPC85XX_PCIE_ERR_CAP_R2 0x0030
+#define MPC85XX_PCIE_ERR_CAP_R3 0x0034
+
+struct mpc85xx_l2_pdata {
+ char *name;
+ int edac_idx;
+ void __iomem *l2_vbase;
+ int irq;
+};
+
+struct mpc85xx_pci_pdata {
+ char *name;
+ bool is_pcie;
+ int edac_idx;
+ void __iomem *pci_vbase;
+ int irq;
+};
+
+#endif
diff --git a/drivers/edac/npcm_edac.c b/drivers/edac/npcm_edac.c
new file mode 100644
index 0000000000..6d15c15502
--- /dev/null
+++ b/drivers/edac/npcm_edac.c
@@ -0,0 +1,544 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright (c) 2022 Nuvoton Technology Corporation
+
+#include <linux/debugfs.h>
+#include <linux/iopoll.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_NAME "npcm-edac"
+#define EDAC_MSG_SIZE 256
+
+/* chip serials */
+#define NPCM7XX_CHIP BIT(0)
+#define NPCM8XX_CHIP BIT(1)
+
+/* syndrome values */
+#define UE_SYNDROME 0x03
+
+/* error injection */
+#define ERROR_TYPE_CORRECTABLE 0
+#define ERROR_TYPE_UNCORRECTABLE 1
+#define ERROR_LOCATION_DATA 0
+#define ERROR_LOCATION_CHECKCODE 1
+#define ERROR_BIT_DATA_MAX 63
+#define ERROR_BIT_CHECKCODE_MAX 7
+
+static char data_synd[] = {
+ 0xf4, 0xf1, 0xec, 0xea, 0xe9, 0xe6, 0xe5, 0xe3,
+ 0xdc, 0xda, 0xd9, 0xd6, 0xd5, 0xd3, 0xce, 0xcb,
+ 0xb5, 0xb0, 0xad, 0xab, 0xa8, 0xa7, 0xa4, 0xa2,
+ 0x9d, 0x9b, 0x98, 0x97, 0x94, 0x92, 0x8f, 0x8a,
+ 0x75, 0x70, 0x6d, 0x6b, 0x68, 0x67, 0x64, 0x62,
+ 0x5e, 0x5b, 0x58, 0x57, 0x54, 0x52, 0x4f, 0x4a,
+ 0x34, 0x31, 0x2c, 0x2a, 0x29, 0x26, 0x25, 0x23,
+ 0x1c, 0x1a, 0x19, 0x16, 0x15, 0x13, 0x0e, 0x0b
+};
+
+static struct regmap *npcm_regmap;
+
+struct npcm_platform_data {
+ /* chip serials */
+ int chip;
+
+ /* memory controller registers */
+ u32 ctl_ecc_en;
+ u32 ctl_int_status;
+ u32 ctl_int_ack;
+ u32 ctl_int_mask_master;
+ u32 ctl_int_mask_ecc;
+ u32 ctl_ce_addr_l;
+ u32 ctl_ce_addr_h;
+ u32 ctl_ce_data_l;
+ u32 ctl_ce_data_h;
+ u32 ctl_ce_synd;
+ u32 ctl_ue_addr_l;
+ u32 ctl_ue_addr_h;
+ u32 ctl_ue_data_l;
+ u32 ctl_ue_data_h;
+ u32 ctl_ue_synd;
+ u32 ctl_source_id;
+ u32 ctl_controller_busy;
+ u32 ctl_xor_check_bits;
+
+ /* masks and shifts */
+ u32 ecc_en_mask;
+ u32 int_status_ce_mask;
+ u32 int_status_ue_mask;
+ u32 int_ack_ce_mask;
+ u32 int_ack_ue_mask;
+ u32 int_mask_master_non_ecc_mask;
+ u32 int_mask_master_global_mask;
+ u32 int_mask_ecc_non_event_mask;
+ u32 ce_addr_h_mask;
+ u32 ce_synd_mask;
+ u32 ce_synd_shift;
+ u32 ue_addr_h_mask;
+ u32 ue_synd_mask;
+ u32 ue_synd_shift;
+ u32 source_id_ce_mask;
+ u32 source_id_ce_shift;
+ u32 source_id_ue_mask;
+ u32 source_id_ue_shift;
+ u32 controller_busy_mask;
+ u32 xor_check_bits_mask;
+ u32 xor_check_bits_shift;
+ u32 writeback_en_mask;
+ u32 fwc_mask;
+};
+
+struct priv_data {
+ void __iomem *reg;
+ char message[EDAC_MSG_SIZE];
+ const struct npcm_platform_data *pdata;
+
+ /* error injection */
+ struct dentry *debugfs;
+ u8 error_type;
+ u8 location;
+ u8 bit;
+};
+
+static void handle_ce(struct mem_ctl_info *mci)
+{
+ struct priv_data *priv = mci->pvt_info;
+ const struct npcm_platform_data *pdata;
+ u32 val_h = 0, val_l, id, synd;
+ u64 addr = 0, data = 0;
+
+ pdata = priv->pdata;
+ regmap_read(npcm_regmap, pdata->ctl_ce_addr_l, &val_l);
+ if (pdata->chip == NPCM8XX_CHIP) {
+ regmap_read(npcm_regmap, pdata->ctl_ce_addr_h, &val_h);
+ val_h &= pdata->ce_addr_h_mask;
+ }
+ addr = ((addr | val_h) << 32) | val_l;
+
+ regmap_read(npcm_regmap, pdata->ctl_ce_data_l, &val_l);
+ if (pdata->chip == NPCM8XX_CHIP)
+ regmap_read(npcm_regmap, pdata->ctl_ce_data_h, &val_h);
+ data = ((data | val_h) << 32) | val_l;
+
+ regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
+ id = (id & pdata->source_id_ce_mask) >> pdata->source_id_ce_shift;
+
+ regmap_read(npcm_regmap, pdata->ctl_ce_synd, &synd);
+ synd = (synd & pdata->ce_synd_mask) >> pdata->ce_synd_shift;
+
+ snprintf(priv->message, EDAC_MSG_SIZE,
+ "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, addr >> PAGE_SHIFT,
+ addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
+}
+
+static void handle_ue(struct mem_ctl_info *mci)
+{
+ struct priv_data *priv = mci->pvt_info;
+ const struct npcm_platform_data *pdata;
+ u32 val_h = 0, val_l, id, synd;
+ u64 addr = 0, data = 0;
+
+ pdata = priv->pdata;
+ regmap_read(npcm_regmap, pdata->ctl_ue_addr_l, &val_l);
+ if (pdata->chip == NPCM8XX_CHIP) {
+ regmap_read(npcm_regmap, pdata->ctl_ue_addr_h, &val_h);
+ val_h &= pdata->ue_addr_h_mask;
+ }
+ addr = ((addr | val_h) << 32) | val_l;
+
+ regmap_read(npcm_regmap, pdata->ctl_ue_data_l, &val_l);
+ if (pdata->chip == NPCM8XX_CHIP)
+ regmap_read(npcm_regmap, pdata->ctl_ue_data_h, &val_h);
+ data = ((data | val_h) << 32) | val_l;
+
+ regmap_read(npcm_regmap, pdata->ctl_source_id, &id);
+ id = (id & pdata->source_id_ue_mask) >> pdata->source_id_ue_shift;
+
+ regmap_read(npcm_regmap, pdata->ctl_ue_synd, &synd);
+ synd = (synd & pdata->ue_synd_mask) >> pdata->ue_synd_shift;
+
+ snprintf(priv->message, EDAC_MSG_SIZE,
+ "addr = 0x%llx, data = 0x%llx, id = 0x%x", addr, data, id);
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, addr >> PAGE_SHIFT,
+ addr & ~PAGE_MASK, synd, 0, 0, -1, priv->message, "");
+}
+
+static irqreturn_t edac_ecc_isr(int irq, void *dev_id)
+{
+ const struct npcm_platform_data *pdata;
+ struct mem_ctl_info *mci = dev_id;
+ u32 status;
+
+ pdata = ((struct priv_data *)mci->pvt_info)->pdata;
+ regmap_read(npcm_regmap, pdata->ctl_int_status, &status);
+ if (status & pdata->int_status_ce_mask) {
+ handle_ce(mci);
+
+ /* acknowledge the CE interrupt */
+ regmap_write(npcm_regmap, pdata->ctl_int_ack,
+ pdata->int_ack_ce_mask);
+ return IRQ_HANDLED;
+ } else if (status & pdata->int_status_ue_mask) {
+ handle_ue(mci);
+
+ /* acknowledge the UE interrupt */
+ regmap_write(npcm_regmap, pdata->ctl_int_ack,
+ pdata->int_ack_ue_mask);
+ return IRQ_HANDLED;
+ }
+
+ WARN_ON_ONCE(1);
+ return IRQ_NONE;
+}
+
+static ssize_t force_ecc_error(struct file *file, const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct device *dev = file->private_data;
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct priv_data *priv = mci->pvt_info;
+ const struct npcm_platform_data *pdata;
+ u32 val, syndrome;
+ int ret;
+
+ pdata = priv->pdata;
+ edac_printk(KERN_INFO, EDAC_MOD_NAME,
+ "force an ECC error, type = %d, location = %d, bit = %d\n",
+ priv->error_type, priv->location, priv->bit);
+
+ /* ensure no pending writes */
+ ret = regmap_read_poll_timeout(npcm_regmap, pdata->ctl_controller_busy,
+ val, !(val & pdata->controller_busy_mask),
+ 1000, 10000);
+ if (ret) {
+ edac_printk(KERN_INFO, EDAC_MOD_NAME,
+ "wait pending writes timeout\n");
+ return count;
+ }
+
+ regmap_read(npcm_regmap, pdata->ctl_xor_check_bits, &val);
+ val &= ~pdata->xor_check_bits_mask;
+
+ /* write syndrome to XOR_CHECK_BITS */
+ if (priv->error_type == ERROR_TYPE_CORRECTABLE) {
+ if (priv->location == ERROR_LOCATION_DATA &&
+ priv->bit > ERROR_BIT_DATA_MAX) {
+ edac_printk(KERN_INFO, EDAC_MOD_NAME,
+ "data bit should not exceed %d (%d)\n",
+ ERROR_BIT_DATA_MAX, priv->bit);
+ return count;
+ }
+
+ if (priv->location == ERROR_LOCATION_CHECKCODE &&
+ priv->bit > ERROR_BIT_CHECKCODE_MAX) {
+ edac_printk(KERN_INFO, EDAC_MOD_NAME,
+ "checkcode bit should not exceed %d (%d)\n",
+ ERROR_BIT_CHECKCODE_MAX, priv->bit);
+ return count;
+ }
+
+ syndrome = priv->location ? 1 << priv->bit
+ : data_synd[priv->bit];
+
+ regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
+ val | (syndrome << pdata->xor_check_bits_shift) |
+ pdata->writeback_en_mask);
+ } else if (priv->error_type == ERROR_TYPE_UNCORRECTABLE) {
+ regmap_write(npcm_regmap, pdata->ctl_xor_check_bits,
+ val | (UE_SYNDROME << pdata->xor_check_bits_shift));
+ }
+
+ /* force write check */
+ regmap_update_bits(npcm_regmap, pdata->ctl_xor_check_bits,
+ pdata->fwc_mask, pdata->fwc_mask);
+
+ return count;
+}
+
+static const struct file_operations force_ecc_error_fops = {
+ .open = simple_open,
+ .write = force_ecc_error,
+ .llseek = generic_file_llseek,
+};
+
+/*
+ * Setup debugfs for error injection.
+ *
+ * Nodes:
+ * error_type - 0: CE, 1: UE
+ * location - 0: data, 1: checkcode
+ * bit - 0 ~ 63 for data and 0 ~ 7 for checkcode
+ * force_ecc_error - trigger
+ *
+ * Examples:
+ * 1. Inject a correctable error (CE) at checkcode bit 7.
+ * ~# echo 0 > /sys/kernel/debug/edac/npcm-edac/error_type
+ * ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/location
+ * ~# echo 7 > /sys/kernel/debug/edac/npcm-edac/bit
+ * ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
+ *
+ * 2. Inject an uncorrectable error (UE).
+ * ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/error_type
+ * ~# echo 1 > /sys/kernel/debug/edac/npcm-edac/force_ecc_error
+ */
+static void setup_debugfs(struct mem_ctl_info *mci)
+{
+ struct priv_data *priv = mci->pvt_info;
+
+ priv->debugfs = edac_debugfs_create_dir(mci->mod_name);
+ if (!priv->debugfs)
+ return;
+
+ edac_debugfs_create_x8("error_type", 0644, priv->debugfs, &priv->error_type);
+ edac_debugfs_create_x8("location", 0644, priv->debugfs, &priv->location);
+ edac_debugfs_create_x8("bit", 0644, priv->debugfs, &priv->bit);
+ edac_debugfs_create_file("force_ecc_error", 0200, priv->debugfs,
+ &mci->dev, &force_ecc_error_fops);
+}
+
+static int setup_irq(struct mem_ctl_info *mci, struct platform_device *pdev)
+{
+ const struct npcm_platform_data *pdata;
+ int ret, irq;
+
+ pdata = ((struct priv_data *)mci->pvt_info)->pdata;
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME, "IRQ not defined in DTS\n");
+ return irq;
+ }
+
+ ret = devm_request_irq(&pdev->dev, irq, edac_ecc_isr, 0,
+ dev_name(&pdev->dev), mci);
+ if (ret < 0) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME, "failed to request IRQ\n");
+ return ret;
+ }
+
+ /* enable the functional group of ECC and mask the others */
+ regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
+ pdata->int_mask_master_non_ecc_mask);
+
+ if (pdata->chip == NPCM8XX_CHIP)
+ regmap_write(npcm_regmap, pdata->ctl_int_mask_ecc,
+ pdata->int_mask_ecc_non_event_mask);
+
+ return 0;
+}
+
+static const struct regmap_config npcm_regmap_cfg = {
+ .reg_bits = 32,
+ .reg_stride = 4,
+ .val_bits = 32,
+};
+
+static int edac_probe(struct platform_device *pdev)
+{
+ const struct npcm_platform_data *pdata;
+ struct device *dev = &pdev->dev;
+ struct edac_mc_layer layers[1];
+ struct mem_ctl_info *mci;
+ struct priv_data *priv;
+ void __iomem *reg;
+ u32 val;
+ int rc;
+
+ reg = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(reg))
+ return PTR_ERR(reg);
+
+ npcm_regmap = devm_regmap_init_mmio(dev, reg, &npcm_regmap_cfg);
+ if (IS_ERR(npcm_regmap))
+ return PTR_ERR(npcm_regmap);
+
+ pdata = of_device_get_match_data(dev);
+ if (!pdata)
+ return -EINVAL;
+
+ /* bail out if ECC is not enabled */
+ regmap_read(npcm_regmap, pdata->ctl_ecc_en, &val);
+ if (!(val & pdata->ecc_en_mask)) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME, "ECC is not enabled\n");
+ return -EPERM;
+ }
+
+ edac_op_state = EDAC_OPSTATE_INT;
+
+ layers[0].type = EDAC_MC_LAYER_ALL_MEM;
+ layers[0].size = 1;
+
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct priv_data));
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ priv = mci->pvt_info;
+ priv->reg = reg;
+ priv->pdata = pdata;
+ platform_set_drvdata(pdev, mci);
+
+ mci->mtype_cap = MEM_FLAG_DDR4;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ mci->scrub_mode = SCRUB_HW_SRC;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->ctl_name = "npcm_ddr_controller";
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->mod_name = EDAC_MOD_NAME;
+ mci->ctl_page_to_phys = NULL;
+
+ rc = setup_irq(mci, pdev);
+ if (rc)
+ goto free_edac_mc;
+
+ rc = edac_mc_add_mc(mci);
+ if (rc)
+ goto free_edac_mc;
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
+ setup_debugfs(mci);
+
+ return rc;
+
+free_edac_mc:
+ edac_mc_free(mci);
+ return rc;
+}
+
+static int edac_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+ struct priv_data *priv = mci->pvt_info;
+ const struct npcm_platform_data *pdata;
+
+ pdata = priv->pdata;
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG) && pdata->chip == NPCM8XX_CHIP)
+ edac_debugfs_remove_recursive(priv->debugfs);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ regmap_write(npcm_regmap, pdata->ctl_int_mask_master,
+ pdata->int_mask_master_global_mask);
+ regmap_update_bits(npcm_regmap, pdata->ctl_ecc_en, pdata->ecc_en_mask, 0);
+
+ return 0;
+}
+
+static const struct npcm_platform_data npcm750_edac = {
+ .chip = NPCM7XX_CHIP,
+
+ /* memory controller registers */
+ .ctl_ecc_en = 0x174,
+ .ctl_int_status = 0x1d0,
+ .ctl_int_ack = 0x1d4,
+ .ctl_int_mask_master = 0x1d8,
+ .ctl_ce_addr_l = 0x188,
+ .ctl_ce_data_l = 0x190,
+ .ctl_ce_synd = 0x18c,
+ .ctl_ue_addr_l = 0x17c,
+ .ctl_ue_data_l = 0x184,
+ .ctl_ue_synd = 0x180,
+ .ctl_source_id = 0x194,
+
+ /* masks and shifts */
+ .ecc_en_mask = BIT(24),
+ .int_status_ce_mask = GENMASK(4, 3),
+ .int_status_ue_mask = GENMASK(6, 5),
+ .int_ack_ce_mask = GENMASK(4, 3),
+ .int_ack_ue_mask = GENMASK(6, 5),
+ .int_mask_master_non_ecc_mask = GENMASK(30, 7) | GENMASK(2, 0),
+ .int_mask_master_global_mask = BIT(31),
+ .ce_synd_mask = GENMASK(6, 0),
+ .ce_synd_shift = 0,
+ .ue_synd_mask = GENMASK(6, 0),
+ .ue_synd_shift = 0,
+ .source_id_ce_mask = GENMASK(29, 16),
+ .source_id_ce_shift = 16,
+ .source_id_ue_mask = GENMASK(13, 0),
+ .source_id_ue_shift = 0,
+};
+
+static const struct npcm_platform_data npcm845_edac = {
+ .chip = NPCM8XX_CHIP,
+
+ /* memory controller registers */
+ .ctl_ecc_en = 0x16c,
+ .ctl_int_status = 0x228,
+ .ctl_int_ack = 0x244,
+ .ctl_int_mask_master = 0x220,
+ .ctl_int_mask_ecc = 0x260,
+ .ctl_ce_addr_l = 0x18c,
+ .ctl_ce_addr_h = 0x190,
+ .ctl_ce_data_l = 0x194,
+ .ctl_ce_data_h = 0x198,
+ .ctl_ce_synd = 0x190,
+ .ctl_ue_addr_l = 0x17c,
+ .ctl_ue_addr_h = 0x180,
+ .ctl_ue_data_l = 0x184,
+ .ctl_ue_data_h = 0x188,
+ .ctl_ue_synd = 0x180,
+ .ctl_source_id = 0x19c,
+ .ctl_controller_busy = 0x20c,
+ .ctl_xor_check_bits = 0x174,
+
+ /* masks and shifts */
+ .ecc_en_mask = GENMASK(17, 16),
+ .int_status_ce_mask = GENMASK(1, 0),
+ .int_status_ue_mask = GENMASK(3, 2),
+ .int_ack_ce_mask = GENMASK(1, 0),
+ .int_ack_ue_mask = GENMASK(3, 2),
+ .int_mask_master_non_ecc_mask = GENMASK(30, 3) | GENMASK(1, 0),
+ .int_mask_master_global_mask = BIT(31),
+ .int_mask_ecc_non_event_mask = GENMASK(8, 4),
+ .ce_addr_h_mask = GENMASK(1, 0),
+ .ce_synd_mask = GENMASK(15, 8),
+ .ce_synd_shift = 8,
+ .ue_addr_h_mask = GENMASK(1, 0),
+ .ue_synd_mask = GENMASK(15, 8),
+ .ue_synd_shift = 8,
+ .source_id_ce_mask = GENMASK(29, 16),
+ .source_id_ce_shift = 16,
+ .source_id_ue_mask = GENMASK(13, 0),
+ .source_id_ue_shift = 0,
+ .controller_busy_mask = BIT(0),
+ .xor_check_bits_mask = GENMASK(23, 16),
+ .xor_check_bits_shift = 16,
+ .writeback_en_mask = BIT(24),
+ .fwc_mask = BIT(8),
+};
+
+static const struct of_device_id npcm_edac_of_match[] = {
+ {
+ .compatible = "nuvoton,npcm750-memory-controller",
+ .data = &npcm750_edac
+ },
+ {
+ .compatible = "nuvoton,npcm845-memory-controller",
+ .data = &npcm845_edac
+ },
+ {},
+};
+
+MODULE_DEVICE_TABLE(of, npcm_edac_of_match);
+
+static struct platform_driver npcm_edac_driver = {
+ .driver = {
+ .name = "npcm-edac",
+ .of_match_table = npcm_edac_of_match,
+ },
+ .probe = edac_probe,
+ .remove = edac_remove,
+};
+
+module_platform_driver(npcm_edac_driver);
+
+MODULE_AUTHOR("Medad CChien <medadyoung@gmail.com>");
+MODULE_AUTHOR("Marvin Lin <kflin@nuvoton.com>");
+MODULE_DESCRIPTION("Nuvoton NPCM EDAC Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/edac/octeon_edac-l2c.c b/drivers/edac/octeon_edac-l2c.c
new file mode 100644
index 0000000000..c33059e9b0
--- /dev/null
+++ b/drivers/edac/octeon_edac-l2c.c
@@ -0,0 +1,207 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2012 Cavium, Inc.
+ *
+ * Copyright (C) 2009 Wind River Systems,
+ * written by Ralf Baechle <ralf@linux-mips.org>
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+#include <linux/edac.h>
+
+#include <asm/octeon/cvmx.h>
+
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "octeon-l2c"
+
+static void octeon_l2c_poll_oct1(struct edac_device_ctl_info *l2c)
+{
+ union cvmx_l2t_err l2t_err, l2t_err_reset;
+ union cvmx_l2d_err l2d_err, l2d_err_reset;
+
+ l2t_err_reset.u64 = 0;
+ l2t_err.u64 = cvmx_read_csr(CVMX_L2T_ERR);
+ if (l2t_err.s.sec_err) {
+ edac_device_handle_ce(l2c, 0, 0,
+ "Tag Single bit error (corrected)");
+ l2t_err_reset.s.sec_err = 1;
+ }
+ if (l2t_err.s.ded_err) {
+ edac_device_handle_ue(l2c, 0, 0,
+ "Tag Double bit error (detected)");
+ l2t_err_reset.s.ded_err = 1;
+ }
+ if (l2t_err_reset.u64)
+ cvmx_write_csr(CVMX_L2T_ERR, l2t_err_reset.u64);
+
+ l2d_err_reset.u64 = 0;
+ l2d_err.u64 = cvmx_read_csr(CVMX_L2D_ERR);
+ if (l2d_err.s.sec_err) {
+ edac_device_handle_ce(l2c, 0, 1,
+ "Data Single bit error (corrected)");
+ l2d_err_reset.s.sec_err = 1;
+ }
+ if (l2d_err.s.ded_err) {
+ edac_device_handle_ue(l2c, 0, 1,
+ "Data Double bit error (detected)");
+ l2d_err_reset.s.ded_err = 1;
+ }
+ if (l2d_err_reset.u64)
+ cvmx_write_csr(CVMX_L2D_ERR, l2d_err_reset.u64);
+
+}
+
+static void _octeon_l2c_poll_oct2(struct edac_device_ctl_info *l2c, int tad)
+{
+ union cvmx_l2c_err_tdtx err_tdtx, err_tdtx_reset;
+ union cvmx_l2c_err_ttgx err_ttgx, err_ttgx_reset;
+ char buf1[64];
+ char buf2[80];
+
+ err_tdtx_reset.u64 = 0;
+ err_tdtx.u64 = cvmx_read_csr(CVMX_L2C_ERR_TDTX(tad));
+ if (err_tdtx.s.dbe || err_tdtx.s.sbe ||
+ err_tdtx.s.vdbe || err_tdtx.s.vsbe)
+ snprintf(buf1, sizeof(buf1),
+ "type:%d, syn:0x%x, way:%d",
+ err_tdtx.s.type, err_tdtx.s.syn, err_tdtx.s.wayidx);
+
+ if (err_tdtx.s.dbe) {
+ snprintf(buf2, sizeof(buf2),
+ "L2D Double bit error (detected):%s", buf1);
+ err_tdtx_reset.s.dbe = 1;
+ edac_device_handle_ue(l2c, tad, 1, buf2);
+ }
+ if (err_tdtx.s.sbe) {
+ snprintf(buf2, sizeof(buf2),
+ "L2D Single bit error (corrected):%s", buf1);
+ err_tdtx_reset.s.sbe = 1;
+ edac_device_handle_ce(l2c, tad, 1, buf2);
+ }
+ if (err_tdtx.s.vdbe) {
+ snprintf(buf2, sizeof(buf2),
+ "VBF Double bit error (detected):%s", buf1);
+ err_tdtx_reset.s.vdbe = 1;
+ edac_device_handle_ue(l2c, tad, 1, buf2);
+ }
+ if (err_tdtx.s.vsbe) {
+ snprintf(buf2, sizeof(buf2),
+ "VBF Single bit error (corrected):%s", buf1);
+ err_tdtx_reset.s.vsbe = 1;
+ edac_device_handle_ce(l2c, tad, 1, buf2);
+ }
+ if (err_tdtx_reset.u64)
+ cvmx_write_csr(CVMX_L2C_ERR_TDTX(tad), err_tdtx_reset.u64);
+
+ err_ttgx_reset.u64 = 0;
+ err_ttgx.u64 = cvmx_read_csr(CVMX_L2C_ERR_TTGX(tad));
+
+ if (err_ttgx.s.dbe || err_ttgx.s.sbe)
+ snprintf(buf1, sizeof(buf1),
+ "type:%d, syn:0x%x, way:%d",
+ err_ttgx.s.type, err_ttgx.s.syn, err_ttgx.s.wayidx);
+
+ if (err_ttgx.s.dbe) {
+ snprintf(buf2, sizeof(buf2),
+ "Tag Double bit error (detected):%s", buf1);
+ err_ttgx_reset.s.dbe = 1;
+ edac_device_handle_ue(l2c, tad, 0, buf2);
+ }
+ if (err_ttgx.s.sbe) {
+ snprintf(buf2, sizeof(buf2),
+ "Tag Single bit error (corrected):%s", buf1);
+ err_ttgx_reset.s.sbe = 1;
+ edac_device_handle_ce(l2c, tad, 0, buf2);
+ }
+ if (err_ttgx_reset.u64)
+ cvmx_write_csr(CVMX_L2C_ERR_TTGX(tad), err_ttgx_reset.u64);
+}
+
+static void octeon_l2c_poll_oct2(struct edac_device_ctl_info *l2c)
+{
+ int i;
+ for (i = 0; i < l2c->nr_instances; i++)
+ _octeon_l2c_poll_oct2(l2c, i);
+}
+
+static int octeon_l2c_probe(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *l2c;
+
+ int num_tads = OCTEON_IS_MODEL(OCTEON_CN68XX) ? 4 : 1;
+
+ /* 'Tags' are block 0, 'Data' is block 1*/
+ l2c = edac_device_alloc_ctl_info(0, "l2c", num_tads, "l2c", 2, 0,
+ NULL, 0, edac_device_alloc_index());
+ if (!l2c)
+ return -ENOMEM;
+
+ l2c->dev = &pdev->dev;
+ platform_set_drvdata(pdev, l2c);
+ l2c->dev_name = dev_name(&pdev->dev);
+
+ l2c->mod_name = "octeon-l2c";
+ l2c->ctl_name = "octeon_l2c_err";
+
+
+ if (OCTEON_IS_OCTEON1PLUS()) {
+ union cvmx_l2t_err l2t_err;
+ union cvmx_l2d_err l2d_err;
+
+ l2t_err.u64 = cvmx_read_csr(CVMX_L2T_ERR);
+ l2t_err.s.sec_intena = 0; /* We poll */
+ l2t_err.s.ded_intena = 0;
+ cvmx_write_csr(CVMX_L2T_ERR, l2t_err.u64);
+
+ l2d_err.u64 = cvmx_read_csr(CVMX_L2D_ERR);
+ l2d_err.s.sec_intena = 0; /* We poll */
+ l2d_err.s.ded_intena = 0;
+ cvmx_write_csr(CVMX_L2T_ERR, l2d_err.u64);
+
+ l2c->edac_check = octeon_l2c_poll_oct1;
+ } else {
+ /* OCTEON II */
+ l2c->edac_check = octeon_l2c_poll_oct2;
+ }
+
+ if (edac_device_add_device(l2c) > 0) {
+ pr_err("%s: edac_device_add_device() failed\n", __func__);
+ goto err;
+ }
+
+
+ return 0;
+
+err:
+ edac_device_free_ctl_info(l2c);
+
+ return -ENXIO;
+}
+
+static int octeon_l2c_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *l2c = platform_get_drvdata(pdev);
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(l2c);
+
+ return 0;
+}
+
+static struct platform_driver octeon_l2c_driver = {
+ .probe = octeon_l2c_probe,
+ .remove = octeon_l2c_remove,
+ .driver = {
+ .name = "octeon_l2c_edac",
+ }
+};
+module_platform_driver(octeon_l2c_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>");
diff --git a/drivers/edac/octeon_edac-lmc.c b/drivers/edac/octeon_edac-lmc.c
new file mode 100644
index 0000000000..aeb222ca3e
--- /dev/null
+++ b/drivers/edac/octeon_edac-lmc.c
@@ -0,0 +1,324 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2009 Wind River Systems,
+ * written by Ralf Baechle <ralf@linux-mips.org>
+ *
+ * Copyright (c) 2013 by Cisco Systems, Inc.
+ * All rights reserved.
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+#include <linux/edac.h>
+#include <linux/ctype.h>
+
+#include <asm/octeon/octeon.h>
+#include <asm/octeon/cvmx-lmcx-defs.h>
+
+#include "edac_module.h"
+
+#define OCTEON_MAX_MC 4
+
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+struct octeon_lmc_pvt {
+ unsigned long inject;
+ unsigned long error_type;
+ unsigned long dimm;
+ unsigned long rank;
+ unsigned long bank;
+ unsigned long row;
+ unsigned long col;
+};
+
+static void octeon_lmc_edac_poll(struct mem_ctl_info *mci)
+{
+ union cvmx_lmcx_mem_cfg0 cfg0;
+ bool do_clear = false;
+ char msg[64];
+
+ cfg0.u64 = cvmx_read_csr(CVMX_LMCX_MEM_CFG0(mci->mc_idx));
+ if (cfg0.s.sec_err || cfg0.s.ded_err) {
+ union cvmx_lmcx_fadr fadr;
+ fadr.u64 = cvmx_read_csr(CVMX_LMCX_FADR(mci->mc_idx));
+ snprintf(msg, sizeof(msg),
+ "DIMM %d rank %d bank %d row %d col %d",
+ fadr.cn30xx.fdimm, fadr.cn30xx.fbunk,
+ fadr.cn30xx.fbank, fadr.cn30xx.frow, fadr.cn30xx.fcol);
+ }
+
+ if (cfg0.s.sec_err) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, msg, "");
+ cfg0.s.sec_err = -1; /* Done, re-arm */
+ do_clear = true;
+ }
+
+ if (cfg0.s.ded_err) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, msg, "");
+ cfg0.s.ded_err = -1; /* Done, re-arm */
+ do_clear = true;
+ }
+ if (do_clear)
+ cvmx_write_csr(CVMX_LMCX_MEM_CFG0(mci->mc_idx), cfg0.u64);
+}
+
+static void octeon_lmc_edac_poll_o2(struct mem_ctl_info *mci)
+{
+ struct octeon_lmc_pvt *pvt = mci->pvt_info;
+ union cvmx_lmcx_int int_reg;
+ bool do_clear = false;
+ char msg[64];
+
+ if (!pvt->inject)
+ int_reg.u64 = cvmx_read_csr(CVMX_LMCX_INT(mci->mc_idx));
+ else {
+ int_reg.u64 = 0;
+ if (pvt->error_type == 1)
+ int_reg.s.sec_err = 1;
+ if (pvt->error_type == 2)
+ int_reg.s.ded_err = 1;
+ }
+
+ if (int_reg.s.sec_err || int_reg.s.ded_err) {
+ union cvmx_lmcx_fadr fadr;
+ if (likely(!pvt->inject))
+ fadr.u64 = cvmx_read_csr(CVMX_LMCX_FADR(mci->mc_idx));
+ else {
+ fadr.cn61xx.fdimm = pvt->dimm;
+ fadr.cn61xx.fbunk = pvt->rank;
+ fadr.cn61xx.fbank = pvt->bank;
+ fadr.cn61xx.frow = pvt->row;
+ fadr.cn61xx.fcol = pvt->col;
+ }
+ snprintf(msg, sizeof(msg),
+ "DIMM %d rank %d bank %d row %d col %d",
+ fadr.cn61xx.fdimm, fadr.cn61xx.fbunk,
+ fadr.cn61xx.fbank, fadr.cn61xx.frow, fadr.cn61xx.fcol);
+ }
+
+ if (int_reg.s.sec_err) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, msg, "");
+ int_reg.s.sec_err = -1; /* Done, re-arm */
+ do_clear = true;
+ }
+
+ if (int_reg.s.ded_err) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1, msg, "");
+ int_reg.s.ded_err = -1; /* Done, re-arm */
+ do_clear = true;
+ }
+
+ if (do_clear) {
+ if (likely(!pvt->inject))
+ cvmx_write_csr(CVMX_LMCX_INT(mci->mc_idx), int_reg.u64);
+ else
+ pvt->inject = 0;
+ }
+}
+
+/************************ MC SYSFS parts ***********************************/
+
+/* Only a couple naming differences per template, so very similar */
+#define TEMPLATE_SHOW(reg) \
+static ssize_t octeon_mc_inject_##reg##_show(struct device *dev, \
+ struct device_attribute *attr, \
+ char *data) \
+{ \
+ struct mem_ctl_info *mci = to_mci(dev); \
+ struct octeon_lmc_pvt *pvt = mci->pvt_info; \
+ return sprintf(data, "%016llu\n", (u64)pvt->reg); \
+}
+
+#define TEMPLATE_STORE(reg) \
+static ssize_t octeon_mc_inject_##reg##_store(struct device *dev, \
+ struct device_attribute *attr, \
+ const char *data, size_t count) \
+{ \
+ struct mem_ctl_info *mci = to_mci(dev); \
+ struct octeon_lmc_pvt *pvt = mci->pvt_info; \
+ if (isdigit(*data)) { \
+ if (!kstrtoul(data, 0, &pvt->reg)) \
+ return count; \
+ } \
+ return 0; \
+}
+
+TEMPLATE_SHOW(inject);
+TEMPLATE_STORE(inject);
+TEMPLATE_SHOW(dimm);
+TEMPLATE_STORE(dimm);
+TEMPLATE_SHOW(bank);
+TEMPLATE_STORE(bank);
+TEMPLATE_SHOW(rank);
+TEMPLATE_STORE(rank);
+TEMPLATE_SHOW(row);
+TEMPLATE_STORE(row);
+TEMPLATE_SHOW(col);
+TEMPLATE_STORE(col);
+
+static ssize_t octeon_mc_inject_error_type_store(struct device *dev,
+ struct device_attribute *attr,
+ const char *data,
+ size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct octeon_lmc_pvt *pvt = mci->pvt_info;
+
+ if (!strncmp(data, "single", 6))
+ pvt->error_type = 1;
+ else if (!strncmp(data, "double", 6))
+ pvt->error_type = 2;
+
+ return count;
+}
+
+static ssize_t octeon_mc_inject_error_type_show(struct device *dev,
+ struct device_attribute *attr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct octeon_lmc_pvt *pvt = mci->pvt_info;
+ if (pvt->error_type == 1)
+ return sprintf(data, "single");
+ else if (pvt->error_type == 2)
+ return sprintf(data, "double");
+
+ return 0;
+}
+
+static DEVICE_ATTR(inject, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_inject_show, octeon_mc_inject_inject_store);
+static DEVICE_ATTR(error_type, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_error_type_show, octeon_mc_inject_error_type_store);
+static DEVICE_ATTR(dimm, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_dimm_show, octeon_mc_inject_dimm_store);
+static DEVICE_ATTR(rank, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_rank_show, octeon_mc_inject_rank_store);
+static DEVICE_ATTR(bank, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_bank_show, octeon_mc_inject_bank_store);
+static DEVICE_ATTR(row, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_row_show, octeon_mc_inject_row_store);
+static DEVICE_ATTR(col, S_IRUGO | S_IWUSR,
+ octeon_mc_inject_col_show, octeon_mc_inject_col_store);
+
+static struct attribute *octeon_dev_attrs[] = {
+ &dev_attr_inject.attr,
+ &dev_attr_error_type.attr,
+ &dev_attr_dimm.attr,
+ &dev_attr_rank.attr,
+ &dev_attr_bank.attr,
+ &dev_attr_row.attr,
+ &dev_attr_col.attr,
+ NULL
+};
+
+ATTRIBUTE_GROUPS(octeon_dev);
+
+static int octeon_lmc_edac_probe(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[1];
+ int mc = pdev->id;
+
+ opstate_init();
+
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = 1;
+ layers[0].is_virt_csrow = false;
+
+ if (OCTEON_IS_OCTEON1PLUS()) {
+ union cvmx_lmcx_mem_cfg0 cfg0;
+
+ cfg0.u64 = cvmx_read_csr(CVMX_LMCX_MEM_CFG0(0));
+ if (!cfg0.s.ecc_ena) {
+ dev_info(&pdev->dev, "Disabled (ECC not enabled)\n");
+ return 0;
+ }
+
+ mci = edac_mc_alloc(mc, ARRAY_SIZE(layers), layers, sizeof(struct octeon_lmc_pvt));
+ if (!mci)
+ return -ENXIO;
+
+ mci->pdev = &pdev->dev;
+ mci->dev_name = dev_name(&pdev->dev);
+
+ mci->mod_name = "octeon-lmc";
+ mci->ctl_name = "octeon-lmc-err";
+ mci->edac_check = octeon_lmc_edac_poll;
+
+ if (edac_mc_add_mc_with_groups(mci, octeon_dev_groups)) {
+ dev_err(&pdev->dev, "edac_mc_add_mc() failed\n");
+ edac_mc_free(mci);
+ return -ENXIO;
+ }
+
+ cfg0.u64 = cvmx_read_csr(CVMX_LMCX_MEM_CFG0(mc));
+ cfg0.s.intr_ded_ena = 0; /* We poll */
+ cfg0.s.intr_sec_ena = 0;
+ cvmx_write_csr(CVMX_LMCX_MEM_CFG0(mc), cfg0.u64);
+ } else {
+ /* OCTEON II */
+ union cvmx_lmcx_int_en en;
+ union cvmx_lmcx_config config;
+
+ config.u64 = cvmx_read_csr(CVMX_LMCX_CONFIG(0));
+ if (!config.s.ecc_ena) {
+ dev_info(&pdev->dev, "Disabled (ECC not enabled)\n");
+ return 0;
+ }
+
+ mci = edac_mc_alloc(mc, ARRAY_SIZE(layers), layers, sizeof(struct octeon_lmc_pvt));
+ if (!mci)
+ return -ENXIO;
+
+ mci->pdev = &pdev->dev;
+ mci->dev_name = dev_name(&pdev->dev);
+
+ mci->mod_name = "octeon-lmc";
+ mci->ctl_name = "co_lmc_err";
+ mci->edac_check = octeon_lmc_edac_poll_o2;
+
+ if (edac_mc_add_mc_with_groups(mci, octeon_dev_groups)) {
+ dev_err(&pdev->dev, "edac_mc_add_mc() failed\n");
+ edac_mc_free(mci);
+ return -ENXIO;
+ }
+
+ en.u64 = cvmx_read_csr(CVMX_LMCX_MEM_CFG0(mc));
+ en.s.intr_ded_ena = 0; /* We poll */
+ en.s.intr_sec_ena = 0;
+ cvmx_write_csr(CVMX_LMCX_MEM_CFG0(mc), en.u64);
+ }
+ platform_set_drvdata(pdev, mci);
+
+ return 0;
+}
+
+static int octeon_lmc_edac_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+ return 0;
+}
+
+static struct platform_driver octeon_lmc_edac_driver = {
+ .probe = octeon_lmc_edac_probe,
+ .remove = octeon_lmc_edac_remove,
+ .driver = {
+ .name = "octeon_lmc_edac",
+ }
+};
+module_platform_driver(octeon_lmc_edac_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>");
diff --git a/drivers/edac/octeon_edac-pc.c b/drivers/edac/octeon_edac-pc.c
new file mode 100644
index 0000000000..754eced59c
--- /dev/null
+++ b/drivers/edac/octeon_edac-pc.c
@@ -0,0 +1,142 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2012 Cavium, Inc.
+ *
+ * Copyright (C) 2009 Wind River Systems,
+ * written by Ralf Baechle <ralf@linux-mips.org>
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/edac.h>
+
+#include "edac_module.h"
+
+#include <asm/octeon/cvmx.h>
+#include <asm/mipsregs.h>
+
+extern int register_co_cache_error_notifier(struct notifier_block *nb);
+extern int unregister_co_cache_error_notifier(struct notifier_block *nb);
+
+extern unsigned long long cache_err_dcache[NR_CPUS];
+
+struct co_cache_error {
+ struct notifier_block notifier;
+ struct edac_device_ctl_info *ed;
+};
+
+/**
+ * EDAC CPU cache error callback
+ *
+ * @event: non-zero if unrecoverable.
+ */
+static int co_cache_error_event(struct notifier_block *this,
+ unsigned long event, void *ptr)
+{
+ struct co_cache_error *p = container_of(this, struct co_cache_error,
+ notifier);
+
+ unsigned int core = cvmx_get_core_num();
+ unsigned int cpu = smp_processor_id();
+ u64 icache_err = read_octeon_c0_icacheerr();
+ u64 dcache_err;
+
+ if (event) {
+ dcache_err = cache_err_dcache[core];
+ cache_err_dcache[core] = 0;
+ } else {
+ dcache_err = read_octeon_c0_dcacheerr();
+ }
+
+ if (icache_err & 1) {
+ edac_device_printk(p->ed, KERN_ERR,
+ "CacheErr (Icache):%llx, core %d/cpu %d, cp0_errorepc == %lx\n",
+ (unsigned long long)icache_err, core, cpu,
+ read_c0_errorepc());
+ write_octeon_c0_icacheerr(0);
+ edac_device_handle_ce(p->ed, cpu, 1, "icache");
+ }
+ if (dcache_err & 1) {
+ edac_device_printk(p->ed, KERN_ERR,
+ "CacheErr (Dcache):%llx, core %d/cpu %d, cp0_errorepc == %lx\n",
+ (unsigned long long)dcache_err, core, cpu,
+ read_c0_errorepc());
+ if (event)
+ edac_device_handle_ue(p->ed, cpu, 0, "dcache");
+ else
+ edac_device_handle_ce(p->ed, cpu, 0, "dcache");
+
+ /* Clear the error indication */
+ if (OCTEON_IS_OCTEON2())
+ write_octeon_c0_dcacheerr(1);
+ else
+ write_octeon_c0_dcacheerr(0);
+ }
+
+ return NOTIFY_STOP;
+}
+
+static int co_cache_error_probe(struct platform_device *pdev)
+{
+ struct co_cache_error *p = devm_kzalloc(&pdev->dev, sizeof(*p),
+ GFP_KERNEL);
+ if (!p)
+ return -ENOMEM;
+
+ p->notifier.notifier_call = co_cache_error_event;
+ platform_set_drvdata(pdev, p);
+
+ p->ed = edac_device_alloc_ctl_info(0, "cpu", num_possible_cpus(),
+ "cache", 2, 0, NULL, 0,
+ edac_device_alloc_index());
+ if (!p->ed)
+ goto err;
+
+ p->ed->dev = &pdev->dev;
+
+ p->ed->dev_name = dev_name(&pdev->dev);
+
+ p->ed->mod_name = "octeon-cpu";
+ p->ed->ctl_name = "cache";
+
+ if (edac_device_add_device(p->ed)) {
+ pr_err("%s: edac_device_add_device() failed\n", __func__);
+ goto err1;
+ }
+
+ register_co_cache_error_notifier(&p->notifier);
+
+ return 0;
+
+err1:
+ edac_device_free_ctl_info(p->ed);
+err:
+ return -ENXIO;
+}
+
+static int co_cache_error_remove(struct platform_device *pdev)
+{
+ struct co_cache_error *p = platform_get_drvdata(pdev);
+
+ unregister_co_cache_error_notifier(&p->notifier);
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(p->ed);
+ return 0;
+}
+
+static struct platform_driver co_cache_error_driver = {
+ .probe = co_cache_error_probe,
+ .remove = co_cache_error_remove,
+ .driver = {
+ .name = "octeon_pc_edac",
+ }
+};
+module_platform_driver(co_cache_error_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>");
diff --git a/drivers/edac/octeon_edac-pci.c b/drivers/edac/octeon_edac-pci.c
new file mode 100644
index 0000000000..28b238eece
--- /dev/null
+++ b/drivers/edac/octeon_edac-pci.c
@@ -0,0 +1,110 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2012 Cavium, Inc.
+ * Copyright (C) 2009 Wind River Systems,
+ * written by Ralf Baechle <ralf@linux-mips.org>
+ */
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+#include <linux/edac.h>
+
+#include <asm/octeon/cvmx.h>
+#include <asm/octeon/cvmx-npi-defs.h>
+#include <asm/octeon/cvmx-pci-defs.h>
+#include <asm/octeon/octeon.h>
+
+#include "edac_module.h"
+
+static void octeon_pci_poll(struct edac_pci_ctl_info *pci)
+{
+ union cvmx_pci_cfg01 cfg01;
+
+ cfg01.u32 = octeon_npi_read32(CVMX_NPI_PCI_CFG01);
+ if (cfg01.s.dpe) { /* Detected parity error */
+ edac_pci_handle_pe(pci, pci->ctl_name);
+ cfg01.s.dpe = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+ if (cfg01.s.sse) {
+ edac_pci_handle_npe(pci, "Signaled System Error");
+ cfg01.s.sse = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+ if (cfg01.s.rma) {
+ edac_pci_handle_npe(pci, "Received Master Abort");
+ cfg01.s.rma = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+ if (cfg01.s.rta) {
+ edac_pci_handle_npe(pci, "Received Target Abort");
+ cfg01.s.rta = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+ if (cfg01.s.sta) {
+ edac_pci_handle_npe(pci, "Signaled Target Abort");
+ cfg01.s.sta = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+ if (cfg01.s.mdpe) {
+ edac_pci_handle_npe(pci, "Master Data Parity Error");
+ cfg01.s.mdpe = 1; /* Reset */
+ octeon_npi_write32(CVMX_NPI_PCI_CFG01, cfg01.u32);
+ }
+}
+
+static int octeon_pci_probe(struct platform_device *pdev)
+{
+ struct edac_pci_ctl_info *pci;
+ int res = 0;
+
+ pci = edac_pci_alloc_ctl_info(0, "octeon_pci_err");
+ if (!pci)
+ return -ENOMEM;
+
+ pci->dev = &pdev->dev;
+ platform_set_drvdata(pdev, pci);
+ pci->dev_name = dev_name(&pdev->dev);
+
+ pci->mod_name = "octeon-pci";
+ pci->ctl_name = "octeon_pci_err";
+ pci->edac_check = octeon_pci_poll;
+
+ if (edac_pci_add_device(pci, 0) > 0) {
+ pr_err("%s: edac_pci_add_device() failed\n", __func__);
+ goto err;
+ }
+
+ return 0;
+
+err:
+ edac_pci_free_ctl_info(pci);
+
+ return res;
+}
+
+static int octeon_pci_remove(struct platform_device *pdev)
+{
+ struct edac_pci_ctl_info *pci = platform_get_drvdata(pdev);
+
+ edac_pci_del_device(&pdev->dev);
+ edac_pci_free_ctl_info(pci);
+
+ return 0;
+}
+
+static struct platform_driver octeon_pci_driver = {
+ .probe = octeon_pci_probe,
+ .remove = octeon_pci_remove,
+ .driver = {
+ .name = "octeon_pci_edac",
+ }
+};
+module_platform_driver(octeon_pci_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>");
diff --git a/drivers/edac/pasemi_edac.c b/drivers/edac/pasemi_edac.c
new file mode 100644
index 0000000000..1a1c3296cc
--- /dev/null
+++ b/drivers/edac/pasemi_edac.c
@@ -0,0 +1,303 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2006-2007 PA Semi, Inc
+ *
+ * Author: Egor Martovetsky <egor@pasemi.com>
+ * Maintained by: Olof Johansson <olof@lixom.net>
+ *
+ * Driver for the PWRficient onchip memory controllers
+ */
+
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define MODULE_NAME "pasemi_edac"
+
+#define MCCFG_MCEN 0x300
+#define MCCFG_MCEN_MMC_EN 0x00000001
+#define MCCFG_ERRCOR 0x388
+#define MCCFG_ERRCOR_RNK_FAIL_DET_EN 0x00000100
+#define MCCFG_ERRCOR_ECC_GEN_EN 0x00000010
+#define MCCFG_ERRCOR_ECC_CRR_EN 0x00000001
+#define MCCFG_SCRUB 0x384
+#define MCCFG_SCRUB_RGLR_SCRB_EN 0x00000001
+#define MCDEBUG_ERRCTL1 0x728
+#define MCDEBUG_ERRCTL1_RFL_LOG_EN 0x00080000
+#define MCDEBUG_ERRCTL1_MBE_LOG_EN 0x00040000
+#define MCDEBUG_ERRCTL1_SBE_LOG_EN 0x00020000
+#define MCDEBUG_ERRSTA 0x730
+#define MCDEBUG_ERRSTA_RFL_STATUS 0x00000004
+#define MCDEBUG_ERRSTA_MBE_STATUS 0x00000002
+#define MCDEBUG_ERRSTA_SBE_STATUS 0x00000001
+#define MCDEBUG_ERRCNT1 0x734
+#define MCDEBUG_ERRCNT1_SBE_CNT_OVRFLO 0x00000080
+#define MCDEBUG_ERRLOG1A 0x738
+#define MCDEBUG_ERRLOG1A_MERR_TYPE_M 0x30000000
+#define MCDEBUG_ERRLOG1A_MERR_TYPE_NONE 0x00000000
+#define MCDEBUG_ERRLOG1A_MERR_TYPE_SBE 0x10000000
+#define MCDEBUG_ERRLOG1A_MERR_TYPE_MBE 0x20000000
+#define MCDEBUG_ERRLOG1A_MERR_TYPE_RFL 0x30000000
+#define MCDEBUG_ERRLOG1A_MERR_BA_M 0x00700000
+#define MCDEBUG_ERRLOG1A_MERR_BA_S 20
+#define MCDEBUG_ERRLOG1A_MERR_CS_M 0x00070000
+#define MCDEBUG_ERRLOG1A_MERR_CS_S 16
+#define MCDEBUG_ERRLOG1A_SYNDROME_M 0x0000ffff
+#define MCDRAM_RANKCFG 0x114
+#define MCDRAM_RANKCFG_EN 0x00000001
+#define MCDRAM_RANKCFG_TYPE_SIZE_M 0x000001c0
+#define MCDRAM_RANKCFG_TYPE_SIZE_S 6
+
+#define PASEMI_EDAC_NR_CSROWS 8
+#define PASEMI_EDAC_NR_CHANS 1
+#define PASEMI_EDAC_ERROR_GRAIN 64
+
+static int last_page_in_mmc;
+static int system_mmc_id;
+
+
+static u32 pasemi_edac_get_error_info(struct mem_ctl_info *mci)
+{
+ struct pci_dev *pdev = to_pci_dev(mci->pdev);
+ u32 tmp;
+
+ pci_read_config_dword(pdev, MCDEBUG_ERRSTA,
+ &tmp);
+
+ tmp &= (MCDEBUG_ERRSTA_RFL_STATUS | MCDEBUG_ERRSTA_MBE_STATUS
+ | MCDEBUG_ERRSTA_SBE_STATUS);
+
+ if (tmp) {
+ if (tmp & MCDEBUG_ERRSTA_SBE_STATUS)
+ pci_write_config_dword(pdev, MCDEBUG_ERRCNT1,
+ MCDEBUG_ERRCNT1_SBE_CNT_OVRFLO);
+ pci_write_config_dword(pdev, MCDEBUG_ERRSTA, tmp);
+ }
+
+ return tmp;
+}
+
+static void pasemi_edac_process_error_info(struct mem_ctl_info *mci, u32 errsta)
+{
+ struct pci_dev *pdev = to_pci_dev(mci->pdev);
+ u32 errlog1a;
+ u32 cs;
+
+ if (!errsta)
+ return;
+
+ pci_read_config_dword(pdev, MCDEBUG_ERRLOG1A, &errlog1a);
+
+ cs = (errlog1a & MCDEBUG_ERRLOG1A_MERR_CS_M) >>
+ MCDEBUG_ERRLOG1A_MERR_CS_S;
+
+ /* uncorrectable/multi-bit errors */
+ if (errsta & (MCDEBUG_ERRSTA_MBE_STATUS |
+ MCDEBUG_ERRSTA_RFL_STATUS)) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ mci->csrows[cs]->first_page, 0, 0,
+ cs, 0, -1, mci->ctl_name, "");
+ }
+
+ /* correctable/single-bit errors */
+ if (errsta & MCDEBUG_ERRSTA_SBE_STATUS)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ mci->csrows[cs]->first_page, 0, 0,
+ cs, 0, -1, mci->ctl_name, "");
+}
+
+static void pasemi_edac_check(struct mem_ctl_info *mci)
+{
+ u32 errsta;
+
+ errsta = pasemi_edac_get_error_info(mci);
+ if (errsta)
+ pasemi_edac_process_error_info(mci, errsta);
+}
+
+static int pasemi_edac_init_csrows(struct mem_ctl_info *mci,
+ struct pci_dev *pdev,
+ enum edac_type edac_mode)
+{
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ u32 rankcfg;
+ int index;
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ pci_read_config_dword(pdev,
+ MCDRAM_RANKCFG + (index * 12),
+ &rankcfg);
+
+ if (!(rankcfg & MCDRAM_RANKCFG_EN))
+ continue;
+
+ switch ((rankcfg & MCDRAM_RANKCFG_TYPE_SIZE_M) >>
+ MCDRAM_RANKCFG_TYPE_SIZE_S) {
+ case 0:
+ dimm->nr_pages = 128 << (20 - PAGE_SHIFT);
+ break;
+ case 1:
+ dimm->nr_pages = 256 << (20 - PAGE_SHIFT);
+ break;
+ case 2:
+ case 3:
+ dimm->nr_pages = 512 << (20 - PAGE_SHIFT);
+ break;
+ case 4:
+ dimm->nr_pages = 1024 << (20 - PAGE_SHIFT);
+ break;
+ case 5:
+ dimm->nr_pages = 2048 << (20 - PAGE_SHIFT);
+ break;
+ default:
+ edac_mc_printk(mci, KERN_ERR,
+ "Unrecognized Rank Config. rankcfg=%u\n",
+ rankcfg);
+ return -EINVAL;
+ }
+
+ csrow->first_page = last_page_in_mmc;
+ csrow->last_page = csrow->first_page + dimm->nr_pages - 1;
+ last_page_in_mmc += dimm->nr_pages;
+ csrow->page_mask = 0;
+ dimm->grain = PASEMI_EDAC_ERROR_GRAIN;
+ dimm->mtype = MEM_DDR;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = edac_mode;
+ }
+ return 0;
+}
+
+static int pasemi_edac_probe(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ u32 errctl1, errcor, scrub, mcen;
+
+ pci_read_config_dword(pdev, MCCFG_MCEN, &mcen);
+ if (!(mcen & MCCFG_MCEN_MMC_EN))
+ return -ENODEV;
+
+ /*
+ * We should think about enabling other error detection later on
+ */
+
+ pci_read_config_dword(pdev, MCDEBUG_ERRCTL1, &errctl1);
+ errctl1 |= MCDEBUG_ERRCTL1_SBE_LOG_EN |
+ MCDEBUG_ERRCTL1_MBE_LOG_EN |
+ MCDEBUG_ERRCTL1_RFL_LOG_EN;
+ pci_write_config_dword(pdev, MCDEBUG_ERRCTL1, errctl1);
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = PASEMI_EDAC_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = PASEMI_EDAC_NR_CHANS;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(system_mmc_id++, ARRAY_SIZE(layers), layers,
+ 0);
+ if (mci == NULL)
+ return -ENOMEM;
+
+ pci_read_config_dword(pdev, MCCFG_ERRCOR, &errcor);
+ errcor |= MCCFG_ERRCOR_RNK_FAIL_DET_EN |
+ MCCFG_ERRCOR_ECC_GEN_EN |
+ MCCFG_ERRCOR_ECC_CRR_EN;
+
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR | MEM_FLAG_RDDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ mci->edac_cap = (errcor & MCCFG_ERRCOR_ECC_GEN_EN) ?
+ ((errcor & MCCFG_ERRCOR_ECC_CRR_EN) ?
+ (EDAC_FLAG_EC | EDAC_FLAG_SECDED) : EDAC_FLAG_EC) :
+ EDAC_FLAG_NONE;
+ mci->mod_name = MODULE_NAME;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_name = "pasemi,pwrficient-mc";
+ mci->edac_check = pasemi_edac_check;
+ mci->ctl_page_to_phys = NULL;
+ pci_read_config_dword(pdev, MCCFG_SCRUB, &scrub);
+ mci->scrub_cap = SCRUB_FLAG_HW_PROG | SCRUB_FLAG_HW_SRC;
+ mci->scrub_mode =
+ ((errcor & MCCFG_ERRCOR_ECC_CRR_EN) ? SCRUB_FLAG_HW_SRC : 0) |
+ ((scrub & MCCFG_SCRUB_RGLR_SCRB_EN) ? SCRUB_FLAG_HW_PROG : 0);
+
+ if (pasemi_edac_init_csrows(mci, pdev,
+ (mci->edac_cap & EDAC_FLAG_SECDED) ?
+ EDAC_SECDED :
+ ((mci->edac_cap & EDAC_FLAG_EC) ?
+ EDAC_EC : EDAC_NONE)))
+ goto fail;
+
+ /*
+ * Clear status
+ */
+ pasemi_edac_get_error_info(mci);
+
+ if (edac_mc_add_mc(mci))
+ goto fail;
+
+ /* get this far and it's successful */
+ return 0;
+
+fail:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+static void pasemi_edac_remove(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci = edac_mc_del_mc(&pdev->dev);
+
+ if (!mci)
+ return;
+
+ edac_mc_free(mci);
+}
+
+
+static const struct pci_device_id pasemi_edac_pci_tbl[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_PASEMI, 0xa00a) },
+ { }
+};
+
+MODULE_DEVICE_TABLE(pci, pasemi_edac_pci_tbl);
+
+static struct pci_driver pasemi_edac_driver = {
+ .name = MODULE_NAME,
+ .probe = pasemi_edac_probe,
+ .remove = pasemi_edac_remove,
+ .id_table = pasemi_edac_pci_tbl,
+};
+
+static int __init pasemi_edac_init(void)
+{
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ return pci_register_driver(&pasemi_edac_driver);
+}
+
+static void __exit pasemi_edac_exit(void)
+{
+ pci_unregister_driver(&pasemi_edac_driver);
+}
+
+module_init(pasemi_edac_init);
+module_exit(pasemi_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Egor Martovetsky <egor@pasemi.com>");
+MODULE_DESCRIPTION("MC support for PA Semi PWRficient memory controller");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
diff --git a/drivers/edac/pnd2_edac.c b/drivers/edac/pnd2_edac.c
new file mode 100644
index 0000000000..2b306f2cc6
--- /dev/null
+++ b/drivers/edac/pnd2_edac.c
@@ -0,0 +1,1587 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Driver for Pondicherry2 memory controller.
+ *
+ * Copyright (c) 2016, Intel Corporation.
+ *
+ * [Derived from sb_edac.c]
+ *
+ * Translation of system physical addresses to DIMM addresses
+ * is a two stage process:
+ *
+ * First the Pondicherry 2 memory controller handles slice and channel interleaving
+ * in "sys2pmi()". This is (almost) completley common between platforms.
+ *
+ * Then a platform specific dunit (DIMM unit) completes the process to provide DIMM,
+ * rank, bank, row and column using the appropriate "dunit_ops" functions/parameters.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+#include <linux/smp.h>
+#include <linux/bitmap.h>
+#include <linux/math64.h>
+#include <linux/mod_devicetable.h>
+#include <linux/platform_data/x86/p2sb.h>
+
+#include <asm/cpu_device_id.h>
+#include <asm/intel-family.h>
+#include <asm/processor.h>
+#include <asm/mce.h>
+
+#include "edac_mc.h"
+#include "edac_module.h"
+#include "pnd2_edac.h"
+
+#define EDAC_MOD_STR "pnd2_edac"
+
+#define APL_NUM_CHANNELS 4
+#define DNV_NUM_CHANNELS 2
+#define DNV_MAX_DIMMS 2 /* Max DIMMs per channel */
+
+enum type {
+ APL,
+ DNV, /* All requests go to PMI CH0 on each slice (CH1 disabled) */
+};
+
+struct dram_addr {
+ int chan;
+ int dimm;
+ int rank;
+ int bank;
+ int row;
+ int col;
+};
+
+struct pnd2_pvt {
+ int dimm_geom[APL_NUM_CHANNELS];
+ u64 tolm, tohm;
+};
+
+/*
+ * System address space is divided into multiple regions with
+ * different interleave rules in each. The as0/as1 regions
+ * have no interleaving at all. The as2 region is interleaved
+ * between two channels. The mot region is magic and may overlap
+ * other regions, with its interleave rules taking precedence.
+ * Addresses not in any of these regions are interleaved across
+ * all four channels.
+ */
+static struct region {
+ u64 base;
+ u64 limit;
+ u8 enabled;
+} mot, as0, as1, as2;
+
+static struct dunit_ops {
+ char *name;
+ enum type type;
+ int pmiaddr_shift;
+ int pmiidx_shift;
+ int channels;
+ int dimms_per_channel;
+ int (*rd_reg)(int port, int off, int op, void *data, size_t sz, char *name);
+ int (*get_registers)(void);
+ int (*check_ecc)(void);
+ void (*mk_region)(char *name, struct region *rp, void *asym);
+ void (*get_dimm_config)(struct mem_ctl_info *mci);
+ int (*pmi2mem)(struct mem_ctl_info *mci, u64 pmiaddr, u32 pmiidx,
+ struct dram_addr *daddr, char *msg);
+} *ops;
+
+static struct mem_ctl_info *pnd2_mci;
+
+#define PND2_MSG_SIZE 256
+
+/* Debug macros */
+#define pnd2_printk(level, fmt, arg...) \
+ edac_printk(level, "pnd2", fmt, ##arg)
+
+#define pnd2_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "pnd2", fmt, ##arg)
+
+#define MOT_CHAN_INTLV_BIT_1SLC_2CH 12
+#define MOT_CHAN_INTLV_BIT_2SLC_2CH 13
+#define SELECTOR_DISABLED (-1)
+#define _4GB (1ul << 32)
+
+#define PMI_ADDRESS_WIDTH 31
+#define PND_MAX_PHYS_BIT 39
+
+#define APL_ASYMSHIFT 28
+#define DNV_ASYMSHIFT 31
+#define CH_HASH_MASK_LSB 6
+#define SLICE_HASH_MASK_LSB 6
+#define MOT_SLC_INTLV_BIT 12
+#define LOG2_PMI_ADDR_GRANULARITY 5
+#define MOT_SHIFT 24
+
+#define GET_BITFIELD(v, lo, hi) (((v) & GENMASK_ULL(hi, lo)) >> (lo))
+#define U64_LSHIFT(val, s) ((u64)(val) << (s))
+
+/*
+ * On Apollo Lake we access memory controller registers via a
+ * side-band mailbox style interface in a hidden PCI device
+ * configuration space.
+ */
+static struct pci_bus *p2sb_bus;
+#define P2SB_DEVFN PCI_DEVFN(0xd, 0)
+#define P2SB_ADDR_OFF 0xd0
+#define P2SB_DATA_OFF 0xd4
+#define P2SB_STAT_OFF 0xd8
+#define P2SB_ROUT_OFF 0xda
+#define P2SB_EADD_OFF 0xdc
+#define P2SB_HIDE_OFF 0xe1
+
+#define P2SB_BUSY 1
+
+#define P2SB_READ(size, off, ptr) \
+ pci_bus_read_config_##size(p2sb_bus, P2SB_DEVFN, off, ptr)
+#define P2SB_WRITE(size, off, val) \
+ pci_bus_write_config_##size(p2sb_bus, P2SB_DEVFN, off, val)
+
+static bool p2sb_is_busy(u16 *status)
+{
+ P2SB_READ(word, P2SB_STAT_OFF, status);
+
+ return !!(*status & P2SB_BUSY);
+}
+
+static int _apl_rd_reg(int port, int off, int op, u32 *data)
+{
+ int retries = 0xff, ret;
+ u16 status;
+ u8 hidden;
+
+ /* Unhide the P2SB device, if it's hidden */
+ P2SB_READ(byte, P2SB_HIDE_OFF, &hidden);
+ if (hidden)
+ P2SB_WRITE(byte, P2SB_HIDE_OFF, 0);
+
+ if (p2sb_is_busy(&status)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+
+ P2SB_WRITE(dword, P2SB_ADDR_OFF, (port << 24) | off);
+ P2SB_WRITE(dword, P2SB_DATA_OFF, 0);
+ P2SB_WRITE(dword, P2SB_EADD_OFF, 0);
+ P2SB_WRITE(word, P2SB_ROUT_OFF, 0);
+ P2SB_WRITE(word, P2SB_STAT_OFF, (op << 8) | P2SB_BUSY);
+
+ while (p2sb_is_busy(&status)) {
+ if (retries-- == 0) {
+ ret = -EBUSY;
+ goto out;
+ }
+ }
+
+ P2SB_READ(dword, P2SB_DATA_OFF, data);
+ ret = (status >> 1) & 0x3;
+out:
+ /* Hide the P2SB device, if it was hidden before */
+ if (hidden)
+ P2SB_WRITE(byte, P2SB_HIDE_OFF, hidden);
+
+ return ret;
+}
+
+static int apl_rd_reg(int port, int off, int op, void *data, size_t sz, char *name)
+{
+ int ret = 0;
+
+ edac_dbg(2, "Read %s port=%x off=%x op=%x\n", name, port, off, op);
+ switch (sz) {
+ case 8:
+ ret = _apl_rd_reg(port, off + 4, op, (u32 *)(data + 4));
+ fallthrough;
+ case 4:
+ ret |= _apl_rd_reg(port, off, op, (u32 *)data);
+ pnd2_printk(KERN_DEBUG, "%s=%x%08x ret=%d\n", name,
+ sz == 8 ? *((u32 *)(data + 4)) : 0, *((u32 *)data), ret);
+ break;
+ }
+
+ return ret;
+}
+
+static u64 get_mem_ctrl_hub_base_addr(void)
+{
+ struct b_cr_mchbar_lo_pci lo;
+ struct b_cr_mchbar_hi_pci hi;
+ struct pci_dev *pdev;
+
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x1980, NULL);
+ if (pdev) {
+ pci_read_config_dword(pdev, 0x48, (u32 *)&lo);
+ pci_read_config_dword(pdev, 0x4c, (u32 *)&hi);
+ pci_dev_put(pdev);
+ } else {
+ return 0;
+ }
+
+ if (!lo.enable) {
+ edac_dbg(2, "MMIO via memory controller hub base address is disabled!\n");
+ return 0;
+ }
+
+ return U64_LSHIFT(hi.base, 32) | U64_LSHIFT(lo.base, 15);
+}
+
+#define DNV_MCHBAR_SIZE 0x8000
+#define DNV_SB_PORT_SIZE 0x10000
+static int dnv_rd_reg(int port, int off, int op, void *data, size_t sz, char *name)
+{
+ struct pci_dev *pdev;
+ void __iomem *base;
+ struct resource r;
+ int ret;
+
+ if (op == 4) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x1980, NULL);
+ if (!pdev)
+ return -ENODEV;
+
+ pci_read_config_dword(pdev, off, data);
+ pci_dev_put(pdev);
+ } else {
+ /* MMIO via memory controller hub base address */
+ if (op == 0 && port == 0x4c) {
+ memset(&r, 0, sizeof(r));
+
+ r.start = get_mem_ctrl_hub_base_addr();
+ if (!r.start)
+ return -ENODEV;
+ r.end = r.start + DNV_MCHBAR_SIZE - 1;
+ } else {
+ /* MMIO via sideband register base address */
+ ret = p2sb_bar(NULL, 0, &r);
+ if (ret)
+ return ret;
+
+ r.start += (port << 16);
+ r.end = r.start + DNV_SB_PORT_SIZE - 1;
+ }
+
+ base = ioremap(r.start, resource_size(&r));
+ if (!base)
+ return -ENODEV;
+
+ if (sz == 8)
+ *(u64 *)data = readq(base + off);
+ else
+ *(u32 *)data = readl(base + off);
+
+ iounmap(base);
+ }
+
+ edac_dbg(2, "Read %s=%.8x_%.8x\n", name,
+ (sz == 8) ? *(u32 *)(data + 4) : 0, *(u32 *)data);
+
+ return 0;
+}
+
+#define RD_REGP(regp, regname, port) \
+ ops->rd_reg(port, \
+ regname##_offset, \
+ regname##_r_opcode, \
+ regp, sizeof(struct regname), \
+ #regname)
+
+#define RD_REG(regp, regname) \
+ ops->rd_reg(regname ## _port, \
+ regname##_offset, \
+ regname##_r_opcode, \
+ regp, sizeof(struct regname), \
+ #regname)
+
+static u64 top_lm, top_hm;
+static bool two_slices;
+static bool two_channels; /* Both PMI channels in one slice enabled */
+
+static u8 sym_chan_mask;
+static u8 asym_chan_mask;
+static u8 chan_mask;
+
+static int slice_selector = -1;
+static int chan_selector = -1;
+static u64 slice_hash_mask;
+static u64 chan_hash_mask;
+
+static void mk_region(char *name, struct region *rp, u64 base, u64 limit)
+{
+ rp->enabled = 1;
+ rp->base = base;
+ rp->limit = limit;
+ edac_dbg(2, "Region:%s [%llx, %llx]\n", name, base, limit);
+}
+
+static void mk_region_mask(char *name, struct region *rp, u64 base, u64 mask)
+{
+ if (mask == 0) {
+ pr_info(FW_BUG "MOT mask cannot be zero\n");
+ return;
+ }
+ if (mask != GENMASK_ULL(PND_MAX_PHYS_BIT, __ffs(mask))) {
+ pr_info(FW_BUG "MOT mask not power of two\n");
+ return;
+ }
+ if (base & ~mask) {
+ pr_info(FW_BUG "MOT region base/mask alignment error\n");
+ return;
+ }
+ rp->base = base;
+ rp->limit = (base | ~mask) & GENMASK_ULL(PND_MAX_PHYS_BIT, 0);
+ rp->enabled = 1;
+ edac_dbg(2, "Region:%s [%llx, %llx]\n", name, base, rp->limit);
+}
+
+static bool in_region(struct region *rp, u64 addr)
+{
+ if (!rp->enabled)
+ return false;
+
+ return rp->base <= addr && addr <= rp->limit;
+}
+
+static int gen_sym_mask(struct b_cr_slice_channel_hash *p)
+{
+ int mask = 0;
+
+ if (!p->slice_0_mem_disabled)
+ mask |= p->sym_slice0_channel_enabled;
+
+ if (!p->slice_1_disabled)
+ mask |= p->sym_slice1_channel_enabled << 2;
+
+ if (p->ch_1_disabled || p->enable_pmi_dual_data_mode)
+ mask &= 0x5;
+
+ return mask;
+}
+
+static int gen_asym_mask(struct b_cr_slice_channel_hash *p,
+ struct b_cr_asym_mem_region0_mchbar *as0,
+ struct b_cr_asym_mem_region1_mchbar *as1,
+ struct b_cr_asym_2way_mem_region_mchbar *as2way)
+{
+ const int intlv[] = { 0x5, 0xA, 0x3, 0xC };
+ int mask = 0;
+
+ if (as2way->asym_2way_interleave_enable)
+ mask = intlv[as2way->asym_2way_intlv_mode];
+ if (as0->slice0_asym_enable)
+ mask |= (1 << as0->slice0_asym_channel_select);
+ if (as1->slice1_asym_enable)
+ mask |= (4 << as1->slice1_asym_channel_select);
+ if (p->slice_0_mem_disabled)
+ mask &= 0xc;
+ if (p->slice_1_disabled)
+ mask &= 0x3;
+ if (p->ch_1_disabled || p->enable_pmi_dual_data_mode)
+ mask &= 0x5;
+
+ return mask;
+}
+
+static struct b_cr_tolud_pci tolud;
+static struct b_cr_touud_lo_pci touud_lo;
+static struct b_cr_touud_hi_pci touud_hi;
+static struct b_cr_asym_mem_region0_mchbar asym0;
+static struct b_cr_asym_mem_region1_mchbar asym1;
+static struct b_cr_asym_2way_mem_region_mchbar asym_2way;
+static struct b_cr_mot_out_base_mchbar mot_base;
+static struct b_cr_mot_out_mask_mchbar mot_mask;
+static struct b_cr_slice_channel_hash chash;
+
+/* Apollo Lake dunit */
+/*
+ * Validated on board with just two DIMMs in the [0] and [2] positions
+ * in this array. Other port number matches documentation, but caution
+ * advised.
+ */
+static const int apl_dports[APL_NUM_CHANNELS] = { 0x18, 0x10, 0x11, 0x19 };
+static struct d_cr_drp0 drp0[APL_NUM_CHANNELS];
+
+/* Denverton dunit */
+static const int dnv_dports[DNV_NUM_CHANNELS] = { 0x10, 0x12 };
+static struct d_cr_dsch dsch;
+static struct d_cr_ecc_ctrl ecc_ctrl[DNV_NUM_CHANNELS];
+static struct d_cr_drp drp[DNV_NUM_CHANNELS];
+static struct d_cr_dmap dmap[DNV_NUM_CHANNELS];
+static struct d_cr_dmap1 dmap1[DNV_NUM_CHANNELS];
+static struct d_cr_dmap2 dmap2[DNV_NUM_CHANNELS];
+static struct d_cr_dmap3 dmap3[DNV_NUM_CHANNELS];
+static struct d_cr_dmap4 dmap4[DNV_NUM_CHANNELS];
+static struct d_cr_dmap5 dmap5[DNV_NUM_CHANNELS];
+
+static void apl_mk_region(char *name, struct region *rp, void *asym)
+{
+ struct b_cr_asym_mem_region0_mchbar *a = asym;
+
+ mk_region(name, rp,
+ U64_LSHIFT(a->slice0_asym_base, APL_ASYMSHIFT),
+ U64_LSHIFT(a->slice0_asym_limit, APL_ASYMSHIFT) +
+ GENMASK_ULL(APL_ASYMSHIFT - 1, 0));
+}
+
+static void dnv_mk_region(char *name, struct region *rp, void *asym)
+{
+ struct b_cr_asym_mem_region_denverton *a = asym;
+
+ mk_region(name, rp,
+ U64_LSHIFT(a->slice_asym_base, DNV_ASYMSHIFT),
+ U64_LSHIFT(a->slice_asym_limit, DNV_ASYMSHIFT) +
+ GENMASK_ULL(DNV_ASYMSHIFT - 1, 0));
+}
+
+static int apl_get_registers(void)
+{
+ int ret = -ENODEV;
+ int i;
+
+ if (RD_REG(&asym_2way, b_cr_asym_2way_mem_region_mchbar))
+ return -ENODEV;
+
+ /*
+ * RD_REGP() will fail for unpopulated or non-existent
+ * DIMM slots. Return success if we find at least one DIMM.
+ */
+ for (i = 0; i < APL_NUM_CHANNELS; i++)
+ if (!RD_REGP(&drp0[i], d_cr_drp0, apl_dports[i]))
+ ret = 0;
+
+ return ret;
+}
+
+static int dnv_get_registers(void)
+{
+ int i;
+
+ if (RD_REG(&dsch, d_cr_dsch))
+ return -ENODEV;
+
+ for (i = 0; i < DNV_NUM_CHANNELS; i++)
+ if (RD_REGP(&ecc_ctrl[i], d_cr_ecc_ctrl, dnv_dports[i]) ||
+ RD_REGP(&drp[i], d_cr_drp, dnv_dports[i]) ||
+ RD_REGP(&dmap[i], d_cr_dmap, dnv_dports[i]) ||
+ RD_REGP(&dmap1[i], d_cr_dmap1, dnv_dports[i]) ||
+ RD_REGP(&dmap2[i], d_cr_dmap2, dnv_dports[i]) ||
+ RD_REGP(&dmap3[i], d_cr_dmap3, dnv_dports[i]) ||
+ RD_REGP(&dmap4[i], d_cr_dmap4, dnv_dports[i]) ||
+ RD_REGP(&dmap5[i], d_cr_dmap5, dnv_dports[i]))
+ return -ENODEV;
+
+ return 0;
+}
+
+/*
+ * Read all the h/w config registers once here (they don't
+ * change at run time. Figure out which address ranges have
+ * which interleave characteristics.
+ */
+static int get_registers(void)
+{
+ const int intlv[] = { 10, 11, 12, 12 };
+
+ if (RD_REG(&tolud, b_cr_tolud_pci) ||
+ RD_REG(&touud_lo, b_cr_touud_lo_pci) ||
+ RD_REG(&touud_hi, b_cr_touud_hi_pci) ||
+ RD_REG(&asym0, b_cr_asym_mem_region0_mchbar) ||
+ RD_REG(&asym1, b_cr_asym_mem_region1_mchbar) ||
+ RD_REG(&mot_base, b_cr_mot_out_base_mchbar) ||
+ RD_REG(&mot_mask, b_cr_mot_out_mask_mchbar) ||
+ RD_REG(&chash, b_cr_slice_channel_hash))
+ return -ENODEV;
+
+ if (ops->get_registers())
+ return -ENODEV;
+
+ if (ops->type == DNV) {
+ /* PMI channel idx (always 0) for asymmetric region */
+ asym0.slice0_asym_channel_select = 0;
+ asym1.slice1_asym_channel_select = 0;
+ /* PMI channel bitmap (always 1) for symmetric region */
+ chash.sym_slice0_channel_enabled = 0x1;
+ chash.sym_slice1_channel_enabled = 0x1;
+ }
+
+ if (asym0.slice0_asym_enable)
+ ops->mk_region("as0", &as0, &asym0);
+
+ if (asym1.slice1_asym_enable)
+ ops->mk_region("as1", &as1, &asym1);
+
+ if (asym_2way.asym_2way_interleave_enable) {
+ mk_region("as2way", &as2,
+ U64_LSHIFT(asym_2way.asym_2way_base, APL_ASYMSHIFT),
+ U64_LSHIFT(asym_2way.asym_2way_limit, APL_ASYMSHIFT) +
+ GENMASK_ULL(APL_ASYMSHIFT - 1, 0));
+ }
+
+ if (mot_base.imr_en) {
+ mk_region_mask("mot", &mot,
+ U64_LSHIFT(mot_base.mot_out_base, MOT_SHIFT),
+ U64_LSHIFT(mot_mask.mot_out_mask, MOT_SHIFT));
+ }
+
+ top_lm = U64_LSHIFT(tolud.tolud, 20);
+ top_hm = U64_LSHIFT(touud_hi.touud, 32) | U64_LSHIFT(touud_lo.touud, 20);
+
+ two_slices = !chash.slice_1_disabled &&
+ !chash.slice_0_mem_disabled &&
+ (chash.sym_slice0_channel_enabled != 0) &&
+ (chash.sym_slice1_channel_enabled != 0);
+ two_channels = !chash.ch_1_disabled &&
+ !chash.enable_pmi_dual_data_mode &&
+ ((chash.sym_slice0_channel_enabled == 3) ||
+ (chash.sym_slice1_channel_enabled == 3));
+
+ sym_chan_mask = gen_sym_mask(&chash);
+ asym_chan_mask = gen_asym_mask(&chash, &asym0, &asym1, &asym_2way);
+ chan_mask = sym_chan_mask | asym_chan_mask;
+
+ if (two_slices && !two_channels) {
+ if (chash.hvm_mode)
+ slice_selector = 29;
+ else
+ slice_selector = intlv[chash.interleave_mode];
+ } else if (!two_slices && two_channels) {
+ if (chash.hvm_mode)
+ chan_selector = 29;
+ else
+ chan_selector = intlv[chash.interleave_mode];
+ } else if (two_slices && two_channels) {
+ if (chash.hvm_mode) {
+ slice_selector = 29;
+ chan_selector = 30;
+ } else {
+ slice_selector = intlv[chash.interleave_mode];
+ chan_selector = intlv[chash.interleave_mode] + 1;
+ }
+ }
+
+ if (two_slices) {
+ if (!chash.hvm_mode)
+ slice_hash_mask = chash.slice_hash_mask << SLICE_HASH_MASK_LSB;
+ if (!two_channels)
+ slice_hash_mask |= BIT_ULL(slice_selector);
+ }
+
+ if (two_channels) {
+ if (!chash.hvm_mode)
+ chan_hash_mask = chash.ch_hash_mask << CH_HASH_MASK_LSB;
+ if (!two_slices)
+ chan_hash_mask |= BIT_ULL(chan_selector);
+ }
+
+ return 0;
+}
+
+/* Get a contiguous memory address (remove the MMIO gap) */
+static u64 remove_mmio_gap(u64 sys)
+{
+ return (sys < _4GB) ? sys : sys - (_4GB - top_lm);
+}
+
+/* Squeeze out one address bit, shift upper part down to fill gap */
+static void remove_addr_bit(u64 *addr, int bitidx)
+{
+ u64 mask;
+
+ if (bitidx == -1)
+ return;
+
+ mask = (1ull << bitidx) - 1;
+ *addr = ((*addr >> 1) & ~mask) | (*addr & mask);
+}
+
+/* XOR all the bits from addr specified in mask */
+static int hash_by_mask(u64 addr, u64 mask)
+{
+ u64 result = addr & mask;
+
+ result = (result >> 32) ^ result;
+ result = (result >> 16) ^ result;
+ result = (result >> 8) ^ result;
+ result = (result >> 4) ^ result;
+ result = (result >> 2) ^ result;
+ result = (result >> 1) ^ result;
+
+ return (int)result & 1;
+}
+
+/*
+ * First stage decode. Take the system address and figure out which
+ * second stage will deal with it based on interleave modes.
+ */
+static int sys2pmi(const u64 addr, u32 *pmiidx, u64 *pmiaddr, char *msg)
+{
+ u64 contig_addr, contig_base, contig_offset, contig_base_adj;
+ int mot_intlv_bit = two_slices ? MOT_CHAN_INTLV_BIT_2SLC_2CH :
+ MOT_CHAN_INTLV_BIT_1SLC_2CH;
+ int slice_intlv_bit_rm = SELECTOR_DISABLED;
+ int chan_intlv_bit_rm = SELECTOR_DISABLED;
+ /* Determine if address is in the MOT region. */
+ bool mot_hit = in_region(&mot, addr);
+ /* Calculate the number of symmetric regions enabled. */
+ int sym_channels = hweight8(sym_chan_mask);
+
+ /*
+ * The amount we need to shift the asym base can be determined by the
+ * number of enabled symmetric channels.
+ * NOTE: This can only work because symmetric memory is not supposed
+ * to do a 3-way interleave.
+ */
+ int sym_chan_shift = sym_channels >> 1;
+
+ /* Give up if address is out of range, or in MMIO gap */
+ if (addr >= (1ul << PND_MAX_PHYS_BIT) ||
+ (addr >= top_lm && addr < _4GB) || addr >= top_hm) {
+ snprintf(msg, PND2_MSG_SIZE, "Error address 0x%llx is not DRAM", addr);
+ return -EINVAL;
+ }
+
+ /* Get a contiguous memory address (remove the MMIO gap) */
+ contig_addr = remove_mmio_gap(addr);
+
+ if (in_region(&as0, addr)) {
+ *pmiidx = asym0.slice0_asym_channel_select;
+
+ contig_base = remove_mmio_gap(as0.base);
+ contig_offset = contig_addr - contig_base;
+ contig_base_adj = (contig_base >> sym_chan_shift) *
+ ((chash.sym_slice0_channel_enabled >> (*pmiidx & 1)) & 1);
+ contig_addr = contig_offset + ((sym_channels > 0) ? contig_base_adj : 0ull);
+ } else if (in_region(&as1, addr)) {
+ *pmiidx = 2u + asym1.slice1_asym_channel_select;
+
+ contig_base = remove_mmio_gap(as1.base);
+ contig_offset = contig_addr - contig_base;
+ contig_base_adj = (contig_base >> sym_chan_shift) *
+ ((chash.sym_slice1_channel_enabled >> (*pmiidx & 1)) & 1);
+ contig_addr = contig_offset + ((sym_channels > 0) ? contig_base_adj : 0ull);
+ } else if (in_region(&as2, addr) && (asym_2way.asym_2way_intlv_mode == 0x3ul)) {
+ bool channel1;
+
+ mot_intlv_bit = MOT_CHAN_INTLV_BIT_1SLC_2CH;
+ *pmiidx = (asym_2way.asym_2way_intlv_mode & 1) << 1;
+ channel1 = mot_hit ? ((bool)((addr >> mot_intlv_bit) & 1)) :
+ hash_by_mask(contig_addr, chan_hash_mask);
+ *pmiidx |= (u32)channel1;
+
+ contig_base = remove_mmio_gap(as2.base);
+ chan_intlv_bit_rm = mot_hit ? mot_intlv_bit : chan_selector;
+ contig_offset = contig_addr - contig_base;
+ remove_addr_bit(&contig_offset, chan_intlv_bit_rm);
+ contig_addr = (contig_base >> sym_chan_shift) + contig_offset;
+ } else {
+ /* Otherwise we're in normal, boring symmetric mode. */
+ *pmiidx = 0u;
+
+ if (two_slices) {
+ bool slice1;
+
+ if (mot_hit) {
+ slice_intlv_bit_rm = MOT_SLC_INTLV_BIT;
+ slice1 = (addr >> MOT_SLC_INTLV_BIT) & 1;
+ } else {
+ slice_intlv_bit_rm = slice_selector;
+ slice1 = hash_by_mask(addr, slice_hash_mask);
+ }
+
+ *pmiidx = (u32)slice1 << 1;
+ }
+
+ if (two_channels) {
+ bool channel1;
+
+ mot_intlv_bit = two_slices ? MOT_CHAN_INTLV_BIT_2SLC_2CH :
+ MOT_CHAN_INTLV_BIT_1SLC_2CH;
+
+ if (mot_hit) {
+ chan_intlv_bit_rm = mot_intlv_bit;
+ channel1 = (addr >> mot_intlv_bit) & 1;
+ } else {
+ chan_intlv_bit_rm = chan_selector;
+ channel1 = hash_by_mask(contig_addr, chan_hash_mask);
+ }
+
+ *pmiidx |= (u32)channel1;
+ }
+ }
+
+ /* Remove the chan_selector bit first */
+ remove_addr_bit(&contig_addr, chan_intlv_bit_rm);
+ /* Remove the slice bit (we remove it second because it must be lower */
+ remove_addr_bit(&contig_addr, slice_intlv_bit_rm);
+ *pmiaddr = contig_addr;
+
+ return 0;
+}
+
+/* Translate PMI address to memory (rank, row, bank, column) */
+#define C(n) (0x10 | (n)) /* column */
+#define B(n) (0x20 | (n)) /* bank */
+#define R(n) (0x40 | (n)) /* row */
+#define RS (0x80) /* rank */
+
+/* addrdec values */
+#define AMAP_1KB 0
+#define AMAP_2KB 1
+#define AMAP_4KB 2
+#define AMAP_RSVD 3
+
+/* dden values */
+#define DEN_4Gb 0
+#define DEN_8Gb 2
+
+/* dwid values */
+#define X8 0
+#define X16 1
+
+static struct dimm_geometry {
+ u8 addrdec;
+ u8 dden;
+ u8 dwid;
+ u8 rowbits, colbits;
+ u16 bits[PMI_ADDRESS_WIDTH];
+} dimms[] = {
+ {
+ .addrdec = AMAP_1KB, .dden = DEN_4Gb, .dwid = X16,
+ .rowbits = 15, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), B(0), B(1), B(2), R(0),
+ R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8), R(9),
+ R(10), C(7), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ 0, 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_1KB, .dden = DEN_4Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), B(0), B(1), B(2), R(0),
+ R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8), R(9),
+ R(10), C(7), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_1KB, .dden = DEN_8Gb, .dwid = X16,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), B(0), B(1), B(2), R(0),
+ R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8), R(9),
+ R(10), C(7), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_1KB, .dden = DEN_8Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 11,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), B(0), B(1), B(2), R(0),
+ R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8), R(9),
+ R(10), C(7), C(8), C(9), R(11), RS, C(11), R(12), R(13),
+ R(14), R(15), 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_2KB, .dden = DEN_4Gb, .dwid = X16,
+ .rowbits = 15, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), B(0), B(1), B(2),
+ R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8),
+ R(9), R(10), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ 0, 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_2KB, .dden = DEN_4Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), B(0), B(1), B(2),
+ R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8),
+ R(9), R(10), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_2KB, .dden = DEN_8Gb, .dwid = X16,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), B(0), B(1), B(2),
+ R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8),
+ R(9), R(10), C(8), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_2KB, .dden = DEN_8Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 11,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), B(0), B(1), B(2),
+ R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7), R(8),
+ R(9), R(10), C(8), C(9), R(11), RS, C(11), R(12), R(13),
+ R(14), R(15), 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_4KB, .dden = DEN_4Gb, .dwid = X16,
+ .rowbits = 15, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), C(8), B(0), B(1),
+ B(2), R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7),
+ R(8), R(9), R(10), C(9), R(11), RS, R(12), R(13), R(14),
+ 0, 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_4KB, .dden = DEN_4Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), C(8), B(0), B(1),
+ B(2), R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7),
+ R(8), R(9), R(10), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_4KB, .dden = DEN_8Gb, .dwid = X16,
+ .rowbits = 16, .colbits = 10,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), C(8), B(0), B(1),
+ B(2), R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7),
+ R(8), R(9), R(10), C(9), R(11), RS, R(12), R(13), R(14),
+ R(15), 0, 0, 0
+ }
+ },
+ {
+ .addrdec = AMAP_4KB, .dden = DEN_8Gb, .dwid = X8,
+ .rowbits = 16, .colbits = 11,
+ .bits = {
+ C(2), C(3), C(4), C(5), C(6), C(7), C(8), B(0), B(1),
+ B(2), R(0), R(1), R(2), R(3), R(4), R(5), R(6), R(7),
+ R(8), R(9), R(10), C(9), R(11), RS, C(11), R(12), R(13),
+ R(14), R(15), 0, 0
+ }
+ }
+};
+
+static int bank_hash(u64 pmiaddr, int idx, int shft)
+{
+ int bhash = 0;
+
+ switch (idx) {
+ case 0:
+ bhash ^= ((pmiaddr >> (12 + shft)) ^ (pmiaddr >> (9 + shft))) & 1;
+ break;
+ case 1:
+ bhash ^= (((pmiaddr >> (10 + shft)) ^ (pmiaddr >> (8 + shft))) & 1) << 1;
+ bhash ^= ((pmiaddr >> 22) & 1) << 1;
+ break;
+ case 2:
+ bhash ^= (((pmiaddr >> (13 + shft)) ^ (pmiaddr >> (11 + shft))) & 1) << 2;
+ break;
+ }
+
+ return bhash;
+}
+
+static int rank_hash(u64 pmiaddr)
+{
+ return ((pmiaddr >> 16) ^ (pmiaddr >> 10)) & 1;
+}
+
+/* Second stage decode. Compute rank, bank, row & column. */
+static int apl_pmi2mem(struct mem_ctl_info *mci, u64 pmiaddr, u32 pmiidx,
+ struct dram_addr *daddr, char *msg)
+{
+ struct d_cr_drp0 *cr_drp0 = &drp0[pmiidx];
+ struct pnd2_pvt *pvt = mci->pvt_info;
+ int g = pvt->dimm_geom[pmiidx];
+ struct dimm_geometry *d = &dimms[g];
+ int column = 0, bank = 0, row = 0, rank = 0;
+ int i, idx, type, skiprs = 0;
+
+ for (i = 0; i < PMI_ADDRESS_WIDTH; i++) {
+ int bit = (pmiaddr >> i) & 1;
+
+ if (i + skiprs >= PMI_ADDRESS_WIDTH) {
+ snprintf(msg, PND2_MSG_SIZE, "Bad dimm_geometry[] table\n");
+ return -EINVAL;
+ }
+
+ type = d->bits[i + skiprs] & ~0xf;
+ idx = d->bits[i + skiprs] & 0xf;
+
+ /*
+ * On single rank DIMMs ignore the rank select bit
+ * and shift remainder of "bits[]" down one place.
+ */
+ if (type == RS && (cr_drp0->rken0 + cr_drp0->rken1) == 1) {
+ skiprs = 1;
+ type = d->bits[i + skiprs] & ~0xf;
+ idx = d->bits[i + skiprs] & 0xf;
+ }
+
+ switch (type) {
+ case C(0):
+ column |= (bit << idx);
+ break;
+ case B(0):
+ bank |= (bit << idx);
+ if (cr_drp0->bahen)
+ bank ^= bank_hash(pmiaddr, idx, d->addrdec);
+ break;
+ case R(0):
+ row |= (bit << idx);
+ break;
+ case RS:
+ rank = bit;
+ if (cr_drp0->rsien)
+ rank ^= rank_hash(pmiaddr);
+ break;
+ default:
+ if (bit) {
+ snprintf(msg, PND2_MSG_SIZE, "Bad translation\n");
+ return -EINVAL;
+ }
+ goto done;
+ }
+ }
+
+done:
+ daddr->col = column;
+ daddr->bank = bank;
+ daddr->row = row;
+ daddr->rank = rank;
+ daddr->dimm = 0;
+
+ return 0;
+}
+
+/* Pluck bit "in" from pmiaddr and return value shifted to bit "out" */
+#define dnv_get_bit(pmi, in, out) ((int)(((pmi) >> (in)) & 1u) << (out))
+
+static int dnv_pmi2mem(struct mem_ctl_info *mci, u64 pmiaddr, u32 pmiidx,
+ struct dram_addr *daddr, char *msg)
+{
+ /* Rank 0 or 1 */
+ daddr->rank = dnv_get_bit(pmiaddr, dmap[pmiidx].rs0 + 13, 0);
+ /* Rank 2 or 3 */
+ daddr->rank |= dnv_get_bit(pmiaddr, dmap[pmiidx].rs1 + 13, 1);
+
+ /*
+ * Normally ranks 0,1 are DIMM0, and 2,3 are DIMM1, but we
+ * flip them if DIMM1 is larger than DIMM0.
+ */
+ daddr->dimm = (daddr->rank >= 2) ^ drp[pmiidx].dimmflip;
+
+ daddr->bank = dnv_get_bit(pmiaddr, dmap[pmiidx].ba0 + 6, 0);
+ daddr->bank |= dnv_get_bit(pmiaddr, dmap[pmiidx].ba1 + 6, 1);
+ daddr->bank |= dnv_get_bit(pmiaddr, dmap[pmiidx].bg0 + 6, 2);
+ if (dsch.ddr4en)
+ daddr->bank |= dnv_get_bit(pmiaddr, dmap[pmiidx].bg1 + 6, 3);
+ if (dmap1[pmiidx].bxor) {
+ if (dsch.ddr4en) {
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap3[pmiidx].row6 + 6, 0);
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap3[pmiidx].row7 + 6, 1);
+ if (dsch.chan_width == 0)
+ /* 64/72 bit dram channel width */
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca3 + 6, 2);
+ else
+ /* 32/40 bit dram channel width */
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca4 + 6, 2);
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap2[pmiidx].row2 + 6, 3);
+ } else {
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap2[pmiidx].row2 + 6, 0);
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap3[pmiidx].row6 + 6, 1);
+ if (dsch.chan_width == 0)
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca3 + 6, 2);
+ else
+ daddr->bank ^= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca4 + 6, 2);
+ }
+ }
+
+ daddr->row = dnv_get_bit(pmiaddr, dmap2[pmiidx].row0 + 6, 0);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap2[pmiidx].row1 + 6, 1);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap2[pmiidx].row2 + 6, 2);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap2[pmiidx].row3 + 6, 3);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap2[pmiidx].row4 + 6, 4);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap2[pmiidx].row5 + 6, 5);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row6 + 6, 6);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row7 + 6, 7);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row8 + 6, 8);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row9 + 6, 9);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row10 + 6, 10);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap3[pmiidx].row11 + 6, 11);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row12 + 6, 12);
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row13 + 6, 13);
+ if (dmap4[pmiidx].row14 != 31)
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row14 + 6, 14);
+ if (dmap4[pmiidx].row15 != 31)
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row15 + 6, 15);
+ if (dmap4[pmiidx].row16 != 31)
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row16 + 6, 16);
+ if (dmap4[pmiidx].row17 != 31)
+ daddr->row |= dnv_get_bit(pmiaddr, dmap4[pmiidx].row17 + 6, 17);
+
+ daddr->col = dnv_get_bit(pmiaddr, dmap5[pmiidx].ca3 + 6, 3);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca4 + 6, 4);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca5 + 6, 5);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca6 + 6, 6);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca7 + 6, 7);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca8 + 6, 8);
+ daddr->col |= dnv_get_bit(pmiaddr, dmap5[pmiidx].ca9 + 6, 9);
+ if (!dsch.ddr4en && dmap1[pmiidx].ca11 != 0x3f)
+ daddr->col |= dnv_get_bit(pmiaddr, dmap1[pmiidx].ca11 + 13, 11);
+
+ return 0;
+}
+
+static int check_channel(int ch)
+{
+ if (drp0[ch].dramtype != 0) {
+ pnd2_printk(KERN_INFO, "Unsupported DIMM in channel %d\n", ch);
+ return 1;
+ } else if (drp0[ch].eccen == 0) {
+ pnd2_printk(KERN_INFO, "ECC disabled on channel %d\n", ch);
+ return 1;
+ }
+ return 0;
+}
+
+static int apl_check_ecc_active(void)
+{
+ int i, ret = 0;
+
+ /* Check dramtype and ECC mode for each present DIMM */
+ for (i = 0; i < APL_NUM_CHANNELS; i++)
+ if (chan_mask & BIT(i))
+ ret += check_channel(i);
+ return ret ? -EINVAL : 0;
+}
+
+#define DIMMS_PRESENT(d) ((d)->rken0 + (d)->rken1 + (d)->rken2 + (d)->rken3)
+
+static int check_unit(int ch)
+{
+ struct d_cr_drp *d = &drp[ch];
+
+ if (DIMMS_PRESENT(d) && !ecc_ctrl[ch].eccen) {
+ pnd2_printk(KERN_INFO, "ECC disabled on channel %d\n", ch);
+ return 1;
+ }
+ return 0;
+}
+
+static int dnv_check_ecc_active(void)
+{
+ int i, ret = 0;
+
+ for (i = 0; i < DNV_NUM_CHANNELS; i++)
+ ret += check_unit(i);
+ return ret ? -EINVAL : 0;
+}
+
+static int get_memory_error_data(struct mem_ctl_info *mci, u64 addr,
+ struct dram_addr *daddr, char *msg)
+{
+ u64 pmiaddr;
+ u32 pmiidx;
+ int ret;
+
+ ret = sys2pmi(addr, &pmiidx, &pmiaddr, msg);
+ if (ret)
+ return ret;
+
+ pmiaddr >>= ops->pmiaddr_shift;
+ /* pmi channel idx to dimm channel idx */
+ pmiidx >>= ops->pmiidx_shift;
+ daddr->chan = pmiidx;
+
+ ret = ops->pmi2mem(mci, pmiaddr, pmiidx, daddr, msg);
+ if (ret)
+ return ret;
+
+ edac_dbg(0, "SysAddr=%llx PmiAddr=%llx Channel=%d DIMM=%d Rank=%d Bank=%d Row=%d Column=%d\n",
+ addr, pmiaddr, daddr->chan, daddr->dimm, daddr->rank, daddr->bank, daddr->row, daddr->col);
+
+ return 0;
+}
+
+static void pnd2_mce_output_error(struct mem_ctl_info *mci, const struct mce *m,
+ struct dram_addr *daddr)
+{
+ enum hw_event_mc_err_type tp_event;
+ char *optype, msg[PND2_MSG_SIZE];
+ bool ripv = m->mcgstatus & MCG_STATUS_RIPV;
+ bool overflow = m->status & MCI_STATUS_OVER;
+ bool uc_err = m->status & MCI_STATUS_UC;
+ bool recov = m->status & MCI_STATUS_S;
+ u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52);
+ u32 mscod = GET_BITFIELD(m->status, 16, 31);
+ u32 errcode = GET_BITFIELD(m->status, 0, 15);
+ u32 optypenum = GET_BITFIELD(m->status, 4, 6);
+ int rc;
+
+ tp_event = uc_err ? (ripv ? HW_EVENT_ERR_UNCORRECTED : HW_EVENT_ERR_FATAL) :
+ HW_EVENT_ERR_CORRECTED;
+
+ /*
+ * According with Table 15-9 of the Intel Architecture spec vol 3A,
+ * memory errors should fit in this mask:
+ * 000f 0000 1mmm cccc (binary)
+ * where:
+ * f = Correction Report Filtering Bit. If 1, subsequent errors
+ * won't be shown
+ * mmm = error type
+ * cccc = channel
+ * If the mask doesn't match, report an error to the parsing logic
+ */
+ if (!((errcode & 0xef80) == 0x80)) {
+ optype = "Can't parse: it is not a mem";
+ } else {
+ switch (optypenum) {
+ case 0:
+ optype = "generic undef request error";
+ break;
+ case 1:
+ optype = "memory read error";
+ break;
+ case 2:
+ optype = "memory write error";
+ break;
+ case 3:
+ optype = "addr/cmd error";
+ break;
+ case 4:
+ optype = "memory scrubbing error";
+ break;
+ default:
+ optype = "reserved";
+ break;
+ }
+ }
+
+ /* Only decode errors with an valid address (ADDRV) */
+ if (!(m->status & MCI_STATUS_ADDRV))
+ return;
+
+ rc = get_memory_error_data(mci, m->addr, daddr, msg);
+ if (rc)
+ goto address_error;
+
+ snprintf(msg, sizeof(msg),
+ "%s%s err_code:%04x:%04x channel:%d DIMM:%d rank:%d row:%d bank:%d col:%d",
+ overflow ? " OVERFLOW" : "", (uc_err && recov) ? " recoverable" : "", mscod,
+ errcode, daddr->chan, daddr->dimm, daddr->rank, daddr->row, daddr->bank, daddr->col);
+
+ edac_dbg(0, "%s\n", msg);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(tp_event, mci, core_err_cnt, m->addr >> PAGE_SHIFT,
+ m->addr & ~PAGE_MASK, 0, daddr->chan, daddr->dimm, -1, optype, msg);
+
+ return;
+
+address_error:
+ edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0, -1, -1, -1, msg, "");
+}
+
+static void apl_get_dimm_config(struct mem_ctl_info *mci)
+{
+ struct pnd2_pvt *pvt = mci->pvt_info;
+ struct dimm_info *dimm;
+ struct d_cr_drp0 *d;
+ u64 capacity;
+ int i, g;
+
+ for (i = 0; i < APL_NUM_CHANNELS; i++) {
+ if (!(chan_mask & BIT(i)))
+ continue;
+
+ dimm = edac_get_dimm(mci, i, 0, 0);
+ if (!dimm) {
+ edac_dbg(0, "No allocated DIMM for channel %d\n", i);
+ continue;
+ }
+
+ d = &drp0[i];
+ for (g = 0; g < ARRAY_SIZE(dimms); g++)
+ if (dimms[g].addrdec == d->addrdec &&
+ dimms[g].dden == d->dden &&
+ dimms[g].dwid == d->dwid)
+ break;
+
+ if (g == ARRAY_SIZE(dimms)) {
+ edac_dbg(0, "Channel %d: unrecognized DIMM\n", i);
+ continue;
+ }
+
+ pvt->dimm_geom[i] = g;
+ capacity = (d->rken0 + d->rken1) * 8 * (1ul << dimms[g].rowbits) *
+ (1ul << dimms[g].colbits);
+ edac_dbg(0, "Channel %d: %lld MByte DIMM\n", i, capacity >> (20 - 3));
+ dimm->nr_pages = MiB_TO_PAGES(capacity >> (20 - 3));
+ dimm->grain = 32;
+ dimm->dtype = (d->dwid == 0) ? DEV_X8 : DEV_X16;
+ dimm->mtype = MEM_DDR3;
+ dimm->edac_mode = EDAC_SECDED;
+ snprintf(dimm->label, sizeof(dimm->label), "Slice#%d_Chan#%d", i / 2, i % 2);
+ }
+}
+
+static const int dnv_dtypes[] = {
+ DEV_X8, DEV_X4, DEV_X16, DEV_UNKNOWN
+};
+
+static void dnv_get_dimm_config(struct mem_ctl_info *mci)
+{
+ int i, j, ranks_of_dimm[DNV_MAX_DIMMS], banks, rowbits, colbits, memtype;
+ struct dimm_info *dimm;
+ struct d_cr_drp *d;
+ u64 capacity;
+
+ if (dsch.ddr4en) {
+ memtype = MEM_DDR4;
+ banks = 16;
+ colbits = 10;
+ } else {
+ memtype = MEM_DDR3;
+ banks = 8;
+ }
+
+ for (i = 0; i < DNV_NUM_CHANNELS; i++) {
+ if (dmap4[i].row14 == 31)
+ rowbits = 14;
+ else if (dmap4[i].row15 == 31)
+ rowbits = 15;
+ else if (dmap4[i].row16 == 31)
+ rowbits = 16;
+ else if (dmap4[i].row17 == 31)
+ rowbits = 17;
+ else
+ rowbits = 18;
+
+ if (memtype == MEM_DDR3) {
+ if (dmap1[i].ca11 != 0x3f)
+ colbits = 12;
+ else
+ colbits = 10;
+ }
+
+ d = &drp[i];
+ /* DIMM0 is present if rank0 and/or rank1 is enabled */
+ ranks_of_dimm[0] = d->rken0 + d->rken1;
+ /* DIMM1 is present if rank2 and/or rank3 is enabled */
+ ranks_of_dimm[1] = d->rken2 + d->rken3;
+
+ for (j = 0; j < DNV_MAX_DIMMS; j++) {
+ if (!ranks_of_dimm[j])
+ continue;
+
+ dimm = edac_get_dimm(mci, i, j, 0);
+ if (!dimm) {
+ edac_dbg(0, "No allocated DIMM for channel %d DIMM %d\n", i, j);
+ continue;
+ }
+
+ capacity = ranks_of_dimm[j] * banks * (1ul << rowbits) * (1ul << colbits);
+ edac_dbg(0, "Channel %d DIMM %d: %lld MByte DIMM\n", i, j, capacity >> (20 - 3));
+ dimm->nr_pages = MiB_TO_PAGES(capacity >> (20 - 3));
+ dimm->grain = 32;
+ dimm->dtype = dnv_dtypes[j ? d->dimmdwid0 : d->dimmdwid1];
+ dimm->mtype = memtype;
+ dimm->edac_mode = EDAC_SECDED;
+ snprintf(dimm->label, sizeof(dimm->label), "Chan#%d_DIMM#%d", i, j);
+ }
+ }
+}
+
+static int pnd2_register_mci(struct mem_ctl_info **ppmci)
+{
+ struct edac_mc_layer layers[2];
+ struct mem_ctl_info *mci;
+ struct pnd2_pvt *pvt;
+ int rc;
+
+ rc = ops->check_ecc();
+ if (rc < 0)
+ return rc;
+
+ /* Allocate a new MC control structure */
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = ops->channels;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = ops->dimms_per_channel;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
+ if (!mci)
+ return -ENOMEM;
+
+ pvt = mci->pvt_info;
+ memset(pvt, 0, sizeof(*pvt));
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->dev_name = ops->name;
+ mci->ctl_name = "Pondicherry2";
+
+ /* Get dimm basic config and the memory layout */
+ ops->get_dimm_config(mci);
+
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ edac_mc_free(mci);
+ return -EINVAL;
+ }
+
+ *ppmci = mci;
+
+ return 0;
+}
+
+static void pnd2_unregister_mci(struct mem_ctl_info *mci)
+{
+ if (unlikely(!mci || !mci->pvt_info)) {
+ pnd2_printk(KERN_ERR, "Couldn't find mci handler\n");
+ return;
+ }
+
+ /* Remove MC sysfs nodes */
+ edac_mc_del_mc(NULL);
+ edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
+ edac_mc_free(mci);
+}
+
+/*
+ * Callback function registered with core kernel mce code.
+ * Called once for each logged error.
+ */
+static int pnd2_mce_check_error(struct notifier_block *nb, unsigned long val, void *data)
+{
+ struct mce *mce = (struct mce *)data;
+ struct mem_ctl_info *mci;
+ struct dram_addr daddr;
+ char *type;
+
+ mci = pnd2_mci;
+ if (!mci || (mce->kflags & MCE_HANDLED_CEC))
+ return NOTIFY_DONE;
+
+ /*
+ * Just let mcelog handle it if the error is
+ * outside the memory controller. A memory error
+ * is indicated by bit 7 = 1 and bits = 8-11,13-15 = 0.
+ * bit 12 has an special meaning.
+ */
+ if ((mce->status & 0xefff) >> 7 != 1)
+ return NOTIFY_DONE;
+
+ if (mce->mcgstatus & MCG_STATUS_MCIP)
+ type = "Exception";
+ else
+ type = "Event";
+
+ pnd2_mc_printk(mci, KERN_INFO, "HANDLING MCE MEMORY ERROR\n");
+ pnd2_mc_printk(mci, KERN_INFO, "CPU %u: Machine Check %s: %llx Bank %u: %llx\n",
+ mce->extcpu, type, mce->mcgstatus, mce->bank, mce->status);
+ pnd2_mc_printk(mci, KERN_INFO, "TSC %llx ", mce->tsc);
+ pnd2_mc_printk(mci, KERN_INFO, "ADDR %llx ", mce->addr);
+ pnd2_mc_printk(mci, KERN_INFO, "MISC %llx ", mce->misc);
+ pnd2_mc_printk(mci, KERN_INFO, "PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n",
+ mce->cpuvendor, mce->cpuid, mce->time, mce->socketid, mce->apicid);
+
+ pnd2_mce_output_error(mci, mce, &daddr);
+
+ /* Advice mcelog that the error were handled */
+ mce->kflags |= MCE_HANDLED_EDAC;
+ return NOTIFY_OK;
+}
+
+static struct notifier_block pnd2_mce_dec = {
+ .notifier_call = pnd2_mce_check_error,
+ .priority = MCE_PRIO_EDAC,
+};
+
+#ifdef CONFIG_EDAC_DEBUG
+/*
+ * Write an address to this file to exercise the address decode
+ * logic in this driver.
+ */
+static u64 pnd2_fake_addr;
+#define PND2_BLOB_SIZE 1024
+static char pnd2_result[PND2_BLOB_SIZE];
+static struct dentry *pnd2_test;
+static struct debugfs_blob_wrapper pnd2_blob = {
+ .data = pnd2_result,
+ .size = 0
+};
+
+static int debugfs_u64_set(void *data, u64 val)
+{
+ struct dram_addr daddr;
+ struct mce m;
+
+ *(u64 *)data = val;
+ m.mcgstatus = 0;
+ /* ADDRV + MemRd + Unknown channel */
+ m.status = MCI_STATUS_ADDRV + 0x9f;
+ m.addr = val;
+ pnd2_mce_output_error(pnd2_mci, &m, &daddr);
+ snprintf(pnd2_blob.data, PND2_BLOB_SIZE,
+ "SysAddr=%llx Channel=%d DIMM=%d Rank=%d Bank=%d Row=%d Column=%d\n",
+ m.addr, daddr.chan, daddr.dimm, daddr.rank, daddr.bank, daddr.row, daddr.col);
+ pnd2_blob.size = strlen(pnd2_blob.data);
+
+ return 0;
+}
+DEFINE_DEBUGFS_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n");
+
+static void setup_pnd2_debug(void)
+{
+ pnd2_test = edac_debugfs_create_dir("pnd2_test");
+ edac_debugfs_create_file("pnd2_debug_addr", 0200, pnd2_test,
+ &pnd2_fake_addr, &fops_u64_wo);
+ debugfs_create_blob("pnd2_debug_results", 0400, pnd2_test, &pnd2_blob);
+}
+
+static void teardown_pnd2_debug(void)
+{
+ debugfs_remove_recursive(pnd2_test);
+}
+#else
+static void setup_pnd2_debug(void) {}
+static void teardown_pnd2_debug(void) {}
+#endif /* CONFIG_EDAC_DEBUG */
+
+
+static int pnd2_probe(void)
+{
+ int rc;
+
+ edac_dbg(2, "\n");
+ rc = get_registers();
+ if (rc)
+ return rc;
+
+ return pnd2_register_mci(&pnd2_mci);
+}
+
+static void pnd2_remove(void)
+{
+ edac_dbg(0, "\n");
+ pnd2_unregister_mci(pnd2_mci);
+}
+
+static struct dunit_ops apl_ops = {
+ .name = "pnd2/apl",
+ .type = APL,
+ .pmiaddr_shift = LOG2_PMI_ADDR_GRANULARITY,
+ .pmiidx_shift = 0,
+ .channels = APL_NUM_CHANNELS,
+ .dimms_per_channel = 1,
+ .rd_reg = apl_rd_reg,
+ .get_registers = apl_get_registers,
+ .check_ecc = apl_check_ecc_active,
+ .mk_region = apl_mk_region,
+ .get_dimm_config = apl_get_dimm_config,
+ .pmi2mem = apl_pmi2mem,
+};
+
+static struct dunit_ops dnv_ops = {
+ .name = "pnd2/dnv",
+ .type = DNV,
+ .pmiaddr_shift = 0,
+ .pmiidx_shift = 1,
+ .channels = DNV_NUM_CHANNELS,
+ .dimms_per_channel = 2,
+ .rd_reg = dnv_rd_reg,
+ .get_registers = dnv_get_registers,
+ .check_ecc = dnv_check_ecc_active,
+ .mk_region = dnv_mk_region,
+ .get_dimm_config = dnv_get_dimm_config,
+ .pmi2mem = dnv_pmi2mem,
+};
+
+static const struct x86_cpu_id pnd2_cpuids[] = {
+ X86_MATCH_INTEL_FAM6_MODEL(ATOM_GOLDMONT, &apl_ops),
+ X86_MATCH_INTEL_FAM6_MODEL(ATOM_GOLDMONT_D, &dnv_ops),
+ { }
+};
+MODULE_DEVICE_TABLE(x86cpu, pnd2_cpuids);
+
+static int __init pnd2_init(void)
+{
+ const struct x86_cpu_id *id;
+ const char *owner;
+ int rc;
+
+ edac_dbg(2, "\n");
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
+ return -ENODEV;
+
+ id = x86_match_cpu(pnd2_cpuids);
+ if (!id)
+ return -ENODEV;
+
+ ops = (struct dunit_ops *)id->driver_data;
+
+ if (ops->type == APL) {
+ p2sb_bus = pci_find_bus(0, 0);
+ if (!p2sb_bus)
+ return -ENODEV;
+ }
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ rc = pnd2_probe();
+ if (rc < 0) {
+ pnd2_printk(KERN_ERR, "Failed to register device with error %d.\n", rc);
+ return rc;
+ }
+
+ if (!pnd2_mci)
+ return -ENODEV;
+
+ mce_register_decode_chain(&pnd2_mce_dec);
+ setup_pnd2_debug();
+
+ return 0;
+}
+
+static void __exit pnd2_exit(void)
+{
+ edac_dbg(2, "\n");
+ teardown_pnd2_debug();
+ mce_unregister_decode_chain(&pnd2_mce_dec);
+ pnd2_remove();
+}
+
+module_init(pnd2_init);
+module_exit(pnd2_exit);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Tony Luck");
+MODULE_DESCRIPTION("MC Driver for Intel SoC using Pondicherry memory controller");
diff --git a/drivers/edac/pnd2_edac.h b/drivers/edac/pnd2_edac.h
new file mode 100644
index 0000000000..028ef70170
--- /dev/null
+++ b/drivers/edac/pnd2_edac.h
@@ -0,0 +1,293 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Register bitfield descriptions for Pondicherry2 memory controller.
+ *
+ * Copyright (c) 2016, Intel Corporation.
+ */
+
+#ifndef _PND2_REGS_H
+#define _PND2_REGS_H
+
+struct b_cr_touud_lo_pci {
+ u32 lock : 1;
+ u32 reserved_1 : 19;
+ u32 touud : 12;
+};
+
+#define b_cr_touud_lo_pci_port 0x4c
+#define b_cr_touud_lo_pci_offset 0xa8
+#define b_cr_touud_lo_pci_r_opcode 0x04
+
+struct b_cr_touud_hi_pci {
+ u32 touud : 7;
+ u32 reserved_0 : 25;
+};
+
+#define b_cr_touud_hi_pci_port 0x4c
+#define b_cr_touud_hi_pci_offset 0xac
+#define b_cr_touud_hi_pci_r_opcode 0x04
+
+struct b_cr_tolud_pci {
+ u32 lock : 1;
+ u32 reserved_0 : 19;
+ u32 tolud : 12;
+};
+
+#define b_cr_tolud_pci_port 0x4c
+#define b_cr_tolud_pci_offset 0xbc
+#define b_cr_tolud_pci_r_opcode 0x04
+
+struct b_cr_mchbar_lo_pci {
+ u32 enable : 1;
+ u32 pad_3_1 : 3;
+ u32 pad_14_4: 11;
+ u32 base: 17;
+};
+
+struct b_cr_mchbar_hi_pci {
+ u32 base : 7;
+ u32 pad_31_7 : 25;
+};
+
+/* Symmetric region */
+struct b_cr_slice_channel_hash {
+ u64 slice_1_disabled : 1;
+ u64 hvm_mode : 1;
+ u64 interleave_mode : 2;
+ u64 slice_0_mem_disabled : 1;
+ u64 reserved_0 : 1;
+ u64 slice_hash_mask : 14;
+ u64 reserved_1 : 11;
+ u64 enable_pmi_dual_data_mode : 1;
+ u64 ch_1_disabled : 1;
+ u64 reserved_2 : 1;
+ u64 sym_slice0_channel_enabled : 2;
+ u64 sym_slice1_channel_enabled : 2;
+ u64 ch_hash_mask : 14;
+ u64 reserved_3 : 11;
+ u64 lock : 1;
+};
+
+#define b_cr_slice_channel_hash_port 0x4c
+#define b_cr_slice_channel_hash_offset 0x4c58
+#define b_cr_slice_channel_hash_r_opcode 0x06
+
+struct b_cr_mot_out_base_mchbar {
+ u32 reserved_0 : 14;
+ u32 mot_out_base : 15;
+ u32 reserved_1 : 1;
+ u32 tr_en : 1;
+ u32 imr_en : 1;
+};
+
+#define b_cr_mot_out_base_mchbar_port 0x4c
+#define b_cr_mot_out_base_mchbar_offset 0x6af0
+#define b_cr_mot_out_base_mchbar_r_opcode 0x00
+
+struct b_cr_mot_out_mask_mchbar {
+ u32 reserved_0 : 14;
+ u32 mot_out_mask : 15;
+ u32 reserved_1 : 1;
+ u32 ia_iwb_en : 1;
+ u32 gt_iwb_en : 1;
+};
+
+#define b_cr_mot_out_mask_mchbar_port 0x4c
+#define b_cr_mot_out_mask_mchbar_offset 0x6af4
+#define b_cr_mot_out_mask_mchbar_r_opcode 0x00
+
+struct b_cr_asym_mem_region0_mchbar {
+ u32 pad : 4;
+ u32 slice0_asym_base : 11;
+ u32 pad_18_15 : 4;
+ u32 slice0_asym_limit : 11;
+ u32 slice0_asym_channel_select : 1;
+ u32 slice0_asym_enable : 1;
+};
+
+#define b_cr_asym_mem_region0_mchbar_port 0x4c
+#define b_cr_asym_mem_region0_mchbar_offset 0x6e40
+#define b_cr_asym_mem_region0_mchbar_r_opcode 0x00
+
+struct b_cr_asym_mem_region1_mchbar {
+ u32 pad : 4;
+ u32 slice1_asym_base : 11;
+ u32 pad_18_15 : 4;
+ u32 slice1_asym_limit : 11;
+ u32 slice1_asym_channel_select : 1;
+ u32 slice1_asym_enable : 1;
+};
+
+#define b_cr_asym_mem_region1_mchbar_port 0x4c
+#define b_cr_asym_mem_region1_mchbar_offset 0x6e44
+#define b_cr_asym_mem_region1_mchbar_r_opcode 0x00
+
+/* Some bit fields moved in above two structs on Denverton */
+struct b_cr_asym_mem_region_denverton {
+ u32 pad : 4;
+ u32 slice_asym_base : 8;
+ u32 pad_19_12 : 8;
+ u32 slice_asym_limit : 8;
+ u32 pad_28_30 : 3;
+ u32 slice_asym_enable : 1;
+};
+
+struct b_cr_asym_2way_mem_region_mchbar {
+ u32 pad : 2;
+ u32 asym_2way_intlv_mode : 2;
+ u32 asym_2way_base : 11;
+ u32 pad_16_15 : 2;
+ u32 asym_2way_limit : 11;
+ u32 pad_30_28 : 3;
+ u32 asym_2way_interleave_enable : 1;
+};
+
+#define b_cr_asym_2way_mem_region_mchbar_port 0x4c
+#define b_cr_asym_2way_mem_region_mchbar_offset 0x6e50
+#define b_cr_asym_2way_mem_region_mchbar_r_opcode 0x00
+
+/* Apollo Lake d-unit */
+
+struct d_cr_drp0 {
+ u32 rken0 : 1;
+ u32 rken1 : 1;
+ u32 ddmen : 1;
+ u32 rsvd3 : 1;
+ u32 dwid : 2;
+ u32 dden : 3;
+ u32 rsvd13_9 : 5;
+ u32 rsien : 1;
+ u32 bahen : 1;
+ u32 rsvd18_16 : 3;
+ u32 caswizzle : 2;
+ u32 eccen : 1;
+ u32 dramtype : 3;
+ u32 blmode : 3;
+ u32 addrdec : 2;
+ u32 dramdevice_pr : 2;
+};
+
+#define d_cr_drp0_offset 0x1400
+#define d_cr_drp0_r_opcode 0x00
+
+/* Denverton d-unit */
+
+struct d_cr_dsch {
+ u32 ch0en : 1;
+ u32 ch1en : 1;
+ u32 ddr4en : 1;
+ u32 coldwake : 1;
+ u32 newbypdis : 1;
+ u32 chan_width : 1;
+ u32 rsvd6_6 : 1;
+ u32 ooodis : 1;
+ u32 rsvd18_8 : 11;
+ u32 ic : 1;
+ u32 rsvd31_20 : 12;
+};
+
+#define d_cr_dsch_port 0x16
+#define d_cr_dsch_offset 0x0
+#define d_cr_dsch_r_opcode 0x0
+
+struct d_cr_ecc_ctrl {
+ u32 eccen : 1;
+ u32 rsvd31_1 : 31;
+};
+
+#define d_cr_ecc_ctrl_offset 0x180
+#define d_cr_ecc_ctrl_r_opcode 0x0
+
+struct d_cr_drp {
+ u32 rken0 : 1;
+ u32 rken1 : 1;
+ u32 rken2 : 1;
+ u32 rken3 : 1;
+ u32 dimmdwid0 : 2;
+ u32 dimmdden0 : 2;
+ u32 dimmdwid1 : 2;
+ u32 dimmdden1 : 2;
+ u32 rsvd15_12 : 4;
+ u32 dimmflip : 1;
+ u32 rsvd31_17 : 15;
+};
+
+#define d_cr_drp_offset 0x158
+#define d_cr_drp_r_opcode 0x0
+
+struct d_cr_dmap {
+ u32 ba0 : 5;
+ u32 ba1 : 5;
+ u32 bg0 : 5; /* if ddr3, ba2 = bg0 */
+ u32 bg1 : 5; /* if ddr3, ba3 = bg1 */
+ u32 rs0 : 5;
+ u32 rs1 : 5;
+ u32 rsvd : 2;
+};
+
+#define d_cr_dmap_offset 0x174
+#define d_cr_dmap_r_opcode 0x0
+
+struct d_cr_dmap1 {
+ u32 ca11 : 6;
+ u32 bxor : 1;
+ u32 rsvd : 25;
+};
+
+#define d_cr_dmap1_offset 0xb4
+#define d_cr_dmap1_r_opcode 0x0
+
+struct d_cr_dmap2 {
+ u32 row0 : 5;
+ u32 row1 : 5;
+ u32 row2 : 5;
+ u32 row3 : 5;
+ u32 row4 : 5;
+ u32 row5 : 5;
+ u32 rsvd : 2;
+};
+
+#define d_cr_dmap2_offset 0x148
+#define d_cr_dmap2_r_opcode 0x0
+
+struct d_cr_dmap3 {
+ u32 row6 : 5;
+ u32 row7 : 5;
+ u32 row8 : 5;
+ u32 row9 : 5;
+ u32 row10 : 5;
+ u32 row11 : 5;
+ u32 rsvd : 2;
+};
+
+#define d_cr_dmap3_offset 0x14c
+#define d_cr_dmap3_r_opcode 0x0
+
+struct d_cr_dmap4 {
+ u32 row12 : 5;
+ u32 row13 : 5;
+ u32 row14 : 5;
+ u32 row15 : 5;
+ u32 row16 : 5;
+ u32 row17 : 5;
+ u32 rsvd : 2;
+};
+
+#define d_cr_dmap4_offset 0x150
+#define d_cr_dmap4_r_opcode 0x0
+
+struct d_cr_dmap5 {
+ u32 ca3 : 4;
+ u32 ca4 : 4;
+ u32 ca5 : 4;
+ u32 ca6 : 4;
+ u32 ca7 : 4;
+ u32 ca8 : 4;
+ u32 ca9 : 4;
+ u32 rsvd : 4;
+};
+
+#define d_cr_dmap5_offset 0x154
+#define d_cr_dmap5_r_opcode 0x0
+
+#endif /* _PND2_REGS_H */
diff --git a/drivers/edac/ppc4xx_edac.c b/drivers/edac/ppc4xx_edac.c
new file mode 100644
index 0000000000..046969b4e8
--- /dev/null
+++ b/drivers/edac/ppc4xx_edac.c
@@ -0,0 +1,1428 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2008 Nuovation System Designs, LLC
+ * Grant Erickson <gerickson@nuovations.com>
+ */
+
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/of_irq.h>
+#include <linux/of_platform.h>
+#include <linux/types.h>
+
+#include <asm/dcr.h>
+
+#include "edac_module.h"
+#include "ppc4xx_edac.h"
+
+/*
+ * This file implements a driver for monitoring and handling events
+ * associated with the IMB DDR2 ECC controller found in the AMCC/IBM
+ * 405EX[r], 440SP, 440SPe, 460EX, 460GT and 460SX.
+ *
+ * As realized in the 405EX[r], this controller features:
+ *
+ * - Support for registered- and non-registered DDR1 and DDR2 memory.
+ * - 32-bit or 16-bit memory interface with optional ECC.
+ *
+ * o ECC support includes:
+ *
+ * - 4-bit SEC/DED
+ * - Aligned-nibble error detect
+ * - Bypass mode
+ *
+ * - Two (2) memory banks/ranks.
+ * - Up to 1 GiB per bank/rank in 32-bit mode and up to 512 MiB per
+ * bank/rank in 16-bit mode.
+ *
+ * As realized in the 440SP and 440SPe, this controller changes/adds:
+ *
+ * - 64-bit or 32-bit memory interface with optional ECC.
+ *
+ * o ECC support includes:
+ *
+ * - 8-bit SEC/DED
+ * - Aligned-nibble error detect
+ * - Bypass mode
+ *
+ * - Up to 4 GiB per bank/rank in 64-bit mode and up to 2 GiB
+ * per bank/rank in 32-bit mode.
+ *
+ * As realized in the 460EX and 460GT, this controller changes/adds:
+ *
+ * - 64-bit or 32-bit memory interface with optional ECC.
+ *
+ * o ECC support includes:
+ *
+ * - 8-bit SEC/DED
+ * - Aligned-nibble error detect
+ * - Bypass mode
+ *
+ * - Four (4) memory banks/ranks.
+ * - Up to 16 GiB per bank/rank in 64-bit mode and up to 8 GiB
+ * per bank/rank in 32-bit mode.
+ *
+ * At present, this driver has ONLY been tested against the controller
+ * realization in the 405EX[r] on the AMCC Kilauea and Haleakala
+ * boards (256 MiB w/o ECC memory soldered onto the board) and a
+ * proprietary board based on those designs (128 MiB ECC memory, also
+ * soldered onto the board).
+ *
+ * Dynamic feature detection and handling needs to be added for the
+ * other realizations of this controller listed above.
+ *
+ * Eventually, this driver will likely be adapted to the above variant
+ * realizations of this controller as well as broken apart to handle
+ * the other known ECC-capable controllers prevalent in other 4xx
+ * processors:
+ *
+ * - IBM SDRAM (405GP, 405CR and 405EP) "ibm,sdram-4xx"
+ * - IBM DDR1 (440GP, 440GX, 440EP and 440GR) "ibm,sdram-4xx-ddr"
+ * - Denali DDR1/DDR2 (440EPX and 440GRX) "denali,sdram-4xx-ddr2"
+ *
+ * For this controller, unfortunately, correctable errors report
+ * nothing more than the beat/cycle and byte/lane the correction
+ * occurred on and the check bit group that covered the error.
+ *
+ * In contrast, uncorrectable errors also report the failing address,
+ * the bus master and the transaction direction (i.e. read or write)
+ *
+ * Regardless of whether the error is a CE or a UE, we report the
+ * following pieces of information in the driver-unique message to the
+ * EDAC subsystem:
+ *
+ * - Device tree path
+ * - Bank(s)
+ * - Check bit error group
+ * - Beat(s)/lane(s)
+ */
+
+/* Preprocessor Definitions */
+
+#define EDAC_OPSTATE_INT_STR "interrupt"
+#define EDAC_OPSTATE_POLL_STR "polled"
+#define EDAC_OPSTATE_UNKNOWN_STR "unknown"
+
+#define PPC4XX_EDAC_MODULE_NAME "ppc4xx_edac"
+#define PPC4XX_EDAC_MODULE_REVISION "v1.0.0"
+
+#define PPC4XX_EDAC_MESSAGE_SIZE 256
+
+/*
+ * Kernel logging without an EDAC instance
+ */
+#define ppc4xx_edac_printk(level, fmt, arg...) \
+ edac_printk(level, "PPC4xx MC", fmt, ##arg)
+
+/*
+ * Kernel logging with an EDAC instance
+ */
+#define ppc4xx_edac_mc_printk(level, mci, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "PPC4xx", fmt, ##arg)
+
+/*
+ * Macros to convert bank configuration size enumerations into MiB and
+ * page values.
+ */
+#define SDRAM_MBCF_SZ_MiB_MIN 4
+#define SDRAM_MBCF_SZ_TO_MiB(n) (SDRAM_MBCF_SZ_MiB_MIN \
+ << (SDRAM_MBCF_SZ_DECODE(n)))
+#define SDRAM_MBCF_SZ_TO_PAGES(n) (SDRAM_MBCF_SZ_MiB_MIN \
+ << (20 - PAGE_SHIFT + \
+ SDRAM_MBCF_SZ_DECODE(n)))
+
+/*
+ * The ibm,sdram-4xx-ddr2 Device Control Registers (DCRs) are
+ * indirectly accessed and have a base and length defined by the
+ * device tree. The base can be anything; however, we expect the
+ * length to be precisely two registers, the first for the address
+ * window and the second for the data window.
+ */
+#define SDRAM_DCR_RESOURCE_LEN 2
+#define SDRAM_DCR_ADDR_OFFSET 0
+#define SDRAM_DCR_DATA_OFFSET 1
+
+/*
+ * Device tree interrupt indices
+ */
+#define INTMAP_ECCDED_INDEX 0 /* Double-bit Error Detect */
+#define INTMAP_ECCSEC_INDEX 1 /* Single-bit Error Correct */
+
+/* Type Definitions */
+
+/*
+ * PPC4xx SDRAM memory controller private instance data
+ */
+struct ppc4xx_edac_pdata {
+ dcr_host_t dcr_host; /* Indirect DCR address/data window mapping */
+ struct {
+ int sec; /* Single-bit correctable error IRQ assigned */
+ int ded; /* Double-bit detectable error IRQ assigned */
+ } irqs;
+};
+
+/*
+ * Various status data gathered and manipulated when checking and
+ * reporting ECC status.
+ */
+struct ppc4xx_ecc_status {
+ u32 ecces;
+ u32 besr;
+ u32 bearh;
+ u32 bearl;
+ u32 wmirq;
+};
+
+/* Global Variables */
+
+/*
+ * Device tree node type and compatible tuples this driver can match
+ * on.
+ */
+static const struct of_device_id ppc4xx_edac_match[] = {
+ {
+ .compatible = "ibm,sdram-4xx-ddr2"
+ },
+ { }
+};
+MODULE_DEVICE_TABLE(of, ppc4xx_edac_match);
+
+/*
+ * TODO: The row and channel parameters likely need to be dynamically
+ * set based on the aforementioned variant controller realizations.
+ */
+static const unsigned ppc4xx_edac_nr_csrows = 2;
+static const unsigned ppc4xx_edac_nr_chans = 1;
+
+/*
+ * Strings associated with PLB master IDs capable of being posted in
+ * SDRAM_BESR or SDRAM_WMIRQ on uncorrectable ECC errors.
+ */
+static const char * const ppc4xx_plb_masters[9] = {
+ [SDRAM_PLB_M0ID_ICU] = "ICU",
+ [SDRAM_PLB_M0ID_PCIE0] = "PCI-E 0",
+ [SDRAM_PLB_M0ID_PCIE1] = "PCI-E 1",
+ [SDRAM_PLB_M0ID_DMA] = "DMA",
+ [SDRAM_PLB_M0ID_DCU] = "DCU",
+ [SDRAM_PLB_M0ID_OPB] = "OPB",
+ [SDRAM_PLB_M0ID_MAL] = "MAL",
+ [SDRAM_PLB_M0ID_SEC] = "SEC",
+ [SDRAM_PLB_M0ID_AHB] = "AHB"
+};
+
+/**
+ * mfsdram - read and return controller register data
+ * @dcr_host: A pointer to the DCR mapping.
+ * @idcr_n: The indirect DCR register to read.
+ *
+ * This routine reads and returns the data associated with the
+ * controller's specified indirect DCR register.
+ *
+ * Returns the read data.
+ */
+static inline u32
+mfsdram(const dcr_host_t *dcr_host, unsigned int idcr_n)
+{
+ return __mfdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
+ dcr_host->base + SDRAM_DCR_DATA_OFFSET,
+ idcr_n);
+}
+
+/**
+ * mtsdram - write controller register data
+ * @dcr_host: A pointer to the DCR mapping.
+ * @idcr_n: The indirect DCR register to write.
+ * @value: The data to write.
+ *
+ * This routine writes the provided data to the controller's specified
+ * indirect DCR register.
+ */
+static inline void
+mtsdram(const dcr_host_t *dcr_host, unsigned int idcr_n, u32 value)
+{
+ return __mtdcri(dcr_host->base + SDRAM_DCR_ADDR_OFFSET,
+ dcr_host->base + SDRAM_DCR_DATA_OFFSET,
+ idcr_n,
+ value);
+}
+
+/**
+ * ppc4xx_edac_check_bank_error - check a bank for an ECC bank error
+ * @status: A pointer to the ECC status structure to check for an
+ * ECC bank error.
+ * @bank: The bank to check for an ECC error.
+ *
+ * This routine determines whether the specified bank has an ECC
+ * error.
+ *
+ * Returns true if the specified bank has an ECC error; otherwise,
+ * false.
+ */
+static bool
+ppc4xx_edac_check_bank_error(const struct ppc4xx_ecc_status *status,
+ unsigned int bank)
+{
+ switch (bank) {
+ case 0:
+ return status->ecces & SDRAM_ECCES_BK0ER;
+ case 1:
+ return status->ecces & SDRAM_ECCES_BK1ER;
+ default:
+ return false;
+ }
+}
+
+/**
+ * ppc4xx_edac_generate_bank_message - generate interpretted bank status message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the bank message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the portion of the
+ * driver-unique report message associated with the ECCESS[BKNER]
+ * field of the specified ECC status.
+ *
+ * Returns the number of characters generated on success; otherwise, <
+ * 0 on error.
+ */
+static int
+ppc4xx_edac_generate_bank_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ int n, total = 0;
+ unsigned int row, rows;
+
+ n = snprintf(buffer, size, "%s: Banks: ", mci->dev_name);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ for (rows = 0, row = 0; row < mci->nr_csrows; row++) {
+ if (ppc4xx_edac_check_bank_error(status, row)) {
+ n = snprintf(buffer, size, "%s%u",
+ (rows++ ? ", " : ""), row);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+ }
+ }
+
+ n = snprintf(buffer, size, "%s; ", rows ? "" : "None");
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ fail:
+ return total;
+}
+
+/**
+ * ppc4xx_edac_generate_checkbit_message - generate interpretted checkbit message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the checkbit message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the portion of the
+ * driver-unique report message associated with the ECCESS[CKBER]
+ * field of the specified ECC status.
+ *
+ * Returns the number of characters generated on success; otherwise, <
+ * 0 on error.
+ */
+static int
+ppc4xx_edac_generate_checkbit_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+ const char *ckber = NULL;
+
+ switch (status->ecces & SDRAM_ECCES_CKBER_MASK) {
+ case SDRAM_ECCES_CKBER_NONE:
+ ckber = "None";
+ break;
+ case SDRAM_ECCES_CKBER_32_ECC_0_3:
+ ckber = "ECC0:3";
+ break;
+ case SDRAM_ECCES_CKBER_32_ECC_4_8:
+ switch (mfsdram(&pdata->dcr_host, SDRAM_MCOPT1) &
+ SDRAM_MCOPT1_WDTH_MASK) {
+ case SDRAM_MCOPT1_WDTH_16:
+ ckber = "ECC0:3";
+ break;
+ case SDRAM_MCOPT1_WDTH_32:
+ ckber = "ECC4:8";
+ break;
+ default:
+ ckber = "Unknown";
+ break;
+ }
+ break;
+ case SDRAM_ECCES_CKBER_32_ECC_0_8:
+ ckber = "ECC0:8";
+ break;
+ default:
+ ckber = "Unknown";
+ break;
+ }
+
+ return snprintf(buffer, size, "Checkbit Error: %s", ckber);
+}
+
+/**
+ * ppc4xx_edac_generate_lane_message - generate interpretted byte lane message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the byte lane message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the portion of the
+ * driver-unique report message associated with the ECCESS[BNCE]
+ * field of the specified ECC status.
+ *
+ * Returns the number of characters generated on success; otherwise, <
+ * 0 on error.
+ */
+static int
+ppc4xx_edac_generate_lane_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ int n, total = 0;
+ unsigned int lane, lanes;
+ const unsigned int first_lane = 0;
+ const unsigned int lane_count = 16;
+
+ n = snprintf(buffer, size, "; Byte Lane Errors: ");
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ for (lanes = 0, lane = first_lane; lane < lane_count; lane++) {
+ if ((status->ecces & SDRAM_ECCES_BNCE_ENCODE(lane)) != 0) {
+ n = snprintf(buffer, size,
+ "%s%u",
+ (lanes++ ? ", " : ""), lane);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+ }
+ }
+
+ n = snprintf(buffer, size, "%s; ", lanes ? "" : "None");
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ fail:
+ return total;
+}
+
+/**
+ * ppc4xx_edac_generate_ecc_message - generate interpretted ECC status message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the ECCES message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the portion of the
+ * driver-unique report message associated with the ECCESS register of
+ * the specified ECC status.
+ *
+ * Returns the number of characters generated on success; otherwise, <
+ * 0 on error.
+ */
+static int
+ppc4xx_edac_generate_ecc_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ int n, total = 0;
+
+ n = ppc4xx_edac_generate_bank_message(mci, status, buffer, size);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ n = ppc4xx_edac_generate_checkbit_message(mci, status, buffer, size);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ n = ppc4xx_edac_generate_lane_message(mci, status, buffer, size);
+
+ if (n < 0 || n >= size)
+ goto fail;
+
+ buffer += n;
+ size -= n;
+ total += n;
+
+ fail:
+ return total;
+}
+
+/**
+ * ppc4xx_edac_generate_plb_message - generate interpretted PLB status message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the PLB message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the portion of the
+ * driver-unique report message associated with the PLB-related BESR
+ * and/or WMIRQ registers of the specified ECC status.
+ *
+ * Returns the number of characters generated on success; otherwise, <
+ * 0 on error.
+ */
+static int
+ppc4xx_edac_generate_plb_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ unsigned int master;
+ bool read;
+
+ if ((status->besr & SDRAM_BESR_MASK) == 0)
+ return 0;
+
+ if ((status->besr & SDRAM_BESR_M0ET_MASK) == SDRAM_BESR_M0ET_NONE)
+ return 0;
+
+ read = ((status->besr & SDRAM_BESR_M0RW_MASK) == SDRAM_BESR_M0RW_READ);
+
+ master = SDRAM_BESR_M0ID_DECODE(status->besr);
+
+ return snprintf(buffer, size,
+ "%s error w/ PLB master %u \"%s\"; ",
+ (read ? "Read" : "Write"),
+ master,
+ (((master >= SDRAM_PLB_M0ID_FIRST) &&
+ (master <= SDRAM_PLB_M0ID_LAST)) ?
+ ppc4xx_plb_masters[master] : "UNKNOWN"));
+}
+
+/**
+ * ppc4xx_edac_generate_message - generate interpretted status message
+ * @mci: A pointer to the EDAC memory controller instance associated
+ * with the driver-unique message being generated.
+ * @status: A pointer to the ECC status structure to generate the
+ * message from.
+ * @buffer: A pointer to the buffer in which to generate the
+ * message.
+ * @size: The size, in bytes, of space available in buffer.
+ *
+ * This routine generates to the provided buffer the driver-unique
+ * EDAC report message from the specified ECC status.
+ */
+static void
+ppc4xx_edac_generate_message(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status,
+ char *buffer,
+ size_t size)
+{
+ int n;
+
+ if (buffer == NULL || size == 0)
+ return;
+
+ n = ppc4xx_edac_generate_ecc_message(mci, status, buffer, size);
+
+ if (n < 0 || n >= size)
+ return;
+
+ buffer += n;
+ size -= n;
+
+ ppc4xx_edac_generate_plb_message(mci, status, buffer, size);
+}
+
+#ifdef DEBUG
+/**
+ * ppc4xx_ecc_dump_status - dump controller ECC status registers
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the status being dumped.
+ * @status: A pointer to the ECC status structure to generate the
+ * dump from.
+ *
+ * This routine dumps to the kernel log buffer the raw and
+ * interpretted specified ECC status.
+ */
+static void
+ppc4xx_ecc_dump_status(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status)
+{
+ char message[PPC4XX_EDAC_MESSAGE_SIZE];
+
+ ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
+
+ ppc4xx_edac_mc_printk(KERN_INFO, mci,
+ "\n"
+ "\tECCES: 0x%08x\n"
+ "\tWMIRQ: 0x%08x\n"
+ "\tBESR: 0x%08x\n"
+ "\tBEAR: 0x%08x%08x\n"
+ "\t%s\n",
+ status->ecces,
+ status->wmirq,
+ status->besr,
+ status->bearh,
+ status->bearl,
+ message);
+}
+#endif /* DEBUG */
+
+/**
+ * ppc4xx_ecc_get_status - get controller ECC status
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the status being retrieved.
+ * @status: A pointer to the ECC status structure to populate the
+ * ECC status with.
+ *
+ * This routine reads and masks, as appropriate, all the relevant
+ * status registers that deal with ibm,sdram-4xx-ddr2 ECC errors.
+ * While we read all of them, for correctable errors, we only expect
+ * to deal with ECCES. For uncorrectable errors, we expect to deal
+ * with all of them.
+ */
+static void
+ppc4xx_ecc_get_status(const struct mem_ctl_info *mci,
+ struct ppc4xx_ecc_status *status)
+{
+ const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+ const dcr_host_t *dcr_host = &pdata->dcr_host;
+
+ status->ecces = mfsdram(dcr_host, SDRAM_ECCES) & SDRAM_ECCES_MASK;
+ status->wmirq = mfsdram(dcr_host, SDRAM_WMIRQ) & SDRAM_WMIRQ_MASK;
+ status->besr = mfsdram(dcr_host, SDRAM_BESR) & SDRAM_BESR_MASK;
+ status->bearl = mfsdram(dcr_host, SDRAM_BEARL);
+ status->bearh = mfsdram(dcr_host, SDRAM_BEARH);
+}
+
+/**
+ * ppc4xx_ecc_clear_status - clear controller ECC status
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the status being cleared.
+ * @status: A pointer to the ECC status structure containing the
+ * values to write to clear the ECC status.
+ *
+ * This routine clears--by writing the masked (as appropriate) status
+ * values back to--the status registers that deal with
+ * ibm,sdram-4xx-ddr2 ECC errors.
+ */
+static void
+ppc4xx_ecc_clear_status(const struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status)
+{
+ const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+ const dcr_host_t *dcr_host = &pdata->dcr_host;
+
+ mtsdram(dcr_host, SDRAM_ECCES, status->ecces & SDRAM_ECCES_MASK);
+ mtsdram(dcr_host, SDRAM_WMIRQ, status->wmirq & SDRAM_WMIRQ_MASK);
+ mtsdram(dcr_host, SDRAM_BESR, status->besr & SDRAM_BESR_MASK);
+ mtsdram(dcr_host, SDRAM_BEARL, 0);
+ mtsdram(dcr_host, SDRAM_BEARH, 0);
+}
+
+/**
+ * ppc4xx_edac_handle_ce - handle controller correctable ECC error (CE)
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the correctable error being handled and reported.
+ * @status: A pointer to the ECC status structure associated with
+ * the correctable error being handled and reported.
+ *
+ * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
+ * correctable error. Per the aforementioned discussion, there's not
+ * enough status available to use the full EDAC correctable error
+ * interface, so we just pass driver-unique message to the "no info"
+ * interface.
+ */
+static void
+ppc4xx_edac_handle_ce(struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status)
+{
+ int row;
+ char message[PPC4XX_EDAC_MESSAGE_SIZE];
+
+ ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
+
+ for (row = 0; row < mci->nr_csrows; row++)
+ if (ppc4xx_edac_check_bank_error(status, row))
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ 0, 0, 0,
+ row, 0, -1,
+ message, "");
+}
+
+/**
+ * ppc4xx_edac_handle_ue - handle controller uncorrectable ECC error (UE)
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the uncorrectable error being handled and
+ * reported.
+ * @status: A pointer to the ECC status structure associated with
+ * the uncorrectable error being handled and reported.
+ *
+ * This routine handles an ibm,sdram-4xx-ddr2 controller ECC
+ * uncorrectable error.
+ */
+static void
+ppc4xx_edac_handle_ue(struct mem_ctl_info *mci,
+ const struct ppc4xx_ecc_status *status)
+{
+ const u64 bear = ((u64)status->bearh << 32 | status->bearl);
+ const unsigned long page = bear >> PAGE_SHIFT;
+ const unsigned long offset = bear & ~PAGE_MASK;
+ int row;
+ char message[PPC4XX_EDAC_MESSAGE_SIZE];
+
+ ppc4xx_edac_generate_message(mci, status, message, sizeof(message));
+
+ for (row = 0; row < mci->nr_csrows; row++)
+ if (ppc4xx_edac_check_bank_error(status, row))
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ page, offset, 0,
+ row, 0, -1,
+ message, "");
+}
+
+/**
+ * ppc4xx_edac_check - check controller for ECC errors
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the ibm,sdram-4xx-ddr2 controller being
+ * checked.
+ *
+ * This routine is used to check and post ECC errors and is called by
+ * both the EDAC polling thread and this driver's CE and UE interrupt
+ * handler.
+ */
+static void
+ppc4xx_edac_check(struct mem_ctl_info *mci)
+{
+#ifdef DEBUG
+ static unsigned int count;
+#endif
+ struct ppc4xx_ecc_status status;
+
+ ppc4xx_ecc_get_status(mci, &status);
+
+#ifdef DEBUG
+ if (count++ % 30 == 0)
+ ppc4xx_ecc_dump_status(mci, &status);
+#endif
+
+ if (status.ecces & SDRAM_ECCES_UE)
+ ppc4xx_edac_handle_ue(mci, &status);
+
+ if (status.ecces & SDRAM_ECCES_CE)
+ ppc4xx_edac_handle_ce(mci, &status);
+
+ ppc4xx_ecc_clear_status(mci, &status);
+}
+
+/**
+ * ppc4xx_edac_isr - SEC (CE) and DED (UE) interrupt service routine
+ * @irq: The virtual interrupt number being serviced.
+ * @dev_id: A pointer to the EDAC memory controller instance
+ * associated with the interrupt being handled.
+ *
+ * This routine implements the interrupt handler for both correctable
+ * (CE) and uncorrectable (UE) ECC errors for the ibm,sdram-4xx-ddr2
+ * controller. It simply calls through to the same routine used during
+ * polling to check, report and clear the ECC status.
+ *
+ * Unconditionally returns IRQ_HANDLED.
+ */
+static irqreturn_t
+ppc4xx_edac_isr(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+
+ ppc4xx_edac_check(mci);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * ppc4xx_edac_get_dtype - return the controller memory width
+ * @mcopt1: The 32-bit Memory Controller Option 1 register value
+ * currently set for the controller, from which the width
+ * is derived.
+ *
+ * This routine returns the EDAC device type width appropriate for the
+ * current controller configuration.
+ *
+ * TODO: This needs to be conditioned dynamically through feature
+ * flags or some such when other controller variants are supported as
+ * the 405EX[r] is 16-/32-bit and the others are 32-/64-bit with the
+ * 16- and 64-bit field definition/value/enumeration (b1) overloaded
+ * among them.
+ *
+ * Returns a device type width enumeration.
+ */
+static enum dev_type ppc4xx_edac_get_dtype(u32 mcopt1)
+{
+ switch (mcopt1 & SDRAM_MCOPT1_WDTH_MASK) {
+ case SDRAM_MCOPT1_WDTH_16:
+ return DEV_X2;
+ case SDRAM_MCOPT1_WDTH_32:
+ return DEV_X4;
+ default:
+ return DEV_UNKNOWN;
+ }
+}
+
+/**
+ * ppc4xx_edac_get_mtype - return controller memory type
+ * @mcopt1: The 32-bit Memory Controller Option 1 register value
+ * currently set for the controller, from which the memory type
+ * is derived.
+ *
+ * This routine returns the EDAC memory type appropriate for the
+ * current controller configuration.
+ *
+ * Returns a memory type enumeration.
+ */
+static enum mem_type ppc4xx_edac_get_mtype(u32 mcopt1)
+{
+ bool rden = ((mcopt1 & SDRAM_MCOPT1_RDEN_MASK) == SDRAM_MCOPT1_RDEN);
+
+ switch (mcopt1 & SDRAM_MCOPT1_DDR_TYPE_MASK) {
+ case SDRAM_MCOPT1_DDR2_TYPE:
+ return rden ? MEM_RDDR2 : MEM_DDR2;
+ case SDRAM_MCOPT1_DDR1_TYPE:
+ return rden ? MEM_RDDR : MEM_DDR;
+ default:
+ return MEM_UNKNOWN;
+ }
+}
+
+/**
+ * ppc4xx_edac_init_csrows - initialize driver instance rows
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the ibm,sdram-4xx-ddr2 controller for which
+ * the csrows (i.e. banks/ranks) are being initialized.
+ * @mcopt1: The 32-bit Memory Controller Option 1 register value
+ * currently set for the controller, from which bank width
+ * and memory typ information is derived.
+ *
+ * This routine initializes the virtual "chip select rows" associated
+ * with the EDAC memory controller instance. An ibm,sdram-4xx-ddr2
+ * controller bank/rank is mapped to a row.
+ *
+ * Returns 0 if OK; otherwise, -EINVAL if the memory bank size
+ * configuration cannot be determined.
+ */
+static int ppc4xx_edac_init_csrows(struct mem_ctl_info *mci, u32 mcopt1)
+{
+ const struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+ int status = 0;
+ enum mem_type mtype;
+ enum dev_type dtype;
+ enum edac_type edac_mode;
+ int row, j;
+ u32 mbxcf, size, nr_pages;
+
+ /* Establish the memory type and width */
+
+ mtype = ppc4xx_edac_get_mtype(mcopt1);
+ dtype = ppc4xx_edac_get_dtype(mcopt1);
+
+ /* Establish EDAC mode */
+
+ if (mci->edac_cap & EDAC_FLAG_SECDED)
+ edac_mode = EDAC_SECDED;
+ else if (mci->edac_cap & EDAC_FLAG_EC)
+ edac_mode = EDAC_EC;
+ else
+ edac_mode = EDAC_NONE;
+
+ /*
+ * Initialize each chip select row structure which correspond
+ * 1:1 with a controller bank/rank.
+ */
+
+ for (row = 0; row < mci->nr_csrows; row++) {
+ struct csrow_info *csi = mci->csrows[row];
+
+ /*
+ * Get the configuration settings for this
+ * row/bank/rank and skip disabled banks.
+ */
+
+ mbxcf = mfsdram(&pdata->dcr_host, SDRAM_MBXCF(row));
+
+ if ((mbxcf & SDRAM_MBCF_BE_MASK) != SDRAM_MBCF_BE_ENABLE)
+ continue;
+
+ /* Map the bank configuration size setting to pages. */
+
+ size = mbxcf & SDRAM_MBCF_SZ_MASK;
+
+ switch (size) {
+ case SDRAM_MBCF_SZ_4MB:
+ case SDRAM_MBCF_SZ_8MB:
+ case SDRAM_MBCF_SZ_16MB:
+ case SDRAM_MBCF_SZ_32MB:
+ case SDRAM_MBCF_SZ_64MB:
+ case SDRAM_MBCF_SZ_128MB:
+ case SDRAM_MBCF_SZ_256MB:
+ case SDRAM_MBCF_SZ_512MB:
+ case SDRAM_MBCF_SZ_1GB:
+ case SDRAM_MBCF_SZ_2GB:
+ case SDRAM_MBCF_SZ_4GB:
+ case SDRAM_MBCF_SZ_8GB:
+ nr_pages = SDRAM_MBCF_SZ_TO_PAGES(size);
+ break;
+ default:
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Unrecognized memory bank %d "
+ "size 0x%08x\n",
+ row, SDRAM_MBCF_SZ_DECODE(size));
+ status = -EINVAL;
+ goto done;
+ }
+
+ /*
+ * It's unclear exactly what grain should be set to
+ * here. The SDRAM_ECCES register allows resolution of
+ * an error down to a nibble which would potentially
+ * argue for a grain of '1' byte, even though we only
+ * know the associated address for uncorrectable
+ * errors. This value is not used at present for
+ * anything other than error reporting so getting it
+ * wrong should be of little consequence. Other
+ * possible values would be the PLB width (16), the
+ * page size (PAGE_SIZE) or the memory width (2 or 4).
+ */
+ for (j = 0; j < csi->nr_channels; j++) {
+ struct dimm_info *dimm = csi->channels[j]->dimm;
+
+ dimm->nr_pages = nr_pages / csi->nr_channels;
+ dimm->grain = 1;
+
+ dimm->mtype = mtype;
+ dimm->dtype = dtype;
+
+ dimm->edac_mode = edac_mode;
+ }
+ }
+
+ done:
+ return status;
+}
+
+/**
+ * ppc4xx_edac_mc_init - initialize driver instance
+ * @mci: A pointer to the EDAC memory controller instance being
+ * initialized.
+ * @op: A pointer to the OpenFirmware device tree node associated
+ * with the controller this EDAC instance is bound to.
+ * @dcr_host: A pointer to the DCR data containing the DCR mapping
+ * for this controller instance.
+ * @mcopt1: The 32-bit Memory Controller Option 1 register value
+ * currently set for the controller, from which ECC capabilities
+ * and scrub mode are derived.
+ *
+ * This routine performs initialization of the EDAC memory controller
+ * instance and related driver-private data associated with the
+ * ibm,sdram-4xx-ddr2 memory controller the instance is bound to.
+ *
+ * Returns 0 if OK; otherwise, < 0 on error.
+ */
+static int ppc4xx_edac_mc_init(struct mem_ctl_info *mci,
+ struct platform_device *op,
+ const dcr_host_t *dcr_host, u32 mcopt1)
+{
+ int status = 0;
+ const u32 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
+ struct ppc4xx_edac_pdata *pdata = NULL;
+ const struct device_node *np = op->dev.of_node;
+
+ if (of_match_device(ppc4xx_edac_match, &op->dev) == NULL)
+ return -EINVAL;
+
+ /* Initial driver pointers and private data */
+
+ mci->pdev = &op->dev;
+
+ dev_set_drvdata(mci->pdev, mci);
+
+ pdata = mci->pvt_info;
+
+ pdata->dcr_host = *dcr_host;
+
+ /* Initialize controller capabilities and configuration */
+
+ mci->mtype_cap = (MEM_FLAG_DDR | MEM_FLAG_RDDR |
+ MEM_FLAG_DDR2 | MEM_FLAG_RDDR2);
+
+ mci->edac_ctl_cap = (EDAC_FLAG_NONE |
+ EDAC_FLAG_EC |
+ EDAC_FLAG_SECDED);
+
+ mci->scrub_cap = SCRUB_NONE;
+ mci->scrub_mode = SCRUB_NONE;
+
+ /*
+ * Update the actual capabilites based on the MCOPT1[MCHK]
+ * settings. Scrubbing is only useful if reporting is enabled.
+ */
+
+ switch (memcheck) {
+ case SDRAM_MCOPT1_MCHK_CHK:
+ mci->edac_cap = EDAC_FLAG_EC;
+ break;
+ case SDRAM_MCOPT1_MCHK_CHK_REP:
+ mci->edac_cap = (EDAC_FLAG_EC | EDAC_FLAG_SECDED);
+ mci->scrub_mode = SCRUB_SW_SRC;
+ break;
+ default:
+ mci->edac_cap = EDAC_FLAG_NONE;
+ break;
+ }
+
+ /* Initialize strings */
+
+ mci->mod_name = PPC4XX_EDAC_MODULE_NAME;
+ mci->ctl_name = ppc4xx_edac_match->compatible;
+ mci->dev_name = np->full_name;
+
+ /* Initialize callbacks */
+
+ mci->edac_check = ppc4xx_edac_check;
+ mci->ctl_page_to_phys = NULL;
+
+ /* Initialize chip select rows */
+
+ status = ppc4xx_edac_init_csrows(mci, mcopt1);
+
+ if (status)
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Failed to initialize rows!\n");
+
+ return status;
+}
+
+/**
+ * ppc4xx_edac_register_irq - setup and register controller interrupts
+ * @op: A pointer to the OpenFirmware device tree node associated
+ * with the controller this EDAC instance is bound to.
+ * @mci: A pointer to the EDAC memory controller instance
+ * associated with the ibm,sdram-4xx-ddr2 controller for which
+ * interrupts are being registered.
+ *
+ * This routine parses the correctable (CE) and uncorrectable error (UE)
+ * interrupts from the device tree node and maps and assigns them to
+ * the associated EDAC memory controller instance.
+ *
+ * Returns 0 if OK; otherwise, -ENODEV if the interrupts could not be
+ * mapped and assigned.
+ */
+static int ppc4xx_edac_register_irq(struct platform_device *op,
+ struct mem_ctl_info *mci)
+{
+ int status = 0;
+ int ded_irq, sec_irq;
+ struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+ struct device_node *np = op->dev.of_node;
+
+ ded_irq = irq_of_parse_and_map(np, INTMAP_ECCDED_INDEX);
+ sec_irq = irq_of_parse_and_map(np, INTMAP_ECCSEC_INDEX);
+
+ if (!ded_irq || !sec_irq) {
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Unable to map interrupts.\n");
+ status = -ENODEV;
+ goto fail;
+ }
+
+ status = request_irq(ded_irq,
+ ppc4xx_edac_isr,
+ 0,
+ "[EDAC] MC ECCDED",
+ mci);
+
+ if (status < 0) {
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Unable to request irq %d for ECC DED",
+ ded_irq);
+ status = -ENODEV;
+ goto fail1;
+ }
+
+ status = request_irq(sec_irq,
+ ppc4xx_edac_isr,
+ 0,
+ "[EDAC] MC ECCSEC",
+ mci);
+
+ if (status < 0) {
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Unable to request irq %d for ECC SEC",
+ sec_irq);
+ status = -ENODEV;
+ goto fail2;
+ }
+
+ ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCDED irq is %d\n", ded_irq);
+ ppc4xx_edac_mc_printk(KERN_INFO, mci, "ECCSEC irq is %d\n", sec_irq);
+
+ pdata->irqs.ded = ded_irq;
+ pdata->irqs.sec = sec_irq;
+
+ return 0;
+
+ fail2:
+ free_irq(sec_irq, mci);
+
+ fail1:
+ free_irq(ded_irq, mci);
+
+ fail:
+ return status;
+}
+
+/**
+ * ppc4xx_edac_map_dcrs - locate and map controller registers
+ * @np: A pointer to the device tree node containing the DCR
+ * resources to map.
+ * @dcr_host: A pointer to the DCR data to populate with the
+ * DCR mapping.
+ *
+ * This routine attempts to locate in the device tree and map the DCR
+ * register resources associated with the controller's indirect DCR
+ * address and data windows.
+ *
+ * Returns 0 if the DCRs were successfully mapped; otherwise, < 0 on
+ * error.
+ */
+static int ppc4xx_edac_map_dcrs(const struct device_node *np,
+ dcr_host_t *dcr_host)
+{
+ unsigned int dcr_base, dcr_len;
+
+ if (np == NULL || dcr_host == NULL)
+ return -EINVAL;
+
+ /* Get the DCR resource extent and sanity check the values. */
+
+ dcr_base = dcr_resource_start(np, 0);
+ dcr_len = dcr_resource_len(np, 0);
+
+ if (dcr_base == 0 || dcr_len == 0) {
+ ppc4xx_edac_printk(KERN_ERR,
+ "Failed to obtain DCR property.\n");
+ return -ENODEV;
+ }
+
+ if (dcr_len != SDRAM_DCR_RESOURCE_LEN) {
+ ppc4xx_edac_printk(KERN_ERR,
+ "Unexpected DCR length %d, expected %d.\n",
+ dcr_len, SDRAM_DCR_RESOURCE_LEN);
+ return -ENODEV;
+ }
+
+ /* Attempt to map the DCR extent. */
+
+ *dcr_host = dcr_map(np, dcr_base, dcr_len);
+
+ if (!DCR_MAP_OK(*dcr_host)) {
+ ppc4xx_edac_printk(KERN_INFO, "Failed to map DCRs.\n");
+ return -ENODEV;
+ }
+
+ return 0;
+}
+
+/**
+ * ppc4xx_edac_probe - check controller and bind driver
+ * @op: A pointer to the OpenFirmware device tree node associated
+ * with the controller being probed for driver binding.
+ *
+ * This routine probes a specific ibm,sdram-4xx-ddr2 controller
+ * instance for binding with the driver.
+ *
+ * Returns 0 if the controller instance was successfully bound to the
+ * driver; otherwise, < 0 on error.
+ */
+static int ppc4xx_edac_probe(struct platform_device *op)
+{
+ int status = 0;
+ u32 mcopt1, memcheck;
+ dcr_host_t dcr_host;
+ const struct device_node *np = op->dev.of_node;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ static int ppc4xx_edac_instance;
+
+ /*
+ * At this point, we only support the controller realized on
+ * the AMCC PPC 405EX[r]. Reject anything else.
+ */
+
+ if (!of_device_is_compatible(np, "ibm,sdram-405ex") &&
+ !of_device_is_compatible(np, "ibm,sdram-405exr")) {
+ ppc4xx_edac_printk(KERN_NOTICE,
+ "Only the PPC405EX[r] is supported.\n");
+ return -ENODEV;
+ }
+
+ /*
+ * Next, get the DCR property and attempt to map it so that we
+ * can probe the controller.
+ */
+
+ status = ppc4xx_edac_map_dcrs(np, &dcr_host);
+
+ if (status)
+ return status;
+
+ /*
+ * First determine whether ECC is enabled at all. If not,
+ * there is no useful checking or monitoring that can be done
+ * for this controller.
+ */
+
+ mcopt1 = mfsdram(&dcr_host, SDRAM_MCOPT1);
+ memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
+
+ if (memcheck == SDRAM_MCOPT1_MCHK_NON) {
+ ppc4xx_edac_printk(KERN_INFO, "%pOF: No ECC memory detected or "
+ "ECC is disabled.\n", np);
+ status = -ENODEV;
+ goto done;
+ }
+
+ /*
+ * At this point, we know ECC is enabled, allocate an EDAC
+ * controller instance and perform the appropriate
+ * initialization.
+ */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = ppc4xx_edac_nr_csrows;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = ppc4xx_edac_nr_chans;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(ppc4xx_edac_instance, ARRAY_SIZE(layers), layers,
+ sizeof(struct ppc4xx_edac_pdata));
+ if (mci == NULL) {
+ ppc4xx_edac_printk(KERN_ERR, "%pOF: "
+ "Failed to allocate EDAC MC instance!\n",
+ np);
+ status = -ENOMEM;
+ goto done;
+ }
+
+ status = ppc4xx_edac_mc_init(mci, op, &dcr_host, mcopt1);
+
+ if (status) {
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Failed to initialize instance!\n");
+ goto fail;
+ }
+
+ /*
+ * We have a valid, initialized EDAC instance bound to the
+ * controller. Attempt to register it with the EDAC subsystem
+ * and, if necessary, register interrupts.
+ */
+
+ if (edac_mc_add_mc(mci)) {
+ ppc4xx_edac_mc_printk(KERN_ERR, mci,
+ "Failed to add instance!\n");
+ status = -ENODEV;
+ goto fail;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ status = ppc4xx_edac_register_irq(op, mci);
+
+ if (status)
+ goto fail1;
+ }
+
+ ppc4xx_edac_instance++;
+
+ return 0;
+
+ fail1:
+ edac_mc_del_mc(mci->pdev);
+
+ fail:
+ edac_mc_free(mci);
+
+ done:
+ return status;
+}
+
+/**
+ * ppc4xx_edac_remove - unbind driver from controller
+ * @op: A pointer to the OpenFirmware device tree node associated
+ * with the controller this EDAC instance is to be unbound/removed
+ * from.
+ *
+ * This routine unbinds the EDAC memory controller instance associated
+ * with the specified ibm,sdram-4xx-ddr2 controller described by the
+ * OpenFirmware device tree node passed as a parameter.
+ *
+ * Unconditionally returns 0.
+ */
+static int
+ppc4xx_edac_remove(struct platform_device *op)
+{
+ struct mem_ctl_info *mci = dev_get_drvdata(&op->dev);
+ struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ free_irq(pdata->irqs.sec, mci);
+ free_irq(pdata->irqs.ded, mci);
+ }
+
+ dcr_unmap(pdata->dcr_host, SDRAM_DCR_RESOURCE_LEN);
+
+ edac_mc_del_mc(mci->pdev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+/**
+ * ppc4xx_edac_opstate_init - initialize EDAC reporting method
+ *
+ * This routine ensures that the EDAC memory controller reporting
+ * method is mapped to a sane value as the EDAC core defines the value
+ * to EDAC_OPSTATE_INVAL by default. We don't call the global
+ * opstate_init as that defaults to polling and we want interrupt as
+ * the default.
+ */
+static inline void __init
+ppc4xx_edac_opstate_init(void)
+{
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_INT:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_INT;
+ break;
+ }
+
+ ppc4xx_edac_printk(KERN_INFO, "Reporting type: %s\n",
+ ((edac_op_state == EDAC_OPSTATE_POLL) ?
+ EDAC_OPSTATE_POLL_STR :
+ ((edac_op_state == EDAC_OPSTATE_INT) ?
+ EDAC_OPSTATE_INT_STR :
+ EDAC_OPSTATE_UNKNOWN_STR)));
+}
+
+static struct platform_driver ppc4xx_edac_driver = {
+ .probe = ppc4xx_edac_probe,
+ .remove = ppc4xx_edac_remove,
+ .driver = {
+ .name = PPC4XX_EDAC_MODULE_NAME,
+ .of_match_table = ppc4xx_edac_match,
+ },
+};
+
+/**
+ * ppc4xx_edac_init - driver/module insertion entry point
+ *
+ * This routine is the driver/module insertion entry point. It
+ * initializes the EDAC memory controller reporting state and
+ * registers the driver as an OpenFirmware device tree platform
+ * driver.
+ */
+static int __init
+ppc4xx_edac_init(void)
+{
+ ppc4xx_edac_printk(KERN_INFO, PPC4XX_EDAC_MODULE_REVISION "\n");
+
+ ppc4xx_edac_opstate_init();
+
+ return platform_driver_register(&ppc4xx_edac_driver);
+}
+
+/**
+ * ppc4xx_edac_exit - driver/module removal entry point
+ *
+ * This routine is the driver/module removal entry point. It
+ * unregisters the driver as an OpenFirmware device tree platform
+ * driver.
+ */
+static void __exit
+ppc4xx_edac_exit(void)
+{
+ platform_driver_unregister(&ppc4xx_edac_driver);
+}
+
+module_init(ppc4xx_edac_init);
+module_exit(ppc4xx_edac_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Grant Erickson <gerickson@nuovations.com>");
+MODULE_DESCRIPTION("EDAC MC Driver for the PPC4xx IBM DDR2 Memory Controller");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting State: "
+ "0=" EDAC_OPSTATE_POLL_STR ", 2=" EDAC_OPSTATE_INT_STR);
diff --git a/drivers/edac/ppc4xx_edac.h b/drivers/edac/ppc4xx_edac.h
new file mode 100644
index 0000000000..b38459aa58
--- /dev/null
+++ b/drivers/edac/ppc4xx_edac.h
@@ -0,0 +1,167 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (c) 2008 Nuovation System Designs, LLC
+ * Grant Erickson <gerickson@nuovations.com>
+ *
+ * This file defines processor mnemonics for accessing and managing
+ * the IBM DDR1/DDR2 ECC controller found in the 405EX[r], 440SP,
+ * 440SPe, 460EX, 460GT and 460SX.
+ */
+
+#ifndef __PPC4XX_EDAC_H
+#define __PPC4XX_EDAC_H
+
+#include <linux/types.h>
+
+/*
+ * Macro for generating register field mnemonics
+ */
+#define PPC_REG_BITS 32
+#define PPC_REG_VAL(bit, val) ((val) << ((PPC_REG_BITS - 1) - (bit)))
+#define PPC_REG_DECODE(bit, val) ((val) >> ((PPC_REG_BITS - 1) - (bit)))
+
+/*
+ * IBM 4xx DDR1/DDR2 SDRAM memory controller registers (at least those
+ * relevant to ECC)
+ */
+#define SDRAM_BESR 0x00 /* Error status (read/clear) */
+#define SDRAM_BESRT 0x01 /* Error statuss (test/set) */
+#define SDRAM_BEARL 0x02 /* Error address low */
+#define SDRAM_BEARH 0x03 /* Error address high */
+#define SDRAM_WMIRQ 0x06 /* Write master (read/clear) */
+#define SDRAM_WMIRQT 0x07 /* Write master (test/set) */
+#define SDRAM_MCOPT1 0x20 /* Controller options 1 */
+#define SDRAM_MBXCF_BASE 0x40 /* Bank n configuration base */
+#define SDRAM_MBXCF(n) (SDRAM_MBXCF_BASE + (4 * (n)))
+#define SDRAM_MB0CF SDRAM_MBXCF(0)
+#define SDRAM_MB1CF SDRAM_MBXCF(1)
+#define SDRAM_MB2CF SDRAM_MBXCF(2)
+#define SDRAM_MB3CF SDRAM_MBXCF(3)
+#define SDRAM_ECCCR 0x98 /* ECC error status */
+#define SDRAM_ECCES SDRAM_ECCCR
+
+/*
+ * PLB Master IDs
+ */
+#define SDRAM_PLB_M0ID_FIRST 0
+#define SDRAM_PLB_M0ID_ICU SDRAM_PLB_M0ID_FIRST
+#define SDRAM_PLB_M0ID_PCIE0 1
+#define SDRAM_PLB_M0ID_PCIE1 2
+#define SDRAM_PLB_M0ID_DMA 3
+#define SDRAM_PLB_M0ID_DCU 4
+#define SDRAM_PLB_M0ID_OPB 5
+#define SDRAM_PLB_M0ID_MAL 6
+#define SDRAM_PLB_M0ID_SEC 7
+#define SDRAM_PLB_M0ID_AHB 8
+#define SDRAM_PLB_M0ID_LAST SDRAM_PLB_M0ID_AHB
+#define SDRAM_PLB_M0ID_COUNT (SDRAM_PLB_M0ID_LAST - \
+ SDRAM_PLB_M0ID_FIRST + 1)
+
+/*
+ * Memory Controller Bus Error Status Register
+ */
+#define SDRAM_BESR_MASK PPC_REG_VAL(7, 0xFF)
+#define SDRAM_BESR_M0ID_MASK PPC_REG_VAL(3, 0xF)
+#define SDRAM_BESR_M0ID_DECODE(n) PPC_REG_DECODE(3, n)
+#define SDRAM_BESR_M0ID_ICU PPC_REG_VAL(3, SDRAM_PLB_M0ID_ICU)
+#define SDRAM_BESR_M0ID_PCIE0 PPC_REG_VAL(3, SDRAM_PLB_M0ID_PCIE0)
+#define SDRAM_BESR_M0ID_PCIE1 PPC_REG_VAL(3, SDRAM_PLB_M0ID_PCIE1)
+#define SDRAM_BESR_M0ID_DMA PPC_REG_VAL(3, SDRAM_PLB_M0ID_DMA)
+#define SDRAM_BESR_M0ID_DCU PPC_REG_VAL(3, SDRAM_PLB_M0ID_DCU)
+#define SDRAM_BESR_M0ID_OPB PPC_REG_VAL(3, SDRAM_PLB_M0ID_OPB)
+#define SDRAM_BESR_M0ID_MAL PPC_REG_VAL(3, SDRAM_PLB_M0ID_MAL)
+#define SDRAM_BESR_M0ID_SEC PPC_REG_VAL(3, SDRAM_PLB_M0ID_SEC)
+#define SDRAM_BESR_M0ID_AHB PPC_REG_VAL(3, SDRAM_PLB_M0ID_AHB)
+#define SDRAM_BESR_M0ET_MASK PPC_REG_VAL(6, 0x7)
+#define SDRAM_BESR_M0ET_NONE PPC_REG_VAL(6, 0)
+#define SDRAM_BESR_M0ET_ECC PPC_REG_VAL(6, 1)
+#define SDRAM_BESR_M0RW_MASK PPC_REG_VAL(7, 1)
+#define SDRAM_BESR_M0RW_WRITE PPC_REG_VAL(7, 0)
+#define SDRAM_BESR_M0RW_READ PPC_REG_VAL(7, 1)
+
+/*
+ * Memory Controller PLB Write Master Interrupt Register
+ */
+#define SDRAM_WMIRQ_MASK PPC_REG_VAL(8, 0x1FF)
+#define SDRAM_WMIRQ_ENCODE(id) PPC_REG_VAL((id % \
+ SDRAM_PLB_M0ID_COUNT), 1)
+#define SDRAM_WMIRQ_ICU PPC_REG_VAL(SDRAM_PLB_M0ID_ICU, 1)
+#define SDRAM_WMIRQ_PCIE0 PPC_REG_VAL(SDRAM_PLB_M0ID_PCIE0, 1)
+#define SDRAM_WMIRQ_PCIE1 PPC_REG_VAL(SDRAM_PLB_M0ID_PCIE1, 1)
+#define SDRAM_WMIRQ_DMA PPC_REG_VAL(SDRAM_PLB_M0ID_DMA, 1)
+#define SDRAM_WMIRQ_DCU PPC_REG_VAL(SDRAM_PLB_M0ID_DCU, 1)
+#define SDRAM_WMIRQ_OPB PPC_REG_VAL(SDRAM_PLB_M0ID_OPB, 1)
+#define SDRAM_WMIRQ_MAL PPC_REG_VAL(SDRAM_PLB_M0ID_MAL, 1)
+#define SDRAM_WMIRQ_SEC PPC_REG_VAL(SDRAM_PLB_M0ID_SEC, 1)
+#define SDRAM_WMIRQ_AHB PPC_REG_VAL(SDRAM_PLB_M0ID_AHB, 1)
+
+/*
+ * Memory Controller Options 1 Register
+ */
+#define SDRAM_MCOPT1_MCHK_MASK PPC_REG_VAL(3, 0x3) /* ECC mask */
+#define SDRAM_MCOPT1_MCHK_NON PPC_REG_VAL(3, 0x0) /* No ECC gen */
+#define SDRAM_MCOPT1_MCHK_GEN PPC_REG_VAL(3, 0x2) /* ECC gen */
+#define SDRAM_MCOPT1_MCHK_CHK PPC_REG_VAL(3, 0x1) /* ECC gen and chk */
+#define SDRAM_MCOPT1_MCHK_CHK_REP PPC_REG_VAL(3, 0x3) /* ECC gen/chk/rpt */
+#define SDRAM_MCOPT1_MCHK_DECODE(n) ((((u32)(n)) >> 28) & 0x3)
+#define SDRAM_MCOPT1_RDEN_MASK PPC_REG_VAL(4, 0x1) /* Rgstrd DIMM mask */
+#define SDRAM_MCOPT1_RDEN PPC_REG_VAL(4, 0x1) /* Rgstrd DIMM enbl */
+#define SDRAM_MCOPT1_WDTH_MASK PPC_REG_VAL(7, 0x1) /* Width mask */
+#define SDRAM_MCOPT1_WDTH_32 PPC_REG_VAL(7, 0x0) /* 32 bits */
+#define SDRAM_MCOPT1_WDTH_16 PPC_REG_VAL(7, 0x1) /* 16 bits */
+#define SDRAM_MCOPT1_DDR_TYPE_MASK PPC_REG_VAL(11, 0x1) /* DDR type mask */
+#define SDRAM_MCOPT1_DDR1_TYPE PPC_REG_VAL(11, 0x0) /* DDR1 type */
+#define SDRAM_MCOPT1_DDR2_TYPE PPC_REG_VAL(11, 0x1) /* DDR2 type */
+
+/*
+ * Memory Bank 0 - n Configuration Register
+ */
+#define SDRAM_MBCF_BA_MASK PPC_REG_VAL(12, 0x1FFF)
+#define SDRAM_MBCF_SZ_MASK PPC_REG_VAL(19, 0xF)
+#define SDRAM_MBCF_SZ_DECODE(mbxcf) PPC_REG_DECODE(19, mbxcf)
+#define SDRAM_MBCF_SZ_4MB PPC_REG_VAL(19, 0x0)
+#define SDRAM_MBCF_SZ_8MB PPC_REG_VAL(19, 0x1)
+#define SDRAM_MBCF_SZ_16MB PPC_REG_VAL(19, 0x2)
+#define SDRAM_MBCF_SZ_32MB PPC_REG_VAL(19, 0x3)
+#define SDRAM_MBCF_SZ_64MB PPC_REG_VAL(19, 0x4)
+#define SDRAM_MBCF_SZ_128MB PPC_REG_VAL(19, 0x5)
+#define SDRAM_MBCF_SZ_256MB PPC_REG_VAL(19, 0x6)
+#define SDRAM_MBCF_SZ_512MB PPC_REG_VAL(19, 0x7)
+#define SDRAM_MBCF_SZ_1GB PPC_REG_VAL(19, 0x8)
+#define SDRAM_MBCF_SZ_2GB PPC_REG_VAL(19, 0x9)
+#define SDRAM_MBCF_SZ_4GB PPC_REG_VAL(19, 0xA)
+#define SDRAM_MBCF_SZ_8GB PPC_REG_VAL(19, 0xB)
+#define SDRAM_MBCF_AM_MASK PPC_REG_VAL(23, 0xF)
+#define SDRAM_MBCF_AM_MODE0 PPC_REG_VAL(23, 0x0)
+#define SDRAM_MBCF_AM_MODE1 PPC_REG_VAL(23, 0x1)
+#define SDRAM_MBCF_AM_MODE2 PPC_REG_VAL(23, 0x2)
+#define SDRAM_MBCF_AM_MODE3 PPC_REG_VAL(23, 0x3)
+#define SDRAM_MBCF_AM_MODE4 PPC_REG_VAL(23, 0x4)
+#define SDRAM_MBCF_AM_MODE5 PPC_REG_VAL(23, 0x5)
+#define SDRAM_MBCF_AM_MODE6 PPC_REG_VAL(23, 0x6)
+#define SDRAM_MBCF_AM_MODE7 PPC_REG_VAL(23, 0x7)
+#define SDRAM_MBCF_AM_MODE8 PPC_REG_VAL(23, 0x8)
+#define SDRAM_MBCF_AM_MODE9 PPC_REG_VAL(23, 0x9)
+#define SDRAM_MBCF_BE_MASK PPC_REG_VAL(31, 0x1)
+#define SDRAM_MBCF_BE_DISABLE PPC_REG_VAL(31, 0x0)
+#define SDRAM_MBCF_BE_ENABLE PPC_REG_VAL(31, 0x1)
+
+/*
+ * ECC Error Status
+ */
+#define SDRAM_ECCES_MASK PPC_REG_VAL(21, 0x3FFFFF)
+#define SDRAM_ECCES_BNCE_MASK PPC_REG_VAL(15, 0xFFFF)
+#define SDRAM_ECCES_BNCE_ENCODE(lane) PPC_REG_VAL(((lane) & 0xF), 1)
+#define SDRAM_ECCES_CKBER_MASK PPC_REG_VAL(17, 0x3)
+#define SDRAM_ECCES_CKBER_NONE PPC_REG_VAL(17, 0)
+#define SDRAM_ECCES_CKBER_16_ECC_0_3 PPC_REG_VAL(17, 2)
+#define SDRAM_ECCES_CKBER_32_ECC_0_3 PPC_REG_VAL(17, 1)
+#define SDRAM_ECCES_CKBER_32_ECC_4_8 PPC_REG_VAL(17, 2)
+#define SDRAM_ECCES_CKBER_32_ECC_0_8 PPC_REG_VAL(17, 3)
+#define SDRAM_ECCES_CE PPC_REG_VAL(18, 1)
+#define SDRAM_ECCES_UE PPC_REG_VAL(19, 1)
+#define SDRAM_ECCES_BKNER_MASK PPC_REG_VAL(21, 0x3)
+#define SDRAM_ECCES_BK0ER PPC_REG_VAL(20, 1)
+#define SDRAM_ECCES_BK1ER PPC_REG_VAL(21, 1)
+
+#endif /* __PPC4XX_EDAC_H */
diff --git a/drivers/edac/qcom_edac.c b/drivers/edac/qcom_edac.c
new file mode 100644
index 0000000000..b2db545c68
--- /dev/null
+++ b/drivers/edac/qcom_edac.c
@@ -0,0 +1,420 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+#include <linux/soc/qcom/llcc-qcom.h>
+
+#include "edac_mc.h"
+#include "edac_device.h"
+
+#define EDAC_LLCC "qcom_llcc"
+
+#define LLCC_ERP_PANIC_ON_UE 1
+
+#define TRP_SYN_REG_CNT 6
+#define DRP_SYN_REG_CNT 8
+
+#define LLCC_LB_CNT_MASK GENMASK(31, 28)
+#define LLCC_LB_CNT_SHIFT 28
+
+/* Mask and shift macros */
+#define ECC_DB_ERR_COUNT_MASK GENMASK(4, 0)
+#define ECC_DB_ERR_WAYS_MASK GENMASK(31, 16)
+#define ECC_DB_ERR_WAYS_SHIFT BIT(4)
+
+#define ECC_SB_ERR_COUNT_MASK GENMASK(23, 16)
+#define ECC_SB_ERR_COUNT_SHIFT BIT(4)
+#define ECC_SB_ERR_WAYS_MASK GENMASK(15, 0)
+
+#define SB_ECC_ERROR BIT(0)
+#define DB_ECC_ERROR BIT(1)
+
+#define DRP_TRP_INT_CLEAR GENMASK(1, 0)
+#define DRP_TRP_CNT_CLEAR GENMASK(1, 0)
+
+#define SB_ERROR_THRESHOLD 0x1
+#define SB_ERROR_THRESHOLD_SHIFT 24
+#define SB_DB_TRP_INTERRUPT_ENABLE 0x3
+#define TRP0_INTERRUPT_ENABLE 0x1
+#define DRP0_INTERRUPT_ENABLE BIT(6)
+#define SB_DB_DRP_INTERRUPT_ENABLE 0x3
+
+#define ECC_POLL_MSEC 5000
+
+enum {
+ LLCC_DRAM_CE = 0,
+ LLCC_DRAM_UE,
+ LLCC_TRAM_CE,
+ LLCC_TRAM_UE,
+};
+
+static const struct llcc_edac_reg_data edac_reg_data[] = {
+ [LLCC_DRAM_CE] = {
+ .name = "DRAM Single-bit",
+ .reg_cnt = DRP_SYN_REG_CNT,
+ .count_mask = ECC_SB_ERR_COUNT_MASK,
+ .ways_mask = ECC_SB_ERR_WAYS_MASK,
+ .count_shift = ECC_SB_ERR_COUNT_SHIFT,
+ },
+ [LLCC_DRAM_UE] = {
+ .name = "DRAM Double-bit",
+ .reg_cnt = DRP_SYN_REG_CNT,
+ .count_mask = ECC_DB_ERR_COUNT_MASK,
+ .ways_mask = ECC_DB_ERR_WAYS_MASK,
+ .ways_shift = ECC_DB_ERR_WAYS_SHIFT,
+ },
+ [LLCC_TRAM_CE] = {
+ .name = "TRAM Single-bit",
+ .reg_cnt = TRP_SYN_REG_CNT,
+ .count_mask = ECC_SB_ERR_COUNT_MASK,
+ .ways_mask = ECC_SB_ERR_WAYS_MASK,
+ .count_shift = ECC_SB_ERR_COUNT_SHIFT,
+ },
+ [LLCC_TRAM_UE] = {
+ .name = "TRAM Double-bit",
+ .reg_cnt = TRP_SYN_REG_CNT,
+ .count_mask = ECC_DB_ERR_COUNT_MASK,
+ .ways_mask = ECC_DB_ERR_WAYS_MASK,
+ .ways_shift = ECC_DB_ERR_WAYS_SHIFT,
+ },
+};
+
+static int qcom_llcc_core_setup(struct llcc_drv_data *drv, struct regmap *llcc_bcast_regmap)
+{
+ u32 sb_err_threshold;
+ int ret;
+
+ /*
+ * Configure interrupt enable registers such that Tag, Data RAM related
+ * interrupts are propagated to interrupt controller for servicing
+ */
+ ret = regmap_update_bits(llcc_bcast_regmap, drv->edac_reg_offset->cmn_interrupt_2_enable,
+ TRP0_INTERRUPT_ENABLE,
+ TRP0_INTERRUPT_ENABLE);
+ if (ret)
+ return ret;
+
+ ret = regmap_update_bits(llcc_bcast_regmap, drv->edac_reg_offset->trp_interrupt_0_enable,
+ SB_DB_TRP_INTERRUPT_ENABLE,
+ SB_DB_TRP_INTERRUPT_ENABLE);
+ if (ret)
+ return ret;
+
+ sb_err_threshold = (SB_ERROR_THRESHOLD << SB_ERROR_THRESHOLD_SHIFT);
+ ret = regmap_write(llcc_bcast_regmap, drv->edac_reg_offset->drp_ecc_error_cfg,
+ sb_err_threshold);
+ if (ret)
+ return ret;
+
+ ret = regmap_update_bits(llcc_bcast_regmap, drv->edac_reg_offset->cmn_interrupt_2_enable,
+ DRP0_INTERRUPT_ENABLE,
+ DRP0_INTERRUPT_ENABLE);
+ if (ret)
+ return ret;
+
+ ret = regmap_write(llcc_bcast_regmap, drv->edac_reg_offset->drp_interrupt_enable,
+ SB_DB_DRP_INTERRUPT_ENABLE);
+ return ret;
+}
+
+/* Clear the error interrupt and counter registers */
+static int
+qcom_llcc_clear_error_status(int err_type, struct llcc_drv_data *drv)
+{
+ int ret;
+
+ switch (err_type) {
+ case LLCC_DRAM_CE:
+ case LLCC_DRAM_UE:
+ ret = regmap_write(drv->bcast_regmap,
+ drv->edac_reg_offset->drp_interrupt_clear,
+ DRP_TRP_INT_CLEAR);
+ if (ret)
+ return ret;
+
+ ret = regmap_write(drv->bcast_regmap,
+ drv->edac_reg_offset->drp_ecc_error_cntr_clear,
+ DRP_TRP_CNT_CLEAR);
+ if (ret)
+ return ret;
+ break;
+ case LLCC_TRAM_CE:
+ case LLCC_TRAM_UE:
+ ret = regmap_write(drv->bcast_regmap,
+ drv->edac_reg_offset->trp_interrupt_0_clear,
+ DRP_TRP_INT_CLEAR);
+ if (ret)
+ return ret;
+
+ ret = regmap_write(drv->bcast_regmap,
+ drv->edac_reg_offset->trp_ecc_error_cntr_clear,
+ DRP_TRP_CNT_CLEAR);
+ if (ret)
+ return ret;
+ break;
+ default:
+ ret = -EINVAL;
+ edac_printk(KERN_CRIT, EDAC_LLCC, "Unexpected error type: %d\n",
+ err_type);
+ }
+ return ret;
+}
+
+struct qcom_llcc_syn_regs {
+ u32 synd_reg;
+ u32 count_status_reg;
+ u32 ways_status_reg;
+};
+
+static void get_reg_offsets(struct llcc_drv_data *drv, int err_type,
+ struct qcom_llcc_syn_regs *syn_regs)
+{
+ const struct llcc_edac_reg_offset *edac_reg_offset = drv->edac_reg_offset;
+
+ switch (err_type) {
+ case LLCC_DRAM_CE:
+ syn_regs->synd_reg = edac_reg_offset->drp_ecc_sb_err_syn0;
+ syn_regs->count_status_reg = edac_reg_offset->drp_ecc_error_status1;
+ syn_regs->ways_status_reg = edac_reg_offset->drp_ecc_error_status0;
+ break;
+ case LLCC_DRAM_UE:
+ syn_regs->synd_reg = edac_reg_offset->drp_ecc_db_err_syn0;
+ syn_regs->count_status_reg = edac_reg_offset->drp_ecc_error_status1;
+ syn_regs->ways_status_reg = edac_reg_offset->drp_ecc_error_status0;
+ break;
+ case LLCC_TRAM_CE:
+ syn_regs->synd_reg = edac_reg_offset->trp_ecc_sb_err_syn0;
+ syn_regs->count_status_reg = edac_reg_offset->trp_ecc_error_status1;
+ syn_regs->ways_status_reg = edac_reg_offset->trp_ecc_error_status0;
+ break;
+ case LLCC_TRAM_UE:
+ syn_regs->synd_reg = edac_reg_offset->trp_ecc_db_err_syn0;
+ syn_regs->count_status_reg = edac_reg_offset->trp_ecc_error_status1;
+ syn_regs->ways_status_reg = edac_reg_offset->trp_ecc_error_status0;
+ break;
+ }
+}
+
+/* Dump Syndrome registers data for Tag RAM, Data RAM bit errors*/
+static int
+dump_syn_reg_values(struct llcc_drv_data *drv, u32 bank, int err_type)
+{
+ struct llcc_edac_reg_data reg_data = edac_reg_data[err_type];
+ struct qcom_llcc_syn_regs regs = { };
+ int err_cnt, err_ways, ret, i;
+ u32 synd_reg, synd_val;
+
+ get_reg_offsets(drv, err_type, &regs);
+
+ for (i = 0; i < reg_data.reg_cnt; i++) {
+ synd_reg = regs.synd_reg + (i * 4);
+ ret = regmap_read(drv->regmaps[bank], synd_reg,
+ &synd_val);
+ if (ret)
+ goto clear;
+
+ edac_printk(KERN_CRIT, EDAC_LLCC, "%s: ECC_SYN%d: 0x%8x\n",
+ reg_data.name, i, synd_val);
+ }
+
+ ret = regmap_read(drv->regmaps[bank], regs.count_status_reg,
+ &err_cnt);
+ if (ret)
+ goto clear;
+
+ err_cnt &= reg_data.count_mask;
+ err_cnt >>= reg_data.count_shift;
+ edac_printk(KERN_CRIT, EDAC_LLCC, "%s: Error count: 0x%4x\n",
+ reg_data.name, err_cnt);
+
+ ret = regmap_read(drv->regmaps[bank], regs.ways_status_reg,
+ &err_ways);
+ if (ret)
+ goto clear;
+
+ err_ways &= reg_data.ways_mask;
+ err_ways >>= reg_data.ways_shift;
+
+ edac_printk(KERN_CRIT, EDAC_LLCC, "%s: Error ways: 0x%4x\n",
+ reg_data.name, err_ways);
+
+clear:
+ return qcom_llcc_clear_error_status(err_type, drv);
+}
+
+static int
+dump_syn_reg(struct edac_device_ctl_info *edev_ctl, int err_type, u32 bank)
+{
+ struct llcc_drv_data *drv = edev_ctl->dev->platform_data;
+ int ret;
+
+ ret = dump_syn_reg_values(drv, bank, err_type);
+ if (ret)
+ return ret;
+
+ switch (err_type) {
+ case LLCC_DRAM_CE:
+ edac_device_handle_ce(edev_ctl, 0, bank,
+ "LLCC Data RAM correctable Error");
+ break;
+ case LLCC_DRAM_UE:
+ edac_device_handle_ue(edev_ctl, 0, bank,
+ "LLCC Data RAM uncorrectable Error");
+ break;
+ case LLCC_TRAM_CE:
+ edac_device_handle_ce(edev_ctl, 0, bank,
+ "LLCC Tag RAM correctable Error");
+ break;
+ case LLCC_TRAM_UE:
+ edac_device_handle_ue(edev_ctl, 0, bank,
+ "LLCC Tag RAM uncorrectable Error");
+ break;
+ default:
+ ret = -EINVAL;
+ edac_printk(KERN_CRIT, EDAC_LLCC, "Unexpected error type: %d\n",
+ err_type);
+ }
+
+ return ret;
+}
+
+static irqreturn_t llcc_ecc_irq_handler(int irq, void *edev_ctl)
+{
+ struct edac_device_ctl_info *edac_dev_ctl = edev_ctl;
+ struct llcc_drv_data *drv = edac_dev_ctl->dev->platform_data;
+ irqreturn_t irq_rc = IRQ_NONE;
+ u32 drp_error, trp_error, i;
+ int ret;
+
+ /* Iterate over the banks and look for Tag RAM or Data RAM errors */
+ for (i = 0; i < drv->num_banks; i++) {
+ ret = regmap_read(drv->regmaps[i], drv->edac_reg_offset->drp_interrupt_status,
+ &drp_error);
+
+ if (!ret && (drp_error & SB_ECC_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_LLCC,
+ "Single Bit Error detected in Data RAM\n");
+ ret = dump_syn_reg(edev_ctl, LLCC_DRAM_CE, i);
+ } else if (!ret && (drp_error & DB_ECC_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_LLCC,
+ "Double Bit Error detected in Data RAM\n");
+ ret = dump_syn_reg(edev_ctl, LLCC_DRAM_UE, i);
+ }
+ if (!ret)
+ irq_rc = IRQ_HANDLED;
+
+ ret = regmap_read(drv->regmaps[i], drv->edac_reg_offset->trp_interrupt_0_status,
+ &trp_error);
+
+ if (!ret && (trp_error & SB_ECC_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_LLCC,
+ "Single Bit Error detected in Tag RAM\n");
+ ret = dump_syn_reg(edev_ctl, LLCC_TRAM_CE, i);
+ } else if (!ret && (trp_error & DB_ECC_ERROR)) {
+ edac_printk(KERN_CRIT, EDAC_LLCC,
+ "Double Bit Error detected in Tag RAM\n");
+ ret = dump_syn_reg(edev_ctl, LLCC_TRAM_UE, i);
+ }
+ if (!ret)
+ irq_rc = IRQ_HANDLED;
+ }
+
+ return irq_rc;
+}
+
+static void llcc_ecc_check(struct edac_device_ctl_info *edev_ctl)
+{
+ llcc_ecc_irq_handler(0, edev_ctl);
+}
+
+static int qcom_llcc_edac_probe(struct platform_device *pdev)
+{
+ struct llcc_drv_data *llcc_driv_data = pdev->dev.platform_data;
+ struct edac_device_ctl_info *edev_ctl;
+ struct device *dev = &pdev->dev;
+ int ecc_irq;
+ int rc;
+
+ rc = qcom_llcc_core_setup(llcc_driv_data, llcc_driv_data->bcast_regmap);
+ if (rc)
+ return rc;
+
+ /* Allocate edac control info */
+ edev_ctl = edac_device_alloc_ctl_info(0, "qcom-llcc", 1, "bank",
+ llcc_driv_data->num_banks, 1,
+ NULL, 0,
+ edac_device_alloc_index());
+
+ if (!edev_ctl)
+ return -ENOMEM;
+
+ edev_ctl->dev = dev;
+ edev_ctl->mod_name = dev_name(dev);
+ edev_ctl->dev_name = dev_name(dev);
+ edev_ctl->ctl_name = "llcc";
+ edev_ctl->panic_on_ue = LLCC_ERP_PANIC_ON_UE;
+
+ /* Check if LLCC driver has passed ECC IRQ */
+ ecc_irq = llcc_driv_data->ecc_irq;
+ if (ecc_irq > 0) {
+ /* Use interrupt mode if IRQ is available */
+ rc = devm_request_irq(dev, ecc_irq, llcc_ecc_irq_handler,
+ IRQF_TRIGGER_HIGH, "llcc_ecc", edev_ctl);
+ if (!rc) {
+ edac_op_state = EDAC_OPSTATE_INT;
+ goto irq_done;
+ }
+ }
+
+ /* Fall back to polling mode otherwise */
+ edev_ctl->poll_msec = ECC_POLL_MSEC;
+ edev_ctl->edac_check = llcc_ecc_check;
+ edac_op_state = EDAC_OPSTATE_POLL;
+
+irq_done:
+ rc = edac_device_add_device(edev_ctl);
+ if (rc) {
+ edac_device_free_ctl_info(edev_ctl);
+ return rc;
+ }
+
+ platform_set_drvdata(pdev, edev_ctl);
+
+ return rc;
+}
+
+static int qcom_llcc_edac_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *edev_ctl = dev_get_drvdata(&pdev->dev);
+
+ edac_device_del_device(edev_ctl->dev);
+ edac_device_free_ctl_info(edev_ctl);
+
+ return 0;
+}
+
+static const struct platform_device_id qcom_llcc_edac_id_table[] = {
+ { .name = "qcom_llcc_edac" },
+ {}
+};
+MODULE_DEVICE_TABLE(platform, qcom_llcc_edac_id_table);
+
+static struct platform_driver qcom_llcc_edac_driver = {
+ .probe = qcom_llcc_edac_probe,
+ .remove = qcom_llcc_edac_remove,
+ .driver = {
+ .name = "qcom_llcc_edac",
+ },
+ .id_table = qcom_llcc_edac_id_table,
+};
+module_platform_driver(qcom_llcc_edac_driver);
+
+MODULE_DESCRIPTION("QCOM EDAC driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/r82600_edac.c b/drivers/edac/r82600_edac.c
new file mode 100644
index 0000000000..61e979d543
--- /dev/null
+++ b/drivers/edac/r82600_edac.c
@@ -0,0 +1,426 @@
+/*
+ * Radisys 82600 Embedded chipset Memory Controller kernel module
+ * (C) 2005 EADS Astrium
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Written by Tim Small <tim@buttersideup.com>, based on work by Thayne
+ * Harbaugh, Dan Hollis <goemon at anime dot net> and others.
+ *
+ * $Id: edac_r82600.c,v 1.1.2.6 2005/10/05 00:43:44 dsp_llnl Exp $
+ *
+ * Written with reference to 82600 High Integration Dual PCI System
+ * Controller Data Book:
+ * www.radisys.com/files/support_downloads/007-01277-0002.82600DataBook.pdf
+ * references to this document given in []
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "r82600_edac"
+
+#define r82600_printk(level, fmt, arg...) \
+ edac_printk(level, "r82600", fmt, ##arg)
+
+#define r82600_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "r82600", fmt, ##arg)
+
+/* Radisys say "The 82600 integrates a main memory SDRAM controller that
+ * supports up to four banks of memory. The four banks can support a mix of
+ * sizes of 64 bit wide (72 bits with ECC) Synchronous DRAM (SDRAM) DIMMs,
+ * each of which can be any size from 16MB to 512MB. Both registered (control
+ * signals buffered) and unbuffered DIMM types are supported. Mixing of
+ * registered and unbuffered DIMMs as well as mixing of ECC and non-ECC DIMMs
+ * is not allowed. The 82600 SDRAM interface operates at the same frequency as
+ * the CPU bus, 66MHz, 100MHz or 133MHz."
+ */
+
+#define R82600_NR_CSROWS 4
+#define R82600_NR_CHANS 1
+#define R82600_NR_DIMMS 4
+
+#define R82600_BRIDGE_ID 0x8200
+
+/* Radisys 82600 register addresses - device 0 function 0 - PCI bridge */
+#define R82600_DRAMC 0x57 /* Various SDRAM related control bits
+ * all bits are R/W
+ *
+ * 7 SDRAM ISA Hole Enable
+ * 6 Flash Page Mode Enable
+ * 5 ECC Enable: 1=ECC 0=noECC
+ * 4 DRAM DIMM Type: 1=
+ * 3 BIOS Alias Disable
+ * 2 SDRAM BIOS Flash Write Enable
+ * 1:0 SDRAM Refresh Rate: 00=Disabled
+ * 01=7.8usec (256Mbit SDRAMs)
+ * 10=15.6us 11=125usec
+ */
+
+#define R82600_SDRAMC 0x76 /* "SDRAM Control Register"
+ * More SDRAM related control bits
+ * all bits are R/W
+ *
+ * 15:8 Reserved.
+ *
+ * 7:5 Special SDRAM Mode Select
+ *
+ * 4 Force ECC
+ *
+ * 1=Drive ECC bits to 0 during
+ * write cycles (i.e. ECC test mode)
+ *
+ * 0=Normal ECC functioning
+ *
+ * 3 Enhanced Paging Enable
+ *
+ * 2 CAS# Latency 0=3clks 1=2clks
+ *
+ * 1 RAS# to CAS# Delay 0=3 1=2
+ *
+ * 0 RAS# Precharge 0=3 1=2
+ */
+
+#define R82600_EAP 0x80 /* ECC Error Address Pointer Register
+ *
+ * 31 Disable Hardware Scrubbing (RW)
+ * 0=Scrub on corrected read
+ * 1=Don't scrub on corrected read
+ *
+ * 30:12 Error Address Pointer (RO)
+ * Upper 19 bits of error address
+ *
+ * 11:4 Syndrome Bits (RO)
+ *
+ * 3 BSERR# on multibit error (RW)
+ * 1=enable 0=disable
+ *
+ * 2 NMI on Single Bit Eror (RW)
+ * 1=NMI triggered by SBE n.b. other
+ * prerequeists
+ * 0=NMI not triggered
+ *
+ * 1 MBE (R/WC)
+ * read 1=MBE at EAP (see above)
+ * read 0=no MBE, or SBE occurred first
+ * write 1=Clear MBE status (must also
+ * clear SBE)
+ * write 0=NOP
+ *
+ * 1 SBE (R/WC)
+ * read 1=SBE at EAP (see above)
+ * read 0=no SBE, or MBE occurred first
+ * write 1=Clear SBE status (must also
+ * clear MBE)
+ * write 0=NOP
+ */
+
+#define R82600_DRBA 0x60 /* + 0x60..0x63 SDRAM Row Boundary Address
+ * Registers
+ *
+ * 7:0 Address lines 30:24 - upper limit of
+ * each row [p57]
+ */
+
+struct r82600_error_info {
+ u32 eapr;
+};
+
+static bool disable_hardware_scrub;
+
+static struct edac_pci_ctl_info *r82600_pci;
+
+static void r82600_get_error_info(struct mem_ctl_info *mci,
+ struct r82600_error_info *info)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+ pci_read_config_dword(pdev, R82600_EAP, &info->eapr);
+
+ if (info->eapr & BIT(0))
+ /* Clear error to allow next error to be reported [p.62] */
+ pci_write_bits32(pdev, R82600_EAP,
+ ((u32) BIT(0) & (u32) BIT(1)),
+ ((u32) BIT(0) & (u32) BIT(1)));
+
+ if (info->eapr & BIT(1))
+ /* Clear error to allow next error to be reported [p.62] */
+ pci_write_bits32(pdev, R82600_EAP,
+ ((u32) BIT(0) & (u32) BIT(1)),
+ ((u32) BIT(0) & (u32) BIT(1)));
+}
+
+static int r82600_process_error_info(struct mem_ctl_info *mci,
+ struct r82600_error_info *info,
+ int handle_errors)
+{
+ int error_found;
+ u32 eapaddr, page;
+ u32 syndrome;
+
+ error_found = 0;
+
+ /* bits 30:12 store the upper 19 bits of the 32 bit error address */
+ eapaddr = ((info->eapr >> 12) & 0x7FFF) << 13;
+ /* Syndrome in bits 11:4 [p.62] */
+ syndrome = (info->eapr >> 4) & 0xFF;
+
+ /* the R82600 reports at less than page *
+ * granularity (upper 19 bits only) */
+ page = eapaddr >> PAGE_SHIFT;
+
+ if (info->eapr & BIT(0)) { /* CE? */
+ error_found = 1;
+
+ if (handle_errors)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ page, 0, syndrome,
+ edac_mc_find_csrow_by_page(mci, page),
+ 0, -1,
+ mci->ctl_name, "");
+ }
+
+ if (info->eapr & BIT(1)) { /* UE? */
+ error_found = 1;
+
+ if (handle_errors)
+ /* 82600 doesn't give enough info */
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ page, 0, 0,
+ edac_mc_find_csrow_by_page(mci, page),
+ 0, -1,
+ mci->ctl_name, "");
+ }
+
+ return error_found;
+}
+
+static void r82600_check(struct mem_ctl_info *mci)
+{
+ struct r82600_error_info info;
+
+ r82600_get_error_info(mci, &info);
+ r82600_process_error_info(mci, &info, 1);
+}
+
+static inline int ecc_enabled(u8 dramcr)
+{
+ return dramcr & BIT(5);
+}
+
+static void r82600_init_csrows(struct mem_ctl_info *mci, struct pci_dev *pdev,
+ u8 dramcr)
+{
+ struct csrow_info *csrow;
+ struct dimm_info *dimm;
+ int index;
+ u8 drbar; /* SDRAM Row Boundary Address Register */
+ u32 row_high_limit, row_high_limit_last;
+ u32 reg_sdram, ecc_on, row_base;
+
+ ecc_on = ecc_enabled(dramcr);
+ reg_sdram = dramcr & BIT(4);
+ row_high_limit_last = 0;
+
+ for (index = 0; index < mci->nr_csrows; index++) {
+ csrow = mci->csrows[index];
+ dimm = csrow->channels[0]->dimm;
+
+ /* find the DRAM Chip Select Base address and mask */
+ pci_read_config_byte(pdev, R82600_DRBA + index, &drbar);
+
+ edac_dbg(1, "Row=%d DRBA = %#0x\n", index, drbar);
+
+ row_high_limit = ((u32) drbar << 24);
+/* row_high_limit = ((u32)drbar << 24) | 0xffffffUL; */
+
+ edac_dbg(1, "Row=%d, Boundary Address=%#0x, Last = %#0x\n",
+ index, row_high_limit, row_high_limit_last);
+
+ /* Empty row [p.57] */
+ if (row_high_limit == row_high_limit_last)
+ continue;
+
+ row_base = row_high_limit_last;
+
+ csrow->first_page = row_base >> PAGE_SHIFT;
+ csrow->last_page = (row_high_limit >> PAGE_SHIFT) - 1;
+
+ dimm->nr_pages = csrow->last_page - csrow->first_page + 1;
+ /* Error address is top 19 bits - so granularity is *
+ * 14 bits */
+ dimm->grain = 1 << 14;
+ dimm->mtype = reg_sdram ? MEM_RDDR : MEM_DDR;
+ /* FIXME - check that this is unknowable with this chipset */
+ dimm->dtype = DEV_UNKNOWN;
+
+ /* Mode is global on 82600 */
+ dimm->edac_mode = ecc_on ? EDAC_SECDED : EDAC_NONE;
+ row_high_limit_last = row_high_limit;
+ }
+}
+
+static int r82600_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ u8 dramcr;
+ u32 eapr;
+ u32 scrub_disabled;
+ u32 sdram_refresh_rate;
+ struct r82600_error_info discard;
+
+ edac_dbg(0, "\n");
+ pci_read_config_byte(pdev, R82600_DRAMC, &dramcr);
+ pci_read_config_dword(pdev, R82600_EAP, &eapr);
+ scrub_disabled = eapr & BIT(31);
+ sdram_refresh_rate = dramcr & (BIT(0) | BIT(1));
+ edac_dbg(2, "sdram refresh rate = %#0x\n", sdram_refresh_rate);
+ edac_dbg(2, "DRAMC register = %#0x\n", dramcr);
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = R82600_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = R82600_NR_CHANS;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (mci == NULL)
+ return -ENOMEM;
+
+ edac_dbg(0, "mci = %p\n", mci);
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_DDR;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+ /* FIXME try to work out if the chip leads have been used for COM2
+ * instead on this board? [MA6?] MAYBE:
+ */
+
+ /* On the R82600, the pins for memory bits 72:65 - i.e. the *
+ * EC bits are shared with the pins for COM2 (!), so if COM2 *
+ * is enabled, we assume COM2 is wired up, and thus no EDAC *
+ * is possible. */
+ mci->edac_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
+
+ if (ecc_enabled(dramcr)) {
+ if (scrub_disabled)
+ edac_dbg(3, "mci = %p - Scrubbing disabled! EAP: %#0x\n",
+ mci, eapr);
+ } else
+ mci->edac_cap = EDAC_FLAG_NONE;
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = "R82600";
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = r82600_check;
+ mci->ctl_page_to_phys = NULL;
+ r82600_init_csrows(mci, pdev, dramcr);
+ r82600_get_error_info(mci, &discard); /* clear counters */
+
+ /* Here we assume that we will never see multiple instances of this
+ * type of memory controller. The ID is therefore hardcoded to 0.
+ */
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* get this far and it's successful */
+
+ if (disable_hardware_scrub) {
+ edac_dbg(3, "Disabling Hardware Scrub (scrub on error)\n");
+ pci_write_bits32(pdev, R82600_EAP, BIT(31), BIT(31));
+ }
+
+ /* allocating generic PCI control info */
+ r82600_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!r82600_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ edac_dbg(3, "success\n");
+ return 0;
+
+fail:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/* returns count (>= 0), or negative on error */
+static int r82600_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ edac_dbg(0, "\n");
+
+ /* don't need to call pci_enable_device() */
+ return r82600_probe1(pdev, ent->driver_data);
+}
+
+static void r82600_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ if (r82600_pci)
+ edac_pci_release_generic_ctl(r82600_pci);
+
+ if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
+ return;
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id r82600_pci_tbl[] = {
+ {
+ PCI_DEVICE(PCI_VENDOR_ID_RADISYS, R82600_BRIDGE_ID)
+ },
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, r82600_pci_tbl);
+
+static struct pci_driver r82600_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = r82600_init_one,
+ .remove = r82600_remove_one,
+ .id_table = r82600_pci_tbl,
+};
+
+static int __init r82600_init(void)
+{
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ return pci_register_driver(&r82600_driver);
+}
+
+static void __exit r82600_exit(void)
+{
+ pci_unregister_driver(&r82600_driver);
+}
+
+module_init(r82600_init);
+module_exit(r82600_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Tim Small <tim@buttersideup.com> - WPAD Ltd. on behalf of EADS Astrium");
+MODULE_DESCRIPTION("MC support for Radisys 82600 memory controllers");
+
+module_param(disable_hardware_scrub, bool, 0644);
+MODULE_PARM_DESC(disable_hardware_scrub,
+ "If set, disable the chipset's automatic scrub for CEs");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/sb_edac.c b/drivers/edac/sb_edac.c
new file mode 100644
index 0000000000..0c779a0326
--- /dev/null
+++ b/drivers/edac/sb_edac.c
@@ -0,0 +1,3688 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Intel Sandy Bridge -EN/-EP/-EX Memory Controller kernel module
+ *
+ * This driver supports the memory controllers found on the Intel
+ * processor family Sandy Bridge.
+ *
+ * Copyright (c) 2011 by:
+ * Mauro Carvalho Chehab
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+#include <linux/smp.h>
+#include <linux/bitmap.h>
+#include <linux/math64.h>
+#include <linux/mod_devicetable.h>
+#include <asm/cpu_device_id.h>
+#include <asm/intel-family.h>
+#include <asm/processor.h>
+#include <asm/mce.h>
+
+#include "edac_module.h"
+
+/* Static vars */
+static LIST_HEAD(sbridge_edac_list);
+
+/*
+ * Alter this version for the module when modifications are made
+ */
+#define SBRIDGE_REVISION " Ver: 1.1.2 "
+#define EDAC_MOD_STR "sb_edac"
+
+/*
+ * Debug macros
+ */
+#define sbridge_printk(level, fmt, arg...) \
+ edac_printk(level, "sbridge", fmt, ##arg)
+
+#define sbridge_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "sbridge", fmt, ##arg)
+
+/*
+ * Get a bit field at register value <v>, from bit <lo> to bit <hi>
+ */
+#define GET_BITFIELD(v, lo, hi) \
+ (((v) & GENMASK_ULL(hi, lo)) >> (lo))
+
+/* Devices 12 Function 6, Offsets 0x80 to 0xcc */
+static const u32 sbridge_dram_rule[] = {
+ 0x80, 0x88, 0x90, 0x98, 0xa0,
+ 0xa8, 0xb0, 0xb8, 0xc0, 0xc8,
+};
+
+static const u32 ibridge_dram_rule[] = {
+ 0x60, 0x68, 0x70, 0x78, 0x80,
+ 0x88, 0x90, 0x98, 0xa0, 0xa8,
+ 0xb0, 0xb8, 0xc0, 0xc8, 0xd0,
+ 0xd8, 0xe0, 0xe8, 0xf0, 0xf8,
+};
+
+static const u32 knl_dram_rule[] = {
+ 0x60, 0x68, 0x70, 0x78, 0x80, /* 0-4 */
+ 0x88, 0x90, 0x98, 0xa0, 0xa8, /* 5-9 */
+ 0xb0, 0xb8, 0xc0, 0xc8, 0xd0, /* 10-14 */
+ 0xd8, 0xe0, 0xe8, 0xf0, 0xf8, /* 15-19 */
+ 0x100, 0x108, 0x110, 0x118, /* 20-23 */
+};
+
+#define DRAM_RULE_ENABLE(reg) GET_BITFIELD(reg, 0, 0)
+#define A7MODE(reg) GET_BITFIELD(reg, 26, 26)
+
+static char *show_dram_attr(u32 attr)
+{
+ switch (attr) {
+ case 0:
+ return "DRAM";
+ case 1:
+ return "MMCFG";
+ case 2:
+ return "NXM";
+ default:
+ return "unknown";
+ }
+}
+
+static const u32 sbridge_interleave_list[] = {
+ 0x84, 0x8c, 0x94, 0x9c, 0xa4,
+ 0xac, 0xb4, 0xbc, 0xc4, 0xcc,
+};
+
+static const u32 ibridge_interleave_list[] = {
+ 0x64, 0x6c, 0x74, 0x7c, 0x84,
+ 0x8c, 0x94, 0x9c, 0xa4, 0xac,
+ 0xb4, 0xbc, 0xc4, 0xcc, 0xd4,
+ 0xdc, 0xe4, 0xec, 0xf4, 0xfc,
+};
+
+static const u32 knl_interleave_list[] = {
+ 0x64, 0x6c, 0x74, 0x7c, 0x84, /* 0-4 */
+ 0x8c, 0x94, 0x9c, 0xa4, 0xac, /* 5-9 */
+ 0xb4, 0xbc, 0xc4, 0xcc, 0xd4, /* 10-14 */
+ 0xdc, 0xe4, 0xec, 0xf4, 0xfc, /* 15-19 */
+ 0x104, 0x10c, 0x114, 0x11c, /* 20-23 */
+};
+#define MAX_INTERLEAVE \
+ (max_t(unsigned int, ARRAY_SIZE(sbridge_interleave_list), \
+ max_t(unsigned int, ARRAY_SIZE(ibridge_interleave_list), \
+ ARRAY_SIZE(knl_interleave_list))))
+
+struct interleave_pkg {
+ unsigned char start;
+ unsigned char end;
+};
+
+static const struct interleave_pkg sbridge_interleave_pkg[] = {
+ { 0, 2 },
+ { 3, 5 },
+ { 8, 10 },
+ { 11, 13 },
+ { 16, 18 },
+ { 19, 21 },
+ { 24, 26 },
+ { 27, 29 },
+};
+
+static const struct interleave_pkg ibridge_interleave_pkg[] = {
+ { 0, 3 },
+ { 4, 7 },
+ { 8, 11 },
+ { 12, 15 },
+ { 16, 19 },
+ { 20, 23 },
+ { 24, 27 },
+ { 28, 31 },
+};
+
+static inline int sad_pkg(const struct interleave_pkg *table, u32 reg,
+ int interleave)
+{
+ return GET_BITFIELD(reg, table[interleave].start,
+ table[interleave].end);
+}
+
+/* Devices 12 Function 7 */
+
+#define TOLM 0x80
+#define TOHM 0x84
+#define HASWELL_TOLM 0xd0
+#define HASWELL_TOHM_0 0xd4
+#define HASWELL_TOHM_1 0xd8
+#define KNL_TOLM 0xd0
+#define KNL_TOHM_0 0xd4
+#define KNL_TOHM_1 0xd8
+
+#define GET_TOLM(reg) ((GET_BITFIELD(reg, 0, 3) << 28) | 0x3ffffff)
+#define GET_TOHM(reg) ((GET_BITFIELD(reg, 0, 20) << 25) | 0x3ffffff)
+
+/* Device 13 Function 6 */
+
+#define SAD_TARGET 0xf0
+
+#define SOURCE_ID(reg) GET_BITFIELD(reg, 9, 11)
+
+#define SOURCE_ID_KNL(reg) GET_BITFIELD(reg, 12, 14)
+
+#define SAD_CONTROL 0xf4
+
+/* Device 14 function 0 */
+
+static const u32 tad_dram_rule[] = {
+ 0x40, 0x44, 0x48, 0x4c,
+ 0x50, 0x54, 0x58, 0x5c,
+ 0x60, 0x64, 0x68, 0x6c,
+};
+#define MAX_TAD ARRAY_SIZE(tad_dram_rule)
+
+#define TAD_LIMIT(reg) ((GET_BITFIELD(reg, 12, 31) << 26) | 0x3ffffff)
+#define TAD_SOCK(reg) GET_BITFIELD(reg, 10, 11)
+#define TAD_CH(reg) GET_BITFIELD(reg, 8, 9)
+#define TAD_TGT3(reg) GET_BITFIELD(reg, 6, 7)
+#define TAD_TGT2(reg) GET_BITFIELD(reg, 4, 5)
+#define TAD_TGT1(reg) GET_BITFIELD(reg, 2, 3)
+#define TAD_TGT0(reg) GET_BITFIELD(reg, 0, 1)
+
+/* Device 15, function 0 */
+
+#define MCMTR 0x7c
+#define KNL_MCMTR 0x624
+
+#define IS_ECC_ENABLED(mcmtr) GET_BITFIELD(mcmtr, 2, 2)
+#define IS_LOCKSTEP_ENABLED(mcmtr) GET_BITFIELD(mcmtr, 1, 1)
+#define IS_CLOSE_PG(mcmtr) GET_BITFIELD(mcmtr, 0, 0)
+
+/* Device 15, function 1 */
+
+#define RASENABLES 0xac
+#define IS_MIRROR_ENABLED(reg) GET_BITFIELD(reg, 0, 0)
+
+/* Device 15, functions 2-5 */
+
+static const int mtr_regs[] = {
+ 0x80, 0x84, 0x88,
+};
+
+static const int knl_mtr_reg = 0xb60;
+
+#define RANK_DISABLE(mtr) GET_BITFIELD(mtr, 16, 19)
+#define IS_DIMM_PRESENT(mtr) GET_BITFIELD(mtr, 14, 14)
+#define RANK_CNT_BITS(mtr) GET_BITFIELD(mtr, 12, 13)
+#define RANK_WIDTH_BITS(mtr) GET_BITFIELD(mtr, 2, 4)
+#define COL_WIDTH_BITS(mtr) GET_BITFIELD(mtr, 0, 1)
+
+static const u32 tad_ch_nilv_offset[] = {
+ 0x90, 0x94, 0x98, 0x9c,
+ 0xa0, 0xa4, 0xa8, 0xac,
+ 0xb0, 0xb4, 0xb8, 0xbc,
+};
+#define CHN_IDX_OFFSET(reg) GET_BITFIELD(reg, 28, 29)
+#define TAD_OFFSET(reg) (GET_BITFIELD(reg, 6, 25) << 26)
+
+static const u32 rir_way_limit[] = {
+ 0x108, 0x10c, 0x110, 0x114, 0x118,
+};
+#define MAX_RIR_RANGES ARRAY_SIZE(rir_way_limit)
+
+#define IS_RIR_VALID(reg) GET_BITFIELD(reg, 31, 31)
+#define RIR_WAY(reg) GET_BITFIELD(reg, 28, 29)
+
+#define MAX_RIR_WAY 8
+
+static const u32 rir_offset[MAX_RIR_RANGES][MAX_RIR_WAY] = {
+ { 0x120, 0x124, 0x128, 0x12c, 0x130, 0x134, 0x138, 0x13c },
+ { 0x140, 0x144, 0x148, 0x14c, 0x150, 0x154, 0x158, 0x15c },
+ { 0x160, 0x164, 0x168, 0x16c, 0x170, 0x174, 0x178, 0x17c },
+ { 0x180, 0x184, 0x188, 0x18c, 0x190, 0x194, 0x198, 0x19c },
+ { 0x1a0, 0x1a4, 0x1a8, 0x1ac, 0x1b0, 0x1b4, 0x1b8, 0x1bc },
+};
+
+#define RIR_RNK_TGT(type, reg) (((type) == BROADWELL) ? \
+ GET_BITFIELD(reg, 20, 23) : GET_BITFIELD(reg, 16, 19))
+
+#define RIR_OFFSET(type, reg) (((type) == HASWELL || (type) == BROADWELL) ? \
+ GET_BITFIELD(reg, 2, 15) : GET_BITFIELD(reg, 2, 14))
+
+/* Device 16, functions 2-7 */
+
+/*
+ * FIXME: Implement the error count reads directly
+ */
+
+#define RANK_ODD_OV(reg) GET_BITFIELD(reg, 31, 31)
+#define RANK_ODD_ERR_CNT(reg) GET_BITFIELD(reg, 16, 30)
+#define RANK_EVEN_OV(reg) GET_BITFIELD(reg, 15, 15)
+#define RANK_EVEN_ERR_CNT(reg) GET_BITFIELD(reg, 0, 14)
+
+#if 0 /* Currently unused*/
+static const u32 correrrcnt[] = {
+ 0x104, 0x108, 0x10c, 0x110,
+};
+
+static const u32 correrrthrsld[] = {
+ 0x11c, 0x120, 0x124, 0x128,
+};
+#endif
+
+#define RANK_ODD_ERR_THRSLD(reg) GET_BITFIELD(reg, 16, 30)
+#define RANK_EVEN_ERR_THRSLD(reg) GET_BITFIELD(reg, 0, 14)
+
+
+/* Device 17, function 0 */
+
+#define SB_RANK_CFG_A 0x0328
+
+#define IB_RANK_CFG_A 0x0320
+
+/*
+ * sbridge structs
+ */
+
+#define NUM_CHANNELS 6 /* Max channels per MC */
+#define MAX_DIMMS 3 /* Max DIMMS per channel */
+#define KNL_MAX_CHAS 38 /* KNL max num. of Cache Home Agents */
+#define KNL_MAX_CHANNELS 6 /* KNL max num. of PCI channels */
+#define KNL_MAX_EDCS 8 /* Embedded DRAM controllers */
+#define CHANNEL_UNSPECIFIED 0xf /* Intel IA32 SDM 15-14 */
+
+enum type {
+ SANDY_BRIDGE,
+ IVY_BRIDGE,
+ HASWELL,
+ BROADWELL,
+ KNIGHTS_LANDING,
+};
+
+enum domain {
+ IMC0 = 0,
+ IMC1,
+ SOCK,
+};
+
+enum mirroring_mode {
+ NON_MIRRORING,
+ ADDR_RANGE_MIRRORING,
+ FULL_MIRRORING,
+};
+
+struct sbridge_pvt;
+struct sbridge_info {
+ enum type type;
+ u32 mcmtr;
+ u32 rankcfgr;
+ u64 (*get_tolm)(struct sbridge_pvt *pvt);
+ u64 (*get_tohm)(struct sbridge_pvt *pvt);
+ u64 (*rir_limit)(u32 reg);
+ u64 (*sad_limit)(u32 reg);
+ u32 (*interleave_mode)(u32 reg);
+ u32 (*dram_attr)(u32 reg);
+ const u32 *dram_rule;
+ const u32 *interleave_list;
+ const struct interleave_pkg *interleave_pkg;
+ u8 max_sad;
+ u8 (*get_node_id)(struct sbridge_pvt *pvt);
+ u8 (*get_ha)(u8 bank);
+ enum mem_type (*get_memory_type)(struct sbridge_pvt *pvt);
+ enum dev_type (*get_width)(struct sbridge_pvt *pvt, u32 mtr);
+ struct pci_dev *pci_vtd;
+};
+
+struct sbridge_channel {
+ u32 ranks;
+ u32 dimms;
+ struct dimm {
+ u32 rowbits;
+ u32 colbits;
+ u32 bank_xor_enable;
+ u32 amap_fine;
+ } dimm[MAX_DIMMS];
+};
+
+struct pci_id_descr {
+ int dev_id;
+ int optional;
+ enum domain dom;
+};
+
+struct pci_id_table {
+ const struct pci_id_descr *descr;
+ int n_devs_per_imc;
+ int n_devs_per_sock;
+ int n_imcs_per_sock;
+ enum type type;
+};
+
+struct sbridge_dev {
+ struct list_head list;
+ int seg;
+ u8 bus, mc;
+ u8 node_id, source_id;
+ struct pci_dev **pdev;
+ enum domain dom;
+ int n_devs;
+ int i_devs;
+ struct mem_ctl_info *mci;
+};
+
+struct knl_pvt {
+ struct pci_dev *pci_cha[KNL_MAX_CHAS];
+ struct pci_dev *pci_channel[KNL_MAX_CHANNELS];
+ struct pci_dev *pci_mc0;
+ struct pci_dev *pci_mc1;
+ struct pci_dev *pci_mc0_misc;
+ struct pci_dev *pci_mc1_misc;
+ struct pci_dev *pci_mc_info; /* tolm, tohm */
+};
+
+struct sbridge_pvt {
+ /* Devices per socket */
+ struct pci_dev *pci_ddrio;
+ struct pci_dev *pci_sad0, *pci_sad1;
+ struct pci_dev *pci_br0, *pci_br1;
+ /* Devices per memory controller */
+ struct pci_dev *pci_ha, *pci_ta, *pci_ras;
+ struct pci_dev *pci_tad[NUM_CHANNELS];
+
+ struct sbridge_dev *sbridge_dev;
+
+ struct sbridge_info info;
+ struct sbridge_channel channel[NUM_CHANNELS];
+
+ /* Memory type detection */
+ bool is_cur_addr_mirrored, is_lockstep, is_close_pg;
+ bool is_chan_hash;
+ enum mirroring_mode mirror_mode;
+
+ /* Memory description */
+ u64 tolm, tohm;
+ struct knl_pvt knl;
+};
+
+#define PCI_DESCR(device_id, opt, domain) \
+ .dev_id = (device_id), \
+ .optional = opt, \
+ .dom = domain
+
+static const struct pci_id_descr pci_dev_descr_sbridge[] = {
+ /* Processor Home Agent */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0, 0, IMC0) },
+
+ /* Memory controller */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_RAS, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD1, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD2, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD3, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO, 1, SOCK) },
+
+ /* System Address Decoder */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0, 0, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1, 0, SOCK) },
+
+ /* Broadcast Registers */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_BR, 0, SOCK) },
+};
+
+#define PCI_ID_TABLE_ENTRY(A, N, M, T) { \
+ .descr = A, \
+ .n_devs_per_imc = N, \
+ .n_devs_per_sock = ARRAY_SIZE(A), \
+ .n_imcs_per_sock = M, \
+ .type = T \
+}
+
+static const struct pci_id_table pci_dev_descr_sbridge_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_sbridge, ARRAY_SIZE(pci_dev_descr_sbridge), 1, SANDY_BRIDGE),
+ {0,} /* 0 terminated list. */
+};
+
+/* This changes depending if 1HA or 2HA:
+ * 1HA:
+ * 0x0eb8 (17.0) is DDRIO0
+ * 2HA:
+ * 0x0ebc (17.4) is DDRIO0
+ */
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0 0x0eb8
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0 0x0ebc
+
+/* pci ids */
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0 0x0ea0
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA 0x0ea8
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS 0x0e71
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0 0x0eaa
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1 0x0eab
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2 0x0eac
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3 0x0ead
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_SAD 0x0ec8
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_BR0 0x0ec9
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_BR1 0x0eca
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1 0x0e60
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA 0x0e68
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS 0x0e79
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0 0x0e6a
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1 0x0e6b
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2 0x0e6c
+#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3 0x0e6d
+
+static const struct pci_id_descr pci_dev_descr_ibridge[] = {
+ /* Processor Home Agent */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1, 1, IMC1) },
+
+ /* Memory controller */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3, 0, IMC0) },
+
+ /* Optional, mode 2HA */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3, 1, IMC1) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0, 1, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0, 1, SOCK) },
+
+ /* System Address Decoder */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_SAD, 0, SOCK) },
+
+ /* Broadcast Registers */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR0, 1, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR1, 0, SOCK) },
+
+};
+
+static const struct pci_id_table pci_dev_descr_ibridge_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_ibridge, 12, 2, IVY_BRIDGE),
+ {0,} /* 0 terminated list. */
+};
+
+/* Haswell support */
+/* EN processor:
+ * - 1 IMC
+ * - 3 DDR3 channels, 2 DPC per channel
+ * EP processor:
+ * - 1 or 2 IMC
+ * - 4 DDR4 channels, 3 DPC per channel
+ * EP 4S processor:
+ * - 2 IMC
+ * - 4 DDR4 channels, 3 DPC per channel
+ * EX processor:
+ * - 2 IMC
+ * - each IMC interfaces with a SMI 2 channel
+ * - each SMI channel interfaces with a scalable memory buffer
+ * - each scalable memory buffer supports 4 DDR3/DDR4 channels, 3 DPC
+ */
+#define HASWELL_DDRCRCLKCONTROLS 0xa10 /* Ditto on Broadwell */
+#define HASWELL_HASYSDEFEATURE2 0x84
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_VTD_MISC 0x2f28
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0 0x2fa0
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1 0x2f60
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA 0x2fa8
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TM 0x2f71
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA 0x2f68
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TM 0x2f79
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0 0x2ffc
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1 0x2ffd
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0 0x2faa
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1 0x2fab
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2 0x2fac
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3 0x2fad
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0 0x2f6a
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1 0x2f6b
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2 0x2f6c
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3 0x2f6d
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0 0x2fbd
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1 0x2fbf
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2 0x2fb9
+#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3 0x2fbb
+static const struct pci_id_descr pci_dev_descr_haswell[] = {
+ /* first item must be the HA */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1, 1, IMC1) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TM, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2, 1, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3, 1, IMC0) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TM, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3, 1, IMC1) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0, 0, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1, 0, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0, 1, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1, 1, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2, 1, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3, 1, SOCK) },
+};
+
+static const struct pci_id_table pci_dev_descr_haswell_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_haswell, 13, 2, HASWELL),
+ {0,} /* 0 terminated list. */
+};
+
+/* Knight's Landing Support */
+/*
+ * KNL's memory channels are swizzled between memory controllers.
+ * MC0 is mapped to CH3,4,5 and MC1 is mapped to CH0,1,2
+ */
+#define knl_channel_remap(mc, chan) ((mc) ? (chan) : (chan) + 3)
+
+/* Memory controller, TAD tables, error injection - 2-8-0, 2-9-0 (2 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_MC 0x7840
+/* DRAM channel stuff; bank addrs, dimmmtr, etc.. 2-8-2 - 2-9-4 (6 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_CHAN 0x7843
+/* kdrwdbu TAD limits/offsets, MCMTR - 2-10-1, 2-11-1 (2 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_TA 0x7844
+/* CHA broadcast registers, dram rules - 1-29-0 (1 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0 0x782a
+/* SAD target - 1-29-1 (1 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1 0x782b
+/* Caching / Home Agent */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_CHA 0x782c
+/* Device with TOLM and TOHM, 0-5-0 (1 of these) */
+#define PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM 0x7810
+
+/*
+ * KNL differs from SB, IB, and Haswell in that it has multiple
+ * instances of the same device with the same device ID, so we handle that
+ * by creating as many copies in the table as we expect to find.
+ * (Like device ID must be grouped together.)
+ */
+
+static const struct pci_id_descr pci_dev_descr_knl[] = {
+ [0 ... 1] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_MC, 0, IMC0)},
+ [2 ... 7] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_CHAN, 0, IMC0) },
+ [8] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_TA, 0, IMC0) },
+ [9] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM, 0, IMC0) },
+ [10] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0, 0, SOCK) },
+ [11] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1, 0, SOCK) },
+ [12 ... 49] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_CHA, 0, SOCK) },
+};
+
+static const struct pci_id_table pci_dev_descr_knl_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_knl, ARRAY_SIZE(pci_dev_descr_knl), 1, KNIGHTS_LANDING),
+ {0,}
+};
+
+/*
+ * Broadwell support
+ *
+ * DE processor:
+ * - 1 IMC
+ * - 2 DDR3 channels, 2 DPC per channel
+ * EP processor:
+ * - 1 or 2 IMC
+ * - 4 DDR4 channels, 3 DPC per channel
+ * EP 4S processor:
+ * - 2 IMC
+ * - 4 DDR4 channels, 3 DPC per channel
+ * EX processor:
+ * - 2 IMC
+ * - each IMC interfaces with a SMI 2 channel
+ * - each SMI channel interfaces with a scalable memory buffer
+ * - each scalable memory buffer supports 4 DDR3/DDR4 channels, 3 DPC
+ */
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_VTD_MISC 0x6f28
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0 0x6fa0
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1 0x6f60
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA 0x6fa8
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TM 0x6f71
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA 0x6f68
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TM 0x6f79
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0 0x6ffc
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1 0x6ffd
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0 0x6faa
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1 0x6fab
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2 0x6fac
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3 0x6fad
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0 0x6f6a
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1 0x6f6b
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2 0x6f6c
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3 0x6f6d
+#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0 0x6faf
+
+static const struct pci_id_descr pci_dev_descr_broadwell[] = {
+ /* first item must be the HA */
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1, 1, IMC1) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TM, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1, 0, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2, 1, IMC0) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3, 1, IMC0) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TM, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2, 1, IMC1) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3, 1, IMC1) },
+
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0, 0, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1, 0, SOCK) },
+ { PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0, 1, SOCK) },
+};
+
+static const struct pci_id_table pci_dev_descr_broadwell_table[] = {
+ PCI_ID_TABLE_ENTRY(pci_dev_descr_broadwell, 10, 2, BROADWELL),
+ {0,} /* 0 terminated list. */
+};
+
+
+/****************************************************************************
+ Ancillary status routines
+ ****************************************************************************/
+
+static inline int numrank(enum type type, u32 mtr)
+{
+ int ranks = (1 << RANK_CNT_BITS(mtr));
+ int max = 4;
+
+ if (type == HASWELL || type == BROADWELL || type == KNIGHTS_LANDING)
+ max = 8;
+
+ if (ranks > max) {
+ edac_dbg(0, "Invalid number of ranks: %d (max = %i) raw value = %x (%04x)\n",
+ ranks, max, (unsigned int)RANK_CNT_BITS(mtr), mtr);
+ return -EINVAL;
+ }
+
+ return ranks;
+}
+
+static inline int numrow(u32 mtr)
+{
+ int rows = (RANK_WIDTH_BITS(mtr) + 12);
+
+ if (rows < 13 || rows > 18) {
+ edac_dbg(0, "Invalid number of rows: %d (should be between 14 and 17) raw value = %x (%04x)\n",
+ rows, (unsigned int)RANK_WIDTH_BITS(mtr), mtr);
+ return -EINVAL;
+ }
+
+ return 1 << rows;
+}
+
+static inline int numcol(u32 mtr)
+{
+ int cols = (COL_WIDTH_BITS(mtr) + 10);
+
+ if (cols > 12) {
+ edac_dbg(0, "Invalid number of cols: %d (max = 4) raw value = %x (%04x)\n",
+ cols, (unsigned int)COL_WIDTH_BITS(mtr), mtr);
+ return -EINVAL;
+ }
+
+ return 1 << cols;
+}
+
+static struct sbridge_dev *get_sbridge_dev(int seg, u8 bus, enum domain dom,
+ int multi_bus,
+ struct sbridge_dev *prev)
+{
+ struct sbridge_dev *sbridge_dev;
+
+ /*
+ * If we have devices scattered across several busses that pertain
+ * to the same memory controller, we'll lump them all together.
+ */
+ if (multi_bus) {
+ return list_first_entry_or_null(&sbridge_edac_list,
+ struct sbridge_dev, list);
+ }
+
+ sbridge_dev = list_entry(prev ? prev->list.next
+ : sbridge_edac_list.next, struct sbridge_dev, list);
+
+ list_for_each_entry_from(sbridge_dev, &sbridge_edac_list, list) {
+ if ((sbridge_dev->seg == seg) && (sbridge_dev->bus == bus) &&
+ (dom == SOCK || dom == sbridge_dev->dom))
+ return sbridge_dev;
+ }
+
+ return NULL;
+}
+
+static struct sbridge_dev *alloc_sbridge_dev(int seg, u8 bus, enum domain dom,
+ const struct pci_id_table *table)
+{
+ struct sbridge_dev *sbridge_dev;
+
+ sbridge_dev = kzalloc(sizeof(*sbridge_dev), GFP_KERNEL);
+ if (!sbridge_dev)
+ return NULL;
+
+ sbridge_dev->pdev = kcalloc(table->n_devs_per_imc,
+ sizeof(*sbridge_dev->pdev),
+ GFP_KERNEL);
+ if (!sbridge_dev->pdev) {
+ kfree(sbridge_dev);
+ return NULL;
+ }
+
+ sbridge_dev->seg = seg;
+ sbridge_dev->bus = bus;
+ sbridge_dev->dom = dom;
+ sbridge_dev->n_devs = table->n_devs_per_imc;
+ list_add_tail(&sbridge_dev->list, &sbridge_edac_list);
+
+ return sbridge_dev;
+}
+
+static void free_sbridge_dev(struct sbridge_dev *sbridge_dev)
+{
+ list_del(&sbridge_dev->list);
+ kfree(sbridge_dev->pdev);
+ kfree(sbridge_dev);
+}
+
+static u64 sbridge_get_tolm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ /* Address range is 32:28 */
+ pci_read_config_dword(pvt->pci_sad1, TOLM, &reg);
+ return GET_TOLM(reg);
+}
+
+static u64 sbridge_get_tohm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->pci_sad1, TOHM, &reg);
+ return GET_TOHM(reg);
+}
+
+static u64 ibridge_get_tolm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->pci_br1, TOLM, &reg);
+
+ return GET_TOLM(reg);
+}
+
+static u64 ibridge_get_tohm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->pci_br1, TOHM, &reg);
+
+ return GET_TOHM(reg);
+}
+
+static u64 rir_limit(u32 reg)
+{
+ return ((u64)GET_BITFIELD(reg, 1, 10) << 29) | 0x1fffffff;
+}
+
+static u64 sad_limit(u32 reg)
+{
+ return (GET_BITFIELD(reg, 6, 25) << 26) | 0x3ffffff;
+}
+
+static u32 interleave_mode(u32 reg)
+{
+ return GET_BITFIELD(reg, 1, 1);
+}
+
+static u32 dram_attr(u32 reg)
+{
+ return GET_BITFIELD(reg, 2, 3);
+}
+
+static u64 knl_sad_limit(u32 reg)
+{
+ return (GET_BITFIELD(reg, 7, 26) << 26) | 0x3ffffff;
+}
+
+static u32 knl_interleave_mode(u32 reg)
+{
+ return GET_BITFIELD(reg, 1, 2);
+}
+
+static const char * const knl_intlv_mode[] = {
+ "[8:6]", "[10:8]", "[14:12]", "[32:30]"
+};
+
+static const char *get_intlv_mode_str(u32 reg, enum type t)
+{
+ if (t == KNIGHTS_LANDING)
+ return knl_intlv_mode[knl_interleave_mode(reg)];
+ else
+ return interleave_mode(reg) ? "[8:6]" : "[8:6]XOR[18:16]";
+}
+
+static u32 dram_attr_knl(u32 reg)
+{
+ return GET_BITFIELD(reg, 3, 4);
+}
+
+
+static enum mem_type get_memory_type(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+ enum mem_type mtype;
+
+ if (pvt->pci_ddrio) {
+ pci_read_config_dword(pvt->pci_ddrio, pvt->info.rankcfgr,
+ &reg);
+ if (GET_BITFIELD(reg, 11, 11))
+ /* FIXME: Can also be LRDIMM */
+ mtype = MEM_RDDR3;
+ else
+ mtype = MEM_DDR3;
+ } else
+ mtype = MEM_UNKNOWN;
+
+ return mtype;
+}
+
+static enum mem_type haswell_get_memory_type(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+ bool registered = false;
+ enum mem_type mtype = MEM_UNKNOWN;
+
+ if (!pvt->pci_ddrio)
+ goto out;
+
+ pci_read_config_dword(pvt->pci_ddrio,
+ HASWELL_DDRCRCLKCONTROLS, &reg);
+ /* Is_Rdimm */
+ if (GET_BITFIELD(reg, 16, 16))
+ registered = true;
+
+ pci_read_config_dword(pvt->pci_ta, MCMTR, &reg);
+ if (GET_BITFIELD(reg, 14, 14)) {
+ if (registered)
+ mtype = MEM_RDDR4;
+ else
+ mtype = MEM_DDR4;
+ } else {
+ if (registered)
+ mtype = MEM_RDDR3;
+ else
+ mtype = MEM_DDR3;
+ }
+
+out:
+ return mtype;
+}
+
+static enum dev_type knl_get_width(struct sbridge_pvt *pvt, u32 mtr)
+{
+ /* for KNL value is fixed */
+ return DEV_X16;
+}
+
+static enum dev_type sbridge_get_width(struct sbridge_pvt *pvt, u32 mtr)
+{
+ /* there's no way to figure out */
+ return DEV_UNKNOWN;
+}
+
+static enum dev_type __ibridge_get_width(u32 mtr)
+{
+ enum dev_type type = DEV_UNKNOWN;
+
+ switch (mtr) {
+ case 2:
+ type = DEV_X16;
+ break;
+ case 1:
+ type = DEV_X8;
+ break;
+ case 0:
+ type = DEV_X4;
+ break;
+ }
+
+ return type;
+}
+
+static enum dev_type ibridge_get_width(struct sbridge_pvt *pvt, u32 mtr)
+{
+ /*
+ * ddr3_width on the documentation but also valid for DDR4 on
+ * Haswell
+ */
+ return __ibridge_get_width(GET_BITFIELD(mtr, 7, 8));
+}
+
+static enum dev_type broadwell_get_width(struct sbridge_pvt *pvt, u32 mtr)
+{
+ /* ddr3_width on the documentation but also valid for DDR4 */
+ return __ibridge_get_width(GET_BITFIELD(mtr, 8, 9));
+}
+
+static enum mem_type knl_get_memory_type(struct sbridge_pvt *pvt)
+{
+ /* DDR4 RDIMMS and LRDIMMS are supported */
+ return MEM_RDDR4;
+}
+
+static u8 get_node_id(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+ pci_read_config_dword(pvt->pci_br0, SAD_CONTROL, &reg);
+ return GET_BITFIELD(reg, 0, 2);
+}
+
+static u8 haswell_get_node_id(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->pci_sad1, SAD_CONTROL, &reg);
+ return GET_BITFIELD(reg, 0, 3);
+}
+
+static u8 knl_get_node_id(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->pci_sad1, SAD_CONTROL, &reg);
+ return GET_BITFIELD(reg, 0, 2);
+}
+
+/*
+ * Use the reporting bank number to determine which memory
+ * controller (also known as "ha" for "home agent"). Sandy
+ * Bridge only has one memory controller per socket, so the
+ * answer is always zero.
+ */
+static u8 sbridge_get_ha(u8 bank)
+{
+ return 0;
+}
+
+/*
+ * On Ivy Bridge, Haswell and Broadwell the error may be in a
+ * home agent bank (7, 8), or one of the per-channel memory
+ * controller banks (9 .. 16).
+ */
+static u8 ibridge_get_ha(u8 bank)
+{
+ switch (bank) {
+ case 7 ... 8:
+ return bank - 7;
+ case 9 ... 16:
+ return (bank - 9) / 4;
+ default:
+ return 0xff;
+ }
+}
+
+/* Not used, but included for safety/symmetry */
+static u8 knl_get_ha(u8 bank)
+{
+ return 0xff;
+}
+
+static u64 haswell_get_tolm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOLM, &reg);
+ return (GET_BITFIELD(reg, 26, 31) << 26) | 0x3ffffff;
+}
+
+static u64 haswell_get_tohm(struct sbridge_pvt *pvt)
+{
+ u64 rc;
+ u32 reg;
+
+ pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOHM_0, &reg);
+ rc = GET_BITFIELD(reg, 26, 31);
+ pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOHM_1, &reg);
+ rc = ((reg << 6) | rc) << 26;
+
+ return rc | 0x3ffffff;
+}
+
+static u64 knl_get_tolm(struct sbridge_pvt *pvt)
+{
+ u32 reg;
+
+ pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOLM, &reg);
+ return (GET_BITFIELD(reg, 26, 31) << 26) | 0x3ffffff;
+}
+
+static u64 knl_get_tohm(struct sbridge_pvt *pvt)
+{
+ u64 rc;
+ u32 reg_lo, reg_hi;
+
+ pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOHM_0, &reg_lo);
+ pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOHM_1, &reg_hi);
+ rc = ((u64)reg_hi << 32) | reg_lo;
+ return rc | 0x3ffffff;
+}
+
+
+static u64 haswell_rir_limit(u32 reg)
+{
+ return (((u64)GET_BITFIELD(reg, 1, 11) + 1) << 29) - 1;
+}
+
+static inline u8 sad_pkg_socket(u8 pkg)
+{
+ /* on Ivy Bridge, nodeID is SASS, where A is HA and S is node id */
+ return ((pkg >> 3) << 2) | (pkg & 0x3);
+}
+
+static inline u8 sad_pkg_ha(u8 pkg)
+{
+ return (pkg >> 2) & 0x1;
+}
+
+static int haswell_chan_hash(int idx, u64 addr)
+{
+ int i;
+
+ /*
+ * XOR even bits from 12:26 to bit0 of idx,
+ * odd bits from 13:27 to bit1
+ */
+ for (i = 12; i < 28; i += 2)
+ idx ^= (addr >> i) & 3;
+
+ return idx;
+}
+
+/* Low bits of TAD limit, and some metadata. */
+static const u32 knl_tad_dram_limit_lo[] = {
+ 0x400, 0x500, 0x600, 0x700,
+ 0x800, 0x900, 0xa00, 0xb00,
+};
+
+/* Low bits of TAD offset. */
+static const u32 knl_tad_dram_offset_lo[] = {
+ 0x404, 0x504, 0x604, 0x704,
+ 0x804, 0x904, 0xa04, 0xb04,
+};
+
+/* High 16 bits of TAD limit and offset. */
+static const u32 knl_tad_dram_hi[] = {
+ 0x408, 0x508, 0x608, 0x708,
+ 0x808, 0x908, 0xa08, 0xb08,
+};
+
+/* Number of ways a tad entry is interleaved. */
+static const u32 knl_tad_ways[] = {
+ 8, 6, 4, 3, 2, 1,
+};
+
+/*
+ * Retrieve the n'th Target Address Decode table entry
+ * from the memory controller's TAD table.
+ *
+ * @pvt: driver private data
+ * @entry: which entry you want to retrieve
+ * @mc: which memory controller (0 or 1)
+ * @offset: output tad range offset
+ * @limit: output address of first byte above tad range
+ * @ways: output number of interleave ways
+ *
+ * The offset value has curious semantics. It's a sort of running total
+ * of the sizes of all the memory regions that aren't mapped in this
+ * tad table.
+ */
+static int knl_get_tad(const struct sbridge_pvt *pvt,
+ const int entry,
+ const int mc,
+ u64 *offset,
+ u64 *limit,
+ int *ways)
+{
+ u32 reg_limit_lo, reg_offset_lo, reg_hi;
+ struct pci_dev *pci_mc;
+ int way_id;
+
+ switch (mc) {
+ case 0:
+ pci_mc = pvt->knl.pci_mc0;
+ break;
+ case 1:
+ pci_mc = pvt->knl.pci_mc1;
+ break;
+ default:
+ WARN_ON(1);
+ return -EINVAL;
+ }
+
+ pci_read_config_dword(pci_mc,
+ knl_tad_dram_limit_lo[entry], &reg_limit_lo);
+ pci_read_config_dword(pci_mc,
+ knl_tad_dram_offset_lo[entry], &reg_offset_lo);
+ pci_read_config_dword(pci_mc,
+ knl_tad_dram_hi[entry], &reg_hi);
+
+ /* Is this TAD entry enabled? */
+ if (!GET_BITFIELD(reg_limit_lo, 0, 0))
+ return -ENODEV;
+
+ way_id = GET_BITFIELD(reg_limit_lo, 3, 5);
+
+ if (way_id < ARRAY_SIZE(knl_tad_ways)) {
+ *ways = knl_tad_ways[way_id];
+ } else {
+ *ways = 0;
+ sbridge_printk(KERN_ERR,
+ "Unexpected value %d in mc_tad_limit_lo wayness field\n",
+ way_id);
+ return -ENODEV;
+ }
+
+ /*
+ * The least significant 6 bits of base and limit are truncated.
+ * For limit, we fill the missing bits with 1s.
+ */
+ *offset = ((u64) GET_BITFIELD(reg_offset_lo, 6, 31) << 6) |
+ ((u64) GET_BITFIELD(reg_hi, 0, 15) << 32);
+ *limit = ((u64) GET_BITFIELD(reg_limit_lo, 6, 31) << 6) | 63 |
+ ((u64) GET_BITFIELD(reg_hi, 16, 31) << 32);
+
+ return 0;
+}
+
+/* Determine which memory controller is responsible for a given channel. */
+static int knl_channel_mc(int channel)
+{
+ WARN_ON(channel < 0 || channel >= 6);
+
+ return channel < 3 ? 1 : 0;
+}
+
+/*
+ * Get the Nth entry from EDC_ROUTE_TABLE register.
+ * (This is the per-tile mapping of logical interleave targets to
+ * physical EDC modules.)
+ *
+ * entry 0: 0:2
+ * 1: 3:5
+ * 2: 6:8
+ * 3: 9:11
+ * 4: 12:14
+ * 5: 15:17
+ * 6: 18:20
+ * 7: 21:23
+ * reserved: 24:31
+ */
+static u32 knl_get_edc_route(int entry, u32 reg)
+{
+ WARN_ON(entry >= KNL_MAX_EDCS);
+ return GET_BITFIELD(reg, entry*3, (entry*3)+2);
+}
+
+/*
+ * Get the Nth entry from MC_ROUTE_TABLE register.
+ * (This is the per-tile mapping of logical interleave targets to
+ * physical DRAM channels modules.)
+ *
+ * entry 0: mc 0:2 channel 18:19
+ * 1: mc 3:5 channel 20:21
+ * 2: mc 6:8 channel 22:23
+ * 3: mc 9:11 channel 24:25
+ * 4: mc 12:14 channel 26:27
+ * 5: mc 15:17 channel 28:29
+ * reserved: 30:31
+ *
+ * Though we have 3 bits to identify the MC, we should only see
+ * the values 0 or 1.
+ */
+
+static u32 knl_get_mc_route(int entry, u32 reg)
+{
+ int mc, chan;
+
+ WARN_ON(entry >= KNL_MAX_CHANNELS);
+
+ mc = GET_BITFIELD(reg, entry*3, (entry*3)+2);
+ chan = GET_BITFIELD(reg, (entry*2) + 18, (entry*2) + 18 + 1);
+
+ return knl_channel_remap(mc, chan);
+}
+
+/*
+ * Render the EDC_ROUTE register in human-readable form.
+ * Output string s should be at least KNL_MAX_EDCS*2 bytes.
+ */
+static void knl_show_edc_route(u32 reg, char *s)
+{
+ int i;
+
+ for (i = 0; i < KNL_MAX_EDCS; i++) {
+ s[i*2] = knl_get_edc_route(i, reg) + '0';
+ s[i*2+1] = '-';
+ }
+
+ s[KNL_MAX_EDCS*2 - 1] = '\0';
+}
+
+/*
+ * Render the MC_ROUTE register in human-readable form.
+ * Output string s should be at least KNL_MAX_CHANNELS*2 bytes.
+ */
+static void knl_show_mc_route(u32 reg, char *s)
+{
+ int i;
+
+ for (i = 0; i < KNL_MAX_CHANNELS; i++) {
+ s[i*2] = knl_get_mc_route(i, reg) + '0';
+ s[i*2+1] = '-';
+ }
+
+ s[KNL_MAX_CHANNELS*2 - 1] = '\0';
+}
+
+#define KNL_EDC_ROUTE 0xb8
+#define KNL_MC_ROUTE 0xb4
+
+/* Is this dram rule backed by regular DRAM in flat mode? */
+#define KNL_EDRAM(reg) GET_BITFIELD(reg, 29, 29)
+
+/* Is this dram rule cached? */
+#define KNL_CACHEABLE(reg) GET_BITFIELD(reg, 28, 28)
+
+/* Is this rule backed by edc ? */
+#define KNL_EDRAM_ONLY(reg) GET_BITFIELD(reg, 29, 29)
+
+/* Is this rule backed by DRAM, cacheable in EDRAM? */
+#define KNL_CACHEABLE(reg) GET_BITFIELD(reg, 28, 28)
+
+/* Is this rule mod3? */
+#define KNL_MOD3(reg) GET_BITFIELD(reg, 27, 27)
+
+/*
+ * Figure out how big our RAM modules are.
+ *
+ * The DIMMMTR register in KNL doesn't tell us the size of the DIMMs, so we
+ * have to figure this out from the SAD rules, interleave lists, route tables,
+ * and TAD rules.
+ *
+ * SAD rules can have holes in them (e.g. the 3G-4G hole), so we have to
+ * inspect the TAD rules to figure out how large the SAD regions really are.
+ *
+ * When we know the real size of a SAD region and how many ways it's
+ * interleaved, we know the individual contribution of each channel to
+ * TAD is size/ways.
+ *
+ * Finally, we have to check whether each channel participates in each SAD
+ * region.
+ *
+ * Fortunately, KNL only supports one DIMM per channel, so once we know how
+ * much memory the channel uses, we know the DIMM is at least that large.
+ * (The BIOS might possibly choose not to map all available memory, in which
+ * case we will underreport the size of the DIMM.)
+ *
+ * In theory, we could try to determine the EDC sizes as well, but that would
+ * only work in flat mode, not in cache mode.
+ *
+ * @mc_sizes: Output sizes of channels (must have space for KNL_MAX_CHANNELS
+ * elements)
+ */
+static int knl_get_dimm_capacity(struct sbridge_pvt *pvt, u64 *mc_sizes)
+{
+ u64 sad_base, sad_limit = 0;
+ u64 tad_base, tad_size, tad_limit, tad_deadspace, tad_livespace;
+ int sad_rule = 0;
+ int tad_rule = 0;
+ int intrlv_ways, tad_ways;
+ u32 first_pkg, pkg;
+ int i;
+ u64 sad_actual_size[2]; /* sad size accounting for holes, per mc */
+ u32 dram_rule, interleave_reg;
+ u32 mc_route_reg[KNL_MAX_CHAS];
+ u32 edc_route_reg[KNL_MAX_CHAS];
+ int edram_only;
+ char edc_route_string[KNL_MAX_EDCS*2];
+ char mc_route_string[KNL_MAX_CHANNELS*2];
+ int cur_reg_start;
+ int mc;
+ int channel;
+ int participants[KNL_MAX_CHANNELS];
+
+ for (i = 0; i < KNL_MAX_CHANNELS; i++)
+ mc_sizes[i] = 0;
+
+ /* Read the EDC route table in each CHA. */
+ cur_reg_start = 0;
+ for (i = 0; i < KNL_MAX_CHAS; i++) {
+ pci_read_config_dword(pvt->knl.pci_cha[i],
+ KNL_EDC_ROUTE, &edc_route_reg[i]);
+
+ if (i > 0 && edc_route_reg[i] != edc_route_reg[i-1]) {
+ knl_show_edc_route(edc_route_reg[i-1],
+ edc_route_string);
+ if (cur_reg_start == i-1)
+ edac_dbg(0, "edc route table for CHA %d: %s\n",
+ cur_reg_start, edc_route_string);
+ else
+ edac_dbg(0, "edc route table for CHA %d-%d: %s\n",
+ cur_reg_start, i-1, edc_route_string);
+ cur_reg_start = i;
+ }
+ }
+ knl_show_edc_route(edc_route_reg[i-1], edc_route_string);
+ if (cur_reg_start == i-1)
+ edac_dbg(0, "edc route table for CHA %d: %s\n",
+ cur_reg_start, edc_route_string);
+ else
+ edac_dbg(0, "edc route table for CHA %d-%d: %s\n",
+ cur_reg_start, i-1, edc_route_string);
+
+ /* Read the MC route table in each CHA. */
+ cur_reg_start = 0;
+ for (i = 0; i < KNL_MAX_CHAS; i++) {
+ pci_read_config_dword(pvt->knl.pci_cha[i],
+ KNL_MC_ROUTE, &mc_route_reg[i]);
+
+ if (i > 0 && mc_route_reg[i] != mc_route_reg[i-1]) {
+ knl_show_mc_route(mc_route_reg[i-1], mc_route_string);
+ if (cur_reg_start == i-1)
+ edac_dbg(0, "mc route table for CHA %d: %s\n",
+ cur_reg_start, mc_route_string);
+ else
+ edac_dbg(0, "mc route table for CHA %d-%d: %s\n",
+ cur_reg_start, i-1, mc_route_string);
+ cur_reg_start = i;
+ }
+ }
+ knl_show_mc_route(mc_route_reg[i-1], mc_route_string);
+ if (cur_reg_start == i-1)
+ edac_dbg(0, "mc route table for CHA %d: %s\n",
+ cur_reg_start, mc_route_string);
+ else
+ edac_dbg(0, "mc route table for CHA %d-%d: %s\n",
+ cur_reg_start, i-1, mc_route_string);
+
+ /* Process DRAM rules */
+ for (sad_rule = 0; sad_rule < pvt->info.max_sad; sad_rule++) {
+ /* previous limit becomes the new base */
+ sad_base = sad_limit;
+
+ pci_read_config_dword(pvt->pci_sad0,
+ pvt->info.dram_rule[sad_rule], &dram_rule);
+
+ if (!DRAM_RULE_ENABLE(dram_rule))
+ break;
+
+ edram_only = KNL_EDRAM_ONLY(dram_rule);
+
+ sad_limit = pvt->info.sad_limit(dram_rule)+1;
+
+ pci_read_config_dword(pvt->pci_sad0,
+ pvt->info.interleave_list[sad_rule], &interleave_reg);
+
+ /*
+ * Find out how many ways this dram rule is interleaved.
+ * We stop when we see the first channel again.
+ */
+ first_pkg = sad_pkg(pvt->info.interleave_pkg,
+ interleave_reg, 0);
+ for (intrlv_ways = 1; intrlv_ways < 8; intrlv_ways++) {
+ pkg = sad_pkg(pvt->info.interleave_pkg,
+ interleave_reg, intrlv_ways);
+
+ if ((pkg & 0x8) == 0) {
+ /*
+ * 0 bit means memory is non-local,
+ * which KNL doesn't support
+ */
+ edac_dbg(0, "Unexpected interleave target %d\n",
+ pkg);
+ return -1;
+ }
+
+ if (pkg == first_pkg)
+ break;
+ }
+ if (KNL_MOD3(dram_rule))
+ intrlv_ways *= 3;
+
+ edac_dbg(3, "dram rule %d (base 0x%llx, limit 0x%llx), %d way interleave%s\n",
+ sad_rule,
+ sad_base,
+ sad_limit,
+ intrlv_ways,
+ edram_only ? ", EDRAM" : "");
+
+ /*
+ * Find out how big the SAD region really is by iterating
+ * over TAD tables (SAD regions may contain holes).
+ * Each memory controller might have a different TAD table, so
+ * we have to look at both.
+ *
+ * Livespace is the memory that's mapped in this TAD table,
+ * deadspace is the holes (this could be the MMIO hole, or it
+ * could be memory that's mapped by the other TAD table but
+ * not this one).
+ */
+ for (mc = 0; mc < 2; mc++) {
+ sad_actual_size[mc] = 0;
+ tad_livespace = 0;
+ for (tad_rule = 0;
+ tad_rule < ARRAY_SIZE(
+ knl_tad_dram_limit_lo);
+ tad_rule++) {
+ if (knl_get_tad(pvt,
+ tad_rule,
+ mc,
+ &tad_deadspace,
+ &tad_limit,
+ &tad_ways))
+ break;
+
+ tad_size = (tad_limit+1) -
+ (tad_livespace + tad_deadspace);
+ tad_livespace += tad_size;
+ tad_base = (tad_limit+1) - tad_size;
+
+ if (tad_base < sad_base) {
+ if (tad_limit > sad_base)
+ edac_dbg(0, "TAD region overlaps lower SAD boundary -- TAD tables may be configured incorrectly.\n");
+ } else if (tad_base < sad_limit) {
+ if (tad_limit+1 > sad_limit) {
+ edac_dbg(0, "TAD region overlaps upper SAD boundary -- TAD tables may be configured incorrectly.\n");
+ } else {
+ /* TAD region is completely inside SAD region */
+ edac_dbg(3, "TAD region %d 0x%llx - 0x%llx (%lld bytes) table%d\n",
+ tad_rule, tad_base,
+ tad_limit, tad_size,
+ mc);
+ sad_actual_size[mc] += tad_size;
+ }
+ }
+ }
+ }
+
+ for (mc = 0; mc < 2; mc++) {
+ edac_dbg(3, " total TAD DRAM footprint in table%d : 0x%llx (%lld bytes)\n",
+ mc, sad_actual_size[mc], sad_actual_size[mc]);
+ }
+
+ /* Ignore EDRAM rule */
+ if (edram_only)
+ continue;
+
+ /* Figure out which channels participate in interleave. */
+ for (channel = 0; channel < KNL_MAX_CHANNELS; channel++)
+ participants[channel] = 0;
+
+ /* For each channel, does at least one CHA have
+ * this channel mapped to the given target?
+ */
+ for (channel = 0; channel < KNL_MAX_CHANNELS; channel++) {
+ int target;
+ int cha;
+
+ for (target = 0; target < KNL_MAX_CHANNELS; target++) {
+ for (cha = 0; cha < KNL_MAX_CHAS; cha++) {
+ if (knl_get_mc_route(target,
+ mc_route_reg[cha]) == channel
+ && !participants[channel]) {
+ participants[channel] = 1;
+ break;
+ }
+ }
+ }
+ }
+
+ for (channel = 0; channel < KNL_MAX_CHANNELS; channel++) {
+ mc = knl_channel_mc(channel);
+ if (participants[channel]) {
+ edac_dbg(4, "mc channel %d contributes %lld bytes via sad entry %d\n",
+ channel,
+ sad_actual_size[mc]/intrlv_ways,
+ sad_rule);
+ mc_sizes[channel] +=
+ sad_actual_size[mc]/intrlv_ways;
+ }
+ }
+ }
+
+ return 0;
+}
+
+static void get_source_id(struct mem_ctl_info *mci)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ u32 reg;
+
+ if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL ||
+ pvt->info.type == KNIGHTS_LANDING)
+ pci_read_config_dword(pvt->pci_sad1, SAD_TARGET, &reg);
+ else
+ pci_read_config_dword(pvt->pci_br0, SAD_TARGET, &reg);
+
+ if (pvt->info.type == KNIGHTS_LANDING)
+ pvt->sbridge_dev->source_id = SOURCE_ID_KNL(reg);
+ else
+ pvt->sbridge_dev->source_id = SOURCE_ID(reg);
+}
+
+static int __populate_dimms(struct mem_ctl_info *mci,
+ u64 knl_mc_sizes[KNL_MAX_CHANNELS],
+ enum edac_type mode)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ int channels = pvt->info.type == KNIGHTS_LANDING ? KNL_MAX_CHANNELS
+ : NUM_CHANNELS;
+ unsigned int i, j, banks, ranks, rows, cols, npages;
+ struct dimm_info *dimm;
+ enum mem_type mtype;
+ u64 size;
+
+ mtype = pvt->info.get_memory_type(pvt);
+ if (mtype == MEM_RDDR3 || mtype == MEM_RDDR4)
+ edac_dbg(0, "Memory is registered\n");
+ else if (mtype == MEM_UNKNOWN)
+ edac_dbg(0, "Cannot determine memory type\n");
+ else
+ edac_dbg(0, "Memory is unregistered\n");
+
+ if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
+ banks = 16;
+ else
+ banks = 8;
+
+ for (i = 0; i < channels; i++) {
+ u32 mtr, amap = 0;
+
+ int max_dimms_per_channel;
+
+ if (pvt->info.type == KNIGHTS_LANDING) {
+ max_dimms_per_channel = 1;
+ if (!pvt->knl.pci_channel[i])
+ continue;
+ } else {
+ max_dimms_per_channel = ARRAY_SIZE(mtr_regs);
+ if (!pvt->pci_tad[i])
+ continue;
+ pci_read_config_dword(pvt->pci_tad[i], 0x8c, &amap);
+ }
+
+ for (j = 0; j < max_dimms_per_channel; j++) {
+ dimm = edac_get_dimm(mci, i, j, 0);
+ if (pvt->info.type == KNIGHTS_LANDING) {
+ pci_read_config_dword(pvt->knl.pci_channel[i],
+ knl_mtr_reg, &mtr);
+ } else {
+ pci_read_config_dword(pvt->pci_tad[i],
+ mtr_regs[j], &mtr);
+ }
+ edac_dbg(4, "Channel #%d MTR%d = %x\n", i, j, mtr);
+
+ if (IS_DIMM_PRESENT(mtr)) {
+ if (!IS_ECC_ENABLED(pvt->info.mcmtr)) {
+ sbridge_printk(KERN_ERR, "CPU SrcID #%d, Ha #%d, Channel #%d has DIMMs, but ECC is disabled\n",
+ pvt->sbridge_dev->source_id,
+ pvt->sbridge_dev->dom, i);
+ return -ENODEV;
+ }
+ pvt->channel[i].dimms++;
+
+ ranks = numrank(pvt->info.type, mtr);
+
+ if (pvt->info.type == KNIGHTS_LANDING) {
+ /* For DDR4, this is fixed. */
+ cols = 1 << 10;
+ rows = knl_mc_sizes[i] /
+ ((u64) cols * ranks * banks * 8);
+ } else {
+ rows = numrow(mtr);
+ cols = numcol(mtr);
+ }
+
+ size = ((u64)rows * cols * banks * ranks) >> (20 - 3);
+ npages = MiB_TO_PAGES(size);
+
+ edac_dbg(0, "mc#%d: ha %d channel %d, dimm %d, %lld MiB (%d pages) bank: %d, rank: %d, row: %#x, col: %#x\n",
+ pvt->sbridge_dev->mc, pvt->sbridge_dev->dom, i, j,
+ size, npages,
+ banks, ranks, rows, cols);
+
+ dimm->nr_pages = npages;
+ dimm->grain = 32;
+ dimm->dtype = pvt->info.get_width(pvt, mtr);
+ dimm->mtype = mtype;
+ dimm->edac_mode = mode;
+ pvt->channel[i].dimm[j].rowbits = order_base_2(rows);
+ pvt->channel[i].dimm[j].colbits = order_base_2(cols);
+ pvt->channel[i].dimm[j].bank_xor_enable =
+ GET_BITFIELD(pvt->info.mcmtr, 9, 9);
+ pvt->channel[i].dimm[j].amap_fine = GET_BITFIELD(amap, 0, 0);
+ snprintf(dimm->label, sizeof(dimm->label),
+ "CPU_SrcID#%u_Ha#%u_Chan#%u_DIMM#%u",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom, i, j);
+ }
+ }
+ }
+
+ return 0;
+}
+
+static int get_dimm_config(struct mem_ctl_info *mci)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ u64 knl_mc_sizes[KNL_MAX_CHANNELS];
+ enum edac_type mode;
+ u32 reg;
+
+ pvt->sbridge_dev->node_id = pvt->info.get_node_id(pvt);
+ edac_dbg(0, "mc#%d: Node ID: %d, source ID: %d\n",
+ pvt->sbridge_dev->mc,
+ pvt->sbridge_dev->node_id,
+ pvt->sbridge_dev->source_id);
+
+ /* KNL doesn't support mirroring or lockstep,
+ * and is always closed page
+ */
+ if (pvt->info.type == KNIGHTS_LANDING) {
+ mode = EDAC_S4ECD4ED;
+ pvt->mirror_mode = NON_MIRRORING;
+ pvt->is_cur_addr_mirrored = false;
+
+ if (knl_get_dimm_capacity(pvt, knl_mc_sizes) != 0)
+ return -1;
+ if (pci_read_config_dword(pvt->pci_ta, KNL_MCMTR, &pvt->info.mcmtr)) {
+ edac_dbg(0, "Failed to read KNL_MCMTR register\n");
+ return -ENODEV;
+ }
+ } else {
+ if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL) {
+ if (pci_read_config_dword(pvt->pci_ha, HASWELL_HASYSDEFEATURE2, &reg)) {
+ edac_dbg(0, "Failed to read HASWELL_HASYSDEFEATURE2 register\n");
+ return -ENODEV;
+ }
+ pvt->is_chan_hash = GET_BITFIELD(reg, 21, 21);
+ if (GET_BITFIELD(reg, 28, 28)) {
+ pvt->mirror_mode = ADDR_RANGE_MIRRORING;
+ edac_dbg(0, "Address range partial memory mirroring is enabled\n");
+ goto next;
+ }
+ }
+ if (pci_read_config_dword(pvt->pci_ras, RASENABLES, &reg)) {
+ edac_dbg(0, "Failed to read RASENABLES register\n");
+ return -ENODEV;
+ }
+ if (IS_MIRROR_ENABLED(reg)) {
+ pvt->mirror_mode = FULL_MIRRORING;
+ edac_dbg(0, "Full memory mirroring is enabled\n");
+ } else {
+ pvt->mirror_mode = NON_MIRRORING;
+ edac_dbg(0, "Memory mirroring is disabled\n");
+ }
+
+next:
+ if (pci_read_config_dword(pvt->pci_ta, MCMTR, &pvt->info.mcmtr)) {
+ edac_dbg(0, "Failed to read MCMTR register\n");
+ return -ENODEV;
+ }
+ if (IS_LOCKSTEP_ENABLED(pvt->info.mcmtr)) {
+ edac_dbg(0, "Lockstep is enabled\n");
+ mode = EDAC_S8ECD8ED;
+ pvt->is_lockstep = true;
+ } else {
+ edac_dbg(0, "Lockstep is disabled\n");
+ mode = EDAC_S4ECD4ED;
+ pvt->is_lockstep = false;
+ }
+ if (IS_CLOSE_PG(pvt->info.mcmtr)) {
+ edac_dbg(0, "address map is on closed page mode\n");
+ pvt->is_close_pg = true;
+ } else {
+ edac_dbg(0, "address map is on open page mode\n");
+ pvt->is_close_pg = false;
+ }
+ }
+
+ return __populate_dimms(mci, knl_mc_sizes, mode);
+}
+
+static void get_memory_layout(const struct mem_ctl_info *mci)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ int i, j, k, n_sads, n_tads, sad_interl;
+ u32 reg;
+ u64 limit, prv = 0;
+ u64 tmp_mb;
+ u32 gb, mb;
+ u32 rir_way;
+
+ /*
+ * Step 1) Get TOLM/TOHM ranges
+ */
+
+ pvt->tolm = pvt->info.get_tolm(pvt);
+ tmp_mb = (1 + pvt->tolm) >> 20;
+
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "TOLM: %u.%03u GB (0x%016Lx)\n",
+ gb, (mb*1000)/1024, (u64)pvt->tolm);
+
+ /* Address range is already 45:25 */
+ pvt->tohm = pvt->info.get_tohm(pvt);
+ tmp_mb = (1 + pvt->tohm) >> 20;
+
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "TOHM: %u.%03u GB (0x%016Lx)\n",
+ gb, (mb*1000)/1024, (u64)pvt->tohm);
+
+ /*
+ * Step 2) Get SAD range and SAD Interleave list
+ * TAD registers contain the interleave wayness. However, it
+ * seems simpler to just discover it indirectly, with the
+ * algorithm bellow.
+ */
+ prv = 0;
+ for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
+ /* SAD_LIMIT Address range is 45:26 */
+ pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
+ &reg);
+ limit = pvt->info.sad_limit(reg);
+
+ if (!DRAM_RULE_ENABLE(reg))
+ continue;
+
+ if (limit <= prv)
+ break;
+
+ tmp_mb = (limit + 1) >> 20;
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "SAD#%d %s up to %u.%03u GB (0x%016Lx) Interleave: %s reg=0x%08x\n",
+ n_sads,
+ show_dram_attr(pvt->info.dram_attr(reg)),
+ gb, (mb*1000)/1024,
+ ((u64)tmp_mb) << 20L,
+ get_intlv_mode_str(reg, pvt->info.type),
+ reg);
+ prv = limit;
+
+ pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
+ &reg);
+ sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
+ for (j = 0; j < 8; j++) {
+ u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, j);
+ if (j > 0 && sad_interl == pkg)
+ break;
+
+ edac_dbg(0, "SAD#%d, interleave #%d: %d\n",
+ n_sads, j, pkg);
+ }
+ }
+
+ if (pvt->info.type == KNIGHTS_LANDING)
+ return;
+
+ /*
+ * Step 3) Get TAD range
+ */
+ prv = 0;
+ for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
+ pci_read_config_dword(pvt->pci_ha, tad_dram_rule[n_tads], &reg);
+ limit = TAD_LIMIT(reg);
+ if (limit <= prv)
+ break;
+ tmp_mb = (limit + 1) >> 20;
+
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "TAD#%d: up to %u.%03u GB (0x%016Lx), socket interleave %d, memory interleave %d, TGT: %d, %d, %d, %d, reg=0x%08x\n",
+ n_tads, gb, (mb*1000)/1024,
+ ((u64)tmp_mb) << 20L,
+ (u32)(1 << TAD_SOCK(reg)),
+ (u32)TAD_CH(reg) + 1,
+ (u32)TAD_TGT0(reg),
+ (u32)TAD_TGT1(reg),
+ (u32)TAD_TGT2(reg),
+ (u32)TAD_TGT3(reg),
+ reg);
+ prv = limit;
+ }
+
+ /*
+ * Step 4) Get TAD offsets, per each channel
+ */
+ for (i = 0; i < NUM_CHANNELS; i++) {
+ if (!pvt->channel[i].dimms)
+ continue;
+ for (j = 0; j < n_tads; j++) {
+ pci_read_config_dword(pvt->pci_tad[i],
+ tad_ch_nilv_offset[j],
+ &reg);
+ tmp_mb = TAD_OFFSET(reg) >> 20;
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "TAD CH#%d, offset #%d: %u.%03u GB (0x%016Lx), reg=0x%08x\n",
+ i, j,
+ gb, (mb*1000)/1024,
+ ((u64)tmp_mb) << 20L,
+ reg);
+ }
+ }
+
+ /*
+ * Step 6) Get RIR Wayness/Limit, per each channel
+ */
+ for (i = 0; i < NUM_CHANNELS; i++) {
+ if (!pvt->channel[i].dimms)
+ continue;
+ for (j = 0; j < MAX_RIR_RANGES; j++) {
+ pci_read_config_dword(pvt->pci_tad[i],
+ rir_way_limit[j],
+ &reg);
+
+ if (!IS_RIR_VALID(reg))
+ continue;
+
+ tmp_mb = pvt->info.rir_limit(reg) >> 20;
+ rir_way = 1 << RIR_WAY(reg);
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "CH#%d RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d, reg=0x%08x\n",
+ i, j,
+ gb, (mb*1000)/1024,
+ ((u64)tmp_mb) << 20L,
+ rir_way,
+ reg);
+
+ for (k = 0; k < rir_way; k++) {
+ pci_read_config_dword(pvt->pci_tad[i],
+ rir_offset[j][k],
+ &reg);
+ tmp_mb = RIR_OFFSET(pvt->info.type, reg) << 6;
+
+ gb = div_u64_rem(tmp_mb, 1024, &mb);
+ edac_dbg(0, "CH#%d RIR#%d INTL#%d, offset %u.%03u GB (0x%016Lx), tgt: %d, reg=0x%08x\n",
+ i, j, k,
+ gb, (mb*1000)/1024,
+ ((u64)tmp_mb) << 20L,
+ (u32)RIR_RNK_TGT(pvt->info.type, reg),
+ reg);
+ }
+ }
+ }
+}
+
+static struct mem_ctl_info *get_mci_for_node_id(u8 node_id, u8 ha)
+{
+ struct sbridge_dev *sbridge_dev;
+
+ list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
+ if (sbridge_dev->node_id == node_id && sbridge_dev->dom == ha)
+ return sbridge_dev->mci;
+ }
+ return NULL;
+}
+
+static u8 sb_close_row[] = {
+ 15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33
+};
+
+static u8 sb_close_column[] = {
+ 3, 4, 5, 14, 19, 23, 24, 25, 26, 27
+};
+
+static u8 sb_open_row[] = {
+ 14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33
+};
+
+static u8 sb_open_column[] = {
+ 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
+};
+
+static u8 sb_open_fine_column[] = {
+ 3, 4, 5, 7, 8, 9, 10, 11, 12, 13
+};
+
+static int sb_bits(u64 addr, int nbits, u8 *bits)
+{
+ int i, res = 0;
+
+ for (i = 0; i < nbits; i++)
+ res |= ((addr >> bits[i]) & 1) << i;
+ return res;
+}
+
+static int sb_bank_bits(u64 addr, int b0, int b1, int do_xor, int x0, int x1)
+{
+ int ret = GET_BITFIELD(addr, b0, b0) | (GET_BITFIELD(addr, b1, b1) << 1);
+
+ if (do_xor)
+ ret ^= GET_BITFIELD(addr, x0, x0) | (GET_BITFIELD(addr, x1, x1) << 1);
+
+ return ret;
+}
+
+static bool sb_decode_ddr4(struct mem_ctl_info *mci, int ch, u8 rank,
+ u64 rank_addr, char *msg)
+{
+ int dimmno = 0;
+ int row, col, bank_address, bank_group;
+ struct sbridge_pvt *pvt;
+ u32 bg0 = 0, rowbits = 0, colbits = 0;
+ u32 amap_fine = 0, bank_xor_enable = 0;
+
+ dimmno = (rank < 12) ? rank / 4 : 2;
+ pvt = mci->pvt_info;
+ amap_fine = pvt->channel[ch].dimm[dimmno].amap_fine;
+ bg0 = amap_fine ? 6 : 13;
+ rowbits = pvt->channel[ch].dimm[dimmno].rowbits;
+ colbits = pvt->channel[ch].dimm[dimmno].colbits;
+ bank_xor_enable = pvt->channel[ch].dimm[dimmno].bank_xor_enable;
+
+ if (pvt->is_lockstep) {
+ pr_warn_once("LockStep row/column decode is not supported yet!\n");
+ msg[0] = '\0';
+ return false;
+ }
+
+ if (pvt->is_close_pg) {
+ row = sb_bits(rank_addr, rowbits, sb_close_row);
+ col = sb_bits(rank_addr, colbits, sb_close_column);
+ col |= 0x400; /* C10 is autoprecharge, always set */
+ bank_address = sb_bank_bits(rank_addr, 8, 9, bank_xor_enable, 22, 28);
+ bank_group = sb_bank_bits(rank_addr, 6, 7, bank_xor_enable, 20, 21);
+ } else {
+ row = sb_bits(rank_addr, rowbits, sb_open_row);
+ if (amap_fine)
+ col = sb_bits(rank_addr, colbits, sb_open_fine_column);
+ else
+ col = sb_bits(rank_addr, colbits, sb_open_column);
+ bank_address = sb_bank_bits(rank_addr, 18, 19, bank_xor_enable, 22, 23);
+ bank_group = sb_bank_bits(rank_addr, bg0, 17, bank_xor_enable, 20, 21);
+ }
+
+ row &= (1u << rowbits) - 1;
+
+ sprintf(msg, "row:0x%x col:0x%x bank_addr:%d bank_group:%d",
+ row, col, bank_address, bank_group);
+ return true;
+}
+
+static bool sb_decode_ddr3(struct mem_ctl_info *mci, int ch, u8 rank,
+ u64 rank_addr, char *msg)
+{
+ pr_warn_once("DDR3 row/column decode not support yet!\n");
+ msg[0] = '\0';
+ return false;
+}
+
+static int get_memory_error_data(struct mem_ctl_info *mci,
+ u64 addr,
+ u8 *socket, u8 *ha,
+ long *channel_mask,
+ u8 *rank,
+ char **area_type, char *msg)
+{
+ struct mem_ctl_info *new_mci;
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pci_ha;
+ int n_rir, n_sads, n_tads, sad_way, sck_xch;
+ int sad_interl, idx, base_ch;
+ int interleave_mode, shiftup = 0;
+ unsigned int sad_interleave[MAX_INTERLEAVE];
+ u32 reg, dram_rule;
+ u8 ch_way, sck_way, pkg, sad_ha = 0, rankid = 0;
+ u32 tad_offset;
+ u32 rir_way;
+ u32 mb, gb;
+ u64 ch_addr, offset, limit = 0, prv = 0;
+ u64 rank_addr;
+ enum mem_type mtype;
+
+ /*
+ * Step 0) Check if the address is at special memory ranges
+ * The check bellow is probably enough to fill all cases where
+ * the error is not inside a memory, except for the legacy
+ * range (e. g. VGA addresses). It is unlikely, however, that the
+ * memory controller would generate an error on that range.
+ */
+ if ((addr > (u64) pvt->tolm) && (addr < (1LL << 32))) {
+ sprintf(msg, "Error at TOLM area, on addr 0x%08Lx", addr);
+ return -EINVAL;
+ }
+ if (addr >= (u64)pvt->tohm) {
+ sprintf(msg, "Error at MMIOH area, on addr 0x%016Lx", addr);
+ return -EINVAL;
+ }
+
+ /*
+ * Step 1) Get socket
+ */
+ for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
+ pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
+ &reg);
+
+ if (!DRAM_RULE_ENABLE(reg))
+ continue;
+
+ limit = pvt->info.sad_limit(reg);
+ if (limit <= prv) {
+ sprintf(msg, "Can't discover the memory socket");
+ return -EINVAL;
+ }
+ if (addr <= limit)
+ break;
+ prv = limit;
+ }
+ if (n_sads == pvt->info.max_sad) {
+ sprintf(msg, "Can't discover the memory socket");
+ return -EINVAL;
+ }
+ dram_rule = reg;
+ *area_type = show_dram_attr(pvt->info.dram_attr(dram_rule));
+ interleave_mode = pvt->info.interleave_mode(dram_rule);
+
+ pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
+ &reg);
+
+ if (pvt->info.type == SANDY_BRIDGE) {
+ sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
+ for (sad_way = 0; sad_way < 8; sad_way++) {
+ u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, sad_way);
+ if (sad_way > 0 && sad_interl == pkg)
+ break;
+ sad_interleave[sad_way] = pkg;
+ edac_dbg(0, "SAD interleave #%d: %d\n",
+ sad_way, sad_interleave[sad_way]);
+ }
+ edac_dbg(0, "mc#%d: Error detected on SAD#%d: address 0x%016Lx < 0x%016Lx, Interleave [%d:6]%s\n",
+ pvt->sbridge_dev->mc,
+ n_sads,
+ addr,
+ limit,
+ sad_way + 7,
+ !interleave_mode ? "" : "XOR[18:16]");
+ if (interleave_mode)
+ idx = ((addr >> 6) ^ (addr >> 16)) & 7;
+ else
+ idx = (addr >> 6) & 7;
+ switch (sad_way) {
+ case 1:
+ idx = 0;
+ break;
+ case 2:
+ idx = idx & 1;
+ break;
+ case 4:
+ idx = idx & 3;
+ break;
+ case 8:
+ break;
+ default:
+ sprintf(msg, "Can't discover socket interleave");
+ return -EINVAL;
+ }
+ *socket = sad_interleave[idx];
+ edac_dbg(0, "SAD interleave index: %d (wayness %d) = CPU socket %d\n",
+ idx, sad_way, *socket);
+ } else if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL) {
+ int bits, a7mode = A7MODE(dram_rule);
+
+ if (a7mode) {
+ /* A7 mode swaps P9 with P6 */
+ bits = GET_BITFIELD(addr, 7, 8) << 1;
+ bits |= GET_BITFIELD(addr, 9, 9);
+ } else
+ bits = GET_BITFIELD(addr, 6, 8);
+
+ if (interleave_mode == 0) {
+ /* interleave mode will XOR {8,7,6} with {18,17,16} */
+ idx = GET_BITFIELD(addr, 16, 18);
+ idx ^= bits;
+ } else
+ idx = bits;
+
+ pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
+ *socket = sad_pkg_socket(pkg);
+ sad_ha = sad_pkg_ha(pkg);
+
+ if (a7mode) {
+ /* MCChanShiftUpEnable */
+ pci_read_config_dword(pvt->pci_ha, HASWELL_HASYSDEFEATURE2, &reg);
+ shiftup = GET_BITFIELD(reg, 22, 22);
+ }
+
+ edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %i, shiftup: %i\n",
+ idx, *socket, sad_ha, shiftup);
+ } else {
+ /* Ivy Bridge's SAD mode doesn't support XOR interleave mode */
+ idx = (addr >> 6) & 7;
+ pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
+ *socket = sad_pkg_socket(pkg);
+ sad_ha = sad_pkg_ha(pkg);
+ edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %d\n",
+ idx, *socket, sad_ha);
+ }
+
+ *ha = sad_ha;
+
+ /*
+ * Move to the proper node structure, in order to access the
+ * right PCI registers
+ */
+ new_mci = get_mci_for_node_id(*socket, sad_ha);
+ if (!new_mci) {
+ sprintf(msg, "Struct for socket #%u wasn't initialized",
+ *socket);
+ return -EINVAL;
+ }
+ mci = new_mci;
+ pvt = mci->pvt_info;
+
+ /*
+ * Step 2) Get memory channel
+ */
+ prv = 0;
+ pci_ha = pvt->pci_ha;
+ for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
+ pci_read_config_dword(pci_ha, tad_dram_rule[n_tads], &reg);
+ limit = TAD_LIMIT(reg);
+ if (limit <= prv) {
+ sprintf(msg, "Can't discover the memory channel");
+ return -EINVAL;
+ }
+ if (addr <= limit)
+ break;
+ prv = limit;
+ }
+ if (n_tads == MAX_TAD) {
+ sprintf(msg, "Can't discover the memory channel");
+ return -EINVAL;
+ }
+
+ ch_way = TAD_CH(reg) + 1;
+ sck_way = TAD_SOCK(reg);
+
+ if (ch_way == 3)
+ idx = addr >> 6;
+ else {
+ idx = (addr >> (6 + sck_way + shiftup)) & 0x3;
+ if (pvt->is_chan_hash)
+ idx = haswell_chan_hash(idx, addr);
+ }
+ idx = idx % ch_way;
+
+ /*
+ * FIXME: Shouldn't we use CHN_IDX_OFFSET() here, when ch_way == 3 ???
+ */
+ switch (idx) {
+ case 0:
+ base_ch = TAD_TGT0(reg);
+ break;
+ case 1:
+ base_ch = TAD_TGT1(reg);
+ break;
+ case 2:
+ base_ch = TAD_TGT2(reg);
+ break;
+ case 3:
+ base_ch = TAD_TGT3(reg);
+ break;
+ default:
+ sprintf(msg, "Can't discover the TAD target");
+ return -EINVAL;
+ }
+ *channel_mask = 1 << base_ch;
+
+ pci_read_config_dword(pvt->pci_tad[base_ch], tad_ch_nilv_offset[n_tads], &tad_offset);
+
+ if (pvt->mirror_mode == FULL_MIRRORING ||
+ (pvt->mirror_mode == ADDR_RANGE_MIRRORING && n_tads == 0)) {
+ *channel_mask |= 1 << ((base_ch + 2) % 4);
+ switch(ch_way) {
+ case 2:
+ case 4:
+ sck_xch = (1 << sck_way) * (ch_way >> 1);
+ break;
+ default:
+ sprintf(msg, "Invalid mirror set. Can't decode addr");
+ return -EINVAL;
+ }
+
+ pvt->is_cur_addr_mirrored = true;
+ } else {
+ sck_xch = (1 << sck_way) * ch_way;
+ pvt->is_cur_addr_mirrored = false;
+ }
+
+ if (pvt->is_lockstep)
+ *channel_mask |= 1 << ((base_ch + 1) % 4);
+
+ offset = TAD_OFFSET(tad_offset);
+
+ edac_dbg(0, "TAD#%d: address 0x%016Lx < 0x%016Lx, socket interleave %d, channel interleave %d (offset 0x%08Lx), index %d, base ch: %d, ch mask: 0x%02lx\n",
+ n_tads,
+ addr,
+ limit,
+ sck_way,
+ ch_way,
+ offset,
+ idx,
+ base_ch,
+ *channel_mask);
+
+ /* Calculate channel address */
+ /* Remove the TAD offset */
+
+ if (offset > addr) {
+ sprintf(msg, "Can't calculate ch addr: TAD offset 0x%08Lx is too high for addr 0x%08Lx!",
+ offset, addr);
+ return -EINVAL;
+ }
+
+ ch_addr = addr - offset;
+ ch_addr >>= (6 + shiftup);
+ ch_addr /= sck_xch;
+ ch_addr <<= (6 + shiftup);
+ ch_addr |= addr & ((1 << (6 + shiftup)) - 1);
+
+ /*
+ * Step 3) Decode rank
+ */
+ for (n_rir = 0; n_rir < MAX_RIR_RANGES; n_rir++) {
+ pci_read_config_dword(pvt->pci_tad[base_ch], rir_way_limit[n_rir], &reg);
+
+ if (!IS_RIR_VALID(reg))
+ continue;
+
+ limit = pvt->info.rir_limit(reg);
+ gb = div_u64_rem(limit >> 20, 1024, &mb);
+ edac_dbg(0, "RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d\n",
+ n_rir,
+ gb, (mb*1000)/1024,
+ limit,
+ 1 << RIR_WAY(reg));
+ if (ch_addr <= limit)
+ break;
+ }
+ if (n_rir == MAX_RIR_RANGES) {
+ sprintf(msg, "Can't discover the memory rank for ch addr 0x%08Lx",
+ ch_addr);
+ return -EINVAL;
+ }
+ rir_way = RIR_WAY(reg);
+
+ if (pvt->is_close_pg)
+ idx = (ch_addr >> 6);
+ else
+ idx = (ch_addr >> 13); /* FIXME: Datasheet says to shift by 15 */
+ idx %= 1 << rir_way;
+
+ pci_read_config_dword(pvt->pci_tad[base_ch], rir_offset[n_rir][idx], &reg);
+ *rank = RIR_RNK_TGT(pvt->info.type, reg);
+
+ if (pvt->info.type == BROADWELL) {
+ if (pvt->is_close_pg)
+ shiftup = 6;
+ else
+ shiftup = 13;
+
+ rank_addr = ch_addr >> shiftup;
+ rank_addr /= (1 << rir_way);
+ rank_addr <<= shiftup;
+ rank_addr |= ch_addr & GENMASK_ULL(shiftup - 1, 0);
+ rank_addr -= RIR_OFFSET(pvt->info.type, reg);
+
+ mtype = pvt->info.get_memory_type(pvt);
+ rankid = *rank;
+ if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
+ sb_decode_ddr4(mci, base_ch, rankid, rank_addr, msg);
+ else
+ sb_decode_ddr3(mci, base_ch, rankid, rank_addr, msg);
+ } else {
+ msg[0] = '\0';
+ }
+
+ edac_dbg(0, "RIR#%d: channel address 0x%08Lx < 0x%08Lx, RIR interleave %d, index %d\n",
+ n_rir,
+ ch_addr,
+ limit,
+ rir_way,
+ idx);
+
+ return 0;
+}
+
+static int get_memory_error_data_from_mce(struct mem_ctl_info *mci,
+ const struct mce *m, u8 *socket,
+ u8 *ha, long *channel_mask,
+ char *msg)
+{
+ u32 reg, channel = GET_BITFIELD(m->status, 0, 3);
+ struct mem_ctl_info *new_mci;
+ struct sbridge_pvt *pvt;
+ struct pci_dev *pci_ha;
+ bool tad0;
+
+ if (channel >= NUM_CHANNELS) {
+ sprintf(msg, "Invalid channel 0x%x", channel);
+ return -EINVAL;
+ }
+
+ pvt = mci->pvt_info;
+ if (!pvt->info.get_ha) {
+ sprintf(msg, "No get_ha()");
+ return -EINVAL;
+ }
+ *ha = pvt->info.get_ha(m->bank);
+ if (*ha != 0 && *ha != 1) {
+ sprintf(msg, "Impossible bank %d", m->bank);
+ return -EINVAL;
+ }
+
+ *socket = m->socketid;
+ new_mci = get_mci_for_node_id(*socket, *ha);
+ if (!new_mci) {
+ strcpy(msg, "mci socket got corrupted!");
+ return -EINVAL;
+ }
+
+ pvt = new_mci->pvt_info;
+ pci_ha = pvt->pci_ha;
+ pci_read_config_dword(pci_ha, tad_dram_rule[0], &reg);
+ tad0 = m->addr <= TAD_LIMIT(reg);
+
+ *channel_mask = 1 << channel;
+ if (pvt->mirror_mode == FULL_MIRRORING ||
+ (pvt->mirror_mode == ADDR_RANGE_MIRRORING && tad0)) {
+ *channel_mask |= 1 << ((channel + 2) % 4);
+ pvt->is_cur_addr_mirrored = true;
+ } else {
+ pvt->is_cur_addr_mirrored = false;
+ }
+
+ if (pvt->is_lockstep)
+ *channel_mask |= 1 << ((channel + 1) % 4);
+
+ return 0;
+}
+
+/****************************************************************************
+ Device initialization routines: put/get, init/exit
+ ****************************************************************************/
+
+/*
+ * sbridge_put_all_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void sbridge_put_devices(struct sbridge_dev *sbridge_dev)
+{
+ int i;
+
+ edac_dbg(0, "\n");
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ struct pci_dev *pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+ edac_dbg(0, "Removing dev %02x:%02x.%d\n",
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+ pci_dev_put(pdev);
+ }
+}
+
+static void sbridge_put_all_devices(void)
+{
+ struct sbridge_dev *sbridge_dev, *tmp;
+
+ list_for_each_entry_safe(sbridge_dev, tmp, &sbridge_edac_list, list) {
+ sbridge_put_devices(sbridge_dev);
+ free_sbridge_dev(sbridge_dev);
+ }
+}
+
+static int sbridge_get_onedevice(struct pci_dev **prev,
+ u8 *num_mc,
+ const struct pci_id_table *table,
+ const unsigned devno,
+ const int multi_bus)
+{
+ struct sbridge_dev *sbridge_dev = NULL;
+ const struct pci_id_descr *dev_descr = &table->descr[devno];
+ struct pci_dev *pdev = NULL;
+ int seg = 0;
+ u8 bus = 0;
+ int i = 0;
+
+ sbridge_printk(KERN_DEBUG,
+ "Seeking for: PCI ID %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ dev_descr->dev_id, *prev);
+
+ if (!pdev) {
+ if (*prev) {
+ *prev = pdev;
+ return 0;
+ }
+
+ if (dev_descr->optional)
+ return 0;
+
+ /* if the HA wasn't found */
+ if (devno == 0)
+ return -ENODEV;
+
+ sbridge_printk(KERN_INFO,
+ "Device not found: %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+ /* End of list, leave */
+ return -ENODEV;
+ }
+ seg = pci_domain_nr(pdev->bus);
+ bus = pdev->bus->number;
+
+next_imc:
+ sbridge_dev = get_sbridge_dev(seg, bus, dev_descr->dom,
+ multi_bus, sbridge_dev);
+ if (!sbridge_dev) {
+ /* If the HA1 wasn't found, don't create EDAC second memory controller */
+ if (dev_descr->dom == IMC1 && devno != 1) {
+ edac_dbg(0, "Skip IMC1: %04x:%04x (since HA1 was absent)\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+ pci_dev_put(pdev);
+ return 0;
+ }
+
+ if (dev_descr->dom == SOCK)
+ goto out_imc;
+
+ sbridge_dev = alloc_sbridge_dev(seg, bus, dev_descr->dom, table);
+ if (!sbridge_dev) {
+ pci_dev_put(pdev);
+ return -ENOMEM;
+ }
+ (*num_mc)++;
+ }
+
+ if (sbridge_dev->pdev[sbridge_dev->i_devs]) {
+ sbridge_printk(KERN_ERR,
+ "Duplicated device for %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+ pci_dev_put(pdev);
+ return -ENODEV;
+ }
+
+ sbridge_dev->pdev[sbridge_dev->i_devs++] = pdev;
+
+ /* pdev belongs to more than one IMC, do extra gets */
+ if (++i > 1)
+ pci_dev_get(pdev);
+
+ if (dev_descr->dom == SOCK && i < table->n_imcs_per_sock)
+ goto next_imc;
+
+out_imc:
+ /* Be sure that the device is enabled */
+ if (unlikely(pci_enable_device(pdev) < 0)) {
+ sbridge_printk(KERN_ERR,
+ "Couldn't enable %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+ return -ENODEV;
+ }
+
+ edac_dbg(0, "Detected %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+ /*
+ * As stated on drivers/pci/search.c, the reference count for
+ * @from is always decremented if it is not %NULL. So, as we need
+ * to get all devices up to null, we need to do a get for the device
+ */
+ pci_dev_get(pdev);
+
+ *prev = pdev;
+
+ return 0;
+}
+
+/*
+ * sbridge_get_all_devices - Find and perform 'get' operation on the MCH's
+ * devices we want to reference for this driver.
+ * @num_mc: pointer to the memory controllers count, to be incremented in case
+ * of success.
+ * @table: model specific table
+ *
+ * returns 0 in case of success or error code
+ */
+static int sbridge_get_all_devices(u8 *num_mc,
+ const struct pci_id_table *table)
+{
+ int i, rc;
+ struct pci_dev *pdev = NULL;
+ int allow_dups = 0;
+ int multi_bus = 0;
+
+ if (table->type == KNIGHTS_LANDING)
+ allow_dups = multi_bus = 1;
+ while (table && table->descr) {
+ for (i = 0; i < table->n_devs_per_sock; i++) {
+ if (!allow_dups || i == 0 ||
+ table->descr[i].dev_id !=
+ table->descr[i-1].dev_id) {
+ pdev = NULL;
+ }
+ do {
+ rc = sbridge_get_onedevice(&pdev, num_mc,
+ table, i, multi_bus);
+ if (rc < 0) {
+ if (i == 0) {
+ i = table->n_devs_per_sock;
+ break;
+ }
+ sbridge_put_all_devices();
+ return -ENODEV;
+ }
+ } while (pdev && !allow_dups);
+ }
+ table++;
+ }
+
+ return 0;
+}
+
+/*
+ * Device IDs for {SBRIDGE,IBRIDGE,HASWELL,BROADWELL}_IMC_HA0_TAD0 are in
+ * the format: XXXa. So we can convert from a device to the corresponding
+ * channel like this
+ */
+#define TAD_DEV_TO_CHAN(dev) (((dev) & 0xf) - 0xa)
+
+static int sbridge_mci_bind_devs(struct mem_ctl_info *mci,
+ struct sbridge_dev *sbridge_dev)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ u8 saw_chan_mask = 0;
+ int i;
+
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0:
+ pvt->pci_sad0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1:
+ pvt->pci_sad1 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_BR:
+ pvt->pci_br0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
+ pvt->pci_ha = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA:
+ pvt->pci_ta = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_RAS:
+ pvt->pci_ras = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0:
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD1:
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD2:
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD3:
+ {
+ int id = TAD_DEV_TO_CHAN(pdev->device);
+ pvt->pci_tad[id] = pdev;
+ saw_chan_mask |= 1 << id;
+ }
+ break;
+ case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO:
+ pvt->pci_ddrio = pdev;
+ break;
+ default:
+ goto error;
+ }
+
+ edac_dbg(0, "Associated PCI %02x:%02x, bus %d with dev = %p\n",
+ pdev->vendor, pdev->device,
+ sbridge_dev->bus,
+ pdev);
+ }
+
+ /* Check if everything were registered */
+ if (!pvt->pci_sad0 || !pvt->pci_sad1 || !pvt->pci_ha ||
+ !pvt->pci_ras || !pvt->pci_ta)
+ goto enodev;
+
+ if (saw_chan_mask != 0x0f)
+ goto enodev;
+ return 0;
+
+enodev:
+ sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
+ return -ENODEV;
+
+error:
+ sbridge_printk(KERN_ERR, "Unexpected device %02x:%02x\n",
+ PCI_VENDOR_ID_INTEL, pdev->device);
+ return -EINVAL;
+}
+
+static int ibridge_mci_bind_devs(struct mem_ctl_info *mci,
+ struct sbridge_dev *sbridge_dev)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ u8 saw_chan_mask = 0;
+ int i;
+
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1:
+ pvt->pci_ha = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA:
+ pvt->pci_ta = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS:
+ pvt->pci_ras = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2:
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3:
+ {
+ int id = TAD_DEV_TO_CHAN(pdev->device);
+ pvt->pci_tad[id] = pdev;
+ saw_chan_mask |= 1 << id;
+ }
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0:
+ pvt->pci_ddrio = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0:
+ pvt->pci_ddrio = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_SAD:
+ pvt->pci_sad0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_BR0:
+ pvt->pci_br0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_IBRIDGE_BR1:
+ pvt->pci_br1 = pdev;
+ break;
+ default:
+ goto error;
+ }
+
+ edac_dbg(0, "Associated PCI %02x.%02d.%d with dev = %p\n",
+ sbridge_dev->bus,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+ pdev);
+ }
+
+ /* Check if everything were registered */
+ if (!pvt->pci_sad0 || !pvt->pci_ha || !pvt->pci_br0 ||
+ !pvt->pci_br1 || !pvt->pci_ras || !pvt->pci_ta)
+ goto enodev;
+
+ if (saw_chan_mask != 0x0f && /* -EN/-EX */
+ saw_chan_mask != 0x03) /* -EP */
+ goto enodev;
+ return 0;
+
+enodev:
+ sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
+ return -ENODEV;
+
+error:
+ sbridge_printk(KERN_ERR,
+ "Unexpected device %02x:%02x\n", PCI_VENDOR_ID_INTEL,
+ pdev->device);
+ return -EINVAL;
+}
+
+static int haswell_mci_bind_devs(struct mem_ctl_info *mci,
+ struct sbridge_dev *sbridge_dev)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ u8 saw_chan_mask = 0;
+ int i;
+
+ /* there's only one device per system; not tied to any bus */
+ if (pvt->info.pci_vtd == NULL)
+ /* result will be checked later */
+ pvt->info.pci_vtd = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_HASWELL_IMC_VTD_MISC,
+ NULL);
+
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0:
+ pvt->pci_sad0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1:
+ pvt->pci_sad1 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1:
+ pvt->pci_ha = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA:
+ pvt->pci_ta = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TM:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TM:
+ pvt->pci_ras = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3:
+ {
+ int id = TAD_DEV_TO_CHAN(pdev->device);
+ pvt->pci_tad[id] = pdev;
+ saw_chan_mask |= 1 << id;
+ }
+ break;
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2:
+ case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3:
+ if (!pvt->pci_ddrio)
+ pvt->pci_ddrio = pdev;
+ break;
+ default:
+ break;
+ }
+
+ edac_dbg(0, "Associated PCI %02x.%02d.%d with dev = %p\n",
+ sbridge_dev->bus,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+ pdev);
+ }
+
+ /* Check if everything were registered */
+ if (!pvt->pci_sad0 || !pvt->pci_ha || !pvt->pci_sad1 ||
+ !pvt->pci_ras || !pvt->pci_ta || !pvt->info.pci_vtd)
+ goto enodev;
+
+ if (saw_chan_mask != 0x0f && /* -EN/-EX */
+ saw_chan_mask != 0x03) /* -EP */
+ goto enodev;
+ return 0;
+
+enodev:
+ sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
+ return -ENODEV;
+}
+
+static int broadwell_mci_bind_devs(struct mem_ctl_info *mci,
+ struct sbridge_dev *sbridge_dev)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ u8 saw_chan_mask = 0;
+ int i;
+
+ /* there's only one device per system; not tied to any bus */
+ if (pvt->info.pci_vtd == NULL)
+ /* result will be checked later */
+ pvt->info.pci_vtd = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_BROADWELL_IMC_VTD_MISC,
+ NULL);
+
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0:
+ pvt->pci_sad0 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1:
+ pvt->pci_sad1 = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1:
+ pvt->pci_ha = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA:
+ pvt->pci_ta = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TM:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TM:
+ pvt->pci_ras = pdev;
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2:
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3:
+ {
+ int id = TAD_DEV_TO_CHAN(pdev->device);
+ pvt->pci_tad[id] = pdev;
+ saw_chan_mask |= 1 << id;
+ }
+ break;
+ case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0:
+ pvt->pci_ddrio = pdev;
+ break;
+ default:
+ break;
+ }
+
+ edac_dbg(0, "Associated PCI %02x.%02d.%d with dev = %p\n",
+ sbridge_dev->bus,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+ pdev);
+ }
+
+ /* Check if everything were registered */
+ if (!pvt->pci_sad0 || !pvt->pci_ha || !pvt->pci_sad1 ||
+ !pvt->pci_ras || !pvt->pci_ta || !pvt->info.pci_vtd)
+ goto enodev;
+
+ if (saw_chan_mask != 0x0f && /* -EN/-EX */
+ saw_chan_mask != 0x03) /* -EP */
+ goto enodev;
+ return 0;
+
+enodev:
+ sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
+ return -ENODEV;
+}
+
+static int knl_mci_bind_devs(struct mem_ctl_info *mci,
+ struct sbridge_dev *sbridge_dev)
+{
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ struct pci_dev *pdev;
+ int dev, func;
+
+ int i;
+ int devidx;
+
+ for (i = 0; i < sbridge_dev->n_devs; i++) {
+ pdev = sbridge_dev->pdev[i];
+ if (!pdev)
+ continue;
+
+ /* Extract PCI device and function. */
+ dev = (pdev->devfn >> 3) & 0x1f;
+ func = pdev->devfn & 0x7;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_MC:
+ if (dev == 8)
+ pvt->knl.pci_mc0 = pdev;
+ else if (dev == 9)
+ pvt->knl.pci_mc1 = pdev;
+ else {
+ sbridge_printk(KERN_ERR,
+ "Memory controller in unexpected place! (dev %d, fn %d)\n",
+ dev, func);
+ continue;
+ }
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0:
+ pvt->pci_sad0 = pdev;
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1:
+ pvt->pci_sad1 = pdev;
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_CHA:
+ /* There are one of these per tile, and range from
+ * 1.14.0 to 1.18.5.
+ */
+ devidx = ((dev-14)*8)+func;
+
+ if (devidx < 0 || devidx >= KNL_MAX_CHAS) {
+ sbridge_printk(KERN_ERR,
+ "Caching and Home Agent in unexpected place! (dev %d, fn %d)\n",
+ dev, func);
+ continue;
+ }
+
+ WARN_ON(pvt->knl.pci_cha[devidx] != NULL);
+
+ pvt->knl.pci_cha[devidx] = pdev;
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_CHAN:
+ devidx = -1;
+
+ /*
+ * MC0 channels 0-2 are device 9 function 2-4,
+ * MC1 channels 3-5 are device 8 function 2-4.
+ */
+
+ if (dev == 9)
+ devidx = func-2;
+ else if (dev == 8)
+ devidx = 3 + (func-2);
+
+ if (devidx < 0 || devidx >= KNL_MAX_CHANNELS) {
+ sbridge_printk(KERN_ERR,
+ "DRAM Channel Registers in unexpected place! (dev %d, fn %d)\n",
+ dev, func);
+ continue;
+ }
+
+ WARN_ON(pvt->knl.pci_channel[devidx] != NULL);
+ pvt->knl.pci_channel[devidx] = pdev;
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM:
+ pvt->knl.pci_mc_info = pdev;
+ break;
+
+ case PCI_DEVICE_ID_INTEL_KNL_IMC_TA:
+ pvt->pci_ta = pdev;
+ break;
+
+ default:
+ sbridge_printk(KERN_ERR, "Unexpected device %d\n",
+ pdev->device);
+ break;
+ }
+ }
+
+ if (!pvt->knl.pci_mc0 || !pvt->knl.pci_mc1 ||
+ !pvt->pci_sad0 || !pvt->pci_sad1 ||
+ !pvt->pci_ta) {
+ goto enodev;
+ }
+
+ for (i = 0; i < KNL_MAX_CHANNELS; i++) {
+ if (!pvt->knl.pci_channel[i]) {
+ sbridge_printk(KERN_ERR, "Missing channel %d\n", i);
+ goto enodev;
+ }
+ }
+
+ for (i = 0; i < KNL_MAX_CHAS; i++) {
+ if (!pvt->knl.pci_cha[i]) {
+ sbridge_printk(KERN_ERR, "Missing CHA %d\n", i);
+ goto enodev;
+ }
+ }
+
+ return 0;
+
+enodev:
+ sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
+ return -ENODEV;
+}
+
+/****************************************************************************
+ Error check routines
+ ****************************************************************************/
+
+/*
+ * While Sandy Bridge has error count registers, SMI BIOS read values from
+ * and resets the counters. So, they are not reliable for the OS to read
+ * from them. So, we have no option but to just trust on whatever MCE is
+ * telling us about the errors.
+ */
+static void sbridge_mce_output_error(struct mem_ctl_info *mci,
+ const struct mce *m)
+{
+ struct mem_ctl_info *new_mci;
+ struct sbridge_pvt *pvt = mci->pvt_info;
+ enum hw_event_mc_err_type tp_event;
+ char *optype, msg[256], msg_full[512];
+ bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
+ bool overflow = GET_BITFIELD(m->status, 62, 62);
+ bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
+ bool recoverable;
+ u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52);
+ u32 mscod = GET_BITFIELD(m->status, 16, 31);
+ u32 errcode = GET_BITFIELD(m->status, 0, 15);
+ u32 channel = GET_BITFIELD(m->status, 0, 3);
+ u32 optypenum = GET_BITFIELD(m->status, 4, 6);
+ /*
+ * Bits 5-0 of MCi_MISC give the least significant bit that is valid.
+ * A value 6 is for cache line aligned address, a value 12 is for page
+ * aligned address reported by patrol scrubber.
+ */
+ u32 lsb = GET_BITFIELD(m->misc, 0, 5);
+ long channel_mask, first_channel;
+ u8 rank = 0xff, socket, ha;
+ int rc, dimm;
+ char *area_type = "DRAM";
+
+ if (pvt->info.type != SANDY_BRIDGE)
+ recoverable = true;
+ else
+ recoverable = GET_BITFIELD(m->status, 56, 56);
+
+ if (uncorrected_error) {
+ core_err_cnt = 1;
+ if (ripv) {
+ tp_event = HW_EVENT_ERR_UNCORRECTED;
+ } else {
+ tp_event = HW_EVENT_ERR_FATAL;
+ }
+ } else {
+ tp_event = HW_EVENT_ERR_CORRECTED;
+ }
+
+ /*
+ * According with Table 15-9 of the Intel Architecture spec vol 3A,
+ * memory errors should fit in this mask:
+ * 000f 0000 1mmm cccc (binary)
+ * where:
+ * f = Correction Report Filtering Bit. If 1, subsequent errors
+ * won't be shown
+ * mmm = error type
+ * cccc = channel
+ * If the mask doesn't match, report an error to the parsing logic
+ */
+ switch (optypenum) {
+ case 0:
+ optype = "generic undef request error";
+ break;
+ case 1:
+ optype = "memory read error";
+ break;
+ case 2:
+ optype = "memory write error";
+ break;
+ case 3:
+ optype = "addr/cmd error";
+ break;
+ case 4:
+ optype = "memory scrubbing error";
+ break;
+ default:
+ optype = "reserved";
+ break;
+ }
+
+ if (pvt->info.type == KNIGHTS_LANDING) {
+ if (channel == 14) {
+ edac_dbg(0, "%s%s err_code:%04x:%04x EDRAM bank %d\n",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable)
+ ? " recoverable" : "",
+ mscod, errcode,
+ m->bank);
+ } else {
+ char A = *("A");
+
+ /*
+ * Reported channel is in range 0-2, so we can't map it
+ * back to mc. To figure out mc we check machine check
+ * bank register that reported this error.
+ * bank15 means mc0 and bank16 means mc1.
+ */
+ channel = knl_channel_remap(m->bank == 16, channel);
+ channel_mask = 1 << channel;
+
+ snprintf(msg, sizeof(msg),
+ "%s%s err_code:%04x:%04x channel:%d (DIMM_%c)",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable)
+ ? " recoverable" : " ",
+ mscod, errcode, channel, A + channel);
+ edac_mc_handle_error(tp_event, mci, core_err_cnt,
+ m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
+ channel, 0, -1,
+ optype, msg);
+ }
+ return;
+ } else if (lsb < 12) {
+ rc = get_memory_error_data(mci, m->addr, &socket, &ha,
+ &channel_mask, &rank,
+ &area_type, msg);
+ } else {
+ rc = get_memory_error_data_from_mce(mci, m, &socket, &ha,
+ &channel_mask, msg);
+ }
+
+ if (rc < 0)
+ goto err_parsing;
+ new_mci = get_mci_for_node_id(socket, ha);
+ if (!new_mci) {
+ strcpy(msg, "Error: socket got corrupted!");
+ goto err_parsing;
+ }
+ mci = new_mci;
+ pvt = mci->pvt_info;
+
+ first_channel = find_first_bit(&channel_mask, NUM_CHANNELS);
+
+ if (rank == 0xff)
+ dimm = -1;
+ else if (rank < 4)
+ dimm = 0;
+ else if (rank < 8)
+ dimm = 1;
+ else
+ dimm = 2;
+
+ /*
+ * FIXME: On some memory configurations (mirror, lockstep), the
+ * Memory Controller can't point the error to a single DIMM. The
+ * EDAC core should be handling the channel mask, in order to point
+ * to the group of dimm's where the error may be happening.
+ */
+ if (!pvt->is_lockstep && !pvt->is_cur_addr_mirrored && !pvt->is_close_pg)
+ channel = first_channel;
+ snprintf(msg_full, sizeof(msg_full),
+ "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d %s",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable) ? " recoverable" : "",
+ area_type,
+ mscod, errcode,
+ socket, ha,
+ channel_mask,
+ rank, msg);
+
+ edac_dbg(0, "%s\n", msg_full);
+
+ /* FIXME: need support for channel mask */
+
+ if (channel == CHANNEL_UNSPECIFIED)
+ channel = -1;
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(tp_event, mci, core_err_cnt,
+ m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
+ channel, dimm, -1,
+ optype, msg_full);
+ return;
+err_parsing:
+ edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
+ -1, -1, -1,
+ msg, "");
+
+}
+
+/*
+ * Check that logging is enabled and that this is the right type
+ * of error for us to handle.
+ */
+static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct mce *mce = (struct mce *)data;
+ struct mem_ctl_info *mci;
+ char *type;
+
+ if (mce->kflags & MCE_HANDLED_CEC)
+ return NOTIFY_DONE;
+
+ /*
+ * Just let mcelog handle it if the error is
+ * outside the memory controller. A memory error
+ * is indicated by bit 7 = 1 and bits = 8-11,13-15 = 0.
+ * bit 12 has an special meaning.
+ */
+ if ((mce->status & 0xefff) >> 7 != 1)
+ return NOTIFY_DONE;
+
+ /* Check ADDRV bit in STATUS */
+ if (!GET_BITFIELD(mce->status, 58, 58))
+ return NOTIFY_DONE;
+
+ /* Check MISCV bit in STATUS */
+ if (!GET_BITFIELD(mce->status, 59, 59))
+ return NOTIFY_DONE;
+
+ /* Check address type in MISC (physical address only) */
+ if (GET_BITFIELD(mce->misc, 6, 8) != 2)
+ return NOTIFY_DONE;
+
+ mci = get_mci_for_node_id(mce->socketid, IMC0);
+ if (!mci)
+ return NOTIFY_DONE;
+
+ if (mce->mcgstatus & MCG_STATUS_MCIP)
+ type = "Exception";
+ else
+ type = "Event";
+
+ sbridge_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
+
+ sbridge_mc_printk(mci, KERN_DEBUG, "CPU %d: Machine Check %s: %Lx "
+ "Bank %d: %016Lx\n", mce->extcpu, type,
+ mce->mcgstatus, mce->bank, mce->status);
+ sbridge_mc_printk(mci, KERN_DEBUG, "TSC %llx ", mce->tsc);
+ sbridge_mc_printk(mci, KERN_DEBUG, "ADDR %llx ", mce->addr);
+ sbridge_mc_printk(mci, KERN_DEBUG, "MISC %llx ", mce->misc);
+
+ sbridge_mc_printk(mci, KERN_DEBUG, "PROCESSOR %u:%x TIME %llu SOCKET "
+ "%u APIC %x\n", mce->cpuvendor, mce->cpuid,
+ mce->time, mce->socketid, mce->apicid);
+
+ sbridge_mce_output_error(mci, mce);
+
+ /* Advice mcelog that the error were handled */
+ mce->kflags |= MCE_HANDLED_EDAC;
+ return NOTIFY_OK;
+}
+
+static struct notifier_block sbridge_mce_dec = {
+ .notifier_call = sbridge_mce_check_error,
+ .priority = MCE_PRIO_EDAC,
+};
+
+/****************************************************************************
+ EDAC register/unregister logic
+ ****************************************************************************/
+
+static void sbridge_unregister_mci(struct sbridge_dev *sbridge_dev)
+{
+ struct mem_ctl_info *mci = sbridge_dev->mci;
+
+ if (unlikely(!mci || !mci->pvt_info)) {
+ edac_dbg(0, "MC: dev = %p\n", &sbridge_dev->pdev[0]->dev);
+
+ sbridge_printk(KERN_ERR, "Couldn't find mci handler\n");
+ return;
+ }
+
+ edac_dbg(0, "MC: mci = %p, dev = %p\n",
+ mci, &sbridge_dev->pdev[0]->dev);
+
+ /* Remove MC sysfs nodes */
+ edac_mc_del_mc(mci->pdev);
+
+ edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
+ kfree(mci->ctl_name);
+ edac_mc_free(mci);
+ sbridge_dev->mci = NULL;
+}
+
+static int sbridge_register_mci(struct sbridge_dev *sbridge_dev, enum type type)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct sbridge_pvt *pvt;
+ struct pci_dev *pdev = sbridge_dev->pdev[0];
+ int rc;
+
+ /* allocate a new MC control structure */
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = type == KNIGHTS_LANDING ?
+ KNL_MAX_CHANNELS : NUM_CHANNELS;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = type == KNIGHTS_LANDING ? 1 : MAX_DIMMS;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(sbridge_dev->mc, ARRAY_SIZE(layers), layers,
+ sizeof(*pvt));
+
+ if (unlikely(!mci))
+ return -ENOMEM;
+
+ edac_dbg(0, "MC: mci = %p, dev = %p\n",
+ mci, &pdev->dev);
+
+ pvt = mci->pvt_info;
+ memset(pvt, 0, sizeof(*pvt));
+
+ /* Associate sbridge_dev and mci for future usage */
+ pvt->sbridge_dev = sbridge_dev;
+ sbridge_dev->mci = mci;
+
+ mci->mtype_cap = type == KNIGHTS_LANDING ?
+ MEM_FLAG_DDR4 : MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ pvt->info.type = type;
+ switch (type) {
+ case IVY_BRIDGE:
+ pvt->info.rankcfgr = IB_RANK_CFG_A;
+ pvt->info.get_tolm = ibridge_get_tolm;
+ pvt->info.get_tohm = ibridge_get_tohm;
+ pvt->info.dram_rule = ibridge_dram_rule;
+ pvt->info.get_memory_type = get_memory_type;
+ pvt->info.get_node_id = get_node_id;
+ pvt->info.get_ha = ibridge_get_ha;
+ pvt->info.rir_limit = rir_limit;
+ pvt->info.sad_limit = sad_limit;
+ pvt->info.interleave_mode = interleave_mode;
+ pvt->info.dram_attr = dram_attr;
+ pvt->info.max_sad = ARRAY_SIZE(ibridge_dram_rule);
+ pvt->info.interleave_list = ibridge_interleave_list;
+ pvt->info.interleave_pkg = ibridge_interleave_pkg;
+ pvt->info.get_width = ibridge_get_width;
+
+ /* Store pci devices at mci for faster access */
+ rc = ibridge_mci_bind_devs(mci, sbridge_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+ get_source_id(mci);
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Ivy Bridge SrcID#%d_Ha#%d",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom);
+ break;
+ case SANDY_BRIDGE:
+ pvt->info.rankcfgr = SB_RANK_CFG_A;
+ pvt->info.get_tolm = sbridge_get_tolm;
+ pvt->info.get_tohm = sbridge_get_tohm;
+ pvt->info.dram_rule = sbridge_dram_rule;
+ pvt->info.get_memory_type = get_memory_type;
+ pvt->info.get_node_id = get_node_id;
+ pvt->info.get_ha = sbridge_get_ha;
+ pvt->info.rir_limit = rir_limit;
+ pvt->info.sad_limit = sad_limit;
+ pvt->info.interleave_mode = interleave_mode;
+ pvt->info.dram_attr = dram_attr;
+ pvt->info.max_sad = ARRAY_SIZE(sbridge_dram_rule);
+ pvt->info.interleave_list = sbridge_interleave_list;
+ pvt->info.interleave_pkg = sbridge_interleave_pkg;
+ pvt->info.get_width = sbridge_get_width;
+
+ /* Store pci devices at mci for faster access */
+ rc = sbridge_mci_bind_devs(mci, sbridge_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+ get_source_id(mci);
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Sandy Bridge SrcID#%d_Ha#%d",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom);
+ break;
+ case HASWELL:
+ /* rankcfgr isn't used */
+ pvt->info.get_tolm = haswell_get_tolm;
+ pvt->info.get_tohm = haswell_get_tohm;
+ pvt->info.dram_rule = ibridge_dram_rule;
+ pvt->info.get_memory_type = haswell_get_memory_type;
+ pvt->info.get_node_id = haswell_get_node_id;
+ pvt->info.get_ha = ibridge_get_ha;
+ pvt->info.rir_limit = haswell_rir_limit;
+ pvt->info.sad_limit = sad_limit;
+ pvt->info.interleave_mode = interleave_mode;
+ pvt->info.dram_attr = dram_attr;
+ pvt->info.max_sad = ARRAY_SIZE(ibridge_dram_rule);
+ pvt->info.interleave_list = ibridge_interleave_list;
+ pvt->info.interleave_pkg = ibridge_interleave_pkg;
+ pvt->info.get_width = ibridge_get_width;
+
+ /* Store pci devices at mci for faster access */
+ rc = haswell_mci_bind_devs(mci, sbridge_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+ get_source_id(mci);
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Haswell SrcID#%d_Ha#%d",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom);
+ break;
+ case BROADWELL:
+ /* rankcfgr isn't used */
+ pvt->info.get_tolm = haswell_get_tolm;
+ pvt->info.get_tohm = haswell_get_tohm;
+ pvt->info.dram_rule = ibridge_dram_rule;
+ pvt->info.get_memory_type = haswell_get_memory_type;
+ pvt->info.get_node_id = haswell_get_node_id;
+ pvt->info.get_ha = ibridge_get_ha;
+ pvt->info.rir_limit = haswell_rir_limit;
+ pvt->info.sad_limit = sad_limit;
+ pvt->info.interleave_mode = interleave_mode;
+ pvt->info.dram_attr = dram_attr;
+ pvt->info.max_sad = ARRAY_SIZE(ibridge_dram_rule);
+ pvt->info.interleave_list = ibridge_interleave_list;
+ pvt->info.interleave_pkg = ibridge_interleave_pkg;
+ pvt->info.get_width = broadwell_get_width;
+
+ /* Store pci devices at mci for faster access */
+ rc = broadwell_mci_bind_devs(mci, sbridge_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+ get_source_id(mci);
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Broadwell SrcID#%d_Ha#%d",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom);
+ break;
+ case KNIGHTS_LANDING:
+ /* pvt->info.rankcfgr == ??? */
+ pvt->info.get_tolm = knl_get_tolm;
+ pvt->info.get_tohm = knl_get_tohm;
+ pvt->info.dram_rule = knl_dram_rule;
+ pvt->info.get_memory_type = knl_get_memory_type;
+ pvt->info.get_node_id = knl_get_node_id;
+ pvt->info.get_ha = knl_get_ha;
+ pvt->info.rir_limit = NULL;
+ pvt->info.sad_limit = knl_sad_limit;
+ pvt->info.interleave_mode = knl_interleave_mode;
+ pvt->info.dram_attr = dram_attr_knl;
+ pvt->info.max_sad = ARRAY_SIZE(knl_dram_rule);
+ pvt->info.interleave_list = knl_interleave_list;
+ pvt->info.interleave_pkg = ibridge_interleave_pkg;
+ pvt->info.get_width = knl_get_width;
+
+ rc = knl_mci_bind_devs(mci, sbridge_dev);
+ if (unlikely(rc < 0))
+ goto fail0;
+ get_source_id(mci);
+ mci->ctl_name = kasprintf(GFP_KERNEL, "Knights Landing SrcID#%d_Ha#%d",
+ pvt->sbridge_dev->source_id, pvt->sbridge_dev->dom);
+ break;
+ }
+
+ if (!mci->ctl_name) {
+ rc = -ENOMEM;
+ goto fail0;
+ }
+
+ /* Get dimm basic config and the memory layout */
+ rc = get_dimm_config(mci);
+ if (rc < 0) {
+ edac_dbg(0, "MC: failed to get_dimm_config()\n");
+ goto fail;
+ }
+ get_memory_layout(mci);
+
+ /* record ptr to the generic device */
+ mci->pdev = &pdev->dev;
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (unlikely(edac_mc_add_mc(mci))) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ rc = -EINVAL;
+ goto fail;
+ }
+
+ return 0;
+
+fail:
+ kfree(mci->ctl_name);
+fail0:
+ edac_mc_free(mci);
+ sbridge_dev->mci = NULL;
+ return rc;
+}
+
+static const struct x86_cpu_id sbridge_cpuids[] = {
+ X86_MATCH_INTEL_FAM6_MODEL(SANDYBRIDGE_X, &pci_dev_descr_sbridge_table),
+ X86_MATCH_INTEL_FAM6_MODEL(IVYBRIDGE_X, &pci_dev_descr_ibridge_table),
+ X86_MATCH_INTEL_FAM6_MODEL(HASWELL_X, &pci_dev_descr_haswell_table),
+ X86_MATCH_INTEL_FAM6_MODEL(BROADWELL_X, &pci_dev_descr_broadwell_table),
+ X86_MATCH_INTEL_FAM6_MODEL(BROADWELL_D, &pci_dev_descr_broadwell_table),
+ X86_MATCH_INTEL_FAM6_MODEL(XEON_PHI_KNL, &pci_dev_descr_knl_table),
+ X86_MATCH_INTEL_FAM6_MODEL(XEON_PHI_KNM, &pci_dev_descr_knl_table),
+ { }
+};
+MODULE_DEVICE_TABLE(x86cpu, sbridge_cpuids);
+
+/*
+ * sbridge_probe Get all devices and register memory controllers
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+
+static int sbridge_probe(const struct x86_cpu_id *id)
+{
+ int rc;
+ u8 mc, num_mc = 0;
+ struct sbridge_dev *sbridge_dev;
+ struct pci_id_table *ptable = (struct pci_id_table *)id->driver_data;
+
+ /* get the pci devices we want to reserve for our use */
+ rc = sbridge_get_all_devices(&num_mc, ptable);
+
+ if (unlikely(rc < 0)) {
+ edac_dbg(0, "couldn't get all devices\n");
+ goto fail0;
+ }
+
+ mc = 0;
+
+ list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
+ edac_dbg(0, "Registering MC#%d (%d of %d)\n",
+ mc, mc + 1, num_mc);
+
+ sbridge_dev->mc = mc++;
+ rc = sbridge_register_mci(sbridge_dev, ptable->type);
+ if (unlikely(rc < 0))
+ goto fail1;
+ }
+
+ sbridge_printk(KERN_INFO, "%s\n", SBRIDGE_REVISION);
+
+ return 0;
+
+fail1:
+ list_for_each_entry(sbridge_dev, &sbridge_edac_list, list)
+ sbridge_unregister_mci(sbridge_dev);
+
+ sbridge_put_all_devices();
+fail0:
+ return rc;
+}
+
+/*
+ * sbridge_remove cleanup
+ *
+ */
+static void sbridge_remove(void)
+{
+ struct sbridge_dev *sbridge_dev;
+
+ edac_dbg(0, "\n");
+
+ list_for_each_entry(sbridge_dev, &sbridge_edac_list, list)
+ sbridge_unregister_mci(sbridge_dev);
+
+ /* Release PCI resources */
+ sbridge_put_all_devices();
+}
+
+/*
+ * sbridge_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init sbridge_init(void)
+{
+ const struct x86_cpu_id *id;
+ const char *owner;
+ int rc;
+
+ edac_dbg(2, "\n");
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
+ return -ENODEV;
+
+ id = x86_match_cpu(sbridge_cpuids);
+ if (!id)
+ return -ENODEV;
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ rc = sbridge_probe(id);
+
+ if (rc >= 0) {
+ mce_register_decode_chain(&sbridge_mce_dec);
+ return 0;
+ }
+
+ sbridge_printk(KERN_ERR, "Failed to register device with error %d.\n",
+ rc);
+
+ return rc;
+}
+
+/*
+ * sbridge_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit sbridge_exit(void)
+{
+ edac_dbg(2, "\n");
+ sbridge_remove();
+ mce_unregister_decode_chain(&sbridge_mce_dec);
+}
+
+module_init(sbridge_init);
+module_exit(sbridge_exit);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab");
+MODULE_AUTHOR("Red Hat Inc. (https://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel Sandy Bridge and Ivy Bridge memory controllers - "
+ SBRIDGE_REVISION);
diff --git a/drivers/edac/sifive_edac.c b/drivers/edac/sifive_edac.c
new file mode 100644
index 0000000000..b844e2626f
--- /dev/null
+++ b/drivers/edac/sifive_edac.c
@@ -0,0 +1,119 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * SiFive Platform EDAC Driver
+ *
+ * Copyright (C) 2018-2022 SiFive, Inc.
+ *
+ * This driver is partially based on octeon_edac-pc.c
+ *
+ */
+#include <linux/edac.h>
+#include <linux/platform_device.h>
+#include "edac_module.h"
+#include <soc/sifive/sifive_ccache.h>
+
+#define DRVNAME "sifive_edac"
+
+struct sifive_edac_priv {
+ struct notifier_block notifier;
+ struct edac_device_ctl_info *dci;
+};
+
+/*
+ * EDAC error callback
+ *
+ * @event: non-zero if unrecoverable.
+ */
+static
+int ecc_err_event(struct notifier_block *this, unsigned long event, void *ptr)
+{
+ const char *msg = (char *)ptr;
+ struct sifive_edac_priv *p;
+
+ p = container_of(this, struct sifive_edac_priv, notifier);
+
+ if (event == SIFIVE_CCACHE_ERR_TYPE_UE)
+ edac_device_handle_ue(p->dci, 0, 0, msg);
+ else if (event == SIFIVE_CCACHE_ERR_TYPE_CE)
+ edac_device_handle_ce(p->dci, 0, 0, msg);
+
+ return NOTIFY_OK;
+}
+
+static int ecc_register(struct platform_device *pdev)
+{
+ struct sifive_edac_priv *p;
+
+ p = devm_kzalloc(&pdev->dev, sizeof(*p), GFP_KERNEL);
+ if (!p)
+ return -ENOMEM;
+
+ p->notifier.notifier_call = ecc_err_event;
+ platform_set_drvdata(pdev, p);
+
+ p->dci = edac_device_alloc_ctl_info(0, "sifive_ecc", 1, "sifive_ecc",
+ 1, 1, NULL, 0,
+ edac_device_alloc_index());
+ if (!p->dci)
+ return -ENOMEM;
+
+ p->dci->dev = &pdev->dev;
+ p->dci->mod_name = "Sifive ECC Manager";
+ p->dci->ctl_name = dev_name(&pdev->dev);
+ p->dci->dev_name = dev_name(&pdev->dev);
+
+ if (edac_device_add_device(p->dci)) {
+ dev_err(p->dci->dev, "failed to register with EDAC core\n");
+ goto err;
+ }
+
+ register_sifive_ccache_error_notifier(&p->notifier);
+
+ return 0;
+
+err:
+ edac_device_free_ctl_info(p->dci);
+
+ return -ENXIO;
+}
+
+static int ecc_unregister(struct platform_device *pdev)
+{
+ struct sifive_edac_priv *p = platform_get_drvdata(pdev);
+
+ unregister_sifive_ccache_error_notifier(&p->notifier);
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(p->dci);
+
+ return 0;
+}
+
+static struct platform_device *sifive_pdev;
+
+static int __init sifive_edac_init(void)
+{
+ int ret;
+
+ sifive_pdev = platform_device_register_simple(DRVNAME, 0, NULL, 0);
+ if (IS_ERR(sifive_pdev))
+ return PTR_ERR(sifive_pdev);
+
+ ret = ecc_register(sifive_pdev);
+ if (ret)
+ platform_device_unregister(sifive_pdev);
+
+ return ret;
+}
+
+static void __exit sifive_edac_exit(void)
+{
+ ecc_unregister(sifive_pdev);
+ platform_device_unregister(sifive_pdev);
+}
+
+module_init(sifive_edac_init);
+module_exit(sifive_edac_exit);
+
+MODULE_AUTHOR("SiFive Inc.");
+MODULE_DESCRIPTION("SiFive platform EDAC driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/skx_base.c b/drivers/edac/skx_base.c
new file mode 100644
index 0000000000..0a862336a7
--- /dev/null
+++ b/drivers/edac/skx_base.c
@@ -0,0 +1,759 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * EDAC driver for Intel(R) Xeon(R) Skylake processors
+ * Copyright (c) 2016, Intel Corporation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/processor.h>
+#include <asm/cpu_device_id.h>
+#include <asm/intel-family.h>
+#include <asm/mce.h>
+
+#include "edac_module.h"
+#include "skx_common.h"
+
+#define EDAC_MOD_STR "skx_edac"
+
+/*
+ * Debug macros
+ */
+#define skx_printk(level, fmt, arg...) \
+ edac_printk(level, "skx", fmt, ##arg)
+
+#define skx_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "skx", fmt, ##arg)
+
+static struct list_head *skx_edac_list;
+
+static u64 skx_tolm, skx_tohm;
+static int skx_num_sockets;
+static unsigned int nvdimm_count;
+
+#define MASK26 0x3FFFFFF /* Mask for 2^26 */
+#define MASK29 0x1FFFFFFF /* Mask for 2^29 */
+
+static struct skx_dev *get_skx_dev(struct pci_bus *bus, u8 idx)
+{
+ struct skx_dev *d;
+
+ list_for_each_entry(d, skx_edac_list, list) {
+ if (d->seg == pci_domain_nr(bus) && d->bus[idx] == bus->number)
+ return d;
+ }
+
+ return NULL;
+}
+
+enum munittype {
+ CHAN0, CHAN1, CHAN2, SAD_ALL, UTIL_ALL, SAD,
+ ERRCHAN0, ERRCHAN1, ERRCHAN2,
+};
+
+struct munit {
+ u16 did;
+ u16 devfn[SKX_NUM_IMC];
+ u8 busidx;
+ u8 per_socket;
+ enum munittype mtype;
+};
+
+/*
+ * List of PCI device ids that we need together with some device
+ * number and function numbers to tell which memory controller the
+ * device belongs to.
+ */
+static const struct munit skx_all_munits[] = {
+ { 0x2054, { }, 1, 1, SAD_ALL },
+ { 0x2055, { }, 1, 1, UTIL_ALL },
+ { 0x2040, { PCI_DEVFN(10, 0), PCI_DEVFN(12, 0) }, 2, 2, CHAN0 },
+ { 0x2044, { PCI_DEVFN(10, 4), PCI_DEVFN(12, 4) }, 2, 2, CHAN1 },
+ { 0x2048, { PCI_DEVFN(11, 0), PCI_DEVFN(13, 0) }, 2, 2, CHAN2 },
+ { 0x2043, { PCI_DEVFN(10, 3), PCI_DEVFN(12, 3) }, 2, 2, ERRCHAN0 },
+ { 0x2047, { PCI_DEVFN(10, 7), PCI_DEVFN(12, 7) }, 2, 2, ERRCHAN1 },
+ { 0x204b, { PCI_DEVFN(11, 3), PCI_DEVFN(13, 3) }, 2, 2, ERRCHAN2 },
+ { 0x208e, { }, 1, 0, SAD },
+ { }
+};
+
+static int get_all_munits(const struct munit *m)
+{
+ struct pci_dev *pdev, *prev;
+ struct skx_dev *d;
+ u32 reg;
+ int i = 0, ndev = 0;
+
+ prev = NULL;
+ for (;;) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, m->did, prev);
+ if (!pdev)
+ break;
+ ndev++;
+ if (m->per_socket == SKX_NUM_IMC) {
+ for (i = 0; i < SKX_NUM_IMC; i++)
+ if (m->devfn[i] == pdev->devfn)
+ break;
+ if (i == SKX_NUM_IMC)
+ goto fail;
+ }
+ d = get_skx_dev(pdev->bus, m->busidx);
+ if (!d)
+ goto fail;
+
+ /* Be sure that the device is enabled */
+ if (unlikely(pci_enable_device(pdev) < 0)) {
+ skx_printk(KERN_ERR, "Couldn't enable device %04x:%04x\n",
+ PCI_VENDOR_ID_INTEL, m->did);
+ goto fail;
+ }
+
+ switch (m->mtype) {
+ case CHAN0:
+ case CHAN1:
+ case CHAN2:
+ pci_dev_get(pdev);
+ d->imc[i].chan[m->mtype].cdev = pdev;
+ break;
+ case ERRCHAN0:
+ case ERRCHAN1:
+ case ERRCHAN2:
+ pci_dev_get(pdev);
+ d->imc[i].chan[m->mtype - ERRCHAN0].edev = pdev;
+ break;
+ case SAD_ALL:
+ pci_dev_get(pdev);
+ d->sad_all = pdev;
+ break;
+ case UTIL_ALL:
+ pci_dev_get(pdev);
+ d->util_all = pdev;
+ break;
+ case SAD:
+ /*
+ * one of these devices per core, including cores
+ * that don't exist on this SKU. Ignore any that
+ * read a route table of zero, make sure all the
+ * non-zero values match.
+ */
+ pci_read_config_dword(pdev, 0xB4, &reg);
+ if (reg != 0) {
+ if (d->mcroute == 0) {
+ d->mcroute = reg;
+ } else if (d->mcroute != reg) {
+ skx_printk(KERN_ERR, "mcroute mismatch\n");
+ goto fail;
+ }
+ }
+ ndev--;
+ break;
+ }
+
+ prev = pdev;
+ }
+
+ return ndev;
+fail:
+ pci_dev_put(pdev);
+ return -ENODEV;
+}
+
+static struct res_config skx_cfg = {
+ .type = SKX,
+ .decs_did = 0x2016,
+ .busno_cfg_offset = 0xcc,
+};
+
+static const struct x86_cpu_id skx_cpuids[] = {
+ X86_MATCH_INTEL_FAM6_MODEL_STEPPINGS(SKYLAKE_X, X86_STEPPINGS(0x0, 0xf), &skx_cfg),
+ { }
+};
+MODULE_DEVICE_TABLE(x86cpu, skx_cpuids);
+
+static bool skx_check_ecc(u32 mcmtr)
+{
+ return !!GET_BITFIELD(mcmtr, 2, 2);
+}
+
+static int skx_get_dimm_config(struct mem_ctl_info *mci, struct res_config *cfg)
+{
+ struct skx_pvt *pvt = mci->pvt_info;
+ u32 mtr, mcmtr, amap, mcddrtcfg;
+ struct skx_imc *imc = pvt->imc;
+ struct dimm_info *dimm;
+ int i, j;
+ int ndimms;
+
+ /* Only the mcmtr on the first channel is effective */
+ pci_read_config_dword(imc->chan[0].cdev, 0x87c, &mcmtr);
+
+ for (i = 0; i < SKX_NUM_CHANNELS; i++) {
+ ndimms = 0;
+ pci_read_config_dword(imc->chan[i].cdev, 0x8C, &amap);
+ pci_read_config_dword(imc->chan[i].cdev, 0x400, &mcddrtcfg);
+ for (j = 0; j < SKX_NUM_DIMMS; j++) {
+ dimm = edac_get_dimm(mci, i, j, 0);
+ pci_read_config_dword(imc->chan[i].cdev,
+ 0x80 + 4 * j, &mtr);
+ if (IS_DIMM_PRESENT(mtr)) {
+ ndimms += skx_get_dimm_info(mtr, mcmtr, amap, dimm, imc, i, j, cfg);
+ } else if (IS_NVDIMM_PRESENT(mcddrtcfg, j)) {
+ ndimms += skx_get_nvdimm_info(dimm, imc, i, j,
+ EDAC_MOD_STR);
+ nvdimm_count++;
+ }
+ }
+ if (ndimms && !skx_check_ecc(mcmtr)) {
+ skx_printk(KERN_ERR, "ECC is disabled on imc %d\n", imc->mc);
+ return -ENODEV;
+ }
+ }
+
+ return 0;
+}
+
+#define SKX_MAX_SAD 24
+
+#define SKX_GET_SAD(d, i, reg) \
+ pci_read_config_dword((d)->sad_all, 0x60 + 8 * (i), &(reg))
+#define SKX_GET_ILV(d, i, reg) \
+ pci_read_config_dword((d)->sad_all, 0x64 + 8 * (i), &(reg))
+
+#define SKX_SAD_MOD3MODE(sad) GET_BITFIELD((sad), 30, 31)
+#define SKX_SAD_MOD3(sad) GET_BITFIELD((sad), 27, 27)
+#define SKX_SAD_LIMIT(sad) (((u64)GET_BITFIELD((sad), 7, 26) << 26) | MASK26)
+#define SKX_SAD_MOD3ASMOD2(sad) GET_BITFIELD((sad), 5, 6)
+#define SKX_SAD_ATTR(sad) GET_BITFIELD((sad), 3, 4)
+#define SKX_SAD_INTERLEAVE(sad) GET_BITFIELD((sad), 1, 2)
+#define SKX_SAD_ENABLE(sad) GET_BITFIELD((sad), 0, 0)
+
+#define SKX_ILV_REMOTE(tgt) (((tgt) & 8) == 0)
+#define SKX_ILV_TARGET(tgt) ((tgt) & 7)
+
+static void skx_show_retry_rd_err_log(struct decoded_addr *res,
+ char *msg, int len,
+ bool scrub_err)
+{
+ u32 log0, log1, log2, log3, log4;
+ u32 corr0, corr1, corr2, corr3;
+ struct pci_dev *edev;
+ int n;
+
+ edev = res->dev->imc[res->imc].chan[res->channel].edev;
+
+ pci_read_config_dword(edev, 0x154, &log0);
+ pci_read_config_dword(edev, 0x148, &log1);
+ pci_read_config_dword(edev, 0x150, &log2);
+ pci_read_config_dword(edev, 0x15c, &log3);
+ pci_read_config_dword(edev, 0x114, &log4);
+
+ n = snprintf(msg, len, " retry_rd_err_log[%.8x %.8x %.8x %.8x %.8x]",
+ log0, log1, log2, log3, log4);
+
+ pci_read_config_dword(edev, 0x104, &corr0);
+ pci_read_config_dword(edev, 0x108, &corr1);
+ pci_read_config_dword(edev, 0x10c, &corr2);
+ pci_read_config_dword(edev, 0x110, &corr3);
+
+ if (len - n > 0)
+ snprintf(msg + n, len - n,
+ " correrrcnt[%.4x %.4x %.4x %.4x %.4x %.4x %.4x %.4x]",
+ corr0 & 0xffff, corr0 >> 16,
+ corr1 & 0xffff, corr1 >> 16,
+ corr2 & 0xffff, corr2 >> 16,
+ corr3 & 0xffff, corr3 >> 16);
+}
+
+static bool skx_sad_decode(struct decoded_addr *res)
+{
+ struct skx_dev *d = list_first_entry(skx_edac_list, typeof(*d), list);
+ u64 addr = res->addr;
+ int i, idx, tgt, lchan, shift;
+ u32 sad, ilv;
+ u64 limit, prev_limit;
+ int remote = 0;
+
+ /* Simple sanity check for I/O space or out of range */
+ if (addr >= skx_tohm || (addr >= skx_tolm && addr < BIT_ULL(32))) {
+ edac_dbg(0, "Address 0x%llx out of range\n", addr);
+ return false;
+ }
+
+restart:
+ prev_limit = 0;
+ for (i = 0; i < SKX_MAX_SAD; i++) {
+ SKX_GET_SAD(d, i, sad);
+ limit = SKX_SAD_LIMIT(sad);
+ if (SKX_SAD_ENABLE(sad)) {
+ if (addr >= prev_limit && addr <= limit)
+ goto sad_found;
+ }
+ prev_limit = limit + 1;
+ }
+ edac_dbg(0, "No SAD entry for 0x%llx\n", addr);
+ return false;
+
+sad_found:
+ SKX_GET_ILV(d, i, ilv);
+
+ switch (SKX_SAD_INTERLEAVE(sad)) {
+ case 0:
+ idx = GET_BITFIELD(addr, 6, 8);
+ break;
+ case 1:
+ idx = GET_BITFIELD(addr, 8, 10);
+ break;
+ case 2:
+ idx = GET_BITFIELD(addr, 12, 14);
+ break;
+ case 3:
+ idx = GET_BITFIELD(addr, 30, 32);
+ break;
+ }
+
+ tgt = GET_BITFIELD(ilv, 4 * idx, 4 * idx + 3);
+
+ /* If point to another node, find it and start over */
+ if (SKX_ILV_REMOTE(tgt)) {
+ if (remote) {
+ edac_dbg(0, "Double remote!\n");
+ return false;
+ }
+ remote = 1;
+ list_for_each_entry(d, skx_edac_list, list) {
+ if (d->imc[0].src_id == SKX_ILV_TARGET(tgt))
+ goto restart;
+ }
+ edac_dbg(0, "Can't find node %d\n", SKX_ILV_TARGET(tgt));
+ return false;
+ }
+
+ if (SKX_SAD_MOD3(sad) == 0) {
+ lchan = SKX_ILV_TARGET(tgt);
+ } else {
+ switch (SKX_SAD_MOD3MODE(sad)) {
+ case 0:
+ shift = 6;
+ break;
+ case 1:
+ shift = 8;
+ break;
+ case 2:
+ shift = 12;
+ break;
+ default:
+ edac_dbg(0, "illegal mod3mode\n");
+ return false;
+ }
+ switch (SKX_SAD_MOD3ASMOD2(sad)) {
+ case 0:
+ lchan = (addr >> shift) % 3;
+ break;
+ case 1:
+ lchan = (addr >> shift) % 2;
+ break;
+ case 2:
+ lchan = (addr >> shift) % 2;
+ lchan = (lchan << 1) | !lchan;
+ break;
+ case 3:
+ lchan = ((addr >> shift) % 2) << 1;
+ break;
+ }
+ lchan = (lchan << 1) | (SKX_ILV_TARGET(tgt) & 1);
+ }
+
+ res->dev = d;
+ res->socket = d->imc[0].src_id;
+ res->imc = GET_BITFIELD(d->mcroute, lchan * 3, lchan * 3 + 2);
+ res->channel = GET_BITFIELD(d->mcroute, lchan * 2 + 18, lchan * 2 + 19);
+
+ edac_dbg(2, "0x%llx: socket=%d imc=%d channel=%d\n",
+ res->addr, res->socket, res->imc, res->channel);
+ return true;
+}
+
+#define SKX_MAX_TAD 8
+
+#define SKX_GET_TADBASE(d, mc, i, reg) \
+ pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x850 + 4 * (i), &(reg))
+#define SKX_GET_TADWAYNESS(d, mc, i, reg) \
+ pci_read_config_dword((d)->imc[mc].chan[0].cdev, 0x880 + 4 * (i), &(reg))
+#define SKX_GET_TADCHNILVOFFSET(d, mc, ch, i, reg) \
+ pci_read_config_dword((d)->imc[mc].chan[ch].cdev, 0x90 + 4 * (i), &(reg))
+
+#define SKX_TAD_BASE(b) ((u64)GET_BITFIELD((b), 12, 31) << 26)
+#define SKX_TAD_SKT_GRAN(b) GET_BITFIELD((b), 4, 5)
+#define SKX_TAD_CHN_GRAN(b) GET_BITFIELD((b), 6, 7)
+#define SKX_TAD_LIMIT(b) (((u64)GET_BITFIELD((b), 12, 31) << 26) | MASK26)
+#define SKX_TAD_OFFSET(b) ((u64)GET_BITFIELD((b), 4, 23) << 26)
+#define SKX_TAD_SKTWAYS(b) (1 << GET_BITFIELD((b), 10, 11))
+#define SKX_TAD_CHNWAYS(b) (GET_BITFIELD((b), 8, 9) + 1)
+
+/* which bit used for both socket and channel interleave */
+static int skx_granularity[] = { 6, 8, 12, 30 };
+
+static u64 skx_do_interleave(u64 addr, int shift, int ways, u64 lowbits)
+{
+ addr >>= shift;
+ addr /= ways;
+ addr <<= shift;
+
+ return addr | (lowbits & ((1ull << shift) - 1));
+}
+
+static bool skx_tad_decode(struct decoded_addr *res)
+{
+ int i;
+ u32 base, wayness, chnilvoffset;
+ int skt_interleave_bit, chn_interleave_bit;
+ u64 channel_addr;
+
+ for (i = 0; i < SKX_MAX_TAD; i++) {
+ SKX_GET_TADBASE(res->dev, res->imc, i, base);
+ SKX_GET_TADWAYNESS(res->dev, res->imc, i, wayness);
+ if (SKX_TAD_BASE(base) <= res->addr && res->addr <= SKX_TAD_LIMIT(wayness))
+ goto tad_found;
+ }
+ edac_dbg(0, "No TAD entry for 0x%llx\n", res->addr);
+ return false;
+
+tad_found:
+ res->sktways = SKX_TAD_SKTWAYS(wayness);
+ res->chanways = SKX_TAD_CHNWAYS(wayness);
+ skt_interleave_bit = skx_granularity[SKX_TAD_SKT_GRAN(base)];
+ chn_interleave_bit = skx_granularity[SKX_TAD_CHN_GRAN(base)];
+
+ SKX_GET_TADCHNILVOFFSET(res->dev, res->imc, res->channel, i, chnilvoffset);
+ channel_addr = res->addr - SKX_TAD_OFFSET(chnilvoffset);
+
+ if (res->chanways == 3 && skt_interleave_bit > chn_interleave_bit) {
+ /* Must handle channel first, then socket */
+ channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit,
+ res->chanways, channel_addr);
+ channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit,
+ res->sktways, channel_addr);
+ } else {
+ /* Handle socket then channel. Preserve low bits from original address */
+ channel_addr = skx_do_interleave(channel_addr, skt_interleave_bit,
+ res->sktways, res->addr);
+ channel_addr = skx_do_interleave(channel_addr, chn_interleave_bit,
+ res->chanways, res->addr);
+ }
+
+ res->chan_addr = channel_addr;
+
+ edac_dbg(2, "0x%llx: chan_addr=0x%llx sktways=%d chanways=%d\n",
+ res->addr, res->chan_addr, res->sktways, res->chanways);
+ return true;
+}
+
+#define SKX_MAX_RIR 4
+
+#define SKX_GET_RIRWAYNESS(d, mc, ch, i, reg) \
+ pci_read_config_dword((d)->imc[mc].chan[ch].cdev, \
+ 0x108 + 4 * (i), &(reg))
+#define SKX_GET_RIRILV(d, mc, ch, idx, i, reg) \
+ pci_read_config_dword((d)->imc[mc].chan[ch].cdev, \
+ 0x120 + 16 * (idx) + 4 * (i), &(reg))
+
+#define SKX_RIR_VALID(b) GET_BITFIELD((b), 31, 31)
+#define SKX_RIR_LIMIT(b) (((u64)GET_BITFIELD((b), 1, 11) << 29) | MASK29)
+#define SKX_RIR_WAYS(b) (1 << GET_BITFIELD((b), 28, 29))
+#define SKX_RIR_CHAN_RANK(b) GET_BITFIELD((b), 16, 19)
+#define SKX_RIR_OFFSET(b) ((u64)(GET_BITFIELD((b), 2, 15) << 26))
+
+static bool skx_rir_decode(struct decoded_addr *res)
+{
+ int i, idx, chan_rank;
+ int shift;
+ u32 rirway, rirlv;
+ u64 rank_addr, prev_limit = 0, limit;
+
+ if (res->dev->imc[res->imc].chan[res->channel].dimms[0].close_pg)
+ shift = 6;
+ else
+ shift = 13;
+
+ for (i = 0; i < SKX_MAX_RIR; i++) {
+ SKX_GET_RIRWAYNESS(res->dev, res->imc, res->channel, i, rirway);
+ limit = SKX_RIR_LIMIT(rirway);
+ if (SKX_RIR_VALID(rirway)) {
+ if (prev_limit <= res->chan_addr &&
+ res->chan_addr <= limit)
+ goto rir_found;
+ }
+ prev_limit = limit;
+ }
+ edac_dbg(0, "No RIR entry for 0x%llx\n", res->addr);
+ return false;
+
+rir_found:
+ rank_addr = res->chan_addr >> shift;
+ rank_addr /= SKX_RIR_WAYS(rirway);
+ rank_addr <<= shift;
+ rank_addr |= res->chan_addr & GENMASK_ULL(shift - 1, 0);
+
+ res->rank_address = rank_addr;
+ idx = (res->chan_addr >> shift) % SKX_RIR_WAYS(rirway);
+
+ SKX_GET_RIRILV(res->dev, res->imc, res->channel, idx, i, rirlv);
+ res->rank_address = rank_addr - SKX_RIR_OFFSET(rirlv);
+ chan_rank = SKX_RIR_CHAN_RANK(rirlv);
+ res->channel_rank = chan_rank;
+ res->dimm = chan_rank / 4;
+ res->rank = chan_rank % 4;
+
+ edac_dbg(2, "0x%llx: dimm=%d rank=%d chan_rank=%d rank_addr=0x%llx\n",
+ res->addr, res->dimm, res->rank,
+ res->channel_rank, res->rank_address);
+ return true;
+}
+
+static u8 skx_close_row[] = {
+ 15, 16, 17, 18, 20, 21, 22, 28, 10, 11, 12, 13, 29, 30, 31, 32, 33, 34
+};
+
+static u8 skx_close_column[] = {
+ 3, 4, 5, 14, 19, 23, 24, 25, 26, 27
+};
+
+static u8 skx_open_row[] = {
+ 14, 15, 16, 20, 28, 21, 22, 23, 24, 25, 26, 27, 29, 30, 31, 32, 33, 34
+};
+
+static u8 skx_open_column[] = {
+ 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
+};
+
+static u8 skx_open_fine_column[] = {
+ 3, 4, 5, 7, 8, 9, 10, 11, 12, 13
+};
+
+static int skx_bits(u64 addr, int nbits, u8 *bits)
+{
+ int i, res = 0;
+
+ for (i = 0; i < nbits; i++)
+ res |= ((addr >> bits[i]) & 1) << i;
+ return res;
+}
+
+static int skx_bank_bits(u64 addr, int b0, int b1, int do_xor, int x0, int x1)
+{
+ int ret = GET_BITFIELD(addr, b0, b0) | (GET_BITFIELD(addr, b1, b1) << 1);
+
+ if (do_xor)
+ ret ^= GET_BITFIELD(addr, x0, x0) | (GET_BITFIELD(addr, x1, x1) << 1);
+
+ return ret;
+}
+
+static bool skx_mad_decode(struct decoded_addr *r)
+{
+ struct skx_dimm *dimm = &r->dev->imc[r->imc].chan[r->channel].dimms[r->dimm];
+ int bg0 = dimm->fine_grain_bank ? 6 : 13;
+
+ if (dimm->close_pg) {
+ r->row = skx_bits(r->rank_address, dimm->rowbits, skx_close_row);
+ r->column = skx_bits(r->rank_address, dimm->colbits, skx_close_column);
+ r->column |= 0x400; /* C10 is autoprecharge, always set */
+ r->bank_address = skx_bank_bits(r->rank_address, 8, 9, dimm->bank_xor_enable, 22, 28);
+ r->bank_group = skx_bank_bits(r->rank_address, 6, 7, dimm->bank_xor_enable, 20, 21);
+ } else {
+ r->row = skx_bits(r->rank_address, dimm->rowbits, skx_open_row);
+ if (dimm->fine_grain_bank)
+ r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_fine_column);
+ else
+ r->column = skx_bits(r->rank_address, dimm->colbits, skx_open_column);
+ r->bank_address = skx_bank_bits(r->rank_address, 18, 19, dimm->bank_xor_enable, 22, 23);
+ r->bank_group = skx_bank_bits(r->rank_address, bg0, 17, dimm->bank_xor_enable, 20, 21);
+ }
+ r->row &= (1u << dimm->rowbits) - 1;
+
+ edac_dbg(2, "0x%llx: row=0x%x col=0x%x bank_addr=%d bank_group=%d\n",
+ r->addr, r->row, r->column, r->bank_address,
+ r->bank_group);
+ return true;
+}
+
+static bool skx_decode(struct decoded_addr *res)
+{
+ return skx_sad_decode(res) && skx_tad_decode(res) &&
+ skx_rir_decode(res) && skx_mad_decode(res);
+}
+
+static struct notifier_block skx_mce_dec = {
+ .notifier_call = skx_mce_check_error,
+ .priority = MCE_PRIO_EDAC,
+};
+
+#ifdef CONFIG_EDAC_DEBUG
+/*
+ * Debug feature.
+ * Exercise the address decode logic by writing an address to
+ * /sys/kernel/debug/edac/skx_test/addr.
+ */
+static struct dentry *skx_test;
+
+static int debugfs_u64_set(void *data, u64 val)
+{
+ struct mce m;
+
+ pr_warn_once("Fake error to 0x%llx injected via debugfs\n", val);
+
+ memset(&m, 0, sizeof(m));
+ /* ADDRV + MemRd + Unknown channel */
+ m.status = MCI_STATUS_ADDRV + 0x90;
+ /* One corrected error */
+ m.status |= BIT_ULL(MCI_STATUS_CEC_SHIFT);
+ m.addr = val;
+ skx_mce_check_error(NULL, 0, &m);
+
+ return 0;
+}
+DEFINE_SIMPLE_ATTRIBUTE(fops_u64_wo, NULL, debugfs_u64_set, "%llu\n");
+
+static void setup_skx_debug(void)
+{
+ skx_test = edac_debugfs_create_dir("skx_test");
+ if (!skx_test)
+ return;
+
+ if (!edac_debugfs_create_file("addr", 0200, skx_test,
+ NULL, &fops_u64_wo)) {
+ debugfs_remove(skx_test);
+ skx_test = NULL;
+ }
+}
+
+static void teardown_skx_debug(void)
+{
+ debugfs_remove_recursive(skx_test);
+}
+#else
+static inline void setup_skx_debug(void) {}
+static inline void teardown_skx_debug(void) {}
+#endif /*CONFIG_EDAC_DEBUG*/
+
+/*
+ * skx_init:
+ * make sure we are running on the correct cpu model
+ * search for all the devices we need
+ * check which DIMMs are present.
+ */
+static int __init skx_init(void)
+{
+ const struct x86_cpu_id *id;
+ struct res_config *cfg;
+ const struct munit *m;
+ const char *owner;
+ int rc = 0, i, off[3] = {0xd0, 0xd4, 0xd8};
+ u8 mc = 0, src_id, node_id;
+ struct skx_dev *d;
+
+ edac_dbg(2, "\n");
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ owner = edac_get_owner();
+ if (owner && strncmp(owner, EDAC_MOD_STR, sizeof(EDAC_MOD_STR)))
+ return -EBUSY;
+
+ if (cpu_feature_enabled(X86_FEATURE_HYPERVISOR))
+ return -ENODEV;
+
+ id = x86_match_cpu(skx_cpuids);
+ if (!id)
+ return -ENODEV;
+
+ cfg = (struct res_config *)id->driver_data;
+
+ rc = skx_get_hi_lo(0x2034, off, &skx_tolm, &skx_tohm);
+ if (rc)
+ return rc;
+
+ rc = skx_get_all_bus_mappings(cfg, &skx_edac_list);
+ if (rc < 0)
+ goto fail;
+ if (rc == 0) {
+ edac_dbg(2, "No memory controllers found\n");
+ return -ENODEV;
+ }
+ skx_num_sockets = rc;
+
+ for (m = skx_all_munits; m->did; m++) {
+ rc = get_all_munits(m);
+ if (rc < 0)
+ goto fail;
+ if (rc != m->per_socket * skx_num_sockets) {
+ edac_dbg(2, "Expected %d, got %d of 0x%x\n",
+ m->per_socket * skx_num_sockets, rc, m->did);
+ rc = -ENODEV;
+ goto fail;
+ }
+ }
+
+ list_for_each_entry(d, skx_edac_list, list) {
+ rc = skx_get_src_id(d, 0xf0, &src_id);
+ if (rc < 0)
+ goto fail;
+ rc = skx_get_node_id(d, &node_id);
+ if (rc < 0)
+ goto fail;
+ edac_dbg(2, "src_id=%d node_id=%d\n", src_id, node_id);
+ for (i = 0; i < SKX_NUM_IMC; i++) {
+ d->imc[i].mc = mc++;
+ d->imc[i].lmc = i;
+ d->imc[i].src_id = src_id;
+ d->imc[i].node_id = node_id;
+ rc = skx_register_mci(&d->imc[i], d->imc[i].chan[0].cdev,
+ "Skylake Socket", EDAC_MOD_STR,
+ skx_get_dimm_config, cfg);
+ if (rc < 0)
+ goto fail;
+ }
+ }
+
+ skx_set_decode(skx_decode, skx_show_retry_rd_err_log);
+
+ if (nvdimm_count && skx_adxl_get() != -ENODEV) {
+ skx_set_decode(NULL, skx_show_retry_rd_err_log);
+ } else {
+ if (nvdimm_count)
+ skx_printk(KERN_NOTICE, "Only decoding DDR4 address!\n");
+ skx_set_decode(skx_decode, skx_show_retry_rd_err_log);
+ }
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ setup_skx_debug();
+
+ mce_register_decode_chain(&skx_mce_dec);
+
+ return 0;
+fail:
+ skx_remove();
+ return rc;
+}
+
+static void __exit skx_exit(void)
+{
+ edac_dbg(2, "\n");
+ mce_unregister_decode_chain(&skx_mce_dec);
+ teardown_skx_debug();
+ if (nvdimm_count)
+ skx_adxl_put();
+ skx_remove();
+}
+
+module_init(skx_init);
+module_exit(skx_exit);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Tony Luck");
+MODULE_DESCRIPTION("MC Driver for Intel Skylake server processors");
diff --git a/drivers/edac/skx_common.c b/drivers/edac/skx_common.c
new file mode 100644
index 0000000000..ce3e0069e0
--- /dev/null
+++ b/drivers/edac/skx_common.c
@@ -0,0 +1,723 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Shared code by both skx_edac and i10nm_edac. Originally split out
+ * from the skx_edac driver.
+ *
+ * This file is linked into both skx_edac and i10nm_edac drivers. In
+ * order to avoid link errors, this file must be like a pure library
+ * without including symbols and defines which would otherwise conflict,
+ * when linked once into a module and into a built-in object, at the
+ * same time. For example, __this_module symbol references when that
+ * file is being linked into a built-in object.
+ *
+ * Copyright (c) 2018, Intel Corporation.
+ */
+
+#include <linux/acpi.h>
+#include <linux/dmi.h>
+#include <linux/adxl.h>
+#include <acpi/nfit.h>
+#include <asm/mce.h>
+#include "edac_module.h"
+#include "skx_common.h"
+
+static const char * const component_names[] = {
+ [INDEX_SOCKET] = "ProcessorSocketId",
+ [INDEX_MEMCTRL] = "MemoryControllerId",
+ [INDEX_CHANNEL] = "ChannelId",
+ [INDEX_DIMM] = "DimmSlotId",
+ [INDEX_CS] = "ChipSelect",
+ [INDEX_NM_MEMCTRL] = "NmMemoryControllerId",
+ [INDEX_NM_CHANNEL] = "NmChannelId",
+ [INDEX_NM_DIMM] = "NmDimmSlotId",
+ [INDEX_NM_CS] = "NmChipSelect",
+};
+
+static int component_indices[ARRAY_SIZE(component_names)];
+static int adxl_component_count;
+static const char * const *adxl_component_names;
+static u64 *adxl_values;
+static char *adxl_msg;
+static unsigned long adxl_nm_bitmap;
+
+static char skx_msg[MSG_SIZE];
+static skx_decode_f driver_decode;
+static skx_show_retry_log_f skx_show_retry_rd_err_log;
+static u64 skx_tolm, skx_tohm;
+static LIST_HEAD(dev_edac_list);
+static bool skx_mem_cfg_2lm;
+
+int __init skx_adxl_get(void)
+{
+ const char * const *names;
+ int i, j;
+
+ names = adxl_get_component_names();
+ if (!names) {
+ skx_printk(KERN_NOTICE, "No firmware support for address translation.\n");
+ return -ENODEV;
+ }
+
+ for (i = 0; i < INDEX_MAX; i++) {
+ for (j = 0; names[j]; j++) {
+ if (!strcmp(component_names[i], names[j])) {
+ component_indices[i] = j;
+
+ if (i >= INDEX_NM_FIRST)
+ adxl_nm_bitmap |= 1 << i;
+
+ break;
+ }
+ }
+
+ if (!names[j] && i < INDEX_NM_FIRST)
+ goto err;
+ }
+
+ if (skx_mem_cfg_2lm) {
+ if (!adxl_nm_bitmap)
+ skx_printk(KERN_NOTICE, "Not enough ADXL components for 2-level memory.\n");
+ else
+ edac_dbg(2, "adxl_nm_bitmap: 0x%lx\n", adxl_nm_bitmap);
+ }
+
+ adxl_component_names = names;
+ while (*names++)
+ adxl_component_count++;
+
+ adxl_values = kcalloc(adxl_component_count, sizeof(*adxl_values),
+ GFP_KERNEL);
+ if (!adxl_values) {
+ adxl_component_count = 0;
+ return -ENOMEM;
+ }
+
+ adxl_msg = kzalloc(MSG_SIZE, GFP_KERNEL);
+ if (!adxl_msg) {
+ adxl_component_count = 0;
+ kfree(adxl_values);
+ return -ENOMEM;
+ }
+
+ return 0;
+err:
+ skx_printk(KERN_ERR, "'%s' is not matched from DSM parameters: ",
+ component_names[i]);
+ for (j = 0; names[j]; j++)
+ skx_printk(KERN_CONT, "%s ", names[j]);
+ skx_printk(KERN_CONT, "\n");
+
+ return -ENODEV;
+}
+
+void __exit skx_adxl_put(void)
+{
+ kfree(adxl_values);
+ kfree(adxl_msg);
+}
+
+static bool skx_adxl_decode(struct decoded_addr *res, bool error_in_1st_level_mem)
+{
+ struct skx_dev *d;
+ int i, len = 0;
+
+ if (res->addr >= skx_tohm || (res->addr >= skx_tolm &&
+ res->addr < BIT_ULL(32))) {
+ edac_dbg(0, "Address 0x%llx out of range\n", res->addr);
+ return false;
+ }
+
+ if (adxl_decode(res->addr, adxl_values)) {
+ edac_dbg(0, "Failed to decode 0x%llx\n", res->addr);
+ return false;
+ }
+
+ res->socket = (int)adxl_values[component_indices[INDEX_SOCKET]];
+ if (error_in_1st_level_mem) {
+ res->imc = (adxl_nm_bitmap & BIT_NM_MEMCTRL) ?
+ (int)adxl_values[component_indices[INDEX_NM_MEMCTRL]] : -1;
+ res->channel = (adxl_nm_bitmap & BIT_NM_CHANNEL) ?
+ (int)adxl_values[component_indices[INDEX_NM_CHANNEL]] : -1;
+ res->dimm = (adxl_nm_bitmap & BIT_NM_DIMM) ?
+ (int)adxl_values[component_indices[INDEX_NM_DIMM]] : -1;
+ res->cs = (adxl_nm_bitmap & BIT_NM_CS) ?
+ (int)adxl_values[component_indices[INDEX_NM_CS]] : -1;
+ } else {
+ res->imc = (int)adxl_values[component_indices[INDEX_MEMCTRL]];
+ res->channel = (int)adxl_values[component_indices[INDEX_CHANNEL]];
+ res->dimm = (int)adxl_values[component_indices[INDEX_DIMM]];
+ res->cs = (int)adxl_values[component_indices[INDEX_CS]];
+ }
+
+ if (res->imc > NUM_IMC - 1 || res->imc < 0) {
+ skx_printk(KERN_ERR, "Bad imc %d\n", res->imc);
+ return false;
+ }
+
+ list_for_each_entry(d, &dev_edac_list, list) {
+ if (d->imc[0].src_id == res->socket) {
+ res->dev = d;
+ break;
+ }
+ }
+
+ if (!res->dev) {
+ skx_printk(KERN_ERR, "No device for src_id %d imc %d\n",
+ res->socket, res->imc);
+ return false;
+ }
+
+ for (i = 0; i < adxl_component_count; i++) {
+ if (adxl_values[i] == ~0x0ull)
+ continue;
+
+ len += snprintf(adxl_msg + len, MSG_SIZE - len, " %s:0x%llx",
+ adxl_component_names[i], adxl_values[i]);
+ if (MSG_SIZE - len <= 0)
+ break;
+ }
+
+ res->decoded_by_adxl = true;
+
+ return true;
+}
+
+void skx_set_mem_cfg(bool mem_cfg_2lm)
+{
+ skx_mem_cfg_2lm = mem_cfg_2lm;
+}
+
+void skx_set_decode(skx_decode_f decode, skx_show_retry_log_f show_retry_log)
+{
+ driver_decode = decode;
+ skx_show_retry_rd_err_log = show_retry_log;
+}
+
+int skx_get_src_id(struct skx_dev *d, int off, u8 *id)
+{
+ u32 reg;
+
+ if (pci_read_config_dword(d->util_all, off, &reg)) {
+ skx_printk(KERN_ERR, "Failed to read src id\n");
+ return -ENODEV;
+ }
+
+ *id = GET_BITFIELD(reg, 12, 14);
+ return 0;
+}
+
+int skx_get_node_id(struct skx_dev *d, u8 *id)
+{
+ u32 reg;
+
+ if (pci_read_config_dword(d->util_all, 0xf4, &reg)) {
+ skx_printk(KERN_ERR, "Failed to read node id\n");
+ return -ENODEV;
+ }
+
+ *id = GET_BITFIELD(reg, 0, 2);
+ return 0;
+}
+
+static int get_width(u32 mtr)
+{
+ switch (GET_BITFIELD(mtr, 8, 9)) {
+ case 0:
+ return DEV_X4;
+ case 1:
+ return DEV_X8;
+ case 2:
+ return DEV_X16;
+ }
+ return DEV_UNKNOWN;
+}
+
+/*
+ * We use the per-socket device @cfg->did to count how many sockets are present,
+ * and to detemine which PCI buses are associated with each socket. Allocate
+ * and build the full list of all the skx_dev structures that we need here.
+ */
+int skx_get_all_bus_mappings(struct res_config *cfg, struct list_head **list)
+{
+ struct pci_dev *pdev, *prev;
+ struct skx_dev *d;
+ u32 reg;
+ int ndev = 0;
+
+ prev = NULL;
+ for (;;) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, cfg->decs_did, prev);
+ if (!pdev)
+ break;
+ ndev++;
+ d = kzalloc(sizeof(*d), GFP_KERNEL);
+ if (!d) {
+ pci_dev_put(pdev);
+ return -ENOMEM;
+ }
+
+ if (pci_read_config_dword(pdev, cfg->busno_cfg_offset, &reg)) {
+ kfree(d);
+ pci_dev_put(pdev);
+ skx_printk(KERN_ERR, "Failed to read bus idx\n");
+ return -ENODEV;
+ }
+
+ d->bus[0] = GET_BITFIELD(reg, 0, 7);
+ d->bus[1] = GET_BITFIELD(reg, 8, 15);
+ if (cfg->type == SKX) {
+ d->seg = pci_domain_nr(pdev->bus);
+ d->bus[2] = GET_BITFIELD(reg, 16, 23);
+ d->bus[3] = GET_BITFIELD(reg, 24, 31);
+ } else {
+ d->seg = GET_BITFIELD(reg, 16, 23);
+ }
+
+ edac_dbg(2, "busses: 0x%x, 0x%x, 0x%x, 0x%x\n",
+ d->bus[0], d->bus[1], d->bus[2], d->bus[3]);
+ list_add_tail(&d->list, &dev_edac_list);
+ prev = pdev;
+ }
+
+ if (list)
+ *list = &dev_edac_list;
+ return ndev;
+}
+
+int skx_get_hi_lo(unsigned int did, int off[], u64 *tolm, u64 *tohm)
+{
+ struct pci_dev *pdev;
+ u32 reg;
+
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL, did, NULL);
+ if (!pdev) {
+ edac_dbg(2, "Can't get tolm/tohm\n");
+ return -ENODEV;
+ }
+
+ if (pci_read_config_dword(pdev, off[0], &reg)) {
+ skx_printk(KERN_ERR, "Failed to read tolm\n");
+ goto fail;
+ }
+ skx_tolm = reg;
+
+ if (pci_read_config_dword(pdev, off[1], &reg)) {
+ skx_printk(KERN_ERR, "Failed to read lower tohm\n");
+ goto fail;
+ }
+ skx_tohm = reg;
+
+ if (pci_read_config_dword(pdev, off[2], &reg)) {
+ skx_printk(KERN_ERR, "Failed to read upper tohm\n");
+ goto fail;
+ }
+ skx_tohm |= (u64)reg << 32;
+
+ pci_dev_put(pdev);
+ *tolm = skx_tolm;
+ *tohm = skx_tohm;
+ edac_dbg(2, "tolm = 0x%llx tohm = 0x%llx\n", skx_tolm, skx_tohm);
+ return 0;
+fail:
+ pci_dev_put(pdev);
+ return -ENODEV;
+}
+
+static int skx_get_dimm_attr(u32 reg, int lobit, int hibit, int add,
+ int minval, int maxval, const char *name)
+{
+ u32 val = GET_BITFIELD(reg, lobit, hibit);
+
+ if (val < minval || val > maxval) {
+ edac_dbg(2, "bad %s = %d (raw=0x%x)\n", name, val, reg);
+ return -EINVAL;
+ }
+ return val + add;
+}
+
+#define numrank(reg) skx_get_dimm_attr(reg, 12, 13, 0, 0, 2, "ranks")
+#define numrow(reg) skx_get_dimm_attr(reg, 2, 4, 12, 1, 6, "rows")
+#define numcol(reg) skx_get_dimm_attr(reg, 0, 1, 10, 0, 2, "cols")
+
+int skx_get_dimm_info(u32 mtr, u32 mcmtr, u32 amap, struct dimm_info *dimm,
+ struct skx_imc *imc, int chan, int dimmno,
+ struct res_config *cfg)
+{
+ int banks, ranks, rows, cols, npages;
+ enum mem_type mtype;
+ u64 size;
+
+ ranks = numrank(mtr);
+ rows = numrow(mtr);
+ cols = imc->hbm_mc ? 6 : numcol(mtr);
+
+ if (imc->hbm_mc) {
+ banks = 32;
+ mtype = MEM_HBM2;
+ } else if (cfg->support_ddr5 && (amap & 0x8)) {
+ banks = 32;
+ mtype = MEM_DDR5;
+ } else {
+ banks = 16;
+ mtype = MEM_DDR4;
+ }
+
+ /*
+ * Compute size in 8-byte (2^3) words, then shift to MiB (2^20)
+ */
+ size = ((1ull << (rows + cols + ranks)) * banks) >> (20 - 3);
+ npages = MiB_TO_PAGES(size);
+
+ edac_dbg(0, "mc#%d: channel %d, dimm %d, %lld MiB (%d pages) bank: %d, rank: %d, row: 0x%x, col: 0x%x\n",
+ imc->mc, chan, dimmno, size, npages,
+ banks, 1 << ranks, rows, cols);
+
+ imc->chan[chan].dimms[dimmno].close_pg = GET_BITFIELD(mcmtr, 0, 0);
+ imc->chan[chan].dimms[dimmno].bank_xor_enable = GET_BITFIELD(mcmtr, 9, 9);
+ imc->chan[chan].dimms[dimmno].fine_grain_bank = GET_BITFIELD(amap, 0, 0);
+ imc->chan[chan].dimms[dimmno].rowbits = rows;
+ imc->chan[chan].dimms[dimmno].colbits = cols;
+
+ dimm->nr_pages = npages;
+ dimm->grain = 32;
+ dimm->dtype = get_width(mtr);
+ dimm->mtype = mtype;
+ dimm->edac_mode = EDAC_SECDED; /* likely better than this */
+
+ if (imc->hbm_mc)
+ snprintf(dimm->label, sizeof(dimm->label), "CPU_SrcID#%u_HBMC#%u_Chan#%u",
+ imc->src_id, imc->lmc, chan);
+ else
+ snprintf(dimm->label, sizeof(dimm->label), "CPU_SrcID#%u_MC#%u_Chan#%u_DIMM#%u",
+ imc->src_id, imc->lmc, chan, dimmno);
+
+ return 1;
+}
+
+int skx_get_nvdimm_info(struct dimm_info *dimm, struct skx_imc *imc,
+ int chan, int dimmno, const char *mod_str)
+{
+ int smbios_handle;
+ u32 dev_handle;
+ u16 flags;
+ u64 size = 0;
+
+ dev_handle = ACPI_NFIT_BUILD_DEVICE_HANDLE(dimmno, chan, imc->lmc,
+ imc->src_id, 0);
+
+ smbios_handle = nfit_get_smbios_id(dev_handle, &flags);
+ if (smbios_handle == -EOPNOTSUPP) {
+ pr_warn_once("%s: Can't find size of NVDIMM. Try enabling CONFIG_ACPI_NFIT\n", mod_str);
+ goto unknown_size;
+ }
+
+ if (smbios_handle < 0) {
+ skx_printk(KERN_ERR, "Can't find handle for NVDIMM ADR=0x%x\n", dev_handle);
+ goto unknown_size;
+ }
+
+ if (flags & ACPI_NFIT_MEM_MAP_FAILED) {
+ skx_printk(KERN_ERR, "NVDIMM ADR=0x%x is not mapped\n", dev_handle);
+ goto unknown_size;
+ }
+
+ size = dmi_memdev_size(smbios_handle);
+ if (size == ~0ull)
+ skx_printk(KERN_ERR, "Can't find size for NVDIMM ADR=0x%x/SMBIOS=0x%x\n",
+ dev_handle, smbios_handle);
+
+unknown_size:
+ dimm->nr_pages = size >> PAGE_SHIFT;
+ dimm->grain = 32;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->mtype = MEM_NVDIMM;
+ dimm->edac_mode = EDAC_SECDED; /* likely better than this */
+
+ edac_dbg(0, "mc#%d: channel %d, dimm %d, %llu MiB (%u pages)\n",
+ imc->mc, chan, dimmno, size >> 20, dimm->nr_pages);
+
+ snprintf(dimm->label, sizeof(dimm->label), "CPU_SrcID#%u_MC#%u_Chan#%u_DIMM#%u",
+ imc->src_id, imc->lmc, chan, dimmno);
+
+ return (size == 0 || size == ~0ull) ? 0 : 1;
+}
+
+int skx_register_mci(struct skx_imc *imc, struct pci_dev *pdev,
+ const char *ctl_name, const char *mod_str,
+ get_dimm_config_f get_dimm_config,
+ struct res_config *cfg)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct skx_pvt *pvt;
+ int rc;
+
+ /* Allocate a new MC control structure */
+ layers[0].type = EDAC_MC_LAYER_CHANNEL;
+ layers[0].size = NUM_CHANNELS;
+ layers[0].is_virt_csrow = false;
+ layers[1].type = EDAC_MC_LAYER_SLOT;
+ layers[1].size = NUM_DIMMS;
+ layers[1].is_virt_csrow = true;
+ mci = edac_mc_alloc(imc->mc, ARRAY_SIZE(layers), layers,
+ sizeof(struct skx_pvt));
+
+ if (unlikely(!mci))
+ return -ENOMEM;
+
+ edac_dbg(0, "MC#%d: mci = %p\n", imc->mc, mci);
+
+ /* Associate skx_dev and mci for future usage */
+ imc->mci = mci;
+ pvt = mci->pvt_info;
+ pvt->imc = imc;
+
+ mci->ctl_name = kasprintf(GFP_KERNEL, "%s#%d IMC#%d", ctl_name,
+ imc->node_id, imc->lmc);
+ if (!mci->ctl_name) {
+ rc = -ENOMEM;
+ goto fail0;
+ }
+
+ mci->mtype_cap = MEM_FLAG_DDR4 | MEM_FLAG_NVDIMM;
+ if (cfg->support_ddr5)
+ mci->mtype_cap |= MEM_FLAG_DDR5;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = mod_str;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ rc = get_dimm_config(mci, cfg);
+ if (rc < 0)
+ goto fail;
+
+ /* Record ptr to the generic device */
+ mci->pdev = &pdev->dev;
+
+ /* Add this new MC control structure to EDAC's list of MCs */
+ if (unlikely(edac_mc_add_mc(mci))) {
+ edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+ rc = -EINVAL;
+ goto fail;
+ }
+
+ return 0;
+
+fail:
+ kfree(mci->ctl_name);
+fail0:
+ edac_mc_free(mci);
+ imc->mci = NULL;
+ return rc;
+}
+
+static void skx_unregister_mci(struct skx_imc *imc)
+{
+ struct mem_ctl_info *mci = imc->mci;
+
+ if (!mci)
+ return;
+
+ edac_dbg(0, "MC%d: mci = %p\n", imc->mc, mci);
+
+ /* Remove MC sysfs nodes */
+ edac_mc_del_mc(mci->pdev);
+
+ edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
+ kfree(mci->ctl_name);
+ edac_mc_free(mci);
+}
+
+static void skx_mce_output_error(struct mem_ctl_info *mci,
+ const struct mce *m,
+ struct decoded_addr *res)
+{
+ enum hw_event_mc_err_type tp_event;
+ char *optype;
+ bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
+ bool overflow = GET_BITFIELD(m->status, 62, 62);
+ bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
+ bool scrub_err = false;
+ bool recoverable;
+ int len;
+ u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52);
+ u32 mscod = GET_BITFIELD(m->status, 16, 31);
+ u32 errcode = GET_BITFIELD(m->status, 0, 15);
+ u32 optypenum = GET_BITFIELD(m->status, 4, 6);
+
+ recoverable = GET_BITFIELD(m->status, 56, 56);
+
+ if (uncorrected_error) {
+ core_err_cnt = 1;
+ if (ripv) {
+ tp_event = HW_EVENT_ERR_UNCORRECTED;
+ } else {
+ tp_event = HW_EVENT_ERR_FATAL;
+ }
+ } else {
+ tp_event = HW_EVENT_ERR_CORRECTED;
+ }
+
+ switch (optypenum) {
+ case 0:
+ optype = "generic undef request error";
+ break;
+ case 1:
+ optype = "memory read error";
+ break;
+ case 2:
+ optype = "memory write error";
+ break;
+ case 3:
+ optype = "addr/cmd error";
+ break;
+ case 4:
+ optype = "memory scrubbing error";
+ scrub_err = true;
+ break;
+ default:
+ optype = "reserved";
+ break;
+ }
+
+ if (res->decoded_by_adxl) {
+ len = snprintf(skx_msg, MSG_SIZE, "%s%s err_code:0x%04x:0x%04x %s",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable) ? " recoverable" : "",
+ mscod, errcode, adxl_msg);
+ } else {
+ len = snprintf(skx_msg, MSG_SIZE,
+ "%s%s err_code:0x%04x:0x%04x ProcessorSocketId:0x%x MemoryControllerId:0x%x PhysicalRankId:0x%x Row:0x%x Column:0x%x Bank:0x%x BankGroup:0x%x",
+ overflow ? " OVERFLOW" : "",
+ (uncorrected_error && recoverable) ? " recoverable" : "",
+ mscod, errcode,
+ res->socket, res->imc, res->rank,
+ res->row, res->column, res->bank_address, res->bank_group);
+ }
+
+ if (skx_show_retry_rd_err_log)
+ skx_show_retry_rd_err_log(res, skx_msg + len, MSG_SIZE - len, scrub_err);
+
+ edac_dbg(0, "%s\n", skx_msg);
+
+ /* Call the helper to output message */
+ edac_mc_handle_error(tp_event, mci, core_err_cnt,
+ m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
+ res->channel, res->dimm, -1,
+ optype, skx_msg);
+}
+
+static bool skx_error_in_1st_level_mem(const struct mce *m)
+{
+ u32 errcode;
+
+ if (!skx_mem_cfg_2lm)
+ return false;
+
+ errcode = GET_BITFIELD(m->status, 0, 15) & MCACOD_MEM_ERR_MASK;
+
+ return errcode == MCACOD_EXT_MEM_ERR;
+}
+
+static bool skx_error_in_mem(const struct mce *m)
+{
+ u32 errcode;
+
+ errcode = GET_BITFIELD(m->status, 0, 15) & MCACOD_MEM_ERR_MASK;
+
+ return (errcode == MCACOD_MEM_CTL_ERR || errcode == MCACOD_EXT_MEM_ERR);
+}
+
+int skx_mce_check_error(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct mce *mce = (struct mce *)data;
+ struct decoded_addr res;
+ struct mem_ctl_info *mci;
+ char *type;
+
+ if (mce->kflags & MCE_HANDLED_CEC)
+ return NOTIFY_DONE;
+
+ /* Ignore unless this is memory related with an address */
+ if (!skx_error_in_mem(mce) || !(mce->status & MCI_STATUS_ADDRV))
+ return NOTIFY_DONE;
+
+ memset(&res, 0, sizeof(res));
+ res.mce = mce;
+ res.addr = mce->addr & MCI_ADDR_PHYSADDR;
+
+ /* Try driver decoder first */
+ if (!(driver_decode && driver_decode(&res))) {
+ /* Then try firmware decoder (ACPI DSM methods) */
+ if (!(adxl_component_count && skx_adxl_decode(&res, skx_error_in_1st_level_mem(mce))))
+ return NOTIFY_DONE;
+ }
+
+ mci = res.dev->imc[res.imc].mci;
+
+ if (!mci)
+ return NOTIFY_DONE;
+
+ if (mce->mcgstatus & MCG_STATUS_MCIP)
+ type = "Exception";
+ else
+ type = "Event";
+
+ skx_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
+
+ skx_mc_printk(mci, KERN_DEBUG, "CPU %d: Machine Check %s: 0x%llx "
+ "Bank %d: 0x%llx\n", mce->extcpu, type,
+ mce->mcgstatus, mce->bank, mce->status);
+ skx_mc_printk(mci, KERN_DEBUG, "TSC 0x%llx ", mce->tsc);
+ skx_mc_printk(mci, KERN_DEBUG, "ADDR 0x%llx ", mce->addr);
+ skx_mc_printk(mci, KERN_DEBUG, "MISC 0x%llx ", mce->misc);
+
+ skx_mc_printk(mci, KERN_DEBUG, "PROCESSOR %u:0x%x TIME %llu SOCKET "
+ "%u APIC 0x%x\n", mce->cpuvendor, mce->cpuid,
+ mce->time, mce->socketid, mce->apicid);
+
+ skx_mce_output_error(mci, mce, &res);
+
+ mce->kflags |= MCE_HANDLED_EDAC;
+ return NOTIFY_DONE;
+}
+
+void skx_remove(void)
+{
+ int i, j;
+ struct skx_dev *d, *tmp;
+
+ edac_dbg(0, "\n");
+
+ list_for_each_entry_safe(d, tmp, &dev_edac_list, list) {
+ list_del(&d->list);
+ for (i = 0; i < NUM_IMC; i++) {
+ if (d->imc[i].mci)
+ skx_unregister_mci(&d->imc[i]);
+
+ if (d->imc[i].mdev)
+ pci_dev_put(d->imc[i].mdev);
+
+ if (d->imc[i].mbase)
+ iounmap(d->imc[i].mbase);
+
+ for (j = 0; j < NUM_CHANNELS; j++) {
+ if (d->imc[i].chan[j].cdev)
+ pci_dev_put(d->imc[i].chan[j].cdev);
+ }
+ }
+ if (d->util_all)
+ pci_dev_put(d->util_all);
+ if (d->pcu_cr3)
+ pci_dev_put(d->pcu_cr3);
+ if (d->sad_all)
+ pci_dev_put(d->sad_all);
+ if (d->uracu)
+ pci_dev_put(d->uracu);
+
+ kfree(d);
+ }
+}
diff --git a/drivers/edac/skx_common.h b/drivers/edac/skx_common.h
new file mode 100644
index 0000000000..b6d3607dff
--- /dev/null
+++ b/drivers/edac/skx_common.h
@@ -0,0 +1,263 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Common codes for both the skx_edac driver and Intel 10nm server EDAC driver.
+ * Originally split out from the skx_edac driver.
+ *
+ * Copyright (c) 2018, Intel Corporation.
+ */
+
+#ifndef _SKX_COMM_EDAC_H
+#define _SKX_COMM_EDAC_H
+
+#include <linux/bits.h>
+#include <asm/mce.h>
+
+#define MSG_SIZE 1024
+
+/*
+ * Debug macros
+ */
+#define skx_printk(level, fmt, arg...) \
+ edac_printk(level, "skx", fmt, ##arg)
+
+#define skx_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "skx", fmt, ##arg)
+
+/*
+ * Get a bit field at register value <v>, from bit <lo> to bit <hi>
+ */
+#define GET_BITFIELD(v, lo, hi) \
+ (((v) & GENMASK_ULL((hi), (lo))) >> (lo))
+
+#define SKX_NUM_IMC 2 /* Memory controllers per socket */
+#define SKX_NUM_CHANNELS 3 /* Channels per memory controller */
+#define SKX_NUM_DIMMS 2 /* Max DIMMS per channel */
+
+#define I10NM_NUM_DDR_IMC 12
+#define I10NM_NUM_DDR_CHANNELS 2
+#define I10NM_NUM_DDR_DIMMS 2
+
+#define I10NM_NUM_HBM_IMC 16
+#define I10NM_NUM_HBM_CHANNELS 2
+#define I10NM_NUM_HBM_DIMMS 1
+
+#define I10NM_NUM_IMC (I10NM_NUM_DDR_IMC + I10NM_NUM_HBM_IMC)
+#define I10NM_NUM_CHANNELS MAX(I10NM_NUM_DDR_CHANNELS, I10NM_NUM_HBM_CHANNELS)
+#define I10NM_NUM_DIMMS MAX(I10NM_NUM_DDR_DIMMS, I10NM_NUM_HBM_DIMMS)
+
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#define NUM_IMC MAX(SKX_NUM_IMC, I10NM_NUM_IMC)
+#define NUM_CHANNELS MAX(SKX_NUM_CHANNELS, I10NM_NUM_CHANNELS)
+#define NUM_DIMMS MAX(SKX_NUM_DIMMS, I10NM_NUM_DIMMS)
+
+#define IS_DIMM_PRESENT(r) GET_BITFIELD(r, 15, 15)
+#define IS_NVDIMM_PRESENT(r, i) GET_BITFIELD(r, i, i)
+
+#define MCI_MISC_ECC_MODE(m) (((m) >> 59) & 15)
+#define MCI_MISC_ECC_DDRT 8 /* read from DDRT */
+
+/*
+ * According to Intel Architecture spec vol 3B,
+ * Table 15-10 "IA32_MCi_Status [15:0] Compound Error Code Encoding"
+ * memory errors should fit one of these masks:
+ * 000f 0000 1mmm cccc (binary)
+ * 000f 0010 1mmm cccc (binary) [RAM used as cache]
+ * where:
+ * f = Correction Report Filtering Bit. If 1, subsequent errors
+ * won't be shown
+ * mmm = error type
+ * cccc = channel
+ */
+#define MCACOD_MEM_ERR_MASK 0xef80
+/*
+ * Errors from either the memory of the 1-level memory system or the
+ * 2nd level memory (the slow "far" memory) of the 2-level memory system.
+ */
+#define MCACOD_MEM_CTL_ERR 0x80
+/*
+ * Errors from the 1st level memory (the fast "near" memory as cache)
+ * of the 2-level memory system.
+ */
+#define MCACOD_EXT_MEM_ERR 0x280
+
+/*
+ * Each cpu socket contains some pci devices that provide global
+ * information, and also some that are local to each of the two
+ * memory controllers on the die.
+ */
+struct skx_dev {
+ struct list_head list;
+ u8 bus[4];
+ int seg;
+ struct pci_dev *sad_all;
+ struct pci_dev *util_all;
+ struct pci_dev *uracu; /* for i10nm CPU */
+ struct pci_dev *pcu_cr3; /* for HBM memory detection */
+ u32 mcroute;
+ struct skx_imc {
+ struct mem_ctl_info *mci;
+ struct pci_dev *mdev; /* for i10nm CPU */
+ void __iomem *mbase; /* for i10nm CPU */
+ int chan_mmio_sz; /* for i10nm CPU */
+ int num_channels; /* channels per memory controller */
+ int num_dimms; /* dimms per channel */
+ bool hbm_mc;
+ u8 mc; /* system wide mc# */
+ u8 lmc; /* socket relative mc# */
+ u8 src_id, node_id;
+ struct skx_channel {
+ struct pci_dev *cdev;
+ struct pci_dev *edev;
+ u32 retry_rd_err_log_s;
+ u32 retry_rd_err_log_d;
+ u32 retry_rd_err_log_d2;
+ struct skx_dimm {
+ u8 close_pg;
+ u8 bank_xor_enable;
+ u8 fine_grain_bank;
+ u8 rowbits;
+ u8 colbits;
+ } dimms[NUM_DIMMS];
+ } chan[NUM_CHANNELS];
+ } imc[NUM_IMC];
+};
+
+struct skx_pvt {
+ struct skx_imc *imc;
+};
+
+enum type {
+ SKX,
+ I10NM,
+ SPR,
+ GNR
+};
+
+enum {
+ INDEX_SOCKET,
+ INDEX_MEMCTRL,
+ INDEX_CHANNEL,
+ INDEX_DIMM,
+ INDEX_CS,
+ INDEX_NM_FIRST,
+ INDEX_NM_MEMCTRL = INDEX_NM_FIRST,
+ INDEX_NM_CHANNEL,
+ INDEX_NM_DIMM,
+ INDEX_NM_CS,
+ INDEX_MAX
+};
+
+#define BIT_NM_MEMCTRL BIT_ULL(INDEX_NM_MEMCTRL)
+#define BIT_NM_CHANNEL BIT_ULL(INDEX_NM_CHANNEL)
+#define BIT_NM_DIMM BIT_ULL(INDEX_NM_DIMM)
+#define BIT_NM_CS BIT_ULL(INDEX_NM_CS)
+
+struct decoded_addr {
+ struct mce *mce;
+ struct skx_dev *dev;
+ u64 addr;
+ int socket;
+ int imc;
+ int channel;
+ u64 chan_addr;
+ int sktways;
+ int chanways;
+ int dimm;
+ int cs;
+ int rank;
+ int channel_rank;
+ u64 rank_address;
+ int row;
+ int column;
+ int bank_address;
+ int bank_group;
+ bool decoded_by_adxl;
+};
+
+struct pci_bdf {
+ u32 bus : 8;
+ u32 dev : 5;
+ u32 fun : 3;
+};
+
+struct res_config {
+ enum type type;
+ /* Configuration agent device ID */
+ unsigned int decs_did;
+ /* Default bus number configuration register offset */
+ int busno_cfg_offset;
+ /* DDR memory controllers per socket */
+ int ddr_imc_num;
+ /* DDR channels per DDR memory controller */
+ int ddr_chan_num;
+ /* DDR DIMMs per DDR memory channel */
+ int ddr_dimm_num;
+ /* Per DDR channel memory-mapped I/O size */
+ int ddr_chan_mmio_sz;
+ /* HBM memory controllers per socket */
+ int hbm_imc_num;
+ /* HBM channels per HBM memory controller */
+ int hbm_chan_num;
+ /* HBM DIMMs per HBM memory channel */
+ int hbm_dimm_num;
+ /* Per HBM channel memory-mapped I/O size */
+ int hbm_chan_mmio_sz;
+ bool support_ddr5;
+ /* SAD device BDF */
+ struct pci_bdf sad_all_bdf;
+ /* PCU device BDF */
+ struct pci_bdf pcu_cr3_bdf;
+ /* UTIL device BDF */
+ struct pci_bdf util_all_bdf;
+ /* URACU device BDF */
+ struct pci_bdf uracu_bdf;
+ /* DDR mdev device BDF */
+ struct pci_bdf ddr_mdev_bdf;
+ /* HBM mdev device BDF */
+ struct pci_bdf hbm_mdev_bdf;
+ int sad_all_offset;
+ /* Offsets of retry_rd_err_log registers */
+ u32 *offsets_scrub;
+ u32 *offsets_scrub_hbm0;
+ u32 *offsets_scrub_hbm1;
+ u32 *offsets_demand;
+ u32 *offsets_demand2;
+ u32 *offsets_demand_hbm0;
+ u32 *offsets_demand_hbm1;
+};
+
+typedef int (*get_dimm_config_f)(struct mem_ctl_info *mci,
+ struct res_config *cfg);
+typedef bool (*skx_decode_f)(struct decoded_addr *res);
+typedef void (*skx_show_retry_log_f)(struct decoded_addr *res, char *msg, int len, bool scrub_err);
+
+int __init skx_adxl_get(void);
+void __exit skx_adxl_put(void);
+void skx_set_decode(skx_decode_f decode, skx_show_retry_log_f show_retry_log);
+void skx_set_mem_cfg(bool mem_cfg_2lm);
+
+int skx_get_src_id(struct skx_dev *d, int off, u8 *id);
+int skx_get_node_id(struct skx_dev *d, u8 *id);
+
+int skx_get_all_bus_mappings(struct res_config *cfg, struct list_head **list);
+
+int skx_get_hi_lo(unsigned int did, int off[], u64 *tolm, u64 *tohm);
+
+int skx_get_dimm_info(u32 mtr, u32 mcmtr, u32 amap, struct dimm_info *dimm,
+ struct skx_imc *imc, int chan, int dimmno,
+ struct res_config *cfg);
+
+int skx_get_nvdimm_info(struct dimm_info *dimm, struct skx_imc *imc,
+ int chan, int dimmno, const char *mod_str);
+
+int skx_register_mci(struct skx_imc *imc, struct pci_dev *pdev,
+ const char *ctl_name, const char *mod_str,
+ get_dimm_config_f get_dimm_config,
+ struct res_config *cfg);
+
+int skx_mce_check_error(struct notifier_block *nb, unsigned long val,
+ void *data);
+
+void skx_remove(void);
+
+#endif /* _SKX_COMM_EDAC_H */
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
new file mode 100644
index 0000000000..c4fc64cbec
--- /dev/null
+++ b/drivers/edac/synopsys_edac.c
@@ -0,0 +1,1445 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Synopsys DDR ECC Driver
+ * This driver is based on ppc4xx_edac.c drivers
+ *
+ * Copyright (C) 2012 - 2014 Xilinx, Inc.
+ */
+
+#include <linux/edac.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/interrupt.h>
+#include <linux/of.h>
+
+#include "edac_module.h"
+
+/* Number of cs_rows needed per memory controller */
+#define SYNPS_EDAC_NR_CSROWS 1
+
+/* Number of channels per memory controller */
+#define SYNPS_EDAC_NR_CHANS 1
+
+/* Granularity of reported error in bytes */
+#define SYNPS_EDAC_ERR_GRAIN 1
+
+#define SYNPS_EDAC_MSG_SIZE 256
+
+#define SYNPS_EDAC_MOD_STRING "synps_edac"
+#define SYNPS_EDAC_MOD_VER "1"
+
+/* Synopsys DDR memory controller registers that are relevant to ECC */
+#define CTRL_OFST 0x0
+#define T_ZQ_OFST 0xA4
+
+/* ECC control register */
+#define ECC_CTRL_OFST 0xC4
+/* ECC log register */
+#define CE_LOG_OFST 0xC8
+/* ECC address register */
+#define CE_ADDR_OFST 0xCC
+/* ECC data[31:0] register */
+#define CE_DATA_31_0_OFST 0xD0
+
+/* Uncorrectable error info registers */
+#define UE_LOG_OFST 0xDC
+#define UE_ADDR_OFST 0xE0
+#define UE_DATA_31_0_OFST 0xE4
+
+#define STAT_OFST 0xF0
+#define SCRUB_OFST 0xF4
+
+/* Control register bit field definitions */
+#define CTRL_BW_MASK 0xC
+#define CTRL_BW_SHIFT 2
+
+#define DDRCTL_WDTH_16 1
+#define DDRCTL_WDTH_32 0
+
+/* ZQ register bit field definitions */
+#define T_ZQ_DDRMODE_MASK 0x2
+
+/* ECC control register bit field definitions */
+#define ECC_CTRL_CLR_CE_ERR 0x2
+#define ECC_CTRL_CLR_UE_ERR 0x1
+
+/* ECC correctable/uncorrectable error log register definitions */
+#define LOG_VALID 0x1
+#define CE_LOG_BITPOS_MASK 0xFE
+#define CE_LOG_BITPOS_SHIFT 1
+
+/* ECC correctable/uncorrectable error address register definitions */
+#define ADDR_COL_MASK 0xFFF
+#define ADDR_ROW_MASK 0xFFFF000
+#define ADDR_ROW_SHIFT 12
+#define ADDR_BANK_MASK 0x70000000
+#define ADDR_BANK_SHIFT 28
+
+/* ECC statistic register definitions */
+#define STAT_UECNT_MASK 0xFF
+#define STAT_CECNT_MASK 0xFF00
+#define STAT_CECNT_SHIFT 8
+
+/* ECC scrub register definitions */
+#define SCRUB_MODE_MASK 0x7
+#define SCRUB_MODE_SECDED 0x4
+
+/* DDR ECC Quirks */
+#define DDR_ECC_INTR_SUPPORT BIT(0)
+#define DDR_ECC_DATA_POISON_SUPPORT BIT(1)
+#define DDR_ECC_INTR_SELF_CLEAR BIT(2)
+
+/* ZynqMP Enhanced DDR memory controller registers that are relevant to ECC */
+/* ECC Configuration Registers */
+#define ECC_CFG0_OFST 0x70
+#define ECC_CFG1_OFST 0x74
+
+/* ECC Status Register */
+#define ECC_STAT_OFST 0x78
+
+/* ECC Clear Register */
+#define ECC_CLR_OFST 0x7C
+
+/* ECC Error count Register */
+#define ECC_ERRCNT_OFST 0x80
+
+/* ECC Corrected Error Address Register */
+#define ECC_CEADDR0_OFST 0x84
+#define ECC_CEADDR1_OFST 0x88
+
+/* ECC Syndrome Registers */
+#define ECC_CSYND0_OFST 0x8C
+#define ECC_CSYND1_OFST 0x90
+#define ECC_CSYND2_OFST 0x94
+
+/* ECC Bit Mask0 Address Register */
+#define ECC_BITMASK0_OFST 0x98
+#define ECC_BITMASK1_OFST 0x9C
+#define ECC_BITMASK2_OFST 0xA0
+
+/* ECC UnCorrected Error Address Register */
+#define ECC_UEADDR0_OFST 0xA4
+#define ECC_UEADDR1_OFST 0xA8
+
+/* ECC Syndrome Registers */
+#define ECC_UESYND0_OFST 0xAC
+#define ECC_UESYND1_OFST 0xB0
+#define ECC_UESYND2_OFST 0xB4
+
+/* ECC Poison Address Reg */
+#define ECC_POISON0_OFST 0xB8
+#define ECC_POISON1_OFST 0xBC
+
+#define ECC_ADDRMAP0_OFFSET 0x200
+
+/* Control register bitfield definitions */
+#define ECC_CTRL_BUSWIDTH_MASK 0x3000
+#define ECC_CTRL_BUSWIDTH_SHIFT 12
+#define ECC_CTRL_CLR_CE_ERRCNT BIT(2)
+#define ECC_CTRL_CLR_UE_ERRCNT BIT(3)
+
+/* DDR Control Register width definitions */
+#define DDRCTL_EWDTH_16 2
+#define DDRCTL_EWDTH_32 1
+#define DDRCTL_EWDTH_64 0
+
+/* ECC status register definitions */
+#define ECC_STAT_UECNT_MASK 0xF0000
+#define ECC_STAT_UECNT_SHIFT 16
+#define ECC_STAT_CECNT_MASK 0xF00
+#define ECC_STAT_CECNT_SHIFT 8
+#define ECC_STAT_BITNUM_MASK 0x7F
+
+/* ECC error count register definitions */
+#define ECC_ERRCNT_UECNT_MASK 0xFFFF0000
+#define ECC_ERRCNT_UECNT_SHIFT 16
+#define ECC_ERRCNT_CECNT_MASK 0xFFFF
+
+/* DDR QOS Interrupt register definitions */
+#define DDR_QOS_IRQ_STAT_OFST 0x20200
+#define DDR_QOSUE_MASK 0x4
+#define DDR_QOSCE_MASK 0x2
+#define ECC_CE_UE_INTR_MASK 0x6
+#define DDR_QOS_IRQ_EN_OFST 0x20208
+#define DDR_QOS_IRQ_DB_OFST 0x2020C
+
+/* DDR QOS Interrupt register definitions */
+#define DDR_UE_MASK BIT(9)
+#define DDR_CE_MASK BIT(8)
+
+/* ECC Corrected Error Register Mask and Shifts*/
+#define ECC_CEADDR0_RW_MASK 0x3FFFF
+#define ECC_CEADDR0_RNK_MASK BIT(24)
+#define ECC_CEADDR1_BNKGRP_MASK 0x3000000
+#define ECC_CEADDR1_BNKNR_MASK 0x70000
+#define ECC_CEADDR1_BLKNR_MASK 0xFFF
+#define ECC_CEADDR1_BNKGRP_SHIFT 24
+#define ECC_CEADDR1_BNKNR_SHIFT 16
+
+/* ECC Poison register shifts */
+#define ECC_POISON0_RANK_SHIFT 24
+#define ECC_POISON0_RANK_MASK BIT(24)
+#define ECC_POISON0_COLUMN_SHIFT 0
+#define ECC_POISON0_COLUMN_MASK 0xFFF
+#define ECC_POISON1_BG_SHIFT 28
+#define ECC_POISON1_BG_MASK 0x30000000
+#define ECC_POISON1_BANKNR_SHIFT 24
+#define ECC_POISON1_BANKNR_MASK 0x7000000
+#define ECC_POISON1_ROW_SHIFT 0
+#define ECC_POISON1_ROW_MASK 0x3FFFF
+
+/* DDR Memory type defines */
+#define MEM_TYPE_DDR3 0x1
+#define MEM_TYPE_LPDDR3 0x8
+#define MEM_TYPE_DDR2 0x4
+#define MEM_TYPE_DDR4 0x10
+#define MEM_TYPE_LPDDR4 0x20
+
+/* DDRC Software control register */
+#define DDRC_SWCTL 0x320
+
+/* DDRC ECC CE & UE poison mask */
+#define ECC_CEPOISON_MASK 0x3
+#define ECC_UEPOISON_MASK 0x1
+
+/* DDRC Device config masks */
+#define DDRC_MSTR_CFG_MASK 0xC0000000
+#define DDRC_MSTR_CFG_SHIFT 30
+#define DDRC_MSTR_CFG_X4_MASK 0x0
+#define DDRC_MSTR_CFG_X8_MASK 0x1
+#define DDRC_MSTR_CFG_X16_MASK 0x2
+#define DDRC_MSTR_CFG_X32_MASK 0x3
+
+#define DDR_MAX_ROW_SHIFT 18
+#define DDR_MAX_COL_SHIFT 14
+#define DDR_MAX_BANK_SHIFT 3
+#define DDR_MAX_BANKGRP_SHIFT 2
+
+#define ROW_MAX_VAL_MASK 0xF
+#define COL_MAX_VAL_MASK 0xF
+#define BANK_MAX_VAL_MASK 0x1F
+#define BANKGRP_MAX_VAL_MASK 0x1F
+#define RANK_MAX_VAL_MASK 0x1F
+
+#define ROW_B0_BASE 6
+#define ROW_B1_BASE 7
+#define ROW_B2_BASE 8
+#define ROW_B3_BASE 9
+#define ROW_B4_BASE 10
+#define ROW_B5_BASE 11
+#define ROW_B6_BASE 12
+#define ROW_B7_BASE 13
+#define ROW_B8_BASE 14
+#define ROW_B9_BASE 15
+#define ROW_B10_BASE 16
+#define ROW_B11_BASE 17
+#define ROW_B12_BASE 18
+#define ROW_B13_BASE 19
+#define ROW_B14_BASE 20
+#define ROW_B15_BASE 21
+#define ROW_B16_BASE 22
+#define ROW_B17_BASE 23
+
+#define COL_B2_BASE 2
+#define COL_B3_BASE 3
+#define COL_B4_BASE 4
+#define COL_B5_BASE 5
+#define COL_B6_BASE 6
+#define COL_B7_BASE 7
+#define COL_B8_BASE 8
+#define COL_B9_BASE 9
+#define COL_B10_BASE 10
+#define COL_B11_BASE 11
+#define COL_B12_BASE 12
+#define COL_B13_BASE 13
+
+#define BANK_B0_BASE 2
+#define BANK_B1_BASE 3
+#define BANK_B2_BASE 4
+
+#define BANKGRP_B0_BASE 2
+#define BANKGRP_B1_BASE 3
+
+#define RANK_B0_BASE 6
+
+/**
+ * struct ecc_error_info - ECC error log information.
+ * @row: Row number.
+ * @col: Column number.
+ * @bank: Bank number.
+ * @bitpos: Bit position.
+ * @data: Data causing the error.
+ * @bankgrpnr: Bank group number.
+ * @blknr: Block number.
+ */
+struct ecc_error_info {
+ u32 row;
+ u32 col;
+ u32 bank;
+ u32 bitpos;
+ u32 data;
+ u32 bankgrpnr;
+ u32 blknr;
+};
+
+/**
+ * struct synps_ecc_status - ECC status information to report.
+ * @ce_cnt: Correctable error count.
+ * @ue_cnt: Uncorrectable error count.
+ * @ceinfo: Correctable error log information.
+ * @ueinfo: Uncorrectable error log information.
+ */
+struct synps_ecc_status {
+ u32 ce_cnt;
+ u32 ue_cnt;
+ struct ecc_error_info ceinfo;
+ struct ecc_error_info ueinfo;
+};
+
+/**
+ * struct synps_edac_priv - DDR memory controller private instance data.
+ * @baseaddr: Base address of the DDR controller.
+ * @message: Buffer for framing the event specific info.
+ * @stat: ECC status information.
+ * @p_data: Platform data.
+ * @ce_cnt: Correctable Error count.
+ * @ue_cnt: Uncorrectable Error count.
+ * @poison_addr: Data poison address.
+ * @row_shift: Bit shifts for row bit.
+ * @col_shift: Bit shifts for column bit.
+ * @bank_shift: Bit shifts for bank bit.
+ * @bankgrp_shift: Bit shifts for bank group bit.
+ * @rank_shift: Bit shifts for rank bit.
+ */
+struct synps_edac_priv {
+ void __iomem *baseaddr;
+ char message[SYNPS_EDAC_MSG_SIZE];
+ struct synps_ecc_status stat;
+ const struct synps_platform_data *p_data;
+ u32 ce_cnt;
+ u32 ue_cnt;
+#ifdef CONFIG_EDAC_DEBUG
+ ulong poison_addr;
+ u32 row_shift[18];
+ u32 col_shift[14];
+ u32 bank_shift[3];
+ u32 bankgrp_shift[2];
+ u32 rank_shift[1];
+#endif
+};
+
+/**
+ * struct synps_platform_data - synps platform data structure.
+ * @get_error_info: Get EDAC error info.
+ * @get_mtype: Get mtype.
+ * @get_dtype: Get dtype.
+ * @get_ecc_state: Get ECC state.
+ * @quirks: To differentiate IPs.
+ */
+struct synps_platform_data {
+ int (*get_error_info)(struct synps_edac_priv *priv);
+ enum mem_type (*get_mtype)(const void __iomem *base);
+ enum dev_type (*get_dtype)(const void __iomem *base);
+ bool (*get_ecc_state)(void __iomem *base);
+ int quirks;
+};
+
+/**
+ * zynq_get_error_info - Get the current ECC error info.
+ * @priv: DDR memory controller private instance data.
+ *
+ * Return: one if there is no error, otherwise zero.
+ */
+static int zynq_get_error_info(struct synps_edac_priv *priv)
+{
+ struct synps_ecc_status *p;
+ u32 regval, clearval = 0;
+ void __iomem *base;
+
+ base = priv->baseaddr;
+ p = &priv->stat;
+
+ regval = readl(base + STAT_OFST);
+ if (!regval)
+ return 1;
+
+ p->ce_cnt = (regval & STAT_CECNT_MASK) >> STAT_CECNT_SHIFT;
+ p->ue_cnt = regval & STAT_UECNT_MASK;
+
+ regval = readl(base + CE_LOG_OFST);
+ if (!(p->ce_cnt && (regval & LOG_VALID)))
+ goto ue_err;
+
+ p->ceinfo.bitpos = (regval & CE_LOG_BITPOS_MASK) >> CE_LOG_BITPOS_SHIFT;
+ regval = readl(base + CE_ADDR_OFST);
+ p->ceinfo.row = (regval & ADDR_ROW_MASK) >> ADDR_ROW_SHIFT;
+ p->ceinfo.col = regval & ADDR_COL_MASK;
+ p->ceinfo.bank = (regval & ADDR_BANK_MASK) >> ADDR_BANK_SHIFT;
+ p->ceinfo.data = readl(base + CE_DATA_31_0_OFST);
+ edac_dbg(3, "CE bit position: %d data: %d\n", p->ceinfo.bitpos,
+ p->ceinfo.data);
+ clearval = ECC_CTRL_CLR_CE_ERR;
+
+ue_err:
+ regval = readl(base + UE_LOG_OFST);
+ if (!(p->ue_cnt && (regval & LOG_VALID)))
+ goto out;
+
+ regval = readl(base + UE_ADDR_OFST);
+ p->ueinfo.row = (regval & ADDR_ROW_MASK) >> ADDR_ROW_SHIFT;
+ p->ueinfo.col = regval & ADDR_COL_MASK;
+ p->ueinfo.bank = (regval & ADDR_BANK_MASK) >> ADDR_BANK_SHIFT;
+ p->ueinfo.data = readl(base + UE_DATA_31_0_OFST);
+ clearval |= ECC_CTRL_CLR_UE_ERR;
+
+out:
+ writel(clearval, base + ECC_CTRL_OFST);
+ writel(0x0, base + ECC_CTRL_OFST);
+
+ return 0;
+}
+
+/**
+ * zynqmp_get_error_info - Get the current ECC error info.
+ * @priv: DDR memory controller private instance data.
+ *
+ * Return: one if there is no error otherwise returns zero.
+ */
+static int zynqmp_get_error_info(struct synps_edac_priv *priv)
+{
+ struct synps_ecc_status *p;
+ u32 regval, clearval = 0;
+ void __iomem *base;
+
+ base = priv->baseaddr;
+ p = &priv->stat;
+
+ regval = readl(base + ECC_ERRCNT_OFST);
+ p->ce_cnt = regval & ECC_ERRCNT_CECNT_MASK;
+ p->ue_cnt = (regval & ECC_ERRCNT_UECNT_MASK) >> ECC_ERRCNT_UECNT_SHIFT;
+ if (!p->ce_cnt)
+ goto ue_err;
+
+ regval = readl(base + ECC_STAT_OFST);
+ if (!regval)
+ return 1;
+
+ p->ceinfo.bitpos = (regval & ECC_STAT_BITNUM_MASK);
+
+ regval = readl(base + ECC_CEADDR0_OFST);
+ p->ceinfo.row = (regval & ECC_CEADDR0_RW_MASK);
+ regval = readl(base + ECC_CEADDR1_OFST);
+ p->ceinfo.bank = (regval & ECC_CEADDR1_BNKNR_MASK) >>
+ ECC_CEADDR1_BNKNR_SHIFT;
+ p->ceinfo.bankgrpnr = (regval & ECC_CEADDR1_BNKGRP_MASK) >>
+ ECC_CEADDR1_BNKGRP_SHIFT;
+ p->ceinfo.blknr = (regval & ECC_CEADDR1_BLKNR_MASK);
+ p->ceinfo.data = readl(base + ECC_CSYND0_OFST);
+ edac_dbg(2, "ECCCSYN0: 0x%08X ECCCSYN1: 0x%08X ECCCSYN2: 0x%08X\n",
+ readl(base + ECC_CSYND0_OFST), readl(base + ECC_CSYND1_OFST),
+ readl(base + ECC_CSYND2_OFST));
+ue_err:
+ if (!p->ue_cnt)
+ goto out;
+
+ regval = readl(base + ECC_UEADDR0_OFST);
+ p->ueinfo.row = (regval & ECC_CEADDR0_RW_MASK);
+ regval = readl(base + ECC_UEADDR1_OFST);
+ p->ueinfo.bankgrpnr = (regval & ECC_CEADDR1_BNKGRP_MASK) >>
+ ECC_CEADDR1_BNKGRP_SHIFT;
+ p->ueinfo.bank = (regval & ECC_CEADDR1_BNKNR_MASK) >>
+ ECC_CEADDR1_BNKNR_SHIFT;
+ p->ueinfo.blknr = (regval & ECC_CEADDR1_BLKNR_MASK);
+ p->ueinfo.data = readl(base + ECC_UESYND0_OFST);
+out:
+ clearval = ECC_CTRL_CLR_CE_ERR | ECC_CTRL_CLR_CE_ERRCNT;
+ clearval |= ECC_CTRL_CLR_UE_ERR | ECC_CTRL_CLR_UE_ERRCNT;
+ writel(clearval, base + ECC_CLR_OFST);
+ writel(0x0, base + ECC_CLR_OFST);
+
+ return 0;
+}
+
+/**
+ * handle_error - Handle Correctable and Uncorrectable errors.
+ * @mci: EDAC memory controller instance.
+ * @p: Synopsys ECC status structure.
+ *
+ * Handles ECC correctable and uncorrectable errors.
+ */
+static void handle_error(struct mem_ctl_info *mci, struct synps_ecc_status *p)
+{
+ struct synps_edac_priv *priv = mci->pvt_info;
+ struct ecc_error_info *pinf;
+
+ if (p->ce_cnt) {
+ pinf = &p->ceinfo;
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT) {
+ snprintf(priv->message, SYNPS_EDAC_MSG_SIZE,
+ "DDR ECC error type:%s Row %d Bank %d BankGroup Number %d Block Number %d Bit Position: %d Data: 0x%08x",
+ "CE", pinf->row, pinf->bank,
+ pinf->bankgrpnr, pinf->blknr,
+ pinf->bitpos, pinf->data);
+ } else {
+ snprintf(priv->message, SYNPS_EDAC_MSG_SIZE,
+ "DDR ECC error type:%s Row %d Bank %d Col %d Bit Position: %d Data: 0x%08x",
+ "CE", pinf->row, pinf->bank, pinf->col,
+ pinf->bitpos, pinf->data);
+ }
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
+ p->ce_cnt, 0, 0, 0, 0, 0, -1,
+ priv->message, "");
+ }
+
+ if (p->ue_cnt) {
+ pinf = &p->ueinfo;
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT) {
+ snprintf(priv->message, SYNPS_EDAC_MSG_SIZE,
+ "DDR ECC error type :%s Row %d Bank %d BankGroup Number %d Block Number %d",
+ "UE", pinf->row, pinf->bank,
+ pinf->bankgrpnr, pinf->blknr);
+ } else {
+ snprintf(priv->message, SYNPS_EDAC_MSG_SIZE,
+ "DDR ECC error type :%s Row %d Bank %d Col %d ",
+ "UE", pinf->row, pinf->bank, pinf->col);
+ }
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
+ p->ue_cnt, 0, 0, 0, 0, 0, -1,
+ priv->message, "");
+ }
+
+ memset(p, 0, sizeof(*p));
+}
+
+static void enable_intr(struct synps_edac_priv *priv)
+{
+ /* Enable UE/CE Interrupts */
+ if (priv->p_data->quirks & DDR_ECC_INTR_SELF_CLEAR)
+ writel(DDR_UE_MASK | DDR_CE_MASK,
+ priv->baseaddr + ECC_CLR_OFST);
+ else
+ writel(DDR_QOSUE_MASK | DDR_QOSCE_MASK,
+ priv->baseaddr + DDR_QOS_IRQ_EN_OFST);
+
+}
+
+static void disable_intr(struct synps_edac_priv *priv)
+{
+ /* Disable UE/CE Interrupts */
+ if (priv->p_data->quirks & DDR_ECC_INTR_SELF_CLEAR)
+ writel(0x0, priv->baseaddr + ECC_CLR_OFST);
+ else
+ writel(DDR_QOSUE_MASK | DDR_QOSCE_MASK,
+ priv->baseaddr + DDR_QOS_IRQ_DB_OFST);
+}
+
+/**
+ * intr_handler - Interrupt Handler for ECC interrupts.
+ * @irq: IRQ number.
+ * @dev_id: Device ID.
+ *
+ * Return: IRQ_NONE, if interrupt not set or IRQ_HANDLED otherwise.
+ */
+static irqreturn_t intr_handler(int irq, void *dev_id)
+{
+ const struct synps_platform_data *p_data;
+ struct mem_ctl_info *mci = dev_id;
+ struct synps_edac_priv *priv;
+ int status, regval;
+
+ priv = mci->pvt_info;
+ p_data = priv->p_data;
+
+ /*
+ * v3.0 of the controller has the ce/ue bits cleared automatically,
+ * so this condition does not apply.
+ */
+ if (!(priv->p_data->quirks & DDR_ECC_INTR_SELF_CLEAR)) {
+ regval = readl(priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+ regval &= (DDR_QOSCE_MASK | DDR_QOSUE_MASK);
+ if (!(regval & ECC_CE_UE_INTR_MASK))
+ return IRQ_NONE;
+ }
+
+ status = p_data->get_error_info(priv);
+ if (status)
+ return IRQ_NONE;
+
+ priv->ce_cnt += priv->stat.ce_cnt;
+ priv->ue_cnt += priv->stat.ue_cnt;
+ handle_error(mci, &priv->stat);
+
+ edac_dbg(3, "Total error count CE %d UE %d\n",
+ priv->ce_cnt, priv->ue_cnt);
+ /* v3.0 of the controller does not have this register */
+ if (!(priv->p_data->quirks & DDR_ECC_INTR_SELF_CLEAR))
+ writel(regval, priv->baseaddr + DDR_QOS_IRQ_STAT_OFST);
+ else
+ enable_intr(priv);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * check_errors - Check controller for ECC errors.
+ * @mci: EDAC memory controller instance.
+ *
+ * Check and post ECC errors. Called by the polling thread.
+ */
+static void check_errors(struct mem_ctl_info *mci)
+{
+ const struct synps_platform_data *p_data;
+ struct synps_edac_priv *priv;
+ int status;
+
+ priv = mci->pvt_info;
+ p_data = priv->p_data;
+
+ status = p_data->get_error_info(priv);
+ if (status)
+ return;
+
+ priv->ce_cnt += priv->stat.ce_cnt;
+ priv->ue_cnt += priv->stat.ue_cnt;
+ handle_error(mci, &priv->stat);
+
+ edac_dbg(3, "Total error count CE %d UE %d\n",
+ priv->ce_cnt, priv->ue_cnt);
+}
+
+/**
+ * zynq_get_dtype - Return the controller memory width.
+ * @base: DDR memory controller base address.
+ *
+ * Get the EDAC device type width appropriate for the current controller
+ * configuration.
+ *
+ * Return: a device type width enumeration.
+ */
+static enum dev_type zynq_get_dtype(const void __iomem *base)
+{
+ enum dev_type dt;
+ u32 width;
+
+ width = readl(base + CTRL_OFST);
+ width = (width & CTRL_BW_MASK) >> CTRL_BW_SHIFT;
+
+ switch (width) {
+ case DDRCTL_WDTH_16:
+ dt = DEV_X2;
+ break;
+ case DDRCTL_WDTH_32:
+ dt = DEV_X4;
+ break;
+ default:
+ dt = DEV_UNKNOWN;
+ }
+
+ return dt;
+}
+
+/**
+ * zynqmp_get_dtype - Return the controller memory width.
+ * @base: DDR memory controller base address.
+ *
+ * Get the EDAC device type width appropriate for the current controller
+ * configuration.
+ *
+ * Return: a device type width enumeration.
+ */
+static enum dev_type zynqmp_get_dtype(const void __iomem *base)
+{
+ enum dev_type dt;
+ u32 width;
+
+ width = readl(base + CTRL_OFST);
+ width = (width & ECC_CTRL_BUSWIDTH_MASK) >> ECC_CTRL_BUSWIDTH_SHIFT;
+ switch (width) {
+ case DDRCTL_EWDTH_16:
+ dt = DEV_X2;
+ break;
+ case DDRCTL_EWDTH_32:
+ dt = DEV_X4;
+ break;
+ case DDRCTL_EWDTH_64:
+ dt = DEV_X8;
+ break;
+ default:
+ dt = DEV_UNKNOWN;
+ }
+
+ return dt;
+}
+
+/**
+ * zynq_get_ecc_state - Return the controller ECC enable/disable status.
+ * @base: DDR memory controller base address.
+ *
+ * Get the ECC enable/disable status of the controller.
+ *
+ * Return: true if enabled, otherwise false.
+ */
+static bool zynq_get_ecc_state(void __iomem *base)
+{
+ enum dev_type dt;
+ u32 ecctype;
+
+ dt = zynq_get_dtype(base);
+ if (dt == DEV_UNKNOWN)
+ return false;
+
+ ecctype = readl(base + SCRUB_OFST) & SCRUB_MODE_MASK;
+ if ((ecctype == SCRUB_MODE_SECDED) && (dt == DEV_X2))
+ return true;
+
+ return false;
+}
+
+/**
+ * zynqmp_get_ecc_state - Return the controller ECC enable/disable status.
+ * @base: DDR memory controller base address.
+ *
+ * Get the ECC enable/disable status for the controller.
+ *
+ * Return: a ECC status boolean i.e true/false - enabled/disabled.
+ */
+static bool zynqmp_get_ecc_state(void __iomem *base)
+{
+ enum dev_type dt;
+ u32 ecctype;
+
+ dt = zynqmp_get_dtype(base);
+ if (dt == DEV_UNKNOWN)
+ return false;
+
+ ecctype = readl(base + ECC_CFG0_OFST) & SCRUB_MODE_MASK;
+ if ((ecctype == SCRUB_MODE_SECDED) &&
+ ((dt == DEV_X2) || (dt == DEV_X4) || (dt == DEV_X8)))
+ return true;
+
+ return false;
+}
+
+/**
+ * get_memsize - Read the size of the attached memory device.
+ *
+ * Return: the memory size in bytes.
+ */
+static u32 get_memsize(void)
+{
+ struct sysinfo inf;
+
+ si_meminfo(&inf);
+
+ return inf.totalram * inf.mem_unit;
+}
+
+/**
+ * zynq_get_mtype - Return the controller memory type.
+ * @base: Synopsys ECC status structure.
+ *
+ * Get the EDAC memory type appropriate for the current controller
+ * configuration.
+ *
+ * Return: a memory type enumeration.
+ */
+static enum mem_type zynq_get_mtype(const void __iomem *base)
+{
+ enum mem_type mt;
+ u32 memtype;
+
+ memtype = readl(base + T_ZQ_OFST);
+
+ if (memtype & T_ZQ_DDRMODE_MASK)
+ mt = MEM_DDR3;
+ else
+ mt = MEM_DDR2;
+
+ return mt;
+}
+
+/**
+ * zynqmp_get_mtype - Returns controller memory type.
+ * @base: Synopsys ECC status structure.
+ *
+ * Get the EDAC memory type appropriate for the current controller
+ * configuration.
+ *
+ * Return: a memory type enumeration.
+ */
+static enum mem_type zynqmp_get_mtype(const void __iomem *base)
+{
+ enum mem_type mt;
+ u32 memtype;
+
+ memtype = readl(base + CTRL_OFST);
+
+ if ((memtype & MEM_TYPE_DDR3) || (memtype & MEM_TYPE_LPDDR3))
+ mt = MEM_DDR3;
+ else if (memtype & MEM_TYPE_DDR2)
+ mt = MEM_RDDR2;
+ else if ((memtype & MEM_TYPE_LPDDR4) || (memtype & MEM_TYPE_DDR4))
+ mt = MEM_DDR4;
+ else
+ mt = MEM_EMPTY;
+
+ return mt;
+}
+
+/**
+ * init_csrows - Initialize the csrow data.
+ * @mci: EDAC memory controller instance.
+ *
+ * Initialize the chip select rows associated with the EDAC memory
+ * controller instance.
+ */
+static void init_csrows(struct mem_ctl_info *mci)
+{
+ struct synps_edac_priv *priv = mci->pvt_info;
+ const struct synps_platform_data *p_data;
+ struct csrow_info *csi;
+ struct dimm_info *dimm;
+ u32 size, row;
+ int j;
+
+ p_data = priv->p_data;
+
+ for (row = 0; row < mci->nr_csrows; row++) {
+ csi = mci->csrows[row];
+ size = get_memsize();
+
+ for (j = 0; j < csi->nr_channels; j++) {
+ dimm = csi->channels[j]->dimm;
+ dimm->edac_mode = EDAC_SECDED;
+ dimm->mtype = p_data->get_mtype(priv->baseaddr);
+ dimm->nr_pages = (size >> PAGE_SHIFT) / csi->nr_channels;
+ dimm->grain = SYNPS_EDAC_ERR_GRAIN;
+ dimm->dtype = p_data->get_dtype(priv->baseaddr);
+ }
+ }
+}
+
+/**
+ * mc_init - Initialize one driver instance.
+ * @mci: EDAC memory controller instance.
+ * @pdev: platform device.
+ *
+ * Perform initialization of the EDAC memory controller instance and
+ * related driver-private data associated with the memory controller the
+ * instance is bound to.
+ */
+static void mc_init(struct mem_ctl_info *mci, struct platform_device *pdev)
+{
+ struct synps_edac_priv *priv;
+
+ mci->pdev = &pdev->dev;
+ priv = mci->pvt_info;
+ platform_set_drvdata(pdev, mci);
+
+ /* Initialize controller capabilities and configuration */
+ mci->mtype_cap = MEM_FLAG_DDR3 | MEM_FLAG_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->scrub_cap = SCRUB_HW_SRC;
+ mci->scrub_mode = SCRUB_NONE;
+
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->ctl_name = "synps_ddr_controller";
+ mci->dev_name = SYNPS_EDAC_MOD_STRING;
+ mci->mod_name = SYNPS_EDAC_MOD_VER;
+
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT) {
+ edac_op_state = EDAC_OPSTATE_INT;
+ } else {
+ edac_op_state = EDAC_OPSTATE_POLL;
+ mci->edac_check = check_errors;
+ }
+
+ mci->ctl_page_to_phys = NULL;
+
+ init_csrows(mci);
+}
+
+static int setup_irq(struct mem_ctl_info *mci,
+ struct platform_device *pdev)
+{
+ struct synps_edac_priv *priv = mci->pvt_info;
+ int ret, irq;
+
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "No IRQ %d in DT\n", irq);
+ return irq;
+ }
+
+ ret = devm_request_irq(&pdev->dev, irq, intr_handler,
+ 0, dev_name(&pdev->dev), mci);
+ if (ret < 0) {
+ edac_printk(KERN_ERR, EDAC_MC, "Failed to request IRQ\n");
+ return ret;
+ }
+
+ enable_intr(priv);
+
+ return 0;
+}
+
+static const struct synps_platform_data zynq_edac_def = {
+ .get_error_info = zynq_get_error_info,
+ .get_mtype = zynq_get_mtype,
+ .get_dtype = zynq_get_dtype,
+ .get_ecc_state = zynq_get_ecc_state,
+ .quirks = 0,
+};
+
+static const struct synps_platform_data zynqmp_edac_def = {
+ .get_error_info = zynqmp_get_error_info,
+ .get_mtype = zynqmp_get_mtype,
+ .get_dtype = zynqmp_get_dtype,
+ .get_ecc_state = zynqmp_get_ecc_state,
+ .quirks = (DDR_ECC_INTR_SUPPORT
+#ifdef CONFIG_EDAC_DEBUG
+ | DDR_ECC_DATA_POISON_SUPPORT
+#endif
+ ),
+};
+
+static const struct synps_platform_data synopsys_edac_def = {
+ .get_error_info = zynqmp_get_error_info,
+ .get_mtype = zynqmp_get_mtype,
+ .get_dtype = zynqmp_get_dtype,
+ .get_ecc_state = zynqmp_get_ecc_state,
+ .quirks = (DDR_ECC_INTR_SUPPORT | DDR_ECC_INTR_SELF_CLEAR
+#ifdef CONFIG_EDAC_DEBUG
+ | DDR_ECC_DATA_POISON_SUPPORT
+#endif
+ ),
+};
+
+
+static const struct of_device_id synps_edac_match[] = {
+ {
+ .compatible = "xlnx,zynq-ddrc-a05",
+ .data = (void *)&zynq_edac_def
+ },
+ {
+ .compatible = "xlnx,zynqmp-ddrc-2.40a",
+ .data = (void *)&zynqmp_edac_def
+ },
+ {
+ .compatible = "snps,ddrc-3.80a",
+ .data = (void *)&synopsys_edac_def
+ },
+ {
+ /* end of table */
+ }
+};
+
+MODULE_DEVICE_TABLE(of, synps_edac_match);
+
+#ifdef CONFIG_EDAC_DEBUG
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
+/**
+ * ddr_poison_setup - Update poison registers.
+ * @priv: DDR memory controller private instance data.
+ *
+ * Update poison registers as per DDR mapping.
+ * Return: none.
+ */
+static void ddr_poison_setup(struct synps_edac_priv *priv)
+{
+ int col = 0, row = 0, bank = 0, bankgrp = 0, rank = 0, regval;
+ int index;
+ ulong hif_addr = 0;
+
+ hif_addr = priv->poison_addr >> 3;
+
+ for (index = 0; index < DDR_MAX_ROW_SHIFT; index++) {
+ if (priv->row_shift[index])
+ row |= (((hif_addr >> priv->row_shift[index]) &
+ BIT(0)) << index);
+ else
+ break;
+ }
+
+ for (index = 0; index < DDR_MAX_COL_SHIFT; index++) {
+ if (priv->col_shift[index] || index < 3)
+ col |= (((hif_addr >> priv->col_shift[index]) &
+ BIT(0)) << index);
+ else
+ break;
+ }
+
+ for (index = 0; index < DDR_MAX_BANK_SHIFT; index++) {
+ if (priv->bank_shift[index])
+ bank |= (((hif_addr >> priv->bank_shift[index]) &
+ BIT(0)) << index);
+ else
+ break;
+ }
+
+ for (index = 0; index < DDR_MAX_BANKGRP_SHIFT; index++) {
+ if (priv->bankgrp_shift[index])
+ bankgrp |= (((hif_addr >> priv->bankgrp_shift[index])
+ & BIT(0)) << index);
+ else
+ break;
+ }
+
+ if (priv->rank_shift[0])
+ rank = (hif_addr >> priv->rank_shift[0]) & BIT(0);
+
+ regval = (rank << ECC_POISON0_RANK_SHIFT) & ECC_POISON0_RANK_MASK;
+ regval |= (col << ECC_POISON0_COLUMN_SHIFT) & ECC_POISON0_COLUMN_MASK;
+ writel(regval, priv->baseaddr + ECC_POISON0_OFST);
+
+ regval = (bankgrp << ECC_POISON1_BG_SHIFT) & ECC_POISON1_BG_MASK;
+ regval |= (bank << ECC_POISON1_BANKNR_SHIFT) & ECC_POISON1_BANKNR_MASK;
+ regval |= (row << ECC_POISON1_ROW_SHIFT) & ECC_POISON1_ROW_MASK;
+ writel(regval, priv->baseaddr + ECC_POISON1_OFST);
+}
+
+static ssize_t inject_data_error_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct synps_edac_priv *priv = mci->pvt_info;
+
+ return sprintf(data, "Poison0 Addr: 0x%08x\n\rPoison1 Addr: 0x%08x\n\r"
+ "Error injection Address: 0x%lx\n\r",
+ readl(priv->baseaddr + ECC_POISON0_OFST),
+ readl(priv->baseaddr + ECC_POISON1_OFST),
+ priv->poison_addr);
+}
+
+static ssize_t inject_data_error_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct synps_edac_priv *priv = mci->pvt_info;
+
+ if (kstrtoul(data, 0, &priv->poison_addr))
+ return -EINVAL;
+
+ ddr_poison_setup(priv);
+
+ return count;
+}
+
+static ssize_t inject_data_poison_show(struct device *dev,
+ struct device_attribute *mattr,
+ char *data)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct synps_edac_priv *priv = mci->pvt_info;
+
+ return sprintf(data, "Data Poisoning: %s\n\r",
+ (((readl(priv->baseaddr + ECC_CFG1_OFST)) & 0x3) == 0x3)
+ ? ("Correctable Error") : ("UnCorrectable Error"));
+}
+
+static ssize_t inject_data_poison_store(struct device *dev,
+ struct device_attribute *mattr,
+ const char *data, size_t count)
+{
+ struct mem_ctl_info *mci = to_mci(dev);
+ struct synps_edac_priv *priv = mci->pvt_info;
+
+ writel(0, priv->baseaddr + DDRC_SWCTL);
+ if (strncmp(data, "CE", 2) == 0)
+ writel(ECC_CEPOISON_MASK, priv->baseaddr + ECC_CFG1_OFST);
+ else
+ writel(ECC_UEPOISON_MASK, priv->baseaddr + ECC_CFG1_OFST);
+ writel(1, priv->baseaddr + DDRC_SWCTL);
+
+ return count;
+}
+
+static DEVICE_ATTR_RW(inject_data_error);
+static DEVICE_ATTR_RW(inject_data_poison);
+
+static int edac_create_sysfs_attributes(struct mem_ctl_info *mci)
+{
+ int rc;
+
+ rc = device_create_file(&mci->dev, &dev_attr_inject_data_error);
+ if (rc < 0)
+ return rc;
+ rc = device_create_file(&mci->dev, &dev_attr_inject_data_poison);
+ if (rc < 0)
+ return rc;
+ return 0;
+}
+
+static void edac_remove_sysfs_attributes(struct mem_ctl_info *mci)
+{
+ device_remove_file(&mci->dev, &dev_attr_inject_data_error);
+ device_remove_file(&mci->dev, &dev_attr_inject_data_poison);
+}
+
+static void setup_row_address_map(struct synps_edac_priv *priv, u32 *addrmap)
+{
+ u32 addrmap_row_b2_10;
+ int index;
+
+ priv->row_shift[0] = (addrmap[5] & ROW_MAX_VAL_MASK) + ROW_B0_BASE;
+ priv->row_shift[1] = ((addrmap[5] >> 8) &
+ ROW_MAX_VAL_MASK) + ROW_B1_BASE;
+
+ addrmap_row_b2_10 = (addrmap[5] >> 16) & ROW_MAX_VAL_MASK;
+ if (addrmap_row_b2_10 != ROW_MAX_VAL_MASK) {
+ for (index = 2; index < 11; index++)
+ priv->row_shift[index] = addrmap_row_b2_10 +
+ index + ROW_B0_BASE;
+
+ } else {
+ priv->row_shift[2] = (addrmap[9] &
+ ROW_MAX_VAL_MASK) + ROW_B2_BASE;
+ priv->row_shift[3] = ((addrmap[9] >> 8) &
+ ROW_MAX_VAL_MASK) + ROW_B3_BASE;
+ priv->row_shift[4] = ((addrmap[9] >> 16) &
+ ROW_MAX_VAL_MASK) + ROW_B4_BASE;
+ priv->row_shift[5] = ((addrmap[9] >> 24) &
+ ROW_MAX_VAL_MASK) + ROW_B5_BASE;
+ priv->row_shift[6] = (addrmap[10] &
+ ROW_MAX_VAL_MASK) + ROW_B6_BASE;
+ priv->row_shift[7] = ((addrmap[10] >> 8) &
+ ROW_MAX_VAL_MASK) + ROW_B7_BASE;
+ priv->row_shift[8] = ((addrmap[10] >> 16) &
+ ROW_MAX_VAL_MASK) + ROW_B8_BASE;
+ priv->row_shift[9] = ((addrmap[10] >> 24) &
+ ROW_MAX_VAL_MASK) + ROW_B9_BASE;
+ priv->row_shift[10] = (addrmap[11] &
+ ROW_MAX_VAL_MASK) + ROW_B10_BASE;
+ }
+
+ priv->row_shift[11] = (((addrmap[5] >> 24) & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : (((addrmap[5] >> 24) &
+ ROW_MAX_VAL_MASK) + ROW_B11_BASE);
+ priv->row_shift[12] = ((addrmap[6] & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : ((addrmap[6] &
+ ROW_MAX_VAL_MASK) + ROW_B12_BASE);
+ priv->row_shift[13] = (((addrmap[6] >> 8) & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : (((addrmap[6] >> 8) &
+ ROW_MAX_VAL_MASK) + ROW_B13_BASE);
+ priv->row_shift[14] = (((addrmap[6] >> 16) & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : (((addrmap[6] >> 16) &
+ ROW_MAX_VAL_MASK) + ROW_B14_BASE);
+ priv->row_shift[15] = (((addrmap[6] >> 24) & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : (((addrmap[6] >> 24) &
+ ROW_MAX_VAL_MASK) + ROW_B15_BASE);
+ priv->row_shift[16] = ((addrmap[7] & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : ((addrmap[7] &
+ ROW_MAX_VAL_MASK) + ROW_B16_BASE);
+ priv->row_shift[17] = (((addrmap[7] >> 8) & ROW_MAX_VAL_MASK) ==
+ ROW_MAX_VAL_MASK) ? 0 : (((addrmap[7] >> 8) &
+ ROW_MAX_VAL_MASK) + ROW_B17_BASE);
+}
+
+static void setup_column_address_map(struct synps_edac_priv *priv, u32 *addrmap)
+{
+ u32 width, memtype;
+ int index;
+
+ memtype = readl(priv->baseaddr + CTRL_OFST);
+ width = (memtype & ECC_CTRL_BUSWIDTH_MASK) >> ECC_CTRL_BUSWIDTH_SHIFT;
+
+ priv->col_shift[0] = 0;
+ priv->col_shift[1] = 1;
+ priv->col_shift[2] = (addrmap[2] & COL_MAX_VAL_MASK) + COL_B2_BASE;
+ priv->col_shift[3] = ((addrmap[2] >> 8) &
+ COL_MAX_VAL_MASK) + COL_B3_BASE;
+ priv->col_shift[4] = (((addrmap[2] >> 16) & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : (((addrmap[2] >> 16) &
+ COL_MAX_VAL_MASK) + COL_B4_BASE);
+ priv->col_shift[5] = (((addrmap[2] >> 24) & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : (((addrmap[2] >> 24) &
+ COL_MAX_VAL_MASK) + COL_B5_BASE);
+ priv->col_shift[6] = ((addrmap[3] & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : ((addrmap[3] &
+ COL_MAX_VAL_MASK) + COL_B6_BASE);
+ priv->col_shift[7] = (((addrmap[3] >> 8) & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : (((addrmap[3] >> 8) &
+ COL_MAX_VAL_MASK) + COL_B7_BASE);
+ priv->col_shift[8] = (((addrmap[3] >> 16) & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : (((addrmap[3] >> 16) &
+ COL_MAX_VAL_MASK) + COL_B8_BASE);
+ priv->col_shift[9] = (((addrmap[3] >> 24) & COL_MAX_VAL_MASK) ==
+ COL_MAX_VAL_MASK) ? 0 : (((addrmap[3] >> 24) &
+ COL_MAX_VAL_MASK) + COL_B9_BASE);
+ if (width == DDRCTL_EWDTH_64) {
+ if (memtype & MEM_TYPE_LPDDR3) {
+ priv->col_shift[10] = ((addrmap[4] &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ ((addrmap[4] & COL_MAX_VAL_MASK) +
+ COL_B10_BASE);
+ priv->col_shift[11] = (((addrmap[4] >> 8) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[4] >> 8) & COL_MAX_VAL_MASK) +
+ COL_B11_BASE);
+ } else {
+ priv->col_shift[11] = ((addrmap[4] &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ ((addrmap[4] & COL_MAX_VAL_MASK) +
+ COL_B10_BASE);
+ priv->col_shift[13] = (((addrmap[4] >> 8) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[4] >> 8) & COL_MAX_VAL_MASK) +
+ COL_B11_BASE);
+ }
+ } else if (width == DDRCTL_EWDTH_32) {
+ if (memtype & MEM_TYPE_LPDDR3) {
+ priv->col_shift[10] = (((addrmap[3] >> 24) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 24) & COL_MAX_VAL_MASK) +
+ COL_B9_BASE);
+ priv->col_shift[11] = ((addrmap[4] &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ ((addrmap[4] & COL_MAX_VAL_MASK) +
+ COL_B10_BASE);
+ } else {
+ priv->col_shift[11] = (((addrmap[3] >> 24) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 24) & COL_MAX_VAL_MASK) +
+ COL_B9_BASE);
+ priv->col_shift[13] = ((addrmap[4] &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ ((addrmap[4] & COL_MAX_VAL_MASK) +
+ COL_B10_BASE);
+ }
+ } else {
+ if (memtype & MEM_TYPE_LPDDR3) {
+ priv->col_shift[10] = (((addrmap[3] >> 16) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 16) & COL_MAX_VAL_MASK) +
+ COL_B8_BASE);
+ priv->col_shift[11] = (((addrmap[3] >> 24) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 24) & COL_MAX_VAL_MASK) +
+ COL_B9_BASE);
+ priv->col_shift[13] = ((addrmap[4] &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ ((addrmap[4] & COL_MAX_VAL_MASK) +
+ COL_B10_BASE);
+ } else {
+ priv->col_shift[11] = (((addrmap[3] >> 16) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 16) & COL_MAX_VAL_MASK) +
+ COL_B8_BASE);
+ priv->col_shift[13] = (((addrmap[3] >> 24) &
+ COL_MAX_VAL_MASK) == COL_MAX_VAL_MASK) ? 0 :
+ (((addrmap[3] >> 24) & COL_MAX_VAL_MASK) +
+ COL_B9_BASE);
+ }
+ }
+
+ if (width) {
+ for (index = 9; index > width; index--) {
+ priv->col_shift[index] = priv->col_shift[index - width];
+ priv->col_shift[index - width] = 0;
+ }
+ }
+
+}
+
+static void setup_bank_address_map(struct synps_edac_priv *priv, u32 *addrmap)
+{
+ priv->bank_shift[0] = (addrmap[1] & BANK_MAX_VAL_MASK) + BANK_B0_BASE;
+ priv->bank_shift[1] = ((addrmap[1] >> 8) &
+ BANK_MAX_VAL_MASK) + BANK_B1_BASE;
+ priv->bank_shift[2] = (((addrmap[1] >> 16) &
+ BANK_MAX_VAL_MASK) == BANK_MAX_VAL_MASK) ? 0 :
+ (((addrmap[1] >> 16) & BANK_MAX_VAL_MASK) +
+ BANK_B2_BASE);
+
+}
+
+static void setup_bg_address_map(struct synps_edac_priv *priv, u32 *addrmap)
+{
+ priv->bankgrp_shift[0] = (addrmap[8] &
+ BANKGRP_MAX_VAL_MASK) + BANKGRP_B0_BASE;
+ priv->bankgrp_shift[1] = (((addrmap[8] >> 8) & BANKGRP_MAX_VAL_MASK) ==
+ BANKGRP_MAX_VAL_MASK) ? 0 : (((addrmap[8] >> 8)
+ & BANKGRP_MAX_VAL_MASK) + BANKGRP_B1_BASE);
+
+}
+
+static void setup_rank_address_map(struct synps_edac_priv *priv, u32 *addrmap)
+{
+ priv->rank_shift[0] = ((addrmap[0] & RANK_MAX_VAL_MASK) ==
+ RANK_MAX_VAL_MASK) ? 0 : ((addrmap[0] &
+ RANK_MAX_VAL_MASK) + RANK_B0_BASE);
+}
+
+/**
+ * setup_address_map - Set Address Map by querying ADDRMAP registers.
+ * @priv: DDR memory controller private instance data.
+ *
+ * Set Address Map by querying ADDRMAP registers.
+ *
+ * Return: none.
+ */
+static void setup_address_map(struct synps_edac_priv *priv)
+{
+ u32 addrmap[12];
+ int index;
+
+ for (index = 0; index < 12; index++) {
+ u32 addrmap_offset;
+
+ addrmap_offset = ECC_ADDRMAP0_OFFSET + (index * 4);
+ addrmap[index] = readl(priv->baseaddr + addrmap_offset);
+ }
+
+ setup_row_address_map(priv, addrmap);
+
+ setup_column_address_map(priv, addrmap);
+
+ setup_bank_address_map(priv, addrmap);
+
+ setup_bg_address_map(priv, addrmap);
+
+ setup_rank_address_map(priv, addrmap);
+}
+#endif /* CONFIG_EDAC_DEBUG */
+
+/**
+ * mc_probe - Check controller and bind driver.
+ * @pdev: platform device.
+ *
+ * Probe a specific controller instance for binding with the driver.
+ *
+ * Return: 0 if the controller instance was successfully bound to the
+ * driver; otherwise, < 0 on error.
+ */
+static int mc_probe(struct platform_device *pdev)
+{
+ const struct synps_platform_data *p_data;
+ struct edac_mc_layer layers[2];
+ struct synps_edac_priv *priv;
+ struct mem_ctl_info *mci;
+ void __iomem *baseaddr;
+ struct resource *res;
+ int rc;
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ baseaddr = devm_ioremap_resource(&pdev->dev, res);
+ if (IS_ERR(baseaddr))
+ return PTR_ERR(baseaddr);
+
+ p_data = of_device_get_match_data(&pdev->dev);
+ if (!p_data)
+ return -ENODEV;
+
+ if (!p_data->get_ecc_state(baseaddr)) {
+ edac_printk(KERN_INFO, EDAC_MC, "ECC not enabled\n");
+ return -ENXIO;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = SYNPS_EDAC_NR_CSROWS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = SYNPS_EDAC_NR_CHANS;
+ layers[1].is_virt_csrow = false;
+
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+ sizeof(struct synps_edac_priv));
+ if (!mci) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Failed memory allocation for mc instance\n");
+ return -ENOMEM;
+ }
+
+ priv = mci->pvt_info;
+ priv->baseaddr = baseaddr;
+ priv->p_data = p_data;
+
+ mc_init(mci, pdev);
+
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT) {
+ rc = setup_irq(mci, pdev);
+ if (rc)
+ goto free_edac_mc;
+ }
+
+ rc = edac_mc_add_mc(mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Failed to register with EDAC core\n");
+ goto free_edac_mc;
+ }
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (priv->p_data->quirks & DDR_ECC_DATA_POISON_SUPPORT) {
+ rc = edac_create_sysfs_attributes(mci);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MC,
+ "Failed to create sysfs entries\n");
+ goto free_edac_mc;
+ }
+ }
+
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT)
+ setup_address_map(priv);
+#endif
+
+ /*
+ * Start capturing the correctable and uncorrectable errors. A write of
+ * 0 starts the counters.
+ */
+ if (!(priv->p_data->quirks & DDR_ECC_INTR_SUPPORT))
+ writel(0x0, baseaddr + ECC_CTRL_OFST);
+
+ return rc;
+
+free_edac_mc:
+ edac_mc_free(mci);
+
+ return rc;
+}
+
+/**
+ * mc_remove - Unbind driver from controller.
+ * @pdev: Platform device.
+ *
+ * Return: Unconditionally 0
+ */
+static int mc_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+ struct synps_edac_priv *priv = mci->pvt_info;
+
+ if (priv->p_data->quirks & DDR_ECC_INTR_SUPPORT)
+ disable_intr(priv);
+
+#ifdef CONFIG_EDAC_DEBUG
+ if (priv->p_data->quirks & DDR_ECC_DATA_POISON_SUPPORT)
+ edac_remove_sysfs_attributes(mci);
+#endif
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+static struct platform_driver synps_edac_mc_driver = {
+ .driver = {
+ .name = "synopsys-edac",
+ .of_match_table = synps_edac_match,
+ },
+ .probe = mc_probe,
+ .remove = mc_remove,
+};
+
+module_platform_driver(synps_edac_mc_driver);
+
+MODULE_AUTHOR("Xilinx Inc");
+MODULE_DESCRIPTION("Synopsys DDR ECC driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/thunderx_edac.c b/drivers/edac/thunderx_edac.c
new file mode 100644
index 0000000000..90d46e5c4f
--- /dev/null
+++ b/drivers/edac/thunderx_edac.c
@@ -0,0 +1,2154 @@
+/*
+ * Cavium ThunderX memory controller kernel module
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright Cavium, Inc. (C) 2015-2017. All rights reserved.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/string.h>
+#include <linux/stop_machine.h>
+#include <linux/delay.h>
+#include <linux/sizes.h>
+#include <linux/atomic.h>
+#include <linux/bitfield.h>
+#include <linux/circ_buf.h>
+
+#include <asm/page.h>
+
+#include "edac_module.h"
+
+#define phys_to_pfn(phys) (PFN_DOWN(phys))
+
+#define THUNDERX_NODE GENMASK(45, 44)
+
+enum {
+ ERR_CORRECTED = 1,
+ ERR_UNCORRECTED = 2,
+ ERR_UNKNOWN = 3,
+};
+
+#define MAX_SYNDROME_REGS 4
+
+struct error_syndrome {
+ u64 reg[MAX_SYNDROME_REGS];
+};
+
+struct error_descr {
+ int type;
+ u64 mask;
+ char *descr;
+};
+
+static void decode_register(char *str, size_t size,
+ const struct error_descr *descr,
+ const uint64_t reg)
+{
+ int ret = 0;
+
+ while (descr->type && descr->mask && descr->descr) {
+ if (reg & descr->mask) {
+ ret = snprintf(str, size, "\n\t%s, %s",
+ descr->type == ERR_CORRECTED ?
+ "Corrected" : "Uncorrected",
+ descr->descr);
+ str += ret;
+ size -= ret;
+ }
+ descr++;
+ }
+}
+
+static unsigned long get_bits(unsigned long data, int pos, int width)
+{
+ return (data >> pos) & ((1 << width) - 1);
+}
+
+#define L2C_CTL 0x87E080800000
+#define L2C_CTL_DISIDXALIAS BIT(0)
+
+#define PCI_DEVICE_ID_THUNDER_LMC 0xa022
+
+#define LMC_FADR 0x20
+#define LMC_FADR_FDIMM(x) ((x >> 37) & 0x1)
+#define LMC_FADR_FBUNK(x) ((x >> 36) & 0x1)
+#define LMC_FADR_FBANK(x) ((x >> 32) & 0xf)
+#define LMC_FADR_FROW(x) ((x >> 14) & 0xffff)
+#define LMC_FADR_FCOL(x) ((x >> 0) & 0x1fff)
+
+#define LMC_NXM_FADR 0x28
+#define LMC_ECC_SYND 0x38
+
+#define LMC_ECC_PARITY_TEST 0x108
+
+#define LMC_INT_W1S 0x150
+
+#define LMC_INT_ENA_W1C 0x158
+#define LMC_INT_ENA_W1S 0x160
+
+#define LMC_CONFIG 0x188
+
+#define LMC_CONFIG_BG2 BIT(62)
+#define LMC_CONFIG_RANK_ENA BIT(42)
+#define LMC_CONFIG_PBANK_LSB(x) (((x) >> 5) & 0xF)
+#define LMC_CONFIG_ROW_LSB(x) (((x) >> 2) & 0x7)
+
+#define LMC_CONTROL 0x190
+#define LMC_CONTROL_XOR_BANK BIT(16)
+
+#define LMC_INT 0x1F0
+
+#define LMC_INT_DDR_ERR BIT(11)
+#define LMC_INT_DED_ERR (0xFUL << 5)
+#define LMC_INT_SEC_ERR (0xFUL << 1)
+#define LMC_INT_NXM_WR_MASK BIT(0)
+
+#define LMC_DDR_PLL_CTL 0x258
+#define LMC_DDR_PLL_CTL_DDR4 BIT(29)
+
+#define LMC_FADR_SCRAMBLED 0x330
+
+#define LMC_INT_UE (LMC_INT_DDR_ERR | LMC_INT_DED_ERR | \
+ LMC_INT_NXM_WR_MASK)
+
+#define LMC_INT_CE (LMC_INT_SEC_ERR)
+
+static const struct error_descr lmc_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = LMC_INT_SEC_ERR,
+ .descr = "Single-bit ECC error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = LMC_INT_DDR_ERR,
+ .descr = "DDR chip error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = LMC_INT_DED_ERR,
+ .descr = "Double-bit ECC error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = LMC_INT_NXM_WR_MASK,
+ .descr = "Non-existent memory write",
+ },
+ {0, 0, NULL},
+};
+
+#define LMC_INT_EN_DDR_ERROR_ALERT_ENA BIT(5)
+#define LMC_INT_EN_DLCRAM_DED_ERR BIT(4)
+#define LMC_INT_EN_DLCRAM_SEC_ERR BIT(3)
+#define LMC_INT_INTR_DED_ENA BIT(2)
+#define LMC_INT_INTR_SEC_ENA BIT(1)
+#define LMC_INT_INTR_NXM_WR_ENA BIT(0)
+
+#define LMC_INT_ENA_ALL GENMASK(5, 0)
+
+#define LMC_DDR_PLL_CTL 0x258
+#define LMC_DDR_PLL_CTL_DDR4 BIT(29)
+
+#define LMC_CONTROL 0x190
+#define LMC_CONTROL_RDIMM BIT(0)
+
+#define LMC_SCRAM_FADR 0x330
+
+#define LMC_CHAR_MASK0 0x228
+#define LMC_CHAR_MASK2 0x238
+
+#define RING_ENTRIES 8
+
+struct debugfs_entry {
+ const char *name;
+ umode_t mode;
+ const struct file_operations fops;
+};
+
+struct lmc_err_ctx {
+ u64 reg_int;
+ u64 reg_fadr;
+ u64 reg_nxm_fadr;
+ u64 reg_scram_fadr;
+ u64 reg_ecc_synd;
+};
+
+struct thunderx_lmc {
+ void __iomem *regs;
+ struct pci_dev *pdev;
+ struct msix_entry msix_ent;
+
+ atomic_t ecc_int;
+
+ u64 mask0;
+ u64 mask2;
+ u64 parity_test;
+ u64 node;
+
+ int xbits;
+ int bank_width;
+ int pbank_lsb;
+ int dimm_lsb;
+ int rank_lsb;
+ int bank_lsb;
+ int row_lsb;
+ int col_hi_lsb;
+
+ int xor_bank;
+ int l2c_alias;
+
+ struct page *mem;
+
+ struct lmc_err_ctx err_ctx[RING_ENTRIES];
+ unsigned long ring_head;
+ unsigned long ring_tail;
+};
+
+#define ring_pos(pos, size) ((pos) & (size - 1))
+
+#define DEBUGFS_STRUCT(_name, _mode, _write, _read) \
+static struct debugfs_entry debugfs_##_name = { \
+ .name = __stringify(_name), \
+ .mode = VERIFY_OCTAL_PERMISSIONS(_mode), \
+ .fops = { \
+ .open = simple_open, \
+ .write = _write, \
+ .read = _read, \
+ .llseek = generic_file_llseek, \
+ }, \
+}
+
+#define DEBUGFS_FIELD_ATTR(_type, _field) \
+static ssize_t thunderx_##_type##_##_field##_read(struct file *file, \
+ char __user *data, \
+ size_t count, loff_t *ppos) \
+{ \
+ struct thunderx_##_type *pdata = file->private_data; \
+ char buf[20]; \
+ \
+ snprintf(buf, count, "0x%016llx", pdata->_field); \
+ return simple_read_from_buffer(data, count, ppos, \
+ buf, sizeof(buf)); \
+} \
+ \
+static ssize_t thunderx_##_type##_##_field##_write(struct file *file, \
+ const char __user *data, \
+ size_t count, loff_t *ppos) \
+{ \
+ struct thunderx_##_type *pdata = file->private_data; \
+ int res; \
+ \
+ res = kstrtoull_from_user(data, count, 0, &pdata->_field); \
+ \
+ return res ? res : count; \
+} \
+ \
+DEBUGFS_STRUCT(_field, 0600, \
+ thunderx_##_type##_##_field##_write, \
+ thunderx_##_type##_##_field##_read) \
+
+#define DEBUGFS_REG_ATTR(_type, _name, _reg) \
+static ssize_t thunderx_##_type##_##_name##_read(struct file *file, \
+ char __user *data, \
+ size_t count, loff_t *ppos) \
+{ \
+ struct thunderx_##_type *pdata = file->private_data; \
+ char buf[20]; \
+ \
+ sprintf(buf, "0x%016llx", readq(pdata->regs + _reg)); \
+ return simple_read_from_buffer(data, count, ppos, \
+ buf, sizeof(buf)); \
+} \
+ \
+static ssize_t thunderx_##_type##_##_name##_write(struct file *file, \
+ const char __user *data, \
+ size_t count, loff_t *ppos) \
+{ \
+ struct thunderx_##_type *pdata = file->private_data; \
+ u64 val; \
+ int res; \
+ \
+ res = kstrtoull_from_user(data, count, 0, &val); \
+ \
+ if (!res) { \
+ writeq(val, pdata->regs + _reg); \
+ res = count; \
+ } \
+ \
+ return res; \
+} \
+ \
+DEBUGFS_STRUCT(_name, 0600, \
+ thunderx_##_type##_##_name##_write, \
+ thunderx_##_type##_##_name##_read)
+
+#define LMC_DEBUGFS_ENT(_field) DEBUGFS_FIELD_ATTR(lmc, _field)
+
+/*
+ * To get an ECC error injected, the following steps are needed:
+ * - Setup the ECC injection by writing the appropriate parameters:
+ * echo <bit mask value> > /sys/kernel/debug/<device number>/ecc_mask0
+ * echo <bit mask value> > /sys/kernel/debug/<device number>/ecc_mask2
+ * echo 0x802 > /sys/kernel/debug/<device number>/ecc_parity_test
+ * - Do the actual injection:
+ * echo 1 > /sys/kernel/debug/<device number>/inject_ecc
+ */
+static ssize_t thunderx_lmc_inject_int_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct thunderx_lmc *lmc = file->private_data;
+ u64 val;
+ int res;
+
+ res = kstrtoull_from_user(data, count, 0, &val);
+
+ if (!res) {
+ /* Trigger the interrupt */
+ writeq(val, lmc->regs + LMC_INT_W1S);
+ res = count;
+ }
+
+ return res;
+}
+
+static ssize_t thunderx_lmc_int_read(struct file *file,
+ char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct thunderx_lmc *lmc = file->private_data;
+ char buf[20];
+ u64 lmc_int = readq(lmc->regs + LMC_INT);
+
+ snprintf(buf, sizeof(buf), "0x%016llx", lmc_int);
+ return simple_read_from_buffer(data, count, ppos, buf, sizeof(buf));
+}
+
+#define TEST_PATTERN 0xa5
+
+static int inject_ecc_fn(void *arg)
+{
+ struct thunderx_lmc *lmc = arg;
+ uintptr_t addr, phys;
+ unsigned int cline_size = cache_line_size();
+ const unsigned int lines = PAGE_SIZE / cline_size;
+ unsigned int i, cl_idx;
+
+ addr = (uintptr_t)page_address(lmc->mem);
+ phys = (uintptr_t)page_to_phys(lmc->mem);
+
+ cl_idx = (phys & 0x7f) >> 4;
+ lmc->parity_test &= ~(7ULL << 8);
+ lmc->parity_test |= (cl_idx << 8);
+
+ writeq(lmc->mask0, lmc->regs + LMC_CHAR_MASK0);
+ writeq(lmc->mask2, lmc->regs + LMC_CHAR_MASK2);
+ writeq(lmc->parity_test, lmc->regs + LMC_ECC_PARITY_TEST);
+
+ readq(lmc->regs + LMC_CHAR_MASK0);
+ readq(lmc->regs + LMC_CHAR_MASK2);
+ readq(lmc->regs + LMC_ECC_PARITY_TEST);
+
+ for (i = 0; i < lines; i++) {
+ memset((void *)addr, TEST_PATTERN, cline_size);
+ barrier();
+
+ /*
+ * Flush L1 cachelines to the PoC (L2).
+ * This will cause cacheline eviction to the L2.
+ */
+ asm volatile("dc civac, %0\n"
+ "dsb sy\n"
+ : : "r"(addr + i * cline_size));
+ }
+
+ for (i = 0; i < lines; i++) {
+ /*
+ * Flush L2 cachelines to the DRAM.
+ * This will cause cacheline eviction to the DRAM
+ * and ECC corruption according to the masks set.
+ */
+ __asm__ volatile("sys #0,c11,C1,#2, %0\n"
+ : : "r"(phys + i * cline_size));
+ }
+
+ for (i = 0; i < lines; i++) {
+ /*
+ * Invalidate L2 cachelines.
+ * The subsequent load will cause cacheline fetch
+ * from the DRAM and an error interrupt
+ */
+ __asm__ volatile("sys #0,c11,C1,#1, %0"
+ : : "r"(phys + i * cline_size));
+ }
+
+ for (i = 0; i < lines; i++) {
+ /*
+ * Invalidate L1 cachelines.
+ * The subsequent load will cause cacheline fetch
+ * from the L2 and/or DRAM
+ */
+ asm volatile("dc ivac, %0\n"
+ "dsb sy\n"
+ : : "r"(addr + i * cline_size));
+ }
+
+ return 0;
+}
+
+static ssize_t thunderx_lmc_inject_ecc_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct thunderx_lmc *lmc = file->private_data;
+ unsigned int cline_size = cache_line_size();
+ u8 *tmp;
+ void __iomem *addr;
+ unsigned int offs, timeout = 100000;
+
+ atomic_set(&lmc->ecc_int, 0);
+
+ lmc->mem = alloc_pages_node(lmc->node, GFP_KERNEL, 0);
+ if (!lmc->mem)
+ return -ENOMEM;
+
+ tmp = kmalloc(cline_size, GFP_KERNEL);
+ if (!tmp) {
+ __free_pages(lmc->mem, 0);
+ return -ENOMEM;
+ }
+
+ addr = page_address(lmc->mem);
+
+ while (!atomic_read(&lmc->ecc_int) && timeout--) {
+ stop_machine(inject_ecc_fn, lmc, NULL);
+
+ for (offs = 0; offs < PAGE_SIZE; offs += cline_size) {
+ /*
+ * Do a load from the previously rigged location
+ * This should generate an error interrupt.
+ */
+ memcpy(tmp, addr + offs, cline_size);
+ asm volatile("dsb ld\n");
+ }
+ }
+
+ kfree(tmp);
+ __free_pages(lmc->mem, 0);
+
+ return count;
+}
+
+LMC_DEBUGFS_ENT(mask0);
+LMC_DEBUGFS_ENT(mask2);
+LMC_DEBUGFS_ENT(parity_test);
+
+DEBUGFS_STRUCT(inject_int, 0200, thunderx_lmc_inject_int_write, NULL);
+DEBUGFS_STRUCT(inject_ecc, 0200, thunderx_lmc_inject_ecc_write, NULL);
+DEBUGFS_STRUCT(int_w1c, 0400, NULL, thunderx_lmc_int_read);
+
+static struct debugfs_entry *lmc_dfs_ents[] = {
+ &debugfs_mask0,
+ &debugfs_mask2,
+ &debugfs_parity_test,
+ &debugfs_inject_ecc,
+ &debugfs_inject_int,
+ &debugfs_int_w1c,
+};
+
+static int thunderx_create_debugfs_nodes(struct dentry *parent,
+ struct debugfs_entry *attrs[],
+ void *data,
+ size_t num)
+{
+ int i;
+ struct dentry *ent;
+
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+ return 0;
+
+ if (!parent)
+ return -ENOENT;
+
+ for (i = 0; i < num; i++) {
+ ent = edac_debugfs_create_file(attrs[i]->name, attrs[i]->mode,
+ parent, data, &attrs[i]->fops);
+
+ if (IS_ERR(ent))
+ break;
+ }
+
+ return i;
+}
+
+static phys_addr_t thunderx_faddr_to_phys(u64 faddr, struct thunderx_lmc *lmc)
+{
+ phys_addr_t addr = 0;
+ int bank, xbits;
+
+ addr |= lmc->node << 40;
+ addr |= LMC_FADR_FDIMM(faddr) << lmc->dimm_lsb;
+ addr |= LMC_FADR_FBUNK(faddr) << lmc->rank_lsb;
+ addr |= LMC_FADR_FROW(faddr) << lmc->row_lsb;
+ addr |= (LMC_FADR_FCOL(faddr) >> 4) << lmc->col_hi_lsb;
+
+ bank = LMC_FADR_FBANK(faddr) << lmc->bank_lsb;
+
+ if (lmc->xor_bank)
+ bank ^= get_bits(addr, 12 + lmc->xbits, lmc->bank_width);
+
+ addr |= bank << lmc->bank_lsb;
+
+ xbits = PCI_FUNC(lmc->pdev->devfn);
+
+ if (lmc->l2c_alias)
+ xbits ^= get_bits(addr, 20, lmc->xbits) ^
+ get_bits(addr, 12, lmc->xbits);
+
+ addr |= xbits << 7;
+
+ return addr;
+}
+
+static unsigned int thunderx_get_num_lmcs(unsigned int node)
+{
+ unsigned int number = 0;
+ struct pci_dev *pdev = NULL;
+
+ do {
+ pdev = pci_get_device(PCI_VENDOR_ID_CAVIUM,
+ PCI_DEVICE_ID_THUNDER_LMC,
+ pdev);
+ if (pdev) {
+#ifdef CONFIG_NUMA
+ if (pdev->dev.numa_node == node)
+ number++;
+#else
+ number++;
+#endif
+ }
+ } while (pdev);
+
+ return number;
+}
+
+#define LMC_MESSAGE_SIZE 120
+#define LMC_OTHER_SIZE (50 * ARRAY_SIZE(lmc_errors))
+
+static irqreturn_t thunderx_lmc_err_isr(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct thunderx_lmc *lmc = mci->pvt_info;
+
+ unsigned long head = ring_pos(lmc->ring_head, ARRAY_SIZE(lmc->err_ctx));
+ struct lmc_err_ctx *ctx = &lmc->err_ctx[head];
+
+ writeq(0, lmc->regs + LMC_CHAR_MASK0);
+ writeq(0, lmc->regs + LMC_CHAR_MASK2);
+ writeq(0x2, lmc->regs + LMC_ECC_PARITY_TEST);
+
+ ctx->reg_int = readq(lmc->regs + LMC_INT);
+ ctx->reg_fadr = readq(lmc->regs + LMC_FADR);
+ ctx->reg_nxm_fadr = readq(lmc->regs + LMC_NXM_FADR);
+ ctx->reg_scram_fadr = readq(lmc->regs + LMC_SCRAM_FADR);
+ ctx->reg_ecc_synd = readq(lmc->regs + LMC_ECC_SYND);
+
+ lmc->ring_head++;
+
+ atomic_set(&lmc->ecc_int, 1);
+
+ /* Clear the interrupt */
+ writeq(ctx->reg_int, lmc->regs + LMC_INT);
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_lmc_threaded_isr(int irq, void *dev_id)
+{
+ struct mem_ctl_info *mci = dev_id;
+ struct thunderx_lmc *lmc = mci->pvt_info;
+ phys_addr_t phys_addr;
+
+ unsigned long tail;
+ struct lmc_err_ctx *ctx;
+
+ irqreturn_t ret = IRQ_NONE;
+
+ char *msg;
+ char *other;
+
+ msg = kmalloc(LMC_MESSAGE_SIZE, GFP_KERNEL);
+ other = kmalloc(LMC_OTHER_SIZE, GFP_KERNEL);
+
+ if (!msg || !other)
+ goto err_free;
+
+ while (CIRC_CNT(lmc->ring_head, lmc->ring_tail,
+ ARRAY_SIZE(lmc->err_ctx))) {
+ tail = ring_pos(lmc->ring_tail, ARRAY_SIZE(lmc->err_ctx));
+
+ ctx = &lmc->err_ctx[tail];
+
+ dev_dbg(&lmc->pdev->dev, "LMC_INT: %016llx\n",
+ ctx->reg_int);
+ dev_dbg(&lmc->pdev->dev, "LMC_FADR: %016llx\n",
+ ctx->reg_fadr);
+ dev_dbg(&lmc->pdev->dev, "LMC_NXM_FADR: %016llx\n",
+ ctx->reg_nxm_fadr);
+ dev_dbg(&lmc->pdev->dev, "LMC_SCRAM_FADR: %016llx\n",
+ ctx->reg_scram_fadr);
+ dev_dbg(&lmc->pdev->dev, "LMC_ECC_SYND: %016llx\n",
+ ctx->reg_ecc_synd);
+
+ snprintf(msg, LMC_MESSAGE_SIZE,
+ "DIMM %lld rank %lld bank %lld row %lld col %lld",
+ LMC_FADR_FDIMM(ctx->reg_scram_fadr),
+ LMC_FADR_FBUNK(ctx->reg_scram_fadr),
+ LMC_FADR_FBANK(ctx->reg_scram_fadr),
+ LMC_FADR_FROW(ctx->reg_scram_fadr),
+ LMC_FADR_FCOL(ctx->reg_scram_fadr));
+
+ decode_register(other, LMC_OTHER_SIZE, lmc_errors,
+ ctx->reg_int);
+
+ phys_addr = thunderx_faddr_to_phys(ctx->reg_fadr, lmc);
+
+ if (ctx->reg_int & LMC_INT_UE)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ phys_to_pfn(phys_addr),
+ offset_in_page(phys_addr),
+ 0, -1, -1, -1, msg, other);
+ else if (ctx->reg_int & LMC_INT_CE)
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ phys_to_pfn(phys_addr),
+ offset_in_page(phys_addr),
+ 0, -1, -1, -1, msg, other);
+
+ lmc->ring_tail++;
+ }
+
+ ret = IRQ_HANDLED;
+
+err_free:
+ kfree(msg);
+ kfree(other);
+
+ return ret;
+}
+
+static const struct pci_device_id thunderx_lmc_pci_tbl[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVICE_ID_THUNDER_LMC) },
+ { 0, },
+};
+
+static inline int pci_dev_to_mc_idx(struct pci_dev *pdev)
+{
+ int node = dev_to_node(&pdev->dev);
+ int ret = PCI_FUNC(pdev->devfn);
+
+ ret += max(node, 0) << 3;
+
+ return ret;
+}
+
+static int thunderx_lmc_probe(struct pci_dev *pdev,
+ const struct pci_device_id *id)
+{
+ struct thunderx_lmc *lmc;
+ struct edac_mc_layer layer;
+ struct mem_ctl_info *mci;
+ u64 lmc_control, lmc_ddr_pll_ctl, lmc_config;
+ int ret;
+ u64 lmc_int;
+ void *l2c_ioaddr;
+
+ layer.type = EDAC_MC_LAYER_SLOT;
+ layer.size = 2;
+ layer.is_virt_csrow = false;
+
+ ret = pcim_enable_device(pdev);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable PCI device: %d\n", ret);
+ return ret;
+ }
+
+ ret = pcim_iomap_regions(pdev, BIT(0), "thunderx_lmc");
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot map PCI resources: %d\n", ret);
+ return ret;
+ }
+
+ mci = edac_mc_alloc(pci_dev_to_mc_idx(pdev), 1, &layer,
+ sizeof(struct thunderx_lmc));
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ lmc = mci->pvt_info;
+
+ pci_set_drvdata(pdev, mci);
+
+ lmc->regs = pcim_iomap_table(pdev)[0];
+
+ lmc_control = readq(lmc->regs + LMC_CONTROL);
+ lmc_ddr_pll_ctl = readq(lmc->regs + LMC_DDR_PLL_CTL);
+ lmc_config = readq(lmc->regs + LMC_CONFIG);
+
+ if (lmc_control & LMC_CONTROL_RDIMM) {
+ mci->mtype_cap = FIELD_GET(LMC_DDR_PLL_CTL_DDR4,
+ lmc_ddr_pll_ctl) ?
+ MEM_RDDR4 : MEM_RDDR3;
+ } else {
+ mci->mtype_cap = FIELD_GET(LMC_DDR_PLL_CTL_DDR4,
+ lmc_ddr_pll_ctl) ?
+ MEM_DDR4 : MEM_DDR3;
+ }
+
+ mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+
+ mci->mod_name = "thunderx-lmc";
+ mci->ctl_name = "thunderx-lmc";
+ mci->dev_name = dev_name(&pdev->dev);
+ mci->scrub_mode = SCRUB_NONE;
+
+ lmc->pdev = pdev;
+ lmc->msix_ent.entry = 0;
+
+ lmc->ring_head = 0;
+ lmc->ring_tail = 0;
+
+ ret = pci_enable_msix_exact(pdev, &lmc->msix_ent, 1);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable interrupt: %d\n", ret);
+ goto err_free;
+ }
+
+ ret = devm_request_threaded_irq(&pdev->dev, lmc->msix_ent.vector,
+ thunderx_lmc_err_isr,
+ thunderx_lmc_threaded_isr, 0,
+ "[EDAC] ThunderX LMC", mci);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot set ISR: %d\n", ret);
+ goto err_free;
+ }
+
+ lmc->node = FIELD_GET(THUNDERX_NODE, pci_resource_start(pdev, 0));
+
+ lmc->xbits = thunderx_get_num_lmcs(lmc->node) >> 1;
+ lmc->bank_width = (FIELD_GET(LMC_DDR_PLL_CTL_DDR4, lmc_ddr_pll_ctl) &&
+ FIELD_GET(LMC_CONFIG_BG2, lmc_config)) ? 4 : 3;
+
+ lmc->pbank_lsb = (lmc_config >> 5) & 0xf;
+ lmc->dimm_lsb = 28 + lmc->pbank_lsb + lmc->xbits;
+ lmc->rank_lsb = lmc->dimm_lsb;
+ lmc->rank_lsb -= FIELD_GET(LMC_CONFIG_RANK_ENA, lmc_config) ? 1 : 0;
+ lmc->bank_lsb = 7 + lmc->xbits;
+ lmc->row_lsb = 14 + LMC_CONFIG_ROW_LSB(lmc_config) + lmc->xbits;
+
+ lmc->col_hi_lsb = lmc->bank_lsb + lmc->bank_width;
+
+ lmc->xor_bank = lmc_control & LMC_CONTROL_XOR_BANK;
+
+ l2c_ioaddr = ioremap(L2C_CTL | FIELD_PREP(THUNDERX_NODE, lmc->node), PAGE_SIZE);
+ if (!l2c_ioaddr) {
+ dev_err(&pdev->dev, "Cannot map L2C_CTL\n");
+ ret = -ENOMEM;
+ goto err_free;
+ }
+
+ lmc->l2c_alias = !(readq(l2c_ioaddr) & L2C_CTL_DISIDXALIAS);
+
+ iounmap(l2c_ioaddr);
+
+ ret = edac_mc_add_mc(mci);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot add the MC: %d\n", ret);
+ goto err_free;
+ }
+
+ lmc_int = readq(lmc->regs + LMC_INT);
+ writeq(lmc_int, lmc->regs + LMC_INT);
+
+ writeq(LMC_INT_ENA_ALL, lmc->regs + LMC_INT_ENA_W1S);
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG)) {
+ ret = thunderx_create_debugfs_nodes(mci->debugfs,
+ lmc_dfs_ents,
+ lmc,
+ ARRAY_SIZE(lmc_dfs_ents));
+
+ if (ret != ARRAY_SIZE(lmc_dfs_ents)) {
+ dev_warn(&pdev->dev, "Error creating debugfs entries: %d%s\n",
+ ret, ret >= 0 ? " created" : "");
+ }
+ }
+
+ return 0;
+
+err_free:
+ pci_set_drvdata(pdev, NULL);
+ edac_mc_free(mci);
+
+ return ret;
+}
+
+static void thunderx_lmc_remove(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci = pci_get_drvdata(pdev);
+ struct thunderx_lmc *lmc = mci->pvt_info;
+
+ writeq(LMC_INT_ENA_ALL, lmc->regs + LMC_INT_ENA_W1C);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+}
+
+MODULE_DEVICE_TABLE(pci, thunderx_lmc_pci_tbl);
+
+static struct pci_driver thunderx_lmc_driver = {
+ .name = "thunderx_lmc_edac",
+ .probe = thunderx_lmc_probe,
+ .remove = thunderx_lmc_remove,
+ .id_table = thunderx_lmc_pci_tbl,
+};
+
+/*---------------------- OCX driver ---------------------------------*/
+
+#define PCI_DEVICE_ID_THUNDER_OCX 0xa013
+
+#define OCX_LINK_INTS 3
+#define OCX_INTS (OCX_LINK_INTS + 1)
+#define OCX_RX_LANES 24
+#define OCX_RX_LANE_STATS 15
+
+#define OCX_COM_INT 0x100
+#define OCX_COM_INT_W1S 0x108
+#define OCX_COM_INT_ENA_W1S 0x110
+#define OCX_COM_INT_ENA_W1C 0x118
+
+#define OCX_COM_IO_BADID BIT(54)
+#define OCX_COM_MEM_BADID BIT(53)
+#define OCX_COM_COPR_BADID BIT(52)
+#define OCX_COM_WIN_REQ_BADID BIT(51)
+#define OCX_COM_WIN_REQ_TOUT BIT(50)
+#define OCX_COM_RX_LANE GENMASK(23, 0)
+
+#define OCX_COM_INT_CE (OCX_COM_IO_BADID | \
+ OCX_COM_MEM_BADID | \
+ OCX_COM_COPR_BADID | \
+ OCX_COM_WIN_REQ_BADID | \
+ OCX_COM_WIN_REQ_TOUT)
+
+static const struct error_descr ocx_com_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_IO_BADID,
+ .descr = "Invalid IO transaction node ID",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_MEM_BADID,
+ .descr = "Invalid memory transaction node ID",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_COPR_BADID,
+ .descr = "Invalid coprocessor transaction node ID",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_WIN_REQ_BADID,
+ .descr = "Invalid SLI transaction node ID",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_WIN_REQ_TOUT,
+ .descr = "Window/core request timeout",
+ },
+ {0, 0, NULL},
+};
+
+#define OCX_COM_LINKX_INT(x) (0x120 + (x) * 8)
+#define OCX_COM_LINKX_INT_W1S(x) (0x140 + (x) * 8)
+#define OCX_COM_LINKX_INT_ENA_W1S(x) (0x160 + (x) * 8)
+#define OCX_COM_LINKX_INT_ENA_W1C(x) (0x180 + (x) * 8)
+
+#define OCX_COM_LINK_BAD_WORD BIT(13)
+#define OCX_COM_LINK_ALIGN_FAIL BIT(12)
+#define OCX_COM_LINK_ALIGN_DONE BIT(11)
+#define OCX_COM_LINK_UP BIT(10)
+#define OCX_COM_LINK_STOP BIT(9)
+#define OCX_COM_LINK_BLK_ERR BIT(8)
+#define OCX_COM_LINK_REINIT BIT(7)
+#define OCX_COM_LINK_LNK_DATA BIT(6)
+#define OCX_COM_LINK_RXFIFO_DBE BIT(5)
+#define OCX_COM_LINK_RXFIFO_SBE BIT(4)
+#define OCX_COM_LINK_TXFIFO_DBE BIT(3)
+#define OCX_COM_LINK_TXFIFO_SBE BIT(2)
+#define OCX_COM_LINK_REPLAY_DBE BIT(1)
+#define OCX_COM_LINK_REPLAY_SBE BIT(0)
+
+static const struct error_descr ocx_com_link_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_REPLAY_SBE,
+ .descr = "Replay buffer single-bit error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_TXFIFO_SBE,
+ .descr = "TX FIFO single-bit error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_RXFIFO_SBE,
+ .descr = "RX FIFO single-bit error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_BLK_ERR,
+ .descr = "Block code error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_ALIGN_FAIL,
+ .descr = "Link alignment failure",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_COM_LINK_BAD_WORD,
+ .descr = "Bad code word",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = OCX_COM_LINK_REPLAY_DBE,
+ .descr = "Replay buffer double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = OCX_COM_LINK_TXFIFO_DBE,
+ .descr = "TX FIFO double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = OCX_COM_LINK_RXFIFO_DBE,
+ .descr = "RX FIFO double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = OCX_COM_LINK_STOP,
+ .descr = "Link stopped",
+ },
+ {0, 0, NULL},
+};
+
+#define OCX_COM_LINK_INT_UE (OCX_COM_LINK_REPLAY_DBE | \
+ OCX_COM_LINK_TXFIFO_DBE | \
+ OCX_COM_LINK_RXFIFO_DBE | \
+ OCX_COM_LINK_STOP)
+
+#define OCX_COM_LINK_INT_CE (OCX_COM_LINK_REPLAY_SBE | \
+ OCX_COM_LINK_TXFIFO_SBE | \
+ OCX_COM_LINK_RXFIFO_SBE | \
+ OCX_COM_LINK_BLK_ERR | \
+ OCX_COM_LINK_ALIGN_FAIL | \
+ OCX_COM_LINK_BAD_WORD)
+
+#define OCX_LNE_INT(x) (0x8018 + (x) * 0x100)
+#define OCX_LNE_INT_EN(x) (0x8020 + (x) * 0x100)
+#define OCX_LNE_BAD_CNT(x) (0x8028 + (x) * 0x100)
+#define OCX_LNE_CFG(x) (0x8000 + (x) * 0x100)
+#define OCX_LNE_STAT(x, y) (0x8040 + (x) * 0x100 + (y) * 8)
+
+#define OCX_LNE_CFG_RX_BDRY_LOCK_DIS BIT(8)
+#define OCX_LNE_CFG_RX_STAT_WRAP_DIS BIT(2)
+#define OCX_LNE_CFG_RX_STAT_RDCLR BIT(1)
+#define OCX_LNE_CFG_RX_STAT_ENA BIT(0)
+
+
+#define OCX_LANE_BAD_64B67B BIT(8)
+#define OCX_LANE_DSKEW_FIFO_OVFL BIT(5)
+#define OCX_LANE_SCRM_SYNC_LOSS BIT(4)
+#define OCX_LANE_UKWN_CNTL_WORD BIT(3)
+#define OCX_LANE_CRC32_ERR BIT(2)
+#define OCX_LANE_BDRY_SYNC_LOSS BIT(1)
+#define OCX_LANE_SERDES_LOCK_LOSS BIT(0)
+
+#define OCX_COM_LANE_INT_UE (0)
+#define OCX_COM_LANE_INT_CE (OCX_LANE_SERDES_LOCK_LOSS | \
+ OCX_LANE_BDRY_SYNC_LOSS | \
+ OCX_LANE_CRC32_ERR | \
+ OCX_LANE_UKWN_CNTL_WORD | \
+ OCX_LANE_SCRM_SYNC_LOSS | \
+ OCX_LANE_DSKEW_FIFO_OVFL | \
+ OCX_LANE_BAD_64B67B)
+
+static const struct error_descr ocx_lane_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_SERDES_LOCK_LOSS,
+ .descr = "RX SerDes lock lost",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_BDRY_SYNC_LOSS,
+ .descr = "RX word boundary lost",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_CRC32_ERR,
+ .descr = "CRC32 error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_UKWN_CNTL_WORD,
+ .descr = "Unknown control word",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_SCRM_SYNC_LOSS,
+ .descr = "Scrambler synchronization lost",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_DSKEW_FIFO_OVFL,
+ .descr = "RX deskew FIFO overflow",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = OCX_LANE_BAD_64B67B,
+ .descr = "Bad 64B/67B codeword",
+ },
+ {0, 0, NULL},
+};
+
+#define OCX_LNE_INT_ENA_ALL (GENMASK(9, 8) | GENMASK(6, 0))
+#define OCX_COM_INT_ENA_ALL (GENMASK(54, 50) | GENMASK(23, 0))
+#define OCX_COM_LINKX_INT_ENA_ALL (GENMASK(13, 12) | \
+ GENMASK(9, 7) | GENMASK(5, 0))
+
+#define OCX_TLKX_ECC_CTL(x) (0x10018 + (x) * 0x2000)
+#define OCX_RLKX_ECC_CTL(x) (0x18018 + (x) * 0x2000)
+
+struct ocx_com_err_ctx {
+ u64 reg_com_int;
+ u64 reg_lane_int[OCX_RX_LANES];
+ u64 reg_lane_stat11[OCX_RX_LANES];
+};
+
+struct ocx_link_err_ctx {
+ u64 reg_com_link_int;
+ int link;
+};
+
+struct thunderx_ocx {
+ void __iomem *regs;
+ int com_link;
+ struct pci_dev *pdev;
+ struct edac_device_ctl_info *edac_dev;
+
+ struct dentry *debugfs;
+ struct msix_entry msix_ent[OCX_INTS];
+
+ struct ocx_com_err_ctx com_err_ctx[RING_ENTRIES];
+ struct ocx_link_err_ctx link_err_ctx[RING_ENTRIES];
+
+ unsigned long com_ring_head;
+ unsigned long com_ring_tail;
+
+ unsigned long link_ring_head;
+ unsigned long link_ring_tail;
+};
+
+#define OCX_MESSAGE_SIZE SZ_1K
+#define OCX_OTHER_SIZE (50 * ARRAY_SIZE(ocx_com_link_errors))
+
+/* This handler is threaded */
+static irqreturn_t thunderx_ocx_com_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_ocx *ocx = container_of(msix, struct thunderx_ocx,
+ msix_ent[msix->entry]);
+
+ int lane;
+ unsigned long head = ring_pos(ocx->com_ring_head,
+ ARRAY_SIZE(ocx->com_err_ctx));
+ struct ocx_com_err_ctx *ctx = &ocx->com_err_ctx[head];
+
+ ctx->reg_com_int = readq(ocx->regs + OCX_COM_INT);
+
+ for (lane = 0; lane < OCX_RX_LANES; lane++) {
+ ctx->reg_lane_int[lane] =
+ readq(ocx->regs + OCX_LNE_INT(lane));
+ ctx->reg_lane_stat11[lane] =
+ readq(ocx->regs + OCX_LNE_STAT(lane, 11));
+
+ writeq(ctx->reg_lane_int[lane], ocx->regs + OCX_LNE_INT(lane));
+ }
+
+ writeq(ctx->reg_com_int, ocx->regs + OCX_COM_INT);
+
+ ocx->com_ring_head++;
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_ocx_com_threaded_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_ocx *ocx = container_of(msix, struct thunderx_ocx,
+ msix_ent[msix->entry]);
+
+ irqreturn_t ret = IRQ_NONE;
+
+ unsigned long tail;
+ struct ocx_com_err_ctx *ctx;
+ int lane;
+ char *msg;
+ char *other;
+
+ msg = kmalloc(OCX_MESSAGE_SIZE, GFP_KERNEL);
+ other = kmalloc(OCX_OTHER_SIZE, GFP_KERNEL);
+
+ if (!msg || !other)
+ goto err_free;
+
+ while (CIRC_CNT(ocx->com_ring_head, ocx->com_ring_tail,
+ ARRAY_SIZE(ocx->com_err_ctx))) {
+ tail = ring_pos(ocx->com_ring_tail,
+ ARRAY_SIZE(ocx->com_err_ctx));
+ ctx = &ocx->com_err_ctx[tail];
+
+ snprintf(msg, OCX_MESSAGE_SIZE, "%s: OCX_COM_INT: %016llx",
+ ocx->edac_dev->ctl_name, ctx->reg_com_int);
+
+ decode_register(other, OCX_OTHER_SIZE,
+ ocx_com_errors, ctx->reg_com_int);
+
+ strlcat(msg, other, OCX_MESSAGE_SIZE);
+
+ for (lane = 0; lane < OCX_RX_LANES; lane++)
+ if (ctx->reg_com_int & BIT(lane)) {
+ snprintf(other, OCX_OTHER_SIZE,
+ "\n\tOCX_LNE_INT[%02d]: %016llx OCX_LNE_STAT11[%02d]: %016llx",
+ lane, ctx->reg_lane_int[lane],
+ lane, ctx->reg_lane_stat11[lane]);
+
+ strlcat(msg, other, OCX_MESSAGE_SIZE);
+
+ decode_register(other, OCX_OTHER_SIZE,
+ ocx_lane_errors,
+ ctx->reg_lane_int[lane]);
+ strlcat(msg, other, OCX_MESSAGE_SIZE);
+ }
+
+ if (ctx->reg_com_int & OCX_COM_INT_CE)
+ edac_device_handle_ce(ocx->edac_dev, 0, 0, msg);
+
+ ocx->com_ring_tail++;
+ }
+
+ ret = IRQ_HANDLED;
+
+err_free:
+ kfree(other);
+ kfree(msg);
+
+ return ret;
+}
+
+static irqreturn_t thunderx_ocx_lnk_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_ocx *ocx = container_of(msix, struct thunderx_ocx,
+ msix_ent[msix->entry]);
+ unsigned long head = ring_pos(ocx->link_ring_head,
+ ARRAY_SIZE(ocx->link_err_ctx));
+ struct ocx_link_err_ctx *ctx = &ocx->link_err_ctx[head];
+
+ ctx->link = msix->entry;
+ ctx->reg_com_link_int = readq(ocx->regs + OCX_COM_LINKX_INT(ctx->link));
+
+ writeq(ctx->reg_com_link_int, ocx->regs + OCX_COM_LINKX_INT(ctx->link));
+
+ ocx->link_ring_head++;
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_ocx_lnk_threaded_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_ocx *ocx = container_of(msix, struct thunderx_ocx,
+ msix_ent[msix->entry]);
+ irqreturn_t ret = IRQ_NONE;
+ unsigned long tail;
+ struct ocx_link_err_ctx *ctx;
+
+ char *msg;
+ char *other;
+
+ msg = kmalloc(OCX_MESSAGE_SIZE, GFP_KERNEL);
+ other = kmalloc(OCX_OTHER_SIZE, GFP_KERNEL);
+
+ if (!msg || !other)
+ goto err_free;
+
+ while (CIRC_CNT(ocx->link_ring_head, ocx->link_ring_tail,
+ ARRAY_SIZE(ocx->link_err_ctx))) {
+ tail = ring_pos(ocx->link_ring_head,
+ ARRAY_SIZE(ocx->link_err_ctx));
+
+ ctx = &ocx->link_err_ctx[tail];
+
+ snprintf(msg, OCX_MESSAGE_SIZE,
+ "%s: OCX_COM_LINK_INT[%d]: %016llx",
+ ocx->edac_dev->ctl_name,
+ ctx->link, ctx->reg_com_link_int);
+
+ decode_register(other, OCX_OTHER_SIZE,
+ ocx_com_link_errors, ctx->reg_com_link_int);
+
+ strlcat(msg, other, OCX_MESSAGE_SIZE);
+
+ if (ctx->reg_com_link_int & OCX_COM_LINK_INT_UE)
+ edac_device_handle_ue(ocx->edac_dev, 0, 0, msg);
+ else if (ctx->reg_com_link_int & OCX_COM_LINK_INT_CE)
+ edac_device_handle_ce(ocx->edac_dev, 0, 0, msg);
+
+ ocx->link_ring_tail++;
+ }
+
+ ret = IRQ_HANDLED;
+err_free:
+ kfree(other);
+ kfree(msg);
+
+ return ret;
+}
+
+#define OCX_DEBUGFS_ATTR(_name, _reg) DEBUGFS_REG_ATTR(ocx, _name, _reg)
+
+OCX_DEBUGFS_ATTR(tlk0_ecc_ctl, OCX_TLKX_ECC_CTL(0));
+OCX_DEBUGFS_ATTR(tlk1_ecc_ctl, OCX_TLKX_ECC_CTL(1));
+OCX_DEBUGFS_ATTR(tlk2_ecc_ctl, OCX_TLKX_ECC_CTL(2));
+
+OCX_DEBUGFS_ATTR(rlk0_ecc_ctl, OCX_RLKX_ECC_CTL(0));
+OCX_DEBUGFS_ATTR(rlk1_ecc_ctl, OCX_RLKX_ECC_CTL(1));
+OCX_DEBUGFS_ATTR(rlk2_ecc_ctl, OCX_RLKX_ECC_CTL(2));
+
+OCX_DEBUGFS_ATTR(com_link0_int, OCX_COM_LINKX_INT_W1S(0));
+OCX_DEBUGFS_ATTR(com_link1_int, OCX_COM_LINKX_INT_W1S(1));
+OCX_DEBUGFS_ATTR(com_link2_int, OCX_COM_LINKX_INT_W1S(2));
+
+OCX_DEBUGFS_ATTR(lne00_badcnt, OCX_LNE_BAD_CNT(0));
+OCX_DEBUGFS_ATTR(lne01_badcnt, OCX_LNE_BAD_CNT(1));
+OCX_DEBUGFS_ATTR(lne02_badcnt, OCX_LNE_BAD_CNT(2));
+OCX_DEBUGFS_ATTR(lne03_badcnt, OCX_LNE_BAD_CNT(3));
+OCX_DEBUGFS_ATTR(lne04_badcnt, OCX_LNE_BAD_CNT(4));
+OCX_DEBUGFS_ATTR(lne05_badcnt, OCX_LNE_BAD_CNT(5));
+OCX_DEBUGFS_ATTR(lne06_badcnt, OCX_LNE_BAD_CNT(6));
+OCX_DEBUGFS_ATTR(lne07_badcnt, OCX_LNE_BAD_CNT(7));
+
+OCX_DEBUGFS_ATTR(lne08_badcnt, OCX_LNE_BAD_CNT(8));
+OCX_DEBUGFS_ATTR(lne09_badcnt, OCX_LNE_BAD_CNT(9));
+OCX_DEBUGFS_ATTR(lne10_badcnt, OCX_LNE_BAD_CNT(10));
+OCX_DEBUGFS_ATTR(lne11_badcnt, OCX_LNE_BAD_CNT(11));
+OCX_DEBUGFS_ATTR(lne12_badcnt, OCX_LNE_BAD_CNT(12));
+OCX_DEBUGFS_ATTR(lne13_badcnt, OCX_LNE_BAD_CNT(13));
+OCX_DEBUGFS_ATTR(lne14_badcnt, OCX_LNE_BAD_CNT(14));
+OCX_DEBUGFS_ATTR(lne15_badcnt, OCX_LNE_BAD_CNT(15));
+
+OCX_DEBUGFS_ATTR(lne16_badcnt, OCX_LNE_BAD_CNT(16));
+OCX_DEBUGFS_ATTR(lne17_badcnt, OCX_LNE_BAD_CNT(17));
+OCX_DEBUGFS_ATTR(lne18_badcnt, OCX_LNE_BAD_CNT(18));
+OCX_DEBUGFS_ATTR(lne19_badcnt, OCX_LNE_BAD_CNT(19));
+OCX_DEBUGFS_ATTR(lne20_badcnt, OCX_LNE_BAD_CNT(20));
+OCX_DEBUGFS_ATTR(lne21_badcnt, OCX_LNE_BAD_CNT(21));
+OCX_DEBUGFS_ATTR(lne22_badcnt, OCX_LNE_BAD_CNT(22));
+OCX_DEBUGFS_ATTR(lne23_badcnt, OCX_LNE_BAD_CNT(23));
+
+OCX_DEBUGFS_ATTR(com_int, OCX_COM_INT_W1S);
+
+static struct debugfs_entry *ocx_dfs_ents[] = {
+ &debugfs_tlk0_ecc_ctl,
+ &debugfs_tlk1_ecc_ctl,
+ &debugfs_tlk2_ecc_ctl,
+
+ &debugfs_rlk0_ecc_ctl,
+ &debugfs_rlk1_ecc_ctl,
+ &debugfs_rlk2_ecc_ctl,
+
+ &debugfs_com_link0_int,
+ &debugfs_com_link1_int,
+ &debugfs_com_link2_int,
+
+ &debugfs_lne00_badcnt,
+ &debugfs_lne01_badcnt,
+ &debugfs_lne02_badcnt,
+ &debugfs_lne03_badcnt,
+ &debugfs_lne04_badcnt,
+ &debugfs_lne05_badcnt,
+ &debugfs_lne06_badcnt,
+ &debugfs_lne07_badcnt,
+ &debugfs_lne08_badcnt,
+ &debugfs_lne09_badcnt,
+ &debugfs_lne10_badcnt,
+ &debugfs_lne11_badcnt,
+ &debugfs_lne12_badcnt,
+ &debugfs_lne13_badcnt,
+ &debugfs_lne14_badcnt,
+ &debugfs_lne15_badcnt,
+ &debugfs_lne16_badcnt,
+ &debugfs_lne17_badcnt,
+ &debugfs_lne18_badcnt,
+ &debugfs_lne19_badcnt,
+ &debugfs_lne20_badcnt,
+ &debugfs_lne21_badcnt,
+ &debugfs_lne22_badcnt,
+ &debugfs_lne23_badcnt,
+
+ &debugfs_com_int,
+};
+
+static const struct pci_device_id thunderx_ocx_pci_tbl[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVICE_ID_THUNDER_OCX) },
+ { 0, },
+};
+
+static void thunderx_ocx_clearstats(struct thunderx_ocx *ocx)
+{
+ int lane, stat, cfg;
+
+ for (lane = 0; lane < OCX_RX_LANES; lane++) {
+ cfg = readq(ocx->regs + OCX_LNE_CFG(lane));
+ cfg |= OCX_LNE_CFG_RX_STAT_RDCLR;
+ cfg &= ~OCX_LNE_CFG_RX_STAT_ENA;
+ writeq(cfg, ocx->regs + OCX_LNE_CFG(lane));
+
+ for (stat = 0; stat < OCX_RX_LANE_STATS; stat++)
+ readq(ocx->regs + OCX_LNE_STAT(lane, stat));
+ }
+}
+
+static int thunderx_ocx_probe(struct pci_dev *pdev,
+ const struct pci_device_id *id)
+{
+ struct thunderx_ocx *ocx;
+ struct edac_device_ctl_info *edac_dev;
+ char name[32];
+ int idx;
+ int i;
+ int ret;
+ u64 reg;
+
+ ret = pcim_enable_device(pdev);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable PCI device: %d\n", ret);
+ return ret;
+ }
+
+ ret = pcim_iomap_regions(pdev, BIT(0), "thunderx_ocx");
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot map PCI resources: %d\n", ret);
+ return ret;
+ }
+
+ idx = edac_device_alloc_index();
+ snprintf(name, sizeof(name), "OCX%d", idx);
+ edac_dev = edac_device_alloc_ctl_info(sizeof(struct thunderx_ocx),
+ name, 1, "CCPI", 1,
+ 0, NULL, 0, idx);
+ if (!edac_dev) {
+ dev_err(&pdev->dev, "Cannot allocate EDAC device\n");
+ return -ENOMEM;
+ }
+ ocx = edac_dev->pvt_info;
+ ocx->edac_dev = edac_dev;
+ ocx->com_ring_head = 0;
+ ocx->com_ring_tail = 0;
+ ocx->link_ring_head = 0;
+ ocx->link_ring_tail = 0;
+
+ ocx->regs = pcim_iomap_table(pdev)[0];
+ if (!ocx->regs) {
+ dev_err(&pdev->dev, "Cannot map PCI resources\n");
+ ret = -ENODEV;
+ goto err_free;
+ }
+
+ ocx->pdev = pdev;
+
+ for (i = 0; i < OCX_INTS; i++) {
+ ocx->msix_ent[i].entry = i;
+ ocx->msix_ent[i].vector = 0;
+ }
+
+ ret = pci_enable_msix_exact(pdev, ocx->msix_ent, OCX_INTS);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable interrupt: %d\n", ret);
+ goto err_free;
+ }
+
+ for (i = 0; i < OCX_INTS; i++) {
+ ret = devm_request_threaded_irq(&pdev->dev,
+ ocx->msix_ent[i].vector,
+ (i == 3) ?
+ thunderx_ocx_com_isr :
+ thunderx_ocx_lnk_isr,
+ (i == 3) ?
+ thunderx_ocx_com_threaded_isr :
+ thunderx_ocx_lnk_threaded_isr,
+ 0, "[EDAC] ThunderX OCX",
+ &ocx->msix_ent[i]);
+ if (ret)
+ goto err_free;
+ }
+
+ edac_dev->dev = &pdev->dev;
+ edac_dev->dev_name = dev_name(&pdev->dev);
+ edac_dev->mod_name = "thunderx-ocx";
+ edac_dev->ctl_name = "thunderx-ocx";
+
+ ret = edac_device_add_device(edac_dev);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot add EDAC device: %d\n", ret);
+ goto err_free;
+ }
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG)) {
+ ocx->debugfs = edac_debugfs_create_dir(pdev->dev.kobj.name);
+
+ ret = thunderx_create_debugfs_nodes(ocx->debugfs,
+ ocx_dfs_ents,
+ ocx,
+ ARRAY_SIZE(ocx_dfs_ents));
+ if (ret != ARRAY_SIZE(ocx_dfs_ents)) {
+ dev_warn(&pdev->dev, "Error creating debugfs entries: %d%s\n",
+ ret, ret >= 0 ? " created" : "");
+ }
+ }
+
+ pci_set_drvdata(pdev, edac_dev);
+
+ thunderx_ocx_clearstats(ocx);
+
+ for (i = 0; i < OCX_RX_LANES; i++) {
+ writeq(OCX_LNE_INT_ENA_ALL,
+ ocx->regs + OCX_LNE_INT_EN(i));
+
+ reg = readq(ocx->regs + OCX_LNE_INT(i));
+ writeq(reg, ocx->regs + OCX_LNE_INT(i));
+
+ }
+
+ for (i = 0; i < OCX_LINK_INTS; i++) {
+ reg = readq(ocx->regs + OCX_COM_LINKX_INT(i));
+ writeq(reg, ocx->regs + OCX_COM_LINKX_INT(i));
+
+ writeq(OCX_COM_LINKX_INT_ENA_ALL,
+ ocx->regs + OCX_COM_LINKX_INT_ENA_W1S(i));
+ }
+
+ reg = readq(ocx->regs + OCX_COM_INT);
+ writeq(reg, ocx->regs + OCX_COM_INT);
+
+ writeq(OCX_COM_INT_ENA_ALL, ocx->regs + OCX_COM_INT_ENA_W1S);
+
+ return 0;
+err_free:
+ edac_device_free_ctl_info(edac_dev);
+
+ return ret;
+}
+
+static void thunderx_ocx_remove(struct pci_dev *pdev)
+{
+ struct edac_device_ctl_info *edac_dev = pci_get_drvdata(pdev);
+ struct thunderx_ocx *ocx = edac_dev->pvt_info;
+ int i;
+
+ writeq(OCX_COM_INT_ENA_ALL, ocx->regs + OCX_COM_INT_ENA_W1C);
+
+ for (i = 0; i < OCX_INTS; i++) {
+ writeq(OCX_COM_LINKX_INT_ENA_ALL,
+ ocx->regs + OCX_COM_LINKX_INT_ENA_W1C(i));
+ }
+
+ edac_debugfs_remove_recursive(ocx->debugfs);
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(edac_dev);
+}
+
+MODULE_DEVICE_TABLE(pci, thunderx_ocx_pci_tbl);
+
+static struct pci_driver thunderx_ocx_driver = {
+ .name = "thunderx_ocx_edac",
+ .probe = thunderx_ocx_probe,
+ .remove = thunderx_ocx_remove,
+ .id_table = thunderx_ocx_pci_tbl,
+};
+
+/*---------------------- L2C driver ---------------------------------*/
+
+#define PCI_DEVICE_ID_THUNDER_L2C_TAD 0xa02e
+#define PCI_DEVICE_ID_THUNDER_L2C_CBC 0xa02f
+#define PCI_DEVICE_ID_THUNDER_L2C_MCI 0xa030
+
+#define L2C_TAD_INT_W1C 0x40000
+#define L2C_TAD_INT_W1S 0x40008
+
+#define L2C_TAD_INT_ENA_W1C 0x40020
+#define L2C_TAD_INT_ENA_W1S 0x40028
+
+
+#define L2C_TAD_INT_L2DDBE BIT(1)
+#define L2C_TAD_INT_SBFSBE BIT(2)
+#define L2C_TAD_INT_SBFDBE BIT(3)
+#define L2C_TAD_INT_FBFSBE BIT(4)
+#define L2C_TAD_INT_FBFDBE BIT(5)
+#define L2C_TAD_INT_TAGDBE BIT(9)
+#define L2C_TAD_INT_RDDISLMC BIT(15)
+#define L2C_TAD_INT_WRDISLMC BIT(16)
+#define L2C_TAD_INT_LFBTO BIT(17)
+#define L2C_TAD_INT_GSYNCTO BIT(18)
+#define L2C_TAD_INT_RTGSBE BIT(32)
+#define L2C_TAD_INT_RTGDBE BIT(33)
+#define L2C_TAD_INT_RDDISOCI BIT(34)
+#define L2C_TAD_INT_WRDISOCI BIT(35)
+
+#define L2C_TAD_INT_ECC (L2C_TAD_INT_L2DDBE | \
+ L2C_TAD_INT_SBFSBE | L2C_TAD_INT_SBFDBE | \
+ L2C_TAD_INT_FBFSBE | L2C_TAD_INT_FBFDBE)
+
+#define L2C_TAD_INT_CE (L2C_TAD_INT_SBFSBE | \
+ L2C_TAD_INT_FBFSBE)
+
+#define L2C_TAD_INT_UE (L2C_TAD_INT_L2DDBE | \
+ L2C_TAD_INT_SBFDBE | \
+ L2C_TAD_INT_FBFDBE | \
+ L2C_TAD_INT_TAGDBE | \
+ L2C_TAD_INT_RTGDBE | \
+ L2C_TAD_INT_WRDISOCI | \
+ L2C_TAD_INT_RDDISOCI | \
+ L2C_TAD_INT_WRDISLMC | \
+ L2C_TAD_INT_RDDISLMC | \
+ L2C_TAD_INT_LFBTO | \
+ L2C_TAD_INT_GSYNCTO)
+
+static const struct error_descr l2_tad_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = L2C_TAD_INT_SBFSBE,
+ .descr = "SBF single-bit error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = L2C_TAD_INT_FBFSBE,
+ .descr = "FBF single-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_L2DDBE,
+ .descr = "L2D double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_SBFDBE,
+ .descr = "SBF double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_FBFDBE,
+ .descr = "FBF double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_TAGDBE,
+ .descr = "TAG double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_RTGDBE,
+ .descr = "RTG double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_WRDISOCI,
+ .descr = "Write to a disabled CCPI",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_RDDISOCI,
+ .descr = "Read from a disabled CCPI",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_WRDISLMC,
+ .descr = "Write to a disabled LMC",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_RDDISLMC,
+ .descr = "Read from a disabled LMC",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_LFBTO,
+ .descr = "LFB entry timeout",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_TAD_INT_GSYNCTO,
+ .descr = "Global sync CCPI timeout",
+ },
+ {0, 0, NULL},
+};
+
+#define L2C_TAD_INT_TAG (L2C_TAD_INT_TAGDBE)
+
+#define L2C_TAD_INT_RTG (L2C_TAD_INT_RTGDBE)
+
+#define L2C_TAD_INT_DISLMC (L2C_TAD_INT_WRDISLMC | L2C_TAD_INT_RDDISLMC)
+
+#define L2C_TAD_INT_DISOCI (L2C_TAD_INT_WRDISOCI | L2C_TAD_INT_RDDISOCI)
+
+#define L2C_TAD_INT_ENA_ALL (L2C_TAD_INT_ECC | L2C_TAD_INT_TAG | \
+ L2C_TAD_INT_RTG | \
+ L2C_TAD_INT_DISLMC | L2C_TAD_INT_DISOCI | \
+ L2C_TAD_INT_LFBTO)
+
+#define L2C_TAD_TIMETWO 0x50000
+#define L2C_TAD_TIMEOUT 0x50100
+#define L2C_TAD_ERR 0x60000
+#define L2C_TAD_TQD_ERR 0x60100
+#define L2C_TAD_TTG_ERR 0x60200
+
+
+#define L2C_CBC_INT_W1C 0x60000
+
+#define L2C_CBC_INT_RSDSBE BIT(0)
+#define L2C_CBC_INT_RSDDBE BIT(1)
+
+#define L2C_CBC_INT_RSD (L2C_CBC_INT_RSDSBE | L2C_CBC_INT_RSDDBE)
+
+#define L2C_CBC_INT_MIBSBE BIT(4)
+#define L2C_CBC_INT_MIBDBE BIT(5)
+
+#define L2C_CBC_INT_MIB (L2C_CBC_INT_MIBSBE | L2C_CBC_INT_MIBDBE)
+
+#define L2C_CBC_INT_IORDDISOCI BIT(6)
+#define L2C_CBC_INT_IOWRDISOCI BIT(7)
+
+#define L2C_CBC_INT_IODISOCI (L2C_CBC_INT_IORDDISOCI | \
+ L2C_CBC_INT_IOWRDISOCI)
+
+#define L2C_CBC_INT_CE (L2C_CBC_INT_RSDSBE | L2C_CBC_INT_MIBSBE)
+#define L2C_CBC_INT_UE (L2C_CBC_INT_RSDDBE | L2C_CBC_INT_MIBDBE)
+
+
+static const struct error_descr l2_cbc_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = L2C_CBC_INT_RSDSBE,
+ .descr = "RSD single-bit error",
+ },
+ {
+ .type = ERR_CORRECTED,
+ .mask = L2C_CBC_INT_MIBSBE,
+ .descr = "MIB single-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_CBC_INT_RSDDBE,
+ .descr = "RSD double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_CBC_INT_MIBDBE,
+ .descr = "MIB double-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_CBC_INT_IORDDISOCI,
+ .descr = "Read from a disabled CCPI",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_CBC_INT_IOWRDISOCI,
+ .descr = "Write to a disabled CCPI",
+ },
+ {0, 0, NULL},
+};
+
+#define L2C_CBC_INT_W1S 0x60008
+#define L2C_CBC_INT_ENA_W1C 0x60020
+
+#define L2C_CBC_INT_ENA_ALL (L2C_CBC_INT_RSD | L2C_CBC_INT_MIB | \
+ L2C_CBC_INT_IODISOCI)
+
+#define L2C_CBC_INT_ENA_W1S 0x60028
+
+#define L2C_CBC_IODISOCIERR 0x80008
+#define L2C_CBC_IOCERR 0x80010
+#define L2C_CBC_RSDERR 0x80018
+#define L2C_CBC_MIBERR 0x80020
+
+
+#define L2C_MCI_INT_W1C 0x0
+
+#define L2C_MCI_INT_VBFSBE BIT(0)
+#define L2C_MCI_INT_VBFDBE BIT(1)
+
+static const struct error_descr l2_mci_errors[] = {
+ {
+ .type = ERR_CORRECTED,
+ .mask = L2C_MCI_INT_VBFSBE,
+ .descr = "VBF single-bit error",
+ },
+ {
+ .type = ERR_UNCORRECTED,
+ .mask = L2C_MCI_INT_VBFDBE,
+ .descr = "VBF double-bit error",
+ },
+ {0, 0, NULL},
+};
+
+#define L2C_MCI_INT_W1S 0x8
+#define L2C_MCI_INT_ENA_W1C 0x20
+
+#define L2C_MCI_INT_ENA_ALL (L2C_MCI_INT_VBFSBE | L2C_MCI_INT_VBFDBE)
+
+#define L2C_MCI_INT_ENA_W1S 0x28
+
+#define L2C_MCI_ERR 0x10000
+
+#define L2C_MESSAGE_SIZE SZ_1K
+#define L2C_OTHER_SIZE (50 * ARRAY_SIZE(l2_tad_errors))
+
+struct l2c_err_ctx {
+ char *reg_ext_name;
+ u64 reg_int;
+ u64 reg_ext;
+};
+
+struct thunderx_l2c {
+ void __iomem *regs;
+ struct pci_dev *pdev;
+ struct edac_device_ctl_info *edac_dev;
+
+ struct dentry *debugfs;
+
+ int index;
+
+ struct msix_entry msix_ent;
+
+ struct l2c_err_ctx err_ctx[RING_ENTRIES];
+ unsigned long ring_head;
+ unsigned long ring_tail;
+};
+
+static irqreturn_t thunderx_l2c_tad_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_l2c *tad = container_of(msix, struct thunderx_l2c,
+ msix_ent);
+
+ unsigned long head = ring_pos(tad->ring_head, ARRAY_SIZE(tad->err_ctx));
+ struct l2c_err_ctx *ctx = &tad->err_ctx[head];
+
+ ctx->reg_int = readq(tad->regs + L2C_TAD_INT_W1C);
+
+ if (ctx->reg_int & L2C_TAD_INT_ECC) {
+ ctx->reg_ext_name = "TQD_ERR";
+ ctx->reg_ext = readq(tad->regs + L2C_TAD_TQD_ERR);
+ } else if (ctx->reg_int & L2C_TAD_INT_TAG) {
+ ctx->reg_ext_name = "TTG_ERR";
+ ctx->reg_ext = readq(tad->regs + L2C_TAD_TTG_ERR);
+ } else if (ctx->reg_int & L2C_TAD_INT_LFBTO) {
+ ctx->reg_ext_name = "TIMEOUT";
+ ctx->reg_ext = readq(tad->regs + L2C_TAD_TIMEOUT);
+ } else if (ctx->reg_int & L2C_TAD_INT_DISOCI) {
+ ctx->reg_ext_name = "ERR";
+ ctx->reg_ext = readq(tad->regs + L2C_TAD_ERR);
+ }
+
+ writeq(ctx->reg_int, tad->regs + L2C_TAD_INT_W1C);
+
+ tad->ring_head++;
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_l2c_cbc_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_l2c *cbc = container_of(msix, struct thunderx_l2c,
+ msix_ent);
+
+ unsigned long head = ring_pos(cbc->ring_head, ARRAY_SIZE(cbc->err_ctx));
+ struct l2c_err_ctx *ctx = &cbc->err_ctx[head];
+
+ ctx->reg_int = readq(cbc->regs + L2C_CBC_INT_W1C);
+
+ if (ctx->reg_int & L2C_CBC_INT_RSD) {
+ ctx->reg_ext_name = "RSDERR";
+ ctx->reg_ext = readq(cbc->regs + L2C_CBC_RSDERR);
+ } else if (ctx->reg_int & L2C_CBC_INT_MIB) {
+ ctx->reg_ext_name = "MIBERR";
+ ctx->reg_ext = readq(cbc->regs + L2C_CBC_MIBERR);
+ } else if (ctx->reg_int & L2C_CBC_INT_IODISOCI) {
+ ctx->reg_ext_name = "IODISOCIERR";
+ ctx->reg_ext = readq(cbc->regs + L2C_CBC_IODISOCIERR);
+ }
+
+ writeq(ctx->reg_int, cbc->regs + L2C_CBC_INT_W1C);
+
+ cbc->ring_head++;
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_l2c_mci_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_l2c *mci = container_of(msix, struct thunderx_l2c,
+ msix_ent);
+
+ unsigned long head = ring_pos(mci->ring_head, ARRAY_SIZE(mci->err_ctx));
+ struct l2c_err_ctx *ctx = &mci->err_ctx[head];
+
+ ctx->reg_int = readq(mci->regs + L2C_MCI_INT_W1C);
+ ctx->reg_ext = readq(mci->regs + L2C_MCI_ERR);
+
+ writeq(ctx->reg_int, mci->regs + L2C_MCI_INT_W1C);
+
+ ctx->reg_ext_name = "ERR";
+
+ mci->ring_head++;
+
+ return IRQ_WAKE_THREAD;
+}
+
+static irqreturn_t thunderx_l2c_threaded_isr(int irq, void *irq_id)
+{
+ struct msix_entry *msix = irq_id;
+ struct thunderx_l2c *l2c = container_of(msix, struct thunderx_l2c,
+ msix_ent);
+
+ unsigned long tail = ring_pos(l2c->ring_tail, ARRAY_SIZE(l2c->err_ctx));
+ struct l2c_err_ctx *ctx = &l2c->err_ctx[tail];
+ irqreturn_t ret = IRQ_NONE;
+
+ u64 mask_ue, mask_ce;
+ const struct error_descr *l2_errors;
+ char *reg_int_name;
+
+ char *msg;
+ char *other;
+
+ msg = kmalloc(OCX_MESSAGE_SIZE, GFP_KERNEL);
+ other = kmalloc(OCX_OTHER_SIZE, GFP_KERNEL);
+
+ if (!msg || !other)
+ goto err_free;
+
+ switch (l2c->pdev->device) {
+ case PCI_DEVICE_ID_THUNDER_L2C_TAD:
+ reg_int_name = "L2C_TAD_INT";
+ mask_ue = L2C_TAD_INT_UE;
+ mask_ce = L2C_TAD_INT_CE;
+ l2_errors = l2_tad_errors;
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_CBC:
+ reg_int_name = "L2C_CBC_INT";
+ mask_ue = L2C_CBC_INT_UE;
+ mask_ce = L2C_CBC_INT_CE;
+ l2_errors = l2_cbc_errors;
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_MCI:
+ reg_int_name = "L2C_MCI_INT";
+ mask_ue = L2C_MCI_INT_VBFDBE;
+ mask_ce = L2C_MCI_INT_VBFSBE;
+ l2_errors = l2_mci_errors;
+ break;
+ default:
+ dev_err(&l2c->pdev->dev, "Unsupported device: %04x\n",
+ l2c->pdev->device);
+ goto err_free;
+ }
+
+ while (CIRC_CNT(l2c->ring_head, l2c->ring_tail,
+ ARRAY_SIZE(l2c->err_ctx))) {
+ snprintf(msg, L2C_MESSAGE_SIZE,
+ "%s: %s: %016llx, %s: %016llx",
+ l2c->edac_dev->ctl_name, reg_int_name, ctx->reg_int,
+ ctx->reg_ext_name, ctx->reg_ext);
+
+ decode_register(other, L2C_OTHER_SIZE, l2_errors, ctx->reg_int);
+
+ strlcat(msg, other, L2C_MESSAGE_SIZE);
+
+ if (ctx->reg_int & mask_ue)
+ edac_device_handle_ue(l2c->edac_dev, 0, 0, msg);
+ else if (ctx->reg_int & mask_ce)
+ edac_device_handle_ce(l2c->edac_dev, 0, 0, msg);
+
+ l2c->ring_tail++;
+ }
+
+ ret = IRQ_HANDLED;
+
+err_free:
+ kfree(other);
+ kfree(msg);
+
+ return ret;
+}
+
+#define L2C_DEBUGFS_ATTR(_name, _reg) DEBUGFS_REG_ATTR(l2c, _name, _reg)
+
+L2C_DEBUGFS_ATTR(tad_int, L2C_TAD_INT_W1S);
+
+static struct debugfs_entry *l2c_tad_dfs_ents[] = {
+ &debugfs_tad_int,
+};
+
+L2C_DEBUGFS_ATTR(cbc_int, L2C_CBC_INT_W1S);
+
+static struct debugfs_entry *l2c_cbc_dfs_ents[] = {
+ &debugfs_cbc_int,
+};
+
+L2C_DEBUGFS_ATTR(mci_int, L2C_MCI_INT_W1S);
+
+static struct debugfs_entry *l2c_mci_dfs_ents[] = {
+ &debugfs_mci_int,
+};
+
+static const struct pci_device_id thunderx_l2c_pci_tbl[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVICE_ID_THUNDER_L2C_TAD), },
+ { PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVICE_ID_THUNDER_L2C_CBC), },
+ { PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVICE_ID_THUNDER_L2C_MCI), },
+ { 0, },
+};
+
+static int thunderx_l2c_probe(struct pci_dev *pdev,
+ const struct pci_device_id *id)
+{
+ struct thunderx_l2c *l2c;
+ struct edac_device_ctl_info *edac_dev;
+ struct debugfs_entry **l2c_devattr;
+ size_t dfs_entries;
+ irqreturn_t (*thunderx_l2c_isr)(int, void *) = NULL;
+ char name[32];
+ const char *fmt;
+ u64 reg_en_offs, reg_en_mask;
+ int idx;
+ int ret;
+
+ ret = pcim_enable_device(pdev);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable PCI device: %d\n", ret);
+ return ret;
+ }
+
+ ret = pcim_iomap_regions(pdev, BIT(0), "thunderx_l2c");
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot map PCI resources: %d\n", ret);
+ return ret;
+ }
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_THUNDER_L2C_TAD:
+ thunderx_l2c_isr = thunderx_l2c_tad_isr;
+ l2c_devattr = l2c_tad_dfs_ents;
+ dfs_entries = ARRAY_SIZE(l2c_tad_dfs_ents);
+ fmt = "L2C-TAD%d";
+ reg_en_offs = L2C_TAD_INT_ENA_W1S;
+ reg_en_mask = L2C_TAD_INT_ENA_ALL;
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_CBC:
+ thunderx_l2c_isr = thunderx_l2c_cbc_isr;
+ l2c_devattr = l2c_cbc_dfs_ents;
+ dfs_entries = ARRAY_SIZE(l2c_cbc_dfs_ents);
+ fmt = "L2C-CBC%d";
+ reg_en_offs = L2C_CBC_INT_ENA_W1S;
+ reg_en_mask = L2C_CBC_INT_ENA_ALL;
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_MCI:
+ thunderx_l2c_isr = thunderx_l2c_mci_isr;
+ l2c_devattr = l2c_mci_dfs_ents;
+ dfs_entries = ARRAY_SIZE(l2c_mci_dfs_ents);
+ fmt = "L2C-MCI%d";
+ reg_en_offs = L2C_MCI_INT_ENA_W1S;
+ reg_en_mask = L2C_MCI_INT_ENA_ALL;
+ break;
+ default:
+ //Should never ever get here
+ dev_err(&pdev->dev, "Unsupported PCI device: %04x\n",
+ pdev->device);
+ return -EINVAL;
+ }
+
+ idx = edac_device_alloc_index();
+ snprintf(name, sizeof(name), fmt, idx);
+
+ edac_dev = edac_device_alloc_ctl_info(sizeof(struct thunderx_l2c),
+ name, 1, "L2C", 1, 0,
+ NULL, 0, idx);
+ if (!edac_dev) {
+ dev_err(&pdev->dev, "Cannot allocate EDAC device\n");
+ return -ENOMEM;
+ }
+
+ l2c = edac_dev->pvt_info;
+ l2c->edac_dev = edac_dev;
+
+ l2c->regs = pcim_iomap_table(pdev)[0];
+ if (!l2c->regs) {
+ dev_err(&pdev->dev, "Cannot map PCI resources\n");
+ ret = -ENODEV;
+ goto err_free;
+ }
+
+ l2c->pdev = pdev;
+
+ l2c->ring_head = 0;
+ l2c->ring_tail = 0;
+
+ l2c->msix_ent.entry = 0;
+ l2c->msix_ent.vector = 0;
+
+ ret = pci_enable_msix_exact(pdev, &l2c->msix_ent, 1);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot enable interrupt: %d\n", ret);
+ goto err_free;
+ }
+
+ ret = devm_request_threaded_irq(&pdev->dev, l2c->msix_ent.vector,
+ thunderx_l2c_isr,
+ thunderx_l2c_threaded_isr,
+ 0, "[EDAC] ThunderX L2C",
+ &l2c->msix_ent);
+ if (ret)
+ goto err_free;
+
+ edac_dev->dev = &pdev->dev;
+ edac_dev->dev_name = dev_name(&pdev->dev);
+ edac_dev->mod_name = "thunderx-l2c";
+ edac_dev->ctl_name = "thunderx-l2c";
+
+ ret = edac_device_add_device(edac_dev);
+ if (ret) {
+ dev_err(&pdev->dev, "Cannot add EDAC device: %d\n", ret);
+ goto err_free;
+ }
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG)) {
+ l2c->debugfs = edac_debugfs_create_dir(pdev->dev.kobj.name);
+
+ ret = thunderx_create_debugfs_nodes(l2c->debugfs, l2c_devattr,
+ l2c, dfs_entries);
+
+ if (ret != dfs_entries) {
+ dev_warn(&pdev->dev, "Error creating debugfs entries: %d%s\n",
+ ret, ret >= 0 ? " created" : "");
+ }
+ }
+
+ pci_set_drvdata(pdev, edac_dev);
+
+ writeq(reg_en_mask, l2c->regs + reg_en_offs);
+
+ return 0;
+
+err_free:
+ edac_device_free_ctl_info(edac_dev);
+
+ return ret;
+}
+
+static void thunderx_l2c_remove(struct pci_dev *pdev)
+{
+ struct edac_device_ctl_info *edac_dev = pci_get_drvdata(pdev);
+ struct thunderx_l2c *l2c = edac_dev->pvt_info;
+
+ switch (pdev->device) {
+ case PCI_DEVICE_ID_THUNDER_L2C_TAD:
+ writeq(L2C_TAD_INT_ENA_ALL, l2c->regs + L2C_TAD_INT_ENA_W1C);
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_CBC:
+ writeq(L2C_CBC_INT_ENA_ALL, l2c->regs + L2C_CBC_INT_ENA_W1C);
+ break;
+ case PCI_DEVICE_ID_THUNDER_L2C_MCI:
+ writeq(L2C_MCI_INT_ENA_ALL, l2c->regs + L2C_MCI_INT_ENA_W1C);
+ break;
+ }
+
+ edac_debugfs_remove_recursive(l2c->debugfs);
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(edac_dev);
+}
+
+MODULE_DEVICE_TABLE(pci, thunderx_l2c_pci_tbl);
+
+static struct pci_driver thunderx_l2c_driver = {
+ .name = "thunderx_l2c_edac",
+ .probe = thunderx_l2c_probe,
+ .remove = thunderx_l2c_remove,
+ .id_table = thunderx_l2c_pci_tbl,
+};
+
+static int __init thunderx_edac_init(void)
+{
+ int rc = 0;
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ rc = pci_register_driver(&thunderx_lmc_driver);
+ if (rc)
+ return rc;
+
+ rc = pci_register_driver(&thunderx_ocx_driver);
+ if (rc)
+ goto err_lmc;
+
+ rc = pci_register_driver(&thunderx_l2c_driver);
+ if (rc)
+ goto err_ocx;
+
+ return rc;
+err_ocx:
+ pci_unregister_driver(&thunderx_ocx_driver);
+err_lmc:
+ pci_unregister_driver(&thunderx_lmc_driver);
+
+ return rc;
+}
+
+static void __exit thunderx_edac_exit(void)
+{
+ pci_unregister_driver(&thunderx_l2c_driver);
+ pci_unregister_driver(&thunderx_ocx_driver);
+ pci_unregister_driver(&thunderx_lmc_driver);
+
+}
+
+module_init(thunderx_edac_init);
+module_exit(thunderx_edac_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Cavium, Inc.");
+MODULE_DESCRIPTION("EDAC Driver for Cavium ThunderX");
diff --git a/drivers/edac/ti_edac.c b/drivers/edac/ti_edac.c
new file mode 100644
index 0000000000..6971ded598
--- /dev/null
+++ b/drivers/edac/ti_edac.c
@@ -0,0 +1,338 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2017 Texas Instruments Incorporated - https://www.ti.com/
+ *
+ * Texas Instruments DDR3 ECC error correction and detection driver
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/init.h>
+#include <linux/edac.h>
+#include <linux/io.h>
+#include <linux/interrupt.h>
+#include <linux/of_address.h>
+#include <linux/of_device.h>
+#include <linux/module.h>
+
+#include "edac_module.h"
+
+/* EMIF controller registers */
+#define EMIF_SDRAM_CONFIG 0x008
+#define EMIF_IRQ_STATUS 0x0ac
+#define EMIF_IRQ_ENABLE_SET 0x0b4
+#define EMIF_ECC_CTRL 0x110
+#define EMIF_1B_ECC_ERR_CNT 0x130
+#define EMIF_1B_ECC_ERR_THRSH 0x134
+#define EMIF_1B_ECC_ERR_ADDR_LOG 0x13c
+#define EMIF_2B_ECC_ERR_ADDR_LOG 0x140
+
+/* Bit definitions for EMIF_SDRAM_CONFIG */
+#define SDRAM_TYPE_SHIFT 29
+#define SDRAM_TYPE_MASK GENMASK(31, 29)
+#define SDRAM_TYPE_DDR3 (3 << SDRAM_TYPE_SHIFT)
+#define SDRAM_TYPE_DDR2 (2 << SDRAM_TYPE_SHIFT)
+#define SDRAM_NARROW_MODE_MASK GENMASK(15, 14)
+#define SDRAM_K2_NARROW_MODE_SHIFT 12
+#define SDRAM_K2_NARROW_MODE_MASK GENMASK(13, 12)
+#define SDRAM_ROWSIZE_SHIFT 7
+#define SDRAM_ROWSIZE_MASK GENMASK(9, 7)
+#define SDRAM_IBANK_SHIFT 4
+#define SDRAM_IBANK_MASK GENMASK(6, 4)
+#define SDRAM_K2_IBANK_SHIFT 5
+#define SDRAM_K2_IBANK_MASK GENMASK(6, 5)
+#define SDRAM_K2_EBANK_SHIFT 3
+#define SDRAM_K2_EBANK_MASK BIT(SDRAM_K2_EBANK_SHIFT)
+#define SDRAM_PAGESIZE_SHIFT 0
+#define SDRAM_PAGESIZE_MASK GENMASK(2, 0)
+#define SDRAM_K2_PAGESIZE_SHIFT 0
+#define SDRAM_K2_PAGESIZE_MASK GENMASK(1, 0)
+
+#define EMIF_1B_ECC_ERR_THRSH_SHIFT 24
+
+/* IRQ bit definitions */
+#define EMIF_1B_ECC_ERR BIT(5)
+#define EMIF_2B_ECC_ERR BIT(4)
+#define EMIF_WR_ECC_ERR BIT(3)
+#define EMIF_SYS_ERR BIT(0)
+/* Bit 31 enables ECC and 28 enables RMW */
+#define ECC_ENABLED (BIT(31) | BIT(28))
+
+#define EDAC_MOD_NAME "ti-emif-edac"
+
+enum {
+ EMIF_TYPE_DRA7,
+ EMIF_TYPE_K2
+};
+
+struct ti_edac {
+ void __iomem *reg;
+};
+
+static u32 ti_edac_readl(struct ti_edac *edac, u16 offset)
+{
+ return readl_relaxed(edac->reg + offset);
+}
+
+static void ti_edac_writel(struct ti_edac *edac, u32 val, u16 offset)
+{
+ writel_relaxed(val, edac->reg + offset);
+}
+
+static irqreturn_t ti_edac_isr(int irq, void *data)
+{
+ struct mem_ctl_info *mci = data;
+ struct ti_edac *edac = mci->pvt_info;
+ u32 irq_status;
+ u32 err_addr;
+ int err_count;
+
+ irq_status = ti_edac_readl(edac, EMIF_IRQ_STATUS);
+
+ if (irq_status & EMIF_1B_ECC_ERR) {
+ err_addr = ti_edac_readl(edac, EMIF_1B_ECC_ERR_ADDR_LOG);
+ err_count = ti_edac_readl(edac, EMIF_1B_ECC_ERR_CNT);
+ ti_edac_writel(edac, err_count, EMIF_1B_ECC_ERR_CNT);
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, err_count,
+ err_addr >> PAGE_SHIFT,
+ err_addr & ~PAGE_MASK, -1, 0, 0, 0,
+ mci->ctl_name, "1B");
+ }
+
+ if (irq_status & EMIF_2B_ECC_ERR) {
+ err_addr = ti_edac_readl(edac, EMIF_2B_ECC_ERR_ADDR_LOG);
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ err_addr >> PAGE_SHIFT,
+ err_addr & ~PAGE_MASK, -1, 0, 0, 0,
+ mci->ctl_name, "2B");
+ }
+
+ if (irq_status & EMIF_WR_ECC_ERR)
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ 0, 0, -1, 0, 0, 0,
+ mci->ctl_name, "WR");
+
+ ti_edac_writel(edac, irq_status, EMIF_IRQ_STATUS);
+
+ return IRQ_HANDLED;
+}
+
+static void ti_edac_setup_dimm(struct mem_ctl_info *mci, u32 type)
+{
+ struct dimm_info *dimm;
+ struct ti_edac *edac = mci->pvt_info;
+ int bits;
+ u32 val;
+ u32 memsize;
+
+ dimm = edac_get_dimm(mci, 0, 0, 0);
+
+ val = ti_edac_readl(edac, EMIF_SDRAM_CONFIG);
+
+ if (type == EMIF_TYPE_DRA7) {
+ bits = ((val & SDRAM_PAGESIZE_MASK) >> SDRAM_PAGESIZE_SHIFT) + 8;
+ bits += ((val & SDRAM_ROWSIZE_MASK) >> SDRAM_ROWSIZE_SHIFT) + 9;
+ bits += (val & SDRAM_IBANK_MASK) >> SDRAM_IBANK_SHIFT;
+
+ if (val & SDRAM_NARROW_MODE_MASK) {
+ bits++;
+ dimm->dtype = DEV_X16;
+ } else {
+ bits += 2;
+ dimm->dtype = DEV_X32;
+ }
+ } else {
+ bits = 16;
+ bits += ((val & SDRAM_K2_PAGESIZE_MASK) >>
+ SDRAM_K2_PAGESIZE_SHIFT) + 8;
+ bits += (val & SDRAM_K2_IBANK_MASK) >> SDRAM_K2_IBANK_SHIFT;
+ bits += (val & SDRAM_K2_EBANK_MASK) >> SDRAM_K2_EBANK_SHIFT;
+
+ val = (val & SDRAM_K2_NARROW_MODE_MASK) >>
+ SDRAM_K2_NARROW_MODE_SHIFT;
+ switch (val) {
+ case 0:
+ bits += 3;
+ dimm->dtype = DEV_X64;
+ break;
+ case 1:
+ bits += 2;
+ dimm->dtype = DEV_X32;
+ break;
+ case 2:
+ bits++;
+ dimm->dtype = DEV_X16;
+ break;
+ }
+ }
+
+ memsize = 1 << bits;
+
+ dimm->nr_pages = memsize >> PAGE_SHIFT;
+ dimm->grain = 4;
+ if ((val & SDRAM_TYPE_MASK) == SDRAM_TYPE_DDR2)
+ dimm->mtype = MEM_DDR2;
+ else
+ dimm->mtype = MEM_DDR3;
+
+ val = ti_edac_readl(edac, EMIF_ECC_CTRL);
+ if (val & ECC_ENABLED)
+ dimm->edac_mode = EDAC_SECDED;
+ else
+ dimm->edac_mode = EDAC_NONE;
+}
+
+static const struct of_device_id ti_edac_of_match[] = {
+ { .compatible = "ti,emif-keystone", .data = (void *)EMIF_TYPE_K2 },
+ { .compatible = "ti,emif-dra7xx", .data = (void *)EMIF_TYPE_DRA7 },
+ {},
+};
+MODULE_DEVICE_TABLE(of, ti_edac_of_match);
+
+static int _emif_get_id(struct device_node *node)
+{
+ struct device_node *np;
+ const __be32 *addrp;
+ u32 addr, my_addr;
+ int my_id = 0;
+
+ addrp = of_get_address(node, 0, NULL, NULL);
+ my_addr = (u32)of_translate_address(node, addrp);
+
+ for_each_matching_node(np, ti_edac_of_match) {
+ if (np == node)
+ continue;
+
+ addrp = of_get_address(np, 0, NULL, NULL);
+ addr = (u32)of_translate_address(np, addrp);
+
+ edac_printk(KERN_INFO, EDAC_MOD_NAME,
+ "addr=%x, my_addr=%x\n",
+ addr, my_addr);
+
+ if (addr < my_addr)
+ my_id++;
+ }
+
+ return my_id;
+}
+
+static int ti_edac_probe(struct platform_device *pdev)
+{
+ int error_irq = 0, ret = -ENODEV;
+ struct device *dev = &pdev->dev;
+ struct resource *res;
+ void __iomem *reg;
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[1];
+ const struct of_device_id *id;
+ struct ti_edac *edac;
+ int emif_id;
+
+ id = of_match_device(ti_edac_of_match, &pdev->dev);
+ if (!id)
+ return -ENODEV;
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ reg = devm_ioremap_resource(dev, res);
+ if (IS_ERR(reg))
+ return PTR_ERR(reg);
+
+ layers[0].type = EDAC_MC_LAYER_ALL_MEM;
+ layers[0].size = 1;
+
+ /* Allocate ID number for our EMIF controller */
+ emif_id = _emif_get_id(pdev->dev.of_node);
+ if (emif_id < 0)
+ return -EINVAL;
+
+ mci = edac_mc_alloc(emif_id, 1, layers, sizeof(*edac));
+ if (!mci)
+ return -ENOMEM;
+
+ mci->pdev = &pdev->dev;
+ edac = mci->pvt_info;
+ edac->reg = reg;
+ platform_set_drvdata(pdev, mci);
+
+ mci->mtype_cap = MEM_FLAG_DDR3 | MEM_FLAG_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED | EDAC_FLAG_NONE;
+ mci->mod_name = EDAC_MOD_NAME;
+ mci->ctl_name = id->compatible;
+ mci->dev_name = dev_name(&pdev->dev);
+
+ /* Setup memory layout */
+ ti_edac_setup_dimm(mci, (u32)(id->data));
+
+ /* add EMIF ECC error handler */
+ error_irq = platform_get_irq(pdev, 0);
+ if (error_irq < 0) {
+ ret = error_irq;
+ goto err;
+ }
+
+ ret = devm_request_irq(dev, error_irq, ti_edac_isr, 0,
+ "emif-edac-irq", mci);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME,
+ "request_irq fail for EMIF EDAC irq\n");
+ goto err;
+ }
+
+ ret = edac_mc_add_mc(mci);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_MOD_NAME,
+ "Failed to register mci: %d.\n", ret);
+ goto err;
+ }
+
+ /* Generate an interrupt with each 1b error */
+ ti_edac_writel(edac, 1 << EMIF_1B_ECC_ERR_THRSH_SHIFT,
+ EMIF_1B_ECC_ERR_THRSH);
+
+ /* Enable interrupts */
+ ti_edac_writel(edac,
+ EMIF_1B_ECC_ERR | EMIF_2B_ECC_ERR | EMIF_WR_ECC_ERR,
+ EMIF_IRQ_ENABLE_SET);
+
+ return 0;
+
+err:
+ edac_mc_free(mci);
+ return ret;
+}
+
+static int ti_edac_remove(struct platform_device *pdev)
+{
+ struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+ edac_mc_del_mc(&pdev->dev);
+ edac_mc_free(mci);
+
+ return 0;
+}
+
+static struct platform_driver ti_edac_driver = {
+ .probe = ti_edac_probe,
+ .remove = ti_edac_remove,
+ .driver = {
+ .name = EDAC_MOD_NAME,
+ .of_match_table = ti_edac_of_match,
+ },
+};
+
+module_platform_driver(ti_edac_driver);
+
+MODULE_AUTHOR("Texas Instruments Inc.");
+MODULE_DESCRIPTION("EDAC Driver for Texas Instruments DDR3 MC");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/edac/wq.c b/drivers/edac/wq.c
new file mode 100644
index 0000000000..ad3f516627
--- /dev/null
+++ b/drivers/edac/wq.c
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include "edac_module.h"
+
+static struct workqueue_struct *wq;
+
+bool edac_queue_work(struct delayed_work *work, unsigned long delay)
+{
+ return queue_delayed_work(wq, work, delay);
+}
+EXPORT_SYMBOL_GPL(edac_queue_work);
+
+bool edac_mod_work(struct delayed_work *work, unsigned long delay)
+{
+ return mod_delayed_work(wq, work, delay);
+}
+EXPORT_SYMBOL_GPL(edac_mod_work);
+
+bool edac_stop_work(struct delayed_work *work)
+{
+ bool ret;
+
+ ret = cancel_delayed_work_sync(work);
+ flush_workqueue(wq);
+
+ return ret;
+}
+EXPORT_SYMBOL_GPL(edac_stop_work);
+
+int edac_workqueue_setup(void)
+{
+ wq = alloc_ordered_workqueue("edac-poller", WQ_MEM_RECLAIM);
+ if (!wq)
+ return -ENODEV;
+ else
+ return 0;
+}
+
+void edac_workqueue_teardown(void)
+{
+ destroy_workqueue(wq);
+ wq = NULL;
+}
diff --git a/drivers/edac/x38_edac.c b/drivers/edac/x38_edac.c
new file mode 100644
index 0000000000..49ab5721aa
--- /dev/null
+++ b/drivers/edac/x38_edac.c
@@ -0,0 +1,523 @@
+/*
+ * Intel X38 Memory Controller kernel module
+ * Copyright (C) 2008 Cluster Computing, Inc.
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * This file is based on i3200_edac.c
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/edac.h>
+
+#include <linux/io-64-nonatomic-lo-hi.h>
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "x38_edac"
+
+#define PCI_DEVICE_ID_INTEL_X38_HB 0x29e0
+
+#define X38_RANKS 8
+#define X38_RANKS_PER_CHANNEL 4
+#define X38_CHANNELS 2
+
+/* Intel X38 register addresses - device 0 function 0 - DRAM Controller */
+
+#define X38_MCHBAR_LOW 0x48 /* MCH Memory Mapped Register BAR */
+#define X38_MCHBAR_HIGH 0x4c
+#define X38_MCHBAR_MASK 0xfffffc000ULL /* bits 35:14 */
+#define X38_MMR_WINDOW_SIZE 16384
+
+#define X38_TOM 0xa0 /* Top of Memory (16b)
+ *
+ * 15:10 reserved
+ * 9:0 total populated physical memory
+ */
+#define X38_TOM_MASK 0x3ff /* bits 9:0 */
+#define X38_TOM_SHIFT 26 /* 64MiB grain */
+
+#define X38_ERRSTS 0xc8 /* Error Status Register (16b)
+ *
+ * 15 reserved
+ * 14 Isochronous TBWRR Run Behind FIFO Full
+ * (ITCV)
+ * 13 Isochronous TBWRR Run Behind FIFO Put
+ * (ITSTV)
+ * 12 reserved
+ * 11 MCH Thermal Sensor Event
+ * for SMI/SCI/SERR (GTSE)
+ * 10 reserved
+ * 9 LOCK to non-DRAM Memory Flag (LCKF)
+ * 8 reserved
+ * 7 DRAM Throttle Flag (DTF)
+ * 6:2 reserved
+ * 1 Multi-bit DRAM ECC Error Flag (DMERR)
+ * 0 Single-bit DRAM ECC Error Flag (DSERR)
+ */
+#define X38_ERRSTS_UE 0x0002
+#define X38_ERRSTS_CE 0x0001
+#define X38_ERRSTS_BITS (X38_ERRSTS_UE | X38_ERRSTS_CE)
+
+
+/* Intel MMIO register space - device 0 function 0 - MMR space */
+
+#define X38_C0DRB 0x200 /* Channel 0 DRAM Rank Boundary (16b x 4)
+ *
+ * 15:10 reserved
+ * 9:0 Channel 0 DRAM Rank Boundary Address
+ */
+#define X38_C1DRB 0x600 /* Channel 1 DRAM Rank Boundary (16b x 4) */
+#define X38_DRB_MASK 0x3ff /* bits 9:0 */
+#define X38_DRB_SHIFT 26 /* 64MiB grain */
+
+#define X38_C0ECCERRLOG 0x280 /* Channel 0 ECC Error Log (64b)
+ *
+ * 63:48 Error Column Address (ERRCOL)
+ * 47:32 Error Row Address (ERRROW)
+ * 31:29 Error Bank Address (ERRBANK)
+ * 28:27 Error Rank Address (ERRRANK)
+ * 26:24 reserved
+ * 23:16 Error Syndrome (ERRSYND)
+ * 15: 2 reserved
+ * 1 Multiple Bit Error Status (MERRSTS)
+ * 0 Correctable Error Status (CERRSTS)
+ */
+#define X38_C1ECCERRLOG 0x680 /* Channel 1 ECC Error Log (64b) */
+#define X38_ECCERRLOG_CE 0x1
+#define X38_ECCERRLOG_UE 0x2
+#define X38_ECCERRLOG_RANK_BITS 0x18000000
+#define X38_ECCERRLOG_SYNDROME_BITS 0xff0000
+
+#define X38_CAPID0 0xe0 /* see P.94 of spec for details */
+
+static int x38_channel_num;
+
+static int how_many_channel(struct pci_dev *pdev)
+{
+ unsigned char capid0_8b; /* 8th byte of CAPID0 */
+
+ pci_read_config_byte(pdev, X38_CAPID0 + 8, &capid0_8b);
+ if (capid0_8b & 0x20) { /* check DCD: Dual Channel Disable */
+ edac_dbg(0, "In single channel mode\n");
+ x38_channel_num = 1;
+ } else {
+ edac_dbg(0, "In dual channel mode\n");
+ x38_channel_num = 2;
+ }
+
+ return x38_channel_num;
+}
+
+static unsigned long eccerrlog_syndrome(u64 log)
+{
+ return (log & X38_ECCERRLOG_SYNDROME_BITS) >> 16;
+}
+
+static int eccerrlog_row(int channel, u64 log)
+{
+ return ((log & X38_ECCERRLOG_RANK_BITS) >> 27) |
+ (channel * X38_RANKS_PER_CHANNEL);
+}
+
+enum x38_chips {
+ X38 = 0,
+};
+
+struct x38_dev_info {
+ const char *ctl_name;
+};
+
+struct x38_error_info {
+ u16 errsts;
+ u16 errsts2;
+ u64 eccerrlog[X38_CHANNELS];
+};
+
+static const struct x38_dev_info x38_devs[] = {
+ [X38] = {
+ .ctl_name = "x38"},
+};
+
+static struct pci_dev *mci_pdev;
+static int x38_registered = 1;
+
+
+static void x38_clear_error_info(struct mem_ctl_info *mci)
+{
+ struct pci_dev *pdev;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * Clear any error bits.
+ * (Yes, we really clear bits by writing 1 to them.)
+ */
+ pci_write_bits16(pdev, X38_ERRSTS, X38_ERRSTS_BITS,
+ X38_ERRSTS_BITS);
+}
+
+static void x38_get_and_clear_error_info(struct mem_ctl_info *mci,
+ struct x38_error_info *info)
+{
+ struct pci_dev *pdev;
+ void __iomem *window = mci->pvt_info;
+
+ pdev = to_pci_dev(mci->pdev);
+
+ /*
+ * This is a mess because there is no atomic way to read all the
+ * registers at once and the registers can transition from CE being
+ * overwritten by UE.
+ */
+ pci_read_config_word(pdev, X38_ERRSTS, &info->errsts);
+ if (!(info->errsts & X38_ERRSTS_BITS))
+ return;
+
+ info->eccerrlog[0] = lo_hi_readq(window + X38_C0ECCERRLOG);
+ if (x38_channel_num == 2)
+ info->eccerrlog[1] = lo_hi_readq(window + X38_C1ECCERRLOG);
+
+ pci_read_config_word(pdev, X38_ERRSTS, &info->errsts2);
+
+ /*
+ * If the error is the same for both reads then the first set
+ * of reads is valid. If there is a change then there is a CE
+ * with no info and the second set of reads is valid and
+ * should be UE info.
+ */
+ if ((info->errsts ^ info->errsts2) & X38_ERRSTS_BITS) {
+ info->eccerrlog[0] = lo_hi_readq(window + X38_C0ECCERRLOG);
+ if (x38_channel_num == 2)
+ info->eccerrlog[1] =
+ lo_hi_readq(window + X38_C1ECCERRLOG);
+ }
+
+ x38_clear_error_info(mci);
+}
+
+static void x38_process_error_info(struct mem_ctl_info *mci,
+ struct x38_error_info *info)
+{
+ int channel;
+ u64 log;
+
+ if (!(info->errsts & X38_ERRSTS_BITS))
+ return;
+
+ if ((info->errsts ^ info->errsts2) & X38_ERRSTS_BITS) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
+ -1, -1, -1,
+ "UE overwrote CE", "");
+ info->errsts = info->errsts2;
+ }
+
+ for (channel = 0; channel < x38_channel_num; channel++) {
+ log = info->eccerrlog[channel];
+ if (log & X38_ECCERRLOG_UE) {
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
+ 0, 0, 0,
+ eccerrlog_row(channel, log),
+ -1, -1,
+ "x38 UE", "");
+ } else if (log & X38_ECCERRLOG_CE) {
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
+ 0, 0, eccerrlog_syndrome(log),
+ eccerrlog_row(channel, log),
+ -1, -1,
+ "x38 CE", "");
+ }
+ }
+}
+
+static void x38_check(struct mem_ctl_info *mci)
+{
+ struct x38_error_info info;
+
+ x38_get_and_clear_error_info(mci, &info);
+ x38_process_error_info(mci, &info);
+}
+
+static void __iomem *x38_map_mchbar(struct pci_dev *pdev)
+{
+ union {
+ u64 mchbar;
+ struct {
+ u32 mchbar_low;
+ u32 mchbar_high;
+ };
+ } u;
+ void __iomem *window;
+
+ pci_read_config_dword(pdev, X38_MCHBAR_LOW, &u.mchbar_low);
+ pci_write_config_dword(pdev, X38_MCHBAR_LOW, u.mchbar_low | 0x1);
+ pci_read_config_dword(pdev, X38_MCHBAR_HIGH, &u.mchbar_high);
+ u.mchbar &= X38_MCHBAR_MASK;
+
+ if (u.mchbar != (resource_size_t)u.mchbar) {
+ printk(KERN_ERR
+ "x38: mmio space beyond accessible range (0x%llx)\n",
+ (unsigned long long)u.mchbar);
+ return NULL;
+ }
+
+ window = ioremap(u.mchbar, X38_MMR_WINDOW_SIZE);
+ if (!window)
+ printk(KERN_ERR "x38: cannot map mmio space at 0x%llx\n",
+ (unsigned long long)u.mchbar);
+
+ return window;
+}
+
+
+static void x38_get_drbs(void __iomem *window,
+ u16 drbs[X38_CHANNELS][X38_RANKS_PER_CHANNEL])
+{
+ int i;
+
+ for (i = 0; i < X38_RANKS_PER_CHANNEL; i++) {
+ drbs[0][i] = readw(window + X38_C0DRB + 2*i) & X38_DRB_MASK;
+ drbs[1][i] = readw(window + X38_C1DRB + 2*i) & X38_DRB_MASK;
+ }
+}
+
+static bool x38_is_stacked(struct pci_dev *pdev,
+ u16 drbs[X38_CHANNELS][X38_RANKS_PER_CHANNEL])
+{
+ u16 tom;
+
+ pci_read_config_word(pdev, X38_TOM, &tom);
+ tom &= X38_TOM_MASK;
+
+ return drbs[X38_CHANNELS - 1][X38_RANKS_PER_CHANNEL - 1] == tom;
+}
+
+static unsigned long drb_to_nr_pages(
+ u16 drbs[X38_CHANNELS][X38_RANKS_PER_CHANNEL],
+ bool stacked, int channel, int rank)
+{
+ int n;
+
+ n = drbs[channel][rank];
+ if (rank > 0)
+ n -= drbs[channel][rank - 1];
+ if (stacked && (channel == 1) && drbs[channel][rank] ==
+ drbs[channel][X38_RANKS_PER_CHANNEL - 1]) {
+ n -= drbs[0][X38_RANKS_PER_CHANNEL - 1];
+ }
+
+ n <<= (X38_DRB_SHIFT - PAGE_SHIFT);
+ return n;
+}
+
+static int x38_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ int rc;
+ int i, j;
+ struct mem_ctl_info *mci = NULL;
+ struct edac_mc_layer layers[2];
+ u16 drbs[X38_CHANNELS][X38_RANKS_PER_CHANNEL];
+ bool stacked;
+ void __iomem *window;
+
+ edac_dbg(0, "MC:\n");
+
+ window = x38_map_mchbar(pdev);
+ if (!window)
+ return -ENODEV;
+
+ x38_get_drbs(window, drbs);
+
+ how_many_channel(pdev);
+
+ /* FIXME: unconventional pvt_info usage */
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = X38_RANKS;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = x38_channel_num;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, 0);
+ if (!mci)
+ return -ENOMEM;
+
+ edac_dbg(3, "MC: init mci\n");
+
+ mci->pdev = &pdev->dev;
+ mci->mtype_cap = MEM_FLAG_DDR2;
+
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_name = x38_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->edac_check = x38_check;
+ mci->ctl_page_to_phys = NULL;
+ mci->pvt_info = window;
+
+ stacked = x38_is_stacked(pdev, drbs);
+
+ /*
+ * The dram rank boundary (DRB) reg values are boundary addresses
+ * for each DRAM rank with a granularity of 64MB. DRB regs are
+ * cumulative; the last one will contain the total memory
+ * contained in all ranks.
+ */
+ for (i = 0; i < mci->nr_csrows; i++) {
+ unsigned long nr_pages;
+ struct csrow_info *csrow = mci->csrows[i];
+
+ nr_pages = drb_to_nr_pages(drbs, stacked,
+ i / X38_RANKS_PER_CHANNEL,
+ i % X38_RANKS_PER_CHANNEL);
+
+ if (nr_pages == 0)
+ continue;
+
+ for (j = 0; j < x38_channel_num; j++) {
+ struct dimm_info *dimm = csrow->channels[j]->dimm;
+
+ dimm->nr_pages = nr_pages / x38_channel_num;
+ dimm->grain = nr_pages << PAGE_SHIFT;
+ dimm->mtype = MEM_DDR2;
+ dimm->dtype = DEV_UNKNOWN;
+ dimm->edac_mode = EDAC_UNKNOWN;
+ }
+ }
+
+ x38_clear_error_info(mci);
+
+ rc = -ENODEV;
+ if (edac_mc_add_mc(mci)) {
+ edac_dbg(3, "MC: failed edac_mc_add_mc()\n");
+ goto fail;
+ }
+
+ /* get this far and it's successful */
+ edac_dbg(3, "MC: success\n");
+ return 0;
+
+fail:
+ iounmap(window);
+ if (mci)
+ edac_mc_free(mci);
+
+ return rc;
+}
+
+static int x38_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ int rc;
+
+ edac_dbg(0, "MC:\n");
+
+ if (pci_enable_device(pdev) < 0)
+ return -EIO;
+
+ rc = x38_probe1(pdev, ent->driver_data);
+ if (!mci_pdev)
+ mci_pdev = pci_dev_get(pdev);
+
+ return rc;
+}
+
+static void x38_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ edac_dbg(0, "\n");
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ iounmap(mci->pvt_info);
+
+ edac_mc_free(mci);
+}
+
+static const struct pci_device_id x38_pci_tbl[] = {
+ {
+ PCI_VEND_DEV(INTEL, X38_HB), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
+ X38},
+ {
+ 0,
+ } /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, x38_pci_tbl);
+
+static struct pci_driver x38_driver = {
+ .name = EDAC_MOD_STR,
+ .probe = x38_init_one,
+ .remove = x38_remove_one,
+ .id_table = x38_pci_tbl,
+};
+
+static int __init x38_init(void)
+{
+ int pci_rc;
+
+ edac_dbg(3, "MC:\n");
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&x38_driver);
+ if (pci_rc < 0)
+ goto fail0;
+
+ if (!mci_pdev) {
+ x38_registered = 0;
+ mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_X38_HB, NULL);
+ if (!mci_pdev) {
+ edac_dbg(0, "x38 pci_get_device fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+
+ pci_rc = x38_init_one(mci_pdev, x38_pci_tbl);
+ if (pci_rc < 0) {
+ edac_dbg(0, "x38 init fail\n");
+ pci_rc = -ENODEV;
+ goto fail1;
+ }
+ }
+
+ return 0;
+
+fail1:
+ pci_unregister_driver(&x38_driver);
+
+fail0:
+ pci_dev_put(mci_pdev);
+
+ return pci_rc;
+}
+
+static void __exit x38_exit(void)
+{
+ edac_dbg(3, "MC:\n");
+
+ pci_unregister_driver(&x38_driver);
+ if (!x38_registered) {
+ x38_remove_one(mci_pdev);
+ pci_dev_put(mci_pdev);
+ }
+}
+
+module_init(x38_init);
+module_exit(x38_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cluster Computing, Inc. Hitoshi Mitake");
+MODULE_DESCRIPTION("MC support for Intel X38 memory hub controllers");
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/xgene_edac.c b/drivers/edac/xgene_edac.c
new file mode 100644
index 0000000000..c52b9dd915
--- /dev/null
+++ b/drivers/edac/xgene_edac.c
@@ -0,0 +1,2045 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * APM X-Gene SoC EDAC (error detection and correction)
+ *
+ * Copyright (c) 2015, Applied Micro Circuits Corporation
+ * Author: Feng Kan <fkan@apm.com>
+ * Loc Ho <lho@apm.com>
+ */
+
+#include <linux/ctype.h>
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/mfd/syscon.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/regmap.h>
+
+#include "edac_module.h"
+
+#define EDAC_MOD_STR "xgene_edac"
+
+/* Global error configuration status registers (CSR) */
+#define PCPHPERRINTSTS 0x0000
+#define PCPHPERRINTMSK 0x0004
+#define MCU_CTL_ERR_MASK BIT(12)
+#define IOB_PA_ERR_MASK BIT(11)
+#define IOB_BA_ERR_MASK BIT(10)
+#define IOB_XGIC_ERR_MASK BIT(9)
+#define IOB_RB_ERR_MASK BIT(8)
+#define L3C_UNCORR_ERR_MASK BIT(5)
+#define MCU_UNCORR_ERR_MASK BIT(4)
+#define PMD3_MERR_MASK BIT(3)
+#define PMD2_MERR_MASK BIT(2)
+#define PMD1_MERR_MASK BIT(1)
+#define PMD0_MERR_MASK BIT(0)
+#define PCPLPERRINTSTS 0x0008
+#define PCPLPERRINTMSK 0x000C
+#define CSW_SWITCH_TRACE_ERR_MASK BIT(2)
+#define L3C_CORR_ERR_MASK BIT(1)
+#define MCU_CORR_ERR_MASK BIT(0)
+#define MEMERRINTSTS 0x0010
+#define MEMERRINTMSK 0x0014
+
+struct xgene_edac {
+ struct device *dev;
+ struct regmap *csw_map;
+ struct regmap *mcba_map;
+ struct regmap *mcbb_map;
+ struct regmap *efuse_map;
+ struct regmap *rb_map;
+ void __iomem *pcp_csr;
+ spinlock_t lock;
+ struct dentry *dfs;
+
+ struct list_head mcus;
+ struct list_head pmds;
+ struct list_head l3s;
+ struct list_head socs;
+
+ struct mutex mc_lock;
+ int mc_active_mask;
+ int mc_registered_mask;
+};
+
+static void xgene_edac_pcp_rd(struct xgene_edac *edac, u32 reg, u32 *val)
+{
+ *val = readl(edac->pcp_csr + reg);
+}
+
+static void xgene_edac_pcp_clrbits(struct xgene_edac *edac, u32 reg,
+ u32 bits_mask)
+{
+ u32 val;
+
+ spin_lock(&edac->lock);
+ val = readl(edac->pcp_csr + reg);
+ val &= ~bits_mask;
+ writel(val, edac->pcp_csr + reg);
+ spin_unlock(&edac->lock);
+}
+
+static void xgene_edac_pcp_setbits(struct xgene_edac *edac, u32 reg,
+ u32 bits_mask)
+{
+ u32 val;
+
+ spin_lock(&edac->lock);
+ val = readl(edac->pcp_csr + reg);
+ val |= bits_mask;
+ writel(val, edac->pcp_csr + reg);
+ spin_unlock(&edac->lock);
+}
+
+/* Memory controller error CSR */
+#define MCU_MAX_RANK 8
+#define MCU_RANK_STRIDE 0x40
+
+#define MCUGECR 0x0110
+#define MCU_GECR_DEMANDUCINTREN_MASK BIT(0)
+#define MCU_GECR_BACKUCINTREN_MASK BIT(1)
+#define MCU_GECR_CINTREN_MASK BIT(2)
+#define MUC_GECR_MCUADDRERREN_MASK BIT(9)
+#define MCUGESR 0x0114
+#define MCU_GESR_ADDRNOMATCH_ERR_MASK BIT(7)
+#define MCU_GESR_ADDRMULTIMATCH_ERR_MASK BIT(6)
+#define MCU_GESR_PHYP_ERR_MASK BIT(3)
+#define MCUESRR0 0x0314
+#define MCU_ESRR_MULTUCERR_MASK BIT(3)
+#define MCU_ESRR_BACKUCERR_MASK BIT(2)
+#define MCU_ESRR_DEMANDUCERR_MASK BIT(1)
+#define MCU_ESRR_CERR_MASK BIT(0)
+#define MCUESRRA0 0x0318
+#define MCUEBLRR0 0x031c
+#define MCU_EBLRR_ERRBANK_RD(src) (((src) & 0x00000007) >> 0)
+#define MCUERCRR0 0x0320
+#define MCU_ERCRR_ERRROW_RD(src) (((src) & 0xFFFF0000) >> 16)
+#define MCU_ERCRR_ERRCOL_RD(src) ((src) & 0x00000FFF)
+#define MCUSBECNT0 0x0324
+#define MCU_SBECNT_COUNT(src) ((src) & 0xFFFF)
+
+#define CSW_CSWCR 0x0000
+#define CSW_CSWCR_DUALMCB_MASK BIT(0)
+
+#define MCBADDRMR 0x0000
+#define MCBADDRMR_MCU_INTLV_MODE_MASK BIT(3)
+#define MCBADDRMR_DUALMCU_MODE_MASK BIT(2)
+#define MCBADDRMR_MCB_INTLV_MODE_MASK BIT(1)
+#define MCBADDRMR_ADDRESS_MODE_MASK BIT(0)
+
+struct xgene_edac_mc_ctx {
+ struct list_head next;
+ char *name;
+ struct mem_ctl_info *mci;
+ struct xgene_edac *edac;
+ void __iomem *mcu_csr;
+ u32 mcu_id;
+};
+
+static ssize_t xgene_edac_mc_err_inject_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct mem_ctl_info *mci = file->private_data;
+ struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
+ int i;
+
+ for (i = 0; i < MCU_MAX_RANK; i++) {
+ writel(MCU_ESRR_MULTUCERR_MASK | MCU_ESRR_BACKUCERR_MASK |
+ MCU_ESRR_DEMANDUCERR_MASK | MCU_ESRR_CERR_MASK,
+ ctx->mcu_csr + MCUESRRA0 + i * MCU_RANK_STRIDE);
+ }
+ return count;
+}
+
+static const struct file_operations xgene_edac_mc_debug_inject_fops = {
+ .open = simple_open,
+ .write = xgene_edac_mc_err_inject_write,
+ .llseek = generic_file_llseek,
+};
+
+static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci)
+{
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+ return;
+
+ if (!mci->debugfs)
+ return;
+
+ edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
+ &xgene_edac_mc_debug_inject_fops);
+}
+
+static void xgene_edac_mc_check(struct mem_ctl_info *mci)
+{
+ struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
+ unsigned int pcp_hp_stat;
+ unsigned int pcp_lp_stat;
+ u32 reg;
+ u32 rank;
+ u32 bank;
+ u32 count;
+ u32 col_row;
+
+ xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
+ xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat);
+ if (!((MCU_UNCORR_ERR_MASK & pcp_hp_stat) ||
+ (MCU_CTL_ERR_MASK & pcp_hp_stat) ||
+ (MCU_CORR_ERR_MASK & pcp_lp_stat)))
+ return;
+
+ for (rank = 0; rank < MCU_MAX_RANK; rank++) {
+ reg = readl(ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE);
+
+ /* Detect uncorrectable memory error */
+ if (reg & (MCU_ESRR_DEMANDUCERR_MASK |
+ MCU_ESRR_BACKUCERR_MASK)) {
+ /* Detected uncorrectable memory error */
+ edac_mc_chipset_printk(mci, KERN_ERR, "X-Gene",
+ "MCU uncorrectable error at rank %d\n", rank);
+
+ edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci,
+ 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
+ }
+
+ /* Detect correctable memory error */
+ if (reg & MCU_ESRR_CERR_MASK) {
+ bank = readl(ctx->mcu_csr + MCUEBLRR0 +
+ rank * MCU_RANK_STRIDE);
+ col_row = readl(ctx->mcu_csr + MCUERCRR0 +
+ rank * MCU_RANK_STRIDE);
+ count = readl(ctx->mcu_csr + MCUSBECNT0 +
+ rank * MCU_RANK_STRIDE);
+ edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
+ "MCU correctable error at rank %d bank %d column %d row %d count %d\n",
+ rank, MCU_EBLRR_ERRBANK_RD(bank),
+ MCU_ERCRR_ERRCOL_RD(col_row),
+ MCU_ERCRR_ERRROW_RD(col_row),
+ MCU_SBECNT_COUNT(count));
+
+ edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci,
+ 1, 0, 0, 0, 0, 0, -1, mci->ctl_name, "");
+ }
+
+ /* Clear all error registers */
+ writel(0x0, ctx->mcu_csr + MCUEBLRR0 + rank * MCU_RANK_STRIDE);
+ writel(0x0, ctx->mcu_csr + MCUERCRR0 + rank * MCU_RANK_STRIDE);
+ writel(0x0, ctx->mcu_csr + MCUSBECNT0 +
+ rank * MCU_RANK_STRIDE);
+ writel(reg, ctx->mcu_csr + MCUESRR0 + rank * MCU_RANK_STRIDE);
+ }
+
+ /* Detect memory controller error */
+ reg = readl(ctx->mcu_csr + MCUGESR);
+ if (reg) {
+ if (reg & MCU_GESR_ADDRNOMATCH_ERR_MASK)
+ edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
+ "MCU address miss-match error\n");
+ if (reg & MCU_GESR_ADDRMULTIMATCH_ERR_MASK)
+ edac_mc_chipset_printk(mci, KERN_WARNING, "X-Gene",
+ "MCU address multi-match error\n");
+
+ writel(reg, ctx->mcu_csr + MCUGESR);
+ }
+}
+
+static void xgene_edac_mc_irq_ctl(struct mem_ctl_info *mci, bool enable)
+{
+ struct xgene_edac_mc_ctx *ctx = mci->pvt_info;
+ unsigned int val;
+
+ if (edac_op_state != EDAC_OPSTATE_INT)
+ return;
+
+ mutex_lock(&ctx->edac->mc_lock);
+
+ /*
+ * As there is only single bit for enable error and interrupt mask,
+ * we must only enable top level interrupt after all MCUs are
+ * registered. Otherwise, if there is an error and the corresponding
+ * MCU has not registered, the interrupt will never get cleared. To
+ * determine all MCU have registered, we will keep track of active
+ * MCUs and registered MCUs.
+ */
+ if (enable) {
+ /* Set registered MCU bit */
+ ctx->edac->mc_registered_mask |= 1 << ctx->mcu_id;
+
+ /* Enable interrupt after all active MCU registered */
+ if (ctx->edac->mc_registered_mask ==
+ ctx->edac->mc_active_mask) {
+ /* Enable memory controller top level interrupt */
+ xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+ MCU_UNCORR_ERR_MASK |
+ MCU_CTL_ERR_MASK);
+ xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
+ MCU_CORR_ERR_MASK);
+ }
+
+ /* Enable MCU interrupt and error reporting */
+ val = readl(ctx->mcu_csr + MCUGECR);
+ val |= MCU_GECR_DEMANDUCINTREN_MASK |
+ MCU_GECR_BACKUCINTREN_MASK |
+ MCU_GECR_CINTREN_MASK |
+ MUC_GECR_MCUADDRERREN_MASK;
+ writel(val, ctx->mcu_csr + MCUGECR);
+ } else {
+ /* Disable MCU interrupt */
+ val = readl(ctx->mcu_csr + MCUGECR);
+ val &= ~(MCU_GECR_DEMANDUCINTREN_MASK |
+ MCU_GECR_BACKUCINTREN_MASK |
+ MCU_GECR_CINTREN_MASK |
+ MUC_GECR_MCUADDRERREN_MASK);
+ writel(val, ctx->mcu_csr + MCUGECR);
+
+ /* Disable memory controller top level interrupt */
+ xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+ MCU_UNCORR_ERR_MASK | MCU_CTL_ERR_MASK);
+ xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
+ MCU_CORR_ERR_MASK);
+
+ /* Clear registered MCU bit */
+ ctx->edac->mc_registered_mask &= ~(1 << ctx->mcu_id);
+ }
+
+ mutex_unlock(&ctx->edac->mc_lock);
+}
+
+static int xgene_edac_mc_is_active(struct xgene_edac_mc_ctx *ctx, int mc_idx)
+{
+ unsigned int reg;
+ u32 mcu_mask;
+
+ if (regmap_read(ctx->edac->csw_map, CSW_CSWCR, &reg))
+ return 0;
+
+ if (reg & CSW_CSWCR_DUALMCB_MASK) {
+ /*
+ * Dual MCB active - Determine if all 4 active or just MCU0
+ * and MCU2 active
+ */
+ if (regmap_read(ctx->edac->mcbb_map, MCBADDRMR, &reg))
+ return 0;
+ mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0xF : 0x5;
+ } else {
+ /*
+ * Single MCB active - Determine if MCU0/MCU1 or just MCU0
+ * active
+ */
+ if (regmap_read(ctx->edac->mcba_map, MCBADDRMR, &reg))
+ return 0;
+ mcu_mask = (reg & MCBADDRMR_DUALMCU_MODE_MASK) ? 0x3 : 0x1;
+ }
+
+ /* Save active MC mask if hasn't set already */
+ if (!ctx->edac->mc_active_mask)
+ ctx->edac->mc_active_mask = mcu_mask;
+
+ return (mcu_mask & (1 << mc_idx)) ? 1 : 0;
+}
+
+static int xgene_edac_mc_add(struct xgene_edac *edac, struct device_node *np)
+{
+ struct mem_ctl_info *mci;
+ struct edac_mc_layer layers[2];
+ struct xgene_edac_mc_ctx tmp_ctx;
+ struct xgene_edac_mc_ctx *ctx;
+ struct resource res;
+ int rc;
+
+ memset(&tmp_ctx, 0, sizeof(tmp_ctx));
+ tmp_ctx.edac = edac;
+
+ if (!devres_open_group(edac->dev, xgene_edac_mc_add, GFP_KERNEL))
+ return -ENOMEM;
+
+ rc = of_address_to_resource(np, 0, &res);
+ if (rc < 0) {
+ dev_err(edac->dev, "no MCU resource address\n");
+ goto err_group;
+ }
+ tmp_ctx.mcu_csr = devm_ioremap_resource(edac->dev, &res);
+ if (IS_ERR(tmp_ctx.mcu_csr)) {
+ dev_err(edac->dev, "unable to map MCU resource\n");
+ rc = PTR_ERR(tmp_ctx.mcu_csr);
+ goto err_group;
+ }
+
+ /* Ignore non-active MCU */
+ if (of_property_read_u32(np, "memory-controller", &tmp_ctx.mcu_id)) {
+ dev_err(edac->dev, "no memory-controller property\n");
+ rc = -ENODEV;
+ goto err_group;
+ }
+ if (!xgene_edac_mc_is_active(&tmp_ctx, tmp_ctx.mcu_id)) {
+ rc = -ENODEV;
+ goto err_group;
+ }
+
+ layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
+ layers[0].size = 4;
+ layers[0].is_virt_csrow = true;
+ layers[1].type = EDAC_MC_LAYER_CHANNEL;
+ layers[1].size = 2;
+ layers[1].is_virt_csrow = false;
+ mci = edac_mc_alloc(tmp_ctx.mcu_id, ARRAY_SIZE(layers), layers,
+ sizeof(*ctx));
+ if (!mci) {
+ rc = -ENOMEM;
+ goto err_group;
+ }
+
+ ctx = mci->pvt_info;
+ *ctx = tmp_ctx; /* Copy over resource value */
+ ctx->name = "xgene_edac_mc_err";
+ ctx->mci = mci;
+ mci->pdev = &mci->dev;
+ mci->ctl_name = ctx->name;
+ mci->dev_name = ctx->name;
+
+ mci->mtype_cap = MEM_FLAG_RDDR | MEM_FLAG_RDDR2 | MEM_FLAG_RDDR3 |
+ MEM_FLAG_DDR | MEM_FLAG_DDR2 | MEM_FLAG_DDR3;
+ mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+ mci->edac_cap = EDAC_FLAG_SECDED;
+ mci->mod_name = EDAC_MOD_STR;
+ mci->ctl_page_to_phys = NULL;
+ mci->scrub_cap = SCRUB_FLAG_HW_SRC;
+ mci->scrub_mode = SCRUB_HW_SRC;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ mci->edac_check = xgene_edac_mc_check;
+
+ if (edac_mc_add_mc(mci)) {
+ dev_err(edac->dev, "edac_mc_add_mc failed\n");
+ rc = -EINVAL;
+ goto err_free;
+ }
+
+ xgene_edac_mc_create_debugfs_node(mci);
+
+ list_add(&ctx->next, &edac->mcus);
+
+ xgene_edac_mc_irq_ctl(mci, true);
+
+ devres_remove_group(edac->dev, xgene_edac_mc_add);
+
+ dev_info(edac->dev, "X-Gene EDAC MC registered\n");
+ return 0;
+
+err_free:
+ edac_mc_free(mci);
+err_group:
+ devres_release_group(edac->dev, xgene_edac_mc_add);
+ return rc;
+}
+
+static int xgene_edac_mc_remove(struct xgene_edac_mc_ctx *mcu)
+{
+ xgene_edac_mc_irq_ctl(mcu->mci, false);
+ edac_mc_del_mc(&mcu->mci->dev);
+ edac_mc_free(mcu->mci);
+ return 0;
+}
+
+/* CPU L1/L2 error CSR */
+#define MAX_CPU_PER_PMD 2
+#define CPU_CSR_STRIDE 0x00100000
+#define CPU_L2C_PAGE 0x000D0000
+#define CPU_MEMERR_L2C_PAGE 0x000E0000
+#define CPU_MEMERR_CPU_PAGE 0x000F0000
+
+#define MEMERR_CPU_ICFECR_PAGE_OFFSET 0x0000
+#define MEMERR_CPU_ICFESR_PAGE_OFFSET 0x0004
+#define MEMERR_CPU_ICFESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
+#define MEMERR_CPU_ICFESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
+#define MEMERR_CPU_ICFESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
+#define MEMERR_CPU_ICFESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
+#define MEMERR_CPU_ICFESR_MULTCERR_MASK BIT(2)
+#define MEMERR_CPU_ICFESR_CERR_MASK BIT(0)
+#define MEMERR_CPU_LSUESR_PAGE_OFFSET 0x000c
+#define MEMERR_CPU_LSUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
+#define MEMERR_CPU_LSUESR_ERRINDEX_RD(src) (((src) & 0x003F0000) >> 16)
+#define MEMERR_CPU_LSUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
+#define MEMERR_CPU_LSUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
+#define MEMERR_CPU_LSUESR_MULTCERR_MASK BIT(2)
+#define MEMERR_CPU_LSUESR_CERR_MASK BIT(0)
+#define MEMERR_CPU_LSUECR_PAGE_OFFSET 0x0008
+#define MEMERR_CPU_MMUECR_PAGE_OFFSET 0x0010
+#define MEMERR_CPU_MMUESR_PAGE_OFFSET 0x0014
+#define MEMERR_CPU_MMUESR_ERRWAY_RD(src) (((src) & 0xFF000000) >> 24)
+#define MEMERR_CPU_MMUESR_ERRINDEX_RD(src) (((src) & 0x007F0000) >> 16)
+#define MEMERR_CPU_MMUESR_ERRINFO_RD(src) (((src) & 0x0000FF00) >> 8)
+#define MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK BIT(7)
+#define MEMERR_CPU_MMUESR_ERRTYPE_RD(src) (((src) & 0x00000070) >> 4)
+#define MEMERR_CPU_MMUESR_MULTCERR_MASK BIT(2)
+#define MEMERR_CPU_MMUESR_CERR_MASK BIT(0)
+#define MEMERR_CPU_ICFESRA_PAGE_OFFSET 0x0804
+#define MEMERR_CPU_LSUESRA_PAGE_OFFSET 0x080c
+#define MEMERR_CPU_MMUESRA_PAGE_OFFSET 0x0814
+
+#define MEMERR_L2C_L2ECR_PAGE_OFFSET 0x0000
+#define MEMERR_L2C_L2ESR_PAGE_OFFSET 0x0004
+#define MEMERR_L2C_L2ESR_ERRSYN_RD(src) (((src) & 0xFF000000) >> 24)
+#define MEMERR_L2C_L2ESR_ERRWAY_RD(src) (((src) & 0x00FC0000) >> 18)
+#define MEMERR_L2C_L2ESR_ERRCPU_RD(src) (((src) & 0x00020000) >> 17)
+#define MEMERR_L2C_L2ESR_ERRGROUP_RD(src) (((src) & 0x0000E000) >> 13)
+#define MEMERR_L2C_L2ESR_ERRACTION_RD(src) (((src) & 0x00001C00) >> 10)
+#define MEMERR_L2C_L2ESR_ERRTYPE_RD(src) (((src) & 0x00000300) >> 8)
+#define MEMERR_L2C_L2ESR_MULTUCERR_MASK BIT(3)
+#define MEMERR_L2C_L2ESR_MULTICERR_MASK BIT(2)
+#define MEMERR_L2C_L2ESR_UCERR_MASK BIT(1)
+#define MEMERR_L2C_L2ESR_ERR_MASK BIT(0)
+#define MEMERR_L2C_L2EALR_PAGE_OFFSET 0x0008
+#define CPUX_L2C_L2RTOCR_PAGE_OFFSET 0x0010
+#define MEMERR_L2C_L2EAHR_PAGE_OFFSET 0x000c
+#define CPUX_L2C_L2RTOSR_PAGE_OFFSET 0x0014
+#define MEMERR_L2C_L2RTOSR_MULTERR_MASK BIT(1)
+#define MEMERR_L2C_L2RTOSR_ERR_MASK BIT(0)
+#define CPUX_L2C_L2RTOALR_PAGE_OFFSET 0x0018
+#define CPUX_L2C_L2RTOAHR_PAGE_OFFSET 0x001c
+#define MEMERR_L2C_L2ESRA_PAGE_OFFSET 0x0804
+
+/*
+ * Processor Module Domain (PMD) context - Context for a pair of processors.
+ * Each PMD consists of 2 CPUs and a shared L2 cache. Each CPU consists of
+ * its own L1 cache.
+ */
+struct xgene_edac_pmd_ctx {
+ struct list_head next;
+ struct device ddev;
+ char *name;
+ struct xgene_edac *edac;
+ struct edac_device_ctl_info *edac_dev;
+ void __iomem *pmd_csr;
+ u32 pmd;
+ int version;
+};
+
+static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev,
+ int cpu_idx)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *pg_f;
+ u32 val;
+
+ pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE;
+
+ val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
+ if (!val)
+ goto chk_lsu;
+ dev_err(edac_dev->dev,
+ "CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
+ ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+ MEMERR_CPU_ICFESR_ERRWAY_RD(val),
+ MEMERR_CPU_ICFESR_ERRINDEX_RD(val),
+ MEMERR_CPU_ICFESR_ERRINFO_RD(val));
+ if (val & MEMERR_CPU_ICFESR_CERR_MASK)
+ dev_err(edac_dev->dev, "One or more correctable error\n");
+ if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK)
+ dev_err(edac_dev->dev, "Multiple correctable error\n");
+ switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) {
+ case 1:
+ dev_err(edac_dev->dev, "L1 TLB multiple hit\n");
+ break;
+ case 2:
+ dev_err(edac_dev->dev, "Way select multiple hit\n");
+ break;
+ case 3:
+ dev_err(edac_dev->dev, "Physical tag parity error\n");
+ break;
+ case 4:
+ case 5:
+ dev_err(edac_dev->dev, "L1 data parity error\n");
+ break;
+ case 6:
+ dev_err(edac_dev->dev, "L1 pre-decode parity error\n");
+ break;
+ }
+
+ /* Clear any HW errors */
+ writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
+
+ if (val & (MEMERR_CPU_ICFESR_CERR_MASK |
+ MEMERR_CPU_ICFESR_MULTCERR_MASK))
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+
+chk_lsu:
+ val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
+ if (!val)
+ goto chk_mmu;
+ dev_err(edac_dev->dev,
+ "CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
+ ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+ MEMERR_CPU_LSUESR_ERRWAY_RD(val),
+ MEMERR_CPU_LSUESR_ERRINDEX_RD(val),
+ MEMERR_CPU_LSUESR_ERRINFO_RD(val));
+ if (val & MEMERR_CPU_LSUESR_CERR_MASK)
+ dev_err(edac_dev->dev, "One or more correctable error\n");
+ if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK)
+ dev_err(edac_dev->dev, "Multiple correctable error\n");
+ switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) {
+ case 0:
+ dev_err(edac_dev->dev, "Load tag error\n");
+ break;
+ case 1:
+ dev_err(edac_dev->dev, "Load data error\n");
+ break;
+ case 2:
+ dev_err(edac_dev->dev, "WSL multihit error\n");
+ break;
+ case 3:
+ dev_err(edac_dev->dev, "Store tag error\n");
+ break;
+ case 4:
+ dev_err(edac_dev->dev,
+ "DTB multihit from load pipeline error\n");
+ break;
+ case 5:
+ dev_err(edac_dev->dev,
+ "DTB multihit from store pipeline error\n");
+ break;
+ }
+
+ /* Clear any HW errors */
+ writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
+
+ if (val & (MEMERR_CPU_LSUESR_CERR_MASK |
+ MEMERR_CPU_LSUESR_MULTCERR_MASK))
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+
+chk_mmu:
+ val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
+ if (!val)
+ return;
+ dev_err(edac_dev->dev,
+ "CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n",
+ ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+ MEMERR_CPU_MMUESR_ERRWAY_RD(val),
+ MEMERR_CPU_MMUESR_ERRINDEX_RD(val),
+ MEMERR_CPU_MMUESR_ERRINFO_RD(val),
+ val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" : "ICF");
+ if (val & MEMERR_CPU_MMUESR_CERR_MASK)
+ dev_err(edac_dev->dev, "One or more correctable error\n");
+ if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK)
+ dev_err(edac_dev->dev, "Multiple correctable error\n");
+ switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) {
+ case 0:
+ dev_err(edac_dev->dev, "Stage 1 UTB hit error\n");
+ break;
+ case 1:
+ dev_err(edac_dev->dev, "Stage 1 UTB miss error\n");
+ break;
+ case 2:
+ dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n");
+ break;
+ case 3:
+ dev_err(edac_dev->dev, "TMO operation single bank error\n");
+ break;
+ case 4:
+ dev_err(edac_dev->dev, "Stage 2 UTB error\n");
+ break;
+ case 5:
+ dev_err(edac_dev->dev, "Stage 2 UTB miss error\n");
+ break;
+ case 6:
+ dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n");
+ break;
+ case 7:
+ dev_err(edac_dev->dev, "TMO operation multiple bank error\n");
+ break;
+ }
+
+ /* Clear any HW errors */
+ writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
+
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *pg_d;
+ void __iomem *pg_e;
+ u32 val_hi;
+ u32 val_lo;
+ u32 val;
+
+ /* Check L2 */
+ pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
+ val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
+ if (!val)
+ goto chk_l2c;
+ val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET);
+ val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET);
+ dev_err(edac_dev->dev,
+ "PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n",
+ ctx->pmd, val, val_hi, val_lo);
+ dev_err(edac_dev->dev,
+ "ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n",
+ MEMERR_L2C_L2ESR_ERRSYN_RD(val),
+ MEMERR_L2C_L2ESR_ERRWAY_RD(val),
+ MEMERR_L2C_L2ESR_ERRCPU_RD(val),
+ MEMERR_L2C_L2ESR_ERRGROUP_RD(val),
+ MEMERR_L2C_L2ESR_ERRACTION_RD(val));
+
+ if (val & MEMERR_L2C_L2ESR_ERR_MASK)
+ dev_err(edac_dev->dev, "One or more correctable error\n");
+ if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK)
+ dev_err(edac_dev->dev, "Multiple correctable error\n");
+ if (val & MEMERR_L2C_L2ESR_UCERR_MASK)
+ dev_err(edac_dev->dev, "One or more uncorrectable error\n");
+ if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK)
+ dev_err(edac_dev->dev, "Multiple uncorrectable error\n");
+
+ switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) {
+ case 0:
+ dev_err(edac_dev->dev, "Outbound SDB parity error\n");
+ break;
+ case 1:
+ dev_err(edac_dev->dev, "Inbound SDB parity error\n");
+ break;
+ case 2:
+ dev_err(edac_dev->dev, "Tag ECC error\n");
+ break;
+ case 3:
+ dev_err(edac_dev->dev, "Data ECC error\n");
+ break;
+ }
+
+ /* Clear any HW errors */
+ writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
+
+ if (val & (MEMERR_L2C_L2ESR_ERR_MASK |
+ MEMERR_L2C_L2ESR_MULTICERR_MASK))
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+ if (val & (MEMERR_L2C_L2ESR_UCERR_MASK |
+ MEMERR_L2C_L2ESR_MULTUCERR_MASK))
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+
+chk_l2c:
+ /* Check if any memory request timed out on L2 cache */
+ pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
+ val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
+ if (val) {
+ val_lo = readl(pg_d + CPUX_L2C_L2RTOALR_PAGE_OFFSET);
+ val_hi = readl(pg_d + CPUX_L2C_L2RTOAHR_PAGE_OFFSET);
+ dev_err(edac_dev->dev,
+ "PMD%d L2C error L2C RTOSR 0x%08X @ 0x%08X.%08X\n",
+ ctx->pmd, val, val_hi, val_lo);
+ writel(val, pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
+ }
+}
+
+static void xgene_edac_pmd_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ unsigned int pcp_hp_stat;
+ int i;
+
+ xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
+ if (!((PMD0_MERR_MASK << ctx->pmd) & pcp_hp_stat))
+ return;
+
+ /* Check CPU L1 error */
+ for (i = 0; i < MAX_CPU_PER_PMD; i++)
+ xgene_edac_pmd_l1_check(edac_dev, i);
+
+ /* Check CPU L2 error */
+ xgene_edac_pmd_l2_check(edac_dev);
+}
+
+static void xgene_edac_pmd_cpu_hw_cfg(struct edac_device_ctl_info *edac_dev,
+ int cpu)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *pg_f = ctx->pmd_csr + cpu * CPU_CSR_STRIDE +
+ CPU_MEMERR_CPU_PAGE;
+
+ /*
+ * Enable CPU memory error:
+ * MEMERR_CPU_ICFESRA, MEMERR_CPU_LSUESRA, and MEMERR_CPU_MMUESRA
+ */
+ writel(0x00000301, pg_f + MEMERR_CPU_ICFECR_PAGE_OFFSET);
+ writel(0x00000301, pg_f + MEMERR_CPU_LSUECR_PAGE_OFFSET);
+ writel(0x00000101, pg_f + MEMERR_CPU_MMUECR_PAGE_OFFSET);
+}
+
+static void xgene_edac_pmd_hw_cfg(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
+ void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
+
+ /* Enable PMD memory error - MEMERR_L2C_L2ECR and L2C_L2RTOCR */
+ writel(0x00000703, pg_e + MEMERR_L2C_L2ECR_PAGE_OFFSET);
+ /* Configure L2C HW request time out feature if supported */
+ if (ctx->version > 1)
+ writel(0x00000119, pg_d + CPUX_L2C_L2RTOCR_PAGE_OFFSET);
+}
+
+static void xgene_edac_pmd_hw_ctl(struct edac_device_ctl_info *edac_dev,
+ bool enable)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ int i;
+
+ /* Enable PMD error interrupt */
+ if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+ if (enable)
+ xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+ PMD0_MERR_MASK << ctx->pmd);
+ else
+ xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+ PMD0_MERR_MASK << ctx->pmd);
+ }
+
+ if (enable) {
+ xgene_edac_pmd_hw_cfg(edac_dev);
+
+ /* Two CPUs per a PMD */
+ for (i = 0; i < MAX_CPU_PER_PMD; i++)
+ xgene_edac_pmd_cpu_hw_cfg(edac_dev, i);
+ }
+}
+
+static ssize_t xgene_edac_pmd_l1_inject_ctrl_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dev = file->private_data;
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *cpux_pg_f;
+ int i;
+
+ for (i = 0; i < MAX_CPU_PER_PMD; i++) {
+ cpux_pg_f = ctx->pmd_csr + i * CPU_CSR_STRIDE +
+ CPU_MEMERR_CPU_PAGE;
+
+ writel(MEMERR_CPU_ICFESR_MULTCERR_MASK |
+ MEMERR_CPU_ICFESR_CERR_MASK,
+ cpux_pg_f + MEMERR_CPU_ICFESRA_PAGE_OFFSET);
+ writel(MEMERR_CPU_LSUESR_MULTCERR_MASK |
+ MEMERR_CPU_LSUESR_CERR_MASK,
+ cpux_pg_f + MEMERR_CPU_LSUESRA_PAGE_OFFSET);
+ writel(MEMERR_CPU_MMUESR_MULTCERR_MASK |
+ MEMERR_CPU_MMUESR_CERR_MASK,
+ cpux_pg_f + MEMERR_CPU_MMUESRA_PAGE_OFFSET);
+ }
+ return count;
+}
+
+static ssize_t xgene_edac_pmd_l2_inject_ctrl_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dev = file->private_data;
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ void __iomem *pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
+
+ writel(MEMERR_L2C_L2ESR_MULTUCERR_MASK |
+ MEMERR_L2C_L2ESR_MULTICERR_MASK |
+ MEMERR_L2C_L2ESR_UCERR_MASK |
+ MEMERR_L2C_L2ESR_ERR_MASK,
+ pg_e + MEMERR_L2C_L2ESRA_PAGE_OFFSET);
+ return count;
+}
+
+static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = {
+ {
+ .open = simple_open,
+ .write = xgene_edac_pmd_l1_inject_ctrl_write,
+ .llseek = generic_file_llseek, },
+ {
+ .open = simple_open,
+ .write = xgene_edac_pmd_l2_inject_ctrl_write,
+ .llseek = generic_file_llseek, },
+ { }
+};
+
+static void
+xgene_edac_pmd_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
+ struct dentry *dbgfs_dir;
+ char name[10];
+
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs)
+ return;
+
+ snprintf(name, sizeof(name), "PMD%d", ctx->pmd);
+ dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
+ if (!dbgfs_dir)
+ return;
+
+ edac_debugfs_create_file("l1_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
+ &xgene_edac_pmd_debug_inject_fops[0]);
+ edac_debugfs_create_file("l2_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
+ &xgene_edac_pmd_debug_inject_fops[1]);
+}
+
+static int xgene_edac_pmd_available(u32 efuse, int pmd)
+{
+ return (efuse & (1 << pmd)) ? 0 : 1;
+}
+
+static int xgene_edac_pmd_add(struct xgene_edac *edac, struct device_node *np,
+ int version)
+{
+ struct edac_device_ctl_info *edac_dev;
+ struct xgene_edac_pmd_ctx *ctx;
+ struct resource res;
+ char edac_name[10];
+ u32 pmd;
+ int rc;
+ u32 val;
+
+ if (!devres_open_group(edac->dev, xgene_edac_pmd_add, GFP_KERNEL))
+ return -ENOMEM;
+
+ /* Determine if this PMD is disabled */
+ if (of_property_read_u32(np, "pmd-controller", &pmd)) {
+ dev_err(edac->dev, "no pmd-controller property\n");
+ rc = -ENODEV;
+ goto err_group;
+ }
+ rc = regmap_read(edac->efuse_map, 0, &val);
+ if (rc)
+ goto err_group;
+ if (!xgene_edac_pmd_available(val, pmd)) {
+ rc = -ENODEV;
+ goto err_group;
+ }
+
+ snprintf(edac_name, sizeof(edac_name), "l2c%d", pmd);
+ edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
+ edac_name, 1, "l2c", 1, 2, NULL,
+ 0, edac_device_alloc_index());
+ if (!edac_dev) {
+ rc = -ENOMEM;
+ goto err_group;
+ }
+
+ ctx = edac_dev->pvt_info;
+ ctx->name = "xgene_pmd_err";
+ ctx->pmd = pmd;
+ ctx->edac = edac;
+ ctx->edac_dev = edac_dev;
+ ctx->ddev = *edac->dev;
+ ctx->version = version;
+ edac_dev->dev = &ctx->ddev;
+ edac_dev->ctl_name = ctx->name;
+ edac_dev->dev_name = ctx->name;
+ edac_dev->mod_name = EDAC_MOD_STR;
+
+ rc = of_address_to_resource(np, 0, &res);
+ if (rc < 0) {
+ dev_err(edac->dev, "no PMD resource address\n");
+ goto err_free;
+ }
+ ctx->pmd_csr = devm_ioremap_resource(edac->dev, &res);
+ if (IS_ERR(ctx->pmd_csr)) {
+ dev_err(edac->dev,
+ "devm_ioremap_resource failed for PMD resource address\n");
+ rc = PTR_ERR(ctx->pmd_csr);
+ goto err_free;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ edac_dev->edac_check = xgene_edac_pmd_check;
+
+ xgene_edac_pmd_create_debugfs_nodes(edac_dev);
+
+ rc = edac_device_add_device(edac_dev);
+ if (rc > 0) {
+ dev_err(edac->dev, "edac_device_add_device failed\n");
+ rc = -ENOMEM;
+ goto err_free;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT)
+ edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+ list_add(&ctx->next, &edac->pmds);
+
+ xgene_edac_pmd_hw_ctl(edac_dev, 1);
+
+ devres_remove_group(edac->dev, xgene_edac_pmd_add);
+
+ dev_info(edac->dev, "X-Gene EDAC PMD%d registered\n", ctx->pmd);
+ return 0;
+
+err_free:
+ edac_device_free_ctl_info(edac_dev);
+err_group:
+ devres_release_group(edac->dev, xgene_edac_pmd_add);
+ return rc;
+}
+
+static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd)
+{
+ struct edac_device_ctl_info *edac_dev = pmd->edac_dev;
+
+ xgene_edac_pmd_hw_ctl(edac_dev, 0);
+ edac_device_del_device(edac_dev->dev);
+ edac_device_free_ctl_info(edac_dev);
+ return 0;
+}
+
+/* L3 Error device */
+#define L3C_ESR (0x0A * 4)
+#define L3C_ESR_DATATAG_MASK BIT(9)
+#define L3C_ESR_MULTIHIT_MASK BIT(8)
+#define L3C_ESR_UCEVICT_MASK BIT(6)
+#define L3C_ESR_MULTIUCERR_MASK BIT(5)
+#define L3C_ESR_MULTICERR_MASK BIT(4)
+#define L3C_ESR_UCERR_MASK BIT(3)
+#define L3C_ESR_CERR_MASK BIT(2)
+#define L3C_ESR_UCERRINTR_MASK BIT(1)
+#define L3C_ESR_CERRINTR_MASK BIT(0)
+#define L3C_ECR (0x0B * 4)
+#define L3C_ECR_UCINTREN BIT(3)
+#define L3C_ECR_CINTREN BIT(2)
+#define L3C_UCERREN BIT(1)
+#define L3C_CERREN BIT(0)
+#define L3C_ELR (0x0C * 4)
+#define L3C_ELR_ERRSYN(src) ((src & 0xFF800000) >> 23)
+#define L3C_ELR_ERRWAY(src) ((src & 0x007E0000) >> 17)
+#define L3C_ELR_AGENTID(src) ((src & 0x0001E000) >> 13)
+#define L3C_ELR_ERRGRP(src) ((src & 0x00000F00) >> 8)
+#define L3C_ELR_OPTYPE(src) ((src & 0x000000F0) >> 4)
+#define L3C_ELR_PADDRHIGH(src) (src & 0x0000000F)
+#define L3C_AELR (0x0D * 4)
+#define L3C_BELR (0x0E * 4)
+#define L3C_BELR_BANK(src) (src & 0x0000000F)
+
+struct xgene_edac_dev_ctx {
+ struct list_head next;
+ struct device ddev;
+ char *name;
+ struct xgene_edac *edac;
+ struct edac_device_ctl_info *edac_dev;
+ int edac_idx;
+ void __iomem *dev_csr;
+ int version;
+};
+
+/*
+ * Version 1 of the L3 controller has broken single bit correctable logic for
+ * certain error syndromes. Log them as uncorrectable in that case.
+ */
+static bool xgene_edac_l3_promote_to_uc_err(u32 l3cesr, u32 l3celr)
+{
+ if (l3cesr & L3C_ESR_DATATAG_MASK) {
+ switch (L3C_ELR_ERRSYN(l3celr)) {
+ case 0x13C:
+ case 0x0B4:
+ case 0x007:
+ case 0x00D:
+ case 0x00E:
+ case 0x019:
+ case 0x01A:
+ case 0x01C:
+ case 0x04E:
+ case 0x041:
+ return true;
+ }
+ } else if (L3C_ELR_ERRWAY(l3celr) == 9)
+ return true;
+
+ return false;
+}
+
+static void xgene_edac_l3_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ u32 l3cesr;
+ u32 l3celr;
+ u32 l3caelr;
+ u32 l3cbelr;
+
+ l3cesr = readl(ctx->dev_csr + L3C_ESR);
+ if (!(l3cesr & (L3C_ESR_UCERR_MASK | L3C_ESR_CERR_MASK)))
+ return;
+
+ if (l3cesr & L3C_ESR_UCERR_MASK)
+ dev_err(edac_dev->dev, "L3C uncorrectable error\n");
+ if (l3cesr & L3C_ESR_CERR_MASK)
+ dev_warn(edac_dev->dev, "L3C correctable error\n");
+
+ l3celr = readl(ctx->dev_csr + L3C_ELR);
+ l3caelr = readl(ctx->dev_csr + L3C_AELR);
+ l3cbelr = readl(ctx->dev_csr + L3C_BELR);
+ if (l3cesr & L3C_ESR_MULTIHIT_MASK)
+ dev_err(edac_dev->dev, "L3C multiple hit error\n");
+ if (l3cesr & L3C_ESR_UCEVICT_MASK)
+ dev_err(edac_dev->dev,
+ "L3C dropped eviction of line with error\n");
+ if (l3cesr & L3C_ESR_MULTIUCERR_MASK)
+ dev_err(edac_dev->dev, "L3C multiple uncorrectable error\n");
+ if (l3cesr & L3C_ESR_DATATAG_MASK)
+ dev_err(edac_dev->dev,
+ "L3C data error syndrome 0x%X group 0x%X\n",
+ L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRGRP(l3celr));
+ else
+ dev_err(edac_dev->dev,
+ "L3C tag error syndrome 0x%X Way of Tag 0x%X Agent ID 0x%X Operation type 0x%X\n",
+ L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRWAY(l3celr),
+ L3C_ELR_AGENTID(l3celr), L3C_ELR_OPTYPE(l3celr));
+ /*
+ * NOTE: Address [41:38] in L3C_ELR_PADDRHIGH(l3celr).
+ * Address [37:6] in l3caelr. Lower 6 bits are zero.
+ */
+ dev_err(edac_dev->dev, "L3C error address 0x%08X.%08X bank %d\n",
+ L3C_ELR_PADDRHIGH(l3celr) << 6 | (l3caelr >> 26),
+ (l3caelr & 0x3FFFFFFF) << 6, L3C_BELR_BANK(l3cbelr));
+ dev_err(edac_dev->dev,
+ "L3C error status register value 0x%X\n", l3cesr);
+
+ /* Clear L3C error interrupt */
+ writel(0, ctx->dev_csr + L3C_ESR);
+
+ if (ctx->version <= 1 &&
+ xgene_edac_l3_promote_to_uc_err(l3cesr, l3celr)) {
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+ return;
+ }
+ if (l3cesr & L3C_ESR_CERR_MASK)
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+ if (l3cesr & L3C_ESR_UCERR_MASK)
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static void xgene_edac_l3_hw_init(struct edac_device_ctl_info *edac_dev,
+ bool enable)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ u32 val;
+
+ val = readl(ctx->dev_csr + L3C_ECR);
+ val |= L3C_UCERREN | L3C_CERREN;
+ /* On disable, we just disable interrupt but keep error enabled */
+ if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+ if (enable)
+ val |= L3C_ECR_UCINTREN | L3C_ECR_CINTREN;
+ else
+ val &= ~(L3C_ECR_UCINTREN | L3C_ECR_CINTREN);
+ }
+ writel(val, ctx->dev_csr + L3C_ECR);
+
+ if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+ /* Enable/disable L3 error top level interrupt */
+ if (enable) {
+ xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+ L3C_UNCORR_ERR_MASK);
+ xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
+ L3C_CORR_ERR_MASK);
+ } else {
+ xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+ L3C_UNCORR_ERR_MASK);
+ xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
+ L3C_CORR_ERR_MASK);
+ }
+ }
+}
+
+static ssize_t xgene_edac_l3_inject_ctrl_write(struct file *file,
+ const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dev = file->private_data;
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+
+ /* Generate all errors */
+ writel(0xFFFFFFFF, ctx->dev_csr + L3C_ESR);
+ return count;
+}
+
+static const struct file_operations xgene_edac_l3_debug_inject_fops = {
+ .open = simple_open,
+ .write = xgene_edac_l3_inject_ctrl_write,
+ .llseek = generic_file_llseek
+};
+
+static void
+xgene_edac_l3_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ struct dentry *dbgfs_dir;
+ char name[10];
+
+ if (!IS_ENABLED(CONFIG_EDAC_DEBUG) || !ctx->edac->dfs)
+ return;
+
+ snprintf(name, sizeof(name), "l3c%d", ctx->edac_idx);
+ dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
+ if (!dbgfs_dir)
+ return;
+
+ debugfs_create_file("l3_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
+ &xgene_edac_l3_debug_inject_fops);
+}
+
+static int xgene_edac_l3_add(struct xgene_edac *edac, struct device_node *np,
+ int version)
+{
+ struct edac_device_ctl_info *edac_dev;
+ struct xgene_edac_dev_ctx *ctx;
+ struct resource res;
+ void __iomem *dev_csr;
+ int edac_idx;
+ int rc = 0;
+
+ if (!devres_open_group(edac->dev, xgene_edac_l3_add, GFP_KERNEL))
+ return -ENOMEM;
+
+ rc = of_address_to_resource(np, 0, &res);
+ if (rc < 0) {
+ dev_err(edac->dev, "no L3 resource address\n");
+ goto err_release_group;
+ }
+ dev_csr = devm_ioremap_resource(edac->dev, &res);
+ if (IS_ERR(dev_csr)) {
+ dev_err(edac->dev,
+ "devm_ioremap_resource failed for L3 resource address\n");
+ rc = PTR_ERR(dev_csr);
+ goto err_release_group;
+ }
+
+ edac_idx = edac_device_alloc_index();
+ edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
+ "l3c", 1, "l3c", 1, 0, NULL, 0,
+ edac_idx);
+ if (!edac_dev) {
+ rc = -ENOMEM;
+ goto err_release_group;
+ }
+
+ ctx = edac_dev->pvt_info;
+ ctx->dev_csr = dev_csr;
+ ctx->name = "xgene_l3_err";
+ ctx->edac_idx = edac_idx;
+ ctx->edac = edac;
+ ctx->edac_dev = edac_dev;
+ ctx->ddev = *edac->dev;
+ ctx->version = version;
+ edac_dev->dev = &ctx->ddev;
+ edac_dev->ctl_name = ctx->name;
+ edac_dev->dev_name = ctx->name;
+ edac_dev->mod_name = EDAC_MOD_STR;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ edac_dev->edac_check = xgene_edac_l3_check;
+
+ xgene_edac_l3_create_debugfs_nodes(edac_dev);
+
+ rc = edac_device_add_device(edac_dev);
+ if (rc > 0) {
+ dev_err(edac->dev, "failed edac_device_add_device()\n");
+ rc = -ENOMEM;
+ goto err_ctl_free;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT)
+ edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+ list_add(&ctx->next, &edac->l3s);
+
+ xgene_edac_l3_hw_init(edac_dev, 1);
+
+ devres_remove_group(edac->dev, xgene_edac_l3_add);
+
+ dev_info(edac->dev, "X-Gene EDAC L3 registered\n");
+ return 0;
+
+err_ctl_free:
+ edac_device_free_ctl_info(edac_dev);
+err_release_group:
+ devres_release_group(edac->dev, xgene_edac_l3_add);
+ return rc;
+}
+
+static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3)
+{
+ struct edac_device_ctl_info *edac_dev = l3->edac_dev;
+
+ xgene_edac_l3_hw_init(edac_dev, 0);
+ edac_device_del_device(l3->edac->dev);
+ edac_device_free_ctl_info(edac_dev);
+ return 0;
+}
+
+/* SoC error device */
+#define IOBAXIS0TRANSERRINTSTS 0x0000
+#define IOBAXIS0_M_ILLEGAL_ACCESS_MASK BIT(1)
+#define IOBAXIS0_ILLEGAL_ACCESS_MASK BIT(0)
+#define IOBAXIS0TRANSERRINTMSK 0x0004
+#define IOBAXIS0TRANSERRREQINFOL 0x0008
+#define IOBAXIS0TRANSERRREQINFOH 0x000c
+#define REQTYPE_RD(src) (((src) & BIT(0)))
+#define ERRADDRH_RD(src) (((src) & 0xffc00000) >> 22)
+#define IOBAXIS1TRANSERRINTSTS 0x0010
+#define IOBAXIS1TRANSERRINTMSK 0x0014
+#define IOBAXIS1TRANSERRREQINFOL 0x0018
+#define IOBAXIS1TRANSERRREQINFOH 0x001c
+#define IOBPATRANSERRINTSTS 0x0020
+#define IOBPA_M_REQIDRAM_CORRUPT_MASK BIT(7)
+#define IOBPA_REQIDRAM_CORRUPT_MASK BIT(6)
+#define IOBPA_M_TRANS_CORRUPT_MASK BIT(5)
+#define IOBPA_TRANS_CORRUPT_MASK BIT(4)
+#define IOBPA_M_WDATA_CORRUPT_MASK BIT(3)
+#define IOBPA_WDATA_CORRUPT_MASK BIT(2)
+#define IOBPA_M_RDATA_CORRUPT_MASK BIT(1)
+#define IOBPA_RDATA_CORRUPT_MASK BIT(0)
+#define IOBBATRANSERRINTSTS 0x0030
+#define M_ILLEGAL_ACCESS_MASK BIT(15)
+#define ILLEGAL_ACCESS_MASK BIT(14)
+#define M_WIDRAM_CORRUPT_MASK BIT(13)
+#define WIDRAM_CORRUPT_MASK BIT(12)
+#define M_RIDRAM_CORRUPT_MASK BIT(11)
+#define RIDRAM_CORRUPT_MASK BIT(10)
+#define M_TRANS_CORRUPT_MASK BIT(9)
+#define TRANS_CORRUPT_MASK BIT(8)
+#define M_WDATA_CORRUPT_MASK BIT(7)
+#define WDATA_CORRUPT_MASK BIT(6)
+#define M_RBM_POISONED_REQ_MASK BIT(5)
+#define RBM_POISONED_REQ_MASK BIT(4)
+#define M_XGIC_POISONED_REQ_MASK BIT(3)
+#define XGIC_POISONED_REQ_MASK BIT(2)
+#define M_WRERR_RESP_MASK BIT(1)
+#define WRERR_RESP_MASK BIT(0)
+#define IOBBATRANSERRREQINFOL 0x0038
+#define IOBBATRANSERRREQINFOH 0x003c
+#define REQTYPE_F2_RD(src) ((src) & BIT(0))
+#define ERRADDRH_F2_RD(src) (((src) & 0xffc00000) >> 22)
+#define IOBBATRANSERRCSWREQID 0x0040
+#define XGICTRANSERRINTSTS 0x0050
+#define M_WR_ACCESS_ERR_MASK BIT(3)
+#define WR_ACCESS_ERR_MASK BIT(2)
+#define M_RD_ACCESS_ERR_MASK BIT(1)
+#define RD_ACCESS_ERR_MASK BIT(0)
+#define XGICTRANSERRINTMSK 0x0054
+#define XGICTRANSERRREQINFO 0x0058
+#define REQTYPE_MASK BIT(26)
+#define ERRADDR_RD(src) ((src) & 0x03ffffff)
+#define GLBL_ERR_STS 0x0800
+#define MDED_ERR_MASK BIT(3)
+#define DED_ERR_MASK BIT(2)
+#define MSEC_ERR_MASK BIT(1)
+#define SEC_ERR_MASK BIT(0)
+#define GLBL_SEC_ERRL 0x0810
+#define GLBL_SEC_ERRH 0x0818
+#define GLBL_MSEC_ERRL 0x0820
+#define GLBL_MSEC_ERRH 0x0828
+#define GLBL_DED_ERRL 0x0830
+#define GLBL_DED_ERRLMASK 0x0834
+#define GLBL_DED_ERRH 0x0838
+#define GLBL_DED_ERRHMASK 0x083c
+#define GLBL_MDED_ERRL 0x0840
+#define GLBL_MDED_ERRLMASK 0x0844
+#define GLBL_MDED_ERRH 0x0848
+#define GLBL_MDED_ERRHMASK 0x084c
+
+/* IO Bus Registers */
+#define RBCSR 0x0000
+#define STICKYERR_MASK BIT(0)
+#define RBEIR 0x0008
+#define AGENT_OFFLINE_ERR_MASK BIT(30)
+#define UNIMPL_RBPAGE_ERR_MASK BIT(29)
+#define WORD_ALIGNED_ERR_MASK BIT(28)
+#define PAGE_ACCESS_ERR_MASK BIT(27)
+#define WRITE_ACCESS_MASK BIT(26)
+
+static const char * const soc_mem_err_v1[] = {
+ "10GbE0",
+ "10GbE1",
+ "Security",
+ "SATA45",
+ "SATA23/ETH23",
+ "SATA01/ETH01",
+ "USB1",
+ "USB0",
+ "QML",
+ "QM0",
+ "QM1 (XGbE01)",
+ "PCIE4",
+ "PCIE3",
+ "PCIE2",
+ "PCIE1",
+ "PCIE0",
+ "CTX Manager",
+ "OCM",
+ "1GbE",
+ "CLE",
+ "AHBC",
+ "PktDMA",
+ "GFC",
+ "MSLIM",
+ "10GbE2",
+ "10GbE3",
+ "QM2 (XGbE23)",
+ "IOB",
+ "unknown",
+ "unknown",
+ "unknown",
+ "unknown",
+};
+
+static void xgene_edac_iob_gic_report(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ u32 err_addr_lo;
+ u32 err_addr_hi;
+ u32 reg;
+ u32 info;
+
+ /* GIC transaction error interrupt */
+ reg = readl(ctx->dev_csr + XGICTRANSERRINTSTS);
+ if (!reg)
+ goto chk_iob_err;
+ dev_err(edac_dev->dev, "XGIC transaction error\n");
+ if (reg & RD_ACCESS_ERR_MASK)
+ dev_err(edac_dev->dev, "XGIC read size error\n");
+ if (reg & M_RD_ACCESS_ERR_MASK)
+ dev_err(edac_dev->dev, "Multiple XGIC read size error\n");
+ if (reg & WR_ACCESS_ERR_MASK)
+ dev_err(edac_dev->dev, "XGIC write size error\n");
+ if (reg & M_WR_ACCESS_ERR_MASK)
+ dev_err(edac_dev->dev, "Multiple XGIC write size error\n");
+ info = readl(ctx->dev_csr + XGICTRANSERRREQINFO);
+ dev_err(edac_dev->dev, "XGIC %s access @ 0x%08X (0x%08X)\n",
+ info & REQTYPE_MASK ? "read" : "write", ERRADDR_RD(info),
+ info);
+ writel(reg, ctx->dev_csr + XGICTRANSERRINTSTS);
+
+chk_iob_err:
+ /* IOB memory error */
+ reg = readl(ctx->dev_csr + GLBL_ERR_STS);
+ if (!reg)
+ return;
+ if (reg & SEC_ERR_MASK) {
+ err_addr_lo = readl(ctx->dev_csr + GLBL_SEC_ERRL);
+ err_addr_hi = readl(ctx->dev_csr + GLBL_SEC_ERRH);
+ dev_err(edac_dev->dev,
+ "IOB single-bit correctable memory at 0x%08X.%08X error\n",
+ err_addr_lo, err_addr_hi);
+ writel(err_addr_lo, ctx->dev_csr + GLBL_SEC_ERRL);
+ writel(err_addr_hi, ctx->dev_csr + GLBL_SEC_ERRH);
+ }
+ if (reg & MSEC_ERR_MASK) {
+ err_addr_lo = readl(ctx->dev_csr + GLBL_MSEC_ERRL);
+ err_addr_hi = readl(ctx->dev_csr + GLBL_MSEC_ERRH);
+ dev_err(edac_dev->dev,
+ "IOB multiple single-bit correctable memory at 0x%08X.%08X error\n",
+ err_addr_lo, err_addr_hi);
+ writel(err_addr_lo, ctx->dev_csr + GLBL_MSEC_ERRL);
+ writel(err_addr_hi, ctx->dev_csr + GLBL_MSEC_ERRH);
+ }
+ if (reg & (SEC_ERR_MASK | MSEC_ERR_MASK))
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+
+ if (reg & DED_ERR_MASK) {
+ err_addr_lo = readl(ctx->dev_csr + GLBL_DED_ERRL);
+ err_addr_hi = readl(ctx->dev_csr + GLBL_DED_ERRH);
+ dev_err(edac_dev->dev,
+ "IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
+ err_addr_lo, err_addr_hi);
+ writel(err_addr_lo, ctx->dev_csr + GLBL_DED_ERRL);
+ writel(err_addr_hi, ctx->dev_csr + GLBL_DED_ERRH);
+ }
+ if (reg & MDED_ERR_MASK) {
+ err_addr_lo = readl(ctx->dev_csr + GLBL_MDED_ERRL);
+ err_addr_hi = readl(ctx->dev_csr + GLBL_MDED_ERRH);
+ dev_err(edac_dev->dev,
+ "Multiple IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
+ err_addr_lo, err_addr_hi);
+ writel(err_addr_lo, ctx->dev_csr + GLBL_MDED_ERRL);
+ writel(err_addr_hi, ctx->dev_csr + GLBL_MDED_ERRH);
+ }
+ if (reg & (DED_ERR_MASK | MDED_ERR_MASK))
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static void xgene_edac_rb_report(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ u32 err_addr_lo;
+ u32 err_addr_hi;
+ u32 reg;
+
+ /* If the register bus resource isn't available, just skip it */
+ if (!ctx->edac->rb_map)
+ goto rb_skip;
+
+ /*
+ * Check RB access errors
+ * 1. Out of range
+ * 2. Un-implemented page
+ * 3. Un-aligned access
+ * 4. Offline slave IP
+ */
+ if (regmap_read(ctx->edac->rb_map, RBCSR, &reg))
+ return;
+ if (reg & STICKYERR_MASK) {
+ bool write;
+
+ dev_err(edac_dev->dev, "IOB bus access error(s)\n");
+ if (regmap_read(ctx->edac->rb_map, RBEIR, &reg))
+ return;
+ write = reg & WRITE_ACCESS_MASK ? 1 : 0;
+ if (reg & AGENT_OFFLINE_ERR_MASK)
+ dev_err(edac_dev->dev,
+ "IOB bus %s access to offline agent error\n",
+ write ? "write" : "read");
+ if (reg & UNIMPL_RBPAGE_ERR_MASK)
+ dev_err(edac_dev->dev,
+ "IOB bus %s access to unimplemented page error\n",
+ write ? "write" : "read");
+ if (reg & WORD_ALIGNED_ERR_MASK)
+ dev_err(edac_dev->dev,
+ "IOB bus %s word aligned access error\n",
+ write ? "write" : "read");
+ if (reg & PAGE_ACCESS_ERR_MASK)
+ dev_err(edac_dev->dev,
+ "IOB bus %s to page out of range access error\n",
+ write ? "write" : "read");
+ if (regmap_write(ctx->edac->rb_map, RBEIR, 0))
+ return;
+ if (regmap_write(ctx->edac->rb_map, RBCSR, 0))
+ return;
+ }
+rb_skip:
+
+ /* IOB Bridge agent transaction error interrupt */
+ reg = readl(ctx->dev_csr + IOBBATRANSERRINTSTS);
+ if (!reg)
+ return;
+
+ dev_err(edac_dev->dev, "IOB bridge agent (BA) transaction error\n");
+ if (reg & WRERR_RESP_MASK)
+ dev_err(edac_dev->dev, "IOB BA write response error\n");
+ if (reg & M_WRERR_RESP_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA write response error\n");
+ if (reg & XGIC_POISONED_REQ_MASK)
+ dev_err(edac_dev->dev, "IOB BA XGIC poisoned write error\n");
+ if (reg & M_XGIC_POISONED_REQ_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA XGIC poisoned write error\n");
+ if (reg & RBM_POISONED_REQ_MASK)
+ dev_err(edac_dev->dev, "IOB BA RBM poisoned write error\n");
+ if (reg & M_RBM_POISONED_REQ_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA RBM poisoned write error\n");
+ if (reg & WDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB BA write error\n");
+ if (reg & M_WDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "Multiple IOB BA write error\n");
+ if (reg & TRANS_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB BA transaction error\n");
+ if (reg & M_TRANS_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "Multiple IOB BA transaction error\n");
+ if (reg & RIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "IOB BA RDIDRAM read transaction ID error\n");
+ if (reg & M_RIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA RDIDRAM read transaction ID error\n");
+ if (reg & WIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "IOB BA RDIDRAM write transaction ID error\n");
+ if (reg & M_WIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA RDIDRAM write transaction ID error\n");
+ if (reg & ILLEGAL_ACCESS_MASK)
+ dev_err(edac_dev->dev,
+ "IOB BA XGIC/RB illegal access error\n");
+ if (reg & M_ILLEGAL_ACCESS_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB BA XGIC/RB illegal access error\n");
+
+ err_addr_lo = readl(ctx->dev_csr + IOBBATRANSERRREQINFOL);
+ err_addr_hi = readl(ctx->dev_csr + IOBBATRANSERRREQINFOH);
+ dev_err(edac_dev->dev, "IOB BA %s access at 0x%02X.%08X (0x%08X)\n",
+ REQTYPE_F2_RD(err_addr_hi) ? "read" : "write",
+ ERRADDRH_F2_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+ if (reg & WRERR_RESP_MASK)
+ dev_err(edac_dev->dev, "IOB BA requestor ID 0x%08X\n",
+ readl(ctx->dev_csr + IOBBATRANSERRCSWREQID));
+ writel(reg, ctx->dev_csr + IOBBATRANSERRINTSTS);
+}
+
+static void xgene_edac_pa_report(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ u32 err_addr_lo;
+ u32 err_addr_hi;
+ u32 reg;
+
+ /* IOB Processing agent transaction error interrupt */
+ reg = readl(ctx->dev_csr + IOBPATRANSERRINTSTS);
+ if (!reg)
+ goto chk_iob_axi0;
+ dev_err(edac_dev->dev, "IOB processing agent (PA) transaction error\n");
+ if (reg & IOBPA_RDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB PA read data RAM error\n");
+ if (reg & IOBPA_M_RDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB PA read data RAM error\n");
+ if (reg & IOBPA_WDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB PA write data RAM error\n");
+ if (reg & IOBPA_M_WDATA_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB PA write data RAM error\n");
+ if (reg & IOBPA_TRANS_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB PA transaction error\n");
+ if (reg & IOBPA_M_TRANS_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "Multiple IOB PA transaction error\n");
+ if (reg & IOBPA_REQIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev, "IOB PA transaction ID RAM error\n");
+ if (reg & IOBPA_M_REQIDRAM_CORRUPT_MASK)
+ dev_err(edac_dev->dev,
+ "Multiple IOB PA transaction ID RAM error\n");
+ writel(reg, ctx->dev_csr + IOBPATRANSERRINTSTS);
+
+chk_iob_axi0:
+ /* IOB AXI0 Error */
+ reg = readl(ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
+ if (!reg)
+ goto chk_iob_axi1;
+ err_addr_lo = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOL);
+ err_addr_hi = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOH);
+ dev_err(edac_dev->dev,
+ "%sAXI slave 0 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
+ reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
+ REQTYPE_RD(err_addr_hi) ? "read" : "write",
+ ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+ writel(reg, ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
+
+chk_iob_axi1:
+ /* IOB AXI1 Error */
+ reg = readl(ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
+ if (!reg)
+ return;
+ err_addr_lo = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOL);
+ err_addr_hi = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOH);
+ dev_err(edac_dev->dev,
+ "%sAXI slave 1 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
+ reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
+ REQTYPE_RD(err_addr_hi) ? "read" : "write",
+ ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+ writel(reg, ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
+}
+
+static void xgene_edac_soc_check(struct edac_device_ctl_info *edac_dev)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+ const char * const *soc_mem_err = NULL;
+ u32 pcp_hp_stat;
+ u32 pcp_lp_stat;
+ u32 reg;
+ int i;
+
+ xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
+ xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat);
+ xgene_edac_pcp_rd(ctx->edac, MEMERRINTSTS, &reg);
+ if (!((pcp_hp_stat & (IOB_PA_ERR_MASK | IOB_BA_ERR_MASK |
+ IOB_XGIC_ERR_MASK | IOB_RB_ERR_MASK)) ||
+ (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) || reg))
+ return;
+
+ if (pcp_hp_stat & IOB_XGIC_ERR_MASK)
+ xgene_edac_iob_gic_report(edac_dev);
+
+ if (pcp_hp_stat & (IOB_RB_ERR_MASK | IOB_BA_ERR_MASK))
+ xgene_edac_rb_report(edac_dev);
+
+ if (pcp_hp_stat & IOB_PA_ERR_MASK)
+ xgene_edac_pa_report(edac_dev);
+
+ if (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) {
+ dev_info(edac_dev->dev,
+ "CSW switch trace correctable memory parity error\n");
+ edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+ }
+
+ if (!reg)
+ return;
+ if (ctx->version == 1)
+ soc_mem_err = soc_mem_err_v1;
+ if (!soc_mem_err) {
+ dev_err(edac_dev->dev, "SoC memory parity error 0x%08X\n",
+ reg);
+ edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+ return;
+ }
+ for (i = 0; i < 31; i++) {
+ if (reg & (1 << i)) {
+ dev_err(edac_dev->dev, "%s memory parity error\n",
+ soc_mem_err[i]);
+ edac_device_handle_ue(edac_dev, 0, 0,
+ edac_dev->ctl_name);
+ }
+ }
+}
+
+static void xgene_edac_soc_hw_init(struct edac_device_ctl_info *edac_dev,
+ bool enable)
+{
+ struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+
+ /* Enable SoC IP error interrupt */
+ if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+ if (enable) {
+ xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+ IOB_PA_ERR_MASK |
+ IOB_BA_ERR_MASK |
+ IOB_XGIC_ERR_MASK |
+ IOB_RB_ERR_MASK);
+ xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
+ CSW_SWITCH_TRACE_ERR_MASK);
+ } else {
+ xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+ IOB_PA_ERR_MASK |
+ IOB_BA_ERR_MASK |
+ IOB_XGIC_ERR_MASK |
+ IOB_RB_ERR_MASK);
+ xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
+ CSW_SWITCH_TRACE_ERR_MASK);
+ }
+
+ writel(enable ? 0x0 : 0xFFFFFFFF,
+ ctx->dev_csr + IOBAXIS0TRANSERRINTMSK);
+ writel(enable ? 0x0 : 0xFFFFFFFF,
+ ctx->dev_csr + IOBAXIS1TRANSERRINTMSK);
+ writel(enable ? 0x0 : 0xFFFFFFFF,
+ ctx->dev_csr + XGICTRANSERRINTMSK);
+
+ xgene_edac_pcp_setbits(ctx->edac, MEMERRINTMSK,
+ enable ? 0x0 : 0xFFFFFFFF);
+ }
+}
+
+static int xgene_edac_soc_add(struct xgene_edac *edac, struct device_node *np,
+ int version)
+{
+ struct edac_device_ctl_info *edac_dev;
+ struct xgene_edac_dev_ctx *ctx;
+ void __iomem *dev_csr;
+ struct resource res;
+ int edac_idx;
+ int rc;
+
+ if (!devres_open_group(edac->dev, xgene_edac_soc_add, GFP_KERNEL))
+ return -ENOMEM;
+
+ rc = of_address_to_resource(np, 0, &res);
+ if (rc < 0) {
+ dev_err(edac->dev, "no SoC resource address\n");
+ goto err_release_group;
+ }
+ dev_csr = devm_ioremap_resource(edac->dev, &res);
+ if (IS_ERR(dev_csr)) {
+ dev_err(edac->dev,
+ "devm_ioremap_resource failed for soc resource address\n");
+ rc = PTR_ERR(dev_csr);
+ goto err_release_group;
+ }
+
+ edac_idx = edac_device_alloc_index();
+ edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
+ "SOC", 1, "SOC", 1, 2, NULL, 0,
+ edac_idx);
+ if (!edac_dev) {
+ rc = -ENOMEM;
+ goto err_release_group;
+ }
+
+ ctx = edac_dev->pvt_info;
+ ctx->dev_csr = dev_csr;
+ ctx->name = "xgene_soc_err";
+ ctx->edac_idx = edac_idx;
+ ctx->edac = edac;
+ ctx->edac_dev = edac_dev;
+ ctx->ddev = *edac->dev;
+ ctx->version = version;
+ edac_dev->dev = &ctx->ddev;
+ edac_dev->ctl_name = ctx->name;
+ edac_dev->dev_name = ctx->name;
+ edac_dev->mod_name = EDAC_MOD_STR;
+
+ if (edac_op_state == EDAC_OPSTATE_POLL)
+ edac_dev->edac_check = xgene_edac_soc_check;
+
+ rc = edac_device_add_device(edac_dev);
+ if (rc > 0) {
+ dev_err(edac->dev, "failed edac_device_add_device()\n");
+ rc = -ENOMEM;
+ goto err_ctl_free;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT)
+ edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+ list_add(&ctx->next, &edac->socs);
+
+ xgene_edac_soc_hw_init(edac_dev, 1);
+
+ devres_remove_group(edac->dev, xgene_edac_soc_add);
+
+ dev_info(edac->dev, "X-Gene EDAC SoC registered\n");
+
+ return 0;
+
+err_ctl_free:
+ edac_device_free_ctl_info(edac_dev);
+err_release_group:
+ devres_release_group(edac->dev, xgene_edac_soc_add);
+ return rc;
+}
+
+static int xgene_edac_soc_remove(struct xgene_edac_dev_ctx *soc)
+{
+ struct edac_device_ctl_info *edac_dev = soc->edac_dev;
+
+ xgene_edac_soc_hw_init(edac_dev, 0);
+ edac_device_del_device(soc->edac->dev);
+ edac_device_free_ctl_info(edac_dev);
+ return 0;
+}
+
+static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
+{
+ struct xgene_edac *ctx = dev_id;
+ struct xgene_edac_pmd_ctx *pmd;
+ struct xgene_edac_dev_ctx *node;
+ unsigned int pcp_hp_stat;
+ unsigned int pcp_lp_stat;
+
+ xgene_edac_pcp_rd(ctx, PCPHPERRINTSTS, &pcp_hp_stat);
+ xgene_edac_pcp_rd(ctx, PCPLPERRINTSTS, &pcp_lp_stat);
+ if ((MCU_UNCORR_ERR_MASK & pcp_hp_stat) ||
+ (MCU_CTL_ERR_MASK & pcp_hp_stat) ||
+ (MCU_CORR_ERR_MASK & pcp_lp_stat)) {
+ struct xgene_edac_mc_ctx *mcu;
+
+ list_for_each_entry(mcu, &ctx->mcus, next)
+ xgene_edac_mc_check(mcu->mci);
+ }
+
+ list_for_each_entry(pmd, &ctx->pmds, next) {
+ if ((PMD0_MERR_MASK << pmd->pmd) & pcp_hp_stat)
+ xgene_edac_pmd_check(pmd->edac_dev);
+ }
+
+ list_for_each_entry(node, &ctx->l3s, next)
+ xgene_edac_l3_check(node->edac_dev);
+
+ list_for_each_entry(node, &ctx->socs, next)
+ xgene_edac_soc_check(node->edac_dev);
+
+ return IRQ_HANDLED;
+}
+
+static int xgene_edac_probe(struct platform_device *pdev)
+{
+ struct xgene_edac *edac;
+ struct device_node *child;
+ struct resource *res;
+ int rc;
+
+ edac = devm_kzalloc(&pdev->dev, sizeof(*edac), GFP_KERNEL);
+ if (!edac)
+ return -ENOMEM;
+
+ edac->dev = &pdev->dev;
+ platform_set_drvdata(pdev, edac);
+ INIT_LIST_HEAD(&edac->mcus);
+ INIT_LIST_HEAD(&edac->pmds);
+ INIT_LIST_HEAD(&edac->l3s);
+ INIT_LIST_HEAD(&edac->socs);
+ spin_lock_init(&edac->lock);
+ mutex_init(&edac->mc_lock);
+
+ edac->csw_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "regmap-csw");
+ if (IS_ERR(edac->csw_map)) {
+ dev_err(edac->dev, "unable to get syscon regmap csw\n");
+ rc = PTR_ERR(edac->csw_map);
+ goto out_err;
+ }
+
+ edac->mcba_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "regmap-mcba");
+ if (IS_ERR(edac->mcba_map)) {
+ dev_err(edac->dev, "unable to get syscon regmap mcba\n");
+ rc = PTR_ERR(edac->mcba_map);
+ goto out_err;
+ }
+
+ edac->mcbb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "regmap-mcbb");
+ if (IS_ERR(edac->mcbb_map)) {
+ dev_err(edac->dev, "unable to get syscon regmap mcbb\n");
+ rc = PTR_ERR(edac->mcbb_map);
+ goto out_err;
+ }
+ edac->efuse_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "regmap-efuse");
+ if (IS_ERR(edac->efuse_map)) {
+ dev_err(edac->dev, "unable to get syscon regmap efuse\n");
+ rc = PTR_ERR(edac->efuse_map);
+ goto out_err;
+ }
+
+ /*
+ * NOTE: The register bus resource is optional for compatibility
+ * reason.
+ */
+ edac->rb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+ "regmap-rb");
+ if (IS_ERR(edac->rb_map)) {
+ dev_warn(edac->dev, "missing syscon regmap rb\n");
+ edac->rb_map = NULL;
+ }
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res);
+ if (IS_ERR(edac->pcp_csr)) {
+ dev_err(&pdev->dev, "no PCP resource address\n");
+ rc = PTR_ERR(edac->pcp_csr);
+ goto out_err;
+ }
+
+ if (edac_op_state == EDAC_OPSTATE_INT) {
+ int irq;
+ int i;
+
+ for (i = 0; i < 3; i++) {
+ irq = platform_get_irq_optional(pdev, i);
+ if (irq < 0) {
+ dev_err(&pdev->dev, "No IRQ resource\n");
+ rc = irq;
+ goto out_err;
+ }
+ rc = devm_request_irq(&pdev->dev, irq,
+ xgene_edac_isr, IRQF_SHARED,
+ dev_name(&pdev->dev), edac);
+ if (rc) {
+ dev_err(&pdev->dev,
+ "Could not request IRQ %d\n", irq);
+ goto out_err;
+ }
+ }
+ }
+
+ edac->dfs = edac_debugfs_create_dir(pdev->dev.kobj.name);
+
+ for_each_child_of_node(pdev->dev.of_node, child) {
+ if (!of_device_is_available(child))
+ continue;
+ if (of_device_is_compatible(child, "apm,xgene-edac-mc"))
+ xgene_edac_mc_add(edac, child);
+ if (of_device_is_compatible(child, "apm,xgene-edac-pmd"))
+ xgene_edac_pmd_add(edac, child, 1);
+ if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2"))
+ xgene_edac_pmd_add(edac, child, 2);
+ if (of_device_is_compatible(child, "apm,xgene-edac-l3"))
+ xgene_edac_l3_add(edac, child, 1);
+ if (of_device_is_compatible(child, "apm,xgene-edac-l3-v2"))
+ xgene_edac_l3_add(edac, child, 2);
+ if (of_device_is_compatible(child, "apm,xgene-edac-soc"))
+ xgene_edac_soc_add(edac, child, 0);
+ if (of_device_is_compatible(child, "apm,xgene-edac-soc-v1"))
+ xgene_edac_soc_add(edac, child, 1);
+ }
+
+ return 0;
+
+out_err:
+ return rc;
+}
+
+static int xgene_edac_remove(struct platform_device *pdev)
+{
+ struct xgene_edac *edac = dev_get_drvdata(&pdev->dev);
+ struct xgene_edac_mc_ctx *mcu;
+ struct xgene_edac_mc_ctx *temp_mcu;
+ struct xgene_edac_pmd_ctx *pmd;
+ struct xgene_edac_pmd_ctx *temp_pmd;
+ struct xgene_edac_dev_ctx *node;
+ struct xgene_edac_dev_ctx *temp_node;
+
+ list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next)
+ xgene_edac_mc_remove(mcu);
+
+ list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next)
+ xgene_edac_pmd_remove(pmd);
+
+ list_for_each_entry_safe(node, temp_node, &edac->l3s, next)
+ xgene_edac_l3_remove(node);
+
+ list_for_each_entry_safe(node, temp_node, &edac->socs, next)
+ xgene_edac_soc_remove(node);
+
+ return 0;
+}
+
+static const struct of_device_id xgene_edac_of_match[] = {
+ { .compatible = "apm,xgene-edac" },
+ {},
+};
+MODULE_DEVICE_TABLE(of, xgene_edac_of_match);
+
+static struct platform_driver xgene_edac_driver = {
+ .probe = xgene_edac_probe,
+ .remove = xgene_edac_remove,
+ .driver = {
+ .name = "xgene-edac",
+ .of_match_table = xgene_edac_of_match,
+ },
+};
+
+static int __init xgene_edac_init(void)
+{
+ int rc;
+
+ if (ghes_get_devices())
+ return -EBUSY;
+
+ /* Make sure error reporting method is sane */
+ switch (edac_op_state) {
+ case EDAC_OPSTATE_POLL:
+ case EDAC_OPSTATE_INT:
+ break;
+ default:
+ edac_op_state = EDAC_OPSTATE_INT;
+ break;
+ }
+
+ rc = platform_driver_register(&xgene_edac_driver);
+ if (rc) {
+ edac_printk(KERN_ERR, EDAC_MOD_STR,
+ "EDAC fails to register\n");
+ goto reg_failed;
+ }
+
+ return 0;
+
+reg_failed:
+ return rc;
+}
+module_init(xgene_edac_init);
+
+static void __exit xgene_edac_exit(void)
+{
+ platform_driver_unregister(&xgene_edac_driver);
+}
+module_exit(xgene_edac_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Feng Kan <fkan@apm.com>");
+MODULE_DESCRIPTION("APM X-Gene EDAC driver");
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state,
+ "EDAC error reporting state: 0=Poll, 2=Interrupt");
diff --git a/drivers/edac/zynqmp_edac.c b/drivers/edac/zynqmp_edac.c
new file mode 100644
index 0000000000..ac7d1e0b32
--- /dev/null
+++ b/drivers/edac/zynqmp_edac.c
@@ -0,0 +1,467 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Xilinx ZynqMP OCM ECC Driver
+ *
+ * Copyright (C) 2022 Advanced Micro Devices, Inc.
+ */
+
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
+
+#include "edac_module.h"
+
+#define ZYNQMP_OCM_EDAC_MSG_SIZE 256
+
+#define ZYNQMP_OCM_EDAC_STRING "zynqmp_ocm"
+
+/* Error/Interrupt registers */
+#define ERR_CTRL_OFST 0x0
+#define OCM_ISR_OFST 0x04
+#define OCM_IMR_OFST 0x08
+#define OCM_IEN_OFST 0x0C
+#define OCM_IDS_OFST 0x10
+
+/* ECC control register */
+#define ECC_CTRL_OFST 0x14
+
+/* Correctable error info registers */
+#define CE_FFA_OFST 0x1C
+#define CE_FFD0_OFST 0x20
+#define CE_FFD1_OFST 0x24
+#define CE_FFD2_OFST 0x28
+#define CE_FFD3_OFST 0x2C
+#define CE_FFE_OFST 0x30
+
+/* Uncorrectable error info registers */
+#define UE_FFA_OFST 0x34
+#define UE_FFD0_OFST 0x38
+#define UE_FFD1_OFST 0x3C
+#define UE_FFD2_OFST 0x40
+#define UE_FFD3_OFST 0x44
+#define UE_FFE_OFST 0x48
+
+/* ECC control register bit field definitions */
+#define ECC_CTRL_CLR_CE_ERR 0x40
+#define ECC_CTRL_CLR_UE_ERR 0x80
+
+/* Fault injection data and count registers */
+#define OCM_FID0_OFST 0x4C
+#define OCM_FID1_OFST 0x50
+#define OCM_FID2_OFST 0x54
+#define OCM_FID3_OFST 0x58
+#define OCM_FIC_OFST 0x74
+
+#define UE_MAX_BITPOS_LOWER 31
+#define UE_MIN_BITPOS_UPPER 32
+#define UE_MAX_BITPOS_UPPER 63
+
+/* Interrupt masks */
+#define OCM_CEINTR_MASK BIT(6)
+#define OCM_UEINTR_MASK BIT(7)
+#define OCM_ECC_ENABLE_MASK BIT(0)
+
+#define OCM_FICOUNT_MASK GENMASK(23, 0)
+#define OCM_NUM_UE_BITPOS 2
+#define OCM_BASEVAL 0xFFFC0000
+#define EDAC_DEVICE "ZynqMP-OCM"
+
+/**
+ * struct ecc_error_info - ECC error log information
+ * @addr: Fault generated at this address
+ * @fault_lo: Generated fault data (lower 32-bit)
+ * @fault_hi: Generated fault data (upper 32-bit)
+ */
+struct ecc_error_info {
+ u32 addr;
+ u32 fault_lo;
+ u32 fault_hi;
+};
+
+/**
+ * struct ecc_status - ECC status information to report
+ * @ce_cnt: Correctable error count
+ * @ue_cnt: Uncorrectable error count
+ * @ceinfo: Correctable error log information
+ * @ueinfo: Uncorrectable error log information
+ */
+struct ecc_status {
+ u32 ce_cnt;
+ u32 ue_cnt;
+ struct ecc_error_info ceinfo;
+ struct ecc_error_info ueinfo;
+};
+
+/**
+ * struct edac_priv - OCM private instance data
+ * @baseaddr: Base address of the OCM
+ * @message: Buffer for framing the event specific info
+ * @stat: ECC status information
+ * @ce_cnt: Correctable Error count
+ * @ue_cnt: Uncorrectable Error count
+ * @debugfs_dir: Directory entry for debugfs
+ * @ce_bitpos: Bit position for Correctable Error
+ * @ue_bitpos: Array to store UnCorrectable Error bit positions
+ * @fault_injection_cnt: Fault Injection Counter value
+ */
+struct edac_priv {
+ void __iomem *baseaddr;
+ char message[ZYNQMP_OCM_EDAC_MSG_SIZE];
+ struct ecc_status stat;
+ u32 ce_cnt;
+ u32 ue_cnt;
+#ifdef CONFIG_EDAC_DEBUG
+ struct dentry *debugfs_dir;
+ u8 ce_bitpos;
+ u8 ue_bitpos[OCM_NUM_UE_BITPOS];
+ u32 fault_injection_cnt;
+#endif
+};
+
+/**
+ * get_error_info - Get the current ECC error info
+ * @base: Pointer to the base address of the OCM
+ * @p: Pointer to the OCM ECC status structure
+ * @mask: Status register mask value
+ *
+ * Determines there is any ECC error or not
+ *
+ */
+static void get_error_info(void __iomem *base, struct ecc_status *p, int mask)
+{
+ if (mask & OCM_CEINTR_MASK) {
+ p->ce_cnt++;
+ p->ceinfo.fault_lo = readl(base + CE_FFD0_OFST);
+ p->ceinfo.fault_hi = readl(base + CE_FFD1_OFST);
+ p->ceinfo.addr = (OCM_BASEVAL | readl(base + CE_FFA_OFST));
+ writel(ECC_CTRL_CLR_CE_ERR, base + OCM_ISR_OFST);
+ } else if (mask & OCM_UEINTR_MASK) {
+ p->ue_cnt++;
+ p->ueinfo.fault_lo = readl(base + UE_FFD0_OFST);
+ p->ueinfo.fault_hi = readl(base + UE_FFD1_OFST);
+ p->ueinfo.addr = (OCM_BASEVAL | readl(base + UE_FFA_OFST));
+ writel(ECC_CTRL_CLR_UE_ERR, base + OCM_ISR_OFST);
+ }
+}
+
+/**
+ * handle_error - Handle error types CE and UE
+ * @dci: Pointer to the EDAC device instance
+ * @p: Pointer to the OCM ECC status structure
+ *
+ * Handles correctable and uncorrectable errors.
+ */
+static void handle_error(struct edac_device_ctl_info *dci, struct ecc_status *p)
+{
+ struct edac_priv *priv = dci->pvt_info;
+ struct ecc_error_info *pinf;
+
+ if (p->ce_cnt) {
+ pinf = &p->ceinfo;
+ snprintf(priv->message, ZYNQMP_OCM_EDAC_MSG_SIZE,
+ "\nOCM ECC error type :%s\nAddr: [0x%x]\nFault Data[0x%08x%08x]",
+ "CE", pinf->addr, pinf->fault_hi, pinf->fault_lo);
+ edac_device_handle_ce(dci, 0, 0, priv->message);
+ }
+
+ if (p->ue_cnt) {
+ pinf = &p->ueinfo;
+ snprintf(priv->message, ZYNQMP_OCM_EDAC_MSG_SIZE,
+ "\nOCM ECC error type :%s\nAddr: [0x%x]\nFault Data[0x%08x%08x]",
+ "UE", pinf->addr, pinf->fault_hi, pinf->fault_lo);
+ edac_device_handle_ue(dci, 0, 0, priv->message);
+ }
+
+ memset(p, 0, sizeof(*p));
+}
+
+/**
+ * intr_handler - ISR routine
+ * @irq: irq number
+ * @dev_id: device id pointer
+ *
+ * Return: IRQ_NONE, if CE/UE interrupt not set or IRQ_HANDLED otherwise
+ */
+static irqreturn_t intr_handler(int irq, void *dev_id)
+{
+ struct edac_device_ctl_info *dci = dev_id;
+ struct edac_priv *priv = dci->pvt_info;
+ int regval;
+
+ regval = readl(priv->baseaddr + OCM_ISR_OFST);
+ if (!(regval & (OCM_CEINTR_MASK | OCM_UEINTR_MASK))) {
+ WARN_ONCE(1, "Unhandled IRQ%d, ISR: 0x%x", irq, regval);
+ return IRQ_NONE;
+ }
+
+ get_error_info(priv->baseaddr, &priv->stat, regval);
+
+ priv->ce_cnt += priv->stat.ce_cnt;
+ priv->ue_cnt += priv->stat.ue_cnt;
+ handle_error(dci, &priv->stat);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * get_eccstate - Return the ECC status
+ * @base: Pointer to the OCM base address
+ *
+ * Get the ECC enable/disable status
+ *
+ * Return: ECC status 0/1.
+ */
+static bool get_eccstate(void __iomem *base)
+{
+ return readl(base + ECC_CTRL_OFST) & OCM_ECC_ENABLE_MASK;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/**
+ * write_fault_count - write fault injection count
+ * @priv: Pointer to the EDAC private struct
+ *
+ * Update the fault injection count register, once the counter reaches
+ * zero, it injects errors
+ */
+static void write_fault_count(struct edac_priv *priv)
+{
+ u32 ficount = priv->fault_injection_cnt;
+
+ if (ficount & ~OCM_FICOUNT_MASK) {
+ ficount &= OCM_FICOUNT_MASK;
+ edac_printk(KERN_INFO, EDAC_DEVICE,
+ "Fault injection count value truncated to %d\n", ficount);
+ }
+
+ writel(ficount, priv->baseaddr + OCM_FIC_OFST);
+}
+
+/*
+ * To get the Correctable Error injected, the following steps are needed:
+ * - Setup the optional Fault Injection Count:
+ * echo <fault_count val> > /sys/kernel/debug/edac/ocm/inject_fault_count
+ * - Write the Correctable Error bit position value:
+ * echo <bit_pos val> > /sys/kernel/debug/edac/ocm/inject_ce_bitpos
+ */
+static ssize_t inject_ce_write(struct file *file, const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dev = file->private_data;
+ struct edac_priv *priv = edac_dev->pvt_info;
+ int ret;
+
+ if (!data)
+ return -EFAULT;
+
+ ret = kstrtou8_from_user(data, count, 0, &priv->ce_bitpos);
+ if (ret)
+ return ret;
+
+ if (priv->ce_bitpos > UE_MAX_BITPOS_UPPER)
+ return -EINVAL;
+
+ if (priv->ce_bitpos <= UE_MAX_BITPOS_LOWER) {
+ writel(BIT(priv->ce_bitpos), priv->baseaddr + OCM_FID0_OFST);
+ writel(0, priv->baseaddr + OCM_FID1_OFST);
+ } else {
+ writel(BIT(priv->ce_bitpos - UE_MIN_BITPOS_UPPER),
+ priv->baseaddr + OCM_FID1_OFST);
+ writel(0, priv->baseaddr + OCM_FID0_OFST);
+ }
+
+ write_fault_count(priv);
+
+ return count;
+}
+
+static const struct file_operations inject_ce_fops = {
+ .open = simple_open,
+ .write = inject_ce_write,
+ .llseek = generic_file_llseek,
+};
+
+/*
+ * To get the Uncorrectable Error injected, the following steps are needed:
+ * - Setup the optional Fault Injection Count:
+ * echo <fault_count val> > /sys/kernel/debug/edac/ocm/inject_fault_count
+ * - Write the Uncorrectable Error bit position values:
+ * echo <bit_pos0 val>,<bit_pos1 val> > /sys/kernel/debug/edac/ocm/inject_ue_bitpos
+ */
+static ssize_t inject_ue_write(struct file *file, const char __user *data,
+ size_t count, loff_t *ppos)
+{
+ struct edac_device_ctl_info *edac_dev = file->private_data;
+ struct edac_priv *priv = edac_dev->pvt_info;
+ char buf[6], *pbuf, *token[2];
+ u64 ue_bitpos;
+ int i, ret;
+ u8 len;
+
+ if (!data)
+ return -EFAULT;
+
+ len = min_t(size_t, count, sizeof(buf));
+ if (copy_from_user(buf, data, len))
+ return -EFAULT;
+
+ buf[len] = '\0';
+ pbuf = &buf[0];
+ for (i = 0; i < OCM_NUM_UE_BITPOS; i++)
+ token[i] = strsep(&pbuf, ",");
+
+ ret = kstrtou8(token[0], 0, &priv->ue_bitpos[0]);
+ if (ret)
+ return ret;
+
+ ret = kstrtou8(token[1], 0, &priv->ue_bitpos[1]);
+ if (ret)
+ return ret;
+
+ if (priv->ue_bitpos[0] > UE_MAX_BITPOS_UPPER ||
+ priv->ue_bitpos[1] > UE_MAX_BITPOS_UPPER)
+ return -EINVAL;
+
+ if (priv->ue_bitpos[0] == priv->ue_bitpos[1]) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Bit positions should not be equal\n");
+ return -EINVAL;
+ }
+
+ ue_bitpos = BIT(priv->ue_bitpos[0]) | BIT(priv->ue_bitpos[1]);
+
+ writel((u32)ue_bitpos, priv->baseaddr + OCM_FID0_OFST);
+ writel((u32)(ue_bitpos >> 32), priv->baseaddr + OCM_FID1_OFST);
+
+ write_fault_count(priv);
+
+ return count;
+}
+
+static const struct file_operations inject_ue_fops = {
+ .open = simple_open,
+ .write = inject_ue_write,
+ .llseek = generic_file_llseek,
+};
+
+static void setup_debugfs(struct edac_device_ctl_info *edac_dev)
+{
+ struct edac_priv *priv = edac_dev->pvt_info;
+
+ priv->debugfs_dir = edac_debugfs_create_dir("ocm");
+ if (!priv->debugfs_dir)
+ return;
+
+ edac_debugfs_create_x32("inject_fault_count", 0644, priv->debugfs_dir,
+ &priv->fault_injection_cnt);
+ edac_debugfs_create_file("inject_ue_bitpos", 0644, priv->debugfs_dir,
+ edac_dev, &inject_ue_fops);
+ edac_debugfs_create_file("inject_ce_bitpos", 0644, priv->debugfs_dir,
+ edac_dev, &inject_ce_fops);
+}
+#endif
+
+static int edac_probe(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci;
+ struct edac_priv *priv;
+ void __iomem *baseaddr;
+ struct resource *res;
+ int irq, ret;
+
+ baseaddr = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
+ if (IS_ERR(baseaddr))
+ return PTR_ERR(baseaddr);
+
+ if (!get_eccstate(baseaddr)) {
+ edac_printk(KERN_INFO, EDAC_DEVICE, "ECC not enabled\n");
+ return -ENXIO;
+ }
+
+ dci = edac_device_alloc_ctl_info(sizeof(*priv), ZYNQMP_OCM_EDAC_STRING,
+ 1, ZYNQMP_OCM_EDAC_STRING, 1, 0, NULL, 0,
+ edac_device_alloc_index());
+ if (!dci)
+ return -ENOMEM;
+
+ priv = dci->pvt_info;
+ platform_set_drvdata(pdev, dci);
+ dci->dev = &pdev->dev;
+ priv->baseaddr = baseaddr;
+ dci->mod_name = pdev->dev.driver->name;
+ dci->ctl_name = ZYNQMP_OCM_EDAC_STRING;
+ dci->dev_name = dev_name(&pdev->dev);
+
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ ret = irq;
+ goto free_dev_ctl;
+ }
+
+ ret = devm_request_irq(&pdev->dev, irq, intr_handler, 0,
+ dev_name(&pdev->dev), dci);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Failed to request Irq\n");
+ goto free_dev_ctl;
+ }
+
+ /* Enable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IEN_OFST);
+
+#ifdef CONFIG_EDAC_DEBUG
+ setup_debugfs(dci);
+#endif
+
+ ret = edac_device_add_device(dci);
+ if (ret)
+ goto free_dev_ctl;
+
+ return 0;
+
+free_dev_ctl:
+ edac_device_free_ctl_info(dci);
+
+ return ret;
+}
+
+static int edac_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
+ struct edac_priv *priv = dci->pvt_info;
+
+ /* Disable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IDS_OFST);
+
+#ifdef CONFIG_EDAC_DEBUG
+ debugfs_remove_recursive(priv->debugfs_dir);
+#endif
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);
+
+ return 0;
+}
+
+static const struct of_device_id zynqmp_ocm_edac_match[] = {
+ { .compatible = "xlnx,zynqmp-ocmc-1.0"},
+ { /* end of table */ }
+};
+
+MODULE_DEVICE_TABLE(of, zynqmp_ocm_edac_match);
+
+static struct platform_driver zynqmp_ocm_edac_driver = {
+ .driver = {
+ .name = "zynqmp-ocm-edac",
+ .of_match_table = zynqmp_ocm_edac_match,
+ },
+ .probe = edac_probe,
+ .remove = edac_remove,
+};
+
+module_platform_driver(zynqmp_ocm_edac_driver);
+
+MODULE_AUTHOR("Advanced Micro Devices, Inc");
+MODULE_DESCRIPTION("Xilinx ZynqMP OCM ECC driver");
+MODULE_LICENSE("GPL");