diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2023-08-30 23:34:34 +0300 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2023-08-30 23:34:34 +0300 |
| commit | 461f35f014466c4e26dca6be0f431f57297df3f2 (patch) | |
| tree | 0bd2fded69ba0752ca16c304d3e1880d5f1eb30b /drivers/gpu/drm/amd/amdkfd | |
| parent | 53ea7f624fb91074c2f9458832ed74975ee5d64c (diff) | |
| parent | 3698a75f5a98d0a6599e2878ab25d30a82dd836a (diff) | |
| download | linux-461f35f014466c4e26dca6be0f431f57297df3f2.tar.xz | |
Merge tag 'drm-next-2023-08-30' of git://anongit.freedesktop.org/drm/drm
Pull drm updates from Dave Airlie:
"The drm core grew a new generic gpu virtual address manager, and new
execution locking helpers. These are used by nouveau now to provide
uAPI support for the userspace Vulkan driver. AMD had a bunch of new
IP core support, loads of refactoring around fbdev, but mostly just
the usual amount of stuff across the board.
core:
- fix gfp flags in drmm_kmalloc
gpuva:
- add new generic GPU VA manager (for nouveau initially)
syncobj:
- add new DRM_IOCTL_SYNCOBJ_EVENTFD ioctl
dma-buf:
- acquire resv lock for mmap() in exporters
- support dma-buf self import automatically
- docs fixes
backlight:
- fix fbdev interactions
atomic:
- improve logging
prime:
- remove struct gem_prim_mmap plus driver updates
gem:
- drm_exec: add locking over multiple GEM objects
- fix lockdep checking
fbdev:
- make fbdev userspace interfaces optional
- use linux device instead of fbdev device
- use deferred i/o helper macros in various drivers
- Make FB core selectable without drivers
- Remove obsolete flags FBINFO_DEFAULT and FBINFO_FLAG_DEFAULT
- Add helper macros and Kconfig tokens for DMA-allocated framebuffer
ttm:
- support init_on_free
- swapout fixes
panel:
- panel-edp: Support AUO B116XAB01.4
- Support Visionox R66451 plus DT bindings
- ld9040:
- Backlight support
- magic improved
- Kconfig fix
- Convert to of_device_get_match_data()
- Fix Kconfig dependencies
- simple:
- Set bpc value to fix warning
- Set connector type for AUO T215HVN01
- Support Innolux G156HCE-L01 plus DT bindings
- ili9881: Support TDO TL050HDV35 LCD panel plus DT bindings
- startek: Support KD070FHFID015 MIPI-DSI panel plus DT bindings
- sitronix-st7789v:
- Support Inanbo T28CP45TN89 plus DT bindings
- Support EDT ET028013DMA plus DT bindings
- Various cleanups
- edp: Add timings for N140HCA-EAC
- Allow panels and touchscreens to power sequence together
- Fix Innolux G156HCE-L01 LVDS clock
bridge:
- debugfs for chains support
- dw-hdmi:
- Improve support for YUV420 bus format
- CEC suspend/resume
- update EDID on HDMI detect
- dw-mipi-dsi: Fix enable/disable of DSI controller
- lt9611uxc: Use MODULE_FIRMWARE()
- ps8640: Remove broken EDID code
- samsung-dsim: Fix command transfer
- tc358764:
- Handle HS/VS polarity
- Use BIT() macro
- Various cleanups
- adv7511: Fix low refresh rate
- anx7625:
- Switch to macros instead of hardcoded values
- locking fixes
- tc358767: fix hardware delays
- sitronix-st7789v:
- Support panel orientation
- Support rotation property
- Add support for Jasonic JT240MHQS-HWT-EK-E3 plus DT bindings
amdgpu:
- SDMA 6.1.0 support
- HDP 6.1 support
- SMUIO 14.0 support
- PSP 14.0 support
- IH 6.1 support
- Lots of checkpatch cleanups
- GFX 9.4.3 updates
- Add USB PD and IFWI flashing documentation
- GPUVM updates
- RAS fixes
- DRR fixes
- FAMS fixes
- Virtual display fixes
- Soft IH fixes
- SMU13 fixes
- Rework PSP firmware loading for other IPs
- Kernel doc fixes
- DCN 3.0.1 fixes
- LTTPR fixes
- DP MST fixes
- DCN 3.1.6 fixes
- SMU 13.x fixes
- PSP 13.x fixes
- SubVP fixes
- GC 9.4.3 fixes
- Display bandwidth calculation fixes
- VCN4 secure submission fixes
- Allow building DC on RISC-V
- Add visible FB info to bo_print_info
- HBR3 fixes
- GFX9 MCBP fix
- GMC10 vmhub index fix
- GMC11 vmhub index fix
- Create a new doorbell manager
- SR-IOV fixes
- initial freesync panel replay support
- revert zpos properly until igt regression is fixeed
- use TTM to manage doorbell BAR
- Expose both current and average power via hwmon if supported
amdkfd:
- Cleanup CRIU dma-buf handling
- Use KIQ to unmap HIQ
- GFX 9.4.3 debugger updates
- GFX 9.4.2 debugger fixes
- Enable cooperative groups fof gfx11
- SVM fixes
- Convert older APUs to use dGPU path like newer APUs
- Drop IOMMUv2 path as it is no longer used
- TBA fix for aldebaran
i915:
- ICL+ DSI modeset sequence
- HDCP improvements
- MTL display fixes and cleanups
- HSW/BDW PSR1 restored
- Init DDI ports in VBT order
- General display refactors
- Start using plane scale factor for relative data rate
- Use shmem for dpt objects
- Expose RPS thresholds in sysfs
- Apply GuC SLPC min frequency softlimit correctly
- Extend Wa_14015795083 to TGL, RKL, DG1 and ADL
- Fix a VMA UAF for multi-gt platform
- Do not use stolen on MTL due to HW bug
- Check HuC and GuC version compatibility on MTL
- avoid infinite GPU waits due to premature release of request memory
- Fixes and updates for GSC memory allocation
- Display SDVO fixes
- Take stolen handling out of FBC code
- Make i915_coherent_map_type GT-centric
- Simplify shmem_create_from_object map_type
msm:
- SM6125 MDSS support
- DPU: SM6125 DPU support
- DSI: runtime PM support, burst mode support
- DSI PHY: SM6125 support in 14nm DSI PHY driver
- GPU: prepare for a7xx
- fix a690 firmware
- disable relocs on a6xx and newer
radeon:
- Lots of checkpatch cleanups
ast:
- improve device-model detection
- Represent BMV as virtual connector
- Report DP connection status
nouveau:
- add new exec/bind interface to support Vulkan
- document some getparam ioctls
- improve VRAM detection
- various fixes/cleanups
- workraound DPCD issues
ivpu:
- MMU updates
- debugfs support
- Support vpu4
virtio:
- add sync object support
atmel-hlcdc:
- Support inverted pixclock polarity
etnaviv:
- runtime PM cleanups
- hang handling fixes
exynos:
- use fbdev DMA helpers
- fix possible NULL ptr dereference
komeda:
- always attach encoder
omapdrm:
- use fbdev DMA helpers
ingenic:
- kconfig regmap fixes
loongson:
- support display controller
mediatek:
- Small mtk-dpi cleanups
- DisplayPort: support eDP and aux-bus
- Fix coverity issues
- Fix potential memory leak if vmap() fail
mgag200:
- minor fixes
mxsfb:
- support disabling overlay planes
panfrost:
- fix sync in IRQ handling
ssd130x:
- Support per-controller default resolution plus DT bindings
- Reduce memory-allocation overhead
- Improve intermediate buffer size computation
- Fix allocation of temporary buffers
- Fix pitch computation
- Fix shadow plane allocation
tegra:
- use fbdev DMA helpers
- Convert to devm_platform_ioremap_resource()
- support bridge/connector
- enable PM
tidss:
- Support TI AM625 plus DT bindings
- Implement new connector model plus driver updates
vkms:
- improve write back support
- docs fixes
- support gamma LUT
zynqmp-dpsub:
- misc fixes"
* tag 'drm-next-2023-08-30' of git://anongit.freedesktop.org/drm/drm: (1327 commits)
drm/gpuva_mgr: remove unused prev pointer in __drm_gpuva_sm_map()
drm/tests/drm_kunit_helpers: Place correct function name in the comment header
drm/nouveau: uapi: don't pass NO_PREFETCH flag implicitly
drm/nouveau: uvmm: fix unset region pointer on remap
drm/nouveau: sched: avoid job races between entities
drm/i915: Fix HPD polling, reenabling the output poll work as needed
drm: Add an HPD poll helper to reschedule the poll work
drm/i915: Fix TLB-Invalidation seqno store
drm/ttm/tests: Fix type conversion in ttm_pool_test
drm/msm/a6xx: Bail out early if setting GPU OOB fails
drm/msm/a6xx: Move LLC accessors to the common header
drm/msm/a6xx: Introduce a6xx_llc_read
drm/ttm/tests: Require MMU when testing
drm/panel: simple: Fix Innolux G156HCE-L01 LVDS clock
Revert "Revert "drm/amdgpu/display: change pipe policy for DCN 2.0""
drm/amdgpu: Add memory vendor information
drm/amd: flush any delayed gfxoff on suspend entry
drm/amdgpu: skip fence GFX interrupts disable/enable for S0ix
drm/amdgpu: Remove gfxoff check in GFX v9.4.3
drm/amd/pm: Update pci link speed for smu v13.0.6
...
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd')
37 files changed, 891 insertions, 1761 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig index 93bd4eda0d94..d3c3d3ab7225 100644 --- a/drivers/gpu/drm/amd/amdkfd/Kconfig +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -6,7 +6,6 @@ config HSA_AMD bool "HSA kernel driver for AMD GPU devices" depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64) - imply AMD_IOMMU_V2 if X86_64 select HMM_MIRROR select MMU_NOTIFIER select DRM_AMDGPU_USERPTR diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 2ec8f27c5366..a5ae7bcf44eb 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -59,10 +59,6 @@ AMDKFD_FILES := $(AMDKFD_PATH)/kfd_module.o \ $(AMDKFD_PATH)/kfd_crat.o \ $(AMDKFD_PATH)/kfd_debug.o -ifneq ($(CONFIG_AMD_IOMMU_V2),) -AMDKFD_FILES += $(AMDKFD_PATH)/kfd_iommu.o -endif - ifneq ($(CONFIG_DEBUG_FS),) AMDKFD_FILES += $(AMDKFD_PATH)/kfd_debugfs.o endif diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h index 73ca9aebf086..d7cd5fa313ff 100644 --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h @@ -274,16 +274,16 @@ static const uint32_t cwsr_trap_gfx8_hex[] = { static const uint32_t cwsr_trap_gfx9_hex[] = { - 0xbf820001, 0xbf820254, + 0xbf820001, 0xbf820258, 0xb8f8f802, 0x8978ff78, 0x00020006, 0xb8fbf803, 0x866eff78, 0x00002000, 0xbf840009, 0x866eff6d, 0x00ff0000, 0xbf85001e, 0x866eff7b, 0x00000400, - 0xbf850051, 0xbf8e0010, + 0xbf850055, 0xbf8e0010, 0xb8fbf803, 0xbf82fffa, - 0x866eff7b, 0x00000900, + 0x866eff7b, 0x03c00900, 0xbf850015, 0x866eff7b, 0x000071ff, 0xbf840008, 0x866fff7b, 0x00007080, @@ -294,13 +294,15 @@ static const uint32_t cwsr_trap_gfx9_hex[] = { 0xbf850007, 0xb8eef801, 0x866eff6e, 0x00000800, 0xbf850003, 0x866eff7b, - 0x00000400, 0xbf850036, + 0x00000400, 0xbf85003a, 0xb8faf807, 0x867aff7a, 0x001f8000, 0x8e7a8b7a, 0x8977ff77, 0xfc000000, 0x87777a77, 0xba7ff807, 0x00000000, 0xb8faf812, 0xb8fbf813, 0x8efa887a, + 0xbf0d8f7b, 0xbf840002, + 0x877bff7b, 0xffff0000, 0xc0031bbd, 0x00000010, 0xbf8cc07f, 0x8e6e976e, 0x8977ff77, 0x00800000, @@ -676,14 +678,14 @@ static const uint32_t cwsr_trap_gfx9_hex[] = { }; static const uint32_t cwsr_trap_nv1x_hex[] = { - 0xbf820001, 0xbf8201f1, + 0xbf820001, 0xbf8201f5, 0xb0804004, 0xb978f802, 0x8a78ff78, 0x00020006, 0xb97bf803, 0x876eff78, 0x00002000, 0xbf840009, 0x876eff6d, 0x00ff0000, 0xbf85001e, 0x876eff7b, - 0x00000400, 0xbf850057, + 0x00000400, 0xbf85005b, 0xbf8e0010, 0xb97bf803, 0xbf82fffa, 0x876eff7b, 0x00000900, 0xbf850015, @@ -697,7 +699,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 0xb96ef801, 0x876eff6e, 0x00000800, 0xbf850003, 0x876eff7b, 0x00000400, - 0xbf85003c, 0x8a77ff77, + 0xbf850040, 0x8a77ff77, 0xff000000, 0xb97af807, 0x877bff7a, 0x02000000, 0x8f7b867b, 0x88777b77, @@ -706,6 +708,8 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 0x8a7aff7a, 0x023f8000, 0xb9faf807, 0xb97af812, 0xb97bf813, 0x8ffa887a, + 0xbf0d8f7b, 0xbf840002, + 0x887bff7b, 0xffff0000, 0xf4011bbd, 0xfa000010, 0xbf8cc07f, 0x8f6e976e, 0x8a77ff77, 0x00800000, @@ -1094,16 +1098,16 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { }; static const uint32_t cwsr_trap_arcturus_hex[] = { - 0xbf820001, 0xbf8202d0, + 0xbf820001, 0xbf8202d4, 0xb8f8f802, 0x8978ff78, 0x00020006, 0xb8fbf803, 0x866eff78, 0x00002000, 0xbf840009, 0x866eff6d, 0x00ff0000, 0xbf85001e, 0x866eff7b, 0x00000400, - 0xbf850051, 0xbf8e0010, + 0xbf850055, 0xbf8e0010, 0xb8fbf803, 0xbf82fffa, - 0x866eff7b, 0x00000900, + 0x866eff7b, 0x03c00900, 0xbf850015, 0x866eff7b, 0x000071ff, 0xbf840008, 0x866fff7b, 0x00007080, @@ -1114,13 +1118,15 @@ static const uint32_t cwsr_trap_arcturus_hex[] = { 0xbf850007, 0xb8eef801, 0x866eff6e, 0x00000800, 0xbf850003, 0x866eff7b, - 0x00000400, 0xbf850036, + 0x00000400, 0xbf85003a, 0xb8faf807, 0x867aff7a, 0x001f8000, 0x8e7a8b7a, 0x8977ff77, 0xfc000000, 0x87777a77, 0xba7ff807, 0x00000000, 0xb8faf812, 0xb8fbf813, 0x8efa887a, + 0xbf0d8f7b, 0xbf840002, + 0x877bff7b, 0xffff0000, 0xc0031bbd, 0x00000010, 0xbf8cc07f, 0x8e6e976e, 0x8977ff77, 0x00800000, @@ -1572,16 +1578,16 @@ static const uint32_t cwsr_trap_arcturus_hex[] = { }; static const uint32_t cwsr_trap_aldebaran_hex[] = { - 0xbf820001, 0xbf8202db, + 0xbf820001, 0xbf8202df, 0xb8f8f802, 0x8978ff78, 0x00020006, 0xb8fbf803, 0x866eff78, 0x00002000, 0xbf840009, 0x866eff6d, 0x00ff0000, 0xbf85001e, 0x866eff7b, 0x00000400, - 0xbf850051, 0xbf8e0010, + 0xbf850055, 0xbf8e0010, 0xb8fbf803, 0xbf82fffa, - 0x866eff7b, 0x00000900, + 0x866eff7b, 0x03c00900, 0xbf850015, 0x866eff7b, 0x000071ff, 0xbf840008, 0x866fff7b, 0x00007080, @@ -1592,13 +1598,15 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = { 0xbf850007, 0xb8eef801, 0x866eff6e, 0x00000800, 0xbf850003, 0x866eff7b, - 0x00000400, 0xbf850036, + 0x00000400, 0xbf85003a, 0xb8faf807, 0x867aff7a, 0x001f8000, 0x8e7a8b7a, 0x8977ff77, 0xfc000000, 0x87777a77, 0xba7ff807, 0x00000000, 0xb8faf812, 0xb8fbf813, 0x8efa887a, + 0xbf0d8f7b, 0xbf840002, + 0x877bff7b, 0xffff0000, 0xc0031bbd, 0x00000010, 0xbf8cc07f, 0x8e6e976e, 0x8977ff77, 0x00800000, @@ -2061,14 +2069,14 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = { }; static const uint32_t cwsr_trap_gfx10_hex[] = { - 0xbf820001, 0xbf82021c, + 0xbf820001, 0xbf820220, 0xb0804004, 0xb978f802, 0x8a78ff78, 0x00020006, 0xb97bf803, 0x876eff78, 0x00002000, 0xbf840009, 0x876eff6d, 0x00ff0000, 0xbf85001e, 0x876eff7b, - 0x00000400, 0xbf850041, + 0x00000400, 0xbf850045, 0xbf8e0010, 0xb97bf803, 0xbf82fffa, 0x876eff7b, 0x00000900, 0xbf850015, @@ -2082,8 +2090,10 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 0xb96ef801, 0x876eff6e, 0x00000800, 0xbf850003, 0x876eff7b, 0x00000400, - 0xbf850026, 0xb97af812, + 0xbf85002a, 0xb97af812, 0xb97bf813, 0x8ffa887a, + 0xbf0d8f7b, 0xbf840002, + 0x887bff7b, 0xffff0000, 0xf4011bbd, 0xfa000010, 0xbf8cc07f, 0x8f6e976e, 0x8a77ff77, 0x00800000, @@ -2494,8 +2504,9 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 0xbf9f0000, 0xbf9f0000, 0xbf9f0000, 0x00000000, }; + static const uint32_t cwsr_trap_gfx11_hex[] = { - 0xbfa00001, 0xbfa00221, + 0xbfa00001, 0xbfa00225, 0xb0804006, 0xb8f8f802, 0x9178ff78, 0x00020006, 0xb8fbf803, 0xbf0d9e6d, @@ -2505,7 +2516,7 @@ static const uint32_t cwsr_trap_gfx11_hex[] = { 0xbfa10009, 0x8b6eff6d, 0x00ff0000, 0xbfa2001e, 0x8b6eff7b, 0x00000400, - 0xbfa20041, 0xbf830010, + 0xbfa20045, 0xbf830010, 0xb8fbf803, 0xbfa0fffa, 0x8b6eff7b, 0x00000900, 0xbfa20015, 0x8b6eff7b, @@ -2518,9 +2529,11 @@ static const uint32_t cwsr_trap_gfx11_hex[] = { 0xbfa20007, 0xb8eef801, 0x8b6eff6e, 0x00000800, 0xbfa20003, 0x8b6eff7b, - 0x00000400, 0xbfa20026, + 0x00000400, 0xbfa2002a, 0xbefa4d82, 0xbf89fc07, - 0x84fa887a, 0xf4005bbd, + 0x84fa887a, 0xbf0d8f7b, + 0xbfa10002, 0x8c7bff7b, + 0xffff0000, 0xf4005bbd, 0xf8000010, 0xbf89fc07, 0x846e976e, 0x9177ff77, 0x00800000, 0x8c776e77, @@ -2938,211 +2951,257 @@ static const uint32_t cwsr_trap_gfx11_hex[] = { }; static const uint32_t cwsr_trap_gfx9_4_3_hex[] = { - 0xbf820001, 0xbf8202d6, - 0xb8f8f802, 0x89788678, - 0xb8fbf803, 0x866eff78, - 0x00002000, 0xbf840009, - 0x866eff6d, 0x00ff0000, - 0xbf85001a, 0x866eff7b, - 0x00000400, 0xbf85004d, - 0xbf8e0010, 0xb8fbf803, - 0xbf82fffa, 0x866eff7b, - 0x03c00900, 0xbf850011, - 0x866eff7b, 0x000071ff, - 0xbf840008, 0x866fff7b, - 0x00007080, 0xbf840001, - 0xbeee1a87, 0xb8eff801, - 0x8e6e8c6e, 0x866e6f6e, - 0xbf850006, 0x866eff6d, - 0x00ff0000, 0xbf850003, + 0xbf820001, 0xbf8202db, + 0xb8f8f802, 0x8978ff78, + 0x00020006, 0xb8fbf803, + 0x866eff78, 0x00002000, + 0xbf840009, 0x866eff6d, + 0x00ff0000, 0xbf85001a, 0x866eff7b, 0x00000400, - 0xbf850036, 0xb8faf807, + 0xbf850051, 0xbf8e0010, + 0xb8fbf803, 0xbf82fffa, + 0x866eff7b, 0x03c00900, + 0xbf850011, 0x866eff7b, + 0x000071ff, 0xbf840008, + 0x866fff7b, 0x00007080, + 0xbf840001, 0xbeee1a87, + 0xb8eff801, 0x8e6e8c6e, + 0x866e6f6e, 0xbf850006, + 0x866eff6d, 0x00ff0000, + 0xbf850003, 0x866eff7b, + 0x00000400, 0xbf85003a, + 0xb8faf807, 0x867aff7a, + 0x001f8000, 0x8e7a8b7a, + 0x8979ff79, 0xfc000000, + 0x87797a79, 0xba7ff807, + 0x00000000, 0xb8faf812, + 0xb8fbf813, 0x8efa887a, + 0xbf0d8f7b, 0xbf840002, + 0x877bff7b, 0xffff0000, + 0xc0031bbd, 0x00000010, + 0xbf8cc07f, 0x8e6e976e, + 0x8979ff79, 0x00800000, + 0x87796e79, 0xc0071bbd, + 0x00000000, 0xbf8cc07f, + 0xc0071ebd, 0x00000008, + 0xbf8cc07f, 0x86ee6e6e, + 0xbf840001, 0xbe801d6e, + 0x866eff6d, 0x01ff0000, + 0xbf850005, 0x8778ff78, + 0x00002000, 0x80ec886c, + 0x82ed806d, 0xbf820005, + 0x866eff6d, 0x01000000, + 0xbf850002, 0x806c846c, + 0x826d806d, 0x866dff6d, + 0x0000ffff, 0x8f7a8b79, 0x867aff7a, 0x001f8000, - 0x8e7a8b7a, 0x8979ff79, - 0xfc000000, 0x87797a79, - 0xba7ff807, 0x00000000, - 0xb8faf812, 0xb8fbf813, - 0x8efa887a, 0xc0031bbd, - 0x00000010, 0xbf8cc07f, - 0x8e6e976e, 0x8979ff79, - 0x00800000, 0x87796e79, - 0xc0071bbd, 0x00000000, - 0xbf8cc07f, 0xc0071ebd, - 0x00000008, 0xbf8cc07f, - 0x86ee6e6e, 0xbf840001, - 0xbe801d6e, 0x866eff6d, - 0x01ff0000, 0xbf850005, - 0x8778ff78, 0x00002000, - 0x80ec886c, 0x82ed806d, - 0xbf820005, 0x866eff6d, - 0x01000000, 0xbf850002, - 0x806c846c, 0x826d806d, + 0xb97af807, 0x86fe7e7e, + 0x86ea6a6a, 0x8f6e8378, + 0xb96ee0c2, 0xbf800002, + 0xb9780002, 0xbe801f6c, 0x866dff6d, 0x0000ffff, - 0x8f7a8b79, 0x867aff7a, - 0x001f8000, 0xb97af807, - 0x86fe7e7e, 0x86ea6a6a, - 0x8f6e8378, 0xb96ee0c2, - 0xbf800002, 0xb9780002, - 0xbe801f6c, 0x866dff6d, - 0x0000ffff, 0xbefa0080, - 0xb97a0283, 0xb8faf807, - 0x867aff7a, 0x001f8000, - 0x8e7a8b7a, 0x8979ff79, - 0xfc000000, 0x87797a79, - 0xba7ff807, 0x00000000, - 0xbeee007e, 0xbeef007f, - 0xbefe0180, 0xbf900004, - 0x877a8478, 0xb97af802, - 0xbf8e0002, 0xbf88fffe, - 0xb8fa2985, 0x807a817a, - 0x8e7a8a7a, 0x8e7a817a, - 0xb8fb1605, 0x807b817b, - 0x8e7b867b, 0x807a7b7a, - 0x807a7e7a, 0x827b807f, - 0x867bff7b, 0x0000ffff, - 0xc04b1c3d, 0x00000050, - 0xbf8cc07f, 0xc04b1d3d, - 0x00000060, 0xbf8cc07f, - 0xc0431e7d, 0x00000074, - 0xbf8cc07f, 0xbef4007e, - 0x8675ff7f, 0x0000ffff, - 0x8775ff75, 0x00040000, - 0xbef60080, 0xbef700ff, - 0x00807fac, 0xbef1007c, - 0xbef00080, 0xb8f02985, - 0x80708170, 0x8e708a70, - 0x8e708170, 0xb8fa1605, - 0x807a817a, 0x8e7a867a, - 0x80707a70, 0xbef60084, - 0xbef600ff, 0x01000000, - 0xbefe007c, 0xbefc0070, - 0xc0611c7a, 0x0000007c, - 0xbf8cc07f, 0x80708470, - 0xbefc007e, 0xbefe007c, - 0xbefc0070, 0xc0611b3a, + 0xbefa0080, 0xb97a0283, + 0xb8faf807, 0x867aff7a, + 0x001f8000, 0x8e7a8b7a, + 0x8979ff79, 0xfc000000, + 0x87797a79, 0xba7ff807, + 0x00000000, 0xbeee007e, + 0xbeef007f, 0xbefe0180, + 0xbf900004, 0x877a8478, + 0xb97af802, 0xbf8e0002, + 0xbf88fffe, 0xb8fa2985, + 0x807a817a, 0x8e7a8a7a, + 0x8e7a817a, 0xb8fb1605, + 0x807b817b, 0x8e7b867b, + 0x807a7b7a, 0x807a7e7a, + 0x827b807f, 0x867bff7b, + 0x0000ffff, 0xc04b1c3d, + 0x00000050, 0xbf8cc07f, + 0xc04b1d3d, 0x00000060, + 0xbf8cc07f, 0xc0431e7d, + 0x00000074, 0xbf8cc07f, + 0xbef4007e, 0x8675ff7f, + 0x0000ffff, 0x8775ff75, + 0x00040000, 0xbef60080, + 0xbef700ff, 0x00807fac, + 0xbef1007c, 0xbef00080, + 0xb8f02985, 0x80708170, + 0x8e708a70, 0x8e708170, + 0xb8fa1605, 0x807a817a, + 0x8e7a867a, 0x80707a70, + 0xbef60084, 0xbef600ff, + 0x01000000, 0xbefe007c, + 0xbefc0070, 0xc0611c7a, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, 0xbefc0070, - 0xc0611b7a, 0x0000007c, + 0xc0611b3a, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, - 0xbefc0070, 0xc0611bba, + 0xbefc0070, 0xc0611b7a, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, 0xbefc0070, - 0xc0611bfa, 0x0000007c, + 0xc0611bba, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, - 0xbefc0070, 0xc0611e3a, - 0x0000007c, 0xbf8cc07f, - 0x80708470, 0xbefc007e, - 0xb8fbf803, 0xbefe007c, - 0xbefc0070, 0xc0611efa, + 0xbefc0070, 0xc0611bfa, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, 0xbefc0070, - 0xc0611a3a, 0x0000007c, + 0xc0611e3a, 0x0000007c, + 0xbf8cc07f, 0x80708470, + 0xbefc007e, 0xb8fbf803, + 0xbefe007c, 0xbefc0070, + 0xc0611efa, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, 0xbefe007c, - 0xbefc0070, 0xc0611a7a, - 0x0000007c, 0xbf8cc07f, - 0x80708470, 0xbefc007e, - 0xb8f1f801, 0xbefe007c, - 0xbefc0070, 0xc0611c7a, + 0xbefc0070, 0xc0611a3a, 0x0000007c, 0xbf8cc07f, 0x80708470, 0xbefc007e, - 0x867aff7f, 0x04000000, - 0xbeef0080, 0x876f6f7a, - 0xb8f02985, 0x80708170, - 0x8e708a70, 0x8e708170, - 0xb8fb1605, 0x807b817b, - 0x8e7b847b, 0x8e76827b, - 0xbef600ff, 0x01000000, - 0xbef20174, 0x80747074, - 0x82758075, 0xbefc0080, - 0xbf800000, 0xbe802b00, - 0xbe822b02, 0xbe842b04, - 0xbe862b06, 0xbe882b08, - 0xbe8a2b0a, 0xbe8c2b0c, - 0xbe8e2b0e, 0xc06b003a, - 0x00000000, 0xbf8cc07f, - 0xc06b013a, 0x00000010, - 0xbf8cc07f, 0xc06b023a, - 0x00000020, 0xbf8cc07f, - 0xc06b033a, 0x00000030, - 0xbf8cc07f, 0x8074c074, - 0x82758075, 0x807c907c, - 0xbf0a7b7c, 0xbf85ffe7, - 0xbef40172, 0xbef00080, - 0xbefe00c1, 0xbeff00c1, - 0xbee80080, 0xbee90080, - 0xbef600ff, 0x01000000, - 0x867aff78, 0x00400000, - 0xbf850003, 0xb8faf803, - 0x897a7aff, 0x10000000, - 0xbf85004d, 0xbe840080, - 0xd2890000, 0x00000900, - 0x80048104, 0xd2890001, + 0xbefe007c, 0xbefc0070, + 0xc0611a7a, 0x0000007c, + 0xbf8cc07f, 0x80708470, + 0xbefc007e, 0xb8f1f801, + 0xbefe007c, 0xbefc0070, + 0xc0611c7a, 0x0000007c, + 0xbf8cc07f, 0x80708470, + 0xbefc007e, 0x867aff7f, + 0x04000000, 0xbeef0080, + 0x876f6f7a, 0xb8f02985, + 0x80708170, 0x8e708a70, + 0x8e708170, 0xb8fb1605, + 0x807b817b, 0x8e7b847b, + 0x8e76827b, 0xbef600ff, + 0x01000000, 0xbef20174, + 0x80747074, 0x82758075, + 0xbefc0080, 0xbf800000, + 0xbe802b00, 0xbe822b02, + 0xbe842b04, 0xbe862b06, + 0xbe882b08, 0xbe8a2b0a, + 0xbe8c2b0c, 0xbe8e2b0e, + 0xc06b003a, 0x00000000, + 0xbf8cc07f, 0xc06b013a, + 0x00000010, 0xbf8cc07f, + 0xc06b023a, 0x00000020, + 0xbf8cc07f, 0xc06b033a, + 0x00000030, 0xbf8cc07f, + 0x8074c074, 0x82758075, + 0x807c907c, 0xbf0a7b7c, + 0xbf85ffe7, 0xbef40172, + 0xbef00080, 0xbefe00c1, + 0xbeff00c1, 0xbee80080, + 0xbee90080, 0xbef600ff, + 0x01000000, 0x867aff78, + 0x00400000, 0xbf850003, + 0xb8faf803, 0x897a7aff, + 0x10000000, 0xbf85004d, + 0xbe840080, 0xd2890000, 0x00000900, 0x80048104, - 0xd2890002, 0x00000900, - 0x80048104, 0xd2890003, + 0xd2890001, 0x00000900, + 0x80048104, 0xd2890002, 0x00000900, 0x80048104, + 0xd2890003, 0x00000900, + 0x80048104, 0xc069003a, + 0x00000070, 0xbf8cc07f, + 0x80709070, 0xbf06c004, + 0xbf84ffee, 0xbe840080, + 0xd2890000, 0x00000901, + 0x80048104, 0xd2890001, + 0x00000901, 0x80048104, + 0xd2890002, 0x00000901, + 0x80048104, 0xd2890003, + 0x00000901, 0x80048104, 0xc069003a, 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, 0xbf84ffee, 0xbe840080, 0xd2890000, - 0x00000901, 0x80048104, - 0xd2890001, 0x00000901, + 0x00000902, 0x80048104, + 0xd2890001, 0x00000902, 0x80048104, 0xd2890002, - 0x00000901, 0x80048104, - 0xd2890003, 0x00000901, + 0x00000902, 0x80048104, + 0xd2890003, 0x00000902, 0x80048104, 0xc069003a, 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, 0xbf84ffee, 0xbe840080, - 0xd2890000, 0x00000902, + 0xd2890000, 0x00000903, 0x80048104, 0xd2890001, - 0x00000902, 0x80048104, - 0xd2890002, 0x00000902, + 0x00000903, 0x80048104, + 0xd2890002, 0x00000903, 0x80048104, 0xd2890003, - 0x00000902, 0x80048104, + 0x00000903, 0x80048104, + 0xc069003a, 0x00000070, + 0xbf8cc07f, 0x80709070, + 0xbf06c004, 0xbf84ffee, + 0xbf820008, 0xe0724000, + 0x701d0000, 0xe0724100, + 0x701d0100, 0xe0724200, + 0x701d0200, 0xe0724300, + 0x701d0300, 0xbefe00c1, + 0xbeff00c1, 0xb8fb4306, + 0x867bc17b, 0xbf840064, + 0xbf8a0000, 0x867aff6f, + 0x04000000, 0xbf840060, + 0x8e7b867b, 0x8e7b827b, + 0xbef6007b, 0xb8f02985, + 0x80708170, 0x8e708a70, + 0x8e708170, 0xb8fa1605, + 0x807a817a, 0x8e7a867a, + 0x80707a70, 0x8070ff70, + 0x00000080, 0xbef600ff, + 0x01000000, 0xbefc0080, + 0xd28c0002, 0x000100c1, + 0xd28d0003, 0x000204c1, + 0x867aff78, 0x00400000, + 0xbf850003, 0xb8faf803, + 0x897a7aff, 0x10000000, + 0xbf850030, 0x24040682, + 0xd86e4000, 0x00000002, + 0xbf8cc07f, 0xbe840080, + 0xd2890000, 0x00000900, + 0x80048104, 0xd2890001, + 0x00000900, 0x80048104, + 0xd2890002, 0x00000900, + 0x80048104, 0xd2890003, + 0x00000900, 0x80048104, 0xc069003a, 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, 0xbf84ffee, 0xbe840080, 0xd2890000, - 0x00000903, 0x80048104, - 0xd2890001, 0x00000903, + 0x00000901, 0x80048104, + 0xd2890001, 0x00000901, 0x80048104, 0xd2890002, - 0x00000903, 0x80048104, - 0xd2890003, 0x00000903, + 0x00000901, 0x80048104, + 0xd2890003, 0x00000901, 0x80048104, 0xc069003a, 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, - 0xbf84ffee, 0xbf820008, - 0xe0724000, 0x701d0000, - 0xe0724100, 0x701d0100, - 0xe0724200, 0x701d0200, - 0xe0724300, 0x701d0300, + 0xbf84ffee, 0x680404ff, + 0x00000200, 0xd0c9006a, + 0x0000f702, 0xbf87ffd2, + 0xbf820015, 0xd1060002, + 0x00011103, 0x7e0602ff, + 0x00000200, 0xbefc00ff, + 0x00010000, 0xbe800077, + 0x8677ff77, 0xff7fffff, + 0x8777ff77, 0x00058000, + 0xd8ec0000, 0x00000002, + 0xbf8cc07f, 0xe0765000, + 0x701d0002, 0x68040702, + 0xd0c9006a, 0x0000f702, + 0xbf87fff7, 0xbef70000, + 0xbef000ff, 0x00000400, 0xbefe00c1, 0xbeff00c1, - 0xb8fb4306, 0x867bc17b, - 0xbf840064, 0xbf8a0000, - 0x867aff6f, 0x04000000, - 0xbf840060, 0x8e7b867b, - 0x8e7b827b, 0xbef6007b, - 0xb8f02985, 0x80708170, - 0x8e708a70, 0x8e708170, - 0xb8fa1605, 0x807a817a, - 0x8e7a867a, 0x80707a70, - 0x8070ff70, 0x00000080, - 0xbef600ff, 0x01000000, - 0xbefc0080, 0xd28c0002, - 0x000100c1, 0xd28d0003, - 0x000204c1, 0x867aff78, + 0xb8fb2b05, 0x807b817b, + 0x8e7b827b, 0xbef600ff, + 0x01000000, 0xbefc0084, + 0xbf0a7b7c, 0xbf84006d, + 0xbf11017c, 0x807bff7b, + 0x00001000, 0x867aff78, 0x00400000, 0xbf850003, 0xb8faf803, 0x897a7aff, - 0x10000000, 0xbf850030, - 0x24040682, 0xd86e4000, - 0x00000002, 0xbf8cc07f, + 0x10000000, 0xbf850051, 0xbe840080, 0xd2890000, 0x00000900, 0x80048104, 0xd2890001, 0x00000900, @@ -3162,31 +3221,51 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = { 0xc069003a, 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, 0xbf84ffee, - 0x680404ff, 0x00000200, - 0xd0c9006a, 0x0000f702, - 0xbf87ffd2, 0xbf820015, - 0xd1060002, 0x00011103, - 0x7e0602ff, 0x00000200, - 0xbefc00ff, 0x00010000, - 0xbe800077, 0x8677ff77, - 0xff7fffff, 0x8777ff77, - 0x00058000, 0xd8ec0000, - 0x00000002, 0xbf8cc07f, - 0xe0765000, 0x701d0002, - 0x68040702, 0xd0c9006a, - 0x0000f702, 0xbf87fff7, - 0xbef70000, 0xbef000ff, - 0x00000400, 0xbefe00c1, - 0xbeff00c1, 0xb8fb2b05, - 0x807b817b, 0x8e7b827b, - 0xbef600ff, 0x01000000, - 0xbefc0084, 0xbf0a7b7c, - 0xbf84006d, 0xbf11017c, + 0xbe840080, 0xd2890000, + 0x00000902, 0x80048104, + 0xd2890001, 0x00000902, + 0x80048104, 0xd2890002, + 0x00000902, 0x80048104, + 0xd2890003, 0x00000902, + 0x80048104, 0xc069003a, + 0x00000070, 0xbf8cc07f, + 0x80709070, 0xbf06c004, + 0xbf84ffee, 0xbe840080, + 0xd2890000, 0x00000903, + 0x80048104, 0xd2890001, + 0x00000903, 0x80048104, + 0xd2890002, 0x00000903, + 0x80048104, 0xd2890003, + 0x00000903, 0x80048104, + 0xc069003a, 0x00000070, + 0xbf8cc07f, 0x80709070, + 0xbf06c004, 0xbf84ffee, + 0x807c847c, 0xbf0a7b7c, + 0xbf85ffb1, 0xbf9c0000, + 0xbf820012, 0x7e000300, + 0x7e020301, 0x7e040302, + 0x7e060303, 0xe0724000, + 0x701d0000, 0xe0724100, + 0x701d0100, 0xe0724200, + 0x701d0200, 0xe0724300, + 0x701d0300, 0x807c847c, + 0x8070ff70, 0x00000400, + 0xbf0a7b7c, 0xbf85ffef, + 0xbf9c0000, 0xb8fb2985, + 0x807b817b, 0x8e7b837b, + 0xb8fa2b05, 0x807a817a, + 0x8e7a827a, 0x80fb7a7b, + 0x867b7b7b, 0xbf84007a, 0x807bff7b, 0x00001000, + 0xbefc0080, 0xbf11017c, 0x867aff78, 0x00400000, 0xbf850003, 0xb8faf803, 0x897a7aff, 0x10000000, - 0xbf850051, 0xbe840080, + 0xbf850059, 0xd3d84000, + 0x18000100, 0xd3d84001, + 0x18000101, 0xd3d84002, + 0x18000102, 0xd3d84003, + 0x18000103, 0xbe840080, 0xd2890000, 0x00000900, 0x80048104, 0xd2890001, 0x00000900, 0x80048104, @@ -3225,201 +3304,137 @@ static const uint32_t cwsr_trap_gfx9_4_3_hex[] = { 0x00000070, 0xbf8cc07f, 0x80709070, 0xbf06c004, 0xbf84ffee, 0x807c847c, - 0xbf0a7b7c, 0xbf85ffb1, - 0xbf9c0000, 0xbf820012, - 0x7e000300, 0x7e020301, - 0x7e040302, 0x7e060303, + 0xbf0a7b7c, 0xbf85ffa9, + 0xbf9c0000, 0xbf820016, + 0xd3d84000, 0x18000100, + 0xd3d84001, 0x18000101, + 0xd3d84002, 0x18000102, + 0xd3d84003, 0x18000103, 0xe0724000, 0x701d0000, 0xe0724100, 0x701d0100, 0xe0724200, 0x701d0200, 0xe0724300, 0x701d0300, 0x807c847c, 0x8070ff70, 0x00000400, 0xbf0a7b7c, - 0xbf85ffef, 0xbf9c0000, - 0xb8fb2985, 0x807b817b, - 0x8e7b837b, 0xb8fa2b05, - 0x807a817a, 0x8e7a827a, - 0x80fb7a7b, 0x867b7b7b, - 0xbf84007a, 0x807bff7b, - 0x00001000, 0xbefc0080, - 0xbf11017c, 0x867aff78, - 0x00400000, 0xbf850003, - 0xb8faf803, 0x897a7aff, - 0x10000000, 0xbf850059, - 0xd3d84000, 0x18000100, - 0xd3d84001, 0x18000101, - 0xd3d84002, 0x18000102, - 0xd3d84003, 0x18000103, - 0xbe840080, 0xd2890000, - 0x00000900, 0x80048104, - 0xd2890001, 0x00000900, - 0x80048104, 0xd2890002, - 0x00000900, 0x80048104, - 0xd2890003, 0x00000900, - 0x80048104, 0xc069003a, - 0x00000070, 0xbf8cc07f, - 0x80709070, 0xbf06c004, - 0xbf84ffee, 0xbe840080, - 0xd2890000, 0x00000901, - 0x80048104, 0xd2890001, - 0x00000901, 0x80048104, - 0xd2890002, 0x00000901, - 0x80048104, 0xd2890003, - 0x00000901, 0x80048104, - 0xc069003a, 0x00000070, - 0xbf8cc07f, 0x80709070, - 0xbf06c004, 0xbf84ffee, - 0xbe840080, 0xd2890000, - 0x00000902, 0x80048104, - 0xd2890001, 0x00000902, - 0x80048104, 0xd2890002, - 0x00000902, 0x80048104, - 0xd2890003, 0x00000902, - 0x80048104, 0xc069003a, - 0x00000070, 0xbf8cc07f, - 0x80709070, 0xbf06c004, - 0xbf84ffee, 0xbe840080, - 0xd2890000, 0x00000903, - 0x80048104, 0xd2890001, - 0x00000903, 0x80048104, - 0xd2890002, 0x00000903, - 0x80048104, 0xd2890003, - 0x00000903, 0x80048104, - 0xc069003a, 0x00000070, - 0xbf8cc07f, 0x80709070, - 0xbf06c004, 0xbf84ffee, - 0x807c847c, 0xbf0a7b7c, - 0xbf85ffa9, 0xbf9c0000, - 0xbf820016, 0xd3d84000, - 0x18000100, 0xd3d84001, - 0x18000101, 0xd3d84002, - 0x18000102, 0xd3d84003, - 0x18000103, 0xe0724000, - 0x701d0000, 0xe0724100, - 0x701d0100, 0xe0724200, - 0x701d0200, 0xe0724300, - 0x701d0300, 0x807c847c, - 0x8070ff70, 0x00000400, - 0xbf0a7b7c, 0xbf85ffeb, - 0xbf9c0000, 0xbf8200ee, - 0xbef4007e, 0x8675ff7f, - 0x0000ffff, 0x8775ff75, - 0x00040000, 0xbef60080, - 0xbef700ff, 0x00807fac, - 0x866eff7f, 0x04000000, - 0xbf84001f, 0xbefe00c1, - 0xbeff00c1, 0xb8ef4306, - 0x866fc16f, 0xbf84001a, - 0x8e6f866f, 0x8e6f826f, - 0xbef6006f, 0xb8f82985, - 0x80788178, 0x8e788a78, - 0x8e788178, 0xb8ee1605, - 0x806e816e, 0x8e6e866e, - 0x80786e78, 0x8078ff78, - 0x00000080, 0xbef600ff, - 0x01000000, 0xbefc0080, - 0xe0510000, 0x781d0000, - 0xe0510100, 0x781d0000, - 0x807cff7c, 0x00000200, - 0x8078ff78, 0x00000200, - 0xbf0a6f7c, 0xbf85fff6, + 0xbf85ffeb, 0xbf9c0000, + 0xbf8200ee, 0xbef4007e, + 0x8675ff7f, 0x0000ffff, + 0x8775ff75, 0x00040000, + 0xbef60080, 0xbef700ff, + 0x00807fac, 0x866eff7f, + 0x04000000, 0xbf84001f, 0xbefe00c1, 0xbeff00c1, + 0xb8ef4306, 0x866fc16f, + 0xbf84001a, 0x8e6f866f, + 0x8e6f826f, 0xbef6006f, + 0xb8f82985, 0x80788178, + 0x8e788a78, 0x8e788178, + 0xb8ee1605, 0x806e816e, + 0x8e6e866e, 0x80786e78, + 0x8078ff78, 0x00000080, 0xbef600ff, 0x01000000, - 0xb8ef2b05, 0x806f816f, - 0x8e6f826f, 0x806fff6f, - 0x00008000, 0xbef80080, - 0xbeee0078, 0x8078ff78, - 0x00000400, 0xbefc0084, + 0xbefc0080, 0xe0510000, + 0x781d0000, 0xe0510100, + 0x781d0000, 0x807cff7c, + 0x00000200, 0x8078ff78, + 0x00000200, 0xbf0a6f7c, + 0xbf85fff6, 0xbefe00c1, + 0xbeff00c1, 0xbef600ff, + 0x01000000, 0xb8ef2b05, + 0x806f816f, 0x8e6f826f, + 0x806fff6f, 0x00008000, + 0xbef80080, 0xbeee0078, + 0x8078ff78, 0x00000400, + 0xbefc0084, 0xbf11087c, + 0xe0524000, 0x781d0000, + 0xe0524100, 0x781d0100, + 0xe0524200, 0x781d0200, + 0xe0524300, 0x781d0300, + 0xbf8c0f70, 0x7e000300, + 0x7e020301, 0x7e040302, + 0x7e060303, 0x807c847c, + 0x8078ff78, 0x00000400, + 0xbf0a6f7c, 0xbf85ffee, + 0xb8ef2985, 0x806f816f, + 0x8e6f836f, 0xb8f92b05, + 0x80798179, 0x8e798279, + 0x80ef796f, 0x866f6f6f, + 0xbf84001a, 0x806fff6f, + 0x00008000, 0xbefc0080, 0xbf11087c, 0xe0524000, 0x781d0000, 0xe0524100, 0x781d0100, 0xe0524200, 0x781d0200, 0xe0524300, 0x781d0300, 0xbf8c0f70, - 0x7e000300, 0x7e020301, - 0x7e040302, 0x7e060303, + 0xd3d94000, 0x18000100, + 0xd3d94001, 0x18000101, + 0xd3d94002, 0x18000102, + 0xd3d94003, 0x18000103, 0x807c847c, 0x8078ff78, 0x00000400, 0xbf0a6f7c, - 0xbf85ffee, 0xb8ef2985, - 0x806f816f, 0x8e6f836f, - 0xb8f92b05, 0x80798179, - 0x8e798279, 0x80ef796f, - 0x866f6f6f, 0xbf84001a, - 0x806fff6f, 0x00008000, - 0xbefc0080, 0xbf11087c, - 0xe0524000, 0x781d0000, - 0xe0524100, 0x781d0100, - 0xe0524200, 0x781d0200, - 0xe0524300, 0x781d0300, - 0xbf8c0f70, 0xd3d94000, - 0x18000100, 0xd3d94001, - 0x18000101, 0xd3d94002, - 0x18000102, 0xd3d94003, - 0x18000103, 0x807c847c, - 0x8078ff78, 0x00000400, - 0xbf0a6f7c, 0xbf85ffea, - 0xbf9c0000, 0xe0524000, - 0x6e1d0000, 0xe0524100, - 0x6e1d0100, 0xe0524200, - 0x6e1d0200, 0xe0524300, - 0x6e1d0300, 0xbf8c0f70, - 0xb8f82985, 0x80788178, - 0x8e788a78, 0x8e788178, - 0xb8ee1605, 0x806e816e, - 0x8e6e866e, 0x80786e78, - 0x80f8c078, 0xb8ef1605, - 0x806f816f, 0x8e6f846f, - 0x8e76826f, 0xbef600ff, - 0x01000000, 0xbefc006f, - 0xc031003a, 0x00000078, - 0x80f8c078, 0xbf8cc07f, - 0x80fc907c, 0xbf800000, - 0xbe802d00, 0xbe822d02, - 0xbe842d04, 0xbe862d06, - 0xbe882d08, 0xbe8a2d0a, - 0xbe8c2d0c, 0xbe8e2d0e, - 0xbf06807c, 0xbf84fff0, - 0xb8f82985, 0x80788178, - 0x8e788a78, 0x8e788178, - 0xb8ee1605, 0x806e816e, - 0x8e6e866e, 0x80786e78, - 0xbef60084, 0xbef600ff, - 0x01000000, 0xc0211bfa, + 0xbf85ffea, 0xbf9c0000, + 0xe0524000, 0x6e1d0000, + 0xe0524100, 0x6e1d0100, + 0xe0524200, 0x6e1d0200, + 0xe0524300, 0x6e1d0300, + 0xbf8c0f70, 0xb8f82985, + 0x80788178, 0x8e788a78, + 0x8e788178, 0xb8ee1605, + 0x806e816e, 0x8e6e866e, + 0x80786e78, 0x80f8c078, + 0xb8ef1605, 0x806f816f, + 0x8e6f846f, 0x8e76826f, + 0xbef600ff, 0x01000000, + 0xbefc006f, 0xc031003a, + 0x00000078, 0x80f8c078, + 0xbf8cc07f, 0x80fc907c, + 0xbf800000, 0xbe802d00, + 0xbe822d02, 0xbe842d04, + 0xbe862d06, 0xbe882d08, + 0xbe8a2d0a, 0xbe8c2d0c, + 0xbe8e2d0e, 0xbf06807c, + 0xbf84fff0, 0xb8f82985, + 0x80788178, 0x8e788a78, + 0x8e788178, 0xb8ee1605, + 0x806e816e, 0x8e6e866e, + 0x80786e78, 0xbef60084, + 0xbef600ff, 0x01000000, + 0xc0211bfa, 0x00000078, + 0x80788478, 0xc0211b3a, 0x00000078, 0x80788478, - 0xc0211b3a, 0x00000078, - 0x80788478, 0xc0211b7a, + 0xc0211b7a, 0x00000078, + 0x80788478, 0xc0211c3a, 0x00000078, 0x80788478, - 0xc0211c3a, 0x00000078, - 0x80788478, 0xc0211c7a, + 0xc0211c7a, 0x00000078, + 0x80788478, 0xc0211eba, 0x00000078, 0x80788478, - 0xc0211eba, 0x00000078, - 0x80788478, 0xc0211efa, + 0xc0211efa, 0x00000078, + 0x80788478, 0xc0211a3a, 0x00000078, 0x80788478, - 0xc0211a3a, 0x00000078, - 0x80788478, 0xc0211a7a, + 0xc0211a7a, 0x00000078, + 0x80788478, 0xc0211cfa, 0x00000078, 0x80788478, - 0xc0211cfa, 0x00000078, - 0x80788478, 0xbf8cc07f, - 0xbefc006f, 0xbefe0070, - 0xbeff0071, 0x866f7bff, - 0x000003ff, 0xb96f4803, - 0x866f7bff, 0xfffff800, - 0x8f6f8b6f, 0xb96fa2c3, - 0xb973f801, 0xb8ee2985, - 0x806e816e, 0x8e6e8a6e, - 0x8e6e816e, 0xb8ef1605, - 0x806f816f, 0x8e6f866f, - 0x806e6f6e, 0x806e746e, - 0x826f8075, 0x866fff6f, - 0x0000ffff, 0xc00b1c37, - 0x00000050, 0xc00b1d37, - 0x00000060, 0xc0031e77, - 0x00000074, 0xbf8cc07f, - 0x8f6e8b79, 0x866eff6e, - 0x001f8000, 0xb96ef807, - 0x866dff6d, 0x0000ffff, - 0x86fe7e7e, 0x86ea6a6a, - 0x8f6e837a, 0xb96ee0c2, - 0xbf800002, 0xb97a0002, - 0xbf8a0000, 0xbe801f6c, - 0xbf810000, 0x00000000, + 0xbf8cc07f, 0xbefc006f, + 0xbefe0070, 0xbeff0071, + 0x866f7bff, 0x000003ff, + 0xb96f4803, 0x866f7bff, + 0xfffff800, 0x8f6f8b6f, + 0xb96fa2c3, 0xb973f801, + 0xb8ee2985, 0x806e816e, + 0x8e6e8a6e, 0x8e6e816e, + 0xb8ef1605, 0x806f816f, + 0x8e6f866f, 0x806e6f6e, + 0x806e746e, 0x826f8075, + 0x866fff6f, 0x0000ffff, + 0xc00b1c37, 0x00000050, + 0xc00b1d37, 0x00000060, + 0xc0031e77, 0x00000074, + 0xbf8cc07f, 0x8f6e8b79, + 0x866eff6e, 0x001f8000, + 0xb96ef807, 0x866dff6d, + 0x0000ffff, 0x86fe7e7e, + 0x86ea6a6a, 0x8f6e837a, + 0xb96ee0c2, 0xbf800002, + 0xb97a0002, 0xbf8a0000, + 0xbe801f6c, 0xbf810000, }; diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm index 8b92c33c2a7c..fdab64624422 100644 --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm @@ -276,6 +276,11 @@ L_FETCH_2ND_TRAP: #endif s_lshl_b64 [ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 + s_bitcmp1_b32 ttmp15, 0xF + s_cbranch_scc0 L_NO_SIGN_EXTEND_TMA + s_or_b32 ttmp15, ttmp15, 0xFFFF0000 +L_NO_SIGN_EXTEND_TMA: + s_load_dword ttmp2, [ttmp14, ttmp15], 0x10 glc:1 // debug trap enabled flag s_waitcnt lgkmcnt(0) s_lshl_b32 ttmp2, ttmp2, TTMP11_DEBUG_TRAP_ENABLED_SHIFT diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm index f2087cc2e89d..e506411ad28a 100644 --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm @@ -283,6 +283,11 @@ L_FETCH_2ND_TRAP: s_getreg_b32 ttmp15, hwreg(HW_REG_SQ_SHADER_TMA_HI) s_lshl_b64 [ttmp14, ttmp15], [ttmp14, ttmp15], 0x8 + s_bitcmp1_b32 ttmp15, 0xF + s_cbranch_scc0 L_NO_SIGN_EXTEND_TMA + s_or_b32 ttmp15, ttmp15, 0xFFFF0000 +L_NO_SIGN_EXTEND_TMA: + s_load_dword ttmp2, [ttmp14, ttmp15], 0x10 glc:1 // debug trap enabled flag s_waitcnt lgkmcnt(0) s_lshl_b32 ttmp2, ttmp2, TTMP_DEBUG_TRAP_ENABLED_SHIFT diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 6a27b000a246..c37f1fcd2165 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -333,10 +333,12 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, goto err_bind_process; } - if (!pdd->doorbell_index && - kfd_alloc_process_doorbells(dev->kfd, &pdd->doorbell_index) < 0) { - err = -ENOMEM; - goto err_alloc_doorbells; + if (!pdd->qpd.proc_doorbells) { + err = kfd_alloc_process_doorbells(dev->kfd, pdd); + if (err) { + pr_debug("failed to allocate process doorbells\n"); + goto err_bind_process; + } } /* Starting with GFX11, wptr BOs must be mapped to GART for MES to determine work @@ -417,7 +419,6 @@ err_create_queue: if (wptr_bo) amdgpu_amdkfd_free_gtt_mem(dev->adev, wptr_bo); err_wptr_map_gart: -err_alloc_doorbells: err_bind_process: err_pdd: mutex_unlock(&p->mutex); @@ -1025,9 +1026,6 @@ bool kfd_dev_is_large_bar(struct kfd_node *dev) return true; } - if (dev->kfd->use_iommu_v2) - return false; - if (dev->local_mem_info.local_mem_size_private == 0 && dev->local_mem_info.local_mem_size_public > 0) return true; @@ -1487,7 +1485,8 @@ static int kfd_ioctl_alloc_queue_gws(struct file *filep, goto out_unlock; } - if (!kfd_dbg_has_gws_support(dev) && p->debug_trap_enabled) { + if (p->debug_trap_enabled && (!kfd_dbg_has_gws_support(dev) || + kfd_dbg_has_cwsr_workaround(dev))) { retval = -EBUSY; goto out_unlock; } @@ -1845,22 +1844,21 @@ static uint32_t get_process_num_bos(struct kfd_process *p) idr_for_each_entry(&pdd->alloc_idr, mem, id) { struct kgd_mem *kgd_mem = (struct kgd_mem *)mem; - if ((uint64_t)kgd_mem->va > pdd->gpuvm_base) + if (!kgd_mem->va || kgd_mem->va > pdd->gpuvm_base) num_of_bos++; } } return num_of_bos; } -static int criu_get_prime_handle(struct drm_gem_object *gobj, int flags, +static int criu_get_prime_handle(struct kgd_mem *mem, int flags, u32 *shared_fd) { struct dma_buf *dmabuf; int ret; - dmabuf = amdgpu_gem_prime_export(gobj, flags); - if (IS_ERR(dmabuf)) { - ret = PTR_ERR(dmabuf); + ret = amdgpu_amdkfd_gpuvm_export_dmabuf(mem, &dmabuf); + if (ret) { pr_err("dmabuf export failed for the BO\n"); return ret; } @@ -1918,7 +1916,11 @@ static int criu_checkpoint_bos(struct kfd_process *p, kgd_mem = (struct kgd_mem *)mem; dumper_bo = kgd_mem->bo; - if ((uint64_t)kgd_mem->va <= pdd->gpuvm_base) + /* Skip checkpointing BOs that are used for Trap handler + * code and state. Currently, these BOs have a VA that + * is less GPUVM Base + */ + if (kgd_mem->va && kgd_mem->va <= pdd->gpuvm_base) continue; bo_bucket = &bo_buckets[bo_index]; @@ -1940,7 +1942,7 @@ static int criu_checkpoint_bos(struct kfd_process *p, } if (bo_bucket->alloc_flags & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT)) { - ret = criu_get_prime_handle(&dumper_bo->tbo.base, + ret = criu_get_prime_handle(kgd_mem, bo_bucket->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ? DRM_RDWR : 0, &bo_bucket->dmabuf_fd); @@ -2262,10 +2264,10 @@ static int criu_restore_devices(struct kfd_process *p, goto exit; } - if (!pdd->doorbell_index && - kfd_alloc_process_doorbells(pdd->dev->kfd, &pdd->doorbell_index) < 0) { - ret = -ENOMEM; - goto exit; + if (!pdd->qpd.proc_doorbells) { + ret = kfd_alloc_process_doorbells(dev->kfd, pdd); + if (ret) + goto exit; } } @@ -2402,7 +2404,7 @@ static int criu_restore_bo(struct kfd_process *p, /* create the dmabuf object and export the bo */ if (bo_bucket->alloc_flags & (KFD_IOC_ALLOC_MEM_FLAGS_VRAM | KFD_IOC_ALLOC_MEM_FLAGS_GTT)) { - ret = criu_get_prime_handle(&kgd_mem->bo->tbo.base, DRM_RDWR, + ret = criu_get_prime_handle(kgd_mem, DRM_RDWR, &bo_bucket->dmabuf_fd); if (ret) return ret; @@ -2755,6 +2757,16 @@ static int runtime_enable(struct kfd_process *p, uint64_t r_debug, if (pdd->qpd.queue_count) return -EEXIST; + + /* + * Setup TTMPs by default. + * Note that this call must remain here for MES ADD QUEUE to + * skip_process_ctx_clear unconditionally as the first call to + * SET_SHADER_DEBUGGER clears any stale process context data + * saved in MES. + */ + if (pdd->dev->kfd->shared_resources.enable_mes) + kfd_dbg_set_mes_debug_mode(pdd, !kfd_dbg_has_cwsr_workaround(pdd->dev)); } p->runtime_info.runtime_state = DEBUG_RUNTIME_STATE_ENABLED; @@ -2848,7 +2860,8 @@ static int runtime_disable(struct kfd_process *p) if (!pdd->dev->kfd->shared_resources.enable_mes) debug_refresh_runlist(pdd->dev->dqm); else - kfd_dbg_set_mes_debug_mode(pdd); + kfd_dbg_set_mes_debug_mode(pdd, + !kfd_dbg_has_cwsr_workaround(pdd->dev)); } } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index f5a6f562e2a8..86fb7ac7982a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -26,7 +26,6 @@ #include "kfd_crat.h" #include "kfd_priv.h" #include "kfd_topology.h" -#include "kfd_iommu.h" #include "amdgpu.h" #include "amdgpu_amdkfd.h" @@ -1536,72 +1535,6 @@ int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pc return num_of_cache_types; } -static bool kfd_ignore_crat(void) -{ - bool ret; - - if (ignore_crat) - return true; - - ret = true; - - return ret; -} - -/* - * kfd_create_crat_image_acpi - Allocates memory for CRAT image and - * copies CRAT from ACPI (if available). - * NOTE: Call kfd_destroy_crat_image to free CRAT image memory - * - * @crat_image: CRAT read from ACPI. If no CRAT in ACPI then - * crat_image will be NULL - * @size: [OUT] size of crat_image - * - * Return 0 if successful else return error code - */ -int kfd_create_crat_image_acpi(void **crat_image, size_t *size) -{ - struct acpi_table_header *crat_table; - acpi_status status; - void *pcrat_image; - int rc = 0; - - if (!crat_image) - return -EINVAL; - - *crat_image = NULL; - - if (kfd_ignore_crat()) { - pr_info("CRAT table disabled by module option\n"); - return -ENODATA; - } - - /* Fetch the CRAT table from ACPI */ - status = acpi_get_table(CRAT_SIGNATURE, 0, &crat_table); - if (status == AE_NOT_FOUND) { - pr_info("CRAT table not found\n"); - return -ENODATA; - } else if (ACPI_FAILURE(status)) { - const char *err = acpi_format_exception(status); - - pr_err("CRAT table error: %s\n", err); - return -EINVAL; - } - - pcrat_image = kvmalloc(crat_table->length, GFP_KERNEL); - if (!pcrat_image) { - rc = -ENOMEM; - goto out; - } - - memcpy(pcrat_image, crat_table, crat_table->length); - *crat_image = pcrat_image; - *size = crat_table->length; -out: - acpi_put_table(crat_table); - return rc; -} - /* Memory required to create Virtual CRAT. * Since there is no easy way to predict the amount of memory required, the * following amount is allocated for GPU Virtual CRAT. This is @@ -2169,12 +2102,6 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image, cu->hsa_capability = 0; - /* Check if this node supports IOMMU. During parsing this flag will - * translate to HSA_CAP_ATS_PRESENT - */ - if (!kfd_iommu_check_device(kdev->kfd)) - cu->hsa_capability |= CRAT_CU_FLAGS_IOMMU_PRESENT; - crat_table->length += sub_type_hdr->length; crat_table->total_entries++; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h index fc719389b5d6..387a8ef49385 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.h @@ -307,7 +307,6 @@ struct kfd_gpu_cache_info { }; int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pcache_info); -int kfd_create_crat_image_acpi(void **crat_image, size_t *size); void kfd_destroy_crat_image(void *crat_image); int kfd_parse_crat_table(void *crat_image, struct list_head *device_list, uint32_t proximity_domain); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c index 9766076e9ec4..9ec750666382 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c @@ -344,11 +344,10 @@ unwind: return r; } -int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd) +int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd, bool sq_trap_en) { uint32_t spi_dbg_cntl = pdd->spi_dbg_override | pdd->spi_dbg_launch_mode; uint32_t flags = pdd->process->dbg_flags; - bool sq_trap_en = !!spi_dbg_cntl || !kfd_dbg_has_cwsr_workaround(pdd->dev); if (!kfd_dbg_is_per_vmid_supported(pdd->dev)) return 0; @@ -432,7 +431,7 @@ int kfd_dbg_trap_clear_dev_address_watch(struct kfd_process_device *pdd, if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_map_and_unlock(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); kfd_dbg_clear_dev_watch_id(pdd, watch_id); @@ -445,7 +444,8 @@ int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd, uint32_t *watch_id, uint32_t watch_mode) { - int r = kfd_dbg_get_dev_watch_id(pdd, watch_id); + int xcc_id, r = kfd_dbg_get_dev_watch_id(pdd, watch_id); + uint32_t xcc_mask = pdd->dev->xcc_mask; if (r) return r; @@ -459,19 +459,21 @@ int kfd_dbg_trap_set_dev_address_watch(struct kfd_process_device *pdd, } amdgpu_gfx_off_ctrl(pdd->dev->adev, false); - pdd->watch_points[*watch_id] = pdd->dev->kfd2kgd->set_address_watch( + for_each_inst(xcc_id, xcc_mask) + pdd->watch_points[*watch_id] = pdd->dev->kfd2kgd->set_address_watch( pdd->dev->adev, watch_address, watch_address_mask, *watch_id, watch_mode, - pdd->dev->vm_info.last_vmid_kfd); + pdd->dev->vm_info.last_vmid_kfd, + xcc_id); amdgpu_gfx_off_ctrl(pdd->dev->adev, true); if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_map_and_unlock(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); /* HWS is broken so no point in HW rollback but release the watchpoint anyways */ if (r) @@ -513,7 +515,7 @@ int kfd_dbg_trap_set_flags(struct kfd_process *target, uint32_t *flags) if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_refresh_runlist(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); if (r) { target->dbg_flags = prev_flags; @@ -536,7 +538,7 @@ int kfd_dbg_trap_set_flags(struct kfd_process *target, uint32_t *flags) if (!pdd->dev->kfd->shared_resources.enable_mes) debug_refresh_runlist(pdd->dev->dqm); else - kfd_dbg_set_mes_debug_mode(pdd); + kfd_dbg_set_mes_debug_mode(pdd, true); } } @@ -598,7 +600,7 @@ void kfd_dbg_trap_deactivate(struct kfd_process *target, bool unwind, int unwind if (!pdd->dev->kfd->shared_resources.enable_mes) debug_refresh_runlist(pdd->dev->dqm); else - kfd_dbg_set_mes_debug_mode(pdd); + kfd_dbg_set_mes_debug_mode(pdd, !kfd_dbg_has_cwsr_workaround(pdd->dev)); } kfd_dbg_set_workaround(target, false); @@ -714,7 +716,7 @@ int kfd_dbg_trap_activate(struct kfd_process *target) if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_refresh_runlist(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); if (r) { target->runtime_info.runtime_state = @@ -750,7 +752,8 @@ int kfd_dbg_trap_enable(struct kfd_process *target, uint32_t fd, if (!KFD_IS_SOC15(pdd->dev)) return -ENODEV; - if (!kfd_dbg_has_gws_support(pdd->dev) && pdd->qpd.num_gws) + if (pdd->qpd.num_gws && (!kfd_dbg_has_gws_support(pdd->dev) || + kfd_dbg_has_cwsr_workaround(pdd->dev))) return -EBUSY; } @@ -847,7 +850,7 @@ int kfd_dbg_trap_set_wave_launch_override(struct kfd_process *target, if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_refresh_runlist(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); if (r) break; @@ -879,7 +882,7 @@ int kfd_dbg_trap_set_wave_launch_mode(struct kfd_process *target, if (!pdd->dev->kfd->shared_resources.enable_mes) r = debug_refresh_runlist(pdd->dev->dqm); else - r = kfd_dbg_set_mes_debug_mode(pdd); + r = kfd_dbg_set_mes_debug_mode(pdd, true); if (r) break; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h index 662a13a0d582..fd0ff64d4184 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h @@ -76,8 +76,9 @@ int kfd_dbg_send_exception_to_runtime(struct kfd_process *p, static inline bool kfd_dbg_is_per_vmid_supported(struct kfd_node *dev) { - return KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2) || - KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0); + return (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2) || + KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3) || + KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0)); } void debug_event_write_work_handler(struct work_struct *work); @@ -125,5 +126,14 @@ static inline bool kfd_dbg_has_gws_support(struct kfd_node *dev) return true; } -int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd); +int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd, bool sq_trap_en); + +static inline bool kfd_dbg_has_ttmps_always_setup(struct kfd_node *dev) +{ + return (KFD_GC_VERSION(dev) < IP_VERSION(11, 0, 0) && + KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 2)) || + (KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0) && + KFD_GC_VERSION(dev) < IP_VERSION(12, 0, 0) && + (dev->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 70); +} #endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index a53e0757fe64..93ce181eb3ba 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -29,7 +29,6 @@ #include "kfd_pm4_headers_vi.h" #include "kfd_pm4_headers_aldebaran.h" #include "cwsr_trap_handler.h" -#include "kfd_iommu.h" #include "amdgpu_amdkfd.h" #include "kfd_smi_events.h" #include "kfd_svm.h" @@ -62,7 +61,6 @@ static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size, unsigned int chunk_size); static void kfd_gtt_sa_fini(struct kfd_dev *kfd); -static int kfd_resume_iommu(struct kfd_dev *kfd); static int kfd_resume(struct kfd_node *kfd); static void kfd_device_info_set_sdma_info(struct kfd_dev *kfd) @@ -442,8 +440,6 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf) atomic_set(&kfd->compute_profile, 0); mutex_init(&kfd->doorbell_mutex); - memset(&kfd->doorbell_available_index, 0, - sizeof(kfd->doorbell_available_index)); ida_init(&kfd->doorbell_ida); @@ -495,6 +491,7 @@ static int kfd_gws_init(struct kfd_node *node) { int ret = 0; struct kfd_dev *kfd = node->kfd; + uint32_t mes_rev = node->adev->mes.sched_version & AMDGPU_MES_VERSION_MASK; if (node->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS) return 0; @@ -511,7 +508,10 @@ static int kfd_gws_init(struct kfd_node *node) (KFD_GC_VERSION(node) == IP_VERSION(9, 4, 3)) || (KFD_GC_VERSION(node) >= IP_VERSION(10, 3, 0) && KFD_GC_VERSION(node) < IP_VERSION(11, 0, 0) - && kfd->mec2_fw_version >= 0x6b)))) + && kfd->mec2_fw_version >= 0x6b) || + (KFD_GC_VERSION(node) >= IP_VERSION(11, 0, 0) + && KFD_GC_VERSION(node) < IP_VERSION(12, 0, 0) + && mes_rev >= 68)))) ret = amdgpu_amdkfd_alloc_gws(node->adev, node->adev->gds.gws_size, &node->gws); @@ -753,15 +753,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, kfd->noretry = kfd->adev->gmc.noretry; - /* If CRAT is broken, won't set iommu enabled */ - kfd_double_confirm_iommu_support(kfd); - - if (kfd_iommu_device_init(kfd)) { - kfd->use_iommu_v2 = false; - dev_err(kfd_device, "Error initializing iommuv2\n"); - goto device_iommu_error; - } - kfd_cwsr_init(kfd); dev_info(kfd_device, "Total number of KFD nodes to be created: %d\n", @@ -836,9 +827,6 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, svm_range_set_max_pages(kfd->adev); - if (kfd_resume_iommu(kfd)) - goto kfd_resume_iommu_error; - spin_lock_init(&kfd->watch_points_lock); kfd->init_complete = true; @@ -850,11 +838,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto out; -kfd_resume_iommu_error: node_init_error: node_alloc_error: kfd_cleanup_nodes(kfd, i); -device_iommu_error: kfd_doorbell_fini(kfd); kfd_doorbell_error: kfd_gtt_sa_fini(kfd); @@ -969,7 +955,6 @@ void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm) node = kfd->nodes[i]; node->dqm->ops.stop(node->dqm); } - kfd_iommu_suspend(kfd); } int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm) @@ -999,26 +984,6 @@ int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm) return ret; } -int kgd2kfd_resume_iommu(struct kfd_dev *kfd) -{ - if (!kfd->init_complete) - return 0; - - return kfd_resume_iommu(kfd); -} - -static int kfd_resume_iommu(struct kfd_dev *kfd) -{ - int err = 0; - - err = kfd_iommu_resume(kfd); - if (err) - dev_err(kfd_device, - "Failed to resume IOMMU for device %x:%x\n", - kfd->adev->pdev->vendor, kfd->adev->pdev->device); - return err; -} - static int kfd_resume(struct kfd_node *node) { int err = 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 01192f5abe46..b166f30f083e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -227,7 +227,8 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, queue_input.tba_addr = qpd->tba_addr; queue_input.tma_addr = qpd->tma_addr; queue_input.trap_en = !kfd_dbg_has_cwsr_workaround(q->device); - queue_input.skip_process_ctx_clear = qpd->pqm->process->debug_trap_enabled; + queue_input.skip_process_ctx_clear = qpd->pqm->process->debug_trap_enabled || + kfd_dbg_has_ttmps_always_setup(q->device); queue_type = convert_to_mes_queue_type(q->properties.type); if (queue_type < 0) { @@ -237,10 +238,7 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, } queue_input.queue_type = (uint32_t)queue_type; - if (q->gws) { - queue_input.gws_base = 0; - queue_input.gws_size = qpd->num_gws; - } + queue_input.exclusively_scheduled = q->properties.is_gws; amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input); @@ -250,7 +248,7 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, q->properties.doorbell_off); pr_err("MES might be in unrecoverable state, issue a GPU reset\n"); kfd_hws_hang(dqm); -} + } return r; } @@ -397,7 +395,7 @@ static int allocate_doorbell(struct qcm_process_device *qpd, unsigned int found; found = find_first_zero_bit(qpd->doorbell_bitmap, - KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) { pr_debug("No doorbells available"); return -EBUSY; @@ -407,9 +405,9 @@ static int allocate_doorbell(struct qcm_process_device *qpd, } } - q->properties.doorbell_off = - kfd_get_doorbell_dw_offset_in_bar(dev->kfd, qpd_to_pdd(qpd), - q->doorbell_id); + q->properties.doorbell_off = amdgpu_doorbell_index_on_bar(dev->adev, + qpd->proc_doorbells, + q->doorbell_id); return 0; } @@ -1620,7 +1618,8 @@ static int initialize_cpsch(struct device_queue_manager *dqm) if (dqm->dev->kfd2kgd->get_iq_wait_times) dqm->dev->kfd2kgd->get_iq_wait_times(dqm->dev->adev, - &dqm->wait_times); + &dqm->wait_times, + ffs(dqm->dev->xcc_mask) - 1); return 0; } @@ -1662,6 +1661,26 @@ static int start_cpsch(struct device_queue_manager *dqm) if (!dqm->dev->kfd->shared_resources.enable_mes) execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD); + + /* Set CWSR grace period to 1x1000 cycle for GFX9.4.3 APU */ + if (amdgpu_emu_mode == 0 && dqm->dev->adev->gmc.is_app_apu && + (KFD_GC_VERSION(dqm->dev) == IP_VERSION(9, 4, 3))) { + uint32_t reg_offset = 0; + uint32_t grace_period = 1; + + retval = pm_update_grace_period(&dqm->packet_mgr, + grace_period); + if (retval) + pr_err("Setting grace timeout failed\n"); + else if (dqm->dev->kfd2kgd->build_grace_period_packet_info) + /* Update dqm->wait_times maintained in software */ + dqm->dev->kfd2kgd->build_grace_period_packet_info( + dqm->dev->adev, dqm->wait_times, + grace_period, ®_offset, + &dqm->wait_times, + ffs(dqm->dev->xcc_mask) - 1); + } + dqm_unlock(dqm); return 0; @@ -2540,7 +2559,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev) switch (dev->adev->asic_type) { case CHIP_KAVERI: case CHIP_HAWAII: - device_queue_manager_init_cik_hawaii(&dqm->asic_ops); + device_queue_manager_init_cik(&dqm->asic_ops); break; case CHIP_CARRIZO: @@ -2550,14 +2569,14 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_node *dev) case CHIP_POLARIS11: case CHIP_POLARIS12: case CHIP_VEGAM: - device_queue_manager_init_vi_tonga(&dqm->asic_ops); + device_queue_manager_init_vi(&dqm->asic_ops); break; default: if (KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0)) device_queue_manager_init_v11(&dqm->asic_ops); else if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1)) - device_queue_manager_init_v10_navi10(&dqm->asic_ops); + device_queue_manager_init_v10(&dqm->asic_ops); else if (KFD_GC_VERSION(dev) >= IP_VERSION(9, 0, 1)) device_queue_manager_init_v9(&dqm->asic_ops); else { @@ -2797,19 +2816,11 @@ static void copy_context_work_handler (struct work_struct *work) static uint32_t *get_queue_ids(uint32_t num_queues, uint32_t *usr_queue_id_array) { size_t array_size = num_queues * sizeof(uint32_t); - uint32_t *queue_ids = NULL; if (!usr_queue_id_array) return NULL; - queue_ids = kzalloc(array_size, GFP_KERNEL); - if (!queue_ids) - return ERR_PTR(-ENOMEM); - - if (copy_from_user(queue_ids, usr_queue_id_array, array_size)) - return ERR_PTR(-EFAULT); - - return queue_ids; + return memdup_user(usr_queue_id_array, array_size); } int resume_queues(struct kfd_process *p, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index 7dd4b177219d..cf7e182588f8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -269,15 +269,11 @@ struct device_queue_manager { void device_queue_manager_init_cik( struct device_queue_manager_asic_ops *asic_ops); -void device_queue_manager_init_cik_hawaii( - struct device_queue_manager_asic_ops *asic_ops); void device_queue_manager_init_vi( struct device_queue_manager_asic_ops *asic_ops); -void device_queue_manager_init_vi_tonga( - struct device_queue_manager_asic_ops *asic_ops); void device_queue_manager_init_v9( struct device_queue_manager_asic_ops *asic_ops); -void device_queue_manager_init_v10_navi10( +void device_queue_manager_init_v10( struct device_queue_manager_asic_ops *asic_ops); void device_queue_manager_init_v11( struct device_queue_manager_asic_ops *asic_ops); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c index b1ab5b0775e1..d4d95c7f2e5d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c @@ -34,17 +34,13 @@ static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm, void __user *alternate_aperture_base, uint64_t alternate_aperture_size); static int update_qpd_cik(struct device_queue_manager *dqm, - struct qcm_process_device *qpd); -static int update_qpd_cik_hawaii(struct device_queue_manager *dqm, - struct qcm_process_device *qpd); -static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd); -static void init_sdma_vm_hawaii(struct device_queue_manager *dqm, - struct queue *q, - struct qcm_process_device *qpd); + struct qcm_process_device *qpd); +static void init_sdma_vm(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd); void device_queue_manager_init_cik( - struct device_queue_manager_asic_ops *asic_ops) + struct device_queue_manager_asic_ops *asic_ops) { asic_ops->set_cache_memory_policy = set_cache_memory_policy_cik; asic_ops->update_qpd = update_qpd_cik; @@ -52,15 +48,6 @@ void device_queue_manager_init_cik( asic_ops->mqd_manager_init = mqd_manager_init_cik; } -void device_queue_manager_init_cik_hawaii( - struct device_queue_manager_asic_ops *asic_ops) -{ - asic_ops->set_cache_memory_policy = set_cache_memory_policy_cik; - asic_ops->update_qpd = update_qpd_cik_hawaii; - asic_ops->init_sdma_vm = init_sdma_vm_hawaii; - asic_ops->mqd_manager_init = mqd_manager_init_cik_hawaii; -} - static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble) { /* In 64-bit mode, we can only control the top 3 bits of the LDS, @@ -115,41 +102,7 @@ static bool set_cache_memory_policy_cik(struct device_queue_manager *dqm, } static int update_qpd_cik(struct device_queue_manager *dqm, - struct qcm_process_device *qpd) -{ - struct kfd_process_device *pdd; - unsigned int temp; - - pdd = qpd_to_pdd(qpd); - - /* check if sh_mem_config register already configured */ - if (qpd->sh_mem_config == 0) { - qpd->sh_mem_config = - ALIGNMENT_MODE(SH_MEM_ALIGNMENT_MODE_UNALIGNED) | - DEFAULT_MTYPE(MTYPE_NONCACHED) | - APE1_MTYPE(MTYPE_NONCACHED); - qpd->sh_mem_ape1_limit = 0; - qpd->sh_mem_ape1_base = 0; - } - - if (qpd->pqm->process->is_32bit_user_mode) { - temp = get_sh_mem_bases_32(pdd); - qpd->sh_mem_bases = SHARED_BASE(temp); - qpd->sh_mem_config |= PTR32; - } else { - temp = get_sh_mem_bases_nybble_64(pdd); - qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp); - qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__PRIVATE_ATC__SHIFT; - } - - pr_debug("is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n", - qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases); - - return 0; -} - -static int update_qpd_cik_hawaii(struct device_queue_manager *dqm, - struct qcm_process_device *qpd) + struct qcm_process_device *qpd) { struct kfd_process_device *pdd; unsigned int temp; @@ -178,25 +131,9 @@ static int update_qpd_cik_hawaii(struct device_queue_manager *dqm, return 0; } -static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd) -{ - uint32_t value = (1 << SDMA0_RLC0_VIRTUAL_ADDR__ATC__SHIFT); - - if (q->process->is_32bit_user_mode) - value |= (1 << SDMA0_RLC0_VIRTUAL_ADDR__PTR32__SHIFT) | - get_sh_mem_bases_32(qpd_to_pdd(qpd)); - else - value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) << - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK; - - q->properties.sdma_vm_addr = value; -} - -static void init_sdma_vm_hawaii(struct device_queue_manager *dqm, - struct queue *q, - struct qcm_process_device *qpd) +static void init_sdma_vm(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd) { /* On dGPU we're always in GPUVM64 addressing mode with 64-bit * aperture addresses. diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c index f1a1f5753e65..245a90dfc2f6 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v10.c @@ -32,7 +32,7 @@ static int update_qpd_v10(struct device_queue_manager *dqm, static void init_sdma_vm_v10(struct device_queue_manager *dqm, struct queue *q, struct qcm_process_device *qpd); -void device_queue_manager_init_v10_navi10( +void device_queue_manager_init_v10( struct device_queue_manager_asic_ops *asic_ops) { asic_ops->update_qpd = update_qpd_v10; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c index 8af643388768..54eb1bff903c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c @@ -60,7 +60,7 @@ static int update_qpd_v9(struct device_queue_manager *dqm, qpd->sh_mem_config = SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT; - if (dqm->dev->kfd->noretry && !dqm->dev->kfd->use_iommu_v2) + if (dqm->dev->kfd->noretry) qpd->sh_mem_config |= 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT; if (KFD_GC_VERSION(dqm->dev->kfd) == IP_VERSION(9, 4, 3)) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c index d7d45832df0f..b291ee0fab94 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c @@ -28,29 +28,19 @@ #include "oss/oss_3_0_sh_mask.h" static bool set_cache_memory_policy_vi(struct device_queue_manager *dqm, - struct qcm_process_device *qpd, - enum cache_policy default_policy, - enum cache_policy alternate_policy, - void __user *alternate_aperture_base, - uint64_t alternate_aperture_size); -static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm, - struct qcm_process_device *qpd, - enum cache_policy default_policy, - enum cache_policy alternate_policy, - void __user *alternate_aperture_base, - uint64_t alternate_aperture_size); + struct qcm_process_device *qpd, + enum cache_policy default_policy, + enum cache_policy alternate_policy, + void __user *alternate_aperture_base, + uint64_t alternate_aperture_size); static int update_qpd_vi(struct device_queue_manager *dqm, - struct qcm_process_device *qpd); -static int update_qpd_vi_tonga(struct device_queue_manager *dqm, - struct qcm_process_device *qpd); -static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd); -static void init_sdma_vm_tonga(struct device_queue_manager *dqm, - struct queue *q, - struct qcm_process_device *qpd); + struct qcm_process_device *qpd); +static void init_sdma_vm(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd); void device_queue_manager_init_vi( - struct device_queue_manager_asic_ops *asic_ops) + struct device_queue_manager_asic_ops *asic_ops) { asic_ops->set_cache_memory_policy = set_cache_memory_policy_vi; asic_ops->update_qpd = update_qpd_vi; @@ -58,15 +48,6 @@ void device_queue_manager_init_vi( asic_ops->mqd_manager_init = mqd_manager_init_vi; } -void device_queue_manager_init_vi_tonga( - struct device_queue_manager_asic_ops *asic_ops) -{ - asic_ops->set_cache_memory_policy = set_cache_memory_policy_vi_tonga; - asic_ops->update_qpd = update_qpd_vi_tonga; - asic_ops->init_sdma_vm = init_sdma_vm_tonga; - asic_ops->mqd_manager_init = mqd_manager_init_vi_tonga; -} - static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble) { /* In 64-bit mode, we can only control the top 3 bits of the LDS, @@ -96,35 +77,6 @@ static uint32_t compute_sh_mem_bases_64bit(unsigned int top_address_nybble) } static bool set_cache_memory_policy_vi(struct device_queue_manager *dqm, - struct qcm_process_device *qpd, - enum cache_policy default_policy, - enum cache_policy alternate_policy, - void __user *alternate_aperture_base, - uint64_t alternate_aperture_size) -{ - uint32_t default_mtype; - uint32_t ape1_mtype; - - default_mtype = (default_policy == cache_policy_coherent) ? - MTYPE_CC : - MTYPE_NC; - - ape1_mtype = (alternate_policy == cache_policy_coherent) ? - MTYPE_CC : - MTYPE_NC; - - qpd->sh_mem_config = (qpd->sh_mem_config & - SH_MEM_CONFIG__ADDRESS_MODE_MASK) | - SH_MEM_ALIGNMENT_MODE_UNALIGNED << - SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT | - default_mtype << SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT | - ape1_mtype << SH_MEM_CONFIG__APE1_MTYPE__SHIFT | - SH_MEM_CONFIG__PRIVATE_ATC_MASK; - - return true; -} - -static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm, struct qcm_process_device *qpd, enum cache_policy default_policy, enum cache_policy alternate_policy, @@ -152,48 +104,7 @@ static bool set_cache_memory_policy_vi_tonga(struct device_queue_manager *dqm, } static int update_qpd_vi(struct device_queue_manager *dqm, - struct qcm_process_device *qpd) -{ - struct kfd_process_device *pdd; - unsigned int temp; - - pdd = qpd_to_pdd(qpd); - - /* check if sh_mem_config register already configured */ - if (qpd->sh_mem_config == 0) { - qpd->sh_mem_config = - SH_MEM_ALIGNMENT_MODE_UNALIGNED << - SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT | - MTYPE_CC << SH_MEM_CONFIG__DEFAULT_MTYPE__SHIFT | - MTYPE_CC << SH_MEM_CONFIG__APE1_MTYPE__SHIFT | - SH_MEM_CONFIG__PRIVATE_ATC_MASK; - - qpd->sh_mem_ape1_limit = 0; - qpd->sh_mem_ape1_base = 0; - } - - if (qpd->pqm->process->is_32bit_user_mode) { - temp = get_sh_mem_bases_32(pdd); - qpd->sh_mem_bases = temp << SH_MEM_BASES__SHARED_BASE__SHIFT; - qpd->sh_mem_config |= SH_MEM_ADDRESS_MODE_HSA32 << - SH_MEM_CONFIG__ADDRESS_MODE__SHIFT; - } else { - temp = get_sh_mem_bases_nybble_64(pdd); - qpd->sh_mem_bases = compute_sh_mem_bases_64bit(temp); - qpd->sh_mem_config |= SH_MEM_ADDRESS_MODE_HSA64 << - SH_MEM_CONFIG__ADDRESS_MODE__SHIFT; - qpd->sh_mem_config |= 1 << - SH_MEM_CONFIG__PRIVATE_ATC__SHIFT; - } - - pr_debug("is32bit process: %d sh_mem_bases nybble: 0x%X and register 0x%X\n", - qpd->pqm->process->is_32bit_user_mode, temp, qpd->sh_mem_bases); - - return 0; -} - -static int update_qpd_vi_tonga(struct device_queue_manager *dqm, - struct qcm_process_device *qpd) + struct qcm_process_device *qpd) { struct kfd_process_device *pdd; unsigned int temp; @@ -226,25 +137,9 @@ static int update_qpd_vi_tonga(struct device_queue_manager *dqm, return 0; } -static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd) -{ - uint32_t value = (1 << SDMA0_RLC0_VIRTUAL_ADDR__ATC__SHIFT); - - if (q->process->is_32bit_user_mode) - value |= (1 << SDMA0_RLC0_VIRTUAL_ADDR__PTR32__SHIFT) | - get_sh_mem_bases_32(qpd_to_pdd(qpd)); - else - value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) << - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK; - - q->properties.sdma_vm_addr = value; -} - -static void init_sdma_vm_tonga(struct device_queue_manager *dqm, - struct queue *q, - struct qcm_process_device *qpd) +static void init_sdma_vm(struct device_queue_manager *dqm, + struct queue *q, + struct qcm_process_device *qpd) { /* On dGPU we're always in GPUVM64 addressing mode with 64-bit * aperture addresses. diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c index 6421b620388d..c2e0b79dcc6d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c @@ -61,81 +61,46 @@ size_t kfd_doorbell_process_slice(struct kfd_dev *kfd) /* Doorbell calculations for device init. */ int kfd_doorbell_init(struct kfd_dev *kfd) { - size_t doorbell_start_offset; - size_t doorbell_aperture_size; - size_t doorbell_process_limit; + int size = PAGE_SIZE; + int r; /* - * With MES enabled, just set the doorbell base as it is needed - * to calculate doorbell physical address. - */ - if (kfd->shared_resources.enable_mes) { - kfd->doorbell_base = - kfd->shared_resources.doorbell_physical_address; - return 0; - } - - /* - * We start with calculations in bytes because the input data might - * only be byte-aligned. - * Only after we have done the rounding can we assume any alignment. + * Todo: KFD kernel level operations need only one doorbell for + * ring test/HWS. So instead of reserving a whole page here for + * kernel, reserve and consume a doorbell from existing KGD kernel + * doorbell page. */ - doorbell_start_offset = - roundup(kfd->shared_resources.doorbell_start_offset, - kfd_doorbell_process_slice(kfd)); - - doorbell_aperture_size = - rounddown(kfd->shared_resources.doorbell_aperture_size, - kfd_doorbell_process_slice(kfd)); - - if (doorbell_aperture_size > doorbell_start_offset) - doorbell_process_limit = - (doorbell_aperture_size - doorbell_start_offset) / - kfd_doorbell_process_slice(kfd); - else - return -ENOSPC; - - if (!kfd->max_doorbell_slices || - doorbell_process_limit < kfd->max_doorbell_slices) - kfd->max_doorbell_slices = doorbell_process_limit; - - kfd->doorbell_base = kfd->shared_resources.doorbell_physical_address + - doorbell_start_offset; - - kfd->doorbell_base_dw_offset = doorbell_start_offset / sizeof(u32); - - kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base, - kfd_doorbell_process_slice(kfd)); - - if (!kfd->doorbell_kernel_ptr) + /* Bitmap to dynamically allocate doorbells from kernel page */ + kfd->doorbell_bitmap = bitmap_zalloc(size / sizeof(u32), GFP_KERNEL); + if (!kfd->doorbell_bitmap) { + DRM_ERROR("Failed to allocate kernel doorbell bitmap\n"); return -ENOMEM; + } - pr_debug("Doorbell initialization:\n"); - pr_debug("doorbell base == 0x%08lX\n", - (uintptr_t)kfd->doorbell_base); - - pr_debug("doorbell_base_dw_offset == 0x%08lX\n", - kfd->doorbell_base_dw_offset); - - pr_debug("doorbell_process_limit == 0x%08lX\n", - doorbell_process_limit); - - pr_debug("doorbell_kernel_offset == 0x%08lX\n", - (uintptr_t)kfd->doorbell_base); - - pr_debug("doorbell aperture size == 0x%08lX\n", - kfd->shared_resources.doorbell_aperture_size); - - pr_debug("doorbell kernel address == %p\n", kfd->doorbell_kernel_ptr); + /* Alloc a doorbell page for KFD kernel usages */ + r = amdgpu_bo_create_kernel(kfd->adev, + size, + PAGE_SIZE, + AMDGPU_GEM_DOMAIN_DOORBELL, + &kfd->doorbells, + NULL, + (void **)&kfd->doorbell_kernel_ptr); + if (r) { + pr_err("failed to allocate kernel doorbells\n"); + bitmap_free(kfd->doorbell_bitmap); + return r; + } + pr_debug("Doorbell kernel address == %p\n", kfd->doorbell_kernel_ptr); return 0; } void kfd_doorbell_fini(struct kfd_dev *kfd) { - if (kfd->doorbell_kernel_ptr) - iounmap(kfd->doorbell_kernel_ptr); + bitmap_free(kfd->doorbell_bitmap); + amdgpu_bo_free_kernel(&kfd->doorbells, NULL, + (void **)&kfd->doorbell_kernel_ptr); } int kfd_doorbell_mmap(struct kfd_node *dev, struct kfd_process *process, @@ -188,22 +153,15 @@ void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, u32 inx; mutex_lock(&kfd->doorbell_mutex); - inx = find_first_zero_bit(kfd->doorbell_available_index, - KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); + inx = find_first_zero_bit(kfd->doorbell_bitmap, PAGE_SIZE / sizeof(u32)); - __set_bit(inx, kfd->doorbell_available_index); + __set_bit(inx, kfd->doorbell_bitmap); mutex_unlock(&kfd->doorbell_mutex); if (inx >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) return NULL; - inx *= kfd->device_info.doorbell_size / sizeof(u32); - - /* - * Calculating the kernel doorbell offset using the first - * doorbell page. - */ - *doorbell_off = kfd->doorbell_base_dw_offset + inx; + *doorbell_off = amdgpu_doorbell_index_on_bar(kfd->adev, kfd->doorbells, inx); pr_debug("Get kernel queue doorbell\n" " doorbell offset == 0x%08X\n" @@ -217,11 +175,10 @@ void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr) { unsigned int inx; - inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr) - * sizeof(u32) / kfd->device_info.doorbell_size; + inx = (unsigned int)(db_addr - kfd->doorbell_kernel_ptr); mutex_lock(&kfd->doorbell_mutex); - __clear_bit(inx, kfd->doorbell_available_index); + __clear_bit(inx, kfd->doorbell_bitmap); mutex_unlock(&kfd->doorbell_mutex); } @@ -243,80 +200,96 @@ void write_kernel_doorbell64(void __iomem *db, u64 value) } } -unsigned int kfd_get_doorbell_dw_offset_in_bar(struct kfd_dev *kfd, - struct kfd_process_device *pdd, - unsigned int doorbell_id) +static int init_doorbell_bitmap(struct qcm_process_device *qpd, + struct kfd_dev *dev) { - /* - * doorbell_base_dw_offset accounts for doorbells taken by KGD. - * index * kfd_doorbell_process_slice/sizeof(u32) adjusts to - * the process's doorbells. The offset returned is in dword - * units regardless of the ASIC-dependent doorbell size. - */ - if (!kfd->shared_resources.enable_mes) - return kfd->doorbell_base_dw_offset + - pdd->doorbell_index - * kfd_doorbell_process_slice(kfd) / sizeof(u32) + - doorbell_id * - kfd->device_info.doorbell_size / sizeof(u32); - else - return amdgpu_mes_get_doorbell_dw_offset_in_bar( - (struct amdgpu_device *)kfd->adev, - pdd->doorbell_index, doorbell_id); -} + unsigned int i; + int range_start = dev->shared_resources.non_cp_doorbells_start; + int range_end = dev->shared_resources.non_cp_doorbells_end; -uint64_t kfd_get_number_elems(struct kfd_dev *kfd) -{ - uint64_t num_of_elems = (kfd->shared_resources.doorbell_aperture_size - - kfd->shared_resources.doorbell_start_offset) / - kfd_doorbell_process_slice(kfd) + 1; + if (!KFD_IS_SOC15(dev)) + return 0; - return num_of_elems; + /* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */ + pr_debug("reserved doorbell 0x%03x - 0x%03x\n", range_start, range_end); + pr_debug("reserved doorbell 0x%03x - 0x%03x\n", + range_start + KFD_QUEUE_DOORBELL_MIRROR_OFFSET, + range_end + KFD_QUEUE_DOORBELL_MIRROR_OFFSET); + + for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) { + if (i >= range_start && i <= range_end) { + __set_bit(i, qpd->doorbell_bitmap); + __set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET, + qpd->doorbell_bitmap); + } + } + return 0; } phys_addr_t kfd_get_process_doorbells(struct kfd_process_device *pdd) { - if (!pdd->doorbell_index) { - int r = kfd_alloc_process_doorbells(pdd->dev->kfd, - &pdd->doorbell_index); - if (r < 0) + struct amdgpu_device *adev = pdd->dev->adev; + uint32_t first_db_index; + + if (!pdd->qpd.proc_doorbells) { + if (kfd_alloc_process_doorbells(pdd->dev->kfd, pdd)) + /* phys_addr_t 0 is error */ return 0; } - return pdd->dev->kfd->doorbell_base + - pdd->doorbell_index * kfd_doorbell_process_slice(pdd->dev->kfd); + first_db_index = amdgpu_doorbell_index_on_bar(adev, pdd->qpd.proc_doorbells, 0); + return adev->doorbell.base + first_db_index * sizeof(uint32_t); } -int kfd_alloc_process_doorbells(struct kfd_dev *kfd, unsigned int *doorbell_index) +int kfd_alloc_process_doorbells(struct kfd_dev *kfd, struct kfd_process_device *pdd) { - int r = 0; - - if (!kfd->shared_resources.enable_mes) - r = ida_simple_get(&kfd->doorbell_ida, 1, - kfd->max_doorbell_slices, GFP_KERNEL); - else - r = amdgpu_mes_alloc_process_doorbells( - (struct amdgpu_device *)kfd->adev, - doorbell_index); + int r; + struct qcm_process_device *qpd = &pdd->qpd; + + /* Allocate bitmap for dynamic doorbell allocation */ + qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, + GFP_KERNEL); + if (!qpd->doorbell_bitmap) { + DRM_ERROR("Failed to allocate process doorbell bitmap\n"); + return -ENOMEM; + } - if (r > 0) - *doorbell_index = r; + r = init_doorbell_bitmap(&pdd->qpd, kfd); + if (r) { + DRM_ERROR("Failed to initialize process doorbells\n"); + r = -ENOMEM; + goto err; + } - if (r < 0) - pr_err("Failed to allocate process doorbells\n"); + /* Allocate doorbells for this process */ + r = amdgpu_bo_create_kernel(kfd->adev, + kfd_doorbell_process_slice(kfd), + PAGE_SIZE, + AMDGPU_GEM_DOMAIN_DOORBELL, + &qpd->proc_doorbells, + NULL, + NULL); + if (r) { + DRM_ERROR("Failed to allocate process doorbells\n"); + goto err; + } + return 0; +err: + bitmap_free(qpd->doorbell_bitmap); + qpd->doorbell_bitmap = NULL; return r; } -void kfd_free_process_doorbells(struct kfd_dev *kfd, unsigned int doorbell_index) +void kfd_free_process_doorbells(struct kfd_dev *kfd, struct kfd_process_device *pdd) { - if (doorbell_index) { - if (!kfd->shared_resources.enable_mes) - ida_simple_remove(&kfd->doorbell_ida, doorbell_index); - else - amdgpu_mes_free_process_doorbells( - (struct amdgpu_device *)kfd->adev, - doorbell_index); + struct qcm_process_device *qpd = &pdd->qpd; + + if (qpd->doorbell_bitmap) { + bitmap_free(qpd->doorbell_bitmap); + qpd->doorbell_bitmap = NULL; } + + amdgpu_bo_free_kernel(&qpd->proc_doorbells, NULL, NULL); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c index 8081a9408006..0f58be65132f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -31,7 +31,6 @@ #include <linux/memory.h> #include "kfd_priv.h" #include "kfd_events.h" -#include "kfd_iommu.h" #include <linux/device.h> /* @@ -1146,87 +1145,6 @@ static void lookup_events_by_type_and_signal(struct kfd_process *p, rcu_read_unlock(); } -#ifdef KFD_SUPPORT_IOMMU_V2 -void kfd_signal_iommu_event(struct kfd_node *dev, u32 pasid, - unsigned long address, bool is_write_requested, - bool is_execute_requested) -{ - struct kfd_hsa_memory_exception_data memory_exception_data; - struct vm_area_struct *vma; - int user_gpu_id; - - /* - * Because we are called from arbitrary context (workqueue) as opposed - * to process context, kfd_process could attempt to exit while we are - * running so the lookup function increments the process ref count. - */ - struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); - struct mm_struct *mm; - - if (!p) - return; /* Presumably process exited. */ - - /* Take a safe reference to the mm_struct, which may otherwise - * disappear even while the kfd_process is still referenced. - */ - mm = get_task_mm(p->lead_thread); - if (!mm) { - kfd_unref_process(p); - return; /* Process is exiting */ - } - - user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id); - if (unlikely(user_gpu_id == -EINVAL)) { - WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id); - return; - } - memset(&memory_exception_data, 0, sizeof(memory_exception_data)); - - mmap_read_lock(mm); - vma = find_vma(mm, address); - - memory_exception_data.gpu_id = user_gpu_id; - memory_exception_data.va = address; - /* Set failure reason */ - memory_exception_data.failure.NotPresent = 1; - memory_exception_data.failure.NoExecute = 0; - memory_exception_data.failure.ReadOnly = 0; - if (vma && address >= vma->vm_start) { - memory_exception_data.failure.NotPresent = 0; - - if (is_write_requested && !(vma->vm_flags & VM_WRITE)) - memory_exception_data.failure.ReadOnly = 1; - else - memory_exception_data.failure.ReadOnly = 0; - - if (is_execute_requested && !(vma->vm_flags & VM_EXEC)) - memory_exception_data.failure.NoExecute = 1; - else - memory_exception_data.failure.NoExecute = 0; - } - - mmap_read_unlock(mm); - mmput(mm); - - pr_debug("notpresent %d, noexecute %d, readonly %d\n", - memory_exception_data.failure.NotPresent, - memory_exception_data.failure.NoExecute, - memory_exception_data.failure.ReadOnly); - - /* Workaround on Raven to not kill the process when memory is freed - * before IOMMU is able to finish processing all the excessive PPRs - */ - - if (KFD_GC_VERSION(dev) != IP_VERSION(9, 1, 0) && - KFD_GC_VERSION(dev) != IP_VERSION(9, 2, 2) && - KFD_GC_VERSION(dev) != IP_VERSION(9, 3, 0)) - lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_MEMORY, - &memory_exception_data); - - kfd_unref_process(p); -} -#endif /* KFD_SUPPORT_IOMMU_V2 */ - void kfd_signal_hw_exception_event(u32 pasid) { /* diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index da2ca00d79e5..62b205dac63a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -322,22 +322,19 @@ static void kfd_init_apertures_vi(struct kfd_process_device *pdd, uint8_t id) pdd->lds_base = MAKE_LDS_APP_BASE_VI(); pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); - if (!pdd->dev->kfd->use_iommu_v2) { - /* dGPUs: SVM aperture starting at 0 - * with small reserved space for kernel. - * Set them to CANONICAL addresses. - */ - pdd->gpuvm_base = SVM_USER_BASE; - pdd->gpuvm_limit = - pdd->dev->kfd->shared_resources.gpuvm_size - 1; - } else { - /* set them to non CANONICAL addresses, and no SVM is - * allocated. - */ - pdd->gpuvm_base = MAKE_GPUVM_APP_BASE_VI(id + 1); - pdd->gpuvm_limit = MAKE_GPUVM_APP_LIMIT(pdd->gpuvm_base, - pdd->dev->kfd->shared_resources.gpuvm_size); - } + /* dGPUs: SVM aperture starting at 0 + * with small reserved space for kernel. + * Set them to CANONICAL addresses. + */ + pdd->gpuvm_base = SVM_USER_BASE; + pdd->gpuvm_limit = + pdd->dev->kfd->shared_resources.gpuvm_size - 1; + + /* dGPUs: the reserved space for kernel + * before SVM + */ + pdd->qpd.cwsr_base = SVM_CWSR_BASE; + pdd->qpd.ib_base = SVM_IB_BASE; pdd->scratch_base = MAKE_SCRATCH_APP_BASE_VI(); pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); @@ -348,18 +345,18 @@ static void kfd_init_apertures_v9(struct kfd_process_device *pdd, uint8_t id) pdd->lds_base = MAKE_LDS_APP_BASE_V9(); pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); - /* Raven needs SVM to support graphic handle, etc. Leave the small - * reserved space before SVM on Raven as well, even though we don't - * have to. - * Set gpuvm_base and gpuvm_limit to CANONICAL addresses so that they - * are used in Thunk to reserve SVM. - */ - pdd->gpuvm_base = SVM_USER_BASE; + pdd->gpuvm_base = PAGE_SIZE; pdd->gpuvm_limit = pdd->dev->kfd->shared_resources.gpuvm_size - 1; pdd->scratch_base = MAKE_SCRATCH_APP_BASE_V9(); pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); + + /* + * Place TBA/TMA on opposite side of VM hole to prevent + * stray faults from triggering SVM on these pages. + */ + pdd->qpd.cwsr_base = pdd->dev->kfd->shared_resources.gpuvm_size; } int kfd_init_apertures(struct kfd_process *process) @@ -416,14 +413,6 @@ int kfd_init_apertures(struct kfd_process *process) return -EINVAL; } } - - if (!dev->kfd->use_iommu_v2) { - /* dGPUs: the reserved space for kernel - * before SVM - */ - pdd->qpd.cwsr_base = SVM_CWSR_BASE; - pdd->qpd.ib_base = SVM_IB_BASE; - } } dev_dbg(kfd_device, "node id %u\n", id); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c deleted file mode 100644 index 808ee010520a..000000000000 --- a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.c +++ /dev/null @@ -1,356 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 OR MIT -/* - * Copyright 2018-2022 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - */ - -#include <linux/kconfig.h> - -#if IS_REACHABLE(CONFIG_AMD_IOMMU_V2) - -#include <linux/printk.h> -#include <linux/device.h> -#include <linux/slab.h> -#include <linux/pci.h> -#include <linux/amd-iommu.h> -#include "kfd_priv.h" -#include "kfd_topology.h" -#include "kfd_iommu.h" - -static const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP | - AMD_IOMMU_DEVICE_FLAG_PRI_SUP | - AMD_IOMMU_DEVICE_FLAG_PASID_SUP; - -/** kfd_iommu_check_device - Check whether IOMMU is available for device - */ -int kfd_iommu_check_device(struct kfd_dev *kfd) -{ - struct amd_iommu_device_info iommu_info; - int err; - - if (!kfd->use_iommu_v2) - return -ENODEV; - - iommu_info.flags = 0; - err = amd_iommu_device_info(kfd->adev->pdev, &iommu_info); - if (err) - return err; - - if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) - return -ENODEV; - - return 0; -} - -/** kfd_iommu_device_init - Initialize IOMMU for device - */ -int kfd_iommu_device_init(struct kfd_dev *kfd) -{ - struct amd_iommu_device_info iommu_info; - unsigned int pasid_limit; - int err; - - if (!kfd->use_iommu_v2) - return 0; - - iommu_info.flags = 0; - err = amd_iommu_device_info(kfd->adev->pdev, &iommu_info); - if (err < 0) { - dev_err(kfd_device, - "error getting iommu info. is the iommu enabled?\n"); - return -ENODEV; - } - - if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) { - dev_err(kfd_device, - "error required iommu flags ats %i, pri %i, pasid %i\n", - (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0, - (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0, - (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP) - != 0); - return -ENODEV; - } - - pasid_limit = min_t(unsigned int, - (unsigned int)(1 << kfd->device_info.max_pasid_bits), - iommu_info.max_pasids); - - if (!kfd_set_pasid_limit(pasid_limit)) { - dev_err(kfd_device, "error setting pasid limit\n"); - return -EBUSY; - } - - return 0; -} - -/** kfd_iommu_bind_process_to_device - Have the IOMMU bind a process - * - * Binds the given process to the given device using its PASID. This - * enables IOMMUv2 address translation for the process on the device. - * - * This function assumes that the process mutex is held. - */ -int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd) -{ - struct kfd_node *dev = pdd->dev; - struct kfd_process *p = pdd->process; - int err; - - if (!dev->kfd->use_iommu_v2 || pdd->bound == PDD_BOUND) - return 0; - - if (unlikely(pdd->bound == PDD_BOUND_SUSPENDED)) { - pr_err("Binding PDD_BOUND_SUSPENDED pdd is unexpected!\n"); - return -EINVAL; - } - - if (!kfd_is_first_node(dev)) { - dev_warn_once(kfd_device, - "IOMMU supported only on first node\n"); - return 0; - } - - err = amd_iommu_bind_pasid(dev->adev->pdev, p->pasid, p->lead_thread); - if (!err) - pdd->bound = PDD_BOUND; - - return err; -} - -/** kfd_iommu_unbind_process - Unbind process from all devices - * - * This removes all IOMMU device bindings of the process. To be used - * before process termination. - */ -void kfd_iommu_unbind_process(struct kfd_process *p) -{ - int i; - - for (i = 0; i < p->n_pdds; i++) - if ((p->pdds[i]->bound == PDD_BOUND) && - (kfd_is_first_node((p->pdds[i]->dev)))) - amd_iommu_unbind_pasid(p->pdds[i]->dev->adev->pdev, - p->pasid); -} - -/* Callback for process shutdown invoked by the IOMMU driver */ -static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, u32 pasid) -{ - struct kfd_node *dev = kfd_device_by_pci_dev(pdev); - struct kfd_process *p; - struct kfd_process_device *pdd; - - if (!dev) - return; - - /* - * Look for the process that matches the pasid. If there is no such - * process, we either released it in amdkfd's own notifier, or there - * is a bug. Unfortunately, there is no way to tell... - */ - p = kfd_lookup_process_by_pasid(pasid); - if (!p) - return; - - pr_debug("Unbinding process 0x%x from IOMMU\n", pasid); - - mutex_lock(&p->mutex); - - pdd = kfd_get_process_device_data(dev, p); - if (pdd) - /* For GPU relying on IOMMU, we need to dequeue here - * when PASID is still bound. - */ - kfd_process_dequeue_from_device(pdd); - - mutex_unlock(&p->mutex); - - kfd_unref_process(p); -} - -/* This function called by IOMMU driver on PPR failure */ -static int iommu_invalid_ppr_cb(struct pci_dev *pdev, u32 pasid, - unsigned long address, u16 flags) -{ - struct kfd_node *dev; - - dev_warn_ratelimited(kfd_device, - "Invalid PPR device %x:%x.%x pasid 0x%x address 0x%lX flags 0x%X", - pdev->bus->number, - PCI_SLOT(pdev->devfn), - PCI_FUNC(pdev->devfn), - pasid, - address, - flags); - - dev = kfd_device_by_pci_dev(pdev); - if (!WARN_ON(!dev)) - kfd_signal_iommu_event(dev, pasid, address, - flags & PPR_FAULT_WRITE, flags & PPR_FAULT_EXEC); - - return AMD_IOMMU_INV_PRI_RSP_INVALID; -} - -/* - * Bind processes do the device that have been temporarily unbound - * (PDD_BOUND_SUSPENDED) in kfd_unbind_processes_from_device. - */ -static int kfd_bind_processes_to_device(struct kfd_node *knode) -{ - struct kfd_process_device *pdd; - struct kfd_process *p; - unsigned int temp; - int err = 0; - - int idx = srcu_read_lock(&kfd_processes_srcu); - - hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { - mutex_lock(&p->mutex); - pdd = kfd_get_process_device_data(knode, p); - - if (WARN_ON(!pdd) || pdd->bound != PDD_BOUND_SUSPENDED) { - mutex_unlock(&p->mutex); - continue; - } - - err = amd_iommu_bind_pasid(knode->adev->pdev, p->pasid, - p->lead_thread); - if (err < 0) { - pr_err("Unexpected pasid 0x%x binding failure\n", - p->pasid); - mutex_unlock(&p->mutex); - break; - } - - pdd->bound = PDD_BOUND; - mutex_unlock(&p->mutex); - } - - srcu_read_unlock(&kfd_processes_srcu, idx); - - return err; -} - -/* - * Mark currently bound processes as PDD_BOUND_SUSPENDED. These - * processes will be restored to PDD_BOUND state in - * kfd_bind_processes_to_device. - */ -static void kfd_unbind_processes_from_device(struct kfd_node *knode) -{ - struct kfd_process_device *pdd; - struct kfd_process *p; - unsigned int temp; - - int idx = srcu_read_lock(&kfd_processes_srcu); - - hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { - mutex_lock(&p->mutex); - pdd = kfd_get_process_device_data(knode, p); - - if (WARN_ON(!pdd)) { - mutex_unlock(&p->mutex); - continue; - } - - if (pdd->bound == PDD_BOUND) - pdd->bound = PDD_BOUND_SUSPENDED; - mutex_unlock(&p->mutex); - } - - srcu_read_unlock(&kfd_processes_srcu, idx); -} - -/** kfd_iommu_suspend - Prepare IOMMU for suspend - * - * This unbinds processes from the device and disables the IOMMU for - * the device. - */ -void kfd_iommu_suspend(struct kfd_dev *kfd) -{ - if (!kfd->use_iommu_v2) - return; - - kfd_unbind_processes_from_device(kfd->nodes[0]); - - amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev, NULL); - amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev, NULL); - amd_iommu_free_device(kfd->adev->pdev); -} - -/** kfd_iommu_resume - Restore IOMMU after resume - * - * This reinitializes the IOMMU for the device and re-binds previously - * suspended processes to the device. - */ -int kfd_iommu_resume(struct kfd_dev *kfd) -{ - unsigned int pasid_limit; - int err; - - if (!kfd->use_iommu_v2) - return 0; - - pasid_limit = kfd_get_pasid_limit(); - - err = amd_iommu_init_device(kfd->adev->pdev, pasid_limit); - if (err) - return -ENXIO; - - amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev, - iommu_pasid_shutdown_callback); - amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev, - iommu_invalid_ppr_cb); - - err = kfd_bind_processes_to_device(kfd->nodes[0]); - if (err) { - amd_iommu_set_invalidate_ctx_cb(kfd->adev->pdev, NULL); - amd_iommu_set_invalid_ppr_cb(kfd->adev->pdev, NULL); - amd_iommu_free_device(kfd->adev->pdev); - return err; - } - - return 0; -} - -/** kfd_iommu_add_perf_counters - Add IOMMU performance counters to topology - */ -int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev) -{ - struct kfd_perf_properties *props; - - if (!(kdev->node_props.capability & HSA_CAP_ATS_PRESENT)) - return 0; - - if (!amd_iommu_pc_supported()) - return 0; - - props = kfd_alloc_struct(props); - if (!props) - return -ENOMEM; - strcpy(props->block_name, "iommu"); - props->max_concurrent = amd_iommu_pc_get_max_banks(0) * - amd_iommu_pc_get_max_counters(0); /* assume one iommu */ - list_add_tail(&props->list, &kdev->perf_props); - - return 0; -} - -#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h b/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h deleted file mode 100644 index 8cf0fcbe87c2..000000000000 --- a/drivers/gpu/drm/amd/amdkfd/kfd_iommu.h +++ /dev/null @@ -1,84 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 OR MIT */ -/* - * Copyright 2018-2022 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef __KFD_IOMMU_H__ -#define __KFD_IOMMU_H__ - -#include <linux/kconfig.h> - -#if IS_REACHABLE(CONFIG_AMD_IOMMU_V2) - -#define KFD_SUPPORT_IOMMU_V2 - -int kfd_iommu_check_device(struct kfd_dev *kfd); -int kfd_iommu_device_init(struct kfd_dev *kfd); - -int kfd_iommu_bind_process_to_device(struct kfd_process_device *pdd); -void kfd_iommu_unbind_process(struct kfd_process *p); - -void kfd_iommu_suspend(struct kfd_dev *kfd); -int kfd_iommu_resume(struct kfd_dev *kfd); - -int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev); - -#else - -static inline int kfd_iommu_check_device(struct kfd_dev *kfd) -{ - return -ENODEV; -} -static inline int kfd_iommu_device_init(struct kfd_dev *kfd) -{ -#if IS_MODULE(CONFIG_AMD_IOMMU_V2) - WARN_ONCE(1, "iommu_v2 module is not usable by built-in KFD"); -#endif - return 0; -} - -static inline int kfd_iommu_bind_process_to_device( - struct kfd_process_device *pdd) -{ - return 0; -} -static inline void kfd_iommu_unbind_process(struct kfd_process *p) -{ - /* empty */ -} - -static inline void kfd_iommu_suspend(struct kfd_dev *kfd) -{ - /* empty */ -} -static inline int kfd_iommu_resume(struct kfd_dev *kfd) -{ - return 0; -} - -static inline int kfd_iommu_add_perf_counters(struct kfd_topology_device *kdev) -{ - return 0; -} - -#endif /* IS_REACHABLE(CONFIG_AMD_IOMMU_V2) */ - -#endif /* __KFD_IOMMU_H__ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c index 709ac885ca6d..7d82c7da223a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c @@ -461,7 +461,6 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange, 0, node->id, trigger); svm_range_dma_unmap(adev->dev, scratch, 0, npages); - svm_range_free_dma_mappings(prange); out_free: kvfree(buf); @@ -543,10 +542,12 @@ svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc, addr = next; } - if (cpages) + if (cpages) { prange->actual_loc = best_loc; - else + svm_range_free_dma_mappings(prange, true); + } else { svm_range_vram_node_free(prange); + } return r < 0 ? r : 0; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c index 863cf060af48..d01bb57733b3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -48,7 +48,7 @@ int pipe_priority_map[] = { struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_node *dev, struct queue_properties *q) { - struct kfd_mem_obj *mqd_mem_obj = NULL; + struct kfd_mem_obj *mqd_mem_obj; mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL); if (!mqd_mem_obj) @@ -64,7 +64,7 @@ struct kfd_mem_obj *allocate_hiq_mqd(struct kfd_node *dev, struct queue_properti struct kfd_mem_obj *allocate_sdma_mqd(struct kfd_node *dev, struct queue_properties *q) { - struct kfd_mem_obj *mqd_mem_obj = NULL; + struct kfd_mem_obj *mqd_mem_obj; uint64_t offset; mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 65c9f01a1f86..ee1d32d957f2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -206,13 +206,6 @@ static void __update_mqd(struct mqd_manager *mm, void *mqd, q->is_active = QUEUE_IS_ACTIVE(*q); } -static void update_mqd(struct mqd_manager *mm, void *mqd, - struct queue_properties *q, - struct mqd_update_info *minfo) -{ - __update_mqd(mm, mqd, q, minfo, 1); -} - static uint32_t read_doorbell_id(void *mqd) { struct cik_mqd *m = (struct cik_mqd *)mqd; @@ -220,9 +213,9 @@ static uint32_t read_doorbell_id(void *mqd) return m->queue_doorbell_id0; } -static void update_mqd_hawaii(struct mqd_manager *mm, void *mqd, - struct queue_properties *q, - struct mqd_update_info *minfo) +static void update_mqd(struct mqd_manager *mm, void *mqd, + struct queue_properties *q, + struct mqd_update_info *minfo) { __update_mqd(mm, mqd, q, minfo, 0); } @@ -387,7 +380,6 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data) #endif - struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, struct kfd_node *dev) { @@ -470,16 +462,3 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, return mqd; } - -struct mqd_manager *mqd_manager_init_cik_hawaii(enum KFD_MQD_TYPE type, - struct kfd_node *dev) -{ - struct mqd_manager *mqd; - - mqd = mqd_manager_init_cik(type, dev); - if (!mqd) - return NULL; - if (type == KFD_MQD_TYPE_CP) - mqd->update_mqd = update_mqd_hawaii; - return mqd; -} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c index 94c0fc2e57b7..83699392c808 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c @@ -318,6 +318,26 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd, 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; } +static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, unsigned int timeout, + uint32_t pipe_id, uint32_t queue_id) +{ + int err; + struct v10_compute_mqd *m; + u32 doorbell_off; + + m = get_mqd(mqd); + + doorbell_off = m->cp_hqd_pq_doorbell_control >> + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + + err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0); + if (err) + pr_debug("Destroy HIQ MQD failed: %d\n", err); + + return err; +} + static void init_mqd_sdma(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) @@ -460,7 +480,7 @@ struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, mqd->free_mqd = free_mqd_hiq_sdma; mqd->load_mqd = kfd_hiq_load_mqd_kiq; mqd->update_mqd = update_mqd; - mqd->destroy_mqd = kfd_destroy_mqd_cp; + mqd->destroy_mqd = destroy_hiq_mqd; mqd->is_occupied = kfd_is_occupied_cp; mqd->mqd_size = sizeof(struct v10_compute_mqd); mqd->mqd_stride = kfd_mqd_stride; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c index 31fec5e70d13..2319467d2d95 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v11.c @@ -335,6 +335,26 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd, 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; } +static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, unsigned int timeout, + uint32_t pipe_id, uint32_t queue_id) +{ + int err; + struct v11_compute_mqd *m; + u32 doorbell_off; + + m = get_mqd(mqd); + + doorbell_off = m->cp_hqd_pq_doorbell_control >> + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + + err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0); + if (err) + pr_debug("Destroy HIQ MQD failed: %d\n", err); + + return err; +} + static void init_mqd_sdma(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) @@ -449,7 +469,7 @@ struct mqd_manager *mqd_manager_init_v11(enum KFD_MQD_TYPE type, mqd->free_mqd = free_mqd_hiq_sdma; mqd->load_mqd = kfd_hiq_load_mqd_kiq; mqd->update_mqd = update_mqd; - mqd->destroy_mqd = kfd_destroy_mqd_cp; + mqd->destroy_mqd = destroy_hiq_mqd; mqd->is_occupied = kfd_is_occupied_cp; mqd->mqd_size = sizeof(struct v11_compute_mqd); #if defined(CONFIG_DEBUG_FS) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c index 601bb9f68048..e23d32f35607 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -405,6 +405,25 @@ static void init_mqd_hiq(struct mqd_manager *mm, void **mqd, 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; } +static int destroy_hiq_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, unsigned int timeout, + uint32_t pipe_id, uint32_t queue_id) +{ + int err; + struct v9_mqd *m; + u32 doorbell_off; + + m = get_mqd(mqd); + + doorbell_off = m->cp_hqd_pq_doorbell_control >> + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, 0); + if (err) + pr_debug("Destroy HIQ MQD failed: %d\n", err); + + return err; +} + static void init_mqd_sdma(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) @@ -548,16 +567,19 @@ static int destroy_hiq_mqd_v9_4_3(struct mqd_manager *mm, void *mqd, { uint32_t xcc_mask = mm->dev->xcc_mask; int xcc_id, err, inst = 0; - void *xcc_mqd; uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev); + struct v9_mqd *m; + u32 doorbell_off; for_each_inst(xcc_id, xcc_mask) { - xcc_mqd = mqd + hiq_mqd_size * inst; - err = mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, xcc_mqd, - type, timeout, pipe_id, - queue_id, xcc_id); + m = get_mqd(mqd + hiq_mqd_size * inst); + + doorbell_off = m->cp_hqd_pq_doorbell_control >> + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + + err = amdgpu_amdkfd_unmap_hiq(mm->dev->adev, doorbell_off, xcc_id); if (err) { - pr_debug("Destroy MQD failed for xcc: %d\n", inst); + pr_debug("Destroy HIQ MQD failed for xcc: %d\n", inst); break; } ++inst; @@ -846,7 +868,7 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, } else { mqd->init_mqd = init_mqd_hiq; mqd->load_mqd = kfd_hiq_load_mqd_kiq; - mqd->destroy_mqd = kfd_destroy_mqd_cp; + mqd->destroy_mqd = destroy_hiq_mqd; } break; case KFD_MQD_TYPE_DIQ: diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index d1e962da51dd..657c37822980 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -237,14 +237,6 @@ static void __update_mqd(struct mqd_manager *mm, void *mqd, q->is_active = QUEUE_IS_ACTIVE(*q); } - -static void update_mqd(struct mqd_manager *mm, void *mqd, - struct queue_properties *q, - struct mqd_update_info *minfo) -{ - __update_mqd(mm, mqd, q, minfo, MTYPE_CC, 1); -} - static uint32_t read_doorbell_id(void *mqd) { struct vi_mqd *m = (struct vi_mqd *)mqd; @@ -252,9 +244,9 @@ static uint32_t read_doorbell_id(void *mqd) return m->queue_doorbell_id0; } -static void update_mqd_tonga(struct mqd_manager *mm, void *mqd, - struct queue_properties *q, - struct mqd_update_info *minfo) +static void update_mqd(struct mqd_manager *mm, void *mqd, + struct queue_properties *q, + struct mqd_update_info *minfo) { __update_mqd(mm, mqd, q, minfo, MTYPE_UC, 0); } @@ -529,16 +521,3 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, return mqd; } - -struct mqd_manager *mqd_manager_init_vi_tonga(enum KFD_MQD_TYPE type, - struct kfd_node *dev) -{ - struct mqd_manager *mqd; - - mqd = mqd_manager_init_vi(type, dev); - if (!mqd) - return NULL; - if (type == KFD_MQD_TYPE_CP) - mqd->update_mqd = update_mqd_tonga; - return mqd; -} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c index 29a2d0499b67..8ce6f5200905 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c @@ -121,6 +121,7 @@ static int pm_map_process_aldebaran(struct packet_manager *pm, packet->sh_mem_bases = qpd->sh_mem_bases; if (qpd->tba_addr) { packet->sq_shader_tba_lo = lower_32_bits(qpd->tba_addr >> 8); + packet->sq_shader_tba_hi = upper_32_bits(qpd->tba_addr >> 8); packet->sq_shader_tma_lo = lower_32_bits(qpd->tma_addr >> 8); packet->sq_shader_tma_hi = upper_32_bits(qpd->tma_addr >> 8); } @@ -298,7 +299,8 @@ static int pm_set_grace_period_v9(struct packet_manager *pm, pm->dqm->wait_times, grace_period, ®_offset, - ®_data); + ®_data, + 0); if (grace_period == USE_DEFAULT_GRACE_PERIOD) reg_data = pm->dqm->wait_times; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index d4c9ee3f9953..3d9ce44d88da 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -175,12 +175,6 @@ extern int send_sigterm; */ extern int debug_largebar; -/* - * Ignore CRAT table during KFD initialization, can be used to work around - * broken CRAT tables on some AMD systems - */ -extern int ignore_crat; - /* Set sh_mem_config.retry_disable on GFX v9 */ extern int amdgpu_noretry; @@ -234,7 +228,6 @@ struct kfd_device_info { uint8_t num_of_watch_points; uint16_t mqd_size_aligned; bool supports_cwsr; - bool needs_iommu_device; bool needs_pci_atomics; uint32_t no_atomic_fw_version; unsigned int num_sdma_queues_per_engine; @@ -323,15 +316,6 @@ struct kfd_dev { struct kfd_device_info device_info; - phys_addr_t doorbell_base; /* Start of actual doorbells used by - * KFD. It is aligned for mapping - * into user mode - */ - size_t doorbell_base_dw_offset; /* Offset from the start of the PCI - * doorbell BAR to the first KFD - * doorbell in dwords. GFX reserves - * the segment before this offset. - */ u32 __iomem *doorbell_kernel_ptr; /* This is a pointer for a doorbells * page used by kernel queue */ @@ -340,8 +324,6 @@ struct kfd_dev { const struct kfd2kgd_calls *kfd2kgd; struct mutex doorbell_mutex; - DECLARE_BITMAP(doorbell_available_index, - KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); void *gtt_mem; uint64_t gtt_start_gpu_addr; @@ -368,9 +350,6 @@ struct kfd_dev { bool pci_atomic_requested; - /* Use IOMMU v2 flag */ - bool use_iommu_v2; - /* Compute Profile ref. count */ atomic_t compute_profile; @@ -385,6 +364,12 @@ struct kfd_dev { /* Track per device allocated watch points */ uint32_t alloc_watch_ids; spinlock_t watch_points_lock; + + /* Kernel doorbells for KFD device */ + struct amdgpu_bo *doorbells; + + /* bitmap for dynamic doorbell allocation from doorbell object */ + unsigned long *doorbell_bitmap; }; enum kfd_mempool { @@ -702,7 +687,10 @@ struct qcm_process_device { uint64_t ib_base; void *ib_kaddr; - /* doorbell resources per process per device */ + /* doorbells for kfd process */ + struct amdgpu_bo *proc_doorbells; + + /* bitmap for dynamic doorbell allocation from the bo */ unsigned long *doorbell_bitmap; }; @@ -792,7 +780,6 @@ struct kfd_process_device { struct attribute attr_evict; struct kobject *kobj_stats; - unsigned int doorbell_index; /* * @cu_occupancy: Reports occupancy of Compute Units (CU) of a process @@ -1100,9 +1087,9 @@ unsigned int kfd_get_doorbell_dw_offset_in_bar(struct kfd_dev *kfd, unsigned int doorbell_id); phys_addr_t kfd_get_process_doorbells(struct kfd_process_device *pdd); int kfd_alloc_process_doorbells(struct kfd_dev *kfd, - unsigned int *doorbell_index); + struct kfd_process_device *pdd); void kfd_free_process_doorbells(struct kfd_dev *kfd, - unsigned int doorbell_index); + struct kfd_process_device *pdd); /* GTT Sub-Allocator */ int kfd_gtt_sa_allocate(struct kfd_node *node, unsigned int size, @@ -1152,7 +1139,6 @@ static inline struct kfd_node *kfd_node_by_irq_ids(struct amdgpu_device *adev, } int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_node **kdev); int kfd_numa_node_to_apic_id(int numa_node_id); -void kfd_double_confirm_iommu_support(struct kfd_dev *gpu); /* Interrupts */ #define KFD_IRQ_FENCE_CLIENTID 0xff @@ -1299,12 +1285,8 @@ void print_queue(struct queue *q); struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, struct kfd_node *dev); -struct mqd_manager *mqd_manager_init_cik_hawaii(enum KFD_MQD_TYPE type, - struct kfd_node *dev); struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, struct kfd_node *dev); -struct mqd_manager *mqd_manager_init_vi_tonga(enum KFD_MQD_TYPE type, - struct kfd_node *dev); struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, struct kfd_node *dev); struct mqd_manager *mqd_manager_init_v10(enum KFD_MQD_TYPE type, @@ -1459,9 +1441,6 @@ int kfd_wait_on_events(struct kfd_process *p, uint32_t *wait_result); void kfd_signal_event_interrupt(u32 pasid, uint32_t partial_id, uint32_t valid_id_bits); -void kfd_signal_iommu_event(struct kfd_node *dev, - u32 pasid, unsigned long address, - bool is_write_requested, bool is_execute_requested); void kfd_signal_hw_exception_event(u32 pasid); int kfd_set_event(struct kfd_process *p, uint32_t event_id); int kfd_reset_event(struct kfd_process *p, uint32_t event_id); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index a844e68211ac..fbf053001af9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -28,7 +28,6 @@ #include <linux/sched/task.h> #include <linux/mmu_context.h> #include <linux/slab.h> -#include <linux/amd-iommu.h> #include <linux/notifier.h> #include <linux/compat.h> #include <linux/mman.h> @@ -41,7 +40,6 @@ struct mm_struct; #include "kfd_priv.h" #include "kfd_device_queue_manager.h" -#include "kfd_iommu.h" #include "kfd_svm.h" #include "kfd_smi_events.h" #include "kfd_debug.h" @@ -1035,10 +1033,9 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) free_pages((unsigned long)pdd->qpd.cwsr_kaddr, get_order(KFD_CWSR_TBA_TMA_SIZE)); - bitmap_free(pdd->qpd.doorbell_bitmap); idr_destroy(&pdd->alloc_idr); - kfd_free_process_doorbells(pdd->dev->kfd, pdd->doorbell_index); + kfd_free_process_doorbells(pdd->dev->kfd, pdd); if (pdd->dev->kfd->shared_resources.enable_mes) amdgpu_amdkfd_free_gtt_mem(pdd->dev->adev, @@ -1123,7 +1120,6 @@ static void kfd_process_wq_release(struct work_struct *work) dma_fence_signal(p->ef); kfd_process_remove_sysfs(p); - kfd_iommu_unbind_process(p); kfd_process_kunmap_signal_bo(p); kfd_process_free_outstanding_kfd_bos(p); @@ -1550,38 +1546,6 @@ err_alloc_process: return ERR_PTR(err); } -static int init_doorbell_bitmap(struct qcm_process_device *qpd, - struct kfd_dev *dev) -{ - unsigned int i; - int range_start = dev->shared_resources.non_cp_doorbells_start; - int range_end = dev->shared_resources.non_cp_doorbells_end; - - if (!KFD_IS_SOC15(dev)) - return 0; - - qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, - GFP_KERNEL); - if (!qpd->doorbell_bitmap) - return -ENOMEM; - - /* Mask out doorbells reserved for SDMA, IH, and VCN on SOC15. */ - pr_debug("reserved doorbell 0x%03x - 0x%03x\n", range_start, range_end); - pr_debug("reserved doorbell 0x%03x - 0x%03x\n", - range_start + KFD_QUEUE_DOORBELL_MIRROR_OFFSET, - range_end + KFD_QUEUE_DOORBELL_MIRROR_OFFSET); - - for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) { - if (i >= range_start && i <= range_end) { - __set_bit(i, qpd->doorbell_bitmap); - __set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET, - qpd->doorbell_bitmap); - } - } - - return 0; -} - struct kfd_process_device *kfd_get_process_device_data(struct kfd_node *dev, struct kfd_process *p) { @@ -1606,11 +1570,6 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev, if (!pdd) return NULL; - if (init_doorbell_bitmap(&pdd->qpd, dev->kfd)) { - pr_err("Failed to init doorbell for process\n"); - goto err_free_pdd; - } - pdd->dev = dev; INIT_LIST_HEAD(&pdd->qpd.queues_list); INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); @@ -1766,10 +1725,6 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_node *dev, } } - err = kfd_iommu_bind_process_to_device(pdd); - if (err) - goto out; - /* * make sure that runtime_usage counter is incremented just once * per pdd @@ -1777,15 +1732,6 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_node *dev, pdd->runtime_inuse = true; return pdd; - -out: - /* balance runpm reference count and exit with error */ - if (!pdd->runtime_inuse) { - pm_runtime_mark_last_busy(adev_to_drm(dev->adev)->dev); - pm_runtime_put_autosuspend(adev_to_drm(dev->adev)->dev); - } - - return ERR_PTR(err); } /* Create specific handle mapped to mem from process local memory idr diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index ba9d69054119..adb5e4bdc0b2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -123,7 +123,7 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid, if (!gws && pdd->qpd.num_gws == 0) return -EINVAL; - if (KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 3)) { + if (KFD_GC_VERSION(dev) != IP_VERSION(9, 4, 3) && !dev->kfd->shared_resources.enable_mes) { if (gws) ret = amdgpu_amdkfd_add_gws_to_process(pdd->process->kgd_process_info, gws, &mem); @@ -136,7 +136,9 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid, } else { /* * Intentionally set GWS to a non-NULL value - * for GFX 9.4.3. + * for devices that do not use GWS for global wave + * synchronization but require the formality + * of setting GWS for cooperative groups. */ pqn->q->gws = gws ? ERR_PTR(-ENOMEM) : NULL; } @@ -173,7 +175,8 @@ void pqm_uninit(struct process_queue_manager *pqm) list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) { if (pqn->q && pqn->q->gws && - KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 3)) + KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 3) && + !pqn->q->device->kfd->shared_resources.enable_mes) amdgpu_amdkfd_remove_gws_from_process(pqm->process->kgd_process_info, pqn->q->gws); kfd_procfs_del_queue(pqn->q); @@ -365,17 +368,20 @@ int pqm_create_queue(struct process_queue_manager *pqm, goto err_create_queue; } - if (q && p_doorbell_offset_in_process) + if (q && p_doorbell_offset_in_process) { /* Return the doorbell offset within the doorbell page * to the caller so it can be passed up to user mode * (in bytes). - * There are always 1024 doorbells per process, so in case - * of 8-byte doorbells, there are two doorbell pages per - * process. + * relative doorbell index = Absolute doorbell index - + * absolute index of first doorbell in the page. */ - *p_doorbell_offset_in_process = - (q->properties.doorbell_off * sizeof(uint32_t)) & - (kfd_doorbell_process_slice(dev->kfd) - 1); + uint32_t first_db_index = amdgpu_doorbell_index_on_bar(pdd->dev->adev, + pdd->qpd.proc_doorbells, + 0); + + *p_doorbell_offset_in_process = (q->properties.doorbell_off + - first_db_index) * sizeof(uint32_t); + } pr_debug("PQM After DQM create queue\n"); @@ -455,7 +461,8 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) } if (pqn->q->gws) { - if (KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 3)) + if (KFD_GC_VERSION(pqn->q->device) != IP_VERSION(9, 4, 3) && + !dev->kfd->shared_resources.enable_mes) amdgpu_amdkfd_remove_gws_from_process( pqm->process->kgd_process_info, pqn->q->gws); @@ -929,12 +936,6 @@ int kfd_criu_restore_queue(struct kfd_process *p, goto exit; } - if (!pdd->doorbell_index && - kfd_alloc_process_doorbells(pdd->dev->kfd, &pdd->doorbell_index) < 0) { - ret = -ENOMEM; - goto exit; - } - /* data stored in this order: mqd, ctl_stack */ mqd = q_extra_data; ctl_stack = mqd + q_data->mqd_size; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index 0b7bfbd0cb66..011561605983 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -23,7 +23,10 @@ #include <linux/types.h> #include <linux/sched/task.h> +#include <linux/dynamic_debug.h> #include <drm/ttm/ttm_tt.h> +#include <drm/drm_exec.h> + #include "amdgpu_sync.h" #include "amdgpu_object.h" #include "amdgpu_vm.h" @@ -46,6 +49,13 @@ * page table is updated. */ #define AMDGPU_SVM_RANGE_RETRY_FAULT_PENDING (2UL * NSEC_PER_MSEC) +#if IS_ENABLED(CONFIG_DYNAMIC_DEBUG) +#define dynamic_svm_range_dump(svms) \ + _dynamic_func_call_no_desc("svm_range_dump", svm_range_debug_dump, svms) +#else +#define dynamic_svm_range_dump(svms) \ + do { if (0) svm_range_debug_dump(svms); } while (0) +#endif /* Giant svm range split into smaller ranges based on this, it is decided using * minimum of all dGPU/APU 1/32 VRAM size, between 2MB to 1GB and alignment to @@ -239,7 +249,7 @@ void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr, } } -void svm_range_free_dma_mappings(struct svm_range *prange) +void svm_range_free_dma_mappings(struct svm_range *prange, bool unmap_dma) { struct kfd_process_device *pdd; dma_addr_t *dma_addr; @@ -260,13 +270,14 @@ void svm_range_free_dma_mappings(struct svm_range *prange) continue; } dev = &pdd->dev->adev->pdev->dev; - svm_range_dma_unmap(dev, dma_addr, 0, prange->npages); + if (unmap_dma) + svm_range_dma_unmap(dev, dma_addr, 0, prange->npages); kvfree(dma_addr); prange->dma_addr[gpuidx] = NULL; } } -static void svm_range_free(struct svm_range *prange, bool update_mem_usage) +static void svm_range_free(struct svm_range *prange, bool do_unmap) { uint64_t size = (prange->last - prange->start + 1) << PAGE_SHIFT; struct kfd_process *p = container_of(prange->svms, struct kfd_process, svms); @@ -275,9 +286,9 @@ static void svm_range_free(struct svm_range *prange, bool update_mem_usage) prange->start, prange->last); svm_range_vram_node_free(prange); - svm_range_free_dma_mappings(prange); + svm_range_free_dma_mappings(prange, do_unmap); - if (update_mem_usage && !p->xnack_enabled) { + if (do_unmap && !p->xnack_enabled) { pr_debug("unreserve prange 0x%p size: 0x%llx\n", prange, size); amdgpu_amdkfd_unreserve_mem_limit(NULL, size, KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0); @@ -849,6 +860,37 @@ static void svm_range_debug_dump(struct svm_range_list *svms) } } +static void * +svm_range_copy_array(void *psrc, size_t size, uint64_t num_elements, + uint64_t offset) +{ + unsigned char *dst; + + dst = kvmalloc_array(num_elements, size, GFP_KERNEL); + if (!dst) + return NULL; + memcpy(dst, (unsigned char *)psrc + offset, num_elements * size); + + return (void *)dst; +} + +static int +svm_range_copy_dma_addrs(struct svm_range *dst, struct svm_range *src) +{ + int i; + + for (i = 0; i < MAX_GPU_INSTANCE; i++) { + if (!src->dma_addr[i]) + continue; + dst->dma_addr[i] = svm_range_copy_array(src->dma_addr[i], + sizeof(*src->dma_addr[i]), src->npages, 0); + if (!dst->dma_addr[i]) + return -ENOMEM; + } + + return 0; +} + static int svm_range_split_array(void *ppnew, void *ppold, size_t size, uint64_t old_start, uint64_t old_n, @@ -863,22 +905,16 @@ svm_range_split_array(void *ppnew, void *ppold, size_t size, if (!pold) return 0; - new = kvmalloc_array(new_n, size, GFP_KERNEL); + d = (new_start - old_start) * size; + new = svm_range_copy_array(pold, size, new_n, d); if (!new) return -ENOMEM; - - d = (new_start - old_start) * size; - memcpy(new, pold + d, new_n * size); - - old = kvmalloc_array(old_n, size, GFP_KERNEL); + d = (new_start == old_start) ? new_n * size : 0; + old = svm_range_copy_array(pold, size, old_n, d); if (!old) { kvfree(new); return -ENOMEM; } - - d = (new_start == old_start) ? new_n * size : 0; - memcpy(old, pold + d, old_n * size); - kvfree(pold); *(void **)ppold = old; *(void **)ppnew = new; @@ -1455,37 +1491,34 @@ struct svm_validate_context { struct svm_range *prange; bool intr; DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE); - struct ttm_validate_buffer tv[MAX_GPU_INSTANCE]; - struct list_head validate_list; - struct ww_acquire_ctx ticket; + struct drm_exec exec; }; -static int svm_range_reserve_bos(struct svm_validate_context *ctx) +static int svm_range_reserve_bos(struct svm_validate_context *ctx, bool intr) { struct kfd_process_device *pdd; struct amdgpu_vm *vm; uint32_t gpuidx; int r; - INIT_LIST_HEAD(&ctx->validate_list); - for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) { - pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx); - if (!pdd) { - pr_debug("failed to find device idx %d\n", gpuidx); - return -EINVAL; - } - vm = drm_priv_to_vm(pdd->drm_priv); - - ctx->tv[gpuidx].bo = &vm->root.bo->tbo; - ctx->tv[gpuidx].num_shared = 4; - list_add(&ctx->tv[gpuidx].head, &ctx->validate_list); - } + drm_exec_init(&ctx->exec, intr ? DRM_EXEC_INTERRUPTIBLE_WAIT: 0); + drm_exec_until_all_locked(&ctx->exec) { + for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) { + pdd = kfd_process_device_from_gpuidx(ctx->process, gpuidx); + if (!pdd) { + pr_debug("failed to find device idx %d\n", gpuidx); + r = -EINVAL; + goto unreserve_out; + } + vm = drm_priv_to_vm(pdd->drm_priv); - r = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->validate_list, - ctx->intr, NULL); - if (r) { - pr_debug("failed %d to reserve bo\n", r); - return r; + r = amdgpu_vm_lock_pd(vm, &ctx->exec, 2); + drm_exec_retry_on_contention(&ctx->exec); + if (unlikely(r)) { + pr_debug("failed %d to reserve bo\n", r); + goto unreserve_out; + } + } } for_each_set_bit(gpuidx, ctx->bitmap, MAX_GPU_INSTANCE) { @@ -1508,13 +1541,13 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx) return 0; unreserve_out: - ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list); + drm_exec_fini(&ctx->exec); return r; } static void svm_range_unreserve_bos(struct svm_validate_context *ctx) { - ttm_eu_backoff_reservation(&ctx->ticket, &ctx->validate_list); + drm_exec_fini(&ctx->exec); } static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx) @@ -1522,6 +1555,8 @@ static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx) struct kfd_process_device *pdd; pdd = kfd_process_device_from_gpuidx(p, gpuidx); + if (!pdd) + return NULL; return SVM_ADEV_PGMAP_OWNER(pdd->dev->adev); } @@ -1596,12 +1631,12 @@ static int svm_range_validate_and_map(struct mm_struct *mm, } if (bitmap_empty(ctx->bitmap, MAX_GPU_INSTANCE)) { - if (!prange->mapped_to_gpu) { + bitmap_copy(ctx->bitmap, prange->bitmap_access, MAX_GPU_INSTANCE); + if (!prange->mapped_to_gpu || + bitmap_empty(ctx->bitmap, MAX_GPU_INSTANCE)) { r = 0; goto free_ctx; } - - bitmap_copy(ctx->bitmap, prange->bitmap_access, MAX_GPU_INSTANCE); } if (prange->actual_loc && !prange->ttm_res) { @@ -1613,7 +1648,7 @@ static int svm_range_validate_and_map(struct mm_struct *mm, goto free_ctx; } - svm_range_reserve_bos(ctx); + svm_range_reserve_bos(ctx, intr); p = container_of(prange->svms, struct kfd_process, svms); owner = kfd_svm_page_owner(p, find_first_bit(ctx->bitmap, @@ -1925,7 +1960,10 @@ static struct svm_range *svm_range_clone(struct svm_range *old) new = svm_range_new(old->svms, old->start, old->last, false); if (!new) return NULL; - + if (svm_range_copy_dma_addrs(new, old)) { + svm_range_free(new, false); + return NULL; + } if (old->svm_bo) { new->ttm_res = old->ttm_res; new->offset = old->offset; @@ -3558,7 +3596,7 @@ out_unlock_range: break; } - svm_range_debug_dump(svms); + dynamic_svm_range_dump(svms); mutex_unlock(&svms->lock); mmap_read_unlock(mm); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h index 21b14510882b..9e668eeefb32 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h @@ -183,7 +183,7 @@ void svm_range_add_list_work(struct svm_range_list *svms, void schedule_deferred_list_work(struct svm_range_list *svms); void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr, unsigned long offset, unsigned long npages); -void svm_range_free_dma_mappings(struct svm_range *prange); +void svm_range_free_dma_mappings(struct svm_range *prange, bool unmap_dma); int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, uint64_t *svm_priv_data_size); int kfd_criu_checkpoint_svm(struct kfd_process *p, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index 4a17bb7c7b27..ff98fded9534 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -36,8 +36,8 @@ #include "kfd_crat.h" #include "kfd_topology.h" #include "kfd_device_queue_manager.h" -#include "kfd_iommu.h" #include "kfd_svm.h" +#include "kfd_debug.h" #include "amdgpu_amdkfd.h" #include "amdgpu_ras.h" #include "amdgpu.h" @@ -988,17 +988,6 @@ static void find_system_memory(const struct dmi_header *dm, } } -/* - * Performance counters information is not part of CRAT but we would like to - * put them in the sysfs under topology directory for Thunk to get the data. - * This function is called before updating the sysfs. - */ -static int kfd_add_perf_to_topology(struct kfd_topology_device *kdev) -{ - /* These are the only counters supported so far */ - return kfd_iommu_add_perf_counters(kdev); -} - /* kfd_add_non_crat_information - Add information that is not currently * defined in CRAT but is necessary for KFD topology * @dev - topology device to which addition info is added @@ -1013,25 +1002,6 @@ static void kfd_add_non_crat_information(struct kfd_topology_device *kdev) /* TODO: For GPU node, rearrange code from kfd_topology_add_device */ } -/* kfd_is_acpi_crat_invalid - CRAT from ACPI is valid only for AMD APU devices. - * Ignore CRAT for all other devices. AMD APU is identified if both CPU - * and GPU cores are present. - * @device_list - topology device list created by parsing ACPI CRAT table. - * @return - TRUE if invalid, FALSE is valid. - */ -static bool kfd_is_acpi_crat_invalid(struct list_head *device_list) -{ - struct kfd_topology_device *dev; - - list_for_each_entry(dev, device_list, list) { - if (dev->node_props.cpu_cores_count && - dev->node_props.simd_count) - return false; - } - pr_info("Ignoring ACPI CRAT on non-APU system\n"); - return true; -} - int kfd_topology_init(void) { void *crat_image = NULL; @@ -1062,48 +1032,25 @@ int kfd_topology_init(void) */ proximity_domain = 0; - /* - * Get the CRAT image from the ACPI. If ACPI doesn't have one - * or if ACPI CRAT is invalid create a virtual CRAT. - * NOTE: The current implementation expects all AMD APUs to have - * CRAT. If no CRAT is available, it is assumed to be a CPU - */ - ret = kfd_create_crat_image_acpi(&crat_image, &image_size); - if (!ret) { - ret = kfd_parse_crat_table(crat_image, - &temp_topology_device_list, - proximity_domain); - if (ret || - kfd_is_acpi_crat_invalid(&temp_topology_device_list)) { - kfd_release_topology_device_list( - &temp_topology_device_list); - kfd_destroy_crat_image(crat_image); - crat_image = NULL; - } + ret = kfd_create_crat_image_virtual(&crat_image, &image_size, + COMPUTE_UNIT_CPU, NULL, + proximity_domain); + cpu_only_node = 1; + if (ret) { + pr_err("Error creating VCRAT table for CPU\n"); + return ret; } - if (!crat_image) { - ret = kfd_create_crat_image_virtual(&crat_image, &image_size, - COMPUTE_UNIT_CPU, NULL, - proximity_domain); - cpu_only_node = 1; - if (ret) { - pr_err("Error creating VCRAT table for CPU\n"); - return ret; - } - - ret = kfd_parse_crat_table(crat_image, - &temp_topology_device_list, - proximity_domain); - if (ret) { - pr_err("Error parsing VCRAT table for CPU\n"); - goto err; - } + ret = kfd_parse_crat_table(crat_image, + &temp_topology_device_list, + proximity_domain); + if (ret) { + pr_err("Error parsing VCRAT table for CPU\n"); + goto err; } kdev = list_first_entry(&temp_topology_device_list, struct kfd_topology_device, list); - kfd_add_perf_to_topology(kdev); down_write(&topology_lock); kfd_topology_update_device_list(&temp_topology_device_list, @@ -1189,8 +1136,7 @@ static struct kfd_topology_device *kfd_assign_gpu(struct kfd_node *gpu) /* Discrete GPUs need their own topology device list * entries. Don't assign them to CPU/APU nodes. */ - if (!gpu->kfd->use_iommu_v2 && - dev->node_props.cpu_cores_count) + if (dev->node_props.cpu_cores_count) continue; if (!dev->gpu && (dev->node_props.simd_count > 0)) { @@ -1931,23 +1877,27 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev) HSA_CAP_TRAP_DEBUG_WAVE_LAUNCH_TRAP_OVERRIDE_SUPPORTED | HSA_CAP_TRAP_DEBUG_WAVE_LAUNCH_MODE_SUPPORTED; - if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0)) { - dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9 | - HSA_DBG_WATCH_ADDR_MASK_HI_BIT; + if (kfd_dbg_has_ttmps_always_setup(dev->gpu)) + dev->node_props.debug_prop |= HSA_DBG_DISPATCH_INFO_ALWAYS_VALID; - if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(9, 4, 2)) + if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0)) { + if (KFD_GC_VERSION(dev->gpu) == IP_VERSION(9, 4, 3)) dev->node_props.debug_prop |= - HSA_DBG_DISPATCH_INFO_ALWAYS_VALID; + HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9_4_3 | + HSA_DBG_WATCH_ADDR_MASK_HI_BIT_GFX9_4_3; else + dev->node_props.debug_prop |= + HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9 | + HSA_DBG_WATCH_ADDR_MASK_HI_BIT; + + if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(9, 4, 2)) dev->node_props.capability |= HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED; } else { dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 | HSA_DBG_WATCH_ADDR_MASK_HI_BIT; - if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(11, 0, 0)) - dev->node_props.debug_prop |= HSA_DBG_DISPATCH_INFO_ALWAYS_VALID; - else + if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(11, 0, 0)) dev->node_props.capability |= HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED; } @@ -2083,10 +2033,7 @@ int kfd_topology_add_device(struct kfd_node *gpu) * Overwrite ATS capability according to needs_iommu_device to fix * potential missing corresponding bit in CRAT of BIOS. */ - if (dev->gpu->kfd->use_iommu_v2) - dev->node_props.capability |= HSA_CAP_ATS_PRESENT; - else - dev->node_props.capability &= ~HSA_CAP_ATS_PRESENT; + dev->node_props.capability &= ~HSA_CAP_ATS_PRESENT; /* Fix errors in CZ CRAT. * simd_count: Carrizo CRAT reports wrong simd_count, probably @@ -2281,29 +2228,6 @@ int kfd_numa_node_to_apic_id(int numa_node_id) return kfd_cpumask_to_apic_id(cpumask_of_node(numa_node_id)); } -void kfd_double_confirm_iommu_support(struct kfd_dev *gpu) -{ - struct kfd_topology_device *dev; - - gpu->use_iommu_v2 = false; - - if (!gpu->device_info.needs_iommu_device) - return; - - down_read(&topology_lock); - - /* Only use IOMMUv2 if there is an APU topology node with no GPU - * assigned yet. This GPU will be assigned to it. - */ - list_for_each_entry(dev, &topology_device_list, list) - if (dev->node_props.cpu_cores_count && - dev->node_props.simd_count && - !dev->gpu) - gpu->use_iommu_v2 = true; - - up_read(&topology_lock); -} - #if defined(CONFIG_DEBUG_FS) int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h index cba2cd5ed9d1..dea32a9e5506 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h @@ -32,9 +32,12 @@ #define KFD_TOPOLOGY_PUBLIC_NAME_SIZE 32 #define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9 6 +#define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX9_4_3 7 #define HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 7 #define HSA_DBG_WATCH_ADDR_MASK_HI_BIT \ (29 << HSA_DBG_WATCH_ADDR_MASK_HI_BIT_SHIFT) +#define HSA_DBG_WATCH_ADDR_MASK_HI_BIT_GFX9_4_3 \ + (30 << HSA_DBG_WATCH_ADDR_MASK_HI_BIT_SHIFT) struct kfd_node_properties { uint64_t hive_id; |