======================================================
WARNING: possible circular locking dependency detected
6.5.0-kfd-fkuehlin #276 Not tainted
------------------------------------------------------
kworker/8:2/2676 is trying to acquire lock:
ffff9435aae95c88 ((work_completion)(&svm_bo->eviction_work)){+.+.}-{0:0}, at: __flush_work+0x52/0x550
but task is already holding lock:
ffff9435cd8e1720 (&svms->lock){+.+.}-{3:3}, at: svm_range_deferred_list_work+0xe8/0x340 [amdgpu]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&svms->lock){+.+.}-{3:3}:
__mutex_lock+0x97/0xd30
kfd_ioctl_alloc_memory_of_gpu+0x6d/0x3c0 [amdgpu]
kfd_ioctl+0x1b2/0x5d0 [amdgpu]
__x64_sys_ioctl+0x86/0xc0
do_syscall_64+0x39/0x80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
-> #1 (&mm->mmap_lock){++++}-{3:3}:
down_read+0x42/0x160
svm_range_evict_svm_bo_worker+0x8b/0x340 [amdgpu]
process_one_work+0x27a/0x540
worker_thread+0x53/0x3e0
kthread+0xeb/0x120
ret_from_fork+0x31/0x50
ret_from_fork_asm+0x11/0x20
-> #0 ((work_completion)(&svm_bo->eviction_work)){+.+.}-{0:0}:
__lock_acquire+0x1426/0x2200
lock_acquire+0xc1/0x2b0
__flush_work+0x80/0x550
__cancel_work_timer+0x109/0x190
svm_range_bo_release+0xdc/0x1c0 [amdgpu]
svm_range_free+0x175/0x180 [amdgpu]
svm_range_deferred_list_work+0x15d/0x340 [amdgpu]
process_one_work+0x27a/0x540
worker_thread+0x53/0x3e0
kthread+0xeb/0x120
ret_from_fork+0x31/0x50
ret_from_fork_asm+0x11/0x20
other info that might help us debug this:
Chain exists of:
(work_completion)(&svm_bo->eviction_work) --> &mm->mmap_lock --> &svms->lock
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&svms->lock);
lock(&mm->mmap_lock);
lock(&svms->lock);
lock((work_completion)(&svm_bo->eviction_work));
I believe this cannot really lead to a deadlock in practice, because
svm_range_evict_svm_bo_worker only takes the mmap_read_lock if the BO
refcount is non-0. That means it's impossible that svm_range_bo_release
is running concurrently. However, there is no good way to annotate this.
To avoid the problem, take a BO reference in
svm_range_schedule_evict_svm_bo instead of in the worker. That way it's
impossible for a BO to get freed while eviction work is pending and the
cancel_work_sync call in svm_range_bo_release can be eliminated.
v2: Use svm_bo_ref_unless_zero and explained why that's safe. Also
removed redundant checks that are already done in
amdkfd_fence_enable_signaling.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Philip Yang <philip.yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Updates for v6.8:
Core:
- Add support for SDM670, SM8650
- Handle the CFG interconnect to fix the obscure hangs / timeouts
on register write
- Kconfig fix for QMP dependency
- DT schema fixes
DPU:
- Add support for SDM670, SM8650
- Enable SmartDMA on SM8350 and SM8450
- Correct UBWC settings for SC8280XP
- Fix catalog settings for SC8180X
- Actually make use of the version to switch between QSEED3/3LITE/4
scalers
- Use devres-managed and drm-managed allocations where appropriate
- misc other fixes
- Enabled YUV writeback on SC7280, SM8250
- Enabled writeback on SM8350, SM8450
- CRC fix when encoder is selected as the input source
- other misc fixes
MDP4:
- Use devres-managed and drm-managed allocations where appropriate
- flush vblank event on CRTC disable
MDP5:
- Use devres-managed and drm-managed allocations where appropriate
DP:
- Add support for SM8650
- Enable PM runtime support
- Merge msm-specific debugfs dir with the generic one
- Described DisplayPort on SM8150 in DeviceTree bindings
- Moved dp_display_get_next_bridge() to probe()
DSI:
- Add support for SM8650
- Enable PM runtime support
GPU/GEM:
- demote userspace triggerable warnings to debug
- add GEM object metadata UAPI
- move GPU devcoredumps to GPU device
- fix hangcheck to skip retired submits
- expose UBWC config to userspace
- fix a680 chip-id
- drm_exec conversion
- drm/ci: remove rebase-merge directory (to unblock CI)
[airlied: fix drm_exec/amd interaction]
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGs9auYqmo-7NSd9FsbNBCDf7aBevd=4xkcF3A5G_OGvMQ@mail.gmail.com
SVM uses hmm page walk to valid buffer before map to gpu vm. After have partial
migration/mapping do validation on same vm range as migration/map do instead of
whole svm range that can be very large. This change is expected to improve svm
code performance.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Philip Yang <philip.yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On gfx943 APU there is no VRAM and page migration, queue CWSR area, svm
range with always mapped flag, is not mapped to GPU correctly. This
works fine if retry fault on CWSR area can be recovered, but could cause
deadlock if there is another retry fault recover waiting for CWSR to
finish.
Fix this by mapping svm range with always mapped flag to GPU with ACCESS
attribute if XNACK ON.
There is side effect, because all GPUs have ACCESS attribute by default
on new svm range with XNACK on, the CWSR area will be mapped to all GPUs
after this change. This side effect will be fixed with Thunk change to
set CWSR svm range with ACCESS_IN_PLACE attribute on the GPU that user
queue is created.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch implements partial migration/mapping for gpu/cpu page faults in SVM
according to migration granularity(default 2MB). A svm range may include pages
from both system ram and vram of one gpu now. These chagnes are expected to
improve migration performance and reduce mmu callback and TLB flush workloads.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make restore workers freezable so we don't have to explicitly flush them
in suspend and GPU reset code paths, and we don't accidentally try to
restore BOs while the GPU is suspended. Not having to flush restore_work
also helps avoid lock/fence dependencies in the GPU reset case where we're
not allowed to wait for fences.
A side effect of this is, that we can now have multiple concurrent threads
trying to signal the same eviction fence. Rework eviction fence signaling
and replacement to account for that.
The GPU reset path can no longer rely on restore_process_worker to resume
queues because evict/restore workers can run independently of it. Instead
call a new restore_process_helper directly.
This is an RFC and request for testing.
v2:
- Reworked eviction fence signaling
- Introduced restore_process_helper
v3:
- Handle unsignaled eviction fences in restore_process_bos
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Tested-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On gfx943 APU, EXT_COHERENT should give MTYPE_CC for local and
MTYPE_UC for nonlocal memory.
On NUMA systems, local memory gets the local mtype, set by an
override callback. If EXT_COHERENT is set, memory will be set as
MTYPE_UC by default, with local memory MTYPE_CC.
Add an option in the override function for this case, and
add a check to ensure it is not used on UNCACHED memory.
V2: Combined APU and NUMA code into one patch
V3: Fixed a potential nullptr in amdgpu_vm_bo_update
Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit dc427a473e.
The change prevents migrating the entire range to VRAM because retry
fault restore_pages map the remaining system memory range to GPUs. It
will work correctly to submit together with partial mapping to GPU
patch later.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Split SVM ranges that have been mapped into 2MB page table entries,
require to be remap in case the split has happened in a non-aligned
VA.
[WHY]:
This condition causes the 2MB page table entries be split into 4KB
PTEs.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch implements partial migration in gpu page fault according to migration
granularity(default 2MB) and not split svm range in cpu page fault handling.
A svm range may include pages from both system ram and vram of one gpu now.
These chagnes are expected to improve migration performance and reduce mmu
callback and TLB flush workloads.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If there is no VRAM domain, bo_node is NULL and this causes crash.
Refactor the change, and use the module parameter as higher privilege.
Need another patch to support override PTE flag on APU.
Fixes: 5f248462c6 ("drm/amdgpu: Add EXT_COHERENT memory allocation flags")
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
prange->svm_bo unref can happen in both mmu callback and a callback after
migrate to system ram. Both are async call in different tasks. Sync svm_bo
unref operation to avoid random "use-after-free".
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Tested-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If new range is splited to multiple pranges with max_svm_range_pages
alignment and added to update_list, svm validate and map should keep
going after error to make sure prange->mapped_to_gpu flag is up to date
for the whole range.
svm validate and map update set prange->mapped_to_gpu after mapping to
GPUs successfully, otherwise clear prange->mapped_to_gpu flag (for
update mapping case) instead of setting error flag, we can remove
the redundant error flag to simpliy code.
Refactor to remove goto and update prange->mapped_to_gpu flag inside
svm_range_lock, to guarant we always evict queues or unmap from GPUs if
there are invalid ranges.
After svm validate and map return error -EAGIN, the caller retry will
update the mapping for the whole range again.
Fixes: c22b044070 ("drm/amdkfd: flag added to handle errors from svm validate and map")
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Tested-by: James Zhu <james.zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch fixes:
1: ref number of prange's svm_bo got decreased by an async call from hmm. When
wait svm_bo of prange got released we shoul also wait prang->svm_bo become NULL,
otherwise prange->svm_bo may be set to null after allocate new vram buffer.
2: During waiting svm_bo of prange got released in a while loop should reschedule
current task to give other tasks oppotunity to run, specially the the workque
task that handles svm_bo ref release, otherwise we may enter to softlock.
Signed-off-by: Xiaogang.Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The validated_once flag is not used after the prefault was removed, The
prefault was needed to ensure validate all system memory pages at least
once before mapping or migrating the range to GPU.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We do not need free dma address array of svm_range each time we do dma unmap
for pages in svm_range as we can reuse the same array. Only free it when free
svm_range. Separate these two operations and use them accordingly.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pull drm fixes from Dave Airlie:
"Regular rounds of rc1 fixes, a large bunch for amdgpu since it's three
weeks in one go, one i915, one nouveau and one ivpu.
I think there might be a few more fixes in misc that I haven't pulled
in yet, but we should get them all for rc2.
amdgpu:
- Display replay fixes
- Fixes for headless boards
- Fix documentation breakage
- RAS fixes
- Handle newer IP discovery tables
- SMU 13.0.6 fixes
- SR-IOV fixes
- Display vstartup fixes
- NBIO 7.9 fixes
- Display scaling mode fixes
- Debugfs power reporting fix
- GC 9.4.3 fixes
- Dirty framebuffer fixes for fbcon
- eDP fixes
- DCN 3.1.5 fix
- Display ODM fixes
- GPU core dump fix
- Re-enable zops property now that IGT test is fixed
- Fix possible UAF in CS code
- Cursor degamma fix
amdkfd:
- HMM fixes
- Interrupt masking fix
- GFX11 MQD fixes
i915:
- Mark requests for GuC virtual engines to avoid use-after-free
nouveau:
- Fix fence state in nouveau_fence_emit()
ivpu:
- replace strncpy"
* tag 'drm-next-2023-09-08' of git://anongit.freedesktop.org/drm/drm: (51 commits)
drm/amdgpu: Restrict bootloader wait to SMUv13.0.6
drm/amd/display: prevent potential division by zero errors
drm/amd/display: enable cursor degamma for DCN3+ DRM legacy gamma
drm/amd/display: limit the v_startup workaround to ASICs older than DCN3.1
Revert "drm/amd/display: Remove v_startup workaround for dcn3+"
drm/amdgpu: fix amdgpu_cs_p1_user_fence
Revert "Revert "drm/amd/display: Implement zpos property""
drm/amdkfd: Add missing gfx11 MQD manager callbacks
drm/amdgpu: Free ras cmd input buffer properly
drm/amdgpu: Hide xcp partition sysfs under SRIOV
drm/amdgpu: use read-modify-write mode for gfx v9_4_3 SQ setting
drm/amdkfd: use mask to get v9 interrupt sq data bits correctly
drm/amdgpu: Allocate coredump memory in a nonblocking way
drm/amdgpu: Support query ecc cap for aqua_vanjaram
drm/amdgpu: Add umc_info v4_0 structure
drm/amd/display: always switch off ODM before committing more streams
drm/amd/display: Remove wait while locked
drm/amd/display: update blank state on ODM changes
drm/amd/display: Add smu write msg id fail retry process
drm/amdgpu: Add SMU v13.0.6 default reset methods
...
if hmm_range_get_pages returns EBUSY error during
svm_range_validate_and_map, within the context of a page fault
interrupt. This should retry through svm_range_restore_pages
callback. Therefore we treat this as EAGAIN error instead, and defer
it to restore pages fallback.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
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
...
if hmm_range_get_pages returns EBUSY error during
svm_range_validate_and_map, within the context of a page fault
interrupt. This should retry through svm_range_restore_pages
callback. Therefore we treat this as EAGAIN error instead, and defer
it to restore pages fallback.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When CONFIG_DYNAMIC_DEBUG is disabled altogether, calling
_dynamic_func_call_no_desc() does not work:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c: In function 'svm_range_set_attr':
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c:52:9: error: implicit declaration of function '_dynamic_func_call_no_desc' [-Werror=implicit-function-declaration]
52 | _dynamic_func_call_no_desc("svm_range_dump", svm_range_debug_dump, svms)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c:3564:9: note: in expansion of macro 'dynamic_svm_range_dump'
3564 | dynamic_svm_range_dump(svms);
| ^~~~~~~~~~~~~~~~~~~~~~
Add a compile-time conditional in addition to the runtime check.
Fixes: 8923137dbe ("drm/amdkfd: avoid svm dump when dynamic debug disabled")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
DMA address reference within svm_ranges should be unmapped only after
the memory has been released from the system. In case of range
splitting, the DMA address information should be copied to the
corresponding range after this has split. But leaving dma mapping
intact.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set dynamic_svm_range_dump macro to avoid iterating over SVM lists
from svm_range_debug_dump when dynamic debug is disabled. Otherwise,
it could drop performance, specially with big number of SVM ranges.
Make sure both svm_range_set_attr and svm_range_debug_dump functions
are dynamically enabled to print svm_range_debug_dump debug traces.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Tested-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If the SVM range has no GPU access nor access-in-place attribute,
validate and map to GPU should skip the range.
Add NULL pointer check if find_first_bit(ctx->bitmap, MAX_GPU_INSTANCE)
returns MAX_GPU_INSTANCE as gpuidx if ctx->bitmap is empty.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If a return error is raised during validation and mapping of a
prange, this flag is set. It is a rare occurrence, but it could happen
when `amdgpu_hmm_range_get_pages_done` returns true. In such cases,
the caller should retry. However, it is important to ensure that the
prange is updated correctly during the retry.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Allocate large local variable on heap to avoid exceeding the
stack size:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c: In function ‘svm_range_validate_and_map’:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c:1690:1: warning: the frame size of 2360 bytes is larger than 2048 bytes [-Wframe-larger-than=]
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rework logic or use do_div() to avoid problems on 32 bit.
v2: add a missing case for XCP macro
v3: fix out of bounds array access
v4: fix xcp handling harder
Acked-by: Guchun Chen <guchun.chen@amd.com> (v1)
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com> (v3)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
svm_migrate_init set the max svm range pages based on the KFD nodes
partition size. APU mode don't init pgmap because there is no migration.
kgd2kfd_device_init calls svm_migrate_init after KFD nodes allocation
and initialization.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch fixes memory reporting on the GFX 9.4.3 APU and dGPU
by reporting available memory on a per partition basis. If its an
APU, available and used memory calculations take into account
system and TTM memory.
v2: squash in fix ("drm/amdkfd: Fix array out of bound warning")
squash in fix ("drm/amdgpu: Update memory reporting for GFX9.4.3")
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For memory accounting per compute partition and export drm amdgpu bo and
then import to KFD, we need the xcp id to account the memory usage or
find the KFD node of the original amdgpu bo to create the KFD bo on the
correct adev KFD node.
Set xcp_id_plus1 of amdgpu_bo_param to create bo and store xcp_id to
amddgpu bo. Add helper macro to get the mem_id from adev and xcp_id.
v2: squash in fix ("drm/amdgpu: Fix BO creation failure on GFX 9.4.3 dGPU")
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use MTYPE RW/MTYPE_CC for mapping system memory or VRAM to KFD node
within the same memory partition, use MTYPE_NC for mapping on KFD node
from the far memory partition of the same socket or from another socket
on same XGMI hive.
On NPS4 or 4P system, MTYPE will be overridden per page depending on
the memory NUMA node id and vm->mem_id.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pass kfd node->xcp->mem_id to amdgpu bo create parameter mem_id_plus1 to
allocate new svm_bo on the specified memory partition.
This is only for dGPU mode as we don't migrate with APU mode.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
