linux-stable.git/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h, branch linux-6.3.y

drm/amdkfd: Page aligned memory reserve size

2023-01-11T21:41:03+00:00

Use page aligned size to reserve memory usage because page aligned TTM
BO size is used to unreserve memory usage, otherwise no page aligned
size causes memory usage accounting unbalanced.

Change vram_used definition type to int64_t to be able to trigger
WARN_ONCE(adev && adev->kfd.vram_used < 0, "..."), to help debug the
accounting issue with warning and backtrace.

Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Cleanup vm process info if init vm failed

2023-01-11T21:40:45+00:00

If acquire_vm failed when initializing KFD vm, set vm->process_info to
NULL and free process info, otherwise, the future acquire_vm will
always fail as vm->process_info is not NULL.

Pass avm as parameter to remove the duplicate code.

Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Fix double release compute pasid

2022-12-20T17:46:20+00:00

If kfd_process_device_init_vm returns failure after vm is converted to
compute vm and vm->pasid set to compute pasid, KFD will not take
pdd->drm_file reference. As a result, drm close file handler maybe
called to release the compute pasid before KFD process destroy worker to
release the same pasid and set vm->pasid to zero, this generates below
WARNING backtrace and NULL pointer access.

Add helper amdgpu_amdkfd_gpuvm_set_vm_pasid and call it at the last step
of kfd_process_device_init_vm, to ensure vm pasid is the original pasid
if acquiring vm failed or is the compute pasid with pdd->drm_file
reference taken to avoid double release same pasid.

 amdgpu: Failed to create process VM object
 ida_free called for id=32770 which is not allocated.
 WARNING: CPU: 57 PID: 72542 at ../lib/idr.c:522 ida_free+0x96/0x140
 RIP: 0010:ida_free+0x96/0x140
 Call Trace:
  amdgpu_pasid_free_delayed+0xe1/0x2a0 [amdgpu]
  amdgpu_driver_postclose_kms+0x2d8/0x340 [amdgpu]
  drm_file_free.part.13+0x216/0x270 [drm]
  drm_close_helper.isra.14+0x60/0x70 [drm]
  drm_release+0x6e/0xf0 [drm]
  __fput+0xcc/0x280
  ____fput+0xe/0x20
  task_work_run+0x96/0xc0
  do_exit+0x3d0/0xc10

 BUG: kernel NULL pointer dereference, address: 0000000000000000
 RIP: 0010:ida_free+0x76/0x140
 Call Trace:
  amdgpu_pasid_free_delayed+0xe1/0x2a0 [amdgpu]
  amdgpu_driver_postclose_kms+0x2d8/0x340 [amdgpu]
  drm_file_free.part.13+0x216/0x270 [drm]
  drm_close_helper.isra.14+0x60/0x70 [drm]
  drm_release+0x6e/0xf0 [drm]
  __fput+0xcc/0x280
  ____fput+0xe/0x20
  task_work_run+0x96/0xc0
  do_exit+0x3d0/0xc10

Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdgpu: Add notifier lock for KFD userptrs

2022-12-14T14:48:05+00:00

Add a per-process MMU notifier lock for processing notifiers from
userptrs. Use that lock to properly synchronize page table updates with
MMU notifiers.

Signed-off-by: Felix Kuehling 
Reviewed-by: Xiaogang Chen
Signed-off-by: Alex Deucher

drm/amdkfd: Cleanup kfd_dev struct

2022-10-27T19:12:09+00:00

Cleanup kfd_dev struct by removing ddev and pdev as both
drm_device and pci_dev can be fetched from amdgpu_device.

Signed-off-by: Mukul Joshi 
Tested-by: Amber Lin 
Reviewed-by: Felix Kuehling 
Acked-by: Christian König 
Signed-off-by: Alex Deucher

drm/amdgpu: Pessimistic availability based on rounded up allocations

2022-08-10T18:58:57+00:00

Separately accumulate a statistic of rounded up allocations to use
to report availability, with a view to increasing the likelihood a
buffer object can be successfully allocated at exactly the size
reported by the availability API.

Signed-off-by: Daniel Phillips 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdgpu: add debugfs for kfd system and ttm mem used

2022-07-28T20:05:16+00:00

This keeps track of kfd system mem used and kfd ttm mem used.

Signed-off-by: Alex Sierra 
Reviewed-by: Philip Yang 
Signed-off-by: Alex Deucher

drm/amdkfd: track unified memory reservation with xnack off

2022-07-28T20:05:16+00:00

[WHY]
Unified memory with xnack off should be tracked, as userptr mappings
and legacy allocations do. To avoid oversuscribe system memory when
xnack off.
[How]
Exposing functions reserve_mem_limit and unreserve_mem_limit to SVM
API and call them on every prange creation and free.

Signed-off-by: Alex Sierra 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Add user queue eviction restore SMI event

2022-06-30T19:31:14+00:00

Output user queue eviction and restore event. User queue eviction may be
triggered by svm or userptr MMU notifier, TTM eviction, device suspend
and CRIU checkpoint and restore.

User queue restore may be rescheduled if eviction happens again while
restore.

Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher

drm/amdkfd: Enable GFX11 usermode queue oversubscription

2022-06-23T21:22:12+00:00

Starting with GFX11, MES requires wptr BOs to be GTT allocated/mapped to
GART for usermode queues in order to support oversubscription. In the
case that work is submitted to an unmapped queue, MES must have a GART
wptr address to determine whether the queue should be mapped.

This change is accompanied with changes in MES and is applicable for
MES_API_VERSION >= 2.

v3:
- Use amdgpu_vm_bo_lookup_mapping for wptr_bo mapping lookup
- Move wptr_bo refcount increment to amdgpu_amdkfd_map_gtt_bo_to_gart
- Remove list_del_init from amdgpu_amdkfd_map_gtt_bo_to_gart
- Cleanup/fix create_queue wptr_bo error handling
v4:
- Add MES version shift/mask defines to amdgpu_mes.h
- Change version check from MES_VERSION to MES_API_VERSION
- Add check in kfd_ioctl_create_queue before wptr bo pin/GART map to
ensure bo is a single page.

Signed-off-by: Graham Sider 
Acked-by: Alex Deucher 
Acked-by: Christian König 
Reviewed-by: Philip Yang 
Signed-off-by: Alex Deucher