linux-toradex.git/drivers/gpu/drm/amd, branch v6.16-rc5

drm/amd/display: Don't allow OLED to go down to fully off

2025-06-30T18:01:05+00:00

[Why]
OLED panels can be fully off, but this behavior is unexpected.

[How]
Ensure that minimum luminance is at least 1.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4338
Reviewed-by: Alex Hung 
Signed-off-by: Mario Limonciello 
Signed-off-by: Ray Wu 
Tested-by: Daniel Wheeler 
Signed-off-by: Alex Deucher 
(cherry picked from commit 51496c7737d06a74b599d0aa7974c3d5a4b1162e)

drm/amd/display: Added case for when RR equals panel's max RR using freesync

2025-06-30T18:00:51+00:00

[WHY]
Rounding error sometimes occurs when the refresh rate is equal to a panel's
max refresh rate, causing HDMI compliance failures.

[HOW]
Added a case so that we round up to avoid v_total_min to be below a panel's
minimum bound.

Reviewed-by: Jun Lei 
Signed-off-by: Harold Sun 
Signed-off-by: Ray Wu 
Tested-by: Daniel Wheeler 
Signed-off-by: Alex Deucher 
(cherry picked from commit fe7645d22bc0f7c1558296538ec49987bf268ef6)

drm/amdkfd: add hqd_sdma_get_doorbell callbacks for gfx7/8

2025-06-30T17:57:54+00:00

These were missed when support was added for other generations.
The callbacks are called unconditionally so we need to make
sure all generations have them.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4304
Link: https://github.com/ROCm/ROCm/issues/4965
Fixes: bac38ca8c475 ("drm/amdkfd: implement per queue sdma reset for gfx 9.4+")
Cc: Jonathan Kim 
Reported-by: Johl Brown 
Reviewed-by: Jonathan Kim 
Signed-off-by: Alex Deucher 
(cherry picked from commit 1e9d17a5dcf1242e9518e461d8e63ad35240e49e)
Cc: stable@vger.kernel.org

drm/amdgpu: Fix memory leak in amdgpu_ctx_mgr_entity_fini

2025-06-30T17:57:31+00:00

patch dd64956685fa ("drm/amdgpu: Remove duplicated "context still
alive" check") removed ctx put, which will cause amdgpu_ctx_fini()
cannot be called and then cause some finished fence that added by
amdgpu_ctx_add_fence() cannot be released and cause memleak.

Fixes: dd64956685fa ("drm/amdgpu: Remove duplicated "context still alive" check")
Signed-off-by: Lin.Cao 
Reviewed-by: Tvrtko Ursulin 
Acked-by: Christian König 
Signed-off-by: Alex Deucher 
(cherry picked from commit 8cf66089e28108dedd47e6156a48489303cf525c)
Cc: stable@vger.kernel.org

drm/amdkfd: Don't call mmput from MMU notifier callback

2025-06-30T17:57:12+00:00

If the process is exiting, the mmput inside mmu notifier callback from
compactd or fork or numa balancing could release the last reference
of mm struct to call exit_mmap and free_pgtable, this triggers deadlock
with below backtrace.

The deadlock will leak kfd process as mmu notifier release is not called
and cause VRAM leaking.

The fix is to take mm reference mmget_non_zero when adding prange to the
deferred list to pair with mmput in deferred list work.

If prange split and add into pchild list, the pchild work_item.mm is not
used, so remove the mm parameter from svm_range_unmap_split and
svm_range_add_child.

The backtrace of hung task:

 INFO: task python:348105 blocked for more than 64512 seconds.
 Call Trace:
  __schedule+0x1c3/0x550
  schedule+0x46/0xb0
  rwsem_down_write_slowpath+0x24b/0x4c0
  unlink_anon_vmas+0xb1/0x1c0
  free_pgtables+0xa9/0x130
  exit_mmap+0xbc/0x1a0
  mmput+0x5a/0x140
  svm_range_cpu_invalidate_pagetables+0x2b/0x40 [amdgpu]
  mn_itree_invalidate+0x72/0xc0
  __mmu_notifier_invalidate_range_start+0x48/0x60
  try_to_unmap_one+0x10fa/0x1400
  rmap_walk_anon+0x196/0x460
  try_to_unmap+0xbb/0x210
  migrate_page_unmap+0x54d/0x7e0
  migrate_pages_batch+0x1c3/0xae0
  migrate_pages_sync+0x98/0x240
  migrate_pages+0x25c/0x520
  compact_zone+0x29d/0x590
  compact_zone_order+0xb6/0xf0
  try_to_compact_pages+0xbe/0x220
  __alloc_pages_direct_compact+0x96/0x1a0
  __alloc_pages_slowpath+0x410/0x930
  __alloc_pages_nodemask+0x3a9/0x3e0
  do_huge_pmd_anonymous_page+0xd7/0x3e0
  __handle_mm_fault+0x5e3/0x5f0
  handle_mm_fault+0xf7/0x2e0
  hmm_vma_fault.isra.0+0x4d/0xa0
  walk_pmd_range.isra.0+0xa8/0x310
  walk_pud_range+0x167/0x240
  walk_pgd_range+0x55/0x100
  __walk_page_range+0x87/0x90
  walk_page_range+0xf6/0x160
  hmm_range_fault+0x4f/0x90
  amdgpu_hmm_range_get_pages+0x123/0x230 [amdgpu]
  amdgpu_ttm_tt_get_user_pages+0xb1/0x150 [amdgpu]
  init_user_pages+0xb1/0x2a0 [amdgpu]
  amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu+0x543/0x7d0 [amdgpu]
  kfd_ioctl_alloc_memory_of_gpu+0x24c/0x4e0 [amdgpu]
  kfd_ioctl+0x29d/0x500 [amdgpu]

Fixes: fa582c6f3684 ("drm/amdkfd: Use mmget_not_zero in MMU notifier")
Signed-off-by: Philip Yang 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher 
(cherry picked from commit a29e067bd38946f752b0ef855f3dfff87e77bec7)
Cc: stable@vger.kernel.org

drm/amdgpu: Include sdma_4_4_4.bin

2025-06-30T17:52:01+00:00

This got missed during SDMA 4.4.4 support.

Fixes: 968e3811c3e8 ("drm/amdgpu: add initial support for sdma444")
Signed-off-by: Kent Russell 
Reviewed-by: Alex Deucher 
Signed-off-by: Alex Deucher 
(cherry picked from commit 51526efe02714339ed6139f7bc348377d363200a)
Cc: stable@vger.kernel.org

amdkfd: MTYPE_UC for ext-coherent system memory

2025-06-30T17:51:46+00:00

Set memory mtype to UC host memory when ext-coherent
flag is set and memory is registered as a SVM allocation.

Reviewed-by: Amber Lin 
Signed-off-by: David Yat Sin 
Signed-off-by: Alex Deucher 
(cherry picked from commit 5d14fdab4778c29cfd39e62c3ce84d232b4a7d8c)

drm/amdgpu/sdma5.x: suspend KFD queues in ring reset

2025-06-30T17:51:07+00:00

SDMA 5.x only supports engine soft reset which resets
all queues on the engine.  As such, we need to suspend
KFD queues around resets like we do for SDMA 4.x.

Reviewed-by: Jesse Zhang 
Signed-off-by: Alex Deucher 
(cherry picked from commit 61feed0baa1a0d094af0e07e968b1e6e875f07d0)

drm/amd/display: Add sanity checks for drm_edid_raw()

2025-06-24T14:39:24+00:00

When EDID is retrieved via drm_edid_raw(), it doesn't guarantee to
return proper EDID bytes the caller wants: it may be either NULL (that
leads to an Oops) or with too long bytes over the fixed size raw_edid
array (that may lead to memory corruption).  The latter was reported
actually when connected with a bad adapter.

Add sanity checks for drm_edid_raw() to address the above corner
cases, and return EDID_BAD_INPUT accordingly.

Fixes: 48edb2a4256e ("drm/amd/display: switch amdgpu_dm_connector to use struct drm_edid")
Link: https://bugzilla.suse.com/show_bug.cgi?id=1236415
Signed-off-by: Takashi Iwai 
Signed-off-by: Alex Deucher 
(cherry picked from commit 648d3f4d209725d51900d6a3ed46b7b600140cdf)
Cc: stable@vger.kernel.org

drm/amd/display: Fix AMDGPU_MAX_BL_LEVEL value

2025-06-24T14:39:13+00:00

[Why]
commit 16dc8bc27c2a ("drm/amd/display: Export full brightness range to
userspace") adjusted the brightness range to scale to larger values, but
missed updating AMDGPU_MAX_BL_LEVEL which is needed to make sure that
scaling works properly with custom brightness curves.

[How]
As the change for max brightness of 0xFFFF only applies to devices
supporting DC, use existing DC define MAX_BACKLIGHT_LEVEL.

Fixes: 16dc8bc27c2a ("drm/amd/display: Export full brightness range to userspace")
Acked-by: Alex Deucher 
Link: https://lore.kernel.org/r/20250623171114.1156451-1-mario.limonciello@amd.com
Signed-off-by: Mario Limonciello 
Signed-off-by: Alex Deucher 
(cherry picked from commit 5b852044eb0d3e1f1c946d32e05fcb068e0a20a0)
Cc: stable@vger.kernel.org