linux-stable.git/drivers/gpu/drm/amd, branch linux-5.4.y

drm/amdkfd: Tie UNMAP_LATENCY to queue_preemption

2025-12-03T11:45:10+00:00

[ Upstream commit f3820e9d356132e18405cd7606e22dc87ccfa6d1 ]

When KFD asks CP to preempt queues, other than preempt CP queues, CP
also requests SDMA to preempt SDMA queues with UNMAP_LATENCY timeout.
Currently queue_preemption_timeout_ms is 9000 ms by default but can be
configured via module parameter. KFD_UNMAP_LATENCY_MS is hard coded as
4000 ms though. This patch ties KFD_UNMAP_LATENCY_MS to
queue_preemption_timeout_ms so in a slow system such as emulator, both
CP and SDMA slowness are taken into account.

Signed-off-by: Amber Lin 
Reviewed-by: Harish Kasiviswanathan 
Signed-off-by: Alex Deucher 
Signed-off-by: Sasha Levin

drm/amdkfd: return -ENOTTY for unsupported IOCTLs

2025-12-03T11:45:09+00:00

[ Upstream commit 57af162bfc8c05332a28c4d458d246cc46d2746d ]

Some kfd ioctls may not be available depending on the kernel version the
user is running, as such we need to report -ENOTTY so userland can
determine the cause of the ioctl failure.

Signed-off-by: Geoffrey McRae 
Acked-by: Alex Deucher 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher 
Signed-off-by: Sasha Levin

drm/amdgpu: use atomic functions with memory barriers for vm fault info

2025-10-29T13:00:01+00:00

[ Upstream commit 6df8e84aa6b5b1812cc2cacd6b3f5ccbb18cda2b ]

The atomic variable vm_fault_info_updated is used to synchronize access to
adev->gmc.vm_fault_info between the interrupt handler and
get_vm_fault_info().

The default atomic functions like atomic_set() and atomic_read() do not
provide memory barriers. This allows for CPU instruction reordering,
meaning the memory accesses to vm_fault_info and the vm_fault_info_updated
flag are not guaranteed to occur in the intended order. This creates a
race condition that can lead to inconsistent or stale data being used.

The previous implementation, which used an explicit mb(), was incomplete
and inefficient. It failed to account for all potential CPU reorderings,
such as the access of vm_fault_info being reordered before the atomic_read
of the flag. This approach is also more verbose and less performant than
using the proper atomic functions with acquire/release semantics.

Fix this by switching to atomic_set_release() and atomic_read_acquire().
These functions provide the necessary acquire and release semantics,
which act as memory barriers to ensure the correct order of operations.
It is also more efficient and idiomatic than using explicit full memory
barriers.

Fixes: b97dfa27ef3a ("drm/amdgpu: save vm fault information for amdkfd")
Cc: stable@vger.kernel.org
Signed-off-by: Gui-Dong Han 
Signed-off-by: Felix Kuehling 
Reviewed-by: Felix Kuehling 
Signed-off-by: Alex Deucher 
[ kept kgd_dev parameter and adev cast in amdgpu_amdkfd_gpuvm_get_vm_fault_info ]
Signed-off-by: Sasha Levin 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: drop hw access in non-DC audio fini

2025-09-09T16:43:59+00:00

commit 71403f58b4bb6c13b71c05505593a355f697fd94 upstream.

We already disable the audio pins in hw_fini so
there is no need to do it again in sw_fini.

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4481
Cc: oushixiong 
Signed-off-by: Alex Deucher 
(cherry picked from commit 5eeb16ca727f11278b2917fd4311a7d7efb0bbd6)
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman

Revert "drm/amdgpu: fix incorrect vm flags to map bo"

2025-09-04T12:05:56+00:00

commit ac4ed2da4c1305a1a002415058aa7deaf49ffe3e upstream.

This reverts commit b08425fa77ad2f305fe57a33dceb456be03b653f.

Revert this to align with 6.17 because the fixes tag
was wrong on this commit.

Signed-off-by: Alex Deucher 
(cherry picked from commit be33e8a239aac204d7e9e673c4220ef244eb1ba3)
Signed-off-by: Greg Kroah-Hartman

drm/amd/display: Find first CRTC and its line time in dce110_fill_display_configs

2025-08-28T14:21:32+00:00

commit 669f73a26f6112eedbadac53a2f2707ac6d0b9c8 upstream.

dce110_fill_display_configs is shared between DCE 6-11, and
finding the first CRTC and its line time is relevant to DCE 6 too.
Move the code to find it from DCE 11 specific code.

Signed-off-by: Timur Kristóf 
Acked-by: Alex Deucher 
Reviewed-by: Rodrigo Siqueira 
Reviewed-by: Alex Hung 
Signed-off-by: Alex Deucher 
(cherry picked from commit 4ab09785f8d5d03df052827af073d5c508ff5f63)
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman

drm/amd/display: Fix fractional fb divider in set_pixel_clock_v3

2025-08-28T14:21:32+00:00

commit 10507478468f165ea681605d133991ed05cdff62 upstream.

For later VBIOS versions, the fractional feedback divider is
calculated as the remainder of dividing the feedback divider by
a factor, which is set to 1000000. For reference, see:
- calculate_fb_and_fractional_fb_divider
- calc_pll_max_vco_construct

However, in case of old VBIOS versions that have
set_pixel_clock_v3, they only have 1 byte available for the
fractional feedback divider, and it's expected to be set to the
remainder from dividing the feedback divider by 10.
For reference see the legacy display code:
- amdgpu_pll_compute
- amdgpu_atombios_crtc_program_pll

This commit fixes set_pixel_clock_v3 by dividing the fractional
feedback divider passed to the function by 100000.

Fixes: 4562236b3bc0 ("drm/amd/dc: Add dc display driver (v2)")
Signed-off-by: Timur Kristóf 
Acked-by: Alex Deucher 
Reviewed-by: Rodrigo Siqueira 
Reviewed-by: Alex Hung 
Signed-off-by: Alex Deucher 
(cherry picked from commit 027e7acc7e17802ebf28e1edb88a404836ad50d6)
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix incorrect vm flags to map bo

2025-08-28T14:21:30+00:00

[ Upstream commit 040bc6d0e0e9c814c9c663f6f1544ebaff6824a8 ]

It should use vm flags instead of pte flags
to specify bo vm attributes.

Fixes: 7946340fa389 ("drm/amdgpu: Move csa related code to separate file")
Signed-off-by: Jack Xiao 
Reviewed-by: Likun Gao 
Signed-off-by: Alex Deucher 
(cherry picked from commit b08425fa77ad2f305fe57a33dceb456be03b653f)
Signed-off-by: Sasha Levin

drm/amd/pm/powerplay/hwmgr/smu_helper: fix order of mask and value

2025-08-28T14:21:19+00:00

[ Upstream commit a54e4639c4ef37a0241bac7d2a77f2e6ffb57099 ]

There is a small typo in phm_wait_on_indirect_register().

Swap mask and value arguments provided to phm_wait_on_register() so that
they satisfy the function signature and actual usage scheme.

Found by Linux Verification Center (linuxtesting.org) with Svace static
analysis tool.

In practice this doesn't fix any issues because the only place this
function is used uses the same value for the value and mask.

Fixes: 3bace3591493 ("drm/amd/powerplay: add hardware manager sub-component")
Signed-off-by: Fedor Pchelkin 
Signed-off-by: Alex Deucher 
Signed-off-by: Sasha Levin

drm/amdkfd: Set SDMA_RLCx_IB_CNTL/SWITCH_INSIDE_IB

2025-06-27T10:02:54+00:00

[ Upstream commit ab9fcc6362e0699fc1150aa1d8503c40fce2c1e1 ]

When submitting MQD to CP, set SDMA_RLCx_IB_CNTL/SWITCH_INSIDE_IB bit so
it'll allow SDMA preemption if there is a massive command buffer of
long-running SDMA commands.

Signed-off-by: Amber Lin 
Acked-by: Alex Deucher 
Signed-off-by: Alex Deucher 
Signed-off-by: Sasha Levin