kernel.git/drivers/gpu/drm/amd/amdgpu/amdgpu.h, branch linux-6.14.y

drm/amdgpu: rework how isolation is enforced v2

2025-05-29T09:12:30Z

[ Upstream commit bd22e44ad415ac22e3a4f9a983d2a085f6cb4427 ] Limiting the number of available VMIDs to enforce isolation causes some issues with gang submit and applying certain HW workarounds which require multiple VMIDs to work correctly. So instead start to track all submissions to the relevant engines in a per partition data structure and use the dma_fences of the submissions to enforce isolation similar to what a VMID limit does. v2: use ~0l for jobs without isolation to distinct it from kernel submissions which uses NULL for the owner. Add some warning when we are OOM. Signed-off-by: Christian König Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

Revert "drm/amd: Stop evicting resources on APUs in suspend"

2025-05-18T06:26:03Z

commit d0ce1aaa8531a4a4707711cab5721374751c51b0 upstream. This reverts commit 3a9626c816db901def438dc2513622e281186d39. This breaks S4 because we end up setting the s3/s0ix flags even when we are entering s4 since prepare is used by both flows. The causes both the S3/s0ix and s4 flags to be set which breaks several checks in the driver which assume they are mutually exclusive. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3634 Cc: Mario Limonciello Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher (cherry picked from commit ce8f7d95899c2869b47ea6ce0b3e5bf304b2fff4) Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman

drm/amd: Forbid suspending into non-default suspend states

2025-05-02T06:02:13Z

[ Upstream commit 1657793def101dac7c9d3b2250391f6a3dd934ba ] On systems that default to 'deep' some userspace software likes to try to suspend in 'deep' first. If there is a failure for any reason (such as -ENOMEM) the failure is ignored and then it will try to use 's2idle' as a fallback. This fails, but more importantly it leads to graphical problems. Forbid this behavior and only allow suspending in the last state supported by the system. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4093 Acked-by: Alex Deucher Link: https://lore.kernel.org/r/20250408180957.4027643-1-superm1@kernel.org Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher (cherry picked from commit 2aabd44aa8a3c08da3d43264c168370f6da5e81d) Signed-off-by: Sasha Levin

drm/amdgpu: Increase KIQ invalidate_tlbs timeout

2025-05-02T06:02:07Z

[ Upstream commit 3666ed821832f42baaf25f362680dda603cde732 ] KIQ invalidate_tlbs request has been seen to marginally exceed the configured 100 ms timeout on systems under load. All other KIQ requests in the driver use a 10 second timeout. Use a similar timeout implementation on the invalidate_tlbs path. v2: Poll once before msleep v3: Fix return value Signed-off-by: Jay Cornwall Cc: Kent Russell Reviewed-by: Harish Kasiviswanathan Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdgpu: partially revert "reduce reset time"

2024-12-18T17:39:07Z

This partially reverts commit 194eb174cbe4fe2b3376ac30acca2dc8c8beca00. This commit introduced a new state variable into adev without even remotely worrying about CPU barriers. Since we already have the amdgpu_in_reset() function exactly for this use case partially revert that. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher

drm/amd: Add Suspend/Hibernate notification callback support

2024-12-10T15:26:50Z

As part of the suspend sequence VRAM needs to be evicted on dGPUs. In order to make suspend/resume more reliable we moved this into the pmops prepare() callback so that the suspend sequence would fail but the system could remain operational under high memory usage suspend. Another class of issues exist though where due to memory fragementation there isn't a large enough contiguous space and swap isn't accessible. Add support for a suspend/hibernate notification callback that could evict VRAM before tasks are frozen. This should allow paging out to swap if necessary. Link: https://github.com/ROCm/ROCK-Kernel-Driver/issues/174 Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3476 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2362 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3781 Reviewed-by: Lijo Lazar Link: https://lore.kernel.org/r/20241128032656.2090059-2-superm1@kernel.org Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher

drm/amdgpu: Add init level for post reset reinit

2024-11-20T15:03:05Z

When device needs to be reset before initialization, it's not required for all IPs to be initialized before a reset. In such cases, it needs to identify whether the IP/feature is initialized for the first time or whether it's reinitialized after a reset. Add RESET_RECOVERY init level to identify post reset reinitialization phase. This only provides a device level identification, IP/features may choose to track their state independently also. Signed-off-by: Lijo Lazar Acked-by: Tao Zhou Signed-off-by: Alex Deucher

drm/amdgpu: Add sysfs interface for gc reset mask

2024-11-08T16:08:01Z

Add two sysfs interfaces for gfx and compute: gfx_reset_mask compute_reset_mask These interfaces are read-only and show the resets supported by the IP. For example, full adapter reset (mode1/mode2/BACO/etc), soft reset, queue reset, and pipe reset. V2: the sysfs node returns a text string instead of some flags (Christian) v3: add a generic helper which takes the ring as parameter and print the strings in the order they are applied (Christian) check amdgpu_gpu_recovery before creating sysfs file itself, and initialize supported_reset_types in IP version files (Lijo) v4: Fixing uninitialized variables (Tim) Signed-off-by: Jesse Zhang Suggested-by: Alex Deucher Reviewed-by: Tim Huang Signed-off-by: Alex Deucher

drm/amdgpu: fix fairness in enforce isolation handling

2024-11-04T16:27:04Z

Make sure KFD gets a turn when serializing access to the GC IP. Currently non-KFD jobs can starve KFD if they submit often enough. This patch prevents that by stalling non-KFD if its time period has elapsed. v2: fix units v3: check enablement properly Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher

drm/amdgpu: clean up the suspend_complete

2024-10-28T20:40:58Z

To check the status of S3 suspend completion, use the PM core pm_suspend_global_flags bit(1) to detect S3 abort events. Therefore, clean up the AMDGPU driver's private flag suspend_complete. Signed-off-by: Prike Liang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher