* [PATCH 6.6.y 2/3] drm/amdgpu: plumb timedout fence through to force completion
2026-09-12 20:59 ` [PATCH 6.6.y 1/3] drm/amdgpu: Fix missing unwind in amdgpu_ib_schedule() error path Sasha Levin
@ 2026-09-12 20:59 ` Sasha Levin
2026-09-12 20:59 ` [PATCH 6.6.y 3/3] drm/amdgpu: avoid force-completing uninitialized UVD rings Sasha Levin
1 sibling, 0 replies; 4+ messages in thread
From: Sasha Levin @ 2026-09-12 20:59 UTC (permalink / raw)
To: stable; +Cc: Alex Deucher, Christian König, Sasha Levin
From: Alex Deucher <alexander.deucher@amd.com>
[ Upstream commit c184df870db1e328691ea0fbb7d0e59efd9d3f9f ]
When we do a full adapter reset, if we know the timedout fence
mark the fence with -ETIME rather than -ECANCELED so it
gets properly handled by userspace.
v2: rebase
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Stable backport notes:
- Keep the two-argument force-completion API and its existing callers so
6760f5cb12d2 can add the UVD fence-driver initialization check unchanged.
- Use &job->hw_fence.base for 6.6's embedded hardware fence. Preserve the
existing scheduler reset and job-fence clearing flow used for resubmission.
- Drop SDMA, VCN and JPEG reset hunks whose functions do not exist here;
vcn_v5_0_1.c is also absent. Do not import newer queue-reset helpers or
their declarations. No functions are added by this backport.
- Restore the direct return after amdgpu_ring_undo() on VM-flush failure.
The earlier ccea36c819df backport introduced a goto free_fence without
the label or the newer fence-allocation model it depends on. In 6.6,
amdgpu_ib_schedule() has no locally allocated fence to free on this
path. This restores the original 6.6 error handling and fixes the
build without importing the unrelated fence-lifetime conversion.
[ sashal: Reduced backport -- upstream c184df870db1e touches 9 file(s), this
backport carries 6. Not backported here:
drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c
This note is generated from the file lists only; see the resolution record
for the reasoning. ]
Stable-dep-of: 6760f5cb12d2 ("drm/amdgpu: avoid force-completing uninitialized UVD rings")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 25 ++++++++++++++++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 3 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 2 +-
6 files changed, 32 insertions(+), 8 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 256cc15fc9b5a..2044f5cca3514 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -1947,7 +1947,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
/* swap out the old fences */
amdgpu_ib_preempt_fences_swap(ring, fences);
- amdgpu_fence_driver_force_completion(ring);
+ amdgpu_fence_driver_force_completion(ring, NULL);
/* resubmit unfinished jobs */
amdgpu_ib_preempt_job_recovery(&ring->sched);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 892a40c439637..60e1d9cda3a8b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4894,6 +4894,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
{
int i, r = 0;
struct amdgpu_job *job = NULL;
+ struct dma_fence *fence = NULL;
bool need_full_reset =
test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
@@ -4907,6 +4908,9 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
amdgpu_fence_driver_isr_toggle(adev, true);
+ if (job)
+ fence = &job->hw_fence.base;
+
/* block all schedulers and reset given job's ring */
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
@@ -4920,7 +4924,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
amdgpu_fence_driver_clear_job_fences(ring);
/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
- amdgpu_fence_driver_force_completion(ring);
+ amdgpu_fence_driver_force_completion(ring, fence);
}
amdgpu_fence_driver_isr_toggle(adev, false);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index 017dd494d0a2f..93208bd5631af 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -595,7 +595,7 @@ void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev)
r = -ENODEV;
/* no need to trigger GPU reset as we are unloading */
if (r)
- amdgpu_fence_driver_force_completion(ring);
+ amdgpu_fence_driver_force_completion(ring, NULL);
if (!drm_dev_is_unplugged(adev_to_drm(adev)) &&
ring->fence_drv.irq_src &&
@@ -740,11 +740,30 @@ void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error)
* amdgpu_fence_driver_force_completion - force signal latest fence of ring
*
* @ring: fence of the ring to signal
+ * @timedout_fence: fence of the timedout job
*
*/
-void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring)
+void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring,
+ struct dma_fence *timedout_fence)
{
- amdgpu_fence_driver_set_error(ring, -ECANCELED);
+ struct amdgpu_fence_driver *drv = &ring->fence_drv;
+ unsigned long flags;
+
+ spin_lock_irqsave(&drv->lock, flags);
+ for (unsigned int i = 0; i <= drv->num_fences_mask; ++i) {
+ struct dma_fence *fence;
+
+ fence = rcu_dereference_protected(drv->fences[i],
+ lockdep_is_held(&drv->lock));
+ if (fence && !dma_fence_is_signaled_locked(fence)) {
+ if (fence == timedout_fence)
+ dma_fence_set_error(fence, -ETIME);
+ else
+ dma_fence_set_error(fence, -ECANCELED);
+ }
+ }
+ spin_unlock_irqrestore(&drv->lock, flags);
+
amdgpu_fence_write(ring, ring->fence_drv.sync_seq);
amdgpu_fence_process(ring);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 2cce0a378ecb3..806ec5d021995 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -217,7 +217,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
r = amdgpu_vm_flush(ring, job, need_pipe_sync);
if (r) {
amdgpu_ring_undo(ring);
- goto free_fence;
+ return r;
}
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index dc17e2ce84d07..b40dbe2f4fc25 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -143,7 +143,8 @@ extern const struct drm_sched_backend_ops amdgpu_sched_ops;
void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring);
void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error);
-void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
+void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring,
+ struct dma_fence *timedout_fence);
int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);
int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index aab293ec3417e..b6047dea0a1c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -511,7 +511,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev)
}
memset_io(ptr, 0, size);
/* to restore uvd fence seq */
- amdgpu_fence_driver_force_completion(&adev->uvd.inst[i].ring);
+ amdgpu_fence_driver_force_completion(&adev->uvd.inst[i].ring, NULL);
}
}
return 0;
--
2.53.0
^ permalink raw reply related [flat|nested] 4+ messages in thread* [PATCH 6.6.y 3/3] drm/amdgpu: avoid force-completing uninitialized UVD rings
2026-09-12 20:59 ` [PATCH 6.6.y 1/3] drm/amdgpu: Fix missing unwind in amdgpu_ib_schedule() error path Sasha Levin
2026-09-12 20:59 ` [PATCH 6.6.y 2/3] drm/amdgpu: plumb timedout fence through to force completion Sasha Levin
@ 2026-09-12 20:59 ` Sasha Levin
1 sibling, 0 replies; 4+ messages in thread
From: Sasha Levin @ 2026-09-12 20:59 UTC (permalink / raw)
To: stable; +Cc: Bob Zhou, Leo Liu, Frank Min, Alex Deucher, Sasha Levin
From: Bob Zhou <bobzhou2@amd.com>
[ Upstream commit 6760f5cb12d2366ddd58a2d8637f7583d73f596b ]
uvd_v7_0_sw_init() does not initialize the UVD decode ring for an
SR-IOV VF. However, amdgpu_uvd_resume() unconditionally force-completes
the decode ring when restoring its fence sequence.
Skip fence completion when the fence driver is not initialized.
Fixes: 0a33b11d26c6 ("drm/amdgpu: mark force completed fences with -ECANCELED")
Cc: stable@vger.kernel.org
Signed-off-by: Bob Zhou <bobzhou2@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Acked-by: Frank Min <Frank.Min@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index b6047dea0a1c6..48d7a5b514151 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -511,7 +511,8 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev)
}
memset_io(ptr, 0, size);
/* to restore uvd fence seq */
- amdgpu_fence_driver_force_completion(&adev->uvd.inst[i].ring, NULL);
+ if (adev->uvd.inst[i].ring.fence_drv.initialized)
+ amdgpu_fence_driver_force_completion(&adev->uvd.inst[i].ring, NULL);
}
}
return 0;
--
2.53.0
^ permalink raw reply related [flat|nested] 4+ messages in thread