Skip to contentSkip to conversation

FAILED: patch "[PATCH] drm/amdgpu: avoid force-completing uninitialized UVD rings" failed to apply to 6.12-stable tree

PATCH New [email protected] 2 participants 5 messages Sep 9 – Sep 12

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkout FETCH_HEAD
git cherry-pick -x 6760f5cb12d2366ddd58a2d8637f7583d73f596b
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026090944-nutmeg-wasp-168d@gregkh' --subject-prefix 'PATCH 6.12.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 6760f5cb12d2366ddd58a2d8637f7583d73f596b Mon Sep 17 00:00:00 2001
From: Bob Zhou <[email protected]>
Date: Wed, 19 Aug 2026 14:23:49 +0800
Subject: [PATCH] drm/amdgpu: avoid force-completing uninitialized UVD rings

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")
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +2 -1
index 228a405a94c4..ecd7caa95d4b 100644
@@ -517,7 +517,8 @@ 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;
maintainer
#1 lore raw

PATCH 1/4 drm/amdgpu: Fix missing unwind in amdgpu_ib_schedule() error path

From: Srinivasan Shanmugam <[email protected]>

[ Upstream commit ba038065655c45728be346d0b174a6da08d8a5c5 ]

amdgpu_ib_schedule() returns early after calling amdgpu_ring_undo(). This skips the common free_fence cleanup path. Other error paths were already changed to use goto free_fence, but this one was missed.

Change the early return to goto free_fence so all error paths clean up the same way.

Fixes the below:
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c:232 amdgpu_ib_schedule()
warn: missing unwind goto?

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

    124 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
    125                        struct amdgpu_ib *ibs, struct amdgpu_job *job,
    126                        struct dma_fence **f)
    127 {

    ...

    224
    225         if (ring->funcs->insert_start)
    226                 ring->funcs->insert_start(ring);
    227
    228         if (job) {
    229                 r = amdgpu_vm_flush(ring, job, need_pipe_sync);
    230                 if (r) {
    231                         amdgpu_ring_undo(ring);

--> 232 return r;

	The patch changed the other error paths to goto free_fence but
	this one was accidentally skipped.

    233                 }
    234         }
    235
    236         amdgpu_ring_ib_begin(ring);

    ...

    338
    339 free_fence:
    340         if (!job)
    341                 kfree(af);
    342         return r;
    343 }
Fixes f903b85ed0f1 ("drm/amdgpu: fix possible fence leaks from job structure")
Cc Alex Deucher <[email protected]>
Cc Christian König <[email protected]>

Stable-dep-of: 6760f5cb12d2 ("drm/amdgpu: avoid force-completing uninitialized UVD rings")

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +1 -1
index dea7cf04fc984..a60b7b5596be1 100644
@@ -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);
-			return r;
+			goto free_fence;
 		}
 	}
 
Signature · 1 line
2.53.0
maintainer
#2 lore raw

PATCH 2/4 drm/amdgpu: add a helper to calculate ring distance

From: Alex Deucher <[email protected]>

[ Upstream commit 08e7d6c3cee880bd3ec3eb14c478f9fa805b0bdc ]

Add a helper to calculate the distance in DWs between two wptrs.

Stable-dep-of: 6760f5cb12d2 ("drm/amdgpu: avoid force-completing uninitialized UVD rings")

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +12 -6
index 3eb74f8457802..743699df8ae7a 100644
@@ -431,6 +431,17 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 	ring->count_dw -= count_dw;
 }
 
+static inline unsigned int amdgpu_ring_get_dw_distance(struct amdgpu_ring *ring,
+						       u64 start_wptr, u64 end_wptr)
+{
+	unsigned int start = start_wptr & ring->buf_mask;
+	unsigned int end = end_wptr & ring->buf_mask;
+
+	if (end < start)
+		end += ring->ring_size >> 2;
+	return end - start;
+}
+
 /**
  * amdgpu_ring_patch_cond_exec - patch dw count of conditional execute
  * @ring: amdgpu_ring structure
@@ -441,18 +452,13 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 static inline void amdgpu_ring_patch_cond_exec(struct amdgpu_ring *ring,
 					       unsigned int offset)
 {
-	unsigned cur;
-
 	if (!ring->funcs->init_cond_exec)
 		return;
 
 	WARN_ON(offset > ring->buf_mask);
 	WARN_ON(ring->ring[offset] != 0);
 
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur < offset)
-		cur += ring->ring_size >> 2;
-	ring->ring[offset] = cur - offset;
+	ring->ring[offset] = amdgpu_ring_get_dw_distance(ring, offset, ring->wptr - 1);
 }
 
 #define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset)			\
Signature · 1 line
2.53.0
maintainer
#3 lore raw

PATCH 3/4 drm/amdgpu: plumb timedout fence through to force completion

In reply to Sasha Levin

From: Alex Deucher <[email protected]>

[ 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

Adapt this dependency to 6.12 without importing newer queue-reset infrastructure. Drop the SDMA/VCN/JPEG reset hunks and the adjacent ring-reset helper context because those functions do not exist here. Keep the two-argument force-completion API and its fence error handling.

Use &job->hw_fence.base for this branch's embedded job fence, and update the additional per-ring timeout caller in amdgpu_job.c to pass that fence. Preserve the existing full-reset job-fence clearing and resubmission flow. The UVD resume caller now matches the context required by the target fix. No functions are added.

The prerequisite ba038065655c also introduced a goto to the absent free_fence label in amdgpu_ib_schedule(). Restore the original return r:
6.12 embeds job fences and allocates non-job fences later, in amdgpu_fence_emit(), so this VM-flush failure path has no new fence to free. This removes the pre-existing build failure without importing the newer fence allocation and cleanup infrastructure.

[ sashal: Reduced backport -- upstream c184df870db1e touches 9 file(s), this
backport carries 7. 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")

---
 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_job.c     |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h    |  3 ++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c     |  2 +-
 7 files changed, 33 insertions(+), 9 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +1 -1
index 81f16e4447f8e..6de83a422b982 100644
@@ -1960,7 +1960,7 @@ 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);
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +5 -1
index 20a13db7cd882..240a4e3cbee0b 100644
@@ -5464,6 +5464,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 {
 	int i, r = 0;
 	struct amdgpu_job *job = NULL;
+	struct dma_fence *fence = NULL;
 	struct amdgpu_device *tmp_adev = reset_context->reset_req_dev;
 	bool need_full_reset =
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
@@ -5476,6 +5477,9 @@ 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];
@@ -5489,7 +5493,7 @@ 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);
drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +22 -3
index 569e0e5373927..7c7da452dfabd 100644
@@ -594,7 +594,7 @@ 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 &&
@@ -739,11 +739,30 @@ 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);
 }
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +1 -1
index a60b7b5596be1..dea7cf04fc984 100644
@@ -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;
 		}
 	}
 
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +1 -1
index 99e9e96f5b7fd..0abc38bb242ca 100644
@@ -148,7 +148,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 			if (amdgpu_ring_sched_ready(ring))
 				drm_sched_stop(&ring->sched, s_job);
 			atomic_inc(&ring->adev->gpu_reset_counter);
-			amdgpu_fence_driver_force_completion(ring);
+			amdgpu_fence_driver_force_completion(ring, &job->hw_fence.base);
 			if (amdgpu_ring_sched_ready(ring))
 				drm_sched_start(&ring->sched);
 			goto exit;
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +2 -1
index 743699df8ae7a..30252cd45840b 100644
@@ -146,7 +146,8 @@ 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,
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +1 -1
index fc6b0650ee7a2..b0ed8e122c3b8 100644
@@ -516,7 +516,7 @@ 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;
Signature · 1 line
2.53.0
maintainer
#4 lore raw

PATCH 4/4 drm/amdgpu: avoid force-completing uninitialized UVD rings

In reply to Sasha Levin

From: Bob Zhou <[email protected]>

[ 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")
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +2 -1
index b0ed8e122c3b8..ea37ac3bc4ea1 100644
@@ -516,7 +516,8 @@ 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;
Signature · 1 line
2.53.0