Skip to contentSkip to conversation

dma-fence: Fix dma_fence_timeline_name() to call get_timeline_name()

PATCH In review Baineng Shou 3 participants 3 messages Jun 25

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 dma-fence: Fix dma_fence_timeline_name() to call get_timeline_name()

dma_fence_timeline_name() incorrectly invokes ops->get_driver_name() instead of ops->get_timeline_name(), so every caller receives the driver name where the timeline name was expected.

This is a copy-paste regression that has resurfaced twice. It was originally introduced by commit 62918542b7bf ("dma-fence: Fix sparse warnings due __rcu annotations") when adding the __rcu casts, fixed by commit 033559473dd3 ("dma-fence: Fix safe access wrapper to call timeline name method"), and then accidentally reintroduced by commit e58b4dea9054 ("dma-buf/dma-fence: Add dma_fence_test_signaled_flag()") when both wrappers were refactored to use the new helper.

---
 drivers/dma-buf/dma-fence.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/dma-buf/dma-fence.c +1 -1
index b3bfa6943a8e..5292d714419b 100644
@@ -1202,7 +1202,7 @@ const char __rcu *dma_fence_timeline_name(struct dma_fence *fence)
 	/* RCU protection is required for safe access to returned string */
 	ops = rcu_dereference(fence->ops);
 	if (!dma_fence_test_signaled_flag(fence))
-		return (const char __rcu *)ops->get_driver_name(fence);
+		return (const char __rcu *)ops->get_timeline_name(fence);
 	else
 		return (const char __rcu *)"signaled-timeline";
 }
Signature · 1 line
2.34.1
#1 lore raw
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [Critical] Potential NULL pointer dereference in dma_fence_timeline_name() and dma_fence_driver_name() due to a missing ops NULL check when the fence is concurrently signaled without a read memory barrier. --

commit 118120d707b8593d71c08a5c22b1149fdc6d4229
Author: Baineng Shou <[email protected]>

dma-fence: Fix dma_fence_timeline_name() to call get_timeline_name()

This patch fixes a copy-paste error where dma_fence_timeline_name() incorrectly invoked get_driver_name() instead of get_timeline_name().

drivers/dma-buf/dma-fence.c const char __rcu *dma_fence_timeline_name(struct dma_fence *fence) · line 1205

Quoted 10 lines
drivers/dma-buf/dma-fence.c +1 -1
index b3bfa6943a8e1..5292d714419b1 100644
@@ -1202,7 +1202,7 @@ const char __rcu *dma_fence_timeline_name(struct dma_fence *fence)
 	/* RCU protection is required for safe access to returned string */
 	ops = rcu_dereference(fence->ops);
 	if (!dma_fence_test_signaled_flag(fence))
-		return (const char __rcu *)ops->get_driver_name(fence);
+		return (const char __rcu *)ops->get_timeline_name(fence);

[Severity: Critical]
This is a pre-existing issue, but is it possible for ops to be NULL here due to memory reordering?

If dma_fence_signal_timestamp_locked() runs concurrently, it sets DMA_FENCE_FLAG_SIGNALED_BIT (which includes a full memory barrier via test_and_set_bit()) and then sets fence->ops = NULL via RCU_INIT_POINTER().

Since rcu_dereference() and test_bit() inside dma_fence_test_signaled_flag() are independent memory loads without an intervening read memory barrier, could a reader on weakly-ordered architectures experience reordering?

Reader CPU:
1. Loads fence->flags -> 0 (sees not signaled)
2. Loads fence->ops -> NULL (sees new value)
3. !dma_fence_test_signaled_flag(fence) evaluates to true
4. ops->get_timeline_name(fence) dereferences NULL

Would adding a NULL check for ops, or an smp_rmb() between the reads, be necessary to prevent a potential kernel panic if this is reachable via concurrent ioctls like SYNC_IOC_FILE_INFO?

Does dma_fence_driver_name() also share this same structural vulnerability when dereferencing ops->get_driver_name?

Quoted 3 lines
drivers/dma-buf/dma-fence.c +0 -0
 	else
 		return (const char __rcu *)"signaled-timeline";
 }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#2 lore raw
In reply to Baineng Shou

+Cc André

Quoted 12 lines · Baineng Shou

dma_fence_timeline_name() incorrectly invokes ops->get_driver_name() instead of ops->get_timeline_name(), so every caller receives the driver name where the timeline name was expected.

This is a copy-paste regression that has resurfaced twice. It was originally introduced by commit 62918542b7bf ("dma-fence: Fix sparse warnings due __rcu annotations") when adding the __rcu casts, fixed by commit 033559473dd3 ("dma-fence: Fix safe access wrapper to call timeline name method"), and then accidentally reintroduced by commit e58b4dea9054 ("dma-buf/dma-fence: Add dma_fence_test_signaled_flag()") when both wrappers were refactored to use the new helper.

Quoted 1 line

André has caught this a few days ago already:

https://lore.kernel.org/dri-devel/[email protected]/

To save my honor I want to state that e58b4dea9054 did not actually *do this* change. But merging it must have caused it somehow, since that commit was still based on the old bug. I guess somewhere in conflict resolution somehow tho old code must have entered.

P.

Quoted 17 lines

---
 drivers/dma-buf/dma-fence.c | 2 +-

 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/dma-buf/dma-fence.c +1 -1
index b3bfa6943a8e..5292d714419b 100644
@@ -1202,7 +1202,7 @@ const char __rcu *dma_fence_timeline_name(struct dma_fence *fence)
  	/* RCU protection is required for safe access to returned string */
  	ops = rcu_dereference(fence->ops);
  	if (!dma_fence_test_signaled_flag(fence))
-		return (const char __rcu *)ops->get_driver_name(fence);
+		return (const char __rcu *)ops->get_timeline_name(fence);

  else
  return (const char __rcu *)"signaled-timeline"; }