CVE-2026-99726: drm/v3d: Widen cache_clean_lock over the whole L2TCACTL sequence
In the Linux kernel, the following vulnerability has been resolved:
drm/v3d: Widen cachecleanlock over the whole L2TCACTL sequence
v3dcleancaches() and v3dflushl2t() both write the single L2TCACTL register and poll its status bits. The mutex cachecleanlock exists to serialize them, but v3dcleancaches() only took the lock around its final FLMCLEAN write.
These functions run concurrently: v3dflushl2t() is issued from the BIN/RENDER/CSD invalidate path while v3dcleancaches() runs from the CACHECLEAN queue, and each queue's scheduler uses its own ordered workqueue, so their runjob callbacks execute in parallel.
Because clean locked only its final write, a concurrent flush can write L2TCACTL during clean's unlocked phase. Both use non read-modify-write writes to the one register, so whichever lands last wins: clean's TMUWCF write can land on the flush's in-flight L2TFLS invalidate, triggering the GFXH-1897 hazard of writing L2TCACTL while a flush is pending.
Hold cachecleanlock across the entire L2TCACTL access sequence so it is fully mutually exclusive with v3dflushl2t(), which already takes the lock around its own write.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 6.18.64 - Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.1 - Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.3.5 - Upgrade
Upgrade
debian/linuxto a version that resolves this vulnerability.Fixed in 6.1.176-1Fixed in 6.1.192-2Fixed in 6.12.107-1Fixed in 6.12.117-1Fixed in 7.3.5-1 - Compensating control
Hold cache_clean_lock across the entire L2TCACTL access sequence in v3d_clean_caches(), making it mutually exclusive with v3d_flush_l2t().
Event History
Frequently Asked Questions
Which systems are exposed to this race condition?
Systems using the Linux kernel's V3D DRM driver are exposed when cache-clean and L2T flush operations can run concurrently. The affected paths use separate ordered workqueues, allowing their job callbacks to execute in parallel.
What condition is required to trigger the issue?
The race requires v3d_clean_caches() to be in its unlocked L2TCACTL sequence while v3d_flush_l2t() performs an L2T flush. A subsequent TMUWCF write from the cache-clean path can then overwrite the in-flight L2TFLS invalidate command.
What does the fix change?
The fix holds cache_clean_lock across the complete L2TCACTL access sequence in v3d_clean_caches(). This makes it mutually exclusive with v3d_flush_l2t(), which already locks its own register write.