CVE-2026-99604: drm/etnaviv: Flush the tile status cache through both the PE and BLT
In the Linux kernel, the following vulnerability has been resolved:
drm/etnaviv: Flush the tile status cache through both the PE and BLT
Since commit f232d9ec029c ("drm/etnaviv: fix TS cache flushing on GPUs with BLT engine"), jobs on GPUs with a BLT engine use a BLT command to flush the tile status cache instead of TSFLUSHCACHE, which is broken on most of these GPUs.
GC7000 r6008/r6009 and GC8200 r6401 are the other way around: TSFLUSHCACHE works, but the BLT command does not flush the cache. The vendor driver handles these GPUs using TSFLUSHCACHE, but the etnaviv kernel driver has no way to identify them.
As a result, dirty tile status cache lines survive the end of a job. They are written back by the next cache flush, using the page table of the job issuing it. If that job belongs to another process, the write goes to the wrong address space.
On GC7000 r6009 this causes MMU faults and GPU recoveries blamed on an innocent process. I see around 1100 per run of a multisample depth/stencil test running alongside two other GPU clients.
Emit both flushes at the end of every job on GPUs with a BLT engine. Whichever does not work on a given GPU is a no-op. Since dirty lines are only written back once, this costs one additional state load per job.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 5.5 - Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 6.12.118 - 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 7.3.5 - Upgrade
Upgrade
debian/linuxto a version that resolves this vulnerability.Fixed in 7.3.5-1 - Compensating control
On GPUs with a BLT engine, flush the tile status cache through both the PE and BLT at the end of every job.
Event History
Frequently Asked Questions
Which systems are exposed to this issue?
Systems using the Linux etnaviv DRM driver on GPUs with a BLT engine are affected by the faulty cache-flush handling. The description specifically identifies GC7000 r6008/r6009 and GC8200 r6401 as GPUs where the BLT flush does not clear the tile status cache.
What conditions are needed for the issue to cause cross-process impact?
A job must leave dirty tile status cache lines behind, and a later cache flush must be issued by a job using another process's page table. The surviving lines can then be written into the wrong address space; on GC7000 r6009, this was observed as MMU faults and GPU recoveries attributed to an innocent process.
How is the issue addressed?
The fix emits both the PE TS_FLUSH_CACHE operation and the BLT cache-flush command at the end of every job on GPUs with a BLT engine. On a GPU where one mechanism is ineffective, that operation is treated as a no-op while the working flush clears the cache.
How might an administrator recognize the problem before applying the fix?
On GC7000 r6009, the issue can manifest as GPU MMU faults and GPU recoveries blamed on a process that did not cause the stale cache state. The reported reproduction involved a multisample depth/stencil workload running alongside two other GPU clients.