CVE-2026-99630: drm/amdgpu/userq: validate the moved individual BOs on restore
In the Linux kernel, the following vulnerability has been resolved:
drm/amdgpu/userq: validate the moved individual BOs on restore
An individual BO of a user queue VM, i.e. one which doesn't share the root reservation, goes to the individual needsupdate list when it is evicted. The user queue restore looked at vm->invalidated for them, which the VM state machine rework turned into the alwaysvalid evicted list and commit 18decfc9e8cc ("drm/amdgpu/userq: actually validate userq BOs") into the individual evicted one. That list only holds imported DMA-bufs of compute VMs though, so for a user queue VM nothing is validated.
The restore then neither locks such a BO nor moves it back. Its PTEs are updated to wherever it is now, cleared if it went to system memory, and it doesn't get the new eviction fence, so it can move again under the running queue.
On navi48, once the BOs of an SDMA user queue are evicted from GTT (debugfs amdgpuevictgtt) the queue faults on its ring and rptr:
amdgpu 0000:03:00.0: in page starting at address 0x0000000100000000 from client 10 amdgpu 0000:03:00.0: Faulty UTCL2 client ID: SDMA0 (0xd) amdgpu 0000:03:00.0: in page starting at address 0x0000000100101000 from client 10
Validate the BOs on the individual needsupdate list like before.
(cherry picked from commit eb0d94765d4be18d41e93ad09947dd9cc06ed31a)
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.3.5
Frequently Asked Questions
Which workloads are affected by this issue?
It affects AMDGPU user queue VMs with individual buffer objects that do not share the root reservation. The described failure concerns SDMA user queues after their buffer objects are evicted from GTT.
What conditions trigger the failure?
The affected individual buffer objects must be evicted and then processed during user queue restore. Because they are not validated, locked, moved back, or assigned a new eviction fence, they can move again while the queue is running.
How can I identify a likely occurrence?
On Navi48, the reported symptom is a page fault on the SDMA queue ring and read pointer after GTT eviction, such as messages identifying a faulty UTCL2 client ID of SDMA0. The example also reports faults at addresses beginning 0x0000000100000000 and 0x0000000100101000.