CVE-2026-97928: drm/amdgpu: skip the VMID 0 flush for VRAM
In the Linux kernel, the following vulnerability has been resolved:
drm/amdgpu: skip the VMID 0 flush for VRAM
Clear-on-release only runs on VRAM, which amdgputtmmapbuffer() reaches via its direct MC address without programming a GART window, yet the wipe still forces a VMID 0 flush. On GFX11 (e.g. Navi33) that spurious SDMA flush can wedge the engine; only flush when a GART window is actually used.
v2: Let amdgputtmmapbuffer() return whether the VMID 0 flush is needed, and drive the clear and copy paths from that. (Christian) v3: Make the vmneedsflush output parameter mandatory instead of allowing NULL. (Christian)
(cherry picked from commit a306e406e570b74318ff7d80e5b07b540ca1d3a9)
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
Linux kernel (drm/amdgpu)to a version that resolves this vulnerability.Patch a306e406e570b74318ff7d80e5b07b540ca1d3a9
Event History
Frequently Asked Questions
Does the available information describe an attacker-controlled or remote exploitation path?
No. The available information describes an AMDGPU driver failure during VRAM clear-on-release handling and does not identify a remote interface, unprivileged trigger, or attacker-controlled input.
What systems and activity are associated with the failure?
The issue is associated with GFX11 hardware, with Navi33 given as an example, when the clear-on-release path operates directly on VRAM. The unnecessary VMID 0 SDMA flush can wedge the engine in that situation.
Are affected kernel version ranges provided?
No affected or fixed version ranges are provided. The data includes two stable-kernel references and identifies the originating commit as a306e406e570b74318ff7d80e5b07b540ca1d3a9.