CVE-2026-99619: drm/amdgpu: don't drop the gang dependency fence twice
In the Linux kernel, the following vulnerability has been resolved:
drm/amdgpu: don't drop the gang dependency fence twice
drmschedjobadddependency() drops the fence reference when it fails, but amdgpucssubmit() drops it again, so the scheduled fence of a gang member loses a reference it still needs.
A GFX+COMPUTE gang waiting for another context, with the xarray node allocation for the leader's dependency failed through failslab:
WARNING: drivers/dma-buf/dma-fence.c:589 at dmafencerelease+0x34d/0x730, CPU#1: amduapiregres/2993 drmschedjobcleanup+0x2c2/0x340 [gpusched] amdgpujobfree+0xcb/0x3e0 [amdgpu] amdgpucsparserfini+0x435/0x710 [amdgpu] amdgpucsioctl+0xcd2/0x7a60 [amdgpu] Oops: general protection fault, probably for non-canonical address 0xdffffc0000000001: 0000 [#1] SMP KASAN NOPTI KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f] RIP: 0010:dmafencesignaltimestamplocked+0x1b8/0x610
(cherry picked from commit 5aa070a83c8378810b906cc8bfc650ce69822a7c)
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 6.12.118 - 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 to a fixed release to a version that resolves this vulnerability.
Patch 5aa070a83c8378810b906cc8bfc650ce69822a7c
Frequently Asked Questions
Which systems and workloads are exposed to this issue?
The affected code path is in the Linux kernel's amdgpu DRM driver. The reported trigger involves a GFX+COMPUTE gang submission waiting on another context.
What must occur for the failure to be triggered?
The leader gang member's dependency xarray-node allocation must fail while adding a scheduler job dependency. The report demonstrates this condition using failslab fault injection.
How can I identify a system that has encountered this issue?
Reported symptoms include a dma_fence_release warning, followed by a general protection fault and a KASAN null-pointer dereference in dma_fence_signal_timestamp_locked. Relevant cleanup-path frames include drm_sched_job_cleanup, amdgpu_job_free, amdgpu_cs_parser_fini, and amdgpu_cs_ioctl.