{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-99625","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-25T10:25:14.425Z","datePublished":"2026-11-20T05:49:44.491Z","dateUpdated":"2026-11-22T11:09:19.419Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-11-22T11:09:19.419Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\ndrm/amdgpu: Fix duplicate dependency fence releases\n\ndrm_sched_job_add_dependency() consumes its fence reference on success\nand failure. Do not release the fence again if\ndrm_sched_job_add_dependency() fails.\n\n(cherry picked from commit e04f3dfb2071451633e6440a3911a9bb65df9144)"}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H","baseScore":7,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - The extra dma_fence_put is in amdgpu_sync_push_to_job, called from amdgpu_cs_sync_rings during AMDGPU_CS (amdgpu_cs_ioctl), and in amdgpu_vm_sdma_update, called from amdgpu_vm_bo_update on that same ioctl. Both paths handle a local dma_fence installed by the submitting process.\nAC:H - The extra put runs only when drm_sched_job_add_dependency fails, which is only xa_alloc returning ENOMEM (xa_limit_32b is not exhaustible). The caller cannot force that GFP_KERNEL xarray-node allocation to fail at will; amdgpu_vm_eviction_lock also sets PF_MEMALLOC, so that path skips reclaim and draws on reserves.\nPR:L - AMDGPU_CS is registered DRM_AUTH|DRM_RENDER_ALLOW, and drm_ioctl_permit accepts a render-node fd without CAP_SYS_ADMIN or DRM master. A local user who can open /dev/dri/renderD* reaches amdgpu_cs_ioctl.\nUI:N - The attacker opens the render node and submits AMDGPU_CS with an unsignaled syncobj dependency themselves. No other user has to mount media or act on the submission.\nS:U - The unbalanced put frees a dma_fence still referenced by an amdgpu_sync_entry, a drm_syncobj, or a page-table BO dma_resv in the caller's kernel, the same security authority as AMDGPU_CS.\nC:H - amdgpu_sync_free's dma_fence_put, after the extra put in amdgpu_sync_push_to_job, can drive the kref to zero and dma_fence_release kfree_rcu's the fence while a drm_syncobj or dma_resv still points at it. Later waits read the freed fence, including ops and error.\nI:H - dma_fence_release calls fence->ops->release, and a later wait invokes ops->wait or ops->enable_signaling on the reclaimed dma_fence. Those are function pointers (drm_sched_fence_ops_scheduled/finished), so reclaiming the slot is kernel code execution.\nA:H - Freeing the fence from the unbalanced dma_fence_put, or using it afterward from amdgpu_sync_entry_free or the PT BO reservation object, dereferences freed memory and oopses the kernel."}]}],"affected":[{"product":"Linux","vendor":"Linux","defaultStatus":"unaffected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c","drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c"],"versions":[{"version":"c1c4a8b217213c1924eabf4f28385bbee9cc50c1","lessThan":"9ef784728cfbfcccfed5d3ea92f7861571efc064","status":"affected","versionType":"git"},{"version":"c1c4a8b217213c1924eabf4f28385bbee9cc50c1","lessThan":"5d6e0c2dccf210838ed4eaf4d736a48532bf7bc5","status":"affected","versionType":"git"},{"version":"c1c4a8b217213c1924eabf4f28385bbee9cc50c1","lessThan":"4cb2aeb7f23d25b8349ee87048415818b94320ad","status":"affected","versionType":"git"},{"version":"c1c4a8b217213c1924eabf4f28385bbee9cc50c1","lessThan":"ea47f84199241dd083d40a241beea665415d4f6c","status":"affected","versionType":"git"}]},{"product":"Linux","vendor":"Linux","defaultStatus":"affected","repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","programFiles":["drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c","drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c"],"versions":[{"version":"6.2","status":"affected"},{"version":"0","lessThan":"6.2","status":"unaffected","versionType":"semver"},{"version":"6.12.118","lessThanOrEqual":"6.12.*","status":"unaffected","versionType":"semver"},{"version":"6.18.64","lessThanOrEqual":"6.18.*","status":"unaffected","versionType":"semver"},{"version":"7.3.5","lessThanOrEqual":"7.3.*","status":"unaffected","versionType":"semver"},{"version":"7.4-rc2","lessThanOrEqual":"*","status":"unaffected","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"operator":"OR","negate":false,"cpeMatch":[{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.2","versionEndExcluding":"6.12.118"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.2","versionEndExcluding":"6.18.64"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.2","versionEndExcluding":"7.3.5"},{"vulnerable":true,"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionStartIncluding":"6.2","versionEndExcluding":"7.4-rc2"}]}]}],"references":[{"url":"https://git.kernel.org/stable/c/9ef784728cfbfcccfed5d3ea92f7861571efc064"},{"url":"https://git.kernel.org/stable/c/5d6e0c2dccf210838ed4eaf4d736a48532bf7bc5"},{"url":"https://git.kernel.org/stable/c/4cb2aeb7f23d25b8349ee87048415818b94320ad"},{"url":"https://git.kernel.org/stable/c/ea47f84199241dd083d40a241beea665415d4f6c"}],"title":"drm/amdgpu: Fix duplicate dependency fence releases","x_generator":{"engine":"bippy-1.2.0"}}}}