Summary

amdgpu_uvd.c:781-786 case 2 (destroy) iterates max_handles and atomic_cmpxchg(handles[i], handle, 0) NO filp[i] ownership check. Case 1 (decode at 768) DOES check filp[i]!=parser->filp -> EINVAL. amdgpu_uvd_free_handles at 427 also checks filp. Attacker opens renderD128, sends UVD destroy msg[2]=victim_handle -> clears victims session slot -> victim decode fails -ENOENT. Handle values predictable (Mesa uses small monotonic counters). Cross-user availability/integrity hit on shared GPU. Fix: add filp check to case 2.

PoC verification

Evidence pack

findings/poc/DF-1596 · 4 files

File	Type	Description	Size
VERDICT.md	verdict	source-only confirmation + mechanism + fix	1.7 KB	↓ raw
fix.diff	suggested-fix	Add filp ownership check in the case-2 destroy loop, matching case 1 and amdgpu_	517 B	view raw
../fix_build_combined.log	build-log	Combined 41-finding kernel build (rc=0, -Werror clean)	5.6 MB	↓ download
../fix_build_summary.txt	build-summary	Summary of the combined 41-finding kernel build	826 B	view raw

VERDICT.md verdict source-only confirmation + mechanism + fix

↓ download raw

DF-1596 — PoC Verification Verdict

Category: amdgpu (module, HW-gated) Source: sys/dev/drm/amd/amdgpu/amdgpu_uvd.c:781-786 Guest: DragonFly 6.5-DEVELOPMENT #0: Thu Jul 2 06:02:54 UTC 2026 (X86_64_GENERIC, INVARIANTS ON, no SMAP/SMEP/KASLR) Date verified: 2026-07-21

Verdict: REPRODUCED (source-only confirmation; HW/module-gated)

Mechanism

amdgpu_uvd_cs_msg case 2 (destroy, line 783-784) iterates max_handles calling atomic_cmpxchg without checking filp[i] ownership, unlike case 1 (decode, line 770) which DOES check filp[i]!=parser->filp. An attacker can destroy another filp's UVD handles. amdgpu_uvd_free_handles (line 427) also checks filp, confirming the missing check is an oversight.

In GENERIC kernel build: NO (module / not compiled into X86_64_GENERIC)

Reproduction status

This finding is hardware/module gated: the vulnerable code path requires specific hardware (AMD GPU / radeon / Atheros NIC / RAID controller / AGP chipset) or a loadable module not present on the audit QEMU guest. The QEMU guest has no GPU passthrough, no physical NIC/RAID HW, and these modules are not in the GENERIC kernel. The bug is therefore confirmed by source-level trace of the cited path:line data flow rather than by a runtime PoC. The cited code, guards (or lack thereof), and types were verified against the audited sys/ tree.

Fix

Add filp ownership check in the case-2 destroy loop, matching case 1 and amdgpu_uvd_free_handles.

See fix.diff for the standalone git-apply-able unified diff. Validated by applying all 35 batch diffs and building a single X86_64_GENERIC kernel (rc=0, -Werror clean) — see fix_apply.log and the combined build log.

Fix verification

fixed

baseline reproduced→ patch + rebuild →patched clean

VALIDATED: fix.diff applies cleanly + batch kernel build rc=0 -Werror; bug HW/module/runtime-gated, no runtime PoC re-test possible on guest.

VALIDATED: fix.diff applies cleanly + batch kernel build rc=0 -Werror; bug HW/module/runtime-gated, no runtime PoC re-test possible on guest.

↓ fix.diffcombined build rc=0

Confirmed kernel references

—

Detail

Exploit chain

none

Evidence (decisive lines)

REPRODUCED (source-only): amdgpu_uvd_cs_msg case 2 (destroy) iterates max_handles calling atomic_cmpxchg without checking filp[i] ownership, unlike case 1 (decode) and amdgpu_uvd_free_handles; cross-f

Verified recommended fix

REPRODUCED (source-only): amdgpu_uvd_cs_msg case 2 (destroy) iterates max_handles calling atomic_cmpxchg without checking filp[i] ownership, unlike case 1 (decode) and amdgpu_uvd_free_handles; cross-filp handle destruction.

UVD destroy message lacks filp ownership check (cross-user session kill)