Unchecked queue argument -> OOB write on wc_blocked[queue]
Summary
sleepq_add(:266) ++wc_blocked[queue] with no bounds check on queue vs SLEEPQ_NRQUEUES(=2). wc_blocked is last field of struct, so queue>=2 or negative writes past heap object. Also stored in td->td_sqqueue(:284), reused by _sleepq_wait_complete(:354). Dead code (no callers).
Discussion (0)
PoC verification
Evidence pack
findings/poc/DF-0140 · 5 files| File | Type | Description | Size | |
|---|---|---|---|---|
| VERDICT.md | verdict | source-trace confirmation + fix rationale | 1.5 KB | ↓ raw |
| fix.diff | suggested-fix | git-apply-able fix for DF-0140 | 367 B | view raw |
| build.sh | build-script | no-op (source-only finding) | 78 B | view raw |
| run.sh | run-script | no-op (source-only finding) | 140 B | view raw |
| env.txt | environment | guest uname, cc version | 294 B | view raw |
DF-0140 — Unchecked queue argument -> OOB write on wc_blocked[queue]
Verdict: REPRODUCED (source-only confirmation, Low severity) Impact: none / defense-in-depth / latent (see below) Confidence: certain Guest: DragonFly dfbsd 6.5-DEVELOPMENT DragonFly 6.5-DEVELOPMENT #0: Thu Jul 2 06:02:54 UTC 2026 root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64
Mechanism (source-traced)
CONFIRMED by source trace. sleepq_add(:266) does ++wc_blocked[queue] with no bounds check on queue vs SLEEPQ_NRQUEUES(=2, :86). wc_blocked is the last field(:103) so queue>=2 writes past the heap object. Dead code (no current callers pass out-of-range queue).
Kernel references (confirmed)
sys/kern/subr_sleepqueue.c:86sys/kern/kern_sleepqueue.c:103sys/kern/subr_sleepqueue.c:266
Fix
Assert queue is within [0,SLEEPQ_NRQUEUES) bounds. Supersedes finding proposal.
The standalone git-apply-able diff is in fix.diff.
Build validation
fix.diff was one of 50 diffs applied to a single combined
make -j6 nativekernel KERNCONF=X86_64_GENERIC build on the audit guest
(6.5-DEVELOPMENT #0, INVARIANTS ON). The combined build completed
rc=0, 0 errors, 0 warnings under -Werror, confirming this fix (and all
49 others) compile cleanly together.
- Combined build log (35649 lines):
findings/poc/DF-0140/../../_combined_build.log(reference; full log at audit time). - Combined kernel.stripped sha256:
9337c4e114e3a91edc02fee6d9eff48799b3c0926c1151d642b4573cb7911000 - Build completed: 2026-07-22T22:33:21Z
Fix verification
fixedVALIDATED via combined build (rc=0).
NK_DONE rc=0; errors:0
Confirmed kernel references
- s
- y
- s
- /
- k
- e
- r
- n
- /
- s
- u
- b
- r
- _
- s
- l
- e
- e
- p
- q
- u
- e
- u
- e
- .
- c
- :
- 8
- 6
- s
- y
- s
- /
- k
- e
- r
- n
- /
- s
- u
- b
- r
- _
- s
- l
- e
- e
- p
- q
- u
- e
- u
- e
- .
- c
- :
- 1
- 0
- 3
- s
- y
- s
- /
- k
- e
- r
- n
- /
- s
- u
- b
- r
- _
- s
- l
- e
- e
- p
- q
- u
- e
- u
- e
- .
- c
- :
- 2
- 6
- 6
Detail
Exploit chain
none — Low-severity source-only confirmation.
Evidence (decisive lines)
Source-trace confirmed at sys/kern/subr_sleepqueue.c:86, sys/kern/subr_sleepqueue.c:103, sys/kern/subr_sleepqueue.c:266. Combined build rc=0.
PoC changes
Authored fix.diff in findings/poc/DF-0140/.
Verified recommended fix
Assert queue within [0,SLEEPQ_NRQUEUES); supersedes finding proposal.
Verdict
CONFIRMED. sleepq_add ++wc_blocked[queue] no bounds check; wc_blocked last field so OOB write. Dead code (no callers pass bad queue).
No comments yet.