⬢ DragonFlyBSD Kernel Audit
DF-4789 / verdict.json
← back to finding ↓ download raw 
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
{
  "finding_id": "DF-4789",
  "status": "untested",
  "reproduced": 0,
  "impact": "none",
  "confidence": "likely",
  "verdict": "mpr_print_sgl(), reached unconditionally from mprsas_scsiio_complete on any MPI2_IOCSTATUS_INVALID_SGL completion (mpr_sas.c:2814), walks the request-frame SGL with no per-frame bound (ieee_sge++/sge++, mpr_table.c:555/581), dereferences chain->chain without a NULL check when the chain TAILQ is exhausted (mpr_table.c:568), and hexdumps a fixed 128 bytes regardless of the IOC-facts request-frame size (mpr_table.c:547). With a buggy or malicious controller this is an OOB heap read printed to console plus a NULL-deref panic (local DoS); no write primitive. Not exercised on the audit guest: it is a virtio-only KVM guest with no LSI MPT3 controller (env.txt), so the mpr driver never attaches and the IOC status cannot be produced. Verified by line-accurate code tracing.",
  "exploit_chain": "",
  "evidence": "README.md (trigger analysis), VERDICT.md (full trace), env.txt (guest has no vendor 0x1000 device), fix.diff (chain NULL guard + walk caps)",
  "kernel_refs": [
    "sys/dev/raid/mpr/mpr_table.c:547",
    "sys/dev/raid/mpr/mpr_table.c:555",
    "sys/dev/raid/mpr/mpr_table.c:557",
    "sys/dev/raid/mpr/mpr_table.c:568",
    "sys/dev/raid/mpr/mpr_table.c:581",
    "sys/dev/raid/mpr/mpr_sas.c:2814"
  ],
  "poc_changes": "no runnable PoC produced: requires Fusion-MPT3 hardware returning INVALID_SGL; audit guest has none",
  "attempts": 0,
  "guest_uname": "DragonFly dfbsd 6.5-DEVELOPMENT #0: Thu Jul  2 06:02:54 UTC 2026 root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64",
  "runtime_sec": 0,
  "guest_dirty": 0,
  "build_cmd": "",
  "run_cmd": "",
  "code_hash": "",
  "notes": "Same class/trust boundary as pass-1 DF-3816/DF-3817 (debug-path OOB reads, Low). The walked SGL is driver-built, so the trigger requires firmware misbehavior or DMA corruption; severity stays Low for that reason.",
  "recommended_fix": "Break the walk when the chain list is exhausted and cap SGEs walked per frame (see fix.diff)",
  "fix_status": "not_testable",
  "fix_baseline_reproduced": 0,
  "fix_patched_reproduced": 0,
  "fix_verdict": "mechanical guards (NULL break + iteration caps); cannot be boot-validated without mpr hardware in the guest",
  "fix_evidence": "findings/poc/DF-4789/fix.diff"
}