⬢ DragonFlyBSD Kernel Audit
DF-3818 / verdict.json
← back to finding ↓ download raw
{
  "finding_id": "DF-3818",
  "status": "inconclusive",
  "reproduced": 0,
  "impact": "none",
  "confidence": "speculative",
  "verdict": "Code defect certain: the PRP pool is fixed at NVME_QDEPTH=2048 pages (mpr.c:1593-1631) while the command pool reaches 2176 (mpr.c:375-381); mpr.c:2988-2991 (I/O path via mpr_data_cb mpr.c:3483-3487) and mpr.c:2716-2720 (ioctl path) dereference mpr_alloc_prp_page() NULL guarded only by KASSERT — panic on INVARIANTS kernels (X86_64_GENERIC ships INVARIANTS) or NULL-offset page fault otherwise. Achievability of >=2049 concurrent native-SGL NVMe I/Os from an unprivileged user is plausible but unproven (needs NVMe-behind-SAS3 hardware). Runtime inconclusive: guest has no such hardware (probe.log).",
  "exploit_chain": "",
  "evidence": "VERDICT.md (pool-vs-demand arithmetic); fix.diff (IEEE-SGL fallback / early return).",
  "kernel_refs": ["sys/dev/raid/mpr/mpr.c:2988", "sys/dev/raid/mpr/mpr.c:2716"],
  "poc_changes": "No runnable PoC without hardware.",
  "attempts": 1,
  "guest_uname": "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",
  "runtime_sec": 5,
  "guest_dirty": 0,
  "build_cmd": "",
  "run_cmd": "",
  "notes": "Pure DoS, no corruption primitive (fixed NULL deref). Related: DF-1227 covered the separate unbounded-length overflow in mpr_build_nvme_prp.",
  "recommended_fix": "Handle mpr_alloc_prp_page() == NULL: fall back to IEEE SGL in mpr_check_pcie_native_sgl(), fail gracefully in mpr_build_nvme_prp(); long term make pool >= num_reqs.",
  "fix_status": "not_testable",
  "fix_baseline_reproduced": 0,
  "fix_patched_reproduced": 0,
  "fix_verdict": "No NVMe-behind-SAS3 hardware on the guest; the PRP path cannot be driven.",
  "fix_evidence": "fix.diff; probe.log."
}