{
  "finding_id": "DF-4637",
  "status": "untested",
  "reproduced": 0,
  "impact": "Multi-second uninterruptible CPU pin (preemption disabled, interrupts deferred) per I/O against a malicious/faulty parallel-port ZIP emulator; repeatable per command \u2014 local/physical DoS (DF-4618 family, vpio instance)",
  "confidence": "likely",
  "verdict": "not_run: no ppbus/ppc/lpt hardware on the audit guest (env.txt: dmesg/pciconf empty, vpo.ko present but unattachable); mechanism source-certain \u2014 vpoio_do_scsis wait/select handshakes are MS_BRSET/MS_DBRA no-sleep port-I/O spins (ppc.c:1447-1458) with budgets of 500000/5000000 iterations (vpoio.c:59-60), executed at vpoio.c:705/716/733/774 entirely inside vpo_intrs crit_enter/crit_exit (vpo.c:215/323, autosense round vpo.c:278-284 included). Worst case ~13M status-register polls per command under one critical section. Follows family convention of DF-4618/DF-4508.",
  "exploit_chain": [],
  "evidence": "VERDICT.md (line-by-line trace: vpo.c:215 crit_enter; vpoio.c:59-60 ceilings; vpoio.c:103-113 wait_microseq; ppc.c:1447-1451 DBRA + 1453-1458 BRSET = raw spin; call sites vpoio.c:716/733/774); trigger.sh documents the hardware trigger; env.txt documents guest hardware absence; fix.diff = 10x ceiling cut (short-term family mitigation)",
  "kernel_refs": [
    "sys/dev/disk/vpo/vpoio.c:59",
    "sys/dev/disk/vpo/vpoio.c:60",
    "sys/dev/disk/vpo/vpoio.c:525",
    "sys/dev/disk/vpo/vpoio.c:716",
    "sys/dev/disk/vpo/vpoio.c:733",
    "sys/dev/disk/vpo/vpoio.c:774",
    "sys/dev/disk/vpo/vpo.c:215",
    "sys/dev/misc/ppc/ppc.c:1447",
    "sys/dev/misc/ppc/ppc.c:1453"
  ],
  "poc_changes": "n/a \u2014 hardware-gated; trigger.sh documents the setup, never executed on guest",
  "attempts": 0,
  "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": 0,
  "guest_dirty": 0,
  "notes": "vpoio.c instance of the DF-4618 family (immio.c). vpo_intrs crit_enter covers both imm_do_scsi and vpoio_do_scsi; this finding covers vpoios own spin sites (select VP0_SELTMO, cmd VP0_FAST_SPINTMO, data VP0_LOW_SPINTMO, status VP0_FAST_SPINTMO).",
  "recommended_fix": "Short-term: cut spin ceilings 10x (fix.diff). Proper: remove crit_enter/crit_exit around the synchronous transaction in vpo_intr (vpo.c) \u2014 ppbus ownership already serializes \u2014 and/or make vpoio_wait() interruptible via ppb_poll_device() (stub at vpoio.c:533-538).",
  "fix_status": "not_testable",
  "fix_baseline_reproduced": 0,
  "fix_patched_reproduced": 0,
  "fix_verdict": "not_testable: no parallel-port hardware on guest to baseline or patch-test; fix.diff is the familys short-term 10x ceiling cut, mirrors DF-4618 guidance, trivially reviewable",
  "fix_evidence": "fix.diff"
}