⬢ DragonFlyBSD Kernel Audit
DF-5734 / run.log
← back to finding ↓ download raw
== DF-5734 run: Sat Oct 10 22:38:06 UTC 2026 ==
== guest: DragonFly 6.5-DEVELOPMENT X86_64_GENERIC ==
== trigger: sysctl hw.hcih_fire_numcompl=1 after hcih_setup+hcih_arm_pdu ==
== EXPECTED: console silence after injection marker + guest wedge (no panic) ==

--- serial console (decisive tail) ---
[harness] <- RssiResult(16) (244 bytes): 04 22 e1 10 e0 17 c0 c1 c2 c3 01 00 00 22 22 22 11 11 c0 e1 17 c0 c1 c2 c3 01 00 00 22 22 22 11 ...
[harness] <- RssiResult(16) (244 bytes): 04 22 e1 10 f0 17 c0 c1 c2 c3 01 00 00 22 22 22 11 11 c0 f1 17 c0 c1 c2 c3 01 00 00 22 22 22 11 ...
[harness] memos after: 3184 (allocated 336 for 2048 responses, none reclaimed: expiry reaps only on hci_memo_find)
[harness] pre-fire state: txq PDUs nonempty, head lp_pending=0, head lp_data qlen=1 (stuck-loop precondition: pending==0 && qlen>0)
[harness] === DF-5734: Num_Compl_Pkts(handle=5, num=1) on link with queued-but-unsent PDU ===
[harness] EXPECT: netisr btintr() -> hci_intr() -> hci_event_num_compl_pkts() -> hci_acl_complete() spins forever (no progress: head PDU has lp_pending=0 + data queued)
[harness] <- Num_Compl_Pkts (8 bytes): 04 13 05 01 05 00 01 00
[harness] injected; if btintr does not return, the console goes quiet here and the CPU burns in hci_acl_complete while holding the mplock

--- post-trigger probes ---
run_root probe 1 (40s): 'Connection timed out during banner exchange'
run_root probe 2 (30s): 'Connection timed out during banner exchange'
vm.sh status: down