Because we do not drain the hdac callout during detach, hot-unloading
can result in a panic if the callback fires after we have freed the
resources it uses.

pcm0: detached
pcm1: detached
hdaa0: detached
hdacc0: detached
Kernel page fault with the following non-sleepable locks held:
exclusive sleep mutex hdac0 (HDA driver mutex) r = 0 (0xfffffe000212c820) locked @ /mnt/src/sys/dev/sound/pci/hda/hdac.c:400

[...]

db_trace_self_wrapper() at db_trace_self_wrapper+0xa5/frame 0xfffffe0046b04490
kdb_backtrace() at kdb_backtrace+0xc6/frame 0xfffffe0046b045f0
vpanic() at vpanic+0x226/frame 0xfffffe0046b04790
panic() at panic+0xb5/frame 0xfffffe0046b04850
trap_fatal() at trap_fatal+0x65b/frame 0xfffffe0046b04950
trap_pfault() at trap_pfault+0x12b/frame 0xfffffe0046b04a70
trap() at trap+0x54c/frame 0xfffffe0046b04c50
calltrap() at calltrap+0x8/frame 0xfffffe0046b04c50

• trap 0xc, rip = 0xffffffff8412e7ec, rsp = 0xfffffe0046b04d20, rbp = 0xfffffe0046b04d40 ---

hdacc_stream_intr() at hdacc_stream_intr+0x3c/frame 0xfffffe0046b04d40
hdac_intr_handler() at hdac_intr_handler+0x15c/frame 0xfffffe0046b04d90
ithread_loop() at ithread_loop+0x387/frame 0xfffffe0046b04ef0
fork_exit() at fork_exit+0xa3/frame 0xfffffe0046b04f30
fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe0046b04f30

• trap 0xc, rip = 0x829cc531a, rsp = 0x82b8b7a88, rbp = 0x82b8b7aa0 ---

KDB: enter: panic
[ thread pid 12 tid 100211 ]
Stopped at kdb_enter+0x34: movq $0,0x1f09af1(%rip)
db>

Sponsored by: The FreeBSD Foundation
MFC after: 1 week