How is supposed to be protected the units list?

[Matthew Jacob]

On 3/3/2010 1:57 PM, Attilio Rao wrote:
> 2010/3/3 Matthew Jacob<mj at feral.com>:
>    
>> On static review, the only code that makes me nervous are
>> ata_shutdown/da_shutdown.
>> Those are the only places where you hold that lock across an uncertain
>> interval.
>>      
> Please note that a def mutex is already held (the cam_periph_lock),
> so, unless LORs I'm not thinking about, I don't expect too much
> surprises for that codepath.
>
> Thanks,
> Attilio
>
>
>    
The only potential operational difference was on reattach (power SAN 
disks on, they attach. Power them off. Wait for 60 seconds. The 
deattach. Power them on again). I've run this test many times before and 
haven't seen this.

da7: 34732MB (71132959 512 byte sectors: 255H 63S/T 4427C)
(da7:isp0:0:6:0): removing device entry
panic: Bad link elm 0xffffff0018fbbc00 next->prev != elm
cpuid = 0
KDB: enter: panic
[ thread pid 0 tid 100044 ]
Stopped at      kdb_enter+0x3d: movq    $0,0x6b02d0(%rip)
db> bt
Tracing pid 0 tid 100044 td 0xffffff0002fbe000
kdb_enter() at kdb_enter+0x3d
panic() at panic+0x17b
camperiphfree() at camperiphfree+0x1c2
cam_periph_release_locked() at cam_periph_release_locked+0x48
cam_periph_release() at cam_periph_release+0x53
dasysctlinit() at dasysctlinit+0x153
taskqueue_run() at taskqueue_run+0x91
taskqueue_thread_loop() at taskqueue_thread_loop+0x3f
fork_exit() at fork_exit+0x12a
fork_trampoline() at fork_trampoline+0xe
--- trap 0, rip = 0, rsp = 0xffffff8000188d30, rbp = 0 ---