Date: Thu, 25 Apr 2002 14:06:51 -0400 (EDT) From: John Baldwin <jhb@FreeBSD.org> To: Kris Kennaway <kris@obsecurity.org> Cc: current@FreeBSD.org Subject: Re: 4 panics under 5.0 on ports cluster Message-ID: <XFMail.20020425140651.jhb@FreeBSD.org> In-Reply-To: <20020425110133.A48383@xor.obsecurity.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 25-Apr-2002 Kris Kennaway wrote:
> On Thu, Apr 25, 2002 at 01:10:50PM -0400, John Baldwin wrote:
>>
>> On 25-Apr-2002 Kris Kennaway wrote:
>> > panic: sleeping thread owns a mutex
>> > Uptime: 8h52m59s
>> > (kgdb) bt
>> >#0 dumpsys () at /local0/scratch/usr/src/sys/kern/kern_shutdown.c:506
>> >#1 0xc0207cb0 in boot (howto=260) at
>> >#/local0/scratch/usr/src/sys/kern/kern_shutdown.c:338
>> >#2 0xc020814f in panic (fmt=0xc0367580 "sleeping thread owns a mutex")
>> > at /local0/scratch/usr/src/sys/kern/kern_shutdown.c:648
>> >#3 0xc0200a35 in propagate_priority (td=0xcd23d400)
>> > at /local0/scratch/usr/src/sys/kern/kern_mutex.c:110
>>
>> Can you go to this frame, then dump the contents of 'm' to find the
>> misbehaving
>> mutex?
>
> $2 = {mtx_object = {lo_class = 0xc03b1fc0, lo_name = 0xc76785a0 "PCPU 16384",
> lo_flags = 4390912,
> lo_list = {stqe_next = 0x0}, lo_witness = 0x0}, mtx_lock = 3453968642,
> mtx_recurse = 0,
> mtx_blocked = {tqh_first = 0xcd23d400, tqh_last = 0xcd23d428},
> mtx_contested = {le_next = 0x0,
> le_prev = 0xcddf6160}}
Ok, so uma is misbehaving it seems and sleeping while holding a lock.
>> WITNESS should have warned about this but I guess getting people to run
>> WITNESS is too much to ask.
>
> WITNESS disables itself after about 30 seconds under load on the ports
> cluster..something about some resource becoming exhausted. I tried
> increasing the number of whatever resource it was by a factor of 8,
> but it was still occurring after a longer time interval, so I gave up
> and just turned it off.
You have stale kernels. This was fixed a couple of weeks ago with all the
mtx_init() changes. If the kernels were up to date, there would be a lo_type
in struct lock_object above. Try updating to newer kernels, turning WITNESS
and WITNESS_DDB on. Then when uma misbehaves it will drop into ddb.
> Kris
--
John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve!" - http://www.FreeBSD.org/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.20020425140651.jhb>