RELENG_6 Livelock

[Frode Nordahl]

On 17. sep. 2006, at 04.42, John Baldwin wrote:

> On Saturday 16 September 2006 16:55, Frode Nordahl wrote:
>> On 16. sep. 2006, at 22.22, Frode Nordahl wrote:
>>
>>> On 16. sep. 2006, at 22.09, John Baldwin wrote:
>>>
>>>> On Saturday 16 September 2006 07:02, Frode Nordahl wrote:
>>>>> Hello,
>>>>>
>>>>> FreeBSD localhost.localdomain 6.2-PRERELEASE FreeBSD 6.2- 
>>>>> PRERELEASE
>>>>> #1: Wed Sep 13 00:10:04 CEST 2006      
>>>>> frode at localhost.localdomain:/
>>>>> usr/obj/usr/src/sys/PT  i386
>>>>>
>>>>> After running some stress tests for 3 days, I wanted to remove  
>>>>> some
>>>>> large directories.
>>>>
>>>> Do you have a coredump?  I assume you do from your debug  
>>>> output.  Can
>>>> you download http://www.FreeBSD.org/~jhb/gdb/gdb6, fire up kgdb,  
>>>> and
>>>> once in kgdb, do 'source /path/to/gdb6' and then run 'ps' and reply
>>>> with the output from that?
>>>
>>> I am sorry, I have not. I tried to call doadump, but there was no
>>> dumpdevice configured :-(
>>>
>>> Somehow I have convinced myself that this was turned on by default
>>> now, so I have not enabled it explicitly in rc.conf. Is there any
>>> way to tell DDB what dumpdevice to use directly?
>>>
>>> I will configure a dumpdevice and try really hard to make it happen
>>> again.
>>
>> I was able to reproduce the livelock again, and this time I had the
>> system armed with dumpon :-)
>>
>> Here is the output you requested:
>> (kgdb) ps
>>    pid  ppid  pgrp   uid   state   wmesg     wchan    cmd
>> 2535  2499  2535     0  R+      CPU  0              rm
>> 2534  2499  2534     0  L+     *Giant    0xc6704580 rm
>> 2533  2499  2533     0  L+     *Giant    0xc6704580 rm
>> 2532  2499  2532     0  R+                          rm
>> 2531  2499  2531     0  L+     *Giant    0xc6704580 rm
>> 2499  2496  2499     0  Ss+     ttyin    0xc655d810 bash
>> 2496   784  2496     0  Rs                          sshd
>
> Ok, do 'lockchain 2534' in kgdb (with gdb6 sourced) and let me see the
> output from that.

(kgdb) lockchain 2534
thread 100038 (pid 2534, rm) blocked on lock 0xc09e6800 "Giant"
thread 100091 (pid 2535, rm) running on CPU 0

--
Frode Nordahl