Re: deadlkres: possible deadlock detected for 0xe000000012aed200, blocked for 900014 ticks

From: Anton Shterenlikht <mexas_at_bris.ac.uk>
Date: Thu, 24 Oct 2013 11:09:24 +0100 (BST)
>From jhb_at_freebsd.org Wed Oct 23 21:10:30 2013
>> >> 
>> >> http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/183007
>> >
>> >Hmm, unfortunately it seems like all the stack traces did not work.  There 
>are 
>> >lots of threads blocked on VM-related locks, and CPU 0 is running 
>vm_daemon.
>> >Probably would need a stack trace of that thread to see what it is doing 
>(this
>> >is part of why a real crash dump would be far better than a textdump as you 
>> >can get more info after the crash instead of having to know in advance 
>> >everything you want).  I saw earlier you had a thread to get textdumps to 
>> >work.  Did you ever have regular crashdumps working?
>> 
>> No. I have another deadlock panic when savecore
>> is running. That panic dumps core, then on reboot
>> when savecore is running I get another deadlock
>> panic, and so on. However, I do get about 8-10gb vmcore*
>> files before the panic happens. Presumably these
>> are truncated somehow.
>> Do you think these incomplete vmcores can still
>> be useful? If you send some instructions on what
>> to try, I'll give it a go, or I can post a vmcore
>> somewhere.
>
>Does you get corresponding core.txt.N files from crashinfo?

yes:

# ls -al /var/crash/*6*
-rw-------  1 root  wheel       27206 Oct 24 09:57 /var/crash/core.txt.6
-rw-------  1 root  wheel         543 Oct 24 09:26 /var/crash/info.6
-rw-------  1 root  wheel  5853138944 Oct 24 09:31 /var/crash/vmcore.6
# 

I added it to the PR:

http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/183007

Is this any use?

Also, since the panic is perfectly reproducible,
I can cause it again, and run some commands in ddb,
if you give explicit instructions.

Many thanks

Anton
Received on Thu Oct 24 2013 - 08:09:56 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:43 UTC