5.4 panic at lockmgr

Kachun Lee kachun at pathlink.com
Mon May 23 10:18:20 PDT 2005

"Kris Kennaway" <kris at obsecurity.org> wrote in message 
news:<20050521063233.GA51618 at xor.obsecurity.org>...
 >On Fri, May 20, 2005 at 04:39:19PM -0700, Kachun Lee wrote:
 >> I upgraded a server from 4.10 to 5.4-rel and it paniced at soon as I 
put >> some load on it...
 >> panic: lockmgr: thread 0xfffffffe, not exclusive lock holder 
0x76fe0300 >> unlocking
 >> The hardware is a Intel 7501 with a single 3G Xeon 6G mem. I tried >> 
disable option MP and made no difference. Any help or suggestion!?

 > Run 'show lockedvnods' and then 'trace <pid>' for each pid listed.
 > Also, if you can explain how to reproduce this it would be greatly
 > helpful.
 > Kris

Thanks for the info. I will try 5.4 on that server again on this Thur/Fri 
(I am not in the office for these few days).

It was hard to tell what really triggered that. The server is one of our 
NFS servers. With that first panic, I did not notice it right away. The 2nd 
panic, I had a top running and was staring at it since reboot. When it 
panic'ed (at ~6 minute uptime), the load climbed to only around 2-3. There 
is nothing weird showing on top. Since this is one of our production 
servers, I could not do too much experiment with it.

I had set up a similar system with the same MB/CPU on the bench. And it ran 
the same 5.4-rel kernel, with some benchmark programs, over the weekend 
without problem. However, I do not have a way to simulate all the network 
traffic to that box. If you have any idea what I can try to isolate the 
problem a little more, let me know.


More information about the freebsd-stable mailing list