FS#24480 - [kernel26] Sudden failure of process management in Kernel 2.6.38

Attached to Project: Arch Linux
Opened by Roy Zuo (roylez) - Saturday, 28 May 2011, 03:02 GMT
Last edited by Tobias Powalowski (tpowa) - Thursday, 16 February 2012, 17:58 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Tobias Powalowski (tpowa)
Thomas Bächler (brain0)
Architecture i686
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

I do not know how and why this happens. When I try to kill my frozen firefox, pkill command hangs, then I find every process related command hangs, including kill, ps, top and even w. However, at the same time, I am still chatting on IRC without any problem and other commands, e.g. ifconfig, do not have any problem. The computer cannot be properly rebooted as well.

Here I attach dmesg output captured. It contains some stack trace that can be relevant.


   dmesg.log (334.2 KiB)
This task depends upon

Closed by  Tobias Powalowski (tpowa)
Thursday, 16 February 2012, 17:58 GMT
Reason for closing:  Fixed
Comment by Jelle van der Waa (jelly) - Saturday, 28 May 2011, 10:30 GMT
which exact kernel are you using?
pacman -Q kernel26
Comment by Roy Zuo (roylez) - Saturday, 28 May 2011, 10:39 GMT
It is 2.6.38.7-1 .
Comment by Gerardo Exequiel Pozzi (djgera) - Saturday, 28 May 2011, 13:36 GMT
Are these tasks in "D" state? If true then try disabling transparent hugepages, its causes some problems at this moment.
echo never > /sys/kernel/mm/transparent_hugepage/enabled
echo never > /sys/kernel/mm/transparent_hugepage/defrag
Comment by Roy Zuo (roylez) - Saturday, 28 May 2011, 13:45 GMT
Hi djgera,

I really cannot tell what were the status of those jobs when it happened. Every process related command hung at that moment. Also, it is not very likely that I could recreate it. I just hope someone can get some hint out of the dmesg log. The part that I believe is useful is from line 4278 to the end.
Comment by Gerardo Exequiel Pozzi (djgera) - Saturday, 28 May 2011, 13:53 GMT
According to the dmesg.log, these task are in D state, so maybe THP was the problem.

Loading...