[PATCH][SRU][precise/quantal] bug #1056746 kernel thread hang on iscsi target disconnect

Louis Bouchard caribou at canonical.com
Fri Oct 19 10:48:56 UTC 2012


SRU justification:

Impact:
Shutting down a server when iscsi targets are active and multipath is in use will result in a hang of the shutdown process and kernel threads locked. It can also be triggered by a simple "iscsiadm -m node --logout".

Fix:
Cherry-picking commit bc3f02a795d3b4faa99d37390174be2a75d091bd fix this issue identified upstream.

Testcase:

Run a VM with two network interfaces connecting to an iscsi target. Enable multipath for those devices and execute :

 $ iscsiadm -m node --logout

After a few seconds, the console will start to display the following :

[   56.060031] BUG: soft lockup - CPU#1 stuck for 23s! [kworker/u:3:44]
[   56.060997] Stack:
[   56.061276] Call Trace:
[   56.061654] Code: 89 df e8 ab 12 ff ff 49 8b 7d 58 e8 22 93 22 00 49 89 c6 eb 89 0f 1f 44 00 00 49 8b 7d 58 4c 89 f6 e8 9c 92 22 00 5b 41 5c 41 5d <41> 5e 5d c3 0f 1f 00 45 31 ed e9 58 ff ff ff 0f 1f 84 00 00 00 

BugLink: http://bugs.launchpad.net/bugs/1056746

Thanks,
..Louis Bouchard




More information about the kernel-team mailing list