bug in fs/xfs/xfs_aops.c:959! (kernel 3.1.5)

Jeremy Sanders jeremy at jeremysanders.net
Tue Dec 20 03:54:42 CST 2011


Dave Chinner wrote:

 > There was a line above this about an assert failure? What was it?

Unfortunately there's are no other messages above this in 
/var/log/messages. I checked carefully for any asserts.

 > If it tripped the assert I think it did, it looks like the system
 > was trying to do write IO into a hole in the file. What workload
 > were you running, and what are the details of your machine + storage
 > config?

It's xfs (nobarrier opts) running from a Linux MD software raid device 
(no LVM). It uses raid6 on 11 1TB discs, using a chunk size of 32 kB. 
The filesystem is 88% full. The discs are on a 3ware 9650SE controller.

The array has a set of backup "snapshots" with hard links created using 
the rsync --link-dest feature. We also used --sparse when doing the 
rsyncs. We have snapshots for a set of user directories, one for each day.

The message occured while rsync was writing to the array (creating the 
snapshots) and it was being read by rsync or a tar to tar pipe to a 
second btrfs array.

Jeremy




More information about the xfs mailing list