Stable Linux kernel hit by ext4 data corruption bug
"A bug introduced with version 3.6.2 of the Linux kernel can cause an ext4 filesystem to lose data if the system is rebooted twice in a short period of time. After the bug was reported, kernel developer Ted Ts'o quickly published a patch." -- source: http://h-online.com/-1736110 Cheers, Peter -- Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ http://www.cms.waikato.ac.nz/~fracpete/ Ph. +64 (7) 858-5174
After discussion at the last workshop I think we concluded that there aren't really any advantages to ext4 over ext3. Cairo also had a bit of a problem last year with entire openoffice documents vanishing (truncated to 0 bytes); turned out to be a combination of ext4's very long delayed allocation, Openoffice not using fsync() when it should, and Cairo hitting the reset button to reboot into Windows rather than shutting down properly ... I think from now on I'm going to just use ext3 on all my installs. On 25 October 2012 10:37, Peter Reutemann <fracpete(a)waikato.ac.nz> wrote:
"A bug introduced with version 3.6.2 of the Linux kernel can cause an ext4 filesystem to lose data if the system is rebooted twice in a short period of time. After the bug was reported, kernel developer Ted Ts'o quickly published a patch."
-- source: http://h-online.com/-1736110
Cheers, Peter -- Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ http://www.cms.waikato.ac.nz/~fracpete/ Ph. +64 (7) 858-5174 _______________________________________________ wlug mailing list | wlug(a)list.waikato.ac.nz Unsubscribe: http://list.waikato.ac.nz/mailman/listinfo/wlug
Ext4 is worth it if only for extents (and preallocation, sadly under used thou), and multi-block allocation, to prevent tail fragmentation, ext4 has some pretty big wins, to tide us over, until we have a stable post 90's filesystem. The OODoc is almost certainly a bug in Openoffice/Libreoffice where they have expected the semantics of ext3 data=ordered mode rather then being agnostic. Thou you would assume, that hitting the save button would imply an fsync, autosave I could understand not. And a Lance Armstong bug, that made me actually lol
Looking like a very rare trigger case (from Ted Tso Google +) *Update*: It now looks like the reproduction involved something very esoteric indeed, involving using umount -l and shutdowns while the file system was still being unmounted --- and the user had nobarrier specified in the mount options as well. See: http://thread.gmane.org/gmane.linux.kernel/1379725/focus=34994 and http://thread.gmane.org/gmane.linux.kernel/1379725/focus=35003 and my reply here http://thread.gmane.org/gmane.linux.kernel/1379725/focus=35006
Hmm, perhaps that's a bit misleading. For what I do, I concluded that there wasn't really much reason to switch. There are improvements in the filesystem which other users will find valuable. On 25 October 2012 11:07, Glenn Ramsey <gr(a)componic.co.nz> wrote:
Hi Bruce,
On 25/10/12 10:58, Bruce Kingsbury wrote:
After discussion at the last workshop I think we concluded that there aren't really any advantages to ext4 over ext3.
Could you summarise the reasons for this?
Glenn
______________________________**_________________ wlug mailing list | wlug(a)list.waikato.ac.nz Unsubscribe: http://list.waikato.ac.nz/**mailman/listinfo/wlug<http://list.waikato.ac.nz/mailman/listinfo/wlug>
participants (4)
-
Bruce Kingsbury -
Glenn Ramsey -
Peter Reutemann -
Ronnie Collinson