View previous topic :: View next topic |
Author |
Message |
yoshi314 l33t
Joined: 30 Dec 2004 Posts: 850 Location: PL
|
Posted: Sun Jan 21, 2007 8:40 pm Post subject: reiserfs data corruption with gentoo-sources 2.6.19-r5 |
|
|
well, as you know there was an infamous ext3-data-corruption in 2.6.19 kernel, esp, visible with rtorrent or other mmap heavy apps.
i though it was supposed to be fixed in 2.6.19.2 patch, which was included into gentoo-sources 2.6.19-r5 patchset, right?
i'm using this kernel right now, and i'm getting corruption on 220GB reiserfs (v3) partition.
the effect of corruption appears with rtorrent, and it's considerably less serious than with 2.6.19 vanilla kernel (i used to get 30% of broken chunks, now i get ~5 bad chunks). also files that seem to be 100% correct (in rtorrent) appear to have bad chunks after a few hours from downloading - i reverify them with rtorrent and some chunks appear incorrect
everything works perfectly, with no errors on 2.6.18 kernel (gentoo-sources 2.6.18-r5)
i'd like to know - is the reiserfs corruption a well known issue, and is there a fix for it? (since all the buzz was about ext3 errors)
i already ran thorough fsck on the partition and memtest on my ram. no errors.
can anyone reproduce/verify this problem? maybe there is something odd with my hardware? _________________ ~amd64
shrink your /usr/portage with squashfs+aufs |
|
Back to top |
|
|
volkris n00b
Joined: 26 May 2002 Posts: 36
|
Posted: Mon Feb 05, 2007 6:08 am Post subject: |
|
|
I am running rtorrent on reiserfs (3) and am seeing the same things.
(for what THAT's worth...) |
|
Back to top |
|
|
yoshi314 l33t
Joined: 30 Dec 2004 Posts: 850 Location: PL
|
Posted: Mon Feb 05, 2007 11:46 am Post subject: |
|
|
the funny thing is - this very same patchset, rebuilt for amd64 doesn't experience any corruptions. what's going on? _________________ ~amd64
shrink your /usr/portage with squashfs+aufs |
|
Back to top |
|
|
volkris n00b
Joined: 26 May 2002 Posts: 36
|
Posted: Mon Feb 05, 2007 2:38 pm Post subject: |
|
|
It looks like this is now a know problem that stumped the kernel devs for a good while. Sounds like it's been fixed in 2.6.20.
Until then I hear you can mount with the option data=writeback. Trying out this workaround myself I believe it's successful.
Sorry for the ambiguities |
|
Back to top |
|
|
|