Hi,

Maybe I'm wrong, but I think the probability of the fault is not kernel
dependent as there have been similar problems on 2.6.28.8.

The error is reproducible after a reboot and mount -i.

Bye,
David Arendt

> Hi David,
>
> On Fri, 27 Mar 2009 05:18:11 +0000, David Arendt <[email protected]> wrote:
>> Hi,
>>
>> There seems to be some bug in the kernel. On another partition
>> reformatted on week ago, I had again the following error:
>>
>> NILFS error (device sda3): nilfs_check_page: bad entry in directory
>> #28261: unaligned directory entry - offset=4096, inode=1647255843,
>> rec_len=29537, name_len=104
>> NILFS error (device sda3): nilfs_check_page: bad entry in directory
>> #28261: unaligned directory entry - offset=4096, inode=1647255843,
>> rec_len=29537, name_len=104
>> NILFS warning (device sda3): nilfs_ifile_get_inode_block: unable to read
>> inode: 42880
> <snip>
>> NILFS warning (device sda3): nilfs_ifile_get_inode_block: unable to read
>> inode: 42910
>> NILFS warning (device sda3): nilfs_ifile_get_inode_block: unable to read
>> inode: 42911
>> init_special_inode: bogus i_mode (35070)
>> init_special_inode: bogus i_mode (30055)
> <snip>
>> init_special_inode: bogus i_mode (36504)
>
> Uum, this time, ifile (i.e. inode index file) seems to be broken.
>
> Do you think probability of the fault depends on the kernel version?
>
> And, is it reproducible after umount(or reboot) and mount -i (= mount
> without GC) ?
>
> We partially succeeded to reproduce corrpution under a near disk full
> condition, and are trying to narrow down the occurrence condition.  I
> now suspect cache coherence violation between GC cache and regular
> page caches, but it's uncorroborated so far.
>
> With regards,
> Ryusuke Konishi
>


_______________________________________________
users mailing list
[email protected]
https://www.nilfs.org/mailman/listinfo/users

Reply via email to