File type is occasionally wrong

Tue Jan 6 08:23:31 CST 2015

On Tue, Jan 06, 2015 at 02:56:08PM +0100, Jan Kara wrote:
>   Hello,
> 
>   I've got two reports of XFS filesystem corruption where file type is
> wrong. Both machines are running fairly recent kernels (currently
> 3.19-rc2) but it's hard to say when the corruption started to happen since
> it isn't easily observable.
> 

No theories off the top of my head, but is some kind of unclean shutdown
involved? I ask because the repair output also warns of populated agi
unlinked buckets. IIRC, that should all be cleaned up on umount.

Brian

> Attached is xfs_repair log. The interesting part there is:
> would fix ftype mismatch (7/1) in directory/child inode 805307401/805388292
> would fix ftype mismatch (1/7) in directory/child inode 805307401/805388295
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939541565
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939540212
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939539662
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939525162
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939525165
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939529260
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939540177
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939539749
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939529266
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939541566
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540205
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540211
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939525182
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540269
> 
> So it seems that the file type between a regular file and a symlink gets
> swapped. Since both directories where corruption happens are relatively
> large (13 and 9 blocks respectively - they are /usr/bin and
> /usr/share/man/man8) it seems as if file type wasn't properly updated when
> the directory entry tree is updated or something like that. I have briefly
> looked into the code but from a first look everything looks fine there...
> 
> I have also metadump without obfuscated names available so I can
> provide that on request. Interestingly, in the second directory all the
> directory entries for regular files that have file type set as symlink are
> in the first directory block while inconsistency in other direction is in
> other directory blocks.
> 
> I will be looking into this further but if some more knowledgeable has an
> idea it would be welcome.
> 
> 								Honza
> 
> -- 
> Jan Kara <jack at suse.cz>
> SUSE Labs, CR

> Phase 1 - find and verify superblock...
>         - reporting progress in intervals of 15 minutes
> Phase 2 - using internal log
>         - scan filesystem freespace and inode maps...
> agi unlinked bucket 47 is 73775 in ag 15 (inode=2013339695)
> agi unlinked bucket 49 is 73777 in ag 15 (inode=2013339697)
> agi unlinked bucket 52 is 73780 in ag 15 (inode=2013339700)
> agi unlinked bucket 53 is 73781 in ag 15 (inode=2013339701)
>         - 09:47:30: scanning filesystem freespace - 32 of 32 allocation groups done
>         - found root inode chunk
> Phase 3 - for each AG...
>         - scan (but don't clear) agi unlinked lists...
>         - 09:47:30: scanning agi unlinked lists - 32 of 32 allocation groups done
>         - process known inodes and perform inode discovery...
>         - agno = 0
>         - agno = 30
>         - agno = 15
>         - agno = 16
>         - agno = 1
>         - agno = 31
>         - agno = 17
>         - agno = 2
>         - agno = 18
>         - agno = 3
>         - agno = 19
>         - agno = 4
>         - agno = 20
>         - agno = 5
>         - agno = 21
>         - agno = 22
>         - agno = 6
>         - agno = 23
>         - agno = 7
>         - agno = 24
>         - agno = 8
>         - agno = 9
>         - agno = 25
>         - agno = 10
>         - agno = 26
>         - agno = 11
>         - agno = 27
>         - agno = 28
>         - agno = 12
>         - agno = 29
>         - agno = 13
>         - agno = 14
>         - 09:47:38: process known inodes and inode discovery - 133312 of 133312 inodes done
>         - process newly discovered inodes...
>         - 09:47:38: process newly discovered inodes - 32 of 32 allocation groups done
> Phase 4 - check for duplicate blocks...
>         - setting up duplicate extent list...
>         - 09:47:38: setting up duplicate extent list - 32 of 32 allocation groups done
>         - check for inodes claiming duplicate blocks...
>         - agno = 0
>         - agno = 1
>         - agno = 2
>         - agno = 3
>         - agno = 4
>         - agno = 5
>         - agno = 6
>         - agno = 7
>         - agno = 8
>         - agno = 9
>         - agno = 10
>         - agno = 11
>         - agno = 12
>         - agno = 13
>         - agno = 14
>         - agno = 15
>         - agno = 16
>         - agno = 17
>         - agno = 18
>         - agno = 19
>         - agno = 20
>         - agno = 21
>         - agno = 22
>         - agno = 23
>         - agno = 24
>         - agno = 25
>         - agno = 26
>         - agno = 27
>         - agno = 28
>         - agno = 29
>         - agno = 30
>         - agno = 31
>         - 09:47:38: check for inodes claiming duplicate blocks - 133312 of 133312 inodes done
> No modify flag set, skipping phase 5
> Phase 6 - check inode connectivity...
>         - traversing filesystem ...
> would fix ftype mismatch (7/1) in directory/child inode 805307401/805388292
> would fix ftype mismatch (1/7) in directory/child inode 805307401/805388295
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939541565
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939540212
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939539662
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939525162
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939525165
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939529260
> would fix ftype mismatch (7/1) in directory/child inode 939525153/939540177
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939539749
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939529266
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939541566
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540205
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540211
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939525182
> would fix ftype mismatch (1/7) in directory/child inode 939525153/939540269
>         - traversal finished ...
>         - moving disconnected inodes to lost+found ...
> disconnected inode 2013339695, would move to lost+found
> disconnected inode 2013339697, would move to lost+found
> disconnected inode 2013339700, would move to lost+found
> disconnected inode 2013339701, would move to lost+found
> Phase 7 - verify link counts...
> would have reset inode 2013339695 nlinks from 0 to 1
> would have reset inode 2013339697 nlinks from 0 to 1
> would have reset inode 2013339700 nlinks from 0 to 1
> would have reset inode 2013339701 nlinks from 0 to 1
> No modify flag set, skipping filesystem flush and exiting.

> _______________________________________________
> xfs mailing list
> xfs at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs