On Apr 14, 2008 07:50 -0700, Mingming Cao wrote: > On Sat, 2008-04-12 at 22:57 +0200, Andi Kleen wrote: > > FYI, a system here running various 2.6.25rc kernels (latest upto rc7-git6) > > with longer uptimes suddenly decided to fsck one of its file systems > > due to an error after reboot. > > > > The error causing this was: > > > > kernel: EXT3-fs error (device dm-0): ext3_valid_block_bitmap: Invalid block bitmap - block_group = 285, block = 9338882 > > > > detected by the 2.6.25rc7-git6 kernel. > > > > I don't see any ill effects from it and fsck didn't find anything wrong > > so it must have been something spurious in memory only (or fsck > > fails to check for this condition, but that is hard to imagine) > > The ext3_valid_block_bitmap() is to check whether the block or inode > bitmap block is marked as "used" in the block group bitmap, to prevent > allocating blocks from these system meta data blocks. Right. > The error messages seems indicating that one of the block group meta > data is corrupted, but I don't why fsck doesn't catch this, Andreas? It might have been corrupted on read (e.g. bad cable, or bad/wrong data read from disk the first time). The message itself isn't very useful though. It should report what it thinks is wrong with the bitmap (e.g. whether block/inode bitmaps are unallocated, which/how many itable blocks are unallocated). > Mingming > > The system never showed anything like this on earlier kernel versions. This is a new check, to catch allocation bitmap corruption before it causes the corruption to spread into the rest of the filesystem by double-allocating blocks, etc. Having a checksum would also be good, but even then memory corruption can lead to a valid checksum of bad data in memory so a validity check is still useful for such important and rarely-read data. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
