Josh, I haven't forgotten about you. I've been looking at this problem off and on and I think I might have figured out what caused the problem.
It looks like what should have been a list of data extents in an inode had been overridden by the value of a symlink. After trying to figure out how this could have been done during runtime, I remembered a problem I had fixed in the journal-replay code in fsck. If your system had crashed or been been abnormally shut down prior to seeing this problem, it was probably caused by older metadata in the journal overwriting newer metadata when running fsck. The problem is fixed in jfsutils-1.1.5, but I see from the fsck output that you are running version 1.1.3. The errors were reported against a file that was being deleted, so running fsck after the fact did not find any additional problems. Your file system is probably okay now, but updating jfsutils will prevent this problem from being seen again (if I'm correct in my assumptions). Thanks for reporting the problem. Shaggy On Mon, 2004-03-29 at 03:31, Josh Jore wrote: > I recently encountered some errors from my JFS file system and thought > I'd consult this list for any experience on it. Should I change my JFS > version or something? Rebuild my file system? Not use this version of > JFS? > > I built the 2.6.3 kernel.org image, probably upgraded the JFS in the > kernel with a download from the IBM web page, also updated the > jfs-tools package. Some issues related to my sound card led me to > upgrade my kernel to 2.6.4-rc1 and for sanity's sake had me upgrading > to the final 2.6.4 kernel. After the upgrade some files in /etc were > found to be corrupted. > > I don't know when or why the corruption happened or if fsck.jfs is > correct that things are now correct. Why might this have gone bad? As > far as I'm aware, there isn't an hardware problem. Its all stored on a > new serial ATA drive which is why I'm using the 2.6 kernel. > > I've included all the potential debugging information that I could dig > up. I have a dd copy of the partition on which I can hack if that's > something that desired. > > Josh Jore > > The transcript from jfs_logdump are available in a 280K gzipped text > file at http://lik.grenekatz.org/jfs/jfslog.dmp.gz. > > The syslog errors: > > blkno = 65616b656e, nblocks = 72672e > ERROR: (device hde2): dbUpdatePMap: blocks are outside the map > ERROR: (device hde2): remounting filesystem as read-only > blkno = 3a30383031, nblocks = 383439 > ERROR: (device hde2): dbUpdatePMap: blocks are outside the map > blkno = 65616b656e, nblocks = 72672e > ERROR: (device hde2): dbFree: block to be freed is outside the map > blkno = 3a30383031, nblocks = 383439 > ERROR: (device hde2): dbFree: block to be freed is outside the map -- David Kleikamp IBM Linux Technology Center _______________________________________________ Jfs-discussion mailing list [EMAIL PROTECTED] http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion
