On Monday 07 October 2002 04:34, Bernhard Mogens Ege wrote: > Running kernel 2.4.19 patched with JFS-1.0.23, I just saw the > following in hour log file: > > Oct 6 04:02:22 boston kernel: diRead: i_ino != di_number > Oct 7 01:02:31 boston kernel: diRead: i_ino != di_number
This warning means that when reading an inode from disk, the inode was found to be invalid. JFS should have marked the partition dirty, so the next time fsck is run it will do a full check. > I am not sure what to read into this, or if this even is a JFS bug of > some sort. I can tell that prior to these messages, no kernel error > has been logged since reboot (at which a full fsck on the jfs > partition was done). Also, at 04:00, the locate database is updated > and at 01:00 the backup is started, both creating some load on the > filesystem. System uptime right now is 6 days. I can't guess how and when the damage was done, but both slocate and the backup probably only discovered the bad inode. I don't think they are the source of the problem. We have been getting a few reports of this problem, and I think it is related to SMP, but it's a tough one to debug. I'm wondering if the problem found by CONFIG_PREEMPT in the 2.5 kernel could be causing this on SMP systems? I posted the patch last week to this list. I can resend it if you'd like to give it a try. Thanks, Shaggy -- David Kleikamp IBM Linux Technology Center _______________________________________________ Jfs-discussion mailing list [EMAIL PROTECTED] http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion
