Nov 4 19:27:18 news kernel: EXT2-fs error (device 09:00): ext2_add_entry: bad entry in directory #1863681: rec_len % 4 != 0 - offset=0, inode=1836020294, rec_len=8250, name_len=105
Nov 4 19:27:18 news kernel: EXT2-fs error (device 09:00): ext2_add_entry: bad entry in directory #1863681: rec_len % 4 != 0 - offset=0, inode=1836020294, rec_len=8250, name_len=105
the device in question is a RAID 5 array (using MD 19981005 &
raidtools 0.90 1005). the fs was created with 2048 bytes per inode.
I saw other errors when using a "normal" disk w/ the same settings,
but unfortunately the logs have been purged already, so I'm wondering
if there are any known errors w/ ext2 and non-4k/inode sizes?
What happens if you run unmount the filesystem and run e2fsck? If you
don't find any errors, then there's probably a with either your hardware
(flakey cabling being the most likely cause) or a flakey DMA or memory.
This is probably the #1 cause of this kind of error. It's possible that
it's caused by an actual filesystem corruption, but in practice it's
very rarely caused by that.
Especially if you take a look at the inode numbers, which are
ridiculously large. If you translate them into hex, you get:
6D6F7246 or in ASCII (LSB first): "From"
This seems to be caused by the wrong block getting loaded into memory,
so that block containing text is getting loaded instead of the actual
correct directory block. My theory has always been a bit getting
dropped when the sector number is sent to the disk controller. It's
possible that there's a race condition or bug in the buffer cache code
or the MD driver, but it happens to very few people --- and those people
who do see, report it that it happens fairly frequently to them. That
tends to indicate that it's a hardware problem.
- Ted
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/