On Mon, Jan 21, 2002 at 04:57:34PM -0600, Skylar Thompson wrote: > On Mon, Jan 21, 2002 at 01:13:42PM -0500, Theodore Tso wrote: > > Try booting for a rescue floppy, and then run badblocks on the disk. > > If you see media errors reported in that case, then you'll know what's > > up. > > I tried that, but still a no-go. Basically, running fsck two or three times > will clear all the errors except for the I/O errors, but upon reboot all > the errors are back in full force. Doing an fsync(2) a few times without > rebooting won't create the errors, though. That sounds very much like a hardware problem, then. I'd have to see the fsck transcript and check to see if the kernel print any messages from the IDE device driver, but your symptoms make me 99% sure that it's a hardware problem. In any case, what I'd suggest that you do ASAP is to find another disk which is at least as big as your disk which is failing, and then do the following command: dd if=/dev/hda of=/dev/hdb bs=1k conv=sync,noerror (replace /dev/hda with your source drive, and /dev/hdb as your destination drive). This will do a disk-to-disk image copy; noerror tells dd not to stop when it hits an error, and sync means that it should replace a block which can't be read with zero-filled block. I suspect then you'll be able to run fsck on the replacement disk, and the filesystem will stay fixed, which will confirm the hardware problem. The other reason why you should do this ASAP is that when disks start failing, there's an tendency for the number of bad blocks to go up exponentially. So if you haven't done backups recently (but that's probably not true; you ***HAVE*** been doing regular backups, RIGHT? :-), you should do the disk-to-disk image copy ASAP, to make sure you reduce the risk that you lose any data. - Ted
