Date: Wed, 10 Dec 1997 14:05:19 -0800 From: Julian Elischer <julian@whistle.com> To: Jeremy Nelson <jnelson@acronet.net> Cc: freebsd-questions@FreeBSD.ORG Subject: Re: Figuring out why my filesystems spontaneously broke Message-ID: <348F121F.28CC1042@whistle.com> References: <199712101713.LAA04378@nemesis.acronet.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Jeremy Nelson wrote: > > After doing a non-component related hardware maintanance (replaced the > fan in my power supply), i booted up the machine. It booted ok, I logged > in, started X windows and pppd. At an xterm, i 'cd'd into a subdirectory, > and did an 'ls'. The machine froze for about 15 seconds and rebooted. That's probably while the disk was having a hernia. what is your layout? (disklabel/fdisk) output.. > > So i move the harddrive to another machine that i know is working ok. > The harddrive appears to have lost its primary superblock. No problem, i'll > use a backup superblock. Fsck emits hundreds of DUP and BAD errors, and > a few various errors about "EXCESSIVE DUP BLKS", "INCORRECT BLOCK COUNT", > After making it through pass 1, pass 2 says that a few first-level directories > are empty. It then dies when it cant find an inode. Both the / partition > (partition a) and the /usr partition (partition e) have the exact same damage > signature. sounds like one of: (in increasing order of likelyhood) 1/ the drive remapped a whole bunch of blocks 2/ the drive FORGOT that it had remapped a whole bunch of blocks. 3/ the drive was writing bad data on every write you asked it to do. > > All of the file data is still there ('dd' has no problem finding it on the > raw devices), and disklabel(8) shows that the disk label is still intact. > Since *both* of the fs's went down and since all the data still seems to > be there, that would indicate that it wasnt a hardware failure. Nevertheless, > as someone on #freebsd suggested, my filesystems are toast. > > This then is a two part question: > 1) Is there any "easy" way to recover the data without having to resort to > my backups which i have to confess are probably a bit older than they > should be? we can maybe do some tests using un-used sectors and see if you can write to them. (simple C program.. open raw disk lseek to free block (e.g. block 2?) write pattern.. close device, check if pattern is on device.. > 2) What might have happened? Is there something i can do to avoid having > it happen again? no idea.. > > System specifics: > * 486/66, 40 megs of ram, a Conner CFS1621A IDE drive, generally the rest of > my system componants are unspectacular (built from parts). > * FreeBSD 2.2.2 (built from CD sources) The machine has been in contiuous > service since 1994 through a variety of harddrives and systems. This > particular harddrive has been running FreeBSD happily for about 18 months > now. I do not mount my drives async. > > Thanks to anyone who might be willing to offer help or suggestions. > Thanks to everyone for their hard work. I (and lots of others) appreciate it. > > Jeremy Nelson
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?348F121F.28CC1042>