There's a problem in e2fsprogs 1.15. Occasionally while doing a parallel file system check during init, fsck.ext2 will return a signal 10 error, and then require a reboot. All the filesystems actually check out alright, and on reboot sometimes this will happen again, sometimes it won't. The following is an e-mail from Ted Tso (e2fsprogs maintainer/author) explaining the problem. He also sent me an e-mail today stating that it'll be fixed in a new version to be released ASAP. ----------------------- start pasted e-mail --------------- OK, if you're using e2fsck 1.15, I think I know what's going on. There's a race condition where fsck starts one fsck.ext2 to start just after another fsck finshes, and fsck sends a SIGUSR1 to the first fsck as a signal to tell it that it can start displaying its progress bar. If this happens before the first fsck has had a chance to set up the signal handler, then it will exit when it receives the SIGUSR1. It's a pretty narrow window, which must be why I haven't seen it. I'll put in some safeguards to prevent it from happening, though. If you aren't using the 1.15 e2fsck, and you're not seeing progress bars when fsck needs to check a filesystem, then it's something else. Let me know, and I'll see if I can find another explanation for what's going on. ------------- end pasted e-mail. ------------------- ------- Additional Comments From 10/19/99 11:04 ------- This came up on cartman-list over the weekend. The suggested work around from Bill Nottingham for the time being was to remove the -C from the invocation of fsck in /etc/rc.d/rc.sysinit so that they do not run in parallel.
I have exactly this problem on my system, with the wrinke that this only happens for my largest partition (4.9 gigs or so) and only if I am bringing the system up single-user. Needless to say, this is quite a disconcerting bug and I hope that a fix is issued promptly.
Actually, the -C says not to show progress bars. They still do run in parallel,so removing the -C isn't that bad.
Fixed in the latest e2fsprogs.