Red Hat Bugzilla – Bug 35072
fsck fails at boot time
Last modified: 2007-04-18 12:32:34 EDT
From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows 98) When forcing a fsck on boot up, or when maximum count is reached and fsck is forced, it always fails while checking root: exited with signal 9 and drops me to the shell. Upon running e2fsck, no errors are found, even checked for bad blocks. I have a big 375M root partition and a small 3.9M /boot on a 408M disk and it always fails around 87% completion. Apparently the process is being killed or timedout at that point. Package versions are: initscripts-5.49-1; e2fsprogs-1.18-16 Reproducible: Always Steps to Reproduce: 1./sbin/shutdown -F -r now 2. 3.
signal 9 is sigkill, which sort of implies the kernel did it. How much memory is in the machine?
If you can reproduce this with the rescue-cd from ftp://ftp.redhat.de/pub/rh-addons/rescue-cd/, then this would still be a problem with current glibc/e2fsprogs. Can you please test this out? Thanks a lot.
The machine is a 486sx with 11,904KB. I will try out the rescue-cd although this machine doesn't have a bootable CD, so I am not sure how to run it. I have tried Tomsrtbt disk and used e2fsck on /dev/hda2 (my root) and it showed no error. I will try installing this disk on a 486DX with 36M to see if it shows the problem and I will let you know.
I decided not to swap the disk to another machine, instead installed 20Mb RAM and tried things out with shutdown -F. On bootup e2fsck checked OK with no problems. Changed back to 12Mb and problem came back. It seems problem is memory related. Decided to try out changing fsck options in rc.sysint. Instead of the completion bar option (-C), I changed it to verbose (-V) and tried again with 12Mb. This time fsck checked OK without the sigkill. Would it be the completion bar option is using up extra memory?
Ok, so not enough RAM seems to be the problem. Thanks, Florian La Roche