From Bugzilla Helper: User-Agent: Mozilla/4.78 [en] (X11; U; Linux 2.4.7-10enterprise i686) Description of problem: I used to have /dev/sdf1 as a swap partition and mount /dev/sdg1 as /var. Following a disk failure, the fstab was modified to mount sdf1 as /var but the swap entry was not modified to point to a different partition. The init processes mounts / and /var then run kudzu which not finding the swap signature override the on-disk info leading to an nearly immediate crash of the journald running on that partition. Version-Release number of selected component (if applicable): How reproducible: Didn't try Steps to Reproduce: 1. use the same /dev/ device for both and ext3 and swap partition in /etc/fstab 2. reboot 3. enjoy your boot crashing when kudzu get started Actual Results: the system won't boot. The filesystem need an fsck. Expected Results: kudzu to not touch partitions mounted or labeelled as mountable in /etc/fstab Additional info: RedHat 7.1 setup with security updated + kernel and fs updates to run ext3 and TUX (fr.rpmfind.net).
kudzu doesn't touch swap partitions, or any partitions. I believe it's just a delayed crash from it running swapon.
Hum, swapon would check whether the partition is mounted first, isn't it ? I followed the following sequence (from memory): - boot single, notice one disk is dead, change all device name for SCSI entries post the dead drive but forget the swap partition. reboot. - try to boot, mount fs, kudzu, journald crash, reset the box - reboot single, fsck manually the corrupted /var, mount it, ^D, init launch kudzu start, journald crash, reset the box - reboot single, fsck, mount /var, rpm -e kudzu, ^D, system boots normally. I think I did comment off the swap partition only at that point but I don't have a precise log to be 100% sure :-\ Daniel
Kudzu does not mount any filesystems, except for a CD for installing the X server. Kudzu has no code that calls swappon/swapoff/etc. So, I don't see *how* it can do any of this.
FWIW, swapon -a is called *before* the partition would be mounted...
Okay, then I have no idea of what's happened. If the swap had been activated, the mount could certainly not have worked (swap signature and ext3 disk layout to "coexist" seems doubtful). The error was probably mine. Daniel