Summary: | initscripts fail during boot because /var not mounted | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Cheyenne Wills <cwills> |
Component: | [OLD] baselayout | Assignee: | Gentoo's Team for Core System packages <base-system> |
Status: | VERIFIED INVALID | ||
Severity: | major | ||
Priority: | Highest | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | sep-var-boot.png |
Description
Cheyenne Wills
2005-02-07 17:11:15 UTC
can you post the output somehow ? did previous 1.11.x versions work ? if you read /sbin/rc you'll see that /var is supposed to be mounted before $svcdir is utilized 2nd question first -- yes this did work with older versions. 1st question hmm -- since /var/log is not yet mounted, I don't have a log. Let me see if I can play around and try to capture something. Won't be till late this evening however. One possible solution is during rc (sysinit) processing, is to see if /var is in the fstab and do the checkfs/mount for it explicitly there. In fact, it might be a good idea to have a list of the different "critical" directories that are needed by rc and have them checked and mounted in rc (sysinit). Just took a look at something... mkdir /tmp/fooroot I did a mount --bind -o ro / /tmp/fooroot and there was /tmp/fooroot/var/lib/init.d/... directory structure So... I think in addition part of the problem is: In checkroot, root gets mounted read only (and it looks like there is a path out of there were root could be left readonly) In addition /var directory that is part of the physical "root" partition gets replaced by the /var partition when it is mounted. Lots of twisty little passages. I will try to play around this evening Created attachment 50882 [details]
sep-var-boot.png
ok, i'm unable to reproduce here ... this is my setup:
/dev/hda / ext2 0 1
/dev/hdb /var ext2 0 2
my / has a completely blank /var/, so there's no way for that to be an issue
bootup with baselayout-1.11.9-r1 is exactly as expected, see the attached
screen shot of qemu test env
Did the following: mount --bind / /mnt/fixroot cd /mnt/fixroot/var rm -r * # Clean out everything on the root partition var directory cd / umount /mnt/fixroot I was still having problems.. changed /etc/conf.d/rc to set svcdir to /etc/runlib and to mount a tmpfs for it. Was still having problems After doing quite abit of tracing, I was able to finally track down that I had a "stray" entry in fstab for none /devfs /dev -- I am using udev So What I think was happening was that when rc hit localmount, it mounted /devfs on /dev and screwing up what udev had set up. removing the /devfs entry from fstab cleared up the boot problem. I suspect that earlier versions of baselayout either masked the problem, or I was just getting lucky. I am going to check one last thing.. I'm going to remove the svcdir mount statement. I suspect that in the root partition I will find the svcdir data there. did the change -- root partition var directory still clean.. So.. looks like I had a config problem that was masked by the earlier versions of baselayout. Sorry for the wild goose chase.. Closing as user configuration error |