Devfsd isn't starting!

Figaro · Posted: Tue Mar 25, 2003 4:33 am Post subject: Devfsd isn't starting!

OK...this system has been up and running for several months. Last night I did an emerge -u world. Everything went fine, I did the etc-update....and apparently replaced something I shouldn't have.

My general rule is if I haven't touched it, I replace it. Anyway, I rebooted today and apparently the devfs aliases (/dev/sda1, /dev/sda2, etc) aren't being created.

I can boot up in single user mode and run '/sbin/devfsd /dev' and the devices are being created with no problem.

I did recompile the kernel, but I kept the old one as a backup and both exhibit the same behavior.

I assume there's a config file somewhere, but I can't find it. Please help!

flukemeister · n00b Joined: 25 Mar 2003 Posts: 3

Well, this frankly irritates me, because my entire system just got hosed by the same problem, and I've had to spend the past 5 hours attempting to work out why, and how to recover from it.

So, the offending package appears to be baselayout 1.8.6.4-r1, and the file is /etc/conf.d/rc

The ebuild script for 1.8.6.4-r1 changes the options for svcdir and svcfstype, but also seems to set svcmnt (i think that's the parameter) to "no". Much badness follows.

I'm going to raise this as a bugt in bugzilla now that I have my net access back

sethbc · n00b Joined: 03 Dec 2002 Posts: 16

i'm gonna suggest you fellas rsync, and update (read downgrade) gawk to version 3.1.1-r1

we've masked -* gawk 3.1.2, as this appears to be the problem...

Thanks to lostlogic that is =)

seth

good luck

sethbc · n00b Joined: 03 Dec 2002 Posts: 16

i lied...you guys getting this message?
rm /dev/.devsd operation not permitted message

?

looks like its the baselayout

di1bert · l33t Joined: 16 May 2002 Posts: 963 Location: Oslo, Norway

Yeah. I got that error when I booted this morning after emerging gawk last night. And I updated baselayout earlier this week and that error didn't appear then.

So I downgraded gawk to 3.1.1-r1 and all is fine and dandy again

I would just like to thank the Gentoo guys and the forums as they have saved my ass on so many occasions..Great job guys

di1bert (~matt)
_________________
choff.

flukemeister · n00b Joined: 25 Mar 2003 Posts: 3

Hi Seth,

Yeah, I'm getting that, as one of the first actions in the initscripts. At this point, only the root fs is mounted, and it's read only, so no joy.

Downgrading to baselayout 1.8.5.8 got me back in, but there are still *huge* problems all over the initscripts.

It would appear that everything in /var/lib/init.d/ gets wiped out every reboot, and that the initscripts are trying to write state information to non-existent subdirectories of /var/lib/init.d/

I'm going to stick with my guess that it's /etc/conf.d/rc that's missing a line. There needs to be a lot more sanity checking in the post-install for baselayout. It might be better to make this package restart portage so that users have time to see how important the changes are.

jcw · n00b Joined: 27 Nov 2002 Posts: 10

Same problem here, error is "rm /dev/.devsd operation not permitted".

I've been able to boot from a secondary partition containing a full (old!) system (phew!), but my question is now:

Since I cannot boot normally, how do I emerge fixes/downgrades/whatever?

(I can get to a root shell after reiserfsck fails due to /dev/hda7 missing - not sure how much of the system is in place at that point, though, a quick look around ran into all sorts of funniness)

derelm · Tux's lil' helper Joined: 26 Nov 2002 Posts: 95

hi, i got the same problem,

last night i updated modutils, binutils, gawk and today i couldn't get in ... so hopefully downgrading gawk will do the trick (i already downgraded modutils and binutils, but without the desired effect)

Figaro · Posted: Tue Mar 25, 2003 12:01 pm Post subject: well that didn't work

Booting in single user mode.

Manually started devfsd & net.eth0. emerge sync ran with no problems, but the gawk build won't run.

Comes up with:
ACCESS VIOLATION SUMMARY
open_wr: /dev/console
open_wr: /dev/console
open_wr: /dev/console

next?

jcw · n00b Joined: 27 Nov 2002 Posts: 10

How do y'all boot into single user mode?

I added "S" to end of gentoo kernel line during the boot - but it doesn't get me past the devfs problems?

Any tips how I can get Gentoo booting up again?

Figaro · Posted: Tue Mar 25, 2003 12:39 pm Post subject:

I'm currently booting with

'<original kernel line> -b rw sushell=/bin/sh'

Figaro · Posted: Tue Mar 25, 2003 1:01 pm Post subject: Moved svcdir?

In /etc/conf.d/rc, SVCDIR has changed from /mnt/.init.d to /var/lib/init.d.

I copied all of the files from /mnt/.init.d to /var/lib/init.d since the latter was empty.

After rebooting, the system came up with the 'rm /dev/.devfsd' mentioned above, but after pressing CTRL+D the system booted pretty normally but wouldn't throw up a login prompt.

I can ssh into it and all my services are back up and running.

EDIT: I forced a kill of:
1229 ? D 0:00 \_ /usr/sbin/rpc.nfsd 8

and after a couple more errors, the system is up at a login prompt.

jcw · n00b Joined: 27 Nov 2002 Posts: 10

moose · n00b Joined: 21 Mar 2003 Posts: 3 Location: Colorado

Hey all-
Same problem here. I had emerge -u world a couple times since my last reboot, and both baselayout and gawk had been updated. I had all the fun CTRL-D or root password stuff, but CTRL-D would reboot.
I used the 'a' option in grup to add "-b rw sushell=/bin/sh" (thanks figaro) to boot the system. Of course, I couldn't emerge anything, because of the "open_wr: /dev/console" error. I got around this by using FEATURES="-sandbox" emerge gawk-xxxx.ebuild. I downgraded to 3.1.1-r1.
I tried to reboot then, but couldn't due to some timeout error that escapes me at the moment. So after a hard-reboot, the system booted, and got through the init-scripts with MASSIVE errors, but i could login normally. The errors all seemed related to /var/lib/init.d/* stuff, but copying from /mnt/.init.d didn't fix it. Sooo, I downgraded my baselayout, etc-updated, rebooted. Everything works fine, but for a couple etc file errors (such as modules.autoload, but I prolly screwed it up myself and didn't realize it.) If anyone knows the latest versions you can run without everything going to hell, could you post it? or if there's another set of posts about it, could you post a link? Thankee kindly.

derk · Posted: Wed Mar 26, 2003 5:40 am Post subject:

Hi I got nailed by this too .. but .. I used the gentoo install CD to fix things .. first I booted and mounted all my partitions as you would for an install (once you have your root partition mounted all your regular mount points exist)

don't forget to

mount -o bind /proc /mnt/gentoo/proc

chroot /mnt/gentoo /bin/bash
env-update
source /etc/profile

then I did emerge sync ; emerge world;

This pulled in gawk as well as a few other things

next unmount everything

then I did: e2fsck -f /dev/hda10
etc on all my partitions to force fixes .. there were a number of errors

reboot, remembering to remove the CD from the drive and so far all is well.

Now that I think of it I did do a e2fsck check on all the partitions before hand but without the force and this did not complete the fix so i had to go back and use the force option '-f'

the best order for this would be to do the e2fsck -f command on all your linux partitions first, then the mount, chroot stuff, and emerge.

hope this helps

derk

BigBug · Posted: Wed Mar 26, 2003 7:26 am Post subject:

Seems problem is in the gawk ebuild script ...
because I'v got the same problem after upgrading gawk, but downgrading didn't help me.

To fix it i have done the following:
cd /usr/portage/sys-apps/gawk/files/filefuncs/
make
cp filefuncs.so /lib/rcscripts/

and now all is working fine ;)

morck · Posted: Wed Mar 26, 2003 9:47 am Post subject:

thanx BigBug!!!

this really works ...
1. downgrade gawk
2. remake filefuncs.so.0.0.1 and copy it to /lib/rcscripts/

Disquiet · Posted: Wed Mar 26, 2003 7:48 pm Post subject:

I gave that a try, but no luck. I'm still not getting my devices created in /dev.

I just tried upgrading to gawk 3.1.1-r2 but that hasn't made a difference either.

Any more ideas?

Thanks
Simon

smulloni · n00b Joined: 26 Mar 2003 Posts: 6

It is amazing that the problems we've been having with this have manifested themselves so differently for different users, and required such different solutions. In my case, I was always able to boot, but before I upgraded gawk got a warning about my kernel not supporting devfs, and then a ton of errors about init scripts. After reading these notices, I downgraded gawk and still got the error notices, but not the devfs message. So I downgraded baselayout to stable, and everything worked again. I didn't edit any files; my /usr/portage/sys-apps/gawk/files directory does not even contain any subdirectory called "filefuncs" that I could run make in.
_________________
Jacob Smullyan

Disquiet · Posted: Wed Mar 26, 2003 9:19 pm Post subject:

Yeah...I'm running a stable system (except for Xfree 4.3 and Mozilla 1.3), which seems strange as everyone else that's been having this problem seems to be running the unstable versions of baselayout etc.

I think gawk only has the filefuncs directory for 3.1.1-r1 and above.

From the packages.mask file it seems like this is due to the developers of gawk changing the api.

wHAcKer · Posted: Thu Mar 27, 2003 1:45 am Post subject:

OH MAN!!!!!!!!!!!!!
i'm reading this now from my parents' 300 mhz @ windows f'in XP... I thought i d0rked everything up since my install didn't recognise my /dev/hda, hdc, hdb,... I tried recompiling my kernel with some other options (i recently removed agpgart, so i thought I made a mistake or removed devfs and this caused the hassle) at my other gentoo machine (which isn't able to emerge anymore :/) a couple of times but nothing worked so i began my 5th reinstall or sth...
DAMN, this sucks... I hope it's fixed fast because i love unstable too much...

Disquiet · Posted: Thu Mar 27, 2003 8:02 pm Post subject:

Man do I feel dumb. Along the way of trying to fix this problem I re-emerged baselayout
and by mistake overwrote my modules.autoload file. I re-did the file, but forgot
to load mousedev...

But the initial problem was caused by the update world...all I can say is that the
problem must have somehow been fixed by me emerging gawk, baselayout and devfsd again.

So my system is up and running again

seto · n00b Joined: 09 Apr 2003 Posts: 25 Location: Switzerland

I tried to look into this bug myself, but could find nothing.
Anyone has an idea why gawk-3.1.2 can't access files in /proc any more?

btw: My "bugfix" was editing /etc/rc and changing all
gawk 'someting' /proc/something
to
cat /proc/something | gawk 'something'
which works fine... But that's really something one shouldn't need to do.
The "update" made gawk working correctly again.

When my system couldn't boot any more I was cursing you gentoo guys badly - really, really sorry for that. It wasn't your fault after all. :oops: