[gentoo-user] Shutdown pauses partway with Give root password

2006-05-09 Thread glen martin
Resending ... anyone have a clue as to why the Give root password for
maintenance ... prompt would come up occasionally at shutdown time?



Well, this is weird.

We've all seen Give root password for maintenance (or type Control-D to
continue):, usually after an unclean shutdown.

I'm getting it on shutdown itself. I've never even heard of this, and
searching google I haven't found any reference to it.

It isn't happening every time - it'll go 2 or 3 or 6  (but certainly not
4 or 5 :b) times with clean predictable shutdowns. I can't tell that
anything is different in the times I shutdown fails - there aren't any
symptoms of any misbehaviour until the message itself.

This is a home theatre PC running MythTV.  Frontend only, so there are
no fancy drivers in the system, just video (nvidia 8756 driver), sound
(snd_hda_intel), and lirc with streamzap.  System was build from scratch
for this purpose, recently, so it is pretty up to date.  P4 with
hyperthreading. Normally runs with no keyboard or mouse or VGA, only an
SVIDEO-out from an NVidia card. BIOS has obviously been set to ignore
post errors given the lack of keyboard.

Myth running or not running is irrelevant, I've seen the shutdown
problem in both situations.

I've also seen the problem whether shutting down with a quick press to
the power button, or using ssh to run a shutdown command.

Here's a sequence:
  - press power button
  - this is presumably caught by acpid, which turns it into an init 0
command
  - hdd lights blink, eventually X is stopped
  - when X stops, that initial login prompt that came out before X
started is now displayed again, and right there I get the Give root
password ... prompt.

Sometimes the svideo-out console hasn't been restored so I can't see any
of this, it is only on a connected monitor (if there is a connected
monitor). Usually the svideo-out console is restored on X shutdown, so
this does in fact display.

After this happened a few times and didn't seem to be going away on its
own :) I grumbled, dug up a keyboard and plugged it in, entered the root
password.

The log shows ntpd, sshd and syslog-ng shutdowns. There are no messages
following the syslog-ng shutdown. :)

ifconfig returns nothing.

runlevel says 3 0

rc-status exhibits poor grammar * Could not local current runlevel in
/var/lib/init.d/softlevel * Assuming current runlevel is 'single'
Sure enough, there is no softlevel file at this point in a partway
shutdown system.

df shows local partitions (/, /var, /usr) still mounted, and the sole
NFS mount has been taken down. local partitions are all ext2, no lvm or
anything fancy.

kernel is 2.6.16-gentoo-r3
System is vanilla, except for ~x86 keywords on nvidia-kernel and
nvidia-driver to get the recent 8756 versions.

Any suggestions on how to further debug? Or suggestions as to what may
be going on?

Thanks,

glen



-- 
gentoo-user@gentoo.org mailing list



Re: [gentoo-user] Shutdown pauses partway with Give root password

2006-05-09 Thread Hans-Werner Hilse
Hi,

On Tue, 09 May 2006 07:20:57 -0700 glen martin [EMAIL PROTECTED]
wrote:

 Resending ... anyone have a clue as to why the Give root password for
 maintenance ... prompt would come up occasionally at shutdown time?

That's sulogin. Did you mess up your /etc/inittab (like uncommenting
that line referring to sulogin)?

But I rather guess its an unclean umount and sulogin is spawned
from /etc/init.d/halt.sh (l.189). Maybe you can cat your /proc/mounts
next time you're in that single-user mode? It might make things more
clear...

-hwh
-- 
gentoo-user@gentoo.org mailing list



Re: [gentoo-user] Shutdown pauses partway with Give root password

2006-05-09 Thread glen martin
Hans-Werner Hilse wrote:
 Hi,

 On Tue, 09 May 2006 07:20:57 -0700 glen martin [EMAIL PROTECTED]
 wrote:
   
 Resending ... anyone have a clue as to why the Give root password for
 maintenance ... prompt would come up occasionally at shutdown time?
 
 That's sulogin. Did you mess up your /etc/inittab (like uncommenting
 that line referring to sulogin)?
   
Nope, I don't recall ever changing that file on this system, and
checking I find that this line is still commented. Though that would
probably have been a problem. :)
 But I rather guess its an unclean umount and sulogin is spawned
 from /etc/init.d/halt.sh (l.189). 
   
Possibly. Looking at that file it seems as if there's a 10 second(?)
timeout on sulogin spawned from halt.sh. This one doesn't go away in any
reasonable period of time. Also it looks as if there should be some
messages about remounting and such before that sulogin would spawn, and
I don't see such messages (presuming they should show up on this console).

If it is starting from halt.sh, is there any chance this could be a race
condition thing, in which some processes aren't fully shut down yet when
halt tries to umount?
 Maybe you can cat your /proc/mounts
 next time you're in that single-user mode? It might make things more
 clear...
I'll try this.

Thanks for the suggestions,

glen

-- 
gentoo-user@gentoo.org mailing list



Re: [gentoo-user] Shutdown pauses partway with Give root password

2006-05-09 Thread glen martin
Hans-Werner Hilse wrote:
 Maybe you can cat your /proc/mounts
 next time you're in that single-user mode? It might make things more
 clear...
   
3 power cycles later I duplicated the problem. Here is /proc/mounts,
transcribed by hand. There is nothing obvious wrong here (to me) except
that the filesystems are still rw.

/proc/mounts:
rootfs / rootfs rw 0 0
/dev/root / ext2 rw,noatime,nogrpid 0 0
proc /proc proc rw 0 0
sysfs /sys sysfs rw 0 0
udev /dev tmpfs rw,nosuid 0 0
devpts /dev/pts devpts rw 0 0
/dev/hda6 /usr ext2 rw,noatime,nogrpid 0 0
/dev/hda7 /var ext2 rw,noatime,nogrpid 0 0
shm /dev/shm tmpfs rw,nosuid,nodev,noexec 0 0
usbfs /proc/bus/usb usbfs rw 0 0

At this point, I manually remounted the 3 local partitions ro
mount -n -o remount,ro /
etc
which went cleanly, and /proc/mounts now shows them ro.

Is there any chance this could be a race condition thing, in which some 
processes aren't fully shut down yet when
halt.sh tries to umount or remount? But they're all shut down now (a couple of 
minutes later) so the remounts go cleanly?

Finally, after remounting the partitions (above), I pressed Ctrl-D to kill the 
sulogin shell, and the machine rebooted. It didn't power off, as I might have 
expected.

glen




-- 
gentoo-user@gentoo.org mailing list