System does not reboot or poweroff

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

Since 3 days ago I have a strange issue: both my Gentoo systems does not poweroff or reboot. I use systemd. In openrc everything is just fine, but, due to configuration, I'm unable to fully use graphical interface with openrc.
I downgraded systemd (207 currently), gentoo-sources, completely disable Gentoo section in kernel, but no success. I found some topics dated somewhere in 2010 regarding wmi on arch, but was an dead road.
The only untested possible culprit remain glibc (2.18 ), but I can't downgrade it

One of system is UEFI, one is normal BIOS. The BIOS one sometimes act normally on systemctl poweroff, one from 3-4 tries, the UEFI one stop on Storage is finalized.
acpid is started. started or not does not make any difference.

kopp · Posted: Tue Sep 17, 2013 2:54 pm Post subject:

I'm also experiencing problem to shutdown/reboot/logoff with GDM. Can't find a way to reproduce it yet.
When it does so, trying to reboot in command line indicates that /dev/initctl cannot be accessed.

I don't know if this is related to your problem.

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

Yes, definitely is related. /dev/initctl is missing due to glibc-2.18 - I also found that reference in journalctl.
Please, can you confirm that this is your glibc version ?

Later: gentoo systemd-initctl[2769]: Received environment initctl request. This is not implemented in systemd.
Can you, please, confirm ?
_________________
Sorry for my English. I'm still learning this language.

kopp · Posted: Tue Sep 17, 2013 9:52 pm Post subject:

No, I'm using glibc 2.17. 2.18 seems hardmasked for me

I haven't checked journalctl
I'll try when this happens again. I still haven't found the conditions to reproduce it, but I guess it has to do with system suspend, wakeups etc and dual screen (laptop + vga output). Or maybe it was because of incomplete system update, since I've experienced with downgrades, rebuilds of some parts, etc. Now everything is up-to-date, we'll see ...

The message about /dev/initctl was appaering on tty when I entered commands such as reboot or shutdown or init 0. Something about too much delay.
The system was really slow once i tried to shut it down through Gnome. Long to log in on tty and to get to the command prompt after that.

kopp · Posted: Wed Sep 18, 2013 9:25 am Post subject:

This is the error I get :

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

Well, I tried a lot of combinatios, even restored oldest backup and incrementally update it.
In the end, I can report just a half victory: the computer with BIOS stop/reboot corect. I no longer can reproduce the issue on it.
The UEFI one, I'm still trying...

On machine with BIOS keeping sysvinit-2.88-r3 (-r5 is the latest) did the trick. But I can not say for sure that it is the culprit because masked -r5 version didn't solve the problem on second computer.
_________________
Sorry for my English. I'm still learning this language.

kopp · Posted: Wed Sep 18, 2013 3:50 pm Post subject:

My computer has BIOS.
Still hasn't reproduced. I really don't see what can trigger the problem so far.

sysvinit should not have influence as this systemd is supposed to not require it anymore. Even though reboot, halt and stuff comes from sysvinit

kopp · Posted: Thu Sep 19, 2013 7:09 am Post subject:

Seems like it happens after a long time in suspend. Or maybe only a long time running.

and FYI, killing the dbus process is not a solution :roll:

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

Just a head up: the computer with BIOS seems to be all right. The UEFI one still in trouble. I think I'm waiting for gnome-3.10 to hit the portage (and hopefully gcc-4.8.2 as about the same time) to do a scrach build. I'm out of ideeas, unable to point the finger to the culprit package. If you know what is wrong you can report it in bugzilla and wait for a fix. But this seems to be a very rare case, or, perhaphs, a big bistake made somehow by me, or a unfortunate mix of hardware, or ... I think you understand my frustration

_________________
Sorry for my English. I'm still learning this language.

kopp · Posted: Sat Sep 21, 2013 8:55 am Post subject:

Well, the last dbus update didn't change a thing.

When I try to use systemctl, i get this :

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

I don't have that kind of messages in journal, but:

kopp · Posted: Wed Sep 25, 2013 9:50 am Post subject:

Well, I tried downgrading systemd (which went down to 204) and it seems to resolve the problem. I haven't experienced the problem yet.

Tony0945 · Posted: Wed Sep 25, 2013 9:13 pm Post subject:

Not using systemd. I have this problem with gentoo-sources 3.10.7 and vanilla-sources 3.10.7. I do not have the problem with 3.8.13 (gentoo) but it is slower and has problems with USB.

I'm thinking the problem is the new CONFIG_GENTOO flag that is not there in 3.8.13

Problems with two systems, one an old Athlon64 x2 (Manchester?) with Biostar board, the other is Phenom II x6 with Gigabyte board.

kopp · Posted: Thu Sep 26, 2013 7:14 am Post subject:

Well, actually, problem came up with old systemd too.
so, yeah, maybe it's kernel related.

I don't think it's because of the new option, which should just select what you need, but maybe I'll give a try to an older kernel.

Tony0945 · Posted: Fri Sep 27, 2013 1:27 am Post subject:

Happened with vanilla 3.10.7 too, so it's not the Gentoo options. Perhaps I'll try the latest kernel and see if it works. Or start going backwards between 3.10.7 and 2.8.13 and see where it broke down. That will take a long time. I'll do it by binary search.

Tony0945 · Posted: Mon Sep 30, 2013 12:51 am Post subject:

Upgraded all my boxes to latest ~amd64 gentoo-sources kernel and latest ~amd64 sysvinit. Two boxes shut down and reboot fine. This box hangs after printing "remounting / as read-only".

Tonight I manually unmounted all drives and partitions except '/' (mount point /dev/sdb3). I killed X with "/etc/init.d/xdm stop" which responded with "stopping slim". I then typed "reboot" and a shutdown sequence started ending with "remounting / as read-only". I went out to shop at three stores returning forty-five minutes later to find the screen in the same spot. I power-cycled with the power button. This box triple boots with Windows XP and Mythbuntu, both of which reboot with no problem. That rules out a motherboard problem. Gentoo is on the second hard drive a Western Digital Velociraptor.

Kernel 3.8.13 still reboots but I don't want to run it because it is hard masked for errors and it gives USB errors when booting.

Manually powering off may be OK because I never get the boot message about unmounting uncleanly, but it is disconcerting and inconveniant to not just command a reboot, walk away, and come back to the rebooted system.

Any ideas on further troubleshooting? What does 'init' do after "remounting / as read-only"? If it is waiting for a process to end, how can I tell what process? After '/' is read only, there can't be any log messages." Is there a debug or verbose version of init that prints more messages to the screen.

costel78 · Guru Joined: 20 Apr 2007 Posts: 407

For me was two problems (solved now) php-fpm which take almost 3 minutes to stop/fail and the error suddenly disappear with recent 3.11.2 kernel. Before kernel, disabled acpid allowed computer to shutdown/reboot normaly.
Unfortunately it was an try & error procedure, and, finally, no exact culprit was identified.
_________________
Sorry for my English. I'm still learning this language.

kopp · Posted: Mon Oct 14, 2013 1:46 pm Post subject:

I downgraded to kernel 3.10.*, and it seems that it solved the problem. On the last 3.11 kernel, I wasn't even able to suspend to ram sometimes. Haven't experienced troubles since the downgrade.

kopp · Posted: Mon Nov 18, 2013 10:05 am Post subject:

Problem seems to have solved with kernel 3.12 or others update. Haven't tester 3.11 lately.