[SOLVED] boot stops after mounting /root, goes emergency mode, no logs written

JZL240I-U · 12-18-2011, 07:37 AM

This is openSuSE 12.1 which comes with systemd.

These are the last messages I get:

Code:

...
fsck succeeded. Mounting root read-write.
Mounting root /dev/sda10
mount -o rw,data=journal,alc,usr_xattr -t ext4 /dev/sda10 /root
[nn] shpchp 000:00:01.0: Cannot reserve MMIO region
[nn] SP5100 TCO timer: mmio address 0xb8fe00 already in use
Welcome to emergency mode. Please login.
root password:

I tried "init 5" and <Ctrl D> with no change.

Only thing I can think of is that I tried starting it with the home partition of my still working 11.4 installation. I took that back of course but the error stays. As far as I can say the home partition is not mounted at this point in the boot procedure anyways...

I had a look at /var/log but the files there are old. So there is no logging done yet.

Anybody with any ideas

?

<edit>
I booted in failsafe mode. More information during boot was

Code:

...
systemd[1]: Job dev-disk-by\x2did-ata....device/start timed out.
Starting /home aborted because a dependency failed.
{some more messages ending with 'dependency'}
failed with result timeout.

I fsck'ed that home partition with option "-f". All is "well". Boot still stops here...

.
</edit>

EricTRA · 12-19-2011, 12:22 AM

Hello,

I'm encountering similar problems with my Debian testing in combination with systemd. Normal boot (SysV init) works fine but when trying to boot with systemd enabled it hangs, albeit at another point then your installation. I've found a bypass, starting in single user mode and then running:

Code:

systemctl default

which gets me into the desired runlevel without problems. I don't know openSUSE so maybe a dumb question: does openSUSE come with systemd as the only init manager? Or did you install it alongside sysvinit? If you did install it along sysvinit, you'll need to go through the sysvinit scripts and check the dependencies listed inside those scripts. That's how I'm trying to fix my systemd environment. For example one of my sysvinit scripts listed acct (process and login accounting) as a dependency but I don't have that installed, so when systemd processed my sysvinit scripts it errored out on that one.

Hope my findings help you out in some way.

Kind regards,

Eric

JZL240I-U · 12-19-2011, 01:39 AM

Thanks Eric. Yes, openSuSE comes with an option to once boot with sysvinit (I'll just have to dig up how to initialize it).

What I don't understand is, why booting with systemd was possible until now and all of a sudden it won't work anymore ...

What could possibly force the mount-process for /home to timeout?

EricTRA · 12-19-2011, 01:53 AM

Hi,

I've had the same experience. When I first installed systemd I got version 29-1 and was able to boot with systemd without problems. About a week later I got an upgrade to 37-1 for systemd and the trouble started. If I now uninstall the latest version and re-install 29-1 I got everything working and booting again, so I imagine something went wrong between the two versions. Right now I'm sticking with the latest version and trying to troubleshoot/fix everything myself manually. Did you had a 'version change' on your system with regards to systemd?

Kind regards,

Eric

JZL240I-U · 12-19-2011, 01:58 AM

No, I had lots of trouble due to a faulty USB-stick. openSuSE 12.1 is now freshly installed (about one week). I want to use the /home of my 11.4, but I created an extra /home (sda11) for 12.1 so as to not overwrite anything vital during installation. I had no connection to the net yet, so I am absolutely sure not to have updated anything. It's really baffling.

JZL240I-U · 12-20-2011, 01:15 AM

One more thing: From the emergency shell I mounted /dev/sda11 into the /home directory -- flawlessly and with absolutely no delay. On issueing an "init 5" KDE started like expected. One more puzzle...

And I currently don't find the command to start with sysvinit

. I'll try tonight with "init=/sbin/init" as a kernel parameter and report later.

EricTRA · 12-20-2011, 01:25 AM

Hi,

Looks similar behavior to what I experienced in my opinion. Booting into single user mode with systemd enabled works as expected without errors. Then running as indicated before

Code:

systemctl default

(which is the systemd equivalent to init 5) starts everything as expected. I'm also puzzled by that one. I'm still in the progress of migrating all my sysvinit scripts to systemd to see if that solves my boot problem. I'll keep you up to date too, but it might take some time since I'm swamped in work issues. I'm not sure but I think passing init=/sbin/init is the only way to 'bypass' systemd, providing you still have your sysvinit scripts in place. I haven't found any 'legacy' boot option for systemd.

Kind regards,

Eric

JZL240I-U · 12-22-2011, 03:56 AM

Well, I did try with "init=/sbin/init" as a kernel parameter. /sbin/init is a symbolic link to /usr/bin/systemd. *hrmph*

.

In case anybody else is wondering, please use "init=/sbin/sysvinit".

I did that and SuSE booted as expected with the standard init process -- until the moment where /home was to be mounted. There it timed out and opened an emergency login prompting me to repair the file system. I ran "fsck -f /dev/sda11" again (no errors again) and I could mount /dev/sda11 with no problem by hand again. I'm completely stymied. Perhaps a new thread in the hardware forum...

Anybody with the slightest idea what is happening here? Or do I pass by 12.1 and wait for 12.2...

JZL240I-U · 12-22-2011, 10:56 AM

Erm. Well. Really, this is embarrassing

.

You know, when I prepared the switch of the 12.1/home (sda11) to the 11.4/home (sda6) working horse, I used the vi-editor in a hurry to edit /etc/fstab. I or rather openSuSE uses the UUID-scheme to identify partitions. So I changed /dev/disk/by-id/ata-ST31000520AS_5VX0VRDP-part11 /home to /dev/disk/by-id/ata-ST31000520AS_5VX0VRDP-par6 /home.

Little wonder mount didn't find that partition.

Thanks for your help and support, Eric

.

EricTRA · 12-23-2011, 12:55 AM

Hi,

Thank you for sharing your findings and it's not that embarrassing. To err is human after all or isn't it? Glad you've got it fixed. I'm enjoying an extended 5-day weekend now and will continue to convert my Debian testing sysvinit scripts to systemd. Let's see if I can find 'my' blocking point. Have fun with Linux.

Kind regards,

Eric

JZL240I-U · 12-23-2011, 08:36 AM

Hi Eric.

I hope you'll enjoy your prolonged weekend so I wish you a merry Christmas and a happy new year right here and now. Please let me know later what you found out during conversion of your scripts. systemd is coming to all of us and the more one is prepared the better

. Good luck.

EricTRA · 12-24-2011, 02:54 AM

Hi,

Merry Christmas and a Happy New Year to you and your family too. I'll be sure to post my findings. And you're right, systemd is coming and I believe it has lots of advantages although it means another adaptation from the user/administrator's point of view.

Kind regards,

Eric

wogo · 02-13-2012, 06:36 PM

I had the same problem when the boot process was mounting my raid-1 volumes:
In one of three times (randomly!) not all volumes were mounted and the process stopped with "Welcome to emergency mode".
And a lot more problems occurred:
message flood from dbus daemon makes /var/log/messages unreadable.
dovecot terminated immediately with error and nfs didn't mount all nfs disks (0-4 out of 5 volumes) and terminated with errors:
systemd[1]: dovecot.service: control process exited, code=exited status=7
nfs[1428]: Mounting network file systems ...mount.nfs: Failed to resolve server wogolix: Name or service not known

I had to return to sysvinit again (added "init=/sbin/sysvinit" in grub boot menu) and the problems disappeared.
My system is openSUSE 12.1 64Bit running on Intel motherboard with core-2-duo CPU and 2 disks (2 terabytes each with 5 partitions) in raid-1 mode.