mountall fails with "mountall: fsck /home [369] terminated with status 8"/ Non-existent device - Possible race ?

Bug #712811 reported by Jean-Baptiste Lallement on 2011-02-03
28
This bug affects 5 people
Affects Status Importance Assigned to Milestone
mountall (Ubuntu)
Undecided
Unassigned

Bug Description

Binary package hint: e2fsprogs

On each boot I get the following message from fsck, which interrupts the boot.
fsck.ext4: No such file or directory while trying to open /dev/disk/by-uuid/a14daa42-496b-425a-bc9f-d59cca6a0216
Possibly non-existent device?
mountall: fsck /home [369] terminated with status 8
mountall: Unrecoverable fsck error: /home

If I press 'I' the the boot continues and the device is correctly mounted.

mountall doesn't fail when:
- quiet splash is not present on the boot command line
- mountall verbose mode is on (-v to the exec line in /etc/init/mountall.conf)
- small delay before starting mountall (sleep .1 before exec line in /etc/init/mountall.conf)

$ ls -l /dev/disk/by-uuid/
total 0
lrwxrwxrwx 1 root root 10 2011-02-04 00:05 0cf40974-73f1-4398-a9a4-8df51aec1d64 -> ../../sdb1
lrwxrwxrwx 1 root root 10 2011-02-04 00:05 30fc9e49-b3c8-4ea6-aec8-d504895ea7eb -> ../../sda1
lrwxrwxrwx 1 root root 10 2011-02-04 00:05 87833b0f-dd6d-4d93-822b-3fe67ea78d69 -> ../../sdb2
lrwxrwxrwx 1 root root 10 2011-02-04 00:05 a14daa42-496b-425a-bc9f-d59cca6a0216 -> ../../sdb3

ProblemType: Bug
DistroRelease: Ubuntu 11.04
Package: e2fsprogs 1.41.14-1ubuntu1
ProcVersionSignature: Ubuntu 2.6.38-1.28-generic 2.6.38-rc2
Uname: Linux 2.6.38-1-generic x86_64
Architecture: amd64
Date: Fri Feb 4 00:09:53 2011
ProcEnviron:
 LANGUAGE=en_US:en
 PATH=(custom, user)
 LANG=en_US.UTF-8
 LC_MESSAGES=en_US.utf8
 SHELL=/bin/bash
SourcePackage: e2fsprogs

Jean-Baptiste Lallement (jibel) wrote :
Jean-Baptiste Lallement (jibel) wrote :
Jean-Baptiste Lallement (jibel) wrote :
Jean-Baptiste Lallement (jibel) wrote :
description: updated
Theodore Ts'o (tytso) wrote :

My guess is that /dev/disk/by-uuid wasn't properly populated at the time that fsck was run.

I've always thought the /dev/disk/by-uuid was a horribly bug, and its use should be stamped out. Fortunately with the blkid library now being incorporated into util-linux-ng, and this is what is being used in all major distributions, it *can* be stamped out. Don't use /dev/disk/by-uuid; it's horribly fragile and not guaranteed to work all the time.

The far better thing to use is to put the following in /etc/fstab instead:

UUID=a14daa42-496b-425a-bc9f-d59cca6a0216

Its use is documented in the fstab man page.

-- Ted

Jean-Baptiste Lallement (jibel) wrote :

Thanks for the quick reply.

fstab already use the syntax UUID= and this is a fresh installation. The difference with a fresh installation are the disk partitions.
I attach fstab to the report.

Theodore Ts'o (tytso) wrote :

Huh. Well, the pathname /dev/disk/by-uuid/.... that e2fsck was complaining about was passed to it by fsck. And fsck is part of the util-linux-ng package, which as far as I knew translated UUID=... into a /dev pathname directly, and which didn't use /dev/disk/by-uuid/... any more.

Hm.... can you show the output of "fsck -ANV" (it doesn't have to run as root; this is a debugging command that doesn't actually do any thing but print what fsck would execute).

Jean-Baptiste Lallement (jibel) wrote :

output of

$ fsck -ANV
fsck from util-linux-ng 2.17.2
Checking all file systems.
[/sbin/fsck.ext4 (1) -- /] fsck.ext4 /dev/sda1
[/sbin/fsck.ext4 (1) -- /home] fsck.ext4 /dev/sdb3
[/sbin/fsck.ext4 (1) -- /var] fsck.ext4 /dev/sdb2

Jean-Baptiste Lallement (jibel) wrote :

This only happens if 'quiet splash' is present on the boot command line.
Definitely not an fsck issue, I'm affecting to mountall.

affects: e2fsprogs (Ubuntu) → mountall (Ubuntu)
summary: - at boot fsck complains that a device is not present but it is
+ mountall fails with "mountall: fsck /home [369] terminated with status
+ 8"/ Non-existent device when booting with arguments "quiet splash"

Also note that the boot doesn't fail if mountall verbose mode is on, or when there is a small delay (sleep .1) before starting mountall in /etc/init/mountall.conf

This looks like a race of some kind.

summary: mountall fails with "mountall: fsck /home [369] terminated with status
- 8"/ Non-existent device when booting with arguments "quiet splash"
+ 8"/ Non-existent device - Possible race ?
description: updated
Rune K. Svendsen (runeks) wrote :

I experience the same issue, but for me it doesn't happen on every boot, only occasionally.
I'm attaching my /var/log/boot.log which claims there are several problems, both with the swap disk (which seems to work fine) and with a disk that is a RAID array (which is present when I get to a desktop, unlike boot.log claims). But it doesn't seems to be associated with any particular partition. Except for my root and home partition, this seems to happen to a different disk each time it occurs.

Milton Y. (milton-0) wrote :

Same here on a dmcrypt /home partition. A sleep 0.1 does not do it, nor does a sleep 1. But if I ignore the error (pressing I) then I can continue normally...
I have checked this partition and it is perfectly fine fsck-wise.

This is a real show stopper for anybody who has set up encryption on a partition and does not know how to troubleshoot using the console (hopefully not my case).

Removing quiet splash does not change anything for me.

Launchpad Janitor (janitor) wrote :

Status changed to 'Confirmed' because the bug affects multiple users.

Changed in mountall (Ubuntu):
status: New → Confirmed
Steve Langasek (vorlon) wrote :

Hi Jean-Baptiste,

Is this issue still affecting you?

Every default Ubuntu system uses UUID= syntax in /etc/fstab. So if there is a race with the creation of the /dev/disk/by-uuid/ links, it's strange that only a handful of people should be reporting it.

If you're still seeing this in 12.04, can you attach /var/log/udev log? There may be something we can see there.

Changed in mountall (Ubuntu):
status: Confirmed → Incomplete

Hi Steve,
I can't tell, I changed the partitioning of this system due to this error.
I'm closing this report because I cannot provide any additional useful information.

Changed in mountall (Ubuntu):
status: Incomplete → Invalid
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers