mountall hangs boot if first field of an fstab entry starts with the name of a filesystem

Bug #695404 reported by James Troup
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
mountall (Ubuntu)
Fix Released
Low
Unassigned

Bug Description

Binary package hint: mountall

After upgrading a server to Ubuntu 10.04, I found it hanging early in
the boot process with no error messages. Eventually I discovered the
the problem: entries in fstab that have a first field (fs_spec) which
starts with the name of a filesystem type caused the boot to hang.
Renaming or simply prefixing them to avoid clashes with a filesystem
type avoided the problem.

So for example, the following fstab would hang during boot:

| proc /proc proc defaults 0 0
| UUID=e8e54459-86eb-41f9-ae03-a8c576fb0c9d / ext3 noatime,errors=remount-ro 0 1
| UUID=dbe2dc82-a339-4b86-b1e6-de1060e09845 none swap sw 0 0
| proc-epic-i386 /srv/epic.canonical.com/chroot-i386/proc proc defaults 0 0
| devpts-epic-i386 /srv/epic.canonical.com/chroot-i386/dev/pts devpts defaults 0 0

While the following works:

| proc /proc proc defaults 0 0
| UUID=e8e54459-86eb-41f9-ae03-a8c576fb0c9d / ext3 noatime,errors=remount-ro 0 1
| UUID=dbe2dc82-a339-4b86-b1e6-de1060e09845 none swap sw 0 0
| lolproc-epic-i386 /srv/epic.canonical.com/chroot-i386/proc proc defaults 0 0
| loldevpts-epic-i386 /srv/epic.canonical.com/chroot-i386/dev/pts devpts defaults 0 0

Arguably our use of fs_spec is 'out of spec' (sorry), but FWIW I've
been doing this for years on both Debian and Ubuntu systems and I've
not run into a problem with it until now. In any event, silently
hanging on boot is a fairly user hostile way to react and (IMO) a bug,
regardless of the validity of our use of fs_spec.

Revision history for this message
James Troup (elmo) wrote :

So, the prefixing trick doesn't seem to work all the time. Some boots
still hang. I'm 99% sure I didn't entirely imagine it working and
that it really did boot for me with the prefixed-entries in there.

But in any event it doesn't anymore. Setting the first field to be
'none' for both the proc + devpts chroot mounts appears to work
reliably over the course of 3-4 reboots.

tags: added: regression-release
removed: regression-potential
Steve Langasek (vorlon)
Changed in mountall (Ubuntu):
status: New → Triaged
importance: Undecided → Low
Revision history for this message
Steve Langasek (vorlon) wrote :

This bug appears to have been fixed somewhere along the way; not reproducible for me with mountall 2.51.

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

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.