Ubuntu
autofs package

hal interferes with umount autofs using SIGUSR1

Bug #63720 reported by Hein Zelle
4
Affects Status Importance Assigned to Milestone
autofs (Ubuntu)
Invalid
Undecided
Unassigned
hal (Debian)
Fix Released
Unknown
hal (Ubuntu)
Fix Released
Low
Unassigned

Bug Description

The automount process is supposed to unmount all mounted filesystems when it receives a SIGUSR1. I use this extensively to unmount things like usb-sticks. On a fresh dapper system with the xubuntu desktop, this doesn't work:

- the automount daemon DOES unmount the filesystem, as can be seen by doing a 'mount' command right after the kill command
- some other process (?) then accesses the filesystem that was just unmounted
- automount notices this and immediately remounts the filesystem

Effectively, within 0.5-1 second all unmounted filesystems are remounted again. This seems to always occur with nfs mounted filesystems, but It also regularly happens with usb sticks and other removable media. (but not 100% reliably).

On a debian unstable system I saw the same thing happening, there it was due to the hald trying to access the filesystem once it was unmounted. I have tried to stop the hald and reproduce the bug - no visible changes in behaviour. The bug also occurs from a text console right after a fresh reboot, so it appears that it is not due to KDE or GNOME services (or XFCE in my case). I have tried shutting down as many system daemons as possible, but could not locate the culprit.

I realize this bug is likely not the fault of the autofs package, but it is related and very annoying. It would be nice if it could be investigated, as it renders the autofs package nearly useless for removable media.

Revision history for this message
Hein Zelle (zelle) wrote :

It appears that hal is the culprit after all: after stopping the dbus service completely using /etc/init.d/dbus stop, the problem has gone away. This means that this report is probably not for autofs but for the hal package, and it could confirm bug 55223.

Revision history for this message
Paul Sladen (sladen) wrote :

Hello Hein, could you attach all the *autofs* files that you are using so that the situation can be replicated?

Revision history for this message
Paul Sladen (sladen) wrote :

Appears that 'hal' (Hardware Abstraction Layer) is causing the partitions to be automatically remounted. Could you test what happens if you also call:

  eject /dev/somewhere

Does that prevent the device from being immediately remounted?

Changed in hal:
importance: Undecided → Low
status: New → Incomplete
Revision history for this message
Paul Sladen (sladen) wrote :

There's a reference to the issue in:

  http://lists.freedesktop.org/pipermail/hal/2004-September/001042.html

comment:

+++ ./hald/hald_conf.h 2004-09-22 23:06:42.000000000 +0100
+ /** If true, then the behaviour of hald matches the use of
+ * autofs (automount - /etc/auto.hal) instead of a static
+ * fstab expectations.
+ *
+ * Basically it doesn't keep on generating mount notifications
+ * just because autofs automatically unmounts a volume, which
+ * would get _excruciatingly_ boring for users when the autofs
+ * timeout is set to oh, say, two seconds.
+ */

changelog:

  - added new config option "autofs_in_use" to hald.conf please think of a
    better name!

Possibly also handled by:

hal (0.5.7.1-1) unstable; urgency=low

  * debian/patches/16_dont_stat_autofs.patch:
    - Added. Don't stat autofs filesystems as that will remount.
      (Closes: #361785) (From upstream git)

Changed in autofs:
status: New → Invalid
Revision history for this message
Paul Sladen (sladen) wrote :

AFAICT, this has been fixed and the change imported via upstream.

Please confirm if it's not the case and reopen.

Changed in hal:
status: Incomplete → Fix Released
Bug Watch Updater (bug-watch-updater)
Changed in hal:
status: Unknown → Fix Released
Revision history for this message
Hein Zelle (zelle) wrote : Re: [Bug 63720] Re: hal interferes with umount autofs using SIGUSR1

Paul Sladen wrote:

> AFAICT, this has been fixed and the change imported via upstream.
>
> Please confirm if it's not the case and reopen.
>
> ** Changed in: hal (Ubuntu)
> Status: Incomplete => Fix Released

Will do as soon as I can find time for it, should be soon.
Should I confirm using 7.10? If so it will wait until the release, I
cannot install it before that at work.

Regards,
 Hein

--

  dr. ir. Hein Zelle

  Project manager

  ARGOSS: Atmospheric, marine & coastal information, systems and consultancy.

  ARGOSS BV
  PO Box 61
  8325ZH Vollenhove
  The Netherlands
  Tel +31 (0)527242299
  Fax +31 (0)527242016
  Web http://www.argoss.nl/

  Confidentiality Notice & Disclaimer

  This message and its attachments can contain confidential and/or
  proprietary information which could be subject to legal privilege or
  immunity. Unless you are a named addressee or receive this email on
  behalf of a named addressee you may not use, copy or disclose the
  contents in any manner. Should you receive this message in error, you
  are kindly requested to inform the sender and to delete all electronic
  or paper copies. ARGOSS BV does not accept any liability for any
  errors or omissions in the context of this email or its attachments
  which arise as a result of Internet transmission, nor accept liability
  for statements which are those of the author and not clearly made on
  behalf of ARGOSS.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

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