2 swap partitions each mounted twice in 2.6.20-8

[Thomas Templin]

Binary package hint: linux-source-2.6.20

After bootup a 'swapon -s' shows:

/dev/hda2
dev/hdb1
/dev/mapper/hda2
/dev/mapper/hdb1

This always results in a system crash only a few minutes after loged in to a GNOME session!

But when I do a 'swapoff -a' followed by a 'swapon -a' a 'swapon -s' shows:

/dev/hda2
dev/hdb1

And the system doesn't crash after log in to a GNOME session, using a lot of swap (~600MB).
Kernel 2.6.20-8-generic and lowlatency

[Kyle McMartin (kyle)]

Yes, having two instances of the same swap partition would be problematic. Can you please paste or attach (in seperate attachments please) the output of the following:

While you have the multiple swap devices:
dmesg
dmsetup ls
dmsetup info
cat /etc/fstab

Thanks!
  Kyle

[Thomas Templin (coastgnu)]

grep -v ^# /etc/fstab

proc /proc proc defaults 0 0
none /proc/bus/usb usbfs defaults 0 0
UUID=1b95fc8e-2fed-4ff6-a727-917f9ce9bb97 / ext3 defaults,errors=remount-ro,user_xattr,acl 0 1
UUID=4683d8da-a917-4b5d-8b4f-03f3f3ad2936 /data ext3 defaults,user_xattr,acl 0 2
UUID=96ff845c-1af0-4c42-a7fa-c8868ea8e59c /home xfs defaults 0 2
UUID=0130e7ef-9add-44ca-9110-0db346af1ef3 /opt xfs defaults 0 2
UUID=361d04ca-d197-419a-89cf-8a1121dddbe6 /var ext3 defaults,user_xattr,acl 0 2
UUID=27991276-db58-4ec1-a999-17dd4b3cd3bb none swap sw,pri=2 0 0

UUID=6c26c5be-d9c5-4b5c-b284-c0b1001c6760 none swap sw,pri=2 0 0
/dev/cdrom /media/cdrom0 udf,iso9660 user,noauto 0 0
/dev/cdrw /media/cdrom1 udf,iso9660 user,noauto 0 0

[Scott Hilleard (scottles-deactivatedaccount)]

I have noticed that my system has the same problem (currently running the 2.6.20-14-generic kernel). The swap is only listed once in my fstab, yet gets mounted twice (as shown by swapon -s):

Filename Type Size Used Priority
/dev/sda1 partition 4883720 0 -1
/dev/mapper/sda1 partition 4883720 0 -2

[Sod75 (launchpad-net-vernaillen)]

same issue here on 2.6.20-15-generic

after a swapoff -a and a swapon -a it's only listed once

[Thomas Templin (coastgnu)]

In [gutsy] this bug is fixed I asume by dbus 1.0.2-5ubuntu1 (23.may.07)

In [feisty] the bug still exists

Changes:
 dbus (1.0.2-5ubuntu1) gutsy; urgency=low
 .
   * Merge with Debian unstable; remaining Ubuntu changes:
     - debian/patches/80-dbus-change-at-console-policy.patch: Make at_console
       policy work with libpam-foreground.
     - debian/patches/81-session.conf-timeout.patch: Raise the service startup
       timeout from 25 to 60 seconds. It may be too short on the live CD with
       slow machines.
     - debian/rules: Start dbus at runlevel priority 12, so that it comes
       before gdm. This eliminates the race condition of starting the X session
       before hal is running. Also use 'multiuser' mode of update-rc.d.
     - debian/dbus.postinst: Migrate rc?.d symlinks from 20 to 12 on upgrades
       to this version. This needs to be kept until the next Ubuntu LTS.
     - debian/dbus.{postinst,prerm}: Reverted Debian's change to always restart
       dbus on upgrades.
   * debian/control: Drop dbus dependency to dbus-x11, since we crave to get
     rid of that Xsession.d script. (LP: #62163)
   * debian/dbus.preinst: Improve rm_conffile() to get the unmodified md5sum
     from dpkg instead of hardcoding it, and remove unchanged
     /etc/X11/Xsession.d/75dbus_dbus-launch on upgrade.
 .
 dbus (1.0.2-5) unstable; urgency=low
 .
   [ Michael Biebl ]
   * debian/dbus.init
     + Add some safety checks and support for file-rc. (Closes: #419984)
   * Drop dbus-1-utils package and move dbus-monitor into the dbus package.
     (Closes: #404981)
   * Create dbus-x11 package and move dbus-launch and the Xsession start script
     from dbus into this package. This removes the X11 dependency from the dbus
     package.
     Add a dependency on dbus-x11 to the dbus package, to not break packages
     which still depend on dbus but should be updated to depend on dbus-x11
     instead if they require a D-Bus session bus.
 .
   [ Sjoerd Simons ]
   * debian/control: Add libselinux1-dev to build depends
   * debian/rules: Clean up some more generated files (From the ubuntu
     packaging)

[otzenpunk (reisswolf-nospam)]

Replacing the UUID in the fstab with the "real" devicename á la /dev/sda2 fixes this problem for me, too.

The swapon command is called twice during the boot process. Once in checkroot.sh and once in mountall.sh, the latter because:

        # Execute swapon command again, in case we want to swap to
        # a file on a now mounted filesystem.

This shouldn't be a problem, because swapon -a normally ignores partitions that are already activated.

The following is my speculation: I think for some reason the /dev/mapper-devices aren't present at the time of the first swapon, so the regular device belonging to the UUID get's activated. When swapon gets executed the second time, the /dev/mapper-device is present and is going to be used, when the translation UUID->device is performed. As the /dev/mapper-device isn't in the swap -list yet, it's going to get activated although the underlying partition is already running under a different name.

[otzenpunk (reisswolf-nospam)]

Just backed up my theory a little bit.

I dumped swapon -s into a file near the end of checkroot.sh and confirmed that there was only one swap entry (/dev/sda9) at this time. When the system was fully up and running, there were the two entrys with and without 'mapper' again.

[Thomas Templin (coastgnu)]

I can confirm this, replacinf UUID with the /dev entry in fstab works.

Also replacing dbus from 7.04 Feisty with dbus (1.0.2-5ubuntu1) gutsy does the job.
So updating feisty with the gutsy dbus package will be a solution for a bugfix upgrade.

So we have two solutions now to fix this bug + updating dbus will be a solution for novice users by offering an unattended upgrade via update-manager.

Kyle, it's your turn now, I would say. :-)

[taj (othertaj)]

Just to keep the last post alive:
I also have this problem, and I know how to work around it.

But... The fix is known for more than a month now, so how many more time is needed (and novice users put off) before the update?

[Michael R. Head (burner)]

This seems to have been fixed in gutsy. I can now use the UUID form in /etc/fstab for my swap partition.

[Launchpad Janitor (janitor)]

Per a decision made by the Ubuntu Kernel Team, bugs will longer be assigned to the ubuntu-kernel-team in Launchpad as part of the bug triage process. The ubuntu-kernel-team is being unassigned from this bug report. Refer to https://wiki.ubuntu.com/KernelTeamBugPolicies for more information. Thanks.