gnome-system-monitor spams dbus with requests for volume information

[Alistair Buxton]

Binary package hint: gnome-system-monitor

Steps to reproduce:

1. open gnome-system-monitor
2. switch to the "file systems" tab
3. in edit->preferences tick "show all filesystems" and set update interval to 1 second

gnome-system-monitor will now spam dbus with volume information requests until dbus eventually refuses to honour any more. At this point it will start filling .xsession-errors with error messages. This behaviour also seems to cause gfvsd to use incredible amounts of memor.

Showing all filesystems seems to be required to trigger the bug on some systems - notably clean VM installs with not much mounted filesystems. But when it's turned on the bug happens on lucid, maverick and natty clean/fully patched installs - although the leak happens very slowly in this case - about 1mb per minute.

On my main workstation (which is maverick) it's a different story - with "optimal" g-s-m settings, gvfs can leak 1GB in 10 minutes.

ProblemType: Bug
DistroRelease: Ubuntu 10.10
Package: gnome-system-monitor 2.28.1-1ubuntu1
ProcVersionSignature: Ubuntu 2.6.35-25.44-generic 2.6.35.10
Uname: Linux 2.6.35-25-generic x86_64
NonfreeKernelModules: nvidia
Architecture: amd64
Date: Tue Apr 5 15:54:06 2011
InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release amd64 (20101007)
ProcEnviron:
 PATH=(custom, user)
 LANG=en_GB.utf8
 SHELL=/bin/bash
SourcePackage: gnome-system-monitor

[Alistair Buxton (a-j-buxton)]

.xsession-errors showing the error spam.

[Alistair Buxton (a-j-buxton)]

This problem seems to very quickly expose the memory leaks in the various gvfs backends since it constantly hammers them with requests, see https://bugs.launchpad.net/ubuntu/+source/gvfs/+bug/433500

[Alistair Buxton (a-j-buxton)]

Mounted gvfs filesystems on my main workstation where the problem is much worse than on a clean install.

[Alistair Buxton (a-j-buxton)]

This is the valgrind log of gvfsd from my maverick workstation with the bug in full effect. Leaking 1.5GB took about 30 minutes. Here's the relevant part:

==18126== 1,590,801,664 (1,795,104 direct, 1,589,006,560 indirect) bytes in 9,756 blocks are definitely lost in loss record 480 of 480
==18126== at 0x4C2815C: malloc (vg_replace_malloc.c:236)
==18126== by 0x57B965A: dbus_message_new_empty_header (dbus-message.c:1079)
==18126== by 0x57BD1C7: dbus_message_new_method_return (dbus-message.c:1241)
==18126== by 0x4085E7: dbus_message_function (mount.c:927)
==18126== by 0x57BEE4D: _dbus_object_tree_dispatch_and_unlock (dbus-object-tree.c:858)
==18126== by 0x57B248B: dbus_connection_dispatch (dbus-connection.c:4691)
==18126== by 0x4175E4: message_queue_dispatch (dbus-gmain.c:127)
==18126== by 0x5C41341: g_main_context_dispatch (gmain.c:2149)
==18126== by 0x5C452A7: g_main_context_iterate (gmain.c:2780)
==18126== by 0x5C457B4: g_main_loop_run (gmain.c:2988)
==18126== by 0x409099: main (main.c:128)

[Alistair Buxton (a-j-buxton)]

I was watching the session bus with dbus-monitor and i noticed that every URL i've clicked on in pidgin at least since the last reboot shows up as a dbus object under gvfs, and g-s-m constantly polls this stuff. That probably explains why the problem is so much worse on my workstation, i've clicked on a *lot* of URLs in the past month...

[Alistair Buxton (a-j-buxton)]

Some more investigation... valgrind points to "listMounts" dbus call as the culprit of the leaks. This is the same call that is full of old pidgin URLs and other cruft. I wrote a python script to call it and gvfsd leaks 200kb every time. But the real killer.....

gnome-system-monitor calls "listMounts" 17 times for every update, causing nearly 2mb to leak for every update.

I've sent the bug report and testcase for gvfsd upstream.

[Launchpad Janitor (janitor)]

This bug was fixed in the package gvfs - 1.8.0-0ubuntu2

---------------
gvfs (1.8.0-0ubuntu2) natty; urgency=low

  * 90_git_leaks_fixes.patch: git commit to fix leaks (lp: #751523)
 -- Sebastien Bacher <email address hidden> Fri, 15 Apr 2011 15:39:56 +0200

[Néstor A.C. (nestorac)]

Is this bug going to be fixed also in Ubuntu 10.04 LTS? It's also happening to me, and as you know, it's serious.

Thanks!

[Andrew Somerville (andy16666)]

Same question about the LTS release. Will we get a fix?

[Dmitri Bachtin (damg)]

The reported patch [2] applies successfully on lucid from bzr repo [1] and the gvfs source builds successfully on i386. I could not verify whether the patch fixes the problem due to my time constraints.

[1] - lp:ubuntu/lucid/gvfs
[2] - http://git.gnome.org/browse/gvfs/commit/?id=7460f26f9243e6b1745a448b8dc05b15d12aa86e

[Robert Roth (evfool)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. However, I am closing it because the bug has been fixed in the latest development version of Ubuntu - Precise Pangolin.

If you need a fix for the bug in previous versions of Ubuntu, please follow the instructions for "How to request new packages" at https://help.ubuntu.com/community/UbuntuBackports#request-new-packages