Unknown/unaccounted memory leak

[idyllic]

I have HP ProBook 4310s loaded with Linux Mint 7 Gloria (Jaunty 9.04) with all the latest updates and ATI Proprietary driver (for ATI Mobility Radeon 4330).

I am currently having some weird memory leak issue. Basically, the total memory usage from `free -m` or gnome-system-monitor doesn't tally with the sum of all processes' memory. The difference is a lot, from 100mb to 1gb. It kept on growing out of no where.

I have attached the screenshots of top and free.
---
Architecture: i386
DistroRelease: Ubuntu 10.04
DkmsStatus:
 fglrx, 8.723.1, 2.6.32-21-generic-pae, i686: built
 fglrx, 8.723.1, 2.6.32-22-generic-pae, i686: installed
InstallationMedia: Kubuntu 9.10 "Karmic Koala" - Release i386 (20091028.5)
MachineType: Hewlett-Packard HP ProBook 4310s
NonfreeKernelModules: fglrx
Package: fglrx-installer (not installed)
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.32-22-generic-pae root=UUID=44c7c661-1f5e-49e6-a14e-7a35f1ec3f9f ro quiet splash
ProcEnviron:
 LANGUAGE=
 PATH=(custom, user)
 LANG=en_GB.utf8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.32-22.33-generic-pae 2.6.32.11+drm33.2
Tags: lucid lucid
Uname: Linux 2.6.32-22-generic-pae i686
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
dmi.bios.date: 01/14/2010
dmi.bios.vendor: Hewlett-Packard
dmi.bios.version: 68PHT Ver. F.0C
dmi.board.name: 7016
dmi.board.vendor: Hewlett-Packard
dmi.board.version: KBC Version 14.10
dmi.chassis.asset.tag: CNU9287VXQ
dmi.chassis.type: 10
dmi.chassis.vendor: Hewlett-Packard
dmi.modalias: dmi:bvnHewlett-Packard:bvr68PHTVer.F.0C:bd01/14/2010:svnHewlett-Packard:pnHPProBook4310s:pvrF.0C:rvnHewlett-Packard:rn7016:rvrKBCVersion14.10:cvnHewlett-Packard:ct10:cvr:
dmi.product.name: HP ProBook 4310s
dmi.product.version: F.0C
dmi.sys.vendor: Hewlett-Packard
system:
 distro: Ubuntu
 codename: lucid
 architecture: i686
 kernel: 2.6.32-22-generic-pae

[Vancouverite (sethgilchrist)]

This bug affects me too. I posted about this on the ubuntu forums many months ago and got a "linux memory use is different from what you are used to" except that linux is really all I'm used to.

If I examine 'free', 'top', 'htop', system monitor, '/proc/meminfo', etc, they all say that I'm using about 2 GB (NOT including buffers or cache, with those it tops 3.3 GB). However, when I print 'ps aux' to a file and add up all the memory listed in there it only comes to 950 MB. So where has the rest gone?

I am running a fresh (within 24 hrs), up-to-date Karmic official release install and this really bums me out! I was hoping that the problem would vanish when I up graded to 9.10. I had the same issue in Jaunty.

I have attached the output of 'cat /proc/meminfo', 'free -t' and 'ps aux'. I imported the ps_aux data into OOCalc and added up all the process memory and it came to 25% or 993 MB. Free shows nearly 2 GB in use (3.3 GB with buffers).

Karmic 64 bit
Lenovo T400
4 GB RAM
ATI Discreet graphics with fglrx driver (note: when I had this problem in 9.04, I was using integrated intel graphics)

[idyllic (idyllic)]

Reproducible in a fresh-install Karmic.

[idyllic (idyllic)]

Vancouverite, yes I am still having the same problem in a fresh install Karmic 32bit (Kubuntu) with fglrx driver. I will attached my free -m again later after leaving for an hour or so.

[Vancouverite (sethgilchrist)]

I discovered an interesting behavior last night. If I let the memory get really high, then start the program ParaView (compiled from source www.paraview.org), even if I don't do anything - literally start it and quit it - the captured memory is released.

To display this I played Freedoom for about an hour and when I quit that I was using 1.7 GB memory with just the desktop open. I saved the output of 'ps aux' and 'free' and then started and quit ParaView as soon as it was loaded. Low and behold, 800 MB of memory spontaneously freed up! I saved the output of the same commands immediately.

I inspected the process list in each case and there were no differences between the two, so wherever that memory was tied up it was not accounted for in the process tree.

I don't know enough about this stuff to even postulate on what is going on, but in my view this represents a fairly serious bug. Is there another place I should look for where this memory may be accounted?

[Vancouverite (sethgilchrist)]

Hi idyllic,
I have isolated the issue to ATI's powerplay. If I turn off powerplay the leak goes away.

Van

[idyllic (idyllic)]

Hi Van,

Disable Powerplay will force the graphic card to run at its maximum clock. Is there a way to always set to lowest powerstate before disable powerplay?

idyllic

[Vancouverite (sethgilchrist)]

Hi idyllic,
I found that if I boot with the power saving off the leak does not appear. The PowerSmart feature can be enabled dynamically, so I can just switch it on after booting and be leak free the whole time! Now I just have to remember to turn it off before shutting down.

Van

[Felix Kuehling (felix-kuehling)]

Anyone experiencing this problem on Karmic, check if you're affected by https://bugs.launchpad.net/ubuntu/+source/ureadahead/+bug/501715

The "memory leak" occurs after system updates that trigger ureadahead to reprofile. Rebooting fixes the problem. A workaround is to reboot the system twice after updates.

[idyllic (idyllic)]

Hi Felix,

It is a separate issue. I have no problem with readahead. The problem lies in POWERPLAY. Only workaround now is to follow Vancouverite's advice.

Cheers,

[Bryce Harrington (bryce)]

[This is an automatic notification.]

Hi idyllic,

This bug was reported against an earlier version of Ubuntu, can you
test if it still occurs on Lucid?

Please note we also provide technical support for older versions of
Ubuntu, but not in the bug tracker. Instead, to raise the issue through
normal support channels, please see:

    http://www.ubuntu.com/support

If you are the original reporter and can still reproduce the issue on
Lucid, please run the following command to refresh the report:

  apport-collect 432531

If you are not the original reporter, please file a new bug report, so
we can work with you as the original reporter instead (you can reference
bug 432531 in your report if you think it may be related):

  ubuntu-bug xorg

If by chance you can no longer reproduce the issue on Lucid or if you
feel it is no longer relevant, please mark the bug report 'Fix Released'
or 'Invalid' as appropriate, at the following URL:

  https://bugs.launchpad.net/ubuntu/+bug/432531

[Allcolor-g (allcolor)]

It still affects my system... in about two hours, my 4Gb are entirely filled (but not by any process, memory is just lost) and swap start to fill... when that happen iowait grows to 90% more and render the system unresponsive. That does not happen with the free open source radeon driver, only with the proprietary fglrx. But since compiz works now with the radeon driver and my ati 4650, it's not so much of a problem...

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[idyllic (idyllic)]

apport information

[Launchpad Janitor (janitor)]

[Expired for fglrx-installer (Ubuntu) because there has been no activity for 60 days.]