When I run compiz, the cached memory grows continuously until the system becomes unusable. I suspect this to be a memory leak. I've found that this is related to UXA, KMS used or not, and happens with all the driver versions I have tested: -- The official Jaunty 2.6.1 -- The 2.7.0 from ubuntu-x-swat (launchpad) -- The 2.7.99 from xorg-edgers (launchpad) OpenGL version string: 1.4 Mesa 7.5-rc1 (happens with default Jaunty too) X.Org X Server 1.6.0 This doesn't happen when using EXA or without compiz. To reproduce: Do some window open/close/resize/… I use this bash line: while true; do firefox & sleep 3; pkill firefox; done You could see the cached memory grow quickly and about one minute after I'm out of memory (on 1GB). The attachments and details of how I investigated are on the original bug report given as URL: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/376092 Summary: This is neither related to architecture (both i386, amd64) nor (actual) kernel version (Jaunty official / 2.6.30-rc). The external monitor doesn't play any role here. It seems that the leak go with the resolution, bigger screen, quicker run out. I hope to be complete…
Created attachment 25990 [details] Xorg log without external screen Here is the Xorg.0.log when starting X using KMS and the 2.7.99 driver.
lscpi from downstream bug report: 00:00.0 Host bridge [0600]: Intel Corporation Mobile 945GM/PM/GMS, 943/940GML and 945GT Express Memory Controller Hub [8086:27a0] (rev 03) Subsystem: Micro-Star International Co., Ltd. Device [1462:0341] 00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller [8086:27a2] (rev 03) Subsystem: Micro-Star International Co., Ltd. Device [1462:0341]
Seeing the same dramatic symptoms with a i915 card, kernels 2.6.30rc5 and rc6, drivers 2.6.99 (still Ubuntu). UXA is working fast with this release, though.
Thanks for the specifics on how to reproduce it! I'll try that.
Suffering from the same memory leaks, using 2.6.30-rc7, gm965/GL960 rev0c, font glyph patch applied (see bug 21790). Xorg-server 1.6.1.901, everything else current git versions. And like stated in comment #3, UXA is really fast ;-)
Just an idea: may a ssh connexion help you find the specificity of my hardware that make this happen? I've a DynDns account and ssh enabled but my machine must be on, so just ask if this could help and I will give you my DNS name and let the computer on when you want. BTW did you success in reproducing this issue?
Any idea about how a mem leak can lead to growing caches? That's not obvious to debug, since X.org's mem usage doesn't really increase. But those caches can't be dropped, to the point that hibernating is no longer possible.
this issue really looks like bug #20704
Still an issue with 2.7.99.901+git20090610.9d3c3b05-0ubuntu0sarvatt~jaunty package from xorg-edgers. Yes symptoms are the same as in the other bug... but this is only with UXA, KMS or not. I'm not sure this is a duplicate but may. Had the fix for bug #20704 been pushed and integrated before git 20090610? If yes then this is a different issue... Yesterday, I've seen this problem with metacity too: playing some time with google maps' Street View reproduces this. But it's only slowly growing.
(In reply to comment #9) > Still an issue with 2.7.99.901+git20090610.9d3c3b05-0ubuntu0sarvatt~jaunty > package from xorg-edgers. > > Yes symptoms are the same as in the other bug... but this is only with UXA, KMS > or not. I'm not sure this is a duplicate but may. > Had the fix for bug #20704 been pushed and integrated before git 20090610? If > yes then this is a different issue... > > Yesterday, I've seen this problem with metacity too: playing some time with > google maps' Street View reproduces this. But it's only slowly growing. > the fix for bug #20704 is still waiting for review, so not pushed yet
The fix for bug #20704 has been committed into mesa. Please verify.
Verified. As I reported on the downstream bug report the patch(es) fix the issues I saw in this bug report as well as in #20704. (I'm using the 7.5 branch of mesa).
commit d027e8feff7d38cccadc6aaccc0454b21ce4dca0 Author: Shuang He <shuang.he@intel.com> Date: Mon Jun 15 16:19:30 2009 -0600 intel: Release fb backing regions in intelDestroyBuffer() Fixes memory leak when destroying framebuffers.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.