Bug 65121

Summary: Possible memory leak in qxl drm driver
Product: Drivers Reporter: Matthew Stapleton (matthew4196)
Component: Video(DRI - non Intel)Assignee: drivers_video-dri
Status: RESOLVED CODE_FIX    
Severity: normal CC: airlied, alan, xerofoify
Priority: P1    
Hardware: x86-64   
OS: Linux   
Kernel Version: 3.11.6 Subsystem:
Regression: No Bisected commit-id:
Attachments: 3.11.9-config
fix

Description Matthew Stapleton 2013-11-18 02:24:58 UTC
The QXL DRM driver seems to have a memory leak on my current custom kernel build using Gentoo Hardened kernel: 3.11.6-hardened-r2 (also on 3.11.3-hardened).  This can be seen by running "watch 'cat /proc/slabinfo  | grep ^kmalloc'", and on my qemu virtual machine running on qemu-1.6.0, kmalloc-64 allocation will constantly increase until system runs out of ram and then OOM killer kills everything and causes the system to crash.  I should be able to test with a vanilla kernel in a few days if needed.
Comment 1 Alan 2013-11-18 14:42:15 UTC
Please do test a standard kernel, and if you can see if you can figure out if the leak is always occurring or seems to be triggered by some specific application or activity.
Comment 2 Matthew Stapleton 2013-11-21 05:27:36 UTC
I just tested with standard kernel: 3.11.9 and booted into a basic rescue initrd image with qxl enabled and "cat /proc/slabinfo  | grep ^kmalloc-64" still shows constantly increasing active_objs and num_objs.
Comment 3 Matthew Stapleton 2013-11-21 05:30:34 UTC
Created attachment 115361 [details]
3.11.9-config

This is the kernel config that was used
Comment 4 Alan 2013-11-26 22:06:02 UTC
Thanks for confirming
Comment 5 Dave Airlie 2013-11-28 05:29:33 UTC
what are you running just console?

or is X running?
Comment 6 Dave Airlie 2013-11-28 05:41:36 UTC
Created attachment 116501 [details]
fix

does this fix it?
Comment 7 Matthew Stapleton 2013-11-29 05:27:59 UTC
It looks like that fixes it.  Thanks.  Also, I'm only running on console, X isn't running.
Comment 8 xerofoify 2014-06-20 04:09:06 UTC
IF the attachment here seems to fix this I would close this.
Just a friendly reminder,
Nick
Comment 9 Dave Airlie 2014-06-20 04:34:45 UTC
kernel bugzilla sucks I can't change bug states.
Comment 10 xerofoify 2014-06-20 04:43:46 UTC
Our you sure, other people seem to 
be able to do it with other bugs.
Thanks Nick
Comment 11 xerofoify 2014-06-20 04:45:32 UTC
It seems to be only the opener or maintainer of this subsystem who can close it.
Cheers Nick
Comment 12 Alan 2014-06-20 15:35:28 UTC
Dave - ask the bugzilla contact address for your superman powers. I'm sure you'll have no trouble then