Mem leaks with last 260 and 270 series

[wodor]

Hi,
I have huge mem leaks in Xorg memory space when I use close drivers. It is possible that its depend of suspend or wine.

Some information about my hardrawe and software, and logs.

hardware:
01:00.0 VGA compatible controller: nVidia Corporation GT218 [NVS 3100M] (rev a2)

software:
xorg 1.9.4
kernel : 2.6.37
dist: arch linux 64 bits (updated stable)
I use compiz last compilation
.

kernel parameters: linux /boot/vmlinuz-2.6.37 root=/dev/sda1 ro nouveau.modeset=0 acpi_sleep=nonvs resume=swap:/dev/sda7 resume=/dev/sda7

graphic mem usage:
Memory Usage
Total : 511 Mb
Used : 301 Mb
Free : 209 Mb

xorg process:

12348 root 20 0 835m 720m 21m R 6 12.1 48:19.03 Xorg

[root@h2 Desktop]# free
total used free shared buffers cached
Mem: 6103408 5501452 601956 0 746896 2583812
-/+ buffers/cache: 2170744 3932664
Swap: 2050044 4600 2045444

Xorg logs:
http://wklej.org/id/491889/

[root@h2 Desktop]# dmesg | grep nv
nvidia 0000:01:00.0: restoring config space at offset 0xc (was 0xe3000000, writing 0x0)
nvidia 0000:01:00.0: restoring config space at offset 0x3 (was 0x800010, writing 0x800000)


best regards,

[Licaon]

can you read http://www.nvnews.net/vbulletin/showthread.php?t=46678 and generate a good report with the debug ( -logverbose 6 ) info as instructed after the issue appears ?

[danix]

Please try to isolate a workflow that reproduces the issue and provide detailed instructions on how to reproduce it. Thanks.

[wodor]

Hi, I still have this problem on my linux machine. I use xfce4 and applications like conky, opera and psi. I attached bug log.


Three months ago I found additional one bug. When I used tool to change resolution from gnome2 on toolbar I had big mem leaks in Xserver when content menu from toolbar apper.


ps. reproduction: leave computer alone for some time (4 or 5 hours) ;-)

best regards,

[DeiF]

Just this morning I noticed that my system was being sluggish (a while after resuming from suspend IIRC), so I opened up "top" to see what was happening.
I saw Xorg using 1400MB of virtual memory, 1300MB of "res" (and a little less shared, I think ~1200MB). That, coupled with other programs (e.g. Firefox) was triggering swap usage, since I only have 2GB of RAM.
So I launched xrestop to see what was the program storing so much pixmaps into Xorg. To my surprise the most demanding app was "plasma-desktop" eating only 50MB, and the rest being much much less.
After a quick googling, I then found the pmap command, so I did "pmap -d xorgpid" and found 2 lines near the top to be the most consuming. One was marked as [heap] and the other [anon].
I stopped there, as I didn't know how to further investigate this.

If I'm able to reproduce this I'll comment further.

[AaronP]

Originally Posted by DeiF

Just this morning I noticed that my system was being sluggish (a while after resuming from suspend IIRC), so I opened up "top" to see what was happening.
I saw Xorg using 1400MB of virtual memory, 1300MB of "res" (and a little less shared, I think ~1200MB). That, coupled with other programs (e.g. Firefox) was triggering swap usage, since I only have 2GB of RAM.
So I launched xrestop to see what was the program storing so much pixmaps into Xorg. To my surprise the most demanding app was "plasma-desktop" eating only 50MB, and the rest being much much less.
After a quick googling, I then found the pmap command, so I did "pmap -d xorgpid" and found 2 lines near the top to be the most consuming. One was marked as [heap] and the other [anon].
I stopped there, as I didn't know how to further investigate this.

If I'm able to reproduce this I'll comment further.

How many pixmaps did xrestop say plasma-desktop was using? The total memory size number reported by xrestop is not always accurate. It would be interesting the next time this happens to try killing plasma-desktop (and other applications, one by one) to see if any one of them causes the X server to free that memory.

There are resources other than pixmaps that could be taking up that memory too.

[DeiF]

Originally Posted by AaronP

How many pixmaps did xrestop say plasma-desktop was using?

Sorry, didn't remember.

Originally Posted by AaronP

The total memory size number reported by xrestop is not always accurate. It would be interesting the next time this happens to try killing plasma-desktop (and other applications, one by one) to see if any one of them causes the X server to free that memory.

There are resources other than pixmaps that could be taking up that memory too.

I think I killed plasma-desktop IIRC but the memory stayed almost the same.
Thanks for the tips, though. I will try that the next time it happens.

[wodor]

Originally Posted by DeiF

Sorry, didn't remember.


I think I killed plasma-desktop IIRC but the memory stayed almost the same.
Thanks for the tips, though. I will try that the next time it happens.

Hi,
I had that same problem with gtk-window-decorator from compiz. I killed this proces and Xserver didnt release memory.

best regards,

[wodor]

Today I had another one huge memleak.

Howto reproduce : maximalize and unmaximalize windows like opera, terminal, gvim etc.
I made it under compiz too and I just inform compiz developers.
ps.

11861 root 20 0 927M 834M 19368 S 12.0 14.0 7:36.92 /usr/bin/X :0 vt7 -nolisten tcp -auth /var/run/xauth/A:0-Mj0ylc

best regards,

[jeisom]

Don't know if this is useful, but running 270.41.19 with 2.6.39.1 kernel on amd64.
free with most everything shutdown(X and most unrequired daemons)

             total       used       free     shared    buffers     cached
Mem:       6114788    3769372    2345416          0        440    1462240
-/+ buffers/cache:    2306692    3808096
Swap:      7802876         44    7802832

Attached is a log from kmemleak. Don't know what took the ram though 3.7 gigs is alot for only what was running at the time.

I'll try to reproduce.

[Licaon]

bad post

[wodor]

With 280.04 too

best regards,