Xorg eats my memory til its all gone.

turtles · Veteran Joined: 31 Dec 2004 Posts: 1655

I am trying to trouble shoot why X eats up all my memory and slowly soft locks up.
I have done all the usual stuff. emerge -e world, check kernel etc.. downgrade to stable stuff.
Can I startx in a memory container so I can recover from the lock up? I am able to get the following meaning full data:

Top, vmstat, and /proc/meminfo are what I have been using.

Before X:

eccerr0r · Posted: Thu Nov 10, 2011 7:52 pm Post subject:

What makes you think it's X11 that's eating all the memory?

By your two 'top' reports, the first one X11 is eating 9.1M, the second, 10M -- not that big a difference.

What you want to do is sort by 'memory used' - usually bound to shift-M.

From your firefox, I can only see firefox eating 42MB RAM, which isn't a whole lot in 2GB... there must be something else. Could be a memory leak or kernel slabs... slabtop will show these.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

turtles · Veteran Joined: 31 Dec 2004 Posts: 1655

I am not sure how to decipher this:
Here is slabtop before logging in:

chithanh · Posted: Thu Nov 10, 2011 9:34 pm Post subject:

Whether really X or some X application is at fault you can check with xrestop.
If you suspect the kernel DRM side, enable the appropriate sysfs debug options and look at /sys/kernel/debug/dri/0/gem_objects

eccerr0r · Posted: Fri Nov 11, 2011 3:50 pm Post subject:

The slabs usage doesn't look way too out of line, does not look like a kernel memory leak. So, have to look at user space memory usage.

Still do not have conclusive evidence as to what's eating so much RAM. Running 'top' and sort via memory (default is by CPU use) may give some more information.

BTW if you're using that much RAM, you're probably running out of RAM. Is the OOM killer killing anything?

OH... BTW. When killing individual processes by PID you should be using 'kill' not 'killall'. The latter is for killing by process name. Also I think baselayout/openrc is probably doing a good job with killing processes and -9'ing them probably wouldn't help... this does imply there are processes in uninterruptable sleep... "killall -9 kdm" is probably what you're trying to do.

turtles · Veteran Joined: 31 Dec 2004 Posts: 1655

looks like zillions of mount.crypt processes eating up my memory only after login using kdm to xfce or kde.
How do I run fsck or e2fsck on an encrypted partition? my /home/dir is encrypted and may be the issue.
I tried unmounting and using cryptsetup lukesopen /dev/sda* dev-name
and got errors
any advice on this?
I cant seem to log a error I get on boot up about a kernel supplied name conflict with a dm something.

Is there a way to get rc.log to capture all of the boot up?
or get this error somewhere from udev?

I have looked in rc.log mesages and dmesg.

I am using kdm to login (I have used gdm and slim in the past).
Seems like some config is borked with the dbus udisks or the encrypted filesystem has gone buggy. How can I tell when it is being checked on mount?
Odd thing is I cant seem to get another test user account to be able to get past kdm and get a xsession and I am not able to startx as root, however everything seems fine running TWM.

turtles · Veteran Joined: 31 Dec 2004 Posts: 1655

Seems like it is related to this: