[Bug 90492] Xorg unresponsive under heavy system-load

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=90492

--- Comment #5 from Chris Wilson <chris at chris-wilson.co.uk> ---
(In reply to alexander.maznev from comment #4)
> low on resources?
> 
> > Your dmesg indicated the first failure with completely full swap. Though it
> > is a known issue that the kernel will start killing processes with swap
> > still available. One of the problems here is during the early oom, the GPU
> > is still active and itself waiting upon allocations (which is preventing it
> > from yielding memory back to the system).
> 
> Yes, I turned swap off afterwards to get more consistent behaviour (the
> thing with having swap on is if there is some very small amount of space, it
> will freeze while aggressively trying to swap stuff around to get memory, so
> the system hangs instead of erroring out of xserver). 

But X still can't do very much if the kernel is preventing the driver from
doing anything whilst it is processing oom.

> > You have oom_kill_allocating_task set in sysctl.
> 
> cat /proc/sys/vm/oom_kill_allocating_task 
> 0

$ grep oom_kill_allocating_task dmesg.txt 
[ 1481.886894] Out of memory (oom_kill_allocating_task): Kill process 540
(tilda) score 0 or sacrifice child
[ 1486.885295] Out of memory (oom_kill_allocating_task): Kill process 540
(tilda) score 0 or sacrifice child
[ 1491.883715] Out of memory (oom_kill_allocating_task): Kill process 540
(tilda) score 0 or sacrifice child
[ 1495.103841] Out of memory (oom_kill_allocating_task): Kill process 538
(tint2) score 0 or sacrifice child
[ 1495.104251] Out of memory (oom_kill_allocating_task): Kill process 1138
(Timer) score 0 or sacrifice child
[ 3405.538346] Out of memory (oom_kill_allocating_task): Kill process 5293
(systemd-coredum) score 0 or sacrifice child

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20150517/eda1420f/attachment.html>