oom_adj

[Jordi Polo]

I am pretty confident that if someone can forward this to the kernel mailing
list some developers there will be very interested in the OOM making wrong
decisions in KDE pathology.


On Tue, Feb 17, 2009 at 11:40 PM, Lubos Lunak <l.lunak at suse.cz> wrote:

> On Saturday 14 of February 2009, John Tapsell wrote:
> > Hi all!
> >
> >  We could identify important processes (xorg, kwin, krunner) that we
> > wouldn't want to be killed in a low memory situation, and adjust their
> > oom_adj property.  This is used by the kernel out-of-memory handler
> > when it tries to kill processes because the system is low on memory.
> >
> > Thoughts?
>
>  We already had this in KDE3: The OOM-murderer in Linux sucks in a special
> way
> when it comes to KDE, since it selects processes that launch other
> processes.
> That means kdeinit, and, in KDE4's case when there are many
> session-restored
> apps, also ksmserver. That also means there's no point in protecting KWin
> or
> KRunner, since those should never accumulate high enough score unless they
> are really broken. It is a question if we should and can protect Xorg,
> since
> the kernel can't tell which X client is the one creating too many X
> resources
> and causing Xorg to use a lot of memory.
>
>  As for kdeinit, we had protection in KDE3, see
>
> http://websvn.kde.org/branches/KDE/3.5/kdelibs/kinit/start_kdeinit.c?view=log#rev579164,
> this commit and the following fixes could be ported to KDE4. At least I
> don't
> think that anything has changed since KDE3. Ksmserver could use something
> similar.
>
> --
> Lubos Lunak
> KDE developer
> --------------------------------------------------------------
> SUSE LINUX, s.r.o.   e-mail: l.lunak at suse.cz , l.lunak at kde.org
> Lihovarska 1060/12   tel: +420 284 028 972
> 190 00 Prague 9      fax: +420 284 028 951
> Czech Republic       http://www.suse.cz
>



-- 
Jordi Polo Carres
NLP laboratory - NAIST
http://www.bahasara.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20090218/11acf877/attachment.htm>