Compiz or Firefox

[Leonard Smith]

On Tue, 2008-07-29 at 21:49 +0200, Mario Vukelic wrote:
> On Tue, 2008-07-29 at 14:01 -0500, Leonard Smith wrote:
> > Yeah, let me clarify.
> 
> Thanks
> 
> >  I first noticed the problem with firefox and
> >   vmware. I quit using vmware (as it was virtually unusable) and tried
> >  virtualbox. Seemed fine, at first, but then started having the same
> >   issue with it. Now I've noticed it happen anytime I have two or more
> >  of these applications running at the same time.
> 
> Ok, I start to see the picture now. 
> 
> Can you rule out completely that if you run any one of these program on
> their own, the CPU issue does not occur?
> Did you maybe test by first running each program without the others,
> then running two of them in all possible pairings, and after that maybe
> even three of them in the possible triplets?
> 
> I find it highly unlikely that arbitrary programs interact in such a
> way, after all their address spaces are kept separate by the kernel.
> IMHO the most likely culprit is one of the kernel modules. vmware has
> one, your graphics driver module has one (since you run hardware
> acceleration for compiz; if your card is not Intel, these drivers are
> also proprietary, making them impossible to debug for kernel
> developers). And Virtualbox also has one.
> 
> You could try reducing these module to the minimum: unload the vmware
> module(s) and virtualbox module(s). Also, disable the proprietary driver
> for X (if you use one. this is easiest done from menu System >
> Administration > Hardware Drivers).
> 
> Then try to trigger the issue with one, two, three of the programs
> running. I know it's a lot of work, but I have no idea how else to
> systematically narrow it down.
> 
> 

I did find a problem. The GUI on my laptop locked up today, but I
managed to switch to a console run top. I couldn't pinpoint the
offending process(es), so I did the ctrl+alt+backspace to restart X. I
noticed in the process that the system was not using any swap. Did a
little checking and found out that it was having some trouble with UUID
on the swap partition listed in fstab. I made sure swap was off (swapoff
-a), reformatted the swap partition, and tried to turn swap back on
(swapon -a). Still had trouble with the UUID, so I edited fstab to point
swap to /dev/sda6 instead of the UUID. Bingo. The system recognized the
swap partition and started using a little bit. I watched it for a while
and seems to be working now.

I don't know whether this actually fixed the problem or not, but we'll
see.

Leonard