Re: hard-lock with CPU spinning

On Friday 22 June 2007 08:41:54 am Giorgos Keramidas wrote:
> On 2007-06-17 19:32, Giorgos Keramidas <keramida_at_freebsd.org> wrote:
> >On 2007-06-14 20:02, Giorgos Keramidas <keramida_at_ceid.upatras.gr> wrote:
> >>On 2007-06-14 18:36, Attilio Rao <attilio_at_freebsd.org> wrote:
> >>>2007/6/14, Giorgos Keramidas <keramida_at_ceid.upatras.gr>:
> >>>> If I leave my laptop idle for a long period of time, it tends to lock 
up
> >>>> with the CPU fan spinning fast (presumambly because some part of the
> >>>> kernel tries to acquire a lock and spins constantly for it).
> >>>>
> >>>> Unfortunately, this happens when X11 is running and I can't break into
> >>>> DDB to snoop around.
> >
> > Hi Attilio,
> >
> > thanks for the eagerness to help, but I was too quick in assuming this
> > was a hard-lock.  The kernel hasn't deadlocked, but the laptop is almost
> > unresponsive because the X server eats up an enormous amount of CPU.
> >
> > I left an xterm window running:
> >
> >     > cd /home/keramida
> >     > ( while true ; do \
> >             uptime ; ps xaur | head -20 ; \
> >             sleep 5 ; echo ; \
> >         done ) 2>&1 | tee logfile
> >
> > and when hte CPU fan started spinning fast, I managed to shutdown
> > normally by pressing the laptop's power-off button and waiting long
> > enough for the X process to die.
> >
> > The ~/logfile file contains near its end entries like:
> >
> > %  6:43PM  up  2:05, 1 user, load averages: 0.76, 0.39, 0.24
> > % USER       PID %CPU %MEM   VSZ   RSS  TT  STAT STARTED      TIME COMMAND
> > % root      1234 97.8  4.2 285648 21428  v1  R     4:41PM   3:22.41 
X :0 -dpi 96 (Xorg)
> > % root        12 97.1  0.0     0     8  ??  RL    4:37PM 112:19.80 [idle: 
cpu0]
> > % root        11  2.2  0.0     0     8  ??  RL    4:37PM 110:16.80 [idle: 
cpu1]
> 
> Finally, more progress :)
> 
> This seems to kick in only when I use:
> 
>     % xset +dpms
>     % xset s on
>     % xset b 100 800 20
> 
> By disabling DPMS with '-dpms' there is no CPU-eating behavior
> even after leaving my laptop on for hours.
> 
> So this seems to be a bug in the +dpms part of X11 :-)

I'll have to try that.  My laptop has a similar issue except it seems to be 
cpufreq related (if I disable powerd I don't see the behavior).  I don't use 
xset, but I probably have dpms enabled in KDE.

-- 
John Baldwin