KDE process is unkillable

Matthew Seaman m.seaman at infracaninophile.co.uk
Tue Jul 15 14:36:47 PDT 2003


On Tue, Jul 15, 2003 at 10:29:00PM +0200, daniela5743 at gmx.net wrote:
> > On 15 Jul daniela5743 at gmx.net wrote:
> > > I tried it multiple times. I was able to kill all the other KDE
> > > related processes (even the parent), but this one just doesn't die.
> > 
> > Then kill (-9) the login session itself (the one kde came from in the
> > first place)
> > 
> > And if that too does not help: a home server is very easely rebooted.
> > Nobody will notice or at least hardly..
> 
> Thanks for the advice, but I really don't want to kick my users off.
> If it is really necessary, I want to get a core dump before rebooting.
> Do you know how I could do this?
> 
> I'm trying to install lsof, but it takes forever (KDE takes up all the CPU
> time).
> Here's the output from fstat, maybe this solves the problem:
> 
> root     kdeinit    62100 root /             2 drwxr-xr-x     512  r
> root     kdeinit    62100   wd /usr     6690817 drwxr-xr-x    2048  r
> root     kdeinit    62100 text /usr     341416 -r-xr-xr-x  412176  r
> root     kdeinit    62100    0 /         65411 crw-------   ttyv3 rw
> root     kdeinit    62100    1 /         65411 crw-------   ttyv3 rw
> root     kdeinit    62100    2 /         65411 crw-------   ttyv3 rw
> root     kdeinit    62100    7 /         64909 crw-rw----  #C145:0 rw
> root     kdeinit    62100    8* pipe e390a2a0 <-> e390a520      0 rw
> root     kdeinit    62100    9* pipe e390a520 <-> e390a2a0      0 rw
> root     kdeinit    62100   10* pipe df266260 <-> df2652c0      0 rw
> root     kdeinit    62100   11* pipe df2652c0 <-> df266260      0 rw

Are you by any chance using the NVIDIA supplied drivers?  I had
exactly the same symptoms with X freezing, especially when doing
something that puts a bit of load on the system.  The recent upgrade
to the drivers actually made things worse.  In the short term, the
only way to get the console screen back was to reboot.

However, in the long term, it seems the problem was a bad interaction
with the FreeBSD AGP GART -- ie. the agp.ko kernel module.  I
recompiled the driver to use it's own built in AGP GART and since then
everything has been rock solid stable --- the x11/nvidia port makes it
easy to switch, but remember to edit /boot/loader.conf or your kernel
config to take out (or add) the agp module.  NVIDIA docs are a bit
equivocal about recommending one way or the other -- which way to
choose depends on the precise hardware you have.  Only way to tell is
by experiment.  There's also a sysctl hw.nvidia.agp.status.rate that
lets you control the agp card rate: it should all work automatically,
but you may find that dropping the rate will help stabilize things.
Of course, that will have a deleterious impact on graphics
performance.

	Cheers,

	Matthew

-- 
Dr Matthew J Seaman MA, D.Phil.                       26 The Paddocks
                                                      Savill Way
PGP: http://www.infracaninophile.co.uk/pgpkey         Marlow
Tel: +44 1628 476614                                  Bucks., SL7 1TH UK
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20030715/f5979e43/attachment.bin


More information about the freebsd-questions mailing list