[CentOS] Xorg goes bonkers, bazillion NVIDIA errors...

Mon Dec 7 23:58:40 UTC 2009
fred smith <fredex at fcshome.stoneham.ma.us>

on one of my Centos 5.4 boxes, a machine that's around 5 years old and
has always run Centos and X, and has had the same Nvidia card in it for
its entire life, I'm suddenly getting these lines in the xorg log file:

(WW) NVIDIA(0): WAIT (2, 6, 0x8000, 0x00000000, 0x000003c0)
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***
(EE) NVIDIA(0): Failed to allocate clip rectangle
(EE) NVIDIA(0):  *** Aborting ***

followed by millions more like them. it's currently sitting mostly
unused, except that it's running a Folding At Home client. it'll run
for a day or a week then I'll note the screen is in powersave mode, will
attempt to wake it up by moving the mouse or touching a key and nothing
happens. (for purposes of this testing I've disabled the screensaver,
so the screen should never shut off.) I can SOMETIMES ssh into it, if
it hasn't completely died yet, and if I can, top shows that Xorg is
consuming about 99% of the CPU instead of the folding at home client
which should be consuming most of it. I guess it's busy spewing errors
to the Xorg log file, or something.

The last time I found it like this I tried to kill xorg. kill -15 wouldn't
kill it so I tried kill -9, then the whole box seized up, requiring a
reset.

sometimes it's already unresponsive, even to ssh or ping, by the time I
notice the problem.

it's running an old Nvidia Geforce 4 MX400 card and the Nvidia proprietary
driver NVIDIA-Linux-x86-96.43.01-pkg1.run, downloaded directly from the 
nvidia web site. the date on that file is Oct 2007, so that means that's
the one that it has been using since then. but it's only the last 6 weeks
or so it's started doing this.

other than some vaguely-defined "nvidia driver problems", what may be
going on here?  (I can buy an nvidia conflict with the stock centos
kernel, but I'd be dubious, since it's been running this driver for
two years and earlier versions for 4 or 5. any such conflict would seem
likely a kernel issue, to me, since it's newly exhibited.)

Thanks!

Fred
-- 
---- Fred Smith -- fredex at fcshome.stoneham.ma.us -----------------------------
               But God demonstrates his own love for us in this: 
                         While we were still sinners, 
                              Christ died for us.
------------------------------- Romans 5:8 (niv) ------------------------------