CentOS - May 2010 - trying to rebuild an old nvidia driver

Let me start by saying that no, I can't get a new card for my user.

So, I've got the 96 nvidia driver. I've rebuilt it several times before,
and Nvidia's installer's always worked fine. Actually, I have the 73 on
my
system, and just updated to 5.5, and no problem.

HOWEVER, I last rebuilt my user's driver back in mid-April, after the last
kernel update (was that mid-March?), and everything was fine... until the
other day, when he walked in to tell me he'd had to reboot, and it
wasn't
working, and dropped him to command line.

I rebuilt it... and it dies. As in, sometimes, I have to go to another
machine and ssh in to reboot, and sometimes, I have to power cycle his
system. I get garbage with the nvidia driver; with nv, I can at least get
one of his two monitors working.

When I try x with the nvidia driver, it just keeps printing out
initializing GART, I think it is (I've lost the log), and then that it has
failed to allocate ROP.

Anyone see anything like this? I've unpacked the executable, and started
working with the makefiles (and yes, I do understand makefiles), but see
things like a lot of warning in the compile, and the one that makes me
nervous is
  make CC=cc KBUILD_OUTPUT=/lib/modules/2.6.18-194.3.1.el5/build
KBUILD_VERBOS
   E=1 -C /lib/modules/2.6.18-194.3.1.el5/source
SUBDIRS=/scratch/NVIDIA-Linux-
   x86-96.43.13-pkg1/usr/src/nv modules
   make -C /lib/modules/2.6.18-194.3.1.el5/build \
        KBUILD_SRC=/usr/src/kernels/2.6.18-194.3.1.el5-i686 \
       
KBUILD_EXTMOD="/scratch/NVIDIA-Linux-x86-96.43.13-pkg1/usr/src/nv"
-f /usr/
   src/kernels/2.6.18-194.3.1.el5-i686/Makefile modules
   test -e include/linux/autoconf.h -a -e include/config/auto.conf || (   
     \
        echo;                                                           \
        echo "  ERROR: Kernel configuration is invalid.";             
\
        echo "         include/linux/autoconf.h or
include/config/auto.conf are mis
   sing.";      \
        echo "         Run 'make oldconfig && make prepare'
on kernel src
to fix it
   .";  \
        echo;                                                           \
        /bin/false)

Even though both those files exist and are where they should be. I've
googled, and find no references to those, and really hope it's not the
card going bad.... So, any clues?

         mark

On Wed, May 19, 2010 at 2:05 PM,  <m.roth at 5-cent.us>
wrote:
> Let me start by saying that no, I can't get a new card for my user.
No problem with that.
> So, I've got the 96 nvidia driver. I've rebuilt it several times
before,
> and Nvidia's installer's always worked fine. Actually, I have the
73 on my
> system, and just updated to 5.5, and no problem.
I suggest you give the ELRepo package a try:

http://elrepo.org/tiki/kmod-nvidia-96xx

Akemi

mark wrote:
Akemi wrote:
>> On Wed, May 19, 2010 at 2:05 PM,  <m.roth at 5-cent.us> wrote:
>>> Let me start by saying that no, I can't get a new card for my
user.
>
>> No problem with that.
>
>>> So, I've got the 96 nvidia driver. I've rebuilt it several
times before,
>>> and Nvidia's installer's always worked fine. Actually, I
have the 73
on my
>>> system, and just updated to 5.5, and no problem.
>
>> I suggest you give the ELRepo package a try:
>
>> http://elrepo.org/tiki/kmod-nvidia-96xx
> Thanks! I'll try that tomorrow.
*sigh*
Well, I did that, and then went through a song-and-dance uninstalling the
stuff the nvidia installer had put in, and uninstalling this so that would
succeed, and then *really* getting rid of the stuff it wouldn't, and then
using yum again.

And it still got trash. My user's sure he had not turned it off, that he
just came in one morning and it had lost it, and I have every reason to
believe him (he *does* know what he's doing).

So, at this point, we've given up - his machine wasn't on a UPS, and
given
the power in this building, there's a high probability of a surge, and
something on the card got fried - so he's got another machine, and we're
probably going to call this one dead.

Thanks, though, for the el repo link. I've got to remember that one.

       mark