Debian nvidia video problem

[Giles Orr]

On 11 February 2010 10:57, Lennart Sorensen
<lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org> wrote:
> On Thu, Feb 11, 2010 at 07:35:33AM -0500, Giles Orr wrote:
>> I'm running a Debian testing system on my main desktop.  Up until a
>> couple days ago it was running three monitors on two video cards (one
>> PCI-E and one PCI, both Nvidia 6200) on kernel 2.6.30-1-amd64 with the
>> nvidia drivers.  I have the latest nvidia-glx, module-assistant,
>> nvidia-kernel-common, and the headers for all installed kernels.
>> After an upgrade to kernel 2.6.32-trunk-amd64 I ran "m-a clean, a-i -t
>> nvidia -l 2.6.32-trunk-amd64" as I had with the previous kernel.  It
>> built without any obvious errors, but an attempt to run X gives this:
>>
>> (==) Using config file: "/etc/X11/xorg.conf"
>> dlopen: /usr/lib/xorg/modules/drivers/nvidia_drv.so: undefined symbol:
>> resVgaShared
>> (EE) Failed to load /usr/lib/xorg/modules/drivers/nvidia_drv.so
>> (EE) Failed to load module "nvidia" (loader failed, 7)
>> (EE) No drivers available.
>>
>> Fatal server error:
>> no screens found
>>
>> All attempts to build the module for other kernels have given the same
>> results: even reverting to the 2.6.30-1 kernel, and despite the fact
>> that "m-a ..." seems to simply unpack the previously built module,
>> which should be identical and should work?
>>
>> Am I missing a step?  Or has something else recently upgraded in
>> Debian messed up my system?  Google hasn't helped much: references to
>> the errors I'm seeing are for Fedora or Arch three years ago.
>> Attempts to revert to the "nv" driver (even using only a single
>> screen) get me bizarre colour bars on the screen, so I'm currently
>> working on a laptop.
>
> What does this give:
>
> dpkg -l |grep nvidia
>
> I suspect you updated the nvidia-kernel-source, compiled the module and
> installed it, but did not then finish the upgrade to get a new matching
> nvidia-glx.
>
> Of course since testing doesn't yet have ndidia packages in it, you will
> have to get those from unstable.

I had the nvidia-glx and nvidia-kernel-source packages v185.18.36-2 -
since testing is still at 173.14.09-5 and unstable is now at 190.53-2,
I had clearly downloaded unstable a while ago.  So I fetched those two
packages from unstable and installed them.  Then built the module
against 2.6.32-trunk.  startx brought up all three screens ... but
very slowly, and shortly thereafter locked up solid.  This is
repeatable: sometimes I can get to a console, sometimes not.  The OS
is still available, so I ssh in and reboot - as killing X doesn't seem
to do the trick.

Not quite sure where to go from here ...  Should I be grabbing the
unstable kernel, or would that only make things worse?  Compile
against one of the 2.6.30 kernels?  Should I cut back to just one
screen?

Here are the error lines (nothing but, grabbed with grep) from the Xorg.0.log:

(EE) ioctl EVIOCGNAME failed: Inappropriate ioctl for device
(EE) PreInit returned NULL for ""Microsoft Microsoft 3-Button Mouse
with IntelliEye(TM)""
(EE) ioctl EVIOCGNAME failed: Inappropriate ioctl for device
(EE) PreInit returned NULL for ""Macintosh mouse button emulation""

-- 
Giles
http://www.gilesorr.com/
gilesorr-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists