nVidia driver died, can't reinstall due to some kernel interface issue

[colinjones]

Hi!

Wondered if anybody could give me some pointers on this one, as I really do not want to have to rebuild my system. A couple of days aog I installed dkms and a couple of test v4l debs - all of which should be compatible with both 0710 (which I am using) and 0810. Unfortunately, it killed my machine. After a reboot, X will not start.

I tried removing all three components and rebooted... same issue. So I thought something must have screwed up the nVidia driver (its failing at the point it tried to load the nVidia module). So I have tried reinstalling the drivers (180.06 and the latest 185 versions), but it always fails as it tried to make the kernel interfaces between kernel and module.

Code: [Select]

Kernel module load error: insmod: error inserting './usr/src/nv/nvidia.ko':
   -1 File exists

<snip>

   [ 1193.364000] NVRM: API mismatch: the client has the version 169.12, but
   [ 1193.364000] NVRM: this kernel module has the version 180.06.  Please
   [ 1193.364000] NVRM: make sure that this kernel module and all NVIDIA driver
   [ 1193.364000] NVRM: components have the same version.
   [ 1200.096000] Linux agpgart interface v0.102 (c) Dave Jones
   [ 1200.400000] PCI: Setting latency timer of device 0000:00:12.0 to 64
   [ 1200.400000] PCI: Setting latency timer of device 0000:02:00.0 to 64
   [ 1200.404000] NVRM: loading NVIDIA UNIX x86 Kernel Module  180.06  Sat Nov
   8 12:13:58 PST 2008
In the install log and similar messages in dmesg. I understand that the nVidia drivers use dkms in some way to build the kernel interfaces, so I can only assume that installing that in someway is now preventing the driver install from doing this properly.

Has any body any idea how I can fix this up??

[tkmedia]

cant you just apt-get a different driver?

if need be look into
dkms -remove driver name

hth



Tim

[totallymaxed]

Hi!

Wondered if anybody could give me some pointers on this one, as I really do not want to have to rebuild my system. A couple of days aog I installed dkms and a couple of test v4l debs - all of which should be compatible with both 0710 (which I am using) and 0810. Unfortunately, it killed my machine. After a reboot, X will not start.

I tried removing all three components and rebooted... same issue. So I thought something must have screwed up the nVidia driver (its failing at the point it tried to load the nVidia module). So I have tried reinstalling the drivers (180.06 and the latest 185 versions), but it always fails as it tried to make the kernel interfaces between kernel and module.

Code: [Select]

Kernel module load error: insmod: error inserting './usr/src/nv/nvidia.ko':
   -1 File exists

<snip>

   [ 1193.364000] NVRM: API mismatch: the client has the version 169.12, but
   [ 1193.364000] NVRM: this kernel module has the version 180.06.  Please
   [ 1193.364000] NVRM: make sure that this kernel module and all NVIDIA driver
   [ 1193.364000] NVRM: components have the same version.
   [ 1200.096000] Linux agpgart interface v0.102 (c) Dave Jones
   [ 1200.400000] PCI: Setting latency timer of device 0000:00:12.0 to 64
   [ 1200.400000] PCI: Setting latency timer of device 0000:02:00.0 to 64
   [ 1200.404000] NVRM: loading NVIDIA UNIX x86 Kernel Module  180.06  Sat Nov
   8 12:13:58 PST 2008
In the install log and similar messages in dmesg. I understand that the nVidia drivers use dkms in some way to build the kernel interfaces, so I can only assume that installing that in someway is now preventing the driver install from doing this properly.

Has any body any idea how I can fix this up??

Hi Colin,

If you can PM me details so that we can ssh into your Core we will take a look at the problem on Sat/Sun for you.

All the best

Andrew

[colinjones]

Thanks Andrew, but I ended up completely screwing it up by repeatedly trying to install/uninstall dkms, the drivers, etc and then it said it couldn't even start the driver install because there were no kernel sources, and I couldn't install them using sudo apt-get install kernel-sources.... I gave up, and blew it all away and I'm reinstalling. I don't suppose anybody has ever successfully backed up an MD (from /usr/pluto/diskless...) and restored it have they?

I want to avoid going through the pain of getting the 8168 driver working again for this MD, so I tarred the entire folder. I figured I would turn on the MD, let it get discovered and build the image/DB details. Then blow away that directory and restore from the tar... do you think that would work?