Temporary sticky: Nvidia driver & kernel 3.3

Knurpht has found that the new kernel 3.3 will not allow the Nvidia driver run file to install invidia in the normal way.
He has started a thread giving the workaround on this link: http://forums.opensuse.org/english/get-technical-help-here/tumbleweed/473658-kernel-3-3-nvidia-driver-fix.html

In summary (at least for 64 bit), make an executable script file with these contents:

#!/bin/sh 
cd /usr/src/linux
make cloneconfig
make prepare
cd /lib/modules/`uname -r`/source/arch/x86/include/
cp -v generated/asm/unistd*.h ./asm/
# the last line (below) must contain the correct path and file name
sh /root/NVIDIA-Linux-x86_64-295.33.run

Notice that the last line is a call to the driver run file, so make sure you give the run file the right name and path in the last line.
Locate the script and the run (driver) file in the directory /root.
After the kernel upgrade (via zypper dup) reboot to runlevel 3, and log in as root. You will be in the direcoty /root. Just run the command ./nvidia.sh and the script will execute the workaround and run the driver installer (the run file).
If you want the latest driver (NVIDIA-Linux-x86_64-295.33.run) it’s discussed and linked by jdmcdaniel3 in the thread, or this link: http://forums.opensuse.org/english/get-technical-help-here/tumbleweed/473658-kernel-3-3-nvidia-driver-fix-3.html#post2450565

A fuller howto on “the hard way” including this workaround is to be found here: http://forums.opensuse.org/content/38-swerdna-nvidia-opensuse-tumbleweed-hard-way.html

This sticky is temporary and locked. It’s just a pointer. Please make comments and inquire in Knurpht’s thread: http://forums.opensuse.org/english/get-technical-help-here/tumbleweed/473658-kernel-3-3-nvidia-driver-fix.html

This fix is no longer needed
(date 5 May 2012)

There are problems now with kernel 3.4.x
Fix here: Index of /nvidia_patches

Well, no need to. Driver 295.53 will install fine in kernel 3.4.0-25-desktop.

For other readers: like sergiom99, I also found no need to install any fix or patch, the old method works fine, see here: http://forums.opensuse.org/content/38-swerdna-nvidia-opensuse-tumbleweed-hard-way.html

Well i got some weirdness with 295.53 , compiling nvidias blob goes well , now the problem persist with kernel 3.4
As soon i enter X its a no blinking cursor upperleft corner and after ten seconds its monitor off and the hole system is hanging
So no X release tru SSH , have to reset on the dumb way (press the resetbutton wich i hate but that aside)

Xorg.log ends midway

1268.273] (II) NVIDIA: Using 768.00 MB of virtual memory for indirect memory access.
1271.281] (II) NVIDIA(0): Setting mode “1024x768+0+0”

And /var/log/warn gives this

Jun 5 22:27:02 nixan kernel: 1217.909464] nvidia: module license ‘NVIDIA’ taints kernel.
Jun 5 22:27:02 nixan kernel: 1217.909468] Disabling lock debugging due to kernel taint
Jun 5 22:27:02 nixan kernel: 1217.920316] NVRM: loading NVIDIA UNIX x86_64 Kernel Module 295.53 Fri May 11 23:28:10 PDT 2012
Jun 5 22:27:37 nixan kernel: 1253.702795] NVRM: loading NVIDIA UNIX x86_64 Kernel Module 295.53 Fri May 11 23:28:10 PDT 2012
Jun 5 22:27:39 nixan kernel: 1255.141597] AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0000 address=0x00000002120fb600 flags=0x0050]
Jun 5 22:27:56 nixan kernel: 1272.406347] NVRM: Xid (0000:01:00): 6, PE0001
Jun 5 22:27:56 nixan kernel: 1272.425564] NVRM: Xid (0000:01:00): 6, PE0001
Jun 5 22:27:59 nixan kernel: 1275.246798] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:27:59 nixan kernel: 1275.253304] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:27:59 nixan kernel: 1275.259823] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:27:59 nixan kernel: 1275.266339] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:27:59 nixan kernel: 1275.272851] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:27:59 nixan kernel: 1275.279355] NVRM: Xid (0000:01:00): 6, PE007e
Jun 5 22:28:06 nixan kernel: 1282.527435] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:08 nixan kernel: 1284.524278] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:11 nixan kernel: 1287.016173] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:13 nixan kernel: 1289.013016] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:16 nixan kernel: 1292.094905] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:18 nixan kernel: 1294.091748] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Jun 5 22:28:18 nixan kernel: 1294.133820] [sched_delayed] sched: RT throttling activated
Jun 5 22:28:21 nixan kernel: 1297.071307] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context

This happens to me with all the tumbleweeds 3.4 kernels , and even with the /kernel:/stable/standard/

So now iam on Tumbleweed with a 3.1.10 kernel from openSuSE-12.1 wich works fine

If anyone has any idea whats wrong ?

On 06/06/2012 08:06 AM, unnamedopa wrote:
>
> sergiom99;2464681 Wrote:
>> Well, no need to. Driver 295.53 will install fine in kernel
>> 3.4.0-25-desktop.
>
> Well i got some weirdness with 295.53 , compiling nvidias blob goes
> well , now the problem persist with kernel 3.4

Snip…
>
> So now iam on Tumbleweed with a 3.1.10 kernel from openSuSE-12.1 wich
> works fine
>
> If anyone has any idea whats wrong ?
>
>

Hi

What do u mean by “compiling Nvidia blob”. Do you mean running the
script NVIDIA-Linux-x86_64-295.53.run ?

Regards
swerdna

Yes by NVIDIA-Linux-x86_64-295.53.run , but starting X then its black screen and a total hangup even with the new 3.4.1 kernel from yesterday

It is fixed i got iommu in bios set to ON , wich worked flawless in 3.1.x kernels
But in 3.4.x kernels seems a iommu=pt is needed in grub bootline
Problem solved for me