Hello everyone,

I wonder if anyone else is having a similar problem: my system no longer
boots properly after kernel upgrade. The old kernel was 2.6.5-7.111.30-smp,
the new one (picked by YAST/YOU) is 2.6.5-7.145-smp.

The system has an NVIDIA card and is using the NVIDIA driver (offered via
YAST).

The symptoms are the following: Upon normal boot, the system will display
the splash screen and will continue to boot normally until the point when
it attempts to start the X. At this point the screen goes blank and the
system seems to be locked up. Booting into runlevel 3 with all frills
removed from kernel command line does allow it to boot without a problem.

If that matters, the last line in the log before the lock up is:

xxx kernel: agpgart: Putting AGP V3 device at 0000:01:00.0 into 8x mode

Is anyone experiencing something similar?

Thanks in advance,
-- Sasha

AFAIK you need to re-install the NVIDIA driver after a kernel update. Since the module is built for the old kernel, it no longer works after this update. You may run the installation from command-line yast.

Do I have to? I've never had to do it in the past (there were some 4-5 kernel upgrades since 9.1 release). My understanding was that the kernel install procedure itself took care of rebuilding the nvidia driver - at least the kernel upgrades have always been seamless in the past.

In the meantime, I've hunted down the packages for the previous kernel version and downgraded my kernel which fixed the problem.

Anyway, how do I "reinstall" the nvidia driver? It no longer shows in the list of updates in YAST.

Thanks.

Sorry, I am not an expert (I have to deal with Intel onboard graphics ), but it sounds logical to me, since the nvidia driver is an external module that is not part of the SUSE distribution and needs to be built for the current architecture. If you update the kernel, also the modules will be updated and the nvidia module may not longer exist. There is a module with less functionality that comes with the kernel so maybe you ran this one without knowing...
When you are running YOU, select 'Installable and Installed Patches' from the 'Show Patch Category' tab. Now it should show up again.

OK. I have the same problem as Sasha, but have not backed the Kernal upgrade out. I am afraid I am a bit more basic. I get the locked up blank screen. Failsafe boot from Grub does not work either (though I don't get a blank screen, it stops after one of the Reiser script lines).

How do I jump-start this puppy so that I can re-install the vid driver?

Sorry for the newbie question...

tdonoughue - first of all, make it boot into the "raw" mode 3:

1. When you get the GRUB boot menu, hit Esc and exit the graphical mode.
2. Select the boot record for Linux and hit 'e' to edit it.
3. Edit the boot command line: remove all parameters except for the actual kernel and the root argument and add '3' as the last argument to force the runlevel.
4. Boot off the modified command line (I believe, it is a 'b' key).
5. You will boot using text mode (no fancy X or framebuffer stuff) up to the text mode login.

Once there, you should be able to downgrade the kernel.

This is even easier: on the graphical grub screen, press '3' and select your normal kernel. You will now boot into command-line.

Thanks. I had already followed the more advanced instructions (but learned lots along the way).

Next stupid question: I was able to get into YAST. Is the old Kernel found somewhere? Or is there some other easy way to uninstall the new Kernel?

More intelligent question: I installed the new Kernel and went back and ran the NVidia install utility. It said something about that the Kernel was wrong and did I want to check their website for an update. No update was found. So it asked me if I wanted to rebuild the interface. I said yes; it said it could not find the Kernel source.

Many steps later, I have the Kernel source and the new Kernel installed. I re-ran the install utility and it would not install. Gave an error (I know I wrote it down here, but I can't find it).

Any idea what I am missing? Or is the old NVidia driver just incompatable with the new Kernel?

TIA

I've been asking in a number of places and the overall response was that sadly SuSE does not offer a feature allowing one to have multiple kernels installed (strange: RedHat/Fedora/Debian have been offering this feature for years). What I did was a search for a full rpm package name for the previous "good" kernel. I found it through Google on ftp.suse.com and downloaded it from there (along w/ other kernel-XXX packages installed along with the main one). Follow this plan:

1. Check your 'last' output and system logs to determine previous "good" kernel. I don't have the name handy right now.
2. Do rpm -qa | grep ^kernel to find out what kernel packages you need. Also check the architecture of those packages.
3. Search for the complete RPM file name in google.
4. Download all necessary packages from a place you trust.
5. Perform rpm -Uvh --oldpackage ... on all downloaded packages.
6. Reboot.

My thanks to one and all. I joyfully write this to you from Linux! It works. The final resolution:

1) Upon booting to a blank screen, I was forced to turn off the machine. I rebooted to level 3 via Grub (as described).
2) From the command prompt, I did a yast (newbies: just type it at the command prompt after logging in as root). I was unable to update/install the source for the new Kernel, so I installed the source for the old Kernel.
3) After rebooting (again to level 3), I used yast again to do an upgrade on the Kernel package. I also forced an upgrade of fetchnvidia.
4) I don't know why, but I had to reboot and do step 3 again (there was a failed attempt at step 5 in here, but since it failed, I figure I can skip it for you).
5) Go to /usr/share/doc/nvidia and run the installer found there (./NVIDIA-Linux-x86-1.0-5336-pkg1.run). Follow the yellow brick road through the installer. It (finally) reported success and advised to run sax2 to complete the install.
6) Exiting yast, type sax2 at the command prompt. Configure the card, center it, etc.
7) Upon reboot, it all works.

Now, I haven't gotten brave enough to enable the 3D accelleration yet. But someday soon.

Again, thanks to all for the help.

Great that it worked!

Just a few comments:

1. it is possible in SuSE to install several kernel in parallel (as you can seen in my profile ). The trick is to delete the actual kernel from the rpm database (only the database, not the kernel itself): rpm -e --justdb <your-kernel-rpm> and installation of the new kernel with rpm -ihv <new-kernel-rpm>. Then you need to create an entry in menu.lst for the new kernel (specify the version) and that's it.

2. the reboot you mentioned above was necessary to load the new kernel into memory. When you ran the YOU update, the new kernel was installed, but the old was still running. That's why you got the message about the wrong kernel, because uname -r and rpm -q kernel-default gave different versions.

I had the same problem. Thanks to this thread I used solution without downloading anything.
1) Boot to runlevel 3.
2) There is installer from previous fetchnvidia installation in /usr/share/doc/nvidia. So I run that. It reinstalled the driver.
3) Run SaX2 a configure display.
4) Reboot without params.

Everything seems fine

I had that problem as well. And this is what I did: Boot to run level 3, "sax2 –m 0=nv". There I adjusted everything more or less to please the eye (the flickering was terrible, you know ). Then, I restarted the GUI to reinstall the NVIDIA drivers. I did this via YaST.

The result: NONE!

Well, I've got my GUI back now, but without NVIDIA drivers.

Doe's anybody have an idea what I have done wrong? I mean it worked fine for Sasha and maque, didn't it?

I believe you need to reinstall the nvidia driver (see step 2 of post #12 above). That is what you missed. It needs to relink the driver to the new kernel. Without that, you have only reconfigured the driver that does not link to the kernel correctly.

(said the relative newbie, hoping that those pros would not shoot down his theory)

Well, this is finally over. The newly released .147 kernel acknowledges and fixes the nvidia problem.