I’m using opensuse 11.3 with kde 4.5 (recently installed from opensuse repository)
I updated the nvidia driver after a prompt from yast and now:
I boot up
Go into the Grub screen which lets me choose between opensuse and Windows.
Click on Opensuse
All the usual text messages show up on the screen, but they finish with “Nvidia taints kernel”
The screen then goes blank with a blinking cursor at the top right hand corner - no prompt for username or password. The only thing that works is control alt delete, which re-starts the system.
I’m presuming this has something to do with the nvidia upgrade - but what do I do about it? Safe mode does not work either.
Do you have more than one entry to load like failsafe or more than one kernel version to try? If you can get any version to load, you could run YaST in the terminal mode and perhaps get rid of the repository or something. Here is a way to run YaST from Terminal (if you can get a prompt). caf4926 did this when we were having kernel issues. Now, you can’t get your openSUSE to run at all which is a kernel issue for sure.
If NO kernel will load, you will need to find your installation disk. Once again, caf4926 shows here how to fix a busted kernel by doing an upgrade, but only replacing your kernel. Again, something to think about if openSUSE is all blown up.
You could try one other fix first. You’ll probably have to do this anyway, even if the kernel has to be replaced.
Boot to windows. Go to the nvidia site and download the 250.53 driver (the latest before the 260.19 that is causing trouble)
Follow the method to get to the command line as shown in the first link above.
Copy the nvidia driver to your linux partition (to be honest, I’m not sure you have to do this. It may run from the windows partition, but I didn’t try that. Instead, I logged out and when I rebooted, I got the graphical login screen. The “session” menu in the lower left let me pick a different desktop -Icewm- from there I ran the file manager -Midnight Commander- to move the driver to where I wanted it in the suse partition. Now, I always have a known-good driver to use.)
Change directories to wherever you have the nvidia driver.
Don’t use the quotes - but type “.sh nvidia” and hit tab. That will automatically complete the long file name of the driver.
Hit enter, and the old driver will install.
If that doesn’t get your desktop back and you have to reinstall the kernel, at least a working driver will be there when you do.
Thanks for everybody’s comments
I should clearly have waited for the new kernel before installing new Nvidia drivers - lesson learnt.
Anyhow, this is where I have got to:
I removed my Nvidia card and am now just using my Shuttle’s built-in graphics - slow though.
By typing “3” into the options line I can now boot to the command prompt- in spite of getting the red message “Starting up sensors failed”.
I have run Yast and removed the Nvidia drivers and disabled the repository.
I ran Zypper on the off chance that that might help, as it has done in the past.
However, if I now try to boot from the Grub menu without putting the “3” in, the booting process grinds to a halt at the line :
[17.897780] ADDRCONF(NETDEV.CHANGE): eth0:link becomes ready.
I’ve done this dance and it makes me believe there is at least a part of the problem resting with kde. I logged out and was brought to the graphical login screen. From there, I selected the lxde desktop from the session menu. It ran fine, not as pretty as kde, but fine. That allowed me to download the old driver so I could roll it back. I tried the nomodeset solution but did not go so far as to edit that initrd file. Did you?
Not sure about kde - certainly not alone. I tried to launch an XFCE session from the command line and got the message “cannot open display”. I seem to remember that in OpenSuse 11.2 there was a display manager that one got fire up from the command line - does this still exist?
I don’t quite understand what you’re trying to do.
Use Yast - System - Syconfig Editor, search for “KMS”, one item is found. Set it’s value to “yes”. This makes using the nomodeset option no longer needed. Correct me if I’m wrong. Run ‘su -c mkinitrd’ afterwards to apply the changes to the system’s ‘initrd’.
But why not install the driver ‘the hard way’? Install the ‘Linux Kernel Development Pattern’, download the driver from ftp://download.nvidia.com/XFree86/ Doing so at least gives you a choice of drivers in case of trouble like this. I have no issues with the latest driver AFAIK now, have to say though I hardly ever use the drivers from the repo.
Just a recap:
I have used the nvidia repo for some time. When the kernel update and the driver update came, I got the black screen with cursor that is widely reported. One fix I tried was to put nomodeset on the boot line. I did not go to the second step and edit initrd file. This fix failed. But, in my defense, people reported varying results of doing both steps. There was universal agreement that rolling the driver back cured the problem.
I rolled it back by downloading the older driver and installing it the hard way. I’m now going to take your tip and try 260.19 again.
No luck for me. I edited the KMS value. Actually, it already was “yes” but I did mkinitrd anyway. Downloaded the 260.19 driver from nvidia (I could not get your ftp link to work). Dropped to init 3 and ran the nvidia file and rebooted. The problem is identical. I dropped back to init 3 and installed the old driver the hard way.
I may have missed one thing. I don’t know what the Linux Kernel Development Pattern is or how to find it in yast–> software. But everything else seemed to work, except for the actual driver!