@pallaswept well they are Third party packages, not openSUSE packages… as in not from the Tumbleweed, oss, non-oss and update repositories
PS, I’ve changed the Thread Title to clarify…
@pallaswept well they are Third party packages, not openSUSE packages… as in not from the Tumbleweed, oss, non-oss and update repositories
PS, I’ve changed the Thread Title to clarify…
This wouldn’t be happening if the default Nvidia repo wasn’t pushing the jurassic 550 drivers. Users are forced to try the other Nvidia repo made for CUDA users just to have an up-to-date driver package.
Usually, this is not a problem, but being parked on the 550 is especially egregious, as it lacks the rather vital explicit sync feature implemented on 555, without which just about impedes the use of Wayland on the distro for Nvidia users.
@alghost The release schedule is based on Production Branch releases, if folks want others as explained on the Mailing List thread, step up, build and maintain and they will be released… Those that do, decide…
Hi,
I had the same problem.
After some work I found out the latest driver of Nvidia is not compatible yet with th Kernel 12.6, so I had to go back to the driver version 550.153 (Nvidia homepage indicate it). Now its working.
Blockquote
I did manage to compile and install de modules of the latest version of the Nvidia driver, but when starting X an error ocurred, no display available, because it did not handle correctly the DRM.
Hope it helps. (OpenSUSE Tumbleed, Nvidia Quadro M4000)
In the grub menu, I select advanced option and boot to 6.11 kernel. As stated in this thread, that boots to a gui. A related question: how do I assure that booting to 6.11 remains an option through new zypper dup’s?
Lock the kernel version you want to
keep.
@Prexy Hi, first let me preface that there is some expectation that Tumbleweed users know/can handle issues like this to move forward with their setup…
So as indicated by @hui locking the kernel, that IMHO is only part and can be avoided by;
/etc/zypp/zypp.conf
multiversion to the version you wish to retainIf you boot to working kernel and run yast2 boot loader and save it, it will make the kernel you booted from as your default.
Be careful, that alone will not save the kernel packages from being deleted by ‘zypper dup’ (see points 1. and 2. from @malcolmlewis above) and so will not
Literally can’t even decide how it went down when he’s making it up. Literally can’t caunt to 3, thinks it’s 20. Refers to reading the contents of a forum thread as “wasting extra efforts”.
Watch out folks we got a real genius on our hands here. Don’t worry, he’ll be sure to visit every thread that has nothing to do with him and let you know.
Just saying that it will become the default. As a user of the machine take care of it by locking the working kernel until such time that the new one is ok to install.
I never ended up "zypper dup"ing again out of fear of bricking the PC, so when the kernel 6.12 killed the graphics session thanks to nvidia’s goofy drivers, I just used snapper to rollback to before the “zypper dup” to get my PC back. As a consequence, functionally, I never installed kernel 6.12, so I think the solution proposed doesn’t apply one to one.
In the end, I added the locks “kernel-default*” and “kernel-devel*” to zypper, and tried "zypper dup"ing again.
So far so good, did “zypper dup”, updated my programs, and kept my graphics session.
A question to the specialists: How wrong was this solution of mine?
If I understand correctly, I will keep the locks until nvidia releases a new driver version or something, yes?
@alghost It’s not wrong to lock when something isn’t working.
Nividia works here. The affected on this thread are the ones using cuda. I can’t comment on that cuda thing
I am using the .run driver version 565.77 from nividia
I’m not using cuda per se, only the drivers from the cuda repo, which are more up to date than the base nvidia repo.
@alghost well neither are if you use the run file So if you have an issue with the Nvidia repo, at least you could create an openSUSE bug report. For the cuda repo, you would need to hit the Nvidia CUDA forums for support…
@malcolmlewis How about something like this in /etc/zypp/zypp.conf
line 581 here < multiversion.kernels = latest,latest-1,latest-2,latest-3,latest-4,latest-5,latest-6,latest-7,latest-8,latest-9,running,6.6.11-1,6.9.9-1,5.0.13-1,5.1.16-1
This may not be good practice then? I was hesitant to ask because the post is solved
now.
@panorain should only need the versions, but if compiling a kernel module, then you need to add a lock to the respective kernel-default-devel… Again, there should be no need, that’s just procrastinating around the issue…
Thanks to those of you that helped me with the procedure to lock the older kernel. As expected, it worked to keep the gui. Another zypper dup wanted to update to 6.12 kernel. I followed directions and kept 6.11 as default. One unexpected thing: when I got to the grub menu, the pre-selected option was not the top line option of Tumbleweed, but the second option for Advanced, where the 6.11 kernel remained listed.
Thanks again.
I use CUDA with latest driver on tumbleweed.
kernel 6.11.8-1-default
Driver Version: 565.77 CUDA Version: 12.7
I know about “compatibility issues” with kernel and CUDA-Nvidia driver from Nvidia repo.
(basically you will be stuck to Leap stable kernel if you want driver from Nvidia-cuda-repo, and there no way to use any current ML stuff that require everything latest with 2-3-4 years old Leap kernel)
I do known “junk way solution”:
P.S. yes on every kernel update - you need to reinstall or update nvidia driver from run file, and obviously you must not have any other nvidia-rpm drivers installed.
Same issue here. I have been running with nvidia-open-driver-G06-signed-cuda-kmp-default 565.57.01_k6.12.6_1-2.2 from the oss repo, and the nvidia-video-G06
, nvidia-gl-G06
and nvidia-compute-G06
from the cuda repo, however this is not working on the new kernel