nvidia-driver-xxx requires reboot?

[Hund]

On May 2, 2021 7:30:54 PM GMT+02:00, Florian Oppermann <florian.oppermann at itp.uni-hannover.de> wrote:
>Hi all,
>
>Our workstations run basically 24/7 and can’t be rebooted arbitrarily. 
>On recent workstations with Nvidia GPU the following error messages 
>appear after upgrading nvidia-driver-460. The system is then unusable at 
>least via GUI / DE until it’s being rebooted. Apparently the updated 
>userspace driver is put into use before the updated kernel module could 
>be loaded.
>
>> kernel: [574824.181928] NVRM: API mismatch: the client has the version 460.73.01, but
>> kernel: [574824.181928] NVRM: this kernel module has the version 460.56.  Please
>> kernel: [574824.181928] NVRM: make sure that this kernel module and all 
>NVIDIA driver
>> kernel: [574824.181928] NVRM: components have the same version.
>
>As far as I can see, the package does not even touch 
>/var/run/reboot-required as kernel updates do. Also, I couldn’t find 
>anything about updates crashing running desktop environments (here: xdm 
>+ Gnome) in the documentation.
>
>Is this supposed to happen? In this case our idea was to pin the nvidia 
>package and update it only when a reboot is due anyway. Is there a more 
>elegant solution?
>
>If this is not the expected behaviour: Who is to blame here (Nvidia? 
>package maintainer? xdm?) and where would I post a bug report?
>
>Many thanks and best regards,
>Florian
>

Have you tried unloading the old kernel module and then lodging the new one after an update? You need to kill X, but you have to restart the computer.

--
Hund