[SOLVED] kernel 5.1.14 Nvidia modules mismatch

[rychok]

I have one installation started as MGA 7 beta 3 now with all updates. Last updates I had to make manually, probably because they are in core not update repo, anyway this installed kernel 5.1.14 and all nvidia 430.26 needed for mageia-prime. Booting to my previous kernel 5.1.6 is still fine, but X is not starting with 5.1.14, dmesg shows in the end lot of lines saying that client nvidia module is 430.26 but kernel version is 430.14 - mismatch - dkms status shows for all kernels nvidia 430.26 installed, I have checked installed nvidia packages, all of them are 430.26.
Is this kernel comes with build in nvidia 430.14?

[benmc]

dkms would imply non-free.
do you have non-free release and non-free-update enabled?

if, so run urpmi --auto-update.

regards

[rychok]

Non-free are enabled, and dkms is installed, nvidia modules for all kernels have same date - compiled in time of upgrade, and booting to 5.1.6 is fine, the question is where trace to nvidia 430.14 is coming from in kernel 5.1.14? unless it was somehow compiled where previous nvidia was not removed yet, so I will uninstall and install back kernel 5.1.14 knowing that there is no old nvidia to let dkms compile modules again.

[rychok]

I do not know what is the order for updates to apply, especially when some scripts are running, but it must have something gone wrong during the update I have done.
When I uninstalled kernel 5.1.14 and rebooted it was not removed from grub and it was the default choice, so next boot was unsuccessful changing to 5.1.6 in the list was fine.
I fixed the GRUB entry by going thru boot option in MCC, the I have install kernel 5.1.14 and realized I have not seen last time message informing that reboot is needed to run new kernel.
After that system booted to kernel 5.1.14 without any problems.
It looks like there is some specific updates configuration that may get you into not boot able system by compiling older driver for kernel when updating to new one at the same time.