On Mon, Aug 17, 2015 at 02:52:01PM -0400, Paul Cartwright wrote:
On 08/17/2015 02:45 PM, David A. De Graaf wrote:
When an nvidia card is present, the akmod-nvidia package is present, and a new kernel is installed, it is necessary to ensure that a new nvidia.ko module is created. One might expect that this would be a natural requirement of the dnf process, which would not terminate until that task had been accomplished.
Sadly, it is NOT. Until this is fixed, what I do to protect myself against a failed reboot is this:
- Near the end of the dnf update, run top and watch for abnormally
intense cpu activity AFTER dnf concludes, looking especially for instances of cc1, rpmbuild, and any other processes being run by user 'akmods'. These suggest that the building of nvidia.ko is underway. Do NOT reboot until this is done.
- Look for the existence of the new module for the NEW kernel: find /lib/modules -name nvidia.ko
Only if it exists, is it safe to reboot.
I see the one for my current kernel:
/lib/modules/4.1.4-200.fc22.x86_64/extra/nvidia/nvidia.ko
- If not, try to build it manually (as root, of course): akmods
or akmods --force
I thought the /usr/sbin/akmods --force WAS the manual way.. what is another way?
The only other way I know is the automatic way - which occurs after a dnf update or, I think, during boot if the nvidia.ko module is missing.
- If this fails and an error message says why, the new kernel must
NOT be used. Edit /boot/grub2/grub.cfg to change the default from 0 to 1. Wait for a solution.
I have other kernel options, but I don't think any have the nvidia module. can I create a file for those?
According to the akmods man page --kernels <kernel> -- build and install only for kernel <kernel> (formatted the same as 'uname -r' would produce) so something like this should work: akmods --force --kernels 4.1.3-200.fc22.x86_64