Comment 32 for bug 1431753

Alberto is rolling this back into older releases over time. To be fair it is an annoying issue but not one that is so fatal one would not be able to recover a system. In the majority of cases this ends up being just a false system issue notification. Because on kernel update the parallel builds are actually started 3 times. By now I just made it a habit of doing the following whenever kernel updates come in:

- run "dkms status" to see whether both nvidia modules appear for the latest kernel (bbswitch is always there)
- if one(or both) is missing, do "sudo dkms install -m <module> -v <module version> -k <kernel version>
   (the name and versions taken from the dkms status report)
- "sudo rm /var/crash/*nvidia*" to ignore the build failures before

When the fixed version of the driver hits this can be noticed by having only one module reported by dkms for nvidia. The *-uvm one will be gone and the module it produced will be part of the remaining dkms module which then produces two kernel modules.