akmod isn't reliable

[Richard Shaw]

On Thu, Jan 2, 2014 at 10:31 PM, Marko Vojinovic <vvmarko at gmail.com> wrote:

> On Thu, 2 Jan 2014 21:48:32 +0000
> "Powell, Michael" <Michael_Powell at mentor.com> wrote:
>
> > I guess this is more of a general question, but sometimes after
> > updating the kernel or nvidia drivers an akmod isn't regenerated and
> > my system will begin to boot, fedora logo will show, but eventually
> > it will dump to the systemd log of services being started and just
> > sit there. I have all the required dependencies before the update
> > because I can simply reboot to runlevel 1 or if I have an older
> > kernel boot it and then manually `akmods --kernels`.
> >
> > So the question is... why isn't regeneration of the akmod reliable?
>
> I think it is reliable, you just need to wait it out. The rebuilding of
> akmod is being done for a given kernel while that kernel is running, so
> when you update the kernel, the akmod doesn't get built until you boot
> into it. And when you boot into it, systemd will at some point try to
> activate the akmod, find out that it doesn't exist, fail, and initiate
> a rebuild.


Well, that's partially true. akmods also tries to build the module after
kernel installation using the kernel posttrans trigger or something like
that, there's a special directory where you can put script which will be
run after a kernel is installed. DKMS uses the same method. There is where
it SHOULD happen. The problem is that it's totally non-interactive and
there's no notification to use user if it fails...

It also attempts to build kernel modules on startup AND shutdown. So there
is more or less 3 attempts. The problem is if it fails one of them it will
usually fail all of them.

Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/users/attachments/20140103/d1a53e3b/attachment.html>