dkms vs later kernels

gene heskett gheskett at wdtv.com
Thu Oct 21 13:03:51 CDT 2010


On Thursday, October 21, 2010 01:53:34 pm Florian Hubold did opine:

>   Am 21.10.2010 19:19, schrieb gene heskett:
> > On Thursday, October 21, 2010 01:00:29 pm Florian Hubold did opine:
> >>    Am 20.10.2010 19:58, schrieb gene heskett:
> >>> So, whats next here?
> >> 
> >> Don't want to be rude, but maybe you should try to solve this
> >> issue with the help of your preferred pclos user community,
> >> as this is not even remotely a dkms issue.
> >> 
> >> Regards
> >> 
> >> 
> >> Florian Hubold
> > 
> > Since it seems to be pretty vaporous (community support) and as that
> > avenue has been explored previously, I thought I would go to the
> > source and ask.
> > 
> > Why do you say this is not even remotely a dkms issue?  That answer
> > could be very enlightening to me, the missing clue so to speak.
> > 
> > What I am seeing here if I strace the dkms_autoinstaller, is 10 to 30
> > second waits at the waitpid( lines, several of them like the child is
> > never being launched.  But the strace output doesn't contain the
> > name, pathto or sub-script being launched in a form recognizable to
> > me prior to that waitpid( call.  Hence the puzzlement.  If I am
> > missing a support function in the current kernel, it would be very
> > enlightening if it was reported as I am quite capable of building
> > these kernels, 2.6.36 3 times so far this morning because a make
> > oldconfig from the 2.6.35.7 .config threw away a goodly bit of the
> > stuff for my hardware and enabled other stuff it had no business
> > doing. Like switching my audio from emu10k1 to intel_hd on a video
> > card that no longer exists in the system and which hasn't been built
> > here for months.  Not that this is the first time a make oldconfig
> > has screwed me, but I should be used to it by now after 12+ years of
> > linux.
> > 
> > Thank you Florian Hubold.
> 
> AFAIK and comparing your make.log to mine (mandriva 2010.1,
> kernel-2.6.33.7, nvidia-current-260.19.12) the nvidia module was built
> succesfully on your box. Attaching mine sou you can compare a bit here
> and there.
> So this is another issue, and has nothing to do with dkms, as dkms is
> *only* for dynamically compiling kernel modules when a new kernel is
> run the first time.
> 
> (please don't beat on me too hardly for this simplification)
> 
> Seems your kernel "problems" are the source of this, you should try to
> get the latest
> of the official pclos packages and try again with that one.

The last time I booted to that old kernel, it worked nicely, so I am 
attempting to deduce why it stopped working after 2.6.35-rc1 and hasn't 
worked since.

I think now I would have to nuke/rename xorg.conf in order to boot to that 
kernel now because the present one calls for the vesa driver, which works 
stably, but no special effects.

What I need most is a list of "must haves" that it needs to work with any 
kernel.

Forinstance, for a 2.6.36 test boot, fglrx and nvidia-current both fail in 
the build with an error 10, but the vbox stuff that I can't use with the 
vesa driver builds and installs just fine.  I have at times also seen error 
5's, but the limited man pages do not discuss the error numbers and what 
they mean.

Perhaps you have a better set of man pages that pclos leaves out.  URL?

Thanks very much Florian Hubold.

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
The two things that can get you into trouble quicker than anything else
are fast women and slow horses.



More information about the DKMS-devel mailing list