[PlanetCCRMA] Life is Not Good with 3.x-rt and Nvidia Graphics

[Fernando Lopez-Lezcano]

On 03/17/2012 05:50 PM, S C Rigler wrote:
>
>
> On Sat, Mar 17, 2012 at 6:31 PM, Brendan Jones
> <brendan.jones.it at gmail.com <mailto:brendan.jones.it at gmail.com>> wrote:
>
>     On 03/17/2012 09:16 PM, Fernando Lopez-Lezcano wrote:
>     >  On 03/17/2012 08:58 AM, S C Rigler wrote:
>     > > Yesterday I decided to go from a working F14 installation using a
>     > > 2.6.33-rt x86_64 kernel using the proprietary Nvidia driver to
>     F16 using
>     > > whatever Nvidia driver would work.  So far I have yet to find
>     something
>     > > that works.
>     > >
>     > > The stock Fedora kernel works fine running either nouveau or the
>     > > proprietary Nvidia driver.  I'm typing this while I am booted into
>     > > 3.2.9-2 running nouveau.
>     >
>     >  Good to hear that at least that is working...
>     >
>
>     FWIW, I am having no issues using the latest CCRMA kernels, nouveau or
>     the patched 290.10 prop driver (hw is 1040/rev a1).
>
>     I do get kernel oops'es in VirtualBox/KVM that I don't with the stock
>     kernel but that is another (for upstream).
>
>     One recommendation though, is you should really ensure your system clean
>     of the prop drivers (and your initrd) when falling back to the nouveau
>     drivers. Use the --uninstall option from nvidia, then reinstall
>     mesa-GL*, otherwise everything becomes a little confused.
>
> I pretty much started out trying to make nouveau work since I don't like
> having to patch the proprietary driver.
>
> FWIW, both nouveau and the proprietary driver show similar behavior on
> the stock kernel when I try to boot it with "threadirqs".
>
> I'm still trying to get some good log messages. I have plenty from
> earlier today, but they weren't consistent with each incident.

The ones related to nouveau would be most useful. I just took a look at 
my last 3.2.x rt kernel build and there are no extra Fedora patches 
(apparently) that deal with nouveau. Maybe with the proper log messages 
the kernel gurus will be able to catch this one...

-- Fernando