Enabling NVAGP and disabling GART_IOMMU?

[Axel Thimm]

On Thu, Jun 15, 2006 at 10:01:14AM -0400, Ric Moore wrote:
> On Thu, 2006-06-15 at 00:53 +0100, Ben Lancaster wrote:
> > Hey,
> > 
> > I cannot for the life of me get nvidia-graphics8762(-kmdl) from  
> > atrpms working with the x86_64 2.6.16-1.2115_FC4 kernel from fedora- 
> > updates, or any other relevant x86_64 kernel combo for that matter.
> > 
> > Here's what happens with nvidia-graphics8762 and 2.6.16-1.2115_FC4:
> > 
> > # tail /var/log/messages
> > Jun 14 23:42:26 benlancaster kernel: NVRM: not using NVAGP, kernel  
> > was compiled with GART_IOMMU support!!

Since the kernel from Fedora Core is indeed with AGP built in
non-modular you can't turn it off. If you really need to, try the
kernel-suspend2 kernels at ATrpms.

Other than that you would really have to build your own kernel (not
recommended).

> > If I "make menuconfig" in /usr/src/kernels/2.6.16-1.2115_FC4,
> > disable GART_IOMMU in "Processor type and features" and then
> > enable /dev/agpgart in Device Drivers -> Character Devices ->
> > "/dev/agpgart (AGP Support)" and "AMD Opteron/Athlon64 on-CPU GART
> > support"
> > 
> > ... then save and exit, "make" then fails:
> > 
> > # make
> >    CHK     include/linux/version.h
> >    SPLIT   include/linux/autoconf.h -> include/config/*
> > make[1]: *** No rule to make target `init/main.o', needed by `init/ 
> > built-in.o'.  Stop.
> > make: *** [init] Error 2
> > 
> > I have to confess, I'm not entirely sure I know what I'm doing, so if  
> > anyone can aid me in getting this all working, I'd be eternally  
> > grateful!
> 
> If you don't know what you are doing, then don't burn a kernel. I've
> been using Linux for some years, 
> and burning a kernel can be highly problematic. I just successfully
> followed the instructions at:
> http://rpm.livna.org/configuration.html
> http://www.stanton-finley.net/fedora_core_5_installation_notes.html#nVidia
> 
>   ...and it works. Read both sites to get all the information for your
> setup. The livna install requires that the kernel version match the
> nVidia driver version. So, the process will install a kernel for you if
> you don't have the correct kernel installed. All in all, it's pretty
> straight forward, IF you follow the instructions. and allow it to work.
> I'm happy as a clam now that I have it up and running. Ric

Same is true for the ATrpms versions, but if you edit xorg.conf to try
to use nvidia's AGP you will see the error/warning above. Unless you
use the kernel-suspend2 kernels.

But I think Ben isn't really after replacing the kernel's AGP with
nvidia's, so he can still use the stock kernel and ATrpms kmdls (or
livna's equivalents). He just shouldn't turn NvAGP to "nvidia's
internal support" in xorg.conf (better drop that configure line
altogether).

Ben, if you really need NvAGP, then kernel-suspend2 & ATrpms' kmdl are
the only sane way to go.
-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20060615/1e6910f0/attachment-0001.sig>