[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: drm and radeon module load problems

Sorry about the delay in my reply :( My ISP did something with the
address space and the ip-address, I had for a long time, changed to
something else. I really need to start working on a script which would
take care of the settings in the case where my ip-address changes...

On Wed, Oct 05, 2005 at 08:04:58PM +0200, Lubos Vrbka wrote:
> >Hmm, that is weird!
> yes, it is!
> >Yes I think it is ok. Unfortunately I don't have any further info about
> >IOMMU, but the help for IOMMU says that it includes a driver for the AMD
> >Opteron/Athlon64 IOMMU, so I guess that you can have agp driver for AMD
> >on-CPU GART with or without IOMMU.
> ok, it's clear now.
> >>i have via k8t890. i just realized that the kernel recognizes many 
> >>features that the last one missed... regarding the agp, it worked fine 
> >>with 2.6.11...
> >At least 2.6.14 has something for the K8T890. You can just `grep -i
> >k8t890 /usr/src/linux/drivers/char/agp/amd64-agp.c` to see if it is in
> >the 2.6.12.
> that file contains following "struct", so i guess it's ok
>         /* VIA K8T890 */
>         {
>         .class          = (PCI_CLASS_BRIDGE_HOST << 8),
>         .class_mask     = ~0,
>         .vendor         = PCI_VENDOR_ID_VIA,
>         .device         = PCI_DEVICE_ID_VIA_3238_0,
>         .subvendor      = PCI_ANY_ID,
>         .subdevice      = PCI_ANY_ID,
>         },
> >If you have the agp compiled in the kernel, you pass a kernel parameter
> >during a boot.
> in case i don't have it in kernel for some reason (config surely had it
> enabled), it should complain at that point, right?
Yes. It is really weird, that you had it in the .config but there was
absolutely nothing about the agp in the dmesg!?

From your trials it seems that the module gets laoded and your chipset
is supported by the kernel, so everything should be ok from the AGP
point of view. Also I think it means that agp_try_unsupported=1 would
not help in your case.

> >If you use GRUB, edit the /boot/grub/menu.lst.
> i use lilo, but that's no problem, modification is easy...
> >If you compile it as a module, just make a file into the
> >/etc/modprobe.d/ directory and add a line
> >options agpgart-amd64 agp_try_unsupported=1
> >e.g. `sudo echo "options agpgart-amd64 agp_try_unsupported=1"
> i'm not sitting in front of my box right now, but what if i change the
> config and try to build the agpgart as a module and force to load it?
> can it conflict with the code in the kernel, or would it just refuse to
> load?

If you have the IOMMU or the AGP support compiled in the kernel, you
should get an error if you try to force modprobe to load the agp module.

My grep -i agp /usr/src/linux/.config has
# CONFIG_AGP_INTEL is not set
and grep -i iommu /usr/src/linux/.config 
# CONFIG_GART_IOMMU is not set

I.e. no IOMMU and AGP compiled into the kernel.
dmesg |grep -i agp gives:
Linux agpgart interface v0.101 (c) Dave Jones
agpgart: Detected AGP bridge 0
agpgart: AGP aperture is 256M @ 0xd0000000

256M is the size of the aperture (from the BIOS settings)

When I start X, the drm and, in my case, mga get loaded automatically.

> might this be some kind of kernel bug?
> thanks for help. regards,
> -- 
> Lubos
> _@_"

If you haven't already checked the trouble shooting guide, have a look

:r ~/.signature

Attachment: signature.asc
Description: Digital signature

Reply to: