HELP: 2.6 + radeon problems [was: Kernel 2.6 build error]

[Stefan Bruda]

Hi.

The setup looks similar to mine.

At 21:01 +0100 on 2004-2-10 Albrecht Dre=DF wrote:
 >
 > Section "Monitor"
 >         Identifier      "Monitor0"
 >         VendorName      "Formac"
 >         ModelName       "Gallery 1740"
 >         HorizSync       55-75
 >         VertRefresh     60
 >         Modeline "1280x1024"  108.0  1280 1328 1440 1688  1024 1025 =
1028 1066

You may want to try to leave out the modeline altogether.  XFree86
4.3.0 and above is supposed to interogate the monitor and obtain a
modeline of its own.

 > EndSection
 > Section "Device"
 >         Identifier "Card0"
 >         Driver "radeon"
 >         BusID "PCI:0:16:0"
 > EndSection

I have here more options such as AGPMode and AGPFastWrite which are
probably irrelevant to the problem at hand.  There is only one option
that could be of interest, namely

    Option      "UseFBDev"

The effect is that X uses the (hardware) framebuffer device, which is
a nice thing to have (it is also needed in my case even with the 2.4
kernel).

The BusID should not be needed, you may want to leave it out (although
it does not hurt if the correct PCI ID is provided in there).

By the way, the "no valid screen found" error is kind of misleading,
since an invalid screen may result because of a faulty Device section
that is used by the said screen.

 > This led me to the assumption that there might be a problem in
 > X11/kernel interaction.

Yes, this seems the most likely cause.  As I said, XFree86 4.3.0
predates the stable 2.6 kernel, which increases the likelyhood of such
a problem.  This could be fixable by kernel arguments (see below).

 > What would be the kernel argument I had to use here=3F

On my G5 I am using `video=3Dradeon:dfp' which I would assume to become=

`video=3Dradeon:crt' for a CRT (somebody please correct me if I am
wrong).

Gee, this may provide a solution, my G5 (running a 2.6) definitely
needs it before it even starts to think about a usable X.

At 21:35 +0100 on 2004-2-10 Albrecht Dre=DF wrote:
 >
 > O.k., I'm one step further: replaced the driver by "fbdev" and
 > removed the BusID statement, and X *does* work - ssslllooowww
 > though.

fbdev is the software framebuffer, so it is supposed to be slower.
Basically you sidestep the video card's processor and do everithing
with the main processor (aka unaccelerated video).

In any case, other than the kernel arguments I cannot think of any
easy solution.  A new XFree86 may work, but this is a timewise costly
process.  You may nonetheless have to do it sooner or later.  By the
way, the compilation from sources is mainly uneventful, though it does
take time.

Stefan

--=20
If it was so, it might be; and if it were so, it would be; but as
it isn't, it ain't. That's logic.
    --Lewis Carroll, Through the Looking-Glass