Re: (non) "bug" in 810 (kernel 2.6.9)

Re: (non) "bug" in 810 (kernel 2.6.9)

[John McGowan]

Re: (non) "bug" in 810 (kernel 2.6.9)

I had reported a problem with getting a working screen in X with kernel
2.6.9. It is not a bug, but a change.

It may have annoyed you a bit, but it annoyed me far more, I assure you
(though I was sure one could use Knoppix to patch, recompile and reinstall a
kernel one had played around with on a hard disk and which would only boot
into a "panic"- now I know one can do it).

In Fedora Core2, the xorg-rpm has a dependency of "kernel-drm" and while I
like to configure and compile a kernel with my options, I left in the
DRI/drm item for the Intel 810 chip. I like 24 bit color and since kernel
2.2.? there was no problem. One might not have acceleration, but one could
use a 24 bit screen with drm compiled in. It was true in 2.4. In 2.6.[1-8].
It is not true in kernel 2.6.9. While it used to be that if the driver did
not support it, acceleration simply was not used. Now, even commenting out
any mention of dri in xorg.conf is not sufficient in 2.6.9 (at least for my
system) to get a workable screen. Adding the option,
 Option "NoAccel" "true",
however, does get me a working screen at 1024x768, 24 bit color.

However, I don't seem to get a working screen with acceleration enabled for
a 1024x768 16 bit screen.

On the other hand, I do get a working 1024x768 24 bit screen using the
Intel 8x0 framebuffer driver (with generalized timing to get better than a
60Hz refresh rate).

I had tried quite a bit, from modifying the 2.6.8->2.6.9 kernel patch (which
is how I wound up with a system with a single kernel which would only
"panic") through rebuilding the Xorg RPM.

I feel like rather a fool to find out that while there may be a problem (as
I don't seem to have a working 16 bit screen with DRI) what I use (24 bit)
was just a matter of the proper configuration (though a change from what
worked for all prior 2.x kernels I had used).

Re: (non) "bug" in 810 (kernel 2.6.9)

[Dave Airlie]

> I had reported a problem with getting a working screen in X with kernel
> 2.6.9. It is not a bug, but a change.
> 

I can't reproduce this on my system at all, I've got an i815 (slightly
newer chipset than i810 - maybe something in that...) but I've just
compiled a kernel with the i810 drm built in and started X in 24-bit
color on 2.6.9 without incident...

I'm running Xorg 6.8.0-rc3.... its wierd I've no idea what could be
causing it ...

Dave.

Re: (non) "bug" in 810 (kernel 2.6.9)

[John McGowan]

On Mon, 8 Nov 2004, Dave Airlie wrote:

> > I had reported a problem with getting a working screen in X with kernel
> > 2.6.9. It is not a bug, but a change.
> >
>
> I can't reproduce this on my system at all, I've got an i815 (slightly
> newer chipset than i810 - maybe something in that...) but I've just
> compiled a kernel with the i810 drm built in and started X in 24-bit
> color on 2.6.9 without incident...

I have been finding more problems. Even with DRM not configured in the
kernel, I have explicitly to kill (NoAccel) all 2D acceleration to
get things working.

I put DRM back in.

I have managed to avoid having to kill ALL 2D acceleration by selectively
disabling parts of it. My xorg.conf now has:

        Option "XaaNoScanlineCPUToScreenColorExpandFill" "true"
        Option "XaaNoSolidFillRect" "true"
        Option "XaaNoScreenToScreenCopy" "true"
        Option "XaaNoMono8x8PatternFillRect" "true"

(the first blocked access to the xterm windows ice put up,
 the second caused the junk on the screen,
 the third was to get text inside Mozilla to scroll and
 the last was to get the MonoPattern in the xterm scrollbar)
(perhaps there will be more things I have to disable)
(I found that disabling all individual options worked and so
then enabled then one at a time until a problem arose to find
out what had to be disabled to get things working)

I can then get DRI working in 16 bit mode (actually I can get it
working with just the first two disabling options but put in the
others for general use).

This is for xorg-x11-6.8.1-12 (rpm, fedora core2 development branch,
rebuilt under kernel 2.6.9).

> I'm running Xorg 6.8.0-rc3.... its wierd I've no idea what could be
> causing it ...

Nor I.

Maybe that's why they brought out the 815 chipset :-)


Regards from:

    John McGowan  |  jmcgowan@inch.com                [Internet Channel]
                  |  jmcgowan@coin.org                [COIN]
    --------------+-----------------------------------------------------