On Wed, Jan 11, 2006 at 11:28:27AM +0100, maximilian attems wrote:

quoted

quoted

quoted

Sigh, can't get a break with alpha kernel support around here.  After
upgrading to 2.6.14 (from 2.4.27), the Matrox framebuffer no longer works
correctly on my alpha with a Matrox Millenium II.  The matroxfb_base module
loads without error, but gives me corrupt video output only.

quoted

quoted

Try turning off acceleration.

quoted

Doesn't make a difference.

quoted

What did make a difference was, after googling, loading fbcon manually
before loading matroxfb_base.  Given that I'm loading matroxfb_base by hand
(/etc/modules), it's not getting loaded via udev or anything like that, it
seems to me that it's my responsibility to load fbcon by hand as well, but
it's still something of an unexpected change from 2.4.  It might be nice to
have these modules all autoloaded by something, but it's not strictly
necessary, and some users may not want the framebuffer activated
automatically?

quoted

The other issue (and the first thing I was trying to get work, which led me
to believe the fb was completely broken) is that, even though console works
on the framebuffer now, X does not.  This breakage corresponds to the kernel
upgrade, not to any changes in X, so still looks like a kernel bug to me.

quoted

If I turn off "UseFBDev" in my xorg.conf, X displays correctly.  I haven't
poked yet to see what this does performance-wise.

2.6.15 has matroxfb patches, did it make a difference on your box?

Don't know yet, thanks to bug #347556. :)

On Wed, Jan 11, 2006 at 07:24:04PM +0800, Antonino A. Daplas wrote:

quoted

quoted

 It might be nice to
have these modules all autoloaded by something, but it's not strictly
necessary, and some users may not want the framebuffer activated
automatically?

As long as fbcon is compiled statically, you'll get the same 2.4 behavior.

<nod>  I leave it to Norbert to decide what the policy should be for fbcon
in the alpha kernel images.

quoted

quoted

The other issue (and the first thing I was trying to get work, which led me
to believe the fb was completely broken) is that, even though console works
on the framebuffer now, X does not.  This breakage corresponds to the kernel
upgrade, not to any changes in X, so still looks like a kernel bug to me.

quoted

quoted

If I turn off "UseFBDev" in my xorg.conf, X displays correctly.  I haven't
poked yet to see what this does performance-wise.

The "UseFBDev" option was added to X so it can cooperate with fbcon (ie, allows
X to restore the console state by using the fbdev API ).  In 2.6, fbcon has its
own way of restoring its own state so the "UseFBDev" option is not needed, and in
your case, is counterproductive.

Aha, didn't know that.  It had been my impression that this option meant
using the fbdev API for all vidcard access; but I guess that's the fbdev
driver, anyway.

So the real target for all of this was to get matroxfb working to the point
of running bterm successfully from the debian installer; and even with fbcon
working, in 2.6.14 running bterm or fbi gives me a whole lot of nothing
(console blanks, nothing gets displayed).  I'll give this another try with
2.6.15 once I've got my /usr partition mounting again.

-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon@debian.org                                   http://www.debian.org/