Bugzilla Bug 174063

Hello,

i own an ATI PCIE X800 card and was unable to use radeonfb at all. The flat
panel didn't initialize and the display stays definitely black or goes
immediately in standby mode.

I made two tiny patches to get the radeon console framebuffer working with that
video card. I hope it works for a lot of cards and not just only for a pcie
X800 card.

My 1680x1050 flat panel is now recognized at its native resolution and without
forcing the video output to TMDS. Also i'm able to use gensplash as well.

Jj




Reproducible: Always




See attachments above

Created an attachment (id=115919) [edit]
radeon_monitor.c patch

Created an attachment (id=115921) [edit]
radeon_accel.c

Necessary to get dri working for the xorg server 

The patches should work with 2.6.21 kernel as well

Created an attachment (id=115962) [edit]
radeonfb.h

Created an attachment (id=115963) [edit]
radeon_base.c

without it ATOM BIOS won't be detected

Created an attachment (id=115964) [edit]
radeonfb.h

I checked twice and i think i added all the necessary patches this time to make
to video card working. 

Indeed, you need to apply radeon_base.c, radeon_accel.c, radeon_monitor.c,
radeonfb.h and ati_ids.h patches to the kernel.

Jj

Thanks for the patches. Did you write them yourself?

If so, can you explain the information sources? It seems to touch upon some
low-level aspects of this hardware.

(In reply to comment #8)
> Thanks for the patches. Did you write them yourself?
> 
> If so, can you explain the information sources? It seems to touch upon some
> low-level aspects of this hardware.
> 

Hello dsd,

unfortunately, it would be quite difficult for me to comment all these lines. I
don't have special knowledge in videocards hardware.

The most i have done was to copy-paste xf86-video-ati in the different kernel
files and to find out what was going wrong with my videocard. Most of the
issues i have solved were: the kernel hardlock, the black screen when switching
from x11 to the console, to make possible to use dri and to integrate monid
among dvi connectors. I made as few modifications as possible in order not to
mess the original kernel code too much.
Anyway if you are interested in a more complete version, that integrates more
x11 codes (but takes more place in the kernel), i can send it to you.  That
version is nearer from the one the maintainers have published. I have modified
it to make it compatible with my videocard too.

Most of my work was around the ati atom bios integration.

My card is a genuine ATI card (indeed ! :)) and the display is an acer
1680x1050 flat screen connected to a dvi-i connector. As a remark, when the
driver failed to probe the connector, it defaulted with a vga connector. That
could explain why many ati cards work flawlessly, when others like mine won't
;).


Jj

There are licensing issues when copying code from xf86-video-ati into the
kernel. 

> Anyway if you are interested in a more complete version, that integrates more
> x11 codes (but takes more place in the kernel), i can send it to you.  That
> version is nearer from the one the maintainers have published. I have modified
> it to make it compatible with my videocard too.

Can you expand on this? Are you saying that there is an alternative patch not
written by yourself which does almost the same thing?

(In reply to comment #10)
> There are licensing issues when copying code from xf86-video-ati into the
> kernel. 
> 

If i remember well the xf86-video-ati file is provided "as is".
Anyway, i prefer my own version of code. I have addressed the atom bios and got
the PLL infos in the same manner as in radeon_bios.c xorg file.

> > Anyway if you are interested in a more complete version, that integrates more
> > x11 codes (but takes more place in the kernel), i can send it to you.  That
> > version is nearer from the one the maintainers have published. I have modified
> > it to make it compatible with my videocard too.
> 
> Can you expand on this? Are you saying that there is an alternative patch not
> written by yourself which does almost the same thing?
> 

You ask me rather strange questions. If you are interested, you will find them
on the linux-fbdev-devel archive list,  http://lwn.net/Articles/215965/ or on 
http://search.gmane.org/?query=radeonfb+X800&author=&group=&sort=relevance&DEFAULTOP=and&xP=radeonfb%09x800&xFILTERS=--A

Moreover, it wouldn't be fair to say i have written them by myself. My own
patches are basic patches over more difficult patches that belong to their
respective owners :)

Anyway, it would be great to know if they are working for other people.

Jj

(In reply to comment #11)
> If i remember well the xf86-video-ati file is provided "as is".

You're right, there are no issues here.

> You ask me rather strange questions. If you are interested, you will find them
> on the linux-fbdev-devel archive list,  http://lwn.net/Articles/215965/ or on 
> http://search.gmane.org/?query=radeonfb+X800&author=&group=&sort=relevance&DEFAULTOP=and&xP=radeonfb%09x800&xFILTERS=--A

Thanks. It's just important to know the origins of patches and to credit the
appropriate people.

I'll follow this up upstream.

(In reply to comment #12) 
> Thanks. It's just important to know the origins of patches and to credit the
> appropriate people.
> 
> I'll follow this up upstream.
> 

I agree with you. As far as i know since 2005 the maintainer is
benh@kernel.crashing.org but it seems to be a lot more unofficial co
maintainers and i didn't find out who was the real initiator of radeon_accel.c
for example. That code is really important if you want the kernel driver to
work with x11. 
Anyway it is provided in the kernel, so it should be GPL.

I created an other diff file (aty-2.6.21.diff) for linux vanilla 2.6.21 kernel
this time. There is even less modifications in it than ever :) If you follow
the code upstream, this one will certainly be more appropriate. The patch will
succeed without an hitch. 

In fact, most of the issues now come from the kernel driver unable to auto
detect the monitor. It didn't check ddc_monid.

Created an attachment (id=119507) [edit]
patch for 2.6.21 vanilla kernel

the patch is build for the 2.6.21 vanilla kernel. It should certainly work with
the gentoo kernel 2.6.21 as well, but not with 2.6.20 series.

Jimmy, have you made any progress upstream with these patches? Where do things
stand now?

Closing this bug, if you have any further progress to report, please reopen it.