aty128fb problems on iMac, kernel 2.8.1 (fwd)

aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Geert Uytterhoeven]

---------- Forwarded message ----------
Date: Wed, 22 Sep 2004 21:14:55 -0500 (CDT)
From: Theodore Kilgore <kilgota@banach.math.auburn.edu>
To: geert@linux-m68k.org
Subject: aty128fb problems on iMac, kernel 2.8.1 (fwd)

Hello.

I suspect that this problem does not come directly under your
specialization, but the mail to the person whose name is on the aty128fb
code ( below ) was returned as undeliverable. Therefore, after doing
enough Google searches to discover that this seems to be an old problem
but I could never come up with anyone who gave the answer, I am back to
square one. I do not ask you more than to help me get help from the right
person about this, realizing that it is not necessarily you yourself who
would want to get involved with the problem.

Assuming that this message reaches the right party, whoever that is, I
will be glad to send any information which is required, or to answer any
questions that I know how to answer. I also willingly volunteer my iMac
for testing new stuff for that hardware platform. That's why I got it, is
for experimentation of this very sort. Of course, I was expecting to
experiment with things like libgphoto2-gphoto2, on MaOS-X and on Linux on
a powerpd, but this is not a problem for me, if it helps to solve a
problem for some others.

Theodore Kilgore



---------- Forwarded message ----------
Date: Mon, 20 Sep 2004 22:13:36 -0500 (CDT)
From: Theodore Kilgore <kilgota@banach.math.auburn.edu>
To: Brad Douglas <brad@neruo.com>
Subject: aty128fb problems on iMac, kernel 2.8.1

Brad,

Sorry to disturb you, but I cannot seem to find any "live" leads to this
problem on Google -- all searches seem to lead to old, dead links.

I have recently decided to learn something about PowerPC architecture, so
I managed to obtain an old iMac, which has in it the model 750 CPU,
running at 350mhz.

I wanted to put both OS-X and Linux on it, with the idea of doing some
software development on both systems (I do some device support for
libgphoto2 and also some other USB stuff, in addition to my "day" job as a
math professor). I am not in a big hurry and being a Slackware lover, I
decided to try to install Slackintosh.

Well, after overcoming many obstacles I have been successful in much of
what I wanted to do, but now it comes to supporting X. Which was compiled
from source on the machine. But it seems that X needs more than what it
sees in order to run properly. I can only get 8-bit color, with a few
problems about artifacts (incidentally, these problems disappear if one
switches the VT back and forth).

I gather that the problem is, I do not have any proper kernel module to
support the framebuffer features of the video chip. So here now is
the problem I had with this.

Doing the kernel config, I enable framebuffer support, and I choose the
support for the ATI Rage 128, which is the chip in the iMac. It also says
in lspci that the ATI Rage 128 is an AGP 4x setup, running off the
uni-north bridge (I wonder if this has anything to do with a framebuffer
setup? I don't know but I enabled it anyway, as a character device module,
and also enabled the agp support module).

So, when I do make modules it says that it compiled the aty128fb module,
but it gives two error messages:

mac_find_mode  undefined

and

mac_vmode_to_var  undefined.


Unless these problems are resolved, I seriously suspect that the module
will not do its proper job. But when it is not available (I did not do
make modules_install after seeing these errors) then the X will not start
unless I fall back to the vga driver with fbdev support. If other than vga
is attempted, the errors say it cannot open any /dev/fbx where x=1,2,...

But if one does not try to run X, then fbset will give any information
wanted -- about /dev/fb0 -- including the fact that it will only give
8-bit color.

I would appreciate it very much if you could let me know what is the cure
for these error messages. It is probably something obvious, but I am sorry
it is a little bit out of my area of knowledge. And my problems in this
regard are compounded by the fact that for me the iMac is an unfamiliar
hardware platform.


Theodore Kilgore


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Antonino A. Daplas]

>
> mac_find_mode  undefined
>
> and
>
> mac_vmode_to_var  undefined.
>

Those functions are in macmodes.c but they are not exportable.

You can try changing this line in drivers/video/Makefile:

obj-$(CONFIG_FB_ATY128)           += aty128fb.o

to

obj-$(CONFIG_FB_ATY128)           += aty128fb.o macmodes.o

Possibly, you may need to remove these lines if the compile complains
of duplicate symbols:

# Only include macmodes.o if we have FB support and are PPC
ifeq ($(CONFIG_FB),y)
obj-$(CONFIG_PPC)                 += macmodes.o
endif

Tony




-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Michal 'hramrach' Suchanek]

On Thu, Sep 23, 2004 at 10:27:34AM +0200, Geert Uytterhoeven wrote:
> 
> 
> ---------- Forwarded message ----------
> Date: Mon, 20 Sep 2004 22:13:36 -0500 (CDT)
> From: Theodore Kilgore <kilgota@banach.math.auburn.edu>
> To: Brad Douglas <brad@neruo.com>
> Subject: aty128fb problems on iMac, kernel 2.8.1
> 
> Brad,
> 
> Sorry to disturb you, but I cannot seem to find any "live" leads to this
> problem on Google -- all searches seem to lead to old, dead links.
> 
> I have recently decided to learn something about PowerPC architecture, so
> I managed to obtain an old iMac, which has in it the model 750 CPU,
> running at 350mhz.
> 
> I wanted to put both OS-X and Linux on it, with the idea of doing some
> software development on both systems (I do some device support for
> libgphoto2 and also some other USB stuff, in addition to my "day" job as a
> math professor). I am not in a big hurry and being a Slackware lover, I
> decided to try to install Slackintosh.
> 
> Well, after overcoming many obstacles I have been successful in much of
> what I wanted to do, but now it comes to supporting X. Which was compiled
> from source on the machine. But it seems that X needs more than what it
> sees in order to run properly. I can only get 8-bit color, with a few
> problems about artifacts (incidentally, these problems disappear if one
> switches the VT back and forth).
> 
> I gather that the problem is, I do not have any proper kernel module to
> support the framebuffer features of the video chip. So here now is
> the problem I had with this.
> 
> Doing the kernel config, I enable framebuffer support, and I choose the
Also enable console on framebuffer support
> support for the ATI Rage 128, which is the chip in the iMac. It also says
> in lspci that the ATI Rage 128 is an AGP 4x setup, running off the
> uni-north bridge (I wonder if this has anything to do with a framebuffer
> setup? I don't know but I enabled it anyway, as a character device module,
> and also enabled the agp support module).
> 
> So, when I do make modules it says that it compiled the aty128fb module,
> but it gives two error messages:
> 
> mac_find_mode  undefined
> 
> and
> 
> mac_vmode_to_var  undefined.
Does it give the error messages during compilation or during module
loading?
If during module loading, you are probably missing some additional
module.
I had no problem with 2.6.7 kernel but I compiled the ati framebuffer
directly into kernel to get some usable console from the start.
I cannot curruntly say what exact ati chip the iMac has or what was my
kernel config because it is running OS X right now and I am not
physically at the machine.
You should definitely try to compile the aty128fb into kernel to avoid
problems with current fbcons.

If you cannot solve the problem I can look at my config later. It would
run X fine.

Hope this helps

Michal Suchanek


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Theodore Kilgore]

No time to reply to both messages, but this is to let you know I am still
here.


> Does it give the error messages during compilation or during module
> loading?

Both. At this point I _did_ try installing and using the module.

> If during module loading, you are probably missing some additional
> module.

Easily possible, but which one in that case? That brings up the question
of exactly what I _do_ need. Some examples (all of which are currently
thre, to be loaded as modules -- see discussion below about what chip is
there)

agpgart
r128
(the above under "Character devices" but the system ignores them)

and under Video

framebuffer support
aty128fb suport
(and the following again I have no idea if they are needed, or not)
aty support
radeon support
and then there are three different kinds of Apple support --palatinum and
Valkyrie and something else, and I have no idea of the relevance or
irrelevance of these, either)

I guess you can see  that one of the problems is that I am proceeding by
guesswork, because I cannot seem to find any relevant documentation which
would provide a roadmap. For example, does the AGP support have anything
at all to do with the price of eggs in China, on a system like mine? Or
are AGP support and framebuffer support sort of orthoghonal to one
another? I don't know and there does not seem to be any place to find out.

> I had no problem with 2.6.7 kernel but I compiled the ati framebuffer
> directly into kernel to get some usable console from the start.

A possibility. But I have no problem to get a usable text console and
fbset does report a setting if one runs it. And BTW I can get a lousy X
support by choosing for my drivers "vga" and "fbdev" (though with lots of
artifacts which will disappear if one does some VT-flipping, which seems
to release and reset something).

> I cannot curruntly say what exact ati chip the iMac has

Well, I could. But I am not in front of the machine, either. It is at home
and I am at work. I can only quote what I vaguely remember from lspci
output, at this point:

(paraphrasing) it says it has an Ati Rage 128 running as an AGP 4X card.

or what was my
> kernel config because it is running OS X right now and I am not
> physically at the machine.
> You should definitely try to compile the aty128fb into kernel to avoid
> problems with current fbcons.
>
a possibility. Byt as I understand the module is supposed to work and to
be automatically loaded by X?

> If you cannot solve the problem I can look at my config later. It would
> run X fine.
>
As far as my ignorant mind can see, the problem is not in X nor in the X
configuration. I could be wrong, of course.

Theodore Kilgore


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Benjamin Herrenschmidt]

> So, when I do make modules it says that it compiled the aty128fb module,
> but it gives two error messages:
> 
> mac_find_mode  undefined
> 
> and
> 
> mac_vmode_to_var  undefined.

Build the fbdev in the kernel, not as a module, there are a number of
issues with fbdev's as module, especially on Macs where offb (the
default 8 bits framebuffer) will take over, preventing you from loading
the specific ATI driver on top of it anyway.

Then, use X "ati" driver with Option "UseFBDev" "true" in the "Device"
section.

Ben.




-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Michal 'hramrach' Suchanek]

Yes, it's reported to to work tihs way.


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php

Re: aty128fb problems on iMac, kernel 2.8.1 (fwd)

[Theodore Kilgore]

Thanks.

Someone else made pretty much the same suggestion (to compile the support
into the kernel) two days ago. I did that and the problem is fixed. Now I
am tryint to compile KDE and can't figure out why in h--- I am able to
compile the qt libraries and then when I go on to the nex thing I get an
error message that the qt libraries are not there. But that's nothing to
do with this list.

In case anyone is curious:

This iMac has an ati rage 128, which lspci says is a 4X agp card. So, I
put in the kernel:

under Character devices:
AGP support
r128 support

and then
framebuffer support
aty128fb support

and (because my eyes are 63 years old just like the rest of me)
font support, with sun12x22 fonts installed.

so now I get a nice bootup with those nice big fonts, right away, instead
of the built-in Apple fonts where the characters are the size of tiny
black ants.

Then in xorg.cfg I chose r128 support with framebuffer option. Then X
maxes out at 1240x1024-75, with 24-bit color.

So as I said these problems are solved, and thanks for the help on the
mailing list because I needed for someont to tell me straight-up that it
isn't going to work if all this is done as modules. I still do not
understand completely why not, however, and if anyone has any ideas to try
out, to make that work, then as I said this machine is being built as a
test machine. Just let me know.

I do have one remaining question. This is not an X problem but a console
problem. Sometimes in the console I get an artifact on the right side and
on the bottom of the screen. This happens especially if Midnight Commander
is running, and I switch it to the "background" with Cntrl-O and then
switch the VT and then switch back again. Then the backwards "L" shaped
artifact may become quite visible and is bright blue. I have looked in
/etc/fb.modes to see if the settings there (which are for another card)
might be causing this, due to the settings for "bottom" and "right" being
a bit off, but it does not seem so. Indeed, the settings which show up in
console after bootup, when one types "fbset" are for 1024x768-75. But
the settings which fbset reports, do not agree with the 1024x768-75
settings which are given in /etc/fb.modes. (?).

Do not misunderstand. This is not a burning issue from the perspective of
a user. But it might be a good laboratory test in which one can learn
better how to control the video hardware.

Thanks again for the help.

Theodore Kilgore


On Sun, 26 Sep 2004, Benjamin Herrenschmidt wrote:

>
>
> > So, when I do make modules it says that it compiled the aty128fb module,
> > but it gives two error messages:
> >
> > mac_find_mode  undefined
> >
> > and
> >
> > mac_vmode_to_var  undefined.
>
> Build the fbdev in the kernel, not as a module, there are a number of
> issues with fbdev's as module, especially on Macs where offb (the
> default 8 bits framebuffer) will take over, preventing you from loading
> the specific ATI driver on top of it anyway.
>
> Then, use X "ati" driver with Option "UseFBDev" "true" in the "Device"
> section.
>
> Ben.
>
>


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php