Subject: Re: Help: 3.0.1->4.0 screwed X display on iBook
To: Jan Henrik Sylvester <me@janh.de>
From: Michael Lorenz <macallan@netbsd.org>
List: port-macppc
Date: 12/29/2007 13:10:16
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Dec 29, 2007, at 12:44, Jan Henrik Sylvester wrote:

>> On Dec 29, 2007, at 07:24, Jan Henrik Sylvester wrote:
>>> I have an old iBook (clamshell) that has an "ATI 3D Rage Mobility  
>>> graphics controller" with "ATI Mach64 adapter" according to  
>>> XFree86.0.log, which matches what is reported by Open Firmware.
>>>
>>> On NetBSD 3.0.1, XFree86 did not work but died immediately.  
>>> Xmacppc did work (though 8 bit colors did not look great). I  
>>> postponed looking for the error that made XFree86 die to 4.0.
>>>
>>> On 4.0, both Xmacppc and XFree86 start, but both give totally  
>>> distorted graphics (in different ways). I played a lot with  
>>> XF86Config, but I cannot fix XFree86. (XF86Config seems to have  
>>> no influence on Xmacppc.)
>> Yikes, I bet it's a Rage LT - neither machfb nor XFree86's atimisc  
>> driver can handle it properly and I don't have the hardware so I  
>> can't really look into it.
>> What you can do is to use XFree86 with the wsfb driver, it is  
>> unaccelerated and limited to whatever video mode OpenFirmware sets  
>> up but at least it should give a usable X.
>
> As Bruce noted, it gives:
>
> (EE) wsfb(0): ioctl WSDISPLAYIO_LINEBYTES: Inappropriate ioctl for  
> device
> (EE) Screen(s) found, but none have a usable configuration.

Which console driver do you use? Adding that ioctl is easy enough.

>>> What the display exactly looks like with XFree86 depends on the  
>>> color depth. I describe it for 16 bit:
>>>
>>> The top quarter (full width) shows something that looks like  
>>> xterms on the left, but with some "bars". Typing something into  
>>> one of them, I only get a few dots. If that were fonts, they  
>>> would be way to small. On the right, I see the clock, which  
>>> basically looks correct, maybe too small, too. The mouse cursor  
>>> looks ok.
>> Looks like X and the hardware disagree about colour depth. Try 8  
>> bit, X may fail to tell the DAC about pixel format change for some  
>> reason.
>
> With 8 bit, the upper quarter looks similar, but with a little more  
> distortion, the next section is not only a quarter but almost the  
> rest of the screen with some black dots on green (not blue text),  
> and the last section is similar (to 16 bit) but smaller and with  
> black distorting the color bars.

This is weird. Same thing with 24bit colour?

>>> (--) PCI:*(0:16:0) ATI Technologies Inc Rage Mobility L AGP 2x  
>>> rev 100, Mem @ 0x91000000/24, 0x90000000/12, I/O @ 0x0400/8, BIOS  
>>> @ 0x90020000/17
>> Hmm, not an LT after all, so it's probably the whole L series :/
>
> In Open Firmware, it is /pci@f0000000/ATY,RageM_Lp@10, which has  
> the model ATY,RageMobilityL.
>
> Any more ideas?

I'll fix whatever console driver you use so at least wsfb will work.

> Did anything change for Xmacppc from 3.0.1 to 4.0 that broke it for  
> me?

Nothing that I'm aware of.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBR3aNicpnzkX8Yg2nAQJK7Qf8CrK1Pe+xy25Zf2IXMji0vvyOsXbHxYA9
zKNe8CxF/Ys0rfchF1TyN7J0iJcDkSKSkdP66SfqB44THSzs3Il62OKC2FhWmJ/R
HtTf/ygeW0CB24qBO99qa7scR7gzo/uOashdfs/SCCPzZJ6e4AWInYFWOmsm7Jlg
UWjGAvWNHssnE68LhC01Cw24/LgPF/uTJbvaEr5Sns++46ag5WyPAo2iAyzfty7l
hVCQseQbA3MP2gOG4EAQSbtPPLvA8ylZdnh0BRLGRaePoHxfK97MOQQ9z7g0gOWf
VZVuxmYSY/ETxbUN3i0T+WQ6LQBimp0nKg0QBjL2+Djlf84G0sWbHA==
=3R6w
-----END PGP SIGNATURE-----