Re: Getting Dual Independent Heads to work on Debian(sid) on iBook
Ben,
Thanks for the information. I used fbset -x and updated XF86Config-4
accordingly.
All of the test results from my original post were without using the
kernel fbdev driver. I just now went ahead and added Option "UseFBDev"
to the lcd-related device section in XF86Config-4 and set BusID to
0:16:0 in both device sections and now the lcd flickering is gone but
both screens are yellow and I am still in clone mode.If I just comment
out the Option "UseFBDev" then restart the X server then the lcd
flickering returns (and both screens are still yellowish etc.).
Given this result I tried comparing the sections of code in the kernel
fbdev driver and the XF86 radeon driver that sets up the PLL and there
was just a few minor diff's in the default min/max values. I went ahead
and changed these to match and recompiled/installed XFree86 4.3 etc. but
it didn't make any difference...the flickering still occured. Maybe I am
looking at the wrong parts of code....hmmm....
I guess the other weird thing is the color issue (yellowing). It also
seems to be related to the BusID setting. If I use 0:16:0 in the device
sections then its yellow. If I use 0:10:0 in just the external crt
device section (an incorrect value per your earlier comments) then the
colors are all ok on both screens etc...
In any case I appreiciate your comments. I guess I really want to get
down to enabling the dual-head mode first and then I can worry about any
flickering and color issues secondarily. On that note I looked at my
XF86 log file and noticed the following error:
(EE) RADEON (0): FBIOPAN_DISPLAY: Invalid argument
This end up being in the fb_ioctl function in the kernel fbmem.c
module.I am not sure whats going on in this particular section so maybe
you can comment.
thanks again,
-dean andreakis
On Fri, 2003-06-13 at 09:06, Benjamin Herrenschmidt wrote:
> > hing is ok.
> >
> > 2. On the iBook if I set both BusID's in both device sections to 0:10:0
> > then X won't startup even though this is the ID reported by lspci for
> > the ATI chip. If I just set the BusID in the device section associated
> > with the external CRT to 0:10:0 then X will start and the colors are ok
> > but there are scrolling lines in the lcd panel. Also, if I run glxinfo
> > then it reports information on just one screen even though I have
> > defined two of them.
>
> I think lspci has hexadecimal output while XF86Config expects decimal,
> try 0:16:0
>
> > 3. If both are set to BusID 0:16:0 (or just the device section
> > associated with the external CRT) then X starts but both screens are
> > yellowish and there are scrolling lines in the lcd panel. Also, if I run
> > glxinfo then it reports information on two screens (as I expect).
>
> First get the native modeline of your panel with fbset -x before you
> launch X and make sure X uses that for the panel.
>
> Also, there is a problem with the LCD panel on some iBooks, possibly
> yours. On these, for an unknown reason, the PLL values we calcule
> provide a flickering display. I have hard coded values in radeonfb
> but of course that's not a solution for you. Keeping the firmware set
> values would work too if XFree driver can be instructed to do that.
>
> I'm not too sure how to solve that last problem properly. The
> interesting fact is that if I use the same PLL values as MacOS uses,
> I also have flickering display. So there is something else that is
> interfering with the PLL in some way. It seems though that MacOS
> reconfigure the MPLL and SPLL (to obtain the same final clocks but
> using a different pair of dividers), there might be something related
> to the flickering in there... Only ATI can tell.
>
> > >From previous mailing list responses around this issue and my test
> > results so far I have thought that further development work is required
> > on the radeon XFree86 4.3.0 drivers in order to properly enable this
> > dual-headed mode on the iBook and relieve some of the other odd issues
> > defined above.
> >
> > Any help in pointing me in the right direction as far as changes
> > required etc. with the driver(s) would be appreciated.
> >
> > -dean andreakis
> --
> Benjamin Herrenschmidt <benh@kernel.crashing.org>
>
Reply to: