Created attachment 17215 [details]
X output (with extra debugging) of a crashed session
When trying to set up a Zaphod configuration (:0.0 from an nVidia GeForce
7300 SE, :0.1 and :0.2 from an ATI RV610. The monitors are all Samsung
206NW, with analog input.), I get a crash in radeon_atombios.c:CailReadATIRegister because info->MMIO is NULL.
(Example session with some extra debugging output is attached.)
After commenting out all but the initial info->MMIO = NULL assignments
in radeon_driver.c, X comes up, but there seems to be some confusion
about where the pixel data for each screen comes from.
By the way, the info->MMIO = NULL after RADEONUnmapMMIO in
radeon_driver.c:RADEONPreInit seem to be redundant.
The scenario I get when the X server comes up is as follows: :0.0 is
on the nVidia card and works as expected. The other two screens also
display an image, but only :0.1 has the expected content. :0.2 shows
black and grey horizontal stripes, each perhaps about 8 pixels high,
overlaid with some color noise.
When trying to display anything on :0.2, nothing shows, but the command
is accepted. So X seems to think that there's a perfectly good :0.2.
:0.2 responds to xrandr, I just can't get it to display anything.
Further investigation reveals that the driver selected DVI-0 for :0.1
and VGA-0 for :0.2, and that both access the same monitor.
All this is with git revision eed241553748125e902c44dcc8cf8e0605be071b.
Any suggestions ?
Created attachment 17216 [details]
xorg.conf used for all this
Created attachment 17217 [details]
X server log for a session without crash but confused framebuffer
Dave Airlie fixed the crash in xf86-video-ati, git revision
and the framebuffer confusion in git revision
My triple-head Zaphod is now happy :-) Thanks !