I do have a problem with my Apple PowerMac G5 running Gentoo Linux/PPC64 with
64bit userland. The machine has a Radeon 9600. The problem is that dri does not
work very stable. I can reproduce this with kernel 2.6.18-rc2 and kernel
184.108.40.206. I am using X version 7.1.
Sometimes X does not even start: the screen turns black and then the fans of my
machine begin to spin faster and faster (ssh into the machine and I see X uses
100% of my CPU). not every time my monitor sais 'no signal'.
If X starts then
a) the upper thirt of the screen is black, but I can move the mouse.
b) X starts normal, but after a while (sometimes minutes, sometimes hours) the
machine freezes. Sometimes I can move the cursor for a while, but I am not able
to click anything. then later even moving the mouse will not work. (ssh into
the machine and I see X uses 100% of my CPU)
I'll attach my xorg.conf, strace output and the terminal output when I execute
a plain 'X' as root.
Created attachment 6368 [details]
Created attachment 6369 [details]
strace output is too big for bugzilla attachment it seems.
I uploaded it: http://www.unixforces.net/upload/strace_X.txt
oh.. and one last note: this does not happen, if I comment out the line in my
xorg.conf which loads the dri module!
Could you try with git version of ati driver and drm driver.
yes, will try it this weekend.
I did a fresh
git clone git://anongit.freedesktop.org/mesa/drm/ and
git clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati
and installed both.
-> same problem
FYI Markus, I've got ebuilds for both in my overlay.
A commit to drm git to solve a bug maybe similar to your one.
Could you try with lastest drm.
(In reply to comment #9)
> A commit to drm git to solve a bug maybe similar to your one.
> Could you try with lastest drm.
hmm.. the last commit to git I see is from 3 days ago (according to gitweb).
and that one I'm using already. still same problem.
what kind of information do you need to get this fixed?
Have you tried enabling colour tiling (won't work with Option "UseFBDev")? Some
people have reported this to increase stability with R300 class hardware.
If I don't use the Option "UseFBDev" then I get a realy screwed up picture. Not
easy to describe. If you want I'll make a picture.
But for now I cannot use colour tiling then.
Please attach a log file from trying to start without Option "UseFBDev".
Also, you may want to use Option "ReverseDDC" on the G5's 9600 or it might not
get the right EDID info from the monitor
thanks, will try it. unfortunatly I'm not near my G5 at the moment so it could
take a few days before I can test.
Created attachment 6637 [details]
the new xorg.conf with suggested options
Created attachment 6638 [details]
unfortunatly X won't even start with that as you can see.
I am showing a similar or related problem. I don't get X starting due to now
modes being valid. I also get negative values in Clock range. However, if X is
started in a 32bit userland (chroot) then I can run everything fine. Clock
range is fine from there. Looks to me like there is a 32/64bit problem in the
Is this still an issue? The bogus clocks should have been fixed with
c979c7f87fc449cc9631820101edff82c3013e66 maybe the other problem too.
(In reply to comment #20)
> Is this still an issue? The bogus clocks should have been fixed with
> c979c7f87fc449cc9631820101edff82c3013e66 maybe the other problem too.
yes, the bogus clocks are fixed now and DRI works on ppc64 again. I don't have
black screens on X startup any more. Unfortunately my G5 still locks up in some
situations. look at b) in comment #0: the machine freezes irregularly.
closing. _this_ problem seems to be fixed for quite some time now.