What to blame for ATI lockups after upgrade ?

[Xavier Bestel]

On Mon, 2008-01-07 at 15:10 +0100, Jimmy Jazz wrote:
> Xavier Bestel a écrit :
> > On Sun, 2008-01-06 at 23:46 -0200, Frédéric L. W. Meunier wrote:
> >   
> >> On Mon, 31 Dec 2007, Alex Deucher wrote:
> >>
> >>     
> >>> On Dec 28, 2007 5:31 PM, Frédéric L. W. Meunier wrote:
> >>>       
> >>>> On Fri, 28 Dec 2007, Alex Deucher wrote:
> >>>>
> >>>>         
> >>>>> On Dec 28, 2007 5:00 PM, Frédéric L. W. Meunier wrote:
> >>>>>
> >>>>>           
> >>>>>> I used XOrg 1.4 with ATI driver 6.7.196 without any problems
> >>>>>> for a month. I upgraded to XOrg 1.4.0.90 on december 13 and to
> >>>>>> ATI 6.7.197 on the 21.
> >>>>>>
> >>>>>> On the 22, suddenly, while moving the mouse cursor, nothing
> >>>>>> worked anymore. I could move the mouse, but it was like an "I"
> >>>>>> and didn't do anything. The screen was apparently locked, as
> >>>>>> the clock didn't get updated.
> >>>>>>             
> >>>>> Try disabling the DRI or changing the AGPMode.  the newer ati driver
> >>>>> switched back to AGP 1x mode by default, previous versions defaulted
> >>>>> to whatever mode the bios set.  Both seem to have stability problems
> >>>>> on different cards.
> >>>>>
> >>>>> Option "AGPMode" "4"
> >>>>>           
> >>>> I always had that option set in my xorg.conf (the card supports
> >>>> 8x, but the motherboard only 4x), so the solution would be to
> >>>> disable DRI and, if I need DRI (I don't for now), return to
> >>>> 6.7.196 ?
> >>>>         
> >>> Well, it should act the same as previous versions with that option.
> >>> If it doesn't then you'll probably want to file a bug and attach your
> >>> xorg log and config along with any other relevant info.
> >>>       
> >> You're right. I tried without DRI and got the same after 6
> >> days. Then, I went back to 6.7.196, but the same happened
> >> today, after 3 days. So, it isn't the ATI driver.
> >>     
> >
> > FWIW I see lockups regularly too (it happens more often with compiz, but
> > I remember it happened with metacity (non-composited)).
> > As I've changed the whole PC, maybe it's the card failing.
> > How could I test for that ?
> >
> >   
> >> The strange thing is that it always happens while I move the 
> >> mouse. And in the four until now, while the system was idle and 
> >> I moved the mouse inside a window running screen.
> >>     
> 
> Are you using avant-window-navigator 0.2.1 as dock for gnome 2.20
> with/without compiz fusion, python 2.5 and kernel 2.6.24 ?
> I get the same symptom as you described until I definitely stopped using
> awn. Sad sad sad :(
> Xorg just freezes except the mouse. I still be able to run ssh but
> restarting gdm slowdown hard the computer
> It is like agony. Also, none of the processes are eating the cpu. If i'm
> lucky init 6 will reboot the computer ... 1 hour later.
> 
> I'm not able to know which of the program is really buggy. It is not
> necessarily awn, it just points out the issue.
> Anyway without it, everything is working well.

I've never used awn, only compiz. I've seen various lockup behaviors,
from the soft ones (ssh works, mouse still moves, only the GPU is stuck)
to the hard ones (the machine is freezed solid, only a hard reboot
works).
It looks like my soft-lockup issues (the ones resembling yours) don't
happen anymore since a few months. I only have the hard ones nowadays,
and they seem to be correlated with 3D use.

I'm using kernel 2.6.23 BTW. 2.6.24 is still at rc-7.
If I were less lazy, I'd try fglrx for a while to see if it's a driver
problem. :/

	Xav