XOrg freeze that affects a lot of people

[Michel Dänzer]

On Sun, 2005-03-20 at 23:58 +0000, Charles Goodwin wrote:
> On Sun, 2005-03-20 at 16:48 -0500, Randall Nortman wrote:
> > FYI: I have the same symptom (not necessarily the same problem) with
> > the open-source radeon driver shipped by X.org.  (I have not tried
> > reducing the AGP bus speed, but I will.)  I think the issue here isn't
> > about specific bugs in specific drivers, but rather about modifying
> > the base code such that driver and hardware bugs are easier to
> > diagnose.
> 
> Right... finally the debate synthesises to the topic I intended; the way
> XOrg handles itself when a bad driver complains.  I've tried to stress
> multiple times that although the problem does not stem from XOrg, the
> end result is that XOrg locks up consuming CPU - i.e. the driver doesn't
> lock up, the kernel doesn't lock up, the only process locking up is
> XOrg.  So whilst the driver is the source of the bug, the end result
> (i.e. the effect on the end user) is XOrg hanging.  Note that XOrg isn't
> crashing, it's just getting confused because whatever operation it
> expected of the driver did not work.  To me, and I'm sure to anybody
> else (even those 'this is off topic' nay-sayers) at some point it could
> surely be possible for XOrg to realise something went wrong and deal
> with the problem gracefully.  There will (most likely) always be buggy
> drivers.  Does that mean that the way XOrg currently behaves under the
> influence of buggy graphics drivers is acceptable?

I'm afraid it doesn't matter whether it's acceptable; buggy drivers will
always be able to kill the system beyond recovery. In particular, as has
been pointed out already in this thread, if Alt-SysRq doesn't work,
there's no way for software to recover from that.


Michel, 'this is off topic' nay-sayer


-- 
Earthling Michel Dänzer      |     Debian (powerpc), X and DRI developer
Libre software enthusiast    |   http://svcs.affero.net/rm.php?r=daenzer