[ubuntu-x] x.org cpu spin / hang while using XAA on G45 with 2.4.1 driver (stack included)

Bryce Harrington bryce at canonical.com
Fri Sep 26 22:10:37 BST 2008


On Fri, Sep 26, 2008 at 10:43:03PM +0200, Martin Olsson wrote:
> So, I was running X.org using XAA (and no compiz of course) on an Ubuntu (intrepid) machine with a G45 chipset (GMA X4500HD) when X.org just froze.
> I could still login through ssh and found the X process spinning with 100% CPU. I got the 2.4.1-1ubuntu6 version of the intel driver installed.
> I grabbed a couple of stacks out it (see below); they actually look similar (identical?) so maybe the machine was actually spinning inside dixLookupPrivate somehow?
> 
> One thing I found strange is the call to AbortServer() in frame 17 eventually seems
> to cause another call to AbortServer() in frame 7? Isn't that really weird?
> 
> If anyone actually finds this info useful I also used the "gcore" utility to save a crash dump (ping me and I'll send it).
> For now I will not file a bug about this because I can't reproduce it, I was just a one off random error.

Thanks for collecting a backtrace.  As you noticed, with hangs
backtraces don't seem to get much useful info - generally X is just
stuck inside some loop or other, and the trace just shows different
points within it.  In theory, with enough data samples you can identify
the outermost loop and analyze why it is not exiting the loop - manually
stepping through the code in gdb is the only way I know of for doing
this.

For the -intel driver, upstream does not support XAA any longer, and
probably wouldn't be surprised to hear there are bugs when using it.  ;-)

So, I'd be more interested in why you're using XAA instead of EXA.  If
there are other bugs with EXA, those will have some priority; XAA bugs
will probably be filed as won't-fix.

Bryce




More information about the Ubuntu-x mailing list