Created attachment 337920 [details] sysprof save of slowness Description of problem: I've been testing the latest rawhide on my x1950pro with kms + composting the last few days. Using a standard gnome desktops with 3d effects enabled. And I've noticed that sometimes X freezes for upto 2 seconds, the mouse pointer still moves, but doesn't change when it should, clicking on / in non of the windows works, clicks do get registered when it unfreezes. The easiest way to reproduce is to open 2 (or more tabs) in firefox and switch between them, for example these 2 work for me to reproduce: http://koji.fedoraproject.org/koji/buildinfo?buildID=93907 http://koji.fedoraproject.org/koji/buildinfo?buildID=95894 Open these 2 and then switching between the 2 tabs, (or away / from firefox), causes the freezes. Note that the freeze does not always happen, but only sometimes (circa 50% of the time). This is similar in many ways to bug 469378, yet very different, as it does not always happen (where as 469378 did have 100% reproducability). Also interesting is that when switching tabs, the webpage gets successfully rendered, and one can even enter for example 1 or 2 chars in a textinput field, and then X freezes. I've done some sysprof-ing of the stalls and this clearly points to a software fall back happening. But why only sometimes? I guess that some special memory allocation operation (GEM / TTM memory) fails and then a software fallback is used to work around this. I can compile the driver with software fallback debugging turned on and get an Xorg.log with that in if you want (what was the define again and in which file does it live?).
Could we get /var/log/Xorg.0.log at least from the current configuration, please?
Created attachment 338665 [details] xorg.log
Note this happens without compositing too.
This bug appears to have been reported against 'rawhide' during the Fedora 11 development cycle. Changing version to '11'. More information and reason for this action is here: http://fedoraproject.org/wiki/BugZappers/HouseKeeping
Can you test with fedora 12 livecd and report if it works with it. Here with my x1950 it does.
No need for the livecd, I've been running F-12 (rawhide) for a couple of months already, and have not seen this for a while. I just tried my own reproducer, and indeed it seems fixed. Closing this. If you're looking for radeon bugs to work on though, this one is easy to reproduce and quite nasty (X crashes): bug 526692