Created attachment 46018 [details] [review]
dmesg on lockup
Here's the dmesg output I just hit when trying the latest mainline kernel from 24 April on the Radeon HD 6570... I didn't check to see if it's similar to the lockups I've previously talked about for HD 6570 / HD 6850 in a slightly older kernel before the latest DRM merge. For this log attached, the GPU locked up right after starting the GNOME desktop with Compiz, soft reset, and recovered. Running Mesa Git master as of this morning.
Created attachment 46019 [details]
dmesg from nexuiz, reseting keeps happening
Here's another dmesg output (I have another one at drm.debug=15 I'll post also) that may or may not be the same exact problem (there's some similarities in the trace) as the other one I just posted, but this one is always been reproducible for me running Nexuiz on the HD 6570 and HD 6850 (anything pre-NI works fine in same configuration). When the GPU resets with Nexuiz in this environment, it basically gets stuck in an infinite loop of the GPU soft resetting and never able to properly restore itself.
Created attachment 46020 [details]
*** Bug 37151 has been marked as a duplicate of this bug. ***
are you running nexuiz in any particular way?
I'm wondering if this is TURKS specific, my BARTS can run gnome-shell + nexuiz timedemo demos/demo1.dem in full here with no hangs.
Don't know "nexuiz's timedemo demos/demo1.dem" but my barts (HD6850) hanging up while I am in desktop (with compiz) too. So I don't believe if this is related with just TURKS but BARTS...
*** Bug 38367 has been marked as a duplicate of this bug. ***
Now that I can boot at all with my 6850, I'm getting this bug, too. Seems to be triggered by 3d, I don't think gnome shell has triggered it alone yet.
Do the following patches help?
Yes! With the two patches (the vddc one was new - I compiled my first fedora kernel!) everything seems to work fine. I tested nexuiz and xbmc.
Fix in next kernel
on Dec 02, 2016 at 20:01:02.
(provided by the Example extension).