Trying to boot a 3.16 kernel post-DRM merge - 3.16.0-0.rc0.git10.2.fc21 - on my main desktop, with the GeForce 9600 GT [10de:0622] (G94) adapter, results in the display dying when X should start up (bootsplash displays OK, and runlevel 3 boot works). Cannot reach a VT. Tested twice, first time the left head (of my two monitors) displayed a distinctive artifact pattern I've also seen when hitting a resume-from-suspend bug in 3.15 - see https://bugs.freedesktop.org/show_bug.cgi?id=77599 - second time both heads simply went into powersaving mode. Log of a boot with drm.debug=15 set will be attached. Aside from the kernel, I have: xorg-x11-drv-nouveau-1.0.10-4.fc21.x86_64 xorg-x11-server-common-1.15.99.902-8.20140428.fc21.x86_64
Created attachment 908798 [details] journal from a boot which hits the failure, with drm.debug=15 set
Created attachment 908799 [details] distinctive noise pattern associated with the bug This is the noise pattern I saw with the first boot attempt, which I've also seen with the resume-from-suspend bug on 3.15. It's not entirely visible from my phone cam, but both the black and white blocks have red, green and blue 'noise speckles' in them.
Ben reports on IRC that he's fixed this - I'm guessing http://cgit.freedesktop.org/nouveau/linux-2.6/commit/?h=drm-nouveau-next&id=ba5e01b0d0146c2d2bb5475cf1d6efec51bf2669 is the fix. It has to percolate through airlied up to Linus and then back into the Rawhide kernel before we can close this bug, though.
I did a scratch kernel build with the fix applied, and it looks good: I'm typing this comment from that kernel. Just have to wait for it to arrive in the official kernel build and we can close this.
3.16.0-0.rc2.git0.1.fc21.x86_64 has the fix for this, it boots fine on my affected system. Marking fixed.