https://bugs.freedesktop.org/show_bug.cgi?id=79488
Priority: medium Bug ID: 79488 Assignee: dri-devel@lists.freedesktop.org Summary: Checkerboard tearing on Radeon card Severity: normal Classification: Unclassified OS: All Reporter: dan.doel@gmail.com Hardware: Other Status: NEW Version: unspecified Component: DRM/Radeon Product: DRI
Created attachment 100211 --> https://bugs.freedesktop.org/attachment.cgi?id=100211&action=edit dmesg while the machine is in the 'tearing' state
For several months now, I've occasionally run into situations where my screen displays a strange artifact or tearing pattern. It is only visible when things are changing, of course, and it looks like the monitor is divided into small squares, and two different 'frames' are drawn alternately. Here is a video of the problem I took back at the end of February:
https://www.youtube.com/watch?v=e2T8BeE0j74
I've just encountered this again, and it reminded me to finally post a bug.
You'll notice that the video was recorded on a cell phone. I tried using a desktop recorder (the one built into gnome-shell) and the resulting video did not display the artifacts. So it would seem to be happening at a level below what capture can detect.
I'll attach dmesg and xorg logs from when the problem is happening. You'll see that there are lots of page flip errors in both. However, I think that actually might be a red herring. Doing a tail -f on the logs, I can cause artifacts to display (for instance, selecting a different channel in my irc client, causing highlighting to change, which briefly displays checkerboard highlights) without a corresponding xorg error. And I also frequently see those errors accompanied by ordinary tearing. But I can't definitively say that they aren't related, either.
I'm using:
mesa from git as of about a week ago (rev 63069) latest glamor (as of 2014-03-13) xf86-video-ati 7.3.0 kernel 3.14.4
But this has been happening for significantly longer than just those versions.
I have a Radeon HD 7870 (pitcairn).
I'm not certain that I've assigned this bug to the right project, although at least some of the visible errors involve drm. Apologies if it's the fault of something else, though.