Noticed while reviewing Matt's atomic wm revert patch that we have fifo underruns pre-existing on this platform. Somehow only happens on this testcase. For context the long-term view:
I suppose I missed some other ilk/ibx corner case when trying to fix it all. Sigh.
Got it on basic-flip-vs-modeset also
Also shows intermittently in hang-read-crc-pipe-a and read-crc-pipe-b.
c.f. also https://bugs.freedesktop.org/show_bug.cgi?id=93640
Also seen in read-crc-pipe-a-frame-sequence.
In fact I think it can happen in pretty much any of the kms_* tests.
Raising priority due to BAT.
Hopefully fixed now
Author: Ville Syrjälä <email@example.com>
Date: Fri Apr 1 21:53:18 2016 +0300
drm/i915: Make sure LP1+ watermarks levels are preserved when going from 1 to 2 pipes