On Tue, Apr 24, 2012 at 06:53:39PM +0100, Chris Wilson wrote: > We seem to trigger an occasional error when the GPU tries to dereference > a PTE for an overlay plane prior to us populating the offsets in the > overlay registers. This is despite the Overlay Command being set to off. > Close this window delaying the switch on until after those registers are > initialised. > > In conjunction with inspecting the state of how we switch on the > overlay, we review when we mark the overlay as active. The overlay > becomes active when we queue the commands to change state in the > ringbuffer, and not before a potential failure. > > References: https://bugs.freedesktop.org/show_bug.cgi?id=48939 > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk> I've tested this on my i855gm, and the overlay seems to still work. But since this changes the way we enable the overlay (the overlay flip ON now directly enables the overlay) I'd like to wait for another tested-by from someone actually hitting an overlay issue before I merge this. -Daniel -- Daniel Vetter Mail: daniel at ffwll.ch Mobile: +41 (0)79 365 57 48
