[Intel-gfx] [PATCH] drm/i915: Initialize ring->obj last and use it as a marker
chris at chris-wilson.co.uk
Sun Apr 15 00:27:39 CEST 2012
On Sun, 15 Apr 2012 00:15:24 +0200, Daniel Vetter <daniel at ffwll.ch> wrote:
> On Sat, Apr 14, 2012 at 12:39:44PM +0100, Chris Wilson wrote:
> > Use the assignment of ring->obj as a marker that the ring is active, and
> > so be careful not to initialise that value too early in case we need to
> > perform some workarounds that would ordinarily require touching the ring
> > whilst prepping the object.
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> I guess we don't need this with the actual workaround that needs this
> patch (i.e. that try-to-make-i845 patch)? Or is other stuff in the works?
It's relevant for the quirk where we might need to idle the gpu whilst
manipulating the GATT, but I'm still chasing the magic to make it work
for 845g. I also thought it relevant in the light of your other
ringbuffer cleanups. There's probably a few other places where we
could make the code more architecture agnostic and just operate on what
has been setup.
Chris Wilson, Intel Open Source Technology Centre
More information about the Intel-gfx