[Intel-gfx] [PATCH v3 2/5] drm/i915: Use batch pools with the command parser

[Daniel Vetter]

On Tue, Nov 04, 2014 at 08:46:16AM -0800, Volkin, Bradley D wrote:
> On Tue, Nov 04, 2014 at 02:30:14AM -0800, Daniel Vetter wrote:
> > On Mon, Nov 03, 2014 at 11:19:42AM -0800, bradley.d.volkin at intel.com wrote:
> > > +			flags |= I915_DISPATCH_SECURE;
> > 
> > I've forgotten one: You must have a full ppgtt check here since the
> > binding for aliasing ppgtt is still broken - all PIN_GLOBAL stuff still
> > ends up in the ppgtt ptes too, so yourspace can still get at your shadow
> > batch.
> 
> Hmm. We have a check for USES_PPGTT in i915_needs_cmd_parser(). Should we
> change that to USES_FULL_PPGTT and just not do any parser stuff without
> full ppgtt? I guess I'm just wondering if there's a benefit to running the
> parser in a case where we can't enable secure dispatch.

Given the situation we're in (ugly) I think the USES_PPGTT is ok (since
that's kinda indicating whether the cmd parser is technically possible).
But granting mode should have an additional USES_FULL_PPGTT with a FIXME
comment that until the aliasing ppgtt binding is fixed we can only hide
the shadow batch with full ppgtt.

Wrt still keeping it running: Atm we can't enable it yet by default (since
neither full ppgtt is solid nor the aliasing ppgtt binding fixed). So I
think it makes sense to keep it in scan-only mode for now. Hence also why
I think an additional check is better - we can then just bin that when
aliasing ppgtt is fixed.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch