On Wed, 15 Feb 2012 12:23:44 -0800
Eric Anholt <eric at anholt.net> wrote:
> On Tue, 14 Feb 2012 22:09:12 +0100, Ben Widawsky <ben at bwidawsk.net> wrote:
> > From: Ben Widawsky <bwidawsk at gmail.com>
> >
> > Create all the necessary data structures and hooks to support a context
> > API for userspace. Also plumb in the calls from the i915 core into the
> > context subsystem. There is a new file i915_context.c which will hold a
> > majority of the context related code. This file has nulled functions
> > that get called by the proper part of the i915 driver. This of course
> > requires a Makefile addition, as well as calls to the new nulled
> > functions from the other partsd of the driver (i915_dma.c).
> >
> > With that, there are 2 new ioctls defined , one for creating a context,
> > and one for destroying a context. These are defined in the necessary
> > places, and again have null functionality
> >
> > It's likely that a third ioctl will be required to associate the context
> > with an execbuf (likely execbuf3). Since that part is mostly trivial,
> > and will surely be contentious, it's ignored for now.
>
> > +struct drm_i915_gem_context_create {
> > + /* output: id of new context*/
> > + __u32 ctx_id;
> > +
> > + __u16 pad;
> > +};
>
> The pad should be made u32 at this point instead of fixing it up in a
> later commit.
Fixed. Thanks.
