[Intel-gfx] [PATCH] drm/i915: Add I915_PARAM_MMAP_GTT_VERSION to advertise unlimited mmaps

Joonas Lahtinen joonas.lahtinen at linux.intel.com
Thu Aug 25 12:15:36 UTC 2016


On to, 2016-08-25 at 09:27 +0100, Chris Wilson wrote:
> On Thu, Aug 25, 2016 at 11:00:54AM +0300, Joonas Lahtinen wrote:
> > 
> > On ke, 2016-08-24 at 20:42 +0100, Chris Wilson wrote:
> > > 
> > > +++ b/drivers/gpu/drm/i915/i915_drv.c
> > > @@ -355,6 +355,14 @@ static int i915_getparam(struct drm_device *dev, void *data,
> > >  	case I915_PARAM_MIN_EU_IN_POOL:
> > >  		value = INTEL_INFO(dev)->min_eu_in_pool;
> > >  		break;
> > > +	case I915_PARAM_MMAP_GTT_VERSION:
> > > +		/* 0 - Objects have to be smaller than aperture,
> > > +		 *     all simultaneous users have to fit within the
> > > +		 *     available space within the aperture.
> > > +		 * 1 - Objects can any size, and X,Y or untiled
> > > +		 */
> > > +		value = 1;
> > The actual test would be if this function call succeeds, though.
> > 
> > value = 0 would never be returned, so I'm not sure it'll be useful to
> > document as such. So I might document as "Failure to execute the query
> > means..."
> It's also going to get crowded if we put all snippets of ABI information
> here.

Yep, but 0 is never returned so the documentation is incorrect as is.
Better clearly describe the that before the parameter was introduced
conditions were as above.

> 
> If I 
> 
> #define I915_MMAP_GTT_VERSION 1
> /* ... */
> 
> at the start of i915_gem_fault() then hopefully we might remember to
> update it.

+1 on this.

Regards, Joonas

> -Chris
> 
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation


More information about the Intel-gfx mailing list