On Sun, Jun 30, 2013 at 1:31 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote: > I respectfully disagree. The semantics of the pin ioctl remain useful > even with the ggtt/ppgtt split, and I think barring its use forever more > is unwise. Not that pinning is a good solution, just in some cases it > may be the only solution. (It has proven useful in the past, it is > likely to do so again.) All that we need to do is note that the offset > returned by pin is ggtt and the offsets used by execbuffer are ppgtt. So > keep pin-ioctl and fix the test not to assume that pin.offset is > meaningful with execbuffer after HAS_FULL_PPGTT. I was eyeing for the most minimal fix, but this is ok with me, too. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
