[Intel-gfx] [RFC i-g-t] igt/gem_ringfill: Adds full ringbuffer preemption test

[Chris Wilson]

Quoting Antonio Argenziano (2017-08-15 22:44:21)
> Sending as RFC to get feedback on what would be the correct behaviour of
> the driver and, therefore, if the test is valid.
> 
> We do a wait while holding the mutex if we are adding a request and figure
> out that there is no more space in the ring buffer.
> Is that considered a bug?

Hmm, note that we may have contention on a struct drm_i915_file_private
lock for execbuf (as handles are in a per-fd namespace). To be 100% sure
that the clients are independent, use separate fd, and also note that we
can only have client separation for execbuf on full-ppgtt.
-Chris