Re: [Intel-gfx] [PATCH 17/17] drm/i915: Support explicit fencing for execbuf

2016-08-30 Thread Joonas Lahtinen
On su, 2016-08-28 at 21:46 +0100, Chris Wilson wrote: > @@ -1915,6 +1979,16 @@ i915_gem_do_execbuffer(struct drm_device *dev, void > *data, >   ret = execbuf_submit(params, args, &eb->vmas); >  err_request: >   __i915_add_request(params->request, ret == 0); > + if (out_fence) { > +

Re: [Intel-gfx] [PATCH 17/17] drm/i915: Support explicit fencing for execbuf

2016-08-29 Thread Chris Wilson
On Sun, Aug 28, 2016 at 09:46:24PM +0100, Chris Wilson wrote: > +/* Setting I915_EXEC_FENCE_OUT causes the ioctl to return a sync_file fd > + * in the upper_32_bits(rsvd2) upon success. Ownership of the fd is given > + * to the caller, and it should be close() after use. (The fd is a regular > + *

[Intel-gfx] [PATCH 17/17] drm/i915: Support explicit fencing for execbuf

2016-08-28 Thread Chris Wilson
Now that the user can opt-out of implicit fencing, we need to give them back control over the fencing. We employ sync_file to wrap our drm_i915_gem_request and provide an fd that userspace can merge with other sync_file fds and pass back to the kernel to wait upon before future execution. Testcase