Commit 851ba2d6 authored Sep 09, 2016 by Chris Wilson

drm/i915: Serialise execbuf operation after a dma-buf reservation object



Now that we can wait upon fences before emitting the request, it becomes
trivial to wait upon any implicit fence provided by the dma-buf
reservation object.

To protect against failure, we force any asynchronous waits on a foreign
fence to timeout after 10s - so that a stall in another driver does not
permanently cripple ourselves. Still unpleasant though!

Testcase: igt/prime_vgem/fence-wait
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: John Harrison <john.c.harrison@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-21-chris@chris-wilson.co.uk

parent 0a046a0e

drivers/gpu/drm/i915/i915_gem_execbuffer.c

+11 −0

Original line number	Diff line number	Diff line
		@@ -1131,6 +1131,7 @@ i915_gem_execbuffer_move_to_gpu(struct drm_i915_gem_request *req,

		list_for_each_entry(vma, vmas, exec_list) {
		struct drm_i915_gem_object *obj = vma->obj;
		struct reservation_object *resv;

		if (obj->flags & other_rings) {
		ret = i915_gem_request_await_object
		@@ -1139,6 +1140,16 @@ i915_gem_execbuffer_move_to_gpu(struct drm_i915_gem_request *req,
		return ret;
		}

		resv = i915_gem_object_get_dmabuf_resv(obj);
		if (resv) {
		ret = i915_sw_fence_await_reservation
		(&req->submit, resv, &i915_fence_ops,
		obj->base.pending_write_domain, 10*HZ,
		GFP_KERNEL \| __GFP_NOWARN);
		if (ret < 0)
		return ret;
		}

		if (obj->base.write_domain & I915_GEM_DOMAIN_CPU)
		i915_gem_clflush_object(obj, false);
		}