drm/i915: Peel dma-fence-chains for await
diff mbox series

Message ID 20200508184641.20524-1-chris@chris-wilson.co.uk
State New
Headers show
Series
  • drm/i915: Peel dma-fence-chains for await
Related show

Commit Message

Chris Wilson May 8, 2020, 6:46 p.m. UTC
From: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

To allow faster engine to engine synchronization, peel the layer of
dma-fence-chain to expose potential i915 fences so that the
i915_request code can emit HW semaphore wait/signal operations in the
ring which is faster than waking up the host to submit unblocked
workloads after interrupt notification.

This is similar to the peeling we do for e.g. dma_fence_array.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
---
I wanted to avoid the recursion, but the timeline squashing is useful
here.
---
 drivers/gpu/drm/i915/i915_request.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

Patch
diff mbox series

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 94189c7d43cd..d0321e990ad1 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -23,6 +23,7 @@ 
  */
 
 #include <linux/dma-fence-array.h>
+#include <linux/dma-fence-chain.h>
 #include <linux/irq_work.h>
 #include <linux/prefetch.h>
 #include <linux/sched.h>
@@ -1068,13 +1069,37 @@  i915_request_await_request(struct i915_request *to, struct i915_request *from)
 }
 
 static int
-i915_request_await_external(struct i915_request *rq, struct dma_fence *fence)
+__i915_request_await_external(struct i915_request *rq, struct dma_fence *fence)
 {
 	return i915_sw_fence_await_dma_fence(&rq->submit, fence,
 					     fence->context ? I915_FENCE_TIMEOUT : 0,
 					     I915_FENCE_GFP);
 }
 
+static int
+i915_request_await_external(struct i915_request *rq, struct dma_fence *fence)
+{
+	struct dma_fence *iter;
+
+	if (!to_dma_fence_chain(fence))
+		return __i915_request_await_external(rq, fence);
+
+	dma_fence_chain_for_each(iter, fence) {
+		struct dma_fence_chain *chain = to_dma_fence_chain(iter);
+		int err;
+
+		GEM_BUG_ON(!chain);
+
+		err = i915_request_await_dma_fence(rq, chain->fence);
+		if (err) {
+			dma_fence_put(iter);
+			return err;
+		}
+	}
+
+	return 0;
+}
+
 int
 i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)
 {