drm/i915: Track the context's seqno in its own timeline HWSP

Now that we have allocated ourselves a cacheline to store a breadcrumb,
we can emit a write from the GPU into the timeline's HWSP of the
per-context seqno as we complete each request. This drops the mirroring
of the per-engine HWSP and allows each context to operate independently.
We do not need to unwind the per-context timeline, and so requests are
always consistent with the timeline breadcrumb, greatly simplifying the
completion checks as we no longer need to be concerned about the
global_seqno changing mid check.

One complication though is that we have to be wary that the request may
outlive the HWSP and so avoid touching the potentially danging pointer
after we have retired the fence. We also have to guard our access of the
HWSP with RCU, the release of the obj->mm.pages should already be RCU-safe.

At this point, we are emitting both per-context and global seqno and
still using the single per-engine execution timeline for resolving
interrupts.

v2: s/fake_complete/mark_complete/

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190128181812.22804-5-chris@chris-wilson.co.uk

1 files changed, 20 insertions, 11 deletions

diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index e388f37743a2..fdbb3fe8eac9 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -832,10 +832,10 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
 	list_for_each_entry(rq, &engine->timeline.requests, link) {
 		GEM_BUG_ON(!rq->global_seqno);
 
-		if (i915_request_signaled(rq))
-			continue;
+		if (!i915_request_signaled(rq))
+			dma_fence_set_error(&rq->fence, -EIO);
 
-		dma_fence_set_error(&rq->fence, -EIO);
+		i915_request_mark_complete(rq);
 	}
 
 	/* Flush the queued requests to the timeline list (for retiring). */
@@ -845,9 +845,9 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
 
 		priolist_for_each_request_consume(rq, rn, p, i) {
 			list_del_init(&rq->sched.link);
-
-			dma_fence_set_error(&rq->fence, -EIO);
 			__i915_request_submit(rq);
+			dma_fence_set_error(&rq->fence, -EIO);
+			i915_request_mark_complete(rq);
 		}
 
 		rb_erase_cached(&p->node, &execlists->queue);
@@ -2044,10 +2044,17 @@ static u32 *gen8_emit_breadcrumb(struct i915_request *request, u32 *cs)
 	/* w/a: bit 5 needs to be zero for MI_FLUSH_DW address. */
 	BUILD_BUG_ON(I915_GEM_HWS_INDEX_ADDR & (1 << 5));
 
-	cs = gen8_emit_ggtt_write(cs, request->global_seqno,
+	cs = gen8_emit_ggtt_write(cs,
+				  request->fence.seqno,
+				  request->timeline->hwsp_offset);
+
+	cs = gen8_emit_ggtt_write(cs,
+				  request->global_seqno,
 				  intel_hws_seqno_address(request->engine));
+
 	*cs++ = MI_USER_INTERRUPT;
 	*cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+
 	request->tail = intel_ring_offset(request, cs);
 	assert_ring_tail_valid(request->ring, request->tail);
 
@@ -2056,18 +2063,20 @@ static u32 *gen8_emit_breadcrumb(struct i915_request *request, u32 *cs)
 
 static u32 *gen8_emit_breadcrumb_rcs(struct i915_request *request, u32 *cs)
 {
-	/* We're using qword write, seqno should be aligned to 8 bytes. */
-	BUILD_BUG_ON(I915_GEM_HWS_INDEX & 1);
-
 	cs = gen8_emit_ggtt_write_rcs(cs,
-				      request->global_seqno,
-				      intel_hws_seqno_address(request->engine),
+				      request->fence.seqno,
+				      request->timeline->hwsp_offset,
 				      PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
 				      PIPE_CONTROL_DEPTH_CACHE_FLUSH |
 				      PIPE_CONTROL_DC_FLUSH_ENABLE |
 				      PIPE_CONTROL_FLUSH_ENABLE |
 				      PIPE_CONTROL_CS_STALL);
 
+	cs = gen8_emit_ggtt_write_rcs(cs,
+				      request->global_seqno,
+				      intel_hws_seqno_address(request->engine),
+				      PIPE_CONTROL_CS_STALL);
+
 	*cs++ = MI_USER_INTERRUPT;
 	*cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;