drm/i915: Keep timeline HWSP allocated until idle across the system

In preparation for enabling HW semaphores, we need to keep in flight
timeline HWSP alive until its use across entire system has completed,
as any other timeline active on the GPU may still refer back to the
already retired timeline. We both have to delay recycling available
cachelines and unpinning old HWSP until the next idle point.

An easy option would be to simply keep all used HWSP until the system as
a whole was idle, i.e. we could release them all at once on parking.
However, on a busy system, we may never see a global idle point,
essentially meaning the resource will be leaked until we are forced to
do a GC pass. We already employ a fine-grained idle detection mechanism
for vma, which we can reuse here so that each cacheline can be freed
immediately after the last request using it is retired.

v3: Keep track of the activity of each cacheline.
v4: cacheline_free() on canceling the seqno tracking
v5: Finally with a testcase to exercise wraparound
v6: Pack cacheline into empty bits of page-aligned vaddr
v7: Use i915_utils to hide the pointer casting around bit manipulation

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/20190301170901.8340-2-chris@chris-wilson.co.uk

1 files changed, 16 insertions, 15 deletions

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 719d1a5ab082..d354967d6ae8 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -325,11 +325,6 @@ void i915_request_retire_upto(struct i915_request *rq)
 	} while (tmp != rq);
 }
 
-static u32 timeline_get_seqno(struct i915_timeline *tl)
-{
-	return tl->seqno += 1 + tl->has_initial_breadcrumb;
-}
-
 static void move_to_timeline(struct i915_request *request,
 			     struct i915_timeline *timeline)
 {
@@ -532,8 +527,10 @@ struct i915_request *
 i915_request_alloc(struct intel_engine_cs *engine, struct i915_gem_context *ctx)
 {
 	struct drm_i915_private *i915 = engine->i915;
-	struct i915_request *rq;
 	struct intel_context *ce;
+	struct i915_timeline *tl;
+	struct i915_request *rq;
+	u32 seqno;
 	int ret;
 
 	lockdep_assert_held(&i915->drm.struct_mutex);
@@ -610,24 +607,27 @@ i915_request_alloc(struct intel_engine_cs *engine, struct i915_gem_context *ctx)
 		}
 	}
 
-	rq->rcustate = get_state_synchronize_rcu();
-
 	INIT_LIST_HEAD(&rq->active_list);
+
+	tl = ce->ring->timeline;
+	ret = i915_timeline_get_seqno(tl, rq, &seqno);
+	if (ret)
+		goto err_free;
+
 	rq->i915 = i915;
 	rq->engine = engine;
 	rq->gem_context = ctx;
 	rq->hw_context = ce;
 	rq->ring = ce->ring;
-	rq->timeline = ce->ring->timeline;
+	rq->timeline = tl;
 	GEM_BUG_ON(rq->timeline == &engine->timeline);
-	rq->hwsp_seqno = rq->timeline->hwsp_seqno;
+	rq->hwsp_seqno = tl->hwsp_seqno;
+	rq->hwsp_cacheline = tl->hwsp_cacheline;
+	rq->rcustate = get_state_synchronize_rcu(); /* acts as smp_mb() */
 
 	spin_lock_init(&rq->lock);
-	dma_fence_init(&rq->fence,
-		       &i915_fence_ops,
-		       &rq->lock,
-		       rq->timeline->fence_context,
-		       timeline_get_seqno(rq->timeline));
+	dma_fence_init(&rq->fence, &i915_fence_ops, &rq->lock,
+		       tl->fence_context, seqno);
 
 	/* We bump the ref for the fence chain */
 	i915_sw_fence_init(&i915_request_get(rq)->submit, submit_notify);
@@ -687,6 +687,7 @@ err_unwind:
 	GEM_BUG_ON(!list_empty(&rq->sched.signalers_list));
 	GEM_BUG_ON(!list_empty(&rq->sched.waiters_list));
 
+err_free:
 	kmem_cache_free(global.slab_requests, rq);
 err_unreserve:
 	mutex_unlock(&ce->ring->timeline->mutex);