1 files changed, 175 insertions, 95 deletions

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index bbd71af00a91..be185886e4fc 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -57,11 +57,13 @@ static struct i915_global_request {
 
 static const char *i915_fence_get_driver_name(struct dma_fence *fence)
 {
-	return "i915";
+	return dev_name(to_request(fence)->i915->drm.dev);
 }
 
 static const char *i915_fence_get_timeline_name(struct dma_fence *fence)
 {
+	const struct i915_gem_context *ctx;
+
 	/*
 	 * The timeline struct (as part of the ppgtt underneath a context)
 	 * may be freed when the request is no longer in use by the GPU.
@@ -74,7 +76,11 @@ static const char *i915_fence_get_timeline_name(struct dma_fence *fence)
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return "signaled";
 
-	return to_request(fence)->gem_context->name ?: "[i915]";
+	ctx = i915_request_gem_context(to_request(fence));
+	if (!ctx)
+		return "[" DRIVER_NAME "]";
+
+	return ctx->name;
 }
 
 static bool i915_fence_signaled(struct dma_fence *fence)
@@ -188,7 +194,7 @@ static void free_capture_list(struct i915_request *request)
 {
 	struct i915_capture_list *capture;
 
-	capture = request->capture_list;
+	capture = fetch_and_zero(&request->capture_list);
 	while (capture) {
 		struct i915_capture_list *next = capture->next;
 
@@ -214,7 +220,7 @@ static void remove_from_engine(struct i915_request *rq)
 		spin_lock(&engine->active.lock);
 		locked = engine;
 	}
-	list_del(&rq->sched.link);
+	list_del_init(&rq->sched.link);
 	spin_unlock_irq(&locked->active.lock);
 }
 
@@ -223,10 +229,7 @@ bool i915_request_retire(struct i915_request *rq)
 	if (!i915_request_completed(rq))
 		return false;
 
-	GEM_TRACE("%s fence %llx:%lld, current %d\n",
-		  rq->engine->name,
-		  rq->fence.context, rq->fence.seqno,
-		  hwsp_seqno(rq));
+	RQ_TRACE(rq, "\n");
 
 	GEM_BUG_ON(!i915_sw_fence_signaled(&rq->submit));
 	trace_i915_request_retire(rq);
@@ -272,8 +275,8 @@ bool i915_request_retire(struct i915_request *rq)
 	remove_from_client(rq);
 	list_del(&rq->link);
 
-	intel_context_exit(rq->hw_context);
-	intel_context_unpin(rq->hw_context);
+	intel_context_exit(rq->context);
+	intel_context_unpin(rq->context);
 
 	free_capture_list(rq);
 	i915_sched_node_fini(&rq->sched);
@@ -287,10 +290,7 @@ void i915_request_retire_upto(struct i915_request *rq)
 	struct intel_timeline * const tl = i915_request_timeline(rq);
 	struct i915_request *tmp;
 
-	GEM_TRACE("%s fence %llx:%lld, current %d\n",
-		  rq->engine->name,
-		  rq->fence.context, rq->fence.seqno,
-		  hwsp_seqno(rq));
+	RQ_TRACE(rq, "\n");
 
 	GEM_BUG_ON(!i915_request_completed(rq));
 
@@ -300,11 +300,11 @@ void i915_request_retire_upto(struct i915_request *rq)
 }
 
 static int
-__i915_request_await_execution(struct i915_request *rq,
-			       struct i915_request *signal,
-			       void (*hook)(struct i915_request *rq,
-					    struct dma_fence *signal),
-			       gfp_t gfp)
+__await_execution(struct i915_request *rq,
+		  struct i915_request *signal,
+		  void (*hook)(struct i915_request *rq,
+			       struct dma_fence *signal),
+		  gfp_t gfp)
 {
 	struct execute_cb *cb;
 
@@ -341,6 +341,8 @@ __i915_request_await_execution(struct i915_request *rq,
 	}
 	spin_unlock_irq(&signal->lock);
 
+	/* Copy across semaphore status as we need the same behaviour */
+	rq->sched.flags |= signal->sched.flags;
 	return 0;
 }
 
@@ -349,10 +351,7 @@ bool __i915_request_submit(struct i915_request *request)
 	struct intel_engine_cs *engine = request->engine;
 	bool result = false;
 
-	GEM_TRACE("%s fence %llx:%lld, current %d\n",
-		  engine->name,
-		  request->fence.context, request->fence.seqno,
-		  hwsp_seqno(request));
+	RQ_TRACE(request, "\n");
 
 	GEM_BUG_ON(!irqs_disabled());
 	lockdep_assert_held(&engine->active.lock);
@@ -376,7 +375,7 @@ bool __i915_request_submit(struct i915_request *request)
 	if (i915_request_completed(request))
 		goto xfer;
 
-	if (i915_gem_context_is_banned(request->gem_context))
+	if (intel_context_is_banned(request->context))
 		i915_request_skip(request, -EIO);
 
 	/*
@@ -415,7 +414,7 @@ xfer:	/* We may be recursing from the signal callback of another i915 fence */
 	if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) &&
 	    !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) &&
 	    !i915_request_enable_breadcrumb(request))
-		intel_engine_queue_breadcrumbs(engine);
+		intel_engine_signal_breadcrumbs(engine);
 
 	__notify_execute_cb(request);
 
@@ -441,10 +440,7 @@ void __i915_request_unsubmit(struct i915_request *request)
 {
 	struct intel_engine_cs *engine = request->engine;
 
-	GEM_TRACE("%s fence %llx:%lld, current %d\n",
-		  engine->name,
-		  request->fence.context, request->fence.seqno,
-		  hwsp_seqno(request));
+	RQ_TRACE(request, "\n");
 
 	GEM_BUG_ON(!irqs_disabled());
 	lockdep_assert_held(&engine->active.lock);
@@ -586,6 +582,21 @@ out:
 	return kmem_cache_alloc(global.slab_requests, gfp);
 }
 
+static void __i915_request_ctor(void *arg)
+{
+	struct i915_request *rq = arg;
+
+	spin_lock_init(&rq->lock);
+	i915_sched_node_init(&rq->sched);
+	i915_sw_fence_init(&rq->submit, submit_notify);
+	i915_sw_fence_init(&rq->semaphore, semaphore_notify);
+
+	rq->file_priv = NULL;
+	rq->capture_list = NULL;
+
+	INIT_LIST_HEAD(&rq->execute_cb);
+}
+
 struct i915_request *
 __i915_request_create(struct intel_context *ce, gfp_t gfp)
 {
@@ -643,35 +654,31 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
 		goto err_free;
 
 	rq->i915 = ce->engine->i915;
-	rq->hw_context = ce;
-	rq->gem_context = ce->gem_context;
+	rq->context = ce;
 	rq->engine = ce->engine;
 	rq->ring = ce->ring;
 	rq->execution_mask = ce->engine->mask;
 
-	rcu_assign_pointer(rq->timeline, tl);
+	RCU_INIT_POINTER(rq->timeline, tl);
+	RCU_INIT_POINTER(rq->hwsp_cacheline, tl->hwsp_cacheline);
 	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,
 		       tl->fence_context, seqno);
 
 	/* We bump the ref for the fence chain */
-	i915_sw_fence_init(&i915_request_get(rq)->submit, submit_notify);
-	i915_sw_fence_init(&i915_request_get(rq)->semaphore, semaphore_notify);
+	i915_sw_fence_reinit(&i915_request_get(rq)->submit);
+	i915_sw_fence_reinit(&i915_request_get(rq)->semaphore);
 
-	i915_sched_node_init(&rq->sched);
+	i915_sched_node_reinit(&rq->sched);
 
-	/* No zalloc, must clear what we need by hand */
-	rq->file_priv = NULL;
+	/* No zalloc, everything must be cleared after use */
 	rq->batch = NULL;
-	rq->capture_list = NULL;
-	rq->flags = 0;
-
-	INIT_LIST_HEAD(&rq->execute_cb);
+	GEM_BUG_ON(rq->file_priv);
+	GEM_BUG_ON(rq->capture_list);
+	GEM_BUG_ON(!list_empty(&rq->execute_cb));
 
 	/*
 	 * Reserve space in the ring buffer for all the commands required to
@@ -753,34 +760,37 @@ err_unlock:
 static int
 i915_request_await_start(struct i915_request *rq, struct i915_request *signal)
 {
-	struct intel_timeline *tl;
 	struct dma_fence *fence;
 	int err;
 
 	GEM_BUG_ON(i915_request_timeline(rq) ==
 		   rcu_access_pointer(signal->timeline));
 
+	fence = NULL;
 	rcu_read_lock();
-	tl = rcu_dereference(signal->timeline);
-	if (i915_request_started(signal) || !kref_get_unless_zero(&tl->kref))
-		tl = NULL;
-	rcu_read_unlock();
-	if (!tl) /* already started or maybe even completed */
-		return 0;
+	spin_lock_irq(&signal->lock);
+	if (!i915_request_started(signal) &&
+	    !list_is_first(&signal->link,
+			   &rcu_dereference(signal->timeline)->requests)) {
+		struct i915_request *prev = list_prev_entry(signal, link);
 
-	fence = ERR_PTR(-EBUSY);
-	if (mutex_trylock(&tl->mutex)) {
-		fence = NULL;
-		if (!i915_request_started(signal) &&
-		    !list_is_first(&signal->link, &tl->requests)) {
-			signal = list_prev_entry(signal, link);
-			fence = dma_fence_get(&signal->fence);
+		/*
+		 * Peek at the request before us in the timeline. That
+		 * request will only be valid before it is retired, so
+		 * after acquiring a reference to it, confirm that it is
+		 * still part of the signaler's timeline.
+		 */
+		if (i915_request_get_rcu(prev)) {
+			if (list_next_entry(prev, link) == signal)
+				fence = &prev->fence;
+			else
+				i915_request_put(prev);
 		}
-		mutex_unlock(&tl->mutex);
 	}
-	intel_timeline_put(tl);
-	if (IS_ERR_OR_NULL(fence))
-		return PTR_ERR_OR_ZERO(fence);
+	spin_unlock_irq(&signal->lock);
+	rcu_read_unlock();
+	if (!fence)
+		return 0;
 
 	err = 0;
 	if (intel_timeline_sync_is_later(i915_request_timeline(rq), fence))
@@ -811,31 +821,21 @@ already_busywaiting(struct i915_request *rq)
 }
 
 static int
-emit_semaphore_wait(struct i915_request *to,
-		    struct i915_request *from,
-		    gfp_t gfp)
+__emit_semaphore_wait(struct i915_request *to,
+		      struct i915_request *from,
+		      u32 seqno)
 {
 	const int has_token = INTEL_GEN(to->i915) >= 12;
 	u32 hwsp_offset;
-	int len;
+	int len, err;
 	u32 *cs;
 
 	GEM_BUG_ON(INTEL_GEN(to->i915) < 8);
 
-	/* Just emit the first semaphore we see as request space is limited. */
-	if (already_busywaiting(to) & from->engine->mask)
-		goto await_fence;
-
-	if (i915_request_await_start(to, from) < 0)
-		goto await_fence;
-
-	/* Only submit our spinner after the signaler is running! */
-	if (__i915_request_await_execution(to, from, NULL, gfp))
-		goto await_fence;
-
 	/* We need to pin the signaler's HWSP until we are finished reading. */
-	if (intel_timeline_read_hwsp(from, to, &hwsp_offset))
-		goto await_fence;
+	err = intel_timeline_read_hwsp(from, to, &hwsp_offset);
+	if (err)
+		return err;
 
 	len = 4;
 	if (has_token)
@@ -858,7 +858,7 @@ emit_semaphore_wait(struct i915_request *to,
 		 MI_SEMAPHORE_POLL |
 		 MI_SEMAPHORE_SAD_GTE_SDD) +
 		has_token;
-	*cs++ = from->fence.seqno;
+	*cs++ = seqno;
 	*cs++ = hwsp_offset;
 	*cs++ = 0;
 	if (has_token) {
@@ -867,6 +867,28 @@ emit_semaphore_wait(struct i915_request *to,
 	}
 
 	intel_ring_advance(to, cs);
+	return 0;
+}
+
+static int
+emit_semaphore_wait(struct i915_request *to,
+		    struct i915_request *from,
+		    gfp_t gfp)
+{
+	/* Just emit the first semaphore we see as request space is limited. */
+	if (already_busywaiting(to) & from->engine->mask)
+		goto await_fence;
+
+	if (i915_request_await_start(to, from) < 0)
+		goto await_fence;
+
+	/* Only submit our spinner after the signaler is running! */
+	if (__await_execution(to, from, NULL, gfp))
+		goto await_fence;
+
+	if (__emit_semaphore_wait(to, from, from->fence.seqno))
+		goto await_fence;
+
 	to->sched.semaphores |= from->engine->mask;
 	to->sched.flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
 	return 0;
@@ -894,18 +916,16 @@ i915_request_await_request(struct i915_request *to, struct i915_request *from)
 			return ret;
 	}
 
-	if (to->engine == from->engine) {
+	if (to->engine == from->engine)
 		ret = i915_sw_fence_await_sw_fence_gfp(&to->submit,
 						       &from->submit,
 						       I915_FENCE_GFP);
-	} else if (intel_engine_has_semaphores(to->engine) &&
-		   to->gem_context->sched.priority >= I915_PRIORITY_NORMAL) {
+	else if (intel_context_use_semaphores(to->context))
 		ret = emit_semaphore_wait(to, from, I915_FENCE_GFP);
-	} else {
+	else
 		ret = i915_sw_fence_await_dma_fence(&to->submit,
 						    &from->fence, 0,
 						    I915_FENCE_GFP);
-	}
 	if (ret < 0)
 		return ret;
 
@@ -945,8 +965,10 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)
 
 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
+			i915_sw_fence_set_error_once(&rq->submit, fence->error);
 			continue;
+		}
 
 		/*
 		 * Requests on the same timeline are explicitly ordered, along
@@ -980,6 +1002,57 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)
 	return 0;
 }
 
+static bool intel_timeline_sync_has_start(struct intel_timeline *tl,
+					  struct dma_fence *fence)
+{
+	return __intel_timeline_sync_is_later(tl,
+					      fence->context,
+					      fence->seqno - 1);
+}
+
+static int intel_timeline_sync_set_start(struct intel_timeline *tl,
+					 const struct dma_fence *fence)
+{
+	return __intel_timeline_sync_set(tl, fence->context, fence->seqno - 1);
+}
+
+static int
+__i915_request_await_execution(struct i915_request *to,
+			       struct i915_request *from,
+			       void (*hook)(struct i915_request *rq,
+					    struct dma_fence *signal))
+{
+	int err;
+
+	/* Submit both requests at the same time */
+	err = __await_execution(to, from, hook, I915_FENCE_GFP);
+	if (err)
+		return err;
+
+	/* Squash repeated depenendices to the same timelines */
+	if (intel_timeline_sync_has_start(i915_request_timeline(to),
+					  &from->fence))
+		return 0;
+
+	/* Ensure both start together [after all semaphores in signal] */
+	if (intel_engine_has_semaphores(to->engine))
+		err = __emit_semaphore_wait(to, from, from->fence.seqno - 1);
+	else
+		err = i915_request_await_start(to, from);
+	if (err < 0)
+		return err;
+
+	/* Couple the dependency tree for PI on this exposed to->fence */
+	if (to->engine->schedule) {
+		err = i915_sched_node_add_dependency(&to->sched, &from->sched);
+		if (err < 0)
+			return err;
+	}
+
+	return intel_timeline_sync_set_start(i915_request_timeline(to),
+					     &from->fence);
+}
+
 int
 i915_request_await_execution(struct i915_request *rq,
 			     struct dma_fence *fence,
@@ -1002,8 +1075,10 @@ i915_request_await_execution(struct i915_request *rq,
 
 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
+			i915_sw_fence_set_error_once(&rq->submit, fence->error);
 			continue;
+		}
 
 		/*
 		 * We don't squash repeated fence dependencies here as we
@@ -1013,8 +1088,7 @@ i915_request_await_execution(struct i915_request *rq,
 		if (dma_fence_is_i915(fence))
 			ret = __i915_request_await_execution(rq,
 							     to_request(fence),
-							     hook,
-							     I915_FENCE_GFP);
+							     hook);
 		else
 			ret = i915_sw_fence_await_dma_fence(&rq->submit, fence,
 							    I915_FENCE_TIMEOUT,
@@ -1180,8 +1254,7 @@ struct i915_request *__i915_request_commit(struct i915_request *rq)
 	struct intel_ring *ring = rq->ring;
 	u32 *cs;
 
-	GEM_TRACE("%s fence %llx:%lld\n",
-		  engine->name, rq->fence.context, rq->fence.seqno);
+	RQ_TRACE(rq, "\n");
 
 	/*
 	 * To ensure that this call will not fail, space for its emissions
@@ -1227,8 +1300,8 @@ void __i915_request_queue(struct i915_request *rq,
 
 void i915_request_add(struct i915_request *rq)
 {
-	struct i915_sched_attr attr = rq->gem_context->sched;
 	struct intel_timeline * const tl = i915_request_timeline(rq);
+	struct i915_sched_attr attr = {};
 	struct i915_request *prev;
 
 	lockdep_assert_held(&tl->mutex);
@@ -1238,6 +1311,9 @@ void i915_request_add(struct i915_request *rq)
 
 	prev = __i915_request_commit(rq);
 
+	if (rcu_access_pointer(rq->context->gem_context))
+		attr = i915_request_gem_context(rq)->sched;
+
 	/*
 	 * Boost actual workloads past semaphores!
 	 *
@@ -1533,10 +1609,14 @@ static struct i915_global_request global = { {
 
 int __init i915_global_request_init(void)
 {
-	global.slab_requests = KMEM_CACHE(i915_request,
-					  SLAB_HWCACHE_ALIGN |
-					  SLAB_RECLAIM_ACCOUNT |
-					  SLAB_TYPESAFE_BY_RCU);
+	global.slab_requests =
+		kmem_cache_create("i915_request",
+				  sizeof(struct i915_request),
+				  __alignof__(struct i915_request),
+				  SLAB_HWCACHE_ALIGN |
+				  SLAB_RECLAIM_ACCOUNT |
+				  SLAB_TYPESAFE_BY_RCU,
+				  __i915_request_ctor);
 	if (!global.slab_requests)
 		return -ENOMEM;