drm/i915: Lift timeline into intel_context

Move the timeline from being inside the intel_ring to intel_context
itself. This saves much pointer dancing and makes the relations of the
context to its timeline much clearer.

Signed-off-by: Chris Wilson <[email protected]>
Reviewed-by: Matthew Auld <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

Author: ickle

drivers/gpu/drm/i915/gem/i915_gem_context.c

@@ -489,6 +489,29 @@ static void __assign_ppgtt(struct i915_gem_context *ctx,
 		i915_vm_put(vm);
 }
 
+static void __set_timeline(struct intel_timeline **dst,
+			   struct intel_timeline *src)
+{
+	struct intel_timeline *old = *dst;
+
+	*dst = src ? intel_timeline_get(src) : NULL;
+
+	if (old)
+		intel_timeline_put(old);
+}
+
+static void __apply_timeline(struct intel_context *ce, void *timeline)
+{
+	__set_timeline(&ce->timeline, timeline);
+}
+
+static void __assign_timeline(struct i915_gem_context *ctx,
+			      struct intel_timeline *timeline)
+{
+	__set_timeline(&ctx->timeline, timeline);
+	context_apply_all(ctx, __apply_timeline, timeline);
+}
+
 static struct i915_gem_context *
 i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
 {
@@ -531,7 +554,8 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
 			return ERR_CAST(timeline);
 		}
 
-		ctx->timeline = timeline;
+		__assign_timeline(ctx, timeline);
+		intel_timeline_put(timeline);
 	}
 
 	trace_i915_context_create(ctx);
@@ -1931,13 +1955,8 @@ static int clone_sseu(struct i915_gem_context *dst,
 static int clone_timeline(struct i915_gem_context *dst,
 			  struct i915_gem_context *src)
 {
-	if (src->timeline) {
-		GEM_BUG_ON(src->timeline == dst->timeline);
-
-		if (dst->timeline)
-			intel_timeline_put(dst->timeline);
-		dst->timeline = intel_timeline_get(src->timeline);
-	}
+	if (src->timeline)
+		__assign_timeline(dst, src->timeline);
 
 	return 0;
 }

drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c

@@ -2182,7 +2182,7 @@ static int eb_pin_context(struct i915_execbuffer *eb, struct intel_context *ce)
 static void eb_unpin_context(struct i915_execbuffer *eb)
 {
 	struct intel_context *ce = eb->context;
-	struct intel_timeline *tl = ce->ring->timeline;
+	struct intel_timeline *tl = ce->timeline;
 
 	mutex_lock(&tl->mutex);
 	intel_context_exit(ce);

drivers/gpu/drm/i915/gt/intel_context.c

@@ -68,7 +68,7 @@ int __intel_context_do_pin(struct intel_context *ce)
 			goto err;
 
 		GEM_TRACE("%s context:%llx pin ring:{head:%04x, tail:%04x}\n",
-			  ce->engine->name, ce->ring->timeline->fence_context,
+			  ce->engine->name, ce->timeline->fence_context,
 			  ce->ring->head, ce->ring->tail);
 
 		i915_gem_context_get(ce->gem_context); /* for ctx->ppgtt */
@@ -98,7 +98,7 @@ void intel_context_unpin(struct intel_context *ce)
 
 	if (likely(atomic_dec_and_test(&ce->pin_count))) {
 		GEM_TRACE("%s context:%llx retire\n",
-			  ce->engine->name, ce->ring->timeline->fence_context);
+			  ce->engine->name, ce->timeline->fence_context);
 
 		ce->ops->unpin(ce);
 
@@ -143,11 +143,12 @@ static void __intel_context_retire(struct i915_active *active)
 	struct intel_context *ce = container_of(active, typeof(*ce), active);
 
 	GEM_TRACE("%s context:%llx retire\n",
-		  ce->engine->name, ce->ring->timeline->fence_context);
+		  ce->engine->name, ce->timeline->fence_context);
 
 	if (ce->state)
 		__context_unpin_state(ce->state);
 
+	intel_timeline_unpin(ce->timeline);
 	intel_ring_unpin(ce->ring);
 	intel_context_put(ce);
 }
@@ -163,15 +164,21 @@ static int __intel_context_active(struct i915_active *active)
 	if (err)
 		goto err_put;
 
+	err = intel_timeline_pin(ce->timeline);
+	if (err)
+		goto err_ring;
+
 	if (!ce->state)
 		return 0;
 
 	err = __context_pin_state(ce->state);
 	if (err)
-		goto err_ring;
+		goto err_timeline;
 
 	return 0;
 
+err_timeline:
+	intel_timeline_unpin(ce->timeline);
 err_ring:
 	intel_ring_unpin(ce->ring);
 err_put:
@@ -218,6 +225,8 @@ intel_context_init(struct intel_context *ce,
 
 	ce->gem_context = ctx;
 	ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm);
+	if (ctx->timeline)
+		ce->timeline = intel_timeline_get(ctx->timeline);
 
 	ce->engine = engine;
 	ce->ops = engine->cops;
@@ -235,6 +244,8 @@ intel_context_init(struct intel_context *ce,
 
 void intel_context_fini(struct intel_context *ce)
 {
+	if (ce->timeline)
+		intel_timeline_put(ce->timeline);
 	i915_vm_put(ce->vm);
 
 	mutex_destroy(&ce->pin_mutex);
@@ -279,7 +290,7 @@ void intel_context_exit_engine(struct intel_context *ce)
 int intel_context_prepare_remote_request(struct intel_context *ce,
 					 struct i915_request *rq)
 {
-	struct intel_timeline *tl = ce->ring->timeline;
+	struct intel_timeline *tl = ce->timeline;
 	int err;
 
 	/* Only suitable for use in remotely modifying this context */

drivers/gpu/drm/i915/gt/intel_context.h

@@ -120,15 +120,15 @@ static inline void intel_context_put(struct intel_context *ce)
 
 static inline int __must_check
 intel_context_timeline_lock(struct intel_context *ce)
-	__acquires(&ce->ring->timeline->mutex)
+	__acquires(&ce->timeline->mutex)
 {
-	return mutex_lock_interruptible(&ce->ring->timeline->mutex);
+	return mutex_lock_interruptible(&ce->timeline->mutex);
 }
 
 static inline void intel_context_timeline_unlock(struct intel_context *ce)
-	__releases(&ce->ring->timeline->mutex)
+	__releases(&ce->timeline->mutex)
 {
-	mutex_unlock(&ce->ring->timeline->mutex);
+	mutex_unlock(&ce->timeline->mutex);
 }
 
 int intel_context_prepare_remote_request(struct intel_context *ce,

drivers/gpu/drm/i915/gt/intel_context_types.h

@@ -53,6 +53,7 @@ struct intel_context {
 
 	struct i915_vma *state;
 	struct intel_ring *ring;
+	struct intel_timeline *timeline;
 
 	unsigned long flags;
 #define CONTEXT_ALLOC_BIT 0

drivers/gpu/drm/i915/gt/intel_engine.h

@@ -196,9 +196,7 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value)
 #define CNL_HWS_CSB_WRITE_INDEX		0x2f
 
 struct intel_ring *
-intel_engine_create_ring(struct intel_engine_cs *engine,
-			 struct intel_timeline *timeline,
-			 int size);
+intel_engine_create_ring(struct intel_engine_cs *engine, int size);
 int intel_ring_pin(struct intel_ring *ring);
 void intel_ring_reset(struct intel_ring *ring, u32 tail);
 unsigned int intel_ring_update_space(struct intel_ring *ring);

drivers/gpu/drm/i915/gt/intel_engine_cs.c

@@ -680,7 +680,6 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine)
 		goto out_frame;
 
 	INIT_LIST_HEAD(&frame->ring.request_list);
-	frame->ring.timeline = &frame->timeline;
 	frame->ring.vaddr = frame->cs;
 	frame->ring.size = sizeof(frame->cs);
 	frame->ring.effective_size = frame->ring.size;

drivers/gpu/drm/i915/gt/intel_engine_types.h

@@ -69,7 +69,6 @@ struct intel_ring {
 	struct i915_vma *vma;
 	void *vaddr;
 
-	struct intel_timeline *timeline;
 	struct list_head request_list;
 	struct list_head active_link;
 
@@ -286,8 +285,6 @@ struct intel_engine_cs {
 
 	struct intel_sseu sseu;
 
-	struct intel_ring *buffer;
-
 	struct {
 		spinlock_t lock;
 		struct list_head requests;
@@ -306,6 +303,11 @@ struct intel_engine_cs {
 	struct drm_i915_gem_object *default_state;
 	void *pinned_default_state;
 
+	struct {
+		struct intel_ring *ring;
+		struct intel_timeline *timeline;
+	} legacy;
+
 	/* Rather than have every client wait upon all user interrupts,
 	 * with the herd waking after every interrupt and each doing the
 	 * heavyweight seqno dance, we delegate the task (of being the

drivers/gpu/drm/i915/gt/intel_lrc.c

@@ -2821,9 +2821,6 @@ logical_ring_default_irqs(struct intel_engine_cs *engine)
 
 int intel_execlists_submission_setup(struct intel_engine_cs *engine)
 {
-	/* Intentionally left blank. */
-	engine->buffer = NULL;
-
 	tasklet_init(&engine->execlists.tasklet,
 		     execlists_submission_tasklet, (unsigned long)engine);
 	timer_setup(&engine->execlists.timer, execlists_submission_timer, 0);
@@ -3071,23 +3068,13 @@ populate_lr_context(struct intel_context *ce,
 	return ret;
 }
 
-static struct intel_timeline *
-get_timeline(struct i915_gem_context *ctx, struct intel_gt *gt)
-{
-	if (ctx->timeline)
-		return intel_timeline_get(ctx->timeline);
-	else
-		return intel_timeline_create(gt, NULL);
-}
-
 static int __execlists_context_alloc(struct intel_context *ce,
 				     struct intel_engine_cs *engine)
 {
 	struct drm_i915_gem_object *ctx_obj;
+	struct intel_ring *ring;
 	struct i915_vma *vma;
 	u32 context_size;
-	struct intel_ring *ring;
-	struct intel_timeline *timeline;
 	int ret;
 
 	GEM_BUG_ON(ce->state);
@@ -3109,15 +3096,19 @@ static int __execlists_context_alloc(struct intel_context *ce,
 		goto error_deref_obj;
 	}
 
-	timeline = get_timeline(ce->gem_context, engine->gt);
-	if (IS_ERR(timeline)) {
-		ret = PTR_ERR(timeline);
-		goto error_deref_obj;
+	if (!ce->timeline) {
+		struct intel_timeline *tl;
+
+		tl = intel_timeline_create(engine->gt, NULL);
+		if (IS_ERR(tl)) {
+			ret = PTR_ERR(tl);
+			goto error_deref_obj;
+		}
+
+		ce->timeline = tl;
 	}
 
-	ring = intel_engine_create_ring(engine, timeline,
-					(unsigned long)ce->ring);
-	intel_timeline_put(timeline);
+	ring = intel_engine_create_ring(engine, (unsigned long)ce->ring);
 	if (IS_ERR(ring)) {
 		ret = PTR_ERR(ring);
 		goto error_deref_obj;