11 files changed, 54 insertions, 25 deletions

diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index 3c0f490ff2c7..7269c87c1372 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -177,7 +177,8 @@ static int __engine_park(struct intel_wakeref *wf)
 
 	engine->execlists.no_priolist = false;
 
-	intel_gt_pm_put(engine->gt);
+	/* While gt calls i915_vma_parked(), we have to break the lock cycle */
+	intel_gt_pm_put_async(engine->gt);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.h b/drivers/gpu/drm/i915/gt/intel_engine_pm.h
index 739c50fefcef..24e20344dc22 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.h
@@ -31,6 +31,16 @@ static inline void intel_engine_pm_put(struct intel_engine_cs *engine)
 	intel_wakeref_put(&engine->wakeref);
 }
 
+static inline void intel_engine_pm_put_async(struct intel_engine_cs *engine)
+{
+	intel_wakeref_put_async(&engine->wakeref);
+}
+
+static inline void intel_engine_pm_flush(struct intel_engine_cs *engine)
+{
+	intel_wakeref_unlock_wait(&engine->wakeref);
+}
+
 void intel_engine_init__pm(struct intel_engine_cs *engine);
 
 #endif /* INTEL_ENGINE_PM_H */
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
index 470fbdc30e5a..f6b5169d623f 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
@@ -105,7 +105,6 @@ static int __gt_park(struct intel_wakeref *wf)
 static const struct intel_wakeref_ops wf_ops = {
 	.get = __gt_unpark,
 	.put = __gt_park,
-	.flags = INTEL_WAKEREF_PUT_ASYNC,
 };
 
 void intel_gt_pm_init_early(struct intel_gt *gt)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.h b/drivers/gpu/drm/i915/gt/intel_gt_pm.h
index b3e17399be9b..990efc27a4e4 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_pm.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.h
@@ -32,6 +32,11 @@ static inline void intel_gt_pm_put(struct intel_gt *gt)
 	intel_wakeref_put(&gt->wakeref);
 }
 
+static inline void intel_gt_pm_put_async(struct intel_gt *gt)
+{
+	intel_wakeref_put_async(&gt->wakeref);
+}
+
 static inline int intel_gt_pm_wait_for_idle(struct intel_gt *gt)
 {
 	return intel_wakeref_wait_for_idle(&gt->wakeref);
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 33ce258d484f..b65bc06855b0 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1172,7 +1172,7 @@ __execlists_schedule_out(struct i915_request *rq,
 
 	intel_engine_context_out(engine);
 	execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT);
-	intel_gt_pm_put(engine->gt);
+	intel_gt_pm_put_async(engine->gt);
 
 	/*
 	 * If this is part of a virtual engine, its next request may
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index b7007cd78c6f..0388f9375366 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -1125,7 +1125,7 @@ int intel_engine_reset(struct intel_engine_cs *engine, const char *msg)
 out:
 	intel_engine_cancel_stop_cs(engine);
 	reset_finish_engine(engine);
-	intel_engine_pm_put(engine);
+	intel_engine_pm_put_async(engine);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c
index 20b9c83f43ad..cbf6b0735272 100644
--- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c
@@ -51,11 +51,12 @@ static int live_engine_pm(void *arg)
 				pr_err("intel_engine_pm_get_if_awake(%s) failed under %s\n",
 				       engine->name, p->name);
 			else
-				intel_engine_pm_put(engine);
-			intel_engine_pm_put(engine);
+				intel_engine_pm_put_async(engine);
+			intel_engine_pm_put_async(engine);
 			p->critical_section_end();
 
-			/* engine wakeref is sync (instant) */
+			intel_engine_pm_flush(engine);
+
 			if (intel_engine_pm_is_awake(engine)) {
 				pr_err("%s is still awake after flushing pm\n",
 				       engine->name);
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index 5448f37c8102..dca15ace88f6 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -672,12 +672,13 @@ void i915_active_acquire_barrier(struct i915_active *ref)
 	 * populated by i915_request_add_active_barriers() to point to the
 	 * request that will eventually release them.
 	 */
-	spin_lock_irqsave_nested(&ref->tree_lock, flags, SINGLE_DEPTH_NESTING);
 	llist_for_each_safe(pos, next, take_preallocated_barriers(ref)) {
 		struct active_node *node = barrier_from_ll(pos);
 		struct intel_engine_cs *engine = barrier_to_engine(node);
 		struct rb_node **p, *parent;
 
+		spin_lock_irqsave_nested(&ref->tree_lock, flags,
+					 SINGLE_DEPTH_NESTING);
 		parent = NULL;
 		p = &ref->tree.rb_node;
 		while (*p) {
@@ -693,12 +694,12 @@ void i915_active_acquire_barrier(struct i915_active *ref)
 		}
 		rb_link_node(&node->node, parent, p);
 		rb_insert_color(&node->node, &ref->tree);
+		spin_unlock_irqrestore(&ref->tree_lock, flags);
 
 		GEM_BUG_ON(!intel_engine_pm_is_awake(engine));
 		llist_add(barrier_to_ll(node), &engine->barrier_tasks);
 		intel_engine_pm_put(engine);
 	}
-	spin_unlock_irqrestore(&ref->tree_lock, flags);
 }
 
 void i915_request_add_active_barriers(struct i915_request *rq)
diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
index 9b02be0ad4e6..95e824a78d4d 100644
--- a/drivers/gpu/drm/i915/i915_pmu.c
+++ b/drivers/gpu/drm/i915/i915_pmu.c
@@ -190,7 +190,7 @@ static u64 get_rc6(struct intel_gt *gt)
 	val = 0;
 	if (intel_gt_pm_get_if_awake(gt)) {
 		val = __get_rc6(gt);
-		intel_gt_pm_put(gt);
+		intel_gt_pm_put_async(gt);
 	}
 
 	spin_lock_irqsave(&pmu->lock, flags);
@@ -360,7 +360,7 @@ engines_sample(struct intel_gt *gt, unsigned int period_ns)
 skip:
 		if (unlikely(mmio_lock))
 			spin_unlock_irqrestore(mmio_lock, flags);
-		intel_engine_pm_put(engine);
+		intel_engine_pm_put_async(engine);
 	}
 }
 
@@ -398,7 +398,7 @@ frequency_sample(struct intel_gt *gt, unsigned int period_ns)
 			if (stat)
 				val = intel_get_cagf(rps, stat);
 
-			intel_gt_pm_put(gt);
+			intel_gt_pm_put_async(gt);
 		}
 
 		add_sample_mult(&pmu->sample[__I915_SAMPLE_FREQ_ACT],
diff --git a/drivers/gpu/drm/i915/intel_wakeref.c b/drivers/gpu/drm/i915/intel_wakeref.c
index ad26d7f4ca3d..59aa1b6f1827 100644
--- a/drivers/gpu/drm/i915/intel_wakeref.c
+++ b/drivers/gpu/drm/i915/intel_wakeref.c
@@ -54,7 +54,8 @@ int __intel_wakeref_get_first(struct intel_wakeref *wf)
 
 static void ____intel_wakeref_put_last(struct intel_wakeref *wf)
 {
-	if (!atomic_dec_and_test(&wf->count))
+	INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count) <= 0);
+	if (unlikely(!atomic_dec_and_test(&wf->count)))
 		goto unlock;
 
 	/* ops->put() must reschedule its own release on error/deferral */
@@ -67,13 +68,12 @@ unlock:
 	mutex_unlock(&wf->mutex);
 }
 
-void __intel_wakeref_put_last(struct intel_wakeref *wf)
+void __intel_wakeref_put_last(struct intel_wakeref *wf, unsigned long flags)
 {
 	INTEL_WAKEREF_BUG_ON(work_pending(&wf->work));
 
 	/* Assume we are not in process context and so cannot sleep. */
-	if (wf->ops->flags & INTEL_WAKEREF_PUT_ASYNC ||
-	    !mutex_trylock(&wf->mutex)) {
+	if (flags & INTEL_WAKEREF_PUT_ASYNC || !mutex_trylock(&wf->mutex)) {
 		schedule_work(&wf->work);
 		return;
 	}
diff --git a/drivers/gpu/drm/i915/intel_wakeref.h b/drivers/gpu/drm/i915/intel_wakeref.h
index affe4de3746b..da6e8fd506e6 100644
--- a/drivers/gpu/drm/i915/intel_wakeref.h
+++ b/drivers/gpu/drm/i915/intel_wakeref.h
@@ -9,6 +9,7 @@
 
 #include <linux/atomic.h>
 #include <linux/bits.h>
+#include <linux/lockdep.h>
 #include <linux/mutex.h>
 #include <linux/refcount.h>
 #include <linux/stackdepot.h>
@@ -29,9 +30,6 @@ typedef depot_stack_handle_t intel_wakeref_t;
 struct intel_wakeref_ops {
 	int (*get)(struct intel_wakeref *wf);
 	int (*put)(struct intel_wakeref *wf);
-
-	unsigned long flags;
-#define INTEL_WAKEREF_PUT_ASYNC BIT(0)
 };
 
 struct intel_wakeref {
@@ -57,7 +55,7 @@ void __intel_wakeref_init(struct intel_wakeref *wf,
 } while (0)
 
 int __intel_wakeref_get_first(struct intel_wakeref *wf);
-void __intel_wakeref_put_last(struct intel_wakeref *wf);
+void __intel_wakeref_put_last(struct intel_wakeref *wf, unsigned long flags);
 
 /**
  * intel_wakeref_get: Acquire the wakeref
@@ -100,10 +98,9 @@ intel_wakeref_get_if_active(struct intel_wakeref *wf)
 }
 
 /**
- * intel_wakeref_put: Release the wakeref
- * @i915: the drm_i915_private device
+ * intel_wakeref_put_flags: Release the wakeref
  * @wf: the wakeref
- * @fn: callback for releasing the wakeref, called only on final release.
+ * @flags: control flags
  *
  * Release our hold on the wakeref. When there are no more users,
  * the runtime pm wakeref will be released after the @fn callback is called
@@ -116,11 +113,25 @@ intel_wakeref_get_if_active(struct intel_wakeref *wf)
  * code otherwise.
  */
 static inline void
-intel_wakeref_put(struct intel_wakeref *wf)
+__intel_wakeref_put(struct intel_wakeref *wf, unsigned long flags)
+#define INTEL_WAKEREF_PUT_ASYNC BIT(0)
 {
 	INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count) <= 0);
 	if (unlikely(!atomic_add_unless(&wf->count, -1, 1)))
-		__intel_wakeref_put_last(wf);
+		__intel_wakeref_put_last(wf, flags);
+}
+
+static inline void
+intel_wakeref_put(struct intel_wakeref *wf)
+{
+	might_sleep();
+	__intel_wakeref_put(wf, 0);
+}
+
+static inline void
+intel_wakeref_put_async(struct intel_wakeref *wf)
+{
+	__intel_wakeref_put(wf, INTEL_WAKEREF_PUT_ASYNC);
 }
 
 /**
@@ -185,6 +196,7 @@ intel_wakeref_is_active(const struct intel_wakeref *wf)
 static inline void
 __intel_wakeref_defer_park(struct intel_wakeref *wf)
 {
+	lockdep_assert_held(&wf->mutex);
 	INTEL_WAKEREF_BUG_ON(atomic_read(&wf->count));
 	atomic_set_release(&wf->count, 1);
 }