drm/i915: Push the i915_active.retire into a worker

As we need to use a mutex to serialise i915_active activation
(because we want to allow the callback to sleep), we need to push the
i915_active.retire into a worker callback in case we get need to retire
from an atomic context.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-5-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/display/intel_frontbuffer.c

@@ -206,6 +206,7 @@ static int frontbuffer_active(struct i915_active *ref)
 	return 0;
 }
 
+__i915_active_call
 static void frontbuffer_retire(struct i915_active *ref)
 {
 	struct intel_frontbuffer *front =
@@ -257,7 +258,8 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
 	kref_init(&front->ref);
 	atomic_set(&front->bits, 0);
 	i915_active_init(i915, &front->write,
-			 frontbuffer_active, frontbuffer_retire);
+			 frontbuffer_active,
+			 i915_active_may_sleep(frontbuffer_retire));
 
 	spin_lock(&i915->fb_tracking.lock);
 	if (obj->frontbuffer) {

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

@@ -848,6 +848,7 @@ struct context_barrier_task {
 	void *data;
 };
 
+__i915_active_call
 static void cb_retire(struct i915_active *base)
 {
 	struct context_barrier_task *cb = container_of(base, typeof(*cb), base);

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

@@ -138,6 +138,7 @@ static void __context_unpin_state(struct i915_vma *vma)
 	__i915_vma_unpin(vma);
 }
 
+__i915_active_call
 static void __intel_context_retire(struct i915_active *active)
 {
 	struct intel_context *ce = container_of(active, typeof(*ce), active);
@@ -150,6 +151,7 @@ static void __intel_context_retire(struct i915_active *active)
 
 	intel_timeline_unpin(ce->timeline);
 	intel_ring_unpin(ce->ring);
+
 	intel_context_put(ce);
 }
 

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

@@ -61,6 +61,7 @@ static int pool_active(struct i915_active *ref)
 	return 0;
 }
 
+__i915_active_call
 static void pool_retire(struct i915_active *ref)
 {
 	struct intel_engine_pool_node *node =

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

@@ -136,6 +136,7 @@ static void __idle_cacheline_free(struct intel_timeline_cacheline *cl)
 	kfree(cl);
 }
 
+__i915_active_call
 static void __cacheline_retire(struct i915_active *active)
 {
 	struct intel_timeline_cacheline *cl =

drivers/gpu/drm/i915/i915_active.c

@@ -132,6 +132,7 @@ __active_retire(struct i915_active *ref)
 	bool retire = false;
 
 	lockdep_assert_held(&ref->mutex);
+	GEM_BUG_ON(i915_active_is_idle(ref));
 
 	/* return the unused nodes to our slabcache -- flushing the allocator */
 	if (atomic_dec_and_test(&ref->count)) {
@@ -157,15 +158,33 @@ __active_retire(struct i915_active *ref)
 		ref->retire(ref);
 }
 
+static void
+active_work(struct work_struct *wrk)
+{
+	struct i915_active *ref = container_of(wrk, typeof(*ref), work);
+
+	GEM_BUG_ON(!atomic_read(&ref->count));
+	if (atomic_add_unless(&ref->count, -1, 1))
+		return;
+
+	mutex_lock(&ref->mutex);
+	__active_retire(ref);
+}
+
 static void
 active_retire(struct i915_active *ref)
 {
 	GEM_BUG_ON(!atomic_read(&ref->count));
 	if (atomic_add_unless(&ref->count, -1, 1))
 		return;
 
-	/* One active may be flushed from inside the acquire of another */
-	mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
+	/* If we are inside interrupt context (fence signaling), defer */
+	if (ref->flags & I915_ACTIVE_RETIRE_SLEEPS ||
+	    !mutex_trylock(&ref->mutex)) {
+		queue_work(system_unbound_wq, &ref->work);
+		return;
+	}
+
 	__active_retire(ref);
 }
 
@@ -240,19 +259,24 @@ void __i915_active_init(struct drm_i915_private *i915,
 			void (*retire)(struct i915_active *ref),
 			struct lock_class_key *key)
 {
+	unsigned long bits;
+
 	debug_active_init(ref);
 
 	ref->i915 = i915;
 	ref->flags = 0;
 	ref->active = active;
-	ref->retire = retire;
+	ref->retire = ptr_unpack_bits(retire, &bits, 2);
+	if (bits & I915_ACTIVE_MAY_SLEEP)
+		ref->flags |= I915_ACTIVE_RETIRE_SLEEPS;
 
 	ref->excl = NULL;
 	ref->tree = RB_ROOT;
 	ref->cache = NULL;
 	init_llist_head(&ref->preallocated_barriers);
 	atomic_set(&ref->count, 0);
 	__mutex_init(&ref->mutex, "i915_active", key);
+	INIT_WORK(&ref->work, active_work);
 }
 
 static bool ____active_del_barrier(struct i915_active *ref,
@@ -504,6 +528,7 @@ int i915_active_wait(struct i915_active *ref)
 	if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE))
 		return -EINTR;
 
+	flush_work(&ref->work);
 	if (!i915_active_is_idle(ref))
 		return -EBUSY;
 
@@ -544,8 +569,9 @@ int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
 void i915_active_fini(struct i915_active *ref)
 {
 	debug_active_fini(ref);
-	GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
 	GEM_BUG_ON(atomic_read(&ref->count));
+	GEM_BUG_ON(work_pending(&ref->work));
+	GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
 	mutex_destroy(&ref->mutex);
 }
 #endif

drivers/gpu/drm/i915/i915_active_types.h

@@ -13,6 +13,9 @@
 #include <linux/mutex.h>
 #include <linux/rbtree.h>
 #include <linux/rcupdate.h>
+#include <linux/workqueue.h>
+
+#include "i915_utils.h"
 
 struct drm_i915_private;
 struct i915_active_request;
@@ -44,6 +47,11 @@ struct i915_active_request {
 
 struct active_node;
 
+#define I915_ACTIVE_MAY_SLEEP BIT(0)
+
+#define __i915_active_call __aligned(4)
+#define i915_active_may_sleep(fn) ptr_pack_bits(&(fn), I915_ACTIVE_MAY_SLEEP, 2)
+
 struct i915_active {
 	struct drm_i915_private *i915;
 
@@ -57,11 +65,14 @@ struct i915_active {
 	struct dma_fence_cb excl_cb;
 
 	unsigned long flags;
-#define I915_ACTIVE_GRAB_BIT 0
+#define I915_ACTIVE_RETIRE_SLEEPS BIT(0)
+#define I915_ACTIVE_GRAB_BIT 1
 
 	int (*active)(struct i915_active *ref);
 	void (*retire)(struct i915_active *ref);
 
+	struct work_struct work;
+
 	struct llist_head preallocated_barriers;
 };
 

drivers/gpu/drm/i915/i915_vma.c

@@ -91,6 +91,7 @@ static int __i915_vma_active(struct i915_active *ref)
 	return i915_vma_tryget(active_to_vma(ref)) ? 0 : -ENOENT;
 }
 
+__i915_active_call
 static void __i915_vma_retire(struct i915_active *ref)
 {
 	i915_vma_put(active_to_vma(ref));
@@ -1152,6 +1153,7 @@ int __i915_vma_unbind(struct i915_vma *vma)
 		return -EBUSY;
 	}
 
+	GEM_BUG_ON(i915_vma_is_active(vma));
 	if (!drm_mm_node_allocated(&vma->node))
 		return 0;
 

drivers/gpu/drm/i915/selftests/i915_active.c

@@ -121,7 +121,7 @@ __live_active_setup(struct drm_i915_private *i915)
 	}
 
 	i915_active_release(&active->base);
-	if (active->retired && count) {
+	if (READ_ONCE(active->retired) && count) {
 		pr_err("i915_active retired before submission!\n");
 		err = -EINVAL;
 	}
@@ -161,7 +161,7 @@ static int live_active_wait(void *arg)
 	}
 
 	i915_active_wait(&active->base);
-	if (!active->retired) {
+	if (!READ_ONCE(active->retired)) {
 		pr_err("i915_active not retired after waiting!\n");
 		err = -EINVAL;
 	}
@@ -200,7 +200,7 @@ static int live_active_retire(void *arg)
 	if (igt_flush_test(i915, I915_WAIT_LOCKED))
 		err = -EIO;
 
-	if (!active->retired) {
+	if (!READ_ONCE(active->retired)) {
 		pr_err("i915_active not retired after flushing!\n");
 		err = -EINVAL;
 	}