Merge tag 'drm-intel-fixes-2019-12-23' of...

Merge tag 'drm-intel-fixes-2019-12-23' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

i915 power and frontbuffer tracking fixes
Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/87r20vdlrs.fsf@intel.com

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

@@ -15112,7 +15112,7 @@ intel_prepare_plane_fb(struct drm_plane *plane,
 		return ret;
 
 	fb_obj_bump_render_priority(obj);
-	intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_DIRTYFB);
+	i915_gem_object_flush_frontbuffer(obj, ORIGIN_DIRTYFB);
 
 	if (!new_plane_state->base.fence) { /* implicit fencing */
 		struct dma_fence *fence;

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

@@ -229,11 +229,11 @@ static void frontbuffer_release(struct kref *ref)
 		vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
 	spin_unlock(&obj->vma.lock);
 
-	obj->frontbuffer = NULL;
+	RCU_INIT_POINTER(obj->frontbuffer, NULL);
 	spin_unlock(&to_i915(obj->base.dev)->fb_tracking.lock);
 
 	i915_gem_object_put(obj);
-	kfree(front);
+	kfree_rcu(front, rcu);
 }
 
 struct intel_frontbuffer *
@@ -242,11 +242,7 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
 	struct drm_i915_private *i915 = to_i915(obj->base.dev);
 	struct intel_frontbuffer *front;
 
-	spin_lock(&i915->fb_tracking.lock);
-	front = obj->frontbuffer;
-	if (front)
-		kref_get(&front->ref);
-	spin_unlock(&i915->fb_tracking.lock);
+	front = __intel_frontbuffer_get(obj);
 	if (front)
 		return front;
 
@@ -262,13 +258,13 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
 			 i915_active_may_sleep(frontbuffer_retire));
 
 	spin_lock(&i915->fb_tracking.lock);
-	if (obj->frontbuffer) {
+	if (rcu_access_pointer(obj->frontbuffer)) {
 		kfree(front);
-		front = obj->frontbuffer;
+		front = rcu_dereference_protected(obj->frontbuffer, true);
 		kref_get(&front->ref);
 	} else {
 		i915_gem_object_get(obj);
-		obj->frontbuffer = front;
+		rcu_assign_pointer(obj->frontbuffer, front);
 	}
 	spin_unlock(&i915->fb_tracking.lock);
 

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

@@ -27,10 +27,10 @@
 #include <linux/atomic.h>
 #include <linux/kref.h>
 
+#include "gem/i915_gem_object_types.h"
 #include "i915_active.h"
 
 struct drm_i915_private;
-struct drm_i915_gem_object;
 
 enum fb_op_origin {
 	ORIGIN_GTT,
@@ -45,6 +45,7 @@ struct intel_frontbuffer {
 	atomic_t bits;
 	struct i915_active write;
 	struct drm_i915_gem_object *obj;
+	struct rcu_head rcu;
 };
 
 void intel_frontbuffer_flip_prepare(struct drm_i915_private *i915,
@@ -54,6 +55,35 @@ void intel_frontbuffer_flip_complete(struct drm_i915_private *i915,
 void intel_frontbuffer_flip(struct drm_i915_private *i915,
 			    unsigned frontbuffer_bits);
 
+void intel_frontbuffer_put(struct intel_frontbuffer *front);
+
+static inline struct intel_frontbuffer *
+__intel_frontbuffer_get(const struct drm_i915_gem_object *obj)
+{
+	struct intel_frontbuffer *front;
+
+	if (likely(!rcu_access_pointer(obj->frontbuffer)))
+		return NULL;
+
+	rcu_read_lock();
+	do {
+		front = rcu_dereference(obj->frontbuffer);
+		if (!front)
+			break;
+
+		if (unlikely(!kref_get_unless_zero(&front->ref)))
+			continue;
+
+		if (likely(front == rcu_access_pointer(obj->frontbuffer)))
+			break;
+
+		intel_frontbuffer_put(front);
+	} while (1);
+	rcu_read_unlock();
+
+	return front;
+}
+
 struct intel_frontbuffer *
 intel_frontbuffer_get(struct drm_i915_gem_object *obj);
 
@@ -119,6 +149,4 @@ void intel_frontbuffer_track(struct intel_frontbuffer *old,
 			     struct intel_frontbuffer *new,
 			     unsigned int frontbuffer_bits);
 
-void intel_frontbuffer_put(struct intel_frontbuffer *front);
-
 #endif /* __INTEL_FRONTBUFFER_H__ */

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

@@ -279,12 +279,21 @@ static void intel_overlay_flip_prepare(struct intel_overlay *overlay,
 				       struct i915_vma *vma)
 {
 	enum pipe pipe = overlay->crtc->pipe;
+	struct intel_frontbuffer *from = NULL, *to = NULL;
 
 	WARN_ON(overlay->old_vma);
 
-	intel_frontbuffer_track(overlay->vma ? overlay->vma->obj->frontbuffer : NULL,
-				vma ? vma->obj->frontbuffer : NULL,
-				INTEL_FRONTBUFFER_OVERLAY(pipe));
+	if (overlay->vma)
+		from = intel_frontbuffer_get(overlay->vma->obj);
+	if (vma)
+		to = intel_frontbuffer_get(vma->obj);
+
+	intel_frontbuffer_track(from, to, INTEL_FRONTBUFFER_OVERLAY(pipe));
+
+	if (to)
+		intel_frontbuffer_put(to);
+	if (from)
+		intel_frontbuffer_put(from);
 
 	intel_frontbuffer_flip_prepare(overlay->i915,
 				       INTEL_FRONTBUFFER_OVERLAY(pipe));
@@ -766,7 +775,7 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 		ret = PTR_ERR(vma);
 		goto out_pin_section;
 	}
-	intel_frontbuffer_flush(new_bo->frontbuffer, ORIGIN_DIRTYFB);
+	i915_gem_object_flush_frontbuffer(new_bo, ORIGIN_DIRTYFB);
 
 	if (!overlay->active) {
 		u32 oconfig;

drivers/gpu/drm/i915/gem/i915_gem_clflush.c

@@ -20,7 +20,8 @@ static void __do_clflush(struct drm_i915_gem_object *obj)
 {
 	GEM_BUG_ON(!i915_gem_object_has_pages(obj));
 	drm_clflush_sg(obj->mm.pages);
-	intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_CPU);
+
+	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 }
 
 static int clflush_work(struct dma_fence_work *base)

drivers/gpu/drm/i915/gem/i915_gem_domain.c

@@ -664,7 +664,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
 	i915_gem_object_unlock(obj);
 
 	if (write_domain)
-		intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CPU);
+		i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
 
 out_unpin:
 	i915_gem_object_unpin_pages(obj);
@@ -784,7 +784,7 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
 	}
 
 out:
-	intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
 	obj->mm.dirty = true;
 	/* return with the pages pinned */
 	return 0;

drivers/gpu/drm/i915/gem/i915_gem_object.c

@@ -280,7 +280,7 @@ i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
 		for_each_ggtt_vma(vma, obj)
 			intel_gt_flush_ggtt_writes(vma->vm->gt);
 
-		intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_CPU);
+		i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 
 		for_each_ggtt_vma(vma, obj) {
 			if (vma->iomap)
@@ -308,6 +308,30 @@ i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
 	obj->write_domain = 0;
 }
 
+void __i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
+					 enum fb_op_origin origin)
+{
+	struct intel_frontbuffer *front;
+
+	front = __intel_frontbuffer_get(obj);
+	if (front) {
+		intel_frontbuffer_flush(front, origin);
+		intel_frontbuffer_put(front);
+	}
+}
+
+void __i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
+					      enum fb_op_origin origin)
+{
+	struct intel_frontbuffer *front;
+
+	front = __intel_frontbuffer_get(obj);
+	if (front) {
+		intel_frontbuffer_invalidate(front, origin);
+		intel_frontbuffer_put(front);
+	}
+}
+
 void i915_gem_init__objects(struct drm_i915_private *i915)
 {
 	INIT_WORK(&i915->mm.free_work, __i915_gem_free_work);

drivers/gpu/drm/i915/gem/i915_gem_object.h

@@ -13,8 +13,8 @@
 
 #include <drm/i915_drm.h>
 
+#include "display/intel_frontbuffer.h"
 #include "i915_gem_object_types.h"
-
 #include "i915_gem_gtt.h"
 
 void i915_gem_init__objects(struct drm_i915_private *i915);
@@ -463,4 +463,25 @@ int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
 				  unsigned int flags,
 				  const struct i915_sched_attr *attr);
 
+void __i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
+					 enum fb_op_origin origin);
+void __i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
+					      enum fb_op_origin origin);
+
+static inline void
+i915_gem_object_flush_frontbuffer(struct drm_i915_gem_object *obj,
+				  enum fb_op_origin origin)
+{
+	if (unlikely(rcu_access_pointer(obj->frontbuffer)))
+		__i915_gem_object_flush_frontbuffer(obj, origin);
+}
+
+static inline void
+i915_gem_object_invalidate_frontbuffer(struct drm_i915_gem_object *obj,
+				       enum fb_op_origin origin)
+{
+	if (unlikely(rcu_access_pointer(obj->frontbuffer)))
+		__i915_gem_object_invalidate_frontbuffer(obj, origin);
+}
+
 #endif

drivers/gpu/drm/i915/gem/i915_gem_object_types.h

@@ -150,7 +150,7 @@ struct drm_i915_gem_object {
 	 */
 	u16 write_domain;
 
-	struct intel_frontbuffer *frontbuffer;
+	struct intel_frontbuffer __rcu *frontbuffer;
 
 	/** Current tiling stride for the object, if it's tiled. */
 	unsigned int tiling_and_stride;

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

@@ -94,8 +94,9 @@ static int __gt_park(struct intel_wakeref *wf)
 		intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL);
 	}
 
+	/* Defer dropping the display power well for 100ms, it's slow! */
 	GEM_BUG_ON(!wakeref);
-	intel_display_power_put(i915, POWER_DOMAIN_GT_IRQ, wakeref);
+	intel_display_power_put_async(i915, POWER_DOMAIN_GT_IRQ, wakeref);
 
 	i915_globals_park();
 

drivers/gpu/drm/i915/i915_gem.c

@@ -161,7 +161,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,
 	 * We manually control the domain here and pretend that it
 	 * remains coherent i.e. in the GTT domain, like shmem_pwrite.
 	 */
-	intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
 
 	if (copy_from_user(vaddr, user_data, args->size))
 		return -EFAULT;
@@ -169,7 +169,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,
 	drm_clflush_virt_range(vaddr, args->size);
 	intel_gt_chipset_flush(&to_i915(obj->base.dev)->gt);
 
-	intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 	return 0;
 }
 
@@ -589,7 +589,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 		goto out_unpin;
 	}
 
-	intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
 
 	user_data = u64_to_user_ptr(args->data_ptr);
 	offset = args->offset;
@@ -631,7 +631,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 		user_data += page_length;
 		offset += page_length;
 	}
-	intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 
 	i915_gem_object_unlock_fence(obj, fence);
 out_unpin:
@@ -721,7 +721,7 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
 		offset = 0;
 	}
 
-	intel_frontbuffer_flush(obj->frontbuffer, ORIGIN_CPU);
+	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 	i915_gem_object_unlock_fence(obj, fence);
 
 	return ret;

drivers/gpu/drm/i915/i915_pmu.c

@@ -144,61 +144,40 @@ static inline s64 ktime_since(const ktime_t kt)
 	return ktime_to_ns(ktime_sub(ktime_get(), kt));
 }
 
-static u64 __pmu_estimate_rc6(struct i915_pmu *pmu)
-{
-	u64 val;
-
-	/*
-	 * We think we are runtime suspended.
-	 *
-	 * Report the delta from when the device was suspended to now,
-	 * on top of the last known real value, as the approximated RC6
-	 * counter value.
-	 */
-	val = ktime_since(pmu->sleep_last);
-	val += pmu->sample[__I915_SAMPLE_RC6].cur;
-
-	pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = val;
-
-	return val;
-}
-
-static u64 __pmu_update_rc6(struct i915_pmu *pmu, u64 val)
-{
-	/*
-	 * If we are coming back from being runtime suspended we must
-	 * be careful not to report a larger value than returned
-	 * previously.
-	 */
-	if (val >= pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur) {
-		pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur = 0;
-		pmu->sample[__I915_SAMPLE_RC6].cur = val;
-	} else {
-		val = pmu->sample[__I915_SAMPLE_RC6_ESTIMATED].cur;
-	}
-
-	return val;
-}
-
 static u64 get_rc6(struct intel_gt *gt)
 {
 	struct drm_i915_private *i915 = gt->i915;
 	struct i915_pmu *pmu = &i915->pmu;
 	unsigned long flags;
+	bool awake = false;
 	u64 val;
 
-	val = 0;
 	if (intel_gt_pm_get_if_awake(gt)) {
 		val = __get_rc6(gt);
 		intel_gt_pm_put_async(gt);
+		awake = true;
 	}
 
 	spin_lock_irqsave(&pmu->lock, flags);
 
-	if (val)
-		val = __pmu_update_rc6(pmu, val);
+	if (awake) {
+		pmu->sample[__I915_SAMPLE_RC6].cur = val;
+	} else {
+		/*
+		 * We think we are runtime suspended.
+		 *
+		 * Report the delta from when the device was suspended to now,
+		 * on top of the last known real value, as the approximated RC6
+		 * counter value.
+		 */
+		val = ktime_since(pmu->sleep_last);
+		val += pmu->sample[__I915_SAMPLE_RC6].cur;
+	}
+
+	if (val < pmu->sample[__I915_SAMPLE_RC6_LAST_REPORTED].cur)
+		val = pmu->sample[__I915_SAMPLE_RC6_LAST_REPORTED].cur;
 	else
-		val = __pmu_estimate_rc6(pmu);
+		pmu->sample[__I915_SAMPLE_RC6_LAST_REPORTED].cur = val;
 
 	spin_unlock_irqrestore(&pmu->lock, flags);
 
@@ -210,20 +189,11 @@ static void park_rc6(struct drm_i915_private *i915)
 	struct i915_pmu *pmu = &i915->pmu;
 
 	if (pmu->enable & config_enabled_mask(I915_PMU_RC6_RESIDENCY))
-		__pmu_update_rc6(pmu, __get_rc6(&i915->gt));
+		pmu->sample[__I915_SAMPLE_RC6].cur = __get_rc6(&i915->gt);
 
 	pmu->sleep_last = ktime_get();
 }
 
-static void unpark_rc6(struct drm_i915_private *i915)
-{
-	struct i915_pmu *pmu = &i915->pmu;
-
-	/* Estimate how long we slept and accumulate that into rc6 counters */
-	if (pmu->enable & config_enabled_mask(I915_PMU_RC6_RESIDENCY))
-		__pmu_estimate_rc6(pmu);
-}
-
 #else
 
 static u64 get_rc6(struct intel_gt *gt)
@@ -232,7 +202,6 @@ static u64 get_rc6(struct intel_gt *gt)
 }
 
 static void park_rc6(struct drm_i915_private *i915) {}
-static void unpark_rc6(struct drm_i915_private *i915) {}
 
 #endif
 
@@ -281,8 +250,6 @@ void i915_pmu_gt_unparked(struct drm_i915_private *i915)
 	 */
 	__i915_pmu_maybe_start_timer(pmu);
 
-	unpark_rc6(i915);
-
 	spin_unlock_irq(&pmu->lock);
 }
 

drivers/gpu/drm/i915/i915_pmu.h

@@ -18,7 +18,7 @@ enum {
 	__I915_SAMPLE_FREQ_ACT = 0,
 	__I915_SAMPLE_FREQ_REQ,
 	__I915_SAMPLE_RC6,
-	__I915_SAMPLE_RC6_ESTIMATED,
+	__I915_SAMPLE_RC6_LAST_REPORTED,
 	__I915_NUM_PMU_SAMPLERS
 };
 

drivers/gpu/drm/i915/i915_vma.c

@@ -1104,8 +1104,14 @@ int i915_vma_move_to_active(struct i915_vma *vma,
 		return err;
 
 	if (flags & EXEC_OBJECT_WRITE) {
-		if (intel_frontbuffer_invalidate(obj->frontbuffer, ORIGIN_CS))
-			i915_active_add_request(&obj->frontbuffer->write, rq);
+		struct intel_frontbuffer *front;
+
+		front = __intel_frontbuffer_get(obj);
+		if (unlikely(front)) {
+			if (intel_frontbuffer_invalidate(front, ORIGIN_CS))
+				i915_active_add_request(&front->write, rq);
+			intel_frontbuffer_put(front);
+		}
 
 		dma_resv_add_excl_fence(vma->resv, &rq->fence);
 		obj->write_domain = I915_GEM_DOMAIN_RENDER;