1 files changed, 119 insertions, 51 deletions

diff --git a/drivers/gpu/drm/radeon/radeon_semaphore.c b/drivers/gpu/drm/radeon/radeon_semaphore.c
index e2ace5dce11..dbd6bcde92d 100644
--- a/drivers/gpu/drm/radeon/radeon_semaphore.c
+++ b/drivers/gpu/drm/radeon/radeon_semaphore.c
@@ -27,22 +27,22 @@
  * Authors:
  *    Christian König <deathsimple@vodafone.de>
  */
-#include "drmP.h"
-#include "drm.h"
+#include <drm/drmP.h>
 #include "radeon.h"
-
+#include "radeon_trace.h"
 
 int radeon_semaphore_create(struct radeon_device *rdev,
 			    struct radeon_semaphore **semaphore)
 {
-	int r;
+	uint32_t *cpu_addr;
+	int i, r;
 
 	*semaphore = kmalloc(sizeof(struct radeon_semaphore), GFP_KERNEL);
 	if (*semaphore == NULL) {
 		return -ENOMEM;
 	}
-	r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo,
-			     &(*semaphore)->sa_bo, 8, 8, true);
+	r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, &(*semaphore)->sa_bo,
+			     8 * RADEON_NUM_SYNCS, 8);
 	if (r) {
 		kfree(*semaphore);
 		*semaphore = NULL;
@@ -50,88 +50,156 @@ int radeon_semaphore_create(struct radeon_device *rdev,
 	}
 	(*semaphore)->waiters = 0;
 	(*semaphore)->gpu_addr = radeon_sa_bo_gpu_addr((*semaphore)->sa_bo);
-	*((uint64_t*)radeon_sa_bo_cpu_addr((*semaphore)->sa_bo)) = 0;
+
+	cpu_addr = radeon_sa_bo_cpu_addr((*semaphore)->sa_bo);
+	for (i = 0; i < RADEON_NUM_SYNCS; ++i)
+		cpu_addr[i] = 0;
+
+	for (i = 0; i < RADEON_NUM_RINGS; ++i)
+		(*semaphore)->sync_to[i] = NULL;
+
 	return 0;
 }
 
-void radeon_semaphore_emit_signal(struct radeon_device *rdev, int ring,
+bool radeon_semaphore_emit_signal(struct radeon_device *rdev, int ridx,
 			          struct radeon_semaphore *semaphore)
 {
-	--semaphore->waiters;
-	radeon_semaphore_ring_emit(rdev, ring, &rdev->ring[ring], semaphore, false);
+	struct radeon_ring *ring = &rdev->ring[ridx];
+
+	trace_radeon_semaphore_signale(ridx, semaphore);
+
+	if (radeon_semaphore_ring_emit(rdev, ridx, ring, semaphore, false)) {
+		--semaphore->waiters;
+
+		/* for debugging lockup only, used by sysfs debug files */
+		ring->last_semaphore_signal_addr = semaphore->gpu_addr;
+		return true;
+	}
+	return false;
 }
 
-void radeon_semaphore_emit_wait(struct radeon_device *rdev, int ring,
+bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ridx,
 			        struct radeon_semaphore *semaphore)
 {
-	++semaphore->waiters;
-	radeon_semaphore_ring_emit(rdev, ring, &rdev->ring[ring], semaphore, true);
+	struct radeon_ring *ring = &rdev->ring[ridx];
+
+	trace_radeon_semaphore_wait(ridx, semaphore);
+
+	if (radeon_semaphore_ring_emit(rdev, ridx, ring, semaphore, true)) {
+		++semaphore->waiters;
+
+		/* for debugging lockup only, used by sysfs debug files */
+		ring->last_semaphore_wait_addr = semaphore->gpu_addr;
+		return true;
+	}
+	return false;
+}
+
+/**
+ * radeon_semaphore_sync_to - use the semaphore to sync to a fence
+ *
+ * @semaphore: semaphore object to add fence to
+ * @fence: fence to sync to
+ *
+ * Sync to the fence using this semaphore object
+ */
+void radeon_semaphore_sync_to(struct radeon_semaphore *semaphore,
+			      struct radeon_fence *fence)
+{
+        struct radeon_fence *other;
+
+        if (!fence)
+                return;
+
+        other = semaphore->sync_to[fence->ring];
+        semaphore->sync_to[fence->ring] = radeon_fence_later(fence, other);
 }
 
+/**
+ * radeon_semaphore_sync_rings - sync ring to all registered fences
+ *
+ * @rdev: radeon_device pointer
+ * @semaphore: semaphore object to use for sync
+ * @ring: ring that needs sync
+ *
+ * Ensure that all registered fences are signaled before letting
+ * the ring continue. The caller must hold the ring lock.
+ */
 int radeon_semaphore_sync_rings(struct radeon_device *rdev,
 				struct radeon_semaphore *semaphore,
-				bool sync_to[RADEON_NUM_RINGS],
-				int dst_ring)
+				int ring)
 {
-	int i = 0, r;
+	unsigned count = 0;
+	int i, r;
 
-	mutex_lock(&rdev->ring_lock);
-	r = radeon_ring_alloc(rdev, &rdev->ring[dst_ring], RADEON_NUM_RINGS * 8);
-	if (r) {
-		goto error;
-	}
+        for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+		struct radeon_fence *fence = semaphore->sync_to[i];
 
-	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
-		/* no need to sync to our own or unused rings */
-		if (!sync_to[i] || i == dst_ring)
+		/* check if we really need to sync */
+                if (!radeon_fence_need_sync(fence, ring))
 			continue;
 
 		/* prevent GPU deadlocks */
 		if (!rdev->ring[i].ready) {
-			dev_err(rdev->dev, "Trying to sync to a disabled ring!");
-			r = -EINVAL;
-			goto error;
+			dev_err(rdev->dev, "Syncing to a disabled ring!");
+			return -EINVAL;
+		}
+
+		if (++count > RADEON_NUM_SYNCS) {
+			/* not enough room, wait manually */
+			r = radeon_fence_wait(fence, false);
+			if (r)
+				return r;
+			continue;
 		}
 
-		r = radeon_ring_alloc(rdev, &rdev->ring[i], 8);
+		/* allocate enough space for sync command */
+		r = radeon_ring_alloc(rdev, &rdev->ring[i], 16);
 		if (r) {
-			goto error;
+			return r;
+		}
+
+		/* emit the signal semaphore */
+		if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) {
+			/* signaling wasn't successful wait manually */
+			radeon_ring_undo(&rdev->ring[i]);
+			r = radeon_fence_wait(fence, false);
+			if (r)
+				return r;
+			continue;
 		}
 
-		radeon_semaphore_emit_signal(rdev, i, semaphore);
-		radeon_semaphore_emit_wait(rdev, dst_ring, semaphore);
+		/* we assume caller has already allocated space on waiters ring */
+		if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) {
+			/* waiting wasn't successful wait manually */
+			radeon_ring_undo(&rdev->ring[i]);
+			r = radeon_fence_wait(fence, false);
+			if (r)
+				return r;
+			continue;
+		}
 
 		radeon_ring_commit(rdev, &rdev->ring[i]);
-	}
+		radeon_fence_note_sync(fence, ring);
 
-	radeon_ring_commit(rdev, &rdev->ring[dst_ring]);
-	mutex_unlock(&rdev->ring_lock);
+		semaphore->gpu_addr += 8;
+	}
 
 	return 0;
-
-error:
-	/* unlock all locks taken so far */
-	for (--i; i >= 0; --i) {
-		if (sync_to[i] || i == dst_ring) {
-			radeon_ring_undo(&rdev->ring[i]);
-		}
-	}
-	radeon_ring_undo(&rdev->ring[dst_ring]);
-	mutex_unlock(&rdev->ring_lock);
-	return r;
 }
 
 void radeon_semaphore_free(struct radeon_device *rdev,
-			   struct radeon_semaphore *semaphore,
+			   struct radeon_semaphore **semaphore,
 			   struct radeon_fence *fence)
 {
-	if (semaphore == NULL) {
+	if (semaphore == NULL || *semaphore == NULL) {
 		return;
 	}
-	if (semaphore->waiters > 0) {
+	if ((*semaphore)->waiters > 0) {
 		dev_err(rdev->dev, "semaphore %p has more waiters than signalers,"
-			" hardware lockup imminent!\n", semaphore);
+			" hardware lockup imminent!\n", *semaphore);
 	}
-	radeon_sa_bo_free(rdev, &semaphore->sa_bo, fence);
-	kfree(semaphore);
+	radeon_sa_bo_free(rdev, &(*semaphore)->sa_bo, fence);
+	kfree(*semaphore);
+	*semaphore = NULL;
 }