drm/radeon/kms: add support for multiple fence queues v2

For supporting multiple CP ring buffers, async DMA
engines and UVD.  We still need a way to synchronize
between engines.

v2 initialize unused fence driver ring to avoid issue in
   suspend/unload

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

20 files changed, 201 insertions, 150 deletions

diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
index 5e00d1670aa..1c7af565433 100644
--- a/drivers/gpu/drm/radeon/evergreen.c
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -3018,11 +3018,11 @@ restart_ih:
 		case 177: /* CP_INT in IB1 */
 		case 178: /* CP_INT in IB2 */
 			DRM_DEBUG("IH: CP int: 0x%08x\n", src_data);
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;
 		case 181: /* CP EOP event */
 			DRM_DEBUG("IH: CP EOP\n");
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;
 		case 233: /* GUI IDLE */
 			DRM_DEBUG("IH: GUI idle\n");
@@ -3221,7 +3221,7 @@ int evergreen_init(struct radeon_device *rdev)
 	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	/* initialize AGP */
diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c
index 0e579985746..710ad8c8276 100644
--- a/drivers/gpu/drm/radeon/ni.c
+++ b/drivers/gpu/drm/radeon/ni.c
@@ -1484,7 +1484,7 @@ int cayman_init(struct radeon_device *rdev)
 	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 3);
 	if (r)
 		return r;
 	/* initialize memory controller */
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index bfc08f6320f..d2dced5679b 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -739,7 +739,7 @@ int r100_irq_process(struct radeon_device *rdev)
 	while (status) {
 		/* SW interrupt */
 		if (status & RADEON_SW_INT_TEST) {
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 		}
 		/* gui idle interrupt */
 		if (status & RADEON_GUI_IDLE_STAT) {
@@ -826,7 +826,7 @@ void r100_fence_ring_emit(struct radeon_device *rdev,
 	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
 	radeon_ring_write(rdev, rdev->config.r100.hdp_cntl);
 	/* Emit fence sequence & fire IRQ */
-	radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
+	radeon_ring_write(rdev, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
 	radeon_ring_write(rdev, fence->seq);
 	radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
 	radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
@@ -4048,7 +4048,7 @@ int r100_init(struct radeon_device *rdev)
 	/* initialize VRAM */
 	r100_mc_init(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index c93bc64707e..b0473120646 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -198,7 +198,7 @@ void r300_fence_ring_emit(struct radeon_device *rdev,
 	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
 	radeon_ring_write(rdev, rdev->config.r300.hdp_cntl);
 	/* Emit fence sequence & fire IRQ */
-	radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
+	radeon_ring_write(rdev, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
 	radeon_ring_write(rdev, fence->seq);
 	radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
 	radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
@@ -1518,7 +1518,7 @@ int r300_init(struct radeon_device *rdev)
 	/* initialize memory controller */
 	r300_mc_init(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c
index 417fab81812..5dbc378d3c2 100644
--- a/drivers/gpu/drm/radeon/r420.c
+++ b/drivers/gpu/drm/radeon/r420.c
@@ -387,7 +387,7 @@ int r420_init(struct radeon_device *rdev)
 	r300_mc_init(rdev);
 	r420_debugfs(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r) {
 		return r;
 	}
diff --git a/drivers/gpu/drm/radeon/r520.c b/drivers/gpu/drm/radeon/r520.c
index 3081d07f8de..cb96a51f7f7 100644
--- a/drivers/gpu/drm/radeon/r520.c
+++ b/drivers/gpu/drm/radeon/r520.c
@@ -278,7 +278,7 @@ int r520_init(struct radeon_device *rdev)
 	r520_mc_init(rdev);
 	rv515_debugfs(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 9cdda0b3b08..2fff8cec723 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -2316,7 +2316,7 @@ void r600_fence_ring_emit(struct radeon_device *rdev,
 {
 	if (rdev->wb.use_event) {
 		u64 addr = rdev->wb.gpu_addr + R600_WB_EVENT_OFFSET +
-			(u64)(rdev->fence_drv.scratch_reg - rdev->scratch.reg_base);
+			(u64)(rdev->fence_drv[fence->ring].scratch_reg - rdev->scratch.reg_base);
 		/* flush read cache over gart */
 		radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
 		radeon_ring_write(rdev, PACKET3_TC_ACTION_ENA |
@@ -2349,7 +2349,7 @@ void r600_fence_ring_emit(struct radeon_device *rdev,
 		radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
 		/* Emit fence sequence & fire IRQ */
 		radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
-		radeon_ring_write(rdev, ((rdev->fence_drv.scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
+		radeon_ring_write(rdev, ((rdev->fence_drv[fence->ring].scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
 		radeon_ring_write(rdev, fence->seq);
 		/* CP_INTERRUPT packet 3 no longer exists, use packet 0 */
 		radeon_ring_write(rdev, PACKET0(CP_INT_STATUS, 0));
@@ -2575,7 +2575,7 @@ int r600_init(struct radeon_device *rdev)
 	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	if (rdev->flags & RADEON_IS_AGP) {
@@ -3459,11 +3459,11 @@ restart_ih:
 		case 177: /* CP_INT in IB1 */
 		case 178: /* CP_INT in IB2 */
 			DRM_DEBUG("IH: CP int: 0x%08x\n", src_data);
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;
 		case 181: /* CP EOP event */
 			DRM_DEBUG("IH: CP EOP\n");
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;
 		case 233: /* GUI IDLE */
 			DRM_DEBUG("IH: GUI idle\n");
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 33f3be369a2..8b93dec66ec 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -197,7 +197,6 @@ struct radeon_fence_driver {
 	unsigned long			last_jiffies;
 	unsigned long			last_timeout;
 	wait_queue_head_t		queue;
-	rwlock_t			lock;
 	struct list_head		created;
 	struct list_head		emitted;
 	struct list_head		signaled;
@@ -212,17 +211,19 @@ struct radeon_fence {
 	uint32_t			seq;
 	bool				emitted;
 	bool				signaled;
+	/* RB, DMA, etc. */
+	int				ring;
 };
 
-int radeon_fence_driver_init(struct radeon_device *rdev);
+int radeon_fence_driver_init(struct radeon_device *rdev, int num_rings);
 void radeon_fence_driver_fini(struct radeon_device *rdev);
-int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence);
+int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence, int ring);
 int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence *fence);
-void radeon_fence_process(struct radeon_device *rdev);
+void radeon_fence_process(struct radeon_device *rdev, int ring);
 bool radeon_fence_signaled(struct radeon_fence *fence);
 int radeon_fence_wait(struct radeon_fence *fence, bool interruptible);
-int radeon_fence_wait_next(struct radeon_device *rdev);
-int radeon_fence_wait_last(struct radeon_device *rdev);
+int radeon_fence_wait_next(struct radeon_device *rdev, int ring);
+int radeon_fence_wait_last(struct radeon_device *rdev, int ring);
 struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence);
 void radeon_fence_unref(struct radeon_fence **fence);
 
@@ -459,6 +460,18 @@ void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc);
 /*
  * CP & ring.
  */
+
+/* max number of rings */
+#define RADEON_NUM_RINGS 3
+
+/* internal ring indices */
+/* r1xx+ has gfx CP ring */
+#define RADEON_RING_TYPE_GFX_INDEX  0
+
+/* cayman has 2 compute CP rings */
+#define CAYMAN_RING_TYPE_CP1_INDEX 1
+#define CAYMAN_RING_TYPE_CP2_INDEX 2
+
 struct radeon_ib {
 	struct list_head	list;
 	unsigned		idx;
@@ -1235,7 +1248,8 @@ struct radeon_device {
 	struct radeon_mode_info		mode_info;
 	struct radeon_scratch		scratch;
 	struct radeon_mman		mman;
-	struct radeon_fence_driver	fence_drv;
+	rwlock_t			fence_lock;
+	struct radeon_fence_driver	fence_drv[RADEON_NUM_RINGS];
 	struct radeon_cp		cp;
 	/* cayman compute rings */
 	struct radeon_cp		cp1;
diff --git a/drivers/gpu/drm/radeon/radeon_benchmark.c b/drivers/gpu/drm/radeon/radeon_benchmark.c
index 17e1a9b2d8f..4f749a69c9c 100644
--- a/drivers/gpu/drm/radeon/radeon_benchmark.c
+++ b/drivers/gpu/drm/radeon/radeon_benchmark.c
@@ -43,7 +43,7 @@ static int radeon_benchmark_do_move(struct radeon_device *rdev, unsigned size,
 
 	start_jiffies = jiffies;
 	for (i = 0; i < n; i++) {
-		r = radeon_fence_create(rdev, &fence);
+		r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 		if (r)
 			return r;
 
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index 4d66d686857..4ed4eeb62cd 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -725,7 +725,7 @@ int radeon_device_init(struct radeon_device *rdev,
 	mutex_init(&rdev->gem.mutex);
 	mutex_init(&rdev->pm.mutex);
 	mutex_init(&rdev->vram_mutex);
-	rwlock_init(&rdev->fence_drv.lock);
+	rwlock_init(&rdev->fence_lock);
 	INIT_LIST_HEAD(&rdev->gem.objects);
 	init_waitqueue_head(&rdev->irq.vblank_queue);
 	init_waitqueue_head(&rdev->irq.idle_queue);
@@ -857,7 +857,7 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 	struct radeon_device *rdev;
 	struct drm_crtc *crtc;
 	struct drm_connector *connector;
-	int r;
+	int i, r;
 
 	if (dev == NULL || dev->dev_private == NULL) {
 		return -ENODEV;
@@ -896,7 +896,8 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 	/* evict vram memory */
 	radeon_bo_evict_vram(rdev);
 	/* wait for gpu to finish processing current batch */
-	radeon_fence_wait_last(rdev);
+	for (i = 0; i < RADEON_NUM_RINGS; i++)
+		radeon_fence_wait_last(rdev, i);
 
 	radeon_save_bios_scratch_regs(rdev);
 
diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c
index 7027766ec2a..086b8a39911 100644
--- a/drivers/gpu/drm/radeon/radeon_fence.c
+++ b/drivers/gpu/drm/radeon/radeon_fence.c
@@ -40,32 +40,37 @@
 #include "radeon.h"
 #include "radeon_trace.h"
 
-static void radeon_fence_write(struct radeon_device *rdev, u32 seq)
+static void radeon_fence_write(struct radeon_device *rdev, u32 seq, int ring)
 {
+	u32 scratch_index;
+
 	if (rdev->wb.enabled) {
-		u32 scratch_index;
 		if (rdev->wb.use_event)
-			scratch_index = R600_WB_EVENT_OFFSET + rdev->fence_drv.scratch_reg - rdev->scratch.reg_base;
+			scratch_index = R600_WB_EVENT_OFFSET +
+				rdev->fence_drv[ring].scratch_reg - rdev->scratch.reg_base;
 		else
-			scratch_index = RADEON_WB_SCRATCH_OFFSET + rdev->fence_drv.scratch_reg - rdev->scratch.reg_base;
+			scratch_index = RADEON_WB_SCRATCH_OFFSET +
+				rdev->fence_drv[ring].scratch_reg - rdev->scratch.reg_base;
 		rdev->wb.wb[scratch_index/4] = cpu_to_le32(seq);
 	} else
-		WREG32(rdev->fence_drv.scratch_reg, seq);
+		WREG32(rdev->fence_drv[ring].scratch_reg, seq);
 }
 
-static u32 radeon_fence_read(struct radeon_device *rdev)
+static u32 radeon_fence_read(struct radeon_device *rdev, int ring)
 {
-	u32 seq;
+	u32 seq = 0;
+	u32 scratch_index;
 
 	if (rdev->wb.enabled) {
-		u32 scratch_index;
 		if (rdev->wb.use_event)
-			scratch_index = R600_WB_EVENT_OFFSET + rdev->fence_drv.scratch_reg - rdev->scratch.reg_base;
+			scratch_index = R600_WB_EVENT_OFFSET +
+				rdev->fence_drv[ring].scratch_reg - rdev->scratch.reg_base;
 		else
-			scratch_index = RADEON_WB_SCRATCH_OFFSET + rdev->fence_drv.scratch_reg - rdev->scratch.reg_base;
+			scratch_index = RADEON_WB_SCRATCH_OFFSET +
+				rdev->fence_drv[ring].scratch_reg - rdev->scratch.reg_base;
 		seq = le32_to_cpu(rdev->wb.wb[scratch_index/4]);
 	} else
-		seq = RREG32(rdev->fence_drv.scratch_reg);
+		seq = RREG32(rdev->fence_drv[ring].scratch_reg);
 	return seq;
 }
 
@@ -73,28 +78,28 @@ int radeon_fence_emit(struct radeon_device *rdev, struct radeon_fence *fence)
 {
 	unsigned long irq_flags;
 
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&rdev->fence_lock, irq_flags);
 	if (fence->emitted) {
-		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 		return 0;
 	}
-	fence->seq = atomic_add_return(1, &rdev->fence_drv.seq);
+	fence->seq = atomic_add_return(1, &rdev->fence_drv[fence->ring].seq);
 	if (!rdev->cp.ready)
 		/* FIXME: cp is not running assume everythings is done right
 		 * away
 		 */
-		radeon_fence_write(rdev, fence->seq);
+		radeon_fence_write(rdev, fence->seq, fence->ring);
 	else
 		radeon_fence_ring_emit(rdev, fence);
 
 	trace_radeon_fence_emit(rdev->ddev, fence->seq);
 	fence->emitted = true;
-	list_move_tail(&fence->list, &rdev->fence_drv.emitted);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	list_move_tail(&fence->list, &rdev->fence_drv[fence->ring].emitted);
+	write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	return 0;
 }
 
-static bool radeon_fence_poll_locked(struct radeon_device *rdev)
+static bool radeon_fence_poll_locked(struct radeon_device *rdev, int ring)
 {
 	struct radeon_fence *fence;
 	struct list_head *i, *n;
@@ -102,34 +107,34 @@ static bool radeon_fence_poll_locked(struct radeon_device *rdev)
 	bool wake = false;
 	unsigned long cjiffies;
 
-	seq = radeon_fence_read(rdev);
-	if (seq != rdev->fence_drv.last_seq) {
-		rdev->fence_drv.last_seq = seq;
-		rdev->fence_drv.last_jiffies = jiffies;
-		rdev->fence_drv.last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT;
+	seq = radeon_fence_read(rdev, ring);
+	if (seq != rdev->fence_drv[ring].last_seq) {
+		rdev->fence_drv[ring].last_seq = seq;
+		rdev->fence_drv[ring].last_jiffies = jiffies;
+		rdev->fence_drv[ring].last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT;
 	} else {
 		cjiffies = jiffies;
-		if (time_after(cjiffies, rdev->fence_drv.last_jiffies)) {
-			cjiffies -= rdev->fence_drv.last_jiffies;
-			if (time_after(rdev->fence_drv.last_timeout, cjiffies)) {
+		if (time_after(cjiffies, rdev->fence_drv[ring].last_jiffies)) {
+			cjiffies -= rdev->fence_drv[ring].last_jiffies;
+			if (time_after(rdev->fence_drv[ring].last_timeout, cjiffies)) {
 				/* update the timeout */
-				rdev->fence_drv.last_timeout -= cjiffies;
+				rdev->fence_drv[ring].last_timeout -= cjiffies;
 			} else {
 				/* the 500ms timeout is elapsed we should test
 				 * for GPU lockup
 				 */
-				rdev->fence_drv.last_timeout = 1;
+				rdev->fence_drv[ring].last_timeout = 1;
 			}
 		} else {
 			/* wrap around update last jiffies, we will just wait
 			 * a little longer
 			 */
-			rdev->fence_drv.last_jiffies = cjiffies;
+			rdev->fence_drv[ring].last_jiffies = cjiffies;
 		}
 		return false;
 	}
 	n = NULL;
-	list_for_each(i, &rdev->fence_drv.emitted) {
+	list_for_each(i, &rdev->fence_drv[ring].emitted) {
 		fence = list_entry(i, struct radeon_fence, list);
 		if (fence->seq == seq) {
 			n = i;
@@ -141,11 +146,11 @@ static bool radeon_fence_poll_locked(struct radeon_device *rdev)
 		i = n;
 		do {
 			n = i->prev;
-			list_move_tail(i, &rdev->fence_drv.signaled);
+			list_move_tail(i, &rdev->fence_drv[ring].signaled);
 			fence = list_entry(i, struct radeon_fence, list);
 			fence->signaled = true;
 			i = n;
-		} while (i != &rdev->fence_drv.emitted);
+		} while (i != &rdev->fence_drv[ring].emitted);
 		wake = true;
 	}
 	return wake;
@@ -157,14 +162,16 @@ static void radeon_fence_destroy(struct kref *kref)
         struct radeon_fence *fence;
 
 	fence = container_of(kref, struct radeon_fence, kref);
-	write_lock_irqsave(&fence->rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&fence->rdev->fence_lock, irq_flags);
 	list_del(&fence->list);
 	fence->emitted = false;
-	write_unlock_irqrestore(&fence->rdev->fence_drv.lock, irq_flags);
+	write_unlock_irqrestore(&fence->rdev->fence_lock, irq_flags);
 	kfree(fence);
 }
 
-int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence)
+int radeon_fence_create(struct radeon_device *rdev,
+			struct radeon_fence **fence,
+			int ring)
 {
 	unsigned long irq_flags;
 
@@ -177,15 +184,15 @@ int radeon_fence_create(struct radeon_device *rdev, struct radeon_fence **fence)
 	(*fence)->emitted = false;
 	(*fence)->signaled = false;
 	(*fence)->seq = 0;
+	(*fence)->ring = ring;
 	INIT_LIST_HEAD(&(*fence)->list);
 
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	list_add_tail(&(*fence)->list, &rdev->fence_drv.created);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&rdev->fence_lock, irq_flags);
+	list_add_tail(&(*fence)->list, &rdev->fence_drv[ring].created);
+	write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	return 0;
 }
 
-
 bool radeon_fence_signaled(struct radeon_fence *fence)
 {
 	unsigned long irq_flags;
@@ -197,7 +204,7 @@ bool radeon_fence_signaled(struct radeon_fence *fence)
 	if (fence->rdev->gpu_lockup)
 		return true;
 
-	write_lock_irqsave(&fence->rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&fence->rdev->fence_lock, irq_flags);
 	signaled = fence->signaled;
 	/* if we are shuting down report all fence as signaled */
 	if (fence->rdev->shutdown) {
@@ -208,10 +215,10 @@ bool radeon_fence_signaled(struct radeon_fence *fence)
 		signaled = true;
 	}
 	if (!signaled) {
-		radeon_fence_poll_locked(fence->rdev);
+		radeon_fence_poll_locked(fence->rdev, fence->ring);
 		signaled = fence->signaled;
 	}
-	write_unlock_irqrestore(&fence->rdev->fence_drv.lock, irq_flags);
+	write_unlock_irqrestore(&fence->rdev->fence_lock, irq_flags);
 	return signaled;
 }
 
@@ -230,14 +237,14 @@ int radeon_fence_wait(struct radeon_fence *fence, bool intr)
 	if (radeon_fence_signaled(fence)) {
 		return 0;
 	}
-	timeout = rdev->fence_drv.last_timeout;
+	timeout = rdev->fence_drv[fence->ring].last_timeout;
 retry:
 	/* save current sequence used to check for GPU lockup */
-	seq = rdev->fence_drv.last_seq;
+	seq = rdev->fence_drv[fence->ring].last_seq;
 	trace_radeon_fence_wait_begin(rdev->ddev, seq);
 	if (intr) {
 		radeon_irq_kms_sw_irq_get(rdev);
-		r = wait_event_interruptible_timeout(rdev->fence_drv.queue,
+		r = wait_event_interruptible_timeout(rdev->fence_drv[fence->ring].queue,
 				radeon_fence_signaled(fence), timeout);
 		radeon_irq_kms_sw_irq_put(rdev);
 		if (unlikely(r < 0)) {
@@ -245,7 +252,7 @@ retry:
 		}
 	} else {
 		radeon_irq_kms_sw_irq_get(rdev);
-		r = wait_event_timeout(rdev->fence_drv.queue,
+		r = wait_event_timeout(rdev->fence_drv[fence->ring].queue,
 			 radeon_fence_signaled(fence), timeout);
 		radeon_irq_kms_sw_irq_put(rdev);
 	}
@@ -258,10 +265,11 @@ retry:
 			timeout = r;
 			goto retry;
 		}
-		/* don't protect read access to rdev->fence_drv.last_seq
+		/* don't protect read access to rdev->fence_drv[t].last_seq
 		 * if we experiencing a lockup the value doesn't change
 		 */
-		if (seq == rdev->fence_drv.last_seq && radeon_gpu_is_lockup(rdev)) {
+		if (seq == rdev->fence_drv[fence->ring].last_seq &&
+		    radeon_gpu_is_lockup(rdev)) {
 			/* good news we believe it's a lockup */
 			printk(KERN_WARNING "GPU lockup (waiting for 0x%08X last fence id 0x%08X)\n",
 			     fence->seq, seq);
@@ -272,20 +280,20 @@ retry:
 			r = radeon_gpu_reset(rdev);
 			if (r)
 				return r;
-			radeon_fence_write(rdev, fence->seq);
+			radeon_fence_write(rdev, fence->seq, fence->ring);
 			rdev->gpu_lockup = false;
 		}
 		timeout = RADEON_FENCE_JIFFIES_TIMEOUT;
-		write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-		rdev->fence_drv.last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT;
-		rdev->fence_drv.last_jiffies = jiffies;
-		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+		write_lock_irqsave(&rdev->fence_lock, irq_flags);
+		rdev->fence_drv[fence->ring].last_timeout = RADEON_FENCE_JIFFIES_TIMEOUT;
+		rdev->fence_drv[fence->ring].last_jiffies = jiffies;
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 		goto retry;
 	}
 	return 0;
 }
 
-int radeon_fence_wait_next(struct radeon_device *rdev)
+int radeon_fence_wait_next(struct radeon_device *rdev, int ring)
 {
 	unsigned long irq_flags;
 	struct radeon_fence *fence;
@@ -294,21 +302,21 @@ int radeon_fence_wait_next(struct radeon_device *rdev)
 	if (rdev->gpu_lockup) {
 		return 0;
 	}
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	if (list_empty(&rdev->fence_drv.emitted)) {
-		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&rdev->fence_lock, irq_flags);
+	if (list_empty(&rdev->fence_drv[ring].emitted)) {
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 		return 0;
 	}
-	fence = list_entry(rdev->fence_drv.emitted.next,
+	fence = list_entry(rdev->fence_drv[ring].emitted.next,
 			   struct radeon_fence, list);
 	radeon_fence_ref(fence);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	r = radeon_fence_wait(fence, false);
 	radeon_fence_unref(&fence);
 	return r;
 }
 
-int radeon_fence_wait_last(struct radeon_device *rdev)
+int radeon_fence_wait_last(struct radeon_device *rdev, int ring)
 {
 	unsigned long irq_flags;
 	struct radeon_fence *fence;
@@ -317,15 +325,15 @@ int radeon_fence_wait_last(struct radeon_device *rdev)
 	if (rdev->gpu_lockup) {
 		return 0;
 	}
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	if (list_empty(&rdev->fence_drv.emitted)) {
-		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&rdev->fence_lock, irq_flags);
+	if (list_empty(&rdev->fence_drv[ring].emitted)) {
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 		return 0;
 	}
-	fence = list_entry(rdev->fence_drv.emitted.prev,
+	fence = list_entry(rdev->fence_drv[ring].emitted.prev,
 			   struct radeon_fence, list);
 	radeon_fence_ref(fence);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	r = radeon_fence_wait(fence, false);
 	radeon_fence_unref(&fence);
 	return r;
@@ -347,39 +355,49 @@ void radeon_fence_unref(struct radeon_fence **fence)
 	}
 }
 
-void radeon_fence_process(struct radeon_device *rdev)
+void radeon_fence_process(struct radeon_device *rdev, int ring)
 {
 	unsigned long irq_flags;
 	bool wake;
 
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	wake = radeon_fence_poll_locked(rdev);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+	write_lock_irqsave(&rdev->fence_lock, irq_flags);
+	wake = radeon_fence_poll_locked(rdev, ring);
+	write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	if (wake) {
-		wake_up_all(&rdev->fence_drv.queue);
+		wake_up_all(&rdev->fence_drv[ring].queue);
 	}
 }
 
-int radeon_fence_driver_init(struct radeon_device *rdev)
+int radeon_fence_driver_init(struct radeon_device *rdev, int num_rings)
 {
 	unsigned long irq_flags;
-	int r;
+	int r, ring;
 
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	r = radeon_scratch_get(rdev, &rdev->fence_drv.scratch_reg);
-	if (r) {
-		dev_err(rdev->dev, "fence failed to get scratch register\n");
-		write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
-		return r;
+	for (ring = 0; ring < num_rings; ring++) {
+		write_lock_irqsave(&rdev->fence_lock, irq_flags);
+		r = radeon_scratch_get(rdev, &rdev->fence_drv[ring].scratch_reg);
+		if (r) {
+			dev_err(rdev->dev, "fence failed to get scratch register\n");
+			write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
+			return r;
+		}
+		radeon_fence_write(rdev, 0, ring);
+		atomic_set(&rdev->fence_drv[ring].seq, 0);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].created);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].emitted);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].signaled);
+		init_waitqueue_head(&rdev->fence_drv[ring].queue);
+		rdev->fence_drv[ring].initialized = true;
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
+	}
+	for (ring = num_rings; ring < RADEON_NUM_RINGS; ring++) {
+		write_lock_irqsave(&rdev->fence_lock, irq_flags);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].created);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].emitted);
+		INIT_LIST_HEAD(&rdev->fence_drv[ring].signaled);
+		rdev->fence_drv[ring].initialized = false;
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 	}
-	radeon_fence_write(rdev, 0);
-	atomic_set(&rdev->fence_drv.seq, 0);
-	INIT_LIST_HEAD(&rdev->fence_drv.created);
-	INIT_LIST_HEAD(&rdev->fence_drv.emitted);
-	INIT_LIST_HEAD(&rdev->fence_drv.signaled);
-	init_waitqueue_head(&rdev->fence_drv.queue);
-	rdev->fence_drv.initialized = true;
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
 	if (radeon_debugfs_fence_init(rdev)) {
 		dev_err(rdev->dev, "fence debugfs file creation failed\n");
 	}
@@ -389,14 +407,17 @@ int radeon_fence_driver_init(struct radeon_device *rdev)
 void radeon_fence_driver_fini(struct radeon_device *rdev)
 {
 	unsigned long irq_flags;
-
-	if (!rdev->fence_drv.initialized)
-		return;
-	wake_up_all(&rdev->fence_drv.queue);
-	write_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-	radeon_scratch_free(rdev, rdev->fence_drv.scratch_reg);
-	write_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
-	rdev->fence_drv.initialized = false;
+	int ring;
+
+	for (ring = 0; ring < RADEON_NUM_RINGS; ring++) {
+		if (!rdev->fence_drv[ring].initialized)
+			continue;
+		wake_up_all(&rdev->fence_drv[ring].queue);
+		write_lock_irqsave(&rdev->fence_lock, irq_flags);
+		radeon_scratch_free(rdev, rdev->fence_drv[ring].scratch_reg);
+		write_unlock_irqrestore(&rdev->fence_lock, irq_flags);
+		rdev->fence_drv[ring].initialized = false;
+	}
 }
 
 
@@ -410,14 +431,21 @@ static int radeon_debugfs_fence_info(struct seq_file *m, void *data)
 	struct drm_device *dev = node->minor->dev;
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_fence *fence;
-
-	seq_printf(m, "Last signaled fence 0x%08X\n",
-		   radeon_fence_read(rdev));
-	if (!list_empty(&rdev->fence_drv.emitted)) {
-		   fence = list_entry(rdev->fence_drv.emitted.prev,
-				      struct radeon_fence, list);
-		   seq_printf(m, "Last emitted fence %p with 0x%08X\n",
-			      fence,  fence->seq);
+	int i;
+
+	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+		if (!rdev->fence_drv[i].initialized)
+			continue;
+
+		seq_printf(m, "--- ring %d ---\n", i);
+		seq_printf(m, "Last signaled fence 0x%08X\n",
+			   radeon_fence_read(rdev, i));
+		if (!list_empty(&rdev->fence_drv[i].emitted)) {
+			fence = list_entry(rdev->fence_drv[i].emitted.prev,
+					   struct radeon_fence, list);
+			seq_printf(m, "Last emitted fence %p with 0x%08X\n",
+				   fence,  fence->seq);
+		}
 	}
 	return 0;
 }
diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c
index 1fb84676afe..19ed2c6c424 100644
--- a/drivers/gpu/drm/radeon/radeon_pm.c
+++ b/drivers/gpu/drm/radeon/radeon_pm.c
@@ -271,7 +271,7 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
 		if (rdev->cp.ready) {
 			struct radeon_fence *fence;
 			radeon_ring_alloc(rdev, 64);
-			radeon_fence_create(rdev, &fence);
+			radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 			radeon_fence_emit(rdev, fence);
 			radeon_ring_commit(rdev);
 			radeon_fence_wait(fence, false);
@@ -797,17 +797,25 @@ static void radeon_dynpm_idle_work_handler(struct work_struct *work)
 	if (rdev->pm.dynpm_state == DYNPM_STATE_ACTIVE) {
 		unsigned long irq_flags;
 		int not_processed = 0;
-
-		read_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-		if (!list_empty(&rdev->fence_drv.emitted)) {
-			struct list_head *ptr;
-			list_for_each(ptr, &rdev->fence_drv.emitted) {
-				/* count up to 3, that's enought info */
-				if (++not_processed >= 3)
-					break;
+		int i;
+
+		read_lock_irqsave(&rdev->fence_lock, irq_flags);
+		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+			if (!rdev->fence_drv[i].initialized)
+				continue;
+
+			if (!list_empty(&rdev->fence_drv[i].emitted)) {
+				struct list_head *ptr;
+				list_for_each(ptr, &rdev->fence_drv[i].emitted) {
+					/* count up to 3, that's enought info */
+					if (++not_processed >= 3)
+						break;
+				}
 			}
+			if (not_processed >= 3)
+				break;
 		}
-		read_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+		read_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 
 		if (not_processed >= 3) { /* should upclock */
 			if (rdev->pm.dynpm_planned_action == DYNPM_ACTION_DOWNCLOCK) {
diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c
index f3d7d224eeb..c232317b1dd 100644
--- a/drivers/gpu/drm/radeon/radeon_ring.c
+++ b/drivers/gpu/drm/radeon/radeon_ring.c
@@ -113,7 +113,7 @@ int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib)
 	int r = 0, i, c;
 
 	*ib = NULL;
-	r = radeon_fence_create(rdev, &fence);
+	r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 	if (r) {
 		dev_err(rdev->dev, "failed to create fence for new IB\n");
 		return r;
@@ -314,7 +314,7 @@ int radeon_ring_alloc(struct radeon_device *rdev, unsigned ndw)
 		if (ndw < rdev->cp.ring_free_dw) {
 			break;
 		}
-		r = radeon_fence_wait_next(rdev);
+		r = radeon_fence_wait_next(rdev, RADEON_RING_TYPE_GFX_INDEX);
 		if (r)
 			return r;
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c
index 602fa3541c4..37f7acb6d5f 100644
--- a/drivers/gpu/drm/radeon/radeon_test.c
+++ b/drivers/gpu/drm/radeon/radeon_test.c
@@ -104,7 +104,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 
 		radeon_bo_kunmap(gtt_obj[i]);
 
-		r = radeon_fence_create(rdev, &fence);
+		r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 		if (r) {
 			DRM_ERROR("Failed to create GTT->VRAM fence %d\n", i);
 			goto out_cleanup;
@@ -153,7 +153,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 
 		radeon_bo_kunmap(vram_obj);
 
-		r = radeon_fence_create(rdev, &fence);
+		r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 		if (r) {
 			DRM_ERROR("Failed to create VRAM->GTT fence %d\n", i);
 			goto out_cleanup;
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index e111a381243..112ecaa6362 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -226,7 +226,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 	int r;
 
 	rdev = radeon_get_rdev(bo->bdev);
-	r = radeon_fence_create(rdev, &fence);
+	r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 	if (unlikely(r)) {
 		return r;
 	}
diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c
index 06b90c87f8f..c71fa16106c 100644
--- a/drivers/gpu/drm/radeon/rs400.c
+++ b/drivers/gpu/drm/radeon/rs400.c
@@ -516,7 +516,7 @@ int rs400_init(struct radeon_device *rdev)
 	/* initialize memory controller */
 	rs400_mc_init(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c
index b1053d64042..1c9ab940953 100644
--- a/drivers/gpu/drm/radeon/rs600.c
+++ b/drivers/gpu/drm/radeon/rs600.c
@@ -642,7 +642,7 @@ int rs600_irq_process(struct radeon_device *rdev)
 	while (status || rdev->irq.stat_regs.r500.disp_int) {
 		/* SW interrupt */
 		if (G_000044_SW_INT(status)) {
-			radeon_fence_process(rdev);
+			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 		}
 		/* GUI idle */
 		if (G_000040_GUI_IDLE(status)) {
@@ -962,7 +962,7 @@ int rs600_init(struct radeon_device *rdev)
 	rs600_mc_init(rdev);
 	rs600_debugfs(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c
index a9049ed1a51..8aa5e7ef2ef 100644
--- a/drivers/gpu/drm/radeon/rs690.c
+++ b/drivers/gpu/drm/radeon/rs690.c
@@ -735,7 +735,7 @@ int rs690_init(struct radeon_device *rdev)
 	rs690_mc_init(rdev);
 	rv515_debugfs(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c
index 6613ee9ecca..fd8da02e1ca 100644
--- a/drivers/gpu/drm/radeon/rv515.c
+++ b/drivers/gpu/drm/radeon/rv515.c
@@ -510,7 +510,7 @@ int rv515_init(struct radeon_device *rdev)
 	rv515_mc_init(rdev);
 	rv515_debugfs(rdev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	r = radeon_irq_kms_init(rdev);
diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c
index 23ae1c60ab3..8637fd84e49 100644
--- a/drivers/gpu/drm/radeon/rv770.c
+++ b/drivers/gpu/drm/radeon/rv770.c
@@ -1194,7 +1194,7 @@ int rv770_init(struct radeon_device *rdev)
 	/* Initialize clocks */
 	radeon_get_clock_info(rdev->ddev);
 	/* Fence driver */
-	r = radeon_fence_driver_init(rdev);
+	r = radeon_fence_driver_init(rdev, 1);
 	if (r)
 		return r;
 	/* initialize AGP */