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>

drivers/gpu/drm/radeon/evergreen.c

@@ -3018,11 +3018,11 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		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 */

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 */

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);

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);

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;
 	}

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);

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 @@ int r600_irq_process(struct radeon_device *rdev)
 		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");

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;

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;
 

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);
 

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;
+		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;
 
-		read_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-		if (!list_empty(&rdev->fence_drv.emitted)) {
+			if (!list_empty(&rdev->fence_drv[i].emitted)) {
 				struct list_head *ptr;
-			list_for_each(ptr, &rdev->fence_drv.emitted) {
+				list_for_each(ptr, &rdev->fence_drv[i].emitted) {
 					/* count up to 3, that's enought info */
 					if (++not_processed >= 3)
 						break;
 				}
 			}
-		read_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
+			if (not_processed >= 3)
+				break;
+		}
+		read_unlock_irqrestore(&rdev->fence_lock, irq_flags);
 
 		if (not_processed >= 3) { /* should upclock */
 			if (rdev->pm.dynpm_planned_action == DYNPM_ACTION_DOWNCLOCK) {

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;
 	}

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;

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;
 	}

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);

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);

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);

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);

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 */