drm/radeon/kms: reorganize copy callbacks

tidy up the radeon_asic struct, handle multiple
rings better.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König<christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drivers/gpu/drm/radeon/evergreen.c

@@ -3197,7 +3197,7 @@ static int evergreen_startup(struct radeon_device *rdev)
 	r = evergreen_blit_init(rdev);
 	if (r) {
 		r600_blit_fini(rdev);
-		rdev->asic->copy = NULL;
+		rdev->asic->copy.copy = NULL;
 		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
 	}
 

drivers/gpu/drm/radeon/ni.c

@@ -1466,7 +1466,7 @@ static int cayman_startup(struct radeon_device *rdev)
 	r = evergreen_blit_init(rdev);
 	if (r) {
 		r600_blit_fini(rdev);
-		rdev->asic->copy = NULL;
+		rdev->asic->copy.copy = NULL;
 		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
 	}
 

drivers/gpu/drm/radeon/r600.c

@@ -2449,7 +2449,7 @@ int r600_startup(struct radeon_device *rdev)
 	r = r600_blit_init(rdev);
 	if (r) {
 		r600_blit_fini(rdev);
-		rdev->asic->copy = NULL;
+		rdev->asic->copy.copy = NULL;
 		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
 	}
 

drivers/gpu/drm/radeon/radeon.h

@@ -1154,21 +1154,30 @@ struct radeon_asic {
 	int (*irq_set)(struct radeon_device *rdev);
 	int (*irq_process)(struct radeon_device *rdev);
 	u32 (*get_vblank_counter)(struct radeon_device *rdev, int crtc);
-	int (*copy_blit)(struct radeon_device *rdev,
+
+	struct {
+		int (*blit)(struct radeon_device *rdev,
 			    uint64_t src_offset,
 			    uint64_t dst_offset,
 			    unsigned num_gpu_pages,
 			    struct radeon_fence *fence);
-	int (*copy_dma)(struct radeon_device *rdev,
+		u32 blit_ring_index;
+		int (*dma)(struct radeon_device *rdev,
 			   uint64_t src_offset,
 			   uint64_t dst_offset,
 			   unsigned num_gpu_pages,
 			   struct radeon_fence *fence);
+		u32 dma_ring_index;
+		/* method used for bo copy */
 		int (*copy)(struct radeon_device *rdev,
 			    uint64_t src_offset,
 			    uint64_t dst_offset,
 			    unsigned num_gpu_pages,
 			    struct radeon_fence *fence);
+		/* ring used for bo copies */
+		u32 copy_ring_index;
+	} copy;
+
 	uint32_t (*get_engine_clock)(struct radeon_device *rdev);
 	void (*set_engine_clock)(struct radeon_device *rdev, uint32_t eng_clock);
 	uint32_t (*get_memory_clock)(struct radeon_device *rdev);
@@ -1505,8 +1514,6 @@ struct radeon_device {
 	unsigned 		debugfs_count;
 	/* virtual memory */
 	struct radeon_vm_manager	vm_manager;
-	/* ring used for bo copies */
-	u32				copy_ring;
 };
 
 int radeon_device_init(struct radeon_device *rdev,
@@ -1677,9 +1684,12 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v);
 #define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->get_vblank_counter((rdev), (crtc))
 #define radeon_fence_ring_emit(rdev, r, fence) (rdev)->asic->ring[(r)].emit_fence((rdev), (fence))
 #define radeon_semaphore_ring_emit(rdev, r, cp, semaphore, emit_wait) (rdev)->asic->ring[(r)].emit_semaphore((rdev), (cp), (semaphore), (emit_wait))
-#define radeon_copy_blit(rdev, s, d, np, f) (rdev)->asic->copy_blit((rdev), (s), (d), (np), (f))
-#define radeon_copy_dma(rdev, s, d, np, f) (rdev)->asic->copy_dma((rdev), (s), (d), (np), (f))
-#define radeon_copy(rdev, s, d, np, f) (rdev)->asic->copy((rdev), (s), (d), (np), (f))
+#define radeon_copy_blit(rdev, s, d, np, f) (rdev)->asic->copy.blit((rdev), (s), (d), (np), (f))
+#define radeon_copy_dma(rdev, s, d, np, f) (rdev)->asic->copy.dma((rdev), (s), (d), (np), (f))
+#define radeon_copy(rdev, s, d, np, f) (rdev)->asic->copy.copy((rdev), (s), (d), (np), (f))
+#define radeon_copy_blit_ring_index(rdev) (rdev)->asic->copy.blit_ring_index
+#define radeon_copy_dma_ring_index(rdev) (rdev)->asic->copy.dma_ring_index
+#define radeon_copy_ring_index(rdev) (rdev)->asic->copy.copy_ring_index
 #define radeon_get_engine_clock(rdev) (rdev)->asic->get_engine_clock((rdev))
 #define radeon_set_engine_clock(rdev, e) (rdev)->asic->set_engine_clock((rdev), (e))
 #define radeon_get_memory_clock(rdev) (rdev)->asic->get_memory_clock((rdev))

drivers/gpu/drm/radeon/radeon_benchmark.c

@@ -43,17 +43,19 @@ 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, RADEON_RING_TYPE_GFX_INDEX);
-		if (r)
-			return r;
-
 		switch (flag) {
 		case RADEON_BENCHMARK_COPY_DMA:
+			r = radeon_fence_create(rdev, &fence, radeon_copy_dma_ring_index(rdev));
+			if (r)
+				return r;
 			r = radeon_copy_dma(rdev, saddr, daddr,
 					    size / RADEON_GPU_PAGE_SIZE,
 					    fence);
 			break;
 		case RADEON_BENCHMARK_COPY_BLIT:
+			r = radeon_fence_create(rdev, &fence, radeon_copy_blit_ring_index(rdev));
+			if (r)
+				return r;
 			r = radeon_copy_blit(rdev, saddr, daddr,
 					     size / RADEON_GPU_PAGE_SIZE,
 					     fence);
@@ -129,7 +131,7 @@ static void radeon_benchmark_move(struct radeon_device *rdev, unsigned size,
 	/* r100 doesn't have dma engine so skip the test */
 	/* also, VRAM-to-VRAM test doesn't make much sense for DMA */
 	/* skip it as well if domains are the same */
-	if ((rdev->asic->copy_dma) && (sdomain != ddomain)) {
+	if ((rdev->asic->copy.dma) && (sdomain != ddomain)) {
 		time = radeon_benchmark_do_move(rdev, size, saddr, daddr,
 						RADEON_BENCHMARK_COPY_DMA, n);
 		if (time < 0)

drivers/gpu/drm/radeon/radeon_ttm.c

@@ -226,7 +226,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 	int r, i;
 
 	rdev = radeon_get_rdev(bo->bdev);
-	r = radeon_fence_create(rdev, &fence, rdev->copy_ring);
+	r = radeon_fence_create(rdev, &fence, radeon_copy_ring_index(rdev));
 	if (unlikely(r)) {
 		return r;
 	}
@@ -255,7 +255,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 		DRM_ERROR("Unknown placement %d\n", old_mem->mem_type);
 		return -EINVAL;
 	}
-	if (!rdev->ring[rdev->copy_ring].ready) {
+	if (!rdev->ring[radeon_copy_ring_index(rdev)].ready) {
 		DRM_ERROR("Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
@@ -266,7 +266,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 	if (rdev->family >= CHIP_R600) {
 		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
 			/* no need to sync to our own or unused rings */
-			if (i == rdev->copy_ring || !rdev->ring[i].ready)
+			if (i == radeon_copy_ring_index(rdev) || !rdev->ring[i].ready)
 				continue;
 
 			if (!fence->semaphore) {
@@ -283,12 +283,12 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 			radeon_semaphore_emit_signal(rdev, i, fence->semaphore);
 			radeon_ring_unlock_commit(rdev, &rdev->ring[i]);
 
-			r = radeon_ring_lock(rdev, &rdev->ring[rdev->copy_ring], 3);
+			r = radeon_ring_lock(rdev, &rdev->ring[radeon_copy_ring_index(rdev)], 3);
 			/* FIXME: handle ring lock error */
 			if (r)
 				continue;
-			radeon_semaphore_emit_wait(rdev, rdev->copy_ring, fence->semaphore);
-			radeon_ring_unlock_commit(rdev, &rdev->ring[rdev->copy_ring]);
+			radeon_semaphore_emit_wait(rdev, radeon_copy_ring_index(rdev), fence->semaphore);
+			radeon_ring_unlock_commit(rdev, &rdev->ring[radeon_copy_ring_index(rdev)]);
 		}
 	}
 
@@ -410,7 +410,8 @@ static int radeon_bo_move(struct ttm_buffer_object *bo,
 		radeon_move_null(bo, new_mem);
 		return 0;
 	}
-	if (!rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready || rdev->asic->copy == NULL) {
+	if (!rdev->ring[radeon_copy_ring_index(rdev)].ready ||
+	    rdev->asic->copy.copy == NULL) {
 		/* use memcpy */
 		goto memcpy;
 	}

drivers/gpu/drm/radeon/rv770.c

@@ -1074,7 +1074,7 @@ static int rv770_startup(struct radeon_device *rdev)
 	r = r600_blit_init(rdev);
 	if (r) {
 		r600_blit_fini(rdev);
-		rdev->asic->copy = NULL;
+		rdev->asic->copy.copy = NULL;
 		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
 	}