drm/nvc0/ttm: use copy engines for async buffer moves

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

1 files changed, 51 insertions, 9 deletions

diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index fc463ef92d4..7f80ed52356 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -523,6 +523,44 @@ nvc0_bo_move_init(struct nouveau_channel *chan, u32 handle)
 }
 
 static int
+nvc0_bo_move_copy(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
+		  struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
+{
+	struct nouveau_mem *node = old_mem->mm_node;
+	u64 src_offset = node->vma[0].offset;
+	u64 dst_offset = node->vma[1].offset;
+	u32 page_count = new_mem->num_pages;
+	int ret;
+
+	page_count = new_mem->num_pages;
+	while (page_count) {
+		int line_count = (page_count > 8191) ? 8191 : page_count;
+
+		ret = RING_SPACE(chan, 11);
+		if (ret)
+			return ret;
+
+		BEGIN_NVC0(chan, NvSubCopy, 0x030c, 8);
+		OUT_RING  (chan, upper_32_bits(src_offset));
+		OUT_RING  (chan, lower_32_bits(src_offset));
+		OUT_RING  (chan, upper_32_bits(dst_offset));
+		OUT_RING  (chan, lower_32_bits(dst_offset));
+		OUT_RING  (chan, PAGE_SIZE);
+		OUT_RING  (chan, PAGE_SIZE);
+		OUT_RING  (chan, PAGE_SIZE);
+		OUT_RING  (chan, line_count);
+		BEGIN_NVC0(chan, NvSubCopy, 0x0300, 1);
+		OUT_RING  (chan, 0x00000110);
+
+		page_count -= line_count;
+		src_offset += (PAGE_SIZE * line_count);
+		dst_offset += (PAGE_SIZE * line_count);
+	}
+
+	return 0;
+}
+
+static int
 nvc0_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 		  struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
 {
@@ -887,28 +925,32 @@ nouveau_bo_move_init(struct nouveau_channel *chan)
 	struct drm_nouveau_private *dev_priv = chan->dev->dev_private;
 	static const struct {
 		const char *name;
+		int engine;
 		u32 oclass;
 		int (*exec)(struct nouveau_channel *,
 			    struct ttm_buffer_object *,
 			    struct ttm_mem_reg *, struct ttm_mem_reg *);
 		int (*init)(struct nouveau_channel *, u32 handle);
 	} _methods[] = {
-		{  "COPY", 0xa0b5, nve0_bo_move_copy, nvc0_bo_move_init },
-		{  "COPY", 0x85b5, nva3_bo_move_copy, nv50_bo_move_init },
-		{ "CRYPT", 0x74c1, nv84_bo_move_exec, nv50_bo_move_init },
-		{  "M2MF", 0x9039, nvc0_bo_move_m2mf, nvc0_bo_move_init },
-		{  "M2MF", 0x5039, nv50_bo_move_m2mf, nv50_bo_move_init },
-		{  "M2MF", 0x0039, nv04_bo_move_m2mf, nv04_bo_move_init },
+		{  "COPY", 0, 0xa0b5, nve0_bo_move_copy, nvc0_bo_move_init },
+		{ "COPY1", 5, 0x90b8, nvc0_bo_move_copy, nvc0_bo_move_init },
+		{ "COPY0", 4, 0x90b5, nvc0_bo_move_copy, nvc0_bo_move_init },
+		{  "COPY", 0, 0x85b5, nva3_bo_move_copy, nv50_bo_move_init },
+		{ "CRYPT", 0, 0x74c1, nv84_bo_move_exec, nv50_bo_move_init },
+		{  "M2MF", 0, 0x9039, nvc0_bo_move_m2mf, nvc0_bo_move_init },
+		{  "M2MF", 0, 0x5039, nv50_bo_move_m2mf, nv50_bo_move_init },
+		{  "M2MF", 0, 0x0039, nv04_bo_move_m2mf, nv04_bo_move_init },
 		{},
-		{ "CRYPT", 0x88b4, nv98_bo_move_exec, nv50_bo_move_init },
+		{ "CRYPT", 0, 0x88b4, nv98_bo_move_exec, nv50_bo_move_init },
 	}, *mthd = _methods;
 	const char *name = "CPU";
 	int ret;
 
 	do {
-		ret = nouveau_gpuobj_gr_new(chan, mthd->oclass, mthd->oclass);
+		u32 handle = (mthd->engine << 16) | mthd->oclass;
+		ret = nouveau_gpuobj_gr_new(chan, handle, mthd->oclass);
 		if (ret == 0) {
-			ret = mthd->init(chan, mthd->oclass);
+			ret = mthd->init(chan, handle);
 			if (ret == 0) {
 				dev_priv->ttm.move = mthd->exec;
 				name = mthd->name;