[RFC PATCH] drm/amdgpu: alloc uvd msg from IB pool

[xinhui pan <xinhui.pan@amd.com>]

There is one dedicated IB pool for IB test. So lets use it for uvd msg
too.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 223 +++++++++++++-----------
 1 file changed, 126 insertions(+), 97 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index d451c359606a..29fbe976a300 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -1066,37 +1066,65 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx)
 	return 0;
 }
 
-static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
-			       bool direct, struct dma_fence **fence)
+typedef void (*uvd_msg_cb) (struct amdgpu_ring *ring, uint32_t handle,
+		struct amdgpu_ib *ib, uint32_t *msg, uint64_t msg_gpu_addr);
+
+static void amdgpu_uvd_get_create_msg_cb(struct amdgpu_ring *ring, uint32_t handle,
+		struct amdgpu_ib *ib, uint32_t *msg, uint64_t addr)
 {
 	struct amdgpu_device *adev = ring->adev;
-	struct dma_fence *f = NULL;
-	struct amdgpu_job *job;
-	struct amdgpu_ib *ib;
 	uint32_t data[4];
-	uint64_t addr;
-	long r;
+	unsigned int offset_idx = 0;
+	unsigned int offset[3] = { UVD_BASE_SI, 0, 0 };
 	int i;
-	unsigned offset_idx = 0;
-	unsigned offset[3] = { UVD_BASE_SI, 0, 0 };
 
-	amdgpu_bo_kunmap(bo);
-	amdgpu_bo_unpin(bo);
+	if (adev->asic_type >= CHIP_VEGA10) {
+		offset_idx = 1 + ring->me;
+		offset[1] = adev->reg_offset[UVD_HWIP][0][1];
+		offset[2] = adev->reg_offset[UVD_HWIP][1][1];
+	}
 
-	if (!ring->adev->uvd.address_64_bit) {
-		struct ttm_operation_ctx ctx = { true, false };
+	data[0] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_DATA0, 0);
+	data[1] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_DATA1, 0);
+	data[2] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_CMD, 0);
+	data[3] = PACKET0(offset[offset_idx] + UVD_NO_OP, 0);
 
-		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM);
-		amdgpu_uvd_force_into_uvd_segment(bo);
-		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
-		if (r)
-			goto err;
+	ib->ptr[0] = data[0];
+	ib->ptr[1] = addr;
+	ib->ptr[2] = data[1];
+	ib->ptr[3] = addr >> 32;
+	ib->ptr[4] = data[2];
+	ib->ptr[5] = 0;
+	for (i = 6; i < 16; i += 2) {
+		ib->ptr[i] = data[3];
+		ib->ptr[i+1] = 0;
 	}
+	ib->length_dw = 16;
 
-	r = amdgpu_job_alloc_with_ib(adev, 64, direct ? AMDGPU_IB_POOL_DIRECT :
-				     AMDGPU_IB_POOL_DELAYED, &job);
-	if (r)
-		goto err;
+	/* stitch together an UVD create msg */
+	msg[0] = cpu_to_le32(0x00000de4);
+	msg[1] = cpu_to_le32(0x00000000);
+	msg[2] = cpu_to_le32(handle);
+	msg[3] = cpu_to_le32(0x00000000);
+	msg[4] = cpu_to_le32(0x00000000);
+	msg[5] = cpu_to_le32(0x00000000);
+	msg[6] = cpu_to_le32(0x00000000);
+	msg[7] = cpu_to_le32(0x00000780);
+	msg[8] = cpu_to_le32(0x00000440);
+	msg[9] = cpu_to_le32(0x00000000);
+	msg[10] = cpu_to_le32(0x01b37000);
+	for (i = 11; i < 1024; ++i)
+		msg[i] = cpu_to_le32(0x0);
+}
+
+static void amdgpu_uvd_get_destroy_msg_cb(struct amdgpu_ring *ring, uint32_t handle,
+		struct amdgpu_ib *ib, uint32_t *msg, uint64_t addr)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t data[4];
+	unsigned int offset_idx = 0;
+	unsigned int offset[3] = { UVD_BASE_SI, 0, 0 };
+	int i;
 
 	if (adev->asic_type >= CHIP_VEGA10) {
 		offset_idx = 1 + ring->me;
@@ -1109,8 +1137,6 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 	data[2] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_CMD, 0);
 	data[3] = PACKET0(offset[offset_idx] + UVD_NO_OP, 0);
 
-	ib = &job->ibs[0];
-	addr = amdgpu_bo_gpu_offset(bo);
 	ib->ptr[0] = data[0];
 	ib->ptr[1] = addr;
 	ib->ptr[2] = data[1];
@@ -1123,46 +1149,92 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 	}
 	ib->length_dw = 16;
 
-	if (direct) {
+	/* stitch together an UVD destroy msg */
+	msg[0] = cpu_to_le32(0x00000de4);
+	msg[1] = cpu_to_le32(0x00000002);
+	msg[2] = cpu_to_le32(handle);
+	msg[3] = cpu_to_le32(0x00000000);
+	for (i = 4; i < 1024; ++i)
+		msg[i] = cpu_to_le32(0x0);
+}
+
+static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, bool direct,
+		uvd_msg_cb cb, uint32_t handle, uint32_t ib_size,
+		uint32_t msg_size, struct dma_fence **fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct dma_fence *f = NULL;
+	struct amdgpu_job *job;
+	struct amdgpu_bo *bo = NULL;
+	struct amdgpu_ib msg;
+	int r;
+
+	if (!ring->adev->uvd.address_64_bit) {
+		struct ttm_operation_ctx ctx = { true, false };
+
+		r = amdgpu_bo_create_reserved(adev, msg_size, PAGE_SIZE,
+				AMDGPU_GEM_DOMAIN_VRAM,
+				&bo, NULL, (void **)&msg.ptr);
+		if (r)
+			return r;
+		amdgpu_bo_kunmap(bo);
+		amdgpu_bo_unpin(bo);
+		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM);
+		amdgpu_uvd_force_into_uvd_segment(bo);
+		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+		if (r)
+			goto error;
 		r = dma_resv_wait_timeout(bo->tbo.base.resv, true, false,
-					  msecs_to_jiffies(10));
+				msecs_to_jiffies(10));
 		if (r == 0)
 			r = -ETIMEDOUT;
 		if (r < 0)
-			goto err_free;
-
-		r = amdgpu_job_submit_direct(job, ring, &f);
-		if (r)
-			goto err_free;
+			goto error;
 	} else {
-		r = amdgpu_sync_resv(adev, &job->sync, bo->tbo.base.resv,
-				     AMDGPU_SYNC_ALWAYS,
-				     AMDGPU_FENCE_OWNER_UNDEFINED);
+		memset(&msg, 0, sizeof(msg));
+		r = amdgpu_ib_get(adev, NULL, PAGE_ALIGN(msg_size),
+				direct ?
+				AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_DELAYED,
+				&msg);
 		if (r)
-			goto err_free;
+			goto error;
+	}
 
+	r = amdgpu_job_alloc_with_ib(adev, ib_size,
+			direct ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_DELAYED,
+			&job);
+	if (r)
+		goto error;
+
+	cb(ring, handle, &(job->ibs[0]), msg.ptr,
+			bo ? amdgpu_bo_gpu_offset(bo) : msg.gpu_addr);
+
+	if (direct)
+		r = amdgpu_job_submit_direct(job, ring, &f);
+	else
 		r = amdgpu_job_submit(job, &adev->uvd.entity,
 				      AMDGPU_FENCE_OWNER_UNDEFINED, &f);
-		if (r)
-			goto err_free;
+	if (r) {
+		amdgpu_job_free(job);
+		goto error;
 	}
 
-	amdgpu_bo_fence(bo, f, false);
-	amdgpu_bo_unreserve(bo);
-	amdgpu_bo_unref(&bo);
+	if (bo)
+		amdgpu_bo_fence(bo, f, false);
+	else
+		amdgpu_ib_free(adev, &msg, f);
 
 	if (fence)
 		*fence = dma_fence_get(f);
 	dma_fence_put(f);
-
-	return 0;
-
-err_free:
-	amdgpu_job_free(job);
-
-err:
-	amdgpu_bo_unreserve(bo);
-	amdgpu_bo_unref(&bo);
+error:
+	if (bo) {
+		amdgpu_bo_unreserve(bo);
+		amdgpu_bo_unref(&bo);
+	} else {
+		if (r)
+			amdgpu_ib_free(adev, &msg, NULL);
+	}
 	return r;
 }
 
@@ -1172,58 +1244,15 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 			      struct dma_fence **fence)
 {
-	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_bo *bo = NULL;
-	uint32_t *msg;
-	int r, i;
-
-	r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_GTT,
-				      &bo, NULL, (void **)&msg);
-	if (r)
-		return r;
-
-	/* stitch together an UVD create msg */
-	msg[0] = cpu_to_le32(0x00000de4);
-	msg[1] = cpu_to_le32(0x00000000);
-	msg[2] = cpu_to_le32(handle);
-	msg[3] = cpu_to_le32(0x00000000);
-	msg[4] = cpu_to_le32(0x00000000);
-	msg[5] = cpu_to_le32(0x00000000);
-	msg[6] = cpu_to_le32(0x00000000);
-	msg[7] = cpu_to_le32(0x00000780);
-	msg[8] = cpu_to_le32(0x00000440);
-	msg[9] = cpu_to_le32(0x00000000);
-	msg[10] = cpu_to_le32(0x01b37000);
-	for (i = 11; i < 1024; ++i)
-		msg[i] = cpu_to_le32(0x0);
-
-	return amdgpu_uvd_send_msg(ring, bo, true, fence);
+	return amdgpu_uvd_send_msg(ring, true, amdgpu_uvd_get_create_msg_cb,
+			handle, 64, PAGE_SIZE, fence);
 }
 
 int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
-			       bool direct, struct dma_fence **fence)
+			      bool direct, struct dma_fence **fence)
 {
-	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_bo *bo = NULL;
-	uint32_t *msg;
-	int r, i;
-
-	r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_GTT,
-				      &bo, NULL, (void **)&msg);
-	if (r)
-		return r;
-
-	/* stitch together an UVD destroy msg */
-	msg[0] = cpu_to_le32(0x00000de4);
-	msg[1] = cpu_to_le32(0x00000002);
-	msg[2] = cpu_to_le32(handle);
-	msg[3] = cpu_to_le32(0x00000000);
-	for (i = 4; i < 1024; ++i)
-		msg[i] = cpu_to_le32(0x0);
-
-	return amdgpu_uvd_send_msg(ring, bo, direct, fence);
+	return amdgpu_uvd_send_msg(ring, direct, amdgpu_uvd_get_destroy_msg_cb,
+			handle, 64, PAGE_SIZE, fence);
 }
 
 static void amdgpu_uvd_idle_work_handler(struct work_struct *work)
-- 
2.25.1