[PATCH 3/5] drm/amdgpu/vcn: add test for dec vcn software ring

[Christian König]

Am 11.11.20 um 23:27 schrieb James Zhu:
> Add vcn software ring decode ring test and decode ib test.
>
> Signed-off-by: James Zhu <James.Zhu at amd.com>
> Reviewed-by: Leo Liu <leo.liu at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 138 ++++++++++++++++++++++++++++++--
>   1 file changed, 132 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> index 7e19a66..e3d54fa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
> @@ -456,6 +456,37 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
>   	return r;
>   }
>   
> +int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring)
> +{
> +	struct amdgpu_device *adev = ring->adev;
> +	uint32_t rptr;
> +	unsigned int i;
> +	int r;
> +
> +	if (amdgpu_sriov_vf(adev))
> +		return 0;
> +
> +	r = amdgpu_ring_alloc(ring, 16);
> +	if (r)
> +		return r;
> +
> +	rptr = amdgpu_ring_get_rptr(ring);
> +
> +	amdgpu_ring_write(ring, VCN_DEC_SW_CMD_END);
> +	amdgpu_ring_commit(ring);
> +
> +	for (i = 0; i < adev->usec_timeout; i++) {
> +		if (amdgpu_ring_get_rptr(ring) != rptr)
> +			break;
> +		udelay(1);
> +	}
> +
> +	if (i >= adev->usec_timeout)
> +		r = -ETIMEDOUT;
> +
> +	return r;
> +}
> +
>   static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
>   				   struct amdgpu_bo *bo,
>   				   struct dma_fence **fence)
> @@ -510,7 +541,10 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
>   }
>   
>   static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
> -			      struct dma_fence **fence)
> +			      struct dma_fence **fence,
> +			      int (*send_msg)(struct amdgpu_ring *,
> +	                              struct amdgpu_bo *,
> +	                              struct dma_fence **))
>   {
>   	struct amdgpu_device *adev = ring->adev;
>   	struct amdgpu_bo *bo = NULL;
> @@ -540,13 +574,17 @@ static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
>   	for (i = 14; i < 1024; ++i)
>   		msg[i] = cpu_to_le32(0x0);
>   
> -	return amdgpu_vcn_dec_send_msg(ring, bo, fence);
> +	return send_msg(ring, bo, fence);
>   }
>   
>   static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
> -			       struct dma_fence **fence)
> +			      struct dma_fence **fence,
> +			      int (*send_msg)(struct amdgpu_ring *,
> +	                              struct amdgpu_bo *,
> +	                              struct dma_fence **))
>   {
>   	struct amdgpu_device *adev = ring->adev;
> +	struct amdgpu_send_msg *dec;
>   	struct amdgpu_bo *bo = NULL;
>   	uint32_t *msg;
>   	int r, i;
> @@ -566,19 +604,107 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
>   	for (i = 6; i < 1024; ++i)
>   		msg[i] = cpu_to_le32(0x0);
>   
> -	return amdgpu_vcn_dec_send_msg(ring, bo, fence);
> +	return send_msg(ring, bo, fence);

You can avoid the send_msg parameter altogether by just doing this in 
the caller of the function.

Christian.

>   }
>   
>   int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout)
>   {
> +	struct amdgpu_device *adev = ring->adev;
>   	struct dma_fence *fence;
>   	long r;
>   
> -	r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL);
> +	r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL,
> +				amdgpu_vcn_dec_send_msg);
> +	if (r)
> +		goto error;
> +
> +	r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence,
> +				amdgpu_vcn_dec_send_msg);
> +	if (r)
> +		goto error;
> +
> +	r = dma_fence_wait_timeout(fence, false, timeout);
> +	if (r == 0)
> +		r = -ETIMEDOUT;
> +	else if (r > 0)
> +		r = 0;
> +
> +	dma_fence_put(fence);
> +error:
> +	return r;
> +}
> +
> +static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
> +				   struct amdgpu_bo *bo,
> +				   struct dma_fence **fence)
> +{
> +	struct amdgpu_vcn_decode_buffer *decode_buffer = NULL;
> +	const unsigned int ib_size_dw = 64;
> +	struct amdgpu_device *adev = ring->adev;
> +	struct dma_fence *f = NULL;
> +	struct amdgpu_job *job;
> +	struct amdgpu_ib *ib;
> +	uint64_t addr;
> +	int i, r;
> +
> +	r = amdgpu_job_alloc_with_ib(adev, ib_size_dw * 4,
> +				AMDGPU_IB_POOL_DIRECT, &job);
> +	if (r)
> +		goto err;
> +
> +	ib = &job->ibs[0];
> +	addr = amdgpu_bo_gpu_offset(bo);
> +	ib->length_dw = 0;
> +
> +	ib->ptr[ib->length_dw++] = sizeof(struct amdgpu_vcn_decode_buffer) + 8;
> +	ib->ptr[ib->length_dw++] = cpu_to_le32(AMDGPU_VCN_IB_FLAG_DECODE_BUFFER);
> +	decode_buffer = (struct amdgpu_vcn_decode_buffer *)&(ib->ptr[ib->length_dw]);
> +	ib->length_dw += sizeof(struct amdgpu_vcn_decode_buffer) / 4;
> +	memset(decode_buffer, 0, sizeof(struct amdgpu_vcn_decode_buffer));
> +
> +	decode_buffer->valid_buf_flag |= cpu_to_le32(AMDGPU_VCN_CMD_FLAG_MSG_BUFFER);
> +	decode_buffer->msg_buffer_address_hi = cpu_to_le32(addr >> 32);
> +	decode_buffer->msg_buffer_address_lo = cpu_to_le32(addr);
> +
> +	for (i = ib->length_dw; i < ib_size_dw; ++i)
> +		ib->ptr[i] = 0x0;
> +
> +	r = amdgpu_job_submit_direct(job, ring, &f);
> +	if (r)
> +		goto err_free;
> +
> +	amdgpu_bo_fence(bo, f, false);
> +	amdgpu_bo_unreserve(bo);
> +	amdgpu_bo_unref(&bo);
> +
> +	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);
> +	return r;
> +}
> +
> +int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout)
> +{
> +	struct amdgpu_device *adev = ring->adev;
> +	struct dma_fence *fence = NULL;
> +	long r;
> +
> +	r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL,
> +				amdgpu_vcn_dec_sw_send_msg);
>   	if (r)
>   		goto error;
>   
> -	r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence);
> +	r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence,
> +				amdgpu_vcn_dec_sw_send_msg);
>   	if (r)
>   		goto error;
>