[PATCH 1/2] drm/amdgpu: add support for inplace IB patching for MM engines v2

[Christian König]

Hi James,

can you give those two patches another try on a Vega20? At least in 
theory it should work now.

Thanks in advance,
Christian.

Am 25.07.2018 um 13:05 schrieb Christian König:
> We are going to need that for the second UVD instance on Vega20.
>
> v2: rename to patch_cs_in_place
>
> Signed-off-by: Christian König <christian.koenig at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu.h      |  1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c   | 22 +++++++++++++++-------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h |  1 +
>   3 files changed, 17 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index a9f09daeffa3..3ec1f31b892c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -1751,6 +1751,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
>   #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))
>   #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))
>   #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
> +#define amdgpu_ring_patch_cs_in_place(r, p, ib) ((r)->funcs->patch_cs_in_place((p), (ib)))
>   #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r))
>   #define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t))
>   #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> index 7c5cc33d0cda..566e2c51931b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> @@ -916,7 +916,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,
>   	int r;
>   
>   	/* Only for UVD/VCE VM emulation */
> -	if (p->ring->funcs->parse_cs) {
> +	if (p->ring->funcs->parse_cs || p->ring->funcs->patch_cs_in_place) {
>   		unsigned i, j;
>   
>   		for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) {
> @@ -957,12 +957,20 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,
>   			offset = m->start * AMDGPU_GPU_PAGE_SIZE;
>   			kptr += va_start - offset;
>   
> -			memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
> -			amdgpu_bo_kunmap(aobj);
> -
> -			r = amdgpu_ring_parse_cs(ring, p, j);
> -			if (r)
> -				return r;
> +			if (p->ring->funcs->parse_cs) {
> +				memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
> +				amdgpu_bo_kunmap(aobj);
> +
> +				r = amdgpu_ring_parse_cs(ring, p, j);
> +				if (r)
> +					return r;
> +			} else {
> +				ib->ptr = (uint32_t *)kptr;
> +				r = amdgpu_ring_patch_cs_in_place(ring, p, j);
> +				amdgpu_bo_kunmap(aobj);
> +				if (r)
> +					return r;
> +			}
>   
>   			j++;
>   		}
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
> index 5018c0b6bf1a..d242b9a51e90 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
> @@ -123,6 +123,7 @@ struct amdgpu_ring_funcs {
>   	void (*set_wptr)(struct amdgpu_ring *ring);
>   	/* validating and patching of IBs */
>   	int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx);
> +	int (*patch_cs_in_place)(struct amdgpu_cs_parser *p, uint32_t ib_idx);
>   	/* constants to calculate how many DW are needed for an emit */
>   	unsigned emit_frame_size;
>   	unsigned emit_ib_size;