[PATCH v2 6/6] drm/amdgpu: do not allocate entries separately

[Christian König]

Am 28.05.21 um 12:56 schrieb Nirmoy Das:
> Allocate PD/PT entries while allocating VM BOs and use that
> instead of allocating those entries separately.
>
> v2: create a new var for num entries.
>
> Signed-off-by: Nirmoy Das <nirmoy.das at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 34 +++++++++++++++-----------
>   1 file changed, 20 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 223c63342ecd..7e478ffb7fdf 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -877,6 +877,7 @@ static int amdgpu_vm_pt_create(struct amdgpu_device *adev,
>   	struct amdgpu_bo *bo;
>   	struct dma_resv *resv;
>   	int r;
> +	unsigned int num_entries;

Move that a more up, variables like r and i should be always declared last.

Apart from that Reviewed-by: Christian König <christian.koenig at amd.com>

Next step should probably be to remove entry->entries.

Regards,
Christian.

>
>   	memset(&bp, 0, sizeof(bp));
>
> @@ -886,7 +887,14 @@ static int amdgpu_vm_pt_create(struct amdgpu_device *adev,
>   	bp.domain = amdgpu_bo_get_preferred_pin_domain(adev, bp.domain);
>   	bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
>   		AMDGPU_GEM_CREATE_CPU_GTT_USWC;
> -	bp.bo_ptr_size = sizeof(struct amdgpu_bo_vm);
> +
> +	if (level < AMDGPU_VM_PTB)
> +		num_entries = amdgpu_vm_num_entries(adev, level);
> +	else
> +		num_entries = 0;
> +
> +	bp.bo_ptr_size = struct_size((*vmbo), entries, num_entries);
> +
>   	if (vm->use_cpu_for_update)
>   		bp.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
>
> @@ -957,19 +965,14 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
>   	struct amdgpu_bo_vm *pt;
>   	int r;
>
> -	if (cursor->level < AMDGPU_VM_PTB && !entry->entries) {
> -		unsigned num_entries;
> -
> -		num_entries = amdgpu_vm_num_entries(adev, cursor->level);
> -		entry->entries = kvmalloc_array(num_entries,
> -						sizeof(*entry->entries),
> -						GFP_KERNEL | __GFP_ZERO);
> -		if (!entry->entries)
> -			return -ENOMEM;
> -	}
> -
> -	if (entry->base.bo)
> +	if (entry->base.bo) {
> +		if (cursor->level < AMDGPU_VM_PTB)
> +			entry->entries =
> +				to_amdgpu_bo_vm(entry->base.bo)->entries;
> +		else
> +			entry->entries = NULL;
>   		return 0;
> +	}
>
>   	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt);
>   	if (r)
> @@ -981,6 +984,10 @@ static int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
>   	pt_bo = &pt->bo;
>   	pt_bo->parent = amdgpu_bo_ref(cursor->parent->base.bo);
>   	amdgpu_vm_bo_base_init(&entry->base, vm, pt_bo);
> +	if (cursor->level < AMDGPU_VM_PTB)
> +		entry->entries = pt->entries;
> +	else
> +		entry->entries = NULL;
>
>   	r = amdgpu_vm_clear_bo(adev, vm, pt, immediate);
>   	if (r)
> @@ -1010,7 +1017,6 @@ static void amdgpu_vm_free_table(struct amdgpu_vm_pt *entry)
>   		amdgpu_bo_unref(&shadow);
>   		amdgpu_bo_unref(&entry->base.bo);
>   	}
> -	kvfree(entry->entries);
>   	entry->entries = NULL;
>   }
>
> --
> 2.31.1
>