[PATCH 2/3] drm/amdgpu: add support to configure MALL for sienna_cichlid (v2)

[Luben Tuikov]

On 2020-10-20 4:26 p.m., Alex Deucher wrote:
> From: Likun Gao <Likun.Gao at amd.com>
> 
> Enable Memory Access at Last Level (MALL) feature for sienna_cichlid.
> 
> v2: drop module option.  We need to add UAPI so userspace can
> request MALL per buffer.
> 
> Signed-off-by: Likun Gao <Likun.Gao at amd.com>
> Reviewed-by: Hawking Zhang <Hawking.Zhang at amd.com>
> Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +++
>  drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 3 ++-
>  drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 2 +-
>  3 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> index ffea3b89b9da..929d7cb92dc0 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> @@ -76,6 +76,9 @@ struct amdgpu_bo_list_entry;
>  /* PTE is handled as PDE for VEGA10 (Translate Further) */
>  #define AMDGPU_PTE_TF		(1ULL << 56)
>  
> +/* MALL noalloc for sienna_cichlid, reserved for older ASICs  */
> +#define AMDGPU_PTE_NOALLOC	(1ULL << 58)
> +

Would've been good to define "MALL", as it is being
done in the message of this commit above, here
in this comment to the macro.

Else, what "MALL" means is lost, unless
one does git-blame and finds the commit which
introduced it.

Regards,
Luben

>  /* PDE Block Fragment Size for VEGA10 */
>  #define AMDGPU_PDE_BFS(a)	((uint64_t)a << 59)
>  
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> index 40af17610207..ef385a529013 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> @@ -486,7 +486,8 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
>  /*
>   * PTE format on NAVI 10:
>   * 63:59 reserved
> - * 58:57 reserved
> + * 58 reserved and for sienna_cichlid is used for MALL noalloc
> + * 57 reserved
>   * 56 F
>   * 55 L
>   * 54 reserved
> diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> index 18eca0d4dbcc..ae6158456094 100644
> --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
> @@ -707,7 +707,7 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
>  		temp &= 0xFF0FFF;
>  		temp |= ((CACHE_READ_POLICY_L2__DEFAULT << 12) |
>  			 (CACHE_WRITE_POLICY_L2__DEFAULT << 14) |
> -			 0x01000000);
> +			 SDMA0_UTCL1_PAGE__LLC_NOALLOC_MASK);
>  		WREG32(sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_UTCL1_PAGE), temp);
>  
>  		if (!amdgpu_sriov_vf(adev)) {
>