[PATCH v2 3/3] drm/xe: Return 2MB page size for compact 64k PTEs

[Thomas Hellström]

On Mon, 2024-02-19 at 13:19 -0800, Matthew Brost wrote:
> Compact 64k PTEs are only intended to be used within a single VMA
> which
> covers the entire 2MB range of the compact 64k PTEs. Add
> XE_VMA_PTE_COMPACT VMA flag to indicate compact 64k PTEs are used and
> update xe_vma_max_pte_size to return at least 2MB if set.
> 
> v2: Include missing changes
> 
> Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/758
> Fixes: 8f33b4f054fc ("drm/xe: Avoid doing rebinds")
> Fixes: c47794bdd63d ("drm/xe: Set max pte size when skipping
> rebinds")
> Reported-by: Paulo Zanoni <paulo.r.zanoni at intel.com>
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>

Reviewed-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>

> ---
>  drivers/gpu/drm/xe/xe_pt.c       | 5 ++++-
>  drivers/gpu/drm/xe/xe_vm.c       | 2 +-
>  drivers/gpu/drm/xe/xe_vm_types.h | 1 +
>  3 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
> index 46cea37bdac5..7f54bc3e389d 100644
> --- a/drivers/gpu/drm/xe/xe_pt.c
> +++ b/drivers/gpu/drm/xe/xe_pt.c
> @@ -547,13 +547,16 @@ xe_pt_stage_bind_entry(struct xe_ptw *parent,
> pgoff_t offset,
>  		*child = &xe_child->base;
>  
>  		/*
> -		 * Prefer the compact pagetable layout for L0 if
> possible.
> +		 * Prefer the compact pagetable layout for L0 if
> possible. Only
> +		 * possible if VMA covers entire 2MB region as
> compact 64k and
> +		 * 4k pages cannot be mixed within a 2MB region.
>  		 * TODO: Suballocate the pt bo to avoid wasting a
> lot of
>  		 * memory.
>  		 */
>  		if (GRAPHICS_VERx100(tile_to_xe(xe_walk->tile)) >=
> 1250 && level == 1 &&
>  		    covers && xe_pt_scan_64K(addr, next, xe_walk)) {
>  			walk->shifts = xe_compact_pt_shifts;
> +			xe_walk->vma->gpuva.flags |=
> XE_VMA_PTE_COMPACT;
>  			flags |= XE_PDE_64K;
>  			xe_child->is_compact = true;
>  		}
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 7781290d2b99..23a44ef85aa4 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -2190,7 +2190,7 @@ static u64 xe_vma_max_pte_size(struct xe_vma
> *vma)
>  {
>  	if (vma->gpuva.flags & XE_VMA_PTE_1G)
>  		return SZ_1G;
> -	else if (vma->gpuva.flags & XE_VMA_PTE_2M)
> +	else if (vma->gpuva.flags & (XE_VMA_PTE_2M |
> XE_VMA_PTE_COMPACT))
>  		return SZ_2M;
>  	else if (vma->gpuva.flags & XE_VMA_PTE_64K)
>  		return SZ_64K;
> diff --git a/drivers/gpu/drm/xe/xe_vm_types.h
> b/drivers/gpu/drm/xe/xe_vm_types.h
> index 91800ce70845..a603cc2eb56b 100644
> --- a/drivers/gpu/drm/xe/xe_vm_types.h
> +++ b/drivers/gpu/drm/xe/xe_vm_types.h
> @@ -30,6 +30,7 @@ struct xe_vm;
>  #define XE_VMA_PTE_2M		(DRM_GPUVA_USERBITS << 6)
>  #define XE_VMA_PTE_1G		(DRM_GPUVA_USERBITS << 7)
>  #define XE_VMA_PTE_64K		(DRM_GPUVA_USERBITS << 8)
> +#define XE_VMA_PTE_COMPACT	(DRM_GPUVA_USERBITS << 9)
>  
>  /** struct xe_userptr - User pointer */
>  struct xe_userptr {