[PATCH v6 11/26] mm: Allow compound zone device pages

[David Hildenbrand]

On 10.01.25 07:00, Alistair Popple wrote:
> Zone device pages are used to represent various type of device memory
> managed by device drivers. Currently compound zone device pages are
> not supported. This is because MEMORY_DEVICE_FS_DAX pages are the only
> user of higher order zone device pages and have their own page
> reference counting.
> 
> A future change will unify FS DAX reference counting with normal page
> reference counting rules and remove the special FS DAX reference
> counting. Supporting that requires compound zone device pages.
> 
> Supporting compound zone device pages requires compound_head() to
> distinguish between head and tail pages whilst still preserving the
> special struct page fields that are specific to zone device pages.
> 
> A tail page is distinguished by having bit zero being set in
> page->compound_head, with the remaining bits pointing to the head
> page. For zone device pages page->compound_head is shared with
> page->pgmap.
> 
> The page->pgmap field is common to all pages within a memory section.
> Therefore pgmap is the same for both head and tail pages and can be
> moved into the folio and we can use the standard scheme to find
> compound_head from a tail page.

The more relevant thing is that the pgmap field must be common to all 
pages in a folio, even if a folio exceeds memory sections (e.g., 128 MiB 
on x86_64 where we have 1 GiB folios).

 > > Signed-off-by: Alistair Popple <apopple at nvidia.com>
> Reviewed-by: Jason Gunthorpe <jgg at nvidia.com>
> Reviewed-by: Dan Williams <dan.j.williams at intel.com>
> 
> ---
> 
> Changes for v4:
>   - Fix build breakages reported by kernel test robot
> 
> Changes since v2:
> 
>   - Indentation fix
>   - Rename page_dev_pagemap() to page_pgmap()
>   - Rename folio _unused field to _unused_pgmap_compound_head
>   - s/WARN_ON/VM_WARN_ON_ONCE_PAGE/
> 
> Changes since v1:
> 
>   - Move pgmap to the folio as suggested by Matthew Wilcox
> ---

[...]

>   static inline bool folio_is_device_coherent(const struct folio *folio)
> diff --git a/include/linux/migrate.h b/include/linux/migrate.h
> index 29919fa..61899ec 100644
> --- a/include/linux/migrate.h
> +++ b/include/linux/migrate.h
> @@ -205,8 +205,8 @@ struct migrate_vma {
>   	unsigned long		end;
>   
>   	/*
> -	 * Set to the owner value also stored in page->pgmap->owner for
> -	 * migrating out of device private memory. The flags also need to
> +	 * Set to the owner value also stored in page_pgmap(page)->owner
> +	 * for migrating out of device private memory. The flags also need to
>   	 * be set to MIGRATE_VMA_SELECT_DEVICE_PRIVATE.
>   	 * The caller should always set this field when using mmu notifier
>   	 * callbacks to avoid device MMU invalidations for device private
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index df8f515..54b59b8 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -129,8 +129,11 @@ struct page {
>   			unsigned long compound_head;	/* Bit zero is set */
>   		};
>   		struct {	/* ZONE_DEVICE pages */
> -			/** @pgmap: Points to the hosting device page map. */
> -			struct dev_pagemap *pgmap;
> +			/*
> +			 * The first word is used for compound_head or folio
> +			 * pgmap
> +			 */
> +			void *_unused_pgmap_compound_head;
>   			void *zone_device_data;
>   			/*
>   			 * ZONE_DEVICE private pages are counted as being
> @@ -299,6 +302,7 @@ typedef struct {
>    * @_refcount: Do not access this member directly.  Use folio_ref_count()
>    *    to find how many references there are to this folio.
>    * @memcg_data: Memory Control Group data.
> + * @pgmap: Metadata for ZONE_DEVICE mappings
>    * @virtual: Virtual address in the kernel direct map.
>    * @_last_cpupid: IDs of last CPU and last process that accessed the folio.
>    * @_entire_mapcount: Do not use directly, call folio_entire_mapcount().
> @@ -337,6 +341,7 @@ struct folio {
>   	/* private: */
>   				};
>   	/* public: */
> +				struct dev_pagemap *pgmap;

Agreed, that should work.

Acked-by: David Hildenbrand <david at redhat.com>

-- 
Cheers,

David / dhildenb