Re: [PATCH 1/2] mm/gup: pass an end address to follow_page_mask() and return a page count

[Rik van Riel]

On Wed, 2026-07-29 at 09:16 +0200, David Hildenbrand (Arm) wrote:
> 
> This patch moves us closer towards the GUP-fast path, however, that
> is about to
> change:
> 
> 	
> https://lore.kernel.org/all/20260219050250.266876166@xxxxxxxxx/
> 
> With that in mind, I guess in the future we want to have:
> 
> static long follow_page_mask(struct vm_area_struct *vma,
> 		unsigned long address, unsigned long end,
> 		unsigned int flags, struct page **pages)
> 
> or even better, now that follow_page() no longer exists, in light of
> gup_fast():
> 
> 
> static long gup_slow(struct vm_area_struct *vma,
> 		unsigned long address, unsigned long end,
> 		unsigned int flags, struct page **pages)
> 
> Returning the number of pinned pages.
> 
> 
> In particular, such an interface would allows us to process multiple
> ptes/pmds/... in one go, without having to restart the page table
> walk all the
> time for each non-batchable folio.
> 
Let me give that a try. This looks like a very
nice optimization, and cleanup.

-- 
All Rights Reversed.