Re: [PATCH] mm: Compute mTHP order efficiently

[Barry Song <baohua@kernel.org>]

On Fri, Sep 13, 2024 at 5:19 PM Dev Jain <dev.jain@arm.com> wrote:
>
> We use pte_range_none() to determine whether contiguous PTEs are empty
> for an mTHP allocation. Instead of iterating the while loop for every
> order, use some information, which is the first set PTE found, from the
> previous iteration, to eliminate some cases. The key to understanding
> the correctness of the patch is that the ranges we want to examine
> form a strictly decreasing sequence of nested intervals.
>
> Suggested-by: Ryan Roberts <ryan.roberts@arm.com>
> Signed-off-by: Dev Jain <dev.jain@arm.com>

I like this patch, but could we come up with a better subject for
pte_range_none()?
The subject is really incorrect.

Also, I'd prefer the change for alloc_anon_folio() to be separated
into its own patch.
So, one patchset with two patches, please.

> ---
>  mm/memory.c | 30 +++++++++++++++++++++++-------
>  1 file changed, 23 insertions(+), 7 deletions(-)
>
> diff --git a/mm/memory.c b/mm/memory.c
> index 3c01d68065be..ffc24a48ef15 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -4409,26 +4409,27 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>         return ret;
>  }
>
> -static bool pte_range_none(pte_t *pte, int nr_pages)
> +static int pte_range_none(pte_t *pte, int nr_pages)
>  {
>         int i;
>
>         for (i = 0; i < nr_pages; i++) {
>                 if (!pte_none(ptep_get_lockless(pte + i)))
> -                       return false;
> +                       return i;
>         }
>
> -       return true;
> +       return nr_pages;
>  }
>
>  static struct folio *alloc_anon_folio(struct vm_fault *vmf)
>  {
>         struct vm_area_struct *vma = vmf->vma;
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +       pte_t *first_set_pte = NULL, *align_pte, *pte;
>         unsigned long orders;
>         struct folio *folio;
>         unsigned long addr;
> -       pte_t *pte;
> +       int max_empty;
>         gfp_t gfp;
>         int order;
>
> @@ -4463,8 +4464,23 @@ static struct folio *alloc_anon_folio(struct vm_fault *vmf)
>         order = highest_order(orders);
>         while (orders) {
>                 addr = ALIGN_DOWN(vmf->address, PAGE_SIZE << order);
> -               if (pte_range_none(pte + pte_index(addr), 1 << order))
> +               align_pte = pte + pte_index(addr);
> +
> +               /* Range to be scanned known to be empty */
> +               if (align_pte + (1 << order) <= first_set_pte)
>                         break;
> +
> +               /* Range to be scanned contains first_set_pte */
> +               if (align_pte <= first_set_pte)
> +                       goto repeat;
> +
> +               /* align_pte > first_set_pte, so need to check properly */
> +               max_empty = pte_range_none(align_pte, 1 << order);
> +               if (max_empty == 1 << order)
> +                       break;
> +
> +               first_set_pte = align_pte + max_empty;
> +repeat:
>                 order = next_order(&orders, order);
>         }
>
> @@ -4579,7 +4595,7 @@ static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
>         if (nr_pages == 1 && vmf_pte_changed(vmf)) {
>                 update_mmu_tlb(vma, addr, vmf->pte);
>                 goto release;
> -       } else if (nr_pages > 1 && !pte_range_none(vmf->pte, nr_pages)) {
> +       } else if (nr_pages > 1 && pte_range_none(vmf->pte, nr_pages) != nr_pages) {
>                 update_mmu_tlb_range(vma, addr, vmf->pte, nr_pages);
>                 goto release;
>         }
> @@ -4915,7 +4931,7 @@ vm_fault_t finish_fault(struct vm_fault *vmf)
>                 update_mmu_tlb(vma, addr, vmf->pte);
>                 ret = VM_FAULT_NOPAGE;
>                 goto unlock;
> -       } else if (nr_pages > 1 && !pte_range_none(vmf->pte, nr_pages)) {
> +       } else if (nr_pages > 1 && pte_range_none(vmf->pte, nr_pages) != nr_pages) {
>                 update_mmu_tlb_range(vma, addr, vmf->pte, nr_pages);
>                 ret = VM_FAULT_NOPAGE;
>                 goto unlock;
> --
> 2.30.2
>

Thanks
Barry