Re: [PATCH v2 3/9] mm/hugetlb: make pud_huge() and huge_pud() aware of non-present pud entry

[Miaohe Lin <linmiaohe@huawei.com>]

On 2022/6/24 7:51, Naoya Horiguchi wrote:
> From: Naoya Horiguchi <naoya.horiguchi@nec.com>
> 
> follow_pud_mask() does not support non-present pud entry now.  As long as
> I tested on x86_64 server, follow_pud_mask() still simply returns
> no_page_table() for non-present_pud_entry() due to pud_bad(), so no severe
> user-visible effect should happen.  But generally we should call
> follow_huge_pud() for non-present pud entry for 1GB hugetlb page.
> 
> Update pud_huge() and huge_pud() to handle non-present pud entries.  The
> changes are similar to previous works for pud entries commit e66f17ff7177
> ("mm/hugetlb: take page table lock in follow_huge_pmd()") and commit
> cbef8478bee5 ("mm/hugetlb: pmd_huge() returns true for non-present hugepage").
> 
> Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
> ---
>  arch/x86/mm/hugetlbpage.c |  3 ++-
>  mm/hugetlb.c              | 26 +++++++++++++++++++++++++-
>  2 files changed, 27 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/mm/hugetlbpage.c b/arch/x86/mm/hugetlbpage.c
> index a0d023cb4292..5fb86fb49ba8 100644
> --- a/arch/x86/mm/hugetlbpage.c
> +++ b/arch/x86/mm/hugetlbpage.c
> @@ -70,7 +70,8 @@ int pmd_huge(pmd_t pmd)
>  

No strong opinion but a comment similar to pmd_huge might be better?

/*
 * pmd_huge() returns 1 if @pmd is hugetlb related entry, that is normal
 * hugetlb entry or non-present (migration or hwpoisoned) hugetlb entry.
 * Otherwise, returns 0.
 */

>  int pud_huge(pud_t pud)
>  {
> -	return !!(pud_val(pud) & _PAGE_PSE);
> +	return !pud_none(pud) &&
> +		(pud_val(pud) & (_PAGE_PRESENT|_PAGE_PSE)) != _PAGE_PRESENT;
>  }
>  #endif
>  
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index f59f43c06601..b7ae5f73f3b2 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -6946,10 +6946,34 @@ struct page * __weak
>  follow_huge_pud(struct mm_struct *mm, unsigned long address,
>  		pud_t *pud, int flags)
>  {
> +	struct page *page = NULL;
> +	spinlock_t *ptl;
> +	pte_t pte;
> +
>  	if (flags & (FOLL_GET | FOLL_PIN))
>  		return NULL;

Should the above check be modified? It seems the below try_grab_page might not grab the page as
expected (as Mike pointed out). Or the extra page refcnt is unneeded?

>  
> -	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
> +retry:
> +	ptl = huge_pte_lock(hstate_sizelog(PUD_SHIFT), mm, (pte_t *)pud);
> +	if (!pud_huge(*pud))
> +		goto out;
> +	pte = huge_ptep_get((pte_t *)pud);
> +	if (pte_present(pte)) {
> +		page = pud_page(*pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
> +		if (WARN_ON_ONCE(!try_grab_page(page, flags))) {
> +			page = NULL;
> +			goto out;
> +		}
> +	} else {
> +		if (is_hugetlb_entry_migration(pte)) {
> +			spin_unlock(ptl);
> +			__migration_entry_wait(mm, (pte_t *)pud, ptl);
> +			goto retry;
> +		}

Again. No strong opinion but a comment similar to follow_huge_pmd might be better?

/*
 * hwpoisoned entry is treated as no_page_table in
 * follow_page_mask().
 */

Thanks!

> +	}
> +out:
> +	spin_unlock(ptl);
> +	return page;
>  }
>  
>  struct page * __weak
>