diff mbox series

[4/7] khugepaged: Allow to callapse a page shared across fork

Message ID 20200327170601.18563-5-kirill.shutemov@linux.intel.com
State New, archived
Headers show
Series thp/khugepaged improvements and CoW semantics | expand

Commit Message

Kirill A. Shutemov March 27, 2020, 5:05 p.m. UTC
The page can be included into collapse as long as it doesn't have extra
pins (from GUP or otherwise).

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
---
 mm/khugepaged.c | 28 ++++++++++++++++------------
 1 file changed, 16 insertions(+), 12 deletions(-)

Comments

Zi Yan March 27, 2020, 6:19 p.m. UTC | #1
On 27 Mar 2020, at 13:05, Kirill A. Shutemov wrote:

> The page can be included into collapse as long as it doesn't have extra
> pins (from GUP or otherwise).
>
> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> ---
>  mm/khugepaged.c | 28 ++++++++++++++++------------
>  1 file changed, 16 insertions(+), 12 deletions(-)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 39e0994abeb8..b47edfe57f7b 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -581,18 +581,26 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>  		}
>
>  		/*
> -		 * cannot use mapcount: can't collapse if there's a gup pin.
> -		 * The page must only be referenced by the scanned process
> -		 * and page swap cache.
> +		 * Check if the page has any GUP (or other external) pins.
> +		 *
> +		 * The page table that maps the page has been already unlinked
> +		 * from the page table tree and this process cannot get
> +		 * additinal pin on the page.
> +		 *
> +		 * New pins can come later if the page is shared across fork,
> +		 * but not for the this process. It is fine. The other process
> +		 * cannot write to the page, only trigger CoW.
>  		 */
> -		if (page_count(page) != 1 + PageSwapCache(page)) {
> +		if (total_mapcount(page) + PageSwapCache(page) !=
> +				page_count(page)) {

Do you think having a function for this check would be better? Since the check is used three times.

>  			/*
>  			 * Drain pagevec and retry just in case we can get rid
>  			 * of the extra pin, like in swapin case.
>  			 */
>  			lru_add_drain();
>  		}
> -		if (page_count(page) != 1 + PageSwapCache(page)) {
> +		if (total_mapcount(page) + PageSwapCache(page) !=
> +				page_count(page)) {
>  			unlock_page(page);
>  			result = SCAN_PAGE_COUNT;
>  			goto out;
> @@ -680,7 +688,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>  		} else {
>  			src_page = pte_page(pteval);
>  			copy_user_highpage(page, src_page, address, vma);
> -			VM_BUG_ON_PAGE(page_mapcount(src_page) != 1, src_page);

Maybe replace it with this?

VM_BUG_ON_PAGE(page_mapcount(src_page) + PageSwapCache(src_page) != page_count(src_page), src_page);


>  			release_pte_page(src_page);
>  			/*
>  			 * ptl mostly unnecessary, but preempt has to
> @@ -1209,12 +1216,9 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
>  			goto out_unmap;
>  		}
>
> -		/*
> -		 * cannot use mapcount: can't collapse if there's a gup pin.
> -		 * The page must only be referenced by the scanned process
> -		 * and page swap cache.
> -		 */
> -		if (page_count(page) != 1 + PageSwapCache(page)) {
> +		/* Check if the page has any GUP (or other external) pins */
> +		if (total_mapcount(page) + PageSwapCache(page) !=
> +				page_count(page)) {
>  			result = SCAN_PAGE_COUNT;
>  			goto out_unmap;
>  		}
> -- 
> 2.26.0

Thanks.

—
Best Regards,
Yan Zi
Yang Shi March 27, 2020, 9:31 p.m. UTC | #2
On Fri, Mar 27, 2020 at 11:20 AM Zi Yan <zi.yan@sent.com> wrote:
>
> On 27 Mar 2020, at 13:05, Kirill A. Shutemov wrote:
>
> > The page can be included into collapse as long as it doesn't have extra
> > pins (from GUP or otherwise).
> >
> > Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> > ---
> >  mm/khugepaged.c | 28 ++++++++++++++++------------
> >  1 file changed, 16 insertions(+), 12 deletions(-)
> >
> > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > index 39e0994abeb8..b47edfe57f7b 100644
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -581,18 +581,26 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
> >               }
> >
> >               /*
> > -              * cannot use mapcount: can't collapse if there's a gup pin.
> > -              * The page must only be referenced by the scanned process
> > -              * and page swap cache.
> > +              * Check if the page has any GUP (or other external) pins.
> > +              *
> > +              * The page table that maps the page has been already unlinked
> > +              * from the page table tree and this process cannot get
> > +              * additinal pin on the page.
> > +              *
> > +              * New pins can come later if the page is shared across fork,
> > +              * but not for the this process. It is fine. The other process
> > +              * cannot write to the page, only trigger CoW.
> >                */
> > -             if (page_count(page) != 1 + PageSwapCache(page)) {
> > +             if (total_mapcount(page) + PageSwapCache(page) !=
> > +                             page_count(page)) {
>
> Do you think having a function for this check would be better? Since the check is used three times.
>
> >                       /*
> >                        * Drain pagevec and retry just in case we can get rid
> >                        * of the extra pin, like in swapin case.
> >                        */
> >                       lru_add_drain();
> >               }
> > -             if (page_count(page) != 1 + PageSwapCache(page)) {
> > +             if (total_mapcount(page) + PageSwapCache(page) !=
> > +                             page_count(page)) {
> >                       unlock_page(page);
> >                       result = SCAN_PAGE_COUNT;
> >                       goto out;
> > @@ -680,7 +688,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> >               } else {
> >                       src_page = pte_page(pteval);
> >                       copy_user_highpage(page, src_page, address, vma);
> > -                     VM_BUG_ON_PAGE(page_mapcount(src_page) != 1, src_page);
>
> Maybe replace it with this?
>
> VM_BUG_ON_PAGE(page_mapcount(src_page) + PageSwapCache(src_page) != page_count(src_page), src_page);

I don't think this is correct either. If a THP is PTE mapped its
refcount would be bumped by the number of PTE mapped subpages. But
page_mapcount() would just return the mapcount of that specific
subpage. So, total_mapcount() should be used, but the same check has
been done before reaching here.

>
>
> >                       release_pte_page(src_page);
> >                       /*
> >                        * ptl mostly unnecessary, but preempt has to
> > @@ -1209,12 +1216,9 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
> >                       goto out_unmap;
> >               }
> >
> > -             /*
> > -              * cannot use mapcount: can't collapse if there's a gup pin.
> > -              * The page must only be referenced by the scanned process
> > -              * and page swap cache.
> > -              */
> > -             if (page_count(page) != 1 + PageSwapCache(page)) {
> > +             /* Check if the page has any GUP (or other external) pins */
> > +             if (total_mapcount(page) + PageSwapCache(page) !=
> > +                             page_count(page)) {
> >                       result = SCAN_PAGE_COUNT;
> >                       goto out_unmap;
> >               }
> > --
> > 2.26.0
>
> Thanks.
>
> —
> Best Regards,
> Yan Zi
Zi Yan March 27, 2020, 9:44 p.m. UTC | #3
<snip>
>>>                       /*
>>>                        * Drain pagevec and retry just in case we can get rid
>>>                        * of the extra pin, like in swapin case.
>>>                        */
>>>                       lru_add_drain();
>>>               }
>>> -             if (page_count(page) != 1 + PageSwapCache(page)) {
>>> +             if (total_mapcount(page) + PageSwapCache(page) !=
>>> +                             page_count(page)) {
>>>                       unlock_page(page);
>>>                       result = SCAN_PAGE_COUNT;
>>>                       goto out;
>>> @@ -680,7 +688,6 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>>>               } else {
>>>                       src_page = pte_page(pteval);
>>>                       copy_user_highpage(page, src_page, address, vma);
>>> -                     VM_BUG_ON_PAGE(page_mapcount(src_page) != 1, src_page);
>>
>> Maybe replace it with this?
>>
>> VM_BUG_ON_PAGE(page_mapcount(src_page) + PageSwapCache(src_page) != page_count(src_page), src_page);
>
> I don't think this is correct either. If a THP is PTE mapped its
> refcount would be bumped by the number of PTE mapped subpages. But
> page_mapcount() would just return the mapcount of that specific
> subpage. So, total_mapcount() should be used, but the same check has
> been done before reaching here.

Yes, you are right. Thanks.

Please disregard this comment.

—
Best Regards,
Yan Zi
diff mbox series

Patch

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 39e0994abeb8..b47edfe57f7b 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -581,18 +581,26 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 		}
 
 		/*
-		 * cannot use mapcount: can't collapse if there's a gup pin.
-		 * The page must only be referenced by the scanned process
-		 * and page swap cache.
+		 * Check if the page has any GUP (or other external) pins.
+		 *
+		 * The page table that maps the page has been already unlinked
+		 * from the page table tree and this process cannot get
+		 * additinal pin on the page.
+		 *
+		 * New pins can come later if the page is shared across fork,
+		 * but not for the this process. It is fine. The other process
+		 * cannot write to the page, only trigger CoW.
 		 */
-		if (page_count(page) != 1 + PageSwapCache(page)) {
+		if (total_mapcount(page) + PageSwapCache(page) !=
+				page_count(page)) {
 			/*
 			 * Drain pagevec and retry just in case we can get rid
 			 * of the extra pin, like in swapin case.
 			 */
 			lru_add_drain();
 		}
-		if (page_count(page) != 1 + PageSwapCache(page)) {
+		if (total_mapcount(page) + PageSwapCache(page) !=
+				page_count(page)) {
 			unlock_page(page);
 			result = SCAN_PAGE_COUNT;
 			goto out;
@@ -680,7 +688,6 @@  static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
 		} else {
 			src_page = pte_page(pteval);
 			copy_user_highpage(page, src_page, address, vma);
-			VM_BUG_ON_PAGE(page_mapcount(src_page) != 1, src_page);
 			release_pte_page(src_page);
 			/*
 			 * ptl mostly unnecessary, but preempt has to
@@ -1209,12 +1216,9 @@  static int khugepaged_scan_pmd(struct mm_struct *mm,
 			goto out_unmap;
 		}
 
-		/*
-		 * cannot use mapcount: can't collapse if there's a gup pin.
-		 * The page must only be referenced by the scanned process
-		 * and page swap cache.
-		 */
-		if (page_count(page) != 1 + PageSwapCache(page)) {
+		/* Check if the page has any GUP (or other external) pins */
+		if (total_mapcount(page) + PageSwapCache(page) !=
+				page_count(page)) {
 			result = SCAN_PAGE_COUNT;
 			goto out_unmap;
 		}