[PATCH] mm/gup: speed up check_and_migrate_cma_pages() on huge page

[PATCH] mm/gup: speed up check_and_migrate_cma_pages() on huge page

[Pingfan Liu]

Both hugetlb and thp locate on the same migration type of pageblock, since
they are allocated from a free_list[]. Based on this fact, it is enough to
check on a single subpage to decide the migration type of the whole huge
page. By this way, it saves (2M/4K - 1) times loop for pmd_huge on x86,
similar on other archs.

Furthermore, when executing isolate_huge_page(), it avoid taking global
hugetlb_lock many times, and meanless remove/add to the local link list
cma_page_list.

Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Linux-kernel@vger.kernel.org
---
 mm/gup.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/mm/gup.c b/mm/gup.c
index ddde097..2eecb16 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1342,16 +1342,19 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
 	LIST_HEAD(cma_page_list);
 
 check_again:
-	for (i = 0; i < nr_pages; i++) {
+	for (i = 0; i < nr_pages;) {
+
+		struct page *head = compound_head(pages[i]);
+		long step = 1;
+
+		if (PageCompound(head))
+			step = compound_order(head) - (pages[i] - head);
 		/*
 		 * If we get a page from the CMA zone, since we are going to
 		 * be pinning these entries, we might as well move them out
 		 * of the CMA zone if possible.
 		 */
 		if (is_migrate_cma_page(pages[i])) {
-
-			struct page *head = compound_head(pages[i]);
-
 			if (PageHuge(head)) {
 				isolate_huge_page(head, &cma_page_list);
 			} else {
@@ -1369,6 +1372,8 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
 				}
 			}
 		}
+
+		i += step;
 	}
 
 	if (!list_empty(&cma_page_list)) {
-- 
2.7.5

Re: [PATCH] mm/gup: speed up check_and_migrate_cma_pages() on huge page

[Ira Weiny]

On Fri, Jun 21, 2019 at 06:15:16PM +0800, Pingfan Liu wrote:
> Both hugetlb and thp locate on the same migration type of pageblock, since
> they are allocated from a free_list[]. Based on this fact, it is enough to
> check on a single subpage to decide the migration type of the whole huge
> page. By this way, it saves (2M/4K - 1) times loop for pmd_huge on x86,
> similar on other archs.
> 
> Furthermore, when executing isolate_huge_page(), it avoid taking global
> hugetlb_lock many times, and meanless remove/add to the local link list
> cma_page_list.
> 
> Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Ira Weiny <ira.weiny@intel.com>
> Cc: Mike Rapoport <rppt@linux.ibm.com>
> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: John Hubbard <jhubbard@nvidia.com>
> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Keith Busch <keith.busch@intel.com>
> Cc: Mike Kravetz <mike.kravetz@oracle.com>
> Cc: Linux-kernel@vger.kernel.org
> ---
>  mm/gup.c | 13 +++++++++----
>  1 file changed, 9 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/gup.c b/mm/gup.c
> index ddde097..2eecb16 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -1342,16 +1342,19 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
>  	LIST_HEAD(cma_page_list);
>  
>  check_again:
> -	for (i = 0; i < nr_pages; i++) {
> +	for (i = 0; i < nr_pages;) {
> +
> +		struct page *head = compound_head(pages[i]);
> +		long step = 1;
> +
> +		if (PageCompound(head))
> +			step = compound_order(head) - (pages[i] - head);
>  		/*
>  		 * If we get a page from the CMA zone, since we are going to
>  		 * be pinning these entries, we might as well move them out
>  		 * of the CMA zone if possible.
>  		 */
>  		if (is_migrate_cma_page(pages[i])) {

I like this but I think for consistency I would change this pages[i] to be
head.  Even though it is not required.

Ira

> -
> -			struct page *head = compound_head(pages[i]);
> -
>  			if (PageHuge(head)) {
>  				isolate_huge_page(head, &cma_page_list);
>  			} else {
> @@ -1369,6 +1372,8 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
>  				}
>  			}
>  		}
> +
> +		i += step;
>  	}
>  
>  	if (!list_empty(&cma_page_list)) {
> -- 
> 2.7.5
> 

Re: [PATCH] mm/gup: speed up check_and_migrate_cma_pages() on huge page

[Pingfan Liu]

On Sat, Jun 22, 2019 at 2:13 AM Ira Weiny <ira.weiny@intel.com> wrote:
>
> On Fri, Jun 21, 2019 at 06:15:16PM +0800, Pingfan Liu wrote:
> > Both hugetlb and thp locate on the same migration type of pageblock, since
> > they are allocated from a free_list[]. Based on this fact, it is enough to
> > check on a single subpage to decide the migration type of the whole huge
> > page. By this way, it saves (2M/4K - 1) times loop for pmd_huge on x86,
> > similar on other archs.
> >
> > Furthermore, when executing isolate_huge_page(), it avoid taking global
> > hugetlb_lock many times, and meanless remove/add to the local link list
> > cma_page_list.
> >
> > Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
> > Cc: Andrew Morton <akpm@linux-foundation.org>
> > Cc: Ira Weiny <ira.weiny@intel.com>
> > Cc: Mike Rapoport <rppt@linux.ibm.com>
> > Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: John Hubbard <jhubbard@nvidia.com>
> > Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
> > Cc: Christoph Hellwig <hch@lst.de>
> > Cc: Keith Busch <keith.busch@intel.com>
> > Cc: Mike Kravetz <mike.kravetz@oracle.com>
> > Cc: Linux-kernel@vger.kernel.org
> > ---
> >  mm/gup.c | 13 +++++++++----
> >  1 file changed, 9 insertions(+), 4 deletions(-)
> >
> > diff --git a/mm/gup.c b/mm/gup.c
> > index ddde097..2eecb16 100644
> > --- a/mm/gup.c
> > +++ b/mm/gup.c
> > @@ -1342,16 +1342,19 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >       LIST_HEAD(cma_page_list);
> >
> >  check_again:
> > -     for (i = 0; i < nr_pages; i++) {
> > +     for (i = 0; i < nr_pages;) {
> > +
> > +             struct page *head = compound_head(pages[i]);
> > +             long step = 1;
> > +
> > +             if (PageCompound(head))
> > +                     step = compound_order(head) - (pages[i] - head);
> >               /*
> >                * If we get a page from the CMA zone, since we are going to
> >                * be pinning these entries, we might as well move them out
> >                * of the CMA zone if possible.
> >                */
> >               if (is_migrate_cma_page(pages[i])) {
>
> I like this but I think for consistency I would change this pages[i] to be
> head.  Even though it is not required.
Yes, agree. Thank you for your good suggestion.

Regards,
  Pingfan
>
> Ira
>
> > -
> > -                     struct page *head = compound_head(pages[i]);
> > -
> >                       if (PageHuge(head)) {
> >                               isolate_huge_page(head, &cma_page_list);
> >                       } else {
> > @@ -1369,6 +1372,8 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >                               }
> >                       }
> >               }
> > +
> > +             i += step;
> >       }
> >
> >       if (!list_empty(&cma_page_list)) {
> > --
> > 2.7.5
> >