diff mbox series

[resend,3/3] mm: fix LRU balancing effect of new transparent huge pages

Message ID 20200527182958.252402-1-shakeelb@google.com (mailing list archive)
State New, archived
Headers show
Series [resend,1/3] mm: swap: fix vmstats for huge pages | expand

Commit Message

Shakeel Butt May 27, 2020, 6:29 p.m. UTC 
From: Johannes Weiner <hannes@cmpxchg.org>

Currently, THP are counted as single pages until they are split right
before being swapped out. However, at that point the VM is already in
the middle of reclaim, and adjusting the LRU balance then is useless.

Always account THP by the number of basepages, and remove the fixup
from the splitting path.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Shakeel Butt <shakeelb@google.com>
---
 mm/swap.c | 23 +++++++++--------------
 1 file changed, 9 insertions(+), 14 deletions(-)

Comments

Johannes Weiner May 27, 2020, 7:41 p.m. UTC | #1 
On Wed, May 27, 2020 at 11:29:58AM -0700, Shakeel Butt wrote:
> From: Johannes Weiner <hannes@cmpxchg.org>
> 
> Currently, THP are counted as single pages until they are split right
> before being swapped out. However, at that point the VM is already in
> the middle of reclaim, and adjusting the LRU balance then is useless.
> 
> Always account THP by the number of basepages, and remove the fixup
> from the splitting path.
> 
> Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
> Signed-off-by: Shakeel Butt <shakeelb@google.com>

This is now already in mm as part of the "mm: balance LRU lists based
on relative thrashing" series that I sent out last week and where it
was originally from.
Shakeel Butt May 27, 2020, 7:46 p.m. UTC | #2 
On Wed, May 27, 2020 at 12:42 PM Johannes Weiner <hannes@cmpxchg.org> wrote:
>
> On Wed, May 27, 2020 at 11:29:58AM -0700, Shakeel Butt wrote:
> > From: Johannes Weiner <hannes@cmpxchg.org>
> >
> > Currently, THP are counted as single pages until they are split right
> > before being swapped out. However, at that point the VM is already in
> > the middle of reclaim, and adjusting the LRU balance then is useless.
> >
> > Always account THP by the number of basepages, and remove the fixup
> > from the splitting path.
> >
> > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
> > Signed-off-by: Shakeel Butt <shakeelb@google.com>
>
> This is now already in mm as part of the "mm: balance LRU lists based
> on relative thrashing" series that I sent out last week and where it
> was originally from.

Oh sorry, I missed it.
Andrew Morton May 27, 2020, 8:46 p.m. UTC | #3 
On Wed, 27 May 2020 15:41:48 -0400 Johannes Weiner <hannes@cmpxchg.org> wrote:

> On Wed, May 27, 2020 at 11:29:58AM -0700, Shakeel Butt wrote:
> > From: Johannes Weiner <hannes@cmpxchg.org>
> > 
> > Currently, THP are counted as single pages until they are split right
> > before being swapped out. However, at that point the VM is already in
> > the middle of reclaim, and adjusting the LRU balance then is useless.
> > 
> > Always account THP by the number of basepages, and remove the fixup
> > from the splitting path.
> > 
> > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
> > Signed-off-by: Shakeel Butt <shakeelb@google.com>
> 
> This is now already in mm as part of the "mm: balance LRU lists based
> on relative thrashing" series that I sent out last week and where it
> was originally from.

Yup.  I hope [1/3] and [2/3] weren't urgent?  Due to a horrid merge
mismatchmishmashmess I've staged them behind lots of other things,
notably

thp/khugepaged improvements and CoW semantics
http://lkml.kernel.org/r/20200416160026.16538-1-kirill.shutemov@linux.intel.com

and mm: memcontrol: charge swapin pages on instantiation
http://lkml.kernel.org/r/20200508183105.225460-1-hannes@cmpxchg.org

and mm: balance LRU lists based on relative thrashing
http://lkml.kernel.org/r/20200520232525.798933-1-hannes@cmpxchg.org
Shakeel Butt May 27, 2020, 9:35 p.m. UTC | #4 
On Wed, May 27, 2020 at 1:46 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Wed, 27 May 2020 15:41:48 -0400 Johannes Weiner <hannes@cmpxchg.org> wrote:
>
> > On Wed, May 27, 2020 at 11:29:58AM -0700, Shakeel Butt wrote:
> > > From: Johannes Weiner <hannes@cmpxchg.org>
> > >
> > > Currently, THP are counted as single pages until they are split right
> > > before being swapped out. However, at that point the VM is already in
> > > the middle of reclaim, and adjusting the LRU balance then is useless.
> > >
> > > Always account THP by the number of basepages, and remove the fixup
> > > from the splitting path.
> > >
> > > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
> > > Signed-off-by: Shakeel Butt <shakeelb@google.com>
> >
> > This is now already in mm as part of the "mm: balance LRU lists based
> > on relative thrashing" series that I sent out last week and where it
> > was originally from.
>
> Yup.  I hope [1/3] and [2/3] weren't urgent?

No urgency.
diff mbox series

Patch

diff --git a/mm/swap.c b/mm/swap.c
index 4eb179ee0b72..b75c0ce90418 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -262,14 +262,14 @@  void rotate_reclaimable_page(struct page *page)
 	}
 }
 
-static void update_page_reclaim_stat(struct lruvec *lruvec,
-				     int file, int rotated)
+static void update_page_reclaim_stat(struct lruvec *lruvec, int file,
+				     int rotated, int nr_pages)
 {
 	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
 
-	reclaim_stat->recent_scanned[file]++;
+	reclaim_stat->recent_scanned[file] += nr_pages;
 	if (rotated)
-		reclaim_stat->recent_rotated[file]++;
+		reclaim_stat->recent_rotated[file] += nr_pages;
 }
 
 static void __activate_page(struct page *page, struct lruvec *lruvec,
@@ -288,7 +288,7 @@  static void __activate_page(struct page *page, struct lruvec *lruvec,
 
 		__count_vm_events(PGACTIVATE, nr_pages);
 		__count_memcg_events(lruvec_memcg(lruvec), PGACTIVATE, nr_pages);
-		update_page_reclaim_stat(lruvec, file, 1);
+		update_page_reclaim_stat(lruvec, file, 1, nr_pages);
 	}
 }
 
@@ -546,7 +546,7 @@  static void lru_deactivate_file_fn(struct page *page, struct lruvec *lruvec,
 		__count_vm_events(PGDEACTIVATE, nr_pages);
 		__count_memcg_events(lruvec_memcg(lruvec), PGDEACTIVATE, nr_pages);
 	}
-	update_page_reclaim_stat(lruvec, file, 0);
+	update_page_reclaim_stat(lruvec, file, 0, nr_pages);
 }
 
 static void lru_deactivate_fn(struct page *page, struct lruvec *lruvec,
@@ -564,7 +564,7 @@  static void lru_deactivate_fn(struct page *page, struct lruvec *lruvec,
 
 		__count_vm_events(PGDEACTIVATE, nr_pages);
 		__count_memcg_events(lruvec_memcg(lruvec), PGDEACTIVATE, nr_pages);
-		update_page_reclaim_stat(lruvec, file, 0);
+		update_page_reclaim_stat(lruvec, file, 0, nr_pages);
 	}
 }
 
@@ -590,7 +590,7 @@  static void lru_lazyfree_fn(struct page *page, struct lruvec *lruvec,
 
 		__count_vm_events(PGLAZYFREE, nr_pages);
 		__count_memcg_events(lruvec_memcg(lruvec), PGLAZYFREE, nr_pages);
-		update_page_reclaim_stat(lruvec, 1, 0);
+		update_page_reclaim_stat(lruvec, 1, 0, nr_pages);
 	}
 }
 
@@ -899,8 +899,6 @@  EXPORT_SYMBOL(__pagevec_release);
 void lru_add_page_tail(struct page *page, struct page *page_tail,
 		       struct lruvec *lruvec, struct list_head *list)
 {
-	const int file = 0;
-
 	VM_BUG_ON_PAGE(!PageHead(page), page);
 	VM_BUG_ON_PAGE(PageCompound(page_tail), page);
 	VM_BUG_ON_PAGE(PageLRU(page_tail), page);
@@ -926,9 +924,6 @@  void lru_add_page_tail(struct page *page, struct page *page_tail,
 		add_page_to_lru_list_tail(page_tail, lruvec,
 					  page_lru(page_tail));
 	}
-
-	if (!PageUnevictable(page))
-		update_page_reclaim_stat(lruvec, file, PageActive(page_tail));
 }
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 
@@ -973,7 +968,7 @@  static void __pagevec_lru_add_fn(struct page *page, struct lruvec *lruvec,
 	if (page_evictable(page)) {
 		lru = page_lru(page);
 		update_page_reclaim_stat(lruvec, page_is_file_lru(page),
-					 PageActive(page));
+					 PageActive(page), nr_pages);
 		if (was_unevictable)
 			__count_vm_events(UNEVICTABLE_PGRESCUED, nr_pages);
 	} else {