[PATCH] mm/mlock: abstract a function to decide whether necessary to drain mlock_pvec

[PATCH] mm/mlock: abstract a function to decide whether necessary to drain mlock_pvec

[Wei Yang]

Just like the function pagevec_add_and_need_flush does, we can abstract
a function mlock_add_and_need_flush to decide whether it is necessary to
drain mlock_pvec.

No functional change is expected.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
CC: Matthew Wilcox (Oracle) <willy@infradead.org>
CC: Hugh Dickins <hughd@google.com>
---
 mm/mlock.c | 27 +++++++++++++++++++++------
 1 file changed, 21 insertions(+), 6 deletions(-)

diff --git a/mm/mlock.c b/mm/mlock.c
index efd2dd2943de..26125d733934 100644
--- a/mm/mlock.c
+++ b/mm/mlock.c
@@ -169,6 +169,24 @@ static inline struct page *mlock_new(struct page *page)
 	return (struct page *)((unsigned long)page + NEW_PAGE);
 }
 
+static inline struct page *mlock_dummy(struct page *page)
+{
+	return page;
+}
+
+/* return true if mlock_pvec needs to drain */
+static bool mlock_add_and_need_flush(struct pagevec *pvec, struct folio *folio,
+		struct page *(*mlock)(struct page *page))
+{
+	bool ret = false;
+
+	if (!pagevec_add(pvec, mlock(&folio->page)) ||
+			folio_test_large(folio) || lru_cache_disabled())
+		ret = true;
+
+	return ret;
+}
+
 /*
  * mlock_pagevec() is derived from pagevec_lru_move_fn():
  * perhaps that can make use of such page pointer flags in future,
@@ -233,8 +251,7 @@ void mlock_folio(struct folio *folio)
 	}
 
 	folio_get(folio);
-	if (!pagevec_add(pvec, mlock_lru(&folio->page)) ||
-	    folio_test_large(folio) || lru_cache_disabled())
+	if (mlock_add_and_need_flush(pvec, folio, mlock_lru))
 		mlock_pagevec(pvec);
 	put_cpu_var(mlock_pvec);
 }
@@ -253,8 +270,7 @@ void mlock_new_page(struct page *page)
 	__count_vm_events(UNEVICTABLE_PGMLOCKED, nr_pages);
 
 	get_page(page);
-	if (!pagevec_add(pvec, mlock_new(page)) ||
-	    PageHead(page) || lru_cache_disabled())
+	if (mlock_add_and_need_flush(pvec, page_folio(page), mlock_new))
 		mlock_pagevec(pvec);
 	put_cpu_var(mlock_pvec);
 }
@@ -273,8 +289,7 @@ void munlock_page(struct page *page)
 	 */
 
 	get_page(page);
-	if (!pagevec_add(pvec, page) ||
-	    PageHead(page) || lru_cache_disabled())
+	if (mlock_add_and_need_flush(pvec, page_folio(page), mlock_dummy))
 		mlock_pagevec(pvec);
 	put_cpu_var(mlock_pvec);
 }
-- 
2.33.1

Re: [PATCH] mm/mlock: abstract a function to decide whether necessary to drain mlock_pvec

[Matthew Wilcox]

On Fri, Apr 22, 2022 at 02:20:59AM +0000, Wei Yang wrote:
> Just like the function pagevec_add_and_need_flush does, we can abstract
> a function mlock_add_and_need_flush to decide whether it is necessary to
> drain mlock_pvec.
> 
> No functional change is expected.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> CC: Matthew Wilcox (Oracle) <willy@infradead.org>
> CC: Hugh Dickins <hughd@google.com>
> ---
>  mm/mlock.c | 27 +++++++++++++++++++++------
>  1 file changed, 21 insertions(+), 6 deletions(-)

This seems like a needless abstraction to me that makes understanding
the code flow harder, not simpler.