[PATCH] batched freeing of anon pages

A reworked version of the batched page freeing and lock amortisation
for VMA teardown.

It walks the existing 507-page list in the mmu_gather_t in 16-page
chunks, drops their refcounts in 16-page chunks, and de-LRUs and
frees any resulting zero-count pages in up-to-16 page chunks.

include/asm-generic/tlb.h

@@ -79,10 +79,8 @@ static inline void tlb_flush_mmu(mmu_gather_t *tlb, unsigned long start, unsigne
 	tlb_flush(tlb);
 	nr = tlb->nr;
 	if (!tlb_fast_mode(tlb)) {
-		unsigned long i;
+		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
-		for (i=0; i < nr; i++)
-			free_page_and_swap_cache(tlb->pages[i]);
 	}
 }
 

include/linux/pagemap.h

@@ -24,6 +24,7 @@
 
 #define page_cache_get(page)		get_page(page)
 #define page_cache_release(page)	put_page(page)
+void release_pages(struct page **pages, int nr);
 
 static inline struct page *page_cache_alloc(struct address_space *x)
 {

include/linux/swap.h

@@ -182,6 +182,7 @@ extern int move_to_swap_cache(struct page *page, swp_entry_t entry);
 extern int move_from_swap_cache(struct page *page, unsigned long index,
 		struct address_space *mapping);
 extern void free_page_and_swap_cache(struct page *page);
+extern void free_pages_and_swap_cache(struct page **pages, int nr);
 extern struct page * lookup_swap_cache(swp_entry_t);
 extern struct page * read_swap_cache_async(swp_entry_t);
 

mm/swap.c

@@ -87,7 +87,7 @@ void __page_cache_release(struct page *page)
 
 /*
  * Batched page_cache_release().  Decrement the reference count on all the
- * pagevec's pages.  If it fell to zero then remove the page from the LRU and
+ * passed pages.  If it fell to zero then remove the page from the LRU and
  * free it.
  *
  * Avoid taking zone->lru_lock if possible, but if it is taken, retain it
@@ -96,18 +96,16 @@ void __page_cache_release(struct page *page)
  * The locking in this function is against shrink_cache(): we recheck the
  * page count inside the lock to see whether shrink_cache grabbed the page
  * via the LRU.  If it did, give up: shrink_cache will free it.
- *
- * This function reinitialises the caller's pagevec.
  */
-void __pagevec_release(struct pagevec *pvec)
+void release_pages(struct page **pages, int nr)
 {
 	int i;
 	struct pagevec pages_to_free;
 	struct zone *zone = NULL;
 
 	pagevec_init(&pages_to_free);
-	for (i = 0; i < pagevec_count(pvec); i++) {
-		struct page *page = pvec->pages[i];
+	for (i = 0; i < nr; i++) {
+		struct page *page = pages[i];
 		struct zone *pagezone;
 
 		if (PageReserved(page) || !put_page_testzero(page))
@@ -122,13 +120,24 @@ void __pagevec_release(struct pagevec *pvec)
 		}
 		if (TestClearPageLRU(page))
 			del_page_from_lru(zone, page);
-		if (page_count(page) == 0)
-			pagevec_add(&pages_to_free, page);
+		if (page_count(page) == 0) {
+			if (!pagevec_add(&pages_to_free, page)) {
+				spin_unlock_irq(&zone->lru_lock);
+				pagevec_free(&pages_to_free);
+				pagevec_init(&pages_to_free);
+				spin_lock_irq(&zone->lru_lock);
+			}
+		}
 	}
 	if (zone)
 		spin_unlock_irq(&zone->lru_lock);
 
 	pagevec_free(&pages_to_free);
+}
+
+void __pagevec_release(struct pagevec *pvec)
+{
+	release_pages(pvec->pages, pagevec_count(pvec));
 	pagevec_init(pvec);
 }
 

mm/swap_state.c

@@ -292,28 +292,55 @@ int move_from_swap_cache(struct page *page, unsigned long index,
 	return err;
 }
 
+
 /* 
- * Perform a free_page(), also freeing any swap cache associated with
- * this page if it is the last user of the page. Can not do a lock_page,
- * as we are holding the page_table_lock spinlock.
+ * If we are the only user, then try to free up the swap cache. 
+ * 
+ * Its ok to check for PageSwapCache without the page lock
+ * here because we are going to recheck again inside 
+ * exclusive_swap_page() _with_ the lock. 
+ * 					- Marcelo
  */
-void free_page_and_swap_cache(struct page *page)
+static inline void free_swap_cache(struct page *page)
 {
-	/* 
-	 * If we are the only user, then try to free up the swap cache. 
-	 * 
-	 * Its ok to check for PageSwapCache without the page lock
-	 * here because we are going to recheck again inside 
-	 * exclusive_swap_page() _with_ the lock. 
-	 * 					- Marcelo
-	 */
 	if (PageSwapCache(page) && !TestSetPageLocked(page)) {
 		remove_exclusive_swap_page(page);
 		unlock_page(page);
 	}
+}
+
+/* 
+ * Perform a free_page(), also freeing any swap cache associated with
+ * this page if it is the last user of the page. Can not do a lock_page,
+ * as we are holding the page_table_lock spinlock.
+ */
+void free_page_and_swap_cache(struct page *page)
+{
+	free_swap_cache(page);
 	page_cache_release(page);
 }
 
+/*
+ * Passed an array of pages, drop them all from swapcache and then release
+ * them.  They are removed from the LRU and freed if this is their last use.
+ */
+void free_pages_and_swap_cache(struct page **pages, int nr)
+{
+	const int chunk = 16;
+	struct page **pagep = pages;
+
+	while (nr) {
+		int todo = min(chunk, nr);
+		int i;
+
+		for (i = 0; i < todo; i++)
+			free_swap_cache(pagep[i]);
+		release_pages(pagep, todo);
+		pagep += todo;
+		nr -= todo;
+	}
+}
+
 /*
  * Lookup a swap entry in the swap cache. A found page will be returned
  * unlocked and with its refcount incremented - we rely on the kernel