vmscan: free swap space on swap-in/activation

If vm_swap_full() (swap space more than 50% full), the system will free swap space at swapin time. With this patch, the system will also free the swap space in the pageout code, when we decide that the page is not a candidate for swapout (and just wasting swap space). Signed-off-by: Rik van Riel <riel@redhat.com> Signed-off-by: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Signed-off-by: MinChan Kim <minchan.kim@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vmscan: free swap space on swap-in/activation
If vm_swap_full() (swap space more than 50% full), the system will free swap space at swapin time. With this patch, the system will also free the swap space in the pageout code, when we decide that the page is not a candidate for swapout (and just wasting swap space). Signed-off-by: Rik van Riel <riel@redhat.com> Signed-off-by: Lee Schermerhorn <Lee.Schermerhorn@hp.com> Signed-off-by: MinChan Kim <minchan.kim@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rik van Riel · Linus Torvalds
1 parent f04e9ebbe4
Showing 5 changed files with 60 additions and 3 deletions Side-by-side Diff
include/linux/pagevec.h
include/linux/swap.h
mm/swap.c
mm/swapfile.c
mm/vmscan.c
@@ -25,6 +25,7 @@
 void __pagevec_free(struct pagevec *pvec);
 void ____pagevec_lru_add(struct pagevec *pvec, enum lru_list lru);
 void pagevec_strip(struct pagevec *pvec);
+void pagevec_swap_free(struct pagevec *pvec);
 unsigned pagevec_lookup(struct pagevec *pvec, struct address_space *mapping,
 		pgoff_t start, unsigned nr_pages);
 unsigned pagevec_lookup_tag(struct pagevec *pvec,
@@ -265,6 +265,7 @@
 extern struct swap_info_struct *get_swap_info_struct(unsigned);
 extern int can_share_swap_page(struct page *);
 extern int remove_exclusive_swap_page(struct page *);
+extern int remove_exclusive_swap_page_ref(struct page *);
 struct backing_dev_info;
  
 /* linux/mm/thrash.c */
@@ -349,6 +350,11 @@
 #define swap_token_default_timeout		0
  
 static inline int remove_exclusive_swap_page(struct page *p)
+{
+	return 0;
+}
+
+static inline int remove_exclusive_swap_page_ref(struct page *page)
 {
 	return 0;
 }
@@ -428,6 +428,30 @@
 }
  
 /**
+ * pagevec_swap_free - try to free swap space from the pages in a pagevec
+ * @pvec: pagevec with swapcache pages to free the swap space of
+ *
+ * The caller needs to hold an extra reference to each page and
+ * not hold the page lock on the pages.  This function uses a
+ * trylock on the page lock so it may not always free the swap
+ * space associated with a page.
+ */
+void pagevec_swap_free(struct pagevec *pvec)
+{
+	int i;
+
+	for (i = 0; i < pagevec_count(pvec); i++) {
+		struct page *page = pvec->pages[i];
+
+		if (PageSwapCache(page) && trylock_page(page)) {
+			if (PageSwapCache(page))
+				remove_exclusive_swap_page_ref(page);
+			unlock_page(page);
+		}
+	}
+}
+
+/**
  * pagevec_lookup - gang pagecache lookup
  * @pvec:	Where the resulting pages are placed
  * @mapping:	The address_space to search
@@ -344,7 +344,7 @@
  * Work out if there are any other processes sharing this
  * swap cache page. Free it if you can. Return success.
  */
-int remove_exclusive_swap_page(struct page *page)
+static int remove_exclusive_swap_page_count(struct page *page, int count)
 {
 	int retval;
 	struct swap_info_struct * p;
@@ -357,7 +357,7 @@
 		return 0;
 	if (PageWriteback(page))
 		return 0;
-	if (page_count(page) != 2) /* 2: us + cache */
+	if (page_count(page) != count) /* us + cache + ptes */
 		return 0;
  
 	entry.val = page_private(page);
@@ -370,7 +370,7 @@
 	if (p->swap_map[swp_offset(entry)] == 1) {
 		/* Recheck the page count with the swapcache lock held.. */
 		spin_lock_irq(&swapper_space.tree_lock);
-		if ((page_count(page) == 2) && !PageWriteback(page)) {
+		if ((page_count(page) == count) && !PageWriteback(page)) {
 			__delete_from_swap_cache(page);
 			SetPageDirty(page);
 			retval = 1;
@@ -385,6 +385,25 @@
 	}
  
 	return retval;
+}
+
+/*
+ * Most of the time the page should have two references: one for the
+ * process and one for the swap cache.
+ */
+int remove_exclusive_swap_page(struct page *page)
+{
+	return remove_exclusive_swap_page_count(page, 2);
+}
+
+/*
+ * The pageout code holds an extra reference to the page.  That raises
+ * the reference count to test for to 2 for a page that is only in the
+ * swap cache plus 1 for each process that maps the page.
+ */
+int remove_exclusive_swap_page_ref(struct page *page)
+{
+	return remove_exclusive_swap_page_count(page, 2 + page_mapcount(page));
 }
  
 /*
@@ -647,6 +647,9 @@
 		continue;
  
 activate_locked:
+		/* Not a candidate for swapping, so reclaim swap space. */
+		if (PageSwapCache(page) && vm_swap_full())
+			remove_exclusive_swap_page_ref(page);
 		SetPageActive(page);
 		pgactivate++;
 keep_locked:
@@ -1228,6 +1231,8 @@
 			__mod_zone_page_state(zone, NR_ACTIVE, pgmoved);
 			pgmoved = 0;
 			spin_unlock_irq(&zone->lru_lock);
+			if (vm_swap_full())
+				pagevec_swap_free(&pvec);
 			__pagevec_release(&pvec);
 			spin_lock_irq(&zone->lru_lock);
 		}
@@ -1237,6 +1242,8 @@
 	__count_zone_vm_events(PGREFILL, zone, pgscanned);
 	__count_vm_events(PGDEACTIVATE, pgdeactivate);
 	spin_unlock_irq(&zone->lru_lock);
+	if (vm_swap_full())
+		pagevec_swap_free(&pvec);
  
 	pagevec_release(&pvec);
 }
...	...	@@ -25,6 +25,7 @@
25	25	void __pagevec_free(struct pagevec *pvec);
26	26	void ____pagevec_lru_add(struct pagevec *pvec, enum lru_list lru);
27	27	void pagevec_strip(struct pagevec *pvec);
	28	+void pagevec_swap_free(struct pagevec *pvec);
28	29	unsigned pagevec_lookup(struct pagevec pvec, struct address_space mapping,
29	30	pgoff_t start, unsigned nr_pages);
30	31	unsigned pagevec_lookup_tag(struct pagevec *pvec,
...	...	@@ -265,6 +265,7 @@
265	265	extern struct swap_info_struct *get_swap_info_struct(unsigned);
266	266	extern int can_share_swap_page(struct page *);
267	267	extern int remove_exclusive_swap_page(struct page *);
	268	+extern int remove_exclusive_swap_page_ref(struct page *);
268	269	struct backing_dev_info;
269	270
270	271	/* linux/mm/thrash.c */
...	...	@@ -349,6 +350,11 @@
349	350	#define swap_token_default_timeout 0
350	351
351	352	static inline int remove_exclusive_swap_page(struct page *p)
	353	+{
	354	+ return 0;
	355	+}
	356	+
	357	+static inline int remove_exclusive_swap_page_ref(struct page *page)
352	358	{
353	359	return 0;
354	360	}
...	...	@@ -428,6 +428,30 @@
428	428	}
429	429
430	430	/**
	431	+ * pagevec_swap_free - try to free swap space from the pages in a pagevec
	432	+ * @pvec: pagevec with swapcache pages to free the swap space of
	433	+ *
	434	+ * The caller needs to hold an extra reference to each page and
	435	+ * not hold the page lock on the pages. This function uses a
	436	+ * trylock on the page lock so it may not always free the swap
	437	+ * space associated with a page.
	438	+ */
	439	+void pagevec_swap_free(struct pagevec *pvec)
	440	+{
	441	+ int i;
	442	+
	443	+ for (i = 0; i < pagevec_count(pvec); i++) {
	444	+ struct page *page = pvec->pages[i];
	445	+
	446	+ if (PageSwapCache(page) && trylock_page(page)) {
	447	+ if (PageSwapCache(page))
	448	+ remove_exclusive_swap_page_ref(page);
	449	+ unlock_page(page);
	450	+ }
	451	+ }
	452	+}
	453	+
	454	+/**
431	455	* pagevec_lookup - gang pagecache lookup
432	456	* @pvec: Where the resulting pages are placed
433	457	* @mapping: The address_space to search
...	...	@@ -344,7 +344,7 @@
344	344	* Work out if there are any other processes sharing this
345	345	* swap cache page. Free it if you can. Return success.
346	346	*/
347		-int remove_exclusive_swap_page(struct page *page)
	347	+static int remove_exclusive_swap_page_count(struct page *page, int count)
348	348	{
349	349	int retval;
350	350	struct swap_info_struct * p;
...	...	@@ -357,7 +357,7 @@
357	357	return 0;
358	358	if (PageWriteback(page))
359	359	return 0;
360		- if (page_count(page) != 2) /* 2: us + cache */
	360	+ if (page_count(page) != count) /* us + cache + ptes */
361	361	return 0;
362	362
363	363	entry.val = page_private(page);
...	...	@@ -370,7 +370,7 @@
370	370	if (p->swap_map[swp_offset(entry)] == 1) {
371	371	/* Recheck the page count with the swapcache lock held.. */
372	372	spin_lock_irq(&swapper_space.tree_lock);
373		- if ((page_count(page) == 2) && !PageWriteback(page)) {
	373	+ if ((page_count(page) == count) && !PageWriteback(page)) {
374	374	__delete_from_swap_cache(page);
375	375	SetPageDirty(page);
376	376	retval = 1;
...	...	@@ -385,6 +385,25 @@
385	385	}
386	386
387	387	return retval;
	388	+}
	389	+
	390	+/*
	391	+ * Most of the time the page should have two references: one for the
	392	+ * process and one for the swap cache.
	393	+ */
	394	+int remove_exclusive_swap_page(struct page *page)
	395	+{
	396	+ return remove_exclusive_swap_page_count(page, 2);
	397	+}
	398	+
	399	+/*
	400	+ * The pageout code holds an extra reference to the page. That raises
	401	+ * the reference count to test for to 2 for a page that is only in the
	402	+ * swap cache plus 1 for each process that maps the page.
	403	+ */
	404	+int remove_exclusive_swap_page_ref(struct page *page)
	405	+{
	406	+ return remove_exclusive_swap_page_count(page, 2 + page_mapcount(page));
388	407	}
389	408
390	409	/*
...	...	@@ -647,6 +647,9 @@
647	647	continue;
648	648
649	649	activate_locked:
	650	+ /* Not a candidate for swapping, so reclaim swap space. */
	651	+ if (PageSwapCache(page) && vm_swap_full())
	652	+ remove_exclusive_swap_page_ref(page);
650	653	SetPageActive(page);
651	654	pgactivate++;
652	655	keep_locked:
...	...	@@ -1228,6 +1231,8 @@
1228	1231	__mod_zone_page_state(zone, NR_ACTIVE, pgmoved);
1229	1232	pgmoved = 0;
1230	1233	spin_unlock_irq(&zone->lru_lock);
	1234	+ if (vm_swap_full())
	1235	+ pagevec_swap_free(&pvec);
1231	1236	__pagevec_release(&pvec);
1232	1237	spin_lock_irq(&zone->lru_lock);
1233	1238	}
...	...	@@ -1237,6 +1242,8 @@
1237	1242	__count_zone_vm_events(PGREFILL, zone, pgscanned);
1238	1243	__count_vm_events(PGDEACTIVATE, pgdeactivate);
1239	1244	spin_unlock_irq(&zone->lru_lock);
	1245	+ if (vm_swap_full())
	1246	+ pagevec_swap_free(&pvec);
1240	1247
1241	1248	pagevec_release(&pvec);
1242	1249	}