mm: disable user interface to manually rescue unevictable pages

At one point, anonymous pages were supposed to go on the unevictable list when no swap space was configured, and the idea was to manually rescue those pages after adding swap and making them evictable again. But nowadays, swap-backed pages on the anon LRU list are not scanned without available swap space anyway, so there is no point in moving them to a separate list anymore. The manual rescue could also be used in case pages were stranded on the unevictable list due to race conditions. But the code has been around for a while now and newly discovered bugs should be properly reported and dealt with instead of relying on such a manual fixup. In addition to the lack of a usecase, the sysfs interface to rescue pages from a specific NUMA node has been broken since its introduction, so it's unlikely that anybody ever relied on that. This patch removes the functionality behind the sysctl and the node-interface and emits a one-time warning when somebody tries to access either of them. Signed-off-by: Johannes Weiner <jweiner@redhat.com> Reported-by: Kautuk Consul <consul.kautuk@gmail.com> Reviewed-by: Minchan Kim <minchan.kim@gmail.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: disable user interface to manually rescue unevictable pages
At one point, anonymous pages were supposed to go on the unevictable list when no swap space was configured, and the idea was to manually rescue those pages after adding swap and making them evictable again. But nowadays, swap-backed pages on the anon LRU list are not scanned without available swap space anyway, so there is no point in moving them to a separate list anymore. The manual rescue could also be used in case pages were stranded on the unevictable list due to race conditions. But the code has been around for a while now and newly discovered bugs should be properly reported and dealt with instead of relying on such a manual fixup. In addition to the lack of a usecase, the sysfs interface to rescue pages from a specific NUMA node has been broken since its introduction, so it's unlikely that anybody ever relied on that. This patch removes the functionality behind the sysctl and the node-interface and emits a one-time warning when somebody tries to access either of them. Signed-off-by: Johannes Weiner <jweiner@redhat.com> Reported-by: Kautuk Consul <consul.kautuk@gmail.com> Reviewed-by: Minchan Kim <minchan.kim@gmail.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Johannes Weiner · Linus Torvalds
1 parent 3f380998ae
Showing 1 changed file with 8 additions and 76 deletions Side-by-side Diff
mm/vmscan.c
@@ -3417,68 +3417,14 @@
  
 }
  
-/**
- * scan_zone_unevictable_pages - check unevictable list for evictable pages
- * @zone - zone of which to scan the unevictable list
- *
- * Scan @zone's unevictable LRU lists to check for pages that have become
- * evictable.  Move those that have to @zone's inactive list where they
- * become candidates for reclaim, unless shrink_inactive_zone() decides
- * to reactivate them.  Pages that are still unevictable are rotated
- * back onto @zone's unevictable list.
- */
-#define SCAN_UNEVICTABLE_BATCH_SIZE 16UL /* arbitrary lock hold batch size */
-static void scan_zone_unevictable_pages(struct zone *zone)
+static void warn_scan_unevictable_pages(void)
 {
-	struct list_head *l_unevictable = &zone->lru[LRU_UNEVICTABLE].list;
-	unsigned long scan;
-	unsigned long nr_to_scan = zone_page_state(zone, NR_UNEVICTABLE);
-
-	while (nr_to_scan > 0) {
-		unsigned long batch_size = min(nr_to_scan,
-						SCAN_UNEVICTABLE_BATCH_SIZE);
-
-		spin_lock_irq(&zone->lru_lock);
-		for (scan = 0;  scan < batch_size; scan++) {
-			struct page *page = lru_to_page(l_unevictable);
-
-			if (!trylock_page(page))
-				continue;
-
-			prefetchw_prev_lru_page(page, l_unevictable, flags);
-
-			if (likely(PageLRU(page) && PageUnevictable(page)))
-				check_move_unevictable_page(page, zone);
-
-			unlock_page(page);
-		}
-		spin_unlock_irq(&zone->lru_lock);
-
-		nr_to_scan -= batch_size;
-	}
+	printk_once(KERN_WARNING
+		    "The scan_unevictable_pages sysctl/node-interface has been "
+		    "disabled for lack of a legitimate use case.  If you have "
+		    "one, please send an email to linux-mm@kvack.org.\n");
 }
  
-
-/**
- * scan_all_zones_unevictable_pages - scan all unevictable lists for evictable pages
- *
- * A really big hammer:  scan all zones' unevictable LRU lists to check for
- * pages that have become evictable.  Move those back to the zones'
- * inactive list where they become candidates for reclaim.
- * This occurs when, e.g., we have unswappable pages on the unevictable lists,
- * and we add swap to the system.  As such, it runs in the context of a task
- * that has possibly/probably made some previously unevictable pages
- * evictable.
- */
-static void scan_all_zones_unevictable_pages(void)
-{
-	struct zone *zone;
-
-	for_each_zone(zone) {
-		scan_zone_unevictable_pages(zone);
-	}
-}
-
 /*
  * scan_unevictable_pages [vm] sysctl handler.  On demand re-scan of
  * all nodes' unevictable lists for evictable pages
  
@@ -3489,11 +3435,8 @@
 			   void __user *buffer,
 			   size_t *length, loff_t *ppos)
 {
+	warn_scan_unevictable_pages();
 	proc_doulongvec_minmax(table, write, buffer, length, ppos);
-
-	if (write && *(unsigned long *)table->data)
-		scan_all_zones_unevictable_pages();
-
 	scan_unevictable_pages = 0;
 	return 0;
 }
@@ -3508,6 +3451,7 @@
 					  struct sysdev_attribute *attr,
 					  char *buf)
 {
+	warn_scan_unevictable_pages();
 	return sprintf(buf, "0\n");	/* always zero; should fit... */
 }
  
@@ -3515,19 +3459,7 @@
 					   struct sysdev_attribute *attr,
 					const char *buf, size_t count)
 {
-	struct zone *node_zones = NODE_DATA(dev->id)->node_zones;
-	struct zone *zone;
-	unsigned long res;
-	unsigned long req = strict_strtoul(buf, 10, &res);
-
-	if (req || !res)
-		return 1; /* Invalid input or zero is no-op */
-
-	for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
-		if (!populated_zone(zone))
-			continue;
-		scan_zone_unevictable_pages(zone);
-	}
+	warn_scan_unevictable_pages();
 	return 1;
 }
...	...	@@ -3417,68 +3417,14 @@
3417	3417
3418	3418	}
3419	3419
3420		-/**
3421		- * scan_zone_unevictable_pages - check unevictable list for evictable pages
3422		- * @zone - zone of which to scan the unevictable list
3423		- *
3424		- * Scan @zone's unevictable LRU lists to check for pages that have become
3425		- * evictable. Move those that have to @zone's inactive list where they
3426		- * become candidates for reclaim, unless shrink_inactive_zone() decides
3427		- * to reactivate them. Pages that are still unevictable are rotated
3428		- * back onto @zone's unevictable list.
3429		- */
3430		-#define SCAN_UNEVICTABLE_BATCH_SIZE 16UL /* arbitrary lock hold batch size */
3431		-static void scan_zone_unevictable_pages(struct zone *zone)
	3420	+static void warn_scan_unevictable_pages(void)
3432	3421	{
3433		- struct list_head *l_unevictable = &zone->lru[LRU_UNEVICTABLE].list;
3434		- unsigned long scan;
3435		- unsigned long nr_to_scan = zone_page_state(zone, NR_UNEVICTABLE);
3436		-
3437		- while (nr_to_scan > 0) {
3438		- unsigned long batch_size = min(nr_to_scan,
3439		- SCAN_UNEVICTABLE_BATCH_SIZE);
3440		-
3441		- spin_lock_irq(&zone->lru_lock);
3442		- for (scan = 0; scan < batch_size; scan++) {
3443		- struct page *page = lru_to_page(l_unevictable);
3444		-
3445		- if (!trylock_page(page))
3446		- continue;
3447		-
3448		- prefetchw_prev_lru_page(page, l_unevictable, flags);
3449		-
3450		- if (likely(PageLRU(page) && PageUnevictable(page)))
3451		- check_move_unevictable_page(page, zone);
3452		-
3453		- unlock_page(page);
3454		- }
3455		- spin_unlock_irq(&zone->lru_lock);
3456		-
3457		- nr_to_scan -= batch_size;
3458		- }
	3422	+ printk_once(KERN_WARNING
	3423	+ "The scan_unevictable_pages sysctl/node-interface has been "
	3424	+ "disabled for lack of a legitimate use case. If you have "
	3425	+ "one, please send an email to linux-mm@kvack.org.\n");
3459	3426	}
3460	3427
3461		-
3462		-/**
3463		- * scan_all_zones_unevictable_pages - scan all unevictable lists for evictable pages
3464		- *
3465		- * A really big hammer: scan all zones' unevictable LRU lists to check for
3466		- * pages that have become evictable. Move those back to the zones'
3467		- * inactive list where they become candidates for reclaim.
3468		- * This occurs when, e.g., we have unswappable pages on the unevictable lists,
3469		- * and we add swap to the system. As such, it runs in the context of a task
3470		- * that has possibly/probably made some previously unevictable pages
3471		- * evictable.
3472		- */
3473		-static void scan_all_zones_unevictable_pages(void)
3474		-{
3475		- struct zone *zone;
3476		-
3477		- for_each_zone(zone) {
3478		- scan_zone_unevictable_pages(zone);
3479		- }
3480		-}
3481		-
3482	3428	/*
3483	3429	* scan_unevictable_pages [vm] sysctl handler. On demand re-scan of
3484	3430	* all nodes' unevictable lists for evictable pages
3485	3431
...	...	@@ -3489,11 +3435,8 @@
3489	3435	void __user *buffer,
3490	3436	size_t length, loff_t ppos)
3491	3437	{
	3438	+ warn_scan_unevictable_pages();
3492	3439	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3493		-
3494		- if (write && (unsigned long )table->data)
3495		- scan_all_zones_unevictable_pages();
3496		-
3497	3440	scan_unevictable_pages = 0;
3498	3441	return 0;
3499	3442	}
...	...	@@ -3508,6 +3451,7 @@
3508	3451	struct sysdev_attribute *attr,
3509	3452	char *buf)
3510	3453	{
	3454	+ warn_scan_unevictable_pages();
3511	3455	return sprintf(buf, "0\n"); /* always zero; should fit... */
3512	3456	}
3513	3457
...	...	@@ -3515,19 +3459,7 @@
3515	3459	struct sysdev_attribute *attr,
3516	3460	const char *buf, size_t count)
3517	3461	{
3518		- struct zone *node_zones = NODE_DATA(dev->id)->node_zones;
3519		- struct zone *zone;
3520		- unsigned long res;
3521		- unsigned long req = strict_strtoul(buf, 10, &res);
3522		-
3523		- if (req \|\| !res)
3524		- return 1; /* Invalid input or zero is no-op */
3525		-
3526		- for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
3527		- if (!populated_zone(zone))
3528		- continue;
3529		- scan_zone_unevictable_pages(zone);
3530		- }
	3462	+ warn_scan_unevictable_pages();
3531	3463	return 1;
3532	3464	}
3533	3465