HWPOISON: introduce delete_from_lru_cache()

Introduce delete_from_lru_cache() to - clear PG_active, PG_unevictable to avoid complains at unpoison time - move the isolate_lru_page() call back to the handlers instead of the entrance of __memory_failure(), this is more hwpoison filter friendly Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andi Kleen <ak@linux.intel.com>

HWPOISON: introduce delete_from_lru_cache()
Introduce delete_from_lru_cache() to - clear PG_active, PG_unevictable to avoid complains at unpoison time - move the isolate_lru_page() call back to the handlers instead of the entrance of __memory_failure(), this is more hwpoison filter friendly Signed-off-by: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andi Kleen <ak@linux.intel.com>
Wu Fengguang · Andi Kleen
1 parent 71f72525df
Showing 1 changed file with 37 additions and 8 deletions Side-by-side Diff
mm/memory-failure.c
@@ -350,6 +350,30 @@
 };
  
 /*
+ * XXX: It is possible that a page is isolated from LRU cache,
+ * and then kept in swap cache or failed to remove from page cache.
+ * The page count will stop it from being freed by unpoison.
+ * Stress tests should be aware of this memory leak problem.
+ */
+static int delete_from_lru_cache(struct page *p)
+{
+	if (!isolate_lru_page(p)) {
+		/*
+		 * Clear sensible page flags, so that the buddy system won't
+		 * complain when the page is unpoison-and-freed.
+		 */
+		ClearPageActive(p);
+		ClearPageUnevictable(p);
+		/*
+		 * drop the page count elevated by isolate_lru_page()
+		 */
+		page_cache_release(p);
+		return 0;
+	}
+	return -EIO;
+}
+
+/*
  * Error hit kernel page.
  * Do nothing, try to be lucky and not touch this instead. For a few cases we
  * could be more sophisticated.
@@ -393,6 +417,8 @@
 	int ret = FAILED;
 	struct address_space *mapping;
  
+	delete_from_lru_cache(p);
+
 	/*
 	 * For anonymous pages we're done the only reference left
 	 * should be the one m_f() holds.
  
@@ -522,14 +548,20 @@
 	/* Trigger EIO in shmem: */
 	ClearPageUptodate(p);
  
-	return DELAYED;
+	if (!delete_from_lru_cache(p))
+		return DELAYED;
+	else
+		return FAILED;
 }
  
 static int me_swapcache_clean(struct page *p, unsigned long pfn)
 {
 	delete_from_swap_cache(p);
  
-	return RECOVERED;
+	if (!delete_from_lru_cache(p))
+		return RECOVERED;
+	else
+		return FAILED;
 }
  
 /*
@@ -746,7 +778,6 @@
  
 int __memory_failure(unsigned long pfn, int trapno, int flags)
 {
-	unsigned long lru_flag;
 	struct page_state *ps;
 	struct page *p;
 	int res;
  
@@ -796,13 +827,11 @@
 	 */
 	if (!PageLRU(p))
 		lru_add_drain_all();
-	lru_flag = p->flags & lru;
-	if (isolate_lru_page(p)) {
+	if (!PageLRU(p)) {
 		action_result(pfn, "non LRU", IGNORED);
 		put_page(p);
 		return -EBUSY;
 	}
-	page_cache_release(p);
  
 	/*
 	 * Lock the page and wait for writeback to finish.
@@ -825,7 +854,7 @@
 	/*
 	 * Torn down by someone else?
 	 */
-	if ((lru_flag & lru) && !PageSwapCache(p) && p->mapping == NULL) {
+	if (PageLRU(p) && !PageSwapCache(p) && p->mapping == NULL) {
 		action_result(pfn, "already truncated LRU", IGNORED);
 		res = 0;
 		goto out;
@@ -833,7 +862,7 @@
  
 	res = -EBUSY;
 	for (ps = error_states;; ps++) {
-		if (((p->flags | lru_flag)& ps->mask) == ps->res) {
+		if ((p->flags & ps->mask) == ps->res) {
 			res = page_action(ps, p, pfn);
 			break;
 		}
...	...	@@ -350,6 +350,30 @@
350	350	};
351	351
352	352	/*
	353	+ * XXX: It is possible that a page is isolated from LRU cache,
	354	+ * and then kept in swap cache or failed to remove from page cache.
	355	+ * The page count will stop it from being freed by unpoison.
	356	+ * Stress tests should be aware of this memory leak problem.
	357	+ */
	358	+static int delete_from_lru_cache(struct page *p)
	359	+{
	360	+ if (!isolate_lru_page(p)) {
	361	+ /*
	362	+ * Clear sensible page flags, so that the buddy system won't
	363	+ * complain when the page is unpoison-and-freed.
	364	+ */
	365	+ ClearPageActive(p);
	366	+ ClearPageUnevictable(p);
	367	+ /*
	368	+ * drop the page count elevated by isolate_lru_page()
	369	+ */
	370	+ page_cache_release(p);
	371	+ return 0;
	372	+ }
	373	+ return -EIO;
	374	+}
	375	+
	376	+/*
353	377	* Error hit kernel page.
354	378	* Do nothing, try to be lucky and not touch this instead. For a few cases we
355	379	* could be more sophisticated.
...	...	@@ -393,6 +417,8 @@
393	417	int ret = FAILED;
394	418	struct address_space *mapping;
395	419
	420	+ delete_from_lru_cache(p);
	421	+
396	422	/*
397	423	* For anonymous pages we're done the only reference left
398	424	* should be the one m_f() holds.
399	425
...	...	@@ -522,14 +548,20 @@
522	548	/* Trigger EIO in shmem: */
523	549	ClearPageUptodate(p);
524	550
525		- return DELAYED;
	551	+ if (!delete_from_lru_cache(p))
	552	+ return DELAYED;
	553	+ else
	554	+ return FAILED;
526	555	}
527	556
528	557	static int me_swapcache_clean(struct page *p, unsigned long pfn)
529	558	{
530	559	delete_from_swap_cache(p);
531	560
532		- return RECOVERED;
	561	+ if (!delete_from_lru_cache(p))
	562	+ return RECOVERED;
	563	+ else
	564	+ return FAILED;
533	565	}
534	566
535	567	/*
...	...	@@ -746,7 +778,6 @@
746	778
747	779	int __memory_failure(unsigned long pfn, int trapno, int flags)
748	780	{
749		- unsigned long lru_flag;
750	781	struct page_state *ps;
751	782	struct page *p;
752	783	int res;
753	784
...	...	@@ -796,13 +827,11 @@
796	827	*/
797	828	if (!PageLRU(p))
798	829	lru_add_drain_all();
799		- lru_flag = p->flags & lru;
800		- if (isolate_lru_page(p)) {
	830	+ if (!PageLRU(p)) {
801	831	action_result(pfn, "non LRU", IGNORED);
802	832	put_page(p);
803	833	return -EBUSY;
804	834	}
805		- page_cache_release(p);
806	835
807	836	/*
808	837	* Lock the page and wait for writeback to finish.
...	...	@@ -825,7 +854,7 @@
825	854	/*
826	855	* Torn down by someone else?
827	856	*/
828		- if ((lru_flag & lru) && !PageSwapCache(p) && p->mapping == NULL) {
	857	+ if (PageLRU(p) && !PageSwapCache(p) && p->mapping == NULL) {
829	858	action_result(pfn, "already truncated LRU", IGNORED);
830	859	res = 0;
831	860	goto out;
...	...	@@ -833,7 +862,7 @@
833	862
834	863	res = -EBUSY;
835	864	for (ps = error_states;; ps++) {
836		- if (((p->flags \| lru_flag)& ps->mask) == ps->res) {
	865	+ if ((p->flags & ps->mask) == ps->res) {
837	866	res = page_action(ps, p, pfn);
838	867	break;
839	868	}