HWPOISON: introduce delete_from_lru_cache()
authorWu Fengguang <fengguang.wu@intel.com>
Wed, 16 Dec 2009 11:19:58 +0000 (12:19 +0100)
committerAndi Kleen <ak@linux.intel.com>
Wed, 16 Dec 2009 11:19:58 +0000 (12:19 +0100)
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>
mm/memory-failure.c

index dc47415a5511bfd657a84e2e444ddb23c65a79b1..9a285f8cdbe19879fdaaa2c8381334e3c19d738a 100644 (file)
@@ -349,6 +349,30 @@ static const char *action_name[] = {
        [RECOVERED] = "Recovered",
 };
 
+/*
+ * 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
@@ -393,6 +417,8 @@ static int me_pagecache_clean(struct page *p, unsigned long pfn)
        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 @@ static int me_swapcache_dirty(struct page *p, unsigned long pfn)
        /* 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 @@ static int hwpoison_user_mappings(struct page *p, unsigned long pfn,
 
 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 @@ int __memory_failure(unsigned long pfn, int trapno, int flags)
         */
        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 @@ int __memory_failure(unsigned long pfn, int trapno, int flags)
        /*
         * 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 @@ int __memory_failure(unsigned long pfn, int trapno, int flags)
 
        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;
                }