ksm: get_ksm_page locked
[~shefty/rdma-dev.git] / mm / ksm.c
index 70daa35..e02430f 100644 (file)
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -514,15 +514,14 @@ static void remove_node_from_stable_tree(struct stable_node *stable_node)
  * but this is different - made simpler by ksm_thread_mutex being held, but
  * interesting for assuming that no other use of the struct page could ever
  * put our expected_mapping into page->mapping (or a field of the union which
- * coincides with page->mapping).  The RCU calls are not for KSM at all, but
- * to keep the page_count protocol described with page_cache_get_speculative.
+ * coincides with page->mapping).
  *
  * Note: it is possible that get_ksm_page() will return NULL one moment,
  * then page the next, if the page is in between page_freeze_refs() and
  * page_unfreeze_refs(): this shouldn't be a problem anywhere, the page
  * is on its way to being freed; but it is an anomaly to bear in mind.
  */
-static struct page *get_ksm_page(struct stable_node *stable_node)
+static struct page *get_ksm_page(struct stable_node *stable_node, bool locked)
 {
        struct page *page;
        void *expected_mapping;
@@ -530,7 +529,6 @@ static struct page *get_ksm_page(struct stable_node *stable_node)
        page = pfn_to_page(stable_node->kpfn);
        expected_mapping = (void *)stable_node +
                                (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM);
-       rcu_read_lock();
        if (page->mapping != expected_mapping)
                goto stale;
        if (!get_page_unless_zero(page))
@@ -539,10 +537,16 @@ static struct page *get_ksm_page(struct stable_node *stable_node)
                put_page(page);
                goto stale;
        }
-       rcu_read_unlock();
+       if (locked) {
+               lock_page(page);
+               if (page->mapping != expected_mapping) {
+                       unlock_page(page);
+                       put_page(page);
+                       goto stale;
+               }
+       }
        return page;
 stale:
-       rcu_read_unlock();
        remove_node_from_stable_tree(stable_node);
        return NULL;
 }
@@ -558,11 +562,10 @@ static void remove_rmap_item_from_tree(struct rmap_item *rmap_item)
                struct page *page;
 
                stable_node = rmap_item->head;
-               page = get_ksm_page(stable_node);
+               page = get_ksm_page(stable_node, true);
                if (!page)
                        goto out;
 
-               lock_page(page);
                hlist_del(&rmap_item->hlist);
                unlock_page(page);
                put_page(page);
@@ -1042,7 +1045,7 @@ static struct page *stable_tree_search(struct page *page)
 
                cond_resched();
                stable_node = rb_entry(node, struct stable_node, node);
-               tree_page = get_ksm_page(stable_node);
+               tree_page = get_ksm_page(stable_node, false);
                if (!tree_page)
                        return NULL;
 
@@ -1086,7 +1089,7 @@ static struct stable_node *stable_tree_insert(struct page *kpage)
 
                cond_resched();
                stable_node = rb_entry(*new, struct stable_node, node);
-               tree_page = get_ksm_page(stable_node);
+               tree_page = get_ksm_page(stable_node, false);
                if (!tree_page)
                        return NULL;