mm: extend KSM refcounts to the anon_vma root

KSM reference counts can cause an anon_vma to exist after the processe it
belongs to have already exited.  Because the anon_vma lock now lives in
the root anon_vma, we need to ensure that the root anon_vma stays around
until after all the "child" anon_vmas have been freed.

The obvious way to do this is to have a "child" anon_vma take a reference
to the root in anon_vma_fork.  When the anon_vma is freed at munmap or
process exit, we drop the refcount in anon_vma_unlink and possibly free
the root anon_vma.

The KSM anon_vma reference count function also needs to be modified to
deal with the possibility of freeing 2 levels of anon_vma.  The easiest
way to do this is to break out the KSM magic and make it generic.

When compiling without CONFIG_KSM, this code is compiled out.

Signed-off-by: Rik van Riel <riel@redhat.com>
Tested-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Tested-by: Dave Young <hidave.darkstar@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

3 files changed, 54 insertions, 19 deletions

diff --git a/mm/ksm.c b/mm/ksm.c
index da6037c261f1..9f2acc998a37 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -318,19 +318,14 @@ static void hold_anon_vma(struct rmap_item *rmap_item,
 			  struct anon_vma *anon_vma)
 {
 	rmap_item->anon_vma = anon_vma;
-	atomic_inc(&anon_vma->external_refcount);
+	get_anon_vma(anon_vma);
 }
 
-static void drop_anon_vma(struct rmap_item *rmap_item)
+static void ksm_drop_anon_vma(struct rmap_item *rmap_item)
 {
 	struct anon_vma *anon_vma = rmap_item->anon_vma;
 
-	if (atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->root->lock)) {
-		int empty = list_empty(&anon_vma->head);
-		anon_vma_unlock(anon_vma);
-		if (empty)
-			anon_vma_free(anon_vma);
-	}
+	drop_anon_vma(anon_vma);
 }
 
 /*
@@ -415,7 +410,7 @@ static void break_cow(struct rmap_item *rmap_item)
 	 * It is not an accident that whenever we want to break COW
 	 * to undo, we also need to drop a reference to the anon_vma.
 	 */
-	drop_anon_vma(rmap_item);
+	ksm_drop_anon_vma(rmap_item);
 
 	down_read(&mm->mmap_sem);
 	if (ksm_test_exit(mm))
@@ -470,7 +465,7 @@ static void remove_node_from_stable_tree(struct stable_node *stable_node)
 			ksm_pages_sharing--;
 		else
 			ksm_pages_shared--;
-		drop_anon_vma(rmap_item);
+		ksm_drop_anon_vma(rmap_item);
 		rmap_item->address &= PAGE_MASK;
 		cond_resched();
 	}
@@ -558,7 +553,7 @@ static void remove_rmap_item_from_tree(struct rmap_item *rmap_item)
 		else
 			ksm_pages_shared--;
 
-		drop_anon_vma(rmap_item);
+		ksm_drop_anon_vma(rmap_item);
 		rmap_item->address &= PAGE_MASK;
 
 	} else if (rmap_item->address & UNSTABLE_FLAG) {
diff --git a/mm/migrate.c b/mm/migrate.c
index 5208fa1d9712..38e7cad782f4 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -639,7 +639,7 @@ static int unmap_and_move(new_page_t get_new_page, unsigned long private,
 			 * exist when the page is remapped later
 			 */
 			anon_vma = page_anon_vma(page);
-			atomic_inc(&anon_vma->external_refcount);
+			get_anon_vma(anon_vma);
 		}
 	}
 
@@ -682,12 +682,8 @@ skip_unmap:
 rcu_unlock:
 
 	/* Drop an anon_vma reference if we took one */
-	if (anon_vma && atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->root->lock)) {
-		int empty = list_empty(&anon_vma->head);
-		anon_vma_unlock(anon_vma);
-		if (empty)
-			anon_vma_free(anon_vma);
-	}
+	if (anon_vma)
+		drop_anon_vma(anon_vma);
 
 	if (rcu_locked)
 		rcu_read_unlock();
diff --git a/mm/rmap.c b/mm/rmap.c
index caa48b27371b..07e9814c7a41 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -235,6 +235,12 @@ int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma)
 	 * lock any of the anon_vmas in this anon_vma tree.
 	 */
 	anon_vma->root = pvma->anon_vma->root;
+	/*
+	 * With KSM refcounts, an anon_vma can stay around longer than the
+	 * process it belongs to.  The root anon_vma needs to be pinned
+	 * until this anon_vma is freed, because the lock lives in the root.
+	 */
+	get_anon_vma(anon_vma->root);
 	/* Mark this anon_vma as the one where our new (COWed) pages go. */
 	vma->anon_vma = anon_vma;
 	anon_vma_chain_link(vma, avc, anon_vma);
@@ -264,8 +270,12 @@ static void anon_vma_unlink(struct anon_vma_chain *anon_vma_chain)
 	empty = list_empty(&anon_vma->head) && !anonvma_external_refcount(anon_vma);
 	anon_vma_unlock(anon_vma);
 
-	if (empty)
+	if (empty) {
+		/* We no longer need the root anon_vma */
+		if (anon_vma->root != anon_vma)
+			drop_anon_vma(anon_vma->root);
 		anon_vma_free(anon_vma);
+	}
 }
 
 void unlink_anon_vmas(struct vm_area_struct *vma)
@@ -1382,6 +1392,40 @@ int try_to_munlock(struct page *page)
 		return try_to_unmap_file(page, TTU_MUNLOCK);
 }
 
+#if defined(CONFIG_KSM) || defined(CONFIG_MIGRATION)
+/*
+ * Drop an anon_vma refcount, freeing the anon_vma and anon_vma->root
+ * if necessary.  Be careful to do all the tests under the lock.  Once
+ * we know we are the last user, nobody else can get a reference and we
+ * can do the freeing without the lock.
+ */
+void drop_anon_vma(struct anon_vma *anon_vma)
+{
+	if (atomic_dec_and_lock(&anon_vma->external_refcount, &anon_vma->root->lock)) {
+		struct anon_vma *root = anon_vma->root;
+		int empty = list_empty(&anon_vma->head);
+		int last_root_user = 0;
+		int root_empty = 0;
+
+		/*
+		 * The refcount on a non-root anon_vma got dropped.  Drop
+		 * the refcount on the root and check if we need to free it.
+		 */
+		if (empty && anon_vma != root) {
+			last_root_user = atomic_dec_and_test(&root->external_refcount);
+			root_empty = list_empty(&root->head);
+		}
+		anon_vma_unlock(anon_vma);
+
+		if (empty) {
+			anon_vma_free(anon_vma);
+			if (root_empty && last_root_user)
+				anon_vma_free(root);
+		}
+	}
+}
+#endif
+
 #ifdef CONFIG_MIGRATION
 /*
  * rmap_walk() and its helpers rmap_walk_anon() and rmap_walk_file():