1 files changed, 44 insertions, 13 deletions

diff --git a/fs/dcache.c b/fs/dcache.c
index c71e3732e53b..37b5afdaf698 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -269,6 +269,41 @@ static inline int dname_external(const struct dentry *dentry)
 	return dentry->d_name.name != dentry->d_iname;
 }
 
+/*
+ * Make sure other CPUs see the inode attached before the type is set.
+ */
+static inline void __d_set_inode_and_type(struct dentry *dentry,
+					  struct inode *inode,
+					  unsigned type_flags)
+{
+	unsigned flags;
+
+	dentry->d_inode = inode;
+	smp_wmb();
+	flags = READ_ONCE(dentry->d_flags);
+	flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
+	flags |= type_flags;
+	WRITE_ONCE(dentry->d_flags, flags);
+}
+
+/*
+ * Ideally, we want to make sure that other CPUs see the flags cleared before
+ * the inode is detached, but this is really a violation of RCU principles
+ * since the ordering suggests we should always set inode before flags.
+ *
+ * We should instead replace or discard the entire dentry - but that sucks
+ * performancewise on mass deletion/rename.
+ */
+static inline void __d_clear_type_and_inode(struct dentry *dentry)
+{
+	unsigned flags = READ_ONCE(dentry->d_flags);
+
+	flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
+	WRITE_ONCE(dentry->d_flags, flags);
+	smp_wmb();
+	dentry->d_inode = NULL;
+}
+
 static void dentry_free(struct dentry *dentry)
 {
 	WARN_ON(!hlist_unhashed(&dentry->d_u.d_alias));
@@ -311,7 +346,7 @@ static void dentry_iput(struct dentry * dentry)
 {
 	struct inode *inode = dentry->d_inode;
 	if (inode) {
-		dentry->d_inode = NULL;
+		__d_clear_type_and_inode(dentry);
 		hlist_del_init(&dentry->d_u.d_alias);
 		spin_unlock(&dentry->d_lock);
 		spin_unlock(&inode->i_lock);
@@ -335,8 +370,7 @@ static void dentry_unlink_inode(struct dentry * dentry)
 	__releases(dentry->d_inode->i_lock)
 {
 	struct inode *inode = dentry->d_inode;
-	__d_clear_type(dentry);
-	dentry->d_inode = NULL;
+	__d_clear_type_and_inode(dentry);
 	hlist_del_init(&dentry->d_u.d_alias);
 	dentry_rcuwalk_barrier(dentry);
 	spin_unlock(&dentry->d_lock);
@@ -1205,13 +1239,13 @@ ascend:
 		/* might go back up the wrong parent if we have had a rename. */
 		if (need_seqretry(&rename_lock, seq))
 			goto rename_retry;
-		next = child->d_child.next;
-		while (unlikely(child->d_flags & DCACHE_DENTRY_KILLED)) {
+		/* go into the first sibling still alive */
+		do {
+			next = child->d_child.next;
 			if (next == &this_parent->d_subdirs)
 				goto ascend;
 			child = list_entry(next, struct dentry, d_child);
-			next = next->next;
-		}
+		} while (unlikely(child->d_flags & DCACHE_DENTRY_KILLED));
 		rcu_read_unlock();
 		goto resume;
 	}
@@ -1715,11 +1749,9 @@ static void __d_instantiate(struct dentry *dentry, struct inode *inode)
 	unsigned add_flags = d_flags_for_inode(inode);
 
 	spin_lock(&dentry->d_lock);
-	dentry->d_flags &= ~(DCACHE_ENTRY_TYPE | DCACHE_FALLTHRU);
-	dentry->d_flags |= add_flags;
 	if (inode)
 		hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
-	dentry->d_inode = inode;
+	__d_set_inode_and_type(dentry, inode, add_flags);
 	dentry_rcuwalk_barrier(dentry);
 	spin_unlock(&dentry->d_lock);
 	fsnotify_d_instantiate(dentry, inode);
@@ -1937,8 +1969,7 @@ static struct dentry *__d_obtain_alias(struct inode *inode, int disconnected)
 		add_flags |= DCACHE_DISCONNECTED;
 
 	spin_lock(&tmp->d_lock);
-	tmp->d_inode = inode;
-	tmp->d_flags |= add_flags;
+	__d_set_inode_and_type(tmp, inode, add_flags);
 	hlist_add_head(&tmp->d_u.d_alias, &inode->i_dentry);
 	hlist_bl_lock(&tmp->d_sb->s_anon);
 	hlist_bl_add_head(&tmp->d_hash, &tmp->d_sb->s_anon);
@@ -2690,7 +2721,7 @@ static int __d_unalias(struct inode *inode,
 		struct dentry *dentry, struct dentry *alias)
 {
 	struct mutex *m1 = NULL, *m2 = NULL;
-	int ret = -EBUSY;
+	int ret = -ESTALE;
 
 	/* If alias and dentry share a parent, then no extra locks required */
 	if (alias->d_parent == dentry->d_parent)