Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
  [PATCH] ocfs2: Release mutex in error handling code
  [PATCH] ocfs2: Fix oops when racing files truncates with writes into an mmap region
  [PATCH 2/2] ocfs2: Fix race between mount and recovery
  [PATCH 1/2] ocfs2: Add counter in struct ocfs2_dinode to track journal replays
  [PATCH] configfs: Convenience macros for attribute definition.
  [PATCH] configfs: Pin configfs subsystems separately from new config_items.
  [PATCH] configfs: Fix open directory making rmdir() fail
  [PATCH] configfs: Lock new directory inodes before removing on cleanup after failure
  [PATCH] configfs: Prevent userspace from creating new entries under attaching directories
  [PATCH] configfs: Fix failing symlink() making rmdir() fail
  [PATCH] configfs: Fix symlink() to a removing item
  [PATCH] configfs: Include linux/err.h in linux/configfs.h

10 files changed, 367 insertions, 98 deletions

diff --git a/fs/configfs/configfs_internal.h b/fs/configfs/configfs_internal.h
index da015c12e3e..762d287123c 100644
--- a/fs/configfs/configfs_internal.h
+++ b/fs/configfs/configfs_internal.h
@@ -49,8 +49,10 @@ struct configfs_dirent {
 #define CONFIGFS_USET_DEFAULT	0x0080
 #define CONFIGFS_USET_DROPPING	0x0100
 #define CONFIGFS_USET_IN_MKDIR	0x0200
+#define CONFIGFS_USET_CREATING	0x0400
 #define CONFIGFS_NOT_PINNED	(CONFIGFS_ITEM_ATTR)
 
+extern struct mutex configfs_symlink_mutex;
 extern spinlock_t configfs_dirent_lock;
 
 extern struct vfsmount * configfs_mount;
@@ -66,6 +68,7 @@ extern void configfs_inode_exit(void);
 extern int configfs_create_file(struct config_item *, const struct configfs_attribute *);
 extern int configfs_make_dirent(struct configfs_dirent *,
 				struct dentry *, void *, umode_t, int);
+extern int configfs_dirent_is_ready(struct configfs_dirent *);
 
 extern int configfs_add_file(struct dentry *, const struct configfs_attribute *, int);
 extern void configfs_hash_and_remove(struct dentry * dir, const char * name);
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index 179589be063..7a8db78a91d 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -185,7 +185,7 @@ static int create_dir(struct config_item * k, struct dentry * p,
 	error = configfs_dirent_exists(p->d_fsdata, d->d_name.name);
 	if (!error)
 		error = configfs_make_dirent(p->d_fsdata, d, k, mode,
-					     CONFIGFS_DIR);
+					     CONFIGFS_DIR | CONFIGFS_USET_CREATING);
 	if (!error) {
 		error = configfs_create(d, mode, init_dir);
 		if (!error) {
@@ -209,6 +209,9 @@ static int create_dir(struct config_item * k, struct dentry * p,
  *	configfs_create_dir - create a directory for an config_item.
  *	@item:		config_itemwe're creating directory for.
  *	@dentry:	config_item's dentry.
+ *
+ *	Note: user-created entries won't be allowed under this new directory
+ *	until it is validated by configfs_dir_set_ready()
  */
 
 static int configfs_create_dir(struct config_item * item, struct dentry *dentry)
@@ -231,6 +234,44 @@ static int configfs_create_dir(struct config_item * item, struct dentry *dentry)
 	return error;
 }
 
+/*
+ * Allow userspace to create new entries under a new directory created with
+ * configfs_create_dir(), and under all of its chidlren directories recursively.
+ * @sd		configfs_dirent of the new directory to validate
+ *
+ * Caller must hold configfs_dirent_lock.
+ */
+static void configfs_dir_set_ready(struct configfs_dirent *sd)
+{
+	struct configfs_dirent *child_sd;
+
+	sd->s_type &= ~CONFIGFS_USET_CREATING;
+	list_for_each_entry(child_sd, &sd->s_children, s_sibling)
+		if (child_sd->s_type & CONFIGFS_USET_CREATING)
+			configfs_dir_set_ready(child_sd);
+}
+
+/*
+ * Check that a directory does not belong to a directory hierarchy being
+ * attached and not validated yet.
+ * @sd		configfs_dirent of the directory to check
+ *
+ * @return	non-zero iff the directory was validated
+ *
+ * Note: takes configfs_dirent_lock, so the result may change from false to true
+ * in two consecutive calls, but never from true to false.
+ */
+int configfs_dirent_is_ready(struct configfs_dirent *sd)
+{
+	int ret;
+
+	spin_lock(&configfs_dirent_lock);
+	ret = !(sd->s_type & CONFIGFS_USET_CREATING);
+	spin_unlock(&configfs_dirent_lock);
+
+	return ret;
+}
+
 int configfs_create_link(struct configfs_symlink *sl,
 			 struct dentry *parent,
 			 struct dentry *dentry)
@@ -283,6 +324,8 @@ static void remove_dir(struct dentry * d)
  * The only thing special about this is that we remove any files in
  * the directory before we remove the directory, and we've inlined
  * what used to be configfs_rmdir() below, instead of calling separately.
+ *
+ * Caller holds the mutex of the item's inode
  */
 
 static void configfs_remove_dir(struct config_item * item)
@@ -330,7 +373,19 @@ static struct dentry * configfs_lookup(struct inode *dir,
 	struct configfs_dirent * parent_sd = dentry->d_parent->d_fsdata;
 	struct configfs_dirent * sd;
 	int found = 0;
-	int err = 0;
+	int err;
+
+	/*
+	 * Fake invisibility if dir belongs to a group/default groups hierarchy
+	 * being attached
+	 *
+	 * This forbids userspace to read/write attributes of items which may
+	 * not complete their initialization, since the dentries of the
+	 * attributes won't be instantiated.
+	 */
+	err = -ENOENT;
+	if (!configfs_dirent_is_ready(parent_sd))
+		goto out;
 
 	list_for_each_entry(sd, &parent_sd->s_children, s_sibling) {
 		if (sd->s_type & CONFIGFS_NOT_PINNED) {
@@ -353,6 +408,7 @@ static struct dentry * configfs_lookup(struct inode *dir,
 		return simple_lookup(dir, dentry, nd);
 	}
 
+out:
 	return ERR_PTR(err);
 }
 
@@ -370,13 +426,17 @@ static int configfs_detach_prep(struct dentry *dentry, struct mutex **wait_mutex
 	struct configfs_dirent *sd;
 	int ret;
 
+	/* Mark that we're trying to drop the group */
+	parent_sd->s_type |= CONFIGFS_USET_DROPPING;
+
 	ret = -EBUSY;
 	if (!list_empty(&parent_sd->s_links))
 		goto out;
 
 	ret = 0;
 	list_for_each_entry(sd, &parent_sd->s_children, s_sibling) {
-		if (sd->s_type & CONFIGFS_NOT_PINNED)
+		if (!sd->s_element ||
+		    (sd->s_type & CONFIGFS_NOT_PINNED))
 			continue;
 		if (sd->s_type & CONFIGFS_USET_DEFAULT) {
 			/* Abort if racing with mkdir() */
@@ -385,8 +445,6 @@ static int configfs_detach_prep(struct dentry *dentry, struct mutex **wait_mutex
 					*wait_mutex = &sd->s_dentry->d_inode->i_mutex;
 				return -EAGAIN;
 			}
-			/* Mark that we're trying to drop the group */
-			sd->s_type |= CONFIGFS_USET_DROPPING;
 
 			/*
 			 * Yup, recursive.  If there's a problem, blame
@@ -414,12 +472,11 @@ static void configfs_detach_rollback(struct dentry *dentry)
 	struct configfs_dirent *parent_sd = dentry->d_fsdata;
 	struct configfs_dirent *sd;
 
-	list_for_each_entry(sd, &parent_sd->s_children, s_sibling) {
-		if (sd->s_type & CONFIGFS_USET_DEFAULT) {
+	parent_sd->s_type &= ~CONFIGFS_USET_DROPPING;
+
+	list_for_each_entry(sd, &parent_sd->s_children, s_sibling)
+		if (sd->s_type & CONFIGFS_USET_DEFAULT)
 			configfs_detach_rollback(sd->s_dentry);
-			sd->s_type &= ~CONFIGFS_USET_DROPPING;
-		}
-	}
 }
 
 static void detach_attrs(struct config_item * item)
@@ -558,36 +615,21 @@ static int create_default_group(struct config_group *parent_group,
 static int populate_groups(struct config_group *group)
 {
 	struct config_group *new_group;
-	struct dentry *dentry = group->cg_item.ci_dentry;
 	int ret = 0;
 	int i;
 
 	if (group->default_groups) {
-		/*
-		 * FYI, we're faking mkdir here
-		 * I'm not sure we need this semaphore, as we're called
-		 * from our parent's mkdir.  That holds our parent's
-		 * i_mutex, so afaik lookup cannot continue through our
-		 * parent to find us, let alone mess with our tree.
-		 * That said, taking our i_mutex is closer to mkdir
-		 * emulation, and shouldn't hurt.
-		 */
-		mutex_lock_nested(&dentry->d_inode->i_mutex, I_MUTEX_CHILD);
-
 		for (i = 0; group->default_groups[i]; i++) {
 			new_group = group->default_groups[i];
 
 			ret = create_default_group(group, new_group);
-			if (ret)
+			if (ret) {
+				detach_groups(group);
 				break;
+			}
 		}
-
-		mutex_unlock(&dentry->d_inode->i_mutex);
 	}
 
-	if (ret)
-		detach_groups(group);
-
 	return ret;
 }
 
@@ -702,7 +744,15 @@ static int configfs_attach_item(struct config_item *parent_item,
 	if (!ret) {
 		ret = populate_attrs(item);
 		if (ret) {
+			/*
+			 * We are going to remove an inode and its dentry but
+			 * the VFS may already have hit and used them. Thus,
+			 * we must lock them as rmdir() would.
+			 */
+			mutex_lock(&dentry->d_inode->i_mutex);
 			configfs_remove_dir(item);
+			dentry->d_inode->i_flags |= S_DEAD;
+			mutex_unlock(&dentry->d_inode->i_mutex);
 			d_delete(dentry);
 		}
 	}
@@ -710,6 +760,7 @@ static int configfs_attach_item(struct config_item *parent_item,
 	return ret;
 }
 
+/* Caller holds the mutex of the item's inode */
 static void configfs_detach_item(struct config_item *item)
 {
 	detach_attrs(item);
@@ -728,16 +779,30 @@ static int configfs_attach_group(struct config_item *parent_item,
 		sd = dentry->d_fsdata;
 		sd->s_type |= CONFIGFS_USET_DIR;
 
+		/*
+		 * FYI, we're faking mkdir in populate_groups()
+		 * We must lock the group's inode to avoid races with the VFS
+		 * which can already hit the inode and try to add/remove entries
+		 * under it.
+		 *
+		 * We must also lock the inode to remove it safely in case of
+		 * error, as rmdir() would.
+		 */
+		mutex_lock_nested(&dentry->d_inode->i_mutex, I_MUTEX_CHILD);
 		ret = populate_groups(to_config_group(item));
 		if (ret) {
 			configfs_detach_item(item);
-			d_delete(dentry);
+			dentry->d_inode->i_flags |= S_DEAD;
 		}
+		mutex_unlock(&dentry->d_inode->i_mutex);
+		if (ret)
+			d_delete(dentry);
 	}
 
 	return ret;
 }
 
+/* Caller holds the mutex of the group's inode */
 static void configfs_detach_group(struct config_item *item)
 {
 	detach_groups(to_config_group(item));
@@ -1035,7 +1100,7 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	struct configfs_subsystem *subsys;
 	struct configfs_dirent *sd;
 	struct config_item_type *type;
-	struct module *owner = NULL;
+	struct module *subsys_owner = NULL, *new_item_owner = NULL;
 	char *name;
 
 	if (dentry->d_parent == configfs_sb->s_root) {
@@ -1044,6 +1109,16 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	}
 
 	sd = dentry->d_parent->d_fsdata;
+
+	/*
+	 * Fake invisibility if dir belongs to a group/default groups hierarchy
+	 * being attached
+	 */
+	if (!configfs_dirent_is_ready(sd)) {
+		ret = -ENOENT;
+		goto out;
+	}
+
 	if (!(sd->s_type & CONFIGFS_USET_DIR)) {
 		ret = -EPERM;
 		goto out;
@@ -1062,10 +1137,25 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 		goto out_put;
 	}
 
+	/*
+	 * The subsystem may belong to a different module than the item
+	 * being created.  We don't want to safely pin the new item but
+	 * fail to pin the subsystem it sits under.
+	 */
+	if (!subsys->su_group.cg_item.ci_type) {
+		ret = -EINVAL;
+		goto out_put;
+	}
+	subsys_owner = subsys->su_group.cg_item.ci_type->ct_owner;
+	if (!try_module_get(subsys_owner)) {
+		ret = -EINVAL;
+		goto out_put;
+	}
+
 	name = kmalloc(dentry->d_name.len + 1, GFP_KERNEL);
 	if (!name) {
 		ret = -ENOMEM;
-		goto out_put;
+		goto out_subsys_put;
 	}
 
 	snprintf(name, dentry->d_name.len + 1, "%s", dentry->d_name.name);
@@ -1094,10 +1184,10 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	kfree(name);
 	if (ret) {
 		/*
-		 * If item == NULL, then link_obj() was never called.
+		 * If ret != 0, then link_obj() was never called.
 		 * There are no extra references to clean up.
 		 */
-		goto out_put;
+		goto out_subsys_put;
 	}
 
 	/*
@@ -1111,8 +1201,8 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 		goto out_unlink;
 	}
 
-	owner = type->ct_owner;
-	if (!try_module_get(owner)) {
+	new_item_owner = type->ct_owner;
+	if (!try_module_get(new_item_owner)) {
 		ret = -EINVAL;
 		goto out_unlink;
 	}
@@ -1142,6 +1232,8 @@ static int configfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 
 	spin_lock(&configfs_dirent_lock);
 	sd->s_type &= ~CONFIGFS_USET_IN_MKDIR;
+	if (!ret)
+		configfs_dir_set_ready(dentry->d_fsdata);
 	spin_unlock(&configfs_dirent_lock);
 
 out_unlink:
@@ -1159,9 +1251,13 @@ out_unlink:
 		mutex_unlock(&subsys->su_mutex);
 
 		if (module_got)
-			module_put(owner);
+			module_put(new_item_owner);
 	}
 
+out_subsys_put:
+	if (ret)
+		module_put(subsys_owner);
+
 out_put:
 	/*
 	 * link_obj()/link_group() took a reference from child->parent,
@@ -1180,7 +1276,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 	struct config_item *item;
 	struct configfs_subsystem *subsys;
 	struct configfs_dirent *sd;
-	struct module *owner = NULL;
+	struct module *subsys_owner = NULL, *dead_item_owner = NULL;
 	int ret;
 
 	if (dentry->d_parent == configfs_sb->s_root)
@@ -1207,6 +1303,15 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 		return -EINVAL;
 	}
 
+	/* configfs_mkdir() shouldn't have allowed this */
+	BUG_ON(!subsys->su_group.cg_item.ci_type);
+	subsys_owner = subsys->su_group.cg_item.ci_type->ct_owner;
+
+	/*
+	 * Ensure that no racing symlink() will make detach_prep() fail while
+	 * the new link is temporarily attached
+	 */
+	mutex_lock(&configfs_symlink_mutex);
 	spin_lock(&configfs_dirent_lock);
 	do {
 		struct mutex *wait_mutex;
@@ -1215,6 +1320,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 		if (ret) {
 			configfs_detach_rollback(dentry);
 			spin_unlock(&configfs_dirent_lock);
+			mutex_unlock(&configfs_symlink_mutex);
 			if (ret != -EAGAIN) {
 				config_item_put(parent_item);
 				return ret;
@@ -1224,10 +1330,12 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 			mutex_lock(wait_mutex);
 			mutex_unlock(wait_mutex);
 
+			mutex_lock(&configfs_symlink_mutex);
 			spin_lock(&configfs_dirent_lock);
 		}
 	} while (ret == -EAGAIN);
 	spin_unlock(&configfs_dirent_lock);
+	mutex_unlock(&configfs_symlink_mutex);
 
 	/* Get a working ref for the duration of this function */
 	item = configfs_get_config_item(dentry);
@@ -1236,7 +1344,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 	config_item_put(parent_item);
 
 	if (item->ci_type)
-		owner = item->ci_type->ct_owner;
+		dead_item_owner = item->ci_type->ct_owner;
 
 	if (sd->s_type & CONFIGFS_USET_DIR) {
 		configfs_detach_group(item);
@@ -1258,7 +1366,8 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
 	/* Drop our reference from above */
 	config_item_put(item);
 
-	module_put(owner);
+	module_put(dead_item_owner);
+	module_put(subsys_owner);
 
 	return 0;
 }
@@ -1314,13 +1423,24 @@ static int configfs_dir_open(struct inode *inode, struct file *file)
 {
 	struct dentry * dentry = file->f_path.dentry;
 	struct configfs_dirent * parent_sd = dentry->d_fsdata;
+	int err;
 
 	mutex_lock(&dentry->d_inode->i_mutex);
-	file->private_data = configfs_new_dirent(parent_sd, NULL);
+	/*
+	 * Fake invisibility if dir belongs to a group/default groups hierarchy
+	 * being attached
+	 */
+	err = -ENOENT;
+	if (configfs_dirent_is_ready(parent_sd)) {
+		file->private_data = configfs_new_dirent(parent_sd, NULL);
+		if (IS_ERR(file->private_data))
+			err = PTR_ERR(file->private_data);
+		else
+			err = 0;
+	}
 	mutex_unlock(&dentry->d_inode->i_mutex);
 
-	return IS_ERR(file->private_data) ? PTR_ERR(file->private_data) : 0;
-
+	return err;
 }
 
 static int configfs_dir_close(struct inode *inode, struct file *file)
@@ -1491,6 +1611,10 @@ int configfs_register_subsystem(struct configfs_subsystem *subsys)
 		if (err) {
 			d_delete(dentry);
 			dput(dentry);
+		} else {
+			spin_lock(&configfs_dirent_lock);
+			configfs_dir_set_ready(dentry->d_fsdata);
+			spin_unlock(&configfs_dirent_lock);
 		}
 	}
 
@@ -1517,11 +1641,13 @@ void configfs_unregister_subsystem(struct configfs_subsystem *subsys)
 	mutex_lock_nested(&configfs_sb->s_root->d_inode->i_mutex,
 			  I_MUTEX_PARENT);
 	mutex_lock_nested(&dentry->d_inode->i_mutex, I_MUTEX_CHILD);
+	mutex_lock(&configfs_symlink_mutex);
 	spin_lock(&configfs_dirent_lock);
 	if (configfs_detach_prep(dentry, NULL)) {
 		printk(KERN_ERR "configfs: Tried to unregister non-empty subsystem!\n");
 	}
 	spin_unlock(&configfs_dirent_lock);
+	mutex_unlock(&configfs_symlink_mutex);
 	configfs_detach_group(&group->cg_item);
 	dentry->d_inode->i_flags |= S_DEAD;
 	mutex_unlock(&dentry->d_inode->i_mutex);
diff --git a/fs/configfs/symlink.c b/fs/configfs/symlink.c
index 0004d18c40a..bf74973b049 100644
--- a/fs/configfs/symlink.c
+++ b/fs/configfs/symlink.c
@@ -31,6 +31,9 @@
 #include <linux/configfs.h>
 #include "configfs_internal.h"
 
+/* Protects attachments of new symlinks */
+DEFINE_MUTEX(configfs_symlink_mutex);
+
 static int item_depth(struct config_item * item)
 {
 	struct config_item * p = item;
@@ -73,11 +76,20 @@ static int create_link(struct config_item *parent_item,
 	struct configfs_symlink *sl;
 	int ret;
 
+	ret = -ENOENT;
+	if (!configfs_dirent_is_ready(target_sd))
+		goto out;
 	ret = -ENOMEM;
 	sl = kmalloc(sizeof(struct configfs_symlink), GFP_KERNEL);
 	if (sl) {
 		sl->sl_target = config_item_get(item);
 		spin_lock(&configfs_dirent_lock);
+		if (target_sd->s_type & CONFIGFS_USET_DROPPING) {
+			spin_unlock(&configfs_dirent_lock);
+			config_item_put(item);
+			kfree(sl);
+			return -ENOENT;
+		}
 		list_add(&sl->sl_list, &target_sd->s_links);
 		spin_unlock(&configfs_dirent_lock);
 		ret = configfs_create_link(sl, parent_item->ci_dentry,
@@ -91,6 +103,7 @@ static int create_link(struct config_item *parent_item,
 		}
 	}
 
+out:
 	return ret;
 }
 
@@ -120,6 +133,7 @@ int configfs_symlink(struct inode *dir, struct dentry *dentry, const char *symna
 {
 	int ret;
 	struct nameidata nd;
+	struct configfs_dirent *sd;
 	struct config_item *parent_item;
 	struct config_item *target_item;
 	struct config_item_type *type;
@@ -128,9 +142,19 @@ int configfs_symlink(struct inode *dir, struct dentry *dentry, const char *symna
 	if (dentry->d_parent == configfs_sb->s_root)
 		goto out;
 
+	sd = dentry->d_parent->d_fsdata;
+	/*
+	 * Fake invisibility if dir belongs to a group/default groups hierarchy
+	 * being attached
+	 */
+	ret = -ENOENT;
+	if (!configfs_dirent_is_ready(sd))
+		goto out;
+
 	parent_item = configfs_get_config_item(dentry->d_parent);
 	type = parent_item->ci_type;
 
+	ret = -EPERM;
 	if (!type || !type->ct_item_ops ||
 	    !type->ct_item_ops->allow_link)
 		goto out_put;
@@ -141,7 +165,9 @@ int configfs_symlink(struct inode *dir, struct dentry *dentry, const char *symna
 
 	ret = type->ct_item_ops->allow_link(parent_item, target_item);
 	if (!ret) {
+		mutex_lock(&configfs_symlink_mutex);
 		ret = create_link(parent_item, target_item, dentry);
+		mutex_unlock(&configfs_symlink_mutex);
 		if (ret && type->ct_item_ops->drop_link)
 			type->ct_item_ops->drop_link(parent_item,
 						     target_item);
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 1db080135c6..506c24fb507 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -1073,12 +1073,15 @@ static void ocfs2_write_failure(struct inode *inode,
 	for(i = 0; i < wc->w_num_pages; i++) {
 		tmppage = wc->w_pages[i];
 
-		if (ocfs2_should_order_data(inode))
-			walk_page_buffers(wc->w_handle, page_buffers(tmppage),
-					  from, to, NULL,
-					  ocfs2_journal_dirty_data);
-
-		block_commit_write(tmppage, from, to);
+		if (page_has_buffers(tmppage)) {
+			if (ocfs2_should_order_data(inode))
+				walk_page_buffers(wc->w_handle,
+						  page_buffers(tmppage),
+						  from, to, NULL,
+						  ocfs2_journal_dirty_data);
+
+			block_commit_write(tmppage, from, to);
+		}
 	}
 }
 
@@ -1901,12 +1904,14 @@ int ocfs2_write_end_nolock(struct address_space *mapping,
 			to = PAGE_CACHE_SIZE;
 		}
 
-		if (ocfs2_should_order_data(inode))
-			walk_page_buffers(wc->w_handle, page_buffers(tmppage),
-					  from, to, NULL,
-					  ocfs2_journal_dirty_data);
-
-		block_commit_write(tmppage, from, to);
+		if (page_has_buffers(tmppage)) {
+			if (ocfs2_should_order_data(inode))
+				walk_page_buffers(wc->w_handle,
+						  page_buffers(tmppage),
+						  from, to, NULL,
+						  ocfs2_journal_dirty_data);
+			block_commit_write(tmppage, from, to);
+		}
 	}
 
 out_write_size:
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index be2dd95d3a1..ec2ed15c3da 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -1766,8 +1766,8 @@ out_inode_unlock:
 out_rw_unlock:
 	ocfs2_rw_unlock(inode, 1);
 
-	mutex_unlock(&inode->i_mutex);
 out:
+	mutex_unlock(&inode->i_mutex);
 	return ret;
 }
 
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index a8c19cb3cfd..7a37240f7a3 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -57,7 +57,7 @@ static int __ocfs2_recovery_thread(void *arg);
 static int ocfs2_commit_cache(struct ocfs2_super *osb);
 static int ocfs2_wait_on_mount(struct ocfs2_super *osb);
 static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
-				      int dirty);
+				      int dirty, int replayed);
 static int ocfs2_trylock_journal(struct ocfs2_super *osb,
 				 int slot_num);
 static int ocfs2_recover_orphans(struct ocfs2_super *osb,
@@ -562,8 +562,18 @@ done:
 	return status;
 }
 
+static void ocfs2_bump_recovery_generation(struct ocfs2_dinode *di)
+{
+	le32_add_cpu(&(di->id1.journal1.ij_recovery_generation), 1);
+}
+
+static u32 ocfs2_get_recovery_generation(struct ocfs2_dinode *di)
+{
+	return le32_to_cpu(di->id1.journal1.ij_recovery_generation);
+}
+
 static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
-				      int dirty)
+				      int dirty, int replayed)
 {
 	int status;
 	unsigned int flags;
@@ -593,6 +603,9 @@ static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
 		flags &= ~OCFS2_JOURNAL_DIRTY_FL;
 	fe->id1.journal1.ij_flags = cpu_to_le32(flags);
 
+	if (replayed)
+		ocfs2_bump_recovery_generation(fe);
+
 	status = ocfs2_write_block(osb, bh, journal->j_inode);
 	if (status < 0)
 		mlog_errno(status);
@@ -667,7 +680,7 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb)
 		 * Do not toggle if flush was unsuccessful otherwise
 		 * will leave dirty metadata in a "clean" journal
 		 */
-		status = ocfs2_journal_toggle_dirty(osb, 0);
+		status = ocfs2_journal_toggle_dirty(osb, 0, 0);
 		if (status < 0)
 			mlog_errno(status);
 	}
@@ -710,7 +723,7 @@ static void ocfs2_clear_journal_error(struct super_block *sb,
 	}
 }
 
-int ocfs2_journal_load(struct ocfs2_journal *journal, int local)
+int ocfs2_journal_load(struct ocfs2_journal *journal, int local, int replayed)
 {
 	int status = 0;
 	struct ocfs2_super *osb;
@@ -729,7 +742,7 @@ int ocfs2_journal_load(struct ocfs2_journal *journal, int local)
 
 	ocfs2_clear_journal_error(osb->sb, journal->j_journal, osb->slot_num);
 
-	status = ocfs2_journal_toggle_dirty(osb, 1);
+	status = ocfs2_journal_toggle_dirty(osb, 1, replayed);
 	if (status < 0) {
 		mlog_errno(status);
 		goto done;
@@ -771,7 +784,7 @@ int ocfs2_journal_wipe(struct ocfs2_journal *journal, int full)
 		goto bail;
 	}
 
-	status = ocfs2_journal_toggle_dirty(journal->j_osb, 0);
+	status = ocfs2_journal_toggle_dirty(journal->j_osb, 0, 0);
 	if (status < 0)
 		mlog_errno(status);
 
@@ -1034,6 +1047,12 @@ restart:
 	spin_unlock(&osb->osb_lock);
 	mlog(0, "All nodes recovered\n");
 
+	/* Refresh all journal recovery generations from disk */
+	status = ocfs2_check_journals_nolocks(osb);
+	status = (status == -EROFS) ? 0 : status;
+	if (status < 0)
+		mlog_errno(status);
+
 	ocfs2_super_unlock(osb, 1);
 
 	/* We always run recovery on our own orphan dir - the dead
@@ -1096,6 +1115,42 @@ out:
 	mlog_exit_void();
 }
 
+static int ocfs2_read_journal_inode(struct ocfs2_super *osb,
+				    int slot_num,
+				    struct buffer_head **bh,
+				    struct inode **ret_inode)
+{
+	int status = -EACCES;
+	struct inode *inode = NULL;
+
+	BUG_ON(slot_num >= osb->max_slots);
+
+	inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
+					    slot_num);
+	if (!inode || is_bad_inode(inode)) {
+		mlog_errno(status);
+		goto bail;
+	}
+	SET_INODE_JOURNAL(inode);
+
+	status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, bh, 0, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = 0;
+
+bail:
+	if (inode) {
+		if (status || !ret_inode)
+			iput(inode);
+		else
+			*ret_inode = inode;
+	}
+	return status;
+}
+
 /* Does the actual journal replay and marks the journal inode as
  * clean. Will only replay if the journal inode is marked dirty. */
 static int ocfs2_replay_journal(struct ocfs2_super *osb,
@@ -1109,22 +1164,36 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb,
 	struct ocfs2_dinode *fe;
 	journal_t *journal = NULL;
 	struct buffer_head *bh = NULL;
+	u32 slot_reco_gen;
 
-	inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
-					    slot_num);
-	if (inode == NULL) {
-		status = -EACCES;
+	status = ocfs2_read_journal_inode(osb, slot_num, &bh, &inode);
+	if (status) {
 		mlog_errno(status);
 		goto done;
 	}
-	if (is_bad_inode(inode)) {
-		status = -EACCES;
-		iput(inode);
-		inode = NULL;
-		mlog_errno(status);
+
+	fe = (struct ocfs2_dinode *)bh->b_data;
+	slot_reco_gen = ocfs2_get_recovery_generation(fe);
+	brelse(bh);
+	bh = NULL;
+
+	/*
+	 * As the fs recovery is asynchronous, there is a small chance that
+	 * another node mounted (and recovered) the slot before the recovery
+	 * thread could get the lock. To handle that, we dirty read the journal
+	 * inode for that slot to get the recovery generation. If it is
+	 * different than what we expected, the slot has been recovered.
+	 * If not, it needs recovery.
+	 */
+	if (osb->slot_recovery_generations[slot_num] != slot_reco_gen) {
+		mlog(0, "Slot %u already recovered (old/new=%u/%u)\n", slot_num,
+		     osb->slot_recovery_generations[slot_num], slot_reco_gen);
+		osb->slot_recovery_generations[slot_num] = slot_reco_gen;
+		status = -EBUSY;
 		goto done;
 	}
-	SET_INODE_JOURNAL(inode);
+
+	/* Continue with recovery as the journal has not yet been recovered */
 
 	status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY);
 	if (status < 0) {
@@ -1138,9 +1207,12 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb,
 	fe = (struct ocfs2_dinode *) bh->b_data;
 
 	flags = le32_to_cpu(fe->id1.journal1.ij_flags);
+	slot_reco_gen = ocfs2_get_recovery_generation(fe);
 
 	if (!(flags & OCFS2_JOURNAL_DIRTY_FL)) {
 		mlog(0, "No recovery required for node %d\n", node_num);
+		/* Refresh recovery generation for the slot */
+		osb->slot_recovery_generations[slot_num] = slot_reco_gen;
 		goto done;
 	}
 
@@ -1188,6 +1260,11 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb,
 	flags &= ~OCFS2_JOURNAL_DIRTY_FL;
 	fe->id1.journal1.ij_flags = cpu_to_le32(flags);
 
+	/* Increment recovery generation to indicate successful recovery */
+	ocfs2_bump_recovery_generation(fe);
+	osb->slot_recovery_generations[slot_num] =
+					ocfs2_get_recovery_generation(fe);
+
 	status = ocfs2_write_block(osb, bh, inode);
 	if (status < 0)
 		mlog_errno(status);
@@ -1252,6 +1329,13 @@ static int ocfs2_recover_node(struct ocfs2_super *osb,
 
 	status = ocfs2_replay_journal(osb, node_num, slot_num);
 	if (status < 0) {
+		if (status == -EBUSY) {
+			mlog(0, "Skipping recovery for slot %u (node %u) "
+			     "as another node has recovered it\n", slot_num,
+			     node_num);
+			status = 0;
+			goto done;
+		}
 		mlog_errno(status);
 		goto done;
 	}
@@ -1334,12 +1418,29 @@ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb)
 {
 	unsigned int node_num;
 	int status, i;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_dinode *di;
 
 	/* This is called with the super block cluster lock, so we
 	 * know that the slot map can't change underneath us. */
 
 	spin_lock(&osb->osb_lock);
 	for (i = 0; i < osb->max_slots; i++) {
+		/* Read journal inode to get the recovery generation */
+		status = ocfs2_read_journal_inode(osb, i, &bh, NULL);
+		if (status) {
+			mlog_errno(status);
+			goto bail;
+		}
+		di = (struct ocfs2_dinode *)bh->b_data;
+		osb->slot_recovery_generations[i] =
+					ocfs2_get_recovery_generation(di);
+		brelse(bh);
+		bh = NULL;
+
+		mlog(0, "Slot %u recovery generation is %u\n", i,
+		     osb->slot_recovery_generations[i]);
+
 		if (i == osb->slot_num)
 			continue;
 
@@ -1603,49 +1704,41 @@ static int ocfs2_commit_thread(void *arg)
 	return 0;
 }
 
-/* Look for a dirty journal without taking any cluster locks. Used for
- * hard readonly access to determine whether the file system journals
- * require recovery. */
+/* Reads all the journal inodes without taking any cluster locks. Used
+ * for hard readonly access to determine whether any journal requires
+ * recovery. Also used to refresh the recovery generation numbers after
+ * a journal has been recovered by another node.
+ */
 int ocfs2_check_journals_nolocks(struct ocfs2_super *osb)
 {
 	int ret = 0;
 	unsigned int slot;
-	struct buffer_head *di_bh;
+	struct buffer_head *di_bh = NULL;
 	struct ocfs2_dinode *di;
-	struct inode *journal = NULL;
+	int journal_dirty = 0;
 
 	for(slot = 0; slot < osb->max_slots; slot++) {
-		journal = ocfs2_get_system_file_inode(osb,
-						      JOURNAL_SYSTEM_INODE,
-						      slot);
-		if (!journal || is_bad_inode(journal)) {
-			ret = -EACCES;
-			mlog_errno(ret);
-			goto out;
-		}
-
-		di_bh = NULL;
-		ret = ocfs2_read_block(osb, OCFS2_I(journal)->ip_blkno, &di_bh,
-				       0, journal);
-		if (ret < 0) {
+		ret = ocfs2_read_journal_inode(osb, slot, &di_bh, NULL);
+		if (ret) {
 			mlog_errno(ret);
 			goto out;
 		}
 
 		di = (struct ocfs2_dinode *) di_bh->b_data;
 
+		osb->slot_recovery_generations[slot] =
+					ocfs2_get_recovery_generation(di);
+
 		if (le32_to_cpu(di->id1.journal1.ij_flags) &
 		    OCFS2_JOURNAL_DIRTY_FL)
-			ret = -EROFS;
+			journal_dirty = 1;
 
 		brelse(di_bh);
-		if (ret)
-			break;
+		di_bh = NULL;
 	}
 
 out:
-	if (journal)
-		iput(journal);
-
+	if (journal_dirty)
+		ret = -EROFS;
 	return ret;
 }
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index db82be2532e..2178ebffa05 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -161,7 +161,8 @@ int    ocfs2_journal_init(struct ocfs2_journal *journal,
 void   ocfs2_journal_shutdown(struct ocfs2_super *osb);
 int    ocfs2_journal_wipe(struct ocfs2_journal *journal,
 			  int full);
-int    ocfs2_journal_load(struct ocfs2_journal *journal, int local);
+int    ocfs2_journal_load(struct ocfs2_journal *journal, int local,
+			  int replayed);
 int    ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
 void   ocfs2_recovery_thread(struct ocfs2_super *osb,
 			     int node_num);
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 1cb814be8ef..7f625f2b111 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -204,6 +204,8 @@ struct ocfs2_super
 
 	struct ocfs2_slot_info *slot_info;
 
+	u32 *slot_recovery_generations;
+
 	spinlock_t node_map_lock;
 
 	u64 root_blkno;
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 3f194517762..4f619850ccf 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -660,7 +660,10 @@ struct ocfs2_dinode {
 		struct {		/* Info for journal system
 					   inodes */
 			__le32 ij_flags;	/* Mounted, version, etc. */
-			__le32 ij_pad;
+			__le32 ij_recovery_generation; /* Incremented when the
+							  journal is recovered
+							  after an unclean
+							  shutdown */
 		} journal1;
 	} id1;				/* Inode type dependant 1 */
 /*C0*/	union {
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index 2560b33889a..88255d3f52b 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -1442,6 +1442,15 @@ static int ocfs2_initialize_super(struct super_block *sb,
 	}
 	mlog(0, "max_slots for this device: %u\n", osb->max_slots);
 
+	osb->slot_recovery_generations =
+		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
+			GFP_KERNEL);
+	if (!osb->slot_recovery_generations) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
 	init_waitqueue_head(&osb->osb_wipe_event);
 	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
 					sizeof(*osb->osb_orphan_wipes),
@@ -1703,7 +1712,7 @@ static int ocfs2_check_volume(struct ocfs2_super *osb)
 	local = ocfs2_mount_local(osb);
 
 	/* will play back anything left in the journal. */
-	status = ocfs2_journal_load(osb->journal, local);
+	status = ocfs2_journal_load(osb->journal, local, dirty);
 	if (status < 0) {
 		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
 		goto finally;
@@ -1768,6 +1777,7 @@ static void ocfs2_delete_osb(struct ocfs2_super *osb)
 	ocfs2_free_slot_info(osb);
 
 	kfree(osb->osb_orphan_wipes);
+	kfree(osb->slot_recovery_generations);
 	/* FIXME
 	 * This belongs in journal shutdown, but because we have to
 	 * allocate osb->journal at the start of ocfs2_initalize_osb(),