[PATCH v2] ntfs: fix mrec_lock ABBA deadlock in rename

[Peiyang He]

ntfs_file_fsync(), ntfs_dir_fsync() and __ntfs_write_inode() lock an
inode's mrec_lock before taking the mrec_lock of its parent directory.

ntfs_rename() takes old_ni->mrec_lock and old_dir_ni->mrec_lock
before taking new_ni->mrec_lock for an existing target, or
new_dir_ni->mrec_lock for a cross-directory rename.
This can deadlock when ntfs_file_fsync() or __ntfs_write_inode() holds
the target inode, or when ntfs_dir_fsync() holds a child target
directory, while rename() holds the parent directory and waits for the
target.

Fix this by locking the existing target inode before taking any parent
directory mrec_lock. For cross-directory renames where the target parent
is a descendant of the source parent, lock the target parent before the
source parent so the directory order matches the child-to-parent order used
by ntfs_file_fsync(), ntfs_dir_fsync(), and __ntfs_write_inode().

Reported-by: Peiyang He <peiyang_he@xxxxxxxxxxxxxxxx>
Closes: https://lore.kernel.org/all/C4D296F0E9F3D66C+9397ffbc-eb55-44bb-9b3f-5da4809e7955@xxxxxxxxxxxxxxxx/
Fixes: af0db57d4293 ("ntfs: update inode operations")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Peiyang He <peiyang_he@xxxxxxxxxxxxxxxx>
Assisted-by: Codex:gpt-5.5
---
Changes in v2:
 - fix the NTFS module build by using exported is_subdir() instead of
   unexported d_ancestor() (suggested by kernel test robot)
 - move the ancestor relationship check before taking any mrec_lock

 fs/ntfs/namei.c | 62 ++++++++++++++++++++++++-------------------------
 1 file changed, 31 insertions(+), 31 deletions(-)

diff --git a/fs/ntfs/namei.c b/fs/ntfs/namei.c
index a19626a135bd..5ff25e9aaa32 100644
--- a/fs/ntfs/namei.c
+++ b/fs/ntfs/namei.c
@@ -1266,6 +1266,7 @@ static int ntfs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
 	struct ntfs_volume *vol = NTFS_SB(sb);
 	struct ntfs_inode *old_ni, *new_ni = NULL;
 	struct ntfs_inode *old_dir_ni = NTFS_I(old_dir), *new_dir_ni = NTFS_I(new_dir);
+	bool new_dir_first = false;
 
 	if (NVolShutdown(old_dir_ni->vol))
 		return -EIO;
@@ -1301,36 +1302,39 @@ static int ntfs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
 	old_inode = old_dentry->d_inode;
 	new_inode = new_dentry->d_inode;
 	old_ni = NTFS_I(old_inode);
+	if (new_inode)
+		new_ni = NTFS_I(new_inode);
+	if (old_dir != new_dir)
+		new_dir_first = is_subdir(new_dentry->d_parent,
+					  old_dentry->d_parent);
 
 	if (!(vol->vol_flags & VOLUME_IS_DIRTY))
 		ntfs_set_volume_flags(vol, VOLUME_IS_DIRTY);
 
 	mutex_lock_nested(&old_ni->mrec_lock, NTFS_INODE_MUTEX_NORMAL);
-	mutex_lock_nested(&old_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT);
+	if (new_ni)
+		mutex_lock_nested(&new_ni->mrec_lock, NTFS_INODE_MUTEX_NORMAL_2);
+
+	if (old_dir == new_dir) {
+		mutex_lock_nested(&old_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT);
+	} else if (new_dir_first) {
+		mutex_lock_nested(&new_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT);
+		mutex_lock_nested(&old_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT_2);
+	} else {
+		mutex_lock_nested(&old_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT);
+		mutex_lock_nested(&new_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT_2);
+	}
 
-	if (NInoBeingDeleted(old_ni) || NInoBeingDeleted(old_dir_ni)) {
+	if (NInoBeingDeleted(old_ni) || NInoBeingDeleted(old_dir_ni) ||
+	    (new_ni && NInoBeingDeleted(new_ni)) ||
+	    (old_dir != new_dir && NInoBeingDeleted(new_dir_ni))) {
 		err = -ENOENT;
-		goto unlock_old;
+		goto err_out;
 	}
 
 	is_dir = S_ISDIR(old_inode->i_mode);
 
 	if (new_inode) {
-		new_ni = NTFS_I(new_inode);
-		mutex_lock_nested(&new_ni->mrec_lock, NTFS_INODE_MUTEX_NORMAL_2);
-		if (old_dir != new_dir) {
-			mutex_lock_nested(&new_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT_2);
-			if (NInoBeingDeleted(new_dir_ni)) {
-				err = -ENOENT;
-				goto err_out;
-			}
-		}
-
-		if (NInoBeingDeleted(new_ni)) {
-			err = -ENOENT;
-			goto err_out;
-		}
-
 		if (is_dir) {
 			struct mft_record *ni_mrec;
 
@@ -1348,14 +1352,6 @@ static int ntfs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
 		err = ntfs_delete(new_ni, new_dir_ni, uname_new, new_name_len, false);
 		if (err)
 			goto err_out;
-	} else {
-		if (old_dir != new_dir) {
-			mutex_lock_nested(&new_dir_ni->mrec_lock, NTFS_INODE_MUTEX_PARENT_2);
-			if (NInoBeingDeleted(new_dir_ni)) {
-				err = -ENOENT;
-				goto err_out;
-			}
-		}
 	}
 
 	err = __ntfs_link(old_ni, new_dir_ni, uname_new, new_name_len);
@@ -1386,13 +1382,17 @@ static int ntfs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
 	inode_inc_iversion(new_dir);
 
 err_out:
-	if (old_dir != new_dir)
+	if (old_dir == new_dir) {
+		mutex_unlock(&old_dir_ni->mrec_lock);
+	} else if (new_dir_first) {
+		mutex_unlock(&old_dir_ni->mrec_lock);
 		mutex_unlock(&new_dir_ni->mrec_lock);
-	if (new_inode)
+	} else {
+		mutex_unlock(&new_dir_ni->mrec_lock);
+		mutex_unlock(&old_dir_ni->mrec_lock);
+	}
+	if (new_ni)
 		mutex_unlock(&new_ni->mrec_lock);
-
-unlock_old:
-	mutex_unlock(&old_dir_ni->mrec_lock);
 	mutex_unlock(&old_ni->mrec_lock);
 	if (uname_new)
 		kmem_cache_free(ntfs_name_cache, uname_new);

base-commit: 1a3746ccbb0a97bed3c06ccde6b880013b1dddc1
-- 
2.43.0