configfs_rmdir() needs to lock recursively a whole tree of i_mutex in order to
provide userspace and client sub-systems with atomic semantics. However,
vfs_rmdir() locks the to-be-removed inode with sub-class I_MUTEX_NORMAL, which
does not allow futher I_MUTEX_CHILD + x nested locks. Luckily this is a case of
I_MUTEX_PARENT -> I_MUTEX_CHILD lock dependency pattern, where do_rmdir() takes
parent's inode with sub-class I_MUTEX_PARENT.

This patch changes the sub-class of i_mutex lock in vfs_rmdir() from
I_MUTEX_NORMAL to I_MUTEX_CHILD, which allows further nested i_mutex locking as
configfs needs. Existing lock dependencies should not be impacted since
I_MUTEX_CHILD can only preceed I_MUTEX_NORMAL in locking order, and nobody locks
with sub-class I_MUTEX_CHILD before calling vfs_rmdir().

Signed-off-by: Louis Rilling
fs/namei.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

Index: b/fs/namei.c
================================================== =================
--- a/fs/namei.c 2008-05-21 09:40:26.000000000 +0200
+++ b/fs/namei.c 2008-05-22 12:21:27.000000000 +0200
@@ -2232,7 +2232,7 @@ int vfs_rmdir(struct inode *dir, struct


- mutex_lock(&dentry->d_inode->i_mutex);
+ mutex_lock_nested(&dentry->d_inode->i_mutex, I_MUTEX_CHILD);
if (d_mountpoint(dentry))
error = -EBUSY;

Dr Louis Rilling Kerlabs
Skype: louis.rilling Batiment Germanium
Phone: (+33|0) 6 80 89 08 23 80 avenue des Buttes de Coesmes 35700 Rennes

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at