lookup_open(): lock the parent shared unless O_CREAT is given

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 12c57abdaac934d9c7a37c2ef42d2b44de4d19d4..46f3bb7a02f5fc783327b724227926b774d432dd 100644 (file)
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -575,3 +575,6 @@ in your dentry operations instead.
 
        Old method is only used if the new one is absent; eventually it will
        be removed.  Switch while you still can; the old one won't stay.
+--
+[mandatory]
+       ->atomic_open() calls without O_CREAT may happen in parallel.

diff --git a/fs/namei.c b/fs/namei.c
index b84e6b2e19e98a438fc7dea997eb852ec2841aa6..01069ddda3e3cf7c36ef6037cd050411d7ad4045 100644 (file)
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -3084,7 +3084,7 @@ static int do_last(struct nameidata *nd,
        }
 
 retry_lookup:
-       if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) {
+       if (open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) {
                error = mnt_want_write(nd->path.mnt);
                if (!error)
                        got_write = true;
@@ -3094,9 +3094,15 @@ retry_lookup:
                 * dropping this one anyway.
                 */
        }
-       inode_lock(dir->d_inode);
+       if (open_flag & O_CREAT)
+               inode_lock(dir->d_inode);
+       else
+               inode_lock_shared(dir->d_inode);
        error = lookup_open(nd, &path, file, op, got_write, opened);
-       inode_unlock(dir->d_inode);
+       if (open_flag & O_CREAT)
+               inode_unlock(dir->d_inode);
+       else
+               inode_unlock_shared(dir->d_inode);
 
        if (error <= 0) {
                if (error)