leases: fix write-open/read-lease race

In setlease, we use i_writecount to decide whether we can give out a
read lease.

In open, we break leases before incrementing i_writecount.

There is therefore a window between the break lease and the i_writecount
increment when setlease could add a new read lease.

This would leave us with a simultaneous write open and read lease, which
shouldn't happen.
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

fs/namei.c

@@ -2035,10 +2035,7 @@ static int may_open(struct path *path, int acc_mode, int flag)
 	if (flag & O_NOATIME && !inode_owner_or_capable(inode))
 		return -EPERM;
 
-	/*
-	 * Ensure there are no outstanding leases on the file.
-	 */
-	return break_lease(inode, flag);
+	return 0;
 }
 
 static int handle_truncate(struct file *filp)

fs/open.c

@@ -685,6 +685,10 @@ static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 	if (error)
 		goto cleanup_all;
 
+	error = break_lease(inode, f->f_flags);
+	if (error)
+		goto cleanup_all;
+
 	if (!open && f->f_op)
 		open = f->f_op->open;
 	if (open) {