re: ceph: move encode_fh to new API

[Dan Carpenter <dan.carpenter@oracle.com>]

Hello Sage Weil,

This is a semi-automatic email about new static checker warnings.

The patch f59919a07e03: "ceph: move encode_fh to new API" from Apr 5, 
2012, leads to the following Smatch complaint:

fs/ceph/export.c:85 ceph_encode_fh()
	 error: we previously assumed 'dentry' could be null (see line 67)

fs/ceph/export.c
    66		/* if we found an alias, generate a connectable fh */
    67		if (*max_len >= connected_handle_length && dentry) {
                                                           ^^^^^^
New check.

    68			dout("encode_fh %p connectable\n", dentry);
    69			spin_lock(&dentry->d_lock);
    70			parent = dentry->d_parent;
    71			cfh->ino = ceph_ino(inode);
    72			cfh->parent_ino = ceph_ino(parent->d_inode);
    73			cfh->parent_name_hash = ceph_dentry_hash(parent->d_inode,
    74								 dentry);
    75			*max_len = connected_handle_length;
    76			type = 2;
    77			spin_unlock(&dentry->d_lock);
    78		} else if (*max_len >= handle_length) {
    79			if (parent_inode) {
    80				/* nfsd wants connectable */
    81				*max_len = connected_handle_length;
    82				type = 255;
    83			} else {
    84				dout("encode_fh %p\n", dentry);
    85				fh->ino = ceph_ino(dentry->d_inode);
                                                   ^^^^^^^^^^^^^^^
Old dereference.

    86				*max_len = handle_length;
    87				type = 1;

These emails really are mostly automated...  So if it's a false positive
then I blame the script.  Hope it's not too much spam.

regards,
dan carpenter