Infiniband: Fix potential NULL d_inode dereference
authorDavid Howells <dhowells@redhat.com>
Tue, 27 Jan 2015 15:01:18 +0000 (15:01 +0000)
committerAl Viro <viro@zeniv.linux.org.uk>
Fri, 20 Feb 2015 09:56:45 +0000 (04:56 -0500)
Code that does this:

if (!(d_unhashed(tmp) && tmp->d_inode)) {
...
simple_unlink(parent->d_inode, tmp);
}

is broken because:

!(d_unhashed(tmp) && tmp->d_inode)

is equivalent to:

!d_unhashed(tmp) || !tmp->d_inode

so it is possible to get into simple_unlink() with tmp->d_inode == NULL.

simple_unlink(), however, assumes tmp->d_inode cannot be NULL.

I think that what was meant is this:

!d_unhashed(tmp) && tmp->d_inode

and that the logical-not operator or the final close-bracket was misplaced.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Bryan O'Sullivan <bos@pathscale.com>
cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
drivers/infiniband/hw/ipath/ipath_fs.c
drivers/infiniband/hw/qib/qib_fs.c

index 4977082e081f2303542d34ef230c5d199b2b2f31..33c45dfcbd88cb11a05c637e22159cc34008a718 100644 (file)
@@ -277,7 +277,7 @@ static int remove_file(struct dentry *parent, char *name)
        }
 
        spin_lock(&tmp->d_lock);
-       if (!(d_unhashed(tmp) && tmp->d_inode)) {
+       if (!d_unhashed(tmp) && tmp->d_inode) {
                dget_dlock(tmp);
                __d_drop(tmp);
                spin_unlock(&tmp->d_lock);
index 81854586c081fee37e0b3933a26d08e5e8b4e815..d2427640d7e99a7cdab8329dd0f8f94b3a10bad0 100644 (file)
@@ -455,7 +455,7 @@ static int remove_file(struct dentry *parent, char *name)
        }
 
        spin_lock(&tmp->d_lock);
-       if (!(d_unhashed(tmp) && tmp->d_inode)) {
+       if (!d_unhashed(tmp) && tmp->d_inode) {
                __d_drop(tmp);
                spin_unlock(&tmp->d_lock);
                simple_unlink(parent->d_inode, tmp);