[BLUETOOTH]: Fix endianness bug in l2cap_sock_listen()

We loop through psm values, calling __l2cap_get_sock_by_addr(psm, ...)
until we get NULL; then we set ->psm of our socket to htobs(psm).
IOW, we find unused psm value and put it into our socket.  So far, so
good, but...  __l2cap_get_sock_by_addr() compares its argument with
->psm of sockets.  IOW, the entire thing works correctly only on
little-endian.  On big-endian we'll get "no socket with such psm"
on the first iteration, since we won't find a socket with ->psm == 0x1001.
We will happily conclude that 0x1001 is unused and slap htobs(0x1001)
(i.e. 0x110) into ->psm of our socket.  Of course, the next time around
the same thing will repeat and we'll just get a fsckload of sockets
with the same ->psm assigned.

Fix: pass htobs(psm) to __l2cap_get_sock_by_addr() there.  All other
callers are already passing little-endian values and all places that
store something in ->psm are storing little-endian.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/bluetooth/l2cap.c b/net/bluetooth/l2cap.c
index 670ff95ca64b1b30048ce72eb82584ca5b17baa3..b82cbdd1fcb1a6969c2bb7032ed9ca2f77239367 100644 (file)
--- a/net/bluetooth/l2cap.c
+++ b/net/bluetooth/l2cap.c
@@ -748,7 +748,7 @@ static int l2cap_sock_listen(struct socket *sock, int backlog)
                write_lock_bh(&l2cap_sk_list.lock);
 
                for (psm = 0x1001; psm < 0x1100; psm += 2)
-                       if (!__l2cap_get_sock_by_addr(psm, src)) {
+                       if (!__l2cap_get_sock_by_addr(htobs(psm), src)) {
                                l2cap_pi(sk)->psm   = htobs(psm);
                                l2cap_pi(sk)->sport = htobs(psm);
                                err = 0;