Linux Ethernet Bridging - Jun 2018 - [Bridge] [PATCH] net: bridge: fix potential null pointer dereference on return from br_port_get_rtnl()

br_port_get_rtnl() can return NULL if the network device is not a bridge
port (IFF_BRIDGE_PORT flag not set). br_port_slave_changelink() and
br_port_fill_slave_info() callbacks dereference this pointer without
checking. Currently this is not a problem because slave devices always
set this flag. Add null check in case these conditions ever change.

Detected by CoverityScan, CID 1339613 ("Dereference null return
value")

Signed-off-by: Garry McNulty <garrmcnu at gmail.com>
---
 net/bridge/br_netlink.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
index 9f5eb05b0373..b3ad135b7157 100644
--- a/net/bridge/br_netlink.c
+++ b/net/bridge/br_netlink.c
@@ -947,13 +947,14 @@ static int br_port_slave_changelink(struct net_device
*brdev,
 				    struct netlink_ext_ack *extack)
 {
 	struct net_bridge *br = netdev_priv(brdev);
+	struct net_bridge_port *p = br_port_get_rtnl(dev);
 	int ret;
 
-	if (!data)
+	if (!data || !p)
 		return 0;
 
 	spin_lock_bh(&br->lock);
-	ret = br_setport(br_port_get_rtnl(dev), data);
+	ret = br_setport(p, data);
 	spin_unlock_bh(&br->lock);
 
 	return ret;
@@ -963,7 +964,9 @@ static int br_port_fill_slave_info(struct sk_buff *skb,
 				   const struct net_device *brdev,
 				   const struct net_device *dev)
 {
-	return br_port_fill_attrs(skb, br_port_get_rtnl(dev));
+	struct net_bridge_port *p = br_port_get_rtnl(dev);
+
+	return p ? br_port_fill_attrs(skb, p) : -EINVAL;
 }
 
 static size_t br_port_get_slave_size(const struct net_device *brdev,
-- 
2.14.4

From: Garry McNulty <garrmcnu at gmail.com>
Date: Thu, 21 Jun 2018 21:14:27 +0100
> br_port_get_rtnl() can return NULL if the network device is not a bridge
> port (IFF_BRIDGE_PORT flag not set). br_port_slave_changelink() and
> br_port_fill_slave_info() callbacks dereference this pointer without
> checking. Currently this is not a problem because slave devices always
> set this flag. Add null check in case these conditions ever change.
> 
> Detected by CoverityScan, CID 1339613 ("Dereference null return
value")
> 
> Signed-off-by: Garry McNulty <garrmcnu at gmail.com>
I don't think this is reasonable.

The bridge code will never, ever, install a slave that doesn't have
that bit set.  It's the most fundamental aspect of how these objects
are managed.