net: fix NULL dereferences in check_peer_redir()

[ Upstream commit d3aaeb38c40e5a6c08dd31a1b64da65c4352be36, along
  with dependent backports of commits:
     69cce1d1404968f78b177a0314f5822d5afdbbfb
     9de79c127cccecb11ae6a21ab1499e87aa222880
     218fa90f072e4aeff9003d57e390857f4f35513e
     580da35a31f91a594f3090b7a2c39b85cb051a12
     f7e57044eeb1841847c24aa06766c8290c202583
     e049f28883126c689cf95859480d9ee4ab23b7fa ]

Gergely Kalman reported crashes in check_peer_redir().

It appears commit f39925dbde778 (ipv4: Cache learned redirect
information in inetpeer.) added a race, leading to possible NULL ptr
dereference.

Since we can now change dst neighbour, we should make sure a reader can
safely use a neighbour.

Add RCU protection to dst neighbour, and make sure check_peer_redir()
can be called safely by different cpus in parallel.

As neighbours are already freed after one RCU grace period, this patch
should not add typical RCU penalty (cache cold effects)

Many thanks to Gergely for providing a pretty report pointing to the
bug.

Reported-by: Gergely Kalman <synapse@hippy.csoma.elte.hu>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

2 files changed, 55 insertions, 28 deletions

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 86addca9ddf..a98c414978e 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -555,14 +555,17 @@ static int path_rec_start(struct net_device *dev,
 	return 0;
 }
 
+/* called with rcu_read_lock */
 static void neigh_add_path(struct sk_buff *skb, struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_path *path;
 	struct ipoib_neigh *neigh;
+	struct neighbour *n;
 	unsigned long flags;
 
-	neigh = ipoib_neigh_alloc(skb_dst(skb)->neighbour, skb->dev);
+	n = dst_get_neighbour(skb_dst(skb));
+	neigh = ipoib_neigh_alloc(n, skb->dev);
 	if (!neigh) {
 		++dev->stats.tx_dropped;
 		dev_kfree_skb_any(skb);
@@ -571,9 +574,9 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev)
 
 	spin_lock_irqsave(&priv->lock, flags);
 
-	path = __path_find(dev, skb_dst(skb)->neighbour->ha + 4);
+	path = __path_find(dev, n->ha + 4);
 	if (!path) {
-		path = path_rec_create(dev, skb_dst(skb)->neighbour->ha + 4);
+		path = path_rec_create(dev, n->ha + 4);
 		if (!path)
 			goto err_path;
 
@@ -607,7 +610,7 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev)
 			}
 		} else {
 			spin_unlock_irqrestore(&priv->lock, flags);
-			ipoib_send(dev, skb, path->ah, IPOIB_QPN(skb_dst(skb)->neighbour->ha));
+			ipoib_send(dev, skb, path->ah, IPOIB_QPN(n->ha));
 			return;
 		}
 	} else {
@@ -634,20 +637,24 @@ err_drop:
 	spin_unlock_irqrestore(&priv->lock, flags);
 }
 
+/* called with rcu_read_lock */
 static void ipoib_path_lookup(struct sk_buff *skb, struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(skb->dev);
+	struct dst_entry *dst = skb_dst(skb);
+	struct neighbour *n;
 
 	/* Look up path record for unicasts */
-	if (skb_dst(skb)->neighbour->ha[4] != 0xff) {
+	n = dst_get_neighbour(dst);
+	if (n->ha[4] != 0xff) {
 		neigh_add_path(skb, dev);
 		return;
 	}
 
 	/* Add in the P_Key for multicasts */
-	skb_dst(skb)->neighbour->ha[8] = (priv->pkey >> 8) & 0xff;
-	skb_dst(skb)->neighbour->ha[9] = priv->pkey & 0xff;
-	ipoib_mcast_send(dev, skb_dst(skb)->neighbour->ha + 4, skb);
+	n->ha[8] = (priv->pkey >> 8) & 0xff;
+	n->ha[9] = priv->pkey & 0xff;
+	ipoib_mcast_send(dev, n->ha + 4, skb);
 }
 
 static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev,
@@ -712,18 +719,23 @@ static int ipoib_start_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_neigh *neigh;
+	struct neighbour *n = NULL;
 	unsigned long flags;
 
-	if (likely(skb_dst(skb) && skb_dst(skb)->neighbour)) {
-		if (unlikely(!*to_ipoib_neigh(skb_dst(skb)->neighbour))) {
+	rcu_read_lock();
+	if (likely(skb_dst(skb)))
+		n = dst_get_neighbour(skb_dst(skb));
+
+	if (likely(n)) {
+		if (unlikely(!*to_ipoib_neigh(n))) {
 			ipoib_path_lookup(skb, dev);
-			return NETDEV_TX_OK;
+			goto unlock;
 		}
 
-		neigh = *to_ipoib_neigh(skb_dst(skb)->neighbour);
+		neigh = *to_ipoib_neigh(n);
 
 		if (unlikely((memcmp(&neigh->dgid.raw,
-				     skb_dst(skb)->neighbour->ha + 4,
+				     n->ha + 4,
 				     sizeof(union ib_gid))) ||
 			     (neigh->dev != dev))) {
 			spin_lock_irqsave(&priv->lock, flags);
@@ -740,17 +752,17 @@ static int ipoib_start_xmit(struct sk_buff *skb, struct net_device *dev)
 			ipoib_neigh_free(dev, neigh);
 			spin_unlock_irqrestore(&priv->lock, flags);
 			ipoib_path_lookup(skb, dev);
-			return NETDEV_TX_OK;
+			goto unlock;
 		}
 
 		if (ipoib_cm_get(neigh)) {
 			if (ipoib_cm_up(neigh)) {
 				ipoib_cm_send(dev, skb, ipoib_cm_get(neigh));
-				return NETDEV_TX_OK;
+				goto unlock;
 			}
 		} else if (neigh->ah) {
-			ipoib_send(dev, skb, neigh->ah, IPOIB_QPN(skb_dst(skb)->neighbour->ha));
-			return NETDEV_TX_OK;
+			ipoib_send(dev, skb, neigh->ah, IPOIB_QPN(n->ha));
+			goto unlock;
 		}
 
 		if (skb_queue_len(&neigh->queue) < IPOIB_MAX_PATH_REC_QUEUE) {
@@ -784,13 +796,14 @@ static int ipoib_start_xmit(struct sk_buff *skb, struct net_device *dev)
 					   phdr->hwaddr + 4);
 				dev_kfree_skb_any(skb);
 				++dev->stats.tx_dropped;
-				return NETDEV_TX_OK;
+				goto unlock;
 			}
 
 			unicast_arp_send(skb, dev, phdr);
 		}
 	}
-
+unlock:
+	rcu_read_unlock();
 	return NETDEV_TX_OK;
 }
 
@@ -812,6 +825,8 @@ static int ipoib_hard_header(struct sk_buff *skb,
 			     const void *daddr, const void *saddr, unsigned len)
 {
 	struct ipoib_header *header;
+	struct dst_entry *dst;
+	struct neighbour *n;
 
 	header = (struct ipoib_header *) skb_push(skb, sizeof *header);
 
@@ -823,7 +838,11 @@ static int ipoib_hard_header(struct sk_buff *skb,
 	 * destination address onto the front of the skb so we can
 	 * figure out where to send the packet later.
 	 */
-	if ((!skb_dst(skb) || !skb_dst(skb)->neighbour) && daddr) {
+	dst = skb_dst(skb);
+	n = NULL;
+	if (dst)
+		n = dst_get_neighbour_raw(dst);
+	if ((!dst || !n) && daddr) {
 		struct ipoib_pseudoheader *phdr =
 			(struct ipoib_pseudoheader *) skb_push(skb, sizeof *phdr);
 		memcpy(phdr->hwaddr, daddr, INFINIBAND_ALEN);
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
index 3871ac66355..a8d2a891b84 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
@@ -258,11 +258,15 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast,
 	netif_tx_lock_bh(dev);
 	while (!skb_queue_empty(&mcast->pkt_queue)) {
 		struct sk_buff *skb = skb_dequeue(&mcast->pkt_queue);
+		struct dst_entry *dst = skb_dst(skb);
+		struct neighbour *n = NULL;
+
 		netif_tx_unlock_bh(dev);
 
 		skb->dev = dev;
-
-		if (!skb_dst(skb) || !skb_dst(skb)->neighbour) {
+		if (dst)
+			n = dst_get_neighbour_raw(dst);
+		if (!dst || !n) {
 			/* put pseudoheader back on for next time */
 			skb_push(skb, sizeof (struct ipoib_pseudoheader));
 		}
@@ -715,11 +719,15 @@ void ipoib_mcast_send(struct net_device *dev, void *mgid, struct sk_buff *skb)
 
 out:
 	if (mcast && mcast->ah) {
-		if (skb_dst(skb)		&&
-		    skb_dst(skb)->neighbour &&
-		    !*to_ipoib_neigh(skb_dst(skb)->neighbour)) {
-			struct ipoib_neigh *neigh = ipoib_neigh_alloc(skb_dst(skb)->neighbour,
-									skb->dev);
+		struct dst_entry *dst = skb_dst(skb);
+		struct neighbour *n = NULL;
+
+		rcu_read_lock();
+		if (dst)
+			n = dst_get_neighbour(dst);
+		if (n && !*to_ipoib_neigh(n)) {
+			struct ipoib_neigh *neigh = ipoib_neigh_alloc(n,
+								      skb->dev);
 
 			if (neigh) {
 				kref_get(&mcast->ah->ref);
@@ -727,7 +735,7 @@ out:
 				list_add_tail(&neigh->list, &mcast->neigh_list);
 			}
 		}
-
+		rcu_read_unlock();
 		spin_unlock_irqrestore(&priv->lock, flags);
 		ipoib_send(dev, skb, mcast->ah, IB_MULTICAST_QPN);
 		return;