4 files changed, 63 insertions, 8 deletions

diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c
index 038f511c73fa..0f1beceb47d5 100644
--- a/net/ipv4/fib_frontend.c
+++ b/net/ipv4/fib_frontend.c
@@ -1291,7 +1291,8 @@ static int fib_inetaddr_event(struct notifier_block *this, unsigned long event,
 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
 {
 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
-	struct netdev_notifier_changeupper_info *info;
+	struct netdev_notifier_changeupper_info *upper_info = ptr;
+	struct netdev_notifier_info_ext *info_ext = ptr;
 	struct in_device *in_dev;
 	struct net *net = dev_net(dev);
 	unsigned int flags;
@@ -1326,16 +1327,19 @@ static int fib_netdev_event(struct notifier_block *this, unsigned long event, vo
 			fib_sync_up(dev, RTNH_F_LINKDOWN);
 		else
 			fib_sync_down_dev(dev, event, false);
-		/* fall through */
+		rt_cache_flush(net);
+		break;
 	case NETDEV_CHANGEMTU:
+		fib_sync_mtu(dev, info_ext->ext.mtu);
 		rt_cache_flush(net);
 		break;
 	case NETDEV_CHANGEUPPER:
-		info = ptr;
+		upper_info = ptr;
 		/* flush all routes if dev is linked to or unlinked from
 		 * an L3 master device (e.g., VRF)
 		 */
-		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
+		if (upper_info->upper_dev &&
+		    netif_is_l3_master(upper_info->upper_dev))
 			fib_disable_ip(dev, NETDEV_DOWN, true);
 		break;
 	}
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index f8c7ec8171a8..b5c3937ca6ec 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -1457,6 +1457,56 @@ static int call_fib_nh_notifiers(struct fib_nh *fib_nh,
 	return NOTIFY_DONE;
 }
 
+/* Update the PMTU of exceptions when:
+ * - the new MTU of the first hop becomes smaller than the PMTU
+ * - the old MTU was the same as the PMTU, and it limited discovery of
+ *   larger MTUs on the path. With that limit raised, we can now
+ *   discover larger MTUs
+ * A special case is locked exceptions, for which the PMTU is smaller
+ * than the minimal accepted PMTU:
+ * - if the new MTU is greater than the PMTU, don't make any change
+ * - otherwise, unlock and set PMTU
+ */
+static void nh_update_mtu(struct fib_nh *nh, u32 new, u32 orig)
+{
+	struct fnhe_hash_bucket *bucket;
+	int i;
+
+	bucket = rcu_dereference_protected(nh->nh_exceptions, 1);
+	if (!bucket)
+		return;
+
+	for (i = 0; i < FNHE_HASH_SIZE; i++) {
+		struct fib_nh_exception *fnhe;
+
+		for (fnhe = rcu_dereference_protected(bucket[i].chain, 1);
+		     fnhe;
+		     fnhe = rcu_dereference_protected(fnhe->fnhe_next, 1)) {
+			if (fnhe->fnhe_mtu_locked) {
+				if (new <= fnhe->fnhe_pmtu) {
+					fnhe->fnhe_pmtu = new;
+					fnhe->fnhe_mtu_locked = false;
+				}
+			} else if (new < fnhe->fnhe_pmtu ||
+				   orig == fnhe->fnhe_pmtu) {
+				fnhe->fnhe_pmtu = new;
+			}
+		}
+	}
+}
+
+void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
+{
+	unsigned int hash = fib_devindex_hashfn(dev->ifindex);
+	struct hlist_head *head = &fib_info_devhash[hash];
+	struct fib_nh *nh;
+
+	hlist_for_each_entry(nh, head, nh_hash) {
+		if (nh->nh_dev == dev)
+			nh_update_mtu(nh, dev->mtu, orig_mtu);
+	}
+}
+
 /* Event              force Flags           Description
  * NETDEV_CHANGE      0     LINKDOWN        Carrier OFF, not for scope host
  * NETDEV_DOWN        0     LINKDOWN|DEAD   Link down, not for scope host
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index f71d2395c428..c0a9d26c06ce 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -1001,21 +1001,22 @@ out:	kfree_skb(skb);
 static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu)
 {
 	struct dst_entry *dst = &rt->dst;
+	u32 old_mtu = ipv4_mtu(dst);
 	struct fib_result res;
 	bool lock = false;
 
 	if (ip_mtu_locked(dst))
 		return;
 
-	if (ipv4_mtu(dst) < mtu)
+	if (old_mtu < mtu)
 		return;
 
 	if (mtu < ip_rt_min_pmtu) {
 		lock = true;
-		mtu = ip_rt_min_pmtu;
+		mtu = min(old_mtu, ip_rt_min_pmtu);
 	}
 
-	if (rt->rt_pmtu == mtu &&
+	if (rt->rt_pmtu == mtu && !lock &&
 	    time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2))
 		return;
 
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 1bec2203d558..cf8252d05a01 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -1627,7 +1627,7 @@ busy_check:
 	*err = error;
 	return NULL;
 }
-EXPORT_SYMBOL_GPL(__skb_recv_udp);
+EXPORT_SYMBOL(__skb_recv_udp);
 
 /*
  * 	This should be easy, if there is something there we