net: ipv4: update fnhe_pmtu when first hop's MTU changes

Since commit 5aad1de5ea2c ("ipv4: use separate genid for next hop
exceptions"), exceptions get deprecated separately from cached
routes. In particular, administrative changes don't clear PMTU anymore.

As Stefano described in commit e9fa1495d738 ("ipv6: Reflect MTU changes
on PMTU of exceptions for MTU-less routes"), the PMTU discovered before
the local MTU change can become stale:
 - if the local MTU is now lower than the PMTU, that PMTU is now
   incorrect
 - if the local MTU was the lowest value in the path, and is increased,
   we might discover a higher PMTU

Similarly to what commit e9fa1495d738 did for IPv6, update PMTU in those
cases.

If the exception was locked, the discovered PMTU was smaller than the
minimal accepted PMTU. In that case, if the new local MTU is smaller
than the current PMTU, let PMTU discovery figure out if locking of the
exception is still needed.

To do this, we need to know the old link MTU in the NETDEV_CHANGEMTU
notifier. By the time the notifier is called, dev->mtu has been
changed. This patch adds the old MTU as additional information in the
notifier structure, and a new call_netdevice_notifiers_u32() function.

Fixes: 5aad1de5ea2c ("ipv4: use separate genid for next hop exceptions")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/netdevice.h

@@ -2458,6 +2458,13 @@ struct netdev_notifier_info {
 	struct netlink_ext_ack	*extack;
 };
 
+struct netdev_notifier_info_ext {
+	struct netdev_notifier_info info; /* must be first */
+	union {
+		u32 mtu;
+	} ext;
+};
+
 struct netdev_notifier_change_info {
 	struct netdev_notifier_info info; /* must be first */
 	unsigned int flags_changed;

include/net/ip_fib.h

@@ -394,6 +394,7 @@ int ip_fib_check_default(__be32 gw, struct net_device *dev);
 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force);
 int fib_sync_down_addr(struct net_device *dev, __be32 local);
 int fib_sync_up(struct net_device *dev, unsigned int nh_flags);
+void fib_sync_mtu(struct net_device *dev, u32 orig_mtu);
 
 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,

net/core/dev.c

@@ -1752,6 +1752,28 @@ int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
 }
 EXPORT_SYMBOL(call_netdevice_notifiers);
 
+/**
+ *	call_netdevice_notifiers_mtu - call all network notifier blocks
+ *	@val: value passed unmodified to notifier function
+ *	@dev: net_device pointer passed unmodified to notifier function
+ *	@arg: additional u32 argument passed to the notifier function
+ *
+ *	Call all network notifier blocks.  Parameters and return value
+ *	are as for raw_notifier_call_chain().
+ */
+static int call_netdevice_notifiers_mtu(unsigned long val,
+					struct net_device *dev, u32 arg)
+{
+	struct netdev_notifier_info_ext info = {
+		.info.dev = dev,
+		.ext.mtu = arg,
+	};
+
+	BUILD_BUG_ON(offsetof(struct netdev_notifier_info_ext, info) != 0);
+
+	return call_netdevice_notifiers_info(val, &info.info);
+}
+
 #ifdef CONFIG_NET_INGRESS
 static DEFINE_STATIC_KEY_FALSE(ingress_needed_key);
 
@@ -7574,14 +7596,16 @@ int dev_set_mtu_ext(struct net_device *dev, int new_mtu,
 	err = __dev_set_mtu(dev, new_mtu);
 
 	if (!err) {
-		err = call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
+		err = call_netdevice_notifiers_mtu(NETDEV_CHANGEMTU, dev,
+						   orig_mtu);
 		err = notifier_to_errno(err);
 		if (err) {
 			/* setting mtu back and notifying everyone again,
 			 * so that they have a chance to revert changes.
 			 */
 			__dev_set_mtu(dev, orig_mtu);
-			call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
+			call_netdevice_notifiers_mtu(NETDEV_CHANGEMTU, dev,
+						     new_mtu);
 		}
 	}
 	return err;

net/ipv4/fib_frontend.c

@@ -1243,7 +1243,8 @@ 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;
@@ -1278,16 +1279,19 @@ 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;
 	}

net/ipv4/fib_semantics.c

@@ -1470,6 +1470,56 @@ 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