net: bonding: use atomic instead of rtnl_mutex, to make sure peer notify updated

Using atomic to protect the send_peer_notif instead of rtnl_mutex.
This approach allows safe updates in both interrupt and process
contexts, while avoiding code complexity.

In lacp mode, the rtnl might be locked, preventing ad_cond_set_peer_notif()
from acquiring the lock and updating send_peer_notif. This patch addresses
the issue by using a atomic. Since updating send_peer_notif does not
require high real-time performance, such atomic updates are acceptable.

After coverting the rtnl lock for send_peer_notif to atomic, in bond_mii_monitor(),
we should check the should_notify_peers (rtnllock required) instead of
send_peer_notif. By the way, to avoid peer notify event loss, we check
again whether to send peer notify, such as active-backup mode failover.

Cc: Jay Vosburgh <jv@jvosburgh.net>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Simon Horman <horms@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Andrew Lunn <andrew+netdev@lunn.ch>
Cc: Nikolay Aleksandrov <razor@blackwall.org>
Cc: Hangbin Liu <liuhangbin@gmail.com>
Suggested-by: Jay Vosburgh <jv@jvosburgh.net>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
Signed-off-by: NipaLocal <nipa@local>

Author: hao022

drivers/net/bonding/bond_3ad.c

@@ -999,11 +999,8 @@ static void ad_cond_set_peer_notif(struct port *port)
 {
 	struct bonding *bond = port->slave->bond;
 
-	if (bond->params.broadcast_neighbor && rtnl_trylock()) {
-		bond->send_peer_notif = bond->params.num_peer_notif *
-			max(1, bond->params.peer_notif_delay);
-		rtnl_unlock();
-	}
+	if (bond->params.broadcast_neighbor)
+		bond_peer_notify_reset(bond);
 }
 
 /**

drivers/net/bonding/bond_main.c

@@ -1167,10 +1167,11 @@ static bool bond_should_notify_peers(struct bonding *bond)
 {
 	struct bond_up_slave *usable;
 	struct slave *slave = NULL;
+	int send_peer_notif;
 
-	if (!bond->send_peer_notif ||
-	    bond->send_peer_notif %
-	    max(1, bond->params.peer_notif_delay) != 0 ||
+	send_peer_notif = atomic_read(&bond->send_peer_notif);
+	if (!send_peer_notif ||
+	    send_peer_notif % max(1, bond->params.peer_notif_delay) != 0 ||
 	    !netif_carrier_ok(bond->dev))
 		return false;
 
@@ -1270,8 +1271,6 @@ void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
 						      BOND_SLAVE_NOTIFY_NOW);
 
 		if (new_active) {
-			bool should_notify_peers = false;
-
 			bond_set_slave_active_flags(new_active,
 						    BOND_SLAVE_NOTIFY_NOW);
 
@@ -1280,19 +1279,17 @@ void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
 						      old_active);
 
 			if (netif_running(bond->dev)) {
-				bond->send_peer_notif =
-					bond->params.num_peer_notif *
-					max(1, bond->params.peer_notif_delay);
-				should_notify_peers =
-					bond_should_notify_peers(bond);
+				bond_peer_notify_reset(bond);
+
+				if (bond_should_notify_peers(bond)) {
+					atomic_dec(&bond->send_peer_notif);
+					call_netdevice_notifiers(
+							NETDEV_NOTIFY_PEERS,
+							bond->dev);
+				}
 			}
 
 			call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
-			if (should_notify_peers) {
-				bond->send_peer_notif--;
-				call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
-							 bond->dev);
-			}
 		}
 	}
 
@@ -2801,7 +2798,7 @@ static void bond_mii_monitor(struct work_struct *work)
 
 	rcu_read_unlock();
 
-	if (commit || bond->send_peer_notif) {
+	if (commit || should_notify_peers) {
 		/* Race avoidance with bond_close cancel of workqueue */
 		if (!rtnl_trylock()) {
 			delay = 1;
@@ -2816,16 +2813,15 @@ static void bond_mii_monitor(struct work_struct *work)
 			bond_miimon_commit(bond);
 		}
 
-		if (bond->send_peer_notif) {
-			bond->send_peer_notif--;
-			if (should_notify_peers)
-				call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
-							 bond->dev);
-		}
+		/* check again to avoid send_peer_notif has been changed. */
+		if (bond_should_notify_peers(bond))
+			call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
 
 		rtnl_unlock();	/* might sleep, hold no other locks */
 	}
 
+	atomic_dec_if_positive(&bond->send_peer_notif);
+
 re_arm:
 	if (bond->params.miimon)
 		queue_delayed_work(bond->wq, &bond->mii_work, delay);
@@ -3773,7 +3769,7 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 			return;
 
 		if (should_notify_peers) {
-			bond->send_peer_notif--;
+			atomic_dec(&bond->send_peer_notif);
 			call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
 						 bond->dev);
 		}
@@ -4267,6 +4263,8 @@ static int bond_open(struct net_device *bond_dev)
 			queue_delayed_work(bond->wq, &bond->alb_work, 0);
 	}
 
+	atomic_set(&bond->send_peer_notif, 0);
+
 	if (bond->params.miimon)  /* link check interval, in milliseconds. */
 		queue_delayed_work(bond->wq, &bond->mii_work, 0);
 
@@ -4300,7 +4298,7 @@ static int bond_close(struct net_device *bond_dev)
 	struct slave *slave;
 
 	bond_work_cancel_all(bond);
-	bond->send_peer_notif = 0;
+	atomic_set(&bond->send_peer_notif, 0);
 	if (bond_is_lb(bond))
 		bond_alb_deinitialize(bond);
 	bond->recv_probe = NULL;

include/net/bonding.h

@@ -236,7 +236,7 @@ struct bonding {
 	 */
 	spinlock_t mode_lock;
 	spinlock_t stats_lock;
-	u32	 send_peer_notif;
+	atomic_t send_peer_notif;
 	u8       igmp_retrans;
 #ifdef CONFIG_PROC_FS
 	struct   proc_dir_entry *proc_entry;
@@ -814,4 +814,11 @@ static inline netdev_tx_t bond_tx_drop(struct net_device *dev, struct sk_buff *s
 	return NET_XMIT_DROP;
 }
 
+static inline void bond_peer_notify_reset(struct bonding *bond)
+{
+	atomic_set(&bond->send_peer_notif,
+		bond->params.num_peer_notif *
+		max(1, bond->params.peer_notif_delay));
+}
+
 #endif /* _NET_BONDING_H */