[PATCH net-next v4 0/4] A series of minor optimizations of the bonding module

[PATCH net-next v4 0/4] A series of minor optimizations of the bonding module

[Tonghao Zhang]

These patches mainly target the peer notify mechanism of the bonding module.
Including updates of peer notify, lock races, etc. For more information, please
refer to the patch.

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>
Cc: Jason Xing <kerneljasonxing@gmail.com>

v4: patch1 keeps the netdevice notifier order
v3: drop the 5/5 patch, net: bonding: combine rtnl lock block for arp monitor in activebackup mode 

Tonghao Zhang (4):
  net: bonding: use workqueue to make sure peer notify updated in lacp
    mode
  net: bonding: move bond_should_notify_peers, e.g. into rtnl lock block
  net: bonding: skip the 2nd trylock when first one fail
  net: bonding: add the READ_ONCE/WRITE_ONCE for outside lock accessing

 drivers/net/bonding/bond_3ad.c  |   7 +--
 drivers/net/bonding/bond_main.c | 101 ++++++++++++++++++++------------
 include/net/bonding.h           |   2 +
 3 files changed, 69 insertions(+), 41 deletions(-)

-- 
2.34.1

[PATCH net-next v4 1/4] net: bonding: use workqueue to make sure peer notify updated in lacp mode

[Tonghao Zhang]

The rtnl lock 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 workqueue. Since updating send_peer_notif does
not require high real-time performance, such delayed updates are entirely
acceptable.

In fact, checking this value and using it in multiple places, all operations
are protected at the same time by rtnl lock, such as
- read send_peer_notif
- send_peer_notif--
- bond_should_notify_peers

By the way, rtnl lock is still required, when accessing bond.params.* for
updating send_peer_notif. In lacp mode, resetting send_peer_notif in
workqueue is safe, simple and effective way.

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>
Cc: Jason Xing <kerneljasonxing@gmail.com>
Suggested-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
---
v4:
- keep the netdevice notifier order.
v2/3:
- no change
v1:
- This patch is actually version v3, https://patchwork.kernel.org/project/netdevbpf/patch/20251118090305.35558-1-tonghao@bamaicloud.com/
- add a comment why we use the trylock.
- add this patch to series
---
 drivers/net/bonding/bond_3ad.c  |  7 ++--
 drivers/net/bonding/bond_main.c | 57 +++++++++++++++++++++++++--------
 include/net/bonding.h           |  2 ++
 3 files changed, 48 insertions(+), 18 deletions(-)

diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
index 1a8de2bf8655..01ae0269a138 100644
--- a/drivers/net/bonding/bond_3ad.c
+++ b/drivers/net/bonding/bond_3ad.c
@@ -1008,11 +1008,8 @@ 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_work_rearm(bond, 0);
 }
 
 /**
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 3d56339a8a10..edf6dac8a98f 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1195,6 +1195,35 @@ static bool bond_should_notify_peers(struct bonding *bond)
 	return true;
 }
 
+/* Use this to update send_peer_notif when RTNL may be held in other places. */
+void bond_peer_notify_work_rearm(struct bonding *bond, unsigned long delay)
+{
+	queue_delayed_work(bond->wq, &bond->peer_notify_work, delay);
+}
+
+/* Peer notify update handler. Holds only RTNL */
+static void bond_peer_notify_reset(struct bonding *bond)
+{
+	bond->send_peer_notif = bond->params.num_peer_notif *
+		max(1, bond->params.peer_notif_delay);
+}
+
+static void bond_peer_notify_handler(struct work_struct *work)
+{
+	struct bonding *bond = container_of(work, struct bonding,
+					    peer_notify_work.work);
+
+	if (!rtnl_trylock()) {
+		bond_peer_notify_work_rearm(bond, 1);
+		return;
+	}
+
+	bond_peer_notify_reset(bond);
+
+	rtnl_unlock();
+	return;
+}
+
 /**
  * bond_change_active_slave - change the active slave into the specified one
  * @bond: our bonding struct
@@ -1270,8 +1299,6 @@ 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);
 
@@ -1279,19 +1306,17 @@ void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
 				bond_do_fail_over_mac(bond, new_active,
 						      old_active);
 
+			call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
+
 			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);
 
-			call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
-			if (should_notify_peers) {
-				bond->send_peer_notif--;
-				call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
-							 bond->dev);
+				if (bond_should_notify_peers(bond)) {
+					bond->send_peer_notif--;
+					call_netdevice_notifiers(
+							NETDEV_NOTIFY_PEERS,
+							bond->dev);
+				}
 			}
 		}
 	}
@@ -4213,6 +4238,10 @@ static u32 bond_xmit_hash_xdp(struct bonding *bond, struct xdp_buff *xdp)
 
 void bond_work_init_all(struct bonding *bond)
 {
+	/* ndo_stop, bond_close() will try to flush the work under
+	 * the rtnl lock. The workqueue must not block on rtnl lock
+	 * to avoid deadlock.
+	 */
 	INIT_DELAYED_WORK(&bond->mcast_work,
 			  bond_resend_igmp_join_requests_delayed);
 	INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
@@ -4220,6 +4249,7 @@ void bond_work_init_all(struct bonding *bond)
 	INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor);
 	INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
 	INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler);
+	INIT_DELAYED_WORK(&bond->peer_notify_work, bond_peer_notify_handler);
 }
 
 void bond_work_cancel_all(struct bonding *bond)
@@ -4230,6 +4260,7 @@ void bond_work_cancel_all(struct bonding *bond)
 	cancel_delayed_work_sync(&bond->ad_work);
 	cancel_delayed_work_sync(&bond->mcast_work);
 	cancel_delayed_work_sync(&bond->slave_arr_work);
+	cancel_delayed_work_sync(&bond->peer_notify_work);
 }
 
 static int bond_open(struct net_device *bond_dev)
diff --git a/include/net/bonding.h b/include/net/bonding.h
index 49edc7da0586..63d08056a4a4 100644
--- a/include/net/bonding.h
+++ b/include/net/bonding.h
@@ -254,6 +254,7 @@ struct bonding {
 	struct   delayed_work ad_work;
 	struct   delayed_work mcast_work;
 	struct   delayed_work slave_arr_work;
+	struct   delayed_work peer_notify_work;
 #ifdef CONFIG_DEBUG_FS
 	/* debugging support via debugfs */
 	struct	 dentry *debug_dir;
@@ -709,6 +710,7 @@ struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev,
 					      int level);
 int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave);
 void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay);
+void bond_peer_notify_work_rearm(struct bonding *bond, unsigned long delay);
 void bond_work_init_all(struct bonding *bond);
 void bond_work_cancel_all(struct bonding *bond);
 
-- 
2.34.1

[PATCH net-next v4 2/4] net: bonding: move bond_should_notify_peers, e.g. into rtnl lock block

[Tonghao Zhang]

This patch tries to avoid the possible peer notify event loss.

In bond_mii_monitor()/bond_activebackup_arp_mon(), when we hold the rtnl lock:
- check send_peer_notif again to avoid unconditionally reducing this value.
- send_peer_notif may have been reset. Therefore, it is necessary to check
  whether to send peer notify via bond_should_notify_peers() to avoid the
  loss of notification events.

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>
Cc: Jason Xing <kerneljasonxing@gmail.com>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
---
v2-4:
- no change
v1:
- splitted from: https://patchwork.kernel.org/project/netdevbpf/patch/20251118090431.35654-1-tonghao@bamaicloud.com/
- this patch only move the bond_should_notify_peers to rtnl lock.
- add this patch to series
---
 drivers/net/bonding/bond_main.c | 25 ++++++++++---------------
 1 file changed, 10 insertions(+), 15 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index edf6dac8a98f..8be7f52e847c 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -2809,11 +2809,10 @@ static void bond_mii_monitor(struct work_struct *work)
 {
 	struct bonding *bond = container_of(work, struct bonding,
 					    mii_work.work);
-	bool should_notify_peers;
-	bool commit;
-	unsigned long delay;
-	struct slave *slave;
 	struct list_head *iter;
+	struct slave *slave;
+	unsigned long delay;
+	bool commit;
 
 	delay = msecs_to_jiffies(bond->params.miimon);
 
@@ -2822,7 +2821,6 @@ static void bond_mii_monitor(struct work_struct *work)
 
 	rcu_read_lock();
 
-	should_notify_peers = bond_should_notify_peers(bond);
 	commit = !!bond_miimon_inspect(bond);
 
 	rcu_read_unlock();
@@ -2843,10 +2841,10 @@ static void bond_mii_monitor(struct work_struct *work)
 		}
 
 		if (bond->send_peer_notif) {
-			bond->send_peer_notif--;
-			if (should_notify_peers)
+			if (bond_should_notify_peers(bond))
 				call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
 							 bond->dev);
+			bond->send_peer_notif--;
 		}
 
 		rtnl_unlock();	/* might sleep, hold no other locks */
@@ -3758,7 +3756,6 @@ static bool bond_ab_arp_probe(struct bonding *bond)
 
 static void bond_activebackup_arp_mon(struct bonding *bond)
 {
-	bool should_notify_peers = false;
 	bool should_notify_rtnl = false;
 	int delta_in_ticks;
 
@@ -3769,15 +3766,12 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 
 	rcu_read_lock();
 
-	should_notify_peers = bond_should_notify_peers(bond);
-
 	if (bond_ab_arp_inspect(bond)) {
 		rcu_read_unlock();
 
 		/* Race avoidance with bond_close flush of workqueue */
 		if (!rtnl_trylock()) {
 			delta_in_ticks = 1;
-			should_notify_peers = false;
 			goto re_arm;
 		}
 
@@ -3794,14 +3788,15 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 	if (bond->params.arp_interval)
 		queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
 
-	if (should_notify_peers || should_notify_rtnl) {
+	if (bond->send_peer_notif || should_notify_rtnl) {
 		if (!rtnl_trylock())
 			return;
 
-		if (should_notify_peers) {
+		if (bond->send_peer_notif) {
+			if (bond_should_notify_peers(bond))
+				call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
+							 bond->dev);
 			bond->send_peer_notif--;
-			call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
-						 bond->dev);
 		}
 		if (should_notify_rtnl) {
 			bond_slave_state_notify(bond);
-- 
2.34.1

[PATCH net-next v4 3/4] net: bonding: skip the 2nd trylock when first one fail

[Tonghao Zhang]

After the first trylock fail, retrying immediately is
not advised as there is a high probability of failing
to acquire the lock again. This optimization makes sense.

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>
Cc: Jason Xing <kerneljasonxing@gmail.com>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
---
v2-4:
- no change
v1:
- splitted from: https://patchwork.kernel.org/project/netdevbpf/patch/20251118090431.35654-1-tonghao@bamaicloud.com/
- this patch only skip the 2nd rtnl lock.
- add this patch to series
---
 drivers/net/bonding/bond_main.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 8be7f52e847c..b835f63d2871 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -3756,7 +3756,7 @@ static bool bond_ab_arp_probe(struct bonding *bond)
 
 static void bond_activebackup_arp_mon(struct bonding *bond)
 {
-	bool should_notify_rtnl = false;
+	bool should_notify_rtnl;
 	int delta_in_ticks;
 
 	delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
@@ -3784,13 +3784,11 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 	should_notify_rtnl = bond_ab_arp_probe(bond);
 	rcu_read_unlock();
 
-re_arm:
-	if (bond->params.arp_interval)
-		queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
-
 	if (bond->send_peer_notif || should_notify_rtnl) {
-		if (!rtnl_trylock())
-			return;
+		if (!rtnl_trylock()) {
+			delta_in_ticks = 1;
+			goto re_arm;
+		}
 
 		if (bond->send_peer_notif) {
 			if (bond_should_notify_peers(bond))
@@ -3805,6 +3803,10 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 
 		rtnl_unlock();
 	}
+
+re_arm:
+	if (bond->params.arp_interval)
+		queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
 }
 
 static void bond_arp_monitor(struct work_struct *work)
-- 
2.34.1

[PATCH net-next v4 4/4] net: bonding: add the READ_ONCE/WRITE_ONCE for outside lock accessing

[Tonghao Zhang]

Although operations on the variable send_peer_notif are already within
a lock-protected critical section, there are cases where it is accessed
outside the lock. Therefore, READ_ONCE() and WRITE_ONCE() should be
added to it.

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>
Cc: Jason Xing <kerneljasonxing@gmail.com>
Signed-off-by: Tonghao Zhang <tonghao@bamaicloud.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
---
v3/4:
- no change
v2: fix compilation errors
---
 drivers/net/bonding/bond_main.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index b835f63d2871..909c01f55744 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1204,8 +1204,9 @@ void bond_peer_notify_work_rearm(struct bonding *bond, unsigned long delay)
 /* Peer notify update handler. Holds only RTNL */
 static void bond_peer_notify_reset(struct bonding *bond)
 {
-	bond->send_peer_notif = bond->params.num_peer_notif *
-		max(1, bond->params.peer_notif_delay);
+	WRITE_ONCE(bond->send_peer_notif,
+		   bond->params.num_peer_notif *
+		   max(1, bond->params.peer_notif_delay));
 }
 
 static void bond_peer_notify_handler(struct work_struct *work)
@@ -2825,7 +2826,7 @@ static void bond_mii_monitor(struct work_struct *work)
 
 	rcu_read_unlock();
 
-	if (commit || bond->send_peer_notif) {
+	if (commit || READ_ONCE(bond->send_peer_notif)) {
 		/* Race avoidance with bond_close cancel of workqueue */
 		if (!rtnl_trylock()) {
 			delay = 1;
@@ -3784,7 +3785,7 @@ static void bond_activebackup_arp_mon(struct bonding *bond)
 	should_notify_rtnl = bond_ab_arp_probe(bond);
 	rcu_read_unlock();
 
-	if (bond->send_peer_notif || should_notify_rtnl) {
+	if (READ_ONCE(bond->send_peer_notif) || should_notify_rtnl) {
 		if (!rtnl_trylock()) {
 			delta_in_ticks = 1;
 			goto re_arm;
-- 
2.34.1

Re: [PATCH net-next v4 0/4] A series of minor optimizations of the bonding module

[Paolo Abeni]

On 12/29/25 10:50 AM, Tonghao Zhang wrote:
> These patches mainly target the peer notify mechanism of the bonding module.
> Including updates of peer notify, lock races, etc. For more information, please
> refer to the patch.
> 
> 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>
> Cc: Jason Xing <kerneljasonxing@gmail.com>

## Form letter - net-next-closed

The net-next tree is closed for new drivers, features, code refactoring
and optimizations due to the merge window and the winter break. We are
currently accepting bug fixes only.

Please repost when net-next reopens after Jan 2nd.

RFC patches sent for review only are obviously welcome at any time.