[Intel-wired-lan] [PATCH net-next v5 0/6] udp_tunnel: remove rtnl_lock dependency

[Intel-wired-lan] [PATCH net-next v5 0/6] udp_tunnel: remove rtnl_lock dependency

[Stanislav Fomichev]

Recently bnxt had to grow back a bunch of rtnl dependencies because
of udp_tunnel's infra. Add separate (global) mutext to protect
udp_tunnel state.

v5:
- remove unused variable (lkp@intel.com)

v4:
- grab lock in more places, specifically netlink and notifiers (Jakub)
- convert geneve and vxlan notifiers to (sleepable) rtnl lock

v3:
- fix 2 test failures (Jakub NIPA)

v2:
- move the lock into udp_tunnel_nic (Jakub)
- reorder the lock ordering (Jakub)
- move udp_ports_sleep removal into separate patch and update the test
(Jakub)

Cc: Michael Chan <michael.chan@broadcom.com>

Stanislav Fomichev (6):
  geneve: rely on rtnl lock in geneve_offload_rx_ports
  vxlan: drop sock_lock
  udp_tunnel: remove rtnl_lock dependency
  net: remove redundant ASSERT_RTNL() in queue setup functions
  netdevsim: remove udp_ports_sleep
  Revert "bnxt_en: bring back rtnl_lock() in the bnxt_open() path"

 .../net/ethernet/broadcom/bnx2x/bnx2x_main.c  |  3 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 42 ++-------
 drivers/net/ethernet/emulex/benet/be_main.c   |  3 +-
 drivers/net/ethernet/intel/i40e/i40e_main.c   |  1 -
 drivers/net/ethernet/intel/ice/ice_main.c     |  1 -
 .../net/ethernet/mellanox/mlx4/en_netdev.c    |  3 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  3 +-
 .../ethernet/netronome/nfp/nfp_net_common.c   |  3 +-
 .../net/ethernet/qlogic/qede/qede_filter.c    |  3 -
 .../net/ethernet/qlogic/qlcnic/qlcnic_main.c  |  1 -
 drivers/net/ethernet/sfc/ef10.c               |  1 -
 drivers/net/geneve.c                          |  7 +-
 drivers/net/netdevsim/netdevsim.h             |  2 -
 drivers/net/netdevsim/udp_tunnels.c           | 12 ---
 drivers/net/vxlan/vxlan_core.c                | 35 +++-----
 drivers/net/vxlan/vxlan_private.h             |  2 +-
 drivers/net/vxlan/vxlan_vnifilter.c           | 18 ++--
 include/net/udp_tunnel.h                      | 87 ++++++++++++++-----
 net/core/dev.c                                |  4 +-
 net/ipv4/udp_tunnel_core.c                    | 16 ++--
 net/ipv4/udp_tunnel_nic.c                     | 78 +++++++++++++----
 .../drivers/net/netdevsim/udp_tunnel_nic.sh   | 23 +----
 22 files changed, 176 insertions(+), 172 deletions(-)

-- 
2.49.0

[Intel-wired-lan] [PATCH net-next v5 1/6] geneve: rely on rtnl lock in geneve_offload_rx_ports

[Stanislav Fomichev]

udp_tunnel_push_rx_port will grab mutex in the next patch so
we can't use rcu. geneve_offload_rx_ports is called
from geneve_netdevice_event for NETDEV_UDP_TUNNEL_PUSH_INFO and
NETDEV_UDP_TUNNEL_DROP_INFO which both have ASSERT_RTNL.
Entries are added to and removed from the sock_list under rtnl
lock as well (when adding or removing a tunneling device).

Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 drivers/net/geneve.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
index ffc15a432689..9efedc6758bf 100644
--- a/drivers/net/geneve.c
+++ b/drivers/net/geneve.c
@@ -41,6 +41,7 @@ MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
 /* per-network namespace private data for this module */
 struct geneve_net {
 	struct list_head	geneve_list;
+	/* sock_list is protected by rtnl lock */
 	struct list_head	sock_list;
 };
 
@@ -1179,8 +1180,9 @@ static void geneve_offload_rx_ports(struct net_device *dev, bool push)
 	struct geneve_net *gn = net_generic(net, geneve_net_id);
 	struct geneve_sock *gs;
 
-	rcu_read_lock();
-	list_for_each_entry_rcu(gs, &gn->sock_list, list) {
+	ASSERT_RTNL();
+
+	list_for_each_entry(gs, &gn->sock_list, list) {
 		if (push) {
 			udp_tunnel_push_rx_port(dev, gs->sock,
 						UDP_TUNNEL_TYPE_GENEVE);
@@ -1189,7 +1191,6 @@ static void geneve_offload_rx_ports(struct net_device *dev, bool push)
 						UDP_TUNNEL_TYPE_GENEVE);
 		}
 	}
-	rcu_read_unlock();
 }
 
 /* Initialize the device structure. */
-- 
2.49.0

Re: [Intel-wired-lan] [PATCH net-next v5 1/6] geneve: rely on rtnl lock in geneve_offload_rx_ports

[Nikolay Aleksandrov]

On 6/16/25 19:21, Stanislav Fomichev wrote:
> udp_tunnel_push_rx_port will grab mutex in the next patch so
> we can't use rcu. geneve_offload_rx_ports is called
> from geneve_netdevice_event for NETDEV_UDP_TUNNEL_PUSH_INFO and
> NETDEV_UDP_TUNNEL_DROP_INFO which both have ASSERT_RTNL.
> Entries are added to and removed from the sock_list under rtnl
> lock as well (when adding or removing a tunneling device).
> 
> Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
> ---
>  drivers/net/geneve.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
> index ffc15a432689..9efedc6758bf 100644
> --- a/drivers/net/geneve.c
> +++ b/drivers/net/geneve.c
> @@ -41,6 +41,7 @@ MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
>  /* per-network namespace private data for this module */
>  struct geneve_net {
>  	struct list_head	geneve_list;
> +	/* sock_list is protected by rtnl lock */
>  	struct list_head	sock_list;
>  };
>  
> @@ -1179,8 +1180,9 @@ static void geneve_offload_rx_ports(struct net_device *dev, bool push)
>  	struct geneve_net *gn = net_generic(net, geneve_net_id);
>  	struct geneve_sock *gs;
>  
> -	rcu_read_lock();
> -	list_for_each_entry_rcu(gs, &gn->sock_list, list) {
> +	ASSERT_RTNL();
> +
> +	list_for_each_entry(gs, &gn->sock_list, list) {
>  		if (push) {
>  			udp_tunnel_push_rx_port(dev, gs->sock,
>  						UDP_TUNNEL_TYPE_GENEVE);
> @@ -1189,7 +1191,6 @@ static void geneve_offload_rx_ports(struct net_device *dev, bool push)
>  						UDP_TUNNEL_TYPE_GENEVE);
>  		}
>  	}
> -	rcu_read_unlock();
>  }
>  
>  /* Initialize the device structure. */

Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>

[Intel-wired-lan] [PATCH net-next v5 2/6] vxlan: drop sock_lock

[Stanislav Fomichev]

We won't be able to sleep soon in vxlan_offload_rx_ports and won't be
able to grab sock_lock. Instead of having separate spinlock to
manage sockets, rely on rtnl lock. This is similar to how geneve
manages its sockets.

Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 drivers/net/vxlan/vxlan_core.c      | 35 ++++++++++++-----------------
 drivers/net/vxlan/vxlan_private.h   |  2 +-
 drivers/net/vxlan/vxlan_vnifilter.c | 18 ++++++---------
 3 files changed, 22 insertions(+), 33 deletions(-)

diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c
index 97792de896b7..9d7249caf137 100644
--- a/drivers/net/vxlan/vxlan_core.c
+++ b/drivers/net/vxlan/vxlan_core.c
@@ -1485,21 +1485,18 @@ static enum skb_drop_reason vxlan_snoop(struct net_device *dev,
 
 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
 {
-	struct vxlan_net *vn;
+	ASSERT_RTNL();
 
 	if (!vs)
 		return false;
 	if (!refcount_dec_and_test(&vs->refcnt))
 		return false;
 
-	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
-	spin_lock(&vn->sock_lock);
 	hlist_del_rcu(&vs->hlist);
 	udp_tunnel_notify_del_rx_port(vs->sock,
 				      (vs->flags & VXLAN_F_GPE) ?
 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
 				      UDP_TUNNEL_TYPE_VXLAN);
-	spin_unlock(&vn->sock_lock);
 
 	return true;
 }
@@ -2847,26 +2844,23 @@ static void vxlan_cleanup(struct timer_list *t)
 
 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
 {
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
+	ASSERT_RTNL();
 
-	spin_lock(&vn->sock_lock);
 	hlist_del_init_rcu(&vxlan->hlist4.hlist);
 #if IS_ENABLED(CONFIG_IPV6)
 	hlist_del_init_rcu(&vxlan->hlist6.hlist);
 #endif
-	spin_unlock(&vn->sock_lock);
 }
 
 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
 			     struct vxlan_dev_node *node)
 {
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 	__be32 vni = vxlan->default_dst.remote_vni;
 
+	ASSERT_RTNL();
+
 	node->vxlan = vxlan;
-	spin_lock(&vn->sock_lock);
 	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
-	spin_unlock(&vn->sock_lock);
 }
 
 /* Setup stats when device is created */
@@ -3291,9 +3285,10 @@ static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
 	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 	unsigned int i;
 
-	spin_lock(&vn->sock_lock);
+	ASSERT_RTNL();
+
 	for (i = 0; i < PORT_HASH_SIZE; ++i) {
-		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
+		hlist_for_each_entry(vs, &vn->sock_list[i], hlist) {
 			unsigned short type;
 
 			if (vs->flags & VXLAN_F_GPE)
@@ -3307,7 +3302,6 @@ static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
 				udp_tunnel_drop_rx_port(dev, vs->sock, type);
 		}
 	}
-	spin_unlock(&vn->sock_lock);
 }
 
 /* Initialize the device structure. */
@@ -3537,12 +3531,13 @@ static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
 					      __be16 port, u32 flags,
 					      int ifindex)
 {
-	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 	struct vxlan_sock *vs;
 	struct socket *sock;
 	unsigned int h;
 	struct udp_tunnel_sock_cfg tunnel_cfg;
 
+	ASSERT_RTNL();
+
 	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
 	if (!vs)
 		return ERR_PTR(-ENOMEM);
@@ -3560,13 +3555,11 @@ static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
 	refcount_set(&vs->refcnt, 1);
 	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
 
-	spin_lock(&vn->sock_lock);
 	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
 	udp_tunnel_notify_add_rx_port(sock,
 				      (vs->flags & VXLAN_F_GPE) ?
 				      UDP_TUNNEL_TYPE_VXLAN_GPE :
 				      UDP_TUNNEL_TYPE_VXLAN);
-	spin_unlock(&vn->sock_lock);
 
 	/* Mark socket as an encapsulation socket. */
 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
@@ -3590,26 +3583,27 @@ static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
 
 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
 {
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
 	struct vxlan_sock *vs = NULL;
 	struct vxlan_dev_node *node;
 	int l3mdev_index = 0;
 
+	ASSERT_RTNL();
+
 	if (vxlan->cfg.remote_ifindex)
 		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
 			vxlan->net, vxlan->cfg.remote_ifindex);
 
 	if (!vxlan->cfg.no_share) {
-		spin_lock(&vn->sock_lock);
+		rcu_read_lock();
 		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
 				     vxlan->cfg.dst_port, vxlan->cfg.flags,
 				     l3mdev_index);
 		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
-			spin_unlock(&vn->sock_lock);
+			rcu_read_unlock();
 			return -EBUSY;
 		}
-		spin_unlock(&vn->sock_lock);
+		rcu_read_unlock();
 	}
 	if (!vs)
 		vs = vxlan_socket_create(vxlan->net, ipv6,
@@ -4875,7 +4869,6 @@ static __net_init int vxlan_init_net(struct net *net)
 	unsigned int h;
 
 	INIT_LIST_HEAD(&vn->vxlan_list);
-	spin_lock_init(&vn->sock_lock);
 	vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
 
 	for (h = 0; h < PORT_HASH_SIZE; ++h)
diff --git a/drivers/net/vxlan/vxlan_private.h b/drivers/net/vxlan/vxlan_private.h
index d328aed9feef..6c625fb29c6c 100644
--- a/drivers/net/vxlan/vxlan_private.h
+++ b/drivers/net/vxlan/vxlan_private.h
@@ -19,8 +19,8 @@ extern const struct rhashtable_params vxlan_vni_rht_params;
 /* per-network namespace private data for this module */
 struct vxlan_net {
 	struct list_head  vxlan_list;
+	/* sock_list is protected by rtnl lock */
 	struct hlist_head sock_list[PORT_HASH_SIZE];
-	spinlock_t	  sock_lock;
 	struct notifier_block nexthop_notifier_block;
 };
 
diff --git a/drivers/net/vxlan/vxlan_vnifilter.c b/drivers/net/vxlan/vxlan_vnifilter.c
index 186d0660669a..4ff56d9f8f28 100644
--- a/drivers/net/vxlan/vxlan_vnifilter.c
+++ b/drivers/net/vxlan/vxlan_vnifilter.c
@@ -40,11 +40,11 @@ static void vxlan_vs_add_del_vninode(struct vxlan_dev *vxlan,
 				     struct vxlan_vni_node *v,
 				     bool del)
 {
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 	struct vxlan_dev_node *node;
 	struct vxlan_sock *vs;
 
-	spin_lock(&vn->sock_lock);
+	ASSERT_RTNL();
+
 	if (del) {
 		if (!hlist_unhashed(&v->hlist4.hlist))
 			hlist_del_init_rcu(&v->hlist4.hlist);
@@ -52,7 +52,7 @@ static void vxlan_vs_add_del_vninode(struct vxlan_dev *vxlan,
 		if (!hlist_unhashed(&v->hlist6.hlist))
 			hlist_del_init_rcu(&v->hlist6.hlist);
 #endif
-		goto out;
+		return;
 	}
 
 #if IS_ENABLED(CONFIG_IPV6)
@@ -67,23 +67,21 @@ static void vxlan_vs_add_del_vninode(struct vxlan_dev *vxlan,
 		node = &v->hlist4;
 		hlist_add_head_rcu(&node->hlist, vni_head(vs, v->vni));
 	}
-out:
-	spin_unlock(&vn->sock_lock);
 }
 
 void vxlan_vs_add_vnigrp(struct vxlan_dev *vxlan,
 			 struct vxlan_sock *vs,
 			 bool ipv6)
 {
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 	struct vxlan_vni_group *vg = rtnl_dereference(vxlan->vnigrp);
 	struct vxlan_vni_node *v, *tmp;
 	struct vxlan_dev_node *node;
 
+	ASSERT_RTNL();
+
 	if (!vg)
 		return;
 
-	spin_lock(&vn->sock_lock);
 	list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
 #if IS_ENABLED(CONFIG_IPV6)
 		if (ipv6)
@@ -94,26 +92,24 @@ void vxlan_vs_add_vnigrp(struct vxlan_dev *vxlan,
 		node->vxlan = vxlan;
 		hlist_add_head_rcu(&node->hlist, vni_head(vs, v->vni));
 	}
-	spin_unlock(&vn->sock_lock);
 }
 
 void vxlan_vs_del_vnigrp(struct vxlan_dev *vxlan)
 {
 	struct vxlan_vni_group *vg = rtnl_dereference(vxlan->vnigrp);
-	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 	struct vxlan_vni_node *v, *tmp;
 
+	ASSERT_RTNL();
+
 	if (!vg)
 		return;
 
-	spin_lock(&vn->sock_lock);
 	list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
 		hlist_del_init_rcu(&v->hlist4.hlist);
 #if IS_ENABLED(CONFIG_IPV6)
 		hlist_del_init_rcu(&v->hlist6.hlist);
 #endif
 	}
-	spin_unlock(&vn->sock_lock);
 }
 
 static void vxlan_vnifilter_stats_get(const struct vxlan_vni_node *vninode,
-- 
2.49.0

Re: [Intel-wired-lan] [PATCH net-next v5 2/6] vxlan: drop sock_lock

[Ido Schimmel]

On Mon, Jun 16, 2025 at 09:21:13AM -0700, Stanislav Fomichev wrote:
> We won't be able to sleep soon in vxlan_offload_rx_ports and won't be
> able to grab sock_lock. Instead of having separate spinlock to
> manage sockets, rely on rtnl lock. This is similar to how geneve
> manages its sockets.
> 
> Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>

Reviewed-by: Ido Schimmel <idosch@nvidia.com>

Re: [Intel-wired-lan] [PATCH net-next v5 2/6] vxlan: drop sock_lock

[Nikolay Aleksandrov]

On 6/16/25 19:21, Stanislav Fomichev wrote:
> We won't be able to sleep soon in vxlan_offload_rx_ports and won't be
> able to grab sock_lock. Instead of having separate spinlock to
> manage sockets, rely on rtnl lock. This is similar to how geneve
> manages its sockets.
> 
> Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
> ---
>  drivers/net/vxlan/vxlan_core.c      | 35 ++++++++++++-----------------
>  drivers/net/vxlan/vxlan_private.h   |  2 +-
>  drivers/net/vxlan/vxlan_vnifilter.c | 18 ++++++---------
>  3 files changed, 22 insertions(+), 33 deletions(-)
> 

Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>

[Intel-wired-lan] [PATCH net-next v5 3/6] udp_tunnel: remove rtnl_lock dependency

[Stanislav Fomichev]

Drivers that are using ops lock and don't depend on RTNL lock
still need to manage it because udp_tunnel's RTNL dependency.
Introduce new udp_tunnel_nic_lock and use it instead of
rtnl_lock. Drop non-UDP_TUNNEL_NIC_INFO_MAY_SLEEP mode from
udp_tunnel infra (udp_tunnel_nic_device_sync_work needs to
grab udp_tunnel_nic_lock mutex and might sleep).

Cover more places in v4:

- netlink
  - udp_tunnel_notify_add_rx_port (ndo_open)
    - triggers udp_tunnel_nic_device_sync_work
  - udp_tunnel_notify_del_rx_port (ndo_stop)
    - triggers udp_tunnel_nic_device_sync_work
  - udp_tunnel_get_rx_info (__netdev_update_features)
    - triggers NETDEV_UDP_TUNNEL_PUSH_INFO
  - udp_tunnel_drop_rx_info (__netdev_update_features)
    - triggers NETDEV_UDP_TUNNEL_DROP_INFO
  - udp_tunnel_nic_reset_ntf (ndo_open)

- notifiers
  - udp_tunnel_nic_netdevice_event, depending on the event:
    - triggers NETDEV_UDP_TUNNEL_PUSH_INFO
    - triggers NETDEV_UDP_TUNNEL_DROP_INFO

- ethnl_tunnel_info_reply_size
- udp_tunnel_nic_set_port_priv (two intel drivers)

Cc: Michael Chan <michael.chan@broadcom.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 .../net/ethernet/broadcom/bnx2x/bnx2x_main.c  |  3 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     |  6 +-
 drivers/net/ethernet/emulex/benet/be_main.c   |  3 +-
 drivers/net/ethernet/intel/i40e/i40e_main.c   |  1 -
 drivers/net/ethernet/intel/ice/ice_main.c     |  1 -
 .../net/ethernet/mellanox/mlx4/en_netdev.c    |  3 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  3 +-
 .../ethernet/netronome/nfp/nfp_net_common.c   |  3 +-
 .../net/ethernet/qlogic/qede/qede_filter.c    |  3 -
 .../net/ethernet/qlogic/qlcnic/qlcnic_main.c  |  1 -
 drivers/net/ethernet/sfc/ef10.c               |  1 -
 drivers/net/netdevsim/udp_tunnels.c           |  4 -
 include/net/udp_tunnel.h                      | 87 ++++++++++++++-----
 net/core/dev.c                                |  2 +
 net/ipv4/udp_tunnel_core.c                    | 16 ++--
 net/ipv4/udp_tunnel_nic.c                     | 78 +++++++++++++----
 16 files changed, 142 insertions(+), 73 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index c9a1a1d504c0..3ee4b848ef53 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -10219,8 +10219,7 @@ static int bnx2x_udp_tunnel_sync(struct net_device *netdev, unsigned int table)
 
 static const struct udp_tunnel_nic_info bnx2x_udp_tunnels = {
 	.sync_table	= bnx2x_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
+	.flags		= UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 869580b6f70d..7946586802af 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -15573,8 +15573,7 @@ static int bnxt_udp_tunnel_unset_port(struct net_device *netdev, unsigned int ta
 static const struct udp_tunnel_nic_info bnxt_udp_tunnels = {
 	.set_port	= bnxt_udp_tunnel_set_port,
 	.unset_port	= bnxt_udp_tunnel_unset_port,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
+	.flags		= UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
@@ -15582,8 +15581,7 @@ static const struct udp_tunnel_nic_info bnxt_udp_tunnels = {
 }, bnxt_udp_tunnels_p7 = {
 	.set_port	= bnxt_udp_tunnel_set_port,
 	.unset_port	= bnxt_udp_tunnel_unset_port,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
+	.flags		= UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 3d2e21592119..f49400ba9729 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -4031,8 +4031,7 @@ static int be_vxlan_unset_port(struct net_device *netdev, unsigned int table,
 static const struct udp_tunnel_nic_info be_udp_tunnels = {
 	.set_port	= be_vxlan_set_port,
 	.unset_port	= be_vxlan_unset_port,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
+	.flags		= UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN, },
 	},
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 1156a5b3055c..3b4f59d978a5 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -15895,7 +15895,6 @@ static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	pf->udp_tunnel_nic.set_port = i40e_udp_tunnel_set_port;
 	pf->udp_tunnel_nic.unset_port = i40e_udp_tunnel_unset_port;
-	pf->udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
 	pf->udp_tunnel_nic.shared = &pf->udp_tunnel_shared;
 	pf->udp_tunnel_nic.tables[0].n_entries = I40E_MAX_PF_UDP_OFFLOAD_PORTS;
 	pf->udp_tunnel_nic.tables[0].tunnel_types = UDP_TUNNEL_TYPE_VXLAN |
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 7959a65c0903..f8ef80069e3d 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -4767,7 +4767,6 @@ int ice_init_dev(struct ice_pf *pf)
 
 	pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
 	pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
-	pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
 	pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
 	if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
 		pf->hw.udp_tunnel_nic.tables[0].n_entries =
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index 281b34af0bb4..d2071aff7b8f 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -2670,8 +2670,7 @@ static int mlx4_udp_tunnel_sync(struct net_device *dev, unsigned int table)
 
 static const struct udp_tunnel_nic_info mlx4_udp_tunnels = {
 	.sync_table	= mlx4_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
+	.flags		= UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN, },
 	},
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index ea822c69d137..ccde53f94045 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -5303,8 +5303,7 @@ void mlx5e_vxlan_set_netdev_info(struct mlx5e_priv *priv)
 
 	priv->nic_info.set_port = mlx5e_vxlan_set_port;
 	priv->nic_info.unset_port = mlx5e_vxlan_unset_port;
-	priv->nic_info.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-				UDP_TUNNEL_NIC_INFO_STATIC_IANA_VXLAN;
+	priv->nic_info.flags = UDP_TUNNEL_NIC_INFO_STATIC_IANA_VXLAN;
 	priv->nic_info.tables[0].tunnel_types = UDP_TUNNEL_TYPE_VXLAN;
 	/* Don't count the space hard-coded to the IANA port */
 	priv->nic_info.tables[0].n_entries =
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
index 932f59d70f41..132626a3f9f7 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
@@ -2394,8 +2394,7 @@ static int nfp_udp_tunnel_sync(struct net_device *netdev, unsigned int table)
 
 static const struct udp_tunnel_nic_info nfp_udp_tunnels = {
 	.sync_table     = nfp_udp_tunnel_sync,
-	.flags          = UDP_TUNNEL_NIC_INFO_MAY_SLEEP |
-			  UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
+	.flags          = UDP_TUNNEL_NIC_INFO_OPEN_ONLY,
 	.tables         = {
 		{
 			.n_entries      = NFP_NET_N_VXLAN_PORTS,
diff --git a/drivers/net/ethernet/qlogic/qede/qede_filter.c b/drivers/net/ethernet/qlogic/qede/qede_filter.c
index 985026dd816f..7e341e026489 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_filter.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_filter.c
@@ -987,20 +987,17 @@ static int qede_udp_tunnel_sync(struct net_device *dev, unsigned int table)
 
 static const struct udp_tunnel_nic_info qede_udp_tunnels_both = {
 	.sync_table	= qede_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
 	},
 }, qede_udp_tunnels_vxlan = {
 	.sync_table	= qede_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
 	},
 }, qede_udp_tunnels_geneve = {
 	.sync_table	= qede_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
 	},
diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
index eb69121df726..53cdd36c4123 100644
--- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
+++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
@@ -486,7 +486,6 @@ static int qlcnic_udp_tunnel_sync(struct net_device *dev, unsigned int table)
 
 static const struct udp_tunnel_nic_info qlcnic_udp_tunnels = {
 	.sync_table	= qlcnic_udp_tunnel_sync,
-	.flags		= UDP_TUNNEL_NIC_INFO_MAY_SLEEP,
 	.tables		= {
 		{ .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN, },
 	},
diff --git a/drivers/net/ethernet/sfc/ef10.c b/drivers/net/ethernet/sfc/ef10.c
index 47349c148c0c..fcec81f862ec 100644
--- a/drivers/net/ethernet/sfc/ef10.c
+++ b/drivers/net/ethernet/sfc/ef10.c
@@ -3985,7 +3985,6 @@ static int efx_ef10_udp_tnl_unset_port(struct net_device *dev,
 static const struct udp_tunnel_nic_info efx_ef10_udp_tunnels = {
 	.set_port	= efx_ef10_udp_tnl_set_port,
 	.unset_port	= efx_ef10_udp_tnl_unset_port,
-	.flags          = UDP_TUNNEL_NIC_INFO_MAY_SLEEP,
 	.tables         = {
 		{
 			.n_entries = 16,
diff --git a/drivers/net/netdevsim/udp_tunnels.c b/drivers/net/netdevsim/udp_tunnels.c
index 640b4983a9a0..10cbbf1c584b 100644
--- a/drivers/net/netdevsim/udp_tunnels.c
+++ b/drivers/net/netdevsim/udp_tunnels.c
@@ -112,12 +112,10 @@ nsim_udp_tunnels_info_reset_write(struct file *file, const char __user *data,
 	struct net_device *dev = file->private_data;
 	struct netdevsim *ns = netdev_priv(dev);
 
-	rtnl_lock();
 	if (dev->reg_state == NETREG_REGISTERED) {
 		memset(ns->udp_ports.ports, 0, sizeof(ns->udp_ports.__ports));
 		udp_tunnel_nic_reset_ntf(dev);
 	}
-	rtnl_unlock();
 
 	return count;
 }
@@ -181,8 +179,6 @@ int nsim_udp_tunnels_info_create(struct nsim_dev *nsim_dev,
 		info->sync_table = NULL;
 	}
 
-	if (ns->udp_ports.sleep)
-		info->flags |= UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
 	if (nsim_dev->udp_ports.open_only)
 		info->flags |= UDP_TUNNEL_NIC_INFO_OPEN_ONLY;
 	if (nsim_dev->udp_ports.ipv4_only)
diff --git a/include/net/udp_tunnel.h b/include/net/udp_tunnel.h
index 2df3b8344eb5..0106e2dc2af1 100644
--- a/include/net/udp_tunnel.h
+++ b/include/net/udp_tunnel.h
@@ -130,22 +130,6 @@ void udp_tunnel_drop_rx_port(struct net_device *dev, struct socket *sock,
 void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type);
 void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type);
 
-static inline void udp_tunnel_get_rx_info(struct net_device *dev)
-{
-	ASSERT_RTNL();
-	if (!(dev->features & NETIF_F_RX_UDP_TUNNEL_PORT))
-		return;
-	call_netdevice_notifiers(NETDEV_UDP_TUNNEL_PUSH_INFO, dev);
-}
-
-static inline void udp_tunnel_drop_rx_info(struct net_device *dev)
-{
-	ASSERT_RTNL();
-	if (!(dev->features & NETIF_F_RX_UDP_TUNNEL_PORT))
-		return;
-	call_netdevice_notifiers(NETDEV_UDP_TUNNEL_DROP_INFO, dev);
-}
-
 /* Transmit the skb using UDP encapsulation. */
 void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
 			 __be32 src, __be32 dst, __u8 tos, __u8 ttl,
@@ -221,19 +205,17 @@ static inline void udp_tunnel_encap_enable(struct sock *sk)
 #define UDP_TUNNEL_NIC_MAX_TABLES	4
 
 enum udp_tunnel_nic_info_flags {
-	/* Device callbacks may sleep */
-	UDP_TUNNEL_NIC_INFO_MAY_SLEEP	= BIT(0),
 	/* Device only supports offloads when it's open, all ports
 	 * will be removed before close and re-added after open.
 	 */
-	UDP_TUNNEL_NIC_INFO_OPEN_ONLY	= BIT(1),
+	UDP_TUNNEL_NIC_INFO_OPEN_ONLY	= BIT(0),
 	/* Device supports only IPv4 tunnels */
-	UDP_TUNNEL_NIC_INFO_IPV4_ONLY	= BIT(2),
+	UDP_TUNNEL_NIC_INFO_IPV4_ONLY	= BIT(1),
 	/* Device has hard-coded the IANA VXLAN port (4789) as VXLAN.
 	 * This port must not be counted towards n_entries of any table.
 	 * Driver will not receive any callback associated with port 4789.
 	 */
-	UDP_TUNNEL_NIC_INFO_STATIC_IANA_VXLAN	= BIT(3),
+	UDP_TUNNEL_NIC_INFO_STATIC_IANA_VXLAN	= BIT(2),
 };
 
 struct udp_tunnel_nic;
@@ -324,6 +306,9 @@ struct udp_tunnel_nic_ops {
 	size_t (*dump_size)(struct net_device *dev, unsigned int table);
 	int (*dump_write)(struct net_device *dev, unsigned int table,
 			  struct sk_buff *skb);
+	void (*assert_locked)(struct net_device *dev);
+	void (*lock)(struct net_device *dev);
+	void (*unlock)(struct net_device *dev);
 };
 
 #ifdef CONFIG_INET
@@ -352,8 +337,29 @@ static inline void
 udp_tunnel_nic_set_port_priv(struct net_device *dev, unsigned int table,
 			     unsigned int idx, u8 priv)
 {
-	if (udp_tunnel_nic_ops)
+	if (udp_tunnel_nic_ops) {
+		udp_tunnel_nic_ops->lock(dev);
 		udp_tunnel_nic_ops->set_port_priv(dev, table, idx, priv);
+		udp_tunnel_nic_ops->unlock(dev);
+	}
+}
+
+static inline void udp_tunnel_nic_assert_locked(struct net_device *dev)
+{
+	if (udp_tunnel_nic_ops)
+		udp_tunnel_nic_ops->assert_locked(dev);
+}
+
+static inline void udp_tunnel_nic_lock(struct net_device *dev)
+{
+	if (udp_tunnel_nic_ops)
+		udp_tunnel_nic_ops->lock(dev);
+}
+
+static inline void udp_tunnel_nic_unlock(struct net_device *dev)
+{
+	if (udp_tunnel_nic_ops)
+		udp_tunnel_nic_ops->unlock(dev);
 }
 
 static inline void
@@ -395,17 +401,50 @@ static inline void udp_tunnel_nic_reset_ntf(struct net_device *dev)
 static inline size_t
 udp_tunnel_nic_dump_size(struct net_device *dev, unsigned int table)
 {
+	size_t ret;
+
 	if (!udp_tunnel_nic_ops)
 		return 0;
-	return udp_tunnel_nic_ops->dump_size(dev, table);
+
+	udp_tunnel_nic_ops->lock(dev);
+	ret = udp_tunnel_nic_ops->dump_size(dev, table);
+	udp_tunnel_nic_ops->unlock(dev);
+
+	return ret;
 }
 
 static inline int
 udp_tunnel_nic_dump_write(struct net_device *dev, unsigned int table,
 			  struct sk_buff *skb)
 {
+	int ret;
+
 	if (!udp_tunnel_nic_ops)
 		return 0;
-	return udp_tunnel_nic_ops->dump_write(dev, table, skb);
+
+	udp_tunnel_nic_ops->lock(dev);
+	ret = udp_tunnel_nic_ops->dump_write(dev, table, skb);
+	udp_tunnel_nic_ops->unlock(dev);
+
+	return ret;
+}
+
+static inline void udp_tunnel_get_rx_info(struct net_device *dev)
+{
+	ASSERT_RTNL();
+	if (!(dev->features & NETIF_F_RX_UDP_TUNNEL_PORT))
+		return;
+	udp_tunnel_nic_assert_locked(dev);
+	call_netdevice_notifiers(NETDEV_UDP_TUNNEL_PUSH_INFO, dev);
 }
+
+static inline void udp_tunnel_drop_rx_info(struct net_device *dev)
+{
+	ASSERT_RTNL();
+	if (!(dev->features & NETIF_F_RX_UDP_TUNNEL_PORT))
+		return;
+	udp_tunnel_nic_assert_locked(dev);
+	call_netdevice_notifiers(NETDEV_UDP_TUNNEL_DROP_INFO, dev);
+}
+
 #endif
diff --git a/net/core/dev.c b/net/core/dev.c
index 5baa4691074f..43f56b44f351 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10771,12 +10771,14 @@ int __netdev_update_features(struct net_device *dev)
 			 * *before* calling udp_tunnel_get_rx_info,
 			 * but *after* calling udp_tunnel_drop_rx_info.
 			 */
+			udp_tunnel_nic_lock(dev);
 			if (features & NETIF_F_RX_UDP_TUNNEL_PORT) {
 				dev->features = features;
 				udp_tunnel_get_rx_info(dev);
 			} else {
 				udp_tunnel_drop_rx_info(dev);
 			}
+			udp_tunnel_nic_unlock(dev);
 		}
 
 		if (diff & NETIF_F_HW_VLAN_CTAG_FILTER) {
diff --git a/net/ipv4/udp_tunnel_core.c b/net/ipv4/udp_tunnel_core.c
index 2326548997d3..433283dd0e70 100644
--- a/net/ipv4/udp_tunnel_core.c
+++ b/net/ipv4/udp_tunnel_core.c
@@ -134,15 +134,17 @@ void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type)
 	struct udp_tunnel_info ti;
 	struct net_device *dev;
 
+	ASSERT_RTNL();
+
 	ti.type = type;
 	ti.sa_family = sk->sk_family;
 	ti.port = inet_sk(sk)->inet_sport;
 
-	rcu_read_lock();
-	for_each_netdev_rcu(net, dev) {
+	for_each_netdev(net, dev) {
+		udp_tunnel_nic_lock(dev);
 		udp_tunnel_nic_add_port(dev, &ti);
+		udp_tunnel_nic_unlock(dev);
 	}
-	rcu_read_unlock();
 }
 EXPORT_SYMBOL_GPL(udp_tunnel_notify_add_rx_port);
 
@@ -154,15 +156,17 @@ void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type)
 	struct udp_tunnel_info ti;
 	struct net_device *dev;
 
+	ASSERT_RTNL();
+
 	ti.type = type;
 	ti.sa_family = sk->sk_family;
 	ti.port = inet_sk(sk)->inet_sport;
 
-	rcu_read_lock();
-	for_each_netdev_rcu(net, dev) {
+	for_each_netdev(net, dev) {
+		udp_tunnel_nic_lock(dev);
 		udp_tunnel_nic_del_port(dev, &ti);
+		udp_tunnel_nic_unlock(dev);
 	}
-	rcu_read_unlock();
 }
 EXPORT_SYMBOL_GPL(udp_tunnel_notify_del_rx_port);
 
diff --git a/net/ipv4/udp_tunnel_nic.c b/net/ipv4/udp_tunnel_nic.c
index b6d2d16189c0..ff66db48453c 100644
--- a/net/ipv4/udp_tunnel_nic.c
+++ b/net/ipv4/udp_tunnel_nic.c
@@ -29,6 +29,7 @@ struct udp_tunnel_nic_table_entry {
  * struct udp_tunnel_nic - UDP tunnel port offload state
  * @work:	async work for talking to hardware from process context
  * @dev:	netdev pointer
+ * @lock:	protects all fields
  * @need_sync:	at least one port start changed
  * @need_replay: space was freed, we need a replay of all ports
  * @work_pending: @work is currently scheduled
@@ -41,6 +42,8 @@ struct udp_tunnel_nic {
 
 	struct net_device *dev;
 
+	struct mutex lock;
+
 	u8 need_sync:1;
 	u8 need_replay:1;
 	u8 work_pending:1;
@@ -298,22 +301,11 @@ __udp_tunnel_nic_device_sync(struct net_device *dev, struct udp_tunnel_nic *utn)
 static void
 udp_tunnel_nic_device_sync(struct net_device *dev, struct udp_tunnel_nic *utn)
 {
-	const struct udp_tunnel_nic_info *info = dev->udp_tunnel_nic_info;
-	bool may_sleep;
-
 	if (!utn->need_sync)
 		return;
 
-	/* Drivers which sleep in the callback need to update from
-	 * the workqueue, if we come from the tunnel driver's notification.
-	 */
-	may_sleep = info->flags & UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
-	if (!may_sleep)
-		__udp_tunnel_nic_device_sync(dev, utn);
-	if (may_sleep || utn->need_replay) {
-		queue_work(udp_tunnel_nic_workqueue, &utn->work);
-		utn->work_pending = 1;
-	}
+	queue_work(udp_tunnel_nic_workqueue, &utn->work);
+	utn->work_pending = 1;
 }
 
 static bool
@@ -554,12 +546,12 @@ static void __udp_tunnel_nic_reset_ntf(struct net_device *dev)
 	struct udp_tunnel_nic *utn;
 	unsigned int i, j;
 
-	ASSERT_RTNL();
-
 	utn = dev->udp_tunnel_nic;
 	if (!utn)
 		return;
 
+	mutex_lock(&utn->lock);
+
 	utn->need_sync = false;
 	for (i = 0; i < utn->n_tables; i++)
 		for (j = 0; j < info->tables[i].n_entries; j++) {
@@ -569,7 +561,7 @@ static void __udp_tunnel_nic_reset_ntf(struct net_device *dev)
 
 			entry->flags &= ~(UDP_TUNNEL_NIC_ENTRY_DEL |
 					  UDP_TUNNEL_NIC_ENTRY_OP_FAIL);
-			/* We don't release rtnl across ops */
+			/* We don't release utn lock across ops */
 			WARN_ON(entry->flags & UDP_TUNNEL_NIC_ENTRY_FROZEN);
 			if (!entry->use_cnt)
 				continue;
@@ -579,6 +571,8 @@ static void __udp_tunnel_nic_reset_ntf(struct net_device *dev)
 		}
 
 	__udp_tunnel_nic_device_sync(dev, utn);
+
+	mutex_unlock(&utn->lock);
 }
 
 static size_t
@@ -643,6 +637,33 @@ __udp_tunnel_nic_dump_write(struct net_device *dev, unsigned int table,
 	return -EMSGSIZE;
 }
 
+static void __udp_tunnel_nic_assert_locked(struct net_device *dev)
+{
+	struct udp_tunnel_nic *utn;
+
+	utn = dev->udp_tunnel_nic;
+	if (utn)
+		lockdep_assert_held(&utn->lock);
+}
+
+static void __udp_tunnel_nic_lock(struct net_device *dev)
+{
+	struct udp_tunnel_nic *utn;
+
+	utn = dev->udp_tunnel_nic;
+	if (utn)
+		mutex_lock(&utn->lock);
+}
+
+static void __udp_tunnel_nic_unlock(struct net_device *dev)
+{
+	struct udp_tunnel_nic *utn;
+
+	utn = dev->udp_tunnel_nic;
+	if (utn)
+		mutex_unlock(&utn->lock);
+}
+
 static const struct udp_tunnel_nic_ops __udp_tunnel_nic_ops = {
 	.get_port	= __udp_tunnel_nic_get_port,
 	.set_port_priv	= __udp_tunnel_nic_set_port_priv,
@@ -651,6 +672,9 @@ static const struct udp_tunnel_nic_ops __udp_tunnel_nic_ops = {
 	.reset_ntf	= __udp_tunnel_nic_reset_ntf,
 	.dump_size	= __udp_tunnel_nic_dump_size,
 	.dump_write	= __udp_tunnel_nic_dump_write,
+	.assert_locked	= __udp_tunnel_nic_assert_locked,
+	.lock		= __udp_tunnel_nic_lock,
+	.unlock		= __udp_tunnel_nic_unlock,
 };
 
 static void
@@ -710,11 +734,15 @@ static void udp_tunnel_nic_device_sync_work(struct work_struct *work)
 		container_of(work, struct udp_tunnel_nic, work);
 
 	rtnl_lock();
+	mutex_lock(&utn->lock);
+
 	utn->work_pending = 0;
 	__udp_tunnel_nic_device_sync(utn->dev, utn);
 
 	if (utn->need_replay)
 		udp_tunnel_nic_replay(utn->dev, utn);
+
+	mutex_unlock(&utn->lock);
 	rtnl_unlock();
 }
 
@@ -730,6 +758,7 @@ udp_tunnel_nic_alloc(const struct udp_tunnel_nic_info *info,
 		return NULL;
 	utn->n_tables = n_tables;
 	INIT_WORK(&utn->work, udp_tunnel_nic_device_sync_work);
+	mutex_init(&utn->lock);
 
 	for (i = 0; i < n_tables; i++) {
 		utn->entries[i] = kcalloc(info->tables[i].n_entries,
@@ -821,8 +850,11 @@ static int udp_tunnel_nic_register(struct net_device *dev)
 	dev_hold(dev);
 	dev->udp_tunnel_nic = utn;
 
-	if (!(info->flags & UDP_TUNNEL_NIC_INFO_OPEN_ONLY))
+	if (!(info->flags & UDP_TUNNEL_NIC_INFO_OPEN_ONLY)) {
+		udp_tunnel_nic_lock(dev);
 		udp_tunnel_get_rx_info(dev);
+		udp_tunnel_nic_unlock(dev);
+	}
 
 	return 0;
 }
@@ -832,6 +864,8 @@ udp_tunnel_nic_unregister(struct net_device *dev, struct udp_tunnel_nic *utn)
 {
 	const struct udp_tunnel_nic_info *info = dev->udp_tunnel_nic_info;
 
+	udp_tunnel_nic_lock(dev);
+
 	/* For a shared table remove this dev from the list of sharing devices
 	 * and if there are other devices just detach.
 	 */
@@ -841,8 +875,10 @@ udp_tunnel_nic_unregister(struct net_device *dev, struct udp_tunnel_nic *utn)
 		list_for_each_entry(node, &info->shared->devices, list)
 			if (node->dev == dev)
 				break;
-		if (list_entry_is_head(node, &info->shared->devices, list))
+		if (list_entry_is_head(node, &info->shared->devices, list)) {
+			udp_tunnel_nic_unlock(dev);
 			return;
+		}
 
 		list_del(&node->list);
 		kfree(node);
@@ -852,6 +888,7 @@ udp_tunnel_nic_unregister(struct net_device *dev, struct udp_tunnel_nic *utn)
 		if (first) {
 			udp_tunnel_drop_rx_info(dev);
 			utn->dev = first->dev;
+			udp_tunnel_nic_unlock(dev);
 			goto release_dev;
 		}
 
@@ -862,6 +899,7 @@ udp_tunnel_nic_unregister(struct net_device *dev, struct udp_tunnel_nic *utn)
 	 * from the work which we will boot immediately.
 	 */
 	udp_tunnel_nic_flush(dev, utn);
+	udp_tunnel_nic_unlock(dev);
 
 	/* Wait for the work to be done using the state, netdev core will
 	 * retry unregister until we give up our reference on this device.
@@ -910,12 +948,16 @@ udp_tunnel_nic_netdevice_event(struct notifier_block *unused,
 		return NOTIFY_DONE;
 
 	if (event == NETDEV_UP) {
+		udp_tunnel_nic_lock(dev);
 		WARN_ON(!udp_tunnel_nic_is_empty(dev, utn));
 		udp_tunnel_get_rx_info(dev);
+		udp_tunnel_nic_unlock(dev);
 		return NOTIFY_OK;
 	}
 	if (event == NETDEV_GOING_DOWN) {
+		udp_tunnel_nic_lock(dev);
 		udp_tunnel_nic_flush(dev, utn);
+		udp_tunnel_nic_unlock(dev);
 		return NOTIFY_OK;
 	}
 
-- 
2.49.0

[Intel-wired-lan] [PATCH net-next v5 4/6] net: remove redundant ASSERT_RTNL() in queue setup functions

[Stanislav Fomichev]

The existing netdev_ops_assert_locked() already asserts that either
the RTNL lock or the per-device lock is held, making the explicit
ASSERT_RTNL() redundant.

Cc: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 net/core/dev.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 43f56b44f351..df1678b1fe24 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3179,7 +3179,6 @@ int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
 
 	if (dev->reg_state == NETREG_REGISTERED ||
 	    dev->reg_state == NETREG_UNREGISTERING) {
-		ASSERT_RTNL();
 		netdev_ops_assert_locked(dev);
 
 		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
@@ -3229,7 +3228,6 @@ int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
 		return -EINVAL;
 
 	if (dev->reg_state == NETREG_REGISTERED) {
-		ASSERT_RTNL();
 		netdev_ops_assert_locked(dev);
 
 		rc = net_rx_queue_update_kobjects(dev, dev->real_num_rx_queues,
-- 
2.49.0

[Intel-wired-lan] [PATCH net-next v5 5/6] netdevsim: remove udp_ports_sleep

[Stanislav Fomichev]

Now that there is only one path in udp_tunnel, there is no need to
have udp_ports_sleep knob. Remove it and adjust the test.

Cc: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 drivers/net/netdevsim/netdevsim.h             |  2 --
 drivers/net/netdevsim/udp_tunnels.c           |  8 -------
 .../drivers/net/netdevsim/udp_tunnel_nic.sh   | 23 +------------------
 3 files changed, 1 insertion(+), 32 deletions(-)

diff --git a/drivers/net/netdevsim/netdevsim.h b/drivers/net/netdevsim/netdevsim.h
index d04401f0bdf7..511ed72a93ce 100644
--- a/drivers/net/netdevsim/netdevsim.h
+++ b/drivers/net/netdevsim/netdevsim.h
@@ -131,7 +131,6 @@ struct netdevsim {
 	struct nsim_macsec macsec;
 	struct {
 		u32 inject_error;
-		u32 sleep;
 		u32 __ports[2][NSIM_UDP_TUNNEL_N_PORTS];
 		u32 (*ports)[NSIM_UDP_TUNNEL_N_PORTS];
 		struct dentry *ddir;
@@ -342,7 +341,6 @@ struct nsim_dev {
 		bool ipv4_only;
 		bool shared;
 		bool static_iana_vxlan;
-		u32 sleep;
 	} udp_ports;
 	struct nsim_dev_psample *psample;
 	u16 esw_mode;
diff --git a/drivers/net/netdevsim/udp_tunnels.c b/drivers/net/netdevsim/udp_tunnels.c
index 10cbbf1c584b..89fff76e51cf 100644
--- a/drivers/net/netdevsim/udp_tunnels.c
+++ b/drivers/net/netdevsim/udp_tunnels.c
@@ -18,9 +18,6 @@ nsim_udp_tunnel_set_port(struct net_device *dev, unsigned int table,
 	ret = -ns->udp_ports.inject_error;
 	ns->udp_ports.inject_error = 0;
 
-	if (ns->udp_ports.sleep)
-		msleep(ns->udp_ports.sleep);
-
 	if (!ret) {
 		if (ns->udp_ports.ports[table][entry]) {
 			WARN(1, "entry already in use\n");
@@ -47,8 +44,6 @@ nsim_udp_tunnel_unset_port(struct net_device *dev, unsigned int table,
 	ret = -ns->udp_ports.inject_error;
 	ns->udp_ports.inject_error = 0;
 
-	if (ns->udp_ports.sleep)
-		msleep(ns->udp_ports.sleep);
 	if (!ret) {
 		u32 val = be16_to_cpu(ti->port) << 16 | ti->type;
 
@@ -170,7 +165,6 @@ int nsim_udp_tunnels_info_create(struct nsim_dev *nsim_dev,
 		       GFP_KERNEL);
 	if (!info)
 		return -ENOMEM;
-	ns->udp_ports.sleep = nsim_dev->udp_ports.sleep;
 
 	if (nsim_dev->udp_ports.sync_all) {
 		info->set_port = NULL;
@@ -213,6 +207,4 @@ void nsim_udp_tunnels_debugfs_create(struct nsim_dev *nsim_dev)
 			    &nsim_dev->udp_ports.shared);
 	debugfs_create_bool("udp_ports_static_iana_vxlan", 0600, nsim_dev->ddir,
 			    &nsim_dev->udp_ports.static_iana_vxlan);
-	debugfs_create_u32("udp_ports_sleep", 0600, nsim_dev->ddir,
-			   &nsim_dev->udp_ports.sleep);
 }
diff --git a/tools/testing/selftests/drivers/net/netdevsim/udp_tunnel_nic.sh b/tools/testing/selftests/drivers/net/netdevsim/udp_tunnel_nic.sh
index 92c2f0376c08..4c859ecdad94 100755
--- a/tools/testing/selftests/drivers/net/netdevsim/udp_tunnel_nic.sh
+++ b/tools/testing/selftests/drivers/net/netdevsim/udp_tunnel_nic.sh
@@ -266,7 +266,6 @@ for port in 0 1; do
 	echo $NSIM_ID > /sys/bus/netdevsim/new_device
     else
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
 	echo 1 > $NSIM_DEV_SYS/new_port
     fi
     NSIM_NETDEV=`get_netdev_name old_netdevs`
@@ -350,23 +349,11 @@ old_netdevs=$(ls /sys/class/net)
 port=0
 echo $NSIM_ID > /sys/bus/netdevsim/new_device
 echo 0 > $NSIM_DEV_SYS/del_port
-echo 1000 > $NSIM_DEV_DFS/udp_ports_sleep
 echo 0 > $NSIM_DEV_SYS/new_port
 NSIM_NETDEV=`get_netdev_name old_netdevs`
 
 msg="create VxLANs"
-exp0=( 0 0 0 0 ) # sleep is longer than out wait
-new_vxlan vxlan0 10000 $NSIM_NETDEV
-
-modprobe -r vxlan
-modprobe -r udp_tunnel
-
-msg="remove tunnels"
-exp0=( 0 0 0 0 )
-check_tables
-
-msg="create VxLANs"
-exp0=( 0 0 0 0 ) # sleep is longer than out wait
+exp0=( `mke 10000 1` 0 0 0 )
 new_vxlan vxlan0 10000 $NSIM_NETDEV
 
 exp0=( 0 0 0 0 )
@@ -428,7 +415,6 @@ echo 0 > $NSIM_DEV_SYS/del_port
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -486,7 +472,6 @@ echo 1 > $NSIM_DEV_DFS/udp_ports_sync_all
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -543,7 +528,6 @@ echo 0 > $NSIM_DEV_SYS/del_port
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -573,7 +557,6 @@ echo 1 > $NSIM_DEV_DFS/udp_ports_ipv4_only
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -634,7 +617,6 @@ echo 0 > $NSIM_DEV_SYS/del_port
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -690,7 +672,6 @@ echo 0 > $NSIM_DEV_SYS/del_port
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -750,7 +731,6 @@ echo 0 > $NSIM_DEV_SYS/del_port
 for port in 0 1; do
     if [ $port -ne 0 ]; then
 	echo 1 > $NSIM_DEV_DFS/udp_ports_open_only
-	echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
     fi
 
     echo $port > $NSIM_DEV_SYS/new_port
@@ -809,7 +789,6 @@ echo $NSIM_ID > /sys/bus/netdevsim/new_device
 echo 0 > $NSIM_DEV_SYS/del_port
 
 echo 0 > $NSIM_DEV_DFS/udp_ports_open_only
-echo 1 > $NSIM_DEV_DFS/udp_ports_sleep
 echo 1 > $NSIM_DEV_DFS/udp_ports_shared
 
 old_netdevs=$(ls /sys/class/net)
-- 
2.49.0

[Intel-wired-lan] [PATCH net-next v5 6/6] Revert "bnxt_en: bring back rtnl_lock() in the bnxt_open() path"

[Stanislav Fomichev]

This reverts commit 325eb217e41fa14f307c7cc702bd18d0bb38fe84.

udp_tunnel infra doesn't need RTNL, should be safe to get back
to only netdev instance lock.

Cc: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Stanislav Fomichev <stfomichev@gmail.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 36 +++++------------------
 1 file changed, 7 insertions(+), 29 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 7946586802af..b359ef4b78a9 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -14055,28 +14055,13 @@ static void bnxt_unlock_sp(struct bnxt *bp)
 	netdev_unlock(bp->dev);
 }
 
-/* Same as bnxt_lock_sp() with additional rtnl_lock */
-static void bnxt_rtnl_lock_sp(struct bnxt *bp)
-{
-	clear_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
-	rtnl_lock();
-	netdev_lock(bp->dev);
-}
-
-static void bnxt_rtnl_unlock_sp(struct bnxt *bp)
-{
-	set_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
-	netdev_unlock(bp->dev);
-	rtnl_unlock();
-}
-
 /* Only called from bnxt_sp_task() */
 static void bnxt_reset(struct bnxt *bp, bool silent)
 {
-	bnxt_rtnl_lock_sp(bp);
+	bnxt_lock_sp(bp);
 	if (test_bit(BNXT_STATE_OPEN, &bp->state))
 		bnxt_reset_task(bp, silent);
-	bnxt_rtnl_unlock_sp(bp);
+	bnxt_unlock_sp(bp);
 }
 
 /* Only called from bnxt_sp_task() */
@@ -14084,9 +14069,9 @@ static void bnxt_rx_ring_reset(struct bnxt *bp)
 {
 	int i;
 
-	bnxt_rtnl_lock_sp(bp);
+	bnxt_lock_sp(bp);
 	if (!test_bit(BNXT_STATE_OPEN, &bp->state)) {
-		bnxt_rtnl_unlock_sp(bp);
+		bnxt_unlock_sp(bp);
 		return;
 	}
 	/* Disable and flush TPA before resetting the RX ring */
@@ -14125,7 +14110,7 @@ static void bnxt_rx_ring_reset(struct bnxt *bp)
 	}
 	if (bp->flags & BNXT_FLAG_TPA)
 		bnxt_set_tpa(bp, true);
-	bnxt_rtnl_unlock_sp(bp);
+	bnxt_unlock_sp(bp);
 }
 
 static void bnxt_fw_fatal_close(struct bnxt *bp)
@@ -15017,17 +15002,15 @@ static void bnxt_fw_reset_task(struct work_struct *work)
 		bp->fw_reset_state = BNXT_FW_RESET_STATE_OPENING;
 		fallthrough;
 	case BNXT_FW_RESET_STATE_OPENING:
-		while (!rtnl_trylock()) {
+		while (!netdev_trylock(bp->dev)) {
 			bnxt_queue_fw_reset_work(bp, HZ / 10);
 			return;
 		}
-		netdev_lock(bp->dev);
 		rc = bnxt_open(bp->dev);
 		if (rc) {
 			netdev_err(bp->dev, "bnxt_open() failed during FW reset\n");
 			bnxt_fw_reset_abort(bp, rc);
 			netdev_unlock(bp->dev);
-			rtnl_unlock();
 			goto ulp_start;
 		}
 
@@ -15047,7 +15030,6 @@ static void bnxt_fw_reset_task(struct work_struct *work)
 			bnxt_dl_health_fw_status_update(bp, true);
 		}
 		netdev_unlock(bp->dev);
-		rtnl_unlock();
 		bnxt_ulp_start(bp, 0);
 		bnxt_reenable_sriov(bp);
 		netdev_lock(bp->dev);
@@ -15996,7 +15978,7 @@ static int bnxt_queue_start(struct net_device *dev, void *qmem, int idx)
 		   rc);
 	napi_enable_locked(&bnapi->napi);
 	bnxt_db_nq_arm(bp, &cpr->cp_db, cpr->cp_raw_cons);
-	netif_close(dev);
+	bnxt_reset_task(bp, true);
 	return rc;
 }
 
@@ -16812,7 +16794,6 @@ static int bnxt_resume(struct device *device)
 	struct bnxt *bp = netdev_priv(dev);
 	int rc = 0;
 
-	rtnl_lock();
 	netdev_lock(dev);
 	rc = pci_enable_device(bp->pdev);
 	if (rc) {
@@ -16857,7 +16838,6 @@ static int bnxt_resume(struct device *device)
 
 resume_exit:
 	netdev_unlock(bp->dev);
-	rtnl_unlock();
 	bnxt_ulp_start(bp, rc);
 	if (!rc)
 		bnxt_reenable_sriov(bp);
@@ -17023,7 +17003,6 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 	int err;
 
 	netdev_info(bp->dev, "PCI Slot Resume\n");
-	rtnl_lock();
 	netdev_lock(netdev);
 
 	err = bnxt_hwrm_func_qcaps(bp);
@@ -17041,7 +17020,6 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 		netif_device_attach(netdev);
 
 	netdev_unlock(netdev);
-	rtnl_unlock();
 	bnxt_ulp_start(bp, err);
 	if (!err)
 		bnxt_reenable_sriov(bp);
-- 
2.49.0

Re: [Intel-wired-lan] [PATCH net-next v5 0/6] udp_tunnel: remove rtnl_lock dependency

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Mon, 16 Jun 2025 09:21:11 -0700 you wrote:
> Recently bnxt had to grow back a bunch of rtnl dependencies because
> of udp_tunnel's infra. Add separate (global) mutext to protect
> udp_tunnel state.
> 
> v5:
> - remove unused variable (lkp@intel.com)
> 
> [...]

Here is the summary with links:
  - [net-next,v5,1/6] geneve: rely on rtnl lock in geneve_offload_rx_ports
    https://git.kernel.org/netdev/net-next/c/3e14960f3bd2
  - [net-next,v5,2/6] vxlan: drop sock_lock
    https://git.kernel.org/netdev/net-next/c/df5425b3bd85
  - [net-next,v5,3/6] udp_tunnel: remove rtnl_lock dependency
    https://git.kernel.org/netdev/net-next/c/1ead7501094c
  - [net-next,v5,4/6] net: remove redundant ASSERT_RTNL() in queue setup functions
    https://git.kernel.org/netdev/net-next/c/3a321b6b1f76
  - [net-next,v5,5/6] netdevsim: remove udp_ports_sleep
    https://git.kernel.org/netdev/net-next/c/e054c8ba3bce
  - [net-next,v5,6/6] Revert "bnxt_en: bring back rtnl_lock() in the bnxt_open() path"
    https://git.kernel.org/netdev/net-next/c/850d9248d2ea

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html