Merge tag 'batadv-next-pullrequest-20250313' of git://git.open-mesh.org/linux-merge
authorPaolo Abeni <pabeni@redhat.com>
Tue, 18 Mar 2025 11:10:20 +0000 (12:10 +0100)
committerPaolo Abeni <pabeni@redhat.com>
Tue, 18 Mar 2025 11:10:20 +0000 (12:10 +0100)
Simon Wunderlich says:

====================
This feature/cleanup patchset includes the following patches:

 - bump version strings, by Simon Wunderlich

 - drop batadv_priv_debug_log struct, by Sven Eckelmann

 - adopt netdev_hold() / netdev_put(), by Eric Dumazet

 - add support for jumbo frames, by Sven Eckelmann

 - use consistent name for mesh interface, by Sven Eckelmann

 - cleanup B.A.T.M.A.N. IV OGM aggregation handling,
   by Sven Eckelmann (4 patches)

 - add missing newlines for log macros, by Sven Eckelmann

* tag 'batadv-next-pullrequest-20250313' of git://git.open-mesh.org/linux-merge:
  batman-adv: add missing newlines for log macros
  batman-adv: Limit aggregation size to outgoing MTU
  batman-adv: Use actual packet count for aggregated packets
  batman-adv: Switch to bitmap helper for aggregation handling
  batman-adv: Limit number of aggregated packets directly
  batman-adv: Use consistent name for mesh interface
  batman-adv: Add support for jumbo frames
  batman-adv: adopt netdev_hold() / netdev_put()
  batman-adv: Drop batadv_priv_debug_log struct
  batman-adv: Start new development cycle
====================

Link: https://patch.msgid.link/20250313164519.72808-1-sw@simonwunderlich.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
1  2 
net/batman-adv/mesh-interface.c

index 0000000000000000000000000000000000000000,c0a2a2a8fbb533d8ad08f1ef3a9bb3a8dee75f39..59e7b5aacbc93554be32062eb1096f3424641188
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1161 +1,1161 @@@
 -      dev->netns_local = true;
+ // SPDX-License-Identifier: GPL-2.0
+ /* Copyright (C) B.A.T.M.A.N. contributors:
+  *
+  * Marek Lindner, Simon Wunderlich
+  */
+ #include "mesh-interface.h"
+ #include "main.h"
+ #include <linux/atomic.h>
+ #include <linux/byteorder/generic.h>
+ #include <linux/cache.h>
+ #include <linux/compiler.h>
+ #include <linux/container_of.h>
+ #include <linux/cpumask.h>
+ #include <linux/errno.h>
+ #include <linux/etherdevice.h>
+ #include <linux/ethtool.h>
+ #include <linux/gfp.h>
+ #include <linux/if_ether.h>
+ #include <linux/if_vlan.h>
+ #include <linux/jiffies.h>
+ #include <linux/kref.h>
+ #include <linux/list.h>
+ #include <linux/lockdep.h>
+ #include <linux/netdevice.h>
+ #include <linux/netlink.h>
+ #include <linux/percpu.h>
+ #include <linux/random.h>
+ #include <linux/rculist.h>
+ #include <linux/rcupdate.h>
+ #include <linux/skbuff.h>
+ #include <linux/slab.h>
+ #include <linux/socket.h>
+ #include <linux/spinlock.h>
+ #include <linux/stddef.h>
+ #include <linux/string.h>
+ #include <linux/types.h>
+ #include <net/net_namespace.h>
+ #include <net/netlink.h>
+ #include <uapi/linux/batadv_packet.h>
+ #include <uapi/linux/batman_adv.h>
+ #include "bat_algo.h"
+ #include "bridge_loop_avoidance.h"
+ #include "distributed-arp-table.h"
+ #include "gateway_client.h"
+ #include "hard-interface.h"
+ #include "multicast.h"
+ #include "network-coding.h"
+ #include "send.h"
+ #include "translation-table.h"
+ /**
+  * batadv_skb_head_push() - Increase header size and move (push) head pointer
+  * @skb: packet buffer which should be modified
+  * @len: number of bytes to add
+  *
+  * Return: 0 on success or negative error number in case of failure
+  */
+ int batadv_skb_head_push(struct sk_buff *skb, unsigned int len)
+ {
+       int result;
+       /* TODO: We must check if we can release all references to non-payload
+        * data using __skb_header_release in our skbs to allow skb_cow_header
+        * to work optimally. This means that those skbs are not allowed to read
+        * or write any data which is before the current position of skb->data
+        * after that call and thus allow other skbs with the same data buffer
+        * to write freely in that area.
+        */
+       result = skb_cow_head(skb, len);
+       if (result < 0)
+               return result;
+       skb_push(skb, len);
+       return 0;
+ }
+ static int batadv_interface_open(struct net_device *dev)
+ {
+       netif_start_queue(dev);
+       return 0;
+ }
+ static int batadv_interface_release(struct net_device *dev)
+ {
+       netif_stop_queue(dev);
+       return 0;
+ }
+ /**
+  * batadv_sum_counter() - Sum the cpu-local counters for index 'idx'
+  * @bat_priv: the bat priv with all the mesh interface information
+  * @idx: index of counter to sum up
+  *
+  * Return: sum of all cpu-local counters
+  */
+ static u64 batadv_sum_counter(struct batadv_priv *bat_priv,  size_t idx)
+ {
+       u64 *counters, sum = 0;
+       int cpu;
+       for_each_possible_cpu(cpu) {
+               counters = per_cpu_ptr(bat_priv->bat_counters, cpu);
+               sum += counters[idx];
+       }
+       return sum;
+ }
+ static struct net_device_stats *batadv_interface_stats(struct net_device *dev)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       struct net_device_stats *stats = &dev->stats;
+       stats->tx_packets = batadv_sum_counter(bat_priv, BATADV_CNT_TX);
+       stats->tx_bytes = batadv_sum_counter(bat_priv, BATADV_CNT_TX_BYTES);
+       stats->tx_dropped = batadv_sum_counter(bat_priv, BATADV_CNT_TX_DROPPED);
+       stats->rx_packets = batadv_sum_counter(bat_priv, BATADV_CNT_RX);
+       stats->rx_bytes = batadv_sum_counter(bat_priv, BATADV_CNT_RX_BYTES);
+       return stats;
+ }
+ static int batadv_interface_set_mac_addr(struct net_device *dev, void *p)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       struct batadv_meshif_vlan *vlan;
+       struct sockaddr *addr = p;
+       u8 old_addr[ETH_ALEN];
+       if (!is_valid_ether_addr(addr->sa_data))
+               return -EADDRNOTAVAIL;
+       ether_addr_copy(old_addr, dev->dev_addr);
+       eth_hw_addr_set(dev, addr->sa_data);
+       /* only modify transtable if it has been initialized before */
+       if (atomic_read(&bat_priv->mesh_state) != BATADV_MESH_ACTIVE)
+               return 0;
+       rcu_read_lock();
+       hlist_for_each_entry_rcu(vlan, &bat_priv->meshif_vlan_list, list) {
+               batadv_tt_local_remove(bat_priv, old_addr, vlan->vid,
+                                      "mac address changed", false);
+               batadv_tt_local_add(dev, addr->sa_data, vlan->vid,
+                                   BATADV_NULL_IFINDEX, BATADV_NO_MARK);
+       }
+       rcu_read_unlock();
+       return 0;
+ }
+ static int batadv_interface_change_mtu(struct net_device *dev, int new_mtu)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       /* check ranges */
+       if (new_mtu < ETH_MIN_MTU || new_mtu > batadv_hardif_min_mtu(dev))
+               return -EINVAL;
+       WRITE_ONCE(dev->mtu, new_mtu);
+       bat_priv->mtu_set_by_user = new_mtu;
+       return 0;
+ }
+ /**
+  * batadv_interface_set_rx_mode() - set the rx mode of a device
+  * @dev: registered network device to modify
+  *
+  * We do not actually need to set any rx filters for the virtual batman
+  * mesh interface. However a dummy handler enables a user to set static
+  * multicast listeners for instance.
+  */
+ static void batadv_interface_set_rx_mode(struct net_device *dev)
+ {
+ }
+ static netdev_tx_t batadv_interface_tx(struct sk_buff *skb,
+                                      struct net_device *mesh_iface)
+ {
+       struct ethhdr *ethhdr;
+       struct batadv_priv *bat_priv = netdev_priv(mesh_iface);
+       struct batadv_hard_iface *primary_if = NULL;
+       struct batadv_bcast_packet *bcast_packet;
+       static const u8 stp_addr[ETH_ALEN] = {0x01, 0x80, 0xC2, 0x00,
+                                             0x00, 0x00};
+       static const u8 ectp_addr[ETH_ALEN] = {0xCF, 0x00, 0x00, 0x00,
+                                              0x00, 0x00};
+       enum batadv_dhcp_recipient dhcp_rcp = BATADV_DHCP_NO;
+       u8 *dst_hint = NULL, chaddr[ETH_ALEN];
+       struct vlan_ethhdr *vhdr;
+       unsigned int header_len = 0;
+       int data_len = skb->len, ret;
+       unsigned long brd_delay = 0;
+       bool do_bcast = false, client_added;
+       unsigned short vid;
+       u32 seqno;
+       int gw_mode;
+       enum batadv_forw_mode forw_mode = BATADV_FORW_BCAST;
+       int mcast_is_routable = 0;
+       int network_offset = ETH_HLEN;
+       __be16 proto;
+       if (atomic_read(&bat_priv->mesh_state) != BATADV_MESH_ACTIVE)
+               goto dropped;
+       /* reset control block to avoid left overs from previous users */
+       memset(skb->cb, 0, sizeof(struct batadv_skb_cb));
+       netif_trans_update(mesh_iface);
+       vid = batadv_get_vid(skb, 0);
+       skb_reset_mac_header(skb);
+       ethhdr = eth_hdr(skb);
+       proto = ethhdr->h_proto;
+       switch (ntohs(proto)) {
+       case ETH_P_8021Q:
+               if (!pskb_may_pull(skb, sizeof(*vhdr)))
+                       goto dropped;
+               vhdr = vlan_eth_hdr(skb);
+               proto = vhdr->h_vlan_encapsulated_proto;
+               /* drop batman-in-batman packets to prevent loops */
+               if (proto != htons(ETH_P_BATMAN)) {
+                       network_offset += VLAN_HLEN;
+                       break;
+               }
+               fallthrough;
+       case ETH_P_BATMAN:
+               goto dropped;
+       }
+       skb_set_network_header(skb, network_offset);
+       if (batadv_bla_tx(bat_priv, skb, vid))
+               goto dropped;
+       /* skb->data might have been reallocated by batadv_bla_tx() */
+       ethhdr = eth_hdr(skb);
+       /* Register the client MAC in the transtable */
+       if (!is_multicast_ether_addr(ethhdr->h_source) &&
+           !batadv_bla_is_loopdetect_mac(ethhdr->h_source)) {
+               client_added = batadv_tt_local_add(mesh_iface, ethhdr->h_source,
+                                                  vid, skb->skb_iif,
+                                                  skb->mark);
+               if (!client_added)
+                       goto dropped;
+       }
+       /* Snoop address candidates from DHCPACKs for early DAT filling */
+       batadv_dat_snoop_outgoing_dhcp_ack(bat_priv, skb, proto, vid);
+       /* don't accept stp packets. STP does not help in meshes.
+        * better use the bridge loop avoidance ...
+        *
+        * The same goes for ECTP sent at least by some Cisco Switches,
+        * it might confuse the mesh when used with bridge loop avoidance.
+        */
+       if (batadv_compare_eth(ethhdr->h_dest, stp_addr))
+               goto dropped;
+       if (batadv_compare_eth(ethhdr->h_dest, ectp_addr))
+               goto dropped;
+       gw_mode = atomic_read(&bat_priv->gw.mode);
+       if (is_multicast_ether_addr(ethhdr->h_dest)) {
+               /* if gw mode is off, broadcast every packet */
+               if (gw_mode == BATADV_GW_MODE_OFF) {
+                       do_bcast = true;
+                       goto send;
+               }
+               dhcp_rcp = batadv_gw_dhcp_recipient_get(skb, &header_len,
+                                                       chaddr);
+               /* skb->data may have been modified by
+                * batadv_gw_dhcp_recipient_get()
+                */
+               ethhdr = eth_hdr(skb);
+               /* if gw_mode is on, broadcast any non-DHCP message.
+                * All the DHCP packets are going to be sent as unicast
+                */
+               if (dhcp_rcp == BATADV_DHCP_NO) {
+                       do_bcast = true;
+                       goto send;
+               }
+               if (dhcp_rcp == BATADV_DHCP_TO_CLIENT)
+                       dst_hint = chaddr;
+               else if ((gw_mode == BATADV_GW_MODE_SERVER) &&
+                        (dhcp_rcp == BATADV_DHCP_TO_SERVER))
+                       /* gateways should not forward any DHCP message if
+                        * directed to a DHCP server
+                        */
+                       goto dropped;
+ send:
+               if (do_bcast && !is_broadcast_ether_addr(ethhdr->h_dest)) {
+                       forw_mode = batadv_mcast_forw_mode(bat_priv, skb, vid,
+                                                          &mcast_is_routable);
+                       switch (forw_mode) {
+                       case BATADV_FORW_BCAST:
+                               break;
+                       case BATADV_FORW_UCASTS:
+                       case BATADV_FORW_MCAST:
+                               do_bcast = false;
+                               break;
+                       case BATADV_FORW_NONE:
+                               fallthrough;
+                       default:
+                               goto dropped;
+                       }
+               }
+       }
+       batadv_skb_set_priority(skb, 0);
+       /* ethernet packet should be broadcasted */
+       if (do_bcast) {
+               primary_if = batadv_primary_if_get_selected(bat_priv);
+               if (!primary_if)
+                       goto dropped;
+               /* in case of ARP request, we do not immediately broadcasti the
+                * packet, instead we first wait for DAT to try to retrieve the
+                * correct ARP entry
+                */
+               if (batadv_dat_snoop_outgoing_arp_request(bat_priv, skb))
+                       brd_delay = msecs_to_jiffies(ARP_REQ_DELAY);
+               if (batadv_skb_head_push(skb, sizeof(*bcast_packet)) < 0)
+                       goto dropped;
+               bcast_packet = (struct batadv_bcast_packet *)skb->data;
+               bcast_packet->version = BATADV_COMPAT_VERSION;
+               bcast_packet->ttl = BATADV_TTL - 1;
+               /* batman packet type: broadcast */
+               bcast_packet->packet_type = BATADV_BCAST;
+               bcast_packet->reserved = 0;
+               /* hw address of first interface is the orig mac because only
+                * this mac is known throughout the mesh
+                */
+               ether_addr_copy(bcast_packet->orig,
+                               primary_if->net_dev->dev_addr);
+               /* set broadcast sequence number */
+               seqno = atomic_inc_return(&bat_priv->bcast_seqno);
+               bcast_packet->seqno = htonl(seqno);
+               batadv_send_bcast_packet(bat_priv, skb, brd_delay, true);
+       /* unicast packet */
+       } else {
+               /* DHCP packets going to a server will use the GW feature */
+               if (dhcp_rcp == BATADV_DHCP_TO_SERVER) {
+                       ret = batadv_gw_out_of_range(bat_priv, skb);
+                       if (ret)
+                               goto dropped;
+                       ret = batadv_send_skb_via_gw(bat_priv, skb, vid);
+               } else if (forw_mode == BATADV_FORW_UCASTS) {
+                       ret = batadv_mcast_forw_send(bat_priv, skb, vid,
+                                                    mcast_is_routable);
+               } else if (forw_mode == BATADV_FORW_MCAST) {
+                       ret = batadv_mcast_forw_mcsend(bat_priv, skb);
+               } else {
+                       if (batadv_dat_snoop_outgoing_arp_request(bat_priv,
+                                                                 skb))
+                               goto dropped;
+                       batadv_dat_snoop_outgoing_arp_reply(bat_priv, skb);
+                       ret = batadv_send_skb_via_tt(bat_priv, skb, dst_hint,
+                                                    vid);
+               }
+               if (ret != NET_XMIT_SUCCESS)
+                       goto dropped_freed;
+       }
+       batadv_inc_counter(bat_priv, BATADV_CNT_TX);
+       batadv_add_counter(bat_priv, BATADV_CNT_TX_BYTES, data_len);
+       goto end;
+ dropped:
+       kfree_skb(skb);
+ dropped_freed:
+       batadv_inc_counter(bat_priv, BATADV_CNT_TX_DROPPED);
+ end:
+       batadv_hardif_put(primary_if);
+       return NETDEV_TX_OK;
+ }
+ /**
+  * batadv_interface_rx() - receive ethernet frame on local batman-adv interface
+  * @mesh_iface: local interface which will receive the ethernet frame
+  * @skb: ethernet frame for @mesh_iface
+  * @hdr_size: size of already parsed batman-adv header
+  * @orig_node: originator from which the batman-adv packet was sent
+  *
+  * Sends an ethernet frame to the receive path of the local @mesh_iface.
+  * skb->data has still point to the batman-adv header with the size @hdr_size.
+  * The caller has to have parsed this header already and made sure that at least
+  * @hdr_size bytes are still available for pull in @skb.
+  *
+  * The packet may still get dropped. This can happen when the encapsulated
+  * ethernet frame is invalid or contains again an batman-adv packet. Also
+  * unicast packets will be dropped directly when it was sent between two
+  * isolated clients.
+  */
+ void batadv_interface_rx(struct net_device *mesh_iface,
+                        struct sk_buff *skb, int hdr_size,
+                        struct batadv_orig_node *orig_node)
+ {
+       struct batadv_bcast_packet *batadv_bcast_packet;
+       struct batadv_priv *bat_priv = netdev_priv(mesh_iface);
+       struct vlan_ethhdr *vhdr;
+       struct ethhdr *ethhdr;
+       unsigned short vid;
+       int packet_type;
+       batadv_bcast_packet = (struct batadv_bcast_packet *)skb->data;
+       packet_type = batadv_bcast_packet->packet_type;
+       skb_pull_rcsum(skb, hdr_size);
+       skb_reset_mac_header(skb);
+       /* clean the netfilter state now that the batman-adv header has been
+        * removed
+        */
+       nf_reset_ct(skb);
+       if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
+               goto dropped;
+       vid = batadv_get_vid(skb, 0);
+       ethhdr = eth_hdr(skb);
+       switch (ntohs(ethhdr->h_proto)) {
+       case ETH_P_8021Q:
+               if (!pskb_may_pull(skb, VLAN_ETH_HLEN))
+                       goto dropped;
+               vhdr = skb_vlan_eth_hdr(skb);
+               /* drop batman-in-batman packets to prevent loops */
+               if (vhdr->h_vlan_encapsulated_proto != htons(ETH_P_BATMAN))
+                       break;
+               fallthrough;
+       case ETH_P_BATMAN:
+               goto dropped;
+       }
+       /* skb->dev & skb->pkt_type are set here */
+       skb->protocol = eth_type_trans(skb, mesh_iface);
+       skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
+       batadv_inc_counter(bat_priv, BATADV_CNT_RX);
+       batadv_add_counter(bat_priv, BATADV_CNT_RX_BYTES,
+                          skb->len + ETH_HLEN);
+       /* Let the bridge loop avoidance check the packet. If will
+        * not handle it, we can safely push it up.
+        */
+       if (batadv_bla_rx(bat_priv, skb, vid, packet_type))
+               goto out;
+       if (orig_node)
+               batadv_tt_add_temporary_global_entry(bat_priv, orig_node,
+                                                    ethhdr->h_source, vid);
+       if (is_multicast_ether_addr(ethhdr->h_dest)) {
+               /* set the mark on broadcast packets if AP isolation is ON and
+                * the packet is coming from an "isolated" client
+                */
+               if (batadv_vlan_ap_isola_get(bat_priv, vid) &&
+                   batadv_tt_global_is_isolated(bat_priv, ethhdr->h_source,
+                                                vid)) {
+                       /* save bits in skb->mark not covered by the mask and
+                        * apply the mark on the rest
+                        */
+                       skb->mark &= ~bat_priv->isolation_mark_mask;
+                       skb->mark |= bat_priv->isolation_mark;
+               }
+       } else if (batadv_is_ap_isolated(bat_priv, ethhdr->h_source,
+                                        ethhdr->h_dest, vid)) {
+               goto dropped;
+       }
+       netif_rx(skb);
+       goto out;
+ dropped:
+       kfree_skb(skb);
+ out:
+       return;
+ }
+ /**
+  * batadv_meshif_vlan_release() - release vlan from lists and queue for free
+  *  after rcu grace period
+  * @ref: kref pointer of the vlan object
+  */
+ void batadv_meshif_vlan_release(struct kref *ref)
+ {
+       struct batadv_meshif_vlan *vlan;
+       vlan = container_of(ref, struct batadv_meshif_vlan, refcount);
+       spin_lock_bh(&vlan->bat_priv->meshif_vlan_list_lock);
+       hlist_del_rcu(&vlan->list);
+       spin_unlock_bh(&vlan->bat_priv->meshif_vlan_list_lock);
+       kfree_rcu(vlan, rcu);
+ }
+ /**
+  * batadv_meshif_vlan_get() - get the vlan object for a specific vid
+  * @bat_priv: the bat priv with all the mesh interface information
+  * @vid: the identifier of the vlan object to retrieve
+  *
+  * Return: the private data of the vlan matching the vid passed as argument or
+  * NULL otherwise. The refcounter of the returned object is incremented by 1.
+  */
+ struct batadv_meshif_vlan *batadv_meshif_vlan_get(struct batadv_priv *bat_priv,
+                                                 unsigned short vid)
+ {
+       struct batadv_meshif_vlan *vlan_tmp, *vlan = NULL;
+       rcu_read_lock();
+       hlist_for_each_entry_rcu(vlan_tmp, &bat_priv->meshif_vlan_list, list) {
+               if (vlan_tmp->vid != vid)
+                       continue;
+               if (!kref_get_unless_zero(&vlan_tmp->refcount))
+                       continue;
+               vlan = vlan_tmp;
+               break;
+       }
+       rcu_read_unlock();
+       return vlan;
+ }
+ /**
+  * batadv_meshif_create_vlan() - allocate the needed resources for a new vlan
+  * @bat_priv: the bat priv with all the mesh interface information
+  * @vid: the VLAN identifier
+  *
+  * Return: 0 on success, a negative error otherwise.
+  */
+ int batadv_meshif_create_vlan(struct batadv_priv *bat_priv, unsigned short vid)
+ {
+       struct batadv_meshif_vlan *vlan;
+       spin_lock_bh(&bat_priv->meshif_vlan_list_lock);
+       vlan = batadv_meshif_vlan_get(bat_priv, vid);
+       if (vlan) {
+               batadv_meshif_vlan_put(vlan);
+               spin_unlock_bh(&bat_priv->meshif_vlan_list_lock);
+               return -EEXIST;
+       }
+       vlan = kzalloc(sizeof(*vlan), GFP_ATOMIC);
+       if (!vlan) {
+               spin_unlock_bh(&bat_priv->meshif_vlan_list_lock);
+               return -ENOMEM;
+       }
+       vlan->bat_priv = bat_priv;
+       vlan->vid = vid;
+       kref_init(&vlan->refcount);
+       atomic_set(&vlan->ap_isolation, 0);
+       kref_get(&vlan->refcount);
+       hlist_add_head_rcu(&vlan->list, &bat_priv->meshif_vlan_list);
+       spin_unlock_bh(&bat_priv->meshif_vlan_list_lock);
+       /* add a new TT local entry. This one will be marked with the NOPURGE
+        * flag
+        */
+       batadv_tt_local_add(bat_priv->mesh_iface,
+                           bat_priv->mesh_iface->dev_addr, vid,
+                           BATADV_NULL_IFINDEX, BATADV_NO_MARK);
+       /* don't return reference to new meshif_vlan */
+       batadv_meshif_vlan_put(vlan);
+       return 0;
+ }
+ /**
+  * batadv_meshif_destroy_vlan() - remove and destroy a meshif_vlan object
+  * @bat_priv: the bat priv with all the mesh interface information
+  * @vlan: the object to remove
+  */
+ static void batadv_meshif_destroy_vlan(struct batadv_priv *bat_priv,
+                                      struct batadv_meshif_vlan *vlan)
+ {
+       /* explicitly remove the associated TT local entry because it is marked
+        * with the NOPURGE flag
+        */
+       batadv_tt_local_remove(bat_priv, bat_priv->mesh_iface->dev_addr,
+                              vlan->vid, "vlan interface destroyed", false);
+       batadv_meshif_vlan_put(vlan);
+ }
+ /**
+  * batadv_interface_add_vid() - ndo_add_vid API implementation
+  * @dev: the netdev of the mesh interface
+  * @proto: protocol of the vlan id
+  * @vid: identifier of the new vlan
+  *
+  * Set up all the internal structures for handling the new vlan on top of the
+  * mesh interface
+  *
+  * Return: 0 on success or a negative error code in case of failure.
+  */
+ static int batadv_interface_add_vid(struct net_device *dev, __be16 proto,
+                                   unsigned short vid)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       struct batadv_meshif_vlan *vlan;
+       /* only 802.1Q vlans are supported.
+        * batman-adv does not know how to handle other types
+        */
+       if (proto != htons(ETH_P_8021Q))
+               return -EINVAL;
+       /* VID 0 is only used to indicate "priority tag" frames which only
+        * contain priority information and no VID. No management structures
+        * should be created for this VID and it should be handled like an
+        * untagged frame.
+        */
+       if (vid == 0)
+               return 0;
+       vid |= BATADV_VLAN_HAS_TAG;
+       /* if a new vlan is getting created and it already exists, it means that
+        * it was not deleted yet. batadv_meshif_vlan_get() increases the
+        * refcount in order to revive the object.
+        *
+        * if it does not exist then create it.
+        */
+       vlan = batadv_meshif_vlan_get(bat_priv, vid);
+       if (!vlan)
+               return batadv_meshif_create_vlan(bat_priv, vid);
+       /* add a new TT local entry. This one will be marked with the NOPURGE
+        * flag. This must be added again, even if the vlan object already
+        * exists, because the entry was deleted by kill_vid()
+        */
+       batadv_tt_local_add(bat_priv->mesh_iface,
+                           bat_priv->mesh_iface->dev_addr, vid,
+                           BATADV_NULL_IFINDEX, BATADV_NO_MARK);
+       return 0;
+ }
+ /**
+  * batadv_interface_kill_vid() - ndo_kill_vid API implementation
+  * @dev: the netdev of the mesh interface
+  * @proto: protocol of the vlan id
+  * @vid: identifier of the deleted vlan
+  *
+  * Destroy all the internal structures used to handle the vlan identified by vid
+  * on top of the mesh interface
+  *
+  * Return: 0 on success, -EINVAL if the specified prototype is not ETH_P_8021Q
+  * or -ENOENT if the specified vlan id wasn't registered.
+  */
+ static int batadv_interface_kill_vid(struct net_device *dev, __be16 proto,
+                                    unsigned short vid)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       struct batadv_meshif_vlan *vlan;
+       /* only 802.1Q vlans are supported. batman-adv does not know how to
+        * handle other types
+        */
+       if (proto != htons(ETH_P_8021Q))
+               return -EINVAL;
+       /* "priority tag" frames are handled like "untagged" frames
+        * and no meshif_vlan needs to be destroyed
+        */
+       if (vid == 0)
+               return 0;
+       vlan = batadv_meshif_vlan_get(bat_priv, vid | BATADV_VLAN_HAS_TAG);
+       if (!vlan)
+               return -ENOENT;
+       batadv_meshif_destroy_vlan(bat_priv, vlan);
+       /* finally free the vlan object */
+       batadv_meshif_vlan_put(vlan);
+       return 0;
+ }
+ /* batman-adv network devices have devices nesting below it and are a special
+  * "super class" of normal network devices; split their locks off into a
+  * separate class since they always nest.
+  */
+ static struct lock_class_key batadv_netdev_xmit_lock_key;
+ static struct lock_class_key batadv_netdev_addr_lock_key;
+ /**
+  * batadv_set_lockdep_class_one() - Set lockdep class for a single tx queue
+  * @dev: device which owns the tx queue
+  * @txq: tx queue to modify
+  * @_unused: always NULL
+  */
+ static void batadv_set_lockdep_class_one(struct net_device *dev,
+                                        struct netdev_queue *txq,
+                                        void *_unused)
+ {
+       lockdep_set_class(&txq->_xmit_lock, &batadv_netdev_xmit_lock_key);
+ }
+ /**
+  * batadv_set_lockdep_class() - Set txq and addr_list lockdep class
+  * @dev: network device to modify
+  */
+ static void batadv_set_lockdep_class(struct net_device *dev)
+ {
+       lockdep_set_class(&dev->addr_list_lock, &batadv_netdev_addr_lock_key);
+       netdev_for_each_tx_queue(dev, batadv_set_lockdep_class_one, NULL);
+ }
+ /**
+  * batadv_meshif_init_late() - late stage initialization of mesh interface
+  * @dev: registered network device to modify
+  *
+  * Return: error code on failures
+  */
+ static int batadv_meshif_init_late(struct net_device *dev)
+ {
+       struct batadv_priv *bat_priv;
+       u32 random_seqno;
+       int ret;
+       size_t cnt_len = sizeof(u64) * BATADV_CNT_NUM;
+       batadv_set_lockdep_class(dev);
+       bat_priv = netdev_priv(dev);
+       bat_priv->mesh_iface = dev;
+       /* batadv_interface_stats() needs to be available as soon as
+        * register_netdevice() has been called
+        */
+       bat_priv->bat_counters = __alloc_percpu(cnt_len, __alignof__(u64));
+       if (!bat_priv->bat_counters)
+               return -ENOMEM;
+       atomic_set(&bat_priv->aggregated_ogms, 1);
+       atomic_set(&bat_priv->bonding, 0);
+ #ifdef CONFIG_BATMAN_ADV_BLA
+       atomic_set(&bat_priv->bridge_loop_avoidance, 1);
+ #endif
+ #ifdef CONFIG_BATMAN_ADV_DAT
+       atomic_set(&bat_priv->distributed_arp_table, 1);
+ #endif
+ #ifdef CONFIG_BATMAN_ADV_MCAST
+       atomic_set(&bat_priv->multicast_mode, 1);
+       atomic_set(&bat_priv->multicast_fanout, 16);
+       atomic_set(&bat_priv->mcast.num_want_all_unsnoopables, 0);
+       atomic_set(&bat_priv->mcast.num_want_all_ipv4, 0);
+       atomic_set(&bat_priv->mcast.num_want_all_ipv6, 0);
+       atomic_set(&bat_priv->mcast.num_no_mc_ptype_capa, 0);
+ #endif
+       atomic_set(&bat_priv->gw.mode, BATADV_GW_MODE_OFF);
+       atomic_set(&bat_priv->gw.bandwidth_down, 100);
+       atomic_set(&bat_priv->gw.bandwidth_up, 20);
+       atomic_set(&bat_priv->orig_interval, 1000);
+       atomic_set(&bat_priv->hop_penalty, 30);
+ #ifdef CONFIG_BATMAN_ADV_DEBUG
+       atomic_set(&bat_priv->log_level, 0);
+ #endif
+       atomic_set(&bat_priv->fragmentation, 1);
+       atomic_set(&bat_priv->packet_size_max, BATADV_MAX_MTU);
+       atomic_set(&bat_priv->bcast_queue_left, BATADV_BCAST_QUEUE_LEN);
+       atomic_set(&bat_priv->batman_queue_left, BATADV_BATMAN_QUEUE_LEN);
+       atomic_set(&bat_priv->mesh_state, BATADV_MESH_INACTIVE);
+       atomic_set(&bat_priv->bcast_seqno, 1);
+       atomic_set(&bat_priv->tt.vn, 0);
+       atomic_set(&bat_priv->tt.ogm_append_cnt, 0);
+ #ifdef CONFIG_BATMAN_ADV_BLA
+       atomic_set(&bat_priv->bla.num_requests, 0);
+ #endif
+       atomic_set(&bat_priv->tp_num, 0);
+       WRITE_ONCE(bat_priv->tt.local_changes, 0);
+       bat_priv->tt.last_changeset = NULL;
+       bat_priv->tt.last_changeset_len = 0;
+       bat_priv->isolation_mark = 0;
+       bat_priv->isolation_mark_mask = 0;
+       /* randomize initial seqno to avoid collision */
+       get_random_bytes(&random_seqno, sizeof(random_seqno));
+       atomic_set(&bat_priv->frag_seqno, random_seqno);
+       bat_priv->primary_if = NULL;
+       batadv_nc_init_bat_priv(bat_priv);
+       if (!bat_priv->algo_ops) {
+               ret = batadv_algo_select(bat_priv, batadv_routing_algo);
+               if (ret < 0)
+                       goto free_bat_counters;
+       }
+       ret = batadv_mesh_init(dev);
+       if (ret < 0)
+               goto free_bat_counters;
+       return 0;
+ free_bat_counters:
+       free_percpu(bat_priv->bat_counters);
+       bat_priv->bat_counters = NULL;
+       return ret;
+ }
+ /**
+  * batadv_meshif_slave_add() - Add a slave interface to a batadv_mesh_interface
+  * @dev: batadv_mesh_interface used as master interface
+  * @slave_dev: net_device which should become the slave interface
+  * @extack: extended ACK report struct
+  *
+  * Return: 0 if successful or error otherwise.
+  */
+ static int batadv_meshif_slave_add(struct net_device *dev,
+                                  struct net_device *slave_dev,
+                                  struct netlink_ext_ack *extack)
+ {
+       struct batadv_hard_iface *hard_iface;
+       int ret = -EINVAL;
+       hard_iface = batadv_hardif_get_by_netdev(slave_dev);
+       if (!hard_iface || hard_iface->mesh_iface)
+               goto out;
+       ret = batadv_hardif_enable_interface(hard_iface, dev);
+ out:
+       batadv_hardif_put(hard_iface);
+       return ret;
+ }
+ /**
+  * batadv_meshif_slave_del() - Delete a slave iface from a batadv_mesh_interface
+  * @dev: batadv_mesh_interface used as master interface
+  * @slave_dev: net_device which should be removed from the master interface
+  *
+  * Return: 0 if successful or error otherwise.
+  */
+ static int batadv_meshif_slave_del(struct net_device *dev,
+                                  struct net_device *slave_dev)
+ {
+       struct batadv_hard_iface *hard_iface;
+       int ret = -EINVAL;
+       hard_iface = batadv_hardif_get_by_netdev(slave_dev);
+       if (!hard_iface || hard_iface->mesh_iface != dev)
+               goto out;
+       batadv_hardif_disable_interface(hard_iface);
+       ret = 0;
+ out:
+       batadv_hardif_put(hard_iface);
+       return ret;
+ }
+ static const struct net_device_ops batadv_netdev_ops = {
+       .ndo_init = batadv_meshif_init_late,
+       .ndo_open = batadv_interface_open,
+       .ndo_stop = batadv_interface_release,
+       .ndo_get_stats = batadv_interface_stats,
+       .ndo_vlan_rx_add_vid = batadv_interface_add_vid,
+       .ndo_vlan_rx_kill_vid = batadv_interface_kill_vid,
+       .ndo_set_mac_address = batadv_interface_set_mac_addr,
+       .ndo_change_mtu = batadv_interface_change_mtu,
+       .ndo_set_rx_mode = batadv_interface_set_rx_mode,
+       .ndo_start_xmit = batadv_interface_tx,
+       .ndo_validate_addr = eth_validate_addr,
+       .ndo_add_slave = batadv_meshif_slave_add,
+       .ndo_del_slave = batadv_meshif_slave_del,
+ };
+ static void batadv_get_drvinfo(struct net_device *dev,
+                              struct ethtool_drvinfo *info)
+ {
+       strscpy(info->driver, "B.A.T.M.A.N. advanced", sizeof(info->driver));
+       strscpy(info->version, BATADV_SOURCE_VERSION, sizeof(info->version));
+       strscpy(info->fw_version, "N/A", sizeof(info->fw_version));
+       strscpy(info->bus_info, "batman", sizeof(info->bus_info));
+ }
+ /* Inspired by drivers/net/ethernet/dlink/sundance.c:1702
+  * Declare each description string in struct.name[] to get fixed sized buffer
+  * and compile time checking for strings longer than ETH_GSTRING_LEN.
+  */
+ static const struct {
+       const char name[ETH_GSTRING_LEN];
+ } batadv_counters_strings[] = {
+       { "tx" },
+       { "tx_bytes" },
+       { "tx_dropped" },
+       { "rx" },
+       { "rx_bytes" },
+       { "forward" },
+       { "forward_bytes" },
+       { "mgmt_tx" },
+       { "mgmt_tx_bytes" },
+       { "mgmt_rx" },
+       { "mgmt_rx_bytes" },
+       { "frag_tx" },
+       { "frag_tx_bytes" },
+       { "frag_rx" },
+       { "frag_rx_bytes" },
+       { "frag_fwd" },
+       { "frag_fwd_bytes" },
+       { "tt_request_tx" },
+       { "tt_request_rx" },
+       { "tt_response_tx" },
+       { "tt_response_rx" },
+       { "tt_roam_adv_tx" },
+       { "tt_roam_adv_rx" },
+ #ifdef CONFIG_BATMAN_ADV_MCAST
+       { "mcast_tx" },
+       { "mcast_tx_bytes" },
+       { "mcast_tx_local" },
+       { "mcast_tx_local_bytes" },
+       { "mcast_rx" },
+       { "mcast_rx_bytes" },
+       { "mcast_rx_local" },
+       { "mcast_rx_local_bytes" },
+       { "mcast_fwd" },
+       { "mcast_fwd_bytes" },
+ #endif
+ #ifdef CONFIG_BATMAN_ADV_DAT
+       { "dat_get_tx" },
+       { "dat_get_rx" },
+       { "dat_put_tx" },
+       { "dat_put_rx" },
+       { "dat_cached_reply_tx" },
+ #endif
+ #ifdef CONFIG_BATMAN_ADV_NC
+       { "nc_code" },
+       { "nc_code_bytes" },
+       { "nc_recode" },
+       { "nc_recode_bytes" },
+       { "nc_buffer" },
+       { "nc_decode" },
+       { "nc_decode_bytes" },
+       { "nc_decode_failed" },
+       { "nc_sniffed" },
+ #endif
+ };
+ static void batadv_get_strings(struct net_device *dev, u32 stringset, u8 *data)
+ {
+       if (stringset == ETH_SS_STATS)
+               memcpy(data, batadv_counters_strings,
+                      sizeof(batadv_counters_strings));
+ }
+ static void batadv_get_ethtool_stats(struct net_device *dev,
+                                    struct ethtool_stats *stats, u64 *data)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       int i;
+       for (i = 0; i < BATADV_CNT_NUM; i++)
+               data[i] = batadv_sum_counter(bat_priv, i);
+ }
+ static int batadv_get_sset_count(struct net_device *dev, int stringset)
+ {
+       if (stringset == ETH_SS_STATS)
+               return BATADV_CNT_NUM;
+       return -EOPNOTSUPP;
+ }
+ static const struct ethtool_ops batadv_ethtool_ops = {
+       .get_drvinfo = batadv_get_drvinfo,
+       .get_link = ethtool_op_get_link,
+       .get_strings = batadv_get_strings,
+       .get_ethtool_stats = batadv_get_ethtool_stats,
+       .get_sset_count = batadv_get_sset_count,
+ };
+ /**
+  * batadv_meshif_free() - Deconstructor of batadv_mesh_interface
+  * @dev: Device to cleanup and remove
+  */
+ static void batadv_meshif_free(struct net_device *dev)
+ {
+       batadv_mesh_free(dev);
+       /* some scheduled RCU callbacks need the bat_priv struct to accomplish
+        * their tasks. Wait for them all to be finished before freeing the
+        * netdev and its private data (bat_priv)
+        */
+       rcu_barrier();
+ }
+ /**
+  * batadv_meshif_init_early() - early stage initialization of mesh interface
+  * @dev: registered network device to modify
+  */
+ static void batadv_meshif_init_early(struct net_device *dev)
+ {
+       ether_setup(dev);
+       dev->netdev_ops = &batadv_netdev_ops;
+       dev->needs_free_netdev = true;
+       dev->priv_destructor = batadv_meshif_free;
+       dev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
+       dev->priv_flags |= IFF_NO_QUEUE;
+       dev->lltx = true;
++      dev->netns_immutable = true;
+       /* can't call min_mtu, because the needed variables
+        * have not been initialized yet
+        */
+       dev->mtu = ETH_DATA_LEN;
+       dev->max_mtu = BATADV_MAX_MTU;
+       /* generate random address */
+       eth_hw_addr_random(dev);
+       dev->ethtool_ops = &batadv_ethtool_ops;
+ }
+ /**
+  * batadv_meshif_validate() - validate configuration of new batadv link
+  * @tb: IFLA_INFO_DATA netlink attributes
+  * @data: enum batadv_ifla_attrs attributes
+  * @extack: extended ACK report struct
+  *
+  * Return: 0 if successful or error otherwise.
+  */
+ static int batadv_meshif_validate(struct nlattr *tb[], struct nlattr *data[],
+                                 struct netlink_ext_ack *extack)
+ {
+       struct batadv_algo_ops *algo_ops;
+       if (!data)
+               return 0;
+       if (data[IFLA_BATADV_ALGO_NAME]) {
+               algo_ops = batadv_algo_get(nla_data(data[IFLA_BATADV_ALGO_NAME]));
+               if (!algo_ops)
+                       return -EINVAL;
+       }
+       return 0;
+ }
+ /**
+  * batadv_meshif_newlink() - pre-initialize and register new batadv link
+  * @dev: network device to register
+  * @params: rtnl newlink parameters
+  * @extack: extended ACK report struct
+  *
+  * Return: 0 if successful or error otherwise.
+  */
+ static int batadv_meshif_newlink(struct net_device *dev,
+                                struct rtnl_newlink_params *params,
+                                struct netlink_ext_ack *extack)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       struct nlattr **data = params->data;
+       const char *algo_name;
+       int err;
+       if (data && data[IFLA_BATADV_ALGO_NAME]) {
+               algo_name = nla_data(data[IFLA_BATADV_ALGO_NAME]);
+               err = batadv_algo_select(bat_priv, algo_name);
+               if (err)
+                       return -EINVAL;
+       }
+       return register_netdevice(dev);
+ }
+ /**
+  * batadv_meshif_destroy_netlink() - deletion of batadv_mesh_interface via
+  *  netlink
+  * @mesh_iface: the to-be-removed batman-adv interface
+  * @head: list pointer
+  */
+ static void batadv_meshif_destroy_netlink(struct net_device *mesh_iface,
+                                         struct list_head *head)
+ {
+       struct batadv_priv *bat_priv = netdev_priv(mesh_iface);
+       struct batadv_hard_iface *hard_iface;
+       struct batadv_meshif_vlan *vlan;
+       list_for_each_entry(hard_iface, &batadv_hardif_list, list) {
+               if (hard_iface->mesh_iface == mesh_iface)
+                       batadv_hardif_disable_interface(hard_iface);
+       }
+       /* destroy the "untagged" VLAN */
+       vlan = batadv_meshif_vlan_get(bat_priv, BATADV_NO_FLAGS);
+       if (vlan) {
+               batadv_meshif_destroy_vlan(bat_priv, vlan);
+               batadv_meshif_vlan_put(vlan);
+       }
+       unregister_netdevice_queue(mesh_iface, head);
+ }
+ /**
+  * batadv_meshif_is_valid() - Check whether device is a batadv mesh interface
+  * @net_dev: device which should be checked
+  *
+  * Return: true when net_dev is a batman-adv interface, false otherwise
+  */
+ bool batadv_meshif_is_valid(const struct net_device *net_dev)
+ {
+       if (net_dev->netdev_ops->ndo_start_xmit == batadv_interface_tx)
+               return true;
+       return false;
+ }
+ static const struct nla_policy batadv_ifla_policy[IFLA_BATADV_MAX + 1] = {
+       [IFLA_BATADV_ALGO_NAME] = { .type = NLA_NUL_STRING },
+ };
+ struct rtnl_link_ops batadv_link_ops __read_mostly = {
+       .kind           = "batadv",
+       .priv_size      = sizeof(struct batadv_priv),
+       .setup          = batadv_meshif_init_early,
+       .maxtype        = IFLA_BATADV_MAX,
+       .policy         = batadv_ifla_policy,
+       .validate       = batadv_meshif_validate,
+       .newlink        = batadv_meshif_newlink,
+       .dellink        = batadv_meshif_destroy_netlink,
+ };