net/bridge/br_input.c

   1 /*
   2  *      Handle incoming frames
   3  *      Linux ethernet bridge
   4  *
   5  *      Authors:
   6  *      Lennert Buytenhek               <buytenh@gnu.org>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 #include <linux/slab.h>
  15 #include <linux/kernel.h>
  16 #include <linux/netdevice.h>
  17 #include <linux/etherdevice.h>
  18 #include <linux/netfilter_bridge.h>
  19 #include <linux/neighbour.h>
  20 #include <net/arp.h>
  21 #include <linux/export.h>
  22 #include <linux/rculist.h>
  23 #include "br_private.h"
  24 #include "br_private_tunnel.h"
  25
  26 /* Hook for brouter */
  27 br_should_route_hook_t __rcu *br_should_route_hook __read_mostly;
  28 EXPORT_SYMBOL(br_should_route_hook);
  29
  30 static int
  31 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
  32 {
  33         br_drop_fake_rtable(skb);
  34         return netif_receive_skb(skb);
  35 }
  36
  37 static int br_pass_frame_up(struct sk_buff *skb)
  38 {
  39         struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
  40         struct net_bridge *br = netdev_priv(brdev);
  41         struct net_bridge_vlan_group *vg;
  42         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  43
  44         u64_stats_update_begin(&brstats->syncp);
  45         brstats->rx_packets++;
  46         brstats->rx_bytes += skb->len;
  47         u64_stats_update_end(&brstats->syncp);
  48
  49         vg = br_vlan_group_rcu(br);
  50         /* Bridge is just like any other port.  Make sure the
  51          * packet is allowed except in promisc modue when someone
  52          * may be running packet capture.
  53          */
  54         if (!(brdev->flags & IFF_PROMISC) &&
  55             !br_allowed_egress(vg, skb)) {
  56                 kfree_skb(skb);
  57                 return NET_RX_DROP;
  58         }
  59
  60         indev = skb->dev;
  61         skb->dev = brdev;
  62         skb = br_handle_vlan(br, NULL, vg, skb);
  63         if (!skb)
  64                 return NET_RX_DROP;
  65         /* update the multicast stats if the packet is IGMP/MLD */
  66         br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
  67                            BR_MCAST_DIR_TX);
  68
  69         return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
  70                        dev_net(indev), NULL, skb, indev, NULL,
  71                        br_netif_receive_skb);
  72 }
  73
  74 /* note: already called with rcu_read_lock */
  75 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
  76 {
  77         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
  78         enum br_pkt_type pkt_type = BR_PKT_UNICAST;
  79         struct net_bridge_fdb_entry *dst = NULL;
  80         struct net_bridge_mdb_entry *mdst;
  81         bool local_rcv, mcast_hit = false;
  82         struct net_bridge *br;
  83         u16 vid = 0;
  84
  85         if (!p || p->state == BR_STATE_DISABLED)
  86                 goto drop;
  87
  88         if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
  89                 goto out;
  90
  91         nbp_switchdev_frame_mark(p, skb);
  92
  93         /* insert into forwarding database after filtering to avoid spoofing */
  94         br = p->br;
  95         if (p->flags & BR_LEARNING)
  96                 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
  97
  98         local_rcv = !!(br->dev->flags & IFF_PROMISC);
  99         if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) {
 100                 /* by definition the broadcast is also a multicast address */
 101                 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) {
 102                         pkt_type = BR_PKT_BROADCAST;
 103                         local_rcv = true;
 104                 } else {
 105                         pkt_type = BR_PKT_MULTICAST;
 106                         if (br_multicast_rcv(br, p, skb, vid))
 107                                 goto drop;
 108                 }
 109         }
 110
 111         if (p->state == BR_STATE_LEARNING)
 112                 goto drop;
 113
 114         BR_INPUT_SKB_CB(skb)->brdev = br->dev;
 115         BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
 116
 117         if (IS_ENABLED(CONFIG_INET) &&
 118             (skb->protocol == htons(ETH_P_ARP) ||
 119              skb->protocol == htons(ETH_P_RARP))) {
 120                 br_do_proxy_suppress_arp(skb, br, vid, p);
 121         } else if (IS_ENABLED(CONFIG_IPV6) &&
 122                    skb->protocol == htons(ETH_P_IPV6) &&
 123                    br->neigh_suppress_enabled &&
 124                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
 125                                  sizeof(struct nd_msg)) &&
 126                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
 127                         struct nd_msg *msg, _msg;
 128
 129                         msg = br_is_nd_neigh_msg(skb, &_msg);
 130                         if (msg)
 131                                 br_do_suppress_nd(skb, br, vid, p, msg);
 132         }
 133
 134         switch (pkt_type) {
 135         case BR_PKT_MULTICAST:
 136                 mdst = br_mdb_get(br, skb, vid);
 137                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 138                     br_multicast_querier_exists(br, eth_hdr(skb))) {
 139                         if ((mdst && mdst->host_joined) ||
 140                             br_multicast_is_router(br)) {
 141                                 local_rcv = true;
 142                                 br->dev->stats.multicast++;
 143                         }
 144                         mcast_hit = true;
 145                 } else {
 146                         local_rcv = true;
 147                         br->dev->stats.multicast++;
 148                 }
 149                 break;
 150         case BR_PKT_UNICAST:
 151                 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid);
 152         default:
 153                 break;
 154         }
 155
 156         if (dst) {
 157                 unsigned long now = jiffies;
 158
 159                 if (dst->is_local)
 160                         return br_pass_frame_up(skb);
 161
 162                 if (now != dst->used)
 163                         dst->used = now;
 164                 br_forward(dst->dst, skb, local_rcv, false);
 165         } else {
 166                 if (!mcast_hit)
 167                         br_flood(br, skb, pkt_type, local_rcv, false);
 168                 else
 169                         br_multicast_flood(mdst, skb, local_rcv, false);
 170         }
 171
 172         if (local_rcv)
 173                 return br_pass_frame_up(skb);
 174
 175 out:
 176         return 0;
 177 drop:
 178         kfree_skb(skb);
 179         goto out;
 180 }
 181 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 182
 183 static void __br_handle_local_finish(struct sk_buff *skb)
 184 {
 185         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 186         u16 vid = 0;
 187
 188         /* check if vlan is allowed, to avoid spoofing */
 189         if (p->flags & BR_LEARNING && br_should_learn(p, skb, &vid))
 190                 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
 191 }
 192
 193 /* note: already called with rcu_read_lock */
 194 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 195 {
 196         __br_handle_local_finish(skb);
 197
 198         /* return 1 to signal the okfn() was called so it's ok to use the skb */
 199         return 1;
 200 }
 201
 202 /*
 203  * Return NULL if skb is handled
 204  * note: already called with rcu_read_lock
 205  */
 206 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 207 {
 208         struct net_bridge_port *p;
 209         struct sk_buff *skb = *pskb;
 210         const unsigned char *dest = eth_hdr(skb)->h_dest;
 211         br_should_route_hook_t *rhook;
 212
 213         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 214                 return RX_HANDLER_PASS;
 215
 216         if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 217                 goto drop;
 218
 219         skb = skb_share_check(skb, GFP_ATOMIC);
 220         if (!skb)
 221                 return RX_HANDLER_CONSUMED;
 222
 223         p = br_port_get_rcu(skb->dev);
 224         if (p->flags & BR_VLAN_TUNNEL) {
 225                 if (br_handle_ingress_vlan_tunnel(skb, p,
 226                                                   nbp_vlan_group_rcu(p)))
 227                         goto drop;
 228         }
 229
 230         if (unlikely(is_link_local_ether_addr(dest))) {
 231                 u16 fwd_mask = p->br->group_fwd_mask_required;
 232
 233                 /*
 234                  * See IEEE 802.1D Table 7-10 Reserved addresses
 235                  *
 236                  * Assignment                           Value
 237                  * Bridge Group Address         01-80-C2-00-00-00
 238                  * (MAC Control) 802.3          01-80-C2-00-00-01
 239                  * (Link Aggregation) 802.3     01-80-C2-00-00-02
 240                  * 802.1X PAE address           01-80-C2-00-00-03
 241                  *
 242                  * 802.1AB LLDP                 01-80-C2-00-00-0E
 243                  *
 244                  * Others reserved for future standardization
 245                  */
 246                 fwd_mask |= p->group_fwd_mask;
 247                 switch (dest[5]) {
 248                 case 0x00:      /* Bridge Group Address */
 249                         /* If STP is turned off,
 250                            then must forward to keep loop detection */
 251                         if (p->br->stp_enabled == BR_NO_STP ||
 252                             fwd_mask & (1u << dest[5]))
 253                                 goto forward;
 254                         *pskb = skb;
 255                         __br_handle_local_finish(skb);
 256                         return RX_HANDLER_PASS;
 257
 258                 case 0x01:      /* IEEE MAC (Pause) */
 259                         goto drop;
 260
 261                 case 0x0E:      /* 802.1AB LLDP */
 262                         fwd_mask |= p->br->group_fwd_mask;
 263                         if (fwd_mask & (1u << dest[5]))
 264                                 goto forward;
 265                         *pskb = skb;
 266                         __br_handle_local_finish(skb);
 267                         return RX_HANDLER_PASS;
 268
 269                 default:
 270                         /* Allow selective forwarding for most other protocols */
 271                         fwd_mask |= p->br->group_fwd_mask;
 272                         if (fwd_mask & (1u << dest[5]))
 273                                 goto forward;
 274                 }
 275
 276                 /* The else clause should be hit when nf_hook():
 277                  *   - returns < 0 (drop/error)
 278                  *   - returns = 0 (stolen/nf_queue)
 279                  * Thus return 1 from the okfn() to signal the skb is ok to pass
 280                  */
 281                 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
 282                             dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 283                             br_handle_local_finish) == 1) {
 284                         return RX_HANDLER_PASS;
 285                 } else {
 286                         return RX_HANDLER_CONSUMED;
 287                 }
 288         }
 289
 290 forward:
 291         switch (p->state) {
 292         case BR_STATE_FORWARDING:
 293                 rhook = rcu_dereference(br_should_route_hook);
 294                 if (rhook) {
 295                         if ((*rhook)(skb)) {
 296                                 *pskb = skb;
 297                                 return RX_HANDLER_PASS;
 298                         }
 299                         dest = eth_hdr(skb)->h_dest;
 300                 }
 301                 /* fall through */
 302         case BR_STATE_LEARNING:
 303                 if (ether_addr_equal(p->br->dev->dev_addr, dest))
 304                         skb->pkt_type = PACKET_HOST;
 305
 306                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 307                         dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 308                         br_handle_frame_finish);
 309                 break;
 310         default:
 311 drop:
 312                 kfree_skb(skb);
 313         }
 314         return RX_HANDLER_CONSUMED;
 315 }