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         const unsigned char *dest;
  83         struct net_bridge *br;
  84         u16 vid = 0;
  85
  86         if (!p || p->state == BR_STATE_DISABLED)
  87                 goto drop;
  88
  89         if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
  90                 goto out;
  91
  92         nbp_switchdev_frame_mark(p, skb);
  93
  94         /* insert into forwarding database after filtering to avoid spoofing */
  95         br = p->br;
  96         if (p->flags & BR_LEARNING)
  97                 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
  98
  99         local_rcv = !!(br->dev->flags & IFF_PROMISC);
 100         dest = eth_hdr(skb)->h_dest;
 101         if (is_multicast_ether_addr(dest)) {
 102                 /* by definition the broadcast is also a multicast address */
 103                 if (is_broadcast_ether_addr(dest)) {
 104                         pkt_type = BR_PKT_BROADCAST;
 105                         local_rcv = true;
 106                 } else {
 107                         pkt_type = BR_PKT_MULTICAST;
 108                         if (br_multicast_rcv(br, p, skb, vid))
 109                                 goto drop;
 110                 }
 111         }
 112
 113         if (p->state == BR_STATE_LEARNING)
 114                 goto drop;
 115
 116         BR_INPUT_SKB_CB(skb)->brdev = br->dev;
 117         BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
 118
 119         if (IS_ENABLED(CONFIG_INET) &&
 120             (skb->protocol == htons(ETH_P_ARP) ||
 121              skb->protocol == htons(ETH_P_RARP))) {
 122                 br_do_proxy_suppress_arp(skb, br, vid, p);
 123         } else if (IS_ENABLED(CONFIG_IPV6) &&
 124                    skb->protocol == htons(ETH_P_IPV6) &&
 125                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
 126                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
 127                                  sizeof(struct nd_msg)) &&
 128                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
 129                         struct nd_msg *msg, _msg;
 130
 131                         msg = br_is_nd_neigh_msg(skb, &_msg);
 132                         if (msg)
 133                                 br_do_suppress_nd(skb, br, vid, p, msg);
 134         }
 135
 136         switch (pkt_type) {
 137         case BR_PKT_MULTICAST:
 138                 mdst = br_mdb_get(br, skb, vid);
 139                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 140                     br_multicast_querier_exists(br, eth_hdr(skb))) {
 141                         if ((mdst && mdst->host_joined) ||
 142                             br_multicast_is_router(br)) {
 143                                 local_rcv = true;
 144                                 br->dev->stats.multicast++;
 145                         }
 146                         mcast_hit = true;
 147                 } else {
 148                         local_rcv = true;
 149                         br->dev->stats.multicast++;
 150                 }
 151                 break;
 152         case BR_PKT_UNICAST:
 153                 dst = br_fdb_find_rcu(br, dest, vid);
 154         default:
 155                 break;
 156         }
 157
 158         if (dst) {
 159                 unsigned long now = jiffies;
 160
 161                 if (dst->is_local)
 162                         return br_pass_frame_up(skb);
 163
 164                 if (now != dst->used)
 165                         dst->used = now;
 166                 br_forward(dst->dst, skb, local_rcv, false);
 167         } else {
 168                 if (!mcast_hit)
 169                         br_flood(br, skb, pkt_type, local_rcv, false);
 170                 else
 171                         br_multicast_flood(mdst, skb, local_rcv, false);
 172         }
 173
 174         if (local_rcv)
 175                 return br_pass_frame_up(skb);
 176
 177 out:
 178         return 0;
 179 drop:
 180         kfree_skb(skb);
 181         goto out;
 182 }
 183 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 184
 185 static void __br_handle_local_finish(struct sk_buff *skb)
 186 {
 187         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 188         u16 vid = 0;
 189
 190         /* check if vlan is allowed, to avoid spoofing */
 191         if ((p->flags & BR_LEARNING) &&
 192             !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
 193             br_should_learn(p, skb, &vid))
 194                 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
 195 }
 196
 197 /* note: already called with rcu_read_lock */
 198 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 199 {
 200         __br_handle_local_finish(skb);
 201
 202         /* return 1 to signal the okfn() was called so it's ok to use the skb */
 203         return 1;
 204 }
 205
 206 /*
 207  * Return NULL if skb is handled
 208  * note: already called with rcu_read_lock
 209  */
 210 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 211 {
 212         struct net_bridge_port *p;
 213         struct sk_buff *skb = *pskb;
 214         const unsigned char *dest = eth_hdr(skb)->h_dest;
 215         br_should_route_hook_t *rhook;
 216
 217         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 218                 return RX_HANDLER_PASS;
 219
 220         if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 221                 goto drop;
 222
 223         skb = skb_share_check(skb, GFP_ATOMIC);
 224         if (!skb)
 225                 return RX_HANDLER_CONSUMED;
 226
 227         p = br_port_get_rcu(skb->dev);
 228         if (p->flags & BR_VLAN_TUNNEL) {
 229                 if (br_handle_ingress_vlan_tunnel(skb, p,
 230                                                   nbp_vlan_group_rcu(p)))
 231                         goto drop;
 232         }
 233
 234         if (unlikely(is_link_local_ether_addr(dest))) {
 235                 u16 fwd_mask = p->br->group_fwd_mask_required;
 236
 237                 /*
 238                  * See IEEE 802.1D Table 7-10 Reserved addresses
 239                  *
 240                  * Assignment                           Value
 241                  * Bridge Group Address         01-80-C2-00-00-00
 242                  * (MAC Control) 802.3          01-80-C2-00-00-01
 243                  * (Link Aggregation) 802.3     01-80-C2-00-00-02
 244                  * 802.1X PAE address           01-80-C2-00-00-03
 245                  *
 246                  * 802.1AB LLDP                 01-80-C2-00-00-0E
 247                  *
 248                  * Others reserved for future standardization
 249                  */
 250                 fwd_mask |= p->group_fwd_mask;
 251                 switch (dest[5]) {
 252                 case 0x00:      /* Bridge Group Address */
 253                         /* If STP is turned off,
 254                            then must forward to keep loop detection */
 255                         if (p->br->stp_enabled == BR_NO_STP ||
 256                             fwd_mask & (1u << dest[5]))
 257                                 goto forward;
 258                         *pskb = skb;
 259                         __br_handle_local_finish(skb);
 260                         return RX_HANDLER_PASS;
 261
 262                 case 0x01:      /* IEEE MAC (Pause) */
 263                         goto drop;
 264
 265                 case 0x0E:      /* 802.1AB LLDP */
 266                         fwd_mask |= p->br->group_fwd_mask;
 267                         if (fwd_mask & (1u << dest[5]))
 268                                 goto forward;
 269                         *pskb = skb;
 270                         __br_handle_local_finish(skb);
 271                         return RX_HANDLER_PASS;
 272
 273                 default:
 274                         /* Allow selective forwarding for most other protocols */
 275                         fwd_mask |= p->br->group_fwd_mask;
 276                         if (fwd_mask & (1u << dest[5]))
 277                                 goto forward;
 278                 }
 279
 280                 /* The else clause should be hit when nf_hook():
 281                  *   - returns < 0 (drop/error)
 282                  *   - returns = 0 (stolen/nf_queue)
 283                  * Thus return 1 from the okfn() to signal the skb is ok to pass
 284                  */
 285                 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
 286                             dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 287                             br_handle_local_finish) == 1) {
 288                         return RX_HANDLER_PASS;
 289                 } else {
 290                         return RX_HANDLER_CONSUMED;
 291                 }
 292         }
 293
 294 forward:
 295         switch (p->state) {
 296         case BR_STATE_FORWARDING:
 297                 rhook = rcu_dereference(br_should_route_hook);
 298                 if (rhook) {
 299                         if ((*rhook)(skb)) {
 300                                 *pskb = skb;
 301                                 return RX_HANDLER_PASS;
 302                         }
 303                         dest = eth_hdr(skb)->h_dest;
 304                 }
 305                 /* fall through */
 306         case BR_STATE_LEARNING:
 307                 if (ether_addr_equal(p->br->dev->dev_addr, dest))
 308                         skb->pkt_type = PACKET_HOST;
 309
 310                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 311                         dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 312                         br_handle_frame_finish);
 313                 break;
 314         default:
 315 drop:
 316                 kfree_skb(skb);
 317         }
 318         return RX_HANDLER_CONSUMED;
 319 }