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