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
  51         /* Reset the offload_fwd_mark because there could be a stacked
  52          * bridge above, and it should not think this bridge it doing
  53          * that bridge's work forwarding out its ports.
  54          */
  55         br_switchdev_frame_unmark(skb);
  56
  57         /* Bridge is just like any other port.  Make sure the
  58          * packet is allowed except in promisc modue when someone
  59          * may be running packet capture.
  60          */
  61         if (!(brdev->flags & IFF_PROMISC) &&
  62             !br_allowed_egress(vg, skb)) {
  63                 kfree_skb(skb);
  64                 return NET_RX_DROP;
  65         }
  66
  67         indev = skb->dev;
  68         skb->dev = brdev;
  69         skb = br_handle_vlan(br, NULL, vg, skb);
  70         if (!skb)
  71                 return NET_RX_DROP;
  72         /* update the multicast stats if the packet is IGMP/MLD */
  73         br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
  74                            BR_MCAST_DIR_TX);
  75
  76         return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
  77                        dev_net(indev), NULL, skb, indev, NULL,
  78                        br_netif_receive_skb);
  79 }
  80
  81 static void br_do_proxy_arp(struct sk_buff *skb, struct net_bridge *br,
  82                             u16 vid, struct net_bridge_port *p)
  83 {
  84         struct net_device *dev = br->dev;
  85         struct neighbour *n;
  86         struct arphdr *parp;
  87         u8 *arpptr, *sha;
  88         __be32 sip, tip;
  89
  90         BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
  91
  92         if ((dev->flags & IFF_NOARP) ||
  93             !pskb_may_pull(skb, arp_hdr_len(dev)))
  94                 return;
  95
  96         parp = arp_hdr(skb);
  97
  98         if (parp->ar_pro != htons(ETH_P_IP) ||
  99             parp->ar_op != htons(ARPOP_REQUEST) ||
 100             parp->ar_hln != dev->addr_len ||
 101             parp->ar_pln != 4)
 102                 return;
 103
 104         arpptr = (u8 *)parp + sizeof(struct arphdr);
 105         sha = arpptr;
 106         arpptr += dev->addr_len;        /* sha */
 107         memcpy(&sip, arpptr, sizeof(sip));
 108         arpptr += sizeof(sip);
 109         arpptr += dev->addr_len;        /* tha */
 110         memcpy(&tip, arpptr, sizeof(tip));
 111
 112         if (ipv4_is_loopback(tip) ||
 113             ipv4_is_multicast(tip))
 114                 return;
 115
 116         n = neigh_lookup(&arp_tbl, &tip, dev);
 117         if (n) {
 118                 struct net_bridge_fdb_entry *f;
 119
 120                 if (!(n->nud_state & NUD_VALID)) {
 121                         neigh_release(n);
 122                         return;
 123                 }
 124
 125                 f = br_fdb_find_rcu(br, n->ha, vid);
 126                 if (f && ((p->flags & BR_PROXYARP) ||
 127                           (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)))) {
 128                         arp_send(ARPOP_REPLY, ETH_P_ARP, sip, skb->dev, tip,
 129                                  sha, n->ha, sha);
 130                         BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
 131                 }
 132
 133                 neigh_release(n);
 134         }
 135 }
 136
 137 /* note: already called with rcu_read_lock */
 138 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 139 {
 140         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 141         enum br_pkt_type pkt_type = BR_PKT_UNICAST;
 142         struct net_bridge_fdb_entry *dst = NULL;
 143         struct net_bridge_mdb_entry *mdst;
 144         bool local_rcv, mcast_hit = false;
 145         const unsigned char *dest;
 146         struct net_bridge *br;
 147         u16 vid = 0;
 148
 149         if (!p || p->state == BR_STATE_DISABLED)
 150                 goto drop;
 151
 152         if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
 153                 goto out;
 154
 155         nbp_switchdev_frame_mark(p, skb);
 156
 157         /* insert into forwarding database after filtering to avoid spoofing */
 158         br = p->br;
 159         if (p->flags & BR_LEARNING)
 160                 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
 161
 162         local_rcv = !!(br->dev->flags & IFF_PROMISC);
 163         dest = eth_hdr(skb)->h_dest;
 164         if (is_multicast_ether_addr(dest)) {
 165                 /* by definition the broadcast is also a multicast address */
 166                 if (is_broadcast_ether_addr(dest)) {
 167                         pkt_type = BR_PKT_BROADCAST;
 168                         local_rcv = true;
 169                 } else {
 170                         pkt_type = BR_PKT_MULTICAST;
 171                         if (br_multicast_rcv(br, p, skb, vid))
 172                                 goto drop;
 173                 }
 174         }
 175
 176         if (p->state == BR_STATE_LEARNING)
 177                 goto drop;
 178
 179         BR_INPUT_SKB_CB(skb)->brdev = br->dev;
 180
 181         if (IS_ENABLED(CONFIG_INET) && skb->protocol == htons(ETH_P_ARP))
 182                 br_do_proxy_arp(skb, br, vid, p);
 183
 184         switch (pkt_type) {
 185         case BR_PKT_MULTICAST:
 186                 mdst = br_mdb_get(br, skb, vid);
 187                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 188                     br_multicast_querier_exists(br, eth_hdr(skb))) {
 189                         if ((mdst && mdst->mglist) ||
 190                             br_multicast_is_router(br)) {
 191                                 local_rcv = true;
 192                                 br->dev->stats.multicast++;
 193                         }
 194                         mcast_hit = true;
 195                 } else {
 196                         local_rcv = true;
 197                         br->dev->stats.multicast++;
 198                 }
 199                 break;
 200         case BR_PKT_UNICAST:
 201                 dst = br_fdb_find_rcu(br, dest, vid);
 202         default:
 203                 break;
 204         }
 205
 206         if (dst) {
 207                 unsigned long now = jiffies;
 208
 209                 if (dst->is_local)
 210                         return br_pass_frame_up(skb);
 211
 212                 if (now != dst->used)
 213                         dst->used = now;
 214                 br_forward(dst->dst, skb, local_rcv, false);
 215         } else {
 216                 if (!mcast_hit)
 217                         br_flood(br, skb, pkt_type, local_rcv, false);
 218                 else
 219                         br_multicast_flood(mdst, skb, local_rcv, false);
 220         }
 221
 222         if (local_rcv)
 223                 return br_pass_frame_up(skb);
 224
 225 out:
 226         return 0;
 227 drop:
 228         kfree_skb(skb);
 229         goto out;
 230 }
 231 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 232
 233 static void __br_handle_local_finish(struct sk_buff *skb)
 234 {
 235         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 236         u16 vid = 0;
 237
 238         /* check if vlan is allowed, to avoid spoofing */
 239         if (p->flags & BR_LEARNING && br_should_learn(p, skb, &vid))
 240                 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
 241 }
 242
 243 /* note: already called with rcu_read_lock */
 244 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 245 {
 246         __br_handle_local_finish(skb);
 247
 248         /* return 1 to signal the okfn() was called so it's ok to use the skb */
 249         return 1;
 250 }
 251
 252 /*
 253  * Return NULL if skb is handled
 254  * note: already called with rcu_read_lock
 255  */
 256 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 257 {
 258         struct net_bridge_port *p;
 259         struct sk_buff *skb = *pskb;
 260         const unsigned char *dest = eth_hdr(skb)->h_dest;
 261         br_should_route_hook_t *rhook;
 262
 263         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 264                 return RX_HANDLER_PASS;
 265
 266         if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 267                 goto drop;
 268
 269         skb = skb_share_check(skb, GFP_ATOMIC);
 270         if (!skb)
 271                 return RX_HANDLER_CONSUMED;
 272
 273         p = br_port_get_rcu(skb->dev);
 274         if (p->flags & BR_VLAN_TUNNEL) {
 275                 if (br_handle_ingress_vlan_tunnel(skb, p,
 276                                                   nbp_vlan_group_rcu(p)))
 277                         goto drop;
 278         }
 279
 280         if (unlikely(is_link_local_ether_addr(dest))) {
 281                 u16 fwd_mask = p->br->group_fwd_mask_required;
 282
 283                 /*
 284                  * See IEEE 802.1D Table 7-10 Reserved addresses
 285                  *
 286                  * Assignment                           Value
 287                  * Bridge Group Address         01-80-C2-00-00-00
 288                  * (MAC Control) 802.3          01-80-C2-00-00-01
 289                  * (Link Aggregation) 802.3     01-80-C2-00-00-02
 290                  * 802.1X PAE address           01-80-C2-00-00-03
 291                  *
 292                  * 802.1AB LLDP                 01-80-C2-00-00-0E
 293                  *
 294                  * Others reserved for future standardization
 295                  */
 296                 switch (dest[5]) {
 297                 case 0x00:      /* Bridge Group Address */
 298                         /* If STP is turned off,
 299                            then must forward to keep loop detection */
 300                         if (p->br->stp_enabled == BR_NO_STP ||
 301                             fwd_mask & (1u << dest[5]))
 302                                 goto forward;
 303                         *pskb = skb;
 304                         __br_handle_local_finish(skb);
 305                         return RX_HANDLER_PASS;
 306
 307                 case 0x01:      /* IEEE MAC (Pause) */
 308                         goto drop;
 309
 310                 case 0x0E:      /* 802.1AB LLDP */
 311                         fwd_mask |= p->br->group_fwd_mask;
 312                         if (fwd_mask & (1u << dest[5]))
 313                                 goto forward;
 314                         *pskb = skb;
 315                         __br_handle_local_finish(skb);
 316                         return RX_HANDLER_PASS;
 317
 318                 default:
 319                         /* Allow selective forwarding for most other protocols */
 320                         fwd_mask |= p->br->group_fwd_mask;
 321                         if (fwd_mask & (1u << dest[5]))
 322                                 goto forward;
 323                 }
 324
 325                 /* The else clause should be hit when nf_hook():
 326                  *   - returns < 0 (drop/error)
 327                  *   - returns = 0 (stolen/nf_queue)
 328                  * Thus return 1 from the okfn() to signal the skb is ok to pass
 329                  */
 330                 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
 331                             dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 332                             br_handle_local_finish) == 1) {
 333                         return RX_HANDLER_PASS;
 334                 } else {
 335                         return RX_HANDLER_CONSUMED;
 336                 }
 337         }
 338
 339 forward:
 340         switch (p->state) {
 341         case BR_STATE_FORWARDING:
 342                 rhook = rcu_dereference(br_should_route_hook);
 343                 if (rhook) {
 344                         if ((*rhook)(skb)) {
 345                                 *pskb = skb;
 346                                 return RX_HANDLER_PASS;
 347                         }
 348                         dest = eth_hdr(skb)->h_dest;
 349                 }
 350                 /* fall through */
 351         case BR_STATE_LEARNING:
 352                 if (ether_addr_equal(p->br->dev->dev_addr, dest))
 353                         skb->pkt_type = PACKET_HOST;
 354
 355                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 356                         dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 357                         br_handle_frame_finish);
 358                 break;
 359         default:
 360 drop:
 361                 kfree_skb(skb);
 362         }
 363         return RX_HANDLER_CONSUMED;
 364 }