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