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
  25 /* Hook for brouter */
  26 br_should_route_hook_t __rcu *br_should_route_hook __read_mostly;
  27 EXPORT_SYMBOL(br_should_route_hook);
  28
  29 static int
  30 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
  31 {
  32         br_drop_fake_rtable(skb);
  33         return netif_receive_skb(skb);
  34 }
  35
  36 static int br_pass_frame_up(struct sk_buff *skb)
  37 {
  38         struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
  39         struct net_bridge *br = netdev_priv(brdev);
  40         struct net_bridge_vlan_group *vg;
  41         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  42
  43         u64_stats_update_begin(&brstats->syncp);
  44         brstats->rx_packets++;
  45         brstats->rx_bytes += skb->len;
  46         u64_stats_update_end(&brstats->syncp);
  47
  48         vg = br_vlan_group_rcu(br);
  49         /* Bridge is just like any other port.  Make sure the
  50          * packet is allowed except in promisc modue when someone
  51          * may be running packet capture.
  52          */
  53         if (!(brdev->flags & IFF_PROMISC) &&
  54             !br_allowed_egress(vg, skb)) {
  55                 kfree_skb(skb);
  56                 return NET_RX_DROP;
  57         }
  58
  59         indev = skb->dev;
  60         skb->dev = brdev;
  61         skb = br_handle_vlan(br, vg, skb);
  62         if (!skb)
  63                 return NET_RX_DROP;
  64         /* update the multicast stats if the packet is IGMP/MLD */
  65         br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
  66                            BR_MCAST_DIR_TX);
  67
  68         return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
  69                        dev_net(indev), NULL, skb, indev, NULL,
  70                        br_netif_receive_skb);
  71 }
  72
  73 static void br_do_proxy_arp(struct sk_buff *skb, struct net_bridge *br,
  74                             u16 vid, struct net_bridge_port *p)
  75 {
  76         struct net_device *dev = br->dev;
  77         struct neighbour *n;
  78         struct arphdr *parp;
  79         u8 *arpptr, *sha;
  80         __be32 sip, tip;
  81
  82         BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
  83
  84         if ((dev->flags & IFF_NOARP) ||
  85             !pskb_may_pull(skb, arp_hdr_len(dev)))
  86                 return;
  87
  88         parp = arp_hdr(skb);
  89
  90         if (parp->ar_pro != htons(ETH_P_IP) ||
  91             parp->ar_op != htons(ARPOP_REQUEST) ||
  92             parp->ar_hln != dev->addr_len ||
  93             parp->ar_pln != 4)
  94                 return;
  95
  96         arpptr = (u8 *)parp + sizeof(struct arphdr);
  97         sha = arpptr;
  98         arpptr += dev->addr_len;        /* sha */
  99         memcpy(&sip, arpptr, sizeof(sip));
 100         arpptr += sizeof(sip);
 101         arpptr += dev->addr_len;        /* tha */
 102         memcpy(&tip, arpptr, sizeof(tip));
 103
 104         if (ipv4_is_loopback(tip) ||
 105             ipv4_is_multicast(tip))
 106                 return;
 107
 108         n = neigh_lookup(&arp_tbl, &tip, dev);
 109         if (n) {
 110                 struct net_bridge_fdb_entry *f;
 111
 112                 if (!(n->nud_state & NUD_VALID)) {
 113                         neigh_release(n);
 114                         return;
 115                 }
 116
 117                 f = __br_fdb_get(br, n->ha, vid);
 118                 if (f && ((p->flags & BR_PROXYARP) ||
 119                           (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)))) {
 120                         arp_send(ARPOP_REPLY, ETH_P_ARP, sip, skb->dev, tip,
 121                                  sha, n->ha, sha);
 122                         BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
 123                 }
 124
 125                 neigh_release(n);
 126         }
 127 }
 128
 129 /* note: already called with rcu_read_lock */
 130 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 131 {
 132         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 133         const unsigned char *dest = eth_hdr(skb)->h_dest;
 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         struct net_bridge *br;
 139         u16 vid = 0;
 140
 141         if (!p || p->state == BR_STATE_DISABLED)
 142                 goto drop;
 143
 144         if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
 145                 goto out;
 146
 147         nbp_switchdev_frame_mark(p, skb);
 148
 149         /* insert into forwarding database after filtering to avoid spoofing */
 150         br = p->br;
 151         if (p->flags & BR_LEARNING)
 152                 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
 153
 154         local_rcv = !!(br->dev->flags & IFF_PROMISC);
 155         if (is_multicast_ether_addr(dest)) {
 156                 /* by definition the broadcast is also a multicast address */
 157                 if (is_broadcast_ether_addr(dest)) {
 158                         pkt_type = BR_PKT_BROADCAST;
 159                         local_rcv = true;
 160                 } else {
 161                         pkt_type = BR_PKT_MULTICAST;
 162                         if (br_multicast_rcv(br, p, skb, vid))
 163                                 goto drop;
 164                 }
 165         }
 166
 167         if (p->state == BR_STATE_LEARNING)
 168                 goto drop;
 169
 170         BR_INPUT_SKB_CB(skb)->brdev = br->dev;
 171
 172         if (IS_ENABLED(CONFIG_INET) && skb->protocol == htons(ETH_P_ARP))
 173                 br_do_proxy_arp(skb, br, vid, p);
 174
 175         switch (pkt_type) {
 176         case BR_PKT_MULTICAST:
 177                 mdst = br_mdb_get(br, skb, vid);
 178                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 179                     br_multicast_querier_exists(br, eth_hdr(skb))) {
 180                         if ((mdst && mdst->mglist) ||
 181                             br_multicast_is_router(br)) {
 182                                 local_rcv = true;
 183                                 br->dev->stats.multicast++;
 184                         }
 185                         mcast_hit = true;
 186                 } else {
 187                         local_rcv = true;
 188                         br->dev->stats.multicast++;
 189                 }
 190                 break;
 191         case BR_PKT_UNICAST:
 192                 dst = __br_fdb_get(br, dest, vid);
 193         default:
 194                 break;
 195         }
 196
 197         if (dst) {
 198                 if (dst->is_local)
 199                         return br_pass_frame_up(skb);
 200
 201                 dst->used = jiffies;
 202                 br_forward(dst->dst, skb, local_rcv, false);
 203         } else {
 204                 if (!mcast_hit)
 205                         br_flood(br, skb, pkt_type, local_rcv, false);
 206                 else
 207                         br_multicast_flood(mdst, skb, local_rcv, false);
 208         }
 209
 210         if (local_rcv)
 211                 return br_pass_frame_up(skb);
 212
 213 out:
 214         return 0;
 215 drop:
 216         kfree_skb(skb);
 217         goto out;
 218 }
 219 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 220
 221 static void __br_handle_local_finish(struct sk_buff *skb)
 222 {
 223         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 224         u16 vid = 0;
 225
 226         /* check if vlan is allowed, to avoid spoofing */
 227         if (p->flags & BR_LEARNING && br_should_learn(p, skb, &vid))
 228                 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, false);
 229 }
 230
 231 /* note: already called with rcu_read_lock */
 232 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 233 {
 234         __br_handle_local_finish(skb);
 235
 236         /* return 1 to signal the okfn() was called so it's ok to use the skb */
 237         return 1;
 238 }
 239
 240 /*
 241  * Return NULL if skb is handled
 242  * note: already called with rcu_read_lock
 243  */
 244 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 245 {
 246         struct net_bridge_port *p;
 247         struct sk_buff *skb = *pskb;
 248         const unsigned char *dest = eth_hdr(skb)->h_dest;
 249         br_should_route_hook_t *rhook;
 250
 251         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 252                 return RX_HANDLER_PASS;
 253
 254         if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 255                 goto drop;
 256
 257         skb = skb_share_check(skb, GFP_ATOMIC);
 258         if (!skb)
 259                 return RX_HANDLER_CONSUMED;
 260
 261         p = br_port_get_rcu(skb->dev);
 262
 263         if (unlikely(is_link_local_ether_addr(dest))) {
 264                 u16 fwd_mask = p->br->group_fwd_mask_required;
 265
 266                 /*
 267                  * See IEEE 802.1D Table 7-10 Reserved addresses
 268                  *
 269                  * Assignment                           Value
 270                  * Bridge Group Address         01-80-C2-00-00-00
 271                  * (MAC Control) 802.3          01-80-C2-00-00-01
 272                  * (Link Aggregation) 802.3     01-80-C2-00-00-02
 273                  * 802.1X PAE address           01-80-C2-00-00-03
 274                  *
 275                  * 802.1AB LLDP                 01-80-C2-00-00-0E
 276                  *
 277                  * Others reserved for future standardization
 278                  */
 279                 switch (dest[5]) {
 280                 case 0x00:      /* Bridge Group Address */
 281                         /* If STP is turned off,
 282                            then must forward to keep loop detection */
 283                         if (p->br->stp_enabled == BR_NO_STP ||
 284                             fwd_mask & (1u << dest[5]))
 285                                 goto forward;
 286                         *pskb = skb;
 287                         __br_handle_local_finish(skb);
 288                         return RX_HANDLER_PASS;
 289
 290                 case 0x01:      /* IEEE MAC (Pause) */
 291                         goto drop;
 292
 293                 case 0x0E:      /* 802.1AB LLDP */
 294                         fwd_mask |= p->br->group_fwd_mask;
 295                         if (fwd_mask & (1u << dest[5]))
 296                                 goto forward;
 297                         *pskb = skb;
 298                         __br_handle_local_finish(skb);
 299                         return RX_HANDLER_PASS;
 300
 301                 default:
 302                         /* Allow selective forwarding for most other protocols */
 303                         fwd_mask |= p->br->group_fwd_mask;
 304                         if (fwd_mask & (1u << dest[5]))
 305                                 goto forward;
 306                 }
 307
 308                 /* The else clause should be hit when nf_hook():
 309                  *   - returns < 0 (drop/error)
 310                  *   - returns = 0 (stolen/nf_queue)
 311                  * Thus return 1 from the okfn() to signal the skb is ok to pass
 312                  */
 313                 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
 314                             dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 315                             br_handle_local_finish) == 1) {
 316                         return RX_HANDLER_PASS;
 317                 } else {
 318                         return RX_HANDLER_CONSUMED;
 319                 }
 320         }
 321
 322 forward:
 323         switch (p->state) {
 324         case BR_STATE_FORWARDING:
 325                 rhook = rcu_dereference(br_should_route_hook);
 326                 if (rhook) {
 327                         if ((*rhook)(skb)) {
 328                                 *pskb = skb;
 329                                 return RX_HANDLER_PASS;
 330                         }
 331                         dest = eth_hdr(skb)->h_dest;
 332                 }
 333                 /* fall through */
 334         case BR_STATE_LEARNING:
 335                 if (ether_addr_equal(p->br->dev->dev_addr, dest))
 336                         skb->pkt_type = PACKET_HOST;
 337
 338                 NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 339                         dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 340                         br_handle_frame_finish);
 341                 break;
 342         default:
 343 drop:
 344                 kfree_skb(skb);
 345         }
 346         return RX_HANDLER_CONSUMED;
 347 }