net/bridge/br_input.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      Handle incoming frames
   4  *      Linux ethernet bridge
   5  *
   6  *      Authors:
   7  *      Lennert Buytenhek               <buytenh@gnu.org>
   8  */
   9
  10 #include <linux/slab.h>
  11 #include <linux/kernel.h>
  12 #include <linux/netdevice.h>
  13 #include <linux/etherdevice.h>
  14 #include <linux/netfilter_bridge.h>
  15 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
  16 #include <net/netfilter/nf_queue.h>
  17 #endif
  18 #include <linux/neighbour.h>
  19 #include <net/arp.h>
  20 #include <linux/export.h>
  21 #include <linux/rculist.h>
  22 #include "br_private.h"
  23 #include "br_private_tunnel.h"
  24
  25 static int
  26 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb)
  27 {
  28         br_drop_fake_rtable(skb);
  29         return netif_receive_skb(skb);
  30 }
  31
  32 static int br_pass_frame_up(struct sk_buff *skb)
  33 {
  34         struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev;
  35         struct net_bridge *br = netdev_priv(brdev);
  36         struct net_bridge_vlan_group *vg;
  37         struct pcpu_sw_netstats *brstats = this_cpu_ptr(br->stats);
  38
  39         u64_stats_update_begin(&brstats->syncp);
  40         brstats->rx_packets++;
  41         brstats->rx_bytes += skb->len;
  42         u64_stats_update_end(&brstats->syncp);
  43
  44         vg = br_vlan_group_rcu(br);
  45         /* Bridge is just like any other port.  Make sure the
  46          * packet is allowed except in promisc modue when someone
  47          * may be running packet capture.
  48          */
  49         if (!(brdev->flags & IFF_PROMISC) &&
  50             !br_allowed_egress(vg, skb)) {
  51                 kfree_skb(skb);
  52                 return NET_RX_DROP;
  53         }
  54
  55         indev = skb->dev;
  56         skb->dev = brdev;
  57         skb = br_handle_vlan(br, NULL, vg, skb);
  58         if (!skb)
  59                 return NET_RX_DROP;
  60         /* update the multicast stats if the packet is IGMP/MLD */
  61         br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb),
  62                            BR_MCAST_DIR_TX);
  63
  64         return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
  65                        dev_net(indev), NULL, skb, indev, NULL,
  66                        br_netif_receive_skb);
  67 }
  68
  69 /* note: already called with rcu_read_lock */
  70 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
  71 {
  72         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
  73         enum br_pkt_type pkt_type = BR_PKT_UNICAST;
  74         struct net_bridge_fdb_entry *dst = NULL;
  75         struct net_bridge_mdb_entry *mdst;
  76         bool local_rcv, mcast_hit = false;
  77         const unsigned char *dest;
  78         struct net_bridge *br;
  79         u16 vid = 0;
  80
  81         if (!p || p->state == BR_STATE_DISABLED)
  82                 goto drop;
  83
  84         if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid))
  85                 goto out;
  86
  87         nbp_switchdev_frame_mark(p, skb);
  88
  89         /* insert into forwarding database after filtering to avoid spoofing */
  90         br = p->br;
  91         if (p->flags & BR_LEARNING)
  92                 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, false);
  93
  94         local_rcv = !!(br->dev->flags & IFF_PROMISC);
  95         dest = eth_hdr(skb)->h_dest;
  96         if (is_multicast_ether_addr(dest)) {
  97                 /* by definition the broadcast is also a multicast address */
  98                 if (is_broadcast_ether_addr(dest)) {
  99                         pkt_type = BR_PKT_BROADCAST;
 100                         local_rcv = true;
 101                 } else {
 102                         pkt_type = BR_PKT_MULTICAST;
 103                         if (br_multicast_rcv(br, p, skb, vid))
 104                                 goto drop;
 105                 }
 106         }
 107
 108         if (p->state == BR_STATE_LEARNING)
 109                 goto drop;
 110
 111         BR_INPUT_SKB_CB(skb)->brdev = br->dev;
 112         BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED);
 113
 114         if (IS_ENABLED(CONFIG_INET) &&
 115             (skb->protocol == htons(ETH_P_ARP) ||
 116              skb->protocol == htons(ETH_P_RARP))) {
 117                 br_do_proxy_suppress_arp(skb, br, vid, p);
 118         } else if (IS_ENABLED(CONFIG_IPV6) &&
 119                    skb->protocol == htons(ETH_P_IPV6) &&
 120                    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
 121                    pskb_may_pull(skb, sizeof(struct ipv6hdr) +
 122                                  sizeof(struct nd_msg)) &&
 123                    ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
 124                         struct nd_msg *msg, _msg;
 125
 126                         msg = br_is_nd_neigh_msg(skb, &_msg);
 127                         if (msg)
 128                                 br_do_suppress_nd(skb, br, vid, p, msg);
 129         }
 130
 131         switch (pkt_type) {
 132         case BR_PKT_MULTICAST:
 133                 mdst = br_mdb_get(br, skb, vid);
 134                 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) &&
 135                     br_multicast_querier_exists(br, eth_hdr(skb))) {
 136                         if ((mdst && mdst->host_joined) ||
 137                             br_multicast_is_router(br)) {
 138                                 local_rcv = true;
 139                                 br->dev->stats.multicast++;
 140                         }
 141                         mcast_hit = true;
 142                 } else {
 143                         local_rcv = true;
 144                         br->dev->stats.multicast++;
 145                 }
 146                 break;
 147         case BR_PKT_UNICAST:
 148                 dst = br_fdb_find_rcu(br, dest, vid);
 149         default:
 150                 break;
 151         }
 152
 153         if (dst) {
 154                 unsigned long now = jiffies;
 155
 156                 if (dst->is_local)
 157                         return br_pass_frame_up(skb);
 158
 159                 if (now != dst->used)
 160                         dst->used = now;
 161                 br_forward(dst->dst, skb, local_rcv, false);
 162         } else {
 163                 if (!mcast_hit)
 164                         br_flood(br, skb, pkt_type, local_rcv, false);
 165                 else
 166                         br_multicast_flood(mdst, skb, local_rcv, false);
 167         }
 168
 169         if (local_rcv)
 170                 return br_pass_frame_up(skb);
 171
 172 out:
 173         return 0;
 174 drop:
 175         kfree_skb(skb);
 176         goto out;
 177 }
 178 EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 179
 180 static void __br_handle_local_finish(struct sk_buff *skb)
 181 {
 182         struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 183         u16 vid = 0;
 184
 185         /* check if vlan is allowed, to avoid spoofing */
 186         if ((p->flags & BR_LEARNING) &&
 187             !br_opt_get(p->br, BROPT_NO_LL_LEARN) &&
 188             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 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb)
 202 {
 203 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
 204         struct nf_hook_entries *e = NULL;
 205         struct nf_hook_state state;
 206         unsigned int verdict, i;
 207         struct net *net;
 208         int ret;
 209
 210         net = dev_net(skb->dev);
 211 #ifdef HAVE_JUMP_LABEL
 212         if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING]))
 213                 goto frame_finish;
 214 #endif
 215
 216         e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]);
 217         if (!e)
 218                 goto frame_finish;
 219
 220         nf_hook_state_init(&state, NF_BR_PRE_ROUTING,
 221                            NFPROTO_BRIDGE, skb->dev, NULL, NULL,
 222                            net, br_handle_frame_finish);
 223
 224         for (i = 0; i < e->num_hook_entries; i++) {
 225                 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state);
 226                 switch (verdict & NF_VERDICT_MASK) {
 227                 case NF_ACCEPT:
 228                         if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) {
 229                                 *pskb = skb;
 230                                 return RX_HANDLER_PASS;
 231                         }
 232                         break;
 233                 case NF_DROP:
 234                         kfree_skb(skb);
 235                         return RX_HANDLER_CONSUMED;
 236                 case NF_QUEUE:
 237                         ret = nf_queue(skb, &state, e, i, verdict);
 238                         if (ret == 1)
 239                                 continue;
 240                         return RX_HANDLER_CONSUMED;
 241                 default: /* STOLEN */
 242                         return RX_HANDLER_CONSUMED;
 243                 }
 244         }
 245 frame_finish:
 246         net = dev_net(skb->dev);
 247         br_handle_frame_finish(net, NULL, skb);
 248 #else
 249         br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
 250 #endif
 251         return RX_HANDLER_CONSUMED;
 252 }
 253
 254 /*
 255  * Return NULL if skb is handled
 256  * note: already called with rcu_read_lock
 257  */
 258 rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 259 {
 260         struct net_bridge_port *p;
 261         struct sk_buff *skb = *pskb;
 262         const unsigned char *dest = eth_hdr(skb)->h_dest;
 263
 264         if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 265                 return RX_HANDLER_PASS;
 266
 267         if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 268                 goto drop;
 269
 270         skb = skb_share_check(skb, GFP_ATOMIC);
 271         if (!skb)
 272                 return RX_HANDLER_CONSUMED;
 273
 274         memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
 275
 276         p = br_port_get_rcu(skb->dev);
 277         if (p->flags & BR_VLAN_TUNNEL) {
 278                 if (br_handle_ingress_vlan_tunnel(skb, p,
 279                                                   nbp_vlan_group_rcu(p)))
 280                         goto drop;
 281         }
 282
 283         if (unlikely(is_link_local_ether_addr(dest))) {
 284                 u16 fwd_mask = p->br->group_fwd_mask_required;
 285
 286                 /*
 287                  * See IEEE 802.1D Table 7-10 Reserved addresses
 288                  *
 289                  * Assignment                           Value
 290                  * Bridge Group Address         01-80-C2-00-00-00
 291                  * (MAC Control) 802.3          01-80-C2-00-00-01
 292                  * (Link Aggregation) 802.3     01-80-C2-00-00-02
 293                  * 802.1X PAE address           01-80-C2-00-00-03
 294                  *
 295                  * 802.1AB LLDP                 01-80-C2-00-00-0E
 296                  *
 297                  * Others reserved for future standardization
 298                  */
 299                 fwd_mask |= p->group_fwd_mask;
 300                 switch (dest[5]) {
 301                 case 0x00:      /* Bridge Group Address */
 302                         /* If STP is turned off,
 303                            then must forward to keep loop detection */
 304                         if (p->br->stp_enabled == BR_NO_STP ||
 305                             fwd_mask & (1u << dest[5]))
 306                                 goto forward;
 307                         *pskb = skb;
 308                         __br_handle_local_finish(skb);
 309                         return RX_HANDLER_PASS;
 310
 311                 case 0x01:      /* IEEE MAC (Pause) */
 312                         goto drop;
 313
 314                 case 0x0E:      /* 802.1AB LLDP */
 315                         fwd_mask |= p->br->group_fwd_mask;
 316                         if (fwd_mask & (1u << dest[5]))
 317                                 goto forward;
 318                         *pskb = skb;
 319                         __br_handle_local_finish(skb);
 320                         return RX_HANDLER_PASS;
 321
 322                 default:
 323                         /* Allow selective forwarding for most other protocols */
 324                         fwd_mask |= p->br->group_fwd_mask;
 325                         if (fwd_mask & (1u << dest[5]))
 326                                 goto forward;
 327                 }
 328
 329                 /* The else clause should be hit when nf_hook():
 330                  *   - returns < 0 (drop/error)
 331                  *   - returns = 0 (stolen/nf_queue)
 332                  * Thus return 1 from the okfn() to signal the skb is ok to pass
 333                  */
 334                 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN,
 335                             dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 336                             br_handle_local_finish) == 1) {
 337                         return RX_HANDLER_PASS;
 338                 } else {
 339                         return RX_HANDLER_CONSUMED;
 340                 }
 341         }
 342
 343 forward:
 344         switch (p->state) {
 345         case BR_STATE_FORWARDING:
 346         case BR_STATE_LEARNING:
 347                 if (ether_addr_equal(p->br->dev->dev_addr, dest))
 348                         skb->pkt_type = PACKET_HOST;
 349
 350                 return nf_hook_bridge_pre(skb, pskb);
 351         default:
 352 drop:
 353                 kfree_skb(skb);
 354         }
 355         return RX_HANDLER_CONSUMED;
 356 }