LXR linux/net/bridge/br

   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 */
  27br_should_route_hook_t __rcu *br_should_route_hook __read_mostly;
  28EXPORT_SYMBOL(br_should_route_hook);
  29
  30static int
  31br_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
  37static 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
  74static 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 */
 131int 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
 218out:
 219        return 0;
 220drop:
 221        kfree_skb(skb);
 222        goto out;
 223}
 224EXPORT_SYMBOL_GPL(br_handle_frame_finish);
 225
 226static 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 */
 237static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 238{
 239        struct net_bridge_port *p = br_port_get_rcu(skb->dev);
 240
 241        __br_handle_local_finish(skb);
 242
 243        BR_INPUT_SKB_CB(skb)->brdev = p->br->dev;
 244        br_pass_frame_up(skb);
 245        return 0;
 246}
 247
 248/*
 249 * Return NULL if skb is handled
 250 * note: already called with rcu_read_lock
 251 */
 252rx_handler_result_t br_handle_frame(struct sk_buff **pskb)
 253{
 254        struct net_bridge_port *p;
 255        struct sk_buff *skb = *pskb;
 256        const unsigned char *dest = eth_hdr(skb)->h_dest;
 257        br_should_route_hook_t *rhook;
 258
 259        if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 260                return RX_HANDLER_PASS;
 261
 262        if (!is_valid_ether_addr(eth_hdr(skb)->h_source))
 263                goto drop;
 264
 265        skb = skb_share_check(skb, GFP_ATOMIC);
 266        if (!skb)
 267                return RX_HANDLER_CONSUMED;
 268
 269        p = br_port_get_rcu(skb->dev);
 270        if (p->flags & BR_VLAN_TUNNEL) {
 271                if (br_handle_ingress_vlan_tunnel(skb, p,
 272                                                  nbp_vlan_group_rcu(p)))
 273                        goto drop;
 274        }
 275
 276        if (unlikely(is_link_local_ether_addr(dest))) {
 277                u16 fwd_mask = p->br->group_fwd_mask_required;
 278
 279                /*
 280                 * See IEEE 802.1D Table 7-10 Reserved addresses
 281                 *
 282                 * Assignment                           Value
 283                 * Bridge Group Address         01-80-C2-00-00-00
 284                 * (MAC Control) 802.3          01-80-C2-00-00-01
 285                 * (Link Aggregation) 802.3     01-80-C2-00-00-02
 286                 * 802.1X PAE address           01-80-C2-00-00-03
 287                 *
 288                 * 802.1AB LLDP                 01-80-C2-00-00-0E
 289                 *
 290                 * Others reserved for future standardization
 291                 */
 292                switch (dest[5]) {
 293                case 0x00:      /* Bridge Group Address */
 294                        /* If STP is turned off,
 295                           then must forward to keep loop detection */
 296                        if (p->br->stp_enabled == BR_NO_STP ||
 297                            fwd_mask & (1u << dest[5]))
 298                                goto forward;
 299                        *pskb = skb;
 300                        __br_handle_local_finish(skb);
 301                        return RX_HANDLER_PASS;
 302
 303                case 0x01:      /* IEEE MAC (Pause) */
 304                        goto drop;
 305
 306                case 0x0E:      /* 802.1AB LLDP */
 307                        fwd_mask |= p->br->group_fwd_mask;
 308                        if (fwd_mask & (1u << dest[5]))
 309                                goto forward;
 310                        *pskb = skb;
 311                        __br_handle_local_finish(skb);
 312                        return RX_HANDLER_PASS;
 313
 314                default:
 315                        /* Allow selective forwarding for most other protocols */
 316                        fwd_mask |= p->br->group_fwd_mask;
 317                        if (fwd_mask & (1u << dest[5]))
 318                                goto forward;
 319                }
 320
 321                /* Deliver packet to local host only */
 322                NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, dev_net(skb->dev),
 323                        NULL, skb, skb->dev, NULL, br_handle_local_finish);
 324                return RX_HANDLER_CONSUMED;
 325        }
 326
 327forward:
 328        switch (p->state) {
 329        case BR_STATE_FORWARDING:
 330                rhook = rcu_dereference(br_should_route_hook);
 331                if (rhook) {
 332                        if ((*rhook)(skb)) {
 333                                *pskb = skb;
 334                                return RX_HANDLER_PASS;
 335                        }
 336                        dest = eth_hdr(skb)->h_dest;
 337                }
 338                /* fall through */
 339        case BR_STATE_LEARNING:
 340                if (ether_addr_equal(p->br->dev->dev_addr, dest))
 341                        skb->pkt_type = PACKET_HOST;
 342
 343                NF_HOOK(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 344                        dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 345                        br_handle_frame_finish);
 346                break;
 347        default:
 348drop:
 349                kfree_skb(skb);
 350        }
 351        return RX_HANDLER_CONSUMED;
 352}
 353