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 <net/dsa.h> 21 #include <linux/export.h> 22 #include <linux/rculist.h> 23 #include "br_private.h" 24 #include "br_private_tunnel.h" 25 26 static int 27 br_netif_receive_skb(struct net *net, struct sock *sk, struct sk_buff *skb) 28 { 29 br_drop_fake_rtable(skb); 30 return netif_receive_skb(skb); 31 } 32 33 static int br_pass_frame_up(struct sk_buff *skb) 34 { 35 struct net_device *indev, *brdev = BR_INPUT_SKB_CB(skb)->brdev; 36 struct net_bridge *br = netdev_priv(brdev); 37 struct net_bridge_vlan_group *vg; 38 39 dev_sw_netstats_rx_add(brdev, skb->len); 40 41 vg = br_vlan_group_rcu(br); 42 /* Bridge is just like any other port. Make sure the 43 * packet is allowed except in promisc mode when someone 44 * may be running packet capture. 45 */ 46 if (!(brdev->flags & IFF_PROMISC) && 47 !br_allowed_egress(vg, skb)) { 48 kfree_skb(skb); 49 return NET_RX_DROP; 50 } 51 52 indev = skb->dev; 53 skb->dev = brdev; 54 skb = br_handle_vlan(br, NULL, vg, skb); 55 if (!skb) 56 return NET_RX_DROP; 57 /* update the multicast stats if the packet is IGMP/MLD */ 58 br_multicast_count(br, NULL, skb, br_multicast_igmp_type(skb), 59 BR_MCAST_DIR_TX); 60 61 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, 62 dev_net(indev), NULL, skb, indev, NULL, 63 br_netif_receive_skb); 64 } 65 66 /* note: already called with rcu_read_lock */ 67 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 68 { 69 struct net_bridge_port *p = br_port_get_rcu(skb->dev); 70 enum br_pkt_type pkt_type = BR_PKT_UNICAST; 71 struct net_bridge_fdb_entry *dst = NULL; 72 struct net_bridge_mcast_port *pmctx; 73 struct net_bridge_mdb_entry *mdst; 74 bool local_rcv, mcast_hit = false; 75 struct net_bridge_mcast *brmctx; 76 struct net_bridge_vlan *vlan; 77 struct net_bridge *br; 78 u16 vid = 0; 79 u8 state; 80 81 if (!p || p->state == BR_STATE_DISABLED) 82 goto drop; 83 84 br = p->br; 85 brmctx = &p->br->multicast_ctx; 86 pmctx = &p->multicast_ctx; 87 state = p->state; 88 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &vid, 89 &state, &vlan)) 90 goto out; 91 92 if (p->flags & BR_PORT_LOCKED) { 93 struct net_bridge_fdb_entry *fdb_src = 94 br_fdb_find_rcu(br, eth_hdr(skb)->h_source, vid); 95 96 if (!fdb_src || READ_ONCE(fdb_src->dst) != p || 97 test_bit(BR_FDB_LOCAL, &fdb_src->flags)) 98 goto drop; 99 } 100 101 nbp_switchdev_frame_mark(p, skb); 102 103 /* insert into forwarding database after filtering to avoid spoofing */ 104 if (p->flags & BR_LEARNING) 105 br_fdb_update(br, p, eth_hdr(skb)->h_source, vid, 0); 106 107 local_rcv = !!(br->dev->flags & IFF_PROMISC); 108 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) { 109 /* by definition the broadcast is also a multicast address */ 110 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) { 111 pkt_type = BR_PKT_BROADCAST; 112 local_rcv = true; 113 } else { 114 pkt_type = BR_PKT_MULTICAST; 115 if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid)) 116 goto drop; 117 } 118 } 119 120 if (state == BR_STATE_LEARNING) 121 goto drop; 122 123 BR_INPUT_SKB_CB(skb)->brdev = br->dev; 124 BR_INPUT_SKB_CB(skb)->src_port_isolated = !!(p->flags & BR_ISOLATED); 125 126 if (IS_ENABLED(CONFIG_INET) && 127 (skb->protocol == htons(ETH_P_ARP) || 128 skb->protocol == htons(ETH_P_RARP))) { 129 br_do_proxy_suppress_arp(skb, br, vid, p); 130 } else if (IS_ENABLED(CONFIG_IPV6) && 131 skb->protocol == htons(ETH_P_IPV6) && 132 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 133 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 134 sizeof(struct nd_msg)) && 135 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 136 struct nd_msg *msg, _msg; 137 138 msg = br_is_nd_neigh_msg(skb, &_msg); 139 if (msg) 140 br_do_suppress_nd(skb, br, vid, p, msg); 141 } 142 143 switch (pkt_type) { 144 case BR_PKT_MULTICAST: 145 mdst = br_mdb_get(brmctx, skb, vid); 146 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 147 br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst)) { 148 if ((mdst && mdst->host_joined) || 149 br_multicast_is_router(brmctx, skb)) { 150 local_rcv = true; 151 br->dev->stats.multicast++; 152 } 153 mcast_hit = true; 154 } else { 155 local_rcv = true; 156 br->dev->stats.multicast++; 157 } 158 break; 159 case BR_PKT_UNICAST: 160 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid); 161 break; 162 default: 163 break; 164 } 165 166 if (dst) { 167 unsigned long now = jiffies; 168 169 if (test_bit(BR_FDB_LOCAL, &dst->flags)) 170 return br_pass_frame_up(skb); 171 172 if (now != dst->used) 173 dst->used = now; 174 br_forward(dst->dst, skb, local_rcv, false); 175 } else { 176 if (!mcast_hit) 177 br_flood(br, skb, pkt_type, local_rcv, false); 178 else 179 br_multicast_flood(mdst, skb, brmctx, local_rcv, false); 180 } 181 182 if (local_rcv) 183 return br_pass_frame_up(skb); 184 185 out: 186 return 0; 187 drop: 188 kfree_skb(skb); 189 goto out; 190 } 191 EXPORT_SYMBOL_GPL(br_handle_frame_finish); 192 193 static void __br_handle_local_finish(struct sk_buff *skb) 194 { 195 struct net_bridge_port *p = br_port_get_rcu(skb->dev); 196 u16 vid = 0; 197 198 /* check if vlan is allowed, to avoid spoofing */ 199 if ((p->flags & BR_LEARNING) && 200 nbp_state_should_learn(p) && 201 !br_opt_get(p->br, BROPT_NO_LL_LEARN) && 202 br_should_learn(p, skb, &vid)) 203 br_fdb_update(p->br, p, eth_hdr(skb)->h_source, vid, 0); 204 } 205 206 /* note: already called with rcu_read_lock */ 207 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 208 { 209 __br_handle_local_finish(skb); 210 211 /* return 1 to signal the okfn() was called so it's ok to use the skb */ 212 return 1; 213 } 214 215 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb) 216 { 217 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE 218 struct nf_hook_entries *e = NULL; 219 struct nf_hook_state state; 220 unsigned int verdict, i; 221 struct net *net; 222 int ret; 223 224 net = dev_net(skb->dev); 225 #ifdef HAVE_JUMP_LABEL 226 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING])) 227 goto frame_finish; 228 #endif 229 230 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]); 231 if (!e) 232 goto frame_finish; 233 234 nf_hook_state_init(&state, NF_BR_PRE_ROUTING, 235 NFPROTO_BRIDGE, skb->dev, NULL, NULL, 236 net, br_handle_frame_finish); 237 238 for (i = 0; i < e->num_hook_entries; i++) { 239 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state); 240 switch (verdict & NF_VERDICT_MASK) { 241 case NF_ACCEPT: 242 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) { 243 *pskb = skb; 244 return RX_HANDLER_PASS; 245 } 246 break; 247 case NF_DROP: 248 kfree_skb(skb); 249 return RX_HANDLER_CONSUMED; 250 case NF_QUEUE: 251 ret = nf_queue(skb, &state, i, verdict); 252 if (ret == 1) 253 continue; 254 return RX_HANDLER_CONSUMED; 255 default: /* STOLEN */ 256 return RX_HANDLER_CONSUMED; 257 } 258 } 259 frame_finish: 260 net = dev_net(skb->dev); 261 br_handle_frame_finish(net, NULL, skb); 262 #else 263 br_handle_frame_finish(dev_net(skb->dev), NULL, skb); 264 #endif 265 return RX_HANDLER_CONSUMED; 266 } 267 268 /* Return 0 if the frame was not processed otherwise 1 269 * note: already called with rcu_read_lock 270 */ 271 static int br_process_frame_type(struct net_bridge_port *p, 272 struct sk_buff *skb) 273 { 274 struct br_frame_type *tmp; 275 276 hlist_for_each_entry_rcu(tmp, &p->br->frame_type_list, list) 277 if (unlikely(tmp->type == skb->protocol)) 278 return tmp->frame_handler(p, skb); 279 280 return 0; 281 } 282 283 /* 284 * Return NULL if skb is handled 285 * note: already called with rcu_read_lock 286 */ 287 static rx_handler_result_t br_handle_frame(struct sk_buff **pskb) 288 { 289 struct net_bridge_port *p; 290 struct sk_buff *skb = *pskb; 291 const unsigned char *dest = eth_hdr(skb)->h_dest; 292 293 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 294 return RX_HANDLER_PASS; 295 296 if (!is_valid_ether_addr(eth_hdr(skb)->h_source)) 297 goto drop; 298 299 skb = skb_share_check(skb, GFP_ATOMIC); 300 if (!skb) 301 return RX_HANDLER_CONSUMED; 302 303 memset(skb->cb, 0, sizeof(struct br_input_skb_cb)); 304 305 p = br_port_get_rcu(skb->dev); 306 if (p->flags & BR_VLAN_TUNNEL) 307 br_handle_ingress_vlan_tunnel(skb, p, nbp_vlan_group_rcu(p)); 308 309 if (unlikely(is_link_local_ether_addr(dest))) { 310 u16 fwd_mask = p->br->group_fwd_mask_required; 311 312 /* 313 * See IEEE 802.1D Table 7-10 Reserved addresses 314 * 315 * Assignment Value 316 * Bridge Group Address 01-80-C2-00-00-00 317 * (MAC Control) 802.3 01-80-C2-00-00-01 318 * (Link Aggregation) 802.3 01-80-C2-00-00-02 319 * 802.1X PAE address 01-80-C2-00-00-03 320 * 321 * 802.1AB LLDP 01-80-C2-00-00-0E 322 * 323 * Others reserved for future standardization 324 */ 325 fwd_mask |= p->group_fwd_mask; 326 switch (dest[5]) { 327 case 0x00: /* Bridge Group Address */ 328 /* If STP is turned off, 329 then must forward to keep loop detection */ 330 if (p->br->stp_enabled == BR_NO_STP || 331 fwd_mask & (1u << dest[5])) 332 goto forward; 333 *pskb = skb; 334 __br_handle_local_finish(skb); 335 return RX_HANDLER_PASS; 336 337 case 0x01: /* IEEE MAC (Pause) */ 338 goto drop; 339 340 case 0x0E: /* 802.1AB LLDP */ 341 fwd_mask |= p->br->group_fwd_mask; 342 if (fwd_mask & (1u << dest[5])) 343 goto forward; 344 *pskb = skb; 345 __br_handle_local_finish(skb); 346 return RX_HANDLER_PASS; 347 348 default: 349 /* Allow selective forwarding for most other protocols */ 350 fwd_mask |= p->br->group_fwd_mask; 351 if (fwd_mask & (1u << dest[5])) 352 goto forward; 353 } 354 355 /* The else clause should be hit when nf_hook(): 356 * - returns < 0 (drop/error) 357 * - returns = 0 (stolen/nf_queue) 358 * Thus return 1 from the okfn() to signal the skb is ok to pass 359 */ 360 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, 361 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 362 br_handle_local_finish) == 1) { 363 return RX_HANDLER_PASS; 364 } else { 365 return RX_HANDLER_CONSUMED; 366 } 367 } 368 369 if (unlikely(br_process_frame_type(p, skb))) 370 return RX_HANDLER_PASS; 371 372 forward: 373 switch (p->state) { 374 case BR_STATE_FORWARDING: 375 case BR_STATE_LEARNING: 376 if (ether_addr_equal(p->br->dev->dev_addr, dest)) 377 skb->pkt_type = PACKET_HOST; 378 379 return nf_hook_bridge_pre(skb, pskb); 380 default: 381 drop: 382 kfree_skb(skb); 383 } 384 return RX_HANDLER_CONSUMED; 385 } 386 387 /* This function has no purpose other than to appease the br_port_get_rcu/rtnl 388 * helpers which identify bridged ports according to the rx_handler installed 389 * on them (so there _needs_ to be a bridge rx_handler even if we don't need it 390 * to do anything useful). This bridge won't support traffic to/from the stack, 391 * but only hardware bridging. So return RX_HANDLER_PASS so we don't steal 392 * frames from the ETH_P_XDSA packet_type handler. 393 */ 394 static rx_handler_result_t br_handle_frame_dummy(struct sk_buff **pskb) 395 { 396 return RX_HANDLER_PASS; 397 } 398 399 rx_handler_func_t *br_get_rx_handler(const struct net_device *dev) 400 { 401 if (netdev_uses_dsa(dev)) 402 return br_handle_frame_dummy; 403 404 return br_handle_frame; 405 } 406 407 void br_add_frame(struct net_bridge *br, struct br_frame_type *ft) 408 { 409 hlist_add_head_rcu(&ft->list, &br->frame_type_list); 410 } 411 412 void br_del_frame(struct net_bridge *br, struct br_frame_type *ft) 413 { 414 struct br_frame_type *tmp; 415 416 hlist_for_each_entry(tmp, &br->frame_type_list, list) 417 if (ft == tmp) { 418 hlist_del_rcu(&ft->list); 419 return; 420 } 421 } 422