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, bool promisc) 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 43 /* Reset the offload_fwd_mark because there could be a stacked 44 * bridge above, and it should not think this bridge it doing 45 * that bridge's work forwarding out its ports. 46 */ 47 br_switchdev_frame_unmark(skb); 48 49 /* Bridge is just like any other port. Make sure the 50 * packet is allowed except in promisc mode 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, NULL, vg, NULL, 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 BR_INPUT_SKB_CB(skb)->promisc = promisc; 69 70 return NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, 71 dev_net(indev), NULL, skb, indev, NULL, 72 br_netif_receive_skb); 73 } 74 75 /* note: already called with rcu_read_lock */ 76 int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 77 { 78 enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED; 79 struct net_bridge_port *p = br_port_get_rcu(skb->dev); 80 enum br_pkt_type pkt_type = BR_PKT_UNICAST; 81 struct net_bridge_fdb_entry *dst = NULL; 82 struct net_bridge_mcast_port *pmctx; 83 struct net_bridge_mdb_entry *mdst; 84 bool local_rcv, mcast_hit = false; 85 struct net_bridge_mcast *brmctx; 86 struct net_bridge_vlan *vlan; 87 struct net_bridge *br; 88 bool promisc; 89 u8 state; 90 u16 vid; 91 92 if (!p) 93 goto drop; 94 95 br = p->br; 96 97 if (br_mst_is_enabled(p)) { 98 state = BR_STATE_FORWARDING; 99 } else { 100 if (p->state == BR_STATE_DISABLED) { 101 reason = SKB_DROP_REASON_BRIDGE_INGRESS_STP_STATE; 102 goto drop; 103 } 104 105 state = p->state; 106 } 107 108 brmctx = &p->br->multicast_ctx; 109 pmctx = &p->multicast_ctx; 110 if (!br_allowed_ingress(p->br, nbp_vlan_group_rcu(p), skb, &state, 111 &vlan)) 112 goto out; 113 vid = vlan ? vlan->vid : 0; 114 115 if (test_bit(BR_PORT_LOCKED_BIT, &p->flags)) { 116 struct net_bridge_fdb_entry *fdb_src = 117 br_fdb_find_rcu(br, eth_hdr(skb)->h_source, vid); 118 119 if (!fdb_src) { 120 /* FDB miss. Create locked FDB entry if MAB is enabled 121 * and drop the packet. 122 */ 123 if (test_bit(BR_PORT_MAB_BIT, &p->flags)) 124 br_fdb_update(br, p, vlan, 125 eth_hdr(skb)->h_source, 126 BIT(BR_FDB_LOCKED)); 127 goto drop; 128 } else if (br_fdb_dst_port(fdb_src) != p || 129 test_bit(BR_FDB_LOCAL, &fdb_src->flags)) { 130 /* FDB mismatch. Drop the packet without roaming. */ 131 goto drop; 132 } else if (test_bit(BR_FDB_LOCKED, &fdb_src->flags)) { 133 /* FDB match, but entry is locked. Refresh it and drop 134 * the packet. 135 */ 136 br_fdb_update(br, p, vlan, eth_hdr(skb)->h_source, 137 BIT(BR_FDB_LOCKED)); 138 goto drop; 139 } 140 } 141 142 nbp_switchdev_frame_mark(p, skb); 143 144 /* insert into forwarding database after filtering to avoid spoofing */ 145 if (test_bit(BR_LEARNING_BIT, &p->flags)) 146 br_fdb_update(br, p, vlan, eth_hdr(skb)->h_source, 0); 147 148 promisc = !!(br->dev->flags & IFF_PROMISC); 149 local_rcv = promisc; 150 151 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) { 152 /* by definition the broadcast is also a multicast address */ 153 if (is_broadcast_ether_addr(eth_hdr(skb)->h_dest)) { 154 pkt_type = BR_PKT_BROADCAST; 155 local_rcv = true; 156 } else { 157 pkt_type = BR_PKT_MULTICAST; 158 if (br_multicast_rcv(&brmctx, &pmctx, vlan, skb, vid)) 159 goto drop; 160 } 161 } 162 163 if (state == BR_STATE_LEARNING) { 164 reason = SKB_DROP_REASON_BRIDGE_INGRESS_STP_STATE; 165 goto drop; 166 } 167 168 BR_INPUT_SKB_CB(skb)->brdev = br->dev; 169 BR_INPUT_SKB_CB(skb)->src_port_isolated = test_bit(BR_ISOLATED_BIT, &p->flags); 170 171 if (IS_ENABLED(CONFIG_INET) && 172 (skb->protocol == htons(ETH_P_ARP) || 173 skb->protocol == htons(ETH_P_RARP))) { 174 br_do_proxy_suppress_arp(skb, br, vid, p); 175 } else if (ipv6_mod_enabled() && 176 skb->protocol == htons(ETH_P_IPV6) && 177 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 178 pskb_may_pull(skb, sizeof(struct ipv6hdr) + 179 sizeof(struct nd_msg)) && 180 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { 181 struct nd_msg *msg; 182 183 msg = br_is_nd_neigh_msg(skb); 184 if (msg) 185 br_do_suppress_nd(skb, br, vid, p, msg); 186 } 187 188 switch (pkt_type) { 189 case BR_PKT_MULTICAST: 190 mdst = br_mdb_entry_skb_get(brmctx, skb, vid); 191 if ((mdst || BR_INPUT_SKB_CB_MROUTERS_ONLY(skb)) && 192 br_multicast_querier_exists(brmctx, eth_hdr(skb), mdst)) { 193 if ((mdst && mdst->host_joined) || 194 br_multicast_is_router(brmctx, skb) || 195 br->dev->flags & IFF_ALLMULTI) { 196 local_rcv = true; 197 DEV_STATS_INC(br->dev, multicast); 198 } 199 mcast_hit = true; 200 } else { 201 local_rcv = true; 202 DEV_STATS_INC(br->dev, multicast); 203 } 204 break; 205 case BR_PKT_UNICAST: 206 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, vid); 207 if (unlikely(!dst && vid && 208 br_opt_get(br, BROPT_FDB_LOCAL_VLAN_0))) { 209 dst = br_fdb_find_rcu(br, eth_hdr(skb)->h_dest, 0); 210 if (dst && 211 (!test_bit(BR_FDB_LOCAL, &dst->flags) || 212 test_bit(BR_FDB_ADDED_BY_USER, &dst->flags))) 213 dst = NULL; 214 } 215 break; 216 default: 217 break; 218 } 219 220 if (dst) { 221 unsigned long now = jiffies; 222 223 if (test_bit(BR_FDB_LOCAL, &dst->flags)) 224 return br_pass_frame_up(skb, false); 225 226 if (now != READ_ONCE(dst->used)) 227 WRITE_ONCE(dst->used, now); 228 br_forward(br_fdb_dst_read(dst), skb, local_rcv, false); 229 } else { 230 if (!mcast_hit) 231 br_flood(br, vlan, skb, pkt_type, local_rcv, false); 232 else 233 br_multicast_flood(mdst, skb, brmctx, local_rcv, false); 234 } 235 236 if (local_rcv) 237 return br_pass_frame_up(skb, promisc); 238 239 out: 240 return 0; 241 drop: 242 kfree_skb_reason(skb, reason); 243 goto out; 244 } 245 EXPORT_SYMBOL_GPL(br_handle_frame_finish); 246 247 static void __br_handle_local_finish(struct sk_buff *skb) 248 { 249 struct net_bridge_port *p = br_port_get_rcu(skb->dev); 250 struct net_bridge_vlan *vlan; 251 252 /* check if vlan is allowed, to avoid spoofing */ 253 if (test_bit(BR_LEARNING_BIT, &p->flags) && 254 nbp_state_should_learn(p) && 255 !br_opt_get(p->br, BROPT_NO_LL_LEARN) && 256 br_should_learn(p, skb, &vlan)) 257 br_fdb_update(p->br, p, vlan, eth_hdr(skb)->h_source, 0); 258 } 259 260 /* note: already called with rcu_read_lock */ 261 static int br_handle_local_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 262 { 263 __br_handle_local_finish(skb); 264 265 /* return 1 to signal the okfn() was called so it's ok to use the skb */ 266 return 1; 267 } 268 269 static int nf_hook_bridge_pre(struct sk_buff *skb, struct sk_buff **pskb) 270 { 271 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE 272 struct nf_hook_entries *e = NULL; 273 struct nf_hook_state state; 274 unsigned int verdict, i; 275 struct net *net; 276 int ret; 277 278 net = dev_net(skb->dev); 279 #ifdef CONFIG_JUMP_LABEL 280 if (!static_key_false(&nf_hooks_needed[NFPROTO_BRIDGE][NF_BR_PRE_ROUTING])) 281 goto frame_finish; 282 #endif 283 284 e = rcu_dereference(net->nf.hooks_bridge[NF_BR_PRE_ROUTING]); 285 if (!e) 286 goto frame_finish; 287 288 nf_hook_state_init(&state, NF_BR_PRE_ROUTING, 289 NFPROTO_BRIDGE, skb->dev, NULL, NULL, 290 net, br_handle_frame_finish); 291 292 for (i = 0; i < e->num_hook_entries; i++) { 293 verdict = nf_hook_entry_hookfn(&e->hooks[i], skb, &state); 294 switch (verdict & NF_VERDICT_MASK) { 295 case NF_ACCEPT: 296 if (BR_INPUT_SKB_CB(skb)->br_netfilter_broute) { 297 *pskb = skb; 298 return RX_HANDLER_PASS; 299 } 300 break; 301 case NF_DROP: 302 kfree_skb(skb); 303 return RX_HANDLER_CONSUMED; 304 case NF_QUEUE: 305 ret = nf_queue(skb, &state, i, verdict); 306 if (ret == 1) 307 continue; 308 return RX_HANDLER_CONSUMED; 309 default: /* STOLEN */ 310 return RX_HANDLER_CONSUMED; 311 } 312 } 313 frame_finish: 314 net = dev_net(skb->dev); 315 br_handle_frame_finish(net, NULL, skb); 316 #else 317 br_handle_frame_finish(dev_net(skb->dev), NULL, skb); 318 #endif 319 return RX_HANDLER_CONSUMED; 320 } 321 322 #define BR_CFM_MRP_OPTS \ 323 ((IS_ENABLED(CONFIG_BRIDGE_CFM) ? BIT(BROPT_CFM_ENABLED) : 0UL) | \ 324 (IS_ENABLED(CONFIG_BRIDGE_MRP) ? BIT(BROPT_MRP_ENABLED) : 0UL)) 325 326 /* Return 0 if the frame was not processed otherwise 1 327 * note: already called with rcu_read_lock 328 */ 329 static int br_process_frame_type(struct net_bridge_port *p, 330 struct sk_buff *skb) 331 { 332 struct net_bridge *br = p->br; 333 334 if (skb->protocol == htons(ETH_P_CFM) && 335 br_opt_get(br, BROPT_CFM_ENABLED)) 336 return br_cfm_frame_rx(p, skb); 337 338 if (skb->protocol == htons(ETH_P_MRP) && 339 br_opt_get(br, BROPT_MRP_ENABLED)) 340 return br_mrp_process(p, skb); 341 342 return 0; 343 } 344 345 /* 346 * Return NULL if skb is handled 347 * note: already called with rcu_read_lock 348 */ 349 static rx_handler_result_t br_handle_frame(struct sk_buff **pskb) 350 { 351 enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED; 352 struct net_bridge_port *p; 353 struct sk_buff *skb = *pskb; 354 const unsigned char *dest = eth_hdr(skb)->h_dest; 355 356 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 357 return RX_HANDLER_PASS; 358 359 if (!is_valid_ether_addr(eth_hdr(skb)->h_source)) { 360 reason = SKB_DROP_REASON_MAC_INVALID_SOURCE; 361 goto drop; 362 } 363 364 skb = skb_share_check(skb, GFP_ATOMIC); 365 if (!skb) 366 return RX_HANDLER_CONSUMED; 367 368 memset(skb->cb, 0, sizeof(struct br_input_skb_cb)); 369 br_tc_skb_miss_set(skb, false); 370 371 p = br_port_get_rcu(skb->dev); 372 if (test_bit(BR_VLAN_TUNNEL_BIT, &p->flags)) 373 br_handle_ingress_vlan_tunnel(skb, p, nbp_vlan_group_rcu(p)); 374 375 if (unlikely(is_link_local_ether_addr(dest))) { 376 u16 fwd_mask = p->br->group_fwd_mask_required; 377 378 /* 379 * See IEEE 802.1D Table 7-10 Reserved addresses 380 * 381 * Assignment Value 382 * Bridge Group Address 01-80-C2-00-00-00 383 * (MAC Control) 802.3 01-80-C2-00-00-01 384 * (Link Aggregation) 802.3 01-80-C2-00-00-02 385 * 802.1X PAE address 01-80-C2-00-00-03 386 * 387 * 802.1AB LLDP 01-80-C2-00-00-0E 388 * 389 * Others reserved for future standardization 390 */ 391 fwd_mask |= p->group_fwd_mask; 392 switch (dest[5]) { 393 case 0x00: /* Bridge Group Address */ 394 /* If STP is turned off, 395 then must forward to keep loop detection */ 396 if (p->br->stp_enabled == BR_NO_STP || 397 fwd_mask & (1u << dest[5])) 398 goto forward; 399 *pskb = skb; 400 __br_handle_local_finish(skb); 401 return RX_HANDLER_PASS; 402 403 case 0x01: /* IEEE MAC (Pause) */ 404 reason = SKB_DROP_REASON_MAC_IEEE_MAC_CONTROL; 405 goto drop; 406 407 case 0x0E: /* 802.1AB LLDP */ 408 fwd_mask |= p->br->group_fwd_mask; 409 if (fwd_mask & (1u << dest[5])) 410 goto forward; 411 *pskb = skb; 412 __br_handle_local_finish(skb); 413 return RX_HANDLER_PASS; 414 415 default: 416 /* Allow selective forwarding for most other protocols */ 417 fwd_mask |= p->br->group_fwd_mask; 418 if (fwd_mask & (1u << dest[5])) 419 goto forward; 420 } 421 422 BR_INPUT_SKB_CB(skb)->promisc = false; 423 424 /* The else clause should be hit when nf_hook(): 425 * - returns < 0 (drop/error) 426 * - returns = 0 (stolen/nf_queue) 427 * Thus return 1 from the okfn() to signal the skb is ok to pass 428 */ 429 if (NF_HOOK(NFPROTO_BRIDGE, NF_BR_LOCAL_IN, 430 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 431 br_handle_local_finish) == 1) { 432 return RX_HANDLER_PASS; 433 } else { 434 return RX_HANDLER_CONSUMED; 435 } 436 } 437 438 if (unlikely((READ_ONCE(p->br->options) & BR_CFM_MRP_OPTS) && 439 br_process_frame_type(p, skb))) 440 return RX_HANDLER_PASS; 441 442 forward: 443 if (br_mst_is_enabled(p)) 444 goto defer_stp_filtering; 445 446 switch (p->state) { 447 case BR_STATE_FORWARDING: 448 case BR_STATE_LEARNING: 449 defer_stp_filtering: 450 if (ether_addr_equal(p->br->dev->dev_addr, dest)) 451 skb->pkt_type = PACKET_HOST; 452 453 return nf_hook_bridge_pre(skb, pskb); 454 default: 455 reason = SKB_DROP_REASON_BRIDGE_INGRESS_STP_STATE; 456 drop: 457 kfree_skb_reason(skb, reason); 458 } 459 return RX_HANDLER_CONSUMED; 460 } 461 462 /* This function has no purpose other than to appease the br_port_get_rcu/rtnl 463 * helpers which identify bridged ports according to the rx_handler installed 464 * on them (so there _needs_ to be a bridge rx_handler even if we don't need it 465 * to do anything useful). This bridge won't support traffic to/from the stack, 466 * but only hardware bridging. So return RX_HANDLER_PASS so we don't steal 467 * frames from the ETH_P_XDSA packet_type handler. 468 */ 469 static rx_handler_result_t br_handle_frame_dummy(struct sk_buff **pskb) 470 { 471 return RX_HANDLER_PASS; 472 } 473 474 rx_handler_func_t *br_get_rx_handler(const struct net_device *dev) 475 { 476 if (netdev_uses_dsa(dev)) 477 return br_handle_frame_dummy; 478 479 return br_handle_frame; 480 } 481