1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Handle bridge arp/nd proxy/suppress 4 * 5 * Copyright (C) 2017 Cumulus Networks 6 * Copyright (c) 2017 Roopa Prabhu <roopa@cumulusnetworks.com> 7 * 8 * Authors: 9 * Roopa Prabhu <roopa@cumulusnetworks.com> 10 */ 11 12 #include <linux/kernel.h> 13 #include <linux/netdevice.h> 14 #include <linux/etherdevice.h> 15 #include <linux/neighbour.h> 16 #include <net/arp.h> 17 #include <linux/if_vlan.h> 18 #include <linux/inetdevice.h> 19 #include <net/addrconf.h> 20 #include <net/ipv6_stubs.h> 21 #if IS_ENABLED(CONFIG_IPV6) 22 #include <net/ip6_checksum.h> 23 #endif 24 25 #include "br_private.h" 26 27 void br_recalculate_neigh_suppress_enabled(struct net_bridge *br) 28 { 29 struct net_bridge_port *p; 30 bool neigh_suppress = false; 31 32 list_for_each_entry(p, &br->port_list, list) { 33 if (p->flags & BR_NEIGH_SUPPRESS) { 34 neigh_suppress = true; 35 break; 36 } 37 } 38 39 br_opt_toggle(br, BROPT_NEIGH_SUPPRESS_ENABLED, neigh_suppress); 40 } 41 42 #if IS_ENABLED(CONFIG_INET) 43 static void br_arp_send(struct net_bridge *br, struct net_bridge_port *p, 44 struct net_device *dev, __be32 dest_ip, __be32 src_ip, 45 const unsigned char *dest_hw, 46 const unsigned char *src_hw, 47 const unsigned char *target_hw, 48 __be16 vlan_proto, u16 vlan_tci) 49 { 50 struct net_bridge_vlan_group *vg; 51 struct sk_buff *skb; 52 u16 pvid; 53 54 netdev_dbg(dev, "arp send dev %s dst %pI4 dst_hw %pM src %pI4 src_hw %pM\n", 55 dev->name, &dest_ip, dest_hw, &src_ip, src_hw); 56 57 if (!vlan_tci) { 58 arp_send(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip, 59 dest_hw, src_hw, target_hw); 60 return; 61 } 62 63 skb = arp_create(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip, 64 dest_hw, src_hw, target_hw); 65 if (!skb) 66 return; 67 68 if (p) 69 vg = nbp_vlan_group_rcu(p); 70 else 71 vg = br_vlan_group_rcu(br); 72 pvid = br_get_pvid(vg); 73 if (pvid == (vlan_tci & VLAN_VID_MASK)) 74 vlan_tci = 0; 75 76 if (vlan_tci) 77 __vlan_hwaccel_put_tag(skb, vlan_proto, vlan_tci); 78 79 if (p) { 80 arp_xmit(skb); 81 } else { 82 skb_reset_mac_header(skb); 83 __skb_pull(skb, skb_network_offset(skb)); 84 skb->ip_summed = CHECKSUM_UNNECESSARY; 85 skb->pkt_type = PACKET_HOST; 86 87 netif_rx_ni(skb); 88 } 89 } 90 91 static int br_chk_addr_ip(struct net_device *dev, 92 struct netdev_nested_priv *priv) 93 { 94 __be32 ip = *(__be32 *)priv->data; 95 struct in_device *in_dev; 96 __be32 addr = 0; 97 98 in_dev = __in_dev_get_rcu(dev); 99 if (in_dev) 100 addr = inet_confirm_addr(dev_net(dev), in_dev, 0, ip, 101 RT_SCOPE_HOST); 102 103 if (addr == ip) 104 return 1; 105 106 return 0; 107 } 108 109 static bool br_is_local_ip(struct net_device *dev, __be32 ip) 110 { 111 struct netdev_nested_priv priv = { 112 .data = (void *)&ip, 113 }; 114 115 if (br_chk_addr_ip(dev, &priv)) 116 return true; 117 118 /* check if ip is configured on upper dev */ 119 if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip, &priv)) 120 return true; 121 122 return false; 123 } 124 125 void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br, 126 u16 vid, struct net_bridge_port *p) 127 { 128 struct net_device *dev = br->dev; 129 struct net_device *vlandev = dev; 130 struct neighbour *n; 131 struct arphdr *parp; 132 u8 *arpptr, *sha; 133 __be32 sip, tip; 134 135 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0; 136 137 if ((dev->flags & IFF_NOARP) || 138 !pskb_may_pull(skb, arp_hdr_len(dev))) 139 return; 140 141 parp = arp_hdr(skb); 142 143 if (parp->ar_pro != htons(ETH_P_IP) || 144 parp->ar_hln != dev->addr_len || 145 parp->ar_pln != 4) 146 return; 147 148 arpptr = (u8 *)parp + sizeof(struct arphdr); 149 sha = arpptr; 150 arpptr += dev->addr_len; /* sha */ 151 memcpy(&sip, arpptr, sizeof(sip)); 152 arpptr += sizeof(sip); 153 arpptr += dev->addr_len; /* tha */ 154 memcpy(&tip, arpptr, sizeof(tip)); 155 156 if (ipv4_is_loopback(tip) || 157 ipv4_is_multicast(tip)) 158 return; 159 160 if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) { 161 if (p && (p->flags & BR_NEIGH_SUPPRESS)) 162 return; 163 if (ipv4_is_zeronet(sip) || sip == tip) { 164 /* prevent flooding to neigh suppress ports */ 165 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 166 return; 167 } 168 } 169 170 if (parp->ar_op != htons(ARPOP_REQUEST)) 171 return; 172 173 if (vid != 0) { 174 vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto, 175 vid); 176 if (!vlandev) 177 return; 178 } 179 180 if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) && 181 br_is_local_ip(vlandev, tip)) { 182 /* its our local ip, so don't proxy reply 183 * and don't forward to neigh suppress ports 184 */ 185 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 186 return; 187 } 188 189 n = neigh_lookup(&arp_tbl, &tip, vlandev); 190 if (n) { 191 struct net_bridge_fdb_entry *f; 192 193 if (!(n->nud_state & NUD_VALID)) { 194 neigh_release(n); 195 return; 196 } 197 198 f = br_fdb_find_rcu(br, n->ha, vid); 199 if (f) { 200 bool replied = false; 201 202 if ((p && (p->flags & BR_PROXYARP)) || 203 (f->dst && (f->dst->flags & (BR_PROXYARP_WIFI | 204 BR_NEIGH_SUPPRESS)))) { 205 if (!vid) 206 br_arp_send(br, p, skb->dev, sip, tip, 207 sha, n->ha, sha, 0, 0); 208 else 209 br_arp_send(br, p, skb->dev, sip, tip, 210 sha, n->ha, sha, 211 skb->vlan_proto, 212 skb_vlan_tag_get(skb)); 213 replied = true; 214 } 215 216 /* If we have replied or as long as we know the 217 * mac, indicate to arp replied 218 */ 219 if (replied || 220 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) 221 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 222 } 223 224 neigh_release(n); 225 } 226 } 227 #endif 228 229 #if IS_ENABLED(CONFIG_IPV6) 230 struct nd_msg *br_is_nd_neigh_msg(struct sk_buff *skb, struct nd_msg *msg) 231 { 232 struct nd_msg *m; 233 234 m = skb_header_pointer(skb, skb_network_offset(skb) + 235 sizeof(struct ipv6hdr), sizeof(*msg), msg); 236 if (!m) 237 return NULL; 238 239 if (m->icmph.icmp6_code != 0 || 240 (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION && 241 m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT)) 242 return NULL; 243 244 return m; 245 } 246 247 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p, 248 struct sk_buff *request, struct neighbour *n, 249 __be16 vlan_proto, u16 vlan_tci, struct nd_msg *ns) 250 { 251 struct net_device *dev = request->dev; 252 struct net_bridge_vlan_group *vg; 253 struct sk_buff *reply; 254 struct nd_msg *na; 255 struct ipv6hdr *pip6; 256 int na_olen = 8; /* opt hdr + ETH_ALEN for target */ 257 int ns_olen; 258 int i, len; 259 u8 *daddr; 260 u16 pvid; 261 262 if (!dev) 263 return; 264 265 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + 266 sizeof(*na) + na_olen + dev->needed_tailroom; 267 268 reply = alloc_skb(len, GFP_ATOMIC); 269 if (!reply) 270 return; 271 272 reply->protocol = htons(ETH_P_IPV6); 273 reply->dev = dev; 274 skb_reserve(reply, LL_RESERVED_SPACE(dev)); 275 skb_push(reply, sizeof(struct ethhdr)); 276 skb_set_mac_header(reply, 0); 277 278 daddr = eth_hdr(request)->h_source; 279 280 /* Do we need option processing ? */ 281 ns_olen = request->len - (skb_network_offset(request) + 282 sizeof(struct ipv6hdr)) - sizeof(*ns); 283 for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) { 284 if (!ns->opt[i + 1]) { 285 kfree_skb(reply); 286 return; 287 } 288 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { 289 daddr = ns->opt + i + sizeof(struct nd_opt_hdr); 290 break; 291 } 292 } 293 294 /* Ethernet header */ 295 ether_addr_copy(eth_hdr(reply)->h_dest, daddr); 296 ether_addr_copy(eth_hdr(reply)->h_source, n->ha); 297 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); 298 reply->protocol = htons(ETH_P_IPV6); 299 300 skb_pull(reply, sizeof(struct ethhdr)); 301 skb_set_network_header(reply, 0); 302 skb_put(reply, sizeof(struct ipv6hdr)); 303 304 /* IPv6 header */ 305 pip6 = ipv6_hdr(reply); 306 memset(pip6, 0, sizeof(struct ipv6hdr)); 307 pip6->version = 6; 308 pip6->priority = ipv6_hdr(request)->priority; 309 pip6->nexthdr = IPPROTO_ICMPV6; 310 pip6->hop_limit = 255; 311 pip6->daddr = ipv6_hdr(request)->saddr; 312 pip6->saddr = *(struct in6_addr *)n->primary_key; 313 314 skb_pull(reply, sizeof(struct ipv6hdr)); 315 skb_set_transport_header(reply, 0); 316 317 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen); 318 319 /* Neighbor Advertisement */ 320 memset(na, 0, sizeof(*na) + na_olen); 321 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; 322 na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0; 323 na->icmph.icmp6_override = 1; 324 na->icmph.icmp6_solicited = 1; 325 na->target = ns->target; 326 ether_addr_copy(&na->opt[2], n->ha); 327 na->opt[0] = ND_OPT_TARGET_LL_ADDR; 328 na->opt[1] = na_olen >> 3; 329 330 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, 331 &pip6->daddr, 332 sizeof(*na) + na_olen, 333 IPPROTO_ICMPV6, 334 csum_partial(na, sizeof(*na) + na_olen, 0)); 335 336 pip6->payload_len = htons(sizeof(*na) + na_olen); 337 338 skb_push(reply, sizeof(struct ipv6hdr)); 339 skb_push(reply, sizeof(struct ethhdr)); 340 341 reply->ip_summed = CHECKSUM_UNNECESSARY; 342 343 if (p) 344 vg = nbp_vlan_group_rcu(p); 345 else 346 vg = br_vlan_group_rcu(br); 347 pvid = br_get_pvid(vg); 348 if (pvid == (vlan_tci & VLAN_VID_MASK)) 349 vlan_tci = 0; 350 351 if (vlan_tci) 352 __vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci); 353 354 netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n", 355 dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha); 356 357 if (p) { 358 dev_queue_xmit(reply); 359 } else { 360 skb_reset_mac_header(reply); 361 __skb_pull(reply, skb_network_offset(reply)); 362 reply->ip_summed = CHECKSUM_UNNECESSARY; 363 reply->pkt_type = PACKET_HOST; 364 365 netif_rx_ni(reply); 366 } 367 } 368 369 static int br_chk_addr_ip6(struct net_device *dev, 370 struct netdev_nested_priv *priv) 371 { 372 struct in6_addr *addr = (struct in6_addr *)priv->data; 373 374 if (ipv6_chk_addr(dev_net(dev), addr, dev, 0)) 375 return 1; 376 377 return 0; 378 } 379 380 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr) 381 382 { 383 struct netdev_nested_priv priv = { 384 .data = (void *)addr, 385 }; 386 387 if (br_chk_addr_ip6(dev, &priv)) 388 return true; 389 390 /* check if ip is configured on upper dev */ 391 if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv)) 392 return true; 393 394 return false; 395 } 396 397 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br, 398 u16 vid, struct net_bridge_port *p, struct nd_msg *msg) 399 { 400 struct net_device *dev = br->dev; 401 struct net_device *vlandev = NULL; 402 struct in6_addr *saddr, *daddr; 403 struct ipv6hdr *iphdr; 404 struct neighbour *n; 405 406 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0; 407 408 if (p && (p->flags & BR_NEIGH_SUPPRESS)) 409 return; 410 411 if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT && 412 !msg->icmph.icmp6_solicited) { 413 /* prevent flooding to neigh suppress ports */ 414 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 415 return; 416 } 417 418 if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION) 419 return; 420 421 iphdr = ipv6_hdr(skb); 422 saddr = &iphdr->saddr; 423 daddr = &iphdr->daddr; 424 425 if (ipv6_addr_any(saddr) || !ipv6_addr_cmp(saddr, daddr)) { 426 /* prevent flooding to neigh suppress ports */ 427 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 428 return; 429 } 430 431 if (vid != 0) { 432 /* build neigh table lookup on the vlan device */ 433 vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto, 434 vid); 435 if (!vlandev) 436 return; 437 } else { 438 vlandev = dev; 439 } 440 441 if (br_is_local_ip6(vlandev, &msg->target)) { 442 /* its our own ip, so don't proxy reply 443 * and don't forward to arp suppress ports 444 */ 445 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 446 return; 447 } 448 449 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, vlandev); 450 if (n) { 451 struct net_bridge_fdb_entry *f; 452 453 if (!(n->nud_state & NUD_VALID)) { 454 neigh_release(n); 455 return; 456 } 457 458 f = br_fdb_find_rcu(br, n->ha, vid); 459 if (f) { 460 bool replied = false; 461 462 if (f->dst && (f->dst->flags & BR_NEIGH_SUPPRESS)) { 463 if (vid != 0) 464 br_nd_send(br, p, skb, n, 465 skb->vlan_proto, 466 skb_vlan_tag_get(skb), msg); 467 else 468 br_nd_send(br, p, skb, n, 0, 0, msg); 469 replied = true; 470 } 471 472 /* If we have replied or as long as we know the 473 * mac, indicate to NEIGH_SUPPRESS ports that we 474 * have replied 475 */ 476 if (replied || 477 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) 478 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1; 479 } 480 neigh_release(n); 481 } 482 } 483 #endif 484