net/bridge/br_arp_nd_proxy.c

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