1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Copyright (c) 2013 Nicira, Inc. 4 */ 5 6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 7 8 #include <linux/types.h> 9 #include <linux/kernel.h> 10 #include <linux/skbuff.h> 11 #include <linux/netdevice.h> 12 #include <linux/in.h> 13 #include <linux/if_arp.h> 14 #include <linux/init.h> 15 #include <linux/in6.h> 16 #include <linux/inetdevice.h> 17 #include <linux/netfilter_ipv4.h> 18 #include <linux/etherdevice.h> 19 #include <linux/if_ether.h> 20 #include <linux/if_vlan.h> 21 #include <linux/static_key.h> 22 23 #include <net/ip.h> 24 #include <net/icmp.h> 25 #include <net/protocol.h> 26 #include <net/ip_tunnels.h> 27 #include <net/ip6_tunnel.h> 28 #include <net/arp.h> 29 #include <net/checksum.h> 30 #include <net/dsfield.h> 31 #include <net/inet_ecn.h> 32 #include <net/xfrm.h> 33 #include <net/net_namespace.h> 34 #include <net/netns/generic.h> 35 #include <net/rtnetlink.h> 36 #include <net/dst_metadata.h> 37 #include <net/geneve.h> 38 #include <net/vxlan.h> 39 #include <net/erspan.h> 40 41 const struct ip_tunnel_encap_ops __rcu * 42 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly; 43 EXPORT_SYMBOL(iptun_encaps); 44 45 const struct ip6_tnl_encap_ops __rcu * 46 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly; 47 EXPORT_SYMBOL(ip6tun_encaps); 48 49 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, 50 __be32 src, __be32 dst, __u8 proto, 51 __u8 tos, __u8 ttl, __be16 df, bool xnet) 52 { 53 int pkt_len = skb->len - skb_inner_network_offset(skb); 54 struct net *net = dev_net(rt->dst.dev); 55 struct net_device *dev = skb->dev; 56 struct iphdr *iph; 57 int err; 58 59 skb_scrub_packet(skb, xnet); 60 61 skb_clear_hash_if_not_l4(skb); 62 skb_dst_set(skb, &rt->dst); 63 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 64 65 /* Push down and install the IP header. */ 66 skb_push(skb, sizeof(struct iphdr)); 67 skb_reset_network_header(skb); 68 69 iph = ip_hdr(skb); 70 71 iph->version = 4; 72 iph->ihl = sizeof(struct iphdr) >> 2; 73 iph->frag_off = ip_mtu_locked(&rt->dst) ? 0 : df; 74 iph->protocol = proto; 75 iph->tos = tos; 76 iph->daddr = dst; 77 iph->saddr = src; 78 iph->ttl = ttl; 79 __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); 80 81 err = ip_local_out(net, sk, skb); 82 83 if (dev) { 84 if (unlikely(net_xmit_eval(err))) 85 pkt_len = 0; 86 iptunnel_xmit_stats(dev, pkt_len); 87 } 88 } 89 EXPORT_SYMBOL_GPL(iptunnel_xmit); 90 91 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 92 __be16 inner_proto, bool raw_proto, bool xnet) 93 { 94 if (unlikely(!pskb_may_pull(skb, hdr_len))) 95 return -ENOMEM; 96 97 skb_pull_rcsum(skb, hdr_len); 98 99 if (!raw_proto && inner_proto == htons(ETH_P_TEB)) { 100 struct ethhdr *eh; 101 102 if (unlikely(!pskb_may_pull(skb, ETH_HLEN))) 103 return -ENOMEM; 104 105 eh = (struct ethhdr *)skb->data; 106 if (likely(eth_proto_is_802_3(eh->h_proto))) 107 skb->protocol = eh->h_proto; 108 else 109 skb->protocol = htons(ETH_P_802_2); 110 111 } else { 112 skb->protocol = inner_proto; 113 } 114 115 skb_clear_hash_if_not_l4(skb); 116 __vlan_hwaccel_clear_tag(skb); 117 skb_set_queue_mapping(skb, 0); 118 skb_scrub_packet(skb, xnet); 119 120 return iptunnel_pull_offloads(skb); 121 } 122 EXPORT_SYMBOL_GPL(__iptunnel_pull_header); 123 124 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md, 125 gfp_t flags) 126 { 127 struct metadata_dst *res; 128 struct ip_tunnel_info *dst, *src; 129 130 if (!md || md->type != METADATA_IP_TUNNEL || 131 md->u.tun_info.mode & IP_TUNNEL_INFO_TX) 132 return NULL; 133 134 src = &md->u.tun_info; 135 res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags); 136 if (!res) 137 return NULL; 138 139 dst = &res->u.tun_info; 140 dst->key.tun_id = src->key.tun_id; 141 if (src->mode & IP_TUNNEL_INFO_IPV6) 142 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src, 143 sizeof(struct in6_addr)); 144 else 145 dst->key.u.ipv4.dst = src->key.u.ipv4.src; 146 dst->key.tun_flags = src->key.tun_flags; 147 dst->mode = src->mode | IP_TUNNEL_INFO_TX; 148 ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src), 149 src->options_len, 0); 150 151 return res; 152 } 153 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply); 154 155 int iptunnel_handle_offloads(struct sk_buff *skb, 156 int gso_type_mask) 157 { 158 int err; 159 160 if (likely(!skb->encapsulation)) { 161 skb_reset_inner_headers(skb); 162 skb->encapsulation = 1; 163 } 164 165 if (skb_is_gso(skb)) { 166 err = skb_header_unclone(skb, GFP_ATOMIC); 167 if (unlikely(err)) 168 return err; 169 skb_shinfo(skb)->gso_type |= gso_type_mask; 170 return 0; 171 } 172 173 if (skb->ip_summed != CHECKSUM_PARTIAL) { 174 skb->ip_summed = CHECKSUM_NONE; 175 /* We clear encapsulation here to prevent badly-written 176 * drivers potentially deciding to offload an inner checksum 177 * if we set CHECKSUM_PARTIAL on the outer header. 178 * This should go away when the drivers are all fixed. 179 */ 180 skb->encapsulation = 0; 181 } 182 183 return 0; 184 } 185 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads); 186 187 /* Often modified stats are per cpu, other are shared (netdev->stats) */ 188 void ip_tunnel_get_stats64(struct net_device *dev, 189 struct rtnl_link_stats64 *tot) 190 { 191 int i; 192 193 netdev_stats_to_stats64(tot, &dev->stats); 194 195 for_each_possible_cpu(i) { 196 const struct pcpu_sw_netstats *tstats = 197 per_cpu_ptr(dev->tstats, i); 198 u64 rx_packets, rx_bytes, tx_packets, tx_bytes; 199 unsigned int start; 200 201 do { 202 start = u64_stats_fetch_begin_irq(&tstats->syncp); 203 rx_packets = tstats->rx_packets; 204 tx_packets = tstats->tx_packets; 205 rx_bytes = tstats->rx_bytes; 206 tx_bytes = tstats->tx_bytes; 207 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start)); 208 209 tot->rx_packets += rx_packets; 210 tot->tx_packets += tx_packets; 211 tot->rx_bytes += rx_bytes; 212 tot->tx_bytes += tx_bytes; 213 } 214 } 215 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64); 216 217 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = { 218 [LWTUNNEL_IP_ID] = { .type = NLA_U64 }, 219 [LWTUNNEL_IP_DST] = { .type = NLA_U32 }, 220 [LWTUNNEL_IP_SRC] = { .type = NLA_U32 }, 221 [LWTUNNEL_IP_TTL] = { .type = NLA_U8 }, 222 [LWTUNNEL_IP_TOS] = { .type = NLA_U8 }, 223 [LWTUNNEL_IP_FLAGS] = { .type = NLA_U16 }, 224 [LWTUNNEL_IP_OPTS] = { .type = NLA_NESTED }, 225 }; 226 227 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = { 228 [LWTUNNEL_IP_OPTS_GENEVE] = { .type = NLA_NESTED }, 229 [LWTUNNEL_IP_OPTS_VXLAN] = { .type = NLA_NESTED }, 230 [LWTUNNEL_IP_OPTS_ERSPAN] = { .type = NLA_NESTED }, 231 }; 232 233 static const struct nla_policy 234 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = { 235 [LWTUNNEL_IP_OPT_GENEVE_CLASS] = { .type = NLA_U16 }, 236 [LWTUNNEL_IP_OPT_GENEVE_TYPE] = { .type = NLA_U8 }, 237 [LWTUNNEL_IP_OPT_GENEVE_DATA] = { .type = NLA_BINARY, .len = 128 }, 238 }; 239 240 static const struct nla_policy 241 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = { 242 [LWTUNNEL_IP_OPT_VXLAN_GBP] = { .type = NLA_U32 }, 243 }; 244 245 static const struct nla_policy 246 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = { 247 [LWTUNNEL_IP_OPT_ERSPAN_VER] = { .type = NLA_U8 }, 248 [LWTUNNEL_IP_OPT_ERSPAN_INDEX] = { .type = NLA_U32 }, 249 [LWTUNNEL_IP_OPT_ERSPAN_DIR] = { .type = NLA_U8 }, 250 [LWTUNNEL_IP_OPT_ERSPAN_HWID] = { .type = NLA_U8 }, 251 }; 252 253 static int ip_tun_parse_opts_geneve(struct nlattr *attr, 254 struct ip_tunnel_info *info, int opts_len, 255 struct netlink_ext_ack *extack) 256 { 257 struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1]; 258 int data_len, err; 259 260 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr, 261 geneve_opt_policy, extack); 262 if (err) 263 return err; 264 265 if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] || 266 !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] || 267 !tb[LWTUNNEL_IP_OPT_GENEVE_DATA]) 268 return -EINVAL; 269 270 attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA]; 271 data_len = nla_len(attr); 272 if (data_len % 4) 273 return -EINVAL; 274 275 if (info) { 276 struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len; 277 278 memcpy(opt->opt_data, nla_data(attr), data_len); 279 opt->length = data_len / 4; 280 attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS]; 281 opt->opt_class = nla_get_be16(attr); 282 attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE]; 283 opt->type = nla_get_u8(attr); 284 info->key.tun_flags |= TUNNEL_GENEVE_OPT; 285 } 286 287 return sizeof(struct geneve_opt) + data_len; 288 } 289 290 static int ip_tun_parse_opts_vxlan(struct nlattr *attr, 291 struct ip_tunnel_info *info, int opts_len, 292 struct netlink_ext_ack *extack) 293 { 294 struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1]; 295 int err; 296 297 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr, 298 vxlan_opt_policy, extack); 299 if (err) 300 return err; 301 302 if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP]) 303 return -EINVAL; 304 305 if (info) { 306 struct vxlan_metadata *md = 307 ip_tunnel_info_opts(info) + opts_len; 308 309 attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP]; 310 md->gbp = nla_get_u32(attr); 311 info->key.tun_flags |= TUNNEL_VXLAN_OPT; 312 } 313 314 return sizeof(struct vxlan_metadata); 315 } 316 317 static int ip_tun_parse_opts_erspan(struct nlattr *attr, 318 struct ip_tunnel_info *info, int opts_len, 319 struct netlink_ext_ack *extack) 320 { 321 struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1]; 322 int err; 323 324 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr, 325 erspan_opt_policy, extack); 326 if (err) 327 return err; 328 329 if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER]) 330 return -EINVAL; 331 332 if (info) { 333 struct erspan_metadata *md = 334 ip_tunnel_info_opts(info) + opts_len; 335 336 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_VER]; 337 md->version = nla_get_u8(attr); 338 339 if (md->version == 1 && tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX]) { 340 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX]; 341 md->u.index = nla_get_be32(attr); 342 } else if (md->version == 2 && tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] && 343 tb[LWTUNNEL_IP_OPT_ERSPAN_HWID]) { 344 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR]; 345 md->u.md2.dir = nla_get_u8(attr); 346 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID]; 347 set_hwid(&md->u.md2, nla_get_u8(attr)); 348 } else { 349 return -EINVAL; 350 } 351 352 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 353 } 354 355 return sizeof(struct erspan_metadata); 356 } 357 358 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info, 359 struct netlink_ext_ack *extack) 360 { 361 int err, rem, opt_len, opts_len = 0, type = 0; 362 struct nlattr *nla; 363 364 if (!attr) 365 return 0; 366 367 err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX, 368 ip_opts_policy, extack); 369 if (err) 370 return err; 371 372 nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) { 373 switch (nla_type(nla)) { 374 case LWTUNNEL_IP_OPTS_GENEVE: 375 if (type && type != TUNNEL_GENEVE_OPT) 376 return -EINVAL; 377 opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len, 378 extack); 379 if (opt_len < 0) 380 return opt_len; 381 opts_len += opt_len; 382 if (opts_len > IP_TUNNEL_OPTS_MAX) 383 return -EINVAL; 384 type = TUNNEL_GENEVE_OPT; 385 break; 386 case LWTUNNEL_IP_OPTS_VXLAN: 387 if (type) 388 return -EINVAL; 389 opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len, 390 extack); 391 if (opt_len < 0) 392 return opt_len; 393 opts_len += opt_len; 394 type = TUNNEL_VXLAN_OPT; 395 break; 396 case LWTUNNEL_IP_OPTS_ERSPAN: 397 if (type) 398 return -EINVAL; 399 opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len, 400 extack); 401 if (opt_len < 0) 402 return opt_len; 403 opts_len += opt_len; 404 type = TUNNEL_ERSPAN_OPT; 405 break; 406 default: 407 return -EINVAL; 408 } 409 } 410 411 return opts_len; 412 } 413 414 static int ip_tun_get_optlen(struct nlattr *attr, 415 struct netlink_ext_ack *extack) 416 { 417 return ip_tun_parse_opts(attr, NULL, extack); 418 } 419 420 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info, 421 struct netlink_ext_ack *extack) 422 { 423 return ip_tun_parse_opts(attr, info, extack); 424 } 425 426 static int ip_tun_build_state(struct nlattr *attr, 427 unsigned int family, const void *cfg, 428 struct lwtunnel_state **ts, 429 struct netlink_ext_ack *extack) 430 { 431 struct nlattr *tb[LWTUNNEL_IP_MAX + 1]; 432 struct lwtunnel_state *new_state; 433 struct ip_tunnel_info *tun_info; 434 int err, opt_len; 435 436 err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr, 437 ip_tun_policy, extack); 438 if (err < 0) 439 return err; 440 441 opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack); 442 if (opt_len < 0) 443 return opt_len; 444 445 new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len); 446 if (!new_state) 447 return -ENOMEM; 448 449 new_state->type = LWTUNNEL_ENCAP_IP; 450 451 tun_info = lwt_tun_info(new_state); 452 453 err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack); 454 if (err < 0) { 455 lwtstate_free(new_state); 456 return err; 457 } 458 459 #ifdef CONFIG_DST_CACHE 460 err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL); 461 if (err) { 462 lwtstate_free(new_state); 463 return err; 464 } 465 #endif 466 467 if (tb[LWTUNNEL_IP_ID]) 468 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]); 469 470 if (tb[LWTUNNEL_IP_DST]) 471 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]); 472 473 if (tb[LWTUNNEL_IP_SRC]) 474 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]); 475 476 if (tb[LWTUNNEL_IP_TTL]) 477 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]); 478 479 if (tb[LWTUNNEL_IP_TOS]) 480 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]); 481 482 if (tb[LWTUNNEL_IP_FLAGS]) 483 tun_info->key.tun_flags |= 484 (nla_get_be16(tb[LWTUNNEL_IP_FLAGS]) & 485 ~TUNNEL_OPTIONS_PRESENT); 486 487 tun_info->mode = IP_TUNNEL_INFO_TX; 488 tun_info->options_len = opt_len; 489 490 *ts = new_state; 491 492 return 0; 493 } 494 495 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate) 496 { 497 #ifdef CONFIG_DST_CACHE 498 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate); 499 500 dst_cache_destroy(&tun_info->dst_cache); 501 #endif 502 } 503 504 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb, 505 struct ip_tunnel_info *tun_info) 506 { 507 struct geneve_opt *opt; 508 struct nlattr *nest; 509 int offset = 0; 510 511 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE); 512 if (!nest) 513 return -ENOMEM; 514 515 while (tun_info->options_len > offset) { 516 opt = ip_tunnel_info_opts(tun_info) + offset; 517 if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS, 518 opt->opt_class) || 519 nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) || 520 nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4, 521 opt->opt_data)) { 522 nla_nest_cancel(skb, nest); 523 return -ENOMEM; 524 } 525 offset += sizeof(*opt) + opt->length * 4; 526 } 527 528 nla_nest_end(skb, nest); 529 return 0; 530 } 531 532 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb, 533 struct ip_tunnel_info *tun_info) 534 { 535 struct vxlan_metadata *md; 536 struct nlattr *nest; 537 538 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN); 539 if (!nest) 540 return -ENOMEM; 541 542 md = ip_tunnel_info_opts(tun_info); 543 if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) { 544 nla_nest_cancel(skb, nest); 545 return -ENOMEM; 546 } 547 548 nla_nest_end(skb, nest); 549 return 0; 550 } 551 552 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb, 553 struct ip_tunnel_info *tun_info) 554 { 555 struct erspan_metadata *md; 556 struct nlattr *nest; 557 558 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN); 559 if (!nest) 560 return -ENOMEM; 561 562 md = ip_tunnel_info_opts(tun_info); 563 if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version)) 564 goto err; 565 566 if (md->version == 1 && 567 nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index)) 568 goto err; 569 570 if (md->version == 2 && 571 (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) || 572 nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID, 573 get_hwid(&md->u.md2)))) 574 goto err; 575 576 nla_nest_end(skb, nest); 577 return 0; 578 err: 579 nla_nest_cancel(skb, nest); 580 return -ENOMEM; 581 } 582 583 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type, 584 struct ip_tunnel_info *tun_info) 585 { 586 struct nlattr *nest; 587 int err = 0; 588 589 if (!(tun_info->key.tun_flags & TUNNEL_OPTIONS_PRESENT)) 590 return 0; 591 592 nest = nla_nest_start_noflag(skb, type); 593 if (!nest) 594 return -ENOMEM; 595 596 if (tun_info->key.tun_flags & TUNNEL_GENEVE_OPT) 597 err = ip_tun_fill_encap_opts_geneve(skb, tun_info); 598 else if (tun_info->key.tun_flags & TUNNEL_VXLAN_OPT) 599 err = ip_tun_fill_encap_opts_vxlan(skb, tun_info); 600 else if (tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT) 601 err = ip_tun_fill_encap_opts_erspan(skb, tun_info); 602 603 if (err) { 604 nla_nest_cancel(skb, nest); 605 return err; 606 } 607 608 nla_nest_end(skb, nest); 609 return 0; 610 } 611 612 static int ip_tun_fill_encap_info(struct sk_buff *skb, 613 struct lwtunnel_state *lwtstate) 614 { 615 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate); 616 617 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id, 618 LWTUNNEL_IP_PAD) || 619 nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) || 620 nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) || 621 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) || 622 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) || 623 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags) || 624 ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info)) 625 return -ENOMEM; 626 627 return 0; 628 } 629 630 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info) 631 { 632 int opt_len; 633 634 if (!(info->key.tun_flags & TUNNEL_OPTIONS_PRESENT)) 635 return 0; 636 637 opt_len = nla_total_size(0); /* LWTUNNEL_IP_OPTS */ 638 if (info->key.tun_flags & TUNNEL_GENEVE_OPT) { 639 struct geneve_opt *opt; 640 int offset = 0; 641 642 opt_len += nla_total_size(0); /* LWTUNNEL_IP_OPTS_GENEVE */ 643 while (info->options_len > offset) { 644 opt = ip_tunnel_info_opts(info) + offset; 645 opt_len += nla_total_size(2) /* OPT_GENEVE_CLASS */ 646 + nla_total_size(1) /* OPT_GENEVE_TYPE */ 647 + nla_total_size(opt->length * 4); 648 /* OPT_GENEVE_DATA */ 649 offset += sizeof(*opt) + opt->length * 4; 650 } 651 } else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) { 652 opt_len += nla_total_size(0) /* LWTUNNEL_IP_OPTS_VXLAN */ 653 + nla_total_size(4); /* OPT_VXLAN_GBP */ 654 } else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) { 655 struct erspan_metadata *md = ip_tunnel_info_opts(info); 656 657 opt_len += nla_total_size(0) /* LWTUNNEL_IP_OPTS_ERSPAN */ 658 + nla_total_size(1) /* OPT_ERSPAN_VER */ 659 + (md->version == 1 ? nla_total_size(4) 660 /* OPT_ERSPAN_INDEX (v1) */ 661 : nla_total_size(1) + 662 nla_total_size(1)); 663 /* OPT_ERSPAN_DIR + HWID (v2) */ 664 } 665 666 return opt_len; 667 } 668 669 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate) 670 { 671 return nla_total_size_64bit(8) /* LWTUNNEL_IP_ID */ 672 + nla_total_size(4) /* LWTUNNEL_IP_DST */ 673 + nla_total_size(4) /* LWTUNNEL_IP_SRC */ 674 + nla_total_size(1) /* LWTUNNEL_IP_TOS */ 675 + nla_total_size(1) /* LWTUNNEL_IP_TTL */ 676 + nla_total_size(2) /* LWTUNNEL_IP_FLAGS */ 677 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate)); 678 /* LWTUNNEL_IP_OPTS */ 679 } 680 681 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b) 682 { 683 struct ip_tunnel_info *info_a = lwt_tun_info(a); 684 struct ip_tunnel_info *info_b = lwt_tun_info(b); 685 686 return memcmp(info_a, info_b, sizeof(info_a->key)) || 687 info_a->mode != info_b->mode || 688 info_a->options_len != info_b->options_len || 689 memcmp(ip_tunnel_info_opts(info_a), 690 ip_tunnel_info_opts(info_b), info_a->options_len); 691 } 692 693 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = { 694 .build_state = ip_tun_build_state, 695 .destroy_state = ip_tun_destroy_state, 696 .fill_encap = ip_tun_fill_encap_info, 697 .get_encap_size = ip_tun_encap_nlsize, 698 .cmp_encap = ip_tun_cmp_encap, 699 .owner = THIS_MODULE, 700 }; 701 702 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = { 703 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 }, 704 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) }, 705 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) }, 706 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 }, 707 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 }, 708 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 }, 709 }; 710 711 static int ip6_tun_build_state(struct nlattr *attr, 712 unsigned int family, const void *cfg, 713 struct lwtunnel_state **ts, 714 struct netlink_ext_ack *extack) 715 { 716 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1]; 717 struct lwtunnel_state *new_state; 718 struct ip_tunnel_info *tun_info; 719 int err, opt_len; 720 721 err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr, 722 ip6_tun_policy, extack); 723 if (err < 0) 724 return err; 725 726 opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack); 727 if (opt_len < 0) 728 return opt_len; 729 730 new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len); 731 if (!new_state) 732 return -ENOMEM; 733 734 new_state->type = LWTUNNEL_ENCAP_IP6; 735 736 tun_info = lwt_tun_info(new_state); 737 738 err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack); 739 if (err < 0) { 740 lwtstate_free(new_state); 741 return err; 742 } 743 744 if (tb[LWTUNNEL_IP6_ID]) 745 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]); 746 747 if (tb[LWTUNNEL_IP6_DST]) 748 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]); 749 750 if (tb[LWTUNNEL_IP6_SRC]) 751 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]); 752 753 if (tb[LWTUNNEL_IP6_HOPLIMIT]) 754 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]); 755 756 if (tb[LWTUNNEL_IP6_TC]) 757 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]); 758 759 if (tb[LWTUNNEL_IP6_FLAGS]) 760 tun_info->key.tun_flags |= 761 (nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]) & 762 ~TUNNEL_OPTIONS_PRESENT); 763 764 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6; 765 tun_info->options_len = opt_len; 766 767 *ts = new_state; 768 769 return 0; 770 } 771 772 static int ip6_tun_fill_encap_info(struct sk_buff *skb, 773 struct lwtunnel_state *lwtstate) 774 { 775 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate); 776 777 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id, 778 LWTUNNEL_IP6_PAD) || 779 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) || 780 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) || 781 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) || 782 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) || 783 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags) || 784 ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info)) 785 return -ENOMEM; 786 787 return 0; 788 } 789 790 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate) 791 { 792 return nla_total_size_64bit(8) /* LWTUNNEL_IP6_ID */ 793 + nla_total_size(16) /* LWTUNNEL_IP6_DST */ 794 + nla_total_size(16) /* LWTUNNEL_IP6_SRC */ 795 + nla_total_size(1) /* LWTUNNEL_IP6_HOPLIMIT */ 796 + nla_total_size(1) /* LWTUNNEL_IP6_TC */ 797 + nla_total_size(2) /* LWTUNNEL_IP6_FLAGS */ 798 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate)); 799 /* LWTUNNEL_IP6_OPTS */ 800 } 801 802 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = { 803 .build_state = ip6_tun_build_state, 804 .fill_encap = ip6_tun_fill_encap_info, 805 .get_encap_size = ip6_tun_encap_nlsize, 806 .cmp_encap = ip_tun_cmp_encap, 807 .owner = THIS_MODULE, 808 }; 809 810 void __init ip_tunnel_core_init(void) 811 { 812 /* If you land here, make sure whether increasing ip_tunnel_info's 813 * options_len is a reasonable choice with its usage in front ends 814 * (f.e., it's part of flow keys, etc). 815 */ 816 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255); 817 818 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP); 819 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6); 820 } 821 822 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt); 823 EXPORT_SYMBOL(ip_tunnel_metadata_cnt); 824 825 void ip_tunnel_need_metadata(void) 826 { 827 static_branch_inc(&ip_tunnel_metadata_cnt); 828 } 829 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata); 830 831 void ip_tunnel_unneed_metadata(void) 832 { 833 static_branch_dec(&ip_tunnel_metadata_cnt); 834 } 835 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata); 836