1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef __NET_IP_TUNNELS_H 3 #define __NET_IP_TUNNELS_H 1 4 5 #include <linux/if_tunnel.h> 6 #include <linux/netdevice.h> 7 #include <linux/skbuff.h> 8 #include <linux/socket.h> 9 #include <linux/types.h> 10 #include <linux/u64_stats_sync.h> 11 #include <linux/bitops.h> 12 13 #include <net/dsfield.h> 14 #include <net/gro_cells.h> 15 #include <net/inet_ecn.h> 16 #include <net/netns/generic.h> 17 #include <net/rtnetlink.h> 18 #include <net/lwtunnel.h> 19 #include <net/dst_cache.h> 20 21 #if IS_ENABLED(CONFIG_IPV6) 22 #include <net/ipv6.h> 23 #include <net/ip6_fib.h> 24 #include <net/ip6_route.h> 25 #endif 26 27 /* Keep error state on tunnel for 30 sec */ 28 #define IPTUNNEL_ERR_TIMEO (30*HZ) 29 30 /* Used to memset ip_tunnel padding. */ 31 #define IP_TUNNEL_KEY_SIZE offsetofend(struct ip_tunnel_key, tp_dst) 32 33 /* Used to memset ipv4 address padding. */ 34 #define IP_TUNNEL_KEY_IPV4_PAD offsetofend(struct ip_tunnel_key, u.ipv4.dst) 35 #define IP_TUNNEL_KEY_IPV4_PAD_LEN \ 36 (sizeof_field(struct ip_tunnel_key, u) - \ 37 sizeof_field(struct ip_tunnel_key, u.ipv4)) 38 39 struct ip_tunnel_key { 40 __be64 tun_id; 41 union { 42 struct { 43 __be32 src; 44 __be32 dst; 45 } ipv4; 46 struct { 47 struct in6_addr src; 48 struct in6_addr dst; 49 } ipv6; 50 } u; 51 __be16 tun_flags; 52 u8 tos; /* TOS for IPv4, TC for IPv6 */ 53 u8 ttl; /* TTL for IPv4, HL for IPv6 */ 54 __be32 label; /* Flow Label for IPv6 */ 55 __be16 tp_src; 56 __be16 tp_dst; 57 __u8 flow_flags; 58 }; 59 60 /* Flags for ip_tunnel_info mode. */ 61 #define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */ 62 #define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */ 63 #define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */ 64 65 /* Maximum tunnel options length. */ 66 #define IP_TUNNEL_OPTS_MAX \ 67 GENMASK((sizeof_field(struct ip_tunnel_info, \ 68 options_len) * BITS_PER_BYTE) - 1, 0) 69 70 struct ip_tunnel_info { 71 struct ip_tunnel_key key; 72 #ifdef CONFIG_DST_CACHE 73 struct dst_cache dst_cache; 74 #endif 75 u8 options_len; 76 u8 mode; 77 }; 78 79 /* 6rd prefix/relay information */ 80 #ifdef CONFIG_IPV6_SIT_6RD 81 struct ip_tunnel_6rd_parm { 82 struct in6_addr prefix; 83 __be32 relay_prefix; 84 u16 prefixlen; 85 u16 relay_prefixlen; 86 }; 87 #endif 88 89 struct ip_tunnel_encap { 90 u16 type; 91 u16 flags; 92 __be16 sport; 93 __be16 dport; 94 }; 95 96 struct ip_tunnel_prl_entry { 97 struct ip_tunnel_prl_entry __rcu *next; 98 __be32 addr; 99 u16 flags; 100 struct rcu_head rcu_head; 101 }; 102 103 struct metadata_dst; 104 105 struct ip_tunnel { 106 struct ip_tunnel __rcu *next; 107 struct hlist_node hash_node; 108 109 struct net_device *dev; 110 netdevice_tracker dev_tracker; 111 112 struct net *net; /* netns for packet i/o */ 113 114 unsigned long err_time; /* Time when the last ICMP error 115 * arrived */ 116 int err_count; /* Number of arrived ICMP errors */ 117 118 /* These four fields used only by GRE */ 119 u32 i_seqno; /* The last seen seqno */ 120 atomic_t o_seqno; /* The last output seqno */ 121 int tun_hlen; /* Precalculated header length */ 122 123 /* These four fields used only by ERSPAN */ 124 u32 index; /* ERSPAN type II index */ 125 u8 erspan_ver; /* ERSPAN version */ 126 u8 dir; /* ERSPAN direction */ 127 u16 hwid; /* ERSPAN hardware ID */ 128 129 struct dst_cache dst_cache; 130 131 struct ip_tunnel_parm parms; 132 133 int mlink; 134 int encap_hlen; /* Encap header length (FOU,GUE) */ 135 int hlen; /* tun_hlen + encap_hlen */ 136 struct ip_tunnel_encap encap; 137 138 /* for SIT */ 139 #ifdef CONFIG_IPV6_SIT_6RD 140 struct ip_tunnel_6rd_parm ip6rd; 141 #endif 142 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */ 143 unsigned int prl_count; /* # of entries in PRL */ 144 unsigned int ip_tnl_net_id; 145 struct gro_cells gro_cells; 146 __u32 fwmark; 147 bool collect_md; 148 bool ignore_df; 149 }; 150 151 struct tnl_ptk_info { 152 __be16 flags; 153 __be16 proto; 154 __be32 key; 155 __be32 seq; 156 int hdr_len; 157 }; 158 159 #define PACKET_RCVD 0 160 #define PACKET_REJECT 1 161 #define PACKET_NEXT 2 162 163 #define IP_TNL_HASH_BITS 7 164 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS) 165 166 struct ip_tunnel_net { 167 struct net_device *fb_tunnel_dev; 168 struct rtnl_link_ops *rtnl_link_ops; 169 struct hlist_head tunnels[IP_TNL_HASH_SIZE]; 170 struct ip_tunnel __rcu *collect_md_tun; 171 int type; 172 }; 173 174 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key, 175 __be32 saddr, __be32 daddr, 176 u8 tos, u8 ttl, __be32 label, 177 __be16 tp_src, __be16 tp_dst, 178 __be64 tun_id, __be16 tun_flags) 179 { 180 key->tun_id = tun_id; 181 key->u.ipv4.src = saddr; 182 key->u.ipv4.dst = daddr; 183 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD, 184 0, IP_TUNNEL_KEY_IPV4_PAD_LEN); 185 key->tos = tos; 186 key->ttl = ttl; 187 key->label = label; 188 key->tun_flags = tun_flags; 189 190 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of 191 * the upper tunnel are used. 192 * E.g: GRE over IPSEC, the tp_src and tp_port are zero. 193 */ 194 key->tp_src = tp_src; 195 key->tp_dst = tp_dst; 196 197 /* Clear struct padding. */ 198 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE) 199 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE, 200 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE); 201 } 202 203 static inline bool 204 ip_tunnel_dst_cache_usable(const struct sk_buff *skb, 205 const struct ip_tunnel_info *info) 206 { 207 if (skb->mark) 208 return false; 209 if (!info) 210 return true; 211 if (info->key.tun_flags & TUNNEL_NOCACHE) 212 return false; 213 214 return true; 215 } 216 217 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info 218 *tun_info) 219 { 220 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET; 221 } 222 223 static inline __be64 key32_to_tunnel_id(__be32 key) 224 { 225 #ifdef __BIG_ENDIAN 226 return (__force __be64)key; 227 #else 228 return (__force __be64)((__force u64)key << 32); 229 #endif 230 } 231 232 /* Returns the least-significant 32 bits of a __be64. */ 233 static inline __be32 tunnel_id_to_key32(__be64 tun_id) 234 { 235 #ifdef __BIG_ENDIAN 236 return (__force __be32)tun_id; 237 #else 238 return (__force __be32)((__force u64)tun_id >> 32); 239 #endif 240 } 241 242 #ifdef CONFIG_INET 243 244 static inline void ip_tunnel_init_flow(struct flowi4 *fl4, 245 int proto, 246 __be32 daddr, __be32 saddr, 247 __be32 key, __u8 tos, 248 struct net *net, int oif, 249 __u32 mark, __u32 tun_inner_hash, 250 __u8 flow_flags) 251 { 252 memset(fl4, 0, sizeof(*fl4)); 253 254 if (oif) { 255 fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index_rcu(net, oif); 256 /* Legacy VRF/l3mdev use case */ 257 fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif; 258 } 259 260 fl4->daddr = daddr; 261 fl4->saddr = saddr; 262 fl4->flowi4_tos = tos; 263 fl4->flowi4_proto = proto; 264 fl4->fl4_gre_key = key; 265 fl4->flowi4_mark = mark; 266 fl4->flowi4_multipath_hash = tun_inner_hash; 267 fl4->flowi4_flags = flow_flags; 268 } 269 270 int ip_tunnel_init(struct net_device *dev); 271 void ip_tunnel_uninit(struct net_device *dev); 272 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head); 273 struct net *ip_tunnel_get_link_net(const struct net_device *dev); 274 int ip_tunnel_get_iflink(const struct net_device *dev); 275 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id, 276 struct rtnl_link_ops *ops, char *devname); 277 278 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id, 279 struct rtnl_link_ops *ops); 280 281 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 282 const struct iphdr *tnl_params, const u8 protocol); 283 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 284 const u8 proto, int tunnel_hlen); 285 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd); 286 int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr, 287 void __user *data, int cmd); 288 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict); 289 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu); 290 291 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn, 292 int link, __be16 flags, 293 __be32 remote, __be32 local, 294 __be32 key); 295 296 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb, 297 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst, 298 bool log_ecn_error); 299 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[], 300 struct ip_tunnel_parm *p, __u32 fwmark); 301 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[], 302 struct ip_tunnel_parm *p, __u32 fwmark); 303 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id); 304 305 extern const struct header_ops ip_tunnel_header_ops; 306 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb); 307 308 struct ip_tunnel_encap_ops { 309 size_t (*encap_hlen)(struct ip_tunnel_encap *e); 310 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e, 311 u8 *protocol, struct flowi4 *fl4); 312 int (*err_handler)(struct sk_buff *skb, u32 info); 313 }; 314 315 #define MAX_IPTUN_ENCAP_OPS 8 316 317 extern const struct ip_tunnel_encap_ops __rcu * 318 iptun_encaps[MAX_IPTUN_ENCAP_OPS]; 319 320 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op, 321 unsigned int num); 322 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op, 323 unsigned int num); 324 325 int ip_tunnel_encap_setup(struct ip_tunnel *t, 326 struct ip_tunnel_encap *ipencap); 327 328 static inline bool pskb_inet_may_pull(struct sk_buff *skb) 329 { 330 int nhlen; 331 332 switch (skb->protocol) { 333 #if IS_ENABLED(CONFIG_IPV6) 334 case htons(ETH_P_IPV6): 335 nhlen = sizeof(struct ipv6hdr); 336 break; 337 #endif 338 case htons(ETH_P_IP): 339 nhlen = sizeof(struct iphdr); 340 break; 341 default: 342 nhlen = 0; 343 } 344 345 return pskb_network_may_pull(skb, nhlen); 346 } 347 348 static inline int ip_encap_hlen(struct ip_tunnel_encap *e) 349 { 350 const struct ip_tunnel_encap_ops *ops; 351 int hlen = -EINVAL; 352 353 if (e->type == TUNNEL_ENCAP_NONE) 354 return 0; 355 356 if (e->type >= MAX_IPTUN_ENCAP_OPS) 357 return -EINVAL; 358 359 rcu_read_lock(); 360 ops = rcu_dereference(iptun_encaps[e->type]); 361 if (likely(ops && ops->encap_hlen)) 362 hlen = ops->encap_hlen(e); 363 rcu_read_unlock(); 364 365 return hlen; 366 } 367 368 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t, 369 u8 *protocol, struct flowi4 *fl4) 370 { 371 const struct ip_tunnel_encap_ops *ops; 372 int ret = -EINVAL; 373 374 if (t->encap.type == TUNNEL_ENCAP_NONE) 375 return 0; 376 377 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS) 378 return -EINVAL; 379 380 rcu_read_lock(); 381 ops = rcu_dereference(iptun_encaps[t->encap.type]); 382 if (likely(ops && ops->build_header)) 383 ret = ops->build_header(skb, &t->encap, protocol, fl4); 384 rcu_read_unlock(); 385 386 return ret; 387 } 388 389 /* Extract dsfield from inner protocol */ 390 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph, 391 const struct sk_buff *skb) 392 { 393 __be16 payload_protocol = skb_protocol(skb, true); 394 395 if (payload_protocol == htons(ETH_P_IP)) 396 return iph->tos; 397 else if (payload_protocol == htons(ETH_P_IPV6)) 398 return ipv6_get_dsfield((const struct ipv6hdr *)iph); 399 else 400 return 0; 401 } 402 403 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph, 404 const struct sk_buff *skb) 405 { 406 __be16 payload_protocol = skb_protocol(skb, true); 407 408 if (payload_protocol == htons(ETH_P_IP)) 409 return iph->ttl; 410 else if (payload_protocol == htons(ETH_P_IPV6)) 411 return ((const struct ipv6hdr *)iph)->hop_limit; 412 else 413 return 0; 414 } 415 416 /* Propogate ECN bits out */ 417 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph, 418 const struct sk_buff *skb) 419 { 420 u8 inner = ip_tunnel_get_dsfield(iph, skb); 421 422 return INET_ECN_encapsulate(tos, inner); 423 } 424 425 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 426 __be16 inner_proto, bool raw_proto, bool xnet); 427 428 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 429 __be16 inner_proto, bool xnet) 430 { 431 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet); 432 } 433 434 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, 435 __be32 src, __be32 dst, u8 proto, 436 u8 tos, u8 ttl, __be16 df, bool xnet); 437 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md, 438 gfp_t flags); 439 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst, 440 int headroom, bool reply); 441 442 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask); 443 444 static inline int iptunnel_pull_offloads(struct sk_buff *skb) 445 { 446 if (skb_is_gso(skb)) { 447 int err; 448 449 err = skb_unclone(skb, GFP_ATOMIC); 450 if (unlikely(err)) 451 return err; 452 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >> 453 NETIF_F_GSO_SHIFT); 454 } 455 456 skb->encapsulation = 0; 457 return 0; 458 } 459 460 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len) 461 { 462 if (pkt_len > 0) { 463 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats); 464 465 u64_stats_update_begin(&tstats->syncp); 466 u64_stats_add(&tstats->tx_bytes, pkt_len); 467 u64_stats_inc(&tstats->tx_packets); 468 u64_stats_update_end(&tstats->syncp); 469 put_cpu_ptr(tstats); 470 } else { 471 struct net_device_stats *err_stats = &dev->stats; 472 473 if (pkt_len < 0) { 474 err_stats->tx_errors++; 475 err_stats->tx_aborted_errors++; 476 } else { 477 err_stats->tx_dropped++; 478 } 479 } 480 } 481 482 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info) 483 { 484 return info + 1; 485 } 486 487 static inline void ip_tunnel_info_opts_get(void *to, 488 const struct ip_tunnel_info *info) 489 { 490 memcpy(to, info + 1, info->options_len); 491 } 492 493 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 494 const void *from, int len, 495 __be16 flags) 496 { 497 info->options_len = len; 498 if (len > 0) { 499 memcpy(ip_tunnel_info_opts(info), from, len); 500 info->key.tun_flags |= flags; 501 } 502 } 503 504 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 505 { 506 return (struct ip_tunnel_info *)lwtstate->data; 507 } 508 509 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt); 510 511 /* Returns > 0 if metadata should be collected */ 512 static inline int ip_tunnel_collect_metadata(void) 513 { 514 return static_branch_unlikely(&ip_tunnel_metadata_cnt); 515 } 516 517 void __init ip_tunnel_core_init(void); 518 519 void ip_tunnel_need_metadata(void); 520 void ip_tunnel_unneed_metadata(void); 521 522 #else /* CONFIG_INET */ 523 524 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 525 { 526 return NULL; 527 } 528 529 static inline void ip_tunnel_need_metadata(void) 530 { 531 } 532 533 static inline void ip_tunnel_unneed_metadata(void) 534 { 535 } 536 537 static inline void ip_tunnel_info_opts_get(void *to, 538 const struct ip_tunnel_info *info) 539 { 540 } 541 542 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 543 const void *from, int len, 544 __be16 flags) 545 { 546 info->options_len = 0; 547 } 548 549 #endif /* CONFIG_INET */ 550 551 #endif /* __NET_IP_TUNNELS_H */ 552