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