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 (FIELD_SIZEOF(struct ip_tunnel_key, u) - \ 37 FIELD_SIZEOF(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((FIELD_SIZEOF(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 struct net_device *dev; 108 struct net *net; /* netns for packet i/o */ 109 110 unsigned long err_time; /* Time when the last ICMP error 111 * arrived */ 112 int err_count; /* Number of arrived ICMP errors */ 113 114 /* These four fields used only by GRE */ 115 u32 i_seqno; /* The last seen seqno */ 116 u32 o_seqno; /* The last output seqno */ 117 int tun_hlen; /* Precalculated header length */ 118 119 /* These four fields used only by ERSPAN */ 120 u32 index; /* ERSPAN type II index */ 121 u8 erspan_ver; /* ERSPAN version */ 122 u8 dir; /* ERSPAN direction */ 123 u16 hwid; /* ERSPAN hardware ID */ 124 125 struct dst_cache dst_cache; 126 127 struct ip_tunnel_parm parms; 128 129 int mlink; 130 int encap_hlen; /* Encap header length (FOU,GUE) */ 131 int hlen; /* tun_hlen + encap_hlen */ 132 struct ip_tunnel_encap encap; 133 134 /* for SIT */ 135 #ifdef CONFIG_IPV6_SIT_6RD 136 struct ip_tunnel_6rd_parm ip6rd; 137 #endif 138 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */ 139 unsigned int prl_count; /* # of entries in PRL */ 140 unsigned int ip_tnl_net_id; 141 struct gro_cells gro_cells; 142 __u32 fwmark; 143 bool collect_md; 144 bool ignore_df; 145 }; 146 147 #define TUNNEL_CSUM __cpu_to_be16(0x01) 148 #define TUNNEL_ROUTING __cpu_to_be16(0x02) 149 #define TUNNEL_KEY __cpu_to_be16(0x04) 150 #define TUNNEL_SEQ __cpu_to_be16(0x08) 151 #define TUNNEL_STRICT __cpu_to_be16(0x10) 152 #define TUNNEL_REC __cpu_to_be16(0x20) 153 #define TUNNEL_VERSION __cpu_to_be16(0x40) 154 #define TUNNEL_NO_KEY __cpu_to_be16(0x80) 155 #define TUNNEL_DONT_FRAGMENT __cpu_to_be16(0x0100) 156 #define TUNNEL_OAM __cpu_to_be16(0x0200) 157 #define TUNNEL_CRIT_OPT __cpu_to_be16(0x0400) 158 #define TUNNEL_GENEVE_OPT __cpu_to_be16(0x0800) 159 #define TUNNEL_VXLAN_OPT __cpu_to_be16(0x1000) 160 #define TUNNEL_NOCACHE __cpu_to_be16(0x2000) 161 #define TUNNEL_ERSPAN_OPT __cpu_to_be16(0x4000) 162 163 #define TUNNEL_OPTIONS_PRESENT \ 164 (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT | TUNNEL_ERSPAN_OPT) 165 166 struct tnl_ptk_info { 167 __be16 flags; 168 __be16 proto; 169 __be32 key; 170 __be32 seq; 171 int hdr_len; 172 }; 173 174 #define PACKET_RCVD 0 175 #define PACKET_REJECT 1 176 #define PACKET_NEXT 2 177 178 #define IP_TNL_HASH_BITS 7 179 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS) 180 181 struct ip_tunnel_net { 182 struct net_device *fb_tunnel_dev; 183 struct rtnl_link_ops *rtnl_link_ops; 184 struct hlist_head tunnels[IP_TNL_HASH_SIZE]; 185 struct ip_tunnel __rcu *collect_md_tun; 186 int type; 187 }; 188 189 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key, 190 __be32 saddr, __be32 daddr, 191 u8 tos, u8 ttl, __be32 label, 192 __be16 tp_src, __be16 tp_dst, 193 __be64 tun_id, __be16 tun_flags) 194 { 195 key->tun_id = tun_id; 196 key->u.ipv4.src = saddr; 197 key->u.ipv4.dst = daddr; 198 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD, 199 0, IP_TUNNEL_KEY_IPV4_PAD_LEN); 200 key->tos = tos; 201 key->ttl = ttl; 202 key->label = label; 203 key->tun_flags = tun_flags; 204 205 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of 206 * the upper tunnel are used. 207 * E.g: GRE over IPSEC, the tp_src and tp_port are zero. 208 */ 209 key->tp_src = tp_src; 210 key->tp_dst = tp_dst; 211 212 /* Clear struct padding. */ 213 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE) 214 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE, 215 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE); 216 } 217 218 static inline bool 219 ip_tunnel_dst_cache_usable(const struct sk_buff *skb, 220 const struct ip_tunnel_info *info) 221 { 222 if (skb->mark) 223 return false; 224 if (!info) 225 return true; 226 if (info->key.tun_flags & TUNNEL_NOCACHE) 227 return false; 228 229 return true; 230 } 231 232 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info 233 *tun_info) 234 { 235 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET; 236 } 237 238 static inline __be64 key32_to_tunnel_id(__be32 key) 239 { 240 #ifdef __BIG_ENDIAN 241 return (__force __be64)key; 242 #else 243 return (__force __be64)((__force u64)key << 32); 244 #endif 245 } 246 247 /* Returns the least-significant 32 bits of a __be64. */ 248 static inline __be32 tunnel_id_to_key32(__be64 tun_id) 249 { 250 #ifdef __BIG_ENDIAN 251 return (__force __be32)tun_id; 252 #else 253 return (__force __be32)((__force u64)tun_id >> 32); 254 #endif 255 } 256 257 #ifdef CONFIG_INET 258 259 static inline void ip_tunnel_init_flow(struct flowi4 *fl4, 260 int proto, 261 __be32 daddr, __be32 saddr, 262 __be32 key, __u8 tos, int oif, 263 __u32 mark) 264 { 265 memset(fl4, 0, sizeof(*fl4)); 266 fl4->flowi4_oif = oif; 267 fl4->daddr = daddr; 268 fl4->saddr = saddr; 269 fl4->flowi4_tos = tos; 270 fl4->flowi4_proto = proto; 271 fl4->fl4_gre_key = key; 272 fl4->flowi4_mark = mark; 273 } 274 275 int ip_tunnel_init(struct net_device *dev); 276 void ip_tunnel_uninit(struct net_device *dev); 277 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head); 278 struct net *ip_tunnel_get_link_net(const struct net_device *dev); 279 int ip_tunnel_get_iflink(const struct net_device *dev); 280 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id, 281 struct rtnl_link_ops *ops, char *devname); 282 283 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id, 284 struct rtnl_link_ops *ops); 285 286 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 287 const struct iphdr *tnl_params, const u8 protocol); 288 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 289 const u8 proto); 290 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd); 291 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict); 292 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu); 293 294 void ip_tunnel_get_stats64(struct net_device *dev, 295 struct rtnl_link_stats64 *tot); 296 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn, 297 int link, __be16 flags, 298 __be32 remote, __be32 local, 299 __be32 key); 300 301 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb, 302 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst, 303 bool log_ecn_error); 304 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[], 305 struct ip_tunnel_parm *p, __u32 fwmark); 306 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[], 307 struct ip_tunnel_parm *p, __u32 fwmark); 308 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id); 309 310 struct ip_tunnel_encap_ops { 311 size_t (*encap_hlen)(struct ip_tunnel_encap *e); 312 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e, 313 u8 *protocol, struct flowi4 *fl4); 314 int (*err_handler)(struct sk_buff *skb, u32 info); 315 }; 316 317 #define MAX_IPTUN_ENCAP_OPS 8 318 319 extern const struct ip_tunnel_encap_ops __rcu * 320 iptun_encaps[MAX_IPTUN_ENCAP_OPS]; 321 322 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op, 323 unsigned int num); 324 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op, 325 unsigned int num); 326 327 int ip_tunnel_encap_setup(struct ip_tunnel *t, 328 struct ip_tunnel_encap *ipencap); 329 330 static inline int ip_encap_hlen(struct ip_tunnel_encap *e) 331 { 332 const struct ip_tunnel_encap_ops *ops; 333 int hlen = -EINVAL; 334 335 if (e->type == TUNNEL_ENCAP_NONE) 336 return 0; 337 338 if (e->type >= MAX_IPTUN_ENCAP_OPS) 339 return -EINVAL; 340 341 rcu_read_lock(); 342 ops = rcu_dereference(iptun_encaps[e->type]); 343 if (likely(ops && ops->encap_hlen)) 344 hlen = ops->encap_hlen(e); 345 rcu_read_unlock(); 346 347 return hlen; 348 } 349 350 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t, 351 u8 *protocol, struct flowi4 *fl4) 352 { 353 const struct ip_tunnel_encap_ops *ops; 354 int ret = -EINVAL; 355 356 if (t->encap.type == TUNNEL_ENCAP_NONE) 357 return 0; 358 359 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS) 360 return -EINVAL; 361 362 rcu_read_lock(); 363 ops = rcu_dereference(iptun_encaps[t->encap.type]); 364 if (likely(ops && ops->build_header)) 365 ret = ops->build_header(skb, &t->encap, protocol, fl4); 366 rcu_read_unlock(); 367 368 return ret; 369 } 370 371 /* Extract dsfield from inner protocol */ 372 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph, 373 const struct sk_buff *skb) 374 { 375 if (skb->protocol == htons(ETH_P_IP)) 376 return iph->tos; 377 else if (skb->protocol == htons(ETH_P_IPV6)) 378 return ipv6_get_dsfield((const struct ipv6hdr *)iph); 379 else 380 return 0; 381 } 382 383 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph, 384 const struct sk_buff *skb) 385 { 386 if (skb->protocol == htons(ETH_P_IP)) 387 return iph->ttl; 388 else if (skb->protocol == htons(ETH_P_IPV6)) 389 return ((const struct ipv6hdr *)iph)->hop_limit; 390 else 391 return 0; 392 } 393 394 /* Propogate ECN bits out */ 395 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph, 396 const struct sk_buff *skb) 397 { 398 u8 inner = ip_tunnel_get_dsfield(iph, skb); 399 400 return INET_ECN_encapsulate(tos, inner); 401 } 402 403 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 404 __be16 inner_proto, bool raw_proto, bool xnet); 405 406 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 407 __be16 inner_proto, bool xnet) 408 { 409 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet); 410 } 411 412 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, 413 __be32 src, __be32 dst, u8 proto, 414 u8 tos, u8 ttl, __be16 df, bool xnet); 415 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md, 416 gfp_t flags); 417 418 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask); 419 420 static inline int iptunnel_pull_offloads(struct sk_buff *skb) 421 { 422 if (skb_is_gso(skb)) { 423 int err; 424 425 err = skb_unclone(skb, GFP_ATOMIC); 426 if (unlikely(err)) 427 return err; 428 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >> 429 NETIF_F_GSO_SHIFT); 430 } 431 432 skb->encapsulation = 0; 433 return 0; 434 } 435 436 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len) 437 { 438 if (pkt_len > 0) { 439 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats); 440 441 u64_stats_update_begin(&tstats->syncp); 442 tstats->tx_bytes += pkt_len; 443 tstats->tx_packets++; 444 u64_stats_update_end(&tstats->syncp); 445 put_cpu_ptr(tstats); 446 } else { 447 struct net_device_stats *err_stats = &dev->stats; 448 449 if (pkt_len < 0) { 450 err_stats->tx_errors++; 451 err_stats->tx_aborted_errors++; 452 } else { 453 err_stats->tx_dropped++; 454 } 455 } 456 } 457 458 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info) 459 { 460 return info + 1; 461 } 462 463 static inline void ip_tunnel_info_opts_get(void *to, 464 const struct ip_tunnel_info *info) 465 { 466 memcpy(to, info + 1, info->options_len); 467 } 468 469 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 470 const void *from, int len, 471 __be16 flags) 472 { 473 memcpy(ip_tunnel_info_opts(info), from, len); 474 info->options_len = len; 475 info->key.tun_flags |= flags; 476 } 477 478 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 479 { 480 return (struct ip_tunnel_info *)lwtstate->data; 481 } 482 483 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt); 484 485 /* Returns > 0 if metadata should be collected */ 486 static inline int ip_tunnel_collect_metadata(void) 487 { 488 return static_branch_unlikely(&ip_tunnel_metadata_cnt); 489 } 490 491 void __init ip_tunnel_core_init(void); 492 493 void ip_tunnel_need_metadata(void); 494 void ip_tunnel_unneed_metadata(void); 495 496 #else /* CONFIG_INET */ 497 498 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 499 { 500 return NULL; 501 } 502 503 static inline void ip_tunnel_need_metadata(void) 504 { 505 } 506 507 static inline void ip_tunnel_unneed_metadata(void) 508 { 509 } 510 511 static inline void ip_tunnel_info_opts_get(void *to, 512 const struct ip_tunnel_info *info) 513 { 514 } 515 516 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 517 const void *from, int len, 518 __be16 flags) 519 { 520 info->options_len = 0; 521 info->key.tun_flags |= flags; 522 } 523 524 #endif /* CONFIG_INET */ 525 526 #endif /* __NET_IP_TUNNELS_H */ 527