1 #ifndef __NET_IP_TUNNELS_H 2 #define __NET_IP_TUNNELS_H 1 3 4 #include <linux/if_tunnel.h> 5 #include <linux/netdevice.h> 6 #include <linux/skbuff.h> 7 #include <linux/socket.h> 8 #include <linux/types.h> 9 #include <linux/u64_stats_sync.h> 10 #include <linux/bitops.h> 11 12 #include <net/dsfield.h> 13 #include <net/gro_cells.h> 14 #include <net/inet_ecn.h> 15 #include <net/netns/generic.h> 16 #include <net/rtnetlink.h> 17 #include <net/lwtunnel.h> 18 #include <net/dst_cache.h> 19 20 #if IS_ENABLED(CONFIG_IPV6) 21 #include <net/ipv6.h> 22 #include <net/ip6_fib.h> 23 #include <net/ip6_route.h> 24 #endif 25 26 /* Keep error state on tunnel for 30 sec */ 27 #define IPTUNNEL_ERR_TIMEO (30*HZ) 28 29 /* Used to memset ip_tunnel padding. */ 30 #define IP_TUNNEL_KEY_SIZE offsetofend(struct ip_tunnel_key, tp_dst) 31 32 /* Used to memset ipv4 address padding. */ 33 #define IP_TUNNEL_KEY_IPV4_PAD offsetofend(struct ip_tunnel_key, u.ipv4.dst) 34 #define IP_TUNNEL_KEY_IPV4_PAD_LEN \ 35 (FIELD_SIZEOF(struct ip_tunnel_key, u) - \ 36 FIELD_SIZEOF(struct ip_tunnel_key, u.ipv4)) 37 38 struct ip_tunnel_key { 39 __be64 tun_id; 40 union { 41 struct { 42 __be32 src; 43 __be32 dst; 44 } ipv4; 45 struct { 46 struct in6_addr src; 47 struct in6_addr dst; 48 } ipv6; 49 } u; 50 __be16 tun_flags; 51 u8 tos; /* TOS for IPv4, TC for IPv6 */ 52 u8 ttl; /* TTL for IPv4, HL for IPv6 */ 53 __be32 label; /* Flow Label for IPv6 */ 54 __be16 tp_src; 55 __be16 tp_dst; 56 }; 57 58 /* Flags for ip_tunnel_info mode. */ 59 #define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */ 60 #define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */ 61 #define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */ 62 63 /* Maximum tunnel options length. */ 64 #define IP_TUNNEL_OPTS_MAX \ 65 GENMASK((FIELD_SIZEOF(struct ip_tunnel_info, \ 66 options_len) * BITS_PER_BYTE) - 1, 0) 67 68 struct ip_tunnel_info { 69 struct ip_tunnel_key key; 70 #ifdef CONFIG_DST_CACHE 71 struct dst_cache dst_cache; 72 #endif 73 u8 options_len; 74 u8 mode; 75 }; 76 77 /* 6rd prefix/relay information */ 78 #ifdef CONFIG_IPV6_SIT_6RD 79 struct ip_tunnel_6rd_parm { 80 struct in6_addr prefix; 81 __be32 relay_prefix; 82 u16 prefixlen; 83 u16 relay_prefixlen; 84 }; 85 #endif 86 87 struct ip_tunnel_encap { 88 u16 type; 89 u16 flags; 90 __be16 sport; 91 __be16 dport; 92 }; 93 94 struct ip_tunnel_prl_entry { 95 struct ip_tunnel_prl_entry __rcu *next; 96 __be32 addr; 97 u16 flags; 98 struct rcu_head rcu_head; 99 }; 100 101 struct metadata_dst; 102 103 struct ip_tunnel { 104 struct ip_tunnel __rcu *next; 105 struct hlist_node hash_node; 106 struct net_device *dev; 107 struct net *net; /* netns for packet i/o */ 108 109 unsigned long err_time; /* Time when the last ICMP error 110 * arrived */ 111 int err_count; /* Number of arrived ICMP errors */ 112 113 /* These four fields used only by GRE */ 114 u32 i_seqno; /* The last seen seqno */ 115 u32 o_seqno; /* The last output seqno */ 116 int tun_hlen; /* Precalculated header length */ 117 118 /* This field used only by ERSPAN */ 119 u32 index; /* ERSPAN type II index */ 120 121 struct dst_cache dst_cache; 122 123 struct ip_tunnel_parm parms; 124 125 int mlink; 126 int encap_hlen; /* Encap header length (FOU,GUE) */ 127 int hlen; /* tun_hlen + encap_hlen */ 128 struct ip_tunnel_encap encap; 129 130 /* for SIT */ 131 #ifdef CONFIG_IPV6_SIT_6RD 132 struct ip_tunnel_6rd_parm ip6rd; 133 #endif 134 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */ 135 unsigned int prl_count; /* # of entries in PRL */ 136 unsigned int ip_tnl_net_id; 137 struct gro_cells gro_cells; 138 __u32 fwmark; 139 bool collect_md; 140 bool ignore_df; 141 }; 142 143 #define TUNNEL_CSUM __cpu_to_be16(0x01) 144 #define TUNNEL_ROUTING __cpu_to_be16(0x02) 145 #define TUNNEL_KEY __cpu_to_be16(0x04) 146 #define TUNNEL_SEQ __cpu_to_be16(0x08) 147 #define TUNNEL_STRICT __cpu_to_be16(0x10) 148 #define TUNNEL_REC __cpu_to_be16(0x20) 149 #define TUNNEL_VERSION __cpu_to_be16(0x40) 150 #define TUNNEL_NO_KEY __cpu_to_be16(0x80) 151 #define TUNNEL_DONT_FRAGMENT __cpu_to_be16(0x0100) 152 #define TUNNEL_OAM __cpu_to_be16(0x0200) 153 #define TUNNEL_CRIT_OPT __cpu_to_be16(0x0400) 154 #define TUNNEL_GENEVE_OPT __cpu_to_be16(0x0800) 155 #define TUNNEL_VXLAN_OPT __cpu_to_be16(0x1000) 156 #define TUNNEL_NOCACHE __cpu_to_be16(0x2000) 157 #define TUNNEL_ERSPAN_OPT __cpu_to_be16(0x4000) 158 159 #define TUNNEL_OPTIONS_PRESENT \ 160 (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT | TUNNEL_ERSPAN_OPT) 161 162 struct tnl_ptk_info { 163 __be16 flags; 164 __be16 proto; 165 __be32 key; 166 __be32 seq; 167 int hdr_len; 168 }; 169 170 #define PACKET_RCVD 0 171 #define PACKET_REJECT 1 172 #define PACKET_NEXT 2 173 174 #define IP_TNL_HASH_BITS 7 175 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS) 176 177 struct ip_tunnel_net { 178 struct net_device *fb_tunnel_dev; 179 struct hlist_head tunnels[IP_TNL_HASH_SIZE]; 180 struct ip_tunnel __rcu *collect_md_tun; 181 }; 182 183 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key, 184 __be32 saddr, __be32 daddr, 185 u8 tos, u8 ttl, __be32 label, 186 __be16 tp_src, __be16 tp_dst, 187 __be64 tun_id, __be16 tun_flags) 188 { 189 key->tun_id = tun_id; 190 key->u.ipv4.src = saddr; 191 key->u.ipv4.dst = daddr; 192 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD, 193 0, IP_TUNNEL_KEY_IPV4_PAD_LEN); 194 key->tos = tos; 195 key->ttl = ttl; 196 key->label = label; 197 key->tun_flags = tun_flags; 198 199 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of 200 * the upper tunnel are used. 201 * E.g: GRE over IPSEC, the tp_src and tp_port are zero. 202 */ 203 key->tp_src = tp_src; 204 key->tp_dst = tp_dst; 205 206 /* Clear struct padding. */ 207 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE) 208 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE, 209 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE); 210 } 211 212 static inline bool 213 ip_tunnel_dst_cache_usable(const struct sk_buff *skb, 214 const struct ip_tunnel_info *info) 215 { 216 if (skb->mark) 217 return false; 218 if (!info) 219 return true; 220 if (info->key.tun_flags & TUNNEL_NOCACHE) 221 return false; 222 223 return true; 224 } 225 226 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info 227 *tun_info) 228 { 229 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET; 230 } 231 232 static inline __be64 key32_to_tunnel_id(__be32 key) 233 { 234 #ifdef __BIG_ENDIAN 235 return (__force __be64)key; 236 #else 237 return (__force __be64)((__force u64)key << 32); 238 #endif 239 } 240 241 /* Returns the least-significant 32 bits of a __be64. */ 242 static inline __be32 tunnel_id_to_key32(__be64 tun_id) 243 { 244 #ifdef __BIG_ENDIAN 245 return (__force __be32)tun_id; 246 #else 247 return (__force __be32)((__force u64)tun_id >> 32); 248 #endif 249 } 250 251 #ifdef CONFIG_INET 252 253 int ip_tunnel_init(struct net_device *dev); 254 void ip_tunnel_uninit(struct net_device *dev); 255 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head); 256 struct net *ip_tunnel_get_link_net(const struct net_device *dev); 257 int ip_tunnel_get_iflink(const struct net_device *dev); 258 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id, 259 struct rtnl_link_ops *ops, char *devname); 260 261 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops); 262 263 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 264 const struct iphdr *tnl_params, const u8 protocol); 265 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 266 const u8 proto); 267 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd); 268 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict); 269 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu); 270 271 void ip_tunnel_get_stats64(struct net_device *dev, 272 struct rtnl_link_stats64 *tot); 273 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn, 274 int link, __be16 flags, 275 __be32 remote, __be32 local, 276 __be32 key); 277 278 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb, 279 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst, 280 bool log_ecn_error); 281 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[], 282 struct ip_tunnel_parm *p, __u32 fwmark); 283 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[], 284 struct ip_tunnel_parm *p, __u32 fwmark); 285 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id); 286 287 struct ip_tunnel_encap_ops { 288 size_t (*encap_hlen)(struct ip_tunnel_encap *e); 289 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e, 290 u8 *protocol, struct flowi4 *fl4); 291 }; 292 293 #define MAX_IPTUN_ENCAP_OPS 8 294 295 extern const struct ip_tunnel_encap_ops __rcu * 296 iptun_encaps[MAX_IPTUN_ENCAP_OPS]; 297 298 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op, 299 unsigned int num); 300 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op, 301 unsigned int num); 302 303 int ip_tunnel_encap_setup(struct ip_tunnel *t, 304 struct ip_tunnel_encap *ipencap); 305 306 static inline int ip_encap_hlen(struct ip_tunnel_encap *e) 307 { 308 const struct ip_tunnel_encap_ops *ops; 309 int hlen = -EINVAL; 310 311 if (e->type == TUNNEL_ENCAP_NONE) 312 return 0; 313 314 if (e->type >= MAX_IPTUN_ENCAP_OPS) 315 return -EINVAL; 316 317 rcu_read_lock(); 318 ops = rcu_dereference(iptun_encaps[e->type]); 319 if (likely(ops && ops->encap_hlen)) 320 hlen = ops->encap_hlen(e); 321 rcu_read_unlock(); 322 323 return hlen; 324 } 325 326 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t, 327 u8 *protocol, struct flowi4 *fl4) 328 { 329 const struct ip_tunnel_encap_ops *ops; 330 int ret = -EINVAL; 331 332 if (t->encap.type == TUNNEL_ENCAP_NONE) 333 return 0; 334 335 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS) 336 return -EINVAL; 337 338 rcu_read_lock(); 339 ops = rcu_dereference(iptun_encaps[t->encap.type]); 340 if (likely(ops && ops->build_header)) 341 ret = ops->build_header(skb, &t->encap, protocol, fl4); 342 rcu_read_unlock(); 343 344 return ret; 345 } 346 347 /* Extract dsfield from inner protocol */ 348 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph, 349 const struct sk_buff *skb) 350 { 351 if (skb->protocol == htons(ETH_P_IP)) 352 return iph->tos; 353 else if (skb->protocol == htons(ETH_P_IPV6)) 354 return ipv6_get_dsfield((const struct ipv6hdr *)iph); 355 else 356 return 0; 357 } 358 359 /* Propogate ECN bits out */ 360 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph, 361 const struct sk_buff *skb) 362 { 363 u8 inner = ip_tunnel_get_dsfield(iph, skb); 364 365 return INET_ECN_encapsulate(tos, inner); 366 } 367 368 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 369 __be16 inner_proto, bool raw_proto, bool xnet); 370 371 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, 372 __be16 inner_proto, bool xnet) 373 { 374 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet); 375 } 376 377 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, 378 __be32 src, __be32 dst, u8 proto, 379 u8 tos, u8 ttl, __be16 df, bool xnet); 380 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md, 381 gfp_t flags); 382 383 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask); 384 385 static inline int iptunnel_pull_offloads(struct sk_buff *skb) 386 { 387 if (skb_is_gso(skb)) { 388 int err; 389 390 err = skb_unclone(skb, GFP_ATOMIC); 391 if (unlikely(err)) 392 return err; 393 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >> 394 NETIF_F_GSO_SHIFT); 395 } 396 397 skb->encapsulation = 0; 398 return 0; 399 } 400 401 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len) 402 { 403 if (pkt_len > 0) { 404 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats); 405 406 u64_stats_update_begin(&tstats->syncp); 407 tstats->tx_bytes += pkt_len; 408 tstats->tx_packets++; 409 u64_stats_update_end(&tstats->syncp); 410 put_cpu_ptr(tstats); 411 } else { 412 struct net_device_stats *err_stats = &dev->stats; 413 414 if (pkt_len < 0) { 415 err_stats->tx_errors++; 416 err_stats->tx_aborted_errors++; 417 } else { 418 err_stats->tx_dropped++; 419 } 420 } 421 } 422 423 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info) 424 { 425 return info + 1; 426 } 427 428 static inline void ip_tunnel_info_opts_get(void *to, 429 const struct ip_tunnel_info *info) 430 { 431 memcpy(to, info + 1, info->options_len); 432 } 433 434 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 435 const void *from, int len) 436 { 437 memcpy(ip_tunnel_info_opts(info), from, len); 438 info->options_len = len; 439 } 440 441 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 442 { 443 return (struct ip_tunnel_info *)lwtstate->data; 444 } 445 446 extern struct static_key ip_tunnel_metadata_cnt; 447 448 /* Returns > 0 if metadata should be collected */ 449 static inline int ip_tunnel_collect_metadata(void) 450 { 451 return static_key_false(&ip_tunnel_metadata_cnt); 452 } 453 454 void __init ip_tunnel_core_init(void); 455 456 void ip_tunnel_need_metadata(void); 457 void ip_tunnel_unneed_metadata(void); 458 459 #else /* CONFIG_INET */ 460 461 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 462 { 463 return NULL; 464 } 465 466 static inline void ip_tunnel_need_metadata(void) 467 { 468 } 469 470 static inline void ip_tunnel_unneed_metadata(void) 471 { 472 } 473 474 static inline void ip_tunnel_info_opts_get(void *to, 475 const struct ip_tunnel_info *info) 476 { 477 } 478 479 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info, 480 const void *from, int len) 481 { 482 info->options_len = 0; 483 } 484 485 #endif /* CONFIG_INET */ 486 487 #endif /* __NET_IP_TUNNELS_H */ 488