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