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/types.h> 8 #include <linux/u64_stats_sync.h> 9 #include <net/dsfield.h> 10 #include <net/gro_cells.h> 11 #include <net/inet_ecn.h> 12 #include <net/netns/generic.h> 13 #include <net/rtnetlink.h> 14 #include <net/lwtunnel.h> 15 16 #if IS_ENABLED(CONFIG_IPV6) 17 #include <net/ipv6.h> 18 #include <net/ip6_fib.h> 19 #include <net/ip6_route.h> 20 #endif 21 22 /* Keep error state on tunnel for 30 sec */ 23 #define IPTUNNEL_ERR_TIMEO (30*HZ) 24 25 /* Used to memset ip_tunnel padding. */ 26 #define IP_TUNNEL_KEY_SIZE \ 27 (offsetof(struct ip_tunnel_key, tp_dst) + \ 28 FIELD_SIZEOF(struct ip_tunnel_key, tp_dst)) 29 30 struct ip_tunnel_key { 31 __be64 tun_id; 32 __be32 ipv4_src; 33 __be32 ipv4_dst; 34 __be16 tun_flags; 35 __u8 ipv4_tos; 36 __u8 ipv4_ttl; 37 __be16 tp_src; 38 __be16 tp_dst; 39 } __packed __aligned(4); /* Minimize padding. */ 40 41 /* Indicates whether the tunnel info structure represents receive 42 * or transmit tunnel parameters. 43 */ 44 enum { 45 IP_TUNNEL_INFO_RX, 46 IP_TUNNEL_INFO_TX, 47 }; 48 49 struct ip_tunnel_info { 50 struct ip_tunnel_key key; 51 const void *options; 52 u8 options_len; 53 u8 mode; 54 }; 55 56 /* 6rd prefix/relay information */ 57 #ifdef CONFIG_IPV6_SIT_6RD 58 struct ip_tunnel_6rd_parm { 59 struct in6_addr prefix; 60 __be32 relay_prefix; 61 u16 prefixlen; 62 u16 relay_prefixlen; 63 }; 64 #endif 65 66 struct ip_tunnel_encap { 67 __u16 type; 68 __u16 flags; 69 __be16 sport; 70 __be16 dport; 71 }; 72 73 struct ip_tunnel_prl_entry { 74 struct ip_tunnel_prl_entry __rcu *next; 75 __be32 addr; 76 u16 flags; 77 struct rcu_head rcu_head; 78 }; 79 80 struct ip_tunnel_dst { 81 struct dst_entry __rcu *dst; 82 __be32 saddr; 83 }; 84 85 struct ip_tunnel { 86 struct ip_tunnel __rcu *next; 87 struct hlist_node hash_node; 88 struct net_device *dev; 89 struct net *net; /* netns for packet i/o */ 90 91 int err_count; /* Number of arrived ICMP errors */ 92 unsigned long err_time; /* Time when the last ICMP error 93 * arrived */ 94 95 /* These four fields used only by GRE */ 96 __u32 i_seqno; /* The last seen seqno */ 97 __u32 o_seqno; /* The last output seqno */ 98 int tun_hlen; /* Precalculated header length */ 99 int mlink; 100 101 struct ip_tunnel_dst __percpu *dst_cache; 102 103 struct ip_tunnel_parm parms; 104 105 int encap_hlen; /* Encap header length (FOU,GUE) */ 106 struct ip_tunnel_encap encap; 107 108 int hlen; /* tun_hlen + encap_hlen */ 109 110 /* for SIT */ 111 #ifdef CONFIG_IPV6_SIT_6RD 112 struct ip_tunnel_6rd_parm ip6rd; 113 #endif 114 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */ 115 unsigned int prl_count; /* # of entries in PRL */ 116 int ip_tnl_net_id; 117 struct gro_cells gro_cells; 118 }; 119 120 #define TUNNEL_CSUM __cpu_to_be16(0x01) 121 #define TUNNEL_ROUTING __cpu_to_be16(0x02) 122 #define TUNNEL_KEY __cpu_to_be16(0x04) 123 #define TUNNEL_SEQ __cpu_to_be16(0x08) 124 #define TUNNEL_STRICT __cpu_to_be16(0x10) 125 #define TUNNEL_REC __cpu_to_be16(0x20) 126 #define TUNNEL_VERSION __cpu_to_be16(0x40) 127 #define TUNNEL_NO_KEY __cpu_to_be16(0x80) 128 #define TUNNEL_DONT_FRAGMENT __cpu_to_be16(0x0100) 129 #define TUNNEL_OAM __cpu_to_be16(0x0200) 130 #define TUNNEL_CRIT_OPT __cpu_to_be16(0x0400) 131 #define TUNNEL_GENEVE_OPT __cpu_to_be16(0x0800) 132 #define TUNNEL_VXLAN_OPT __cpu_to_be16(0x1000) 133 134 #define TUNNEL_OPTIONS_PRESENT (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT) 135 136 struct tnl_ptk_info { 137 __be16 flags; 138 __be16 proto; 139 __be32 key; 140 __be32 seq; 141 }; 142 143 #define PACKET_RCVD 0 144 #define PACKET_REJECT 1 145 146 #define IP_TNL_HASH_BITS 7 147 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS) 148 149 struct ip_tunnel_net { 150 struct net_device *fb_tunnel_dev; 151 struct hlist_head tunnels[IP_TNL_HASH_SIZE]; 152 }; 153 154 struct ip_tunnel_encap_ops { 155 size_t (*encap_hlen)(struct ip_tunnel_encap *e); 156 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e, 157 u8 *protocol, struct flowi4 *fl4); 158 }; 159 160 #define MAX_IPTUN_ENCAP_OPS 8 161 162 extern const struct ip_tunnel_encap_ops __rcu * 163 iptun_encaps[MAX_IPTUN_ENCAP_OPS]; 164 165 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op, 166 unsigned int num); 167 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op, 168 unsigned int num); 169 170 static inline void __ip_tunnel_info_init(struct ip_tunnel_info *tun_info, 171 __be32 saddr, __be32 daddr, 172 u8 tos, u8 ttl, 173 __be16 tp_src, __be16 tp_dst, 174 __be64 tun_id, __be16 tun_flags, 175 const void *opts, u8 opts_len) 176 { 177 tun_info->key.tun_id = tun_id; 178 tun_info->key.ipv4_src = saddr; 179 tun_info->key.ipv4_dst = daddr; 180 tun_info->key.ipv4_tos = tos; 181 tun_info->key.ipv4_ttl = ttl; 182 tun_info->key.tun_flags = tun_flags; 183 184 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of 185 * the upper tunnel are used. 186 * E.g: GRE over IPSEC, the tp_src and tp_port are zero. 187 */ 188 tun_info->key.tp_src = tp_src; 189 tun_info->key.tp_dst = tp_dst; 190 191 /* Clear struct padding. */ 192 if (sizeof(tun_info->key) != IP_TUNNEL_KEY_SIZE) 193 memset((unsigned char *)&tun_info->key + IP_TUNNEL_KEY_SIZE, 194 0, sizeof(tun_info->key) - IP_TUNNEL_KEY_SIZE); 195 196 tun_info->options = opts; 197 tun_info->options_len = opts_len; 198 } 199 200 static inline void ip_tunnel_info_init(struct ip_tunnel_info *tun_info, 201 const struct iphdr *iph, 202 __be16 tp_src, __be16 tp_dst, 203 __be64 tun_id, __be16 tun_flags, 204 const void *opts, u8 opts_len) 205 { 206 __ip_tunnel_info_init(tun_info, iph->saddr, iph->daddr, 207 iph->tos, iph->ttl, tp_src, tp_dst, 208 tun_id, tun_flags, opts, opts_len); 209 } 210 211 #ifdef CONFIG_INET 212 213 int ip_tunnel_init(struct net_device *dev); 214 void ip_tunnel_uninit(struct net_device *dev); 215 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head); 216 struct net *ip_tunnel_get_link_net(const struct net_device *dev); 217 int ip_tunnel_get_iflink(const struct net_device *dev); 218 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id, 219 struct rtnl_link_ops *ops, char *devname); 220 221 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops); 222 223 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, 224 const struct iphdr *tnl_params, const u8 protocol); 225 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd); 226 int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t, 227 u8 *protocol, struct flowi4 *fl4); 228 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu); 229 230 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev, 231 struct rtnl_link_stats64 *tot); 232 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn, 233 int link, __be16 flags, 234 __be32 remote, __be32 local, 235 __be32 key); 236 237 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb, 238 const struct tnl_ptk_info *tpi, bool log_ecn_error); 239 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[], 240 struct ip_tunnel_parm *p); 241 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[], 242 struct ip_tunnel_parm *p); 243 void ip_tunnel_setup(struct net_device *dev, int net_id); 244 void ip_tunnel_dst_reset_all(struct ip_tunnel *t); 245 int ip_tunnel_encap_setup(struct ip_tunnel *t, 246 struct ip_tunnel_encap *ipencap); 247 248 /* Extract dsfield from inner protocol */ 249 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph, 250 const struct sk_buff *skb) 251 { 252 if (skb->protocol == htons(ETH_P_IP)) 253 return iph->tos; 254 else if (skb->protocol == htons(ETH_P_IPV6)) 255 return ipv6_get_dsfield((const struct ipv6hdr *)iph); 256 else 257 return 0; 258 } 259 260 /* Propogate ECN bits out */ 261 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph, 262 const struct sk_buff *skb) 263 { 264 u8 inner = ip_tunnel_get_dsfield(iph, skb); 265 266 return INET_ECN_encapsulate(tos, inner); 267 } 268 269 int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, __be16 inner_proto); 270 int iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, 271 __be32 src, __be32 dst, __u8 proto, 272 __u8 tos, __u8 ttl, __be16 df, bool xnet); 273 274 struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb, bool gre_csum, 275 int gso_type_mask); 276 277 static inline void iptunnel_xmit_stats(int err, 278 struct net_device_stats *err_stats, 279 struct pcpu_sw_netstats __percpu *stats) 280 { 281 if (err > 0) { 282 struct pcpu_sw_netstats *tstats = this_cpu_ptr(stats); 283 284 u64_stats_update_begin(&tstats->syncp); 285 tstats->tx_bytes += err; 286 tstats->tx_packets++; 287 u64_stats_update_end(&tstats->syncp); 288 } else if (err < 0) { 289 err_stats->tx_errors++; 290 err_stats->tx_aborted_errors++; 291 } else { 292 err_stats->tx_dropped++; 293 } 294 } 295 296 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info, size_t n) 297 { 298 return info + 1; 299 } 300 301 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 302 { 303 return (struct ip_tunnel_info *)lwtstate->data; 304 } 305 306 extern struct static_key ip_tunnel_metadata_cnt; 307 308 /* Returns > 0 if metadata should be collected */ 309 static inline int ip_tunnel_collect_metadata(void) 310 { 311 return static_key_false(&ip_tunnel_metadata_cnt); 312 } 313 314 void __init ip_tunnel_core_init(void); 315 316 void ip_tunnel_need_metadata(void); 317 void ip_tunnel_unneed_metadata(void); 318 319 #else /* CONFIG_INET */ 320 321 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate) 322 { 323 return NULL; 324 } 325 326 static inline void ip_tunnel_need_metadata(void) 327 { 328 } 329 330 static inline void ip_tunnel_unneed_metadata(void) 331 { 332 } 333 334 #endif /* CONFIG_INET */ 335 336 #endif /* __NET_IP_TUNNELS_H */ 337