include/net/ip_tunnels.h

   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 /* A fan overlay /8 (250.0.0.0/8, for example) maps to exactly one /16
 103  * underlay (10.88.0.0/16, for example).  Multiple local addresses within
 104  * the /16 may be used, but a particular overlay may not span
 105  * multiple underlay subnets.
 106  *
 107  * We store one underlay, indexed by the overlay's high order octet.
 108  */
 109 #define FAN_OVERLAY_CNT         256
 110
 111 struct ip_fan_map {
 112         __be32                  underlay;
 113         __be32                  overlay;
 114         u16                     underlay_prefix;
 115         u16                     overlay_prefix;
 116         u32                     overlay_mask;
 117         struct list_head        list;
 118         struct rcu_head         rcu;
 119 };
 120
 121 struct ip_tunnel_fan {
 122         struct list_head        fan_maps;
 123 };
 124
 125 struct ip_tunnel {
 126         struct ip_tunnel __rcu  *next;
 127         struct hlist_node hash_node;
 128         struct net_device       *dev;
 129         struct net              *net;   /* netns for packet i/o */
 130
 131         unsigned long   err_time;       /* Time when the last ICMP error
 132                                          * arrived */
 133         int             err_count;      /* Number of arrived ICMP errors */
 134
 135         /* These four fields used only by GRE */
 136         u32             i_seqno;        /* The last seen seqno  */
 137         u32             o_seqno;        /* The last output seqno */
 138         int             tun_hlen;       /* Precalculated header length */
 139
 140         struct dst_cache dst_cache;
 141
 142         struct ip_tunnel_parm parms;
 143
 144         int             mlink;
 145         int             encap_hlen;     /* Encap header length (FOU,GUE) */
 146         int             hlen;           /* tun_hlen + encap_hlen */
 147         struct ip_tunnel_encap encap;
 148
 149         /* for SIT */
 150 #ifdef CONFIG_IPV6_SIT_6RD
 151         struct ip_tunnel_6rd_parm ip6rd;
 152 #endif
 153         struct ip_tunnel_prl_entry __rcu *prl;  /* potential router list */
 154         unsigned int            prl_count;      /* # of entries in PRL */
 155         struct ip_tunnel_fan    fan;
 156         unsigned int            ip_tnl_net_id;
 157         struct gro_cells        gro_cells;
 158         __u32                   fwmark;
 159         bool                    collect_md;
 160         bool                    ignore_df;
 161 };
 162
 163 #define TUNNEL_CSUM             __cpu_to_be16(0x01)
 164 #define TUNNEL_ROUTING          __cpu_to_be16(0x02)
 165 #define TUNNEL_KEY              __cpu_to_be16(0x04)
 166 #define TUNNEL_SEQ              __cpu_to_be16(0x08)
 167 #define TUNNEL_STRICT           __cpu_to_be16(0x10)
 168 #define TUNNEL_REC              __cpu_to_be16(0x20)
 169 #define TUNNEL_VERSION          __cpu_to_be16(0x40)
 170 #define TUNNEL_NO_KEY           __cpu_to_be16(0x80)
 171 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
 172 #define TUNNEL_OAM              __cpu_to_be16(0x0200)
 173 #define TUNNEL_CRIT_OPT         __cpu_to_be16(0x0400)
 174 #define TUNNEL_GENEVE_OPT       __cpu_to_be16(0x0800)
 175 #define TUNNEL_VXLAN_OPT        __cpu_to_be16(0x1000)
 176 #define TUNNEL_NOCACHE          __cpu_to_be16(0x2000)
 177
 178 #define TUNNEL_OPTIONS_PRESENT  (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT)
 179
 180 static inline int fan_has_map(const struct ip_tunnel_fan *fan)
 181 {
 182         return !list_empty(&fan->fan_maps);
 183 }
 184
 185 struct tnl_ptk_info {
 186         __be16 flags;
 187         __be16 proto;
 188         __be32 key;
 189         __be32 seq;
 190         int hdr_len;
 191 };
 192
 193 #define PACKET_RCVD     0
 194 #define PACKET_REJECT   1
 195 #define PACKET_NEXT     2
 196
 197 #define IP_TNL_HASH_BITS   7
 198 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
 199
 200 struct ip_tunnel_net {
 201         struct net_device *fb_tunnel_dev;
 202         struct hlist_head tunnels[IP_TNL_HASH_SIZE];
 203         struct ip_tunnel __rcu *collect_md_tun;
 204 };
 205
 206 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
 207                                       __be32 saddr, __be32 daddr,
 208                                       u8 tos, u8 ttl, __be32 label,
 209                                       __be16 tp_src, __be16 tp_dst,
 210                                       __be64 tun_id, __be16 tun_flags)
 211 {
 212         key->tun_id = tun_id;
 213         key->u.ipv4.src = saddr;
 214         key->u.ipv4.dst = daddr;
 215         memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
 216                0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
 217         key->tos = tos;
 218         key->ttl = ttl;
 219         key->label = label;
 220         key->tun_flags = tun_flags;
 221
 222         /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
 223          * the upper tunnel are used.
 224          * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
 225          */
 226         key->tp_src = tp_src;
 227         key->tp_dst = tp_dst;
 228
 229         /* Clear struct padding. */
 230         if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
 231                 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
 232                        0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
 233 }
 234
 235 static inline bool
 236 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
 237                            const struct ip_tunnel_info *info)
 238 {
 239         if (skb->mark)
 240                 return false;
 241         if (!info)
 242                 return true;
 243         if (info->key.tun_flags & TUNNEL_NOCACHE)
 244                 return false;
 245
 246         return true;
 247 }
 248
 249 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
 250                                                *tun_info)
 251 {
 252         return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
 253 }
 254
 255 static inline __be64 key32_to_tunnel_id(__be32 key)
 256 {
 257 #ifdef __BIG_ENDIAN
 258         return (__force __be64)key;
 259 #else
 260         return (__force __be64)((__force u64)key << 32);
 261 #endif
 262 }
 263
 264 /* Returns the least-significant 32 bits of a __be64. */
 265 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
 266 {
 267 #ifdef __BIG_ENDIAN
 268         return (__force __be32)tun_id;
 269 #else
 270         return (__force __be32)((__force u64)tun_id >> 32);
 271 #endif
 272 }
 273
 274 #ifdef CONFIG_INET
 275
 276 int ip_tunnel_init(struct net_device *dev);
 277 void ip_tunnel_uninit(struct net_device *dev);
 278 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
 279 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
 280 int ip_tunnel_get_iflink(const struct net_device *dev);
 281 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
 282                        struct rtnl_link_ops *ops, char *devname);
 283
 284 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, 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 };
 315
 316 #define MAX_IPTUN_ENCAP_OPS 8
 317
 318 extern const struct ip_tunnel_encap_ops __rcu *
 319                 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
 320
 321 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
 322                             unsigned int num);
 323 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
 324                             unsigned int num);
 325
 326 int ip_tunnel_encap_setup(struct ip_tunnel *t,
 327                           struct ip_tunnel_encap *ipencap);
 328
 329 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
 330 {
 331         const struct ip_tunnel_encap_ops *ops;
 332         int hlen = -EINVAL;
 333
 334         if (e->type == TUNNEL_ENCAP_NONE)
 335                 return 0;
 336
 337         if (e->type >= MAX_IPTUN_ENCAP_OPS)
 338                 return -EINVAL;
 339
 340         rcu_read_lock();
 341         ops = rcu_dereference(iptun_encaps[e->type]);
 342         if (likely(ops && ops->encap_hlen))
 343                 hlen = ops->encap_hlen(e);
 344         rcu_read_unlock();
 345
 346         return hlen;
 347 }
 348
 349 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
 350                                   u8 *protocol, struct flowi4 *fl4)
 351 {
 352         const struct ip_tunnel_encap_ops *ops;
 353         int ret = -EINVAL;
 354
 355         if (t->encap.type == TUNNEL_ENCAP_NONE)
 356                 return 0;
 357
 358         if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
 359                 return -EINVAL;
 360
 361         rcu_read_lock();
 362         ops = rcu_dereference(iptun_encaps[t->encap.type]);
 363         if (likely(ops && ops->build_header))
 364                 ret = ops->build_header(skb, &t->encap, protocol, fl4);
 365         rcu_read_unlock();
 366
 367         return ret;
 368 }
 369
 370 /* Extract dsfield from inner protocol */
 371 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
 372                                        const struct sk_buff *skb)
 373 {
 374         if (skb->protocol == htons(ETH_P_IP))
 375                 return iph->tos;
 376         else if (skb->protocol == htons(ETH_P_IPV6))
 377                 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
 378         else
 379                 return 0;
 380 }
 381
 382 /* Propogate ECN bits out */
 383 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
 384                                      const struct sk_buff *skb)
 385 {
 386         u8 inner = ip_tunnel_get_dsfield(iph, skb);
 387
 388         return INET_ECN_encapsulate(tos, inner);
 389 }
 390
 391 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 392                            __be16 inner_proto, bool raw_proto, bool xnet);
 393
 394 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 395                                        __be16 inner_proto, bool xnet)
 396 {
 397         return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
 398 }
 399
 400 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
 401                    __be32 src, __be32 dst, u8 proto,
 402                    u8 tos, u8 ttl, __be16 df, bool xnet);
 403 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 404                                              gfp_t flags);
 405
 406 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
 407
 408 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
 409 {
 410         if (skb_is_gso(skb)) {
 411                 int err;
 412
 413                 err = skb_unclone(skb, GFP_ATOMIC);
 414                 if (unlikely(err))
 415                         return err;
 416                 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
 417                                                NETIF_F_GSO_SHIFT);
 418         }
 419
 420         skb->encapsulation = 0;
 421         return 0;
 422 }
 423
 424 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
 425 {
 426         if (pkt_len > 0) {
 427                 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
 428
 429                 u64_stats_update_begin(&tstats->syncp);
 430                 tstats->tx_bytes += pkt_len;
 431                 tstats->tx_packets++;
 432                 u64_stats_update_end(&tstats->syncp);
 433                 put_cpu_ptr(tstats);
 434         } else {
 435                 struct net_device_stats *err_stats = &dev->stats;
 436
 437                 if (pkt_len < 0) {
 438                         err_stats->tx_errors++;
 439                         err_stats->tx_aborted_errors++;
 440                 } else {
 441                         err_stats->tx_dropped++;
 442                 }
 443         }
 444 }
 445
 446 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
 447 {
 448         return info + 1;
 449 }
 450
 451 static inline void ip_tunnel_info_opts_get(void *to,
 452                                            const struct ip_tunnel_info *info)
 453 {
 454         memcpy(to, info + 1, info->options_len);
 455 }
 456
 457 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 458                                            const void *from, int len)
 459 {
 460         memcpy(ip_tunnel_info_opts(info), from, len);
 461         info->options_len = len;
 462 }
 463
 464 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 465 {
 466         return (struct ip_tunnel_info *)lwtstate->data;
 467 }
 468
 469 extern struct static_key ip_tunnel_metadata_cnt;
 470
 471 /* Returns > 0 if metadata should be collected */
 472 static inline int ip_tunnel_collect_metadata(void)
 473 {
 474         return static_key_false(&ip_tunnel_metadata_cnt);
 475 }
 476
 477 void __init ip_tunnel_core_init(void);
 478
 479 void ip_tunnel_need_metadata(void);
 480 void ip_tunnel_unneed_metadata(void);
 481
 482 #else /* CONFIG_INET */
 483
 484 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 485 {
 486         return NULL;
 487 }
 488
 489 static inline void ip_tunnel_need_metadata(void)
 490 {
 491 }
 492
 493 static inline void ip_tunnel_unneed_metadata(void)
 494 {
 495 }
 496
 497 static inline void ip_tunnel_info_opts_get(void *to,
 498                                            const struct ip_tunnel_info *info)
 499 {
 500 }
 501
 502 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 503                                            const void *from, int len)
 504 {
 505         info->options_len = 0;
 506 }
 507
 508 #endif /* CONFIG_INET */
 509
 510 #endif /* __NET_IP_TUNNELS_H */