include/net/ip_tunnels.h

   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 /* A fan overlay /8 (250.0.0.0/8, for example) maps to exactly one /16
 104  * underlay (10.88.0.0/16, for example).  Multiple local addresses within
 105  * the /16 may be used, but a particular overlay may not span
 106  * multiple underlay subnets.
 107  *
 108  * We store one underlay, indexed by the overlay's high order octet.
 109  */
 110 #define FAN_OVERLAY_CNT         256
 111
 112 struct ip_fan_map {
 113         __be32                  underlay;
 114         __be32                  overlay;
 115         u16                     underlay_prefix;
 116         u16                     overlay_prefix;
 117         u32                     overlay_mask;
 118         struct list_head        list;
 119         struct rcu_head         rcu;
 120 };
 121
 122 struct ip_tunnel_fan {
 123         struct list_head        fan_maps;
 124 };
 125
 126 struct ip_tunnel {
 127         struct ip_tunnel __rcu  *next;
 128         struct hlist_node hash_node;
 129         struct net_device       *dev;
 130         struct net              *net;   /* netns for packet i/o */
 131
 132         unsigned long   err_time;       /* Time when the last ICMP error
 133                                          * arrived */
 134         int             err_count;      /* Number of arrived ICMP errors */
 135
 136         /* These four fields used only by GRE */
 137         u32             i_seqno;        /* The last seen seqno  */
 138         u32             o_seqno;        /* The last output seqno */
 139         int             tun_hlen;       /* Precalculated header length */
 140
 141         /* This field used only by ERSPAN */
 142         u32             index;          /* ERSPAN type II index */
 143
 144         struct dst_cache dst_cache;
 145
 146         struct ip_tunnel_parm parms;
 147
 148         int             mlink;
 149         int             encap_hlen;     /* Encap header length (FOU,GUE) */
 150         int             hlen;           /* tun_hlen + encap_hlen */
 151         struct ip_tunnel_encap encap;
 152
 153         /* for SIT */
 154 #ifdef CONFIG_IPV6_SIT_6RD
 155         struct ip_tunnel_6rd_parm ip6rd;
 156 #endif
 157         struct ip_tunnel_prl_entry __rcu *prl;  /* potential router list */
 158         unsigned int            prl_count;      /* # of entries in PRL */
 159         struct ip_tunnel_fan    fan;
 160         unsigned int            ip_tnl_net_id;
 161         struct gro_cells        gro_cells;
 162         __u32                   fwmark;
 163         bool                    collect_md;
 164         bool                    ignore_df;
 165 };
 166
 167 #define TUNNEL_CSUM             __cpu_to_be16(0x01)
 168 #define TUNNEL_ROUTING          __cpu_to_be16(0x02)
 169 #define TUNNEL_KEY              __cpu_to_be16(0x04)
 170 #define TUNNEL_SEQ              __cpu_to_be16(0x08)
 171 #define TUNNEL_STRICT           __cpu_to_be16(0x10)
 172 #define TUNNEL_REC              __cpu_to_be16(0x20)
 173 #define TUNNEL_VERSION          __cpu_to_be16(0x40)
 174 #define TUNNEL_NO_KEY           __cpu_to_be16(0x80)
 175 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
 176 #define TUNNEL_OAM              __cpu_to_be16(0x0200)
 177 #define TUNNEL_CRIT_OPT         __cpu_to_be16(0x0400)
 178 #define TUNNEL_GENEVE_OPT       __cpu_to_be16(0x0800)
 179 #define TUNNEL_VXLAN_OPT        __cpu_to_be16(0x1000)
 180 #define TUNNEL_NOCACHE          __cpu_to_be16(0x2000)
 181 #define TUNNEL_ERSPAN_OPT       __cpu_to_be16(0x4000)
 182
 183 #define TUNNEL_OPTIONS_PRESENT \
 184                 (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT | TUNNEL_ERSPAN_OPT)
 185
 186 static inline int fan_has_map(const struct ip_tunnel_fan *fan)
 187 {
 188         return !list_empty(&fan->fan_maps);
 189 }
 190
 191 struct tnl_ptk_info {
 192         __be16 flags;
 193         __be16 proto;
 194         __be32 key;
 195         __be32 seq;
 196         int hdr_len;
 197 };
 198
 199 #define PACKET_RCVD     0
 200 #define PACKET_REJECT   1
 201 #define PACKET_NEXT     2
 202
 203 #define IP_TNL_HASH_BITS   7
 204 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
 205
 206 struct ip_tunnel_net {
 207         struct net_device *fb_tunnel_dev;
 208         struct hlist_head tunnels[IP_TNL_HASH_SIZE];
 209         struct ip_tunnel __rcu *collect_md_tun;
 210 };
 211
 212 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
 213                                       __be32 saddr, __be32 daddr,
 214                                       u8 tos, u8 ttl, __be32 label,
 215                                       __be16 tp_src, __be16 tp_dst,
 216                                       __be64 tun_id, __be16 tun_flags)
 217 {
 218         key->tun_id = tun_id;
 219         key->u.ipv4.src = saddr;
 220         key->u.ipv4.dst = daddr;
 221         memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
 222                0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
 223         key->tos = tos;
 224         key->ttl = ttl;
 225         key->label = label;
 226         key->tun_flags = tun_flags;
 227
 228         /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
 229          * the upper tunnel are used.
 230          * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
 231          */
 232         key->tp_src = tp_src;
 233         key->tp_dst = tp_dst;
 234
 235         /* Clear struct padding. */
 236         if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
 237                 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
 238                        0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
 239 }
 240
 241 static inline bool
 242 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
 243                            const struct ip_tunnel_info *info)
 244 {
 245         if (skb->mark)
 246                 return false;
 247         if (!info)
 248                 return true;
 249         if (info->key.tun_flags & TUNNEL_NOCACHE)
 250                 return false;
 251
 252         return true;
 253 }
 254
 255 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
 256                                                *tun_info)
 257 {
 258         return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
 259 }
 260
 261 static inline __be64 key32_to_tunnel_id(__be32 key)
 262 {
 263 #ifdef __BIG_ENDIAN
 264         return (__force __be64)key;
 265 #else
 266         return (__force __be64)((__force u64)key << 32);
 267 #endif
 268 }
 269
 270 /* Returns the least-significant 32 bits of a __be64. */
 271 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
 272 {
 273 #ifdef __BIG_ENDIAN
 274         return (__force __be32)tun_id;
 275 #else
 276         return (__force __be32)((__force u64)tun_id >> 32);
 277 #endif
 278 }
 279
 280 #ifdef CONFIG_INET
 281
 282 int ip_tunnel_init(struct net_device *dev);
 283 void ip_tunnel_uninit(struct net_device *dev);
 284 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
 285 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
 286 int ip_tunnel_get_iflink(const struct net_device *dev);
 287 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
 288                        struct rtnl_link_ops *ops, char *devname);
 289
 290 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
 291                            struct rtnl_link_ops *ops);
 292
 293 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 294                     const struct iphdr *tnl_params, const u8 protocol);
 295 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 296                        const u8 proto);
 297 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
 298 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
 299 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
 300
 301 void ip_tunnel_get_stats64(struct net_device *dev,
 302                            struct rtnl_link_stats64 *tot);
 303 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 304                                    int link, __be16 flags,
 305                                    __be32 remote, __be32 local,
 306                                    __be32 key);
 307
 308 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
 309                   const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
 310                   bool log_ecn_error);
 311 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
 312                          struct ip_tunnel_parm *p, __u32 fwmark);
 313 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
 314                       struct ip_tunnel_parm *p, __u32 fwmark);
 315 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
 316
 317 struct ip_tunnel_encap_ops {
 318         size_t (*encap_hlen)(struct ip_tunnel_encap *e);
 319         int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
 320                             u8 *protocol, struct flowi4 *fl4);
 321 };
 322
 323 #define MAX_IPTUN_ENCAP_OPS 8
 324
 325 extern const struct ip_tunnel_encap_ops __rcu *
 326                 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
 327
 328 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
 329                             unsigned int num);
 330 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
 331                             unsigned int num);
 332
 333 int ip_tunnel_encap_setup(struct ip_tunnel *t,
 334                           struct ip_tunnel_encap *ipencap);
 335
 336 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
 337 {
 338         int nhlen;
 339
 340         switch (skb->protocol) {
 341 #if IS_ENABLED(CONFIG_IPV6)
 342         case htons(ETH_P_IPV6):
 343                 nhlen = sizeof(struct ipv6hdr);
 344                 break;
 345 #endif
 346         case htons(ETH_P_IP):
 347                 nhlen = sizeof(struct iphdr);
 348                 break;
 349         default:
 350                 nhlen = 0;
 351         }
 352
 353         return pskb_network_may_pull(skb, nhlen);
 354 }
 355
 356 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
 357 {
 358         const struct ip_tunnel_encap_ops *ops;
 359         int hlen = -EINVAL;
 360
 361         if (e->type == TUNNEL_ENCAP_NONE)
 362                 return 0;
 363
 364         if (e->type >= MAX_IPTUN_ENCAP_OPS)
 365                 return -EINVAL;
 366
 367         rcu_read_lock();
 368         ops = rcu_dereference(iptun_encaps[e->type]);
 369         if (likely(ops && ops->encap_hlen))
 370                 hlen = ops->encap_hlen(e);
 371         rcu_read_unlock();
 372
 373         return hlen;
 374 }
 375
 376 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
 377                                   u8 *protocol, struct flowi4 *fl4)
 378 {
 379         const struct ip_tunnel_encap_ops *ops;
 380         int ret = -EINVAL;
 381
 382         if (t->encap.type == TUNNEL_ENCAP_NONE)
 383                 return 0;
 384
 385         if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
 386                 return -EINVAL;
 387
 388         rcu_read_lock();
 389         ops = rcu_dereference(iptun_encaps[t->encap.type]);
 390         if (likely(ops && ops->build_header))
 391                 ret = ops->build_header(skb, &t->encap, protocol, fl4);
 392         rcu_read_unlock();
 393
 394         return ret;
 395 }
 396
 397 /* Extract dsfield from inner protocol */
 398 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
 399                                        const struct sk_buff *skb)
 400 {
 401         if (skb->protocol == htons(ETH_P_IP))
 402                 return iph->tos;
 403         else if (skb->protocol == htons(ETH_P_IPV6))
 404                 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
 405         else
 406                 return 0;
 407 }
 408
 409 /* Propogate ECN bits out */
 410 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
 411                                      const struct sk_buff *skb)
 412 {
 413         u8 inner = ip_tunnel_get_dsfield(iph, skb);
 414
 415         return INET_ECN_encapsulate(tos, inner);
 416 }
 417
 418 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 419                            __be16 inner_proto, bool raw_proto, bool xnet);
 420
 421 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 422                                        __be16 inner_proto, bool xnet)
 423 {
 424         return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
 425 }
 426
 427 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
 428                    __be32 src, __be32 dst, u8 proto,
 429                    u8 tos, u8 ttl, __be16 df, bool xnet);
 430 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 431                                              gfp_t flags);
 432
 433 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
 434
 435 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
 436 {
 437         if (skb_is_gso(skb)) {
 438                 int err;
 439
 440                 err = skb_unclone(skb, GFP_ATOMIC);
 441                 if (unlikely(err))
 442                         return err;
 443                 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
 444                                                NETIF_F_GSO_SHIFT);
 445         }
 446
 447         skb->encapsulation = 0;
 448         return 0;
 449 }
 450
 451 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
 452 {
 453         if (pkt_len > 0) {
 454                 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
 455
 456                 u64_stats_update_begin(&tstats->syncp);
 457                 tstats->tx_bytes += pkt_len;
 458                 tstats->tx_packets++;
 459                 u64_stats_update_end(&tstats->syncp);
 460                 put_cpu_ptr(tstats);
 461         } else {
 462                 struct net_device_stats *err_stats = &dev->stats;
 463
 464                 if (pkt_len < 0) {
 465                         err_stats->tx_errors++;
 466                         err_stats->tx_aborted_errors++;
 467                 } else {
 468                         err_stats->tx_dropped++;
 469                 }
 470         }
 471 }
 472
 473 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
 474 {
 475         return info + 1;
 476 }
 477
 478 static inline void ip_tunnel_info_opts_get(void *to,
 479                                            const struct ip_tunnel_info *info)
 480 {
 481         memcpy(to, info + 1, info->options_len);
 482 }
 483
 484 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 485                                            const void *from, int len)
 486 {
 487         memcpy(ip_tunnel_info_opts(info), from, len);
 488         info->options_len = len;
 489 }
 490
 491 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 492 {
 493         return (struct ip_tunnel_info *)lwtstate->data;
 494 }
 495
 496 extern struct static_key ip_tunnel_metadata_cnt;
 497
 498 /* Returns > 0 if metadata should be collected */
 499 static inline int ip_tunnel_collect_metadata(void)
 500 {
 501         return static_key_false(&ip_tunnel_metadata_cnt);
 502 }
 503
 504 void __init ip_tunnel_core_init(void);
 505
 506 void ip_tunnel_need_metadata(void);
 507 void ip_tunnel_unneed_metadata(void);
 508
 509 #else /* CONFIG_INET */
 510
 511 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 512 {
 513         return NULL;
 514 }
 515
 516 static inline void ip_tunnel_need_metadata(void)
 517 {
 518 }
 519
 520 static inline void ip_tunnel_unneed_metadata(void)
 521 {
 522 }
 523
 524 static inline void ip_tunnel_info_opts_get(void *to,
 525                                            const struct ip_tunnel_info *info)
 526 {
 527 }
 528
 529 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 530                                            const void *from, int len)
 531 {
 532         info->options_len = 0;
 533 }
 534
 535 #endif /* CONFIG_INET */
 536
 537 #endif /* __NET_IP_TUNNELS_H */