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         /* These four fields used only by ERSPAN */
 142         u32             index;          /* ERSPAN type II index */
 143         u8              erspan_ver;     /* ERSPAN version */
 144         u8              dir;            /* ERSPAN direction */
 145         u16             hwid;           /* ERSPAN hardware ID */
 146
 147         struct dst_cache dst_cache;
 148
 149         struct ip_tunnel_parm parms;
 150
 151         int             mlink;
 152         int             encap_hlen;     /* Encap header length (FOU,GUE) */
 153         int             hlen;           /* tun_hlen + encap_hlen */
 154         struct ip_tunnel_encap encap;
 155
 156         /* for SIT */
 157 #ifdef CONFIG_IPV6_SIT_6RD
 158         struct ip_tunnel_6rd_parm ip6rd;
 159 #endif
 160         struct ip_tunnel_prl_entry __rcu *prl;  /* potential router list */
 161         unsigned int            prl_count;      /* # of entries in PRL */
 162         struct ip_tunnel_fan    fan;
 163         unsigned int            ip_tnl_net_id;
 164         struct gro_cells        gro_cells;
 165         __u32                   fwmark;
 166         bool                    collect_md;
 167         bool                    ignore_df;
 168 };
 169
 170 static inline int fan_has_map(const struct ip_tunnel_fan *fan)
 171 {
 172         return !list_empty(&fan->fan_maps);
 173 }
 174
 175 struct tnl_ptk_info {
 176         __be16 flags;
 177         __be16 proto;
 178         __be32 key;
 179         __be32 seq;
 180         int hdr_len;
 181 };
 182
 183 #define PACKET_RCVD     0
 184 #define PACKET_REJECT   1
 185 #define PACKET_NEXT     2
 186
 187 #define IP_TNL_HASH_BITS   7
 188 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
 189
 190 struct ip_tunnel_net {
 191         struct net_device *fb_tunnel_dev;
 192         struct rtnl_link_ops *rtnl_link_ops;
 193         struct hlist_head tunnels[IP_TNL_HASH_SIZE];
 194         struct ip_tunnel __rcu *collect_md_tun;
 195         int type;
 196 };
 197
 198 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
 199                                       __be32 saddr, __be32 daddr,
 200                                       u8 tos, u8 ttl, __be32 label,
 201                                       __be16 tp_src, __be16 tp_dst,
 202                                       __be64 tun_id, __be16 tun_flags)
 203 {
 204         key->tun_id = tun_id;
 205         key->u.ipv4.src = saddr;
 206         key->u.ipv4.dst = daddr;
 207         memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
 208                0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
 209         key->tos = tos;
 210         key->ttl = ttl;
 211         key->label = label;
 212         key->tun_flags = tun_flags;
 213
 214         /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
 215          * the upper tunnel are used.
 216          * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
 217          */
 218         key->tp_src = tp_src;
 219         key->tp_dst = tp_dst;
 220
 221         /* Clear struct padding. */
 222         if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
 223                 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
 224                        0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
 225 }
 226
 227 static inline bool
 228 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
 229                            const struct ip_tunnel_info *info)
 230 {
 231         if (skb->mark)
 232                 return false;
 233         if (!info)
 234                 return true;
 235         if (info->key.tun_flags & TUNNEL_NOCACHE)
 236                 return false;
 237
 238         return true;
 239 }
 240
 241 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
 242                                                *tun_info)
 243 {
 244         return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
 245 }
 246
 247 static inline __be64 key32_to_tunnel_id(__be32 key)
 248 {
 249 #ifdef __BIG_ENDIAN
 250         return (__force __be64)key;
 251 #else
 252         return (__force __be64)((__force u64)key << 32);
 253 #endif
 254 }
 255
 256 /* Returns the least-significant 32 bits of a __be64. */
 257 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
 258 {
 259 #ifdef __BIG_ENDIAN
 260         return (__force __be32)tun_id;
 261 #else
 262         return (__force __be32)((__force u64)tun_id >> 32);
 263 #endif
 264 }
 265
 266 #ifdef CONFIG_INET
 267
 268 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
 269                                        int proto,
 270                                        __be32 daddr, __be32 saddr,
 271                                        __be32 key, __u8 tos, int oif,
 272                                        __u32 mark, __u32 tun_inner_hash)
 273 {
 274         memset(fl4, 0, sizeof(*fl4));
 275         fl4->flowi4_oif = oif;
 276         fl4->daddr = daddr;
 277         fl4->saddr = saddr;
 278         fl4->flowi4_tos = tos;
 279         fl4->flowi4_proto = proto;
 280         fl4->fl4_gre_key = key;
 281         fl4->flowi4_mark = mark;
 282         fl4->flowi4_multipath_hash = tun_inner_hash;
 283 }
 284
 285 int ip_tunnel_init(struct net_device *dev);
 286 void ip_tunnel_uninit(struct net_device *dev);
 287 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
 288 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
 289 int ip_tunnel_get_iflink(const struct net_device *dev);
 290 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
 291                        struct rtnl_link_ops *ops, char *devname);
 292
 293 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
 294                            struct rtnl_link_ops *ops);
 295
 296 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 297                     const struct iphdr *tnl_params, const u8 protocol);
 298 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 299                        const u8 proto, int tunnel_hlen);
 300 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
 301 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
 302 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
 303
 304 void ip_tunnel_get_stats64(struct net_device *dev,
 305                            struct rtnl_link_stats64 *tot);
 306 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 307                                    int link, __be16 flags,
 308                                    __be32 remote, __be32 local,
 309                                    __be32 key);
 310
 311 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
 312                   const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
 313                   bool log_ecn_error);
 314 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
 315                          struct ip_tunnel_parm *p, __u32 fwmark);
 316 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
 317                       struct ip_tunnel_parm *p, __u32 fwmark);
 318 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
 319
 320 struct ip_tunnel_encap_ops {
 321         size_t (*encap_hlen)(struct ip_tunnel_encap *e);
 322         int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
 323                             u8 *protocol, struct flowi4 *fl4);
 324         int (*err_handler)(struct sk_buff *skb, u32 info);
 325 };
 326
 327 #define MAX_IPTUN_ENCAP_OPS 8
 328
 329 extern const struct ip_tunnel_encap_ops __rcu *
 330                 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
 331
 332 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
 333                             unsigned int num);
 334 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
 335                             unsigned int num);
 336
 337 int ip_tunnel_encap_setup(struct ip_tunnel *t,
 338                           struct ip_tunnel_encap *ipencap);
 339
 340 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
 341 {
 342         int nhlen;
 343
 344         switch (skb->protocol) {
 345 #if IS_ENABLED(CONFIG_IPV6)
 346         case htons(ETH_P_IPV6):
 347                 nhlen = sizeof(struct ipv6hdr);
 348                 break;
 349 #endif
 350         case htons(ETH_P_IP):
 351                 nhlen = sizeof(struct iphdr);
 352                 break;
 353         default:
 354                 nhlen = 0;
 355         }
 356
 357         return pskb_network_may_pull(skb, nhlen);
 358 }
 359
 360 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
 361 {
 362         const struct ip_tunnel_encap_ops *ops;
 363         int hlen = -EINVAL;
 364
 365         if (e->type == TUNNEL_ENCAP_NONE)
 366                 return 0;
 367
 368         if (e->type >= MAX_IPTUN_ENCAP_OPS)
 369                 return -EINVAL;
 370
 371         rcu_read_lock();
 372         ops = rcu_dereference(iptun_encaps[e->type]);
 373         if (likely(ops && ops->encap_hlen))
 374                 hlen = ops->encap_hlen(e);
 375         rcu_read_unlock();
 376
 377         return hlen;
 378 }
 379
 380 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
 381                                   u8 *protocol, struct flowi4 *fl4)
 382 {
 383         const struct ip_tunnel_encap_ops *ops;
 384         int ret = -EINVAL;
 385
 386         if (t->encap.type == TUNNEL_ENCAP_NONE)
 387                 return 0;
 388
 389         if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
 390                 return -EINVAL;
 391
 392         rcu_read_lock();
 393         ops = rcu_dereference(iptun_encaps[t->encap.type]);
 394         if (likely(ops && ops->build_header))
 395                 ret = ops->build_header(skb, &t->encap, protocol, fl4);
 396         rcu_read_unlock();
 397
 398         return ret;
 399 }
 400
 401 /* Extract dsfield from inner protocol */
 402 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
 403                                        const struct sk_buff *skb)
 404 {
 405         if (skb->protocol == htons(ETH_P_IP))
 406                 return iph->tos;
 407         else if (skb->protocol == htons(ETH_P_IPV6))
 408                 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
 409         else
 410                 return 0;
 411 }
 412
 413 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
 414                                        const struct sk_buff *skb)
 415 {
 416         if (skb->protocol == htons(ETH_P_IP))
 417                 return iph->ttl;
 418         else if (skb->protocol == htons(ETH_P_IPV6))
 419                 return ((const struct ipv6hdr *)iph)->hop_limit;
 420         else
 421                 return 0;
 422 }
 423
 424 /* Propogate ECN bits out */
 425 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
 426                                      const struct sk_buff *skb)
 427 {
 428         u8 inner = ip_tunnel_get_dsfield(iph, skb);
 429
 430         return INET_ECN_encapsulate(tos, inner);
 431 }
 432
 433 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 434                            __be16 inner_proto, bool raw_proto, bool xnet);
 435
 436 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 437                                        __be16 inner_proto, bool xnet)
 438 {
 439         return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
 440 }
 441
 442 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
 443                    __be32 src, __be32 dst, u8 proto,
 444                    u8 tos, u8 ttl, __be16 df, bool xnet);
 445 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 446                                              gfp_t flags);
 447
 448 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
 449
 450 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
 451 {
 452         if (skb_is_gso(skb)) {
 453                 int err;
 454
 455                 err = skb_unclone(skb, GFP_ATOMIC);
 456                 if (unlikely(err))
 457                         return err;
 458                 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
 459                                                NETIF_F_GSO_SHIFT);
 460         }
 461
 462         skb->encapsulation = 0;
 463         return 0;
 464 }
 465
 466 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
 467 {
 468         if (pkt_len > 0) {
 469                 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
 470
 471                 u64_stats_update_begin(&tstats->syncp);
 472                 tstats->tx_bytes += pkt_len;
 473                 tstats->tx_packets++;
 474                 u64_stats_update_end(&tstats->syncp);
 475                 put_cpu_ptr(tstats);
 476         } else {
 477                 struct net_device_stats *err_stats = &dev->stats;
 478
 479                 if (pkt_len < 0) {
 480                         err_stats->tx_errors++;
 481                         err_stats->tx_aborted_errors++;
 482                 } else {
 483                         err_stats->tx_dropped++;
 484                 }
 485         }
 486 }
 487
 488 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
 489 {
 490         return info + 1;
 491 }
 492
 493 static inline void ip_tunnel_info_opts_get(void *to,
 494                                            const struct ip_tunnel_info *info)
 495 {
 496         memcpy(to, info + 1, info->options_len);
 497 }
 498
 499 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 500                                            const void *from, int len,
 501                                            __be16 flags)
 502 {
 503         memcpy(ip_tunnel_info_opts(info), from, len);
 504         info->options_len = len;
 505         info->key.tun_flags |= flags;
 506 }
 507
 508 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 509 {
 510         return (struct ip_tunnel_info *)lwtstate->data;
 511 }
 512
 513 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
 514
 515 /* Returns > 0 if metadata should be collected */
 516 static inline int ip_tunnel_collect_metadata(void)
 517 {
 518         return static_branch_unlikely(&ip_tunnel_metadata_cnt);
 519 }
 520
 521 void __init ip_tunnel_core_init(void);
 522
 523 void ip_tunnel_need_metadata(void);
 524 void ip_tunnel_unneed_metadata(void);
 525
 526 #else /* CONFIG_INET */
 527
 528 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 529 {
 530         return NULL;
 531 }
 532
 533 static inline void ip_tunnel_need_metadata(void)
 534 {
 535 }
 536
 537 static inline void ip_tunnel_unneed_metadata(void)
 538 {
 539 }
 540
 541 static inline void ip_tunnel_info_opts_get(void *to,
 542                                            const struct ip_tunnel_info *info)
 543 {
 544 }
 545
 546 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 547                                            const void *from, int len,
 548                                            __be16 flags)
 549 {
 550         info->options_len = 0;
 551         info->key.tun_flags |= flags;
 552 }
 553
 554 #endif /* CONFIG_INET */
 555
 556 #endif /* __NET_IP_TUNNELS_H */