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