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