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
  62 /* Maximum tunnel options length. */
  63 #define IP_TUNNEL_OPTS_MAX                                      \
  64         GENMASK((FIELD_SIZEOF(struct ip_tunnel_info,            \
  65                               options_len) * BITS_PER_BYTE) - 1, 0)
  66
  67 struct ip_tunnel_info {
  68         struct ip_tunnel_key    key;
  69 #ifdef CONFIG_DST_CACHE
  70         struct dst_cache        dst_cache;
  71 #endif
  72         u8                      options_len;
  73         u8                      mode;
  74 };
  75
  76 /* 6rd prefix/relay information */
  77 #ifdef CONFIG_IPV6_SIT_6RD
  78 struct ip_tunnel_6rd_parm {
  79         struct in6_addr         prefix;
  80         __be32                  relay_prefix;
  81         u16                     prefixlen;
  82         u16                     relay_prefixlen;
  83 };
  84 #endif
  85
  86 struct ip_tunnel_encap {
  87         u16                     type;
  88         u16                     flags;
  89         __be16                  sport;
  90         __be16                  dport;
  91 };
  92
  93 struct ip_tunnel_prl_entry {
  94         struct ip_tunnel_prl_entry __rcu *next;
  95         __be32                          addr;
  96         u16                             flags;
  97         struct rcu_head                 rcu_head;
  98 };
  99
 100 struct metadata_dst;
 101
 102 struct ip_tunnel {
 103         struct ip_tunnel __rcu  *next;
 104         struct hlist_node hash_node;
 105         struct net_device       *dev;
 106         struct net              *net;   /* netns for packet i/o */
 107
 108         unsigned long   err_time;       /* Time when the last ICMP error
 109                                          * arrived */
 110         int             err_count;      /* Number of arrived ICMP errors */
 111
 112         /* These four fields used only by GRE */
 113         u32             i_seqno;        /* The last seen seqno  */
 114         u32             o_seqno;        /* The last output seqno */
 115         int             tun_hlen;       /* Precalculated header length */
 116
 117         struct dst_cache dst_cache;
 118
 119         struct ip_tunnel_parm parms;
 120
 121         int             mlink;
 122         int             encap_hlen;     /* Encap header length (FOU,GUE) */
 123         int             hlen;           /* tun_hlen + encap_hlen */
 124         struct ip_tunnel_encap encap;
 125
 126         /* for SIT */
 127 #ifdef CONFIG_IPV6_SIT_6RD
 128         struct ip_tunnel_6rd_parm ip6rd;
 129 #endif
 130         struct ip_tunnel_prl_entry __rcu *prl;  /* potential router list */
 131         unsigned int            prl_count;      /* # of entries in PRL */
 132         int                     ip_tnl_net_id;
 133         struct gro_cells        gro_cells;
 134         bool                    collect_md;
 135 };
 136
 137 #define TUNNEL_CSUM             __cpu_to_be16(0x01)
 138 #define TUNNEL_ROUTING          __cpu_to_be16(0x02)
 139 #define TUNNEL_KEY              __cpu_to_be16(0x04)
 140 #define TUNNEL_SEQ              __cpu_to_be16(0x08)
 141 #define TUNNEL_STRICT           __cpu_to_be16(0x10)
 142 #define TUNNEL_REC              __cpu_to_be16(0x20)
 143 #define TUNNEL_VERSION          __cpu_to_be16(0x40)
 144 #define TUNNEL_NO_KEY           __cpu_to_be16(0x80)
 145 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
 146 #define TUNNEL_OAM              __cpu_to_be16(0x0200)
 147 #define TUNNEL_CRIT_OPT         __cpu_to_be16(0x0400)
 148 #define TUNNEL_GENEVE_OPT       __cpu_to_be16(0x0800)
 149 #define TUNNEL_VXLAN_OPT        __cpu_to_be16(0x1000)
 150 #define TUNNEL_NOCACHE          __cpu_to_be16(0x2000)
 151
 152 #define TUNNEL_OPTIONS_PRESENT  (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT)
 153
 154 struct tnl_ptk_info {
 155         __be16 flags;
 156         __be16 proto;
 157         __be32 key;
 158         __be32 seq;
 159 };
 160
 161 #define PACKET_RCVD     0
 162 #define PACKET_REJECT   1
 163 #define PACKET_NEXT     2
 164
 165 #define IP_TNL_HASH_BITS   7
 166 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
 167
 168 struct ip_tunnel_net {
 169         struct net_device *fb_tunnel_dev;
 170         struct hlist_head tunnels[IP_TNL_HASH_SIZE];
 171         struct ip_tunnel __rcu *collect_md_tun;
 172 };
 173
 174 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
 175                                       __be32 saddr, __be32 daddr,
 176                                       u8 tos, u8 ttl, __be32 label,
 177                                       __be16 tp_src, __be16 tp_dst,
 178                                       __be64 tun_id, __be16 tun_flags)
 179 {
 180         key->tun_id = tun_id;
 181         key->u.ipv4.src = saddr;
 182         key->u.ipv4.dst = daddr;
 183         memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
 184                0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
 185         key->tos = tos;
 186         key->ttl = ttl;
 187         key->label = label;
 188         key->tun_flags = tun_flags;
 189
 190         /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
 191          * the upper tunnel are used.
 192          * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
 193          */
 194         key->tp_src = tp_src;
 195         key->tp_dst = tp_dst;
 196
 197         /* Clear struct padding. */
 198         if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
 199                 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
 200                        0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
 201 }
 202
 203 static inline bool
 204 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
 205                            const struct ip_tunnel_info *info)
 206 {
 207         if (skb->mark)
 208                 return false;
 209         if (!info)
 210                 return true;
 211         if (info->key.tun_flags & TUNNEL_NOCACHE)
 212                 return false;
 213
 214         return true;
 215 }
 216
 217 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
 218                                                *tun_info)
 219 {
 220         return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
 221 }
 222
 223 #ifdef CONFIG_INET
 224
 225 int ip_tunnel_init(struct net_device *dev);
 226 void ip_tunnel_uninit(struct net_device *dev);
 227 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
 228 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
 229 int ip_tunnel_get_iflink(const struct net_device *dev);
 230 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id,
 231                        struct rtnl_link_ops *ops, char *devname);
 232
 233 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops);
 234
 235 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 236                     const struct iphdr *tnl_params, const u8 protocol);
 237 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
 238 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
 239 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
 240
 241 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
 242                                                 struct rtnl_link_stats64 *tot);
 243 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
 244                                    int link, __be16 flags,
 245                                    __be32 remote, __be32 local,
 246                                    __be32 key);
 247
 248 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
 249                   const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
 250                   bool log_ecn_error);
 251 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
 252                          struct ip_tunnel_parm *p);
 253 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
 254                       struct ip_tunnel_parm *p);
 255 void ip_tunnel_setup(struct net_device *dev, int net_id);
 256
 257 struct ip_tunnel_encap_ops {
 258         size_t (*encap_hlen)(struct ip_tunnel_encap *e);
 259         int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
 260                             u8 *protocol, struct flowi4 *fl4);
 261 };
 262
 263 #define MAX_IPTUN_ENCAP_OPS 8
 264
 265 extern const struct ip_tunnel_encap_ops __rcu *
 266                 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
 267
 268 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
 269                             unsigned int num);
 270 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
 271                             unsigned int num);
 272
 273 int ip_tunnel_encap_setup(struct ip_tunnel *t,
 274                           struct ip_tunnel_encap *ipencap);
 275
 276 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
 277 {
 278         const struct ip_tunnel_encap_ops *ops;
 279         int hlen = -EINVAL;
 280
 281         if (e->type == TUNNEL_ENCAP_NONE)
 282                 return 0;
 283
 284         if (e->type >= MAX_IPTUN_ENCAP_OPS)
 285                 return -EINVAL;
 286
 287         rcu_read_lock();
 288         ops = rcu_dereference(iptun_encaps[e->type]);
 289         if (likely(ops && ops->encap_hlen))
 290                 hlen = ops->encap_hlen(e);
 291         rcu_read_unlock();
 292
 293         return hlen;
 294 }
 295
 296 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
 297                                   u8 *protocol, struct flowi4 *fl4)
 298 {
 299         const struct ip_tunnel_encap_ops *ops;
 300         int ret = -EINVAL;
 301
 302         if (t->encap.type == TUNNEL_ENCAP_NONE)
 303                 return 0;
 304
 305         if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
 306                 return -EINVAL;
 307
 308         rcu_read_lock();
 309         ops = rcu_dereference(iptun_encaps[t->encap.type]);
 310         if (likely(ops && ops->build_header))
 311                 ret = ops->build_header(skb, &t->encap, protocol, fl4);
 312         rcu_read_unlock();
 313
 314         return ret;
 315 }
 316
 317 /* Extract dsfield from inner protocol */
 318 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
 319                                        const struct sk_buff *skb)
 320 {
 321         if (skb->protocol == htons(ETH_P_IP))
 322                 return iph->tos;
 323         else if (skb->protocol == htons(ETH_P_IPV6))
 324                 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
 325         else
 326                 return 0;
 327 }
 328
 329 /* Propogate ECN bits out */
 330 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
 331                                      const struct sk_buff *skb)
 332 {
 333         u8 inner = ip_tunnel_get_dsfield(iph, skb);
 334
 335         return INET_ECN_encapsulate(tos, inner);
 336 }
 337
 338 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 339                            __be16 inner_proto, bool raw_proto, bool xnet);
 340
 341 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 342                                        __be16 inner_proto, bool xnet)
 343 {
 344         return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
 345 }
 346
 347 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
 348                    __be32 src, __be32 dst, u8 proto,
 349                    u8 tos, u8 ttl, __be16 df, bool xnet);
 350 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 351                                              gfp_t flags);
 352
 353 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
 354
 355 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
 356 {
 357         if (skb_is_gso(skb)) {
 358                 int err;
 359
 360                 err = skb_unclone(skb, GFP_ATOMIC);
 361                 if (unlikely(err))
 362                         return err;
 363                 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
 364                                                NETIF_F_GSO_SHIFT);
 365         }
 366
 367         skb->encapsulation = 0;
 368         return 0;
 369 }
 370
 371 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
 372 {
 373         if (pkt_len > 0) {
 374                 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
 375
 376                 u64_stats_update_begin(&tstats->syncp);
 377                 tstats->tx_bytes += pkt_len;
 378                 tstats->tx_packets++;
 379                 u64_stats_update_end(&tstats->syncp);
 380                 put_cpu_ptr(tstats);
 381         } else {
 382                 struct net_device_stats *err_stats = &dev->stats;
 383
 384                 if (pkt_len < 0) {
 385                         err_stats->tx_errors++;
 386                         err_stats->tx_aborted_errors++;
 387                 } else {
 388                         err_stats->tx_dropped++;
 389                 }
 390         }
 391 }
 392
 393 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
 394 {
 395         return info + 1;
 396 }
 397
 398 static inline void ip_tunnel_info_opts_get(void *to,
 399                                            const struct ip_tunnel_info *info)
 400 {
 401         memcpy(to, info + 1, info->options_len);
 402 }
 403
 404 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 405                                            const void *from, int len)
 406 {
 407         memcpy(ip_tunnel_info_opts(info), from, len);
 408         info->options_len = len;
 409 }
 410
 411 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 412 {
 413         return (struct ip_tunnel_info *)lwtstate->data;
 414 }
 415
 416 extern struct static_key ip_tunnel_metadata_cnt;
 417
 418 /* Returns > 0 if metadata should be collected */
 419 static inline int ip_tunnel_collect_metadata(void)
 420 {
 421         return static_key_false(&ip_tunnel_metadata_cnt);
 422 }
 423
 424 void __init ip_tunnel_core_init(void);
 425
 426 void ip_tunnel_need_metadata(void);
 427 void ip_tunnel_unneed_metadata(void);
 428
 429 #else /* CONFIG_INET */
 430
 431 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
 432 {
 433         return NULL;
 434 }
 435
 436 static inline void ip_tunnel_need_metadata(void)
 437 {
 438 }
 439
 440 static inline void ip_tunnel_unneed_metadata(void)
 441 {
 442 }
 443
 444 static inline void ip_tunnel_info_opts_get(void *to,
 445                                            const struct ip_tunnel_info *info)
 446 {
 447 }
 448
 449 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
 450                                            const void *from, int len)
 451 {
 452         info->options_len = 0;
 453 }
 454
 455 #endif /* CONFIG_INET */
 456
 457 #endif /* __NET_IP_TUNNELS_H */