net/ipv4/ip_tunnel_core.c

   1 /*
   2  * Copyright (c) 2013 Nicira, Inc.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License
  14  * along with this program; if not, write to the Free Software
  15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16  * 02110-1301, USA
  17  */
  18
  19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  20
  21 #include <linux/types.h>
  22 #include <linux/kernel.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/netdevice.h>
  25 #include <linux/in.h>
  26 #include <linux/if_arp.h>
  27 #include <linux/init.h>
  28 #include <linux/in6.h>
  29 #include <linux/inetdevice.h>
  30 #include <linux/netfilter_ipv4.h>
  31 #include <linux/etherdevice.h>
  32 #include <linux/if_ether.h>
  33 #include <linux/if_vlan.h>
  34 #include <linux/static_key.h>
  35
  36 #include <net/ip.h>
  37 #include <net/icmp.h>
  38 #include <net/protocol.h>
  39 #include <net/ip_tunnels.h>
  40 #include <net/ip6_tunnel.h>
  41 #include <net/arp.h>
  42 #include <net/checksum.h>
  43 #include <net/dsfield.h>
  44 #include <net/inet_ecn.h>
  45 #include <net/xfrm.h>
  46 #include <net/net_namespace.h>
  47 #include <net/netns/generic.h>
  48 #include <net/rtnetlink.h>
  49 #include <net/dst_metadata.h>
  50
  51 const struct ip_tunnel_encap_ops __rcu *
  52                 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
  53 EXPORT_SYMBOL(iptun_encaps);
  54
  55 const struct ip6_tnl_encap_ops __rcu *
  56                 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
  57 EXPORT_SYMBOL(ip6tun_encaps);
  58
  59 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
  60                    __be32 src, __be32 dst, __u8 proto,
  61                    __u8 tos, __u8 ttl, __be16 df, bool xnet)
  62 {
  63         int pkt_len = skb->len - skb_inner_network_offset(skb);
  64         struct net *net = dev_net(rt->dst.dev);
  65         struct net_device *dev = skb->dev;
  66         int skb_iif = skb->skb_iif;
  67         struct iphdr *iph;
  68         int err;
  69
  70         skb_scrub_packet(skb, xnet);
  71
  72         skb_clear_hash(skb);
  73         skb_dst_set(skb, &rt->dst);
  74         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
  75
  76         if (skb_iif && proto == IPPROTO_UDP) {
  77                 /* Arrived from an ingress interface and got udp encapuslated.
  78                  * The encapsulated network segment length may exceed dst mtu.
  79                  * Allow IP Fragmentation of segments.
  80                  */
  81                 IPCB(skb)->flags |= IPSKB_FRAG_SEGS;
  82         }
  83
  84         /* Push down and install the IP header. */
  85         skb_push(skb, sizeof(struct iphdr));
  86         skb_reset_network_header(skb);
  87
  88         iph = ip_hdr(skb);
  89
  90         iph->version    =       4;
  91         iph->ihl        =       sizeof(struct iphdr) >> 2;
  92         iph->frag_off   =       df;
  93         iph->protocol   =       proto;
  94         iph->tos        =       tos;
  95         iph->daddr      =       dst;
  96         iph->saddr      =       src;
  97         iph->ttl        =       ttl;
  98         __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
  99
 100         err = ip_local_out(net, sk, skb);
 101         if (unlikely(net_xmit_eval(err)))
 102                 pkt_len = 0;
 103         iptunnel_xmit_stats(dev, pkt_len);
 104 }
 105 EXPORT_SYMBOL_GPL(iptunnel_xmit);
 106
 107 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 108                            __be16 inner_proto, bool raw_proto, bool xnet)
 109 {
 110         if (unlikely(!pskb_may_pull(skb, hdr_len)))
 111                 return -ENOMEM;
 112
 113         skb_pull_rcsum(skb, hdr_len);
 114
 115         if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
 116                 struct ethhdr *eh;
 117
 118                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
 119                         return -ENOMEM;
 120
 121                 eh = (struct ethhdr *)skb->data;
 122                 if (likely(eth_proto_is_802_3(eh->h_proto)))
 123                         skb->protocol = eh->h_proto;
 124                 else
 125                         skb->protocol = htons(ETH_P_802_2);
 126
 127         } else {
 128                 skb->protocol = inner_proto;
 129         }
 130
 131         skb_clear_hash_if_not_l4(skb);
 132         skb->vlan_tci = 0;
 133         skb_set_queue_mapping(skb, 0);
 134         skb_scrub_packet(skb, xnet);
 135
 136         return iptunnel_pull_offloads(skb);
 137 }
 138 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
 139
 140 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 141                                              gfp_t flags)
 142 {
 143         struct metadata_dst *res;
 144         struct ip_tunnel_info *dst, *src;
 145
 146         if (!md || md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
 147                 return NULL;
 148
 149         res = metadata_dst_alloc(0, flags);
 150         if (!res)
 151                 return NULL;
 152
 153         dst = &res->u.tun_info;
 154         src = &md->u.tun_info;
 155         dst->key.tun_id = src->key.tun_id;
 156         if (src->mode & IP_TUNNEL_INFO_IPV6)
 157                 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
 158                        sizeof(struct in6_addr));
 159         else
 160                 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
 161         dst->mode = src->mode | IP_TUNNEL_INFO_TX;
 162
 163         return res;
 164 }
 165 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
 166
 167 int iptunnel_handle_offloads(struct sk_buff *skb,
 168                              int gso_type_mask)
 169 {
 170         int err;
 171
 172         if (likely(!skb->encapsulation)) {
 173                 skb_reset_inner_headers(skb);
 174                 skb->encapsulation = 1;
 175         }
 176
 177         if (skb_is_gso(skb)) {
 178                 err = skb_header_unclone(skb, GFP_ATOMIC);
 179                 if (unlikely(err))
 180                         return err;
 181                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 182                 return 0;
 183         }
 184
 185         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 186                 skb->ip_summed = CHECKSUM_NONE;
 187                 /* We clear encapsulation here to prevent badly-written
 188                  * drivers potentially deciding to offload an inner checksum
 189                  * if we set CHECKSUM_PARTIAL on the outer header.
 190                  * This should go away when the drivers are all fixed.
 191                  */
 192                 skb->encapsulation = 0;
 193         }
 194
 195         return 0;
 196 }
 197 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
 198
 199 /* Often modified stats are per cpu, other are shared (netdev->stats) */
 200 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
 201                                                 struct rtnl_link_stats64 *tot)
 202 {
 203         int i;
 204
 205         netdev_stats_to_stats64(tot, &dev->stats);
 206
 207         for_each_possible_cpu(i) {
 208                 const struct pcpu_sw_netstats *tstats =
 209                                                    per_cpu_ptr(dev->tstats, i);
 210                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 211                 unsigned int start;
 212
 213                 do {
 214                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 215                         rx_packets = tstats->rx_packets;
 216                         tx_packets = tstats->tx_packets;
 217                         rx_bytes = tstats->rx_bytes;
 218                         tx_bytes = tstats->tx_bytes;
 219                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 220
 221                 tot->rx_packets += rx_packets;
 222                 tot->tx_packets += tx_packets;
 223                 tot->rx_bytes   += rx_bytes;
 224                 tot->tx_bytes   += tx_bytes;
 225         }
 226
 227         return tot;
 228 }
 229 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
 230
 231 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
 232         [LWTUNNEL_IP_ID]        = { .type = NLA_U64 },
 233         [LWTUNNEL_IP_DST]       = { .type = NLA_U32 },
 234         [LWTUNNEL_IP_SRC]       = { .type = NLA_U32 },
 235         [LWTUNNEL_IP_TTL]       = { .type = NLA_U8 },
 236         [LWTUNNEL_IP_TOS]       = { .type = NLA_U8 },
 237         [LWTUNNEL_IP_FLAGS]     = { .type = NLA_U16 },
 238 };
 239
 240 static int ip_tun_build_state(struct net_device *dev, struct nlattr *attr,
 241                               unsigned int family, const void *cfg,
 242                               struct lwtunnel_state **ts)
 243 {
 244         struct ip_tunnel_info *tun_info;
 245         struct lwtunnel_state *new_state;
 246         struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
 247         int err;
 248
 249         err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy);
 250         if (err < 0)
 251                 return err;
 252
 253         new_state = lwtunnel_state_alloc(sizeof(*tun_info));
 254         if (!new_state)
 255                 return -ENOMEM;
 256
 257         new_state->type = LWTUNNEL_ENCAP_IP;
 258
 259         tun_info = lwt_tun_info(new_state);
 260
 261         if (tb[LWTUNNEL_IP_ID])
 262                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
 263
 264         if (tb[LWTUNNEL_IP_DST])
 265                 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
 266
 267         if (tb[LWTUNNEL_IP_SRC])
 268                 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
 269
 270         if (tb[LWTUNNEL_IP_TTL])
 271                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
 272
 273         if (tb[LWTUNNEL_IP_TOS])
 274                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
 275
 276         if (tb[LWTUNNEL_IP_FLAGS])
 277                 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
 278
 279         tun_info->mode = IP_TUNNEL_INFO_TX;
 280         tun_info->options_len = 0;
 281
 282         *ts = new_state;
 283
 284         return 0;
 285 }
 286
 287 static int ip_tun_fill_encap_info(struct sk_buff *skb,
 288                                   struct lwtunnel_state *lwtstate)
 289 {
 290         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 291
 292         if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
 293                          LWTUNNEL_IP_PAD) ||
 294             nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
 295             nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
 296             nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
 297             nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
 298             nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
 299                 return -ENOMEM;
 300
 301         return 0;
 302 }
 303
 304 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 305 {
 306         return nla_total_size_64bit(8)  /* LWTUNNEL_IP_ID */
 307                 + nla_total_size(4)     /* LWTUNNEL_IP_DST */
 308                 + nla_total_size(4)     /* LWTUNNEL_IP_SRC */
 309                 + nla_total_size(1)     /* LWTUNNEL_IP_TOS */
 310                 + nla_total_size(1)     /* LWTUNNEL_IP_TTL */
 311                 + nla_total_size(2);    /* LWTUNNEL_IP_FLAGS */
 312 }
 313
 314 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
 315 {
 316         return memcmp(lwt_tun_info(a), lwt_tun_info(b),
 317                       sizeof(struct ip_tunnel_info));
 318 }
 319
 320 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
 321         .build_state = ip_tun_build_state,
 322         .fill_encap = ip_tun_fill_encap_info,
 323         .get_encap_size = ip_tun_encap_nlsize,
 324         .cmp_encap = ip_tun_cmp_encap,
 325 };
 326
 327 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
 328         [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
 329         [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
 330         [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
 331         [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
 332         [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
 333         [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
 334 };
 335
 336 static int ip6_tun_build_state(struct net_device *dev, struct nlattr *attr,
 337                                unsigned int family, const void *cfg,
 338                                struct lwtunnel_state **ts)
 339 {
 340         struct ip_tunnel_info *tun_info;
 341         struct lwtunnel_state *new_state;
 342         struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
 343         int err;
 344
 345         err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy);
 346         if (err < 0)
 347                 return err;
 348
 349         new_state = lwtunnel_state_alloc(sizeof(*tun_info));
 350         if (!new_state)
 351                 return -ENOMEM;
 352
 353         new_state->type = LWTUNNEL_ENCAP_IP6;
 354
 355         tun_info = lwt_tun_info(new_state);
 356
 357         if (tb[LWTUNNEL_IP6_ID])
 358                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
 359
 360         if (tb[LWTUNNEL_IP6_DST])
 361                 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
 362
 363         if (tb[LWTUNNEL_IP6_SRC])
 364                 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
 365
 366         if (tb[LWTUNNEL_IP6_HOPLIMIT])
 367                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
 368
 369         if (tb[LWTUNNEL_IP6_TC])
 370                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
 371
 372         if (tb[LWTUNNEL_IP6_FLAGS])
 373                 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
 374
 375         tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
 376         tun_info->options_len = 0;
 377
 378         *ts = new_state;
 379
 380         return 0;
 381 }
 382
 383 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
 384                                    struct lwtunnel_state *lwtstate)
 385 {
 386         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 387
 388         if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
 389                          LWTUNNEL_IP6_PAD) ||
 390             nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
 391             nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
 392             nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
 393             nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
 394             nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
 395                 return -ENOMEM;
 396
 397         return 0;
 398 }
 399
 400 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 401 {
 402         return nla_total_size_64bit(8)  /* LWTUNNEL_IP6_ID */
 403                 + nla_total_size(16)    /* LWTUNNEL_IP6_DST */
 404                 + nla_total_size(16)    /* LWTUNNEL_IP6_SRC */
 405                 + nla_total_size(1)     /* LWTUNNEL_IP6_HOPLIMIT */
 406                 + nla_total_size(1)     /* LWTUNNEL_IP6_TC */
 407                 + nla_total_size(2);    /* LWTUNNEL_IP6_FLAGS */
 408 }
 409
 410 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
 411         .build_state = ip6_tun_build_state,
 412         .fill_encap = ip6_tun_fill_encap_info,
 413         .get_encap_size = ip6_tun_encap_nlsize,
 414         .cmp_encap = ip_tun_cmp_encap,
 415 };
 416
 417 void __init ip_tunnel_core_init(void)
 418 {
 419         /* If you land here, make sure whether increasing ip_tunnel_info's
 420          * options_len is a reasonable choice with its usage in front ends
 421          * (f.e., it's part of flow keys, etc).
 422          */
 423         BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
 424
 425         lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
 426         lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
 427 }
 428
 429 struct static_key ip_tunnel_metadata_cnt = STATIC_KEY_INIT_FALSE;
 430 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
 431
 432 void ip_tunnel_need_metadata(void)
 433 {
 434         static_key_slow_inc(&ip_tunnel_metadata_cnt);
 435 }
 436 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
 437
 438 void ip_tunnel_unneed_metadata(void)
 439 {
 440         static_key_slow_dec(&ip_tunnel_metadata_cnt);
 441 }
 442 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);