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         struct iphdr *iph;
  67         int err;
  68
  69         skb_scrub_packet(skb, xnet);
  70
  71         skb_clear_hash_if_not_l4(skb);
  72         skb_dst_set(skb, &rt->dst);
  73         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
  74
  75         /* Push down and install the IP header. */
  76         skb_push(skb, sizeof(struct iphdr));
  77         skb_reset_network_header(skb);
  78
  79         iph = ip_hdr(skb);
  80
  81         iph->version    =       4;
  82         iph->ihl        =       sizeof(struct iphdr) >> 2;
  83         iph->frag_off   =       ip_mtu_locked(&rt->dst) ? 0 : df;
  84         iph->protocol   =       proto;
  85         iph->tos        =       tos;
  86         iph->daddr      =       dst;
  87         iph->saddr      =       src;
  88         iph->ttl        =       ttl;
  89         __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
  90
  91         err = ip_local_out(net, sk, skb);
  92         if (unlikely(net_xmit_eval(err)))
  93                 pkt_len = 0;
  94         iptunnel_xmit_stats(dev, pkt_len);
  95 }
  96 EXPORT_SYMBOL_GPL(iptunnel_xmit);
  97
  98 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
  99                            __be16 inner_proto, bool raw_proto, bool xnet)
 100 {
 101         if (unlikely(!pskb_may_pull(skb, hdr_len)))
 102                 return -ENOMEM;
 103
 104         skb_pull_rcsum(skb, hdr_len);
 105
 106         if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
 107                 struct ethhdr *eh;
 108
 109                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
 110                         return -ENOMEM;
 111
 112                 eh = (struct ethhdr *)skb->data;
 113                 if (likely(eth_proto_is_802_3(eh->h_proto)))
 114                         skb->protocol = eh->h_proto;
 115                 else
 116                         skb->protocol = htons(ETH_P_802_2);
 117
 118         } else {
 119                 skb->protocol = inner_proto;
 120         }
 121
 122         skb_clear_hash_if_not_l4(skb);
 123         skb->vlan_tci = 0;
 124         skb_set_queue_mapping(skb, 0);
 125         skb_scrub_packet(skb, xnet);
 126
 127         return iptunnel_pull_offloads(skb);
 128 }
 129 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
 130
 131 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 132                                              gfp_t flags)
 133 {
 134         struct metadata_dst *res;
 135         struct ip_tunnel_info *dst, *src;
 136
 137         if (!md || md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
 138                 return NULL;
 139
 140         res = metadata_dst_alloc(0, flags);
 141         if (!res)
 142                 return NULL;
 143
 144         dst = &res->u.tun_info;
 145         src = &md->u.tun_info;
 146         dst->key.tun_id = src->key.tun_id;
 147         if (src->mode & IP_TUNNEL_INFO_IPV6)
 148                 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
 149                        sizeof(struct in6_addr));
 150         else
 151                 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
 152         dst->mode = src->mode | IP_TUNNEL_INFO_TX;
 153
 154         return res;
 155 }
 156 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
 157
 158 int iptunnel_handle_offloads(struct sk_buff *skb,
 159                              int gso_type_mask)
 160 {
 161         int err;
 162
 163         if (likely(!skb->encapsulation)) {
 164                 skb_reset_inner_headers(skb);
 165                 skb->encapsulation = 1;
 166         }
 167
 168         if (skb_is_gso(skb)) {
 169                 err = skb_header_unclone(skb, GFP_ATOMIC);
 170                 if (unlikely(err))
 171                         return err;
 172                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 173                 return 0;
 174         }
 175
 176         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 177                 skb->ip_summed = CHECKSUM_NONE;
 178                 /* We clear encapsulation here to prevent badly-written
 179                  * drivers potentially deciding to offload an inner checksum
 180                  * if we set CHECKSUM_PARTIAL on the outer header.
 181                  * This should go away when the drivers are all fixed.
 182                  */
 183                 skb->encapsulation = 0;
 184         }
 185
 186         return 0;
 187 }
 188 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
 189
 190 /* Often modified stats are per cpu, other are shared (netdev->stats) */
 191 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
 192                                                 struct rtnl_link_stats64 *tot)
 193 {
 194         int i;
 195
 196         netdev_stats_to_stats64(tot, &dev->stats);
 197
 198         for_each_possible_cpu(i) {
 199                 const struct pcpu_sw_netstats *tstats =
 200                                                    per_cpu_ptr(dev->tstats, i);
 201                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 202                 unsigned int start;
 203
 204                 do {
 205                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 206                         rx_packets = tstats->rx_packets;
 207                         tx_packets = tstats->tx_packets;
 208                         rx_bytes = tstats->rx_bytes;
 209                         tx_bytes = tstats->tx_bytes;
 210                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 211
 212                 tot->rx_packets += rx_packets;
 213                 tot->tx_packets += tx_packets;
 214                 tot->rx_bytes   += rx_bytes;
 215                 tot->tx_bytes   += tx_bytes;
 216         }
 217
 218         return tot;
 219 }
 220 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
 221
 222 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
 223         [LWTUNNEL_IP_ID]        = { .type = NLA_U64 },
 224         [LWTUNNEL_IP_DST]       = { .type = NLA_U32 },
 225         [LWTUNNEL_IP_SRC]       = { .type = NLA_U32 },
 226         [LWTUNNEL_IP_TTL]       = { .type = NLA_U8 },
 227         [LWTUNNEL_IP_TOS]       = { .type = NLA_U8 },
 228         [LWTUNNEL_IP_FLAGS]     = { .type = NLA_U16 },
 229 };
 230
 231 static int ip_tun_build_state(struct net_device *dev, struct nlattr *attr,
 232                               unsigned int family, const void *cfg,
 233                               struct lwtunnel_state **ts)
 234 {
 235         struct ip_tunnel_info *tun_info;
 236         struct lwtunnel_state *new_state;
 237         struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
 238         int err;
 239
 240         err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy);
 241         if (err < 0)
 242                 return err;
 243
 244         new_state = lwtunnel_state_alloc(sizeof(*tun_info));
 245         if (!new_state)
 246                 return -ENOMEM;
 247
 248         new_state->type = LWTUNNEL_ENCAP_IP;
 249
 250         tun_info = lwt_tun_info(new_state);
 251
 252         if (tb[LWTUNNEL_IP_ID])
 253                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
 254
 255         if (tb[LWTUNNEL_IP_DST])
 256                 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
 257
 258         if (tb[LWTUNNEL_IP_SRC])
 259                 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
 260
 261         if (tb[LWTUNNEL_IP_TTL])
 262                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
 263
 264         if (tb[LWTUNNEL_IP_TOS])
 265                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
 266
 267         if (tb[LWTUNNEL_IP_FLAGS])
 268                 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
 269
 270         tun_info->mode = IP_TUNNEL_INFO_TX;
 271         tun_info->options_len = 0;
 272
 273         *ts = new_state;
 274
 275         return 0;
 276 }
 277
 278 static int ip_tun_fill_encap_info(struct sk_buff *skb,
 279                                   struct lwtunnel_state *lwtstate)
 280 {
 281         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 282
 283         if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
 284                          LWTUNNEL_IP_PAD) ||
 285             nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
 286             nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
 287             nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
 288             nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
 289             nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
 290                 return -ENOMEM;
 291
 292         return 0;
 293 }
 294
 295 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 296 {
 297         return nla_total_size_64bit(8)  /* LWTUNNEL_IP_ID */
 298                 + nla_total_size(4)     /* LWTUNNEL_IP_DST */
 299                 + nla_total_size(4)     /* LWTUNNEL_IP_SRC */
 300                 + nla_total_size(1)     /* LWTUNNEL_IP_TOS */
 301                 + nla_total_size(1)     /* LWTUNNEL_IP_TTL */
 302                 + nla_total_size(2);    /* LWTUNNEL_IP_FLAGS */
 303 }
 304
 305 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
 306 {
 307         return memcmp(lwt_tun_info(a), lwt_tun_info(b),
 308                       sizeof(struct ip_tunnel_info));
 309 }
 310
 311 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
 312         .build_state = ip_tun_build_state,
 313         .fill_encap = ip_tun_fill_encap_info,
 314         .get_encap_size = ip_tun_encap_nlsize,
 315         .cmp_encap = ip_tun_cmp_encap,
 316         .owner = THIS_MODULE,
 317 };
 318
 319 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
 320         [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
 321         [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
 322         [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
 323         [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
 324         [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
 325         [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
 326 };
 327
 328 static int ip6_tun_build_state(struct net_device *dev, struct nlattr *attr,
 329                                unsigned int family, const void *cfg,
 330                                struct lwtunnel_state **ts)
 331 {
 332         struct ip_tunnel_info *tun_info;
 333         struct lwtunnel_state *new_state;
 334         struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
 335         int err;
 336
 337         err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy);
 338         if (err < 0)
 339                 return err;
 340
 341         new_state = lwtunnel_state_alloc(sizeof(*tun_info));
 342         if (!new_state)
 343                 return -ENOMEM;
 344
 345         new_state->type = LWTUNNEL_ENCAP_IP6;
 346
 347         tun_info = lwt_tun_info(new_state);
 348
 349         if (tb[LWTUNNEL_IP6_ID])
 350                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
 351
 352         if (tb[LWTUNNEL_IP6_DST])
 353                 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
 354
 355         if (tb[LWTUNNEL_IP6_SRC])
 356                 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
 357
 358         if (tb[LWTUNNEL_IP6_HOPLIMIT])
 359                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
 360
 361         if (tb[LWTUNNEL_IP6_TC])
 362                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
 363
 364         if (tb[LWTUNNEL_IP6_FLAGS])
 365                 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
 366
 367         tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
 368         tun_info->options_len = 0;
 369
 370         *ts = new_state;
 371
 372         return 0;
 373 }
 374
 375 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
 376                                    struct lwtunnel_state *lwtstate)
 377 {
 378         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 379
 380         if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
 381                          LWTUNNEL_IP6_PAD) ||
 382             nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
 383             nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
 384             nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
 385             nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
 386             nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
 387                 return -ENOMEM;
 388
 389         return 0;
 390 }
 391
 392 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 393 {
 394         return nla_total_size_64bit(8)  /* LWTUNNEL_IP6_ID */
 395                 + nla_total_size(16)    /* LWTUNNEL_IP6_DST */
 396                 + nla_total_size(16)    /* LWTUNNEL_IP6_SRC */
 397                 + nla_total_size(1)     /* LWTUNNEL_IP6_HOPLIMIT */
 398                 + nla_total_size(1)     /* LWTUNNEL_IP6_TC */
 399                 + nla_total_size(2);    /* LWTUNNEL_IP6_FLAGS */
 400 }
 401
 402 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
 403         .build_state = ip6_tun_build_state,
 404         .fill_encap = ip6_tun_fill_encap_info,
 405         .get_encap_size = ip6_tun_encap_nlsize,
 406         .cmp_encap = ip_tun_cmp_encap,
 407         .owner = THIS_MODULE,
 408 };
 409
 410 void __init ip_tunnel_core_init(void)
 411 {
 412         /* If you land here, make sure whether increasing ip_tunnel_info's
 413          * options_len is a reasonable choice with its usage in front ends
 414          * (f.e., it's part of flow keys, etc).
 415          */
 416         BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
 417
 418         lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
 419         lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
 420 }
 421
 422 struct static_key ip_tunnel_metadata_cnt = STATIC_KEY_INIT_FALSE;
 423 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
 424
 425 void ip_tunnel_need_metadata(void)
 426 {
 427         static_key_slow_inc(&ip_tunnel_metadata_cnt);
 428 }
 429 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
 430
 431 void ip_tunnel_unneed_metadata(void)
 432 {
 433         static_key_slow_dec(&ip_tunnel_metadata_cnt);
 434 }
 435 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);