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/arp.h>
  41 #include <net/checksum.h>
  42 #include <net/dsfield.h>
  43 #include <net/inet_ecn.h>
  44 #include <net/xfrm.h>
  45 #include <net/net_namespace.h>
  46 #include <net/netns/generic.h>
  47 #include <net/rtnetlink.h>
  48 #include <net/dst_metadata.h>
  49
  50 int iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
  51                   __be32 src, __be32 dst, __u8 proto,
  52                   __u8 tos, __u8 ttl, __be16 df, bool xnet)
  53 {
  54         int pkt_len = skb->len - skb_inner_network_offset(skb);
  55         struct net *net = dev_net(rt->dst.dev);
  56         struct iphdr *iph;
  57         int err;
  58
  59         skb_scrub_packet(skb, xnet);
  60
  61         skb_clear_hash(skb);
  62         skb_dst_set(skb, &rt->dst);
  63         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
  64
  65         /* Push down and install the IP header. */
  66         skb_push(skb, sizeof(struct iphdr));
  67         skb_reset_network_header(skb);
  68
  69         iph = ip_hdr(skb);
  70
  71         iph->version    =       4;
  72         iph->ihl        =       sizeof(struct iphdr) >> 2;
  73         iph->frag_off   =       df;
  74         iph->protocol   =       proto;
  75         iph->tos        =       tos;
  76         iph->daddr      =       dst;
  77         iph->saddr      =       src;
  78         iph->ttl        =       ttl;
  79         __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
  80
  81         err = ip_local_out(net, sk, skb);
  82         if (unlikely(net_xmit_eval(err)))
  83                 pkt_len = 0;
  84         return pkt_len;
  85 }
  86 EXPORT_SYMBOL_GPL(iptunnel_xmit);
  87
  88 int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, __be16 inner_proto)
  89 {
  90         if (unlikely(!pskb_may_pull(skb, hdr_len)))
  91                 return -ENOMEM;
  92
  93         skb_pull_rcsum(skb, hdr_len);
  94
  95         if (inner_proto == htons(ETH_P_TEB)) {
  96                 struct ethhdr *eh;
  97
  98                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
  99                         return -ENOMEM;
 100
 101                 eh = (struct ethhdr *)skb->data;
 102                 if (likely(eth_proto_is_802_3(eh->h_proto)))
 103                         skb->protocol = eh->h_proto;
 104                 else
 105                         skb->protocol = htons(ETH_P_802_2);
 106
 107         } else {
 108                 skb->protocol = inner_proto;
 109         }
 110
 111         nf_reset(skb);
 112         secpath_reset(skb);
 113         skb_clear_hash_if_not_l4(skb);
 114         skb_dst_drop(skb);
 115         skb->vlan_tci = 0;
 116         skb_set_queue_mapping(skb, 0);
 117         skb->pkt_type = PACKET_HOST;
 118         return 0;
 119 }
 120 EXPORT_SYMBOL_GPL(iptunnel_pull_header);
 121
 122 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 123                                              gfp_t flags)
 124 {
 125         struct metadata_dst *res;
 126         struct ip_tunnel_info *dst, *src;
 127
 128         if (!md || md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
 129                 return NULL;
 130
 131         res = metadata_dst_alloc(0, flags);
 132         if (!res)
 133                 return NULL;
 134
 135         dst = &res->u.tun_info;
 136         src = &md->u.tun_info;
 137         dst->key.tun_id = src->key.tun_id;
 138         if (src->mode & IP_TUNNEL_INFO_IPV6)
 139                 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
 140                        sizeof(struct in6_addr));
 141         else
 142                 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
 143         dst->mode = src->mode | IP_TUNNEL_INFO_TX;
 144
 145         return res;
 146 }
 147 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
 148
 149 struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb,
 150                                          bool csum_help,
 151                                          int gso_type_mask)
 152 {
 153         int err;
 154
 155         if (likely(!skb->encapsulation)) {
 156                 skb_reset_inner_headers(skb);
 157                 skb->encapsulation = 1;
 158         }
 159
 160         if (skb_is_gso(skb)) {
 161                 err = skb_unclone(skb, GFP_ATOMIC);
 162                 if (unlikely(err))
 163                         goto error;
 164                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 165                 return skb;
 166         }
 167
 168         /* If packet is not gso and we are resolving any partial checksum,
 169          * clear encapsulation flag. This allows setting CHECKSUM_PARTIAL
 170          * on the outer header without confusing devices that implement
 171          * NETIF_F_IP_CSUM with encapsulation.
 172          */
 173         if (csum_help)
 174                 skb->encapsulation = 0;
 175
 176         if (skb->ip_summed == CHECKSUM_PARTIAL && csum_help) {
 177                 err = skb_checksum_help(skb);
 178                 if (unlikely(err))
 179                         goto error;
 180         } else if (skb->ip_summed != CHECKSUM_PARTIAL)
 181                 skb->ip_summed = CHECKSUM_NONE;
 182
 183         return skb;
 184 error:
 185         kfree_skb(skb);
 186         return ERR_PTR(err);
 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_u64(tb[LWTUNNEL_IP_ID]);
 254
 255         if (tb[LWTUNNEL_IP_DST])
 256                 tun_info->key.u.ipv4.dst = nla_get_be32(tb[LWTUNNEL_IP_DST]);
 257
 258         if (tb[LWTUNNEL_IP_SRC])
 259                 tun_info->key.u.ipv4.src = nla_get_be32(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_u16(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_u64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id) ||
 284             nla_put_be32(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
 285             nla_put_be32(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
 286             nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
 287             nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
 288             nla_put_u16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
 289                 return -ENOMEM;
 290
 291         return 0;
 292 }
 293
 294 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 295 {
 296         return nla_total_size(8)        /* LWTUNNEL_IP_ID */
 297                 + nla_total_size(4)     /* LWTUNNEL_IP_DST */
 298                 + nla_total_size(4)     /* LWTUNNEL_IP_SRC */
 299                 + nla_total_size(1)     /* LWTUNNEL_IP_TOS */
 300                 + nla_total_size(1)     /* LWTUNNEL_IP_TTL */
 301                 + nla_total_size(2);    /* LWTUNNEL_IP_FLAGS */
 302 }
 303
 304 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
 305 {
 306         return memcmp(lwt_tun_info(a), lwt_tun_info(b),
 307                       sizeof(struct ip_tunnel_info));
 308 }
 309
 310 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
 311         .build_state = ip_tun_build_state,
 312         .fill_encap = ip_tun_fill_encap_info,
 313         .get_encap_size = ip_tun_encap_nlsize,
 314         .cmp_encap = ip_tun_cmp_encap,
 315 };
 316
 317 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
 318         [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
 319         [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
 320         [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
 321         [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
 322         [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
 323         [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
 324 };
 325
 326 static int ip6_tun_build_state(struct net_device *dev, struct nlattr *attr,
 327                                unsigned int family, const void *cfg,
 328                                struct lwtunnel_state **ts)
 329 {
 330         struct ip_tunnel_info *tun_info;
 331         struct lwtunnel_state *new_state;
 332         struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
 333         int err;
 334
 335         err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy);
 336         if (err < 0)
 337                 return err;
 338
 339         new_state = lwtunnel_state_alloc(sizeof(*tun_info));
 340         if (!new_state)
 341                 return -ENOMEM;
 342
 343         new_state->type = LWTUNNEL_ENCAP_IP6;
 344
 345         tun_info = lwt_tun_info(new_state);
 346
 347         if (tb[LWTUNNEL_IP6_ID])
 348                 tun_info->key.tun_id = nla_get_u64(tb[LWTUNNEL_IP6_ID]);
 349
 350         if (tb[LWTUNNEL_IP6_DST])
 351                 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
 352
 353         if (tb[LWTUNNEL_IP6_SRC])
 354                 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
 355
 356         if (tb[LWTUNNEL_IP6_HOPLIMIT])
 357                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
 358
 359         if (tb[LWTUNNEL_IP6_TC])
 360                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
 361
 362         if (tb[LWTUNNEL_IP6_FLAGS])
 363                 tun_info->key.tun_flags = nla_get_u16(tb[LWTUNNEL_IP6_FLAGS]);
 364
 365         tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
 366         tun_info->options_len = 0;
 367
 368         *ts = new_state;
 369
 370         return 0;
 371 }
 372
 373 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
 374                                    struct lwtunnel_state *lwtstate)
 375 {
 376         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 377
 378         if (nla_put_u64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id) ||
 379             nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
 380             nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
 381             nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.tos) ||
 382             nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.ttl) ||
 383             nla_put_u16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
 384                 return -ENOMEM;
 385
 386         return 0;
 387 }
 388
 389 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 390 {
 391         return nla_total_size(8)        /* LWTUNNEL_IP6_ID */
 392                 + nla_total_size(16)    /* LWTUNNEL_IP6_DST */
 393                 + nla_total_size(16)    /* LWTUNNEL_IP6_SRC */
 394                 + nla_total_size(1)     /* LWTUNNEL_IP6_HOPLIMIT */
 395                 + nla_total_size(1)     /* LWTUNNEL_IP6_TC */
 396                 + nla_total_size(2);    /* LWTUNNEL_IP6_FLAGS */
 397 }
 398
 399 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
 400         .build_state = ip6_tun_build_state,
 401         .fill_encap = ip6_tun_fill_encap_info,
 402         .get_encap_size = ip6_tun_encap_nlsize,
 403         .cmp_encap = ip_tun_cmp_encap,
 404 };
 405
 406 void __init ip_tunnel_core_init(void)
 407 {
 408         lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
 409         lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
 410 }
 411
 412 struct static_key ip_tunnel_metadata_cnt = STATIC_KEY_INIT_FALSE;
 413 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
 414
 415 void ip_tunnel_need_metadata(void)
 416 {
 417         static_key_slow_inc(&ip_tunnel_metadata_cnt);
 418 }
 419 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
 420
 421 void ip_tunnel_unneed_metadata(void)
 422 {
 423         static_key_slow_dec(&ip_tunnel_metadata_cnt);
 424 }
 425 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);