2 * Copyright (c) 2013 Nicira, Inc.
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.
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.
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
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/types.h>
22 #include <linux/kernel.h>
23 #include <linux/skbuff.h>
24 #include <linux/netdevice.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>
38 #include <net/protocol.h>
39 #include <net/ip_tunnels.h>
40 #include <net/ip6_tunnel.h>
42 #include <net/checksum.h>
43 #include <net/dsfield.h>
44 #include <net/inet_ecn.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <net/rtnetlink.h>
49 #include <net/dst_metadata.h>
51 const struct ip_tunnel_encap_ops __rcu *
52 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
53 EXPORT_SYMBOL(iptun_encaps);
55 const struct ip6_tnl_encap_ops __rcu *
56 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
57 EXPORT_SYMBOL(ip6tun_encaps);
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)
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;
69 skb_scrub_packet(skb, xnet);
71 skb_clear_hash_if_not_l4(skb);
72 skb_dst_set(skb, &rt->dst);
73 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
75 /* Push down and install the IP header. */
76 skb_push(skb, sizeof(struct iphdr));
77 skb_reset_network_header(skb);
82 iph->ihl = sizeof(struct iphdr) >> 2;
83 iph->frag_off = ip_mtu_locked(&rt->dst) ? 0 : df;
84 iph->protocol = proto;
89 __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
91 err = ip_local_out(net, sk, skb);
94 if (unlikely(net_xmit_eval(err)))
96 iptunnel_xmit_stats(dev, pkt_len);
99 EXPORT_SYMBOL_GPL(iptunnel_xmit);
101 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
102 __be16 inner_proto, bool raw_proto, bool xnet)
104 if (unlikely(!pskb_may_pull(skb, hdr_len)))
107 skb_pull_rcsum(skb, hdr_len);
109 if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
112 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
115 eh = (struct ethhdr *)skb->data;
116 if (likely(eth_proto_is_802_3(eh->h_proto)))
117 skb->protocol = eh->h_proto;
119 skb->protocol = htons(ETH_P_802_2);
122 skb->protocol = inner_proto;
125 skb_clear_hash_if_not_l4(skb);
127 skb_set_queue_mapping(skb, 0);
128 skb_scrub_packet(skb, xnet);
130 return iptunnel_pull_offloads(skb);
132 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
134 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
137 struct metadata_dst *res;
138 struct ip_tunnel_info *dst, *src;
140 if (!md || md->type != METADATA_IP_TUNNEL ||
141 md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
145 res = metadata_dst_alloc(0, METADATA_IP_TUNNEL, flags);
149 dst = &res->u.tun_info;
150 src = &md->u.tun_info;
151 dst->key.tun_id = src->key.tun_id;
152 if (src->mode & IP_TUNNEL_INFO_IPV6)
153 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
154 sizeof(struct in6_addr));
156 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
157 dst->mode = src->mode | IP_TUNNEL_INFO_TX;
161 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
163 int iptunnel_handle_offloads(struct sk_buff *skb,
168 if (likely(!skb->encapsulation)) {
169 skb_reset_inner_headers(skb);
170 skb->encapsulation = 1;
173 if (skb_is_gso(skb)) {
174 err = skb_header_unclone(skb, GFP_ATOMIC);
177 skb_shinfo(skb)->gso_type |= gso_type_mask;
181 if (skb->ip_summed != CHECKSUM_PARTIAL) {
182 skb->ip_summed = CHECKSUM_NONE;
183 /* We clear encapsulation here to prevent badly-written
184 * drivers potentially deciding to offload an inner checksum
185 * if we set CHECKSUM_PARTIAL on the outer header.
186 * This should go away when the drivers are all fixed.
188 skb->encapsulation = 0;
193 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
195 /* Often modified stats are per cpu, other are shared (netdev->stats) */
196 void ip_tunnel_get_stats64(struct net_device *dev,
197 struct rtnl_link_stats64 *tot)
201 netdev_stats_to_stats64(tot, &dev->stats);
203 for_each_possible_cpu(i) {
204 const struct pcpu_sw_netstats *tstats =
205 per_cpu_ptr(dev->tstats, i);
206 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
210 start = u64_stats_fetch_begin_irq(&tstats->syncp);
211 rx_packets = tstats->rx_packets;
212 tx_packets = tstats->tx_packets;
213 rx_bytes = tstats->rx_bytes;
214 tx_bytes = tstats->tx_bytes;
215 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
217 tot->rx_packets += rx_packets;
218 tot->tx_packets += tx_packets;
219 tot->rx_bytes += rx_bytes;
220 tot->tx_bytes += tx_bytes;
223 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
225 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
226 [LWTUNNEL_IP_ID] = { .type = NLA_U64 },
227 [LWTUNNEL_IP_DST] = { .type = NLA_U32 },
228 [LWTUNNEL_IP_SRC] = { .type = NLA_U32 },
229 [LWTUNNEL_IP_TTL] = { .type = NLA_U8 },
230 [LWTUNNEL_IP_TOS] = { .type = NLA_U8 },
231 [LWTUNNEL_IP_FLAGS] = { .type = NLA_U16 },
234 static int ip_tun_build_state(struct nlattr *attr,
235 unsigned int family, const void *cfg,
236 struct lwtunnel_state **ts,
237 struct netlink_ext_ack *extack)
239 struct ip_tunnel_info *tun_info;
240 struct lwtunnel_state *new_state;
241 struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
244 err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy,
249 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
253 new_state->type = LWTUNNEL_ENCAP_IP;
255 tun_info = lwt_tun_info(new_state);
257 if (tb[LWTUNNEL_IP_ID])
258 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
260 if (tb[LWTUNNEL_IP_DST])
261 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
263 if (tb[LWTUNNEL_IP_SRC])
264 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
266 if (tb[LWTUNNEL_IP_TTL])
267 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
269 if (tb[LWTUNNEL_IP_TOS])
270 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
272 if (tb[LWTUNNEL_IP_FLAGS])
273 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
275 tun_info->mode = IP_TUNNEL_INFO_TX;
276 tun_info->options_len = 0;
283 static int ip_tun_fill_encap_info(struct sk_buff *skb,
284 struct lwtunnel_state *lwtstate)
286 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
288 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
290 nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
291 nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
292 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
293 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
294 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
300 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
302 return nla_total_size_64bit(8) /* LWTUNNEL_IP_ID */
303 + nla_total_size(4) /* LWTUNNEL_IP_DST */
304 + nla_total_size(4) /* LWTUNNEL_IP_SRC */
305 + nla_total_size(1) /* LWTUNNEL_IP_TOS */
306 + nla_total_size(1) /* LWTUNNEL_IP_TTL */
307 + nla_total_size(2); /* LWTUNNEL_IP_FLAGS */
310 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
312 return memcmp(lwt_tun_info(a), lwt_tun_info(b),
313 sizeof(struct ip_tunnel_info));
316 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
317 .build_state = ip_tun_build_state,
318 .fill_encap = ip_tun_fill_encap_info,
319 .get_encap_size = ip_tun_encap_nlsize,
320 .cmp_encap = ip_tun_cmp_encap,
321 .owner = THIS_MODULE,
324 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
325 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 },
326 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) },
327 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) },
328 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 },
329 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 },
330 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 },
333 static int ip6_tun_build_state(struct nlattr *attr,
334 unsigned int family, const void *cfg,
335 struct lwtunnel_state **ts,
336 struct netlink_ext_ack *extack)
338 struct ip_tunnel_info *tun_info;
339 struct lwtunnel_state *new_state;
340 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
343 err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy,
348 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
352 new_state->type = LWTUNNEL_ENCAP_IP6;
354 tun_info = lwt_tun_info(new_state);
356 if (tb[LWTUNNEL_IP6_ID])
357 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
359 if (tb[LWTUNNEL_IP6_DST])
360 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
362 if (tb[LWTUNNEL_IP6_SRC])
363 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
365 if (tb[LWTUNNEL_IP6_HOPLIMIT])
366 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
368 if (tb[LWTUNNEL_IP6_TC])
369 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
371 if (tb[LWTUNNEL_IP6_FLAGS])
372 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
374 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
375 tun_info->options_len = 0;
382 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
383 struct lwtunnel_state *lwtstate)
385 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
387 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
389 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
390 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
391 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
392 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
393 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
399 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
401 return nla_total_size_64bit(8) /* LWTUNNEL_IP6_ID */
402 + nla_total_size(16) /* LWTUNNEL_IP6_DST */
403 + nla_total_size(16) /* LWTUNNEL_IP6_SRC */
404 + nla_total_size(1) /* LWTUNNEL_IP6_HOPLIMIT */
405 + nla_total_size(1) /* LWTUNNEL_IP6_TC */
406 + nla_total_size(2); /* LWTUNNEL_IP6_FLAGS */
409 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
410 .build_state = ip6_tun_build_state,
411 .fill_encap = ip6_tun_fill_encap_info,
412 .get_encap_size = ip6_tun_encap_nlsize,
413 .cmp_encap = ip_tun_cmp_encap,
414 .owner = THIS_MODULE,
417 void __init ip_tunnel_core_init(void)
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).
423 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
425 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
426 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
429 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
430 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
432 void ip_tunnel_need_metadata(void)
434 static_branch_inc(&ip_tunnel_metadata_cnt);
436 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
438 void ip_tunnel_unneed_metadata(void)
440 static_branch_dec(&ip_tunnel_metadata_cnt);
442 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);