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>
41 #include <net/checksum.h>
42 #include <net/dsfield.h>
43 #include <net/inet_ecn.h>
45 #include <net/net_namespace.h>
46 #include <net/netns/generic.h>
47 #include <net/rtnetlink.h>
48 #include <net/dst_metadata.h>
50 void 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)
54 int pkt_len = skb->len - skb_inner_network_offset(skb);
55 struct net *net = dev_net(rt->dst.dev);
56 struct net_device *dev = skb->dev;
60 skb_scrub_packet(skb, xnet);
63 skb_dst_set(skb, &rt->dst);
64 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
66 /* Push down and install the IP header. */
67 skb_push(skb, sizeof(struct iphdr));
68 skb_reset_network_header(skb);
73 iph->ihl = sizeof(struct iphdr) >> 2;
75 iph->protocol = proto;
80 __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
82 err = ip_local_out(net, sk, skb);
83 if (unlikely(net_xmit_eval(err)))
85 iptunnel_xmit_stats(dev, pkt_len);
87 EXPORT_SYMBOL_GPL(iptunnel_xmit);
89 int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, __be16 inner_proto,
92 if (unlikely(!pskb_may_pull(skb, hdr_len)))
95 skb_pull_rcsum(skb, hdr_len);
97 if (inner_proto == htons(ETH_P_TEB)) {
100 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
103 eh = (struct ethhdr *)skb->data;
104 if (likely(eth_proto_is_802_3(eh->h_proto)))
105 skb->protocol = eh->h_proto;
107 skb->protocol = htons(ETH_P_802_2);
110 skb->protocol = inner_proto;
113 skb_clear_hash_if_not_l4(skb);
115 skb_set_queue_mapping(skb, 0);
116 skb_scrub_packet(skb, xnet);
119 EXPORT_SYMBOL_GPL(iptunnel_pull_header);
121 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
124 struct metadata_dst *res;
125 struct ip_tunnel_info *dst, *src;
127 if (!md || md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
130 res = metadata_dst_alloc(0, flags);
134 dst = &res->u.tun_info;
135 src = &md->u.tun_info;
136 dst->key.tun_id = src->key.tun_id;
137 if (src->mode & IP_TUNNEL_INFO_IPV6)
138 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
139 sizeof(struct in6_addr));
141 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
142 dst->mode = src->mode | IP_TUNNEL_INFO_TX;
146 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
148 struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb,
153 if (likely(!skb->encapsulation)) {
154 skb_reset_inner_headers(skb);
155 skb->encapsulation = 1;
158 if (skb_is_gso(skb)) {
159 err = skb_unclone(skb, GFP_ATOMIC);
162 skb_shinfo(skb)->gso_type |= gso_type_mask;
166 if (skb->ip_summed != CHECKSUM_PARTIAL) {
167 skb->ip_summed = CHECKSUM_NONE;
168 /* We clear encapsulation here to prevent badly-written
169 * drivers potentially deciding to offload an inner checksum
170 * if we set CHECKSUM_PARTIAL on the outer header.
171 * This should go away when the drivers are all fixed.
173 skb->encapsulation = 0;
181 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
183 /* Often modified stats are per cpu, other are shared (netdev->stats) */
184 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
185 struct rtnl_link_stats64 *tot)
189 netdev_stats_to_stats64(tot, &dev->stats);
191 for_each_possible_cpu(i) {
192 const struct pcpu_sw_netstats *tstats =
193 per_cpu_ptr(dev->tstats, i);
194 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
198 start = u64_stats_fetch_begin_irq(&tstats->syncp);
199 rx_packets = tstats->rx_packets;
200 tx_packets = tstats->tx_packets;
201 rx_bytes = tstats->rx_bytes;
202 tx_bytes = tstats->tx_bytes;
203 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
205 tot->rx_packets += rx_packets;
206 tot->tx_packets += tx_packets;
207 tot->rx_bytes += rx_bytes;
208 tot->tx_bytes += tx_bytes;
213 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
215 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
216 [LWTUNNEL_IP_ID] = { .type = NLA_U64 },
217 [LWTUNNEL_IP_DST] = { .type = NLA_U32 },
218 [LWTUNNEL_IP_SRC] = { .type = NLA_U32 },
219 [LWTUNNEL_IP_TTL] = { .type = NLA_U8 },
220 [LWTUNNEL_IP_TOS] = { .type = NLA_U8 },
221 [LWTUNNEL_IP_FLAGS] = { .type = NLA_U16 },
224 static int ip_tun_build_state(struct net_device *dev, struct nlattr *attr,
225 unsigned int family, const void *cfg,
226 struct lwtunnel_state **ts)
228 struct ip_tunnel_info *tun_info;
229 struct lwtunnel_state *new_state;
230 struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
233 err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy);
237 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
241 new_state->type = LWTUNNEL_ENCAP_IP;
243 tun_info = lwt_tun_info(new_state);
245 if (tb[LWTUNNEL_IP_ID])
246 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
248 if (tb[LWTUNNEL_IP_DST])
249 tun_info->key.u.ipv4.dst = nla_get_be32(tb[LWTUNNEL_IP_DST]);
251 if (tb[LWTUNNEL_IP_SRC])
252 tun_info->key.u.ipv4.src = nla_get_be32(tb[LWTUNNEL_IP_SRC]);
254 if (tb[LWTUNNEL_IP_TTL])
255 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
257 if (tb[LWTUNNEL_IP_TOS])
258 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
260 if (tb[LWTUNNEL_IP_FLAGS])
261 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
263 tun_info->mode = IP_TUNNEL_INFO_TX;
264 tun_info->options_len = 0;
271 static int ip_tun_fill_encap_info(struct sk_buff *skb,
272 struct lwtunnel_state *lwtstate)
274 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
276 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id) ||
277 nla_put_be32(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
278 nla_put_be32(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
279 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
280 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
281 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
287 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
289 return nla_total_size(8) /* LWTUNNEL_IP_ID */
290 + nla_total_size(4) /* LWTUNNEL_IP_DST */
291 + nla_total_size(4) /* LWTUNNEL_IP_SRC */
292 + nla_total_size(1) /* LWTUNNEL_IP_TOS */
293 + nla_total_size(1) /* LWTUNNEL_IP_TTL */
294 + nla_total_size(2); /* LWTUNNEL_IP_FLAGS */
297 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
299 return memcmp(lwt_tun_info(a), lwt_tun_info(b),
300 sizeof(struct ip_tunnel_info));
303 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
304 .build_state = ip_tun_build_state,
305 .fill_encap = ip_tun_fill_encap_info,
306 .get_encap_size = ip_tun_encap_nlsize,
307 .cmp_encap = ip_tun_cmp_encap,
310 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
311 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 },
312 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) },
313 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) },
314 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 },
315 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 },
316 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 },
319 static int ip6_tun_build_state(struct net_device *dev, struct nlattr *attr,
320 unsigned int family, const void *cfg,
321 struct lwtunnel_state **ts)
323 struct ip_tunnel_info *tun_info;
324 struct lwtunnel_state *new_state;
325 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
328 err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy);
332 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
336 new_state->type = LWTUNNEL_ENCAP_IP6;
338 tun_info = lwt_tun_info(new_state);
340 if (tb[LWTUNNEL_IP6_ID])
341 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
343 if (tb[LWTUNNEL_IP6_DST])
344 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
346 if (tb[LWTUNNEL_IP6_SRC])
347 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
349 if (tb[LWTUNNEL_IP6_HOPLIMIT])
350 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
352 if (tb[LWTUNNEL_IP6_TC])
353 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
355 if (tb[LWTUNNEL_IP6_FLAGS])
356 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
358 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
359 tun_info->options_len = 0;
366 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
367 struct lwtunnel_state *lwtstate)
369 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
371 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id) ||
372 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
373 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
374 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.tos) ||
375 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.ttl) ||
376 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
382 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
384 return nla_total_size(8) /* LWTUNNEL_IP6_ID */
385 + nla_total_size(16) /* LWTUNNEL_IP6_DST */
386 + nla_total_size(16) /* LWTUNNEL_IP6_SRC */
387 + nla_total_size(1) /* LWTUNNEL_IP6_HOPLIMIT */
388 + nla_total_size(1) /* LWTUNNEL_IP6_TC */
389 + nla_total_size(2); /* LWTUNNEL_IP6_FLAGS */
392 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
393 .build_state = ip6_tun_build_state,
394 .fill_encap = ip6_tun_fill_encap_info,
395 .get_encap_size = ip6_tun_encap_nlsize,
396 .cmp_encap = ip_tun_cmp_encap,
399 void __init ip_tunnel_core_init(void)
401 /* If you land here, make sure whether increasing ip_tunnel_info's
402 * options_len is a reasonable choice with its usage in front ends
403 * (f.e., it's part of flow keys, etc).
405 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
407 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
408 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
411 struct static_key ip_tunnel_metadata_cnt = STATIC_KEY_INIT_FALSE;
412 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
414 void ip_tunnel_need_metadata(void)
416 static_key_slow_inc(&ip_tunnel_metadata_cnt);
418 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
420 void ip_tunnel_unneed_metadata(void)
422 static_key_slow_dec(&ip_tunnel_metadata_cnt);
424 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);