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);
92 if (unlikely(net_xmit_eval(err)))
94 iptunnel_xmit_stats(dev, pkt_len);
96 EXPORT_SYMBOL_GPL(iptunnel_xmit);
98 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
99 __be16 inner_proto, bool raw_proto, bool xnet)
101 if (unlikely(!pskb_may_pull(skb, hdr_len)))
104 skb_pull_rcsum(skb, hdr_len);
106 if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
109 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
112 eh = (struct ethhdr *)skb->data;
113 if (likely(eth_proto_is_802_3(eh->h_proto)))
114 skb->protocol = eh->h_proto;
116 skb->protocol = htons(ETH_P_802_2);
119 skb->protocol = inner_proto;
122 skb_clear_hash_if_not_l4(skb);
124 skb_set_queue_mapping(skb, 0);
125 skb_scrub_packet(skb, xnet);
127 return iptunnel_pull_offloads(skb);
129 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
131 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
134 struct metadata_dst *res;
135 struct ip_tunnel_info *dst, *src;
137 if (!md || md->type != METADATA_IP_TUNNEL ||
138 md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
142 res = metadata_dst_alloc(0, METADATA_IP_TUNNEL, flags);
146 dst = &res->u.tun_info;
147 src = &md->u.tun_info;
148 dst->key.tun_id = src->key.tun_id;
149 if (src->mode & IP_TUNNEL_INFO_IPV6)
150 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
151 sizeof(struct in6_addr));
153 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
154 dst->key.tun_flags = src->key.tun_flags;
155 dst->mode = src->mode | IP_TUNNEL_INFO_TX;
159 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
161 int iptunnel_handle_offloads(struct sk_buff *skb,
166 if (likely(!skb->encapsulation)) {
167 skb_reset_inner_headers(skb);
168 skb->encapsulation = 1;
171 if (skb_is_gso(skb)) {
172 err = skb_header_unclone(skb, GFP_ATOMIC);
175 skb_shinfo(skb)->gso_type |= gso_type_mask;
179 if (skb->ip_summed != CHECKSUM_PARTIAL) {
180 skb->ip_summed = CHECKSUM_NONE;
181 /* We clear encapsulation here to prevent badly-written
182 * drivers potentially deciding to offload an inner checksum
183 * if we set CHECKSUM_PARTIAL on the outer header.
184 * This should go away when the drivers are all fixed.
186 skb->encapsulation = 0;
191 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
193 /* Often modified stats are per cpu, other are shared (netdev->stats) */
194 void ip_tunnel_get_stats64(struct net_device *dev,
195 struct rtnl_link_stats64 *tot)
199 netdev_stats_to_stats64(tot, &dev->stats);
201 for_each_possible_cpu(i) {
202 const struct pcpu_sw_netstats *tstats =
203 per_cpu_ptr(dev->tstats, i);
204 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
208 start = u64_stats_fetch_begin_irq(&tstats->syncp);
209 rx_packets = tstats->rx_packets;
210 tx_packets = tstats->tx_packets;
211 rx_bytes = tstats->rx_bytes;
212 tx_bytes = tstats->tx_bytes;
213 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
215 tot->rx_packets += rx_packets;
216 tot->tx_packets += tx_packets;
217 tot->rx_bytes += rx_bytes;
218 tot->tx_bytes += tx_bytes;
221 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
223 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
224 [LWTUNNEL_IP_ID] = { .type = NLA_U64 },
225 [LWTUNNEL_IP_DST] = { .type = NLA_U32 },
226 [LWTUNNEL_IP_SRC] = { .type = NLA_U32 },
227 [LWTUNNEL_IP_TTL] = { .type = NLA_U8 },
228 [LWTUNNEL_IP_TOS] = { .type = NLA_U8 },
229 [LWTUNNEL_IP_FLAGS] = { .type = NLA_U16 },
232 static int ip_tun_build_state(struct nlattr *attr,
233 unsigned int family, const void *cfg,
234 struct lwtunnel_state **ts,
235 struct netlink_ext_ack *extack)
237 struct ip_tunnel_info *tun_info;
238 struct lwtunnel_state *new_state;
239 struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
242 err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy,
247 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
251 new_state->type = LWTUNNEL_ENCAP_IP;
253 tun_info = lwt_tun_info(new_state);
255 if (tb[LWTUNNEL_IP_ID])
256 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
258 if (tb[LWTUNNEL_IP_DST])
259 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
261 if (tb[LWTUNNEL_IP_SRC])
262 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
264 if (tb[LWTUNNEL_IP_TTL])
265 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
267 if (tb[LWTUNNEL_IP_TOS])
268 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
270 if (tb[LWTUNNEL_IP_FLAGS])
271 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
273 tun_info->mode = IP_TUNNEL_INFO_TX;
274 tun_info->options_len = 0;
281 static int ip_tun_fill_encap_info(struct sk_buff *skb,
282 struct lwtunnel_state *lwtstate)
284 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
286 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
288 nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
289 nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
290 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
291 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
292 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
298 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
300 return nla_total_size_64bit(8) /* LWTUNNEL_IP_ID */
301 + nla_total_size(4) /* LWTUNNEL_IP_DST */
302 + nla_total_size(4) /* LWTUNNEL_IP_SRC */
303 + nla_total_size(1) /* LWTUNNEL_IP_TOS */
304 + nla_total_size(1) /* LWTUNNEL_IP_TTL */
305 + nla_total_size(2); /* LWTUNNEL_IP_FLAGS */
308 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
310 return memcmp(lwt_tun_info(a), lwt_tun_info(b),
311 sizeof(struct ip_tunnel_info));
314 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
315 .build_state = ip_tun_build_state,
316 .fill_encap = ip_tun_fill_encap_info,
317 .get_encap_size = ip_tun_encap_nlsize,
318 .cmp_encap = ip_tun_cmp_encap,
319 .owner = THIS_MODULE,
322 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
323 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 },
324 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) },
325 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) },
326 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 },
327 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 },
328 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 },
331 static int ip6_tun_build_state(struct nlattr *attr,
332 unsigned int family, const void *cfg,
333 struct lwtunnel_state **ts,
334 struct netlink_ext_ack *extack)
336 struct ip_tunnel_info *tun_info;
337 struct lwtunnel_state *new_state;
338 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
341 err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy,
346 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
350 new_state->type = LWTUNNEL_ENCAP_IP6;
352 tun_info = lwt_tun_info(new_state);
354 if (tb[LWTUNNEL_IP6_ID])
355 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
357 if (tb[LWTUNNEL_IP6_DST])
358 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
360 if (tb[LWTUNNEL_IP6_SRC])
361 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
363 if (tb[LWTUNNEL_IP6_HOPLIMIT])
364 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
366 if (tb[LWTUNNEL_IP6_TC])
367 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
369 if (tb[LWTUNNEL_IP6_FLAGS])
370 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
372 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
373 tun_info->options_len = 0;
380 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
381 struct lwtunnel_state *lwtstate)
383 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
385 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
387 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
388 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
389 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
390 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
391 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
397 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
399 return nla_total_size_64bit(8) /* LWTUNNEL_IP6_ID */
400 + nla_total_size(16) /* LWTUNNEL_IP6_DST */
401 + nla_total_size(16) /* LWTUNNEL_IP6_SRC */
402 + nla_total_size(1) /* LWTUNNEL_IP6_HOPLIMIT */
403 + nla_total_size(1) /* LWTUNNEL_IP6_TC */
404 + nla_total_size(2); /* LWTUNNEL_IP6_FLAGS */
407 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
408 .build_state = ip6_tun_build_state,
409 .fill_encap = ip6_tun_fill_encap_info,
410 .get_encap_size = ip6_tun_encap_nlsize,
411 .cmp_encap = ip_tun_cmp_encap,
412 .owner = THIS_MODULE,
415 void __init ip_tunnel_core_init(void)
417 /* If you land here, make sure whether increasing ip_tunnel_info's
418 * options_len is a reasonable choice with its usage in front ends
419 * (f.e., it's part of flow keys, etc).
421 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
423 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
424 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
427 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
428 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
430 void ip_tunnel_need_metadata(void)
432 static_branch_inc(&ip_tunnel_metadata_cnt);
434 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
436 void ip_tunnel_unneed_metadata(void)
438 static_branch_dec(&ip_tunnel_metadata_cnt);
440 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);