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
51const struct ip_tunnel_encap_ops __rcu *
52 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
53EXPORT_SYMBOL(iptun_encaps);
54
55const struct ip6_tnl_encap_ops __rcu *
56 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
57EXPORT_SYMBOL(ip6tun_encaps);
58
59void 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}
96EXPORT_SYMBOL_GPL(iptunnel_xmit);
97
98int __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 __vlan_hwaccel_clear_tag(skb);
124 skb_set_queue_mapping(skb, 0);
125 skb_scrub_packet(skb, xnet);
126
127 return iptunnel_pull_offloads(skb);
128}
129EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
130
131struct 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->type != METADATA_IP_TUNNEL ||
138 md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
139
140 return NULL;
141
142 res = metadata_dst_alloc(0, METADATA_IP_TUNNEL, flags);
143 if (!res)
144 return NULL;
145
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));
152 else
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;
156
157 return res;
158}
159EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
160
161int iptunnel_handle_offloads(struct sk_buff *skb,
162 int gso_type_mask)
163{
164 int err;
165
166 if (likely(!skb->encapsulation)) {
167 skb_reset_inner_headers(skb);
168 skb->encapsulation = 1;
169 }
170
171 if (skb_is_gso(skb)) {
172 err = skb_header_unclone(skb, GFP_ATOMIC);
173 if (unlikely(err))
174 return err;
175 skb_shinfo(skb)->gso_type |= gso_type_mask;
176 return 0;
177 }
178
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.
185 */
186 skb->encapsulation = 0;
187 }
188
189 return 0;
190}
191EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
192
193/* Often modified stats are per cpu, other are shared (netdev->stats) */
194void ip_tunnel_get_stats64(struct net_device *dev,
195 struct rtnl_link_stats64 *tot)
196{
197 int i;
198
199 netdev_stats_to_stats64(tot, &dev->stats);
200
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;
205 unsigned int start;
206
207 do {
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));
214
215 tot->rx_packets += rx_packets;
216 tot->tx_packets += tx_packets;
217 tot->rx_bytes += rx_bytes;
218 tot->tx_bytes += tx_bytes;
219 }
220}
221EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
222
223static 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 },
230};
231
232static 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)
236{
237 struct ip_tunnel_info *tun_info;
238 struct lwtunnel_state *new_state;
239 struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
240 int err;
241
242 err = nla_parse_nested(tb, LWTUNNEL_IP_MAX, attr, ip_tun_policy,
243 extack);
244 if (err < 0)
245 return err;
246
247 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
248 if (!new_state)
249 return -ENOMEM;
250
251 new_state->type = LWTUNNEL_ENCAP_IP;
252
253 tun_info = lwt_tun_info(new_state);
254
255#ifdef CONFIG_DST_CACHE
256 err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
257 if (err) {
258 lwtstate_free(new_state);
259 return err;
260 }
261#endif
262
263 if (tb[LWTUNNEL_IP_ID])
264 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
265
266 if (tb[LWTUNNEL_IP_DST])
267 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
268
269 if (tb[LWTUNNEL_IP_SRC])
270 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
271
272 if (tb[LWTUNNEL_IP_TTL])
273 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
274
275 if (tb[LWTUNNEL_IP_TOS])
276 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
277
278 if (tb[LWTUNNEL_IP_FLAGS])
279 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
280
281 tun_info->mode = IP_TUNNEL_INFO_TX;
282 tun_info->options_len = 0;
283
284 *ts = new_state;
285
286 return 0;
287}
288
289static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
290{
291#ifdef CONFIG_DST_CACHE
292 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
293
294 dst_cache_destroy(&tun_info->dst_cache);
295#endif
296}
297
298static int ip_tun_fill_encap_info(struct sk_buff *skb,
299 struct lwtunnel_state *lwtstate)
300{
301 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
302
303 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
304 LWTUNNEL_IP_PAD) ||
305 nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
306 nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
307 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
308 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
309 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags))
310 return -ENOMEM;
311
312 return 0;
313}
314
315static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
316{
317 return nla_total_size_64bit(8) /* LWTUNNEL_IP_ID */
318 + nla_total_size(4) /* LWTUNNEL_IP_DST */
319 + nla_total_size(4) /* LWTUNNEL_IP_SRC */
320 + nla_total_size(1) /* LWTUNNEL_IP_TOS */
321 + nla_total_size(1) /* LWTUNNEL_IP_TTL */
322 + nla_total_size(2); /* LWTUNNEL_IP_FLAGS */
323}
324
325static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
326{
327 return memcmp(lwt_tun_info(a), lwt_tun_info(b),
328 sizeof(struct ip_tunnel_info));
329}
330
331static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
332 .build_state = ip_tun_build_state,
333 .destroy_state = ip_tun_destroy_state,
334 .fill_encap = ip_tun_fill_encap_info,
335 .get_encap_size = ip_tun_encap_nlsize,
336 .cmp_encap = ip_tun_cmp_encap,
337 .owner = THIS_MODULE,
338};
339
340static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
341 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 },
342 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) },
343 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) },
344 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 },
345 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 },
346 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 },
347};
348
349static int ip6_tun_build_state(struct nlattr *attr,
350 unsigned int family, const void *cfg,
351 struct lwtunnel_state **ts,
352 struct netlink_ext_ack *extack)
353{
354 struct ip_tunnel_info *tun_info;
355 struct lwtunnel_state *new_state;
356 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
357 int err;
358
359 err = nla_parse_nested(tb, LWTUNNEL_IP6_MAX, attr, ip6_tun_policy,
360 extack);
361 if (err < 0)
362 return err;
363
364 new_state = lwtunnel_state_alloc(sizeof(*tun_info));
365 if (!new_state)
366 return -ENOMEM;
367
368 new_state->type = LWTUNNEL_ENCAP_IP6;
369
370 tun_info = lwt_tun_info(new_state);
371
372 if (tb[LWTUNNEL_IP6_ID])
373 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
374
375 if (tb[LWTUNNEL_IP6_DST])
376 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
377
378 if (tb[LWTUNNEL_IP6_SRC])
379 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
380
381 if (tb[LWTUNNEL_IP6_HOPLIMIT])
382 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
383
384 if (tb[LWTUNNEL_IP6_TC])
385 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
386
387 if (tb[LWTUNNEL_IP6_FLAGS])
388 tun_info->key.tun_flags = nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
389
390 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
391 tun_info->options_len = 0;
392
393 *ts = new_state;
394
395 return 0;
396}
397
398static int ip6_tun_fill_encap_info(struct sk_buff *skb,
399 struct lwtunnel_state *lwtstate)
400{
401 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
402
403 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
404 LWTUNNEL_IP6_PAD) ||
405 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
406 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
407 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
408 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
409 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags))
410 return -ENOMEM;
411
412 return 0;
413}
414
415static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
416{
417 return nla_total_size_64bit(8) /* LWTUNNEL_IP6_ID */
418 + nla_total_size(16) /* LWTUNNEL_IP6_DST */
419 + nla_total_size(16) /* LWTUNNEL_IP6_SRC */
420 + nla_total_size(1) /* LWTUNNEL_IP6_HOPLIMIT */
421 + nla_total_size(1) /* LWTUNNEL_IP6_TC */
422 + nla_total_size(2); /* LWTUNNEL_IP6_FLAGS */
423}
424
425static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
426 .build_state = ip6_tun_build_state,
427 .fill_encap = ip6_tun_fill_encap_info,
428 .get_encap_size = ip6_tun_encap_nlsize,
429 .cmp_encap = ip_tun_cmp_encap,
430 .owner = THIS_MODULE,
431};
432
433void __init ip_tunnel_core_init(void)
434{
435 /* If you land here, make sure whether increasing ip_tunnel_info's
436 * options_len is a reasonable choice with its usage in front ends
437 * (f.e., it's part of flow keys, etc).
438 */
439 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
440
441 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
442 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
443}
444
445DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
446EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
447
448void ip_tunnel_need_metadata(void)
449{
450 static_branch_inc(&ip_tunnel_metadata_cnt);
451}
452EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
453
454void ip_tunnel_unneed_metadata(void)
455{
456 static_branch_dec(&ip_tunnel_metadata_cnt);
457}
458EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
459