Loading...
1/*
2 * Linux NET3: IP/IP protocol decoder modified to support
3 * virtual tunnel interface
4 *
5 * Authors:
6 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 *
13 */
14
15/*
16 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
17
18 For comments look at net/ipv4/ip_gre.c --ANK
19 */
20
21
22#include <linux/capability.h>
23#include <linux/module.h>
24#include <linux/types.h>
25#include <linux/kernel.h>
26#include <linux/uaccess.h>
27#include <linux/skbuff.h>
28#include <linux/netdevice.h>
29#include <linux/in.h>
30#include <linux/tcp.h>
31#include <linux/udp.h>
32#include <linux/if_arp.h>
33#include <linux/mroute.h>
34#include <linux/init.h>
35#include <linux/netfilter_ipv4.h>
36#include <linux/if_ether.h>
37#include <linux/icmpv6.h>
38
39#include <net/sock.h>
40#include <net/ip.h>
41#include <net/icmp.h>
42#include <net/ip_tunnels.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
48static struct rtnl_link_ops vti_link_ops __read_mostly;
49
50static int vti_net_id __read_mostly;
51static int vti_tunnel_init(struct net_device *dev);
52
53static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
54 int encap_type)
55{
56 struct ip_tunnel *tunnel;
57 const struct iphdr *iph = ip_hdr(skb);
58 struct net *net = dev_net(skb->dev);
59 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
60
61 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
62 iph->saddr, iph->daddr, 0);
63 if (tunnel != NULL) {
64 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
65 goto drop;
66
67 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
68 skb->mark = be32_to_cpu(tunnel->parms.i_key);
69
70 return xfrm_input(skb, nexthdr, spi, encap_type);
71 }
72
73 return -EINVAL;
74drop:
75 kfree_skb(skb);
76 return 0;
77}
78
79static int vti_rcv(struct sk_buff *skb)
80{
81 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
82 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
83
84 return vti_input(skb, ip_hdr(skb)->protocol, 0, 0);
85}
86
87static int vti_rcv_cb(struct sk_buff *skb, int err)
88{
89 unsigned short family;
90 struct net_device *dev;
91 struct pcpu_sw_netstats *tstats;
92 struct xfrm_state *x;
93 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
94
95 if (!tunnel)
96 return 1;
97
98 dev = tunnel->dev;
99
100 if (err) {
101 dev->stats.rx_errors++;
102 dev->stats.rx_dropped++;
103
104 return 0;
105 }
106
107 x = xfrm_input_state(skb);
108 family = x->inner_mode->afinfo->family;
109
110 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
111 return -EPERM;
112
113 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
114 skb->dev = dev;
115
116 tstats = this_cpu_ptr(dev->tstats);
117
118 u64_stats_update_begin(&tstats->syncp);
119 tstats->rx_packets++;
120 tstats->rx_bytes += skb->len;
121 u64_stats_update_end(&tstats->syncp);
122
123 return 0;
124}
125
126static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
127{
128 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
129 xfrm_address_t *saddr = (xfrm_address_t *)&src;
130
131 /* if there is no transform then this tunnel is not functional.
132 * Or if the xfrm is not mode tunnel.
133 */
134 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
135 x->props.family != AF_INET)
136 return false;
137
138 if (!dst)
139 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
140
141 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
142 return false;
143
144 return true;
145}
146
147static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
148 struct flowi *fl)
149{
150 struct ip_tunnel *tunnel = netdev_priv(dev);
151 struct ip_tunnel_parm *parms = &tunnel->parms;
152 struct dst_entry *dst = skb_dst(skb);
153 struct net_device *tdev; /* Device to other host */
154 int err;
155
156 if (!dst) {
157 dev->stats.tx_carrier_errors++;
158 goto tx_error_icmp;
159 }
160
161 dst_hold(dst);
162 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
163 if (IS_ERR(dst)) {
164 dev->stats.tx_carrier_errors++;
165 goto tx_error_icmp;
166 }
167
168 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
169 dev->stats.tx_carrier_errors++;
170 dst_release(dst);
171 goto tx_error_icmp;
172 }
173
174 tdev = dst->dev;
175
176 if (tdev == dev) {
177 dst_release(dst);
178 dev->stats.collisions++;
179 goto tx_error;
180 }
181
182 if (tunnel->err_count > 0) {
183 if (time_before(jiffies,
184 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
185 tunnel->err_count--;
186 dst_link_failure(skb);
187 } else
188 tunnel->err_count = 0;
189 }
190
191 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
192 skb_dst_set(skb, dst);
193 skb->dev = skb_dst(skb)->dev;
194
195 err = dst_output(skb);
196 if (net_xmit_eval(err) == 0)
197 err = skb->len;
198 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
199 return NETDEV_TX_OK;
200
201tx_error_icmp:
202 dst_link_failure(skb);
203tx_error:
204 dev->stats.tx_errors++;
205 kfree_skb(skb);
206 return NETDEV_TX_OK;
207}
208
209/* This function assumes it is being called from dev_queue_xmit()
210 * and that skb is filled properly by that function.
211 */
212static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
213{
214 struct ip_tunnel *tunnel = netdev_priv(dev);
215 struct flowi fl;
216
217 memset(&fl, 0, sizeof(fl));
218
219 skb->mark = be32_to_cpu(tunnel->parms.o_key);
220
221 switch (skb->protocol) {
222 case htons(ETH_P_IP):
223 xfrm_decode_session(skb, &fl, AF_INET);
224 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
225 break;
226 case htons(ETH_P_IPV6):
227 xfrm_decode_session(skb, &fl, AF_INET6);
228 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
229 break;
230 default:
231 dev->stats.tx_errors++;
232 dev_kfree_skb(skb);
233 return NETDEV_TX_OK;
234 }
235
236 return vti_xmit(skb, dev, &fl);
237}
238
239static int vti4_err(struct sk_buff *skb, u32 info)
240{
241 __be32 spi;
242 __u32 mark;
243 struct xfrm_state *x;
244 struct ip_tunnel *tunnel;
245 struct ip_esp_hdr *esph;
246 struct ip_auth_hdr *ah ;
247 struct ip_comp_hdr *ipch;
248 struct net *net = dev_net(skb->dev);
249 const struct iphdr *iph = (const struct iphdr *)skb->data;
250 int protocol = iph->protocol;
251 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
252
253 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
254 iph->daddr, iph->saddr, 0);
255 if (!tunnel)
256 return -1;
257
258 mark = be32_to_cpu(tunnel->parms.o_key);
259
260 switch (protocol) {
261 case IPPROTO_ESP:
262 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
263 spi = esph->spi;
264 break;
265 case IPPROTO_AH:
266 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
267 spi = ah->spi;
268 break;
269 case IPPROTO_COMP:
270 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
271 spi = htonl(ntohs(ipch->cpi));
272 break;
273 default:
274 return 0;
275 }
276
277 switch (icmp_hdr(skb)->type) {
278 case ICMP_DEST_UNREACH:
279 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
280 return 0;
281 case ICMP_REDIRECT:
282 break;
283 default:
284 return 0;
285 }
286
287 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
288 spi, protocol, AF_INET);
289 if (!x)
290 return 0;
291
292 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
293 ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
294 else
295 ipv4_redirect(skb, net, 0, 0, protocol, 0);
296 xfrm_state_put(x);
297
298 return 0;
299}
300
301static int
302vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
303{
304 int err = 0;
305 struct ip_tunnel_parm p;
306
307 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
308 return -EFAULT;
309
310 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
311 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
312 p.iph.ihl != 5)
313 return -EINVAL;
314 }
315
316 p.i_flags |= VTI_ISVTI;
317 err = ip_tunnel_ioctl(dev, &p, cmd);
318 if (err)
319 return err;
320
321 if (cmd != SIOCDELTUNNEL) {
322 p.i_flags |= GRE_KEY;
323 p.o_flags |= GRE_KEY;
324 }
325
326 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
327 return -EFAULT;
328 return 0;
329}
330
331static const struct net_device_ops vti_netdev_ops = {
332 .ndo_init = vti_tunnel_init,
333 .ndo_uninit = ip_tunnel_uninit,
334 .ndo_start_xmit = vti_tunnel_xmit,
335 .ndo_do_ioctl = vti_tunnel_ioctl,
336 .ndo_change_mtu = ip_tunnel_change_mtu,
337 .ndo_get_stats64 = ip_tunnel_get_stats64,
338};
339
340static void vti_tunnel_setup(struct net_device *dev)
341{
342 dev->netdev_ops = &vti_netdev_ops;
343 dev->type = ARPHRD_TUNNEL;
344 ip_tunnel_setup(dev, vti_net_id);
345}
346
347static int vti_tunnel_init(struct net_device *dev)
348{
349 struct ip_tunnel *tunnel = netdev_priv(dev);
350 struct iphdr *iph = &tunnel->parms.iph;
351
352 memcpy(dev->dev_addr, &iph->saddr, 4);
353 memcpy(dev->broadcast, &iph->daddr, 4);
354
355 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
356 dev->mtu = ETH_DATA_LEN;
357 dev->flags = IFF_NOARP;
358 dev->iflink = 0;
359 dev->addr_len = 4;
360 dev->features |= NETIF_F_LLTX;
361 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
362
363 return ip_tunnel_init(dev);
364}
365
366static void __net_init vti_fb_tunnel_init(struct net_device *dev)
367{
368 struct ip_tunnel *tunnel = netdev_priv(dev);
369 struct iphdr *iph = &tunnel->parms.iph;
370
371 iph->version = 4;
372 iph->protocol = IPPROTO_IPIP;
373 iph->ihl = 5;
374}
375
376static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
377 .handler = vti_rcv,
378 .input_handler = vti_input,
379 .cb_handler = vti_rcv_cb,
380 .err_handler = vti4_err,
381 .priority = 100,
382};
383
384static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
385 .handler = vti_rcv,
386 .input_handler = vti_input,
387 .cb_handler = vti_rcv_cb,
388 .err_handler = vti4_err,
389 .priority = 100,
390};
391
392static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
393 .handler = vti_rcv,
394 .input_handler = vti_input,
395 .cb_handler = vti_rcv_cb,
396 .err_handler = vti4_err,
397 .priority = 100,
398};
399
400static int __net_init vti_init_net(struct net *net)
401{
402 int err;
403 struct ip_tunnel_net *itn;
404
405 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
406 if (err)
407 return err;
408 itn = net_generic(net, vti_net_id);
409 vti_fb_tunnel_init(itn->fb_tunnel_dev);
410 return 0;
411}
412
413static void __net_exit vti_exit_net(struct net *net)
414{
415 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
416 ip_tunnel_delete_net(itn, &vti_link_ops);
417}
418
419static struct pernet_operations vti_net_ops = {
420 .init = vti_init_net,
421 .exit = vti_exit_net,
422 .id = &vti_net_id,
423 .size = sizeof(struct ip_tunnel_net),
424};
425
426static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
427{
428 return 0;
429}
430
431static void vti_netlink_parms(struct nlattr *data[],
432 struct ip_tunnel_parm *parms)
433{
434 memset(parms, 0, sizeof(*parms));
435
436 parms->iph.protocol = IPPROTO_IPIP;
437
438 if (!data)
439 return;
440
441 parms->i_flags = VTI_ISVTI;
442
443 if (data[IFLA_VTI_LINK])
444 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
445
446 if (data[IFLA_VTI_IKEY])
447 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
448
449 if (data[IFLA_VTI_OKEY])
450 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
451
452 if (data[IFLA_VTI_LOCAL])
453 parms->iph.saddr = nla_get_be32(data[IFLA_VTI_LOCAL]);
454
455 if (data[IFLA_VTI_REMOTE])
456 parms->iph.daddr = nla_get_be32(data[IFLA_VTI_REMOTE]);
457
458}
459
460static int vti_newlink(struct net *src_net, struct net_device *dev,
461 struct nlattr *tb[], struct nlattr *data[])
462{
463 struct ip_tunnel_parm parms;
464
465 vti_netlink_parms(data, &parms);
466 return ip_tunnel_newlink(dev, tb, &parms);
467}
468
469static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
470 struct nlattr *data[])
471{
472 struct ip_tunnel_parm p;
473
474 vti_netlink_parms(data, &p);
475 return ip_tunnel_changelink(dev, tb, &p);
476}
477
478static size_t vti_get_size(const struct net_device *dev)
479{
480 return
481 /* IFLA_VTI_LINK */
482 nla_total_size(4) +
483 /* IFLA_VTI_IKEY */
484 nla_total_size(4) +
485 /* IFLA_VTI_OKEY */
486 nla_total_size(4) +
487 /* IFLA_VTI_LOCAL */
488 nla_total_size(4) +
489 /* IFLA_VTI_REMOTE */
490 nla_total_size(4) +
491 0;
492}
493
494static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
495{
496 struct ip_tunnel *t = netdev_priv(dev);
497 struct ip_tunnel_parm *p = &t->parms;
498
499 nla_put_u32(skb, IFLA_VTI_LINK, p->link);
500 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
501 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
502 nla_put_be32(skb, IFLA_VTI_LOCAL, p->iph.saddr);
503 nla_put_be32(skb, IFLA_VTI_REMOTE, p->iph.daddr);
504
505 return 0;
506}
507
508static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
509 [IFLA_VTI_LINK] = { .type = NLA_U32 },
510 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
511 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
512 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
513 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
514};
515
516static struct rtnl_link_ops vti_link_ops __read_mostly = {
517 .kind = "vti",
518 .maxtype = IFLA_VTI_MAX,
519 .policy = vti_policy,
520 .priv_size = sizeof(struct ip_tunnel),
521 .setup = vti_tunnel_setup,
522 .validate = vti_tunnel_validate,
523 .newlink = vti_newlink,
524 .changelink = vti_changelink,
525 .get_size = vti_get_size,
526 .fill_info = vti_fill_info,
527};
528
529static int __init vti_init(void)
530{
531 int err;
532
533 pr_info("IPv4 over IPSec tunneling driver\n");
534
535 err = register_pernet_device(&vti_net_ops);
536 if (err < 0)
537 return err;
538 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
539 if (err < 0) {
540 unregister_pernet_device(&vti_net_ops);
541 pr_info("vti init: can't register tunnel\n");
542
543 return err;
544 }
545
546 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
547 if (err < 0) {
548 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
549 unregister_pernet_device(&vti_net_ops);
550 pr_info("vti init: can't register tunnel\n");
551
552 return err;
553 }
554
555 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
556 if (err < 0) {
557 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
558 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
559 unregister_pernet_device(&vti_net_ops);
560 pr_info("vti init: can't register tunnel\n");
561
562 return err;
563 }
564
565 err = rtnl_link_register(&vti_link_ops);
566 if (err < 0)
567 goto rtnl_link_failed;
568
569 return err;
570
571rtnl_link_failed:
572 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
573 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
574 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
575 unregister_pernet_device(&vti_net_ops);
576 return err;
577}
578
579static void __exit vti_fini(void)
580{
581 rtnl_link_unregister(&vti_link_ops);
582 if (xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP))
583 pr_info("vti close: can't deregister tunnel\n");
584 if (xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH))
585 pr_info("vti close: can't deregister tunnel\n");
586 if (xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP))
587 pr_info("vti close: can't deregister tunnel\n");
588
589
590 unregister_pernet_device(&vti_net_ops);
591}
592
593module_init(vti_init);
594module_exit(vti_fini);
595MODULE_LICENSE("GPL");
596MODULE_ALIAS_RTNL_LINK("vti");
597MODULE_ALIAS_NETDEV("ip_vti0");
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Linux NET3: IP/IP protocol decoder modified to support
4 * virtual tunnel interface
5 *
6 * Authors:
7 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
8 */
9
10/*
11 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
12
13 For comments look at net/ipv4/ip_gre.c --ANK
14 */
15
16
17#include <linux/capability.h>
18#include <linux/module.h>
19#include <linux/types.h>
20#include <linux/kernel.h>
21#include <linux/uaccess.h>
22#include <linux/skbuff.h>
23#include <linux/netdevice.h>
24#include <linux/in.h>
25#include <linux/tcp.h>
26#include <linux/udp.h>
27#include <linux/if_arp.h>
28#include <linux/init.h>
29#include <linux/netfilter_ipv4.h>
30#include <linux/if_ether.h>
31#include <linux/icmpv6.h>
32
33#include <net/sock.h>
34#include <net/ip.h>
35#include <net/icmp.h>
36#include <net/ip_tunnels.h>
37#include <net/inet_ecn.h>
38#include <net/xfrm.h>
39#include <net/net_namespace.h>
40#include <net/netns/generic.h>
41
42static struct rtnl_link_ops vti_link_ops __read_mostly;
43
44static unsigned int vti_net_id __read_mostly;
45static int vti_tunnel_init(struct net_device *dev);
46
47static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
48 int encap_type, bool update_skb_dev)
49{
50 struct ip_tunnel *tunnel;
51 const struct iphdr *iph = ip_hdr(skb);
52 struct net *net = dev_net(skb->dev);
53 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
54
55 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
56 iph->saddr, iph->daddr, 0);
57 if (tunnel) {
58 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
59 goto drop;
60
61 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
62
63 if (update_skb_dev)
64 skb->dev = tunnel->dev;
65
66 return xfrm_input(skb, nexthdr, spi, encap_type);
67 }
68
69 return -EINVAL;
70drop:
71 kfree_skb(skb);
72 return 0;
73}
74
75static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
76 int encap_type)
77{
78 return vti_input(skb, nexthdr, spi, encap_type, false);
79}
80
81static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev)
82{
83 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
84 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
85
86 return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev);
87}
88
89static int vti_rcv_proto(struct sk_buff *skb)
90{
91 return vti_rcv(skb, 0, false);
92}
93
94static int vti_rcv_cb(struct sk_buff *skb, int err)
95{
96 unsigned short family;
97 struct net_device *dev;
98 struct xfrm_state *x;
99 const struct xfrm_mode *inner_mode;
100 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
101 u32 orig_mark = skb->mark;
102 int ret;
103
104 if (!tunnel)
105 return 1;
106
107 dev = tunnel->dev;
108
109 if (err) {
110 DEV_STATS_INC(dev, rx_errors);
111 DEV_STATS_INC(dev, rx_dropped);
112
113 return 0;
114 }
115
116 x = xfrm_input_state(skb);
117
118 inner_mode = &x->inner_mode;
119
120 if (x->sel.family == AF_UNSPEC) {
121 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
122 if (inner_mode == NULL) {
123 XFRM_INC_STATS(dev_net(skb->dev),
124 LINUX_MIB_XFRMINSTATEMODEERROR);
125 return -EINVAL;
126 }
127 }
128
129 family = inner_mode->family;
130
131 skb->mark = be32_to_cpu(tunnel->parms.i_key);
132 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
133 skb->mark = orig_mark;
134
135 if (!ret)
136 return -EPERM;
137
138 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
139 skb->dev = dev;
140 dev_sw_netstats_rx_add(dev, skb->len);
141
142 return 0;
143}
144
145static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
146{
147 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
148 xfrm_address_t *saddr = (xfrm_address_t *)&src;
149
150 /* if there is no transform then this tunnel is not functional.
151 * Or if the xfrm is not mode tunnel.
152 */
153 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
154 x->props.family != AF_INET)
155 return false;
156
157 if (!dst)
158 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
159
160 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
161 return false;
162
163 return true;
164}
165
166static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
167 struct flowi *fl)
168{
169 struct ip_tunnel *tunnel = netdev_priv(dev);
170 struct ip_tunnel_parm *parms = &tunnel->parms;
171 struct dst_entry *dst = skb_dst(skb);
172 struct net_device *tdev; /* Device to other host */
173 int pkt_len = skb->len;
174 int err;
175 int mtu;
176
177 if (!dst) {
178 switch (skb->protocol) {
179 case htons(ETH_P_IP): {
180 struct rtable *rt;
181
182 fl->u.ip4.flowi4_oif = dev->ifindex;
183 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
184 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
185 if (IS_ERR(rt)) {
186 DEV_STATS_INC(dev, tx_carrier_errors);
187 goto tx_error_icmp;
188 }
189 dst = &rt->dst;
190 skb_dst_set(skb, dst);
191 break;
192 }
193#if IS_ENABLED(CONFIG_IPV6)
194 case htons(ETH_P_IPV6):
195 fl->u.ip6.flowi6_oif = dev->ifindex;
196 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
197 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
198 if (dst->error) {
199 dst_release(dst);
200 dst = NULL;
201 DEV_STATS_INC(dev, tx_carrier_errors);
202 goto tx_error_icmp;
203 }
204 skb_dst_set(skb, dst);
205 break;
206#endif
207 default:
208 DEV_STATS_INC(dev, tx_carrier_errors);
209 goto tx_error_icmp;
210 }
211 }
212
213 dst_hold(dst);
214 dst = xfrm_lookup_route(tunnel->net, dst, fl, NULL, 0);
215 if (IS_ERR(dst)) {
216 DEV_STATS_INC(dev, tx_carrier_errors);
217 goto tx_error_icmp;
218 }
219
220 if (dst->flags & DST_XFRM_QUEUE)
221 goto xmit;
222
223 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
224 DEV_STATS_INC(dev, tx_carrier_errors);
225 dst_release(dst);
226 goto tx_error_icmp;
227 }
228
229 tdev = dst->dev;
230
231 if (tdev == dev) {
232 dst_release(dst);
233 DEV_STATS_INC(dev, collisions);
234 goto tx_error;
235 }
236
237 mtu = dst_mtu(dst);
238 if (skb->len > mtu) {
239 skb_dst_update_pmtu_no_confirm(skb, mtu);
240 if (skb->protocol == htons(ETH_P_IP)) {
241 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
242 goto xmit;
243 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
244 htonl(mtu));
245 } else {
246 if (mtu < IPV6_MIN_MTU)
247 mtu = IPV6_MIN_MTU;
248
249 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
250 }
251
252 dst_release(dst);
253 goto tx_error;
254 }
255
256xmit:
257 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
258 skb_dst_set(skb, dst);
259 skb->dev = skb_dst(skb)->dev;
260
261 err = dst_output(tunnel->net, skb->sk, skb);
262 if (net_xmit_eval(err) == 0)
263 err = pkt_len;
264 iptunnel_xmit_stats(dev, err);
265 return NETDEV_TX_OK;
266
267tx_error_icmp:
268 dst_link_failure(skb);
269tx_error:
270 DEV_STATS_INC(dev, tx_errors);
271 kfree_skb(skb);
272 return NETDEV_TX_OK;
273}
274
275/* This function assumes it is being called from dev_queue_xmit()
276 * and that skb is filled properly by that function.
277 */
278static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
279{
280 struct ip_tunnel *tunnel = netdev_priv(dev);
281 struct flowi fl;
282
283 if (!pskb_inet_may_pull(skb))
284 goto tx_err;
285
286 memset(&fl, 0, sizeof(fl));
287
288 switch (skb->protocol) {
289 case htons(ETH_P_IP):
290 xfrm_decode_session(skb, &fl, AF_INET);
291 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
292 break;
293 case htons(ETH_P_IPV6):
294 xfrm_decode_session(skb, &fl, AF_INET6);
295 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
296 break;
297 default:
298 goto tx_err;
299 }
300
301 /* override mark with tunnel output key */
302 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
303
304 return vti_xmit(skb, dev, &fl);
305
306tx_err:
307 DEV_STATS_INC(dev, tx_errors);
308 kfree_skb(skb);
309 return NETDEV_TX_OK;
310}
311
312static int vti4_err(struct sk_buff *skb, u32 info)
313{
314 __be32 spi;
315 __u32 mark;
316 struct xfrm_state *x;
317 struct ip_tunnel *tunnel;
318 struct ip_esp_hdr *esph;
319 struct ip_auth_hdr *ah ;
320 struct ip_comp_hdr *ipch;
321 struct net *net = dev_net(skb->dev);
322 const struct iphdr *iph = (const struct iphdr *)skb->data;
323 int protocol = iph->protocol;
324 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
325
326 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
327 iph->daddr, iph->saddr, 0);
328 if (!tunnel)
329 return -1;
330
331 mark = be32_to_cpu(tunnel->parms.o_key);
332
333 switch (protocol) {
334 case IPPROTO_ESP:
335 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
336 spi = esph->spi;
337 break;
338 case IPPROTO_AH:
339 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
340 spi = ah->spi;
341 break;
342 case IPPROTO_COMP:
343 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
344 spi = htonl(ntohs(ipch->cpi));
345 break;
346 default:
347 return 0;
348 }
349
350 switch (icmp_hdr(skb)->type) {
351 case ICMP_DEST_UNREACH:
352 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
353 return 0;
354 break;
355 case ICMP_REDIRECT:
356 break;
357 default:
358 return 0;
359 }
360
361 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
362 spi, protocol, AF_INET);
363 if (!x)
364 return 0;
365
366 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
367 ipv4_update_pmtu(skb, net, info, 0, protocol);
368 else
369 ipv4_redirect(skb, net, 0, protocol);
370 xfrm_state_put(x);
371
372 return 0;
373}
374
375static int
376vti_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
377{
378 int err = 0;
379
380 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
381 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_IPIP ||
382 p->iph.ihl != 5)
383 return -EINVAL;
384 }
385
386 if (!(p->i_flags & GRE_KEY))
387 p->i_key = 0;
388 if (!(p->o_flags & GRE_KEY))
389 p->o_key = 0;
390
391 p->i_flags = VTI_ISVTI;
392
393 err = ip_tunnel_ctl(dev, p, cmd);
394 if (err)
395 return err;
396
397 if (cmd != SIOCDELTUNNEL) {
398 p->i_flags |= GRE_KEY;
399 p->o_flags |= GRE_KEY;
400 }
401 return 0;
402}
403
404static const struct net_device_ops vti_netdev_ops = {
405 .ndo_init = vti_tunnel_init,
406 .ndo_uninit = ip_tunnel_uninit,
407 .ndo_start_xmit = vti_tunnel_xmit,
408 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
409 .ndo_change_mtu = ip_tunnel_change_mtu,
410 .ndo_get_stats64 = dev_get_tstats64,
411 .ndo_get_iflink = ip_tunnel_get_iflink,
412 .ndo_tunnel_ctl = vti_tunnel_ctl,
413};
414
415static void vti_tunnel_setup(struct net_device *dev)
416{
417 dev->netdev_ops = &vti_netdev_ops;
418 dev->header_ops = &ip_tunnel_header_ops;
419 dev->type = ARPHRD_TUNNEL;
420 ip_tunnel_setup(dev, vti_net_id);
421}
422
423static int vti_tunnel_init(struct net_device *dev)
424{
425 struct ip_tunnel *tunnel = netdev_priv(dev);
426 struct iphdr *iph = &tunnel->parms.iph;
427
428 __dev_addr_set(dev, &iph->saddr, 4);
429 memcpy(dev->broadcast, &iph->daddr, 4);
430
431 dev->flags = IFF_NOARP;
432 dev->addr_len = 4;
433 dev->features |= NETIF_F_LLTX;
434 netif_keep_dst(dev);
435
436 return ip_tunnel_init(dev);
437}
438
439static void __net_init vti_fb_tunnel_init(struct net_device *dev)
440{
441 struct ip_tunnel *tunnel = netdev_priv(dev);
442 struct iphdr *iph = &tunnel->parms.iph;
443
444 iph->version = 4;
445 iph->protocol = IPPROTO_IPIP;
446 iph->ihl = 5;
447}
448
449static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
450 .handler = vti_rcv_proto,
451 .input_handler = vti_input_proto,
452 .cb_handler = vti_rcv_cb,
453 .err_handler = vti4_err,
454 .priority = 100,
455};
456
457static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
458 .handler = vti_rcv_proto,
459 .input_handler = vti_input_proto,
460 .cb_handler = vti_rcv_cb,
461 .err_handler = vti4_err,
462 .priority = 100,
463};
464
465static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
466 .handler = vti_rcv_proto,
467 .input_handler = vti_input_proto,
468 .cb_handler = vti_rcv_cb,
469 .err_handler = vti4_err,
470 .priority = 100,
471};
472
473#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
474static int vti_rcv_tunnel(struct sk_buff *skb)
475{
476 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
477 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
478
479 return vti_input(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr, 0, false);
480}
481
482static struct xfrm_tunnel vti_ipip_handler __read_mostly = {
483 .handler = vti_rcv_tunnel,
484 .cb_handler = vti_rcv_cb,
485 .err_handler = vti4_err,
486 .priority = 0,
487};
488
489#if IS_ENABLED(CONFIG_IPV6)
490static struct xfrm_tunnel vti_ipip6_handler __read_mostly = {
491 .handler = vti_rcv_tunnel,
492 .cb_handler = vti_rcv_cb,
493 .err_handler = vti4_err,
494 .priority = 0,
495};
496#endif
497#endif
498
499static int __net_init vti_init_net(struct net *net)
500{
501 int err;
502 struct ip_tunnel_net *itn;
503
504 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
505 if (err)
506 return err;
507 itn = net_generic(net, vti_net_id);
508 if (itn->fb_tunnel_dev)
509 vti_fb_tunnel_init(itn->fb_tunnel_dev);
510 return 0;
511}
512
513static void __net_exit vti_exit_batch_net(struct list_head *list_net)
514{
515 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
516}
517
518static struct pernet_operations vti_net_ops = {
519 .init = vti_init_net,
520 .exit_batch = vti_exit_batch_net,
521 .id = &vti_net_id,
522 .size = sizeof(struct ip_tunnel_net),
523};
524
525static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
526 struct netlink_ext_ack *extack)
527{
528 return 0;
529}
530
531static void vti_netlink_parms(struct nlattr *data[],
532 struct ip_tunnel_parm *parms,
533 __u32 *fwmark)
534{
535 memset(parms, 0, sizeof(*parms));
536
537 parms->iph.protocol = IPPROTO_IPIP;
538
539 if (!data)
540 return;
541
542 parms->i_flags = VTI_ISVTI;
543
544 if (data[IFLA_VTI_LINK])
545 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
546
547 if (data[IFLA_VTI_IKEY])
548 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
549
550 if (data[IFLA_VTI_OKEY])
551 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
552
553 if (data[IFLA_VTI_LOCAL])
554 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
555
556 if (data[IFLA_VTI_REMOTE])
557 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
558
559 if (data[IFLA_VTI_FWMARK])
560 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
561}
562
563static int vti_newlink(struct net *src_net, struct net_device *dev,
564 struct nlattr *tb[], struct nlattr *data[],
565 struct netlink_ext_ack *extack)
566{
567 struct ip_tunnel_parm parms;
568 __u32 fwmark = 0;
569
570 vti_netlink_parms(data, &parms, &fwmark);
571 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
572}
573
574static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
575 struct nlattr *data[],
576 struct netlink_ext_ack *extack)
577{
578 struct ip_tunnel *t = netdev_priv(dev);
579 __u32 fwmark = t->fwmark;
580 struct ip_tunnel_parm p;
581
582 vti_netlink_parms(data, &p, &fwmark);
583 return ip_tunnel_changelink(dev, tb, &p, fwmark);
584}
585
586static size_t vti_get_size(const struct net_device *dev)
587{
588 return
589 /* IFLA_VTI_LINK */
590 nla_total_size(4) +
591 /* IFLA_VTI_IKEY */
592 nla_total_size(4) +
593 /* IFLA_VTI_OKEY */
594 nla_total_size(4) +
595 /* IFLA_VTI_LOCAL */
596 nla_total_size(4) +
597 /* IFLA_VTI_REMOTE */
598 nla_total_size(4) +
599 /* IFLA_VTI_FWMARK */
600 nla_total_size(4) +
601 0;
602}
603
604static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
605{
606 struct ip_tunnel *t = netdev_priv(dev);
607 struct ip_tunnel_parm *p = &t->parms;
608
609 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
610 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
611 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
612 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
613 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
614 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
615 return -EMSGSIZE;
616
617 return 0;
618}
619
620static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
621 [IFLA_VTI_LINK] = { .type = NLA_U32 },
622 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
623 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
624 [IFLA_VTI_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
625 [IFLA_VTI_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
626 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
627};
628
629static struct rtnl_link_ops vti_link_ops __read_mostly = {
630 .kind = "vti",
631 .maxtype = IFLA_VTI_MAX,
632 .policy = vti_policy,
633 .priv_size = sizeof(struct ip_tunnel),
634 .setup = vti_tunnel_setup,
635 .validate = vti_tunnel_validate,
636 .newlink = vti_newlink,
637 .changelink = vti_changelink,
638 .dellink = ip_tunnel_dellink,
639 .get_size = vti_get_size,
640 .fill_info = vti_fill_info,
641 .get_link_net = ip_tunnel_get_link_net,
642};
643
644static int __init vti_init(void)
645{
646 const char *msg;
647 int err;
648
649 pr_info("IPv4 over IPsec tunneling driver\n");
650
651 msg = "tunnel device";
652 err = register_pernet_device(&vti_net_ops);
653 if (err < 0)
654 goto pernet_dev_failed;
655
656 msg = "tunnel protocols";
657 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
658 if (err < 0)
659 goto xfrm_proto_esp_failed;
660 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
661 if (err < 0)
662 goto xfrm_proto_ah_failed;
663 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
664 if (err < 0)
665 goto xfrm_proto_comp_failed;
666
667#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
668 msg = "ipip tunnel";
669 err = xfrm4_tunnel_register(&vti_ipip_handler, AF_INET);
670 if (err < 0)
671 goto xfrm_tunnel_ipip_failed;
672#if IS_ENABLED(CONFIG_IPV6)
673 err = xfrm4_tunnel_register(&vti_ipip6_handler, AF_INET6);
674 if (err < 0)
675 goto xfrm_tunnel_ipip6_failed;
676#endif
677#endif
678
679 msg = "netlink interface";
680 err = rtnl_link_register(&vti_link_ops);
681 if (err < 0)
682 goto rtnl_link_failed;
683
684 return err;
685
686rtnl_link_failed:
687#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
688#if IS_ENABLED(CONFIG_IPV6)
689 xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
690xfrm_tunnel_ipip6_failed:
691#endif
692 xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
693xfrm_tunnel_ipip_failed:
694#endif
695 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
696xfrm_proto_comp_failed:
697 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
698xfrm_proto_ah_failed:
699 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
700xfrm_proto_esp_failed:
701 unregister_pernet_device(&vti_net_ops);
702pernet_dev_failed:
703 pr_err("vti init: failed to register %s\n", msg);
704 return err;
705}
706
707static void __exit vti_fini(void)
708{
709 rtnl_link_unregister(&vti_link_ops);
710#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
711#if IS_ENABLED(CONFIG_IPV6)
712 xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
713#endif
714 xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
715#endif
716 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
717 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
718 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
719 unregister_pernet_device(&vti_net_ops);
720}
721
722module_init(vti_init);
723module_exit(vti_fini);
724MODULE_LICENSE("GPL");
725MODULE_ALIAS_RTNL_LINK("vti");
726MODULE_ALIAS_NETDEV("ip_vti0");