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 pcpu_sw_netstats *tstats;
99 struct xfrm_state *x;
100 const struct xfrm_mode *inner_mode;
101 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
102 u32 orig_mark = skb->mark;
103 int ret;
104
105 if (!tunnel)
106 return 1;
107
108 dev = tunnel->dev;
109
110 if (err) {
111 dev->stats.rx_errors++;
112 dev->stats.rx_dropped++;
113
114 return 0;
115 }
116
117 x = xfrm_input_state(skb);
118
119 inner_mode = &x->inner_mode;
120
121 if (x->sel.family == AF_UNSPEC) {
122 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
123 if (inner_mode == NULL) {
124 XFRM_INC_STATS(dev_net(skb->dev),
125 LINUX_MIB_XFRMINSTATEMODEERROR);
126 return -EINVAL;
127 }
128 }
129
130 family = inner_mode->family;
131
132 skb->mark = be32_to_cpu(tunnel->parms.i_key);
133 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
134 skb->mark = orig_mark;
135
136 if (!ret)
137 return -EPERM;
138
139 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
140 skb->dev = dev;
141
142 tstats = this_cpu_ptr(dev->tstats);
143
144 u64_stats_update_begin(&tstats->syncp);
145 tstats->rx_packets++;
146 tstats->rx_bytes += skb->len;
147 u64_stats_update_end(&tstats->syncp);
148
149 return 0;
150}
151
152static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
153{
154 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
155 xfrm_address_t *saddr = (xfrm_address_t *)&src;
156
157 /* if there is no transform then this tunnel is not functional.
158 * Or if the xfrm is not mode tunnel.
159 */
160 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
161 x->props.family != AF_INET)
162 return false;
163
164 if (!dst)
165 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
166
167 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
168 return false;
169
170 return true;
171}
172
173static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
174 struct flowi *fl)
175{
176 struct ip_tunnel *tunnel = netdev_priv(dev);
177 struct ip_tunnel_parm *parms = &tunnel->parms;
178 struct dst_entry *dst = skb_dst(skb);
179 struct net_device *tdev; /* Device to other host */
180 int pkt_len = skb->len;
181 int err;
182 int mtu;
183
184 if (!dst) {
185 switch (skb->protocol) {
186 case htons(ETH_P_IP): {
187 struct rtable *rt;
188
189 fl->u.ip4.flowi4_oif = dev->ifindex;
190 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
191 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
192 if (IS_ERR(rt)) {
193 dev->stats.tx_carrier_errors++;
194 goto tx_error_icmp;
195 }
196 dst = &rt->dst;
197 skb_dst_set(skb, dst);
198 break;
199 }
200#if IS_ENABLED(CONFIG_IPV6)
201 case htons(ETH_P_IPV6):
202 fl->u.ip6.flowi6_oif = dev->ifindex;
203 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
204 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
205 if (dst->error) {
206 dst_release(dst);
207 dst = NULL;
208 dev->stats.tx_carrier_errors++;
209 goto tx_error_icmp;
210 }
211 skb_dst_set(skb, dst);
212 break;
213#endif
214 default:
215 dev->stats.tx_carrier_errors++;
216 goto tx_error_icmp;
217 }
218 }
219
220 dst_hold(dst);
221 dst = xfrm_lookup_route(tunnel->net, dst, fl, NULL, 0);
222 if (IS_ERR(dst)) {
223 dev->stats.tx_carrier_errors++;
224 goto tx_error_icmp;
225 }
226
227 if (dst->flags & DST_XFRM_QUEUE)
228 goto queued;
229
230 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
231 dev->stats.tx_carrier_errors++;
232 dst_release(dst);
233 goto tx_error_icmp;
234 }
235
236 tdev = dst->dev;
237
238 if (tdev == dev) {
239 dst_release(dst);
240 dev->stats.collisions++;
241 goto tx_error;
242 }
243
244 mtu = dst_mtu(dst);
245 if (skb->len > mtu) {
246 skb_dst_update_pmtu_no_confirm(skb, mtu);
247 if (skb->protocol == htons(ETH_P_IP)) {
248 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
249 htonl(mtu));
250 } else {
251 if (mtu < IPV6_MIN_MTU)
252 mtu = IPV6_MIN_MTU;
253
254 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
255 }
256
257 dst_release(dst);
258 goto tx_error;
259 }
260
261queued:
262 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
263 skb_dst_set(skb, dst);
264 skb->dev = skb_dst(skb)->dev;
265
266 err = dst_output(tunnel->net, skb->sk, skb);
267 if (net_xmit_eval(err) == 0)
268 err = pkt_len;
269 iptunnel_xmit_stats(dev, err);
270 return NETDEV_TX_OK;
271
272tx_error_icmp:
273 dst_link_failure(skb);
274tx_error:
275 dev->stats.tx_errors++;
276 kfree_skb(skb);
277 return NETDEV_TX_OK;
278}
279
280/* This function assumes it is being called from dev_queue_xmit()
281 * and that skb is filled properly by that function.
282 */
283static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
284{
285 struct ip_tunnel *tunnel = netdev_priv(dev);
286 struct flowi fl;
287
288 if (!pskb_inet_may_pull(skb))
289 goto tx_err;
290
291 memset(&fl, 0, sizeof(fl));
292
293 switch (skb->protocol) {
294 case htons(ETH_P_IP):
295 xfrm_decode_session(skb, &fl, AF_INET);
296 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
297 break;
298 case htons(ETH_P_IPV6):
299 xfrm_decode_session(skb, &fl, AF_INET6);
300 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
301 break;
302 default:
303 goto tx_err;
304 }
305
306 /* override mark with tunnel output key */
307 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
308
309 return vti_xmit(skb, dev, &fl);
310
311tx_err:
312 dev->stats.tx_errors++;
313 kfree_skb(skb);
314 return NETDEV_TX_OK;
315}
316
317static int vti4_err(struct sk_buff *skb, u32 info)
318{
319 __be32 spi;
320 __u32 mark;
321 struct xfrm_state *x;
322 struct ip_tunnel *tunnel;
323 struct ip_esp_hdr *esph;
324 struct ip_auth_hdr *ah ;
325 struct ip_comp_hdr *ipch;
326 struct net *net = dev_net(skb->dev);
327 const struct iphdr *iph = (const struct iphdr *)skb->data;
328 int protocol = iph->protocol;
329 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
330
331 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
332 iph->daddr, iph->saddr, 0);
333 if (!tunnel)
334 return -1;
335
336 mark = be32_to_cpu(tunnel->parms.o_key);
337
338 switch (protocol) {
339 case IPPROTO_ESP:
340 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
341 spi = esph->spi;
342 break;
343 case IPPROTO_AH:
344 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
345 spi = ah->spi;
346 break;
347 case IPPROTO_COMP:
348 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
349 spi = htonl(ntohs(ipch->cpi));
350 break;
351 default:
352 return 0;
353 }
354
355 switch (icmp_hdr(skb)->type) {
356 case ICMP_DEST_UNREACH:
357 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
358 return 0;
359 case ICMP_REDIRECT:
360 break;
361 default:
362 return 0;
363 }
364
365 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
366 spi, protocol, AF_INET);
367 if (!x)
368 return 0;
369
370 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
371 ipv4_update_pmtu(skb, net, info, 0, protocol);
372 else
373 ipv4_redirect(skb, net, 0, protocol);
374 xfrm_state_put(x);
375
376 return 0;
377}
378
379static int
380vti_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
381{
382 int err = 0;
383
384 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
385 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_IPIP ||
386 p->iph.ihl != 5)
387 return -EINVAL;
388 }
389
390 if (!(p->i_flags & GRE_KEY))
391 p->i_key = 0;
392 if (!(p->o_flags & GRE_KEY))
393 p->o_key = 0;
394
395 p->i_flags = VTI_ISVTI;
396
397 err = ip_tunnel_ctl(dev, p, cmd);
398 if (err)
399 return err;
400
401 if (cmd != SIOCDELTUNNEL) {
402 p->i_flags |= GRE_KEY;
403 p->o_flags |= GRE_KEY;
404 }
405 return 0;
406}
407
408static const struct net_device_ops vti_netdev_ops = {
409 .ndo_init = vti_tunnel_init,
410 .ndo_uninit = ip_tunnel_uninit,
411 .ndo_start_xmit = vti_tunnel_xmit,
412 .ndo_do_ioctl = ip_tunnel_ioctl,
413 .ndo_change_mtu = ip_tunnel_change_mtu,
414 .ndo_get_stats64 = ip_tunnel_get_stats64,
415 .ndo_get_iflink = ip_tunnel_get_iflink,
416 .ndo_tunnel_ctl = vti_tunnel_ctl,
417};
418
419static void vti_tunnel_setup(struct net_device *dev)
420{
421 dev->netdev_ops = &vti_netdev_ops;
422 dev->header_ops = &ip_tunnel_header_ops;
423 dev->type = ARPHRD_TUNNEL;
424 ip_tunnel_setup(dev, vti_net_id);
425}
426
427static int vti_tunnel_init(struct net_device *dev)
428{
429 struct ip_tunnel *tunnel = netdev_priv(dev);
430 struct iphdr *iph = &tunnel->parms.iph;
431
432 memcpy(dev->dev_addr, &iph->saddr, 4);
433 memcpy(dev->broadcast, &iph->daddr, 4);
434
435 dev->flags = IFF_NOARP;
436 dev->addr_len = 4;
437 dev->features |= NETIF_F_LLTX;
438 netif_keep_dst(dev);
439
440 return ip_tunnel_init(dev);
441}
442
443static void __net_init vti_fb_tunnel_init(struct net_device *dev)
444{
445 struct ip_tunnel *tunnel = netdev_priv(dev);
446 struct iphdr *iph = &tunnel->parms.iph;
447
448 iph->version = 4;
449 iph->protocol = IPPROTO_IPIP;
450 iph->ihl = 5;
451}
452
453static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
454 .handler = vti_rcv_proto,
455 .input_handler = vti_input_proto,
456 .cb_handler = vti_rcv_cb,
457 .err_handler = vti4_err,
458 .priority = 100,
459};
460
461static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
462 .handler = vti_rcv_proto,
463 .input_handler = vti_input_proto,
464 .cb_handler = vti_rcv_cb,
465 .err_handler = vti4_err,
466 .priority = 100,
467};
468
469static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
470 .handler = vti_rcv_proto,
471 .input_handler = vti_input_proto,
472 .cb_handler = vti_rcv_cb,
473 .err_handler = vti4_err,
474 .priority = 100,
475};
476
477#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
478static int vti_rcv_tunnel(struct sk_buff *skb)
479{
480 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
481 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
482
483 return vti_input(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr, 0, false);
484}
485
486static struct xfrm_tunnel vti_ipip_handler __read_mostly = {
487 .handler = vti_rcv_tunnel,
488 .cb_handler = vti_rcv_cb,
489 .err_handler = vti4_err,
490 .priority = 0,
491};
492
493#if IS_ENABLED(CONFIG_IPV6)
494static struct xfrm_tunnel vti_ipip6_handler __read_mostly = {
495 .handler = vti_rcv_tunnel,
496 .cb_handler = vti_rcv_cb,
497 .err_handler = vti4_err,
498 .priority = 0,
499};
500#endif
501#endif
502
503static int __net_init vti_init_net(struct net *net)
504{
505 int err;
506 struct ip_tunnel_net *itn;
507
508 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
509 if (err)
510 return err;
511 itn = net_generic(net, vti_net_id);
512 if (itn->fb_tunnel_dev)
513 vti_fb_tunnel_init(itn->fb_tunnel_dev);
514 return 0;
515}
516
517static void __net_exit vti_exit_batch_net(struct list_head *list_net)
518{
519 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
520}
521
522static struct pernet_operations vti_net_ops = {
523 .init = vti_init_net,
524 .exit_batch = vti_exit_batch_net,
525 .id = &vti_net_id,
526 .size = sizeof(struct ip_tunnel_net),
527};
528
529static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
530 struct netlink_ext_ack *extack)
531{
532 return 0;
533}
534
535static void vti_netlink_parms(struct nlattr *data[],
536 struct ip_tunnel_parm *parms,
537 __u32 *fwmark)
538{
539 memset(parms, 0, sizeof(*parms));
540
541 parms->iph.protocol = IPPROTO_IPIP;
542
543 if (!data)
544 return;
545
546 parms->i_flags = VTI_ISVTI;
547
548 if (data[IFLA_VTI_LINK])
549 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
550
551 if (data[IFLA_VTI_IKEY])
552 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
553
554 if (data[IFLA_VTI_OKEY])
555 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
556
557 if (data[IFLA_VTI_LOCAL])
558 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
559
560 if (data[IFLA_VTI_REMOTE])
561 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
562
563 if (data[IFLA_VTI_FWMARK])
564 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
565}
566
567static int vti_newlink(struct net *src_net, struct net_device *dev,
568 struct nlattr *tb[], struct nlattr *data[],
569 struct netlink_ext_ack *extack)
570{
571 struct ip_tunnel_parm parms;
572 __u32 fwmark = 0;
573
574 vti_netlink_parms(data, &parms, &fwmark);
575 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
576}
577
578static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
579 struct nlattr *data[],
580 struct netlink_ext_ack *extack)
581{
582 struct ip_tunnel *t = netdev_priv(dev);
583 __u32 fwmark = t->fwmark;
584 struct ip_tunnel_parm p;
585
586 vti_netlink_parms(data, &p, &fwmark);
587 return ip_tunnel_changelink(dev, tb, &p, fwmark);
588}
589
590static size_t vti_get_size(const struct net_device *dev)
591{
592 return
593 /* IFLA_VTI_LINK */
594 nla_total_size(4) +
595 /* IFLA_VTI_IKEY */
596 nla_total_size(4) +
597 /* IFLA_VTI_OKEY */
598 nla_total_size(4) +
599 /* IFLA_VTI_LOCAL */
600 nla_total_size(4) +
601 /* IFLA_VTI_REMOTE */
602 nla_total_size(4) +
603 /* IFLA_VTI_FWMARK */
604 nla_total_size(4) +
605 0;
606}
607
608static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
609{
610 struct ip_tunnel *t = netdev_priv(dev);
611 struct ip_tunnel_parm *p = &t->parms;
612
613 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
614 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
615 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
616 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
617 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
618 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
619 return -EMSGSIZE;
620
621 return 0;
622}
623
624static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
625 [IFLA_VTI_LINK] = { .type = NLA_U32 },
626 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
627 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
628 [IFLA_VTI_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
629 [IFLA_VTI_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
630 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
631};
632
633static struct rtnl_link_ops vti_link_ops __read_mostly = {
634 .kind = "vti",
635 .maxtype = IFLA_VTI_MAX,
636 .policy = vti_policy,
637 .priv_size = sizeof(struct ip_tunnel),
638 .setup = vti_tunnel_setup,
639 .validate = vti_tunnel_validate,
640 .newlink = vti_newlink,
641 .changelink = vti_changelink,
642 .dellink = ip_tunnel_dellink,
643 .get_size = vti_get_size,
644 .fill_info = vti_fill_info,
645 .get_link_net = ip_tunnel_get_link_net,
646};
647
648static int __init vti_init(void)
649{
650 const char *msg;
651 int err;
652
653 pr_info("IPv4 over IPsec tunneling driver\n");
654
655 msg = "tunnel device";
656 err = register_pernet_device(&vti_net_ops);
657 if (err < 0)
658 goto pernet_dev_failed;
659
660 msg = "tunnel protocols";
661 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
662 if (err < 0)
663 goto xfrm_proto_esp_failed;
664 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
665 if (err < 0)
666 goto xfrm_proto_ah_failed;
667 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
668 if (err < 0)
669 goto xfrm_proto_comp_failed;
670
671#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
672 msg = "ipip tunnel";
673 err = xfrm4_tunnel_register(&vti_ipip_handler, AF_INET);
674 if (err < 0)
675 goto xfrm_tunnel_ipip_failed;
676#if IS_ENABLED(CONFIG_IPV6)
677 err = xfrm4_tunnel_register(&vti_ipip6_handler, AF_INET6);
678 if (err < 0)
679 goto xfrm_tunnel_ipip6_failed;
680#endif
681#endif
682
683 msg = "netlink interface";
684 err = rtnl_link_register(&vti_link_ops);
685 if (err < 0)
686 goto rtnl_link_failed;
687
688 return err;
689
690rtnl_link_failed:
691#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
692#if IS_ENABLED(CONFIG_IPV6)
693 xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
694xfrm_tunnel_ipip6_failed:
695#endif
696 xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
697xfrm_tunnel_ipip_failed:
698#endif
699 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
700xfrm_proto_comp_failed:
701 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
702xfrm_proto_ah_failed:
703 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
704xfrm_proto_esp_failed:
705 unregister_pernet_device(&vti_net_ops);
706pernet_dev_failed:
707 pr_err("vti init: failed to register %s\n", msg);
708 return err;
709}
710
711static void __exit vti_fini(void)
712{
713 rtnl_link_unregister(&vti_link_ops);
714#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
715#if IS_ENABLED(CONFIG_IPV6)
716 xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
717#endif
718 xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
719#endif
720 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
721 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
722 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
723 unregister_pernet_device(&vti_net_ops);
724}
725
726module_init(vti_init);
727module_exit(vti_fini);
728MODULE_LICENSE("GPL");
729MODULE_ALIAS_RTNL_LINK("vti");
730MODULE_ALIAS_NETDEV("ip_vti0");