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/*
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/init.h>
34#include <linux/netfilter_ipv4.h>
35#include <linux/if_ether.h>
36#include <linux/icmpv6.h>
37
38#include <net/sock.h>
39#include <net/ip.h>
40#include <net/icmp.h>
41#include <net/ip_tunnels.h>
42#include <net/inet_ecn.h>
43#include <net/xfrm.h>
44#include <net/net_namespace.h>
45#include <net/netns/generic.h>
46
47static struct rtnl_link_ops vti_link_ops __read_mostly;
48
49static unsigned int vti_net_id __read_mostly;
50static int vti_tunnel_init(struct net_device *dev);
51
52static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
53 int encap_type)
54{
55 struct ip_tunnel *tunnel;
56 const struct iphdr *iph = ip_hdr(skb);
57 struct net *net = dev_net(skb->dev);
58 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
59
60 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
61 iph->saddr, iph->daddr, 0);
62 if (tunnel) {
63 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
64 goto drop;
65
66 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
67
68 return xfrm_input(skb, nexthdr, spi, encap_type);
69 }
70
71 return -EINVAL;
72drop:
73 kfree_skb(skb);
74 return 0;
75}
76
77static int vti_rcv(struct sk_buff *skb)
78{
79 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
80 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
81
82 return vti_input(skb, ip_hdr(skb)->protocol, 0, 0);
83}
84
85static int vti_rcv_cb(struct sk_buff *skb, int err)
86{
87 unsigned short family;
88 struct net_device *dev;
89 struct pcpu_sw_netstats *tstats;
90 struct xfrm_state *x;
91 struct xfrm_mode *inner_mode;
92 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
93 u32 orig_mark = skb->mark;
94 int ret;
95
96 if (!tunnel)
97 return 1;
98
99 dev = tunnel->dev;
100
101 if (err) {
102 dev->stats.rx_errors++;
103 dev->stats.rx_dropped++;
104
105 return 0;
106 }
107
108 x = xfrm_input_state(skb);
109
110 inner_mode = x->inner_mode;
111
112 if (x->sel.family == AF_UNSPEC) {
113 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
114 if (inner_mode == NULL) {
115 XFRM_INC_STATS(dev_net(skb->dev),
116 LINUX_MIB_XFRMINSTATEMODEERROR);
117 return -EINVAL;
118 }
119 }
120
121 family = inner_mode->afinfo->family;
122
123 skb->mark = be32_to_cpu(tunnel->parms.i_key);
124 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
125 skb->mark = orig_mark;
126
127 if (!ret)
128 return -EPERM;
129
130 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
131 skb->dev = dev;
132
133 tstats = this_cpu_ptr(dev->tstats);
134
135 u64_stats_update_begin(&tstats->syncp);
136 tstats->rx_packets++;
137 tstats->rx_bytes += skb->len;
138 u64_stats_update_end(&tstats->syncp);
139
140 return 0;
141}
142
143static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
144{
145 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
146 xfrm_address_t *saddr = (xfrm_address_t *)&src;
147
148 /* if there is no transform then this tunnel is not functional.
149 * Or if the xfrm is not mode tunnel.
150 */
151 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
152 x->props.family != AF_INET)
153 return false;
154
155 if (!dst)
156 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
157
158 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
159 return false;
160
161 return true;
162}
163
164static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
165 struct flowi *fl)
166{
167 struct ip_tunnel *tunnel = netdev_priv(dev);
168 struct ip_tunnel_parm *parms = &tunnel->parms;
169 struct dst_entry *dst = skb_dst(skb);
170 struct net_device *tdev; /* Device to other host */
171 int pkt_len = skb->len;
172 int err;
173 int mtu;
174
175 if (!dst) {
176 dev->stats.tx_carrier_errors++;
177 goto tx_error_icmp;
178 }
179
180 dst_hold(dst);
181 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
182 if (IS_ERR(dst)) {
183 dev->stats.tx_carrier_errors++;
184 goto tx_error_icmp;
185 }
186
187 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
188 dev->stats.tx_carrier_errors++;
189 dst_release(dst);
190 goto tx_error_icmp;
191 }
192
193 tdev = dst->dev;
194
195 if (tdev == dev) {
196 dst_release(dst);
197 dev->stats.collisions++;
198 goto tx_error;
199 }
200
201 mtu = dst_mtu(dst);
202 if (skb->len > mtu) {
203 skb_dst_update_pmtu(skb, mtu);
204 if (skb->protocol == htons(ETH_P_IP)) {
205 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
206 htonl(mtu));
207 } else {
208 if (mtu < IPV6_MIN_MTU)
209 mtu = IPV6_MIN_MTU;
210
211 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
212 }
213
214 dst_release(dst);
215 goto tx_error;
216 }
217
218 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
219 skb_dst_set(skb, dst);
220 skb->dev = skb_dst(skb)->dev;
221
222 err = dst_output(tunnel->net, skb->sk, skb);
223 if (net_xmit_eval(err) == 0)
224 err = pkt_len;
225 iptunnel_xmit_stats(dev, err);
226 return NETDEV_TX_OK;
227
228tx_error_icmp:
229 dst_link_failure(skb);
230tx_error:
231 dev->stats.tx_errors++;
232 kfree_skb(skb);
233 return NETDEV_TX_OK;
234}
235
236/* This function assumes it is being called from dev_queue_xmit()
237 * and that skb is filled properly by that function.
238 */
239static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
240{
241 struct ip_tunnel *tunnel = netdev_priv(dev);
242 struct flowi fl;
243
244 memset(&fl, 0, sizeof(fl));
245
246 switch (skb->protocol) {
247 case htons(ETH_P_IP):
248 xfrm_decode_session(skb, &fl, AF_INET);
249 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
250 break;
251 case htons(ETH_P_IPV6):
252 xfrm_decode_session(skb, &fl, AF_INET6);
253 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
254 break;
255 default:
256 dev->stats.tx_errors++;
257 dev_kfree_skb(skb);
258 return NETDEV_TX_OK;
259 }
260
261 /* override mark with tunnel output key */
262 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
263
264 return vti_xmit(skb, dev, &fl);
265}
266
267static int vti4_err(struct sk_buff *skb, u32 info)
268{
269 __be32 spi;
270 __u32 mark;
271 struct xfrm_state *x;
272 struct ip_tunnel *tunnel;
273 struct ip_esp_hdr *esph;
274 struct ip_auth_hdr *ah ;
275 struct ip_comp_hdr *ipch;
276 struct net *net = dev_net(skb->dev);
277 const struct iphdr *iph = (const struct iphdr *)skb->data;
278 int protocol = iph->protocol;
279 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
280
281 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
282 iph->daddr, iph->saddr, 0);
283 if (!tunnel)
284 return -1;
285
286 mark = be32_to_cpu(tunnel->parms.o_key);
287
288 switch (protocol) {
289 case IPPROTO_ESP:
290 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
291 spi = esph->spi;
292 break;
293 case IPPROTO_AH:
294 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
295 spi = ah->spi;
296 break;
297 case IPPROTO_COMP:
298 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
299 spi = htonl(ntohs(ipch->cpi));
300 break;
301 default:
302 return 0;
303 }
304
305 switch (icmp_hdr(skb)->type) {
306 case ICMP_DEST_UNREACH:
307 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
308 return 0;
309 case ICMP_REDIRECT:
310 break;
311 default:
312 return 0;
313 }
314
315 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
316 spi, protocol, AF_INET);
317 if (!x)
318 return 0;
319
320 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
321 ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
322 else
323 ipv4_redirect(skb, net, 0, 0, protocol, 0);
324 xfrm_state_put(x);
325
326 return 0;
327}
328
329static int
330vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
331{
332 int err = 0;
333 struct ip_tunnel_parm p;
334
335 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
336 return -EFAULT;
337
338 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
339 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
340 p.iph.ihl != 5)
341 return -EINVAL;
342 }
343
344 if (!(p.i_flags & GRE_KEY))
345 p.i_key = 0;
346 if (!(p.o_flags & GRE_KEY))
347 p.o_key = 0;
348
349 p.i_flags = VTI_ISVTI;
350
351 err = ip_tunnel_ioctl(dev, &p, cmd);
352 if (err)
353 return err;
354
355 if (cmd != SIOCDELTUNNEL) {
356 p.i_flags |= GRE_KEY;
357 p.o_flags |= GRE_KEY;
358 }
359
360 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
361 return -EFAULT;
362 return 0;
363}
364
365static const struct net_device_ops vti_netdev_ops = {
366 .ndo_init = vti_tunnel_init,
367 .ndo_uninit = ip_tunnel_uninit,
368 .ndo_start_xmit = vti_tunnel_xmit,
369 .ndo_do_ioctl = vti_tunnel_ioctl,
370 .ndo_change_mtu = ip_tunnel_change_mtu,
371 .ndo_get_stats64 = ip_tunnel_get_stats64,
372 .ndo_get_iflink = ip_tunnel_get_iflink,
373};
374
375static void vti_tunnel_setup(struct net_device *dev)
376{
377 dev->netdev_ops = &vti_netdev_ops;
378 dev->type = ARPHRD_TUNNEL;
379 ip_tunnel_setup(dev, vti_net_id);
380}
381
382static int vti_tunnel_init(struct net_device *dev)
383{
384 struct ip_tunnel *tunnel = netdev_priv(dev);
385 struct iphdr *iph = &tunnel->parms.iph;
386
387 memcpy(dev->dev_addr, &iph->saddr, 4);
388 memcpy(dev->broadcast, &iph->daddr, 4);
389
390 dev->flags = IFF_NOARP;
391 dev->addr_len = 4;
392 dev->features |= NETIF_F_LLTX;
393 netif_keep_dst(dev);
394
395 return ip_tunnel_init(dev);
396}
397
398static void __net_init vti_fb_tunnel_init(struct net_device *dev)
399{
400 struct ip_tunnel *tunnel = netdev_priv(dev);
401 struct iphdr *iph = &tunnel->parms.iph;
402
403 iph->version = 4;
404 iph->protocol = IPPROTO_IPIP;
405 iph->ihl = 5;
406}
407
408static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
409 .handler = vti_rcv,
410 .input_handler = vti_input,
411 .cb_handler = vti_rcv_cb,
412 .err_handler = vti4_err,
413 .priority = 100,
414};
415
416static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
417 .handler = vti_rcv,
418 .input_handler = vti_input,
419 .cb_handler = vti_rcv_cb,
420 .err_handler = vti4_err,
421 .priority = 100,
422};
423
424static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
425 .handler = vti_rcv,
426 .input_handler = vti_input,
427 .cb_handler = vti_rcv_cb,
428 .err_handler = vti4_err,
429 .priority = 100,
430};
431
432static int __net_init vti_init_net(struct net *net)
433{
434 int err;
435 struct ip_tunnel_net *itn;
436
437 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
438 if (err)
439 return err;
440 itn = net_generic(net, vti_net_id);
441 vti_fb_tunnel_init(itn->fb_tunnel_dev);
442 return 0;
443}
444
445static void __net_exit vti_exit_batch_net(struct list_head *list_net)
446{
447 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
448}
449
450static struct pernet_operations vti_net_ops = {
451 .init = vti_init_net,
452 .exit_batch = vti_exit_batch_net,
453 .id = &vti_net_id,
454 .size = sizeof(struct ip_tunnel_net),
455};
456
457static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
458 struct netlink_ext_ack *extack)
459{
460 return 0;
461}
462
463static void vti_netlink_parms(struct nlattr *data[],
464 struct ip_tunnel_parm *parms,
465 __u32 *fwmark)
466{
467 memset(parms, 0, sizeof(*parms));
468
469 parms->iph.protocol = IPPROTO_IPIP;
470
471 if (!data)
472 return;
473
474 parms->i_flags = VTI_ISVTI;
475
476 if (data[IFLA_VTI_LINK])
477 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
478
479 if (data[IFLA_VTI_IKEY])
480 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
481
482 if (data[IFLA_VTI_OKEY])
483 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
484
485 if (data[IFLA_VTI_LOCAL])
486 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
487
488 if (data[IFLA_VTI_REMOTE])
489 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
490
491 if (data[IFLA_VTI_FWMARK])
492 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
493}
494
495static int vti_newlink(struct net *src_net, struct net_device *dev,
496 struct nlattr *tb[], struct nlattr *data[],
497 struct netlink_ext_ack *extack)
498{
499 struct ip_tunnel_parm parms;
500 __u32 fwmark = 0;
501
502 vti_netlink_parms(data, &parms, &fwmark);
503 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
504}
505
506static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
507 struct nlattr *data[],
508 struct netlink_ext_ack *extack)
509{
510 struct ip_tunnel *t = netdev_priv(dev);
511 __u32 fwmark = t->fwmark;
512 struct ip_tunnel_parm p;
513
514 vti_netlink_parms(data, &p, &fwmark);
515 return ip_tunnel_changelink(dev, tb, &p, fwmark);
516}
517
518static size_t vti_get_size(const struct net_device *dev)
519{
520 return
521 /* IFLA_VTI_LINK */
522 nla_total_size(4) +
523 /* IFLA_VTI_IKEY */
524 nla_total_size(4) +
525 /* IFLA_VTI_OKEY */
526 nla_total_size(4) +
527 /* IFLA_VTI_LOCAL */
528 nla_total_size(4) +
529 /* IFLA_VTI_REMOTE */
530 nla_total_size(4) +
531 /* IFLA_VTI_FWMARK */
532 nla_total_size(4) +
533 0;
534}
535
536static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
537{
538 struct ip_tunnel *t = netdev_priv(dev);
539 struct ip_tunnel_parm *p = &t->parms;
540
541 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
542 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
543 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
544 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
545 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
546 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
547 return -EMSGSIZE;
548
549 return 0;
550}
551
552static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
553 [IFLA_VTI_LINK] = { .type = NLA_U32 },
554 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
555 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
556 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
557 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
558 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
559};
560
561static struct rtnl_link_ops vti_link_ops __read_mostly = {
562 .kind = "vti",
563 .maxtype = IFLA_VTI_MAX,
564 .policy = vti_policy,
565 .priv_size = sizeof(struct ip_tunnel),
566 .setup = vti_tunnel_setup,
567 .validate = vti_tunnel_validate,
568 .newlink = vti_newlink,
569 .changelink = vti_changelink,
570 .dellink = ip_tunnel_dellink,
571 .get_size = vti_get_size,
572 .fill_info = vti_fill_info,
573 .get_link_net = ip_tunnel_get_link_net,
574};
575
576static int __init vti_init(void)
577{
578 const char *msg;
579 int err;
580
581 pr_info("IPv4 over IPsec tunneling driver\n");
582
583 msg = "tunnel device";
584 err = register_pernet_device(&vti_net_ops);
585 if (err < 0)
586 goto pernet_dev_failed;
587
588 msg = "tunnel protocols";
589 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
590 if (err < 0)
591 goto xfrm_proto_esp_failed;
592 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
593 if (err < 0)
594 goto xfrm_proto_ah_failed;
595 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
596 if (err < 0)
597 goto xfrm_proto_comp_failed;
598
599 msg = "netlink interface";
600 err = rtnl_link_register(&vti_link_ops);
601 if (err < 0)
602 goto rtnl_link_failed;
603
604 return err;
605
606rtnl_link_failed:
607 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
608xfrm_proto_comp_failed:
609 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
610xfrm_proto_ah_failed:
611 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
612xfrm_proto_esp_failed:
613 unregister_pernet_device(&vti_net_ops);
614pernet_dev_failed:
615 pr_err("vti init: failed to register %s\n", msg);
616 return err;
617}
618
619static void __exit vti_fini(void)
620{
621 rtnl_link_unregister(&vti_link_ops);
622 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
623 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
624 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
625 unregister_pernet_device(&vti_net_ops);
626}
627
628module_init(vti_init);
629module_exit(vti_fini);
630MODULE_LICENSE("GPL");
631MODULE_ALIAS_RTNL_LINK("vti");
632MODULE_ALIAS_NETDEV("ip_vti0");