?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / net / ipv4 / ip_vti.c
blob51b1669334fe6baeea0045fcfdd631700c1ccbf2
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
47 static struct rtnl_link_ops vti_link_ops __read_mostly;
48
49 static unsigned int vti_net_id __read_mostly;
50 static int vti_tunnel_init(struct net_device *dev);
51
52 static 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;
72 drop:
73 kfree_skb(skb);
74 return 0;
75 }
76
77 static 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
85 static 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
143 static 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
164 static 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
228 tx_error_icmp:
229 dst_link_failure(skb);
230 tx_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 */
239 static 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
267 static 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
329 static int
330 vti_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
365 static 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
375 static 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
382 static 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->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
391 dev->mtu = ETH_DATA_LEN;
392 dev->flags = IFF_NOARP;
393 dev->addr_len = 4;
394 dev->features |= NETIF_F_LLTX;
395 netif_keep_dst(dev);
396
397 return ip_tunnel_init(dev);
398 }
399
400 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
401 {
402 struct ip_tunnel *tunnel = netdev_priv(dev);
403 struct iphdr *iph = &tunnel->parms.iph;
404
405 iph->version = 4;
406 iph->protocol = IPPROTO_IPIP;
407 iph->ihl = 5;
408 }
409
410 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
411 .handler = vti_rcv,
412 .input_handler = vti_input,
413 .cb_handler = vti_rcv_cb,
414 .err_handler = vti4_err,
415 .priority = 100,
416 };
417
418 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
419 .handler = vti_rcv,
420 .input_handler = vti_input,
421 .cb_handler = vti_rcv_cb,
422 .err_handler = vti4_err,
423 .priority = 100,
424 };
425
426 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
427 .handler = vti_rcv,
428 .input_handler = vti_input,
429 .cb_handler = vti_rcv_cb,
430 .err_handler = vti4_err,
431 .priority = 100,
432 };
433
434 static int __net_init vti_init_net(struct net *net)
435 {
436 int err;
437 struct ip_tunnel_net *itn;
438
439 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
440 if (err)
441 return err;
442 itn = net_generic(net, vti_net_id);
443 vti_fb_tunnel_init(itn->fb_tunnel_dev);
444 return 0;
445 }
446
447 static void __net_exit vti_exit_batch_net(struct list_head *list_net)
448 {
449 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
450 }
451
452 static struct pernet_operations vti_net_ops = {
453 .init = vti_init_net,
454 .exit_batch = vti_exit_batch_net,
455 .id = &vti_net_id,
456 .size = sizeof(struct ip_tunnel_net),
457 };
458
459 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
460 struct netlink_ext_ack *extack)
461 {
462 return 0;
463 }
464
465 static void vti_netlink_parms(struct nlattr *data[],
466 struct ip_tunnel_parm *parms,
467 __u32 *fwmark)
468 {
469 memset(parms, 0, sizeof(*parms));
470
471 parms->iph.protocol = IPPROTO_IPIP;
472
473 if (!data)
474 return;
475
476 parms->i_flags = VTI_ISVTI;
477
478 if (data[IFLA_VTI_LINK])
479 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
480
481 if (data[IFLA_VTI_IKEY])
482 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
483
484 if (data[IFLA_VTI_OKEY])
485 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
486
487 if (data[IFLA_VTI_LOCAL])
488 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
489
490 if (data[IFLA_VTI_REMOTE])
491 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
492
493 if (data[IFLA_VTI_FWMARK])
494 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
495 }
496
497 static int vti_newlink(struct net *src_net, struct net_device *dev,
498 struct nlattr *tb[], struct nlattr *data[],
499 struct netlink_ext_ack *extack)
500 {
501 struct ip_tunnel_parm parms;
502 __u32 fwmark = 0;
503
504 vti_netlink_parms(data, &parms, &fwmark);
505 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
506 }
507
508 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
509 struct nlattr *data[],
510 struct netlink_ext_ack *extack)
511 {
512 struct ip_tunnel *t = netdev_priv(dev);
513 __u32 fwmark = t->fwmark;
514 struct ip_tunnel_parm p;
515
516 vti_netlink_parms(data, &p, &fwmark);
517 return ip_tunnel_changelink(dev, tb, &p, fwmark);
518 }
519
520 static size_t vti_get_size(const struct net_device *dev)
521 {
522 return
523 /* IFLA_VTI_LINK */
524 nla_total_size(4) +
525 /* IFLA_VTI_IKEY */
526 nla_total_size(4) +
527 /* IFLA_VTI_OKEY */
528 nla_total_size(4) +
529 /* IFLA_VTI_LOCAL */
530 nla_total_size(4) +
531 /* IFLA_VTI_REMOTE */
532 nla_total_size(4) +
533 /* IFLA_VTI_FWMARK */
534 nla_total_size(4) +
535 0;
536 }
537
538 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
539 {
540 struct ip_tunnel *t = netdev_priv(dev);
541 struct ip_tunnel_parm *p = &t->parms;
542
543 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
544 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
545 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
546 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
547 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
548 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
549 return -EMSGSIZE;
550
551 return 0;
552 }
553
554 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
555 [IFLA_VTI_LINK] = { .type = NLA_U32 },
556 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
557 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
558 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
559 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
560 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
561 };
562
563 static struct rtnl_link_ops vti_link_ops __read_mostly = {
564 .kind = "vti",
565 .maxtype = IFLA_VTI_MAX,
566 .policy = vti_policy,
567 .priv_size = sizeof(struct ip_tunnel),
568 .setup = vti_tunnel_setup,
569 .validate = vti_tunnel_validate,
570 .newlink = vti_newlink,
571 .changelink = vti_changelink,
572 .dellink = ip_tunnel_dellink,
573 .get_size = vti_get_size,
574 .fill_info = vti_fill_info,
575 .get_link_net = ip_tunnel_get_link_net,
576 };
577
578 static int __init vti_init(void)
579 {
580 const char *msg;
581 int err;
582
583 pr_info("IPv4 over IPsec tunneling driver\n");
584
585 msg = "tunnel device";
586 err = register_pernet_device(&vti_net_ops);
587 if (err < 0)
588 goto pernet_dev_failed;
589
590 msg = "tunnel protocols";
591 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
592 if (err < 0)
593 goto xfrm_proto_esp_failed;
594 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
595 if (err < 0)
596 goto xfrm_proto_ah_failed;
597 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
598 if (err < 0)
599 goto xfrm_proto_comp_failed;
600
601 msg = "netlink interface";
602 err = rtnl_link_register(&vti_link_ops);
603 if (err < 0)
604 goto rtnl_link_failed;
605
606 return err;
607
608 rtnl_link_failed:
609 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
610 xfrm_proto_comp_failed:
611 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
612 xfrm_proto_ah_failed:
613 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
614 xfrm_proto_esp_failed:
615 unregister_pernet_device(&vti_net_ops);
616 pernet_dev_failed:
617 pr_err("vti init: failed to register %s\n", msg);
618 return err;
619 }
620
621 static void __exit vti_fini(void)
622 {
623 rtnl_link_unregister(&vti_link_ops);
624 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
625 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
626 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
627 unregister_pernet_device(&vti_net_ops);
628 }
629
630 module_init(vti_init);
631 module_exit(vti_fini);
632 MODULE_LICENSE("GPL");
633 MODULE_ALIAS_RTNL_LINK("vti");
634 MODULE_ALIAS_NETDEV("ip_vti0");
CRIS linux kernel tree