?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
netfilter: nf_nat: don't bug when mapping already exists
[linux-2.6/btrfs-unstable.git] / net / ipv4 / ip_vti.c
blob5ed63d25095062d44dacfd291e227290d24ea0ed
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 err;
172 int mtu;
173
174 if (!dst) {
175 dev->stats.tx_carrier_errors++;
176 goto tx_error_icmp;
177 }
178
179 dst_hold(dst);
180 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
181 if (IS_ERR(dst)) {
182 dev->stats.tx_carrier_errors++;
183 goto tx_error_icmp;
184 }
185
186 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
187 dev->stats.tx_carrier_errors++;
188 dst_release(dst);
189 goto tx_error_icmp;
190 }
191
192 tdev = dst->dev;
193
194 if (tdev == dev) {
195 dst_release(dst);
196 dev->stats.collisions++;
197 goto tx_error;
198 }
199
200 if (tunnel->err_count > 0) {
201 if (time_before(jiffies,
202 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
203 tunnel->err_count--;
204 dst_link_failure(skb);
205 } else
206 tunnel->err_count = 0;
207 }
208
209 mtu = dst_mtu(dst);
210 if (skb->len > mtu) {
211 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
212 if (skb->protocol == htons(ETH_P_IP)) {
213 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
214 htonl(mtu));
215 } else {
216 if (mtu < IPV6_MIN_MTU)
217 mtu = IPV6_MIN_MTU;
218
219 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
220 }
221
222 dst_release(dst);
223 goto tx_error;
224 }
225
226 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
227 skb_dst_set(skb, dst);
228 skb->dev = skb_dst(skb)->dev;
229
230 err = dst_output(tunnel->net, skb->sk, skb);
231 if (net_xmit_eval(err) == 0)
232 err = skb->len;
233 iptunnel_xmit_stats(dev, err);
234 return NETDEV_TX_OK;
235
236 tx_error_icmp:
237 dst_link_failure(skb);
238 tx_error:
239 dev->stats.tx_errors++;
240 kfree_skb(skb);
241 return NETDEV_TX_OK;
242 }
243
244 /* This function assumes it is being called from dev_queue_xmit()
245 * and that skb is filled properly by that function.
246 */
247 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
248 {
249 struct ip_tunnel *tunnel = netdev_priv(dev);
250 struct flowi fl;
251
252 memset(&fl, 0, sizeof(fl));
253
254 switch (skb->protocol) {
255 case htons(ETH_P_IP):
256 xfrm_decode_session(skb, &fl, AF_INET);
257 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
258 break;
259 case htons(ETH_P_IPV6):
260 xfrm_decode_session(skb, &fl, AF_INET6);
261 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
262 break;
263 default:
264 dev->stats.tx_errors++;
265 dev_kfree_skb(skb);
266 return NETDEV_TX_OK;
267 }
268
269 /* override mark with tunnel output key */
270 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
271
272 return vti_xmit(skb, dev, &fl);
273 }
274
275 static int vti4_err(struct sk_buff *skb, u32 info)
276 {
277 __be32 spi;
278 __u32 mark;
279 struct xfrm_state *x;
280 struct ip_tunnel *tunnel;
281 struct ip_esp_hdr *esph;
282 struct ip_auth_hdr *ah ;
283 struct ip_comp_hdr *ipch;
284 struct net *net = dev_net(skb->dev);
285 const struct iphdr *iph = (const struct iphdr *)skb->data;
286 int protocol = iph->protocol;
287 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
288
289 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
290 iph->daddr, iph->saddr, 0);
291 if (!tunnel)
292 return -1;
293
294 mark = be32_to_cpu(tunnel->parms.o_key);
295
296 switch (protocol) {
297 case IPPROTO_ESP:
298 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
299 spi = esph->spi;
300 break;
301 case IPPROTO_AH:
302 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
303 spi = ah->spi;
304 break;
305 case IPPROTO_COMP:
306 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
307 spi = htonl(ntohs(ipch->cpi));
308 break;
309 default:
310 return 0;
311 }
312
313 switch (icmp_hdr(skb)->type) {
314 case ICMP_DEST_UNREACH:
315 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
316 return 0;
317 case ICMP_REDIRECT:
318 break;
319 default:
320 return 0;
321 }
322
323 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
324 spi, protocol, AF_INET);
325 if (!x)
326 return 0;
327
328 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
329 ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
330 else
331 ipv4_redirect(skb, net, 0, 0, protocol, 0);
332 xfrm_state_put(x);
333
334 return 0;
335 }
336
337 static int
338 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
339 {
340 int err = 0;
341 struct ip_tunnel_parm p;
342
343 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
344 return -EFAULT;
345
346 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
347 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
348 p.iph.ihl != 5)
349 return -EINVAL;
350 }
351
352 if (!(p.i_flags & GRE_KEY))
353 p.i_key = 0;
354 if (!(p.o_flags & GRE_KEY))
355 p.o_key = 0;
356
357 p.i_flags = VTI_ISVTI;
358
359 err = ip_tunnel_ioctl(dev, &p, cmd);
360 if (err)
361 return err;
362
363 if (cmd != SIOCDELTUNNEL) {
364 p.i_flags |= GRE_KEY;
365 p.o_flags |= GRE_KEY;
366 }
367
368 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
369 return -EFAULT;
370 return 0;
371 }
372
373 static const struct net_device_ops vti_netdev_ops = {
374 .ndo_init = vti_tunnel_init,
375 .ndo_uninit = ip_tunnel_uninit,
376 .ndo_start_xmit = vti_tunnel_xmit,
377 .ndo_do_ioctl = vti_tunnel_ioctl,
378 .ndo_change_mtu = ip_tunnel_change_mtu,
379 .ndo_get_stats64 = ip_tunnel_get_stats64,
380 .ndo_get_iflink = ip_tunnel_get_iflink,
381 };
382
383 static void vti_tunnel_setup(struct net_device *dev)
384 {
385 dev->netdev_ops = &vti_netdev_ops;
386 dev->type = ARPHRD_TUNNEL;
387 ip_tunnel_setup(dev, vti_net_id);
388 }
389
390 static int vti_tunnel_init(struct net_device *dev)
391 {
392 struct ip_tunnel *tunnel = netdev_priv(dev);
393 struct iphdr *iph = &tunnel->parms.iph;
394
395 memcpy(dev->dev_addr, &iph->saddr, 4);
396 memcpy(dev->broadcast, &iph->daddr, 4);
397
398 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
399 dev->mtu = ETH_DATA_LEN;
400 dev->flags = IFF_NOARP;
401 dev->addr_len = 4;
402 dev->features |= NETIF_F_LLTX;
403 netif_keep_dst(dev);
404
405 return ip_tunnel_init(dev);
406 }
407
408 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
409 {
410 struct ip_tunnel *tunnel = netdev_priv(dev);
411 struct iphdr *iph = &tunnel->parms.iph;
412
413 iph->version = 4;
414 iph->protocol = IPPROTO_IPIP;
415 iph->ihl = 5;
416 }
417
418 static struct xfrm4_protocol vti_esp4_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_ah4_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 struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
435 .handler = vti_rcv,
436 .input_handler = vti_input,
437 .cb_handler = vti_rcv_cb,
438 .err_handler = vti4_err,
439 .priority = 100,
440 };
441
442 static int __net_init vti_init_net(struct net *net)
443 {
444 int err;
445 struct ip_tunnel_net *itn;
446
447 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
448 if (err)
449 return err;
450 itn = net_generic(net, vti_net_id);
451 vti_fb_tunnel_init(itn->fb_tunnel_dev);
452 return 0;
453 }
454
455 static void __net_exit vti_exit_net(struct net *net)
456 {
457 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
458 ip_tunnel_delete_net(itn, &vti_link_ops);
459 }
460
461 static struct pernet_operations vti_net_ops = {
462 .init = vti_init_net,
463 .exit = vti_exit_net,
464 .id = &vti_net_id,
465 .size = sizeof(struct ip_tunnel_net),
466 };
467
468 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
469 struct netlink_ext_ack *extack)
470 {
471 return 0;
472 }
473
474 static void vti_netlink_parms(struct nlattr *data[],
475 struct ip_tunnel_parm *parms,
476 __u32 *fwmark)
477 {
478 memset(parms, 0, sizeof(*parms));
479
480 parms->iph.protocol = IPPROTO_IPIP;
481
482 if (!data)
483 return;
484
485 parms->i_flags = VTI_ISVTI;
486
487 if (data[IFLA_VTI_LINK])
488 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
489
490 if (data[IFLA_VTI_IKEY])
491 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
492
493 if (data[IFLA_VTI_OKEY])
494 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
495
496 if (data[IFLA_VTI_LOCAL])
497 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
498
499 if (data[IFLA_VTI_REMOTE])
500 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
501
502 if (data[IFLA_VTI_FWMARK])
503 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
504 }
505
506 static int vti_newlink(struct net *src_net, struct net_device *dev,
507 struct nlattr *tb[], struct nlattr *data[],
508 struct netlink_ext_ack *extack)
509 {
510 struct ip_tunnel_parm parms;
511 __u32 fwmark = 0;
512
513 vti_netlink_parms(data, &parms, &fwmark);
514 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
515 }
516
517 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
518 struct nlattr *data[],
519 struct netlink_ext_ack *extack)
520 {
521 struct ip_tunnel *t = netdev_priv(dev);
522 __u32 fwmark = t->fwmark;
523 struct ip_tunnel_parm p;
524
525 vti_netlink_parms(data, &p, &fwmark);
526 return ip_tunnel_changelink(dev, tb, &p, fwmark);
527 }
528
529 static size_t vti_get_size(const struct net_device *dev)
530 {
531 return
532 /* IFLA_VTI_LINK */
533 nla_total_size(4) +
534 /* IFLA_VTI_IKEY */
535 nla_total_size(4) +
536 /* IFLA_VTI_OKEY */
537 nla_total_size(4) +
538 /* IFLA_VTI_LOCAL */
539 nla_total_size(4) +
540 /* IFLA_VTI_REMOTE */
541 nla_total_size(4) +
542 /* IFLA_VTI_FWMARK */
543 nla_total_size(4) +
544 0;
545 }
546
547 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
548 {
549 struct ip_tunnel *t = netdev_priv(dev);
550 struct ip_tunnel_parm *p = &t->parms;
551
552 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
553 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
554 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
555 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
556 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
557 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
558 return -EMSGSIZE;
559
560 return 0;
561 }
562
563 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
564 [IFLA_VTI_LINK] = { .type = NLA_U32 },
565 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
566 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
567 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
568 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
569 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
570 };
571
572 static struct rtnl_link_ops vti_link_ops __read_mostly = {
573 .kind = "vti",
574 .maxtype = IFLA_VTI_MAX,
575 .policy = vti_policy,
576 .priv_size = sizeof(struct ip_tunnel),
577 .setup = vti_tunnel_setup,
578 .validate = vti_tunnel_validate,
579 .newlink = vti_newlink,
580 .changelink = vti_changelink,
581 .dellink = ip_tunnel_dellink,
582 .get_size = vti_get_size,
583 .fill_info = vti_fill_info,
584 .get_link_net = ip_tunnel_get_link_net,
585 };
586
587 static int __init vti_init(void)
588 {
589 const char *msg;
590 int err;
591
592 pr_info("IPv4 over IPsec tunneling driver\n");
593
594 msg = "tunnel device";
595 err = register_pernet_device(&vti_net_ops);
596 if (err < 0)
597 goto pernet_dev_failed;
598
599 msg = "tunnel protocols";
600 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
601 if (err < 0)
602 goto xfrm_proto_esp_failed;
603 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
604 if (err < 0)
605 goto xfrm_proto_ah_failed;
606 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
607 if (err < 0)
608 goto xfrm_proto_comp_failed;
609
610 msg = "netlink interface";
611 err = rtnl_link_register(&vti_link_ops);
612 if (err < 0)
613 goto rtnl_link_failed;
614
615 return err;
616
617 rtnl_link_failed:
618 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
619 xfrm_proto_comp_failed:
620 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
621 xfrm_proto_ah_failed:
622 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
623 xfrm_proto_esp_failed:
624 unregister_pernet_device(&vti_net_ops);
625 pernet_dev_failed:
626 pr_err("vti init: failed to register %s\n", msg);
627 return err;
628 }
629
630 static void __exit vti_fini(void)
631 {
632 rtnl_link_unregister(&vti_link_ops);
633 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
634 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
635 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
636 unregister_pernet_device(&vti_net_ops);
637 }
638
639 module_init(vti_init);
640 module_exit(vti_fini);
641 MODULE_LICENSE("GPL");
642 MODULE_ALIAS_RTNL_LINK("vti");
643 MODULE_ALIAS_NETDEV("ip_vti0");
(deprecated) Btrfs devel tree