?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Revert "ALSA: hda: Flush interrupts on disabling"
[linux/fpc-iii.git] / net / ipv6 / ip6_tunnel.c
blobcc28b8646986b186d8168b96ee96b15d831edac9
1 /*
2 * IPv6 tunneling device
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
8 *
9 * Based on:
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11 *
12 * RFC 2473
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
44
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
47
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60 #include <net/dst_metadata.h>
61
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
67
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
70
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 {
77 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78
79 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
80 }
81
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85
86 static int ip6_tnl_net_id __read_mostly;
87 struct ip6_tnl_net {
88 /* the IPv6 tunnel fallback device */
89 struct net_device *fb_tnl_dev;
90 /* lists for storing tunnels in use */
91 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 struct ip6_tnl __rcu *tnls_wc[1];
93 struct ip6_tnl __rcu **tnls[2];
94 struct ip6_tnl __rcu *collect_md_tun;
95 };
96
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
98 {
99 struct pcpu_sw_netstats tmp, sum = { 0 };
100 int i;
101
102 for_each_possible_cpu(i) {
103 unsigned int start;
104 const struct pcpu_sw_netstats *tstats =
105 per_cpu_ptr(dev->tstats, i);
106
107 do {
108 start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 tmp.rx_packets = tstats->rx_packets;
110 tmp.rx_bytes = tstats->rx_bytes;
111 tmp.tx_packets = tstats->tx_packets;
112 tmp.tx_bytes = tstats->tx_bytes;
113 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
114
115 sum.rx_packets += tmp.rx_packets;
116 sum.rx_bytes += tmp.rx_bytes;
117 sum.tx_packets += tmp.tx_packets;
118 sum.tx_bytes += tmp.tx_bytes;
119 }
120 dev->stats.rx_packets = sum.rx_packets;
121 dev->stats.rx_bytes = sum.rx_bytes;
122 dev->stats.tx_packets = sum.tx_packets;
123 dev->stats.tx_bytes = sum.tx_bytes;
124 return &dev->stats;
125 }
126
127 /**
128 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129 * @remote: the address of the tunnel exit-point
130 * @local: the address of the tunnel entry-point
131 *
132 * Return:
133 * tunnel matching given end-points if found,
134 * else fallback tunnel if its device is up,
135 * else %NULL
136 **/
137
138 #define for_each_ip6_tunnel_rcu(start) \
139 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
140
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
143 {
144 unsigned int hash = HASH(remote, local);
145 struct ip6_tnl *t;
146 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147 struct in6_addr any;
148
149 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 ipv6_addr_equal(remote, &t->parms.raddr) &&
152 (t->dev->flags & IFF_UP))
153 return t;
154 }
155
156 memset(&any, 0, sizeof(any));
157 hash = HASH(&any, local);
158 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 ipv6_addr_any(&t->parms.raddr) &&
161 (t->dev->flags & IFF_UP))
162 return t;
163 }
164
165 hash = HASH(remote, &any);
166 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168 ipv6_addr_any(&t->parms.laddr) &&
169 (t->dev->flags & IFF_UP))
170 return t;
171 }
172
173 t = rcu_dereference(ip6n->collect_md_tun);
174 if (t)
175 return t;
176
177 t = rcu_dereference(ip6n->tnls_wc[0]);
178 if (t && (t->dev->flags & IFF_UP))
179 return t;
180
181 return NULL;
182 }
183
184 /**
185 * ip6_tnl_bucket - get head of list matching given tunnel parameters
186 * @p: parameters containing tunnel end-points
187 *
188 * Description:
189 * ip6_tnl_bucket() returns the head of the list matching the
190 * &struct in6_addr entries laddr and raddr in @p.
191 *
192 * Return: head of IPv6 tunnel list
193 **/
194
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
197 {
198 const struct in6_addr *remote = &p->raddr;
199 const struct in6_addr *local = &p->laddr;
200 unsigned int h = 0;
201 int prio = 0;
202
203 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204 prio = 1;
205 h = HASH(remote, local);
206 }
207 return &ip6n->tnls[prio][h];
208 }
209
210 /**
211 * ip6_tnl_link - add tunnel to hash table
212 * @t: tunnel to be added
213 **/
214
215 static void
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
217 {
218 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
219
220 if (t->parms.collect_md)
221 rcu_assign_pointer(ip6n->collect_md_tun, t);
222 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 rcu_assign_pointer(*tp, t);
224 }
225
226 /**
227 * ip6_tnl_unlink - remove tunnel from hash table
228 * @t: tunnel to be removed
229 **/
230
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 struct ip6_tnl __rcu **tp;
235 struct ip6_tnl *iter;
236
237 if (t->parms.collect_md)
238 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
239
240 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241 (iter = rtnl_dereference(*tp)) != NULL;
242 tp = &iter->next) {
243 if (t == iter) {
244 rcu_assign_pointer(*tp, t->next);
245 break;
246 }
247 }
248 }
249
250 static void ip6_dev_free(struct net_device *dev)
251 {
252 struct ip6_tnl *t = netdev_priv(dev);
253
254 gro_cells_destroy(&t->gro_cells);
255 dst_cache_destroy(&t->dst_cache);
256 free_percpu(dev->tstats);
257 free_netdev(dev);
258 }
259
260 static int ip6_tnl_create2(struct net_device *dev)
261 {
262 struct ip6_tnl *t = netdev_priv(dev);
263 struct net *net = dev_net(dev);
264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
265 int err;
266
267 t = netdev_priv(dev);
268
269 dev->rtnl_link_ops = &ip6_link_ops;
270 err = register_netdevice(dev);
271 if (err < 0)
272 goto out;
273
274 strcpy(t->parms.name, dev->name);
275
276 dev_hold(dev);
277 ip6_tnl_link(ip6n, t);
278 return 0;
279
280 out:
281 return err;
282 }
283
284 /**
285 * ip6_tnl_create - create a new tunnel
286 * @p: tunnel parameters
287 * @pt: pointer to new tunnel
288 *
289 * Description:
290 * Create tunnel matching given parameters.
291 *
292 * Return:
293 * created tunnel or error pointer
294 **/
295
296 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
297 {
298 struct net_device *dev;
299 struct ip6_tnl *t;
300 char name[IFNAMSIZ];
301 int err = -E2BIG;
302
303 if (p->name[0]) {
304 if (!dev_valid_name(p->name))
305 goto failed;
306 strlcpy(name, p->name, IFNAMSIZ);
307 } else {
308 sprintf(name, "ip6tnl%%d");
309 }
310 err = -ENOMEM;
311 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
312 ip6_tnl_dev_setup);
313 if (!dev)
314 goto failed;
315
316 dev_net_set(dev, net);
317
318 t = netdev_priv(dev);
319 t->parms = *p;
320 t->net = dev_net(dev);
321 err = ip6_tnl_create2(dev);
322 if (err < 0)
323 goto failed_free;
324
325 return t;
326
327 failed_free:
328 ip6_dev_free(dev);
329 failed:
330 return ERR_PTR(err);
331 }
332
333 /**
334 * ip6_tnl_locate - find or create tunnel matching given parameters
335 * @p: tunnel parameters
336 * @create: != 0 if allowed to create new tunnel if no match found
337 *
338 * Description:
339 * ip6_tnl_locate() first tries to locate an existing tunnel
340 * based on @parms. If this is unsuccessful, but @create is set a new
341 * tunnel device is created and registered for use.
342 *
343 * Return:
344 * matching tunnel or error pointer
345 **/
346
347 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
348 struct __ip6_tnl_parm *p, int create)
349 {
350 const struct in6_addr *remote = &p->raddr;
351 const struct in6_addr *local = &p->laddr;
352 struct ip6_tnl __rcu **tp;
353 struct ip6_tnl *t;
354 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
355
356 for (tp = ip6_tnl_bucket(ip6n, p);
357 (t = rtnl_dereference(*tp)) != NULL;
358 tp = &t->next) {
359 if (ipv6_addr_equal(local, &t->parms.laddr) &&
360 ipv6_addr_equal(remote, &t->parms.raddr)) {
361 if (create)
362 return ERR_PTR(-EEXIST);
363
364 return t;
365 }
366 }
367 if (!create)
368 return ERR_PTR(-ENODEV);
369 return ip6_tnl_create(net, p);
370 }
371
372 /**
373 * ip6_tnl_dev_uninit - tunnel device uninitializer
374 * @dev: the device to be destroyed
375 *
376 * Description:
377 * ip6_tnl_dev_uninit() removes tunnel from its list
378 **/
379
380 static void
381 ip6_tnl_dev_uninit(struct net_device *dev)
382 {
383 struct ip6_tnl *t = netdev_priv(dev);
384 struct net *net = t->net;
385 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
386
387 if (dev == ip6n->fb_tnl_dev)
388 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
389 else
390 ip6_tnl_unlink(ip6n, t);
391 dst_cache_reset(&t->dst_cache);
392 dev_put(dev);
393 }
394
395 /**
396 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
397 * @skb: received socket buffer
398 *
399 * Return:
400 * 0 if none was found,
401 * else index to encapsulation limit
402 **/
403
404 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
405 {
406 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
407 unsigned int nhoff = raw - skb->data;
408 unsigned int off = nhoff + sizeof(*ipv6h);
409 u8 next, nexthdr = ipv6h->nexthdr;
410
411 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
412 struct ipv6_opt_hdr *hdr;
413 u16 optlen;
414
415 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
416 break;
417
418 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
419 if (nexthdr == NEXTHDR_FRAGMENT) {
420 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
421 if (frag_hdr->frag_off)
422 break;
423 optlen = 8;
424 } else if (nexthdr == NEXTHDR_AUTH) {
425 optlen = (hdr->hdrlen + 2) << 2;
426 } else {
427 optlen = ipv6_optlen(hdr);
428 }
429 /* cache hdr->nexthdr, since pskb_may_pull() might
430 * invalidate hdr
431 */
432 next = hdr->nexthdr;
433 if (nexthdr == NEXTHDR_DEST) {
434 u16 i = 2;
435
436 /* Remember : hdr is no longer valid at this point. */
437 if (!pskb_may_pull(skb, off + optlen))
438 break;
439
440 while (1) {
441 struct ipv6_tlv_tnl_enc_lim *tel;
442
443 /* No more room for encapsulation limit */
444 if (i + sizeof(*tel) > optlen)
445 break;
446
447 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
448 /* return index of option if found and valid */
449 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
450 tel->length == 1)
451 return i + off - nhoff;
452 /* else jump to next option */
453 if (tel->type)
454 i += tel->length + 2;
455 else
456 i++;
457 }
458 }
459 nexthdr = next;
460 off += optlen;
461 }
462 return 0;
463 }
464 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
465
466 /**
467 * ip6_tnl_err - tunnel error handler
468 *
469 * Description:
470 * ip6_tnl_err() should handle errors in the tunnel according
471 * to the specifications in RFC 2473.
472 **/
473
474 static int
475 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
476 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
477 {
478 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
479 struct ip6_tnl *t;
480 int rel_msg = 0;
481 u8 rel_type = ICMPV6_DEST_UNREACH;
482 u8 rel_code = ICMPV6_ADDR_UNREACH;
483 u8 tproto;
484 __u32 rel_info = 0;
485 __u16 len;
486 int err = -ENOENT;
487
488 /* If the packet doesn't contain the original IPv6 header we are
489 in trouble since we might need the source address for further
490 processing of the error. */
491
492 rcu_read_lock();
493 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
494 if (!t)
495 goto out;
496
497 tproto = ACCESS_ONCE(t->parms.proto);
498 if (tproto != ipproto && tproto != 0)
499 goto out;
500
501 err = 0;
502
503 switch (*type) {
504 __u32 teli;
505 struct ipv6_tlv_tnl_enc_lim *tel;
506 __u32 mtu;
507 case ICMPV6_DEST_UNREACH:
508 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
509 t->parms.name);
510 rel_msg = 1;
511 break;
512 case ICMPV6_TIME_EXCEED:
513 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
514 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
515 t->parms.name);
516 rel_msg = 1;
517 }
518 break;
519 case ICMPV6_PARAMPROB:
520 teli = 0;
521 if ((*code) == ICMPV6_HDR_FIELD)
522 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
523
524 if (teli && teli == *info - 2) {
525 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
526 if (tel->encap_limit == 0) {
527 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
528 t->parms.name);
529 rel_msg = 1;
530 }
531 } else {
532 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
533 t->parms.name);
534 }
535 break;
536 case ICMPV6_PKT_TOOBIG:
537 mtu = *info - offset;
538 if (mtu < IPV6_MIN_MTU)
539 mtu = IPV6_MIN_MTU;
540 t->dev->mtu = mtu;
541
542 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
543 if (len > mtu) {
544 rel_type = ICMPV6_PKT_TOOBIG;
545 rel_code = 0;
546 rel_info = mtu;
547 rel_msg = 1;
548 }
549 break;
550 }
551
552 *type = rel_type;
553 *code = rel_code;
554 *info = rel_info;
555 *msg = rel_msg;
556
557 out:
558 rcu_read_unlock();
559 return err;
560 }
561
562 static int
563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
564 u8 type, u8 code, int offset, __be32 info)
565 {
566 int rel_msg = 0;
567 u8 rel_type = type;
568 u8 rel_code = code;
569 __u32 rel_info = ntohl(info);
570 int err;
571 struct sk_buff *skb2;
572 const struct iphdr *eiph;
573 struct rtable *rt;
574 struct flowi4 fl4;
575
576 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
577 &rel_msg, &rel_info, offset);
578 if (err < 0)
579 return err;
580
581 if (rel_msg == 0)
582 return 0;
583
584 switch (rel_type) {
585 case ICMPV6_DEST_UNREACH:
586 if (rel_code != ICMPV6_ADDR_UNREACH)
587 return 0;
588 rel_type = ICMP_DEST_UNREACH;
589 rel_code = ICMP_HOST_UNREACH;
590 break;
591 case ICMPV6_PKT_TOOBIG:
592 if (rel_code != 0)
593 return 0;
594 rel_type = ICMP_DEST_UNREACH;
595 rel_code = ICMP_FRAG_NEEDED;
596 break;
597 case NDISC_REDIRECT:
598 rel_type = ICMP_REDIRECT;
599 rel_code = ICMP_REDIR_HOST;
600 default:
601 return 0;
602 }
603
604 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
605 return 0;
606
607 skb2 = skb_clone(skb, GFP_ATOMIC);
608 if (!skb2)
609 return 0;
610
611 skb_dst_drop(skb2);
612
613 skb_pull(skb2, offset);
614 skb_reset_network_header(skb2);
615 eiph = ip_hdr(skb2);
616
617 /* Try to guess incoming interface */
618 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
619 eiph->saddr, 0,
620 0, 0,
621 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
622 if (IS_ERR(rt))
623 goto out;
624
625 skb2->dev = rt->dst.dev;
626
627 /* route "incoming" packet */
628 if (rt->rt_flags & RTCF_LOCAL) {
629 ip_rt_put(rt);
630 rt = NULL;
631 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
632 eiph->daddr, eiph->saddr,
633 0, 0,
634 IPPROTO_IPIP,
635 RT_TOS(eiph->tos), 0);
636 if (IS_ERR(rt) ||
637 rt->dst.dev->type != ARPHRD_TUNNEL6) {
638 if (!IS_ERR(rt))
639 ip_rt_put(rt);
640 goto out;
641 }
642 skb_dst_set(skb2, &rt->dst);
643 } else {
644 ip_rt_put(rt);
645 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
646 skb2->dev) ||
647 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
648 goto out;
649 }
650
651 /* change mtu on this route */
652 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
653 if (rel_info > dst_mtu(skb_dst(skb2)))
654 goto out;
655
656 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
657 }
658 if (rel_type == ICMP_REDIRECT)
659 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
660
661 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
662
663 out:
664 kfree_skb(skb2);
665 return 0;
666 }
667
668 static int
669 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
670 u8 type, u8 code, int offset, __be32 info)
671 {
672 int rel_msg = 0;
673 u8 rel_type = type;
674 u8 rel_code = code;
675 __u32 rel_info = ntohl(info);
676 int err;
677
678 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
679 &rel_msg, &rel_info, offset);
680 if (err < 0)
681 return err;
682
683 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
684 struct rt6_info *rt;
685 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
686
687 if (!skb2)
688 return 0;
689
690 skb_dst_drop(skb2);
691 skb_pull(skb2, offset);
692 skb_reset_network_header(skb2);
693
694 /* Try to guess incoming interface */
695 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
696 NULL, 0, 0);
697
698 if (rt && rt->dst.dev)
699 skb2->dev = rt->dst.dev;
700
701 icmpv6_send(skb2, rel_type, rel_code, rel_info);
702
703 ip6_rt_put(rt);
704
705 kfree_skb(skb2);
706 }
707
708 return 0;
709 }
710
711 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
712 const struct ipv6hdr *ipv6h,
713 struct sk_buff *skb)
714 {
715 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
716
717 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
718 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
719
720 return IP6_ECN_decapsulate(ipv6h, skb);
721 }
722
723 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
724 const struct ipv6hdr *ipv6h,
725 struct sk_buff *skb)
726 {
727 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
728 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
729
730 return IP6_ECN_decapsulate(ipv6h, skb);
731 }
732
733 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
734 const struct in6_addr *laddr,
735 const struct in6_addr *raddr)
736 {
737 struct __ip6_tnl_parm *p = &t->parms;
738 int ltype = ipv6_addr_type(laddr);
739 int rtype = ipv6_addr_type(raddr);
740 __u32 flags = 0;
741
742 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
743 flags = IP6_TNL_F_CAP_PER_PACKET;
744 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
745 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
746 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
747 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
748 if (ltype&IPV6_ADDR_UNICAST)
749 flags |= IP6_TNL_F_CAP_XMIT;
750 if (rtype&IPV6_ADDR_UNICAST)
751 flags |= IP6_TNL_F_CAP_RCV;
752 }
753 return flags;
754 }
755 EXPORT_SYMBOL(ip6_tnl_get_cap);
756
757 /* called with rcu_read_lock() */
758 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
759 const struct in6_addr *laddr,
760 const struct in6_addr *raddr)
761 {
762 struct __ip6_tnl_parm *p = &t->parms;
763 int ret = 0;
764 struct net *net = t->net;
765
766 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
767 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
768 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
769 struct net_device *ldev = NULL;
770
771 if (p->link)
772 ldev = dev_get_by_index_rcu(net, p->link);
773
774 if ((ipv6_addr_is_multicast(laddr) ||
775 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
776 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
777 ret = 1;
778 }
779 return ret;
780 }
781 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
782
783 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
784 const struct tnl_ptk_info *tpi,
785 struct metadata_dst *tun_dst,
786 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
787 const struct ipv6hdr *ipv6h,
788 struct sk_buff *skb),
789 bool log_ecn_err)
790 {
791 struct pcpu_sw_netstats *tstats;
792 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
793 int err;
794
795 if ((!(tpi->flags & TUNNEL_CSUM) &&
796 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
797 ((tpi->flags & TUNNEL_CSUM) &&
798 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
799 tunnel->dev->stats.rx_crc_errors++;
800 tunnel->dev->stats.rx_errors++;
801 goto drop;
802 }
803
804 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
805 if (!(tpi->flags & TUNNEL_SEQ) ||
806 (tunnel->i_seqno &&
807 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
808 tunnel->dev->stats.rx_fifo_errors++;
809 tunnel->dev->stats.rx_errors++;
810 goto drop;
811 }
812 tunnel->i_seqno = ntohl(tpi->seq) + 1;
813 }
814
815 skb->protocol = tpi->proto;
816
817 /* Warning: All skb pointers will be invalidated! */
818 if (tunnel->dev->type == ARPHRD_ETHER) {
819 if (!pskb_may_pull(skb, ETH_HLEN)) {
820 tunnel->dev->stats.rx_length_errors++;
821 tunnel->dev->stats.rx_errors++;
822 goto drop;
823 }
824
825 ipv6h = ipv6_hdr(skb);
826 skb->protocol = eth_type_trans(skb, tunnel->dev);
827 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
828 } else {
829 skb->dev = tunnel->dev;
830 }
831
832 skb_reset_network_header(skb);
833 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
834
835 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
836
837 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
838 if (unlikely(err)) {
839 if (log_ecn_err)
840 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
841 &ipv6h->saddr,
842 ipv6_get_dsfield(ipv6h));
843 if (err > 1) {
844 ++tunnel->dev->stats.rx_frame_errors;
845 ++tunnel->dev->stats.rx_errors;
846 goto drop;
847 }
848 }
849
850 tstats = this_cpu_ptr(tunnel->dev->tstats);
851 u64_stats_update_begin(&tstats->syncp);
852 tstats->rx_packets++;
853 tstats->rx_bytes += skb->len;
854 u64_stats_update_end(&tstats->syncp);
855
856 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
857
858 if (tun_dst)
859 skb_dst_set(skb, (struct dst_entry *)tun_dst);
860
861 gro_cells_receive(&tunnel->gro_cells, skb);
862 return 0;
863
864 drop:
865 kfree_skb(skb);
866 return 0;
867 }
868
869 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
870 const struct tnl_ptk_info *tpi,
871 struct metadata_dst *tun_dst,
872 bool log_ecn_err)
873 {
874 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
875 log_ecn_err);
876 }
877 EXPORT_SYMBOL(ip6_tnl_rcv);
878
879 static const struct tnl_ptk_info tpi_v6 = {
880 /* no tunnel info required for ipxip6. */
881 .proto = htons(ETH_P_IPV6),
882 };
883
884 static const struct tnl_ptk_info tpi_v4 = {
885 /* no tunnel info required for ipxip6. */
886 .proto = htons(ETH_P_IP),
887 };
888
889 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
890 const struct tnl_ptk_info *tpi,
891 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
892 const struct ipv6hdr *ipv6h,
893 struct sk_buff *skb))
894 {
895 struct ip6_tnl *t;
896 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
897 struct metadata_dst *tun_dst = NULL;
898 int ret = -1;
899
900 rcu_read_lock();
901 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
902
903 if (t) {
904 u8 tproto = ACCESS_ONCE(t->parms.proto);
905
906 if (tproto != ipproto && tproto != 0)
907 goto drop;
908 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
909 goto drop;
910 ipv6h = ipv6_hdr(skb);
911 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
912 goto drop;
913 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
914 goto drop;
915 if (t->parms.collect_md) {
916 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
917 if (!tun_dst)
918 goto drop;
919 }
920 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
921 log_ecn_error);
922 }
923
924 rcu_read_unlock();
925
926 return ret;
927
928 drop:
929 rcu_read_unlock();
930 kfree_skb(skb);
931 return 0;
932 }
933
934 static int ip4ip6_rcv(struct sk_buff *skb)
935 {
936 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
937 ip4ip6_dscp_ecn_decapsulate);
938 }
939
940 static int ip6ip6_rcv(struct sk_buff *skb)
941 {
942 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
943 ip6ip6_dscp_ecn_decapsulate);
944 }
945
946 struct ipv6_tel_txoption {
947 struct ipv6_txoptions ops;
948 __u8 dst_opt[8];
949 };
950
951 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
952 {
953 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
954
955 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
956 opt->dst_opt[3] = 1;
957 opt->dst_opt[4] = encap_limit;
958 opt->dst_opt[5] = IPV6_TLV_PADN;
959 opt->dst_opt[6] = 1;
960
961 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
962 opt->ops.opt_nflen = 8;
963 }
964
965 /**
966 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
967 * @t: the outgoing tunnel device
968 * @hdr: IPv6 header from the incoming packet
969 *
970 * Description:
971 * Avoid trivial tunneling loop by checking that tunnel exit-point
972 * doesn't match source of incoming packet.
973 *
974 * Return:
975 * 1 if conflict,
976 * 0 else
977 **/
978
979 static inline bool
980 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
981 {
982 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
983 }
984
985 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
986 const struct in6_addr *laddr,
987 const struct in6_addr *raddr)
988 {
989 struct __ip6_tnl_parm *p = &t->parms;
990 int ret = 0;
991 struct net *net = t->net;
992
993 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
994 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
995 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
996 struct net_device *ldev = NULL;
997
998 rcu_read_lock();
999 if (p->link)
1000 ldev = dev_get_by_index_rcu(net, p->link);
1001
1002 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1003 pr_warn("%s xmit: Local address not yet configured!\n",
1004 p->name);
1005 else if (!ipv6_addr_is_multicast(raddr) &&
1006 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1007 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1008 p->name);
1009 else
1010 ret = 1;
1011 rcu_read_unlock();
1012 }
1013 return ret;
1014 }
1015 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1016
1017 /**
1018 * ip6_tnl_xmit - encapsulate packet and send
1019 * @skb: the outgoing socket buffer
1020 * @dev: the outgoing tunnel device
1021 * @dsfield: dscp code for outer header
1022 * @fl6: flow of tunneled packet
1023 * @encap_limit: encapsulation limit
1024 * @pmtu: Path MTU is stored if packet is too big
1025 * @proto: next header value
1026 *
1027 * Description:
1028 * Build new header and do some sanity checks on the packet before sending
1029 * it.
1030 *
1031 * Return:
1032 * 0 on success
1033 * -1 fail
1034 * %-EMSGSIZE message too big. return mtu in this case.
1035 **/
1036
1037 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1038 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1039 __u8 proto)
1040 {
1041 struct ip6_tnl *t = netdev_priv(dev);
1042 struct net *net = t->net;
1043 struct net_device_stats *stats = &t->dev->stats;
1044 struct ipv6hdr *ipv6h;
1045 struct ipv6_tel_txoption opt;
1046 struct dst_entry *dst = NULL, *ndst = NULL;
1047 struct net_device *tdev;
1048 int mtu;
1049 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1050 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1051 unsigned int max_headroom = psh_hlen;
1052 bool use_cache = false;
1053 u8 hop_limit;
1054 int err = -1;
1055
1056 if (t->parms.collect_md) {
1057 hop_limit = skb_tunnel_info(skb)->key.ttl;
1058 goto route_lookup;
1059 } else {
1060 hop_limit = t->parms.hop_limit;
1061 }
1062
1063 /* NBMA tunnel */
1064 if (ipv6_addr_any(&t->parms.raddr)) {
1065 if (skb->protocol == htons(ETH_P_IPV6)) {
1066 struct in6_addr *addr6;
1067 struct neighbour *neigh;
1068 int addr_type;
1069
1070 if (!skb_dst(skb))
1071 goto tx_err_link_failure;
1072
1073 neigh = dst_neigh_lookup(skb_dst(skb),
1074 &ipv6_hdr(skb)->daddr);
1075 if (!neigh)
1076 goto tx_err_link_failure;
1077
1078 addr6 = (struct in6_addr *)&neigh->primary_key;
1079 addr_type = ipv6_addr_type(addr6);
1080
1081 if (addr_type == IPV6_ADDR_ANY)
1082 addr6 = &ipv6_hdr(skb)->daddr;
1083
1084 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1085 neigh_release(neigh);
1086 }
1087 } else if (t->parms.proto != 0 && !(t->parms.flags &
1088 (IP6_TNL_F_USE_ORIG_TCLASS |
1089 IP6_TNL_F_USE_ORIG_FWMARK))) {
1090 /* enable the cache only if neither the outer protocol nor the
1091 * routing decision depends on the current inner header value
1092 */
1093 use_cache = true;
1094 }
1095
1096 if (use_cache)
1097 dst = dst_cache_get(&t->dst_cache);
1098
1099 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1100 goto tx_err_link_failure;
1101
1102 if (!dst) {
1103 route_lookup:
1104 /* add dsfield to flowlabel for route lookup */
1105 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1106
1107 dst = ip6_route_output(net, NULL, fl6);
1108
1109 if (dst->error)
1110 goto tx_err_link_failure;
1111 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1112 if (IS_ERR(dst)) {
1113 err = PTR_ERR(dst);
1114 dst = NULL;
1115 goto tx_err_link_failure;
1116 }
1117 if (t->parms.collect_md &&
1118 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1119 &fl6->daddr, 0, &fl6->saddr))
1120 goto tx_err_link_failure;
1121 ndst = dst;
1122 }
1123
1124 tdev = dst->dev;
1125
1126 if (tdev == dev) {
1127 stats->collisions++;
1128 net_warn_ratelimited("%s: Local routing loop detected!\n",
1129 t->parms.name);
1130 goto tx_err_dst_release;
1131 }
1132 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1133 if (encap_limit >= 0) {
1134 max_headroom += 8;
1135 mtu -= 8;
1136 }
1137 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1138 IPV6_MIN_MTU : IPV4_MIN_MTU);
1139
1140 if (skb_dst(skb) && !t->parms.collect_md)
1141 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1142 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1143 *pmtu = mtu;
1144 err = -EMSGSIZE;
1145 goto tx_err_dst_release;
1146 }
1147
1148 if (t->err_count > 0) {
1149 if (time_before(jiffies,
1150 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1151 t->err_count--;
1152
1153 dst_link_failure(skb);
1154 } else {
1155 t->err_count = 0;
1156 }
1157 }
1158
1159 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1160
1161 /*
1162 * Okay, now see if we can stuff it in the buffer as-is.
1163 */
1164 max_headroom += LL_RESERVED_SPACE(tdev);
1165
1166 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1167 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1168 struct sk_buff *new_skb;
1169
1170 new_skb = skb_realloc_headroom(skb, max_headroom);
1171 if (!new_skb)
1172 goto tx_err_dst_release;
1173
1174 if (skb->sk)
1175 skb_set_owner_w(new_skb, skb->sk);
1176 consume_skb(skb);
1177 skb = new_skb;
1178 }
1179
1180 if (t->parms.collect_md) {
1181 if (t->encap.type != TUNNEL_ENCAP_NONE)
1182 goto tx_err_dst_release;
1183 } else {
1184 if (use_cache && ndst)
1185 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1186 }
1187 skb_dst_set(skb, dst);
1188
1189 /* Calculate max headroom for all the headers and adjust
1190 * needed_headroom if necessary.
1191 */
1192 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1193 + dst->header_len + t->hlen;
1194 if (max_headroom > dev->needed_headroom)
1195 dev->needed_headroom = max_headroom;
1196
1197 err = ip6_tnl_encap(skb, t, &proto, fl6);
1198 if (err)
1199 return err;
1200
1201 if (encap_limit >= 0) {
1202 init_tel_txopt(&opt, encap_limit);
1203 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1204 }
1205
1206 skb_push(skb, sizeof(struct ipv6hdr));
1207 skb_reset_network_header(skb);
1208 ipv6h = ipv6_hdr(skb);
1209 ip6_flow_hdr(ipv6h, dsfield,
1210 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1211 ipv6h->hop_limit = hop_limit;
1212 ipv6h->nexthdr = proto;
1213 ipv6h->saddr = fl6->saddr;
1214 ipv6h->daddr = fl6->daddr;
1215 ip6tunnel_xmit(NULL, skb, dev);
1216 return 0;
1217 tx_err_link_failure:
1218 stats->tx_carrier_errors++;
1219 dst_link_failure(skb);
1220 tx_err_dst_release:
1221 dst_release(dst);
1222 return err;
1223 }
1224 EXPORT_SYMBOL(ip6_tnl_xmit);
1225
1226 static inline int
1227 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1228 {
1229 struct ip6_tnl *t = netdev_priv(dev);
1230 const struct iphdr *iph;
1231 int encap_limit = -1;
1232 struct flowi6 fl6;
1233 __u8 dsfield;
1234 __u32 mtu;
1235 u8 tproto;
1236 int err;
1237
1238 /* ensure we can access the full inner ip header */
1239 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1240 return -1;
1241
1242 iph = ip_hdr(skb);
1243 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1244
1245 tproto = ACCESS_ONCE(t->parms.proto);
1246 if (tproto != IPPROTO_IPIP && tproto != 0)
1247 return -1;
1248
1249 if (t->parms.collect_md) {
1250 struct ip_tunnel_info *tun_info;
1251 const struct ip_tunnel_key *key;
1252
1253 tun_info = skb_tunnel_info(skb);
1254 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1255 ip_tunnel_info_af(tun_info) != AF_INET6))
1256 return -1;
1257 key = &tun_info->key;
1258 memset(&fl6, 0, sizeof(fl6));
1259 fl6.flowi6_proto = IPPROTO_IPIP;
1260 fl6.daddr = key->u.ipv6.dst;
1261 fl6.flowlabel = key->label;
1262 dsfield = key->tos;
1263 } else {
1264 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1265 encap_limit = t->parms.encap_limit;
1266
1267 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1268 fl6.flowi6_proto = IPPROTO_IPIP;
1269
1270 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1271 dsfield = ipv4_get_dsfield(iph);
1272 else
1273 dsfield = ip6_tclass(t->parms.flowinfo);
1274 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1275 fl6.flowi6_mark = skb->mark;
1276 }
1277
1278 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1279
1280 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1281 return -1;
1282
1283 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1284
1285 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1286 IPPROTO_IPIP);
1287 if (err != 0) {
1288 /* XXX: send ICMP error even if DF is not set. */
1289 if (err == -EMSGSIZE)
1290 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1291 htonl(mtu));
1292 return -1;
1293 }
1294
1295 return 0;
1296 }
1297
1298 static inline int
1299 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1300 {
1301 struct ip6_tnl *t = netdev_priv(dev);
1302 struct ipv6hdr *ipv6h;
1303 int encap_limit = -1;
1304 __u16 offset;
1305 struct flowi6 fl6;
1306 __u8 dsfield;
1307 __u32 mtu;
1308 u8 tproto;
1309 int err;
1310
1311 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1312 return -1;
1313
1314 ipv6h = ipv6_hdr(skb);
1315 tproto = ACCESS_ONCE(t->parms.proto);
1316 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1317 ip6_tnl_addr_conflict(t, ipv6h))
1318 return -1;
1319
1320 if (t->parms.collect_md) {
1321 struct ip_tunnel_info *tun_info;
1322 const struct ip_tunnel_key *key;
1323
1324 tun_info = skb_tunnel_info(skb);
1325 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1326 ip_tunnel_info_af(tun_info) != AF_INET6))
1327 return -1;
1328 key = &tun_info->key;
1329 memset(&fl6, 0, sizeof(fl6));
1330 fl6.flowi6_proto = IPPROTO_IPV6;
1331 fl6.daddr = key->u.ipv6.dst;
1332 fl6.flowlabel = key->label;
1333 dsfield = key->tos;
1334 } else {
1335 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1336 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1337 ipv6h = ipv6_hdr(skb);
1338 if (offset > 0) {
1339 struct ipv6_tlv_tnl_enc_lim *tel;
1340
1341 tel = (void *)&skb_network_header(skb)[offset];
1342 if (tel->encap_limit == 0) {
1343 icmpv6_send(skb, ICMPV6_PARAMPROB,
1344 ICMPV6_HDR_FIELD, offset + 2);
1345 return -1;
1346 }
1347 encap_limit = tel->encap_limit - 1;
1348 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1349 encap_limit = t->parms.encap_limit;
1350 }
1351
1352 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1353 fl6.flowi6_proto = IPPROTO_IPV6;
1354
1355 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1356 dsfield = ipv6_get_dsfield(ipv6h);
1357 else
1358 dsfield = ip6_tclass(t->parms.flowinfo);
1359 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1360 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1361 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1362 fl6.flowi6_mark = skb->mark;
1363 }
1364
1365 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1366
1367 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1368 return -1;
1369
1370 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1371
1372 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1373 IPPROTO_IPV6);
1374 if (err != 0) {
1375 if (err == -EMSGSIZE)
1376 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1377 return -1;
1378 }
1379
1380 return 0;
1381 }
1382
1383 static netdev_tx_t
1384 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1385 {
1386 struct ip6_tnl *t = netdev_priv(dev);
1387 struct net_device_stats *stats = &t->dev->stats;
1388 int ret;
1389
1390 switch (skb->protocol) {
1391 case htons(ETH_P_IP):
1392 ret = ip4ip6_tnl_xmit(skb, dev);
1393 break;
1394 case htons(ETH_P_IPV6):
1395 ret = ip6ip6_tnl_xmit(skb, dev);
1396 break;
1397 default:
1398 goto tx_err;
1399 }
1400
1401 if (ret < 0)
1402 goto tx_err;
1403
1404 return NETDEV_TX_OK;
1405
1406 tx_err:
1407 stats->tx_errors++;
1408 stats->tx_dropped++;
1409 kfree_skb(skb);
1410 return NETDEV_TX_OK;
1411 }
1412
1413 static void ip6_tnl_link_config(struct ip6_tnl *t)
1414 {
1415 struct net_device *dev = t->dev;
1416 struct __ip6_tnl_parm *p = &t->parms;
1417 struct flowi6 *fl6 = &t->fl.u.ip6;
1418 int t_hlen;
1419
1420 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1421 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1422
1423 /* Set up flowi template */
1424 fl6->saddr = p->laddr;
1425 fl6->daddr = p->raddr;
1426 fl6->flowi6_oif = p->link;
1427 fl6->flowlabel = 0;
1428
1429 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1430 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1431 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1432 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1433
1434 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1435 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1436
1437 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1438 dev->flags |= IFF_POINTOPOINT;
1439 else
1440 dev->flags &= ~IFF_POINTOPOINT;
1441
1442 t->tun_hlen = 0;
1443 t->hlen = t->encap_hlen + t->tun_hlen;
1444 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1445
1446 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1447 int strict = (ipv6_addr_type(&p->raddr) &
1448 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1449
1450 struct rt6_info *rt = rt6_lookup(t->net,
1451 &p->raddr, &p->laddr,
1452 p->link, strict);
1453
1454 if (!rt)
1455 return;
1456
1457 if (rt->dst.dev) {
1458 dev->hard_header_len = rt->dst.dev->hard_header_len +
1459 t_hlen;
1460
1461 dev->mtu = rt->dst.dev->mtu - t_hlen;
1462 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1463 dev->mtu -= 8;
1464
1465 if (dev->mtu < IPV6_MIN_MTU)
1466 dev->mtu = IPV6_MIN_MTU;
1467 }
1468 ip6_rt_put(rt);
1469 }
1470 }
1471
1472 /**
1473 * ip6_tnl_change - update the tunnel parameters
1474 * @t: tunnel to be changed
1475 * @p: tunnel configuration parameters
1476 *
1477 * Description:
1478 * ip6_tnl_change() updates the tunnel parameters
1479 **/
1480
1481 static int
1482 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1483 {
1484 t->parms.laddr = p->laddr;
1485 t->parms.raddr = p->raddr;
1486 t->parms.flags = p->flags;
1487 t->parms.hop_limit = p->hop_limit;
1488 t->parms.encap_limit = p->encap_limit;
1489 t->parms.flowinfo = p->flowinfo;
1490 t->parms.link = p->link;
1491 t->parms.proto = p->proto;
1492 dst_cache_reset(&t->dst_cache);
1493 ip6_tnl_link_config(t);
1494 return 0;
1495 }
1496
1497 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1498 {
1499 struct net *net = t->net;
1500 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1501 int err;
1502
1503 ip6_tnl_unlink(ip6n, t);
1504 synchronize_net();
1505 err = ip6_tnl_change(t, p);
1506 ip6_tnl_link(ip6n, t);
1507 netdev_state_change(t->dev);
1508 return err;
1509 }
1510
1511 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1512 {
1513 /* for default tnl0 device allow to change only the proto */
1514 t->parms.proto = p->proto;
1515 netdev_state_change(t->dev);
1516 return 0;
1517 }
1518
1519 static void
1520 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1521 {
1522 p->laddr = u->laddr;
1523 p->raddr = u->raddr;
1524 p->flags = u->flags;
1525 p->hop_limit = u->hop_limit;
1526 p->encap_limit = u->encap_limit;
1527 p->flowinfo = u->flowinfo;
1528 p->link = u->link;
1529 p->proto = u->proto;
1530 memcpy(p->name, u->name, sizeof(u->name));
1531 }
1532
1533 static void
1534 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1535 {
1536 u->laddr = p->laddr;
1537 u->raddr = p->raddr;
1538 u->flags = p->flags;
1539 u->hop_limit = p->hop_limit;
1540 u->encap_limit = p->encap_limit;
1541 u->flowinfo = p->flowinfo;
1542 u->link = p->link;
1543 u->proto = p->proto;
1544 memcpy(u->name, p->name, sizeof(u->name));
1545 }
1546
1547 /**
1548 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1549 * @dev: virtual device associated with tunnel
1550 * @ifr: parameters passed from userspace
1551 * @cmd: command to be performed
1552 *
1553 * Description:
1554 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1555 * from userspace.
1556 *
1557 * The possible commands are the following:
1558 * %SIOCGETTUNNEL: get tunnel parameters for device
1559 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1560 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1561 * %SIOCDELTUNNEL: delete tunnel
1562 *
1563 * The fallback device "ip6tnl0", created during module
1564 * initialization, can be used for creating other tunnel devices.
1565 *
1566 * Return:
1567 * 0 on success,
1568 * %-EFAULT if unable to copy data to or from userspace,
1569 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1570 * %-EINVAL if passed tunnel parameters are invalid,
1571 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1572 * %-ENODEV if attempting to change or delete a nonexisting device
1573 **/
1574
1575 static int
1576 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1577 {
1578 int err = 0;
1579 struct ip6_tnl_parm p;
1580 struct __ip6_tnl_parm p1;
1581 struct ip6_tnl *t = netdev_priv(dev);
1582 struct net *net = t->net;
1583 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1584
1585 memset(&p1, 0, sizeof(p1));
1586
1587 switch (cmd) {
1588 case SIOCGETTUNNEL:
1589 if (dev == ip6n->fb_tnl_dev) {
1590 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1591 err = -EFAULT;
1592 break;
1593 }
1594 ip6_tnl_parm_from_user(&p1, &p);
1595 t = ip6_tnl_locate(net, &p1, 0);
1596 if (IS_ERR(t))
1597 t = netdev_priv(dev);
1598 } else {
1599 memset(&p, 0, sizeof(p));
1600 }
1601 ip6_tnl_parm_to_user(&p, &t->parms);
1602 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1603 err = -EFAULT;
1604 }
1605 break;
1606 case SIOCADDTUNNEL:
1607 case SIOCCHGTUNNEL:
1608 err = -EPERM;
1609 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1610 break;
1611 err = -EFAULT;
1612 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1613 break;
1614 err = -EINVAL;
1615 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1616 p.proto != 0)
1617 break;
1618 ip6_tnl_parm_from_user(&p1, &p);
1619 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1620 if (cmd == SIOCCHGTUNNEL) {
1621 if (!IS_ERR(t)) {
1622 if (t->dev != dev) {
1623 err = -EEXIST;
1624 break;
1625 }
1626 } else
1627 t = netdev_priv(dev);
1628 if (dev == ip6n->fb_tnl_dev)
1629 err = ip6_tnl0_update(t, &p1);
1630 else
1631 err = ip6_tnl_update(t, &p1);
1632 }
1633 if (!IS_ERR(t)) {
1634 err = 0;
1635 ip6_tnl_parm_to_user(&p, &t->parms);
1636 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1637 err = -EFAULT;
1638
1639 } else {
1640 err = PTR_ERR(t);
1641 }
1642 break;
1643 case SIOCDELTUNNEL:
1644 err = -EPERM;
1645 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1646 break;
1647
1648 if (dev == ip6n->fb_tnl_dev) {
1649 err = -EFAULT;
1650 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1651 break;
1652 err = -ENOENT;
1653 ip6_tnl_parm_from_user(&p1, &p);
1654 t = ip6_tnl_locate(net, &p1, 0);
1655 if (IS_ERR(t))
1656 break;
1657 err = -EPERM;
1658 if (t->dev == ip6n->fb_tnl_dev)
1659 break;
1660 dev = t->dev;
1661 }
1662 err = 0;
1663 unregister_netdevice(dev);
1664 break;
1665 default:
1666 err = -EINVAL;
1667 }
1668 return err;
1669 }
1670
1671 /**
1672 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1673 * @dev: virtual device associated with tunnel
1674 * @new_mtu: the new mtu
1675 *
1676 * Return:
1677 * 0 on success,
1678 * %-EINVAL if mtu too small
1679 **/
1680
1681 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1682 {
1683 struct ip6_tnl *tnl = netdev_priv(dev);
1684
1685 if (tnl->parms.proto == IPPROTO_IPIP) {
1686 if (new_mtu < 68)
1687 return -EINVAL;
1688 } else {
1689 if (new_mtu < IPV6_MIN_MTU)
1690 return -EINVAL;
1691 }
1692 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1693 return -EINVAL;
1694 dev->mtu = new_mtu;
1695 return 0;
1696 }
1697 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1698
1699 int ip6_tnl_get_iflink(const struct net_device *dev)
1700 {
1701 struct ip6_tnl *t = netdev_priv(dev);
1702
1703 return t->parms.link;
1704 }
1705 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1706
1707 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1708 unsigned int num)
1709 {
1710 if (num >= MAX_IPTUN_ENCAP_OPS)
1711 return -ERANGE;
1712
1713 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1714 &ip6tun_encaps[num],
1715 NULL, ops) ? 0 : -1;
1716 }
1717 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1718
1719 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1720 unsigned int num)
1721 {
1722 int ret;
1723
1724 if (num >= MAX_IPTUN_ENCAP_OPS)
1725 return -ERANGE;
1726
1727 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1728 &ip6tun_encaps[num],
1729 ops, NULL) == ops) ? 0 : -1;
1730
1731 synchronize_net();
1732
1733 return ret;
1734 }
1735 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1736
1737 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1738 struct ip_tunnel_encap *ipencap)
1739 {
1740 int hlen;
1741
1742 memset(&t->encap, 0, sizeof(t->encap));
1743
1744 hlen = ip6_encap_hlen(ipencap);
1745 if (hlen < 0)
1746 return hlen;
1747
1748 t->encap.type = ipencap->type;
1749 t->encap.sport = ipencap->sport;
1750 t->encap.dport = ipencap->dport;
1751 t->encap.flags = ipencap->flags;
1752
1753 t->encap_hlen = hlen;
1754 t->hlen = t->encap_hlen + t->tun_hlen;
1755
1756 return 0;
1757 }
1758 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1759
1760 static const struct net_device_ops ip6_tnl_netdev_ops = {
1761 .ndo_init = ip6_tnl_dev_init,
1762 .ndo_uninit = ip6_tnl_dev_uninit,
1763 .ndo_start_xmit = ip6_tnl_start_xmit,
1764 .ndo_do_ioctl = ip6_tnl_ioctl,
1765 .ndo_change_mtu = ip6_tnl_change_mtu,
1766 .ndo_get_stats = ip6_get_stats,
1767 .ndo_get_iflink = ip6_tnl_get_iflink,
1768 };
1769
1770 #define IPXIPX_FEATURES (NETIF_F_SG | \
1771 NETIF_F_FRAGLIST | \
1772 NETIF_F_HIGHDMA | \
1773 NETIF_F_GSO_SOFTWARE | \
1774 NETIF_F_HW_CSUM)
1775
1776 /**
1777 * ip6_tnl_dev_setup - setup virtual tunnel device
1778 * @dev: virtual device associated with tunnel
1779 *
1780 * Description:
1781 * Initialize function pointers and device parameters
1782 **/
1783
1784 static void ip6_tnl_dev_setup(struct net_device *dev)
1785 {
1786 dev->netdev_ops = &ip6_tnl_netdev_ops;
1787 dev->destructor = ip6_dev_free;
1788
1789 dev->type = ARPHRD_TUNNEL6;
1790 dev->flags |= IFF_NOARP;
1791 dev->addr_len = sizeof(struct in6_addr);
1792 dev->features |= NETIF_F_LLTX;
1793 netif_keep_dst(dev);
1794
1795 dev->features |= IPXIPX_FEATURES;
1796 dev->hw_features |= IPXIPX_FEATURES;
1797
1798 /* This perm addr will be used as interface identifier by IPv6 */
1799 dev->addr_assign_type = NET_ADDR_RANDOM;
1800 eth_random_addr(dev->perm_addr);
1801 }
1802
1803
1804 /**
1805 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1806 * @dev: virtual device associated with tunnel
1807 **/
1808
1809 static inline int
1810 ip6_tnl_dev_init_gen(struct net_device *dev)
1811 {
1812 struct ip6_tnl *t = netdev_priv(dev);
1813 int ret;
1814 int t_hlen;
1815
1816 t->dev = dev;
1817 t->net = dev_net(dev);
1818 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1819 if (!dev->tstats)
1820 return -ENOMEM;
1821
1822 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1823 if (ret)
1824 goto free_stats;
1825
1826 ret = gro_cells_init(&t->gro_cells, dev);
1827 if (ret)
1828 goto destroy_dst;
1829
1830 t->tun_hlen = 0;
1831 t->hlen = t->encap_hlen + t->tun_hlen;
1832 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1833
1834 dev->type = ARPHRD_TUNNEL6;
1835 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1836 dev->mtu = ETH_DATA_LEN - t_hlen;
1837 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1838 dev->mtu -= 8;
1839
1840 return 0;
1841
1842 destroy_dst:
1843 dst_cache_destroy(&t->dst_cache);
1844 free_stats:
1845 free_percpu(dev->tstats);
1846 dev->tstats = NULL;
1847
1848 return ret;
1849 }
1850
1851 /**
1852 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1853 * @dev: virtual device associated with tunnel
1854 **/
1855
1856 static int ip6_tnl_dev_init(struct net_device *dev)
1857 {
1858 struct ip6_tnl *t = netdev_priv(dev);
1859 int err = ip6_tnl_dev_init_gen(dev);
1860
1861 if (err)
1862 return err;
1863 ip6_tnl_link_config(t);
1864 if (t->parms.collect_md) {
1865 dev->features |= NETIF_F_NETNS_LOCAL;
1866 netif_keep_dst(dev);
1867 }
1868 return 0;
1869 }
1870
1871 /**
1872 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1873 * @dev: fallback device
1874 *
1875 * Return: 0
1876 **/
1877
1878 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1879 {
1880 struct ip6_tnl *t = netdev_priv(dev);
1881 struct net *net = dev_net(dev);
1882 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1883
1884 t->parms.proto = IPPROTO_IPV6;
1885 dev_hold(dev);
1886
1887 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1888 return 0;
1889 }
1890
1891 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1892 {
1893 u8 proto;
1894
1895 if (!data || !data[IFLA_IPTUN_PROTO])
1896 return 0;
1897
1898 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1899 if (proto != IPPROTO_IPV6 &&
1900 proto != IPPROTO_IPIP &&
1901 proto != 0)
1902 return -EINVAL;
1903
1904 return 0;
1905 }
1906
1907 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1908 struct __ip6_tnl_parm *parms)
1909 {
1910 memset(parms, 0, sizeof(*parms));
1911
1912 if (!data)
1913 return;
1914
1915 if (data[IFLA_IPTUN_LINK])
1916 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1917
1918 if (data[IFLA_IPTUN_LOCAL])
1919 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1920
1921 if (data[IFLA_IPTUN_REMOTE])
1922 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1923
1924 if (data[IFLA_IPTUN_TTL])
1925 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1926
1927 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1928 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1929
1930 if (data[IFLA_IPTUN_FLOWINFO])
1931 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1932
1933 if (data[IFLA_IPTUN_FLAGS])
1934 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1935
1936 if (data[IFLA_IPTUN_PROTO])
1937 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1938
1939 if (data[IFLA_IPTUN_COLLECT_METADATA])
1940 parms->collect_md = true;
1941 }
1942
1943 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1944 struct ip_tunnel_encap *ipencap)
1945 {
1946 bool ret = false;
1947
1948 memset(ipencap, 0, sizeof(*ipencap));
1949
1950 if (!data)
1951 return ret;
1952
1953 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1954 ret = true;
1955 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1956 }
1957
1958 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1959 ret = true;
1960 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1961 }
1962
1963 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1964 ret = true;
1965 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1966 }
1967
1968 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1969 ret = true;
1970 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1971 }
1972
1973 return ret;
1974 }
1975
1976 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1977 struct nlattr *tb[], struct nlattr *data[])
1978 {
1979 struct net *net = dev_net(dev);
1980 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1981 struct ip_tunnel_encap ipencap;
1982 struct ip6_tnl *nt, *t;
1983 int err;
1984
1985 nt = netdev_priv(dev);
1986
1987 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1988 err = ip6_tnl_encap_setup(nt, &ipencap);
1989 if (err < 0)
1990 return err;
1991 }
1992
1993 ip6_tnl_netlink_parms(data, &nt->parms);
1994
1995 if (nt->parms.collect_md) {
1996 if (rtnl_dereference(ip6n->collect_md_tun))
1997 return -EEXIST;
1998 } else {
1999 t = ip6_tnl_locate(net, &nt->parms, 0);
2000 if (!IS_ERR(t))
2001 return -EEXIST;
2002 }
2003
2004 err = ip6_tnl_create2(dev);
2005 if (!err && tb[IFLA_MTU])
2006 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2007
2008 return err;
2009 }
2010
2011 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2012 struct nlattr *data[])
2013 {
2014 struct ip6_tnl *t = netdev_priv(dev);
2015 struct __ip6_tnl_parm p;
2016 struct net *net = t->net;
2017 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2018 struct ip_tunnel_encap ipencap;
2019
2020 if (dev == ip6n->fb_tnl_dev)
2021 return -EINVAL;
2022
2023 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2024 int err = ip6_tnl_encap_setup(t, &ipencap);
2025
2026 if (err < 0)
2027 return err;
2028 }
2029 ip6_tnl_netlink_parms(data, &p);
2030 if (p.collect_md)
2031 return -EINVAL;
2032
2033 t = ip6_tnl_locate(net, &p, 0);
2034 if (!IS_ERR(t)) {
2035 if (t->dev != dev)
2036 return -EEXIST;
2037 } else
2038 t = netdev_priv(dev);
2039
2040 return ip6_tnl_update(t, &p);
2041 }
2042
2043 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2044 {
2045 struct net *net = dev_net(dev);
2046 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2047
2048 if (dev != ip6n->fb_tnl_dev)
2049 unregister_netdevice_queue(dev, head);
2050 }
2051
2052 static size_t ip6_tnl_get_size(const struct net_device *dev)
2053 {
2054 return
2055 /* IFLA_IPTUN_LINK */
2056 nla_total_size(4) +
2057 /* IFLA_IPTUN_LOCAL */
2058 nla_total_size(sizeof(struct in6_addr)) +
2059 /* IFLA_IPTUN_REMOTE */
2060 nla_total_size(sizeof(struct in6_addr)) +
2061 /* IFLA_IPTUN_TTL */
2062 nla_total_size(1) +
2063 /* IFLA_IPTUN_ENCAP_LIMIT */
2064 nla_total_size(1) +
2065 /* IFLA_IPTUN_FLOWINFO */
2066 nla_total_size(4) +
2067 /* IFLA_IPTUN_FLAGS */
2068 nla_total_size(4) +
2069 /* IFLA_IPTUN_PROTO */
2070 nla_total_size(1) +
2071 /* IFLA_IPTUN_ENCAP_TYPE */
2072 nla_total_size(2) +
2073 /* IFLA_IPTUN_ENCAP_FLAGS */
2074 nla_total_size(2) +
2075 /* IFLA_IPTUN_ENCAP_SPORT */
2076 nla_total_size(2) +
2077 /* IFLA_IPTUN_ENCAP_DPORT */
2078 nla_total_size(2) +
2079 /* IFLA_IPTUN_COLLECT_METADATA */
2080 nla_total_size(0) +
2081 0;
2082 }
2083
2084 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2085 {
2086 struct ip6_tnl *tunnel = netdev_priv(dev);
2087 struct __ip6_tnl_parm *parm = &tunnel->parms;
2088
2089 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2090 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2091 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2092 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2093 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2094 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2095 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2096 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2097 goto nla_put_failure;
2098
2099 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2100 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2101 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2102 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2103 goto nla_put_failure;
2104
2105 if (parm->collect_md)
2106 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2107 goto nla_put_failure;
2108 return 0;
2109
2110 nla_put_failure:
2111 return -EMSGSIZE;
2112 }
2113
2114 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2115 {
2116 struct ip6_tnl *tunnel = netdev_priv(dev);
2117
2118 return tunnel->net;
2119 }
2120 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2121
2122 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2123 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2124 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2125 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2126 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2127 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2128 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2129 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2130 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2131 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2132 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2133 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2134 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2135 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2136 };
2137
2138 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2139 .kind = "ip6tnl",
2140 .maxtype = IFLA_IPTUN_MAX,
2141 .policy = ip6_tnl_policy,
2142 .priv_size = sizeof(struct ip6_tnl),
2143 .setup = ip6_tnl_dev_setup,
2144 .validate = ip6_tnl_validate,
2145 .newlink = ip6_tnl_newlink,
2146 .changelink = ip6_tnl_changelink,
2147 .dellink = ip6_tnl_dellink,
2148 .get_size = ip6_tnl_get_size,
2149 .fill_info = ip6_tnl_fill_info,
2150 .get_link_net = ip6_tnl_get_link_net,
2151 };
2152
2153 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2154 .handler = ip4ip6_rcv,
2155 .err_handler = ip4ip6_err,
2156 .priority = 1,
2157 };
2158
2159 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2160 .handler = ip6ip6_rcv,
2161 .err_handler = ip6ip6_err,
2162 .priority = 1,
2163 };
2164
2165 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2166 {
2167 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2168 struct net_device *dev, *aux;
2169 int h;
2170 struct ip6_tnl *t;
2171 LIST_HEAD(list);
2172
2173 for_each_netdev_safe(net, dev, aux)
2174 if (dev->rtnl_link_ops == &ip6_link_ops)
2175 unregister_netdevice_queue(dev, &list);
2176
2177 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2178 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2179 while (t) {
2180 /* If dev is in the same netns, it has already
2181 * been added to the list by the previous loop.
2182 */
2183 if (!net_eq(dev_net(t->dev), net))
2184 unregister_netdevice_queue(t->dev, &list);
2185 t = rtnl_dereference(t->next);
2186 }
2187 }
2188
2189 unregister_netdevice_many(&list);
2190 }
2191
2192 static int __net_init ip6_tnl_init_net(struct net *net)
2193 {
2194 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2195 struct ip6_tnl *t = NULL;
2196 int err;
2197
2198 ip6n->tnls[0] = ip6n->tnls_wc;
2199 ip6n->tnls[1] = ip6n->tnls_r_l;
2200
2201 err = -ENOMEM;
2202 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2203 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2204
2205 if (!ip6n->fb_tnl_dev)
2206 goto err_alloc_dev;
2207 dev_net_set(ip6n->fb_tnl_dev, net);
2208 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2209 /* FB netdevice is special: we have one, and only one per netns.
2210 * Allowing to move it to another netns is clearly unsafe.
2211 */
2212 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2213
2214 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2215 if (err < 0)
2216 goto err_register;
2217
2218 err = register_netdev(ip6n->fb_tnl_dev);
2219 if (err < 0)
2220 goto err_register;
2221
2222 t = netdev_priv(ip6n->fb_tnl_dev);
2223
2224 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2225 return 0;
2226
2227 err_register:
2228 ip6_dev_free(ip6n->fb_tnl_dev);
2229 err_alloc_dev:
2230 return err;
2231 }
2232
2233 static void __net_exit ip6_tnl_exit_net(struct net *net)
2234 {
2235 rtnl_lock();
2236 ip6_tnl_destroy_tunnels(net);
2237 rtnl_unlock();
2238 }
2239
2240 static struct pernet_operations ip6_tnl_net_ops = {
2241 .init = ip6_tnl_init_net,
2242 .exit = ip6_tnl_exit_net,
2243 .id = &ip6_tnl_net_id,
2244 .size = sizeof(struct ip6_tnl_net),
2245 };
2246
2247 /**
2248 * ip6_tunnel_init - register protocol and reserve needed resources
2249 *
2250 * Return: 0 on success
2251 **/
2252
2253 static int __init ip6_tunnel_init(void)
2254 {
2255 int err;
2256
2257 if (!ipv6_mod_enabled())
2258 return -EOPNOTSUPP;
2259
2260 err = register_pernet_device(&ip6_tnl_net_ops);
2261 if (err < 0)
2262 goto out_pernet;
2263
2264 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2265 if (err < 0) {
2266 pr_err("%s: can't register ip4ip6\n", __func__);
2267 goto out_ip4ip6;
2268 }
2269
2270 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2271 if (err < 0) {
2272 pr_err("%s: can't register ip6ip6\n", __func__);
2273 goto out_ip6ip6;
2274 }
2275 err = rtnl_link_register(&ip6_link_ops);
2276 if (err < 0)
2277 goto rtnl_link_failed;
2278
2279 return 0;
2280
2281 rtnl_link_failed:
2282 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2283 out_ip6ip6:
2284 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2285 out_ip4ip6:
2286 unregister_pernet_device(&ip6_tnl_net_ops);
2287 out_pernet:
2288 return err;
2289 }
2290
2291 /**
2292 * ip6_tunnel_cleanup - free resources and unregister protocol
2293 **/
2294
2295 static void __exit ip6_tunnel_cleanup(void)
2296 {
2297 rtnl_link_unregister(&ip6_link_ops);
2298 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2299 pr_info("%s: can't deregister ip4ip6\n", __func__);
2300
2301 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2302 pr_info("%s: can't deregister ip6ip6\n", __func__);
2303
2304 unregister_pernet_device(&ip6_tnl_net_ops);
2305 }
2306
2307 module_init(ip6_tunnel_init);
2308 module_exit(ip6_tunnel_cleanup);
Linux with added FPC-III support