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 <linux/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
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)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 unsigned 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
ip6_get_stats(struct net_device * dev)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 *
ip6_tnl_lookup(struct net * net,const struct in6_addr * remote,const struct in6_addr * local)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 && t->dev->flags & IFF_UP)
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 **
ip6_tnl_bucket(struct ip6_tnl_net * ip6n,const struct __ip6_tnl_parm * p)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
ip6_tnl_link(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)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
ip6_tnl_unlink(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)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
ip6_dev_free(struct net_device * dev)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 }
258
ip6_tnl_create2(struct net_device * dev)259 static int ip6_tnl_create2(struct net_device *dev)
260 {
261 struct ip6_tnl *t = netdev_priv(dev);
262 struct net *net = dev_net(dev);
263 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264 int err;
265
266 t = netdev_priv(dev);
267
268 dev->rtnl_link_ops = &ip6_link_ops;
269 err = register_netdevice(dev);
270 if (err < 0)
271 goto out;
272
273 strcpy(t->parms.name, dev->name);
274
275 dev_hold(dev);
276 ip6_tnl_link(ip6n, t);
277 return 0;
278
279 out:
280 return err;
281 }
282
283 /**
284 * ip6_tnl_create - create a new tunnel
285 * @p: tunnel parameters
286 * @pt: pointer to new tunnel
287 *
288 * Description:
289 * Create tunnel matching given parameters.
290 *
291 * Return:
292 * created tunnel or error pointer
293 **/
294
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 {
297 struct net_device *dev;
298 struct ip6_tnl *t;
299 char name[IFNAMSIZ];
300 int err = -E2BIG;
301
302 if (p->name[0]) {
303 if (!dev_valid_name(p->name))
304 goto failed;
305 strlcpy(name, p->name, IFNAMSIZ);
306 } else {
307 sprintf(name, "ip6tnl%%d");
308 }
309 err = -ENOMEM;
310 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
311 ip6_tnl_dev_setup);
312 if (!dev)
313 goto failed;
314
315 dev_net_set(dev, net);
316
317 t = netdev_priv(dev);
318 t->parms = *p;
319 t->net = dev_net(dev);
320 err = ip6_tnl_create2(dev);
321 if (err < 0)
322 goto failed_free;
323
324 return t;
325
326 failed_free:
327 free_netdev(dev);
328 failed:
329 return ERR_PTR(err);
330 }
331
332 /**
333 * ip6_tnl_locate - find or create tunnel matching given parameters
334 * @p: tunnel parameters
335 * @create: != 0 if allowed to create new tunnel if no match found
336 *
337 * Description:
338 * ip6_tnl_locate() first tries to locate an existing tunnel
339 * based on @parms. If this is unsuccessful, but @create is set a new
340 * tunnel device is created and registered for use.
341 *
342 * Return:
343 * matching tunnel or error pointer
344 **/
345
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 struct __ip6_tnl_parm *p, int create)
348 {
349 const struct in6_addr *remote = &p->raddr;
350 const struct in6_addr *local = &p->laddr;
351 struct ip6_tnl __rcu **tp;
352 struct ip6_tnl *t;
353 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354
355 for (tp = ip6_tnl_bucket(ip6n, p);
356 (t = rtnl_dereference(*tp)) != NULL;
357 tp = &t->next) {
358 if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 ipv6_addr_equal(remote, &t->parms.raddr)) {
360 if (create)
361 return ERR_PTR(-EEXIST);
362
363 return t;
364 }
365 }
366 if (!create)
367 return ERR_PTR(-ENODEV);
368 return ip6_tnl_create(net, p);
369 }
370
371 /**
372 * ip6_tnl_dev_uninit - tunnel device uninitializer
373 * @dev: the device to be destroyed
374 *
375 * Description:
376 * ip6_tnl_dev_uninit() removes tunnel from its list
377 **/
378
379 static void
ip6_tnl_dev_uninit(struct net_device * dev)380 ip6_tnl_dev_uninit(struct net_device *dev)
381 {
382 struct ip6_tnl *t = netdev_priv(dev);
383 struct net *net = t->net;
384 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
385
386 if (dev == ip6n->fb_tnl_dev)
387 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
388 else
389 ip6_tnl_unlink(ip6n, t);
390 dst_cache_reset(&t->dst_cache);
391 dev_put(dev);
392 }
393
394 /**
395 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
396 * @skb: received socket buffer
397 *
398 * Return:
399 * 0 if none was found,
400 * else index to encapsulation limit
401 **/
402
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
404 {
405 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
406 unsigned int nhoff = raw - skb->data;
407 unsigned int off = nhoff + sizeof(*ipv6h);
408 u8 next, nexthdr = ipv6h->nexthdr;
409
410 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
411 struct ipv6_opt_hdr *hdr;
412 u16 optlen;
413
414 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
415 break;
416
417 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
418 if (nexthdr == NEXTHDR_FRAGMENT) {
419 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
420 if (frag_hdr->frag_off)
421 break;
422 optlen = 8;
423 } else if (nexthdr == NEXTHDR_AUTH) {
424 optlen = (hdr->hdrlen + 2) << 2;
425 } else {
426 optlen = ipv6_optlen(hdr);
427 }
428 /* cache hdr->nexthdr, since pskb_may_pull() might
429 * invalidate hdr
430 */
431 next = hdr->nexthdr;
432 if (nexthdr == NEXTHDR_DEST) {
433 u16 i = 2;
434
435 /* Remember : hdr is no longer valid at this point. */
436 if (!pskb_may_pull(skb, off + optlen))
437 break;
438
439 while (1) {
440 struct ipv6_tlv_tnl_enc_lim *tel;
441
442 /* No more room for encapsulation limit */
443 if (i + sizeof(*tel) > optlen)
444 break;
445
446 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
447 /* return index of option if found and valid */
448 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
449 tel->length == 1)
450 return i + off - nhoff;
451 /* else jump to next option */
452 if (tel->type)
453 i += tel->length + 2;
454 else
455 i++;
456 }
457 }
458 nexthdr = next;
459 off += optlen;
460 }
461 return 0;
462 }
463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
464
465 /**
466 * ip6_tnl_err - tunnel error handler
467 *
468 * Description:
469 * ip6_tnl_err() should handle errors in the tunnel according
470 * to the specifications in RFC 2473.
471 **/
472
473 static int
ip6_tnl_err(struct sk_buff * skb,__u8 ipproto,struct inet6_skb_parm * opt,u8 * type,u8 * code,int * msg,__u32 * info,int offset)474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
475 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
476 {
477 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
478 struct ip6_tnl *t;
479 int rel_msg = 0;
480 u8 rel_type = ICMPV6_DEST_UNREACH;
481 u8 rel_code = ICMPV6_ADDR_UNREACH;
482 u8 tproto;
483 __u32 rel_info = 0;
484 __u16 len;
485 int err = -ENOENT;
486
487 /* If the packet doesn't contain the original IPv6 header we are
488 in trouble since we might need the source address for further
489 processing of the error. */
490
491 rcu_read_lock();
492 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
493 if (!t)
494 goto out;
495
496 tproto = ACCESS_ONCE(t->parms.proto);
497 if (tproto != ipproto && tproto != 0)
498 goto out;
499
500 err = 0;
501
502 switch (*type) {
503 __u32 teli;
504 struct ipv6_tlv_tnl_enc_lim *tel;
505 __u32 mtu;
506 case ICMPV6_DEST_UNREACH:
507 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
508 t->parms.name);
509 rel_msg = 1;
510 break;
511 case ICMPV6_TIME_EXCEED:
512 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
514 t->parms.name);
515 rel_msg = 1;
516 }
517 break;
518 case ICMPV6_PARAMPROB:
519 teli = 0;
520 if ((*code) == ICMPV6_HDR_FIELD)
521 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522
523 if (teli && teli == *info - 2) {
524 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525 if (tel->encap_limit == 0) {
526 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
527 t->parms.name);
528 rel_msg = 1;
529 }
530 } else {
531 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
532 t->parms.name);
533 }
534 break;
535 case ICMPV6_PKT_TOOBIG:
536 mtu = *info - offset;
537 if (mtu < IPV6_MIN_MTU)
538 mtu = IPV6_MIN_MTU;
539 t->dev->mtu = mtu;
540
541 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
542 if (len > mtu) {
543 rel_type = ICMPV6_PKT_TOOBIG;
544 rel_code = 0;
545 rel_info = mtu;
546 rel_msg = 1;
547 }
548 break;
549 }
550
551 *type = rel_type;
552 *code = rel_code;
553 *info = rel_info;
554 *msg = rel_msg;
555
556 out:
557 rcu_read_unlock();
558 return err;
559 }
560
561 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 u8 type, u8 code, int offset, __be32 info)
564 {
565 int rel_msg = 0;
566 u8 rel_type = type;
567 u8 rel_code = code;
568 __u32 rel_info = ntohl(info);
569 int err;
570 struct sk_buff *skb2;
571 const struct iphdr *eiph;
572 struct rtable *rt;
573 struct flowi4 fl4;
574
575 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576 &rel_msg, &rel_info, offset);
577 if (err < 0)
578 return err;
579
580 if (rel_msg == 0)
581 return 0;
582
583 switch (rel_type) {
584 case ICMPV6_DEST_UNREACH:
585 if (rel_code != ICMPV6_ADDR_UNREACH)
586 return 0;
587 rel_type = ICMP_DEST_UNREACH;
588 rel_code = ICMP_HOST_UNREACH;
589 break;
590 case ICMPV6_PKT_TOOBIG:
591 if (rel_code != 0)
592 return 0;
593 rel_type = ICMP_DEST_UNREACH;
594 rel_code = ICMP_FRAG_NEEDED;
595 break;
596 case NDISC_REDIRECT:
597 rel_type = ICMP_REDIRECT;
598 rel_code = ICMP_REDIR_HOST;
599 default:
600 return 0;
601 }
602
603 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
604 return 0;
605
606 skb2 = skb_clone(skb, GFP_ATOMIC);
607 if (!skb2)
608 return 0;
609
610 skb_dst_drop(skb2);
611
612 skb_pull(skb2, offset);
613 skb_reset_network_header(skb2);
614 eiph = ip_hdr(skb2);
615
616 /* Try to guess incoming interface */
617 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
618 eiph->saddr, 0,
619 0, 0,
620 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
621 if (IS_ERR(rt))
622 goto out;
623
624 skb2->dev = rt->dst.dev;
625
626 /* route "incoming" packet */
627 if (rt->rt_flags & RTCF_LOCAL) {
628 ip_rt_put(rt);
629 rt = NULL;
630 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
631 eiph->daddr, eiph->saddr,
632 0, 0,
633 IPPROTO_IPIP,
634 RT_TOS(eiph->tos), 0);
635 if (IS_ERR(rt) ||
636 rt->dst.dev->type != ARPHRD_TUNNEL6) {
637 if (!IS_ERR(rt))
638 ip_rt_put(rt);
639 goto out;
640 }
641 skb_dst_set(skb2, &rt->dst);
642 } else {
643 ip_rt_put(rt);
644 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
645 skb2->dev) ||
646 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
647 goto out;
648 }
649
650 /* change mtu on this route */
651 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
652 if (rel_info > dst_mtu(skb_dst(skb2)))
653 goto out;
654
655 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
656 }
657 if (rel_type == ICMP_REDIRECT)
658 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
659
660 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
661
662 out:
663 kfree_skb(skb2);
664 return 0;
665 }
666
667 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)668 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
669 u8 type, u8 code, int offset, __be32 info)
670 {
671 int rel_msg = 0;
672 u8 rel_type = type;
673 u8 rel_code = code;
674 __u32 rel_info = ntohl(info);
675 int err;
676
677 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
678 &rel_msg, &rel_info, offset);
679 if (err < 0)
680 return err;
681
682 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
683 struct rt6_info *rt;
684 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
685
686 if (!skb2)
687 return 0;
688
689 skb_dst_drop(skb2);
690 skb_pull(skb2, offset);
691 skb_reset_network_header(skb2);
692
693 /* Try to guess incoming interface */
694 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
695 NULL, 0, 0);
696
697 if (rt && rt->dst.dev)
698 skb2->dev = rt->dst.dev;
699
700 icmpv6_send(skb2, rel_type, rel_code, rel_info);
701
702 ip6_rt_put(rt);
703
704 kfree_skb(skb2);
705 }
706
707 return 0;
708 }
709
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
711 const struct ipv6hdr *ipv6h,
712 struct sk_buff *skb)
713 {
714 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
715
716 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
718
719 return IP6_ECN_decapsulate(ipv6h, skb);
720 }
721
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
723 const struct ipv6hdr *ipv6h,
724 struct sk_buff *skb)
725 {
726 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
727 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
728
729 return IP6_ECN_decapsulate(ipv6h, skb);
730 }
731
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)732 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
733 const struct in6_addr *laddr,
734 const struct in6_addr *raddr)
735 {
736 struct __ip6_tnl_parm *p = &t->parms;
737 int ltype = ipv6_addr_type(laddr);
738 int rtype = ipv6_addr_type(raddr);
739 __u32 flags = 0;
740
741 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
742 flags = IP6_TNL_F_CAP_PER_PACKET;
743 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
744 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
745 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
746 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
747 if (ltype&IPV6_ADDR_UNICAST)
748 flags |= IP6_TNL_F_CAP_XMIT;
749 if (rtype&IPV6_ADDR_UNICAST)
750 flags |= IP6_TNL_F_CAP_RCV;
751 }
752 return flags;
753 }
754 EXPORT_SYMBOL(ip6_tnl_get_cap);
755
756 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)757 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
758 const struct in6_addr *laddr,
759 const struct in6_addr *raddr)
760 {
761 struct __ip6_tnl_parm *p = &t->parms;
762 int ret = 0;
763 struct net *net = t->net;
764
765 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
766 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
767 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
768 struct net_device *ldev = NULL;
769
770 if (p->link)
771 ldev = dev_get_by_index_rcu(net, p->link);
772
773 if ((ipv6_addr_is_multicast(laddr) ||
774 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
775 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
776 ret = 1;
777 }
778 return ret;
779 }
780 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
781
__ip6_tnl_rcv(struct ip6_tnl * tunnel,struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct metadata_dst * tun_dst,int (* dscp_ecn_decapsulate)(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb),bool log_ecn_err)782 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
783 const struct tnl_ptk_info *tpi,
784 struct metadata_dst *tun_dst,
785 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
786 const struct ipv6hdr *ipv6h,
787 struct sk_buff *skb),
788 bool log_ecn_err)
789 {
790 struct pcpu_sw_netstats *tstats;
791 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
792 int err;
793
794 if ((!(tpi->flags & TUNNEL_CSUM) &&
795 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
796 ((tpi->flags & TUNNEL_CSUM) &&
797 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
798 tunnel->dev->stats.rx_crc_errors++;
799 tunnel->dev->stats.rx_errors++;
800 goto drop;
801 }
802
803 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
804 if (!(tpi->flags & TUNNEL_SEQ) ||
805 (tunnel->i_seqno &&
806 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
807 tunnel->dev->stats.rx_fifo_errors++;
808 tunnel->dev->stats.rx_errors++;
809 goto drop;
810 }
811 tunnel->i_seqno = ntohl(tpi->seq) + 1;
812 }
813
814 skb->protocol = tpi->proto;
815
816 /* Warning: All skb pointers will be invalidated! */
817 if (tunnel->dev->type == ARPHRD_ETHER) {
818 if (!pskb_may_pull(skb, ETH_HLEN)) {
819 tunnel->dev->stats.rx_length_errors++;
820 tunnel->dev->stats.rx_errors++;
821 goto drop;
822 }
823
824 ipv6h = ipv6_hdr(skb);
825 skb->protocol = eth_type_trans(skb, tunnel->dev);
826 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
827 } else {
828 skb->dev = tunnel->dev;
829 }
830
831 skb_reset_network_header(skb);
832 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
833
834 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
835
836 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
837 if (unlikely(err)) {
838 if (log_ecn_err)
839 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
840 &ipv6h->saddr,
841 ipv6_get_dsfield(ipv6h));
842 if (err > 1) {
843 ++tunnel->dev->stats.rx_frame_errors;
844 ++tunnel->dev->stats.rx_errors;
845 goto drop;
846 }
847 }
848
849 tstats = this_cpu_ptr(tunnel->dev->tstats);
850 u64_stats_update_begin(&tstats->syncp);
851 tstats->rx_packets++;
852 tstats->rx_bytes += skb->len;
853 u64_stats_update_end(&tstats->syncp);
854
855 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
856
857 if (tun_dst)
858 skb_dst_set(skb, (struct dst_entry *)tun_dst);
859
860 gro_cells_receive(&tunnel->gro_cells, skb);
861 return 0;
862
863 drop:
864 if (tun_dst)
865 dst_release((struct dst_entry *)tun_dst);
866 kfree_skb(skb);
867 return 0;
868 }
869
ip6_tnl_rcv(struct ip6_tnl * t,struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct metadata_dst * tun_dst,bool log_ecn_err)870 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
871 const struct tnl_ptk_info *tpi,
872 struct metadata_dst *tun_dst,
873 bool log_ecn_err)
874 {
875 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
876 log_ecn_err);
877 }
878 EXPORT_SYMBOL(ip6_tnl_rcv);
879
880 static const struct tnl_ptk_info tpi_v6 = {
881 /* no tunnel info required for ipxip6. */
882 .proto = htons(ETH_P_IPV6),
883 };
884
885 static const struct tnl_ptk_info tpi_v4 = {
886 /* no tunnel info required for ipxip6. */
887 .proto = htons(ETH_P_IP),
888 };
889
ipxip6_rcv(struct sk_buff * skb,u8 ipproto,const struct tnl_ptk_info * tpi,int (* dscp_ecn_decapsulate)(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb))890 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
891 const struct tnl_ptk_info *tpi,
892 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
893 const struct ipv6hdr *ipv6h,
894 struct sk_buff *skb))
895 {
896 struct ip6_tnl *t;
897 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
898 struct metadata_dst *tun_dst = NULL;
899 int ret = -1;
900
901 rcu_read_lock();
902 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
903
904 if (t) {
905 u8 tproto = ACCESS_ONCE(t->parms.proto);
906
907 if (tproto != ipproto && tproto != 0)
908 goto drop;
909 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
910 goto drop;
911 ipv6h = ipv6_hdr(skb);
912 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
913 goto drop;
914 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
915 goto drop;
916 if (t->parms.collect_md) {
917 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
918 if (!tun_dst)
919 goto drop;
920 }
921 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
922 log_ecn_error);
923 }
924
925 rcu_read_unlock();
926
927 return ret;
928
929 drop:
930 rcu_read_unlock();
931 kfree_skb(skb);
932 return 0;
933 }
934
ip4ip6_rcv(struct sk_buff * skb)935 static int ip4ip6_rcv(struct sk_buff *skb)
936 {
937 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
938 ip4ip6_dscp_ecn_decapsulate);
939 }
940
ip6ip6_rcv(struct sk_buff * skb)941 static int ip6ip6_rcv(struct sk_buff *skb)
942 {
943 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
944 ip6ip6_dscp_ecn_decapsulate);
945 }
946
947 struct ipv6_tel_txoption {
948 struct ipv6_txoptions ops;
949 __u8 dst_opt[8];
950 };
951
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)952 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
953 {
954 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
955
956 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
957 opt->dst_opt[3] = 1;
958 opt->dst_opt[4] = encap_limit;
959 opt->dst_opt[5] = IPV6_TLV_PADN;
960 opt->dst_opt[6] = 1;
961
962 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
963 opt->ops.opt_nflen = 8;
964 }
965
966 /**
967 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
968 * @t: the outgoing tunnel device
969 * @hdr: IPv6 header from the incoming packet
970 *
971 * Description:
972 * Avoid trivial tunneling loop by checking that tunnel exit-point
973 * doesn't match source of incoming packet.
974 *
975 * Return:
976 * 1 if conflict,
977 * 0 else
978 **/
979
980 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)981 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
982 {
983 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
984 }
985
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)986 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
987 const struct in6_addr *laddr,
988 const struct in6_addr *raddr)
989 {
990 struct __ip6_tnl_parm *p = &t->parms;
991 int ret = 0;
992 struct net *net = t->net;
993
994 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
995 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
996 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
997 struct net_device *ldev = NULL;
998
999 rcu_read_lock();
1000 if (p->link)
1001 ldev = dev_get_by_index_rcu(net, p->link);
1002
1003 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1004 pr_warn("%s xmit: Local address not yet configured!\n",
1005 p->name);
1006 else if (!ipv6_addr_is_multicast(raddr) &&
1007 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1008 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1009 p->name);
1010 else
1011 ret = 1;
1012 rcu_read_unlock();
1013 }
1014 return ret;
1015 }
1016 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1017
1018 /**
1019 * ip6_tnl_xmit - encapsulate packet and send
1020 * @skb: the outgoing socket buffer
1021 * @dev: the outgoing tunnel device
1022 * @dsfield: dscp code for outer header
1023 * @fl6: flow of tunneled packet
1024 * @encap_limit: encapsulation limit
1025 * @pmtu: Path MTU is stored if packet is too big
1026 * @proto: next header value
1027 *
1028 * Description:
1029 * Build new header and do some sanity checks on the packet before sending
1030 * it.
1031 *
1032 * Return:
1033 * 0 on success
1034 * -1 fail
1035 * %-EMSGSIZE message too big. return mtu in this case.
1036 **/
1037
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1038 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1039 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1040 __u8 proto)
1041 {
1042 struct ip6_tnl *t = netdev_priv(dev);
1043 struct net *net = t->net;
1044 struct net_device_stats *stats = &t->dev->stats;
1045 struct ipv6hdr *ipv6h;
1046 struct ipv6_tel_txoption opt;
1047 struct dst_entry *dst = NULL, *ndst = NULL;
1048 struct net_device *tdev;
1049 int mtu;
1050 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1051 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1052 unsigned int max_headroom = psh_hlen;
1053 bool use_cache = false;
1054 u8 hop_limit;
1055 int err = -1;
1056
1057 if (t->parms.collect_md) {
1058 hop_limit = skb_tunnel_info(skb)->key.ttl;
1059 goto route_lookup;
1060 } else {
1061 hop_limit = t->parms.hop_limit;
1062 }
1063
1064 /* NBMA tunnel */
1065 if (ipv6_addr_any(&t->parms.raddr)) {
1066 if (skb->protocol == htons(ETH_P_IPV6)) {
1067 struct in6_addr *addr6;
1068 struct neighbour *neigh;
1069 int addr_type;
1070
1071 if (!skb_dst(skb))
1072 goto tx_err_link_failure;
1073
1074 neigh = dst_neigh_lookup(skb_dst(skb),
1075 &ipv6_hdr(skb)->daddr);
1076 if (!neigh)
1077 goto tx_err_link_failure;
1078
1079 addr6 = (struct in6_addr *)&neigh->primary_key;
1080 addr_type = ipv6_addr_type(addr6);
1081
1082 if (addr_type == IPV6_ADDR_ANY)
1083 addr6 = &ipv6_hdr(skb)->daddr;
1084
1085 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1086 neigh_release(neigh);
1087 }
1088 } else if (t->parms.proto != 0 && !(t->parms.flags &
1089 (IP6_TNL_F_USE_ORIG_TCLASS |
1090 IP6_TNL_F_USE_ORIG_FWMARK))) {
1091 /* enable the cache only if neither the outer protocol nor the
1092 * routing decision depends on the current inner header value
1093 */
1094 use_cache = true;
1095 }
1096
1097 if (use_cache)
1098 dst = dst_cache_get(&t->dst_cache);
1099
1100 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1101 goto tx_err_link_failure;
1102
1103 if (!dst) {
1104 route_lookup:
1105 /* add dsfield to flowlabel for route lookup */
1106 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1107
1108 dst = ip6_route_output(net, NULL, fl6);
1109
1110 if (dst->error)
1111 goto tx_err_link_failure;
1112 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1113 if (IS_ERR(dst)) {
1114 err = PTR_ERR(dst);
1115 dst = NULL;
1116 goto tx_err_link_failure;
1117 }
1118 if (t->parms.collect_md &&
1119 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1120 &fl6->daddr, 0, &fl6->saddr))
1121 goto tx_err_link_failure;
1122 ndst = dst;
1123 }
1124
1125 tdev = dst->dev;
1126
1127 if (tdev == dev) {
1128 stats->collisions++;
1129 net_warn_ratelimited("%s: Local routing loop detected!\n",
1130 t->parms.name);
1131 goto tx_err_dst_release;
1132 }
1133 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1134 if (encap_limit >= 0) {
1135 max_headroom += 8;
1136 mtu -= 8;
1137 }
1138 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1139 IPV6_MIN_MTU : IPV4_MIN_MTU);
1140
1141 skb_dst_update_pmtu_no_confirm(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 hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1190
1191 /* Calculate max headroom for all the headers and adjust
1192 * needed_headroom if necessary.
1193 */
1194 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1195 + dst->header_len + t->hlen;
1196 if (max_headroom > dev->needed_headroom)
1197 dev->needed_headroom = max_headroom;
1198
1199 err = ip6_tnl_encap(skb, t, &proto, fl6);
1200 if (err)
1201 return err;
1202
1203 if (encap_limit >= 0) {
1204 init_tel_txopt(&opt, encap_limit);
1205 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1206 }
1207
1208 skb_push(skb, sizeof(struct ipv6hdr));
1209 skb_reset_network_header(skb);
1210 ipv6h = ipv6_hdr(skb);
1211 ip6_flow_hdr(ipv6h, dsfield,
1212 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1213 ipv6h->hop_limit = hop_limit;
1214 ipv6h->nexthdr = proto;
1215 ipv6h->saddr = fl6->saddr;
1216 ipv6h->daddr = fl6->daddr;
1217 ip6tunnel_xmit(NULL, skb, dev);
1218 return 0;
1219 tx_err_link_failure:
1220 stats->tx_carrier_errors++;
1221 dst_link_failure(skb);
1222 tx_err_dst_release:
1223 dst_release(dst);
1224 return err;
1225 }
1226 EXPORT_SYMBOL(ip6_tnl_xmit);
1227
1228 static inline int
ip4ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)1229 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1230 {
1231 struct ip6_tnl *t = netdev_priv(dev);
1232 const struct iphdr *iph;
1233 int encap_limit = -1;
1234 struct flowi6 fl6;
1235 __u8 dsfield;
1236 __u32 mtu;
1237 u8 tproto;
1238 int err;
1239
1240 /* ensure we can access the full inner ip header */
1241 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1242 return -1;
1243
1244 iph = ip_hdr(skb);
1245 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1246
1247 tproto = ACCESS_ONCE(t->parms.proto);
1248 if (tproto != IPPROTO_IPIP && tproto != 0)
1249 return -1;
1250
1251 if (t->parms.collect_md) {
1252 struct ip_tunnel_info *tun_info;
1253 const struct ip_tunnel_key *key;
1254
1255 tun_info = skb_tunnel_info(skb);
1256 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1257 ip_tunnel_info_af(tun_info) != AF_INET6))
1258 return -1;
1259 key = &tun_info->key;
1260 memset(&fl6, 0, sizeof(fl6));
1261 fl6.flowi6_proto = IPPROTO_IPIP;
1262 fl6.daddr = key->u.ipv6.dst;
1263 fl6.flowlabel = key->label;
1264 dsfield = key->tos;
1265 } else {
1266 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1267 encap_limit = t->parms.encap_limit;
1268
1269 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1270 fl6.flowi6_proto = IPPROTO_IPIP;
1271
1272 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1273 dsfield = ipv4_get_dsfield(iph);
1274 else
1275 dsfield = ip6_tclass(t->parms.flowinfo);
1276 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1277 fl6.flowi6_mark = skb->mark;
1278 else
1279 fl6.flowi6_mark = t->parms.fwmark;
1280 }
1281
1282 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1283 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1284
1285 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1286 return -1;
1287
1288 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1289
1290 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1291 IPPROTO_IPIP);
1292 if (err != 0) {
1293 /* XXX: send ICMP error even if DF is not set. */
1294 if (err == -EMSGSIZE)
1295 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1296 htonl(mtu));
1297 return -1;
1298 }
1299
1300 return 0;
1301 }
1302
1303 static inline int
ip6ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)1304 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1305 {
1306 struct ip6_tnl *t = netdev_priv(dev);
1307 struct ipv6hdr *ipv6h;
1308 int encap_limit = -1;
1309 __u16 offset;
1310 struct flowi6 fl6;
1311 __u8 dsfield;
1312 __u32 mtu;
1313 u8 tproto;
1314 int err;
1315
1316 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1317 return -1;
1318
1319 ipv6h = ipv6_hdr(skb);
1320 tproto = ACCESS_ONCE(t->parms.proto);
1321 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1322 ip6_tnl_addr_conflict(t, ipv6h))
1323 return -1;
1324
1325 if (t->parms.collect_md) {
1326 struct ip_tunnel_info *tun_info;
1327 const struct ip_tunnel_key *key;
1328
1329 tun_info = skb_tunnel_info(skb);
1330 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1331 ip_tunnel_info_af(tun_info) != AF_INET6))
1332 return -1;
1333 key = &tun_info->key;
1334 memset(&fl6, 0, sizeof(fl6));
1335 fl6.flowi6_proto = IPPROTO_IPV6;
1336 fl6.daddr = key->u.ipv6.dst;
1337 fl6.flowlabel = key->label;
1338 dsfield = key->tos;
1339 } else {
1340 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1341 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1342 ipv6h = ipv6_hdr(skb);
1343 if (offset > 0) {
1344 struct ipv6_tlv_tnl_enc_lim *tel;
1345
1346 tel = (void *)&skb_network_header(skb)[offset];
1347 if (tel->encap_limit == 0) {
1348 icmpv6_send(skb, ICMPV6_PARAMPROB,
1349 ICMPV6_HDR_FIELD, offset + 2);
1350 return -1;
1351 }
1352 encap_limit = tel->encap_limit - 1;
1353 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1354 encap_limit = t->parms.encap_limit;
1355 }
1356
1357 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1358 fl6.flowi6_proto = IPPROTO_IPV6;
1359
1360 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1361 dsfield = ipv6_get_dsfield(ipv6h);
1362 else
1363 dsfield = ip6_tclass(t->parms.flowinfo);
1364 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1365 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1366 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1367 fl6.flowi6_mark = skb->mark;
1368 else
1369 fl6.flowi6_mark = t->parms.fwmark;
1370 }
1371
1372 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1373 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1374
1375 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1376 return -1;
1377
1378 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1379
1380 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1381 IPPROTO_IPV6);
1382 if (err != 0) {
1383 if (err == -EMSGSIZE)
1384 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1385 return -1;
1386 }
1387
1388 return 0;
1389 }
1390
1391 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1392 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1393 {
1394 struct ip6_tnl *t = netdev_priv(dev);
1395 struct net_device_stats *stats = &t->dev->stats;
1396 int ret;
1397
1398 switch (skb->protocol) {
1399 case htons(ETH_P_IP):
1400 ret = ip4ip6_tnl_xmit(skb, dev);
1401 break;
1402 case htons(ETH_P_IPV6):
1403 ret = ip6ip6_tnl_xmit(skb, dev);
1404 break;
1405 default:
1406 goto tx_err;
1407 }
1408
1409 if (ret < 0)
1410 goto tx_err;
1411
1412 return NETDEV_TX_OK;
1413
1414 tx_err:
1415 stats->tx_errors++;
1416 stats->tx_dropped++;
1417 kfree_skb(skb);
1418 return NETDEV_TX_OK;
1419 }
1420
ip6_tnl_link_config(struct ip6_tnl * t)1421 static void ip6_tnl_link_config(struct ip6_tnl *t)
1422 {
1423 struct net_device *dev = t->dev;
1424 struct __ip6_tnl_parm *p = &t->parms;
1425 struct flowi6 *fl6 = &t->fl.u.ip6;
1426 int t_hlen;
1427
1428 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1429 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1430
1431 /* Set up flowi template */
1432 fl6->saddr = p->laddr;
1433 fl6->daddr = p->raddr;
1434 fl6->flowi6_oif = p->link;
1435 fl6->flowlabel = 0;
1436
1437 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1438 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1439 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1440 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1441
1442 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1443 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1444
1445 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1446 dev->flags |= IFF_POINTOPOINT;
1447 else
1448 dev->flags &= ~IFF_POINTOPOINT;
1449
1450 t->tun_hlen = 0;
1451 t->hlen = t->encap_hlen + t->tun_hlen;
1452 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1453
1454 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1455 int strict = (ipv6_addr_type(&p->raddr) &
1456 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1457
1458 struct rt6_info *rt = rt6_lookup(t->net,
1459 &p->raddr, &p->laddr,
1460 p->link, strict);
1461
1462 if (!rt)
1463 return;
1464
1465 if (rt->dst.dev) {
1466 dev->hard_header_len = rt->dst.dev->hard_header_len +
1467 t_hlen;
1468
1469 dev->mtu = rt->dst.dev->mtu - t_hlen;
1470 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1471 dev->mtu -= 8;
1472
1473 if (dev->mtu < IPV6_MIN_MTU)
1474 dev->mtu = IPV6_MIN_MTU;
1475 }
1476 ip6_rt_put(rt);
1477 }
1478 }
1479
1480 /**
1481 * ip6_tnl_change - update the tunnel parameters
1482 * @t: tunnel to be changed
1483 * @p: tunnel configuration parameters
1484 *
1485 * Description:
1486 * ip6_tnl_change() updates the tunnel parameters
1487 **/
1488
1489 static int
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1490 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1491 {
1492 t->parms.laddr = p->laddr;
1493 t->parms.raddr = p->raddr;
1494 t->parms.flags = p->flags;
1495 t->parms.hop_limit = p->hop_limit;
1496 t->parms.encap_limit = p->encap_limit;
1497 t->parms.flowinfo = p->flowinfo;
1498 t->parms.link = p->link;
1499 t->parms.proto = p->proto;
1500 t->parms.fwmark = p->fwmark;
1501 dst_cache_reset(&t->dst_cache);
1502 ip6_tnl_link_config(t);
1503 return 0;
1504 }
1505
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1506 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1507 {
1508 struct net *net = t->net;
1509 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1510 int err;
1511
1512 ip6_tnl_unlink(ip6n, t);
1513 synchronize_net();
1514 err = ip6_tnl_change(t, p);
1515 ip6_tnl_link(ip6n, t);
1516 netdev_state_change(t->dev);
1517 return err;
1518 }
1519
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1520 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1521 {
1522 /* for default tnl0 device allow to change only the proto */
1523 t->parms.proto = p->proto;
1524 netdev_state_change(t->dev);
1525 return 0;
1526 }
1527
1528 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1529 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1530 {
1531 p->laddr = u->laddr;
1532 p->raddr = u->raddr;
1533 p->flags = u->flags;
1534 p->hop_limit = u->hop_limit;
1535 p->encap_limit = u->encap_limit;
1536 p->flowinfo = u->flowinfo;
1537 p->link = u->link;
1538 p->proto = u->proto;
1539 memcpy(p->name, u->name, sizeof(u->name));
1540 }
1541
1542 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1543 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1544 {
1545 u->laddr = p->laddr;
1546 u->raddr = p->raddr;
1547 u->flags = p->flags;
1548 u->hop_limit = p->hop_limit;
1549 u->encap_limit = p->encap_limit;
1550 u->flowinfo = p->flowinfo;
1551 u->link = p->link;
1552 u->proto = p->proto;
1553 memcpy(u->name, p->name, sizeof(u->name));
1554 }
1555
1556 /**
1557 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1558 * @dev: virtual device associated with tunnel
1559 * @ifr: parameters passed from userspace
1560 * @cmd: command to be performed
1561 *
1562 * Description:
1563 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1564 * from userspace.
1565 *
1566 * The possible commands are the following:
1567 * %SIOCGETTUNNEL: get tunnel parameters for device
1568 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1569 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1570 * %SIOCDELTUNNEL: delete tunnel
1571 *
1572 * The fallback device "ip6tnl0", created during module
1573 * initialization, can be used for creating other tunnel devices.
1574 *
1575 * Return:
1576 * 0 on success,
1577 * %-EFAULT if unable to copy data to or from userspace,
1578 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1579 * %-EINVAL if passed tunnel parameters are invalid,
1580 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1581 * %-ENODEV if attempting to change or delete a nonexisting device
1582 **/
1583
1584 static int
ip6_tnl_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)1585 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1586 {
1587 int err = 0;
1588 struct ip6_tnl_parm p;
1589 struct __ip6_tnl_parm p1;
1590 struct ip6_tnl *t = netdev_priv(dev);
1591 struct net *net = t->net;
1592 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1593
1594 memset(&p1, 0, sizeof(p1));
1595
1596 switch (cmd) {
1597 case SIOCGETTUNNEL:
1598 if (dev == ip6n->fb_tnl_dev) {
1599 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1600 err = -EFAULT;
1601 break;
1602 }
1603 ip6_tnl_parm_from_user(&p1, &p);
1604 t = ip6_tnl_locate(net, &p1, 0);
1605 if (IS_ERR(t))
1606 t = netdev_priv(dev);
1607 } else {
1608 memset(&p, 0, sizeof(p));
1609 }
1610 ip6_tnl_parm_to_user(&p, &t->parms);
1611 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1612 err = -EFAULT;
1613 }
1614 break;
1615 case SIOCADDTUNNEL:
1616 case SIOCCHGTUNNEL:
1617 err = -EPERM;
1618 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1619 break;
1620 err = -EFAULT;
1621 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1622 break;
1623 err = -EINVAL;
1624 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1625 p.proto != 0)
1626 break;
1627 ip6_tnl_parm_from_user(&p1, &p);
1628 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1629 if (cmd == SIOCCHGTUNNEL) {
1630 if (!IS_ERR(t)) {
1631 if (t->dev != dev) {
1632 err = -EEXIST;
1633 break;
1634 }
1635 } else
1636 t = netdev_priv(dev);
1637 if (dev == ip6n->fb_tnl_dev)
1638 err = ip6_tnl0_update(t, &p1);
1639 else
1640 err = ip6_tnl_update(t, &p1);
1641 }
1642 if (!IS_ERR(t)) {
1643 err = 0;
1644 ip6_tnl_parm_to_user(&p, &t->parms);
1645 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1646 err = -EFAULT;
1647
1648 } else {
1649 err = PTR_ERR(t);
1650 }
1651 break;
1652 case SIOCDELTUNNEL:
1653 err = -EPERM;
1654 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1655 break;
1656
1657 if (dev == ip6n->fb_tnl_dev) {
1658 err = -EFAULT;
1659 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1660 break;
1661 err = -ENOENT;
1662 ip6_tnl_parm_from_user(&p1, &p);
1663 t = ip6_tnl_locate(net, &p1, 0);
1664 if (IS_ERR(t))
1665 break;
1666 err = -EPERM;
1667 if (t->dev == ip6n->fb_tnl_dev)
1668 break;
1669 dev = t->dev;
1670 }
1671 err = 0;
1672 unregister_netdevice(dev);
1673 break;
1674 default:
1675 err = -EINVAL;
1676 }
1677 return err;
1678 }
1679
1680 /**
1681 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1682 * @dev: virtual device associated with tunnel
1683 * @new_mtu: the new mtu
1684 *
1685 * Return:
1686 * 0 on success,
1687 * %-EINVAL if mtu too small
1688 **/
1689
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1690 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1691 {
1692 struct ip6_tnl *tnl = netdev_priv(dev);
1693
1694 if (tnl->parms.proto == IPPROTO_IPV6) {
1695 if (new_mtu < IPV6_MIN_MTU)
1696 return -EINVAL;
1697 } else {
1698 if (new_mtu < ETH_MIN_MTU)
1699 return -EINVAL;
1700 }
1701 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1702 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1703 return -EINVAL;
1704 } else {
1705 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1706 return -EINVAL;
1707 }
1708 dev->mtu = new_mtu;
1709 return 0;
1710 }
1711 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1712
ip6_tnl_get_iflink(const struct net_device * dev)1713 int ip6_tnl_get_iflink(const struct net_device *dev)
1714 {
1715 struct ip6_tnl *t = netdev_priv(dev);
1716
1717 return t->parms.link;
1718 }
1719 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1720
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1721 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1722 unsigned int num)
1723 {
1724 if (num >= MAX_IPTUN_ENCAP_OPS)
1725 return -ERANGE;
1726
1727 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1728 &ip6tun_encaps[num],
1729 NULL, ops) ? 0 : -1;
1730 }
1731 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1732
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1733 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1734 unsigned int num)
1735 {
1736 int ret;
1737
1738 if (num >= MAX_IPTUN_ENCAP_OPS)
1739 return -ERANGE;
1740
1741 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1742 &ip6tun_encaps[num],
1743 ops, NULL) == ops) ? 0 : -1;
1744
1745 synchronize_net();
1746
1747 return ret;
1748 }
1749 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1750
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1751 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1752 struct ip_tunnel_encap *ipencap)
1753 {
1754 int hlen;
1755
1756 memset(&t->encap, 0, sizeof(t->encap));
1757
1758 hlen = ip6_encap_hlen(ipencap);
1759 if (hlen < 0)
1760 return hlen;
1761
1762 t->encap.type = ipencap->type;
1763 t->encap.sport = ipencap->sport;
1764 t->encap.dport = ipencap->dport;
1765 t->encap.flags = ipencap->flags;
1766
1767 t->encap_hlen = hlen;
1768 t->hlen = t->encap_hlen + t->tun_hlen;
1769
1770 return 0;
1771 }
1772 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1773
1774 static const struct net_device_ops ip6_tnl_netdev_ops = {
1775 .ndo_init = ip6_tnl_dev_init,
1776 .ndo_uninit = ip6_tnl_dev_uninit,
1777 .ndo_start_xmit = ip6_tnl_start_xmit,
1778 .ndo_do_ioctl = ip6_tnl_ioctl,
1779 .ndo_change_mtu = ip6_tnl_change_mtu,
1780 .ndo_get_stats = ip6_get_stats,
1781 .ndo_get_iflink = ip6_tnl_get_iflink,
1782 };
1783
1784 #define IPXIPX_FEATURES (NETIF_F_SG | \
1785 NETIF_F_FRAGLIST | \
1786 NETIF_F_HIGHDMA | \
1787 NETIF_F_GSO_SOFTWARE | \
1788 NETIF_F_HW_CSUM)
1789
1790 /**
1791 * ip6_tnl_dev_setup - setup virtual tunnel device
1792 * @dev: virtual device associated with tunnel
1793 *
1794 * Description:
1795 * Initialize function pointers and device parameters
1796 **/
1797
ip6_tnl_dev_setup(struct net_device * dev)1798 static void ip6_tnl_dev_setup(struct net_device *dev)
1799 {
1800 dev->netdev_ops = &ip6_tnl_netdev_ops;
1801 dev->needs_free_netdev = true;
1802 dev->priv_destructor = ip6_dev_free;
1803
1804 dev->type = ARPHRD_TUNNEL6;
1805 dev->flags |= IFF_NOARP;
1806 dev->addr_len = sizeof(struct in6_addr);
1807 dev->features |= NETIF_F_LLTX;
1808 netif_keep_dst(dev);
1809
1810 dev->features |= IPXIPX_FEATURES;
1811 dev->hw_features |= IPXIPX_FEATURES;
1812
1813 /* This perm addr will be used as interface identifier by IPv6 */
1814 dev->addr_assign_type = NET_ADDR_RANDOM;
1815 eth_random_addr(dev->perm_addr);
1816 }
1817
1818
1819 /**
1820 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1821 * @dev: virtual device associated with tunnel
1822 **/
1823
1824 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1825 ip6_tnl_dev_init_gen(struct net_device *dev)
1826 {
1827 struct ip6_tnl *t = netdev_priv(dev);
1828 int ret;
1829 int t_hlen;
1830
1831 t->dev = dev;
1832 t->net = dev_net(dev);
1833 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1834 if (!dev->tstats)
1835 return -ENOMEM;
1836
1837 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1838 if (ret)
1839 goto free_stats;
1840
1841 ret = gro_cells_init(&t->gro_cells, dev);
1842 if (ret)
1843 goto destroy_dst;
1844
1845 t->tun_hlen = 0;
1846 t->hlen = t->encap_hlen + t->tun_hlen;
1847 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1848
1849 dev->type = ARPHRD_TUNNEL6;
1850 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1851 dev->mtu = ETH_DATA_LEN - t_hlen;
1852 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1853 dev->mtu -= 8;
1854 dev->min_mtu = ETH_MIN_MTU;
1855 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1856
1857 return 0;
1858
1859 destroy_dst:
1860 dst_cache_destroy(&t->dst_cache);
1861 free_stats:
1862 free_percpu(dev->tstats);
1863 dev->tstats = NULL;
1864
1865 return ret;
1866 }
1867
1868 /**
1869 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1870 * @dev: virtual device associated with tunnel
1871 **/
1872
ip6_tnl_dev_init(struct net_device * dev)1873 static int ip6_tnl_dev_init(struct net_device *dev)
1874 {
1875 struct ip6_tnl *t = netdev_priv(dev);
1876 int err = ip6_tnl_dev_init_gen(dev);
1877
1878 if (err)
1879 return err;
1880 ip6_tnl_link_config(t);
1881 if (t->parms.collect_md)
1882 netif_keep_dst(dev);
1883 return 0;
1884 }
1885
1886 /**
1887 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1888 * @dev: fallback device
1889 *
1890 * Return: 0
1891 **/
1892
ip6_fb_tnl_dev_init(struct net_device * dev)1893 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1894 {
1895 struct ip6_tnl *t = netdev_priv(dev);
1896 struct net *net = dev_net(dev);
1897 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1898
1899 t->parms.proto = IPPROTO_IPV6;
1900 dev_hold(dev);
1901
1902 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1903 return 0;
1904 }
1905
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1906 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1907 struct netlink_ext_ack *extack)
1908 {
1909 u8 proto;
1910
1911 if (!data || !data[IFLA_IPTUN_PROTO])
1912 return 0;
1913
1914 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1915 if (proto != IPPROTO_IPV6 &&
1916 proto != IPPROTO_IPIP &&
1917 proto != 0)
1918 return -EINVAL;
1919
1920 return 0;
1921 }
1922
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1923 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1924 struct __ip6_tnl_parm *parms)
1925 {
1926 memset(parms, 0, sizeof(*parms));
1927
1928 if (!data)
1929 return;
1930
1931 if (data[IFLA_IPTUN_LINK])
1932 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1933
1934 if (data[IFLA_IPTUN_LOCAL])
1935 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1936
1937 if (data[IFLA_IPTUN_REMOTE])
1938 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1939
1940 if (data[IFLA_IPTUN_TTL])
1941 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1942
1943 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1944 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1945
1946 if (data[IFLA_IPTUN_FLOWINFO])
1947 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1948
1949 if (data[IFLA_IPTUN_FLAGS])
1950 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1951
1952 if (data[IFLA_IPTUN_PROTO])
1953 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1954
1955 if (data[IFLA_IPTUN_COLLECT_METADATA])
1956 parms->collect_md = true;
1957
1958 if (data[IFLA_IPTUN_FWMARK])
1959 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1960 }
1961
ip6_tnl_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)1962 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1963 struct ip_tunnel_encap *ipencap)
1964 {
1965 bool ret = false;
1966
1967 memset(ipencap, 0, sizeof(*ipencap));
1968
1969 if (!data)
1970 return ret;
1971
1972 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1973 ret = true;
1974 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1975 }
1976
1977 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1978 ret = true;
1979 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1980 }
1981
1982 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1983 ret = true;
1984 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1985 }
1986
1987 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1988 ret = true;
1989 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1990 }
1991
1992 return ret;
1993 }
1994
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1995 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1996 struct nlattr *tb[], struct nlattr *data[],
1997 struct netlink_ext_ack *extack)
1998 {
1999 struct net *net = dev_net(dev);
2000 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2001 struct ip_tunnel_encap ipencap;
2002 struct ip6_tnl *nt, *t;
2003 int err;
2004
2005 nt = netdev_priv(dev);
2006
2007 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2008 err = ip6_tnl_encap_setup(nt, &ipencap);
2009 if (err < 0)
2010 return err;
2011 }
2012
2013 ip6_tnl_netlink_parms(data, &nt->parms);
2014
2015 if (nt->parms.collect_md) {
2016 if (rtnl_dereference(ip6n->collect_md_tun))
2017 return -EEXIST;
2018 } else {
2019 t = ip6_tnl_locate(net, &nt->parms, 0);
2020 if (!IS_ERR(t))
2021 return -EEXIST;
2022 }
2023
2024 err = ip6_tnl_create2(dev);
2025 if (!err && tb[IFLA_MTU])
2026 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2027
2028 return err;
2029 }
2030
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2031 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2032 struct nlattr *data[],
2033 struct netlink_ext_ack *extack)
2034 {
2035 struct ip6_tnl *t = netdev_priv(dev);
2036 struct __ip6_tnl_parm p;
2037 struct net *net = t->net;
2038 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2039 struct ip_tunnel_encap ipencap;
2040
2041 if (dev == ip6n->fb_tnl_dev)
2042 return -EINVAL;
2043
2044 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2045 int err = ip6_tnl_encap_setup(t, &ipencap);
2046
2047 if (err < 0)
2048 return err;
2049 }
2050 ip6_tnl_netlink_parms(data, &p);
2051 if (p.collect_md)
2052 return -EINVAL;
2053
2054 t = ip6_tnl_locate(net, &p, 0);
2055 if (!IS_ERR(t)) {
2056 if (t->dev != dev)
2057 return -EEXIST;
2058 } else
2059 t = netdev_priv(dev);
2060
2061 return ip6_tnl_update(t, &p);
2062 }
2063
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2064 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2065 {
2066 struct net *net = dev_net(dev);
2067 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2068
2069 if (dev != ip6n->fb_tnl_dev)
2070 unregister_netdevice_queue(dev, head);
2071 }
2072
ip6_tnl_get_size(const struct net_device * dev)2073 static size_t ip6_tnl_get_size(const struct net_device *dev)
2074 {
2075 return
2076 /* IFLA_IPTUN_LINK */
2077 nla_total_size(4) +
2078 /* IFLA_IPTUN_LOCAL */
2079 nla_total_size(sizeof(struct in6_addr)) +
2080 /* IFLA_IPTUN_REMOTE */
2081 nla_total_size(sizeof(struct in6_addr)) +
2082 /* IFLA_IPTUN_TTL */
2083 nla_total_size(1) +
2084 /* IFLA_IPTUN_ENCAP_LIMIT */
2085 nla_total_size(1) +
2086 /* IFLA_IPTUN_FLOWINFO */
2087 nla_total_size(4) +
2088 /* IFLA_IPTUN_FLAGS */
2089 nla_total_size(4) +
2090 /* IFLA_IPTUN_PROTO */
2091 nla_total_size(1) +
2092 /* IFLA_IPTUN_ENCAP_TYPE */
2093 nla_total_size(2) +
2094 /* IFLA_IPTUN_ENCAP_FLAGS */
2095 nla_total_size(2) +
2096 /* IFLA_IPTUN_ENCAP_SPORT */
2097 nla_total_size(2) +
2098 /* IFLA_IPTUN_ENCAP_DPORT */
2099 nla_total_size(2) +
2100 /* IFLA_IPTUN_COLLECT_METADATA */
2101 nla_total_size(0) +
2102 /* IFLA_IPTUN_FWMARK */
2103 nla_total_size(4) +
2104 0;
2105 }
2106
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2107 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2108 {
2109 struct ip6_tnl *tunnel = netdev_priv(dev);
2110 struct __ip6_tnl_parm *parm = &tunnel->parms;
2111
2112 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2113 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2114 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2115 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2116 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2117 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2118 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2119 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2120 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2121 goto nla_put_failure;
2122
2123 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2124 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2125 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2126 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2127 goto nla_put_failure;
2128
2129 if (parm->collect_md)
2130 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2131 goto nla_put_failure;
2132
2133 return 0;
2134
2135 nla_put_failure:
2136 return -EMSGSIZE;
2137 }
2138
ip6_tnl_get_link_net(const struct net_device * dev)2139 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2140 {
2141 struct ip6_tnl *tunnel = netdev_priv(dev);
2142
2143 return tunnel->net;
2144 }
2145 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2146
2147 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2148 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2149 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2150 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2151 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2152 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2153 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2154 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2155 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2156 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2157 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2158 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2159 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2160 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2161 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
2162 };
2163
2164 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2165 .kind = "ip6tnl",
2166 .maxtype = IFLA_IPTUN_MAX,
2167 .policy = ip6_tnl_policy,
2168 .priv_size = sizeof(struct ip6_tnl),
2169 .setup = ip6_tnl_dev_setup,
2170 .validate = ip6_tnl_validate,
2171 .newlink = ip6_tnl_newlink,
2172 .changelink = ip6_tnl_changelink,
2173 .dellink = ip6_tnl_dellink,
2174 .get_size = ip6_tnl_get_size,
2175 .fill_info = ip6_tnl_fill_info,
2176 .get_link_net = ip6_tnl_get_link_net,
2177 };
2178
2179 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2180 .handler = ip4ip6_rcv,
2181 .err_handler = ip4ip6_err,
2182 .priority = 1,
2183 };
2184
2185 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2186 .handler = ip6ip6_rcv,
2187 .err_handler = ip6ip6_err,
2188 .priority = 1,
2189 };
2190
ip6_tnl_destroy_tunnels(struct net * net)2191 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2192 {
2193 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2194 struct net_device *dev, *aux;
2195 int h;
2196 struct ip6_tnl *t;
2197 LIST_HEAD(list);
2198
2199 for_each_netdev_safe(net, dev, aux)
2200 if (dev->rtnl_link_ops == &ip6_link_ops)
2201 unregister_netdevice_queue(dev, &list);
2202
2203 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2204 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2205 while (t) {
2206 /* If dev is in the same netns, it has already
2207 * been added to the list by the previous loop.
2208 */
2209 if (!net_eq(dev_net(t->dev), net))
2210 unregister_netdevice_queue(t->dev, &list);
2211 t = rtnl_dereference(t->next);
2212 }
2213 }
2214
2215 unregister_netdevice_many(&list);
2216 }
2217
ip6_tnl_init_net(struct net * net)2218 static int __net_init ip6_tnl_init_net(struct net *net)
2219 {
2220 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2221 struct ip6_tnl *t = NULL;
2222 int err;
2223
2224 ip6n->tnls[0] = ip6n->tnls_wc;
2225 ip6n->tnls[1] = ip6n->tnls_r_l;
2226
2227 err = -ENOMEM;
2228 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2229 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2230
2231 if (!ip6n->fb_tnl_dev)
2232 goto err_alloc_dev;
2233 dev_net_set(ip6n->fb_tnl_dev, net);
2234 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2235 /* FB netdevice is special: we have one, and only one per netns.
2236 * Allowing to move it to another netns is clearly unsafe.
2237 */
2238 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2239
2240 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2241 if (err < 0)
2242 goto err_register;
2243
2244 err = register_netdev(ip6n->fb_tnl_dev);
2245 if (err < 0)
2246 goto err_register;
2247
2248 t = netdev_priv(ip6n->fb_tnl_dev);
2249
2250 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2251 return 0;
2252
2253 err_register:
2254 free_netdev(ip6n->fb_tnl_dev);
2255 err_alloc_dev:
2256 return err;
2257 }
2258
ip6_tnl_exit_net(struct net * net)2259 static void __net_exit ip6_tnl_exit_net(struct net *net)
2260 {
2261 rtnl_lock();
2262 ip6_tnl_destroy_tunnels(net);
2263 rtnl_unlock();
2264 }
2265
2266 static struct pernet_operations ip6_tnl_net_ops = {
2267 .init = ip6_tnl_init_net,
2268 .exit = ip6_tnl_exit_net,
2269 .id = &ip6_tnl_net_id,
2270 .size = sizeof(struct ip6_tnl_net),
2271 };
2272
2273 /**
2274 * ip6_tunnel_init - register protocol and reserve needed resources
2275 *
2276 * Return: 0 on success
2277 **/
2278
ip6_tunnel_init(void)2279 static int __init ip6_tunnel_init(void)
2280 {
2281 int err;
2282
2283 if (!ipv6_mod_enabled())
2284 return -EOPNOTSUPP;
2285
2286 err = register_pernet_device(&ip6_tnl_net_ops);
2287 if (err < 0)
2288 goto out_pernet;
2289
2290 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2291 if (err < 0) {
2292 pr_err("%s: can't register ip4ip6\n", __func__);
2293 goto out_ip4ip6;
2294 }
2295
2296 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2297 if (err < 0) {
2298 pr_err("%s: can't register ip6ip6\n", __func__);
2299 goto out_ip6ip6;
2300 }
2301 err = rtnl_link_register(&ip6_link_ops);
2302 if (err < 0)
2303 goto rtnl_link_failed;
2304
2305 return 0;
2306
2307 rtnl_link_failed:
2308 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2309 out_ip6ip6:
2310 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2311 out_ip4ip6:
2312 unregister_pernet_device(&ip6_tnl_net_ops);
2313 out_pernet:
2314 return err;
2315 }
2316
2317 /**
2318 * ip6_tunnel_cleanup - free resources and unregister protocol
2319 **/
2320
ip6_tunnel_cleanup(void)2321 static void __exit ip6_tunnel_cleanup(void)
2322 {
2323 rtnl_link_unregister(&ip6_link_ops);
2324 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2325 pr_info("%s: can't deregister ip4ip6\n", __func__);
2326
2327 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2328 pr_info("%s: can't deregister ip6ip6\n", __func__);
2329
2330 unregister_pernet_device(&ip6_tnl_net_ops);
2331 }
2332
2333 module_init(ip6_tunnel_init);
2334 module_exit(ip6_tunnel_cleanup);
2335