projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ip6_tunnel: disable caching when the traffic class is inherited
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / ip6_tunnel.c
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/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
42 #include <linux/slab.h>
43 #include <linux/hash.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
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
66
67 #ifdef IP6_TNL_DEBUG
68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
69 #else
70 #define IP6_TNL_TRACE(x...) do {;} while(0)
71 #endif
72
73 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
74 #define IPV6_TCLASS_SHIFT 20
75
76 #define HASH_SIZE_SHIFT 5
77 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
78
79 static bool log_ecn_error = true;
80 module_param(log_ecn_error, bool, 0644);
81 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
82
83 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
84 {
85 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
86
87 return hash_32(hash, HASH_SIZE_SHIFT);
88 }
89
90 static int ip6_tnl_dev_init(struct net_device *dev);
91 static void ip6_tnl_dev_setup(struct net_device *dev);
92 static struct rtnl_link_ops ip6_link_ops __read_mostly;
93
94 static int ip6_tnl_net_id __read_mostly;
95 struct ip6_tnl_net {
96 /* the IPv6 tunnel fallback device */
97 struct net_device *fb_tnl_dev;
98 /* lists for storing tunnels in use */
99 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
100 struct ip6_tnl __rcu *tnls_wc[1];
101 struct ip6_tnl __rcu **tnls[2];
102 };
103
104 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
105 {
106 struct pcpu_tstats sum = { 0 };
107 int i;
108
109 for_each_possible_cpu(i) {
110 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
111
112 sum.rx_packets += tstats->rx_packets;
113 sum.rx_bytes += tstats->rx_bytes;
114 sum.tx_packets += tstats->tx_packets;
115 sum.tx_bytes += tstats->tx_bytes;
116 }
117 dev->stats.rx_packets = sum.rx_packets;
118 dev->stats.rx_bytes = sum.rx_bytes;
119 dev->stats.tx_packets = sum.tx_packets;
120 dev->stats.tx_bytes = sum.tx_bytes;
121 return &dev->stats;
122 }
123
124 /*
125 * Locking : hash tables are protected by RCU and RTNL
126 */
127
128 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
129 {
130 struct dst_entry *dst = t->dst_cache;
131
132 if (dst && dst->obsolete &&
133 dst->ops->check(dst, t->dst_cookie) == NULL) {
134 t->dst_cache = NULL;
135 dst_release(dst);
136 return NULL;
137 }
138
139 return dst;
140 }
141 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
142
143 void ip6_tnl_dst_reset(struct ip6_tnl *t)
144 {
145 dst_release(t->dst_cache);
146 t->dst_cache = NULL;
147 }
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
149
150 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
151 {
152 struct rt6_info *rt = (struct rt6_info *) dst;
153 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
154 dst_release(t->dst_cache);
155 t->dst_cache = dst;
156 }
157 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
158
159 /**
160 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
161 * @remote: the address of the tunnel exit-point
162 * @local: the address of the tunnel entry-point
163 *
164 * Return:
165 * tunnel matching given end-points if found,
166 * else fallback tunnel if its device is up,
167 * else %NULL
168 **/
169
170 #define for_each_ip6_tunnel_rcu(start) \
171 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
172
173 static struct ip6_tnl *
174 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
175 {
176 unsigned int hash = HASH(remote, local);
177 struct ip6_tnl *t;
178 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
179
180 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
181 if (ipv6_addr_equal(local, &t->parms.laddr) &&
182 ipv6_addr_equal(remote, &t->parms.raddr) &&
183 (t->dev->flags & IFF_UP))
184 return t;
185 }
186 t = rcu_dereference(ip6n->tnls_wc[0]);
187 if (t && (t->dev->flags & IFF_UP))
188 return t;
189
190 return NULL;
191 }
192
193 /**
194 * ip6_tnl_bucket - get head of list matching given tunnel parameters
195 * @p: parameters containing tunnel end-points
196 *
197 * Description:
198 * ip6_tnl_bucket() returns the head of the list matching the
199 * &struct in6_addr entries laddr and raddr in @p.
200 *
201 * Return: head of IPv6 tunnel list
202 **/
203
204 static struct ip6_tnl __rcu **
205 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
206 {
207 const struct in6_addr *remote = &p->raddr;
208 const struct in6_addr *local = &p->laddr;
209 unsigned int h = 0;
210 int prio = 0;
211
212 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
213 prio = 1;
214 h = HASH(remote, local);
215 }
216 return &ip6n->tnls[prio][h];
217 }
218
219 /**
220 * ip6_tnl_link - add tunnel to hash table
221 * @t: tunnel to be added
222 **/
223
224 static void
225 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
226 {
227 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
228
229 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
230 rcu_assign_pointer(*tp, t);
231 }
232
233 /**
234 * ip6_tnl_unlink - remove tunnel from hash table
235 * @t: tunnel to be removed
236 **/
237
238 static void
239 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
240 {
241 struct ip6_tnl __rcu **tp;
242 struct ip6_tnl *iter;
243
244 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
245 (iter = rtnl_dereference(*tp)) != NULL;
246 tp = &iter->next) {
247 if (t == iter) {
248 rcu_assign_pointer(*tp, t->next);
249 break;
250 }
251 }
252 }
253
254 static void ip6_dev_free(struct net_device *dev)
255 {
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 NULL
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;
302
303 if (p->name[0])
304 strlcpy(name, p->name, IFNAMSIZ);
305 else
306 sprintf(name, "ip6tnl%%d");
307
308 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
309 if (dev == NULL)
310 goto failed;
311
312 dev_net_set(dev, net);
313
314 t = netdev_priv(dev);
315 t->parms = *p;
316 err = ip6_tnl_create2(dev);
317 if (err < 0)
318 goto failed_free;
319
320 return t;
321
322 failed_free:
323 ip6_dev_free(dev);
324 failed:
325 return NULL;
326 }
327
328 /**
329 * ip6_tnl_locate - find or create tunnel matching given parameters
330 * @p: tunnel parameters
331 * @create: != 0 if allowed to create new tunnel if no match found
332 *
333 * Description:
334 * ip6_tnl_locate() first tries to locate an existing tunnel
335 * based on @parms. If this is unsuccessful, but @create is set a new
336 * tunnel device is created and registered for use.
337 *
338 * Return:
339 * matching tunnel or NULL
340 **/
341
342 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
343 struct __ip6_tnl_parm *p, int create)
344 {
345 const struct in6_addr *remote = &p->raddr;
346 const struct in6_addr *local = &p->laddr;
347 struct ip6_tnl __rcu **tp;
348 struct ip6_tnl *t;
349 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
350
351 for (tp = ip6_tnl_bucket(ip6n, p);
352 (t = rtnl_dereference(*tp)) != NULL;
353 tp = &t->next) {
354 if (ipv6_addr_equal(local, &t->parms.laddr) &&
355 ipv6_addr_equal(remote, &t->parms.raddr))
356 return t;
357 }
358 if (!create)
359 return NULL;
360 return ip6_tnl_create(net, p);
361 }
362
363 /**
364 * ip6_tnl_dev_uninit - tunnel device uninitializer
365 * @dev: the device to be destroyed
366 *
367 * Description:
368 * ip6_tnl_dev_uninit() removes tunnel from its list
369 **/
370
371 static void
372 ip6_tnl_dev_uninit(struct net_device *dev)
373 {
374 struct ip6_tnl *t = netdev_priv(dev);
375 struct net *net = dev_net(dev);
376 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
377
378 if (dev == ip6n->fb_tnl_dev)
379 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
380 else
381 ip6_tnl_unlink(ip6n, t);
382 ip6_tnl_dst_reset(t);
383 dev_put(dev);
384 }
385
386 /**
387 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
388 * @skb: received socket buffer
389 *
390 * Return:
391 * 0 if none was found,
392 * else index to encapsulation limit
393 **/
394
395 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
396 {
397 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
398 __u8 nexthdr = ipv6h->nexthdr;
399 __u16 off = sizeof (*ipv6h);
400
401 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
402 __u16 optlen = 0;
403 struct ipv6_opt_hdr *hdr;
404 if (raw + off + sizeof (*hdr) > skb->data &&
405 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
406 break;
407
408 hdr = (struct ipv6_opt_hdr *) (raw + off);
409 if (nexthdr == NEXTHDR_FRAGMENT) {
410 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
411 if (frag_hdr->frag_off)
412 break;
413 optlen = 8;
414 } else if (nexthdr == NEXTHDR_AUTH) {
415 optlen = (hdr->hdrlen + 2) << 2;
416 } else {
417 optlen = ipv6_optlen(hdr);
418 }
419 if (nexthdr == NEXTHDR_DEST) {
420 __u16 i = off + 2;
421 while (1) {
422 struct ipv6_tlv_tnl_enc_lim *tel;
423
424 /* No more room for encapsulation limit */
425 if (i + sizeof (*tel) > off + optlen)
426 break;
427
428 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
429 /* return index of option if found and valid */
430 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
431 tel->length == 1)
432 return i;
433 /* else jump to next option */
434 if (tel->type)
435 i += tel->length + 2;
436 else
437 i++;
438 }
439 }
440 nexthdr = hdr->nexthdr;
441 off += optlen;
442 }
443 return 0;
444 }
445 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
446
447 /**
448 * ip6_tnl_err - tunnel error handler
449 *
450 * Description:
451 * ip6_tnl_err() should handle errors in the tunnel according
452 * to the specifications in RFC 2473.
453 **/
454
455 static int
456 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
457 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
458 {
459 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
460 struct ip6_tnl *t;
461 int rel_msg = 0;
462 u8 rel_type = ICMPV6_DEST_UNREACH;
463 u8 rel_code = ICMPV6_ADDR_UNREACH;
464 __u32 rel_info = 0;
465 __u16 len;
466 int err = -ENOENT;
467
468 /* If the packet doesn't contain the original IPv6 header we are
469 in trouble since we might need the source address for further
470 processing of the error. */
471
472 rcu_read_lock();
473 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
474 &ipv6h->saddr)) == NULL)
475 goto out;
476
477 if (t->parms.proto != ipproto && t->parms.proto != 0)
478 goto out;
479
480 err = 0;
481
482 switch (*type) {
483 __u32 teli;
484 struct ipv6_tlv_tnl_enc_lim *tel;
485 __u32 mtu;
486 case ICMPV6_DEST_UNREACH:
487 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
488 t->parms.name);
489 rel_msg = 1;
490 break;
491 case ICMPV6_TIME_EXCEED:
492 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
493 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
494 t->parms.name);
495 rel_msg = 1;
496 }
497 break;
498 case ICMPV6_PARAMPROB:
499 teli = 0;
500 if ((*code) == ICMPV6_HDR_FIELD)
501 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
502
503 if (teli && teli == *info - 2) {
504 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
505 if (tel->encap_limit == 0) {
506 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
507 t->parms.name);
508 rel_msg = 1;
509 }
510 } else {
511 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
512 t->parms.name);
513 }
514 break;
515 case ICMPV6_PKT_TOOBIG:
516 mtu = *info - offset;
517 if (mtu < IPV6_MIN_MTU)
518 mtu = IPV6_MIN_MTU;
519 t->dev->mtu = mtu;
520
521 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
522 rel_type = ICMPV6_PKT_TOOBIG;
523 rel_code = 0;
524 rel_info = mtu;
525 rel_msg = 1;
526 }
527 break;
528 }
529
530 *type = rel_type;
531 *code = rel_code;
532 *info = rel_info;
533 *msg = rel_msg;
534
535 out:
536 rcu_read_unlock();
537 return err;
538 }
539
540 static int
541 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
542 u8 type, u8 code, int offset, __be32 info)
543 {
544 int rel_msg = 0;
545 u8 rel_type = type;
546 u8 rel_code = code;
547 __u32 rel_info = ntohl(info);
548 int err;
549 struct sk_buff *skb2;
550 const struct iphdr *eiph;
551 struct rtable *rt;
552 struct flowi4 fl4;
553
554 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
555 &rel_msg, &rel_info, offset);
556 if (err < 0)
557 return err;
558
559 if (rel_msg == 0)
560 return 0;
561
562 switch (rel_type) {
563 case ICMPV6_DEST_UNREACH:
564 if (rel_code != ICMPV6_ADDR_UNREACH)
565 return 0;
566 rel_type = ICMP_DEST_UNREACH;
567 rel_code = ICMP_HOST_UNREACH;
568 break;
569 case ICMPV6_PKT_TOOBIG:
570 if (rel_code != 0)
571 return 0;
572 rel_type = ICMP_DEST_UNREACH;
573 rel_code = ICMP_FRAG_NEEDED;
574 break;
575 case NDISC_REDIRECT:
576 rel_type = ICMP_REDIRECT;
577 rel_code = ICMP_REDIR_HOST;
578 default:
579 return 0;
580 }
581
582 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
583 return 0;
584
585 skb2 = skb_clone(skb, GFP_ATOMIC);
586 if (!skb2)
587 return 0;
588
589 skb_dst_drop(skb2);
590
591 skb_pull(skb2, offset);
592 skb_reset_network_header(skb2);
593 eiph = ip_hdr(skb2);
594
595 /* Try to guess incoming interface */
596 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
597 eiph->saddr, 0,
598 0, 0,
599 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
600 if (IS_ERR(rt))
601 goto out;
602
603 skb2->dev = rt->dst.dev;
604
605 /* route "incoming" packet */
606 if (rt->rt_flags & RTCF_LOCAL) {
607 ip_rt_put(rt);
608 rt = NULL;
609 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
610 eiph->daddr, eiph->saddr,
611 0, 0,
612 IPPROTO_IPIP,
613 RT_TOS(eiph->tos), 0);
614 if (IS_ERR(rt) ||
615 rt->dst.dev->type != ARPHRD_TUNNEL) {
616 if (!IS_ERR(rt))
617 ip_rt_put(rt);
618 goto out;
619 }
620 skb_dst_set(skb2, &rt->dst);
621 } else {
622 ip_rt_put(rt);
623 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
624 skb2->dev) ||
625 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
626 goto out;
627 }
628
629 /* change mtu on this route */
630 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
631 if (rel_info > dst_mtu(skb_dst(skb2)))
632 goto out;
633
634 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
635 }
636 if (rel_type == ICMP_REDIRECT)
637 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
638
639 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
640
641 out:
642 kfree_skb(skb2);
643 return 0;
644 }
645
646 static int
647 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
648 u8 type, u8 code, int offset, __be32 info)
649 {
650 int rel_msg = 0;
651 u8 rel_type = type;
652 u8 rel_code = code;
653 __u32 rel_info = ntohl(info);
654 int err;
655
656 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
657 &rel_msg, &rel_info, offset);
658 if (err < 0)
659 return err;
660
661 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
662 struct rt6_info *rt;
663 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
664
665 if (!skb2)
666 return 0;
667
668 skb_dst_drop(skb2);
669 skb_pull(skb2, offset);
670 skb_reset_network_header(skb2);
671
672 /* Try to guess incoming interface */
673 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
674 NULL, 0, 0);
675
676 if (rt && rt->dst.dev)
677 skb2->dev = rt->dst.dev;
678
679 icmpv6_send(skb2, rel_type, rel_code, rel_info);
680
681 ip6_rt_put(rt);
682
683 kfree_skb(skb2);
684 }
685
686 return 0;
687 }
688
689 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
690 const struct ipv6hdr *ipv6h,
691 struct sk_buff *skb)
692 {
693 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
694
695 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
696 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
697
698 return IP6_ECN_decapsulate(ipv6h, skb);
699 }
700
701 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
702 const struct ipv6hdr *ipv6h,
703 struct sk_buff *skb)
704 {
705 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
706 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
707
708 return IP6_ECN_decapsulate(ipv6h, skb);
709 }
710
711 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
712 const struct in6_addr *laddr,
713 const struct in6_addr *raddr)
714 {
715 struct __ip6_tnl_parm *p = &t->parms;
716 int ltype = ipv6_addr_type(laddr);
717 int rtype = ipv6_addr_type(raddr);
718 __u32 flags = 0;
719
720 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
721 flags = IP6_TNL_F_CAP_PER_PACKET;
722 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
723 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
724 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
725 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
726 if (ltype&IPV6_ADDR_UNICAST)
727 flags |= IP6_TNL_F_CAP_XMIT;
728 if (rtype&IPV6_ADDR_UNICAST)
729 flags |= IP6_TNL_F_CAP_RCV;
730 }
731 return flags;
732 }
733 EXPORT_SYMBOL(ip6_tnl_get_cap);
734
735 /* called with rcu_read_lock() */
736 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
737 const struct in6_addr *laddr,
738 const struct in6_addr *raddr)
739 {
740 struct __ip6_tnl_parm *p = &t->parms;
741 int ret = 0;
742 struct net *net = dev_net(t->dev);
743
744 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
745 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
746 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
747 struct net_device *ldev = NULL;
748
749 if (p->link)
750 ldev = dev_get_by_index_rcu(net, p->link);
751
752 if ((ipv6_addr_is_multicast(laddr) ||
753 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
754 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
755 ret = 1;
756 }
757 return ret;
758 }
759 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
760
761 /**
762 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
763 * @skb: received socket buffer
764 * @protocol: ethernet protocol ID
765 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
766 *
767 * Return: 0
768 **/
769
770 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
771 __u8 ipproto,
772 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
773 const struct ipv6hdr *ipv6h,
774 struct sk_buff *skb))
775 {
776 struct ip6_tnl *t;
777 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
778 int err;
779
780 rcu_read_lock();
781
782 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
783 &ipv6h->daddr)) != NULL) {
784 struct pcpu_tstats *tstats;
785
786 if (t->parms.proto != ipproto && t->parms.proto != 0) {
787 rcu_read_unlock();
788 goto discard;
789 }
790
791 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
792 rcu_read_unlock();
793 goto discard;
794 }
795
796 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
797 t->dev->stats.rx_dropped++;
798 rcu_read_unlock();
799 goto discard;
800 }
801 secpath_reset(skb);
802 skb->mac_header = skb->network_header;
803 skb_reset_network_header(skb);
804 skb->protocol = htons(protocol);
805 skb->pkt_type = PACKET_HOST;
806 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
807
808 __skb_tunnel_rx(skb, t->dev);
809
810 err = dscp_ecn_decapsulate(t, ipv6h, skb);
811 if (unlikely(err)) {
812 if (log_ecn_error)
813 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
814 &ipv6h->saddr,
815 ipv6_get_dsfield(ipv6h));
816 if (err > 1) {
817 ++t->dev->stats.rx_frame_errors;
818 ++t->dev->stats.rx_errors;
819 rcu_read_unlock();
820 goto discard;
821 }
822 }
823
824 tstats = this_cpu_ptr(t->dev->tstats);
825 tstats->rx_packets++;
826 tstats->rx_bytes += skb->len;
827
828 netif_rx(skb);
829
830 rcu_read_unlock();
831 return 0;
832 }
833 rcu_read_unlock();
834 return 1;
835
836 discard:
837 kfree_skb(skb);
838 return 0;
839 }
840
841 static int ip4ip6_rcv(struct sk_buff *skb)
842 {
843 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
844 ip4ip6_dscp_ecn_decapsulate);
845 }
846
847 static int ip6ip6_rcv(struct sk_buff *skb)
848 {
849 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
850 ip6ip6_dscp_ecn_decapsulate);
851 }
852
853 struct ipv6_tel_txoption {
854 struct ipv6_txoptions ops;
855 __u8 dst_opt[8];
856 };
857
858 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
859 {
860 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
861
862 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
863 opt->dst_opt[3] = 1;
864 opt->dst_opt[4] = encap_limit;
865 opt->dst_opt[5] = IPV6_TLV_PADN;
866 opt->dst_opt[6] = 1;
867
868 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
869 opt->ops.opt_nflen = 8;
870 }
871
872 /**
873 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
874 * @t: the outgoing tunnel device
875 * @hdr: IPv6 header from the incoming packet
876 *
877 * Description:
878 * Avoid trivial tunneling loop by checking that tunnel exit-point
879 * doesn't match source of incoming packet.
880 *
881 * Return:
882 * 1 if conflict,
883 * 0 else
884 **/
885
886 static inline bool
887 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
888 {
889 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
890 }
891
892 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
893 {
894 struct __ip6_tnl_parm *p = &t->parms;
895 int ret = 0;
896 struct net *net = dev_net(t->dev);
897
898 if (p->flags & IP6_TNL_F_CAP_XMIT) {
899 struct net_device *ldev = NULL;
900
901 rcu_read_lock();
902 if (p->link)
903 ldev = dev_get_by_index_rcu(net, p->link);
904
905 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
906 pr_warn("%s xmit: Local address not yet configured!\n",
907 p->name);
908 else if (!ipv6_addr_is_multicast(&p->raddr) &&
909 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
910 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
911 p->name);
912 else
913 ret = 1;
914 rcu_read_unlock();
915 }
916 return ret;
917 }
918 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
919
920 /**
921 * ip6_tnl_xmit2 - encapsulate packet and send
922 * @skb: the outgoing socket buffer
923 * @dev: the outgoing tunnel device
924 * @dsfield: dscp code for outer header
925 * @fl: flow of tunneled packet
926 * @encap_limit: encapsulation limit
927 * @pmtu: Path MTU is stored if packet is too big
928 *
929 * Description:
930 * Build new header and do some sanity checks on the packet before sending
931 * it.
932 *
933 * Return:
934 * 0 on success
935 * -1 fail
936 * %-EMSGSIZE message too big. return mtu in this case.
937 **/
938
939 static int ip6_tnl_xmit2(struct sk_buff *skb,
940 struct net_device *dev,
941 __u8 dsfield,
942 struct flowi6 *fl6,
943 int encap_limit,
944 __u32 *pmtu)
945 {
946 struct net *net = dev_net(dev);
947 struct ip6_tnl *t = netdev_priv(dev);
948 struct net_device_stats *stats = &t->dev->stats;
949 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
950 struct ipv6_tel_txoption opt;
951 struct dst_entry *dst = NULL, *ndst = NULL;
952 struct net_device *tdev;
953 bool use_cache = false;
954 int mtu;
955 unsigned int max_headroom = sizeof(struct ipv6hdr);
956 u8 proto;
957 int err = -1;
958
959 if (!(t->parms.flags &
960 (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
961 /* enable the cache only only if the routing decision does
962 * not depend on the current inner header value
963 */
964 use_cache = true;
965 }
966
967 if (use_cache)
968 dst = ip6_tnl_dst_check(t);
969 if (!dst) {
970 ndst = ip6_route_output(net, NULL, fl6);
971
972 if (ndst->error)
973 goto tx_err_link_failure;
974 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
975 if (IS_ERR(ndst)) {
976 err = PTR_ERR(ndst);
977 ndst = NULL;
978 goto tx_err_link_failure;
979 }
980 dst = ndst;
981 }
982
983 tdev = dst->dev;
984
985 if (tdev == dev) {
986 stats->collisions++;
987 net_warn_ratelimited("%s: Local routing loop detected!\n",
988 t->parms.name);
989 goto tx_err_dst_release;
990 }
991 mtu = dst_mtu(dst) - sizeof (*ipv6h);
992 if (encap_limit >= 0) {
993 max_headroom += 8;
994 mtu -= 8;
995 }
996 if (mtu < IPV6_MIN_MTU)
997 mtu = IPV6_MIN_MTU;
998 if (skb_dst(skb))
999 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1000 if (skb->len > mtu) {
1001 *pmtu = mtu;
1002 err = -EMSGSIZE;
1003 goto tx_err_dst_release;
1004 }
1005
1006 /*
1007 * Okay, now see if we can stuff it in the buffer as-is.
1008 */
1009 max_headroom += LL_RESERVED_SPACE(tdev);
1010
1011 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1012 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1013 struct sk_buff *new_skb;
1014
1015 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1016 goto tx_err_dst_release;
1017
1018 if (skb->sk)
1019 skb_set_owner_w(new_skb, skb->sk);
1020 consume_skb(skb);
1021 skb = new_skb;
1022 }
1023 skb_dst_drop(skb);
1024 if (!use_cache) {
1025 skb_dst_set(skb, dst);
1026 ndst = NULL;
1027 } else {
1028 skb_dst_set_noref(skb, dst);
1029 }
1030 skb->transport_header = skb->network_header;
1031
1032 proto = fl6->flowi6_proto;
1033 if (encap_limit >= 0) {
1034 init_tel_txopt(&opt, encap_limit);
1035 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1036 }
1037 skb_push(skb, sizeof(struct ipv6hdr));
1038 skb_reset_network_header(skb);
1039 ipv6h = ipv6_hdr(skb);
1040 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
1041 ipv6h->hop_limit = t->parms.hop_limit;
1042 ipv6h->nexthdr = proto;
1043 ipv6h->saddr = fl6->saddr;
1044 ipv6h->daddr = fl6->daddr;
1045 ip6tunnel_xmit(skb, dev);
1046 if (ndst)
1047 ip6_tnl_dst_store(t, ndst);
1048 return 0;
1049 tx_err_link_failure:
1050 stats->tx_carrier_errors++;
1051 dst_link_failure(skb);
1052 tx_err_dst_release:
1053 dst_release(ndst);
1054 return err;
1055 }
1056
1057 static inline int
1058 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1059 {
1060 struct ip6_tnl *t = netdev_priv(dev);
1061 const struct iphdr *iph = ip_hdr(skb);
1062 int encap_limit = -1;
1063 struct flowi6 fl6;
1064 __u8 dsfield;
1065 __u32 mtu;
1066 int err;
1067
1068 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1069 !ip6_tnl_xmit_ctl(t))
1070 return -1;
1071
1072 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1073 encap_limit = t->parms.encap_limit;
1074
1075 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1076 fl6.flowi6_proto = IPPROTO_IPIP;
1077
1078 dsfield = ipv4_get_dsfield(iph);
1079
1080 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1081 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1082 & IPV6_TCLASS_MASK;
1083 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1084 fl6.flowi6_mark = skb->mark;
1085
1086 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1087 if (err != 0) {
1088 /* XXX: send ICMP error even if DF is not set. */
1089 if (err == -EMSGSIZE)
1090 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1091 htonl(mtu));
1092 return -1;
1093 }
1094
1095 return 0;
1096 }
1097
1098 static inline int
1099 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1100 {
1101 struct ip6_tnl *t = netdev_priv(dev);
1102 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1103 int encap_limit = -1;
1104 __u16 offset;
1105 struct flowi6 fl6;
1106 __u8 dsfield;
1107 __u32 mtu;
1108 int err;
1109
1110 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1111 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1112 return -1;
1113
1114 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1115 if (offset > 0) {
1116 struct ipv6_tlv_tnl_enc_lim *tel;
1117 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1118 if (tel->encap_limit == 0) {
1119 icmpv6_send(skb, ICMPV6_PARAMPROB,
1120 ICMPV6_HDR_FIELD, offset + 2);
1121 return -1;
1122 }
1123 encap_limit = tel->encap_limit - 1;
1124 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1125 encap_limit = t->parms.encap_limit;
1126
1127 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1128 fl6.flowi6_proto = IPPROTO_IPV6;
1129
1130 dsfield = ipv6_get_dsfield(ipv6h);
1131 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1132 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1133 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1134 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1135 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1136 fl6.flowi6_mark = skb->mark;
1137
1138 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1139 if (err != 0) {
1140 if (err == -EMSGSIZE)
1141 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1142 return -1;
1143 }
1144
1145 return 0;
1146 }
1147
1148 static netdev_tx_t
1149 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1150 {
1151 struct ip6_tnl *t = netdev_priv(dev);
1152 struct net_device_stats *stats = &t->dev->stats;
1153 int ret;
1154
1155 switch (skb->protocol) {
1156 case htons(ETH_P_IP):
1157 ret = ip4ip6_tnl_xmit(skb, dev);
1158 break;
1159 case htons(ETH_P_IPV6):
1160 ret = ip6ip6_tnl_xmit(skb, dev);
1161 break;
1162 default:
1163 goto tx_err;
1164 }
1165
1166 if (ret < 0)
1167 goto tx_err;
1168
1169 return NETDEV_TX_OK;
1170
1171 tx_err:
1172 stats->tx_errors++;
1173 stats->tx_dropped++;
1174 kfree_skb(skb);
1175 return NETDEV_TX_OK;
1176 }
1177
1178 static void ip6_tnl_link_config(struct ip6_tnl *t)
1179 {
1180 struct net_device *dev = t->dev;
1181 struct __ip6_tnl_parm *p = &t->parms;
1182 struct flowi6 *fl6 = &t->fl.u.ip6;
1183
1184 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1185 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1186
1187 /* Set up flowi template */
1188 fl6->saddr = p->laddr;
1189 fl6->daddr = p->raddr;
1190 fl6->flowi6_oif = p->link;
1191 fl6->flowlabel = 0;
1192
1193 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1194 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1195 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1196 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1197
1198 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1199 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1200
1201 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1202 dev->flags |= IFF_POINTOPOINT;
1203 else
1204 dev->flags &= ~IFF_POINTOPOINT;
1205
1206 dev->iflink = p->link;
1207
1208 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1209 int strict = (ipv6_addr_type(&p->raddr) &
1210 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1211
1212 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1213 &p->raddr, &p->laddr,
1214 p->link, strict);
1215
1216 if (rt == NULL)
1217 return;
1218
1219 if (rt->dst.dev) {
1220 dev->hard_header_len = rt->dst.dev->hard_header_len +
1221 sizeof (struct ipv6hdr);
1222
1223 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1224 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1225 dev->mtu-=8;
1226
1227 if (dev->mtu < IPV6_MIN_MTU)
1228 dev->mtu = IPV6_MIN_MTU;
1229 }
1230 ip6_rt_put(rt);
1231 }
1232 }
1233
1234 /**
1235 * ip6_tnl_change - update the tunnel parameters
1236 * @t: tunnel to be changed
1237 * @p: tunnel configuration parameters
1238 *
1239 * Description:
1240 * ip6_tnl_change() updates the tunnel parameters
1241 **/
1242
1243 static int
1244 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1245 {
1246 t->parms.laddr = p->laddr;
1247 t->parms.raddr = p->raddr;
1248 t->parms.flags = p->flags;
1249 t->parms.hop_limit = p->hop_limit;
1250 t->parms.encap_limit = p->encap_limit;
1251 t->parms.flowinfo = p->flowinfo;
1252 t->parms.link = p->link;
1253 t->parms.proto = p->proto;
1254 ip6_tnl_dst_reset(t);
1255 ip6_tnl_link_config(t);
1256 return 0;
1257 }
1258
1259 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1260 {
1261 struct net *net = dev_net(t->dev);
1262 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1263 int err;
1264
1265 ip6_tnl_unlink(ip6n, t);
1266 synchronize_net();
1267 err = ip6_tnl_change(t, p);
1268 ip6_tnl_link(ip6n, t);
1269 netdev_state_change(t->dev);
1270 return err;
1271 }
1272
1273 static void
1274 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1275 {
1276 p->laddr = u->laddr;
1277 p->raddr = u->raddr;
1278 p->flags = u->flags;
1279 p->hop_limit = u->hop_limit;
1280 p->encap_limit = u->encap_limit;
1281 p->flowinfo = u->flowinfo;
1282 p->link = u->link;
1283 p->proto = u->proto;
1284 memcpy(p->name, u->name, sizeof(u->name));
1285 }
1286
1287 static void
1288 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1289 {
1290 u->laddr = p->laddr;
1291 u->raddr = p->raddr;
1292 u->flags = p->flags;
1293 u->hop_limit = p->hop_limit;
1294 u->encap_limit = p->encap_limit;
1295 u->flowinfo = p->flowinfo;
1296 u->link = p->link;
1297 u->proto = p->proto;
1298 memcpy(u->name, p->name, sizeof(u->name));
1299 }
1300
1301 /**
1302 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1303 * @dev: virtual device associated with tunnel
1304 * @ifr: parameters passed from userspace
1305 * @cmd: command to be performed
1306 *
1307 * Description:
1308 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1309 * from userspace.
1310 *
1311 * The possible commands are the following:
1312 * %SIOCGETTUNNEL: get tunnel parameters for device
1313 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1314 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1315 * %SIOCDELTUNNEL: delete tunnel
1316 *
1317 * The fallback device "ip6tnl0", created during module
1318 * initialization, can be used for creating other tunnel devices.
1319 *
1320 * Return:
1321 * 0 on success,
1322 * %-EFAULT if unable to copy data to or from userspace,
1323 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1324 * %-EINVAL if passed tunnel parameters are invalid,
1325 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1326 * %-ENODEV if attempting to change or delete a nonexisting device
1327 **/
1328
1329 static int
1330 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1331 {
1332 int err = 0;
1333 struct ip6_tnl_parm p;
1334 struct __ip6_tnl_parm p1;
1335 struct ip6_tnl *t = NULL;
1336 struct net *net = dev_net(dev);
1337 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1338
1339 switch (cmd) {
1340 case SIOCGETTUNNEL:
1341 if (dev == ip6n->fb_tnl_dev) {
1342 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1343 err = -EFAULT;
1344 break;
1345 }
1346 ip6_tnl_parm_from_user(&p1, &p);
1347 t = ip6_tnl_locate(net, &p1, 0);
1348 } else {
1349 memset(&p, 0, sizeof(p));
1350 }
1351 if (t == NULL)
1352 t = netdev_priv(dev);
1353 ip6_tnl_parm_to_user(&p, &t->parms);
1354 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1355 err = -EFAULT;
1356 }
1357 break;
1358 case SIOCADDTUNNEL:
1359 case SIOCCHGTUNNEL:
1360 err = -EPERM;
1361 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1362 break;
1363 err = -EFAULT;
1364 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1365 break;
1366 err = -EINVAL;
1367 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1368 p.proto != 0)
1369 break;
1370 ip6_tnl_parm_from_user(&p1, &p);
1371 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1372 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1373 if (t != NULL) {
1374 if (t->dev != dev) {
1375 err = -EEXIST;
1376 break;
1377 }
1378 } else
1379 t = netdev_priv(dev);
1380
1381 err = ip6_tnl_update(t, &p1);
1382 }
1383 if (t) {
1384 err = 0;
1385 ip6_tnl_parm_to_user(&p, &t->parms);
1386 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1387 err = -EFAULT;
1388
1389 } else
1390 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1391 break;
1392 case SIOCDELTUNNEL:
1393 err = -EPERM;
1394 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1395 break;
1396
1397 if (dev == ip6n->fb_tnl_dev) {
1398 err = -EFAULT;
1399 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1400 break;
1401 err = -ENOENT;
1402 ip6_tnl_parm_from_user(&p1, &p);
1403 t = ip6_tnl_locate(net, &p1, 0);
1404 if (t == NULL)
1405 break;
1406 err = -EPERM;
1407 if (t->dev == ip6n->fb_tnl_dev)
1408 break;
1409 dev = t->dev;
1410 }
1411 err = 0;
1412 unregister_netdevice(dev);
1413 break;
1414 default:
1415 err = -EINVAL;
1416 }
1417 return err;
1418 }
1419
1420 /**
1421 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1422 * @dev: virtual device associated with tunnel
1423 * @new_mtu: the new mtu
1424 *
1425 * Return:
1426 * 0 on success,
1427 * %-EINVAL if mtu too small
1428 **/
1429
1430 static int
1431 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1432 {
1433 if (new_mtu < IPV6_MIN_MTU) {
1434 return -EINVAL;
1435 }
1436 dev->mtu = new_mtu;
1437 return 0;
1438 }
1439
1440
1441 static const struct net_device_ops ip6_tnl_netdev_ops = {
1442 .ndo_init = ip6_tnl_dev_init,
1443 .ndo_uninit = ip6_tnl_dev_uninit,
1444 .ndo_start_xmit = ip6_tnl_xmit,
1445 .ndo_do_ioctl = ip6_tnl_ioctl,
1446 .ndo_change_mtu = ip6_tnl_change_mtu,
1447 .ndo_get_stats = ip6_get_stats,
1448 };
1449
1450
1451 /**
1452 * ip6_tnl_dev_setup - setup virtual tunnel device
1453 * @dev: virtual device associated with tunnel
1454 *
1455 * Description:
1456 * Initialize function pointers and device parameters
1457 **/
1458
1459 static void ip6_tnl_dev_setup(struct net_device *dev)
1460 {
1461 struct ip6_tnl *t;
1462
1463 dev->netdev_ops = &ip6_tnl_netdev_ops;
1464 dev->destructor = ip6_dev_free;
1465
1466 dev->type = ARPHRD_TUNNEL6;
1467 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1468 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1469 t = netdev_priv(dev);
1470 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1471 dev->mtu-=8;
1472 dev->flags |= IFF_NOARP;
1473 dev->addr_len = sizeof(struct in6_addr);
1474 dev->features |= NETIF_F_NETNS_LOCAL;
1475 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1476 }
1477
1478
1479 /**
1480 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1481 * @dev: virtual device associated with tunnel
1482 **/
1483
1484 static inline int
1485 ip6_tnl_dev_init_gen(struct net_device *dev)
1486 {
1487 struct ip6_tnl *t = netdev_priv(dev);
1488
1489 t->dev = dev;
1490 dev->tstats = alloc_percpu(struct pcpu_tstats);
1491 if (!dev->tstats)
1492 return -ENOMEM;
1493 return 0;
1494 }
1495
1496 /**
1497 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1498 * @dev: virtual device associated with tunnel
1499 **/
1500
1501 static int ip6_tnl_dev_init(struct net_device *dev)
1502 {
1503 struct ip6_tnl *t = netdev_priv(dev);
1504 int err = ip6_tnl_dev_init_gen(dev);
1505
1506 if (err)
1507 return err;
1508 ip6_tnl_link_config(t);
1509 return 0;
1510 }
1511
1512 /**
1513 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1514 * @dev: fallback device
1515 *
1516 * Return: 0
1517 **/
1518
1519 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1520 {
1521 struct ip6_tnl *t = netdev_priv(dev);
1522 struct net *net = dev_net(dev);
1523 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1524
1525 t->parms.proto = IPPROTO_IPV6;
1526 dev_hold(dev);
1527
1528 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1529 return 0;
1530 }
1531
1532 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1533 {
1534 u8 proto;
1535
1536 if (!data || !data[IFLA_IPTUN_PROTO])
1537 return 0;
1538
1539 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1540 if (proto != IPPROTO_IPV6 &&
1541 proto != IPPROTO_IPIP &&
1542 proto != 0)
1543 return -EINVAL;
1544
1545 return 0;
1546 }
1547
1548 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1549 struct __ip6_tnl_parm *parms)
1550 {
1551 memset(parms, 0, sizeof(*parms));
1552
1553 if (!data)
1554 return;
1555
1556 if (data[IFLA_IPTUN_LINK])
1557 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1558
1559 if (data[IFLA_IPTUN_LOCAL])
1560 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1561 sizeof(struct in6_addr));
1562
1563 if (data[IFLA_IPTUN_REMOTE])
1564 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1565 sizeof(struct in6_addr));
1566
1567 if (data[IFLA_IPTUN_TTL])
1568 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1569
1570 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1571 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1572
1573 if (data[IFLA_IPTUN_FLOWINFO])
1574 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1575
1576 if (data[IFLA_IPTUN_FLAGS])
1577 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1578
1579 if (data[IFLA_IPTUN_PROTO])
1580 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1581 }
1582
1583 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1584 struct nlattr *tb[], struct nlattr *data[])
1585 {
1586 struct net *net = dev_net(dev);
1587 struct ip6_tnl *nt;
1588
1589 nt = netdev_priv(dev);
1590 ip6_tnl_netlink_parms(data, &nt->parms);
1591
1592 if (ip6_tnl_locate(net, &nt->parms, 0))
1593 return -EEXIST;
1594
1595 return ip6_tnl_create2(dev);
1596 }
1597
1598 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1599 struct nlattr *data[])
1600 {
1601 struct ip6_tnl *t;
1602 struct __ip6_tnl_parm p;
1603 struct net *net = dev_net(dev);
1604 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1605
1606 if (dev == ip6n->fb_tnl_dev)
1607 return -EINVAL;
1608
1609 ip6_tnl_netlink_parms(data, &p);
1610
1611 t = ip6_tnl_locate(net, &p, 0);
1612
1613 if (t) {
1614 if (t->dev != dev)
1615 return -EEXIST;
1616 } else
1617 t = netdev_priv(dev);
1618
1619 return ip6_tnl_update(t, &p);
1620 }
1621
1622 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1623 {
1624 struct net *net = dev_net(dev);
1625 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1626
1627 if (dev != ip6n->fb_tnl_dev)
1628 unregister_netdevice_queue(dev, head);
1629 }
1630
1631 static size_t ip6_tnl_get_size(const struct net_device *dev)
1632 {
1633 return
1634 /* IFLA_IPTUN_LINK */
1635 nla_total_size(4) +
1636 /* IFLA_IPTUN_LOCAL */
1637 nla_total_size(sizeof(struct in6_addr)) +
1638 /* IFLA_IPTUN_REMOTE */
1639 nla_total_size(sizeof(struct in6_addr)) +
1640 /* IFLA_IPTUN_TTL */
1641 nla_total_size(1) +
1642 /* IFLA_IPTUN_ENCAP_LIMIT */
1643 nla_total_size(1) +
1644 /* IFLA_IPTUN_FLOWINFO */
1645 nla_total_size(4) +
1646 /* IFLA_IPTUN_FLAGS */
1647 nla_total_size(4) +
1648 /* IFLA_IPTUN_PROTO */
1649 nla_total_size(1) +
1650 0;
1651 }
1652
1653 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1654 {
1655 struct ip6_tnl *tunnel = netdev_priv(dev);
1656 struct __ip6_tnl_parm *parm = &tunnel->parms;
1657
1658 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1659 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1660 &parm->laddr) ||
1661 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1662 &parm->raddr) ||
1663 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1664 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1665 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1666 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1667 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1668 goto nla_put_failure;
1669 return 0;
1670
1671 nla_put_failure:
1672 return -EMSGSIZE;
1673 }
1674
1675 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1676 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1677 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1678 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1679 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1680 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1681 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1682 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1683 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1684 };
1685
1686 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1687 .kind = "ip6tnl",
1688 .maxtype = IFLA_IPTUN_MAX,
1689 .policy = ip6_tnl_policy,
1690 .priv_size = sizeof(struct ip6_tnl),
1691 .setup = ip6_tnl_dev_setup,
1692 .validate = ip6_tnl_validate,
1693 .newlink = ip6_tnl_newlink,
1694 .changelink = ip6_tnl_changelink,
1695 .dellink = ip6_tnl_dellink,
1696 .get_size = ip6_tnl_get_size,
1697 .fill_info = ip6_tnl_fill_info,
1698 };
1699
1700 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1701 .handler = ip4ip6_rcv,
1702 .err_handler = ip4ip6_err,
1703 .priority = 1,
1704 };
1705
1706 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1707 .handler = ip6ip6_rcv,
1708 .err_handler = ip6ip6_err,
1709 .priority = 1,
1710 };
1711
1712 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1713 {
1714 int h;
1715 struct ip6_tnl *t;
1716 LIST_HEAD(list);
1717
1718 for (h = 0; h < HASH_SIZE; h++) {
1719 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1720 while (t != NULL) {
1721 unregister_netdevice_queue(t->dev, &list);
1722 t = rtnl_dereference(t->next);
1723 }
1724 }
1725
1726 t = rtnl_dereference(ip6n->tnls_wc[0]);
1727 unregister_netdevice_queue(t->dev, &list);
1728 unregister_netdevice_many(&list);
1729 }
1730
1731 static int __net_init ip6_tnl_init_net(struct net *net)
1732 {
1733 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1734 struct ip6_tnl *t = NULL;
1735 int err;
1736
1737 ip6n->tnls[0] = ip6n->tnls_wc;
1738 ip6n->tnls[1] = ip6n->tnls_r_l;
1739
1740 err = -ENOMEM;
1741 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1742 ip6_tnl_dev_setup);
1743
1744 if (!ip6n->fb_tnl_dev)
1745 goto err_alloc_dev;
1746 dev_net_set(ip6n->fb_tnl_dev, net);
1747 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1748
1749 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1750 if (err < 0)
1751 goto err_register;
1752
1753 err = register_netdev(ip6n->fb_tnl_dev);
1754 if (err < 0)
1755 goto err_register;
1756
1757 t = netdev_priv(ip6n->fb_tnl_dev);
1758
1759 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1760 return 0;
1761
1762 err_register:
1763 ip6_dev_free(ip6n->fb_tnl_dev);
1764 err_alloc_dev:
1765 return err;
1766 }
1767
1768 static void __net_exit ip6_tnl_exit_net(struct net *net)
1769 {
1770 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1771
1772 rtnl_lock();
1773 ip6_tnl_destroy_tunnels(ip6n);
1774 rtnl_unlock();
1775 }
1776
1777 static struct pernet_operations ip6_tnl_net_ops = {
1778 .init = ip6_tnl_init_net,
1779 .exit = ip6_tnl_exit_net,
1780 .id = &ip6_tnl_net_id,
1781 .size = sizeof(struct ip6_tnl_net),
1782 };
1783
1784 /**
1785 * ip6_tunnel_init - register protocol and reserve needed resources
1786 *
1787 * Return: 0 on success
1788 **/
1789
1790 static int __init ip6_tunnel_init(void)
1791 {
1792 int err;
1793
1794 err = register_pernet_device(&ip6_tnl_net_ops);
1795 if (err < 0)
1796 goto out_pernet;
1797
1798 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1799 if (err < 0) {
1800 pr_err("%s: can't register ip4ip6\n", __func__);
1801 goto out_ip4ip6;
1802 }
1803
1804 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1805 if (err < 0) {
1806 pr_err("%s: can't register ip6ip6\n", __func__);
1807 goto out_ip6ip6;
1808 }
1809 err = rtnl_link_register(&ip6_link_ops);
1810 if (err < 0)
1811 goto rtnl_link_failed;
1812
1813 return 0;
1814
1815 rtnl_link_failed:
1816 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1817 out_ip6ip6:
1818 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1819 out_ip4ip6:
1820 unregister_pernet_device(&ip6_tnl_net_ops);
1821 out_pernet:
1822 return err;
1823 }
1824
1825 /**
1826 * ip6_tunnel_cleanup - free resources and unregister protocol
1827 **/
1828
1829 static void __exit ip6_tunnel_cleanup(void)
1830 {
1831 rtnl_link_unregister(&ip6_link_ops);
1832 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1833 pr_info("%s: can't deregister ip4ip6\n", __func__);
1834
1835 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1836 pr_info("%s: can't deregister ip6ip6\n", __func__);
1837
1838 unregister_pernet_device(&ip6_tnl_net_ops);
1839 }
1840
1841 module_init(ip6_tunnel_init);
1842 module_exit(ip6_tunnel_cleanup);
kernel source for telekom puls (alcatel/mediatek)