]> Pileus Git - ~andy/linux/blob - net/ipv6/ip6_tunnel.c
tcp: trivial: Remove nocache argument from tcp_v4_send_synack
[~andy/linux] / 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_NETDEV("ip6tnl0");
65
66 #ifdef IP6_TNL_DEBUG
67 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
68 #else
69 #define IP6_TNL_TRACE(x...) do {;} while(0)
70 #endif
71
72 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
73 #define IPV6_TCLASS_SHIFT 20
74
75 #define HASH_SIZE_SHIFT  5
76 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
77
78 static bool log_ecn_error = true;
79 module_param(log_ecn_error, bool, 0644);
80 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
81
82 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
83 {
84         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
85
86         return hash_32(hash, HASH_SIZE_SHIFT);
87 }
88
89 static int ip6_tnl_dev_init(struct net_device *dev);
90 static void ip6_tnl_dev_setup(struct net_device *dev);
91 static struct rtnl_link_ops ip6_link_ops __read_mostly;
92
93 static int ip6_tnl_net_id __read_mostly;
94 struct ip6_tnl_net {
95         /* the IPv6 tunnel fallback device */
96         struct net_device *fb_tnl_dev;
97         /* lists for storing tunnels in use */
98         struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
99         struct ip6_tnl __rcu *tnls_wc[1];
100         struct ip6_tnl __rcu **tnls[2];
101 };
102
103 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
104 {
105         struct pcpu_tstats sum = { 0 };
106         int i;
107
108         for_each_possible_cpu(i) {
109                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
110
111                 sum.rx_packets += tstats->rx_packets;
112                 sum.rx_bytes   += tstats->rx_bytes;
113                 sum.tx_packets += tstats->tx_packets;
114                 sum.tx_bytes   += tstats->tx_bytes;
115         }
116         dev->stats.rx_packets = sum.rx_packets;
117         dev->stats.rx_bytes   = sum.rx_bytes;
118         dev->stats.tx_packets = sum.tx_packets;
119         dev->stats.tx_bytes   = sum.tx_bytes;
120         return &dev->stats;
121 }
122
123 /*
124  * Locking : hash tables are protected by RCU and RTNL
125  */
126
127 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
128 {
129         struct dst_entry *dst = t->dst_cache;
130
131         if (dst && dst->obsolete &&
132             dst->ops->check(dst, t->dst_cookie) == NULL) {
133                 t->dst_cache = NULL;
134                 dst_release(dst);
135                 return NULL;
136         }
137
138         return dst;
139 }
140 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
141
142 void ip6_tnl_dst_reset(struct ip6_tnl *t)
143 {
144         dst_release(t->dst_cache);
145         t->dst_cache = NULL;
146 }
147 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
148
149 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
150 {
151         struct rt6_info *rt = (struct rt6_info *) dst;
152         t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
153         dst_release(t->dst_cache);
154         t->dst_cache = dst;
155 }
156 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
157
158 /**
159  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
160  *   @remote: the address of the tunnel exit-point
161  *   @local: the address of the tunnel entry-point
162  *
163  * Return:
164  *   tunnel matching given end-points if found,
165  *   else fallback tunnel if its device is up,
166  *   else %NULL
167  **/
168
169 #define for_each_ip6_tunnel_rcu(start) \
170         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
171
172 static struct ip6_tnl *
173 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
174 {
175         unsigned int hash = HASH(remote, local);
176         struct ip6_tnl *t;
177         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
178
179         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
180                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
181                     ipv6_addr_equal(remote, &t->parms.raddr) &&
182                     (t->dev->flags & IFF_UP))
183                         return t;
184         }
185         t = rcu_dereference(ip6n->tnls_wc[0]);
186         if (t && (t->dev->flags & IFF_UP))
187                 return t;
188
189         return NULL;
190 }
191
192 /**
193  * ip6_tnl_bucket - get head of list matching given tunnel parameters
194  *   @p: parameters containing tunnel end-points
195  *
196  * Description:
197  *   ip6_tnl_bucket() returns the head of the list matching the
198  *   &struct in6_addr entries laddr and raddr in @p.
199  *
200  * Return: head of IPv6 tunnel list
201  **/
202
203 static struct ip6_tnl __rcu **
204 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
205 {
206         const struct in6_addr *remote = &p->raddr;
207         const struct in6_addr *local = &p->laddr;
208         unsigned int h = 0;
209         int prio = 0;
210
211         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
212                 prio = 1;
213                 h = HASH(remote, local);
214         }
215         return &ip6n->tnls[prio][h];
216 }
217
218 /**
219  * ip6_tnl_link - add tunnel to hash table
220  *   @t: tunnel to be added
221  **/
222
223 static void
224 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
225 {
226         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
227
228         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
229         rcu_assign_pointer(*tp, t);
230 }
231
232 /**
233  * ip6_tnl_unlink - remove tunnel from hash table
234  *   @t: tunnel to be removed
235  **/
236
237 static void
238 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
239 {
240         struct ip6_tnl __rcu **tp;
241         struct ip6_tnl *iter;
242
243         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
244              (iter = rtnl_dereference(*tp)) != NULL;
245              tp = &iter->next) {
246                 if (t == iter) {
247                         rcu_assign_pointer(*tp, t->next);
248                         break;
249                 }
250         }
251 }
252
253 static void ip6_dev_free(struct net_device *dev)
254 {
255         free_percpu(dev->tstats);
256         free_netdev(dev);
257 }
258
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         err = ip6_tnl_dev_init(dev);
268         if (err < 0)
269                 goto out;
270
271         err = register_netdevice(dev);
272         if (err < 0)
273                 goto out;
274
275         strcpy(t->parms.name, dev->name);
276         dev->rtnl_link_ops = &ip6_link_ops;
277
278         dev_hold(dev);
279         ip6_tnl_link(ip6n, t);
280         return 0;
281
282 out:
283         return err;
284 }
285
286 /**
287  * ip6_tnl_create - create a new tunnel
288  *   @p: tunnel parameters
289  *   @pt: pointer to new tunnel
290  *
291  * Description:
292  *   Create tunnel matching given parameters.
293  *
294  * Return:
295  *   created tunnel or NULL
296  **/
297
298 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
299 {
300         struct net_device *dev;
301         struct ip6_tnl *t;
302         char name[IFNAMSIZ];
303         int err;
304
305         if (p->name[0])
306                 strlcpy(name, p->name, IFNAMSIZ);
307         else
308                 sprintf(name, "ip6tnl%%d");
309
310         dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
311         if (dev == NULL)
312                 goto failed;
313
314         dev_net_set(dev, net);
315
316         t = netdev_priv(dev);
317         t->parms = *p;
318         t->net = dev_net(dev);
319         err = ip6_tnl_create2(dev);
320         if (err < 0)
321                 goto failed_free;
322
323         return t;
324
325 failed_free:
326         ip6_dev_free(dev);
327 failed:
328         return NULL;
329 }
330
331 /**
332  * ip6_tnl_locate - find or create tunnel matching given parameters
333  *   @p: tunnel parameters
334  *   @create: != 0 if allowed to create new tunnel if no match found
335  *
336  * Description:
337  *   ip6_tnl_locate() first tries to locate an existing tunnel
338  *   based on @parms. If this is unsuccessful, but @create is set a new
339  *   tunnel device is created and registered for use.
340  *
341  * Return:
342  *   matching tunnel or NULL
343  **/
344
345 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
346                 struct __ip6_tnl_parm *p, int create)
347 {
348         const struct in6_addr *remote = &p->raddr;
349         const struct in6_addr *local = &p->laddr;
350         struct ip6_tnl __rcu **tp;
351         struct ip6_tnl *t;
352         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
353
354         for (tp = ip6_tnl_bucket(ip6n, p);
355              (t = rtnl_dereference(*tp)) != NULL;
356              tp = &t->next) {
357                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
358                     ipv6_addr_equal(remote, &t->parms.raddr))
359                         return t;
360         }
361         if (!create)
362                 return NULL;
363         return ip6_tnl_create(net, p);
364 }
365
366 /**
367  * ip6_tnl_dev_uninit - tunnel device uninitializer
368  *   @dev: the device to be destroyed
369  *
370  * Description:
371  *   ip6_tnl_dev_uninit() removes tunnel from its list
372  **/
373
374 static void
375 ip6_tnl_dev_uninit(struct net_device *dev)
376 {
377         struct ip6_tnl *t = netdev_priv(dev);
378         struct net *net = t->net;
379         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
380
381         if (dev == ip6n->fb_tnl_dev)
382                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
383         else
384                 ip6_tnl_unlink(ip6n, t);
385         ip6_tnl_dst_reset(t);
386         dev_put(dev);
387 }
388
389 /**
390  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
391  *   @skb: received socket buffer
392  *
393  * Return:
394  *   0 if none was found,
395  *   else index to encapsulation limit
396  **/
397
398 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
399 {
400         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
401         __u8 nexthdr = ipv6h->nexthdr;
402         __u16 off = sizeof (*ipv6h);
403
404         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
405                 __u16 optlen = 0;
406                 struct ipv6_opt_hdr *hdr;
407                 if (raw + off + sizeof (*hdr) > skb->data &&
408                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
409                         break;
410
411                 hdr = (struct ipv6_opt_hdr *) (raw + off);
412                 if (nexthdr == NEXTHDR_FRAGMENT) {
413                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
414                         if (frag_hdr->frag_off)
415                                 break;
416                         optlen = 8;
417                 } else if (nexthdr == NEXTHDR_AUTH) {
418                         optlen = (hdr->hdrlen + 2) << 2;
419                 } else {
420                         optlen = ipv6_optlen(hdr);
421                 }
422                 if (nexthdr == NEXTHDR_DEST) {
423                         __u16 i = off + 2;
424                         while (1) {
425                                 struct ipv6_tlv_tnl_enc_lim *tel;
426
427                                 /* No more room for encapsulation limit */
428                                 if (i + sizeof (*tel) > off + optlen)
429                                         break;
430
431                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
432                                 /* return index of option if found and valid */
433                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
434                                     tel->length == 1)
435                                         return i;
436                                 /* else jump to next option */
437                                 if (tel->type)
438                                         i += tel->length + 2;
439                                 else
440                                         i++;
441                         }
442                 }
443                 nexthdr = hdr->nexthdr;
444                 off += optlen;
445         }
446         return 0;
447 }
448 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
449
450 /**
451  * ip6_tnl_err - tunnel error handler
452  *
453  * Description:
454  *   ip6_tnl_err() should handle errors in the tunnel according
455  *   to the specifications in RFC 2473.
456  **/
457
458 static int
459 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
460             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
461 {
462         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
463         struct ip6_tnl *t;
464         int rel_msg = 0;
465         u8 rel_type = ICMPV6_DEST_UNREACH;
466         u8 rel_code = ICMPV6_ADDR_UNREACH;
467         __u32 rel_info = 0;
468         __u16 len;
469         int err = -ENOENT;
470
471         /* If the packet doesn't contain the original IPv6 header we are
472            in trouble since we might need the source address for further
473            processing of the error. */
474
475         rcu_read_lock();
476         if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
477                                         &ipv6h->saddr)) == NULL)
478                 goto out;
479
480         if (t->parms.proto != ipproto && t->parms.proto != 0)
481                 goto out;
482
483         err = 0;
484
485         switch (*type) {
486                 __u32 teli;
487                 struct ipv6_tlv_tnl_enc_lim *tel;
488                 __u32 mtu;
489         case ICMPV6_DEST_UNREACH:
490                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
491                                      t->parms.name);
492                 rel_msg = 1;
493                 break;
494         case ICMPV6_TIME_EXCEED:
495                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
496                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
497                                              t->parms.name);
498                         rel_msg = 1;
499                 }
500                 break;
501         case ICMPV6_PARAMPROB:
502                 teli = 0;
503                 if ((*code) == ICMPV6_HDR_FIELD)
504                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
505
506                 if (teli && teli == *info - 2) {
507                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
508                         if (tel->encap_limit == 0) {
509                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
510                                                      t->parms.name);
511                                 rel_msg = 1;
512                         }
513                 } else {
514                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
515                                              t->parms.name);
516                 }
517                 break;
518         case ICMPV6_PKT_TOOBIG:
519                 mtu = *info - offset;
520                 if (mtu < IPV6_MIN_MTU)
521                         mtu = IPV6_MIN_MTU;
522                 t->dev->mtu = mtu;
523
524                 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
525                         rel_type = ICMPV6_PKT_TOOBIG;
526                         rel_code = 0;
527                         rel_info = mtu;
528                         rel_msg = 1;
529                 }
530                 break;
531         }
532
533         *type = rel_type;
534         *code = rel_code;
535         *info = rel_info;
536         *msg = rel_msg;
537
538 out:
539         rcu_read_unlock();
540         return err;
541 }
542
543 static int
544 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
545            u8 type, u8 code, int offset, __be32 info)
546 {
547         int rel_msg = 0;
548         u8 rel_type = type;
549         u8 rel_code = code;
550         __u32 rel_info = ntohl(info);
551         int err;
552         struct sk_buff *skb2;
553         const struct iphdr *eiph;
554         struct rtable *rt;
555         struct flowi4 fl4;
556
557         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
558                           &rel_msg, &rel_info, offset);
559         if (err < 0)
560                 return err;
561
562         if (rel_msg == 0)
563                 return 0;
564
565         switch (rel_type) {
566         case ICMPV6_DEST_UNREACH:
567                 if (rel_code != ICMPV6_ADDR_UNREACH)
568                         return 0;
569                 rel_type = ICMP_DEST_UNREACH;
570                 rel_code = ICMP_HOST_UNREACH;
571                 break;
572         case ICMPV6_PKT_TOOBIG:
573                 if (rel_code != 0)
574                         return 0;
575                 rel_type = ICMP_DEST_UNREACH;
576                 rel_code = ICMP_FRAG_NEEDED;
577                 break;
578         case NDISC_REDIRECT:
579                 rel_type = ICMP_REDIRECT;
580                 rel_code = ICMP_REDIR_HOST;
581         default:
582                 return 0;
583         }
584
585         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
586                 return 0;
587
588         skb2 = skb_clone(skb, GFP_ATOMIC);
589         if (!skb2)
590                 return 0;
591
592         skb_dst_drop(skb2);
593
594         skb_pull(skb2, offset);
595         skb_reset_network_header(skb2);
596         eiph = ip_hdr(skb2);
597
598         /* Try to guess incoming interface */
599         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
600                                    eiph->saddr, 0,
601                                    0, 0,
602                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
603         if (IS_ERR(rt))
604                 goto out;
605
606         skb2->dev = rt->dst.dev;
607
608         /* route "incoming" packet */
609         if (rt->rt_flags & RTCF_LOCAL) {
610                 ip_rt_put(rt);
611                 rt = NULL;
612                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
613                                            eiph->daddr, eiph->saddr,
614                                            0, 0,
615                                            IPPROTO_IPIP,
616                                            RT_TOS(eiph->tos), 0);
617                 if (IS_ERR(rt) ||
618                     rt->dst.dev->type != ARPHRD_TUNNEL) {
619                         if (!IS_ERR(rt))
620                                 ip_rt_put(rt);
621                         goto out;
622                 }
623                 skb_dst_set(skb2, &rt->dst);
624         } else {
625                 ip_rt_put(rt);
626                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
627                                    skb2->dev) ||
628                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
629                         goto out;
630         }
631
632         /* change mtu on this route */
633         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
634                 if (rel_info > dst_mtu(skb_dst(skb2)))
635                         goto out;
636
637                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
638         }
639         if (rel_type == ICMP_REDIRECT)
640                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
641
642         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
643
644 out:
645         kfree_skb(skb2);
646         return 0;
647 }
648
649 static int
650 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
651            u8 type, u8 code, int offset, __be32 info)
652 {
653         int rel_msg = 0;
654         u8 rel_type = type;
655         u8 rel_code = code;
656         __u32 rel_info = ntohl(info);
657         int err;
658
659         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
660                           &rel_msg, &rel_info, offset);
661         if (err < 0)
662                 return err;
663
664         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
665                 struct rt6_info *rt;
666                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
667
668                 if (!skb2)
669                         return 0;
670
671                 skb_dst_drop(skb2);
672                 skb_pull(skb2, offset);
673                 skb_reset_network_header(skb2);
674
675                 /* Try to guess incoming interface */
676                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
677                                 NULL, 0, 0);
678
679                 if (rt && rt->dst.dev)
680                         skb2->dev = rt->dst.dev;
681
682                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
683
684                 ip6_rt_put(rt);
685
686                 kfree_skb(skb2);
687         }
688
689         return 0;
690 }
691
692 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
693                                        const struct ipv6hdr *ipv6h,
694                                        struct sk_buff *skb)
695 {
696         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
697
698         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
699                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
700
701         return IP6_ECN_decapsulate(ipv6h, skb);
702 }
703
704 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
705                                        const struct ipv6hdr *ipv6h,
706                                        struct sk_buff *skb)
707 {
708         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
709                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
710
711         return IP6_ECN_decapsulate(ipv6h, skb);
712 }
713
714 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
715                              const struct in6_addr *laddr,
716                              const struct in6_addr *raddr)
717 {
718         struct __ip6_tnl_parm *p = &t->parms;
719         int ltype = ipv6_addr_type(laddr);
720         int rtype = ipv6_addr_type(raddr);
721         __u32 flags = 0;
722
723         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
724                 flags = IP6_TNL_F_CAP_PER_PACKET;
725         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
726                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
727                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
728                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
729                 if (ltype&IPV6_ADDR_UNICAST)
730                         flags |= IP6_TNL_F_CAP_XMIT;
731                 if (rtype&IPV6_ADDR_UNICAST)
732                         flags |= IP6_TNL_F_CAP_RCV;
733         }
734         return flags;
735 }
736 EXPORT_SYMBOL(ip6_tnl_get_cap);
737
738 /* called with rcu_read_lock() */
739 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
740                                   const struct in6_addr *laddr,
741                                   const struct in6_addr *raddr)
742 {
743         struct __ip6_tnl_parm *p = &t->parms;
744         int ret = 0;
745         struct net *net = t->net;
746
747         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
748             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
749              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
750                 struct net_device *ldev = NULL;
751
752                 if (p->link)
753                         ldev = dev_get_by_index_rcu(net, p->link);
754
755                 if ((ipv6_addr_is_multicast(laddr) ||
756                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
757                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
758                         ret = 1;
759         }
760         return ret;
761 }
762 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
763
764 /**
765  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
766  *   @skb: received socket buffer
767  *   @protocol: ethernet protocol ID
768  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
769  *
770  * Return: 0
771  **/
772
773 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
774                        __u8 ipproto,
775                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
776                                                    const struct ipv6hdr *ipv6h,
777                                                    struct sk_buff *skb))
778 {
779         struct ip6_tnl *t;
780         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
781         int err;
782
783         rcu_read_lock();
784
785         if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
786                                         &ipv6h->daddr)) != NULL) {
787                 struct pcpu_tstats *tstats;
788
789                 if (t->parms.proto != ipproto && t->parms.proto != 0) {
790                         rcu_read_unlock();
791                         goto discard;
792                 }
793
794                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
795                         rcu_read_unlock();
796                         goto discard;
797                 }
798
799                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
800                         t->dev->stats.rx_dropped++;
801                         rcu_read_unlock();
802                         goto discard;
803                 }
804                 secpath_reset(skb);
805                 skb->mac_header = skb->network_header;
806                 skb_reset_network_header(skb);
807                 skb->protocol = htons(protocol);
808                 skb->pkt_type = PACKET_HOST;
809                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
810
811                 __skb_tunnel_rx(skb, t->dev);
812
813                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
814                 if (unlikely(err)) {
815                         if (log_ecn_error)
816                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
817                                                      &ipv6h->saddr,
818                                                      ipv6_get_dsfield(ipv6h));
819                         if (err > 1) {
820                                 ++t->dev->stats.rx_frame_errors;
821                                 ++t->dev->stats.rx_errors;
822                                 rcu_read_unlock();
823                                 goto discard;
824                         }
825                 }
826
827                 tstats = this_cpu_ptr(t->dev->tstats);
828                 tstats->rx_packets++;
829                 tstats->rx_bytes += skb->len;
830
831                 if (!net_eq(t->net, dev_net(t->dev)))
832                         skb_scrub_packet(skb);
833
834                 netif_rx(skb);
835
836                 rcu_read_unlock();
837                 return 0;
838         }
839         rcu_read_unlock();
840         return 1;
841
842 discard:
843         kfree_skb(skb);
844         return 0;
845 }
846
847 static int ip4ip6_rcv(struct sk_buff *skb)
848 {
849         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
850                            ip4ip6_dscp_ecn_decapsulate);
851 }
852
853 static int ip6ip6_rcv(struct sk_buff *skb)
854 {
855         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
856                            ip6ip6_dscp_ecn_decapsulate);
857 }
858
859 struct ipv6_tel_txoption {
860         struct ipv6_txoptions ops;
861         __u8 dst_opt[8];
862 };
863
864 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
865 {
866         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
867
868         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
869         opt->dst_opt[3] = 1;
870         opt->dst_opt[4] = encap_limit;
871         opt->dst_opt[5] = IPV6_TLV_PADN;
872         opt->dst_opt[6] = 1;
873
874         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
875         opt->ops.opt_nflen = 8;
876 }
877
878 /**
879  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
880  *   @t: the outgoing tunnel device
881  *   @hdr: IPv6 header from the incoming packet
882  *
883  * Description:
884  *   Avoid trivial tunneling loop by checking that tunnel exit-point
885  *   doesn't match source of incoming packet.
886  *
887  * Return:
888  *   1 if conflict,
889  *   0 else
890  **/
891
892 static inline bool
893 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
894 {
895         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
896 }
897
898 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
899 {
900         struct __ip6_tnl_parm *p = &t->parms;
901         int ret = 0;
902         struct net *net = t->net;
903
904         if (p->flags & IP6_TNL_F_CAP_XMIT) {
905                 struct net_device *ldev = NULL;
906
907                 rcu_read_lock();
908                 if (p->link)
909                         ldev = dev_get_by_index_rcu(net, p->link);
910
911                 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
912                         pr_warn("%s xmit: Local address not yet configured!\n",
913                                 p->name);
914                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
915                          unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
916                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
917                                 p->name);
918                 else
919                         ret = 1;
920                 rcu_read_unlock();
921         }
922         return ret;
923 }
924 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
925
926 /**
927  * ip6_tnl_xmit2 - encapsulate packet and send
928  *   @skb: the outgoing socket buffer
929  *   @dev: the outgoing tunnel device
930  *   @dsfield: dscp code for outer header
931  *   @fl: flow of tunneled packet
932  *   @encap_limit: encapsulation limit
933  *   @pmtu: Path MTU is stored if packet is too big
934  *
935  * Description:
936  *   Build new header and do some sanity checks on the packet before sending
937  *   it.
938  *
939  * Return:
940  *   0 on success
941  *   -1 fail
942  *   %-EMSGSIZE message too big. return mtu in this case.
943  **/
944
945 static int ip6_tnl_xmit2(struct sk_buff *skb,
946                          struct net_device *dev,
947                          __u8 dsfield,
948                          struct flowi6 *fl6,
949                          int encap_limit,
950                          __u32 *pmtu)
951 {
952         struct ip6_tnl *t = netdev_priv(dev);
953         struct net *net = t->net;
954         struct net_device_stats *stats = &t->dev->stats;
955         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
956         struct ipv6_tel_txoption opt;
957         struct dst_entry *dst = NULL, *ndst = NULL;
958         struct net_device *tdev;
959         int mtu;
960         unsigned int max_headroom = sizeof(struct ipv6hdr);
961         u8 proto;
962         int err = -1;
963
964         if (!fl6->flowi6_mark)
965                 dst = ip6_tnl_dst_check(t);
966         if (!dst) {
967                 ndst = ip6_route_output(net, NULL, fl6);
968
969                 if (ndst->error)
970                         goto tx_err_link_failure;
971                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
972                 if (IS_ERR(ndst)) {
973                         err = PTR_ERR(ndst);
974                         ndst = NULL;
975                         goto tx_err_link_failure;
976                 }
977                 dst = ndst;
978         }
979
980         tdev = dst->dev;
981
982         if (tdev == dev) {
983                 stats->collisions++;
984                 net_warn_ratelimited("%s: Local routing loop detected!\n",
985                                      t->parms.name);
986                 goto tx_err_dst_release;
987         }
988         mtu = dst_mtu(dst) - sizeof (*ipv6h);
989         if (encap_limit >= 0) {
990                 max_headroom += 8;
991                 mtu -= 8;
992         }
993         if (mtu < IPV6_MIN_MTU)
994                 mtu = IPV6_MIN_MTU;
995         if (skb_dst(skb))
996                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
997         if (skb->len > mtu) {
998                 *pmtu = mtu;
999                 err = -EMSGSIZE;
1000                 goto tx_err_dst_release;
1001         }
1002
1003         if (!net_eq(t->net, dev_net(dev)))
1004                 skb_scrub_packet(skb);
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 (fl6->flowi6_mark) {
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(t->net,
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 = t->net;
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_uninit     = ip6_tnl_dev_uninit,
1443         .ndo_start_xmit = ip6_tnl_xmit,
1444         .ndo_do_ioctl   = ip6_tnl_ioctl,
1445         .ndo_change_mtu = ip6_tnl_change_mtu,
1446         .ndo_get_stats  = ip6_get_stats,
1447 };
1448
1449
1450 /**
1451  * ip6_tnl_dev_setup - setup virtual tunnel device
1452  *   @dev: virtual device associated with tunnel
1453  *
1454  * Description:
1455  *   Initialize function pointers and device parameters
1456  **/
1457
1458 static void ip6_tnl_dev_setup(struct net_device *dev)
1459 {
1460         struct ip6_tnl *t;
1461
1462         dev->netdev_ops = &ip6_tnl_netdev_ops;
1463         dev->destructor = ip6_dev_free;
1464
1465         dev->type = ARPHRD_TUNNEL6;
1466         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1467         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1468         t = netdev_priv(dev);
1469         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1470                 dev->mtu-=8;
1471         dev->flags |= IFF_NOARP;
1472         dev->addr_len = sizeof(struct in6_addr);
1473         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1474 }
1475
1476
1477 /**
1478  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1479  *   @dev: virtual device associated with tunnel
1480  **/
1481
1482 static inline int
1483 ip6_tnl_dev_init_gen(struct net_device *dev)
1484 {
1485         struct ip6_tnl *t = netdev_priv(dev);
1486
1487         t->dev = dev;
1488         t->net = dev_net(dev);
1489         dev->tstats = alloc_percpu(struct pcpu_tstats);
1490         if (!dev->tstats)
1491                 return -ENOMEM;
1492         return 0;
1493 }
1494
1495 /**
1496  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1497  *   @dev: virtual device associated with tunnel
1498  **/
1499
1500 static int ip6_tnl_dev_init(struct net_device *dev)
1501 {
1502         struct ip6_tnl *t = netdev_priv(dev);
1503         int err = ip6_tnl_dev_init_gen(dev);
1504
1505         if (err)
1506                 return err;
1507         ip6_tnl_link_config(t);
1508         return 0;
1509 }
1510
1511 /**
1512  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1513  *   @dev: fallback device
1514  *
1515  * Return: 0
1516  **/
1517
1518 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1519 {
1520         struct ip6_tnl *t = netdev_priv(dev);
1521         struct net *net = dev_net(dev);
1522         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1523         int err = ip6_tnl_dev_init_gen(dev);
1524
1525         if (err)
1526                 return err;
1527
1528         t->parms.proto = IPPROTO_IPV6;
1529         dev_hold(dev);
1530
1531         ip6_tnl_link_config(t);
1532
1533         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1534         return 0;
1535 }
1536
1537 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1538 {
1539         u8 proto;
1540
1541         if (!data)
1542                 return 0;
1543
1544         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1545         if (proto != IPPROTO_IPV6 &&
1546             proto != IPPROTO_IPIP &&
1547             proto != 0)
1548                 return -EINVAL;
1549
1550         return 0;
1551 }
1552
1553 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1554                                   struct __ip6_tnl_parm *parms)
1555 {
1556         memset(parms, 0, sizeof(*parms));
1557
1558         if (!data)
1559                 return;
1560
1561         if (data[IFLA_IPTUN_LINK])
1562                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1563
1564         if (data[IFLA_IPTUN_LOCAL])
1565                 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1566                            sizeof(struct in6_addr));
1567
1568         if (data[IFLA_IPTUN_REMOTE])
1569                 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1570                            sizeof(struct in6_addr));
1571
1572         if (data[IFLA_IPTUN_TTL])
1573                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1574
1575         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1576                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1577
1578         if (data[IFLA_IPTUN_FLOWINFO])
1579                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1580
1581         if (data[IFLA_IPTUN_FLAGS])
1582                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1583
1584         if (data[IFLA_IPTUN_PROTO])
1585                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1586 }
1587
1588 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1589                            struct nlattr *tb[], struct nlattr *data[])
1590 {
1591         struct net *net = dev_net(dev);
1592         struct ip6_tnl *nt;
1593
1594         nt = netdev_priv(dev);
1595         ip6_tnl_netlink_parms(data, &nt->parms);
1596
1597         if (ip6_tnl_locate(net, &nt->parms, 0))
1598                 return -EEXIST;
1599
1600         return ip6_tnl_create2(dev);
1601 }
1602
1603 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1604                               struct nlattr *data[])
1605 {
1606         struct ip6_tnl *t = netdev_priv(dev);
1607         struct __ip6_tnl_parm p;
1608         struct net *net = t->net;
1609         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1610
1611         if (dev == ip6n->fb_tnl_dev)
1612                 return -EINVAL;
1613
1614         ip6_tnl_netlink_parms(data, &p);
1615
1616         t = ip6_tnl_locate(net, &p, 0);
1617
1618         if (t) {
1619                 if (t->dev != dev)
1620                         return -EEXIST;
1621         } else
1622                 t = netdev_priv(dev);
1623
1624         return ip6_tnl_update(t, &p);
1625 }
1626
1627 static size_t ip6_tnl_get_size(const struct net_device *dev)
1628 {
1629         return
1630                 /* IFLA_IPTUN_LINK */
1631                 nla_total_size(4) +
1632                 /* IFLA_IPTUN_LOCAL */
1633                 nla_total_size(sizeof(struct in6_addr)) +
1634                 /* IFLA_IPTUN_REMOTE */
1635                 nla_total_size(sizeof(struct in6_addr)) +
1636                 /* IFLA_IPTUN_TTL */
1637                 nla_total_size(1) +
1638                 /* IFLA_IPTUN_ENCAP_LIMIT */
1639                 nla_total_size(1) +
1640                 /* IFLA_IPTUN_FLOWINFO */
1641                 nla_total_size(4) +
1642                 /* IFLA_IPTUN_FLAGS */
1643                 nla_total_size(4) +
1644                 /* IFLA_IPTUN_PROTO */
1645                 nla_total_size(1) +
1646                 0;
1647 }
1648
1649 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1650 {
1651         struct ip6_tnl *tunnel = netdev_priv(dev);
1652         struct __ip6_tnl_parm *parm = &tunnel->parms;
1653
1654         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1655             nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1656                     &parm->raddr) ||
1657             nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1658                     &parm->laddr) ||
1659             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1660             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1661             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1662             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1663             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1664                 goto nla_put_failure;
1665         return 0;
1666
1667 nla_put_failure:
1668         return -EMSGSIZE;
1669 }
1670
1671 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1672         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1673         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1674         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1675         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1676         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1677         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1678         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1679         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1680 };
1681
1682 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1683         .kind           = "ip6tnl",
1684         .maxtype        = IFLA_IPTUN_MAX,
1685         .policy         = ip6_tnl_policy,
1686         .priv_size      = sizeof(struct ip6_tnl),
1687         .setup          = ip6_tnl_dev_setup,
1688         .validate       = ip6_tnl_validate,
1689         .newlink        = ip6_tnl_newlink,
1690         .changelink     = ip6_tnl_changelink,
1691         .get_size       = ip6_tnl_get_size,
1692         .fill_info      = ip6_tnl_fill_info,
1693 };
1694
1695 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1696         .handler        = ip4ip6_rcv,
1697         .err_handler    = ip4ip6_err,
1698         .priority       =       1,
1699 };
1700
1701 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1702         .handler        = ip6ip6_rcv,
1703         .err_handler    = ip6ip6_err,
1704         .priority       =       1,
1705 };
1706
1707 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1708 {
1709         struct net *net = dev_net(ip6n->fb_tnl_dev);
1710         struct net_device *dev, *aux;
1711         int h;
1712         struct ip6_tnl *t;
1713         LIST_HEAD(list);
1714
1715         for_each_netdev_safe(net, dev, aux)
1716                 if (dev->rtnl_link_ops == &ip6_link_ops)
1717                         unregister_netdevice_queue(dev, &list);
1718
1719         for (h = 0; h < HASH_SIZE; h++) {
1720                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1721                 while (t != NULL) {
1722                         /* If dev is in the same netns, it has already
1723                          * been added to the list by the previous loop.
1724                          */
1725                         if (!net_eq(dev_net(t->dev), net))
1726                                 unregister_netdevice_queue(t->dev, &list);
1727                         t = rtnl_dereference(t->next);
1728                 }
1729         }
1730
1731         t = rtnl_dereference(ip6n->tnls_wc[0]);
1732         unregister_netdevice_queue(t->dev, &list);
1733         unregister_netdevice_many(&list);
1734 }
1735
1736 static int __net_init ip6_tnl_init_net(struct net *net)
1737 {
1738         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1739         struct ip6_tnl *t = NULL;
1740         int err;
1741
1742         ip6n->tnls[0] = ip6n->tnls_wc;
1743         ip6n->tnls[1] = ip6n->tnls_r_l;
1744
1745         err = -ENOMEM;
1746         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1747                                       ip6_tnl_dev_setup);
1748
1749         if (!ip6n->fb_tnl_dev)
1750                 goto err_alloc_dev;
1751         dev_net_set(ip6n->fb_tnl_dev, net);
1752         /* FB netdevice is special: we have one, and only one per netns.
1753          * Allowing to move it to another netns is clearly unsafe.
1754          */
1755         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1756
1757         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1758         if (err < 0)
1759                 goto err_register;
1760
1761         err = register_netdev(ip6n->fb_tnl_dev);
1762         if (err < 0)
1763                 goto err_register;
1764
1765         t = netdev_priv(ip6n->fb_tnl_dev);
1766
1767         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1768         return 0;
1769
1770 err_register:
1771         ip6_dev_free(ip6n->fb_tnl_dev);
1772 err_alloc_dev:
1773         return err;
1774 }
1775
1776 static void __net_exit ip6_tnl_exit_net(struct net *net)
1777 {
1778         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1779
1780         rtnl_lock();
1781         ip6_tnl_destroy_tunnels(ip6n);
1782         rtnl_unlock();
1783 }
1784
1785 static struct pernet_operations ip6_tnl_net_ops = {
1786         .init = ip6_tnl_init_net,
1787         .exit = ip6_tnl_exit_net,
1788         .id   = &ip6_tnl_net_id,
1789         .size = sizeof(struct ip6_tnl_net),
1790 };
1791
1792 /**
1793  * ip6_tunnel_init - register protocol and reserve needed resources
1794  *
1795  * Return: 0 on success
1796  **/
1797
1798 static int __init ip6_tunnel_init(void)
1799 {
1800         int  err;
1801
1802         err = register_pernet_device(&ip6_tnl_net_ops);
1803         if (err < 0)
1804                 goto out_pernet;
1805
1806         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1807         if (err < 0) {
1808                 pr_err("%s: can't register ip4ip6\n", __func__);
1809                 goto out_ip4ip6;
1810         }
1811
1812         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1813         if (err < 0) {
1814                 pr_err("%s: can't register ip6ip6\n", __func__);
1815                 goto out_ip6ip6;
1816         }
1817         err = rtnl_link_register(&ip6_link_ops);
1818         if (err < 0)
1819                 goto rtnl_link_failed;
1820
1821         return 0;
1822
1823 rtnl_link_failed:
1824         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1825 out_ip6ip6:
1826         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1827 out_ip4ip6:
1828         unregister_pernet_device(&ip6_tnl_net_ops);
1829 out_pernet:
1830         return err;
1831 }
1832
1833 /**
1834  * ip6_tunnel_cleanup - free resources and unregister protocol
1835  **/
1836
1837 static void __exit ip6_tunnel_cleanup(void)
1838 {
1839         rtnl_link_unregister(&ip6_link_ops);
1840         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1841                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1842
1843         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1844                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1845
1846         unregister_pernet_device(&ip6_tnl_net_ops);
1847 }
1848
1849 module_init(ip6_tunnel_init);
1850 module_exit(ip6_tunnel_cleanup);