Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[linux-drm-fsl-dcu.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 #include <linux/etherdevice.h>
45
46 #include <asm/uaccess.h>
47 #include <linux/atomic.h>
48
49 #include <net/icmp.h>
50 #include <net/ip.h>
51 #include <net/ip_tunnels.h>
52 #include <net/ipv6.h>
53 #include <net/ip6_route.h>
54 #include <net/addrconf.h>
55 #include <net/ip6_tunnel.h>
56 #include <net/xfrm.h>
57 #include <net/dsfield.h>
58 #include <net/inet_ecn.h>
59 #include <net/net_namespace.h>
60 #include <net/netns/generic.h>
61
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
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_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                 skb->mac_header = skb->network_header;
805                 skb_reset_network_header(skb);
806                 skb->protocol = htons(protocol);
807                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
808
809                 __skb_tunnel_rx(skb, t->dev, t->net);
810
811                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
812                 if (unlikely(err)) {
813                         if (log_ecn_error)
814                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
815                                                      &ipv6h->saddr,
816                                                      ipv6_get_dsfield(ipv6h));
817                         if (err > 1) {
818                                 ++t->dev->stats.rx_frame_errors;
819                                 ++t->dev->stats.rx_errors;
820                                 rcu_read_unlock();
821                                 goto discard;
822                         }
823                 }
824
825                 tstats = this_cpu_ptr(t->dev->tstats);
826                 tstats->rx_packets++;
827                 tstats->rx_bytes += skb->len;
828
829                 netif_rx(skb);
830
831                 rcu_read_unlock();
832                 return 0;
833         }
834         rcu_read_unlock();
835         return 1;
836
837 discard:
838         kfree_skb(skb);
839         return 0;
840 }
841
842 static int ip4ip6_rcv(struct sk_buff *skb)
843 {
844         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
845                            ip4ip6_dscp_ecn_decapsulate);
846 }
847
848 static int ip6ip6_rcv(struct sk_buff *skb)
849 {
850         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
851                            ip6ip6_dscp_ecn_decapsulate);
852 }
853
854 struct ipv6_tel_txoption {
855         struct ipv6_txoptions ops;
856         __u8 dst_opt[8];
857 };
858
859 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
860 {
861         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
862
863         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
864         opt->dst_opt[3] = 1;
865         opt->dst_opt[4] = encap_limit;
866         opt->dst_opt[5] = IPV6_TLV_PADN;
867         opt->dst_opt[6] = 1;
868
869         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
870         opt->ops.opt_nflen = 8;
871 }
872
873 /**
874  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
875  *   @t: the outgoing tunnel device
876  *   @hdr: IPv6 header from the incoming packet
877  *
878  * Description:
879  *   Avoid trivial tunneling loop by checking that tunnel exit-point
880  *   doesn't match source of incoming packet.
881  *
882  * Return:
883  *   1 if conflict,
884  *   0 else
885  **/
886
887 static inline bool
888 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
889 {
890         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
891 }
892
893 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
894 {
895         struct __ip6_tnl_parm *p = &t->parms;
896         int ret = 0;
897         struct net *net = t->net;
898
899         if (p->flags & IP6_TNL_F_CAP_XMIT) {
900                 struct net_device *ldev = NULL;
901
902                 rcu_read_lock();
903                 if (p->link)
904                         ldev = dev_get_by_index_rcu(net, p->link);
905
906                 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
907                         pr_warn("%s xmit: Local address not yet configured!\n",
908                                 p->name);
909                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
910                          unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
911                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
912                                 p->name);
913                 else
914                         ret = 1;
915                 rcu_read_unlock();
916         }
917         return ret;
918 }
919 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
920
921 /**
922  * ip6_tnl_xmit2 - encapsulate packet and send
923  *   @skb: the outgoing socket buffer
924  *   @dev: the outgoing tunnel device
925  *   @dsfield: dscp code for outer header
926  *   @fl: flow of tunneled packet
927  *   @encap_limit: encapsulation limit
928  *   @pmtu: Path MTU is stored if packet is too big
929  *
930  * Description:
931  *   Build new header and do some sanity checks on the packet before sending
932  *   it.
933  *
934  * Return:
935  *   0 on success
936  *   -1 fail
937  *   %-EMSGSIZE message too big. return mtu in this case.
938  **/
939
940 static int ip6_tnl_xmit2(struct sk_buff *skb,
941                          struct net_device *dev,
942                          __u8 dsfield,
943                          struct flowi6 *fl6,
944                          int encap_limit,
945                          __u32 *pmtu)
946 {
947         struct ip6_tnl *t = netdev_priv(dev);
948         struct net *net = t->net;
949         struct net_device_stats *stats = &t->dev->stats;
950         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
951         struct ipv6_tel_txoption opt;
952         struct dst_entry *dst = NULL, *ndst = NULL;
953         struct net_device *tdev;
954         int mtu;
955         unsigned int max_headroom = sizeof(struct ipv6hdr);
956         u8 proto;
957         int err = -1;
958
959         if (!fl6->flowi6_mark)
960                 dst = ip6_tnl_dst_check(t);
961         if (!dst) {
962                 ndst = ip6_route_output(net, NULL, fl6);
963
964                 if (ndst->error)
965                         goto tx_err_link_failure;
966                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
967                 if (IS_ERR(ndst)) {
968                         err = PTR_ERR(ndst);
969                         ndst = NULL;
970                         goto tx_err_link_failure;
971                 }
972                 dst = ndst;
973         }
974
975         tdev = dst->dev;
976
977         if (tdev == dev) {
978                 stats->collisions++;
979                 net_warn_ratelimited("%s: Local routing loop detected!\n",
980                                      t->parms.name);
981                 goto tx_err_dst_release;
982         }
983         mtu = dst_mtu(dst) - sizeof (*ipv6h);
984         if (encap_limit >= 0) {
985                 max_headroom += 8;
986                 mtu -= 8;
987         }
988         if (mtu < IPV6_MIN_MTU)
989                 mtu = IPV6_MIN_MTU;
990         if (skb_dst(skb))
991                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
992         if (skb->len > mtu) {
993                 *pmtu = mtu;
994                 err = -EMSGSIZE;
995                 goto tx_err_dst_release;
996         }
997
998         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
999
1000         /*
1001          * Okay, now see if we can stuff it in the buffer as-is.
1002          */
1003         max_headroom += LL_RESERVED_SPACE(tdev);
1004
1005         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1006             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1007                 struct sk_buff *new_skb;
1008
1009                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1010                         goto tx_err_dst_release;
1011
1012                 if (skb->sk)
1013                         skb_set_owner_w(new_skb, skb->sk);
1014                 consume_skb(skb);
1015                 skb = new_skb;
1016         }
1017         if (fl6->flowi6_mark) {
1018                 skb_dst_set(skb, dst);
1019                 ndst = NULL;
1020         } else {
1021                 skb_dst_set_noref(skb, dst);
1022         }
1023         skb->transport_header = skb->network_header;
1024
1025         proto = fl6->flowi6_proto;
1026         if (encap_limit >= 0) {
1027                 init_tel_txopt(&opt, encap_limit);
1028                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1029         }
1030
1031         if (likely(!skb->encapsulation)) {
1032                 skb_reset_inner_headers(skb);
1033                 skb->encapsulation = 1;
1034         }
1035
1036         skb_push(skb, sizeof(struct ipv6hdr));
1037         skb_reset_network_header(skb);
1038         ipv6h = ipv6_hdr(skb);
1039         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
1040         ipv6h->hop_limit = t->parms.hop_limit;
1041         ipv6h->nexthdr = proto;
1042         ipv6h->saddr = fl6->saddr;
1043         ipv6h->daddr = fl6->daddr;
1044         ip6tunnel_xmit(skb, dev);
1045         if (ndst)
1046                 ip6_tnl_dst_store(t, ndst);
1047         return 0;
1048 tx_err_link_failure:
1049         stats->tx_carrier_errors++;
1050         dst_link_failure(skb);
1051 tx_err_dst_release:
1052         dst_release(ndst);
1053         return err;
1054 }
1055
1056 static inline int
1057 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1058 {
1059         struct ip6_tnl *t = netdev_priv(dev);
1060         const struct iphdr  *iph = ip_hdr(skb);
1061         int encap_limit = -1;
1062         struct flowi6 fl6;
1063         __u8 dsfield;
1064         __u32 mtu;
1065         int err;
1066
1067         if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1068             !ip6_tnl_xmit_ctl(t))
1069                 return -1;
1070
1071         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1072                 encap_limit = t->parms.encap_limit;
1073
1074         memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1075         fl6.flowi6_proto = IPPROTO_IPIP;
1076
1077         dsfield = ipv4_get_dsfield(iph);
1078
1079         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1080                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1081                                           & IPV6_TCLASS_MASK;
1082         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1083                 fl6.flowi6_mark = skb->mark;
1084
1085         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1086         if (err != 0) {
1087                 /* XXX: send ICMP error even if DF is not set. */
1088                 if (err == -EMSGSIZE)
1089                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1090                                   htonl(mtu));
1091                 return -1;
1092         }
1093
1094         return 0;
1095 }
1096
1097 static inline int
1098 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1099 {
1100         struct ip6_tnl *t = netdev_priv(dev);
1101         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1102         int encap_limit = -1;
1103         __u16 offset;
1104         struct flowi6 fl6;
1105         __u8 dsfield;
1106         __u32 mtu;
1107         int err;
1108
1109         if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1110             !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1111                 return -1;
1112
1113         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1114         if (offset > 0) {
1115                 struct ipv6_tlv_tnl_enc_lim *tel;
1116                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1117                 if (tel->encap_limit == 0) {
1118                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1119                                     ICMPV6_HDR_FIELD, offset + 2);
1120                         return -1;
1121                 }
1122                 encap_limit = tel->encap_limit - 1;
1123         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1124                 encap_limit = t->parms.encap_limit;
1125
1126         memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1127         fl6.flowi6_proto = IPPROTO_IPV6;
1128
1129         dsfield = ipv6_get_dsfield(ipv6h);
1130         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1131                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1132         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1133                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1134         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1135                 fl6.flowi6_mark = skb->mark;
1136
1137         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1138         if (err != 0) {
1139                 if (err == -EMSGSIZE)
1140                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1141                 return -1;
1142         }
1143
1144         return 0;
1145 }
1146
1147 static netdev_tx_t
1148 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1149 {
1150         struct ip6_tnl *t = netdev_priv(dev);
1151         struct net_device_stats *stats = &t->dev->stats;
1152         int ret;
1153
1154         switch (skb->protocol) {
1155         case htons(ETH_P_IP):
1156                 ret = ip4ip6_tnl_xmit(skb, dev);
1157                 break;
1158         case htons(ETH_P_IPV6):
1159                 ret = ip6ip6_tnl_xmit(skb, dev);
1160                 break;
1161         default:
1162                 goto tx_err;
1163         }
1164
1165         if (ret < 0)
1166                 goto tx_err;
1167
1168         return NETDEV_TX_OK;
1169
1170 tx_err:
1171         stats->tx_errors++;
1172         stats->tx_dropped++;
1173         kfree_skb(skb);
1174         return NETDEV_TX_OK;
1175 }
1176
1177 static void ip6_tnl_link_config(struct ip6_tnl *t)
1178 {
1179         struct net_device *dev = t->dev;
1180         struct __ip6_tnl_parm *p = &t->parms;
1181         struct flowi6 *fl6 = &t->fl.u.ip6;
1182
1183         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1184         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1185
1186         /* Set up flowi template */
1187         fl6->saddr = p->laddr;
1188         fl6->daddr = p->raddr;
1189         fl6->flowi6_oif = p->link;
1190         fl6->flowlabel = 0;
1191
1192         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1193                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1194         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1195                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1196
1197         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1198         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1199
1200         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1201                 dev->flags |= IFF_POINTOPOINT;
1202         else
1203                 dev->flags &= ~IFF_POINTOPOINT;
1204
1205         dev->iflink = p->link;
1206
1207         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1208                 int strict = (ipv6_addr_type(&p->raddr) &
1209                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1210
1211                 struct rt6_info *rt = rt6_lookup(t->net,
1212                                                  &p->raddr, &p->laddr,
1213                                                  p->link, strict);
1214
1215                 if (rt == NULL)
1216                         return;
1217
1218                 if (rt->dst.dev) {
1219                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1220                                 sizeof (struct ipv6hdr);
1221
1222                         dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1223                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1224                                 dev->mtu-=8;
1225
1226                         if (dev->mtu < IPV6_MIN_MTU)
1227                                 dev->mtu = IPV6_MIN_MTU;
1228                 }
1229                 ip6_rt_put(rt);
1230         }
1231 }
1232
1233 /**
1234  * ip6_tnl_change - update the tunnel parameters
1235  *   @t: tunnel to be changed
1236  *   @p: tunnel configuration parameters
1237  *
1238  * Description:
1239  *   ip6_tnl_change() updates the tunnel parameters
1240  **/
1241
1242 static int
1243 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1244 {
1245         t->parms.laddr = p->laddr;
1246         t->parms.raddr = p->raddr;
1247         t->parms.flags = p->flags;
1248         t->parms.hop_limit = p->hop_limit;
1249         t->parms.encap_limit = p->encap_limit;
1250         t->parms.flowinfo = p->flowinfo;
1251         t->parms.link = p->link;
1252         t->parms.proto = p->proto;
1253         ip6_tnl_dst_reset(t);
1254         ip6_tnl_link_config(t);
1255         return 0;
1256 }
1257
1258 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1259 {
1260         struct net *net = t->net;
1261         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1262         int err;
1263
1264         ip6_tnl_unlink(ip6n, t);
1265         synchronize_net();
1266         err = ip6_tnl_change(t, p);
1267         ip6_tnl_link(ip6n, t);
1268         netdev_state_change(t->dev);
1269         return err;
1270 }
1271
1272 static void
1273 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1274 {
1275         p->laddr = u->laddr;
1276         p->raddr = u->raddr;
1277         p->flags = u->flags;
1278         p->hop_limit = u->hop_limit;
1279         p->encap_limit = u->encap_limit;
1280         p->flowinfo = u->flowinfo;
1281         p->link = u->link;
1282         p->proto = u->proto;
1283         memcpy(p->name, u->name, sizeof(u->name));
1284 }
1285
1286 static void
1287 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1288 {
1289         u->laddr = p->laddr;
1290         u->raddr = p->raddr;
1291         u->flags = p->flags;
1292         u->hop_limit = p->hop_limit;
1293         u->encap_limit = p->encap_limit;
1294         u->flowinfo = p->flowinfo;
1295         u->link = p->link;
1296         u->proto = p->proto;
1297         memcpy(u->name, p->name, sizeof(u->name));
1298 }
1299
1300 /**
1301  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1302  *   @dev: virtual device associated with tunnel
1303  *   @ifr: parameters passed from userspace
1304  *   @cmd: command to be performed
1305  *
1306  * Description:
1307  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1308  *   from userspace.
1309  *
1310  *   The possible commands are the following:
1311  *     %SIOCGETTUNNEL: get tunnel parameters for device
1312  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1313  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1314  *     %SIOCDELTUNNEL: delete tunnel
1315  *
1316  *   The fallback device "ip6tnl0", created during module
1317  *   initialization, can be used for creating other tunnel devices.
1318  *
1319  * Return:
1320  *   0 on success,
1321  *   %-EFAULT if unable to copy data to or from userspace,
1322  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1323  *   %-EINVAL if passed tunnel parameters are invalid,
1324  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1325  *   %-ENODEV if attempting to change or delete a nonexisting device
1326  **/
1327
1328 static int
1329 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1330 {
1331         int err = 0;
1332         struct ip6_tnl_parm p;
1333         struct __ip6_tnl_parm p1;
1334         struct ip6_tnl *t = NULL;
1335         struct net *net = dev_net(dev);
1336         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1337
1338         switch (cmd) {
1339         case SIOCGETTUNNEL:
1340                 if (dev == ip6n->fb_tnl_dev) {
1341                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1342                                 err = -EFAULT;
1343                                 break;
1344                         }
1345                         ip6_tnl_parm_from_user(&p1, &p);
1346                         t = ip6_tnl_locate(net, &p1, 0);
1347                 } else {
1348                         memset(&p, 0, sizeof(p));
1349                 }
1350                 if (t == NULL)
1351                         t = netdev_priv(dev);
1352                 ip6_tnl_parm_to_user(&p, &t->parms);
1353                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1354                         err = -EFAULT;
1355                 }
1356                 break;
1357         case SIOCADDTUNNEL:
1358         case SIOCCHGTUNNEL:
1359                 err = -EPERM;
1360                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1361                         break;
1362                 err = -EFAULT;
1363                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1364                         break;
1365                 err = -EINVAL;
1366                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1367                     p.proto != 0)
1368                         break;
1369                 ip6_tnl_parm_from_user(&p1, &p);
1370                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1371                 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1372                         if (t != NULL) {
1373                                 if (t->dev != dev) {
1374                                         err = -EEXIST;
1375                                         break;
1376                                 }
1377                         } else
1378                                 t = netdev_priv(dev);
1379
1380                         err = ip6_tnl_update(t, &p1);
1381                 }
1382                 if (t) {
1383                         err = 0;
1384                         ip6_tnl_parm_to_user(&p, &t->parms);
1385                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1386                                 err = -EFAULT;
1387
1388                 } else
1389                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1390                 break;
1391         case SIOCDELTUNNEL:
1392                 err = -EPERM;
1393                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1394                         break;
1395
1396                 if (dev == ip6n->fb_tnl_dev) {
1397                         err = -EFAULT;
1398                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1399                                 break;
1400                         err = -ENOENT;
1401                         ip6_tnl_parm_from_user(&p1, &p);
1402                         t = ip6_tnl_locate(net, &p1, 0);
1403                         if (t == NULL)
1404                                 break;
1405                         err = -EPERM;
1406                         if (t->dev == ip6n->fb_tnl_dev)
1407                                 break;
1408                         dev = t->dev;
1409                 }
1410                 err = 0;
1411                 unregister_netdevice(dev);
1412                 break;
1413         default:
1414                 err = -EINVAL;
1415         }
1416         return err;
1417 }
1418
1419 /**
1420  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1421  *   @dev: virtual device associated with tunnel
1422  *   @new_mtu: the new mtu
1423  *
1424  * Return:
1425  *   0 on success,
1426  *   %-EINVAL if mtu too small
1427  **/
1428
1429 static int
1430 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1431 {
1432         struct ip6_tnl *tnl = netdev_priv(dev);
1433
1434         if (tnl->parms.proto == IPPROTO_IPIP) {
1435                 if (new_mtu < 68)
1436                         return -EINVAL;
1437         } else {
1438                 if (new_mtu < IPV6_MIN_MTU)
1439                         return -EINVAL;
1440         }
1441         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1442                 return -EINVAL;
1443         dev->mtu = new_mtu;
1444         return 0;
1445 }
1446
1447
1448 static const struct net_device_ops ip6_tnl_netdev_ops = {
1449         .ndo_uninit     = ip6_tnl_dev_uninit,
1450         .ndo_start_xmit = ip6_tnl_xmit,
1451         .ndo_do_ioctl   = ip6_tnl_ioctl,
1452         .ndo_change_mtu = ip6_tnl_change_mtu,
1453         .ndo_get_stats  = ip6_get_stats,
1454 };
1455
1456
1457 /**
1458  * ip6_tnl_dev_setup - setup virtual tunnel device
1459  *   @dev: virtual device associated with tunnel
1460  *
1461  * Description:
1462  *   Initialize function pointers and device parameters
1463  **/
1464
1465 static void ip6_tnl_dev_setup(struct net_device *dev)
1466 {
1467         struct ip6_tnl *t;
1468
1469         dev->netdev_ops = &ip6_tnl_netdev_ops;
1470         dev->destructor = ip6_dev_free;
1471
1472         dev->type = ARPHRD_TUNNEL6;
1473         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1474         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1475         t = netdev_priv(dev);
1476         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1477                 dev->mtu-=8;
1478         dev->flags |= IFF_NOARP;
1479         dev->addr_len = sizeof(struct in6_addr);
1480         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1481         /* This perm addr will be used as interface identifier by IPv6 */
1482         dev->addr_assign_type = NET_ADDR_RANDOM;
1483         eth_random_addr(dev->perm_addr);
1484 }
1485
1486
1487 /**
1488  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1489  *   @dev: virtual device associated with tunnel
1490  **/
1491
1492 static inline int
1493 ip6_tnl_dev_init_gen(struct net_device *dev)
1494 {
1495         struct ip6_tnl *t = netdev_priv(dev);
1496         int i;
1497
1498         t->dev = dev;
1499         t->net = dev_net(dev);
1500         dev->tstats = alloc_percpu(struct pcpu_tstats);
1501         if (!dev->tstats)
1502                 return -ENOMEM;
1503
1504         for_each_possible_cpu(i) {
1505                 struct pcpu_tstats *ip6_tnl_stats;
1506                 ip6_tnl_stats = per_cpu_ptr(dev->tstats, i);
1507                 u64_stats_init(&ip6_tnl_stats->syncp);
1508         }
1509         return 0;
1510 }
1511
1512 /**
1513  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1514  *   @dev: virtual device associated with tunnel
1515  **/
1516
1517 static int ip6_tnl_dev_init(struct net_device *dev)
1518 {
1519         struct ip6_tnl *t = netdev_priv(dev);
1520         int err = ip6_tnl_dev_init_gen(dev);
1521
1522         if (err)
1523                 return err;
1524         ip6_tnl_link_config(t);
1525         return 0;
1526 }
1527
1528 /**
1529  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1530  *   @dev: fallback device
1531  *
1532  * Return: 0
1533  **/
1534
1535 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1536 {
1537         struct ip6_tnl *t = netdev_priv(dev);
1538         struct net *net = dev_net(dev);
1539         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1540         int err = ip6_tnl_dev_init_gen(dev);
1541
1542         if (err)
1543                 return err;
1544
1545         t->parms.proto = IPPROTO_IPV6;
1546         dev_hold(dev);
1547
1548         ip6_tnl_link_config(t);
1549
1550         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1551         return 0;
1552 }
1553
1554 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1555 {
1556         u8 proto;
1557
1558         if (!data)
1559                 return 0;
1560
1561         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1562         if (proto != IPPROTO_IPV6 &&
1563             proto != IPPROTO_IPIP &&
1564             proto != 0)
1565                 return -EINVAL;
1566
1567         return 0;
1568 }
1569
1570 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1571                                   struct __ip6_tnl_parm *parms)
1572 {
1573         memset(parms, 0, sizeof(*parms));
1574
1575         if (!data)
1576                 return;
1577
1578         if (data[IFLA_IPTUN_LINK])
1579                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1580
1581         if (data[IFLA_IPTUN_LOCAL])
1582                 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1583                            sizeof(struct in6_addr));
1584
1585         if (data[IFLA_IPTUN_REMOTE])
1586                 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1587                            sizeof(struct in6_addr));
1588
1589         if (data[IFLA_IPTUN_TTL])
1590                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1591
1592         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1593                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1594
1595         if (data[IFLA_IPTUN_FLOWINFO])
1596                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1597
1598         if (data[IFLA_IPTUN_FLAGS])
1599                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1600
1601         if (data[IFLA_IPTUN_PROTO])
1602                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1603 }
1604
1605 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1606                            struct nlattr *tb[], struct nlattr *data[])
1607 {
1608         struct net *net = dev_net(dev);
1609         struct ip6_tnl *nt;
1610
1611         nt = netdev_priv(dev);
1612         ip6_tnl_netlink_parms(data, &nt->parms);
1613
1614         if (ip6_tnl_locate(net, &nt->parms, 0))
1615                 return -EEXIST;
1616
1617         return ip6_tnl_create2(dev);
1618 }
1619
1620 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1621                               struct nlattr *data[])
1622 {
1623         struct ip6_tnl *t = netdev_priv(dev);
1624         struct __ip6_tnl_parm p;
1625         struct net *net = t->net;
1626         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1627
1628         if (dev == ip6n->fb_tnl_dev)
1629                 return -EINVAL;
1630
1631         ip6_tnl_netlink_parms(data, &p);
1632
1633         t = ip6_tnl_locate(net, &p, 0);
1634
1635         if (t) {
1636                 if (t->dev != dev)
1637                         return -EEXIST;
1638         } else
1639                 t = netdev_priv(dev);
1640
1641         return ip6_tnl_update(t, &p);
1642 }
1643
1644 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1645 {
1646         struct net *net = dev_net(dev);
1647         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1648
1649         if (dev != ip6n->fb_tnl_dev)
1650                 unregister_netdevice_queue(dev, head);
1651 }
1652
1653 static size_t ip6_tnl_get_size(const struct net_device *dev)
1654 {
1655         return
1656                 /* IFLA_IPTUN_LINK */
1657                 nla_total_size(4) +
1658                 /* IFLA_IPTUN_LOCAL */
1659                 nla_total_size(sizeof(struct in6_addr)) +
1660                 /* IFLA_IPTUN_REMOTE */
1661                 nla_total_size(sizeof(struct in6_addr)) +
1662                 /* IFLA_IPTUN_TTL */
1663                 nla_total_size(1) +
1664                 /* IFLA_IPTUN_ENCAP_LIMIT */
1665                 nla_total_size(1) +
1666                 /* IFLA_IPTUN_FLOWINFO */
1667                 nla_total_size(4) +
1668                 /* IFLA_IPTUN_FLAGS */
1669                 nla_total_size(4) +
1670                 /* IFLA_IPTUN_PROTO */
1671                 nla_total_size(1) +
1672                 0;
1673 }
1674
1675 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1676 {
1677         struct ip6_tnl *tunnel = netdev_priv(dev);
1678         struct __ip6_tnl_parm *parm = &tunnel->parms;
1679
1680         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1681             nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1682                     &parm->laddr) ||
1683             nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1684                     &parm->raddr) ||
1685             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1686             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1687             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1688             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1689             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1690                 goto nla_put_failure;
1691         return 0;
1692
1693 nla_put_failure:
1694         return -EMSGSIZE;
1695 }
1696
1697 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1698         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1699         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1700         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1701         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1702         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1703         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1704         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1705         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1706 };
1707
1708 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1709         .kind           = "ip6tnl",
1710         .maxtype        = IFLA_IPTUN_MAX,
1711         .policy         = ip6_tnl_policy,
1712         .priv_size      = sizeof(struct ip6_tnl),
1713         .setup          = ip6_tnl_dev_setup,
1714         .validate       = ip6_tnl_validate,
1715         .newlink        = ip6_tnl_newlink,
1716         .changelink     = ip6_tnl_changelink,
1717         .dellink        = ip6_tnl_dellink,
1718         .get_size       = ip6_tnl_get_size,
1719         .fill_info      = ip6_tnl_fill_info,
1720 };
1721
1722 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1723         .handler        = ip4ip6_rcv,
1724         .err_handler    = ip4ip6_err,
1725         .priority       =       1,
1726 };
1727
1728 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1729         .handler        = ip6ip6_rcv,
1730         .err_handler    = ip6ip6_err,
1731         .priority       =       1,
1732 };
1733
1734 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1735 {
1736         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1737         struct net_device *dev, *aux;
1738         int h;
1739         struct ip6_tnl *t;
1740         LIST_HEAD(list);
1741
1742         for_each_netdev_safe(net, dev, aux)
1743                 if (dev->rtnl_link_ops == &ip6_link_ops)
1744                         unregister_netdevice_queue(dev, &list);
1745
1746         for (h = 0; h < HASH_SIZE; h++) {
1747                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1748                 while (t != NULL) {
1749                         /* If dev is in the same netns, it has already
1750                          * been added to the list by the previous loop.
1751                          */
1752                         if (!net_eq(dev_net(t->dev), net))
1753                                 unregister_netdevice_queue(t->dev, &list);
1754                         t = rtnl_dereference(t->next);
1755                 }
1756         }
1757
1758         unregister_netdevice_many(&list);
1759 }
1760
1761 static int __net_init ip6_tnl_init_net(struct net *net)
1762 {
1763         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1764         struct ip6_tnl *t = NULL;
1765         int err;
1766
1767         ip6n->tnls[0] = ip6n->tnls_wc;
1768         ip6n->tnls[1] = ip6n->tnls_r_l;
1769
1770         err = -ENOMEM;
1771         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1772                                       ip6_tnl_dev_setup);
1773
1774         if (!ip6n->fb_tnl_dev)
1775                 goto err_alloc_dev;
1776         dev_net_set(ip6n->fb_tnl_dev, net);
1777         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1778         /* FB netdevice is special: we have one, and only one per netns.
1779          * Allowing to move it to another netns is clearly unsafe.
1780          */
1781         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1782
1783         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1784         if (err < 0)
1785                 goto err_register;
1786
1787         err = register_netdev(ip6n->fb_tnl_dev);
1788         if (err < 0)
1789                 goto err_register;
1790
1791         t = netdev_priv(ip6n->fb_tnl_dev);
1792
1793         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1794         return 0;
1795
1796 err_register:
1797         ip6_dev_free(ip6n->fb_tnl_dev);
1798 err_alloc_dev:
1799         return err;
1800 }
1801
1802 static void __net_exit ip6_tnl_exit_net(struct net *net)
1803 {
1804         rtnl_lock();
1805         ip6_tnl_destroy_tunnels(net);
1806         rtnl_unlock();
1807 }
1808
1809 static struct pernet_operations ip6_tnl_net_ops = {
1810         .init = ip6_tnl_init_net,
1811         .exit = ip6_tnl_exit_net,
1812         .id   = &ip6_tnl_net_id,
1813         .size = sizeof(struct ip6_tnl_net),
1814 };
1815
1816 /**
1817  * ip6_tunnel_init - register protocol and reserve needed resources
1818  *
1819  * Return: 0 on success
1820  **/
1821
1822 static int __init ip6_tunnel_init(void)
1823 {
1824         int  err;
1825
1826         err = register_pernet_device(&ip6_tnl_net_ops);
1827         if (err < 0)
1828                 goto out_pernet;
1829
1830         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1831         if (err < 0) {
1832                 pr_err("%s: can't register ip4ip6\n", __func__);
1833                 goto out_ip4ip6;
1834         }
1835
1836         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1837         if (err < 0) {
1838                 pr_err("%s: can't register ip6ip6\n", __func__);
1839                 goto out_ip6ip6;
1840         }
1841         err = rtnl_link_register(&ip6_link_ops);
1842         if (err < 0)
1843                 goto rtnl_link_failed;
1844
1845         return 0;
1846
1847 rtnl_link_failed:
1848         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1849 out_ip6ip6:
1850         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1851 out_ip4ip6:
1852         unregister_pernet_device(&ip6_tnl_net_ops);
1853 out_pernet:
1854         return err;
1855 }
1856
1857 /**
1858  * ip6_tunnel_cleanup - free resources and unregister protocol
1859  **/
1860
1861 static void __exit ip6_tunnel_cleanup(void)
1862 {
1863         rtnl_link_unregister(&ip6_link_ops);
1864         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1865                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1866
1867         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1868                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1869
1870         unregister_pernet_device(&ip6_tnl_net_ops);
1871 }
1872
1873 module_init(ip6_tunnel_init);
1874 module_exit(ip6_tunnel_cleanup);