Merge tag 'net-5.12-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
[platform/kernel/linux-rpi.git] / net / ipv6 / ip6_tunnel.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      IPv6 tunneling device
4  *      Linux INET6 implementation
5  *
6  *      Authors:
7  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
8  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
9  *
10  *      Based on:
11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
12  *
13  *      RFC 2473
14  */
15
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 #include <linux/etherdevice.h>
39
40 #include <linux/uaccess.h>
41 #include <linux/atomic.h>
42
43 #include <net/icmp.h>
44 #include <net/ip.h>
45 #include <net/ip_tunnels.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 #include <net/dst_metadata.h>
56
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
61 MODULE_ALIAS_NETDEV("ip6tnl0");
62
63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
65
66 static bool log_ecn_error = true;
67 module_param(log_ecn_error, bool, 0644);
68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
69
70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
71 {
72         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
73
74         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
75 }
76
77 static int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
80
81 static unsigned int ip6_tnl_net_id __read_mostly;
82 struct ip6_tnl_net {
83         /* the IPv6 tunnel fallback device */
84         struct net_device *fb_tnl_dev;
85         /* lists for storing tunnels in use */
86         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
87         struct ip6_tnl __rcu *tnls_wc[1];
88         struct ip6_tnl __rcu **tnls[2];
89         struct ip6_tnl __rcu *collect_md_tun;
90 };
91
92 static inline int ip6_tnl_mpls_supported(void)
93 {
94         return IS_ENABLED(CONFIG_MPLS);
95 }
96
97 #define for_each_ip6_tunnel_rcu(start) \
98         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
99
100 /**
101  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
102  *   @net: network namespace
103  *   @link: ifindex of underlying interface
104  *   @remote: the address of the tunnel exit-point
105  *   @local: the address of the tunnel entry-point
106  *
107  * Return:
108  *   tunnel matching given end-points if found,
109  *   else fallback tunnel if its device is up,
110  *   else %NULL
111  **/
112
113 static struct ip6_tnl *
114 ip6_tnl_lookup(struct net *net, int link,
115                const struct in6_addr *remote, const struct in6_addr *local)
116 {
117         unsigned int hash = HASH(remote, local);
118         struct ip6_tnl *t, *cand = NULL;
119         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
120         struct in6_addr any;
121
122         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
123                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
124                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
125                     !(t->dev->flags & IFF_UP))
126                         continue;
127
128                 if (link == t->parms.link)
129                         return t;
130                 else
131                         cand = t;
132         }
133
134         memset(&any, 0, sizeof(any));
135         hash = HASH(&any, local);
136         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
137                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138                     !ipv6_addr_any(&t->parms.raddr) ||
139                     !(t->dev->flags & IFF_UP))
140                         continue;
141
142                 if (link == t->parms.link)
143                         return t;
144                 else if (!cand)
145                         cand = t;
146         }
147
148         hash = HASH(remote, &any);
149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
151                     !ipv6_addr_any(&t->parms.laddr) ||
152                     !(t->dev->flags & IFF_UP))
153                         continue;
154
155                 if (link == t->parms.link)
156                         return t;
157                 else if (!cand)
158                         cand = t;
159         }
160
161         if (cand)
162                 return cand;
163
164         t = rcu_dereference(ip6n->collect_md_tun);
165         if (t && t->dev->flags & IFF_UP)
166                 return t;
167
168         t = rcu_dereference(ip6n->tnls_wc[0]);
169         if (t && (t->dev->flags & IFF_UP))
170                 return t;
171
172         return NULL;
173 }
174
175 /**
176  * ip6_tnl_bucket - get head of list matching given tunnel parameters
177  *   @ip6n: the private data for ip6_vti in the netns
178  *   @p: parameters containing tunnel end-points
179  *
180  * Description:
181  *   ip6_tnl_bucket() returns the head of the list matching the
182  *   &struct in6_addr entries laddr and raddr in @p.
183  *
184  * Return: head of IPv6 tunnel list
185  **/
186
187 static struct ip6_tnl __rcu **
188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
189 {
190         const struct in6_addr *remote = &p->raddr;
191         const struct in6_addr *local = &p->laddr;
192         unsigned int h = 0;
193         int prio = 0;
194
195         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196                 prio = 1;
197                 h = HASH(remote, local);
198         }
199         return &ip6n->tnls[prio][h];
200 }
201
202 /**
203  * ip6_tnl_link - add tunnel to hash table
204  *   @ip6n: the private data for ip6_vti in the netns
205  *   @t: tunnel to be added
206  **/
207
208 static void
209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 {
211         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
212
213         if (t->parms.collect_md)
214                 rcu_assign_pointer(ip6n->collect_md_tun, t);
215         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
216         rcu_assign_pointer(*tp, t);
217 }
218
219 /**
220  * ip6_tnl_unlink - remove tunnel from hash table
221  *   @ip6n: the private data for ip6_vti in the netns
222  *   @t: tunnel to be removed
223  **/
224
225 static void
226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
227 {
228         struct ip6_tnl __rcu **tp;
229         struct ip6_tnl *iter;
230
231         if (t->parms.collect_md)
232                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
233
234         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
235              (iter = rtnl_dereference(*tp)) != NULL;
236              tp = &iter->next) {
237                 if (t == iter) {
238                         rcu_assign_pointer(*tp, t->next);
239                         break;
240                 }
241         }
242 }
243
244 static void ip6_dev_free(struct net_device *dev)
245 {
246         struct ip6_tnl *t = netdev_priv(dev);
247
248         gro_cells_destroy(&t->gro_cells);
249         dst_cache_destroy(&t->dst_cache);
250         free_percpu(dev->tstats);
251 }
252
253 static int ip6_tnl_create2(struct net_device *dev)
254 {
255         struct ip6_tnl *t = netdev_priv(dev);
256         struct net *net = dev_net(dev);
257         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
258         int err;
259
260         t = netdev_priv(dev);
261
262         dev->rtnl_link_ops = &ip6_link_ops;
263         err = register_netdevice(dev);
264         if (err < 0)
265                 goto out;
266
267         strcpy(t->parms.name, dev->name);
268
269         dev_hold(dev);
270         ip6_tnl_link(ip6n, t);
271         return 0;
272
273 out:
274         return err;
275 }
276
277 /**
278  * ip6_tnl_create - create a new tunnel
279  *   @net: network namespace
280  *   @p: tunnel parameters
281  *
282  * Description:
283  *   Create tunnel matching given parameters.
284  *
285  * Return:
286  *   created tunnel or error pointer
287  **/
288
289 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
290 {
291         struct net_device *dev;
292         struct ip6_tnl *t;
293         char name[IFNAMSIZ];
294         int err = -E2BIG;
295
296         if (p->name[0]) {
297                 if (!dev_valid_name(p->name))
298                         goto failed;
299                 strlcpy(name, p->name, IFNAMSIZ);
300         } else {
301                 sprintf(name, "ip6tnl%%d");
302         }
303         err = -ENOMEM;
304         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
305                            ip6_tnl_dev_setup);
306         if (!dev)
307                 goto failed;
308
309         dev_net_set(dev, net);
310
311         t = netdev_priv(dev);
312         t->parms = *p;
313         t->net = dev_net(dev);
314         err = ip6_tnl_create2(dev);
315         if (err < 0)
316                 goto failed_free;
317
318         return t;
319
320 failed_free:
321         free_netdev(dev);
322 failed:
323         return ERR_PTR(err);
324 }
325
326 /**
327  * ip6_tnl_locate - find or create tunnel matching given parameters
328  *   @net: network namespace
329  *   @p: tunnel parameters
330  *   @create: != 0 if allowed to create new tunnel if no match found
331  *
332  * Description:
333  *   ip6_tnl_locate() first tries to locate an existing tunnel
334  *   based on @parms. If this is unsuccessful, but @create is set a new
335  *   tunnel device is created and registered for use.
336  *
337  * Return:
338  *   matching tunnel or error pointer
339  **/
340
341 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
342                 struct __ip6_tnl_parm *p, int create)
343 {
344         const struct in6_addr *remote = &p->raddr;
345         const struct in6_addr *local = &p->laddr;
346         struct ip6_tnl __rcu **tp;
347         struct ip6_tnl *t;
348         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
349
350         for (tp = ip6_tnl_bucket(ip6n, p);
351              (t = rtnl_dereference(*tp)) != NULL;
352              tp = &t->next) {
353                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
354                     ipv6_addr_equal(remote, &t->parms.raddr) &&
355                     p->link == t->parms.link) {
356                         if (create)
357                                 return ERR_PTR(-EEXIST);
358
359                         return t;
360                 }
361         }
362         if (!create)
363                 return ERR_PTR(-ENODEV);
364         return ip6_tnl_create(net, p);
365 }
366
367 /**
368  * ip6_tnl_dev_uninit - tunnel device uninitializer
369  *   @dev: the device to be destroyed
370  *
371  * Description:
372  *   ip6_tnl_dev_uninit() removes tunnel from its list
373  **/
374
375 static void
376 ip6_tnl_dev_uninit(struct net_device *dev)
377 {
378         struct ip6_tnl *t = netdev_priv(dev);
379         struct net *net = t->net;
380         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
381
382         if (dev == ip6n->fb_tnl_dev)
383                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
384         else
385                 ip6_tnl_unlink(ip6n, t);
386         dst_cache_reset(&t->dst_cache);
387         dev_put(dev);
388 }
389
390 /**
391  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
392  *   @skb: received socket buffer
393  *   @raw: the ICMPv6 error message data
394  *
395  * Return:
396  *   0 if none was found,
397  *   else index to encapsulation limit
398  **/
399
400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
401 {
402         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
403         unsigned int nhoff = raw - skb->data;
404         unsigned int off = nhoff + sizeof(*ipv6h);
405         u8 next, nexthdr = ipv6h->nexthdr;
406
407         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
408                 struct ipv6_opt_hdr *hdr;
409                 u16 optlen;
410
411                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
412                         break;
413
414                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
415                 if (nexthdr == NEXTHDR_FRAGMENT) {
416                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
417                         if (frag_hdr->frag_off)
418                                 break;
419                         optlen = 8;
420                 } else if (nexthdr == NEXTHDR_AUTH) {
421                         optlen = ipv6_authlen(hdr);
422                 } else {
423                         optlen = ipv6_optlen(hdr);
424                 }
425                 /* cache hdr->nexthdr, since pskb_may_pull() might
426                  * invalidate hdr
427                  */
428                 next = hdr->nexthdr;
429                 if (nexthdr == NEXTHDR_DEST) {
430                         u16 i = 2;
431
432                         /* Remember : hdr is no longer valid at this point. */
433                         if (!pskb_may_pull(skb, off + optlen))
434                                 break;
435
436                         while (1) {
437                                 struct ipv6_tlv_tnl_enc_lim *tel;
438
439                                 /* No more room for encapsulation limit */
440                                 if (i + sizeof(*tel) > optlen)
441                                         break;
442
443                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
444                                 /* return index of option if found and valid */
445                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
446                                     tel->length == 1)
447                                         return i + off - nhoff;
448                                 /* else jump to next option */
449                                 if (tel->type)
450                                         i += tel->length + 2;
451                                 else
452                                         i++;
453                         }
454                 }
455                 nexthdr = next;
456                 off += optlen;
457         }
458         return 0;
459 }
460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
461
462 /* ip6_tnl_err() should handle errors in the tunnel according to the
463  * specifications in RFC 2473.
464  */
465 static int
466 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
467             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
468 {
469         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
470         struct net *net = dev_net(skb->dev);
471         u8 rel_type = ICMPV6_DEST_UNREACH;
472         u8 rel_code = ICMPV6_ADDR_UNREACH;
473         __u32 rel_info = 0;
474         struct ip6_tnl *t;
475         int err = -ENOENT;
476         int rel_msg = 0;
477         u8 tproto;
478         __u16 len;
479
480         /* If the packet doesn't contain the original IPv6 header we are
481            in trouble since we might need the source address for further
482            processing of the error. */
483
484         rcu_read_lock();
485         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
486         if (!t)
487                 goto out;
488
489         tproto = READ_ONCE(t->parms.proto);
490         if (tproto != ipproto && tproto != 0)
491                 goto out;
492
493         err = 0;
494
495         switch (*type) {
496         case ICMPV6_DEST_UNREACH:
497                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
498                                     t->parms.name);
499                 rel_msg = 1;
500                 break;
501         case ICMPV6_TIME_EXCEED:
502                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
503                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
504                                             t->parms.name);
505                         rel_msg = 1;
506                 }
507                 break;
508         case ICMPV6_PARAMPROB: {
509                 struct ipv6_tlv_tnl_enc_lim *tel;
510                 __u32 teli;
511
512                 teli = 0;
513                 if ((*code) == ICMPV6_HDR_FIELD)
514                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
515
516                 if (teli && teli == *info - 2) {
517                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
518                         if (tel->encap_limit == 0) {
519                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
520                                                     t->parms.name);
521                                 rel_msg = 1;
522                         }
523                 } else {
524                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
525                                             t->parms.name);
526                 }
527                 break;
528         }
529         case ICMPV6_PKT_TOOBIG: {
530                 __u32 mtu;
531
532                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
533                                 sock_net_uid(net, NULL));
534                 mtu = *info - offset;
535                 if (mtu < IPV6_MIN_MTU)
536                         mtu = IPV6_MIN_MTU;
537                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
538                 if (len > mtu) {
539                         rel_type = ICMPV6_PKT_TOOBIG;
540                         rel_code = 0;
541                         rel_info = mtu;
542                         rel_msg = 1;
543                 }
544                 break;
545         }
546         case NDISC_REDIRECT:
547                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
548                              sock_net_uid(net, NULL));
549                 break;
550         }
551
552         *type = rel_type;
553         *code = rel_code;
554         *info = rel_info;
555         *msg = rel_msg;
556
557 out:
558         rcu_read_unlock();
559         return err;
560 }
561
562 static int
563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
564            u8 type, u8 code, int offset, __be32 info)
565 {
566         __u32 rel_info = ntohl(info);
567         const struct iphdr *eiph;
568         struct sk_buff *skb2;
569         int err, rel_msg = 0;
570         u8 rel_type = type;
571         u8 rel_code = code;
572         struct rtable *rt;
573         struct flowi4 fl4;
574
575         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576                           &rel_msg, &rel_info, offset);
577         if (err < 0)
578                 return err;
579
580         if (rel_msg == 0)
581                 return 0;
582
583         switch (rel_type) {
584         case ICMPV6_DEST_UNREACH:
585                 if (rel_code != ICMPV6_ADDR_UNREACH)
586                         return 0;
587                 rel_type = ICMP_DEST_UNREACH;
588                 rel_code = ICMP_HOST_UNREACH;
589                 break;
590         case ICMPV6_PKT_TOOBIG:
591                 if (rel_code != 0)
592                         return 0;
593                 rel_type = ICMP_DEST_UNREACH;
594                 rel_code = ICMP_FRAG_NEEDED;
595                 break;
596         default:
597                 return 0;
598         }
599
600         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
601                 return 0;
602
603         skb2 = skb_clone(skb, GFP_ATOMIC);
604         if (!skb2)
605                 return 0;
606
607         skb_dst_drop(skb2);
608
609         skb_pull(skb2, offset);
610         skb_reset_network_header(skb2);
611         eiph = ip_hdr(skb2);
612
613         /* Try to guess incoming interface */
614         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
615                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
616         if (IS_ERR(rt))
617                 goto out;
618
619         skb2->dev = rt->dst.dev;
620         ip_rt_put(rt);
621
622         /* route "incoming" packet */
623         if (rt->rt_flags & RTCF_LOCAL) {
624                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
625                                            eiph->daddr, eiph->saddr, 0, 0,
626                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
627                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
628                         if (!IS_ERR(rt))
629                                 ip_rt_put(rt);
630                         goto out;
631                 }
632                 skb_dst_set(skb2, &rt->dst);
633         } else {
634                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
635                                    skb2->dev) ||
636                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
637                         goto out;
638         }
639
640         /* change mtu on this route */
641         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
642                 if (rel_info > dst_mtu(skb_dst(skb2)))
643                         goto out;
644
645                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
646         }
647
648         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
649
650 out:
651         kfree_skb(skb2);
652         return 0;
653 }
654
655 static int
656 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
657            u8 type, u8 code, int offset, __be32 info)
658 {
659         __u32 rel_info = ntohl(info);
660         int err, rel_msg = 0;
661         u8 rel_type = type;
662         u8 rel_code = code;
663
664         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
665                           &rel_msg, &rel_info, offset);
666         if (err < 0)
667                 return err;
668
669         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
670                 struct rt6_info *rt;
671                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
672
673                 if (!skb2)
674                         return 0;
675
676                 skb_dst_drop(skb2);
677                 skb_pull(skb2, offset);
678                 skb_reset_network_header(skb2);
679
680                 /* Try to guess incoming interface */
681                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
682                                 NULL, 0, skb2, 0);
683
684                 if (rt && rt->dst.dev)
685                         skb2->dev = rt->dst.dev;
686
687                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
688
689                 ip6_rt_put(rt);
690
691                 kfree_skb(skb2);
692         }
693
694         return 0;
695 }
696
697 static int
698 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
699             u8 type, u8 code, int offset, __be32 info)
700 {
701         __u32 rel_info = ntohl(info);
702         int err, rel_msg = 0;
703         u8 rel_type = type;
704         u8 rel_code = code;
705
706         err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
707                           &rel_msg, &rel_info, offset);
708         return err;
709 }
710
711 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
712                                        const struct ipv6hdr *ipv6h,
713                                        struct sk_buff *skb)
714 {
715         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
716
717         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
718                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
719
720         return IP6_ECN_decapsulate(ipv6h, skb);
721 }
722
723 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
724                                        const struct ipv6hdr *ipv6h,
725                                        struct sk_buff *skb)
726 {
727         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
728                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
729
730         return IP6_ECN_decapsulate(ipv6h, skb);
731 }
732
733 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
734                                                const struct ipv6hdr *ipv6h,
735                                                struct sk_buff *skb)
736 {
737         /* ECN is not supported in AF_MPLS */
738         return 0;
739 }
740
741 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
742                              const struct in6_addr *laddr,
743                              const struct in6_addr *raddr)
744 {
745         struct __ip6_tnl_parm *p = &t->parms;
746         int ltype = ipv6_addr_type(laddr);
747         int rtype = ipv6_addr_type(raddr);
748         __u32 flags = 0;
749
750         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
751                 flags = IP6_TNL_F_CAP_PER_PACKET;
752         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
753                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
754                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
755                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
756                 if (ltype&IPV6_ADDR_UNICAST)
757                         flags |= IP6_TNL_F_CAP_XMIT;
758                 if (rtype&IPV6_ADDR_UNICAST)
759                         flags |= IP6_TNL_F_CAP_RCV;
760         }
761         return flags;
762 }
763 EXPORT_SYMBOL(ip6_tnl_get_cap);
764
765 /* called with rcu_read_lock() */
766 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
767                                   const struct in6_addr *laddr,
768                                   const struct in6_addr *raddr)
769 {
770         struct __ip6_tnl_parm *p = &t->parms;
771         int ret = 0;
772         struct net *net = t->net;
773
774         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
775             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
776              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
777                 struct net_device *ldev = NULL;
778
779                 if (p->link)
780                         ldev = dev_get_by_index_rcu(net, p->link);
781
782                 if ((ipv6_addr_is_multicast(laddr) ||
783                      likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
784                                                     0, IFA_F_TENTATIVE))) &&
785                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
786                      likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
787                                                      0, IFA_F_TENTATIVE))))
788                         ret = 1;
789         }
790         return ret;
791 }
792 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
793
794 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
795                          const struct tnl_ptk_info *tpi,
796                          struct metadata_dst *tun_dst,
797                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
798                                                 const struct ipv6hdr *ipv6h,
799                                                 struct sk_buff *skb),
800                          bool log_ecn_err)
801 {
802         struct pcpu_sw_netstats *tstats;
803         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
804         int err;
805
806         if ((!(tpi->flags & TUNNEL_CSUM) &&
807              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
808             ((tpi->flags & TUNNEL_CSUM) &&
809              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
810                 tunnel->dev->stats.rx_crc_errors++;
811                 tunnel->dev->stats.rx_errors++;
812                 goto drop;
813         }
814
815         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
816                 if (!(tpi->flags & TUNNEL_SEQ) ||
817                     (tunnel->i_seqno &&
818                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
819                         tunnel->dev->stats.rx_fifo_errors++;
820                         tunnel->dev->stats.rx_errors++;
821                         goto drop;
822                 }
823                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
824         }
825
826         skb->protocol = tpi->proto;
827
828         /* Warning: All skb pointers will be invalidated! */
829         if (tunnel->dev->type == ARPHRD_ETHER) {
830                 if (!pskb_may_pull(skb, ETH_HLEN)) {
831                         tunnel->dev->stats.rx_length_errors++;
832                         tunnel->dev->stats.rx_errors++;
833                         goto drop;
834                 }
835
836                 ipv6h = ipv6_hdr(skb);
837                 skb->protocol = eth_type_trans(skb, tunnel->dev);
838                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
839         } else {
840                 skb->dev = tunnel->dev;
841         }
842
843         skb_reset_network_header(skb);
844         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
845
846         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
847
848         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
849         if (unlikely(err)) {
850                 if (log_ecn_err)
851                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
852                                              &ipv6h->saddr,
853                                              ipv6_get_dsfield(ipv6h));
854                 if (err > 1) {
855                         ++tunnel->dev->stats.rx_frame_errors;
856                         ++tunnel->dev->stats.rx_errors;
857                         goto drop;
858                 }
859         }
860
861         tstats = this_cpu_ptr(tunnel->dev->tstats);
862         u64_stats_update_begin(&tstats->syncp);
863         tstats->rx_packets++;
864         tstats->rx_bytes += skb->len;
865         u64_stats_update_end(&tstats->syncp);
866
867         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
868
869         if (tun_dst)
870                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
871
872         gro_cells_receive(&tunnel->gro_cells, skb);
873         return 0;
874
875 drop:
876         if (tun_dst)
877                 dst_release((struct dst_entry *)tun_dst);
878         kfree_skb(skb);
879         return 0;
880 }
881
882 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
883                 const struct tnl_ptk_info *tpi,
884                 struct metadata_dst *tun_dst,
885                 bool log_ecn_err)
886 {
887         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
888                                     const struct ipv6hdr *ipv6h,
889                                     struct sk_buff *skb);
890
891         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
892         if (tpi->proto == htons(ETH_P_IP))
893                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
894
895         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
896                              log_ecn_err);
897 }
898 EXPORT_SYMBOL(ip6_tnl_rcv);
899
900 static const struct tnl_ptk_info tpi_v6 = {
901         /* no tunnel info required for ipxip6. */
902         .proto = htons(ETH_P_IPV6),
903 };
904
905 static const struct tnl_ptk_info tpi_v4 = {
906         /* no tunnel info required for ipxip6. */
907         .proto = htons(ETH_P_IP),
908 };
909
910 static const struct tnl_ptk_info tpi_mpls = {
911         /* no tunnel info required for mplsip6. */
912         .proto = htons(ETH_P_MPLS_UC),
913 };
914
915 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
916                       const struct tnl_ptk_info *tpi,
917                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
918                                                   const struct ipv6hdr *ipv6h,
919                                                   struct sk_buff *skb))
920 {
921         struct ip6_tnl *t;
922         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
923         struct metadata_dst *tun_dst = NULL;
924         int ret = -1;
925
926         rcu_read_lock();
927         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
928
929         if (t) {
930                 u8 tproto = READ_ONCE(t->parms.proto);
931
932                 if (tproto != ipproto && tproto != 0)
933                         goto drop;
934                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
935                         goto drop;
936                 ipv6h = ipv6_hdr(skb);
937                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
938                         goto drop;
939                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
940                         goto drop;
941                 if (t->parms.collect_md) {
942                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
943                         if (!tun_dst)
944                                 goto drop;
945                 }
946                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
947                                     log_ecn_error);
948         }
949
950         rcu_read_unlock();
951
952         return ret;
953
954 drop:
955         rcu_read_unlock();
956         kfree_skb(skb);
957         return 0;
958 }
959
960 static int ip4ip6_rcv(struct sk_buff *skb)
961 {
962         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
963                           ip4ip6_dscp_ecn_decapsulate);
964 }
965
966 static int ip6ip6_rcv(struct sk_buff *skb)
967 {
968         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
969                           ip6ip6_dscp_ecn_decapsulate);
970 }
971
972 static int mplsip6_rcv(struct sk_buff *skb)
973 {
974         return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
975                           mplsip6_dscp_ecn_decapsulate);
976 }
977
978 struct ipv6_tel_txoption {
979         struct ipv6_txoptions ops;
980         __u8 dst_opt[8];
981 };
982
983 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
984 {
985         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
986
987         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
988         opt->dst_opt[3] = 1;
989         opt->dst_opt[4] = encap_limit;
990         opt->dst_opt[5] = IPV6_TLV_PADN;
991         opt->dst_opt[6] = 1;
992
993         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
994         opt->ops.opt_nflen = 8;
995 }
996
997 /**
998  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
999  *   @t: the outgoing tunnel device
1000  *   @hdr: IPv6 header from the incoming packet
1001  *
1002  * Description:
1003  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1004  *   doesn't match source of incoming packet.
1005  *
1006  * Return:
1007  *   1 if conflict,
1008  *   0 else
1009  **/
1010
1011 static inline bool
1012 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1013 {
1014         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1015 }
1016
1017 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1018                      const struct in6_addr *laddr,
1019                      const struct in6_addr *raddr)
1020 {
1021         struct __ip6_tnl_parm *p = &t->parms;
1022         int ret = 0;
1023         struct net *net = t->net;
1024
1025         if (t->parms.collect_md)
1026                 return 1;
1027
1028         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1029             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1030              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1031                 struct net_device *ldev = NULL;
1032
1033                 rcu_read_lock();
1034                 if (p->link)
1035                         ldev = dev_get_by_index_rcu(net, p->link);
1036
1037                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1038                                                       0, IFA_F_TENTATIVE)))
1039                         pr_warn("%s xmit: Local address not yet configured!\n",
1040                                 p->name);
1041                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1042                          !ipv6_addr_is_multicast(raddr) &&
1043                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1044                                                           true, 0, IFA_F_TENTATIVE)))
1045                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1046                                 p->name);
1047                 else
1048                         ret = 1;
1049                 rcu_read_unlock();
1050         }
1051         return ret;
1052 }
1053 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1054
1055 /**
1056  * ip6_tnl_xmit - encapsulate packet and send
1057  *   @skb: the outgoing socket buffer
1058  *   @dev: the outgoing tunnel device
1059  *   @dsfield: dscp code for outer header
1060  *   @fl6: flow of tunneled packet
1061  *   @encap_limit: encapsulation limit
1062  *   @pmtu: Path MTU is stored if packet is too big
1063  *   @proto: next header value
1064  *
1065  * Description:
1066  *   Build new header and do some sanity checks on the packet before sending
1067  *   it.
1068  *
1069  * Return:
1070  *   0 on success
1071  *   -1 fail
1072  *   %-EMSGSIZE message too big. return mtu in this case.
1073  **/
1074
1075 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1076                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1077                  __u8 proto)
1078 {
1079         struct ip6_tnl *t = netdev_priv(dev);
1080         struct net *net = t->net;
1081         struct net_device_stats *stats = &t->dev->stats;
1082         struct ipv6hdr *ipv6h;
1083         struct ipv6_tel_txoption opt;
1084         struct dst_entry *dst = NULL, *ndst = NULL;
1085         struct net_device *tdev;
1086         int mtu;
1087         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1088         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1089         unsigned int max_headroom = psh_hlen;
1090         bool use_cache = false;
1091         u8 hop_limit;
1092         int err = -1;
1093
1094         if (t->parms.collect_md) {
1095                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1096                 goto route_lookup;
1097         } else {
1098                 hop_limit = t->parms.hop_limit;
1099         }
1100
1101         /* NBMA tunnel */
1102         if (ipv6_addr_any(&t->parms.raddr)) {
1103                 if (skb->protocol == htons(ETH_P_IPV6)) {
1104                         struct in6_addr *addr6;
1105                         struct neighbour *neigh;
1106                         int addr_type;
1107
1108                         if (!skb_dst(skb))
1109                                 goto tx_err_link_failure;
1110
1111                         neigh = dst_neigh_lookup(skb_dst(skb),
1112                                                  &ipv6_hdr(skb)->daddr);
1113                         if (!neigh)
1114                                 goto tx_err_link_failure;
1115
1116                         addr6 = (struct in6_addr *)&neigh->primary_key;
1117                         addr_type = ipv6_addr_type(addr6);
1118
1119                         if (addr_type == IPV6_ADDR_ANY)
1120                                 addr6 = &ipv6_hdr(skb)->daddr;
1121
1122                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1123                         neigh_release(neigh);
1124                 }
1125         } else if (t->parms.proto != 0 && !(t->parms.flags &
1126                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1127                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1128                 /* enable the cache only if neither the outer protocol nor the
1129                  * routing decision depends on the current inner header value
1130                  */
1131                 use_cache = true;
1132         }
1133
1134         if (use_cache)
1135                 dst = dst_cache_get(&t->dst_cache);
1136
1137         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1138                 goto tx_err_link_failure;
1139
1140         if (!dst) {
1141 route_lookup:
1142                 /* add dsfield to flowlabel for route lookup */
1143                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1144
1145                 dst = ip6_route_output(net, NULL, fl6);
1146
1147                 if (dst->error)
1148                         goto tx_err_link_failure;
1149                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1150                 if (IS_ERR(dst)) {
1151                         err = PTR_ERR(dst);
1152                         dst = NULL;
1153                         goto tx_err_link_failure;
1154                 }
1155                 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1156                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1157                                        &fl6->daddr, 0, &fl6->saddr))
1158                         goto tx_err_link_failure;
1159                 ndst = dst;
1160         }
1161
1162         tdev = dst->dev;
1163
1164         if (tdev == dev) {
1165                 stats->collisions++;
1166                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1167                                      t->parms.name);
1168                 goto tx_err_dst_release;
1169         }
1170         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1171         if (encap_limit >= 0) {
1172                 max_headroom += 8;
1173                 mtu -= 8;
1174         }
1175         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1176                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1177
1178         skb_dst_update_pmtu_no_confirm(skb, mtu);
1179         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1180                 *pmtu = mtu;
1181                 err = -EMSGSIZE;
1182                 goto tx_err_dst_release;
1183         }
1184
1185         if (t->err_count > 0) {
1186                 if (time_before(jiffies,
1187                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1188                         t->err_count--;
1189
1190                         dst_link_failure(skb);
1191                 } else {
1192                         t->err_count = 0;
1193                 }
1194         }
1195
1196         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1197
1198         /*
1199          * Okay, now see if we can stuff it in the buffer as-is.
1200          */
1201         max_headroom += LL_RESERVED_SPACE(tdev);
1202
1203         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1204             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1205                 struct sk_buff *new_skb;
1206
1207                 new_skb = skb_realloc_headroom(skb, max_headroom);
1208                 if (!new_skb)
1209                         goto tx_err_dst_release;
1210
1211                 if (skb->sk)
1212                         skb_set_owner_w(new_skb, skb->sk);
1213                 consume_skb(skb);
1214                 skb = new_skb;
1215         }
1216
1217         if (t->parms.collect_md) {
1218                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1219                         goto tx_err_dst_release;
1220         } else {
1221                 if (use_cache && ndst)
1222                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1223         }
1224         skb_dst_set(skb, dst);
1225
1226         if (hop_limit == 0) {
1227                 if (skb->protocol == htons(ETH_P_IP))
1228                         hop_limit = ip_hdr(skb)->ttl;
1229                 else if (skb->protocol == htons(ETH_P_IPV6))
1230                         hop_limit = ipv6_hdr(skb)->hop_limit;
1231                 else
1232                         hop_limit = ip6_dst_hoplimit(dst);
1233         }
1234
1235         /* Calculate max headroom for all the headers and adjust
1236          * needed_headroom if necessary.
1237          */
1238         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1239                         + dst->header_len + t->hlen;
1240         if (max_headroom > dev->needed_headroom)
1241                 dev->needed_headroom = max_headroom;
1242
1243         skb_set_inner_ipproto(skb, proto);
1244
1245         err = ip6_tnl_encap(skb, t, &proto, fl6);
1246         if (err)
1247                 return err;
1248
1249         if (encap_limit >= 0) {
1250                 init_tel_txopt(&opt, encap_limit);
1251                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1252         }
1253
1254         skb_push(skb, sizeof(struct ipv6hdr));
1255         skb_reset_network_header(skb);
1256         ipv6h = ipv6_hdr(skb);
1257         ip6_flow_hdr(ipv6h, dsfield,
1258                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1259         ipv6h->hop_limit = hop_limit;
1260         ipv6h->nexthdr = proto;
1261         ipv6h->saddr = fl6->saddr;
1262         ipv6h->daddr = fl6->daddr;
1263         ip6tunnel_xmit(NULL, skb, dev);
1264         return 0;
1265 tx_err_link_failure:
1266         stats->tx_carrier_errors++;
1267         dst_link_failure(skb);
1268 tx_err_dst_release:
1269         dst_release(dst);
1270         return err;
1271 }
1272 EXPORT_SYMBOL(ip6_tnl_xmit);
1273
1274 static inline int
1275 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1276                 u8 protocol)
1277 {
1278         struct ip6_tnl *t = netdev_priv(dev);
1279         struct ipv6hdr *ipv6h;
1280         const struct iphdr  *iph;
1281         int encap_limit = -1;
1282         __u16 offset;
1283         struct flowi6 fl6;
1284         __u8 dsfield, orig_dsfield;
1285         __u32 mtu;
1286         u8 tproto;
1287         int err;
1288
1289         tproto = READ_ONCE(t->parms.proto);
1290         if (tproto != protocol && tproto != 0)
1291                 return -1;
1292
1293         if (t->parms.collect_md) {
1294                 struct ip_tunnel_info *tun_info;
1295                 const struct ip_tunnel_key *key;
1296
1297                 tun_info = skb_tunnel_info(skb);
1298                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1299                              ip_tunnel_info_af(tun_info) != AF_INET6))
1300                         return -1;
1301                 key = &tun_info->key;
1302                 memset(&fl6, 0, sizeof(fl6));
1303                 fl6.flowi6_proto = protocol;
1304                 fl6.saddr = key->u.ipv6.src;
1305                 fl6.daddr = key->u.ipv6.dst;
1306                 fl6.flowlabel = key->label;
1307                 dsfield =  key->tos;
1308                 switch (protocol) {
1309                 case IPPROTO_IPIP:
1310                         iph = ip_hdr(skb);
1311                         orig_dsfield = ipv4_get_dsfield(iph);
1312                         break;
1313                 case IPPROTO_IPV6:
1314                         ipv6h = ipv6_hdr(skb);
1315                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1316                         break;
1317                 default:
1318                         orig_dsfield = dsfield;
1319                         break;
1320                 }
1321         } else {
1322                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1323                         encap_limit = t->parms.encap_limit;
1324                 if (protocol == IPPROTO_IPV6) {
1325                         offset = ip6_tnl_parse_tlv_enc_lim(skb,
1326                                                 skb_network_header(skb));
1327                         /* ip6_tnl_parse_tlv_enc_lim() might have
1328                          * reallocated skb->head
1329                          */
1330                         if (offset > 0) {
1331                                 struct ipv6_tlv_tnl_enc_lim *tel;
1332
1333                                 tel = (void *)&skb_network_header(skb)[offset];
1334                                 if (tel->encap_limit == 0) {
1335                                         icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1336                                                         ICMPV6_HDR_FIELD, offset + 2);
1337                                         return -1;
1338                                 }
1339                                 encap_limit = tel->encap_limit - 1;
1340                         }
1341                 }
1342
1343                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1344                 fl6.flowi6_proto = protocol;
1345
1346                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1347                         fl6.flowi6_mark = skb->mark;
1348                 else
1349                         fl6.flowi6_mark = t->parms.fwmark;
1350                 switch (protocol) {
1351                 case IPPROTO_IPIP:
1352                         iph = ip_hdr(skb);
1353                         orig_dsfield = ipv4_get_dsfield(iph);
1354                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1355                                 dsfield = orig_dsfield;
1356                         else
1357                                 dsfield = ip6_tclass(t->parms.flowinfo);
1358                         break;
1359                 case IPPROTO_IPV6:
1360                         ipv6h = ipv6_hdr(skb);
1361                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1362                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1363                                 dsfield = orig_dsfield;
1364                         else
1365                                 dsfield = ip6_tclass(t->parms.flowinfo);
1366                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1367                                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1368                         break;
1369                 default:
1370                         orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1371                         break;
1372                 }
1373         }
1374
1375         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1376         dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1377
1378         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1379                 return -1;
1380
1381         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1382                            protocol);
1383         if (err != 0) {
1384                 /* XXX: send ICMP error even if DF is not set. */
1385                 if (err == -EMSGSIZE)
1386                         switch (protocol) {
1387                         case IPPROTO_IPIP:
1388                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1389                                               ICMP_FRAG_NEEDED, htonl(mtu));
1390                                 break;
1391                         case IPPROTO_IPV6:
1392                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1393                                 break;
1394                         default:
1395                                 break;
1396                         }
1397                 return -1;
1398         }
1399
1400         return 0;
1401 }
1402
1403 static netdev_tx_t
1404 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1405 {
1406         struct ip6_tnl *t = netdev_priv(dev);
1407         struct net_device_stats *stats = &t->dev->stats;
1408         u8 ipproto;
1409         int ret;
1410
1411         if (!pskb_inet_may_pull(skb))
1412                 goto tx_err;
1413
1414         switch (skb->protocol) {
1415         case htons(ETH_P_IP):
1416                 ipproto = IPPROTO_IPIP;
1417                 break;
1418         case htons(ETH_P_IPV6):
1419                 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1420                         goto tx_err;
1421                 ipproto = IPPROTO_IPV6;
1422                 break;
1423         case htons(ETH_P_MPLS_UC):
1424                 ipproto = IPPROTO_MPLS;
1425                 break;
1426         default:
1427                 goto tx_err;
1428         }
1429
1430         ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1431         if (ret < 0)
1432                 goto tx_err;
1433
1434         return NETDEV_TX_OK;
1435
1436 tx_err:
1437         stats->tx_errors++;
1438         stats->tx_dropped++;
1439         kfree_skb(skb);
1440         return NETDEV_TX_OK;
1441 }
1442
1443 static void ip6_tnl_link_config(struct ip6_tnl *t)
1444 {
1445         struct net_device *dev = t->dev;
1446         struct net_device *tdev = NULL;
1447         struct __ip6_tnl_parm *p = &t->parms;
1448         struct flowi6 *fl6 = &t->fl.u.ip6;
1449         unsigned int mtu;
1450         int t_hlen;
1451
1452         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1453         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1454
1455         /* Set up flowi template */
1456         fl6->saddr = p->laddr;
1457         fl6->daddr = p->raddr;
1458         fl6->flowi6_oif = p->link;
1459         fl6->flowlabel = 0;
1460
1461         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1462                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1463         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1464                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1465
1466         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1467         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1468
1469         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1470                 dev->flags |= IFF_POINTOPOINT;
1471         else
1472                 dev->flags &= ~IFF_POINTOPOINT;
1473
1474         t->tun_hlen = 0;
1475         t->hlen = t->encap_hlen + t->tun_hlen;
1476         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1477
1478         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1479                 int strict = (ipv6_addr_type(&p->raddr) &
1480                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1481
1482                 struct rt6_info *rt = rt6_lookup(t->net,
1483                                                  &p->raddr, &p->laddr,
1484                                                  p->link, NULL, strict);
1485                 if (rt) {
1486                         tdev = rt->dst.dev;
1487                         ip6_rt_put(rt);
1488                 }
1489
1490                 if (!tdev && p->link)
1491                         tdev = __dev_get_by_index(t->net, p->link);
1492
1493                 if (tdev) {
1494                         dev->hard_header_len = tdev->hard_header_len + t_hlen;
1495                         mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1496
1497                         dev->mtu = mtu - t_hlen;
1498                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1499                                 dev->mtu -= 8;
1500
1501                         if (dev->mtu < IPV6_MIN_MTU)
1502                                 dev->mtu = IPV6_MIN_MTU;
1503                 }
1504         }
1505 }
1506
1507 /**
1508  * ip6_tnl_change - update the tunnel parameters
1509  *   @t: tunnel to be changed
1510  *   @p: tunnel configuration parameters
1511  *
1512  * Description:
1513  *   ip6_tnl_change() updates the tunnel parameters
1514  **/
1515
1516 static int
1517 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1518 {
1519         t->parms.laddr = p->laddr;
1520         t->parms.raddr = p->raddr;
1521         t->parms.flags = p->flags;
1522         t->parms.hop_limit = p->hop_limit;
1523         t->parms.encap_limit = p->encap_limit;
1524         t->parms.flowinfo = p->flowinfo;
1525         t->parms.link = p->link;
1526         t->parms.proto = p->proto;
1527         t->parms.fwmark = p->fwmark;
1528         dst_cache_reset(&t->dst_cache);
1529         ip6_tnl_link_config(t);
1530         return 0;
1531 }
1532
1533 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1534 {
1535         struct net *net = t->net;
1536         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1537         int err;
1538
1539         ip6_tnl_unlink(ip6n, t);
1540         synchronize_net();
1541         err = ip6_tnl_change(t, p);
1542         ip6_tnl_link(ip6n, t);
1543         netdev_state_change(t->dev);
1544         return err;
1545 }
1546
1547 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1548 {
1549         /* for default tnl0 device allow to change only the proto */
1550         t->parms.proto = p->proto;
1551         netdev_state_change(t->dev);
1552         return 0;
1553 }
1554
1555 static void
1556 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1557 {
1558         p->laddr = u->laddr;
1559         p->raddr = u->raddr;
1560         p->flags = u->flags;
1561         p->hop_limit = u->hop_limit;
1562         p->encap_limit = u->encap_limit;
1563         p->flowinfo = u->flowinfo;
1564         p->link = u->link;
1565         p->proto = u->proto;
1566         memcpy(p->name, u->name, sizeof(u->name));
1567 }
1568
1569 static void
1570 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1571 {
1572         u->laddr = p->laddr;
1573         u->raddr = p->raddr;
1574         u->flags = p->flags;
1575         u->hop_limit = p->hop_limit;
1576         u->encap_limit = p->encap_limit;
1577         u->flowinfo = p->flowinfo;
1578         u->link = p->link;
1579         u->proto = p->proto;
1580         memcpy(u->name, p->name, sizeof(u->name));
1581 }
1582
1583 /**
1584  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1585  *   @dev: virtual device associated with tunnel
1586  *   @ifr: parameters passed from userspace
1587  *   @cmd: command to be performed
1588  *
1589  * Description:
1590  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1591  *   from userspace.
1592  *
1593  *   The possible commands are the following:
1594  *     %SIOCGETTUNNEL: get tunnel parameters for device
1595  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1596  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1597  *     %SIOCDELTUNNEL: delete tunnel
1598  *
1599  *   The fallback device "ip6tnl0", created during module
1600  *   initialization, can be used for creating other tunnel devices.
1601  *
1602  * Return:
1603  *   0 on success,
1604  *   %-EFAULT if unable to copy data to or from userspace,
1605  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1606  *   %-EINVAL if passed tunnel parameters are invalid,
1607  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1608  *   %-ENODEV if attempting to change or delete a nonexisting device
1609  **/
1610
1611 static int
1612 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1613 {
1614         int err = 0;
1615         struct ip6_tnl_parm p;
1616         struct __ip6_tnl_parm p1;
1617         struct ip6_tnl *t = netdev_priv(dev);
1618         struct net *net = t->net;
1619         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1620
1621         memset(&p1, 0, sizeof(p1));
1622
1623         switch (cmd) {
1624         case SIOCGETTUNNEL:
1625                 if (dev == ip6n->fb_tnl_dev) {
1626                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1627                                 err = -EFAULT;
1628                                 break;
1629                         }
1630                         ip6_tnl_parm_from_user(&p1, &p);
1631                         t = ip6_tnl_locate(net, &p1, 0);
1632                         if (IS_ERR(t))
1633                                 t = netdev_priv(dev);
1634                 } else {
1635                         memset(&p, 0, sizeof(p));
1636                 }
1637                 ip6_tnl_parm_to_user(&p, &t->parms);
1638                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1639                         err = -EFAULT;
1640                 }
1641                 break;
1642         case SIOCADDTUNNEL:
1643         case SIOCCHGTUNNEL:
1644                 err = -EPERM;
1645                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1646                         break;
1647                 err = -EFAULT;
1648                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1649                         break;
1650                 err = -EINVAL;
1651                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1652                     p.proto != 0)
1653                         break;
1654                 ip6_tnl_parm_from_user(&p1, &p);
1655                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1656                 if (cmd == SIOCCHGTUNNEL) {
1657                         if (!IS_ERR(t)) {
1658                                 if (t->dev != dev) {
1659                                         err = -EEXIST;
1660                                         break;
1661                                 }
1662                         } else
1663                                 t = netdev_priv(dev);
1664                         if (dev == ip6n->fb_tnl_dev)
1665                                 err = ip6_tnl0_update(t, &p1);
1666                         else
1667                                 err = ip6_tnl_update(t, &p1);
1668                 }
1669                 if (!IS_ERR(t)) {
1670                         err = 0;
1671                         ip6_tnl_parm_to_user(&p, &t->parms);
1672                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1673                                 err = -EFAULT;
1674
1675                 } else {
1676                         err = PTR_ERR(t);
1677                 }
1678                 break;
1679         case SIOCDELTUNNEL:
1680                 err = -EPERM;
1681                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1682                         break;
1683
1684                 if (dev == ip6n->fb_tnl_dev) {
1685                         err = -EFAULT;
1686                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1687                                 break;
1688                         err = -ENOENT;
1689                         ip6_tnl_parm_from_user(&p1, &p);
1690                         t = ip6_tnl_locate(net, &p1, 0);
1691                         if (IS_ERR(t))
1692                                 break;
1693                         err = -EPERM;
1694                         if (t->dev == ip6n->fb_tnl_dev)
1695                                 break;
1696                         dev = t->dev;
1697                 }
1698                 err = 0;
1699                 unregister_netdevice(dev);
1700                 break;
1701         default:
1702                 err = -EINVAL;
1703         }
1704         return err;
1705 }
1706
1707 /**
1708  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1709  *   @dev: virtual device associated with tunnel
1710  *   @new_mtu: the new mtu
1711  *
1712  * Return:
1713  *   0 on success,
1714  *   %-EINVAL if mtu too small
1715  **/
1716
1717 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1718 {
1719         struct ip6_tnl *tnl = netdev_priv(dev);
1720
1721         if (tnl->parms.proto == IPPROTO_IPV6) {
1722                 if (new_mtu < IPV6_MIN_MTU)
1723                         return -EINVAL;
1724         } else {
1725                 if (new_mtu < ETH_MIN_MTU)
1726                         return -EINVAL;
1727         }
1728         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1729                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1730                         return -EINVAL;
1731         } else {
1732                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1733                         return -EINVAL;
1734         }
1735         dev->mtu = new_mtu;
1736         return 0;
1737 }
1738 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1739
1740 int ip6_tnl_get_iflink(const struct net_device *dev)
1741 {
1742         struct ip6_tnl *t = netdev_priv(dev);
1743
1744         return t->parms.link;
1745 }
1746 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1747
1748 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1749                           unsigned int num)
1750 {
1751         if (num >= MAX_IPTUN_ENCAP_OPS)
1752                 return -ERANGE;
1753
1754         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1755                         &ip6tun_encaps[num],
1756                         NULL, ops) ? 0 : -1;
1757 }
1758 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1759
1760 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1761                           unsigned int num)
1762 {
1763         int ret;
1764
1765         if (num >= MAX_IPTUN_ENCAP_OPS)
1766                 return -ERANGE;
1767
1768         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1769                        &ip6tun_encaps[num],
1770                        ops, NULL) == ops) ? 0 : -1;
1771
1772         synchronize_net();
1773
1774         return ret;
1775 }
1776 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1777
1778 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1779                         struct ip_tunnel_encap *ipencap)
1780 {
1781         int hlen;
1782
1783         memset(&t->encap, 0, sizeof(t->encap));
1784
1785         hlen = ip6_encap_hlen(ipencap);
1786         if (hlen < 0)
1787                 return hlen;
1788
1789         t->encap.type = ipencap->type;
1790         t->encap.sport = ipencap->sport;
1791         t->encap.dport = ipencap->dport;
1792         t->encap.flags = ipencap->flags;
1793
1794         t->encap_hlen = hlen;
1795         t->hlen = t->encap_hlen + t->tun_hlen;
1796
1797         return 0;
1798 }
1799 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1800
1801 static const struct net_device_ops ip6_tnl_netdev_ops = {
1802         .ndo_init       = ip6_tnl_dev_init,
1803         .ndo_uninit     = ip6_tnl_dev_uninit,
1804         .ndo_start_xmit = ip6_tnl_start_xmit,
1805         .ndo_do_ioctl   = ip6_tnl_ioctl,
1806         .ndo_change_mtu = ip6_tnl_change_mtu,
1807         .ndo_get_stats64 = dev_get_tstats64,
1808         .ndo_get_iflink = ip6_tnl_get_iflink,
1809 };
1810
1811 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1812                          NETIF_F_FRAGLIST |     \
1813                          NETIF_F_HIGHDMA |      \
1814                          NETIF_F_GSO_SOFTWARE | \
1815                          NETIF_F_HW_CSUM)
1816
1817 /**
1818  * ip6_tnl_dev_setup - setup virtual tunnel device
1819  *   @dev: virtual device associated with tunnel
1820  *
1821  * Description:
1822  *   Initialize function pointers and device parameters
1823  **/
1824
1825 static void ip6_tnl_dev_setup(struct net_device *dev)
1826 {
1827         dev->netdev_ops = &ip6_tnl_netdev_ops;
1828         dev->header_ops = &ip_tunnel_header_ops;
1829         dev->needs_free_netdev = true;
1830         dev->priv_destructor = ip6_dev_free;
1831
1832         dev->type = ARPHRD_TUNNEL6;
1833         dev->flags |= IFF_NOARP;
1834         dev->addr_len = sizeof(struct in6_addr);
1835         dev->features |= NETIF_F_LLTX;
1836         netif_keep_dst(dev);
1837
1838         dev->features           |= IPXIPX_FEATURES;
1839         dev->hw_features        |= IPXIPX_FEATURES;
1840
1841         /* This perm addr will be used as interface identifier by IPv6 */
1842         dev->addr_assign_type = NET_ADDR_RANDOM;
1843         eth_random_addr(dev->perm_addr);
1844 }
1845
1846
1847 /**
1848  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1849  *   @dev: virtual device associated with tunnel
1850  **/
1851
1852 static inline int
1853 ip6_tnl_dev_init_gen(struct net_device *dev)
1854 {
1855         struct ip6_tnl *t = netdev_priv(dev);
1856         int ret;
1857         int t_hlen;
1858
1859         t->dev = dev;
1860         t->net = dev_net(dev);
1861         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1862         if (!dev->tstats)
1863                 return -ENOMEM;
1864
1865         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1866         if (ret)
1867                 goto free_stats;
1868
1869         ret = gro_cells_init(&t->gro_cells, dev);
1870         if (ret)
1871                 goto destroy_dst;
1872
1873         t->tun_hlen = 0;
1874         t->hlen = t->encap_hlen + t->tun_hlen;
1875         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1876
1877         dev->type = ARPHRD_TUNNEL6;
1878         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1879         dev->mtu = ETH_DATA_LEN - t_hlen;
1880         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1881                 dev->mtu -= 8;
1882         dev->min_mtu = ETH_MIN_MTU;
1883         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1884
1885         return 0;
1886
1887 destroy_dst:
1888         dst_cache_destroy(&t->dst_cache);
1889 free_stats:
1890         free_percpu(dev->tstats);
1891         dev->tstats = NULL;
1892
1893         return ret;
1894 }
1895
1896 /**
1897  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1898  *   @dev: virtual device associated with tunnel
1899  **/
1900
1901 static int ip6_tnl_dev_init(struct net_device *dev)
1902 {
1903         struct ip6_tnl *t = netdev_priv(dev);
1904         int err = ip6_tnl_dev_init_gen(dev);
1905
1906         if (err)
1907                 return err;
1908         ip6_tnl_link_config(t);
1909         if (t->parms.collect_md)
1910                 netif_keep_dst(dev);
1911         return 0;
1912 }
1913
1914 /**
1915  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1916  *   @dev: fallback device
1917  *
1918  * Return: 0
1919  **/
1920
1921 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1922 {
1923         struct ip6_tnl *t = netdev_priv(dev);
1924         struct net *net = dev_net(dev);
1925         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1926
1927         t->parms.proto = IPPROTO_IPV6;
1928         dev_hold(dev);
1929
1930         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1931         return 0;
1932 }
1933
1934 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1935                             struct netlink_ext_ack *extack)
1936 {
1937         u8 proto;
1938
1939         if (!data || !data[IFLA_IPTUN_PROTO])
1940                 return 0;
1941
1942         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1943         if (proto != IPPROTO_IPV6 &&
1944             proto != IPPROTO_IPIP &&
1945             proto != 0)
1946                 return -EINVAL;
1947
1948         return 0;
1949 }
1950
1951 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1952                                   struct __ip6_tnl_parm *parms)
1953 {
1954         memset(parms, 0, sizeof(*parms));
1955
1956         if (!data)
1957                 return;
1958
1959         if (data[IFLA_IPTUN_LINK])
1960                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1961
1962         if (data[IFLA_IPTUN_LOCAL])
1963                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1964
1965         if (data[IFLA_IPTUN_REMOTE])
1966                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1967
1968         if (data[IFLA_IPTUN_TTL])
1969                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1970
1971         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1972                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1973
1974         if (data[IFLA_IPTUN_FLOWINFO])
1975                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1976
1977         if (data[IFLA_IPTUN_FLAGS])
1978                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1979
1980         if (data[IFLA_IPTUN_PROTO])
1981                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1982
1983         if (data[IFLA_IPTUN_COLLECT_METADATA])
1984                 parms->collect_md = true;
1985
1986         if (data[IFLA_IPTUN_FWMARK])
1987                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1988 }
1989
1990 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1991                                         struct ip_tunnel_encap *ipencap)
1992 {
1993         bool ret = false;
1994
1995         memset(ipencap, 0, sizeof(*ipencap));
1996
1997         if (!data)
1998                 return ret;
1999
2000         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
2001                 ret = true;
2002                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2003         }
2004
2005         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2006                 ret = true;
2007                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2008         }
2009
2010         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2011                 ret = true;
2012                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2013         }
2014
2015         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2016                 ret = true;
2017                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2018         }
2019
2020         return ret;
2021 }
2022
2023 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2024                            struct nlattr *tb[], struct nlattr *data[],
2025                            struct netlink_ext_ack *extack)
2026 {
2027         struct net *net = dev_net(dev);
2028         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2029         struct ip_tunnel_encap ipencap;
2030         struct ip6_tnl *nt, *t;
2031         int err;
2032
2033         nt = netdev_priv(dev);
2034
2035         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2036                 err = ip6_tnl_encap_setup(nt, &ipencap);
2037                 if (err < 0)
2038                         return err;
2039         }
2040
2041         ip6_tnl_netlink_parms(data, &nt->parms);
2042
2043         if (nt->parms.collect_md) {
2044                 if (rtnl_dereference(ip6n->collect_md_tun))
2045                         return -EEXIST;
2046         } else {
2047                 t = ip6_tnl_locate(net, &nt->parms, 0);
2048                 if (!IS_ERR(t))
2049                         return -EEXIST;
2050         }
2051
2052         err = ip6_tnl_create2(dev);
2053         if (!err && tb[IFLA_MTU])
2054                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2055
2056         return err;
2057 }
2058
2059 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2060                               struct nlattr *data[],
2061                               struct netlink_ext_ack *extack)
2062 {
2063         struct ip6_tnl *t = netdev_priv(dev);
2064         struct __ip6_tnl_parm p;
2065         struct net *net = t->net;
2066         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2067         struct ip_tunnel_encap ipencap;
2068
2069         if (dev == ip6n->fb_tnl_dev)
2070                 return -EINVAL;
2071
2072         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2073                 int err = ip6_tnl_encap_setup(t, &ipencap);
2074
2075                 if (err < 0)
2076                         return err;
2077         }
2078         ip6_tnl_netlink_parms(data, &p);
2079         if (p.collect_md)
2080                 return -EINVAL;
2081
2082         t = ip6_tnl_locate(net, &p, 0);
2083         if (!IS_ERR(t)) {
2084                 if (t->dev != dev)
2085                         return -EEXIST;
2086         } else
2087                 t = netdev_priv(dev);
2088
2089         return ip6_tnl_update(t, &p);
2090 }
2091
2092 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2093 {
2094         struct net *net = dev_net(dev);
2095         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2096
2097         if (dev != ip6n->fb_tnl_dev)
2098                 unregister_netdevice_queue(dev, head);
2099 }
2100
2101 static size_t ip6_tnl_get_size(const struct net_device *dev)
2102 {
2103         return
2104                 /* IFLA_IPTUN_LINK */
2105                 nla_total_size(4) +
2106                 /* IFLA_IPTUN_LOCAL */
2107                 nla_total_size(sizeof(struct in6_addr)) +
2108                 /* IFLA_IPTUN_REMOTE */
2109                 nla_total_size(sizeof(struct in6_addr)) +
2110                 /* IFLA_IPTUN_TTL */
2111                 nla_total_size(1) +
2112                 /* IFLA_IPTUN_ENCAP_LIMIT */
2113                 nla_total_size(1) +
2114                 /* IFLA_IPTUN_FLOWINFO */
2115                 nla_total_size(4) +
2116                 /* IFLA_IPTUN_FLAGS */
2117                 nla_total_size(4) +
2118                 /* IFLA_IPTUN_PROTO */
2119                 nla_total_size(1) +
2120                 /* IFLA_IPTUN_ENCAP_TYPE */
2121                 nla_total_size(2) +
2122                 /* IFLA_IPTUN_ENCAP_FLAGS */
2123                 nla_total_size(2) +
2124                 /* IFLA_IPTUN_ENCAP_SPORT */
2125                 nla_total_size(2) +
2126                 /* IFLA_IPTUN_ENCAP_DPORT */
2127                 nla_total_size(2) +
2128                 /* IFLA_IPTUN_COLLECT_METADATA */
2129                 nla_total_size(0) +
2130                 /* IFLA_IPTUN_FWMARK */
2131                 nla_total_size(4) +
2132                 0;
2133 }
2134
2135 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2136 {
2137         struct ip6_tnl *tunnel = netdev_priv(dev);
2138         struct __ip6_tnl_parm *parm = &tunnel->parms;
2139
2140         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2141             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2142             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2143             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2144             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2145             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2146             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2147             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2148             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2149                 goto nla_put_failure;
2150
2151         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2152             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2153             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2154             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2155                 goto nla_put_failure;
2156
2157         if (parm->collect_md)
2158                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2159                         goto nla_put_failure;
2160
2161         return 0;
2162
2163 nla_put_failure:
2164         return -EMSGSIZE;
2165 }
2166
2167 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2168 {
2169         struct ip6_tnl *tunnel = netdev_priv(dev);
2170
2171         return tunnel->net;
2172 }
2173 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2174
2175 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2176         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2177         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2178         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2179         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2180         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2181         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2182         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2183         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2184         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2185         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2186         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2187         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2188         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2189         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2190 };
2191
2192 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2193         .kind           = "ip6tnl",
2194         .maxtype        = IFLA_IPTUN_MAX,
2195         .policy         = ip6_tnl_policy,
2196         .priv_size      = sizeof(struct ip6_tnl),
2197         .setup          = ip6_tnl_dev_setup,
2198         .validate       = ip6_tnl_validate,
2199         .newlink        = ip6_tnl_newlink,
2200         .changelink     = ip6_tnl_changelink,
2201         .dellink        = ip6_tnl_dellink,
2202         .get_size       = ip6_tnl_get_size,
2203         .fill_info      = ip6_tnl_fill_info,
2204         .get_link_net   = ip6_tnl_get_link_net,
2205 };
2206
2207 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2208         .handler        = ip4ip6_rcv,
2209         .err_handler    = ip4ip6_err,
2210         .priority       =       1,
2211 };
2212
2213 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2214         .handler        = ip6ip6_rcv,
2215         .err_handler    = ip6ip6_err,
2216         .priority       =       1,
2217 };
2218
2219 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2220         .handler        = mplsip6_rcv,
2221         .err_handler    = mplsip6_err,
2222         .priority       =       1,
2223 };
2224
2225 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2226 {
2227         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2228         struct net_device *dev, *aux;
2229         int h;
2230         struct ip6_tnl *t;
2231
2232         for_each_netdev_safe(net, dev, aux)
2233                 if (dev->rtnl_link_ops == &ip6_link_ops)
2234                         unregister_netdevice_queue(dev, list);
2235
2236         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2237                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2238                 while (t) {
2239                         /* If dev is in the same netns, it has already
2240                          * been added to the list by the previous loop.
2241                          */
2242                         if (!net_eq(dev_net(t->dev), net))
2243                                 unregister_netdevice_queue(t->dev, list);
2244                         t = rtnl_dereference(t->next);
2245                 }
2246         }
2247
2248         t = rtnl_dereference(ip6n->tnls_wc[0]);
2249         while (t) {
2250                 /* If dev is in the same netns, it has already
2251                  * been added to the list by the previous loop.
2252                  */
2253                 if (!net_eq(dev_net(t->dev), net))
2254                         unregister_netdevice_queue(t->dev, list);
2255                 t = rtnl_dereference(t->next);
2256         }
2257 }
2258
2259 static int __net_init ip6_tnl_init_net(struct net *net)
2260 {
2261         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2262         struct ip6_tnl *t = NULL;
2263         int err;
2264
2265         ip6n->tnls[0] = ip6n->tnls_wc;
2266         ip6n->tnls[1] = ip6n->tnls_r_l;
2267
2268         if (!net_has_fallback_tunnels(net))
2269                 return 0;
2270         err = -ENOMEM;
2271         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2272                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2273
2274         if (!ip6n->fb_tnl_dev)
2275                 goto err_alloc_dev;
2276         dev_net_set(ip6n->fb_tnl_dev, net);
2277         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2278         /* FB netdevice is special: we have one, and only one per netns.
2279          * Allowing to move it to another netns is clearly unsafe.
2280          */
2281         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2282
2283         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2284         if (err < 0)
2285                 goto err_register;
2286
2287         err = register_netdev(ip6n->fb_tnl_dev);
2288         if (err < 0)
2289                 goto err_register;
2290
2291         t = netdev_priv(ip6n->fb_tnl_dev);
2292
2293         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2294         return 0;
2295
2296 err_register:
2297         free_netdev(ip6n->fb_tnl_dev);
2298 err_alloc_dev:
2299         return err;
2300 }
2301
2302 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2303 {
2304         struct net *net;
2305         LIST_HEAD(list);
2306
2307         rtnl_lock();
2308         list_for_each_entry(net, net_list, exit_list)
2309                 ip6_tnl_destroy_tunnels(net, &list);
2310         unregister_netdevice_many(&list);
2311         rtnl_unlock();
2312 }
2313
2314 static struct pernet_operations ip6_tnl_net_ops = {
2315         .init = ip6_tnl_init_net,
2316         .exit_batch = ip6_tnl_exit_batch_net,
2317         .id   = &ip6_tnl_net_id,
2318         .size = sizeof(struct ip6_tnl_net),
2319 };
2320
2321 /**
2322  * ip6_tunnel_init - register protocol and reserve needed resources
2323  *
2324  * Return: 0 on success
2325  **/
2326
2327 static int __init ip6_tunnel_init(void)
2328 {
2329         int  err;
2330
2331         if (!ipv6_mod_enabled())
2332                 return -EOPNOTSUPP;
2333
2334         err = register_pernet_device(&ip6_tnl_net_ops);
2335         if (err < 0)
2336                 goto out_pernet;
2337
2338         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2339         if (err < 0) {
2340                 pr_err("%s: can't register ip4ip6\n", __func__);
2341                 goto out_ip4ip6;
2342         }
2343
2344         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2345         if (err < 0) {
2346                 pr_err("%s: can't register ip6ip6\n", __func__);
2347                 goto out_ip6ip6;
2348         }
2349
2350         if (ip6_tnl_mpls_supported()) {
2351                 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2352                 if (err < 0) {
2353                         pr_err("%s: can't register mplsip6\n", __func__);
2354                         goto out_mplsip6;
2355                 }
2356         }
2357
2358         err = rtnl_link_register(&ip6_link_ops);
2359         if (err < 0)
2360                 goto rtnl_link_failed;
2361
2362         return 0;
2363
2364 rtnl_link_failed:
2365         if (ip6_tnl_mpls_supported())
2366                 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2367 out_mplsip6:
2368         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2369 out_ip6ip6:
2370         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2371 out_ip4ip6:
2372         unregister_pernet_device(&ip6_tnl_net_ops);
2373 out_pernet:
2374         return err;
2375 }
2376
2377 /**
2378  * ip6_tunnel_cleanup - free resources and unregister protocol
2379  **/
2380
2381 static void __exit ip6_tunnel_cleanup(void)
2382 {
2383         rtnl_link_unregister(&ip6_link_ops);
2384         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2385                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2386
2387         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2388                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2389
2390         if (ip6_tnl_mpls_supported() &&
2391             xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2392                 pr_info("%s: can't deregister mplsip6\n", __func__);
2393         unregister_pernet_device(&ip6_tnl_net_ops);
2394 }
2395
2396 module_init(ip6_tunnel_init);
2397 module_exit(ip6_tunnel_cleanup);