~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv6/ip6_tunnel.c

Version: ~ [ linux-5.4-rc3 ] ~ [ linux-5.3.6 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.79 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.149 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.196 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.196 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.75 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *      IPv6 tunneling device
  3  *      Linux INET6 implementation
  4  *
  5  *      Authors:
  6  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
  7  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
  8  *
  9  *      Based on:
 10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
 11  *
 12  *      RFC 2473
 13  *
 14  *      This program is free software; you can redistribute it and/or
 15  *      modify it under the terms of the GNU General Public License
 16  *      as published by the Free Software Foundation; either version
 17  *      2 of the License, or (at your option) any later version.
 18  *
 19  */
 20 
 21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 22 
 23 #include <linux/module.h>
 24 #include <linux/capability.h>
 25 #include <linux/errno.h>
 26 #include <linux/types.h>
 27 #include <linux/sockios.h>
 28 #include <linux/icmp.h>
 29 #include <linux/if.h>
 30 #include <linux/in.h>
 31 #include <linux/ip.h>
 32 #include <linux/net.h>
 33 #include <linux/in6.h>
 34 #include <linux/netdevice.h>
 35 #include <linux/if_arp.h>
 36 #include <linux/icmpv6.h>
 37 #include <linux/init.h>
 38 #include <linux/route.h>
 39 #include <linux/rtnetlink.h>
 40 #include <linux/netfilter_ipv6.h>
 41 #include <linux/slab.h>
 42 #include <linux/hash.h>
 43 #include <linux/etherdevice.h>
 44 
 45 #include <linux/uaccess.h>
 46 #include <linux/atomic.h>
 47 
 48 #include <net/icmp.h>
 49 #include <net/ip.h>
 50 #include <net/ip_tunnels.h>
 51 #include <net/ipv6.h>
 52 #include <net/ip6_route.h>
 53 #include <net/addrconf.h>
 54 #include <net/ip6_tunnel.h>
 55 #include <net/xfrm.h>
 56 #include <net/dsfield.h>
 57 #include <net/inet_ecn.h>
 58 #include <net/net_namespace.h>
 59 #include <net/netns/generic.h>
 60 #include <net/dst_metadata.h>
 61 
 62 MODULE_AUTHOR("Ville Nuorvala");
 63 MODULE_DESCRIPTION("IPv6 tunneling device");
 64 MODULE_LICENSE("GPL");
 65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
 66 MODULE_ALIAS_NETDEV("ip6tnl0");
 67 
 68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
 69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
 70 
 71 static bool log_ecn_error = true;
 72 module_param(log_ecn_error, bool, 0644);
 73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
 74 
 75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
 76 {
 77         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
 78 
 79         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
 80 }
 81 
 82 static int ip6_tnl_dev_init(struct net_device *dev);
 83 static void ip6_tnl_dev_setup(struct net_device *dev);
 84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
 85 
 86 static unsigned int ip6_tnl_net_id __read_mostly;
 87 struct ip6_tnl_net {
 88         /* the IPv6 tunnel fallback device */
 89         struct net_device *fb_tnl_dev;
 90         /* lists for storing tunnels in use */
 91         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
 92         struct ip6_tnl __rcu *tnls_wc[1];
 93         struct ip6_tnl __rcu **tnls[2];
 94         struct ip6_tnl __rcu *collect_md_tun;
 95 };
 96 
 97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
 98 {
 99         struct pcpu_sw_netstats tmp, sum = { 0 };
100         int i;
101 
102         for_each_possible_cpu(i) {
103                 unsigned int start;
104                 const struct pcpu_sw_netstats *tstats =
105                                                    per_cpu_ptr(dev->tstats, i);
106 
107                 do {
108                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
109                         tmp.rx_packets = tstats->rx_packets;
110                         tmp.rx_bytes = tstats->rx_bytes;
111                         tmp.tx_packets = tstats->tx_packets;
112                         tmp.tx_bytes =  tstats->tx_bytes;
113                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
114 
115                 sum.rx_packets += tmp.rx_packets;
116                 sum.rx_bytes   += tmp.rx_bytes;
117                 sum.tx_packets += tmp.tx_packets;
118                 sum.tx_bytes   += tmp.tx_bytes;
119         }
120         dev->stats.rx_packets = sum.rx_packets;
121         dev->stats.rx_bytes   = sum.rx_bytes;
122         dev->stats.tx_packets = sum.tx_packets;
123         dev->stats.tx_bytes   = sum.tx_bytes;
124         return &dev->stats;
125 }
126 
127 /**
128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129  *   @remote: the address of the tunnel exit-point
130  *   @local: the address of the tunnel entry-point
131  *
132  * Return:
133  *   tunnel matching given end-points if found,
134  *   else fallback tunnel if its device is up,
135  *   else %NULL
136  **/
137 
138 #define for_each_ip6_tunnel_rcu(start) \
139         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
140 
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
143 {
144         unsigned int hash = HASH(remote, local);
145         struct ip6_tnl *t;
146         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147         struct in6_addr any;
148 
149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
151                     ipv6_addr_equal(remote, &t->parms.raddr) &&
152                     (t->dev->flags & IFF_UP))
153                         return t;
154         }
155 
156         memset(&any, 0, sizeof(any));
157         hash = HASH(&any, local);
158         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
160                     ipv6_addr_any(&t->parms.raddr) &&
161                     (t->dev->flags & IFF_UP))
162                         return t;
163         }
164 
165         hash = HASH(remote, &any);
166         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168                     ipv6_addr_any(&t->parms.laddr) &&
169                     (t->dev->flags & IFF_UP))
170                         return t;
171         }
172 
173         t = rcu_dereference(ip6n->collect_md_tun);
174         if (t && t->dev->flags & IFF_UP)
175                 return t;
176 
177         t = rcu_dereference(ip6n->tnls_wc[0]);
178         if (t && (t->dev->flags & IFF_UP))
179                 return t;
180 
181         return NULL;
182 }
183 
184 /**
185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
186  *   @p: parameters containing tunnel end-points
187  *
188  * Description:
189  *   ip6_tnl_bucket() returns the head of the list matching the
190  *   &struct in6_addr entries laddr and raddr in @p.
191  *
192  * Return: head of IPv6 tunnel list
193  **/
194 
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
197 {
198         const struct in6_addr *remote = &p->raddr;
199         const struct in6_addr *local = &p->laddr;
200         unsigned int h = 0;
201         int prio = 0;
202 
203         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204                 prio = 1;
205                 h = HASH(remote, local);
206         }
207         return &ip6n->tnls[prio][h];
208 }
209 
210 /**
211  * ip6_tnl_link - add tunnel to hash table
212  *   @t: tunnel to be added
213  **/
214 
215 static void
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
217 {
218         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
219 
220         if (t->parms.collect_md)
221                 rcu_assign_pointer(ip6n->collect_md_tun, t);
222         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223         rcu_assign_pointer(*tp, t);
224 }
225 
226 /**
227  * ip6_tnl_unlink - remove tunnel from hash table
228  *   @t: tunnel to be removed
229  **/
230 
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234         struct ip6_tnl __rcu **tp;
235         struct ip6_tnl *iter;
236 
237         if (t->parms.collect_md)
238                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
239 
240         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241              (iter = rtnl_dereference(*tp)) != NULL;
242              tp = &iter->next) {
243                 if (t == iter) {
244                         rcu_assign_pointer(*tp, t->next);
245                         break;
246                 }
247         }
248 }
249 
250 static void ip6_dev_free(struct net_device *dev)
251 {
252         struct ip6_tnl *t = netdev_priv(dev);
253 
254         gro_cells_destroy(&t->gro_cells);
255         dst_cache_destroy(&t->dst_cache);
256         free_percpu(dev->tstats);
257 }
258 
259 static int ip6_tnl_create2(struct net_device *dev)
260 {
261         struct ip6_tnl *t = netdev_priv(dev);
262         struct net *net = dev_net(dev);
263         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264         int err;
265 
266         t = netdev_priv(dev);
267 
268         dev->rtnl_link_ops = &ip6_link_ops;
269         err = register_netdevice(dev);
270         if (err < 0)
271                 goto out;
272 
273         strcpy(t->parms.name, dev->name);
274 
275         dev_hold(dev);
276         ip6_tnl_link(ip6n, t);
277         return 0;
278 
279 out:
280         return err;
281 }
282 
283 /**
284  * ip6_tnl_create - create a new tunnel
285  *   @p: tunnel parameters
286  *   @pt: pointer to new tunnel
287  *
288  * Description:
289  *   Create tunnel matching given parameters.
290  *
291  * Return:
292  *   created tunnel or error pointer
293  **/
294 
295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 {
297         struct net_device *dev;
298         struct ip6_tnl *t;
299         char name[IFNAMSIZ];
300         int err = -E2BIG;
301 
302         if (p->name[0]) {
303                 if (!dev_valid_name(p->name))
304                         goto failed;
305                 strlcpy(name, p->name, IFNAMSIZ);
306         } else {
307                 sprintf(name, "ip6tnl%%d");
308         }
309         err = -ENOMEM;
310         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
311                            ip6_tnl_dev_setup);
312         if (!dev)
313                 goto failed;
314 
315         dev_net_set(dev, net);
316 
317         t = netdev_priv(dev);
318         t->parms = *p;
319         t->net = dev_net(dev);
320         err = ip6_tnl_create2(dev);
321         if (err < 0)
322                 goto failed_free;
323 
324         return t;
325 
326 failed_free:
327         free_netdev(dev);
328 failed:
329         return ERR_PTR(err);
330 }
331 
332 /**
333  * ip6_tnl_locate - find or create tunnel matching given parameters
334  *   @p: tunnel parameters
335  *   @create: != 0 if allowed to create new tunnel if no match found
336  *
337  * Description:
338  *   ip6_tnl_locate() first tries to locate an existing tunnel
339  *   based on @parms. If this is unsuccessful, but @create is set a new
340  *   tunnel device is created and registered for use.
341  *
342  * Return:
343  *   matching tunnel or error pointer
344  **/
345 
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347                 struct __ip6_tnl_parm *p, int create)
348 {
349         const struct in6_addr *remote = &p->raddr;
350         const struct in6_addr *local = &p->laddr;
351         struct ip6_tnl __rcu **tp;
352         struct ip6_tnl *t;
353         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354 
355         for (tp = ip6_tnl_bucket(ip6n, p);
356              (t = rtnl_dereference(*tp)) != NULL;
357              tp = &t->next) {
358                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
359                     ipv6_addr_equal(remote, &t->parms.raddr)) {
360                         if (create)
361                                 return ERR_PTR(-EEXIST);
362 
363                         return t;
364                 }
365         }
366         if (!create)
367                 return ERR_PTR(-ENODEV);
368         return ip6_tnl_create(net, p);
369 }
370 
371 /**
372  * ip6_tnl_dev_uninit - tunnel device uninitializer
373  *   @dev: the device to be destroyed
374  *
375  * Description:
376  *   ip6_tnl_dev_uninit() removes tunnel from its list
377  **/
378 
379 static void
380 ip6_tnl_dev_uninit(struct net_device *dev)
381 {
382         struct ip6_tnl *t = netdev_priv(dev);
383         struct net *net = t->net;
384         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
385 
386         if (dev == ip6n->fb_tnl_dev)
387                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
388         else
389                 ip6_tnl_unlink(ip6n, t);
390         dst_cache_reset(&t->dst_cache);
391         dev_put(dev);
392 }
393 
394 /**
395  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
396  *   @skb: received socket buffer
397  *
398  * Return:
399  *   0 if none was found,
400  *   else index to encapsulation limit
401  **/
402 
403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
404 {
405         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
406         unsigned int nhoff = raw - skb->data;
407         unsigned int off = nhoff + sizeof(*ipv6h);
408         u8 next, nexthdr = ipv6h->nexthdr;
409 
410         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
411                 struct ipv6_opt_hdr *hdr;
412                 u16 optlen;
413 
414                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
415                         break;
416 
417                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
418                 if (nexthdr == NEXTHDR_FRAGMENT) {
419                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
420                         if (frag_hdr->frag_off)
421                                 break;
422                         optlen = 8;
423                 } else if (nexthdr == NEXTHDR_AUTH) {
424                         optlen = (hdr->hdrlen + 2) << 2;
425                 } else {
426                         optlen = ipv6_optlen(hdr);
427                 }
428                 /* cache hdr->nexthdr, since pskb_may_pull() might
429                  * invalidate hdr
430                  */
431                 next = hdr->nexthdr;
432                 if (nexthdr == NEXTHDR_DEST) {
433                         u16 i = 2;
434 
435                         /* Remember : hdr is no longer valid at this point. */
436                         if (!pskb_may_pull(skb, off + optlen))
437                                 break;
438 
439                         while (1) {
440                                 struct ipv6_tlv_tnl_enc_lim *tel;
441 
442                                 /* No more room for encapsulation limit */
443                                 if (i + sizeof(*tel) > optlen)
444                                         break;
445 
446                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
447                                 /* return index of option if found and valid */
448                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
449                                     tel->length == 1)
450                                         return i + off - nhoff;
451                                 /* else jump to next option */
452                                 if (tel->type)
453                                         i += tel->length + 2;
454                                 else
455                                         i++;
456                         }
457                 }
458                 nexthdr = next;
459                 off += optlen;
460         }
461         return 0;
462 }
463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
464 
465 /**
466  * ip6_tnl_err - tunnel error handler
467  *
468  * Description:
469  *   ip6_tnl_err() should handle errors in the tunnel according
470  *   to the specifications in RFC 2473.
471  **/
472 
473 static int
474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
475             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
476 {
477         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
478         struct net *net = dev_net(skb->dev);
479         u8 rel_type = ICMPV6_DEST_UNREACH;
480         u8 rel_code = ICMPV6_ADDR_UNREACH;
481         __u32 rel_info = 0;
482         struct ip6_tnl *t;
483         int err = -ENOENT;
484         int rel_msg = 0;
485         u8 tproto;
486         __u16 len;
487 
488         /* If the packet doesn't contain the original IPv6 header we are
489            in trouble since we might need the source address for further
490            processing of the error. */
491 
492         rcu_read_lock();
493         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
494         if (!t)
495                 goto out;
496 
497         tproto = READ_ONCE(t->parms.proto);
498         if (tproto != ipproto && tproto != 0)
499                 goto out;
500 
501         err = 0;
502 
503         switch (*type) {
504                 struct ipv6_tlv_tnl_enc_lim *tel;
505                 __u32 mtu, teli;
506         case ICMPV6_DEST_UNREACH:
507                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
508                                     t->parms.name);
509                 rel_msg = 1;
510                 break;
511         case ICMPV6_TIME_EXCEED:
512                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
514                                             t->parms.name);
515                         rel_msg = 1;
516                 }
517                 break;
518         case ICMPV6_PARAMPROB:
519                 teli = 0;
520                 if ((*code) == ICMPV6_HDR_FIELD)
521                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522 
523                 if (teli && teli == *info - 2) {
524                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525                         if (tel->encap_limit == 0) {
526                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
527                                                     t->parms.name);
528                                 rel_msg = 1;
529                         }
530                 } else {
531                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
532                                             t->parms.name);
533                 }
534                 break;
535         case ICMPV6_PKT_TOOBIG:
536                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
537                                 sock_net_uid(net, NULL));
538                 mtu = *info - offset;
539                 if (mtu < IPV6_MIN_MTU)
540                         mtu = IPV6_MIN_MTU;
541                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
542                 if (len > mtu) {
543                         rel_type = ICMPV6_PKT_TOOBIG;
544                         rel_code = 0;
545                         rel_info = mtu;
546                         rel_msg = 1;
547                 }
548                 break;
549         case NDISC_REDIRECT:
550                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
551                              sock_net_uid(net, NULL));
552                 break;
553         }
554 
555         *type = rel_type;
556         *code = rel_code;
557         *info = rel_info;
558         *msg = rel_msg;
559 
560 out:
561         rcu_read_unlock();
562         return err;
563 }
564 
565 static int
566 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
567            u8 type, u8 code, int offset, __be32 info)
568 {
569         __u32 rel_info = ntohl(info);
570         const struct iphdr *eiph;
571         struct sk_buff *skb2;
572         int err, rel_msg = 0;
573         u8 rel_type = type;
574         u8 rel_code = code;
575         struct rtable *rt;
576         struct flowi4 fl4;
577 
578         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
579                           &rel_msg, &rel_info, offset);
580         if (err < 0)
581                 return err;
582 
583         if (rel_msg == 0)
584                 return 0;
585 
586         switch (rel_type) {
587         case ICMPV6_DEST_UNREACH:
588                 if (rel_code != ICMPV6_ADDR_UNREACH)
589                         return 0;
590                 rel_type = ICMP_DEST_UNREACH;
591                 rel_code = ICMP_HOST_UNREACH;
592                 break;
593         case ICMPV6_PKT_TOOBIG:
594                 if (rel_code != 0)
595                         return 0;
596                 rel_type = ICMP_DEST_UNREACH;
597                 rel_code = ICMP_FRAG_NEEDED;
598                 break;
599         default:
600                 return 0;
601         }
602 
603         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
604                 return 0;
605 
606         skb2 = skb_clone(skb, GFP_ATOMIC);
607         if (!skb2)
608                 return 0;
609 
610         skb_dst_drop(skb2);
611 
612         skb_pull(skb2, offset);
613         skb_reset_network_header(skb2);
614         eiph = ip_hdr(skb2);
615 
616         /* Try to guess incoming interface */
617         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
618                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
619         if (IS_ERR(rt))
620                 goto out;
621 
622         skb2->dev = rt->dst.dev;
623         ip_rt_put(rt);
624 
625         /* route "incoming" packet */
626         if (rt->rt_flags & RTCF_LOCAL) {
627                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
628                                            eiph->daddr, eiph->saddr, 0, 0,
629                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
630                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
631                         if (!IS_ERR(rt))
632                                 ip_rt_put(rt);
633                         goto out;
634                 }
635                 skb_dst_set(skb2, &rt->dst);
636         } else {
637                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
638                                    skb2->dev) ||
639                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
640                         goto out;
641         }
642 
643         /* change mtu on this route */
644         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
645                 if (rel_info > dst_mtu(skb_dst(skb2)))
646                         goto out;
647 
648                 skb_dst_update_pmtu(skb2, rel_info);
649         }
650 
651         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
652 
653 out:
654         kfree_skb(skb2);
655         return 0;
656 }
657 
658 static int
659 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
660            u8 type, u8 code, int offset, __be32 info)
661 {
662         __u32 rel_info = ntohl(info);
663         int err, rel_msg = 0;
664         u8 rel_type = type;
665         u8 rel_code = code;
666 
667         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
668                           &rel_msg, &rel_info, offset);
669         if (err < 0)
670                 return err;
671 
672         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
673                 struct rt6_info *rt;
674                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
675 
676                 if (!skb2)
677                         return 0;
678 
679                 skb_dst_drop(skb2);
680                 skb_pull(skb2, offset);
681                 skb_reset_network_header(skb2);
682 
683                 /* Try to guess incoming interface */
684                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
685                                 NULL, 0, 0);
686 
687                 if (rt && rt->dst.dev)
688                         skb2->dev = rt->dst.dev;
689 
690                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
691 
692                 ip6_rt_put(rt);
693 
694                 kfree_skb(skb2);
695         }
696 
697         return 0;
698 }
699 
700 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
701                                        const struct ipv6hdr *ipv6h,
702                                        struct sk_buff *skb)
703 {
704         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
705 
706         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
707                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
708 
709         return IP6_ECN_decapsulate(ipv6h, skb);
710 }
711 
712 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
713                                        const struct ipv6hdr *ipv6h,
714                                        struct sk_buff *skb)
715 {
716         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
718 
719         return IP6_ECN_decapsulate(ipv6h, skb);
720 }
721 
722 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
723                              const struct in6_addr *laddr,
724                              const struct in6_addr *raddr)
725 {
726         struct __ip6_tnl_parm *p = &t->parms;
727         int ltype = ipv6_addr_type(laddr);
728         int rtype = ipv6_addr_type(raddr);
729         __u32 flags = 0;
730 
731         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
732                 flags = IP6_TNL_F_CAP_PER_PACKET;
733         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
734                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
735                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
736                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
737                 if (ltype&IPV6_ADDR_UNICAST)
738                         flags |= IP6_TNL_F_CAP_XMIT;
739                 if (rtype&IPV6_ADDR_UNICAST)
740                         flags |= IP6_TNL_F_CAP_RCV;
741         }
742         return flags;
743 }
744 EXPORT_SYMBOL(ip6_tnl_get_cap);
745 
746 /* called with rcu_read_lock() */
747 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
748                                   const struct in6_addr *laddr,
749                                   const struct in6_addr *raddr)
750 {
751         struct __ip6_tnl_parm *p = &t->parms;
752         int ret = 0;
753         struct net *net = t->net;
754 
755         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
756             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
757              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
758                 struct net_device *ldev = NULL;
759 
760                 if (p->link)
761                         ldev = dev_get_by_index_rcu(net, p->link);
762 
763                 if ((ipv6_addr_is_multicast(laddr) ||
764                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
765                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
766                      likely(!ipv6_chk_addr(net, raddr, NULL, 0))))
767                         ret = 1;
768         }
769         return ret;
770 }
771 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
772 
773 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
774                          const struct tnl_ptk_info *tpi,
775                          struct metadata_dst *tun_dst,
776                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
777                                                 const struct ipv6hdr *ipv6h,
778                                                 struct sk_buff *skb),
779                          bool log_ecn_err)
780 {
781         struct pcpu_sw_netstats *tstats;
782         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
783         int err;
784 
785         if ((!(tpi->flags & TUNNEL_CSUM) &&
786              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
787             ((tpi->flags & TUNNEL_CSUM) &&
788              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
789                 tunnel->dev->stats.rx_crc_errors++;
790                 tunnel->dev->stats.rx_errors++;
791                 goto drop;
792         }
793 
794         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
795                 if (!(tpi->flags & TUNNEL_SEQ) ||
796                     (tunnel->i_seqno &&
797                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
798                         tunnel->dev->stats.rx_fifo_errors++;
799                         tunnel->dev->stats.rx_errors++;
800                         goto drop;
801                 }
802                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
803         }
804 
805         skb->protocol = tpi->proto;
806 
807         /* Warning: All skb pointers will be invalidated! */
808         if (tunnel->dev->type == ARPHRD_ETHER) {
809                 if (!pskb_may_pull(skb, ETH_HLEN)) {
810                         tunnel->dev->stats.rx_length_errors++;
811                         tunnel->dev->stats.rx_errors++;
812                         goto drop;
813                 }
814 
815                 ipv6h = ipv6_hdr(skb);
816                 skb->protocol = eth_type_trans(skb, tunnel->dev);
817                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
818         } else {
819                 skb->dev = tunnel->dev;
820         }
821 
822         skb_reset_network_header(skb);
823         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
824 
825         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
826 
827         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
828         if (unlikely(err)) {
829                 if (log_ecn_err)
830                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
831                                              &ipv6h->saddr,
832                                              ipv6_get_dsfield(ipv6h));
833                 if (err > 1) {
834                         ++tunnel->dev->stats.rx_frame_errors;
835                         ++tunnel->dev->stats.rx_errors;
836                         goto drop;
837                 }
838         }
839 
840         tstats = this_cpu_ptr(tunnel->dev->tstats);
841         u64_stats_update_begin(&tstats->syncp);
842         tstats->rx_packets++;
843         tstats->rx_bytes += skb->len;
844         u64_stats_update_end(&tstats->syncp);
845 
846         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
847 
848         if (tun_dst)
849                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
850 
851         gro_cells_receive(&tunnel->gro_cells, skb);
852         return 0;
853 
854 drop:
855         if (tun_dst)
856                 dst_release((struct dst_entry *)tun_dst);
857         kfree_skb(skb);
858         return 0;
859 }
860 
861 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
862                 const struct tnl_ptk_info *tpi,
863                 struct metadata_dst *tun_dst,
864                 bool log_ecn_err)
865 {
866         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
867                              log_ecn_err);
868 }
869 EXPORT_SYMBOL(ip6_tnl_rcv);
870 
871 static const struct tnl_ptk_info tpi_v6 = {
872         /* no tunnel info required for ipxip6. */
873         .proto = htons(ETH_P_IPV6),
874 };
875 
876 static const struct tnl_ptk_info tpi_v4 = {
877         /* no tunnel info required for ipxip6. */
878         .proto = htons(ETH_P_IP),
879 };
880 
881 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
882                       const struct tnl_ptk_info *tpi,
883                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
884                                                   const struct ipv6hdr *ipv6h,
885                                                   struct sk_buff *skb))
886 {
887         struct ip6_tnl *t;
888         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
889         struct metadata_dst *tun_dst = NULL;
890         int ret = -1;
891 
892         rcu_read_lock();
893         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
894 
895         if (t) {
896                 u8 tproto = READ_ONCE(t->parms.proto);
897 
898                 if (tproto != ipproto && tproto != 0)
899                         goto drop;
900                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
901                         goto drop;
902                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
903                         goto drop;
904                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
905                         goto drop;
906                 if (t->parms.collect_md) {
907                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
908                         if (!tun_dst)
909                                 goto drop;
910                 }
911                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
912                                     log_ecn_error);
913         }
914 
915         rcu_read_unlock();
916 
917         return ret;
918 
919 drop:
920         rcu_read_unlock();
921         kfree_skb(skb);
922         return 0;
923 }
924 
925 static int ip4ip6_rcv(struct sk_buff *skb)
926 {
927         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
928                           ip4ip6_dscp_ecn_decapsulate);
929 }
930 
931 static int ip6ip6_rcv(struct sk_buff *skb)
932 {
933         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
934                           ip6ip6_dscp_ecn_decapsulate);
935 }
936 
937 struct ipv6_tel_txoption {
938         struct ipv6_txoptions ops;
939         __u8 dst_opt[8];
940 };
941 
942 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
943 {
944         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
945 
946         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
947         opt->dst_opt[3] = 1;
948         opt->dst_opt[4] = encap_limit;
949         opt->dst_opt[5] = IPV6_TLV_PADN;
950         opt->dst_opt[6] = 1;
951 
952         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
953         opt->ops.opt_nflen = 8;
954 }
955 
956 /**
957  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
958  *   @t: the outgoing tunnel device
959  *   @hdr: IPv6 header from the incoming packet
960  *
961  * Description:
962  *   Avoid trivial tunneling loop by checking that tunnel exit-point
963  *   doesn't match source of incoming packet.
964  *
965  * Return:
966  *   1 if conflict,
967  *   0 else
968  **/
969 
970 static inline bool
971 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
972 {
973         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
974 }
975 
976 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
977                      const struct in6_addr *laddr,
978                      const struct in6_addr *raddr)
979 {
980         struct __ip6_tnl_parm *p = &t->parms;
981         int ret = 0;
982         struct net *net = t->net;
983 
984         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
985             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
986              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
987                 struct net_device *ldev = NULL;
988 
989                 rcu_read_lock();
990                 if (p->link)
991                         ldev = dev_get_by_index_rcu(net, p->link);
992 
993                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
994                         pr_warn("%s xmit: Local address not yet configured!\n",
995                                 p->name);
996                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
997                          !ipv6_addr_is_multicast(raddr) &&
998                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
999                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1000                                 p->name);
1001                 else
1002                         ret = 1;
1003                 rcu_read_unlock();
1004         }
1005         return ret;
1006 }
1007 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1008 
1009 /**
1010  * ip6_tnl_xmit - encapsulate packet and send
1011  *   @skb: the outgoing socket buffer
1012  *   @dev: the outgoing tunnel device
1013  *   @dsfield: dscp code for outer header
1014  *   @fl6: flow of tunneled packet
1015  *   @encap_limit: encapsulation limit
1016  *   @pmtu: Path MTU is stored if packet is too big
1017  *   @proto: next header value
1018  *
1019  * Description:
1020  *   Build new header and do some sanity checks on the packet before sending
1021  *   it.
1022  *
1023  * Return:
1024  *   0 on success
1025  *   -1 fail
1026  *   %-EMSGSIZE message too big. return mtu in this case.
1027  **/
1028 
1029 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1030                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1031                  __u8 proto)
1032 {
1033         struct ip6_tnl *t = netdev_priv(dev);
1034         struct net *net = t->net;
1035         struct net_device_stats *stats = &t->dev->stats;
1036         struct ipv6hdr *ipv6h;
1037         struct ipv6_tel_txoption opt;
1038         struct dst_entry *dst = NULL, *ndst = NULL;
1039         struct net_device *tdev;
1040         int mtu;
1041         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1042         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1043         unsigned int max_headroom = psh_hlen;
1044         bool use_cache = false;
1045         u8 hop_limit;
1046         int err = -1;
1047 
1048         if (t->parms.collect_md) {
1049                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1050                 goto route_lookup;
1051         } else {
1052                 hop_limit = t->parms.hop_limit;
1053         }
1054 
1055         /* NBMA tunnel */
1056         if (ipv6_addr_any(&t->parms.raddr)) {
1057                 if (skb->protocol == htons(ETH_P_IPV6)) {
1058                         struct in6_addr *addr6;
1059                         struct neighbour *neigh;
1060                         int addr_type;
1061 
1062                         if (!skb_dst(skb))
1063                                 goto tx_err_link_failure;
1064 
1065                         neigh = dst_neigh_lookup(skb_dst(skb),
1066                                                  &ipv6_hdr(skb)->daddr);
1067                         if (!neigh)
1068                                 goto tx_err_link_failure;
1069 
1070                         addr6 = (struct in6_addr *)&neigh->primary_key;
1071                         addr_type = ipv6_addr_type(addr6);
1072 
1073                         if (addr_type == IPV6_ADDR_ANY)
1074                                 addr6 = &ipv6_hdr(skb)->daddr;
1075 
1076                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1077                         neigh_release(neigh);
1078                 }
1079         } else if (t->parms.proto != 0 && !(t->parms.flags &
1080                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1081                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1082                 /* enable the cache only if neither the outer protocol nor the
1083                  * routing decision depends on the current inner header value
1084                  */
1085                 use_cache = true;
1086         }
1087 
1088         if (use_cache)
1089                 dst = dst_cache_get(&t->dst_cache);
1090 
1091         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1092                 goto tx_err_link_failure;
1093 
1094         if (!dst) {
1095 route_lookup:
1096                 /* add dsfield to flowlabel for route lookup */
1097                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1098 
1099                 dst = ip6_route_output(net, NULL, fl6);
1100 
1101                 if (dst->error)
1102                         goto tx_err_link_failure;
1103                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1104                 if (IS_ERR(dst)) {
1105                         err = PTR_ERR(dst);
1106                         dst = NULL;
1107                         goto tx_err_link_failure;
1108                 }
1109                 if (t->parms.collect_md &&
1110                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1111                                        &fl6->daddr, 0, &fl6->saddr))
1112                         goto tx_err_link_failure;
1113                 ndst = dst;
1114         }
1115 
1116         tdev = dst->dev;
1117 
1118         if (tdev == dev) {
1119                 stats->collisions++;
1120                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1121                                      t->parms.name);
1122                 goto tx_err_dst_release;
1123         }
1124         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1125         if (encap_limit >= 0) {
1126                 max_headroom += 8;
1127                 mtu -= 8;
1128         }
1129         if (skb->protocol == htons(ETH_P_IPV6)) {
1130                 if (mtu < IPV6_MIN_MTU)
1131                         mtu = IPV6_MIN_MTU;
1132         } else if (mtu < 576) {
1133                 mtu = 576;
1134         }
1135 
1136         skb_dst_update_pmtu(skb, mtu);
1137         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1138                 *pmtu = mtu;
1139                 err = -EMSGSIZE;
1140                 goto tx_err_dst_release;
1141         }
1142 
1143         if (t->err_count > 0) {
1144                 if (time_before(jiffies,
1145                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1146                         t->err_count--;
1147 
1148                         dst_link_failure(skb);
1149                 } else {
1150                         t->err_count = 0;
1151                 }
1152         }
1153 
1154         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1155 
1156         /*
1157          * Okay, now see if we can stuff it in the buffer as-is.
1158          */
1159         max_headroom += LL_RESERVED_SPACE(tdev);
1160 
1161         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1162             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1163                 struct sk_buff *new_skb;
1164 
1165                 new_skb = skb_realloc_headroom(skb, max_headroom);
1166                 if (!new_skb)
1167                         goto tx_err_dst_release;
1168 
1169                 if (skb->sk)
1170                         skb_set_owner_w(new_skb, skb->sk);
1171                 consume_skb(skb);
1172                 skb = new_skb;
1173         }
1174 
1175         if (t->parms.collect_md) {
1176                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1177                         goto tx_err_dst_release;
1178         } else {
1179                 if (use_cache && ndst)
1180                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1181         }
1182         skb_dst_set(skb, dst);
1183 
1184         if (encap_limit >= 0) {
1185                 init_tel_txopt(&opt, encap_limit);
1186                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1187         }
1188         hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1189 
1190         /* Calculate max headroom for all the headers and adjust
1191          * needed_headroom if necessary.
1192          */
1193         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1194                         + dst->header_len + t->hlen;
1195         if (max_headroom > dev->needed_headroom)
1196                 dev->needed_headroom = max_headroom;
1197 
1198         err = ip6_tnl_encap(skb, t, &proto, fl6);
1199         if (err)
1200                 return err;
1201 
1202         skb_push(skb, sizeof(struct ipv6hdr));
1203         skb_reset_network_header(skb);
1204         ipv6h = ipv6_hdr(skb);
1205         ip6_flow_hdr(ipv6h, dsfield,
1206                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1207         ipv6h->hop_limit = hop_limit;
1208         ipv6h->nexthdr = proto;
1209         ipv6h->saddr = fl6->saddr;
1210         ipv6h->daddr = fl6->daddr;
1211         ip6tunnel_xmit(NULL, skb, dev);
1212         return 0;
1213 tx_err_link_failure:
1214         stats->tx_carrier_errors++;
1215         dst_link_failure(skb);
1216 tx_err_dst_release:
1217         dst_release(dst);
1218         return err;
1219 }
1220 EXPORT_SYMBOL(ip6_tnl_xmit);
1221 
1222 static inline int
1223 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1224 {
1225         struct ip6_tnl *t = netdev_priv(dev);
1226         const struct iphdr  *iph = ip_hdr(skb);
1227         int encap_limit = -1;
1228         struct flowi6 fl6;
1229         __u8 dsfield;
1230         __u32 mtu;
1231         u8 tproto;
1232         int err;
1233 
1234         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1235 
1236         tproto = READ_ONCE(t->parms.proto);
1237         if (tproto != IPPROTO_IPIP && tproto != 0)
1238                 return -1;
1239 
1240         if (t->parms.collect_md) {
1241                 struct ip_tunnel_info *tun_info;
1242                 const struct ip_tunnel_key *key;
1243 
1244                 tun_info = skb_tunnel_info(skb);
1245                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1246                              ip_tunnel_info_af(tun_info) != AF_INET6))
1247                         return -1;
1248                 key = &tun_info->key;
1249                 memset(&fl6, 0, sizeof(fl6));
1250                 fl6.flowi6_proto = IPPROTO_IPIP;
1251                 fl6.daddr = key->u.ipv6.dst;
1252                 fl6.flowlabel = key->label;
1253                 dsfield =  key->tos;
1254         } else {
1255                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1256                         encap_limit = t->parms.encap_limit;
1257 
1258                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1259                 fl6.flowi6_proto = IPPROTO_IPIP;
1260 
1261                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1262                         dsfield = ipv4_get_dsfield(iph);
1263                 else
1264                         dsfield = ip6_tclass(t->parms.flowinfo);
1265                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1266                         fl6.flowi6_mark = skb->mark;
1267                 else
1268                         fl6.flowi6_mark = t->parms.fwmark;
1269         }
1270 
1271         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1272 
1273         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1274                 return -1;
1275 
1276         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1277 
1278         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1279 
1280         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1281                            IPPROTO_IPIP);
1282         if (err != 0) {
1283                 /* XXX: send ICMP error even if DF is not set. */
1284                 if (err == -EMSGSIZE)
1285                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1286                                   htonl(mtu));
1287                 return -1;
1288         }
1289 
1290         return 0;
1291 }
1292 
1293 static inline int
1294 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1295 {
1296         struct ip6_tnl *t = netdev_priv(dev);
1297         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1298         int encap_limit = -1;
1299         __u16 offset;
1300         struct flowi6 fl6;
1301         __u8 dsfield;
1302         __u32 mtu;
1303         u8 tproto;
1304         int err;
1305 
1306         tproto = READ_ONCE(t->parms.proto);
1307         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1308             ip6_tnl_addr_conflict(t, ipv6h))
1309                 return -1;
1310 
1311         if (t->parms.collect_md) {
1312                 struct ip_tunnel_info *tun_info;
1313                 const struct ip_tunnel_key *key;
1314 
1315                 tun_info = skb_tunnel_info(skb);
1316                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1317                              ip_tunnel_info_af(tun_info) != AF_INET6))
1318                         return -1;
1319                 key = &tun_info->key;
1320                 memset(&fl6, 0, sizeof(fl6));
1321                 fl6.flowi6_proto = IPPROTO_IPV6;
1322                 fl6.daddr = key->u.ipv6.dst;
1323                 fl6.flowlabel = key->label;
1324                 dsfield = key->tos;
1325         } else {
1326                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1327                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1328                 ipv6h = ipv6_hdr(skb);
1329                 if (offset > 0) {
1330                         struct ipv6_tlv_tnl_enc_lim *tel;
1331 
1332                         tel = (void *)&skb_network_header(skb)[offset];
1333                         if (tel->encap_limit == 0) {
1334                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1335                                             ICMPV6_HDR_FIELD, offset + 2);
1336                                 return -1;
1337                         }
1338                         encap_limit = tel->encap_limit - 1;
1339                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1340                         encap_limit = t->parms.encap_limit;
1341                 }
1342 
1343                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1344                 fl6.flowi6_proto = IPPROTO_IPV6;
1345 
1346                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1347                         dsfield = ipv6_get_dsfield(ipv6h);
1348                 else
1349                         dsfield = ip6_tclass(t->parms.flowinfo);
1350                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1351                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1352                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1353                         fl6.flowi6_mark = skb->mark;
1354                 else
1355                         fl6.flowi6_mark = t->parms.fwmark;
1356         }
1357 
1358         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1359 
1360         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1361                 return -1;
1362 
1363         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1364 
1365         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1366 
1367         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1368                            IPPROTO_IPV6);
1369         if (err != 0) {
1370                 if (err == -EMSGSIZE)
1371                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1372                 return -1;
1373         }
1374 
1375         return 0;
1376 }
1377 
1378 static netdev_tx_t
1379 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1380 {
1381         struct ip6_tnl *t = netdev_priv(dev);
1382         struct net_device_stats *stats = &t->dev->stats;
1383         int ret;
1384 
1385         switch (skb->protocol) {
1386         case htons(ETH_P_IP):
1387                 ret = ip4ip6_tnl_xmit(skb, dev);
1388                 break;
1389         case htons(ETH_P_IPV6):
1390                 ret = ip6ip6_tnl_xmit(skb, dev);
1391                 break;
1392         default:
1393                 goto tx_err;
1394         }
1395 
1396         if (ret < 0)
1397                 goto tx_err;
1398 
1399         return NETDEV_TX_OK;
1400 
1401 tx_err:
1402         stats->tx_errors++;
1403         stats->tx_dropped++;
1404         kfree_skb(skb);
1405         return NETDEV_TX_OK;
1406 }
1407 
1408 static void ip6_tnl_link_config(struct ip6_tnl *t)
1409 {
1410         struct net_device *dev = t->dev;
1411         struct __ip6_tnl_parm *p = &t->parms;
1412         struct flowi6 *fl6 = &t->fl.u.ip6;
1413         int t_hlen;
1414 
1415         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1416         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1417 
1418         /* Set up flowi template */
1419         fl6->saddr = p->laddr;
1420         fl6->daddr = p->raddr;
1421         fl6->flowi6_oif = p->link;
1422         fl6->flowlabel = 0;
1423 
1424         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1425                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1426         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1427                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1428 
1429         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1430         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1431 
1432         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1433                 dev->flags |= IFF_POINTOPOINT;
1434         else
1435                 dev->flags &= ~IFF_POINTOPOINT;
1436 
1437         t->tun_hlen = 0;
1438         t->hlen = t->encap_hlen + t->tun_hlen;
1439         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1440 
1441         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1442                 int strict = (ipv6_addr_type(&p->raddr) &
1443                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1444 
1445                 struct rt6_info *rt = rt6_lookup(t->net,
1446                                                  &p->raddr, &p->laddr,
1447                                                  p->link, strict);
1448 
1449                 if (!rt)
1450                         return;
1451 
1452                 if (rt->dst.dev) {
1453                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1454                                 t_hlen;
1455 
1456                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1457                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1458                                 dev->mtu -= 8;
1459 
1460                         if (dev->mtu < IPV6_MIN_MTU)
1461                                 dev->mtu = IPV6_MIN_MTU;
1462                 }
1463                 ip6_rt_put(rt);
1464         }
1465 }
1466 
1467 /**
1468  * ip6_tnl_change - update the tunnel parameters
1469  *   @t: tunnel to be changed
1470  *   @p: tunnel configuration parameters
1471  *
1472  * Description:
1473  *   ip6_tnl_change() updates the tunnel parameters
1474  **/
1475 
1476 static int
1477 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1478 {
1479         t->parms.laddr = p->laddr;
1480         t->parms.raddr = p->raddr;
1481         t->parms.flags = p->flags;
1482         t->parms.hop_limit = p->hop_limit;
1483         t->parms.encap_limit = p->encap_limit;
1484         t->parms.flowinfo = p->flowinfo;
1485         t->parms.link = p->link;
1486         t->parms.proto = p->proto;
1487         t->parms.fwmark = p->fwmark;
1488         dst_cache_reset(&t->dst_cache);
1489         ip6_tnl_link_config(t);
1490         return 0;
1491 }
1492 
1493 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1494 {
1495         struct net *net = t->net;
1496         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1497         int err;
1498 
1499         ip6_tnl_unlink(ip6n, t);
1500         synchronize_net();
1501         err = ip6_tnl_change(t, p);
1502         ip6_tnl_link(ip6n, t);
1503         netdev_state_change(t->dev);
1504         return err;
1505 }
1506 
1507 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1508 {
1509         /* for default tnl0 device allow to change only the proto */
1510         t->parms.proto = p->proto;
1511         netdev_state_change(t->dev);
1512         return 0;
1513 }
1514 
1515 static void
1516 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1517 {
1518         p->laddr = u->laddr;
1519         p->raddr = u->raddr;
1520         p->flags = u->flags;
1521         p->hop_limit = u->hop_limit;
1522         p->encap_limit = u->encap_limit;
1523         p->flowinfo = u->flowinfo;
1524         p->link = u->link;
1525         p->proto = u->proto;
1526         memcpy(p->name, u->name, sizeof(u->name));
1527 }
1528 
1529 static void
1530 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1531 {
1532         u->laddr = p->laddr;
1533         u->raddr = p->raddr;
1534         u->flags = p->flags;
1535         u->hop_limit = p->hop_limit;
1536         u->encap_limit = p->encap_limit;
1537         u->flowinfo = p->flowinfo;
1538         u->link = p->link;
1539         u->proto = p->proto;
1540         memcpy(u->name, p->name, sizeof(u->name));
1541 }
1542 
1543 /**
1544  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1545  *   @dev: virtual device associated with tunnel
1546  *   @ifr: parameters passed from userspace
1547  *   @cmd: command to be performed
1548  *
1549  * Description:
1550  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1551  *   from userspace.
1552  *
1553  *   The possible commands are the following:
1554  *     %SIOCGETTUNNEL: get tunnel parameters for device
1555  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1556  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1557  *     %SIOCDELTUNNEL: delete tunnel
1558  *
1559  *   The fallback device "ip6tnl0", created during module
1560  *   initialization, can be used for creating other tunnel devices.
1561  *
1562  * Return:
1563  *   0 on success,
1564  *   %-EFAULT if unable to copy data to or from userspace,
1565  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1566  *   %-EINVAL if passed tunnel parameters are invalid,
1567  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1568  *   %-ENODEV if attempting to change or delete a nonexisting device
1569  **/
1570 
1571 static int
1572 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1573 {
1574         int err = 0;
1575         struct ip6_tnl_parm p;
1576         struct __ip6_tnl_parm p1;
1577         struct ip6_tnl *t = netdev_priv(dev);
1578         struct net *net = t->net;
1579         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1580 
1581         memset(&p1, 0, sizeof(p1));
1582 
1583         switch (cmd) {
1584         case SIOCGETTUNNEL:
1585                 if (dev == ip6n->fb_tnl_dev) {
1586                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1587                                 err = -EFAULT;
1588                                 break;
1589                         }
1590                         ip6_tnl_parm_from_user(&p1, &p);
1591                         t = ip6_tnl_locate(net, &p1, 0);
1592                         if (IS_ERR(t))
1593                                 t = netdev_priv(dev);
1594                 } else {
1595                         memset(&p, 0, sizeof(p));
1596                 }
1597                 ip6_tnl_parm_to_user(&p, &t->parms);
1598                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1599                         err = -EFAULT;
1600                 }
1601                 break;
1602         case SIOCADDTUNNEL:
1603         case SIOCCHGTUNNEL:
1604                 err = -EPERM;
1605                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1606                         break;
1607                 err = -EFAULT;
1608                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1609                         break;
1610                 err = -EINVAL;
1611                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1612                     p.proto != 0)
1613                         break;
1614                 ip6_tnl_parm_from_user(&p1, &p);
1615                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1616                 if (cmd == SIOCCHGTUNNEL) {
1617                         if (!IS_ERR(t)) {
1618                                 if (t->dev != dev) {
1619                                         err = -EEXIST;
1620                                         break;
1621                                 }
1622                         } else
1623                                 t = netdev_priv(dev);
1624                         if (dev == ip6n->fb_tnl_dev)
1625                                 err = ip6_tnl0_update(t, &p1);
1626                         else
1627                                 err = ip6_tnl_update(t, &p1);
1628                 }
1629                 if (!IS_ERR(t)) {
1630                         err = 0;
1631                         ip6_tnl_parm_to_user(&p, &t->parms);
1632                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1633                                 err = -EFAULT;
1634 
1635                 } else {
1636                         err = PTR_ERR(t);
1637                 }
1638                 break;
1639         case SIOCDELTUNNEL:
1640                 err = -EPERM;
1641                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1642                         break;
1643 
1644                 if (dev == ip6n->fb_tnl_dev) {
1645                         err = -EFAULT;
1646                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1647                                 break;
1648                         err = -ENOENT;
1649                         ip6_tnl_parm_from_user(&p1, &p);
1650                         t = ip6_tnl_locate(net, &p1, 0);
1651                         if (IS_ERR(t))
1652                                 break;
1653                         err = -EPERM;
1654                         if (t->dev == ip6n->fb_tnl_dev)
1655                                 break;
1656                         dev = t->dev;
1657                 }
1658                 err = 0;
1659                 unregister_netdevice(dev);
1660                 break;
1661         default:
1662                 err = -EINVAL;
1663         }
1664         return err;
1665 }
1666 
1667 /**
1668  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1669  *   @dev: virtual device associated with tunnel
1670  *   @new_mtu: the new mtu
1671  *
1672  * Return:
1673  *   0 on success,
1674  *   %-EINVAL if mtu too small
1675  **/
1676 
1677 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1678 {
1679         struct ip6_tnl *tnl = netdev_priv(dev);
1680 
1681         if (tnl->parms.proto == IPPROTO_IPV6) {
1682                 if (new_mtu < IPV6_MIN_MTU)
1683                         return -EINVAL;
1684         } else {
1685                 if (new_mtu < ETH_MIN_MTU)
1686                         return -EINVAL;
1687         }
1688         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1689                 return -EINVAL;
1690         dev->mtu = new_mtu;
1691         return 0;
1692 }
1693 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1694 
1695 int ip6_tnl_get_iflink(const struct net_device *dev)
1696 {
1697         struct ip6_tnl *t = netdev_priv(dev);
1698 
1699         return t->parms.link;
1700 }
1701 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1702 
1703 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1704                           unsigned int num)
1705 {
1706         if (num >= MAX_IPTUN_ENCAP_OPS)
1707                 return -ERANGE;
1708 
1709         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1710                         &ip6tun_encaps[num],
1711                         NULL, ops) ? 0 : -1;
1712 }
1713 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1714 
1715 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1716                           unsigned int num)
1717 {
1718         int ret;
1719 
1720         if (num >= MAX_IPTUN_ENCAP_OPS)
1721                 return -ERANGE;
1722 
1723         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1724                        &ip6tun_encaps[num],
1725                        ops, NULL) == ops) ? 0 : -1;
1726 
1727         synchronize_net();
1728 
1729         return ret;
1730 }
1731 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1732 
1733 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1734                         struct ip_tunnel_encap *ipencap)
1735 {
1736         int hlen;
1737 
1738         memset(&t->encap, 0, sizeof(t->encap));
1739 
1740         hlen = ip6_encap_hlen(ipencap);
1741         if (hlen < 0)
1742                 return hlen;
1743 
1744         t->encap.type = ipencap->type;
1745         t->encap.sport = ipencap->sport;
1746         t->encap.dport = ipencap->dport;
1747         t->encap.flags = ipencap->flags;
1748 
1749         t->encap_hlen = hlen;
1750         t->hlen = t->encap_hlen + t->tun_hlen;
1751 
1752         return 0;
1753 }
1754 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1755 
1756 static const struct net_device_ops ip6_tnl_netdev_ops = {
1757         .ndo_init       = ip6_tnl_dev_init,
1758         .ndo_uninit     = ip6_tnl_dev_uninit,
1759         .ndo_start_xmit = ip6_tnl_start_xmit,
1760         .ndo_do_ioctl   = ip6_tnl_ioctl,
1761         .ndo_change_mtu = ip6_tnl_change_mtu,
1762         .ndo_get_stats  = ip6_get_stats,
1763         .ndo_get_iflink = ip6_tnl_get_iflink,
1764 };
1765 
1766 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1767                          NETIF_F_FRAGLIST |     \
1768                          NETIF_F_HIGHDMA |      \
1769                          NETIF_F_GSO_SOFTWARE | \
1770                          NETIF_F_HW_CSUM)
1771 
1772 /**
1773  * ip6_tnl_dev_setup - setup virtual tunnel device
1774  *   @dev: virtual device associated with tunnel
1775  *
1776  * Description:
1777  *   Initialize function pointers and device parameters
1778  **/
1779 
1780 static void ip6_tnl_dev_setup(struct net_device *dev)
1781 {
1782         dev->netdev_ops = &ip6_tnl_netdev_ops;
1783         dev->needs_free_netdev = true;
1784         dev->priv_destructor = ip6_dev_free;
1785 
1786         dev->type = ARPHRD_TUNNEL6;
1787         dev->flags |= IFF_NOARP;
1788         dev->addr_len = sizeof(struct in6_addr);
1789         dev->features |= NETIF_F_LLTX;
1790         netif_keep_dst(dev);
1791 
1792         dev->features           |= IPXIPX_FEATURES;
1793         dev->hw_features        |= IPXIPX_FEATURES;
1794 
1795         /* This perm addr will be used as interface identifier by IPv6 */
1796         dev->addr_assign_type = NET_ADDR_RANDOM;
1797         eth_random_addr(dev->perm_addr);
1798 }
1799 
1800 
1801 /**
1802  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1803  *   @dev: virtual device associated with tunnel
1804  **/
1805 
1806 static inline int
1807 ip6_tnl_dev_init_gen(struct net_device *dev)
1808 {
1809         struct ip6_tnl *t = netdev_priv(dev);
1810         int ret;
1811         int t_hlen;
1812 
1813         t->dev = dev;
1814         t->net = dev_net(dev);
1815         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1816         if (!dev->tstats)
1817                 return -ENOMEM;
1818 
1819         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1820         if (ret)
1821                 goto free_stats;
1822 
1823         ret = gro_cells_init(&t->gro_cells, dev);
1824         if (ret)
1825                 goto destroy_dst;
1826 
1827         t->tun_hlen = 0;
1828         t->hlen = t->encap_hlen + t->tun_hlen;
1829         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1830 
1831         dev->type = ARPHRD_TUNNEL6;
1832         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1833         dev->mtu = ETH_DATA_LEN - t_hlen;
1834         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1835                 dev->mtu -= 8;
1836         dev->min_mtu = ETH_MIN_MTU;
1837         dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1838 
1839         return 0;
1840 
1841 destroy_dst:
1842         dst_cache_destroy(&t->dst_cache);
1843 free_stats:
1844         free_percpu(dev->tstats);
1845         dev->tstats = NULL;
1846 
1847         return ret;
1848 }
1849 
1850 /**
1851  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1852  *   @dev: virtual device associated with tunnel
1853  **/
1854 
1855 static int ip6_tnl_dev_init(struct net_device *dev)
1856 {
1857         struct ip6_tnl *t = netdev_priv(dev);
1858         int err = ip6_tnl_dev_init_gen(dev);
1859 
1860         if (err)
1861                 return err;
1862         ip6_tnl_link_config(t);
1863         if (t->parms.collect_md) {
1864                 dev->features |= NETIF_F_NETNS_LOCAL;
1865                 netif_keep_dst(dev);
1866         }
1867         return 0;
1868 }
1869 
1870 /**
1871  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1872  *   @dev: fallback device
1873  *
1874  * Return: 0
1875  **/
1876 
1877 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1878 {
1879         struct ip6_tnl *t = netdev_priv(dev);
1880         struct net *net = dev_net(dev);
1881         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1882 
1883         t->parms.proto = IPPROTO_IPV6;
1884         dev_hold(dev);
1885 
1886         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1887         return 0;
1888 }
1889 
1890 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1891                             struct netlink_ext_ack *extack)
1892 {
1893         u8 proto;
1894 
1895         if (!data || !data[IFLA_IPTUN_PROTO])
1896                 return 0;
1897 
1898         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1899         if (proto != IPPROTO_IPV6 &&
1900             proto != IPPROTO_IPIP &&
1901             proto != 0)
1902                 return -EINVAL;
1903 
1904         return 0;
1905 }
1906 
1907 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1908                                   struct __ip6_tnl_parm *parms)
1909 {
1910         memset(parms, 0, sizeof(*parms));
1911 
1912         if (!data)
1913                 return;
1914 
1915         if (data[IFLA_IPTUN_LINK])
1916                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1917 
1918         if (data[IFLA_IPTUN_LOCAL])
1919                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1920 
1921         if (data[IFLA_IPTUN_REMOTE])
1922                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1923 
1924         if (data[IFLA_IPTUN_TTL])
1925                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1926 
1927         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1928                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1929 
1930         if (data[IFLA_IPTUN_FLOWINFO])
1931                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1932 
1933         if (data[IFLA_IPTUN_FLAGS])
1934                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1935 
1936         if (data[IFLA_IPTUN_PROTO])
1937                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1938 
1939         if (data[IFLA_IPTUN_COLLECT_METADATA])
1940                 parms->collect_md = true;
1941 
1942         if (data[IFLA_IPTUN_FWMARK])
1943                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1944 }
1945 
1946 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1947                                         struct ip_tunnel_encap *ipencap)
1948 {
1949         bool ret = false;
1950 
1951         memset(ipencap, 0, sizeof(*ipencap));
1952 
1953         if (!data)
1954                 return ret;
1955 
1956         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1957                 ret = true;
1958                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1959         }
1960 
1961         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1962                 ret = true;
1963                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1964         }
1965 
1966         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1967                 ret = true;
1968                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1969         }
1970 
1971         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1972                 ret = true;
1973                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1974         }
1975 
1976         return ret;
1977 }
1978 
1979 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1980                            struct nlattr *tb[], struct nlattr *data[],
1981                            struct netlink_ext_ack *extack)
1982 {
1983         struct net *net = dev_net(dev);
1984         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1985         struct ip6_tnl *nt, *t;
1986         struct ip_tunnel_encap ipencap;
1987 
1988         nt = netdev_priv(dev);
1989 
1990         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1991                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1992 
1993                 if (err < 0)
1994                         return err;
1995         }
1996 
1997         ip6_tnl_netlink_parms(data, &nt->parms);
1998 
1999         if (nt->parms.collect_md) {
2000                 if (rtnl_dereference(ip6n->collect_md_tun))
2001                         return -EEXIST;
2002         } else {
2003                 t = ip6_tnl_locate(net, &nt->parms, 0);
2004                 if (!IS_ERR(t))
2005                         return -EEXIST;
2006         }
2007 
2008         return ip6_tnl_create2(dev);
2009 }
2010 
2011 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2012                               struct nlattr *data[],
2013                               struct netlink_ext_ack *extack)
2014 {
2015         struct ip6_tnl *t = netdev_priv(dev);
2016         struct __ip6_tnl_parm p;
2017         struct net *net = t->net;
2018         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2019         struct ip_tunnel_encap ipencap;
2020 
2021         if (dev == ip6n->fb_tnl_dev)
2022                 return -EINVAL;
2023 
2024         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2025                 int err = ip6_tnl_encap_setup(t, &ipencap);
2026 
2027                 if (err < 0)
2028                         return err;
2029         }
2030         ip6_tnl_netlink_parms(data, &p);
2031         if (p.collect_md)
2032                 return -EINVAL;
2033 
2034         t = ip6_tnl_locate(net, &p, 0);
2035         if (!IS_ERR(t)) {
2036                 if (t->dev != dev)
2037                         return -EEXIST;
2038         } else
2039                 t = netdev_priv(dev);
2040 
2041         return ip6_tnl_update(t, &p);
2042 }
2043 
2044 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2045 {
2046         struct net *net = dev_net(dev);
2047         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2048 
2049         if (dev != ip6n->fb_tnl_dev)
2050                 unregister_netdevice_queue(dev, head);
2051 }
2052 
2053 static size_t ip6_tnl_get_size(const struct net_device *dev)
2054 {
2055         return
2056                 /* IFLA_IPTUN_LINK */
2057                 nla_total_size(4) +
2058                 /* IFLA_IPTUN_LOCAL */
2059                 nla_total_size(sizeof(struct in6_addr)) +
2060                 /* IFLA_IPTUN_REMOTE */
2061                 nla_total_size(sizeof(struct in6_addr)) +
2062                 /* IFLA_IPTUN_TTL */
2063                 nla_total_size(1) +
2064                 /* IFLA_IPTUN_ENCAP_LIMIT */
2065                 nla_total_size(1) +
2066                 /* IFLA_IPTUN_FLOWINFO */
2067                 nla_total_size(4) +
2068                 /* IFLA_IPTUN_FLAGS */
2069                 nla_total_size(4) +
2070                 /* IFLA_IPTUN_PROTO */
2071                 nla_total_size(1) +
2072                 /* IFLA_IPTUN_ENCAP_TYPE */
2073                 nla_total_size(2) +
2074                 /* IFLA_IPTUN_ENCAP_FLAGS */
2075                 nla_total_size(2) +
2076                 /* IFLA_IPTUN_ENCAP_SPORT */
2077                 nla_total_size(2) +
2078                 /* IFLA_IPTUN_ENCAP_DPORT */
2079                 nla_total_size(2) +
2080                 /* IFLA_IPTUN_COLLECT_METADATA */
2081                 nla_total_size(0) +
2082                 /* IFLA_IPTUN_FWMARK */
2083                 nla_total_size(4) +
2084                 0;
2085 }
2086 
2087 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2088 {
2089         struct ip6_tnl *tunnel = netdev_priv(dev);
2090         struct __ip6_tnl_parm *parm = &tunnel->parms;
2091 
2092         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2093             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2094             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2095             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2096             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2097             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2098             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2099             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2100             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2101                 goto nla_put_failure;
2102 
2103         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2104             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2105             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2106             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2107                 goto nla_put_failure;
2108 
2109         if (parm->collect_md)
2110                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2111                         goto nla_put_failure;
2112 
2113         return 0;
2114 
2115 nla_put_failure:
2116         return -EMSGSIZE;
2117 }
2118 
2119 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2120 {
2121         struct ip6_tnl *tunnel = netdev_priv(dev);
2122 
2123         return tunnel->net;
2124 }
2125 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2126 
2127 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2128         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2129         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2130         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2131         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2132         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2133         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2134         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2135         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2136         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2137         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2138         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2139         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2140         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2141         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2142 };
2143 
2144 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2145         .kind           = "ip6tnl",
2146         .maxtype        = IFLA_IPTUN_MAX,
2147         .policy         = ip6_tnl_policy,
2148         .priv_size      = sizeof(struct ip6_tnl),
2149         .setup          = ip6_tnl_dev_setup,
2150         .validate       = ip6_tnl_validate,
2151         .newlink        = ip6_tnl_newlink,
2152         .changelink     = ip6_tnl_changelink,
2153         .dellink        = ip6_tnl_dellink,
2154         .get_size       = ip6_tnl_get_size,
2155         .fill_info      = ip6_tnl_fill_info,
2156         .get_link_net   = ip6_tnl_get_link_net,
2157 };
2158 
2159 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2160         .handler        = ip4ip6_rcv,
2161         .err_handler    = ip4ip6_err,
2162         .priority       =       1,
2163 };
2164 
2165 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2166         .handler        = ip6ip6_rcv,
2167         .err_handler    = ip6ip6_err,
2168         .priority       =       1,
2169 };
2170 
2171 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2172 {
2173         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2174         struct net_device *dev, *aux;
2175         int h;
2176         struct ip6_tnl *t;
2177 
2178         for_each_netdev_safe(net, dev, aux)
2179                 if (dev->rtnl_link_ops == &ip6_link_ops)
2180                         unregister_netdevice_queue(dev, list);
2181 
2182         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2183                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2184                 while (t) {
2185                         /* If dev is in the same netns, it has already
2186                          * been added to the list by the previous loop.
2187                          */
2188                         if (!net_eq(dev_net(t->dev), net))
2189                                 unregister_netdevice_queue(t->dev, list);
2190                         t = rtnl_dereference(t->next);
2191                 }
2192         }
2193 }
2194 
2195 static int __net_init ip6_tnl_init_net(struct net *net)
2196 {
2197         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2198         struct ip6_tnl *t = NULL;
2199         int err;
2200 
2201         ip6n->tnls[0] = ip6n->tnls_wc;
2202         ip6n->tnls[1] = ip6n->tnls_r_l;
2203 
2204         err = -ENOMEM;
2205         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2206                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2207 
2208         if (!ip6n->fb_tnl_dev)
2209                 goto err_alloc_dev;
2210         dev_net_set(ip6n->fb_tnl_dev, net);
2211         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2212         /* FB netdevice is special: we have one, and only one per netns.
2213          * Allowing to move it to another netns is clearly unsafe.
2214          */
2215         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2216 
2217         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2218         if (err < 0)
2219                 goto err_register;
2220 
2221         err = register_netdev(ip6n->fb_tnl_dev);
2222         if (err < 0)
2223                 goto err_register;
2224 
2225         t = netdev_priv(ip6n->fb_tnl_dev);
2226 
2227         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2228         return 0;
2229 
2230 err_register:
2231         free_netdev(ip6n->fb_tnl_dev);
2232 err_alloc_dev:
2233         return err;
2234 }
2235 
2236 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2237 {
2238         struct net *net;
2239         LIST_HEAD(list);
2240 
2241         rtnl_lock();
2242         list_for_each_entry(net, net_list, exit_list)
2243                 ip6_tnl_destroy_tunnels(net, &list);
2244         unregister_netdevice_many(&list);
2245         rtnl_unlock();
2246 }
2247 
2248 static struct pernet_operations ip6_tnl_net_ops = {
2249         .init = ip6_tnl_init_net,
2250         .exit_batch = ip6_tnl_exit_batch_net,
2251         .id   = &ip6_tnl_net_id,
2252         .size = sizeof(struct ip6_tnl_net),
2253 };
2254 
2255 /**
2256  * ip6_tunnel_init - register protocol and reserve needed resources
2257  *
2258  * Return: 0 on success
2259  **/
2260 
2261 static int __init ip6_tunnel_init(void)
2262 {
2263         int  err;
2264 
2265         if (!ipv6_mod_enabled())
2266                 return -EOPNOTSUPP;
2267 
2268         err = register_pernet_device(&ip6_tnl_net_ops);
2269         if (err < 0)
2270                 goto out_pernet;
2271 
2272         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2273         if (err < 0) {
2274                 pr_err("%s: can't register ip4ip6\n", __func__);
2275                 goto out_ip4ip6;
2276         }
2277 
2278         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2279         if (err < 0) {
2280                 pr_err("%s: can't register ip6ip6\n", __func__);
2281                 goto out_ip6ip6;
2282         }
2283         err = rtnl_link_register(&ip6_link_ops);
2284         if (err < 0)
2285                 goto rtnl_link_failed;
2286 
2287         return 0;
2288 
2289 rtnl_link_failed:
2290         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2291 out_ip6ip6:
2292         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2293 out_ip4ip6:
2294         unregister_pernet_device(&ip6_tnl_net_ops);
2295 out_pernet:
2296         return err;
2297 }
2298 
2299 /**
2300  * ip6_tunnel_cleanup - free resources and unregister protocol
2301  **/
2302 
2303 static void __exit ip6_tunnel_cleanup(void)
2304 {
2305         rtnl_link_unregister(&ip6_link_ops);
2306         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2307                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2308 
2309         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2310                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2311 
2312         unregister_pernet_device(&ip6_tnl_net_ops);
2313 }
2314 
2315 module_init(ip6_tunnel_init);
2316 module_exit(ip6_tunnel_cleanup);
2317 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp