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

TOMOYO Linux Cross Reference
Linux/net/ipv6/ip6_gre.c

Version: ~ [ linux-6.0-rc1 ] ~ [ linux-5.19.1 ] ~ [ linux-5.18.17 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.60 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.136 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.210 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.255 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.290 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.325 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.302 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *      GRE over IPv6 protocol decoder.
  3  *
  4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
  5  *
  6  *      This program is free software; you can redistribute it and/or
  7  *      modify it under the terms of the GNU General Public License
  8  *      as published by the Free Software Foundation; either version
  9  *      2 of the License, or (at your option) any later version.
 10  *
 11  */
 12 
 13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 14 
 15 #include <linux/capability.h>
 16 #include <linux/module.h>
 17 #include <linux/types.h>
 18 #include <linux/kernel.h>
 19 #include <linux/slab.h>
 20 #include <linux/uaccess.h>
 21 #include <linux/skbuff.h>
 22 #include <linux/netdevice.h>
 23 #include <linux/in.h>
 24 #include <linux/tcp.h>
 25 #include <linux/udp.h>
 26 #include <linux/if_arp.h>
 27 #include <linux/init.h>
 28 #include <linux/in6.h>
 29 #include <linux/inetdevice.h>
 30 #include <linux/igmp.h>
 31 #include <linux/netfilter_ipv4.h>
 32 #include <linux/etherdevice.h>
 33 #include <linux/if_ether.h>
 34 #include <linux/hash.h>
 35 #include <linux/if_tunnel.h>
 36 #include <linux/ip6_tunnel.h>
 37 
 38 #include <net/sock.h>
 39 #include <net/ip.h>
 40 #include <net/ip_tunnels.h>
 41 #include <net/icmp.h>
 42 #include <net/protocol.h>
 43 #include <net/addrconf.h>
 44 #include <net/arp.h>
 45 #include <net/checksum.h>
 46 #include <net/dsfield.h>
 47 #include <net/inet_ecn.h>
 48 #include <net/xfrm.h>
 49 #include <net/net_namespace.h>
 50 #include <net/netns/generic.h>
 51 #include <net/rtnetlink.h>
 52 
 53 #include <net/ipv6.h>
 54 #include <net/ip6_fib.h>
 55 #include <net/ip6_route.h>
 56 #include <net/ip6_tunnel.h>
 57 
 58 
 59 static bool log_ecn_error = true;
 60 module_param(log_ecn_error, bool, 0644);
 61 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
 62 
 63 #define HASH_SIZE_SHIFT  5
 64 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
 65 
 66 static int ip6gre_net_id __read_mostly;
 67 struct ip6gre_net {
 68         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
 69 
 70         struct net_device *fb_tunnel_dev;
 71 };
 72 
 73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
 74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
 75 static int ip6gre_tunnel_init(struct net_device *dev);
 76 static void ip6gre_tunnel_setup(struct net_device *dev);
 77 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
 78 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
 79 
 80 /* Tunnel hash table */
 81 
 82 /*
 83    4 hash tables:
 84 
 85    3: (remote,local)
 86    2: (remote,*)
 87    1: (*,local)
 88    0: (*,*)
 89 
 90    We require exact key match i.e. if a key is present in packet
 91    it will match only tunnel with the same key; if it is not present,
 92    it will match only keyless tunnel.
 93 
 94    All keysless packets, if not matched configured keyless tunnels
 95    will match fallback tunnel.
 96  */
 97 
 98 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
 99 static u32 HASH_ADDR(const struct in6_addr *addr)
100 {
101         u32 hash = ipv6_addr_hash(addr);
102 
103         return hash_32(hash, HASH_SIZE_SHIFT);
104 }
105 
106 #define tunnels_r_l     tunnels[3]
107 #define tunnels_r       tunnels[2]
108 #define tunnels_l       tunnels[1]
109 #define tunnels_wc      tunnels[0]
110 
111 /* Given src, dst and key, find appropriate for input tunnel. */
112 
113 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
114                 const struct in6_addr *remote, const struct in6_addr *local,
115                 __be32 key, __be16 gre_proto)
116 {
117         struct net *net = dev_net(dev);
118         int link = dev->ifindex;
119         unsigned int h0 = HASH_ADDR(remote);
120         unsigned int h1 = HASH_KEY(key);
121         struct ip6_tnl *t, *cand = NULL;
122         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
123         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
124                        ARPHRD_ETHER : ARPHRD_IP6GRE;
125         int score, cand_score = 4;
126 
127         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
128                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
129                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
130                     key != t->parms.i_key ||
131                     !(t->dev->flags & IFF_UP))
132                         continue;
133 
134                 if (t->dev->type != ARPHRD_IP6GRE &&
135                     t->dev->type != dev_type)
136                         continue;
137 
138                 score = 0;
139                 if (t->parms.link != link)
140                         score |= 1;
141                 if (t->dev->type != dev_type)
142                         score |= 2;
143                 if (score == 0)
144                         return t;
145 
146                 if (score < cand_score) {
147                         cand = t;
148                         cand_score = score;
149                 }
150         }
151 
152         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
153                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
154                     key != t->parms.i_key ||
155                     !(t->dev->flags & IFF_UP))
156                         continue;
157 
158                 if (t->dev->type != ARPHRD_IP6GRE &&
159                     t->dev->type != dev_type)
160                         continue;
161 
162                 score = 0;
163                 if (t->parms.link != link)
164                         score |= 1;
165                 if (t->dev->type != dev_type)
166                         score |= 2;
167                 if (score == 0)
168                         return t;
169 
170                 if (score < cand_score) {
171                         cand = t;
172                         cand_score = score;
173                 }
174         }
175 
176         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
177                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
178                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
179                                  !ipv6_addr_is_multicast(local))) ||
180                     key != t->parms.i_key ||
181                     !(t->dev->flags & IFF_UP))
182                         continue;
183 
184                 if (t->dev->type != ARPHRD_IP6GRE &&
185                     t->dev->type != dev_type)
186                         continue;
187 
188                 score = 0;
189                 if (t->parms.link != link)
190                         score |= 1;
191                 if (t->dev->type != dev_type)
192                         score |= 2;
193                 if (score == 0)
194                         return t;
195 
196                 if (score < cand_score) {
197                         cand = t;
198                         cand_score = score;
199                 }
200         }
201 
202         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
203                 if (t->parms.i_key != key ||
204                     !(t->dev->flags & IFF_UP))
205                         continue;
206 
207                 if (t->dev->type != ARPHRD_IP6GRE &&
208                     t->dev->type != dev_type)
209                         continue;
210 
211                 score = 0;
212                 if (t->parms.link != link)
213                         score |= 1;
214                 if (t->dev->type != dev_type)
215                         score |= 2;
216                 if (score == 0)
217                         return t;
218 
219                 if (score < cand_score) {
220                         cand = t;
221                         cand_score = score;
222                 }
223         }
224 
225         if (cand)
226                 return cand;
227 
228         dev = ign->fb_tunnel_dev;
229         if (dev->flags & IFF_UP)
230                 return netdev_priv(dev);
231 
232         return NULL;
233 }
234 
235 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
236                 const struct __ip6_tnl_parm *p)
237 {
238         const struct in6_addr *remote = &p->raddr;
239         const struct in6_addr *local = &p->laddr;
240         unsigned int h = HASH_KEY(p->i_key);
241         int prio = 0;
242 
243         if (!ipv6_addr_any(local))
244                 prio |= 1;
245         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
246                 prio |= 2;
247                 h ^= HASH_ADDR(remote);
248         }
249 
250         return &ign->tunnels[prio][h];
251 }
252 
253 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
254                 const struct ip6_tnl *t)
255 {
256         return __ip6gre_bucket(ign, &t->parms);
257 }
258 
259 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
260 {
261         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
262 
263         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
264         rcu_assign_pointer(*tp, t);
265 }
266 
267 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269         struct ip6_tnl __rcu **tp;
270         struct ip6_tnl *iter;
271 
272         for (tp = ip6gre_bucket(ign, t);
273              (iter = rtnl_dereference(*tp)) != NULL;
274              tp = &iter->next) {
275                 if (t == iter) {
276                         rcu_assign_pointer(*tp, t->next);
277                         break;
278                 }
279         }
280 }
281 
282 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
283                                            const struct __ip6_tnl_parm *parms,
284                                            int type)
285 {
286         const struct in6_addr *remote = &parms->raddr;
287         const struct in6_addr *local = &parms->laddr;
288         __be32 key = parms->i_key;
289         int link = parms->link;
290         struct ip6_tnl *t;
291         struct ip6_tnl __rcu **tp;
292         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
293 
294         for (tp = __ip6gre_bucket(ign, parms);
295              (t = rtnl_dereference(*tp)) != NULL;
296              tp = &t->next)
297                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
298                     ipv6_addr_equal(remote, &t->parms.raddr) &&
299                     key == t->parms.i_key &&
300                     link == t->parms.link &&
301                     type == t->dev->type)
302                         break;
303 
304         return t;
305 }
306 
307 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
308                 const struct __ip6_tnl_parm *parms, int create)
309 {
310         struct ip6_tnl *t, *nt;
311         struct net_device *dev;
312         char name[IFNAMSIZ];
313         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
314 
315         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
316         if (t && create)
317                 return NULL;
318         if (t || !create)
319                 return t;
320 
321         if (parms->name[0])
322                 strlcpy(name, parms->name, IFNAMSIZ);
323         else
324                 strcpy(name, "ip6gre%d");
325 
326         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
327                            ip6gre_tunnel_setup);
328         if (!dev)
329                 return NULL;
330 
331         dev_net_set(dev, net);
332 
333         nt = netdev_priv(dev);
334         nt->parms = *parms;
335         dev->rtnl_link_ops = &ip6gre_link_ops;
336 
337         nt->dev = dev;
338         nt->net = dev_net(dev);
339         ip6gre_tnl_link_config(nt, 1);
340 
341         if (register_netdevice(dev) < 0)
342                 goto failed_free;
343 
344         /* Can use a lockless transmit, unless we generate output sequences */
345         if (!(nt->parms.o_flags & GRE_SEQ))
346                 dev->features |= NETIF_F_LLTX;
347 
348         dev_hold(dev);
349         ip6gre_tunnel_link(ign, nt);
350         return nt;
351 
352 failed_free:
353         free_netdev(dev);
354         return NULL;
355 }
356 
357 static void ip6gre_tunnel_uninit(struct net_device *dev)
358 {
359         struct ip6_tnl *t = netdev_priv(dev);
360         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
361 
362         ip6gre_tunnel_unlink(ign, t);
363         dst_cache_reset(&t->dst_cache);
364         dev_put(dev);
365 }
366 
367 
368 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
369                 u8 type, u8 code, int offset, __be32 info)
370 {
371         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
372         __be16 *p = (__be16 *)(skb->data + offset);
373         int grehlen = offset + 4;
374         struct ip6_tnl *t;
375         __be16 flags;
376 
377         flags = p[0];
378         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
379                 if (flags&(GRE_VERSION|GRE_ROUTING))
380                         return;
381                 if (flags&GRE_KEY) {
382                         grehlen += 4;
383                         if (flags&GRE_CSUM)
384                                 grehlen += 4;
385                 }
386         }
387 
388         /* If only 8 bytes returned, keyed message will be dropped here */
389         if (!pskb_may_pull(skb, grehlen))
390                 return;
391         ipv6h = (const struct ipv6hdr *)skb->data;
392         p = (__be16 *)(skb->data + offset);
393 
394         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
395                                 flags & GRE_KEY ?
396                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
397                                 p[1]);
398         if (!t)
399                 return;
400 
401         switch (type) {
402                 __u32 teli;
403                 struct ipv6_tlv_tnl_enc_lim *tel;
404                 __u32 mtu;
405         case ICMPV6_DEST_UNREACH:
406                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
407                                     t->parms.name);
408                 break;
409         case ICMPV6_TIME_EXCEED:
410                 if (code == ICMPV6_EXC_HOPLIMIT) {
411                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
412                                             t->parms.name);
413                 }
414                 break;
415         case ICMPV6_PARAMPROB:
416                 teli = 0;
417                 if (code == ICMPV6_HDR_FIELD)
418                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
419 
420                 if (teli && teli == be32_to_cpu(info) - 2) {
421                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
422                         if (tel->encap_limit == 0) {
423                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
424                                                     t->parms.name);
425                         }
426                 } else {
427                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
428                                             t->parms.name);
429                 }
430                 break;
431         case ICMPV6_PKT_TOOBIG:
432                 mtu = be32_to_cpu(info) - offset;
433                 if (mtu < IPV6_MIN_MTU)
434                         mtu = IPV6_MIN_MTU;
435                 t->dev->mtu = mtu;
436                 break;
437         }
438 
439         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
440                 t->err_count++;
441         else
442                 t->err_count = 1;
443         t->err_time = jiffies;
444 }
445 
446 static int ip6gre_rcv(struct sk_buff *skb)
447 {
448         const struct ipv6hdr *ipv6h;
449         u8     *h;
450         __be16    flags;
451         __sum16   csum = 0;
452         __be32 key = 0;
453         u32    seqno = 0;
454         struct ip6_tnl *tunnel;
455         int    offset = 4;
456         __be16 gre_proto;
457         int err;
458 
459         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
460                 goto drop;
461 
462         ipv6h = ipv6_hdr(skb);
463         h = skb->data;
464         flags = *(__be16 *)h;
465 
466         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
467                 /* - Version must be 0.
468                    - We do not support routing headers.
469                  */
470                 if (flags&(GRE_VERSION|GRE_ROUTING))
471                         goto drop;
472 
473                 if (flags&GRE_CSUM) {
474                         csum = skb_checksum_simple_validate(skb);
475                         offset += 4;
476                 }
477                 if (flags&GRE_KEY) {
478                         key = *(__be32 *)(h + offset);
479                         offset += 4;
480                 }
481                 if (flags&GRE_SEQ) {
482                         seqno = ntohl(*(__be32 *)(h + offset));
483                         offset += 4;
484                 }
485         }
486 
487         gre_proto = *(__be16 *)(h + 2);
488 
489         tunnel = ip6gre_tunnel_lookup(skb->dev,
490                                           &ipv6h->saddr, &ipv6h->daddr, key,
491                                           gre_proto);
492         if (tunnel) {
493                 struct pcpu_sw_netstats *tstats;
494 
495                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
496                         goto drop;
497 
498                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
499                         tunnel->dev->stats.rx_dropped++;
500                         goto drop;
501                 }
502 
503                 skb->protocol = gre_proto;
504                 /* WCCP version 1 and 2 protocol decoding.
505                  * - Change protocol to IPv6
506                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
507                  */
508                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
509                         skb->protocol = htons(ETH_P_IPV6);
510                         if ((*(h + offset) & 0xF0) != 0x40)
511                                 offset += 4;
512                 }
513 
514                 skb->mac_header = skb->network_header;
515                 __pskb_pull(skb, offset);
516                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
517 
518                 if (((flags&GRE_CSUM) && csum) ||
519                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
520                         tunnel->dev->stats.rx_crc_errors++;
521                         tunnel->dev->stats.rx_errors++;
522                         goto drop;
523                 }
524                 if (tunnel->parms.i_flags&GRE_SEQ) {
525                         if (!(flags&GRE_SEQ) ||
526                             (tunnel->i_seqno &&
527                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
528                                 tunnel->dev->stats.rx_fifo_errors++;
529                                 tunnel->dev->stats.rx_errors++;
530                                 goto drop;
531                         }
532                         tunnel->i_seqno = seqno + 1;
533                 }
534 
535                 /* Warning: All skb pointers will be invalidated! */
536                 if (tunnel->dev->type == ARPHRD_ETHER) {
537                         if (!pskb_may_pull(skb, ETH_HLEN)) {
538                                 tunnel->dev->stats.rx_length_errors++;
539                                 tunnel->dev->stats.rx_errors++;
540                                 goto drop;
541                         }
542 
543                         ipv6h = ipv6_hdr(skb);
544                         skb->protocol = eth_type_trans(skb, tunnel->dev);
545                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
546                 }
547 
548                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
549 
550                 skb_reset_network_header(skb);
551 
552                 err = IP6_ECN_decapsulate(ipv6h, skb);
553                 if (unlikely(err)) {
554                         if (log_ecn_error)
555                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
556                                                      &ipv6h->saddr,
557                                                      ipv6_get_dsfield(ipv6h));
558                         if (err > 1) {
559                                 ++tunnel->dev->stats.rx_frame_errors;
560                                 ++tunnel->dev->stats.rx_errors;
561                                 goto drop;
562                         }
563                 }
564 
565                 tstats = this_cpu_ptr(tunnel->dev->tstats);
566                 u64_stats_update_begin(&tstats->syncp);
567                 tstats->rx_packets++;
568                 tstats->rx_bytes += skb->len;
569                 u64_stats_update_end(&tstats->syncp);
570 
571                 netif_rx(skb);
572 
573                 return 0;
574         }
575         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
576 
577 drop:
578         kfree_skb(skb);
579         return 0;
580 }
581 
582 struct ipv6_tel_txoption {
583         struct ipv6_txoptions ops;
584         __u8 dst_opt[8];
585 };
586 
587 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
588 {
589         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
590 
591         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
592         opt->dst_opt[3] = 1;
593         opt->dst_opt[4] = encap_limit;
594         opt->dst_opt[5] = IPV6_TLV_PADN;
595         opt->dst_opt[6] = 1;
596 
597         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
598         opt->ops.opt_nflen = 8;
599 }
600 
601 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
602                          struct net_device *dev,
603                          __u8 dsfield,
604                          struct flowi6 *fl6,
605                          int encap_limit,
606                          __u32 *pmtu)
607 {
608         struct ip6_tnl *tunnel = netdev_priv(dev);
609         struct net *net = tunnel->net;
610         struct net_device *tdev;    /* Device to other host */
611         struct ipv6hdr  *ipv6h;     /* Our new IP header */
612         unsigned int max_headroom = 0; /* The extra header space needed */
613         int    gre_hlen;
614         struct ipv6_tel_txoption opt;
615         int    mtu;
616         struct dst_entry *dst = NULL, *ndst = NULL;
617         struct net_device_stats *stats = &tunnel->dev->stats;
618         int err = -1;
619         u8 proto;
620         struct sk_buff *new_skb;
621         __be16 protocol;
622 
623         if (dev->type == ARPHRD_ETHER)
624                 IPCB(skb)->flags = 0;
625 
626         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
627                 gre_hlen = 0;
628                 ipv6h = (struct ipv6hdr *)skb->data;
629                 fl6->daddr = ipv6h->daddr;
630         } else {
631                 gre_hlen = tunnel->hlen;
632                 fl6->daddr = tunnel->parms.raddr;
633         }
634 
635         if (!fl6->flowi6_mark)
636                 dst = dst_cache_get(&tunnel->dst_cache);
637 
638         if (!dst) {
639                 dst = ip6_route_output(net, NULL, fl6);
640 
641                 if (dst->error)
642                         goto tx_err_link_failure;
643                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
644                 if (IS_ERR(dst)) {
645                         err = PTR_ERR(dst);
646                         dst = NULL;
647                         goto tx_err_link_failure;
648                 }
649                 ndst = dst;
650         }
651 
652         tdev = dst->dev;
653 
654         if (tdev == dev) {
655                 stats->collisions++;
656                 net_warn_ratelimited("%s: Local routing loop detected!\n",
657                                      tunnel->parms.name);
658                 goto tx_err_dst_release;
659         }
660 
661         mtu = dst_mtu(dst) - sizeof(*ipv6h);
662         if (encap_limit >= 0) {
663                 max_headroom += 8;
664                 mtu -= 8;
665         }
666         if (mtu < IPV6_MIN_MTU)
667                 mtu = IPV6_MIN_MTU;
668         if (skb_dst(skb))
669                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
670         if (skb->len > mtu) {
671                 *pmtu = mtu;
672                 err = -EMSGSIZE;
673                 goto tx_err_dst_release;
674         }
675 
676         if (tunnel->err_count > 0) {
677                 if (time_before(jiffies,
678                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
679                         tunnel->err_count--;
680 
681                         dst_link_failure(skb);
682                 } else
683                         tunnel->err_count = 0;
684         }
685 
686         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
687 
688         max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
689 
690         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
691             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
692                 new_skb = skb_realloc_headroom(skb, max_headroom);
693                 if (max_headroom > dev->needed_headroom)
694                         dev->needed_headroom = max_headroom;
695                 if (!new_skb)
696                         goto tx_err_dst_release;
697 
698                 if (skb->sk)
699                         skb_set_owner_w(new_skb, skb->sk);
700                 consume_skb(skb);
701                 skb = new_skb;
702         }
703 
704         if (!fl6->flowi6_mark && ndst)
705                 dst_cache_set_ip6(&tunnel->dst_cache, ndst, &fl6->saddr);
706         skb_dst_set(skb, dst);
707 
708         proto = NEXTHDR_GRE;
709         if (encap_limit >= 0) {
710                 init_tel_txopt(&opt, encap_limit);
711                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
712         }
713 
714         if (likely(!skb->encapsulation)) {
715                 skb_reset_inner_headers(skb);
716                 skb->encapsulation = 1;
717         }
718 
719         skb_push(skb, gre_hlen);
720         skb_reset_network_header(skb);
721         skb_set_transport_header(skb, sizeof(*ipv6h));
722 
723         /*
724          *      Push down and install the IP header.
725          */
726         ipv6h = ipv6_hdr(skb);
727         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
728                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
729         ipv6h->hop_limit = tunnel->parms.hop_limit;
730         ipv6h->nexthdr = proto;
731         ipv6h->saddr = fl6->saddr;
732         ipv6h->daddr = fl6->daddr;
733 
734         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
735         protocol = (dev->type == ARPHRD_ETHER) ?
736                     htons(ETH_P_TEB) : skb->protocol;
737         ((__be16 *)(ipv6h + 1))[1] = protocol;
738 
739         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
740                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
741 
742                 if (tunnel->parms.o_flags&GRE_SEQ) {
743                         ++tunnel->o_seqno;
744                         *ptr = htonl(tunnel->o_seqno);
745                         ptr--;
746                 }
747                 if (tunnel->parms.o_flags&GRE_KEY) {
748                         *ptr = tunnel->parms.o_key;
749                         ptr--;
750                 }
751                 if (tunnel->parms.o_flags&GRE_CSUM) {
752                         *ptr = 0;
753                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
754                                 skb->len - sizeof(struct ipv6hdr));
755                 }
756         }
757 
758         skb_set_inner_protocol(skb, protocol);
759 
760         ip6tunnel_xmit(NULL, skb, dev);
761         return 0;
762 tx_err_link_failure:
763         stats->tx_carrier_errors++;
764         dst_link_failure(skb);
765 tx_err_dst_release:
766         dst_release(dst);
767         return err;
768 }
769 
770 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
771 {
772         struct ip6_tnl *t = netdev_priv(dev);
773         const struct iphdr  *iph = ip_hdr(skb);
774         int encap_limit = -1;
775         struct flowi6 fl6;
776         __u8 dsfield;
777         __u32 mtu;
778         int err;
779 
780         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
781 
782         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
783                 encap_limit = t->parms.encap_limit;
784 
785         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
786         fl6.flowi6_proto = IPPROTO_GRE;
787 
788         dsfield = ipv4_get_dsfield(iph);
789 
790         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
791                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
792                                           & IPV6_TCLASS_MASK;
793         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
794                 fl6.flowi6_mark = skb->mark;
795 
796         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
797         if (err != 0) {
798                 /* XXX: send ICMP error even if DF is not set. */
799                 if (err == -EMSGSIZE)
800                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
801                                   htonl(mtu));
802                 return -1;
803         }
804 
805         return 0;
806 }
807 
808 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
809 {
810         struct ip6_tnl *t = netdev_priv(dev);
811         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
812         int encap_limit = -1;
813         __u16 offset;
814         struct flowi6 fl6;
815         __u8 dsfield;
816         __u32 mtu;
817         int err;
818 
819         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
820                 return -1;
821 
822         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
823         if (offset > 0) {
824                 struct ipv6_tlv_tnl_enc_lim *tel;
825                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
826                 if (tel->encap_limit == 0) {
827                         icmpv6_send(skb, ICMPV6_PARAMPROB,
828                                     ICMPV6_HDR_FIELD, offset + 2);
829                         return -1;
830                 }
831                 encap_limit = tel->encap_limit - 1;
832         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
833                 encap_limit = t->parms.encap_limit;
834 
835         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
836         fl6.flowi6_proto = IPPROTO_GRE;
837 
838         dsfield = ipv6_get_dsfield(ipv6h);
839         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
840                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
841         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
842                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
843         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
844                 fl6.flowi6_mark = skb->mark;
845 
846         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
847         if (err != 0) {
848                 if (err == -EMSGSIZE)
849                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
850                 return -1;
851         }
852 
853         return 0;
854 }
855 
856 /**
857  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
858  *   @t: the outgoing tunnel device
859  *   @hdr: IPv6 header from the incoming packet
860  *
861  * Description:
862  *   Avoid trivial tunneling loop by checking that tunnel exit-point
863  *   doesn't match source of incoming packet.
864  *
865  * Return:
866  *   1 if conflict,
867  *   0 else
868  **/
869 
870 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
871         const struct ipv6hdr *hdr)
872 {
873         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
874 }
875 
876 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
877 {
878         struct ip6_tnl *t = netdev_priv(dev);
879         int encap_limit = -1;
880         struct flowi6 fl6;
881         __u32 mtu;
882         int err;
883 
884         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
885                 encap_limit = t->parms.encap_limit;
886 
887         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
888         fl6.flowi6_proto = skb->protocol;
889 
890         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
891 
892         return err;
893 }
894 
895 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
896         struct net_device *dev)
897 {
898         struct ip6_tnl *t = netdev_priv(dev);
899         struct net_device_stats *stats = &t->dev->stats;
900         int ret;
901 
902         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
903                 goto tx_err;
904 
905         switch (skb->protocol) {
906         case htons(ETH_P_IP):
907                 ret = ip6gre_xmit_ipv4(skb, dev);
908                 break;
909         case htons(ETH_P_IPV6):
910                 ret = ip6gre_xmit_ipv6(skb, dev);
911                 break;
912         default:
913                 ret = ip6gre_xmit_other(skb, dev);
914                 break;
915         }
916 
917         if (ret < 0)
918                 goto tx_err;
919 
920         return NETDEV_TX_OK;
921 
922 tx_err:
923         stats->tx_errors++;
924         stats->tx_dropped++;
925         kfree_skb(skb);
926         return NETDEV_TX_OK;
927 }
928 
929 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
930 {
931         struct net_device *dev = t->dev;
932         struct __ip6_tnl_parm *p = &t->parms;
933         struct flowi6 *fl6 = &t->fl.u.ip6;
934         int addend = sizeof(struct ipv6hdr) + 4;
935 
936         if (dev->type != ARPHRD_ETHER) {
937                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
938                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
939         }
940 
941         /* Set up flowi template */
942         fl6->saddr = p->laddr;
943         fl6->daddr = p->raddr;
944         fl6->flowi6_oif = p->link;
945         fl6->flowlabel = 0;
946 
947         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
948                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
949         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
950                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
951 
952         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
953         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
954 
955         if (p->flags&IP6_TNL_F_CAP_XMIT &&
956                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
957                 dev->flags |= IFF_POINTOPOINT;
958         else
959                 dev->flags &= ~IFF_POINTOPOINT;
960 
961         /* Precalculate GRE options length */
962         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
963                 if (t->parms.o_flags&GRE_CSUM)
964                         addend += 4;
965                 if (t->parms.o_flags&GRE_KEY)
966                         addend += 4;
967                 if (t->parms.o_flags&GRE_SEQ)
968                         addend += 4;
969         }
970         t->hlen = addend;
971 
972         if (p->flags & IP6_TNL_F_CAP_XMIT) {
973                 int strict = (ipv6_addr_type(&p->raddr) &
974                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
975 
976                 struct rt6_info *rt = rt6_lookup(t->net,
977                                                  &p->raddr, &p->laddr,
978                                                  p->link, strict);
979 
980                 if (!rt)
981                         return;
982 
983                 if (rt->dst.dev) {
984                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
985 
986                         if (set_mtu) {
987                                 dev->mtu = rt->dst.dev->mtu - addend;
988                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
989                                         dev->mtu -= 8;
990 
991                                 if (dev->mtu < IPV6_MIN_MTU)
992                                         dev->mtu = IPV6_MIN_MTU;
993                         }
994                 }
995                 ip6_rt_put(rt);
996         }
997 }
998 
999 static int ip6gre_tnl_change(struct ip6_tnl *t,
1000         const struct __ip6_tnl_parm *p, int set_mtu)
1001 {
1002         t->parms.laddr = p->laddr;
1003         t->parms.raddr = p->raddr;
1004         t->parms.flags = p->flags;
1005         t->parms.hop_limit = p->hop_limit;
1006         t->parms.encap_limit = p->encap_limit;
1007         t->parms.flowinfo = p->flowinfo;
1008         t->parms.link = p->link;
1009         t->parms.proto = p->proto;
1010         t->parms.i_key = p->i_key;
1011         t->parms.o_key = p->o_key;
1012         t->parms.i_flags = p->i_flags;
1013         t->parms.o_flags = p->o_flags;
1014         dst_cache_reset(&t->dst_cache);
1015         ip6gre_tnl_link_config(t, set_mtu);
1016         return 0;
1017 }
1018 
1019 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1020         const struct ip6_tnl_parm2 *u)
1021 {
1022         p->laddr = u->laddr;
1023         p->raddr = u->raddr;
1024         p->flags = u->flags;
1025         p->hop_limit = u->hop_limit;
1026         p->encap_limit = u->encap_limit;
1027         p->flowinfo = u->flowinfo;
1028         p->link = u->link;
1029         p->i_key = u->i_key;
1030         p->o_key = u->o_key;
1031         p->i_flags = u->i_flags;
1032         p->o_flags = u->o_flags;
1033         memcpy(p->name, u->name, sizeof(u->name));
1034 }
1035 
1036 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1037         const struct __ip6_tnl_parm *p)
1038 {
1039         u->proto = IPPROTO_GRE;
1040         u->laddr = p->laddr;
1041         u->raddr = p->raddr;
1042         u->flags = p->flags;
1043         u->hop_limit = p->hop_limit;
1044         u->encap_limit = p->encap_limit;
1045         u->flowinfo = p->flowinfo;
1046         u->link = p->link;
1047         u->i_key = p->i_key;
1048         u->o_key = p->o_key;
1049         u->i_flags = p->i_flags;
1050         u->o_flags = p->o_flags;
1051         memcpy(u->name, p->name, sizeof(u->name));
1052 }
1053 
1054 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1055         struct ifreq *ifr, int cmd)
1056 {
1057         int err = 0;
1058         struct ip6_tnl_parm2 p;
1059         struct __ip6_tnl_parm p1;
1060         struct ip6_tnl *t = netdev_priv(dev);
1061         struct net *net = t->net;
1062         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1063 
1064         switch (cmd) {
1065         case SIOCGETTUNNEL:
1066                 if (dev == ign->fb_tunnel_dev) {
1067                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1068                                 err = -EFAULT;
1069                                 break;
1070                         }
1071                         ip6gre_tnl_parm_from_user(&p1, &p);
1072                         t = ip6gre_tunnel_locate(net, &p1, 0);
1073                         if (!t)
1074                                 t = netdev_priv(dev);
1075                 }
1076                 memset(&p, 0, sizeof(p));
1077                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1078                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1079                         err = -EFAULT;
1080                 break;
1081 
1082         case SIOCADDTUNNEL:
1083         case SIOCCHGTUNNEL:
1084                 err = -EPERM;
1085                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1086                         goto done;
1087 
1088                 err = -EFAULT;
1089                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1090                         goto done;
1091 
1092                 err = -EINVAL;
1093                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1094                         goto done;
1095 
1096                 if (!(p.i_flags&GRE_KEY))
1097                         p.i_key = 0;
1098                 if (!(p.o_flags&GRE_KEY))
1099                         p.o_key = 0;
1100 
1101                 ip6gre_tnl_parm_from_user(&p1, &p);
1102                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1103 
1104                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1105                         if (t) {
1106                                 if (t->dev != dev) {
1107                                         err = -EEXIST;
1108                                         break;
1109                                 }
1110                         } else {
1111                                 t = netdev_priv(dev);
1112 
1113                                 ip6gre_tunnel_unlink(ign, t);
1114                                 synchronize_net();
1115                                 ip6gre_tnl_change(t, &p1, 1);
1116                                 ip6gre_tunnel_link(ign, t);
1117                                 netdev_state_change(dev);
1118                         }
1119                 }
1120 
1121                 if (t) {
1122                         err = 0;
1123 
1124                         memset(&p, 0, sizeof(p));
1125                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1126                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1127                                 err = -EFAULT;
1128                 } else
1129                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1130                 break;
1131 
1132         case SIOCDELTUNNEL:
1133                 err = -EPERM;
1134                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1135                         goto done;
1136 
1137                 if (dev == ign->fb_tunnel_dev) {
1138                         err = -EFAULT;
1139                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1140                                 goto done;
1141                         err = -ENOENT;
1142                         ip6gre_tnl_parm_from_user(&p1, &p);
1143                         t = ip6gre_tunnel_locate(net, &p1, 0);
1144                         if (!t)
1145                                 goto done;
1146                         err = -EPERM;
1147                         if (t == netdev_priv(ign->fb_tunnel_dev))
1148                                 goto done;
1149                         dev = t->dev;
1150                 }
1151                 unregister_netdevice(dev);
1152                 err = 0;
1153                 break;
1154 
1155         default:
1156                 err = -EINVAL;
1157         }
1158 
1159 done:
1160         return err;
1161 }
1162 
1163 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1164 {
1165         if (new_mtu < 68 ||
1166             new_mtu > 0xFFF8 - dev->hard_header_len)
1167                 return -EINVAL;
1168         dev->mtu = new_mtu;
1169         return 0;
1170 }
1171 
1172 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1173                         unsigned short type,
1174                         const void *daddr, const void *saddr, unsigned int len)
1175 {
1176         struct ip6_tnl *t = netdev_priv(dev);
1177         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1178         __be16 *p = (__be16 *)(ipv6h+1);
1179 
1180         ip6_flow_hdr(ipv6h, 0,
1181                      ip6_make_flowlabel(dev_net(dev), skb,
1182                                         t->fl.u.ip6.flowlabel, true,
1183                                         &t->fl.u.ip6));
1184         ipv6h->hop_limit = t->parms.hop_limit;
1185         ipv6h->nexthdr = NEXTHDR_GRE;
1186         ipv6h->saddr = t->parms.laddr;
1187         ipv6h->daddr = t->parms.raddr;
1188 
1189         p[0]            = t->parms.o_flags;
1190         p[1]            = htons(type);
1191 
1192         /*
1193          *      Set the source hardware address.
1194          */
1195 
1196         if (saddr)
1197                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1198         if (daddr)
1199                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1200         if (!ipv6_addr_any(&ipv6h->daddr))
1201                 return t->hlen;
1202 
1203         return -t->hlen;
1204 }
1205 
1206 static const struct header_ops ip6gre_header_ops = {
1207         .create = ip6gre_header,
1208 };
1209 
1210 static const struct net_device_ops ip6gre_netdev_ops = {
1211         .ndo_init               = ip6gre_tunnel_init,
1212         .ndo_uninit             = ip6gre_tunnel_uninit,
1213         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1214         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1215         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1216         .ndo_get_stats64        = ip_tunnel_get_stats64,
1217         .ndo_get_iflink         = ip6_tnl_get_iflink,
1218 };
1219 
1220 static void ip6gre_dev_free(struct net_device *dev)
1221 {
1222         struct ip6_tnl *t = netdev_priv(dev);
1223 
1224         dst_cache_destroy(&t->dst_cache);
1225         free_percpu(dev->tstats);
1226         free_netdev(dev);
1227 }
1228 
1229 static void ip6gre_tunnel_setup(struct net_device *dev)
1230 {
1231         struct ip6_tnl *t;
1232 
1233         dev->netdev_ops = &ip6gre_netdev_ops;
1234         dev->destructor = ip6gre_dev_free;
1235 
1236         dev->type = ARPHRD_IP6GRE;
1237         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1238         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1239         t = netdev_priv(dev);
1240         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1241                 dev->mtu -= 8;
1242         dev->flags |= IFF_NOARP;
1243         dev->addr_len = sizeof(struct in6_addr);
1244         netif_keep_dst(dev);
1245 }
1246 
1247 static int ip6gre_tunnel_init_common(struct net_device *dev)
1248 {
1249         struct ip6_tnl *tunnel;
1250         int ret;
1251 
1252         tunnel = netdev_priv(dev);
1253 
1254         tunnel->dev = dev;
1255         tunnel->net = dev_net(dev);
1256         strcpy(tunnel->parms.name, dev->name);
1257 
1258         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1259         if (!dev->tstats)
1260                 return -ENOMEM;
1261 
1262         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1263         if (ret) {
1264                 free_percpu(dev->tstats);
1265                 dev->tstats = NULL;
1266                 return ret;
1267         }
1268 
1269         return 0;
1270 }
1271 
1272 static int ip6gre_tunnel_init(struct net_device *dev)
1273 {
1274         struct ip6_tnl *tunnel;
1275         int ret;
1276 
1277         ret = ip6gre_tunnel_init_common(dev);
1278         if (ret)
1279                 return ret;
1280 
1281         tunnel = netdev_priv(dev);
1282 
1283         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1284         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1285 
1286         if (ipv6_addr_any(&tunnel->parms.raddr))
1287                 dev->header_ops = &ip6gre_header_ops;
1288 
1289         return 0;
1290 }
1291 
1292 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1293 {
1294         struct ip6_tnl *tunnel = netdev_priv(dev);
1295 
1296         tunnel->dev = dev;
1297         tunnel->net = dev_net(dev);
1298         strcpy(tunnel->parms.name, dev->name);
1299 
1300         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1301 
1302         dev_hold(dev);
1303 }
1304 
1305 
1306 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1307         .handler     = ip6gre_rcv,
1308         .err_handler = ip6gre_err,
1309         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1310 };
1311 
1312 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1313 {
1314         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1315         struct net_device *dev, *aux;
1316         int prio;
1317 
1318         for_each_netdev_safe(net, dev, aux)
1319                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1320                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1321                         unregister_netdevice_queue(dev, head);
1322 
1323         for (prio = 0; prio < 4; prio++) {
1324                 int h;
1325                 for (h = 0; h < HASH_SIZE; h++) {
1326                         struct ip6_tnl *t;
1327 
1328                         t = rtnl_dereference(ign->tunnels[prio][h]);
1329 
1330                         while (t) {
1331                                 /* If dev is in the same netns, it has already
1332                                  * been added to the list by the previous loop.
1333                                  */
1334                                 if (!net_eq(dev_net(t->dev), net))
1335                                         unregister_netdevice_queue(t->dev,
1336                                                                    head);
1337                                 t = rtnl_dereference(t->next);
1338                         }
1339                 }
1340         }
1341 }
1342 
1343 static int __net_init ip6gre_init_net(struct net *net)
1344 {
1345         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1346         int err;
1347 
1348         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1349                                           NET_NAME_UNKNOWN,
1350                                           ip6gre_tunnel_setup);
1351         if (!ign->fb_tunnel_dev) {
1352                 err = -ENOMEM;
1353                 goto err_alloc_dev;
1354         }
1355         dev_net_set(ign->fb_tunnel_dev, net);
1356         /* FB netdevice is special: we have one, and only one per netns.
1357          * Allowing to move it to another netns is clearly unsafe.
1358          */
1359         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1360 
1361 
1362         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1363         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1364 
1365         err = register_netdev(ign->fb_tunnel_dev);
1366         if (err)
1367                 goto err_reg_dev;
1368 
1369         rcu_assign_pointer(ign->tunnels_wc[0],
1370                            netdev_priv(ign->fb_tunnel_dev));
1371         return 0;
1372 
1373 err_reg_dev:
1374         ip6gre_dev_free(ign->fb_tunnel_dev);
1375 err_alloc_dev:
1376         return err;
1377 }
1378 
1379 static void __net_exit ip6gre_exit_net(struct net *net)
1380 {
1381         LIST_HEAD(list);
1382 
1383         rtnl_lock();
1384         ip6gre_destroy_tunnels(net, &list);
1385         unregister_netdevice_many(&list);
1386         rtnl_unlock();
1387 }
1388 
1389 static struct pernet_operations ip6gre_net_ops = {
1390         .init = ip6gre_init_net,
1391         .exit = ip6gre_exit_net,
1392         .id   = &ip6gre_net_id,
1393         .size = sizeof(struct ip6gre_net),
1394 };
1395 
1396 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1397 {
1398         __be16 flags;
1399 
1400         if (!data)
1401                 return 0;
1402 
1403         flags = 0;
1404         if (data[IFLA_GRE_IFLAGS])
1405                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1406         if (data[IFLA_GRE_OFLAGS])
1407                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1408         if (flags & (GRE_VERSION|GRE_ROUTING))
1409                 return -EINVAL;
1410 
1411         return 0;
1412 }
1413 
1414 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1415 {
1416         struct in6_addr daddr;
1417 
1418         if (tb[IFLA_ADDRESS]) {
1419                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1420                         return -EINVAL;
1421                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1422                         return -EADDRNOTAVAIL;
1423         }
1424 
1425         if (!data)
1426                 goto out;
1427 
1428         if (data[IFLA_GRE_REMOTE]) {
1429                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1430                 if (ipv6_addr_any(&daddr))
1431                         return -EINVAL;
1432         }
1433 
1434 out:
1435         return ip6gre_tunnel_validate(tb, data);
1436 }
1437 
1438 
1439 static void ip6gre_netlink_parms(struct nlattr *data[],
1440                                 struct __ip6_tnl_parm *parms)
1441 {
1442         memset(parms, 0, sizeof(*parms));
1443 
1444         if (!data)
1445                 return;
1446 
1447         if (data[IFLA_GRE_LINK])
1448                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1449 
1450         if (data[IFLA_GRE_IFLAGS])
1451                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1452 
1453         if (data[IFLA_GRE_OFLAGS])
1454                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1455 
1456         if (data[IFLA_GRE_IKEY])
1457                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1458 
1459         if (data[IFLA_GRE_OKEY])
1460                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1461 
1462         if (data[IFLA_GRE_LOCAL])
1463                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1464 
1465         if (data[IFLA_GRE_REMOTE])
1466                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1467 
1468         if (data[IFLA_GRE_TTL])
1469                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1470 
1471         if (data[IFLA_GRE_ENCAP_LIMIT])
1472                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1473 
1474         if (data[IFLA_GRE_FLOWINFO])
1475                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1476 
1477         if (data[IFLA_GRE_FLAGS])
1478                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1479 }
1480 
1481 static int ip6gre_tap_init(struct net_device *dev)
1482 {
1483         struct ip6_tnl *tunnel;
1484         int ret;
1485 
1486         ret = ip6gre_tunnel_init_common(dev);
1487         if (ret)
1488                 return ret;
1489 
1490         tunnel = netdev_priv(dev);
1491 
1492         ip6gre_tnl_link_config(tunnel, 1);
1493 
1494         return 0;
1495 }
1496 
1497 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1498         .ndo_init = ip6gre_tap_init,
1499         .ndo_uninit = ip6gre_tunnel_uninit,
1500         .ndo_start_xmit = ip6gre_tunnel_xmit,
1501         .ndo_set_mac_address = eth_mac_addr,
1502         .ndo_validate_addr = eth_validate_addr,
1503         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1504         .ndo_get_stats64 = ip_tunnel_get_stats64,
1505         .ndo_get_iflink = ip6_tnl_get_iflink,
1506 };
1507 
1508 static void ip6gre_tap_setup(struct net_device *dev)
1509 {
1510 
1511         ether_setup(dev);
1512 
1513         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1514         dev->destructor = ip6gre_dev_free;
1515 
1516         dev->features |= NETIF_F_NETNS_LOCAL;
1517         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1518 }
1519 
1520 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1521         struct nlattr *tb[], struct nlattr *data[])
1522 {
1523         struct ip6_tnl *nt;
1524         struct net *net = dev_net(dev);
1525         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1526         int err;
1527 
1528         nt = netdev_priv(dev);
1529         ip6gre_netlink_parms(data, &nt->parms);
1530 
1531         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1532                 return -EEXIST;
1533 
1534         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1535                 eth_hw_addr_random(dev);
1536 
1537         nt->dev = dev;
1538         nt->net = dev_net(dev);
1539         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1540 
1541         /* Can use a lockless transmit, unless we generate output sequences */
1542         if (!(nt->parms.o_flags & GRE_SEQ))
1543                 dev->features |= NETIF_F_LLTX;
1544 
1545         err = register_netdevice(dev);
1546         if (err)
1547                 goto out;
1548 
1549         dev_hold(dev);
1550         ip6gre_tunnel_link(ign, nt);
1551 
1552 out:
1553         return err;
1554 }
1555 
1556 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1557                             struct nlattr *data[])
1558 {
1559         struct ip6_tnl *t, *nt = netdev_priv(dev);
1560         struct net *net = nt->net;
1561         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1562         struct __ip6_tnl_parm p;
1563 
1564         if (dev == ign->fb_tunnel_dev)
1565                 return -EINVAL;
1566 
1567         ip6gre_netlink_parms(data, &p);
1568 
1569         t = ip6gre_tunnel_locate(net, &p, 0);
1570 
1571         if (t) {
1572                 if (t->dev != dev)
1573                         return -EEXIST;
1574         } else {
1575                 t = nt;
1576         }
1577 
1578         ip6gre_tunnel_unlink(ign, t);
1579         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1580         ip6gre_tunnel_link(ign, t);
1581         return 0;
1582 }
1583 
1584 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1585 {
1586         struct net *net = dev_net(dev);
1587         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1588 
1589         if (dev != ign->fb_tunnel_dev)
1590                 unregister_netdevice_queue(dev, head);
1591 }
1592 
1593 static size_t ip6gre_get_size(const struct net_device *dev)
1594 {
1595         return
1596                 /* IFLA_GRE_LINK */
1597                 nla_total_size(4) +
1598                 /* IFLA_GRE_IFLAGS */
1599                 nla_total_size(2) +
1600                 /* IFLA_GRE_OFLAGS */
1601                 nla_total_size(2) +
1602                 /* IFLA_GRE_IKEY */
1603                 nla_total_size(4) +
1604                 /* IFLA_GRE_OKEY */
1605                 nla_total_size(4) +
1606                 /* IFLA_GRE_LOCAL */
1607                 nla_total_size(sizeof(struct in6_addr)) +
1608                 /* IFLA_GRE_REMOTE */
1609                 nla_total_size(sizeof(struct in6_addr)) +
1610                 /* IFLA_GRE_TTL */
1611                 nla_total_size(1) +
1612                 /* IFLA_GRE_TOS */
1613                 nla_total_size(1) +
1614                 /* IFLA_GRE_ENCAP_LIMIT */
1615                 nla_total_size(1) +
1616                 /* IFLA_GRE_FLOWINFO */
1617                 nla_total_size(4) +
1618                 /* IFLA_GRE_FLAGS */
1619                 nla_total_size(4) +
1620                 0;
1621 }
1622 
1623 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1624 {
1625         struct ip6_tnl *t = netdev_priv(dev);
1626         struct __ip6_tnl_parm *p = &t->parms;
1627 
1628         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1629             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1630             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1631             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1632             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1633             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1634             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1635             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1636             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1637             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1638             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1639             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1640                 goto nla_put_failure;
1641         return 0;
1642 
1643 nla_put_failure:
1644         return -EMSGSIZE;
1645 }
1646 
1647 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1648         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1649         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1650         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1651         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1652         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1653         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1654         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1655         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1656         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1657         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1658         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1659 };
1660 
1661 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1662         .kind           = "ip6gre",
1663         .maxtype        = IFLA_GRE_MAX,
1664         .policy         = ip6gre_policy,
1665         .priv_size      = sizeof(struct ip6_tnl),
1666         .setup          = ip6gre_tunnel_setup,
1667         .validate       = ip6gre_tunnel_validate,
1668         .newlink        = ip6gre_newlink,
1669         .changelink     = ip6gre_changelink,
1670         .dellink        = ip6gre_dellink,
1671         .get_size       = ip6gre_get_size,
1672         .fill_info      = ip6gre_fill_info,
1673         .get_link_net   = ip6_tnl_get_link_net,
1674 };
1675 
1676 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1677         .kind           = "ip6gretap",
1678         .maxtype        = IFLA_GRE_MAX,
1679         .policy         = ip6gre_policy,
1680         .priv_size      = sizeof(struct ip6_tnl),
1681         .setup          = ip6gre_tap_setup,
1682         .validate       = ip6gre_tap_validate,
1683         .newlink        = ip6gre_newlink,
1684         .changelink     = ip6gre_changelink,
1685         .get_size       = ip6gre_get_size,
1686         .fill_info      = ip6gre_fill_info,
1687         .get_link_net   = ip6_tnl_get_link_net,
1688 };
1689 
1690 /*
1691  *      And now the modules code and kernel interface.
1692  */
1693 
1694 static int __init ip6gre_init(void)
1695 {
1696         int err;
1697 
1698         pr_info("GRE over IPv6 tunneling driver\n");
1699 
1700         err = register_pernet_device(&ip6gre_net_ops);
1701         if (err < 0)
1702                 return err;
1703 
1704         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1705         if (err < 0) {
1706                 pr_info("%s: can't add protocol\n", __func__);
1707                 goto add_proto_failed;
1708         }
1709 
1710         err = rtnl_link_register(&ip6gre_link_ops);
1711         if (err < 0)
1712                 goto rtnl_link_failed;
1713 
1714         err = rtnl_link_register(&ip6gre_tap_ops);
1715         if (err < 0)
1716                 goto tap_ops_failed;
1717 
1718 out:
1719         return err;
1720 
1721 tap_ops_failed:
1722         rtnl_link_unregister(&ip6gre_link_ops);
1723 rtnl_link_failed:
1724         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1725 add_proto_failed:
1726         unregister_pernet_device(&ip6gre_net_ops);
1727         goto out;
1728 }
1729 
1730 static void __exit ip6gre_fini(void)
1731 {
1732         rtnl_link_unregister(&ip6gre_tap_ops);
1733         rtnl_link_unregister(&ip6gre_link_ops);
1734         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1735         unregister_pernet_device(&ip6gre_net_ops);
1736 }
1737 
1738 module_init(ip6gre_init);
1739 module_exit(ip6gre_fini);
1740 MODULE_LICENSE("GPL");
1741 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1742 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1743 MODULE_ALIAS_RTNL_LINK("ip6gre");
1744 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1745 MODULE_ALIAS_NETDEV("ip6gre0");
1746 

~ [ 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