Merge tag 'staging-4.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
[linux-2.6-block.git] / net / ipv6 / ip6_gre.c
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 #include <net/gre.h>
58 #include <net/erspan.h>
59 #include <net/dst_metadata.h>
60
61
62 static bool log_ecn_error = true;
63 module_param(log_ecn_error, bool, 0644);
64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
65
66 #define IP6_GRE_HASH_SIZE_SHIFT  5
67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
68
69 static unsigned int ip6gre_net_id __read_mostly;
70 struct ip6gre_net {
71         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
72
73         struct ip6_tnl __rcu *collect_md_tun;
74         struct ip6_tnl __rcu *collect_md_tun_erspan;
75         struct net_device *fb_tunnel_dev;
76 };
77
78 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
80 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
81 static int ip6gre_tunnel_init(struct net_device *dev);
82 static void ip6gre_tunnel_setup(struct net_device *dev);
83 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
84 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
85 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
86
87 /* Tunnel hash table */
88
89 /*
90    4 hash tables:
91
92    3: (remote,local)
93    2: (remote,*)
94    1: (*,local)
95    0: (*,*)
96
97    We require exact key match i.e. if a key is present in packet
98    it will match only tunnel with the same key; if it is not present,
99    it will match only keyless tunnel.
100
101    All keysless packets, if not matched configured keyless tunnels
102    will match fallback tunnel.
103  */
104
105 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
106 static u32 HASH_ADDR(const struct in6_addr *addr)
107 {
108         u32 hash = ipv6_addr_hash(addr);
109
110         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
111 }
112
113 #define tunnels_r_l     tunnels[3]
114 #define tunnels_r       tunnels[2]
115 #define tunnels_l       tunnels[1]
116 #define tunnels_wc      tunnels[0]
117
118 /* Given src, dst and key, find appropriate for input tunnel. */
119
120 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
121                 const struct in6_addr *remote, const struct in6_addr *local,
122                 __be32 key, __be16 gre_proto)
123 {
124         struct net *net = dev_net(dev);
125         int link = dev->ifindex;
126         unsigned int h0 = HASH_ADDR(remote);
127         unsigned int h1 = HASH_KEY(key);
128         struct ip6_tnl *t, *cand = NULL;
129         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
130         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
131                         gre_proto == htons(ETH_P_ERSPAN) ||
132                         gre_proto == htons(ETH_P_ERSPAN2)) ?
133                        ARPHRD_ETHER : ARPHRD_IP6GRE;
134         int score, cand_score = 4;
135
136         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
137                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
139                     key != t->parms.i_key ||
140                     !(t->dev->flags & IFF_UP))
141                         continue;
142
143                 if (t->dev->type != ARPHRD_IP6GRE &&
144                     t->dev->type != dev_type)
145                         continue;
146
147                 score = 0;
148                 if (t->parms.link != link)
149                         score |= 1;
150                 if (t->dev->type != dev_type)
151                         score |= 2;
152                 if (score == 0)
153                         return t;
154
155                 if (score < cand_score) {
156                         cand = t;
157                         cand_score = score;
158                 }
159         }
160
161         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
162                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
163                     key != t->parms.i_key ||
164                     !(t->dev->flags & IFF_UP))
165                         continue;
166
167                 if (t->dev->type != ARPHRD_IP6GRE &&
168                     t->dev->type != dev_type)
169                         continue;
170
171                 score = 0;
172                 if (t->parms.link != link)
173                         score |= 1;
174                 if (t->dev->type != dev_type)
175                         score |= 2;
176                 if (score == 0)
177                         return t;
178
179                 if (score < cand_score) {
180                         cand = t;
181                         cand_score = score;
182                 }
183         }
184
185         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
186                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
187                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
188                                  !ipv6_addr_is_multicast(local))) ||
189                     key != t->parms.i_key ||
190                     !(t->dev->flags & IFF_UP))
191                         continue;
192
193                 if (t->dev->type != ARPHRD_IP6GRE &&
194                     t->dev->type != dev_type)
195                         continue;
196
197                 score = 0;
198                 if (t->parms.link != link)
199                         score |= 1;
200                 if (t->dev->type != dev_type)
201                         score |= 2;
202                 if (score == 0)
203                         return t;
204
205                 if (score < cand_score) {
206                         cand = t;
207                         cand_score = score;
208                 }
209         }
210
211         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
212                 if (t->parms.i_key != key ||
213                     !(t->dev->flags & IFF_UP))
214                         continue;
215
216                 if (t->dev->type != ARPHRD_IP6GRE &&
217                     t->dev->type != dev_type)
218                         continue;
219
220                 score = 0;
221                 if (t->parms.link != link)
222                         score |= 1;
223                 if (t->dev->type != dev_type)
224                         score |= 2;
225                 if (score == 0)
226                         return t;
227
228                 if (score < cand_score) {
229                         cand = t;
230                         cand_score = score;
231                 }
232         }
233
234         if (cand)
235                 return cand;
236
237         if (gre_proto == htons(ETH_P_ERSPAN) ||
238             gre_proto == htons(ETH_P_ERSPAN2))
239                 t = rcu_dereference(ign->collect_md_tun_erspan);
240         else
241                 t = rcu_dereference(ign->collect_md_tun);
242
243         if (t && t->dev->flags & IFF_UP)
244                 return t;
245
246         dev = ign->fb_tunnel_dev;
247         if (dev && dev->flags & IFF_UP)
248                 return netdev_priv(dev);
249
250         return NULL;
251 }
252
253 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
254                 const struct __ip6_tnl_parm *p)
255 {
256         const struct in6_addr *remote = &p->raddr;
257         const struct in6_addr *local = &p->laddr;
258         unsigned int h = HASH_KEY(p->i_key);
259         int prio = 0;
260
261         if (!ipv6_addr_any(local))
262                 prio |= 1;
263         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
264                 prio |= 2;
265                 h ^= HASH_ADDR(remote);
266         }
267
268         return &ign->tunnels[prio][h];
269 }
270
271 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
272 {
273         if (t->parms.collect_md)
274                 rcu_assign_pointer(ign->collect_md_tun, t);
275 }
276
277 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
278 {
279         if (t->parms.collect_md)
280                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
281 }
282
283 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
284 {
285         if (t->parms.collect_md)
286                 rcu_assign_pointer(ign->collect_md_tun, NULL);
287 }
288
289 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
290                                        struct ip6_tnl *t)
291 {
292         if (t->parms.collect_md)
293                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
294 }
295
296 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
297                 const struct ip6_tnl *t)
298 {
299         return __ip6gre_bucket(ign, &t->parms);
300 }
301
302 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
303 {
304         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
305
306         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
307         rcu_assign_pointer(*tp, t);
308 }
309
310 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
311 {
312         struct ip6_tnl __rcu **tp;
313         struct ip6_tnl *iter;
314
315         for (tp = ip6gre_bucket(ign, t);
316              (iter = rtnl_dereference(*tp)) != NULL;
317              tp = &iter->next) {
318                 if (t == iter) {
319                         rcu_assign_pointer(*tp, t->next);
320                         break;
321                 }
322         }
323 }
324
325 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
326                                            const struct __ip6_tnl_parm *parms,
327                                            int type)
328 {
329         const struct in6_addr *remote = &parms->raddr;
330         const struct in6_addr *local = &parms->laddr;
331         __be32 key = parms->i_key;
332         int link = parms->link;
333         struct ip6_tnl *t;
334         struct ip6_tnl __rcu **tp;
335         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
336
337         for (tp = __ip6gre_bucket(ign, parms);
338              (t = rtnl_dereference(*tp)) != NULL;
339              tp = &t->next)
340                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
341                     ipv6_addr_equal(remote, &t->parms.raddr) &&
342                     key == t->parms.i_key &&
343                     link == t->parms.link &&
344                     type == t->dev->type)
345                         break;
346
347         return t;
348 }
349
350 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
351                 const struct __ip6_tnl_parm *parms, int create)
352 {
353         struct ip6_tnl *t, *nt;
354         struct net_device *dev;
355         char name[IFNAMSIZ];
356         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
357
358         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
359         if (t && create)
360                 return NULL;
361         if (t || !create)
362                 return t;
363
364         if (parms->name[0]) {
365                 if (!dev_valid_name(parms->name))
366                         return NULL;
367                 strlcpy(name, parms->name, IFNAMSIZ);
368         } else {
369                 strcpy(name, "ip6gre%d");
370         }
371         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
372                            ip6gre_tunnel_setup);
373         if (!dev)
374                 return NULL;
375
376         dev_net_set(dev, net);
377
378         nt = netdev_priv(dev);
379         nt->parms = *parms;
380         dev->rtnl_link_ops = &ip6gre_link_ops;
381
382         nt->dev = dev;
383         nt->net = dev_net(dev);
384
385         if (register_netdevice(dev) < 0)
386                 goto failed_free;
387
388         ip6gre_tnl_link_config(nt, 1);
389
390         /* Can use a lockless transmit, unless we generate output sequences */
391         if (!(nt->parms.o_flags & TUNNEL_SEQ))
392                 dev->features |= NETIF_F_LLTX;
393
394         dev_hold(dev);
395         ip6gre_tunnel_link(ign, nt);
396         return nt;
397
398 failed_free:
399         free_netdev(dev);
400         return NULL;
401 }
402
403 static void ip6erspan_tunnel_uninit(struct net_device *dev)
404 {
405         struct ip6_tnl *t = netdev_priv(dev);
406         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
407
408         ip6erspan_tunnel_unlink_md(ign, t);
409         ip6gre_tunnel_unlink(ign, t);
410         dst_cache_reset(&t->dst_cache);
411         dev_put(dev);
412 }
413
414 static void ip6gre_tunnel_uninit(struct net_device *dev)
415 {
416         struct ip6_tnl *t = netdev_priv(dev);
417         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
418
419         ip6gre_tunnel_unlink_md(ign, t);
420         ip6gre_tunnel_unlink(ign, t);
421         dst_cache_reset(&t->dst_cache);
422         dev_put(dev);
423 }
424
425
426 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
427                        u8 type, u8 code, int offset, __be32 info)
428 {
429         struct net *net = dev_net(skb->dev);
430         const struct gre_base_hdr *greh;
431         const struct ipv6hdr *ipv6h;
432         int grehlen = sizeof(*greh);
433         struct ip6_tnl *t;
434         int key_off = 0;
435         __be16 flags;
436         __be32 key;
437
438         if (!pskb_may_pull(skb, offset + grehlen))
439                 return;
440         greh = (const struct gre_base_hdr *)(skb->data + offset);
441         flags = greh->flags;
442         if (flags & (GRE_VERSION | GRE_ROUTING))
443                 return;
444         if (flags & GRE_CSUM)
445                 grehlen += 4;
446         if (flags & GRE_KEY) {
447                 key_off = grehlen + offset;
448                 grehlen += 4;
449         }
450
451         if (!pskb_may_pull(skb, offset + grehlen))
452                 return;
453         ipv6h = (const struct ipv6hdr *)skb->data;
454         greh = (const struct gre_base_hdr *)(skb->data + offset);
455         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
456
457         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
458                                  key, greh->protocol);
459         if (!t)
460                 return;
461
462         switch (type) {
463                 struct ipv6_tlv_tnl_enc_lim *tel;
464                 __u32 teli;
465         case ICMPV6_DEST_UNREACH:
466                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
467                                     t->parms.name);
468                 if (code != ICMPV6_PORT_UNREACH)
469                         break;
470                 return;
471         case ICMPV6_TIME_EXCEED:
472                 if (code == ICMPV6_EXC_HOPLIMIT) {
473                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
474                                             t->parms.name);
475                         break;
476                 }
477                 return;
478         case ICMPV6_PARAMPROB:
479                 teli = 0;
480                 if (code == ICMPV6_HDR_FIELD)
481                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
482
483                 if (teli && teli == be32_to_cpu(info) - 2) {
484                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
485                         if (tel->encap_limit == 0) {
486                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
487                                                     t->parms.name);
488                         }
489                 } else {
490                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
491                                             t->parms.name);
492                 }
493                 return;
494         case ICMPV6_PKT_TOOBIG:
495                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
496                 return;
497         case NDISC_REDIRECT:
498                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
499                              sock_net_uid(net, NULL));
500                 return;
501         }
502
503         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
504                 t->err_count++;
505         else
506                 t->err_count = 1;
507         t->err_time = jiffies;
508 }
509
510 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
511 {
512         const struct ipv6hdr *ipv6h;
513         struct ip6_tnl *tunnel;
514
515         ipv6h = ipv6_hdr(skb);
516         tunnel = ip6gre_tunnel_lookup(skb->dev,
517                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
518                                       tpi->proto);
519         if (tunnel) {
520                 if (tunnel->parms.collect_md) {
521                         struct metadata_dst *tun_dst;
522                         __be64 tun_id;
523                         __be16 flags;
524
525                         flags = tpi->flags;
526                         tun_id = key32_to_tunnel_id(tpi->key);
527
528                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
529                         if (!tun_dst)
530                                 return PACKET_REJECT;
531
532                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
533                 } else {
534                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
535                 }
536
537                 return PACKET_RCVD;
538         }
539
540         return PACKET_REJECT;
541 }
542
543 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
544                          struct tnl_ptk_info *tpi)
545 {
546         struct erspan_base_hdr *ershdr;
547         struct erspan_metadata *pkt_md;
548         const struct ipv6hdr *ipv6h;
549         struct erspan_md2 *md2;
550         struct ip6_tnl *tunnel;
551         u8 ver;
552
553         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
554                 return PACKET_REJECT;
555
556         ipv6h = ipv6_hdr(skb);
557         ershdr = (struct erspan_base_hdr *)skb->data;
558         ver = ershdr->ver;
559         tpi->key = cpu_to_be32(get_session_id(ershdr));
560
561         tunnel = ip6gre_tunnel_lookup(skb->dev,
562                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
563                                       tpi->proto);
564         if (tunnel) {
565                 int len = erspan_hdr_len(ver);
566
567                 if (unlikely(!pskb_may_pull(skb, len)))
568                         return PACKET_REJECT;
569
570                 ershdr = (struct erspan_base_hdr *)skb->data;
571                 pkt_md = (struct erspan_metadata *)(ershdr + 1);
572
573                 if (__iptunnel_pull_header(skb, len,
574                                            htons(ETH_P_TEB),
575                                            false, false) < 0)
576                         return PACKET_REJECT;
577
578                 if (tunnel->parms.collect_md) {
579                         struct metadata_dst *tun_dst;
580                         struct ip_tunnel_info *info;
581                         struct erspan_metadata *md;
582                         __be64 tun_id;
583                         __be16 flags;
584
585                         tpi->flags |= TUNNEL_KEY;
586                         flags = tpi->flags;
587                         tun_id = key32_to_tunnel_id(tpi->key);
588
589                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
590                                                   sizeof(*md));
591                         if (!tun_dst)
592                                 return PACKET_REJECT;
593
594                         info = &tun_dst->u.tun_info;
595                         md = ip_tunnel_info_opts(info);
596                         md->version = ver;
597                         md2 = &md->u.md2;
598                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
599                                                        ERSPAN_V2_MDSIZE);
600                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
601                         info->options_len = sizeof(*md);
602
603                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
604
605                 } else {
606                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
607                 }
608
609                 return PACKET_RCVD;
610         }
611
612         return PACKET_REJECT;
613 }
614
615 static int gre_rcv(struct sk_buff *skb)
616 {
617         struct tnl_ptk_info tpi;
618         bool csum_err = false;
619         int hdr_len;
620
621         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
622         if (hdr_len < 0)
623                 goto drop;
624
625         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
626                 goto drop;
627
628         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
629                      tpi.proto == htons(ETH_P_ERSPAN2))) {
630                 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
631                         return 0;
632                 goto out;
633         }
634
635         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
636                 return 0;
637
638 out:
639         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
640 drop:
641         kfree_skb(skb);
642         return 0;
643 }
644
645 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
646 {
647         return iptunnel_handle_offloads(skb,
648                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
649 }
650
651 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
652                                      struct net_device *dev,
653                                      struct flowi6 *fl6, __u8 *dsfield,
654                                      int *encap_limit)
655 {
656         const struct iphdr *iph = ip_hdr(skb);
657         struct ip6_tnl *t = netdev_priv(dev);
658
659         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
660                 *encap_limit = t->parms.encap_limit;
661
662         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
663
664         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
665                 *dsfield = ipv4_get_dsfield(iph);
666         else
667                 *dsfield = ip6_tclass(t->parms.flowinfo);
668
669         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
670                 fl6->flowi6_mark = skb->mark;
671         else
672                 fl6->flowi6_mark = t->parms.fwmark;
673
674         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
675 }
676
677 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
678                                     struct net_device *dev,
679                                     struct flowi6 *fl6, __u8 *dsfield,
680                                     int *encap_limit)
681 {
682         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
683         struct ip6_tnl *t = netdev_priv(dev);
684         __u16 offset;
685
686         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
687         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
688
689         if (offset > 0) {
690                 struct ipv6_tlv_tnl_enc_lim *tel;
691
692                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
693                 if (tel->encap_limit == 0) {
694                         icmpv6_send(skb, ICMPV6_PARAMPROB,
695                                     ICMPV6_HDR_FIELD, offset + 2);
696                         return -1;
697                 }
698                 *encap_limit = tel->encap_limit - 1;
699         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
700                 *encap_limit = t->parms.encap_limit;
701         }
702
703         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
704
705         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
706                 *dsfield = ipv6_get_dsfield(ipv6h);
707         else
708                 *dsfield = ip6_tclass(t->parms.flowinfo);
709
710         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
711                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
712
713         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
714                 fl6->flowi6_mark = skb->mark;
715         else
716                 fl6->flowi6_mark = t->parms.fwmark;
717
718         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
719
720         return 0;
721 }
722
723 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
724                                struct net_device *dev, __u8 dsfield,
725                                struct flowi6 *fl6, int encap_limit,
726                                __u32 *pmtu, __be16 proto)
727 {
728         struct ip6_tnl *tunnel = netdev_priv(dev);
729         __be16 protocol;
730
731         if (dev->type == ARPHRD_ETHER)
732                 IPCB(skb)->flags = 0;
733
734         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
735                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
736         else
737                 fl6->daddr = tunnel->parms.raddr;
738
739         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
740                 return -ENOMEM;
741
742         /* Push GRE header. */
743         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
744
745         if (tunnel->parms.collect_md) {
746                 struct ip_tunnel_info *tun_info;
747                 const struct ip_tunnel_key *key;
748                 __be16 flags;
749
750                 tun_info = skb_tunnel_info(skb);
751                 if (unlikely(!tun_info ||
752                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
753                              ip_tunnel_info_af(tun_info) != AF_INET6))
754                         return -EINVAL;
755
756                 key = &tun_info->key;
757                 memset(fl6, 0, sizeof(*fl6));
758                 fl6->flowi6_proto = IPPROTO_GRE;
759                 fl6->daddr = key->u.ipv6.dst;
760                 fl6->flowlabel = key->label;
761                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
762
763                 dsfield = key->tos;
764                 flags = key->tun_flags &
765                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
766                 tunnel->tun_hlen = gre_calc_hlen(flags);
767
768                 gre_build_header(skb, tunnel->tun_hlen,
769                                  flags, protocol,
770                                  tunnel_id_to_key32(tun_info->key.tun_id),
771                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
772                                                       : 0);
773
774         } else {
775                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
776                         tunnel->o_seqno++;
777
778                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
779                                  protocol, tunnel->parms.o_key,
780                                  htonl(tunnel->o_seqno));
781         }
782
783         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
784                             NEXTHDR_GRE);
785 }
786
787 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
788 {
789         struct ip6_tnl *t = netdev_priv(dev);
790         int encap_limit = -1;
791         struct flowi6 fl6;
792         __u8 dsfield = 0;
793         __u32 mtu;
794         int err;
795
796         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
797
798         if (!t->parms.collect_md)
799                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
800                                          &dsfield, &encap_limit);
801
802         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
803         if (err)
804                 return -1;
805
806         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
807                           skb->protocol);
808         if (err != 0) {
809                 /* XXX: send ICMP error even if DF is not set. */
810                 if (err == -EMSGSIZE)
811                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
812                                   htonl(mtu));
813                 return -1;
814         }
815
816         return 0;
817 }
818
819 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
820 {
821         struct ip6_tnl *t = netdev_priv(dev);
822         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
823         int encap_limit = -1;
824         struct flowi6 fl6;
825         __u8 dsfield = 0;
826         __u32 mtu;
827         int err;
828
829         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
830                 return -1;
831
832         if (!t->parms.collect_md &&
833             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
834                 return -1;
835
836         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
837                 return -1;
838
839         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
840                           &mtu, skb->protocol);
841         if (err != 0) {
842                 if (err == -EMSGSIZE)
843                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
844                 return -1;
845         }
846
847         return 0;
848 }
849
850 /**
851  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
852  *   @t: the outgoing tunnel device
853  *   @hdr: IPv6 header from the incoming packet
854  *
855  * Description:
856  *   Avoid trivial tunneling loop by checking that tunnel exit-point
857  *   doesn't match source of incoming packet.
858  *
859  * Return:
860  *   1 if conflict,
861  *   0 else
862  **/
863
864 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
865         const struct ipv6hdr *hdr)
866 {
867         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
868 }
869
870 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
871 {
872         struct ip6_tnl *t = netdev_priv(dev);
873         int encap_limit = -1;
874         struct flowi6 fl6;
875         __u32 mtu;
876         int err;
877
878         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
879                 encap_limit = t->parms.encap_limit;
880
881         if (!t->parms.collect_md)
882                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
883
884         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
885         if (err)
886                 return err;
887
888         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
889
890         return err;
891 }
892
893 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
894         struct net_device *dev)
895 {
896         struct ip6_tnl *t = netdev_priv(dev);
897         struct net_device_stats *stats = &t->dev->stats;
898         int ret;
899
900         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
901                 goto tx_err;
902
903         switch (skb->protocol) {
904         case htons(ETH_P_IP):
905                 ret = ip6gre_xmit_ipv4(skb, dev);
906                 break;
907         case htons(ETH_P_IPV6):
908                 ret = ip6gre_xmit_ipv6(skb, dev);
909                 break;
910         default:
911                 ret = ip6gre_xmit_other(skb, dev);
912                 break;
913         }
914
915         if (ret < 0)
916                 goto tx_err;
917
918         return NETDEV_TX_OK;
919
920 tx_err:
921         stats->tx_errors++;
922         stats->tx_dropped++;
923         kfree_skb(skb);
924         return NETDEV_TX_OK;
925 }
926
927 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
928                                          struct net_device *dev)
929 {
930         struct ip6_tnl *t = netdev_priv(dev);
931         struct dst_entry *dst = skb_dst(skb);
932         struct net_device_stats *stats;
933         bool truncate = false;
934         int encap_limit = -1;
935         __u8 dsfield = false;
936         struct flowi6 fl6;
937         int err = -EINVAL;
938         __u32 mtu;
939         int nhoff;
940         int thoff;
941
942         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
943                 goto tx_err;
944
945         if (gre_handle_offloads(skb, false))
946                 goto tx_err;
947
948         if (skb->len > dev->mtu + dev->hard_header_len) {
949                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
950                 truncate = true;
951         }
952
953         nhoff = skb_network_header(skb) - skb_mac_header(skb);
954         if (skb->protocol == htons(ETH_P_IP) &&
955             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
956                 truncate = true;
957
958         thoff = skb_transport_header(skb) - skb_mac_header(skb);
959         if (skb->protocol == htons(ETH_P_IPV6) &&
960             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
961                 truncate = true;
962
963         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
964                 goto tx_err;
965
966         t->parms.o_flags &= ~TUNNEL_KEY;
967         IPCB(skb)->flags = 0;
968
969         /* For collect_md mode, derive fl6 from the tunnel key,
970          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
971          */
972         if (t->parms.collect_md) {
973                 struct ip_tunnel_info *tun_info;
974                 const struct ip_tunnel_key *key;
975                 struct erspan_metadata *md;
976                 __be32 tun_id;
977
978                 tun_info = skb_tunnel_info(skb);
979                 if (unlikely(!tun_info ||
980                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
981                              ip_tunnel_info_af(tun_info) != AF_INET6))
982                         return -EINVAL;
983
984                 key = &tun_info->key;
985                 memset(&fl6, 0, sizeof(fl6));
986                 fl6.flowi6_proto = IPPROTO_GRE;
987                 fl6.daddr = key->u.ipv6.dst;
988                 fl6.flowlabel = key->label;
989                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
990
991                 dsfield = key->tos;
992                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
993                         goto tx_err;
994                 md = ip_tunnel_info_opts(tun_info);
995                 if (!md)
996                         goto tx_err;
997
998                 tun_id = tunnel_id_to_key32(key->tun_id);
999                 if (md->version == 1) {
1000                         erspan_build_header(skb,
1001                                             ntohl(tun_id),
1002                                             ntohl(md->u.index), truncate,
1003                                             false);
1004                 } else if (md->version == 2) {
1005                         erspan_build_header_v2(skb,
1006                                                ntohl(tun_id),
1007                                                md->u.md2.dir,
1008                                                get_hwid(&md->u.md2),
1009                                                truncate, false);
1010                 } else {
1011                         goto tx_err;
1012                 }
1013         } else {
1014                 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1015
1016                 switch (skb->protocol) {
1017                 case htons(ETH_P_IP):
1018                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1019                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1020                                                  &dsfield, &encap_limit);
1021                         break;
1022                 case htons(ETH_P_IPV6):
1023                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
1024                                 goto tx_err;
1025                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1026                                                      &dsfield, &encap_limit))
1027                                 goto tx_err;
1028                         break;
1029                 default:
1030                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1031                         break;
1032                 }
1033
1034                 if (t->parms.erspan_ver == 1)
1035                         erspan_build_header(skb, ntohl(t->parms.o_key),
1036                                             t->parms.index,
1037                                             truncate, false);
1038                 else if (t->parms.erspan_ver == 2)
1039                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1040                                                t->parms.dir,
1041                                                t->parms.hwid,
1042                                                truncate, false);
1043                 else
1044                         goto tx_err;
1045
1046                 fl6.daddr = t->parms.raddr;
1047         }
1048
1049         /* Push GRE header. */
1050         gre_build_header(skb, 8, TUNNEL_SEQ,
1051                          htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1052
1053         /* TooBig packet may have updated dst->dev's mtu */
1054         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1055                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1056
1057         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1058                            NEXTHDR_GRE);
1059         if (err != 0) {
1060                 /* XXX: send ICMP error even if DF is not set. */
1061                 if (err == -EMSGSIZE) {
1062                         if (skb->protocol == htons(ETH_P_IP))
1063                                 icmp_send(skb, ICMP_DEST_UNREACH,
1064                                           ICMP_FRAG_NEEDED, htonl(mtu));
1065                         else
1066                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1067                 }
1068
1069                 goto tx_err;
1070         }
1071         return NETDEV_TX_OK;
1072
1073 tx_err:
1074         stats = &t->dev->stats;
1075         stats->tx_errors++;
1076         stats->tx_dropped++;
1077         kfree_skb(skb);
1078         return NETDEV_TX_OK;
1079 }
1080
1081 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1082 {
1083         struct net_device *dev = t->dev;
1084         struct __ip6_tnl_parm *p = &t->parms;
1085         struct flowi6 *fl6 = &t->fl.u.ip6;
1086
1087         if (dev->type != ARPHRD_ETHER) {
1088                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1089                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1090         }
1091
1092         /* Set up flowi template */
1093         fl6->saddr = p->laddr;
1094         fl6->daddr = p->raddr;
1095         fl6->flowi6_oif = p->link;
1096         fl6->flowlabel = 0;
1097         fl6->flowi6_proto = IPPROTO_GRE;
1098
1099         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1100                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1101         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1102                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1103
1104         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1105         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1106
1107         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1108                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1109                 dev->flags |= IFF_POINTOPOINT;
1110         else
1111                 dev->flags &= ~IFF_POINTOPOINT;
1112 }
1113
1114 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1115                                          int t_hlen)
1116 {
1117         const struct __ip6_tnl_parm *p = &t->parms;
1118         struct net_device *dev = t->dev;
1119
1120         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1121                 int strict = (ipv6_addr_type(&p->raddr) &
1122                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1123
1124                 struct rt6_info *rt = rt6_lookup(t->net,
1125                                                  &p->raddr, &p->laddr,
1126                                                  p->link, NULL, strict);
1127
1128                 if (!rt)
1129                         return;
1130
1131                 if (rt->dst.dev) {
1132                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1133                                                t_hlen;
1134
1135                         if (set_mtu) {
1136                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1137                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1138                                         dev->mtu -= 8;
1139                                 if (dev->type == ARPHRD_ETHER)
1140                                         dev->mtu -= ETH_HLEN;
1141
1142                                 if (dev->mtu < IPV6_MIN_MTU)
1143                                         dev->mtu = IPV6_MIN_MTU;
1144                         }
1145                 }
1146                 ip6_rt_put(rt);
1147         }
1148 }
1149
1150 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1151 {
1152         int t_hlen;
1153
1154         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1155         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1156
1157         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1158         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1159         return t_hlen;
1160 }
1161
1162 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1163 {
1164         ip6gre_tnl_link_config_common(t);
1165         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1166 }
1167
1168 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1169                                      const struct __ip6_tnl_parm *p)
1170 {
1171         t->parms.laddr = p->laddr;
1172         t->parms.raddr = p->raddr;
1173         t->parms.flags = p->flags;
1174         t->parms.hop_limit = p->hop_limit;
1175         t->parms.encap_limit = p->encap_limit;
1176         t->parms.flowinfo = p->flowinfo;
1177         t->parms.link = p->link;
1178         t->parms.proto = p->proto;
1179         t->parms.i_key = p->i_key;
1180         t->parms.o_key = p->o_key;
1181         t->parms.i_flags = p->i_flags;
1182         t->parms.o_flags = p->o_flags;
1183         t->parms.fwmark = p->fwmark;
1184         dst_cache_reset(&t->dst_cache);
1185 }
1186
1187 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1188                              int set_mtu)
1189 {
1190         ip6gre_tnl_copy_tnl_parm(t, p);
1191         ip6gre_tnl_link_config(t, set_mtu);
1192         return 0;
1193 }
1194
1195 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1196         const struct ip6_tnl_parm2 *u)
1197 {
1198         p->laddr = u->laddr;
1199         p->raddr = u->raddr;
1200         p->flags = u->flags;
1201         p->hop_limit = u->hop_limit;
1202         p->encap_limit = u->encap_limit;
1203         p->flowinfo = u->flowinfo;
1204         p->link = u->link;
1205         p->i_key = u->i_key;
1206         p->o_key = u->o_key;
1207         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1208         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1209         memcpy(p->name, u->name, sizeof(u->name));
1210 }
1211
1212 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1213         const struct __ip6_tnl_parm *p)
1214 {
1215         u->proto = IPPROTO_GRE;
1216         u->laddr = p->laddr;
1217         u->raddr = p->raddr;
1218         u->flags = p->flags;
1219         u->hop_limit = p->hop_limit;
1220         u->encap_limit = p->encap_limit;
1221         u->flowinfo = p->flowinfo;
1222         u->link = p->link;
1223         u->i_key = p->i_key;
1224         u->o_key = p->o_key;
1225         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1226         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1227         memcpy(u->name, p->name, sizeof(u->name));
1228 }
1229
1230 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1231         struct ifreq *ifr, int cmd)
1232 {
1233         int err = 0;
1234         struct ip6_tnl_parm2 p;
1235         struct __ip6_tnl_parm p1;
1236         struct ip6_tnl *t = netdev_priv(dev);
1237         struct net *net = t->net;
1238         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1239
1240         memset(&p1, 0, sizeof(p1));
1241
1242         switch (cmd) {
1243         case SIOCGETTUNNEL:
1244                 if (dev == ign->fb_tunnel_dev) {
1245                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1246                                 err = -EFAULT;
1247                                 break;
1248                         }
1249                         ip6gre_tnl_parm_from_user(&p1, &p);
1250                         t = ip6gre_tunnel_locate(net, &p1, 0);
1251                         if (!t)
1252                                 t = netdev_priv(dev);
1253                 }
1254                 memset(&p, 0, sizeof(p));
1255                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1256                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1257                         err = -EFAULT;
1258                 break;
1259
1260         case SIOCADDTUNNEL:
1261         case SIOCCHGTUNNEL:
1262                 err = -EPERM;
1263                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1264                         goto done;
1265
1266                 err = -EFAULT;
1267                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1268                         goto done;
1269
1270                 err = -EINVAL;
1271                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1272                         goto done;
1273
1274                 if (!(p.i_flags&GRE_KEY))
1275                         p.i_key = 0;
1276                 if (!(p.o_flags&GRE_KEY))
1277                         p.o_key = 0;
1278
1279                 ip6gre_tnl_parm_from_user(&p1, &p);
1280                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1281
1282                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1283                         if (t) {
1284                                 if (t->dev != dev) {
1285                                         err = -EEXIST;
1286                                         break;
1287                                 }
1288                         } else {
1289                                 t = netdev_priv(dev);
1290
1291                                 ip6gre_tunnel_unlink(ign, t);
1292                                 synchronize_net();
1293                                 ip6gre_tnl_change(t, &p1, 1);
1294                                 ip6gre_tunnel_link(ign, t);
1295                                 netdev_state_change(dev);
1296                         }
1297                 }
1298
1299                 if (t) {
1300                         err = 0;
1301
1302                         memset(&p, 0, sizeof(p));
1303                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1304                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1305                                 err = -EFAULT;
1306                 } else
1307                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1308                 break;
1309
1310         case SIOCDELTUNNEL:
1311                 err = -EPERM;
1312                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1313                         goto done;
1314
1315                 if (dev == ign->fb_tunnel_dev) {
1316                         err = -EFAULT;
1317                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1318                                 goto done;
1319                         err = -ENOENT;
1320                         ip6gre_tnl_parm_from_user(&p1, &p);
1321                         t = ip6gre_tunnel_locate(net, &p1, 0);
1322                         if (!t)
1323                                 goto done;
1324                         err = -EPERM;
1325                         if (t == netdev_priv(ign->fb_tunnel_dev))
1326                                 goto done;
1327                         dev = t->dev;
1328                 }
1329                 unregister_netdevice(dev);
1330                 err = 0;
1331                 break;
1332
1333         default:
1334                 err = -EINVAL;
1335         }
1336
1337 done:
1338         return err;
1339 }
1340
1341 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1342                          unsigned short type, const void *daddr,
1343                          const void *saddr, unsigned int len)
1344 {
1345         struct ip6_tnl *t = netdev_priv(dev);
1346         struct ipv6hdr *ipv6h;
1347         __be16 *p;
1348
1349         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1350         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1351                                                   t->fl.u.ip6.flowlabel,
1352                                                   true, &t->fl.u.ip6));
1353         ipv6h->hop_limit = t->parms.hop_limit;
1354         ipv6h->nexthdr = NEXTHDR_GRE;
1355         ipv6h->saddr = t->parms.laddr;
1356         ipv6h->daddr = t->parms.raddr;
1357
1358         p = (__be16 *)(ipv6h + 1);
1359         p[0] = t->parms.o_flags;
1360         p[1] = htons(type);
1361
1362         /*
1363          *      Set the source hardware address.
1364          */
1365
1366         if (saddr)
1367                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1368         if (daddr)
1369                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1370         if (!ipv6_addr_any(&ipv6h->daddr))
1371                 return t->hlen;
1372
1373         return -t->hlen;
1374 }
1375
1376 static const struct header_ops ip6gre_header_ops = {
1377         .create = ip6gre_header,
1378 };
1379
1380 static const struct net_device_ops ip6gre_netdev_ops = {
1381         .ndo_init               = ip6gre_tunnel_init,
1382         .ndo_uninit             = ip6gre_tunnel_uninit,
1383         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1384         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1385         .ndo_change_mtu         = ip6_tnl_change_mtu,
1386         .ndo_get_stats64        = ip_tunnel_get_stats64,
1387         .ndo_get_iflink         = ip6_tnl_get_iflink,
1388 };
1389
1390 static void ip6gre_dev_free(struct net_device *dev)
1391 {
1392         struct ip6_tnl *t = netdev_priv(dev);
1393
1394         gro_cells_destroy(&t->gro_cells);
1395         dst_cache_destroy(&t->dst_cache);
1396         free_percpu(dev->tstats);
1397 }
1398
1399 static void ip6gre_tunnel_setup(struct net_device *dev)
1400 {
1401         dev->netdev_ops = &ip6gre_netdev_ops;
1402         dev->needs_free_netdev = true;
1403         dev->priv_destructor = ip6gre_dev_free;
1404
1405         dev->type = ARPHRD_IP6GRE;
1406
1407         dev->flags |= IFF_NOARP;
1408         dev->addr_len = sizeof(struct in6_addr);
1409         netif_keep_dst(dev);
1410         /* This perm addr will be used as interface identifier by IPv6 */
1411         dev->addr_assign_type = NET_ADDR_RANDOM;
1412         eth_random_addr(dev->perm_addr);
1413 }
1414
1415 #define GRE6_FEATURES (NETIF_F_SG |             \
1416                        NETIF_F_FRAGLIST |       \
1417                        NETIF_F_HIGHDMA |        \
1418                        NETIF_F_HW_CSUM)
1419
1420 static void ip6gre_tnl_init_features(struct net_device *dev)
1421 {
1422         struct ip6_tnl *nt = netdev_priv(dev);
1423
1424         dev->features           |= GRE6_FEATURES;
1425         dev->hw_features        |= GRE6_FEATURES;
1426
1427         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1428                 /* TCP offload with GRE SEQ is not supported, nor
1429                  * can we support 2 levels of outer headers requiring
1430                  * an update.
1431                  */
1432                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1433                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1434                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1435                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1436                 }
1437
1438                 /* Can use a lockless transmit, unless we generate
1439                  * output sequences
1440                  */
1441                 dev->features |= NETIF_F_LLTX;
1442         }
1443 }
1444
1445 static int ip6gre_tunnel_init_common(struct net_device *dev)
1446 {
1447         struct ip6_tnl *tunnel;
1448         int ret;
1449         int t_hlen;
1450
1451         tunnel = netdev_priv(dev);
1452
1453         tunnel->dev = dev;
1454         tunnel->net = dev_net(dev);
1455         strcpy(tunnel->parms.name, dev->name);
1456
1457         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1458         if (!dev->tstats)
1459                 return -ENOMEM;
1460
1461         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1462         if (ret)
1463                 goto cleanup_alloc_pcpu_stats;
1464
1465         ret = gro_cells_init(&tunnel->gro_cells, dev);
1466         if (ret)
1467                 goto cleanup_dst_cache_init;
1468
1469         t_hlen = ip6gre_calc_hlen(tunnel);
1470         dev->mtu = ETH_DATA_LEN - t_hlen;
1471         if (dev->type == ARPHRD_ETHER)
1472                 dev->mtu -= ETH_HLEN;
1473         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1474                 dev->mtu -= 8;
1475
1476         if (tunnel->parms.collect_md) {
1477                 dev->features |= NETIF_F_NETNS_LOCAL;
1478                 netif_keep_dst(dev);
1479         }
1480         ip6gre_tnl_init_features(dev);
1481
1482         return 0;
1483
1484 cleanup_dst_cache_init:
1485         dst_cache_destroy(&tunnel->dst_cache);
1486 cleanup_alloc_pcpu_stats:
1487         free_percpu(dev->tstats);
1488         dev->tstats = NULL;
1489         return ret;
1490 }
1491
1492 static int ip6gre_tunnel_init(struct net_device *dev)
1493 {
1494         struct ip6_tnl *tunnel;
1495         int ret;
1496
1497         ret = ip6gre_tunnel_init_common(dev);
1498         if (ret)
1499                 return ret;
1500
1501         tunnel = netdev_priv(dev);
1502
1503         if (tunnel->parms.collect_md)
1504                 return 0;
1505
1506         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1507         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1508
1509         if (ipv6_addr_any(&tunnel->parms.raddr))
1510                 dev->header_ops = &ip6gre_header_ops;
1511
1512         return 0;
1513 }
1514
1515 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1516 {
1517         struct ip6_tnl *tunnel = netdev_priv(dev);
1518
1519         tunnel->dev = dev;
1520         tunnel->net = dev_net(dev);
1521         strcpy(tunnel->parms.name, dev->name);
1522
1523         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1524
1525         dev_hold(dev);
1526 }
1527
1528 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1529         .handler     = gre_rcv,
1530         .err_handler = ip6gre_err,
1531         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1532 };
1533
1534 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1535 {
1536         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1537         struct net_device *dev, *aux;
1538         int prio;
1539
1540         for_each_netdev_safe(net, dev, aux)
1541                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1542                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1543                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1544                         unregister_netdevice_queue(dev, head);
1545
1546         for (prio = 0; prio < 4; prio++) {
1547                 int h;
1548                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1549                         struct ip6_tnl *t;
1550
1551                         t = rtnl_dereference(ign->tunnels[prio][h]);
1552
1553                         while (t) {
1554                                 /* If dev is in the same netns, it has already
1555                                  * been added to the list by the previous loop.
1556                                  */
1557                                 if (!net_eq(dev_net(t->dev), net))
1558                                         unregister_netdevice_queue(t->dev,
1559                                                                    head);
1560                                 t = rtnl_dereference(t->next);
1561                         }
1562                 }
1563         }
1564 }
1565
1566 static int __net_init ip6gre_init_net(struct net *net)
1567 {
1568         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1569         int err;
1570
1571         if (!net_has_fallback_tunnels(net))
1572                 return 0;
1573         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1574                                           NET_NAME_UNKNOWN,
1575                                           ip6gre_tunnel_setup);
1576         if (!ign->fb_tunnel_dev) {
1577                 err = -ENOMEM;
1578                 goto err_alloc_dev;
1579         }
1580         dev_net_set(ign->fb_tunnel_dev, net);
1581         /* FB netdevice is special: we have one, and only one per netns.
1582          * Allowing to move it to another netns is clearly unsafe.
1583          */
1584         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1585
1586
1587         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1588         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1589
1590         err = register_netdev(ign->fb_tunnel_dev);
1591         if (err)
1592                 goto err_reg_dev;
1593
1594         rcu_assign_pointer(ign->tunnels_wc[0],
1595                            netdev_priv(ign->fb_tunnel_dev));
1596         return 0;
1597
1598 err_reg_dev:
1599         free_netdev(ign->fb_tunnel_dev);
1600 err_alloc_dev:
1601         return err;
1602 }
1603
1604 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1605 {
1606         struct net *net;
1607         LIST_HEAD(list);
1608
1609         rtnl_lock();
1610         list_for_each_entry(net, net_list, exit_list)
1611                 ip6gre_destroy_tunnels(net, &list);
1612         unregister_netdevice_many(&list);
1613         rtnl_unlock();
1614 }
1615
1616 static struct pernet_operations ip6gre_net_ops = {
1617         .init = ip6gre_init_net,
1618         .exit_batch = ip6gre_exit_batch_net,
1619         .id   = &ip6gre_net_id,
1620         .size = sizeof(struct ip6gre_net),
1621 };
1622
1623 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1624                                   struct netlink_ext_ack *extack)
1625 {
1626         __be16 flags;
1627
1628         if (!data)
1629                 return 0;
1630
1631         flags = 0;
1632         if (data[IFLA_GRE_IFLAGS])
1633                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1634         if (data[IFLA_GRE_OFLAGS])
1635                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1636         if (flags & (GRE_VERSION|GRE_ROUTING))
1637                 return -EINVAL;
1638
1639         return 0;
1640 }
1641
1642 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1643                                struct netlink_ext_ack *extack)
1644 {
1645         struct in6_addr daddr;
1646
1647         if (tb[IFLA_ADDRESS]) {
1648                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1649                         return -EINVAL;
1650                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1651                         return -EADDRNOTAVAIL;
1652         }
1653
1654         if (!data)
1655                 goto out;
1656
1657         if (data[IFLA_GRE_REMOTE]) {
1658                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1659                 if (ipv6_addr_any(&daddr))
1660                         return -EINVAL;
1661         }
1662
1663 out:
1664         return ip6gre_tunnel_validate(tb, data, extack);
1665 }
1666
1667 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1668                                   struct netlink_ext_ack *extack)
1669 {
1670         __be16 flags = 0;
1671         int ret, ver = 0;
1672
1673         if (!data)
1674                 return 0;
1675
1676         ret = ip6gre_tap_validate(tb, data, extack);
1677         if (ret)
1678                 return ret;
1679
1680         /* ERSPAN should only have GRE sequence and key flag */
1681         if (data[IFLA_GRE_OFLAGS])
1682                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1683         if (data[IFLA_GRE_IFLAGS])
1684                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1685         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1686             flags != (GRE_SEQ | GRE_KEY))
1687                 return -EINVAL;
1688
1689         /* ERSPAN Session ID only has 10-bit. Since we reuse
1690          * 32-bit key field as ID, check it's range.
1691          */
1692         if (data[IFLA_GRE_IKEY] &&
1693             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1694                 return -EINVAL;
1695
1696         if (data[IFLA_GRE_OKEY] &&
1697             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1698                 return -EINVAL;
1699
1700         if (data[IFLA_GRE_ERSPAN_VER]) {
1701                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1702                 if (ver != 1 && ver != 2)
1703                         return -EINVAL;
1704         }
1705
1706         if (ver == 1) {
1707                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1708                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1709
1710                         if (index & ~INDEX_MASK)
1711                                 return -EINVAL;
1712                 }
1713         } else if (ver == 2) {
1714                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1715                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1716
1717                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1718                                 return -EINVAL;
1719                 }
1720
1721                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1722                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1723
1724                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1725                                 return -EINVAL;
1726                 }
1727         }
1728
1729         return 0;
1730 }
1731
1732 static void ip6gre_netlink_parms(struct nlattr *data[],
1733                                 struct __ip6_tnl_parm *parms)
1734 {
1735         memset(parms, 0, sizeof(*parms));
1736
1737         if (!data)
1738                 return;
1739
1740         if (data[IFLA_GRE_LINK])
1741                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1742
1743         if (data[IFLA_GRE_IFLAGS])
1744                 parms->i_flags = gre_flags_to_tnl_flags(
1745                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1746
1747         if (data[IFLA_GRE_OFLAGS])
1748                 parms->o_flags = gre_flags_to_tnl_flags(
1749                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1750
1751         if (data[IFLA_GRE_IKEY])
1752                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1753
1754         if (data[IFLA_GRE_OKEY])
1755                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1756
1757         if (data[IFLA_GRE_LOCAL])
1758                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1759
1760         if (data[IFLA_GRE_REMOTE])
1761                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1762
1763         if (data[IFLA_GRE_TTL])
1764                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1765
1766         if (data[IFLA_GRE_ENCAP_LIMIT])
1767                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1768
1769         if (data[IFLA_GRE_FLOWINFO])
1770                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1771
1772         if (data[IFLA_GRE_FLAGS])
1773                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1774
1775         if (data[IFLA_GRE_FWMARK])
1776                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1777
1778         if (data[IFLA_GRE_COLLECT_METADATA])
1779                 parms->collect_md = true;
1780
1781         if (data[IFLA_GRE_ERSPAN_VER])
1782                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1783
1784         if (parms->erspan_ver == 1) {
1785                 if (data[IFLA_GRE_ERSPAN_INDEX])
1786                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1787         } else if (parms->erspan_ver == 2) {
1788                 if (data[IFLA_GRE_ERSPAN_DIR])
1789                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1790                 if (data[IFLA_GRE_ERSPAN_HWID])
1791                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1792         }
1793 }
1794
1795 static int ip6gre_tap_init(struct net_device *dev)
1796 {
1797         int ret;
1798
1799         ret = ip6gre_tunnel_init_common(dev);
1800         if (ret)
1801                 return ret;
1802
1803         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1804
1805         return 0;
1806 }
1807
1808 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1809         .ndo_init = ip6gre_tap_init,
1810         .ndo_uninit = ip6gre_tunnel_uninit,
1811         .ndo_start_xmit = ip6gre_tunnel_xmit,
1812         .ndo_set_mac_address = eth_mac_addr,
1813         .ndo_validate_addr = eth_validate_addr,
1814         .ndo_change_mtu = ip6_tnl_change_mtu,
1815         .ndo_get_stats64 = ip_tunnel_get_stats64,
1816         .ndo_get_iflink = ip6_tnl_get_iflink,
1817 };
1818
1819 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1820 {
1821         int t_hlen;
1822
1823         tunnel->tun_hlen = 8;
1824         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1825                        erspan_hdr_len(tunnel->parms.erspan_ver);
1826
1827         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1828         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1829         return t_hlen;
1830 }
1831
1832 static int ip6erspan_tap_init(struct net_device *dev)
1833 {
1834         struct ip6_tnl *tunnel;
1835         int t_hlen;
1836         int ret;
1837
1838         tunnel = netdev_priv(dev);
1839
1840         tunnel->dev = dev;
1841         tunnel->net = dev_net(dev);
1842         strcpy(tunnel->parms.name, dev->name);
1843
1844         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1845         if (!dev->tstats)
1846                 return -ENOMEM;
1847
1848         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1849         if (ret)
1850                 goto cleanup_alloc_pcpu_stats;
1851
1852         ret = gro_cells_init(&tunnel->gro_cells, dev);
1853         if (ret)
1854                 goto cleanup_dst_cache_init;
1855
1856         t_hlen = ip6erspan_calc_hlen(tunnel);
1857         dev->mtu = ETH_DATA_LEN - t_hlen;
1858         if (dev->type == ARPHRD_ETHER)
1859                 dev->mtu -= ETH_HLEN;
1860         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1861                 dev->mtu -= 8;
1862
1863         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1864         ip6erspan_tnl_link_config(tunnel, 1);
1865
1866         return 0;
1867
1868 cleanup_dst_cache_init:
1869         dst_cache_destroy(&tunnel->dst_cache);
1870 cleanup_alloc_pcpu_stats:
1871         free_percpu(dev->tstats);
1872         dev->tstats = NULL;
1873         return ret;
1874 }
1875
1876 static const struct net_device_ops ip6erspan_netdev_ops = {
1877         .ndo_init =             ip6erspan_tap_init,
1878         .ndo_uninit =           ip6erspan_tunnel_uninit,
1879         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1880         .ndo_set_mac_address =  eth_mac_addr,
1881         .ndo_validate_addr =    eth_validate_addr,
1882         .ndo_change_mtu =       ip6_tnl_change_mtu,
1883         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1884         .ndo_get_iflink =       ip6_tnl_get_iflink,
1885 };
1886
1887 static void ip6gre_tap_setup(struct net_device *dev)
1888 {
1889
1890         ether_setup(dev);
1891
1892         dev->max_mtu = 0;
1893         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1894         dev->needs_free_netdev = true;
1895         dev->priv_destructor = ip6gre_dev_free;
1896
1897         dev->features |= NETIF_F_NETNS_LOCAL;
1898         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1899         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1900         netif_keep_dst(dev);
1901 }
1902
1903 bool is_ip6gretap_dev(const struct net_device *dev)
1904 {
1905         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1906 }
1907 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1908
1909 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1910                                        struct ip_tunnel_encap *ipencap)
1911 {
1912         bool ret = false;
1913
1914         memset(ipencap, 0, sizeof(*ipencap));
1915
1916         if (!data)
1917                 return ret;
1918
1919         if (data[IFLA_GRE_ENCAP_TYPE]) {
1920                 ret = true;
1921                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1922         }
1923
1924         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1925                 ret = true;
1926                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1927         }
1928
1929         if (data[IFLA_GRE_ENCAP_SPORT]) {
1930                 ret = true;
1931                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1932         }
1933
1934         if (data[IFLA_GRE_ENCAP_DPORT]) {
1935                 ret = true;
1936                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1937         }
1938
1939         return ret;
1940 }
1941
1942 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1943                                  struct nlattr *tb[], struct nlattr *data[],
1944                                  struct netlink_ext_ack *extack)
1945 {
1946         struct ip6_tnl *nt;
1947         struct ip_tunnel_encap ipencap;
1948         int err;
1949
1950         nt = netdev_priv(dev);
1951
1952         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1953                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1954
1955                 if (err < 0)
1956                         return err;
1957         }
1958
1959         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1960                 eth_hw_addr_random(dev);
1961
1962         nt->dev = dev;
1963         nt->net = dev_net(dev);
1964
1965         err = register_netdevice(dev);
1966         if (err)
1967                 goto out;
1968
1969         if (tb[IFLA_MTU])
1970                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1971
1972         dev_hold(dev);
1973
1974 out:
1975         return err;
1976 }
1977
1978 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1979                           struct nlattr *tb[], struct nlattr *data[],
1980                           struct netlink_ext_ack *extack)
1981 {
1982         struct ip6_tnl *nt = netdev_priv(dev);
1983         struct net *net = dev_net(dev);
1984         struct ip6gre_net *ign;
1985         int err;
1986
1987         ip6gre_netlink_parms(data, &nt->parms);
1988         ign = net_generic(net, ip6gre_net_id);
1989
1990         if (nt->parms.collect_md) {
1991                 if (rtnl_dereference(ign->collect_md_tun))
1992                         return -EEXIST;
1993         } else {
1994                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1995                         return -EEXIST;
1996         }
1997
1998         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1999         if (!err) {
2000                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2001                 ip6gre_tunnel_link_md(ign, nt);
2002                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2003         }
2004         return err;
2005 }
2006
2007 static struct ip6_tnl *
2008 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2009                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2010                          struct netlink_ext_ack *extack)
2011 {
2012         struct ip6_tnl *t, *nt = netdev_priv(dev);
2013         struct net *net = nt->net;
2014         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2015         struct ip_tunnel_encap ipencap;
2016
2017         if (dev == ign->fb_tunnel_dev)
2018                 return ERR_PTR(-EINVAL);
2019
2020         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2021                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2022
2023                 if (err < 0)
2024                         return ERR_PTR(err);
2025         }
2026
2027         ip6gre_netlink_parms(data, p_p);
2028
2029         t = ip6gre_tunnel_locate(net, p_p, 0);
2030
2031         if (t) {
2032                 if (t->dev != dev)
2033                         return ERR_PTR(-EEXIST);
2034         } else {
2035                 t = nt;
2036         }
2037
2038         return t;
2039 }
2040
2041 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2042                              struct nlattr *data[],
2043                              struct netlink_ext_ack *extack)
2044 {
2045         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2046         struct __ip6_tnl_parm p;
2047         struct ip6_tnl *t;
2048
2049         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2050         if (IS_ERR(t))
2051                 return PTR_ERR(t);
2052
2053         ip6gre_tunnel_unlink_md(ign, t);
2054         ip6gre_tunnel_unlink(ign, t);
2055         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2056         ip6gre_tunnel_link_md(ign, t);
2057         ip6gre_tunnel_link(ign, t);
2058         return 0;
2059 }
2060
2061 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2062 {
2063         struct net *net = dev_net(dev);
2064         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2065
2066         if (dev != ign->fb_tunnel_dev)
2067                 unregister_netdevice_queue(dev, head);
2068 }
2069
2070 static size_t ip6gre_get_size(const struct net_device *dev)
2071 {
2072         return
2073                 /* IFLA_GRE_LINK */
2074                 nla_total_size(4) +
2075                 /* IFLA_GRE_IFLAGS */
2076                 nla_total_size(2) +
2077                 /* IFLA_GRE_OFLAGS */
2078                 nla_total_size(2) +
2079                 /* IFLA_GRE_IKEY */
2080                 nla_total_size(4) +
2081                 /* IFLA_GRE_OKEY */
2082                 nla_total_size(4) +
2083                 /* IFLA_GRE_LOCAL */
2084                 nla_total_size(sizeof(struct in6_addr)) +
2085                 /* IFLA_GRE_REMOTE */
2086                 nla_total_size(sizeof(struct in6_addr)) +
2087                 /* IFLA_GRE_TTL */
2088                 nla_total_size(1) +
2089                 /* IFLA_GRE_ENCAP_LIMIT */
2090                 nla_total_size(1) +
2091                 /* IFLA_GRE_FLOWINFO */
2092                 nla_total_size(4) +
2093                 /* IFLA_GRE_FLAGS */
2094                 nla_total_size(4) +
2095                 /* IFLA_GRE_ENCAP_TYPE */
2096                 nla_total_size(2) +
2097                 /* IFLA_GRE_ENCAP_FLAGS */
2098                 nla_total_size(2) +
2099                 /* IFLA_GRE_ENCAP_SPORT */
2100                 nla_total_size(2) +
2101                 /* IFLA_GRE_ENCAP_DPORT */
2102                 nla_total_size(2) +
2103                 /* IFLA_GRE_COLLECT_METADATA */
2104                 nla_total_size(0) +
2105                 /* IFLA_GRE_FWMARK */
2106                 nla_total_size(4) +
2107                 /* IFLA_GRE_ERSPAN_INDEX */
2108                 nla_total_size(4) +
2109                 0;
2110 }
2111
2112 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2113 {
2114         struct ip6_tnl *t = netdev_priv(dev);
2115         struct __ip6_tnl_parm *p = &t->parms;
2116
2117         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2118             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2119                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2120             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2121                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2122             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2123             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2124             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2125             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2126             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2127             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2128             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2129             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2130             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2131             nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2132                 goto nla_put_failure;
2133
2134         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2135                         t->encap.type) ||
2136             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2137                          t->encap.sport) ||
2138             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2139                          t->encap.dport) ||
2140             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2141                         t->encap.flags))
2142                 goto nla_put_failure;
2143
2144         if (p->collect_md) {
2145                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2146                         goto nla_put_failure;
2147         }
2148
2149         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2150                 goto nla_put_failure;
2151
2152         if (p->erspan_ver == 1) {
2153                 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2154                         goto nla_put_failure;
2155         } else if (p->erspan_ver == 2) {
2156                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2157                         goto nla_put_failure;
2158                 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2159                         goto nla_put_failure;
2160         }
2161
2162         return 0;
2163
2164 nla_put_failure:
2165         return -EMSGSIZE;
2166 }
2167
2168 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2169         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2170         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2171         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2172         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2173         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2174         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2175         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2176         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2177         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2178         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2179         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2180         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2181         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2182         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2183         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2184         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2185         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2186         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2187         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2188         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2189         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2190 };
2191
2192 static void ip6erspan_tap_setup(struct net_device *dev)
2193 {
2194         ether_setup(dev);
2195
2196         dev->netdev_ops = &ip6erspan_netdev_ops;
2197         dev->needs_free_netdev = true;
2198         dev->priv_destructor = ip6gre_dev_free;
2199
2200         dev->features |= NETIF_F_NETNS_LOCAL;
2201         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2202         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2203         netif_keep_dst(dev);
2204 }
2205
2206 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2207                              struct nlattr *tb[], struct nlattr *data[],
2208                              struct netlink_ext_ack *extack)
2209 {
2210         struct ip6_tnl *nt = netdev_priv(dev);
2211         struct net *net = dev_net(dev);
2212         struct ip6gre_net *ign;
2213         int err;
2214
2215         ip6gre_netlink_parms(data, &nt->parms);
2216         ign = net_generic(net, ip6gre_net_id);
2217
2218         if (nt->parms.collect_md) {
2219                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2220                         return -EEXIST;
2221         } else {
2222                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2223                         return -EEXIST;
2224         }
2225
2226         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2227         if (!err) {
2228                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2229                 ip6erspan_tunnel_link_md(ign, nt);
2230                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2231         }
2232         return err;
2233 }
2234
2235 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2236 {
2237         ip6gre_tnl_link_config_common(t);
2238         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2239 }
2240
2241 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2242                                 const struct __ip6_tnl_parm *p, int set_mtu)
2243 {
2244         ip6gre_tnl_copy_tnl_parm(t, p);
2245         ip6erspan_tnl_link_config(t, set_mtu);
2246         return 0;
2247 }
2248
2249 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2250                                 struct nlattr *data[],
2251                                 struct netlink_ext_ack *extack)
2252 {
2253         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2254         struct __ip6_tnl_parm p;
2255         struct ip6_tnl *t;
2256
2257         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2258         if (IS_ERR(t))
2259                 return PTR_ERR(t);
2260
2261         ip6gre_tunnel_unlink_md(ign, t);
2262         ip6gre_tunnel_unlink(ign, t);
2263         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2264         ip6erspan_tunnel_link_md(ign, t);
2265         ip6gre_tunnel_link(ign, t);
2266         return 0;
2267 }
2268
2269 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2270         .kind           = "ip6gre",
2271         .maxtype        = IFLA_GRE_MAX,
2272         .policy         = ip6gre_policy,
2273         .priv_size      = sizeof(struct ip6_tnl),
2274         .setup          = ip6gre_tunnel_setup,
2275         .validate       = ip6gre_tunnel_validate,
2276         .newlink        = ip6gre_newlink,
2277         .changelink     = ip6gre_changelink,
2278         .dellink        = ip6gre_dellink,
2279         .get_size       = ip6gre_get_size,
2280         .fill_info      = ip6gre_fill_info,
2281         .get_link_net   = ip6_tnl_get_link_net,
2282 };
2283
2284 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2285         .kind           = "ip6gretap",
2286         .maxtype        = IFLA_GRE_MAX,
2287         .policy         = ip6gre_policy,
2288         .priv_size      = sizeof(struct ip6_tnl),
2289         .setup          = ip6gre_tap_setup,
2290         .validate       = ip6gre_tap_validate,
2291         .newlink        = ip6gre_newlink,
2292         .changelink     = ip6gre_changelink,
2293         .get_size       = ip6gre_get_size,
2294         .fill_info      = ip6gre_fill_info,
2295         .get_link_net   = ip6_tnl_get_link_net,
2296 };
2297
2298 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2299         .kind           = "ip6erspan",
2300         .maxtype        = IFLA_GRE_MAX,
2301         .policy         = ip6gre_policy,
2302         .priv_size      = sizeof(struct ip6_tnl),
2303         .setup          = ip6erspan_tap_setup,
2304         .validate       = ip6erspan_tap_validate,
2305         .newlink        = ip6erspan_newlink,
2306         .changelink     = ip6erspan_changelink,
2307         .get_size       = ip6gre_get_size,
2308         .fill_info      = ip6gre_fill_info,
2309         .get_link_net   = ip6_tnl_get_link_net,
2310 };
2311
2312 /*
2313  *      And now the modules code and kernel interface.
2314  */
2315
2316 static int __init ip6gre_init(void)
2317 {
2318         int err;
2319
2320         pr_info("GRE over IPv6 tunneling driver\n");
2321
2322         err = register_pernet_device(&ip6gre_net_ops);
2323         if (err < 0)
2324                 return err;
2325
2326         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2327         if (err < 0) {
2328                 pr_info("%s: can't add protocol\n", __func__);
2329                 goto add_proto_failed;
2330         }
2331
2332         err = rtnl_link_register(&ip6gre_link_ops);
2333         if (err < 0)
2334                 goto rtnl_link_failed;
2335
2336         err = rtnl_link_register(&ip6gre_tap_ops);
2337         if (err < 0)
2338                 goto tap_ops_failed;
2339
2340         err = rtnl_link_register(&ip6erspan_tap_ops);
2341         if (err < 0)
2342                 goto erspan_link_failed;
2343
2344 out:
2345         return err;
2346
2347 erspan_link_failed:
2348         rtnl_link_unregister(&ip6gre_tap_ops);
2349 tap_ops_failed:
2350         rtnl_link_unregister(&ip6gre_link_ops);
2351 rtnl_link_failed:
2352         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2353 add_proto_failed:
2354         unregister_pernet_device(&ip6gre_net_ops);
2355         goto out;
2356 }
2357
2358 static void __exit ip6gre_fini(void)
2359 {
2360         rtnl_link_unregister(&ip6gre_tap_ops);
2361         rtnl_link_unregister(&ip6gre_link_ops);
2362         rtnl_link_unregister(&ip6erspan_tap_ops);
2363         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2364         unregister_pernet_device(&ip6gre_net_ops);
2365 }
2366
2367 module_init(ip6gre_init);
2368 module_exit(ip6gre_fini);
2369 MODULE_LICENSE("GPL");
2370 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2371 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2372 MODULE_ALIAS_RTNL_LINK("ip6gre");
2373 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2374 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2375 MODULE_ALIAS_NETDEV("ip6gre0");