Merge tag '5.0-rc3-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
[sfrench/cifs-2.6.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 int 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 ipv6hdr *ipv6h;
431         struct tnl_ptk_info tpi;
432         struct ip6_tnl *t;
433
434         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
435                              offset) < 0)
436                 return -EINVAL;
437
438         ipv6h = (const struct ipv6hdr *)skb->data;
439         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
440                                  tpi.key, tpi.proto);
441         if (!t)
442                 return -ENOENT;
443
444         switch (type) {
445                 struct ipv6_tlv_tnl_enc_lim *tel;
446                 __u32 teli;
447         case ICMPV6_DEST_UNREACH:
448                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
449                                     t->parms.name);
450                 if (code != ICMPV6_PORT_UNREACH)
451                         break;
452                 return 0;
453         case ICMPV6_TIME_EXCEED:
454                 if (code == ICMPV6_EXC_HOPLIMIT) {
455                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
456                                             t->parms.name);
457                         break;
458                 }
459                 return 0;
460         case ICMPV6_PARAMPROB:
461                 teli = 0;
462                 if (code == ICMPV6_HDR_FIELD)
463                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
464
465                 if (teli && teli == be32_to_cpu(info) - 2) {
466                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
467                         if (tel->encap_limit == 0) {
468                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
469                                                     t->parms.name);
470                         }
471                 } else {
472                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
473                                             t->parms.name);
474                 }
475                 return 0;
476         case ICMPV6_PKT_TOOBIG:
477                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
478                 return 0;
479         case NDISC_REDIRECT:
480                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
481                              sock_net_uid(net, NULL));
482                 return 0;
483         }
484
485         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
486                 t->err_count++;
487         else
488                 t->err_count = 1;
489         t->err_time = jiffies;
490
491         return 0;
492 }
493
494 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
495 {
496         const struct ipv6hdr *ipv6h;
497         struct ip6_tnl *tunnel;
498
499         ipv6h = ipv6_hdr(skb);
500         tunnel = ip6gre_tunnel_lookup(skb->dev,
501                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
502                                       tpi->proto);
503         if (tunnel) {
504                 if (tunnel->parms.collect_md) {
505                         struct metadata_dst *tun_dst;
506                         __be64 tun_id;
507                         __be16 flags;
508
509                         flags = tpi->flags;
510                         tun_id = key32_to_tunnel_id(tpi->key);
511
512                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
513                         if (!tun_dst)
514                                 return PACKET_REJECT;
515
516                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
517                 } else {
518                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
519                 }
520
521                 return PACKET_RCVD;
522         }
523
524         return PACKET_REJECT;
525 }
526
527 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
528                          struct tnl_ptk_info *tpi)
529 {
530         struct erspan_base_hdr *ershdr;
531         struct erspan_metadata *pkt_md;
532         const struct ipv6hdr *ipv6h;
533         struct erspan_md2 *md2;
534         struct ip6_tnl *tunnel;
535         u8 ver;
536
537         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
538                 return PACKET_REJECT;
539
540         ipv6h = ipv6_hdr(skb);
541         ershdr = (struct erspan_base_hdr *)skb->data;
542         ver = ershdr->ver;
543         tpi->key = cpu_to_be32(get_session_id(ershdr));
544
545         tunnel = ip6gre_tunnel_lookup(skb->dev,
546                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
547                                       tpi->proto);
548         if (tunnel) {
549                 int len = erspan_hdr_len(ver);
550
551                 if (unlikely(!pskb_may_pull(skb, len)))
552                         return PACKET_REJECT;
553
554                 ershdr = (struct erspan_base_hdr *)skb->data;
555                 pkt_md = (struct erspan_metadata *)(ershdr + 1);
556
557                 if (__iptunnel_pull_header(skb, len,
558                                            htons(ETH_P_TEB),
559                                            false, false) < 0)
560                         return PACKET_REJECT;
561
562                 if (tunnel->parms.collect_md) {
563                         struct metadata_dst *tun_dst;
564                         struct ip_tunnel_info *info;
565                         struct erspan_metadata *md;
566                         __be64 tun_id;
567                         __be16 flags;
568
569                         tpi->flags |= TUNNEL_KEY;
570                         flags = tpi->flags;
571                         tun_id = key32_to_tunnel_id(tpi->key);
572
573                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
574                                                   sizeof(*md));
575                         if (!tun_dst)
576                                 return PACKET_REJECT;
577
578                         info = &tun_dst->u.tun_info;
579                         md = ip_tunnel_info_opts(info);
580                         md->version = ver;
581                         md2 = &md->u.md2;
582                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
583                                                        ERSPAN_V2_MDSIZE);
584                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
585                         info->options_len = sizeof(*md);
586
587                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
588
589                 } else {
590                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
591                 }
592
593                 return PACKET_RCVD;
594         }
595
596         return PACKET_REJECT;
597 }
598
599 static int gre_rcv(struct sk_buff *skb)
600 {
601         struct tnl_ptk_info tpi;
602         bool csum_err = false;
603         int hdr_len;
604
605         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
606         if (hdr_len < 0)
607                 goto drop;
608
609         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
610                 goto drop;
611
612         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
613                      tpi.proto == htons(ETH_P_ERSPAN2))) {
614                 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
615                         return 0;
616                 goto out;
617         }
618
619         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
620                 return 0;
621
622 out:
623         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
624 drop:
625         kfree_skb(skb);
626         return 0;
627 }
628
629 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
630 {
631         return iptunnel_handle_offloads(skb,
632                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
633 }
634
635 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
636                                      struct net_device *dev,
637                                      struct flowi6 *fl6, __u8 *dsfield,
638                                      int *encap_limit)
639 {
640         const struct iphdr *iph = ip_hdr(skb);
641         struct ip6_tnl *t = netdev_priv(dev);
642
643         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
644                 *encap_limit = t->parms.encap_limit;
645
646         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
647
648         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
649                 *dsfield = ipv4_get_dsfield(iph);
650         else
651                 *dsfield = ip6_tclass(t->parms.flowinfo);
652
653         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
654                 fl6->flowi6_mark = skb->mark;
655         else
656                 fl6->flowi6_mark = t->parms.fwmark;
657
658         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
659 }
660
661 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
662                                     struct net_device *dev,
663                                     struct flowi6 *fl6, __u8 *dsfield,
664                                     int *encap_limit)
665 {
666         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
667         struct ip6_tnl *t = netdev_priv(dev);
668         __u16 offset;
669
670         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
671         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
672
673         if (offset > 0) {
674                 struct ipv6_tlv_tnl_enc_lim *tel;
675
676                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
677                 if (tel->encap_limit == 0) {
678                         icmpv6_send(skb, ICMPV6_PARAMPROB,
679                                     ICMPV6_HDR_FIELD, offset + 2);
680                         return -1;
681                 }
682                 *encap_limit = tel->encap_limit - 1;
683         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
684                 *encap_limit = t->parms.encap_limit;
685         }
686
687         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
688
689         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
690                 *dsfield = ipv6_get_dsfield(ipv6h);
691         else
692                 *dsfield = ip6_tclass(t->parms.flowinfo);
693
694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
695                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
696
697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
698                 fl6->flowi6_mark = skb->mark;
699         else
700                 fl6->flowi6_mark = t->parms.fwmark;
701
702         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
703
704         return 0;
705 }
706
707 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
708                                struct net_device *dev, __u8 dsfield,
709                                struct flowi6 *fl6, int encap_limit,
710                                __u32 *pmtu, __be16 proto)
711 {
712         struct ip6_tnl *tunnel = netdev_priv(dev);
713         __be16 protocol;
714
715         if (dev->type == ARPHRD_ETHER)
716                 IPCB(skb)->flags = 0;
717
718         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
719                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
720         else
721                 fl6->daddr = tunnel->parms.raddr;
722
723         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
724                 return -ENOMEM;
725
726         /* Push GRE header. */
727         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
728
729         if (tunnel->parms.collect_md) {
730                 struct ip_tunnel_info *tun_info;
731                 const struct ip_tunnel_key *key;
732                 __be16 flags;
733
734                 tun_info = skb_tunnel_info(skb);
735                 if (unlikely(!tun_info ||
736                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
737                              ip_tunnel_info_af(tun_info) != AF_INET6))
738                         return -EINVAL;
739
740                 key = &tun_info->key;
741                 memset(fl6, 0, sizeof(*fl6));
742                 fl6->flowi6_proto = IPPROTO_GRE;
743                 fl6->daddr = key->u.ipv6.dst;
744                 fl6->flowlabel = key->label;
745                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
746
747                 dsfield = key->tos;
748                 flags = key->tun_flags &
749                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
750                 tunnel->tun_hlen = gre_calc_hlen(flags);
751
752                 gre_build_header(skb, tunnel->tun_hlen,
753                                  flags, protocol,
754                                  tunnel_id_to_key32(tun_info->key.tun_id),
755                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
756                                                       : 0);
757
758         } else {
759                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
760                         tunnel->o_seqno++;
761
762                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
763                                  protocol, tunnel->parms.o_key,
764                                  htonl(tunnel->o_seqno));
765         }
766
767         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
768                             NEXTHDR_GRE);
769 }
770
771 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
772 {
773         struct ip6_tnl *t = netdev_priv(dev);
774         int encap_limit = -1;
775         struct flowi6 fl6;
776         __u8 dsfield = 0;
777         __u32 mtu;
778         int err;
779
780         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
781
782         if (!t->parms.collect_md)
783                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
784                                          &dsfield, &encap_limit);
785
786         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
787         if (err)
788                 return -1;
789
790         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
791                           skb->protocol);
792         if (err != 0) {
793                 /* XXX: send ICMP error even if DF is not set. */
794                 if (err == -EMSGSIZE)
795                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
796                                   htonl(mtu));
797                 return -1;
798         }
799
800         return 0;
801 }
802
803 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
804 {
805         struct ip6_tnl *t = netdev_priv(dev);
806         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
807         int encap_limit = -1;
808         struct flowi6 fl6;
809         __u8 dsfield = 0;
810         __u32 mtu;
811         int err;
812
813         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
814                 return -1;
815
816         if (!t->parms.collect_md &&
817             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
818                 return -1;
819
820         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
821                 return -1;
822
823         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
824                           &mtu, skb->protocol);
825         if (err != 0) {
826                 if (err == -EMSGSIZE)
827                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
828                 return -1;
829         }
830
831         return 0;
832 }
833
834 /**
835  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
836  *   @t: the outgoing tunnel device
837  *   @hdr: IPv6 header from the incoming packet
838  *
839  * Description:
840  *   Avoid trivial tunneling loop by checking that tunnel exit-point
841  *   doesn't match source of incoming packet.
842  *
843  * Return:
844  *   1 if conflict,
845  *   0 else
846  **/
847
848 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
849         const struct ipv6hdr *hdr)
850 {
851         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
852 }
853
854 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
855 {
856         struct ip6_tnl *t = netdev_priv(dev);
857         int encap_limit = -1;
858         struct flowi6 fl6;
859         __u32 mtu;
860         int err;
861
862         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
863                 encap_limit = t->parms.encap_limit;
864
865         if (!t->parms.collect_md)
866                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
867
868         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
869         if (err)
870                 return err;
871
872         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
873
874         return err;
875 }
876
877 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
878         struct net_device *dev)
879 {
880         struct ip6_tnl *t = netdev_priv(dev);
881         struct net_device_stats *stats = &t->dev->stats;
882         int ret;
883
884         if (!pskb_inet_may_pull(skb))
885                 goto tx_err;
886
887         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
888                 goto tx_err;
889
890         switch (skb->protocol) {
891         case htons(ETH_P_IP):
892                 ret = ip6gre_xmit_ipv4(skb, dev);
893                 break;
894         case htons(ETH_P_IPV6):
895                 ret = ip6gre_xmit_ipv6(skb, dev);
896                 break;
897         default:
898                 ret = ip6gre_xmit_other(skb, dev);
899                 break;
900         }
901
902         if (ret < 0)
903                 goto tx_err;
904
905         return NETDEV_TX_OK;
906
907 tx_err:
908         stats->tx_errors++;
909         stats->tx_dropped++;
910         kfree_skb(skb);
911         return NETDEV_TX_OK;
912 }
913
914 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
915                                          struct net_device *dev)
916 {
917         struct ip6_tnl *t = netdev_priv(dev);
918         struct dst_entry *dst = skb_dst(skb);
919         struct net_device_stats *stats;
920         bool truncate = false;
921         int encap_limit = -1;
922         __u8 dsfield = false;
923         struct flowi6 fl6;
924         int err = -EINVAL;
925         __be16 proto;
926         __u32 mtu;
927         int nhoff;
928         int thoff;
929
930         if (!pskb_inet_may_pull(skb))
931                 goto tx_err;
932
933         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
934                 goto tx_err;
935
936         if (gre_handle_offloads(skb, false))
937                 goto tx_err;
938
939         if (skb->len > dev->mtu + dev->hard_header_len) {
940                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
941                 truncate = true;
942         }
943
944         nhoff = skb_network_header(skb) - skb_mac_header(skb);
945         if (skb->protocol == htons(ETH_P_IP) &&
946             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
947                 truncate = true;
948
949         thoff = skb_transport_header(skb) - skb_mac_header(skb);
950         if (skb->protocol == htons(ETH_P_IPV6) &&
951             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
952                 truncate = true;
953
954         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
955                 goto tx_err;
956
957         t->parms.o_flags &= ~TUNNEL_KEY;
958         IPCB(skb)->flags = 0;
959
960         /* For collect_md mode, derive fl6 from the tunnel key,
961          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
962          */
963         if (t->parms.collect_md) {
964                 struct ip_tunnel_info *tun_info;
965                 const struct ip_tunnel_key *key;
966                 struct erspan_metadata *md;
967                 __be32 tun_id;
968
969                 tun_info = skb_tunnel_info(skb);
970                 if (unlikely(!tun_info ||
971                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
972                              ip_tunnel_info_af(tun_info) != AF_INET6))
973                         return -EINVAL;
974
975                 key = &tun_info->key;
976                 memset(&fl6, 0, sizeof(fl6));
977                 fl6.flowi6_proto = IPPROTO_GRE;
978                 fl6.daddr = key->u.ipv6.dst;
979                 fl6.flowlabel = key->label;
980                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
981
982                 dsfield = key->tos;
983                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
984                         goto tx_err;
985                 md = ip_tunnel_info_opts(tun_info);
986                 if (!md)
987                         goto tx_err;
988
989                 tun_id = tunnel_id_to_key32(key->tun_id);
990                 if (md->version == 1) {
991                         erspan_build_header(skb,
992                                             ntohl(tun_id),
993                                             ntohl(md->u.index), truncate,
994                                             false);
995                 } else if (md->version == 2) {
996                         erspan_build_header_v2(skb,
997                                                ntohl(tun_id),
998                                                md->u.md2.dir,
999                                                get_hwid(&md->u.md2),
1000                                                truncate, false);
1001                 } else {
1002                         goto tx_err;
1003                 }
1004         } else {
1005                 switch (skb->protocol) {
1006                 case htons(ETH_P_IP):
1007                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1008                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1009                                                  &dsfield, &encap_limit);
1010                         break;
1011                 case htons(ETH_P_IPV6):
1012                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1013                                 goto tx_err;
1014                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1015                                                      &dsfield, &encap_limit))
1016                                 goto tx_err;
1017                         break;
1018                 default:
1019                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1020                         break;
1021                 }
1022
1023                 if (t->parms.erspan_ver == 1)
1024                         erspan_build_header(skb, ntohl(t->parms.o_key),
1025                                             t->parms.index,
1026                                             truncate, false);
1027                 else if (t->parms.erspan_ver == 2)
1028                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1029                                                t->parms.dir,
1030                                                t->parms.hwid,
1031                                                truncate, false);
1032                 else
1033                         goto tx_err;
1034
1035                 fl6.daddr = t->parms.raddr;
1036         }
1037
1038         /* Push GRE header. */
1039         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1040                                            : htons(ETH_P_ERSPAN2);
1041         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1042
1043         /* TooBig packet may have updated dst->dev's mtu */
1044         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1045                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1046
1047         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1048                            NEXTHDR_GRE);
1049         if (err != 0) {
1050                 /* XXX: send ICMP error even if DF is not set. */
1051                 if (err == -EMSGSIZE) {
1052                         if (skb->protocol == htons(ETH_P_IP))
1053                                 icmp_send(skb, ICMP_DEST_UNREACH,
1054                                           ICMP_FRAG_NEEDED, htonl(mtu));
1055                         else
1056                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1057                 }
1058
1059                 goto tx_err;
1060         }
1061         return NETDEV_TX_OK;
1062
1063 tx_err:
1064         stats = &t->dev->stats;
1065         stats->tx_errors++;
1066         stats->tx_dropped++;
1067         kfree_skb(skb);
1068         return NETDEV_TX_OK;
1069 }
1070
1071 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1072 {
1073         struct net_device *dev = t->dev;
1074         struct __ip6_tnl_parm *p = &t->parms;
1075         struct flowi6 *fl6 = &t->fl.u.ip6;
1076
1077         if (dev->type != ARPHRD_ETHER) {
1078                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1079                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1080         }
1081
1082         /* Set up flowi template */
1083         fl6->saddr = p->laddr;
1084         fl6->daddr = p->raddr;
1085         fl6->flowi6_oif = p->link;
1086         fl6->flowlabel = 0;
1087         fl6->flowi6_proto = IPPROTO_GRE;
1088
1089         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1090                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1091         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1092                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1093
1094         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1095         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1096
1097         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1098                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1099                 dev->flags |= IFF_POINTOPOINT;
1100         else
1101                 dev->flags &= ~IFF_POINTOPOINT;
1102 }
1103
1104 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1105                                          int t_hlen)
1106 {
1107         const struct __ip6_tnl_parm *p = &t->parms;
1108         struct net_device *dev = t->dev;
1109
1110         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1111                 int strict = (ipv6_addr_type(&p->raddr) &
1112                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1113
1114                 struct rt6_info *rt = rt6_lookup(t->net,
1115                                                  &p->raddr, &p->laddr,
1116                                                  p->link, NULL, strict);
1117
1118                 if (!rt)
1119                         return;
1120
1121                 if (rt->dst.dev) {
1122                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1123                                                t_hlen;
1124
1125                         if (set_mtu) {
1126                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1127                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1128                                         dev->mtu -= 8;
1129                                 if (dev->type == ARPHRD_ETHER)
1130                                         dev->mtu -= ETH_HLEN;
1131
1132                                 if (dev->mtu < IPV6_MIN_MTU)
1133                                         dev->mtu = IPV6_MIN_MTU;
1134                         }
1135                 }
1136                 ip6_rt_put(rt);
1137         }
1138 }
1139
1140 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1141 {
1142         int t_hlen;
1143
1144         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1145         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1146
1147         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1148         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1149         return t_hlen;
1150 }
1151
1152 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1153 {
1154         ip6gre_tnl_link_config_common(t);
1155         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1156 }
1157
1158 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1159                                      const struct __ip6_tnl_parm *p)
1160 {
1161         t->parms.laddr = p->laddr;
1162         t->parms.raddr = p->raddr;
1163         t->parms.flags = p->flags;
1164         t->parms.hop_limit = p->hop_limit;
1165         t->parms.encap_limit = p->encap_limit;
1166         t->parms.flowinfo = p->flowinfo;
1167         t->parms.link = p->link;
1168         t->parms.proto = p->proto;
1169         t->parms.i_key = p->i_key;
1170         t->parms.o_key = p->o_key;
1171         t->parms.i_flags = p->i_flags;
1172         t->parms.o_flags = p->o_flags;
1173         t->parms.fwmark = p->fwmark;
1174         t->parms.erspan_ver = p->erspan_ver;
1175         t->parms.index = p->index;
1176         t->parms.dir = p->dir;
1177         t->parms.hwid = p->hwid;
1178         dst_cache_reset(&t->dst_cache);
1179 }
1180
1181 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1182                              int set_mtu)
1183 {
1184         ip6gre_tnl_copy_tnl_parm(t, p);
1185         ip6gre_tnl_link_config(t, set_mtu);
1186         return 0;
1187 }
1188
1189 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1190         const struct ip6_tnl_parm2 *u)
1191 {
1192         p->laddr = u->laddr;
1193         p->raddr = u->raddr;
1194         p->flags = u->flags;
1195         p->hop_limit = u->hop_limit;
1196         p->encap_limit = u->encap_limit;
1197         p->flowinfo = u->flowinfo;
1198         p->link = u->link;
1199         p->i_key = u->i_key;
1200         p->o_key = u->o_key;
1201         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1202         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1203         memcpy(p->name, u->name, sizeof(u->name));
1204 }
1205
1206 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1207         const struct __ip6_tnl_parm *p)
1208 {
1209         u->proto = IPPROTO_GRE;
1210         u->laddr = p->laddr;
1211         u->raddr = p->raddr;
1212         u->flags = p->flags;
1213         u->hop_limit = p->hop_limit;
1214         u->encap_limit = p->encap_limit;
1215         u->flowinfo = p->flowinfo;
1216         u->link = p->link;
1217         u->i_key = p->i_key;
1218         u->o_key = p->o_key;
1219         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1220         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1221         memcpy(u->name, p->name, sizeof(u->name));
1222 }
1223
1224 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1225         struct ifreq *ifr, int cmd)
1226 {
1227         int err = 0;
1228         struct ip6_tnl_parm2 p;
1229         struct __ip6_tnl_parm p1;
1230         struct ip6_tnl *t = netdev_priv(dev);
1231         struct net *net = t->net;
1232         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1233
1234         memset(&p1, 0, sizeof(p1));
1235
1236         switch (cmd) {
1237         case SIOCGETTUNNEL:
1238                 if (dev == ign->fb_tunnel_dev) {
1239                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1240                                 err = -EFAULT;
1241                                 break;
1242                         }
1243                         ip6gre_tnl_parm_from_user(&p1, &p);
1244                         t = ip6gre_tunnel_locate(net, &p1, 0);
1245                         if (!t)
1246                                 t = netdev_priv(dev);
1247                 }
1248                 memset(&p, 0, sizeof(p));
1249                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1250                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1251                         err = -EFAULT;
1252                 break;
1253
1254         case SIOCADDTUNNEL:
1255         case SIOCCHGTUNNEL:
1256                 err = -EPERM;
1257                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1258                         goto done;
1259
1260                 err = -EFAULT;
1261                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1262                         goto done;
1263
1264                 err = -EINVAL;
1265                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1266                         goto done;
1267
1268                 if (!(p.i_flags&GRE_KEY))
1269                         p.i_key = 0;
1270                 if (!(p.o_flags&GRE_KEY))
1271                         p.o_key = 0;
1272
1273                 ip6gre_tnl_parm_from_user(&p1, &p);
1274                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1275
1276                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1277                         if (t) {
1278                                 if (t->dev != dev) {
1279                                         err = -EEXIST;
1280                                         break;
1281                                 }
1282                         } else {
1283                                 t = netdev_priv(dev);
1284
1285                                 ip6gre_tunnel_unlink(ign, t);
1286                                 synchronize_net();
1287                                 ip6gre_tnl_change(t, &p1, 1);
1288                                 ip6gre_tunnel_link(ign, t);
1289                                 netdev_state_change(dev);
1290                         }
1291                 }
1292
1293                 if (t) {
1294                         err = 0;
1295
1296                         memset(&p, 0, sizeof(p));
1297                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1298                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1299                                 err = -EFAULT;
1300                 } else
1301                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1302                 break;
1303
1304         case SIOCDELTUNNEL:
1305                 err = -EPERM;
1306                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1307                         goto done;
1308
1309                 if (dev == ign->fb_tunnel_dev) {
1310                         err = -EFAULT;
1311                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1312                                 goto done;
1313                         err = -ENOENT;
1314                         ip6gre_tnl_parm_from_user(&p1, &p);
1315                         t = ip6gre_tunnel_locate(net, &p1, 0);
1316                         if (!t)
1317                                 goto done;
1318                         err = -EPERM;
1319                         if (t == netdev_priv(ign->fb_tunnel_dev))
1320                                 goto done;
1321                         dev = t->dev;
1322                 }
1323                 unregister_netdevice(dev);
1324                 err = 0;
1325                 break;
1326
1327         default:
1328                 err = -EINVAL;
1329         }
1330
1331 done:
1332         return err;
1333 }
1334
1335 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1336                          unsigned short type, const void *daddr,
1337                          const void *saddr, unsigned int len)
1338 {
1339         struct ip6_tnl *t = netdev_priv(dev);
1340         struct ipv6hdr *ipv6h;
1341         __be16 *p;
1342
1343         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1344         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1345                                                   t->fl.u.ip6.flowlabel,
1346                                                   true, &t->fl.u.ip6));
1347         ipv6h->hop_limit = t->parms.hop_limit;
1348         ipv6h->nexthdr = NEXTHDR_GRE;
1349         ipv6h->saddr = t->parms.laddr;
1350         ipv6h->daddr = t->parms.raddr;
1351
1352         p = (__be16 *)(ipv6h + 1);
1353         p[0] = t->parms.o_flags;
1354         p[1] = htons(type);
1355
1356         /*
1357          *      Set the source hardware address.
1358          */
1359
1360         if (saddr)
1361                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1362         if (daddr)
1363                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1364         if (!ipv6_addr_any(&ipv6h->daddr))
1365                 return t->hlen;
1366
1367         return -t->hlen;
1368 }
1369
1370 static const struct header_ops ip6gre_header_ops = {
1371         .create = ip6gre_header,
1372 };
1373
1374 static const struct net_device_ops ip6gre_netdev_ops = {
1375         .ndo_init               = ip6gre_tunnel_init,
1376         .ndo_uninit             = ip6gre_tunnel_uninit,
1377         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1378         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1379         .ndo_change_mtu         = ip6_tnl_change_mtu,
1380         .ndo_get_stats64        = ip_tunnel_get_stats64,
1381         .ndo_get_iflink         = ip6_tnl_get_iflink,
1382 };
1383
1384 static void ip6gre_dev_free(struct net_device *dev)
1385 {
1386         struct ip6_tnl *t = netdev_priv(dev);
1387
1388         gro_cells_destroy(&t->gro_cells);
1389         dst_cache_destroy(&t->dst_cache);
1390         free_percpu(dev->tstats);
1391 }
1392
1393 static void ip6gre_tunnel_setup(struct net_device *dev)
1394 {
1395         dev->netdev_ops = &ip6gre_netdev_ops;
1396         dev->needs_free_netdev = true;
1397         dev->priv_destructor = ip6gre_dev_free;
1398
1399         dev->type = ARPHRD_IP6GRE;
1400
1401         dev->flags |= IFF_NOARP;
1402         dev->addr_len = sizeof(struct in6_addr);
1403         netif_keep_dst(dev);
1404         /* This perm addr will be used as interface identifier by IPv6 */
1405         dev->addr_assign_type = NET_ADDR_RANDOM;
1406         eth_random_addr(dev->perm_addr);
1407 }
1408
1409 #define GRE6_FEATURES (NETIF_F_SG |             \
1410                        NETIF_F_FRAGLIST |       \
1411                        NETIF_F_HIGHDMA |        \
1412                        NETIF_F_HW_CSUM)
1413
1414 static void ip6gre_tnl_init_features(struct net_device *dev)
1415 {
1416         struct ip6_tnl *nt = netdev_priv(dev);
1417
1418         dev->features           |= GRE6_FEATURES;
1419         dev->hw_features        |= GRE6_FEATURES;
1420
1421         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1422                 /* TCP offload with GRE SEQ is not supported, nor
1423                  * can we support 2 levels of outer headers requiring
1424                  * an update.
1425                  */
1426                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1427                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1428                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1429                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1430                 }
1431
1432                 /* Can use a lockless transmit, unless we generate
1433                  * output sequences
1434                  */
1435                 dev->features |= NETIF_F_LLTX;
1436         }
1437 }
1438
1439 static int ip6gre_tunnel_init_common(struct net_device *dev)
1440 {
1441         struct ip6_tnl *tunnel;
1442         int ret;
1443         int t_hlen;
1444
1445         tunnel = netdev_priv(dev);
1446
1447         tunnel->dev = dev;
1448         tunnel->net = dev_net(dev);
1449         strcpy(tunnel->parms.name, dev->name);
1450
1451         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1452         if (!dev->tstats)
1453                 return -ENOMEM;
1454
1455         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1456         if (ret)
1457                 goto cleanup_alloc_pcpu_stats;
1458
1459         ret = gro_cells_init(&tunnel->gro_cells, dev);
1460         if (ret)
1461                 goto cleanup_dst_cache_init;
1462
1463         t_hlen = ip6gre_calc_hlen(tunnel);
1464         dev->mtu = ETH_DATA_LEN - t_hlen;
1465         if (dev->type == ARPHRD_ETHER)
1466                 dev->mtu -= ETH_HLEN;
1467         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1468                 dev->mtu -= 8;
1469
1470         if (tunnel->parms.collect_md) {
1471                 dev->features |= NETIF_F_NETNS_LOCAL;
1472                 netif_keep_dst(dev);
1473         }
1474         ip6gre_tnl_init_features(dev);
1475
1476         return 0;
1477
1478 cleanup_dst_cache_init:
1479         dst_cache_destroy(&tunnel->dst_cache);
1480 cleanup_alloc_pcpu_stats:
1481         free_percpu(dev->tstats);
1482         dev->tstats = NULL;
1483         return ret;
1484 }
1485
1486 static int ip6gre_tunnel_init(struct net_device *dev)
1487 {
1488         struct ip6_tnl *tunnel;
1489         int ret;
1490
1491         ret = ip6gre_tunnel_init_common(dev);
1492         if (ret)
1493                 return ret;
1494
1495         tunnel = netdev_priv(dev);
1496
1497         if (tunnel->parms.collect_md)
1498                 return 0;
1499
1500         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1501         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1502
1503         if (ipv6_addr_any(&tunnel->parms.raddr))
1504                 dev->header_ops = &ip6gre_header_ops;
1505
1506         return 0;
1507 }
1508
1509 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1510 {
1511         struct ip6_tnl *tunnel = netdev_priv(dev);
1512
1513         tunnel->dev = dev;
1514         tunnel->net = dev_net(dev);
1515         strcpy(tunnel->parms.name, dev->name);
1516
1517         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1518
1519         dev_hold(dev);
1520 }
1521
1522 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1523         .handler     = gre_rcv,
1524         .err_handler = ip6gre_err,
1525         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1526 };
1527
1528 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1529 {
1530         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1531         struct net_device *dev, *aux;
1532         int prio;
1533
1534         for_each_netdev_safe(net, dev, aux)
1535                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1536                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1537                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1538                         unregister_netdevice_queue(dev, head);
1539
1540         for (prio = 0; prio < 4; prio++) {
1541                 int h;
1542                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1543                         struct ip6_tnl *t;
1544
1545                         t = rtnl_dereference(ign->tunnels[prio][h]);
1546
1547                         while (t) {
1548                                 /* If dev is in the same netns, it has already
1549                                  * been added to the list by the previous loop.
1550                                  */
1551                                 if (!net_eq(dev_net(t->dev), net))
1552                                         unregister_netdevice_queue(t->dev,
1553                                                                    head);
1554                                 t = rtnl_dereference(t->next);
1555                         }
1556                 }
1557         }
1558 }
1559
1560 static int __net_init ip6gre_init_net(struct net *net)
1561 {
1562         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1563         int err;
1564
1565         if (!net_has_fallback_tunnels(net))
1566                 return 0;
1567         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1568                                           NET_NAME_UNKNOWN,
1569                                           ip6gre_tunnel_setup);
1570         if (!ign->fb_tunnel_dev) {
1571                 err = -ENOMEM;
1572                 goto err_alloc_dev;
1573         }
1574         dev_net_set(ign->fb_tunnel_dev, net);
1575         /* FB netdevice is special: we have one, and only one per netns.
1576          * Allowing to move it to another netns is clearly unsafe.
1577          */
1578         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1579
1580
1581         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1582         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1583
1584         err = register_netdev(ign->fb_tunnel_dev);
1585         if (err)
1586                 goto err_reg_dev;
1587
1588         rcu_assign_pointer(ign->tunnels_wc[0],
1589                            netdev_priv(ign->fb_tunnel_dev));
1590         return 0;
1591
1592 err_reg_dev:
1593         free_netdev(ign->fb_tunnel_dev);
1594 err_alloc_dev:
1595         return err;
1596 }
1597
1598 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1599 {
1600         struct net *net;
1601         LIST_HEAD(list);
1602
1603         rtnl_lock();
1604         list_for_each_entry(net, net_list, exit_list)
1605                 ip6gre_destroy_tunnels(net, &list);
1606         unregister_netdevice_many(&list);
1607         rtnl_unlock();
1608 }
1609
1610 static struct pernet_operations ip6gre_net_ops = {
1611         .init = ip6gre_init_net,
1612         .exit_batch = ip6gre_exit_batch_net,
1613         .id   = &ip6gre_net_id,
1614         .size = sizeof(struct ip6gre_net),
1615 };
1616
1617 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1618                                   struct netlink_ext_ack *extack)
1619 {
1620         __be16 flags;
1621
1622         if (!data)
1623                 return 0;
1624
1625         flags = 0;
1626         if (data[IFLA_GRE_IFLAGS])
1627                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1628         if (data[IFLA_GRE_OFLAGS])
1629                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1630         if (flags & (GRE_VERSION|GRE_ROUTING))
1631                 return -EINVAL;
1632
1633         return 0;
1634 }
1635
1636 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1637                                struct netlink_ext_ack *extack)
1638 {
1639         struct in6_addr daddr;
1640
1641         if (tb[IFLA_ADDRESS]) {
1642                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1643                         return -EINVAL;
1644                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1645                         return -EADDRNOTAVAIL;
1646         }
1647
1648         if (!data)
1649                 goto out;
1650
1651         if (data[IFLA_GRE_REMOTE]) {
1652                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1653                 if (ipv6_addr_any(&daddr))
1654                         return -EINVAL;
1655         }
1656
1657 out:
1658         return ip6gre_tunnel_validate(tb, data, extack);
1659 }
1660
1661 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1662                                   struct netlink_ext_ack *extack)
1663 {
1664         __be16 flags = 0;
1665         int ret, ver = 0;
1666
1667         if (!data)
1668                 return 0;
1669
1670         ret = ip6gre_tap_validate(tb, data, extack);
1671         if (ret)
1672                 return ret;
1673
1674         /* ERSPAN should only have GRE sequence and key flag */
1675         if (data[IFLA_GRE_OFLAGS])
1676                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1677         if (data[IFLA_GRE_IFLAGS])
1678                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1679         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1680             flags != (GRE_SEQ | GRE_KEY))
1681                 return -EINVAL;
1682
1683         /* ERSPAN Session ID only has 10-bit. Since we reuse
1684          * 32-bit key field as ID, check it's range.
1685          */
1686         if (data[IFLA_GRE_IKEY] &&
1687             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1688                 return -EINVAL;
1689
1690         if (data[IFLA_GRE_OKEY] &&
1691             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1692                 return -EINVAL;
1693
1694         if (data[IFLA_GRE_ERSPAN_VER]) {
1695                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1696                 if (ver != 1 && ver != 2)
1697                         return -EINVAL;
1698         }
1699
1700         if (ver == 1) {
1701                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1702                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1703
1704                         if (index & ~INDEX_MASK)
1705                                 return -EINVAL;
1706                 }
1707         } else if (ver == 2) {
1708                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1709                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1710
1711                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1712                                 return -EINVAL;
1713                 }
1714
1715                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1716                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1717
1718                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1719                                 return -EINVAL;
1720                 }
1721         }
1722
1723         return 0;
1724 }
1725
1726 static void ip6gre_netlink_parms(struct nlattr *data[],
1727                                 struct __ip6_tnl_parm *parms)
1728 {
1729         memset(parms, 0, sizeof(*parms));
1730
1731         if (!data)
1732                 return;
1733
1734         if (data[IFLA_GRE_LINK])
1735                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1736
1737         if (data[IFLA_GRE_IFLAGS])
1738                 parms->i_flags = gre_flags_to_tnl_flags(
1739                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1740
1741         if (data[IFLA_GRE_OFLAGS])
1742                 parms->o_flags = gre_flags_to_tnl_flags(
1743                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1744
1745         if (data[IFLA_GRE_IKEY])
1746                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1747
1748         if (data[IFLA_GRE_OKEY])
1749                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1750
1751         if (data[IFLA_GRE_LOCAL])
1752                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1753
1754         if (data[IFLA_GRE_REMOTE])
1755                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1756
1757         if (data[IFLA_GRE_TTL])
1758                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1759
1760         if (data[IFLA_GRE_ENCAP_LIMIT])
1761                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1762
1763         if (data[IFLA_GRE_FLOWINFO])
1764                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1765
1766         if (data[IFLA_GRE_FLAGS])
1767                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1768
1769         if (data[IFLA_GRE_FWMARK])
1770                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1771
1772         if (data[IFLA_GRE_COLLECT_METADATA])
1773                 parms->collect_md = true;
1774
1775         parms->erspan_ver = 1;
1776         if (data[IFLA_GRE_ERSPAN_VER])
1777                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1778
1779         if (parms->erspan_ver == 1) {
1780                 if (data[IFLA_GRE_ERSPAN_INDEX])
1781                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1782         } else if (parms->erspan_ver == 2) {
1783                 if (data[IFLA_GRE_ERSPAN_DIR])
1784                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1785                 if (data[IFLA_GRE_ERSPAN_HWID])
1786                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1787         }
1788 }
1789
1790 static int ip6gre_tap_init(struct net_device *dev)
1791 {
1792         int ret;
1793
1794         ret = ip6gre_tunnel_init_common(dev);
1795         if (ret)
1796                 return ret;
1797
1798         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1799
1800         return 0;
1801 }
1802
1803 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1804         .ndo_init = ip6gre_tap_init,
1805         .ndo_uninit = ip6gre_tunnel_uninit,
1806         .ndo_start_xmit = ip6gre_tunnel_xmit,
1807         .ndo_set_mac_address = eth_mac_addr,
1808         .ndo_validate_addr = eth_validate_addr,
1809         .ndo_change_mtu = ip6_tnl_change_mtu,
1810         .ndo_get_stats64 = ip_tunnel_get_stats64,
1811         .ndo_get_iflink = ip6_tnl_get_iflink,
1812 };
1813
1814 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1815 {
1816         int t_hlen;
1817
1818         tunnel->tun_hlen = 8;
1819         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1820                        erspan_hdr_len(tunnel->parms.erspan_ver);
1821
1822         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1823         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1824         return t_hlen;
1825 }
1826
1827 static int ip6erspan_tap_init(struct net_device *dev)
1828 {
1829         struct ip6_tnl *tunnel;
1830         int t_hlen;
1831         int ret;
1832
1833         tunnel = netdev_priv(dev);
1834
1835         tunnel->dev = dev;
1836         tunnel->net = dev_net(dev);
1837         strcpy(tunnel->parms.name, dev->name);
1838
1839         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1840         if (!dev->tstats)
1841                 return -ENOMEM;
1842
1843         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1844         if (ret)
1845                 goto cleanup_alloc_pcpu_stats;
1846
1847         ret = gro_cells_init(&tunnel->gro_cells, dev);
1848         if (ret)
1849                 goto cleanup_dst_cache_init;
1850
1851         t_hlen = ip6erspan_calc_hlen(tunnel);
1852         dev->mtu = ETH_DATA_LEN - t_hlen;
1853         if (dev->type == ARPHRD_ETHER)
1854                 dev->mtu -= ETH_HLEN;
1855         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1856                 dev->mtu -= 8;
1857
1858         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1859         ip6erspan_tnl_link_config(tunnel, 1);
1860
1861         return 0;
1862
1863 cleanup_dst_cache_init:
1864         dst_cache_destroy(&tunnel->dst_cache);
1865 cleanup_alloc_pcpu_stats:
1866         free_percpu(dev->tstats);
1867         dev->tstats = NULL;
1868         return ret;
1869 }
1870
1871 static const struct net_device_ops ip6erspan_netdev_ops = {
1872         .ndo_init =             ip6erspan_tap_init,
1873         .ndo_uninit =           ip6erspan_tunnel_uninit,
1874         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1875         .ndo_set_mac_address =  eth_mac_addr,
1876         .ndo_validate_addr =    eth_validate_addr,
1877         .ndo_change_mtu =       ip6_tnl_change_mtu,
1878         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1879         .ndo_get_iflink =       ip6_tnl_get_iflink,
1880 };
1881
1882 static void ip6gre_tap_setup(struct net_device *dev)
1883 {
1884
1885         ether_setup(dev);
1886
1887         dev->max_mtu = 0;
1888         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1889         dev->needs_free_netdev = true;
1890         dev->priv_destructor = ip6gre_dev_free;
1891
1892         dev->features |= NETIF_F_NETNS_LOCAL;
1893         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1894         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1895         netif_keep_dst(dev);
1896 }
1897
1898 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1899                                        struct ip_tunnel_encap *ipencap)
1900 {
1901         bool ret = false;
1902
1903         memset(ipencap, 0, sizeof(*ipencap));
1904
1905         if (!data)
1906                 return ret;
1907
1908         if (data[IFLA_GRE_ENCAP_TYPE]) {
1909                 ret = true;
1910                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1911         }
1912
1913         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1914                 ret = true;
1915                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1916         }
1917
1918         if (data[IFLA_GRE_ENCAP_SPORT]) {
1919                 ret = true;
1920                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1921         }
1922
1923         if (data[IFLA_GRE_ENCAP_DPORT]) {
1924                 ret = true;
1925                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1926         }
1927
1928         return ret;
1929 }
1930
1931 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1932                                  struct nlattr *tb[], struct nlattr *data[],
1933                                  struct netlink_ext_ack *extack)
1934 {
1935         struct ip6_tnl *nt;
1936         struct ip_tunnel_encap ipencap;
1937         int err;
1938
1939         nt = netdev_priv(dev);
1940
1941         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1942                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1943
1944                 if (err < 0)
1945                         return err;
1946         }
1947
1948         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1949                 eth_hw_addr_random(dev);
1950
1951         nt->dev = dev;
1952         nt->net = dev_net(dev);
1953
1954         err = register_netdevice(dev);
1955         if (err)
1956                 goto out;
1957
1958         if (tb[IFLA_MTU])
1959                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1960
1961         dev_hold(dev);
1962
1963 out:
1964         return err;
1965 }
1966
1967 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1968                           struct nlattr *tb[], struct nlattr *data[],
1969                           struct netlink_ext_ack *extack)
1970 {
1971         struct ip6_tnl *nt = netdev_priv(dev);
1972         struct net *net = dev_net(dev);
1973         struct ip6gre_net *ign;
1974         int err;
1975
1976         ip6gre_netlink_parms(data, &nt->parms);
1977         ign = net_generic(net, ip6gre_net_id);
1978
1979         if (nt->parms.collect_md) {
1980                 if (rtnl_dereference(ign->collect_md_tun))
1981                         return -EEXIST;
1982         } else {
1983                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1984                         return -EEXIST;
1985         }
1986
1987         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1988         if (!err) {
1989                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1990                 ip6gre_tunnel_link_md(ign, nt);
1991                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1992         }
1993         return err;
1994 }
1995
1996 static struct ip6_tnl *
1997 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
1998                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
1999                          struct netlink_ext_ack *extack)
2000 {
2001         struct ip6_tnl *t, *nt = netdev_priv(dev);
2002         struct net *net = nt->net;
2003         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2004         struct ip_tunnel_encap ipencap;
2005
2006         if (dev == ign->fb_tunnel_dev)
2007                 return ERR_PTR(-EINVAL);
2008
2009         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2010                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2011
2012                 if (err < 0)
2013                         return ERR_PTR(err);
2014         }
2015
2016         ip6gre_netlink_parms(data, p_p);
2017
2018         t = ip6gre_tunnel_locate(net, p_p, 0);
2019
2020         if (t) {
2021                 if (t->dev != dev)
2022                         return ERR_PTR(-EEXIST);
2023         } else {
2024                 t = nt;
2025         }
2026
2027         return t;
2028 }
2029
2030 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2031                              struct nlattr *data[],
2032                              struct netlink_ext_ack *extack)
2033 {
2034         struct ip6_tnl *t = netdev_priv(dev);
2035         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2036         struct __ip6_tnl_parm p;
2037
2038         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2039         if (IS_ERR(t))
2040                 return PTR_ERR(t);
2041
2042         ip6gre_tunnel_unlink_md(ign, t);
2043         ip6gre_tunnel_unlink(ign, t);
2044         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2045         ip6gre_tunnel_link_md(ign, t);
2046         ip6gre_tunnel_link(ign, t);
2047         return 0;
2048 }
2049
2050 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2051 {
2052         struct net *net = dev_net(dev);
2053         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2054
2055         if (dev != ign->fb_tunnel_dev)
2056                 unregister_netdevice_queue(dev, head);
2057 }
2058
2059 static size_t ip6gre_get_size(const struct net_device *dev)
2060 {
2061         return
2062                 /* IFLA_GRE_LINK */
2063                 nla_total_size(4) +
2064                 /* IFLA_GRE_IFLAGS */
2065                 nla_total_size(2) +
2066                 /* IFLA_GRE_OFLAGS */
2067                 nla_total_size(2) +
2068                 /* IFLA_GRE_IKEY */
2069                 nla_total_size(4) +
2070                 /* IFLA_GRE_OKEY */
2071                 nla_total_size(4) +
2072                 /* IFLA_GRE_LOCAL */
2073                 nla_total_size(sizeof(struct in6_addr)) +
2074                 /* IFLA_GRE_REMOTE */
2075                 nla_total_size(sizeof(struct in6_addr)) +
2076                 /* IFLA_GRE_TTL */
2077                 nla_total_size(1) +
2078                 /* IFLA_GRE_ENCAP_LIMIT */
2079                 nla_total_size(1) +
2080                 /* IFLA_GRE_FLOWINFO */
2081                 nla_total_size(4) +
2082                 /* IFLA_GRE_FLAGS */
2083                 nla_total_size(4) +
2084                 /* IFLA_GRE_ENCAP_TYPE */
2085                 nla_total_size(2) +
2086                 /* IFLA_GRE_ENCAP_FLAGS */
2087                 nla_total_size(2) +
2088                 /* IFLA_GRE_ENCAP_SPORT */
2089                 nla_total_size(2) +
2090                 /* IFLA_GRE_ENCAP_DPORT */
2091                 nla_total_size(2) +
2092                 /* IFLA_GRE_COLLECT_METADATA */
2093                 nla_total_size(0) +
2094                 /* IFLA_GRE_FWMARK */
2095                 nla_total_size(4) +
2096                 /* IFLA_GRE_ERSPAN_INDEX */
2097                 nla_total_size(4) +
2098                 0;
2099 }
2100
2101 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2102 {
2103         struct ip6_tnl *t = netdev_priv(dev);
2104         struct __ip6_tnl_parm *p = &t->parms;
2105
2106         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2107             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2108                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2109             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2110                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2111             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2112             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2113             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2114             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2115             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2116             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2117             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2118             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2119             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2120             nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2121                 goto nla_put_failure;
2122
2123         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2124                         t->encap.type) ||
2125             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2126                          t->encap.sport) ||
2127             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2128                          t->encap.dport) ||
2129             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2130                         t->encap.flags))
2131                 goto nla_put_failure;
2132
2133         if (p->collect_md) {
2134                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2135                         goto nla_put_failure;
2136         }
2137
2138         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2139                 goto nla_put_failure;
2140
2141         if (p->erspan_ver == 1) {
2142                 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2143                         goto nla_put_failure;
2144         } else if (p->erspan_ver == 2) {
2145                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2146                         goto nla_put_failure;
2147                 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2148                         goto nla_put_failure;
2149         }
2150
2151         return 0;
2152
2153 nla_put_failure:
2154         return -EMSGSIZE;
2155 }
2156
2157 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2158         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2159         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2160         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2161         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2162         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2163         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2164         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2165         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2166         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2167         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2168         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2169         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2170         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2171         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2172         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2173         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2174         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2175         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2176         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2177         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2178         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2179 };
2180
2181 static void ip6erspan_tap_setup(struct net_device *dev)
2182 {
2183         ether_setup(dev);
2184
2185         dev->netdev_ops = &ip6erspan_netdev_ops;
2186         dev->needs_free_netdev = true;
2187         dev->priv_destructor = ip6gre_dev_free;
2188
2189         dev->features |= NETIF_F_NETNS_LOCAL;
2190         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2191         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2192         netif_keep_dst(dev);
2193 }
2194
2195 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2196                              struct nlattr *tb[], struct nlattr *data[],
2197                              struct netlink_ext_ack *extack)
2198 {
2199         struct ip6_tnl *nt = netdev_priv(dev);
2200         struct net *net = dev_net(dev);
2201         struct ip6gre_net *ign;
2202         int err;
2203
2204         ip6gre_netlink_parms(data, &nt->parms);
2205         ign = net_generic(net, ip6gre_net_id);
2206
2207         if (nt->parms.collect_md) {
2208                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2209                         return -EEXIST;
2210         } else {
2211                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2212                         return -EEXIST;
2213         }
2214
2215         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2216         if (!err) {
2217                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2218                 ip6erspan_tunnel_link_md(ign, nt);
2219                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2220         }
2221         return err;
2222 }
2223
2224 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2225 {
2226         ip6gre_tnl_link_config_common(t);
2227         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2228 }
2229
2230 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2231                                 const struct __ip6_tnl_parm *p, int set_mtu)
2232 {
2233         ip6gre_tnl_copy_tnl_parm(t, p);
2234         ip6erspan_tnl_link_config(t, set_mtu);
2235         return 0;
2236 }
2237
2238 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2239                                 struct nlattr *data[],
2240                                 struct netlink_ext_ack *extack)
2241 {
2242         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2243         struct __ip6_tnl_parm p;
2244         struct ip6_tnl *t;
2245
2246         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2247         if (IS_ERR(t))
2248                 return PTR_ERR(t);
2249
2250         ip6gre_tunnel_unlink_md(ign, t);
2251         ip6gre_tunnel_unlink(ign, t);
2252         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2253         ip6erspan_tunnel_link_md(ign, t);
2254         ip6gre_tunnel_link(ign, t);
2255         return 0;
2256 }
2257
2258 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2259         .kind           = "ip6gre",
2260         .maxtype        = IFLA_GRE_MAX,
2261         .policy         = ip6gre_policy,
2262         .priv_size      = sizeof(struct ip6_tnl),
2263         .setup          = ip6gre_tunnel_setup,
2264         .validate       = ip6gre_tunnel_validate,
2265         .newlink        = ip6gre_newlink,
2266         .changelink     = ip6gre_changelink,
2267         .dellink        = ip6gre_dellink,
2268         .get_size       = ip6gre_get_size,
2269         .fill_info      = ip6gre_fill_info,
2270         .get_link_net   = ip6_tnl_get_link_net,
2271 };
2272
2273 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2274         .kind           = "ip6gretap",
2275         .maxtype        = IFLA_GRE_MAX,
2276         .policy         = ip6gre_policy,
2277         .priv_size      = sizeof(struct ip6_tnl),
2278         .setup          = ip6gre_tap_setup,
2279         .validate       = ip6gre_tap_validate,
2280         .newlink        = ip6gre_newlink,
2281         .changelink     = ip6gre_changelink,
2282         .get_size       = ip6gre_get_size,
2283         .fill_info      = ip6gre_fill_info,
2284         .get_link_net   = ip6_tnl_get_link_net,
2285 };
2286
2287 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2288         .kind           = "ip6erspan",
2289         .maxtype        = IFLA_GRE_MAX,
2290         .policy         = ip6gre_policy,
2291         .priv_size      = sizeof(struct ip6_tnl),
2292         .setup          = ip6erspan_tap_setup,
2293         .validate       = ip6erspan_tap_validate,
2294         .newlink        = ip6erspan_newlink,
2295         .changelink     = ip6erspan_changelink,
2296         .get_size       = ip6gre_get_size,
2297         .fill_info      = ip6gre_fill_info,
2298         .get_link_net   = ip6_tnl_get_link_net,
2299 };
2300
2301 /*
2302  *      And now the modules code and kernel interface.
2303  */
2304
2305 static int __init ip6gre_init(void)
2306 {
2307         int err;
2308
2309         pr_info("GRE over IPv6 tunneling driver\n");
2310
2311         err = register_pernet_device(&ip6gre_net_ops);
2312         if (err < 0)
2313                 return err;
2314
2315         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2316         if (err < 0) {
2317                 pr_info("%s: can't add protocol\n", __func__);
2318                 goto add_proto_failed;
2319         }
2320
2321         err = rtnl_link_register(&ip6gre_link_ops);
2322         if (err < 0)
2323                 goto rtnl_link_failed;
2324
2325         err = rtnl_link_register(&ip6gre_tap_ops);
2326         if (err < 0)
2327                 goto tap_ops_failed;
2328
2329         err = rtnl_link_register(&ip6erspan_tap_ops);
2330         if (err < 0)
2331                 goto erspan_link_failed;
2332
2333 out:
2334         return err;
2335
2336 erspan_link_failed:
2337         rtnl_link_unregister(&ip6gre_tap_ops);
2338 tap_ops_failed:
2339         rtnl_link_unregister(&ip6gre_link_ops);
2340 rtnl_link_failed:
2341         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2342 add_proto_failed:
2343         unregister_pernet_device(&ip6gre_net_ops);
2344         goto out;
2345 }
2346
2347 static void __exit ip6gre_fini(void)
2348 {
2349         rtnl_link_unregister(&ip6gre_tap_ops);
2350         rtnl_link_unregister(&ip6gre_link_ops);
2351         rtnl_link_unregister(&ip6erspan_tap_ops);
2352         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2353         unregister_pernet_device(&ip6gre_net_ops);
2354 }
2355
2356 module_init(ip6gre_init);
2357 module_exit(ip6gre_fini);
2358 MODULE_LICENSE("GPL");
2359 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2360 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2361 MODULE_ALIAS_RTNL_LINK("ip6gre");
2362 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2363 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2364 MODULE_ALIAS_NETDEV("ip6gre0");