]> asedeno.scripts.mit.edu Git - linux.git/blob - net/ipv6/ip6_gre.c
Merge tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
[linux.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 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;
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;
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;
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;
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;
476         case ICMPV6_PKT_TOOBIG:
477                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
478                 return;
479         case NDISC_REDIRECT:
480                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
481                              sock_net_uid(net, NULL));
482                 return;
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
492 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
493 {
494         const struct ipv6hdr *ipv6h;
495         struct ip6_tnl *tunnel;
496
497         ipv6h = ipv6_hdr(skb);
498         tunnel = ip6gre_tunnel_lookup(skb->dev,
499                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
500                                       tpi->proto);
501         if (tunnel) {
502                 if (tunnel->parms.collect_md) {
503                         struct metadata_dst *tun_dst;
504                         __be64 tun_id;
505                         __be16 flags;
506
507                         flags = tpi->flags;
508                         tun_id = key32_to_tunnel_id(tpi->key);
509
510                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
511                         if (!tun_dst)
512                                 return PACKET_REJECT;
513
514                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
515                 } else {
516                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
517                 }
518
519                 return PACKET_RCVD;
520         }
521
522         return PACKET_REJECT;
523 }
524
525 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
526                          struct tnl_ptk_info *tpi)
527 {
528         struct erspan_base_hdr *ershdr;
529         struct erspan_metadata *pkt_md;
530         const struct ipv6hdr *ipv6h;
531         struct erspan_md2 *md2;
532         struct ip6_tnl *tunnel;
533         u8 ver;
534
535         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
536                 return PACKET_REJECT;
537
538         ipv6h = ipv6_hdr(skb);
539         ershdr = (struct erspan_base_hdr *)skb->data;
540         ver = ershdr->ver;
541         tpi->key = cpu_to_be32(get_session_id(ershdr));
542
543         tunnel = ip6gre_tunnel_lookup(skb->dev,
544                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
545                                       tpi->proto);
546         if (tunnel) {
547                 int len = erspan_hdr_len(ver);
548
549                 if (unlikely(!pskb_may_pull(skb, len)))
550                         return PACKET_REJECT;
551
552                 ershdr = (struct erspan_base_hdr *)skb->data;
553                 pkt_md = (struct erspan_metadata *)(ershdr + 1);
554
555                 if (__iptunnel_pull_header(skb, len,
556                                            htons(ETH_P_TEB),
557                                            false, false) < 0)
558                         return PACKET_REJECT;
559
560                 if (tunnel->parms.collect_md) {
561                         struct metadata_dst *tun_dst;
562                         struct ip_tunnel_info *info;
563                         struct erspan_metadata *md;
564                         __be64 tun_id;
565                         __be16 flags;
566
567                         tpi->flags |= TUNNEL_KEY;
568                         flags = tpi->flags;
569                         tun_id = key32_to_tunnel_id(tpi->key);
570
571                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
572                                                   sizeof(*md));
573                         if (!tun_dst)
574                                 return PACKET_REJECT;
575
576                         info = &tun_dst->u.tun_info;
577                         md = ip_tunnel_info_opts(info);
578                         md->version = ver;
579                         md2 = &md->u.md2;
580                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
581                                                        ERSPAN_V2_MDSIZE);
582                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
583                         info->options_len = sizeof(*md);
584
585                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
586
587                 } else {
588                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
589                 }
590
591                 return PACKET_RCVD;
592         }
593
594         return PACKET_REJECT;
595 }
596
597 static int gre_rcv(struct sk_buff *skb)
598 {
599         struct tnl_ptk_info tpi;
600         bool csum_err = false;
601         int hdr_len;
602
603         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
604         if (hdr_len < 0)
605                 goto drop;
606
607         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
608                 goto drop;
609
610         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
611                      tpi.proto == htons(ETH_P_ERSPAN2))) {
612                 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
613                         return 0;
614                 goto out;
615         }
616
617         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
618                 return 0;
619
620 out:
621         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
622 drop:
623         kfree_skb(skb);
624         return 0;
625 }
626
627 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
628 {
629         return iptunnel_handle_offloads(skb,
630                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
631 }
632
633 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
634                                      struct net_device *dev,
635                                      struct flowi6 *fl6, __u8 *dsfield,
636                                      int *encap_limit)
637 {
638         const struct iphdr *iph = ip_hdr(skb);
639         struct ip6_tnl *t = netdev_priv(dev);
640
641         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
642                 *encap_limit = t->parms.encap_limit;
643
644         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
645
646         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
647                 *dsfield = ipv4_get_dsfield(iph);
648         else
649                 *dsfield = ip6_tclass(t->parms.flowinfo);
650
651         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
652                 fl6->flowi6_mark = skb->mark;
653         else
654                 fl6->flowi6_mark = t->parms.fwmark;
655
656         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
657 }
658
659 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
660                                     struct net_device *dev,
661                                     struct flowi6 *fl6, __u8 *dsfield,
662                                     int *encap_limit)
663 {
664         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
665         struct ip6_tnl *t = netdev_priv(dev);
666         __u16 offset;
667
668         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
669         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
670
671         if (offset > 0) {
672                 struct ipv6_tlv_tnl_enc_lim *tel;
673
674                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
675                 if (tel->encap_limit == 0) {
676                         icmpv6_send(skb, ICMPV6_PARAMPROB,
677                                     ICMPV6_HDR_FIELD, offset + 2);
678                         return -1;
679                 }
680                 *encap_limit = tel->encap_limit - 1;
681         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
682                 *encap_limit = t->parms.encap_limit;
683         }
684
685         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
686
687         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
688                 *dsfield = ipv6_get_dsfield(ipv6h);
689         else
690                 *dsfield = ip6_tclass(t->parms.flowinfo);
691
692         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
693                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
694
695         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
696                 fl6->flowi6_mark = skb->mark;
697         else
698                 fl6->flowi6_mark = t->parms.fwmark;
699
700         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
701
702         return 0;
703 }
704
705 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
706                                struct net_device *dev, __u8 dsfield,
707                                struct flowi6 *fl6, int encap_limit,
708                                __u32 *pmtu, __be16 proto)
709 {
710         struct ip6_tnl *tunnel = netdev_priv(dev);
711         __be16 protocol;
712
713         if (dev->type == ARPHRD_ETHER)
714                 IPCB(skb)->flags = 0;
715
716         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
717                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
718         else
719                 fl6->daddr = tunnel->parms.raddr;
720
721         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
722                 return -ENOMEM;
723
724         /* Push GRE header. */
725         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
726
727         if (tunnel->parms.collect_md) {
728                 struct ip_tunnel_info *tun_info;
729                 const struct ip_tunnel_key *key;
730                 __be16 flags;
731
732                 tun_info = skb_tunnel_info(skb);
733                 if (unlikely(!tun_info ||
734                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
735                              ip_tunnel_info_af(tun_info) != AF_INET6))
736                         return -EINVAL;
737
738                 key = &tun_info->key;
739                 memset(fl6, 0, sizeof(*fl6));
740                 fl6->flowi6_proto = IPPROTO_GRE;
741                 fl6->daddr = key->u.ipv6.dst;
742                 fl6->flowlabel = key->label;
743                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
744
745                 dsfield = key->tos;
746                 flags = key->tun_flags &
747                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
748                 tunnel->tun_hlen = gre_calc_hlen(flags);
749
750                 gre_build_header(skb, tunnel->tun_hlen,
751                                  flags, protocol,
752                                  tunnel_id_to_key32(tun_info->key.tun_id),
753                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
754                                                       : 0);
755
756         } else {
757                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
758                         tunnel->o_seqno++;
759
760                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
761                                  protocol, tunnel->parms.o_key,
762                                  htonl(tunnel->o_seqno));
763         }
764
765         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
766                             NEXTHDR_GRE);
767 }
768
769 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
770 {
771         struct ip6_tnl *t = netdev_priv(dev);
772         int encap_limit = -1;
773         struct flowi6 fl6;
774         __u8 dsfield = 0;
775         __u32 mtu;
776         int err;
777
778         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
779
780         if (!t->parms.collect_md)
781                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
782                                          &dsfield, &encap_limit);
783
784         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
785         if (err)
786                 return -1;
787
788         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
789                           skb->protocol);
790         if (err != 0) {
791                 /* XXX: send ICMP error even if DF is not set. */
792                 if (err == -EMSGSIZE)
793                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
794                                   htonl(mtu));
795                 return -1;
796         }
797
798         return 0;
799 }
800
801 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
802 {
803         struct ip6_tnl *t = netdev_priv(dev);
804         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
805         int encap_limit = -1;
806         struct flowi6 fl6;
807         __u8 dsfield = 0;
808         __u32 mtu;
809         int err;
810
811         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
812                 return -1;
813
814         if (!t->parms.collect_md &&
815             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
816                 return -1;
817
818         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
819                 return -1;
820
821         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
822                           &mtu, skb->protocol);
823         if (err != 0) {
824                 if (err == -EMSGSIZE)
825                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
826                 return -1;
827         }
828
829         return 0;
830 }
831
832 /**
833  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
834  *   @t: the outgoing tunnel device
835  *   @hdr: IPv6 header from the incoming packet
836  *
837  * Description:
838  *   Avoid trivial tunneling loop by checking that tunnel exit-point
839  *   doesn't match source of incoming packet.
840  *
841  * Return:
842  *   1 if conflict,
843  *   0 else
844  **/
845
846 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
847         const struct ipv6hdr *hdr)
848 {
849         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
850 }
851
852 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
853 {
854         struct ip6_tnl *t = netdev_priv(dev);
855         int encap_limit = -1;
856         struct flowi6 fl6;
857         __u32 mtu;
858         int err;
859
860         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
861                 encap_limit = t->parms.encap_limit;
862
863         if (!t->parms.collect_md)
864                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
865
866         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
867         if (err)
868                 return err;
869
870         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
871
872         return err;
873 }
874
875 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
876         struct net_device *dev)
877 {
878         struct ip6_tnl *t = netdev_priv(dev);
879         struct net_device_stats *stats = &t->dev->stats;
880         int ret;
881
882         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
883                 goto tx_err;
884
885         switch (skb->protocol) {
886         case htons(ETH_P_IP):
887                 ret = ip6gre_xmit_ipv4(skb, dev);
888                 break;
889         case htons(ETH_P_IPV6):
890                 ret = ip6gre_xmit_ipv6(skb, dev);
891                 break;
892         default:
893                 ret = ip6gre_xmit_other(skb, dev);
894                 break;
895         }
896
897         if (ret < 0)
898                 goto tx_err;
899
900         return NETDEV_TX_OK;
901
902 tx_err:
903         stats->tx_errors++;
904         stats->tx_dropped++;
905         kfree_skb(skb);
906         return NETDEV_TX_OK;
907 }
908
909 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
910                                          struct net_device *dev)
911 {
912         struct ip6_tnl *t = netdev_priv(dev);
913         struct dst_entry *dst = skb_dst(skb);
914         struct net_device_stats *stats;
915         bool truncate = false;
916         int encap_limit = -1;
917         __u8 dsfield = false;
918         struct flowi6 fl6;
919         int err = -EINVAL;
920         __u32 mtu;
921         int nhoff;
922         int thoff;
923
924         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
925                 goto tx_err;
926
927         if (gre_handle_offloads(skb, false))
928                 goto tx_err;
929
930         if (skb->len > dev->mtu + dev->hard_header_len) {
931                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
932                 truncate = true;
933         }
934
935         nhoff = skb_network_header(skb) - skb_mac_header(skb);
936         if (skb->protocol == htons(ETH_P_IP) &&
937             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
938                 truncate = true;
939
940         thoff = skb_transport_header(skb) - skb_mac_header(skb);
941         if (skb->protocol == htons(ETH_P_IPV6) &&
942             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
943                 truncate = true;
944
945         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
946                 goto tx_err;
947
948         t->parms.o_flags &= ~TUNNEL_KEY;
949         IPCB(skb)->flags = 0;
950
951         /* For collect_md mode, derive fl6 from the tunnel key,
952          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
953          */
954         if (t->parms.collect_md) {
955                 struct ip_tunnel_info *tun_info;
956                 const struct ip_tunnel_key *key;
957                 struct erspan_metadata *md;
958                 __be32 tun_id;
959
960                 tun_info = skb_tunnel_info(skb);
961                 if (unlikely(!tun_info ||
962                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
963                              ip_tunnel_info_af(tun_info) != AF_INET6))
964                         return -EINVAL;
965
966                 key = &tun_info->key;
967                 memset(&fl6, 0, sizeof(fl6));
968                 fl6.flowi6_proto = IPPROTO_GRE;
969                 fl6.daddr = key->u.ipv6.dst;
970                 fl6.flowlabel = key->label;
971                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
972
973                 dsfield = key->tos;
974                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
975                         goto tx_err;
976                 md = ip_tunnel_info_opts(tun_info);
977                 if (!md)
978                         goto tx_err;
979
980                 tun_id = tunnel_id_to_key32(key->tun_id);
981                 if (md->version == 1) {
982                         erspan_build_header(skb,
983                                             ntohl(tun_id),
984                                             ntohl(md->u.index), truncate,
985                                             false);
986                 } else if (md->version == 2) {
987                         erspan_build_header_v2(skb,
988                                                ntohl(tun_id),
989                                                md->u.md2.dir,
990                                                get_hwid(&md->u.md2),
991                                                truncate, false);
992                 } else {
993                         goto tx_err;
994                 }
995         } else {
996                 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
997
998                 switch (skb->protocol) {
999                 case htons(ETH_P_IP):
1000                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1001                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1002                                                  &dsfield, &encap_limit);
1003                         break;
1004                 case htons(ETH_P_IPV6):
1005                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
1006                                 goto tx_err;
1007                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1008                                                      &dsfield, &encap_limit))
1009                                 goto tx_err;
1010                         break;
1011                 default:
1012                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1013                         break;
1014                 }
1015
1016                 if (t->parms.erspan_ver == 1)
1017                         erspan_build_header(skb, ntohl(t->parms.o_key),
1018                                             t->parms.index,
1019                                             truncate, false);
1020                 else if (t->parms.erspan_ver == 2)
1021                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1022                                                t->parms.dir,
1023                                                t->parms.hwid,
1024                                                truncate, false);
1025                 else
1026                         goto tx_err;
1027
1028                 fl6.daddr = t->parms.raddr;
1029         }
1030
1031         /* Push GRE header. */
1032         gre_build_header(skb, 8, TUNNEL_SEQ,
1033                          htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1034
1035         /* TooBig packet may have updated dst->dev's mtu */
1036         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1037                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1038
1039         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1040                            NEXTHDR_GRE);
1041         if (err != 0) {
1042                 /* XXX: send ICMP error even if DF is not set. */
1043                 if (err == -EMSGSIZE) {
1044                         if (skb->protocol == htons(ETH_P_IP))
1045                                 icmp_send(skb, ICMP_DEST_UNREACH,
1046                                           ICMP_FRAG_NEEDED, htonl(mtu));
1047                         else
1048                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1049                 }
1050
1051                 goto tx_err;
1052         }
1053         return NETDEV_TX_OK;
1054
1055 tx_err:
1056         stats = &t->dev->stats;
1057         stats->tx_errors++;
1058         stats->tx_dropped++;
1059         kfree_skb(skb);
1060         return NETDEV_TX_OK;
1061 }
1062
1063 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1064 {
1065         struct net_device *dev = t->dev;
1066         struct __ip6_tnl_parm *p = &t->parms;
1067         struct flowi6 *fl6 = &t->fl.u.ip6;
1068
1069         if (dev->type != ARPHRD_ETHER) {
1070                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1071                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1072         }
1073
1074         /* Set up flowi template */
1075         fl6->saddr = p->laddr;
1076         fl6->daddr = p->raddr;
1077         fl6->flowi6_oif = p->link;
1078         fl6->flowlabel = 0;
1079         fl6->flowi6_proto = IPPROTO_GRE;
1080
1081         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1082                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1083         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1084                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1085
1086         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1087         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1088
1089         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1090                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1091                 dev->flags |= IFF_POINTOPOINT;
1092         else
1093                 dev->flags &= ~IFF_POINTOPOINT;
1094 }
1095
1096 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1097                                          int t_hlen)
1098 {
1099         const struct __ip6_tnl_parm *p = &t->parms;
1100         struct net_device *dev = t->dev;
1101
1102         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1103                 int strict = (ipv6_addr_type(&p->raddr) &
1104                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1105
1106                 struct rt6_info *rt = rt6_lookup(t->net,
1107                                                  &p->raddr, &p->laddr,
1108                                                  p->link, NULL, strict);
1109
1110                 if (!rt)
1111                         return;
1112
1113                 if (rt->dst.dev) {
1114                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1115                                                t_hlen;
1116
1117                         if (set_mtu) {
1118                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1119                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1120                                         dev->mtu -= 8;
1121                                 if (dev->type == ARPHRD_ETHER)
1122                                         dev->mtu -= ETH_HLEN;
1123
1124                                 if (dev->mtu < IPV6_MIN_MTU)
1125                                         dev->mtu = IPV6_MIN_MTU;
1126                         }
1127                 }
1128                 ip6_rt_put(rt);
1129         }
1130 }
1131
1132 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1133 {
1134         int t_hlen;
1135
1136         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1137         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1138
1139         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1140         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1141         return t_hlen;
1142 }
1143
1144 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1145 {
1146         ip6gre_tnl_link_config_common(t);
1147         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1148 }
1149
1150 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1151                                      const struct __ip6_tnl_parm *p)
1152 {
1153         t->parms.laddr = p->laddr;
1154         t->parms.raddr = p->raddr;
1155         t->parms.flags = p->flags;
1156         t->parms.hop_limit = p->hop_limit;
1157         t->parms.encap_limit = p->encap_limit;
1158         t->parms.flowinfo = p->flowinfo;
1159         t->parms.link = p->link;
1160         t->parms.proto = p->proto;
1161         t->parms.i_key = p->i_key;
1162         t->parms.o_key = p->o_key;
1163         t->parms.i_flags = p->i_flags;
1164         t->parms.o_flags = p->o_flags;
1165         t->parms.fwmark = p->fwmark;
1166         dst_cache_reset(&t->dst_cache);
1167 }
1168
1169 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1170                              int set_mtu)
1171 {
1172         ip6gre_tnl_copy_tnl_parm(t, p);
1173         ip6gre_tnl_link_config(t, set_mtu);
1174         return 0;
1175 }
1176
1177 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1178         const struct ip6_tnl_parm2 *u)
1179 {
1180         p->laddr = u->laddr;
1181         p->raddr = u->raddr;
1182         p->flags = u->flags;
1183         p->hop_limit = u->hop_limit;
1184         p->encap_limit = u->encap_limit;
1185         p->flowinfo = u->flowinfo;
1186         p->link = u->link;
1187         p->i_key = u->i_key;
1188         p->o_key = u->o_key;
1189         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1190         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1191         memcpy(p->name, u->name, sizeof(u->name));
1192 }
1193
1194 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1195         const struct __ip6_tnl_parm *p)
1196 {
1197         u->proto = IPPROTO_GRE;
1198         u->laddr = p->laddr;
1199         u->raddr = p->raddr;
1200         u->flags = p->flags;
1201         u->hop_limit = p->hop_limit;
1202         u->encap_limit = p->encap_limit;
1203         u->flowinfo = p->flowinfo;
1204         u->link = p->link;
1205         u->i_key = p->i_key;
1206         u->o_key = p->o_key;
1207         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1208         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1209         memcpy(u->name, p->name, sizeof(u->name));
1210 }
1211
1212 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1213         struct ifreq *ifr, int cmd)
1214 {
1215         int err = 0;
1216         struct ip6_tnl_parm2 p;
1217         struct __ip6_tnl_parm p1;
1218         struct ip6_tnl *t = netdev_priv(dev);
1219         struct net *net = t->net;
1220         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1221
1222         memset(&p1, 0, sizeof(p1));
1223
1224         switch (cmd) {
1225         case SIOCGETTUNNEL:
1226                 if (dev == ign->fb_tunnel_dev) {
1227                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1228                                 err = -EFAULT;
1229                                 break;
1230                         }
1231                         ip6gre_tnl_parm_from_user(&p1, &p);
1232                         t = ip6gre_tunnel_locate(net, &p1, 0);
1233                         if (!t)
1234                                 t = netdev_priv(dev);
1235                 }
1236                 memset(&p, 0, sizeof(p));
1237                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1238                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1239                         err = -EFAULT;
1240                 break;
1241
1242         case SIOCADDTUNNEL:
1243         case SIOCCHGTUNNEL:
1244                 err = -EPERM;
1245                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1246                         goto done;
1247
1248                 err = -EFAULT;
1249                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1250                         goto done;
1251
1252                 err = -EINVAL;
1253                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1254                         goto done;
1255
1256                 if (!(p.i_flags&GRE_KEY))
1257                         p.i_key = 0;
1258                 if (!(p.o_flags&GRE_KEY))
1259                         p.o_key = 0;
1260
1261                 ip6gre_tnl_parm_from_user(&p1, &p);
1262                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1263
1264                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1265                         if (t) {
1266                                 if (t->dev != dev) {
1267                                         err = -EEXIST;
1268                                         break;
1269                                 }
1270                         } else {
1271                                 t = netdev_priv(dev);
1272
1273                                 ip6gre_tunnel_unlink(ign, t);
1274                                 synchronize_net();
1275                                 ip6gre_tnl_change(t, &p1, 1);
1276                                 ip6gre_tunnel_link(ign, t);
1277                                 netdev_state_change(dev);
1278                         }
1279                 }
1280
1281                 if (t) {
1282                         err = 0;
1283
1284                         memset(&p, 0, sizeof(p));
1285                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1286                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1287                                 err = -EFAULT;
1288                 } else
1289                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1290                 break;
1291
1292         case SIOCDELTUNNEL:
1293                 err = -EPERM;
1294                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1295                         goto done;
1296
1297                 if (dev == ign->fb_tunnel_dev) {
1298                         err = -EFAULT;
1299                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1300                                 goto done;
1301                         err = -ENOENT;
1302                         ip6gre_tnl_parm_from_user(&p1, &p);
1303                         t = ip6gre_tunnel_locate(net, &p1, 0);
1304                         if (!t)
1305                                 goto done;
1306                         err = -EPERM;
1307                         if (t == netdev_priv(ign->fb_tunnel_dev))
1308                                 goto done;
1309                         dev = t->dev;
1310                 }
1311                 unregister_netdevice(dev);
1312                 err = 0;
1313                 break;
1314
1315         default:
1316                 err = -EINVAL;
1317         }
1318
1319 done:
1320         return err;
1321 }
1322
1323 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1324                          unsigned short type, const void *daddr,
1325                          const void *saddr, unsigned int len)
1326 {
1327         struct ip6_tnl *t = netdev_priv(dev);
1328         struct ipv6hdr *ipv6h;
1329         __be16 *p;
1330
1331         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1332         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1333                                                   t->fl.u.ip6.flowlabel,
1334                                                   true, &t->fl.u.ip6));
1335         ipv6h->hop_limit = t->parms.hop_limit;
1336         ipv6h->nexthdr = NEXTHDR_GRE;
1337         ipv6h->saddr = t->parms.laddr;
1338         ipv6h->daddr = t->parms.raddr;
1339
1340         p = (__be16 *)(ipv6h + 1);
1341         p[0] = t->parms.o_flags;
1342         p[1] = htons(type);
1343
1344         /*
1345          *      Set the source hardware address.
1346          */
1347
1348         if (saddr)
1349                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1350         if (daddr)
1351                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1352         if (!ipv6_addr_any(&ipv6h->daddr))
1353                 return t->hlen;
1354
1355         return -t->hlen;
1356 }
1357
1358 static const struct header_ops ip6gre_header_ops = {
1359         .create = ip6gre_header,
1360 };
1361
1362 static const struct net_device_ops ip6gre_netdev_ops = {
1363         .ndo_init               = ip6gre_tunnel_init,
1364         .ndo_uninit             = ip6gre_tunnel_uninit,
1365         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1366         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1367         .ndo_change_mtu         = ip6_tnl_change_mtu,
1368         .ndo_get_stats64        = ip_tunnel_get_stats64,
1369         .ndo_get_iflink         = ip6_tnl_get_iflink,
1370 };
1371
1372 static void ip6gre_dev_free(struct net_device *dev)
1373 {
1374         struct ip6_tnl *t = netdev_priv(dev);
1375
1376         gro_cells_destroy(&t->gro_cells);
1377         dst_cache_destroy(&t->dst_cache);
1378         free_percpu(dev->tstats);
1379 }
1380
1381 static void ip6gre_tunnel_setup(struct net_device *dev)
1382 {
1383         dev->netdev_ops = &ip6gre_netdev_ops;
1384         dev->needs_free_netdev = true;
1385         dev->priv_destructor = ip6gre_dev_free;
1386
1387         dev->type = ARPHRD_IP6GRE;
1388
1389         dev->flags |= IFF_NOARP;
1390         dev->addr_len = sizeof(struct in6_addr);
1391         netif_keep_dst(dev);
1392         /* This perm addr will be used as interface identifier by IPv6 */
1393         dev->addr_assign_type = NET_ADDR_RANDOM;
1394         eth_random_addr(dev->perm_addr);
1395 }
1396
1397 #define GRE6_FEATURES (NETIF_F_SG |             \
1398                        NETIF_F_FRAGLIST |       \
1399                        NETIF_F_HIGHDMA |        \
1400                        NETIF_F_HW_CSUM)
1401
1402 static void ip6gre_tnl_init_features(struct net_device *dev)
1403 {
1404         struct ip6_tnl *nt = netdev_priv(dev);
1405
1406         dev->features           |= GRE6_FEATURES;
1407         dev->hw_features        |= GRE6_FEATURES;
1408
1409         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1410                 /* TCP offload with GRE SEQ is not supported, nor
1411                  * can we support 2 levels of outer headers requiring
1412                  * an update.
1413                  */
1414                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1415                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1416                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1417                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1418                 }
1419
1420                 /* Can use a lockless transmit, unless we generate
1421                  * output sequences
1422                  */
1423                 dev->features |= NETIF_F_LLTX;
1424         }
1425 }
1426
1427 static int ip6gre_tunnel_init_common(struct net_device *dev)
1428 {
1429         struct ip6_tnl *tunnel;
1430         int ret;
1431         int t_hlen;
1432
1433         tunnel = netdev_priv(dev);
1434
1435         tunnel->dev = dev;
1436         tunnel->net = dev_net(dev);
1437         strcpy(tunnel->parms.name, dev->name);
1438
1439         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1440         if (!dev->tstats)
1441                 return -ENOMEM;
1442
1443         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1444         if (ret)
1445                 goto cleanup_alloc_pcpu_stats;
1446
1447         ret = gro_cells_init(&tunnel->gro_cells, dev);
1448         if (ret)
1449                 goto cleanup_dst_cache_init;
1450
1451         t_hlen = ip6gre_calc_hlen(tunnel);
1452         dev->mtu = ETH_DATA_LEN - t_hlen;
1453         if (dev->type == ARPHRD_ETHER)
1454                 dev->mtu -= ETH_HLEN;
1455         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1456                 dev->mtu -= 8;
1457
1458         if (tunnel->parms.collect_md) {
1459                 dev->features |= NETIF_F_NETNS_LOCAL;
1460                 netif_keep_dst(dev);
1461         }
1462         ip6gre_tnl_init_features(dev);
1463
1464         return 0;
1465
1466 cleanup_dst_cache_init:
1467         dst_cache_destroy(&tunnel->dst_cache);
1468 cleanup_alloc_pcpu_stats:
1469         free_percpu(dev->tstats);
1470         dev->tstats = NULL;
1471         return ret;
1472 }
1473
1474 static int ip6gre_tunnel_init(struct net_device *dev)
1475 {
1476         struct ip6_tnl *tunnel;
1477         int ret;
1478
1479         ret = ip6gre_tunnel_init_common(dev);
1480         if (ret)
1481                 return ret;
1482
1483         tunnel = netdev_priv(dev);
1484
1485         if (tunnel->parms.collect_md)
1486                 return 0;
1487
1488         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1489         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1490
1491         if (ipv6_addr_any(&tunnel->parms.raddr))
1492                 dev->header_ops = &ip6gre_header_ops;
1493
1494         return 0;
1495 }
1496
1497 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1498 {
1499         struct ip6_tnl *tunnel = netdev_priv(dev);
1500
1501         tunnel->dev = dev;
1502         tunnel->net = dev_net(dev);
1503         strcpy(tunnel->parms.name, dev->name);
1504
1505         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1506
1507         dev_hold(dev);
1508 }
1509
1510 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1511         .handler     = gre_rcv,
1512         .err_handler = ip6gre_err,
1513         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1514 };
1515
1516 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1517 {
1518         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1519         struct net_device *dev, *aux;
1520         int prio;
1521
1522         for_each_netdev_safe(net, dev, aux)
1523                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1524                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1525                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1526                         unregister_netdevice_queue(dev, head);
1527
1528         for (prio = 0; prio < 4; prio++) {
1529                 int h;
1530                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1531                         struct ip6_tnl *t;
1532
1533                         t = rtnl_dereference(ign->tunnels[prio][h]);
1534
1535                         while (t) {
1536                                 /* If dev is in the same netns, it has already
1537                                  * been added to the list by the previous loop.
1538                                  */
1539                                 if (!net_eq(dev_net(t->dev), net))
1540                                         unregister_netdevice_queue(t->dev,
1541                                                                    head);
1542                                 t = rtnl_dereference(t->next);
1543                         }
1544                 }
1545         }
1546 }
1547
1548 static int __net_init ip6gre_init_net(struct net *net)
1549 {
1550         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1551         int err;
1552
1553         if (!net_has_fallback_tunnels(net))
1554                 return 0;
1555         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1556                                           NET_NAME_UNKNOWN,
1557                                           ip6gre_tunnel_setup);
1558         if (!ign->fb_tunnel_dev) {
1559                 err = -ENOMEM;
1560                 goto err_alloc_dev;
1561         }
1562         dev_net_set(ign->fb_tunnel_dev, net);
1563         /* FB netdevice is special: we have one, and only one per netns.
1564          * Allowing to move it to another netns is clearly unsafe.
1565          */
1566         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1567
1568
1569         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1570         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1571
1572         err = register_netdev(ign->fb_tunnel_dev);
1573         if (err)
1574                 goto err_reg_dev;
1575
1576         rcu_assign_pointer(ign->tunnels_wc[0],
1577                            netdev_priv(ign->fb_tunnel_dev));
1578         return 0;
1579
1580 err_reg_dev:
1581         free_netdev(ign->fb_tunnel_dev);
1582 err_alloc_dev:
1583         return err;
1584 }
1585
1586 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1587 {
1588         struct net *net;
1589         LIST_HEAD(list);
1590
1591         rtnl_lock();
1592         list_for_each_entry(net, net_list, exit_list)
1593                 ip6gre_destroy_tunnels(net, &list);
1594         unregister_netdevice_many(&list);
1595         rtnl_unlock();
1596 }
1597
1598 static struct pernet_operations ip6gre_net_ops = {
1599         .init = ip6gre_init_net,
1600         .exit_batch = ip6gre_exit_batch_net,
1601         .id   = &ip6gre_net_id,
1602         .size = sizeof(struct ip6gre_net),
1603 };
1604
1605 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1606                                   struct netlink_ext_ack *extack)
1607 {
1608         __be16 flags;
1609
1610         if (!data)
1611                 return 0;
1612
1613         flags = 0;
1614         if (data[IFLA_GRE_IFLAGS])
1615                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1616         if (data[IFLA_GRE_OFLAGS])
1617                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1618         if (flags & (GRE_VERSION|GRE_ROUTING))
1619                 return -EINVAL;
1620
1621         return 0;
1622 }
1623
1624 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1625                                struct netlink_ext_ack *extack)
1626 {
1627         struct in6_addr daddr;
1628
1629         if (tb[IFLA_ADDRESS]) {
1630                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1631                         return -EINVAL;
1632                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1633                         return -EADDRNOTAVAIL;
1634         }
1635
1636         if (!data)
1637                 goto out;
1638
1639         if (data[IFLA_GRE_REMOTE]) {
1640                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1641                 if (ipv6_addr_any(&daddr))
1642                         return -EINVAL;
1643         }
1644
1645 out:
1646         return ip6gre_tunnel_validate(tb, data, extack);
1647 }
1648
1649 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1650                                   struct netlink_ext_ack *extack)
1651 {
1652         __be16 flags = 0;
1653         int ret, ver = 0;
1654
1655         if (!data)
1656                 return 0;
1657
1658         ret = ip6gre_tap_validate(tb, data, extack);
1659         if (ret)
1660                 return ret;
1661
1662         /* ERSPAN should only have GRE sequence and key flag */
1663         if (data[IFLA_GRE_OFLAGS])
1664                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1665         if (data[IFLA_GRE_IFLAGS])
1666                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1667         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1668             flags != (GRE_SEQ | GRE_KEY))
1669                 return -EINVAL;
1670
1671         /* ERSPAN Session ID only has 10-bit. Since we reuse
1672          * 32-bit key field as ID, check it's range.
1673          */
1674         if (data[IFLA_GRE_IKEY] &&
1675             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1676                 return -EINVAL;
1677
1678         if (data[IFLA_GRE_OKEY] &&
1679             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1680                 return -EINVAL;
1681
1682         if (data[IFLA_GRE_ERSPAN_VER]) {
1683                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1684                 if (ver != 1 && ver != 2)
1685                         return -EINVAL;
1686         }
1687
1688         if (ver == 1) {
1689                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1690                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1691
1692                         if (index & ~INDEX_MASK)
1693                                 return -EINVAL;
1694                 }
1695         } else if (ver == 2) {
1696                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1697                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1698
1699                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1700                                 return -EINVAL;
1701                 }
1702
1703                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1704                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1705
1706                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1707                                 return -EINVAL;
1708                 }
1709         }
1710
1711         return 0;
1712 }
1713
1714 static void ip6gre_netlink_parms(struct nlattr *data[],
1715                                 struct __ip6_tnl_parm *parms)
1716 {
1717         memset(parms, 0, sizeof(*parms));
1718
1719         if (!data)
1720                 return;
1721
1722         if (data[IFLA_GRE_LINK])
1723                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1724
1725         if (data[IFLA_GRE_IFLAGS])
1726                 parms->i_flags = gre_flags_to_tnl_flags(
1727                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1728
1729         if (data[IFLA_GRE_OFLAGS])
1730                 parms->o_flags = gre_flags_to_tnl_flags(
1731                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1732
1733         if (data[IFLA_GRE_IKEY])
1734                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1735
1736         if (data[IFLA_GRE_OKEY])
1737                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1738
1739         if (data[IFLA_GRE_LOCAL])
1740                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1741
1742         if (data[IFLA_GRE_REMOTE])
1743                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1744
1745         if (data[IFLA_GRE_TTL])
1746                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1747
1748         if (data[IFLA_GRE_ENCAP_LIMIT])
1749                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1750
1751         if (data[IFLA_GRE_FLOWINFO])
1752                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1753
1754         if (data[IFLA_GRE_FLAGS])
1755                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1756
1757         if (data[IFLA_GRE_FWMARK])
1758                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1759
1760         if (data[IFLA_GRE_COLLECT_METADATA])
1761                 parms->collect_md = true;
1762
1763         parms->erspan_ver = 1;
1764         if (data[IFLA_GRE_ERSPAN_VER])
1765                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1766
1767         if (parms->erspan_ver == 1) {
1768                 if (data[IFLA_GRE_ERSPAN_INDEX])
1769                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1770         } else if (parms->erspan_ver == 2) {
1771                 if (data[IFLA_GRE_ERSPAN_DIR])
1772                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1773                 if (data[IFLA_GRE_ERSPAN_HWID])
1774                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1775         }
1776 }
1777
1778 static int ip6gre_tap_init(struct net_device *dev)
1779 {
1780         int ret;
1781
1782         ret = ip6gre_tunnel_init_common(dev);
1783         if (ret)
1784                 return ret;
1785
1786         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1787
1788         return 0;
1789 }
1790
1791 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1792         .ndo_init = ip6gre_tap_init,
1793         .ndo_uninit = ip6gre_tunnel_uninit,
1794         .ndo_start_xmit = ip6gre_tunnel_xmit,
1795         .ndo_set_mac_address = eth_mac_addr,
1796         .ndo_validate_addr = eth_validate_addr,
1797         .ndo_change_mtu = ip6_tnl_change_mtu,
1798         .ndo_get_stats64 = ip_tunnel_get_stats64,
1799         .ndo_get_iflink = ip6_tnl_get_iflink,
1800 };
1801
1802 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1803 {
1804         int t_hlen;
1805
1806         tunnel->tun_hlen = 8;
1807         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1808                        erspan_hdr_len(tunnel->parms.erspan_ver);
1809
1810         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1811         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1812         return t_hlen;
1813 }
1814
1815 static int ip6erspan_tap_init(struct net_device *dev)
1816 {
1817         struct ip6_tnl *tunnel;
1818         int t_hlen;
1819         int ret;
1820
1821         tunnel = netdev_priv(dev);
1822
1823         tunnel->dev = dev;
1824         tunnel->net = dev_net(dev);
1825         strcpy(tunnel->parms.name, dev->name);
1826
1827         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1828         if (!dev->tstats)
1829                 return -ENOMEM;
1830
1831         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1832         if (ret)
1833                 goto cleanup_alloc_pcpu_stats;
1834
1835         ret = gro_cells_init(&tunnel->gro_cells, dev);
1836         if (ret)
1837                 goto cleanup_dst_cache_init;
1838
1839         t_hlen = ip6erspan_calc_hlen(tunnel);
1840         dev->mtu = ETH_DATA_LEN - t_hlen;
1841         if (dev->type == ARPHRD_ETHER)
1842                 dev->mtu -= ETH_HLEN;
1843         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1844                 dev->mtu -= 8;
1845
1846         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1847         ip6erspan_tnl_link_config(tunnel, 1);
1848
1849         return 0;
1850
1851 cleanup_dst_cache_init:
1852         dst_cache_destroy(&tunnel->dst_cache);
1853 cleanup_alloc_pcpu_stats:
1854         free_percpu(dev->tstats);
1855         dev->tstats = NULL;
1856         return ret;
1857 }
1858
1859 static const struct net_device_ops ip6erspan_netdev_ops = {
1860         .ndo_init =             ip6erspan_tap_init,
1861         .ndo_uninit =           ip6erspan_tunnel_uninit,
1862         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1863         .ndo_set_mac_address =  eth_mac_addr,
1864         .ndo_validate_addr =    eth_validate_addr,
1865         .ndo_change_mtu =       ip6_tnl_change_mtu,
1866         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1867         .ndo_get_iflink =       ip6_tnl_get_iflink,
1868 };
1869
1870 static void ip6gre_tap_setup(struct net_device *dev)
1871 {
1872
1873         ether_setup(dev);
1874
1875         dev->max_mtu = 0;
1876         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1877         dev->needs_free_netdev = true;
1878         dev->priv_destructor = ip6gre_dev_free;
1879
1880         dev->features |= NETIF_F_NETNS_LOCAL;
1881         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1882         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1883         netif_keep_dst(dev);
1884 }
1885
1886 bool is_ip6gretap_dev(const struct net_device *dev)
1887 {
1888         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1889 }
1890 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1891
1892 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1893                                        struct ip_tunnel_encap *ipencap)
1894 {
1895         bool ret = false;
1896
1897         memset(ipencap, 0, sizeof(*ipencap));
1898
1899         if (!data)
1900                 return ret;
1901
1902         if (data[IFLA_GRE_ENCAP_TYPE]) {
1903                 ret = true;
1904                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1905         }
1906
1907         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1908                 ret = true;
1909                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1910         }
1911
1912         if (data[IFLA_GRE_ENCAP_SPORT]) {
1913                 ret = true;
1914                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1915         }
1916
1917         if (data[IFLA_GRE_ENCAP_DPORT]) {
1918                 ret = true;
1919                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1920         }
1921
1922         return ret;
1923 }
1924
1925 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1926                                  struct nlattr *tb[], struct nlattr *data[],
1927                                  struct netlink_ext_ack *extack)
1928 {
1929         struct ip6_tnl *nt;
1930         struct ip_tunnel_encap ipencap;
1931         int err;
1932
1933         nt = netdev_priv(dev);
1934
1935         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1936                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1937
1938                 if (err < 0)
1939                         return err;
1940         }
1941
1942         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1943                 eth_hw_addr_random(dev);
1944
1945         nt->dev = dev;
1946         nt->net = dev_net(dev);
1947
1948         err = register_netdevice(dev);
1949         if (err)
1950                 goto out;
1951
1952         if (tb[IFLA_MTU])
1953                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1954
1955         dev_hold(dev);
1956
1957 out:
1958         return err;
1959 }
1960
1961 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1962                           struct nlattr *tb[], struct nlattr *data[],
1963                           struct netlink_ext_ack *extack)
1964 {
1965         struct ip6_tnl *nt = netdev_priv(dev);
1966         struct net *net = dev_net(dev);
1967         struct ip6gre_net *ign;
1968         int err;
1969
1970         ip6gre_netlink_parms(data, &nt->parms);
1971         ign = net_generic(net, ip6gre_net_id);
1972
1973         if (nt->parms.collect_md) {
1974                 if (rtnl_dereference(ign->collect_md_tun))
1975                         return -EEXIST;
1976         } else {
1977                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1978                         return -EEXIST;
1979         }
1980
1981         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1982         if (!err) {
1983                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1984                 ip6gre_tunnel_link_md(ign, nt);
1985                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1986         }
1987         return err;
1988 }
1989
1990 static struct ip6_tnl *
1991 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
1992                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
1993                          struct netlink_ext_ack *extack)
1994 {
1995         struct ip6_tnl *t, *nt = netdev_priv(dev);
1996         struct net *net = nt->net;
1997         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1998         struct ip_tunnel_encap ipencap;
1999
2000         if (dev == ign->fb_tunnel_dev)
2001                 return ERR_PTR(-EINVAL);
2002
2003         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2004                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2005
2006                 if (err < 0)
2007                         return ERR_PTR(err);
2008         }
2009
2010         ip6gre_netlink_parms(data, p_p);
2011
2012         t = ip6gre_tunnel_locate(net, p_p, 0);
2013
2014         if (t) {
2015                 if (t->dev != dev)
2016                         return ERR_PTR(-EEXIST);
2017         } else {
2018                 t = nt;
2019         }
2020
2021         return t;
2022 }
2023
2024 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2025                              struct nlattr *data[],
2026                              struct netlink_ext_ack *extack)
2027 {
2028         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2029         struct __ip6_tnl_parm p;
2030         struct ip6_tnl *t;
2031
2032         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2033         if (IS_ERR(t))
2034                 return PTR_ERR(t);
2035
2036         ip6gre_tunnel_unlink_md(ign, t);
2037         ip6gre_tunnel_unlink(ign, t);
2038         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2039         ip6gre_tunnel_link_md(ign, t);
2040         ip6gre_tunnel_link(ign, t);
2041         return 0;
2042 }
2043
2044 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2045 {
2046         struct net *net = dev_net(dev);
2047         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2048
2049         if (dev != ign->fb_tunnel_dev)
2050                 unregister_netdevice_queue(dev, head);
2051 }
2052
2053 static size_t ip6gre_get_size(const struct net_device *dev)
2054 {
2055         return
2056                 /* IFLA_GRE_LINK */
2057                 nla_total_size(4) +
2058                 /* IFLA_GRE_IFLAGS */
2059                 nla_total_size(2) +
2060                 /* IFLA_GRE_OFLAGS */
2061                 nla_total_size(2) +
2062                 /* IFLA_GRE_IKEY */
2063                 nla_total_size(4) +
2064                 /* IFLA_GRE_OKEY */
2065                 nla_total_size(4) +
2066                 /* IFLA_GRE_LOCAL */
2067                 nla_total_size(sizeof(struct in6_addr)) +
2068                 /* IFLA_GRE_REMOTE */
2069                 nla_total_size(sizeof(struct in6_addr)) +
2070                 /* IFLA_GRE_TTL */
2071                 nla_total_size(1) +
2072                 /* IFLA_GRE_ENCAP_LIMIT */
2073                 nla_total_size(1) +
2074                 /* IFLA_GRE_FLOWINFO */
2075                 nla_total_size(4) +
2076                 /* IFLA_GRE_FLAGS */
2077                 nla_total_size(4) +
2078                 /* IFLA_GRE_ENCAP_TYPE */
2079                 nla_total_size(2) +
2080                 /* IFLA_GRE_ENCAP_FLAGS */
2081                 nla_total_size(2) +
2082                 /* IFLA_GRE_ENCAP_SPORT */
2083                 nla_total_size(2) +
2084                 /* IFLA_GRE_ENCAP_DPORT */
2085                 nla_total_size(2) +
2086                 /* IFLA_GRE_COLLECT_METADATA */
2087                 nla_total_size(0) +
2088                 /* IFLA_GRE_FWMARK */
2089                 nla_total_size(4) +
2090                 /* IFLA_GRE_ERSPAN_INDEX */
2091                 nla_total_size(4) +
2092                 0;
2093 }
2094
2095 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2096 {
2097         struct ip6_tnl *t = netdev_priv(dev);
2098         struct __ip6_tnl_parm *p = &t->parms;
2099
2100         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2101             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2102                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2103             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2104                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2105             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2106             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2107             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2108             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2109             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2110             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2111             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2112             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2113             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2114             nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2115                 goto nla_put_failure;
2116
2117         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2118                         t->encap.type) ||
2119             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2120                          t->encap.sport) ||
2121             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2122                          t->encap.dport) ||
2123             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2124                         t->encap.flags))
2125                 goto nla_put_failure;
2126
2127         if (p->collect_md) {
2128                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2129                         goto nla_put_failure;
2130         }
2131
2132         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2133                 goto nla_put_failure;
2134
2135         if (p->erspan_ver == 1) {
2136                 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2137                         goto nla_put_failure;
2138         } else if (p->erspan_ver == 2) {
2139                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2140                         goto nla_put_failure;
2141                 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2142                         goto nla_put_failure;
2143         }
2144
2145         return 0;
2146
2147 nla_put_failure:
2148         return -EMSGSIZE;
2149 }
2150
2151 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2152         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2153         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2154         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2155         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2156         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2157         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2158         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2159         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2160         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2161         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2162         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2163         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2164         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2165         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2166         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2167         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2168         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2169         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2170         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2171         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2172         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2173 };
2174
2175 static void ip6erspan_tap_setup(struct net_device *dev)
2176 {
2177         ether_setup(dev);
2178
2179         dev->netdev_ops = &ip6erspan_netdev_ops;
2180         dev->needs_free_netdev = true;
2181         dev->priv_destructor = ip6gre_dev_free;
2182
2183         dev->features |= NETIF_F_NETNS_LOCAL;
2184         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2185         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2186         netif_keep_dst(dev);
2187 }
2188
2189 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2190                              struct nlattr *tb[], struct nlattr *data[],
2191                              struct netlink_ext_ack *extack)
2192 {
2193         struct ip6_tnl *nt = netdev_priv(dev);
2194         struct net *net = dev_net(dev);
2195         struct ip6gre_net *ign;
2196         int err;
2197
2198         ip6gre_netlink_parms(data, &nt->parms);
2199         ign = net_generic(net, ip6gre_net_id);
2200
2201         if (nt->parms.collect_md) {
2202                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2203                         return -EEXIST;
2204         } else {
2205                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2206                         return -EEXIST;
2207         }
2208
2209         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2210         if (!err) {
2211                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2212                 ip6erspan_tunnel_link_md(ign, nt);
2213                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2214         }
2215         return err;
2216 }
2217
2218 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2219 {
2220         ip6gre_tnl_link_config_common(t);
2221         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2222 }
2223
2224 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2225                                 const struct __ip6_tnl_parm *p, int set_mtu)
2226 {
2227         ip6gre_tnl_copy_tnl_parm(t, p);
2228         ip6erspan_tnl_link_config(t, set_mtu);
2229         return 0;
2230 }
2231
2232 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2233                                 struct nlattr *data[],
2234                                 struct netlink_ext_ack *extack)
2235 {
2236         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2237         struct __ip6_tnl_parm p;
2238         struct ip6_tnl *t;
2239
2240         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2241         if (IS_ERR(t))
2242                 return PTR_ERR(t);
2243
2244         ip6gre_tunnel_unlink_md(ign, t);
2245         ip6gre_tunnel_unlink(ign, t);
2246         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2247         ip6erspan_tunnel_link_md(ign, t);
2248         ip6gre_tunnel_link(ign, t);
2249         return 0;
2250 }
2251
2252 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2253         .kind           = "ip6gre",
2254         .maxtype        = IFLA_GRE_MAX,
2255         .policy         = ip6gre_policy,
2256         .priv_size      = sizeof(struct ip6_tnl),
2257         .setup          = ip6gre_tunnel_setup,
2258         .validate       = ip6gre_tunnel_validate,
2259         .newlink        = ip6gre_newlink,
2260         .changelink     = ip6gre_changelink,
2261         .dellink        = ip6gre_dellink,
2262         .get_size       = ip6gre_get_size,
2263         .fill_info      = ip6gre_fill_info,
2264         .get_link_net   = ip6_tnl_get_link_net,
2265 };
2266
2267 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2268         .kind           = "ip6gretap",
2269         .maxtype        = IFLA_GRE_MAX,
2270         .policy         = ip6gre_policy,
2271         .priv_size      = sizeof(struct ip6_tnl),
2272         .setup          = ip6gre_tap_setup,
2273         .validate       = ip6gre_tap_validate,
2274         .newlink        = ip6gre_newlink,
2275         .changelink     = ip6gre_changelink,
2276         .get_size       = ip6gre_get_size,
2277         .fill_info      = ip6gre_fill_info,
2278         .get_link_net   = ip6_tnl_get_link_net,
2279 };
2280
2281 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2282         .kind           = "ip6erspan",
2283         .maxtype        = IFLA_GRE_MAX,
2284         .policy         = ip6gre_policy,
2285         .priv_size      = sizeof(struct ip6_tnl),
2286         .setup          = ip6erspan_tap_setup,
2287         .validate       = ip6erspan_tap_validate,
2288         .newlink        = ip6erspan_newlink,
2289         .changelink     = ip6erspan_changelink,
2290         .get_size       = ip6gre_get_size,
2291         .fill_info      = ip6gre_fill_info,
2292         .get_link_net   = ip6_tnl_get_link_net,
2293 };
2294
2295 /*
2296  *      And now the modules code and kernel interface.
2297  */
2298
2299 static int __init ip6gre_init(void)
2300 {
2301         int err;
2302
2303         pr_info("GRE over IPv6 tunneling driver\n");
2304
2305         err = register_pernet_device(&ip6gre_net_ops);
2306         if (err < 0)
2307                 return err;
2308
2309         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2310         if (err < 0) {
2311                 pr_info("%s: can't add protocol\n", __func__);
2312                 goto add_proto_failed;
2313         }
2314
2315         err = rtnl_link_register(&ip6gre_link_ops);
2316         if (err < 0)
2317                 goto rtnl_link_failed;
2318
2319         err = rtnl_link_register(&ip6gre_tap_ops);
2320         if (err < 0)
2321                 goto tap_ops_failed;
2322
2323         err = rtnl_link_register(&ip6erspan_tap_ops);
2324         if (err < 0)
2325                 goto erspan_link_failed;
2326
2327 out:
2328         return err;
2329
2330 erspan_link_failed:
2331         rtnl_link_unregister(&ip6gre_tap_ops);
2332 tap_ops_failed:
2333         rtnl_link_unregister(&ip6gre_link_ops);
2334 rtnl_link_failed:
2335         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2336 add_proto_failed:
2337         unregister_pernet_device(&ip6gre_net_ops);
2338         goto out;
2339 }
2340
2341 static void __exit ip6gre_fini(void)
2342 {
2343         rtnl_link_unregister(&ip6gre_tap_ops);
2344         rtnl_link_unregister(&ip6gre_link_ops);
2345         rtnl_link_unregister(&ip6erspan_tap_ops);
2346         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2347         unregister_pernet_device(&ip6gre_net_ops);
2348 }
2349
2350 module_init(ip6gre_init);
2351 module_exit(ip6gre_fini);
2352 MODULE_LICENSE("GPL");
2353 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2354 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2355 MODULE_ALIAS_RTNL_LINK("ip6gre");
2356 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2357 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2358 MODULE_ALIAS_NETDEV("ip6gre0");