]> asedeno.scripts.mit.edu Git - linux.git/blob - net/ipv6/ip6_gre.c
Merge tag 'mips_fixes_4.18_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/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 gre_base_hdr *greh;
431         const struct ipv6hdr *ipv6h;
432         int grehlen = sizeof(*greh);
433         struct ip6_tnl *t;
434         int key_off = 0;
435         __be16 flags;
436         __be32 key;
437
438         if (!pskb_may_pull(skb, offset + grehlen))
439                 return;
440         greh = (const struct gre_base_hdr *)(skb->data + offset);
441         flags = greh->flags;
442         if (flags & (GRE_VERSION | GRE_ROUTING))
443                 return;
444         if (flags & GRE_CSUM)
445                 grehlen += 4;
446         if (flags & GRE_KEY) {
447                 key_off = grehlen + offset;
448                 grehlen += 4;
449         }
450
451         if (!pskb_may_pull(skb, offset + grehlen))
452                 return;
453         ipv6h = (const struct ipv6hdr *)skb->data;
454         greh = (const struct gre_base_hdr *)(skb->data + offset);
455         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
456
457         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
458                                  key, greh->protocol);
459         if (!t)
460                 return;
461
462         switch (type) {
463                 struct ipv6_tlv_tnl_enc_lim *tel;
464                 __u32 teli;
465         case ICMPV6_DEST_UNREACH:
466                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
467                                     t->parms.name);
468                 if (code != ICMPV6_PORT_UNREACH)
469                         break;
470                 return;
471         case ICMPV6_TIME_EXCEED:
472                 if (code == ICMPV6_EXC_HOPLIMIT) {
473                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
474                                             t->parms.name);
475                         break;
476                 }
477                 return;
478         case ICMPV6_PARAMPROB:
479                 teli = 0;
480                 if (code == ICMPV6_HDR_FIELD)
481                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
482
483                 if (teli && teli == be32_to_cpu(info) - 2) {
484                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
485                         if (tel->encap_limit == 0) {
486                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
487                                                     t->parms.name);
488                         }
489                 } else {
490                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
491                                             t->parms.name);
492                 }
493                 return;
494         case ICMPV6_PKT_TOOBIG:
495                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
496                 return;
497         case NDISC_REDIRECT:
498                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
499                              sock_net_uid(net, NULL));
500                 return;
501         }
502
503         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
504                 t->err_count++;
505         else
506                 t->err_count = 1;
507         t->err_time = jiffies;
508 }
509
510 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
511 {
512         const struct ipv6hdr *ipv6h;
513         struct ip6_tnl *tunnel;
514
515         ipv6h = ipv6_hdr(skb);
516         tunnel = ip6gre_tunnel_lookup(skb->dev,
517                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
518                                       tpi->proto);
519         if (tunnel) {
520                 if (tunnel->parms.collect_md) {
521                         struct metadata_dst *tun_dst;
522                         __be64 tun_id;
523                         __be16 flags;
524
525                         flags = tpi->flags;
526                         tun_id = key32_to_tunnel_id(tpi->key);
527
528                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
529                         if (!tun_dst)
530                                 return PACKET_REJECT;
531
532                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
533                 } else {
534                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
535                 }
536
537                 return PACKET_RCVD;
538         }
539
540         return PACKET_REJECT;
541 }
542
543 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
544                          struct tnl_ptk_info *tpi)
545 {
546         struct erspan_base_hdr *ershdr;
547         struct erspan_metadata *pkt_md;
548         const struct ipv6hdr *ipv6h;
549         struct erspan_md2 *md2;
550         struct ip6_tnl *tunnel;
551         u8 ver;
552
553         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
554                 return PACKET_REJECT;
555
556         ipv6h = ipv6_hdr(skb);
557         ershdr = (struct erspan_base_hdr *)skb->data;
558         ver = ershdr->ver;
559         tpi->key = cpu_to_be32(get_session_id(ershdr));
560
561         tunnel = ip6gre_tunnel_lookup(skb->dev,
562                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
563                                       tpi->proto);
564         if (tunnel) {
565                 int len = erspan_hdr_len(ver);
566
567                 if (unlikely(!pskb_may_pull(skb, len)))
568                         return PACKET_REJECT;
569
570                 ershdr = (struct erspan_base_hdr *)skb->data;
571                 pkt_md = (struct erspan_metadata *)(ershdr + 1);
572
573                 if (__iptunnel_pull_header(skb, len,
574                                            htons(ETH_P_TEB),
575                                            false, false) < 0)
576                         return PACKET_REJECT;
577
578                 if (tunnel->parms.collect_md) {
579                         struct metadata_dst *tun_dst;
580                         struct ip_tunnel_info *info;
581                         struct erspan_metadata *md;
582                         __be64 tun_id;
583                         __be16 flags;
584
585                         tpi->flags |= TUNNEL_KEY;
586                         flags = tpi->flags;
587                         tun_id = key32_to_tunnel_id(tpi->key);
588
589                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
590                                                   sizeof(*md));
591                         if (!tun_dst)
592                                 return PACKET_REJECT;
593
594                         info = &tun_dst->u.tun_info;
595                         md = ip_tunnel_info_opts(info);
596                         md->version = ver;
597                         md2 = &md->u.md2;
598                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
599                                                        ERSPAN_V2_MDSIZE);
600                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
601                         info->options_len = sizeof(*md);
602
603                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
604
605                 } else {
606                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
607                 }
608
609                 return PACKET_RCVD;
610         }
611
612         return PACKET_REJECT;
613 }
614
615 static int gre_rcv(struct sk_buff *skb)
616 {
617         struct tnl_ptk_info tpi;
618         bool csum_err = false;
619         int hdr_len;
620
621         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
622         if (hdr_len < 0)
623                 goto drop;
624
625         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
626                 goto drop;
627
628         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
629                      tpi.proto == htons(ETH_P_ERSPAN2))) {
630                 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
631                         return 0;
632                 goto out;
633         }
634
635         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
636                 return 0;
637
638 out:
639         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
640 drop:
641         kfree_skb(skb);
642         return 0;
643 }
644
645 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
646 {
647         return iptunnel_handle_offloads(skb,
648                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
649 }
650
651 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
652                                      struct net_device *dev,
653                                      struct flowi6 *fl6, __u8 *dsfield,
654                                      int *encap_limit)
655 {
656         const struct iphdr *iph = ip_hdr(skb);
657         struct ip6_tnl *t = netdev_priv(dev);
658
659         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
660                 *encap_limit = t->parms.encap_limit;
661
662         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
663
664         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
665                 *dsfield = ipv4_get_dsfield(iph);
666         else
667                 *dsfield = ip6_tclass(t->parms.flowinfo);
668
669         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
670                 fl6->flowi6_mark = skb->mark;
671         else
672                 fl6->flowi6_mark = t->parms.fwmark;
673
674         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
675 }
676
677 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
678                                     struct net_device *dev,
679                                     struct flowi6 *fl6, __u8 *dsfield,
680                                     int *encap_limit)
681 {
682         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
683         struct ip6_tnl *t = netdev_priv(dev);
684         __u16 offset;
685
686         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
687         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
688
689         if (offset > 0) {
690                 struct ipv6_tlv_tnl_enc_lim *tel;
691
692                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
693                 if (tel->encap_limit == 0) {
694                         icmpv6_send(skb, ICMPV6_PARAMPROB,
695                                     ICMPV6_HDR_FIELD, offset + 2);
696                         return -1;
697                 }
698                 *encap_limit = tel->encap_limit - 1;
699         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
700                 *encap_limit = t->parms.encap_limit;
701         }
702
703         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
704
705         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
706                 *dsfield = ipv6_get_dsfield(ipv6h);
707         else
708                 *dsfield = ip6_tclass(t->parms.flowinfo);
709
710         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
711                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
712
713         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
714                 fl6->flowi6_mark = skb->mark;
715         else
716                 fl6->flowi6_mark = t->parms.fwmark;
717
718         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
719
720         return 0;
721 }
722
723 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
724                                struct net_device *dev, __u8 dsfield,
725                                struct flowi6 *fl6, int encap_limit,
726                                __u32 *pmtu, __be16 proto)
727 {
728         struct ip6_tnl *tunnel = netdev_priv(dev);
729         __be16 protocol;
730
731         if (dev->type == ARPHRD_ETHER)
732                 IPCB(skb)->flags = 0;
733
734         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
735                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
736         else
737                 fl6->daddr = tunnel->parms.raddr;
738
739         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
740                 return -ENOMEM;
741
742         /* Push GRE header. */
743         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
744
745         if (tunnel->parms.collect_md) {
746                 struct ip_tunnel_info *tun_info;
747                 const struct ip_tunnel_key *key;
748                 __be16 flags;
749
750                 tun_info = skb_tunnel_info(skb);
751                 if (unlikely(!tun_info ||
752                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
753                              ip_tunnel_info_af(tun_info) != AF_INET6))
754                         return -EINVAL;
755
756                 key = &tun_info->key;
757                 memset(fl6, 0, sizeof(*fl6));
758                 fl6->flowi6_proto = IPPROTO_GRE;
759                 fl6->daddr = key->u.ipv6.dst;
760                 fl6->flowlabel = key->label;
761                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
762
763                 dsfield = key->tos;
764                 flags = key->tun_flags &
765                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
766                 tunnel->tun_hlen = gre_calc_hlen(flags);
767
768                 gre_build_header(skb, tunnel->tun_hlen,
769                                  flags, protocol,
770                                  tunnel_id_to_key32(tun_info->key.tun_id),
771                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
772                                                       : 0);
773
774         } else {
775                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
776                         tunnel->o_seqno++;
777
778                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
779                                  protocol, tunnel->parms.o_key,
780                                  htonl(tunnel->o_seqno));
781         }
782
783         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
784                             NEXTHDR_GRE);
785 }
786
787 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
788 {
789         struct ip6_tnl *t = netdev_priv(dev);
790         int encap_limit = -1;
791         struct flowi6 fl6;
792         __u8 dsfield = 0;
793         __u32 mtu;
794         int err;
795
796         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
797
798         if (!t->parms.collect_md)
799                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
800                                          &dsfield, &encap_limit);
801
802         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
803         if (err)
804                 return -1;
805
806         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
807                           skb->protocol);
808         if (err != 0) {
809                 /* XXX: send ICMP error even if DF is not set. */
810                 if (err == -EMSGSIZE)
811                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
812                                   htonl(mtu));
813                 return -1;
814         }
815
816         return 0;
817 }
818
819 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
820 {
821         struct ip6_tnl *t = netdev_priv(dev);
822         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
823         int encap_limit = -1;
824         struct flowi6 fl6;
825         __u8 dsfield = 0;
826         __u32 mtu;
827         int err;
828
829         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
830                 return -1;
831
832         if (!t->parms.collect_md &&
833             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
834                 return -1;
835
836         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
837                 return -1;
838
839         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
840                           &mtu, skb->protocol);
841         if (err != 0) {
842                 if (err == -EMSGSIZE)
843                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
844                 return -1;
845         }
846
847         return 0;
848 }
849
850 /**
851  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
852  *   @t: the outgoing tunnel device
853  *   @hdr: IPv6 header from the incoming packet
854  *
855  * Description:
856  *   Avoid trivial tunneling loop by checking that tunnel exit-point
857  *   doesn't match source of incoming packet.
858  *
859  * Return:
860  *   1 if conflict,
861  *   0 else
862  **/
863
864 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
865         const struct ipv6hdr *hdr)
866 {
867         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
868 }
869
870 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
871 {
872         struct ip6_tnl *t = netdev_priv(dev);
873         int encap_limit = -1;
874         struct flowi6 fl6;
875         __u32 mtu;
876         int err;
877
878         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
879                 encap_limit = t->parms.encap_limit;
880
881         if (!t->parms.collect_md)
882                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
883
884         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
885         if (err)
886                 return err;
887
888         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
889
890         return err;
891 }
892
893 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
894         struct net_device *dev)
895 {
896         struct ip6_tnl *t = netdev_priv(dev);
897         struct net_device_stats *stats = &t->dev->stats;
898         int ret;
899
900         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
901                 goto tx_err;
902
903         switch (skb->protocol) {
904         case htons(ETH_P_IP):
905                 ret = ip6gre_xmit_ipv4(skb, dev);
906                 break;
907         case htons(ETH_P_IPV6):
908                 ret = ip6gre_xmit_ipv6(skb, dev);
909                 break;
910         default:
911                 ret = ip6gre_xmit_other(skb, dev);
912                 break;
913         }
914
915         if (ret < 0)
916                 goto tx_err;
917
918         return NETDEV_TX_OK;
919
920 tx_err:
921         stats->tx_errors++;
922         stats->tx_dropped++;
923         kfree_skb(skb);
924         return NETDEV_TX_OK;
925 }
926
927 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
928                                          struct net_device *dev)
929 {
930         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
931         struct ip6_tnl *t = netdev_priv(dev);
932         struct dst_entry *dst = skb_dst(skb);
933         struct net_device_stats *stats;
934         bool truncate = false;
935         int encap_limit = -1;
936         __u8 dsfield = false;
937         struct flowi6 fl6;
938         int err = -EINVAL;
939         __u32 mtu;
940         int nhoff;
941         int thoff;
942
943         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
944                 goto tx_err;
945
946         if (gre_handle_offloads(skb, false))
947                 goto tx_err;
948
949         if (skb->len > dev->mtu + dev->hard_header_len) {
950                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
951                 truncate = true;
952         }
953
954         nhoff = skb_network_header(skb) - skb_mac_header(skb);
955         if (skb->protocol == htons(ETH_P_IP) &&
956             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
957                 truncate = true;
958
959         thoff = skb_transport_header(skb) - skb_mac_header(skb);
960         if (skb->protocol == htons(ETH_P_IPV6) &&
961             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
962                 truncate = true;
963
964         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
965                 goto tx_err;
966
967         t->parms.o_flags &= ~TUNNEL_KEY;
968         IPCB(skb)->flags = 0;
969
970         /* For collect_md mode, derive fl6 from the tunnel key,
971          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
972          */
973         if (t->parms.collect_md) {
974                 struct ip_tunnel_info *tun_info;
975                 const struct ip_tunnel_key *key;
976                 struct erspan_metadata *md;
977                 __be32 tun_id;
978
979                 tun_info = skb_tunnel_info(skb);
980                 if (unlikely(!tun_info ||
981                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
982                              ip_tunnel_info_af(tun_info) != AF_INET6))
983                         return -EINVAL;
984
985                 key = &tun_info->key;
986                 memset(&fl6, 0, sizeof(fl6));
987                 fl6.flowi6_proto = IPPROTO_GRE;
988                 fl6.daddr = key->u.ipv6.dst;
989                 fl6.flowlabel = key->label;
990                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
991
992                 dsfield = key->tos;
993                 md = ip_tunnel_info_opts(tun_info);
994                 if (!md)
995                         goto tx_err;
996
997                 tun_id = tunnel_id_to_key32(key->tun_id);
998                 if (md->version == 1) {
999                         erspan_build_header(skb,
1000                                             ntohl(tun_id),
1001                                             ntohl(md->u.index), truncate,
1002                                             false);
1003                 } else if (md->version == 2) {
1004                         erspan_build_header_v2(skb,
1005                                                ntohl(tun_id),
1006                                                md->u.md2.dir,
1007                                                get_hwid(&md->u.md2),
1008                                                truncate, false);
1009                 } else {
1010                         goto tx_err;
1011                 }
1012         } else {
1013                 switch (skb->protocol) {
1014                 case htons(ETH_P_IP):
1015                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1016                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1017                                                  &dsfield, &encap_limit);
1018                         break;
1019                 case htons(ETH_P_IPV6):
1020                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
1021                                 goto tx_err;
1022                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1023                                                      &dsfield, &encap_limit))
1024                                 goto tx_err;
1025                         break;
1026                 default:
1027                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1028                         break;
1029                 }
1030
1031                 if (t->parms.erspan_ver == 1)
1032                         erspan_build_header(skb, ntohl(t->parms.o_key),
1033                                             t->parms.index,
1034                                             truncate, false);
1035                 else if (t->parms.erspan_ver == 2)
1036                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1037                                                t->parms.dir,
1038                                                t->parms.hwid,
1039                                                truncate, false);
1040                 else
1041                         goto tx_err;
1042
1043                 fl6.daddr = t->parms.raddr;
1044         }
1045
1046         /* Push GRE header. */
1047         gre_build_header(skb, 8, TUNNEL_SEQ,
1048                          htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1049
1050         /* TooBig packet may have updated dst->dev's mtu */
1051         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1052                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1053
1054         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1055                            NEXTHDR_GRE);
1056         if (err != 0) {
1057                 /* XXX: send ICMP error even if DF is not set. */
1058                 if (err == -EMSGSIZE) {
1059                         if (skb->protocol == htons(ETH_P_IP))
1060                                 icmp_send(skb, ICMP_DEST_UNREACH,
1061                                           ICMP_FRAG_NEEDED, htonl(mtu));
1062                         else
1063                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1064                 }
1065
1066                 goto tx_err;
1067         }
1068         return NETDEV_TX_OK;
1069
1070 tx_err:
1071         stats = &t->dev->stats;
1072         stats->tx_errors++;
1073         stats->tx_dropped++;
1074         kfree_skb(skb);
1075         return NETDEV_TX_OK;
1076 }
1077
1078 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1079 {
1080         struct net_device *dev = t->dev;
1081         struct __ip6_tnl_parm *p = &t->parms;
1082         struct flowi6 *fl6 = &t->fl.u.ip6;
1083
1084         if (dev->type != ARPHRD_ETHER) {
1085                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1086                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1087         }
1088
1089         /* Set up flowi template */
1090         fl6->saddr = p->laddr;
1091         fl6->daddr = p->raddr;
1092         fl6->flowi6_oif = p->link;
1093         fl6->flowlabel = 0;
1094         fl6->flowi6_proto = IPPROTO_GRE;
1095
1096         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1097                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1098         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1099                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1100
1101         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1102         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1103
1104         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1105                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1106                 dev->flags |= IFF_POINTOPOINT;
1107         else
1108                 dev->flags &= ~IFF_POINTOPOINT;
1109 }
1110
1111 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1112                                          int t_hlen)
1113 {
1114         const struct __ip6_tnl_parm *p = &t->parms;
1115         struct net_device *dev = t->dev;
1116
1117         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1118                 int strict = (ipv6_addr_type(&p->raddr) &
1119                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1120
1121                 struct rt6_info *rt = rt6_lookup(t->net,
1122                                                  &p->raddr, &p->laddr,
1123                                                  p->link, NULL, strict);
1124
1125                 if (!rt)
1126                         return;
1127
1128                 if (rt->dst.dev) {
1129                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1130                                                t_hlen;
1131
1132                         if (set_mtu) {
1133                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1134                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1135                                         dev->mtu -= 8;
1136                                 if (dev->type == ARPHRD_ETHER)
1137                                         dev->mtu -= ETH_HLEN;
1138
1139                                 if (dev->mtu < IPV6_MIN_MTU)
1140                                         dev->mtu = IPV6_MIN_MTU;
1141                         }
1142                 }
1143                 ip6_rt_put(rt);
1144         }
1145 }
1146
1147 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1148 {
1149         int t_hlen;
1150
1151         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1152         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1153
1154         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1155         tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1156         return t_hlen;
1157 }
1158
1159 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1160 {
1161         ip6gre_tnl_link_config_common(t);
1162         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1163 }
1164
1165 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1166                                      const struct __ip6_tnl_parm *p)
1167 {
1168         t->parms.laddr = p->laddr;
1169         t->parms.raddr = p->raddr;
1170         t->parms.flags = p->flags;
1171         t->parms.hop_limit = p->hop_limit;
1172         t->parms.encap_limit = p->encap_limit;
1173         t->parms.flowinfo = p->flowinfo;
1174         t->parms.link = p->link;
1175         t->parms.proto = p->proto;
1176         t->parms.i_key = p->i_key;
1177         t->parms.o_key = p->o_key;
1178         t->parms.i_flags = p->i_flags;
1179         t->parms.o_flags = p->o_flags;
1180         t->parms.fwmark = p->fwmark;
1181         dst_cache_reset(&t->dst_cache);
1182 }
1183
1184 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1185                              int set_mtu)
1186 {
1187         ip6gre_tnl_copy_tnl_parm(t, p);
1188         ip6gre_tnl_link_config(t, set_mtu);
1189         return 0;
1190 }
1191
1192 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1193         const struct ip6_tnl_parm2 *u)
1194 {
1195         p->laddr = u->laddr;
1196         p->raddr = u->raddr;
1197         p->flags = u->flags;
1198         p->hop_limit = u->hop_limit;
1199         p->encap_limit = u->encap_limit;
1200         p->flowinfo = u->flowinfo;
1201         p->link = u->link;
1202         p->i_key = u->i_key;
1203         p->o_key = u->o_key;
1204         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1205         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1206         memcpy(p->name, u->name, sizeof(u->name));
1207 }
1208
1209 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1210         const struct __ip6_tnl_parm *p)
1211 {
1212         u->proto = IPPROTO_GRE;
1213         u->laddr = p->laddr;
1214         u->raddr = p->raddr;
1215         u->flags = p->flags;
1216         u->hop_limit = p->hop_limit;
1217         u->encap_limit = p->encap_limit;
1218         u->flowinfo = p->flowinfo;
1219         u->link = p->link;
1220         u->i_key = p->i_key;
1221         u->o_key = p->o_key;
1222         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1223         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1224         memcpy(u->name, p->name, sizeof(u->name));
1225 }
1226
1227 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1228         struct ifreq *ifr, int cmd)
1229 {
1230         int err = 0;
1231         struct ip6_tnl_parm2 p;
1232         struct __ip6_tnl_parm p1;
1233         struct ip6_tnl *t = netdev_priv(dev);
1234         struct net *net = t->net;
1235         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1236
1237         memset(&p1, 0, sizeof(p1));
1238
1239         switch (cmd) {
1240         case SIOCGETTUNNEL:
1241                 if (dev == ign->fb_tunnel_dev) {
1242                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1243                                 err = -EFAULT;
1244                                 break;
1245                         }
1246                         ip6gre_tnl_parm_from_user(&p1, &p);
1247                         t = ip6gre_tunnel_locate(net, &p1, 0);
1248                         if (!t)
1249                                 t = netdev_priv(dev);
1250                 }
1251                 memset(&p, 0, sizeof(p));
1252                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1253                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1254                         err = -EFAULT;
1255                 break;
1256
1257         case SIOCADDTUNNEL:
1258         case SIOCCHGTUNNEL:
1259                 err = -EPERM;
1260                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1261                         goto done;
1262
1263                 err = -EFAULT;
1264                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1265                         goto done;
1266
1267                 err = -EINVAL;
1268                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1269                         goto done;
1270
1271                 if (!(p.i_flags&GRE_KEY))
1272                         p.i_key = 0;
1273                 if (!(p.o_flags&GRE_KEY))
1274                         p.o_key = 0;
1275
1276                 ip6gre_tnl_parm_from_user(&p1, &p);
1277                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1278
1279                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1280                         if (t) {
1281                                 if (t->dev != dev) {
1282                                         err = -EEXIST;
1283                                         break;
1284                                 }
1285                         } else {
1286                                 t = netdev_priv(dev);
1287
1288                                 ip6gre_tunnel_unlink(ign, t);
1289                                 synchronize_net();
1290                                 ip6gre_tnl_change(t, &p1, 1);
1291                                 ip6gre_tunnel_link(ign, t);
1292                                 netdev_state_change(dev);
1293                         }
1294                 }
1295
1296                 if (t) {
1297                         err = 0;
1298
1299                         memset(&p, 0, sizeof(p));
1300                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1301                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1302                                 err = -EFAULT;
1303                 } else
1304                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1305                 break;
1306
1307         case SIOCDELTUNNEL:
1308                 err = -EPERM;
1309                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1310                         goto done;
1311
1312                 if (dev == ign->fb_tunnel_dev) {
1313                         err = -EFAULT;
1314                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1315                                 goto done;
1316                         err = -ENOENT;
1317                         ip6gre_tnl_parm_from_user(&p1, &p);
1318                         t = ip6gre_tunnel_locate(net, &p1, 0);
1319                         if (!t)
1320                                 goto done;
1321                         err = -EPERM;
1322                         if (t == netdev_priv(ign->fb_tunnel_dev))
1323                                 goto done;
1324                         dev = t->dev;
1325                 }
1326                 unregister_netdevice(dev);
1327                 err = 0;
1328                 break;
1329
1330         default:
1331                 err = -EINVAL;
1332         }
1333
1334 done:
1335         return err;
1336 }
1337
1338 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1339                          unsigned short type, const void *daddr,
1340                          const void *saddr, unsigned int len)
1341 {
1342         struct ip6_tnl *t = netdev_priv(dev);
1343         struct ipv6hdr *ipv6h;
1344         __be16 *p;
1345
1346         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1347         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1348                                                   t->fl.u.ip6.flowlabel,
1349                                                   true, &t->fl.u.ip6));
1350         ipv6h->hop_limit = t->parms.hop_limit;
1351         ipv6h->nexthdr = NEXTHDR_GRE;
1352         ipv6h->saddr = t->parms.laddr;
1353         ipv6h->daddr = t->parms.raddr;
1354
1355         p = (__be16 *)(ipv6h + 1);
1356         p[0] = t->parms.o_flags;
1357         p[1] = htons(type);
1358
1359         /*
1360          *      Set the source hardware address.
1361          */
1362
1363         if (saddr)
1364                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1365         if (daddr)
1366                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1367         if (!ipv6_addr_any(&ipv6h->daddr))
1368                 return t->hlen;
1369
1370         return -t->hlen;
1371 }
1372
1373 static const struct header_ops ip6gre_header_ops = {
1374         .create = ip6gre_header,
1375 };
1376
1377 static const struct net_device_ops ip6gre_netdev_ops = {
1378         .ndo_init               = ip6gre_tunnel_init,
1379         .ndo_uninit             = ip6gre_tunnel_uninit,
1380         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1381         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1382         .ndo_change_mtu         = ip6_tnl_change_mtu,
1383         .ndo_get_stats64        = ip_tunnel_get_stats64,
1384         .ndo_get_iflink         = ip6_tnl_get_iflink,
1385 };
1386
1387 static void ip6gre_dev_free(struct net_device *dev)
1388 {
1389         struct ip6_tnl *t = netdev_priv(dev);
1390
1391         gro_cells_destroy(&t->gro_cells);
1392         dst_cache_destroy(&t->dst_cache);
1393         free_percpu(dev->tstats);
1394 }
1395
1396 static void ip6gre_tunnel_setup(struct net_device *dev)
1397 {
1398         dev->netdev_ops = &ip6gre_netdev_ops;
1399         dev->needs_free_netdev = true;
1400         dev->priv_destructor = ip6gre_dev_free;
1401
1402         dev->type = ARPHRD_IP6GRE;
1403
1404         dev->flags |= IFF_NOARP;
1405         dev->addr_len = sizeof(struct in6_addr);
1406         netif_keep_dst(dev);
1407         /* This perm addr will be used as interface identifier by IPv6 */
1408         dev->addr_assign_type = NET_ADDR_RANDOM;
1409         eth_random_addr(dev->perm_addr);
1410 }
1411
1412 #define GRE6_FEATURES (NETIF_F_SG |             \
1413                        NETIF_F_FRAGLIST |       \
1414                        NETIF_F_HIGHDMA |        \
1415                        NETIF_F_HW_CSUM)
1416
1417 static void ip6gre_tnl_init_features(struct net_device *dev)
1418 {
1419         struct ip6_tnl *nt = netdev_priv(dev);
1420
1421         dev->features           |= GRE6_FEATURES;
1422         dev->hw_features        |= GRE6_FEATURES;
1423
1424         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1425                 /* TCP offload with GRE SEQ is not supported, nor
1426                  * can we support 2 levels of outer headers requiring
1427                  * an update.
1428                  */
1429                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1430                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1431                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1432                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1433                 }
1434
1435                 /* Can use a lockless transmit, unless we generate
1436                  * output sequences
1437                  */
1438                 dev->features |= NETIF_F_LLTX;
1439         }
1440 }
1441
1442 static int ip6gre_tunnel_init_common(struct net_device *dev)
1443 {
1444         struct ip6_tnl *tunnel;
1445         int ret;
1446         int t_hlen;
1447
1448         tunnel = netdev_priv(dev);
1449
1450         tunnel->dev = dev;
1451         tunnel->net = dev_net(dev);
1452         strcpy(tunnel->parms.name, dev->name);
1453
1454         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1455         if (!dev->tstats)
1456                 return -ENOMEM;
1457
1458         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1459         if (ret)
1460                 goto cleanup_alloc_pcpu_stats;
1461
1462         ret = gro_cells_init(&tunnel->gro_cells, dev);
1463         if (ret)
1464                 goto cleanup_dst_cache_init;
1465
1466         t_hlen = ip6gre_calc_hlen(tunnel);
1467         dev->mtu = ETH_DATA_LEN - t_hlen;
1468         if (dev->type == ARPHRD_ETHER)
1469                 dev->mtu -= ETH_HLEN;
1470         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1471                 dev->mtu -= 8;
1472
1473         if (tunnel->parms.collect_md) {
1474                 dev->features |= NETIF_F_NETNS_LOCAL;
1475                 netif_keep_dst(dev);
1476         }
1477         ip6gre_tnl_init_features(dev);
1478
1479         return 0;
1480
1481 cleanup_dst_cache_init:
1482         dst_cache_destroy(&tunnel->dst_cache);
1483 cleanup_alloc_pcpu_stats:
1484         free_percpu(dev->tstats);
1485         dev->tstats = NULL;
1486         return ret;
1487 }
1488
1489 static int ip6gre_tunnel_init(struct net_device *dev)
1490 {
1491         struct ip6_tnl *tunnel;
1492         int ret;
1493
1494         ret = ip6gre_tunnel_init_common(dev);
1495         if (ret)
1496                 return ret;
1497
1498         tunnel = netdev_priv(dev);
1499
1500         if (tunnel->parms.collect_md)
1501                 return 0;
1502
1503         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1504         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1505
1506         if (ipv6_addr_any(&tunnel->parms.raddr))
1507                 dev->header_ops = &ip6gre_header_ops;
1508
1509         return 0;
1510 }
1511
1512 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1513 {
1514         struct ip6_tnl *tunnel = netdev_priv(dev);
1515
1516         tunnel->dev = dev;
1517         tunnel->net = dev_net(dev);
1518         strcpy(tunnel->parms.name, dev->name);
1519
1520         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1521
1522         dev_hold(dev);
1523 }
1524
1525 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1526         .handler     = gre_rcv,
1527         .err_handler = ip6gre_err,
1528         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1529 };
1530
1531 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1532 {
1533         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1534         struct net_device *dev, *aux;
1535         int prio;
1536
1537         for_each_netdev_safe(net, dev, aux)
1538                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1539                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1540                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1541                         unregister_netdevice_queue(dev, head);
1542
1543         for (prio = 0; prio < 4; prio++) {
1544                 int h;
1545                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1546                         struct ip6_tnl *t;
1547
1548                         t = rtnl_dereference(ign->tunnels[prio][h]);
1549
1550                         while (t) {
1551                                 /* If dev is in the same netns, it has already
1552                                  * been added to the list by the previous loop.
1553                                  */
1554                                 if (!net_eq(dev_net(t->dev), net))
1555                                         unregister_netdevice_queue(t->dev,
1556                                                                    head);
1557                                 t = rtnl_dereference(t->next);
1558                         }
1559                 }
1560         }
1561 }
1562
1563 static int __net_init ip6gre_init_net(struct net *net)
1564 {
1565         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1566         int err;
1567
1568         if (!net_has_fallback_tunnels(net))
1569                 return 0;
1570         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1571                                           NET_NAME_UNKNOWN,
1572                                           ip6gre_tunnel_setup);
1573         if (!ign->fb_tunnel_dev) {
1574                 err = -ENOMEM;
1575                 goto err_alloc_dev;
1576         }
1577         dev_net_set(ign->fb_tunnel_dev, net);
1578         /* FB netdevice is special: we have one, and only one per netns.
1579          * Allowing to move it to another netns is clearly unsafe.
1580          */
1581         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1582
1583
1584         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1585         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1586
1587         err = register_netdev(ign->fb_tunnel_dev);
1588         if (err)
1589                 goto err_reg_dev;
1590
1591         rcu_assign_pointer(ign->tunnels_wc[0],
1592                            netdev_priv(ign->fb_tunnel_dev));
1593         return 0;
1594
1595 err_reg_dev:
1596         free_netdev(ign->fb_tunnel_dev);
1597 err_alloc_dev:
1598         return err;
1599 }
1600
1601 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1602 {
1603         struct net *net;
1604         LIST_HEAD(list);
1605
1606         rtnl_lock();
1607         list_for_each_entry(net, net_list, exit_list)
1608                 ip6gre_destroy_tunnels(net, &list);
1609         unregister_netdevice_many(&list);
1610         rtnl_unlock();
1611 }
1612
1613 static struct pernet_operations ip6gre_net_ops = {
1614         .init = ip6gre_init_net,
1615         .exit_batch = ip6gre_exit_batch_net,
1616         .id   = &ip6gre_net_id,
1617         .size = sizeof(struct ip6gre_net),
1618 };
1619
1620 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1621                                   struct netlink_ext_ack *extack)
1622 {
1623         __be16 flags;
1624
1625         if (!data)
1626                 return 0;
1627
1628         flags = 0;
1629         if (data[IFLA_GRE_IFLAGS])
1630                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1631         if (data[IFLA_GRE_OFLAGS])
1632                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1633         if (flags & (GRE_VERSION|GRE_ROUTING))
1634                 return -EINVAL;
1635
1636         return 0;
1637 }
1638
1639 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1640                                struct netlink_ext_ack *extack)
1641 {
1642         struct in6_addr daddr;
1643
1644         if (tb[IFLA_ADDRESS]) {
1645                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1646                         return -EINVAL;
1647                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1648                         return -EADDRNOTAVAIL;
1649         }
1650
1651         if (!data)
1652                 goto out;
1653
1654         if (data[IFLA_GRE_REMOTE]) {
1655                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1656                 if (ipv6_addr_any(&daddr))
1657                         return -EINVAL;
1658         }
1659
1660 out:
1661         return ip6gre_tunnel_validate(tb, data, extack);
1662 }
1663
1664 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1665                                   struct netlink_ext_ack *extack)
1666 {
1667         __be16 flags = 0;
1668         int ret, ver = 0;
1669
1670         if (!data)
1671                 return 0;
1672
1673         ret = ip6gre_tap_validate(tb, data, extack);
1674         if (ret)
1675                 return ret;
1676
1677         /* ERSPAN should only have GRE sequence and key flag */
1678         if (data[IFLA_GRE_OFLAGS])
1679                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1680         if (data[IFLA_GRE_IFLAGS])
1681                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1682         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1683             flags != (GRE_SEQ | GRE_KEY))
1684                 return -EINVAL;
1685
1686         /* ERSPAN Session ID only has 10-bit. Since we reuse
1687          * 32-bit key field as ID, check it's range.
1688          */
1689         if (data[IFLA_GRE_IKEY] &&
1690             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1691                 return -EINVAL;
1692
1693         if (data[IFLA_GRE_OKEY] &&
1694             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1695                 return -EINVAL;
1696
1697         if (data[IFLA_GRE_ERSPAN_VER]) {
1698                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1699                 if (ver != 1 && ver != 2)
1700                         return -EINVAL;
1701         }
1702
1703         if (ver == 1) {
1704                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1705                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1706
1707                         if (index & ~INDEX_MASK)
1708                                 return -EINVAL;
1709                 }
1710         } else if (ver == 2) {
1711                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1712                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1713
1714                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1715                                 return -EINVAL;
1716                 }
1717
1718                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1719                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1720
1721                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1722                                 return -EINVAL;
1723                 }
1724         }
1725
1726         return 0;
1727 }
1728
1729 static void ip6gre_netlink_parms(struct nlattr *data[],
1730                                 struct __ip6_tnl_parm *parms)
1731 {
1732         memset(parms, 0, sizeof(*parms));
1733
1734         if (!data)
1735                 return;
1736
1737         if (data[IFLA_GRE_LINK])
1738                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1739
1740         if (data[IFLA_GRE_IFLAGS])
1741                 parms->i_flags = gre_flags_to_tnl_flags(
1742                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1743
1744         if (data[IFLA_GRE_OFLAGS])
1745                 parms->o_flags = gre_flags_to_tnl_flags(
1746                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1747
1748         if (data[IFLA_GRE_IKEY])
1749                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1750
1751         if (data[IFLA_GRE_OKEY])
1752                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1753
1754         if (data[IFLA_GRE_LOCAL])
1755                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1756
1757         if (data[IFLA_GRE_REMOTE])
1758                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1759
1760         if (data[IFLA_GRE_TTL])
1761                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1762
1763         if (data[IFLA_GRE_ENCAP_LIMIT])
1764                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1765
1766         if (data[IFLA_GRE_FLOWINFO])
1767                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1768
1769         if (data[IFLA_GRE_FLAGS])
1770                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1771
1772         if (data[IFLA_GRE_FWMARK])
1773                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1774
1775         if (data[IFLA_GRE_COLLECT_METADATA])
1776                 parms->collect_md = true;
1777
1778         if (data[IFLA_GRE_ERSPAN_VER])
1779                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1780
1781         if (parms->erspan_ver == 1) {
1782                 if (data[IFLA_GRE_ERSPAN_INDEX])
1783                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1784         } else if (parms->erspan_ver == 2) {
1785                 if (data[IFLA_GRE_ERSPAN_DIR])
1786                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1787                 if (data[IFLA_GRE_ERSPAN_HWID])
1788                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1789         }
1790 }
1791
1792 static int ip6gre_tap_init(struct net_device *dev)
1793 {
1794         int ret;
1795
1796         ret = ip6gre_tunnel_init_common(dev);
1797         if (ret)
1798                 return ret;
1799
1800         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1801
1802         return 0;
1803 }
1804
1805 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1806         .ndo_init = ip6gre_tap_init,
1807         .ndo_uninit = ip6gre_tunnel_uninit,
1808         .ndo_start_xmit = ip6gre_tunnel_xmit,
1809         .ndo_set_mac_address = eth_mac_addr,
1810         .ndo_validate_addr = eth_validate_addr,
1811         .ndo_change_mtu = ip6_tnl_change_mtu,
1812         .ndo_get_stats64 = ip_tunnel_get_stats64,
1813         .ndo_get_iflink = ip6_tnl_get_iflink,
1814 };
1815
1816 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1817 {
1818         int t_hlen;
1819
1820         tunnel->tun_hlen = 8;
1821         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1822                        erspan_hdr_len(tunnel->parms.erspan_ver);
1823
1824         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1825         tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1826         return t_hlen;
1827 }
1828
1829 static int ip6erspan_tap_init(struct net_device *dev)
1830 {
1831         struct ip6_tnl *tunnel;
1832         int t_hlen;
1833         int ret;
1834
1835         tunnel = netdev_priv(dev);
1836
1837         tunnel->dev = dev;
1838         tunnel->net = dev_net(dev);
1839         strcpy(tunnel->parms.name, dev->name);
1840
1841         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1842         if (!dev->tstats)
1843                 return -ENOMEM;
1844
1845         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1846         if (ret)
1847                 goto cleanup_alloc_pcpu_stats;
1848
1849         ret = gro_cells_init(&tunnel->gro_cells, dev);
1850         if (ret)
1851                 goto cleanup_dst_cache_init;
1852
1853         t_hlen = ip6erspan_calc_hlen(tunnel);
1854         dev->mtu = ETH_DATA_LEN - t_hlen;
1855         if (dev->type == ARPHRD_ETHER)
1856                 dev->mtu -= ETH_HLEN;
1857         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1858                 dev->mtu -= 8;
1859
1860         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1861         ip6erspan_tnl_link_config(tunnel, 1);
1862
1863         return 0;
1864
1865 cleanup_dst_cache_init:
1866         dst_cache_destroy(&tunnel->dst_cache);
1867 cleanup_alloc_pcpu_stats:
1868         free_percpu(dev->tstats);
1869         dev->tstats = NULL;
1870         return ret;
1871 }
1872
1873 static const struct net_device_ops ip6erspan_netdev_ops = {
1874         .ndo_init =             ip6erspan_tap_init,
1875         .ndo_uninit =           ip6erspan_tunnel_uninit,
1876         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1877         .ndo_set_mac_address =  eth_mac_addr,
1878         .ndo_validate_addr =    eth_validate_addr,
1879         .ndo_change_mtu =       ip6_tnl_change_mtu,
1880         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1881         .ndo_get_iflink =       ip6_tnl_get_iflink,
1882 };
1883
1884 static void ip6gre_tap_setup(struct net_device *dev)
1885 {
1886
1887         ether_setup(dev);
1888
1889         dev->max_mtu = 0;
1890         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1891         dev->needs_free_netdev = true;
1892         dev->priv_destructor = ip6gre_dev_free;
1893
1894         dev->features |= NETIF_F_NETNS_LOCAL;
1895         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1896         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1897         netif_keep_dst(dev);
1898 }
1899
1900 bool is_ip6gretap_dev(const struct net_device *dev)
1901 {
1902         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1903 }
1904 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1905
1906 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1907                                        struct ip_tunnel_encap *ipencap)
1908 {
1909         bool ret = false;
1910
1911         memset(ipencap, 0, sizeof(*ipencap));
1912
1913         if (!data)
1914                 return ret;
1915
1916         if (data[IFLA_GRE_ENCAP_TYPE]) {
1917                 ret = true;
1918                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1919         }
1920
1921         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1922                 ret = true;
1923                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1924         }
1925
1926         if (data[IFLA_GRE_ENCAP_SPORT]) {
1927                 ret = true;
1928                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1929         }
1930
1931         if (data[IFLA_GRE_ENCAP_DPORT]) {
1932                 ret = true;
1933                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1934         }
1935
1936         return ret;
1937 }
1938
1939 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1940                                  struct nlattr *tb[], struct nlattr *data[],
1941                                  struct netlink_ext_ack *extack)
1942 {
1943         struct ip6_tnl *nt;
1944         struct ip_tunnel_encap ipencap;
1945         int err;
1946
1947         nt = netdev_priv(dev);
1948
1949         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1950                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1951
1952                 if (err < 0)
1953                         return err;
1954         }
1955
1956         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1957                 eth_hw_addr_random(dev);
1958
1959         nt->dev = dev;
1960         nt->net = dev_net(dev);
1961
1962         err = register_netdevice(dev);
1963         if (err)
1964                 goto out;
1965
1966         if (tb[IFLA_MTU])
1967                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1968
1969         dev_hold(dev);
1970
1971 out:
1972         return err;
1973 }
1974
1975 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1976                           struct nlattr *tb[], struct nlattr *data[],
1977                           struct netlink_ext_ack *extack)
1978 {
1979         struct ip6_tnl *nt = netdev_priv(dev);
1980         struct net *net = dev_net(dev);
1981         struct ip6gre_net *ign;
1982         int err;
1983
1984         ip6gre_netlink_parms(data, &nt->parms);
1985         ign = net_generic(net, ip6gre_net_id);
1986
1987         if (nt->parms.collect_md) {
1988                 if (rtnl_dereference(ign->collect_md_tun))
1989                         return -EEXIST;
1990         } else {
1991                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1992                         return -EEXIST;
1993         }
1994
1995         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1996         if (!err) {
1997                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1998                 ip6gre_tunnel_link_md(ign, nt);
1999                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2000         }
2001         return err;
2002 }
2003
2004 static struct ip6_tnl *
2005 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2006                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2007                          struct netlink_ext_ack *extack)
2008 {
2009         struct ip6_tnl *t, *nt = netdev_priv(dev);
2010         struct net *net = nt->net;
2011         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2012         struct ip_tunnel_encap ipencap;
2013
2014         if (dev == ign->fb_tunnel_dev)
2015                 return ERR_PTR(-EINVAL);
2016
2017         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2018                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2019
2020                 if (err < 0)
2021                         return ERR_PTR(err);
2022         }
2023
2024         ip6gre_netlink_parms(data, p_p);
2025
2026         t = ip6gre_tunnel_locate(net, p_p, 0);
2027
2028         if (t) {
2029                 if (t->dev != dev)
2030                         return ERR_PTR(-EEXIST);
2031         } else {
2032                 t = nt;
2033         }
2034
2035         return t;
2036 }
2037
2038 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2039                              struct nlattr *data[],
2040                              struct netlink_ext_ack *extack)
2041 {
2042         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2043         struct __ip6_tnl_parm p;
2044         struct ip6_tnl *t;
2045
2046         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2047         if (IS_ERR(t))
2048                 return PTR_ERR(t);
2049
2050         ip6gre_tunnel_unlink_md(ign, t);
2051         ip6gre_tunnel_unlink(ign, t);
2052         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2053         ip6gre_tunnel_link_md(ign, t);
2054         ip6gre_tunnel_link(ign, t);
2055         return 0;
2056 }
2057
2058 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2059 {
2060         struct net *net = dev_net(dev);
2061         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2062
2063         if (dev != ign->fb_tunnel_dev)
2064                 unregister_netdevice_queue(dev, head);
2065 }
2066
2067 static size_t ip6gre_get_size(const struct net_device *dev)
2068 {
2069         return
2070                 /* IFLA_GRE_LINK */
2071                 nla_total_size(4) +
2072                 /* IFLA_GRE_IFLAGS */
2073                 nla_total_size(2) +
2074                 /* IFLA_GRE_OFLAGS */
2075                 nla_total_size(2) +
2076                 /* IFLA_GRE_IKEY */
2077                 nla_total_size(4) +
2078                 /* IFLA_GRE_OKEY */
2079                 nla_total_size(4) +
2080                 /* IFLA_GRE_LOCAL */
2081                 nla_total_size(sizeof(struct in6_addr)) +
2082                 /* IFLA_GRE_REMOTE */
2083                 nla_total_size(sizeof(struct in6_addr)) +
2084                 /* IFLA_GRE_TTL */
2085                 nla_total_size(1) +
2086                 /* IFLA_GRE_ENCAP_LIMIT */
2087                 nla_total_size(1) +
2088                 /* IFLA_GRE_FLOWINFO */
2089                 nla_total_size(4) +
2090                 /* IFLA_GRE_FLAGS */
2091                 nla_total_size(4) +
2092                 /* IFLA_GRE_ENCAP_TYPE */
2093                 nla_total_size(2) +
2094                 /* IFLA_GRE_ENCAP_FLAGS */
2095                 nla_total_size(2) +
2096                 /* IFLA_GRE_ENCAP_SPORT */
2097                 nla_total_size(2) +
2098                 /* IFLA_GRE_ENCAP_DPORT */
2099                 nla_total_size(2) +
2100                 /* IFLA_GRE_COLLECT_METADATA */
2101                 nla_total_size(0) +
2102                 /* IFLA_GRE_FWMARK */
2103                 nla_total_size(4) +
2104                 /* IFLA_GRE_ERSPAN_INDEX */
2105                 nla_total_size(4) +
2106                 0;
2107 }
2108
2109 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2110 {
2111         struct ip6_tnl *t = netdev_priv(dev);
2112         struct __ip6_tnl_parm *p = &t->parms;
2113
2114         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2115             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2116                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2117             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2118                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2119             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2120             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2121             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2122             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2123             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2124             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2125             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2126             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2127             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2128             nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2129                 goto nla_put_failure;
2130
2131         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2132                         t->encap.type) ||
2133             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2134                          t->encap.sport) ||
2135             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2136                          t->encap.dport) ||
2137             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2138                         t->encap.flags))
2139                 goto nla_put_failure;
2140
2141         if (p->collect_md) {
2142                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2143                         goto nla_put_failure;
2144         }
2145
2146         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2147                 goto nla_put_failure;
2148
2149         if (p->erspan_ver == 1) {
2150                 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2151                         goto nla_put_failure;
2152         } else if (p->erspan_ver == 2) {
2153                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2154                         goto nla_put_failure;
2155                 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2156                         goto nla_put_failure;
2157         }
2158
2159         return 0;
2160
2161 nla_put_failure:
2162         return -EMSGSIZE;
2163 }
2164
2165 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2166         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2167         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2168         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2169         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2170         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2171         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2172         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2173         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2174         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2175         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2176         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2177         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2178         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2179         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2180         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2181         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2182         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2183         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2184         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2185         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2186         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2187 };
2188
2189 static void ip6erspan_tap_setup(struct net_device *dev)
2190 {
2191         ether_setup(dev);
2192
2193         dev->netdev_ops = &ip6erspan_netdev_ops;
2194         dev->needs_free_netdev = true;
2195         dev->priv_destructor = ip6gre_dev_free;
2196
2197         dev->features |= NETIF_F_NETNS_LOCAL;
2198         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2199         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2200         netif_keep_dst(dev);
2201 }
2202
2203 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2204                              struct nlattr *tb[], struct nlattr *data[],
2205                              struct netlink_ext_ack *extack)
2206 {
2207         struct ip6_tnl *nt = netdev_priv(dev);
2208         struct net *net = dev_net(dev);
2209         struct ip6gre_net *ign;
2210         int err;
2211
2212         ip6gre_netlink_parms(data, &nt->parms);
2213         ign = net_generic(net, ip6gre_net_id);
2214
2215         if (nt->parms.collect_md) {
2216                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2217                         return -EEXIST;
2218         } else {
2219                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2220                         return -EEXIST;
2221         }
2222
2223         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2224         if (!err) {
2225                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2226                 ip6erspan_tunnel_link_md(ign, nt);
2227                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2228         }
2229         return err;
2230 }
2231
2232 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2233 {
2234         ip6gre_tnl_link_config_common(t);
2235         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2236 }
2237
2238 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2239                                 const struct __ip6_tnl_parm *p, int set_mtu)
2240 {
2241         ip6gre_tnl_copy_tnl_parm(t, p);
2242         ip6erspan_tnl_link_config(t, set_mtu);
2243         return 0;
2244 }
2245
2246 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2247                                 struct nlattr *data[],
2248                                 struct netlink_ext_ack *extack)
2249 {
2250         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2251         struct __ip6_tnl_parm p;
2252         struct ip6_tnl *t;
2253
2254         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2255         if (IS_ERR(t))
2256                 return PTR_ERR(t);
2257
2258         ip6gre_tunnel_unlink_md(ign, t);
2259         ip6gre_tunnel_unlink(ign, t);
2260         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2261         ip6erspan_tunnel_link_md(ign, t);
2262         ip6gre_tunnel_link(ign, t);
2263         return 0;
2264 }
2265
2266 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2267         .kind           = "ip6gre",
2268         .maxtype        = IFLA_GRE_MAX,
2269         .policy         = ip6gre_policy,
2270         .priv_size      = sizeof(struct ip6_tnl),
2271         .setup          = ip6gre_tunnel_setup,
2272         .validate       = ip6gre_tunnel_validate,
2273         .newlink        = ip6gre_newlink,
2274         .changelink     = ip6gre_changelink,
2275         .dellink        = ip6gre_dellink,
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 ip6gre_tap_ops __read_mostly = {
2282         .kind           = "ip6gretap",
2283         .maxtype        = IFLA_GRE_MAX,
2284         .policy         = ip6gre_policy,
2285         .priv_size      = sizeof(struct ip6_tnl),
2286         .setup          = ip6gre_tap_setup,
2287         .validate       = ip6gre_tap_validate,
2288         .newlink        = ip6gre_newlink,
2289         .changelink     = ip6gre_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 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2296         .kind           = "ip6erspan",
2297         .maxtype        = IFLA_GRE_MAX,
2298         .policy         = ip6gre_policy,
2299         .priv_size      = sizeof(struct ip6_tnl),
2300         .setup          = ip6erspan_tap_setup,
2301         .validate       = ip6erspan_tap_validate,
2302         .newlink        = ip6erspan_newlink,
2303         .changelink     = ip6erspan_changelink,
2304         .get_size       = ip6gre_get_size,
2305         .fill_info      = ip6gre_fill_info,
2306         .get_link_net   = ip6_tnl_get_link_net,
2307 };
2308
2309 /*
2310  *      And now the modules code and kernel interface.
2311  */
2312
2313 static int __init ip6gre_init(void)
2314 {
2315         int err;
2316
2317         pr_info("GRE over IPv6 tunneling driver\n");
2318
2319         err = register_pernet_device(&ip6gre_net_ops);
2320         if (err < 0)
2321                 return err;
2322
2323         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2324         if (err < 0) {
2325                 pr_info("%s: can't add protocol\n", __func__);
2326                 goto add_proto_failed;
2327         }
2328
2329         err = rtnl_link_register(&ip6gre_link_ops);
2330         if (err < 0)
2331                 goto rtnl_link_failed;
2332
2333         err = rtnl_link_register(&ip6gre_tap_ops);
2334         if (err < 0)
2335                 goto tap_ops_failed;
2336
2337         err = rtnl_link_register(&ip6erspan_tap_ops);
2338         if (err < 0)
2339                 goto erspan_link_failed;
2340
2341 out:
2342         return err;
2343
2344 erspan_link_failed:
2345         rtnl_link_unregister(&ip6gre_tap_ops);
2346 tap_ops_failed:
2347         rtnl_link_unregister(&ip6gre_link_ops);
2348 rtnl_link_failed:
2349         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2350 add_proto_failed:
2351         unregister_pernet_device(&ip6gre_net_ops);
2352         goto out;
2353 }
2354
2355 static void __exit ip6gre_fini(void)
2356 {
2357         rtnl_link_unregister(&ip6gre_tap_ops);
2358         rtnl_link_unregister(&ip6gre_link_ops);
2359         rtnl_link_unregister(&ip6erspan_tap_ops);
2360         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2361         unregister_pernet_device(&ip6gre_net_ops);
2362 }
2363
2364 module_init(ip6gre_init);
2365 module_exit(ip6gre_fini);
2366 MODULE_LICENSE("GPL");
2367 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2368 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2369 MODULE_ALIAS_RTNL_LINK("ip6gre");
2370 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2371 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2372 MODULE_ALIAS_NETDEV("ip6gre0");