GNU Linux-libre 4.19.242-gnu1
[releases.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         struct net_device *ndev;
136
137         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
138                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
139                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
140                     key != t->parms.i_key ||
141                     !(t->dev->flags & IFF_UP))
142                         continue;
143
144                 if (t->dev->type != ARPHRD_IP6GRE &&
145                     t->dev->type != dev_type)
146                         continue;
147
148                 score = 0;
149                 if (t->parms.link != link)
150                         score |= 1;
151                 if (t->dev->type != dev_type)
152                         score |= 2;
153                 if (score == 0)
154                         return t;
155
156                 if (score < cand_score) {
157                         cand = t;
158                         cand_score = score;
159                 }
160         }
161
162         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
163                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
164                     key != t->parms.i_key ||
165                     !(t->dev->flags & IFF_UP))
166                         continue;
167
168                 if (t->dev->type != ARPHRD_IP6GRE &&
169                     t->dev->type != dev_type)
170                         continue;
171
172                 score = 0;
173                 if (t->parms.link != link)
174                         score |= 1;
175                 if (t->dev->type != dev_type)
176                         score |= 2;
177                 if (score == 0)
178                         return t;
179
180                 if (score < cand_score) {
181                         cand = t;
182                         cand_score = score;
183                 }
184         }
185
186         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
187                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
188                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
189                                  !ipv6_addr_is_multicast(local))) ||
190                     key != t->parms.i_key ||
191                     !(t->dev->flags & IFF_UP))
192                         continue;
193
194                 if (t->dev->type != ARPHRD_IP6GRE &&
195                     t->dev->type != dev_type)
196                         continue;
197
198                 score = 0;
199                 if (t->parms.link != link)
200                         score |= 1;
201                 if (t->dev->type != dev_type)
202                         score |= 2;
203                 if (score == 0)
204                         return t;
205
206                 if (score < cand_score) {
207                         cand = t;
208                         cand_score = score;
209                 }
210         }
211
212         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
213                 if (t->parms.i_key != key ||
214                     !(t->dev->flags & IFF_UP))
215                         continue;
216
217                 if (t->dev->type != ARPHRD_IP6GRE &&
218                     t->dev->type != dev_type)
219                         continue;
220
221                 score = 0;
222                 if (t->parms.link != link)
223                         score |= 1;
224                 if (t->dev->type != dev_type)
225                         score |= 2;
226                 if (score == 0)
227                         return t;
228
229                 if (score < cand_score) {
230                         cand = t;
231                         cand_score = score;
232                 }
233         }
234
235         if (cand)
236                 return cand;
237
238         if (gre_proto == htons(ETH_P_ERSPAN) ||
239             gre_proto == htons(ETH_P_ERSPAN2))
240                 t = rcu_dereference(ign->collect_md_tun_erspan);
241         else
242                 t = rcu_dereference(ign->collect_md_tun);
243
244         if (t && t->dev->flags & IFF_UP)
245                 return t;
246
247         ndev = READ_ONCE(ign->fb_tunnel_dev);
248         if (ndev && ndev->flags & IFF_UP)
249                 return netdev_priv(ndev);
250
251         return NULL;
252 }
253
254 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
255                 const struct __ip6_tnl_parm *p)
256 {
257         const struct in6_addr *remote = &p->raddr;
258         const struct in6_addr *local = &p->laddr;
259         unsigned int h = HASH_KEY(p->i_key);
260         int prio = 0;
261
262         if (!ipv6_addr_any(local))
263                 prio |= 1;
264         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
265                 prio |= 2;
266                 h ^= HASH_ADDR(remote);
267         }
268
269         return &ign->tunnels[prio][h];
270 }
271
272 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
273 {
274         if (t->parms.collect_md)
275                 rcu_assign_pointer(ign->collect_md_tun, t);
276 }
277
278 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
279 {
280         if (t->parms.collect_md)
281                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
282 }
283
284 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
285 {
286         if (t->parms.collect_md)
287                 rcu_assign_pointer(ign->collect_md_tun, NULL);
288 }
289
290 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
291                                        struct ip6_tnl *t)
292 {
293         if (t->parms.collect_md)
294                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
295 }
296
297 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
298                 const struct ip6_tnl *t)
299 {
300         return __ip6gre_bucket(ign, &t->parms);
301 }
302
303 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
304 {
305         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
306
307         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
308         rcu_assign_pointer(*tp, t);
309 }
310
311 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
312 {
313         struct ip6_tnl __rcu **tp;
314         struct ip6_tnl *iter;
315
316         for (tp = ip6gre_bucket(ign, t);
317              (iter = rtnl_dereference(*tp)) != NULL;
318              tp = &iter->next) {
319                 if (t == iter) {
320                         rcu_assign_pointer(*tp, t->next);
321                         break;
322                 }
323         }
324 }
325
326 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
327                                            const struct __ip6_tnl_parm *parms,
328                                            int type)
329 {
330         const struct in6_addr *remote = &parms->raddr;
331         const struct in6_addr *local = &parms->laddr;
332         __be32 key = parms->i_key;
333         int link = parms->link;
334         struct ip6_tnl *t;
335         struct ip6_tnl __rcu **tp;
336         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
337
338         for (tp = __ip6gre_bucket(ign, parms);
339              (t = rtnl_dereference(*tp)) != NULL;
340              tp = &t->next)
341                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
342                     ipv6_addr_equal(remote, &t->parms.raddr) &&
343                     key == t->parms.i_key &&
344                     link == t->parms.link &&
345                     type == t->dev->type)
346                         break;
347
348         return t;
349 }
350
351 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
352                 const struct __ip6_tnl_parm *parms, int create)
353 {
354         struct ip6_tnl *t, *nt;
355         struct net_device *dev;
356         char name[IFNAMSIZ];
357         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
358
359         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
360         if (t && create)
361                 return NULL;
362         if (t || !create)
363                 return t;
364
365         if (parms->name[0]) {
366                 if (!dev_valid_name(parms->name))
367                         return NULL;
368                 strlcpy(name, parms->name, IFNAMSIZ);
369         } else {
370                 strcpy(name, "ip6gre%d");
371         }
372         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
373                            ip6gre_tunnel_setup);
374         if (!dev)
375                 return NULL;
376
377         dev_net_set(dev, net);
378
379         nt = netdev_priv(dev);
380         nt->parms = *parms;
381         dev->rtnl_link_ops = &ip6gre_link_ops;
382
383         nt->dev = dev;
384         nt->net = dev_net(dev);
385
386         if (register_netdevice(dev) < 0)
387                 goto failed_free;
388
389         ip6gre_tnl_link_config(nt, 1);
390
391         /* Can use a lockless transmit, unless we generate output sequences */
392         if (!(nt->parms.o_flags & TUNNEL_SEQ))
393                 dev->features |= NETIF_F_LLTX;
394
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         if (ign->fb_tunnel_dev == dev)
422                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
423         dst_cache_reset(&t->dst_cache);
424         dev_put(dev);
425 }
426
427
428 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
429                        u8 type, u8 code, int offset, __be32 info)
430 {
431         struct net *net = dev_net(skb->dev);
432         const struct gre_base_hdr *greh;
433         const struct ipv6hdr *ipv6h;
434         int grehlen = sizeof(*greh);
435         struct ip6_tnl *t;
436         int key_off = 0;
437         __be16 flags;
438         __be32 key;
439
440         if (!pskb_may_pull(skb, offset + grehlen))
441                 return;
442         greh = (const struct gre_base_hdr *)(skb->data + offset);
443         flags = greh->flags;
444         if (flags & (GRE_VERSION | GRE_ROUTING))
445                 return;
446         if (flags & GRE_CSUM)
447                 grehlen += 4;
448         if (flags & GRE_KEY) {
449                 key_off = grehlen + offset;
450                 grehlen += 4;
451         }
452
453         if (!pskb_may_pull(skb, offset + grehlen))
454                 return;
455         ipv6h = (const struct ipv6hdr *)skb->data;
456         greh = (const struct gre_base_hdr *)(skb->data + offset);
457         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
458
459         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
460                                  key, greh->protocol);
461         if (!t)
462                 return;
463
464         switch (type) {
465                 struct ipv6_tlv_tnl_enc_lim *tel;
466                 __u32 teli;
467         case ICMPV6_DEST_UNREACH:
468                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
469                                     t->parms.name);
470                 if (code != ICMPV6_PORT_UNREACH)
471                         break;
472                 return;
473         case ICMPV6_TIME_EXCEED:
474                 if (code == ICMPV6_EXC_HOPLIMIT) {
475                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
476                                             t->parms.name);
477                         break;
478                 }
479                 return;
480         case ICMPV6_PARAMPROB:
481                 teli = 0;
482                 if (code == ICMPV6_HDR_FIELD)
483                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
484
485                 if (teli && teli == be32_to_cpu(info) - 2) {
486                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
487                         if (tel->encap_limit == 0) {
488                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
489                                                     t->parms.name);
490                         }
491                 } else {
492                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
493                                             t->parms.name);
494                 }
495                 return;
496         case ICMPV6_PKT_TOOBIG:
497                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
498                 return;
499         case NDISC_REDIRECT:
500                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
501                              sock_net_uid(net, NULL));
502                 return;
503         }
504
505         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
506                 t->err_count++;
507         else
508                 t->err_count = 1;
509         t->err_time = jiffies;
510 }
511
512 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
513 {
514         const struct ipv6hdr *ipv6h;
515         struct ip6_tnl *tunnel;
516
517         ipv6h = ipv6_hdr(skb);
518         tunnel = ip6gre_tunnel_lookup(skb->dev,
519                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
520                                       tpi->proto);
521         if (tunnel) {
522                 if (tunnel->parms.collect_md) {
523                         struct metadata_dst *tun_dst;
524                         __be64 tun_id;
525                         __be16 flags;
526
527                         flags = tpi->flags;
528                         tun_id = key32_to_tunnel_id(tpi->key);
529
530                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
531                         if (!tun_dst)
532                                 return PACKET_REJECT;
533
534                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
535                 } else {
536                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
537                 }
538
539                 return PACKET_RCVD;
540         }
541
542         return PACKET_REJECT;
543 }
544
545 static int ip6erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
546                          int gre_hdr_len)
547 {
548         struct erspan_base_hdr *ershdr;
549         const struct ipv6hdr *ipv6h;
550         struct erspan_md2 *md2;
551         struct ip6_tnl *tunnel;
552         u8 ver;
553
554         ipv6h = ipv6_hdr(skb);
555         ershdr = (struct erspan_base_hdr *)skb->data;
556         ver = ershdr->ver;
557
558         tunnel = ip6gre_tunnel_lookup(skb->dev,
559                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
560                                       tpi->proto);
561         if (tunnel) {
562                 int len = erspan_hdr_len(ver);
563
564                 if (unlikely(!pskb_may_pull(skb, len)))
565                         return PACKET_REJECT;
566
567                 if (__iptunnel_pull_header(skb, len,
568                                            htons(ETH_P_TEB),
569                                            false, false) < 0)
570                         return PACKET_REJECT;
571
572                 if (tunnel->parms.collect_md) {
573                         struct erspan_metadata *pkt_md, *md;
574                         struct metadata_dst *tun_dst;
575                         struct ip_tunnel_info *info;
576                         unsigned char *gh;
577                         __be64 tun_id;
578                         __be16 flags;
579
580                         tpi->flags |= TUNNEL_KEY;
581                         flags = tpi->flags;
582                         tun_id = key32_to_tunnel_id(tpi->key);
583
584                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
585                                                   sizeof(*md));
586                         if (!tun_dst)
587                                 return PACKET_REJECT;
588
589                         /* skb can be uncloned in __iptunnel_pull_header, so
590                          * old pkt_md is no longer valid and we need to reset
591                          * it
592                          */
593                         gh = skb_network_header(skb) +
594                              skb_network_header_len(skb);
595                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
596                                                             sizeof(*ershdr));
597                         info = &tun_dst->u.tun_info;
598                         md = ip_tunnel_info_opts(info);
599                         md->version = ver;
600                         md2 = &md->u.md2;
601                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
602                                                        ERSPAN_V2_MDSIZE);
603                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
604                         info->options_len = sizeof(*md);
605
606                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
607
608                 } else {
609                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
610                 }
611
612                 return PACKET_RCVD;
613         }
614
615         return PACKET_REJECT;
616 }
617
618 static int gre_rcv(struct sk_buff *skb)
619 {
620         struct tnl_ptk_info tpi;
621         bool csum_err = false;
622         int hdr_len;
623
624         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
625         if (hdr_len < 0)
626                 goto drop;
627
628         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
629                 goto drop;
630
631         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
632                      tpi.proto == htons(ETH_P_ERSPAN2))) {
633                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
634                         return 0;
635                 goto out;
636         }
637
638         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
639                 return 0;
640
641 out:
642         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
643 drop:
644         kfree_skb(skb);
645         return 0;
646 }
647
648 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
649 {
650         return iptunnel_handle_offloads(skb,
651                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
652 }
653
654 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
655                                      struct net_device *dev,
656                                      struct flowi6 *fl6, __u8 *dsfield,
657                                      int *encap_limit)
658 {
659         const struct iphdr *iph = ip_hdr(skb);
660         struct ip6_tnl *t = netdev_priv(dev);
661
662         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
663                 *encap_limit = t->parms.encap_limit;
664
665         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
666
667         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
668                 *dsfield = ipv4_get_dsfield(iph);
669         else
670                 *dsfield = ip6_tclass(t->parms.flowinfo);
671
672         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
673                 fl6->flowi6_mark = skb->mark;
674         else
675                 fl6->flowi6_mark = t->parms.fwmark;
676
677         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
678 }
679
680 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
681                                     struct net_device *dev,
682                                     struct flowi6 *fl6, __u8 *dsfield,
683                                     int *encap_limit)
684 {
685         struct ipv6hdr *ipv6h;
686         struct ip6_tnl *t = netdev_priv(dev);
687         __u16 offset;
688
689         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
690         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
691         ipv6h = ipv6_hdr(skb);
692
693         if (offset > 0) {
694                 struct ipv6_tlv_tnl_enc_lim *tel;
695
696                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
697                 if (tel->encap_limit == 0) {
698                         icmpv6_send(skb, ICMPV6_PARAMPROB,
699                                     ICMPV6_HDR_FIELD, offset + 2);
700                         return -1;
701                 }
702                 *encap_limit = tel->encap_limit - 1;
703         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
704                 *encap_limit = t->parms.encap_limit;
705         }
706
707         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
708
709         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
710                 *dsfield = ipv6_get_dsfield(ipv6h);
711         else
712                 *dsfield = ip6_tclass(t->parms.flowinfo);
713
714         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
715                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
716
717         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
718                 fl6->flowi6_mark = skb->mark;
719         else
720                 fl6->flowi6_mark = t->parms.fwmark;
721
722         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
723
724         return 0;
725 }
726
727 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
728                                struct net_device *dev, __u8 dsfield,
729                                struct flowi6 *fl6, int encap_limit,
730                                __u32 *pmtu, __be16 proto)
731 {
732         struct ip6_tnl *tunnel = netdev_priv(dev);
733         __be16 protocol;
734
735         if (dev->type == ARPHRD_ETHER)
736                 IPCB(skb)->flags = 0;
737
738         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
739                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
740         else
741                 fl6->daddr = tunnel->parms.raddr;
742
743         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
744                 return -ENOMEM;
745
746         /* Push GRE header. */
747         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
748
749         if (tunnel->parms.collect_md) {
750                 struct ip_tunnel_info *tun_info;
751                 const struct ip_tunnel_key *key;
752                 __be16 flags;
753                 int tun_hlen;
754
755                 tun_info = skb_tunnel_info(skb);
756                 if (unlikely(!tun_info ||
757                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
758                              ip_tunnel_info_af(tun_info) != AF_INET6))
759                         return -EINVAL;
760
761                 key = &tun_info->key;
762                 memset(fl6, 0, sizeof(*fl6));
763                 fl6->flowi6_proto = IPPROTO_GRE;
764                 fl6->daddr = key->u.ipv6.dst;
765                 fl6->flowlabel = key->label;
766                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
767
768                 dsfield = key->tos;
769                 flags = key->tun_flags &
770                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
771                 tun_hlen = gre_calc_hlen(flags);
772
773                 gre_build_header(skb, tun_hlen,
774                                  flags, protocol,
775                                  tunnel_id_to_key32(tun_info->key.tun_id),
776                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
777                                                       : 0);
778
779         } else {
780                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
781                         tunnel->o_seqno++;
782
783                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
784                                  protocol, tunnel->parms.o_key,
785                                  htonl(tunnel->o_seqno));
786         }
787
788         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
789                             NEXTHDR_GRE);
790 }
791
792 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
793 {
794         struct ip6_tnl *t = netdev_priv(dev);
795         int encap_limit = -1;
796         struct flowi6 fl6;
797         __u8 dsfield = 0;
798         __u32 mtu;
799         int err;
800
801         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
802
803         if (!t->parms.collect_md)
804                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
805                                          &dsfield, &encap_limit);
806
807         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
808         if (err)
809                 return -1;
810
811         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
812                           skb->protocol);
813         if (err != 0) {
814                 /* XXX: send ICMP error even if DF is not set. */
815                 if (err == -EMSGSIZE)
816                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
817                                   htonl(mtu));
818                 return -1;
819         }
820
821         return 0;
822 }
823
824 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
825 {
826         struct ip6_tnl *t = netdev_priv(dev);
827         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
828         int encap_limit = -1;
829         struct flowi6 fl6;
830         __u8 dsfield = 0;
831         __u32 mtu;
832         int err;
833
834         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
835                 return -1;
836
837         if (!t->parms.collect_md &&
838             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
839                 return -1;
840
841         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
842                 return -1;
843
844         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
845                           &mtu, skb->protocol);
846         if (err != 0) {
847                 if (err == -EMSGSIZE)
848                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
849                 return -1;
850         }
851
852         return 0;
853 }
854
855 /**
856  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
857  *   @t: the outgoing tunnel device
858  *   @hdr: IPv6 header from the incoming packet
859  *
860  * Description:
861  *   Avoid trivial tunneling loop by checking that tunnel exit-point
862  *   doesn't match source of incoming packet.
863  *
864  * Return:
865  *   1 if conflict,
866  *   0 else
867  **/
868
869 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
870         const struct ipv6hdr *hdr)
871 {
872         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
873 }
874
875 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
876 {
877         struct ip6_tnl *t = netdev_priv(dev);
878         int encap_limit = -1;
879         struct flowi6 fl6;
880         __u32 mtu;
881         int err;
882
883         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
884                 encap_limit = t->parms.encap_limit;
885
886         if (!t->parms.collect_md)
887                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
888
889         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
890         if (err)
891                 return err;
892
893         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
894
895         return err;
896 }
897
898 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
899         struct net_device *dev)
900 {
901         struct ip6_tnl *t = netdev_priv(dev);
902         struct net_device_stats *stats = &t->dev->stats;
903         int ret;
904
905         if (!pskb_inet_may_pull(skb))
906                 goto tx_err;
907
908         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
909                 goto tx_err;
910
911         switch (skb->protocol) {
912         case htons(ETH_P_IP):
913                 ret = ip6gre_xmit_ipv4(skb, dev);
914                 break;
915         case htons(ETH_P_IPV6):
916                 ret = ip6gre_xmit_ipv6(skb, dev);
917                 break;
918         default:
919                 ret = ip6gre_xmit_other(skb, dev);
920                 break;
921         }
922
923         if (ret < 0)
924                 goto tx_err;
925
926         return NETDEV_TX_OK;
927
928 tx_err:
929         stats->tx_errors++;
930         stats->tx_dropped++;
931         kfree_skb(skb);
932         return NETDEV_TX_OK;
933 }
934
935 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
936                                          struct net_device *dev)
937 {
938         struct ip6_tnl *t = netdev_priv(dev);
939         struct dst_entry *dst = skb_dst(skb);
940         struct net_device_stats *stats;
941         bool truncate = false;
942         int encap_limit = -1;
943         __u8 dsfield = false;
944         struct flowi6 fl6;
945         int err = -EINVAL;
946         __be16 proto;
947         __u32 mtu;
948         int nhoff;
949         int thoff;
950
951         if (!pskb_inet_may_pull(skb))
952                 goto tx_err;
953
954         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
955                 goto tx_err;
956
957         if (gre_handle_offloads(skb, false))
958                 goto tx_err;
959
960         if (skb->len > dev->mtu + dev->hard_header_len) {
961                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
962                 truncate = true;
963         }
964
965         nhoff = skb_network_header(skb) - skb_mac_header(skb);
966         if (skb->protocol == htons(ETH_P_IP) &&
967             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
968                 truncate = true;
969
970         thoff = skb_transport_header(skb) - skb_mac_header(skb);
971         if (skb->protocol == htons(ETH_P_IPV6) &&
972             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
973                 truncate = true;
974
975         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
976                 goto tx_err;
977
978         t->parms.o_flags &= ~TUNNEL_KEY;
979         IPCB(skb)->flags = 0;
980
981         /* For collect_md mode, derive fl6 from the tunnel key,
982          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
983          */
984         if (t->parms.collect_md) {
985                 struct ip_tunnel_info *tun_info;
986                 const struct ip_tunnel_key *key;
987                 struct erspan_metadata *md;
988                 __be32 tun_id;
989
990                 tun_info = skb_tunnel_info(skb);
991                 if (unlikely(!tun_info ||
992                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
993                              ip_tunnel_info_af(tun_info) != AF_INET6))
994                         goto tx_err;
995
996                 key = &tun_info->key;
997                 memset(&fl6, 0, sizeof(fl6));
998                 fl6.flowi6_proto = IPPROTO_GRE;
999                 fl6.daddr = key->u.ipv6.dst;
1000                 fl6.flowlabel = key->label;
1001                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1002
1003                 dsfield = key->tos;
1004                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1005                         goto tx_err;
1006                 if (tun_info->options_len < sizeof(*md))
1007                         goto tx_err;
1008                 md = ip_tunnel_info_opts(tun_info);
1009
1010                 tun_id = tunnel_id_to_key32(key->tun_id);
1011                 if (md->version == 1) {
1012                         erspan_build_header(skb,
1013                                             ntohl(tun_id),
1014                                             ntohl(md->u.index), truncate,
1015                                             false);
1016                 } else if (md->version == 2) {
1017                         erspan_build_header_v2(skb,
1018                                                ntohl(tun_id),
1019                                                md->u.md2.dir,
1020                                                get_hwid(&md->u.md2),
1021                                                truncate, false);
1022                 } else {
1023                         goto tx_err;
1024                 }
1025         } else {
1026                 switch (skb->protocol) {
1027                 case htons(ETH_P_IP):
1028                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1029                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1030                                                  &dsfield, &encap_limit);
1031                         break;
1032                 case htons(ETH_P_IPV6):
1033                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1034                                 goto tx_err;
1035                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1036                                                      &dsfield, &encap_limit))
1037                                 goto tx_err;
1038                         break;
1039                 default:
1040                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1041                         break;
1042                 }
1043
1044                 if (t->parms.erspan_ver == 1)
1045                         erspan_build_header(skb, ntohl(t->parms.o_key),
1046                                             t->parms.index,
1047                                             truncate, false);
1048                 else if (t->parms.erspan_ver == 2)
1049                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1050                                                t->parms.dir,
1051                                                t->parms.hwid,
1052                                                truncate, false);
1053                 else
1054                         goto tx_err;
1055
1056                 fl6.daddr = t->parms.raddr;
1057         }
1058
1059         /* Push GRE header. */
1060         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1061                                            : htons(ETH_P_ERSPAN2);
1062         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1063
1064         /* TooBig packet may have updated dst->dev's mtu */
1065         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1066                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1067
1068         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1069                            NEXTHDR_GRE);
1070         if (err != 0) {
1071                 /* XXX: send ICMP error even if DF is not set. */
1072                 if (err == -EMSGSIZE) {
1073                         if (skb->protocol == htons(ETH_P_IP))
1074                                 icmp_send(skb, ICMP_DEST_UNREACH,
1075                                           ICMP_FRAG_NEEDED, htonl(mtu));
1076                         else
1077                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1078                 }
1079
1080                 goto tx_err;
1081         }
1082         return NETDEV_TX_OK;
1083
1084 tx_err:
1085         stats = &t->dev->stats;
1086         stats->tx_errors++;
1087         stats->tx_dropped++;
1088         kfree_skb(skb);
1089         return NETDEV_TX_OK;
1090 }
1091
1092 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1093 {
1094         struct net_device *dev = t->dev;
1095         struct __ip6_tnl_parm *p = &t->parms;
1096         struct flowi6 *fl6 = &t->fl.u.ip6;
1097
1098         if (dev->type != ARPHRD_ETHER) {
1099                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1100                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1101         }
1102
1103         /* Set up flowi template */
1104         fl6->saddr = p->laddr;
1105         fl6->daddr = p->raddr;
1106         fl6->flowi6_oif = p->link;
1107         fl6->flowlabel = 0;
1108         fl6->flowi6_proto = IPPROTO_GRE;
1109
1110         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1111                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1112         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1113                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1114
1115         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1116         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1117
1118         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1119                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1120                 dev->flags |= IFF_POINTOPOINT;
1121         else
1122                 dev->flags &= ~IFF_POINTOPOINT;
1123 }
1124
1125 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1126                                          int t_hlen)
1127 {
1128         const struct __ip6_tnl_parm *p = &t->parms;
1129         struct net_device *dev = t->dev;
1130
1131         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1132                 int strict = (ipv6_addr_type(&p->raddr) &
1133                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1134
1135                 struct rt6_info *rt = rt6_lookup(t->net,
1136                                                  &p->raddr, &p->laddr,
1137                                                  p->link, NULL, strict);
1138
1139                 if (!rt)
1140                         return;
1141
1142                 if (rt->dst.dev) {
1143                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1144                                                  t_hlen;
1145
1146                         if (t->dev->header_ops)
1147                                 dev->hard_header_len = dst_len;
1148                         else
1149                                 dev->needed_headroom = dst_len;
1150
1151                         if (set_mtu) {
1152                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1153                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1154                                         dev->mtu -= 8;
1155                                 if (dev->type == ARPHRD_ETHER)
1156                                         dev->mtu -= ETH_HLEN;
1157
1158                                 if (dev->mtu < IPV6_MIN_MTU)
1159                                         dev->mtu = IPV6_MIN_MTU;
1160                         }
1161                 }
1162                 ip6_rt_put(rt);
1163         }
1164 }
1165
1166 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1167 {
1168         int t_hlen;
1169
1170         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1171         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1172
1173         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1174
1175         if (tunnel->dev->header_ops)
1176                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1177         else
1178                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1179
1180         return t_hlen;
1181 }
1182
1183 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1184 {
1185         ip6gre_tnl_link_config_common(t);
1186         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1187 }
1188
1189 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1190                                      const struct __ip6_tnl_parm *p)
1191 {
1192         t->parms.laddr = p->laddr;
1193         t->parms.raddr = p->raddr;
1194         t->parms.flags = p->flags;
1195         t->parms.hop_limit = p->hop_limit;
1196         t->parms.encap_limit = p->encap_limit;
1197         t->parms.flowinfo = p->flowinfo;
1198         t->parms.link = p->link;
1199         t->parms.proto = p->proto;
1200         t->parms.i_key = p->i_key;
1201         t->parms.o_key = p->o_key;
1202         t->parms.i_flags = p->i_flags;
1203         t->parms.o_flags = p->o_flags;
1204         t->parms.fwmark = p->fwmark;
1205         t->parms.erspan_ver = p->erspan_ver;
1206         t->parms.index = p->index;
1207         t->parms.dir = p->dir;
1208         t->parms.hwid = p->hwid;
1209         dst_cache_reset(&t->dst_cache);
1210 }
1211
1212 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1213                              int set_mtu)
1214 {
1215         ip6gre_tnl_copy_tnl_parm(t, p);
1216         ip6gre_tnl_link_config(t, set_mtu);
1217         return 0;
1218 }
1219
1220 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1221         const struct ip6_tnl_parm2 *u)
1222 {
1223         p->laddr = u->laddr;
1224         p->raddr = u->raddr;
1225         p->flags = u->flags;
1226         p->hop_limit = u->hop_limit;
1227         p->encap_limit = u->encap_limit;
1228         p->flowinfo = u->flowinfo;
1229         p->link = u->link;
1230         p->i_key = u->i_key;
1231         p->o_key = u->o_key;
1232         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1233         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1234         memcpy(p->name, u->name, sizeof(u->name));
1235 }
1236
1237 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1238         const struct __ip6_tnl_parm *p)
1239 {
1240         u->proto = IPPROTO_GRE;
1241         u->laddr = p->laddr;
1242         u->raddr = p->raddr;
1243         u->flags = p->flags;
1244         u->hop_limit = p->hop_limit;
1245         u->encap_limit = p->encap_limit;
1246         u->flowinfo = p->flowinfo;
1247         u->link = p->link;
1248         u->i_key = p->i_key;
1249         u->o_key = p->o_key;
1250         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1251         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1252         memcpy(u->name, p->name, sizeof(u->name));
1253 }
1254
1255 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1256         struct ifreq *ifr, int cmd)
1257 {
1258         int err = 0;
1259         struct ip6_tnl_parm2 p;
1260         struct __ip6_tnl_parm p1;
1261         struct ip6_tnl *t = netdev_priv(dev);
1262         struct net *net = t->net;
1263         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1264
1265         memset(&p1, 0, sizeof(p1));
1266
1267         switch (cmd) {
1268         case SIOCGETTUNNEL:
1269                 if (dev == ign->fb_tunnel_dev) {
1270                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1271                                 err = -EFAULT;
1272                                 break;
1273                         }
1274                         ip6gre_tnl_parm_from_user(&p1, &p);
1275                         t = ip6gre_tunnel_locate(net, &p1, 0);
1276                         if (!t)
1277                                 t = netdev_priv(dev);
1278                 }
1279                 memset(&p, 0, sizeof(p));
1280                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1281                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1282                         err = -EFAULT;
1283                 break;
1284
1285         case SIOCADDTUNNEL:
1286         case SIOCCHGTUNNEL:
1287                 err = -EPERM;
1288                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1289                         goto done;
1290
1291                 err = -EFAULT;
1292                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1293                         goto done;
1294
1295                 err = -EINVAL;
1296                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1297                         goto done;
1298
1299                 if (!(p.i_flags&GRE_KEY))
1300                         p.i_key = 0;
1301                 if (!(p.o_flags&GRE_KEY))
1302                         p.o_key = 0;
1303
1304                 ip6gre_tnl_parm_from_user(&p1, &p);
1305                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1306
1307                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1308                         if (t) {
1309                                 if (t->dev != dev) {
1310                                         err = -EEXIST;
1311                                         break;
1312                                 }
1313                         } else {
1314                                 t = netdev_priv(dev);
1315
1316                                 ip6gre_tunnel_unlink(ign, t);
1317                                 synchronize_net();
1318                                 ip6gre_tnl_change(t, &p1, 1);
1319                                 ip6gre_tunnel_link(ign, t);
1320                                 netdev_state_change(dev);
1321                         }
1322                 }
1323
1324                 if (t) {
1325                         err = 0;
1326
1327                         memset(&p, 0, sizeof(p));
1328                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1329                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1330                                 err = -EFAULT;
1331                 } else
1332                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1333                 break;
1334
1335         case SIOCDELTUNNEL:
1336                 err = -EPERM;
1337                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1338                         goto done;
1339
1340                 if (dev == ign->fb_tunnel_dev) {
1341                         err = -EFAULT;
1342                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1343                                 goto done;
1344                         err = -ENOENT;
1345                         ip6gre_tnl_parm_from_user(&p1, &p);
1346                         t = ip6gre_tunnel_locate(net, &p1, 0);
1347                         if (!t)
1348                                 goto done;
1349                         err = -EPERM;
1350                         if (t == netdev_priv(ign->fb_tunnel_dev))
1351                                 goto done;
1352                         dev = t->dev;
1353                 }
1354                 unregister_netdevice(dev);
1355                 err = 0;
1356                 break;
1357
1358         default:
1359                 err = -EINVAL;
1360         }
1361
1362 done:
1363         return err;
1364 }
1365
1366 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1367                          unsigned short type, const void *daddr,
1368                          const void *saddr, unsigned int len)
1369 {
1370         struct ip6_tnl *t = netdev_priv(dev);
1371         struct ipv6hdr *ipv6h;
1372         __be16 *p;
1373
1374         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1375         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1376                                                   t->fl.u.ip6.flowlabel,
1377                                                   true, &t->fl.u.ip6));
1378         ipv6h->hop_limit = t->parms.hop_limit;
1379         ipv6h->nexthdr = NEXTHDR_GRE;
1380         ipv6h->saddr = t->parms.laddr;
1381         ipv6h->daddr = t->parms.raddr;
1382
1383         p = (__be16 *)(ipv6h + 1);
1384         p[0] = t->parms.o_flags;
1385         p[1] = htons(type);
1386
1387         /*
1388          *      Set the source hardware address.
1389          */
1390
1391         if (saddr)
1392                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1393         if (daddr)
1394                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1395         if (!ipv6_addr_any(&ipv6h->daddr))
1396                 return t->hlen;
1397
1398         return -t->hlen;
1399 }
1400
1401 static const struct header_ops ip6gre_header_ops = {
1402         .create = ip6gre_header,
1403 };
1404
1405 static const struct net_device_ops ip6gre_netdev_ops = {
1406         .ndo_init               = ip6gre_tunnel_init,
1407         .ndo_uninit             = ip6gre_tunnel_uninit,
1408         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1409         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1410         .ndo_change_mtu         = ip6_tnl_change_mtu,
1411         .ndo_get_stats64        = ip_tunnel_get_stats64,
1412         .ndo_get_iflink         = ip6_tnl_get_iflink,
1413 };
1414
1415 static void ip6gre_dev_free(struct net_device *dev)
1416 {
1417         struct ip6_tnl *t = netdev_priv(dev);
1418
1419         gro_cells_destroy(&t->gro_cells);
1420         dst_cache_destroy(&t->dst_cache);
1421         free_percpu(dev->tstats);
1422 }
1423
1424 static void ip6gre_tunnel_setup(struct net_device *dev)
1425 {
1426         dev->netdev_ops = &ip6gre_netdev_ops;
1427         dev->needs_free_netdev = true;
1428         dev->priv_destructor = ip6gre_dev_free;
1429
1430         dev->type = ARPHRD_IP6GRE;
1431
1432         dev->flags |= IFF_NOARP;
1433         dev->addr_len = sizeof(struct in6_addr);
1434         netif_keep_dst(dev);
1435         /* This perm addr will be used as interface identifier by IPv6 */
1436         dev->addr_assign_type = NET_ADDR_RANDOM;
1437         eth_random_addr(dev->perm_addr);
1438 }
1439
1440 #define GRE6_FEATURES (NETIF_F_SG |             \
1441                        NETIF_F_FRAGLIST |       \
1442                        NETIF_F_HIGHDMA |        \
1443                        NETIF_F_HW_CSUM)
1444
1445 static void ip6gre_tnl_init_features(struct net_device *dev)
1446 {
1447         struct ip6_tnl *nt = netdev_priv(dev);
1448
1449         dev->features           |= GRE6_FEATURES;
1450         dev->hw_features        |= GRE6_FEATURES;
1451
1452         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1453                 /* TCP offload with GRE SEQ is not supported, nor
1454                  * can we support 2 levels of outer headers requiring
1455                  * an update.
1456                  */
1457                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1458                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1459                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1460                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1461                 }
1462
1463                 /* Can use a lockless transmit, unless we generate
1464                  * output sequences
1465                  */
1466                 dev->features |= NETIF_F_LLTX;
1467         }
1468 }
1469
1470 static int ip6gre_tunnel_init_common(struct net_device *dev)
1471 {
1472         struct ip6_tnl *tunnel;
1473         int ret;
1474         int t_hlen;
1475
1476         tunnel = netdev_priv(dev);
1477
1478         tunnel->dev = dev;
1479         tunnel->net = dev_net(dev);
1480         strcpy(tunnel->parms.name, dev->name);
1481
1482         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1483         if (!dev->tstats)
1484                 return -ENOMEM;
1485
1486         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1487         if (ret)
1488                 goto cleanup_alloc_pcpu_stats;
1489
1490         ret = gro_cells_init(&tunnel->gro_cells, dev);
1491         if (ret)
1492                 goto cleanup_dst_cache_init;
1493
1494         t_hlen = ip6gre_calc_hlen(tunnel);
1495         dev->mtu = ETH_DATA_LEN - t_hlen;
1496         if (dev->type == ARPHRD_ETHER)
1497                 dev->mtu -= ETH_HLEN;
1498         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1499                 dev->mtu -= 8;
1500
1501         if (tunnel->parms.collect_md) {
1502                 netif_keep_dst(dev);
1503         }
1504         ip6gre_tnl_init_features(dev);
1505
1506         dev_hold(dev);
1507         return 0;
1508
1509 cleanup_dst_cache_init:
1510         dst_cache_destroy(&tunnel->dst_cache);
1511 cleanup_alloc_pcpu_stats:
1512         free_percpu(dev->tstats);
1513         dev->tstats = NULL;
1514         return ret;
1515 }
1516
1517 static int ip6gre_tunnel_init(struct net_device *dev)
1518 {
1519         struct ip6_tnl *tunnel;
1520         int ret;
1521
1522         ret = ip6gre_tunnel_init_common(dev);
1523         if (ret)
1524                 return ret;
1525
1526         tunnel = netdev_priv(dev);
1527
1528         if (tunnel->parms.collect_md)
1529                 return 0;
1530
1531         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1532         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1533
1534         if (ipv6_addr_any(&tunnel->parms.raddr))
1535                 dev->header_ops = &ip6gre_header_ops;
1536
1537         return 0;
1538 }
1539
1540 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1541 {
1542         struct ip6_tnl *tunnel = netdev_priv(dev);
1543
1544         tunnel->dev = dev;
1545         tunnel->net = dev_net(dev);
1546         strcpy(tunnel->parms.name, dev->name);
1547
1548         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1549 }
1550
1551 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1552         .handler     = gre_rcv,
1553         .err_handler = ip6gre_err,
1554         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1555 };
1556
1557 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1558 {
1559         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1560         struct net_device *dev, *aux;
1561         int prio;
1562
1563         for_each_netdev_safe(net, dev, aux)
1564                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1565                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1566                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1567                         unregister_netdevice_queue(dev, head);
1568
1569         for (prio = 0; prio < 4; prio++) {
1570                 int h;
1571                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1572                         struct ip6_tnl *t;
1573
1574                         t = rtnl_dereference(ign->tunnels[prio][h]);
1575
1576                         while (t) {
1577                                 /* If dev is in the same netns, it has already
1578                                  * been added to the list by the previous loop.
1579                                  */
1580                                 if (!net_eq(dev_net(t->dev), net))
1581                                         unregister_netdevice_queue(t->dev,
1582                                                                    head);
1583                                 t = rtnl_dereference(t->next);
1584                         }
1585                 }
1586         }
1587 }
1588
1589 static int __net_init ip6gre_init_net(struct net *net)
1590 {
1591         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1592         struct net_device *ndev;
1593         int err;
1594
1595         if (!net_has_fallback_tunnels(net))
1596                 return 0;
1597         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1598                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1599         if (!ndev) {
1600                 err = -ENOMEM;
1601                 goto err_alloc_dev;
1602         }
1603         ign->fb_tunnel_dev = ndev;
1604         dev_net_set(ign->fb_tunnel_dev, net);
1605         /* FB netdevice is special: we have one, and only one per netns.
1606          * Allowing to move it to another netns is clearly unsafe.
1607          */
1608         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1609
1610
1611         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1612         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1613
1614         err = register_netdev(ign->fb_tunnel_dev);
1615         if (err)
1616                 goto err_reg_dev;
1617
1618         rcu_assign_pointer(ign->tunnels_wc[0],
1619                            netdev_priv(ign->fb_tunnel_dev));
1620         return 0;
1621
1622 err_reg_dev:
1623         free_netdev(ndev);
1624 err_alloc_dev:
1625         return err;
1626 }
1627
1628 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1629 {
1630         struct net *net;
1631         LIST_HEAD(list);
1632
1633         rtnl_lock();
1634         list_for_each_entry(net, net_list, exit_list)
1635                 ip6gre_destroy_tunnels(net, &list);
1636         unregister_netdevice_many(&list);
1637         rtnl_unlock();
1638 }
1639
1640 static struct pernet_operations ip6gre_net_ops = {
1641         .init = ip6gre_init_net,
1642         .exit_batch = ip6gre_exit_batch_net,
1643         .id   = &ip6gre_net_id,
1644         .size = sizeof(struct ip6gre_net),
1645 };
1646
1647 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1648                                   struct netlink_ext_ack *extack)
1649 {
1650         __be16 flags;
1651
1652         if (!data)
1653                 return 0;
1654
1655         flags = 0;
1656         if (data[IFLA_GRE_IFLAGS])
1657                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1658         if (data[IFLA_GRE_OFLAGS])
1659                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1660         if (flags & (GRE_VERSION|GRE_ROUTING))
1661                 return -EINVAL;
1662
1663         return 0;
1664 }
1665
1666 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1667                                struct netlink_ext_ack *extack)
1668 {
1669         struct in6_addr daddr;
1670
1671         if (tb[IFLA_ADDRESS]) {
1672                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1673                         return -EINVAL;
1674                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1675                         return -EADDRNOTAVAIL;
1676         }
1677
1678         if (!data)
1679                 goto out;
1680
1681         if (data[IFLA_GRE_REMOTE]) {
1682                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1683                 if (ipv6_addr_any(&daddr))
1684                         return -EINVAL;
1685         }
1686
1687 out:
1688         return ip6gre_tunnel_validate(tb, data, extack);
1689 }
1690
1691 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1692                                   struct netlink_ext_ack *extack)
1693 {
1694         __be16 flags = 0;
1695         int ret, ver = 0;
1696
1697         if (!data)
1698                 return 0;
1699
1700         ret = ip6gre_tap_validate(tb, data, extack);
1701         if (ret)
1702                 return ret;
1703
1704         /* ERSPAN should only have GRE sequence and key flag */
1705         if (data[IFLA_GRE_OFLAGS])
1706                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1707         if (data[IFLA_GRE_IFLAGS])
1708                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1709         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1710             flags != (GRE_SEQ | GRE_KEY))
1711                 return -EINVAL;
1712
1713         /* ERSPAN Session ID only has 10-bit. Since we reuse
1714          * 32-bit key field as ID, check it's range.
1715          */
1716         if (data[IFLA_GRE_IKEY] &&
1717             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1718                 return -EINVAL;
1719
1720         if (data[IFLA_GRE_OKEY] &&
1721             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1722                 return -EINVAL;
1723
1724         if (data[IFLA_GRE_ERSPAN_VER]) {
1725                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1726                 if (ver != 1 && ver != 2)
1727                         return -EINVAL;
1728         }
1729
1730         if (ver == 1) {
1731                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1732                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1733
1734                         if (index & ~INDEX_MASK)
1735                                 return -EINVAL;
1736                 }
1737         } else if (ver == 2) {
1738                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1739                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1740
1741                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1742                                 return -EINVAL;
1743                 }
1744
1745                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1746                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1747
1748                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1749                                 return -EINVAL;
1750                 }
1751         }
1752
1753         return 0;
1754 }
1755
1756 static void ip6erspan_set_version(struct nlattr *data[],
1757                                   struct __ip6_tnl_parm *parms)
1758 {
1759         if (!data)
1760                 return;
1761
1762         parms->erspan_ver = 1;
1763         if (data[IFLA_GRE_ERSPAN_VER])
1764                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1765
1766         if (parms->erspan_ver == 1) {
1767                 if (data[IFLA_GRE_ERSPAN_INDEX])
1768                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1769         } else if (parms->erspan_ver == 2) {
1770                 if (data[IFLA_GRE_ERSPAN_DIR])
1771                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1772                 if (data[IFLA_GRE_ERSPAN_HWID])
1773                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1774         }
1775 }
1776
1777 static void ip6gre_netlink_parms(struct nlattr *data[],
1778                                 struct __ip6_tnl_parm *parms)
1779 {
1780         memset(parms, 0, sizeof(*parms));
1781
1782         if (!data)
1783                 return;
1784
1785         if (data[IFLA_GRE_LINK])
1786                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1787
1788         if (data[IFLA_GRE_IFLAGS])
1789                 parms->i_flags = gre_flags_to_tnl_flags(
1790                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1791
1792         if (data[IFLA_GRE_OFLAGS])
1793                 parms->o_flags = gre_flags_to_tnl_flags(
1794                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1795
1796         if (data[IFLA_GRE_IKEY])
1797                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1798
1799         if (data[IFLA_GRE_OKEY])
1800                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1801
1802         if (data[IFLA_GRE_LOCAL])
1803                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1804
1805         if (data[IFLA_GRE_REMOTE])
1806                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1807
1808         if (data[IFLA_GRE_TTL])
1809                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1810
1811         if (data[IFLA_GRE_ENCAP_LIMIT])
1812                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1813
1814         if (data[IFLA_GRE_FLOWINFO])
1815                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1816
1817         if (data[IFLA_GRE_FLAGS])
1818                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1819
1820         if (data[IFLA_GRE_FWMARK])
1821                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1822
1823         if (data[IFLA_GRE_COLLECT_METADATA])
1824                 parms->collect_md = true;
1825 }
1826
1827 static int ip6gre_tap_init(struct net_device *dev)
1828 {
1829         int ret;
1830
1831         ret = ip6gre_tunnel_init_common(dev);
1832         if (ret)
1833                 return ret;
1834
1835         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1836
1837         return 0;
1838 }
1839
1840 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1841         .ndo_init = ip6gre_tap_init,
1842         .ndo_uninit = ip6gre_tunnel_uninit,
1843         .ndo_start_xmit = ip6gre_tunnel_xmit,
1844         .ndo_set_mac_address = eth_mac_addr,
1845         .ndo_validate_addr = eth_validate_addr,
1846         .ndo_change_mtu = ip6_tnl_change_mtu,
1847         .ndo_get_stats64 = ip_tunnel_get_stats64,
1848         .ndo_get_iflink = ip6_tnl_get_iflink,
1849 };
1850
1851 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1852 {
1853         int t_hlen;
1854
1855         tunnel->tun_hlen = 8;
1856         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1857                        erspan_hdr_len(tunnel->parms.erspan_ver);
1858
1859         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1860         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1861         return t_hlen;
1862 }
1863
1864 static int ip6erspan_tap_init(struct net_device *dev)
1865 {
1866         struct ip6_tnl *tunnel;
1867         int t_hlen;
1868         int ret;
1869
1870         tunnel = netdev_priv(dev);
1871
1872         tunnel->dev = dev;
1873         tunnel->net = dev_net(dev);
1874         strcpy(tunnel->parms.name, dev->name);
1875
1876         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1877         if (!dev->tstats)
1878                 return -ENOMEM;
1879
1880         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1881         if (ret)
1882                 goto cleanup_alloc_pcpu_stats;
1883
1884         ret = gro_cells_init(&tunnel->gro_cells, dev);
1885         if (ret)
1886                 goto cleanup_dst_cache_init;
1887
1888         t_hlen = ip6erspan_calc_hlen(tunnel);
1889         dev->mtu = ETH_DATA_LEN - t_hlen;
1890         if (dev->type == ARPHRD_ETHER)
1891                 dev->mtu -= ETH_HLEN;
1892         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1893                 dev->mtu -= 8;
1894
1895         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1896         ip6erspan_tnl_link_config(tunnel, 1);
1897
1898         dev_hold(dev);
1899         return 0;
1900
1901 cleanup_dst_cache_init:
1902         dst_cache_destroy(&tunnel->dst_cache);
1903 cleanup_alloc_pcpu_stats:
1904         free_percpu(dev->tstats);
1905         dev->tstats = NULL;
1906         return ret;
1907 }
1908
1909 static const struct net_device_ops ip6erspan_netdev_ops = {
1910         .ndo_init =             ip6erspan_tap_init,
1911         .ndo_uninit =           ip6erspan_tunnel_uninit,
1912         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1913         .ndo_set_mac_address =  eth_mac_addr,
1914         .ndo_validate_addr =    eth_validate_addr,
1915         .ndo_change_mtu =       ip6_tnl_change_mtu,
1916         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1917         .ndo_get_iflink =       ip6_tnl_get_iflink,
1918 };
1919
1920 static void ip6gre_tap_setup(struct net_device *dev)
1921 {
1922
1923         ether_setup(dev);
1924
1925         dev->max_mtu = 0;
1926         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1927         dev->needs_free_netdev = true;
1928         dev->priv_destructor = ip6gre_dev_free;
1929
1930         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1931         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1932         netif_keep_dst(dev);
1933 }
1934
1935 bool is_ip6gretap_dev(const struct net_device *dev)
1936 {
1937         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1938 }
1939 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1940
1941 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1942                                        struct ip_tunnel_encap *ipencap)
1943 {
1944         bool ret = false;
1945
1946         memset(ipencap, 0, sizeof(*ipencap));
1947
1948         if (!data)
1949                 return ret;
1950
1951         if (data[IFLA_GRE_ENCAP_TYPE]) {
1952                 ret = true;
1953                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1954         }
1955
1956         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1957                 ret = true;
1958                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1959         }
1960
1961         if (data[IFLA_GRE_ENCAP_SPORT]) {
1962                 ret = true;
1963                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1964         }
1965
1966         if (data[IFLA_GRE_ENCAP_DPORT]) {
1967                 ret = true;
1968                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1969         }
1970
1971         return ret;
1972 }
1973
1974 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1975                                  struct nlattr *tb[], struct nlattr *data[],
1976                                  struct netlink_ext_ack *extack)
1977 {
1978         struct ip6_tnl *nt;
1979         struct ip_tunnel_encap ipencap;
1980         int err;
1981
1982         nt = netdev_priv(dev);
1983
1984         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1985                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1986
1987                 if (err < 0)
1988                         return err;
1989         }
1990
1991         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1992                 eth_hw_addr_random(dev);
1993
1994         nt->dev = dev;
1995         nt->net = dev_net(dev);
1996
1997         err = register_netdevice(dev);
1998         if (err)
1999                 goto out;
2000
2001         if (tb[IFLA_MTU])
2002                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2003
2004 out:
2005         return err;
2006 }
2007
2008 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2009                           struct nlattr *tb[], struct nlattr *data[],
2010                           struct netlink_ext_ack *extack)
2011 {
2012         struct ip6_tnl *nt = netdev_priv(dev);
2013         struct net *net = dev_net(dev);
2014         struct ip6gre_net *ign;
2015         int err;
2016
2017         ip6gre_netlink_parms(data, &nt->parms);
2018         ign = net_generic(net, ip6gre_net_id);
2019
2020         if (nt->parms.collect_md) {
2021                 if (rtnl_dereference(ign->collect_md_tun))
2022                         return -EEXIST;
2023         } else {
2024                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2025                         return -EEXIST;
2026         }
2027
2028         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2029         if (!err) {
2030                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2031                 ip6gre_tunnel_link_md(ign, nt);
2032                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2033         }
2034         return err;
2035 }
2036
2037 static struct ip6_tnl *
2038 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2039                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2040                          struct netlink_ext_ack *extack)
2041 {
2042         struct ip6_tnl *t, *nt = netdev_priv(dev);
2043         struct net *net = nt->net;
2044         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2045         struct ip_tunnel_encap ipencap;
2046
2047         if (dev == ign->fb_tunnel_dev)
2048                 return ERR_PTR(-EINVAL);
2049
2050         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2051                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2052
2053                 if (err < 0)
2054                         return ERR_PTR(err);
2055         }
2056
2057         ip6gre_netlink_parms(data, p_p);
2058
2059         t = ip6gre_tunnel_locate(net, p_p, 0);
2060
2061         if (t) {
2062                 if (t->dev != dev)
2063                         return ERR_PTR(-EEXIST);
2064         } else {
2065                 t = nt;
2066         }
2067
2068         return t;
2069 }
2070
2071 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2072                              struct nlattr *data[],
2073                              struct netlink_ext_ack *extack)
2074 {
2075         struct ip6_tnl *t = netdev_priv(dev);
2076         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2077         struct __ip6_tnl_parm p;
2078
2079         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2080         if (IS_ERR(t))
2081                 return PTR_ERR(t);
2082
2083         ip6gre_tunnel_unlink_md(ign, t);
2084         ip6gre_tunnel_unlink(ign, t);
2085         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2086         ip6gre_tunnel_link_md(ign, t);
2087         ip6gre_tunnel_link(ign, t);
2088         return 0;
2089 }
2090
2091 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2092 {
2093         struct net *net = dev_net(dev);
2094         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2095
2096         if (dev != ign->fb_tunnel_dev)
2097                 unregister_netdevice_queue(dev, head);
2098 }
2099
2100 static size_t ip6gre_get_size(const struct net_device *dev)
2101 {
2102         return
2103                 /* IFLA_GRE_LINK */
2104                 nla_total_size(4) +
2105                 /* IFLA_GRE_IFLAGS */
2106                 nla_total_size(2) +
2107                 /* IFLA_GRE_OFLAGS */
2108                 nla_total_size(2) +
2109                 /* IFLA_GRE_IKEY */
2110                 nla_total_size(4) +
2111                 /* IFLA_GRE_OKEY */
2112                 nla_total_size(4) +
2113                 /* IFLA_GRE_LOCAL */
2114                 nla_total_size(sizeof(struct in6_addr)) +
2115                 /* IFLA_GRE_REMOTE */
2116                 nla_total_size(sizeof(struct in6_addr)) +
2117                 /* IFLA_GRE_TTL */
2118                 nla_total_size(1) +
2119                 /* IFLA_GRE_ENCAP_LIMIT */
2120                 nla_total_size(1) +
2121                 /* IFLA_GRE_FLOWINFO */
2122                 nla_total_size(4) +
2123                 /* IFLA_GRE_FLAGS */
2124                 nla_total_size(4) +
2125                 /* IFLA_GRE_ENCAP_TYPE */
2126                 nla_total_size(2) +
2127                 /* IFLA_GRE_ENCAP_FLAGS */
2128                 nla_total_size(2) +
2129                 /* IFLA_GRE_ENCAP_SPORT */
2130                 nla_total_size(2) +
2131                 /* IFLA_GRE_ENCAP_DPORT */
2132                 nla_total_size(2) +
2133                 /* IFLA_GRE_COLLECT_METADATA */
2134                 nla_total_size(0) +
2135                 /* IFLA_GRE_FWMARK */
2136                 nla_total_size(4) +
2137                 /* IFLA_GRE_ERSPAN_INDEX */
2138                 nla_total_size(4) +
2139                 0;
2140 }
2141
2142 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2143 {
2144         struct ip6_tnl *t = netdev_priv(dev);
2145         struct __ip6_tnl_parm *p = &t->parms;
2146         __be16 o_flags = p->o_flags;
2147
2148         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2149                 if (!p->collect_md)
2150                         o_flags |= TUNNEL_KEY;
2151
2152                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2153                         goto nla_put_failure;
2154
2155                 if (p->erspan_ver == 1) {
2156                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2157                                 goto nla_put_failure;
2158                 } else {
2159                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2160                                 goto nla_put_failure;
2161                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2162                                 goto nla_put_failure;
2163                 }
2164         }
2165
2166         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2167             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2168                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2169             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2170                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2171             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2172             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2173             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2174             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2175             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2176             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2177             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2178             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2179             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2180                 goto nla_put_failure;
2181
2182         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2183                         t->encap.type) ||
2184             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2185                          t->encap.sport) ||
2186             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2187                          t->encap.dport) ||
2188             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2189                         t->encap.flags))
2190                 goto nla_put_failure;
2191
2192         if (p->collect_md) {
2193                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2194                         goto nla_put_failure;
2195         }
2196
2197         return 0;
2198
2199 nla_put_failure:
2200         return -EMSGSIZE;
2201 }
2202
2203 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2204         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2205         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2206         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2207         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2208         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2209         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2210         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2211         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2212         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2213         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2214         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2215         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2216         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2217         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2218         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2219         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2220         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2221         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2222         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2223         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2224         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2225 };
2226
2227 static void ip6erspan_tap_setup(struct net_device *dev)
2228 {
2229         ether_setup(dev);
2230
2231         dev->max_mtu = 0;
2232         dev->netdev_ops = &ip6erspan_netdev_ops;
2233         dev->needs_free_netdev = true;
2234         dev->priv_destructor = ip6gre_dev_free;
2235
2236         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2237         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2238         netif_keep_dst(dev);
2239 }
2240
2241 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2242                              struct nlattr *tb[], struct nlattr *data[],
2243                              struct netlink_ext_ack *extack)
2244 {
2245         struct ip6_tnl *nt = netdev_priv(dev);
2246         struct net *net = dev_net(dev);
2247         struct ip6gre_net *ign;
2248         int err;
2249
2250         ip6gre_netlink_parms(data, &nt->parms);
2251         ip6erspan_set_version(data, &nt->parms);
2252         ign = net_generic(net, ip6gre_net_id);
2253
2254         if (nt->parms.collect_md) {
2255                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2256                         return -EEXIST;
2257         } else {
2258                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2259                         return -EEXIST;
2260         }
2261
2262         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2263         if (!err) {
2264                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2265                 ip6erspan_tunnel_link_md(ign, nt);
2266                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2267         }
2268         return err;
2269 }
2270
2271 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2272 {
2273         ip6gre_tnl_link_config_common(t);
2274         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2275 }
2276
2277 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2278                                 const struct __ip6_tnl_parm *p, int set_mtu)
2279 {
2280         ip6gre_tnl_copy_tnl_parm(t, p);
2281         ip6erspan_tnl_link_config(t, set_mtu);
2282         return 0;
2283 }
2284
2285 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2286                                 struct nlattr *data[],
2287                                 struct netlink_ext_ack *extack)
2288 {
2289         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2290         struct __ip6_tnl_parm p;
2291         struct ip6_tnl *t;
2292
2293         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2294         if (IS_ERR(t))
2295                 return PTR_ERR(t);
2296
2297         ip6erspan_set_version(data, &p);
2298         ip6gre_tunnel_unlink_md(ign, t);
2299         ip6gre_tunnel_unlink(ign, t);
2300         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2301         ip6erspan_tunnel_link_md(ign, t);
2302         ip6gre_tunnel_link(ign, t);
2303         return 0;
2304 }
2305
2306 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2307         .kind           = "ip6gre",
2308         .maxtype        = IFLA_GRE_MAX,
2309         .policy         = ip6gre_policy,
2310         .priv_size      = sizeof(struct ip6_tnl),
2311         .setup          = ip6gre_tunnel_setup,
2312         .validate       = ip6gre_tunnel_validate,
2313         .newlink        = ip6gre_newlink,
2314         .changelink     = ip6gre_changelink,
2315         .dellink        = ip6gre_dellink,
2316         .get_size       = ip6gre_get_size,
2317         .fill_info      = ip6gre_fill_info,
2318         .get_link_net   = ip6_tnl_get_link_net,
2319 };
2320
2321 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2322         .kind           = "ip6gretap",
2323         .maxtype        = IFLA_GRE_MAX,
2324         .policy         = ip6gre_policy,
2325         .priv_size      = sizeof(struct ip6_tnl),
2326         .setup          = ip6gre_tap_setup,
2327         .validate       = ip6gre_tap_validate,
2328         .newlink        = ip6gre_newlink,
2329         .changelink     = ip6gre_changelink,
2330         .get_size       = ip6gre_get_size,
2331         .fill_info      = ip6gre_fill_info,
2332         .get_link_net   = ip6_tnl_get_link_net,
2333 };
2334
2335 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2336         .kind           = "ip6erspan",
2337         .maxtype        = IFLA_GRE_MAX,
2338         .policy         = ip6gre_policy,
2339         .priv_size      = sizeof(struct ip6_tnl),
2340         .setup          = ip6erspan_tap_setup,
2341         .validate       = ip6erspan_tap_validate,
2342         .newlink        = ip6erspan_newlink,
2343         .changelink     = ip6erspan_changelink,
2344         .get_size       = ip6gre_get_size,
2345         .fill_info      = ip6gre_fill_info,
2346         .get_link_net   = ip6_tnl_get_link_net,
2347 };
2348
2349 /*
2350  *      And now the modules code and kernel interface.
2351  */
2352
2353 static int __init ip6gre_init(void)
2354 {
2355         int err;
2356
2357         pr_info("GRE over IPv6 tunneling driver\n");
2358
2359         err = register_pernet_device(&ip6gre_net_ops);
2360         if (err < 0)
2361                 return err;
2362
2363         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2364         if (err < 0) {
2365                 pr_info("%s: can't add protocol\n", __func__);
2366                 goto add_proto_failed;
2367         }
2368
2369         err = rtnl_link_register(&ip6gre_link_ops);
2370         if (err < 0)
2371                 goto rtnl_link_failed;
2372
2373         err = rtnl_link_register(&ip6gre_tap_ops);
2374         if (err < 0)
2375                 goto tap_ops_failed;
2376
2377         err = rtnl_link_register(&ip6erspan_tap_ops);
2378         if (err < 0)
2379                 goto erspan_link_failed;
2380
2381 out:
2382         return err;
2383
2384 erspan_link_failed:
2385         rtnl_link_unregister(&ip6gre_tap_ops);
2386 tap_ops_failed:
2387         rtnl_link_unregister(&ip6gre_link_ops);
2388 rtnl_link_failed:
2389         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2390 add_proto_failed:
2391         unregister_pernet_device(&ip6gre_net_ops);
2392         goto out;
2393 }
2394
2395 static void __exit ip6gre_fini(void)
2396 {
2397         rtnl_link_unregister(&ip6gre_tap_ops);
2398         rtnl_link_unregister(&ip6gre_link_ops);
2399         rtnl_link_unregister(&ip6erspan_tap_ops);
2400         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2401         unregister_pernet_device(&ip6gre_net_ops);
2402 }
2403
2404 module_init(ip6gre_init);
2405 module_exit(ip6gre_fini);
2406 MODULE_LICENSE("GPL");
2407 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2408 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2409 MODULE_ALIAS_RTNL_LINK("ip6gre");
2410 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2411 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2412 MODULE_ALIAS_NETDEV("ip6gre0");