Merge remote-tracking branch 'stable/linux-4.19.y' into rpi-4.19.y
[platform/kernel/linux-rpi.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, struct tnl_ptk_info *tpi,
544                          int gre_hdr_len)
545 {
546         struct erspan_base_hdr *ershdr;
547         const struct ipv6hdr *ipv6h;
548         struct erspan_md2 *md2;
549         struct ip6_tnl *tunnel;
550         u8 ver;
551
552         ipv6h = ipv6_hdr(skb);
553         ershdr = (struct erspan_base_hdr *)skb->data;
554         ver = ershdr->ver;
555
556         tunnel = ip6gre_tunnel_lookup(skb->dev,
557                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
558                                       tpi->proto);
559         if (tunnel) {
560                 int len = erspan_hdr_len(ver);
561
562                 if (unlikely(!pskb_may_pull(skb, len)))
563                         return PACKET_REJECT;
564
565                 if (__iptunnel_pull_header(skb, len,
566                                            htons(ETH_P_TEB),
567                                            false, false) < 0)
568                         return PACKET_REJECT;
569
570                 if (tunnel->parms.collect_md) {
571                         struct erspan_metadata *pkt_md, *md;
572                         struct metadata_dst *tun_dst;
573                         struct ip_tunnel_info *info;
574                         unsigned char *gh;
575                         __be64 tun_id;
576                         __be16 flags;
577
578                         tpi->flags |= TUNNEL_KEY;
579                         flags = tpi->flags;
580                         tun_id = key32_to_tunnel_id(tpi->key);
581
582                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
583                                                   sizeof(*md));
584                         if (!tun_dst)
585                                 return PACKET_REJECT;
586
587                         /* skb can be uncloned in __iptunnel_pull_header, so
588                          * old pkt_md is no longer valid and we need to reset
589                          * it
590                          */
591                         gh = skb_network_header(skb) +
592                              skb_network_header_len(skb);
593                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
594                                                             sizeof(*ershdr));
595                         info = &tun_dst->u.tun_info;
596                         md = ip_tunnel_info_opts(info);
597                         md->version = ver;
598                         md2 = &md->u.md2;
599                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
600                                                        ERSPAN_V2_MDSIZE);
601                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
602                         info->options_len = sizeof(*md);
603
604                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
605
606                 } else {
607                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
608                 }
609
610                 return PACKET_RCVD;
611         }
612
613         return PACKET_REJECT;
614 }
615
616 static int gre_rcv(struct sk_buff *skb)
617 {
618         struct tnl_ptk_info tpi;
619         bool csum_err = false;
620         int hdr_len;
621
622         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
623         if (hdr_len < 0)
624                 goto drop;
625
626         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
627                 goto drop;
628
629         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
630                      tpi.proto == htons(ETH_P_ERSPAN2))) {
631                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
632                         return 0;
633                 goto out;
634         }
635
636         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
637                 return 0;
638
639 out:
640         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
641 drop:
642         kfree_skb(skb);
643         return 0;
644 }
645
646 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
647 {
648         return iptunnel_handle_offloads(skb,
649                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
650 }
651
652 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
653                                      struct net_device *dev,
654                                      struct flowi6 *fl6, __u8 *dsfield,
655                                      int *encap_limit)
656 {
657         const struct iphdr *iph = ip_hdr(skb);
658         struct ip6_tnl *t = netdev_priv(dev);
659
660         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
661                 *encap_limit = t->parms.encap_limit;
662
663         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
664
665         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
666                 *dsfield = ipv4_get_dsfield(iph);
667         else
668                 *dsfield = ip6_tclass(t->parms.flowinfo);
669
670         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
671                 fl6->flowi6_mark = skb->mark;
672         else
673                 fl6->flowi6_mark = t->parms.fwmark;
674
675         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
676 }
677
678 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
679                                     struct net_device *dev,
680                                     struct flowi6 *fl6, __u8 *dsfield,
681                                     int *encap_limit)
682 {
683         struct ipv6hdr *ipv6h;
684         struct ip6_tnl *t = netdev_priv(dev);
685         __u16 offset;
686
687         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
688         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
689         ipv6h = ipv6_hdr(skb);
690
691         if (offset > 0) {
692                 struct ipv6_tlv_tnl_enc_lim *tel;
693
694                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
695                 if (tel->encap_limit == 0) {
696                         icmpv6_send(skb, ICMPV6_PARAMPROB,
697                                     ICMPV6_HDR_FIELD, offset + 2);
698                         return -1;
699                 }
700                 *encap_limit = tel->encap_limit - 1;
701         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
702                 *encap_limit = t->parms.encap_limit;
703         }
704
705         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
706
707         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
708                 *dsfield = ipv6_get_dsfield(ipv6h);
709         else
710                 *dsfield = ip6_tclass(t->parms.flowinfo);
711
712         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
713                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
714
715         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
716                 fl6->flowi6_mark = skb->mark;
717         else
718                 fl6->flowi6_mark = t->parms.fwmark;
719
720         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
721
722         return 0;
723 }
724
725 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
726                                struct net_device *dev, __u8 dsfield,
727                                struct flowi6 *fl6, int encap_limit,
728                                __u32 *pmtu, __be16 proto)
729 {
730         struct ip6_tnl *tunnel = netdev_priv(dev);
731         __be16 protocol;
732
733         if (dev->type == ARPHRD_ETHER)
734                 IPCB(skb)->flags = 0;
735
736         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
737                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
738         else
739                 fl6->daddr = tunnel->parms.raddr;
740
741         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
742                 return -ENOMEM;
743
744         /* Push GRE header. */
745         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
746
747         if (tunnel->parms.collect_md) {
748                 struct ip_tunnel_info *tun_info;
749                 const struct ip_tunnel_key *key;
750                 __be16 flags;
751
752                 tun_info = skb_tunnel_info(skb);
753                 if (unlikely(!tun_info ||
754                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
755                              ip_tunnel_info_af(tun_info) != AF_INET6))
756                         return -EINVAL;
757
758                 key = &tun_info->key;
759                 memset(fl6, 0, sizeof(*fl6));
760                 fl6->flowi6_proto = IPPROTO_GRE;
761                 fl6->daddr = key->u.ipv6.dst;
762                 fl6->flowlabel = key->label;
763                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
764
765                 dsfield = key->tos;
766                 flags = key->tun_flags &
767                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
768                 tunnel->tun_hlen = gre_calc_hlen(flags);
769
770                 gre_build_header(skb, tunnel->tun_hlen,
771                                  flags, protocol,
772                                  tunnel_id_to_key32(tun_info->key.tun_id),
773                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
774                                                       : 0);
775
776         } else {
777                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
778                         tunnel->o_seqno++;
779
780                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
781                                  protocol, tunnel->parms.o_key,
782                                  htonl(tunnel->o_seqno));
783         }
784
785         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
786                             NEXTHDR_GRE);
787 }
788
789 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
790 {
791         struct ip6_tnl *t = netdev_priv(dev);
792         int encap_limit = -1;
793         struct flowi6 fl6;
794         __u8 dsfield = 0;
795         __u32 mtu;
796         int err;
797
798         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
799
800         if (!t->parms.collect_md)
801                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
802                                          &dsfield, &encap_limit);
803
804         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
805         if (err)
806                 return -1;
807
808         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
809                           skb->protocol);
810         if (err != 0) {
811                 /* XXX: send ICMP error even if DF is not set. */
812                 if (err == -EMSGSIZE)
813                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
814                                   htonl(mtu));
815                 return -1;
816         }
817
818         return 0;
819 }
820
821 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
822 {
823         struct ip6_tnl *t = netdev_priv(dev);
824         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
825         int encap_limit = -1;
826         struct flowi6 fl6;
827         __u8 dsfield = 0;
828         __u32 mtu;
829         int err;
830
831         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
832                 return -1;
833
834         if (!t->parms.collect_md &&
835             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
836                 return -1;
837
838         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
839                 return -1;
840
841         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
842                           &mtu, skb->protocol);
843         if (err != 0) {
844                 if (err == -EMSGSIZE)
845                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
846                 return -1;
847         }
848
849         return 0;
850 }
851
852 /**
853  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
854  *   @t: the outgoing tunnel device
855  *   @hdr: IPv6 header from the incoming packet
856  *
857  * Description:
858  *   Avoid trivial tunneling loop by checking that tunnel exit-point
859  *   doesn't match source of incoming packet.
860  *
861  * Return:
862  *   1 if conflict,
863  *   0 else
864  **/
865
866 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
867         const struct ipv6hdr *hdr)
868 {
869         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
870 }
871
872 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
873 {
874         struct ip6_tnl *t = netdev_priv(dev);
875         int encap_limit = -1;
876         struct flowi6 fl6;
877         __u32 mtu;
878         int err;
879
880         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
881                 encap_limit = t->parms.encap_limit;
882
883         if (!t->parms.collect_md)
884                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
885
886         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
887         if (err)
888                 return err;
889
890         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
891
892         return err;
893 }
894
895 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
896         struct net_device *dev)
897 {
898         struct ip6_tnl *t = netdev_priv(dev);
899         struct net_device_stats *stats = &t->dev->stats;
900         int ret;
901
902         if (!pskb_inet_may_pull(skb))
903                 goto tx_err;
904
905         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
906                 goto tx_err;
907
908         switch (skb->protocol) {
909         case htons(ETH_P_IP):
910                 ret = ip6gre_xmit_ipv4(skb, dev);
911                 break;
912         case htons(ETH_P_IPV6):
913                 ret = ip6gre_xmit_ipv6(skb, dev);
914                 break;
915         default:
916                 ret = ip6gre_xmit_other(skb, dev);
917                 break;
918         }
919
920         if (ret < 0)
921                 goto tx_err;
922
923         return NETDEV_TX_OK;
924
925 tx_err:
926         stats->tx_errors++;
927         stats->tx_dropped++;
928         kfree_skb(skb);
929         return NETDEV_TX_OK;
930 }
931
932 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
933                                          struct net_device *dev)
934 {
935         struct ip6_tnl *t = netdev_priv(dev);
936         struct dst_entry *dst = skb_dst(skb);
937         struct net_device_stats *stats;
938         bool truncate = false;
939         int encap_limit = -1;
940         __u8 dsfield = false;
941         struct flowi6 fl6;
942         int err = -EINVAL;
943         __be16 proto;
944         __u32 mtu;
945         int nhoff;
946         int thoff;
947
948         if (!pskb_inet_may_pull(skb))
949                 goto tx_err;
950
951         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
952                 goto tx_err;
953
954         if (gre_handle_offloads(skb, false))
955                 goto tx_err;
956
957         if (skb->len > dev->mtu + dev->hard_header_len) {
958                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
959                 truncate = true;
960         }
961
962         nhoff = skb_network_header(skb) - skb_mac_header(skb);
963         if (skb->protocol == htons(ETH_P_IP) &&
964             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
965                 truncate = true;
966
967         thoff = skb_transport_header(skb) - skb_mac_header(skb);
968         if (skb->protocol == htons(ETH_P_IPV6) &&
969             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
970                 truncate = true;
971
972         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
973                 goto tx_err;
974
975         t->parms.o_flags &= ~TUNNEL_KEY;
976         IPCB(skb)->flags = 0;
977
978         /* For collect_md mode, derive fl6 from the tunnel key,
979          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
980          */
981         if (t->parms.collect_md) {
982                 struct ip_tunnel_info *tun_info;
983                 const struct ip_tunnel_key *key;
984                 struct erspan_metadata *md;
985                 __be32 tun_id;
986
987                 tun_info = skb_tunnel_info(skb);
988                 if (unlikely(!tun_info ||
989                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
990                              ip_tunnel_info_af(tun_info) != AF_INET6))
991                         goto tx_err;
992
993                 key = &tun_info->key;
994                 memset(&fl6, 0, sizeof(fl6));
995                 fl6.flowi6_proto = IPPROTO_GRE;
996                 fl6.daddr = key->u.ipv6.dst;
997                 fl6.flowlabel = key->label;
998                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
999
1000                 dsfield = key->tos;
1001                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1002                         goto tx_err;
1003                 if (tun_info->options_len < sizeof(*md))
1004                         goto tx_err;
1005                 md = ip_tunnel_info_opts(tun_info);
1006
1007                 tun_id = tunnel_id_to_key32(key->tun_id);
1008                 if (md->version == 1) {
1009                         erspan_build_header(skb,
1010                                             ntohl(tun_id),
1011                                             ntohl(md->u.index), truncate,
1012                                             false);
1013                 } else if (md->version == 2) {
1014                         erspan_build_header_v2(skb,
1015                                                ntohl(tun_id),
1016                                                md->u.md2.dir,
1017                                                get_hwid(&md->u.md2),
1018                                                truncate, false);
1019                 } else {
1020                         goto tx_err;
1021                 }
1022         } else {
1023                 switch (skb->protocol) {
1024                 case htons(ETH_P_IP):
1025                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1026                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1027                                                  &dsfield, &encap_limit);
1028                         break;
1029                 case htons(ETH_P_IPV6):
1030                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1031                                 goto tx_err;
1032                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1033                                                      &dsfield, &encap_limit))
1034                                 goto tx_err;
1035                         break;
1036                 default:
1037                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1038                         break;
1039                 }
1040
1041                 if (t->parms.erspan_ver == 1)
1042                         erspan_build_header(skb, ntohl(t->parms.o_key),
1043                                             t->parms.index,
1044                                             truncate, false);
1045                 else if (t->parms.erspan_ver == 2)
1046                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1047                                                t->parms.dir,
1048                                                t->parms.hwid,
1049                                                truncate, false);
1050                 else
1051                         goto tx_err;
1052
1053                 fl6.daddr = t->parms.raddr;
1054         }
1055
1056         /* Push GRE header. */
1057         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1058                                            : htons(ETH_P_ERSPAN2);
1059         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1060
1061         /* TooBig packet may have updated dst->dev's mtu */
1062         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1063                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1064
1065         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1066                            NEXTHDR_GRE);
1067         if (err != 0) {
1068                 /* XXX: send ICMP error even if DF is not set. */
1069                 if (err == -EMSGSIZE) {
1070                         if (skb->protocol == htons(ETH_P_IP))
1071                                 icmp_send(skb, ICMP_DEST_UNREACH,
1072                                           ICMP_FRAG_NEEDED, htonl(mtu));
1073                         else
1074                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1075                 }
1076
1077                 goto tx_err;
1078         }
1079         return NETDEV_TX_OK;
1080
1081 tx_err:
1082         stats = &t->dev->stats;
1083         stats->tx_errors++;
1084         stats->tx_dropped++;
1085         kfree_skb(skb);
1086         return NETDEV_TX_OK;
1087 }
1088
1089 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1090 {
1091         struct net_device *dev = t->dev;
1092         struct __ip6_tnl_parm *p = &t->parms;
1093         struct flowi6 *fl6 = &t->fl.u.ip6;
1094
1095         if (dev->type != ARPHRD_ETHER) {
1096                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1097                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1098         }
1099
1100         /* Set up flowi template */
1101         fl6->saddr = p->laddr;
1102         fl6->daddr = p->raddr;
1103         fl6->flowi6_oif = p->link;
1104         fl6->flowlabel = 0;
1105         fl6->flowi6_proto = IPPROTO_GRE;
1106
1107         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1108                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1109         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1110                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1111
1112         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1113         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1114
1115         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1116                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1117                 dev->flags |= IFF_POINTOPOINT;
1118         else
1119                 dev->flags &= ~IFF_POINTOPOINT;
1120 }
1121
1122 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1123                                          int t_hlen)
1124 {
1125         const struct __ip6_tnl_parm *p = &t->parms;
1126         struct net_device *dev = t->dev;
1127
1128         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1129                 int strict = (ipv6_addr_type(&p->raddr) &
1130                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1131
1132                 struct rt6_info *rt = rt6_lookup(t->net,
1133                                                  &p->raddr, &p->laddr,
1134                                                  p->link, NULL, strict);
1135
1136                 if (!rt)
1137                         return;
1138
1139                 if (rt->dst.dev) {
1140                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1141                                                t_hlen;
1142
1143                         if (set_mtu) {
1144                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1145                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1146                                         dev->mtu -= 8;
1147                                 if (dev->type == ARPHRD_ETHER)
1148                                         dev->mtu -= ETH_HLEN;
1149
1150                                 if (dev->mtu < IPV6_MIN_MTU)
1151                                         dev->mtu = IPV6_MIN_MTU;
1152                         }
1153                 }
1154                 ip6_rt_put(rt);
1155         }
1156 }
1157
1158 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1159 {
1160         int t_hlen;
1161
1162         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1163         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1164
1165         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1166         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1167         return t_hlen;
1168 }
1169
1170 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1171 {
1172         ip6gre_tnl_link_config_common(t);
1173         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1174 }
1175
1176 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1177                                      const struct __ip6_tnl_parm *p)
1178 {
1179         t->parms.laddr = p->laddr;
1180         t->parms.raddr = p->raddr;
1181         t->parms.flags = p->flags;
1182         t->parms.hop_limit = p->hop_limit;
1183         t->parms.encap_limit = p->encap_limit;
1184         t->parms.flowinfo = p->flowinfo;
1185         t->parms.link = p->link;
1186         t->parms.proto = p->proto;
1187         t->parms.i_key = p->i_key;
1188         t->parms.o_key = p->o_key;
1189         t->parms.i_flags = p->i_flags;
1190         t->parms.o_flags = p->o_flags;
1191         t->parms.fwmark = p->fwmark;
1192         t->parms.erspan_ver = p->erspan_ver;
1193         t->parms.index = p->index;
1194         t->parms.dir = p->dir;
1195         t->parms.hwid = p->hwid;
1196         dst_cache_reset(&t->dst_cache);
1197 }
1198
1199 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1200                              int set_mtu)
1201 {
1202         ip6gre_tnl_copy_tnl_parm(t, p);
1203         ip6gre_tnl_link_config(t, set_mtu);
1204         return 0;
1205 }
1206
1207 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1208         const struct ip6_tnl_parm2 *u)
1209 {
1210         p->laddr = u->laddr;
1211         p->raddr = u->raddr;
1212         p->flags = u->flags;
1213         p->hop_limit = u->hop_limit;
1214         p->encap_limit = u->encap_limit;
1215         p->flowinfo = u->flowinfo;
1216         p->link = u->link;
1217         p->i_key = u->i_key;
1218         p->o_key = u->o_key;
1219         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1220         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1221         memcpy(p->name, u->name, sizeof(u->name));
1222 }
1223
1224 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1225         const struct __ip6_tnl_parm *p)
1226 {
1227         u->proto = IPPROTO_GRE;
1228         u->laddr = p->laddr;
1229         u->raddr = p->raddr;
1230         u->flags = p->flags;
1231         u->hop_limit = p->hop_limit;
1232         u->encap_limit = p->encap_limit;
1233         u->flowinfo = p->flowinfo;
1234         u->link = p->link;
1235         u->i_key = p->i_key;
1236         u->o_key = p->o_key;
1237         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1238         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1239         memcpy(u->name, p->name, sizeof(u->name));
1240 }
1241
1242 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1243         struct ifreq *ifr, int cmd)
1244 {
1245         int err = 0;
1246         struct ip6_tnl_parm2 p;
1247         struct __ip6_tnl_parm p1;
1248         struct ip6_tnl *t = netdev_priv(dev);
1249         struct net *net = t->net;
1250         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1251
1252         memset(&p1, 0, sizeof(p1));
1253
1254         switch (cmd) {
1255         case SIOCGETTUNNEL:
1256                 if (dev == ign->fb_tunnel_dev) {
1257                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1258                                 err = -EFAULT;
1259                                 break;
1260                         }
1261                         ip6gre_tnl_parm_from_user(&p1, &p);
1262                         t = ip6gre_tunnel_locate(net, &p1, 0);
1263                         if (!t)
1264                                 t = netdev_priv(dev);
1265                 }
1266                 memset(&p, 0, sizeof(p));
1267                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1268                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1269                         err = -EFAULT;
1270                 break;
1271
1272         case SIOCADDTUNNEL:
1273         case SIOCCHGTUNNEL:
1274                 err = -EPERM;
1275                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1276                         goto done;
1277
1278                 err = -EFAULT;
1279                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1280                         goto done;
1281
1282                 err = -EINVAL;
1283                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1284                         goto done;
1285
1286                 if (!(p.i_flags&GRE_KEY))
1287                         p.i_key = 0;
1288                 if (!(p.o_flags&GRE_KEY))
1289                         p.o_key = 0;
1290
1291                 ip6gre_tnl_parm_from_user(&p1, &p);
1292                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1293
1294                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1295                         if (t) {
1296                                 if (t->dev != dev) {
1297                                         err = -EEXIST;
1298                                         break;
1299                                 }
1300                         } else {
1301                                 t = netdev_priv(dev);
1302
1303                                 ip6gre_tunnel_unlink(ign, t);
1304                                 synchronize_net();
1305                                 ip6gre_tnl_change(t, &p1, 1);
1306                                 ip6gre_tunnel_link(ign, t);
1307                                 netdev_state_change(dev);
1308                         }
1309                 }
1310
1311                 if (t) {
1312                         err = 0;
1313
1314                         memset(&p, 0, sizeof(p));
1315                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1316                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1317                                 err = -EFAULT;
1318                 } else
1319                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1320                 break;
1321
1322         case SIOCDELTUNNEL:
1323                 err = -EPERM;
1324                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1325                         goto done;
1326
1327                 if (dev == ign->fb_tunnel_dev) {
1328                         err = -EFAULT;
1329                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1330                                 goto done;
1331                         err = -ENOENT;
1332                         ip6gre_tnl_parm_from_user(&p1, &p);
1333                         t = ip6gre_tunnel_locate(net, &p1, 0);
1334                         if (!t)
1335                                 goto done;
1336                         err = -EPERM;
1337                         if (t == netdev_priv(ign->fb_tunnel_dev))
1338                                 goto done;
1339                         dev = t->dev;
1340                 }
1341                 unregister_netdevice(dev);
1342                 err = 0;
1343                 break;
1344
1345         default:
1346                 err = -EINVAL;
1347         }
1348
1349 done:
1350         return err;
1351 }
1352
1353 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1354                          unsigned short type, const void *daddr,
1355                          const void *saddr, unsigned int len)
1356 {
1357         struct ip6_tnl *t = netdev_priv(dev);
1358         struct ipv6hdr *ipv6h;
1359         __be16 *p;
1360
1361         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1362         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1363                                                   t->fl.u.ip6.flowlabel,
1364                                                   true, &t->fl.u.ip6));
1365         ipv6h->hop_limit = t->parms.hop_limit;
1366         ipv6h->nexthdr = NEXTHDR_GRE;
1367         ipv6h->saddr = t->parms.laddr;
1368         ipv6h->daddr = t->parms.raddr;
1369
1370         p = (__be16 *)(ipv6h + 1);
1371         p[0] = t->parms.o_flags;
1372         p[1] = htons(type);
1373
1374         /*
1375          *      Set the source hardware address.
1376          */
1377
1378         if (saddr)
1379                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1380         if (daddr)
1381                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1382         if (!ipv6_addr_any(&ipv6h->daddr))
1383                 return t->hlen;
1384
1385         return -t->hlen;
1386 }
1387
1388 static const struct header_ops ip6gre_header_ops = {
1389         .create = ip6gre_header,
1390 };
1391
1392 static const struct net_device_ops ip6gre_netdev_ops = {
1393         .ndo_init               = ip6gre_tunnel_init,
1394         .ndo_uninit             = ip6gre_tunnel_uninit,
1395         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1396         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1397         .ndo_change_mtu         = ip6_tnl_change_mtu,
1398         .ndo_get_stats64        = ip_tunnel_get_stats64,
1399         .ndo_get_iflink         = ip6_tnl_get_iflink,
1400 };
1401
1402 static void ip6gre_dev_free(struct net_device *dev)
1403 {
1404         struct ip6_tnl *t = netdev_priv(dev);
1405
1406         gro_cells_destroy(&t->gro_cells);
1407         dst_cache_destroy(&t->dst_cache);
1408         free_percpu(dev->tstats);
1409 }
1410
1411 static void ip6gre_tunnel_setup(struct net_device *dev)
1412 {
1413         dev->netdev_ops = &ip6gre_netdev_ops;
1414         dev->needs_free_netdev = true;
1415         dev->priv_destructor = ip6gre_dev_free;
1416
1417         dev->type = ARPHRD_IP6GRE;
1418
1419         dev->flags |= IFF_NOARP;
1420         dev->addr_len = sizeof(struct in6_addr);
1421         netif_keep_dst(dev);
1422         /* This perm addr will be used as interface identifier by IPv6 */
1423         dev->addr_assign_type = NET_ADDR_RANDOM;
1424         eth_random_addr(dev->perm_addr);
1425 }
1426
1427 #define GRE6_FEATURES (NETIF_F_SG |             \
1428                        NETIF_F_FRAGLIST |       \
1429                        NETIF_F_HIGHDMA |        \
1430                        NETIF_F_HW_CSUM)
1431
1432 static void ip6gre_tnl_init_features(struct net_device *dev)
1433 {
1434         struct ip6_tnl *nt = netdev_priv(dev);
1435
1436         dev->features           |= GRE6_FEATURES;
1437         dev->hw_features        |= GRE6_FEATURES;
1438
1439         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1440                 /* TCP offload with GRE SEQ is not supported, nor
1441                  * can we support 2 levels of outer headers requiring
1442                  * an update.
1443                  */
1444                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1445                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1446                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1447                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1448                 }
1449
1450                 /* Can use a lockless transmit, unless we generate
1451                  * output sequences
1452                  */
1453                 dev->features |= NETIF_F_LLTX;
1454         }
1455 }
1456
1457 static int ip6gre_tunnel_init_common(struct net_device *dev)
1458 {
1459         struct ip6_tnl *tunnel;
1460         int ret;
1461         int t_hlen;
1462
1463         tunnel = netdev_priv(dev);
1464
1465         tunnel->dev = dev;
1466         tunnel->net = dev_net(dev);
1467         strcpy(tunnel->parms.name, dev->name);
1468
1469         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1470         if (!dev->tstats)
1471                 return -ENOMEM;
1472
1473         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1474         if (ret)
1475                 goto cleanup_alloc_pcpu_stats;
1476
1477         ret = gro_cells_init(&tunnel->gro_cells, dev);
1478         if (ret)
1479                 goto cleanup_dst_cache_init;
1480
1481         t_hlen = ip6gre_calc_hlen(tunnel);
1482         dev->mtu = ETH_DATA_LEN - t_hlen;
1483         if (dev->type == ARPHRD_ETHER)
1484                 dev->mtu -= ETH_HLEN;
1485         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1486                 dev->mtu -= 8;
1487
1488         if (tunnel->parms.collect_md) {
1489                 dev->features |= NETIF_F_NETNS_LOCAL;
1490                 netif_keep_dst(dev);
1491         }
1492         ip6gre_tnl_init_features(dev);
1493
1494         return 0;
1495
1496 cleanup_dst_cache_init:
1497         dst_cache_destroy(&tunnel->dst_cache);
1498 cleanup_alloc_pcpu_stats:
1499         free_percpu(dev->tstats);
1500         dev->tstats = NULL;
1501         return ret;
1502 }
1503
1504 static int ip6gre_tunnel_init(struct net_device *dev)
1505 {
1506         struct ip6_tnl *tunnel;
1507         int ret;
1508
1509         ret = ip6gre_tunnel_init_common(dev);
1510         if (ret)
1511                 return ret;
1512
1513         tunnel = netdev_priv(dev);
1514
1515         if (tunnel->parms.collect_md)
1516                 return 0;
1517
1518         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1519         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1520
1521         if (ipv6_addr_any(&tunnel->parms.raddr))
1522                 dev->header_ops = &ip6gre_header_ops;
1523
1524         return 0;
1525 }
1526
1527 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1528 {
1529         struct ip6_tnl *tunnel = netdev_priv(dev);
1530
1531         tunnel->dev = dev;
1532         tunnel->net = dev_net(dev);
1533         strcpy(tunnel->parms.name, dev->name);
1534
1535         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1536
1537         dev_hold(dev);
1538 }
1539
1540 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1541         .handler     = gre_rcv,
1542         .err_handler = ip6gre_err,
1543         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1544 };
1545
1546 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1547 {
1548         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1549         struct net_device *dev, *aux;
1550         int prio;
1551
1552         for_each_netdev_safe(net, dev, aux)
1553                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1554                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1555                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1556                         unregister_netdevice_queue(dev, head);
1557
1558         for (prio = 0; prio < 4; prio++) {
1559                 int h;
1560                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1561                         struct ip6_tnl *t;
1562
1563                         t = rtnl_dereference(ign->tunnels[prio][h]);
1564
1565                         while (t) {
1566                                 /* If dev is in the same netns, it has already
1567                                  * been added to the list by the previous loop.
1568                                  */
1569                                 if (!net_eq(dev_net(t->dev), net))
1570                                         unregister_netdevice_queue(t->dev,
1571                                                                    head);
1572                                 t = rtnl_dereference(t->next);
1573                         }
1574                 }
1575         }
1576 }
1577
1578 static int __net_init ip6gre_init_net(struct net *net)
1579 {
1580         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1581         int err;
1582
1583         if (!net_has_fallback_tunnels(net))
1584                 return 0;
1585         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1586                                           NET_NAME_UNKNOWN,
1587                                           ip6gre_tunnel_setup);
1588         if (!ign->fb_tunnel_dev) {
1589                 err = -ENOMEM;
1590                 goto err_alloc_dev;
1591         }
1592         dev_net_set(ign->fb_tunnel_dev, net);
1593         /* FB netdevice is special: we have one, and only one per netns.
1594          * Allowing to move it to another netns is clearly unsafe.
1595          */
1596         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1597
1598
1599         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1600         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1601
1602         err = register_netdev(ign->fb_tunnel_dev);
1603         if (err)
1604                 goto err_reg_dev;
1605
1606         rcu_assign_pointer(ign->tunnels_wc[0],
1607                            netdev_priv(ign->fb_tunnel_dev));
1608         return 0;
1609
1610 err_reg_dev:
1611         free_netdev(ign->fb_tunnel_dev);
1612 err_alloc_dev:
1613         return err;
1614 }
1615
1616 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1617 {
1618         struct net *net;
1619         LIST_HEAD(list);
1620
1621         rtnl_lock();
1622         list_for_each_entry(net, net_list, exit_list)
1623                 ip6gre_destroy_tunnels(net, &list);
1624         unregister_netdevice_many(&list);
1625         rtnl_unlock();
1626 }
1627
1628 static struct pernet_operations ip6gre_net_ops = {
1629         .init = ip6gre_init_net,
1630         .exit_batch = ip6gre_exit_batch_net,
1631         .id   = &ip6gre_net_id,
1632         .size = sizeof(struct ip6gre_net),
1633 };
1634
1635 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1636                                   struct netlink_ext_ack *extack)
1637 {
1638         __be16 flags;
1639
1640         if (!data)
1641                 return 0;
1642
1643         flags = 0;
1644         if (data[IFLA_GRE_IFLAGS])
1645                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1646         if (data[IFLA_GRE_OFLAGS])
1647                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1648         if (flags & (GRE_VERSION|GRE_ROUTING))
1649                 return -EINVAL;
1650
1651         return 0;
1652 }
1653
1654 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1655                                struct netlink_ext_ack *extack)
1656 {
1657         struct in6_addr daddr;
1658
1659         if (tb[IFLA_ADDRESS]) {
1660                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1661                         return -EINVAL;
1662                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1663                         return -EADDRNOTAVAIL;
1664         }
1665
1666         if (!data)
1667                 goto out;
1668
1669         if (data[IFLA_GRE_REMOTE]) {
1670                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1671                 if (ipv6_addr_any(&daddr))
1672                         return -EINVAL;
1673         }
1674
1675 out:
1676         return ip6gre_tunnel_validate(tb, data, extack);
1677 }
1678
1679 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1680                                   struct netlink_ext_ack *extack)
1681 {
1682         __be16 flags = 0;
1683         int ret, ver = 0;
1684
1685         if (!data)
1686                 return 0;
1687
1688         ret = ip6gre_tap_validate(tb, data, extack);
1689         if (ret)
1690                 return ret;
1691
1692         /* ERSPAN should only have GRE sequence and key flag */
1693         if (data[IFLA_GRE_OFLAGS])
1694                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1695         if (data[IFLA_GRE_IFLAGS])
1696                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1697         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1698             flags != (GRE_SEQ | GRE_KEY))
1699                 return -EINVAL;
1700
1701         /* ERSPAN Session ID only has 10-bit. Since we reuse
1702          * 32-bit key field as ID, check it's range.
1703          */
1704         if (data[IFLA_GRE_IKEY] &&
1705             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1706                 return -EINVAL;
1707
1708         if (data[IFLA_GRE_OKEY] &&
1709             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1710                 return -EINVAL;
1711
1712         if (data[IFLA_GRE_ERSPAN_VER]) {
1713                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1714                 if (ver != 1 && ver != 2)
1715                         return -EINVAL;
1716         }
1717
1718         if (ver == 1) {
1719                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1720                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1721
1722                         if (index & ~INDEX_MASK)
1723                                 return -EINVAL;
1724                 }
1725         } else if (ver == 2) {
1726                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1727                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1728
1729                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1730                                 return -EINVAL;
1731                 }
1732
1733                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1734                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1735
1736                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1737                                 return -EINVAL;
1738                 }
1739         }
1740
1741         return 0;
1742 }
1743
1744 static void ip6erspan_set_version(struct nlattr *data[],
1745                                   struct __ip6_tnl_parm *parms)
1746 {
1747         if (!data)
1748                 return;
1749
1750         parms->erspan_ver = 1;
1751         if (data[IFLA_GRE_ERSPAN_VER])
1752                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1753
1754         if (parms->erspan_ver == 1) {
1755                 if (data[IFLA_GRE_ERSPAN_INDEX])
1756                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1757         } else if (parms->erspan_ver == 2) {
1758                 if (data[IFLA_GRE_ERSPAN_DIR])
1759                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1760                 if (data[IFLA_GRE_ERSPAN_HWID])
1761                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1762         }
1763 }
1764
1765 static void ip6gre_netlink_parms(struct nlattr *data[],
1766                                 struct __ip6_tnl_parm *parms)
1767 {
1768         memset(parms, 0, sizeof(*parms));
1769
1770         if (!data)
1771                 return;
1772
1773         if (data[IFLA_GRE_LINK])
1774                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1775
1776         if (data[IFLA_GRE_IFLAGS])
1777                 parms->i_flags = gre_flags_to_tnl_flags(
1778                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1779
1780         if (data[IFLA_GRE_OFLAGS])
1781                 parms->o_flags = gre_flags_to_tnl_flags(
1782                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1783
1784         if (data[IFLA_GRE_IKEY])
1785                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1786
1787         if (data[IFLA_GRE_OKEY])
1788                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1789
1790         if (data[IFLA_GRE_LOCAL])
1791                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1792
1793         if (data[IFLA_GRE_REMOTE])
1794                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1795
1796         if (data[IFLA_GRE_TTL])
1797                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1798
1799         if (data[IFLA_GRE_ENCAP_LIMIT])
1800                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1801
1802         if (data[IFLA_GRE_FLOWINFO])
1803                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1804
1805         if (data[IFLA_GRE_FLAGS])
1806                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1807
1808         if (data[IFLA_GRE_FWMARK])
1809                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1810
1811         if (data[IFLA_GRE_COLLECT_METADATA])
1812                 parms->collect_md = true;
1813 }
1814
1815 static int ip6gre_tap_init(struct net_device *dev)
1816 {
1817         int ret;
1818
1819         ret = ip6gre_tunnel_init_common(dev);
1820         if (ret)
1821                 return ret;
1822
1823         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1824
1825         return 0;
1826 }
1827
1828 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1829         .ndo_init = ip6gre_tap_init,
1830         .ndo_uninit = ip6gre_tunnel_uninit,
1831         .ndo_start_xmit = ip6gre_tunnel_xmit,
1832         .ndo_set_mac_address = eth_mac_addr,
1833         .ndo_validate_addr = eth_validate_addr,
1834         .ndo_change_mtu = ip6_tnl_change_mtu,
1835         .ndo_get_stats64 = ip_tunnel_get_stats64,
1836         .ndo_get_iflink = ip6_tnl_get_iflink,
1837 };
1838
1839 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1840 {
1841         int t_hlen;
1842
1843         tunnel->tun_hlen = 8;
1844         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1845                        erspan_hdr_len(tunnel->parms.erspan_ver);
1846
1847         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1848         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1849         return t_hlen;
1850 }
1851
1852 static int ip6erspan_tap_init(struct net_device *dev)
1853 {
1854         struct ip6_tnl *tunnel;
1855         int t_hlen;
1856         int ret;
1857
1858         tunnel = netdev_priv(dev);
1859
1860         tunnel->dev = dev;
1861         tunnel->net = dev_net(dev);
1862         strcpy(tunnel->parms.name, dev->name);
1863
1864         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1865         if (!dev->tstats)
1866                 return -ENOMEM;
1867
1868         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1869         if (ret)
1870                 goto cleanup_alloc_pcpu_stats;
1871
1872         ret = gro_cells_init(&tunnel->gro_cells, dev);
1873         if (ret)
1874                 goto cleanup_dst_cache_init;
1875
1876         t_hlen = ip6erspan_calc_hlen(tunnel);
1877         dev->mtu = ETH_DATA_LEN - t_hlen;
1878         if (dev->type == ARPHRD_ETHER)
1879                 dev->mtu -= ETH_HLEN;
1880         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1881                 dev->mtu -= 8;
1882
1883         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1884         ip6erspan_tnl_link_config(tunnel, 1);
1885
1886         return 0;
1887
1888 cleanup_dst_cache_init:
1889         dst_cache_destroy(&tunnel->dst_cache);
1890 cleanup_alloc_pcpu_stats:
1891         free_percpu(dev->tstats);
1892         dev->tstats = NULL;
1893         return ret;
1894 }
1895
1896 static const struct net_device_ops ip6erspan_netdev_ops = {
1897         .ndo_init =             ip6erspan_tap_init,
1898         .ndo_uninit =           ip6erspan_tunnel_uninit,
1899         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1900         .ndo_set_mac_address =  eth_mac_addr,
1901         .ndo_validate_addr =    eth_validate_addr,
1902         .ndo_change_mtu =       ip6_tnl_change_mtu,
1903         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1904         .ndo_get_iflink =       ip6_tnl_get_iflink,
1905 };
1906
1907 static void ip6gre_tap_setup(struct net_device *dev)
1908 {
1909
1910         ether_setup(dev);
1911
1912         dev->max_mtu = 0;
1913         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1914         dev->needs_free_netdev = true;
1915         dev->priv_destructor = ip6gre_dev_free;
1916
1917         dev->features |= NETIF_F_NETNS_LOCAL;
1918         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1919         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1920         netif_keep_dst(dev);
1921 }
1922
1923 bool is_ip6gretap_dev(const struct net_device *dev)
1924 {
1925         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1926 }
1927 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1928
1929 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1930                                        struct ip_tunnel_encap *ipencap)
1931 {
1932         bool ret = false;
1933
1934         memset(ipencap, 0, sizeof(*ipencap));
1935
1936         if (!data)
1937                 return ret;
1938
1939         if (data[IFLA_GRE_ENCAP_TYPE]) {
1940                 ret = true;
1941                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1942         }
1943
1944         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1945                 ret = true;
1946                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1947         }
1948
1949         if (data[IFLA_GRE_ENCAP_SPORT]) {
1950                 ret = true;
1951                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1952         }
1953
1954         if (data[IFLA_GRE_ENCAP_DPORT]) {
1955                 ret = true;
1956                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1957         }
1958
1959         return ret;
1960 }
1961
1962 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1963                                  struct nlattr *tb[], struct nlattr *data[],
1964                                  struct netlink_ext_ack *extack)
1965 {
1966         struct ip6_tnl *nt;
1967         struct ip_tunnel_encap ipencap;
1968         int err;
1969
1970         nt = netdev_priv(dev);
1971
1972         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1973                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1974
1975                 if (err < 0)
1976                         return err;
1977         }
1978
1979         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1980                 eth_hw_addr_random(dev);
1981
1982         nt->dev = dev;
1983         nt->net = dev_net(dev);
1984
1985         err = register_netdevice(dev);
1986         if (err)
1987                 goto out;
1988
1989         if (tb[IFLA_MTU])
1990                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1991
1992         dev_hold(dev);
1993
1994 out:
1995         return err;
1996 }
1997
1998 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1999                           struct nlattr *tb[], struct nlattr *data[],
2000                           struct netlink_ext_ack *extack)
2001 {
2002         struct ip6_tnl *nt = netdev_priv(dev);
2003         struct net *net = dev_net(dev);
2004         struct ip6gre_net *ign;
2005         int err;
2006
2007         ip6gre_netlink_parms(data, &nt->parms);
2008         ign = net_generic(net, ip6gre_net_id);
2009
2010         if (nt->parms.collect_md) {
2011                 if (rtnl_dereference(ign->collect_md_tun))
2012                         return -EEXIST;
2013         } else {
2014                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2015                         return -EEXIST;
2016         }
2017
2018         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2019         if (!err) {
2020                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2021                 ip6gre_tunnel_link_md(ign, nt);
2022                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2023         }
2024         return err;
2025 }
2026
2027 static struct ip6_tnl *
2028 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2029                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2030                          struct netlink_ext_ack *extack)
2031 {
2032         struct ip6_tnl *t, *nt = netdev_priv(dev);
2033         struct net *net = nt->net;
2034         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2035         struct ip_tunnel_encap ipencap;
2036
2037         if (dev == ign->fb_tunnel_dev)
2038                 return ERR_PTR(-EINVAL);
2039
2040         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2041                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2042
2043                 if (err < 0)
2044                         return ERR_PTR(err);
2045         }
2046
2047         ip6gre_netlink_parms(data, p_p);
2048
2049         t = ip6gre_tunnel_locate(net, p_p, 0);
2050
2051         if (t) {
2052                 if (t->dev != dev)
2053                         return ERR_PTR(-EEXIST);
2054         } else {
2055                 t = nt;
2056         }
2057
2058         return t;
2059 }
2060
2061 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2062                              struct nlattr *data[],
2063                              struct netlink_ext_ack *extack)
2064 {
2065         struct ip6_tnl *t = netdev_priv(dev);
2066         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2067         struct __ip6_tnl_parm p;
2068
2069         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2070         if (IS_ERR(t))
2071                 return PTR_ERR(t);
2072
2073         ip6gre_tunnel_unlink_md(ign, t);
2074         ip6gre_tunnel_unlink(ign, t);
2075         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2076         ip6gre_tunnel_link_md(ign, t);
2077         ip6gre_tunnel_link(ign, t);
2078         return 0;
2079 }
2080
2081 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2082 {
2083         struct net *net = dev_net(dev);
2084         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2085
2086         if (dev != ign->fb_tunnel_dev)
2087                 unregister_netdevice_queue(dev, head);
2088 }
2089
2090 static size_t ip6gre_get_size(const struct net_device *dev)
2091 {
2092         return
2093                 /* IFLA_GRE_LINK */
2094                 nla_total_size(4) +
2095                 /* IFLA_GRE_IFLAGS */
2096                 nla_total_size(2) +
2097                 /* IFLA_GRE_OFLAGS */
2098                 nla_total_size(2) +
2099                 /* IFLA_GRE_IKEY */
2100                 nla_total_size(4) +
2101                 /* IFLA_GRE_OKEY */
2102                 nla_total_size(4) +
2103                 /* IFLA_GRE_LOCAL */
2104                 nla_total_size(sizeof(struct in6_addr)) +
2105                 /* IFLA_GRE_REMOTE */
2106                 nla_total_size(sizeof(struct in6_addr)) +
2107                 /* IFLA_GRE_TTL */
2108                 nla_total_size(1) +
2109                 /* IFLA_GRE_ENCAP_LIMIT */
2110                 nla_total_size(1) +
2111                 /* IFLA_GRE_FLOWINFO */
2112                 nla_total_size(4) +
2113                 /* IFLA_GRE_FLAGS */
2114                 nla_total_size(4) +
2115                 /* IFLA_GRE_ENCAP_TYPE */
2116                 nla_total_size(2) +
2117                 /* IFLA_GRE_ENCAP_FLAGS */
2118                 nla_total_size(2) +
2119                 /* IFLA_GRE_ENCAP_SPORT */
2120                 nla_total_size(2) +
2121                 /* IFLA_GRE_ENCAP_DPORT */
2122                 nla_total_size(2) +
2123                 /* IFLA_GRE_COLLECT_METADATA */
2124                 nla_total_size(0) +
2125                 /* IFLA_GRE_FWMARK */
2126                 nla_total_size(4) +
2127                 /* IFLA_GRE_ERSPAN_INDEX */
2128                 nla_total_size(4) +
2129                 0;
2130 }
2131
2132 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2133 {
2134         struct ip6_tnl *t = netdev_priv(dev);
2135         struct __ip6_tnl_parm *p = &t->parms;
2136         __be16 o_flags = p->o_flags;
2137
2138         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2139                 if (!p->collect_md)
2140                         o_flags |= TUNNEL_KEY;
2141
2142                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2143                         goto nla_put_failure;
2144
2145                 if (p->erspan_ver == 1) {
2146                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2147                                 goto nla_put_failure;
2148                 } else {
2149                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2150                                 goto nla_put_failure;
2151                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2152                                 goto nla_put_failure;
2153                 }
2154         }
2155
2156         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2157             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2158                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2159             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2160                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2161             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2162             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2163             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2164             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2165             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2166             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2167             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2168             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2169             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2170                 goto nla_put_failure;
2171
2172         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2173                         t->encap.type) ||
2174             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2175                          t->encap.sport) ||
2176             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2177                          t->encap.dport) ||
2178             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2179                         t->encap.flags))
2180                 goto nla_put_failure;
2181
2182         if (p->collect_md) {
2183                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2184                         goto nla_put_failure;
2185         }
2186
2187         return 0;
2188
2189 nla_put_failure:
2190         return -EMSGSIZE;
2191 }
2192
2193 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2194         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2195         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2196         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2197         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2198         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2199         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2200         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2201         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2202         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2203         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2204         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2205         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2206         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2207         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2208         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2209         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2210         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2211         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2212         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2213         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2214         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2215 };
2216
2217 static void ip6erspan_tap_setup(struct net_device *dev)
2218 {
2219         ether_setup(dev);
2220
2221         dev->netdev_ops = &ip6erspan_netdev_ops;
2222         dev->needs_free_netdev = true;
2223         dev->priv_destructor = ip6gre_dev_free;
2224
2225         dev->features |= NETIF_F_NETNS_LOCAL;
2226         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2227         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2228         netif_keep_dst(dev);
2229 }
2230
2231 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2232                              struct nlattr *tb[], struct nlattr *data[],
2233                              struct netlink_ext_ack *extack)
2234 {
2235         struct ip6_tnl *nt = netdev_priv(dev);
2236         struct net *net = dev_net(dev);
2237         struct ip6gre_net *ign;
2238         int err;
2239
2240         ip6gre_netlink_parms(data, &nt->parms);
2241         ip6erspan_set_version(data, &nt->parms);
2242         ign = net_generic(net, ip6gre_net_id);
2243
2244         if (nt->parms.collect_md) {
2245                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2246                         return -EEXIST;
2247         } else {
2248                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2249                         return -EEXIST;
2250         }
2251
2252         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2253         if (!err) {
2254                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2255                 ip6erspan_tunnel_link_md(ign, nt);
2256                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2257         }
2258         return err;
2259 }
2260
2261 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2262 {
2263         ip6gre_tnl_link_config_common(t);
2264         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2265 }
2266
2267 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2268                                 const struct __ip6_tnl_parm *p, int set_mtu)
2269 {
2270         ip6gre_tnl_copy_tnl_parm(t, p);
2271         ip6erspan_tnl_link_config(t, set_mtu);
2272         return 0;
2273 }
2274
2275 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2276                                 struct nlattr *data[],
2277                                 struct netlink_ext_ack *extack)
2278 {
2279         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2280         struct __ip6_tnl_parm p;
2281         struct ip6_tnl *t;
2282
2283         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2284         if (IS_ERR(t))
2285                 return PTR_ERR(t);
2286
2287         ip6erspan_set_version(data, &p);
2288         ip6gre_tunnel_unlink_md(ign, t);
2289         ip6gre_tunnel_unlink(ign, t);
2290         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2291         ip6erspan_tunnel_link_md(ign, t);
2292         ip6gre_tunnel_link(ign, t);
2293         return 0;
2294 }
2295
2296 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2297         .kind           = "ip6gre",
2298         .maxtype        = IFLA_GRE_MAX,
2299         .policy         = ip6gre_policy,
2300         .priv_size      = sizeof(struct ip6_tnl),
2301         .setup          = ip6gre_tunnel_setup,
2302         .validate       = ip6gre_tunnel_validate,
2303         .newlink        = ip6gre_newlink,
2304         .changelink     = ip6gre_changelink,
2305         .dellink        = ip6gre_dellink,
2306         .get_size       = ip6gre_get_size,
2307         .fill_info      = ip6gre_fill_info,
2308         .get_link_net   = ip6_tnl_get_link_net,
2309 };
2310
2311 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2312         .kind           = "ip6gretap",
2313         .maxtype        = IFLA_GRE_MAX,
2314         .policy         = ip6gre_policy,
2315         .priv_size      = sizeof(struct ip6_tnl),
2316         .setup          = ip6gre_tap_setup,
2317         .validate       = ip6gre_tap_validate,
2318         .newlink        = ip6gre_newlink,
2319         .changelink     = ip6gre_changelink,
2320         .get_size       = ip6gre_get_size,
2321         .fill_info      = ip6gre_fill_info,
2322         .get_link_net   = ip6_tnl_get_link_net,
2323 };
2324
2325 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2326         .kind           = "ip6erspan",
2327         .maxtype        = IFLA_GRE_MAX,
2328         .policy         = ip6gre_policy,
2329         .priv_size      = sizeof(struct ip6_tnl),
2330         .setup          = ip6erspan_tap_setup,
2331         .validate       = ip6erspan_tap_validate,
2332         .newlink        = ip6erspan_newlink,
2333         .changelink     = ip6erspan_changelink,
2334         .get_size       = ip6gre_get_size,
2335         .fill_info      = ip6gre_fill_info,
2336         .get_link_net   = ip6_tnl_get_link_net,
2337 };
2338
2339 /*
2340  *      And now the modules code and kernel interface.
2341  */
2342
2343 static int __init ip6gre_init(void)
2344 {
2345         int err;
2346
2347         pr_info("GRE over IPv6 tunneling driver\n");
2348
2349         err = register_pernet_device(&ip6gre_net_ops);
2350         if (err < 0)
2351                 return err;
2352
2353         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2354         if (err < 0) {
2355                 pr_info("%s: can't add protocol\n", __func__);
2356                 goto add_proto_failed;
2357         }
2358
2359         err = rtnl_link_register(&ip6gre_link_ops);
2360         if (err < 0)
2361                 goto rtnl_link_failed;
2362
2363         err = rtnl_link_register(&ip6gre_tap_ops);
2364         if (err < 0)
2365                 goto tap_ops_failed;
2366
2367         err = rtnl_link_register(&ip6erspan_tap_ops);
2368         if (err < 0)
2369                 goto erspan_link_failed;
2370
2371 out:
2372         return err;
2373
2374 erspan_link_failed:
2375         rtnl_link_unregister(&ip6gre_tap_ops);
2376 tap_ops_failed:
2377         rtnl_link_unregister(&ip6gre_link_ops);
2378 rtnl_link_failed:
2379         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2380 add_proto_failed:
2381         unregister_pernet_device(&ip6gre_net_ops);
2382         goto out;
2383 }
2384
2385 static void __exit ip6gre_fini(void)
2386 {
2387         rtnl_link_unregister(&ip6gre_tap_ops);
2388         rtnl_link_unregister(&ip6gre_link_ops);
2389         rtnl_link_unregister(&ip6erspan_tap_ops);
2390         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2391         unregister_pernet_device(&ip6gre_net_ops);
2392 }
2393
2394 module_init(ip6gre_init);
2395 module_exit(ip6gre_fini);
2396 MODULE_LICENSE("GPL");
2397 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2398 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2399 MODULE_ALIAS_RTNL_LINK("ip6gre");
2400 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2401 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2402 MODULE_ALIAS_NETDEV("ip6gre0");