Merge tag 'zynqmp-soc-for-v5.7' of https://github.com/Xilinx/linux-xlnx into arm/soc
[linux-2.6-microblaze.git] / net / ipv6 / ip6_gre.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      GRE over IPv6 protocol decoder.
4  *
5  *      Authors: Dmitry Kozlov (xeb@mail.ru)
6  */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
18 #include <linux/in.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/rtnetlink.h>
47
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/gre.h>
53 #include <net/erspan.h>
54 #include <net/dst_metadata.h>
55
56
57 static bool log_ecn_error = true;
58 module_param(log_ecn_error, bool, 0644);
59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60
61 #define IP6_GRE_HASH_SIZE_SHIFT  5
62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63
64 static unsigned int ip6gre_net_id __read_mostly;
65 struct ip6gre_net {
66         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67
68         struct ip6_tnl __rcu *collect_md_tun;
69         struct ip6_tnl __rcu *collect_md_tun_erspan;
70         struct net_device *fb_tunnel_dev;
71 };
72
73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85    4 hash tables:
86
87    3: (remote,local)
88    2: (remote,*)
89    1: (*,local)
90    0: (*,*)
91
92    We require exact key match i.e. if a key is present in packet
93    it will match only tunnel with the same key; if it is not present,
94    it will match only keyless tunnel.
95
96    All keysless packets, if not matched configured keyless tunnels
97    will match fallback tunnel.
98  */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103         u32 hash = ipv6_addr_hash(addr);
104
105         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l     tunnels[3]
109 #define tunnels_r       tunnels[2]
110 #define tunnels_l       tunnels[1]
111 #define tunnels_wc      tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116                 const struct in6_addr *remote, const struct in6_addr *local,
117                 __be32 key, __be16 gre_proto)
118 {
119         struct net *net = dev_net(dev);
120         int link = dev->ifindex;
121         unsigned int h0 = HASH_ADDR(remote);
122         unsigned int h1 = HASH_KEY(key);
123         struct ip6_tnl *t, *cand = NULL;
124         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126                         gre_proto == htons(ETH_P_ERSPAN) ||
127                         gre_proto == htons(ETH_P_ERSPAN2)) ?
128                        ARPHRD_ETHER : ARPHRD_IP6GRE;
129         int score, cand_score = 4;
130
131         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
132                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
133                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
134                     key != t->parms.i_key ||
135                     !(t->dev->flags & IFF_UP))
136                         continue;
137
138                 if (t->dev->type != ARPHRD_IP6GRE &&
139                     t->dev->type != dev_type)
140                         continue;
141
142                 score = 0;
143                 if (t->parms.link != link)
144                         score |= 1;
145                 if (t->dev->type != dev_type)
146                         score |= 2;
147                 if (score == 0)
148                         return t;
149
150                 if (score < cand_score) {
151                         cand = t;
152                         cand_score = score;
153                 }
154         }
155
156         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
157                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
158                     key != t->parms.i_key ||
159                     !(t->dev->flags & IFF_UP))
160                         continue;
161
162                 if (t->dev->type != ARPHRD_IP6GRE &&
163                     t->dev->type != dev_type)
164                         continue;
165
166                 score = 0;
167                 if (t->parms.link != link)
168                         score |= 1;
169                 if (t->dev->type != dev_type)
170                         score |= 2;
171                 if (score == 0)
172                         return t;
173
174                 if (score < cand_score) {
175                         cand = t;
176                         cand_score = score;
177                 }
178         }
179
180         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
181                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
182                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
183                                  !ipv6_addr_is_multicast(local))) ||
184                     key != t->parms.i_key ||
185                     !(t->dev->flags & IFF_UP))
186                         continue;
187
188                 if (t->dev->type != ARPHRD_IP6GRE &&
189                     t->dev->type != dev_type)
190                         continue;
191
192                 score = 0;
193                 if (t->parms.link != link)
194                         score |= 1;
195                 if (t->dev->type != dev_type)
196                         score |= 2;
197                 if (score == 0)
198                         return t;
199
200                 if (score < cand_score) {
201                         cand = t;
202                         cand_score = score;
203                 }
204         }
205
206         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
207                 if (t->parms.i_key != key ||
208                     !(t->dev->flags & IFF_UP))
209                         continue;
210
211                 if (t->dev->type != ARPHRD_IP6GRE &&
212                     t->dev->type != dev_type)
213                         continue;
214
215                 score = 0;
216                 if (t->parms.link != link)
217                         score |= 1;
218                 if (t->dev->type != dev_type)
219                         score |= 2;
220                 if (score == 0)
221                         return t;
222
223                 if (score < cand_score) {
224                         cand = t;
225                         cand_score = score;
226                 }
227         }
228
229         if (cand)
230                 return cand;
231
232         if (gre_proto == htons(ETH_P_ERSPAN) ||
233             gre_proto == htons(ETH_P_ERSPAN2))
234                 t = rcu_dereference(ign->collect_md_tun_erspan);
235         else
236                 t = rcu_dereference(ign->collect_md_tun);
237
238         if (t && t->dev->flags & IFF_UP)
239                 return t;
240
241         dev = ign->fb_tunnel_dev;
242         if (dev && dev->flags & IFF_UP)
243                 return netdev_priv(dev);
244
245         return NULL;
246 }
247
248 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
249                 const struct __ip6_tnl_parm *p)
250 {
251         const struct in6_addr *remote = &p->raddr;
252         const struct in6_addr *local = &p->laddr;
253         unsigned int h = HASH_KEY(p->i_key);
254         int prio = 0;
255
256         if (!ipv6_addr_any(local))
257                 prio |= 1;
258         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
259                 prio |= 2;
260                 h ^= HASH_ADDR(remote);
261         }
262
263         return &ign->tunnels[prio][h];
264 }
265
266 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
267 {
268         if (t->parms.collect_md)
269                 rcu_assign_pointer(ign->collect_md_tun, t);
270 }
271
272 static void ip6erspan_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_erspan, t);
276 }
277
278 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
279 {
280         if (t->parms.collect_md)
281                 rcu_assign_pointer(ign->collect_md_tun, NULL);
282 }
283
284 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
285                                        struct ip6_tnl *t)
286 {
287         if (t->parms.collect_md)
288                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
289 }
290
291 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
292                 const struct ip6_tnl *t)
293 {
294         return __ip6gre_bucket(ign, &t->parms);
295 }
296
297 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
298 {
299         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
300
301         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
302         rcu_assign_pointer(*tp, t);
303 }
304
305 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
306 {
307         struct ip6_tnl __rcu **tp;
308         struct ip6_tnl *iter;
309
310         for (tp = ip6gre_bucket(ign, t);
311              (iter = rtnl_dereference(*tp)) != NULL;
312              tp = &iter->next) {
313                 if (t == iter) {
314                         rcu_assign_pointer(*tp, t->next);
315                         break;
316                 }
317         }
318 }
319
320 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
321                                            const struct __ip6_tnl_parm *parms,
322                                            int type)
323 {
324         const struct in6_addr *remote = &parms->raddr;
325         const struct in6_addr *local = &parms->laddr;
326         __be32 key = parms->i_key;
327         int link = parms->link;
328         struct ip6_tnl *t;
329         struct ip6_tnl __rcu **tp;
330         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
331
332         for (tp = __ip6gre_bucket(ign, parms);
333              (t = rtnl_dereference(*tp)) != NULL;
334              tp = &t->next)
335                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
336                     ipv6_addr_equal(remote, &t->parms.raddr) &&
337                     key == t->parms.i_key &&
338                     link == t->parms.link &&
339                     type == t->dev->type)
340                         break;
341
342         return t;
343 }
344
345 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
346                 const struct __ip6_tnl_parm *parms, int create)
347 {
348         struct ip6_tnl *t, *nt;
349         struct net_device *dev;
350         char name[IFNAMSIZ];
351         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
352
353         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
354         if (t && create)
355                 return NULL;
356         if (t || !create)
357                 return t;
358
359         if (parms->name[0]) {
360                 if (!dev_valid_name(parms->name))
361                         return NULL;
362                 strlcpy(name, parms->name, IFNAMSIZ);
363         } else {
364                 strcpy(name, "ip6gre%d");
365         }
366         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
367                            ip6gre_tunnel_setup);
368         if (!dev)
369                 return NULL;
370
371         dev_net_set(dev, net);
372
373         nt = netdev_priv(dev);
374         nt->parms = *parms;
375         dev->rtnl_link_ops = &ip6gre_link_ops;
376
377         nt->dev = dev;
378         nt->net = dev_net(dev);
379
380         if (register_netdevice(dev) < 0)
381                 goto failed_free;
382
383         ip6gre_tnl_link_config(nt, 1);
384
385         /* Can use a lockless transmit, unless we generate output sequences */
386         if (!(nt->parms.o_flags & TUNNEL_SEQ))
387                 dev->features |= NETIF_F_LLTX;
388
389         dev_hold(dev);
390         ip6gre_tunnel_link(ign, nt);
391         return nt;
392
393 failed_free:
394         free_netdev(dev);
395         return NULL;
396 }
397
398 static void ip6erspan_tunnel_uninit(struct net_device *dev)
399 {
400         struct ip6_tnl *t = netdev_priv(dev);
401         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
402
403         ip6erspan_tunnel_unlink_md(ign, t);
404         ip6gre_tunnel_unlink(ign, t);
405         dst_cache_reset(&t->dst_cache);
406         dev_put(dev);
407 }
408
409 static void ip6gre_tunnel_uninit(struct net_device *dev)
410 {
411         struct ip6_tnl *t = netdev_priv(dev);
412         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
413
414         ip6gre_tunnel_unlink_md(ign, t);
415         ip6gre_tunnel_unlink(ign, t);
416         dst_cache_reset(&t->dst_cache);
417         dev_put(dev);
418 }
419
420
421 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
422                        u8 type, u8 code, int offset, __be32 info)
423 {
424         struct net *net = dev_net(skb->dev);
425         const struct ipv6hdr *ipv6h;
426         struct tnl_ptk_info tpi;
427         struct ip6_tnl *t;
428
429         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
430                              offset) < 0)
431                 return -EINVAL;
432
433         ipv6h = (const struct ipv6hdr *)skb->data;
434         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
435                                  tpi.key, tpi.proto);
436         if (!t)
437                 return -ENOENT;
438
439         switch (type) {
440         case ICMPV6_DEST_UNREACH:
441                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
442                                     t->parms.name);
443                 if (code != ICMPV6_PORT_UNREACH)
444                         break;
445                 return 0;
446         case ICMPV6_TIME_EXCEED:
447                 if (code == ICMPV6_EXC_HOPLIMIT) {
448                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
449                                             t->parms.name);
450                         break;
451                 }
452                 return 0;
453         case ICMPV6_PARAMPROB: {
454                 struct ipv6_tlv_tnl_enc_lim *tel;
455                 __u32 teli;
456
457                 teli = 0;
458                 if (code == ICMPV6_HDR_FIELD)
459                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
460
461                 if (teli && teli == be32_to_cpu(info) - 2) {
462                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
463                         if (tel->encap_limit == 0) {
464                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
465                                                     t->parms.name);
466                         }
467                 } else {
468                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
469                                             t->parms.name);
470                 }
471                 return 0;
472         }
473         case ICMPV6_PKT_TOOBIG:
474                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
475                 return 0;
476         case NDISC_REDIRECT:
477                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
478                              sock_net_uid(net, NULL));
479                 return 0;
480         }
481
482         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
483                 t->err_count++;
484         else
485                 t->err_count = 1;
486         t->err_time = jiffies;
487
488         return 0;
489 }
490
491 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
492 {
493         const struct ipv6hdr *ipv6h;
494         struct ip6_tnl *tunnel;
495
496         ipv6h = ipv6_hdr(skb);
497         tunnel = ip6gre_tunnel_lookup(skb->dev,
498                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
499                                       tpi->proto);
500         if (tunnel) {
501                 if (tunnel->parms.collect_md) {
502                         struct metadata_dst *tun_dst;
503                         __be64 tun_id;
504                         __be16 flags;
505
506                         flags = tpi->flags;
507                         tun_id = key32_to_tunnel_id(tpi->key);
508
509                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
510                         if (!tun_dst)
511                                 return PACKET_REJECT;
512
513                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
514                 } else {
515                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
516                 }
517
518                 return PACKET_RCVD;
519         }
520
521         return PACKET_REJECT;
522 }
523
524 static int ip6erspan_rcv(struct sk_buff *skb,
525                          struct tnl_ptk_info *tpi,
526                          int gre_hdr_len)
527 {
528         struct erspan_base_hdr *ershdr;
529         const struct ipv6hdr *ipv6h;
530         struct erspan_md2 *md2;
531         struct ip6_tnl *tunnel;
532         u8 ver;
533
534         ipv6h = ipv6_hdr(skb);
535         ershdr = (struct erspan_base_hdr *)skb->data;
536         ver = ershdr->ver;
537
538         tunnel = ip6gre_tunnel_lookup(skb->dev,
539                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
540                                       tpi->proto);
541         if (tunnel) {
542                 int len = erspan_hdr_len(ver);
543
544                 if (unlikely(!pskb_may_pull(skb, len)))
545                         return PACKET_REJECT;
546
547                 if (__iptunnel_pull_header(skb, len,
548                                            htons(ETH_P_TEB),
549                                            false, false) < 0)
550                         return PACKET_REJECT;
551
552                 if (tunnel->parms.collect_md) {
553                         struct erspan_metadata *pkt_md, *md;
554                         struct metadata_dst *tun_dst;
555                         struct ip_tunnel_info *info;
556                         unsigned char *gh;
557                         __be64 tun_id;
558                         __be16 flags;
559
560                         tpi->flags |= TUNNEL_KEY;
561                         flags = tpi->flags;
562                         tun_id = key32_to_tunnel_id(tpi->key);
563
564                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
565                                                   sizeof(*md));
566                         if (!tun_dst)
567                                 return PACKET_REJECT;
568
569                         /* skb can be uncloned in __iptunnel_pull_header, so
570                          * old pkt_md is no longer valid and we need to reset
571                          * it
572                          */
573                         gh = skb_network_header(skb) +
574                              skb_network_header_len(skb);
575                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
576                                                             sizeof(*ershdr));
577                         info = &tun_dst->u.tun_info;
578                         md = ip_tunnel_info_opts(info);
579                         md->version = ver;
580                         md2 = &md->u.md2;
581                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
582                                                        ERSPAN_V2_MDSIZE);
583                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
584                         info->options_len = sizeof(*md);
585
586                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
587
588                 } else {
589                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
590                 }
591
592                 return PACKET_RCVD;
593         }
594
595         return PACKET_REJECT;
596 }
597
598 static int gre_rcv(struct sk_buff *skb)
599 {
600         struct tnl_ptk_info tpi;
601         bool csum_err = false;
602         int hdr_len;
603
604         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
605         if (hdr_len < 0)
606                 goto drop;
607
608         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
609                 goto drop;
610
611         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
612                      tpi.proto == htons(ETH_P_ERSPAN2))) {
613                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
614                         return 0;
615                 goto out;
616         }
617
618         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
619                 return 0;
620
621 out:
622         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
623 drop:
624         kfree_skb(skb);
625         return 0;
626 }
627
628 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
629 {
630         return iptunnel_handle_offloads(skb,
631                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
632 }
633
634 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
635                                      struct net_device *dev,
636                                      struct flowi6 *fl6, __u8 *dsfield,
637                                      int *encap_limit)
638 {
639         const struct iphdr *iph = ip_hdr(skb);
640         struct ip6_tnl *t = netdev_priv(dev);
641
642         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
643                 *encap_limit = t->parms.encap_limit;
644
645         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
646
647         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
648                 *dsfield = ipv4_get_dsfield(iph);
649         else
650                 *dsfield = ip6_tclass(t->parms.flowinfo);
651
652         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
653                 fl6->flowi6_mark = skb->mark;
654         else
655                 fl6->flowi6_mark = t->parms.fwmark;
656
657         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
658 }
659
660 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
661                                     struct net_device *dev,
662                                     struct flowi6 *fl6, __u8 *dsfield,
663                                     int *encap_limit)
664 {
665         struct ipv6hdr *ipv6h;
666         struct ip6_tnl *t = netdev_priv(dev);
667         __u16 offset;
668
669         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
670         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
671         ipv6h = ipv6_hdr(skb);
672
673         if (offset > 0) {
674                 struct ipv6_tlv_tnl_enc_lim *tel;
675
676                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
677                 if (tel->encap_limit == 0) {
678                         icmpv6_send(skb, ICMPV6_PARAMPROB,
679                                     ICMPV6_HDR_FIELD, offset + 2);
680                         return -1;
681                 }
682                 *encap_limit = tel->encap_limit - 1;
683         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
684                 *encap_limit = t->parms.encap_limit;
685         }
686
687         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
688
689         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
690                 *dsfield = ipv6_get_dsfield(ipv6h);
691         else
692                 *dsfield = ip6_tclass(t->parms.flowinfo);
693
694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
695                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
696
697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
698                 fl6->flowi6_mark = skb->mark;
699         else
700                 fl6->flowi6_mark = t->parms.fwmark;
701
702         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
703
704         return 0;
705 }
706
707 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
708                                struct net_device *dev, __u8 dsfield,
709                                struct flowi6 *fl6, int encap_limit,
710                                __u32 *pmtu, __be16 proto)
711 {
712         struct ip6_tnl *tunnel = netdev_priv(dev);
713         __be16 protocol;
714
715         if (dev->type == ARPHRD_ETHER)
716                 IPCB(skb)->flags = 0;
717
718         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
719                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
720         else
721                 fl6->daddr = tunnel->parms.raddr;
722
723         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
724                 return -ENOMEM;
725
726         /* Push GRE header. */
727         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
728
729         if (tunnel->parms.collect_md) {
730                 struct ip_tunnel_info *tun_info;
731                 const struct ip_tunnel_key *key;
732                 __be16 flags;
733
734                 tun_info = skb_tunnel_info(skb);
735                 if (unlikely(!tun_info ||
736                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
737                              ip_tunnel_info_af(tun_info) != AF_INET6))
738                         return -EINVAL;
739
740                 key = &tun_info->key;
741                 memset(fl6, 0, sizeof(*fl6));
742                 fl6->flowi6_proto = IPPROTO_GRE;
743                 fl6->daddr = key->u.ipv6.dst;
744                 fl6->flowlabel = key->label;
745                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
746
747                 dsfield = key->tos;
748                 flags = key->tun_flags &
749                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
750                 tunnel->tun_hlen = gre_calc_hlen(flags);
751
752                 gre_build_header(skb, tunnel->tun_hlen,
753                                  flags, protocol,
754                                  tunnel_id_to_key32(tun_info->key.tun_id),
755                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
756                                                       : 0);
757
758         } else {
759                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
760                         tunnel->o_seqno++;
761
762                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
763                                  protocol, tunnel->parms.o_key,
764                                  htonl(tunnel->o_seqno));
765         }
766
767         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
768                             NEXTHDR_GRE);
769 }
770
771 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
772 {
773         struct ip6_tnl *t = netdev_priv(dev);
774         int encap_limit = -1;
775         struct flowi6 fl6;
776         __u8 dsfield = 0;
777         __u32 mtu;
778         int err;
779
780         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
781
782         if (!t->parms.collect_md)
783                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
784                                          &dsfield, &encap_limit);
785
786         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
787         if (err)
788                 return -1;
789
790         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
791                           skb->protocol);
792         if (err != 0) {
793                 /* XXX: send ICMP error even if DF is not set. */
794                 if (err == -EMSGSIZE)
795                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
796                                   htonl(mtu));
797                 return -1;
798         }
799
800         return 0;
801 }
802
803 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
804 {
805         struct ip6_tnl *t = netdev_priv(dev);
806         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
807         int encap_limit = -1;
808         struct flowi6 fl6;
809         __u8 dsfield = 0;
810         __u32 mtu;
811         int err;
812
813         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
814                 return -1;
815
816         if (!t->parms.collect_md &&
817             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
818                 return -1;
819
820         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
821                 return -1;
822
823         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
824                           &mtu, skb->protocol);
825         if (err != 0) {
826                 if (err == -EMSGSIZE)
827                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
828                 return -1;
829         }
830
831         return 0;
832 }
833
834 /**
835  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
836  *   @t: the outgoing tunnel device
837  *   @hdr: IPv6 header from the incoming packet
838  *
839  * Description:
840  *   Avoid trivial tunneling loop by checking that tunnel exit-point
841  *   doesn't match source of incoming packet.
842  *
843  * Return:
844  *   1 if conflict,
845  *   0 else
846  **/
847
848 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
849         const struct ipv6hdr *hdr)
850 {
851         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
852 }
853
854 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
855 {
856         struct ip6_tnl *t = netdev_priv(dev);
857         int encap_limit = -1;
858         struct flowi6 fl6;
859         __u32 mtu;
860         int err;
861
862         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
863                 encap_limit = t->parms.encap_limit;
864
865         if (!t->parms.collect_md)
866                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
867
868         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
869         if (err)
870                 return err;
871
872         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
873
874         return err;
875 }
876
877 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
878         struct net_device *dev)
879 {
880         struct ip6_tnl *t = netdev_priv(dev);
881         struct net_device_stats *stats = &t->dev->stats;
882         int ret;
883
884         if (!pskb_inet_may_pull(skb))
885                 goto tx_err;
886
887         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
888                 goto tx_err;
889
890         switch (skb->protocol) {
891         case htons(ETH_P_IP):
892                 ret = ip6gre_xmit_ipv4(skb, dev);
893                 break;
894         case htons(ETH_P_IPV6):
895                 ret = ip6gre_xmit_ipv6(skb, dev);
896                 break;
897         default:
898                 ret = ip6gre_xmit_other(skb, dev);
899                 break;
900         }
901
902         if (ret < 0)
903                 goto tx_err;
904
905         return NETDEV_TX_OK;
906
907 tx_err:
908         stats->tx_errors++;
909         stats->tx_dropped++;
910         kfree_skb(skb);
911         return NETDEV_TX_OK;
912 }
913
914 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
915                                          struct net_device *dev)
916 {
917         struct ip6_tnl *t = netdev_priv(dev);
918         struct dst_entry *dst = skb_dst(skb);
919         struct net_device_stats *stats;
920         bool truncate = false;
921         int encap_limit = -1;
922         __u8 dsfield = false;
923         struct flowi6 fl6;
924         int err = -EINVAL;
925         __be16 proto;
926         __u32 mtu;
927         int nhoff;
928         int thoff;
929
930         if (!pskb_inet_may_pull(skb))
931                 goto tx_err;
932
933         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
934                 goto tx_err;
935
936         if (gre_handle_offloads(skb, false))
937                 goto tx_err;
938
939         if (skb->len > dev->mtu + dev->hard_header_len) {
940                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
941                 truncate = true;
942         }
943
944         nhoff = skb_network_header(skb) - skb_mac_header(skb);
945         if (skb->protocol == htons(ETH_P_IP) &&
946             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
947                 truncate = true;
948
949         thoff = skb_transport_header(skb) - skb_mac_header(skb);
950         if (skb->protocol == htons(ETH_P_IPV6) &&
951             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
952                 truncate = true;
953
954         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
955                 goto tx_err;
956
957         t->parms.o_flags &= ~TUNNEL_KEY;
958         IPCB(skb)->flags = 0;
959
960         /* For collect_md mode, derive fl6 from the tunnel key,
961          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
962          */
963         if (t->parms.collect_md) {
964                 struct ip_tunnel_info *tun_info;
965                 const struct ip_tunnel_key *key;
966                 struct erspan_metadata *md;
967                 __be32 tun_id;
968
969                 tun_info = skb_tunnel_info(skb);
970                 if (unlikely(!tun_info ||
971                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
972                              ip_tunnel_info_af(tun_info) != AF_INET6))
973                         goto tx_err;
974
975                 key = &tun_info->key;
976                 memset(&fl6, 0, sizeof(fl6));
977                 fl6.flowi6_proto = IPPROTO_GRE;
978                 fl6.daddr = key->u.ipv6.dst;
979                 fl6.flowlabel = key->label;
980                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
981
982                 dsfield = key->tos;
983                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
984                         goto tx_err;
985                 if (tun_info->options_len < sizeof(*md))
986                         goto tx_err;
987                 md = ip_tunnel_info_opts(tun_info);
988
989                 tun_id = tunnel_id_to_key32(key->tun_id);
990                 if (md->version == 1) {
991                         erspan_build_header(skb,
992                                             ntohl(tun_id),
993                                             ntohl(md->u.index), truncate,
994                                             false);
995                 } else if (md->version == 2) {
996                         erspan_build_header_v2(skb,
997                                                ntohl(tun_id),
998                                                md->u.md2.dir,
999                                                get_hwid(&md->u.md2),
1000                                                truncate, false);
1001                 } else {
1002                         goto tx_err;
1003                 }
1004         } else {
1005                 switch (skb->protocol) {
1006                 case htons(ETH_P_IP):
1007                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1008                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1009                                                  &dsfield, &encap_limit);
1010                         break;
1011                 case htons(ETH_P_IPV6):
1012                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1013                                 goto tx_err;
1014                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1015                                                      &dsfield, &encap_limit))
1016                                 goto tx_err;
1017                         break;
1018                 default:
1019                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1020                         break;
1021                 }
1022
1023                 if (t->parms.erspan_ver == 1)
1024                         erspan_build_header(skb, ntohl(t->parms.o_key),
1025                                             t->parms.index,
1026                                             truncate, false);
1027                 else if (t->parms.erspan_ver == 2)
1028                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1029                                                t->parms.dir,
1030                                                t->parms.hwid,
1031                                                truncate, false);
1032                 else
1033                         goto tx_err;
1034
1035                 fl6.daddr = t->parms.raddr;
1036         }
1037
1038         /* Push GRE header. */
1039         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1040                                            : htons(ETH_P_ERSPAN2);
1041         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1042
1043         /* TooBig packet may have updated dst->dev's mtu */
1044         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1045                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1046
1047         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1048                            NEXTHDR_GRE);
1049         if (err != 0) {
1050                 /* XXX: send ICMP error even if DF is not set. */
1051                 if (err == -EMSGSIZE) {
1052                         if (skb->protocol == htons(ETH_P_IP))
1053                                 icmp_send(skb, ICMP_DEST_UNREACH,
1054                                           ICMP_FRAG_NEEDED, htonl(mtu));
1055                         else
1056                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1057                 }
1058
1059                 goto tx_err;
1060         }
1061         return NETDEV_TX_OK;
1062
1063 tx_err:
1064         stats = &t->dev->stats;
1065         stats->tx_errors++;
1066         stats->tx_dropped++;
1067         kfree_skb(skb);
1068         return NETDEV_TX_OK;
1069 }
1070
1071 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1072 {
1073         struct net_device *dev = t->dev;
1074         struct __ip6_tnl_parm *p = &t->parms;
1075         struct flowi6 *fl6 = &t->fl.u.ip6;
1076
1077         if (dev->type != ARPHRD_ETHER) {
1078                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1079                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1080         }
1081
1082         /* Set up flowi template */
1083         fl6->saddr = p->laddr;
1084         fl6->daddr = p->raddr;
1085         fl6->flowi6_oif = p->link;
1086         fl6->flowlabel = 0;
1087         fl6->flowi6_proto = IPPROTO_GRE;
1088
1089         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1090                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1091         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1092                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1093
1094         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1095         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1096
1097         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1098                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1099                 dev->flags |= IFF_POINTOPOINT;
1100         else
1101                 dev->flags &= ~IFF_POINTOPOINT;
1102 }
1103
1104 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1105                                          int t_hlen)
1106 {
1107         const struct __ip6_tnl_parm *p = &t->parms;
1108         struct net_device *dev = t->dev;
1109
1110         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1111                 int strict = (ipv6_addr_type(&p->raddr) &
1112                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1113
1114                 struct rt6_info *rt = rt6_lookup(t->net,
1115                                                  &p->raddr, &p->laddr,
1116                                                  p->link, NULL, strict);
1117
1118                 if (!rt)
1119                         return;
1120
1121                 if (rt->dst.dev) {
1122                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1123                                                t_hlen;
1124
1125                         if (set_mtu) {
1126                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1127                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1128                                         dev->mtu -= 8;
1129                                 if (dev->type == ARPHRD_ETHER)
1130                                         dev->mtu -= ETH_HLEN;
1131
1132                                 if (dev->mtu < IPV6_MIN_MTU)
1133                                         dev->mtu = IPV6_MIN_MTU;
1134                         }
1135                 }
1136                 ip6_rt_put(rt);
1137         }
1138 }
1139
1140 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1141 {
1142         int t_hlen;
1143
1144         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1145         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1146
1147         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1148         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1149         return t_hlen;
1150 }
1151
1152 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1153 {
1154         ip6gre_tnl_link_config_common(t);
1155         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1156 }
1157
1158 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1159                                      const struct __ip6_tnl_parm *p)
1160 {
1161         t->parms.laddr = p->laddr;
1162         t->parms.raddr = p->raddr;
1163         t->parms.flags = p->flags;
1164         t->parms.hop_limit = p->hop_limit;
1165         t->parms.encap_limit = p->encap_limit;
1166         t->parms.flowinfo = p->flowinfo;
1167         t->parms.link = p->link;
1168         t->parms.proto = p->proto;
1169         t->parms.i_key = p->i_key;
1170         t->parms.o_key = p->o_key;
1171         t->parms.i_flags = p->i_flags;
1172         t->parms.o_flags = p->o_flags;
1173         t->parms.fwmark = p->fwmark;
1174         t->parms.erspan_ver = p->erspan_ver;
1175         t->parms.index = p->index;
1176         t->parms.dir = p->dir;
1177         t->parms.hwid = p->hwid;
1178         dst_cache_reset(&t->dst_cache);
1179 }
1180
1181 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1182                              int set_mtu)
1183 {
1184         ip6gre_tnl_copy_tnl_parm(t, p);
1185         ip6gre_tnl_link_config(t, set_mtu);
1186         return 0;
1187 }
1188
1189 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1190         const struct ip6_tnl_parm2 *u)
1191 {
1192         p->laddr = u->laddr;
1193         p->raddr = u->raddr;
1194         p->flags = u->flags;
1195         p->hop_limit = u->hop_limit;
1196         p->encap_limit = u->encap_limit;
1197         p->flowinfo = u->flowinfo;
1198         p->link = u->link;
1199         p->i_key = u->i_key;
1200         p->o_key = u->o_key;
1201         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1202         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1203         memcpy(p->name, u->name, sizeof(u->name));
1204 }
1205
1206 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1207         const struct __ip6_tnl_parm *p)
1208 {
1209         u->proto = IPPROTO_GRE;
1210         u->laddr = p->laddr;
1211         u->raddr = p->raddr;
1212         u->flags = p->flags;
1213         u->hop_limit = p->hop_limit;
1214         u->encap_limit = p->encap_limit;
1215         u->flowinfo = p->flowinfo;
1216         u->link = p->link;
1217         u->i_key = p->i_key;
1218         u->o_key = p->o_key;
1219         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1220         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1221         memcpy(u->name, p->name, sizeof(u->name));
1222 }
1223
1224 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1225         struct ifreq *ifr, int cmd)
1226 {
1227         int err = 0;
1228         struct ip6_tnl_parm2 p;
1229         struct __ip6_tnl_parm p1;
1230         struct ip6_tnl *t = netdev_priv(dev);
1231         struct net *net = t->net;
1232         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1233
1234         memset(&p1, 0, sizeof(p1));
1235
1236         switch (cmd) {
1237         case SIOCGETTUNNEL:
1238                 if (dev == ign->fb_tunnel_dev) {
1239                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1240                                 err = -EFAULT;
1241                                 break;
1242                         }
1243                         ip6gre_tnl_parm_from_user(&p1, &p);
1244                         t = ip6gre_tunnel_locate(net, &p1, 0);
1245                         if (!t)
1246                                 t = netdev_priv(dev);
1247                 }
1248                 memset(&p, 0, sizeof(p));
1249                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1250                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1251                         err = -EFAULT;
1252                 break;
1253
1254         case SIOCADDTUNNEL:
1255         case SIOCCHGTUNNEL:
1256                 err = -EPERM;
1257                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1258                         goto done;
1259
1260                 err = -EFAULT;
1261                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1262                         goto done;
1263
1264                 err = -EINVAL;
1265                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1266                         goto done;
1267
1268                 if (!(p.i_flags&GRE_KEY))
1269                         p.i_key = 0;
1270                 if (!(p.o_flags&GRE_KEY))
1271                         p.o_key = 0;
1272
1273                 ip6gre_tnl_parm_from_user(&p1, &p);
1274                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1275
1276                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1277                         if (t) {
1278                                 if (t->dev != dev) {
1279                                         err = -EEXIST;
1280                                         break;
1281                                 }
1282                         } else {
1283                                 t = netdev_priv(dev);
1284
1285                                 ip6gre_tunnel_unlink(ign, t);
1286                                 synchronize_net();
1287                                 ip6gre_tnl_change(t, &p1, 1);
1288                                 ip6gre_tunnel_link(ign, t);
1289                                 netdev_state_change(dev);
1290                         }
1291                 }
1292
1293                 if (t) {
1294                         err = 0;
1295
1296                         memset(&p, 0, sizeof(p));
1297                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1298                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1299                                 err = -EFAULT;
1300                 } else
1301                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1302                 break;
1303
1304         case SIOCDELTUNNEL:
1305                 err = -EPERM;
1306                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1307                         goto done;
1308
1309                 if (dev == ign->fb_tunnel_dev) {
1310                         err = -EFAULT;
1311                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1312                                 goto done;
1313                         err = -ENOENT;
1314                         ip6gre_tnl_parm_from_user(&p1, &p);
1315                         t = ip6gre_tunnel_locate(net, &p1, 0);
1316                         if (!t)
1317                                 goto done;
1318                         err = -EPERM;
1319                         if (t == netdev_priv(ign->fb_tunnel_dev))
1320                                 goto done;
1321                         dev = t->dev;
1322                 }
1323                 unregister_netdevice(dev);
1324                 err = 0;
1325                 break;
1326
1327         default:
1328                 err = -EINVAL;
1329         }
1330
1331 done:
1332         return err;
1333 }
1334
1335 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1336                          unsigned short type, const void *daddr,
1337                          const void *saddr, unsigned int len)
1338 {
1339         struct ip6_tnl *t = netdev_priv(dev);
1340         struct ipv6hdr *ipv6h;
1341         __be16 *p;
1342
1343         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1344         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1345                                                   t->fl.u.ip6.flowlabel,
1346                                                   true, &t->fl.u.ip6));
1347         ipv6h->hop_limit = t->parms.hop_limit;
1348         ipv6h->nexthdr = NEXTHDR_GRE;
1349         ipv6h->saddr = t->parms.laddr;
1350         ipv6h->daddr = t->parms.raddr;
1351
1352         p = (__be16 *)(ipv6h + 1);
1353         p[0] = t->parms.o_flags;
1354         p[1] = htons(type);
1355
1356         /*
1357          *      Set the source hardware address.
1358          */
1359
1360         if (saddr)
1361                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1362         if (daddr)
1363                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1364         if (!ipv6_addr_any(&ipv6h->daddr))
1365                 return t->hlen;
1366
1367         return -t->hlen;
1368 }
1369
1370 static const struct header_ops ip6gre_header_ops = {
1371         .create = ip6gre_header,
1372 };
1373
1374 static const struct net_device_ops ip6gre_netdev_ops = {
1375         .ndo_init               = ip6gre_tunnel_init,
1376         .ndo_uninit             = ip6gre_tunnel_uninit,
1377         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1378         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1379         .ndo_change_mtu         = ip6_tnl_change_mtu,
1380         .ndo_get_stats64        = ip_tunnel_get_stats64,
1381         .ndo_get_iflink         = ip6_tnl_get_iflink,
1382 };
1383
1384 static void ip6gre_dev_free(struct net_device *dev)
1385 {
1386         struct ip6_tnl *t = netdev_priv(dev);
1387
1388         gro_cells_destroy(&t->gro_cells);
1389         dst_cache_destroy(&t->dst_cache);
1390         free_percpu(dev->tstats);
1391 }
1392
1393 static void ip6gre_tunnel_setup(struct net_device *dev)
1394 {
1395         dev->netdev_ops = &ip6gre_netdev_ops;
1396         dev->needs_free_netdev = true;
1397         dev->priv_destructor = ip6gre_dev_free;
1398
1399         dev->type = ARPHRD_IP6GRE;
1400
1401         dev->flags |= IFF_NOARP;
1402         dev->addr_len = sizeof(struct in6_addr);
1403         netif_keep_dst(dev);
1404         /* This perm addr will be used as interface identifier by IPv6 */
1405         dev->addr_assign_type = NET_ADDR_RANDOM;
1406         eth_random_addr(dev->perm_addr);
1407 }
1408
1409 #define GRE6_FEATURES (NETIF_F_SG |             \
1410                        NETIF_F_FRAGLIST |       \
1411                        NETIF_F_HIGHDMA |        \
1412                        NETIF_F_HW_CSUM)
1413
1414 static void ip6gre_tnl_init_features(struct net_device *dev)
1415 {
1416         struct ip6_tnl *nt = netdev_priv(dev);
1417
1418         dev->features           |= GRE6_FEATURES;
1419         dev->hw_features        |= GRE6_FEATURES;
1420
1421         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1422                 /* TCP offload with GRE SEQ is not supported, nor
1423                  * can we support 2 levels of outer headers requiring
1424                  * an update.
1425                  */
1426                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1427                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1428                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1429                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1430                 }
1431
1432                 /* Can use a lockless transmit, unless we generate
1433                  * output sequences
1434                  */
1435                 dev->features |= NETIF_F_LLTX;
1436         }
1437 }
1438
1439 static int ip6gre_tunnel_init_common(struct net_device *dev)
1440 {
1441         struct ip6_tnl *tunnel;
1442         int ret;
1443         int t_hlen;
1444
1445         tunnel = netdev_priv(dev);
1446
1447         tunnel->dev = dev;
1448         tunnel->net = dev_net(dev);
1449         strcpy(tunnel->parms.name, dev->name);
1450
1451         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1452         if (!dev->tstats)
1453                 return -ENOMEM;
1454
1455         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1456         if (ret)
1457                 goto cleanup_alloc_pcpu_stats;
1458
1459         ret = gro_cells_init(&tunnel->gro_cells, dev);
1460         if (ret)
1461                 goto cleanup_dst_cache_init;
1462
1463         t_hlen = ip6gre_calc_hlen(tunnel);
1464         dev->mtu = ETH_DATA_LEN - t_hlen;
1465         if (dev->type == ARPHRD_ETHER)
1466                 dev->mtu -= ETH_HLEN;
1467         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1468                 dev->mtu -= 8;
1469
1470         if (tunnel->parms.collect_md) {
1471                 netif_keep_dst(dev);
1472         }
1473         ip6gre_tnl_init_features(dev);
1474
1475         return 0;
1476
1477 cleanup_dst_cache_init:
1478         dst_cache_destroy(&tunnel->dst_cache);
1479 cleanup_alloc_pcpu_stats:
1480         free_percpu(dev->tstats);
1481         dev->tstats = NULL;
1482         return ret;
1483 }
1484
1485 static int ip6gre_tunnel_init(struct net_device *dev)
1486 {
1487         struct ip6_tnl *tunnel;
1488         int ret;
1489
1490         ret = ip6gre_tunnel_init_common(dev);
1491         if (ret)
1492                 return ret;
1493
1494         tunnel = netdev_priv(dev);
1495
1496         if (tunnel->parms.collect_md)
1497                 return 0;
1498
1499         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1500         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1501
1502         if (ipv6_addr_any(&tunnel->parms.raddr))
1503                 dev->header_ops = &ip6gre_header_ops;
1504
1505         return 0;
1506 }
1507
1508 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1509 {
1510         struct ip6_tnl *tunnel = netdev_priv(dev);
1511
1512         tunnel->dev = dev;
1513         tunnel->net = dev_net(dev);
1514         strcpy(tunnel->parms.name, dev->name);
1515
1516         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1517
1518         dev_hold(dev);
1519 }
1520
1521 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1522         .handler     = gre_rcv,
1523         .err_handler = ip6gre_err,
1524         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1525 };
1526
1527 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1528 {
1529         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1530         struct net_device *dev, *aux;
1531         int prio;
1532
1533         for_each_netdev_safe(net, dev, aux)
1534                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1535                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1536                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1537                         unregister_netdevice_queue(dev, head);
1538
1539         for (prio = 0; prio < 4; prio++) {
1540                 int h;
1541                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1542                         struct ip6_tnl *t;
1543
1544                         t = rtnl_dereference(ign->tunnels[prio][h]);
1545
1546                         while (t) {
1547                                 /* If dev is in the same netns, it has already
1548                                  * been added to the list by the previous loop.
1549                                  */
1550                                 if (!net_eq(dev_net(t->dev), net))
1551                                         unregister_netdevice_queue(t->dev,
1552                                                                    head);
1553                                 t = rtnl_dereference(t->next);
1554                         }
1555                 }
1556         }
1557 }
1558
1559 static int __net_init ip6gre_init_net(struct net *net)
1560 {
1561         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1562         int err;
1563
1564         if (!net_has_fallback_tunnels(net))
1565                 return 0;
1566         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1567                                           NET_NAME_UNKNOWN,
1568                                           ip6gre_tunnel_setup);
1569         if (!ign->fb_tunnel_dev) {
1570                 err = -ENOMEM;
1571                 goto err_alloc_dev;
1572         }
1573         dev_net_set(ign->fb_tunnel_dev, net);
1574         /* FB netdevice is special: we have one, and only one per netns.
1575          * Allowing to move it to another netns is clearly unsafe.
1576          */
1577         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1578
1579
1580         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1581         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1582
1583         err = register_netdev(ign->fb_tunnel_dev);
1584         if (err)
1585                 goto err_reg_dev;
1586
1587         rcu_assign_pointer(ign->tunnels_wc[0],
1588                            netdev_priv(ign->fb_tunnel_dev));
1589         return 0;
1590
1591 err_reg_dev:
1592         free_netdev(ign->fb_tunnel_dev);
1593 err_alloc_dev:
1594         return err;
1595 }
1596
1597 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1598 {
1599         struct net *net;
1600         LIST_HEAD(list);
1601
1602         rtnl_lock();
1603         list_for_each_entry(net, net_list, exit_list)
1604                 ip6gre_destroy_tunnels(net, &list);
1605         unregister_netdevice_many(&list);
1606         rtnl_unlock();
1607 }
1608
1609 static struct pernet_operations ip6gre_net_ops = {
1610         .init = ip6gre_init_net,
1611         .exit_batch = ip6gre_exit_batch_net,
1612         .id   = &ip6gre_net_id,
1613         .size = sizeof(struct ip6gre_net),
1614 };
1615
1616 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1617                                   struct netlink_ext_ack *extack)
1618 {
1619         __be16 flags;
1620
1621         if (!data)
1622                 return 0;
1623
1624         flags = 0;
1625         if (data[IFLA_GRE_IFLAGS])
1626                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1627         if (data[IFLA_GRE_OFLAGS])
1628                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1629         if (flags & (GRE_VERSION|GRE_ROUTING))
1630                 return -EINVAL;
1631
1632         return 0;
1633 }
1634
1635 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1636                                struct netlink_ext_ack *extack)
1637 {
1638         struct in6_addr daddr;
1639
1640         if (tb[IFLA_ADDRESS]) {
1641                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1642                         return -EINVAL;
1643                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1644                         return -EADDRNOTAVAIL;
1645         }
1646
1647         if (!data)
1648                 goto out;
1649
1650         if (data[IFLA_GRE_REMOTE]) {
1651                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1652                 if (ipv6_addr_any(&daddr))
1653                         return -EINVAL;
1654         }
1655
1656 out:
1657         return ip6gre_tunnel_validate(tb, data, extack);
1658 }
1659
1660 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1661                                   struct netlink_ext_ack *extack)
1662 {
1663         __be16 flags = 0;
1664         int ret, ver = 0;
1665
1666         if (!data)
1667                 return 0;
1668
1669         ret = ip6gre_tap_validate(tb, data, extack);
1670         if (ret)
1671                 return ret;
1672
1673         /* ERSPAN should only have GRE sequence and key flag */
1674         if (data[IFLA_GRE_OFLAGS])
1675                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1676         if (data[IFLA_GRE_IFLAGS])
1677                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1678         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1679             flags != (GRE_SEQ | GRE_KEY))
1680                 return -EINVAL;
1681
1682         /* ERSPAN Session ID only has 10-bit. Since we reuse
1683          * 32-bit key field as ID, check it's range.
1684          */
1685         if (data[IFLA_GRE_IKEY] &&
1686             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1687                 return -EINVAL;
1688
1689         if (data[IFLA_GRE_OKEY] &&
1690             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1691                 return -EINVAL;
1692
1693         if (data[IFLA_GRE_ERSPAN_VER]) {
1694                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1695                 if (ver != 1 && ver != 2)
1696                         return -EINVAL;
1697         }
1698
1699         if (ver == 1) {
1700                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1701                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1702
1703                         if (index & ~INDEX_MASK)
1704                                 return -EINVAL;
1705                 }
1706         } else if (ver == 2) {
1707                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1708                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1709
1710                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1711                                 return -EINVAL;
1712                 }
1713
1714                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1715                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1716
1717                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1718                                 return -EINVAL;
1719                 }
1720         }
1721
1722         return 0;
1723 }
1724
1725 static void ip6erspan_set_version(struct nlattr *data[],
1726                                   struct __ip6_tnl_parm *parms)
1727 {
1728         if (!data)
1729                 return;
1730
1731         parms->erspan_ver = 1;
1732         if (data[IFLA_GRE_ERSPAN_VER])
1733                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1734
1735         if (parms->erspan_ver == 1) {
1736                 if (data[IFLA_GRE_ERSPAN_INDEX])
1737                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1738         } else if (parms->erspan_ver == 2) {
1739                 if (data[IFLA_GRE_ERSPAN_DIR])
1740                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1741                 if (data[IFLA_GRE_ERSPAN_HWID])
1742                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1743         }
1744 }
1745
1746 static void ip6gre_netlink_parms(struct nlattr *data[],
1747                                 struct __ip6_tnl_parm *parms)
1748 {
1749         memset(parms, 0, sizeof(*parms));
1750
1751         if (!data)
1752                 return;
1753
1754         if (data[IFLA_GRE_LINK])
1755                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1756
1757         if (data[IFLA_GRE_IFLAGS])
1758                 parms->i_flags = gre_flags_to_tnl_flags(
1759                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1760
1761         if (data[IFLA_GRE_OFLAGS])
1762                 parms->o_flags = gre_flags_to_tnl_flags(
1763                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1764
1765         if (data[IFLA_GRE_IKEY])
1766                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1767
1768         if (data[IFLA_GRE_OKEY])
1769                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1770
1771         if (data[IFLA_GRE_LOCAL])
1772                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1773
1774         if (data[IFLA_GRE_REMOTE])
1775                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1776
1777         if (data[IFLA_GRE_TTL])
1778                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1779
1780         if (data[IFLA_GRE_ENCAP_LIMIT])
1781                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1782
1783         if (data[IFLA_GRE_FLOWINFO])
1784                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1785
1786         if (data[IFLA_GRE_FLAGS])
1787                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1788
1789         if (data[IFLA_GRE_FWMARK])
1790                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1791
1792         if (data[IFLA_GRE_COLLECT_METADATA])
1793                 parms->collect_md = true;
1794 }
1795
1796 static int ip6gre_tap_init(struct net_device *dev)
1797 {
1798         int ret;
1799
1800         ret = ip6gre_tunnel_init_common(dev);
1801         if (ret)
1802                 return ret;
1803
1804         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1805
1806         return 0;
1807 }
1808
1809 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1810         .ndo_init = ip6gre_tap_init,
1811         .ndo_uninit = ip6gre_tunnel_uninit,
1812         .ndo_start_xmit = ip6gre_tunnel_xmit,
1813         .ndo_set_mac_address = eth_mac_addr,
1814         .ndo_validate_addr = eth_validate_addr,
1815         .ndo_change_mtu = ip6_tnl_change_mtu,
1816         .ndo_get_stats64 = ip_tunnel_get_stats64,
1817         .ndo_get_iflink = ip6_tnl_get_iflink,
1818 };
1819
1820 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1821 {
1822         int t_hlen;
1823
1824         tunnel->tun_hlen = 8;
1825         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1826                        erspan_hdr_len(tunnel->parms.erspan_ver);
1827
1828         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1829         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1830         return t_hlen;
1831 }
1832
1833 static int ip6erspan_tap_init(struct net_device *dev)
1834 {
1835         struct ip6_tnl *tunnel;
1836         int t_hlen;
1837         int ret;
1838
1839         tunnel = netdev_priv(dev);
1840
1841         tunnel->dev = dev;
1842         tunnel->net = dev_net(dev);
1843         strcpy(tunnel->parms.name, dev->name);
1844
1845         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1846         if (!dev->tstats)
1847                 return -ENOMEM;
1848
1849         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1850         if (ret)
1851                 goto cleanup_alloc_pcpu_stats;
1852
1853         ret = gro_cells_init(&tunnel->gro_cells, dev);
1854         if (ret)
1855                 goto cleanup_dst_cache_init;
1856
1857         t_hlen = ip6erspan_calc_hlen(tunnel);
1858         dev->mtu = ETH_DATA_LEN - t_hlen;
1859         if (dev->type == ARPHRD_ETHER)
1860                 dev->mtu -= ETH_HLEN;
1861         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1862                 dev->mtu -= 8;
1863
1864         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1865         ip6erspan_tnl_link_config(tunnel, 1);
1866
1867         return 0;
1868
1869 cleanup_dst_cache_init:
1870         dst_cache_destroy(&tunnel->dst_cache);
1871 cleanup_alloc_pcpu_stats:
1872         free_percpu(dev->tstats);
1873         dev->tstats = NULL;
1874         return ret;
1875 }
1876
1877 static const struct net_device_ops ip6erspan_netdev_ops = {
1878         .ndo_init =             ip6erspan_tap_init,
1879         .ndo_uninit =           ip6erspan_tunnel_uninit,
1880         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1881         .ndo_set_mac_address =  eth_mac_addr,
1882         .ndo_validate_addr =    eth_validate_addr,
1883         .ndo_change_mtu =       ip6_tnl_change_mtu,
1884         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1885         .ndo_get_iflink =       ip6_tnl_get_iflink,
1886 };
1887
1888 static void ip6gre_tap_setup(struct net_device *dev)
1889 {
1890
1891         ether_setup(dev);
1892
1893         dev->max_mtu = 0;
1894         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1895         dev->needs_free_netdev = true;
1896         dev->priv_destructor = ip6gre_dev_free;
1897
1898         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1899         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1900         netif_keep_dst(dev);
1901 }
1902
1903 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1904                                        struct ip_tunnel_encap *ipencap)
1905 {
1906         bool ret = false;
1907
1908         memset(ipencap, 0, sizeof(*ipencap));
1909
1910         if (!data)
1911                 return ret;
1912
1913         if (data[IFLA_GRE_ENCAP_TYPE]) {
1914                 ret = true;
1915                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1916         }
1917
1918         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1919                 ret = true;
1920                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1921         }
1922
1923         if (data[IFLA_GRE_ENCAP_SPORT]) {
1924                 ret = true;
1925                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1926         }
1927
1928         if (data[IFLA_GRE_ENCAP_DPORT]) {
1929                 ret = true;
1930                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1931         }
1932
1933         return ret;
1934 }
1935
1936 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1937                                  struct nlattr *tb[], struct nlattr *data[],
1938                                  struct netlink_ext_ack *extack)
1939 {
1940         struct ip6_tnl *nt;
1941         struct ip_tunnel_encap ipencap;
1942         int err;
1943
1944         nt = netdev_priv(dev);
1945
1946         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1947                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1948
1949                 if (err < 0)
1950                         return err;
1951         }
1952
1953         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1954                 eth_hw_addr_random(dev);
1955
1956         nt->dev = dev;
1957         nt->net = dev_net(dev);
1958
1959         err = register_netdevice(dev);
1960         if (err)
1961                 goto out;
1962
1963         if (tb[IFLA_MTU])
1964                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1965
1966         dev_hold(dev);
1967
1968 out:
1969         return err;
1970 }
1971
1972 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1973                           struct nlattr *tb[], struct nlattr *data[],
1974                           struct netlink_ext_ack *extack)
1975 {
1976         struct ip6_tnl *nt = netdev_priv(dev);
1977         struct net *net = dev_net(dev);
1978         struct ip6gre_net *ign;
1979         int err;
1980
1981         ip6gre_netlink_parms(data, &nt->parms);
1982         ign = net_generic(net, ip6gre_net_id);
1983
1984         if (nt->parms.collect_md) {
1985                 if (rtnl_dereference(ign->collect_md_tun))
1986                         return -EEXIST;
1987         } else {
1988                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1989                         return -EEXIST;
1990         }
1991
1992         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1993         if (!err) {
1994                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1995                 ip6gre_tunnel_link_md(ign, nt);
1996                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1997         }
1998         return err;
1999 }
2000
2001 static struct ip6_tnl *
2002 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2003                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2004                          struct netlink_ext_ack *extack)
2005 {
2006         struct ip6_tnl *t, *nt = netdev_priv(dev);
2007         struct net *net = nt->net;
2008         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2009         struct ip_tunnel_encap ipencap;
2010
2011         if (dev == ign->fb_tunnel_dev)
2012                 return ERR_PTR(-EINVAL);
2013
2014         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2015                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2016
2017                 if (err < 0)
2018                         return ERR_PTR(err);
2019         }
2020
2021         ip6gre_netlink_parms(data, p_p);
2022
2023         t = ip6gre_tunnel_locate(net, p_p, 0);
2024
2025         if (t) {
2026                 if (t->dev != dev)
2027                         return ERR_PTR(-EEXIST);
2028         } else {
2029                 t = nt;
2030         }
2031
2032         return t;
2033 }
2034
2035 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2036                              struct nlattr *data[],
2037                              struct netlink_ext_ack *extack)
2038 {
2039         struct ip6_tnl *t = netdev_priv(dev);
2040         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2041         struct __ip6_tnl_parm p;
2042
2043         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2044         if (IS_ERR(t))
2045                 return PTR_ERR(t);
2046
2047         ip6gre_tunnel_unlink_md(ign, t);
2048         ip6gre_tunnel_unlink(ign, t);
2049         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2050         ip6gre_tunnel_link_md(ign, t);
2051         ip6gre_tunnel_link(ign, t);
2052         return 0;
2053 }
2054
2055 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2056 {
2057         struct net *net = dev_net(dev);
2058         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2059
2060         if (dev != ign->fb_tunnel_dev)
2061                 unregister_netdevice_queue(dev, head);
2062 }
2063
2064 static size_t ip6gre_get_size(const struct net_device *dev)
2065 {
2066         return
2067                 /* IFLA_GRE_LINK */
2068                 nla_total_size(4) +
2069                 /* IFLA_GRE_IFLAGS */
2070                 nla_total_size(2) +
2071                 /* IFLA_GRE_OFLAGS */
2072                 nla_total_size(2) +
2073                 /* IFLA_GRE_IKEY */
2074                 nla_total_size(4) +
2075                 /* IFLA_GRE_OKEY */
2076                 nla_total_size(4) +
2077                 /* IFLA_GRE_LOCAL */
2078                 nla_total_size(sizeof(struct in6_addr)) +
2079                 /* IFLA_GRE_REMOTE */
2080                 nla_total_size(sizeof(struct in6_addr)) +
2081                 /* IFLA_GRE_TTL */
2082                 nla_total_size(1) +
2083                 /* IFLA_GRE_ENCAP_LIMIT */
2084                 nla_total_size(1) +
2085                 /* IFLA_GRE_FLOWINFO */
2086                 nla_total_size(4) +
2087                 /* IFLA_GRE_FLAGS */
2088                 nla_total_size(4) +
2089                 /* IFLA_GRE_ENCAP_TYPE */
2090                 nla_total_size(2) +
2091                 /* IFLA_GRE_ENCAP_FLAGS */
2092                 nla_total_size(2) +
2093                 /* IFLA_GRE_ENCAP_SPORT */
2094                 nla_total_size(2) +
2095                 /* IFLA_GRE_ENCAP_DPORT */
2096                 nla_total_size(2) +
2097                 /* IFLA_GRE_COLLECT_METADATA */
2098                 nla_total_size(0) +
2099                 /* IFLA_GRE_FWMARK */
2100                 nla_total_size(4) +
2101                 /* IFLA_GRE_ERSPAN_INDEX */
2102                 nla_total_size(4) +
2103                 0;
2104 }
2105
2106 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2107 {
2108         struct ip6_tnl *t = netdev_priv(dev);
2109         struct __ip6_tnl_parm *p = &t->parms;
2110         __be16 o_flags = p->o_flags;
2111
2112         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2113                 if (!p->collect_md)
2114                         o_flags |= TUNNEL_KEY;
2115
2116                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2117                         goto nla_put_failure;
2118
2119                 if (p->erspan_ver == 1) {
2120                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2121                                 goto nla_put_failure;
2122                 } else {
2123                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2124                                 goto nla_put_failure;
2125                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2126                                 goto nla_put_failure;
2127                 }
2128         }
2129
2130         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2131             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2132                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2133             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2134                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2135             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2136             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2137             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2138             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2139             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2140             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2141             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2142             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2143             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2144                 goto nla_put_failure;
2145
2146         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2147                         t->encap.type) ||
2148             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2149                          t->encap.sport) ||
2150             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2151                          t->encap.dport) ||
2152             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2153                         t->encap.flags))
2154                 goto nla_put_failure;
2155
2156         if (p->collect_md) {
2157                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2158                         goto nla_put_failure;
2159         }
2160
2161         return 0;
2162
2163 nla_put_failure:
2164         return -EMSGSIZE;
2165 }
2166
2167 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2168         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2169         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2170         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2171         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2172         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2173         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2174         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2175         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2176         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2177         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2178         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2179         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2180         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2181         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2182         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2183         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2184         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2185         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2186         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2187         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2188         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2189 };
2190
2191 static void ip6erspan_tap_setup(struct net_device *dev)
2192 {
2193         ether_setup(dev);
2194
2195         dev->max_mtu = 0;
2196         dev->netdev_ops = &ip6erspan_netdev_ops;
2197         dev->needs_free_netdev = true;
2198         dev->priv_destructor = ip6gre_dev_free;
2199
2200         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2201         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2202         netif_keep_dst(dev);
2203 }
2204
2205 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2206                              struct nlattr *tb[], struct nlattr *data[],
2207                              struct netlink_ext_ack *extack)
2208 {
2209         struct ip6_tnl *nt = netdev_priv(dev);
2210         struct net *net = dev_net(dev);
2211         struct ip6gre_net *ign;
2212         int err;
2213
2214         ip6gre_netlink_parms(data, &nt->parms);
2215         ip6erspan_set_version(data, &nt->parms);
2216         ign = net_generic(net, ip6gre_net_id);
2217
2218         if (nt->parms.collect_md) {
2219                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2220                         return -EEXIST;
2221         } else {
2222                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2223                         return -EEXIST;
2224         }
2225
2226         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2227         if (!err) {
2228                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2229                 ip6erspan_tunnel_link_md(ign, nt);
2230                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2231         }
2232         return err;
2233 }
2234
2235 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2236 {
2237         ip6gre_tnl_link_config_common(t);
2238         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2239 }
2240
2241 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2242                                 const struct __ip6_tnl_parm *p, int set_mtu)
2243 {
2244         ip6gre_tnl_copy_tnl_parm(t, p);
2245         ip6erspan_tnl_link_config(t, set_mtu);
2246         return 0;
2247 }
2248
2249 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2250                                 struct nlattr *data[],
2251                                 struct netlink_ext_ack *extack)
2252 {
2253         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2254         struct __ip6_tnl_parm p;
2255         struct ip6_tnl *t;
2256
2257         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2258         if (IS_ERR(t))
2259                 return PTR_ERR(t);
2260
2261         ip6erspan_set_version(data, &p);
2262         ip6gre_tunnel_unlink_md(ign, t);
2263         ip6gre_tunnel_unlink(ign, t);
2264         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2265         ip6erspan_tunnel_link_md(ign, t);
2266         ip6gre_tunnel_link(ign, t);
2267         return 0;
2268 }
2269
2270 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2271         .kind           = "ip6gre",
2272         .maxtype        = IFLA_GRE_MAX,
2273         .policy         = ip6gre_policy,
2274         .priv_size      = sizeof(struct ip6_tnl),
2275         .setup          = ip6gre_tunnel_setup,
2276         .validate       = ip6gre_tunnel_validate,
2277         .newlink        = ip6gre_newlink,
2278         .changelink     = ip6gre_changelink,
2279         .dellink        = ip6gre_dellink,
2280         .get_size       = ip6gre_get_size,
2281         .fill_info      = ip6gre_fill_info,
2282         .get_link_net   = ip6_tnl_get_link_net,
2283 };
2284
2285 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2286         .kind           = "ip6gretap",
2287         .maxtype        = IFLA_GRE_MAX,
2288         .policy         = ip6gre_policy,
2289         .priv_size      = sizeof(struct ip6_tnl),
2290         .setup          = ip6gre_tap_setup,
2291         .validate       = ip6gre_tap_validate,
2292         .newlink        = ip6gre_newlink,
2293         .changelink     = ip6gre_changelink,
2294         .get_size       = ip6gre_get_size,
2295         .fill_info      = ip6gre_fill_info,
2296         .get_link_net   = ip6_tnl_get_link_net,
2297 };
2298
2299 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2300         .kind           = "ip6erspan",
2301         .maxtype        = IFLA_GRE_MAX,
2302         .policy         = ip6gre_policy,
2303         .priv_size      = sizeof(struct ip6_tnl),
2304         .setup          = ip6erspan_tap_setup,
2305         .validate       = ip6erspan_tap_validate,
2306         .newlink        = ip6erspan_newlink,
2307         .changelink     = ip6erspan_changelink,
2308         .get_size       = ip6gre_get_size,
2309         .fill_info      = ip6gre_fill_info,
2310         .get_link_net   = ip6_tnl_get_link_net,
2311 };
2312
2313 /*
2314  *      And now the modules code and kernel interface.
2315  */
2316
2317 static int __init ip6gre_init(void)
2318 {
2319         int err;
2320
2321         pr_info("GRE over IPv6 tunneling driver\n");
2322
2323         err = register_pernet_device(&ip6gre_net_ops);
2324         if (err < 0)
2325                 return err;
2326
2327         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2328         if (err < 0) {
2329                 pr_info("%s: can't add protocol\n", __func__);
2330                 goto add_proto_failed;
2331         }
2332
2333         err = rtnl_link_register(&ip6gre_link_ops);
2334         if (err < 0)
2335                 goto rtnl_link_failed;
2336
2337         err = rtnl_link_register(&ip6gre_tap_ops);
2338         if (err < 0)
2339                 goto tap_ops_failed;
2340
2341         err = rtnl_link_register(&ip6erspan_tap_ops);
2342         if (err < 0)
2343                 goto erspan_link_failed;
2344
2345 out:
2346         return err;
2347
2348 erspan_link_failed:
2349         rtnl_link_unregister(&ip6gre_tap_ops);
2350 tap_ops_failed:
2351         rtnl_link_unregister(&ip6gre_link_ops);
2352 rtnl_link_failed:
2353         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2354 add_proto_failed:
2355         unregister_pernet_device(&ip6gre_net_ops);
2356         goto out;
2357 }
2358
2359 static void __exit ip6gre_fini(void)
2360 {
2361         rtnl_link_unregister(&ip6gre_tap_ops);
2362         rtnl_link_unregister(&ip6gre_link_ops);
2363         rtnl_link_unregister(&ip6erspan_tap_ops);
2364         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2365         unregister_pernet_device(&ip6gre_net_ops);
2366 }
2367
2368 module_init(ip6gre_init);
2369 module_exit(ip6gre_fini);
2370 MODULE_LICENSE("GPL");
2371 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2372 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2373 MODULE_ALIAS_RTNL_LINK("ip6gre");
2374 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2375 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2376 MODULE_ALIAS_NETDEV("ip6gre0");