s390/mm: fix typo in comment
[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         struct net_device *ndev;
131
132         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
133                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
134                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
135                     key != t->parms.i_key ||
136                     !(t->dev->flags & IFF_UP))
137                         continue;
138
139                 if (t->dev->type != ARPHRD_IP6GRE &&
140                     t->dev->type != dev_type)
141                         continue;
142
143                 score = 0;
144                 if (t->parms.link != link)
145                         score |= 1;
146                 if (t->dev->type != dev_type)
147                         score |= 2;
148                 if (score == 0)
149                         return t;
150
151                 if (score < cand_score) {
152                         cand = t;
153                         cand_score = score;
154                 }
155         }
156
157         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
158                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
159                     key != t->parms.i_key ||
160                     !(t->dev->flags & IFF_UP))
161                         continue;
162
163                 if (t->dev->type != ARPHRD_IP6GRE &&
164                     t->dev->type != dev_type)
165                         continue;
166
167                 score = 0;
168                 if (t->parms.link != link)
169                         score |= 1;
170                 if (t->dev->type != dev_type)
171                         score |= 2;
172                 if (score == 0)
173                         return t;
174
175                 if (score < cand_score) {
176                         cand = t;
177                         cand_score = score;
178                 }
179         }
180
181         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
182                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
183                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
184                                  !ipv6_addr_is_multicast(local))) ||
185                     key != t->parms.i_key ||
186                     !(t->dev->flags & IFF_UP))
187                         continue;
188
189                 if (t->dev->type != ARPHRD_IP6GRE &&
190                     t->dev->type != dev_type)
191                         continue;
192
193                 score = 0;
194                 if (t->parms.link != link)
195                         score |= 1;
196                 if (t->dev->type != dev_type)
197                         score |= 2;
198                 if (score == 0)
199                         return t;
200
201                 if (score < cand_score) {
202                         cand = t;
203                         cand_score = score;
204                 }
205         }
206
207         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
208                 if (t->parms.i_key != key ||
209                     !(t->dev->flags & IFF_UP))
210                         continue;
211
212                 if (t->dev->type != ARPHRD_IP6GRE &&
213                     t->dev->type != dev_type)
214                         continue;
215
216                 score = 0;
217                 if (t->parms.link != link)
218                         score |= 1;
219                 if (t->dev->type != dev_type)
220                         score |= 2;
221                 if (score == 0)
222                         return t;
223
224                 if (score < cand_score) {
225                         cand = t;
226                         cand_score = score;
227                 }
228         }
229
230         if (cand)
231                 return cand;
232
233         if (gre_proto == htons(ETH_P_ERSPAN) ||
234             gre_proto == htons(ETH_P_ERSPAN2))
235                 t = rcu_dereference(ign->collect_md_tun_erspan);
236         else
237                 t = rcu_dereference(ign->collect_md_tun);
238
239         if (t && t->dev->flags & IFF_UP)
240                 return t;
241
242         ndev = READ_ONCE(ign->fb_tunnel_dev);
243         if (ndev && ndev->flags & IFF_UP)
244                 return netdev_priv(ndev);
245
246         return NULL;
247 }
248
249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
250                 const struct __ip6_tnl_parm *p)
251 {
252         const struct in6_addr *remote = &p->raddr;
253         const struct in6_addr *local = &p->laddr;
254         unsigned int h = HASH_KEY(p->i_key);
255         int prio = 0;
256
257         if (!ipv6_addr_any(local))
258                 prio |= 1;
259         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
260                 prio |= 2;
261                 h ^= HASH_ADDR(remote);
262         }
263
264         return &ign->tunnels[prio][h];
265 }
266
267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269         if (t->parms.collect_md)
270                 rcu_assign_pointer(ign->collect_md_tun, t);
271 }
272
273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
274 {
275         if (t->parms.collect_md)
276                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
277 }
278
279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
280 {
281         if (t->parms.collect_md)
282                 rcu_assign_pointer(ign->collect_md_tun, NULL);
283 }
284
285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
286                                        struct ip6_tnl *t)
287 {
288         if (t->parms.collect_md)
289                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
290 }
291
292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
293                 const struct ip6_tnl *t)
294 {
295         return __ip6gre_bucket(ign, &t->parms);
296 }
297
298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
299 {
300         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
301
302         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
303         rcu_assign_pointer(*tp, t);
304 }
305
306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
307 {
308         struct ip6_tnl __rcu **tp;
309         struct ip6_tnl *iter;
310
311         for (tp = ip6gre_bucket(ign, t);
312              (iter = rtnl_dereference(*tp)) != NULL;
313              tp = &iter->next) {
314                 if (t == iter) {
315                         rcu_assign_pointer(*tp, t->next);
316                         break;
317                 }
318         }
319 }
320
321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
322                                            const struct __ip6_tnl_parm *parms,
323                                            int type)
324 {
325         const struct in6_addr *remote = &parms->raddr;
326         const struct in6_addr *local = &parms->laddr;
327         __be32 key = parms->i_key;
328         int link = parms->link;
329         struct ip6_tnl *t;
330         struct ip6_tnl __rcu **tp;
331         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
332
333         for (tp = __ip6gre_bucket(ign, parms);
334              (t = rtnl_dereference(*tp)) != NULL;
335              tp = &t->next)
336                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
337                     ipv6_addr_equal(remote, &t->parms.raddr) &&
338                     key == t->parms.i_key &&
339                     link == t->parms.link &&
340                     type == t->dev->type)
341                         break;
342
343         return t;
344 }
345
346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
347                 const struct __ip6_tnl_parm *parms, int create)
348 {
349         struct ip6_tnl *t, *nt;
350         struct net_device *dev;
351         char name[IFNAMSIZ];
352         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
353
354         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
355         if (t && create)
356                 return NULL;
357         if (t || !create)
358                 return t;
359
360         if (parms->name[0]) {
361                 if (!dev_valid_name(parms->name))
362                         return NULL;
363                 strlcpy(name, parms->name, IFNAMSIZ);
364         } else {
365                 strcpy(name, "ip6gre%d");
366         }
367         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
368                            ip6gre_tunnel_setup);
369         if (!dev)
370                 return NULL;
371
372         dev_net_set(dev, net);
373
374         nt = netdev_priv(dev);
375         nt->parms = *parms;
376         dev->rtnl_link_ops = &ip6gre_link_ops;
377
378         nt->dev = dev;
379         nt->net = dev_net(dev);
380
381         if (register_netdevice(dev) < 0)
382                 goto failed_free;
383
384         ip6gre_tnl_link_config(nt, 1);
385
386         /* Can use a lockless transmit, unless we generate output sequences */
387         if (!(nt->parms.o_flags & TUNNEL_SEQ))
388                 dev->features |= NETIF_F_LLTX;
389
390         dev_hold(dev);
391         ip6gre_tunnel_link(ign, nt);
392         return nt;
393
394 failed_free:
395         free_netdev(dev);
396         return NULL;
397 }
398
399 static void ip6erspan_tunnel_uninit(struct net_device *dev)
400 {
401         struct ip6_tnl *t = netdev_priv(dev);
402         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
403
404         ip6erspan_tunnel_unlink_md(ign, t);
405         ip6gre_tunnel_unlink(ign, t);
406         dst_cache_reset(&t->dst_cache);
407         dev_put(dev);
408 }
409
410 static void ip6gre_tunnel_uninit(struct net_device *dev)
411 {
412         struct ip6_tnl *t = netdev_priv(dev);
413         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
414
415         ip6gre_tunnel_unlink_md(ign, t);
416         ip6gre_tunnel_unlink(ign, t);
417         if (ign->fb_tunnel_dev == dev)
418                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
419         dst_cache_reset(&t->dst_cache);
420         dev_put(dev);
421 }
422
423
424 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
425                        u8 type, u8 code, int offset, __be32 info)
426 {
427         struct net *net = dev_net(skb->dev);
428         const struct ipv6hdr *ipv6h;
429         struct tnl_ptk_info tpi;
430         struct ip6_tnl *t;
431
432         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
433                              offset) < 0)
434                 return -EINVAL;
435
436         ipv6h = (const struct ipv6hdr *)skb->data;
437         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
438                                  tpi.key, tpi.proto);
439         if (!t)
440                 return -ENOENT;
441
442         switch (type) {
443         case ICMPV6_DEST_UNREACH:
444                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
445                                     t->parms.name);
446                 if (code != ICMPV6_PORT_UNREACH)
447                         break;
448                 return 0;
449         case ICMPV6_TIME_EXCEED:
450                 if (code == ICMPV6_EXC_HOPLIMIT) {
451                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
452                                             t->parms.name);
453                         break;
454                 }
455                 return 0;
456         case ICMPV6_PARAMPROB: {
457                 struct ipv6_tlv_tnl_enc_lim *tel;
458                 __u32 teli;
459
460                 teli = 0;
461                 if (code == ICMPV6_HDR_FIELD)
462                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
463
464                 if (teli && teli == be32_to_cpu(info) - 2) {
465                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
466                         if (tel->encap_limit == 0) {
467                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
468                                                     t->parms.name);
469                         }
470                 } else {
471                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
472                                             t->parms.name);
473                 }
474                 return 0;
475         }
476         case ICMPV6_PKT_TOOBIG:
477                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
478                 return 0;
479         case NDISC_REDIRECT:
480                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
481                              sock_net_uid(net, NULL));
482                 return 0;
483         }
484
485         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
486                 t->err_count++;
487         else
488                 t->err_count = 1;
489         t->err_time = jiffies;
490
491         return 0;
492 }
493
494 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
495 {
496         const struct ipv6hdr *ipv6h;
497         struct ip6_tnl *tunnel;
498
499         ipv6h = ipv6_hdr(skb);
500         tunnel = ip6gre_tunnel_lookup(skb->dev,
501                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
502                                       tpi->proto);
503         if (tunnel) {
504                 if (tunnel->parms.collect_md) {
505                         struct metadata_dst *tun_dst;
506                         __be64 tun_id;
507                         __be16 flags;
508
509                         flags = tpi->flags;
510                         tun_id = key32_to_tunnel_id(tpi->key);
511
512                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
513                         if (!tun_dst)
514                                 return PACKET_REJECT;
515
516                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
517                 } else {
518                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
519                 }
520
521                 return PACKET_RCVD;
522         }
523
524         return PACKET_REJECT;
525 }
526
527 static int ip6erspan_rcv(struct sk_buff *skb,
528                          struct tnl_ptk_info *tpi,
529                          int gre_hdr_len)
530 {
531         struct erspan_base_hdr *ershdr;
532         const struct ipv6hdr *ipv6h;
533         struct erspan_md2 *md2;
534         struct ip6_tnl *tunnel;
535         u8 ver;
536
537         ipv6h = ipv6_hdr(skb);
538         ershdr = (struct erspan_base_hdr *)skb->data;
539         ver = ershdr->ver;
540
541         tunnel = ip6gre_tunnel_lookup(skb->dev,
542                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
543                                       tpi->proto);
544         if (tunnel) {
545                 int len = erspan_hdr_len(ver);
546
547                 if (unlikely(!pskb_may_pull(skb, len)))
548                         return PACKET_REJECT;
549
550                 if (__iptunnel_pull_header(skb, len,
551                                            htons(ETH_P_TEB),
552                                            false, false) < 0)
553                         return PACKET_REJECT;
554
555                 if (tunnel->parms.collect_md) {
556                         struct erspan_metadata *pkt_md, *md;
557                         struct metadata_dst *tun_dst;
558                         struct ip_tunnel_info *info;
559                         unsigned char *gh;
560                         __be64 tun_id;
561                         __be16 flags;
562
563                         tpi->flags |= TUNNEL_KEY;
564                         flags = tpi->flags;
565                         tun_id = key32_to_tunnel_id(tpi->key);
566
567                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
568                                                   sizeof(*md));
569                         if (!tun_dst)
570                                 return PACKET_REJECT;
571
572                         /* skb can be uncloned in __iptunnel_pull_header, so
573                          * old pkt_md is no longer valid and we need to reset
574                          * it
575                          */
576                         gh = skb_network_header(skb) +
577                              skb_network_header_len(skb);
578                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
579                                                             sizeof(*ershdr));
580                         info = &tun_dst->u.tun_info;
581                         md = ip_tunnel_info_opts(info);
582                         md->version = ver;
583                         md2 = &md->u.md2;
584                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
585                                                        ERSPAN_V2_MDSIZE);
586                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
587                         info->options_len = sizeof(*md);
588
589                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
590
591                 } else {
592                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
593                 }
594
595                 return PACKET_RCVD;
596         }
597
598         return PACKET_REJECT;
599 }
600
601 static int gre_rcv(struct sk_buff *skb)
602 {
603         struct tnl_ptk_info tpi;
604         bool csum_err = false;
605         int hdr_len;
606
607         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
608         if (hdr_len < 0)
609                 goto drop;
610
611         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
612                 goto drop;
613
614         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
615                      tpi.proto == htons(ETH_P_ERSPAN2))) {
616                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
617                         return 0;
618                 goto out;
619         }
620
621         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
622                 return 0;
623
624 out:
625         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
626 drop:
627         kfree_skb(skb);
628         return 0;
629 }
630
631 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
632 {
633         return iptunnel_handle_offloads(skb,
634                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
635 }
636
637 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
638                                      struct net_device *dev,
639                                      struct flowi6 *fl6, __u8 *dsfield,
640                                      int *encap_limit)
641 {
642         const struct iphdr *iph = ip_hdr(skb);
643         struct ip6_tnl *t = netdev_priv(dev);
644
645         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
646                 *encap_limit = t->parms.encap_limit;
647
648         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
649
650         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
651                 *dsfield = ipv4_get_dsfield(iph);
652         else
653                 *dsfield = ip6_tclass(t->parms.flowinfo);
654
655         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
656                 fl6->flowi6_mark = skb->mark;
657         else
658                 fl6->flowi6_mark = t->parms.fwmark;
659
660         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
661 }
662
663 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
664                                     struct net_device *dev,
665                                     struct flowi6 *fl6, __u8 *dsfield,
666                                     int *encap_limit)
667 {
668         struct ipv6hdr *ipv6h;
669         struct ip6_tnl *t = netdev_priv(dev);
670         __u16 offset;
671
672         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
673         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
674         ipv6h = ipv6_hdr(skb);
675
676         if (offset > 0) {
677                 struct ipv6_tlv_tnl_enc_lim *tel;
678
679                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
680                 if (tel->encap_limit == 0) {
681                         icmpv6_send(skb, ICMPV6_PARAMPROB,
682                                     ICMPV6_HDR_FIELD, offset + 2);
683                         return -1;
684                 }
685                 *encap_limit = tel->encap_limit - 1;
686         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
687                 *encap_limit = t->parms.encap_limit;
688         }
689
690         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
691
692         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
693                 *dsfield = ipv6_get_dsfield(ipv6h);
694         else
695                 *dsfield = ip6_tclass(t->parms.flowinfo);
696
697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
698                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
699
700         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
701                 fl6->flowi6_mark = skb->mark;
702         else
703                 fl6->flowi6_mark = t->parms.fwmark;
704
705         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
706
707         return 0;
708 }
709
710 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
711                                struct net_device *dev, __u8 dsfield,
712                                struct flowi6 *fl6, int encap_limit,
713                                __u32 *pmtu, __be16 proto)
714 {
715         struct ip6_tnl *tunnel = netdev_priv(dev);
716         __be16 protocol;
717
718         if (dev->type == ARPHRD_ETHER)
719                 IPCB(skb)->flags = 0;
720
721         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
722                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
723         else
724                 fl6->daddr = tunnel->parms.raddr;
725
726         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
727                 return -ENOMEM;
728
729         /* Push GRE header. */
730         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
731
732         if (tunnel->parms.collect_md) {
733                 struct ip_tunnel_info *tun_info;
734                 const struct ip_tunnel_key *key;
735                 __be16 flags;
736
737                 tun_info = skb_tunnel_info(skb);
738                 if (unlikely(!tun_info ||
739                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
740                              ip_tunnel_info_af(tun_info) != AF_INET6))
741                         return -EINVAL;
742
743                 key = &tun_info->key;
744                 memset(fl6, 0, sizeof(*fl6));
745                 fl6->flowi6_proto = IPPROTO_GRE;
746                 fl6->daddr = key->u.ipv6.dst;
747                 fl6->flowlabel = key->label;
748                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
749
750                 dsfield = key->tos;
751                 flags = key->tun_flags &
752                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
753                 tunnel->tun_hlen = gre_calc_hlen(flags);
754
755                 gre_build_header(skb, tunnel->tun_hlen,
756                                  flags, protocol,
757                                  tunnel_id_to_key32(tun_info->key.tun_id),
758                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
759                                                       : 0);
760
761         } else {
762                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
763                         tunnel->o_seqno++;
764
765                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
766                                  protocol, tunnel->parms.o_key,
767                                  htonl(tunnel->o_seqno));
768         }
769
770         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
771                             NEXTHDR_GRE);
772 }
773
774 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
775 {
776         struct ip6_tnl *t = netdev_priv(dev);
777         int encap_limit = -1;
778         struct flowi6 fl6;
779         __u8 dsfield = 0;
780         __u32 mtu;
781         int err;
782
783         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
784
785         if (!t->parms.collect_md)
786                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
787                                          &dsfield, &encap_limit);
788
789         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
790         if (err)
791                 return -1;
792
793         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
794                           skb->protocol);
795         if (err != 0) {
796                 /* XXX: send ICMP error even if DF is not set. */
797                 if (err == -EMSGSIZE)
798                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
799                                   htonl(mtu));
800                 return -1;
801         }
802
803         return 0;
804 }
805
806 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
807 {
808         struct ip6_tnl *t = netdev_priv(dev);
809         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
810         int encap_limit = -1;
811         struct flowi6 fl6;
812         __u8 dsfield = 0;
813         __u32 mtu;
814         int err;
815
816         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
817                 return -1;
818
819         if (!t->parms.collect_md &&
820             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
821                 return -1;
822
823         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
824                 return -1;
825
826         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
827                           &mtu, skb->protocol);
828         if (err != 0) {
829                 if (err == -EMSGSIZE)
830                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
831                 return -1;
832         }
833
834         return 0;
835 }
836
837 /**
838  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
839  *   @t: the outgoing tunnel device
840  *   @hdr: IPv6 header from the incoming packet
841  *
842  * Description:
843  *   Avoid trivial tunneling loop by checking that tunnel exit-point
844  *   doesn't match source of incoming packet.
845  *
846  * Return:
847  *   1 if conflict,
848  *   0 else
849  **/
850
851 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
852         const struct ipv6hdr *hdr)
853 {
854         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
855 }
856
857 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
858 {
859         struct ip6_tnl *t = netdev_priv(dev);
860         int encap_limit = -1;
861         struct flowi6 fl6;
862         __u32 mtu;
863         int err;
864
865         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
866                 encap_limit = t->parms.encap_limit;
867
868         if (!t->parms.collect_md)
869                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
870
871         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
872         if (err)
873                 return err;
874
875         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
876
877         return err;
878 }
879
880 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
881         struct net_device *dev)
882 {
883         struct ip6_tnl *t = netdev_priv(dev);
884         struct net_device_stats *stats = &t->dev->stats;
885         int ret;
886
887         if (!pskb_inet_may_pull(skb))
888                 goto tx_err;
889
890         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
891                 goto tx_err;
892
893         switch (skb->protocol) {
894         case htons(ETH_P_IP):
895                 ret = ip6gre_xmit_ipv4(skb, dev);
896                 break;
897         case htons(ETH_P_IPV6):
898                 ret = ip6gre_xmit_ipv6(skb, dev);
899                 break;
900         default:
901                 ret = ip6gre_xmit_other(skb, dev);
902                 break;
903         }
904
905         if (ret < 0)
906                 goto tx_err;
907
908         return NETDEV_TX_OK;
909
910 tx_err:
911         stats->tx_errors++;
912         stats->tx_dropped++;
913         kfree_skb(skb);
914         return NETDEV_TX_OK;
915 }
916
917 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
918                                          struct net_device *dev)
919 {
920         struct ip6_tnl *t = netdev_priv(dev);
921         struct dst_entry *dst = skb_dst(skb);
922         struct net_device_stats *stats;
923         bool truncate = false;
924         int encap_limit = -1;
925         __u8 dsfield = false;
926         struct flowi6 fl6;
927         int err = -EINVAL;
928         __be16 proto;
929         __u32 mtu;
930         int nhoff;
931         int thoff;
932
933         if (!pskb_inet_may_pull(skb))
934                 goto tx_err;
935
936         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
937                 goto tx_err;
938
939         if (gre_handle_offloads(skb, false))
940                 goto tx_err;
941
942         if (skb->len > dev->mtu + dev->hard_header_len) {
943                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
944                 truncate = true;
945         }
946
947         nhoff = skb_network_header(skb) - skb_mac_header(skb);
948         if (skb->protocol == htons(ETH_P_IP) &&
949             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
950                 truncate = true;
951
952         thoff = skb_transport_header(skb) - skb_mac_header(skb);
953         if (skb->protocol == htons(ETH_P_IPV6) &&
954             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
955                 truncate = true;
956
957         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
958                 goto tx_err;
959
960         t->parms.o_flags &= ~TUNNEL_KEY;
961         IPCB(skb)->flags = 0;
962
963         /* For collect_md mode, derive fl6 from the tunnel key,
964          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
965          */
966         if (t->parms.collect_md) {
967                 struct ip_tunnel_info *tun_info;
968                 const struct ip_tunnel_key *key;
969                 struct erspan_metadata *md;
970                 __be32 tun_id;
971
972                 tun_info = skb_tunnel_info(skb);
973                 if (unlikely(!tun_info ||
974                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
975                              ip_tunnel_info_af(tun_info) != AF_INET6))
976                         goto tx_err;
977
978                 key = &tun_info->key;
979                 memset(&fl6, 0, sizeof(fl6));
980                 fl6.flowi6_proto = IPPROTO_GRE;
981                 fl6.daddr = key->u.ipv6.dst;
982                 fl6.flowlabel = key->label;
983                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
984
985                 dsfield = key->tos;
986                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
987                         goto tx_err;
988                 if (tun_info->options_len < sizeof(*md))
989                         goto tx_err;
990                 md = ip_tunnel_info_opts(tun_info);
991
992                 tun_id = tunnel_id_to_key32(key->tun_id);
993                 if (md->version == 1) {
994                         erspan_build_header(skb,
995                                             ntohl(tun_id),
996                                             ntohl(md->u.index), truncate,
997                                             false);
998                 } else if (md->version == 2) {
999                         erspan_build_header_v2(skb,
1000                                                ntohl(tun_id),
1001                                                md->u.md2.dir,
1002                                                get_hwid(&md->u.md2),
1003                                                truncate, false);
1004                 } else {
1005                         goto tx_err;
1006                 }
1007         } else {
1008                 switch (skb->protocol) {
1009                 case htons(ETH_P_IP):
1010                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1011                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1012                                                  &dsfield, &encap_limit);
1013                         break;
1014                 case htons(ETH_P_IPV6):
1015                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1016                                 goto tx_err;
1017                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1018                                                      &dsfield, &encap_limit))
1019                                 goto tx_err;
1020                         break;
1021                 default:
1022                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1023                         break;
1024                 }
1025
1026                 if (t->parms.erspan_ver == 1)
1027                         erspan_build_header(skb, ntohl(t->parms.o_key),
1028                                             t->parms.index,
1029                                             truncate, false);
1030                 else if (t->parms.erspan_ver == 2)
1031                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1032                                                t->parms.dir,
1033                                                t->parms.hwid,
1034                                                truncate, false);
1035                 else
1036                         goto tx_err;
1037
1038                 fl6.daddr = t->parms.raddr;
1039         }
1040
1041         /* Push GRE header. */
1042         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1043                                            : htons(ETH_P_ERSPAN2);
1044         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1045
1046         /* TooBig packet may have updated dst->dev's mtu */
1047         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1048                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1049
1050         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1051                            NEXTHDR_GRE);
1052         if (err != 0) {
1053                 /* XXX: send ICMP error even if DF is not set. */
1054                 if (err == -EMSGSIZE) {
1055                         if (skb->protocol == htons(ETH_P_IP))
1056                                 icmp_send(skb, ICMP_DEST_UNREACH,
1057                                           ICMP_FRAG_NEEDED, htonl(mtu));
1058                         else
1059                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1060                 }
1061
1062                 goto tx_err;
1063         }
1064         return NETDEV_TX_OK;
1065
1066 tx_err:
1067         stats = &t->dev->stats;
1068         stats->tx_errors++;
1069         stats->tx_dropped++;
1070         kfree_skb(skb);
1071         return NETDEV_TX_OK;
1072 }
1073
1074 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1075 {
1076         struct net_device *dev = t->dev;
1077         struct __ip6_tnl_parm *p = &t->parms;
1078         struct flowi6 *fl6 = &t->fl.u.ip6;
1079
1080         if (dev->type != ARPHRD_ETHER) {
1081                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1082                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1083         }
1084
1085         /* Set up flowi template */
1086         fl6->saddr = p->laddr;
1087         fl6->daddr = p->raddr;
1088         fl6->flowi6_oif = p->link;
1089         fl6->flowlabel = 0;
1090         fl6->flowi6_proto = IPPROTO_GRE;
1091
1092         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1093                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1094         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1095                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1096
1097         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1098         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1099
1100         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1101                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1102                 dev->flags |= IFF_POINTOPOINT;
1103         else
1104                 dev->flags &= ~IFF_POINTOPOINT;
1105 }
1106
1107 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1108                                          int t_hlen)
1109 {
1110         const struct __ip6_tnl_parm *p = &t->parms;
1111         struct net_device *dev = t->dev;
1112
1113         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1114                 int strict = (ipv6_addr_type(&p->raddr) &
1115                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1116
1117                 struct rt6_info *rt = rt6_lookup(t->net,
1118                                                  &p->raddr, &p->laddr,
1119                                                  p->link, NULL, strict);
1120
1121                 if (!rt)
1122                         return;
1123
1124                 if (rt->dst.dev) {
1125                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1126                                                t_hlen;
1127
1128                         if (set_mtu) {
1129                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1130                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1131                                         dev->mtu -= 8;
1132                                 if (dev->type == ARPHRD_ETHER)
1133                                         dev->mtu -= ETH_HLEN;
1134
1135                                 if (dev->mtu < IPV6_MIN_MTU)
1136                                         dev->mtu = IPV6_MIN_MTU;
1137                         }
1138                 }
1139                 ip6_rt_put(rt);
1140         }
1141 }
1142
1143 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1144 {
1145         int t_hlen;
1146
1147         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1148         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1149
1150         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1151         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1152         return t_hlen;
1153 }
1154
1155 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1156 {
1157         ip6gre_tnl_link_config_common(t);
1158         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1159 }
1160
1161 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1162                                      const struct __ip6_tnl_parm *p)
1163 {
1164         t->parms.laddr = p->laddr;
1165         t->parms.raddr = p->raddr;
1166         t->parms.flags = p->flags;
1167         t->parms.hop_limit = p->hop_limit;
1168         t->parms.encap_limit = p->encap_limit;
1169         t->parms.flowinfo = p->flowinfo;
1170         t->parms.link = p->link;
1171         t->parms.proto = p->proto;
1172         t->parms.i_key = p->i_key;
1173         t->parms.o_key = p->o_key;
1174         t->parms.i_flags = p->i_flags;
1175         t->parms.o_flags = p->o_flags;
1176         t->parms.fwmark = p->fwmark;
1177         t->parms.erspan_ver = p->erspan_ver;
1178         t->parms.index = p->index;
1179         t->parms.dir = p->dir;
1180         t->parms.hwid = p->hwid;
1181         dst_cache_reset(&t->dst_cache);
1182 }
1183
1184 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1185                              int set_mtu)
1186 {
1187         ip6gre_tnl_copy_tnl_parm(t, p);
1188         ip6gre_tnl_link_config(t, set_mtu);
1189         return 0;
1190 }
1191
1192 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1193         const struct ip6_tnl_parm2 *u)
1194 {
1195         p->laddr = u->laddr;
1196         p->raddr = u->raddr;
1197         p->flags = u->flags;
1198         p->hop_limit = u->hop_limit;
1199         p->encap_limit = u->encap_limit;
1200         p->flowinfo = u->flowinfo;
1201         p->link = u->link;
1202         p->i_key = u->i_key;
1203         p->o_key = u->o_key;
1204         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1205         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1206         memcpy(p->name, u->name, sizeof(u->name));
1207 }
1208
1209 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1210         const struct __ip6_tnl_parm *p)
1211 {
1212         u->proto = IPPROTO_GRE;
1213         u->laddr = p->laddr;
1214         u->raddr = p->raddr;
1215         u->flags = p->flags;
1216         u->hop_limit = p->hop_limit;
1217         u->encap_limit = p->encap_limit;
1218         u->flowinfo = p->flowinfo;
1219         u->link = p->link;
1220         u->i_key = p->i_key;
1221         u->o_key = p->o_key;
1222         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1223         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1224         memcpy(u->name, p->name, sizeof(u->name));
1225 }
1226
1227 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1228         struct ifreq *ifr, int cmd)
1229 {
1230         int err = 0;
1231         struct ip6_tnl_parm2 p;
1232         struct __ip6_tnl_parm p1;
1233         struct ip6_tnl *t = netdev_priv(dev);
1234         struct net *net = t->net;
1235         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1236
1237         memset(&p1, 0, sizeof(p1));
1238
1239         switch (cmd) {
1240         case SIOCGETTUNNEL:
1241                 if (dev == ign->fb_tunnel_dev) {
1242                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1243                                 err = -EFAULT;
1244                                 break;
1245                         }
1246                         ip6gre_tnl_parm_from_user(&p1, &p);
1247                         t = ip6gre_tunnel_locate(net, &p1, 0);
1248                         if (!t)
1249                                 t = netdev_priv(dev);
1250                 }
1251                 memset(&p, 0, sizeof(p));
1252                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1253                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1254                         err = -EFAULT;
1255                 break;
1256
1257         case SIOCADDTUNNEL:
1258         case SIOCCHGTUNNEL:
1259                 err = -EPERM;
1260                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1261                         goto done;
1262
1263                 err = -EFAULT;
1264                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1265                         goto done;
1266
1267                 err = -EINVAL;
1268                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1269                         goto done;
1270
1271                 if (!(p.i_flags&GRE_KEY))
1272                         p.i_key = 0;
1273                 if (!(p.o_flags&GRE_KEY))
1274                         p.o_key = 0;
1275
1276                 ip6gre_tnl_parm_from_user(&p1, &p);
1277                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1278
1279                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1280                         if (t) {
1281                                 if (t->dev != dev) {
1282                                         err = -EEXIST;
1283                                         break;
1284                                 }
1285                         } else {
1286                                 t = netdev_priv(dev);
1287
1288                                 ip6gre_tunnel_unlink(ign, t);
1289                                 synchronize_net();
1290                                 ip6gre_tnl_change(t, &p1, 1);
1291                                 ip6gre_tunnel_link(ign, t);
1292                                 netdev_state_change(dev);
1293                         }
1294                 }
1295
1296                 if (t) {
1297                         err = 0;
1298
1299                         memset(&p, 0, sizeof(p));
1300                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1301                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1302                                 err = -EFAULT;
1303                 } else
1304                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1305                 break;
1306
1307         case SIOCDELTUNNEL:
1308                 err = -EPERM;
1309                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1310                         goto done;
1311
1312                 if (dev == ign->fb_tunnel_dev) {
1313                         err = -EFAULT;
1314                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1315                                 goto done;
1316                         err = -ENOENT;
1317                         ip6gre_tnl_parm_from_user(&p1, &p);
1318                         t = ip6gre_tunnel_locate(net, &p1, 0);
1319                         if (!t)
1320                                 goto done;
1321                         err = -EPERM;
1322                         if (t == netdev_priv(ign->fb_tunnel_dev))
1323                                 goto done;
1324                         dev = t->dev;
1325                 }
1326                 unregister_netdevice(dev);
1327                 err = 0;
1328                 break;
1329
1330         default:
1331                 err = -EINVAL;
1332         }
1333
1334 done:
1335         return err;
1336 }
1337
1338 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1339                          unsigned short type, const void *daddr,
1340                          const void *saddr, unsigned int len)
1341 {
1342         struct ip6_tnl *t = netdev_priv(dev);
1343         struct ipv6hdr *ipv6h;
1344         __be16 *p;
1345
1346         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1347         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1348                                                   t->fl.u.ip6.flowlabel,
1349                                                   true, &t->fl.u.ip6));
1350         ipv6h->hop_limit = t->parms.hop_limit;
1351         ipv6h->nexthdr = NEXTHDR_GRE;
1352         ipv6h->saddr = t->parms.laddr;
1353         ipv6h->daddr = t->parms.raddr;
1354
1355         p = (__be16 *)(ipv6h + 1);
1356         p[0] = t->parms.o_flags;
1357         p[1] = htons(type);
1358
1359         /*
1360          *      Set the source hardware address.
1361          */
1362
1363         if (saddr)
1364                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1365         if (daddr)
1366                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1367         if (!ipv6_addr_any(&ipv6h->daddr))
1368                 return t->hlen;
1369
1370         return -t->hlen;
1371 }
1372
1373 static const struct header_ops ip6gre_header_ops = {
1374         .create = ip6gre_header,
1375 };
1376
1377 static const struct net_device_ops ip6gre_netdev_ops = {
1378         .ndo_init               = ip6gre_tunnel_init,
1379         .ndo_uninit             = ip6gre_tunnel_uninit,
1380         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1381         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1382         .ndo_change_mtu         = ip6_tnl_change_mtu,
1383         .ndo_get_stats64        = ip_tunnel_get_stats64,
1384         .ndo_get_iflink         = ip6_tnl_get_iflink,
1385 };
1386
1387 static void ip6gre_dev_free(struct net_device *dev)
1388 {
1389         struct ip6_tnl *t = netdev_priv(dev);
1390
1391         gro_cells_destroy(&t->gro_cells);
1392         dst_cache_destroy(&t->dst_cache);
1393         free_percpu(dev->tstats);
1394 }
1395
1396 static void ip6gre_tunnel_setup(struct net_device *dev)
1397 {
1398         dev->netdev_ops = &ip6gre_netdev_ops;
1399         dev->needs_free_netdev = true;
1400         dev->priv_destructor = ip6gre_dev_free;
1401
1402         dev->type = ARPHRD_IP6GRE;
1403
1404         dev->flags |= IFF_NOARP;
1405         dev->addr_len = sizeof(struct in6_addr);
1406         netif_keep_dst(dev);
1407         /* This perm addr will be used as interface identifier by IPv6 */
1408         dev->addr_assign_type = NET_ADDR_RANDOM;
1409         eth_random_addr(dev->perm_addr);
1410 }
1411
1412 #define GRE6_FEATURES (NETIF_F_SG |             \
1413                        NETIF_F_FRAGLIST |       \
1414                        NETIF_F_HIGHDMA |        \
1415                        NETIF_F_HW_CSUM)
1416
1417 static void ip6gre_tnl_init_features(struct net_device *dev)
1418 {
1419         struct ip6_tnl *nt = netdev_priv(dev);
1420
1421         dev->features           |= GRE6_FEATURES;
1422         dev->hw_features        |= GRE6_FEATURES;
1423
1424         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1425                 /* TCP offload with GRE SEQ is not supported, nor
1426                  * can we support 2 levels of outer headers requiring
1427                  * an update.
1428                  */
1429                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1430                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1431                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1432                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1433                 }
1434
1435                 /* Can use a lockless transmit, unless we generate
1436                  * output sequences
1437                  */
1438                 dev->features |= NETIF_F_LLTX;
1439         }
1440 }
1441
1442 static int ip6gre_tunnel_init_common(struct net_device *dev)
1443 {
1444         struct ip6_tnl *tunnel;
1445         int ret;
1446         int t_hlen;
1447
1448         tunnel = netdev_priv(dev);
1449
1450         tunnel->dev = dev;
1451         tunnel->net = dev_net(dev);
1452         strcpy(tunnel->parms.name, dev->name);
1453
1454         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1455         if (!dev->tstats)
1456                 return -ENOMEM;
1457
1458         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1459         if (ret)
1460                 goto cleanup_alloc_pcpu_stats;
1461
1462         ret = gro_cells_init(&tunnel->gro_cells, dev);
1463         if (ret)
1464                 goto cleanup_dst_cache_init;
1465
1466         t_hlen = ip6gre_calc_hlen(tunnel);
1467         dev->mtu = ETH_DATA_LEN - t_hlen;
1468         if (dev->type == ARPHRD_ETHER)
1469                 dev->mtu -= ETH_HLEN;
1470         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1471                 dev->mtu -= 8;
1472
1473         if (tunnel->parms.collect_md) {
1474                 netif_keep_dst(dev);
1475         }
1476         ip6gre_tnl_init_features(dev);
1477
1478         return 0;
1479
1480 cleanup_dst_cache_init:
1481         dst_cache_destroy(&tunnel->dst_cache);
1482 cleanup_alloc_pcpu_stats:
1483         free_percpu(dev->tstats);
1484         dev->tstats = NULL;
1485         return ret;
1486 }
1487
1488 static int ip6gre_tunnel_init(struct net_device *dev)
1489 {
1490         struct ip6_tnl *tunnel;
1491         int ret;
1492
1493         ret = ip6gre_tunnel_init_common(dev);
1494         if (ret)
1495                 return ret;
1496
1497         tunnel = netdev_priv(dev);
1498
1499         if (tunnel->parms.collect_md)
1500                 return 0;
1501
1502         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1503         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1504
1505         if (ipv6_addr_any(&tunnel->parms.raddr))
1506                 dev->header_ops = &ip6gre_header_ops;
1507
1508         return 0;
1509 }
1510
1511 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1512 {
1513         struct ip6_tnl *tunnel = netdev_priv(dev);
1514
1515         tunnel->dev = dev;
1516         tunnel->net = dev_net(dev);
1517         strcpy(tunnel->parms.name, dev->name);
1518
1519         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1520
1521         dev_hold(dev);
1522 }
1523
1524 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1525         .handler     = gre_rcv,
1526         .err_handler = ip6gre_err,
1527         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1528 };
1529
1530 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1531 {
1532         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1533         struct net_device *dev, *aux;
1534         int prio;
1535
1536         for_each_netdev_safe(net, dev, aux)
1537                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1538                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1539                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1540                         unregister_netdevice_queue(dev, head);
1541
1542         for (prio = 0; prio < 4; prio++) {
1543                 int h;
1544                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1545                         struct ip6_tnl *t;
1546
1547                         t = rtnl_dereference(ign->tunnels[prio][h]);
1548
1549                         while (t) {
1550                                 /* If dev is in the same netns, it has already
1551                                  * been added to the list by the previous loop.
1552                                  */
1553                                 if (!net_eq(dev_net(t->dev), net))
1554                                         unregister_netdevice_queue(t->dev,
1555                                                                    head);
1556                                 t = rtnl_dereference(t->next);
1557                         }
1558                 }
1559         }
1560 }
1561
1562 static int __net_init ip6gre_init_net(struct net *net)
1563 {
1564         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1565         int err;
1566
1567         if (!net_has_fallback_tunnels(net))
1568                 return 0;
1569         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1570                                           NET_NAME_UNKNOWN,
1571                                           ip6gre_tunnel_setup);
1572         if (!ign->fb_tunnel_dev) {
1573                 err = -ENOMEM;
1574                 goto err_alloc_dev;
1575         }
1576         dev_net_set(ign->fb_tunnel_dev, net);
1577         /* FB netdevice is special: we have one, and only one per netns.
1578          * Allowing to move it to another netns is clearly unsafe.
1579          */
1580         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1581
1582
1583         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1584         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1585
1586         err = register_netdev(ign->fb_tunnel_dev);
1587         if (err)
1588                 goto err_reg_dev;
1589
1590         rcu_assign_pointer(ign->tunnels_wc[0],
1591                            netdev_priv(ign->fb_tunnel_dev));
1592         return 0;
1593
1594 err_reg_dev:
1595         free_netdev(ign->fb_tunnel_dev);
1596 err_alloc_dev:
1597         return err;
1598 }
1599
1600 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1601 {
1602         struct net *net;
1603         LIST_HEAD(list);
1604
1605         rtnl_lock();
1606         list_for_each_entry(net, net_list, exit_list)
1607                 ip6gre_destroy_tunnels(net, &list);
1608         unregister_netdevice_many(&list);
1609         rtnl_unlock();
1610 }
1611
1612 static struct pernet_operations ip6gre_net_ops = {
1613         .init = ip6gre_init_net,
1614         .exit_batch = ip6gre_exit_batch_net,
1615         .id   = &ip6gre_net_id,
1616         .size = sizeof(struct ip6gre_net),
1617 };
1618
1619 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1620                                   struct netlink_ext_ack *extack)
1621 {
1622         __be16 flags;
1623
1624         if (!data)
1625                 return 0;
1626
1627         flags = 0;
1628         if (data[IFLA_GRE_IFLAGS])
1629                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1630         if (data[IFLA_GRE_OFLAGS])
1631                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1632         if (flags & (GRE_VERSION|GRE_ROUTING))
1633                 return -EINVAL;
1634
1635         return 0;
1636 }
1637
1638 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1639                                struct netlink_ext_ack *extack)
1640 {
1641         struct in6_addr daddr;
1642
1643         if (tb[IFLA_ADDRESS]) {
1644                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1645                         return -EINVAL;
1646                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1647                         return -EADDRNOTAVAIL;
1648         }
1649
1650         if (!data)
1651                 goto out;
1652
1653         if (data[IFLA_GRE_REMOTE]) {
1654                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1655                 if (ipv6_addr_any(&daddr))
1656                         return -EINVAL;
1657         }
1658
1659 out:
1660         return ip6gre_tunnel_validate(tb, data, extack);
1661 }
1662
1663 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1664                                   struct netlink_ext_ack *extack)
1665 {
1666         __be16 flags = 0;
1667         int ret, ver = 0;
1668
1669         if (!data)
1670                 return 0;
1671
1672         ret = ip6gre_tap_validate(tb, data, extack);
1673         if (ret)
1674                 return ret;
1675
1676         /* ERSPAN should only have GRE sequence and key flag */
1677         if (data[IFLA_GRE_OFLAGS])
1678                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1679         if (data[IFLA_GRE_IFLAGS])
1680                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1681         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1682             flags != (GRE_SEQ | GRE_KEY))
1683                 return -EINVAL;
1684
1685         /* ERSPAN Session ID only has 10-bit. Since we reuse
1686          * 32-bit key field as ID, check it's range.
1687          */
1688         if (data[IFLA_GRE_IKEY] &&
1689             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1690                 return -EINVAL;
1691
1692         if (data[IFLA_GRE_OKEY] &&
1693             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1694                 return -EINVAL;
1695
1696         if (data[IFLA_GRE_ERSPAN_VER]) {
1697                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1698                 if (ver != 1 && ver != 2)
1699                         return -EINVAL;
1700         }
1701
1702         if (ver == 1) {
1703                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1704                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1705
1706                         if (index & ~INDEX_MASK)
1707                                 return -EINVAL;
1708                 }
1709         } else if (ver == 2) {
1710                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1711                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1712
1713                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1714                                 return -EINVAL;
1715                 }
1716
1717                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1718                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1719
1720                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1721                                 return -EINVAL;
1722                 }
1723         }
1724
1725         return 0;
1726 }
1727
1728 static void ip6erspan_set_version(struct nlattr *data[],
1729                                   struct __ip6_tnl_parm *parms)
1730 {
1731         if (!data)
1732                 return;
1733
1734         parms->erspan_ver = 1;
1735         if (data[IFLA_GRE_ERSPAN_VER])
1736                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1737
1738         if (parms->erspan_ver == 1) {
1739                 if (data[IFLA_GRE_ERSPAN_INDEX])
1740                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1741         } else if (parms->erspan_ver == 2) {
1742                 if (data[IFLA_GRE_ERSPAN_DIR])
1743                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1744                 if (data[IFLA_GRE_ERSPAN_HWID])
1745                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1746         }
1747 }
1748
1749 static void ip6gre_netlink_parms(struct nlattr *data[],
1750                                 struct __ip6_tnl_parm *parms)
1751 {
1752         memset(parms, 0, sizeof(*parms));
1753
1754         if (!data)
1755                 return;
1756
1757         if (data[IFLA_GRE_LINK])
1758                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1759
1760         if (data[IFLA_GRE_IFLAGS])
1761                 parms->i_flags = gre_flags_to_tnl_flags(
1762                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1763
1764         if (data[IFLA_GRE_OFLAGS])
1765                 parms->o_flags = gre_flags_to_tnl_flags(
1766                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1767
1768         if (data[IFLA_GRE_IKEY])
1769                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1770
1771         if (data[IFLA_GRE_OKEY])
1772                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1773
1774         if (data[IFLA_GRE_LOCAL])
1775                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1776
1777         if (data[IFLA_GRE_REMOTE])
1778                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1779
1780         if (data[IFLA_GRE_TTL])
1781                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1782
1783         if (data[IFLA_GRE_ENCAP_LIMIT])
1784                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1785
1786         if (data[IFLA_GRE_FLOWINFO])
1787                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1788
1789         if (data[IFLA_GRE_FLAGS])
1790                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1791
1792         if (data[IFLA_GRE_FWMARK])
1793                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1794
1795         if (data[IFLA_GRE_COLLECT_METADATA])
1796                 parms->collect_md = true;
1797 }
1798
1799 static int ip6gre_tap_init(struct net_device *dev)
1800 {
1801         int ret;
1802
1803         ret = ip6gre_tunnel_init_common(dev);
1804         if (ret)
1805                 return ret;
1806
1807         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1808
1809         return 0;
1810 }
1811
1812 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1813         .ndo_init = ip6gre_tap_init,
1814         .ndo_uninit = ip6gre_tunnel_uninit,
1815         .ndo_start_xmit = ip6gre_tunnel_xmit,
1816         .ndo_set_mac_address = eth_mac_addr,
1817         .ndo_validate_addr = eth_validate_addr,
1818         .ndo_change_mtu = ip6_tnl_change_mtu,
1819         .ndo_get_stats64 = ip_tunnel_get_stats64,
1820         .ndo_get_iflink = ip6_tnl_get_iflink,
1821 };
1822
1823 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1824 {
1825         int t_hlen;
1826
1827         tunnel->tun_hlen = 8;
1828         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1829                        erspan_hdr_len(tunnel->parms.erspan_ver);
1830
1831         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1832         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1833         return t_hlen;
1834 }
1835
1836 static int ip6erspan_tap_init(struct net_device *dev)
1837 {
1838         struct ip6_tnl *tunnel;
1839         int t_hlen;
1840         int ret;
1841
1842         tunnel = netdev_priv(dev);
1843
1844         tunnel->dev = dev;
1845         tunnel->net = dev_net(dev);
1846         strcpy(tunnel->parms.name, dev->name);
1847
1848         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1849         if (!dev->tstats)
1850                 return -ENOMEM;
1851
1852         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1853         if (ret)
1854                 goto cleanup_alloc_pcpu_stats;
1855
1856         ret = gro_cells_init(&tunnel->gro_cells, dev);
1857         if (ret)
1858                 goto cleanup_dst_cache_init;
1859
1860         t_hlen = ip6erspan_calc_hlen(tunnel);
1861         dev->mtu = ETH_DATA_LEN - t_hlen;
1862         if (dev->type == ARPHRD_ETHER)
1863                 dev->mtu -= ETH_HLEN;
1864         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1865                 dev->mtu -= 8;
1866
1867         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1868         ip6erspan_tnl_link_config(tunnel, 1);
1869
1870         return 0;
1871
1872 cleanup_dst_cache_init:
1873         dst_cache_destroy(&tunnel->dst_cache);
1874 cleanup_alloc_pcpu_stats:
1875         free_percpu(dev->tstats);
1876         dev->tstats = NULL;
1877         return ret;
1878 }
1879
1880 static const struct net_device_ops ip6erspan_netdev_ops = {
1881         .ndo_init =             ip6erspan_tap_init,
1882         .ndo_uninit =           ip6erspan_tunnel_uninit,
1883         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1884         .ndo_set_mac_address =  eth_mac_addr,
1885         .ndo_validate_addr =    eth_validate_addr,
1886         .ndo_change_mtu =       ip6_tnl_change_mtu,
1887         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1888         .ndo_get_iflink =       ip6_tnl_get_iflink,
1889 };
1890
1891 static void ip6gre_tap_setup(struct net_device *dev)
1892 {
1893
1894         ether_setup(dev);
1895
1896         dev->max_mtu = 0;
1897         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1898         dev->needs_free_netdev = true;
1899         dev->priv_destructor = ip6gre_dev_free;
1900
1901         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1902         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1903         netif_keep_dst(dev);
1904 }
1905
1906 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1907                                        struct ip_tunnel_encap *ipencap)
1908 {
1909         bool ret = false;
1910
1911         memset(ipencap, 0, sizeof(*ipencap));
1912
1913         if (!data)
1914                 return ret;
1915
1916         if (data[IFLA_GRE_ENCAP_TYPE]) {
1917                 ret = true;
1918                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1919         }
1920
1921         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1922                 ret = true;
1923                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1924         }
1925
1926         if (data[IFLA_GRE_ENCAP_SPORT]) {
1927                 ret = true;
1928                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1929         }
1930
1931         if (data[IFLA_GRE_ENCAP_DPORT]) {
1932                 ret = true;
1933                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1934         }
1935
1936         return ret;
1937 }
1938
1939 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1940                                  struct nlattr *tb[], struct nlattr *data[],
1941                                  struct netlink_ext_ack *extack)
1942 {
1943         struct ip6_tnl *nt;
1944         struct ip_tunnel_encap ipencap;
1945         int err;
1946
1947         nt = netdev_priv(dev);
1948
1949         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1950                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1951
1952                 if (err < 0)
1953                         return err;
1954         }
1955
1956         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1957                 eth_hw_addr_random(dev);
1958
1959         nt->dev = dev;
1960         nt->net = dev_net(dev);
1961
1962         err = register_netdevice(dev);
1963         if (err)
1964                 goto out;
1965
1966         if (tb[IFLA_MTU])
1967                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1968
1969         dev_hold(dev);
1970
1971 out:
1972         return err;
1973 }
1974
1975 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1976                           struct nlattr *tb[], struct nlattr *data[],
1977                           struct netlink_ext_ack *extack)
1978 {
1979         struct ip6_tnl *nt = netdev_priv(dev);
1980         struct net *net = dev_net(dev);
1981         struct ip6gre_net *ign;
1982         int err;
1983
1984         ip6gre_netlink_parms(data, &nt->parms);
1985         ign = net_generic(net, ip6gre_net_id);
1986
1987         if (nt->parms.collect_md) {
1988                 if (rtnl_dereference(ign->collect_md_tun))
1989                         return -EEXIST;
1990         } else {
1991                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1992                         return -EEXIST;
1993         }
1994
1995         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1996         if (!err) {
1997                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1998                 ip6gre_tunnel_link_md(ign, nt);
1999                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2000         }
2001         return err;
2002 }
2003
2004 static struct ip6_tnl *
2005 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2006                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2007                          struct netlink_ext_ack *extack)
2008 {
2009         struct ip6_tnl *t, *nt = netdev_priv(dev);
2010         struct net *net = nt->net;
2011         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2012         struct ip_tunnel_encap ipencap;
2013
2014         if (dev == ign->fb_tunnel_dev)
2015                 return ERR_PTR(-EINVAL);
2016
2017         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2018                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2019
2020                 if (err < 0)
2021                         return ERR_PTR(err);
2022         }
2023
2024         ip6gre_netlink_parms(data, p_p);
2025
2026         t = ip6gre_tunnel_locate(net, p_p, 0);
2027
2028         if (t) {
2029                 if (t->dev != dev)
2030                         return ERR_PTR(-EEXIST);
2031         } else {
2032                 t = nt;
2033         }
2034
2035         return t;
2036 }
2037
2038 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2039                              struct nlattr *data[],
2040                              struct netlink_ext_ack *extack)
2041 {
2042         struct ip6_tnl *t = netdev_priv(dev);
2043         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2044         struct __ip6_tnl_parm p;
2045
2046         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2047         if (IS_ERR(t))
2048                 return PTR_ERR(t);
2049
2050         ip6gre_tunnel_unlink_md(ign, t);
2051         ip6gre_tunnel_unlink(ign, t);
2052         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2053         ip6gre_tunnel_link_md(ign, t);
2054         ip6gre_tunnel_link(ign, t);
2055         return 0;
2056 }
2057
2058 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2059 {
2060         struct net *net = dev_net(dev);
2061         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2062
2063         if (dev != ign->fb_tunnel_dev)
2064                 unregister_netdevice_queue(dev, head);
2065 }
2066
2067 static size_t ip6gre_get_size(const struct net_device *dev)
2068 {
2069         return
2070                 /* IFLA_GRE_LINK */
2071                 nla_total_size(4) +
2072                 /* IFLA_GRE_IFLAGS */
2073                 nla_total_size(2) +
2074                 /* IFLA_GRE_OFLAGS */
2075                 nla_total_size(2) +
2076                 /* IFLA_GRE_IKEY */
2077                 nla_total_size(4) +
2078                 /* IFLA_GRE_OKEY */
2079                 nla_total_size(4) +
2080                 /* IFLA_GRE_LOCAL */
2081                 nla_total_size(sizeof(struct in6_addr)) +
2082                 /* IFLA_GRE_REMOTE */
2083                 nla_total_size(sizeof(struct in6_addr)) +
2084                 /* IFLA_GRE_TTL */
2085                 nla_total_size(1) +
2086                 /* IFLA_GRE_ENCAP_LIMIT */
2087                 nla_total_size(1) +
2088                 /* IFLA_GRE_FLOWINFO */
2089                 nla_total_size(4) +
2090                 /* IFLA_GRE_FLAGS */
2091                 nla_total_size(4) +
2092                 /* IFLA_GRE_ENCAP_TYPE */
2093                 nla_total_size(2) +
2094                 /* IFLA_GRE_ENCAP_FLAGS */
2095                 nla_total_size(2) +
2096                 /* IFLA_GRE_ENCAP_SPORT */
2097                 nla_total_size(2) +
2098                 /* IFLA_GRE_ENCAP_DPORT */
2099                 nla_total_size(2) +
2100                 /* IFLA_GRE_COLLECT_METADATA */
2101                 nla_total_size(0) +
2102                 /* IFLA_GRE_FWMARK */
2103                 nla_total_size(4) +
2104                 /* IFLA_GRE_ERSPAN_INDEX */
2105                 nla_total_size(4) +
2106                 0;
2107 }
2108
2109 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2110 {
2111         struct ip6_tnl *t = netdev_priv(dev);
2112         struct __ip6_tnl_parm *p = &t->parms;
2113         __be16 o_flags = p->o_flags;
2114
2115         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2116                 if (!p->collect_md)
2117                         o_flags |= TUNNEL_KEY;
2118
2119                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2120                         goto nla_put_failure;
2121
2122                 if (p->erspan_ver == 1) {
2123                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2124                                 goto nla_put_failure;
2125                 } else {
2126                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2127                                 goto nla_put_failure;
2128                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2129                                 goto nla_put_failure;
2130                 }
2131         }
2132
2133         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2134             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2135                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2136             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2137                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2138             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2139             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2140             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2141             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2142             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2143             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2144             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2145             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2146             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2147                 goto nla_put_failure;
2148
2149         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2150                         t->encap.type) ||
2151             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2152                          t->encap.sport) ||
2153             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2154                          t->encap.dport) ||
2155             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2156                         t->encap.flags))
2157                 goto nla_put_failure;
2158
2159         if (p->collect_md) {
2160                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2161                         goto nla_put_failure;
2162         }
2163
2164         return 0;
2165
2166 nla_put_failure:
2167         return -EMSGSIZE;
2168 }
2169
2170 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2171         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2172         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2173         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2174         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2175         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2176         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2177         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2178         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2179         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2180         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2181         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2182         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2183         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2184         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2185         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2186         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2187         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2188         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2189         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2190         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2191         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2192 };
2193
2194 static void ip6erspan_tap_setup(struct net_device *dev)
2195 {
2196         ether_setup(dev);
2197
2198         dev->max_mtu = 0;
2199         dev->netdev_ops = &ip6erspan_netdev_ops;
2200         dev->needs_free_netdev = true;
2201         dev->priv_destructor = ip6gre_dev_free;
2202
2203         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2204         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2205         netif_keep_dst(dev);
2206 }
2207
2208 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2209                              struct nlattr *tb[], struct nlattr *data[],
2210                              struct netlink_ext_ack *extack)
2211 {
2212         struct ip6_tnl *nt = netdev_priv(dev);
2213         struct net *net = dev_net(dev);
2214         struct ip6gre_net *ign;
2215         int err;
2216
2217         ip6gre_netlink_parms(data, &nt->parms);
2218         ip6erspan_set_version(data, &nt->parms);
2219         ign = net_generic(net, ip6gre_net_id);
2220
2221         if (nt->parms.collect_md) {
2222                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2223                         return -EEXIST;
2224         } else {
2225                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2226                         return -EEXIST;
2227         }
2228
2229         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2230         if (!err) {
2231                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2232                 ip6erspan_tunnel_link_md(ign, nt);
2233                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2234         }
2235         return err;
2236 }
2237
2238 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2239 {
2240         ip6gre_tnl_link_config_common(t);
2241         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2242 }
2243
2244 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2245                                 const struct __ip6_tnl_parm *p, int set_mtu)
2246 {
2247         ip6gre_tnl_copy_tnl_parm(t, p);
2248         ip6erspan_tnl_link_config(t, set_mtu);
2249         return 0;
2250 }
2251
2252 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2253                                 struct nlattr *data[],
2254                                 struct netlink_ext_ack *extack)
2255 {
2256         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2257         struct __ip6_tnl_parm p;
2258         struct ip6_tnl *t;
2259
2260         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2261         if (IS_ERR(t))
2262                 return PTR_ERR(t);
2263
2264         ip6erspan_set_version(data, &p);
2265         ip6gre_tunnel_unlink_md(ign, t);
2266         ip6gre_tunnel_unlink(ign, t);
2267         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2268         ip6erspan_tunnel_link_md(ign, t);
2269         ip6gre_tunnel_link(ign, t);
2270         return 0;
2271 }
2272
2273 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2274         .kind           = "ip6gre",
2275         .maxtype        = IFLA_GRE_MAX,
2276         .policy         = ip6gre_policy,
2277         .priv_size      = sizeof(struct ip6_tnl),
2278         .setup          = ip6gre_tunnel_setup,
2279         .validate       = ip6gre_tunnel_validate,
2280         .newlink        = ip6gre_newlink,
2281         .changelink     = ip6gre_changelink,
2282         .dellink        = ip6gre_dellink,
2283         .get_size       = ip6gre_get_size,
2284         .fill_info      = ip6gre_fill_info,
2285         .get_link_net   = ip6_tnl_get_link_net,
2286 };
2287
2288 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2289         .kind           = "ip6gretap",
2290         .maxtype        = IFLA_GRE_MAX,
2291         .policy         = ip6gre_policy,
2292         .priv_size      = sizeof(struct ip6_tnl),
2293         .setup          = ip6gre_tap_setup,
2294         .validate       = ip6gre_tap_validate,
2295         .newlink        = ip6gre_newlink,
2296         .changelink     = ip6gre_changelink,
2297         .get_size       = ip6gre_get_size,
2298         .fill_info      = ip6gre_fill_info,
2299         .get_link_net   = ip6_tnl_get_link_net,
2300 };
2301
2302 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2303         .kind           = "ip6erspan",
2304         .maxtype        = IFLA_GRE_MAX,
2305         .policy         = ip6gre_policy,
2306         .priv_size      = sizeof(struct ip6_tnl),
2307         .setup          = ip6erspan_tap_setup,
2308         .validate       = ip6erspan_tap_validate,
2309         .newlink        = ip6erspan_newlink,
2310         .changelink     = ip6erspan_changelink,
2311         .get_size       = ip6gre_get_size,
2312         .fill_info      = ip6gre_fill_info,
2313         .get_link_net   = ip6_tnl_get_link_net,
2314 };
2315
2316 /*
2317  *      And now the modules code and kernel interface.
2318  */
2319
2320 static int __init ip6gre_init(void)
2321 {
2322         int err;
2323
2324         pr_info("GRE over IPv6 tunneling driver\n");
2325
2326         err = register_pernet_device(&ip6gre_net_ops);
2327         if (err < 0)
2328                 return err;
2329
2330         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2331         if (err < 0) {
2332                 pr_info("%s: can't add protocol\n", __func__);
2333                 goto add_proto_failed;
2334         }
2335
2336         err = rtnl_link_register(&ip6gre_link_ops);
2337         if (err < 0)
2338                 goto rtnl_link_failed;
2339
2340         err = rtnl_link_register(&ip6gre_tap_ops);
2341         if (err < 0)
2342                 goto tap_ops_failed;
2343
2344         err = rtnl_link_register(&ip6erspan_tap_ops);
2345         if (err < 0)
2346                 goto erspan_link_failed;
2347
2348 out:
2349         return err;
2350
2351 erspan_link_failed:
2352         rtnl_link_unregister(&ip6gre_tap_ops);
2353 tap_ops_failed:
2354         rtnl_link_unregister(&ip6gre_link_ops);
2355 rtnl_link_failed:
2356         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2357 add_proto_failed:
2358         unregister_pernet_device(&ip6gre_net_ops);
2359         goto out;
2360 }
2361
2362 static void __exit ip6gre_fini(void)
2363 {
2364         rtnl_link_unregister(&ip6gre_tap_ops);
2365         rtnl_link_unregister(&ip6gre_link_ops);
2366         rtnl_link_unregister(&ip6erspan_tap_ops);
2367         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2368         unregister_pernet_device(&ip6gre_net_ops);
2369 }
2370
2371 module_init(ip6gre_init);
2372 module_exit(ip6gre_fini);
2373 MODULE_LICENSE("GPL");
2374 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2375 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2376 MODULE_ALIAS_RTNL_LINK("ip6gre");
2377 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2378 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2379 MODULE_ALIAS_NETDEV("ip6gre0");