GNU Linux-libre 4.19.211-gnu1
[releases.git] / net / ipv6 / ip6_gre.c
1 /*
2  *      GRE over IPv6 protocol decoder.
3  *
4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
5  *
6  *      This program is free software; you can redistribute it and/or
7  *      modify it under the terms of the GNU General Public License
8  *      as published by the Free Software Foundation; either version
9  *      2 of the License, or (at your option) any later version.
10  *
11  */
12
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/init.h>
28 #include <linux/in6.h>
29 #include <linux/inetdevice.h>
30 #include <linux/igmp.h>
31 #include <linux/netfilter_ipv4.h>
32 #include <linux/etherdevice.h>
33 #include <linux/if_ether.h>
34 #include <linux/hash.h>
35 #include <linux/if_tunnel.h>
36 #include <linux/ip6_tunnel.h>
37
38 #include <net/sock.h>
39 #include <net/ip.h>
40 #include <net/ip_tunnels.h>
41 #include <net/icmp.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
44 #include <net/arp.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
48 #include <net/xfrm.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
52
53 #include <net/ipv6.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.h>
57 #include <net/gre.h>
58 #include <net/erspan.h>
59 #include <net/dst_metadata.h>
60
61
62 static bool log_ecn_error = true;
63 module_param(log_ecn_error, bool, 0644);
64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
65
66 #define IP6_GRE_HASH_SIZE_SHIFT  5
67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
68
69 static unsigned int ip6gre_net_id __read_mostly;
70 struct ip6gre_net {
71         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
72
73         struct ip6_tnl __rcu *collect_md_tun;
74         struct ip6_tnl __rcu *collect_md_tun_erspan;
75         struct net_device *fb_tunnel_dev;
76 };
77
78 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
80 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
81 static int ip6gre_tunnel_init(struct net_device *dev);
82 static void ip6gre_tunnel_setup(struct net_device *dev);
83 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
84 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
85 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
86
87 /* Tunnel hash table */
88
89 /*
90    4 hash tables:
91
92    3: (remote,local)
93    2: (remote,*)
94    1: (*,local)
95    0: (*,*)
96
97    We require exact key match i.e. if a key is present in packet
98    it will match only tunnel with the same key; if it is not present,
99    it will match only keyless tunnel.
100
101    All keysless packets, if not matched configured keyless tunnels
102    will match fallback tunnel.
103  */
104
105 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
106 static u32 HASH_ADDR(const struct in6_addr *addr)
107 {
108         u32 hash = ipv6_addr_hash(addr);
109
110         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
111 }
112
113 #define tunnels_r_l     tunnels[3]
114 #define tunnels_r       tunnels[2]
115 #define tunnels_l       tunnels[1]
116 #define tunnels_wc      tunnels[0]
117
118 /* Given src, dst and key, find appropriate for input tunnel. */
119
120 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
121                 const struct in6_addr *remote, const struct in6_addr *local,
122                 __be32 key, __be16 gre_proto)
123 {
124         struct net *net = dev_net(dev);
125         int link = dev->ifindex;
126         unsigned int h0 = HASH_ADDR(remote);
127         unsigned int h1 = HASH_KEY(key);
128         struct ip6_tnl *t, *cand = NULL;
129         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
130         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
131                         gre_proto == htons(ETH_P_ERSPAN) ||
132                         gre_proto == htons(ETH_P_ERSPAN2)) ?
133                        ARPHRD_ETHER : ARPHRD_IP6GRE;
134         int score, cand_score = 4;
135         struct net_device *ndev;
136
137         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
138                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
139                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
140                     key != t->parms.i_key ||
141                     !(t->dev->flags & IFF_UP))
142                         continue;
143
144                 if (t->dev->type != ARPHRD_IP6GRE &&
145                     t->dev->type != dev_type)
146                         continue;
147
148                 score = 0;
149                 if (t->parms.link != link)
150                         score |= 1;
151                 if (t->dev->type != dev_type)
152                         score |= 2;
153                 if (score == 0)
154                         return t;
155
156                 if (score < cand_score) {
157                         cand = t;
158                         cand_score = score;
159                 }
160         }
161
162         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
163                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
164                     key != t->parms.i_key ||
165                     !(t->dev->flags & IFF_UP))
166                         continue;
167
168                 if (t->dev->type != ARPHRD_IP6GRE &&
169                     t->dev->type != dev_type)
170                         continue;
171
172                 score = 0;
173                 if (t->parms.link != link)
174                         score |= 1;
175                 if (t->dev->type != dev_type)
176                         score |= 2;
177                 if (score == 0)
178                         return t;
179
180                 if (score < cand_score) {
181                         cand = t;
182                         cand_score = score;
183                 }
184         }
185
186         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
187                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
188                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
189                                  !ipv6_addr_is_multicast(local))) ||
190                     key != t->parms.i_key ||
191                     !(t->dev->flags & IFF_UP))
192                         continue;
193
194                 if (t->dev->type != ARPHRD_IP6GRE &&
195                     t->dev->type != dev_type)
196                         continue;
197
198                 score = 0;
199                 if (t->parms.link != link)
200                         score |= 1;
201                 if (t->dev->type != dev_type)
202                         score |= 2;
203                 if (score == 0)
204                         return t;
205
206                 if (score < cand_score) {
207                         cand = t;
208                         cand_score = score;
209                 }
210         }
211
212         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
213                 if (t->parms.i_key != key ||
214                     !(t->dev->flags & IFF_UP))
215                         continue;
216
217                 if (t->dev->type != ARPHRD_IP6GRE &&
218                     t->dev->type != dev_type)
219                         continue;
220
221                 score = 0;
222                 if (t->parms.link != link)
223                         score |= 1;
224                 if (t->dev->type != dev_type)
225                         score |= 2;
226                 if (score == 0)
227                         return t;
228
229                 if (score < cand_score) {
230                         cand = t;
231                         cand_score = score;
232                 }
233         }
234
235         if (cand)
236                 return cand;
237
238         if (gre_proto == htons(ETH_P_ERSPAN) ||
239             gre_proto == htons(ETH_P_ERSPAN2))
240                 t = rcu_dereference(ign->collect_md_tun_erspan);
241         else
242                 t = rcu_dereference(ign->collect_md_tun);
243
244         if (t && t->dev->flags & IFF_UP)
245                 return t;
246
247         ndev = READ_ONCE(ign->fb_tunnel_dev);
248         if (ndev && ndev->flags & IFF_UP)
249                 return netdev_priv(ndev);
250
251         return NULL;
252 }
253
254 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
255                 const struct __ip6_tnl_parm *p)
256 {
257         const struct in6_addr *remote = &p->raddr;
258         const struct in6_addr *local = &p->laddr;
259         unsigned int h = HASH_KEY(p->i_key);
260         int prio = 0;
261
262         if (!ipv6_addr_any(local))
263                 prio |= 1;
264         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
265                 prio |= 2;
266                 h ^= HASH_ADDR(remote);
267         }
268
269         return &ign->tunnels[prio][h];
270 }
271
272 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
273 {
274         if (t->parms.collect_md)
275                 rcu_assign_pointer(ign->collect_md_tun, t);
276 }
277
278 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
279 {
280         if (t->parms.collect_md)
281                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
282 }
283
284 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
285 {
286         if (t->parms.collect_md)
287                 rcu_assign_pointer(ign->collect_md_tun, NULL);
288 }
289
290 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
291                                        struct ip6_tnl *t)
292 {
293         if (t->parms.collect_md)
294                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
295 }
296
297 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
298                 const struct ip6_tnl *t)
299 {
300         return __ip6gre_bucket(ign, &t->parms);
301 }
302
303 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
304 {
305         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
306
307         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
308         rcu_assign_pointer(*tp, t);
309 }
310
311 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
312 {
313         struct ip6_tnl __rcu **tp;
314         struct ip6_tnl *iter;
315
316         for (tp = ip6gre_bucket(ign, t);
317              (iter = rtnl_dereference(*tp)) != NULL;
318              tp = &iter->next) {
319                 if (t == iter) {
320                         rcu_assign_pointer(*tp, t->next);
321                         break;
322                 }
323         }
324 }
325
326 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
327                                            const struct __ip6_tnl_parm *parms,
328                                            int type)
329 {
330         const struct in6_addr *remote = &parms->raddr;
331         const struct in6_addr *local = &parms->laddr;
332         __be32 key = parms->i_key;
333         int link = parms->link;
334         struct ip6_tnl *t;
335         struct ip6_tnl __rcu **tp;
336         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
337
338         for (tp = __ip6gre_bucket(ign, parms);
339              (t = rtnl_dereference(*tp)) != NULL;
340              tp = &t->next)
341                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
342                     ipv6_addr_equal(remote, &t->parms.raddr) &&
343                     key == t->parms.i_key &&
344                     link == t->parms.link &&
345                     type == t->dev->type)
346                         break;
347
348         return t;
349 }
350
351 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
352                 const struct __ip6_tnl_parm *parms, int create)
353 {
354         struct ip6_tnl *t, *nt;
355         struct net_device *dev;
356         char name[IFNAMSIZ];
357         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
358
359         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
360         if (t && create)
361                 return NULL;
362         if (t || !create)
363                 return t;
364
365         if (parms->name[0]) {
366                 if (!dev_valid_name(parms->name))
367                         return NULL;
368                 strlcpy(name, parms->name, IFNAMSIZ);
369         } else {
370                 strcpy(name, "ip6gre%d");
371         }
372         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
373                            ip6gre_tunnel_setup);
374         if (!dev)
375                 return NULL;
376
377         dev_net_set(dev, net);
378
379         nt = netdev_priv(dev);
380         nt->parms = *parms;
381         dev->rtnl_link_ops = &ip6gre_link_ops;
382
383         nt->dev = dev;
384         nt->net = dev_net(dev);
385
386         if (register_netdevice(dev) < 0)
387                 goto failed_free;
388
389         ip6gre_tnl_link_config(nt, 1);
390
391         /* Can use a lockless transmit, unless we generate output sequences */
392         if (!(nt->parms.o_flags & TUNNEL_SEQ))
393                 dev->features |= NETIF_F_LLTX;
394
395         ip6gre_tunnel_link(ign, nt);
396         return nt;
397
398 failed_free:
399         free_netdev(dev);
400         return NULL;
401 }
402
403 static void ip6erspan_tunnel_uninit(struct net_device *dev)
404 {
405         struct ip6_tnl *t = netdev_priv(dev);
406         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
407
408         ip6erspan_tunnel_unlink_md(ign, t);
409         ip6gre_tunnel_unlink(ign, t);
410         dst_cache_reset(&t->dst_cache);
411         dev_put(dev);
412 }
413
414 static void ip6gre_tunnel_uninit(struct net_device *dev)
415 {
416         struct ip6_tnl *t = netdev_priv(dev);
417         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
418
419         ip6gre_tunnel_unlink_md(ign, t);
420         ip6gre_tunnel_unlink(ign, t);
421         if (ign->fb_tunnel_dev == dev)
422                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
423         dst_cache_reset(&t->dst_cache);
424         dev_put(dev);
425 }
426
427
428 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
429                        u8 type, u8 code, int offset, __be32 info)
430 {
431         struct net *net = dev_net(skb->dev);
432         const struct gre_base_hdr *greh;
433         const struct ipv6hdr *ipv6h;
434         int grehlen = sizeof(*greh);
435         struct ip6_tnl *t;
436         int key_off = 0;
437         __be16 flags;
438         __be32 key;
439
440         if (!pskb_may_pull(skb, offset + grehlen))
441                 return;
442         greh = (const struct gre_base_hdr *)(skb->data + offset);
443         flags = greh->flags;
444         if (flags & (GRE_VERSION | GRE_ROUTING))
445                 return;
446         if (flags & GRE_CSUM)
447                 grehlen += 4;
448         if (flags & GRE_KEY) {
449                 key_off = grehlen + offset;
450                 grehlen += 4;
451         }
452
453         if (!pskb_may_pull(skb, offset + grehlen))
454                 return;
455         ipv6h = (const struct ipv6hdr *)skb->data;
456         greh = (const struct gre_base_hdr *)(skb->data + offset);
457         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
458
459         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
460                                  key, greh->protocol);
461         if (!t)
462                 return;
463
464         switch (type) {
465                 struct ipv6_tlv_tnl_enc_lim *tel;
466                 __u32 teli;
467         case ICMPV6_DEST_UNREACH:
468                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
469                                     t->parms.name);
470                 if (code != ICMPV6_PORT_UNREACH)
471                         break;
472                 return;
473         case ICMPV6_TIME_EXCEED:
474                 if (code == ICMPV6_EXC_HOPLIMIT) {
475                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
476                                             t->parms.name);
477                         break;
478                 }
479                 return;
480         case ICMPV6_PARAMPROB:
481                 teli = 0;
482                 if (code == ICMPV6_HDR_FIELD)
483                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
484
485                 if (teli && teli == be32_to_cpu(info) - 2) {
486                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
487                         if (tel->encap_limit == 0) {
488                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
489                                                     t->parms.name);
490                         }
491                 } else {
492                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
493                                             t->parms.name);
494                 }
495                 return;
496         case ICMPV6_PKT_TOOBIG:
497                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
498                 return;
499         case NDISC_REDIRECT:
500                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
501                              sock_net_uid(net, NULL));
502                 return;
503         }
504
505         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
506                 t->err_count++;
507         else
508                 t->err_count = 1;
509         t->err_time = jiffies;
510 }
511
512 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
513 {
514         const struct ipv6hdr *ipv6h;
515         struct ip6_tnl *tunnel;
516
517         ipv6h = ipv6_hdr(skb);
518         tunnel = ip6gre_tunnel_lookup(skb->dev,
519                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
520                                       tpi->proto);
521         if (tunnel) {
522                 if (tunnel->parms.collect_md) {
523                         struct metadata_dst *tun_dst;
524                         __be64 tun_id;
525                         __be16 flags;
526
527                         flags = tpi->flags;
528                         tun_id = key32_to_tunnel_id(tpi->key);
529
530                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
531                         if (!tun_dst)
532                                 return PACKET_REJECT;
533
534                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
535                 } else {
536                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
537                 }
538
539                 return PACKET_RCVD;
540         }
541
542         return PACKET_REJECT;
543 }
544
545 static int ip6erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
546                          int gre_hdr_len)
547 {
548         struct erspan_base_hdr *ershdr;
549         const struct ipv6hdr *ipv6h;
550         struct erspan_md2 *md2;
551         struct ip6_tnl *tunnel;
552         u8 ver;
553
554         ipv6h = ipv6_hdr(skb);
555         ershdr = (struct erspan_base_hdr *)skb->data;
556         ver = ershdr->ver;
557
558         tunnel = ip6gre_tunnel_lookup(skb->dev,
559                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
560                                       tpi->proto);
561         if (tunnel) {
562                 int len = erspan_hdr_len(ver);
563
564                 if (unlikely(!pskb_may_pull(skb, len)))
565                         return PACKET_REJECT;
566
567                 if (__iptunnel_pull_header(skb, len,
568                                            htons(ETH_P_TEB),
569                                            false, false) < 0)
570                         return PACKET_REJECT;
571
572                 if (tunnel->parms.collect_md) {
573                         struct erspan_metadata *pkt_md, *md;
574                         struct metadata_dst *tun_dst;
575                         struct ip_tunnel_info *info;
576                         unsigned char *gh;
577                         __be64 tun_id;
578                         __be16 flags;
579
580                         tpi->flags |= TUNNEL_KEY;
581                         flags = tpi->flags;
582                         tun_id = key32_to_tunnel_id(tpi->key);
583
584                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
585                                                   sizeof(*md));
586                         if (!tun_dst)
587                                 return PACKET_REJECT;
588
589                         /* skb can be uncloned in __iptunnel_pull_header, so
590                          * old pkt_md is no longer valid and we need to reset
591                          * it
592                          */
593                         gh = skb_network_header(skb) +
594                              skb_network_header_len(skb);
595                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
596                                                             sizeof(*ershdr));
597                         info = &tun_dst->u.tun_info;
598                         md = ip_tunnel_info_opts(info);
599                         md->version = ver;
600                         md2 = &md->u.md2;
601                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
602                                                        ERSPAN_V2_MDSIZE);
603                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
604                         info->options_len = sizeof(*md);
605
606                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
607
608                 } else {
609                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
610                 }
611
612                 return PACKET_RCVD;
613         }
614
615         return PACKET_REJECT;
616 }
617
618 static int gre_rcv(struct sk_buff *skb)
619 {
620         struct tnl_ptk_info tpi;
621         bool csum_err = false;
622         int hdr_len;
623
624         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
625         if (hdr_len < 0)
626                 goto drop;
627
628         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
629                 goto drop;
630
631         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
632                      tpi.proto == htons(ETH_P_ERSPAN2))) {
633                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
634                         return 0;
635                 goto out;
636         }
637
638         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
639                 return 0;
640
641 out:
642         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
643 drop:
644         kfree_skb(skb);
645         return 0;
646 }
647
648 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
649 {
650         return iptunnel_handle_offloads(skb,
651                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
652 }
653
654 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
655                                      struct net_device *dev,
656                                      struct flowi6 *fl6, __u8 *dsfield,
657                                      int *encap_limit)
658 {
659         const struct iphdr *iph = ip_hdr(skb);
660         struct ip6_tnl *t = netdev_priv(dev);
661
662         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
663                 *encap_limit = t->parms.encap_limit;
664
665         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
666
667         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
668                 *dsfield = ipv4_get_dsfield(iph);
669         else
670                 *dsfield = ip6_tclass(t->parms.flowinfo);
671
672         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
673                 fl6->flowi6_mark = skb->mark;
674         else
675                 fl6->flowi6_mark = t->parms.fwmark;
676
677         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
678 }
679
680 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
681                                     struct net_device *dev,
682                                     struct flowi6 *fl6, __u8 *dsfield,
683                                     int *encap_limit)
684 {
685         struct ipv6hdr *ipv6h;
686         struct ip6_tnl *t = netdev_priv(dev);
687         __u16 offset;
688
689         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
690         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
691         ipv6h = ipv6_hdr(skb);
692
693         if (offset > 0) {
694                 struct ipv6_tlv_tnl_enc_lim *tel;
695
696                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
697                 if (tel->encap_limit == 0) {
698                         icmpv6_send(skb, ICMPV6_PARAMPROB,
699                                     ICMPV6_HDR_FIELD, offset + 2);
700                         return -1;
701                 }
702                 *encap_limit = tel->encap_limit - 1;
703         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
704                 *encap_limit = t->parms.encap_limit;
705         }
706
707         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
708
709         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
710                 *dsfield = ipv6_get_dsfield(ipv6h);
711         else
712                 *dsfield = ip6_tclass(t->parms.flowinfo);
713
714         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
715                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
716
717         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
718                 fl6->flowi6_mark = skb->mark;
719         else
720                 fl6->flowi6_mark = t->parms.fwmark;
721
722         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
723
724         return 0;
725 }
726
727 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
728                                struct net_device *dev, __u8 dsfield,
729                                struct flowi6 *fl6, int encap_limit,
730                                __u32 *pmtu, __be16 proto)
731 {
732         struct ip6_tnl *tunnel = netdev_priv(dev);
733         __be16 protocol;
734
735         if (dev->type == ARPHRD_ETHER)
736                 IPCB(skb)->flags = 0;
737
738         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
739                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
740         else
741                 fl6->daddr = tunnel->parms.raddr;
742
743         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
744                 return -ENOMEM;
745
746         /* Push GRE header. */
747         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
748
749         if (tunnel->parms.collect_md) {
750                 struct ip_tunnel_info *tun_info;
751                 const struct ip_tunnel_key *key;
752                 __be16 flags;
753
754                 tun_info = skb_tunnel_info(skb);
755                 if (unlikely(!tun_info ||
756                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
757                              ip_tunnel_info_af(tun_info) != AF_INET6))
758                         return -EINVAL;
759
760                 key = &tun_info->key;
761                 memset(fl6, 0, sizeof(*fl6));
762                 fl6->flowi6_proto = IPPROTO_GRE;
763                 fl6->daddr = key->u.ipv6.dst;
764                 fl6->flowlabel = key->label;
765                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
766
767                 dsfield = key->tos;
768                 flags = key->tun_flags &
769                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
770                 tunnel->tun_hlen = gre_calc_hlen(flags);
771
772                 gre_build_header(skb, tunnel->tun_hlen,
773                                  flags, protocol,
774                                  tunnel_id_to_key32(tun_info->key.tun_id),
775                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
776                                                       : 0);
777
778         } else {
779                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
780                         tunnel->o_seqno++;
781
782                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
783                                  protocol, tunnel->parms.o_key,
784                                  htonl(tunnel->o_seqno));
785         }
786
787         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
788                             NEXTHDR_GRE);
789 }
790
791 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
792 {
793         struct ip6_tnl *t = netdev_priv(dev);
794         int encap_limit = -1;
795         struct flowi6 fl6;
796         __u8 dsfield = 0;
797         __u32 mtu;
798         int err;
799
800         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
801
802         if (!t->parms.collect_md)
803                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
804                                          &dsfield, &encap_limit);
805
806         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
807         if (err)
808                 return -1;
809
810         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
811                           skb->protocol);
812         if (err != 0) {
813                 /* XXX: send ICMP error even if DF is not set. */
814                 if (err == -EMSGSIZE)
815                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
816                                   htonl(mtu));
817                 return -1;
818         }
819
820         return 0;
821 }
822
823 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
824 {
825         struct ip6_tnl *t = netdev_priv(dev);
826         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
827         int encap_limit = -1;
828         struct flowi6 fl6;
829         __u8 dsfield = 0;
830         __u32 mtu;
831         int err;
832
833         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
834                 return -1;
835
836         if (!t->parms.collect_md &&
837             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
838                 return -1;
839
840         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
841                 return -1;
842
843         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
844                           &mtu, skb->protocol);
845         if (err != 0) {
846                 if (err == -EMSGSIZE)
847                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
848                 return -1;
849         }
850
851         return 0;
852 }
853
854 /**
855  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
856  *   @t: the outgoing tunnel device
857  *   @hdr: IPv6 header from the incoming packet
858  *
859  * Description:
860  *   Avoid trivial tunneling loop by checking that tunnel exit-point
861  *   doesn't match source of incoming packet.
862  *
863  * Return:
864  *   1 if conflict,
865  *   0 else
866  **/
867
868 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
869         const struct ipv6hdr *hdr)
870 {
871         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
872 }
873
874 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
875 {
876         struct ip6_tnl *t = netdev_priv(dev);
877         int encap_limit = -1;
878         struct flowi6 fl6;
879         __u32 mtu;
880         int err;
881
882         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
883                 encap_limit = t->parms.encap_limit;
884
885         if (!t->parms.collect_md)
886                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
887
888         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
889         if (err)
890                 return err;
891
892         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
893
894         return err;
895 }
896
897 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
898         struct net_device *dev)
899 {
900         struct ip6_tnl *t = netdev_priv(dev);
901         struct net_device_stats *stats = &t->dev->stats;
902         int ret;
903
904         if (!pskb_inet_may_pull(skb))
905                 goto tx_err;
906
907         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
908                 goto tx_err;
909
910         switch (skb->protocol) {
911         case htons(ETH_P_IP):
912                 ret = ip6gre_xmit_ipv4(skb, dev);
913                 break;
914         case htons(ETH_P_IPV6):
915                 ret = ip6gre_xmit_ipv6(skb, dev);
916                 break;
917         default:
918                 ret = ip6gre_xmit_other(skb, dev);
919                 break;
920         }
921
922         if (ret < 0)
923                 goto tx_err;
924
925         return NETDEV_TX_OK;
926
927 tx_err:
928         stats->tx_errors++;
929         stats->tx_dropped++;
930         kfree_skb(skb);
931         return NETDEV_TX_OK;
932 }
933
934 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
935                                          struct net_device *dev)
936 {
937         struct ip6_tnl *t = netdev_priv(dev);
938         struct dst_entry *dst = skb_dst(skb);
939         struct net_device_stats *stats;
940         bool truncate = false;
941         int encap_limit = -1;
942         __u8 dsfield = false;
943         struct flowi6 fl6;
944         int err = -EINVAL;
945         __be16 proto;
946         __u32 mtu;
947         int nhoff;
948         int thoff;
949
950         if (!pskb_inet_may_pull(skb))
951                 goto tx_err;
952
953         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
954                 goto tx_err;
955
956         if (gre_handle_offloads(skb, false))
957                 goto tx_err;
958
959         if (skb->len > dev->mtu + dev->hard_header_len) {
960                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
961                 truncate = true;
962         }
963
964         nhoff = skb_network_header(skb) - skb_mac_header(skb);
965         if (skb->protocol == htons(ETH_P_IP) &&
966             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
967                 truncate = true;
968
969         thoff = skb_transport_header(skb) - skb_mac_header(skb);
970         if (skb->protocol == htons(ETH_P_IPV6) &&
971             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
972                 truncate = true;
973
974         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
975                 goto tx_err;
976
977         t->parms.o_flags &= ~TUNNEL_KEY;
978         IPCB(skb)->flags = 0;
979
980         /* For collect_md mode, derive fl6 from the tunnel key,
981          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
982          */
983         if (t->parms.collect_md) {
984                 struct ip_tunnel_info *tun_info;
985                 const struct ip_tunnel_key *key;
986                 struct erspan_metadata *md;
987                 __be32 tun_id;
988
989                 tun_info = skb_tunnel_info(skb);
990                 if (unlikely(!tun_info ||
991                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
992                              ip_tunnel_info_af(tun_info) != AF_INET6))
993                         goto tx_err;
994
995                 key = &tun_info->key;
996                 memset(&fl6, 0, sizeof(fl6));
997                 fl6.flowi6_proto = IPPROTO_GRE;
998                 fl6.daddr = key->u.ipv6.dst;
999                 fl6.flowlabel = key->label;
1000                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1001
1002                 dsfield = key->tos;
1003                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1004                         goto tx_err;
1005                 if (tun_info->options_len < sizeof(*md))
1006                         goto tx_err;
1007                 md = ip_tunnel_info_opts(tun_info);
1008
1009                 tun_id = tunnel_id_to_key32(key->tun_id);
1010                 if (md->version == 1) {
1011                         erspan_build_header(skb,
1012                                             ntohl(tun_id),
1013                                             ntohl(md->u.index), truncate,
1014                                             false);
1015                 } else if (md->version == 2) {
1016                         erspan_build_header_v2(skb,
1017                                                ntohl(tun_id),
1018                                                md->u.md2.dir,
1019                                                get_hwid(&md->u.md2),
1020                                                truncate, false);
1021                 } else {
1022                         goto tx_err;
1023                 }
1024         } else {
1025                 switch (skb->protocol) {
1026                 case htons(ETH_P_IP):
1027                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1028                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1029                                                  &dsfield, &encap_limit);
1030                         break;
1031                 case htons(ETH_P_IPV6):
1032                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1033                                 goto tx_err;
1034                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1035                                                      &dsfield, &encap_limit))
1036                                 goto tx_err;
1037                         break;
1038                 default:
1039                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1040                         break;
1041                 }
1042
1043                 if (t->parms.erspan_ver == 1)
1044                         erspan_build_header(skb, ntohl(t->parms.o_key),
1045                                             t->parms.index,
1046                                             truncate, false);
1047                 else if (t->parms.erspan_ver == 2)
1048                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1049                                                t->parms.dir,
1050                                                t->parms.hwid,
1051                                                truncate, false);
1052                 else
1053                         goto tx_err;
1054
1055                 fl6.daddr = t->parms.raddr;
1056         }
1057
1058         /* Push GRE header. */
1059         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1060                                            : htons(ETH_P_ERSPAN2);
1061         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1062
1063         /* TooBig packet may have updated dst->dev's mtu */
1064         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1065                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1066
1067         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1068                            NEXTHDR_GRE);
1069         if (err != 0) {
1070                 /* XXX: send ICMP error even if DF is not set. */
1071                 if (err == -EMSGSIZE) {
1072                         if (skb->protocol == htons(ETH_P_IP))
1073                                 icmp_send(skb, ICMP_DEST_UNREACH,
1074                                           ICMP_FRAG_NEEDED, htonl(mtu));
1075                         else
1076                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1077                 }
1078
1079                 goto tx_err;
1080         }
1081         return NETDEV_TX_OK;
1082
1083 tx_err:
1084         stats = &t->dev->stats;
1085         stats->tx_errors++;
1086         stats->tx_dropped++;
1087         kfree_skb(skb);
1088         return NETDEV_TX_OK;
1089 }
1090
1091 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1092 {
1093         struct net_device *dev = t->dev;
1094         struct __ip6_tnl_parm *p = &t->parms;
1095         struct flowi6 *fl6 = &t->fl.u.ip6;
1096
1097         if (dev->type != ARPHRD_ETHER) {
1098                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1099                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1100         }
1101
1102         /* Set up flowi template */
1103         fl6->saddr = p->laddr;
1104         fl6->daddr = p->raddr;
1105         fl6->flowi6_oif = p->link;
1106         fl6->flowlabel = 0;
1107         fl6->flowi6_proto = IPPROTO_GRE;
1108
1109         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1110                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1111         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1112                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1113
1114         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1115         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1116
1117         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1118                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1119                 dev->flags |= IFF_POINTOPOINT;
1120         else
1121                 dev->flags &= ~IFF_POINTOPOINT;
1122 }
1123
1124 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1125                                          int t_hlen)
1126 {
1127         const struct __ip6_tnl_parm *p = &t->parms;
1128         struct net_device *dev = t->dev;
1129
1130         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1131                 int strict = (ipv6_addr_type(&p->raddr) &
1132                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1133
1134                 struct rt6_info *rt = rt6_lookup(t->net,
1135                                                  &p->raddr, &p->laddr,
1136                                                  p->link, NULL, strict);
1137
1138                 if (!rt)
1139                         return;
1140
1141                 if (rt->dst.dev) {
1142                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1143                                                  t_hlen;
1144
1145                         if (t->dev->header_ops)
1146                                 dev->hard_header_len = dst_len;
1147                         else
1148                                 dev->needed_headroom = dst_len;
1149
1150                         if (set_mtu) {
1151                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1152                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1153                                         dev->mtu -= 8;
1154                                 if (dev->type == ARPHRD_ETHER)
1155                                         dev->mtu -= ETH_HLEN;
1156
1157                                 if (dev->mtu < IPV6_MIN_MTU)
1158                                         dev->mtu = IPV6_MIN_MTU;
1159                         }
1160                 }
1161                 ip6_rt_put(rt);
1162         }
1163 }
1164
1165 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1166 {
1167         int t_hlen;
1168
1169         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1170         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1171
1172         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1173
1174         if (tunnel->dev->header_ops)
1175                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1176         else
1177                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1178
1179         return t_hlen;
1180 }
1181
1182 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1183 {
1184         ip6gre_tnl_link_config_common(t);
1185         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1186 }
1187
1188 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1189                                      const struct __ip6_tnl_parm *p)
1190 {
1191         t->parms.laddr = p->laddr;
1192         t->parms.raddr = p->raddr;
1193         t->parms.flags = p->flags;
1194         t->parms.hop_limit = p->hop_limit;
1195         t->parms.encap_limit = p->encap_limit;
1196         t->parms.flowinfo = p->flowinfo;
1197         t->parms.link = p->link;
1198         t->parms.proto = p->proto;
1199         t->parms.i_key = p->i_key;
1200         t->parms.o_key = p->o_key;
1201         t->parms.i_flags = p->i_flags;
1202         t->parms.o_flags = p->o_flags;
1203         t->parms.fwmark = p->fwmark;
1204         t->parms.erspan_ver = p->erspan_ver;
1205         t->parms.index = p->index;
1206         t->parms.dir = p->dir;
1207         t->parms.hwid = p->hwid;
1208         dst_cache_reset(&t->dst_cache);
1209 }
1210
1211 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1212                              int set_mtu)
1213 {
1214         ip6gre_tnl_copy_tnl_parm(t, p);
1215         ip6gre_tnl_link_config(t, set_mtu);
1216         return 0;
1217 }
1218
1219 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1220         const struct ip6_tnl_parm2 *u)
1221 {
1222         p->laddr = u->laddr;
1223         p->raddr = u->raddr;
1224         p->flags = u->flags;
1225         p->hop_limit = u->hop_limit;
1226         p->encap_limit = u->encap_limit;
1227         p->flowinfo = u->flowinfo;
1228         p->link = u->link;
1229         p->i_key = u->i_key;
1230         p->o_key = u->o_key;
1231         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1232         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1233         memcpy(p->name, u->name, sizeof(u->name));
1234 }
1235
1236 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1237         const struct __ip6_tnl_parm *p)
1238 {
1239         u->proto = IPPROTO_GRE;
1240         u->laddr = p->laddr;
1241         u->raddr = p->raddr;
1242         u->flags = p->flags;
1243         u->hop_limit = p->hop_limit;
1244         u->encap_limit = p->encap_limit;
1245         u->flowinfo = p->flowinfo;
1246         u->link = p->link;
1247         u->i_key = p->i_key;
1248         u->o_key = p->o_key;
1249         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1250         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1251         memcpy(u->name, p->name, sizeof(u->name));
1252 }
1253
1254 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1255         struct ifreq *ifr, int cmd)
1256 {
1257         int err = 0;
1258         struct ip6_tnl_parm2 p;
1259         struct __ip6_tnl_parm p1;
1260         struct ip6_tnl *t = netdev_priv(dev);
1261         struct net *net = t->net;
1262         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1263
1264         memset(&p1, 0, sizeof(p1));
1265
1266         switch (cmd) {
1267         case SIOCGETTUNNEL:
1268                 if (dev == ign->fb_tunnel_dev) {
1269                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1270                                 err = -EFAULT;
1271                                 break;
1272                         }
1273                         ip6gre_tnl_parm_from_user(&p1, &p);
1274                         t = ip6gre_tunnel_locate(net, &p1, 0);
1275                         if (!t)
1276                                 t = netdev_priv(dev);
1277                 }
1278                 memset(&p, 0, sizeof(p));
1279                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1280                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1281                         err = -EFAULT;
1282                 break;
1283
1284         case SIOCADDTUNNEL:
1285         case SIOCCHGTUNNEL:
1286                 err = -EPERM;
1287                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1288                         goto done;
1289
1290                 err = -EFAULT;
1291                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1292                         goto done;
1293
1294                 err = -EINVAL;
1295                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1296                         goto done;
1297
1298                 if (!(p.i_flags&GRE_KEY))
1299                         p.i_key = 0;
1300                 if (!(p.o_flags&GRE_KEY))
1301                         p.o_key = 0;
1302
1303                 ip6gre_tnl_parm_from_user(&p1, &p);
1304                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1305
1306                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1307                         if (t) {
1308                                 if (t->dev != dev) {
1309                                         err = -EEXIST;
1310                                         break;
1311                                 }
1312                         } else {
1313                                 t = netdev_priv(dev);
1314
1315                                 ip6gre_tunnel_unlink(ign, t);
1316                                 synchronize_net();
1317                                 ip6gre_tnl_change(t, &p1, 1);
1318                                 ip6gre_tunnel_link(ign, t);
1319                                 netdev_state_change(dev);
1320                         }
1321                 }
1322
1323                 if (t) {
1324                         err = 0;
1325
1326                         memset(&p, 0, sizeof(p));
1327                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1328                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1329                                 err = -EFAULT;
1330                 } else
1331                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1332                 break;
1333
1334         case SIOCDELTUNNEL:
1335                 err = -EPERM;
1336                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1337                         goto done;
1338
1339                 if (dev == ign->fb_tunnel_dev) {
1340                         err = -EFAULT;
1341                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1342                                 goto done;
1343                         err = -ENOENT;
1344                         ip6gre_tnl_parm_from_user(&p1, &p);
1345                         t = ip6gre_tunnel_locate(net, &p1, 0);
1346                         if (!t)
1347                                 goto done;
1348                         err = -EPERM;
1349                         if (t == netdev_priv(ign->fb_tunnel_dev))
1350                                 goto done;
1351                         dev = t->dev;
1352                 }
1353                 unregister_netdevice(dev);
1354                 err = 0;
1355                 break;
1356
1357         default:
1358                 err = -EINVAL;
1359         }
1360
1361 done:
1362         return err;
1363 }
1364
1365 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1366                          unsigned short type, const void *daddr,
1367                          const void *saddr, unsigned int len)
1368 {
1369         struct ip6_tnl *t = netdev_priv(dev);
1370         struct ipv6hdr *ipv6h;
1371         __be16 *p;
1372
1373         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1374         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1375                                                   t->fl.u.ip6.flowlabel,
1376                                                   true, &t->fl.u.ip6));
1377         ipv6h->hop_limit = t->parms.hop_limit;
1378         ipv6h->nexthdr = NEXTHDR_GRE;
1379         ipv6h->saddr = t->parms.laddr;
1380         ipv6h->daddr = t->parms.raddr;
1381
1382         p = (__be16 *)(ipv6h + 1);
1383         p[0] = t->parms.o_flags;
1384         p[1] = htons(type);
1385
1386         /*
1387          *      Set the source hardware address.
1388          */
1389
1390         if (saddr)
1391                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1392         if (daddr)
1393                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1394         if (!ipv6_addr_any(&ipv6h->daddr))
1395                 return t->hlen;
1396
1397         return -t->hlen;
1398 }
1399
1400 static const struct header_ops ip6gre_header_ops = {
1401         .create = ip6gre_header,
1402 };
1403
1404 static const struct net_device_ops ip6gre_netdev_ops = {
1405         .ndo_init               = ip6gre_tunnel_init,
1406         .ndo_uninit             = ip6gre_tunnel_uninit,
1407         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1408         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1409         .ndo_change_mtu         = ip6_tnl_change_mtu,
1410         .ndo_get_stats64        = ip_tunnel_get_stats64,
1411         .ndo_get_iflink         = ip6_tnl_get_iflink,
1412 };
1413
1414 static void ip6gre_dev_free(struct net_device *dev)
1415 {
1416         struct ip6_tnl *t = netdev_priv(dev);
1417
1418         gro_cells_destroy(&t->gro_cells);
1419         dst_cache_destroy(&t->dst_cache);
1420         free_percpu(dev->tstats);
1421 }
1422
1423 static void ip6gre_tunnel_setup(struct net_device *dev)
1424 {
1425         dev->netdev_ops = &ip6gre_netdev_ops;
1426         dev->needs_free_netdev = true;
1427         dev->priv_destructor = ip6gre_dev_free;
1428
1429         dev->type = ARPHRD_IP6GRE;
1430
1431         dev->flags |= IFF_NOARP;
1432         dev->addr_len = sizeof(struct in6_addr);
1433         netif_keep_dst(dev);
1434         /* This perm addr will be used as interface identifier by IPv6 */
1435         dev->addr_assign_type = NET_ADDR_RANDOM;
1436         eth_random_addr(dev->perm_addr);
1437 }
1438
1439 #define GRE6_FEATURES (NETIF_F_SG |             \
1440                        NETIF_F_FRAGLIST |       \
1441                        NETIF_F_HIGHDMA |        \
1442                        NETIF_F_HW_CSUM)
1443
1444 static void ip6gre_tnl_init_features(struct net_device *dev)
1445 {
1446         struct ip6_tnl *nt = netdev_priv(dev);
1447
1448         dev->features           |= GRE6_FEATURES;
1449         dev->hw_features        |= GRE6_FEATURES;
1450
1451         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1452                 /* TCP offload with GRE SEQ is not supported, nor
1453                  * can we support 2 levels of outer headers requiring
1454                  * an update.
1455                  */
1456                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1457                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1458                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1459                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1460                 }
1461
1462                 /* Can use a lockless transmit, unless we generate
1463                  * output sequences
1464                  */
1465                 dev->features |= NETIF_F_LLTX;
1466         }
1467 }
1468
1469 static int ip6gre_tunnel_init_common(struct net_device *dev)
1470 {
1471         struct ip6_tnl *tunnel;
1472         int ret;
1473         int t_hlen;
1474
1475         tunnel = netdev_priv(dev);
1476
1477         tunnel->dev = dev;
1478         tunnel->net = dev_net(dev);
1479         strcpy(tunnel->parms.name, dev->name);
1480
1481         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1482         if (!dev->tstats)
1483                 return -ENOMEM;
1484
1485         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1486         if (ret)
1487                 goto cleanup_alloc_pcpu_stats;
1488
1489         ret = gro_cells_init(&tunnel->gro_cells, dev);
1490         if (ret)
1491                 goto cleanup_dst_cache_init;
1492
1493         t_hlen = ip6gre_calc_hlen(tunnel);
1494         dev->mtu = ETH_DATA_LEN - t_hlen;
1495         if (dev->type == ARPHRD_ETHER)
1496                 dev->mtu -= ETH_HLEN;
1497         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1498                 dev->mtu -= 8;
1499
1500         if (tunnel->parms.collect_md) {
1501                 netif_keep_dst(dev);
1502         }
1503         ip6gre_tnl_init_features(dev);
1504
1505         dev_hold(dev);
1506         return 0;
1507
1508 cleanup_dst_cache_init:
1509         dst_cache_destroy(&tunnel->dst_cache);
1510 cleanup_alloc_pcpu_stats:
1511         free_percpu(dev->tstats);
1512         dev->tstats = NULL;
1513         return ret;
1514 }
1515
1516 static int ip6gre_tunnel_init(struct net_device *dev)
1517 {
1518         struct ip6_tnl *tunnel;
1519         int ret;
1520
1521         ret = ip6gre_tunnel_init_common(dev);
1522         if (ret)
1523                 return ret;
1524
1525         tunnel = netdev_priv(dev);
1526
1527         if (tunnel->parms.collect_md)
1528                 return 0;
1529
1530         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1531         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1532
1533         if (ipv6_addr_any(&tunnel->parms.raddr))
1534                 dev->header_ops = &ip6gre_header_ops;
1535
1536         return 0;
1537 }
1538
1539 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1540 {
1541         struct ip6_tnl *tunnel = netdev_priv(dev);
1542
1543         tunnel->dev = dev;
1544         tunnel->net = dev_net(dev);
1545         strcpy(tunnel->parms.name, dev->name);
1546
1547         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1548 }
1549
1550 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1551         .handler     = gre_rcv,
1552         .err_handler = ip6gre_err,
1553         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1554 };
1555
1556 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1557 {
1558         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1559         struct net_device *dev, *aux;
1560         int prio;
1561
1562         for_each_netdev_safe(net, dev, aux)
1563                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1564                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1565                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1566                         unregister_netdevice_queue(dev, head);
1567
1568         for (prio = 0; prio < 4; prio++) {
1569                 int h;
1570                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1571                         struct ip6_tnl *t;
1572
1573                         t = rtnl_dereference(ign->tunnels[prio][h]);
1574
1575                         while (t) {
1576                                 /* If dev is in the same netns, it has already
1577                                  * been added to the list by the previous loop.
1578                                  */
1579                                 if (!net_eq(dev_net(t->dev), net))
1580                                         unregister_netdevice_queue(t->dev,
1581                                                                    head);
1582                                 t = rtnl_dereference(t->next);
1583                         }
1584                 }
1585         }
1586 }
1587
1588 static int __net_init ip6gre_init_net(struct net *net)
1589 {
1590         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1591         struct net_device *ndev;
1592         int err;
1593
1594         if (!net_has_fallback_tunnels(net))
1595                 return 0;
1596         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1597                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1598         if (!ndev) {
1599                 err = -ENOMEM;
1600                 goto err_alloc_dev;
1601         }
1602         ign->fb_tunnel_dev = ndev;
1603         dev_net_set(ign->fb_tunnel_dev, net);
1604         /* FB netdevice is special: we have one, and only one per netns.
1605          * Allowing to move it to another netns is clearly unsafe.
1606          */
1607         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1608
1609
1610         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1611         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1612
1613         err = register_netdev(ign->fb_tunnel_dev);
1614         if (err)
1615                 goto err_reg_dev;
1616
1617         rcu_assign_pointer(ign->tunnels_wc[0],
1618                            netdev_priv(ign->fb_tunnel_dev));
1619         return 0;
1620
1621 err_reg_dev:
1622         free_netdev(ndev);
1623 err_alloc_dev:
1624         return err;
1625 }
1626
1627 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1628 {
1629         struct net *net;
1630         LIST_HEAD(list);
1631
1632         rtnl_lock();
1633         list_for_each_entry(net, net_list, exit_list)
1634                 ip6gre_destroy_tunnels(net, &list);
1635         unregister_netdevice_many(&list);
1636         rtnl_unlock();
1637 }
1638
1639 static struct pernet_operations ip6gre_net_ops = {
1640         .init = ip6gre_init_net,
1641         .exit_batch = ip6gre_exit_batch_net,
1642         .id   = &ip6gre_net_id,
1643         .size = sizeof(struct ip6gre_net),
1644 };
1645
1646 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1647                                   struct netlink_ext_ack *extack)
1648 {
1649         __be16 flags;
1650
1651         if (!data)
1652                 return 0;
1653
1654         flags = 0;
1655         if (data[IFLA_GRE_IFLAGS])
1656                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1657         if (data[IFLA_GRE_OFLAGS])
1658                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1659         if (flags & (GRE_VERSION|GRE_ROUTING))
1660                 return -EINVAL;
1661
1662         return 0;
1663 }
1664
1665 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1666                                struct netlink_ext_ack *extack)
1667 {
1668         struct in6_addr daddr;
1669
1670         if (tb[IFLA_ADDRESS]) {
1671                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1672                         return -EINVAL;
1673                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1674                         return -EADDRNOTAVAIL;
1675         }
1676
1677         if (!data)
1678                 goto out;
1679
1680         if (data[IFLA_GRE_REMOTE]) {
1681                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1682                 if (ipv6_addr_any(&daddr))
1683                         return -EINVAL;
1684         }
1685
1686 out:
1687         return ip6gre_tunnel_validate(tb, data, extack);
1688 }
1689
1690 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1691                                   struct netlink_ext_ack *extack)
1692 {
1693         __be16 flags = 0;
1694         int ret, ver = 0;
1695
1696         if (!data)
1697                 return 0;
1698
1699         ret = ip6gre_tap_validate(tb, data, extack);
1700         if (ret)
1701                 return ret;
1702
1703         /* ERSPAN should only have GRE sequence and key flag */
1704         if (data[IFLA_GRE_OFLAGS])
1705                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1706         if (data[IFLA_GRE_IFLAGS])
1707                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1708         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1709             flags != (GRE_SEQ | GRE_KEY))
1710                 return -EINVAL;
1711
1712         /* ERSPAN Session ID only has 10-bit. Since we reuse
1713          * 32-bit key field as ID, check it's range.
1714          */
1715         if (data[IFLA_GRE_IKEY] &&
1716             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1717                 return -EINVAL;
1718
1719         if (data[IFLA_GRE_OKEY] &&
1720             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1721                 return -EINVAL;
1722
1723         if (data[IFLA_GRE_ERSPAN_VER]) {
1724                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1725                 if (ver != 1 && ver != 2)
1726                         return -EINVAL;
1727         }
1728
1729         if (ver == 1) {
1730                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1731                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1732
1733                         if (index & ~INDEX_MASK)
1734                                 return -EINVAL;
1735                 }
1736         } else if (ver == 2) {
1737                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1738                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1739
1740                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1741                                 return -EINVAL;
1742                 }
1743
1744                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1745                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1746
1747                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1748                                 return -EINVAL;
1749                 }
1750         }
1751
1752         return 0;
1753 }
1754
1755 static void ip6erspan_set_version(struct nlattr *data[],
1756                                   struct __ip6_tnl_parm *parms)
1757 {
1758         if (!data)
1759                 return;
1760
1761         parms->erspan_ver = 1;
1762         if (data[IFLA_GRE_ERSPAN_VER])
1763                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1764
1765         if (parms->erspan_ver == 1) {
1766                 if (data[IFLA_GRE_ERSPAN_INDEX])
1767                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1768         } else if (parms->erspan_ver == 2) {
1769                 if (data[IFLA_GRE_ERSPAN_DIR])
1770                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1771                 if (data[IFLA_GRE_ERSPAN_HWID])
1772                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1773         }
1774 }
1775
1776 static void ip6gre_netlink_parms(struct nlattr *data[],
1777                                 struct __ip6_tnl_parm *parms)
1778 {
1779         memset(parms, 0, sizeof(*parms));
1780
1781         if (!data)
1782                 return;
1783
1784         if (data[IFLA_GRE_LINK])
1785                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1786
1787         if (data[IFLA_GRE_IFLAGS])
1788                 parms->i_flags = gre_flags_to_tnl_flags(
1789                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1790
1791         if (data[IFLA_GRE_OFLAGS])
1792                 parms->o_flags = gre_flags_to_tnl_flags(
1793                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1794
1795         if (data[IFLA_GRE_IKEY])
1796                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1797
1798         if (data[IFLA_GRE_OKEY])
1799                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1800
1801         if (data[IFLA_GRE_LOCAL])
1802                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1803
1804         if (data[IFLA_GRE_REMOTE])
1805                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1806
1807         if (data[IFLA_GRE_TTL])
1808                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1809
1810         if (data[IFLA_GRE_ENCAP_LIMIT])
1811                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1812
1813         if (data[IFLA_GRE_FLOWINFO])
1814                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1815
1816         if (data[IFLA_GRE_FLAGS])
1817                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1818
1819         if (data[IFLA_GRE_FWMARK])
1820                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1821
1822         if (data[IFLA_GRE_COLLECT_METADATA])
1823                 parms->collect_md = true;
1824 }
1825
1826 static int ip6gre_tap_init(struct net_device *dev)
1827 {
1828         int ret;
1829
1830         ret = ip6gre_tunnel_init_common(dev);
1831         if (ret)
1832                 return ret;
1833
1834         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1835
1836         return 0;
1837 }
1838
1839 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1840         .ndo_init = ip6gre_tap_init,
1841         .ndo_uninit = ip6gre_tunnel_uninit,
1842         .ndo_start_xmit = ip6gre_tunnel_xmit,
1843         .ndo_set_mac_address = eth_mac_addr,
1844         .ndo_validate_addr = eth_validate_addr,
1845         .ndo_change_mtu = ip6_tnl_change_mtu,
1846         .ndo_get_stats64 = ip_tunnel_get_stats64,
1847         .ndo_get_iflink = ip6_tnl_get_iflink,
1848 };
1849
1850 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1851 {
1852         int t_hlen;
1853
1854         tunnel->tun_hlen = 8;
1855         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1856                        erspan_hdr_len(tunnel->parms.erspan_ver);
1857
1858         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1859         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1860         return t_hlen;
1861 }
1862
1863 static int ip6erspan_tap_init(struct net_device *dev)
1864 {
1865         struct ip6_tnl *tunnel;
1866         int t_hlen;
1867         int ret;
1868
1869         tunnel = netdev_priv(dev);
1870
1871         tunnel->dev = dev;
1872         tunnel->net = dev_net(dev);
1873         strcpy(tunnel->parms.name, dev->name);
1874
1875         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1876         if (!dev->tstats)
1877                 return -ENOMEM;
1878
1879         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1880         if (ret)
1881                 goto cleanup_alloc_pcpu_stats;
1882
1883         ret = gro_cells_init(&tunnel->gro_cells, dev);
1884         if (ret)
1885                 goto cleanup_dst_cache_init;
1886
1887         t_hlen = ip6erspan_calc_hlen(tunnel);
1888         dev->mtu = ETH_DATA_LEN - t_hlen;
1889         if (dev->type == ARPHRD_ETHER)
1890                 dev->mtu -= ETH_HLEN;
1891         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1892                 dev->mtu -= 8;
1893
1894         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1895         ip6erspan_tnl_link_config(tunnel, 1);
1896
1897         dev_hold(dev);
1898         return 0;
1899
1900 cleanup_dst_cache_init:
1901         dst_cache_destroy(&tunnel->dst_cache);
1902 cleanup_alloc_pcpu_stats:
1903         free_percpu(dev->tstats);
1904         dev->tstats = NULL;
1905         return ret;
1906 }
1907
1908 static const struct net_device_ops ip6erspan_netdev_ops = {
1909         .ndo_init =             ip6erspan_tap_init,
1910         .ndo_uninit =           ip6erspan_tunnel_uninit,
1911         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1912         .ndo_set_mac_address =  eth_mac_addr,
1913         .ndo_validate_addr =    eth_validate_addr,
1914         .ndo_change_mtu =       ip6_tnl_change_mtu,
1915         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1916         .ndo_get_iflink =       ip6_tnl_get_iflink,
1917 };
1918
1919 static void ip6gre_tap_setup(struct net_device *dev)
1920 {
1921
1922         ether_setup(dev);
1923
1924         dev->max_mtu = 0;
1925         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1926         dev->needs_free_netdev = true;
1927         dev->priv_destructor = ip6gre_dev_free;
1928
1929         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1930         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1931         netif_keep_dst(dev);
1932 }
1933
1934 bool is_ip6gretap_dev(const struct net_device *dev)
1935 {
1936         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1937 }
1938 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1939
1940 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1941                                        struct ip_tunnel_encap *ipencap)
1942 {
1943         bool ret = false;
1944
1945         memset(ipencap, 0, sizeof(*ipencap));
1946
1947         if (!data)
1948                 return ret;
1949
1950         if (data[IFLA_GRE_ENCAP_TYPE]) {
1951                 ret = true;
1952                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1953         }
1954
1955         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1956                 ret = true;
1957                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1958         }
1959
1960         if (data[IFLA_GRE_ENCAP_SPORT]) {
1961                 ret = true;
1962                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1963         }
1964
1965         if (data[IFLA_GRE_ENCAP_DPORT]) {
1966                 ret = true;
1967                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1968         }
1969
1970         return ret;
1971 }
1972
1973 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1974                                  struct nlattr *tb[], struct nlattr *data[],
1975                                  struct netlink_ext_ack *extack)
1976 {
1977         struct ip6_tnl *nt;
1978         struct ip_tunnel_encap ipencap;
1979         int err;
1980
1981         nt = netdev_priv(dev);
1982
1983         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1984                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1985
1986                 if (err < 0)
1987                         return err;
1988         }
1989
1990         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1991                 eth_hw_addr_random(dev);
1992
1993         nt->dev = dev;
1994         nt->net = dev_net(dev);
1995
1996         err = register_netdevice(dev);
1997         if (err)
1998                 goto out;
1999
2000         if (tb[IFLA_MTU])
2001                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2002
2003 out:
2004         return err;
2005 }
2006
2007 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2008                           struct nlattr *tb[], struct nlattr *data[],
2009                           struct netlink_ext_ack *extack)
2010 {
2011         struct ip6_tnl *nt = netdev_priv(dev);
2012         struct net *net = dev_net(dev);
2013         struct ip6gre_net *ign;
2014         int err;
2015
2016         ip6gre_netlink_parms(data, &nt->parms);
2017         ign = net_generic(net, ip6gre_net_id);
2018
2019         if (nt->parms.collect_md) {
2020                 if (rtnl_dereference(ign->collect_md_tun))
2021                         return -EEXIST;
2022         } else {
2023                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2024                         return -EEXIST;
2025         }
2026
2027         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2028         if (!err) {
2029                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2030                 ip6gre_tunnel_link_md(ign, nt);
2031                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2032         }
2033         return err;
2034 }
2035
2036 static struct ip6_tnl *
2037 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2038                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2039                          struct netlink_ext_ack *extack)
2040 {
2041         struct ip6_tnl *t, *nt = netdev_priv(dev);
2042         struct net *net = nt->net;
2043         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2044         struct ip_tunnel_encap ipencap;
2045
2046         if (dev == ign->fb_tunnel_dev)
2047                 return ERR_PTR(-EINVAL);
2048
2049         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2050                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2051
2052                 if (err < 0)
2053                         return ERR_PTR(err);
2054         }
2055
2056         ip6gre_netlink_parms(data, p_p);
2057
2058         t = ip6gre_tunnel_locate(net, p_p, 0);
2059
2060         if (t) {
2061                 if (t->dev != dev)
2062                         return ERR_PTR(-EEXIST);
2063         } else {
2064                 t = nt;
2065         }
2066
2067         return t;
2068 }
2069
2070 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2071                              struct nlattr *data[],
2072                              struct netlink_ext_ack *extack)
2073 {
2074         struct ip6_tnl *t = netdev_priv(dev);
2075         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2076         struct __ip6_tnl_parm p;
2077
2078         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2079         if (IS_ERR(t))
2080                 return PTR_ERR(t);
2081
2082         ip6gre_tunnel_unlink_md(ign, t);
2083         ip6gre_tunnel_unlink(ign, t);
2084         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2085         ip6gre_tunnel_link_md(ign, t);
2086         ip6gre_tunnel_link(ign, t);
2087         return 0;
2088 }
2089
2090 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2091 {
2092         struct net *net = dev_net(dev);
2093         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2094
2095         if (dev != ign->fb_tunnel_dev)
2096                 unregister_netdevice_queue(dev, head);
2097 }
2098
2099 static size_t ip6gre_get_size(const struct net_device *dev)
2100 {
2101         return
2102                 /* IFLA_GRE_LINK */
2103                 nla_total_size(4) +
2104                 /* IFLA_GRE_IFLAGS */
2105                 nla_total_size(2) +
2106                 /* IFLA_GRE_OFLAGS */
2107                 nla_total_size(2) +
2108                 /* IFLA_GRE_IKEY */
2109                 nla_total_size(4) +
2110                 /* IFLA_GRE_OKEY */
2111                 nla_total_size(4) +
2112                 /* IFLA_GRE_LOCAL */
2113                 nla_total_size(sizeof(struct in6_addr)) +
2114                 /* IFLA_GRE_REMOTE */
2115                 nla_total_size(sizeof(struct in6_addr)) +
2116                 /* IFLA_GRE_TTL */
2117                 nla_total_size(1) +
2118                 /* IFLA_GRE_ENCAP_LIMIT */
2119                 nla_total_size(1) +
2120                 /* IFLA_GRE_FLOWINFO */
2121                 nla_total_size(4) +
2122                 /* IFLA_GRE_FLAGS */
2123                 nla_total_size(4) +
2124                 /* IFLA_GRE_ENCAP_TYPE */
2125                 nla_total_size(2) +
2126                 /* IFLA_GRE_ENCAP_FLAGS */
2127                 nla_total_size(2) +
2128                 /* IFLA_GRE_ENCAP_SPORT */
2129                 nla_total_size(2) +
2130                 /* IFLA_GRE_ENCAP_DPORT */
2131                 nla_total_size(2) +
2132                 /* IFLA_GRE_COLLECT_METADATA */
2133                 nla_total_size(0) +
2134                 /* IFLA_GRE_FWMARK */
2135                 nla_total_size(4) +
2136                 /* IFLA_GRE_ERSPAN_INDEX */
2137                 nla_total_size(4) +
2138                 0;
2139 }
2140
2141 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2142 {
2143         struct ip6_tnl *t = netdev_priv(dev);
2144         struct __ip6_tnl_parm *p = &t->parms;
2145         __be16 o_flags = p->o_flags;
2146
2147         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2148                 if (!p->collect_md)
2149                         o_flags |= TUNNEL_KEY;
2150
2151                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2152                         goto nla_put_failure;
2153
2154                 if (p->erspan_ver == 1) {
2155                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2156                                 goto nla_put_failure;
2157                 } else {
2158                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2159                                 goto nla_put_failure;
2160                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2161                                 goto nla_put_failure;
2162                 }
2163         }
2164
2165         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2166             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2167                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2168             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2169                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2170             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2171             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2172             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2173             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2174             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2175             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2176             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2177             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2178             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2179                 goto nla_put_failure;
2180
2181         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2182                         t->encap.type) ||
2183             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2184                          t->encap.sport) ||
2185             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2186                          t->encap.dport) ||
2187             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2188                         t->encap.flags))
2189                 goto nla_put_failure;
2190
2191         if (p->collect_md) {
2192                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2193                         goto nla_put_failure;
2194         }
2195
2196         return 0;
2197
2198 nla_put_failure:
2199         return -EMSGSIZE;
2200 }
2201
2202 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2203         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2204         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2205         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2206         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2207         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2208         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2209         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2210         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2211         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2212         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2213         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2214         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2215         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2216         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2217         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2218         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2219         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2220         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2221         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2222         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2223         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2224 };
2225
2226 static void ip6erspan_tap_setup(struct net_device *dev)
2227 {
2228         ether_setup(dev);
2229
2230         dev->max_mtu = 0;
2231         dev->netdev_ops = &ip6erspan_netdev_ops;
2232         dev->needs_free_netdev = true;
2233         dev->priv_destructor = ip6gre_dev_free;
2234
2235         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2236         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2237         netif_keep_dst(dev);
2238 }
2239
2240 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2241                              struct nlattr *tb[], struct nlattr *data[],
2242                              struct netlink_ext_ack *extack)
2243 {
2244         struct ip6_tnl *nt = netdev_priv(dev);
2245         struct net *net = dev_net(dev);
2246         struct ip6gre_net *ign;
2247         int err;
2248
2249         ip6gre_netlink_parms(data, &nt->parms);
2250         ip6erspan_set_version(data, &nt->parms);
2251         ign = net_generic(net, ip6gre_net_id);
2252
2253         if (nt->parms.collect_md) {
2254                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2255                         return -EEXIST;
2256         } else {
2257                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2258                         return -EEXIST;
2259         }
2260
2261         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2262         if (!err) {
2263                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2264                 ip6erspan_tunnel_link_md(ign, nt);
2265                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2266         }
2267         return err;
2268 }
2269
2270 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2271 {
2272         ip6gre_tnl_link_config_common(t);
2273         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2274 }
2275
2276 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2277                                 const struct __ip6_tnl_parm *p, int set_mtu)
2278 {
2279         ip6gre_tnl_copy_tnl_parm(t, p);
2280         ip6erspan_tnl_link_config(t, set_mtu);
2281         return 0;
2282 }
2283
2284 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2285                                 struct nlattr *data[],
2286                                 struct netlink_ext_ack *extack)
2287 {
2288         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2289         struct __ip6_tnl_parm p;
2290         struct ip6_tnl *t;
2291
2292         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2293         if (IS_ERR(t))
2294                 return PTR_ERR(t);
2295
2296         ip6erspan_set_version(data, &p);
2297         ip6gre_tunnel_unlink_md(ign, t);
2298         ip6gre_tunnel_unlink(ign, t);
2299         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2300         ip6erspan_tunnel_link_md(ign, t);
2301         ip6gre_tunnel_link(ign, t);
2302         return 0;
2303 }
2304
2305 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2306         .kind           = "ip6gre",
2307         .maxtype        = IFLA_GRE_MAX,
2308         .policy         = ip6gre_policy,
2309         .priv_size      = sizeof(struct ip6_tnl),
2310         .setup          = ip6gre_tunnel_setup,
2311         .validate       = ip6gre_tunnel_validate,
2312         .newlink        = ip6gre_newlink,
2313         .changelink     = ip6gre_changelink,
2314         .dellink        = ip6gre_dellink,
2315         .get_size       = ip6gre_get_size,
2316         .fill_info      = ip6gre_fill_info,
2317         .get_link_net   = ip6_tnl_get_link_net,
2318 };
2319
2320 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2321         .kind           = "ip6gretap",
2322         .maxtype        = IFLA_GRE_MAX,
2323         .policy         = ip6gre_policy,
2324         .priv_size      = sizeof(struct ip6_tnl),
2325         .setup          = ip6gre_tap_setup,
2326         .validate       = ip6gre_tap_validate,
2327         .newlink        = ip6gre_newlink,
2328         .changelink     = ip6gre_changelink,
2329         .get_size       = ip6gre_get_size,
2330         .fill_info      = ip6gre_fill_info,
2331         .get_link_net   = ip6_tnl_get_link_net,
2332 };
2333
2334 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2335         .kind           = "ip6erspan",
2336         .maxtype        = IFLA_GRE_MAX,
2337         .policy         = ip6gre_policy,
2338         .priv_size      = sizeof(struct ip6_tnl),
2339         .setup          = ip6erspan_tap_setup,
2340         .validate       = ip6erspan_tap_validate,
2341         .newlink        = ip6erspan_newlink,
2342         .changelink     = ip6erspan_changelink,
2343         .get_size       = ip6gre_get_size,
2344         .fill_info      = ip6gre_fill_info,
2345         .get_link_net   = ip6_tnl_get_link_net,
2346 };
2347
2348 /*
2349  *      And now the modules code and kernel interface.
2350  */
2351
2352 static int __init ip6gre_init(void)
2353 {
2354         int err;
2355
2356         pr_info("GRE over IPv6 tunneling driver\n");
2357
2358         err = register_pernet_device(&ip6gre_net_ops);
2359         if (err < 0)
2360                 return err;
2361
2362         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2363         if (err < 0) {
2364                 pr_info("%s: can't add protocol\n", __func__);
2365                 goto add_proto_failed;
2366         }
2367
2368         err = rtnl_link_register(&ip6gre_link_ops);
2369         if (err < 0)
2370                 goto rtnl_link_failed;
2371
2372         err = rtnl_link_register(&ip6gre_tap_ops);
2373         if (err < 0)
2374                 goto tap_ops_failed;
2375
2376         err = rtnl_link_register(&ip6erspan_tap_ops);
2377         if (err < 0)
2378                 goto erspan_link_failed;
2379
2380 out:
2381         return err;
2382
2383 erspan_link_failed:
2384         rtnl_link_unregister(&ip6gre_tap_ops);
2385 tap_ops_failed:
2386         rtnl_link_unregister(&ip6gre_link_ops);
2387 rtnl_link_failed:
2388         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2389 add_proto_failed:
2390         unregister_pernet_device(&ip6gre_net_ops);
2391         goto out;
2392 }
2393
2394 static void __exit ip6gre_fini(void)
2395 {
2396         rtnl_link_unregister(&ip6gre_tap_ops);
2397         rtnl_link_unregister(&ip6gre_link_ops);
2398         rtnl_link_unregister(&ip6erspan_tap_ops);
2399         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2400         unregister_pernet_device(&ip6gre_net_ops);
2401 }
2402
2403 module_init(ip6gre_init);
2404 module_exit(ip6gre_fini);
2405 MODULE_LICENSE("GPL");
2406 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2407 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2408 MODULE_ALIAS_RTNL_LINK("ip6gre");
2409 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2410 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2411 MODULE_ALIAS_NETDEV("ip6gre0");