GNU Linux-libre 5.4.207-gnu1
[releases.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         ip6gre_tunnel_link(ign, nt);
391         return nt;
392
393 failed_free:
394         free_netdev(dev);
395         return NULL;
396 }
397
398 static void ip6erspan_tunnel_uninit(struct net_device *dev)
399 {
400         struct ip6_tnl *t = netdev_priv(dev);
401         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
402
403         ip6erspan_tunnel_unlink_md(ign, t);
404         ip6gre_tunnel_unlink(ign, t);
405         dst_cache_reset(&t->dst_cache);
406         dev_put(dev);
407 }
408
409 static void ip6gre_tunnel_uninit(struct net_device *dev)
410 {
411         struct ip6_tnl *t = netdev_priv(dev);
412         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
413
414         ip6gre_tunnel_unlink_md(ign, t);
415         ip6gre_tunnel_unlink(ign, t);
416         if (ign->fb_tunnel_dev == dev)
417                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
418         dst_cache_reset(&t->dst_cache);
419         dev_put(dev);
420 }
421
422
423 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
424                        u8 type, u8 code, int offset, __be32 info)
425 {
426         struct net *net = dev_net(skb->dev);
427         const struct ipv6hdr *ipv6h;
428         struct tnl_ptk_info tpi;
429         struct ip6_tnl *t;
430
431         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
432                              offset) < 0)
433                 return -EINVAL;
434
435         ipv6h = (const struct ipv6hdr *)skb->data;
436         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
437                                  tpi.key, tpi.proto);
438         if (!t)
439                 return -ENOENT;
440
441         switch (type) {
442                 struct ipv6_tlv_tnl_enc_lim *tel;
443                 __u32 teli;
444         case ICMPV6_DEST_UNREACH:
445                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
446                                     t->parms.name);
447                 if (code != ICMPV6_PORT_UNREACH)
448                         break;
449                 return 0;
450         case ICMPV6_TIME_EXCEED:
451                 if (code == ICMPV6_EXC_HOPLIMIT) {
452                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
453                                             t->parms.name);
454                         break;
455                 }
456                 return 0;
457         case ICMPV6_PARAMPROB:
458                 teli = 0;
459                 if (code == ICMPV6_HDR_FIELD)
460                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
461
462                 if (teli && teli == be32_to_cpu(info) - 2) {
463                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
464                         if (tel->encap_limit == 0) {
465                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
466                                                     t->parms.name);
467                         }
468                 } else {
469                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
470                                             t->parms.name);
471                 }
472                 return 0;
473         case ICMPV6_PKT_TOOBIG:
474                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
475                 return 0;
476         case NDISC_REDIRECT:
477                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
478                              sock_net_uid(net, NULL));
479                 return 0;
480         }
481
482         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
483                 t->err_count++;
484         else
485                 t->err_count = 1;
486         t->err_time = jiffies;
487
488         return 0;
489 }
490
491 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
492 {
493         const struct ipv6hdr *ipv6h;
494         struct ip6_tnl *tunnel;
495
496         ipv6h = ipv6_hdr(skb);
497         tunnel = ip6gre_tunnel_lookup(skb->dev,
498                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
499                                       tpi->proto);
500         if (tunnel) {
501                 if (tunnel->parms.collect_md) {
502                         struct metadata_dst *tun_dst;
503                         __be64 tun_id;
504                         __be16 flags;
505
506                         flags = tpi->flags;
507                         tun_id = key32_to_tunnel_id(tpi->key);
508
509                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
510                         if (!tun_dst)
511                                 return PACKET_REJECT;
512
513                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
514                 } else {
515                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
516                 }
517
518                 return PACKET_RCVD;
519         }
520
521         return PACKET_REJECT;
522 }
523
524 static int ip6erspan_rcv(struct sk_buff *skb,
525                          struct tnl_ptk_info *tpi,
526                          int gre_hdr_len)
527 {
528         struct erspan_base_hdr *ershdr;
529         const struct ipv6hdr *ipv6h;
530         struct erspan_md2 *md2;
531         struct ip6_tnl *tunnel;
532         u8 ver;
533
534         ipv6h = ipv6_hdr(skb);
535         ershdr = (struct erspan_base_hdr *)skb->data;
536         ver = ershdr->ver;
537
538         tunnel = ip6gre_tunnel_lookup(skb->dev,
539                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
540                                       tpi->proto);
541         if (tunnel) {
542                 int len = erspan_hdr_len(ver);
543
544                 if (unlikely(!pskb_may_pull(skb, len)))
545                         return PACKET_REJECT;
546
547                 if (__iptunnel_pull_header(skb, len,
548                                            htons(ETH_P_TEB),
549                                            false, false) < 0)
550                         return PACKET_REJECT;
551
552                 if (tunnel->parms.collect_md) {
553                         struct erspan_metadata *pkt_md, *md;
554                         struct metadata_dst *tun_dst;
555                         struct ip_tunnel_info *info;
556                         unsigned char *gh;
557                         __be64 tun_id;
558                         __be16 flags;
559
560                         tpi->flags |= TUNNEL_KEY;
561                         flags = tpi->flags;
562                         tun_id = key32_to_tunnel_id(tpi->key);
563
564                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
565                                                   sizeof(*md));
566                         if (!tun_dst)
567                                 return PACKET_REJECT;
568
569                         /* skb can be uncloned in __iptunnel_pull_header, so
570                          * old pkt_md is no longer valid and we need to reset
571                          * it
572                          */
573                         gh = skb_network_header(skb) +
574                              skb_network_header_len(skb);
575                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
576                                                             sizeof(*ershdr));
577                         info = &tun_dst->u.tun_info;
578                         md = ip_tunnel_info_opts(info);
579                         md->version = ver;
580                         md2 = &md->u.md2;
581                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
582                                                        ERSPAN_V2_MDSIZE);
583                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
584                         info->options_len = sizeof(*md);
585
586                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
587
588                 } else {
589                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
590                 }
591
592                 return PACKET_RCVD;
593         }
594
595         return PACKET_REJECT;
596 }
597
598 static int gre_rcv(struct sk_buff *skb)
599 {
600         struct tnl_ptk_info tpi;
601         bool csum_err = false;
602         int hdr_len;
603
604         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
605         if (hdr_len < 0)
606                 goto drop;
607
608         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
609                 goto drop;
610
611         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
612                      tpi.proto == htons(ETH_P_ERSPAN2))) {
613                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
614                         return 0;
615                 goto out;
616         }
617
618         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
619                 return 0;
620
621 out:
622         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
623 drop:
624         kfree_skb(skb);
625         return 0;
626 }
627
628 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
629 {
630         return iptunnel_handle_offloads(skb,
631                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
632 }
633
634 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
635                                      struct net_device *dev,
636                                      struct flowi6 *fl6, __u8 *dsfield,
637                                      int *encap_limit)
638 {
639         const struct iphdr *iph = ip_hdr(skb);
640         struct ip6_tnl *t = netdev_priv(dev);
641
642         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
643                 *encap_limit = t->parms.encap_limit;
644
645         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
646
647         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
648                 *dsfield = ipv4_get_dsfield(iph);
649         else
650                 *dsfield = ip6_tclass(t->parms.flowinfo);
651
652         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
653                 fl6->flowi6_mark = skb->mark;
654         else
655                 fl6->flowi6_mark = t->parms.fwmark;
656
657         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
658 }
659
660 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
661                                     struct net_device *dev,
662                                     struct flowi6 *fl6, __u8 *dsfield,
663                                     int *encap_limit)
664 {
665         struct ipv6hdr *ipv6h;
666         struct ip6_tnl *t = netdev_priv(dev);
667         __u16 offset;
668
669         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
670         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
671         ipv6h = ipv6_hdr(skb);
672
673         if (offset > 0) {
674                 struct ipv6_tlv_tnl_enc_lim *tel;
675
676                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
677                 if (tel->encap_limit == 0) {
678                         icmpv6_send(skb, ICMPV6_PARAMPROB,
679                                     ICMPV6_HDR_FIELD, offset + 2);
680                         return -1;
681                 }
682                 *encap_limit = tel->encap_limit - 1;
683         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
684                 *encap_limit = t->parms.encap_limit;
685         }
686
687         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
688
689         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
690                 *dsfield = ipv6_get_dsfield(ipv6h);
691         else
692                 *dsfield = ip6_tclass(t->parms.flowinfo);
693
694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
695                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
696
697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
698                 fl6->flowi6_mark = skb->mark;
699         else
700                 fl6->flowi6_mark = t->parms.fwmark;
701
702         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
703
704         return 0;
705 }
706
707 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
708                                struct net_device *dev, __u8 dsfield,
709                                struct flowi6 *fl6, int encap_limit,
710                                __u32 *pmtu, __be16 proto)
711 {
712         struct ip6_tnl *tunnel = netdev_priv(dev);
713         __be16 protocol;
714
715         if (dev->type == ARPHRD_ETHER)
716                 IPCB(skb)->flags = 0;
717
718         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
719                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
720         else
721                 fl6->daddr = tunnel->parms.raddr;
722
723         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
724                 return -ENOMEM;
725
726         /* Push GRE header. */
727         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
728
729         if (tunnel->parms.collect_md) {
730                 struct ip_tunnel_info *tun_info;
731                 const struct ip_tunnel_key *key;
732                 __be16 flags;
733                 int tun_hlen;
734
735                 tun_info = skb_tunnel_info(skb);
736                 if (unlikely(!tun_info ||
737                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
738                              ip_tunnel_info_af(tun_info) != AF_INET6))
739                         return -EINVAL;
740
741                 key = &tun_info->key;
742                 memset(fl6, 0, sizeof(*fl6));
743                 fl6->flowi6_proto = IPPROTO_GRE;
744                 fl6->daddr = key->u.ipv6.dst;
745                 fl6->flowlabel = key->label;
746                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
747                 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
748
749                 dsfield = key->tos;
750                 flags = key->tun_flags &
751                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
752                 tun_hlen = gre_calc_hlen(flags);
753
754                 gre_build_header(skb, tun_hlen,
755                                  flags, protocol,
756                                  tunnel_id_to_key32(tun_info->key.tun_id),
757                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
758                                                       : 0);
759
760         } else {
761                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
762                         tunnel->o_seqno++;
763
764                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
765                                  protocol, tunnel->parms.o_key,
766                                  htonl(tunnel->o_seqno));
767         }
768
769         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
770                             NEXTHDR_GRE);
771 }
772
773 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
774 {
775         struct ip6_tnl *t = netdev_priv(dev);
776         int encap_limit = -1;
777         struct flowi6 fl6;
778         __u8 dsfield = 0;
779         __u32 mtu;
780         int err;
781
782         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
783
784         if (!t->parms.collect_md)
785                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
786                                          &dsfield, &encap_limit);
787
788         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
789         if (err)
790                 return -1;
791
792         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
793                           skb->protocol);
794         if (err != 0) {
795                 /* XXX: send ICMP error even if DF is not set. */
796                 if (err == -EMSGSIZE)
797                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
798                                   htonl(mtu));
799                 return -1;
800         }
801
802         return 0;
803 }
804
805 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
806 {
807         struct ip6_tnl *t = netdev_priv(dev);
808         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
809         int encap_limit = -1;
810         struct flowi6 fl6;
811         __u8 dsfield = 0;
812         __u32 mtu;
813         int err;
814
815         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
816                 return -1;
817
818         if (!t->parms.collect_md &&
819             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
820                 return -1;
821
822         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
823                 return -1;
824
825         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
826                           &mtu, skb->protocol);
827         if (err != 0) {
828                 if (err == -EMSGSIZE)
829                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
830                 return -1;
831         }
832
833         return 0;
834 }
835
836 /**
837  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
838  *   @t: the outgoing tunnel device
839  *   @hdr: IPv6 header from the incoming packet
840  *
841  * Description:
842  *   Avoid trivial tunneling loop by checking that tunnel exit-point
843  *   doesn't match source of incoming packet.
844  *
845  * Return:
846  *   1 if conflict,
847  *   0 else
848  **/
849
850 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
851         const struct ipv6hdr *hdr)
852 {
853         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
854 }
855
856 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
857 {
858         struct ip6_tnl *t = netdev_priv(dev);
859         int encap_limit = -1;
860         struct flowi6 fl6;
861         __u32 mtu;
862         int err;
863
864         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
865                 encap_limit = t->parms.encap_limit;
866
867         if (!t->parms.collect_md)
868                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
869
870         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
871         if (err)
872                 return err;
873
874         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
875
876         return err;
877 }
878
879 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
880         struct net_device *dev)
881 {
882         struct ip6_tnl *t = netdev_priv(dev);
883         struct net_device_stats *stats = &t->dev->stats;
884         int ret;
885
886         if (!pskb_inet_may_pull(skb))
887                 goto tx_err;
888
889         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
890                 goto tx_err;
891
892         switch (skb->protocol) {
893         case htons(ETH_P_IP):
894                 ret = ip6gre_xmit_ipv4(skb, dev);
895                 break;
896         case htons(ETH_P_IPV6):
897                 ret = ip6gre_xmit_ipv6(skb, dev);
898                 break;
899         default:
900                 ret = ip6gre_xmit_other(skb, dev);
901                 break;
902         }
903
904         if (ret < 0)
905                 goto tx_err;
906
907         return NETDEV_TX_OK;
908
909 tx_err:
910         stats->tx_errors++;
911         stats->tx_dropped++;
912         kfree_skb(skb);
913         return NETDEV_TX_OK;
914 }
915
916 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
917                                          struct net_device *dev)
918 {
919         struct ip6_tnl *t = netdev_priv(dev);
920         struct dst_entry *dst = skb_dst(skb);
921         struct net_device_stats *stats;
922         bool truncate = false;
923         int encap_limit = -1;
924         __u8 dsfield = false;
925         struct flowi6 fl6;
926         int err = -EINVAL;
927         __be16 proto;
928         __u32 mtu;
929         int nhoff;
930
931         if (!pskb_inet_may_pull(skb))
932                 goto tx_err;
933
934         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
935                 goto tx_err;
936
937         if (gre_handle_offloads(skb, false))
938                 goto tx_err;
939
940         if (skb->len > dev->mtu + dev->hard_header_len) {
941                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
942                 truncate = true;
943         }
944
945         nhoff = skb_network_header(skb) - skb_mac_header(skb);
946         if (skb->protocol == htons(ETH_P_IP) &&
947             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
948                 truncate = true;
949
950         if (skb->protocol == htons(ETH_P_IPV6)) {
951                 int thoff;
952
953                 if (skb_transport_header_was_set(skb))
954                         thoff = skb_transport_header(skb) - skb_mac_header(skb);
955                 else
956                         thoff = nhoff + sizeof(struct ipv6hdr);
957                 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
958                         truncate = true;
959         }
960
961         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
962                 goto tx_err;
963
964         t->parms.o_flags &= ~TUNNEL_KEY;
965         IPCB(skb)->flags = 0;
966
967         /* For collect_md mode, derive fl6 from the tunnel key,
968          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
969          */
970         if (t->parms.collect_md) {
971                 struct ip_tunnel_info *tun_info;
972                 const struct ip_tunnel_key *key;
973                 struct erspan_metadata *md;
974                 __be32 tun_id;
975
976                 tun_info = skb_tunnel_info(skb);
977                 if (unlikely(!tun_info ||
978                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
979                              ip_tunnel_info_af(tun_info) != AF_INET6))
980                         goto tx_err;
981
982                 key = &tun_info->key;
983                 memset(&fl6, 0, sizeof(fl6));
984                 fl6.flowi6_proto = IPPROTO_GRE;
985                 fl6.daddr = key->u.ipv6.dst;
986                 fl6.flowlabel = key->label;
987                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
988                 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
989
990                 dsfield = key->tos;
991                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
992                         goto tx_err;
993                 if (tun_info->options_len < sizeof(*md))
994                         goto tx_err;
995                 md = ip_tunnel_info_opts(tun_info);
996
997                 tun_id = tunnel_id_to_key32(key->tun_id);
998                 if (md->version == 1) {
999                         erspan_build_header(skb,
1000                                             ntohl(tun_id),
1001                                             ntohl(md->u.index), truncate,
1002                                             false);
1003                 } else if (md->version == 2) {
1004                         erspan_build_header_v2(skb,
1005                                                ntohl(tun_id),
1006                                                md->u.md2.dir,
1007                                                get_hwid(&md->u.md2),
1008                                                truncate, false);
1009                 } else {
1010                         goto tx_err;
1011                 }
1012         } else {
1013                 switch (skb->protocol) {
1014                 case htons(ETH_P_IP):
1015                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1016                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1017                                                  &dsfield, &encap_limit);
1018                         break;
1019                 case htons(ETH_P_IPV6):
1020                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1021                                 goto tx_err;
1022                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1023                                                      &dsfield, &encap_limit))
1024                                 goto tx_err;
1025                         break;
1026                 default:
1027                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1028                         break;
1029                 }
1030
1031                 if (t->parms.erspan_ver == 1)
1032                         erspan_build_header(skb, ntohl(t->parms.o_key),
1033                                             t->parms.index,
1034                                             truncate, false);
1035                 else if (t->parms.erspan_ver == 2)
1036                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1037                                                t->parms.dir,
1038                                                t->parms.hwid,
1039                                                truncate, false);
1040                 else
1041                         goto tx_err;
1042
1043                 fl6.daddr = t->parms.raddr;
1044         }
1045
1046         /* Push GRE header. */
1047         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1048                                            : htons(ETH_P_ERSPAN2);
1049         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1050
1051         /* TooBig packet may have updated dst->dev's mtu */
1052         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1053                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1054
1055         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1056                            NEXTHDR_GRE);
1057         if (err != 0) {
1058                 /* XXX: send ICMP error even if DF is not set. */
1059                 if (err == -EMSGSIZE) {
1060                         if (skb->protocol == htons(ETH_P_IP))
1061                                 icmp_send(skb, ICMP_DEST_UNREACH,
1062                                           ICMP_FRAG_NEEDED, htonl(mtu));
1063                         else
1064                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1065                 }
1066
1067                 goto tx_err;
1068         }
1069         return NETDEV_TX_OK;
1070
1071 tx_err:
1072         stats = &t->dev->stats;
1073         stats->tx_errors++;
1074         stats->tx_dropped++;
1075         kfree_skb(skb);
1076         return NETDEV_TX_OK;
1077 }
1078
1079 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1080 {
1081         struct net_device *dev = t->dev;
1082         struct __ip6_tnl_parm *p = &t->parms;
1083         struct flowi6 *fl6 = &t->fl.u.ip6;
1084
1085         if (dev->type != ARPHRD_ETHER) {
1086                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1087                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1088         }
1089
1090         /* Set up flowi template */
1091         fl6->saddr = p->laddr;
1092         fl6->daddr = p->raddr;
1093         fl6->flowi6_oif = p->link;
1094         fl6->flowlabel = 0;
1095         fl6->flowi6_proto = IPPROTO_GRE;
1096         fl6->fl6_gre_key = t->parms.o_key;
1097
1098         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1099                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1100         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1101                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1102
1103         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1104         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1105
1106         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1107                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1108                 dev->flags |= IFF_POINTOPOINT;
1109         else
1110                 dev->flags &= ~IFF_POINTOPOINT;
1111 }
1112
1113 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1114                                          int t_hlen)
1115 {
1116         const struct __ip6_tnl_parm *p = &t->parms;
1117         struct net_device *dev = t->dev;
1118
1119         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1120                 int strict = (ipv6_addr_type(&p->raddr) &
1121                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1122
1123                 struct rt6_info *rt = rt6_lookup(t->net,
1124                                                  &p->raddr, &p->laddr,
1125                                                  p->link, NULL, strict);
1126
1127                 if (!rt)
1128                         return;
1129
1130                 if (rt->dst.dev) {
1131                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1132                                                  t_hlen;
1133
1134                         if (t->dev->header_ops)
1135                                 dev->hard_header_len = dst_len;
1136                         else
1137                                 dev->needed_headroom = dst_len;
1138
1139                         if (set_mtu) {
1140                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1141                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1142                                         dev->mtu -= 8;
1143                                 if (dev->type == ARPHRD_ETHER)
1144                                         dev->mtu -= ETH_HLEN;
1145
1146                                 if (dev->mtu < IPV6_MIN_MTU)
1147                                         dev->mtu = IPV6_MIN_MTU;
1148                         }
1149                 }
1150                 ip6_rt_put(rt);
1151         }
1152 }
1153
1154 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1155 {
1156         int t_hlen;
1157
1158         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1159         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1160
1161         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1162
1163         if (tunnel->dev->header_ops)
1164                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1165         else
1166                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1167
1168         return t_hlen;
1169 }
1170
1171 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1172 {
1173         ip6gre_tnl_link_config_common(t);
1174         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1175 }
1176
1177 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1178                                      const struct __ip6_tnl_parm *p)
1179 {
1180         t->parms.laddr = p->laddr;
1181         t->parms.raddr = p->raddr;
1182         t->parms.flags = p->flags;
1183         t->parms.hop_limit = p->hop_limit;
1184         t->parms.encap_limit = p->encap_limit;
1185         t->parms.flowinfo = p->flowinfo;
1186         t->parms.link = p->link;
1187         t->parms.proto = p->proto;
1188         t->parms.i_key = p->i_key;
1189         t->parms.o_key = p->o_key;
1190         t->parms.i_flags = p->i_flags;
1191         t->parms.o_flags = p->o_flags;
1192         t->parms.fwmark = p->fwmark;
1193         t->parms.erspan_ver = p->erspan_ver;
1194         t->parms.index = p->index;
1195         t->parms.dir = p->dir;
1196         t->parms.hwid = p->hwid;
1197         dst_cache_reset(&t->dst_cache);
1198 }
1199
1200 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1201                              int set_mtu)
1202 {
1203         ip6gre_tnl_copy_tnl_parm(t, p);
1204         ip6gre_tnl_link_config(t, set_mtu);
1205         return 0;
1206 }
1207
1208 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1209         const struct ip6_tnl_parm2 *u)
1210 {
1211         p->laddr = u->laddr;
1212         p->raddr = u->raddr;
1213         p->flags = u->flags;
1214         p->hop_limit = u->hop_limit;
1215         p->encap_limit = u->encap_limit;
1216         p->flowinfo = u->flowinfo;
1217         p->link = u->link;
1218         p->i_key = u->i_key;
1219         p->o_key = u->o_key;
1220         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1221         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1222         memcpy(p->name, u->name, sizeof(u->name));
1223 }
1224
1225 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1226         const struct __ip6_tnl_parm *p)
1227 {
1228         u->proto = IPPROTO_GRE;
1229         u->laddr = p->laddr;
1230         u->raddr = p->raddr;
1231         u->flags = p->flags;
1232         u->hop_limit = p->hop_limit;
1233         u->encap_limit = p->encap_limit;
1234         u->flowinfo = p->flowinfo;
1235         u->link = p->link;
1236         u->i_key = p->i_key;
1237         u->o_key = p->o_key;
1238         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1239         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1240         memcpy(u->name, p->name, sizeof(u->name));
1241 }
1242
1243 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1244         struct ifreq *ifr, int cmd)
1245 {
1246         int err = 0;
1247         struct ip6_tnl_parm2 p;
1248         struct __ip6_tnl_parm p1;
1249         struct ip6_tnl *t = netdev_priv(dev);
1250         struct net *net = t->net;
1251         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1252
1253         memset(&p1, 0, sizeof(p1));
1254
1255         switch (cmd) {
1256         case SIOCGETTUNNEL:
1257                 if (dev == ign->fb_tunnel_dev) {
1258                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1259                                 err = -EFAULT;
1260                                 break;
1261                         }
1262                         ip6gre_tnl_parm_from_user(&p1, &p);
1263                         t = ip6gre_tunnel_locate(net, &p1, 0);
1264                         if (!t)
1265                                 t = netdev_priv(dev);
1266                 }
1267                 memset(&p, 0, sizeof(p));
1268                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1269                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1270                         err = -EFAULT;
1271                 break;
1272
1273         case SIOCADDTUNNEL:
1274         case SIOCCHGTUNNEL:
1275                 err = -EPERM;
1276                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1277                         goto done;
1278
1279                 err = -EFAULT;
1280                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1281                         goto done;
1282
1283                 err = -EINVAL;
1284                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1285                         goto done;
1286
1287                 if (!(p.i_flags&GRE_KEY))
1288                         p.i_key = 0;
1289                 if (!(p.o_flags&GRE_KEY))
1290                         p.o_key = 0;
1291
1292                 ip6gre_tnl_parm_from_user(&p1, &p);
1293                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1294
1295                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1296                         if (t) {
1297                                 if (t->dev != dev) {
1298                                         err = -EEXIST;
1299                                         break;
1300                                 }
1301                         } else {
1302                                 t = netdev_priv(dev);
1303
1304                                 ip6gre_tunnel_unlink(ign, t);
1305                                 synchronize_net();
1306                                 ip6gre_tnl_change(t, &p1, 1);
1307                                 ip6gre_tunnel_link(ign, t);
1308                                 netdev_state_change(dev);
1309                         }
1310                 }
1311
1312                 if (t) {
1313                         err = 0;
1314
1315                         memset(&p, 0, sizeof(p));
1316                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1317                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1318                                 err = -EFAULT;
1319                 } else
1320                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1321                 break;
1322
1323         case SIOCDELTUNNEL:
1324                 err = -EPERM;
1325                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1326                         goto done;
1327
1328                 if (dev == ign->fb_tunnel_dev) {
1329                         err = -EFAULT;
1330                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1331                                 goto done;
1332                         err = -ENOENT;
1333                         ip6gre_tnl_parm_from_user(&p1, &p);
1334                         t = ip6gre_tunnel_locate(net, &p1, 0);
1335                         if (!t)
1336                                 goto done;
1337                         err = -EPERM;
1338                         if (t == netdev_priv(ign->fb_tunnel_dev))
1339                                 goto done;
1340                         dev = t->dev;
1341                 }
1342                 unregister_netdevice(dev);
1343                 err = 0;
1344                 break;
1345
1346         default:
1347                 err = -EINVAL;
1348         }
1349
1350 done:
1351         return err;
1352 }
1353
1354 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1355                          unsigned short type, const void *daddr,
1356                          const void *saddr, unsigned int len)
1357 {
1358         struct ip6_tnl *t = netdev_priv(dev);
1359         struct ipv6hdr *ipv6h;
1360         __be16 *p;
1361
1362         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1363         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1364                                                   t->fl.u.ip6.flowlabel,
1365                                                   true, &t->fl.u.ip6));
1366         ipv6h->hop_limit = t->parms.hop_limit;
1367         ipv6h->nexthdr = NEXTHDR_GRE;
1368         ipv6h->saddr = t->parms.laddr;
1369         ipv6h->daddr = t->parms.raddr;
1370
1371         p = (__be16 *)(ipv6h + 1);
1372         p[0] = t->parms.o_flags;
1373         p[1] = htons(type);
1374
1375         /*
1376          *      Set the source hardware address.
1377          */
1378
1379         if (saddr)
1380                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1381         if (daddr)
1382                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1383         if (!ipv6_addr_any(&ipv6h->daddr))
1384                 return t->hlen;
1385
1386         return -t->hlen;
1387 }
1388
1389 static const struct header_ops ip6gre_header_ops = {
1390         .create = ip6gre_header,
1391 };
1392
1393 static const struct net_device_ops ip6gre_netdev_ops = {
1394         .ndo_init               = ip6gre_tunnel_init,
1395         .ndo_uninit             = ip6gre_tunnel_uninit,
1396         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1397         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1398         .ndo_change_mtu         = ip6_tnl_change_mtu,
1399         .ndo_get_stats64        = ip_tunnel_get_stats64,
1400         .ndo_get_iflink         = ip6_tnl_get_iflink,
1401 };
1402
1403 static void ip6gre_dev_free(struct net_device *dev)
1404 {
1405         struct ip6_tnl *t = netdev_priv(dev);
1406
1407         gro_cells_destroy(&t->gro_cells);
1408         dst_cache_destroy(&t->dst_cache);
1409         free_percpu(dev->tstats);
1410 }
1411
1412 static void ip6gre_tunnel_setup(struct net_device *dev)
1413 {
1414         dev->netdev_ops = &ip6gre_netdev_ops;
1415         dev->needs_free_netdev = true;
1416         dev->priv_destructor = ip6gre_dev_free;
1417
1418         dev->type = ARPHRD_IP6GRE;
1419
1420         dev->flags |= IFF_NOARP;
1421         dev->addr_len = sizeof(struct in6_addr);
1422         netif_keep_dst(dev);
1423         /* This perm addr will be used as interface identifier by IPv6 */
1424         dev->addr_assign_type = NET_ADDR_RANDOM;
1425         eth_random_addr(dev->perm_addr);
1426 }
1427
1428 #define GRE6_FEATURES (NETIF_F_SG |             \
1429                        NETIF_F_FRAGLIST |       \
1430                        NETIF_F_HIGHDMA |        \
1431                        NETIF_F_HW_CSUM)
1432
1433 static void ip6gre_tnl_init_features(struct net_device *dev)
1434 {
1435         struct ip6_tnl *nt = netdev_priv(dev);
1436
1437         dev->features           |= GRE6_FEATURES;
1438         dev->hw_features        |= GRE6_FEATURES;
1439
1440         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1441                 /* TCP offload with GRE SEQ is not supported, nor
1442                  * can we support 2 levels of outer headers requiring
1443                  * an update.
1444                  */
1445                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1446                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1447                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1448                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1449                 }
1450
1451                 /* Can use a lockless transmit, unless we generate
1452                  * output sequences
1453                  */
1454                 dev->features |= NETIF_F_LLTX;
1455         }
1456 }
1457
1458 static int ip6gre_tunnel_init_common(struct net_device *dev)
1459 {
1460         struct ip6_tnl *tunnel;
1461         int ret;
1462         int t_hlen;
1463
1464         tunnel = netdev_priv(dev);
1465
1466         tunnel->dev = dev;
1467         tunnel->net = dev_net(dev);
1468         strcpy(tunnel->parms.name, dev->name);
1469
1470         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1471         if (!dev->tstats)
1472                 return -ENOMEM;
1473
1474         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1475         if (ret)
1476                 goto cleanup_alloc_pcpu_stats;
1477
1478         ret = gro_cells_init(&tunnel->gro_cells, dev);
1479         if (ret)
1480                 goto cleanup_dst_cache_init;
1481
1482         t_hlen = ip6gre_calc_hlen(tunnel);
1483         dev->mtu = ETH_DATA_LEN - t_hlen;
1484         if (dev->type == ARPHRD_ETHER)
1485                 dev->mtu -= ETH_HLEN;
1486         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1487                 dev->mtu -= 8;
1488
1489         if (tunnel->parms.collect_md) {
1490                 netif_keep_dst(dev);
1491         }
1492         ip6gre_tnl_init_features(dev);
1493
1494         dev_hold(dev);
1495         return 0;
1496
1497 cleanup_dst_cache_init:
1498         dst_cache_destroy(&tunnel->dst_cache);
1499 cleanup_alloc_pcpu_stats:
1500         free_percpu(dev->tstats);
1501         dev->tstats = NULL;
1502         return ret;
1503 }
1504
1505 static int ip6gre_tunnel_init(struct net_device *dev)
1506 {
1507         struct ip6_tnl *tunnel;
1508         int ret;
1509
1510         ret = ip6gre_tunnel_init_common(dev);
1511         if (ret)
1512                 return ret;
1513
1514         tunnel = netdev_priv(dev);
1515
1516         if (tunnel->parms.collect_md)
1517                 return 0;
1518
1519         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1520         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1521
1522         if (ipv6_addr_any(&tunnel->parms.raddr))
1523                 dev->header_ops = &ip6gre_header_ops;
1524
1525         return 0;
1526 }
1527
1528 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1529 {
1530         struct ip6_tnl *tunnel = netdev_priv(dev);
1531
1532         tunnel->dev = dev;
1533         tunnel->net = dev_net(dev);
1534         strcpy(tunnel->parms.name, dev->name);
1535
1536         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1537 }
1538
1539 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1540         .handler     = gre_rcv,
1541         .err_handler = ip6gre_err,
1542         .flags       = INET6_PROTO_FINAL,
1543 };
1544
1545 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1546 {
1547         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1548         struct net_device *dev, *aux;
1549         int prio;
1550
1551         for_each_netdev_safe(net, dev, aux)
1552                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1553                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1554                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1555                         unregister_netdevice_queue(dev, head);
1556
1557         for (prio = 0; prio < 4; prio++) {
1558                 int h;
1559                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1560                         struct ip6_tnl *t;
1561
1562                         t = rtnl_dereference(ign->tunnels[prio][h]);
1563
1564                         while (t) {
1565                                 /* If dev is in the same netns, it has already
1566                                  * been added to the list by the previous loop.
1567                                  */
1568                                 if (!net_eq(dev_net(t->dev), net))
1569                                         unregister_netdevice_queue(t->dev,
1570                                                                    head);
1571                                 t = rtnl_dereference(t->next);
1572                         }
1573                 }
1574         }
1575 }
1576
1577 static int __net_init ip6gre_init_net(struct net *net)
1578 {
1579         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1580         struct net_device *ndev;
1581         int err;
1582
1583         if (!net_has_fallback_tunnels(net))
1584                 return 0;
1585         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1586                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1587         if (!ndev) {
1588                 err = -ENOMEM;
1589                 goto err_alloc_dev;
1590         }
1591         ign->fb_tunnel_dev = ndev;
1592         dev_net_set(ign->fb_tunnel_dev, net);
1593         /* FB netdevice is special: we have one, and only one per netns.
1594          * Allowing to move it to another netns is clearly unsafe.
1595          */
1596         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1597
1598
1599         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1600         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1601
1602         err = register_netdev(ign->fb_tunnel_dev);
1603         if (err)
1604                 goto err_reg_dev;
1605
1606         rcu_assign_pointer(ign->tunnels_wc[0],
1607                            netdev_priv(ign->fb_tunnel_dev));
1608         return 0;
1609
1610 err_reg_dev:
1611         free_netdev(ndev);
1612 err_alloc_dev:
1613         return err;
1614 }
1615
1616 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1617 {
1618         struct net *net;
1619         LIST_HEAD(list);
1620
1621         rtnl_lock();
1622         list_for_each_entry(net, net_list, exit_list)
1623                 ip6gre_destroy_tunnels(net, &list);
1624         unregister_netdevice_many(&list);
1625         rtnl_unlock();
1626 }
1627
1628 static struct pernet_operations ip6gre_net_ops = {
1629         .init = ip6gre_init_net,
1630         .exit_batch = ip6gre_exit_batch_net,
1631         .id   = &ip6gre_net_id,
1632         .size = sizeof(struct ip6gre_net),
1633 };
1634
1635 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1636                                   struct netlink_ext_ack *extack)
1637 {
1638         __be16 flags;
1639
1640         if (!data)
1641                 return 0;
1642
1643         flags = 0;
1644         if (data[IFLA_GRE_IFLAGS])
1645                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1646         if (data[IFLA_GRE_OFLAGS])
1647                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1648         if (flags & (GRE_VERSION|GRE_ROUTING))
1649                 return -EINVAL;
1650
1651         return 0;
1652 }
1653
1654 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1655                                struct netlink_ext_ack *extack)
1656 {
1657         struct in6_addr daddr;
1658
1659         if (tb[IFLA_ADDRESS]) {
1660                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1661                         return -EINVAL;
1662                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1663                         return -EADDRNOTAVAIL;
1664         }
1665
1666         if (!data)
1667                 goto out;
1668
1669         if (data[IFLA_GRE_REMOTE]) {
1670                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1671                 if (ipv6_addr_any(&daddr))
1672                         return -EINVAL;
1673         }
1674
1675 out:
1676         return ip6gre_tunnel_validate(tb, data, extack);
1677 }
1678
1679 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1680                                   struct netlink_ext_ack *extack)
1681 {
1682         __be16 flags = 0;
1683         int ret, ver = 0;
1684
1685         if (!data)
1686                 return 0;
1687
1688         ret = ip6gre_tap_validate(tb, data, extack);
1689         if (ret)
1690                 return ret;
1691
1692         /* ERSPAN should only have GRE sequence and key flag */
1693         if (data[IFLA_GRE_OFLAGS])
1694                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1695         if (data[IFLA_GRE_IFLAGS])
1696                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1697         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1698             flags != (GRE_SEQ | GRE_KEY))
1699                 return -EINVAL;
1700
1701         /* ERSPAN Session ID only has 10-bit. Since we reuse
1702          * 32-bit key field as ID, check it's range.
1703          */
1704         if (data[IFLA_GRE_IKEY] &&
1705             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1706                 return -EINVAL;
1707
1708         if (data[IFLA_GRE_OKEY] &&
1709             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1710                 return -EINVAL;
1711
1712         if (data[IFLA_GRE_ERSPAN_VER]) {
1713                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1714                 if (ver != 1 && ver != 2)
1715                         return -EINVAL;
1716         }
1717
1718         if (ver == 1) {
1719                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1720                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1721
1722                         if (index & ~INDEX_MASK)
1723                                 return -EINVAL;
1724                 }
1725         } else if (ver == 2) {
1726                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1727                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1728
1729                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1730                                 return -EINVAL;
1731                 }
1732
1733                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1734                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1735
1736                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1737                                 return -EINVAL;
1738                 }
1739         }
1740
1741         return 0;
1742 }
1743
1744 static void ip6erspan_set_version(struct nlattr *data[],
1745                                   struct __ip6_tnl_parm *parms)
1746 {
1747         if (!data)
1748                 return;
1749
1750         parms->erspan_ver = 1;
1751         if (data[IFLA_GRE_ERSPAN_VER])
1752                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1753
1754         if (parms->erspan_ver == 1) {
1755                 if (data[IFLA_GRE_ERSPAN_INDEX])
1756                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1757         } else if (parms->erspan_ver == 2) {
1758                 if (data[IFLA_GRE_ERSPAN_DIR])
1759                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1760                 if (data[IFLA_GRE_ERSPAN_HWID])
1761                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1762         }
1763 }
1764
1765 static void ip6gre_netlink_parms(struct nlattr *data[],
1766                                 struct __ip6_tnl_parm *parms)
1767 {
1768         memset(parms, 0, sizeof(*parms));
1769
1770         if (!data)
1771                 return;
1772
1773         if (data[IFLA_GRE_LINK])
1774                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1775
1776         if (data[IFLA_GRE_IFLAGS])
1777                 parms->i_flags = gre_flags_to_tnl_flags(
1778                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1779
1780         if (data[IFLA_GRE_OFLAGS])
1781                 parms->o_flags = gre_flags_to_tnl_flags(
1782                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1783
1784         if (data[IFLA_GRE_IKEY])
1785                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1786
1787         if (data[IFLA_GRE_OKEY])
1788                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1789
1790         if (data[IFLA_GRE_LOCAL])
1791                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1792
1793         if (data[IFLA_GRE_REMOTE])
1794                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1795
1796         if (data[IFLA_GRE_TTL])
1797                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1798
1799         if (data[IFLA_GRE_ENCAP_LIMIT])
1800                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1801
1802         if (data[IFLA_GRE_FLOWINFO])
1803                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1804
1805         if (data[IFLA_GRE_FLAGS])
1806                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1807
1808         if (data[IFLA_GRE_FWMARK])
1809                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1810
1811         if (data[IFLA_GRE_COLLECT_METADATA])
1812                 parms->collect_md = true;
1813 }
1814
1815 static int ip6gre_tap_init(struct net_device *dev)
1816 {
1817         int ret;
1818
1819         ret = ip6gre_tunnel_init_common(dev);
1820         if (ret)
1821                 return ret;
1822
1823         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1824
1825         return 0;
1826 }
1827
1828 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1829         .ndo_init = ip6gre_tap_init,
1830         .ndo_uninit = ip6gre_tunnel_uninit,
1831         .ndo_start_xmit = ip6gre_tunnel_xmit,
1832         .ndo_set_mac_address = eth_mac_addr,
1833         .ndo_validate_addr = eth_validate_addr,
1834         .ndo_change_mtu = ip6_tnl_change_mtu,
1835         .ndo_get_stats64 = ip_tunnel_get_stats64,
1836         .ndo_get_iflink = ip6_tnl_get_iflink,
1837 };
1838
1839 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1840 {
1841         int t_hlen;
1842
1843         tunnel->tun_hlen = 8;
1844         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1845                        erspan_hdr_len(tunnel->parms.erspan_ver);
1846
1847         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1848         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1849         return t_hlen;
1850 }
1851
1852 static int ip6erspan_tap_init(struct net_device *dev)
1853 {
1854         struct ip6_tnl *tunnel;
1855         int t_hlen;
1856         int ret;
1857
1858         tunnel = netdev_priv(dev);
1859
1860         tunnel->dev = dev;
1861         tunnel->net = dev_net(dev);
1862         strcpy(tunnel->parms.name, dev->name);
1863
1864         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1865         if (!dev->tstats)
1866                 return -ENOMEM;
1867
1868         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1869         if (ret)
1870                 goto cleanup_alloc_pcpu_stats;
1871
1872         ret = gro_cells_init(&tunnel->gro_cells, dev);
1873         if (ret)
1874                 goto cleanup_dst_cache_init;
1875
1876         t_hlen = ip6erspan_calc_hlen(tunnel);
1877         dev->mtu = ETH_DATA_LEN - t_hlen;
1878         if (dev->type == ARPHRD_ETHER)
1879                 dev->mtu -= ETH_HLEN;
1880         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1881                 dev->mtu -= 8;
1882
1883         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1884         ip6erspan_tnl_link_config(tunnel, 1);
1885
1886         dev_hold(dev);
1887         return 0;
1888
1889 cleanup_dst_cache_init:
1890         dst_cache_destroy(&tunnel->dst_cache);
1891 cleanup_alloc_pcpu_stats:
1892         free_percpu(dev->tstats);
1893         dev->tstats = NULL;
1894         return ret;
1895 }
1896
1897 static const struct net_device_ops ip6erspan_netdev_ops = {
1898         .ndo_init =             ip6erspan_tap_init,
1899         .ndo_uninit =           ip6erspan_tunnel_uninit,
1900         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1901         .ndo_set_mac_address =  eth_mac_addr,
1902         .ndo_validate_addr =    eth_validate_addr,
1903         .ndo_change_mtu =       ip6_tnl_change_mtu,
1904         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1905         .ndo_get_iflink =       ip6_tnl_get_iflink,
1906 };
1907
1908 static void ip6gre_tap_setup(struct net_device *dev)
1909 {
1910
1911         ether_setup(dev);
1912
1913         dev->max_mtu = 0;
1914         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1915         dev->needs_free_netdev = true;
1916         dev->priv_destructor = ip6gre_dev_free;
1917
1918         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1919         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1920         netif_keep_dst(dev);
1921 }
1922
1923 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1924                                        struct ip_tunnel_encap *ipencap)
1925 {
1926         bool ret = false;
1927
1928         memset(ipencap, 0, sizeof(*ipencap));
1929
1930         if (!data)
1931                 return ret;
1932
1933         if (data[IFLA_GRE_ENCAP_TYPE]) {
1934                 ret = true;
1935                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1936         }
1937
1938         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1939                 ret = true;
1940                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1941         }
1942
1943         if (data[IFLA_GRE_ENCAP_SPORT]) {
1944                 ret = true;
1945                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1946         }
1947
1948         if (data[IFLA_GRE_ENCAP_DPORT]) {
1949                 ret = true;
1950                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1951         }
1952
1953         return ret;
1954 }
1955
1956 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1957                                  struct nlattr *tb[], struct nlattr *data[],
1958                                  struct netlink_ext_ack *extack)
1959 {
1960         struct ip6_tnl *nt;
1961         struct ip_tunnel_encap ipencap;
1962         int err;
1963
1964         nt = netdev_priv(dev);
1965
1966         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1967                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1968
1969                 if (err < 0)
1970                         return err;
1971         }
1972
1973         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1974                 eth_hw_addr_random(dev);
1975
1976         nt->dev = dev;
1977         nt->net = dev_net(dev);
1978
1979         err = register_netdevice(dev);
1980         if (err)
1981                 goto out;
1982
1983         if (tb[IFLA_MTU])
1984                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1985
1986 out:
1987         return err;
1988 }
1989
1990 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1991                           struct nlattr *tb[], struct nlattr *data[],
1992                           struct netlink_ext_ack *extack)
1993 {
1994         struct ip6_tnl *nt = netdev_priv(dev);
1995         struct net *net = dev_net(dev);
1996         struct ip6gre_net *ign;
1997         int err;
1998
1999         ip6gre_netlink_parms(data, &nt->parms);
2000         ign = net_generic(net, ip6gre_net_id);
2001
2002         if (nt->parms.collect_md) {
2003                 if (rtnl_dereference(ign->collect_md_tun))
2004                         return -EEXIST;
2005         } else {
2006                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2007                         return -EEXIST;
2008         }
2009
2010         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2011         if (!err) {
2012                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2013                 ip6gre_tunnel_link_md(ign, nt);
2014                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2015         }
2016         return err;
2017 }
2018
2019 static struct ip6_tnl *
2020 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2021                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2022                          struct netlink_ext_ack *extack)
2023 {
2024         struct ip6_tnl *t, *nt = netdev_priv(dev);
2025         struct net *net = nt->net;
2026         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2027         struct ip_tunnel_encap ipencap;
2028
2029         if (dev == ign->fb_tunnel_dev)
2030                 return ERR_PTR(-EINVAL);
2031
2032         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2033                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2034
2035                 if (err < 0)
2036                         return ERR_PTR(err);
2037         }
2038
2039         ip6gre_netlink_parms(data, p_p);
2040
2041         t = ip6gre_tunnel_locate(net, p_p, 0);
2042
2043         if (t) {
2044                 if (t->dev != dev)
2045                         return ERR_PTR(-EEXIST);
2046         } else {
2047                 t = nt;
2048         }
2049
2050         return t;
2051 }
2052
2053 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2054                              struct nlattr *data[],
2055                              struct netlink_ext_ack *extack)
2056 {
2057         struct ip6_tnl *t = netdev_priv(dev);
2058         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2059         struct __ip6_tnl_parm p;
2060
2061         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2062         if (IS_ERR(t))
2063                 return PTR_ERR(t);
2064
2065         ip6gre_tunnel_unlink_md(ign, t);
2066         ip6gre_tunnel_unlink(ign, t);
2067         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2068         ip6gre_tunnel_link_md(ign, t);
2069         ip6gre_tunnel_link(ign, t);
2070         return 0;
2071 }
2072
2073 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2074 {
2075         struct net *net = dev_net(dev);
2076         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2077
2078         if (dev != ign->fb_tunnel_dev)
2079                 unregister_netdevice_queue(dev, head);
2080 }
2081
2082 static size_t ip6gre_get_size(const struct net_device *dev)
2083 {
2084         return
2085                 /* IFLA_GRE_LINK */
2086                 nla_total_size(4) +
2087                 /* IFLA_GRE_IFLAGS */
2088                 nla_total_size(2) +
2089                 /* IFLA_GRE_OFLAGS */
2090                 nla_total_size(2) +
2091                 /* IFLA_GRE_IKEY */
2092                 nla_total_size(4) +
2093                 /* IFLA_GRE_OKEY */
2094                 nla_total_size(4) +
2095                 /* IFLA_GRE_LOCAL */
2096                 nla_total_size(sizeof(struct in6_addr)) +
2097                 /* IFLA_GRE_REMOTE */
2098                 nla_total_size(sizeof(struct in6_addr)) +
2099                 /* IFLA_GRE_TTL */
2100                 nla_total_size(1) +
2101                 /* IFLA_GRE_ENCAP_LIMIT */
2102                 nla_total_size(1) +
2103                 /* IFLA_GRE_FLOWINFO */
2104                 nla_total_size(4) +
2105                 /* IFLA_GRE_FLAGS */
2106                 nla_total_size(4) +
2107                 /* IFLA_GRE_ENCAP_TYPE */
2108                 nla_total_size(2) +
2109                 /* IFLA_GRE_ENCAP_FLAGS */
2110                 nla_total_size(2) +
2111                 /* IFLA_GRE_ENCAP_SPORT */
2112                 nla_total_size(2) +
2113                 /* IFLA_GRE_ENCAP_DPORT */
2114                 nla_total_size(2) +
2115                 /* IFLA_GRE_COLLECT_METADATA */
2116                 nla_total_size(0) +
2117                 /* IFLA_GRE_FWMARK */
2118                 nla_total_size(4) +
2119                 /* IFLA_GRE_ERSPAN_INDEX */
2120                 nla_total_size(4) +
2121                 0;
2122 }
2123
2124 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2125 {
2126         struct ip6_tnl *t = netdev_priv(dev);
2127         struct __ip6_tnl_parm *p = &t->parms;
2128         __be16 o_flags = p->o_flags;
2129
2130         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2131                 if (!p->collect_md)
2132                         o_flags |= TUNNEL_KEY;
2133
2134                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2135                         goto nla_put_failure;
2136
2137                 if (p->erspan_ver == 1) {
2138                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2139                                 goto nla_put_failure;
2140                 } else {
2141                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2142                                 goto nla_put_failure;
2143                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2144                                 goto nla_put_failure;
2145                 }
2146         }
2147
2148         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2149             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2150                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2151             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2152                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2153             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2154             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2155             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2156             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2157             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2158             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2159             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2160             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2161             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2162                 goto nla_put_failure;
2163
2164         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2165                         t->encap.type) ||
2166             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2167                          t->encap.sport) ||
2168             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2169                          t->encap.dport) ||
2170             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2171                         t->encap.flags))
2172                 goto nla_put_failure;
2173
2174         if (p->collect_md) {
2175                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2176                         goto nla_put_failure;
2177         }
2178
2179         return 0;
2180
2181 nla_put_failure:
2182         return -EMSGSIZE;
2183 }
2184
2185 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2186         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2187         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2188         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2189         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2190         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2191         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2192         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2193         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2194         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2195         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2196         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2197         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2198         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2199         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2200         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2201         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2202         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2203         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2204         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2205         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2206         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2207 };
2208
2209 static void ip6erspan_tap_setup(struct net_device *dev)
2210 {
2211         ether_setup(dev);
2212
2213         dev->max_mtu = 0;
2214         dev->netdev_ops = &ip6erspan_netdev_ops;
2215         dev->needs_free_netdev = true;
2216         dev->priv_destructor = ip6gre_dev_free;
2217
2218         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2219         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2220         netif_keep_dst(dev);
2221 }
2222
2223 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2224                              struct nlattr *tb[], struct nlattr *data[],
2225                              struct netlink_ext_ack *extack)
2226 {
2227         struct ip6_tnl *nt = netdev_priv(dev);
2228         struct net *net = dev_net(dev);
2229         struct ip6gre_net *ign;
2230         int err;
2231
2232         ip6gre_netlink_parms(data, &nt->parms);
2233         ip6erspan_set_version(data, &nt->parms);
2234         ign = net_generic(net, ip6gre_net_id);
2235
2236         if (nt->parms.collect_md) {
2237                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2238                         return -EEXIST;
2239         } else {
2240                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2241                         return -EEXIST;
2242         }
2243
2244         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2245         if (!err) {
2246                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2247                 ip6erspan_tunnel_link_md(ign, nt);
2248                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2249         }
2250         return err;
2251 }
2252
2253 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2254 {
2255         ip6gre_tnl_link_config_common(t);
2256         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2257 }
2258
2259 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2260                                 const struct __ip6_tnl_parm *p, int set_mtu)
2261 {
2262         ip6gre_tnl_copy_tnl_parm(t, p);
2263         ip6erspan_tnl_link_config(t, set_mtu);
2264         return 0;
2265 }
2266
2267 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2268                                 struct nlattr *data[],
2269                                 struct netlink_ext_ack *extack)
2270 {
2271         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2272         struct __ip6_tnl_parm p;
2273         struct ip6_tnl *t;
2274
2275         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2276         if (IS_ERR(t))
2277                 return PTR_ERR(t);
2278
2279         ip6erspan_set_version(data, &p);
2280         ip6gre_tunnel_unlink_md(ign, t);
2281         ip6gre_tunnel_unlink(ign, t);
2282         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2283         ip6erspan_tunnel_link_md(ign, t);
2284         ip6gre_tunnel_link(ign, t);
2285         return 0;
2286 }
2287
2288 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2289         .kind           = "ip6gre",
2290         .maxtype        = IFLA_GRE_MAX,
2291         .policy         = ip6gre_policy,
2292         .priv_size      = sizeof(struct ip6_tnl),
2293         .setup          = ip6gre_tunnel_setup,
2294         .validate       = ip6gre_tunnel_validate,
2295         .newlink        = ip6gre_newlink,
2296         .changelink     = ip6gre_changelink,
2297         .dellink        = ip6gre_dellink,
2298         .get_size       = ip6gre_get_size,
2299         .fill_info      = ip6gre_fill_info,
2300         .get_link_net   = ip6_tnl_get_link_net,
2301 };
2302
2303 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2304         .kind           = "ip6gretap",
2305         .maxtype        = IFLA_GRE_MAX,
2306         .policy         = ip6gre_policy,
2307         .priv_size      = sizeof(struct ip6_tnl),
2308         .setup          = ip6gre_tap_setup,
2309         .validate       = ip6gre_tap_validate,
2310         .newlink        = ip6gre_newlink,
2311         .changelink     = ip6gre_changelink,
2312         .get_size       = ip6gre_get_size,
2313         .fill_info      = ip6gre_fill_info,
2314         .get_link_net   = ip6_tnl_get_link_net,
2315 };
2316
2317 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2318         .kind           = "ip6erspan",
2319         .maxtype        = IFLA_GRE_MAX,
2320         .policy         = ip6gre_policy,
2321         .priv_size      = sizeof(struct ip6_tnl),
2322         .setup          = ip6erspan_tap_setup,
2323         .validate       = ip6erspan_tap_validate,
2324         .newlink        = ip6erspan_newlink,
2325         .changelink     = ip6erspan_changelink,
2326         .get_size       = ip6gre_get_size,
2327         .fill_info      = ip6gre_fill_info,
2328         .get_link_net   = ip6_tnl_get_link_net,
2329 };
2330
2331 /*
2332  *      And now the modules code and kernel interface.
2333  */
2334
2335 static int __init ip6gre_init(void)
2336 {
2337         int err;
2338
2339         pr_info("GRE over IPv6 tunneling driver\n");
2340
2341         err = register_pernet_device(&ip6gre_net_ops);
2342         if (err < 0)
2343                 return err;
2344
2345         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2346         if (err < 0) {
2347                 pr_info("%s: can't add protocol\n", __func__);
2348                 goto add_proto_failed;
2349         }
2350
2351         err = rtnl_link_register(&ip6gre_link_ops);
2352         if (err < 0)
2353                 goto rtnl_link_failed;
2354
2355         err = rtnl_link_register(&ip6gre_tap_ops);
2356         if (err < 0)
2357                 goto tap_ops_failed;
2358
2359         err = rtnl_link_register(&ip6erspan_tap_ops);
2360         if (err < 0)
2361                 goto erspan_link_failed;
2362
2363 out:
2364         return err;
2365
2366 erspan_link_failed:
2367         rtnl_link_unregister(&ip6gre_tap_ops);
2368 tap_ops_failed:
2369         rtnl_link_unregister(&ip6gre_link_ops);
2370 rtnl_link_failed:
2371         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2372 add_proto_failed:
2373         unregister_pernet_device(&ip6gre_net_ops);
2374         goto out;
2375 }
2376
2377 static void __exit ip6gre_fini(void)
2378 {
2379         rtnl_link_unregister(&ip6gre_tap_ops);
2380         rtnl_link_unregister(&ip6gre_link_ops);
2381         rtnl_link_unregister(&ip6erspan_tap_ops);
2382         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2383         unregister_pernet_device(&ip6gre_net_ops);
2384 }
2385
2386 module_init(ip6gre_init);
2387 module_exit(ip6gre_fini);
2388 MODULE_LICENSE("GPL");
2389 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2390 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2391 MODULE_ALIAS_RTNL_LINK("ip6gre");
2392 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2393 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2394 MODULE_ALIAS_NETDEV("ip6gre0");