Merge tag 'keys-namespace-20190627' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-microblaze.git] / net / ipv6 / ip6_gre.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      GRE over IPv6 protocol decoder.
4  *
5  *      Authors: Dmitry Kozlov (xeb@mail.ru)
6  */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
18 #include <linux/in.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/rtnetlink.h>
47
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/gre.h>
53 #include <net/erspan.h>
54 #include <net/dst_metadata.h>
55
56
57 static bool log_ecn_error = true;
58 module_param(log_ecn_error, bool, 0644);
59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60
61 #define IP6_GRE_HASH_SIZE_SHIFT  5
62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63
64 static unsigned int ip6gre_net_id __read_mostly;
65 struct ip6gre_net {
66         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67
68         struct ip6_tnl __rcu *collect_md_tun;
69         struct ip6_tnl __rcu *collect_md_tun_erspan;
70         struct net_device *fb_tunnel_dev;
71 };
72
73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85    4 hash tables:
86
87    3: (remote,local)
88    2: (remote,*)
89    1: (*,local)
90    0: (*,*)
91
92    We require exact key match i.e. if a key is present in packet
93    it will match only tunnel with the same key; if it is not present,
94    it will match only keyless tunnel.
95
96    All keysless packets, if not matched configured keyless tunnels
97    will match fallback tunnel.
98  */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103         u32 hash = ipv6_addr_hash(addr);
104
105         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l     tunnels[3]
109 #define tunnels_r       tunnels[2]
110 #define tunnels_l       tunnels[1]
111 #define tunnels_wc      tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116                 const struct in6_addr *remote, const struct in6_addr *local,
117                 __be32 key, __be16 gre_proto)
118 {
119         struct net *net = dev_net(dev);
120         int link = dev->ifindex;
121         unsigned int h0 = HASH_ADDR(remote);
122         unsigned int h1 = HASH_KEY(key);
123         struct ip6_tnl *t, *cand = NULL;
124         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126                         gre_proto == htons(ETH_P_ERSPAN) ||
127                         gre_proto == htons(ETH_P_ERSPAN2)) ?
128                        ARPHRD_ETHER : ARPHRD_IP6GRE;
129         int score, cand_score = 4;
130
131         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
132                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
133                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
134                     key != t->parms.i_key ||
135                     !(t->dev->flags & IFF_UP))
136                         continue;
137
138                 if (t->dev->type != ARPHRD_IP6GRE &&
139                     t->dev->type != dev_type)
140                         continue;
141
142                 score = 0;
143                 if (t->parms.link != link)
144                         score |= 1;
145                 if (t->dev->type != dev_type)
146                         score |= 2;
147                 if (score == 0)
148                         return t;
149
150                 if (score < cand_score) {
151                         cand = t;
152                         cand_score = score;
153                 }
154         }
155
156         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
157                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
158                     key != t->parms.i_key ||
159                     !(t->dev->flags & IFF_UP))
160                         continue;
161
162                 if (t->dev->type != ARPHRD_IP6GRE &&
163                     t->dev->type != dev_type)
164                         continue;
165
166                 score = 0;
167                 if (t->parms.link != link)
168                         score |= 1;
169                 if (t->dev->type != dev_type)
170                         score |= 2;
171                 if (score == 0)
172                         return t;
173
174                 if (score < cand_score) {
175                         cand = t;
176                         cand_score = score;
177                 }
178         }
179
180         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
181                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
182                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
183                                  !ipv6_addr_is_multicast(local))) ||
184                     key != t->parms.i_key ||
185                     !(t->dev->flags & IFF_UP))
186                         continue;
187
188                 if (t->dev->type != ARPHRD_IP6GRE &&
189                     t->dev->type != dev_type)
190                         continue;
191
192                 score = 0;
193                 if (t->parms.link != link)
194                         score |= 1;
195                 if (t->dev->type != dev_type)
196                         score |= 2;
197                 if (score == 0)
198                         return t;
199
200                 if (score < cand_score) {
201                         cand = t;
202                         cand_score = score;
203                 }
204         }
205
206         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
207                 if (t->parms.i_key != key ||
208                     !(t->dev->flags & IFF_UP))
209                         continue;
210
211                 if (t->dev->type != ARPHRD_IP6GRE &&
212                     t->dev->type != dev_type)
213                         continue;
214
215                 score = 0;
216                 if (t->parms.link != link)
217                         score |= 1;
218                 if (t->dev->type != dev_type)
219                         score |= 2;
220                 if (score == 0)
221                         return t;
222
223                 if (score < cand_score) {
224                         cand = t;
225                         cand_score = score;
226                 }
227         }
228
229         if (cand)
230                 return cand;
231
232         if (gre_proto == htons(ETH_P_ERSPAN) ||
233             gre_proto == htons(ETH_P_ERSPAN2))
234                 t = rcu_dereference(ign->collect_md_tun_erspan);
235         else
236                 t = rcu_dereference(ign->collect_md_tun);
237
238         if (t && t->dev->flags & IFF_UP)
239                 return t;
240
241         dev = ign->fb_tunnel_dev;
242         if (dev && dev->flags & IFF_UP)
243                 return netdev_priv(dev);
244
245         return NULL;
246 }
247
248 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
249                 const struct __ip6_tnl_parm *p)
250 {
251         const struct in6_addr *remote = &p->raddr;
252         const struct in6_addr *local = &p->laddr;
253         unsigned int h = HASH_KEY(p->i_key);
254         int prio = 0;
255
256         if (!ipv6_addr_any(local))
257                 prio |= 1;
258         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
259                 prio |= 2;
260                 h ^= HASH_ADDR(remote);
261         }
262
263         return &ign->tunnels[prio][h];
264 }
265
266 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
267 {
268         if (t->parms.collect_md)
269                 rcu_assign_pointer(ign->collect_md_tun, t);
270 }
271
272 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
273 {
274         if (t->parms.collect_md)
275                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
276 }
277
278 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
279 {
280         if (t->parms.collect_md)
281                 rcu_assign_pointer(ign->collect_md_tun, NULL);
282 }
283
284 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
285                                        struct ip6_tnl *t)
286 {
287         if (t->parms.collect_md)
288                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
289 }
290
291 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
292                 const struct ip6_tnl *t)
293 {
294         return __ip6gre_bucket(ign, &t->parms);
295 }
296
297 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
298 {
299         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
300
301         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
302         rcu_assign_pointer(*tp, t);
303 }
304
305 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
306 {
307         struct ip6_tnl __rcu **tp;
308         struct ip6_tnl *iter;
309
310         for (tp = ip6gre_bucket(ign, t);
311              (iter = rtnl_dereference(*tp)) != NULL;
312              tp = &iter->next) {
313                 if (t == iter) {
314                         rcu_assign_pointer(*tp, t->next);
315                         break;
316                 }
317         }
318 }
319
320 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
321                                            const struct __ip6_tnl_parm *parms,
322                                            int type)
323 {
324         const struct in6_addr *remote = &parms->raddr;
325         const struct in6_addr *local = &parms->laddr;
326         __be32 key = parms->i_key;
327         int link = parms->link;
328         struct ip6_tnl *t;
329         struct ip6_tnl __rcu **tp;
330         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
331
332         for (tp = __ip6gre_bucket(ign, parms);
333              (t = rtnl_dereference(*tp)) != NULL;
334              tp = &t->next)
335                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
336                     ipv6_addr_equal(remote, &t->parms.raddr) &&
337                     key == t->parms.i_key &&
338                     link == t->parms.link &&
339                     type == t->dev->type)
340                         break;
341
342         return t;
343 }
344
345 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
346                 const struct __ip6_tnl_parm *parms, int create)
347 {
348         struct ip6_tnl *t, *nt;
349         struct net_device *dev;
350         char name[IFNAMSIZ];
351         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
352
353         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
354         if (t && create)
355                 return NULL;
356         if (t || !create)
357                 return t;
358
359         if (parms->name[0]) {
360                 if (!dev_valid_name(parms->name))
361                         return NULL;
362                 strlcpy(name, parms->name, IFNAMSIZ);
363         } else {
364                 strcpy(name, "ip6gre%d");
365         }
366         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
367                            ip6gre_tunnel_setup);
368         if (!dev)
369                 return NULL;
370
371         dev_net_set(dev, net);
372
373         nt = netdev_priv(dev);
374         nt->parms = *parms;
375         dev->rtnl_link_ops = &ip6gre_link_ops;
376
377         nt->dev = dev;
378         nt->net = dev_net(dev);
379
380         if (register_netdevice(dev) < 0)
381                 goto failed_free;
382
383         ip6gre_tnl_link_config(nt, 1);
384
385         /* Can use a lockless transmit, unless we generate output sequences */
386         if (!(nt->parms.o_flags & TUNNEL_SEQ))
387                 dev->features |= NETIF_F_LLTX;
388
389         dev_hold(dev);
390         ip6gre_tunnel_link(ign, nt);
391         return nt;
392
393 failed_free:
394         free_netdev(dev);
395         return NULL;
396 }
397
398 static void ip6erspan_tunnel_uninit(struct net_device *dev)
399 {
400         struct ip6_tnl *t = netdev_priv(dev);
401         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
402
403         ip6erspan_tunnel_unlink_md(ign, t);
404         ip6gre_tunnel_unlink(ign, t);
405         dst_cache_reset(&t->dst_cache);
406         dev_put(dev);
407 }
408
409 static void ip6gre_tunnel_uninit(struct net_device *dev)
410 {
411         struct ip6_tnl *t = netdev_priv(dev);
412         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
413
414         ip6gre_tunnel_unlink_md(ign, t);
415         ip6gre_tunnel_unlink(ign, t);
416         dst_cache_reset(&t->dst_cache);
417         dev_put(dev);
418 }
419
420
421 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
422                        u8 type, u8 code, int offset, __be32 info)
423 {
424         struct net *net = dev_net(skb->dev);
425         const struct ipv6hdr *ipv6h;
426         struct tnl_ptk_info tpi;
427         struct ip6_tnl *t;
428
429         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
430                              offset) < 0)
431                 return -EINVAL;
432
433         ipv6h = (const struct ipv6hdr *)skb->data;
434         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
435                                  tpi.key, tpi.proto);
436         if (!t)
437                 return -ENOENT;
438
439         switch (type) {
440                 struct ipv6_tlv_tnl_enc_lim *tel;
441                 __u32 teli;
442         case ICMPV6_DEST_UNREACH:
443                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
444                                     t->parms.name);
445                 if (code != ICMPV6_PORT_UNREACH)
446                         break;
447                 return 0;
448         case ICMPV6_TIME_EXCEED:
449                 if (code == ICMPV6_EXC_HOPLIMIT) {
450                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
451                                             t->parms.name);
452                         break;
453                 }
454                 return 0;
455         case ICMPV6_PARAMPROB:
456                 teli = 0;
457                 if (code == ICMPV6_HDR_FIELD)
458                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
459
460                 if (teli && teli == be32_to_cpu(info) - 2) {
461                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
462                         if (tel->encap_limit == 0) {
463                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
464                                                     t->parms.name);
465                         }
466                 } else {
467                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
468                                             t->parms.name);
469                 }
470                 return 0;
471         case ICMPV6_PKT_TOOBIG:
472                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
473                 return 0;
474         case NDISC_REDIRECT:
475                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
476                              sock_net_uid(net, NULL));
477                 return 0;
478         }
479
480         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
481                 t->err_count++;
482         else
483                 t->err_count = 1;
484         t->err_time = jiffies;
485
486         return 0;
487 }
488
489 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
490 {
491         const struct ipv6hdr *ipv6h;
492         struct ip6_tnl *tunnel;
493
494         ipv6h = ipv6_hdr(skb);
495         tunnel = ip6gre_tunnel_lookup(skb->dev,
496                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
497                                       tpi->proto);
498         if (tunnel) {
499                 if (tunnel->parms.collect_md) {
500                         struct metadata_dst *tun_dst;
501                         __be64 tun_id;
502                         __be16 flags;
503
504                         flags = tpi->flags;
505                         tun_id = key32_to_tunnel_id(tpi->key);
506
507                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
508                         if (!tun_dst)
509                                 return PACKET_REJECT;
510
511                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
512                 } else {
513                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
514                 }
515
516                 return PACKET_RCVD;
517         }
518
519         return PACKET_REJECT;
520 }
521
522 static int ip6erspan_rcv(struct sk_buff *skb,
523                          struct tnl_ptk_info *tpi,
524                          int gre_hdr_len)
525 {
526         struct erspan_base_hdr *ershdr;
527         const struct ipv6hdr *ipv6h;
528         struct erspan_md2 *md2;
529         struct ip6_tnl *tunnel;
530         u8 ver;
531
532         ipv6h = ipv6_hdr(skb);
533         ershdr = (struct erspan_base_hdr *)skb->data;
534         ver = ershdr->ver;
535
536         tunnel = ip6gre_tunnel_lookup(skb->dev,
537                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
538                                       tpi->proto);
539         if (tunnel) {
540                 int len = erspan_hdr_len(ver);
541
542                 if (unlikely(!pskb_may_pull(skb, len)))
543                         return PACKET_REJECT;
544
545                 if (__iptunnel_pull_header(skb, len,
546                                            htons(ETH_P_TEB),
547                                            false, false) < 0)
548                         return PACKET_REJECT;
549
550                 if (tunnel->parms.collect_md) {
551                         struct erspan_metadata *pkt_md, *md;
552                         struct metadata_dst *tun_dst;
553                         struct ip_tunnel_info *info;
554                         unsigned char *gh;
555                         __be64 tun_id;
556                         __be16 flags;
557
558                         tpi->flags |= TUNNEL_KEY;
559                         flags = tpi->flags;
560                         tun_id = key32_to_tunnel_id(tpi->key);
561
562                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
563                                                   sizeof(*md));
564                         if (!tun_dst)
565                                 return PACKET_REJECT;
566
567                         /* skb can be uncloned in __iptunnel_pull_header, so
568                          * old pkt_md is no longer valid and we need to reset
569                          * it
570                          */
571                         gh = skb_network_header(skb) +
572                              skb_network_header_len(skb);
573                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
574                                                             sizeof(*ershdr));
575                         info = &tun_dst->u.tun_info;
576                         md = ip_tunnel_info_opts(info);
577                         md->version = ver;
578                         md2 = &md->u.md2;
579                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
580                                                        ERSPAN_V2_MDSIZE);
581                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
582                         info->options_len = sizeof(*md);
583
584                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
585
586                 } else {
587                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
588                 }
589
590                 return PACKET_RCVD;
591         }
592
593         return PACKET_REJECT;
594 }
595
596 static int gre_rcv(struct sk_buff *skb)
597 {
598         struct tnl_ptk_info tpi;
599         bool csum_err = false;
600         int hdr_len;
601
602         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
603         if (hdr_len < 0)
604                 goto drop;
605
606         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
607                 goto drop;
608
609         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
610                      tpi.proto == htons(ETH_P_ERSPAN2))) {
611                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
612                         return 0;
613                 goto out;
614         }
615
616         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
617                 return 0;
618
619 out:
620         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
621 drop:
622         kfree_skb(skb);
623         return 0;
624 }
625
626 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
627 {
628         return iptunnel_handle_offloads(skb,
629                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
630 }
631
632 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
633                                      struct net_device *dev,
634                                      struct flowi6 *fl6, __u8 *dsfield,
635                                      int *encap_limit)
636 {
637         const struct iphdr *iph = ip_hdr(skb);
638         struct ip6_tnl *t = netdev_priv(dev);
639
640         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
641                 *encap_limit = t->parms.encap_limit;
642
643         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
644
645         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
646                 *dsfield = ipv4_get_dsfield(iph);
647         else
648                 *dsfield = ip6_tclass(t->parms.flowinfo);
649
650         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
651                 fl6->flowi6_mark = skb->mark;
652         else
653                 fl6->flowi6_mark = t->parms.fwmark;
654
655         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
656 }
657
658 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
659                                     struct net_device *dev,
660                                     struct flowi6 *fl6, __u8 *dsfield,
661                                     int *encap_limit)
662 {
663         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
664         struct ip6_tnl *t = netdev_priv(dev);
665         __u16 offset;
666
667         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
668         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
669
670         if (offset > 0) {
671                 struct ipv6_tlv_tnl_enc_lim *tel;
672
673                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
674                 if (tel->encap_limit == 0) {
675                         icmpv6_send(skb, ICMPV6_PARAMPROB,
676                                     ICMPV6_HDR_FIELD, offset + 2);
677                         return -1;
678                 }
679                 *encap_limit = tel->encap_limit - 1;
680         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
681                 *encap_limit = t->parms.encap_limit;
682         }
683
684         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
685
686         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
687                 *dsfield = ipv6_get_dsfield(ipv6h);
688         else
689                 *dsfield = ip6_tclass(t->parms.flowinfo);
690
691         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
692                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
693
694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
695                 fl6->flowi6_mark = skb->mark;
696         else
697                 fl6->flowi6_mark = t->parms.fwmark;
698
699         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
700
701         return 0;
702 }
703
704 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
705                                struct net_device *dev, __u8 dsfield,
706                                struct flowi6 *fl6, int encap_limit,
707                                __u32 *pmtu, __be16 proto)
708 {
709         struct ip6_tnl *tunnel = netdev_priv(dev);
710         __be16 protocol;
711
712         if (dev->type == ARPHRD_ETHER)
713                 IPCB(skb)->flags = 0;
714
715         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
716                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
717         else
718                 fl6->daddr = tunnel->parms.raddr;
719
720         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
721                 return -ENOMEM;
722
723         /* Push GRE header. */
724         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
725
726         if (tunnel->parms.collect_md) {
727                 struct ip_tunnel_info *tun_info;
728                 const struct ip_tunnel_key *key;
729                 __be16 flags;
730
731                 tun_info = skb_tunnel_info(skb);
732                 if (unlikely(!tun_info ||
733                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
734                              ip_tunnel_info_af(tun_info) != AF_INET6))
735                         return -EINVAL;
736
737                 key = &tun_info->key;
738                 memset(fl6, 0, sizeof(*fl6));
739                 fl6->flowi6_proto = IPPROTO_GRE;
740                 fl6->daddr = key->u.ipv6.dst;
741                 fl6->flowlabel = key->label;
742                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
743
744                 dsfield = key->tos;
745                 flags = key->tun_flags &
746                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
747                 tunnel->tun_hlen = gre_calc_hlen(flags);
748
749                 gre_build_header(skb, tunnel->tun_hlen,
750                                  flags, protocol,
751                                  tunnel_id_to_key32(tun_info->key.tun_id),
752                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
753                                                       : 0);
754
755         } else {
756                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
757                         tunnel->o_seqno++;
758
759                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
760                                  protocol, tunnel->parms.o_key,
761                                  htonl(tunnel->o_seqno));
762         }
763
764         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
765                             NEXTHDR_GRE);
766 }
767
768 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
769 {
770         struct ip6_tnl *t = netdev_priv(dev);
771         int encap_limit = -1;
772         struct flowi6 fl6;
773         __u8 dsfield = 0;
774         __u32 mtu;
775         int err;
776
777         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
778
779         if (!t->parms.collect_md)
780                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
781                                          &dsfield, &encap_limit);
782
783         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
784         if (err)
785                 return -1;
786
787         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
788                           skb->protocol);
789         if (err != 0) {
790                 /* XXX: send ICMP error even if DF is not set. */
791                 if (err == -EMSGSIZE)
792                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
793                                   htonl(mtu));
794                 return -1;
795         }
796
797         return 0;
798 }
799
800 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
801 {
802         struct ip6_tnl *t = netdev_priv(dev);
803         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
804         int encap_limit = -1;
805         struct flowi6 fl6;
806         __u8 dsfield = 0;
807         __u32 mtu;
808         int err;
809
810         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
811                 return -1;
812
813         if (!t->parms.collect_md &&
814             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
815                 return -1;
816
817         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
818                 return -1;
819
820         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
821                           &mtu, skb->protocol);
822         if (err != 0) {
823                 if (err == -EMSGSIZE)
824                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
825                 return -1;
826         }
827
828         return 0;
829 }
830
831 /**
832  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
833  *   @t: the outgoing tunnel device
834  *   @hdr: IPv6 header from the incoming packet
835  *
836  * Description:
837  *   Avoid trivial tunneling loop by checking that tunnel exit-point
838  *   doesn't match source of incoming packet.
839  *
840  * Return:
841  *   1 if conflict,
842  *   0 else
843  **/
844
845 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
846         const struct ipv6hdr *hdr)
847 {
848         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
849 }
850
851 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
852 {
853         struct ip6_tnl *t = netdev_priv(dev);
854         int encap_limit = -1;
855         struct flowi6 fl6;
856         __u32 mtu;
857         int err;
858
859         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
860                 encap_limit = t->parms.encap_limit;
861
862         if (!t->parms.collect_md)
863                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
864
865         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
866         if (err)
867                 return err;
868
869         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
870
871         return err;
872 }
873
874 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
875         struct net_device *dev)
876 {
877         struct ip6_tnl *t = netdev_priv(dev);
878         struct net_device_stats *stats = &t->dev->stats;
879         int ret;
880
881         if (!pskb_inet_may_pull(skb))
882                 goto tx_err;
883
884         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
885                 goto tx_err;
886
887         switch (skb->protocol) {
888         case htons(ETH_P_IP):
889                 ret = ip6gre_xmit_ipv4(skb, dev);
890                 break;
891         case htons(ETH_P_IPV6):
892                 ret = ip6gre_xmit_ipv6(skb, dev);
893                 break;
894         default:
895                 ret = ip6gre_xmit_other(skb, dev);
896                 break;
897         }
898
899         if (ret < 0)
900                 goto tx_err;
901
902         return NETDEV_TX_OK;
903
904 tx_err:
905         stats->tx_errors++;
906         stats->tx_dropped++;
907         kfree_skb(skb);
908         return NETDEV_TX_OK;
909 }
910
911 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
912                                          struct net_device *dev)
913 {
914         struct ip6_tnl *t = netdev_priv(dev);
915         struct dst_entry *dst = skb_dst(skb);
916         struct net_device_stats *stats;
917         bool truncate = false;
918         int encap_limit = -1;
919         __u8 dsfield = false;
920         struct flowi6 fl6;
921         int err = -EINVAL;
922         __be16 proto;
923         __u32 mtu;
924         int nhoff;
925         int thoff;
926
927         if (!pskb_inet_may_pull(skb))
928                 goto tx_err;
929
930         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
931                 goto tx_err;
932
933         if (gre_handle_offloads(skb, false))
934                 goto tx_err;
935
936         if (skb->len > dev->mtu + dev->hard_header_len) {
937                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
938                 truncate = true;
939         }
940
941         nhoff = skb_network_header(skb) - skb_mac_header(skb);
942         if (skb->protocol == htons(ETH_P_IP) &&
943             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
944                 truncate = true;
945
946         thoff = skb_transport_header(skb) - skb_mac_header(skb);
947         if (skb->protocol == htons(ETH_P_IPV6) &&
948             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
949                 truncate = true;
950
951         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
952                 goto tx_err;
953
954         t->parms.o_flags &= ~TUNNEL_KEY;
955         IPCB(skb)->flags = 0;
956
957         /* For collect_md mode, derive fl6 from the tunnel key,
958          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
959          */
960         if (t->parms.collect_md) {
961                 struct ip_tunnel_info *tun_info;
962                 const struct ip_tunnel_key *key;
963                 struct erspan_metadata *md;
964                 __be32 tun_id;
965
966                 tun_info = skb_tunnel_info(skb);
967                 if (unlikely(!tun_info ||
968                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
969                              ip_tunnel_info_af(tun_info) != AF_INET6))
970                         return -EINVAL;
971
972                 key = &tun_info->key;
973                 memset(&fl6, 0, sizeof(fl6));
974                 fl6.flowi6_proto = IPPROTO_GRE;
975                 fl6.daddr = key->u.ipv6.dst;
976                 fl6.flowlabel = key->label;
977                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
978
979                 dsfield = key->tos;
980                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
981                         goto tx_err;
982                 md = ip_tunnel_info_opts(tun_info);
983                 if (!md)
984                         goto tx_err;
985
986                 tun_id = tunnel_id_to_key32(key->tun_id);
987                 if (md->version == 1) {
988                         erspan_build_header(skb,
989                                             ntohl(tun_id),
990                                             ntohl(md->u.index), truncate,
991                                             false);
992                 } else if (md->version == 2) {
993                         erspan_build_header_v2(skb,
994                                                ntohl(tun_id),
995                                                md->u.md2.dir,
996                                                get_hwid(&md->u.md2),
997                                                truncate, false);
998                 } else {
999                         goto tx_err;
1000                 }
1001         } else {
1002                 switch (skb->protocol) {
1003                 case htons(ETH_P_IP):
1004                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1005                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1006                                                  &dsfield, &encap_limit);
1007                         break;
1008                 case htons(ETH_P_IPV6):
1009                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1010                                 goto tx_err;
1011                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1012                                                      &dsfield, &encap_limit))
1013                                 goto tx_err;
1014                         break;
1015                 default:
1016                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1017                         break;
1018                 }
1019
1020                 if (t->parms.erspan_ver == 1)
1021                         erspan_build_header(skb, ntohl(t->parms.o_key),
1022                                             t->parms.index,
1023                                             truncate, false);
1024                 else if (t->parms.erspan_ver == 2)
1025                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1026                                                t->parms.dir,
1027                                                t->parms.hwid,
1028                                                truncate, false);
1029                 else
1030                         goto tx_err;
1031
1032                 fl6.daddr = t->parms.raddr;
1033         }
1034
1035         /* Push GRE header. */
1036         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1037                                            : htons(ETH_P_ERSPAN2);
1038         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1039
1040         /* TooBig packet may have updated dst->dev's mtu */
1041         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1042                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1043
1044         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1045                            NEXTHDR_GRE);
1046         if (err != 0) {
1047                 /* XXX: send ICMP error even if DF is not set. */
1048                 if (err == -EMSGSIZE) {
1049                         if (skb->protocol == htons(ETH_P_IP))
1050                                 icmp_send(skb, ICMP_DEST_UNREACH,
1051                                           ICMP_FRAG_NEEDED, htonl(mtu));
1052                         else
1053                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1054                 }
1055
1056                 goto tx_err;
1057         }
1058         return NETDEV_TX_OK;
1059
1060 tx_err:
1061         stats = &t->dev->stats;
1062         stats->tx_errors++;
1063         stats->tx_dropped++;
1064         kfree_skb(skb);
1065         return NETDEV_TX_OK;
1066 }
1067
1068 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1069 {
1070         struct net_device *dev = t->dev;
1071         struct __ip6_tnl_parm *p = &t->parms;
1072         struct flowi6 *fl6 = &t->fl.u.ip6;
1073
1074         if (dev->type != ARPHRD_ETHER) {
1075                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1076                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1077         }
1078
1079         /* Set up flowi template */
1080         fl6->saddr = p->laddr;
1081         fl6->daddr = p->raddr;
1082         fl6->flowi6_oif = p->link;
1083         fl6->flowlabel = 0;
1084         fl6->flowi6_proto = IPPROTO_GRE;
1085
1086         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1087                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1088         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1089                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1090
1091         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1092         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1093
1094         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1095                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1096                 dev->flags |= IFF_POINTOPOINT;
1097         else
1098                 dev->flags &= ~IFF_POINTOPOINT;
1099 }
1100
1101 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1102                                          int t_hlen)
1103 {
1104         const struct __ip6_tnl_parm *p = &t->parms;
1105         struct net_device *dev = t->dev;
1106
1107         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1108                 int strict = (ipv6_addr_type(&p->raddr) &
1109                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1110
1111                 struct rt6_info *rt = rt6_lookup(t->net,
1112                                                  &p->raddr, &p->laddr,
1113                                                  p->link, NULL, strict);
1114
1115                 if (!rt)
1116                         return;
1117
1118                 if (rt->dst.dev) {
1119                         dev->needed_headroom = rt->dst.dev->hard_header_len +
1120                                                t_hlen;
1121
1122                         if (set_mtu) {
1123                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1124                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1125                                         dev->mtu -= 8;
1126                                 if (dev->type == ARPHRD_ETHER)
1127                                         dev->mtu -= ETH_HLEN;
1128
1129                                 if (dev->mtu < IPV6_MIN_MTU)
1130                                         dev->mtu = IPV6_MIN_MTU;
1131                         }
1132                 }
1133                 ip6_rt_put(rt);
1134         }
1135 }
1136
1137 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1138 {
1139         int t_hlen;
1140
1141         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1142         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1143
1144         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1145         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1146         return t_hlen;
1147 }
1148
1149 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1150 {
1151         ip6gre_tnl_link_config_common(t);
1152         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1153 }
1154
1155 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1156                                      const struct __ip6_tnl_parm *p)
1157 {
1158         t->parms.laddr = p->laddr;
1159         t->parms.raddr = p->raddr;
1160         t->parms.flags = p->flags;
1161         t->parms.hop_limit = p->hop_limit;
1162         t->parms.encap_limit = p->encap_limit;
1163         t->parms.flowinfo = p->flowinfo;
1164         t->parms.link = p->link;
1165         t->parms.proto = p->proto;
1166         t->parms.i_key = p->i_key;
1167         t->parms.o_key = p->o_key;
1168         t->parms.i_flags = p->i_flags;
1169         t->parms.o_flags = p->o_flags;
1170         t->parms.fwmark = p->fwmark;
1171         t->parms.erspan_ver = p->erspan_ver;
1172         t->parms.index = p->index;
1173         t->parms.dir = p->dir;
1174         t->parms.hwid = p->hwid;
1175         dst_cache_reset(&t->dst_cache);
1176 }
1177
1178 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1179                              int set_mtu)
1180 {
1181         ip6gre_tnl_copy_tnl_parm(t, p);
1182         ip6gre_tnl_link_config(t, set_mtu);
1183         return 0;
1184 }
1185
1186 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1187         const struct ip6_tnl_parm2 *u)
1188 {
1189         p->laddr = u->laddr;
1190         p->raddr = u->raddr;
1191         p->flags = u->flags;
1192         p->hop_limit = u->hop_limit;
1193         p->encap_limit = u->encap_limit;
1194         p->flowinfo = u->flowinfo;
1195         p->link = u->link;
1196         p->i_key = u->i_key;
1197         p->o_key = u->o_key;
1198         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1199         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1200         memcpy(p->name, u->name, sizeof(u->name));
1201 }
1202
1203 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1204         const struct __ip6_tnl_parm *p)
1205 {
1206         u->proto = IPPROTO_GRE;
1207         u->laddr = p->laddr;
1208         u->raddr = p->raddr;
1209         u->flags = p->flags;
1210         u->hop_limit = p->hop_limit;
1211         u->encap_limit = p->encap_limit;
1212         u->flowinfo = p->flowinfo;
1213         u->link = p->link;
1214         u->i_key = p->i_key;
1215         u->o_key = p->o_key;
1216         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1217         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1218         memcpy(u->name, p->name, sizeof(u->name));
1219 }
1220
1221 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1222         struct ifreq *ifr, int cmd)
1223 {
1224         int err = 0;
1225         struct ip6_tnl_parm2 p;
1226         struct __ip6_tnl_parm p1;
1227         struct ip6_tnl *t = netdev_priv(dev);
1228         struct net *net = t->net;
1229         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1230
1231         memset(&p1, 0, sizeof(p1));
1232
1233         switch (cmd) {
1234         case SIOCGETTUNNEL:
1235                 if (dev == ign->fb_tunnel_dev) {
1236                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1237                                 err = -EFAULT;
1238                                 break;
1239                         }
1240                         ip6gre_tnl_parm_from_user(&p1, &p);
1241                         t = ip6gre_tunnel_locate(net, &p1, 0);
1242                         if (!t)
1243                                 t = netdev_priv(dev);
1244                 }
1245                 memset(&p, 0, sizeof(p));
1246                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1247                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1248                         err = -EFAULT;
1249                 break;
1250
1251         case SIOCADDTUNNEL:
1252         case SIOCCHGTUNNEL:
1253                 err = -EPERM;
1254                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1255                         goto done;
1256
1257                 err = -EFAULT;
1258                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1259                         goto done;
1260
1261                 err = -EINVAL;
1262                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1263                         goto done;
1264
1265                 if (!(p.i_flags&GRE_KEY))
1266                         p.i_key = 0;
1267                 if (!(p.o_flags&GRE_KEY))
1268                         p.o_key = 0;
1269
1270                 ip6gre_tnl_parm_from_user(&p1, &p);
1271                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1272
1273                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1274                         if (t) {
1275                                 if (t->dev != dev) {
1276                                         err = -EEXIST;
1277                                         break;
1278                                 }
1279                         } else {
1280                                 t = netdev_priv(dev);
1281
1282                                 ip6gre_tunnel_unlink(ign, t);
1283                                 synchronize_net();
1284                                 ip6gre_tnl_change(t, &p1, 1);
1285                                 ip6gre_tunnel_link(ign, t);
1286                                 netdev_state_change(dev);
1287                         }
1288                 }
1289
1290                 if (t) {
1291                         err = 0;
1292
1293                         memset(&p, 0, sizeof(p));
1294                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1295                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1296                                 err = -EFAULT;
1297                 } else
1298                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1299                 break;
1300
1301         case SIOCDELTUNNEL:
1302                 err = -EPERM;
1303                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1304                         goto done;
1305
1306                 if (dev == ign->fb_tunnel_dev) {
1307                         err = -EFAULT;
1308                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1309                                 goto done;
1310                         err = -ENOENT;
1311                         ip6gre_tnl_parm_from_user(&p1, &p);
1312                         t = ip6gre_tunnel_locate(net, &p1, 0);
1313                         if (!t)
1314                                 goto done;
1315                         err = -EPERM;
1316                         if (t == netdev_priv(ign->fb_tunnel_dev))
1317                                 goto done;
1318                         dev = t->dev;
1319                 }
1320                 unregister_netdevice(dev);
1321                 err = 0;
1322                 break;
1323
1324         default:
1325                 err = -EINVAL;
1326         }
1327
1328 done:
1329         return err;
1330 }
1331
1332 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1333                          unsigned short type, const void *daddr,
1334                          const void *saddr, unsigned int len)
1335 {
1336         struct ip6_tnl *t = netdev_priv(dev);
1337         struct ipv6hdr *ipv6h;
1338         __be16 *p;
1339
1340         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1341         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1342                                                   t->fl.u.ip6.flowlabel,
1343                                                   true, &t->fl.u.ip6));
1344         ipv6h->hop_limit = t->parms.hop_limit;
1345         ipv6h->nexthdr = NEXTHDR_GRE;
1346         ipv6h->saddr = t->parms.laddr;
1347         ipv6h->daddr = t->parms.raddr;
1348
1349         p = (__be16 *)(ipv6h + 1);
1350         p[0] = t->parms.o_flags;
1351         p[1] = htons(type);
1352
1353         /*
1354          *      Set the source hardware address.
1355          */
1356
1357         if (saddr)
1358                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1359         if (daddr)
1360                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1361         if (!ipv6_addr_any(&ipv6h->daddr))
1362                 return t->hlen;
1363
1364         return -t->hlen;
1365 }
1366
1367 static const struct header_ops ip6gre_header_ops = {
1368         .create = ip6gre_header,
1369 };
1370
1371 static const struct net_device_ops ip6gre_netdev_ops = {
1372         .ndo_init               = ip6gre_tunnel_init,
1373         .ndo_uninit             = ip6gre_tunnel_uninit,
1374         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1375         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1376         .ndo_change_mtu         = ip6_tnl_change_mtu,
1377         .ndo_get_stats64        = ip_tunnel_get_stats64,
1378         .ndo_get_iflink         = ip6_tnl_get_iflink,
1379 };
1380
1381 static void ip6gre_dev_free(struct net_device *dev)
1382 {
1383         struct ip6_tnl *t = netdev_priv(dev);
1384
1385         gro_cells_destroy(&t->gro_cells);
1386         dst_cache_destroy(&t->dst_cache);
1387         free_percpu(dev->tstats);
1388 }
1389
1390 static void ip6gre_tunnel_setup(struct net_device *dev)
1391 {
1392         dev->netdev_ops = &ip6gre_netdev_ops;
1393         dev->needs_free_netdev = true;
1394         dev->priv_destructor = ip6gre_dev_free;
1395
1396         dev->type = ARPHRD_IP6GRE;
1397
1398         dev->flags |= IFF_NOARP;
1399         dev->addr_len = sizeof(struct in6_addr);
1400         netif_keep_dst(dev);
1401         /* This perm addr will be used as interface identifier by IPv6 */
1402         dev->addr_assign_type = NET_ADDR_RANDOM;
1403         eth_random_addr(dev->perm_addr);
1404 }
1405
1406 #define GRE6_FEATURES (NETIF_F_SG |             \
1407                        NETIF_F_FRAGLIST |       \
1408                        NETIF_F_HIGHDMA |        \
1409                        NETIF_F_HW_CSUM)
1410
1411 static void ip6gre_tnl_init_features(struct net_device *dev)
1412 {
1413         struct ip6_tnl *nt = netdev_priv(dev);
1414
1415         dev->features           |= GRE6_FEATURES;
1416         dev->hw_features        |= GRE6_FEATURES;
1417
1418         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1419                 /* TCP offload with GRE SEQ is not supported, nor
1420                  * can we support 2 levels of outer headers requiring
1421                  * an update.
1422                  */
1423                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1424                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1425                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1426                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1427                 }
1428
1429                 /* Can use a lockless transmit, unless we generate
1430                  * output sequences
1431                  */
1432                 dev->features |= NETIF_F_LLTX;
1433         }
1434 }
1435
1436 static int ip6gre_tunnel_init_common(struct net_device *dev)
1437 {
1438         struct ip6_tnl *tunnel;
1439         int ret;
1440         int t_hlen;
1441
1442         tunnel = netdev_priv(dev);
1443
1444         tunnel->dev = dev;
1445         tunnel->net = dev_net(dev);
1446         strcpy(tunnel->parms.name, dev->name);
1447
1448         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1449         if (!dev->tstats)
1450                 return -ENOMEM;
1451
1452         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1453         if (ret)
1454                 goto cleanup_alloc_pcpu_stats;
1455
1456         ret = gro_cells_init(&tunnel->gro_cells, dev);
1457         if (ret)
1458                 goto cleanup_dst_cache_init;
1459
1460         t_hlen = ip6gre_calc_hlen(tunnel);
1461         dev->mtu = ETH_DATA_LEN - t_hlen;
1462         if (dev->type == ARPHRD_ETHER)
1463                 dev->mtu -= ETH_HLEN;
1464         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1465                 dev->mtu -= 8;
1466
1467         if (tunnel->parms.collect_md) {
1468                 dev->features |= NETIF_F_NETNS_LOCAL;
1469                 netif_keep_dst(dev);
1470         }
1471         ip6gre_tnl_init_features(dev);
1472
1473         return 0;
1474
1475 cleanup_dst_cache_init:
1476         dst_cache_destroy(&tunnel->dst_cache);
1477 cleanup_alloc_pcpu_stats:
1478         free_percpu(dev->tstats);
1479         dev->tstats = NULL;
1480         return ret;
1481 }
1482
1483 static int ip6gre_tunnel_init(struct net_device *dev)
1484 {
1485         struct ip6_tnl *tunnel;
1486         int ret;
1487
1488         ret = ip6gre_tunnel_init_common(dev);
1489         if (ret)
1490                 return ret;
1491
1492         tunnel = netdev_priv(dev);
1493
1494         if (tunnel->parms.collect_md)
1495                 return 0;
1496
1497         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1498         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1499
1500         if (ipv6_addr_any(&tunnel->parms.raddr))
1501                 dev->header_ops = &ip6gre_header_ops;
1502
1503         return 0;
1504 }
1505
1506 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1507 {
1508         struct ip6_tnl *tunnel = netdev_priv(dev);
1509
1510         tunnel->dev = dev;
1511         tunnel->net = dev_net(dev);
1512         strcpy(tunnel->parms.name, dev->name);
1513
1514         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1515
1516         dev_hold(dev);
1517 }
1518
1519 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1520         .handler     = gre_rcv,
1521         .err_handler = ip6gre_err,
1522         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1523 };
1524
1525 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1526 {
1527         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1528         struct net_device *dev, *aux;
1529         int prio;
1530
1531         for_each_netdev_safe(net, dev, aux)
1532                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1533                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1534                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1535                         unregister_netdevice_queue(dev, head);
1536
1537         for (prio = 0; prio < 4; prio++) {
1538                 int h;
1539                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1540                         struct ip6_tnl *t;
1541
1542                         t = rtnl_dereference(ign->tunnels[prio][h]);
1543
1544                         while (t) {
1545                                 /* If dev is in the same netns, it has already
1546                                  * been added to the list by the previous loop.
1547                                  */
1548                                 if (!net_eq(dev_net(t->dev), net))
1549                                         unregister_netdevice_queue(t->dev,
1550                                                                    head);
1551                                 t = rtnl_dereference(t->next);
1552                         }
1553                 }
1554         }
1555 }
1556
1557 static int __net_init ip6gre_init_net(struct net *net)
1558 {
1559         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1560         int err;
1561
1562         if (!net_has_fallback_tunnels(net))
1563                 return 0;
1564         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1565                                           NET_NAME_UNKNOWN,
1566                                           ip6gre_tunnel_setup);
1567         if (!ign->fb_tunnel_dev) {
1568                 err = -ENOMEM;
1569                 goto err_alloc_dev;
1570         }
1571         dev_net_set(ign->fb_tunnel_dev, net);
1572         /* FB netdevice is special: we have one, and only one per netns.
1573          * Allowing to move it to another netns is clearly unsafe.
1574          */
1575         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1576
1577
1578         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1579         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1580
1581         err = register_netdev(ign->fb_tunnel_dev);
1582         if (err)
1583                 goto err_reg_dev;
1584
1585         rcu_assign_pointer(ign->tunnels_wc[0],
1586                            netdev_priv(ign->fb_tunnel_dev));
1587         return 0;
1588
1589 err_reg_dev:
1590         free_netdev(ign->fb_tunnel_dev);
1591 err_alloc_dev:
1592         return err;
1593 }
1594
1595 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1596 {
1597         struct net *net;
1598         LIST_HEAD(list);
1599
1600         rtnl_lock();
1601         list_for_each_entry(net, net_list, exit_list)
1602                 ip6gre_destroy_tunnels(net, &list);
1603         unregister_netdevice_many(&list);
1604         rtnl_unlock();
1605 }
1606
1607 static struct pernet_operations ip6gre_net_ops = {
1608         .init = ip6gre_init_net,
1609         .exit_batch = ip6gre_exit_batch_net,
1610         .id   = &ip6gre_net_id,
1611         .size = sizeof(struct ip6gre_net),
1612 };
1613
1614 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1615                                   struct netlink_ext_ack *extack)
1616 {
1617         __be16 flags;
1618
1619         if (!data)
1620                 return 0;
1621
1622         flags = 0;
1623         if (data[IFLA_GRE_IFLAGS])
1624                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1625         if (data[IFLA_GRE_OFLAGS])
1626                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1627         if (flags & (GRE_VERSION|GRE_ROUTING))
1628                 return -EINVAL;
1629
1630         return 0;
1631 }
1632
1633 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1634                                struct netlink_ext_ack *extack)
1635 {
1636         struct in6_addr daddr;
1637
1638         if (tb[IFLA_ADDRESS]) {
1639                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1640                         return -EINVAL;
1641                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1642                         return -EADDRNOTAVAIL;
1643         }
1644
1645         if (!data)
1646                 goto out;
1647
1648         if (data[IFLA_GRE_REMOTE]) {
1649                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1650                 if (ipv6_addr_any(&daddr))
1651                         return -EINVAL;
1652         }
1653
1654 out:
1655         return ip6gre_tunnel_validate(tb, data, extack);
1656 }
1657
1658 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1659                                   struct netlink_ext_ack *extack)
1660 {
1661         __be16 flags = 0;
1662         int ret, ver = 0;
1663
1664         if (!data)
1665                 return 0;
1666
1667         ret = ip6gre_tap_validate(tb, data, extack);
1668         if (ret)
1669                 return ret;
1670
1671         /* ERSPAN should only have GRE sequence and key flag */
1672         if (data[IFLA_GRE_OFLAGS])
1673                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1674         if (data[IFLA_GRE_IFLAGS])
1675                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1676         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1677             flags != (GRE_SEQ | GRE_KEY))
1678                 return -EINVAL;
1679
1680         /* ERSPAN Session ID only has 10-bit. Since we reuse
1681          * 32-bit key field as ID, check it's range.
1682          */
1683         if (data[IFLA_GRE_IKEY] &&
1684             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1685                 return -EINVAL;
1686
1687         if (data[IFLA_GRE_OKEY] &&
1688             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1689                 return -EINVAL;
1690
1691         if (data[IFLA_GRE_ERSPAN_VER]) {
1692                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1693                 if (ver != 1 && ver != 2)
1694                         return -EINVAL;
1695         }
1696
1697         if (ver == 1) {
1698                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1699                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1700
1701                         if (index & ~INDEX_MASK)
1702                                 return -EINVAL;
1703                 }
1704         } else if (ver == 2) {
1705                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1706                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1707
1708                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1709                                 return -EINVAL;
1710                 }
1711
1712                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1713                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1714
1715                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1716                                 return -EINVAL;
1717                 }
1718         }
1719
1720         return 0;
1721 }
1722
1723 static void ip6erspan_set_version(struct nlattr *data[],
1724                                   struct __ip6_tnl_parm *parms)
1725 {
1726         if (!data)
1727                 return;
1728
1729         parms->erspan_ver = 1;
1730         if (data[IFLA_GRE_ERSPAN_VER])
1731                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1732
1733         if (parms->erspan_ver == 1) {
1734                 if (data[IFLA_GRE_ERSPAN_INDEX])
1735                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1736         } else if (parms->erspan_ver == 2) {
1737                 if (data[IFLA_GRE_ERSPAN_DIR])
1738                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1739                 if (data[IFLA_GRE_ERSPAN_HWID])
1740                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1741         }
1742 }
1743
1744 static void ip6gre_netlink_parms(struct nlattr *data[],
1745                                 struct __ip6_tnl_parm *parms)
1746 {
1747         memset(parms, 0, sizeof(*parms));
1748
1749         if (!data)
1750                 return;
1751
1752         if (data[IFLA_GRE_LINK])
1753                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1754
1755         if (data[IFLA_GRE_IFLAGS])
1756                 parms->i_flags = gre_flags_to_tnl_flags(
1757                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1758
1759         if (data[IFLA_GRE_OFLAGS])
1760                 parms->o_flags = gre_flags_to_tnl_flags(
1761                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1762
1763         if (data[IFLA_GRE_IKEY])
1764                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1765
1766         if (data[IFLA_GRE_OKEY])
1767                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1768
1769         if (data[IFLA_GRE_LOCAL])
1770                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1771
1772         if (data[IFLA_GRE_REMOTE])
1773                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1774
1775         if (data[IFLA_GRE_TTL])
1776                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1777
1778         if (data[IFLA_GRE_ENCAP_LIMIT])
1779                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1780
1781         if (data[IFLA_GRE_FLOWINFO])
1782                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1783
1784         if (data[IFLA_GRE_FLAGS])
1785                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1786
1787         if (data[IFLA_GRE_FWMARK])
1788                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1789
1790         if (data[IFLA_GRE_COLLECT_METADATA])
1791                 parms->collect_md = true;
1792 }
1793
1794 static int ip6gre_tap_init(struct net_device *dev)
1795 {
1796         int ret;
1797
1798         ret = ip6gre_tunnel_init_common(dev);
1799         if (ret)
1800                 return ret;
1801
1802         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1803
1804         return 0;
1805 }
1806
1807 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1808         .ndo_init = ip6gre_tap_init,
1809         .ndo_uninit = ip6gre_tunnel_uninit,
1810         .ndo_start_xmit = ip6gre_tunnel_xmit,
1811         .ndo_set_mac_address = eth_mac_addr,
1812         .ndo_validate_addr = eth_validate_addr,
1813         .ndo_change_mtu = ip6_tnl_change_mtu,
1814         .ndo_get_stats64 = ip_tunnel_get_stats64,
1815         .ndo_get_iflink = ip6_tnl_get_iflink,
1816 };
1817
1818 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1819 {
1820         int t_hlen;
1821
1822         tunnel->tun_hlen = 8;
1823         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1824                        erspan_hdr_len(tunnel->parms.erspan_ver);
1825
1826         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1827         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1828         return t_hlen;
1829 }
1830
1831 static int ip6erspan_tap_init(struct net_device *dev)
1832 {
1833         struct ip6_tnl *tunnel;
1834         int t_hlen;
1835         int ret;
1836
1837         tunnel = netdev_priv(dev);
1838
1839         tunnel->dev = dev;
1840         tunnel->net = dev_net(dev);
1841         strcpy(tunnel->parms.name, dev->name);
1842
1843         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1844         if (!dev->tstats)
1845                 return -ENOMEM;
1846
1847         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1848         if (ret)
1849                 goto cleanup_alloc_pcpu_stats;
1850
1851         ret = gro_cells_init(&tunnel->gro_cells, dev);
1852         if (ret)
1853                 goto cleanup_dst_cache_init;
1854
1855         t_hlen = ip6erspan_calc_hlen(tunnel);
1856         dev->mtu = ETH_DATA_LEN - t_hlen;
1857         if (dev->type == ARPHRD_ETHER)
1858                 dev->mtu -= ETH_HLEN;
1859         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1860                 dev->mtu -= 8;
1861
1862         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1863         ip6erspan_tnl_link_config(tunnel, 1);
1864
1865         return 0;
1866
1867 cleanup_dst_cache_init:
1868         dst_cache_destroy(&tunnel->dst_cache);
1869 cleanup_alloc_pcpu_stats:
1870         free_percpu(dev->tstats);
1871         dev->tstats = NULL;
1872         return ret;
1873 }
1874
1875 static const struct net_device_ops ip6erspan_netdev_ops = {
1876         .ndo_init =             ip6erspan_tap_init,
1877         .ndo_uninit =           ip6erspan_tunnel_uninit,
1878         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1879         .ndo_set_mac_address =  eth_mac_addr,
1880         .ndo_validate_addr =    eth_validate_addr,
1881         .ndo_change_mtu =       ip6_tnl_change_mtu,
1882         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1883         .ndo_get_iflink =       ip6_tnl_get_iflink,
1884 };
1885
1886 static void ip6gre_tap_setup(struct net_device *dev)
1887 {
1888
1889         ether_setup(dev);
1890
1891         dev->max_mtu = 0;
1892         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1893         dev->needs_free_netdev = true;
1894         dev->priv_destructor = ip6gre_dev_free;
1895
1896         dev->features |= NETIF_F_NETNS_LOCAL;
1897         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1898         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1899         netif_keep_dst(dev);
1900 }
1901
1902 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1903                                        struct ip_tunnel_encap *ipencap)
1904 {
1905         bool ret = false;
1906
1907         memset(ipencap, 0, sizeof(*ipencap));
1908
1909         if (!data)
1910                 return ret;
1911
1912         if (data[IFLA_GRE_ENCAP_TYPE]) {
1913                 ret = true;
1914                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1915         }
1916
1917         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1918                 ret = true;
1919                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1920         }
1921
1922         if (data[IFLA_GRE_ENCAP_SPORT]) {
1923                 ret = true;
1924                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1925         }
1926
1927         if (data[IFLA_GRE_ENCAP_DPORT]) {
1928                 ret = true;
1929                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1930         }
1931
1932         return ret;
1933 }
1934
1935 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1936                                  struct nlattr *tb[], struct nlattr *data[],
1937                                  struct netlink_ext_ack *extack)
1938 {
1939         struct ip6_tnl *nt;
1940         struct ip_tunnel_encap ipencap;
1941         int err;
1942
1943         nt = netdev_priv(dev);
1944
1945         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1946                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1947
1948                 if (err < 0)
1949                         return err;
1950         }
1951
1952         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1953                 eth_hw_addr_random(dev);
1954
1955         nt->dev = dev;
1956         nt->net = dev_net(dev);
1957
1958         err = register_netdevice(dev);
1959         if (err)
1960                 goto out;
1961
1962         if (tb[IFLA_MTU])
1963                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1964
1965         dev_hold(dev);
1966
1967 out:
1968         return err;
1969 }
1970
1971 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1972                           struct nlattr *tb[], struct nlattr *data[],
1973                           struct netlink_ext_ack *extack)
1974 {
1975         struct ip6_tnl *nt = netdev_priv(dev);
1976         struct net *net = dev_net(dev);
1977         struct ip6gre_net *ign;
1978         int err;
1979
1980         ip6gre_netlink_parms(data, &nt->parms);
1981         ign = net_generic(net, ip6gre_net_id);
1982
1983         if (nt->parms.collect_md) {
1984                 if (rtnl_dereference(ign->collect_md_tun))
1985                         return -EEXIST;
1986         } else {
1987                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1988                         return -EEXIST;
1989         }
1990
1991         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1992         if (!err) {
1993                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1994                 ip6gre_tunnel_link_md(ign, nt);
1995                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1996         }
1997         return err;
1998 }
1999
2000 static struct ip6_tnl *
2001 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2002                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2003                          struct netlink_ext_ack *extack)
2004 {
2005         struct ip6_tnl *t, *nt = netdev_priv(dev);
2006         struct net *net = nt->net;
2007         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2008         struct ip_tunnel_encap ipencap;
2009
2010         if (dev == ign->fb_tunnel_dev)
2011                 return ERR_PTR(-EINVAL);
2012
2013         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2014                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2015
2016                 if (err < 0)
2017                         return ERR_PTR(err);
2018         }
2019
2020         ip6gre_netlink_parms(data, p_p);
2021
2022         t = ip6gre_tunnel_locate(net, p_p, 0);
2023
2024         if (t) {
2025                 if (t->dev != dev)
2026                         return ERR_PTR(-EEXIST);
2027         } else {
2028                 t = nt;
2029         }
2030
2031         return t;
2032 }
2033
2034 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2035                              struct nlattr *data[],
2036                              struct netlink_ext_ack *extack)
2037 {
2038         struct ip6_tnl *t = netdev_priv(dev);
2039         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2040         struct __ip6_tnl_parm p;
2041
2042         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2043         if (IS_ERR(t))
2044                 return PTR_ERR(t);
2045
2046         ip6gre_tunnel_unlink_md(ign, t);
2047         ip6gre_tunnel_unlink(ign, t);
2048         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2049         ip6gre_tunnel_link_md(ign, t);
2050         ip6gre_tunnel_link(ign, t);
2051         return 0;
2052 }
2053
2054 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2055 {
2056         struct net *net = dev_net(dev);
2057         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2058
2059         if (dev != ign->fb_tunnel_dev)
2060                 unregister_netdevice_queue(dev, head);
2061 }
2062
2063 static size_t ip6gre_get_size(const struct net_device *dev)
2064 {
2065         return
2066                 /* IFLA_GRE_LINK */
2067                 nla_total_size(4) +
2068                 /* IFLA_GRE_IFLAGS */
2069                 nla_total_size(2) +
2070                 /* IFLA_GRE_OFLAGS */
2071                 nla_total_size(2) +
2072                 /* IFLA_GRE_IKEY */
2073                 nla_total_size(4) +
2074                 /* IFLA_GRE_OKEY */
2075                 nla_total_size(4) +
2076                 /* IFLA_GRE_LOCAL */
2077                 nla_total_size(sizeof(struct in6_addr)) +
2078                 /* IFLA_GRE_REMOTE */
2079                 nla_total_size(sizeof(struct in6_addr)) +
2080                 /* IFLA_GRE_TTL */
2081                 nla_total_size(1) +
2082                 /* IFLA_GRE_ENCAP_LIMIT */
2083                 nla_total_size(1) +
2084                 /* IFLA_GRE_FLOWINFO */
2085                 nla_total_size(4) +
2086                 /* IFLA_GRE_FLAGS */
2087                 nla_total_size(4) +
2088                 /* IFLA_GRE_ENCAP_TYPE */
2089                 nla_total_size(2) +
2090                 /* IFLA_GRE_ENCAP_FLAGS */
2091                 nla_total_size(2) +
2092                 /* IFLA_GRE_ENCAP_SPORT */
2093                 nla_total_size(2) +
2094                 /* IFLA_GRE_ENCAP_DPORT */
2095                 nla_total_size(2) +
2096                 /* IFLA_GRE_COLLECT_METADATA */
2097                 nla_total_size(0) +
2098                 /* IFLA_GRE_FWMARK */
2099                 nla_total_size(4) +
2100                 /* IFLA_GRE_ERSPAN_INDEX */
2101                 nla_total_size(4) +
2102                 0;
2103 }
2104
2105 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2106 {
2107         struct ip6_tnl *t = netdev_priv(dev);
2108         struct __ip6_tnl_parm *p = &t->parms;
2109         __be16 o_flags = p->o_flags;
2110
2111         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2112                 if (!p->collect_md)
2113                         o_flags |= TUNNEL_KEY;
2114
2115                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2116                         goto nla_put_failure;
2117
2118                 if (p->erspan_ver == 1) {
2119                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2120                                 goto nla_put_failure;
2121                 } else {
2122                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2123                                 goto nla_put_failure;
2124                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2125                                 goto nla_put_failure;
2126                 }
2127         }
2128
2129         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2130             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2131                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2132             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2133                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2134             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2135             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2136             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2137             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2138             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2139             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2140             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2141             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2142             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2143                 goto nla_put_failure;
2144
2145         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2146                         t->encap.type) ||
2147             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2148                          t->encap.sport) ||
2149             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2150                          t->encap.dport) ||
2151             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2152                         t->encap.flags))
2153                 goto nla_put_failure;
2154
2155         if (p->collect_md) {
2156                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2157                         goto nla_put_failure;
2158         }
2159
2160         return 0;
2161
2162 nla_put_failure:
2163         return -EMSGSIZE;
2164 }
2165
2166 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2167         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2168         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2169         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2170         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2171         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2172         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2173         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2174         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2175         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2176         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2177         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2178         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2179         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2180         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2181         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2182         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2183         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2184         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2185         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2186         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2187         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2188 };
2189
2190 static void ip6erspan_tap_setup(struct net_device *dev)
2191 {
2192         ether_setup(dev);
2193
2194         dev->netdev_ops = &ip6erspan_netdev_ops;
2195         dev->needs_free_netdev = true;
2196         dev->priv_destructor = ip6gre_dev_free;
2197
2198         dev->features |= NETIF_F_NETNS_LOCAL;
2199         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2200         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2201         netif_keep_dst(dev);
2202 }
2203
2204 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2205                              struct nlattr *tb[], struct nlattr *data[],
2206                              struct netlink_ext_ack *extack)
2207 {
2208         struct ip6_tnl *nt = netdev_priv(dev);
2209         struct net *net = dev_net(dev);
2210         struct ip6gre_net *ign;
2211         int err;
2212
2213         ip6gre_netlink_parms(data, &nt->parms);
2214         ip6erspan_set_version(data, &nt->parms);
2215         ign = net_generic(net, ip6gre_net_id);
2216
2217         if (nt->parms.collect_md) {
2218                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2219                         return -EEXIST;
2220         } else {
2221                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2222                         return -EEXIST;
2223         }
2224
2225         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2226         if (!err) {
2227                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2228                 ip6erspan_tunnel_link_md(ign, nt);
2229                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2230         }
2231         return err;
2232 }
2233
2234 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2235 {
2236         ip6gre_tnl_link_config_common(t);
2237         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2238 }
2239
2240 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2241                                 const struct __ip6_tnl_parm *p, int set_mtu)
2242 {
2243         ip6gre_tnl_copy_tnl_parm(t, p);
2244         ip6erspan_tnl_link_config(t, set_mtu);
2245         return 0;
2246 }
2247
2248 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2249                                 struct nlattr *data[],
2250                                 struct netlink_ext_ack *extack)
2251 {
2252         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2253         struct __ip6_tnl_parm p;
2254         struct ip6_tnl *t;
2255
2256         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2257         if (IS_ERR(t))
2258                 return PTR_ERR(t);
2259
2260         ip6erspan_set_version(data, &p);
2261         ip6gre_tunnel_unlink_md(ign, t);
2262         ip6gre_tunnel_unlink(ign, t);
2263         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2264         ip6erspan_tunnel_link_md(ign, t);
2265         ip6gre_tunnel_link(ign, t);
2266         return 0;
2267 }
2268
2269 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2270         .kind           = "ip6gre",
2271         .maxtype        = IFLA_GRE_MAX,
2272         .policy         = ip6gre_policy,
2273         .priv_size      = sizeof(struct ip6_tnl),
2274         .setup          = ip6gre_tunnel_setup,
2275         .validate       = ip6gre_tunnel_validate,
2276         .newlink        = ip6gre_newlink,
2277         .changelink     = ip6gre_changelink,
2278         .dellink        = ip6gre_dellink,
2279         .get_size       = ip6gre_get_size,
2280         .fill_info      = ip6gre_fill_info,
2281         .get_link_net   = ip6_tnl_get_link_net,
2282 };
2283
2284 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2285         .kind           = "ip6gretap",
2286         .maxtype        = IFLA_GRE_MAX,
2287         .policy         = ip6gre_policy,
2288         .priv_size      = sizeof(struct ip6_tnl),
2289         .setup          = ip6gre_tap_setup,
2290         .validate       = ip6gre_tap_validate,
2291         .newlink        = ip6gre_newlink,
2292         .changelink     = ip6gre_changelink,
2293         .get_size       = ip6gre_get_size,
2294         .fill_info      = ip6gre_fill_info,
2295         .get_link_net   = ip6_tnl_get_link_net,
2296 };
2297
2298 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2299         .kind           = "ip6erspan",
2300         .maxtype        = IFLA_GRE_MAX,
2301         .policy         = ip6gre_policy,
2302         .priv_size      = sizeof(struct ip6_tnl),
2303         .setup          = ip6erspan_tap_setup,
2304         .validate       = ip6erspan_tap_validate,
2305         .newlink        = ip6erspan_newlink,
2306         .changelink     = ip6erspan_changelink,
2307         .get_size       = ip6gre_get_size,
2308         .fill_info      = ip6gre_fill_info,
2309         .get_link_net   = ip6_tnl_get_link_net,
2310 };
2311
2312 /*
2313  *      And now the modules code and kernel interface.
2314  */
2315
2316 static int __init ip6gre_init(void)
2317 {
2318         int err;
2319
2320         pr_info("GRE over IPv6 tunneling driver\n");
2321
2322         err = register_pernet_device(&ip6gre_net_ops);
2323         if (err < 0)
2324                 return err;
2325
2326         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2327         if (err < 0) {
2328                 pr_info("%s: can't add protocol\n", __func__);
2329                 goto add_proto_failed;
2330         }
2331
2332         err = rtnl_link_register(&ip6gre_link_ops);
2333         if (err < 0)
2334                 goto rtnl_link_failed;
2335
2336         err = rtnl_link_register(&ip6gre_tap_ops);
2337         if (err < 0)
2338                 goto tap_ops_failed;
2339
2340         err = rtnl_link_register(&ip6erspan_tap_ops);
2341         if (err < 0)
2342                 goto erspan_link_failed;
2343
2344 out:
2345         return err;
2346
2347 erspan_link_failed:
2348         rtnl_link_unregister(&ip6gre_tap_ops);
2349 tap_ops_failed:
2350         rtnl_link_unregister(&ip6gre_link_ops);
2351 rtnl_link_failed:
2352         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2353 add_proto_failed:
2354         unregister_pernet_device(&ip6gre_net_ops);
2355         goto out;
2356 }
2357
2358 static void __exit ip6gre_fini(void)
2359 {
2360         rtnl_link_unregister(&ip6gre_tap_ops);
2361         rtnl_link_unregister(&ip6gre_link_ops);
2362         rtnl_link_unregister(&ip6erspan_tap_ops);
2363         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2364         unregister_pernet_device(&ip6gre_net_ops);
2365 }
2366
2367 module_init(ip6gre_init);
2368 module_exit(ip6gre_fini);
2369 MODULE_LICENSE("GPL");
2370 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2371 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2372 MODULE_ALIAS_RTNL_LINK("ip6gre");
2373 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2374 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2375 MODULE_ALIAS_NETDEV("ip6gre0");