GNU Linux-libre 5.19-rc6-gnu
[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         ip6gre_tunnel_link(ign, nt);
386         return nt;
387
388 failed_free:
389         free_netdev(dev);
390         return NULL;
391 }
392
393 static void ip6erspan_tunnel_uninit(struct net_device *dev)
394 {
395         struct ip6_tnl *t = netdev_priv(dev);
396         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
397
398         ip6erspan_tunnel_unlink_md(ign, t);
399         ip6gre_tunnel_unlink(ign, t);
400         dst_cache_reset(&t->dst_cache);
401         dev_put_track(dev, &t->dev_tracker);
402 }
403
404 static void ip6gre_tunnel_uninit(struct net_device *dev)
405 {
406         struct ip6_tnl *t = netdev_priv(dev);
407         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
408
409         ip6gre_tunnel_unlink_md(ign, t);
410         ip6gre_tunnel_unlink(ign, t);
411         if (ign->fb_tunnel_dev == dev)
412                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
413         dst_cache_reset(&t->dst_cache);
414         dev_put_track(dev, &t->dev_tracker);
415 }
416
417
418 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
419                        u8 type, u8 code, int offset, __be32 info)
420 {
421         struct net *net = dev_net(skb->dev);
422         const struct ipv6hdr *ipv6h;
423         struct tnl_ptk_info tpi;
424         struct ip6_tnl *t;
425
426         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
427                              offset) < 0)
428                 return -EINVAL;
429
430         ipv6h = (const struct ipv6hdr *)skb->data;
431         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
432                                  tpi.key, tpi.proto);
433         if (!t)
434                 return -ENOENT;
435
436         switch (type) {
437         case ICMPV6_DEST_UNREACH:
438                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
439                                     t->parms.name);
440                 if (code != ICMPV6_PORT_UNREACH)
441                         break;
442                 return 0;
443         case ICMPV6_TIME_EXCEED:
444                 if (code == ICMPV6_EXC_HOPLIMIT) {
445                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
446                                             t->parms.name);
447                         break;
448                 }
449                 return 0;
450         case ICMPV6_PARAMPROB: {
451                 struct ipv6_tlv_tnl_enc_lim *tel;
452                 __u32 teli;
453
454                 teli = 0;
455                 if (code == ICMPV6_HDR_FIELD)
456                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
457
458                 if (teli && teli == be32_to_cpu(info) - 2) {
459                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
460                         if (tel->encap_limit == 0) {
461                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
462                                                     t->parms.name);
463                         }
464                 } else {
465                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
466                                             t->parms.name);
467                 }
468                 return 0;
469         }
470         case ICMPV6_PKT_TOOBIG:
471                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
472                 return 0;
473         case NDISC_REDIRECT:
474                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
475                              sock_net_uid(net, NULL));
476                 return 0;
477         }
478
479         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
480                 t->err_count++;
481         else
482                 t->err_count = 1;
483         t->err_time = jiffies;
484
485         return 0;
486 }
487
488 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
489 {
490         const struct ipv6hdr *ipv6h;
491         struct ip6_tnl *tunnel;
492
493         ipv6h = ipv6_hdr(skb);
494         tunnel = ip6gre_tunnel_lookup(skb->dev,
495                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
496                                       tpi->proto);
497         if (tunnel) {
498                 if (tunnel->parms.collect_md) {
499                         struct metadata_dst *tun_dst;
500                         __be64 tun_id;
501                         __be16 flags;
502
503                         flags = tpi->flags;
504                         tun_id = key32_to_tunnel_id(tpi->key);
505
506                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
507                         if (!tun_dst)
508                                 return PACKET_REJECT;
509
510                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
511                 } else {
512                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
513                 }
514
515                 return PACKET_RCVD;
516         }
517
518         return PACKET_REJECT;
519 }
520
521 static int ip6erspan_rcv(struct sk_buff *skb,
522                          struct tnl_ptk_info *tpi,
523                          int gre_hdr_len)
524 {
525         struct erspan_base_hdr *ershdr;
526         const struct ipv6hdr *ipv6h;
527         struct erspan_md2 *md2;
528         struct ip6_tnl *tunnel;
529         u8 ver;
530
531         ipv6h = ipv6_hdr(skb);
532         ershdr = (struct erspan_base_hdr *)skb->data;
533         ver = ershdr->ver;
534
535         tunnel = ip6gre_tunnel_lookup(skb->dev,
536                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
537                                       tpi->proto);
538         if (tunnel) {
539                 int len = erspan_hdr_len(ver);
540
541                 if (unlikely(!pskb_may_pull(skb, len)))
542                         return PACKET_REJECT;
543
544                 if (__iptunnel_pull_header(skb, len,
545                                            htons(ETH_P_TEB),
546                                            false, false) < 0)
547                         return PACKET_REJECT;
548
549                 if (tunnel->parms.collect_md) {
550                         struct erspan_metadata *pkt_md, *md;
551                         struct metadata_dst *tun_dst;
552                         struct ip_tunnel_info *info;
553                         unsigned char *gh;
554                         __be64 tun_id;
555                         __be16 flags;
556
557                         tpi->flags |= TUNNEL_KEY;
558                         flags = tpi->flags;
559                         tun_id = key32_to_tunnel_id(tpi->key);
560
561                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
562                                                   sizeof(*md));
563                         if (!tun_dst)
564                                 return PACKET_REJECT;
565
566                         /* skb can be uncloned in __iptunnel_pull_header, so
567                          * old pkt_md is no longer valid and we need to reset
568                          * it
569                          */
570                         gh = skb_network_header(skb) +
571                              skb_network_header_len(skb);
572                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
573                                                             sizeof(*ershdr));
574                         info = &tun_dst->u.tun_info;
575                         md = ip_tunnel_info_opts(info);
576                         md->version = ver;
577                         md2 = &md->u.md2;
578                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
579                                                        ERSPAN_V2_MDSIZE);
580                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
581                         info->options_len = sizeof(*md);
582
583                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
584
585                 } else {
586                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
587                 }
588
589                 return PACKET_RCVD;
590         }
591
592         return PACKET_REJECT;
593 }
594
595 static int gre_rcv(struct sk_buff *skb)
596 {
597         struct tnl_ptk_info tpi;
598         bool csum_err = false;
599         int hdr_len;
600
601         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
602         if (hdr_len < 0)
603                 goto drop;
604
605         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
606                 goto drop;
607
608         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
609                      tpi.proto == htons(ETH_P_ERSPAN2))) {
610                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
611                         return 0;
612                 goto out;
613         }
614
615         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
616                 return 0;
617
618 out:
619         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
620 drop:
621         kfree_skb(skb);
622         return 0;
623 }
624
625 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
626 {
627         return iptunnel_handle_offloads(skb,
628                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
629 }
630
631 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
632                                      struct net_device *dev,
633                                      struct flowi6 *fl6, __u8 *dsfield,
634                                      int *encap_limit)
635 {
636         const struct iphdr *iph = ip_hdr(skb);
637         struct ip6_tnl *t = netdev_priv(dev);
638
639         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
640                 *encap_limit = t->parms.encap_limit;
641
642         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
643
644         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
645                 *dsfield = ipv4_get_dsfield(iph);
646         else
647                 *dsfield = ip6_tclass(t->parms.flowinfo);
648
649         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
650                 fl6->flowi6_mark = skb->mark;
651         else
652                 fl6->flowi6_mark = t->parms.fwmark;
653
654         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
655 }
656
657 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
658                                     struct net_device *dev,
659                                     struct flowi6 *fl6, __u8 *dsfield,
660                                     int *encap_limit)
661 {
662         struct ipv6hdr *ipv6h;
663         struct ip6_tnl *t = netdev_priv(dev);
664         __u16 offset;
665
666         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
667         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
668         ipv6h = ipv6_hdr(skb);
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_ndo_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 struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
705 {
706         struct ip_tunnel_info *tun_info;
707
708         tun_info = skb_tunnel_info(skb);
709         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
710                 return ERR_PTR(-EINVAL);
711
712         return tun_info;
713 }
714
715 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
716                                struct net_device *dev, __u8 dsfield,
717                                struct flowi6 *fl6, int encap_limit,
718                                __u32 *pmtu, __be16 proto)
719 {
720         struct ip6_tnl *tunnel = netdev_priv(dev);
721         __be16 protocol;
722         __be16 flags;
723
724         if (dev->type == ARPHRD_ETHER)
725                 IPCB(skb)->flags = 0;
726
727         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
728                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
729         else
730                 fl6->daddr = tunnel->parms.raddr;
731
732         /* Push GRE header. */
733         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
734
735         if (tunnel->parms.collect_md) {
736                 struct ip_tunnel_info *tun_info;
737                 const struct ip_tunnel_key *key;
738                 int tun_hlen;
739
740                 tun_info = skb_tunnel_info_txcheck(skb);
741                 if (IS_ERR(tun_info) ||
742                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
743                         return -EINVAL;
744
745                 key = &tun_info->key;
746                 memset(fl6, 0, sizeof(*fl6));
747                 fl6->flowi6_proto = IPPROTO_GRE;
748                 fl6->daddr = key->u.ipv6.dst;
749                 fl6->flowlabel = key->label;
750                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
751                 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
752
753                 dsfield = key->tos;
754                 flags = key->tun_flags &
755                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
756                 tun_hlen = gre_calc_hlen(flags);
757
758                 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
759                         return -ENOMEM;
760
761                 gre_build_header(skb, tun_hlen,
762                                  flags, protocol,
763                                  tunnel_id_to_key32(tun_info->key.tun_id),
764                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
765                                                       : 0);
766
767         } else {
768                 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
769                         return -ENOMEM;
770
771                 flags = tunnel->parms.o_flags;
772
773                 gre_build_header(skb, tunnel->tun_hlen, flags,
774                                  protocol, tunnel->parms.o_key,
775                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
776                                                       : 0);
777         }
778
779         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
780                             NEXTHDR_GRE);
781 }
782
783 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
784 {
785         struct ip6_tnl *t = netdev_priv(dev);
786         int encap_limit = -1;
787         struct flowi6 fl6;
788         __u8 dsfield = 0;
789         __u32 mtu;
790         int err;
791
792         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
793
794         if (!t->parms.collect_md)
795                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
796                                          &dsfield, &encap_limit);
797
798         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
799         if (err)
800                 return -1;
801
802         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
803                           skb->protocol);
804         if (err != 0) {
805                 /* XXX: send ICMP error even if DF is not set. */
806                 if (err == -EMSGSIZE)
807                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
808                                       htonl(mtu));
809                 return -1;
810         }
811
812         return 0;
813 }
814
815 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
816 {
817         struct ip6_tnl *t = netdev_priv(dev);
818         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
819         int encap_limit = -1;
820         struct flowi6 fl6;
821         __u8 dsfield = 0;
822         __u32 mtu;
823         int err;
824
825         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
826                 return -1;
827
828         if (!t->parms.collect_md &&
829             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
830                 return -1;
831
832         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
833                 return -1;
834
835         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
836                           &mtu, skb->protocol);
837         if (err != 0) {
838                 if (err == -EMSGSIZE)
839                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
840                 return -1;
841         }
842
843         return 0;
844 }
845
846 /**
847  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
848  *   @t: the outgoing tunnel device
849  *   @hdr: IPv6 header from the incoming packet
850  *
851  * Description:
852  *   Avoid trivial tunneling loop by checking that tunnel exit-point
853  *   doesn't match source of incoming packet.
854  *
855  * Return:
856  *   1 if conflict,
857  *   0 else
858  **/
859
860 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
861         const struct ipv6hdr *hdr)
862 {
863         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
864 }
865
866 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
867 {
868         struct ip6_tnl *t = netdev_priv(dev);
869         int encap_limit = -1;
870         struct flowi6 fl6;
871         __u32 mtu;
872         int err;
873
874         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
875                 encap_limit = t->parms.encap_limit;
876
877         if (!t->parms.collect_md)
878                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
879
880         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
881         if (err)
882                 return err;
883
884         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
885
886         return err;
887 }
888
889 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
890         struct net_device *dev)
891 {
892         struct ip6_tnl *t = netdev_priv(dev);
893         struct net_device_stats *stats = &t->dev->stats;
894         int ret;
895
896         if (!pskb_inet_may_pull(skb))
897                 goto tx_err;
898
899         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
900                 goto tx_err;
901
902         switch (skb->protocol) {
903         case htons(ETH_P_IP):
904                 ret = ip6gre_xmit_ipv4(skb, dev);
905                 break;
906         case htons(ETH_P_IPV6):
907                 ret = ip6gre_xmit_ipv6(skb, dev);
908                 break;
909         default:
910                 ret = ip6gre_xmit_other(skb, dev);
911                 break;
912         }
913
914         if (ret < 0)
915                 goto tx_err;
916
917         return NETDEV_TX_OK;
918
919 tx_err:
920         if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
921                 stats->tx_errors++;
922         stats->tx_dropped++;
923         kfree_skb(skb);
924         return NETDEV_TX_OK;
925 }
926
927 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
928                                          struct net_device *dev)
929 {
930         struct ip_tunnel_info *tun_info = NULL;
931         struct ip6_tnl *t = netdev_priv(dev);
932         struct dst_entry *dst = skb_dst(skb);
933         struct net_device_stats *stats;
934         bool truncate = false;
935         int encap_limit = -1;
936         __u8 dsfield = false;
937         struct flowi6 fl6;
938         int err = -EINVAL;
939         __be16 proto;
940         __u32 mtu;
941         int nhoff;
942
943         if (!pskb_inet_may_pull(skb))
944                 goto tx_err;
945
946         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
947                 goto tx_err;
948
949         if (gre_handle_offloads(skb, false))
950                 goto tx_err;
951
952         if (skb->len > dev->mtu + dev->hard_header_len) {
953                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
954                 truncate = true;
955         }
956
957         nhoff = skb_network_header(skb) - skb_mac_header(skb);
958         if (skb->protocol == htons(ETH_P_IP) &&
959             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
960                 truncate = true;
961
962         if (skb->protocol == htons(ETH_P_IPV6)) {
963                 int thoff;
964
965                 if (skb_transport_header_was_set(skb))
966                         thoff = skb_transport_header(skb) - skb_mac_header(skb);
967                 else
968                         thoff = nhoff + sizeof(struct ipv6hdr);
969                 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
970                         truncate = true;
971         }
972
973         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
974                 goto tx_err;
975
976         t->parms.o_flags &= ~TUNNEL_KEY;
977         IPCB(skb)->flags = 0;
978
979         /* For collect_md mode, derive fl6 from the tunnel key,
980          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
981          */
982         if (t->parms.collect_md) {
983                 const struct ip_tunnel_key *key;
984                 struct erspan_metadata *md;
985                 __be32 tun_id;
986
987                 tun_info = skb_tunnel_info_txcheck(skb);
988                 if (IS_ERR(tun_info) ||
989                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
990                         goto tx_err;
991
992                 key = &tun_info->key;
993                 memset(&fl6, 0, sizeof(fl6));
994                 fl6.flowi6_proto = IPPROTO_GRE;
995                 fl6.daddr = key->u.ipv6.dst;
996                 fl6.flowlabel = key->label;
997                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
998                 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
999
1000                 dsfield = key->tos;
1001                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1002                         goto tx_err;
1003                 if (tun_info->options_len < sizeof(*md))
1004                         goto tx_err;
1005                 md = ip_tunnel_info_opts(tun_info);
1006
1007                 tun_id = tunnel_id_to_key32(key->tun_id);
1008                 if (md->version == 1) {
1009                         erspan_build_header(skb,
1010                                             ntohl(tun_id),
1011                                             ntohl(md->u.index), truncate,
1012                                             false);
1013                 } else if (md->version == 2) {
1014                         erspan_build_header_v2(skb,
1015                                                ntohl(tun_id),
1016                                                md->u.md2.dir,
1017                                                get_hwid(&md->u.md2),
1018                                                truncate, false);
1019                 } else {
1020                         goto tx_err;
1021                 }
1022         } else {
1023                 switch (skb->protocol) {
1024                 case htons(ETH_P_IP):
1025                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1026                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1027                                                  &dsfield, &encap_limit);
1028                         break;
1029                 case htons(ETH_P_IPV6):
1030                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1031                                 goto tx_err;
1032                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1033                                                      &dsfield, &encap_limit))
1034                                 goto tx_err;
1035                         break;
1036                 default:
1037                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1038                         break;
1039                 }
1040
1041                 if (t->parms.erspan_ver == 1)
1042                         erspan_build_header(skb, ntohl(t->parms.o_key),
1043                                             t->parms.index,
1044                                             truncate, false);
1045                 else if (t->parms.erspan_ver == 2)
1046                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1047                                                t->parms.dir,
1048                                                t->parms.hwid,
1049                                                truncate, false);
1050                 else
1051                         goto tx_err;
1052
1053                 fl6.daddr = t->parms.raddr;
1054         }
1055
1056         /* Push GRE header. */
1057         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1058                                            : htons(ETH_P_ERSPAN2);
1059         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1060
1061         /* TooBig packet may have updated dst->dev's mtu */
1062         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1063                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1064
1065         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1066                            NEXTHDR_GRE);
1067         if (err != 0) {
1068                 /* XXX: send ICMP error even if DF is not set. */
1069                 if (err == -EMSGSIZE) {
1070                         if (skb->protocol == htons(ETH_P_IP))
1071                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1072                                               ICMP_FRAG_NEEDED, htonl(mtu));
1073                         else
1074                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1075                 }
1076
1077                 goto tx_err;
1078         }
1079         return NETDEV_TX_OK;
1080
1081 tx_err:
1082         stats = &t->dev->stats;
1083         if (!IS_ERR(tun_info))
1084                 stats->tx_errors++;
1085         stats->tx_dropped++;
1086         kfree_skb(skb);
1087         return NETDEV_TX_OK;
1088 }
1089
1090 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1091 {
1092         struct net_device *dev = t->dev;
1093         struct __ip6_tnl_parm *p = &t->parms;
1094         struct flowi6 *fl6 = &t->fl.u.ip6;
1095
1096         if (dev->type != ARPHRD_ETHER) {
1097                 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1098                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1099         }
1100
1101         /* Set up flowi template */
1102         fl6->saddr = p->laddr;
1103         fl6->daddr = p->raddr;
1104         fl6->flowi6_oif = p->link;
1105         fl6->flowlabel = 0;
1106         fl6->flowi6_proto = IPPROTO_GRE;
1107         fl6->fl6_gre_key = t->parms.o_key;
1108
1109         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1110                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1111         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1112                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1113
1114         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1115         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1116
1117         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1118                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1119                 dev->flags |= IFF_POINTOPOINT;
1120         else
1121                 dev->flags &= ~IFF_POINTOPOINT;
1122 }
1123
1124 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1125                                          int t_hlen)
1126 {
1127         const struct __ip6_tnl_parm *p = &t->parms;
1128         struct net_device *dev = t->dev;
1129
1130         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1131                 int strict = (ipv6_addr_type(&p->raddr) &
1132                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1133
1134                 struct rt6_info *rt = rt6_lookup(t->net,
1135                                                  &p->raddr, &p->laddr,
1136                                                  p->link, NULL, strict);
1137
1138                 if (!rt)
1139                         return;
1140
1141                 if (rt->dst.dev) {
1142                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1143                                                  t_hlen;
1144
1145                         if (t->dev->header_ops)
1146                                 dev->hard_header_len = dst_len;
1147                         else
1148                                 dev->needed_headroom = dst_len;
1149
1150                         if (set_mtu) {
1151                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1152                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1153                                         dev->mtu -= 8;
1154                                 if (dev->type == ARPHRD_ETHER)
1155                                         dev->mtu -= ETH_HLEN;
1156
1157                                 if (dev->mtu < IPV6_MIN_MTU)
1158                                         dev->mtu = IPV6_MIN_MTU;
1159                         }
1160                 }
1161                 ip6_rt_put(rt);
1162         }
1163 }
1164
1165 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1166 {
1167         int t_hlen;
1168
1169         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1170         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1171
1172         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1173
1174         if (tunnel->dev->header_ops)
1175                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1176         else
1177                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1178
1179         return t_hlen;
1180 }
1181
1182 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1183 {
1184         ip6gre_tnl_link_config_common(t);
1185         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1186 }
1187
1188 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1189                                      const struct __ip6_tnl_parm *p)
1190 {
1191         t->parms.laddr = p->laddr;
1192         t->parms.raddr = p->raddr;
1193         t->parms.flags = p->flags;
1194         t->parms.hop_limit = p->hop_limit;
1195         t->parms.encap_limit = p->encap_limit;
1196         t->parms.flowinfo = p->flowinfo;
1197         t->parms.link = p->link;
1198         t->parms.proto = p->proto;
1199         t->parms.i_key = p->i_key;
1200         t->parms.o_key = p->o_key;
1201         t->parms.i_flags = p->i_flags;
1202         t->parms.o_flags = p->o_flags;
1203         t->parms.fwmark = p->fwmark;
1204         t->parms.erspan_ver = p->erspan_ver;
1205         t->parms.index = p->index;
1206         t->parms.dir = p->dir;
1207         t->parms.hwid = p->hwid;
1208         dst_cache_reset(&t->dst_cache);
1209 }
1210
1211 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1212                              int set_mtu)
1213 {
1214         ip6gre_tnl_copy_tnl_parm(t, p);
1215         ip6gre_tnl_link_config(t, set_mtu);
1216         return 0;
1217 }
1218
1219 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1220         const struct ip6_tnl_parm2 *u)
1221 {
1222         p->laddr = u->laddr;
1223         p->raddr = u->raddr;
1224         p->flags = u->flags;
1225         p->hop_limit = u->hop_limit;
1226         p->encap_limit = u->encap_limit;
1227         p->flowinfo = u->flowinfo;
1228         p->link = u->link;
1229         p->i_key = u->i_key;
1230         p->o_key = u->o_key;
1231         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1232         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1233         memcpy(p->name, u->name, sizeof(u->name));
1234 }
1235
1236 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1237         const struct __ip6_tnl_parm *p)
1238 {
1239         u->proto = IPPROTO_GRE;
1240         u->laddr = p->laddr;
1241         u->raddr = p->raddr;
1242         u->flags = p->flags;
1243         u->hop_limit = p->hop_limit;
1244         u->encap_limit = p->encap_limit;
1245         u->flowinfo = p->flowinfo;
1246         u->link = p->link;
1247         u->i_key = p->i_key;
1248         u->o_key = p->o_key;
1249         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1250         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1251         memcpy(u->name, p->name, sizeof(u->name));
1252 }
1253
1254 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1255                                         struct ifreq *ifr, void __user *data,
1256                                         int cmd)
1257 {
1258         int err = 0;
1259         struct ip6_tnl_parm2 p;
1260         struct __ip6_tnl_parm p1;
1261         struct ip6_tnl *t = netdev_priv(dev);
1262         struct net *net = t->net;
1263         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1264
1265         memset(&p1, 0, sizeof(p1));
1266
1267         switch (cmd) {
1268         case SIOCGETTUNNEL:
1269                 if (dev == ign->fb_tunnel_dev) {
1270                         if (copy_from_user(&p, data, sizeof(p))) {
1271                                 err = -EFAULT;
1272                                 break;
1273                         }
1274                         ip6gre_tnl_parm_from_user(&p1, &p);
1275                         t = ip6gre_tunnel_locate(net, &p1, 0);
1276                         if (!t)
1277                                 t = netdev_priv(dev);
1278                 }
1279                 memset(&p, 0, sizeof(p));
1280                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1281                 if (copy_to_user(data, &p, sizeof(p)))
1282                         err = -EFAULT;
1283                 break;
1284
1285         case SIOCADDTUNNEL:
1286         case SIOCCHGTUNNEL:
1287                 err = -EPERM;
1288                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1289                         goto done;
1290
1291                 err = -EFAULT;
1292                 if (copy_from_user(&p, data, sizeof(p)))
1293                         goto done;
1294
1295                 err = -EINVAL;
1296                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1297                         goto done;
1298
1299                 if (!(p.i_flags&GRE_KEY))
1300                         p.i_key = 0;
1301                 if (!(p.o_flags&GRE_KEY))
1302                         p.o_key = 0;
1303
1304                 ip6gre_tnl_parm_from_user(&p1, &p);
1305                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1306
1307                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1308                         if (t) {
1309                                 if (t->dev != dev) {
1310                                         err = -EEXIST;
1311                                         break;
1312                                 }
1313                         } else {
1314                                 t = netdev_priv(dev);
1315
1316                                 ip6gre_tunnel_unlink(ign, t);
1317                                 synchronize_net();
1318                                 ip6gre_tnl_change(t, &p1, 1);
1319                                 ip6gre_tunnel_link(ign, t);
1320                                 netdev_state_change(dev);
1321                         }
1322                 }
1323
1324                 if (t) {
1325                         err = 0;
1326
1327                         memset(&p, 0, sizeof(p));
1328                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1329                         if (copy_to_user(data, &p, sizeof(p)))
1330                                 err = -EFAULT;
1331                 } else
1332                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1333                 break;
1334
1335         case SIOCDELTUNNEL:
1336                 err = -EPERM;
1337                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1338                         goto done;
1339
1340                 if (dev == ign->fb_tunnel_dev) {
1341                         err = -EFAULT;
1342                         if (copy_from_user(&p, data, sizeof(p)))
1343                                 goto done;
1344                         err = -ENOENT;
1345                         ip6gre_tnl_parm_from_user(&p1, &p);
1346                         t = ip6gre_tunnel_locate(net, &p1, 0);
1347                         if (!t)
1348                                 goto done;
1349                         err = -EPERM;
1350                         if (t == netdev_priv(ign->fb_tunnel_dev))
1351                                 goto done;
1352                         dev = t->dev;
1353                 }
1354                 unregister_netdevice(dev);
1355                 err = 0;
1356                 break;
1357
1358         default:
1359                 err = -EINVAL;
1360         }
1361
1362 done:
1363         return err;
1364 }
1365
1366 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1367                          unsigned short type, const void *daddr,
1368                          const void *saddr, unsigned int len)
1369 {
1370         struct ip6_tnl *t = netdev_priv(dev);
1371         struct ipv6hdr *ipv6h;
1372         __be16 *p;
1373
1374         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1375         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1376                                                   t->fl.u.ip6.flowlabel,
1377                                                   true, &t->fl.u.ip6));
1378         ipv6h->hop_limit = t->parms.hop_limit;
1379         ipv6h->nexthdr = NEXTHDR_GRE;
1380         ipv6h->saddr = t->parms.laddr;
1381         ipv6h->daddr = t->parms.raddr;
1382
1383         p = (__be16 *)(ipv6h + 1);
1384         p[0] = t->parms.o_flags;
1385         p[1] = htons(type);
1386
1387         /*
1388          *      Set the source hardware address.
1389          */
1390
1391         if (saddr)
1392                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1393         if (daddr)
1394                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1395         if (!ipv6_addr_any(&ipv6h->daddr))
1396                 return t->hlen;
1397
1398         return -t->hlen;
1399 }
1400
1401 static const struct header_ops ip6gre_header_ops = {
1402         .create = ip6gre_header,
1403 };
1404
1405 static const struct net_device_ops ip6gre_netdev_ops = {
1406         .ndo_init               = ip6gre_tunnel_init,
1407         .ndo_uninit             = ip6gre_tunnel_uninit,
1408         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1409         .ndo_siocdevprivate     = ip6gre_tunnel_siocdevprivate,
1410         .ndo_change_mtu         = ip6_tnl_change_mtu,
1411         .ndo_get_stats64        = dev_get_tstats64,
1412         .ndo_get_iflink         = ip6_tnl_get_iflink,
1413 };
1414
1415 static void ip6gre_dev_free(struct net_device *dev)
1416 {
1417         struct ip6_tnl *t = netdev_priv(dev);
1418
1419         gro_cells_destroy(&t->gro_cells);
1420         dst_cache_destroy(&t->dst_cache);
1421         free_percpu(dev->tstats);
1422 }
1423
1424 static void ip6gre_tunnel_setup(struct net_device *dev)
1425 {
1426         dev->netdev_ops = &ip6gre_netdev_ops;
1427         dev->needs_free_netdev = true;
1428         dev->priv_destructor = ip6gre_dev_free;
1429
1430         dev->type = ARPHRD_IP6GRE;
1431
1432         dev->flags |= IFF_NOARP;
1433         dev->addr_len = sizeof(struct in6_addr);
1434         netif_keep_dst(dev);
1435         /* This perm addr will be used as interface identifier by IPv6 */
1436         dev->addr_assign_type = NET_ADDR_RANDOM;
1437         eth_random_addr(dev->perm_addr);
1438 }
1439
1440 #define GRE6_FEATURES (NETIF_F_SG |             \
1441                        NETIF_F_FRAGLIST |       \
1442                        NETIF_F_HIGHDMA |        \
1443                        NETIF_F_HW_CSUM)
1444
1445 static void ip6gre_tnl_init_features(struct net_device *dev)
1446 {
1447         struct ip6_tnl *nt = netdev_priv(dev);
1448         __be16 flags;
1449
1450         dev->features           |= GRE6_FEATURES | NETIF_F_LLTX;
1451         dev->hw_features        |= GRE6_FEATURES;
1452
1453         flags = nt->parms.o_flags;
1454
1455         /* TCP offload with GRE SEQ is not supported, nor can we support 2
1456          * levels of outer headers requiring an update.
1457          */
1458         if (flags & TUNNEL_SEQ)
1459                 return;
1460         if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE)
1461                 return;
1462
1463         dev->features |= NETIF_F_GSO_SOFTWARE;
1464         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1465 }
1466
1467 static int ip6gre_tunnel_init_common(struct net_device *dev)
1468 {
1469         struct ip6_tnl *tunnel;
1470         int ret;
1471         int t_hlen;
1472
1473         tunnel = netdev_priv(dev);
1474
1475         tunnel->dev = dev;
1476         tunnel->net = dev_net(dev);
1477         strcpy(tunnel->parms.name, dev->name);
1478
1479         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1480         if (!dev->tstats)
1481                 return -ENOMEM;
1482
1483         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1484         if (ret)
1485                 goto cleanup_alloc_pcpu_stats;
1486
1487         ret = gro_cells_init(&tunnel->gro_cells, dev);
1488         if (ret)
1489                 goto cleanup_dst_cache_init;
1490
1491         t_hlen = ip6gre_calc_hlen(tunnel);
1492         dev->mtu = ETH_DATA_LEN - t_hlen;
1493         if (dev->type == ARPHRD_ETHER)
1494                 dev->mtu -= ETH_HLEN;
1495         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1496                 dev->mtu -= 8;
1497
1498         if (tunnel->parms.collect_md) {
1499                 netif_keep_dst(dev);
1500         }
1501         ip6gre_tnl_init_features(dev);
1502
1503         dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1504         return 0;
1505
1506 cleanup_dst_cache_init:
1507         dst_cache_destroy(&tunnel->dst_cache);
1508 cleanup_alloc_pcpu_stats:
1509         free_percpu(dev->tstats);
1510         dev->tstats = NULL;
1511         return ret;
1512 }
1513
1514 static int ip6gre_tunnel_init(struct net_device *dev)
1515 {
1516         struct ip6_tnl *tunnel;
1517         int ret;
1518
1519         ret = ip6gre_tunnel_init_common(dev);
1520         if (ret)
1521                 return ret;
1522
1523         tunnel = netdev_priv(dev);
1524
1525         if (tunnel->parms.collect_md)
1526                 return 0;
1527
1528         __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1529         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1530
1531         if (ipv6_addr_any(&tunnel->parms.raddr))
1532                 dev->header_ops = &ip6gre_header_ops;
1533
1534         return 0;
1535 }
1536
1537 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1538 {
1539         struct ip6_tnl *tunnel = netdev_priv(dev);
1540
1541         tunnel->dev = dev;
1542         tunnel->net = dev_net(dev);
1543         strcpy(tunnel->parms.name, dev->name);
1544
1545         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1546 }
1547
1548 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1549         .handler     = gre_rcv,
1550         .err_handler = ip6gre_err,
1551         .flags       = INET6_PROTO_FINAL,
1552 };
1553
1554 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1555 {
1556         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1557         struct net_device *dev, *aux;
1558         int prio;
1559
1560         for_each_netdev_safe(net, dev, aux)
1561                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1562                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1563                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1564                         unregister_netdevice_queue(dev, head);
1565
1566         for (prio = 0; prio < 4; prio++) {
1567                 int h;
1568                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1569                         struct ip6_tnl *t;
1570
1571                         t = rtnl_dereference(ign->tunnels[prio][h]);
1572
1573                         while (t) {
1574                                 /* If dev is in the same netns, it has already
1575                                  * been added to the list by the previous loop.
1576                                  */
1577                                 if (!net_eq(dev_net(t->dev), net))
1578                                         unregister_netdevice_queue(t->dev,
1579                                                                    head);
1580                                 t = rtnl_dereference(t->next);
1581                         }
1582                 }
1583         }
1584 }
1585
1586 static int __net_init ip6gre_init_net(struct net *net)
1587 {
1588         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1589         struct net_device *ndev;
1590         int err;
1591
1592         if (!net_has_fallback_tunnels(net))
1593                 return 0;
1594         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1595                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1596         if (!ndev) {
1597                 err = -ENOMEM;
1598                 goto err_alloc_dev;
1599         }
1600         ign->fb_tunnel_dev = ndev;
1601         dev_net_set(ign->fb_tunnel_dev, net);
1602         /* FB netdevice is special: we have one, and only one per netns.
1603          * Allowing to move it to another netns is clearly unsafe.
1604          */
1605         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1606
1607
1608         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1609         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1610
1611         err = register_netdev(ign->fb_tunnel_dev);
1612         if (err)
1613                 goto err_reg_dev;
1614
1615         rcu_assign_pointer(ign->tunnels_wc[0],
1616                            netdev_priv(ign->fb_tunnel_dev));
1617         return 0;
1618
1619 err_reg_dev:
1620         free_netdev(ndev);
1621 err_alloc_dev:
1622         return err;
1623 }
1624
1625 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1626 {
1627         struct net *net;
1628         LIST_HEAD(list);
1629
1630         rtnl_lock();
1631         list_for_each_entry(net, net_list, exit_list)
1632                 ip6gre_destroy_tunnels(net, &list);
1633         unregister_netdevice_many(&list);
1634         rtnl_unlock();
1635 }
1636
1637 static struct pernet_operations ip6gre_net_ops = {
1638         .init = ip6gre_init_net,
1639         .exit_batch = ip6gre_exit_batch_net,
1640         .id   = &ip6gre_net_id,
1641         .size = sizeof(struct ip6gre_net),
1642 };
1643
1644 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1645                                   struct netlink_ext_ack *extack)
1646 {
1647         __be16 flags;
1648
1649         if (!data)
1650                 return 0;
1651
1652         flags = 0;
1653         if (data[IFLA_GRE_IFLAGS])
1654                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1655         if (data[IFLA_GRE_OFLAGS])
1656                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1657         if (flags & (GRE_VERSION|GRE_ROUTING))
1658                 return -EINVAL;
1659
1660         return 0;
1661 }
1662
1663 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1664                                struct netlink_ext_ack *extack)
1665 {
1666         struct in6_addr daddr;
1667
1668         if (tb[IFLA_ADDRESS]) {
1669                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1670                         return -EINVAL;
1671                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1672                         return -EADDRNOTAVAIL;
1673         }
1674
1675         if (!data)
1676                 goto out;
1677
1678         if (data[IFLA_GRE_REMOTE]) {
1679                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1680                 if (ipv6_addr_any(&daddr))
1681                         return -EINVAL;
1682         }
1683
1684 out:
1685         return ip6gre_tunnel_validate(tb, data, extack);
1686 }
1687
1688 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1689                                   struct netlink_ext_ack *extack)
1690 {
1691         __be16 flags = 0;
1692         int ret, ver = 0;
1693
1694         if (!data)
1695                 return 0;
1696
1697         ret = ip6gre_tap_validate(tb, data, extack);
1698         if (ret)
1699                 return ret;
1700
1701         /* ERSPAN should only have GRE sequence and key flag */
1702         if (data[IFLA_GRE_OFLAGS])
1703                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1704         if (data[IFLA_GRE_IFLAGS])
1705                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1706         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1707             flags != (GRE_SEQ | GRE_KEY))
1708                 return -EINVAL;
1709
1710         /* ERSPAN Session ID only has 10-bit. Since we reuse
1711          * 32-bit key field as ID, check it's range.
1712          */
1713         if (data[IFLA_GRE_IKEY] &&
1714             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1715                 return -EINVAL;
1716
1717         if (data[IFLA_GRE_OKEY] &&
1718             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1719                 return -EINVAL;
1720
1721         if (data[IFLA_GRE_ERSPAN_VER]) {
1722                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1723                 if (ver != 1 && ver != 2)
1724                         return -EINVAL;
1725         }
1726
1727         if (ver == 1) {
1728                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1729                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1730
1731                         if (index & ~INDEX_MASK)
1732                                 return -EINVAL;
1733                 }
1734         } else if (ver == 2) {
1735                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1736                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1737
1738                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1739                                 return -EINVAL;
1740                 }
1741
1742                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1743                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1744
1745                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1746                                 return -EINVAL;
1747                 }
1748         }
1749
1750         return 0;
1751 }
1752
1753 static void ip6erspan_set_version(struct nlattr *data[],
1754                                   struct __ip6_tnl_parm *parms)
1755 {
1756         if (!data)
1757                 return;
1758
1759         parms->erspan_ver = 1;
1760         if (data[IFLA_GRE_ERSPAN_VER])
1761                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1762
1763         if (parms->erspan_ver == 1) {
1764                 if (data[IFLA_GRE_ERSPAN_INDEX])
1765                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1766         } else if (parms->erspan_ver == 2) {
1767                 if (data[IFLA_GRE_ERSPAN_DIR])
1768                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1769                 if (data[IFLA_GRE_ERSPAN_HWID])
1770                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1771         }
1772 }
1773
1774 static void ip6gre_netlink_parms(struct nlattr *data[],
1775                                 struct __ip6_tnl_parm *parms)
1776 {
1777         memset(parms, 0, sizeof(*parms));
1778
1779         if (!data)
1780                 return;
1781
1782         if (data[IFLA_GRE_LINK])
1783                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1784
1785         if (data[IFLA_GRE_IFLAGS])
1786                 parms->i_flags = gre_flags_to_tnl_flags(
1787                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1788
1789         if (data[IFLA_GRE_OFLAGS])
1790                 parms->o_flags = gre_flags_to_tnl_flags(
1791                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1792
1793         if (data[IFLA_GRE_IKEY])
1794                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1795
1796         if (data[IFLA_GRE_OKEY])
1797                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1798
1799         if (data[IFLA_GRE_LOCAL])
1800                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1801
1802         if (data[IFLA_GRE_REMOTE])
1803                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1804
1805         if (data[IFLA_GRE_TTL])
1806                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1807
1808         if (data[IFLA_GRE_ENCAP_LIMIT])
1809                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1810
1811         if (data[IFLA_GRE_FLOWINFO])
1812                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1813
1814         if (data[IFLA_GRE_FLAGS])
1815                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1816
1817         if (data[IFLA_GRE_FWMARK])
1818                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1819
1820         if (data[IFLA_GRE_COLLECT_METADATA])
1821                 parms->collect_md = true;
1822 }
1823
1824 static int ip6gre_tap_init(struct net_device *dev)
1825 {
1826         int ret;
1827
1828         ret = ip6gre_tunnel_init_common(dev);
1829         if (ret)
1830                 return ret;
1831
1832         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1833
1834         return 0;
1835 }
1836
1837 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1838         .ndo_init = ip6gre_tap_init,
1839         .ndo_uninit = ip6gre_tunnel_uninit,
1840         .ndo_start_xmit = ip6gre_tunnel_xmit,
1841         .ndo_set_mac_address = eth_mac_addr,
1842         .ndo_validate_addr = eth_validate_addr,
1843         .ndo_change_mtu = ip6_tnl_change_mtu,
1844         .ndo_get_stats64 = dev_get_tstats64,
1845         .ndo_get_iflink = ip6_tnl_get_iflink,
1846 };
1847
1848 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1849 {
1850         int t_hlen;
1851
1852         tunnel->tun_hlen = 8;
1853         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1854                        erspan_hdr_len(tunnel->parms.erspan_ver);
1855
1856         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1857         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1858         return t_hlen;
1859 }
1860
1861 static int ip6erspan_tap_init(struct net_device *dev)
1862 {
1863         struct ip6_tnl *tunnel;
1864         int t_hlen;
1865         int ret;
1866
1867         tunnel = netdev_priv(dev);
1868
1869         tunnel->dev = dev;
1870         tunnel->net = dev_net(dev);
1871         strcpy(tunnel->parms.name, dev->name);
1872
1873         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1874         if (!dev->tstats)
1875                 return -ENOMEM;
1876
1877         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1878         if (ret)
1879                 goto cleanup_alloc_pcpu_stats;
1880
1881         ret = gro_cells_init(&tunnel->gro_cells, dev);
1882         if (ret)
1883                 goto cleanup_dst_cache_init;
1884
1885         t_hlen = ip6erspan_calc_hlen(tunnel);
1886         dev->mtu = ETH_DATA_LEN - t_hlen;
1887         if (dev->type == ARPHRD_ETHER)
1888                 dev->mtu -= ETH_HLEN;
1889         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1890                 dev->mtu -= 8;
1891
1892         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1893         ip6erspan_tnl_link_config(tunnel, 1);
1894
1895         dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1896         return 0;
1897
1898 cleanup_dst_cache_init:
1899         dst_cache_destroy(&tunnel->dst_cache);
1900 cleanup_alloc_pcpu_stats:
1901         free_percpu(dev->tstats);
1902         dev->tstats = NULL;
1903         return ret;
1904 }
1905
1906 static const struct net_device_ops ip6erspan_netdev_ops = {
1907         .ndo_init =             ip6erspan_tap_init,
1908         .ndo_uninit =           ip6erspan_tunnel_uninit,
1909         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1910         .ndo_set_mac_address =  eth_mac_addr,
1911         .ndo_validate_addr =    eth_validate_addr,
1912         .ndo_change_mtu =       ip6_tnl_change_mtu,
1913         .ndo_get_stats64 =      dev_get_tstats64,
1914         .ndo_get_iflink =       ip6_tnl_get_iflink,
1915 };
1916
1917 static void ip6gre_tap_setup(struct net_device *dev)
1918 {
1919
1920         ether_setup(dev);
1921
1922         dev->max_mtu = 0;
1923         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1924         dev->needs_free_netdev = true;
1925         dev->priv_destructor = ip6gre_dev_free;
1926
1927         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1928         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1929         netif_keep_dst(dev);
1930 }
1931
1932 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1933                                        struct ip_tunnel_encap *ipencap)
1934 {
1935         bool ret = false;
1936
1937         memset(ipencap, 0, sizeof(*ipencap));
1938
1939         if (!data)
1940                 return ret;
1941
1942         if (data[IFLA_GRE_ENCAP_TYPE]) {
1943                 ret = true;
1944                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1945         }
1946
1947         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1948                 ret = true;
1949                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1950         }
1951
1952         if (data[IFLA_GRE_ENCAP_SPORT]) {
1953                 ret = true;
1954                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1955         }
1956
1957         if (data[IFLA_GRE_ENCAP_DPORT]) {
1958                 ret = true;
1959                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1960         }
1961
1962         return ret;
1963 }
1964
1965 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1966                                  struct nlattr *tb[], struct nlattr *data[],
1967                                  struct netlink_ext_ack *extack)
1968 {
1969         struct ip6_tnl *nt;
1970         struct ip_tunnel_encap ipencap;
1971         int err;
1972
1973         nt = netdev_priv(dev);
1974
1975         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1976                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1977
1978                 if (err < 0)
1979                         return err;
1980         }
1981
1982         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1983                 eth_hw_addr_random(dev);
1984
1985         nt->dev = dev;
1986         nt->net = dev_net(dev);
1987
1988         err = register_netdevice(dev);
1989         if (err)
1990                 goto out;
1991
1992         if (tb[IFLA_MTU])
1993                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1994
1995 out:
1996         return err;
1997 }
1998
1999 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2000                           struct nlattr *tb[], struct nlattr *data[],
2001                           struct netlink_ext_ack *extack)
2002 {
2003         struct ip6_tnl *nt = netdev_priv(dev);
2004         struct net *net = dev_net(dev);
2005         struct ip6gre_net *ign;
2006         int err;
2007
2008         ip6gre_netlink_parms(data, &nt->parms);
2009         ign = net_generic(net, ip6gre_net_id);
2010
2011         if (nt->parms.collect_md) {
2012                 if (rtnl_dereference(ign->collect_md_tun))
2013                         return -EEXIST;
2014         } else {
2015                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2016                         return -EEXIST;
2017         }
2018
2019         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2020         if (!err) {
2021                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2022                 ip6gre_tunnel_link_md(ign, nt);
2023                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2024         }
2025         return err;
2026 }
2027
2028 static struct ip6_tnl *
2029 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2030                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2031                          struct netlink_ext_ack *extack)
2032 {
2033         struct ip6_tnl *t, *nt = netdev_priv(dev);
2034         struct net *net = nt->net;
2035         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2036         struct ip_tunnel_encap ipencap;
2037
2038         if (dev == ign->fb_tunnel_dev)
2039                 return ERR_PTR(-EINVAL);
2040
2041         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2042                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2043
2044                 if (err < 0)
2045                         return ERR_PTR(err);
2046         }
2047
2048         ip6gre_netlink_parms(data, p_p);
2049
2050         t = ip6gre_tunnel_locate(net, p_p, 0);
2051
2052         if (t) {
2053                 if (t->dev != dev)
2054                         return ERR_PTR(-EEXIST);
2055         } else {
2056                 t = nt;
2057         }
2058
2059         return t;
2060 }
2061
2062 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2063                              struct nlattr *data[],
2064                              struct netlink_ext_ack *extack)
2065 {
2066         struct ip6_tnl *t = netdev_priv(dev);
2067         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2068         struct __ip6_tnl_parm p;
2069
2070         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2071         if (IS_ERR(t))
2072                 return PTR_ERR(t);
2073
2074         ip6gre_tunnel_unlink_md(ign, t);
2075         ip6gre_tunnel_unlink(ign, t);
2076         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2077         ip6gre_tunnel_link_md(ign, t);
2078         ip6gre_tunnel_link(ign, t);
2079         return 0;
2080 }
2081
2082 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2083 {
2084         struct net *net = dev_net(dev);
2085         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2086
2087         if (dev != ign->fb_tunnel_dev)
2088                 unregister_netdevice_queue(dev, head);
2089 }
2090
2091 static size_t ip6gre_get_size(const struct net_device *dev)
2092 {
2093         return
2094                 /* IFLA_GRE_LINK */
2095                 nla_total_size(4) +
2096                 /* IFLA_GRE_IFLAGS */
2097                 nla_total_size(2) +
2098                 /* IFLA_GRE_OFLAGS */
2099                 nla_total_size(2) +
2100                 /* IFLA_GRE_IKEY */
2101                 nla_total_size(4) +
2102                 /* IFLA_GRE_OKEY */
2103                 nla_total_size(4) +
2104                 /* IFLA_GRE_LOCAL */
2105                 nla_total_size(sizeof(struct in6_addr)) +
2106                 /* IFLA_GRE_REMOTE */
2107                 nla_total_size(sizeof(struct in6_addr)) +
2108                 /* IFLA_GRE_TTL */
2109                 nla_total_size(1) +
2110                 /* IFLA_GRE_ENCAP_LIMIT */
2111                 nla_total_size(1) +
2112                 /* IFLA_GRE_FLOWINFO */
2113                 nla_total_size(4) +
2114                 /* IFLA_GRE_FLAGS */
2115                 nla_total_size(4) +
2116                 /* IFLA_GRE_ENCAP_TYPE */
2117                 nla_total_size(2) +
2118                 /* IFLA_GRE_ENCAP_FLAGS */
2119                 nla_total_size(2) +
2120                 /* IFLA_GRE_ENCAP_SPORT */
2121                 nla_total_size(2) +
2122                 /* IFLA_GRE_ENCAP_DPORT */
2123                 nla_total_size(2) +
2124                 /* IFLA_GRE_COLLECT_METADATA */
2125                 nla_total_size(0) +
2126                 /* IFLA_GRE_FWMARK */
2127                 nla_total_size(4) +
2128                 /* IFLA_GRE_ERSPAN_INDEX */
2129                 nla_total_size(4) +
2130                 0;
2131 }
2132
2133 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2134 {
2135         struct ip6_tnl *t = netdev_priv(dev);
2136         struct __ip6_tnl_parm *p = &t->parms;
2137         __be16 o_flags = p->o_flags;
2138
2139         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2140                 if (!p->collect_md)
2141                         o_flags |= TUNNEL_KEY;
2142
2143                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2144                         goto nla_put_failure;
2145
2146                 if (p->erspan_ver == 1) {
2147                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2148                                 goto nla_put_failure;
2149                 } else {
2150                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2151                                 goto nla_put_failure;
2152                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2153                                 goto nla_put_failure;
2154                 }
2155         }
2156
2157         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2158             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2159                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2160             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2161                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2162             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2163             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2164             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2165             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2166             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2167             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2168             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2169             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2170             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2171                 goto nla_put_failure;
2172
2173         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2174                         t->encap.type) ||
2175             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2176                          t->encap.sport) ||
2177             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2178                          t->encap.dport) ||
2179             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2180                         t->encap.flags))
2181                 goto nla_put_failure;
2182
2183         if (p->collect_md) {
2184                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2185                         goto nla_put_failure;
2186         }
2187
2188         return 0;
2189
2190 nla_put_failure:
2191         return -EMSGSIZE;
2192 }
2193
2194 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2195         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2196         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2197         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2198         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2199         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2200         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2201         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2202         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2203         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2204         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2205         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2206         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2207         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2208         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2209         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2210         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2211         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2212         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2213         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2214         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2215         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2216 };
2217
2218 static void ip6erspan_tap_setup(struct net_device *dev)
2219 {
2220         ether_setup(dev);
2221
2222         dev->max_mtu = 0;
2223         dev->netdev_ops = &ip6erspan_netdev_ops;
2224         dev->needs_free_netdev = true;
2225         dev->priv_destructor = ip6gre_dev_free;
2226
2227         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2228         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2229         netif_keep_dst(dev);
2230 }
2231
2232 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2233                              struct nlattr *tb[], struct nlattr *data[],
2234                              struct netlink_ext_ack *extack)
2235 {
2236         struct ip6_tnl *nt = netdev_priv(dev);
2237         struct net *net = dev_net(dev);
2238         struct ip6gre_net *ign;
2239         int err;
2240
2241         ip6gre_netlink_parms(data, &nt->parms);
2242         ip6erspan_set_version(data, &nt->parms);
2243         ign = net_generic(net, ip6gre_net_id);
2244
2245         if (nt->parms.collect_md) {
2246                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2247                         return -EEXIST;
2248         } else {
2249                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2250                         return -EEXIST;
2251         }
2252
2253         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2254         if (!err) {
2255                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2256                 ip6erspan_tunnel_link_md(ign, nt);
2257                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2258         }
2259         return err;
2260 }
2261
2262 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2263 {
2264         ip6gre_tnl_link_config_common(t);
2265         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2266 }
2267
2268 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2269                                 const struct __ip6_tnl_parm *p, int set_mtu)
2270 {
2271         ip6gre_tnl_copy_tnl_parm(t, p);
2272         ip6erspan_tnl_link_config(t, set_mtu);
2273         return 0;
2274 }
2275
2276 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2277                                 struct nlattr *data[],
2278                                 struct netlink_ext_ack *extack)
2279 {
2280         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2281         struct __ip6_tnl_parm p;
2282         struct ip6_tnl *t;
2283
2284         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2285         if (IS_ERR(t))
2286                 return PTR_ERR(t);
2287
2288         ip6erspan_set_version(data, &p);
2289         ip6gre_tunnel_unlink_md(ign, t);
2290         ip6gre_tunnel_unlink(ign, t);
2291         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2292         ip6erspan_tunnel_link_md(ign, t);
2293         ip6gre_tunnel_link(ign, t);
2294         return 0;
2295 }
2296
2297 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2298         .kind           = "ip6gre",
2299         .maxtype        = IFLA_GRE_MAX,
2300         .policy         = ip6gre_policy,
2301         .priv_size      = sizeof(struct ip6_tnl),
2302         .setup          = ip6gre_tunnel_setup,
2303         .validate       = ip6gre_tunnel_validate,
2304         .newlink        = ip6gre_newlink,
2305         .changelink     = ip6gre_changelink,
2306         .dellink        = ip6gre_dellink,
2307         .get_size       = ip6gre_get_size,
2308         .fill_info      = ip6gre_fill_info,
2309         .get_link_net   = ip6_tnl_get_link_net,
2310 };
2311
2312 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2313         .kind           = "ip6gretap",
2314         .maxtype        = IFLA_GRE_MAX,
2315         .policy         = ip6gre_policy,
2316         .priv_size      = sizeof(struct ip6_tnl),
2317         .setup          = ip6gre_tap_setup,
2318         .validate       = ip6gre_tap_validate,
2319         .newlink        = ip6gre_newlink,
2320         .changelink     = ip6gre_changelink,
2321         .get_size       = ip6gre_get_size,
2322         .fill_info      = ip6gre_fill_info,
2323         .get_link_net   = ip6_tnl_get_link_net,
2324 };
2325
2326 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2327         .kind           = "ip6erspan",
2328         .maxtype        = IFLA_GRE_MAX,
2329         .policy         = ip6gre_policy,
2330         .priv_size      = sizeof(struct ip6_tnl),
2331         .setup          = ip6erspan_tap_setup,
2332         .validate       = ip6erspan_tap_validate,
2333         .newlink        = ip6erspan_newlink,
2334         .changelink     = ip6erspan_changelink,
2335         .get_size       = ip6gre_get_size,
2336         .fill_info      = ip6gre_fill_info,
2337         .get_link_net   = ip6_tnl_get_link_net,
2338 };
2339
2340 /*
2341  *      And now the modules code and kernel interface.
2342  */
2343
2344 static int __init ip6gre_init(void)
2345 {
2346         int err;
2347
2348         pr_info("GRE over IPv6 tunneling driver\n");
2349
2350         err = register_pernet_device(&ip6gre_net_ops);
2351         if (err < 0)
2352                 return err;
2353
2354         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2355         if (err < 0) {
2356                 pr_info("%s: can't add protocol\n", __func__);
2357                 goto add_proto_failed;
2358         }
2359
2360         err = rtnl_link_register(&ip6gre_link_ops);
2361         if (err < 0)
2362                 goto rtnl_link_failed;
2363
2364         err = rtnl_link_register(&ip6gre_tap_ops);
2365         if (err < 0)
2366                 goto tap_ops_failed;
2367
2368         err = rtnl_link_register(&ip6erspan_tap_ops);
2369         if (err < 0)
2370                 goto erspan_link_failed;
2371
2372 out:
2373         return err;
2374
2375 erspan_link_failed:
2376         rtnl_link_unregister(&ip6gre_tap_ops);
2377 tap_ops_failed:
2378         rtnl_link_unregister(&ip6gre_link_ops);
2379 rtnl_link_failed:
2380         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2381 add_proto_failed:
2382         unregister_pernet_device(&ip6gre_net_ops);
2383         goto out;
2384 }
2385
2386 static void __exit ip6gre_fini(void)
2387 {
2388         rtnl_link_unregister(&ip6gre_tap_ops);
2389         rtnl_link_unregister(&ip6gre_link_ops);
2390         rtnl_link_unregister(&ip6erspan_tap_ops);
2391         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2392         unregister_pernet_device(&ip6gre_net_ops);
2393 }
2394
2395 module_init(ip6gre_init);
2396 module_exit(ip6gre_fini);
2397 MODULE_LICENSE("GPL");
2398 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2399 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2400 MODULE_ALIAS_RTNL_LINK("ip6gre");
2401 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2402 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2403 MODULE_ALIAS_NETDEV("ip6gre0");