Merge branch 'for-6.4/amd-sfh' into for-linus
[linux-block.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                 strscpy(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         netdev_put(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         netdev_put(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 int prepare_ip6gre_xmit_other(struct sk_buff *skb,
705                                      struct net_device *dev,
706                                      struct flowi6 *fl6, __u8 *dsfield,
707                                      int *encap_limit)
708 {
709         struct ip6_tnl *t = netdev_priv(dev);
710
711         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
712                 *encap_limit = t->parms.encap_limit;
713
714         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
715
716         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
717                 *dsfield = 0;
718         else
719                 *dsfield = ip6_tclass(t->parms.flowinfo);
720
721         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
722                 fl6->flowi6_mark = skb->mark;
723         else
724                 fl6->flowi6_mark = t->parms.fwmark;
725
726         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
727
728         return 0;
729 }
730
731 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
732 {
733         struct ip_tunnel_info *tun_info;
734
735         tun_info = skb_tunnel_info(skb);
736         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
737                 return ERR_PTR(-EINVAL);
738
739         return tun_info;
740 }
741
742 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
743                                struct net_device *dev, __u8 dsfield,
744                                struct flowi6 *fl6, int encap_limit,
745                                __u32 *pmtu, __be16 proto)
746 {
747         struct ip6_tnl *tunnel = netdev_priv(dev);
748         __be16 protocol;
749         __be16 flags;
750
751         if (dev->type == ARPHRD_ETHER)
752                 IPCB(skb)->flags = 0;
753
754         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
755                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
756         else
757                 fl6->daddr = tunnel->parms.raddr;
758
759         /* Push GRE header. */
760         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
761
762         if (tunnel->parms.collect_md) {
763                 struct ip_tunnel_info *tun_info;
764                 const struct ip_tunnel_key *key;
765                 int tun_hlen;
766
767                 tun_info = skb_tunnel_info_txcheck(skb);
768                 if (IS_ERR(tun_info) ||
769                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
770                         return -EINVAL;
771
772                 key = &tun_info->key;
773                 memset(fl6, 0, sizeof(*fl6));
774                 fl6->flowi6_proto = IPPROTO_GRE;
775                 fl6->daddr = key->u.ipv6.dst;
776                 fl6->flowlabel = key->label;
777                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
778                 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
779
780                 dsfield = key->tos;
781                 flags = key->tun_flags &
782                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
783                 tun_hlen = gre_calc_hlen(flags);
784
785                 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
786                         return -ENOMEM;
787
788                 gre_build_header(skb, tun_hlen,
789                                  flags, protocol,
790                                  tunnel_id_to_key32(tun_info->key.tun_id),
791                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
792                                                       : 0);
793
794         } else {
795                 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
796                         return -ENOMEM;
797
798                 flags = tunnel->parms.o_flags;
799
800                 gre_build_header(skb, tunnel->tun_hlen, flags,
801                                  protocol, tunnel->parms.o_key,
802                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
803                                                       : 0);
804         }
805
806         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
807                             NEXTHDR_GRE);
808 }
809
810 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
811 {
812         struct ip6_tnl *t = netdev_priv(dev);
813         int encap_limit = -1;
814         struct flowi6 fl6;
815         __u8 dsfield = 0;
816         __u32 mtu;
817         int err;
818
819         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
820
821         if (!t->parms.collect_md)
822                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
823                                          &dsfield, &encap_limit);
824
825         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
826         if (err)
827                 return -1;
828
829         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
830                           skb->protocol);
831         if (err != 0) {
832                 /* XXX: send ICMP error even if DF is not set. */
833                 if (err == -EMSGSIZE)
834                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
835                                       htonl(mtu));
836                 return -1;
837         }
838
839         return 0;
840 }
841
842 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
843 {
844         struct ip6_tnl *t = netdev_priv(dev);
845         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
846         int encap_limit = -1;
847         struct flowi6 fl6;
848         __u8 dsfield = 0;
849         __u32 mtu;
850         int err;
851
852         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
853                 return -1;
854
855         if (!t->parms.collect_md &&
856             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
857                 return -1;
858
859         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
860                 return -1;
861
862         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
863                           &mtu, skb->protocol);
864         if (err != 0) {
865                 if (err == -EMSGSIZE)
866                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
867                 return -1;
868         }
869
870         return 0;
871 }
872
873 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
874 {
875         struct ip6_tnl *t = netdev_priv(dev);
876         int encap_limit = -1;
877         struct flowi6 fl6;
878         __u8 dsfield = 0;
879         __u32 mtu;
880         int err;
881
882         if (!t->parms.collect_md &&
883             prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit))
884                 return -1;
885
886         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
887         if (err)
888                 return err;
889         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol);
890
891         return err;
892 }
893
894 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
895         struct net_device *dev)
896 {
897         struct ip6_tnl *t = netdev_priv(dev);
898         __be16 payload_protocol;
899         int ret;
900
901         if (!pskb_inet_may_pull(skb))
902                 goto tx_err;
903
904         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
905                 goto tx_err;
906
907         payload_protocol = skb_protocol(skb, true);
908         switch (payload_protocol) {
909         case htons(ETH_P_IP):
910                 ret = ip6gre_xmit_ipv4(skb, dev);
911                 break;
912         case htons(ETH_P_IPV6):
913                 ret = ip6gre_xmit_ipv6(skb, dev);
914                 break;
915         default:
916                 ret = ip6gre_xmit_other(skb, dev);
917                 break;
918         }
919
920         if (ret < 0)
921                 goto tx_err;
922
923         return NETDEV_TX_OK;
924
925 tx_err:
926         if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
927                 DEV_STATS_INC(dev, tx_errors);
928         DEV_STATS_INC(dev, tx_dropped);
929         kfree_skb(skb);
930         return NETDEV_TX_OK;
931 }
932
933 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
934                                          struct net_device *dev)
935 {
936         struct ip_tunnel_info *tun_info = NULL;
937         struct ip6_tnl *t = netdev_priv(dev);
938         struct dst_entry *dst = skb_dst(skb);
939         bool truncate = false;
940         int encap_limit = -1;
941         __u8 dsfield = false;
942         struct flowi6 fl6;
943         int err = -EINVAL;
944         __be16 proto;
945         __u32 mtu;
946         int nhoff;
947
948         if (!pskb_inet_may_pull(skb))
949                 goto tx_err;
950
951         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
952                 goto tx_err;
953
954         if (gre_handle_offloads(skb, false))
955                 goto tx_err;
956
957         if (skb->len > dev->mtu + dev->hard_header_len) {
958                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
959                 truncate = true;
960         }
961
962         nhoff = skb_network_offset(skb);
963         if (skb->protocol == htons(ETH_P_IP) &&
964             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
965                 truncate = true;
966
967         if (skb->protocol == htons(ETH_P_IPV6)) {
968                 int thoff;
969
970                 if (skb_transport_header_was_set(skb))
971                         thoff = skb_transport_offset(skb);
972                 else
973                         thoff = nhoff + sizeof(struct ipv6hdr);
974                 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
975                         truncate = true;
976         }
977
978         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
979                 goto tx_err;
980
981         t->parms.o_flags &= ~TUNNEL_KEY;
982         IPCB(skb)->flags = 0;
983
984         /* For collect_md mode, derive fl6 from the tunnel key,
985          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
986          */
987         if (t->parms.collect_md) {
988                 const struct ip_tunnel_key *key;
989                 struct erspan_metadata *md;
990                 __be32 tun_id;
991
992                 tun_info = skb_tunnel_info_txcheck(skb);
993                 if (IS_ERR(tun_info) ||
994                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
995                         goto tx_err;
996
997                 key = &tun_info->key;
998                 memset(&fl6, 0, sizeof(fl6));
999                 fl6.flowi6_proto = IPPROTO_GRE;
1000                 fl6.daddr = key->u.ipv6.dst;
1001                 fl6.flowlabel = key->label;
1002                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1003                 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1004
1005                 dsfield = key->tos;
1006                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1007                         goto tx_err;
1008                 if (tun_info->options_len < sizeof(*md))
1009                         goto tx_err;
1010                 md = ip_tunnel_info_opts(tun_info);
1011
1012                 tun_id = tunnel_id_to_key32(key->tun_id);
1013                 if (md->version == 1) {
1014                         erspan_build_header(skb,
1015                                             ntohl(tun_id),
1016                                             ntohl(md->u.index), truncate,
1017                                             false);
1018                 } else if (md->version == 2) {
1019                         erspan_build_header_v2(skb,
1020                                                ntohl(tun_id),
1021                                                md->u.md2.dir,
1022                                                get_hwid(&md->u.md2),
1023                                                truncate, false);
1024                 } else {
1025                         goto tx_err;
1026                 }
1027         } else {
1028                 switch (skb->protocol) {
1029                 case htons(ETH_P_IP):
1030                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1031                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1032                                                  &dsfield, &encap_limit);
1033                         break;
1034                 case htons(ETH_P_IPV6):
1035                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1036                                 goto tx_err;
1037                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1038                                                      &dsfield, &encap_limit))
1039                                 goto tx_err;
1040                         break;
1041                 default:
1042                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1043                         break;
1044                 }
1045
1046                 if (t->parms.erspan_ver == 1)
1047                         erspan_build_header(skb, ntohl(t->parms.o_key),
1048                                             t->parms.index,
1049                                             truncate, false);
1050                 else if (t->parms.erspan_ver == 2)
1051                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1052                                                t->parms.dir,
1053                                                t->parms.hwid,
1054                                                truncate, false);
1055                 else
1056                         goto tx_err;
1057
1058                 fl6.daddr = t->parms.raddr;
1059         }
1060
1061         /* Push GRE header. */
1062         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1063                                            : htons(ETH_P_ERSPAN2);
1064         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1065
1066         /* TooBig packet may have updated dst->dev's mtu */
1067         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1068                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1069
1070         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1071                            NEXTHDR_GRE);
1072         if (err != 0) {
1073                 /* XXX: send ICMP error even if DF is not set. */
1074                 if (err == -EMSGSIZE) {
1075                         if (skb->protocol == htons(ETH_P_IP))
1076                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1077                                               ICMP_FRAG_NEEDED, htonl(mtu));
1078                         else
1079                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1080                 }
1081
1082                 goto tx_err;
1083         }
1084         return NETDEV_TX_OK;
1085
1086 tx_err:
1087         if (!IS_ERR(tun_info))
1088                 DEV_STATS_INC(dev, tx_errors);
1089         DEV_STATS_INC(dev, tx_dropped);
1090         kfree_skb(skb);
1091         return NETDEV_TX_OK;
1092 }
1093
1094 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1095 {
1096         struct net_device *dev = t->dev;
1097         struct __ip6_tnl_parm *p = &t->parms;
1098         struct flowi6 *fl6 = &t->fl.u.ip6;
1099
1100         if (dev->type != ARPHRD_ETHER) {
1101                 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1102                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1103         }
1104
1105         /* Set up flowi template */
1106         fl6->saddr = p->laddr;
1107         fl6->daddr = p->raddr;
1108         fl6->flowi6_oif = p->link;
1109         fl6->flowlabel = 0;
1110         fl6->flowi6_proto = IPPROTO_GRE;
1111         fl6->fl6_gre_key = t->parms.o_key;
1112
1113         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1114                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1115         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1116                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1117
1118         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1119         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1120
1121         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1122                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1123                 dev->flags |= IFF_POINTOPOINT;
1124         else
1125                 dev->flags &= ~IFF_POINTOPOINT;
1126 }
1127
1128 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1129                                          int t_hlen)
1130 {
1131         const struct __ip6_tnl_parm *p = &t->parms;
1132         struct net_device *dev = t->dev;
1133
1134         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1135                 int strict = (ipv6_addr_type(&p->raddr) &
1136                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1137
1138                 struct rt6_info *rt = rt6_lookup(t->net,
1139                                                  &p->raddr, &p->laddr,
1140                                                  p->link, NULL, strict);
1141
1142                 if (!rt)
1143                         return;
1144
1145                 if (rt->dst.dev) {
1146                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1147                                                  t_hlen;
1148
1149                         if (t->dev->header_ops)
1150                                 dev->hard_header_len = dst_len;
1151                         else
1152                                 dev->needed_headroom = dst_len;
1153
1154                         if (set_mtu) {
1155                                 int mtu = rt->dst.dev->mtu - t_hlen;
1156
1157                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1158                                         mtu -= 8;
1159                                 if (dev->type == ARPHRD_ETHER)
1160                                         mtu -= ETH_HLEN;
1161
1162                                 if (mtu < IPV6_MIN_MTU)
1163                                         mtu = IPV6_MIN_MTU;
1164                                 WRITE_ONCE(dev->mtu, mtu);
1165                         }
1166                 }
1167                 ip6_rt_put(rt);
1168         }
1169 }
1170
1171 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1172 {
1173         int t_hlen;
1174
1175         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1176         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1177
1178         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1179
1180         if (tunnel->dev->header_ops)
1181                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1182         else
1183                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1184
1185         return t_hlen;
1186 }
1187
1188 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1189 {
1190         ip6gre_tnl_link_config_common(t);
1191         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1192 }
1193
1194 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1195                                      const struct __ip6_tnl_parm *p)
1196 {
1197         t->parms.laddr = p->laddr;
1198         t->parms.raddr = p->raddr;
1199         t->parms.flags = p->flags;
1200         t->parms.hop_limit = p->hop_limit;
1201         t->parms.encap_limit = p->encap_limit;
1202         t->parms.flowinfo = p->flowinfo;
1203         t->parms.link = p->link;
1204         t->parms.proto = p->proto;
1205         t->parms.i_key = p->i_key;
1206         t->parms.o_key = p->o_key;
1207         t->parms.i_flags = p->i_flags;
1208         t->parms.o_flags = p->o_flags;
1209         t->parms.fwmark = p->fwmark;
1210         t->parms.erspan_ver = p->erspan_ver;
1211         t->parms.index = p->index;
1212         t->parms.dir = p->dir;
1213         t->parms.hwid = p->hwid;
1214         dst_cache_reset(&t->dst_cache);
1215 }
1216
1217 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1218                              int set_mtu)
1219 {
1220         ip6gre_tnl_copy_tnl_parm(t, p);
1221         ip6gre_tnl_link_config(t, set_mtu);
1222         return 0;
1223 }
1224
1225 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1226         const struct ip6_tnl_parm2 *u)
1227 {
1228         p->laddr = u->laddr;
1229         p->raddr = u->raddr;
1230         p->flags = u->flags;
1231         p->hop_limit = u->hop_limit;
1232         p->encap_limit = u->encap_limit;
1233         p->flowinfo = u->flowinfo;
1234         p->link = u->link;
1235         p->i_key = u->i_key;
1236         p->o_key = u->o_key;
1237         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1238         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1239         memcpy(p->name, u->name, sizeof(u->name));
1240 }
1241
1242 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1243         const struct __ip6_tnl_parm *p)
1244 {
1245         u->proto = IPPROTO_GRE;
1246         u->laddr = p->laddr;
1247         u->raddr = p->raddr;
1248         u->flags = p->flags;
1249         u->hop_limit = p->hop_limit;
1250         u->encap_limit = p->encap_limit;
1251         u->flowinfo = p->flowinfo;
1252         u->link = p->link;
1253         u->i_key = p->i_key;
1254         u->o_key = p->o_key;
1255         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1256         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1257         memcpy(u->name, p->name, sizeof(u->name));
1258 }
1259
1260 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1261                                         struct ifreq *ifr, void __user *data,
1262                                         int cmd)
1263 {
1264         int err = 0;
1265         struct ip6_tnl_parm2 p;
1266         struct __ip6_tnl_parm p1;
1267         struct ip6_tnl *t = netdev_priv(dev);
1268         struct net *net = t->net;
1269         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1270
1271         memset(&p1, 0, sizeof(p1));
1272
1273         switch (cmd) {
1274         case SIOCGETTUNNEL:
1275                 if (dev == ign->fb_tunnel_dev) {
1276                         if (copy_from_user(&p, data, sizeof(p))) {
1277                                 err = -EFAULT;
1278                                 break;
1279                         }
1280                         ip6gre_tnl_parm_from_user(&p1, &p);
1281                         t = ip6gre_tunnel_locate(net, &p1, 0);
1282                         if (!t)
1283                                 t = netdev_priv(dev);
1284                 }
1285                 memset(&p, 0, sizeof(p));
1286                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1287                 if (copy_to_user(data, &p, sizeof(p)))
1288                         err = -EFAULT;
1289                 break;
1290
1291         case SIOCADDTUNNEL:
1292         case SIOCCHGTUNNEL:
1293                 err = -EPERM;
1294                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1295                         goto done;
1296
1297                 err = -EFAULT;
1298                 if (copy_from_user(&p, data, sizeof(p)))
1299                         goto done;
1300
1301                 err = -EINVAL;
1302                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1303                         goto done;
1304
1305                 if (!(p.i_flags&GRE_KEY))
1306                         p.i_key = 0;
1307                 if (!(p.o_flags&GRE_KEY))
1308                         p.o_key = 0;
1309
1310                 ip6gre_tnl_parm_from_user(&p1, &p);
1311                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1312
1313                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1314                         if (t) {
1315                                 if (t->dev != dev) {
1316                                         err = -EEXIST;
1317                                         break;
1318                                 }
1319                         } else {
1320                                 t = netdev_priv(dev);
1321
1322                                 ip6gre_tunnel_unlink(ign, t);
1323                                 synchronize_net();
1324                                 ip6gre_tnl_change(t, &p1, 1);
1325                                 ip6gre_tunnel_link(ign, t);
1326                                 netdev_state_change(dev);
1327                         }
1328                 }
1329
1330                 if (t) {
1331                         err = 0;
1332
1333                         memset(&p, 0, sizeof(p));
1334                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1335                         if (copy_to_user(data, &p, sizeof(p)))
1336                                 err = -EFAULT;
1337                 } else
1338                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1339                 break;
1340
1341         case SIOCDELTUNNEL:
1342                 err = -EPERM;
1343                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1344                         goto done;
1345
1346                 if (dev == ign->fb_tunnel_dev) {
1347                         err = -EFAULT;
1348                         if (copy_from_user(&p, data, sizeof(p)))
1349                                 goto done;
1350                         err = -ENOENT;
1351                         ip6gre_tnl_parm_from_user(&p1, &p);
1352                         t = ip6gre_tunnel_locate(net, &p1, 0);
1353                         if (!t)
1354                                 goto done;
1355                         err = -EPERM;
1356                         if (t == netdev_priv(ign->fb_tunnel_dev))
1357                                 goto done;
1358                         dev = t->dev;
1359                 }
1360                 unregister_netdevice(dev);
1361                 err = 0;
1362                 break;
1363
1364         default:
1365                 err = -EINVAL;
1366         }
1367
1368 done:
1369         return err;
1370 }
1371
1372 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1373                          unsigned short type, const void *daddr,
1374                          const void *saddr, unsigned int len)
1375 {
1376         struct ip6_tnl *t = netdev_priv(dev);
1377         struct ipv6hdr *ipv6h;
1378         __be16 *p;
1379
1380         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1381         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1382                                                   t->fl.u.ip6.flowlabel,
1383                                                   true, &t->fl.u.ip6));
1384         ipv6h->hop_limit = t->parms.hop_limit;
1385         ipv6h->nexthdr = NEXTHDR_GRE;
1386         ipv6h->saddr = t->parms.laddr;
1387         ipv6h->daddr = t->parms.raddr;
1388
1389         p = (__be16 *)(ipv6h + 1);
1390         p[0] = t->parms.o_flags;
1391         p[1] = htons(type);
1392
1393         /*
1394          *      Set the source hardware address.
1395          */
1396
1397         if (saddr)
1398                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1399         if (daddr)
1400                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1401         if (!ipv6_addr_any(&ipv6h->daddr))
1402                 return t->hlen;
1403
1404         return -t->hlen;
1405 }
1406
1407 static const struct header_ops ip6gre_header_ops = {
1408         .create = ip6gre_header,
1409 };
1410
1411 static const struct net_device_ops ip6gre_netdev_ops = {
1412         .ndo_init               = ip6gre_tunnel_init,
1413         .ndo_uninit             = ip6gre_tunnel_uninit,
1414         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1415         .ndo_siocdevprivate     = ip6gre_tunnel_siocdevprivate,
1416         .ndo_change_mtu         = ip6_tnl_change_mtu,
1417         .ndo_get_stats64        = dev_get_tstats64,
1418         .ndo_get_iflink         = ip6_tnl_get_iflink,
1419 };
1420
1421 static void ip6gre_dev_free(struct net_device *dev)
1422 {
1423         struct ip6_tnl *t = netdev_priv(dev);
1424
1425         gro_cells_destroy(&t->gro_cells);
1426         dst_cache_destroy(&t->dst_cache);
1427         free_percpu(dev->tstats);
1428 }
1429
1430 static void ip6gre_tunnel_setup(struct net_device *dev)
1431 {
1432         dev->netdev_ops = &ip6gre_netdev_ops;
1433         dev->needs_free_netdev = true;
1434         dev->priv_destructor = ip6gre_dev_free;
1435
1436         dev->type = ARPHRD_IP6GRE;
1437
1438         dev->flags |= IFF_NOARP;
1439         dev->addr_len = sizeof(struct in6_addr);
1440         netif_keep_dst(dev);
1441         /* This perm addr will be used as interface identifier by IPv6 */
1442         dev->addr_assign_type = NET_ADDR_RANDOM;
1443         eth_random_addr(dev->perm_addr);
1444 }
1445
1446 #define GRE6_FEATURES (NETIF_F_SG |             \
1447                        NETIF_F_FRAGLIST |       \
1448                        NETIF_F_HIGHDMA |        \
1449                        NETIF_F_HW_CSUM)
1450
1451 static void ip6gre_tnl_init_features(struct net_device *dev)
1452 {
1453         struct ip6_tnl *nt = netdev_priv(dev);
1454         __be16 flags;
1455
1456         dev->features           |= GRE6_FEATURES | NETIF_F_LLTX;
1457         dev->hw_features        |= GRE6_FEATURES;
1458
1459         flags = nt->parms.o_flags;
1460
1461         /* TCP offload with GRE SEQ is not supported, nor can we support 2
1462          * levels of outer headers requiring an update.
1463          */
1464         if (flags & TUNNEL_SEQ)
1465                 return;
1466         if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE)
1467                 return;
1468
1469         dev->features |= NETIF_F_GSO_SOFTWARE;
1470         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1471 }
1472
1473 static int ip6gre_tunnel_init_common(struct net_device *dev)
1474 {
1475         struct ip6_tnl *tunnel;
1476         int ret;
1477         int t_hlen;
1478
1479         tunnel = netdev_priv(dev);
1480
1481         tunnel->dev = dev;
1482         tunnel->net = dev_net(dev);
1483         strcpy(tunnel->parms.name, dev->name);
1484
1485         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1486         if (!dev->tstats)
1487                 return -ENOMEM;
1488
1489         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1490         if (ret)
1491                 goto cleanup_alloc_pcpu_stats;
1492
1493         ret = gro_cells_init(&tunnel->gro_cells, dev);
1494         if (ret)
1495                 goto cleanup_dst_cache_init;
1496
1497         t_hlen = ip6gre_calc_hlen(tunnel);
1498         dev->mtu = ETH_DATA_LEN - t_hlen;
1499         if (dev->type == ARPHRD_ETHER)
1500                 dev->mtu -= ETH_HLEN;
1501         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1502                 dev->mtu -= 8;
1503
1504         if (tunnel->parms.collect_md) {
1505                 netif_keep_dst(dev);
1506         }
1507         ip6gre_tnl_init_features(dev);
1508
1509         netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1510         return 0;
1511
1512 cleanup_dst_cache_init:
1513         dst_cache_destroy(&tunnel->dst_cache);
1514 cleanup_alloc_pcpu_stats:
1515         free_percpu(dev->tstats);
1516         dev->tstats = NULL;
1517         return ret;
1518 }
1519
1520 static int ip6gre_tunnel_init(struct net_device *dev)
1521 {
1522         struct ip6_tnl *tunnel;
1523         int ret;
1524
1525         ret = ip6gre_tunnel_init_common(dev);
1526         if (ret)
1527                 return ret;
1528
1529         tunnel = netdev_priv(dev);
1530
1531         if (tunnel->parms.collect_md)
1532                 return 0;
1533
1534         __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1535         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1536
1537         if (ipv6_addr_any(&tunnel->parms.raddr))
1538                 dev->header_ops = &ip6gre_header_ops;
1539
1540         return 0;
1541 }
1542
1543 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1544 {
1545         struct ip6_tnl *tunnel = netdev_priv(dev);
1546
1547         tunnel->dev = dev;
1548         tunnel->net = dev_net(dev);
1549         strcpy(tunnel->parms.name, dev->name);
1550
1551         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1552 }
1553
1554 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1555         .handler     = gre_rcv,
1556         .err_handler = ip6gre_err,
1557         .flags       = INET6_PROTO_FINAL,
1558 };
1559
1560 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1561 {
1562         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1563         struct net_device *dev, *aux;
1564         int prio;
1565
1566         for_each_netdev_safe(net, dev, aux)
1567                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1568                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1569                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1570                         unregister_netdevice_queue(dev, head);
1571
1572         for (prio = 0; prio < 4; prio++) {
1573                 int h;
1574                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1575                         struct ip6_tnl *t;
1576
1577                         t = rtnl_dereference(ign->tunnels[prio][h]);
1578
1579                         while (t) {
1580                                 /* If dev is in the same netns, it has already
1581                                  * been added to the list by the previous loop.
1582                                  */
1583                                 if (!net_eq(dev_net(t->dev), net))
1584                                         unregister_netdevice_queue(t->dev,
1585                                                                    head);
1586                                 t = rtnl_dereference(t->next);
1587                         }
1588                 }
1589         }
1590 }
1591
1592 static int __net_init ip6gre_init_net(struct net *net)
1593 {
1594         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1595         struct net_device *ndev;
1596         int err;
1597
1598         if (!net_has_fallback_tunnels(net))
1599                 return 0;
1600         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1601                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1602         if (!ndev) {
1603                 err = -ENOMEM;
1604                 goto err_alloc_dev;
1605         }
1606         ign->fb_tunnel_dev = ndev;
1607         dev_net_set(ign->fb_tunnel_dev, net);
1608         /* FB netdevice is special: we have one, and only one per netns.
1609          * Allowing to move it to another netns is clearly unsafe.
1610          */
1611         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1612
1613
1614         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1615         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1616
1617         err = register_netdev(ign->fb_tunnel_dev);
1618         if (err)
1619                 goto err_reg_dev;
1620
1621         rcu_assign_pointer(ign->tunnels_wc[0],
1622                            netdev_priv(ign->fb_tunnel_dev));
1623         return 0;
1624
1625 err_reg_dev:
1626         free_netdev(ndev);
1627 err_alloc_dev:
1628         return err;
1629 }
1630
1631 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1632 {
1633         struct net *net;
1634         LIST_HEAD(list);
1635
1636         rtnl_lock();
1637         list_for_each_entry(net, net_list, exit_list)
1638                 ip6gre_destroy_tunnels(net, &list);
1639         unregister_netdevice_many(&list);
1640         rtnl_unlock();
1641 }
1642
1643 static struct pernet_operations ip6gre_net_ops = {
1644         .init = ip6gre_init_net,
1645         .exit_batch = ip6gre_exit_batch_net,
1646         .id   = &ip6gre_net_id,
1647         .size = sizeof(struct ip6gre_net),
1648 };
1649
1650 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1651                                   struct netlink_ext_ack *extack)
1652 {
1653         __be16 flags;
1654
1655         if (!data)
1656                 return 0;
1657
1658         flags = 0;
1659         if (data[IFLA_GRE_IFLAGS])
1660                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1661         if (data[IFLA_GRE_OFLAGS])
1662                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1663         if (flags & (GRE_VERSION|GRE_ROUTING))
1664                 return -EINVAL;
1665
1666         return 0;
1667 }
1668
1669 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1670                                struct netlink_ext_ack *extack)
1671 {
1672         struct in6_addr daddr;
1673
1674         if (tb[IFLA_ADDRESS]) {
1675                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1676                         return -EINVAL;
1677                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1678                         return -EADDRNOTAVAIL;
1679         }
1680
1681         if (!data)
1682                 goto out;
1683
1684         if (data[IFLA_GRE_REMOTE]) {
1685                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1686                 if (ipv6_addr_any(&daddr))
1687                         return -EINVAL;
1688         }
1689
1690 out:
1691         return ip6gre_tunnel_validate(tb, data, extack);
1692 }
1693
1694 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1695                                   struct netlink_ext_ack *extack)
1696 {
1697         __be16 flags = 0;
1698         int ret, ver = 0;
1699
1700         if (!data)
1701                 return 0;
1702
1703         ret = ip6gre_tap_validate(tb, data, extack);
1704         if (ret)
1705                 return ret;
1706
1707         /* ERSPAN should only have GRE sequence and key flag */
1708         if (data[IFLA_GRE_OFLAGS])
1709                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1710         if (data[IFLA_GRE_IFLAGS])
1711                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1712         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1713             flags != (GRE_SEQ | GRE_KEY))
1714                 return -EINVAL;
1715
1716         /* ERSPAN Session ID only has 10-bit. Since we reuse
1717          * 32-bit key field as ID, check it's range.
1718          */
1719         if (data[IFLA_GRE_IKEY] &&
1720             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1721                 return -EINVAL;
1722
1723         if (data[IFLA_GRE_OKEY] &&
1724             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1725                 return -EINVAL;
1726
1727         if (data[IFLA_GRE_ERSPAN_VER]) {
1728                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1729                 if (ver != 1 && ver != 2)
1730                         return -EINVAL;
1731         }
1732
1733         if (ver == 1) {
1734                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1735                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1736
1737                         if (index & ~INDEX_MASK)
1738                                 return -EINVAL;
1739                 }
1740         } else if (ver == 2) {
1741                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1742                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1743
1744                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1745                                 return -EINVAL;
1746                 }
1747
1748                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1749                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1750
1751                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1752                                 return -EINVAL;
1753                 }
1754         }
1755
1756         return 0;
1757 }
1758
1759 static void ip6erspan_set_version(struct nlattr *data[],
1760                                   struct __ip6_tnl_parm *parms)
1761 {
1762         if (!data)
1763                 return;
1764
1765         parms->erspan_ver = 1;
1766         if (data[IFLA_GRE_ERSPAN_VER])
1767                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1768
1769         if (parms->erspan_ver == 1) {
1770                 if (data[IFLA_GRE_ERSPAN_INDEX])
1771                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1772         } else if (parms->erspan_ver == 2) {
1773                 if (data[IFLA_GRE_ERSPAN_DIR])
1774                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1775                 if (data[IFLA_GRE_ERSPAN_HWID])
1776                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1777         }
1778 }
1779
1780 static void ip6gre_netlink_parms(struct nlattr *data[],
1781                                 struct __ip6_tnl_parm *parms)
1782 {
1783         memset(parms, 0, sizeof(*parms));
1784
1785         if (!data)
1786                 return;
1787
1788         if (data[IFLA_GRE_LINK])
1789                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1790
1791         if (data[IFLA_GRE_IFLAGS])
1792                 parms->i_flags = gre_flags_to_tnl_flags(
1793                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1794
1795         if (data[IFLA_GRE_OFLAGS])
1796                 parms->o_flags = gre_flags_to_tnl_flags(
1797                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1798
1799         if (data[IFLA_GRE_IKEY])
1800                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1801
1802         if (data[IFLA_GRE_OKEY])
1803                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1804
1805         if (data[IFLA_GRE_LOCAL])
1806                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1807
1808         if (data[IFLA_GRE_REMOTE])
1809                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1810
1811         if (data[IFLA_GRE_TTL])
1812                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1813
1814         if (data[IFLA_GRE_ENCAP_LIMIT])
1815                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1816
1817         if (data[IFLA_GRE_FLOWINFO])
1818                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1819
1820         if (data[IFLA_GRE_FLAGS])
1821                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1822
1823         if (data[IFLA_GRE_FWMARK])
1824                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1825
1826         if (data[IFLA_GRE_COLLECT_METADATA])
1827                 parms->collect_md = true;
1828 }
1829
1830 static int ip6gre_tap_init(struct net_device *dev)
1831 {
1832         int ret;
1833
1834         ret = ip6gre_tunnel_init_common(dev);
1835         if (ret)
1836                 return ret;
1837
1838         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1839
1840         return 0;
1841 }
1842
1843 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1844         .ndo_init = ip6gre_tap_init,
1845         .ndo_uninit = ip6gre_tunnel_uninit,
1846         .ndo_start_xmit = ip6gre_tunnel_xmit,
1847         .ndo_set_mac_address = eth_mac_addr,
1848         .ndo_validate_addr = eth_validate_addr,
1849         .ndo_change_mtu = ip6_tnl_change_mtu,
1850         .ndo_get_stats64 = dev_get_tstats64,
1851         .ndo_get_iflink = ip6_tnl_get_iflink,
1852 };
1853
1854 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1855 {
1856         int t_hlen;
1857
1858         tunnel->tun_hlen = 8;
1859         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1860                        erspan_hdr_len(tunnel->parms.erspan_ver);
1861
1862         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1863         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1864         return t_hlen;
1865 }
1866
1867 static int ip6erspan_tap_init(struct net_device *dev)
1868 {
1869         struct ip6_tnl *tunnel;
1870         int t_hlen;
1871         int ret;
1872
1873         tunnel = netdev_priv(dev);
1874
1875         tunnel->dev = dev;
1876         tunnel->net = dev_net(dev);
1877         strcpy(tunnel->parms.name, dev->name);
1878
1879         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1880         if (!dev->tstats)
1881                 return -ENOMEM;
1882
1883         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1884         if (ret)
1885                 goto cleanup_alloc_pcpu_stats;
1886
1887         ret = gro_cells_init(&tunnel->gro_cells, dev);
1888         if (ret)
1889                 goto cleanup_dst_cache_init;
1890
1891         t_hlen = ip6erspan_calc_hlen(tunnel);
1892         dev->mtu = ETH_DATA_LEN - t_hlen;
1893         if (dev->type == ARPHRD_ETHER)
1894                 dev->mtu -= ETH_HLEN;
1895         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1896                 dev->mtu -= 8;
1897
1898         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1899         ip6erspan_tnl_link_config(tunnel, 1);
1900
1901         netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1902         return 0;
1903
1904 cleanup_dst_cache_init:
1905         dst_cache_destroy(&tunnel->dst_cache);
1906 cleanup_alloc_pcpu_stats:
1907         free_percpu(dev->tstats);
1908         dev->tstats = NULL;
1909         return ret;
1910 }
1911
1912 static const struct net_device_ops ip6erspan_netdev_ops = {
1913         .ndo_init =             ip6erspan_tap_init,
1914         .ndo_uninit =           ip6erspan_tunnel_uninit,
1915         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1916         .ndo_set_mac_address =  eth_mac_addr,
1917         .ndo_validate_addr =    eth_validate_addr,
1918         .ndo_change_mtu =       ip6_tnl_change_mtu,
1919         .ndo_get_stats64 =      dev_get_tstats64,
1920         .ndo_get_iflink =       ip6_tnl_get_iflink,
1921 };
1922
1923 static void ip6gre_tap_setup(struct net_device *dev)
1924 {
1925
1926         ether_setup(dev);
1927
1928         dev->max_mtu = 0;
1929         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1930         dev->needs_free_netdev = true;
1931         dev->priv_destructor = ip6gre_dev_free;
1932
1933         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1934         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1935         netif_keep_dst(dev);
1936 }
1937
1938 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1939                                        struct ip_tunnel_encap *ipencap)
1940 {
1941         bool ret = false;
1942
1943         memset(ipencap, 0, sizeof(*ipencap));
1944
1945         if (!data)
1946                 return ret;
1947
1948         if (data[IFLA_GRE_ENCAP_TYPE]) {
1949                 ret = true;
1950                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1951         }
1952
1953         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1954                 ret = true;
1955                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1956         }
1957
1958         if (data[IFLA_GRE_ENCAP_SPORT]) {
1959                 ret = true;
1960                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1961         }
1962
1963         if (data[IFLA_GRE_ENCAP_DPORT]) {
1964                 ret = true;
1965                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1966         }
1967
1968         return ret;
1969 }
1970
1971 static int ip6gre_newlink_common(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;
1976         struct ip_tunnel_encap ipencap;
1977         int err;
1978
1979         nt = netdev_priv(dev);
1980
1981         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1982                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1983
1984                 if (err < 0)
1985                         return err;
1986         }
1987
1988         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1989                 eth_hw_addr_random(dev);
1990
1991         nt->dev = dev;
1992         nt->net = dev_net(dev);
1993
1994         err = register_netdevice(dev);
1995         if (err)
1996                 goto out;
1997
1998         if (tb[IFLA_MTU])
1999                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2000
2001 out:
2002         return err;
2003 }
2004
2005 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2006                           struct nlattr *tb[], struct nlattr *data[],
2007                           struct netlink_ext_ack *extack)
2008 {
2009         struct ip6_tnl *nt = netdev_priv(dev);
2010         struct net *net = dev_net(dev);
2011         struct ip6gre_net *ign;
2012         int err;
2013
2014         ip6gre_netlink_parms(data, &nt->parms);
2015         ign = net_generic(net, ip6gre_net_id);
2016
2017         if (nt->parms.collect_md) {
2018                 if (rtnl_dereference(ign->collect_md_tun))
2019                         return -EEXIST;
2020         } else {
2021                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2022                         return -EEXIST;
2023         }
2024
2025         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2026         if (!err) {
2027                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2028                 ip6gre_tunnel_link_md(ign, nt);
2029                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2030         }
2031         return err;
2032 }
2033
2034 static struct ip6_tnl *
2035 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2036                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2037                          struct netlink_ext_ack *extack)
2038 {
2039         struct ip6_tnl *t, *nt = netdev_priv(dev);
2040         struct net *net = nt->net;
2041         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2042         struct ip_tunnel_encap ipencap;
2043
2044         if (dev == ign->fb_tunnel_dev)
2045                 return ERR_PTR(-EINVAL);
2046
2047         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2048                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2049
2050                 if (err < 0)
2051                         return ERR_PTR(err);
2052         }
2053
2054         ip6gre_netlink_parms(data, p_p);
2055
2056         t = ip6gre_tunnel_locate(net, p_p, 0);
2057
2058         if (t) {
2059                 if (t->dev != dev)
2060                         return ERR_PTR(-EEXIST);
2061         } else {
2062                 t = nt;
2063         }
2064
2065         return t;
2066 }
2067
2068 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2069                              struct nlattr *data[],
2070                              struct netlink_ext_ack *extack)
2071 {
2072         struct ip6_tnl *t = netdev_priv(dev);
2073         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2074         struct __ip6_tnl_parm p;
2075
2076         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2077         if (IS_ERR(t))
2078                 return PTR_ERR(t);
2079
2080         ip6gre_tunnel_unlink_md(ign, t);
2081         ip6gre_tunnel_unlink(ign, t);
2082         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2083         ip6gre_tunnel_link_md(ign, t);
2084         ip6gre_tunnel_link(ign, t);
2085         return 0;
2086 }
2087
2088 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2089 {
2090         struct net *net = dev_net(dev);
2091         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2092
2093         if (dev != ign->fb_tunnel_dev)
2094                 unregister_netdevice_queue(dev, head);
2095 }
2096
2097 static size_t ip6gre_get_size(const struct net_device *dev)
2098 {
2099         return
2100                 /* IFLA_GRE_LINK */
2101                 nla_total_size(4) +
2102                 /* IFLA_GRE_IFLAGS */
2103                 nla_total_size(2) +
2104                 /* IFLA_GRE_OFLAGS */
2105                 nla_total_size(2) +
2106                 /* IFLA_GRE_IKEY */
2107                 nla_total_size(4) +
2108                 /* IFLA_GRE_OKEY */
2109                 nla_total_size(4) +
2110                 /* IFLA_GRE_LOCAL */
2111                 nla_total_size(sizeof(struct in6_addr)) +
2112                 /* IFLA_GRE_REMOTE */
2113                 nla_total_size(sizeof(struct in6_addr)) +
2114                 /* IFLA_GRE_TTL */
2115                 nla_total_size(1) +
2116                 /* IFLA_GRE_ENCAP_LIMIT */
2117                 nla_total_size(1) +
2118                 /* IFLA_GRE_FLOWINFO */
2119                 nla_total_size(4) +
2120                 /* IFLA_GRE_FLAGS */
2121                 nla_total_size(4) +
2122                 /* IFLA_GRE_ENCAP_TYPE */
2123                 nla_total_size(2) +
2124                 /* IFLA_GRE_ENCAP_FLAGS */
2125                 nla_total_size(2) +
2126                 /* IFLA_GRE_ENCAP_SPORT */
2127                 nla_total_size(2) +
2128                 /* IFLA_GRE_ENCAP_DPORT */
2129                 nla_total_size(2) +
2130                 /* IFLA_GRE_COLLECT_METADATA */
2131                 nla_total_size(0) +
2132                 /* IFLA_GRE_FWMARK */
2133                 nla_total_size(4) +
2134                 /* IFLA_GRE_ERSPAN_INDEX */
2135                 nla_total_size(4) +
2136                 0;
2137 }
2138
2139 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2140 {
2141         struct ip6_tnl *t = netdev_priv(dev);
2142         struct __ip6_tnl_parm *p = &t->parms;
2143         __be16 o_flags = p->o_flags;
2144
2145         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2146                 if (!p->collect_md)
2147                         o_flags |= TUNNEL_KEY;
2148
2149                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2150                         goto nla_put_failure;
2151
2152                 if (p->erspan_ver == 1) {
2153                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2154                                 goto nla_put_failure;
2155                 } else {
2156                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2157                                 goto nla_put_failure;
2158                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2159                                 goto nla_put_failure;
2160                 }
2161         }
2162
2163         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2164             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2165                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2166             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2167                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2168             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2169             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2170             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2171             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2172             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2173             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2174             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2175             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2176             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2177                 goto nla_put_failure;
2178
2179         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2180                         t->encap.type) ||
2181             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2182                          t->encap.sport) ||
2183             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2184                          t->encap.dport) ||
2185             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2186                         t->encap.flags))
2187                 goto nla_put_failure;
2188
2189         if (p->collect_md) {
2190                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2191                         goto nla_put_failure;
2192         }
2193
2194         return 0;
2195
2196 nla_put_failure:
2197         return -EMSGSIZE;
2198 }
2199
2200 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2201         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2202         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2203         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2204         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2205         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2206         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2207         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2208         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2209         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2210         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2211         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2212         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2213         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2214         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2215         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2216         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2217         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2218         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2219         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2220         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2221         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2222 };
2223
2224 static void ip6erspan_tap_setup(struct net_device *dev)
2225 {
2226         ether_setup(dev);
2227
2228         dev->max_mtu = 0;
2229         dev->netdev_ops = &ip6erspan_netdev_ops;
2230         dev->needs_free_netdev = true;
2231         dev->priv_destructor = ip6gre_dev_free;
2232
2233         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2234         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2235         netif_keep_dst(dev);
2236 }
2237
2238 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2239                              struct nlattr *tb[], struct nlattr *data[],
2240                              struct netlink_ext_ack *extack)
2241 {
2242         struct ip6_tnl *nt = netdev_priv(dev);
2243         struct net *net = dev_net(dev);
2244         struct ip6gre_net *ign;
2245         int err;
2246
2247         ip6gre_netlink_parms(data, &nt->parms);
2248         ip6erspan_set_version(data, &nt->parms);
2249         ign = net_generic(net, ip6gre_net_id);
2250
2251         if (nt->parms.collect_md) {
2252                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2253                         return -EEXIST;
2254         } else {
2255                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2256                         return -EEXIST;
2257         }
2258
2259         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2260         if (!err) {
2261                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2262                 ip6erspan_tunnel_link_md(ign, nt);
2263                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2264         }
2265         return err;
2266 }
2267
2268 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2269 {
2270         ip6gre_tnl_link_config_common(t);
2271         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2272 }
2273
2274 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2275                                 const struct __ip6_tnl_parm *p, int set_mtu)
2276 {
2277         ip6gre_tnl_copy_tnl_parm(t, p);
2278         ip6erspan_tnl_link_config(t, set_mtu);
2279         return 0;
2280 }
2281
2282 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2283                                 struct nlattr *data[],
2284                                 struct netlink_ext_ack *extack)
2285 {
2286         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2287         struct __ip6_tnl_parm p;
2288         struct ip6_tnl *t;
2289
2290         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2291         if (IS_ERR(t))
2292                 return PTR_ERR(t);
2293
2294         ip6erspan_set_version(data, &p);
2295         ip6gre_tunnel_unlink_md(ign, t);
2296         ip6gre_tunnel_unlink(ign, t);
2297         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2298         ip6erspan_tunnel_link_md(ign, t);
2299         ip6gre_tunnel_link(ign, t);
2300         return 0;
2301 }
2302
2303 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2304         .kind           = "ip6gre",
2305         .maxtype        = IFLA_GRE_MAX,
2306         .policy         = ip6gre_policy,
2307         .priv_size      = sizeof(struct ip6_tnl),
2308         .setup          = ip6gre_tunnel_setup,
2309         .validate       = ip6gre_tunnel_validate,
2310         .newlink        = ip6gre_newlink,
2311         .changelink     = ip6gre_changelink,
2312         .dellink        = ip6gre_dellink,
2313         .get_size       = ip6gre_get_size,
2314         .fill_info      = ip6gre_fill_info,
2315         .get_link_net   = ip6_tnl_get_link_net,
2316 };
2317
2318 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2319         .kind           = "ip6gretap",
2320         .maxtype        = IFLA_GRE_MAX,
2321         .policy         = ip6gre_policy,
2322         .priv_size      = sizeof(struct ip6_tnl),
2323         .setup          = ip6gre_tap_setup,
2324         .validate       = ip6gre_tap_validate,
2325         .newlink        = ip6gre_newlink,
2326         .changelink     = ip6gre_changelink,
2327         .get_size       = ip6gre_get_size,
2328         .fill_info      = ip6gre_fill_info,
2329         .get_link_net   = ip6_tnl_get_link_net,
2330 };
2331
2332 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2333         .kind           = "ip6erspan",
2334         .maxtype        = IFLA_GRE_MAX,
2335         .policy         = ip6gre_policy,
2336         .priv_size      = sizeof(struct ip6_tnl),
2337         .setup          = ip6erspan_tap_setup,
2338         .validate       = ip6erspan_tap_validate,
2339         .newlink        = ip6erspan_newlink,
2340         .changelink     = ip6erspan_changelink,
2341         .get_size       = ip6gre_get_size,
2342         .fill_info      = ip6gre_fill_info,
2343         .get_link_net   = ip6_tnl_get_link_net,
2344 };
2345
2346 /*
2347  *      And now the modules code and kernel interface.
2348  */
2349
2350 static int __init ip6gre_init(void)
2351 {
2352         int err;
2353
2354         pr_info("GRE over IPv6 tunneling driver\n");
2355
2356         err = register_pernet_device(&ip6gre_net_ops);
2357         if (err < 0)
2358                 return err;
2359
2360         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2361         if (err < 0) {
2362                 pr_info("%s: can't add protocol\n", __func__);
2363                 goto add_proto_failed;
2364         }
2365
2366         err = rtnl_link_register(&ip6gre_link_ops);
2367         if (err < 0)
2368                 goto rtnl_link_failed;
2369
2370         err = rtnl_link_register(&ip6gre_tap_ops);
2371         if (err < 0)
2372                 goto tap_ops_failed;
2373
2374         err = rtnl_link_register(&ip6erspan_tap_ops);
2375         if (err < 0)
2376                 goto erspan_link_failed;
2377
2378 out:
2379         return err;
2380
2381 erspan_link_failed:
2382         rtnl_link_unregister(&ip6gre_tap_ops);
2383 tap_ops_failed:
2384         rtnl_link_unregister(&ip6gre_link_ops);
2385 rtnl_link_failed:
2386         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2387 add_proto_failed:
2388         unregister_pernet_device(&ip6gre_net_ops);
2389         goto out;
2390 }
2391
2392 static void __exit ip6gre_fini(void)
2393 {
2394         rtnl_link_unregister(&ip6gre_tap_ops);
2395         rtnl_link_unregister(&ip6gre_link_ops);
2396         rtnl_link_unregister(&ip6erspan_tap_ops);
2397         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2398         unregister_pernet_device(&ip6gre_net_ops);
2399 }
2400
2401 module_init(ip6gre_init);
2402 module_exit(ip6gre_fini);
2403 MODULE_LICENSE("GPL");
2404 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2405 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2406 MODULE_ALIAS_RTNL_LINK("ip6gre");
2407 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2408 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2409 MODULE_ALIAS_NETDEV("ip6gre0");