Merge tag 'irq-urgent-2020-12-06' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6-microblaze.git] / net / ipv6 / ip6_gre.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      GRE over IPv6 protocol decoder.
4  *
5  *      Authors: Dmitry Kozlov (xeb@mail.ru)
6  */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
18 #include <linux/in.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/rtnetlink.h>
47
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/gre.h>
53 #include <net/erspan.h>
54 #include <net/dst_metadata.h>
55
56
57 static bool log_ecn_error = true;
58 module_param(log_ecn_error, bool, 0644);
59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60
61 #define IP6_GRE_HASH_SIZE_SHIFT  5
62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63
64 static unsigned int ip6gre_net_id __read_mostly;
65 struct ip6gre_net {
66         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67
68         struct ip6_tnl __rcu *collect_md_tun;
69         struct ip6_tnl __rcu *collect_md_tun_erspan;
70         struct net_device *fb_tunnel_dev;
71 };
72
73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85    4 hash tables:
86
87    3: (remote,local)
88    2: (remote,*)
89    1: (*,local)
90    0: (*,*)
91
92    We require exact key match i.e. if a key is present in packet
93    it will match only tunnel with the same key; if it is not present,
94    it will match only keyless tunnel.
95
96    All keysless packets, if not matched configured keyless tunnels
97    will match fallback tunnel.
98  */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103         u32 hash = ipv6_addr_hash(addr);
104
105         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l     tunnels[3]
109 #define tunnels_r       tunnels[2]
110 #define tunnels_l       tunnels[1]
111 #define tunnels_wc      tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116                 const struct in6_addr *remote, const struct in6_addr *local,
117                 __be32 key, __be16 gre_proto)
118 {
119         struct net *net = dev_net(dev);
120         int link = dev->ifindex;
121         unsigned int h0 = HASH_ADDR(remote);
122         unsigned int h1 = HASH_KEY(key);
123         struct ip6_tnl *t, *cand = NULL;
124         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126                         gre_proto == htons(ETH_P_ERSPAN) ||
127                         gre_proto == htons(ETH_P_ERSPAN2)) ?
128                        ARPHRD_ETHER : ARPHRD_IP6GRE;
129         int score, cand_score = 4;
130         struct net_device *ndev;
131
132         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
133                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
134                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
135                     key != t->parms.i_key ||
136                     !(t->dev->flags & IFF_UP))
137                         continue;
138
139                 if (t->dev->type != ARPHRD_IP6GRE &&
140                     t->dev->type != dev_type)
141                         continue;
142
143                 score = 0;
144                 if (t->parms.link != link)
145                         score |= 1;
146                 if (t->dev->type != dev_type)
147                         score |= 2;
148                 if (score == 0)
149                         return t;
150
151                 if (score < cand_score) {
152                         cand = t;
153                         cand_score = score;
154                 }
155         }
156
157         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
158                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
159                     key != t->parms.i_key ||
160                     !(t->dev->flags & IFF_UP))
161                         continue;
162
163                 if (t->dev->type != ARPHRD_IP6GRE &&
164                     t->dev->type != dev_type)
165                         continue;
166
167                 score = 0;
168                 if (t->parms.link != link)
169                         score |= 1;
170                 if (t->dev->type != dev_type)
171                         score |= 2;
172                 if (score == 0)
173                         return t;
174
175                 if (score < cand_score) {
176                         cand = t;
177                         cand_score = score;
178                 }
179         }
180
181         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
182                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
183                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
184                                  !ipv6_addr_is_multicast(local))) ||
185                     key != t->parms.i_key ||
186                     !(t->dev->flags & IFF_UP))
187                         continue;
188
189                 if (t->dev->type != ARPHRD_IP6GRE &&
190                     t->dev->type != dev_type)
191                         continue;
192
193                 score = 0;
194                 if (t->parms.link != link)
195                         score |= 1;
196                 if (t->dev->type != dev_type)
197                         score |= 2;
198                 if (score == 0)
199                         return t;
200
201                 if (score < cand_score) {
202                         cand = t;
203                         cand_score = score;
204                 }
205         }
206
207         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
208                 if (t->parms.i_key != key ||
209                     !(t->dev->flags & IFF_UP))
210                         continue;
211
212                 if (t->dev->type != ARPHRD_IP6GRE &&
213                     t->dev->type != dev_type)
214                         continue;
215
216                 score = 0;
217                 if (t->parms.link != link)
218                         score |= 1;
219                 if (t->dev->type != dev_type)
220                         score |= 2;
221                 if (score == 0)
222                         return t;
223
224                 if (score < cand_score) {
225                         cand = t;
226                         cand_score = score;
227                 }
228         }
229
230         if (cand)
231                 return cand;
232
233         if (gre_proto == htons(ETH_P_ERSPAN) ||
234             gre_proto == htons(ETH_P_ERSPAN2))
235                 t = rcu_dereference(ign->collect_md_tun_erspan);
236         else
237                 t = rcu_dereference(ign->collect_md_tun);
238
239         if (t && t->dev->flags & IFF_UP)
240                 return t;
241
242         ndev = READ_ONCE(ign->fb_tunnel_dev);
243         if (ndev && ndev->flags & IFF_UP)
244                 return netdev_priv(ndev);
245
246         return NULL;
247 }
248
249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
250                 const struct __ip6_tnl_parm *p)
251 {
252         const struct in6_addr *remote = &p->raddr;
253         const struct in6_addr *local = &p->laddr;
254         unsigned int h = HASH_KEY(p->i_key);
255         int prio = 0;
256
257         if (!ipv6_addr_any(local))
258                 prio |= 1;
259         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
260                 prio |= 2;
261                 h ^= HASH_ADDR(remote);
262         }
263
264         return &ign->tunnels[prio][h];
265 }
266
267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269         if (t->parms.collect_md)
270                 rcu_assign_pointer(ign->collect_md_tun, t);
271 }
272
273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
274 {
275         if (t->parms.collect_md)
276                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
277 }
278
279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
280 {
281         if (t->parms.collect_md)
282                 rcu_assign_pointer(ign->collect_md_tun, NULL);
283 }
284
285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
286                                        struct ip6_tnl *t)
287 {
288         if (t->parms.collect_md)
289                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
290 }
291
292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
293                 const struct ip6_tnl *t)
294 {
295         return __ip6gre_bucket(ign, &t->parms);
296 }
297
298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
299 {
300         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
301
302         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
303         rcu_assign_pointer(*tp, t);
304 }
305
306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
307 {
308         struct ip6_tnl __rcu **tp;
309         struct ip6_tnl *iter;
310
311         for (tp = ip6gre_bucket(ign, t);
312              (iter = rtnl_dereference(*tp)) != NULL;
313              tp = &iter->next) {
314                 if (t == iter) {
315                         rcu_assign_pointer(*tp, t->next);
316                         break;
317                 }
318         }
319 }
320
321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
322                                            const struct __ip6_tnl_parm *parms,
323                                            int type)
324 {
325         const struct in6_addr *remote = &parms->raddr;
326         const struct in6_addr *local = &parms->laddr;
327         __be32 key = parms->i_key;
328         int link = parms->link;
329         struct ip6_tnl *t;
330         struct ip6_tnl __rcu **tp;
331         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
332
333         for (tp = __ip6gre_bucket(ign, parms);
334              (t = rtnl_dereference(*tp)) != NULL;
335              tp = &t->next)
336                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
337                     ipv6_addr_equal(remote, &t->parms.raddr) &&
338                     key == t->parms.i_key &&
339                     link == t->parms.link &&
340                     type == t->dev->type)
341                         break;
342
343         return t;
344 }
345
346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
347                 const struct __ip6_tnl_parm *parms, int create)
348 {
349         struct ip6_tnl *t, *nt;
350         struct net_device *dev;
351         char name[IFNAMSIZ];
352         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
353
354         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
355         if (t && create)
356                 return NULL;
357         if (t || !create)
358                 return t;
359
360         if (parms->name[0]) {
361                 if (!dev_valid_name(parms->name))
362                         return NULL;
363                 strlcpy(name, parms->name, IFNAMSIZ);
364         } else {
365                 strcpy(name, "ip6gre%d");
366         }
367         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
368                            ip6gre_tunnel_setup);
369         if (!dev)
370                 return NULL;
371
372         dev_net_set(dev, net);
373
374         nt = netdev_priv(dev);
375         nt->parms = *parms;
376         dev->rtnl_link_ops = &ip6gre_link_ops;
377
378         nt->dev = dev;
379         nt->net = dev_net(dev);
380
381         if (register_netdevice(dev) < 0)
382                 goto failed_free;
383
384         ip6gre_tnl_link_config(nt, 1);
385
386         /* Can use a lockless transmit, unless we generate output sequences */
387         if (!(nt->parms.o_flags & TUNNEL_SEQ))
388                 dev->features |= NETIF_F_LLTX;
389
390         dev_hold(dev);
391         ip6gre_tunnel_link(ign, nt);
392         return nt;
393
394 failed_free:
395         free_netdev(dev);
396         return NULL;
397 }
398
399 static void ip6erspan_tunnel_uninit(struct net_device *dev)
400 {
401         struct ip6_tnl *t = netdev_priv(dev);
402         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
403
404         ip6erspan_tunnel_unlink_md(ign, t);
405         ip6gre_tunnel_unlink(ign, t);
406         dst_cache_reset(&t->dst_cache);
407         dev_put(dev);
408 }
409
410 static void ip6gre_tunnel_uninit(struct net_device *dev)
411 {
412         struct ip6_tnl *t = netdev_priv(dev);
413         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
414
415         ip6gre_tunnel_unlink_md(ign, t);
416         ip6gre_tunnel_unlink(ign, t);
417         if (ign->fb_tunnel_dev == dev)
418                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
419         dst_cache_reset(&t->dst_cache);
420         dev_put(dev);
421 }
422
423
424 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
425                        u8 type, u8 code, int offset, __be32 info)
426 {
427         struct net *net = dev_net(skb->dev);
428         const struct ipv6hdr *ipv6h;
429         struct tnl_ptk_info tpi;
430         struct ip6_tnl *t;
431
432         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
433                              offset) < 0)
434                 return -EINVAL;
435
436         ipv6h = (const struct ipv6hdr *)skb->data;
437         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
438                                  tpi.key, tpi.proto);
439         if (!t)
440                 return -ENOENT;
441
442         switch (type) {
443         case ICMPV6_DEST_UNREACH:
444                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
445                                     t->parms.name);
446                 if (code != ICMPV6_PORT_UNREACH)
447                         break;
448                 return 0;
449         case ICMPV6_TIME_EXCEED:
450                 if (code == ICMPV6_EXC_HOPLIMIT) {
451                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
452                                             t->parms.name);
453                         break;
454                 }
455                 return 0;
456         case ICMPV6_PARAMPROB: {
457                 struct ipv6_tlv_tnl_enc_lim *tel;
458                 __u32 teli;
459
460                 teli = 0;
461                 if (code == ICMPV6_HDR_FIELD)
462                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
463
464                 if (teli && teli == be32_to_cpu(info) - 2) {
465                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
466                         if (tel->encap_limit == 0) {
467                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
468                                                     t->parms.name);
469                         }
470                 } else {
471                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
472                                             t->parms.name);
473                 }
474                 return 0;
475         }
476         case ICMPV6_PKT_TOOBIG:
477                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
478                 return 0;
479         case NDISC_REDIRECT:
480                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
481                              sock_net_uid(net, NULL));
482                 return 0;
483         }
484
485         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
486                 t->err_count++;
487         else
488                 t->err_count = 1;
489         t->err_time = jiffies;
490
491         return 0;
492 }
493
494 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
495 {
496         const struct ipv6hdr *ipv6h;
497         struct ip6_tnl *tunnel;
498
499         ipv6h = ipv6_hdr(skb);
500         tunnel = ip6gre_tunnel_lookup(skb->dev,
501                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
502                                       tpi->proto);
503         if (tunnel) {
504                 if (tunnel->parms.collect_md) {
505                         struct metadata_dst *tun_dst;
506                         __be64 tun_id;
507                         __be16 flags;
508
509                         flags = tpi->flags;
510                         tun_id = key32_to_tunnel_id(tpi->key);
511
512                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
513                         if (!tun_dst)
514                                 return PACKET_REJECT;
515
516                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
517                 } else {
518                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
519                 }
520
521                 return PACKET_RCVD;
522         }
523
524         return PACKET_REJECT;
525 }
526
527 static int ip6erspan_rcv(struct sk_buff *skb,
528                          struct tnl_ptk_info *tpi,
529                          int gre_hdr_len)
530 {
531         struct erspan_base_hdr *ershdr;
532         const struct ipv6hdr *ipv6h;
533         struct erspan_md2 *md2;
534         struct ip6_tnl *tunnel;
535         u8 ver;
536
537         ipv6h = ipv6_hdr(skb);
538         ershdr = (struct erspan_base_hdr *)skb->data;
539         ver = ershdr->ver;
540
541         tunnel = ip6gre_tunnel_lookup(skb->dev,
542                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
543                                       tpi->proto);
544         if (tunnel) {
545                 int len = erspan_hdr_len(ver);
546
547                 if (unlikely(!pskb_may_pull(skb, len)))
548                         return PACKET_REJECT;
549
550                 if (__iptunnel_pull_header(skb, len,
551                                            htons(ETH_P_TEB),
552                                            false, false) < 0)
553                         return PACKET_REJECT;
554
555                 if (tunnel->parms.collect_md) {
556                         struct erspan_metadata *pkt_md, *md;
557                         struct metadata_dst *tun_dst;
558                         struct ip_tunnel_info *info;
559                         unsigned char *gh;
560                         __be64 tun_id;
561                         __be16 flags;
562
563                         tpi->flags |= TUNNEL_KEY;
564                         flags = tpi->flags;
565                         tun_id = key32_to_tunnel_id(tpi->key);
566
567                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
568                                                   sizeof(*md));
569                         if (!tun_dst)
570                                 return PACKET_REJECT;
571
572                         /* skb can be uncloned in __iptunnel_pull_header, so
573                          * old pkt_md is no longer valid and we need to reset
574                          * it
575                          */
576                         gh = skb_network_header(skb) +
577                              skb_network_header_len(skb);
578                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
579                                                             sizeof(*ershdr));
580                         info = &tun_dst->u.tun_info;
581                         md = ip_tunnel_info_opts(info);
582                         md->version = ver;
583                         md2 = &md->u.md2;
584                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
585                                                        ERSPAN_V2_MDSIZE);
586                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
587                         info->options_len = sizeof(*md);
588
589                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
590
591                 } else {
592                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
593                 }
594
595                 return PACKET_RCVD;
596         }
597
598         return PACKET_REJECT;
599 }
600
601 static int gre_rcv(struct sk_buff *skb)
602 {
603         struct tnl_ptk_info tpi;
604         bool csum_err = false;
605         int hdr_len;
606
607         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
608         if (hdr_len < 0)
609                 goto drop;
610
611         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
612                 goto drop;
613
614         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
615                      tpi.proto == htons(ETH_P_ERSPAN2))) {
616                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
617                         return 0;
618                 goto out;
619         }
620
621         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
622                 return 0;
623
624 out:
625         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
626 drop:
627         kfree_skb(skb);
628         return 0;
629 }
630
631 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
632 {
633         return iptunnel_handle_offloads(skb,
634                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
635 }
636
637 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
638                                      struct net_device *dev,
639                                      struct flowi6 *fl6, __u8 *dsfield,
640                                      int *encap_limit)
641 {
642         const struct iphdr *iph = ip_hdr(skb);
643         struct ip6_tnl *t = netdev_priv(dev);
644
645         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
646                 *encap_limit = t->parms.encap_limit;
647
648         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
649
650         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
651                 *dsfield = ipv4_get_dsfield(iph);
652         else
653                 *dsfield = ip6_tclass(t->parms.flowinfo);
654
655         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
656                 fl6->flowi6_mark = skb->mark;
657         else
658                 fl6->flowi6_mark = t->parms.fwmark;
659
660         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
661 }
662
663 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
664                                     struct net_device *dev,
665                                     struct flowi6 *fl6, __u8 *dsfield,
666                                     int *encap_limit)
667 {
668         struct ipv6hdr *ipv6h;
669         struct ip6_tnl *t = netdev_priv(dev);
670         __u16 offset;
671
672         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
673         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
674         ipv6h = ipv6_hdr(skb);
675
676         if (offset > 0) {
677                 struct ipv6_tlv_tnl_enc_lim *tel;
678
679                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
680                 if (tel->encap_limit == 0) {
681                         icmpv6_send(skb, ICMPV6_PARAMPROB,
682                                     ICMPV6_HDR_FIELD, offset + 2);
683                         return -1;
684                 }
685                 *encap_limit = tel->encap_limit - 1;
686         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
687                 *encap_limit = t->parms.encap_limit;
688         }
689
690         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
691
692         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
693                 *dsfield = ipv6_get_dsfield(ipv6h);
694         else
695                 *dsfield = ip6_tclass(t->parms.flowinfo);
696
697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
698                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
699
700         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
701                 fl6->flowi6_mark = skb->mark;
702         else
703                 fl6->flowi6_mark = t->parms.fwmark;
704
705         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
706
707         return 0;
708 }
709
710 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
711 {
712         struct ip_tunnel_info *tun_info;
713
714         tun_info = skb_tunnel_info(skb);
715         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
716                 return ERR_PTR(-EINVAL);
717
718         return tun_info;
719 }
720
721 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
722                                struct net_device *dev, __u8 dsfield,
723                                struct flowi6 *fl6, int encap_limit,
724                                __u32 *pmtu, __be16 proto)
725 {
726         struct ip6_tnl *tunnel = netdev_priv(dev);
727         __be16 protocol;
728
729         if (dev->type == ARPHRD_ETHER)
730                 IPCB(skb)->flags = 0;
731
732         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
733                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
734         else
735                 fl6->daddr = tunnel->parms.raddr;
736
737         if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
738                 return -ENOMEM;
739
740         /* Push GRE header. */
741         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
742
743         if (tunnel->parms.collect_md) {
744                 struct ip_tunnel_info *tun_info;
745                 const struct ip_tunnel_key *key;
746                 __be16 flags;
747
748                 tun_info = skb_tunnel_info_txcheck(skb);
749                 if (IS_ERR(tun_info) ||
750                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
751                         return -EINVAL;
752
753                 key = &tun_info->key;
754                 memset(fl6, 0, sizeof(*fl6));
755                 fl6->flowi6_proto = IPPROTO_GRE;
756                 fl6->daddr = key->u.ipv6.dst;
757                 fl6->flowlabel = key->label;
758                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
759
760                 dsfield = key->tos;
761                 flags = key->tun_flags &
762                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
763                 tunnel->tun_hlen = gre_calc_hlen(flags);
764
765                 gre_build_header(skb, tunnel->tun_hlen,
766                                  flags, protocol,
767                                  tunnel_id_to_key32(tun_info->key.tun_id),
768                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
769                                                       : 0);
770
771         } else {
772                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
773                         tunnel->o_seqno++;
774
775                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
776                                  protocol, tunnel->parms.o_key,
777                                  htonl(tunnel->o_seqno));
778         }
779
780         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
781                             NEXTHDR_GRE);
782 }
783
784 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
785 {
786         struct ip6_tnl *t = netdev_priv(dev);
787         int encap_limit = -1;
788         struct flowi6 fl6;
789         __u8 dsfield = 0;
790         __u32 mtu;
791         int err;
792
793         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
794
795         if (!t->parms.collect_md)
796                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
797                                          &dsfield, &encap_limit);
798
799         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
800         if (err)
801                 return -1;
802
803         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
804                           skb->protocol);
805         if (err != 0) {
806                 /* XXX: send ICMP error even if DF is not set. */
807                 if (err == -EMSGSIZE)
808                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
809                                   htonl(mtu));
810                 return -1;
811         }
812
813         return 0;
814 }
815
816 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
817 {
818         struct ip6_tnl *t = netdev_priv(dev);
819         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
820         int encap_limit = -1;
821         struct flowi6 fl6;
822         __u8 dsfield = 0;
823         __u32 mtu;
824         int err;
825
826         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
827                 return -1;
828
829         if (!t->parms.collect_md &&
830             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
831                 return -1;
832
833         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
834                 return -1;
835
836         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
837                           &mtu, skb->protocol);
838         if (err != 0) {
839                 if (err == -EMSGSIZE)
840                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
841                 return -1;
842         }
843
844         return 0;
845 }
846
847 /**
848  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
849  *   @t: the outgoing tunnel device
850  *   @hdr: IPv6 header from the incoming packet
851  *
852  * Description:
853  *   Avoid trivial tunneling loop by checking that tunnel exit-point
854  *   doesn't match source of incoming packet.
855  *
856  * Return:
857  *   1 if conflict,
858  *   0 else
859  **/
860
861 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
862         const struct ipv6hdr *hdr)
863 {
864         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
865 }
866
867 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
868 {
869         struct ip6_tnl *t = netdev_priv(dev);
870         int encap_limit = -1;
871         struct flowi6 fl6;
872         __u32 mtu;
873         int err;
874
875         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
876                 encap_limit = t->parms.encap_limit;
877
878         if (!t->parms.collect_md)
879                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
880
881         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
882         if (err)
883                 return err;
884
885         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
886
887         return err;
888 }
889
890 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
891         struct net_device *dev)
892 {
893         struct ip6_tnl *t = netdev_priv(dev);
894         struct net_device_stats *stats = &t->dev->stats;
895         int ret;
896
897         if (!pskb_inet_may_pull(skb))
898                 goto tx_err;
899
900         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
901                 goto tx_err;
902
903         switch (skb->protocol) {
904         case htons(ETH_P_IP):
905                 ret = ip6gre_xmit_ipv4(skb, dev);
906                 break;
907         case htons(ETH_P_IPV6):
908                 ret = ip6gre_xmit_ipv6(skb, dev);
909                 break;
910         default:
911                 ret = ip6gre_xmit_other(skb, dev);
912                 break;
913         }
914
915         if (ret < 0)
916                 goto tx_err;
917
918         return NETDEV_TX_OK;
919
920 tx_err:
921         if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
922                 stats->tx_errors++;
923         stats->tx_dropped++;
924         kfree_skb(skb);
925         return NETDEV_TX_OK;
926 }
927
928 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
929                                          struct net_device *dev)
930 {
931         struct ip_tunnel_info *tun_info = NULL;
932         struct ip6_tnl *t = netdev_priv(dev);
933         struct dst_entry *dst = skb_dst(skb);
934         struct net_device_stats *stats;
935         bool truncate = false;
936         int encap_limit = -1;
937         __u8 dsfield = false;
938         struct flowi6 fl6;
939         int err = -EINVAL;
940         __be16 proto;
941         __u32 mtu;
942         int nhoff;
943         int thoff;
944
945         if (!pskb_inet_may_pull(skb))
946                 goto tx_err;
947
948         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
949                 goto tx_err;
950
951         if (gre_handle_offloads(skb, false))
952                 goto tx_err;
953
954         if (skb->len > dev->mtu + dev->hard_header_len) {
955                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
956                 truncate = true;
957         }
958
959         nhoff = skb_network_header(skb) - skb_mac_header(skb);
960         if (skb->protocol == htons(ETH_P_IP) &&
961             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
962                 truncate = true;
963
964         thoff = skb_transport_header(skb) - skb_mac_header(skb);
965         if (skb->protocol == htons(ETH_P_IPV6) &&
966             (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
967                 truncate = true;
968
969         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
970                 goto tx_err;
971
972         t->parms.o_flags &= ~TUNNEL_KEY;
973         IPCB(skb)->flags = 0;
974
975         /* For collect_md mode, derive fl6 from the tunnel key,
976          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
977          */
978         if (t->parms.collect_md) {
979                 const struct ip_tunnel_key *key;
980                 struct erspan_metadata *md;
981                 __be32 tun_id;
982
983                 tun_info = skb_tunnel_info_txcheck(skb);
984                 if (IS_ERR(tun_info) ||
985                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
986                         goto tx_err;
987
988                 key = &tun_info->key;
989                 memset(&fl6, 0, sizeof(fl6));
990                 fl6.flowi6_proto = IPPROTO_GRE;
991                 fl6.daddr = key->u.ipv6.dst;
992                 fl6.flowlabel = key->label;
993                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
994
995                 dsfield = key->tos;
996                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
997                         goto tx_err;
998                 if (tun_info->options_len < sizeof(*md))
999                         goto tx_err;
1000                 md = ip_tunnel_info_opts(tun_info);
1001
1002                 tun_id = tunnel_id_to_key32(key->tun_id);
1003                 if (md->version == 1) {
1004                         erspan_build_header(skb,
1005                                             ntohl(tun_id),
1006                                             ntohl(md->u.index), truncate,
1007                                             false);
1008                 } else if (md->version == 2) {
1009                         erspan_build_header_v2(skb,
1010                                                ntohl(tun_id),
1011                                                md->u.md2.dir,
1012                                                get_hwid(&md->u.md2),
1013                                                truncate, false);
1014                 } else {
1015                         goto tx_err;
1016                 }
1017         } else {
1018                 switch (skb->protocol) {
1019                 case htons(ETH_P_IP):
1020                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1021                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1022                                                  &dsfield, &encap_limit);
1023                         break;
1024                 case htons(ETH_P_IPV6):
1025                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1026                                 goto tx_err;
1027                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1028                                                      &dsfield, &encap_limit))
1029                                 goto tx_err;
1030                         break;
1031                 default:
1032                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1033                         break;
1034                 }
1035
1036                 if (t->parms.erspan_ver == 1)
1037                         erspan_build_header(skb, ntohl(t->parms.o_key),
1038                                             t->parms.index,
1039                                             truncate, false);
1040                 else if (t->parms.erspan_ver == 2)
1041                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1042                                                t->parms.dir,
1043                                                t->parms.hwid,
1044                                                truncate, false);
1045                 else
1046                         goto tx_err;
1047
1048                 fl6.daddr = t->parms.raddr;
1049         }
1050
1051         /* Push GRE header. */
1052         proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1053                                            : htons(ETH_P_ERSPAN2);
1054         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1055
1056         /* TooBig packet may have updated dst->dev's mtu */
1057         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1058                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1059
1060         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1061                            NEXTHDR_GRE);
1062         if (err != 0) {
1063                 /* XXX: send ICMP error even if DF is not set. */
1064                 if (err == -EMSGSIZE) {
1065                         if (skb->protocol == htons(ETH_P_IP))
1066                                 icmp_send(skb, ICMP_DEST_UNREACH,
1067                                           ICMP_FRAG_NEEDED, htonl(mtu));
1068                         else
1069                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1070                 }
1071
1072                 goto tx_err;
1073         }
1074         return NETDEV_TX_OK;
1075
1076 tx_err:
1077         stats = &t->dev->stats;
1078         if (!IS_ERR(tun_info))
1079                 stats->tx_errors++;
1080         stats->tx_dropped++;
1081         kfree_skb(skb);
1082         return NETDEV_TX_OK;
1083 }
1084
1085 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1086 {
1087         struct net_device *dev = t->dev;
1088         struct __ip6_tnl_parm *p = &t->parms;
1089         struct flowi6 *fl6 = &t->fl.u.ip6;
1090
1091         if (dev->type != ARPHRD_ETHER) {
1092                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1093                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1094         }
1095
1096         /* Set up flowi template */
1097         fl6->saddr = p->laddr;
1098         fl6->daddr = p->raddr;
1099         fl6->flowi6_oif = p->link;
1100         fl6->flowlabel = 0;
1101         fl6->flowi6_proto = IPPROTO_GRE;
1102
1103         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1104                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1105         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1106                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1107
1108         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1109         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1110
1111         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1112                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1113                 dev->flags |= IFF_POINTOPOINT;
1114         else
1115                 dev->flags &= ~IFF_POINTOPOINT;
1116 }
1117
1118 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1119                                          int t_hlen)
1120 {
1121         const struct __ip6_tnl_parm *p = &t->parms;
1122         struct net_device *dev = t->dev;
1123
1124         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1125                 int strict = (ipv6_addr_type(&p->raddr) &
1126                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1127
1128                 struct rt6_info *rt = rt6_lookup(t->net,
1129                                                  &p->raddr, &p->laddr,
1130                                                  p->link, NULL, strict);
1131
1132                 if (!rt)
1133                         return;
1134
1135                 if (rt->dst.dev) {
1136                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1137                                                  t_hlen;
1138
1139                         if (t->dev->header_ops)
1140                                 dev->hard_header_len = dst_len;
1141                         else
1142                                 dev->needed_headroom = dst_len;
1143
1144                         if (set_mtu) {
1145                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1146                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1147                                         dev->mtu -= 8;
1148                                 if (dev->type == ARPHRD_ETHER)
1149                                         dev->mtu -= ETH_HLEN;
1150
1151                                 if (dev->mtu < IPV6_MIN_MTU)
1152                                         dev->mtu = IPV6_MIN_MTU;
1153                         }
1154                 }
1155                 ip6_rt_put(rt);
1156         }
1157 }
1158
1159 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1160 {
1161         int t_hlen;
1162
1163         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1164         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1165
1166         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1167
1168         if (tunnel->dev->header_ops)
1169                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1170         else
1171                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1172
1173         return t_hlen;
1174 }
1175
1176 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1177 {
1178         ip6gre_tnl_link_config_common(t);
1179         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1180 }
1181
1182 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1183                                      const struct __ip6_tnl_parm *p)
1184 {
1185         t->parms.laddr = p->laddr;
1186         t->parms.raddr = p->raddr;
1187         t->parms.flags = p->flags;
1188         t->parms.hop_limit = p->hop_limit;
1189         t->parms.encap_limit = p->encap_limit;
1190         t->parms.flowinfo = p->flowinfo;
1191         t->parms.link = p->link;
1192         t->parms.proto = p->proto;
1193         t->parms.i_key = p->i_key;
1194         t->parms.o_key = p->o_key;
1195         t->parms.i_flags = p->i_flags;
1196         t->parms.o_flags = p->o_flags;
1197         t->parms.fwmark = p->fwmark;
1198         t->parms.erspan_ver = p->erspan_ver;
1199         t->parms.index = p->index;
1200         t->parms.dir = p->dir;
1201         t->parms.hwid = p->hwid;
1202         dst_cache_reset(&t->dst_cache);
1203 }
1204
1205 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1206                              int set_mtu)
1207 {
1208         ip6gre_tnl_copy_tnl_parm(t, p);
1209         ip6gre_tnl_link_config(t, set_mtu);
1210         return 0;
1211 }
1212
1213 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1214         const struct ip6_tnl_parm2 *u)
1215 {
1216         p->laddr = u->laddr;
1217         p->raddr = u->raddr;
1218         p->flags = u->flags;
1219         p->hop_limit = u->hop_limit;
1220         p->encap_limit = u->encap_limit;
1221         p->flowinfo = u->flowinfo;
1222         p->link = u->link;
1223         p->i_key = u->i_key;
1224         p->o_key = u->o_key;
1225         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1226         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1227         memcpy(p->name, u->name, sizeof(u->name));
1228 }
1229
1230 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1231         const struct __ip6_tnl_parm *p)
1232 {
1233         u->proto = IPPROTO_GRE;
1234         u->laddr = p->laddr;
1235         u->raddr = p->raddr;
1236         u->flags = p->flags;
1237         u->hop_limit = p->hop_limit;
1238         u->encap_limit = p->encap_limit;
1239         u->flowinfo = p->flowinfo;
1240         u->link = p->link;
1241         u->i_key = p->i_key;
1242         u->o_key = p->o_key;
1243         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1244         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1245         memcpy(u->name, p->name, sizeof(u->name));
1246 }
1247
1248 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1249         struct ifreq *ifr, int cmd)
1250 {
1251         int err = 0;
1252         struct ip6_tnl_parm2 p;
1253         struct __ip6_tnl_parm p1;
1254         struct ip6_tnl *t = netdev_priv(dev);
1255         struct net *net = t->net;
1256         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1257
1258         memset(&p1, 0, sizeof(p1));
1259
1260         switch (cmd) {
1261         case SIOCGETTUNNEL:
1262                 if (dev == ign->fb_tunnel_dev) {
1263                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1264                                 err = -EFAULT;
1265                                 break;
1266                         }
1267                         ip6gre_tnl_parm_from_user(&p1, &p);
1268                         t = ip6gre_tunnel_locate(net, &p1, 0);
1269                         if (!t)
1270                                 t = netdev_priv(dev);
1271                 }
1272                 memset(&p, 0, sizeof(p));
1273                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1274                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1275                         err = -EFAULT;
1276                 break;
1277
1278         case SIOCADDTUNNEL:
1279         case SIOCCHGTUNNEL:
1280                 err = -EPERM;
1281                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1282                         goto done;
1283
1284                 err = -EFAULT;
1285                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1286                         goto done;
1287
1288                 err = -EINVAL;
1289                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1290                         goto done;
1291
1292                 if (!(p.i_flags&GRE_KEY))
1293                         p.i_key = 0;
1294                 if (!(p.o_flags&GRE_KEY))
1295                         p.o_key = 0;
1296
1297                 ip6gre_tnl_parm_from_user(&p1, &p);
1298                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1299
1300                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1301                         if (t) {
1302                                 if (t->dev != dev) {
1303                                         err = -EEXIST;
1304                                         break;
1305                                 }
1306                         } else {
1307                                 t = netdev_priv(dev);
1308
1309                                 ip6gre_tunnel_unlink(ign, t);
1310                                 synchronize_net();
1311                                 ip6gre_tnl_change(t, &p1, 1);
1312                                 ip6gre_tunnel_link(ign, t);
1313                                 netdev_state_change(dev);
1314                         }
1315                 }
1316
1317                 if (t) {
1318                         err = 0;
1319
1320                         memset(&p, 0, sizeof(p));
1321                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1322                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1323                                 err = -EFAULT;
1324                 } else
1325                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1326                 break;
1327
1328         case SIOCDELTUNNEL:
1329                 err = -EPERM;
1330                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1331                         goto done;
1332
1333                 if (dev == ign->fb_tunnel_dev) {
1334                         err = -EFAULT;
1335                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1336                                 goto done;
1337                         err = -ENOENT;
1338                         ip6gre_tnl_parm_from_user(&p1, &p);
1339                         t = ip6gre_tunnel_locate(net, &p1, 0);
1340                         if (!t)
1341                                 goto done;
1342                         err = -EPERM;
1343                         if (t == netdev_priv(ign->fb_tunnel_dev))
1344                                 goto done;
1345                         dev = t->dev;
1346                 }
1347                 unregister_netdevice(dev);
1348                 err = 0;
1349                 break;
1350
1351         default:
1352                 err = -EINVAL;
1353         }
1354
1355 done:
1356         return err;
1357 }
1358
1359 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1360                          unsigned short type, const void *daddr,
1361                          const void *saddr, unsigned int len)
1362 {
1363         struct ip6_tnl *t = netdev_priv(dev);
1364         struct ipv6hdr *ipv6h;
1365         __be16 *p;
1366
1367         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1368         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1369                                                   t->fl.u.ip6.flowlabel,
1370                                                   true, &t->fl.u.ip6));
1371         ipv6h->hop_limit = t->parms.hop_limit;
1372         ipv6h->nexthdr = NEXTHDR_GRE;
1373         ipv6h->saddr = t->parms.laddr;
1374         ipv6h->daddr = t->parms.raddr;
1375
1376         p = (__be16 *)(ipv6h + 1);
1377         p[0] = t->parms.o_flags;
1378         p[1] = htons(type);
1379
1380         /*
1381          *      Set the source hardware address.
1382          */
1383
1384         if (saddr)
1385                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1386         if (daddr)
1387                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1388         if (!ipv6_addr_any(&ipv6h->daddr))
1389                 return t->hlen;
1390
1391         return -t->hlen;
1392 }
1393
1394 static const struct header_ops ip6gre_header_ops = {
1395         .create = ip6gre_header,
1396 };
1397
1398 static const struct net_device_ops ip6gre_netdev_ops = {
1399         .ndo_init               = ip6gre_tunnel_init,
1400         .ndo_uninit             = ip6gre_tunnel_uninit,
1401         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1402         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1403         .ndo_change_mtu         = ip6_tnl_change_mtu,
1404         .ndo_get_stats64        = ip_tunnel_get_stats64,
1405         .ndo_get_iflink         = ip6_tnl_get_iflink,
1406 };
1407
1408 static void ip6gre_dev_free(struct net_device *dev)
1409 {
1410         struct ip6_tnl *t = netdev_priv(dev);
1411
1412         gro_cells_destroy(&t->gro_cells);
1413         dst_cache_destroy(&t->dst_cache);
1414         free_percpu(dev->tstats);
1415 }
1416
1417 static void ip6gre_tunnel_setup(struct net_device *dev)
1418 {
1419         dev->netdev_ops = &ip6gre_netdev_ops;
1420         dev->needs_free_netdev = true;
1421         dev->priv_destructor = ip6gre_dev_free;
1422
1423         dev->type = ARPHRD_IP6GRE;
1424
1425         dev->flags |= IFF_NOARP;
1426         dev->addr_len = sizeof(struct in6_addr);
1427         netif_keep_dst(dev);
1428         /* This perm addr will be used as interface identifier by IPv6 */
1429         dev->addr_assign_type = NET_ADDR_RANDOM;
1430         eth_random_addr(dev->perm_addr);
1431 }
1432
1433 #define GRE6_FEATURES (NETIF_F_SG |             \
1434                        NETIF_F_FRAGLIST |       \
1435                        NETIF_F_HIGHDMA |        \
1436                        NETIF_F_HW_CSUM)
1437
1438 static void ip6gre_tnl_init_features(struct net_device *dev)
1439 {
1440         struct ip6_tnl *nt = netdev_priv(dev);
1441
1442         dev->features           |= GRE6_FEATURES;
1443         dev->hw_features        |= GRE6_FEATURES;
1444
1445         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1446                 /* TCP offload with GRE SEQ is not supported, nor
1447                  * can we support 2 levels of outer headers requiring
1448                  * an update.
1449                  */
1450                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1451                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1452                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1453                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1454                 }
1455
1456                 /* Can use a lockless transmit, unless we generate
1457                  * output sequences
1458                  */
1459                 dev->features |= NETIF_F_LLTX;
1460         }
1461 }
1462
1463 static int ip6gre_tunnel_init_common(struct net_device *dev)
1464 {
1465         struct ip6_tnl *tunnel;
1466         int ret;
1467         int t_hlen;
1468
1469         tunnel = netdev_priv(dev);
1470
1471         tunnel->dev = dev;
1472         tunnel->net = dev_net(dev);
1473         strcpy(tunnel->parms.name, dev->name);
1474
1475         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1476         if (!dev->tstats)
1477                 return -ENOMEM;
1478
1479         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1480         if (ret)
1481                 goto cleanup_alloc_pcpu_stats;
1482
1483         ret = gro_cells_init(&tunnel->gro_cells, dev);
1484         if (ret)
1485                 goto cleanup_dst_cache_init;
1486
1487         t_hlen = ip6gre_calc_hlen(tunnel);
1488         dev->mtu = ETH_DATA_LEN - t_hlen;
1489         if (dev->type == ARPHRD_ETHER)
1490                 dev->mtu -= ETH_HLEN;
1491         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1492                 dev->mtu -= 8;
1493
1494         if (tunnel->parms.collect_md) {
1495                 netif_keep_dst(dev);
1496         }
1497         ip6gre_tnl_init_features(dev);
1498
1499         return 0;
1500
1501 cleanup_dst_cache_init:
1502         dst_cache_destroy(&tunnel->dst_cache);
1503 cleanup_alloc_pcpu_stats:
1504         free_percpu(dev->tstats);
1505         dev->tstats = NULL;
1506         return ret;
1507 }
1508
1509 static int ip6gre_tunnel_init(struct net_device *dev)
1510 {
1511         struct ip6_tnl *tunnel;
1512         int ret;
1513
1514         ret = ip6gre_tunnel_init_common(dev);
1515         if (ret)
1516                 return ret;
1517
1518         tunnel = netdev_priv(dev);
1519
1520         if (tunnel->parms.collect_md)
1521                 return 0;
1522
1523         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1524         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1525
1526         if (ipv6_addr_any(&tunnel->parms.raddr))
1527                 dev->header_ops = &ip6gre_header_ops;
1528
1529         return 0;
1530 }
1531
1532 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1533 {
1534         struct ip6_tnl *tunnel = netdev_priv(dev);
1535
1536         tunnel->dev = dev;
1537         tunnel->net = dev_net(dev);
1538         strcpy(tunnel->parms.name, dev->name);
1539
1540         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1541
1542         dev_hold(dev);
1543 }
1544
1545 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1546         .handler     = gre_rcv,
1547         .err_handler = ip6gre_err,
1548         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1549 };
1550
1551 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1552 {
1553         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1554         struct net_device *dev, *aux;
1555         int prio;
1556
1557         for_each_netdev_safe(net, dev, aux)
1558                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1559                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1560                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1561                         unregister_netdevice_queue(dev, head);
1562
1563         for (prio = 0; prio < 4; prio++) {
1564                 int h;
1565                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1566                         struct ip6_tnl *t;
1567
1568                         t = rtnl_dereference(ign->tunnels[prio][h]);
1569
1570                         while (t) {
1571                                 /* If dev is in the same netns, it has already
1572                                  * been added to the list by the previous loop.
1573                                  */
1574                                 if (!net_eq(dev_net(t->dev), net))
1575                                         unregister_netdevice_queue(t->dev,
1576                                                                    head);
1577                                 t = rtnl_dereference(t->next);
1578                         }
1579                 }
1580         }
1581 }
1582
1583 static int __net_init ip6gre_init_net(struct net *net)
1584 {
1585         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1586         struct net_device *ndev;
1587         int err;
1588
1589         if (!net_has_fallback_tunnels(net))
1590                 return 0;
1591         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1592                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1593         if (!ndev) {
1594                 err = -ENOMEM;
1595                 goto err_alloc_dev;
1596         }
1597         ign->fb_tunnel_dev = ndev;
1598         dev_net_set(ign->fb_tunnel_dev, net);
1599         /* FB netdevice is special: we have one, and only one per netns.
1600          * Allowing to move it to another netns is clearly unsafe.
1601          */
1602         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1603
1604
1605         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1606         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1607
1608         err = register_netdev(ign->fb_tunnel_dev);
1609         if (err)
1610                 goto err_reg_dev;
1611
1612         rcu_assign_pointer(ign->tunnels_wc[0],
1613                            netdev_priv(ign->fb_tunnel_dev));
1614         return 0;
1615
1616 err_reg_dev:
1617         free_netdev(ndev);
1618 err_alloc_dev:
1619         return err;
1620 }
1621
1622 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1623 {
1624         struct net *net;
1625         LIST_HEAD(list);
1626
1627         rtnl_lock();
1628         list_for_each_entry(net, net_list, exit_list)
1629                 ip6gre_destroy_tunnels(net, &list);
1630         unregister_netdevice_many(&list);
1631         rtnl_unlock();
1632 }
1633
1634 static struct pernet_operations ip6gre_net_ops = {
1635         .init = ip6gre_init_net,
1636         .exit_batch = ip6gre_exit_batch_net,
1637         .id   = &ip6gre_net_id,
1638         .size = sizeof(struct ip6gre_net),
1639 };
1640
1641 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1642                                   struct netlink_ext_ack *extack)
1643 {
1644         __be16 flags;
1645
1646         if (!data)
1647                 return 0;
1648
1649         flags = 0;
1650         if (data[IFLA_GRE_IFLAGS])
1651                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1652         if (data[IFLA_GRE_OFLAGS])
1653                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1654         if (flags & (GRE_VERSION|GRE_ROUTING))
1655                 return -EINVAL;
1656
1657         return 0;
1658 }
1659
1660 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1661                                struct netlink_ext_ack *extack)
1662 {
1663         struct in6_addr daddr;
1664
1665         if (tb[IFLA_ADDRESS]) {
1666                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1667                         return -EINVAL;
1668                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1669                         return -EADDRNOTAVAIL;
1670         }
1671
1672         if (!data)
1673                 goto out;
1674
1675         if (data[IFLA_GRE_REMOTE]) {
1676                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1677                 if (ipv6_addr_any(&daddr))
1678                         return -EINVAL;
1679         }
1680
1681 out:
1682         return ip6gre_tunnel_validate(tb, data, extack);
1683 }
1684
1685 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1686                                   struct netlink_ext_ack *extack)
1687 {
1688         __be16 flags = 0;
1689         int ret, ver = 0;
1690
1691         if (!data)
1692                 return 0;
1693
1694         ret = ip6gre_tap_validate(tb, data, extack);
1695         if (ret)
1696                 return ret;
1697
1698         /* ERSPAN should only have GRE sequence and key flag */
1699         if (data[IFLA_GRE_OFLAGS])
1700                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1701         if (data[IFLA_GRE_IFLAGS])
1702                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1703         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1704             flags != (GRE_SEQ | GRE_KEY))
1705                 return -EINVAL;
1706
1707         /* ERSPAN Session ID only has 10-bit. Since we reuse
1708          * 32-bit key field as ID, check it's range.
1709          */
1710         if (data[IFLA_GRE_IKEY] &&
1711             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1712                 return -EINVAL;
1713
1714         if (data[IFLA_GRE_OKEY] &&
1715             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1716                 return -EINVAL;
1717
1718         if (data[IFLA_GRE_ERSPAN_VER]) {
1719                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1720                 if (ver != 1 && ver != 2)
1721                         return -EINVAL;
1722         }
1723
1724         if (ver == 1) {
1725                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1726                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1727
1728                         if (index & ~INDEX_MASK)
1729                                 return -EINVAL;
1730                 }
1731         } else if (ver == 2) {
1732                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1733                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1734
1735                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1736                                 return -EINVAL;
1737                 }
1738
1739                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1740                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1741
1742                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1743                                 return -EINVAL;
1744                 }
1745         }
1746
1747         return 0;
1748 }
1749
1750 static void ip6erspan_set_version(struct nlattr *data[],
1751                                   struct __ip6_tnl_parm *parms)
1752 {
1753         if (!data)
1754                 return;
1755
1756         parms->erspan_ver = 1;
1757         if (data[IFLA_GRE_ERSPAN_VER])
1758                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1759
1760         if (parms->erspan_ver == 1) {
1761                 if (data[IFLA_GRE_ERSPAN_INDEX])
1762                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1763         } else if (parms->erspan_ver == 2) {
1764                 if (data[IFLA_GRE_ERSPAN_DIR])
1765                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1766                 if (data[IFLA_GRE_ERSPAN_HWID])
1767                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1768         }
1769 }
1770
1771 static void ip6gre_netlink_parms(struct nlattr *data[],
1772                                 struct __ip6_tnl_parm *parms)
1773 {
1774         memset(parms, 0, sizeof(*parms));
1775
1776         if (!data)
1777                 return;
1778
1779         if (data[IFLA_GRE_LINK])
1780                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1781
1782         if (data[IFLA_GRE_IFLAGS])
1783                 parms->i_flags = gre_flags_to_tnl_flags(
1784                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1785
1786         if (data[IFLA_GRE_OFLAGS])
1787                 parms->o_flags = gre_flags_to_tnl_flags(
1788                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1789
1790         if (data[IFLA_GRE_IKEY])
1791                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1792
1793         if (data[IFLA_GRE_OKEY])
1794                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1795
1796         if (data[IFLA_GRE_LOCAL])
1797                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1798
1799         if (data[IFLA_GRE_REMOTE])
1800                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1801
1802         if (data[IFLA_GRE_TTL])
1803                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1804
1805         if (data[IFLA_GRE_ENCAP_LIMIT])
1806                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1807
1808         if (data[IFLA_GRE_FLOWINFO])
1809                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1810
1811         if (data[IFLA_GRE_FLAGS])
1812                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1813
1814         if (data[IFLA_GRE_FWMARK])
1815                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1816
1817         if (data[IFLA_GRE_COLLECT_METADATA])
1818                 parms->collect_md = true;
1819 }
1820
1821 static int ip6gre_tap_init(struct net_device *dev)
1822 {
1823         int ret;
1824
1825         ret = ip6gre_tunnel_init_common(dev);
1826         if (ret)
1827                 return ret;
1828
1829         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1830
1831         return 0;
1832 }
1833
1834 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1835         .ndo_init = ip6gre_tap_init,
1836         .ndo_uninit = ip6gre_tunnel_uninit,
1837         .ndo_start_xmit = ip6gre_tunnel_xmit,
1838         .ndo_set_mac_address = eth_mac_addr,
1839         .ndo_validate_addr = eth_validate_addr,
1840         .ndo_change_mtu = ip6_tnl_change_mtu,
1841         .ndo_get_stats64 = ip_tunnel_get_stats64,
1842         .ndo_get_iflink = ip6_tnl_get_iflink,
1843 };
1844
1845 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1846 {
1847         int t_hlen;
1848
1849         tunnel->tun_hlen = 8;
1850         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1851                        erspan_hdr_len(tunnel->parms.erspan_ver);
1852
1853         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1854         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1855         return t_hlen;
1856 }
1857
1858 static int ip6erspan_tap_init(struct net_device *dev)
1859 {
1860         struct ip6_tnl *tunnel;
1861         int t_hlen;
1862         int ret;
1863
1864         tunnel = netdev_priv(dev);
1865
1866         tunnel->dev = dev;
1867         tunnel->net = dev_net(dev);
1868         strcpy(tunnel->parms.name, dev->name);
1869
1870         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1871         if (!dev->tstats)
1872                 return -ENOMEM;
1873
1874         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1875         if (ret)
1876                 goto cleanup_alloc_pcpu_stats;
1877
1878         ret = gro_cells_init(&tunnel->gro_cells, dev);
1879         if (ret)
1880                 goto cleanup_dst_cache_init;
1881
1882         t_hlen = ip6erspan_calc_hlen(tunnel);
1883         dev->mtu = ETH_DATA_LEN - t_hlen;
1884         if (dev->type == ARPHRD_ETHER)
1885                 dev->mtu -= ETH_HLEN;
1886         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1887                 dev->mtu -= 8;
1888
1889         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1890         ip6erspan_tnl_link_config(tunnel, 1);
1891
1892         return 0;
1893
1894 cleanup_dst_cache_init:
1895         dst_cache_destroy(&tunnel->dst_cache);
1896 cleanup_alloc_pcpu_stats:
1897         free_percpu(dev->tstats);
1898         dev->tstats = NULL;
1899         return ret;
1900 }
1901
1902 static const struct net_device_ops ip6erspan_netdev_ops = {
1903         .ndo_init =             ip6erspan_tap_init,
1904         .ndo_uninit =           ip6erspan_tunnel_uninit,
1905         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1906         .ndo_set_mac_address =  eth_mac_addr,
1907         .ndo_validate_addr =    eth_validate_addr,
1908         .ndo_change_mtu =       ip6_tnl_change_mtu,
1909         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1910         .ndo_get_iflink =       ip6_tnl_get_iflink,
1911 };
1912
1913 static void ip6gre_tap_setup(struct net_device *dev)
1914 {
1915
1916         ether_setup(dev);
1917
1918         dev->max_mtu = 0;
1919         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1920         dev->needs_free_netdev = true;
1921         dev->priv_destructor = ip6gre_dev_free;
1922
1923         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1924         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1925         netif_keep_dst(dev);
1926 }
1927
1928 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1929                                        struct ip_tunnel_encap *ipencap)
1930 {
1931         bool ret = false;
1932
1933         memset(ipencap, 0, sizeof(*ipencap));
1934
1935         if (!data)
1936                 return ret;
1937
1938         if (data[IFLA_GRE_ENCAP_TYPE]) {
1939                 ret = true;
1940                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1941         }
1942
1943         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1944                 ret = true;
1945                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1946         }
1947
1948         if (data[IFLA_GRE_ENCAP_SPORT]) {
1949                 ret = true;
1950                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1951         }
1952
1953         if (data[IFLA_GRE_ENCAP_DPORT]) {
1954                 ret = true;
1955                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1956         }
1957
1958         return ret;
1959 }
1960
1961 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1962                                  struct nlattr *tb[], struct nlattr *data[],
1963                                  struct netlink_ext_ack *extack)
1964 {
1965         struct ip6_tnl *nt;
1966         struct ip_tunnel_encap ipencap;
1967         int err;
1968
1969         nt = netdev_priv(dev);
1970
1971         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1972                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1973
1974                 if (err < 0)
1975                         return err;
1976         }
1977
1978         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1979                 eth_hw_addr_random(dev);
1980
1981         nt->dev = dev;
1982         nt->net = dev_net(dev);
1983
1984         err = register_netdevice(dev);
1985         if (err)
1986                 goto out;
1987
1988         if (tb[IFLA_MTU])
1989                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1990
1991         dev_hold(dev);
1992
1993 out:
1994         return err;
1995 }
1996
1997 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1998                           struct nlattr *tb[], struct nlattr *data[],
1999                           struct netlink_ext_ack *extack)
2000 {
2001         struct ip6_tnl *nt = netdev_priv(dev);
2002         struct net *net = dev_net(dev);
2003         struct ip6gre_net *ign;
2004         int err;
2005
2006         ip6gre_netlink_parms(data, &nt->parms);
2007         ign = net_generic(net, ip6gre_net_id);
2008
2009         if (nt->parms.collect_md) {
2010                 if (rtnl_dereference(ign->collect_md_tun))
2011                         return -EEXIST;
2012         } else {
2013                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2014                         return -EEXIST;
2015         }
2016
2017         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2018         if (!err) {
2019                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2020                 ip6gre_tunnel_link_md(ign, nt);
2021                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2022         }
2023         return err;
2024 }
2025
2026 static struct ip6_tnl *
2027 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2028                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2029                          struct netlink_ext_ack *extack)
2030 {
2031         struct ip6_tnl *t, *nt = netdev_priv(dev);
2032         struct net *net = nt->net;
2033         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2034         struct ip_tunnel_encap ipencap;
2035
2036         if (dev == ign->fb_tunnel_dev)
2037                 return ERR_PTR(-EINVAL);
2038
2039         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2040                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2041
2042                 if (err < 0)
2043                         return ERR_PTR(err);
2044         }
2045
2046         ip6gre_netlink_parms(data, p_p);
2047
2048         t = ip6gre_tunnel_locate(net, p_p, 0);
2049
2050         if (t) {
2051                 if (t->dev != dev)
2052                         return ERR_PTR(-EEXIST);
2053         } else {
2054                 t = nt;
2055         }
2056
2057         return t;
2058 }
2059
2060 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2061                              struct nlattr *data[],
2062                              struct netlink_ext_ack *extack)
2063 {
2064         struct ip6_tnl *t = netdev_priv(dev);
2065         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2066         struct __ip6_tnl_parm p;
2067
2068         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2069         if (IS_ERR(t))
2070                 return PTR_ERR(t);
2071
2072         ip6gre_tunnel_unlink_md(ign, t);
2073         ip6gre_tunnel_unlink(ign, t);
2074         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2075         ip6gre_tunnel_link_md(ign, t);
2076         ip6gre_tunnel_link(ign, t);
2077         return 0;
2078 }
2079
2080 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2081 {
2082         struct net *net = dev_net(dev);
2083         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2084
2085         if (dev != ign->fb_tunnel_dev)
2086                 unregister_netdevice_queue(dev, head);
2087 }
2088
2089 static size_t ip6gre_get_size(const struct net_device *dev)
2090 {
2091         return
2092                 /* IFLA_GRE_LINK */
2093                 nla_total_size(4) +
2094                 /* IFLA_GRE_IFLAGS */
2095                 nla_total_size(2) +
2096                 /* IFLA_GRE_OFLAGS */
2097                 nla_total_size(2) +
2098                 /* IFLA_GRE_IKEY */
2099                 nla_total_size(4) +
2100                 /* IFLA_GRE_OKEY */
2101                 nla_total_size(4) +
2102                 /* IFLA_GRE_LOCAL */
2103                 nla_total_size(sizeof(struct in6_addr)) +
2104                 /* IFLA_GRE_REMOTE */
2105                 nla_total_size(sizeof(struct in6_addr)) +
2106                 /* IFLA_GRE_TTL */
2107                 nla_total_size(1) +
2108                 /* IFLA_GRE_ENCAP_LIMIT */
2109                 nla_total_size(1) +
2110                 /* IFLA_GRE_FLOWINFO */
2111                 nla_total_size(4) +
2112                 /* IFLA_GRE_FLAGS */
2113                 nla_total_size(4) +
2114                 /* IFLA_GRE_ENCAP_TYPE */
2115                 nla_total_size(2) +
2116                 /* IFLA_GRE_ENCAP_FLAGS */
2117                 nla_total_size(2) +
2118                 /* IFLA_GRE_ENCAP_SPORT */
2119                 nla_total_size(2) +
2120                 /* IFLA_GRE_ENCAP_DPORT */
2121                 nla_total_size(2) +
2122                 /* IFLA_GRE_COLLECT_METADATA */
2123                 nla_total_size(0) +
2124                 /* IFLA_GRE_FWMARK */
2125                 nla_total_size(4) +
2126                 /* IFLA_GRE_ERSPAN_INDEX */
2127                 nla_total_size(4) +
2128                 0;
2129 }
2130
2131 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2132 {
2133         struct ip6_tnl *t = netdev_priv(dev);
2134         struct __ip6_tnl_parm *p = &t->parms;
2135         __be16 o_flags = p->o_flags;
2136
2137         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2138                 if (!p->collect_md)
2139                         o_flags |= TUNNEL_KEY;
2140
2141                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2142                         goto nla_put_failure;
2143
2144                 if (p->erspan_ver == 1) {
2145                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2146                                 goto nla_put_failure;
2147                 } else {
2148                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2149                                 goto nla_put_failure;
2150                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2151                                 goto nla_put_failure;
2152                 }
2153         }
2154
2155         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2156             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2157                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2158             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2159                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2160             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2161             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2162             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2163             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2164             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2165             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2166             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2167             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2168             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2169                 goto nla_put_failure;
2170
2171         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2172                         t->encap.type) ||
2173             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2174                          t->encap.sport) ||
2175             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2176                          t->encap.dport) ||
2177             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2178                         t->encap.flags))
2179                 goto nla_put_failure;
2180
2181         if (p->collect_md) {
2182                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2183                         goto nla_put_failure;
2184         }
2185
2186         return 0;
2187
2188 nla_put_failure:
2189         return -EMSGSIZE;
2190 }
2191
2192 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2193         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2194         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2195         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2196         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2197         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2198         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2199         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2200         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2201         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2202         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2203         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2204         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2205         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2206         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2207         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2208         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2209         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2210         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2211         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2212         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2213         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2214 };
2215
2216 static void ip6erspan_tap_setup(struct net_device *dev)
2217 {
2218         ether_setup(dev);
2219
2220         dev->max_mtu = 0;
2221         dev->netdev_ops = &ip6erspan_netdev_ops;
2222         dev->needs_free_netdev = true;
2223         dev->priv_destructor = ip6gre_dev_free;
2224
2225         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2226         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2227         netif_keep_dst(dev);
2228 }
2229
2230 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2231                              struct nlattr *tb[], struct nlattr *data[],
2232                              struct netlink_ext_ack *extack)
2233 {
2234         struct ip6_tnl *nt = netdev_priv(dev);
2235         struct net *net = dev_net(dev);
2236         struct ip6gre_net *ign;
2237         int err;
2238
2239         ip6gre_netlink_parms(data, &nt->parms);
2240         ip6erspan_set_version(data, &nt->parms);
2241         ign = net_generic(net, ip6gre_net_id);
2242
2243         if (nt->parms.collect_md) {
2244                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2245                         return -EEXIST;
2246         } else {
2247                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2248                         return -EEXIST;
2249         }
2250
2251         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2252         if (!err) {
2253                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2254                 ip6erspan_tunnel_link_md(ign, nt);
2255                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2256         }
2257         return err;
2258 }
2259
2260 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2261 {
2262         ip6gre_tnl_link_config_common(t);
2263         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2264 }
2265
2266 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2267                                 const struct __ip6_tnl_parm *p, int set_mtu)
2268 {
2269         ip6gre_tnl_copy_tnl_parm(t, p);
2270         ip6erspan_tnl_link_config(t, set_mtu);
2271         return 0;
2272 }
2273
2274 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2275                                 struct nlattr *data[],
2276                                 struct netlink_ext_ack *extack)
2277 {
2278         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2279         struct __ip6_tnl_parm p;
2280         struct ip6_tnl *t;
2281
2282         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2283         if (IS_ERR(t))
2284                 return PTR_ERR(t);
2285
2286         ip6erspan_set_version(data, &p);
2287         ip6gre_tunnel_unlink_md(ign, t);
2288         ip6gre_tunnel_unlink(ign, t);
2289         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2290         ip6erspan_tunnel_link_md(ign, t);
2291         ip6gre_tunnel_link(ign, t);
2292         return 0;
2293 }
2294
2295 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2296         .kind           = "ip6gre",
2297         .maxtype        = IFLA_GRE_MAX,
2298         .policy         = ip6gre_policy,
2299         .priv_size      = sizeof(struct ip6_tnl),
2300         .setup          = ip6gre_tunnel_setup,
2301         .validate       = ip6gre_tunnel_validate,
2302         .newlink        = ip6gre_newlink,
2303         .changelink     = ip6gre_changelink,
2304         .dellink        = ip6gre_dellink,
2305         .get_size       = ip6gre_get_size,
2306         .fill_info      = ip6gre_fill_info,
2307         .get_link_net   = ip6_tnl_get_link_net,
2308 };
2309
2310 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2311         .kind           = "ip6gretap",
2312         .maxtype        = IFLA_GRE_MAX,
2313         .policy         = ip6gre_policy,
2314         .priv_size      = sizeof(struct ip6_tnl),
2315         .setup          = ip6gre_tap_setup,
2316         .validate       = ip6gre_tap_validate,
2317         .newlink        = ip6gre_newlink,
2318         .changelink     = ip6gre_changelink,
2319         .get_size       = ip6gre_get_size,
2320         .fill_info      = ip6gre_fill_info,
2321         .get_link_net   = ip6_tnl_get_link_net,
2322 };
2323
2324 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2325         .kind           = "ip6erspan",
2326         .maxtype        = IFLA_GRE_MAX,
2327         .policy         = ip6gre_policy,
2328         .priv_size      = sizeof(struct ip6_tnl),
2329         .setup          = ip6erspan_tap_setup,
2330         .validate       = ip6erspan_tap_validate,
2331         .newlink        = ip6erspan_newlink,
2332         .changelink     = ip6erspan_changelink,
2333         .get_size       = ip6gre_get_size,
2334         .fill_info      = ip6gre_fill_info,
2335         .get_link_net   = ip6_tnl_get_link_net,
2336 };
2337
2338 /*
2339  *      And now the modules code and kernel interface.
2340  */
2341
2342 static int __init ip6gre_init(void)
2343 {
2344         int err;
2345
2346         pr_info("GRE over IPv6 tunneling driver\n");
2347
2348         err = register_pernet_device(&ip6gre_net_ops);
2349         if (err < 0)
2350                 return err;
2351
2352         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2353         if (err < 0) {
2354                 pr_info("%s: can't add protocol\n", __func__);
2355                 goto add_proto_failed;
2356         }
2357
2358         err = rtnl_link_register(&ip6gre_link_ops);
2359         if (err < 0)
2360                 goto rtnl_link_failed;
2361
2362         err = rtnl_link_register(&ip6gre_tap_ops);
2363         if (err < 0)
2364                 goto tap_ops_failed;
2365
2366         err = rtnl_link_register(&ip6erspan_tap_ops);
2367         if (err < 0)
2368                 goto erspan_link_failed;
2369
2370 out:
2371         return err;
2372
2373 erspan_link_failed:
2374         rtnl_link_unregister(&ip6gre_tap_ops);
2375 tap_ops_failed:
2376         rtnl_link_unregister(&ip6gre_link_ops);
2377 rtnl_link_failed:
2378         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2379 add_proto_failed:
2380         unregister_pernet_device(&ip6gre_net_ops);
2381         goto out;
2382 }
2383
2384 static void __exit ip6gre_fini(void)
2385 {
2386         rtnl_link_unregister(&ip6gre_tap_ops);
2387         rtnl_link_unregister(&ip6gre_link_ops);
2388         rtnl_link_unregister(&ip6erspan_tap_ops);
2389         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2390         unregister_pernet_device(&ip6gre_net_ops);
2391 }
2392
2393 module_init(ip6gre_init);
2394 module_exit(ip6gre_fini);
2395 MODULE_LICENSE("GPL");
2396 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2397 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2398 MODULE_ALIAS_RTNL_LINK("ip6gre");
2399 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2400 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2401 MODULE_ALIAS_NETDEV("ip6gre0");