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 IP_TUNNEL_DECLARE_FLAGS(flags);
500 struct metadata_dst *tun_dst;
501 __be64 tun_id;
502
503 ip_tunnel_flags_copy(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 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
532 return PACKET_REJECT;
533
534 ipv6h = ipv6_hdr(skb);
535 ershdr = (struct erspan_base_hdr *)skb->data;
536 ver = ershdr->ver;
537
538 tunnel = ip6gre_tunnel_lookup(skb->dev,
539 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
540 tpi->proto);
541 if (tunnel) {
542 int len = erspan_hdr_len(ver);
543
544 if (unlikely(!pskb_may_pull(skb, len)))
545 return PACKET_REJECT;
546
547 if (__iptunnel_pull_header(skb, len,
548 htons(ETH_P_TEB),
549 false, false) < 0)
550 return PACKET_REJECT;
551
552 if (tunnel->parms.collect_md) {
553 struct erspan_metadata *pkt_md, *md;
554 IP_TUNNEL_DECLARE_FLAGS(flags);
555 struct metadata_dst *tun_dst;
556 struct ip_tunnel_info *info;
557 unsigned char *gh;
558 __be64 tun_id;
559
560 __set_bit(IP_TUNNEL_KEY_BIT, tpi->flags);
561 ip_tunnel_flags_copy(flags, tpi->flags);
562 tun_id = key32_to_tunnel_id(tpi->key);
563
564 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
565 sizeof(*md));
566 if (!tun_dst)
567 return PACKET_REJECT;
568
569 /* skb can be uncloned in __iptunnel_pull_header, so
570 * old pkt_md is no longer valid and we need to reset
571 * it
572 */
573 gh = skb_network_header(skb) +
574 skb_network_header_len(skb);
575 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
576 sizeof(*ershdr));
577 info = &tun_dst->u.tun_info;
578 md = ip_tunnel_info_opts(info);
579 md->version = ver;
580 md2 = &md->u.md2;
581 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
582 ERSPAN_V2_MDSIZE);
583 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
584 info->key.tun_flags);
585 info->options_len = sizeof(*md);
586
587 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
588
589 } else {
590 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
591 }
592
593 return PACKET_RCVD;
594 }
595
596 return PACKET_REJECT;
597 }
598
599 static int gre_rcv(struct sk_buff *skb)
600 {
601 struct tnl_ptk_info tpi;
602 bool csum_err = false;
603 int hdr_len;
604
605 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
606 if (hdr_len < 0)
607 goto drop;
608
609 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
610 goto drop;
611
612 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
613 tpi.proto == htons(ETH_P_ERSPAN2))) {
614 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
615 return 0;
616 goto out;
617 }
618
619 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
620 return 0;
621
622 out:
623 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
624 drop:
625 kfree_skb(skb);
626 return 0;
627 }
628
629 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
630 {
631 return iptunnel_handle_offloads(skb,
632 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
633 }
634
635 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
636 struct net_device *dev,
637 struct flowi6 *fl6, __u8 *dsfield,
638 int *encap_limit)
639 {
640 const struct iphdr *iph = ip_hdr(skb);
641 struct ip6_tnl *t = netdev_priv(dev);
642
643 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
644 *encap_limit = t->parms.encap_limit;
645
646 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
647
648 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
649 *dsfield = ipv4_get_dsfield(iph);
650 else
651 *dsfield = ip6_tclass(t->parms.flowinfo);
652
653 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
654 fl6->flowi6_mark = skb->mark;
655 else
656 fl6->flowi6_mark = t->parms.fwmark;
657
658 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
659 }
660
661 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
662 struct net_device *dev,
663 struct flowi6 *fl6, __u8 *dsfield,
664 int *encap_limit)
665 {
666 struct ipv6hdr *ipv6h;
667 struct ip6_tnl *t = netdev_priv(dev);
668 __u16 offset;
669
670 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
671 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
672 ipv6h = ipv6_hdr(skb);
673
674 if (offset > 0) {
675 struct ipv6_tlv_tnl_enc_lim *tel;
676
677 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
678 if (tel->encap_limit == 0) {
679 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
680 ICMPV6_HDR_FIELD, offset + 2);
681 return -1;
682 }
683 *encap_limit = tel->encap_limit - 1;
684 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
685 *encap_limit = t->parms.encap_limit;
686 }
687
688 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
689
690 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
691 *dsfield = ipv6_get_dsfield(ipv6h);
692 else
693 *dsfield = ip6_tclass(t->parms.flowinfo);
694
695 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
696 fl6->flowlabel |= ip6_flowlabel(ipv6h);
697
698 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
699 fl6->flowi6_mark = skb->mark;
700 else
701 fl6->flowi6_mark = t->parms.fwmark;
702
703 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
704
705 return 0;
706 }
707
708 static int prepare_ip6gre_xmit_other(struct sk_buff *skb,
709 struct net_device *dev,
710 struct flowi6 *fl6, __u8 *dsfield,
711 int *encap_limit)
712 {
713 struct ip6_tnl *t = netdev_priv(dev);
714
715 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
716 *encap_limit = t->parms.encap_limit;
717
718 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
719
720 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
721 *dsfield = 0;
722 else
723 *dsfield = ip6_tclass(t->parms.flowinfo);
724
725 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
726 fl6->flowi6_mark = skb->mark;
727 else
728 fl6->flowi6_mark = t->parms.fwmark;
729
730 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
731
732 return 0;
733 }
734
735 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
736 {
737 struct ip_tunnel_info *tun_info;
738
739 tun_info = skb_tunnel_info(skb);
740 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
741 return ERR_PTR(-EINVAL);
742
743 return tun_info;
744 }
745
746 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
747 struct net_device *dev, __u8 dsfield,
748 struct flowi6 *fl6, int encap_limit,
749 __u32 *pmtu, __be16 proto)
750 {
751 struct ip6_tnl *tunnel = netdev_priv(dev);
752 IP_TUNNEL_DECLARE_FLAGS(flags);
753 __be16 protocol;
754
755 if (dev->type == ARPHRD_ETHER)
756 IPCB(skb)->flags = 0;
757
758 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
759 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
760 else
761 fl6->daddr = tunnel->parms.raddr;
762
763 /* Push GRE header. */
764 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
765
766 if (tunnel->parms.collect_md) {
767 struct ip_tunnel_info *tun_info;
768 const struct ip_tunnel_key *key;
769 int tun_hlen;
770
771 tun_info = skb_tunnel_info_txcheck(skb);
772 if (IS_ERR(tun_info) ||
773 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
774 return -EINVAL;
775
776 key = &tun_info->key;
777 memset(fl6, 0, sizeof(*fl6));
778 fl6->flowi6_proto = IPPROTO_GRE;
779 fl6->daddr = key->u.ipv6.dst;
780 fl6->flowlabel = key->label;
781 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
782 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
783
784 dsfield = key->tos;
785 ip_tunnel_flags_zero(flags);
786 __set_bit(IP_TUNNEL_CSUM_BIT, flags);
787 __set_bit(IP_TUNNEL_KEY_BIT, flags);
788 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
789 ip_tunnel_flags_and(flags, flags, key->tun_flags);
790 tun_hlen = gre_calc_hlen(flags);
791
792 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
793 return -ENOMEM;
794
795 gre_build_header(skb, tun_hlen,
796 flags, protocol,
797 tunnel_id_to_key32(tun_info->key.tun_id),
798 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
799 htonl(atomic_fetch_inc(&tunnel->o_seqno)) :
800 0);
801
802 } else {
803 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
804 return -ENOMEM;
805
806 ip_tunnel_flags_copy(flags, tunnel->parms.o_flags);
807
808 gre_build_header(skb, tunnel->tun_hlen, flags,
809 protocol, tunnel->parms.o_key,
810 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
811 htonl(atomic_fetch_inc(&tunnel->o_seqno)) :
812 0);
813 }
814
815 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
816 NEXTHDR_GRE);
817 }
818
819 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
820 {
821 struct ip6_tnl *t = netdev_priv(dev);
822 int encap_limit = -1;
823 struct flowi6 fl6;
824 __u8 dsfield = 0;
825 __u32 mtu;
826 int err;
827
828 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
829
830 if (!t->parms.collect_md)
831 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
832 &dsfield, &encap_limit);
833
834 err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
835 t->parms.o_flags));
836 if (err)
837 return -1;
838
839 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
840 skb->protocol);
841 if (err != 0) {
842 /* XXX: send ICMP error even if DF is not set. */
843 if (err == -EMSGSIZE)
844 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
845 htonl(mtu));
846 return -1;
847 }
848
849 return 0;
850 }
851
852 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
853 {
854 struct ip6_tnl *t = netdev_priv(dev);
855 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
856 int encap_limit = -1;
857 struct flowi6 fl6;
858 __u8 dsfield = 0;
859 __u32 mtu;
860 int err;
861
862 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
863 return -1;
864
865 if (!t->parms.collect_md &&
866 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
867 return -1;
868
869 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
870 t->parms.o_flags)))
871 return -1;
872
873 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
874 &mtu, skb->protocol);
875 if (err != 0) {
876 if (err == -EMSGSIZE)
877 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
878 return -1;
879 }
880
881 return 0;
882 }
883
884 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
885 {
886 struct ip6_tnl *t = netdev_priv(dev);
887 int encap_limit = -1;
888 struct flowi6 fl6;
889 __u8 dsfield = 0;
890 __u32 mtu;
891 int err;
892
893 if (!t->parms.collect_md &&
894 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit))
895 return -1;
896
897 err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
898 t->parms.o_flags));
899 if (err)
900 return err;
901 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol);
902
903 return err;
904 }
905
906 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
907 struct net_device *dev)
908 {
909 struct ip6_tnl *t = netdev_priv(dev);
910 __be16 payload_protocol;
911 int ret;
912
913 if (!pskb_inet_may_pull(skb))
914 goto tx_err;
915
916 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
917 goto tx_err;
918
919 payload_protocol = skb_protocol(skb, true);
920 switch (payload_protocol) {
921 case htons(ETH_P_IP):
922 ret = ip6gre_xmit_ipv4(skb, dev);
923 break;
924 case htons(ETH_P_IPV6):
925 ret = ip6gre_xmit_ipv6(skb, dev);
926 break;
927 default:
928 ret = ip6gre_xmit_other(skb, dev);
929 break;
930 }
931
932 if (ret < 0)
933 goto tx_err;
934
935 return NETDEV_TX_OK;
936
937 tx_err:
938 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
939 DEV_STATS_INC(dev, tx_errors);
940 DEV_STATS_INC(dev, tx_dropped);
941 kfree_skb(skb);
942 return NETDEV_TX_OK;
943 }
944
945 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
946 struct net_device *dev)
947 {
948 struct ip_tunnel_info *tun_info = NULL;
949 struct ip6_tnl *t = netdev_priv(dev);
950 struct dst_entry *dst = skb_dst(skb);
951 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
952 bool truncate = false;
953 int encap_limit = -1;
954 __u8 dsfield = false;
955 struct flowi6 fl6;
956 int err = -EINVAL;
957 __be16 proto;
958 __u32 mtu;
959 int nhoff;
960
961 if (!pskb_inet_may_pull(skb))
962 goto tx_err;
963
964 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
965 goto tx_err;
966
967 if (gre_handle_offloads(skb, false))
968 goto tx_err;
969
970 if (skb->len > dev->mtu + dev->hard_header_len) {
971 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
972 goto tx_err;
973 truncate = true;
974 }
975
976 nhoff = skb_network_offset(skb);
977 if (skb->protocol == htons(ETH_P_IP) &&
978 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
979 truncate = true;
980
981 if (skb->protocol == htons(ETH_P_IPV6)) {
982 int thoff;
983
984 if (skb_transport_header_was_set(skb))
985 thoff = skb_transport_offset(skb);
986 else
987 thoff = nhoff + sizeof(struct ipv6hdr);
988 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
989 truncate = true;
990 }
991
992 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
993 goto tx_err;
994
995 __clear_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags);
996 IPCB(skb)->flags = 0;
997
998 /* For collect_md mode, derive fl6 from the tunnel key,
999 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
1000 */
1001 if (t->parms.collect_md) {
1002 const struct ip_tunnel_key *key;
1003 struct erspan_metadata *md;
1004 __be32 tun_id;
1005
1006 tun_info = skb_tunnel_info_txcheck(skb);
1007 if (IS_ERR(tun_info) ||
1008 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
1009 goto tx_err;
1010
1011 key = &tun_info->key;
1012 memset(&fl6, 0, sizeof(fl6));
1013 fl6.flowi6_proto = IPPROTO_GRE;
1014 fl6.daddr = key->u.ipv6.dst;
1015 fl6.flowlabel = key->label;
1016 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1017 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1018
1019 dsfield = key->tos;
1020 if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
1021 tun_info->key.tun_flags))
1022 goto tx_err;
1023 if (tun_info->options_len < sizeof(*md))
1024 goto tx_err;
1025 md = ip_tunnel_info_opts(tun_info);
1026
1027 tun_id = tunnel_id_to_key32(key->tun_id);
1028 if (md->version == 1) {
1029 erspan_build_header(skb,
1030 ntohl(tun_id),
1031 ntohl(md->u.index), truncate,
1032 false);
1033 proto = htons(ETH_P_ERSPAN);
1034 } else if (md->version == 2) {
1035 erspan_build_header_v2(skb,
1036 ntohl(tun_id),
1037 md->u.md2.dir,
1038 get_hwid(&md->u.md2),
1039 truncate, false);
1040 proto = htons(ETH_P_ERSPAN2);
1041 } else {
1042 goto tx_err;
1043 }
1044 } else {
1045 switch (skb->protocol) {
1046 case htons(ETH_P_IP):
1047 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1048 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1049 &dsfield, &encap_limit);
1050 break;
1051 case htons(ETH_P_IPV6):
1052 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1053 goto tx_err;
1054 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1055 &dsfield, &encap_limit))
1056 goto tx_err;
1057 break;
1058 default:
1059 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1060 break;
1061 }
1062
1063 if (t->parms.erspan_ver == 1) {
1064 erspan_build_header(skb, ntohl(t->parms.o_key),
1065 t->parms.index,
1066 truncate, false);
1067 proto = htons(ETH_P_ERSPAN);
1068 } else if (t->parms.erspan_ver == 2) {
1069 erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1070 t->parms.dir,
1071 t->parms.hwid,
1072 truncate, false);
1073 proto = htons(ETH_P_ERSPAN2);
1074 } else {
1075 goto tx_err;
1076 }
1077
1078 fl6.daddr = t->parms.raddr;
1079 }
1080
1081 /* Push GRE header. */
1082 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
1083 gre_build_header(skb, 8, flags, proto, 0,
1084 htonl(atomic_fetch_inc(&t->o_seqno)));
1085
1086 /* TooBig packet may have updated dst->dev's mtu */
1087 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1088 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1089
1090 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1091 NEXTHDR_GRE);
1092 if (err != 0) {
1093 /* XXX: send ICMP error even if DF is not set. */
1094 if (err == -EMSGSIZE) {
1095 if (skb->protocol == htons(ETH_P_IP))
1096 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1097 ICMP_FRAG_NEEDED, htonl(mtu));
1098 else
1099 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1100 }
1101
1102 goto tx_err;
1103 }
1104 return NETDEV_TX_OK;
1105
1106 tx_err:
1107 if (!IS_ERR(tun_info))
1108 DEV_STATS_INC(dev, tx_errors);
1109 DEV_STATS_INC(dev, tx_dropped);
1110 kfree_skb(skb);
1111 return NETDEV_TX_OK;
1112 }
1113
1114 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1115 {
1116 struct net_device *dev = t->dev;
1117 struct __ip6_tnl_parm *p = &t->parms;
1118 struct flowi6 *fl6 = &t->fl.u.ip6;
1119
1120 if (dev->type != ARPHRD_ETHER) {
1121 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1122 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1123 }
1124
1125 /* Set up flowi template */
1126 fl6->saddr = p->laddr;
1127 fl6->daddr = p->raddr;
1128 fl6->flowi6_oif = p->link;
1129 fl6->flowlabel = 0;
1130 fl6->flowi6_proto = IPPROTO_GRE;
1131 fl6->fl6_gre_key = t->parms.o_key;
1132
1133 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1134 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1135 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1136 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1137
1138 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1139 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1140
1141 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1142 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1143 dev->flags |= IFF_POINTOPOINT;
1144 else
1145 dev->flags &= ~IFF_POINTOPOINT;
1146 }
1147
1148 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1149 int t_hlen)
1150 {
1151 const struct __ip6_tnl_parm *p = &t->parms;
1152 struct net_device *dev = t->dev;
1153
1154 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1155 int strict = (ipv6_addr_type(&p->raddr) &
1156 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1157
1158 struct rt6_info *rt = rt6_lookup(t->net,
1159 &p->raddr, &p->laddr,
1160 p->link, NULL, strict);
1161
1162 if (!rt)
1163 return;
1164
1165 if (rt->dst.dev) {
1166 unsigned short dst_len = rt->dst.dev->hard_header_len +
1167 t_hlen;
1168
1169 if (t->dev->header_ops)
1170 dev->hard_header_len = dst_len;
1171 else
1172 dev->needed_headroom = dst_len;
1173
1174 if (set_mtu) {
1175 int mtu = rt->dst.dev->mtu - t_hlen;
1176
1177 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1178 mtu -= 8;
1179 if (dev->type == ARPHRD_ETHER)
1180 mtu -= ETH_HLEN;
1181
1182 if (mtu < IPV6_MIN_MTU)
1183 mtu = IPV6_MIN_MTU;
1184 WRITE_ONCE(dev->mtu, mtu);
1185 }
1186 }
1187 ip6_rt_put(rt);
1188 }
1189 }
1190
1191 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1192 {
1193 int t_hlen;
1194
1195 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1196 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1197
1198 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1199
1200 if (tunnel->dev->header_ops)
1201 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1202 else
1203 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1204
1205 return t_hlen;
1206 }
1207
1208 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1209 {
1210 ip6gre_tnl_link_config_common(t);
1211 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1212 }
1213
1214 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1215 const struct __ip6_tnl_parm *p)
1216 {
1217 t->parms.laddr = p->laddr;
1218 t->parms.raddr = p->raddr;
1219 t->parms.flags = p->flags;
1220 t->parms.hop_limit = p->hop_limit;
1221 t->parms.encap_limit = p->encap_limit;
1222 t->parms.flowinfo = p->flowinfo;
1223 t->parms.link = p->link;
1224 t->parms.proto = p->proto;
1225 t->parms.i_key = p->i_key;
1226 t->parms.o_key = p->o_key;
1227 ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags);
1228 ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags);
1229 t->parms.fwmark = p->fwmark;
1230 t->parms.erspan_ver = p->erspan_ver;
1231 t->parms.index = p->index;
1232 t->parms.dir = p->dir;
1233 t->parms.hwid = p->hwid;
1234 dst_cache_reset(&t->dst_cache);
1235 }
1236
1237 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1238 int set_mtu)
1239 {
1240 ip6gre_tnl_copy_tnl_parm(t, p);
1241 ip6gre_tnl_link_config(t, set_mtu);
1242 return 0;
1243 }
1244
1245 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1246 const struct ip6_tnl_parm2 *u)
1247 {
1248 p->laddr = u->laddr;
1249 p->raddr = u->raddr;
1250 p->flags = u->flags;
1251 p->hop_limit = u->hop_limit;
1252 p->encap_limit = u->encap_limit;
1253 p->flowinfo = u->flowinfo;
1254 p->link = u->link;
1255 p->i_key = u->i_key;
1256 p->o_key = u->o_key;
1257 gre_flags_to_tnl_flags(p->i_flags, u->i_flags);
1258 gre_flags_to_tnl_flags(p->o_flags, u->o_flags);
1259 memcpy(p->name, u->name, sizeof(u->name));
1260 }
1261
1262 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1263 const struct __ip6_tnl_parm *p)
1264 {
1265 u->proto = IPPROTO_GRE;
1266 u->laddr = p->laddr;
1267 u->raddr = p->raddr;
1268 u->flags = p->flags;
1269 u->hop_limit = p->hop_limit;
1270 u->encap_limit = p->encap_limit;
1271 u->flowinfo = p->flowinfo;
1272 u->link = p->link;
1273 u->i_key = p->i_key;
1274 u->o_key = p->o_key;
1275 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1276 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1277 memcpy(u->name, p->name, sizeof(u->name));
1278 }
1279
1280 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1281 struct ifreq *ifr, void __user *data,
1282 int cmd)
1283 {
1284 int err = 0;
1285 struct ip6_tnl_parm2 p;
1286 struct __ip6_tnl_parm p1;
1287 struct ip6_tnl *t = netdev_priv(dev);
1288 struct net *net = t->net;
1289 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1290
1291 memset(&p1, 0, sizeof(p1));
1292
1293 switch (cmd) {
1294 case SIOCGETTUNNEL:
1295 if (dev == ign->fb_tunnel_dev) {
1296 if (copy_from_user(&p, data, sizeof(p))) {
1297 err = -EFAULT;
1298 break;
1299 }
1300 ip6gre_tnl_parm_from_user(&p1, &p);
1301 t = ip6gre_tunnel_locate(net, &p1, 0);
1302 if (!t)
1303 t = netdev_priv(dev);
1304 }
1305 memset(&p, 0, sizeof(p));
1306 ip6gre_tnl_parm_to_user(&p, &t->parms);
1307 if (copy_to_user(data, &p, sizeof(p)))
1308 err = -EFAULT;
1309 break;
1310
1311 case SIOCADDTUNNEL:
1312 case SIOCCHGTUNNEL:
1313 err = -EPERM;
1314 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1315 goto done;
1316
1317 err = -EFAULT;
1318 if (copy_from_user(&p, data, sizeof(p)))
1319 goto done;
1320
1321 err = -EINVAL;
1322 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1323 goto done;
1324
1325 if (!(p.i_flags&GRE_KEY))
1326 p.i_key = 0;
1327 if (!(p.o_flags&GRE_KEY))
1328 p.o_key = 0;
1329
1330 ip6gre_tnl_parm_from_user(&p1, &p);
1331 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1332
1333 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1334 if (t) {
1335 if (t->dev != dev) {
1336 err = -EEXIST;
1337 break;
1338 }
1339 } else {
1340 t = netdev_priv(dev);
1341
1342 ip6gre_tunnel_unlink(ign, t);
1343 synchronize_net();
1344 ip6gre_tnl_change(t, &p1, 1);
1345 ip6gre_tunnel_link(ign, t);
1346 netdev_state_change(dev);
1347 }
1348 }
1349
1350 if (t) {
1351 err = 0;
1352
1353 memset(&p, 0, sizeof(p));
1354 ip6gre_tnl_parm_to_user(&p, &t->parms);
1355 if (copy_to_user(data, &p, sizeof(p)))
1356 err = -EFAULT;
1357 } else
1358 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1359 break;
1360
1361 case SIOCDELTUNNEL:
1362 err = -EPERM;
1363 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1364 goto done;
1365
1366 if (dev == ign->fb_tunnel_dev) {
1367 err = -EFAULT;
1368 if (copy_from_user(&p, data, sizeof(p)))
1369 goto done;
1370 err = -ENOENT;
1371 ip6gre_tnl_parm_from_user(&p1, &p);
1372 t = ip6gre_tunnel_locate(net, &p1, 0);
1373 if (!t)
1374 goto done;
1375 err = -EPERM;
1376 if (t == netdev_priv(ign->fb_tunnel_dev))
1377 goto done;
1378 dev = t->dev;
1379 }
1380 unregister_netdevice(dev);
1381 err = 0;
1382 break;
1383
1384 default:
1385 err = -EINVAL;
1386 }
1387
1388 done:
1389 return err;
1390 }
1391
1392 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1393 unsigned short type, const void *daddr,
1394 const void *saddr, unsigned int len)
1395 {
1396 struct ip6_tnl *t = netdev_priv(dev);
1397 struct ipv6hdr *ipv6h;
1398 __be16 *p;
1399
1400 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1401 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1402 t->fl.u.ip6.flowlabel,
1403 true, &t->fl.u.ip6));
1404 ipv6h->hop_limit = t->parms.hop_limit;
1405 ipv6h->nexthdr = NEXTHDR_GRE;
1406 ipv6h->saddr = t->parms.laddr;
1407 ipv6h->daddr = t->parms.raddr;
1408
1409 p = (__be16 *)(ipv6h + 1);
1410 p[0] = ip_tunnel_flags_to_be16(t->parms.o_flags);
1411 p[1] = htons(type);
1412
1413 /*
1414 * Set the source hardware address.
1415 */
1416
1417 if (saddr)
1418 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1419 if (daddr)
1420 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1421 if (!ipv6_addr_any(&ipv6h->daddr))
1422 return t->hlen;
1423
1424 return -t->hlen;
1425 }
1426
1427 static const struct header_ops ip6gre_header_ops = {
1428 .create = ip6gre_header,
1429 };
1430
1431 static const struct net_device_ops ip6gre_netdev_ops = {
1432 .ndo_init = ip6gre_tunnel_init,
1433 .ndo_uninit = ip6gre_tunnel_uninit,
1434 .ndo_start_xmit = ip6gre_tunnel_xmit,
1435 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate,
1436 .ndo_change_mtu = ip6_tnl_change_mtu,
1437 .ndo_get_iflink = ip6_tnl_get_iflink,
1438 };
1439
1440 static void ip6gre_dev_free(struct net_device *dev)
1441 {
1442 struct ip6_tnl *t = netdev_priv(dev);
1443
1444 gro_cells_destroy(&t->gro_cells);
1445 dst_cache_destroy(&t->dst_cache);
1446 }
1447
1448 static void ip6gre_tunnel_setup(struct net_device *dev)
1449 {
1450 dev->netdev_ops = &ip6gre_netdev_ops;
1451 dev->needs_free_netdev = true;
1452 dev->priv_destructor = ip6gre_dev_free;
1453
1454 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1455 dev->type = ARPHRD_IP6GRE;
1456
1457 dev->flags |= IFF_NOARP;
1458 dev->addr_len = sizeof(struct in6_addr);
1459 netif_keep_dst(dev);
1460 /* This perm addr will be used as interface identifier by IPv6 */
1461 dev->addr_assign_type = NET_ADDR_RANDOM;
1462 eth_random_addr(dev->perm_addr);
1463 }
1464
1465 #define GRE6_FEATURES (NETIF_F_SG | \
1466 NETIF_F_FRAGLIST | \
1467 NETIF_F_HIGHDMA | \
1468 NETIF_F_HW_CSUM)
1469
1470 static void ip6gre_tnl_init_features(struct net_device *dev)
1471 {
1472 struct ip6_tnl *nt = netdev_priv(dev);
1473
1474 dev->features |= GRE6_FEATURES | NETIF_F_LLTX;
1475 dev->hw_features |= GRE6_FEATURES;
1476
1477 /* TCP offload with GRE SEQ is not supported, nor can we support 2
1478 * levels of outer headers requiring an update.
1479 */
1480 if (test_bit(IP_TUNNEL_SEQ_BIT, nt->parms.o_flags))
1481 return;
1482 if (test_bit(IP_TUNNEL_CSUM_BIT, nt->parms.o_flags) &&
1483 nt->encap.type != TUNNEL_ENCAP_NONE)
1484 return;
1485
1486 dev->features |= NETIF_F_GSO_SOFTWARE;
1487 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1488 }
1489
1490 static int ip6gre_tunnel_init_common(struct net_device *dev)
1491 {
1492 struct ip6_tnl *tunnel;
1493 int ret;
1494 int t_hlen;
1495
1496 tunnel = netdev_priv(dev);
1497
1498 tunnel->dev = dev;
1499 tunnel->net = dev_net(dev);
1500 strcpy(tunnel->parms.name, dev->name);
1501
1502 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1503 if (ret)
1504 return ret;
1505
1506 ret = gro_cells_init(&tunnel->gro_cells, dev);
1507 if (ret)
1508 goto cleanup_dst_cache_init;
1509
1510 t_hlen = ip6gre_calc_hlen(tunnel);
1511 dev->mtu = ETH_DATA_LEN - t_hlen;
1512 if (dev->type == ARPHRD_ETHER)
1513 dev->mtu -= ETH_HLEN;
1514 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1515 dev->mtu -= 8;
1516
1517 if (tunnel->parms.collect_md) {
1518 netif_keep_dst(dev);
1519 }
1520 ip6gre_tnl_init_features(dev);
1521
1522 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1523 netdev_lockdep_set_classes(dev);
1524 return 0;
1525
1526 cleanup_dst_cache_init:
1527 dst_cache_destroy(&tunnel->dst_cache);
1528 return ret;
1529 }
1530
1531 static int ip6gre_tunnel_init(struct net_device *dev)
1532 {
1533 struct ip6_tnl *tunnel;
1534 int ret;
1535
1536 ret = ip6gre_tunnel_init_common(dev);
1537 if (ret)
1538 return ret;
1539
1540 tunnel = netdev_priv(dev);
1541
1542 if (tunnel->parms.collect_md)
1543 return 0;
1544
1545 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1546 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1547
1548 if (ipv6_addr_any(&tunnel->parms.raddr))
1549 dev->header_ops = &ip6gre_header_ops;
1550
1551 return 0;
1552 }
1553
1554 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1555 {
1556 struct ip6_tnl *tunnel = netdev_priv(dev);
1557
1558 tunnel->dev = dev;
1559 tunnel->net = dev_net(dev);
1560 strcpy(tunnel->parms.name, dev->name);
1561
1562 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1563 }
1564
1565 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1566 .handler = gre_rcv,
1567 .err_handler = ip6gre_err,
1568 .flags = INET6_PROTO_FINAL,
1569 };
1570
1571 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1572 {
1573 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1574 struct net_device *dev, *aux;
1575 int prio;
1576
1577 for_each_netdev_safe(net, dev, aux)
1578 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1579 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1580 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1581 unregister_netdevice_queue(dev, head);
1582
1583 for (prio = 0; prio < 4; prio++) {
1584 int h;
1585 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1586 struct ip6_tnl *t;
1587
1588 t = rtnl_dereference(ign->tunnels[prio][h]);
1589
1590 while (t) {
1591 /* If dev is in the same netns, it has already
1592 * been added to the list by the previous loop.
1593 */
1594 if (!net_eq(dev_net(t->dev), net))
1595 unregister_netdevice_queue(t->dev,
1596 head);
1597 t = rtnl_dereference(t->next);
1598 }
1599 }
1600 }
1601 }
1602
1603 static int __net_init ip6gre_init_net(struct net *net)
1604 {
1605 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1606 struct net_device *ndev;
1607 int err;
1608
1609 if (!net_has_fallback_tunnels(net))
1610 return 0;
1611 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1612 NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1613 if (!ndev) {
1614 err = -ENOMEM;
1615 goto err_alloc_dev;
1616 }
1617 ign->fb_tunnel_dev = ndev;
1618 dev_net_set(ign->fb_tunnel_dev, net);
1619 /* FB netdevice is special: we have one, and only one per netns.
1620 * Allowing to move it to another netns is clearly unsafe.
1621 */
1622 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1623
1624
1625 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1626 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1627
1628 err = register_netdev(ign->fb_tunnel_dev);
1629 if (err)
1630 goto err_reg_dev;
1631
1632 rcu_assign_pointer(ign->tunnels_wc[0],
1633 netdev_priv(ign->fb_tunnel_dev));
1634 return 0;
1635
1636 err_reg_dev:
1637 free_netdev(ndev);
1638 err_alloc_dev:
1639 return err;
1640 }
1641
1642 static void __net_exit ip6gre_exit_batch_rtnl(struct list_head *net_list,
1643 struct list_head *dev_to_kill)
1644 {
1645 struct net *net;
1646
1647 ASSERT_RTNL();
1648 list_for_each_entry(net, net_list, exit_list)
1649 ip6gre_destroy_tunnels(net, dev_to_kill);
1650 }
1651
1652 static struct pernet_operations ip6gre_net_ops = {
1653 .init = ip6gre_init_net,
1654 .exit_batch_rtnl = ip6gre_exit_batch_rtnl,
1655 .id = &ip6gre_net_id,
1656 .size = sizeof(struct ip6gre_net),
1657 };
1658
1659 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1660 struct netlink_ext_ack *extack)
1661 {
1662 __be16 flags;
1663
1664 if (!data)
1665 return 0;
1666
1667 flags = 0;
1668 if (data[IFLA_GRE_IFLAGS])
1669 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1670 if (data[IFLA_GRE_OFLAGS])
1671 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1672 if (flags & (GRE_VERSION|GRE_ROUTING))
1673 return -EINVAL;
1674
1675 return 0;
1676 }
1677
1678 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1679 struct netlink_ext_ack *extack)
1680 {
1681 struct in6_addr daddr;
1682
1683 if (tb[IFLA_ADDRESS]) {
1684 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1685 return -EINVAL;
1686 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1687 return -EADDRNOTAVAIL;
1688 }
1689
1690 if (!data)
1691 goto out;
1692
1693 if (data[IFLA_GRE_REMOTE]) {
1694 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1695 if (ipv6_addr_any(&daddr))
1696 return -EINVAL;
1697 }
1698
1699 out:
1700 return ip6gre_tunnel_validate(tb, data, extack);
1701 }
1702
1703 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1704 struct netlink_ext_ack *extack)
1705 {
1706 __be16 flags = 0;
1707 int ret, ver = 0;
1708
1709 if (!data)
1710 return 0;
1711
1712 ret = ip6gre_tap_validate(tb, data, extack);
1713 if (ret)
1714 return ret;
1715
1716 /* ERSPAN should only have GRE sequence and key flag */
1717 if (data[IFLA_GRE_OFLAGS])
1718 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1719 if (data[IFLA_GRE_IFLAGS])
1720 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1721 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1722 flags != (GRE_SEQ | GRE_KEY))
1723 return -EINVAL;
1724
1725 /* ERSPAN Session ID only has 10-bit. Since we reuse
1726 * 32-bit key field as ID, check it's range.
1727 */
1728 if (data[IFLA_GRE_IKEY] &&
1729 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1730 return -EINVAL;
1731
1732 if (data[IFLA_GRE_OKEY] &&
1733 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1734 return -EINVAL;
1735
1736 if (data[IFLA_GRE_ERSPAN_VER]) {
1737 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1738 if (ver != 1 && ver != 2)
1739 return -EINVAL;
1740 }
1741
1742 if (ver == 1) {
1743 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1744 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1745
1746 if (index & ~INDEX_MASK)
1747 return -EINVAL;
1748 }
1749 } else if (ver == 2) {
1750 if (data[IFLA_GRE_ERSPAN_DIR]) {
1751 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1752
1753 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1754 return -EINVAL;
1755 }
1756
1757 if (data[IFLA_GRE_ERSPAN_HWID]) {
1758 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1759
1760 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1761 return -EINVAL;
1762 }
1763 }
1764
1765 return 0;
1766 }
1767
1768 static void ip6erspan_set_version(struct nlattr *data[],
1769 struct __ip6_tnl_parm *parms)
1770 {
1771 if (!data)
1772 return;
1773
1774 parms->erspan_ver = 1;
1775 if (data[IFLA_GRE_ERSPAN_VER])
1776 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1777
1778 if (parms->erspan_ver == 1) {
1779 if (data[IFLA_GRE_ERSPAN_INDEX])
1780 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1781 } else if (parms->erspan_ver == 2) {
1782 if (data[IFLA_GRE_ERSPAN_DIR])
1783 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1784 if (data[IFLA_GRE_ERSPAN_HWID])
1785 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1786 }
1787 }
1788
1789 static void ip6gre_netlink_parms(struct nlattr *data[],
1790 struct __ip6_tnl_parm *parms)
1791 {
1792 memset(parms, 0, sizeof(*parms));
1793
1794 if (!data)
1795 return;
1796
1797 if (data[IFLA_GRE_LINK])
1798 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1799
1800 if (data[IFLA_GRE_IFLAGS])
1801 gre_flags_to_tnl_flags(parms->i_flags,
1802 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1803
1804 if (data[IFLA_GRE_OFLAGS])
1805 gre_flags_to_tnl_flags(parms->o_flags,
1806 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1807
1808 if (data[IFLA_GRE_IKEY])
1809 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1810
1811 if (data[IFLA_GRE_OKEY])
1812 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1813
1814 if (data[IFLA_GRE_LOCAL])
1815 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1816
1817 if (data[IFLA_GRE_REMOTE])
1818 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1819
1820 if (data[IFLA_GRE_TTL])
1821 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1822
1823 if (data[IFLA_GRE_ENCAP_LIMIT])
1824 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1825
1826 if (data[IFLA_GRE_FLOWINFO])
1827 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1828
1829 if (data[IFLA_GRE_FLAGS])
1830 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1831
1832 if (data[IFLA_GRE_FWMARK])
1833 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1834
1835 if (data[IFLA_GRE_COLLECT_METADATA])
1836 parms->collect_md = true;
1837 }
1838
1839 static int ip6gre_tap_init(struct net_device *dev)
1840 {
1841 int ret;
1842
1843 ret = ip6gre_tunnel_init_common(dev);
1844 if (ret)
1845 return ret;
1846
1847 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1848
1849 return 0;
1850 }
1851
1852 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1853 .ndo_init = ip6gre_tap_init,
1854 .ndo_uninit = ip6gre_tunnel_uninit,
1855 .ndo_start_xmit = ip6gre_tunnel_xmit,
1856 .ndo_set_mac_address = eth_mac_addr,
1857 .ndo_validate_addr = eth_validate_addr,
1858 .ndo_change_mtu = ip6_tnl_change_mtu,
1859 .ndo_get_iflink = ip6_tnl_get_iflink,
1860 };
1861
1862 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1863 {
1864 int t_hlen;
1865
1866 tunnel->tun_hlen = 8;
1867 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1868 erspan_hdr_len(tunnel->parms.erspan_ver);
1869
1870 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1871 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1872 return t_hlen;
1873 }
1874
1875 static int ip6erspan_tap_init(struct net_device *dev)
1876 {
1877 struct ip6_tnl *tunnel;
1878 int t_hlen;
1879 int ret;
1880
1881 tunnel = netdev_priv(dev);
1882
1883 tunnel->dev = dev;
1884 tunnel->net = dev_net(dev);
1885 strcpy(tunnel->parms.name, dev->name);
1886
1887 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1888 if (ret)
1889 return ret;
1890
1891 ret = gro_cells_init(&tunnel->gro_cells, dev);
1892 if (ret)
1893 goto cleanup_dst_cache_init;
1894
1895 t_hlen = ip6erspan_calc_hlen(tunnel);
1896 dev->mtu = ETH_DATA_LEN - t_hlen;
1897 if (dev->type == ARPHRD_ETHER)
1898 dev->mtu -= ETH_HLEN;
1899 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1900 dev->mtu -= 8;
1901
1902 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1903 ip6erspan_tnl_link_config(tunnel, 1);
1904
1905 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1906 netdev_lockdep_set_classes(dev);
1907 return 0;
1908
1909 cleanup_dst_cache_init:
1910 dst_cache_destroy(&tunnel->dst_cache);
1911 return ret;
1912 }
1913
1914 static const struct net_device_ops ip6erspan_netdev_ops = {
1915 .ndo_init = ip6erspan_tap_init,
1916 .ndo_uninit = ip6erspan_tunnel_uninit,
1917 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1918 .ndo_set_mac_address = eth_mac_addr,
1919 .ndo_validate_addr = eth_validate_addr,
1920 .ndo_change_mtu = ip6_tnl_change_mtu,
1921 .ndo_get_iflink = ip6_tnl_get_iflink,
1922 };
1923
1924 static void ip6gre_tap_setup(struct net_device *dev)
1925 {
1926
1927 ether_setup(dev);
1928
1929 dev->max_mtu = 0;
1930 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1931 dev->needs_free_netdev = true;
1932 dev->priv_destructor = ip6gre_dev_free;
1933
1934 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1935 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1936 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1937 netif_keep_dst(dev);
1938 }
1939
1940 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1941 struct ip_tunnel_encap *ipencap)
1942 {
1943 bool ret = false;
1944
1945 memset(ipencap, 0, sizeof(*ipencap));
1946
1947 if (!data)
1948 return ret;
1949
1950 if (data[IFLA_GRE_ENCAP_TYPE]) {
1951 ret = true;
1952 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1953 }
1954
1955 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1956 ret = true;
1957 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1958 }
1959
1960 if (data[IFLA_GRE_ENCAP_SPORT]) {
1961 ret = true;
1962 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1963 }
1964
1965 if (data[IFLA_GRE_ENCAP_DPORT]) {
1966 ret = true;
1967 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1968 }
1969
1970 return ret;
1971 }
1972
1973 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1974 struct nlattr *tb[], struct nlattr *data[],
1975 struct netlink_ext_ack *extack)
1976 {
1977 struct ip6_tnl *nt;
1978 struct ip_tunnel_encap ipencap;
1979 int err;
1980
1981 nt = netdev_priv(dev);
1982
1983 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1984 int err = ip6_tnl_encap_setup(nt, &ipencap);
1985
1986 if (err < 0)
1987 return err;
1988 }
1989
1990 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1991 eth_hw_addr_random(dev);
1992
1993 nt->dev = dev;
1994 nt->net = dev_net(dev);
1995
1996 err = register_netdevice(dev);
1997 if (err)
1998 goto out;
1999
2000 if (tb[IFLA_MTU])
2001 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2002
2003 out:
2004 return err;
2005 }
2006
2007 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2008 struct nlattr *tb[], struct nlattr *data[],
2009 struct netlink_ext_ack *extack)
2010 {
2011 struct ip6_tnl *nt = netdev_priv(dev);
2012 struct net *net = dev_net(dev);
2013 struct ip6gre_net *ign;
2014 int err;
2015
2016 ip6gre_netlink_parms(data, &nt->parms);
2017 ign = net_generic(net, ip6gre_net_id);
2018
2019 if (nt->parms.collect_md) {
2020 if (rtnl_dereference(ign->collect_md_tun))
2021 return -EEXIST;
2022 } else {
2023 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2024 return -EEXIST;
2025 }
2026
2027 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2028 if (!err) {
2029 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2030 ip6gre_tunnel_link_md(ign, nt);
2031 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2032 }
2033 return err;
2034 }
2035
2036 static struct ip6_tnl *
2037 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2038 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2039 struct netlink_ext_ack *extack)
2040 {
2041 struct ip6_tnl *t, *nt = netdev_priv(dev);
2042 struct net *net = nt->net;
2043 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2044 struct ip_tunnel_encap ipencap;
2045
2046 if (dev == ign->fb_tunnel_dev)
2047 return ERR_PTR(-EINVAL);
2048
2049 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2050 int err = ip6_tnl_encap_setup(nt, &ipencap);
2051
2052 if (err < 0)
2053 return ERR_PTR(err);
2054 }
2055
2056 ip6gre_netlink_parms(data, p_p);
2057
2058 t = ip6gre_tunnel_locate(net, p_p, 0);
2059
2060 if (t) {
2061 if (t->dev != dev)
2062 return ERR_PTR(-EEXIST);
2063 } else {
2064 t = nt;
2065 }
2066
2067 return t;
2068 }
2069
2070 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2071 struct nlattr *data[],
2072 struct netlink_ext_ack *extack)
2073 {
2074 struct ip6_tnl *t = netdev_priv(dev);
2075 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2076 struct __ip6_tnl_parm p;
2077
2078 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2079 if (IS_ERR(t))
2080 return PTR_ERR(t);
2081
2082 ip6gre_tunnel_unlink_md(ign, t);
2083 ip6gre_tunnel_unlink(ign, t);
2084 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2085 ip6gre_tunnel_link_md(ign, t);
2086 ip6gre_tunnel_link(ign, t);
2087 return 0;
2088 }
2089
2090 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2091 {
2092 struct net *net = dev_net(dev);
2093 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2094
2095 if (dev != ign->fb_tunnel_dev)
2096 unregister_netdevice_queue(dev, head);
2097 }
2098
2099 static size_t ip6gre_get_size(const struct net_device *dev)
2100 {
2101 return
2102 /* IFLA_GRE_LINK */
2103 nla_total_size(4) +
2104 /* IFLA_GRE_IFLAGS */
2105 nla_total_size(2) +
2106 /* IFLA_GRE_OFLAGS */
2107 nla_total_size(2) +
2108 /* IFLA_GRE_IKEY */
2109 nla_total_size(4) +
2110 /* IFLA_GRE_OKEY */
2111 nla_total_size(4) +
2112 /* IFLA_GRE_LOCAL */
2113 nla_total_size(sizeof(struct in6_addr)) +
2114 /* IFLA_GRE_REMOTE */
2115 nla_total_size(sizeof(struct in6_addr)) +
2116 /* IFLA_GRE_TTL */
2117 nla_total_size(1) +
2118 /* IFLA_GRE_ENCAP_LIMIT */
2119 nla_total_size(1) +
2120 /* IFLA_GRE_FLOWINFO */
2121 nla_total_size(4) +
2122 /* IFLA_GRE_FLAGS */
2123 nla_total_size(4) +
2124 /* IFLA_GRE_ENCAP_TYPE */
2125 nla_total_size(2) +
2126 /* IFLA_GRE_ENCAP_FLAGS */
2127 nla_total_size(2) +
2128 /* IFLA_GRE_ENCAP_SPORT */
2129 nla_total_size(2) +
2130 /* IFLA_GRE_ENCAP_DPORT */
2131 nla_total_size(2) +
2132 /* IFLA_GRE_COLLECT_METADATA */
2133 nla_total_size(0) +
2134 /* IFLA_GRE_FWMARK */
2135 nla_total_size(4) +
2136 /* IFLA_GRE_ERSPAN_INDEX */
2137 nla_total_size(4) +
2138 0;
2139 }
2140
2141 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2142 {
2143 struct ip6_tnl *t = netdev_priv(dev);
2144 struct __ip6_tnl_parm *p = &t->parms;
2145 IP_TUNNEL_DECLARE_FLAGS(o_flags);
2146
2147 ip_tunnel_flags_copy(o_flags, p->o_flags);
2148
2149 if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2150 if (!p->collect_md)
2151 __set_bit(IP_TUNNEL_KEY_BIT, o_flags);
2152
2153 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2154 goto nla_put_failure;
2155
2156 if (p->erspan_ver == 1) {
2157 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2158 goto nla_put_failure;
2159 } else {
2160 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2161 goto nla_put_failure;
2162 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2163 goto nla_put_failure;
2164 }
2165 }
2166
2167 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2168 nla_put_be16(skb, IFLA_GRE_IFLAGS,
2169 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2170 nla_put_be16(skb, IFLA_GRE_OFLAGS,
2171 gre_tnl_flags_to_gre_flags(o_flags)) ||
2172 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2173 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2174 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2175 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2176 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2177 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2178 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2179 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2180 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2181 goto nla_put_failure;
2182
2183 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2184 t->encap.type) ||
2185 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2186 t->encap.sport) ||
2187 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2188 t->encap.dport) ||
2189 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2190 t->encap.flags))
2191 goto nla_put_failure;
2192
2193 if (p->collect_md) {
2194 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2195 goto nla_put_failure;
2196 }
2197
2198 return 0;
2199
2200 nla_put_failure:
2201 return -EMSGSIZE;
2202 }
2203
2204 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2205 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2206 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2207 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2208 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2209 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2210 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) },
2211 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) },
2212 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2213 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2214 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2215 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2216 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2217 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2218 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2219 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2220 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2221 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2222 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2223 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2224 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2225 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2226 };
2227
2228 static void ip6erspan_tap_setup(struct net_device *dev)
2229 {
2230 ether_setup(dev);
2231
2232 dev->max_mtu = 0;
2233 dev->netdev_ops = &ip6erspan_netdev_ops;
2234 dev->needs_free_netdev = true;
2235 dev->priv_destructor = ip6gre_dev_free;
2236
2237 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
2238 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2239 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2240 netif_keep_dst(dev);
2241 }
2242
2243 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2244 struct nlattr *tb[], struct nlattr *data[],
2245 struct netlink_ext_ack *extack)
2246 {
2247 struct ip6_tnl *nt = netdev_priv(dev);
2248 struct net *net = dev_net(dev);
2249 struct ip6gre_net *ign;
2250 int err;
2251
2252 ip6gre_netlink_parms(data, &nt->parms);
2253 ip6erspan_set_version(data, &nt->parms);
2254 ign = net_generic(net, ip6gre_net_id);
2255
2256 if (nt->parms.collect_md) {
2257 if (rtnl_dereference(ign->collect_md_tun_erspan))
2258 return -EEXIST;
2259 } else {
2260 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2261 return -EEXIST;
2262 }
2263
2264 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2265 if (!err) {
2266 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2267 ip6erspan_tunnel_link_md(ign, nt);
2268 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2269 }
2270 return err;
2271 }
2272
2273 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2274 {
2275 ip6gre_tnl_link_config_common(t);
2276 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2277 }
2278
2279 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2280 const struct __ip6_tnl_parm *p, int set_mtu)
2281 {
2282 ip6gre_tnl_copy_tnl_parm(t, p);
2283 ip6erspan_tnl_link_config(t, set_mtu);
2284 return 0;
2285 }
2286
2287 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2288 struct nlattr *data[],
2289 struct netlink_ext_ack *extack)
2290 {
2291 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2292 struct __ip6_tnl_parm p;
2293 struct ip6_tnl *t;
2294
2295 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2296 if (IS_ERR(t))
2297 return PTR_ERR(t);
2298
2299 ip6erspan_set_version(data, &p);
2300 ip6gre_tunnel_unlink_md(ign, t);
2301 ip6gre_tunnel_unlink(ign, t);
2302 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2303 ip6erspan_tunnel_link_md(ign, t);
2304 ip6gre_tunnel_link(ign, t);
2305 return 0;
2306 }
2307
2308 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2309 .kind = "ip6gre",
2310 .maxtype = IFLA_GRE_MAX,
2311 .policy = ip6gre_policy,
2312 .priv_size = sizeof(struct ip6_tnl),
2313 .setup = ip6gre_tunnel_setup,
2314 .validate = ip6gre_tunnel_validate,
2315 .newlink = ip6gre_newlink,
2316 .changelink = ip6gre_changelink,
2317 .dellink = ip6gre_dellink,
2318 .get_size = ip6gre_get_size,
2319 .fill_info = ip6gre_fill_info,
2320 .get_link_net = ip6_tnl_get_link_net,
2321 };
2322
2323 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2324 .kind = "ip6gretap",
2325 .maxtype = IFLA_GRE_MAX,
2326 .policy = ip6gre_policy,
2327 .priv_size = sizeof(struct ip6_tnl),
2328 .setup = ip6gre_tap_setup,
2329 .validate = ip6gre_tap_validate,
2330 .newlink = ip6gre_newlink,
2331 .changelink = ip6gre_changelink,
2332 .get_size = ip6gre_get_size,
2333 .fill_info = ip6gre_fill_info,
2334 .get_link_net = ip6_tnl_get_link_net,
2335 };
2336
2337 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2338 .kind = "ip6erspan",
2339 .maxtype = IFLA_GRE_MAX,
2340 .policy = ip6gre_policy,
2341 .priv_size = sizeof(struct ip6_tnl),
2342 .setup = ip6erspan_tap_setup,
2343 .validate = ip6erspan_tap_validate,
2344 .newlink = ip6erspan_newlink,
2345 .changelink = ip6erspan_changelink,
2346 .get_size = ip6gre_get_size,
2347 .fill_info = ip6gre_fill_info,
2348 .get_link_net = ip6_tnl_get_link_net,
2349 };
2350
2351 /*
2352 * And now the modules code and kernel interface.
2353 */
2354
2355 static int __init ip6gre_init(void)
2356 {
2357 int err;
2358
2359 pr_info("GRE over IPv6 tunneling driver\n");
2360
2361 err = register_pernet_device(&ip6gre_net_ops);
2362 if (err < 0)
2363 return err;
2364
2365 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2366 if (err < 0) {
2367 pr_info("%s: can't add protocol\n", __func__);
2368 goto add_proto_failed;
2369 }
2370
2371 err = rtnl_link_register(&ip6gre_link_ops);
2372 if (err < 0)
2373 goto rtnl_link_failed;
2374
2375 err = rtnl_link_register(&ip6gre_tap_ops);
2376 if (err < 0)
2377 goto tap_ops_failed;
2378
2379 err = rtnl_link_register(&ip6erspan_tap_ops);
2380 if (err < 0)
2381 goto erspan_link_failed;
2382
2383 out:
2384 return err;
2385
2386 erspan_link_failed:
2387 rtnl_link_unregister(&ip6gre_tap_ops);
2388 tap_ops_failed:
2389 rtnl_link_unregister(&ip6gre_link_ops);
2390 rtnl_link_failed:
2391 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2392 add_proto_failed:
2393 unregister_pernet_device(&ip6gre_net_ops);
2394 goto out;
2395 }
2396
2397 static void __exit ip6gre_fini(void)
2398 {
2399 rtnl_link_unregister(&ip6gre_tap_ops);
2400 rtnl_link_unregister(&ip6gre_link_ops);
2401 rtnl_link_unregister(&ip6erspan_tap_ops);
2402 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2403 unregister_pernet_device(&ip6gre_net_ops);
2404 }
2405
2406 module_init(ip6gre_init);
2407 module_exit(ip6gre_fini);
2408 MODULE_LICENSE("GPL");
2409 MODULE_AUTHOR("D. Kozlov <xeb@mail.ru>");
2410 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2411 MODULE_ALIAS_RTNL_LINK("ip6gre");
2412 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2413 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2414 MODULE_ALIAS_NETDEV("ip6gre0");
2415
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.