]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blob - net/ipv6/ip6_gre.c
projects / mirror_ubuntu-focal-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ip6_gre: update version related info when changing link
[mirror_ubuntu-focal-kernel.git] / net / ipv6 / ip6_gre.c
1 /*
2 * GRE over IPv6 protocol decoder.
3 *
4 * Authors: Dmitry Kozlov (xeb@mail.ru)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/init.h>
28 #include <linux/in6.h>
29 #include <linux/inetdevice.h>
30 #include <linux/igmp.h>
31 #include <linux/netfilter_ipv4.h>
32 #include <linux/etherdevice.h>
33 #include <linux/if_ether.h>
34 #include <linux/hash.h>
35 #include <linux/if_tunnel.h>
36 #include <linux/ip6_tunnel.h>
37
38 #include <net/sock.h>
39 #include <net/ip.h>
40 #include <net/ip_tunnels.h>
41 #include <net/icmp.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
44 #include <net/arp.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
48 #include <net/xfrm.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
52
53 #include <net/ipv6.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.h>
57 #include <net/gre.h>
58 #include <net/erspan.h>
59 #include <net/dst_metadata.h>
60
61
62 static bool log_ecn_error = true;
63 module_param(log_ecn_error, bool, 0644);
64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
65
66 #define IP6_GRE_HASH_SIZE_SHIFT 5
67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
68
69 static unsigned int ip6gre_net_id __read_mostly;
70 struct ip6gre_net {
71 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
72
73 struct ip6_tnl __rcu *collect_md_tun;
74 struct ip6_tnl __rcu *collect_md_tun_erspan;
75 struct net_device *fb_tunnel_dev;
76 };
77
78 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
80 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
81 static int ip6gre_tunnel_init(struct net_device *dev);
82 static void ip6gre_tunnel_setup(struct net_device *dev);
83 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
84 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
85 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
86
87 /* Tunnel hash table */
88
89 /*
90 4 hash tables:
91
92 3: (remote,local)
93 2: (remote,*)
94 1: (*,local)
95 0: (*,*)
96
97 We require exact key match i.e. if a key is present in packet
98 it will match only tunnel with the same key; if it is not present,
99 it will match only keyless tunnel.
100
101 All keysless packets, if not matched configured keyless tunnels
102 will match fallback tunnel.
103 */
104
105 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
106 static u32 HASH_ADDR(const struct in6_addr *addr)
107 {
108 u32 hash = ipv6_addr_hash(addr);
109
110 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
111 }
112
113 #define tunnels_r_l tunnels[3]
114 #define tunnels_r tunnels[2]
115 #define tunnels_l tunnels[1]
116 #define tunnels_wc tunnels[0]
117
118 /* Given src, dst and key, find appropriate for input tunnel. */
119
120 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
121 const struct in6_addr *remote, const struct in6_addr *local,
122 __be32 key, __be16 gre_proto)
123 {
124 struct net *net = dev_net(dev);
125 int link = dev->ifindex;
126 unsigned int h0 = HASH_ADDR(remote);
127 unsigned int h1 = HASH_KEY(key);
128 struct ip6_tnl *t, *cand = NULL;
129 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
130 int dev_type = (gre_proto == htons(ETH_P_TEB) ||
131 gre_proto == htons(ETH_P_ERSPAN) ||
132 gre_proto == htons(ETH_P_ERSPAN2)) ?
133 ARPHRD_ETHER : ARPHRD_IP6GRE;
134 int score, cand_score = 4;
135
136 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
137 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138 !ipv6_addr_equal(remote, &t->parms.raddr) ||
139 key != t->parms.i_key ||
140 !(t->dev->flags & IFF_UP))
141 continue;
142
143 if (t->dev->type != ARPHRD_IP6GRE &&
144 t->dev->type != dev_type)
145 continue;
146
147 score = 0;
148 if (t->parms.link != link)
149 score |= 1;
150 if (t->dev->type != dev_type)
151 score |= 2;
152 if (score == 0)
153 return t;
154
155 if (score < cand_score) {
156 cand = t;
157 cand_score = score;
158 }
159 }
160
161 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
162 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
163 key != t->parms.i_key ||
164 !(t->dev->flags & IFF_UP))
165 continue;
166
167 if (t->dev->type != ARPHRD_IP6GRE &&
168 t->dev->type != dev_type)
169 continue;
170
171 score = 0;
172 if (t->parms.link != link)
173 score |= 1;
174 if (t->dev->type != dev_type)
175 score |= 2;
176 if (score == 0)
177 return t;
178
179 if (score < cand_score) {
180 cand = t;
181 cand_score = score;
182 }
183 }
184
185 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
186 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
187 (!ipv6_addr_equal(local, &t->parms.raddr) ||
188 !ipv6_addr_is_multicast(local))) ||
189 key != t->parms.i_key ||
190 !(t->dev->flags & IFF_UP))
191 continue;
192
193 if (t->dev->type != ARPHRD_IP6GRE &&
194 t->dev->type != dev_type)
195 continue;
196
197 score = 0;
198 if (t->parms.link != link)
199 score |= 1;
200 if (t->dev->type != dev_type)
201 score |= 2;
202 if (score == 0)
203 return t;
204
205 if (score < cand_score) {
206 cand = t;
207 cand_score = score;
208 }
209 }
210
211 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
212 if (t->parms.i_key != key ||
213 !(t->dev->flags & IFF_UP))
214 continue;
215
216 if (t->dev->type != ARPHRD_IP6GRE &&
217 t->dev->type != dev_type)
218 continue;
219
220 score = 0;
221 if (t->parms.link != link)
222 score |= 1;
223 if (t->dev->type != dev_type)
224 score |= 2;
225 if (score == 0)
226 return t;
227
228 if (score < cand_score) {
229 cand = t;
230 cand_score = score;
231 }
232 }
233
234 if (cand)
235 return cand;
236
237 if (gre_proto == htons(ETH_P_ERSPAN) ||
238 gre_proto == htons(ETH_P_ERSPAN2))
239 t = rcu_dereference(ign->collect_md_tun_erspan);
240 else
241 t = rcu_dereference(ign->collect_md_tun);
242
243 if (t && t->dev->flags & IFF_UP)
244 return t;
245
246 dev = ign->fb_tunnel_dev;
247 if (dev && dev->flags & IFF_UP)
248 return netdev_priv(dev);
249
250 return NULL;
251 }
252
253 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
254 const struct __ip6_tnl_parm *p)
255 {
256 const struct in6_addr *remote = &p->raddr;
257 const struct in6_addr *local = &p->laddr;
258 unsigned int h = HASH_KEY(p->i_key);
259 int prio = 0;
260
261 if (!ipv6_addr_any(local))
262 prio |= 1;
263 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
264 prio |= 2;
265 h ^= HASH_ADDR(remote);
266 }
267
268 return &ign->tunnels[prio][h];
269 }
270
271 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
272 {
273 if (t->parms.collect_md)
274 rcu_assign_pointer(ign->collect_md_tun, t);
275 }
276
277 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
278 {
279 if (t->parms.collect_md)
280 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
281 }
282
283 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
284 {
285 if (t->parms.collect_md)
286 rcu_assign_pointer(ign->collect_md_tun, NULL);
287 }
288
289 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
290 struct ip6_tnl *t)
291 {
292 if (t->parms.collect_md)
293 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
294 }
295
296 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
297 const struct ip6_tnl *t)
298 {
299 return __ip6gre_bucket(ign, &t->parms);
300 }
301
302 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
303 {
304 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
305
306 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
307 rcu_assign_pointer(*tp, t);
308 }
309
310 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
311 {
312 struct ip6_tnl __rcu **tp;
313 struct ip6_tnl *iter;
314
315 for (tp = ip6gre_bucket(ign, t);
316 (iter = rtnl_dereference(*tp)) != NULL;
317 tp = &iter->next) {
318 if (t == iter) {
319 rcu_assign_pointer(*tp, t->next);
320 break;
321 }
322 }
323 }
324
325 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
326 const struct __ip6_tnl_parm *parms,
327 int type)
328 {
329 const struct in6_addr *remote = &parms->raddr;
330 const struct in6_addr *local = &parms->laddr;
331 __be32 key = parms->i_key;
332 int link = parms->link;
333 struct ip6_tnl *t;
334 struct ip6_tnl __rcu **tp;
335 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
336
337 for (tp = __ip6gre_bucket(ign, parms);
338 (t = rtnl_dereference(*tp)) != NULL;
339 tp = &t->next)
340 if (ipv6_addr_equal(local, &t->parms.laddr) &&
341 ipv6_addr_equal(remote, &t->parms.raddr) &&
342 key == t->parms.i_key &&
343 link == t->parms.link &&
344 type == t->dev->type)
345 break;
346
347 return t;
348 }
349
350 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
351 const struct __ip6_tnl_parm *parms, int create)
352 {
353 struct ip6_tnl *t, *nt;
354 struct net_device *dev;
355 char name[IFNAMSIZ];
356 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
357
358 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
359 if (t && create)
360 return NULL;
361 if (t || !create)
362 return t;
363
364 if (parms->name[0]) {
365 if (!dev_valid_name(parms->name))
366 return NULL;
367 strlcpy(name, parms->name, IFNAMSIZ);
368 } else {
369 strcpy(name, "ip6gre%d");
370 }
371 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
372 ip6gre_tunnel_setup);
373 if (!dev)
374 return NULL;
375
376 dev_net_set(dev, net);
377
378 nt = netdev_priv(dev);
379 nt->parms = *parms;
380 dev->rtnl_link_ops = &ip6gre_link_ops;
381
382 nt->dev = dev;
383 nt->net = dev_net(dev);
384
385 if (register_netdevice(dev) < 0)
386 goto failed_free;
387
388 ip6gre_tnl_link_config(nt, 1);
389
390 /* Can use a lockless transmit, unless we generate output sequences */
391 if (!(nt->parms.o_flags & TUNNEL_SEQ))
392 dev->features |= NETIF_F_LLTX;
393
394 dev_hold(dev);
395 ip6gre_tunnel_link(ign, nt);
396 return nt;
397
398 failed_free:
399 free_netdev(dev);
400 return NULL;
401 }
402
403 static void ip6erspan_tunnel_uninit(struct net_device *dev)
404 {
405 struct ip6_tnl *t = netdev_priv(dev);
406 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
407
408 ip6erspan_tunnel_unlink_md(ign, t);
409 ip6gre_tunnel_unlink(ign, t);
410 dst_cache_reset(&t->dst_cache);
411 dev_put(dev);
412 }
413
414 static void ip6gre_tunnel_uninit(struct net_device *dev)
415 {
416 struct ip6_tnl *t = netdev_priv(dev);
417 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
418
419 ip6gre_tunnel_unlink_md(ign, t);
420 ip6gre_tunnel_unlink(ign, t);
421 dst_cache_reset(&t->dst_cache);
422 dev_put(dev);
423 }
424
425
426 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
427 u8 type, u8 code, int offset, __be32 info)
428 {
429 struct net *net = dev_net(skb->dev);
430 const struct ipv6hdr *ipv6h;
431 struct tnl_ptk_info tpi;
432 struct ip6_tnl *t;
433
434 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
435 offset) < 0)
436 return -EINVAL;
437
438 ipv6h = (const struct ipv6hdr *)skb->data;
439 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
440 tpi.key, tpi.proto);
441 if (!t)
442 return -ENOENT;
443
444 switch (type) {
445 struct ipv6_tlv_tnl_enc_lim *tel;
446 __u32 teli;
447 case ICMPV6_DEST_UNREACH:
448 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
449 t->parms.name);
450 if (code != ICMPV6_PORT_UNREACH)
451 break;
452 return 0;
453 case ICMPV6_TIME_EXCEED:
454 if (code == ICMPV6_EXC_HOPLIMIT) {
455 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
456 t->parms.name);
457 break;
458 }
459 return 0;
460 case ICMPV6_PARAMPROB:
461 teli = 0;
462 if (code == ICMPV6_HDR_FIELD)
463 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
464
465 if (teli && teli == be32_to_cpu(info) - 2) {
466 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
467 if (tel->encap_limit == 0) {
468 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
469 t->parms.name);
470 }
471 } else {
472 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
473 t->parms.name);
474 }
475 return 0;
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, int gre_hdr_len,
528 struct tnl_ptk_info *tpi)
529 {
530 struct erspan_base_hdr *ershdr;
531 struct erspan_metadata *pkt_md;
532 const struct ipv6hdr *ipv6h;
533 struct erspan_md2 *md2;
534 struct ip6_tnl *tunnel;
535 u8 ver;
536
537 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
538 return PACKET_REJECT;
539
540 ipv6h = ipv6_hdr(skb);
541 ershdr = (struct erspan_base_hdr *)skb->data;
542 ver = ershdr->ver;
543 tpi->key = cpu_to_be32(get_session_id(ershdr));
544
545 tunnel = ip6gre_tunnel_lookup(skb->dev,
546 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
547 tpi->proto);
548 if (tunnel) {
549 int len = erspan_hdr_len(ver);
550
551 if (unlikely(!pskb_may_pull(skb, len)))
552 return PACKET_REJECT;
553
554 ershdr = (struct erspan_base_hdr *)skb->data;
555 pkt_md = (struct erspan_metadata *)(ershdr + 1);
556
557 if (__iptunnel_pull_header(skb, len,
558 htons(ETH_P_TEB),
559 false, false) < 0)
560 return PACKET_REJECT;
561
562 if (tunnel->parms.collect_md) {
563 struct metadata_dst *tun_dst;
564 struct ip_tunnel_info *info;
565 struct erspan_metadata *md;
566 __be64 tun_id;
567 __be16 flags;
568
569 tpi->flags |= TUNNEL_KEY;
570 flags = tpi->flags;
571 tun_id = key32_to_tunnel_id(tpi->key);
572
573 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
574 sizeof(*md));
575 if (!tun_dst)
576 return PACKET_REJECT;
577
578 info = &tun_dst->u.tun_info;
579 md = ip_tunnel_info_opts(info);
580 md->version = ver;
581 md2 = &md->u.md2;
582 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
583 ERSPAN_V2_MDSIZE);
584 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
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, hdr_len, &tpi) == 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 = ipv6_hdr(skb);
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
673 if (offset > 0) {
674 struct ipv6_tlv_tnl_enc_lim *tel;
675
676 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
677 if (tel->encap_limit == 0) {
678 icmpv6_send(skb, ICMPV6_PARAMPROB,
679 ICMPV6_HDR_FIELD, offset + 2);
680 return -1;
681 }
682 *encap_limit = tel->encap_limit - 1;
683 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
684 *encap_limit = t->parms.encap_limit;
685 }
686
687 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
688
689 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
690 *dsfield = ipv6_get_dsfield(ipv6h);
691 else
692 *dsfield = ip6_tclass(t->parms.flowinfo);
693
694 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
695 fl6->flowlabel |= ip6_flowlabel(ipv6h);
696
697 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
698 fl6->flowi6_mark = skb->mark;
699 else
700 fl6->flowi6_mark = t->parms.fwmark;
701
702 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
703
704 return 0;
705 }
706
707 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
708 struct net_device *dev, __u8 dsfield,
709 struct flowi6 *fl6, int encap_limit,
710 __u32 *pmtu, __be16 proto)
711 {
712 struct ip6_tnl *tunnel = netdev_priv(dev);
713 __be16 protocol;
714
715 if (dev->type == ARPHRD_ETHER)
716 IPCB(skb)->flags = 0;
717
718 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
719 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
720 else
721 fl6->daddr = tunnel->parms.raddr;
722
723 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
724 return -ENOMEM;
725
726 /* Push GRE header. */
727 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
728
729 if (tunnel->parms.collect_md) {
730 struct ip_tunnel_info *tun_info;
731 const struct ip_tunnel_key *key;
732 __be16 flags;
733
734 tun_info = skb_tunnel_info(skb);
735 if (unlikely(!tun_info ||
736 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
737 ip_tunnel_info_af(tun_info) != AF_INET6))
738 return -EINVAL;
739
740 key = &tun_info->key;
741 memset(fl6, 0, sizeof(*fl6));
742 fl6->flowi6_proto = IPPROTO_GRE;
743 fl6->daddr = key->u.ipv6.dst;
744 fl6->flowlabel = key->label;
745 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
746
747 dsfield = key->tos;
748 flags = key->tun_flags &
749 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
750 tunnel->tun_hlen = gre_calc_hlen(flags);
751
752 gre_build_header(skb, tunnel->tun_hlen,
753 flags, protocol,
754 tunnel_id_to_key32(tun_info->key.tun_id),
755 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
756 : 0);
757
758 } else {
759 if (tunnel->parms.o_flags & TUNNEL_SEQ)
760 tunnel->o_seqno++;
761
762 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
763 protocol, tunnel->parms.o_key,
764 htonl(tunnel->o_seqno));
765 }
766
767 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
768 NEXTHDR_GRE);
769 }
770
771 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
772 {
773 struct ip6_tnl *t = netdev_priv(dev);
774 int encap_limit = -1;
775 struct flowi6 fl6;
776 __u8 dsfield = 0;
777 __u32 mtu;
778 int err;
779
780 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
781
782 if (!t->parms.collect_md)
783 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
784 &dsfield, &encap_limit);
785
786 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
787 if (err)
788 return -1;
789
790 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
791 skb->protocol);
792 if (err != 0) {
793 /* XXX: send ICMP error even if DF is not set. */
794 if (err == -EMSGSIZE)
795 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
796 htonl(mtu));
797 return -1;
798 }
799
800 return 0;
801 }
802
803 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
804 {
805 struct ip6_tnl *t = netdev_priv(dev);
806 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
807 int encap_limit = -1;
808 struct flowi6 fl6;
809 __u8 dsfield = 0;
810 __u32 mtu;
811 int err;
812
813 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
814 return -1;
815
816 if (!t->parms.collect_md &&
817 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
818 return -1;
819
820 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
821 return -1;
822
823 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
824 &mtu, skb->protocol);
825 if (err != 0) {
826 if (err == -EMSGSIZE)
827 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
828 return -1;
829 }
830
831 return 0;
832 }
833
834 /**
835 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
836 * @t: the outgoing tunnel device
837 * @hdr: IPv6 header from the incoming packet
838 *
839 * Description:
840 * Avoid trivial tunneling loop by checking that tunnel exit-point
841 * doesn't match source of incoming packet.
842 *
843 * Return:
844 * 1 if conflict,
845 * 0 else
846 **/
847
848 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
849 const struct ipv6hdr *hdr)
850 {
851 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
852 }
853
854 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
855 {
856 struct ip6_tnl *t = netdev_priv(dev);
857 int encap_limit = -1;
858 struct flowi6 fl6;
859 __u32 mtu;
860 int err;
861
862 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
863 encap_limit = t->parms.encap_limit;
864
865 if (!t->parms.collect_md)
866 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
867
868 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
869 if (err)
870 return err;
871
872 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
873
874 return err;
875 }
876
877 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
878 struct net_device *dev)
879 {
880 struct ip6_tnl *t = netdev_priv(dev);
881 struct net_device_stats *stats = &t->dev->stats;
882 int ret;
883
884 if (!pskb_inet_may_pull(skb))
885 goto tx_err;
886
887 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
888 goto tx_err;
889
890 switch (skb->protocol) {
891 case htons(ETH_P_IP):
892 ret = ip6gre_xmit_ipv4(skb, dev);
893 break;
894 case htons(ETH_P_IPV6):
895 ret = ip6gre_xmit_ipv6(skb, dev);
896 break;
897 default:
898 ret = ip6gre_xmit_other(skb, dev);
899 break;
900 }
901
902 if (ret < 0)
903 goto tx_err;
904
905 return NETDEV_TX_OK;
906
907 tx_err:
908 stats->tx_errors++;
909 stats->tx_dropped++;
910 kfree_skb(skb);
911 return NETDEV_TX_OK;
912 }
913
914 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
915 struct net_device *dev)
916 {
917 struct ip6_tnl *t = netdev_priv(dev);
918 struct dst_entry *dst = skb_dst(skb);
919 struct net_device_stats *stats;
920 bool truncate = false;
921 int encap_limit = -1;
922 __u8 dsfield = false;
923 struct flowi6 fl6;
924 int err = -EINVAL;
925 __u32 mtu;
926 int nhoff;
927 int thoff;
928
929 if (!pskb_inet_may_pull(skb))
930 goto tx_err;
931
932 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
933 goto tx_err;
934
935 if (gre_handle_offloads(skb, false))
936 goto tx_err;
937
938 if (skb->len > dev->mtu + dev->hard_header_len) {
939 pskb_trim(skb, dev->mtu + dev->hard_header_len);
940 truncate = true;
941 }
942
943 nhoff = skb_network_header(skb) - skb_mac_header(skb);
944 if (skb->protocol == htons(ETH_P_IP) &&
945 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
946 truncate = true;
947
948 thoff = skb_transport_header(skb) - skb_mac_header(skb);
949 if (skb->protocol == htons(ETH_P_IPV6) &&
950 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
951 truncate = true;
952
953 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
954 goto tx_err;
955
956 t->parms.o_flags &= ~TUNNEL_KEY;
957 IPCB(skb)->flags = 0;
958
959 /* For collect_md mode, derive fl6 from the tunnel key,
960 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
961 */
962 if (t->parms.collect_md) {
963 struct ip_tunnel_info *tun_info;
964 const struct ip_tunnel_key *key;
965 struct erspan_metadata *md;
966 __be32 tun_id;
967
968 tun_info = skb_tunnel_info(skb);
969 if (unlikely(!tun_info ||
970 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
971 ip_tunnel_info_af(tun_info) != AF_INET6))
972 return -EINVAL;
973
974 key = &tun_info->key;
975 memset(&fl6, 0, sizeof(fl6));
976 fl6.flowi6_proto = IPPROTO_GRE;
977 fl6.daddr = key->u.ipv6.dst;
978 fl6.flowlabel = key->label;
979 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
980
981 dsfield = key->tos;
982 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
983 goto tx_err;
984 md = ip_tunnel_info_opts(tun_info);
985 if (!md)
986 goto tx_err;
987
988 tun_id = tunnel_id_to_key32(key->tun_id);
989 if (md->version == 1) {
990 erspan_build_header(skb,
991 ntohl(tun_id),
992 ntohl(md->u.index), truncate,
993 false);
994 } else if (md->version == 2) {
995 erspan_build_header_v2(skb,
996 ntohl(tun_id),
997 md->u.md2.dir,
998 get_hwid(&md->u.md2),
999 truncate, false);
1000 } else {
1001 goto tx_err;
1002 }
1003 } else {
1004 switch (skb->protocol) {
1005 case htons(ETH_P_IP):
1006 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1007 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1008 &dsfield, &encap_limit);
1009 break;
1010 case htons(ETH_P_IPV6):
1011 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1012 goto tx_err;
1013 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1014 &dsfield, &encap_limit))
1015 goto tx_err;
1016 break;
1017 default:
1018 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1019 break;
1020 }
1021
1022 if (t->parms.erspan_ver == 1)
1023 erspan_build_header(skb, ntohl(t->parms.o_key),
1024 t->parms.index,
1025 truncate, false);
1026 else if (t->parms.erspan_ver == 2)
1027 erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1028 t->parms.dir,
1029 t->parms.hwid,
1030 truncate, false);
1031 else
1032 goto tx_err;
1033
1034 fl6.daddr = t->parms.raddr;
1035 }
1036
1037 /* Push GRE header. */
1038 gre_build_header(skb, 8, TUNNEL_SEQ,
1039 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1040
1041 /* TooBig packet may have updated dst->dev's mtu */
1042 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1043 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1044
1045 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1046 NEXTHDR_GRE);
1047 if (err != 0) {
1048 /* XXX: send ICMP error even if DF is not set. */
1049 if (err == -EMSGSIZE) {
1050 if (skb->protocol == htons(ETH_P_IP))
1051 icmp_send(skb, ICMP_DEST_UNREACH,
1052 ICMP_FRAG_NEEDED, htonl(mtu));
1053 else
1054 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1055 }
1056
1057 goto tx_err;
1058 }
1059 return NETDEV_TX_OK;
1060
1061 tx_err:
1062 stats = &t->dev->stats;
1063 stats->tx_errors++;
1064 stats->tx_dropped++;
1065 kfree_skb(skb);
1066 return NETDEV_TX_OK;
1067 }
1068
1069 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1070 {
1071 struct net_device *dev = t->dev;
1072 struct __ip6_tnl_parm *p = &t->parms;
1073 struct flowi6 *fl6 = &t->fl.u.ip6;
1074
1075 if (dev->type != ARPHRD_ETHER) {
1076 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1077 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1078 }
1079
1080 /* Set up flowi template */
1081 fl6->saddr = p->laddr;
1082 fl6->daddr = p->raddr;
1083 fl6->flowi6_oif = p->link;
1084 fl6->flowlabel = 0;
1085 fl6->flowi6_proto = IPPROTO_GRE;
1086
1087 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1088 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1089 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1090 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1091
1092 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1093 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1094
1095 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1096 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1097 dev->flags |= IFF_POINTOPOINT;
1098 else
1099 dev->flags &= ~IFF_POINTOPOINT;
1100 }
1101
1102 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1103 int t_hlen)
1104 {
1105 const struct __ip6_tnl_parm *p = &t->parms;
1106 struct net_device *dev = t->dev;
1107
1108 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1109 int strict = (ipv6_addr_type(&p->raddr) &
1110 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1111
1112 struct rt6_info *rt = rt6_lookup(t->net,
1113 &p->raddr, &p->laddr,
1114 p->link, NULL, strict);
1115
1116 if (!rt)
1117 return;
1118
1119 if (rt->dst.dev) {
1120 dev->needed_headroom = rt->dst.dev->hard_header_len +
1121 t_hlen;
1122
1123 if (set_mtu) {
1124 dev->mtu = rt->dst.dev->mtu - t_hlen;
1125 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1126 dev->mtu -= 8;
1127 if (dev->type == ARPHRD_ETHER)
1128 dev->mtu -= ETH_HLEN;
1129
1130 if (dev->mtu < IPV6_MIN_MTU)
1131 dev->mtu = IPV6_MIN_MTU;
1132 }
1133 }
1134 ip6_rt_put(rt);
1135 }
1136 }
1137
1138 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1139 {
1140 int t_hlen;
1141
1142 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1143 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1144
1145 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1146 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1147 return t_hlen;
1148 }
1149
1150 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1151 {
1152 ip6gre_tnl_link_config_common(t);
1153 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1154 }
1155
1156 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1157 const struct __ip6_tnl_parm *p)
1158 {
1159 t->parms.laddr = p->laddr;
1160 t->parms.raddr = p->raddr;
1161 t->parms.flags = p->flags;
1162 t->parms.hop_limit = p->hop_limit;
1163 t->parms.encap_limit = p->encap_limit;
1164 t->parms.flowinfo = p->flowinfo;
1165 t->parms.link = p->link;
1166 t->parms.proto = p->proto;
1167 t->parms.i_key = p->i_key;
1168 t->parms.o_key = p->o_key;
1169 t->parms.i_flags = p->i_flags;
1170 t->parms.o_flags = p->o_flags;
1171 t->parms.fwmark = p->fwmark;
1172 t->parms.erspan_ver = p->erspan_ver;
1173 t->parms.index = p->index;
1174 t->parms.dir = p->dir;
1175 t->parms.hwid = p->hwid;
1176 dst_cache_reset(&t->dst_cache);
1177 }
1178
1179 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1180 int set_mtu)
1181 {
1182 ip6gre_tnl_copy_tnl_parm(t, p);
1183 ip6gre_tnl_link_config(t, set_mtu);
1184 return 0;
1185 }
1186
1187 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1188 const struct ip6_tnl_parm2 *u)
1189 {
1190 p->laddr = u->laddr;
1191 p->raddr = u->raddr;
1192 p->flags = u->flags;
1193 p->hop_limit = u->hop_limit;
1194 p->encap_limit = u->encap_limit;
1195 p->flowinfo = u->flowinfo;
1196 p->link = u->link;
1197 p->i_key = u->i_key;
1198 p->o_key = u->o_key;
1199 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1200 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1201 memcpy(p->name, u->name, sizeof(u->name));
1202 }
1203
1204 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1205 const struct __ip6_tnl_parm *p)
1206 {
1207 u->proto = IPPROTO_GRE;
1208 u->laddr = p->laddr;
1209 u->raddr = p->raddr;
1210 u->flags = p->flags;
1211 u->hop_limit = p->hop_limit;
1212 u->encap_limit = p->encap_limit;
1213 u->flowinfo = p->flowinfo;
1214 u->link = p->link;
1215 u->i_key = p->i_key;
1216 u->o_key = p->o_key;
1217 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1218 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1219 memcpy(u->name, p->name, sizeof(u->name));
1220 }
1221
1222 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1223 struct ifreq *ifr, int cmd)
1224 {
1225 int err = 0;
1226 struct ip6_tnl_parm2 p;
1227 struct __ip6_tnl_parm p1;
1228 struct ip6_tnl *t = netdev_priv(dev);
1229 struct net *net = t->net;
1230 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1231
1232 memset(&p1, 0, sizeof(p1));
1233
1234 switch (cmd) {
1235 case SIOCGETTUNNEL:
1236 if (dev == ign->fb_tunnel_dev) {
1237 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1238 err = -EFAULT;
1239 break;
1240 }
1241 ip6gre_tnl_parm_from_user(&p1, &p);
1242 t = ip6gre_tunnel_locate(net, &p1, 0);
1243 if (!t)
1244 t = netdev_priv(dev);
1245 }
1246 memset(&p, 0, sizeof(p));
1247 ip6gre_tnl_parm_to_user(&p, &t->parms);
1248 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1249 err = -EFAULT;
1250 break;
1251
1252 case SIOCADDTUNNEL:
1253 case SIOCCHGTUNNEL:
1254 err = -EPERM;
1255 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1256 goto done;
1257
1258 err = -EFAULT;
1259 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1260 goto done;
1261
1262 err = -EINVAL;
1263 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1264 goto done;
1265
1266 if (!(p.i_flags&GRE_KEY))
1267 p.i_key = 0;
1268 if (!(p.o_flags&GRE_KEY))
1269 p.o_key = 0;
1270
1271 ip6gre_tnl_parm_from_user(&p1, &p);
1272 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1273
1274 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1275 if (t) {
1276 if (t->dev != dev) {
1277 err = -EEXIST;
1278 break;
1279 }
1280 } else {
1281 t = netdev_priv(dev);
1282
1283 ip6gre_tunnel_unlink(ign, t);
1284 synchronize_net();
1285 ip6gre_tnl_change(t, &p1, 1);
1286 ip6gre_tunnel_link(ign, t);
1287 netdev_state_change(dev);
1288 }
1289 }
1290
1291 if (t) {
1292 err = 0;
1293
1294 memset(&p, 0, sizeof(p));
1295 ip6gre_tnl_parm_to_user(&p, &t->parms);
1296 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1297 err = -EFAULT;
1298 } else
1299 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1300 break;
1301
1302 case SIOCDELTUNNEL:
1303 err = -EPERM;
1304 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1305 goto done;
1306
1307 if (dev == ign->fb_tunnel_dev) {
1308 err = -EFAULT;
1309 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1310 goto done;
1311 err = -ENOENT;
1312 ip6gre_tnl_parm_from_user(&p1, &p);
1313 t = ip6gre_tunnel_locate(net, &p1, 0);
1314 if (!t)
1315 goto done;
1316 err = -EPERM;
1317 if (t == netdev_priv(ign->fb_tunnel_dev))
1318 goto done;
1319 dev = t->dev;
1320 }
1321 unregister_netdevice(dev);
1322 err = 0;
1323 break;
1324
1325 default:
1326 err = -EINVAL;
1327 }
1328
1329 done:
1330 return err;
1331 }
1332
1333 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1334 unsigned short type, const void *daddr,
1335 const void *saddr, unsigned int len)
1336 {
1337 struct ip6_tnl *t = netdev_priv(dev);
1338 struct ipv6hdr *ipv6h;
1339 __be16 *p;
1340
1341 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1342 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1343 t->fl.u.ip6.flowlabel,
1344 true, &t->fl.u.ip6));
1345 ipv6h->hop_limit = t->parms.hop_limit;
1346 ipv6h->nexthdr = NEXTHDR_GRE;
1347 ipv6h->saddr = t->parms.laddr;
1348 ipv6h->daddr = t->parms.raddr;
1349
1350 p = (__be16 *)(ipv6h + 1);
1351 p[0] = t->parms.o_flags;
1352 p[1] = htons(type);
1353
1354 /*
1355 * Set the source hardware address.
1356 */
1357
1358 if (saddr)
1359 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1360 if (daddr)
1361 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1362 if (!ipv6_addr_any(&ipv6h->daddr))
1363 return t->hlen;
1364
1365 return -t->hlen;
1366 }
1367
1368 static const struct header_ops ip6gre_header_ops = {
1369 .create = ip6gre_header,
1370 };
1371
1372 static const struct net_device_ops ip6gre_netdev_ops = {
1373 .ndo_init = ip6gre_tunnel_init,
1374 .ndo_uninit = ip6gre_tunnel_uninit,
1375 .ndo_start_xmit = ip6gre_tunnel_xmit,
1376 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1377 .ndo_change_mtu = ip6_tnl_change_mtu,
1378 .ndo_get_stats64 = ip_tunnel_get_stats64,
1379 .ndo_get_iflink = ip6_tnl_get_iflink,
1380 };
1381
1382 static void ip6gre_dev_free(struct net_device *dev)
1383 {
1384 struct ip6_tnl *t = netdev_priv(dev);
1385
1386 gro_cells_destroy(&t->gro_cells);
1387 dst_cache_destroy(&t->dst_cache);
1388 free_percpu(dev->tstats);
1389 }
1390
1391 static void ip6gre_tunnel_setup(struct net_device *dev)
1392 {
1393 dev->netdev_ops = &ip6gre_netdev_ops;
1394 dev->needs_free_netdev = true;
1395 dev->priv_destructor = ip6gre_dev_free;
1396
1397 dev->type = ARPHRD_IP6GRE;
1398
1399 dev->flags |= IFF_NOARP;
1400 dev->addr_len = sizeof(struct in6_addr);
1401 netif_keep_dst(dev);
1402 /* This perm addr will be used as interface identifier by IPv6 */
1403 dev->addr_assign_type = NET_ADDR_RANDOM;
1404 eth_random_addr(dev->perm_addr);
1405 }
1406
1407 #define GRE6_FEATURES (NETIF_F_SG | \
1408 NETIF_F_FRAGLIST | \
1409 NETIF_F_HIGHDMA | \
1410 NETIF_F_HW_CSUM)
1411
1412 static void ip6gre_tnl_init_features(struct net_device *dev)
1413 {
1414 struct ip6_tnl *nt = netdev_priv(dev);
1415
1416 dev->features |= GRE6_FEATURES;
1417 dev->hw_features |= GRE6_FEATURES;
1418
1419 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1420 /* TCP offload with GRE SEQ is not supported, nor
1421 * can we support 2 levels of outer headers requiring
1422 * an update.
1423 */
1424 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1425 nt->encap.type == TUNNEL_ENCAP_NONE) {
1426 dev->features |= NETIF_F_GSO_SOFTWARE;
1427 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1428 }
1429
1430 /* Can use a lockless transmit, unless we generate
1431 * output sequences
1432 */
1433 dev->features |= NETIF_F_LLTX;
1434 }
1435 }
1436
1437 static int ip6gre_tunnel_init_common(struct net_device *dev)
1438 {
1439 struct ip6_tnl *tunnel;
1440 int ret;
1441 int t_hlen;
1442
1443 tunnel = netdev_priv(dev);
1444
1445 tunnel->dev = dev;
1446 tunnel->net = dev_net(dev);
1447 strcpy(tunnel->parms.name, dev->name);
1448
1449 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1450 if (!dev->tstats)
1451 return -ENOMEM;
1452
1453 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1454 if (ret)
1455 goto cleanup_alloc_pcpu_stats;
1456
1457 ret = gro_cells_init(&tunnel->gro_cells, dev);
1458 if (ret)
1459 goto cleanup_dst_cache_init;
1460
1461 t_hlen = ip6gre_calc_hlen(tunnel);
1462 dev->mtu = ETH_DATA_LEN - t_hlen;
1463 if (dev->type == ARPHRD_ETHER)
1464 dev->mtu -= ETH_HLEN;
1465 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1466 dev->mtu -= 8;
1467
1468 if (tunnel->parms.collect_md) {
1469 dev->features |= NETIF_F_NETNS_LOCAL;
1470 netif_keep_dst(dev);
1471 }
1472 ip6gre_tnl_init_features(dev);
1473
1474 return 0;
1475
1476 cleanup_dst_cache_init:
1477 dst_cache_destroy(&tunnel->dst_cache);
1478 cleanup_alloc_pcpu_stats:
1479 free_percpu(dev->tstats);
1480 dev->tstats = NULL;
1481 return ret;
1482 }
1483
1484 static int ip6gre_tunnel_init(struct net_device *dev)
1485 {
1486 struct ip6_tnl *tunnel;
1487 int ret;
1488
1489 ret = ip6gre_tunnel_init_common(dev);
1490 if (ret)
1491 return ret;
1492
1493 tunnel = netdev_priv(dev);
1494
1495 if (tunnel->parms.collect_md)
1496 return 0;
1497
1498 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1499 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1500
1501 if (ipv6_addr_any(&tunnel->parms.raddr))
1502 dev->header_ops = &ip6gre_header_ops;
1503
1504 return 0;
1505 }
1506
1507 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1508 {
1509 struct ip6_tnl *tunnel = netdev_priv(dev);
1510
1511 tunnel->dev = dev;
1512 tunnel->net = dev_net(dev);
1513 strcpy(tunnel->parms.name, dev->name);
1514
1515 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1516
1517 dev_hold(dev);
1518 }
1519
1520 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1521 .handler = gre_rcv,
1522 .err_handler = ip6gre_err,
1523 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1524 };
1525
1526 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1527 {
1528 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1529 struct net_device *dev, *aux;
1530 int prio;
1531
1532 for_each_netdev_safe(net, dev, aux)
1533 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1534 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1535 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1536 unregister_netdevice_queue(dev, head);
1537
1538 for (prio = 0; prio < 4; prio++) {
1539 int h;
1540 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1541 struct ip6_tnl *t;
1542
1543 t = rtnl_dereference(ign->tunnels[prio][h]);
1544
1545 while (t) {
1546 /* If dev is in the same netns, it has already
1547 * been added to the list by the previous loop.
1548 */
1549 if (!net_eq(dev_net(t->dev), net))
1550 unregister_netdevice_queue(t->dev,
1551 head);
1552 t = rtnl_dereference(t->next);
1553 }
1554 }
1555 }
1556 }
1557
1558 static int __net_init ip6gre_init_net(struct net *net)
1559 {
1560 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1561 int err;
1562
1563 if (!net_has_fallback_tunnels(net))
1564 return 0;
1565 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1566 NET_NAME_UNKNOWN,
1567 ip6gre_tunnel_setup);
1568 if (!ign->fb_tunnel_dev) {
1569 err = -ENOMEM;
1570 goto err_alloc_dev;
1571 }
1572 dev_net_set(ign->fb_tunnel_dev, net);
1573 /* FB netdevice is special: we have one, and only one per netns.
1574 * Allowing to move it to another netns is clearly unsafe.
1575 */
1576 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1577
1578
1579 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1580 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1581
1582 err = register_netdev(ign->fb_tunnel_dev);
1583 if (err)
1584 goto err_reg_dev;
1585
1586 rcu_assign_pointer(ign->tunnels_wc[0],
1587 netdev_priv(ign->fb_tunnel_dev));
1588 return 0;
1589
1590 err_reg_dev:
1591 free_netdev(ign->fb_tunnel_dev);
1592 err_alloc_dev:
1593 return err;
1594 }
1595
1596 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1597 {
1598 struct net *net;
1599 LIST_HEAD(list);
1600
1601 rtnl_lock();
1602 list_for_each_entry(net, net_list, exit_list)
1603 ip6gre_destroy_tunnels(net, &list);
1604 unregister_netdevice_many(&list);
1605 rtnl_unlock();
1606 }
1607
1608 static struct pernet_operations ip6gre_net_ops = {
1609 .init = ip6gre_init_net,
1610 .exit_batch = ip6gre_exit_batch_net,
1611 .id = &ip6gre_net_id,
1612 .size = sizeof(struct ip6gre_net),
1613 };
1614
1615 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1616 struct netlink_ext_ack *extack)
1617 {
1618 __be16 flags;
1619
1620 if (!data)
1621 return 0;
1622
1623 flags = 0;
1624 if (data[IFLA_GRE_IFLAGS])
1625 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1626 if (data[IFLA_GRE_OFLAGS])
1627 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1628 if (flags & (GRE_VERSION|GRE_ROUTING))
1629 return -EINVAL;
1630
1631 return 0;
1632 }
1633
1634 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1635 struct netlink_ext_ack *extack)
1636 {
1637 struct in6_addr daddr;
1638
1639 if (tb[IFLA_ADDRESS]) {
1640 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1641 return -EINVAL;
1642 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1643 return -EADDRNOTAVAIL;
1644 }
1645
1646 if (!data)
1647 goto out;
1648
1649 if (data[IFLA_GRE_REMOTE]) {
1650 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1651 if (ipv6_addr_any(&daddr))
1652 return -EINVAL;
1653 }
1654
1655 out:
1656 return ip6gre_tunnel_validate(tb, data, extack);
1657 }
1658
1659 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1660 struct netlink_ext_ack *extack)
1661 {
1662 __be16 flags = 0;
1663 int ret, ver = 0;
1664
1665 if (!data)
1666 return 0;
1667
1668 ret = ip6gre_tap_validate(tb, data, extack);
1669 if (ret)
1670 return ret;
1671
1672 /* ERSPAN should only have GRE sequence and key flag */
1673 if (data[IFLA_GRE_OFLAGS])
1674 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1675 if (data[IFLA_GRE_IFLAGS])
1676 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1677 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1678 flags != (GRE_SEQ | GRE_KEY))
1679 return -EINVAL;
1680
1681 /* ERSPAN Session ID only has 10-bit. Since we reuse
1682 * 32-bit key field as ID, check it's range.
1683 */
1684 if (data[IFLA_GRE_IKEY] &&
1685 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1686 return -EINVAL;
1687
1688 if (data[IFLA_GRE_OKEY] &&
1689 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1690 return -EINVAL;
1691
1692 if (data[IFLA_GRE_ERSPAN_VER]) {
1693 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1694 if (ver != 1 && ver != 2)
1695 return -EINVAL;
1696 }
1697
1698 if (ver == 1) {
1699 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1700 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1701
1702 if (index & ~INDEX_MASK)
1703 return -EINVAL;
1704 }
1705 } else if (ver == 2) {
1706 if (data[IFLA_GRE_ERSPAN_DIR]) {
1707 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1708
1709 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1710 return -EINVAL;
1711 }
1712
1713 if (data[IFLA_GRE_ERSPAN_HWID]) {
1714 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1715
1716 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1717 return -EINVAL;
1718 }
1719 }
1720
1721 return 0;
1722 }
1723
1724 static void ip6gre_netlink_parms(struct nlattr *data[],
1725 struct __ip6_tnl_parm *parms)
1726 {
1727 memset(parms, 0, sizeof(*parms));
1728
1729 if (!data)
1730 return;
1731
1732 if (data[IFLA_GRE_LINK])
1733 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1734
1735 if (data[IFLA_GRE_IFLAGS])
1736 parms->i_flags = gre_flags_to_tnl_flags(
1737 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1738
1739 if (data[IFLA_GRE_OFLAGS])
1740 parms->o_flags = gre_flags_to_tnl_flags(
1741 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1742
1743 if (data[IFLA_GRE_IKEY])
1744 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1745
1746 if (data[IFLA_GRE_OKEY])
1747 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1748
1749 if (data[IFLA_GRE_LOCAL])
1750 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1751
1752 if (data[IFLA_GRE_REMOTE])
1753 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1754
1755 if (data[IFLA_GRE_TTL])
1756 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1757
1758 if (data[IFLA_GRE_ENCAP_LIMIT])
1759 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1760
1761 if (data[IFLA_GRE_FLOWINFO])
1762 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1763
1764 if (data[IFLA_GRE_FLAGS])
1765 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1766
1767 if (data[IFLA_GRE_FWMARK])
1768 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1769
1770 if (data[IFLA_GRE_COLLECT_METADATA])
1771 parms->collect_md = true;
1772
1773 parms->erspan_ver = 1;
1774 if (data[IFLA_GRE_ERSPAN_VER])
1775 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1776
1777 if (parms->erspan_ver == 1) {
1778 if (data[IFLA_GRE_ERSPAN_INDEX])
1779 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1780 } else if (parms->erspan_ver == 2) {
1781 if (data[IFLA_GRE_ERSPAN_DIR])
1782 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1783 if (data[IFLA_GRE_ERSPAN_HWID])
1784 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1785 }
1786 }
1787
1788 static int ip6gre_tap_init(struct net_device *dev)
1789 {
1790 int ret;
1791
1792 ret = ip6gre_tunnel_init_common(dev);
1793 if (ret)
1794 return ret;
1795
1796 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1797
1798 return 0;
1799 }
1800
1801 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1802 .ndo_init = ip6gre_tap_init,
1803 .ndo_uninit = ip6gre_tunnel_uninit,
1804 .ndo_start_xmit = ip6gre_tunnel_xmit,
1805 .ndo_set_mac_address = eth_mac_addr,
1806 .ndo_validate_addr = eth_validate_addr,
1807 .ndo_change_mtu = ip6_tnl_change_mtu,
1808 .ndo_get_stats64 = ip_tunnel_get_stats64,
1809 .ndo_get_iflink = ip6_tnl_get_iflink,
1810 };
1811
1812 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1813 {
1814 int t_hlen;
1815
1816 tunnel->tun_hlen = 8;
1817 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1818 erspan_hdr_len(tunnel->parms.erspan_ver);
1819
1820 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1821 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1822 return t_hlen;
1823 }
1824
1825 static int ip6erspan_tap_init(struct net_device *dev)
1826 {
1827 struct ip6_tnl *tunnel;
1828 int t_hlen;
1829 int ret;
1830
1831 tunnel = netdev_priv(dev);
1832
1833 tunnel->dev = dev;
1834 tunnel->net = dev_net(dev);
1835 strcpy(tunnel->parms.name, dev->name);
1836
1837 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1838 if (!dev->tstats)
1839 return -ENOMEM;
1840
1841 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1842 if (ret)
1843 goto cleanup_alloc_pcpu_stats;
1844
1845 ret = gro_cells_init(&tunnel->gro_cells, dev);
1846 if (ret)
1847 goto cleanup_dst_cache_init;
1848
1849 t_hlen = ip6erspan_calc_hlen(tunnel);
1850 dev->mtu = ETH_DATA_LEN - t_hlen;
1851 if (dev->type == ARPHRD_ETHER)
1852 dev->mtu -= ETH_HLEN;
1853 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1854 dev->mtu -= 8;
1855
1856 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1857 ip6erspan_tnl_link_config(tunnel, 1);
1858
1859 return 0;
1860
1861 cleanup_dst_cache_init:
1862 dst_cache_destroy(&tunnel->dst_cache);
1863 cleanup_alloc_pcpu_stats:
1864 free_percpu(dev->tstats);
1865 dev->tstats = NULL;
1866 return ret;
1867 }
1868
1869 static const struct net_device_ops ip6erspan_netdev_ops = {
1870 .ndo_init = ip6erspan_tap_init,
1871 .ndo_uninit = ip6erspan_tunnel_uninit,
1872 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1873 .ndo_set_mac_address = eth_mac_addr,
1874 .ndo_validate_addr = eth_validate_addr,
1875 .ndo_change_mtu = ip6_tnl_change_mtu,
1876 .ndo_get_stats64 = ip_tunnel_get_stats64,
1877 .ndo_get_iflink = ip6_tnl_get_iflink,
1878 };
1879
1880 static void ip6gre_tap_setup(struct net_device *dev)
1881 {
1882
1883 ether_setup(dev);
1884
1885 dev->max_mtu = 0;
1886 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1887 dev->needs_free_netdev = true;
1888 dev->priv_destructor = ip6gre_dev_free;
1889
1890 dev->features |= NETIF_F_NETNS_LOCAL;
1891 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1892 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1893 netif_keep_dst(dev);
1894 }
1895
1896 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1897 struct ip_tunnel_encap *ipencap)
1898 {
1899 bool ret = false;
1900
1901 memset(ipencap, 0, sizeof(*ipencap));
1902
1903 if (!data)
1904 return ret;
1905
1906 if (data[IFLA_GRE_ENCAP_TYPE]) {
1907 ret = true;
1908 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1909 }
1910
1911 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1912 ret = true;
1913 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1914 }
1915
1916 if (data[IFLA_GRE_ENCAP_SPORT]) {
1917 ret = true;
1918 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1919 }
1920
1921 if (data[IFLA_GRE_ENCAP_DPORT]) {
1922 ret = true;
1923 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1924 }
1925
1926 return ret;
1927 }
1928
1929 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1930 struct nlattr *tb[], struct nlattr *data[],
1931 struct netlink_ext_ack *extack)
1932 {
1933 struct ip6_tnl *nt;
1934 struct ip_tunnel_encap ipencap;
1935 int err;
1936
1937 nt = netdev_priv(dev);
1938
1939 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1940 int err = ip6_tnl_encap_setup(nt, &ipencap);
1941
1942 if (err < 0)
1943 return err;
1944 }
1945
1946 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1947 eth_hw_addr_random(dev);
1948
1949 nt->dev = dev;
1950 nt->net = dev_net(dev);
1951
1952 err = register_netdevice(dev);
1953 if (err)
1954 goto out;
1955
1956 if (tb[IFLA_MTU])
1957 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1958
1959 dev_hold(dev);
1960
1961 out:
1962 return err;
1963 }
1964
1965 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1966 struct nlattr *tb[], struct nlattr *data[],
1967 struct netlink_ext_ack *extack)
1968 {
1969 struct ip6_tnl *nt = netdev_priv(dev);
1970 struct net *net = dev_net(dev);
1971 struct ip6gre_net *ign;
1972 int err;
1973
1974 ip6gre_netlink_parms(data, &nt->parms);
1975 ign = net_generic(net, ip6gre_net_id);
1976
1977 if (nt->parms.collect_md) {
1978 if (rtnl_dereference(ign->collect_md_tun))
1979 return -EEXIST;
1980 } else {
1981 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1982 return -EEXIST;
1983 }
1984
1985 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1986 if (!err) {
1987 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1988 ip6gre_tunnel_link_md(ign, nt);
1989 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1990 }
1991 return err;
1992 }
1993
1994 static struct ip6_tnl *
1995 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
1996 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
1997 struct netlink_ext_ack *extack)
1998 {
1999 struct ip6_tnl *t, *nt = netdev_priv(dev);
2000 struct net *net = nt->net;
2001 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2002 struct ip_tunnel_encap ipencap;
2003
2004 if (dev == ign->fb_tunnel_dev)
2005 return ERR_PTR(-EINVAL);
2006
2007 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2008 int err = ip6_tnl_encap_setup(nt, &ipencap);
2009
2010 if (err < 0)
2011 return ERR_PTR(err);
2012 }
2013
2014 ip6gre_netlink_parms(data, p_p);
2015
2016 t = ip6gre_tunnel_locate(net, p_p, 0);
2017
2018 if (t) {
2019 if (t->dev != dev)
2020 return ERR_PTR(-EEXIST);
2021 } else {
2022 t = nt;
2023 }
2024
2025 return t;
2026 }
2027
2028 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2029 struct nlattr *data[],
2030 struct netlink_ext_ack *extack)
2031 {
2032 struct ip6_tnl *t = netdev_priv(dev);
2033 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2034 struct __ip6_tnl_parm p;
2035
2036 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2037 if (IS_ERR(t))
2038 return PTR_ERR(t);
2039
2040 ip6gre_tunnel_unlink_md(ign, t);
2041 ip6gre_tunnel_unlink(ign, t);
2042 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2043 ip6gre_tunnel_link_md(ign, t);
2044 ip6gre_tunnel_link(ign, t);
2045 return 0;
2046 }
2047
2048 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2049 {
2050 struct net *net = dev_net(dev);
2051 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2052
2053 if (dev != ign->fb_tunnel_dev)
2054 unregister_netdevice_queue(dev, head);
2055 }
2056
2057 static size_t ip6gre_get_size(const struct net_device *dev)
2058 {
2059 return
2060 /* IFLA_GRE_LINK */
2061 nla_total_size(4) +
2062 /* IFLA_GRE_IFLAGS */
2063 nla_total_size(2) +
2064 /* IFLA_GRE_OFLAGS */
2065 nla_total_size(2) +
2066 /* IFLA_GRE_IKEY */
2067 nla_total_size(4) +
2068 /* IFLA_GRE_OKEY */
2069 nla_total_size(4) +
2070 /* IFLA_GRE_LOCAL */
2071 nla_total_size(sizeof(struct in6_addr)) +
2072 /* IFLA_GRE_REMOTE */
2073 nla_total_size(sizeof(struct in6_addr)) +
2074 /* IFLA_GRE_TTL */
2075 nla_total_size(1) +
2076 /* IFLA_GRE_ENCAP_LIMIT */
2077 nla_total_size(1) +
2078 /* IFLA_GRE_FLOWINFO */
2079 nla_total_size(4) +
2080 /* IFLA_GRE_FLAGS */
2081 nla_total_size(4) +
2082 /* IFLA_GRE_ENCAP_TYPE */
2083 nla_total_size(2) +
2084 /* IFLA_GRE_ENCAP_FLAGS */
2085 nla_total_size(2) +
2086 /* IFLA_GRE_ENCAP_SPORT */
2087 nla_total_size(2) +
2088 /* IFLA_GRE_ENCAP_DPORT */
2089 nla_total_size(2) +
2090 /* IFLA_GRE_COLLECT_METADATA */
2091 nla_total_size(0) +
2092 /* IFLA_GRE_FWMARK */
2093 nla_total_size(4) +
2094 /* IFLA_GRE_ERSPAN_INDEX */
2095 nla_total_size(4) +
2096 0;
2097 }
2098
2099 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2100 {
2101 struct ip6_tnl *t = netdev_priv(dev);
2102 struct __ip6_tnl_parm *p = &t->parms;
2103
2104 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2105 nla_put_be16(skb, IFLA_GRE_IFLAGS,
2106 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2107 nla_put_be16(skb, IFLA_GRE_OFLAGS,
2108 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2109 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2110 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2111 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2112 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2113 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2114 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2115 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2116 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2117 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2118 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2119 goto nla_put_failure;
2120
2121 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2122 t->encap.type) ||
2123 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2124 t->encap.sport) ||
2125 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2126 t->encap.dport) ||
2127 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2128 t->encap.flags))
2129 goto nla_put_failure;
2130
2131 if (p->collect_md) {
2132 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2133 goto nla_put_failure;
2134 }
2135
2136 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2137 goto nla_put_failure;
2138
2139 if (p->erspan_ver == 1) {
2140 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2141 goto nla_put_failure;
2142 } else if (p->erspan_ver == 2) {
2143 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2144 goto nla_put_failure;
2145 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2146 goto nla_put_failure;
2147 }
2148
2149 return 0;
2150
2151 nla_put_failure:
2152 return -EMSGSIZE;
2153 }
2154
2155 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2156 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2157 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2158 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2159 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2160 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2161 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2162 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2163 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2164 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2165 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2166 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2167 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2168 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2169 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2170 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2171 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2172 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2173 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2174 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2175 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2176 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2177 };
2178
2179 static void ip6erspan_tap_setup(struct net_device *dev)
2180 {
2181 ether_setup(dev);
2182
2183 dev->netdev_ops = &ip6erspan_netdev_ops;
2184 dev->needs_free_netdev = true;
2185 dev->priv_destructor = ip6gre_dev_free;
2186
2187 dev->features |= NETIF_F_NETNS_LOCAL;
2188 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2189 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2190 netif_keep_dst(dev);
2191 }
2192
2193 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2194 struct nlattr *tb[], struct nlattr *data[],
2195 struct netlink_ext_ack *extack)
2196 {
2197 struct ip6_tnl *nt = netdev_priv(dev);
2198 struct net *net = dev_net(dev);
2199 struct ip6gre_net *ign;
2200 int err;
2201
2202 ip6gre_netlink_parms(data, &nt->parms);
2203 ign = net_generic(net, ip6gre_net_id);
2204
2205 if (nt->parms.collect_md) {
2206 if (rtnl_dereference(ign->collect_md_tun_erspan))
2207 return -EEXIST;
2208 } else {
2209 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2210 return -EEXIST;
2211 }
2212
2213 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2214 if (!err) {
2215 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2216 ip6erspan_tunnel_link_md(ign, nt);
2217 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2218 }
2219 return err;
2220 }
2221
2222 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2223 {
2224 ip6gre_tnl_link_config_common(t);
2225 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2226 }
2227
2228 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2229 const struct __ip6_tnl_parm *p, int set_mtu)
2230 {
2231 ip6gre_tnl_copy_tnl_parm(t, p);
2232 ip6erspan_tnl_link_config(t, set_mtu);
2233 return 0;
2234 }
2235
2236 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2237 struct nlattr *data[],
2238 struct netlink_ext_ack *extack)
2239 {
2240 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2241 struct __ip6_tnl_parm p;
2242 struct ip6_tnl *t;
2243
2244 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2245 if (IS_ERR(t))
2246 return PTR_ERR(t);
2247
2248 ip6gre_tunnel_unlink_md(ign, t);
2249 ip6gre_tunnel_unlink(ign, t);
2250 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2251 ip6erspan_tunnel_link_md(ign, t);
2252 ip6gre_tunnel_link(ign, t);
2253 return 0;
2254 }
2255
2256 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2257 .kind = "ip6gre",
2258 .maxtype = IFLA_GRE_MAX,
2259 .policy = ip6gre_policy,
2260 .priv_size = sizeof(struct ip6_tnl),
2261 .setup = ip6gre_tunnel_setup,
2262 .validate = ip6gre_tunnel_validate,
2263 .newlink = ip6gre_newlink,
2264 .changelink = ip6gre_changelink,
2265 .dellink = ip6gre_dellink,
2266 .get_size = ip6gre_get_size,
2267 .fill_info = ip6gre_fill_info,
2268 .get_link_net = ip6_tnl_get_link_net,
2269 };
2270
2271 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2272 .kind = "ip6gretap",
2273 .maxtype = IFLA_GRE_MAX,
2274 .policy = ip6gre_policy,
2275 .priv_size = sizeof(struct ip6_tnl),
2276 .setup = ip6gre_tap_setup,
2277 .validate = ip6gre_tap_validate,
2278 .newlink = ip6gre_newlink,
2279 .changelink = ip6gre_changelink,
2280 .get_size = ip6gre_get_size,
2281 .fill_info = ip6gre_fill_info,
2282 .get_link_net = ip6_tnl_get_link_net,
2283 };
2284
2285 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2286 .kind = "ip6erspan",
2287 .maxtype = IFLA_GRE_MAX,
2288 .policy = ip6gre_policy,
2289 .priv_size = sizeof(struct ip6_tnl),
2290 .setup = ip6erspan_tap_setup,
2291 .validate = ip6erspan_tap_validate,
2292 .newlink = ip6erspan_newlink,
2293 .changelink = ip6erspan_changelink,
2294 .get_size = ip6gre_get_size,
2295 .fill_info = ip6gre_fill_info,
2296 .get_link_net = ip6_tnl_get_link_net,
2297 };
2298
2299 /*
2300 * And now the modules code and kernel interface.
2301 */
2302
2303 static int __init ip6gre_init(void)
2304 {
2305 int err;
2306
2307 pr_info("GRE over IPv6 tunneling driver\n");
2308
2309 err = register_pernet_device(&ip6gre_net_ops);
2310 if (err < 0)
2311 return err;
2312
2313 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2314 if (err < 0) {
2315 pr_info("%s: can't add protocol\n", __func__);
2316 goto add_proto_failed;
2317 }
2318
2319 err = rtnl_link_register(&ip6gre_link_ops);
2320 if (err < 0)
2321 goto rtnl_link_failed;
2322
2323 err = rtnl_link_register(&ip6gre_tap_ops);
2324 if (err < 0)
2325 goto tap_ops_failed;
2326
2327 err = rtnl_link_register(&ip6erspan_tap_ops);
2328 if (err < 0)
2329 goto erspan_link_failed;
2330
2331 out:
2332 return err;
2333
2334 erspan_link_failed:
2335 rtnl_link_unregister(&ip6gre_tap_ops);
2336 tap_ops_failed:
2337 rtnl_link_unregister(&ip6gre_link_ops);
2338 rtnl_link_failed:
2339 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2340 add_proto_failed:
2341 unregister_pernet_device(&ip6gre_net_ops);
2342 goto out;
2343 }
2344
2345 static void __exit ip6gre_fini(void)
2346 {
2347 rtnl_link_unregister(&ip6gre_tap_ops);
2348 rtnl_link_unregister(&ip6gre_link_ops);
2349 rtnl_link_unregister(&ip6erspan_tap_ops);
2350 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2351 unregister_pernet_device(&ip6gre_net_ops);
2352 }
2353
2354 module_init(ip6gre_init);
2355 module_exit(ip6gre_fini);
2356 MODULE_LICENSE("GPL");
2357 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2358 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2359 MODULE_ALIAS_RTNL_LINK("ip6gre");
2360 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2361 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2362 MODULE_ALIAS_NETDEV("ip6gre0");
Ubuntu Focal Linux kernel mirror