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
59
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63
64 #define IP6_GRE_HASH_SIZE_SHIFT 5
65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
66
67 static unsigned int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
70
71 struct net_device *fb_tunnel_dev;
72 };
73
74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75 static struct rtnl_link_ops ip6gre_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
81 /* Tunnel hash table */
82
83 /*
84 4 hash tables:
85
86 3: (remote,local)
87 2: (remote,*)
88 1: (*,local)
89 0: (*,*)
90
91 We require exact key match i.e. if a key is present in packet
92 it will match only tunnel with the same key; if it is not present,
93 it will match only keyless tunnel.
94
95 All keysless packets, if not matched configured keyless tunnels
96 will match fallback tunnel.
97 */
98
99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
100 static u32 HASH_ADDR(const struct in6_addr *addr)
101 {
102 u32 hash = ipv6_addr_hash(addr);
103
104 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
105 }
106
107 #define tunnels_r_l tunnels[3]
108 #define tunnels_r tunnels[2]
109 #define tunnels_l tunnels[1]
110 #define tunnels_wc tunnels[0]
111
112 /* Given src, dst and key, find appropriate for input tunnel. */
113
114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
115 const struct in6_addr *remote, const struct in6_addr *local,
116 __be32 key, __be16 gre_proto)
117 {
118 struct net *net = dev_net(dev);
119 int link = dev->ifindex;
120 unsigned int h0 = HASH_ADDR(remote);
121 unsigned int h1 = HASH_KEY(key);
122 struct ip6_tnl *t, *cand = NULL;
123 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
124 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
125 ARPHRD_ETHER : ARPHRD_IP6GRE;
126 int score, cand_score = 4;
127
128 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
129 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
130 !ipv6_addr_equal(remote, &t->parms.raddr) ||
131 key != t->parms.i_key ||
132 !(t->dev->flags & IFF_UP))
133 continue;
134
135 if (t->dev->type != ARPHRD_IP6GRE &&
136 t->dev->type != dev_type)
137 continue;
138
139 score = 0;
140 if (t->parms.link != link)
141 score |= 1;
142 if (t->dev->type != dev_type)
143 score |= 2;
144 if (score == 0)
145 return t;
146
147 if (score < cand_score) {
148 cand = t;
149 cand_score = score;
150 }
151 }
152
153 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
154 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
155 key != t->parms.i_key ||
156 !(t->dev->flags & IFF_UP))
157 continue;
158
159 if (t->dev->type != ARPHRD_IP6GRE &&
160 t->dev->type != dev_type)
161 continue;
162
163 score = 0;
164 if (t->parms.link != link)
165 score |= 1;
166 if (t->dev->type != dev_type)
167 score |= 2;
168 if (score == 0)
169 return t;
170
171 if (score < cand_score) {
172 cand = t;
173 cand_score = score;
174 }
175 }
176
177 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
178 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
179 (!ipv6_addr_equal(local, &t->parms.raddr) ||
180 !ipv6_addr_is_multicast(local))) ||
181 key != t->parms.i_key ||
182 !(t->dev->flags & IFF_UP))
183 continue;
184
185 if (t->dev->type != ARPHRD_IP6GRE &&
186 t->dev->type != dev_type)
187 continue;
188
189 score = 0;
190 if (t->parms.link != link)
191 score |= 1;
192 if (t->dev->type != dev_type)
193 score |= 2;
194 if (score == 0)
195 return t;
196
197 if (score < cand_score) {
198 cand = t;
199 cand_score = score;
200 }
201 }
202
203 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
204 if (t->parms.i_key != key ||
205 !(t->dev->flags & IFF_UP))
206 continue;
207
208 if (t->dev->type != ARPHRD_IP6GRE &&
209 t->dev->type != dev_type)
210 continue;
211
212 score = 0;
213 if (t->parms.link != link)
214 score |= 1;
215 if (t->dev->type != dev_type)
216 score |= 2;
217 if (score == 0)
218 return t;
219
220 if (score < cand_score) {
221 cand = t;
222 cand_score = score;
223 }
224 }
225
226 if (cand)
227 return cand;
228
229 dev = ign->fb_tunnel_dev;
230 if (dev->flags & IFF_UP)
231 return netdev_priv(dev);
232
233 return NULL;
234 }
235
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237 const struct __ip6_tnl_parm *p)
238 {
239 const struct in6_addr *remote = &p->raddr;
240 const struct in6_addr *local = &p->laddr;
241 unsigned int h = HASH_KEY(p->i_key);
242 int prio = 0;
243
244 if (!ipv6_addr_any(local))
245 prio |= 1;
246 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
247 prio |= 2;
248 h ^= HASH_ADDR(remote);
249 }
250
251 return &ign->tunnels[prio][h];
252 }
253
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255 const struct ip6_tnl *t)
256 {
257 return __ip6gre_bucket(ign, &t->parms);
258 }
259
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
261 {
262 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
263
264 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265 rcu_assign_pointer(*tp, t);
266 }
267
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
269 {
270 struct ip6_tnl __rcu **tp;
271 struct ip6_tnl *iter;
272
273 for (tp = ip6gre_bucket(ign, t);
274 (iter = rtnl_dereference(*tp)) != NULL;
275 tp = &iter->next) {
276 if (t == iter) {
277 rcu_assign_pointer(*tp, t->next);
278 break;
279 }
280 }
281 }
282
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 const struct __ip6_tnl_parm *parms,
285 int type)
286 {
287 const struct in6_addr *remote = &parms->raddr;
288 const struct in6_addr *local = &parms->laddr;
289 __be32 key = parms->i_key;
290 int link = parms->link;
291 struct ip6_tnl *t;
292 struct ip6_tnl __rcu **tp;
293 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
294
295 for (tp = __ip6gre_bucket(ign, parms);
296 (t = rtnl_dereference(*tp)) != NULL;
297 tp = &t->next)
298 if (ipv6_addr_equal(local, &t->parms.laddr) &&
299 ipv6_addr_equal(remote, &t->parms.raddr) &&
300 key == t->parms.i_key &&
301 link == t->parms.link &&
302 type == t->dev->type)
303 break;
304
305 return t;
306 }
307
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309 const struct __ip6_tnl_parm *parms, int create)
310 {
311 struct ip6_tnl *t, *nt;
312 struct net_device *dev;
313 char name[IFNAMSIZ];
314 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315
316 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317 if (t && create)
318 return NULL;
319 if (t || !create)
320 return t;
321
322 if (parms->name[0]) {
323 if (!dev_valid_name(parms->name))
324 return NULL;
325 strlcpy(name, parms->name, IFNAMSIZ);
326 } else {
327 strcpy(name, "ip6gre%d");
328 }
329 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
330 ip6gre_tunnel_setup);
331 if (!dev)
332 return NULL;
333
334 dev_net_set(dev, net);
335
336 nt = netdev_priv(dev);
337 nt->parms = *parms;
338 dev->rtnl_link_ops = &ip6gre_link_ops;
339
340 nt->dev = dev;
341 nt->net = dev_net(dev);
342
343 if (register_netdevice(dev) < 0)
344 goto failed_free;
345
346 ip6gre_tnl_link_config(nt, 1);
347
348 /* Can use a lockless transmit, unless we generate output sequences */
349 if (!(nt->parms.o_flags & TUNNEL_SEQ))
350 dev->features |= NETIF_F_LLTX;
351
352 dev_hold(dev);
353 ip6gre_tunnel_link(ign, nt);
354 return nt;
355
356 failed_free:
357 free_netdev(dev);
358 return NULL;
359 }
360
361 static void ip6gre_tunnel_uninit(struct net_device *dev)
362 {
363 struct ip6_tnl *t = netdev_priv(dev);
364 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
365
366 ip6gre_tunnel_unlink(ign, t);
367 dst_cache_reset(&t->dst_cache);
368 dev_put(dev);
369 }
370
371
372 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
373 u8 type, u8 code, int offset, __be32 info)
374 {
375 struct net *net = dev_net(skb->dev);
376 const struct gre_base_hdr *greh;
377 const struct ipv6hdr *ipv6h;
378 int grehlen = sizeof(*greh);
379 struct ip6_tnl *t;
380 int key_off = 0;
381 __be16 flags;
382 __be32 key;
383
384 if (!pskb_may_pull(skb, offset + grehlen))
385 return;
386 greh = (const struct gre_base_hdr *)(skb->data + offset);
387 flags = greh->flags;
388 if (flags & (GRE_VERSION | GRE_ROUTING))
389 return;
390 if (flags & GRE_CSUM)
391 grehlen += 4;
392 if (flags & GRE_KEY) {
393 key_off = grehlen + offset;
394 grehlen += 4;
395 }
396
397 if (!pskb_may_pull(skb, offset + grehlen))
398 return;
399 ipv6h = (const struct ipv6hdr *)skb->data;
400 greh = (const struct gre_base_hdr *)(skb->data + offset);
401 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
402
403 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
404 key, greh->protocol);
405 if (!t)
406 return;
407
408 switch (type) {
409 struct ipv6_tlv_tnl_enc_lim *tel;
410 __u32 teli;
411 case ICMPV6_DEST_UNREACH:
412 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
413 t->parms.name);
414 if (code != ICMPV6_PORT_UNREACH)
415 break;
416 return;
417 case ICMPV6_TIME_EXCEED:
418 if (code == ICMPV6_EXC_HOPLIMIT) {
419 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
420 t->parms.name);
421 break;
422 }
423 return;
424 case ICMPV6_PARAMPROB:
425 teli = 0;
426 if (code == ICMPV6_HDR_FIELD)
427 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
428
429 if (teli && teli == be32_to_cpu(info) - 2) {
430 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
431 if (tel->encap_limit == 0) {
432 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
433 t->parms.name);
434 }
435 } else {
436 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
437 t->parms.name);
438 }
439 return;
440 case ICMPV6_PKT_TOOBIG:
441 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
442 return;
443 case NDISC_REDIRECT:
444 ip6_redirect(skb, net, skb->dev->ifindex, 0,
445 sock_net_uid(net, NULL));
446 return;
447 }
448
449 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
450 t->err_count++;
451 else
452 t->err_count = 1;
453 t->err_time = jiffies;
454 }
455
456 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
457 {
458 const struct ipv6hdr *ipv6h;
459 struct ip6_tnl *tunnel;
460
461 ipv6h = ipv6_hdr(skb);
462 tunnel = ip6gre_tunnel_lookup(skb->dev,
463 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
464 tpi->proto);
465 if (tunnel) {
466 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
467
468 return PACKET_RCVD;
469 }
470
471 return PACKET_REJECT;
472 }
473
474 static int gre_rcv(struct sk_buff *skb)
475 {
476 struct tnl_ptk_info tpi;
477 bool csum_err = false;
478 int hdr_len;
479
480 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
481 if (hdr_len < 0)
482 goto drop;
483
484 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
485 goto drop;
486
487 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
488 return 0;
489
490 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
491 drop:
492 kfree_skb(skb);
493 return 0;
494 }
495
496 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
497 {
498 return iptunnel_handle_offloads(skb,
499 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
500 }
501
502 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
503 struct net_device *dev, __u8 dsfield,
504 struct flowi6 *fl6, int encap_limit,
505 __u32 *pmtu, __be16 proto)
506 {
507 struct ip6_tnl *tunnel = netdev_priv(dev);
508 __be16 protocol;
509
510 if (dev->type == ARPHRD_ETHER)
511 IPCB(skb)->flags = 0;
512
513 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
514 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
515 else
516 fl6->daddr = tunnel->parms.raddr;
517
518 if (tunnel->parms.o_flags & TUNNEL_SEQ)
519 tunnel->o_seqno++;
520
521 /* Push GRE header. */
522 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
523 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
524 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
525
526 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
527 NEXTHDR_GRE);
528 }
529
530 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
531 {
532 struct ip6_tnl *t = netdev_priv(dev);
533 const struct iphdr *iph = ip_hdr(skb);
534 int encap_limit = -1;
535 struct flowi6 fl6;
536 __u8 dsfield;
537 __u32 mtu;
538 int err;
539
540 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
541
542 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
543 encap_limit = t->parms.encap_limit;
544
545 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
546
547 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
548 dsfield = ipv4_get_dsfield(iph);
549 else
550 dsfield = ip6_tclass(t->parms.flowinfo);
551 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
552 fl6.flowi6_mark = skb->mark;
553 else
554 fl6.flowi6_mark = t->parms.fwmark;
555
556 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
557
558 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
559 if (err)
560 return -1;
561
562 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
563 skb->protocol);
564 if (err != 0) {
565 /* XXX: send ICMP error even if DF is not set. */
566 if (err == -EMSGSIZE)
567 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
568 htonl(mtu));
569 return -1;
570 }
571
572 return 0;
573 }
574
575 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
576 {
577 struct ip6_tnl *t = netdev_priv(dev);
578 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
579 int encap_limit = -1;
580 __u16 offset;
581 struct flowi6 fl6;
582 __u8 dsfield;
583 __u32 mtu;
584 int err;
585
586 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
587 return -1;
588
589 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
590 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
591 ipv6h = ipv6_hdr(skb);
592
593 if (offset > 0) {
594 struct ipv6_tlv_tnl_enc_lim *tel;
595 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
596 if (tel->encap_limit == 0) {
597 icmpv6_send(skb, ICMPV6_PARAMPROB,
598 ICMPV6_HDR_FIELD, offset + 2);
599 return -1;
600 }
601 encap_limit = tel->encap_limit - 1;
602 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
603 encap_limit = t->parms.encap_limit;
604
605 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
606
607 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
608 dsfield = ipv6_get_dsfield(ipv6h);
609 else
610 dsfield = ip6_tclass(t->parms.flowinfo);
611
612 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
613 fl6.flowlabel |= ip6_flowlabel(ipv6h);
614 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
615 fl6.flowi6_mark = skb->mark;
616 else
617 fl6.flowi6_mark = t->parms.fwmark;
618
619 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
620
621 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
622 return -1;
623
624 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
625 &mtu, skb->protocol);
626 if (err != 0) {
627 if (err == -EMSGSIZE)
628 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
629 return -1;
630 }
631
632 return 0;
633 }
634
635 /**
636 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
637 * @t: the outgoing tunnel device
638 * @hdr: IPv6 header from the incoming packet
639 *
640 * Description:
641 * Avoid trivial tunneling loop by checking that tunnel exit-point
642 * doesn't match source of incoming packet.
643 *
644 * Return:
645 * 1 if conflict,
646 * 0 else
647 **/
648
649 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
650 const struct ipv6hdr *hdr)
651 {
652 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
653 }
654
655 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
656 {
657 struct ip6_tnl *t = netdev_priv(dev);
658 int encap_limit = -1;
659 struct flowi6 fl6;
660 __u32 mtu;
661 int err;
662
663 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
664 encap_limit = t->parms.encap_limit;
665
666 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
667
668 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
669 if (err)
670 return err;
671
672 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
673
674 return err;
675 }
676
677 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
678 struct net_device *dev)
679 {
680 struct ip6_tnl *t = netdev_priv(dev);
681 struct net_device_stats *stats = &t->dev->stats;
682 int ret;
683
684 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
685 goto tx_err;
686
687 switch (skb->protocol) {
688 case htons(ETH_P_IP):
689 ret = ip6gre_xmit_ipv4(skb, dev);
690 break;
691 case htons(ETH_P_IPV6):
692 ret = ip6gre_xmit_ipv6(skb, dev);
693 break;
694 default:
695 ret = ip6gre_xmit_other(skb, dev);
696 break;
697 }
698
699 if (ret < 0)
700 goto tx_err;
701
702 return NETDEV_TX_OK;
703
704 tx_err:
705 stats->tx_errors++;
706 stats->tx_dropped++;
707 kfree_skb(skb);
708 return NETDEV_TX_OK;
709 }
710
711 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
712 {
713 struct net_device *dev = t->dev;
714 struct __ip6_tnl_parm *p = &t->parms;
715 struct flowi6 *fl6 = &t->fl.u.ip6;
716 int t_hlen;
717
718 if (dev->type != ARPHRD_ETHER) {
719 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
720 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
721 }
722
723 /* Set up flowi template */
724 fl6->saddr = p->laddr;
725 fl6->daddr = p->raddr;
726 fl6->flowi6_oif = p->link;
727 fl6->flowlabel = 0;
728 fl6->flowi6_proto = IPPROTO_GRE;
729
730 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
731 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
732 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
733 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
734
735 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
736 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
737
738 if (p->flags&IP6_TNL_F_CAP_XMIT &&
739 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
740 dev->flags |= IFF_POINTOPOINT;
741 else
742 dev->flags &= ~IFF_POINTOPOINT;
743
744 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
745
746 t->hlen = t->encap_hlen + t->tun_hlen;
747
748 t_hlen = t->hlen + sizeof(struct ipv6hdr);
749
750 if (p->flags & IP6_TNL_F_CAP_XMIT) {
751 int strict = (ipv6_addr_type(&p->raddr) &
752 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
753
754 struct rt6_info *rt = rt6_lookup(t->net,
755 &p->raddr, &p->laddr,
756 p->link, strict);
757
758 if (!rt)
759 return;
760
761 if (rt->dst.dev) {
762 dev->hard_header_len = rt->dst.dev->hard_header_len +
763 t_hlen;
764
765 if (set_mtu) {
766 dev->mtu = rt->dst.dev->mtu - t_hlen;
767 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
768 dev->mtu -= 8;
769 if (dev->type == ARPHRD_ETHER)
770 dev->mtu -= ETH_HLEN;
771
772 if (dev->mtu < IPV6_MIN_MTU)
773 dev->mtu = IPV6_MIN_MTU;
774 }
775 }
776 ip6_rt_put(rt);
777 }
778 }
779
780 static int ip6gre_tnl_change(struct ip6_tnl *t,
781 const struct __ip6_tnl_parm *p, int set_mtu)
782 {
783 t->parms.laddr = p->laddr;
784 t->parms.raddr = p->raddr;
785 t->parms.flags = p->flags;
786 t->parms.hop_limit = p->hop_limit;
787 t->parms.encap_limit = p->encap_limit;
788 t->parms.flowinfo = p->flowinfo;
789 t->parms.link = p->link;
790 t->parms.proto = p->proto;
791 t->parms.i_key = p->i_key;
792 t->parms.o_key = p->o_key;
793 t->parms.i_flags = p->i_flags;
794 t->parms.o_flags = p->o_flags;
795 t->parms.fwmark = p->fwmark;
796 dst_cache_reset(&t->dst_cache);
797 ip6gre_tnl_link_config(t, set_mtu);
798 return 0;
799 }
800
801 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
802 const struct ip6_tnl_parm2 *u)
803 {
804 p->laddr = u->laddr;
805 p->raddr = u->raddr;
806 p->flags = u->flags;
807 p->hop_limit = u->hop_limit;
808 p->encap_limit = u->encap_limit;
809 p->flowinfo = u->flowinfo;
810 p->link = u->link;
811 p->i_key = u->i_key;
812 p->o_key = u->o_key;
813 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
814 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
815 memcpy(p->name, u->name, sizeof(u->name));
816 }
817
818 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
819 const struct __ip6_tnl_parm *p)
820 {
821 u->proto = IPPROTO_GRE;
822 u->laddr = p->laddr;
823 u->raddr = p->raddr;
824 u->flags = p->flags;
825 u->hop_limit = p->hop_limit;
826 u->encap_limit = p->encap_limit;
827 u->flowinfo = p->flowinfo;
828 u->link = p->link;
829 u->i_key = p->i_key;
830 u->o_key = p->o_key;
831 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
832 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
833 memcpy(u->name, p->name, sizeof(u->name));
834 }
835
836 static int ip6gre_tunnel_ioctl(struct net_device *dev,
837 struct ifreq *ifr, int cmd)
838 {
839 int err = 0;
840 struct ip6_tnl_parm2 p;
841 struct __ip6_tnl_parm p1;
842 struct ip6_tnl *t = netdev_priv(dev);
843 struct net *net = t->net;
844 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
845
846 memset(&p1, 0, sizeof(p1));
847
848 switch (cmd) {
849 case SIOCGETTUNNEL:
850 if (dev == ign->fb_tunnel_dev) {
851 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
852 err = -EFAULT;
853 break;
854 }
855 ip6gre_tnl_parm_from_user(&p1, &p);
856 t = ip6gre_tunnel_locate(net, &p1, 0);
857 if (!t)
858 t = netdev_priv(dev);
859 }
860 memset(&p, 0, sizeof(p));
861 ip6gre_tnl_parm_to_user(&p, &t->parms);
862 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
863 err = -EFAULT;
864 break;
865
866 case SIOCADDTUNNEL:
867 case SIOCCHGTUNNEL:
868 err = -EPERM;
869 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
870 goto done;
871
872 err = -EFAULT;
873 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
874 goto done;
875
876 err = -EINVAL;
877 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
878 goto done;
879
880 if (!(p.i_flags&GRE_KEY))
881 p.i_key = 0;
882 if (!(p.o_flags&GRE_KEY))
883 p.o_key = 0;
884
885 ip6gre_tnl_parm_from_user(&p1, &p);
886 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
887
888 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
889 if (t) {
890 if (t->dev != dev) {
891 err = -EEXIST;
892 break;
893 }
894 } else {
895 t = netdev_priv(dev);
896
897 ip6gre_tunnel_unlink(ign, t);
898 synchronize_net();
899 ip6gre_tnl_change(t, &p1, 1);
900 ip6gre_tunnel_link(ign, t);
901 netdev_state_change(dev);
902 }
903 }
904
905 if (t) {
906 err = 0;
907
908 memset(&p, 0, sizeof(p));
909 ip6gre_tnl_parm_to_user(&p, &t->parms);
910 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
911 err = -EFAULT;
912 } else
913 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
914 break;
915
916 case SIOCDELTUNNEL:
917 err = -EPERM;
918 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
919 goto done;
920
921 if (dev == ign->fb_tunnel_dev) {
922 err = -EFAULT;
923 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
924 goto done;
925 err = -ENOENT;
926 ip6gre_tnl_parm_from_user(&p1, &p);
927 t = ip6gre_tunnel_locate(net, &p1, 0);
928 if (!t)
929 goto done;
930 err = -EPERM;
931 if (t == netdev_priv(ign->fb_tunnel_dev))
932 goto done;
933 dev = t->dev;
934 }
935 unregister_netdevice(dev);
936 err = 0;
937 break;
938
939 default:
940 err = -EINVAL;
941 }
942
943 done:
944 return err;
945 }
946
947 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
948 unsigned short type, const void *daddr,
949 const void *saddr, unsigned int len)
950 {
951 struct ip6_tnl *t = netdev_priv(dev);
952 struct ipv6hdr *ipv6h;
953 __be16 *p;
954
955 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
956 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
957 t->fl.u.ip6.flowlabel,
958 true, &t->fl.u.ip6));
959 ipv6h->hop_limit = t->parms.hop_limit;
960 ipv6h->nexthdr = NEXTHDR_GRE;
961 ipv6h->saddr = t->parms.laddr;
962 ipv6h->daddr = t->parms.raddr;
963
964 p = (__be16 *)(ipv6h + 1);
965 p[0] = t->parms.o_flags;
966 p[1] = htons(type);
967
968 /*
969 * Set the source hardware address.
970 */
971
972 if (saddr)
973 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
974 if (daddr)
975 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
976 if (!ipv6_addr_any(&ipv6h->daddr))
977 return t->hlen;
978
979 return -t->hlen;
980 }
981
982 static const struct header_ops ip6gre_header_ops = {
983 .create = ip6gre_header,
984 };
985
986 static const struct net_device_ops ip6gre_netdev_ops = {
987 .ndo_init = ip6gre_tunnel_init,
988 .ndo_uninit = ip6gre_tunnel_uninit,
989 .ndo_start_xmit = ip6gre_tunnel_xmit,
990 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
991 .ndo_change_mtu = ip6_tnl_change_mtu,
992 .ndo_get_stats64 = ip_tunnel_get_stats64,
993 .ndo_get_iflink = ip6_tnl_get_iflink,
994 };
995
996 static void ip6gre_dev_free(struct net_device *dev)
997 {
998 struct ip6_tnl *t = netdev_priv(dev);
999
1000 dst_cache_destroy(&t->dst_cache);
1001 free_percpu(dev->tstats);
1002 }
1003
1004 static void ip6gre_tunnel_setup(struct net_device *dev)
1005 {
1006 dev->netdev_ops = &ip6gre_netdev_ops;
1007 dev->needs_free_netdev = true;
1008 dev->priv_destructor = ip6gre_dev_free;
1009
1010 dev->type = ARPHRD_IP6GRE;
1011
1012 dev->flags |= IFF_NOARP;
1013 dev->addr_len = sizeof(struct in6_addr);
1014 netif_keep_dst(dev);
1015 /* This perm addr will be used as interface identifier by IPv6 */
1016 dev->addr_assign_type = NET_ADDR_RANDOM;
1017 eth_random_addr(dev->perm_addr);
1018 }
1019
1020 #define GRE6_FEATURES (NETIF_F_SG | \
1021 NETIF_F_FRAGLIST | \
1022 NETIF_F_HIGHDMA | \
1023 NETIF_F_HW_CSUM)
1024
1025 static void ip6gre_tnl_init_features(struct net_device *dev)
1026 {
1027 struct ip6_tnl *nt = netdev_priv(dev);
1028
1029 dev->features |= GRE6_FEATURES;
1030 dev->hw_features |= GRE6_FEATURES;
1031
1032 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1033 /* TCP offload with GRE SEQ is not supported, nor
1034 * can we support 2 levels of outer headers requiring
1035 * an update.
1036 */
1037 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1038 nt->encap.type == TUNNEL_ENCAP_NONE) {
1039 dev->features |= NETIF_F_GSO_SOFTWARE;
1040 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1041 }
1042
1043 /* Can use a lockless transmit, unless we generate
1044 * output sequences
1045 */
1046 dev->features |= NETIF_F_LLTX;
1047 }
1048 }
1049
1050 static int ip6gre_tunnel_init_common(struct net_device *dev)
1051 {
1052 struct ip6_tnl *tunnel;
1053 int ret;
1054 int t_hlen;
1055
1056 tunnel = netdev_priv(dev);
1057
1058 tunnel->dev = dev;
1059 tunnel->net = dev_net(dev);
1060 strcpy(tunnel->parms.name, dev->name);
1061
1062 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1063 if (!dev->tstats)
1064 return -ENOMEM;
1065
1066 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1067 if (ret) {
1068 free_percpu(dev->tstats);
1069 dev->tstats = NULL;
1070 return ret;
1071 }
1072
1073 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1074 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1075 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1076
1077 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1078 dev->mtu = ETH_DATA_LEN - t_hlen;
1079 if (dev->type == ARPHRD_ETHER)
1080 dev->mtu -= ETH_HLEN;
1081 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1082 dev->mtu -= 8;
1083
1084 ip6gre_tnl_init_features(dev);
1085
1086 return 0;
1087 }
1088
1089 static int ip6gre_tunnel_init(struct net_device *dev)
1090 {
1091 struct ip6_tnl *tunnel;
1092 int ret;
1093
1094 ret = ip6gre_tunnel_init_common(dev);
1095 if (ret)
1096 return ret;
1097
1098 tunnel = netdev_priv(dev);
1099
1100 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1101 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1102
1103 if (ipv6_addr_any(&tunnel->parms.raddr))
1104 dev->header_ops = &ip6gre_header_ops;
1105
1106 return 0;
1107 }
1108
1109 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1110 {
1111 struct ip6_tnl *tunnel = netdev_priv(dev);
1112
1113 tunnel->dev = dev;
1114 tunnel->net = dev_net(dev);
1115 strcpy(tunnel->parms.name, dev->name);
1116
1117 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1118
1119 dev_hold(dev);
1120 }
1121
1122
1123 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1124 .handler = gre_rcv,
1125 .err_handler = ip6gre_err,
1126 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1127 };
1128
1129 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1130 {
1131 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1132 struct net_device *dev, *aux;
1133 int prio;
1134
1135 for_each_netdev_safe(net, dev, aux)
1136 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1137 dev->rtnl_link_ops == &ip6gre_tap_ops)
1138 unregister_netdevice_queue(dev, head);
1139
1140 for (prio = 0; prio < 4; prio++) {
1141 int h;
1142 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1143 struct ip6_tnl *t;
1144
1145 t = rtnl_dereference(ign->tunnels[prio][h]);
1146
1147 while (t) {
1148 /* If dev is in the same netns, it has already
1149 * been added to the list by the previous loop.
1150 */
1151 if (!net_eq(dev_net(t->dev), net))
1152 unregister_netdevice_queue(t->dev,
1153 head);
1154 t = rtnl_dereference(t->next);
1155 }
1156 }
1157 }
1158 }
1159
1160 static int __net_init ip6gre_init_net(struct net *net)
1161 {
1162 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1163 int err;
1164
1165 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1166 NET_NAME_UNKNOWN,
1167 ip6gre_tunnel_setup);
1168 if (!ign->fb_tunnel_dev) {
1169 err = -ENOMEM;
1170 goto err_alloc_dev;
1171 }
1172 dev_net_set(ign->fb_tunnel_dev, net);
1173 /* FB netdevice is special: we have one, and only one per netns.
1174 * Allowing to move it to another netns is clearly unsafe.
1175 */
1176 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1177
1178
1179 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1180 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1181
1182 err = register_netdev(ign->fb_tunnel_dev);
1183 if (err)
1184 goto err_reg_dev;
1185
1186 rcu_assign_pointer(ign->tunnels_wc[0],
1187 netdev_priv(ign->fb_tunnel_dev));
1188 return 0;
1189
1190 err_reg_dev:
1191 free_netdev(ign->fb_tunnel_dev);
1192 err_alloc_dev:
1193 return err;
1194 }
1195
1196 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1197 {
1198 struct net *net;
1199 LIST_HEAD(list);
1200
1201 rtnl_lock();
1202 list_for_each_entry(net, net_list, exit_list)
1203 ip6gre_destroy_tunnels(net, &list);
1204 unregister_netdevice_many(&list);
1205 rtnl_unlock();
1206 }
1207
1208 static struct pernet_operations ip6gre_net_ops = {
1209 .init = ip6gre_init_net,
1210 .exit_batch = ip6gre_exit_batch_net,
1211 .id = &ip6gre_net_id,
1212 .size = sizeof(struct ip6gre_net),
1213 };
1214
1215 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1216 struct netlink_ext_ack *extack)
1217 {
1218 __be16 flags;
1219
1220 if (!data)
1221 return 0;
1222
1223 flags = 0;
1224 if (data[IFLA_GRE_IFLAGS])
1225 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1226 if (data[IFLA_GRE_OFLAGS])
1227 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1228 if (flags & (GRE_VERSION|GRE_ROUTING))
1229 return -EINVAL;
1230
1231 return 0;
1232 }
1233
1234 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1235 struct netlink_ext_ack *extack)
1236 {
1237 struct in6_addr daddr;
1238
1239 if (tb[IFLA_ADDRESS]) {
1240 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1241 return -EINVAL;
1242 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1243 return -EADDRNOTAVAIL;
1244 }
1245
1246 if (!data)
1247 goto out;
1248
1249 if (data[IFLA_GRE_REMOTE]) {
1250 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1251 if (ipv6_addr_any(&daddr))
1252 return -EINVAL;
1253 }
1254
1255 out:
1256 return ip6gre_tunnel_validate(tb, data, extack);
1257 }
1258
1259
1260 static void ip6gre_netlink_parms(struct nlattr *data[],
1261 struct __ip6_tnl_parm *parms)
1262 {
1263 memset(parms, 0, sizeof(*parms));
1264
1265 if (!data)
1266 return;
1267
1268 if (data[IFLA_GRE_LINK])
1269 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1270
1271 if (data[IFLA_GRE_IFLAGS])
1272 parms->i_flags = gre_flags_to_tnl_flags(
1273 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1274
1275 if (data[IFLA_GRE_OFLAGS])
1276 parms->o_flags = gre_flags_to_tnl_flags(
1277 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1278
1279 if (data[IFLA_GRE_IKEY])
1280 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1281
1282 if (data[IFLA_GRE_OKEY])
1283 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1284
1285 if (data[IFLA_GRE_LOCAL])
1286 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1287
1288 if (data[IFLA_GRE_REMOTE])
1289 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1290
1291 if (data[IFLA_GRE_TTL])
1292 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1293
1294 if (data[IFLA_GRE_ENCAP_LIMIT])
1295 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1296
1297 if (data[IFLA_GRE_FLOWINFO])
1298 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1299
1300 if (data[IFLA_GRE_FLAGS])
1301 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1302
1303 if (data[IFLA_GRE_FWMARK])
1304 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1305 }
1306
1307 static int ip6gre_tap_init(struct net_device *dev)
1308 {
1309 int ret;
1310
1311 ret = ip6gre_tunnel_init_common(dev);
1312 if (ret)
1313 return ret;
1314
1315 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1316
1317 return 0;
1318 }
1319
1320 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1321 .ndo_init = ip6gre_tap_init,
1322 .ndo_uninit = ip6gre_tunnel_uninit,
1323 .ndo_start_xmit = ip6gre_tunnel_xmit,
1324 .ndo_set_mac_address = eth_mac_addr,
1325 .ndo_validate_addr = eth_validate_addr,
1326 .ndo_change_mtu = ip6_tnl_change_mtu,
1327 .ndo_get_stats64 = ip_tunnel_get_stats64,
1328 .ndo_get_iflink = ip6_tnl_get_iflink,
1329 };
1330
1331 static void ip6gre_tap_setup(struct net_device *dev)
1332 {
1333
1334 ether_setup(dev);
1335
1336 dev->max_mtu = 0;
1337 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1338 dev->needs_free_netdev = true;
1339 dev->priv_destructor = ip6gre_dev_free;
1340
1341 dev->features |= NETIF_F_NETNS_LOCAL;
1342 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1343 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1344 netif_keep_dst(dev);
1345 }
1346
1347 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1348 struct ip_tunnel_encap *ipencap)
1349 {
1350 bool ret = false;
1351
1352 memset(ipencap, 0, sizeof(*ipencap));
1353
1354 if (!data)
1355 return ret;
1356
1357 if (data[IFLA_GRE_ENCAP_TYPE]) {
1358 ret = true;
1359 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1360 }
1361
1362 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1363 ret = true;
1364 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1365 }
1366
1367 if (data[IFLA_GRE_ENCAP_SPORT]) {
1368 ret = true;
1369 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1370 }
1371
1372 if (data[IFLA_GRE_ENCAP_DPORT]) {
1373 ret = true;
1374 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1375 }
1376
1377 return ret;
1378 }
1379
1380 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1381 struct nlattr *tb[], struct nlattr *data[],
1382 struct netlink_ext_ack *extack)
1383 {
1384 struct ip6_tnl *nt;
1385 struct net *net = dev_net(dev);
1386 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1387 struct ip_tunnel_encap ipencap;
1388 int err;
1389
1390 nt = netdev_priv(dev);
1391
1392 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1393 int err = ip6_tnl_encap_setup(nt, &ipencap);
1394
1395 if (err < 0)
1396 return err;
1397 }
1398
1399 ip6gre_netlink_parms(data, &nt->parms);
1400
1401 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1402 return -EEXIST;
1403
1404 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1405 eth_hw_addr_random(dev);
1406
1407 nt->dev = dev;
1408 nt->net = dev_net(dev);
1409
1410 err = register_netdevice(dev);
1411 if (err)
1412 goto out;
1413
1414 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1415
1416 if (tb[IFLA_MTU])
1417 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1418
1419 dev_hold(dev);
1420 ip6gre_tunnel_link(ign, nt);
1421
1422 out:
1423 return err;
1424 }
1425
1426 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1427 struct nlattr *data[],
1428 struct netlink_ext_ack *extack)
1429 {
1430 struct ip6_tnl *t, *nt = netdev_priv(dev);
1431 struct net *net = nt->net;
1432 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1433 struct __ip6_tnl_parm p;
1434 struct ip_tunnel_encap ipencap;
1435
1436 if (dev == ign->fb_tunnel_dev)
1437 return -EINVAL;
1438
1439 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1440 int err = ip6_tnl_encap_setup(nt, &ipencap);
1441
1442 if (err < 0)
1443 return err;
1444 }
1445
1446 ip6gre_netlink_parms(data, &p);
1447
1448 t = ip6gre_tunnel_locate(net, &p, 0);
1449
1450 if (t) {
1451 if (t->dev != dev)
1452 return -EEXIST;
1453 } else {
1454 t = nt;
1455 }
1456
1457 ip6gre_tunnel_unlink(ign, t);
1458 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1459 ip6gre_tunnel_link(ign, t);
1460 return 0;
1461 }
1462
1463 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1464 {
1465 struct net *net = dev_net(dev);
1466 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1467
1468 if (dev != ign->fb_tunnel_dev)
1469 unregister_netdevice_queue(dev, head);
1470 }
1471
1472 static size_t ip6gre_get_size(const struct net_device *dev)
1473 {
1474 return
1475 /* IFLA_GRE_LINK */
1476 nla_total_size(4) +
1477 /* IFLA_GRE_IFLAGS */
1478 nla_total_size(2) +
1479 /* IFLA_GRE_OFLAGS */
1480 nla_total_size(2) +
1481 /* IFLA_GRE_IKEY */
1482 nla_total_size(4) +
1483 /* IFLA_GRE_OKEY */
1484 nla_total_size(4) +
1485 /* IFLA_GRE_LOCAL */
1486 nla_total_size(sizeof(struct in6_addr)) +
1487 /* IFLA_GRE_REMOTE */
1488 nla_total_size(sizeof(struct in6_addr)) +
1489 /* IFLA_GRE_TTL */
1490 nla_total_size(1) +
1491 /* IFLA_GRE_ENCAP_LIMIT */
1492 nla_total_size(1) +
1493 /* IFLA_GRE_FLOWINFO */
1494 nla_total_size(4) +
1495 /* IFLA_GRE_FLAGS */
1496 nla_total_size(4) +
1497 /* IFLA_GRE_ENCAP_TYPE */
1498 nla_total_size(2) +
1499 /* IFLA_GRE_ENCAP_FLAGS */
1500 nla_total_size(2) +
1501 /* IFLA_GRE_ENCAP_SPORT */
1502 nla_total_size(2) +
1503 /* IFLA_GRE_ENCAP_DPORT */
1504 nla_total_size(2) +
1505 /* IFLA_GRE_FWMARK */
1506 nla_total_size(4) +
1507 0;
1508 }
1509
1510 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1511 {
1512 struct ip6_tnl *t = netdev_priv(dev);
1513 struct __ip6_tnl_parm *p = &t->parms;
1514
1515 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1516 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1517 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1518 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1519 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1520 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1521 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1522 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1523 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1524 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1525 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1526 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1527 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1528 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1529 goto nla_put_failure;
1530
1531 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1532 t->encap.type) ||
1533 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1534 t->encap.sport) ||
1535 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1536 t->encap.dport) ||
1537 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1538 t->encap.flags))
1539 goto nla_put_failure;
1540
1541 return 0;
1542
1543 nla_put_failure:
1544 return -EMSGSIZE;
1545 }
1546
1547 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1548 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1549 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1550 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1551 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1552 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1553 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1554 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1555 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1556 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1557 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1558 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1559 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1560 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1561 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1562 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1563 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1564 };
1565
1566 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1567 .kind = "ip6gre",
1568 .maxtype = IFLA_GRE_MAX,
1569 .policy = ip6gre_policy,
1570 .priv_size = sizeof(struct ip6_tnl),
1571 .setup = ip6gre_tunnel_setup,
1572 .validate = ip6gre_tunnel_validate,
1573 .newlink = ip6gre_newlink,
1574 .changelink = ip6gre_changelink,
1575 .dellink = ip6gre_dellink,
1576 .get_size = ip6gre_get_size,
1577 .fill_info = ip6gre_fill_info,
1578 .get_link_net = ip6_tnl_get_link_net,
1579 };
1580
1581 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1582 .kind = "ip6gretap",
1583 .maxtype = IFLA_GRE_MAX,
1584 .policy = ip6gre_policy,
1585 .priv_size = sizeof(struct ip6_tnl),
1586 .setup = ip6gre_tap_setup,
1587 .validate = ip6gre_tap_validate,
1588 .newlink = ip6gre_newlink,
1589 .changelink = ip6gre_changelink,
1590 .get_size = ip6gre_get_size,
1591 .fill_info = ip6gre_fill_info,
1592 .get_link_net = ip6_tnl_get_link_net,
1593 };
1594
1595 /*
1596 * And now the modules code and kernel interface.
1597 */
1598
1599 static int __init ip6gre_init(void)
1600 {
1601 int err;
1602
1603 pr_info("GRE over IPv6 tunneling driver\n");
1604
1605 err = register_pernet_device(&ip6gre_net_ops);
1606 if (err < 0)
1607 return err;
1608
1609 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1610 if (err < 0) {
1611 pr_info("%s: can't add protocol\n", __func__);
1612 goto add_proto_failed;
1613 }
1614
1615 err = rtnl_link_register(&ip6gre_link_ops);
1616 if (err < 0)
1617 goto rtnl_link_failed;
1618
1619 err = rtnl_link_register(&ip6gre_tap_ops);
1620 if (err < 0)
1621 goto tap_ops_failed;
1622
1623 out:
1624 return err;
1625
1626 tap_ops_failed:
1627 rtnl_link_unregister(&ip6gre_link_ops);
1628 rtnl_link_failed:
1629 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1630 add_proto_failed:
1631 unregister_pernet_device(&ip6gre_net_ops);
1632 goto out;
1633 }
1634
1635 static void __exit ip6gre_fini(void)
1636 {
1637 rtnl_link_unregister(&ip6gre_tap_ops);
1638 rtnl_link_unregister(&ip6gre_link_ops);
1639 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1640 unregister_pernet_device(&ip6gre_net_ops);
1641 }
1642
1643 module_init(ip6gre_init);
1644 module_exit(ip6gre_fini);
1645 MODULE_LICENSE("GPL");
1646 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1647 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1648 MODULE_ALIAS_RTNL_LINK("ip6gre");
1649 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1650 MODULE_ALIAS_NETDEV("ip6gre0");
1651
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.