?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/speculation/mds: Fix documentation typo
[linux/fpc-iii.git] / net / ipv6 / ip6_gre.c
blob4228f3b2f34765a01653aac1fb04f0611031c31e
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 const struct gre_base_hdr *greh;
376 const struct ipv6hdr *ipv6h;
377 int grehlen = sizeof(*greh);
378 struct ip6_tnl *t;
379 int key_off = 0;
380 __be16 flags;
381 __be32 key;
382
383 if (!pskb_may_pull(skb, offset + grehlen))
384 return;
385 greh = (const struct gre_base_hdr *)(skb->data + offset);
386 flags = greh->flags;
387 if (flags & (GRE_VERSION | GRE_ROUTING))
388 return;
389 if (flags & GRE_CSUM)
390 grehlen += 4;
391 if (flags & GRE_KEY) {
392 key_off = grehlen + offset;
393 grehlen += 4;
394 }
395
396 if (!pskb_may_pull(skb, offset + grehlen))
397 return;
398 ipv6h = (const struct ipv6hdr *)skb->data;
399 greh = (const struct gre_base_hdr *)(skb->data + offset);
400 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
401
402 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
403 key, greh->protocol);
404 if (!t)
405 return;
406
407 switch (type) {
408 __u32 teli;
409 struct ipv6_tlv_tnl_enc_lim *tel;
410 __u32 mtu;
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 mtu = be32_to_cpu(info) - offset - t->tun_hlen;
442 if (t->dev->type == ARPHRD_ETHER)
443 mtu -= ETH_HLEN;
444 if (mtu < IPV6_MIN_MTU)
445 mtu = IPV6_MIN_MTU;
446 t->dev->mtu = mtu;
447 return;
448 }
449
450 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
451 t->err_count++;
452 else
453 t->err_count = 1;
454 t->err_time = jiffies;
455 }
456
457 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
458 {
459 const struct ipv6hdr *ipv6h;
460 struct ip6_tnl *tunnel;
461
462 ipv6h = ipv6_hdr(skb);
463 tunnel = ip6gre_tunnel_lookup(skb->dev,
464 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
465 tpi->proto);
466 if (tunnel) {
467 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
468
469 return PACKET_RCVD;
470 }
471
472 return PACKET_REJECT;
473 }
474
475 static int gre_rcv(struct sk_buff *skb)
476 {
477 struct tnl_ptk_info tpi;
478 bool csum_err = false;
479 int hdr_len;
480
481 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
482 if (hdr_len < 0)
483 goto drop;
484
485 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
486 goto drop;
487
488 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
489 return 0;
490
491 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
492 drop:
493 kfree_skb(skb);
494 return 0;
495 }
496
497 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
498 {
499 return iptunnel_handle_offloads(skb,
500 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
501 }
502
503 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
504 struct net_device *dev, __u8 dsfield,
505 struct flowi6 *fl6, int encap_limit,
506 __u32 *pmtu, __be16 proto)
507 {
508 struct ip6_tnl *tunnel = netdev_priv(dev);
509 struct dst_entry *dst = skb_dst(skb);
510 __be16 protocol;
511
512 if (dev->type == ARPHRD_ETHER)
513 IPCB(skb)->flags = 0;
514
515 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
516 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
517 else
518 fl6->daddr = tunnel->parms.raddr;
519
520 if (tunnel->parms.o_flags & TUNNEL_SEQ)
521 tunnel->o_seqno++;
522
523 /* Push GRE header. */
524 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
525 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
526 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
527
528 /* TooBig packet may have updated dst->dev's mtu */
529 if (dst && dst_mtu(dst) > dst->dev->mtu)
530 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
531
532 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
533 NEXTHDR_GRE);
534 }
535
536 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
537 {
538 struct ip6_tnl *t = netdev_priv(dev);
539 const struct iphdr *iph = ip_hdr(skb);
540 int encap_limit = -1;
541 struct flowi6 fl6;
542 __u8 dsfield;
543 __u32 mtu;
544 int err;
545
546 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
547
548 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
549 encap_limit = t->parms.encap_limit;
550
551 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
552
553 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
554 dsfield = ipv4_get_dsfield(iph);
555 else
556 dsfield = ip6_tclass(t->parms.flowinfo);
557 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
558 fl6.flowi6_mark = skb->mark;
559 else
560 fl6.flowi6_mark = t->parms.fwmark;
561
562 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
563
564 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
565 if (err)
566 return -1;
567
568 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
569 skb->protocol);
570 if (err != 0) {
571 /* XXX: send ICMP error even if DF is not set. */
572 if (err == -EMSGSIZE)
573 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
574 htonl(mtu));
575 return -1;
576 }
577
578 return 0;
579 }
580
581 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
582 {
583 struct ip6_tnl *t = netdev_priv(dev);
584 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
585 int encap_limit = -1;
586 __u16 offset;
587 struct flowi6 fl6;
588 __u8 dsfield;
589 __u32 mtu;
590 int err;
591
592 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
593 return -1;
594
595 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
596 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
597 ipv6h = ipv6_hdr(skb);
598
599 if (offset > 0) {
600 struct ipv6_tlv_tnl_enc_lim *tel;
601 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
602 if (tel->encap_limit == 0) {
603 icmpv6_send(skb, ICMPV6_PARAMPROB,
604 ICMPV6_HDR_FIELD, offset + 2);
605 return -1;
606 }
607 encap_limit = tel->encap_limit - 1;
608 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
609 encap_limit = t->parms.encap_limit;
610
611 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
612
613 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
614 dsfield = ipv6_get_dsfield(ipv6h);
615 else
616 dsfield = ip6_tclass(t->parms.flowinfo);
617
618 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
619 fl6.flowlabel |= ip6_flowlabel(ipv6h);
620 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
621 fl6.flowi6_mark = skb->mark;
622 else
623 fl6.flowi6_mark = t->parms.fwmark;
624
625 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
626
627 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
628 return -1;
629
630 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
631 &mtu, skb->protocol);
632 if (err != 0) {
633 if (err == -EMSGSIZE)
634 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
635 return -1;
636 }
637
638 return 0;
639 }
640
641 /**
642 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
643 * @t: the outgoing tunnel device
644 * @hdr: IPv6 header from the incoming packet
645 *
646 * Description:
647 * Avoid trivial tunneling loop by checking that tunnel exit-point
648 * doesn't match source of incoming packet.
649 *
650 * Return:
651 * 1 if conflict,
652 * 0 else
653 **/
654
655 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
656 const struct ipv6hdr *hdr)
657 {
658 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
659 }
660
661 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
662 {
663 struct ip6_tnl *t = netdev_priv(dev);
664 int encap_limit = -1;
665 struct flowi6 fl6;
666 __u32 mtu;
667 int err;
668
669 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
670 encap_limit = t->parms.encap_limit;
671
672 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
673
674 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
675 if (err)
676 return err;
677
678 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
679
680 return err;
681 }
682
683 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
684 struct net_device *dev)
685 {
686 struct ip6_tnl *t = netdev_priv(dev);
687 struct net_device_stats *stats = &t->dev->stats;
688 int ret;
689
690 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
691 goto tx_err;
692
693 switch (skb->protocol) {
694 case htons(ETH_P_IP):
695 ret = ip6gre_xmit_ipv4(skb, dev);
696 break;
697 case htons(ETH_P_IPV6):
698 ret = ip6gre_xmit_ipv6(skb, dev);
699 break;
700 default:
701 ret = ip6gre_xmit_other(skb, dev);
702 break;
703 }
704
705 if (ret < 0)
706 goto tx_err;
707
708 return NETDEV_TX_OK;
709
710 tx_err:
711 stats->tx_errors++;
712 stats->tx_dropped++;
713 kfree_skb(skb);
714 return NETDEV_TX_OK;
715 }
716
717 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
718 {
719 struct net_device *dev = t->dev;
720 struct __ip6_tnl_parm *p = &t->parms;
721 struct flowi6 *fl6 = &t->fl.u.ip6;
722 int t_hlen;
723
724 if (dev->type != ARPHRD_ETHER) {
725 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
726 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
727 }
728
729 /* Set up flowi template */
730 fl6->saddr = p->laddr;
731 fl6->daddr = p->raddr;
732 fl6->flowi6_oif = p->link;
733 fl6->flowlabel = 0;
734 fl6->flowi6_proto = IPPROTO_GRE;
735
736 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
737 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
738 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
739 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
740
741 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
742 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
743
744 if (p->flags&IP6_TNL_F_CAP_XMIT &&
745 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
746 dev->flags |= IFF_POINTOPOINT;
747 else
748 dev->flags &= ~IFF_POINTOPOINT;
749
750 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
751
752 t->hlen = t->encap_hlen + t->tun_hlen;
753
754 t_hlen = t->hlen + sizeof(struct ipv6hdr);
755
756 if (p->flags & IP6_TNL_F_CAP_XMIT) {
757 int strict = (ipv6_addr_type(&p->raddr) &
758 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
759
760 struct rt6_info *rt = rt6_lookup(t->net,
761 &p->raddr, &p->laddr,
762 p->link, strict);
763
764 if (!rt)
765 return;
766
767 if (rt->dst.dev) {
768 dev->hard_header_len = rt->dst.dev->hard_header_len +
769 t_hlen;
770
771 if (set_mtu) {
772 dev->mtu = rt->dst.dev->mtu - t_hlen;
773 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
774 dev->mtu -= 8;
775 if (dev->type == ARPHRD_ETHER)
776 dev->mtu -= ETH_HLEN;
777
778 if (dev->mtu < IPV6_MIN_MTU)
779 dev->mtu = IPV6_MIN_MTU;
780 }
781 }
782 ip6_rt_put(rt);
783 }
784 }
785
786 static int ip6gre_tnl_change(struct ip6_tnl *t,
787 const struct __ip6_tnl_parm *p, int set_mtu)
788 {
789 t->parms.laddr = p->laddr;
790 t->parms.raddr = p->raddr;
791 t->parms.flags = p->flags;
792 t->parms.hop_limit = p->hop_limit;
793 t->parms.encap_limit = p->encap_limit;
794 t->parms.flowinfo = p->flowinfo;
795 t->parms.link = p->link;
796 t->parms.proto = p->proto;
797 t->parms.i_key = p->i_key;
798 t->parms.o_key = p->o_key;
799 t->parms.i_flags = p->i_flags;
800 t->parms.o_flags = p->o_flags;
801 t->parms.fwmark = p->fwmark;
802 dst_cache_reset(&t->dst_cache);
803 ip6gre_tnl_link_config(t, set_mtu);
804 return 0;
805 }
806
807 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
808 const struct ip6_tnl_parm2 *u)
809 {
810 p->laddr = u->laddr;
811 p->raddr = u->raddr;
812 p->flags = u->flags;
813 p->hop_limit = u->hop_limit;
814 p->encap_limit = u->encap_limit;
815 p->flowinfo = u->flowinfo;
816 p->link = u->link;
817 p->i_key = u->i_key;
818 p->o_key = u->o_key;
819 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
820 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
821 memcpy(p->name, u->name, sizeof(u->name));
822 }
823
824 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
825 const struct __ip6_tnl_parm *p)
826 {
827 u->proto = IPPROTO_GRE;
828 u->laddr = p->laddr;
829 u->raddr = p->raddr;
830 u->flags = p->flags;
831 u->hop_limit = p->hop_limit;
832 u->encap_limit = p->encap_limit;
833 u->flowinfo = p->flowinfo;
834 u->link = p->link;
835 u->i_key = p->i_key;
836 u->o_key = p->o_key;
837 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
838 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
839 memcpy(u->name, p->name, sizeof(u->name));
840 }
841
842 static int ip6gre_tunnel_ioctl(struct net_device *dev,
843 struct ifreq *ifr, int cmd)
844 {
845 int err = 0;
846 struct ip6_tnl_parm2 p;
847 struct __ip6_tnl_parm p1;
848 struct ip6_tnl *t = netdev_priv(dev);
849 struct net *net = t->net;
850 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
851
852 memset(&p1, 0, sizeof(p1));
853
854 switch (cmd) {
855 case SIOCGETTUNNEL:
856 if (dev == ign->fb_tunnel_dev) {
857 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
858 err = -EFAULT;
859 break;
860 }
861 ip6gre_tnl_parm_from_user(&p1, &p);
862 t = ip6gre_tunnel_locate(net, &p1, 0);
863 if (!t)
864 t = netdev_priv(dev);
865 }
866 memset(&p, 0, sizeof(p));
867 ip6gre_tnl_parm_to_user(&p, &t->parms);
868 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
869 err = -EFAULT;
870 break;
871
872 case SIOCADDTUNNEL:
873 case SIOCCHGTUNNEL:
874 err = -EPERM;
875 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
876 goto done;
877
878 err = -EFAULT;
879 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
880 goto done;
881
882 err = -EINVAL;
883 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
884 goto done;
885
886 if (!(p.i_flags&GRE_KEY))
887 p.i_key = 0;
888 if (!(p.o_flags&GRE_KEY))
889 p.o_key = 0;
890
891 ip6gre_tnl_parm_from_user(&p1, &p);
892 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
893
894 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
895 if (t) {
896 if (t->dev != dev) {
897 err = -EEXIST;
898 break;
899 }
900 } else {
901 t = netdev_priv(dev);
902
903 ip6gre_tunnel_unlink(ign, t);
904 synchronize_net();
905 ip6gre_tnl_change(t, &p1, 1);
906 ip6gre_tunnel_link(ign, t);
907 netdev_state_change(dev);
908 }
909 }
910
911 if (t) {
912 err = 0;
913
914 memset(&p, 0, sizeof(p));
915 ip6gre_tnl_parm_to_user(&p, &t->parms);
916 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
917 err = -EFAULT;
918 } else
919 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
920 break;
921
922 case SIOCDELTUNNEL:
923 err = -EPERM;
924 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
925 goto done;
926
927 if (dev == ign->fb_tunnel_dev) {
928 err = -EFAULT;
929 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
930 goto done;
931 err = -ENOENT;
932 ip6gre_tnl_parm_from_user(&p1, &p);
933 t = ip6gre_tunnel_locate(net, &p1, 0);
934 if (!t)
935 goto done;
936 err = -EPERM;
937 if (t == netdev_priv(ign->fb_tunnel_dev))
938 goto done;
939 dev = t->dev;
940 }
941 unregister_netdevice(dev);
942 err = 0;
943 break;
944
945 default:
946 err = -EINVAL;
947 }
948
949 done:
950 return err;
951 }
952
953 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
954 unsigned short type, const void *daddr,
955 const void *saddr, unsigned int len)
956 {
957 struct ip6_tnl *t = netdev_priv(dev);
958 struct ipv6hdr *ipv6h;
959 __be16 *p;
960
961 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
962 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
963 t->fl.u.ip6.flowlabel,
964 true, &t->fl.u.ip6));
965 ipv6h->hop_limit = t->parms.hop_limit;
966 ipv6h->nexthdr = NEXTHDR_GRE;
967 ipv6h->saddr = t->parms.laddr;
968 ipv6h->daddr = t->parms.raddr;
969
970 p = (__be16 *)(ipv6h + 1);
971 p[0] = t->parms.o_flags;
972 p[1] = htons(type);
973
974 /*
975 * Set the source hardware address.
976 */
977
978 if (saddr)
979 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
980 if (daddr)
981 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
982 if (!ipv6_addr_any(&ipv6h->daddr))
983 return t->hlen;
984
985 return -t->hlen;
986 }
987
988 static const struct header_ops ip6gre_header_ops = {
989 .create = ip6gre_header,
990 };
991
992 static const struct net_device_ops ip6gre_netdev_ops = {
993 .ndo_init = ip6gre_tunnel_init,
994 .ndo_uninit = ip6gre_tunnel_uninit,
995 .ndo_start_xmit = ip6gre_tunnel_xmit,
996 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
997 .ndo_change_mtu = ip6_tnl_change_mtu,
998 .ndo_get_stats64 = ip_tunnel_get_stats64,
999 .ndo_get_iflink = ip6_tnl_get_iflink,
1000 };
1001
1002 static void ip6gre_dev_free(struct net_device *dev)
1003 {
1004 struct ip6_tnl *t = netdev_priv(dev);
1005
1006 dst_cache_destroy(&t->dst_cache);
1007 free_percpu(dev->tstats);
1008 }
1009
1010 static void ip6gre_tunnel_setup(struct net_device *dev)
1011 {
1012 dev->netdev_ops = &ip6gre_netdev_ops;
1013 dev->needs_free_netdev = true;
1014 dev->priv_destructor = ip6gre_dev_free;
1015
1016 dev->type = ARPHRD_IP6GRE;
1017
1018 dev->flags |= IFF_NOARP;
1019 dev->addr_len = sizeof(struct in6_addr);
1020 netif_keep_dst(dev);
1021 /* This perm addr will be used as interface identifier by IPv6 */
1022 dev->addr_assign_type = NET_ADDR_RANDOM;
1023 eth_random_addr(dev->perm_addr);
1024 }
1025
1026 #define GRE6_FEATURES (NETIF_F_SG | \
1027 NETIF_F_FRAGLIST | \
1028 NETIF_F_HIGHDMA | \
1029 NETIF_F_HW_CSUM)
1030
1031 static void ip6gre_tnl_init_features(struct net_device *dev)
1032 {
1033 struct ip6_tnl *nt = netdev_priv(dev);
1034
1035 dev->features |= GRE6_FEATURES;
1036 dev->hw_features |= GRE6_FEATURES;
1037
1038 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1039 /* TCP offload with GRE SEQ is not supported, nor
1040 * can we support 2 levels of outer headers requiring
1041 * an update.
1042 */
1043 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1044 nt->encap.type == TUNNEL_ENCAP_NONE) {
1045 dev->features |= NETIF_F_GSO_SOFTWARE;
1046 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1047 }
1048
1049 /* Can use a lockless transmit, unless we generate
1050 * output sequences
1051 */
1052 dev->features |= NETIF_F_LLTX;
1053 }
1054 }
1055
1056 static int ip6gre_tunnel_init_common(struct net_device *dev)
1057 {
1058 struct ip6_tnl *tunnel;
1059 int ret;
1060 int t_hlen;
1061
1062 tunnel = netdev_priv(dev);
1063
1064 tunnel->dev = dev;
1065 tunnel->net = dev_net(dev);
1066 strcpy(tunnel->parms.name, dev->name);
1067
1068 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1069 if (!dev->tstats)
1070 return -ENOMEM;
1071
1072 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1073 if (ret) {
1074 free_percpu(dev->tstats);
1075 dev->tstats = NULL;
1076 return ret;
1077 }
1078
1079 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1080 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1081 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1082
1083 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1084 dev->mtu = ETH_DATA_LEN - t_hlen;
1085 if (dev->type == ARPHRD_ETHER)
1086 dev->mtu -= ETH_HLEN;
1087 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1088 dev->mtu -= 8;
1089
1090 ip6gre_tnl_init_features(dev);
1091
1092 return 0;
1093 }
1094
1095 static int ip6gre_tunnel_init(struct net_device *dev)
1096 {
1097 struct ip6_tnl *tunnel;
1098 int ret;
1099
1100 ret = ip6gre_tunnel_init_common(dev);
1101 if (ret)
1102 return ret;
1103
1104 tunnel = netdev_priv(dev);
1105
1106 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1107 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1108
1109 if (ipv6_addr_any(&tunnel->parms.raddr))
1110 dev->header_ops = &ip6gre_header_ops;
1111
1112 return 0;
1113 }
1114
1115 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1116 {
1117 struct ip6_tnl *tunnel = netdev_priv(dev);
1118
1119 tunnel->dev = dev;
1120 tunnel->net = dev_net(dev);
1121 strcpy(tunnel->parms.name, dev->name);
1122
1123 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1124
1125 dev_hold(dev);
1126 }
1127
1128
1129 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1130 .handler = gre_rcv,
1131 .err_handler = ip6gre_err,
1132 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1133 };
1134
1135 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1136 {
1137 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1138 struct net_device *dev, *aux;
1139 int prio;
1140
1141 for_each_netdev_safe(net, dev, aux)
1142 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1143 dev->rtnl_link_ops == &ip6gre_tap_ops)
1144 unregister_netdevice_queue(dev, head);
1145
1146 for (prio = 0; prio < 4; prio++) {
1147 int h;
1148 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1149 struct ip6_tnl *t;
1150
1151 t = rtnl_dereference(ign->tunnels[prio][h]);
1152
1153 while (t) {
1154 /* If dev is in the same netns, it has already
1155 * been added to the list by the previous loop.
1156 */
1157 if (!net_eq(dev_net(t->dev), net))
1158 unregister_netdevice_queue(t->dev,
1159 head);
1160 t = rtnl_dereference(t->next);
1161 }
1162 }
1163 }
1164 }
1165
1166 static int __net_init ip6gre_init_net(struct net *net)
1167 {
1168 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1169 int err;
1170
1171 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1172 NET_NAME_UNKNOWN,
1173 ip6gre_tunnel_setup);
1174 if (!ign->fb_tunnel_dev) {
1175 err = -ENOMEM;
1176 goto err_alloc_dev;
1177 }
1178 dev_net_set(ign->fb_tunnel_dev, net);
1179 /* FB netdevice is special: we have one, and only one per netns.
1180 * Allowing to move it to another netns is clearly unsafe.
1181 */
1182 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1183
1184
1185 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1186 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1187
1188 err = register_netdev(ign->fb_tunnel_dev);
1189 if (err)
1190 goto err_reg_dev;
1191
1192 rcu_assign_pointer(ign->tunnels_wc[0],
1193 netdev_priv(ign->fb_tunnel_dev));
1194 return 0;
1195
1196 err_reg_dev:
1197 free_netdev(ign->fb_tunnel_dev);
1198 err_alloc_dev:
1199 return err;
1200 }
1201
1202 static void __net_exit ip6gre_exit_net(struct net *net)
1203 {
1204 LIST_HEAD(list);
1205
1206 rtnl_lock();
1207 ip6gre_destroy_tunnels(net, &list);
1208 unregister_netdevice_many(&list);
1209 rtnl_unlock();
1210 }
1211
1212 static struct pernet_operations ip6gre_net_ops = {
1213 .init = ip6gre_init_net,
1214 .exit = ip6gre_exit_net,
1215 .id = &ip6gre_net_id,
1216 .size = sizeof(struct ip6gre_net),
1217 };
1218
1219 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1220 struct netlink_ext_ack *extack)
1221 {
1222 __be16 flags;
1223
1224 if (!data)
1225 return 0;
1226
1227 flags = 0;
1228 if (data[IFLA_GRE_IFLAGS])
1229 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1230 if (data[IFLA_GRE_OFLAGS])
1231 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1232 if (flags & (GRE_VERSION|GRE_ROUTING))
1233 return -EINVAL;
1234
1235 return 0;
1236 }
1237
1238 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1239 struct netlink_ext_ack *extack)
1240 {
1241 struct in6_addr daddr;
1242
1243 if (tb[IFLA_ADDRESS]) {
1244 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1245 return -EINVAL;
1246 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1247 return -EADDRNOTAVAIL;
1248 }
1249
1250 if (!data)
1251 goto out;
1252
1253 if (data[IFLA_GRE_REMOTE]) {
1254 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1255 if (ipv6_addr_any(&daddr))
1256 return -EINVAL;
1257 }
1258
1259 out:
1260 return ip6gre_tunnel_validate(tb, data, extack);
1261 }
1262
1263
1264 static void ip6gre_netlink_parms(struct nlattr *data[],
1265 struct __ip6_tnl_parm *parms)
1266 {
1267 memset(parms, 0, sizeof(*parms));
1268
1269 if (!data)
1270 return;
1271
1272 if (data[IFLA_GRE_LINK])
1273 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1274
1275 if (data[IFLA_GRE_IFLAGS])
1276 parms->i_flags = gre_flags_to_tnl_flags(
1277 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1278
1279 if (data[IFLA_GRE_OFLAGS])
1280 parms->o_flags = gre_flags_to_tnl_flags(
1281 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1282
1283 if (data[IFLA_GRE_IKEY])
1284 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1285
1286 if (data[IFLA_GRE_OKEY])
1287 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1288
1289 if (data[IFLA_GRE_LOCAL])
1290 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1291
1292 if (data[IFLA_GRE_REMOTE])
1293 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1294
1295 if (data[IFLA_GRE_TTL])
1296 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1297
1298 if (data[IFLA_GRE_ENCAP_LIMIT])
1299 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1300
1301 if (data[IFLA_GRE_FLOWINFO])
1302 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1303
1304 if (data[IFLA_GRE_FLAGS])
1305 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1306
1307 if (data[IFLA_GRE_FWMARK])
1308 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1309 }
1310
1311 static int ip6gre_tap_init(struct net_device *dev)
1312 {
1313 int ret;
1314
1315 ret = ip6gre_tunnel_init_common(dev);
1316 if (ret)
1317 return ret;
1318
1319 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1320
1321 return 0;
1322 }
1323
1324 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1325 .ndo_init = ip6gre_tap_init,
1326 .ndo_uninit = ip6gre_tunnel_uninit,
1327 .ndo_start_xmit = ip6gre_tunnel_xmit,
1328 .ndo_set_mac_address = eth_mac_addr,
1329 .ndo_validate_addr = eth_validate_addr,
1330 .ndo_change_mtu = ip6_tnl_change_mtu,
1331 .ndo_get_stats64 = ip_tunnel_get_stats64,
1332 .ndo_get_iflink = ip6_tnl_get_iflink,
1333 };
1334
1335 static void ip6gre_tap_setup(struct net_device *dev)
1336 {
1337
1338 ether_setup(dev);
1339
1340 dev->max_mtu = 0;
1341 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1342 dev->needs_free_netdev = true;
1343 dev->priv_destructor = ip6gre_dev_free;
1344
1345 dev->features |= NETIF_F_NETNS_LOCAL;
1346 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1347 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1348 netif_keep_dst(dev);
1349 }
1350
1351 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1352 struct ip_tunnel_encap *ipencap)
1353 {
1354 bool ret = false;
1355
1356 memset(ipencap, 0, sizeof(*ipencap));
1357
1358 if (!data)
1359 return ret;
1360
1361 if (data[IFLA_GRE_ENCAP_TYPE]) {
1362 ret = true;
1363 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1364 }
1365
1366 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1367 ret = true;
1368 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1369 }
1370
1371 if (data[IFLA_GRE_ENCAP_SPORT]) {
1372 ret = true;
1373 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1374 }
1375
1376 if (data[IFLA_GRE_ENCAP_DPORT]) {
1377 ret = true;
1378 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1379 }
1380
1381 return ret;
1382 }
1383
1384 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1385 struct nlattr *tb[], struct nlattr *data[],
1386 struct netlink_ext_ack *extack)
1387 {
1388 struct ip6_tnl *nt;
1389 struct net *net = dev_net(dev);
1390 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1391 struct ip_tunnel_encap ipencap;
1392 int err;
1393
1394 nt = netdev_priv(dev);
1395
1396 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1397 int err = ip6_tnl_encap_setup(nt, &ipencap);
1398
1399 if (err < 0)
1400 return err;
1401 }
1402
1403 ip6gre_netlink_parms(data, &nt->parms);
1404
1405 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1406 return -EEXIST;
1407
1408 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1409 eth_hw_addr_random(dev);
1410
1411 nt->dev = dev;
1412 nt->net = dev_net(dev);
1413
1414 err = register_netdevice(dev);
1415 if (err)
1416 goto out;
1417
1418 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1419
1420 if (tb[IFLA_MTU])
1421 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1422
1423 dev_hold(dev);
1424 ip6gre_tunnel_link(ign, nt);
1425
1426 out:
1427 return err;
1428 }
1429
1430 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1431 struct nlattr *data[],
1432 struct netlink_ext_ack *extack)
1433 {
1434 struct ip6_tnl *t, *nt = netdev_priv(dev);
1435 struct net *net = nt->net;
1436 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1437 struct __ip6_tnl_parm p;
1438 struct ip_tunnel_encap ipencap;
1439
1440 if (dev == ign->fb_tunnel_dev)
1441 return -EINVAL;
1442
1443 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1444 int err = ip6_tnl_encap_setup(nt, &ipencap);
1445
1446 if (err < 0)
1447 return err;
1448 }
1449
1450 ip6gre_netlink_parms(data, &p);
1451
1452 t = ip6gre_tunnel_locate(net, &p, 0);
1453
1454 if (t) {
1455 if (t->dev != dev)
1456 return -EEXIST;
1457 } else {
1458 t = nt;
1459 }
1460
1461 ip6gre_tunnel_unlink(ign, t);
1462 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1463 ip6gre_tunnel_link(ign, t);
1464 return 0;
1465 }
1466
1467 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1468 {
1469 struct net *net = dev_net(dev);
1470 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1471
1472 if (dev != ign->fb_tunnel_dev)
1473 unregister_netdevice_queue(dev, head);
1474 }
1475
1476 static size_t ip6gre_get_size(const struct net_device *dev)
1477 {
1478 return
1479 /* IFLA_GRE_LINK */
1480 nla_total_size(4) +
1481 /* IFLA_GRE_IFLAGS */
1482 nla_total_size(2) +
1483 /* IFLA_GRE_OFLAGS */
1484 nla_total_size(2) +
1485 /* IFLA_GRE_IKEY */
1486 nla_total_size(4) +
1487 /* IFLA_GRE_OKEY */
1488 nla_total_size(4) +
1489 /* IFLA_GRE_LOCAL */
1490 nla_total_size(sizeof(struct in6_addr)) +
1491 /* IFLA_GRE_REMOTE */
1492 nla_total_size(sizeof(struct in6_addr)) +
1493 /* IFLA_GRE_TTL */
1494 nla_total_size(1) +
1495 /* IFLA_GRE_ENCAP_LIMIT */
1496 nla_total_size(1) +
1497 /* IFLA_GRE_FLOWINFO */
1498 nla_total_size(4) +
1499 /* IFLA_GRE_FLAGS */
1500 nla_total_size(4) +
1501 /* IFLA_GRE_ENCAP_TYPE */
1502 nla_total_size(2) +
1503 /* IFLA_GRE_ENCAP_FLAGS */
1504 nla_total_size(2) +
1505 /* IFLA_GRE_ENCAP_SPORT */
1506 nla_total_size(2) +
1507 /* IFLA_GRE_ENCAP_DPORT */
1508 nla_total_size(2) +
1509 /* IFLA_GRE_FWMARK */
1510 nla_total_size(4) +
1511 0;
1512 }
1513
1514 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1515 {
1516 struct ip6_tnl *t = netdev_priv(dev);
1517 struct __ip6_tnl_parm *p = &t->parms;
1518
1519 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1520 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1521 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1522 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1523 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1524 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1525 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1526 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1527 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1528 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1529 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1530 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1531 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1532 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1533 goto nla_put_failure;
1534
1535 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1536 t->encap.type) ||
1537 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1538 t->encap.sport) ||
1539 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1540 t->encap.dport) ||
1541 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1542 t->encap.flags))
1543 goto nla_put_failure;
1544
1545 return 0;
1546
1547 nla_put_failure:
1548 return -EMSGSIZE;
1549 }
1550
1551 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1552 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1553 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1554 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1555 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1556 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1557 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1558 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1559 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1560 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1561 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1562 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1563 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1564 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1565 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1566 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1567 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1568 };
1569
1570 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1571 .kind = "ip6gre",
1572 .maxtype = IFLA_GRE_MAX,
1573 .policy = ip6gre_policy,
1574 .priv_size = sizeof(struct ip6_tnl),
1575 .setup = ip6gre_tunnel_setup,
1576 .validate = ip6gre_tunnel_validate,
1577 .newlink = ip6gre_newlink,
1578 .changelink = ip6gre_changelink,
1579 .dellink = ip6gre_dellink,
1580 .get_size = ip6gre_get_size,
1581 .fill_info = ip6gre_fill_info,
1582 .get_link_net = ip6_tnl_get_link_net,
1583 };
1584
1585 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1586 .kind = "ip6gretap",
1587 .maxtype = IFLA_GRE_MAX,
1588 .policy = ip6gre_policy,
1589 .priv_size = sizeof(struct ip6_tnl),
1590 .setup = ip6gre_tap_setup,
1591 .validate = ip6gre_tap_validate,
1592 .newlink = ip6gre_newlink,
1593 .changelink = ip6gre_changelink,
1594 .get_size = ip6gre_get_size,
1595 .fill_info = ip6gre_fill_info,
1596 .get_link_net = ip6_tnl_get_link_net,
1597 };
1598
1599 /*
1600 * And now the modules code and kernel interface.
1601 */
1602
1603 static int __init ip6gre_init(void)
1604 {
1605 int err;
1606
1607 pr_info("GRE over IPv6 tunneling driver\n");
1608
1609 err = register_pernet_device(&ip6gre_net_ops);
1610 if (err < 0)
1611 return err;
1612
1613 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1614 if (err < 0) {
1615 pr_info("%s: can't add protocol\n", __func__);
1616 goto add_proto_failed;
1617 }
1618
1619 err = rtnl_link_register(&ip6gre_link_ops);
1620 if (err < 0)
1621 goto rtnl_link_failed;
1622
1623 err = rtnl_link_register(&ip6gre_tap_ops);
1624 if (err < 0)
1625 goto tap_ops_failed;
1626
1627 out:
1628 return err;
1629
1630 tap_ops_failed:
1631 rtnl_link_unregister(&ip6gre_link_ops);
1632 rtnl_link_failed:
1633 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1634 add_proto_failed:
1635 unregister_pernet_device(&ip6gre_net_ops);
1636 goto out;
1637 }
1638
1639 static void __exit ip6gre_fini(void)
1640 {
1641 rtnl_link_unregister(&ip6gre_tap_ops);
1642 rtnl_link_unregister(&ip6gre_link_ops);
1643 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1644 unregister_pernet_device(&ip6gre_net_ops);
1645 }
1646
1647 module_init(ip6gre_init);
1648 module_exit(ip6gre_fini);
1649 MODULE_LICENSE("GPL");
1650 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1651 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1652 MODULE_ALIAS_RTNL_LINK("ip6gre");
1653 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1654 MODULE_ALIAS_NETDEV("ip6gre0");
Linux with added FPC-III support