?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
dm thin metadata: fix __udivdi3 undefined on 32-bit
[linux/fpc-iii.git] / net / ipv6 / ip6_gre.c
blobb25f4ad28b03276a27c7acc9d0421dc084596dc8
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/mroute.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
35 #include <linux/hash.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38
39 #include <net/sock.h>
40 #include <net/ip.h>
41 #include <net/ip_tunnels.h>
42 #include <net/icmp.h>
43 #include <net/protocol.h>
44 #include <net/addrconf.h>
45 #include <net/arp.h>
46 #include <net/checksum.h>
47 #include <net/dsfield.h>
48 #include <net/inet_ecn.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/rtnetlink.h>
53
54 #include <net/ipv6.h>
55 #include <net/ip6_fib.h>
56 #include <net/ip6_route.h>
57 #include <net/ip6_tunnel.h>
58 #include <net/gre.h>
59
60
61 static bool log_ecn_error = true;
62 module_param(log_ecn_error, bool, 0644);
63 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
64
65 #define HASH_SIZE_SHIFT 5
66 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
67
68 static int ip6gre_net_id __read_mostly;
69 struct ip6gre_net {
70 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
71
72 struct net_device *fb_tunnel_dev;
73 };
74
75 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
76 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
77 static int ip6gre_tunnel_init(struct net_device *dev);
78 static void ip6gre_tunnel_setup(struct net_device *dev);
79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
80 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85 4 hash tables:
86
87 3: (remote,local)
88 2: (remote,*)
89 1: (*,local)
90 0: (*,*)
91
92 We require exact key match i.e. if a key is present in packet
93 it will match only tunnel with the same key; if it is not present,
94 it will match only keyless tunnel.
95
96 All keysless packets, if not matched configured keyless tunnels
97 will match fallback tunnel.
98 */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103 u32 hash = ipv6_addr_hash(addr);
104
105 return hash_32(hash, HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l tunnels[3]
109 #define tunnels_r tunnels[2]
110 #define tunnels_l tunnels[1]
111 #define tunnels_wc tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116 const struct in6_addr *remote, const struct in6_addr *local,
117 __be32 key, __be16 gre_proto)
118 {
119 struct net *net = dev_net(dev);
120 int link = dev->ifindex;
121 unsigned int h0 = HASH_ADDR(remote);
122 unsigned int h1 = HASH_KEY(key);
123 struct ip6_tnl *t, *cand = NULL;
124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
126 ARPHRD_ETHER : ARPHRD_IP6GRE;
127 int score, cand_score = 4;
128
129 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
130 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
131 !ipv6_addr_equal(remote, &t->parms.raddr) ||
132 key != t->parms.i_key ||
133 !(t->dev->flags & IFF_UP))
134 continue;
135
136 if (t->dev->type != ARPHRD_IP6GRE &&
137 t->dev->type != dev_type)
138 continue;
139
140 score = 0;
141 if (t->parms.link != link)
142 score |= 1;
143 if (t->dev->type != dev_type)
144 score |= 2;
145 if (score == 0)
146 return t;
147
148 if (score < cand_score) {
149 cand = t;
150 cand_score = score;
151 }
152 }
153
154 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
155 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
156 key != t->parms.i_key ||
157 !(t->dev->flags & IFF_UP))
158 continue;
159
160 if (t->dev->type != ARPHRD_IP6GRE &&
161 t->dev->type != dev_type)
162 continue;
163
164 score = 0;
165 if (t->parms.link != link)
166 score |= 1;
167 if (t->dev->type != dev_type)
168 score |= 2;
169 if (score == 0)
170 return t;
171
172 if (score < cand_score) {
173 cand = t;
174 cand_score = score;
175 }
176 }
177
178 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
179 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
180 (!ipv6_addr_equal(local, &t->parms.raddr) ||
181 !ipv6_addr_is_multicast(local))) ||
182 key != t->parms.i_key ||
183 !(t->dev->flags & IFF_UP))
184 continue;
185
186 if (t->dev->type != ARPHRD_IP6GRE &&
187 t->dev->type != dev_type)
188 continue;
189
190 score = 0;
191 if (t->parms.link != link)
192 score |= 1;
193 if (t->dev->type != dev_type)
194 score |= 2;
195 if (score == 0)
196 return t;
197
198 if (score < cand_score) {
199 cand = t;
200 cand_score = score;
201 }
202 }
203
204 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
205 if (t->parms.i_key != key ||
206 !(t->dev->flags & IFF_UP))
207 continue;
208
209 if (t->dev->type != ARPHRD_IP6GRE &&
210 t->dev->type != dev_type)
211 continue;
212
213 score = 0;
214 if (t->parms.link != link)
215 score |= 1;
216 if (t->dev->type != dev_type)
217 score |= 2;
218 if (score == 0)
219 return t;
220
221 if (score < cand_score) {
222 cand = t;
223 cand_score = score;
224 }
225 }
226
227 if (cand)
228 return cand;
229
230 dev = ign->fb_tunnel_dev;
231 if (dev->flags & IFF_UP)
232 return netdev_priv(dev);
233
234 return NULL;
235 }
236
237 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
238 const struct __ip6_tnl_parm *p)
239 {
240 const struct in6_addr *remote = &p->raddr;
241 const struct in6_addr *local = &p->laddr;
242 unsigned int h = HASH_KEY(p->i_key);
243 int prio = 0;
244
245 if (!ipv6_addr_any(local))
246 prio |= 1;
247 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
248 prio |= 2;
249 h ^= HASH_ADDR(remote);
250 }
251
252 return &ign->tunnels[prio][h];
253 }
254
255 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
256 const struct ip6_tnl *t)
257 {
258 return __ip6gre_bucket(ign, &t->parms);
259 }
260
261 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
262 {
263 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
264
265 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
266 rcu_assign_pointer(*tp, t);
267 }
268
269 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
270 {
271 struct ip6_tnl __rcu **tp;
272 struct ip6_tnl *iter;
273
274 for (tp = ip6gre_bucket(ign, t);
275 (iter = rtnl_dereference(*tp)) != NULL;
276 tp = &iter->next) {
277 if (t == iter) {
278 rcu_assign_pointer(*tp, t->next);
279 break;
280 }
281 }
282 }
283
284 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
285 const struct __ip6_tnl_parm *parms,
286 int type)
287 {
288 const struct in6_addr *remote = &parms->raddr;
289 const struct in6_addr *local = &parms->laddr;
290 __be32 key = parms->i_key;
291 int link = parms->link;
292 struct ip6_tnl *t;
293 struct ip6_tnl __rcu **tp;
294 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
295
296 for (tp = __ip6gre_bucket(ign, parms);
297 (t = rtnl_dereference(*tp)) != NULL;
298 tp = &t->next)
299 if (ipv6_addr_equal(local, &t->parms.laddr) &&
300 ipv6_addr_equal(remote, &t->parms.raddr) &&
301 key == t->parms.i_key &&
302 link == t->parms.link &&
303 type == t->dev->type)
304 break;
305
306 return t;
307 }
308
309 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
310 const struct __ip6_tnl_parm *parms, int create)
311 {
312 struct ip6_tnl *t, *nt;
313 struct net_device *dev;
314 char name[IFNAMSIZ];
315 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
316
317 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
318 if (t && create)
319 return NULL;
320 if (t || !create)
321 return t;
322
323 if (parms->name[0]) {
324 if (!dev_valid_name(parms->name))
325 return NULL;
326 strlcpy(name, parms->name, IFNAMSIZ);
327 } else {
328 strcpy(name, "ip6gre%d");
329 }
330 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
331 ip6gre_tunnel_setup);
332 if (!dev)
333 return NULL;
334
335 dev_net_set(dev, net);
336
337 nt = netdev_priv(dev);
338 nt->parms = *parms;
339 dev->rtnl_link_ops = &ip6gre_link_ops;
340
341 nt->dev = dev;
342 nt->net = dev_net(dev);
343 ip6gre_tnl_link_config(nt, 1);
344
345 if (register_netdevice(dev) < 0)
346 goto failed_free;
347
348 /* Can use a lockless transmit, unless we generate output sequences */
349 if (!(nt->parms.o_flags & GRE_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;
442 if (mtu < IPV6_MIN_MTU)
443 mtu = IPV6_MIN_MTU;
444 t->dev->mtu = mtu;
445 return;
446 }
447
448 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
449 t->err_count++;
450 else
451 t->err_count = 1;
452 t->err_time = jiffies;
453 }
454
455 static int ip6gre_rcv(struct sk_buff *skb)
456 {
457 const struct ipv6hdr *ipv6h;
458 u8 *h;
459 __be16 flags;
460 __sum16 csum = 0;
461 __be32 key = 0;
462 u32 seqno = 0;
463 struct ip6_tnl *tunnel;
464 int offset = 4;
465 __be16 gre_proto;
466 int err;
467
468 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
469 goto drop;
470
471 ipv6h = ipv6_hdr(skb);
472 h = skb->data;
473 flags = *(__be16 *)h;
474
475 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
476 /* - Version must be 0.
477 - We do not support routing headers.
478 */
479 if (flags&(GRE_VERSION|GRE_ROUTING))
480 goto drop;
481
482 if (flags&GRE_CSUM) {
483 csum = skb_checksum_simple_validate(skb);
484 offset += 4;
485 }
486 if (flags&GRE_KEY) {
487 key = *(__be32 *)(h + offset);
488 offset += 4;
489 }
490 if (flags&GRE_SEQ) {
491 seqno = ntohl(*(__be32 *)(h + offset));
492 offset += 4;
493 }
494 }
495
496 gre_proto = *(__be16 *)(h + 2);
497
498 tunnel = ip6gre_tunnel_lookup(skb->dev,
499 &ipv6h->saddr, &ipv6h->daddr, key,
500 gre_proto);
501 if (tunnel) {
502 struct pcpu_sw_netstats *tstats;
503
504 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
505 goto drop;
506
507 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
508 tunnel->dev->stats.rx_dropped++;
509 goto drop;
510 }
511
512 skb->protocol = gre_proto;
513 /* WCCP version 1 and 2 protocol decoding.
514 * - Change protocol to IPv6
515 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
516 */
517 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
518 skb->protocol = htons(ETH_P_IPV6);
519 if ((*(h + offset) & 0xF0) != 0x40)
520 offset += 4;
521 }
522
523 skb->mac_header = skb->network_header;
524 __pskb_pull(skb, offset);
525 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
526
527 if (((flags&GRE_CSUM) && csum) ||
528 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
529 tunnel->dev->stats.rx_crc_errors++;
530 tunnel->dev->stats.rx_errors++;
531 goto drop;
532 }
533 if (tunnel->parms.i_flags&GRE_SEQ) {
534 if (!(flags&GRE_SEQ) ||
535 (tunnel->i_seqno &&
536 (s32)(seqno - tunnel->i_seqno) < 0)) {
537 tunnel->dev->stats.rx_fifo_errors++;
538 tunnel->dev->stats.rx_errors++;
539 goto drop;
540 }
541 tunnel->i_seqno = seqno + 1;
542 }
543
544 /* Warning: All skb pointers will be invalidated! */
545 if (tunnel->dev->type == ARPHRD_ETHER) {
546 if (!pskb_may_pull(skb, ETH_HLEN)) {
547 tunnel->dev->stats.rx_length_errors++;
548 tunnel->dev->stats.rx_errors++;
549 goto drop;
550 }
551
552 ipv6h = ipv6_hdr(skb);
553 skb->protocol = eth_type_trans(skb, tunnel->dev);
554 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
555 }
556
557 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
558
559 skb_reset_network_header(skb);
560
561 err = IP6_ECN_decapsulate(ipv6h, skb);
562 if (unlikely(err)) {
563 if (log_ecn_error)
564 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
565 &ipv6h->saddr,
566 ipv6_get_dsfield(ipv6h));
567 if (err > 1) {
568 ++tunnel->dev->stats.rx_frame_errors;
569 ++tunnel->dev->stats.rx_errors;
570 goto drop;
571 }
572 }
573
574 tstats = this_cpu_ptr(tunnel->dev->tstats);
575 u64_stats_update_begin(&tstats->syncp);
576 tstats->rx_packets++;
577 tstats->rx_bytes += skb->len;
578 u64_stats_update_end(&tstats->syncp);
579
580 netif_rx(skb);
581
582 return 0;
583 }
584 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
585
586 drop:
587 kfree_skb(skb);
588 return 0;
589 }
590
591 struct ipv6_tel_txoption {
592 struct ipv6_txoptions ops;
593 __u8 dst_opt[8];
594 };
595
596 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
597 {
598 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
599
600 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
601 opt->dst_opt[3] = 1;
602 opt->dst_opt[4] = encap_limit;
603 opt->dst_opt[5] = IPV6_TLV_PADN;
604 opt->dst_opt[6] = 1;
605
606 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
607 opt->ops.opt_nflen = 8;
608 }
609
610 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
611 struct net_device *dev,
612 __u8 dsfield,
613 struct flowi6 *fl6,
614 int encap_limit,
615 __u32 *pmtu)
616 {
617 struct ip6_tnl *tunnel = netdev_priv(dev);
618 struct net *net = tunnel->net;
619 struct net_device *tdev; /* Device to other host */
620 struct ipv6hdr *ipv6h; /* Our new IP header */
621 unsigned int max_headroom = 0; /* The extra header space needed */
622 int gre_hlen;
623 struct ipv6_tel_txoption opt;
624 int mtu;
625 struct dst_entry *dst = NULL, *ndst = NULL;
626 struct net_device_stats *stats = &tunnel->dev->stats;
627 int err = -1;
628 u8 proto;
629 struct sk_buff *new_skb;
630 __be16 protocol;
631
632 if (dev->type == ARPHRD_ETHER)
633 IPCB(skb)->flags = 0;
634
635 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
636 gre_hlen = 0;
637 ipv6h = (struct ipv6hdr *)skb->data;
638 fl6->daddr = ipv6h->daddr;
639 } else {
640 gre_hlen = tunnel->hlen;
641 fl6->daddr = tunnel->parms.raddr;
642 }
643
644 if (!fl6->flowi6_mark)
645 dst = dst_cache_get(&tunnel->dst_cache);
646
647 if (!dst) {
648 dst = ip6_route_output(net, NULL, fl6);
649
650 if (dst->error)
651 goto tx_err_link_failure;
652 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
653 if (IS_ERR(dst)) {
654 err = PTR_ERR(dst);
655 dst = NULL;
656 goto tx_err_link_failure;
657 }
658 ndst = dst;
659 }
660
661 tdev = dst->dev;
662
663 if (tdev == dev) {
664 stats->collisions++;
665 net_warn_ratelimited("%s: Local routing loop detected!\n",
666 tunnel->parms.name);
667 goto tx_err_dst_release;
668 }
669
670 mtu = dst_mtu(dst) - sizeof(*ipv6h);
671 if (encap_limit >= 0) {
672 max_headroom += 8;
673 mtu -= 8;
674 }
675 if (mtu < IPV6_MIN_MTU)
676 mtu = IPV6_MIN_MTU;
677 if (skb_dst(skb))
678 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
679 if (skb->len > mtu) {
680 *pmtu = mtu;
681 err = -EMSGSIZE;
682 goto tx_err_dst_release;
683 }
684
685 if (tunnel->err_count > 0) {
686 if (time_before(jiffies,
687 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
688 tunnel->err_count--;
689
690 dst_link_failure(skb);
691 } else
692 tunnel->err_count = 0;
693 }
694
695 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
696
697 max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
698
699 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
700 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
701 new_skb = skb_realloc_headroom(skb, max_headroom);
702 if (max_headroom > dev->needed_headroom)
703 dev->needed_headroom = max_headroom;
704 if (!new_skb)
705 goto tx_err_dst_release;
706
707 if (skb->sk)
708 skb_set_owner_w(new_skb, skb->sk);
709 consume_skb(skb);
710 skb = new_skb;
711 }
712
713 if (!fl6->flowi6_mark && ndst)
714 dst_cache_set_ip6(&tunnel->dst_cache, ndst, &fl6->saddr);
715 skb_dst_set(skb, dst);
716
717 proto = NEXTHDR_GRE;
718 if (encap_limit >= 0) {
719 init_tel_txopt(&opt, encap_limit);
720 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
721 }
722
723 if (likely(!skb->encapsulation)) {
724 skb_reset_inner_headers(skb);
725 skb->encapsulation = 1;
726 }
727
728 skb_push(skb, gre_hlen);
729 skb_reset_network_header(skb);
730 skb_set_transport_header(skb, sizeof(*ipv6h));
731
732 /*
733 * Push down and install the IP header.
734 */
735 ipv6h = ipv6_hdr(skb);
736 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
737 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
738 ipv6h->hop_limit = tunnel->parms.hop_limit;
739 ipv6h->nexthdr = proto;
740 ipv6h->saddr = fl6->saddr;
741 ipv6h->daddr = fl6->daddr;
742
743 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
744 protocol = (dev->type == ARPHRD_ETHER) ?
745 htons(ETH_P_TEB) : skb->protocol;
746 ((__be16 *)(ipv6h + 1))[1] = protocol;
747
748 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
749 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
750
751 if (tunnel->parms.o_flags&GRE_SEQ) {
752 ++tunnel->o_seqno;
753 *ptr = htonl(tunnel->o_seqno);
754 ptr--;
755 }
756 if (tunnel->parms.o_flags&GRE_KEY) {
757 *ptr = tunnel->parms.o_key;
758 ptr--;
759 }
760 if (tunnel->parms.o_flags&GRE_CSUM) {
761 *ptr = 0;
762 *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
763 skb->len - sizeof(struct ipv6hdr));
764 }
765 }
766
767 skb_set_inner_protocol(skb, protocol);
768
769 ip6tunnel_xmit(NULL, skb, dev);
770 return 0;
771 tx_err_link_failure:
772 stats->tx_carrier_errors++;
773 dst_link_failure(skb);
774 tx_err_dst_release:
775 dst_release(dst);
776 return err;
777 }
778
779 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
780 {
781 struct ip6_tnl *t = netdev_priv(dev);
782 const struct iphdr *iph = ip_hdr(skb);
783 int encap_limit = -1;
784 struct flowi6 fl6;
785 __u8 dsfield;
786 __u32 mtu;
787 int err;
788
789 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
790
791 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
792 encap_limit = t->parms.encap_limit;
793
794 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
795 fl6.flowi6_proto = IPPROTO_GRE;
796
797 dsfield = ipv4_get_dsfield(iph);
798
799 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
800 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
801 & IPV6_TCLASS_MASK;
802 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
803 fl6.flowi6_mark = skb->mark;
804
805 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
806 if (err != 0) {
807 /* XXX: send ICMP error even if DF is not set. */
808 if (err == -EMSGSIZE)
809 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
810 htonl(mtu));
811 return -1;
812 }
813
814 return 0;
815 }
816
817 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
818 {
819 struct ip6_tnl *t = netdev_priv(dev);
820 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
821 int encap_limit = -1;
822 __u16 offset;
823 struct flowi6 fl6;
824 __u8 dsfield;
825 __u32 mtu;
826 int err;
827
828 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
829 return -1;
830
831 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
832 if (offset > 0) {
833 struct ipv6_tlv_tnl_enc_lim *tel;
834 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
835 if (tel->encap_limit == 0) {
836 icmpv6_send(skb, ICMPV6_PARAMPROB,
837 ICMPV6_HDR_FIELD, offset + 2);
838 return -1;
839 }
840 encap_limit = tel->encap_limit - 1;
841 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
842 encap_limit = t->parms.encap_limit;
843
844 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
845 fl6.flowi6_proto = IPPROTO_GRE;
846
847 dsfield = ipv6_get_dsfield(ipv6h);
848 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
849 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
850 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
851 fl6.flowlabel |= ip6_flowlabel(ipv6h);
852 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
853 fl6.flowi6_mark = skb->mark;
854
855 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
856 if (err != 0) {
857 if (err == -EMSGSIZE)
858 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
859 return -1;
860 }
861
862 return 0;
863 }
864
865 /**
866 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
867 * @t: the outgoing tunnel device
868 * @hdr: IPv6 header from the incoming packet
869 *
870 * Description:
871 * Avoid trivial tunneling loop by checking that tunnel exit-point
872 * doesn't match source of incoming packet.
873 *
874 * Return:
875 * 1 if conflict,
876 * 0 else
877 **/
878
879 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
880 const struct ipv6hdr *hdr)
881 {
882 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
883 }
884
885 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
886 {
887 struct ip6_tnl *t = netdev_priv(dev);
888 int encap_limit = -1;
889 struct flowi6 fl6;
890 __u32 mtu;
891 int err;
892
893 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
894 encap_limit = t->parms.encap_limit;
895
896 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
897
898 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
899
900 return err;
901 }
902
903 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
904 struct net_device *dev)
905 {
906 struct ip6_tnl *t = netdev_priv(dev);
907 struct net_device_stats *stats = &t->dev->stats;
908 int ret;
909
910 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
911 goto tx_err;
912
913 switch (skb->protocol) {
914 case htons(ETH_P_IP):
915 ret = ip6gre_xmit_ipv4(skb, dev);
916 break;
917 case htons(ETH_P_IPV6):
918 ret = ip6gre_xmit_ipv6(skb, dev);
919 break;
920 default:
921 ret = ip6gre_xmit_other(skb, dev);
922 break;
923 }
924
925 if (ret < 0)
926 goto tx_err;
927
928 return NETDEV_TX_OK;
929
930 tx_err:
931 stats->tx_errors++;
932 stats->tx_dropped++;
933 kfree_skb(skb);
934 return NETDEV_TX_OK;
935 }
936
937 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
938 {
939 struct net_device *dev = t->dev;
940 struct __ip6_tnl_parm *p = &t->parms;
941 struct flowi6 *fl6 = &t->fl.u.ip6;
942 int addend = sizeof(struct ipv6hdr) + 4;
943
944 if (dev->type != ARPHRD_ETHER) {
945 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
946 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
947 }
948
949 /* Set up flowi template */
950 fl6->saddr = p->laddr;
951 fl6->daddr = p->raddr;
952 fl6->flowi6_oif = p->link;
953 fl6->flowlabel = 0;
954
955 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
956 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
957 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
958 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
959
960 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
961 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
962
963 if (p->flags&IP6_TNL_F_CAP_XMIT &&
964 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
965 dev->flags |= IFF_POINTOPOINT;
966 else
967 dev->flags &= ~IFF_POINTOPOINT;
968
969 /* Precalculate GRE options length */
970 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
971 if (t->parms.o_flags&GRE_CSUM)
972 addend += 4;
973 if (t->parms.o_flags&GRE_KEY)
974 addend += 4;
975 if (t->parms.o_flags&GRE_SEQ)
976 addend += 4;
977 }
978 t->hlen = addend;
979
980 if (p->flags & IP6_TNL_F_CAP_XMIT) {
981 int strict = (ipv6_addr_type(&p->raddr) &
982 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
983
984 struct rt6_info *rt = rt6_lookup(t->net,
985 &p->raddr, &p->laddr,
986 p->link, strict);
987
988 if (!rt)
989 return;
990
991 if (rt->dst.dev) {
992 dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
993
994 if (set_mtu) {
995 dev->mtu = rt->dst.dev->mtu - addend;
996 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
997 dev->mtu -= 8;
998
999 if (dev->mtu < IPV6_MIN_MTU)
1000 dev->mtu = IPV6_MIN_MTU;
1001 }
1002 }
1003 ip6_rt_put(rt);
1004 }
1005 }
1006
1007 static int ip6gre_tnl_change(struct ip6_tnl *t,
1008 const struct __ip6_tnl_parm *p, int set_mtu)
1009 {
1010 t->parms.laddr = p->laddr;
1011 t->parms.raddr = p->raddr;
1012 t->parms.flags = p->flags;
1013 t->parms.hop_limit = p->hop_limit;
1014 t->parms.encap_limit = p->encap_limit;
1015 t->parms.flowinfo = p->flowinfo;
1016 t->parms.link = p->link;
1017 t->parms.proto = p->proto;
1018 t->parms.i_key = p->i_key;
1019 t->parms.o_key = p->o_key;
1020 t->parms.i_flags = p->i_flags;
1021 t->parms.o_flags = p->o_flags;
1022 dst_cache_reset(&t->dst_cache);
1023 ip6gre_tnl_link_config(t, set_mtu);
1024 return 0;
1025 }
1026
1027 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1028 const struct ip6_tnl_parm2 *u)
1029 {
1030 p->laddr = u->laddr;
1031 p->raddr = u->raddr;
1032 p->flags = u->flags;
1033 p->hop_limit = u->hop_limit;
1034 p->encap_limit = u->encap_limit;
1035 p->flowinfo = u->flowinfo;
1036 p->link = u->link;
1037 p->i_key = u->i_key;
1038 p->o_key = u->o_key;
1039 p->i_flags = u->i_flags;
1040 p->o_flags = u->o_flags;
1041 memcpy(p->name, u->name, sizeof(u->name));
1042 }
1043
1044 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1045 const struct __ip6_tnl_parm *p)
1046 {
1047 u->proto = IPPROTO_GRE;
1048 u->laddr = p->laddr;
1049 u->raddr = p->raddr;
1050 u->flags = p->flags;
1051 u->hop_limit = p->hop_limit;
1052 u->encap_limit = p->encap_limit;
1053 u->flowinfo = p->flowinfo;
1054 u->link = p->link;
1055 u->i_key = p->i_key;
1056 u->o_key = p->o_key;
1057 u->i_flags = p->i_flags;
1058 u->o_flags = p->o_flags;
1059 memcpy(u->name, p->name, sizeof(u->name));
1060 }
1061
1062 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1063 struct ifreq *ifr, int cmd)
1064 {
1065 int err = 0;
1066 struct ip6_tnl_parm2 p;
1067 struct __ip6_tnl_parm p1;
1068 struct ip6_tnl *t = netdev_priv(dev);
1069 struct net *net = t->net;
1070 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1071
1072 switch (cmd) {
1073 case SIOCGETTUNNEL:
1074 if (dev == ign->fb_tunnel_dev) {
1075 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1076 err = -EFAULT;
1077 break;
1078 }
1079 ip6gre_tnl_parm_from_user(&p1, &p);
1080 t = ip6gre_tunnel_locate(net, &p1, 0);
1081 if (!t)
1082 t = netdev_priv(dev);
1083 }
1084 memset(&p, 0, sizeof(p));
1085 ip6gre_tnl_parm_to_user(&p, &t->parms);
1086 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1087 err = -EFAULT;
1088 break;
1089
1090 case SIOCADDTUNNEL:
1091 case SIOCCHGTUNNEL:
1092 err = -EPERM;
1093 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1094 goto done;
1095
1096 err = -EFAULT;
1097 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1098 goto done;
1099
1100 err = -EINVAL;
1101 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1102 goto done;
1103
1104 if (!(p.i_flags&GRE_KEY))
1105 p.i_key = 0;
1106 if (!(p.o_flags&GRE_KEY))
1107 p.o_key = 0;
1108
1109 ip6gre_tnl_parm_from_user(&p1, &p);
1110 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1111
1112 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1113 if (t) {
1114 if (t->dev != dev) {
1115 err = -EEXIST;
1116 break;
1117 }
1118 } else {
1119 t = netdev_priv(dev);
1120
1121 ip6gre_tunnel_unlink(ign, t);
1122 synchronize_net();
1123 ip6gre_tnl_change(t, &p1, 1);
1124 ip6gre_tunnel_link(ign, t);
1125 netdev_state_change(dev);
1126 }
1127 }
1128
1129 if (t) {
1130 err = 0;
1131
1132 memset(&p, 0, sizeof(p));
1133 ip6gre_tnl_parm_to_user(&p, &t->parms);
1134 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1135 err = -EFAULT;
1136 } else
1137 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1138 break;
1139
1140 case SIOCDELTUNNEL:
1141 err = -EPERM;
1142 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1143 goto done;
1144
1145 if (dev == ign->fb_tunnel_dev) {
1146 err = -EFAULT;
1147 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1148 goto done;
1149 err = -ENOENT;
1150 ip6gre_tnl_parm_from_user(&p1, &p);
1151 t = ip6gre_tunnel_locate(net, &p1, 0);
1152 if (!t)
1153 goto done;
1154 err = -EPERM;
1155 if (t == netdev_priv(ign->fb_tunnel_dev))
1156 goto done;
1157 dev = t->dev;
1158 }
1159 unregister_netdevice(dev);
1160 err = 0;
1161 break;
1162
1163 default:
1164 err = -EINVAL;
1165 }
1166
1167 done:
1168 return err;
1169 }
1170
1171 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1172 {
1173 if (new_mtu < 68 ||
1174 new_mtu > 0xFFF8 - dev->hard_header_len)
1175 return -EINVAL;
1176 dev->mtu = new_mtu;
1177 return 0;
1178 }
1179
1180 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1181 unsigned short type, const void *daddr,
1182 const void *saddr, unsigned int len)
1183 {
1184 struct ip6_tnl *t = netdev_priv(dev);
1185 struct ipv6hdr *ipv6h;
1186 __be16 *p;
1187
1188 ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen + sizeof(*ipv6h));
1189 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1190 t->fl.u.ip6.flowlabel,
1191 true, &t->fl.u.ip6));
1192 ipv6h->hop_limit = t->parms.hop_limit;
1193 ipv6h->nexthdr = NEXTHDR_GRE;
1194 ipv6h->saddr = t->parms.laddr;
1195 ipv6h->daddr = t->parms.raddr;
1196
1197 p = (__be16 *)(ipv6h + 1);
1198 p[0] = t->parms.o_flags;
1199 p[1] = htons(type);
1200
1201 /*
1202 * Set the source hardware address.
1203 */
1204
1205 if (saddr)
1206 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1207 if (daddr)
1208 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1209 if (!ipv6_addr_any(&ipv6h->daddr))
1210 return t->hlen;
1211
1212 return -t->hlen;
1213 }
1214
1215 static const struct header_ops ip6gre_header_ops = {
1216 .create = ip6gre_header,
1217 };
1218
1219 static const struct net_device_ops ip6gre_netdev_ops = {
1220 .ndo_init = ip6gre_tunnel_init,
1221 .ndo_uninit = ip6gre_tunnel_uninit,
1222 .ndo_start_xmit = ip6gre_tunnel_xmit,
1223 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1224 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1225 .ndo_get_stats64 = ip_tunnel_get_stats64,
1226 .ndo_get_iflink = ip6_tnl_get_iflink,
1227 };
1228
1229 static void ip6gre_dev_free(struct net_device *dev)
1230 {
1231 struct ip6_tnl *t = netdev_priv(dev);
1232
1233 dst_cache_destroy(&t->dst_cache);
1234 free_percpu(dev->tstats);
1235 free_netdev(dev);
1236 }
1237
1238 static void ip6gre_tunnel_setup(struct net_device *dev)
1239 {
1240 struct ip6_tnl *t;
1241
1242 dev->netdev_ops = &ip6gre_netdev_ops;
1243 dev->destructor = ip6gre_dev_free;
1244
1245 dev->type = ARPHRD_IP6GRE;
1246 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1247 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1248 t = netdev_priv(dev);
1249 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1250 dev->mtu -= 8;
1251 dev->flags |= IFF_NOARP;
1252 dev->addr_len = sizeof(struct in6_addr);
1253 netif_keep_dst(dev);
1254 }
1255
1256 static int ip6gre_tunnel_init_common(struct net_device *dev)
1257 {
1258 struct ip6_tnl *tunnel;
1259 int ret;
1260
1261 tunnel = netdev_priv(dev);
1262
1263 tunnel->dev = dev;
1264 tunnel->net = dev_net(dev);
1265 strcpy(tunnel->parms.name, dev->name);
1266
1267 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1268 if (!dev->tstats)
1269 return -ENOMEM;
1270
1271 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1272 if (ret) {
1273 free_percpu(dev->tstats);
1274 dev->tstats = NULL;
1275 return ret;
1276 }
1277
1278 return 0;
1279 }
1280
1281 static int ip6gre_tunnel_init(struct net_device *dev)
1282 {
1283 struct ip6_tnl *tunnel;
1284 int ret;
1285
1286 ret = ip6gre_tunnel_init_common(dev);
1287 if (ret)
1288 return ret;
1289
1290 tunnel = netdev_priv(dev);
1291
1292 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1293 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1294
1295 if (ipv6_addr_any(&tunnel->parms.raddr))
1296 dev->header_ops = &ip6gre_header_ops;
1297
1298 return 0;
1299 }
1300
1301 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1302 {
1303 struct ip6_tnl *tunnel = netdev_priv(dev);
1304
1305 tunnel->dev = dev;
1306 tunnel->net = dev_net(dev);
1307 strcpy(tunnel->parms.name, dev->name);
1308
1309 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1310
1311 dev_hold(dev);
1312 }
1313
1314
1315 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1316 .handler = ip6gre_rcv,
1317 .err_handler = ip6gre_err,
1318 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1319 };
1320
1321 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1322 {
1323 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1324 struct net_device *dev, *aux;
1325 int prio;
1326
1327 for_each_netdev_safe(net, dev, aux)
1328 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1329 dev->rtnl_link_ops == &ip6gre_tap_ops)
1330 unregister_netdevice_queue(dev, head);
1331
1332 for (prio = 0; prio < 4; prio++) {
1333 int h;
1334 for (h = 0; h < HASH_SIZE; h++) {
1335 struct ip6_tnl *t;
1336
1337 t = rtnl_dereference(ign->tunnels[prio][h]);
1338
1339 while (t) {
1340 /* If dev is in the same netns, it has already
1341 * been added to the list by the previous loop.
1342 */
1343 if (!net_eq(dev_net(t->dev), net))
1344 unregister_netdevice_queue(t->dev,
1345 head);
1346 t = rtnl_dereference(t->next);
1347 }
1348 }
1349 }
1350 }
1351
1352 static int __net_init ip6gre_init_net(struct net *net)
1353 {
1354 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1355 int err;
1356
1357 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1358 NET_NAME_UNKNOWN,
1359 ip6gre_tunnel_setup);
1360 if (!ign->fb_tunnel_dev) {
1361 err = -ENOMEM;
1362 goto err_alloc_dev;
1363 }
1364 dev_net_set(ign->fb_tunnel_dev, net);
1365 /* FB netdevice is special: we have one, and only one per netns.
1366 * Allowing to move it to another netns is clearly unsafe.
1367 */
1368 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1369
1370
1371 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1372 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1373
1374 err = register_netdev(ign->fb_tunnel_dev);
1375 if (err)
1376 goto err_reg_dev;
1377
1378 rcu_assign_pointer(ign->tunnels_wc[0],
1379 netdev_priv(ign->fb_tunnel_dev));
1380 return 0;
1381
1382 err_reg_dev:
1383 ip6gre_dev_free(ign->fb_tunnel_dev);
1384 err_alloc_dev:
1385 return err;
1386 }
1387
1388 static void __net_exit ip6gre_exit_net(struct net *net)
1389 {
1390 LIST_HEAD(list);
1391
1392 rtnl_lock();
1393 ip6gre_destroy_tunnels(net, &list);
1394 unregister_netdevice_many(&list);
1395 rtnl_unlock();
1396 }
1397
1398 static struct pernet_operations ip6gre_net_ops = {
1399 .init = ip6gre_init_net,
1400 .exit = ip6gre_exit_net,
1401 .id = &ip6gre_net_id,
1402 .size = sizeof(struct ip6gre_net),
1403 };
1404
1405 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1406 {
1407 __be16 flags;
1408
1409 if (!data)
1410 return 0;
1411
1412 flags = 0;
1413 if (data[IFLA_GRE_IFLAGS])
1414 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1415 if (data[IFLA_GRE_OFLAGS])
1416 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1417 if (flags & (GRE_VERSION|GRE_ROUTING))
1418 return -EINVAL;
1419
1420 return 0;
1421 }
1422
1423 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1424 {
1425 struct in6_addr daddr;
1426
1427 if (tb[IFLA_ADDRESS]) {
1428 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1429 return -EINVAL;
1430 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1431 return -EADDRNOTAVAIL;
1432 }
1433
1434 if (!data)
1435 goto out;
1436
1437 if (data[IFLA_GRE_REMOTE]) {
1438 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1439 if (ipv6_addr_any(&daddr))
1440 return -EINVAL;
1441 }
1442
1443 out:
1444 return ip6gre_tunnel_validate(tb, data);
1445 }
1446
1447
1448 static void ip6gre_netlink_parms(struct nlattr *data[],
1449 struct __ip6_tnl_parm *parms)
1450 {
1451 memset(parms, 0, sizeof(*parms));
1452
1453 if (!data)
1454 return;
1455
1456 if (data[IFLA_GRE_LINK])
1457 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1458
1459 if (data[IFLA_GRE_IFLAGS])
1460 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1461
1462 if (data[IFLA_GRE_OFLAGS])
1463 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1464
1465 if (data[IFLA_GRE_IKEY])
1466 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1467
1468 if (data[IFLA_GRE_OKEY])
1469 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1470
1471 if (data[IFLA_GRE_LOCAL])
1472 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1473
1474 if (data[IFLA_GRE_REMOTE])
1475 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1476
1477 if (data[IFLA_GRE_TTL])
1478 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1479
1480 if (data[IFLA_GRE_ENCAP_LIMIT])
1481 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1482
1483 if (data[IFLA_GRE_FLOWINFO])
1484 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1485
1486 if (data[IFLA_GRE_FLAGS])
1487 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1488 }
1489
1490 static int ip6gre_tap_init(struct net_device *dev)
1491 {
1492 struct ip6_tnl *tunnel;
1493 int ret;
1494
1495 ret = ip6gre_tunnel_init_common(dev);
1496 if (ret)
1497 return ret;
1498
1499 tunnel = netdev_priv(dev);
1500
1501 ip6gre_tnl_link_config(tunnel, 1);
1502
1503 return 0;
1504 }
1505
1506 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1507 .ndo_init = ip6gre_tap_init,
1508 .ndo_uninit = ip6gre_tunnel_uninit,
1509 .ndo_start_xmit = ip6gre_tunnel_xmit,
1510 .ndo_set_mac_address = eth_mac_addr,
1511 .ndo_validate_addr = eth_validate_addr,
1512 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1513 .ndo_get_stats64 = ip_tunnel_get_stats64,
1514 .ndo_get_iflink = ip6_tnl_get_iflink,
1515 };
1516
1517 static void ip6gre_tap_setup(struct net_device *dev)
1518 {
1519
1520 ether_setup(dev);
1521
1522 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1523 dev->destructor = ip6gre_dev_free;
1524
1525 dev->features |= NETIF_F_NETNS_LOCAL;
1526 }
1527
1528 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1529 struct nlattr *tb[], struct nlattr *data[])
1530 {
1531 struct ip6_tnl *nt;
1532 struct net *net = dev_net(dev);
1533 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1534 int err;
1535
1536 nt = netdev_priv(dev);
1537 ip6gre_netlink_parms(data, &nt->parms);
1538
1539 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1540 return -EEXIST;
1541
1542 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1543 eth_hw_addr_random(dev);
1544
1545 nt->dev = dev;
1546 nt->net = dev_net(dev);
1547 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1548
1549 /* Can use a lockless transmit, unless we generate output sequences */
1550 if (!(nt->parms.o_flags & GRE_SEQ))
1551 dev->features |= NETIF_F_LLTX;
1552
1553 err = register_netdevice(dev);
1554 if (err)
1555 goto out;
1556
1557 dev_hold(dev);
1558 ip6gre_tunnel_link(ign, nt);
1559
1560 out:
1561 return err;
1562 }
1563
1564 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1565 struct nlattr *data[])
1566 {
1567 struct ip6_tnl *t, *nt = netdev_priv(dev);
1568 struct net *net = nt->net;
1569 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1570 struct __ip6_tnl_parm p;
1571
1572 if (dev == ign->fb_tunnel_dev)
1573 return -EINVAL;
1574
1575 ip6gre_netlink_parms(data, &p);
1576
1577 t = ip6gre_tunnel_locate(net, &p, 0);
1578
1579 if (t) {
1580 if (t->dev != dev)
1581 return -EEXIST;
1582 } else {
1583 t = nt;
1584 }
1585
1586 ip6gre_tunnel_unlink(ign, t);
1587 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1588 ip6gre_tunnel_link(ign, t);
1589 return 0;
1590 }
1591
1592 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1593 {
1594 struct net *net = dev_net(dev);
1595 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1596
1597 if (dev != ign->fb_tunnel_dev)
1598 unregister_netdevice_queue(dev, head);
1599 }
1600
1601 static size_t ip6gre_get_size(const struct net_device *dev)
1602 {
1603 return
1604 /* IFLA_GRE_LINK */
1605 nla_total_size(4) +
1606 /* IFLA_GRE_IFLAGS */
1607 nla_total_size(2) +
1608 /* IFLA_GRE_OFLAGS */
1609 nla_total_size(2) +
1610 /* IFLA_GRE_IKEY */
1611 nla_total_size(4) +
1612 /* IFLA_GRE_OKEY */
1613 nla_total_size(4) +
1614 /* IFLA_GRE_LOCAL */
1615 nla_total_size(sizeof(struct in6_addr)) +
1616 /* IFLA_GRE_REMOTE */
1617 nla_total_size(sizeof(struct in6_addr)) +
1618 /* IFLA_GRE_TTL */
1619 nla_total_size(1) +
1620 /* IFLA_GRE_TOS */
1621 nla_total_size(1) +
1622 /* IFLA_GRE_ENCAP_LIMIT */
1623 nla_total_size(1) +
1624 /* IFLA_GRE_FLOWINFO */
1625 nla_total_size(4) +
1626 /* IFLA_GRE_FLAGS */
1627 nla_total_size(4) +
1628 0;
1629 }
1630
1631 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1632 {
1633 struct ip6_tnl *t = netdev_priv(dev);
1634 struct __ip6_tnl_parm *p = &t->parms;
1635
1636 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1637 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1638 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1639 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1640 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1641 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1642 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1643 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1644 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1645 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1646 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1647 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1648 goto nla_put_failure;
1649 return 0;
1650
1651 nla_put_failure:
1652 return -EMSGSIZE;
1653 }
1654
1655 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1656 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1657 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1658 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1659 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1660 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1661 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1662 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1663 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1664 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1665 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1666 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1667 };
1668
1669 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1670 .kind = "ip6gre",
1671 .maxtype = IFLA_GRE_MAX,
1672 .policy = ip6gre_policy,
1673 .priv_size = sizeof(struct ip6_tnl),
1674 .setup = ip6gre_tunnel_setup,
1675 .validate = ip6gre_tunnel_validate,
1676 .newlink = ip6gre_newlink,
1677 .changelink = ip6gre_changelink,
1678 .dellink = ip6gre_dellink,
1679 .get_size = ip6gre_get_size,
1680 .fill_info = ip6gre_fill_info,
1681 .get_link_net = ip6_tnl_get_link_net,
1682 };
1683
1684 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1685 .kind = "ip6gretap",
1686 .maxtype = IFLA_GRE_MAX,
1687 .policy = ip6gre_policy,
1688 .priv_size = sizeof(struct ip6_tnl),
1689 .setup = ip6gre_tap_setup,
1690 .validate = ip6gre_tap_validate,
1691 .newlink = ip6gre_newlink,
1692 .changelink = ip6gre_changelink,
1693 .get_size = ip6gre_get_size,
1694 .fill_info = ip6gre_fill_info,
1695 .get_link_net = ip6_tnl_get_link_net,
1696 };
1697
1698 /*
1699 * And now the modules code and kernel interface.
1700 */
1701
1702 static int __init ip6gre_init(void)
1703 {
1704 int err;
1705
1706 pr_info("GRE over IPv6 tunneling driver\n");
1707
1708 err = register_pernet_device(&ip6gre_net_ops);
1709 if (err < 0)
1710 return err;
1711
1712 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1713 if (err < 0) {
1714 pr_info("%s: can't add protocol\n", __func__);
1715 goto add_proto_failed;
1716 }
1717
1718 err = rtnl_link_register(&ip6gre_link_ops);
1719 if (err < 0)
1720 goto rtnl_link_failed;
1721
1722 err = rtnl_link_register(&ip6gre_tap_ops);
1723 if (err < 0)
1724 goto tap_ops_failed;
1725
1726 out:
1727 return err;
1728
1729 tap_ops_failed:
1730 rtnl_link_unregister(&ip6gre_link_ops);
1731 rtnl_link_failed:
1732 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1733 add_proto_failed:
1734 unregister_pernet_device(&ip6gre_net_ops);
1735 goto out;
1736 }
1737
1738 static void __exit ip6gre_fini(void)
1739 {
1740 rtnl_link_unregister(&ip6gre_tap_ops);
1741 rtnl_link_unregister(&ip6gre_link_ops);
1742 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1743 unregister_pernet_device(&ip6gre_net_ops);
1744 }
1745
1746 module_init(ip6gre_init);
1747 module_exit(ip6gre_fini);
1748 MODULE_LICENSE("GPL");
1749 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1750 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1751 MODULE_ALIAS_RTNL_LINK("ip6gre");
1752 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1753 MODULE_ALIAS_NETDEV("ip6gre0");
Linux with added FPC-III support