projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'acpi-assorted'
[deliverable/linux.git] / net / ipv6 / ip6_gre.c
1 /*
2 * GRE over IPv6 protocol decoder.
3 *
4 * Authors: Dmitry Kozlov (xeb@mail.ru)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/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
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 IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
65 #define IPV6_TCLASS_SHIFT 20
66
67 #define HASH_SIZE_SHIFT 5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
69
70 static int ip6gre_net_id __read_mostly;
71 struct ip6gre_net {
72 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
73
74 struct net_device *fb_tunnel_dev;
75 };
76
77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
78 static int ip6gre_tunnel_init(struct net_device *dev);
79 static void ip6gre_tunnel_setup(struct net_device *dev);
80 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
81 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
82
83 /* Tunnel hash table */
84
85 /*
86 4 hash tables:
87
88 3: (remote,local)
89 2: (remote,*)
90 1: (*,local)
91 0: (*,*)
92
93 We require exact key match i.e. if a key is present in packet
94 it will match only tunnel with the same key; if it is not present,
95 it will match only keyless tunnel.
96
97 All keysless packets, if not matched configured keyless tunnels
98 will match fallback tunnel.
99 */
100
101 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
102 static u32 HASH_ADDR(const struct in6_addr *addr)
103 {
104 u32 hash = ipv6_addr_hash(addr);
105
106 return hash_32(hash, HASH_SIZE_SHIFT);
107 }
108
109 #define tunnels_r_l tunnels[3]
110 #define tunnels_r tunnels[2]
111 #define tunnels_l tunnels[1]
112 #define tunnels_wc tunnels[0]
113
114 /* Given src, dst and key, find appropriate for input tunnel. */
115
116 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
117 const struct in6_addr *remote, const struct in6_addr *local,
118 __be32 key, __be16 gre_proto)
119 {
120 struct net *net = dev_net(dev);
121 int link = dev->ifindex;
122 unsigned int h0 = HASH_ADDR(remote);
123 unsigned int h1 = HASH_KEY(key);
124 struct ip6_tnl *t, *cand = NULL;
125 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
126 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
127 ARPHRD_ETHER : ARPHRD_IP6GRE;
128 int score, cand_score = 4;
129
130 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
131 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
132 !ipv6_addr_equal(remote, &t->parms.raddr) ||
133 key != t->parms.i_key ||
134 !(t->dev->flags & IFF_UP))
135 continue;
136
137 if (t->dev->type != ARPHRD_IP6GRE &&
138 t->dev->type != dev_type)
139 continue;
140
141 score = 0;
142 if (t->parms.link != link)
143 score |= 1;
144 if (t->dev->type != dev_type)
145 score |= 2;
146 if (score == 0)
147 return t;
148
149 if (score < cand_score) {
150 cand = t;
151 cand_score = score;
152 }
153 }
154
155 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
156 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
157 key != t->parms.i_key ||
158 !(t->dev->flags & IFF_UP))
159 continue;
160
161 if (t->dev->type != ARPHRD_IP6GRE &&
162 t->dev->type != dev_type)
163 continue;
164
165 score = 0;
166 if (t->parms.link != link)
167 score |= 1;
168 if (t->dev->type != dev_type)
169 score |= 2;
170 if (score == 0)
171 return t;
172
173 if (score < cand_score) {
174 cand = t;
175 cand_score = score;
176 }
177 }
178
179 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
180 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
181 (!ipv6_addr_equal(local, &t->parms.raddr) ||
182 !ipv6_addr_is_multicast(local))) ||
183 key != t->parms.i_key ||
184 !(t->dev->flags & IFF_UP))
185 continue;
186
187 if (t->dev->type != ARPHRD_IP6GRE &&
188 t->dev->type != dev_type)
189 continue;
190
191 score = 0;
192 if (t->parms.link != link)
193 score |= 1;
194 if (t->dev->type != dev_type)
195 score |= 2;
196 if (score == 0)
197 return t;
198
199 if (score < cand_score) {
200 cand = t;
201 cand_score = score;
202 }
203 }
204
205 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
206 if (t->parms.i_key != key ||
207 !(t->dev->flags & IFF_UP))
208 continue;
209
210 if (t->dev->type != ARPHRD_IP6GRE &&
211 t->dev->type != dev_type)
212 continue;
213
214 score = 0;
215 if (t->parms.link != link)
216 score |= 1;
217 if (t->dev->type != dev_type)
218 score |= 2;
219 if (score == 0)
220 return t;
221
222 if (score < cand_score) {
223 cand = t;
224 cand_score = score;
225 }
226 }
227
228 if (cand != NULL)
229 return cand;
230
231 dev = ign->fb_tunnel_dev;
232 if (dev->flags & IFF_UP)
233 return netdev_priv(dev);
234
235 return NULL;
236 }
237
238 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
239 const struct __ip6_tnl_parm *p)
240 {
241 const struct in6_addr *remote = &p->raddr;
242 const struct in6_addr *local = &p->laddr;
243 unsigned int h = HASH_KEY(p->i_key);
244 int prio = 0;
245
246 if (!ipv6_addr_any(local))
247 prio |= 1;
248 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
249 prio |= 2;
250 h ^= HASH_ADDR(remote);
251 }
252
253 return &ign->tunnels[prio][h];
254 }
255
256 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
257 const struct ip6_tnl *t)
258 {
259 return __ip6gre_bucket(ign, &t->parms);
260 }
261
262 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
263 {
264 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
265
266 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
267 rcu_assign_pointer(*tp, t);
268 }
269
270 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
271 {
272 struct ip6_tnl __rcu **tp;
273 struct ip6_tnl *iter;
274
275 for (tp = ip6gre_bucket(ign, t);
276 (iter = rtnl_dereference(*tp)) != NULL;
277 tp = &iter->next) {
278 if (t == iter) {
279 rcu_assign_pointer(*tp, t->next);
280 break;
281 }
282 }
283 }
284
285 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
286 const struct __ip6_tnl_parm *parms,
287 int type)
288 {
289 const struct in6_addr *remote = &parms->raddr;
290 const struct in6_addr *local = &parms->laddr;
291 __be32 key = parms->i_key;
292 int link = parms->link;
293 struct ip6_tnl *t;
294 struct ip6_tnl __rcu **tp;
295 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
296
297 for (tp = __ip6gre_bucket(ign, parms);
298 (t = rtnl_dereference(*tp)) != NULL;
299 tp = &t->next)
300 if (ipv6_addr_equal(local, &t->parms.laddr) &&
301 ipv6_addr_equal(remote, &t->parms.raddr) &&
302 key == t->parms.i_key &&
303 link == t->parms.link &&
304 type == t->dev->type)
305 break;
306
307 return t;
308 }
309
310 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
311 const struct __ip6_tnl_parm *parms, int create)
312 {
313 struct ip6_tnl *t, *nt;
314 struct net_device *dev;
315 char name[IFNAMSIZ];
316 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
317
318 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
319 if (t || !create)
320 return t;
321
322 if (parms->name[0])
323 strlcpy(name, parms->name, IFNAMSIZ);
324 else
325 strcpy(name, "ip6gre%d");
326
327 dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
328 if (!dev)
329 return NULL;
330
331 dev_net_set(dev, net);
332
333 nt = netdev_priv(dev);
334 nt->parms = *parms;
335 dev->rtnl_link_ops = &ip6gre_link_ops;
336
337 nt->dev = dev;
338 ip6gre_tnl_link_config(nt, 1);
339
340 if (register_netdevice(dev) < 0)
341 goto failed_free;
342
343 /* Can use a lockless transmit, unless we generate output sequences */
344 if (!(nt->parms.o_flags & GRE_SEQ))
345 dev->features |= NETIF_F_LLTX;
346
347 dev_hold(dev);
348 ip6gre_tunnel_link(ign, nt);
349 return nt;
350
351 failed_free:
352 free_netdev(dev);
353 return NULL;
354 }
355
356 static void ip6gre_tunnel_uninit(struct net_device *dev)
357 {
358 struct net *net = dev_net(dev);
359 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
360
361 ip6gre_tunnel_unlink(ign, netdev_priv(dev));
362 dev_put(dev);
363 }
364
365
366 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
367 u8 type, u8 code, int offset, __be32 info)
368 {
369 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
370 __be16 *p = (__be16 *)(skb->data + offset);
371 int grehlen = offset + 4;
372 struct ip6_tnl *t;
373 __be16 flags;
374
375 flags = p[0];
376 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
377 if (flags&(GRE_VERSION|GRE_ROUTING))
378 return;
379 if (flags&GRE_KEY) {
380 grehlen += 4;
381 if (flags&GRE_CSUM)
382 grehlen += 4;
383 }
384 }
385
386 /* If only 8 bytes returned, keyed message will be dropped here */
387 if (!pskb_may_pull(skb, grehlen))
388 return;
389 ipv6h = (const struct ipv6hdr *)skb->data;
390 p = (__be16 *)(skb->data + offset);
391
392 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
393 flags & GRE_KEY ?
394 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
395 p[1]);
396 if (t == NULL)
397 return;
398
399 switch (type) {
400 __u32 teli;
401 struct ipv6_tlv_tnl_enc_lim *tel;
402 __u32 mtu;
403 case ICMPV6_DEST_UNREACH:
404 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
405 t->parms.name);
406 break;
407 case ICMPV6_TIME_EXCEED:
408 if (code == ICMPV6_EXC_HOPLIMIT) {
409 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
410 t->parms.name);
411 }
412 break;
413 case ICMPV6_PARAMPROB:
414 teli = 0;
415 if (code == ICMPV6_HDR_FIELD)
416 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
417
418 if (teli && teli == info - 2) {
419 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
420 if (tel->encap_limit == 0) {
421 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
422 t->parms.name);
423 }
424 } else {
425 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
426 t->parms.name);
427 }
428 break;
429 case ICMPV6_PKT_TOOBIG:
430 mtu = info - offset;
431 if (mtu < IPV6_MIN_MTU)
432 mtu = IPV6_MIN_MTU;
433 t->dev->mtu = mtu;
434 break;
435 }
436
437 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
438 t->err_count++;
439 else
440 t->err_count = 1;
441 t->err_time = jiffies;
442 }
443
444 static int ip6gre_rcv(struct sk_buff *skb)
445 {
446 const struct ipv6hdr *ipv6h;
447 u8 *h;
448 __be16 flags;
449 __sum16 csum = 0;
450 __be32 key = 0;
451 u32 seqno = 0;
452 struct ip6_tnl *tunnel;
453 int offset = 4;
454 __be16 gre_proto;
455 int err;
456
457 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
458 goto drop;
459
460 ipv6h = ipv6_hdr(skb);
461 h = skb->data;
462 flags = *(__be16 *)h;
463
464 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
465 /* - Version must be 0.
466 - We do not support routing headers.
467 */
468 if (flags&(GRE_VERSION|GRE_ROUTING))
469 goto drop;
470
471 if (flags&GRE_CSUM) {
472 switch (skb->ip_summed) {
473 case CHECKSUM_COMPLETE:
474 csum = csum_fold(skb->csum);
475 if (!csum)
476 break;
477 /* fall through */
478 case CHECKSUM_NONE:
479 skb->csum = 0;
480 csum = __skb_checksum_complete(skb);
481 skb->ip_summed = CHECKSUM_COMPLETE;
482 }
483 offset += 4;
484 }
485 if (flags&GRE_KEY) {
486 key = *(__be32 *)(h + offset);
487 offset += 4;
488 }
489 if (flags&GRE_SEQ) {
490 seqno = ntohl(*(__be32 *)(h + offset));
491 offset += 4;
492 }
493 }
494
495 gre_proto = *(__be16 *)(h + 2);
496
497 tunnel = ip6gre_tunnel_lookup(skb->dev,
498 &ipv6h->saddr, &ipv6h->daddr, key,
499 gre_proto);
500 if (tunnel) {
501 struct pcpu_tstats *tstats;
502
503 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
504 goto drop;
505
506 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
507 tunnel->dev->stats.rx_dropped++;
508 goto drop;
509 }
510
511 secpath_reset(skb);
512
513 skb->protocol = gre_proto;
514 /* WCCP version 1 and 2 protocol decoding.
515 * - Change protocol to IP
516 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
517 */
518 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
519 skb->protocol = htons(ETH_P_IP);
520 if ((*(h + offset) & 0xF0) != 0x40)
521 offset += 4;
522 }
523
524 skb->mac_header = skb->network_header;
525 __pskb_pull(skb, offset);
526 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
527 skb->pkt_type = PACKET_HOST;
528
529 if (((flags&GRE_CSUM) && csum) ||
530 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
531 tunnel->dev->stats.rx_crc_errors++;
532 tunnel->dev->stats.rx_errors++;
533 goto drop;
534 }
535 if (tunnel->parms.i_flags&GRE_SEQ) {
536 if (!(flags&GRE_SEQ) ||
537 (tunnel->i_seqno &&
538 (s32)(seqno - tunnel->i_seqno) < 0)) {
539 tunnel->dev->stats.rx_fifo_errors++;
540 tunnel->dev->stats.rx_errors++;
541 goto drop;
542 }
543 tunnel->i_seqno = seqno + 1;
544 }
545
546 /* Warning: All skb pointers will be invalidated! */
547 if (tunnel->dev->type == ARPHRD_ETHER) {
548 if (!pskb_may_pull(skb, ETH_HLEN)) {
549 tunnel->dev->stats.rx_length_errors++;
550 tunnel->dev->stats.rx_errors++;
551 goto drop;
552 }
553
554 ipv6h = ipv6_hdr(skb);
555 skb->protocol = eth_type_trans(skb, tunnel->dev);
556 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
557 }
558
559 __skb_tunnel_rx(skb, tunnel->dev);
560
561 skb_reset_network_header(skb);
562
563 err = IP6_ECN_decapsulate(ipv6h, skb);
564 if (unlikely(err)) {
565 if (log_ecn_error)
566 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
567 &ipv6h->saddr,
568 ipv6_get_dsfield(ipv6h));
569 if (err > 1) {
570 ++tunnel->dev->stats.rx_frame_errors;
571 ++tunnel->dev->stats.rx_errors;
572 goto drop;
573 }
574 }
575
576 tstats = this_cpu_ptr(tunnel->dev->tstats);
577 u64_stats_update_begin(&tstats->syncp);
578 tstats->rx_packets++;
579 tstats->rx_bytes += skb->len;
580 u64_stats_update_end(&tstats->syncp);
581
582 netif_rx(skb);
583
584 return 0;
585 }
586 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
587
588 drop:
589 kfree_skb(skb);
590 return 0;
591 }
592
593 struct ipv6_tel_txoption {
594 struct ipv6_txoptions ops;
595 __u8 dst_opt[8];
596 };
597
598 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
599 {
600 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
601
602 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
603 opt->dst_opt[3] = 1;
604 opt->dst_opt[4] = encap_limit;
605 opt->dst_opt[5] = IPV6_TLV_PADN;
606 opt->dst_opt[6] = 1;
607
608 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
609 opt->ops.opt_nflen = 8;
610 }
611
612 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
613 struct net_device *dev,
614 __u8 dsfield,
615 struct flowi6 *fl6,
616 int encap_limit,
617 __u32 *pmtu)
618 {
619 struct net *net = dev_net(dev);
620 struct ip6_tnl *tunnel = netdev_priv(dev);
621 struct net_device *tdev; /* Device to other host */
622 struct ipv6hdr *ipv6h; /* Our new IP header */
623 unsigned int max_headroom; /* The extra header space needed */
624 int gre_hlen;
625 struct ipv6_tel_txoption opt;
626 int mtu;
627 struct dst_entry *dst = NULL, *ndst = NULL;
628 struct net_device_stats *stats = &tunnel->dev->stats;
629 int err = -1;
630 u8 proto;
631 struct sk_buff *new_skb;
632
633 if (dev->type == ARPHRD_ETHER)
634 IPCB(skb)->flags = 0;
635
636 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
637 gre_hlen = 0;
638 ipv6h = (struct ipv6hdr *)skb->data;
639 fl6->daddr = ipv6h->daddr;
640 } else {
641 gre_hlen = tunnel->hlen;
642 fl6->daddr = tunnel->parms.raddr;
643 }
644
645 if (!fl6->flowi6_mark)
646 dst = ip6_tnl_dst_check(tunnel);
647
648 if (!dst) {
649 ndst = ip6_route_output(net, NULL, fl6);
650
651 if (ndst->error)
652 goto tx_err_link_failure;
653 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
654 if (IS_ERR(ndst)) {
655 err = PTR_ERR(ndst);
656 ndst = NULL;
657 goto tx_err_link_failure;
658 }
659 dst = ndst;
660 }
661
662 tdev = dst->dev;
663
664 if (tdev == dev) {
665 stats->collisions++;
666 net_warn_ratelimited("%s: Local routing loop detected!\n",
667 tunnel->parms.name);
668 goto tx_err_dst_release;
669 }
670
671 mtu = dst_mtu(dst) - sizeof(*ipv6h);
672 if (encap_limit >= 0) {
673 max_headroom += 8;
674 mtu -= 8;
675 }
676 if (mtu < IPV6_MIN_MTU)
677 mtu = IPV6_MIN_MTU;
678 if (skb_dst(skb))
679 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
680 if (skb->len > mtu) {
681 *pmtu = mtu;
682 err = -EMSGSIZE;
683 goto tx_err_dst_release;
684 }
685
686 if (tunnel->err_count > 0) {
687 if (time_before(jiffies,
688 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
689 tunnel->err_count--;
690
691 dst_link_failure(skb);
692 } else
693 tunnel->err_count = 0;
694 }
695
696 max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
697
698 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
699 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
700 new_skb = skb_realloc_headroom(skb, max_headroom);
701 if (max_headroom > dev->needed_headroom)
702 dev->needed_headroom = max_headroom;
703 if (!new_skb)
704 goto tx_err_dst_release;
705
706 if (skb->sk)
707 skb_set_owner_w(new_skb, skb->sk);
708 consume_skb(skb);
709 skb = new_skb;
710 }
711
712 skb_dst_drop(skb);
713
714 if (fl6->flowi6_mark) {
715 skb_dst_set(skb, dst);
716 ndst = NULL;
717 } else {
718 skb_dst_set_noref(skb, dst);
719 }
720
721 proto = NEXTHDR_GRE;
722 if (encap_limit >= 0) {
723 init_tel_txopt(&opt, encap_limit);
724 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
725 }
726
727 if (likely(!skb->encapsulation)) {
728 skb_reset_inner_headers(skb);
729 skb->encapsulation = 1;
730 }
731
732 skb_push(skb, gre_hlen);
733 skb_reset_network_header(skb);
734 skb_set_transport_header(skb, sizeof(*ipv6h));
735
736 /*
737 * Push down and install the IP header.
738 */
739 ipv6h = ipv6_hdr(skb);
740 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
741 ipv6h->hop_limit = tunnel->parms.hop_limit;
742 ipv6h->nexthdr = proto;
743 ipv6h->saddr = fl6->saddr;
744 ipv6h->daddr = fl6->daddr;
745
746 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
747 ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
748 htons(ETH_P_TEB) : skb->protocol;
749
750 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
751 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
752
753 if (tunnel->parms.o_flags&GRE_SEQ) {
754 ++tunnel->o_seqno;
755 *ptr = htonl(tunnel->o_seqno);
756 ptr--;
757 }
758 if (tunnel->parms.o_flags&GRE_KEY) {
759 *ptr = tunnel->parms.o_key;
760 ptr--;
761 }
762 if (tunnel->parms.o_flags&GRE_CSUM) {
763 *ptr = 0;
764 *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
765 skb->len - sizeof(struct ipv6hdr));
766 }
767 }
768
769 ip6tunnel_xmit(skb, dev);
770 if (ndst)
771 ip6_tnl_dst_store(tunnel, ndst);
772 return 0;
773 tx_err_link_failure:
774 stats->tx_carrier_errors++;
775 dst_link_failure(skb);
776 tx_err_dst_release:
777 dst_release(ndst);
778 return err;
779 }
780
781 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
782 {
783 struct ip6_tnl *t = netdev_priv(dev);
784 const struct iphdr *iph = ip_hdr(skb);
785 int encap_limit = -1;
786 struct flowi6 fl6;
787 __u8 dsfield;
788 __u32 mtu;
789 int err;
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_IPIP;
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_IPV6;
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 |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
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 fl6.flowi6_proto = skb->protocol;
898
899 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
900
901 return err;
902 }
903
904 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
905 struct net_device *dev)
906 {
907 struct ip6_tnl *t = netdev_priv(dev);
908 struct net_device_stats *stats = &t->dev->stats;
909 int ret;
910
911 if (!ip6_tnl_xmit_ctl(t))
912 goto tx_err;
913
914 switch (skb->protocol) {
915 case htons(ETH_P_IP):
916 ret = ip6gre_xmit_ipv4(skb, dev);
917 break;
918 case htons(ETH_P_IPV6):
919 ret = ip6gre_xmit_ipv6(skb, dev);
920 break;
921 default:
922 ret = ip6gre_xmit_other(skb, dev);
923 break;
924 }
925
926 if (ret < 0)
927 goto tx_err;
928
929 return NETDEV_TX_OK;
930
931 tx_err:
932 stats->tx_errors++;
933 stats->tx_dropped++;
934 kfree_skb(skb);
935 return NETDEV_TX_OK;
936 }
937
938 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
939 {
940 struct net_device *dev = t->dev;
941 struct __ip6_tnl_parm *p = &t->parms;
942 struct flowi6 *fl6 = &t->fl.u.ip6;
943 int addend = sizeof(struct ipv6hdr) + 4;
944
945 if (dev->type != ARPHRD_ETHER) {
946 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
947 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
948 }
949
950 /* Set up flowi template */
951 fl6->saddr = p->laddr;
952 fl6->daddr = p->raddr;
953 fl6->flowi6_oif = p->link;
954 fl6->flowlabel = 0;
955
956 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
957 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
958 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
959 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
960
961 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
962 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
963
964 if (p->flags&IP6_TNL_F_CAP_XMIT &&
965 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
966 dev->flags |= IFF_POINTOPOINT;
967 else
968 dev->flags &= ~IFF_POINTOPOINT;
969
970 dev->iflink = p->link;
971
972 /* Precalculate GRE options length */
973 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
974 if (t->parms.o_flags&GRE_CSUM)
975 addend += 4;
976 if (t->parms.o_flags&GRE_KEY)
977 addend += 4;
978 if (t->parms.o_flags&GRE_SEQ)
979 addend += 4;
980 }
981
982 if (p->flags & IP6_TNL_F_CAP_XMIT) {
983 int strict = (ipv6_addr_type(&p->raddr) &
984 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
985
986 struct rt6_info *rt = rt6_lookup(dev_net(dev),
987 &p->raddr, &p->laddr,
988 p->link, strict);
989
990 if (rt == NULL)
991 return;
992
993 if (rt->dst.dev) {
994 dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
995
996 if (set_mtu) {
997 dev->mtu = rt->dst.dev->mtu - addend;
998 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
999 dev->mtu -= 8;
1000
1001 if (dev->mtu < IPV6_MIN_MTU)
1002 dev->mtu = IPV6_MIN_MTU;
1003 }
1004 }
1005 ip6_rt_put(rt);
1006 }
1007
1008 t->hlen = addend;
1009 }
1010
1011 static int ip6gre_tnl_change(struct ip6_tnl *t,
1012 const struct __ip6_tnl_parm *p, int set_mtu)
1013 {
1014 t->parms.laddr = p->laddr;
1015 t->parms.raddr = p->raddr;
1016 t->parms.flags = p->flags;
1017 t->parms.hop_limit = p->hop_limit;
1018 t->parms.encap_limit = p->encap_limit;
1019 t->parms.flowinfo = p->flowinfo;
1020 t->parms.link = p->link;
1021 t->parms.proto = p->proto;
1022 t->parms.i_key = p->i_key;
1023 t->parms.o_key = p->o_key;
1024 t->parms.i_flags = p->i_flags;
1025 t->parms.o_flags = p->o_flags;
1026 ip6_tnl_dst_reset(t);
1027 ip6gre_tnl_link_config(t, set_mtu);
1028 return 0;
1029 }
1030
1031 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1032 const struct ip6_tnl_parm2 *u)
1033 {
1034 p->laddr = u->laddr;
1035 p->raddr = u->raddr;
1036 p->flags = u->flags;
1037 p->hop_limit = u->hop_limit;
1038 p->encap_limit = u->encap_limit;
1039 p->flowinfo = u->flowinfo;
1040 p->link = u->link;
1041 p->i_key = u->i_key;
1042 p->o_key = u->o_key;
1043 p->i_flags = u->i_flags;
1044 p->o_flags = u->o_flags;
1045 memcpy(p->name, u->name, sizeof(u->name));
1046 }
1047
1048 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1049 const struct __ip6_tnl_parm *p)
1050 {
1051 u->proto = IPPROTO_GRE;
1052 u->laddr = p->laddr;
1053 u->raddr = p->raddr;
1054 u->flags = p->flags;
1055 u->hop_limit = p->hop_limit;
1056 u->encap_limit = p->encap_limit;
1057 u->flowinfo = p->flowinfo;
1058 u->link = p->link;
1059 u->i_key = p->i_key;
1060 u->o_key = p->o_key;
1061 u->i_flags = p->i_flags;
1062 u->o_flags = p->o_flags;
1063 memcpy(u->name, p->name, sizeof(u->name));
1064 }
1065
1066 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1067 struct ifreq *ifr, int cmd)
1068 {
1069 int err = 0;
1070 struct ip6_tnl_parm2 p;
1071 struct __ip6_tnl_parm p1;
1072 struct ip6_tnl *t;
1073 struct net *net = dev_net(dev);
1074 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1075
1076 switch (cmd) {
1077 case SIOCGETTUNNEL:
1078 t = NULL;
1079 if (dev == ign->fb_tunnel_dev) {
1080 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1081 err = -EFAULT;
1082 break;
1083 }
1084 ip6gre_tnl_parm_from_user(&p1, &p);
1085 t = ip6gre_tunnel_locate(net, &p1, 0);
1086 }
1087 if (t == NULL)
1088 t = netdev_priv(dev);
1089 memset(&p, 0, sizeof(p));
1090 ip6gre_tnl_parm_to_user(&p, &t->parms);
1091 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1092 err = -EFAULT;
1093 break;
1094
1095 case SIOCADDTUNNEL:
1096 case SIOCCHGTUNNEL:
1097 err = -EPERM;
1098 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1099 goto done;
1100
1101 err = -EFAULT;
1102 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1103 goto done;
1104
1105 err = -EINVAL;
1106 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1107 goto done;
1108
1109 if (!(p.i_flags&GRE_KEY))
1110 p.i_key = 0;
1111 if (!(p.o_flags&GRE_KEY))
1112 p.o_key = 0;
1113
1114 ip6gre_tnl_parm_from_user(&p1, &p);
1115 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1116
1117 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1118 if (t != NULL) {
1119 if (t->dev != dev) {
1120 err = -EEXIST;
1121 break;
1122 }
1123 } else {
1124 t = netdev_priv(dev);
1125
1126 ip6gre_tunnel_unlink(ign, t);
1127 synchronize_net();
1128 ip6gre_tnl_change(t, &p1, 1);
1129 ip6gre_tunnel_link(ign, t);
1130 netdev_state_change(dev);
1131 }
1132 }
1133
1134 if (t) {
1135 err = 0;
1136
1137 memset(&p, 0, sizeof(p));
1138 ip6gre_tnl_parm_to_user(&p, &t->parms);
1139 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1140 err = -EFAULT;
1141 } else
1142 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1143 break;
1144
1145 case SIOCDELTUNNEL:
1146 err = -EPERM;
1147 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1148 goto done;
1149
1150 if (dev == ign->fb_tunnel_dev) {
1151 err = -EFAULT;
1152 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1153 goto done;
1154 err = -ENOENT;
1155 ip6gre_tnl_parm_from_user(&p1, &p);
1156 t = ip6gre_tunnel_locate(net, &p1, 0);
1157 if (t == NULL)
1158 goto done;
1159 err = -EPERM;
1160 if (t == netdev_priv(ign->fb_tunnel_dev))
1161 goto done;
1162 dev = t->dev;
1163 }
1164 unregister_netdevice(dev);
1165 err = 0;
1166 break;
1167
1168 default:
1169 err = -EINVAL;
1170 }
1171
1172 done:
1173 return err;
1174 }
1175
1176 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1177 {
1178 struct ip6_tnl *tunnel = netdev_priv(dev);
1179 if (new_mtu < 68 ||
1180 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1181 return -EINVAL;
1182 dev->mtu = new_mtu;
1183 return 0;
1184 }
1185
1186 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1187 unsigned short type,
1188 const void *daddr, const void *saddr, unsigned int len)
1189 {
1190 struct ip6_tnl *t = netdev_priv(dev);
1191 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1192 __be16 *p = (__be16 *)(ipv6h+1);
1193
1194 ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1195 ipv6h->hop_limit = t->parms.hop_limit;
1196 ipv6h->nexthdr = NEXTHDR_GRE;
1197 ipv6h->saddr = t->parms.laddr;
1198 ipv6h->daddr = t->parms.raddr;
1199
1200 p[0] = t->parms.o_flags;
1201 p[1] = htons(type);
1202
1203 /*
1204 * Set the source hardware address.
1205 */
1206
1207 if (saddr)
1208 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1209 if (daddr)
1210 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1211 if (!ipv6_addr_any(&ipv6h->daddr))
1212 return t->hlen;
1213
1214 return -t->hlen;
1215 }
1216
1217 static const struct header_ops ip6gre_header_ops = {
1218 .create = ip6gre_header,
1219 };
1220
1221 static const struct net_device_ops ip6gre_netdev_ops = {
1222 .ndo_init = ip6gre_tunnel_init,
1223 .ndo_uninit = ip6gre_tunnel_uninit,
1224 .ndo_start_xmit = ip6gre_tunnel_xmit,
1225 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1226 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1227 .ndo_get_stats64 = ip_tunnel_get_stats64,
1228 };
1229
1230 static void ip6gre_dev_free(struct net_device *dev)
1231 {
1232 free_percpu(dev->tstats);
1233 free_netdev(dev);
1234 }
1235
1236 static void ip6gre_tunnel_setup(struct net_device *dev)
1237 {
1238 struct ip6_tnl *t;
1239
1240 dev->netdev_ops = &ip6gre_netdev_ops;
1241 dev->destructor = ip6gre_dev_free;
1242
1243 dev->type = ARPHRD_IP6GRE;
1244 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1245 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1246 t = netdev_priv(dev);
1247 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1248 dev->mtu -= 8;
1249 dev->flags |= IFF_NOARP;
1250 dev->iflink = 0;
1251 dev->addr_len = sizeof(struct in6_addr);
1252 dev->features |= NETIF_F_NETNS_LOCAL;
1253 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1254 }
1255
1256 static int ip6gre_tunnel_init(struct net_device *dev)
1257 {
1258 struct ip6_tnl *tunnel;
1259
1260 tunnel = netdev_priv(dev);
1261
1262 tunnel->dev = dev;
1263 strcpy(tunnel->parms.name, dev->name);
1264
1265 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1266 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1267
1268 if (ipv6_addr_any(&tunnel->parms.raddr))
1269 dev->header_ops = &ip6gre_header_ops;
1270
1271 dev->tstats = alloc_percpu(struct pcpu_tstats);
1272 if (!dev->tstats)
1273 return -ENOMEM;
1274
1275 return 0;
1276 }
1277
1278 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1279 {
1280 struct ip6_tnl *tunnel = netdev_priv(dev);
1281
1282 tunnel->dev = dev;
1283 strcpy(tunnel->parms.name, dev->name);
1284
1285 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1286
1287 dev_hold(dev);
1288 }
1289
1290
1291 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1292 .handler = ip6gre_rcv,
1293 .err_handler = ip6gre_err,
1294 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1295 };
1296
1297 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1298 struct list_head *head)
1299 {
1300 int prio;
1301
1302 for (prio = 0; prio < 4; prio++) {
1303 int h;
1304 for (h = 0; h < HASH_SIZE; h++) {
1305 struct ip6_tnl *t;
1306
1307 t = rtnl_dereference(ign->tunnels[prio][h]);
1308
1309 while (t != NULL) {
1310 unregister_netdevice_queue(t->dev, head);
1311 t = rtnl_dereference(t->next);
1312 }
1313 }
1314 }
1315 }
1316
1317 static int __net_init ip6gre_init_net(struct net *net)
1318 {
1319 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1320 int err;
1321
1322 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1323 ip6gre_tunnel_setup);
1324 if (!ign->fb_tunnel_dev) {
1325 err = -ENOMEM;
1326 goto err_alloc_dev;
1327 }
1328 dev_net_set(ign->fb_tunnel_dev, net);
1329
1330 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1331 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1332
1333 err = register_netdev(ign->fb_tunnel_dev);
1334 if (err)
1335 goto err_reg_dev;
1336
1337 rcu_assign_pointer(ign->tunnels_wc[0],
1338 netdev_priv(ign->fb_tunnel_dev));
1339 return 0;
1340
1341 err_reg_dev:
1342 ip6gre_dev_free(ign->fb_tunnel_dev);
1343 err_alloc_dev:
1344 return err;
1345 }
1346
1347 static void __net_exit ip6gre_exit_net(struct net *net)
1348 {
1349 struct ip6gre_net *ign;
1350 LIST_HEAD(list);
1351
1352 ign = net_generic(net, ip6gre_net_id);
1353 rtnl_lock();
1354 ip6gre_destroy_tunnels(ign, &list);
1355 unregister_netdevice_many(&list);
1356 rtnl_unlock();
1357 }
1358
1359 static struct pernet_operations ip6gre_net_ops = {
1360 .init = ip6gre_init_net,
1361 .exit = ip6gre_exit_net,
1362 .id = &ip6gre_net_id,
1363 .size = sizeof(struct ip6gre_net),
1364 };
1365
1366 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1367 {
1368 __be16 flags;
1369
1370 if (!data)
1371 return 0;
1372
1373 flags = 0;
1374 if (data[IFLA_GRE_IFLAGS])
1375 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1376 if (data[IFLA_GRE_OFLAGS])
1377 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1378 if (flags & (GRE_VERSION|GRE_ROUTING))
1379 return -EINVAL;
1380
1381 return 0;
1382 }
1383
1384 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1385 {
1386 struct in6_addr daddr;
1387
1388 if (tb[IFLA_ADDRESS]) {
1389 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1390 return -EINVAL;
1391 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1392 return -EADDRNOTAVAIL;
1393 }
1394
1395 if (!data)
1396 goto out;
1397
1398 if (data[IFLA_GRE_REMOTE]) {
1399 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1400 if (ipv6_addr_any(&daddr))
1401 return -EINVAL;
1402 }
1403
1404 out:
1405 return ip6gre_tunnel_validate(tb, data);
1406 }
1407
1408
1409 static void ip6gre_netlink_parms(struct nlattr *data[],
1410 struct __ip6_tnl_parm *parms)
1411 {
1412 memset(parms, 0, sizeof(*parms));
1413
1414 if (!data)
1415 return;
1416
1417 if (data[IFLA_GRE_LINK])
1418 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1419
1420 if (data[IFLA_GRE_IFLAGS])
1421 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1422
1423 if (data[IFLA_GRE_OFLAGS])
1424 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1425
1426 if (data[IFLA_GRE_IKEY])
1427 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1428
1429 if (data[IFLA_GRE_OKEY])
1430 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1431
1432 if (data[IFLA_GRE_LOCAL])
1433 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1434
1435 if (data[IFLA_GRE_REMOTE])
1436 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1437
1438 if (data[IFLA_GRE_TTL])
1439 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1440
1441 if (data[IFLA_GRE_ENCAP_LIMIT])
1442 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1443
1444 if (data[IFLA_GRE_FLOWINFO])
1445 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1446
1447 if (data[IFLA_GRE_FLAGS])
1448 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1449 }
1450
1451 static int ip6gre_tap_init(struct net_device *dev)
1452 {
1453 struct ip6_tnl *tunnel;
1454
1455 tunnel = netdev_priv(dev);
1456
1457 tunnel->dev = dev;
1458 strcpy(tunnel->parms.name, dev->name);
1459
1460 ip6gre_tnl_link_config(tunnel, 1);
1461
1462 dev->tstats = alloc_percpu(struct pcpu_tstats);
1463 if (!dev->tstats)
1464 return -ENOMEM;
1465
1466 return 0;
1467 }
1468
1469 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1470 .ndo_init = ip6gre_tap_init,
1471 .ndo_uninit = ip6gre_tunnel_uninit,
1472 .ndo_start_xmit = ip6gre_tunnel_xmit,
1473 .ndo_set_mac_address = eth_mac_addr,
1474 .ndo_validate_addr = eth_validate_addr,
1475 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1476 .ndo_get_stats64 = ip_tunnel_get_stats64,
1477 };
1478
1479 static void ip6gre_tap_setup(struct net_device *dev)
1480 {
1481
1482 ether_setup(dev);
1483
1484 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1485 dev->destructor = ip6gre_dev_free;
1486
1487 dev->iflink = 0;
1488 dev->features |= NETIF_F_NETNS_LOCAL;
1489 }
1490
1491 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1492 struct nlattr *tb[], struct nlattr *data[])
1493 {
1494 struct ip6_tnl *nt;
1495 struct net *net = dev_net(dev);
1496 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1497 int err;
1498
1499 nt = netdev_priv(dev);
1500 ip6gre_netlink_parms(data, &nt->parms);
1501
1502 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1503 return -EEXIST;
1504
1505 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1506 eth_hw_addr_random(dev);
1507
1508 nt->dev = dev;
1509 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1510
1511 /* Can use a lockless transmit, unless we generate output sequences */
1512 if (!(nt->parms.o_flags & GRE_SEQ))
1513 dev->features |= NETIF_F_LLTX;
1514
1515 err = register_netdevice(dev);
1516 if (err)
1517 goto out;
1518
1519 dev_hold(dev);
1520 ip6gre_tunnel_link(ign, nt);
1521
1522 out:
1523 return err;
1524 }
1525
1526 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1527 struct nlattr *data[])
1528 {
1529 struct ip6_tnl *t, *nt;
1530 struct net *net = dev_net(dev);
1531 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1532 struct __ip6_tnl_parm p;
1533
1534 if (dev == ign->fb_tunnel_dev)
1535 return -EINVAL;
1536
1537 nt = netdev_priv(dev);
1538 ip6gre_netlink_parms(data, &p);
1539
1540 t = ip6gre_tunnel_locate(net, &p, 0);
1541
1542 if (t) {
1543 if (t->dev != dev)
1544 return -EEXIST;
1545 } else {
1546 t = nt;
1547
1548 ip6gre_tunnel_unlink(ign, t);
1549 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1550 ip6gre_tunnel_link(ign, t);
1551 netdev_state_change(dev);
1552 }
1553
1554 return 0;
1555 }
1556
1557 static size_t ip6gre_get_size(const struct net_device *dev)
1558 {
1559 return
1560 /* IFLA_GRE_LINK */
1561 nla_total_size(4) +
1562 /* IFLA_GRE_IFLAGS */
1563 nla_total_size(2) +
1564 /* IFLA_GRE_OFLAGS */
1565 nla_total_size(2) +
1566 /* IFLA_GRE_IKEY */
1567 nla_total_size(4) +
1568 /* IFLA_GRE_OKEY */
1569 nla_total_size(4) +
1570 /* IFLA_GRE_LOCAL */
1571 nla_total_size(sizeof(struct in6_addr)) +
1572 /* IFLA_GRE_REMOTE */
1573 nla_total_size(sizeof(struct in6_addr)) +
1574 /* IFLA_GRE_TTL */
1575 nla_total_size(1) +
1576 /* IFLA_GRE_TOS */
1577 nla_total_size(1) +
1578 /* IFLA_GRE_ENCAP_LIMIT */
1579 nla_total_size(1) +
1580 /* IFLA_GRE_FLOWINFO */
1581 nla_total_size(4) +
1582 /* IFLA_GRE_FLAGS */
1583 nla_total_size(4) +
1584 0;
1585 }
1586
1587 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1588 {
1589 struct ip6_tnl *t = netdev_priv(dev);
1590 struct __ip6_tnl_parm *p = &t->parms;
1591
1592 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1593 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1594 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1595 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1596 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1597 nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1598 nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1599 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1600 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1601 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1602 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1603 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1604 goto nla_put_failure;
1605 return 0;
1606
1607 nla_put_failure:
1608 return -EMSGSIZE;
1609 }
1610
1611 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1612 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1613 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1614 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1615 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1616 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1617 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1618 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1619 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1620 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1621 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1622 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1623 };
1624
1625 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1626 .kind = "ip6gre",
1627 .maxtype = IFLA_GRE_MAX,
1628 .policy = ip6gre_policy,
1629 .priv_size = sizeof(struct ip6_tnl),
1630 .setup = ip6gre_tunnel_setup,
1631 .validate = ip6gre_tunnel_validate,
1632 .newlink = ip6gre_newlink,
1633 .changelink = ip6gre_changelink,
1634 .get_size = ip6gre_get_size,
1635 .fill_info = ip6gre_fill_info,
1636 };
1637
1638 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1639 .kind = "ip6gretap",
1640 .maxtype = IFLA_GRE_MAX,
1641 .policy = ip6gre_policy,
1642 .priv_size = sizeof(struct ip6_tnl),
1643 .setup = ip6gre_tap_setup,
1644 .validate = ip6gre_tap_validate,
1645 .newlink = ip6gre_newlink,
1646 .changelink = ip6gre_changelink,
1647 .get_size = ip6gre_get_size,
1648 .fill_info = ip6gre_fill_info,
1649 };
1650
1651 /*
1652 * And now the modules code and kernel interface.
1653 */
1654
1655 static int __init ip6gre_init(void)
1656 {
1657 int err;
1658
1659 pr_info("GRE over IPv6 tunneling driver\n");
1660
1661 err = register_pernet_device(&ip6gre_net_ops);
1662 if (err < 0)
1663 return err;
1664
1665 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1666 if (err < 0) {
1667 pr_info("%s: can't add protocol\n", __func__);
1668 goto add_proto_failed;
1669 }
1670
1671 err = rtnl_link_register(&ip6gre_link_ops);
1672 if (err < 0)
1673 goto rtnl_link_failed;
1674
1675 err = rtnl_link_register(&ip6gre_tap_ops);
1676 if (err < 0)
1677 goto tap_ops_failed;
1678
1679 out:
1680 return err;
1681
1682 tap_ops_failed:
1683 rtnl_link_unregister(&ip6gre_link_ops);
1684 rtnl_link_failed:
1685 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1686 add_proto_failed:
1687 unregister_pernet_device(&ip6gre_net_ops);
1688 goto out;
1689 }
1690
1691 static void __exit ip6gre_fini(void)
1692 {
1693 rtnl_link_unregister(&ip6gre_tap_ops);
1694 rtnl_link_unregister(&ip6gre_link_ops);
1695 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1696 unregister_pernet_device(&ip6gre_net_ops);
1697 }
1698
1699 module_init(ip6gre_init);
1700 module_exit(ip6gre_fini);
1701 MODULE_LICENSE("GPL");
1702 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1703 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1704 MODULE_ALIAS_RTNL_LINK("ip6gre");
1705 MODULE_ALIAS_NETDEV("ip6gre0");
Linux kernel - Deliverables
This page took 0.064136 seconds and 6 git commands to generate.