]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv6/sit.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipv4: Make caller provide on-stack flow key to ip_route_output_ports().
[mirror_ubuntu-artful-kernel.git] / net / ipv6 / sit.c
1 /*
2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 *
14 * Changes:
15 * Roger Venning <r.venning@telstra.com>: 6to4 support
16 * Nate Thompson <nate@thebog.net>: 6to4 support
17 * Fred Templin <fred.l.templin@boeing.com>: isatap support
18 */
19
20 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <linux/slab.h>
32 #include <asm/uaccess.h>
33 #include <linux/init.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
36
37 #include <net/sock.h>
38 #include <net/snmp.h>
39
40 #include <net/ipv6.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_fib.h>
44 #include <net/ip6_route.h>
45 #include <net/ndisc.h>
46 #include <net/addrconf.h>
47 #include <net/ip.h>
48 #include <net/udp.h>
49 #include <net/icmp.h>
50 #include <net/ipip.h>
51 #include <net/inet_ecn.h>
52 #include <net/xfrm.h>
53 #include <net/dsfield.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56
57 /*
58 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
59
60 For comments look at net/ipv4/ip_gre.c --ANK
61 */
62
63 #define HASH_SIZE 16
64 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
65
66 static int ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68 static void ipip6_dev_free(struct net_device *dev);
69
70 static int sit_net_id __read_mostly;
71 struct sit_net {
72 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
73 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
74 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_wc[1];
76 struct ip_tunnel __rcu **tunnels[4];
77
78 struct net_device *fb_tunnel_dev;
79 };
80
81 /*
82 * Locking : hash tables are protected by RCU and RTNL
83 */
84
85 #define for_each_ip_tunnel_rcu(start) \
86 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
87
88 /* often modified stats are per cpu, other are shared (netdev->stats) */
89 struct pcpu_tstats {
90 unsigned long rx_packets;
91 unsigned long rx_bytes;
92 unsigned long tx_packets;
93 unsigned long tx_bytes;
94 };
95
96 static struct net_device_stats *ipip6_get_stats(struct net_device *dev)
97 {
98 struct pcpu_tstats sum = { 0 };
99 int i;
100
101 for_each_possible_cpu(i) {
102 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
103
104 sum.rx_packets += tstats->rx_packets;
105 sum.rx_bytes += tstats->rx_bytes;
106 sum.tx_packets += tstats->tx_packets;
107 sum.tx_bytes += tstats->tx_bytes;
108 }
109 dev->stats.rx_packets = sum.rx_packets;
110 dev->stats.rx_bytes = sum.rx_bytes;
111 dev->stats.tx_packets = sum.tx_packets;
112 dev->stats.tx_bytes = sum.tx_bytes;
113 return &dev->stats;
114 }
115 /*
116 * Must be invoked with rcu_read_lock
117 */
118 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
119 struct net_device *dev, __be32 remote, __be32 local)
120 {
121 unsigned int h0 = HASH(remote);
122 unsigned int h1 = HASH(local);
123 struct ip_tunnel *t;
124 struct sit_net *sitn = net_generic(net, sit_net_id);
125
126 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
127 if (local == t->parms.iph.saddr &&
128 remote == t->parms.iph.daddr &&
129 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
130 (t->dev->flags & IFF_UP))
131 return t;
132 }
133 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
134 if (remote == t->parms.iph.daddr &&
135 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
136 (t->dev->flags & IFF_UP))
137 return t;
138 }
139 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
140 if (local == t->parms.iph.saddr &&
141 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
142 (t->dev->flags & IFF_UP))
143 return t;
144 }
145 t = rcu_dereference(sitn->tunnels_wc[0]);
146 if ((t != NULL) && (t->dev->flags & IFF_UP))
147 return t;
148 return NULL;
149 }
150
151 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
152 struct ip_tunnel_parm *parms)
153 {
154 __be32 remote = parms->iph.daddr;
155 __be32 local = parms->iph.saddr;
156 unsigned int h = 0;
157 int prio = 0;
158
159 if (remote) {
160 prio |= 2;
161 h ^= HASH(remote);
162 }
163 if (local) {
164 prio |= 1;
165 h ^= HASH(local);
166 }
167 return &sitn->tunnels[prio][h];
168 }
169
170 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
171 struct ip_tunnel *t)
172 {
173 return __ipip6_bucket(sitn, &t->parms);
174 }
175
176 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
177 {
178 struct ip_tunnel __rcu **tp;
179 struct ip_tunnel *iter;
180
181 for (tp = ipip6_bucket(sitn, t);
182 (iter = rtnl_dereference(*tp)) != NULL;
183 tp = &iter->next) {
184 if (t == iter) {
185 rcu_assign_pointer(*tp, t->next);
186 break;
187 }
188 }
189 }
190
191 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
192 {
193 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
194
195 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
196 rcu_assign_pointer(*tp, t);
197 }
198
199 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
200 {
201 #ifdef CONFIG_IPV6_SIT_6RD
202 struct ip_tunnel *t = netdev_priv(dev);
203
204 if (t->dev == sitn->fb_tunnel_dev) {
205 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
206 t->ip6rd.relay_prefix = 0;
207 t->ip6rd.prefixlen = 16;
208 t->ip6rd.relay_prefixlen = 0;
209 } else {
210 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
211 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
212 }
213 #endif
214 }
215
216 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
217 struct ip_tunnel_parm *parms, int create)
218 {
219 __be32 remote = parms->iph.daddr;
220 __be32 local = parms->iph.saddr;
221 struct ip_tunnel *t, *nt;
222 struct ip_tunnel __rcu **tp;
223 struct net_device *dev;
224 char name[IFNAMSIZ];
225 struct sit_net *sitn = net_generic(net, sit_net_id);
226
227 for (tp = __ipip6_bucket(sitn, parms);
228 (t = rtnl_dereference(*tp)) != NULL;
229 tp = &t->next) {
230 if (local == t->parms.iph.saddr &&
231 remote == t->parms.iph.daddr &&
232 parms->link == t->parms.link) {
233 if (create)
234 return NULL;
235 else
236 return t;
237 }
238 }
239 if (!create)
240 goto failed;
241
242 if (parms->name[0])
243 strlcpy(name, parms->name, IFNAMSIZ);
244 else
245 strcpy(name, "sit%d");
246
247 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
248 if (dev == NULL)
249 return NULL;
250
251 dev_net_set(dev, net);
252
253 if (strchr(name, '%')) {
254 if (dev_alloc_name(dev, name) < 0)
255 goto failed_free;
256 }
257
258 nt = netdev_priv(dev);
259
260 nt->parms = *parms;
261 if (ipip6_tunnel_init(dev) < 0)
262 goto failed_free;
263 ipip6_tunnel_clone_6rd(dev, sitn);
264
265 if (parms->i_flags & SIT_ISATAP)
266 dev->priv_flags |= IFF_ISATAP;
267
268 if (register_netdevice(dev) < 0)
269 goto failed_free;
270
271 dev_hold(dev);
272
273 ipip6_tunnel_link(sitn, nt);
274 return nt;
275
276 failed_free:
277 ipip6_dev_free(dev);
278 failed:
279 return NULL;
280 }
281
282 #define for_each_prl_rcu(start) \
283 for (prl = rcu_dereference(start); \
284 prl; \
285 prl = rcu_dereference(prl->next))
286
287 static struct ip_tunnel_prl_entry *
288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
289 {
290 struct ip_tunnel_prl_entry *prl;
291
292 for_each_prl_rcu(t->prl)
293 if (prl->addr == addr)
294 break;
295 return prl;
296
297 }
298
299 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
300 struct ip_tunnel_prl __user *a)
301 {
302 struct ip_tunnel_prl kprl, *kp;
303 struct ip_tunnel_prl_entry *prl;
304 unsigned int cmax, c = 0, ca, len;
305 int ret = 0;
306
307 if (copy_from_user(&kprl, a, sizeof(kprl)))
308 return -EFAULT;
309 cmax = kprl.datalen / sizeof(kprl);
310 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
311 cmax = 1;
312
313 /* For simple GET or for root users,
314 * we try harder to allocate.
315 */
316 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
317 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
318 NULL;
319
320 rcu_read_lock();
321
322 ca = t->prl_count < cmax ? t->prl_count : cmax;
323
324 if (!kp) {
325 /* We don't try hard to allocate much memory for
326 * non-root users.
327 * For root users, retry allocating enough memory for
328 * the answer.
329 */
330 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
331 if (!kp) {
332 ret = -ENOMEM;
333 goto out;
334 }
335 }
336
337 c = 0;
338 for_each_prl_rcu(t->prl) {
339 if (c >= cmax)
340 break;
341 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
342 continue;
343 kp[c].addr = prl->addr;
344 kp[c].flags = prl->flags;
345 c++;
346 if (kprl.addr != htonl(INADDR_ANY))
347 break;
348 }
349 out:
350 rcu_read_unlock();
351
352 len = sizeof(*kp) * c;
353 ret = 0;
354 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
355 ret = -EFAULT;
356
357 kfree(kp);
358
359 return ret;
360 }
361
362 static int
363 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
364 {
365 struct ip_tunnel_prl_entry *p;
366 int err = 0;
367
368 if (a->addr == htonl(INADDR_ANY))
369 return -EINVAL;
370
371 ASSERT_RTNL();
372
373 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
374 if (p->addr == a->addr) {
375 if (chg) {
376 p->flags = a->flags;
377 goto out;
378 }
379 err = -EEXIST;
380 goto out;
381 }
382 }
383
384 if (chg) {
385 err = -ENXIO;
386 goto out;
387 }
388
389 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
390 if (!p) {
391 err = -ENOBUFS;
392 goto out;
393 }
394
395 p->next = t->prl;
396 p->addr = a->addr;
397 p->flags = a->flags;
398 t->prl_count++;
399 rcu_assign_pointer(t->prl, p);
400 out:
401 return err;
402 }
403
404 static void prl_entry_destroy_rcu(struct rcu_head *head)
405 {
406 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
407 }
408
409 static void prl_list_destroy_rcu(struct rcu_head *head)
410 {
411 struct ip_tunnel_prl_entry *p, *n;
412
413 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
414 do {
415 n = rcu_dereference_protected(p->next, 1);
416 kfree(p);
417 p = n;
418 } while (p);
419 }
420
421 static int
422 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
423 {
424 struct ip_tunnel_prl_entry *x;
425 struct ip_tunnel_prl_entry __rcu **p;
426 int err = 0;
427
428 ASSERT_RTNL();
429
430 if (a && a->addr != htonl(INADDR_ANY)) {
431 for (p = &t->prl;
432 (x = rtnl_dereference(*p)) != NULL;
433 p = &x->next) {
434 if (x->addr == a->addr) {
435 *p = x->next;
436 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
437 t->prl_count--;
438 goto out;
439 }
440 }
441 err = -ENXIO;
442 } else {
443 x = rtnl_dereference(t->prl);
444 if (x) {
445 t->prl_count = 0;
446 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
447 t->prl = NULL;
448 }
449 }
450 out:
451 return err;
452 }
453
454 static int
455 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
456 {
457 struct ip_tunnel_prl_entry *p;
458 int ok = 1;
459
460 rcu_read_lock();
461 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
462 if (p) {
463 if (p->flags & PRL_DEFAULT)
464 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
465 else
466 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
467 } else {
468 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
469
470 if (ipv6_addr_is_isatap(addr6) &&
471 (addr6->s6_addr32[3] == iph->saddr) &&
472 ipv6_chk_prefix(addr6, t->dev))
473 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
474 else
475 ok = 0;
476 }
477 rcu_read_unlock();
478 return ok;
479 }
480
481 static void ipip6_tunnel_uninit(struct net_device *dev)
482 {
483 struct net *net = dev_net(dev);
484 struct sit_net *sitn = net_generic(net, sit_net_id);
485
486 if (dev == sitn->fb_tunnel_dev) {
487 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
488 } else {
489 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
490 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
491 }
492 dev_put(dev);
493 }
494
495
496 static int ipip6_err(struct sk_buff *skb, u32 info)
497 {
498
499 /* All the routers (except for Linux) return only
500 8 bytes of packet payload. It means, that precise relaying of
501 ICMP in the real Internet is absolutely infeasible.
502 */
503 const struct iphdr *iph = (const struct iphdr *)skb->data;
504 const int type = icmp_hdr(skb)->type;
505 const int code = icmp_hdr(skb)->code;
506 struct ip_tunnel *t;
507 int err;
508
509 switch (type) {
510 default:
511 case ICMP_PARAMETERPROB:
512 return 0;
513
514 case ICMP_DEST_UNREACH:
515 switch (code) {
516 case ICMP_SR_FAILED:
517 case ICMP_PORT_UNREACH:
518 /* Impossible event. */
519 return 0;
520 case ICMP_FRAG_NEEDED:
521 /* Soft state for pmtu is maintained by IP core. */
522 return 0;
523 default:
524 /* All others are translated to HOST_UNREACH.
525 rfc2003 contains "deep thoughts" about NET_UNREACH,
526 I believe they are just ether pollution. --ANK
527 */
528 break;
529 }
530 break;
531 case ICMP_TIME_EXCEEDED:
532 if (code != ICMP_EXC_TTL)
533 return 0;
534 break;
535 }
536
537 err = -ENOENT;
538
539 rcu_read_lock();
540 t = ipip6_tunnel_lookup(dev_net(skb->dev),
541 skb->dev,
542 iph->daddr,
543 iph->saddr);
544 if (t == NULL || t->parms.iph.daddr == 0)
545 goto out;
546
547 err = 0;
548 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
549 goto out;
550
551 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
552 t->err_count++;
553 else
554 t->err_count = 1;
555 t->err_time = jiffies;
556 out:
557 rcu_read_unlock();
558 return err;
559 }
560
561 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
562 {
563 if (INET_ECN_is_ce(iph->tos))
564 IP6_ECN_set_ce(ipv6_hdr(skb));
565 }
566
567 static int ipip6_rcv(struct sk_buff *skb)
568 {
569 const struct iphdr *iph;
570 struct ip_tunnel *tunnel;
571
572 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
573 goto out;
574
575 iph = ip_hdr(skb);
576
577 rcu_read_lock();
578 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
579 iph->saddr, iph->daddr);
580 if (tunnel != NULL) {
581 struct pcpu_tstats *tstats;
582
583 secpath_reset(skb);
584 skb->mac_header = skb->network_header;
585 skb_reset_network_header(skb);
586 IPCB(skb)->flags = 0;
587 skb->protocol = htons(ETH_P_IPV6);
588 skb->pkt_type = PACKET_HOST;
589
590 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
591 !isatap_chksrc(skb, iph, tunnel)) {
592 tunnel->dev->stats.rx_errors++;
593 rcu_read_unlock();
594 kfree_skb(skb);
595 return 0;
596 }
597
598 tstats = this_cpu_ptr(tunnel->dev->tstats);
599 tstats->rx_packets++;
600 tstats->rx_bytes += skb->len;
601
602 __skb_tunnel_rx(skb, tunnel->dev);
603
604 ipip6_ecn_decapsulate(iph, skb);
605
606 netif_rx(skb);
607
608 rcu_read_unlock();
609 return 0;
610 }
611
612 /* no tunnel matched, let upstream know, ipsec may handle it */
613 rcu_read_unlock();
614 return 1;
615 out:
616 kfree_skb(skb);
617 return 0;
618 }
619
620 /*
621 * Returns the embedded IPv4 address if the IPv6 address
622 * comes from 6rd / 6to4 (RFC 3056) addr space.
623 */
624 static inline
625 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
626 {
627 __be32 dst = 0;
628
629 #ifdef CONFIG_IPV6_SIT_6RD
630 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
631 tunnel->ip6rd.prefixlen)) {
632 unsigned int pbw0, pbi0;
633 int pbi1;
634 u32 d;
635
636 pbw0 = tunnel->ip6rd.prefixlen >> 5;
637 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
638
639 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
640 tunnel->ip6rd.relay_prefixlen;
641
642 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
643 if (pbi1 > 0)
644 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
645 (32 - pbi1);
646
647 dst = tunnel->ip6rd.relay_prefix | htonl(d);
648 }
649 #else
650 if (v6dst->s6_addr16[0] == htons(0x2002)) {
651 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
652 memcpy(&dst, &v6dst->s6_addr16[1], 4);
653 }
654 #endif
655 return dst;
656 }
657
658 /*
659 * This function assumes it is being called from dev_queue_xmit()
660 * and that skb is filled properly by that function.
661 */
662
663 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
664 struct net_device *dev)
665 {
666 struct ip_tunnel *tunnel = netdev_priv(dev);
667 struct pcpu_tstats *tstats;
668 const struct iphdr *tiph = &tunnel->parms.iph;
669 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
670 u8 tos = tunnel->parms.iph.tos;
671 __be16 df = tiph->frag_off;
672 struct rtable *rt; /* Route to the other host */
673 struct net_device *tdev; /* Device to other host */
674 struct iphdr *iph; /* Our new IP header */
675 unsigned int max_headroom; /* The extra header space needed */
676 __be32 dst = tiph->daddr;
677 struct flowi4 fl4;
678 int mtu;
679 const struct in6_addr *addr6;
680 int addr_type;
681
682 if (skb->protocol != htons(ETH_P_IPV6))
683 goto tx_error;
684
685 /* ISATAP (RFC4214) - must come before 6to4 */
686 if (dev->priv_flags & IFF_ISATAP) {
687 struct neighbour *neigh = NULL;
688
689 if (skb_dst(skb))
690 neigh = skb_dst(skb)->neighbour;
691
692 if (neigh == NULL) {
693 if (net_ratelimit())
694 printk(KERN_DEBUG "sit: nexthop == NULL\n");
695 goto tx_error;
696 }
697
698 addr6 = (const struct in6_addr*)&neigh->primary_key;
699 addr_type = ipv6_addr_type(addr6);
700
701 if ((addr_type & IPV6_ADDR_UNICAST) &&
702 ipv6_addr_is_isatap(addr6))
703 dst = addr6->s6_addr32[3];
704 else
705 goto tx_error;
706 }
707
708 if (!dst)
709 dst = try_6rd(&iph6->daddr, tunnel);
710
711 if (!dst) {
712 struct neighbour *neigh = NULL;
713
714 if (skb_dst(skb))
715 neigh = skb_dst(skb)->neighbour;
716
717 if (neigh == NULL) {
718 if (net_ratelimit())
719 printk(KERN_DEBUG "sit: nexthop == NULL\n");
720 goto tx_error;
721 }
722
723 addr6 = (const struct in6_addr*)&neigh->primary_key;
724 addr_type = ipv6_addr_type(addr6);
725
726 if (addr_type == IPV6_ADDR_ANY) {
727 addr6 = &ipv6_hdr(skb)->daddr;
728 addr_type = ipv6_addr_type(addr6);
729 }
730
731 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
732 goto tx_error_icmp;
733
734 dst = addr6->s6_addr32[3];
735 }
736
737 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
738 dst, tiph->saddr,
739 0, 0,
740 IPPROTO_IPV6, RT_TOS(tos),
741 tunnel->parms.link);
742 if (IS_ERR(rt)) {
743 dev->stats.tx_carrier_errors++;
744 goto tx_error_icmp;
745 }
746 if (rt->rt_type != RTN_UNICAST) {
747 ip_rt_put(rt);
748 dev->stats.tx_carrier_errors++;
749 goto tx_error_icmp;
750 }
751 tdev = rt->dst.dev;
752
753 if (tdev == dev) {
754 ip_rt_put(rt);
755 dev->stats.collisions++;
756 goto tx_error;
757 }
758
759 if (df) {
760 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
761
762 if (mtu < 68) {
763 dev->stats.collisions++;
764 ip_rt_put(rt);
765 goto tx_error;
766 }
767
768 if (mtu < IPV6_MIN_MTU) {
769 mtu = IPV6_MIN_MTU;
770 df = 0;
771 }
772
773 if (tunnel->parms.iph.daddr && skb_dst(skb))
774 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
775
776 if (skb->len > mtu) {
777 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
778 ip_rt_put(rt);
779 goto tx_error;
780 }
781 }
782
783 if (tunnel->err_count > 0) {
784 if (time_before(jiffies,
785 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
786 tunnel->err_count--;
787 dst_link_failure(skb);
788 } else
789 tunnel->err_count = 0;
790 }
791
792 /*
793 * Okay, now see if we can stuff it in the buffer as-is.
794 */
795 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
796
797 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
798 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
799 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
800 if (!new_skb) {
801 ip_rt_put(rt);
802 dev->stats.tx_dropped++;
803 dev_kfree_skb(skb);
804 return NETDEV_TX_OK;
805 }
806 if (skb->sk)
807 skb_set_owner_w(new_skb, skb->sk);
808 dev_kfree_skb(skb);
809 skb = new_skb;
810 iph6 = ipv6_hdr(skb);
811 }
812
813 skb->transport_header = skb->network_header;
814 skb_push(skb, sizeof(struct iphdr));
815 skb_reset_network_header(skb);
816 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
817 IPCB(skb)->flags = 0;
818 skb_dst_drop(skb);
819 skb_dst_set(skb, &rt->dst);
820
821 /*
822 * Push down and install the IPIP header.
823 */
824
825 iph = ip_hdr(skb);
826 iph->version = 4;
827 iph->ihl = sizeof(struct iphdr)>>2;
828 iph->frag_off = df;
829 iph->protocol = IPPROTO_IPV6;
830 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
831 iph->daddr = rt->rt_dst;
832 iph->saddr = rt->rt_src;
833
834 if ((iph->ttl = tiph->ttl) == 0)
835 iph->ttl = iph6->hop_limit;
836
837 nf_reset(skb);
838 tstats = this_cpu_ptr(dev->tstats);
839 __IPTUNNEL_XMIT(tstats, &dev->stats);
840 return NETDEV_TX_OK;
841
842 tx_error_icmp:
843 dst_link_failure(skb);
844 tx_error:
845 dev->stats.tx_errors++;
846 dev_kfree_skb(skb);
847 return NETDEV_TX_OK;
848 }
849
850 static void ipip6_tunnel_bind_dev(struct net_device *dev)
851 {
852 struct net_device *tdev = NULL;
853 struct ip_tunnel *tunnel;
854 const struct iphdr *iph;
855 struct flowi4 fl4;
856
857 tunnel = netdev_priv(dev);
858 iph = &tunnel->parms.iph;
859
860 if (iph->daddr) {
861 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
862 iph->daddr, iph->saddr,
863 0, 0,
864 IPPROTO_IPV6,
865 RT_TOS(iph->tos),
866 tunnel->parms.link);
867
868 if (!IS_ERR(rt)) {
869 tdev = rt->dst.dev;
870 ip_rt_put(rt);
871 }
872 dev->flags |= IFF_POINTOPOINT;
873 }
874
875 if (!tdev && tunnel->parms.link)
876 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
877
878 if (tdev) {
879 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
880 dev->mtu = tdev->mtu - sizeof(struct iphdr);
881 if (dev->mtu < IPV6_MIN_MTU)
882 dev->mtu = IPV6_MIN_MTU;
883 }
884 dev->iflink = tunnel->parms.link;
885 }
886
887 static int
888 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
889 {
890 int err = 0;
891 struct ip_tunnel_parm p;
892 struct ip_tunnel_prl prl;
893 struct ip_tunnel *t;
894 struct net *net = dev_net(dev);
895 struct sit_net *sitn = net_generic(net, sit_net_id);
896 #ifdef CONFIG_IPV6_SIT_6RD
897 struct ip_tunnel_6rd ip6rd;
898 #endif
899
900 switch (cmd) {
901 case SIOCGETTUNNEL:
902 #ifdef CONFIG_IPV6_SIT_6RD
903 case SIOCGET6RD:
904 #endif
905 t = NULL;
906 if (dev == sitn->fb_tunnel_dev) {
907 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
908 err = -EFAULT;
909 break;
910 }
911 t = ipip6_tunnel_locate(net, &p, 0);
912 }
913 if (t == NULL)
914 t = netdev_priv(dev);
915
916 err = -EFAULT;
917 if (cmd == SIOCGETTUNNEL) {
918 memcpy(&p, &t->parms, sizeof(p));
919 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
920 sizeof(p)))
921 goto done;
922 #ifdef CONFIG_IPV6_SIT_6RD
923 } else {
924 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
925 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
926 ip6rd.prefixlen = t->ip6rd.prefixlen;
927 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
928 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
929 sizeof(ip6rd)))
930 goto done;
931 #endif
932 }
933 err = 0;
934 break;
935
936 case SIOCADDTUNNEL:
937 case SIOCCHGTUNNEL:
938 err = -EPERM;
939 if (!capable(CAP_NET_ADMIN))
940 goto done;
941
942 err = -EFAULT;
943 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
944 goto done;
945
946 err = -EINVAL;
947 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
948 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
949 goto done;
950 if (p.iph.ttl)
951 p.iph.frag_off |= htons(IP_DF);
952
953 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
954
955 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
956 if (t != NULL) {
957 if (t->dev != dev) {
958 err = -EEXIST;
959 break;
960 }
961 } else {
962 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
963 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
964 err = -EINVAL;
965 break;
966 }
967 t = netdev_priv(dev);
968 ipip6_tunnel_unlink(sitn, t);
969 synchronize_net();
970 t->parms.iph.saddr = p.iph.saddr;
971 t->parms.iph.daddr = p.iph.daddr;
972 memcpy(dev->dev_addr, &p.iph.saddr, 4);
973 memcpy(dev->broadcast, &p.iph.daddr, 4);
974 ipip6_tunnel_link(sitn, t);
975 netdev_state_change(dev);
976 }
977 }
978
979 if (t) {
980 err = 0;
981 if (cmd == SIOCCHGTUNNEL) {
982 t->parms.iph.ttl = p.iph.ttl;
983 t->parms.iph.tos = p.iph.tos;
984 if (t->parms.link != p.link) {
985 t->parms.link = p.link;
986 ipip6_tunnel_bind_dev(dev);
987 netdev_state_change(dev);
988 }
989 }
990 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
991 err = -EFAULT;
992 } else
993 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
994 break;
995
996 case SIOCDELTUNNEL:
997 err = -EPERM;
998 if (!capable(CAP_NET_ADMIN))
999 goto done;
1000
1001 if (dev == sitn->fb_tunnel_dev) {
1002 err = -EFAULT;
1003 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1004 goto done;
1005 err = -ENOENT;
1006 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1007 goto done;
1008 err = -EPERM;
1009 if (t == netdev_priv(sitn->fb_tunnel_dev))
1010 goto done;
1011 dev = t->dev;
1012 }
1013 unregister_netdevice(dev);
1014 err = 0;
1015 break;
1016
1017 case SIOCGETPRL:
1018 err = -EINVAL;
1019 if (dev == sitn->fb_tunnel_dev)
1020 goto done;
1021 err = -ENOENT;
1022 if (!(t = netdev_priv(dev)))
1023 goto done;
1024 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1025 break;
1026
1027 case SIOCADDPRL:
1028 case SIOCDELPRL:
1029 case SIOCCHGPRL:
1030 err = -EPERM;
1031 if (!capable(CAP_NET_ADMIN))
1032 goto done;
1033 err = -EINVAL;
1034 if (dev == sitn->fb_tunnel_dev)
1035 goto done;
1036 err = -EFAULT;
1037 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1038 goto done;
1039 err = -ENOENT;
1040 if (!(t = netdev_priv(dev)))
1041 goto done;
1042
1043 switch (cmd) {
1044 case SIOCDELPRL:
1045 err = ipip6_tunnel_del_prl(t, &prl);
1046 break;
1047 case SIOCADDPRL:
1048 case SIOCCHGPRL:
1049 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1050 break;
1051 }
1052 netdev_state_change(dev);
1053 break;
1054
1055 #ifdef CONFIG_IPV6_SIT_6RD
1056 case SIOCADD6RD:
1057 case SIOCCHG6RD:
1058 case SIOCDEL6RD:
1059 err = -EPERM;
1060 if (!capable(CAP_NET_ADMIN))
1061 goto done;
1062
1063 err = -EFAULT;
1064 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1065 sizeof(ip6rd)))
1066 goto done;
1067
1068 t = netdev_priv(dev);
1069
1070 if (cmd != SIOCDEL6RD) {
1071 struct in6_addr prefix;
1072 __be32 relay_prefix;
1073
1074 err = -EINVAL;
1075 if (ip6rd.relay_prefixlen > 32 ||
1076 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1077 goto done;
1078
1079 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1080 ip6rd.prefixlen);
1081 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1082 goto done;
1083 if (ip6rd.relay_prefixlen)
1084 relay_prefix = ip6rd.relay_prefix &
1085 htonl(0xffffffffUL <<
1086 (32 - ip6rd.relay_prefixlen));
1087 else
1088 relay_prefix = 0;
1089 if (relay_prefix != ip6rd.relay_prefix)
1090 goto done;
1091
1092 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1093 t->ip6rd.relay_prefix = relay_prefix;
1094 t->ip6rd.prefixlen = ip6rd.prefixlen;
1095 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1096 } else
1097 ipip6_tunnel_clone_6rd(dev, sitn);
1098
1099 err = 0;
1100 break;
1101 #endif
1102
1103 default:
1104 err = -EINVAL;
1105 }
1106
1107 done:
1108 return err;
1109 }
1110
1111 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1112 {
1113 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1114 return -EINVAL;
1115 dev->mtu = new_mtu;
1116 return 0;
1117 }
1118
1119 static const struct net_device_ops ipip6_netdev_ops = {
1120 .ndo_uninit = ipip6_tunnel_uninit,
1121 .ndo_start_xmit = ipip6_tunnel_xmit,
1122 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1123 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1124 .ndo_get_stats = ipip6_get_stats,
1125 };
1126
1127 static void ipip6_dev_free(struct net_device *dev)
1128 {
1129 free_percpu(dev->tstats);
1130 free_netdev(dev);
1131 }
1132
1133 static void ipip6_tunnel_setup(struct net_device *dev)
1134 {
1135 dev->netdev_ops = &ipip6_netdev_ops;
1136 dev->destructor = ipip6_dev_free;
1137
1138 dev->type = ARPHRD_SIT;
1139 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1140 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1141 dev->flags = IFF_NOARP;
1142 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1143 dev->iflink = 0;
1144 dev->addr_len = 4;
1145 dev->features |= NETIF_F_NETNS_LOCAL;
1146 dev->features |= NETIF_F_LLTX;
1147 }
1148
1149 static int ipip6_tunnel_init(struct net_device *dev)
1150 {
1151 struct ip_tunnel *tunnel = netdev_priv(dev);
1152
1153 tunnel->dev = dev;
1154 strcpy(tunnel->parms.name, dev->name);
1155
1156 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1157 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1158
1159 ipip6_tunnel_bind_dev(dev);
1160 dev->tstats = alloc_percpu(struct pcpu_tstats);
1161 if (!dev->tstats)
1162 return -ENOMEM;
1163
1164 return 0;
1165 }
1166
1167 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1168 {
1169 struct ip_tunnel *tunnel = netdev_priv(dev);
1170 struct iphdr *iph = &tunnel->parms.iph;
1171 struct net *net = dev_net(dev);
1172 struct sit_net *sitn = net_generic(net, sit_net_id);
1173
1174 tunnel->dev = dev;
1175 strcpy(tunnel->parms.name, dev->name);
1176
1177 iph->version = 4;
1178 iph->protocol = IPPROTO_IPV6;
1179 iph->ihl = 5;
1180 iph->ttl = 64;
1181
1182 dev->tstats = alloc_percpu(struct pcpu_tstats);
1183 if (!dev->tstats)
1184 return -ENOMEM;
1185 dev_hold(dev);
1186 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1187 return 0;
1188 }
1189
1190 static struct xfrm_tunnel sit_handler __read_mostly = {
1191 .handler = ipip6_rcv,
1192 .err_handler = ipip6_err,
1193 .priority = 1,
1194 };
1195
1196 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1197 {
1198 int prio;
1199
1200 for (prio = 1; prio < 4; prio++) {
1201 int h;
1202 for (h = 0; h < HASH_SIZE; h++) {
1203 struct ip_tunnel *t;
1204
1205 t = rtnl_dereference(sitn->tunnels[prio][h]);
1206 while (t != NULL) {
1207 unregister_netdevice_queue(t->dev, head);
1208 t = rtnl_dereference(t->next);
1209 }
1210 }
1211 }
1212 }
1213
1214 static int __net_init sit_init_net(struct net *net)
1215 {
1216 struct sit_net *sitn = net_generic(net, sit_net_id);
1217 int err;
1218
1219 sitn->tunnels[0] = sitn->tunnels_wc;
1220 sitn->tunnels[1] = sitn->tunnels_l;
1221 sitn->tunnels[2] = sitn->tunnels_r;
1222 sitn->tunnels[3] = sitn->tunnels_r_l;
1223
1224 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1225 ipip6_tunnel_setup);
1226 if (!sitn->fb_tunnel_dev) {
1227 err = -ENOMEM;
1228 goto err_alloc_dev;
1229 }
1230 dev_net_set(sitn->fb_tunnel_dev, net);
1231
1232 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1233 if (err)
1234 goto err_dev_free;
1235
1236 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1237
1238 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1239 goto err_reg_dev;
1240
1241 return 0;
1242
1243 err_reg_dev:
1244 dev_put(sitn->fb_tunnel_dev);
1245 err_dev_free:
1246 ipip6_dev_free(sitn->fb_tunnel_dev);
1247 err_alloc_dev:
1248 return err;
1249 }
1250
1251 static void __net_exit sit_exit_net(struct net *net)
1252 {
1253 struct sit_net *sitn = net_generic(net, sit_net_id);
1254 LIST_HEAD(list);
1255
1256 rtnl_lock();
1257 sit_destroy_tunnels(sitn, &list);
1258 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1259 unregister_netdevice_many(&list);
1260 rtnl_unlock();
1261 }
1262
1263 static struct pernet_operations sit_net_ops = {
1264 .init = sit_init_net,
1265 .exit = sit_exit_net,
1266 .id = &sit_net_id,
1267 .size = sizeof(struct sit_net),
1268 };
1269
1270 static void __exit sit_cleanup(void)
1271 {
1272 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1273
1274 unregister_pernet_device(&sit_net_ops);
1275 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1276 }
1277
1278 static int __init sit_init(void)
1279 {
1280 int err;
1281
1282 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1283
1284 err = register_pernet_device(&sit_net_ops);
1285 if (err < 0)
1286 return err;
1287 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1288 if (err < 0) {
1289 unregister_pernet_device(&sit_net_ops);
1290 printk(KERN_INFO "sit init: Can't add protocol\n");
1291 }
1292 return err;
1293 }
1294
1295 module_init(sit_init);
1296 module_exit(sit_cleanup);
1297 MODULE_LICENSE("GPL");
1298 MODULE_ALIAS_NETDEV("sit0");
Ubuntu Artful kernel mirror