projects / deliverable / linux.git / blob
? search:


8cdcc2ad048c916e54ddeb0048a66f8574ad64a5
[deliverable/linux.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 <asm/uaccess.h>
32 #include <linux/init.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/if_ether.h>
35
36 #include <net/sock.h>
37 #include <net/snmp.h>
38
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_fib.h>
43 #include <net/ip6_route.h>
44 #include <net/ndisc.h>
45 #include <net/addrconf.h>
46 #include <net/ip.h>
47 #include <net/udp.h>
48 #include <net/icmp.h>
49 #include <net/ipip.h>
50 #include <net/inet_ecn.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55
56 /*
57 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58
59 For comments look at net/ipv4/ip_gre.c --ANK
60 */
61
62 #define HASH_SIZE 16
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64
65 static void ipip6_fb_tunnel_init(struct net_device *dev);
66 static void ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68
69 static int sit_net_id;
70 struct sit_net {
71 struct ip_tunnel *tunnels_r_l[HASH_SIZE];
72 struct ip_tunnel *tunnels_r[HASH_SIZE];
73 struct ip_tunnel *tunnels_l[HASH_SIZE];
74 struct ip_tunnel *tunnels_wc[1];
75 struct ip_tunnel **tunnels[4];
76
77 struct net_device *fb_tunnel_dev;
78 };
79
80 static DEFINE_RWLOCK(ipip6_lock);
81
82 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
83 struct net_device *dev, __be32 remote, __be32 local)
84 {
85 unsigned h0 = HASH(remote);
86 unsigned h1 = HASH(local);
87 struct ip_tunnel *t;
88 struct sit_net *sitn = net_generic(net, sit_net_id);
89
90 for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
91 if (local == t->parms.iph.saddr &&
92 remote == t->parms.iph.daddr &&
93 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
94 (t->dev->flags & IFF_UP))
95 return t;
96 }
97 for (t = sitn->tunnels_r[h0]; t; t = t->next) {
98 if (remote == t->parms.iph.daddr &&
99 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
100 (t->dev->flags & IFF_UP))
101 return t;
102 }
103 for (t = sitn->tunnels_l[h1]; t; t = t->next) {
104 if (local == t->parms.iph.saddr &&
105 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
106 (t->dev->flags & IFF_UP))
107 return t;
108 }
109 t = sitn->tunnels_wc[0];
110 if ((t != NULL) && (t->dev->flags & IFF_UP))
111 return t;
112 return NULL;
113 }
114
115 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
116 struct ip_tunnel_parm *parms)
117 {
118 __be32 remote = parms->iph.daddr;
119 __be32 local = parms->iph.saddr;
120 unsigned h = 0;
121 int prio = 0;
122
123 if (remote) {
124 prio |= 2;
125 h ^= HASH(remote);
126 }
127 if (local) {
128 prio |= 1;
129 h ^= HASH(local);
130 }
131 return &sitn->tunnels[prio][h];
132 }
133
134 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
135 struct ip_tunnel *t)
136 {
137 return __ipip6_bucket(sitn, &t->parms);
138 }
139
140 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
141 {
142 struct ip_tunnel **tp;
143
144 for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
145 if (t == *tp) {
146 write_lock_bh(&ipip6_lock);
147 *tp = t->next;
148 write_unlock_bh(&ipip6_lock);
149 break;
150 }
151 }
152 }
153
154 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
155 {
156 struct ip_tunnel **tp = ipip6_bucket(sitn, t);
157
158 t->next = *tp;
159 write_lock_bh(&ipip6_lock);
160 *tp = t;
161 write_unlock_bh(&ipip6_lock);
162 }
163
164 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
165 {
166 #ifdef CONFIG_IPV6_SIT_6RD
167 struct ip_tunnel *t = netdev_priv(dev);
168
169 if (t->dev == sitn->fb_tunnel_dev) {
170 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
171 t->ip6rd.relay_prefix = 0;
172 t->ip6rd.prefixlen = 16;
173 t->ip6rd.relay_prefixlen = 0;
174 } else {
175 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
176 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
177 }
178 #endif
179 }
180
181 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
182 struct ip_tunnel_parm *parms, int create)
183 {
184 __be32 remote = parms->iph.daddr;
185 __be32 local = parms->iph.saddr;
186 struct ip_tunnel *t, **tp, *nt;
187 struct net_device *dev;
188 char name[IFNAMSIZ];
189 struct sit_net *sitn = net_generic(net, sit_net_id);
190
191 for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
192 if (local == t->parms.iph.saddr &&
193 remote == t->parms.iph.daddr &&
194 parms->link == t->parms.link) {
195 if (create)
196 return NULL;
197 else
198 return t;
199 }
200 }
201 if (!create)
202 goto failed;
203
204 if (parms->name[0])
205 strlcpy(name, parms->name, IFNAMSIZ);
206 else
207 sprintf(name, "sit%%d");
208
209 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
210 if (dev == NULL)
211 return NULL;
212
213 dev_net_set(dev, net);
214
215 if (strchr(name, '%')) {
216 if (dev_alloc_name(dev, name) < 0)
217 goto failed_free;
218 }
219
220 nt = netdev_priv(dev);
221
222 nt->parms = *parms;
223 ipip6_tunnel_init(dev);
224 ipip6_tunnel_clone_6rd(dev, sitn);
225
226 if (parms->i_flags & SIT_ISATAP)
227 dev->priv_flags |= IFF_ISATAP;
228
229 if (register_netdevice(dev) < 0)
230 goto failed_free;
231
232 dev_hold(dev);
233
234 ipip6_tunnel_link(sitn, nt);
235 return nt;
236
237 failed_free:
238 free_netdev(dev);
239 failed:
240 return NULL;
241 }
242
243 static DEFINE_SPINLOCK(ipip6_prl_lock);
244
245 #define for_each_prl_rcu(start) \
246 for (prl = rcu_dereference(start); \
247 prl; \
248 prl = rcu_dereference(prl->next))
249
250 static struct ip_tunnel_prl_entry *
251 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
252 {
253 struct ip_tunnel_prl_entry *prl;
254
255 for_each_prl_rcu(t->prl)
256 if (prl->addr == addr)
257 break;
258 return prl;
259
260 }
261
262 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
263 struct ip_tunnel_prl __user *a)
264 {
265 struct ip_tunnel_prl kprl, *kp;
266 struct ip_tunnel_prl_entry *prl;
267 unsigned int cmax, c = 0, ca, len;
268 int ret = 0;
269
270 if (copy_from_user(&kprl, a, sizeof(kprl)))
271 return -EFAULT;
272 cmax = kprl.datalen / sizeof(kprl);
273 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
274 cmax = 1;
275
276 /* For simple GET or for root users,
277 * we try harder to allocate.
278 */
279 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
280 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
281 NULL;
282
283 rcu_read_lock();
284
285 ca = t->prl_count < cmax ? t->prl_count : cmax;
286
287 if (!kp) {
288 /* We don't try hard to allocate much memory for
289 * non-root users.
290 * For root users, retry allocating enough memory for
291 * the answer.
292 */
293 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
294 if (!kp) {
295 ret = -ENOMEM;
296 goto out;
297 }
298 }
299
300 c = 0;
301 for_each_prl_rcu(t->prl) {
302 if (c >= cmax)
303 break;
304 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
305 continue;
306 kp[c].addr = prl->addr;
307 kp[c].flags = prl->flags;
308 c++;
309 if (kprl.addr != htonl(INADDR_ANY))
310 break;
311 }
312 out:
313 rcu_read_unlock();
314
315 len = sizeof(*kp) * c;
316 ret = 0;
317 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
318 ret = -EFAULT;
319
320 kfree(kp);
321
322 return ret;
323 }
324
325 static int
326 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
327 {
328 struct ip_tunnel_prl_entry *p;
329 int err = 0;
330
331 if (a->addr == htonl(INADDR_ANY))
332 return -EINVAL;
333
334 spin_lock(&ipip6_prl_lock);
335
336 for (p = t->prl; p; p = p->next) {
337 if (p->addr == a->addr) {
338 if (chg) {
339 p->flags = a->flags;
340 goto out;
341 }
342 err = -EEXIST;
343 goto out;
344 }
345 }
346
347 if (chg) {
348 err = -ENXIO;
349 goto out;
350 }
351
352 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
353 if (!p) {
354 err = -ENOBUFS;
355 goto out;
356 }
357
358 INIT_RCU_HEAD(&p->rcu_head);
359 p->next = t->prl;
360 p->addr = a->addr;
361 p->flags = a->flags;
362 t->prl_count++;
363 rcu_assign_pointer(t->prl, p);
364 out:
365 spin_unlock(&ipip6_prl_lock);
366 return err;
367 }
368
369 static void prl_entry_destroy_rcu(struct rcu_head *head)
370 {
371 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
372 }
373
374 static void prl_list_destroy_rcu(struct rcu_head *head)
375 {
376 struct ip_tunnel_prl_entry *p, *n;
377
378 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
379 do {
380 n = p->next;
381 kfree(p);
382 p = n;
383 } while (p);
384 }
385
386 static int
387 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
388 {
389 struct ip_tunnel_prl_entry *x, **p;
390 int err = 0;
391
392 spin_lock(&ipip6_prl_lock);
393
394 if (a && a->addr != htonl(INADDR_ANY)) {
395 for (p = &t->prl; *p; p = &(*p)->next) {
396 if ((*p)->addr == a->addr) {
397 x = *p;
398 *p = x->next;
399 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
400 t->prl_count--;
401 goto out;
402 }
403 }
404 err = -ENXIO;
405 } else {
406 if (t->prl) {
407 t->prl_count = 0;
408 x = t->prl;
409 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
410 t->prl = NULL;
411 }
412 }
413 out:
414 spin_unlock(&ipip6_prl_lock);
415 return err;
416 }
417
418 static int
419 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
420 {
421 struct ip_tunnel_prl_entry *p;
422 int ok = 1;
423
424 rcu_read_lock();
425 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
426 if (p) {
427 if (p->flags & PRL_DEFAULT)
428 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
429 else
430 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
431 } else {
432 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
433 if (ipv6_addr_is_isatap(addr6) &&
434 (addr6->s6_addr32[3] == iph->saddr) &&
435 ipv6_chk_prefix(addr6, t->dev))
436 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
437 else
438 ok = 0;
439 }
440 rcu_read_unlock();
441 return ok;
442 }
443
444 static void ipip6_tunnel_uninit(struct net_device *dev)
445 {
446 struct net *net = dev_net(dev);
447 struct sit_net *sitn = net_generic(net, sit_net_id);
448
449 if (dev == sitn->fb_tunnel_dev) {
450 write_lock_bh(&ipip6_lock);
451 sitn->tunnels_wc[0] = NULL;
452 write_unlock_bh(&ipip6_lock);
453 dev_put(dev);
454 } else {
455 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
456 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
457 dev_put(dev);
458 }
459 }
460
461
462 static int ipip6_err(struct sk_buff *skb, u32 info)
463 {
464
465 /* All the routers (except for Linux) return only
466 8 bytes of packet payload. It means, that precise relaying of
467 ICMP in the real Internet is absolutely infeasible.
468 */
469 struct iphdr *iph = (struct iphdr*)skb->data;
470 const int type = icmp_hdr(skb)->type;
471 const int code = icmp_hdr(skb)->code;
472 struct ip_tunnel *t;
473 int err;
474
475 switch (type) {
476 default:
477 case ICMP_PARAMETERPROB:
478 return 0;
479
480 case ICMP_DEST_UNREACH:
481 switch (code) {
482 case ICMP_SR_FAILED:
483 case ICMP_PORT_UNREACH:
484 /* Impossible event. */
485 return 0;
486 case ICMP_FRAG_NEEDED:
487 /* Soft state for pmtu is maintained by IP core. */
488 return 0;
489 default:
490 /* All others are translated to HOST_UNREACH.
491 rfc2003 contains "deep thoughts" about NET_UNREACH,
492 I believe they are just ether pollution. --ANK
493 */
494 break;
495 }
496 break;
497 case ICMP_TIME_EXCEEDED:
498 if (code != ICMP_EXC_TTL)
499 return 0;
500 break;
501 }
502
503 err = -ENOENT;
504
505 read_lock(&ipip6_lock);
506 t = ipip6_tunnel_lookup(dev_net(skb->dev),
507 skb->dev,
508 iph->daddr,
509 iph->saddr);
510 if (t == NULL || t->parms.iph.daddr == 0)
511 goto out;
512
513 err = 0;
514 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
515 goto out;
516
517 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
518 t->err_count++;
519 else
520 t->err_count = 1;
521 t->err_time = jiffies;
522 out:
523 read_unlock(&ipip6_lock);
524 return err;
525 }
526
527 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
528 {
529 if (INET_ECN_is_ce(iph->tos))
530 IP6_ECN_set_ce(ipv6_hdr(skb));
531 }
532
533 static int ipip6_rcv(struct sk_buff *skb)
534 {
535 struct iphdr *iph;
536 struct ip_tunnel *tunnel;
537
538 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
539 goto out;
540
541 iph = ip_hdr(skb);
542
543 read_lock(&ipip6_lock);
544 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
545 iph->saddr, iph->daddr);
546 if (tunnel != NULL) {
547 secpath_reset(skb);
548 skb->mac_header = skb->network_header;
549 skb_reset_network_header(skb);
550 IPCB(skb)->flags = 0;
551 skb->protocol = htons(ETH_P_IPV6);
552 skb->pkt_type = PACKET_HOST;
553
554 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
555 !isatap_chksrc(skb, iph, tunnel)) {
556 tunnel->dev->stats.rx_errors++;
557 read_unlock(&ipip6_lock);
558 kfree_skb(skb);
559 return 0;
560 }
561 tunnel->dev->stats.rx_packets++;
562 tunnel->dev->stats.rx_bytes += skb->len;
563 skb->dev = tunnel->dev;
564 skb_dst_drop(skb);
565 nf_reset(skb);
566 ipip6_ecn_decapsulate(iph, skb);
567 netif_rx(skb);
568 read_unlock(&ipip6_lock);
569 return 0;
570 }
571
572 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
573 read_unlock(&ipip6_lock);
574 out:
575 kfree_skb(skb);
576 return 0;
577 }
578
579 /*
580 * Returns the embedded IPv4 address if the IPv6 address
581 * comes from 6rd / 6to4 (RFC 3056) addr space.
582 */
583 static inline
584 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
585 {
586 __be32 dst = 0;
587
588 #ifdef CONFIG_IPV6_SIT_6RD
589 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
590 tunnel->ip6rd.prefixlen)) {
591 unsigned pbw0, pbi0;
592 int pbi1;
593 u32 d;
594
595 pbw0 = tunnel->ip6rd.prefixlen >> 5;
596 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
597
598 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
599 tunnel->ip6rd.relay_prefixlen;
600
601 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
602 if (pbi1 > 0)
603 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
604 (32 - pbi1);
605
606 dst = tunnel->ip6rd.relay_prefix | htonl(d);
607 }
608 #else
609 if (v6dst->s6_addr16[0] == htons(0x2002)) {
610 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
611 memcpy(&dst, &v6dst->s6_addr16[1], 4);
612 }
613 #endif
614 return dst;
615 }
616
617 /*
618 * This function assumes it is being called from dev_queue_xmit()
619 * and that skb is filled properly by that function.
620 */
621
622 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
623 struct net_device *dev)
624 {
625 struct ip_tunnel *tunnel = netdev_priv(dev);
626 struct net_device_stats *stats = &dev->stats;
627 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
628 struct iphdr *tiph = &tunnel->parms.iph;
629 struct ipv6hdr *iph6 = ipv6_hdr(skb);
630 u8 tos = tunnel->parms.iph.tos;
631 struct rtable *rt; /* Route to the other host */
632 struct net_device *tdev; /* Device to other host */
633 struct iphdr *iph; /* Our new IP header */
634 unsigned int max_headroom; /* The extra header space needed */
635 __be32 dst = tiph->daddr;
636 int mtu;
637 struct in6_addr *addr6;
638 int addr_type;
639
640 if (skb->protocol != htons(ETH_P_IPV6))
641 goto tx_error;
642
643 /* ISATAP (RFC4214) - must come before 6to4 */
644 if (dev->priv_flags & IFF_ISATAP) {
645 struct neighbour *neigh = NULL;
646
647 if (skb_dst(skb))
648 neigh = skb_dst(skb)->neighbour;
649
650 if (neigh == NULL) {
651 if (net_ratelimit())
652 printk(KERN_DEBUG "sit: nexthop == NULL\n");
653 goto tx_error;
654 }
655
656 addr6 = (struct in6_addr*)&neigh->primary_key;
657 addr_type = ipv6_addr_type(addr6);
658
659 if ((addr_type & IPV6_ADDR_UNICAST) &&
660 ipv6_addr_is_isatap(addr6))
661 dst = addr6->s6_addr32[3];
662 else
663 goto tx_error;
664 }
665
666 if (!dst)
667 dst = try_6rd(&iph6->daddr, tunnel);
668
669 if (!dst) {
670 struct neighbour *neigh = NULL;
671
672 if (skb_dst(skb))
673 neigh = skb_dst(skb)->neighbour;
674
675 if (neigh == NULL) {
676 if (net_ratelimit())
677 printk(KERN_DEBUG "sit: nexthop == NULL\n");
678 goto tx_error;
679 }
680
681 addr6 = (struct in6_addr*)&neigh->primary_key;
682 addr_type = ipv6_addr_type(addr6);
683
684 if (addr_type == IPV6_ADDR_ANY) {
685 addr6 = &ipv6_hdr(skb)->daddr;
686 addr_type = ipv6_addr_type(addr6);
687 }
688
689 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
690 goto tx_error_icmp;
691
692 dst = addr6->s6_addr32[3];
693 }
694
695 {
696 struct flowi fl = { .nl_u = { .ip4_u =
697 { .daddr = dst,
698 .saddr = tiph->saddr,
699 .tos = RT_TOS(tos) } },
700 .oif = tunnel->parms.link,
701 .proto = IPPROTO_IPV6 };
702 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
703 stats->tx_carrier_errors++;
704 goto tx_error_icmp;
705 }
706 }
707 if (rt->rt_type != RTN_UNICAST) {
708 ip_rt_put(rt);
709 stats->tx_carrier_errors++;
710 goto tx_error_icmp;
711 }
712 tdev = rt->u.dst.dev;
713
714 if (tdev == dev) {
715 ip_rt_put(rt);
716 stats->collisions++;
717 goto tx_error;
718 }
719
720 if (tiph->frag_off)
721 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
722 else
723 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
724
725 if (mtu < 68) {
726 stats->collisions++;
727 ip_rt_put(rt);
728 goto tx_error;
729 }
730 if (mtu < IPV6_MIN_MTU)
731 mtu = IPV6_MIN_MTU;
732 if (tunnel->parms.iph.daddr && skb_dst(skb))
733 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
734
735 if (skb->len > mtu) {
736 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
737 ip_rt_put(rt);
738 goto tx_error;
739 }
740
741 if (tunnel->err_count > 0) {
742 if (time_before(jiffies,
743 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
744 tunnel->err_count--;
745 dst_link_failure(skb);
746 } else
747 tunnel->err_count = 0;
748 }
749
750 /*
751 * Okay, now see if we can stuff it in the buffer as-is.
752 */
753 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
754
755 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
756 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
757 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
758 if (!new_skb) {
759 ip_rt_put(rt);
760 txq->tx_dropped++;
761 dev_kfree_skb(skb);
762 return NETDEV_TX_OK;
763 }
764 if (skb->sk)
765 skb_set_owner_w(new_skb, skb->sk);
766 dev_kfree_skb(skb);
767 skb = new_skb;
768 iph6 = ipv6_hdr(skb);
769 }
770
771 skb->transport_header = skb->network_header;
772 skb_push(skb, sizeof(struct iphdr));
773 skb_reset_network_header(skb);
774 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
775 IPCB(skb)->flags = 0;
776 skb_dst_drop(skb);
777 skb_dst_set(skb, &rt->u.dst);
778
779 /*
780 * Push down and install the IPIP header.
781 */
782
783 iph = ip_hdr(skb);
784 iph->version = 4;
785 iph->ihl = sizeof(struct iphdr)>>2;
786 if (mtu > IPV6_MIN_MTU)
787 iph->frag_off = tiph->frag_off;
788 else
789 iph->frag_off = 0;
790
791 iph->protocol = IPPROTO_IPV6;
792 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
793 iph->daddr = rt->rt_dst;
794 iph->saddr = rt->rt_src;
795
796 if ((iph->ttl = tiph->ttl) == 0)
797 iph->ttl = iph6->hop_limit;
798
799 nf_reset(skb);
800
801 IPTUNNEL_XMIT();
802 return NETDEV_TX_OK;
803
804 tx_error_icmp:
805 dst_link_failure(skb);
806 tx_error:
807 stats->tx_errors++;
808 dev_kfree_skb(skb);
809 return NETDEV_TX_OK;
810 }
811
812 static void ipip6_tunnel_bind_dev(struct net_device *dev)
813 {
814 struct net_device *tdev = NULL;
815 struct ip_tunnel *tunnel;
816 struct iphdr *iph;
817
818 tunnel = netdev_priv(dev);
819 iph = &tunnel->parms.iph;
820
821 if (iph->daddr) {
822 struct flowi fl = { .nl_u = { .ip4_u =
823 { .daddr = iph->daddr,
824 .saddr = iph->saddr,
825 .tos = RT_TOS(iph->tos) } },
826 .oif = tunnel->parms.link,
827 .proto = IPPROTO_IPV6 };
828 struct rtable *rt;
829 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
830 tdev = rt->u.dst.dev;
831 ip_rt_put(rt);
832 }
833 dev->flags |= IFF_POINTOPOINT;
834 }
835
836 if (!tdev && tunnel->parms.link)
837 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
838
839 if (tdev) {
840 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
841 dev->mtu = tdev->mtu - sizeof(struct iphdr);
842 if (dev->mtu < IPV6_MIN_MTU)
843 dev->mtu = IPV6_MIN_MTU;
844 }
845 dev->iflink = tunnel->parms.link;
846 }
847
848 static int
849 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
850 {
851 int err = 0;
852 struct ip_tunnel_parm p;
853 struct ip_tunnel_prl prl;
854 struct ip_tunnel *t;
855 struct net *net = dev_net(dev);
856 struct sit_net *sitn = net_generic(net, sit_net_id);
857 #ifdef CONFIG_IPV6_SIT_6RD
858 struct ip_tunnel_6rd ip6rd;
859 #endif
860
861 switch (cmd) {
862 case SIOCGETTUNNEL:
863 #ifdef CONFIG_IPV6_SIT_6RD
864 case SIOCGET6RD:
865 #endif
866 t = NULL;
867 if (dev == sitn->fb_tunnel_dev) {
868 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
869 err = -EFAULT;
870 break;
871 }
872 t = ipip6_tunnel_locate(net, &p, 0);
873 }
874 if (t == NULL)
875 t = netdev_priv(dev);
876
877 err = -EFAULT;
878 if (cmd == SIOCGETTUNNEL) {
879 memcpy(&p, &t->parms, sizeof(p));
880 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
881 sizeof(p)))
882 goto done;
883 #ifdef CONFIG_IPV6_SIT_6RD
884 } else {
885 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
886 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
887 ip6rd.prefixlen = t->ip6rd.prefixlen;
888 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
889 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
890 sizeof(ip6rd)))
891 goto done;
892 #endif
893 }
894 err = 0;
895 break;
896
897 case SIOCADDTUNNEL:
898 case SIOCCHGTUNNEL:
899 err = -EPERM;
900 if (!capable(CAP_NET_ADMIN))
901 goto done;
902
903 err = -EFAULT;
904 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
905 goto done;
906
907 err = -EINVAL;
908 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
909 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
910 goto done;
911 if (p.iph.ttl)
912 p.iph.frag_off |= htons(IP_DF);
913
914 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
915
916 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
917 if (t != NULL) {
918 if (t->dev != dev) {
919 err = -EEXIST;
920 break;
921 }
922 } else {
923 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
924 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
925 err = -EINVAL;
926 break;
927 }
928 t = netdev_priv(dev);
929 ipip6_tunnel_unlink(sitn, t);
930 t->parms.iph.saddr = p.iph.saddr;
931 t->parms.iph.daddr = p.iph.daddr;
932 memcpy(dev->dev_addr, &p.iph.saddr, 4);
933 memcpy(dev->broadcast, &p.iph.daddr, 4);
934 ipip6_tunnel_link(sitn, t);
935 netdev_state_change(dev);
936 }
937 }
938
939 if (t) {
940 err = 0;
941 if (cmd == SIOCCHGTUNNEL) {
942 t->parms.iph.ttl = p.iph.ttl;
943 t->parms.iph.tos = p.iph.tos;
944 if (t->parms.link != p.link) {
945 t->parms.link = p.link;
946 ipip6_tunnel_bind_dev(dev);
947 netdev_state_change(dev);
948 }
949 }
950 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
951 err = -EFAULT;
952 } else
953 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
954 break;
955
956 case SIOCDELTUNNEL:
957 err = -EPERM;
958 if (!capable(CAP_NET_ADMIN))
959 goto done;
960
961 if (dev == sitn->fb_tunnel_dev) {
962 err = -EFAULT;
963 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
964 goto done;
965 err = -ENOENT;
966 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
967 goto done;
968 err = -EPERM;
969 if (t == netdev_priv(sitn->fb_tunnel_dev))
970 goto done;
971 dev = t->dev;
972 }
973 unregister_netdevice(dev);
974 err = 0;
975 break;
976
977 case SIOCGETPRL:
978 err = -EINVAL;
979 if (dev == sitn->fb_tunnel_dev)
980 goto done;
981 err = -ENOENT;
982 if (!(t = netdev_priv(dev)))
983 goto done;
984 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
985 break;
986
987 case SIOCADDPRL:
988 case SIOCDELPRL:
989 case SIOCCHGPRL:
990 err = -EPERM;
991 if (!capable(CAP_NET_ADMIN))
992 goto done;
993 err = -EINVAL;
994 if (dev == sitn->fb_tunnel_dev)
995 goto done;
996 err = -EFAULT;
997 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
998 goto done;
999 err = -ENOENT;
1000 if (!(t = netdev_priv(dev)))
1001 goto done;
1002
1003 switch (cmd) {
1004 case SIOCDELPRL:
1005 err = ipip6_tunnel_del_prl(t, &prl);
1006 break;
1007 case SIOCADDPRL:
1008 case SIOCCHGPRL:
1009 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1010 break;
1011 }
1012 netdev_state_change(dev);
1013 break;
1014
1015 #ifdef CONFIG_IPV6_SIT_6RD
1016 case SIOCADD6RD:
1017 case SIOCCHG6RD:
1018 case SIOCDEL6RD:
1019 err = -EPERM;
1020 if (!capable(CAP_NET_ADMIN))
1021 goto done;
1022
1023 err = -EFAULT;
1024 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1025 sizeof(ip6rd)))
1026 goto done;
1027
1028 t = netdev_priv(dev);
1029
1030 if (cmd != SIOCDEL6RD) {
1031 struct in6_addr prefix;
1032 __be32 relay_prefix;
1033
1034 err = -EINVAL;
1035 if (ip6rd.relay_prefixlen > 32 ||
1036 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1037 goto done;
1038
1039 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1040 ip6rd.prefixlen);
1041 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1042 goto done;
1043 if (ip6rd.relay_prefixlen)
1044 relay_prefix = ip6rd.relay_prefix &
1045 htonl(0xffffffffUL <<
1046 (32 - ip6rd.relay_prefixlen));
1047 else
1048 relay_prefix = 0;
1049 if (relay_prefix != ip6rd.relay_prefix)
1050 goto done;
1051
1052 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1053 t->ip6rd.relay_prefix = relay_prefix;
1054 t->ip6rd.prefixlen = ip6rd.prefixlen;
1055 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1056 } else
1057 ipip6_tunnel_clone_6rd(dev, sitn);
1058
1059 err = 0;
1060 break;
1061 #endif
1062
1063 default:
1064 err = -EINVAL;
1065 }
1066
1067 done:
1068 return err;
1069 }
1070
1071 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1072 {
1073 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1074 return -EINVAL;
1075 dev->mtu = new_mtu;
1076 return 0;
1077 }
1078
1079 static const struct net_device_ops ipip6_netdev_ops = {
1080 .ndo_uninit = ipip6_tunnel_uninit,
1081 .ndo_start_xmit = ipip6_tunnel_xmit,
1082 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1083 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1084 };
1085
1086 static void ipip6_tunnel_setup(struct net_device *dev)
1087 {
1088 dev->netdev_ops = &ipip6_netdev_ops;
1089 dev->destructor = free_netdev;
1090
1091 dev->type = ARPHRD_SIT;
1092 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1093 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1094 dev->flags = IFF_NOARP;
1095 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1096 dev->iflink = 0;
1097 dev->addr_len = 4;
1098 dev->features |= NETIF_F_NETNS_LOCAL;
1099 }
1100
1101 static void ipip6_tunnel_init(struct net_device *dev)
1102 {
1103 struct ip_tunnel *tunnel = netdev_priv(dev);
1104
1105 tunnel->dev = dev;
1106 strcpy(tunnel->parms.name, dev->name);
1107
1108 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1109 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1110
1111 ipip6_tunnel_bind_dev(dev);
1112 }
1113
1114 static void ipip6_fb_tunnel_init(struct net_device *dev)
1115 {
1116 struct ip_tunnel *tunnel = netdev_priv(dev);
1117 struct iphdr *iph = &tunnel->parms.iph;
1118 struct net *net = dev_net(dev);
1119 struct sit_net *sitn = net_generic(net, sit_net_id);
1120
1121 tunnel->dev = dev;
1122 strcpy(tunnel->parms.name, dev->name);
1123
1124 iph->version = 4;
1125 iph->protocol = IPPROTO_IPV6;
1126 iph->ihl = 5;
1127 iph->ttl = 64;
1128
1129 dev_hold(dev);
1130 sitn->tunnels_wc[0] = tunnel;
1131 }
1132
1133 static struct xfrm_tunnel sit_handler = {
1134 .handler = ipip6_rcv,
1135 .err_handler = ipip6_err,
1136 .priority = 1,
1137 };
1138
1139 static void sit_destroy_tunnels(struct sit_net *sitn)
1140 {
1141 int prio;
1142
1143 for (prio = 1; prio < 4; prio++) {
1144 int h;
1145 for (h = 0; h < HASH_SIZE; h++) {
1146 struct ip_tunnel *t;
1147 while ((t = sitn->tunnels[prio][h]) != NULL)
1148 unregister_netdevice(t->dev);
1149 }
1150 }
1151 }
1152
1153 static int sit_init_net(struct net *net)
1154 {
1155 int err;
1156 struct sit_net *sitn;
1157
1158 err = -ENOMEM;
1159 sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1160 if (sitn == NULL)
1161 goto err_alloc;
1162
1163 err = net_assign_generic(net, sit_net_id, sitn);
1164 if (err < 0)
1165 goto err_assign;
1166
1167 sitn->tunnels[0] = sitn->tunnels_wc;
1168 sitn->tunnels[1] = sitn->tunnels_l;
1169 sitn->tunnels[2] = sitn->tunnels_r;
1170 sitn->tunnels[3] = sitn->tunnels_r_l;
1171
1172 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1173 ipip6_tunnel_setup);
1174 if (!sitn->fb_tunnel_dev) {
1175 err = -ENOMEM;
1176 goto err_alloc_dev;
1177 }
1178 dev_net_set(sitn->fb_tunnel_dev, net);
1179
1180 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1181 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1182
1183 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1184 goto err_reg_dev;
1185
1186 return 0;
1187
1188 err_reg_dev:
1189 dev_put(sitn->fb_tunnel_dev);
1190 free_netdev(sitn->fb_tunnel_dev);
1191 err_alloc_dev:
1192 /* nothing */
1193 err_assign:
1194 kfree(sitn);
1195 err_alloc:
1196 return err;
1197 }
1198
1199 static void sit_exit_net(struct net *net)
1200 {
1201 struct sit_net *sitn;
1202
1203 sitn = net_generic(net, sit_net_id);
1204 rtnl_lock();
1205 sit_destroy_tunnels(sitn);
1206 unregister_netdevice(sitn->fb_tunnel_dev);
1207 rtnl_unlock();
1208 kfree(sitn);
1209 }
1210
1211 static struct pernet_operations sit_net_ops = {
1212 .init = sit_init_net,
1213 .exit = sit_exit_net,
1214 };
1215
1216 static void __exit sit_cleanup(void)
1217 {
1218 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1219
1220 unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1221 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1222 }
1223
1224 static int __init sit_init(void)
1225 {
1226 int err;
1227
1228 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1229
1230 if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1231 printk(KERN_INFO "sit init: Can't add protocol\n");
1232 return -EAGAIN;
1233 }
1234
1235 err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1236 if (err < 0)
1237 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1238
1239 return err;
1240 }
1241
1242 module_init(sit_init);
1243 module_exit(sit_cleanup);
1244 MODULE_LICENSE("GPL");
1245 MODULE_ALIAS("sit0");
Linux kernel - Deliverables
This page took 0.056278 seconds and 4 git commands to generate.