1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
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
59
60
61
62
63#define HASH_SIZE 16
64#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
65
66static int ipip6_tunnel_init(struct net_device *dev);
67static void ipip6_tunnel_setup(struct net_device *dev);
68static void ipip6_dev_free(struct net_device *dev);
69
70static int sit_net_id __read_mostly;
71struct 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
83
84
85#define for_each_ip_tunnel_rcu(start) \
86 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
87
88
89struct pcpu_tstats {
90 unsigned long rx_packets;
91 unsigned long rx_bytes;
92 unsigned long tx_packets;
93 unsigned long tx_bytes;
94};
95
96static 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
117
118static 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
151static 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
170static 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
176static 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
191static 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
199static 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
216static 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
276failed_free:
277 ipip6_dev_free(dev);
278failed:
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
287static 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
299static 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
314
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
326
327
328
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 }
349out:
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
362static int
363ipip6_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);
400out:
401 return err;
402}
403
404static void prl_entry_destroy_rcu(struct rcu_head *head)
405{
406 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
407}
408
409static 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 = p->next;
416 kfree(p);
417 p = n;
418 } while (p);
419}
420
421static int
422ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
423{
424 struct ip_tunnel_prl_entry *x, **p;
425 int err = 0;
426
427 ASSERT_RTNL();
428
429 if (a && a->addr != htonl(INADDR_ANY)) {
430 for (p = &t->prl; *p; p = &(*p)->next) {
431 if ((*p)->addr == a->addr) {
432 x = *p;
433 *p = x->next;
434 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
435 t->prl_count--;
436 goto out;
437 }
438 }
439 err = -ENXIO;
440 } else {
441 if (t->prl) {
442 t->prl_count = 0;
443 x = t->prl;
444 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
445 t->prl = NULL;
446 }
447 }
448out:
449 return err;
450}
451
452static int
453isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
454{
455 struct ip_tunnel_prl_entry *p;
456 int ok = 1;
457
458 rcu_read_lock();
459 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
460 if (p) {
461 if (p->flags & PRL_DEFAULT)
462 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
463 else
464 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
465 } else {
466 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
467 if (ipv6_addr_is_isatap(addr6) &&
468 (addr6->s6_addr32[3] == iph->saddr) &&
469 ipv6_chk_prefix(addr6, t->dev))
470 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
471 else
472 ok = 0;
473 }
474 rcu_read_unlock();
475 return ok;
476}
477
478static void ipip6_tunnel_uninit(struct net_device *dev)
479{
480 struct net *net = dev_net(dev);
481 struct sit_net *sitn = net_generic(net, sit_net_id);
482
483 if (dev == sitn->fb_tunnel_dev) {
484 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
485 } else {
486 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
487 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
488 }
489 dev_put(dev);
490}
491
492
493static int ipip6_err(struct sk_buff *skb, u32 info)
494{
495
496
497
498
499
500 struct iphdr *iph = (struct iphdr*)skb->data;
501 const int type = icmp_hdr(skb)->type;
502 const int code = icmp_hdr(skb)->code;
503 struct ip_tunnel *t;
504 int err;
505
506 switch (type) {
507 default:
508 case ICMP_PARAMETERPROB:
509 return 0;
510
511 case ICMP_DEST_UNREACH:
512 switch (code) {
513 case ICMP_SR_FAILED:
514 case ICMP_PORT_UNREACH:
515
516 return 0;
517 case ICMP_FRAG_NEEDED:
518
519 return 0;
520 default:
521
522
523
524
525 break;
526 }
527 break;
528 case ICMP_TIME_EXCEEDED:
529 if (code != ICMP_EXC_TTL)
530 return 0;
531 break;
532 }
533
534 err = -ENOENT;
535
536 rcu_read_lock();
537 t = ipip6_tunnel_lookup(dev_net(skb->dev),
538 skb->dev,
539 iph->daddr,
540 iph->saddr);
541 if (t == NULL || t->parms.iph.daddr == 0)
542 goto out;
543
544 err = 0;
545 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
546 goto out;
547
548 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
549 t->err_count++;
550 else
551 t->err_count = 1;
552 t->err_time = jiffies;
553out:
554 rcu_read_unlock();
555 return err;
556}
557
558static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
559{
560 if (INET_ECN_is_ce(iph->tos))
561 IP6_ECN_set_ce(ipv6_hdr(skb));
562}
563
564static int ipip6_rcv(struct sk_buff *skb)
565{
566 struct iphdr *iph;
567 struct ip_tunnel *tunnel;
568
569 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
570 goto out;
571
572 iph = ip_hdr(skb);
573
574 rcu_read_lock();
575 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
576 iph->saddr, iph->daddr);
577 if (tunnel != NULL) {
578 struct pcpu_tstats *tstats;
579
580 secpath_reset(skb);
581 skb->mac_header = skb->network_header;
582 skb_reset_network_header(skb);
583 IPCB(skb)->flags = 0;
584 skb->protocol = htons(ETH_P_IPV6);
585 skb->pkt_type = PACKET_HOST;
586
587 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
588 !isatap_chksrc(skb, iph, tunnel)) {
589 tunnel->dev->stats.rx_errors++;
590 rcu_read_unlock();
591 kfree_skb(skb);
592 return 0;
593 }
594
595 tstats = this_cpu_ptr(tunnel->dev->tstats);
596 tstats->rx_packets++;
597 tstats->rx_bytes += skb->len;
598
599 __skb_tunnel_rx(skb, tunnel->dev);
600
601 ipip6_ecn_decapsulate(iph, skb);
602
603 netif_rx(skb);
604
605 rcu_read_unlock();
606 return 0;
607 }
608
609
610 rcu_read_unlock();
611 return 1;
612out:
613 kfree_skb(skb);
614 return 0;
615}
616
617
618
619
620
621static inline
622__be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
623{
624 __be32 dst = 0;
625
626#ifdef CONFIG_IPV6_SIT_6RD
627 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
628 tunnel->ip6rd.prefixlen)) {
629 unsigned int pbw0, pbi0;
630 int pbi1;
631 u32 d;
632
633 pbw0 = tunnel->ip6rd.prefixlen >> 5;
634 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
635
636 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
637 tunnel->ip6rd.relay_prefixlen;
638
639 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
640 if (pbi1 > 0)
641 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
642 (32 - pbi1);
643
644 dst = tunnel->ip6rd.relay_prefix | htonl(d);
645 }
646#else
647 if (v6dst->s6_addr16[0] == htons(0x2002)) {
648
649 memcpy(&dst, &v6dst->s6_addr16[1], 4);
650 }
651#endif
652 return dst;
653}
654
655
656
657
658
659
660static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
661 struct net_device *dev)
662{
663 struct ip_tunnel *tunnel = netdev_priv(dev);
664 struct pcpu_tstats *tstats;
665 struct iphdr *tiph = &tunnel->parms.iph;
666 struct ipv6hdr *iph6 = ipv6_hdr(skb);
667 u8 tos = tunnel->parms.iph.tos;
668 __be16 df = tiph->frag_off;
669 struct rtable *rt;
670 struct net_device *tdev;
671 struct iphdr *iph;
672 unsigned int max_headroom;
673 __be32 dst = tiph->daddr;
674 int mtu;
675 struct in6_addr *addr6;
676 int addr_type;
677
678 if (skb->protocol != htons(ETH_P_IPV6))
679 goto tx_error;
680
681
682 if (dev->priv_flags & IFF_ISATAP) {
683 struct neighbour *neigh = NULL;
684
685 if (skb_dst(skb))
686 neigh = skb_dst(skb)->neighbour;
687
688 if (neigh == NULL) {
689 if (net_ratelimit())
690 printk(KERN_DEBUG "sit: nexthop == NULL\n");
691 goto tx_error;
692 }
693
694 addr6 = (struct in6_addr*)&neigh->primary_key;
695 addr_type = ipv6_addr_type(addr6);
696
697 if ((addr_type & IPV6_ADDR_UNICAST) &&
698 ipv6_addr_is_isatap(addr6))
699 dst = addr6->s6_addr32[3];
700 else
701 goto tx_error;
702 }
703
704 if (!dst)
705 dst = try_6rd(&iph6->daddr, tunnel);
706
707 if (!dst) {
708 struct neighbour *neigh = NULL;
709
710 if (skb_dst(skb))
711 neigh = skb_dst(skb)->neighbour;
712
713 if (neigh == NULL) {
714 if (net_ratelimit())
715 printk(KERN_DEBUG "sit: nexthop == NULL\n");
716 goto tx_error;
717 }
718
719 addr6 = (struct in6_addr*)&neigh->primary_key;
720 addr_type = ipv6_addr_type(addr6);
721
722 if (addr_type == IPV6_ADDR_ANY) {
723 addr6 = &ipv6_hdr(skb)->daddr;
724 addr_type = ipv6_addr_type(addr6);
725 }
726
727 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
728 goto tx_error_icmp;
729
730 dst = addr6->s6_addr32[3];
731 }
732
733 {
734 struct flowi fl = { .fl4_dst = dst,
735 .fl4_src = tiph->saddr,
736 .fl4_tos = RT_TOS(tos),
737 .oif = tunnel->parms.link,
738 .proto = IPPROTO_IPV6 };
739 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
740 dev->stats.tx_carrier_errors++;
741 goto tx_error_icmp;
742 }
743 }
744 if (rt->rt_type != RTN_UNICAST) {
745 ip_rt_put(rt);
746 dev->stats.tx_carrier_errors++;
747 goto tx_error_icmp;
748 }
749 tdev = rt->dst.dev;
750
751 if (tdev == dev) {
752 ip_rt_put(rt);
753 dev->stats.collisions++;
754 goto tx_error;
755 }
756
757 if (df) {
758 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
759
760 if (mtu < 68) {
761 dev->stats.collisions++;
762 ip_rt_put(rt);
763 goto tx_error;
764 }
765
766 if (mtu < IPV6_MIN_MTU) {
767 mtu = IPV6_MIN_MTU;
768 df = 0;
769 }
770
771 if (tunnel->parms.iph.daddr && skb_dst(skb))
772 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
773
774 if (skb->len > mtu) {
775 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
776 ip_rt_put(rt);
777 goto tx_error;
778 }
779 }
780
781 if (tunnel->err_count > 0) {
782 if (time_before(jiffies,
783 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
784 tunnel->err_count--;
785 dst_link_failure(skb);
786 } else
787 tunnel->err_count = 0;
788 }
789
790
791
792
793 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
794
795 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
796 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
797 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
798 if (!new_skb) {
799 ip_rt_put(rt);
800 dev->stats.tx_dropped++;
801 dev_kfree_skb(skb);
802 return NETDEV_TX_OK;
803 }
804 if (skb->sk)
805 skb_set_owner_w(new_skb, skb->sk);
806 dev_kfree_skb(skb);
807 skb = new_skb;
808 iph6 = ipv6_hdr(skb);
809 }
810
811 skb->transport_header = skb->network_header;
812 skb_push(skb, sizeof(struct iphdr));
813 skb_reset_network_header(skb);
814 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
815 IPCB(skb)->flags = 0;
816 skb_dst_drop(skb);
817 skb_dst_set(skb, &rt->dst);
818
819
820
821
822
823 iph = ip_hdr(skb);
824 iph->version = 4;
825 iph->ihl = sizeof(struct iphdr)>>2;
826 iph->frag_off = df;
827 iph->protocol = IPPROTO_IPV6;
828 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
829 iph->daddr = rt->rt_dst;
830 iph->saddr = rt->rt_src;
831
832 if ((iph->ttl = tiph->ttl) == 0)
833 iph->ttl = iph6->hop_limit;
834
835 nf_reset(skb);
836 tstats = this_cpu_ptr(dev->tstats);
837 __IPTUNNEL_XMIT(tstats, &dev->stats);
838 return NETDEV_TX_OK;
839
840tx_error_icmp:
841 dst_link_failure(skb);
842tx_error:
843 dev->stats.tx_errors++;
844 dev_kfree_skb(skb);
845 return NETDEV_TX_OK;
846}
847
848static void ipip6_tunnel_bind_dev(struct net_device *dev)
849{
850 struct net_device *tdev = NULL;
851 struct ip_tunnel *tunnel;
852 struct iphdr *iph;
853
854 tunnel = netdev_priv(dev);
855 iph = &tunnel->parms.iph;
856
857 if (iph->daddr) {
858 struct flowi fl = { .fl4_dst = iph->daddr,
859 .fl4_src = iph->saddr,
860 .fl4_tos = RT_TOS(iph->tos),
861 .oif = tunnel->parms.link,
862 .proto = IPPROTO_IPV6 };
863 struct rtable *rt;
864 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
865 tdev = rt->dst.dev;
866 ip_rt_put(rt);
867 }
868 dev->flags |= IFF_POINTOPOINT;
869 }
870
871 if (!tdev && tunnel->parms.link)
872 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
873
874 if (tdev) {
875 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
876 dev->mtu = tdev->mtu - sizeof(struct iphdr);
877 if (dev->mtu < IPV6_MIN_MTU)
878 dev->mtu = IPV6_MIN_MTU;
879 }
880 dev->iflink = tunnel->parms.link;
881}
882
883static int
884ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
885{
886 int err = 0;
887 struct ip_tunnel_parm p;
888 struct ip_tunnel_prl prl;
889 struct ip_tunnel *t;
890 struct net *net = dev_net(dev);
891 struct sit_net *sitn = net_generic(net, sit_net_id);
892#ifdef CONFIG_IPV6_SIT_6RD
893 struct ip_tunnel_6rd ip6rd;
894#endif
895
896 switch (cmd) {
897 case SIOCGETTUNNEL:
898#ifdef CONFIG_IPV6_SIT_6RD
899 case SIOCGET6RD:
900#endif
901 t = NULL;
902 if (dev == sitn->fb_tunnel_dev) {
903 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
904 err = -EFAULT;
905 break;
906 }
907 t = ipip6_tunnel_locate(net, &p, 0);
908 }
909 if (t == NULL)
910 t = netdev_priv(dev);
911
912 err = -EFAULT;
913 if (cmd == SIOCGETTUNNEL) {
914 memcpy(&p, &t->parms, sizeof(p));
915 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
916 sizeof(p)))
917 goto done;
918#ifdef CONFIG_IPV6_SIT_6RD
919 } else {
920 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
921 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
922 ip6rd.prefixlen = t->ip6rd.prefixlen;
923 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
924 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
925 sizeof(ip6rd)))
926 goto done;
927#endif
928 }
929 err = 0;
930 break;
931
932 case SIOCADDTUNNEL:
933 case SIOCCHGTUNNEL:
934 err = -EPERM;
935 if (!capable(CAP_NET_ADMIN))
936 goto done;
937
938 err = -EFAULT;
939 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
940 goto done;
941
942 err = -EINVAL;
943 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
944 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
945 goto done;
946 if (p.iph.ttl)
947 p.iph.frag_off |= htons(IP_DF);
948
949 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
950
951 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
952 if (t != NULL) {
953 if (t->dev != dev) {
954 err = -EEXIST;
955 break;
956 }
957 } else {
958 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
959 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
960 err = -EINVAL;
961 break;
962 }
963 t = netdev_priv(dev);
964 ipip6_tunnel_unlink(sitn, t);
965 synchronize_net();
966 t->parms.iph.saddr = p.iph.saddr;
967 t->parms.iph.daddr = p.iph.daddr;
968 memcpy(dev->dev_addr, &p.iph.saddr, 4);
969 memcpy(dev->broadcast, &p.iph.daddr, 4);
970 ipip6_tunnel_link(sitn, t);
971 netdev_state_change(dev);
972 }
973 }
974
975 if (t) {
976 err = 0;
977 if (cmd == SIOCCHGTUNNEL) {
978 t->parms.iph.ttl = p.iph.ttl;
979 t->parms.iph.tos = p.iph.tos;
980 if (t->parms.link != p.link) {
981 t->parms.link = p.link;
982 ipip6_tunnel_bind_dev(dev);
983 netdev_state_change(dev);
984 }
985 }
986 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
987 err = -EFAULT;
988 } else
989 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
990 break;
991
992 case SIOCDELTUNNEL:
993 err = -EPERM;
994 if (!capable(CAP_NET_ADMIN))
995 goto done;
996
997 if (dev == sitn->fb_tunnel_dev) {
998 err = -EFAULT;
999 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1000 goto done;
1001 err = -ENOENT;
1002 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1003 goto done;
1004 err = -EPERM;
1005 if (t == netdev_priv(sitn->fb_tunnel_dev))
1006 goto done;
1007 dev = t->dev;
1008 }
1009 unregister_netdevice(dev);
1010 err = 0;
1011 break;
1012
1013 case SIOCGETPRL:
1014 err = -EINVAL;
1015 if (dev == sitn->fb_tunnel_dev)
1016 goto done;
1017 err = -ENOENT;
1018 if (!(t = netdev_priv(dev)))
1019 goto done;
1020 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1021 break;
1022
1023 case SIOCADDPRL:
1024 case SIOCDELPRL:
1025 case SIOCCHGPRL:
1026 err = -EPERM;
1027 if (!capable(CAP_NET_ADMIN))
1028 goto done;
1029 err = -EINVAL;
1030 if (dev == sitn->fb_tunnel_dev)
1031 goto done;
1032 err = -EFAULT;
1033 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1034 goto done;
1035 err = -ENOENT;
1036 if (!(t = netdev_priv(dev)))
1037 goto done;
1038
1039 switch (cmd) {
1040 case SIOCDELPRL:
1041 err = ipip6_tunnel_del_prl(t, &prl);
1042 break;
1043 case SIOCADDPRL:
1044 case SIOCCHGPRL:
1045 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1046 break;
1047 }
1048 netdev_state_change(dev);
1049 break;
1050
1051#ifdef CONFIG_IPV6_SIT_6RD
1052 case SIOCADD6RD:
1053 case SIOCCHG6RD:
1054 case SIOCDEL6RD:
1055 err = -EPERM;
1056 if (!capable(CAP_NET_ADMIN))
1057 goto done;
1058
1059 err = -EFAULT;
1060 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1061 sizeof(ip6rd)))
1062 goto done;
1063
1064 t = netdev_priv(dev);
1065
1066 if (cmd != SIOCDEL6RD) {
1067 struct in6_addr prefix;
1068 __be32 relay_prefix;
1069
1070 err = -EINVAL;
1071 if (ip6rd.relay_prefixlen > 32 ||
1072 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1073 goto done;
1074
1075 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1076 ip6rd.prefixlen);
1077 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1078 goto done;
1079 if (ip6rd.relay_prefixlen)
1080 relay_prefix = ip6rd.relay_prefix &
1081 htonl(0xffffffffUL <<
1082 (32 - ip6rd.relay_prefixlen));
1083 else
1084 relay_prefix = 0;
1085 if (relay_prefix != ip6rd.relay_prefix)
1086 goto done;
1087
1088 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1089 t->ip6rd.relay_prefix = relay_prefix;
1090 t->ip6rd.prefixlen = ip6rd.prefixlen;
1091 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1092 } else
1093 ipip6_tunnel_clone_6rd(dev, sitn);
1094
1095 err = 0;
1096 break;
1097#endif
1098
1099 default:
1100 err = -EINVAL;
1101 }
1102
1103done:
1104 return err;
1105}
1106
1107static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1108{
1109 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1110 return -EINVAL;
1111 dev->mtu = new_mtu;
1112 return 0;
1113}
1114
1115static const struct net_device_ops ipip6_netdev_ops = {
1116 .ndo_uninit = ipip6_tunnel_uninit,
1117 .ndo_start_xmit = ipip6_tunnel_xmit,
1118 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1119 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1120 .ndo_get_stats = ipip6_get_stats,
1121};
1122
1123static void ipip6_dev_free(struct net_device *dev)
1124{
1125 free_percpu(dev->tstats);
1126 free_netdev(dev);
1127}
1128
1129static void ipip6_tunnel_setup(struct net_device *dev)
1130{
1131 dev->netdev_ops = &ipip6_netdev_ops;
1132 dev->destructor = ipip6_dev_free;
1133
1134 dev->type = ARPHRD_SIT;
1135 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1136 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1137 dev->flags = IFF_NOARP;
1138 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1139 dev->iflink = 0;
1140 dev->addr_len = 4;
1141 dev->features |= NETIF_F_NETNS_LOCAL;
1142 dev->features |= NETIF_F_LLTX;
1143}
1144
1145static int ipip6_tunnel_init(struct net_device *dev)
1146{
1147 struct ip_tunnel *tunnel = netdev_priv(dev);
1148
1149 tunnel->dev = dev;
1150 strcpy(tunnel->parms.name, dev->name);
1151
1152 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1153 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1154
1155 ipip6_tunnel_bind_dev(dev);
1156 dev->tstats = alloc_percpu(struct pcpu_tstats);
1157 if (!dev->tstats)
1158 return -ENOMEM;
1159
1160 return 0;
1161}
1162
1163static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1164{
1165 struct ip_tunnel *tunnel = netdev_priv(dev);
1166 struct iphdr *iph = &tunnel->parms.iph;
1167 struct net *net = dev_net(dev);
1168 struct sit_net *sitn = net_generic(net, sit_net_id);
1169
1170 tunnel->dev = dev;
1171 strcpy(tunnel->parms.name, dev->name);
1172
1173 iph->version = 4;
1174 iph->protocol = IPPROTO_IPV6;
1175 iph->ihl = 5;
1176 iph->ttl = 64;
1177
1178 dev->tstats = alloc_percpu(struct pcpu_tstats);
1179 if (!dev->tstats)
1180 return -ENOMEM;
1181 dev_hold(dev);
1182 sitn->tunnels_wc[0] = tunnel;
1183 return 0;
1184}
1185
1186static struct xfrm_tunnel sit_handler __read_mostly = {
1187 .handler = ipip6_rcv,
1188 .err_handler = ipip6_err,
1189 .priority = 1,
1190};
1191
1192static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1193{
1194 int prio;
1195
1196 for (prio = 1; prio < 4; prio++) {
1197 int h;
1198 for (h = 0; h < HASH_SIZE; h++) {
1199 struct ip_tunnel *t = sitn->tunnels[prio][h];
1200
1201 while (t != NULL) {
1202 unregister_netdevice_queue(t->dev, head);
1203 t = t->next;
1204 }
1205 }
1206 }
1207}
1208
1209static int __net_init sit_init_net(struct net *net)
1210{
1211 struct sit_net *sitn = net_generic(net, sit_net_id);
1212 int err;
1213
1214 sitn->tunnels[0] = sitn->tunnels_wc;
1215 sitn->tunnels[1] = sitn->tunnels_l;
1216 sitn->tunnels[2] = sitn->tunnels_r;
1217 sitn->tunnels[3] = sitn->tunnels_r_l;
1218
1219 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1220 ipip6_tunnel_setup);
1221 if (!sitn->fb_tunnel_dev) {
1222 err = -ENOMEM;
1223 goto err_alloc_dev;
1224 }
1225 dev_net_set(sitn->fb_tunnel_dev, net);
1226
1227 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1228 if (err)
1229 goto err_dev_free;
1230
1231 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1232
1233 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1234 goto err_reg_dev;
1235
1236 return 0;
1237
1238err_reg_dev:
1239 dev_put(sitn->fb_tunnel_dev);
1240err_dev_free:
1241 ipip6_dev_free(sitn->fb_tunnel_dev);
1242err_alloc_dev:
1243 return err;
1244}
1245
1246static void __net_exit sit_exit_net(struct net *net)
1247{
1248 struct sit_net *sitn = net_generic(net, sit_net_id);
1249 LIST_HEAD(list);
1250
1251 rtnl_lock();
1252 sit_destroy_tunnels(sitn, &list);
1253 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1254 unregister_netdevice_many(&list);
1255 rtnl_unlock();
1256}
1257
1258static struct pernet_operations sit_net_ops = {
1259 .init = sit_init_net,
1260 .exit = sit_exit_net,
1261 .id = &sit_net_id,
1262 .size = sizeof(struct sit_net),
1263};
1264
1265static void __exit sit_cleanup(void)
1266{
1267 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1268
1269 unregister_pernet_device(&sit_net_ops);
1270 rcu_barrier();
1271}
1272
1273static int __init sit_init(void)
1274{
1275 int err;
1276
1277 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1278
1279 err = register_pernet_device(&sit_net_ops);
1280 if (err < 0)
1281 return err;
1282 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1283 if (err < 0) {
1284 unregister_pernet_device(&sit_net_ops);
1285 printk(KERN_INFO "sit init: Can't add protocol\n");
1286 }
1287 return err;
1288}
1289
1290module_init(sit_init);
1291module_exit(sit_cleanup);
1292MODULE_LICENSE("GPL");
1293MODULE_ALIAS_NETDEV("sit0");
1294
