]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv6/sit.c
sit: rename rtnl functions for consistency
[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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38
39 #include <net/sock.h>
40 #include <net/snmp.h>
41
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/ipip.h>
53 #include <net/inet_ecn.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
58
59 /*
60 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
61
62 For comments look at net/ipv4/ip_gre.c --ANK
63 */
64
65 #define HASH_SIZE 16
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
67
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
71 static struct rtnl_link_ops sit_link_ops __read_mostly;
72
73 static int sit_net_id __read_mostly;
74 struct sit_net {
75 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
76 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
77 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
78 struct ip_tunnel __rcu *tunnels_wc[1];
79 struct ip_tunnel __rcu **tunnels[4];
80
81 struct net_device *fb_tunnel_dev;
82 };
83
84 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
85 struct rtnl_link_stats64 *tot)
86 {
87 int i;
88
89 for_each_possible_cpu(i) {
90 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
91 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
92 unsigned int start;
93
94 do {
95 start = u64_stats_fetch_begin_bh(&tstats->syncp);
96 rx_packets = tstats->rx_packets;
97 tx_packets = tstats->tx_packets;
98 rx_bytes = tstats->rx_bytes;
99 tx_bytes = tstats->tx_bytes;
100 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
101
102 tot->rx_packets += rx_packets;
103 tot->tx_packets += tx_packets;
104 tot->rx_bytes += rx_bytes;
105 tot->tx_bytes += tx_bytes;
106 }
107
108 tot->rx_errors = dev->stats.rx_errors;
109 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
110 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
111 tot->tx_dropped = dev->stats.tx_dropped;
112 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
113 tot->tx_errors = dev->stats.tx_errors;
114
115 return tot;
116 }
117
118 /*
119 * Must be invoked with rcu_read_lock
120 */
121 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
122 struct net_device *dev, __be32 remote, __be32 local)
123 {
124 unsigned int h0 = HASH(remote);
125 unsigned int h1 = HASH(local);
126 struct ip_tunnel *t;
127 struct sit_net *sitn = net_generic(net, sit_net_id);
128
129 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
130 if (local == t->parms.iph.saddr &&
131 remote == t->parms.iph.daddr &&
132 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
133 (t->dev->flags & IFF_UP))
134 return t;
135 }
136 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
137 if (remote == t->parms.iph.daddr &&
138 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
139 (t->dev->flags & IFF_UP))
140 return t;
141 }
142 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
143 if (local == t->parms.iph.saddr &&
144 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
145 (t->dev->flags & IFF_UP))
146 return t;
147 }
148 t = rcu_dereference(sitn->tunnels_wc[0]);
149 if ((t != NULL) && (t->dev->flags & IFF_UP))
150 return t;
151 return NULL;
152 }
153
154 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
155 struct ip_tunnel_parm *parms)
156 {
157 __be32 remote = parms->iph.daddr;
158 __be32 local = parms->iph.saddr;
159 unsigned int h = 0;
160 int prio = 0;
161
162 if (remote) {
163 prio |= 2;
164 h ^= HASH(remote);
165 }
166 if (local) {
167 prio |= 1;
168 h ^= HASH(local);
169 }
170 return &sitn->tunnels[prio][h];
171 }
172
173 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
174 struct ip_tunnel *t)
175 {
176 return __ipip6_bucket(sitn, &t->parms);
177 }
178
179 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
180 {
181 struct ip_tunnel __rcu **tp;
182 struct ip_tunnel *iter;
183
184 for (tp = ipip6_bucket(sitn, t);
185 (iter = rtnl_dereference(*tp)) != NULL;
186 tp = &iter->next) {
187 if (t == iter) {
188 rcu_assign_pointer(*tp, t->next);
189 break;
190 }
191 }
192 }
193
194 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
195 {
196 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
197
198 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
199 rcu_assign_pointer(*tp, t);
200 }
201
202 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
203 {
204 #ifdef CONFIG_IPV6_SIT_6RD
205 struct ip_tunnel *t = netdev_priv(dev);
206
207 if (t->dev == sitn->fb_tunnel_dev) {
208 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
209 t->ip6rd.relay_prefix = 0;
210 t->ip6rd.prefixlen = 16;
211 t->ip6rd.relay_prefixlen = 0;
212 } else {
213 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
214 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
215 }
216 #endif
217 }
218
219 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
220 struct ip_tunnel_parm *parms, int create)
221 {
222 __be32 remote = parms->iph.daddr;
223 __be32 local = parms->iph.saddr;
224 struct ip_tunnel *t, *nt;
225 struct ip_tunnel __rcu **tp;
226 struct net_device *dev;
227 char name[IFNAMSIZ];
228 struct sit_net *sitn = net_generic(net, sit_net_id);
229
230 for (tp = __ipip6_bucket(sitn, parms);
231 (t = rtnl_dereference(*tp)) != NULL;
232 tp = &t->next) {
233 if (local == t->parms.iph.saddr &&
234 remote == t->parms.iph.daddr &&
235 parms->link == t->parms.link) {
236 if (create)
237 return NULL;
238 else
239 return t;
240 }
241 }
242 if (!create)
243 goto failed;
244
245 if (parms->name[0])
246 strlcpy(name, parms->name, IFNAMSIZ);
247 else
248 strcpy(name, "sit%d");
249
250 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
251 if (dev == NULL)
252 return NULL;
253
254 dev_net_set(dev, net);
255
256 nt = netdev_priv(dev);
257
258 nt->parms = *parms;
259 if (ipip6_tunnel_init(dev) < 0)
260 goto failed_free;
261 ipip6_tunnel_clone_6rd(dev, sitn);
262
263 if (parms->i_flags & SIT_ISATAP)
264 dev->priv_flags |= IFF_ISATAP;
265
266 if (register_netdevice(dev) < 0)
267 goto failed_free;
268
269 strcpy(nt->parms.name, dev->name);
270 dev->rtnl_link_ops = &sit_link_ops;
271
272 dev_hold(dev);
273
274 ipip6_tunnel_link(sitn, nt);
275 return nt;
276
277 failed_free:
278 ipip6_dev_free(dev);
279 failed:
280 return NULL;
281 }
282
283 #define for_each_prl_rcu(start) \
284 for (prl = rcu_dereference(start); \
285 prl; \
286 prl = rcu_dereference(prl->next))
287
288 static struct ip_tunnel_prl_entry *
289 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
290 {
291 struct ip_tunnel_prl_entry *prl;
292
293 for_each_prl_rcu(t->prl)
294 if (prl->addr == addr)
295 break;
296 return prl;
297
298 }
299
300 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
301 struct ip_tunnel_prl __user *a)
302 {
303 struct ip_tunnel_prl kprl, *kp;
304 struct ip_tunnel_prl_entry *prl;
305 unsigned int cmax, c = 0, ca, len;
306 int ret = 0;
307
308 if (copy_from_user(&kprl, a, sizeof(kprl)))
309 return -EFAULT;
310 cmax = kprl.datalen / sizeof(kprl);
311 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
312 cmax = 1;
313
314 /* For simple GET or for root users,
315 * we try harder to allocate.
316 */
317 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
318 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
319 NULL;
320
321 rcu_read_lock();
322
323 ca = t->prl_count < cmax ? t->prl_count : cmax;
324
325 if (!kp) {
326 /* We don't try hard to allocate much memory for
327 * non-root users.
328 * For root users, retry allocating enough memory for
329 * the answer.
330 */
331 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
332 if (!kp) {
333 ret = -ENOMEM;
334 goto out;
335 }
336 }
337
338 c = 0;
339 for_each_prl_rcu(t->prl) {
340 if (c >= cmax)
341 break;
342 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
343 continue;
344 kp[c].addr = prl->addr;
345 kp[c].flags = prl->flags;
346 c++;
347 if (kprl.addr != htonl(INADDR_ANY))
348 break;
349 }
350 out:
351 rcu_read_unlock();
352
353 len = sizeof(*kp) * c;
354 ret = 0;
355 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
356 ret = -EFAULT;
357
358 kfree(kp);
359
360 return ret;
361 }
362
363 static int
364 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
365 {
366 struct ip_tunnel_prl_entry *p;
367 int err = 0;
368
369 if (a->addr == htonl(INADDR_ANY))
370 return -EINVAL;
371
372 ASSERT_RTNL();
373
374 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
375 if (p->addr == a->addr) {
376 if (chg) {
377 p->flags = a->flags;
378 goto out;
379 }
380 err = -EEXIST;
381 goto out;
382 }
383 }
384
385 if (chg) {
386 err = -ENXIO;
387 goto out;
388 }
389
390 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
391 if (!p) {
392 err = -ENOBUFS;
393 goto out;
394 }
395
396 p->next = t->prl;
397 p->addr = a->addr;
398 p->flags = a->flags;
399 t->prl_count++;
400 rcu_assign_pointer(t->prl, p);
401 out:
402 return err;
403 }
404
405 static void prl_list_destroy_rcu(struct rcu_head *head)
406 {
407 struct ip_tunnel_prl_entry *p, *n;
408
409 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
410 do {
411 n = rcu_dereference_protected(p->next, 1);
412 kfree(p);
413 p = n;
414 } while (p);
415 }
416
417 static int
418 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
419 {
420 struct ip_tunnel_prl_entry *x;
421 struct ip_tunnel_prl_entry __rcu **p;
422 int err = 0;
423
424 ASSERT_RTNL();
425
426 if (a && a->addr != htonl(INADDR_ANY)) {
427 for (p = &t->prl;
428 (x = rtnl_dereference(*p)) != NULL;
429 p = &x->next) {
430 if (x->addr == a->addr) {
431 *p = x->next;
432 kfree_rcu(x, rcu_head);
433 t->prl_count--;
434 goto out;
435 }
436 }
437 err = -ENXIO;
438 } else {
439 x = rtnl_dereference(t->prl);
440 if (x) {
441 t->prl_count = 0;
442 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
443 t->prl = NULL;
444 }
445 }
446 out:
447 return err;
448 }
449
450 static int
451 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
452 {
453 struct ip_tunnel_prl_entry *p;
454 int ok = 1;
455
456 rcu_read_lock();
457 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
458 if (p) {
459 if (p->flags & PRL_DEFAULT)
460 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
461 else
462 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
463 } else {
464 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
465
466 if (ipv6_addr_is_isatap(addr6) &&
467 (addr6->s6_addr32[3] == iph->saddr) &&
468 ipv6_chk_prefix(addr6, t->dev))
469 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
470 else
471 ok = 0;
472 }
473 rcu_read_unlock();
474 return ok;
475 }
476
477 static void ipip6_tunnel_uninit(struct net_device *dev)
478 {
479 struct net *net = dev_net(dev);
480 struct sit_net *sitn = net_generic(net, sit_net_id);
481
482 if (dev == sitn->fb_tunnel_dev) {
483 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
484 } else {
485 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
486 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
487 }
488 dev_put(dev);
489 }
490
491
492 static int ipip6_err(struct sk_buff *skb, u32 info)
493 {
494
495 /* All the routers (except for Linux) return only
496 8 bytes of packet payload. It means, that precise relaying of
497 ICMP in the real Internet is absolutely infeasible.
498 */
499 const struct iphdr *iph = (const struct iphdr *)skb->data;
500 const int type = icmp_hdr(skb)->type;
501 const int code = icmp_hdr(skb)->code;
502 struct ip_tunnel *t;
503 int err;
504
505 switch (type) {
506 default:
507 case ICMP_PARAMETERPROB:
508 return 0;
509
510 case ICMP_DEST_UNREACH:
511 switch (code) {
512 case ICMP_SR_FAILED:
513 case ICMP_PORT_UNREACH:
514 /* Impossible event. */
515 return 0;
516 default:
517 /* All others are translated to HOST_UNREACH.
518 rfc2003 contains "deep thoughts" about NET_UNREACH,
519 I believe they are just ether pollution. --ANK
520 */
521 break;
522 }
523 break;
524 case ICMP_TIME_EXCEEDED:
525 if (code != ICMP_EXC_TTL)
526 return 0;
527 break;
528 case ICMP_REDIRECT:
529 break;
530 }
531
532 err = -ENOENT;
533
534 t = ipip6_tunnel_lookup(dev_net(skb->dev),
535 skb->dev,
536 iph->daddr,
537 iph->saddr);
538 if (t == NULL)
539 goto out;
540
541 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
542 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
543 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
544 err = 0;
545 goto out;
546 }
547 if (type == ICMP_REDIRECT) {
548 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
549 IPPROTO_IPV6, 0);
550 err = 0;
551 goto out;
552 }
553
554 if (t->parms.iph.daddr == 0)
555 goto out;
556
557 err = 0;
558 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
559 goto out;
560
561 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
562 t->err_count++;
563 else
564 t->err_count = 1;
565 t->err_time = jiffies;
566 out:
567 return err;
568 }
569
570 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
571 {
572 if (INET_ECN_is_ce(iph->tos))
573 IP6_ECN_set_ce(ipv6_hdr(skb));
574 }
575
576 static int ipip6_rcv(struct sk_buff *skb)
577 {
578 const struct iphdr *iph;
579 struct ip_tunnel *tunnel;
580
581 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
582 goto out;
583
584 iph = ip_hdr(skb);
585
586 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
587 iph->saddr, iph->daddr);
588 if (tunnel != NULL) {
589 struct pcpu_tstats *tstats;
590
591 secpath_reset(skb);
592 skb->mac_header = skb->network_header;
593 skb_reset_network_header(skb);
594 IPCB(skb)->flags = 0;
595 skb->protocol = htons(ETH_P_IPV6);
596 skb->pkt_type = PACKET_HOST;
597
598 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
599 !isatap_chksrc(skb, iph, tunnel)) {
600 tunnel->dev->stats.rx_errors++;
601 kfree_skb(skb);
602 return 0;
603 }
604
605 tstats = this_cpu_ptr(tunnel->dev->tstats);
606 tstats->rx_packets++;
607 tstats->rx_bytes += skb->len;
608
609 __skb_tunnel_rx(skb, tunnel->dev);
610
611 ipip6_ecn_decapsulate(iph, skb);
612
613 netif_rx(skb);
614
615 return 0;
616 }
617
618 /* no tunnel matched, let upstream know, ipsec may handle it */
619 return 1;
620 out:
621 kfree_skb(skb);
622 return 0;
623 }
624
625 /*
626 * Returns the embedded IPv4 address if the IPv6 address
627 * comes from 6rd / 6to4 (RFC 3056) addr space.
628 */
629 static inline
630 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
631 {
632 __be32 dst = 0;
633
634 #ifdef CONFIG_IPV6_SIT_6RD
635 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
636 tunnel->ip6rd.prefixlen)) {
637 unsigned int pbw0, pbi0;
638 int pbi1;
639 u32 d;
640
641 pbw0 = tunnel->ip6rd.prefixlen >> 5;
642 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
643
644 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
645 tunnel->ip6rd.relay_prefixlen;
646
647 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
648 if (pbi1 > 0)
649 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
650 (32 - pbi1);
651
652 dst = tunnel->ip6rd.relay_prefix | htonl(d);
653 }
654 #else
655 if (v6dst->s6_addr16[0] == htons(0x2002)) {
656 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
657 memcpy(&dst, &v6dst->s6_addr16[1], 4);
658 }
659 #endif
660 return dst;
661 }
662
663 /*
664 * This function assumes it is being called from dev_queue_xmit()
665 * and that skb is filled properly by that function.
666 */
667
668 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
669 struct net_device *dev)
670 {
671 struct ip_tunnel *tunnel = netdev_priv(dev);
672 const struct iphdr *tiph = &tunnel->parms.iph;
673 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
674 u8 tos = tunnel->parms.iph.tos;
675 __be16 df = tiph->frag_off;
676 struct rtable *rt; /* Route to the other host */
677 struct net_device *tdev; /* Device to other host */
678 struct iphdr *iph; /* Our new IP header */
679 unsigned int max_headroom; /* The extra header space needed */
680 __be32 dst = tiph->daddr;
681 struct flowi4 fl4;
682 int mtu;
683 const struct in6_addr *addr6;
684 int addr_type;
685
686 if (skb->protocol != htons(ETH_P_IPV6))
687 goto tx_error;
688
689 if (tos == 1)
690 tos = ipv6_get_dsfield(iph6);
691
692 /* ISATAP (RFC4214) - must come before 6to4 */
693 if (dev->priv_flags & IFF_ISATAP) {
694 struct neighbour *neigh = NULL;
695 bool do_tx_error = false;
696
697 if (skb_dst(skb))
698 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
699
700 if (neigh == NULL) {
701 net_dbg_ratelimited("sit: nexthop == NULL\n");
702 goto tx_error;
703 }
704
705 addr6 = (const struct in6_addr *)&neigh->primary_key;
706 addr_type = ipv6_addr_type(addr6);
707
708 if ((addr_type & IPV6_ADDR_UNICAST) &&
709 ipv6_addr_is_isatap(addr6))
710 dst = addr6->s6_addr32[3];
711 else
712 do_tx_error = true;
713
714 neigh_release(neigh);
715 if (do_tx_error)
716 goto tx_error;
717 }
718
719 if (!dst)
720 dst = try_6rd(&iph6->daddr, tunnel);
721
722 if (!dst) {
723 struct neighbour *neigh = NULL;
724 bool do_tx_error = false;
725
726 if (skb_dst(skb))
727 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
728
729 if (neigh == NULL) {
730 net_dbg_ratelimited("sit: nexthop == NULL\n");
731 goto tx_error;
732 }
733
734 addr6 = (const struct in6_addr *)&neigh->primary_key;
735 addr_type = ipv6_addr_type(addr6);
736
737 if (addr_type == IPV6_ADDR_ANY) {
738 addr6 = &ipv6_hdr(skb)->daddr;
739 addr_type = ipv6_addr_type(addr6);
740 }
741
742 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
743 dst = addr6->s6_addr32[3];
744 else
745 do_tx_error = true;
746
747 neigh_release(neigh);
748 if (do_tx_error)
749 goto tx_error;
750 }
751
752 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
753 dst, tiph->saddr,
754 0, 0,
755 IPPROTO_IPV6, RT_TOS(tos),
756 tunnel->parms.link);
757 if (IS_ERR(rt)) {
758 dev->stats.tx_carrier_errors++;
759 goto tx_error_icmp;
760 }
761 if (rt->rt_type != RTN_UNICAST) {
762 ip_rt_put(rt);
763 dev->stats.tx_carrier_errors++;
764 goto tx_error_icmp;
765 }
766 tdev = rt->dst.dev;
767
768 if (tdev == dev) {
769 ip_rt_put(rt);
770 dev->stats.collisions++;
771 goto tx_error;
772 }
773
774 if (df) {
775 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
776
777 if (mtu < 68) {
778 dev->stats.collisions++;
779 ip_rt_put(rt);
780 goto tx_error;
781 }
782
783 if (mtu < IPV6_MIN_MTU) {
784 mtu = IPV6_MIN_MTU;
785 df = 0;
786 }
787
788 if (tunnel->parms.iph.daddr && skb_dst(skb))
789 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
790
791 if (skb->len > mtu) {
792 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
793 ip_rt_put(rt);
794 goto tx_error;
795 }
796 }
797
798 if (tunnel->err_count > 0) {
799 if (time_before(jiffies,
800 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
801 tunnel->err_count--;
802 dst_link_failure(skb);
803 } else
804 tunnel->err_count = 0;
805 }
806
807 /*
808 * Okay, now see if we can stuff it in the buffer as-is.
809 */
810 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
811
812 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
813 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
814 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
815 if (!new_skb) {
816 ip_rt_put(rt);
817 dev->stats.tx_dropped++;
818 dev_kfree_skb(skb);
819 return NETDEV_TX_OK;
820 }
821 if (skb->sk)
822 skb_set_owner_w(new_skb, skb->sk);
823 dev_kfree_skb(skb);
824 skb = new_skb;
825 iph6 = ipv6_hdr(skb);
826 }
827
828 skb->transport_header = skb->network_header;
829 skb_push(skb, sizeof(struct iphdr));
830 skb_reset_network_header(skb);
831 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
832 IPCB(skb)->flags = 0;
833 skb_dst_drop(skb);
834 skb_dst_set(skb, &rt->dst);
835
836 /*
837 * Push down and install the IPIP header.
838 */
839
840 iph = ip_hdr(skb);
841 iph->version = 4;
842 iph->ihl = sizeof(struct iphdr)>>2;
843 iph->frag_off = df;
844 iph->protocol = IPPROTO_IPV6;
845 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
846 iph->daddr = fl4.daddr;
847 iph->saddr = fl4.saddr;
848
849 if ((iph->ttl = tiph->ttl) == 0)
850 iph->ttl = iph6->hop_limit;
851
852 iptunnel_xmit(skb, dev);
853 return NETDEV_TX_OK;
854
855 tx_error_icmp:
856 dst_link_failure(skb);
857 tx_error:
858 dev->stats.tx_errors++;
859 dev_kfree_skb(skb);
860 return NETDEV_TX_OK;
861 }
862
863 static void ipip6_tunnel_bind_dev(struct net_device *dev)
864 {
865 struct net_device *tdev = NULL;
866 struct ip_tunnel *tunnel;
867 const struct iphdr *iph;
868 struct flowi4 fl4;
869
870 tunnel = netdev_priv(dev);
871 iph = &tunnel->parms.iph;
872
873 if (iph->daddr) {
874 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
875 iph->daddr, iph->saddr,
876 0, 0,
877 IPPROTO_IPV6,
878 RT_TOS(iph->tos),
879 tunnel->parms.link);
880
881 if (!IS_ERR(rt)) {
882 tdev = rt->dst.dev;
883 ip_rt_put(rt);
884 }
885 dev->flags |= IFF_POINTOPOINT;
886 }
887
888 if (!tdev && tunnel->parms.link)
889 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
890
891 if (tdev) {
892 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
893 dev->mtu = tdev->mtu - sizeof(struct iphdr);
894 if (dev->mtu < IPV6_MIN_MTU)
895 dev->mtu = IPV6_MIN_MTU;
896 }
897 dev->iflink = tunnel->parms.link;
898 }
899
900 static int
901 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
902 {
903 int err = 0;
904 struct ip_tunnel_parm p;
905 struct ip_tunnel_prl prl;
906 struct ip_tunnel *t;
907 struct net *net = dev_net(dev);
908 struct sit_net *sitn = net_generic(net, sit_net_id);
909 #ifdef CONFIG_IPV6_SIT_6RD
910 struct ip_tunnel_6rd ip6rd;
911 #endif
912
913 switch (cmd) {
914 case SIOCGETTUNNEL:
915 #ifdef CONFIG_IPV6_SIT_6RD
916 case SIOCGET6RD:
917 #endif
918 t = NULL;
919 if (dev == sitn->fb_tunnel_dev) {
920 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
921 err = -EFAULT;
922 break;
923 }
924 t = ipip6_tunnel_locate(net, &p, 0);
925 }
926 if (t == NULL)
927 t = netdev_priv(dev);
928
929 err = -EFAULT;
930 if (cmd == SIOCGETTUNNEL) {
931 memcpy(&p, &t->parms, sizeof(p));
932 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
933 sizeof(p)))
934 goto done;
935 #ifdef CONFIG_IPV6_SIT_6RD
936 } else {
937 ip6rd.prefix = t->ip6rd.prefix;
938 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
939 ip6rd.prefixlen = t->ip6rd.prefixlen;
940 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
941 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
942 sizeof(ip6rd)))
943 goto done;
944 #endif
945 }
946 err = 0;
947 break;
948
949 case SIOCADDTUNNEL:
950 case SIOCCHGTUNNEL:
951 err = -EPERM;
952 if (!capable(CAP_NET_ADMIN))
953 goto done;
954
955 err = -EFAULT;
956 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
957 goto done;
958
959 err = -EINVAL;
960 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
961 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
962 goto done;
963 if (p.iph.ttl)
964 p.iph.frag_off |= htons(IP_DF);
965
966 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
967
968 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
969 if (t != NULL) {
970 if (t->dev != dev) {
971 err = -EEXIST;
972 break;
973 }
974 } else {
975 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
976 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
977 err = -EINVAL;
978 break;
979 }
980 t = netdev_priv(dev);
981 }
982
983 ipip6_tunnel_unlink(sitn, t);
984 synchronize_net();
985 t->parms.iph.saddr = p.iph.saddr;
986 t->parms.iph.daddr = p.iph.daddr;
987 memcpy(dev->dev_addr, &p.iph.saddr, 4);
988 memcpy(dev->broadcast, &p.iph.daddr, 4);
989 ipip6_tunnel_link(sitn, t);
990 t->parms.iph.ttl = p.iph.ttl;
991 t->parms.iph.tos = p.iph.tos;
992 if (t->parms.link != p.link) {
993 t->parms.link = p.link;
994 ipip6_tunnel_bind_dev(dev);
995 }
996 netdev_state_change(dev);
997 }
998
999 if (t) {
1000 err = 0;
1001 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1002 err = -EFAULT;
1003 } else
1004 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1005 break;
1006
1007 case SIOCDELTUNNEL:
1008 err = -EPERM;
1009 if (!capable(CAP_NET_ADMIN))
1010 goto done;
1011
1012 if (dev == sitn->fb_tunnel_dev) {
1013 err = -EFAULT;
1014 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1015 goto done;
1016 err = -ENOENT;
1017 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1018 goto done;
1019 err = -EPERM;
1020 if (t == netdev_priv(sitn->fb_tunnel_dev))
1021 goto done;
1022 dev = t->dev;
1023 }
1024 unregister_netdevice(dev);
1025 err = 0;
1026 break;
1027
1028 case SIOCGETPRL:
1029 err = -EINVAL;
1030 if (dev == sitn->fb_tunnel_dev)
1031 goto done;
1032 err = -ENOENT;
1033 if (!(t = netdev_priv(dev)))
1034 goto done;
1035 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1036 break;
1037
1038 case SIOCADDPRL:
1039 case SIOCDELPRL:
1040 case SIOCCHGPRL:
1041 err = -EPERM;
1042 if (!capable(CAP_NET_ADMIN))
1043 goto done;
1044 err = -EINVAL;
1045 if (dev == sitn->fb_tunnel_dev)
1046 goto done;
1047 err = -EFAULT;
1048 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1049 goto done;
1050 err = -ENOENT;
1051 if (!(t = netdev_priv(dev)))
1052 goto done;
1053
1054 switch (cmd) {
1055 case SIOCDELPRL:
1056 err = ipip6_tunnel_del_prl(t, &prl);
1057 break;
1058 case SIOCADDPRL:
1059 case SIOCCHGPRL:
1060 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1061 break;
1062 }
1063 netdev_state_change(dev);
1064 break;
1065
1066 #ifdef CONFIG_IPV6_SIT_6RD
1067 case SIOCADD6RD:
1068 case SIOCCHG6RD:
1069 case SIOCDEL6RD:
1070 err = -EPERM;
1071 if (!capable(CAP_NET_ADMIN))
1072 goto done;
1073
1074 err = -EFAULT;
1075 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1076 sizeof(ip6rd)))
1077 goto done;
1078
1079 t = netdev_priv(dev);
1080
1081 if (cmd != SIOCDEL6RD) {
1082 struct in6_addr prefix;
1083 __be32 relay_prefix;
1084
1085 err = -EINVAL;
1086 if (ip6rd.relay_prefixlen > 32 ||
1087 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1088 goto done;
1089
1090 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1091 ip6rd.prefixlen);
1092 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1093 goto done;
1094 if (ip6rd.relay_prefixlen)
1095 relay_prefix = ip6rd.relay_prefix &
1096 htonl(0xffffffffUL <<
1097 (32 - ip6rd.relay_prefixlen));
1098 else
1099 relay_prefix = 0;
1100 if (relay_prefix != ip6rd.relay_prefix)
1101 goto done;
1102
1103 t->ip6rd.prefix = prefix;
1104 t->ip6rd.relay_prefix = relay_prefix;
1105 t->ip6rd.prefixlen = ip6rd.prefixlen;
1106 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1107 } else
1108 ipip6_tunnel_clone_6rd(dev, sitn);
1109
1110 err = 0;
1111 break;
1112 #endif
1113
1114 default:
1115 err = -EINVAL;
1116 }
1117
1118 done:
1119 return err;
1120 }
1121
1122 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1123 {
1124 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1125 return -EINVAL;
1126 dev->mtu = new_mtu;
1127 return 0;
1128 }
1129
1130 static const struct net_device_ops ipip6_netdev_ops = {
1131 .ndo_uninit = ipip6_tunnel_uninit,
1132 .ndo_start_xmit = ipip6_tunnel_xmit,
1133 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1134 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1135 .ndo_get_stats64= ipip6_get_stats64,
1136 };
1137
1138 static void ipip6_dev_free(struct net_device *dev)
1139 {
1140 free_percpu(dev->tstats);
1141 free_netdev(dev);
1142 }
1143
1144 static void ipip6_tunnel_setup(struct net_device *dev)
1145 {
1146 dev->netdev_ops = &ipip6_netdev_ops;
1147 dev->destructor = ipip6_dev_free;
1148
1149 dev->type = ARPHRD_SIT;
1150 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1151 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1152 dev->flags = IFF_NOARP;
1153 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1154 dev->iflink = 0;
1155 dev->addr_len = 4;
1156 dev->features |= NETIF_F_NETNS_LOCAL;
1157 dev->features |= NETIF_F_LLTX;
1158 }
1159
1160 static int ipip6_tunnel_init(struct net_device *dev)
1161 {
1162 struct ip_tunnel *tunnel = netdev_priv(dev);
1163
1164 tunnel->dev = dev;
1165
1166 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1167 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1168
1169 ipip6_tunnel_bind_dev(dev);
1170 dev->tstats = alloc_percpu(struct pcpu_tstats);
1171 if (!dev->tstats)
1172 return -ENOMEM;
1173
1174 return 0;
1175 }
1176
1177 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1178 {
1179 struct ip_tunnel *tunnel = netdev_priv(dev);
1180 struct iphdr *iph = &tunnel->parms.iph;
1181 struct net *net = dev_net(dev);
1182 struct sit_net *sitn = net_generic(net, sit_net_id);
1183
1184 tunnel->dev = dev;
1185 strcpy(tunnel->parms.name, dev->name);
1186
1187 iph->version = 4;
1188 iph->protocol = IPPROTO_IPV6;
1189 iph->ihl = 5;
1190 iph->ttl = 64;
1191
1192 dev->tstats = alloc_percpu(struct pcpu_tstats);
1193 if (!dev->tstats)
1194 return -ENOMEM;
1195 dev_hold(dev);
1196 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1197 return 0;
1198 }
1199
1200 static size_t ipip6_get_size(const struct net_device *dev)
1201 {
1202 return
1203 /* IFLA_IPTUN_LINK */
1204 nla_total_size(4) +
1205 /* IFLA_IPTUN_LOCAL */
1206 nla_total_size(4) +
1207 /* IFLA_IPTUN_REMOTE */
1208 nla_total_size(4) +
1209 /* IFLA_IPTUN_TTL */
1210 nla_total_size(1) +
1211 /* IFLA_IPTUN_TOS */
1212 nla_total_size(1) +
1213 /* IFLA_IPTUN_PMTUDISC */
1214 nla_total_size(1) +
1215 /* IFLA_IPTUN_FLAGS */
1216 nla_total_size(2) +
1217 0;
1218 }
1219
1220 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1221 {
1222 struct ip_tunnel *tunnel = netdev_priv(dev);
1223 struct ip_tunnel_parm *parm = &tunnel->parms;
1224
1225 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1226 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1227 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1228 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1229 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1230 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1231 !!(parm->iph.frag_off & htons(IP_DF))) ||
1232 nla_put_u16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1233 goto nla_put_failure;
1234 return 0;
1235
1236 nla_put_failure:
1237 return -EMSGSIZE;
1238 }
1239
1240 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1241 .kind = "sit",
1242 .maxtype = IFLA_IPTUN_MAX,
1243 .priv_size = sizeof(struct ip_tunnel),
1244 .get_size = ipip6_get_size,
1245 .fill_info = ipip6_fill_info,
1246 };
1247
1248 static struct xfrm_tunnel sit_handler __read_mostly = {
1249 .handler = ipip6_rcv,
1250 .err_handler = ipip6_err,
1251 .priority = 1,
1252 };
1253
1254 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1255 {
1256 int prio;
1257
1258 for (prio = 1; prio < 4; prio++) {
1259 int h;
1260 for (h = 0; h < HASH_SIZE; h++) {
1261 struct ip_tunnel *t;
1262
1263 t = rtnl_dereference(sitn->tunnels[prio][h]);
1264 while (t != NULL) {
1265 unregister_netdevice_queue(t->dev, head);
1266 t = rtnl_dereference(t->next);
1267 }
1268 }
1269 }
1270 }
1271
1272 static int __net_init sit_init_net(struct net *net)
1273 {
1274 struct sit_net *sitn = net_generic(net, sit_net_id);
1275 struct ip_tunnel *t;
1276 int err;
1277
1278 sitn->tunnels[0] = sitn->tunnels_wc;
1279 sitn->tunnels[1] = sitn->tunnels_l;
1280 sitn->tunnels[2] = sitn->tunnels_r;
1281 sitn->tunnels[3] = sitn->tunnels_r_l;
1282
1283 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1284 ipip6_tunnel_setup);
1285 if (!sitn->fb_tunnel_dev) {
1286 err = -ENOMEM;
1287 goto err_alloc_dev;
1288 }
1289 dev_net_set(sitn->fb_tunnel_dev, net);
1290
1291 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1292 if (err)
1293 goto err_dev_free;
1294
1295 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1296
1297 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1298 goto err_reg_dev;
1299
1300 t = netdev_priv(sitn->fb_tunnel_dev);
1301
1302 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1303 return 0;
1304
1305 err_reg_dev:
1306 dev_put(sitn->fb_tunnel_dev);
1307 err_dev_free:
1308 ipip6_dev_free(sitn->fb_tunnel_dev);
1309 err_alloc_dev:
1310 return err;
1311 }
1312
1313 static void __net_exit sit_exit_net(struct net *net)
1314 {
1315 struct sit_net *sitn = net_generic(net, sit_net_id);
1316 LIST_HEAD(list);
1317
1318 rtnl_lock();
1319 sit_destroy_tunnels(sitn, &list);
1320 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1321 unregister_netdevice_many(&list);
1322 rtnl_unlock();
1323 }
1324
1325 static struct pernet_operations sit_net_ops = {
1326 .init = sit_init_net,
1327 .exit = sit_exit_net,
1328 .id = &sit_net_id,
1329 .size = sizeof(struct sit_net),
1330 };
1331
1332 static void __exit sit_cleanup(void)
1333 {
1334 rtnl_link_unregister(&sit_link_ops);
1335 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1336
1337 unregister_pernet_device(&sit_net_ops);
1338 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1339 }
1340
1341 static int __init sit_init(void)
1342 {
1343 int err;
1344
1345 pr_info("IPv6 over IPv4 tunneling driver\n");
1346
1347 err = register_pernet_device(&sit_net_ops);
1348 if (err < 0)
1349 return err;
1350 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1351 if (err < 0) {
1352 pr_info("%s: can't add protocol\n", __func__);
1353 goto xfrm_tunnel_failed;
1354 }
1355 err = rtnl_link_register(&sit_link_ops);
1356 if (err < 0)
1357 goto rtnl_link_failed;
1358
1359 out:
1360 return err;
1361
1362 rtnl_link_failed:
1363 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1364 xfrm_tunnel_failed:
1365 unregister_pernet_device(&sit_net_ops);
1366 goto out;
1367 }
1368
1369 module_init(sit_init);
1370 module_exit(sit_cleanup);
1371 MODULE_LICENSE("GPL");
1372 MODULE_ALIAS_NETDEV("sit0");