projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'drbd-8.4_ed6' into for-3.8-drivers-drbd-8.4_ed6
[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 #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
72 static int sit_net_id __read_mostly;
73 struct sit_net {
74 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
76 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
77 struct ip_tunnel __rcu *tunnels_wc[1];
78 struct ip_tunnel __rcu **tunnels[4];
79
80 struct net_device *fb_tunnel_dev;
81 };
82
83 /*
84 * Locking : hash tables are protected by RCU and RTNL
85 */
86
87 #define for_each_ip_tunnel_rcu(start) \
88 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
89
90 /* often modified stats are per cpu, other are shared (netdev->stats) */
91 struct pcpu_tstats {
92 u64 rx_packets;
93 u64 rx_bytes;
94 u64 tx_packets;
95 u64 tx_bytes;
96 struct u64_stats_sync syncp;
97 };
98
99 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
100 struct rtnl_link_stats64 *tot)
101 {
102 int i;
103
104 for_each_possible_cpu(i) {
105 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
106 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
107 unsigned int start;
108
109 do {
110 start = u64_stats_fetch_begin_bh(&tstats->syncp);
111 rx_packets = tstats->rx_packets;
112 tx_packets = tstats->tx_packets;
113 rx_bytes = tstats->rx_bytes;
114 tx_bytes = tstats->tx_bytes;
115 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
116
117 tot->rx_packets += rx_packets;
118 tot->tx_packets += tx_packets;
119 tot->rx_bytes += rx_bytes;
120 tot->tx_bytes += tx_bytes;
121 }
122
123 tot->rx_errors = dev->stats.rx_errors;
124 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
125 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
126 tot->tx_dropped = dev->stats.tx_dropped;
127 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
128 tot->tx_errors = dev->stats.tx_errors;
129
130 return tot;
131 }
132
133 /*
134 * Must be invoked with rcu_read_lock
135 */
136 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
137 struct net_device *dev, __be32 remote, __be32 local)
138 {
139 unsigned int h0 = HASH(remote);
140 unsigned int h1 = HASH(local);
141 struct ip_tunnel *t;
142 struct sit_net *sitn = net_generic(net, sit_net_id);
143
144 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
145 if (local == t->parms.iph.saddr &&
146 remote == t->parms.iph.daddr &&
147 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
148 (t->dev->flags & IFF_UP))
149 return t;
150 }
151 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
152 if (remote == t->parms.iph.daddr &&
153 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
154 (t->dev->flags & IFF_UP))
155 return t;
156 }
157 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
158 if (local == t->parms.iph.saddr &&
159 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
160 (t->dev->flags & IFF_UP))
161 return t;
162 }
163 t = rcu_dereference(sitn->tunnels_wc[0]);
164 if ((t != NULL) && (t->dev->flags & IFF_UP))
165 return t;
166 return NULL;
167 }
168
169 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
170 struct ip_tunnel_parm *parms)
171 {
172 __be32 remote = parms->iph.daddr;
173 __be32 local = parms->iph.saddr;
174 unsigned int h = 0;
175 int prio = 0;
176
177 if (remote) {
178 prio |= 2;
179 h ^= HASH(remote);
180 }
181 if (local) {
182 prio |= 1;
183 h ^= HASH(local);
184 }
185 return &sitn->tunnels[prio][h];
186 }
187
188 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
189 struct ip_tunnel *t)
190 {
191 return __ipip6_bucket(sitn, &t->parms);
192 }
193
194 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
195 {
196 struct ip_tunnel __rcu **tp;
197 struct ip_tunnel *iter;
198
199 for (tp = ipip6_bucket(sitn, t);
200 (iter = rtnl_dereference(*tp)) != NULL;
201 tp = &iter->next) {
202 if (t == iter) {
203 rcu_assign_pointer(*tp, t->next);
204 break;
205 }
206 }
207 }
208
209 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
210 {
211 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
212
213 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
214 rcu_assign_pointer(*tp, t);
215 }
216
217 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
218 {
219 #ifdef CONFIG_IPV6_SIT_6RD
220 struct ip_tunnel *t = netdev_priv(dev);
221
222 if (t->dev == sitn->fb_tunnel_dev) {
223 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
224 t->ip6rd.relay_prefix = 0;
225 t->ip6rd.prefixlen = 16;
226 t->ip6rd.relay_prefixlen = 0;
227 } else {
228 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
229 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
230 }
231 #endif
232 }
233
234 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
235 struct ip_tunnel_parm *parms, int create)
236 {
237 __be32 remote = parms->iph.daddr;
238 __be32 local = parms->iph.saddr;
239 struct ip_tunnel *t, *nt;
240 struct ip_tunnel __rcu **tp;
241 struct net_device *dev;
242 char name[IFNAMSIZ];
243 struct sit_net *sitn = net_generic(net, sit_net_id);
244
245 for (tp = __ipip6_bucket(sitn, parms);
246 (t = rtnl_dereference(*tp)) != NULL;
247 tp = &t->next) {
248 if (local == t->parms.iph.saddr &&
249 remote == t->parms.iph.daddr &&
250 parms->link == t->parms.link) {
251 if (create)
252 return NULL;
253 else
254 return t;
255 }
256 }
257 if (!create)
258 goto failed;
259
260 if (parms->name[0])
261 strlcpy(name, parms->name, IFNAMSIZ);
262 else
263 strcpy(name, "sit%d");
264
265 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
266 if (dev == NULL)
267 return NULL;
268
269 dev_net_set(dev, net);
270
271 nt = netdev_priv(dev);
272
273 nt->parms = *parms;
274 if (ipip6_tunnel_init(dev) < 0)
275 goto failed_free;
276 ipip6_tunnel_clone_6rd(dev, sitn);
277
278 if (parms->i_flags & SIT_ISATAP)
279 dev->priv_flags |= IFF_ISATAP;
280
281 if (register_netdevice(dev) < 0)
282 goto failed_free;
283
284 strcpy(nt->parms.name, dev->name);
285
286 dev_hold(dev);
287
288 ipip6_tunnel_link(sitn, nt);
289 return nt;
290
291 failed_free:
292 ipip6_dev_free(dev);
293 failed:
294 return NULL;
295 }
296
297 #define for_each_prl_rcu(start) \
298 for (prl = rcu_dereference(start); \
299 prl; \
300 prl = rcu_dereference(prl->next))
301
302 static struct ip_tunnel_prl_entry *
303 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
304 {
305 struct ip_tunnel_prl_entry *prl;
306
307 for_each_prl_rcu(t->prl)
308 if (prl->addr == addr)
309 break;
310 return prl;
311
312 }
313
314 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
315 struct ip_tunnel_prl __user *a)
316 {
317 struct ip_tunnel_prl kprl, *kp;
318 struct ip_tunnel_prl_entry *prl;
319 unsigned int cmax, c = 0, ca, len;
320 int ret = 0;
321
322 if (copy_from_user(&kprl, a, sizeof(kprl)))
323 return -EFAULT;
324 cmax = kprl.datalen / sizeof(kprl);
325 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
326 cmax = 1;
327
328 /* For simple GET or for root users,
329 * we try harder to allocate.
330 */
331 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
332 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
333 NULL;
334
335 rcu_read_lock();
336
337 ca = t->prl_count < cmax ? t->prl_count : cmax;
338
339 if (!kp) {
340 /* We don't try hard to allocate much memory for
341 * non-root users.
342 * For root users, retry allocating enough memory for
343 * the answer.
344 */
345 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
346 if (!kp) {
347 ret = -ENOMEM;
348 goto out;
349 }
350 }
351
352 c = 0;
353 for_each_prl_rcu(t->prl) {
354 if (c >= cmax)
355 break;
356 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
357 continue;
358 kp[c].addr = prl->addr;
359 kp[c].flags = prl->flags;
360 c++;
361 if (kprl.addr != htonl(INADDR_ANY))
362 break;
363 }
364 out:
365 rcu_read_unlock();
366
367 len = sizeof(*kp) * c;
368 ret = 0;
369 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
370 ret = -EFAULT;
371
372 kfree(kp);
373
374 return ret;
375 }
376
377 static int
378 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
379 {
380 struct ip_tunnel_prl_entry *p;
381 int err = 0;
382
383 if (a->addr == htonl(INADDR_ANY))
384 return -EINVAL;
385
386 ASSERT_RTNL();
387
388 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
389 if (p->addr == a->addr) {
390 if (chg) {
391 p->flags = a->flags;
392 goto out;
393 }
394 err = -EEXIST;
395 goto out;
396 }
397 }
398
399 if (chg) {
400 err = -ENXIO;
401 goto out;
402 }
403
404 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
405 if (!p) {
406 err = -ENOBUFS;
407 goto out;
408 }
409
410 p->next = t->prl;
411 p->addr = a->addr;
412 p->flags = a->flags;
413 t->prl_count++;
414 rcu_assign_pointer(t->prl, p);
415 out:
416 return err;
417 }
418
419 static void prl_list_destroy_rcu(struct rcu_head *head)
420 {
421 struct ip_tunnel_prl_entry *p, *n;
422
423 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
424 do {
425 n = rcu_dereference_protected(p->next, 1);
426 kfree(p);
427 p = n;
428 } while (p);
429 }
430
431 static int
432 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
433 {
434 struct ip_tunnel_prl_entry *x;
435 struct ip_tunnel_prl_entry __rcu **p;
436 int err = 0;
437
438 ASSERT_RTNL();
439
440 if (a && a->addr != htonl(INADDR_ANY)) {
441 for (p = &t->prl;
442 (x = rtnl_dereference(*p)) != NULL;
443 p = &x->next) {
444 if (x->addr == a->addr) {
445 *p = x->next;
446 kfree_rcu(x, rcu_head);
447 t->prl_count--;
448 goto out;
449 }
450 }
451 err = -ENXIO;
452 } else {
453 x = rtnl_dereference(t->prl);
454 if (x) {
455 t->prl_count = 0;
456 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
457 t->prl = NULL;
458 }
459 }
460 out:
461 return err;
462 }
463
464 static int
465 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
466 {
467 struct ip_tunnel_prl_entry *p;
468 int ok = 1;
469
470 rcu_read_lock();
471 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
472 if (p) {
473 if (p->flags & PRL_DEFAULT)
474 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
475 else
476 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
477 } else {
478 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
479
480 if (ipv6_addr_is_isatap(addr6) &&
481 (addr6->s6_addr32[3] == iph->saddr) &&
482 ipv6_chk_prefix(addr6, t->dev))
483 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
484 else
485 ok = 0;
486 }
487 rcu_read_unlock();
488 return ok;
489 }
490
491 static void ipip6_tunnel_uninit(struct net_device *dev)
492 {
493 struct net *net = dev_net(dev);
494 struct sit_net *sitn = net_generic(net, sit_net_id);
495
496 if (dev == sitn->fb_tunnel_dev) {
497 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
498 } else {
499 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
500 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
501 }
502 dev_put(dev);
503 }
504
505
506 static int ipip6_err(struct sk_buff *skb, u32 info)
507 {
508
509 /* All the routers (except for Linux) return only
510 8 bytes of packet payload. It means, that precise relaying of
511 ICMP in the real Internet is absolutely infeasible.
512 */
513 const struct iphdr *iph = (const struct iphdr *)skb->data;
514 const int type = icmp_hdr(skb)->type;
515 const int code = icmp_hdr(skb)->code;
516 struct ip_tunnel *t;
517 int err;
518
519 switch (type) {
520 default:
521 case ICMP_PARAMETERPROB:
522 return 0;
523
524 case ICMP_DEST_UNREACH:
525 switch (code) {
526 case ICMP_SR_FAILED:
527 case ICMP_PORT_UNREACH:
528 /* Impossible event. */
529 return 0;
530 default:
531 /* All others are translated to HOST_UNREACH.
532 rfc2003 contains "deep thoughts" about NET_UNREACH,
533 I believe they are just ether pollution. --ANK
534 */
535 break;
536 }
537 break;
538 case ICMP_TIME_EXCEEDED:
539 if (code != ICMP_EXC_TTL)
540 return 0;
541 break;
542 case ICMP_REDIRECT:
543 break;
544 }
545
546 err = -ENOENT;
547
548 t = ipip6_tunnel_lookup(dev_net(skb->dev),
549 skb->dev,
550 iph->daddr,
551 iph->saddr);
552 if (t == NULL)
553 goto out;
554
555 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
556 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
557 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
558 err = 0;
559 goto out;
560 }
561 if (type == ICMP_REDIRECT) {
562 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
563 IPPROTO_IPV6, 0);
564 err = 0;
565 goto out;
566 }
567
568 if (t->parms.iph.daddr == 0)
569 goto out;
570
571 err = 0;
572 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
573 goto out;
574
575 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
576 t->err_count++;
577 else
578 t->err_count = 1;
579 t->err_time = jiffies;
580 out:
581 return err;
582 }
583
584 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
585 {
586 if (INET_ECN_is_ce(iph->tos))
587 IP6_ECN_set_ce(ipv6_hdr(skb));
588 }
589
590 static int ipip6_rcv(struct sk_buff *skb)
591 {
592 const struct iphdr *iph;
593 struct ip_tunnel *tunnel;
594
595 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
596 goto out;
597
598 iph = ip_hdr(skb);
599
600 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
601 iph->saddr, iph->daddr);
602 if (tunnel != NULL) {
603 struct pcpu_tstats *tstats;
604
605 secpath_reset(skb);
606 skb->mac_header = skb->network_header;
607 skb_reset_network_header(skb);
608 IPCB(skb)->flags = 0;
609 skb->protocol = htons(ETH_P_IPV6);
610 skb->pkt_type = PACKET_HOST;
611
612 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
613 !isatap_chksrc(skb, iph, tunnel)) {
614 tunnel->dev->stats.rx_errors++;
615 kfree_skb(skb);
616 return 0;
617 }
618
619 tstats = this_cpu_ptr(tunnel->dev->tstats);
620 tstats->rx_packets++;
621 tstats->rx_bytes += skb->len;
622
623 __skb_tunnel_rx(skb, tunnel->dev);
624
625 ipip6_ecn_decapsulate(iph, skb);
626
627 netif_rx(skb);
628
629 return 0;
630 }
631
632 /* no tunnel matched, let upstream know, ipsec may handle it */
633 return 1;
634 out:
635 kfree_skb(skb);
636 return 0;
637 }
638
639 /*
640 * Returns the embedded IPv4 address if the IPv6 address
641 * comes from 6rd / 6to4 (RFC 3056) addr space.
642 */
643 static inline
644 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
645 {
646 __be32 dst = 0;
647
648 #ifdef CONFIG_IPV6_SIT_6RD
649 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
650 tunnel->ip6rd.prefixlen)) {
651 unsigned int pbw0, pbi0;
652 int pbi1;
653 u32 d;
654
655 pbw0 = tunnel->ip6rd.prefixlen >> 5;
656 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
657
658 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
659 tunnel->ip6rd.relay_prefixlen;
660
661 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
662 if (pbi1 > 0)
663 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
664 (32 - pbi1);
665
666 dst = tunnel->ip6rd.relay_prefix | htonl(d);
667 }
668 #else
669 if (v6dst->s6_addr16[0] == htons(0x2002)) {
670 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
671 memcpy(&dst, &v6dst->s6_addr16[1], 4);
672 }
673 #endif
674 return dst;
675 }
676
677 /*
678 * This function assumes it is being called from dev_queue_xmit()
679 * and that skb is filled properly by that function.
680 */
681
682 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
683 struct net_device *dev)
684 {
685 struct ip_tunnel *tunnel = netdev_priv(dev);
686 struct pcpu_tstats *tstats;
687 const struct iphdr *tiph = &tunnel->parms.iph;
688 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
689 u8 tos = tunnel->parms.iph.tos;
690 __be16 df = tiph->frag_off;
691 struct rtable *rt; /* Route to the other host */
692 struct net_device *tdev; /* Device to other host */
693 struct iphdr *iph; /* Our new IP header */
694 unsigned int max_headroom; /* The extra header space needed */
695 __be32 dst = tiph->daddr;
696 struct flowi4 fl4;
697 int mtu;
698 const struct in6_addr *addr6;
699 int addr_type;
700
701 if (skb->protocol != htons(ETH_P_IPV6))
702 goto tx_error;
703
704 if (tos == 1)
705 tos = ipv6_get_dsfield(iph6);
706
707 /* ISATAP (RFC4214) - must come before 6to4 */
708 if (dev->priv_flags & IFF_ISATAP) {
709 struct neighbour *neigh = NULL;
710 bool do_tx_error = false;
711
712 if (skb_dst(skb))
713 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
714
715 if (neigh == NULL) {
716 net_dbg_ratelimited("sit: nexthop == NULL\n");
717 goto tx_error;
718 }
719
720 addr6 = (const struct in6_addr *)&neigh->primary_key;
721 addr_type = ipv6_addr_type(addr6);
722
723 if ((addr_type & IPV6_ADDR_UNICAST) &&
724 ipv6_addr_is_isatap(addr6))
725 dst = addr6->s6_addr32[3];
726 else
727 do_tx_error = true;
728
729 neigh_release(neigh);
730 if (do_tx_error)
731 goto tx_error;
732 }
733
734 if (!dst)
735 dst = try_6rd(&iph6->daddr, tunnel);
736
737 if (!dst) {
738 struct neighbour *neigh = NULL;
739 bool do_tx_error = false;
740
741 if (skb_dst(skb))
742 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
743
744 if (neigh == NULL) {
745 net_dbg_ratelimited("sit: nexthop == NULL\n");
746 goto tx_error;
747 }
748
749 addr6 = (const struct in6_addr *)&neigh->primary_key;
750 addr_type = ipv6_addr_type(addr6);
751
752 if (addr_type == IPV6_ADDR_ANY) {
753 addr6 = &ipv6_hdr(skb)->daddr;
754 addr_type = ipv6_addr_type(addr6);
755 }
756
757 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
758 dst = addr6->s6_addr32[3];
759 else
760 do_tx_error = true;
761
762 neigh_release(neigh);
763 if (do_tx_error)
764 goto tx_error;
765 }
766
767 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
768 dst, tiph->saddr,
769 0, 0,
770 IPPROTO_IPV6, RT_TOS(tos),
771 tunnel->parms.link);
772 if (IS_ERR(rt)) {
773 dev->stats.tx_carrier_errors++;
774 goto tx_error_icmp;
775 }
776 if (rt->rt_type != RTN_UNICAST) {
777 ip_rt_put(rt);
778 dev->stats.tx_carrier_errors++;
779 goto tx_error_icmp;
780 }
781 tdev = rt->dst.dev;
782
783 if (tdev == dev) {
784 ip_rt_put(rt);
785 dev->stats.collisions++;
786 goto tx_error;
787 }
788
789 if (df) {
790 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
791
792 if (mtu < 68) {
793 dev->stats.collisions++;
794 ip_rt_put(rt);
795 goto tx_error;
796 }
797
798 if (mtu < IPV6_MIN_MTU) {
799 mtu = IPV6_MIN_MTU;
800 df = 0;
801 }
802
803 if (tunnel->parms.iph.daddr && skb_dst(skb))
804 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
805
806 if (skb->len > mtu) {
807 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
808 ip_rt_put(rt);
809 goto tx_error;
810 }
811 }
812
813 if (tunnel->err_count > 0) {
814 if (time_before(jiffies,
815 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
816 tunnel->err_count--;
817 dst_link_failure(skb);
818 } else
819 tunnel->err_count = 0;
820 }
821
822 /*
823 * Okay, now see if we can stuff it in the buffer as-is.
824 */
825 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
826
827 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
828 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
829 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
830 if (!new_skb) {
831 ip_rt_put(rt);
832 dev->stats.tx_dropped++;
833 dev_kfree_skb(skb);
834 return NETDEV_TX_OK;
835 }
836 if (skb->sk)
837 skb_set_owner_w(new_skb, skb->sk);
838 dev_kfree_skb(skb);
839 skb = new_skb;
840 iph6 = ipv6_hdr(skb);
841 }
842
843 skb->transport_header = skb->network_header;
844 skb_push(skb, sizeof(struct iphdr));
845 skb_reset_network_header(skb);
846 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
847 IPCB(skb)->flags = 0;
848 skb_dst_drop(skb);
849 skb_dst_set(skb, &rt->dst);
850
851 /*
852 * Push down and install the IPIP header.
853 */
854
855 iph = ip_hdr(skb);
856 iph->version = 4;
857 iph->ihl = sizeof(struct iphdr)>>2;
858 iph->frag_off = df;
859 iph->protocol = IPPROTO_IPV6;
860 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
861 iph->daddr = fl4.daddr;
862 iph->saddr = fl4.saddr;
863
864 if ((iph->ttl = tiph->ttl) == 0)
865 iph->ttl = iph6->hop_limit;
866
867 nf_reset(skb);
868 tstats = this_cpu_ptr(dev->tstats);
869 __IPTUNNEL_XMIT(tstats, &dev->stats);
870 return NETDEV_TX_OK;
871
872 tx_error_icmp:
873 dst_link_failure(skb);
874 tx_error:
875 dev->stats.tx_errors++;
876 dev_kfree_skb(skb);
877 return NETDEV_TX_OK;
878 }
879
880 static void ipip6_tunnel_bind_dev(struct net_device *dev)
881 {
882 struct net_device *tdev = NULL;
883 struct ip_tunnel *tunnel;
884 const struct iphdr *iph;
885 struct flowi4 fl4;
886
887 tunnel = netdev_priv(dev);
888 iph = &tunnel->parms.iph;
889
890 if (iph->daddr) {
891 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
892 iph->daddr, iph->saddr,
893 0, 0,
894 IPPROTO_IPV6,
895 RT_TOS(iph->tos),
896 tunnel->parms.link);
897
898 if (!IS_ERR(rt)) {
899 tdev = rt->dst.dev;
900 ip_rt_put(rt);
901 }
902 dev->flags |= IFF_POINTOPOINT;
903 }
904
905 if (!tdev && tunnel->parms.link)
906 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
907
908 if (tdev) {
909 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
910 dev->mtu = tdev->mtu - sizeof(struct iphdr);
911 if (dev->mtu < IPV6_MIN_MTU)
912 dev->mtu = IPV6_MIN_MTU;
913 }
914 dev->iflink = tunnel->parms.link;
915 }
916
917 static int
918 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
919 {
920 int err = 0;
921 struct ip_tunnel_parm p;
922 struct ip_tunnel_prl prl;
923 struct ip_tunnel *t;
924 struct net *net = dev_net(dev);
925 struct sit_net *sitn = net_generic(net, sit_net_id);
926 #ifdef CONFIG_IPV6_SIT_6RD
927 struct ip_tunnel_6rd ip6rd;
928 #endif
929
930 switch (cmd) {
931 case SIOCGETTUNNEL:
932 #ifdef CONFIG_IPV6_SIT_6RD
933 case SIOCGET6RD:
934 #endif
935 t = NULL;
936 if (dev == sitn->fb_tunnel_dev) {
937 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
938 err = -EFAULT;
939 break;
940 }
941 t = ipip6_tunnel_locate(net, &p, 0);
942 }
943 if (t == NULL)
944 t = netdev_priv(dev);
945
946 err = -EFAULT;
947 if (cmd == SIOCGETTUNNEL) {
948 memcpy(&p, &t->parms, sizeof(p));
949 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
950 sizeof(p)))
951 goto done;
952 #ifdef CONFIG_IPV6_SIT_6RD
953 } else {
954 ip6rd.prefix = t->ip6rd.prefix;
955 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
956 ip6rd.prefixlen = t->ip6rd.prefixlen;
957 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
958 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
959 sizeof(ip6rd)))
960 goto done;
961 #endif
962 }
963 err = 0;
964 break;
965
966 case SIOCADDTUNNEL:
967 case SIOCCHGTUNNEL:
968 err = -EPERM;
969 if (!capable(CAP_NET_ADMIN))
970 goto done;
971
972 err = -EFAULT;
973 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
974 goto done;
975
976 err = -EINVAL;
977 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
978 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
979 goto done;
980 if (p.iph.ttl)
981 p.iph.frag_off |= htons(IP_DF);
982
983 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
984
985 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
986 if (t != NULL) {
987 if (t->dev != dev) {
988 err = -EEXIST;
989 break;
990 }
991 } else {
992 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
993 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
994 err = -EINVAL;
995 break;
996 }
997 t = netdev_priv(dev);
998 ipip6_tunnel_unlink(sitn, t);
999 synchronize_net();
1000 t->parms.iph.saddr = p.iph.saddr;
1001 t->parms.iph.daddr = p.iph.daddr;
1002 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1003 memcpy(dev->broadcast, &p.iph.daddr, 4);
1004 ipip6_tunnel_link(sitn, t);
1005 netdev_state_change(dev);
1006 }
1007 }
1008
1009 if (t) {
1010 err = 0;
1011 if (cmd == SIOCCHGTUNNEL) {
1012 t->parms.iph.ttl = p.iph.ttl;
1013 t->parms.iph.tos = p.iph.tos;
1014 if (t->parms.link != p.link) {
1015 t->parms.link = p.link;
1016 ipip6_tunnel_bind_dev(dev);
1017 netdev_state_change(dev);
1018 }
1019 }
1020 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1021 err = -EFAULT;
1022 } else
1023 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1024 break;
1025
1026 case SIOCDELTUNNEL:
1027 err = -EPERM;
1028 if (!capable(CAP_NET_ADMIN))
1029 goto done;
1030
1031 if (dev == sitn->fb_tunnel_dev) {
1032 err = -EFAULT;
1033 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1034 goto done;
1035 err = -ENOENT;
1036 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1037 goto done;
1038 err = -EPERM;
1039 if (t == netdev_priv(sitn->fb_tunnel_dev))
1040 goto done;
1041 dev = t->dev;
1042 }
1043 unregister_netdevice(dev);
1044 err = 0;
1045 break;
1046
1047 case SIOCGETPRL:
1048 err = -EINVAL;
1049 if (dev == sitn->fb_tunnel_dev)
1050 goto done;
1051 err = -ENOENT;
1052 if (!(t = netdev_priv(dev)))
1053 goto done;
1054 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1055 break;
1056
1057 case SIOCADDPRL:
1058 case SIOCDELPRL:
1059 case SIOCCHGPRL:
1060 err = -EPERM;
1061 if (!capable(CAP_NET_ADMIN))
1062 goto done;
1063 err = -EINVAL;
1064 if (dev == sitn->fb_tunnel_dev)
1065 goto done;
1066 err = -EFAULT;
1067 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1068 goto done;
1069 err = -ENOENT;
1070 if (!(t = netdev_priv(dev)))
1071 goto done;
1072
1073 switch (cmd) {
1074 case SIOCDELPRL:
1075 err = ipip6_tunnel_del_prl(t, &prl);
1076 break;
1077 case SIOCADDPRL:
1078 case SIOCCHGPRL:
1079 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1080 break;
1081 }
1082 netdev_state_change(dev);
1083 break;
1084
1085 #ifdef CONFIG_IPV6_SIT_6RD
1086 case SIOCADD6RD:
1087 case SIOCCHG6RD:
1088 case SIOCDEL6RD:
1089 err = -EPERM;
1090 if (!capable(CAP_NET_ADMIN))
1091 goto done;
1092
1093 err = -EFAULT;
1094 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1095 sizeof(ip6rd)))
1096 goto done;
1097
1098 t = netdev_priv(dev);
1099
1100 if (cmd != SIOCDEL6RD) {
1101 struct in6_addr prefix;
1102 __be32 relay_prefix;
1103
1104 err = -EINVAL;
1105 if (ip6rd.relay_prefixlen > 32 ||
1106 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1107 goto done;
1108
1109 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1110 ip6rd.prefixlen);
1111 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1112 goto done;
1113 if (ip6rd.relay_prefixlen)
1114 relay_prefix = ip6rd.relay_prefix &
1115 htonl(0xffffffffUL <<
1116 (32 - ip6rd.relay_prefixlen));
1117 else
1118 relay_prefix = 0;
1119 if (relay_prefix != ip6rd.relay_prefix)
1120 goto done;
1121
1122 t->ip6rd.prefix = prefix;
1123 t->ip6rd.relay_prefix = relay_prefix;
1124 t->ip6rd.prefixlen = ip6rd.prefixlen;
1125 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1126 } else
1127 ipip6_tunnel_clone_6rd(dev, sitn);
1128
1129 err = 0;
1130 break;
1131 #endif
1132
1133 default:
1134 err = -EINVAL;
1135 }
1136
1137 done:
1138 return err;
1139 }
1140
1141 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1142 {
1143 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1144 return -EINVAL;
1145 dev->mtu = new_mtu;
1146 return 0;
1147 }
1148
1149 static const struct net_device_ops ipip6_netdev_ops = {
1150 .ndo_uninit = ipip6_tunnel_uninit,
1151 .ndo_start_xmit = ipip6_tunnel_xmit,
1152 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1153 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1154 .ndo_get_stats64= ipip6_get_stats64,
1155 };
1156
1157 static void ipip6_dev_free(struct net_device *dev)
1158 {
1159 free_percpu(dev->tstats);
1160 free_netdev(dev);
1161 }
1162
1163 static void ipip6_tunnel_setup(struct net_device *dev)
1164 {
1165 dev->netdev_ops = &ipip6_netdev_ops;
1166 dev->destructor = ipip6_dev_free;
1167
1168 dev->type = ARPHRD_SIT;
1169 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1170 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1171 dev->flags = IFF_NOARP;
1172 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1173 dev->iflink = 0;
1174 dev->addr_len = 4;
1175 dev->features |= NETIF_F_NETNS_LOCAL;
1176 dev->features |= NETIF_F_LLTX;
1177 }
1178
1179 static int ipip6_tunnel_init(struct net_device *dev)
1180 {
1181 struct ip_tunnel *tunnel = netdev_priv(dev);
1182
1183 tunnel->dev = dev;
1184
1185 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1186 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1187
1188 ipip6_tunnel_bind_dev(dev);
1189 dev->tstats = alloc_percpu(struct pcpu_tstats);
1190 if (!dev->tstats)
1191 return -ENOMEM;
1192
1193 return 0;
1194 }
1195
1196 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1197 {
1198 struct ip_tunnel *tunnel = netdev_priv(dev);
1199 struct iphdr *iph = &tunnel->parms.iph;
1200 struct net *net = dev_net(dev);
1201 struct sit_net *sitn = net_generic(net, sit_net_id);
1202
1203 tunnel->dev = dev;
1204 strcpy(tunnel->parms.name, dev->name);
1205
1206 iph->version = 4;
1207 iph->protocol = IPPROTO_IPV6;
1208 iph->ihl = 5;
1209 iph->ttl = 64;
1210
1211 dev->tstats = alloc_percpu(struct pcpu_tstats);
1212 if (!dev->tstats)
1213 return -ENOMEM;
1214 dev_hold(dev);
1215 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1216 return 0;
1217 }
1218
1219 static struct xfrm_tunnel sit_handler __read_mostly = {
1220 .handler = ipip6_rcv,
1221 .err_handler = ipip6_err,
1222 .priority = 1,
1223 };
1224
1225 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1226 {
1227 int prio;
1228
1229 for (prio = 1; prio < 4; prio++) {
1230 int h;
1231 for (h = 0; h < HASH_SIZE; h++) {
1232 struct ip_tunnel *t;
1233
1234 t = rtnl_dereference(sitn->tunnels[prio][h]);
1235 while (t != NULL) {
1236 unregister_netdevice_queue(t->dev, head);
1237 t = rtnl_dereference(t->next);
1238 }
1239 }
1240 }
1241 }
1242
1243 static int __net_init sit_init_net(struct net *net)
1244 {
1245 struct sit_net *sitn = net_generic(net, sit_net_id);
1246 struct ip_tunnel *t;
1247 int err;
1248
1249 sitn->tunnels[0] = sitn->tunnels_wc;
1250 sitn->tunnels[1] = sitn->tunnels_l;
1251 sitn->tunnels[2] = sitn->tunnels_r;
1252 sitn->tunnels[3] = sitn->tunnels_r_l;
1253
1254 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1255 ipip6_tunnel_setup);
1256 if (!sitn->fb_tunnel_dev) {
1257 err = -ENOMEM;
1258 goto err_alloc_dev;
1259 }
1260 dev_net_set(sitn->fb_tunnel_dev, net);
1261
1262 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1263 if (err)
1264 goto err_dev_free;
1265
1266 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1267
1268 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1269 goto err_reg_dev;
1270
1271 t = netdev_priv(sitn->fb_tunnel_dev);
1272
1273 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1274 return 0;
1275
1276 err_reg_dev:
1277 dev_put(sitn->fb_tunnel_dev);
1278 err_dev_free:
1279 ipip6_dev_free(sitn->fb_tunnel_dev);
1280 err_alloc_dev:
1281 return err;
1282 }
1283
1284 static void __net_exit sit_exit_net(struct net *net)
1285 {
1286 struct sit_net *sitn = net_generic(net, sit_net_id);
1287 LIST_HEAD(list);
1288
1289 rtnl_lock();
1290 sit_destroy_tunnels(sitn, &list);
1291 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1292 unregister_netdevice_many(&list);
1293 rtnl_unlock();
1294 }
1295
1296 static struct pernet_operations sit_net_ops = {
1297 .init = sit_init_net,
1298 .exit = sit_exit_net,
1299 .id = &sit_net_id,
1300 .size = sizeof(struct sit_net),
1301 };
1302
1303 static void __exit sit_cleanup(void)
1304 {
1305 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1306
1307 unregister_pernet_device(&sit_net_ops);
1308 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1309 }
1310
1311 static int __init sit_init(void)
1312 {
1313 int err;
1314
1315 pr_info("IPv6 over IPv4 tunneling driver\n");
1316
1317 err = register_pernet_device(&sit_net_ops);
1318 if (err < 0)
1319 return err;
1320 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1321 if (err < 0) {
1322 unregister_pernet_device(&sit_net_ops);
1323 pr_info("%s: can't add protocol\n", __func__);
1324 }
1325 return err;
1326 }
1327
1328 module_init(sit_init);
1329 module_exit(sit_cleanup);
1330 MODULE_LICENSE("GPL");
1331 MODULE_ALIAS_NETDEV("sit0");
Linux kernel - Deliverables
This page took 0.056292 seconds and 5 git commands to generate.