?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
KEYS: add missing permission check for request_key() destination
[linux/fpc-iii.git] / net / ipv6 / sit.c
blob96b442ea904af00be71f4b3e3088aa8341f22825
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/ip_tunnels.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 bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78
79 static int sit_net_id __read_mostly;
80 struct sit_net {
81 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 struct ip_tunnel __rcu *tunnels_wc[1];
85 struct ip_tunnel __rcu **tunnels[4];
86
87 struct net_device *fb_tunnel_dev;
88 };
89
90 /*
91 * Must be invoked with rcu_read_lock
92 */
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 struct net_device *dev, __be32 remote, __be32 local)
95 {
96 unsigned int h0 = HASH(remote);
97 unsigned int h1 = HASH(local);
98 struct ip_tunnel *t;
99 struct sit_net *sitn = net_generic(net, sit_net_id);
100
101 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 if (local == t->parms.iph.saddr &&
103 remote == t->parms.iph.daddr &&
104 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
106 return t;
107 }
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111 (t->dev->flags & IFF_UP))
112 return t;
113 }
114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 if (local == t->parms.iph.saddr &&
116 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
118 return t;
119 }
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if ((t != NULL) && (t->dev->flags & IFF_UP))
122 return t;
123 return NULL;
124 }
125
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 struct ip_tunnel_parm *parms)
128 {
129 __be32 remote = parms->iph.daddr;
130 __be32 local = parms->iph.saddr;
131 unsigned int h = 0;
132 int prio = 0;
133
134 if (remote) {
135 prio |= 2;
136 h ^= HASH(remote);
137 }
138 if (local) {
139 prio |= 1;
140 h ^= HASH(local);
141 }
142 return &sitn->tunnels[prio][h];
143 }
144
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146 struct ip_tunnel *t)
147 {
148 return __ipip6_bucket(sitn, &t->parms);
149 }
150
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152 {
153 struct ip_tunnel __rcu **tp;
154 struct ip_tunnel *iter;
155
156 for (tp = ipip6_bucket(sitn, t);
157 (iter = rtnl_dereference(*tp)) != NULL;
158 tp = &iter->next) {
159 if (t == iter) {
160 rcu_assign_pointer(*tp, t->next);
161 break;
162 }
163 }
164 }
165
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167 {
168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169
170 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 rcu_assign_pointer(*tp, t);
172 }
173
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175 {
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel *t = netdev_priv(dev);
178
179 if (t->dev == sitn->fb_tunnel_dev) {
180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 t->ip6rd.relay_prefix = 0;
182 t->ip6rd.prefixlen = 16;
183 t->ip6rd.relay_prefixlen = 0;
184 } else {
185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187 }
188 #endif
189 }
190
191 static int ipip6_tunnel_create(struct net_device *dev)
192 {
193 struct ip_tunnel *t = netdev_priv(dev);
194 struct net *net = dev_net(dev);
195 struct sit_net *sitn = net_generic(net, sit_net_id);
196 int err;
197
198 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
199 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
200
201 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
202 dev->priv_flags |= IFF_ISATAP;
203
204 err = register_netdevice(dev);
205 if (err < 0)
206 goto out;
207
208 ipip6_tunnel_clone_6rd(dev, sitn);
209
210 dev->rtnl_link_ops = &sit_link_ops;
211
212 dev_hold(dev);
213
214 ipip6_tunnel_link(sitn, t);
215 return 0;
216
217 out:
218 return err;
219 }
220
221 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
222 struct ip_tunnel_parm *parms, int create)
223 {
224 __be32 remote = parms->iph.daddr;
225 __be32 local = parms->iph.saddr;
226 struct ip_tunnel *t, *nt;
227 struct ip_tunnel __rcu **tp;
228 struct net_device *dev;
229 char name[IFNAMSIZ];
230 struct sit_net *sitn = net_generic(net, sit_net_id);
231
232 for (tp = __ipip6_bucket(sitn, parms);
233 (t = rtnl_dereference(*tp)) != NULL;
234 tp = &t->next) {
235 if (local == t->parms.iph.saddr &&
236 remote == t->parms.iph.daddr &&
237 parms->link == t->parms.link) {
238 if (create)
239 return NULL;
240 else
241 return t;
242 }
243 }
244 if (!create)
245 goto failed;
246
247 if (parms->name[0])
248 strlcpy(name, parms->name, IFNAMSIZ);
249 else
250 strcpy(name, "sit%d");
251
252 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
253 if (dev == NULL)
254 return NULL;
255
256 dev_net_set(dev, net);
257
258 nt = netdev_priv(dev);
259
260 nt->parms = *parms;
261 if (ipip6_tunnel_create(dev) < 0)
262 goto failed_free;
263
264 return nt;
265
266 failed_free:
267 ipip6_dev_free(dev);
268 failed:
269 return NULL;
270 }
271
272 #define for_each_prl_rcu(start) \
273 for (prl = rcu_dereference(start); \
274 prl; \
275 prl = rcu_dereference(prl->next))
276
277 static struct ip_tunnel_prl_entry *
278 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
279 {
280 struct ip_tunnel_prl_entry *prl;
281
282 for_each_prl_rcu(t->prl)
283 if (prl->addr == addr)
284 break;
285 return prl;
286
287 }
288
289 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
290 struct ip_tunnel_prl __user *a)
291 {
292 struct ip_tunnel_prl kprl, *kp;
293 struct ip_tunnel_prl_entry *prl;
294 unsigned int cmax, c = 0, ca, len;
295 int ret = 0;
296
297 if (copy_from_user(&kprl, a, sizeof(kprl)))
298 return -EFAULT;
299 cmax = kprl.datalen / sizeof(kprl);
300 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
301 cmax = 1;
302
303 /* For simple GET or for root users,
304 * we try harder to allocate.
305 */
306 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
307 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
308 NULL;
309
310 rcu_read_lock();
311
312 ca = t->prl_count < cmax ? t->prl_count : cmax;
313
314 if (!kp) {
315 /* We don't try hard to allocate much memory for
316 * non-root users.
317 * For root users, retry allocating enough memory for
318 * the answer.
319 */
320 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
321 if (!kp) {
322 ret = -ENOMEM;
323 goto out;
324 }
325 }
326
327 c = 0;
328 for_each_prl_rcu(t->prl) {
329 if (c >= cmax)
330 break;
331 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
332 continue;
333 kp[c].addr = prl->addr;
334 kp[c].flags = prl->flags;
335 c++;
336 if (kprl.addr != htonl(INADDR_ANY))
337 break;
338 }
339 out:
340 rcu_read_unlock();
341
342 len = sizeof(*kp) * c;
343 ret = 0;
344 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
345 ret = -EFAULT;
346
347 kfree(kp);
348
349 return ret;
350 }
351
352 static int
353 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
354 {
355 struct ip_tunnel_prl_entry *p;
356 int err = 0;
357
358 if (a->addr == htonl(INADDR_ANY))
359 return -EINVAL;
360
361 ASSERT_RTNL();
362
363 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
364 if (p->addr == a->addr) {
365 if (chg) {
366 p->flags = a->flags;
367 goto out;
368 }
369 err = -EEXIST;
370 goto out;
371 }
372 }
373
374 if (chg) {
375 err = -ENXIO;
376 goto out;
377 }
378
379 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
380 if (!p) {
381 err = -ENOBUFS;
382 goto out;
383 }
384
385 p->next = t->prl;
386 p->addr = a->addr;
387 p->flags = a->flags;
388 t->prl_count++;
389 rcu_assign_pointer(t->prl, p);
390 out:
391 return err;
392 }
393
394 static void prl_list_destroy_rcu(struct rcu_head *head)
395 {
396 struct ip_tunnel_prl_entry *p, *n;
397
398 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
399 do {
400 n = rcu_dereference_protected(p->next, 1);
401 kfree(p);
402 p = n;
403 } while (p);
404 }
405
406 static int
407 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
408 {
409 struct ip_tunnel_prl_entry *x;
410 struct ip_tunnel_prl_entry __rcu **p;
411 int err = 0;
412
413 ASSERT_RTNL();
414
415 if (a && a->addr != htonl(INADDR_ANY)) {
416 for (p = &t->prl;
417 (x = rtnl_dereference(*p)) != NULL;
418 p = &x->next) {
419 if (x->addr == a->addr) {
420 *p = x->next;
421 kfree_rcu(x, rcu_head);
422 t->prl_count--;
423 goto out;
424 }
425 }
426 err = -ENXIO;
427 } else {
428 x = rtnl_dereference(t->prl);
429 if (x) {
430 t->prl_count = 0;
431 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
432 t->prl = NULL;
433 }
434 }
435 out:
436 return err;
437 }
438
439 static int
440 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
441 {
442 struct ip_tunnel_prl_entry *p;
443 int ok = 1;
444
445 rcu_read_lock();
446 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
447 if (p) {
448 if (p->flags & PRL_DEFAULT)
449 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
450 else
451 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
452 } else {
453 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
454
455 if (ipv6_addr_is_isatap(addr6) &&
456 (addr6->s6_addr32[3] == iph->saddr) &&
457 ipv6_chk_prefix(addr6, t->dev))
458 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
459 else
460 ok = 0;
461 }
462 rcu_read_unlock();
463 return ok;
464 }
465
466 static void ipip6_tunnel_uninit(struct net_device *dev)
467 {
468 struct ip_tunnel *tunnel = netdev_priv(dev);
469 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
470
471 if (dev == sitn->fb_tunnel_dev) {
472 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
473 } else {
474 ipip6_tunnel_unlink(sitn, tunnel);
475 ipip6_tunnel_del_prl(tunnel, NULL);
476 }
477 ip_tunnel_dst_reset_all(tunnel);
478 dev_put(dev);
479 }
480
481 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
482 * if sufficient data bytes are available
483 */
484 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
485 {
486 const struct iphdr *iph = (const struct iphdr *) skb->data;
487 struct rt6_info *rt;
488 struct sk_buff *skb2;
489
490 if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
491 return 1;
492
493 skb2 = skb_clone(skb, GFP_ATOMIC);
494
495 if (!skb2)
496 return 1;
497
498 skb_dst_drop(skb2);
499 skb_pull(skb2, iph->ihl * 4);
500 skb_reset_network_header(skb2);
501
502 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
503
504 if (rt && rt->dst.dev)
505 skb2->dev = rt->dst.dev;
506
507 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
508
509 if (rt)
510 ip6_rt_put(rt);
511
512 kfree_skb(skb2);
513
514 return 0;
515 }
516
517 static int ipip6_err(struct sk_buff *skb, u32 info)
518 {
519 const struct iphdr *iph = (const struct iphdr *)skb->data;
520 const int type = icmp_hdr(skb)->type;
521 const int code = icmp_hdr(skb)->code;
522 struct ip_tunnel *t;
523 int err;
524
525 switch (type) {
526 default:
527 case ICMP_PARAMETERPROB:
528 return 0;
529
530 case ICMP_DEST_UNREACH:
531 switch (code) {
532 case ICMP_SR_FAILED:
533 /* Impossible event. */
534 return 0;
535 default:
536 /* All others are translated to HOST_UNREACH.
537 rfc2003 contains "deep thoughts" about NET_UNREACH,
538 I believe they are just ether pollution. --ANK
539 */
540 break;
541 }
542 break;
543 case ICMP_TIME_EXCEEDED:
544 if (code != ICMP_EXC_TTL)
545 return 0;
546 break;
547 case ICMP_REDIRECT:
548 break;
549 }
550
551 err = -ENOENT;
552
553 t = ipip6_tunnel_lookup(dev_net(skb->dev),
554 skb->dev,
555 iph->daddr,
556 iph->saddr);
557 if (t == NULL)
558 goto out;
559
560 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
561 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
562 t->parms.link, 0, iph->protocol, 0);
563 err = 0;
564 goto out;
565 }
566 if (type == ICMP_REDIRECT) {
567 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
568 iph->protocol, 0);
569 err = 0;
570 goto out;
571 }
572
573 if (t->parms.iph.daddr == 0)
574 goto out;
575
576 err = 0;
577 if (!ipip6_err_gen_icmpv6_unreach(skb))
578 goto out;
579
580 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
581 goto out;
582
583 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
584 t->err_count++;
585 else
586 t->err_count = 1;
587 t->err_time = jiffies;
588 out:
589 return err;
590 }
591
592 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
593 const struct in6_addr *v6addr)
594 {
595 __be32 v4embed = 0;
596 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
597 return true;
598 return false;
599 }
600
601 /* Checks if an address matches an address on the tunnel interface.
602 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
603 * Long story:
604 * This function is called after we considered the packet as spoofed
605 * in is_spoofed_6rd.
606 * We may have a router that is doing NAT for proto 41 packets
607 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
608 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
609 * function will return true, dropping the packet.
610 * But, we can still check if is spoofed against the IP
611 * addresses associated with the interface.
612 */
613 static bool only_dnatted(const struct ip_tunnel *tunnel,
614 const struct in6_addr *v6dst)
615 {
616 int prefix_len;
617
618 #ifdef CONFIG_IPV6_SIT_6RD
619 prefix_len = tunnel->ip6rd.prefixlen + 32
620 - tunnel->ip6rd.relay_prefixlen;
621 #else
622 prefix_len = 48;
623 #endif
624 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
625 }
626
627 /* Returns true if a packet is spoofed */
628 static bool packet_is_spoofed(struct sk_buff *skb,
629 const struct iphdr *iph,
630 struct ip_tunnel *tunnel)
631 {
632 const struct ipv6hdr *ipv6h;
633
634 if (tunnel->dev->priv_flags & IFF_ISATAP) {
635 if (!isatap_chksrc(skb, iph, tunnel))
636 return true;
637
638 return false;
639 }
640
641 if (tunnel->dev->flags & IFF_POINTOPOINT)
642 return false;
643
644 ipv6h = ipv6_hdr(skb);
645
646 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
647 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
648 &iph->saddr, &ipv6h->saddr,
649 &iph->daddr, &ipv6h->daddr);
650 return true;
651 }
652
653 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
654 return false;
655
656 if (only_dnatted(tunnel, &ipv6h->daddr))
657 return false;
658
659 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
660 &iph->saddr, &ipv6h->saddr,
661 &iph->daddr, &ipv6h->daddr);
662 return true;
663 }
664
665 static int ipip6_rcv(struct sk_buff *skb)
666 {
667 const struct iphdr *iph = ip_hdr(skb);
668 struct ip_tunnel *tunnel;
669 int err;
670
671 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
672 iph->saddr, iph->daddr);
673 if (tunnel != NULL) {
674 struct pcpu_sw_netstats *tstats;
675
676 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
677 tunnel->parms.iph.protocol != 0)
678 goto out;
679
680 skb->mac_header = skb->network_header;
681 skb_reset_network_header(skb);
682 IPCB(skb)->flags = 0;
683 skb->protocol = htons(ETH_P_IPV6);
684
685 if (packet_is_spoofed(skb, iph, tunnel)) {
686 tunnel->dev->stats.rx_errors++;
687 goto out;
688 }
689
690 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
691
692 err = IP_ECN_decapsulate(iph, skb);
693 if (unlikely(err)) {
694 if (log_ecn_error)
695 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
696 &iph->saddr, iph->tos);
697 if (err > 1) {
698 ++tunnel->dev->stats.rx_frame_errors;
699 ++tunnel->dev->stats.rx_errors;
700 goto out;
701 }
702 }
703
704 tstats = this_cpu_ptr(tunnel->dev->tstats);
705 u64_stats_update_begin(&tstats->syncp);
706 tstats->rx_packets++;
707 tstats->rx_bytes += skb->len;
708 u64_stats_update_end(&tstats->syncp);
709
710 netif_rx(skb);
711
712 return 0;
713 }
714
715 /* no tunnel matched, let upstream know, ipsec may handle it */
716 return 1;
717 out:
718 kfree_skb(skb);
719 return 0;
720 }
721
722 static const struct tnl_ptk_info tpi = {
723 /* no tunnel info required for ipip. */
724 .proto = htons(ETH_P_IP),
725 };
726
727 static int ipip_rcv(struct sk_buff *skb)
728 {
729 const struct iphdr *iph;
730 struct ip_tunnel *tunnel;
731
732 iph = ip_hdr(skb);
733 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
734 iph->saddr, iph->daddr);
735 if (tunnel != NULL) {
736 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
737 tunnel->parms.iph.protocol != 0)
738 goto drop;
739
740 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
741 goto drop;
742 if (iptunnel_pull_header(skb, 0, tpi.proto))
743 goto drop;
744 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
745 }
746
747 return 1;
748
749 drop:
750 kfree_skb(skb);
751 return 0;
752 }
753
754 /*
755 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
756 * stores the embedded IPv4 address in v4dst and returns true.
757 */
758 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
759 __be32 *v4dst)
760 {
761 #ifdef CONFIG_IPV6_SIT_6RD
762 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
763 tunnel->ip6rd.prefixlen)) {
764 unsigned int pbw0, pbi0;
765 int pbi1;
766 u32 d;
767
768 pbw0 = tunnel->ip6rd.prefixlen >> 5;
769 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
770
771 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
772 tunnel->ip6rd.relay_prefixlen;
773
774 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
775 if (pbi1 > 0)
776 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
777 (32 - pbi1);
778
779 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
780 return true;
781 }
782 #else
783 if (v6dst->s6_addr16[0] == htons(0x2002)) {
784 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
785 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
786 return true;
787 }
788 #endif
789 return false;
790 }
791
792 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
793 const struct in6_addr *v6dst)
794 {
795 __be32 dst = 0;
796 check_6rd(tunnel, v6dst, &dst);
797 return dst;
798 }
799
800 /*
801 * This function assumes it is being called from dev_queue_xmit()
802 * and that skb is filled properly by that function.
803 */
804
805 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
806 struct net_device *dev)
807 {
808 struct ip_tunnel *tunnel = netdev_priv(dev);
809 const struct iphdr *tiph = &tunnel->parms.iph;
810 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
811 u8 tos = tunnel->parms.iph.tos;
812 __be16 df = tiph->frag_off;
813 struct rtable *rt; /* Route to the other host */
814 struct net_device *tdev; /* Device to other host */
815 unsigned int max_headroom; /* The extra header space needed */
816 __be32 dst = tiph->daddr;
817 struct flowi4 fl4;
818 int mtu;
819 const struct in6_addr *addr6;
820 int addr_type;
821 u8 ttl;
822 int err;
823
824 if (skb->protocol != htons(ETH_P_IPV6))
825 goto tx_error;
826
827 if (tos == 1)
828 tos = ipv6_get_dsfield(iph6);
829
830 /* ISATAP (RFC4214) - must come before 6to4 */
831 if (dev->priv_flags & IFF_ISATAP) {
832 struct neighbour *neigh = NULL;
833 bool do_tx_error = false;
834
835 if (skb_dst(skb))
836 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
837
838 if (neigh == NULL) {
839 net_dbg_ratelimited("nexthop == NULL\n");
840 goto tx_error;
841 }
842
843 addr6 = (const struct in6_addr *)&neigh->primary_key;
844 addr_type = ipv6_addr_type(addr6);
845
846 if ((addr_type & IPV6_ADDR_UNICAST) &&
847 ipv6_addr_is_isatap(addr6))
848 dst = addr6->s6_addr32[3];
849 else
850 do_tx_error = true;
851
852 neigh_release(neigh);
853 if (do_tx_error)
854 goto tx_error;
855 }
856
857 if (!dst)
858 dst = try_6rd(tunnel, &iph6->daddr);
859
860 if (!dst) {
861 struct neighbour *neigh = NULL;
862 bool do_tx_error = false;
863
864 if (skb_dst(skb))
865 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
866
867 if (neigh == NULL) {
868 net_dbg_ratelimited("nexthop == NULL\n");
869 goto tx_error;
870 }
871
872 addr6 = (const struct in6_addr *)&neigh->primary_key;
873 addr_type = ipv6_addr_type(addr6);
874
875 if (addr_type == IPV6_ADDR_ANY) {
876 addr6 = &ipv6_hdr(skb)->daddr;
877 addr_type = ipv6_addr_type(addr6);
878 }
879
880 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
881 dst = addr6->s6_addr32[3];
882 else
883 do_tx_error = true;
884
885 neigh_release(neigh);
886 if (do_tx_error)
887 goto tx_error;
888 }
889
890 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
891 dst, tiph->saddr,
892 0, 0,
893 IPPROTO_IPV6, RT_TOS(tos),
894 tunnel->parms.link);
895 if (IS_ERR(rt)) {
896 dev->stats.tx_carrier_errors++;
897 goto tx_error_icmp;
898 }
899 if (rt->rt_type != RTN_UNICAST) {
900 ip_rt_put(rt);
901 dev->stats.tx_carrier_errors++;
902 goto tx_error_icmp;
903 }
904 tdev = rt->dst.dev;
905
906 if (tdev == dev) {
907 ip_rt_put(rt);
908 dev->stats.collisions++;
909 goto tx_error;
910 }
911
912 if (df) {
913 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
914
915 if (mtu < 68) {
916 dev->stats.collisions++;
917 ip_rt_put(rt);
918 goto tx_error;
919 }
920
921 if (mtu < IPV6_MIN_MTU) {
922 mtu = IPV6_MIN_MTU;
923 df = 0;
924 }
925
926 if (tunnel->parms.iph.daddr && skb_dst(skb))
927 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
928
929 if (skb->len > mtu && !skb_is_gso(skb)) {
930 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
931 ip_rt_put(rt);
932 goto tx_error;
933 }
934 }
935
936 if (tunnel->err_count > 0) {
937 if (time_before(jiffies,
938 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
939 tunnel->err_count--;
940 dst_link_failure(skb);
941 } else
942 tunnel->err_count = 0;
943 }
944
945 /*
946 * Okay, now see if we can stuff it in the buffer as-is.
947 */
948 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
949
950 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
951 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
952 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
953 if (!new_skb) {
954 ip_rt_put(rt);
955 dev->stats.tx_dropped++;
956 kfree_skb(skb);
957 return NETDEV_TX_OK;
958 }
959 if (skb->sk)
960 skb_set_owner_w(new_skb, skb->sk);
961 dev_kfree_skb(skb);
962 skb = new_skb;
963 iph6 = ipv6_hdr(skb);
964 }
965 ttl = tiph->ttl;
966 if (ttl == 0)
967 ttl = iph6->hop_limit;
968 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
969
970 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
971 if (IS_ERR(skb)) {
972 ip_rt_put(rt);
973 goto out;
974 }
975
976 err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr,
977 IPPROTO_IPV6, tos, ttl, df,
978 !net_eq(tunnel->net, dev_net(dev)));
979 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
980 return NETDEV_TX_OK;
981
982 tx_error_icmp:
983 dst_link_failure(skb);
984 tx_error:
985 kfree_skb(skb);
986 out:
987 dev->stats.tx_errors++;
988 return NETDEV_TX_OK;
989 }
990
991 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
992 {
993 struct ip_tunnel *tunnel = netdev_priv(dev);
994 const struct iphdr *tiph = &tunnel->parms.iph;
995
996 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
997 if (IS_ERR(skb))
998 goto out;
999
1000 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1001 return NETDEV_TX_OK;
1002 out:
1003 dev->stats.tx_errors++;
1004 return NETDEV_TX_OK;
1005 }
1006
1007 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1008 struct net_device *dev)
1009 {
1010 switch (skb->protocol) {
1011 case htons(ETH_P_IP):
1012 ipip_tunnel_xmit(skb, dev);
1013 break;
1014 case htons(ETH_P_IPV6):
1015 ipip6_tunnel_xmit(skb, dev);
1016 break;
1017 default:
1018 goto tx_err;
1019 }
1020
1021 return NETDEV_TX_OK;
1022
1023 tx_err:
1024 dev->stats.tx_errors++;
1025 kfree_skb(skb);
1026 return NETDEV_TX_OK;
1027
1028 }
1029
1030 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1031 {
1032 struct net_device *tdev = NULL;
1033 struct ip_tunnel *tunnel;
1034 const struct iphdr *iph;
1035 struct flowi4 fl4;
1036
1037 tunnel = netdev_priv(dev);
1038 iph = &tunnel->parms.iph;
1039
1040 if (iph->daddr) {
1041 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1042 NULL,
1043 iph->daddr, iph->saddr,
1044 0, 0,
1045 IPPROTO_IPV6,
1046 RT_TOS(iph->tos),
1047 tunnel->parms.link);
1048
1049 if (!IS_ERR(rt)) {
1050 tdev = rt->dst.dev;
1051 ip_rt_put(rt);
1052 }
1053 dev->flags |= IFF_POINTOPOINT;
1054 }
1055
1056 if (!tdev && tunnel->parms.link)
1057 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1058
1059 if (tdev) {
1060 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1061 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1062 if (dev->mtu < IPV6_MIN_MTU)
1063 dev->mtu = IPV6_MIN_MTU;
1064 }
1065 dev->iflink = tunnel->parms.link;
1066 }
1067
1068 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1069 {
1070 struct net *net = t->net;
1071 struct sit_net *sitn = net_generic(net, sit_net_id);
1072
1073 ipip6_tunnel_unlink(sitn, t);
1074 synchronize_net();
1075 t->parms.iph.saddr = p->iph.saddr;
1076 t->parms.iph.daddr = p->iph.daddr;
1077 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1078 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1079 ipip6_tunnel_link(sitn, t);
1080 t->parms.iph.ttl = p->iph.ttl;
1081 t->parms.iph.tos = p->iph.tos;
1082 if (t->parms.link != p->link) {
1083 t->parms.link = p->link;
1084 ipip6_tunnel_bind_dev(t->dev);
1085 }
1086 ip_tunnel_dst_reset_all(t);
1087 netdev_state_change(t->dev);
1088 }
1089
1090 #ifdef CONFIG_IPV6_SIT_6RD
1091 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1092 struct ip_tunnel_6rd *ip6rd)
1093 {
1094 struct in6_addr prefix;
1095 __be32 relay_prefix;
1096
1097 if (ip6rd->relay_prefixlen > 32 ||
1098 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1099 return -EINVAL;
1100
1101 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1102 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1103 return -EINVAL;
1104 if (ip6rd->relay_prefixlen)
1105 relay_prefix = ip6rd->relay_prefix &
1106 htonl(0xffffffffUL <<
1107 (32 - ip6rd->relay_prefixlen));
1108 else
1109 relay_prefix = 0;
1110 if (relay_prefix != ip6rd->relay_prefix)
1111 return -EINVAL;
1112
1113 t->ip6rd.prefix = prefix;
1114 t->ip6rd.relay_prefix = relay_prefix;
1115 t->ip6rd.prefixlen = ip6rd->prefixlen;
1116 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1117 ip_tunnel_dst_reset_all(t);
1118 netdev_state_change(t->dev);
1119 return 0;
1120 }
1121 #endif
1122
1123 static int
1124 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1125 {
1126 int err = 0;
1127 struct ip_tunnel_parm p;
1128 struct ip_tunnel_prl prl;
1129 struct ip_tunnel *t = netdev_priv(dev);
1130 struct net *net = t->net;
1131 struct sit_net *sitn = net_generic(net, sit_net_id);
1132 #ifdef CONFIG_IPV6_SIT_6RD
1133 struct ip_tunnel_6rd ip6rd;
1134 #endif
1135
1136 switch (cmd) {
1137 case SIOCGETTUNNEL:
1138 #ifdef CONFIG_IPV6_SIT_6RD
1139 case SIOCGET6RD:
1140 #endif
1141 if (dev == sitn->fb_tunnel_dev) {
1142 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1143 err = -EFAULT;
1144 break;
1145 }
1146 t = ipip6_tunnel_locate(net, &p, 0);
1147 if (t == NULL)
1148 t = netdev_priv(dev);
1149 }
1150
1151 err = -EFAULT;
1152 if (cmd == SIOCGETTUNNEL) {
1153 memcpy(&p, &t->parms, sizeof(p));
1154 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1155 sizeof(p)))
1156 goto done;
1157 #ifdef CONFIG_IPV6_SIT_6RD
1158 } else {
1159 ip6rd.prefix = t->ip6rd.prefix;
1160 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1161 ip6rd.prefixlen = t->ip6rd.prefixlen;
1162 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1163 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1164 sizeof(ip6rd)))
1165 goto done;
1166 #endif
1167 }
1168 err = 0;
1169 break;
1170
1171 case SIOCADDTUNNEL:
1172 case SIOCCHGTUNNEL:
1173 err = -EPERM;
1174 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1175 goto done;
1176
1177 err = -EFAULT;
1178 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1179 goto done;
1180
1181 err = -EINVAL;
1182 if (p.iph.protocol != IPPROTO_IPV6 &&
1183 p.iph.protocol != IPPROTO_IPIP &&
1184 p.iph.protocol != 0)
1185 goto done;
1186 if (p.iph.version != 4 ||
1187 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1188 goto done;
1189 if (p.iph.ttl)
1190 p.iph.frag_off |= htons(IP_DF);
1191
1192 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1193
1194 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1195 if (t != NULL) {
1196 if (t->dev != dev) {
1197 err = -EEXIST;
1198 break;
1199 }
1200 } else {
1201 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1202 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1203 err = -EINVAL;
1204 break;
1205 }
1206 t = netdev_priv(dev);
1207 }
1208
1209 ipip6_tunnel_update(t, &p);
1210 }
1211
1212 if (t) {
1213 err = 0;
1214 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1215 err = -EFAULT;
1216 } else
1217 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1218 break;
1219
1220 case SIOCDELTUNNEL:
1221 err = -EPERM;
1222 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1223 goto done;
1224
1225 if (dev == sitn->fb_tunnel_dev) {
1226 err = -EFAULT;
1227 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1228 goto done;
1229 err = -ENOENT;
1230 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1231 goto done;
1232 err = -EPERM;
1233 if (t == netdev_priv(sitn->fb_tunnel_dev))
1234 goto done;
1235 dev = t->dev;
1236 }
1237 unregister_netdevice(dev);
1238 err = 0;
1239 break;
1240
1241 case SIOCGETPRL:
1242 err = -EINVAL;
1243 if (dev == sitn->fb_tunnel_dev)
1244 goto done;
1245 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1246 break;
1247
1248 case SIOCADDPRL:
1249 case SIOCDELPRL:
1250 case SIOCCHGPRL:
1251 err = -EPERM;
1252 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1253 goto done;
1254 err = -EINVAL;
1255 if (dev == sitn->fb_tunnel_dev)
1256 goto done;
1257 err = -EFAULT;
1258 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1259 goto done;
1260
1261 switch (cmd) {
1262 case SIOCDELPRL:
1263 err = ipip6_tunnel_del_prl(t, &prl);
1264 break;
1265 case SIOCADDPRL:
1266 case SIOCCHGPRL:
1267 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1268 break;
1269 }
1270 ip_tunnel_dst_reset_all(t);
1271 netdev_state_change(dev);
1272 break;
1273
1274 #ifdef CONFIG_IPV6_SIT_6RD
1275 case SIOCADD6RD:
1276 case SIOCCHG6RD:
1277 case SIOCDEL6RD:
1278 err = -EPERM;
1279 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1280 goto done;
1281
1282 err = -EFAULT;
1283 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1284 sizeof(ip6rd)))
1285 goto done;
1286
1287 if (cmd != SIOCDEL6RD) {
1288 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1289 if (err < 0)
1290 goto done;
1291 } else
1292 ipip6_tunnel_clone_6rd(dev, sitn);
1293
1294 err = 0;
1295 break;
1296 #endif
1297
1298 default:
1299 err = -EINVAL;
1300 }
1301
1302 done:
1303 return err;
1304 }
1305
1306 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1307 {
1308 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1309 return -EINVAL;
1310 dev->mtu = new_mtu;
1311 return 0;
1312 }
1313
1314 static const struct net_device_ops ipip6_netdev_ops = {
1315 .ndo_init = ipip6_tunnel_init,
1316 .ndo_uninit = ipip6_tunnel_uninit,
1317 .ndo_start_xmit = sit_tunnel_xmit,
1318 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1319 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1320 .ndo_get_stats64 = ip_tunnel_get_stats64,
1321 };
1322
1323 static void ipip6_dev_free(struct net_device *dev)
1324 {
1325 struct ip_tunnel *tunnel = netdev_priv(dev);
1326
1327 free_percpu(tunnel->dst_cache);
1328 free_percpu(dev->tstats);
1329 free_netdev(dev);
1330 }
1331
1332 #define SIT_FEATURES (NETIF_F_SG | \
1333 NETIF_F_FRAGLIST | \
1334 NETIF_F_HIGHDMA | \
1335 NETIF_F_GSO_SOFTWARE | \
1336 NETIF_F_HW_CSUM)
1337
1338 static void ipip6_tunnel_setup(struct net_device *dev)
1339 {
1340 dev->netdev_ops = &ipip6_netdev_ops;
1341 dev->destructor = ipip6_dev_free;
1342
1343 dev->type = ARPHRD_SIT;
1344 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1345 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1346 dev->flags = IFF_NOARP;
1347 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1348 dev->iflink = 0;
1349 dev->addr_len = 4;
1350 dev->features |= NETIF_F_LLTX;
1351 dev->features |= SIT_FEATURES;
1352 dev->hw_features |= SIT_FEATURES;
1353 }
1354
1355 static int ipip6_tunnel_init(struct net_device *dev)
1356 {
1357 struct ip_tunnel *tunnel = netdev_priv(dev);
1358
1359 tunnel->dev = dev;
1360 tunnel->net = dev_net(dev);
1361 strcpy(tunnel->parms.name, dev->name);
1362
1363 ipip6_tunnel_bind_dev(dev);
1364 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1365 if (!dev->tstats)
1366 return -ENOMEM;
1367
1368 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1369 if (!tunnel->dst_cache) {
1370 free_percpu(dev->tstats);
1371 dev->tstats = NULL;
1372 return -ENOMEM;
1373 }
1374
1375 return 0;
1376 }
1377
1378 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1379 {
1380 struct ip_tunnel *tunnel = netdev_priv(dev);
1381 struct iphdr *iph = &tunnel->parms.iph;
1382 struct net *net = dev_net(dev);
1383 struct sit_net *sitn = net_generic(net, sit_net_id);
1384
1385 iph->version = 4;
1386 iph->protocol = IPPROTO_IPV6;
1387 iph->ihl = 5;
1388 iph->ttl = 64;
1389
1390 dev_hold(dev);
1391 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1392 }
1393
1394 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1395 {
1396 u8 proto;
1397
1398 if (!data || !data[IFLA_IPTUN_PROTO])
1399 return 0;
1400
1401 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1402 if (proto != IPPROTO_IPV6 &&
1403 proto != IPPROTO_IPIP &&
1404 proto != 0)
1405 return -EINVAL;
1406
1407 return 0;
1408 }
1409
1410 static void ipip6_netlink_parms(struct nlattr *data[],
1411 struct ip_tunnel_parm *parms)
1412 {
1413 memset(parms, 0, sizeof(*parms));
1414
1415 parms->iph.version = 4;
1416 parms->iph.protocol = IPPROTO_IPV6;
1417 parms->iph.ihl = 5;
1418 parms->iph.ttl = 64;
1419
1420 if (!data)
1421 return;
1422
1423 if (data[IFLA_IPTUN_LINK])
1424 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1425
1426 if (data[IFLA_IPTUN_LOCAL])
1427 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1428
1429 if (data[IFLA_IPTUN_REMOTE])
1430 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1431
1432 if (data[IFLA_IPTUN_TTL]) {
1433 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1434 if (parms->iph.ttl)
1435 parms->iph.frag_off = htons(IP_DF);
1436 }
1437
1438 if (data[IFLA_IPTUN_TOS])
1439 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1440
1441 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1442 parms->iph.frag_off = htons(IP_DF);
1443
1444 if (data[IFLA_IPTUN_FLAGS])
1445 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1446
1447 if (data[IFLA_IPTUN_PROTO])
1448 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1449
1450 }
1451
1452 #ifdef CONFIG_IPV6_SIT_6RD
1453 /* This function returns true when 6RD attributes are present in the nl msg */
1454 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1455 struct ip_tunnel_6rd *ip6rd)
1456 {
1457 bool ret = false;
1458 memset(ip6rd, 0, sizeof(*ip6rd));
1459
1460 if (!data)
1461 return ret;
1462
1463 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1464 ret = true;
1465 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1466 sizeof(struct in6_addr));
1467 }
1468
1469 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1470 ret = true;
1471 ip6rd->relay_prefix =
1472 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1473 }
1474
1475 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1476 ret = true;
1477 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1478 }
1479
1480 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1481 ret = true;
1482 ip6rd->relay_prefixlen =
1483 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1484 }
1485
1486 return ret;
1487 }
1488 #endif
1489
1490 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1491 struct nlattr *tb[], struct nlattr *data[])
1492 {
1493 struct net *net = dev_net(dev);
1494 struct ip_tunnel *nt;
1495 #ifdef CONFIG_IPV6_SIT_6RD
1496 struct ip_tunnel_6rd ip6rd;
1497 #endif
1498 int err;
1499
1500 nt = netdev_priv(dev);
1501 ipip6_netlink_parms(data, &nt->parms);
1502
1503 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1504 return -EEXIST;
1505
1506 err = ipip6_tunnel_create(dev);
1507 if (err < 0)
1508 return err;
1509
1510 #ifdef CONFIG_IPV6_SIT_6RD
1511 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1512 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1513 #endif
1514
1515 return err;
1516 }
1517
1518 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1519 struct nlattr *data[])
1520 {
1521 struct ip_tunnel *t = netdev_priv(dev);
1522 struct ip_tunnel_parm p;
1523 struct net *net = t->net;
1524 struct sit_net *sitn = net_generic(net, sit_net_id);
1525 #ifdef CONFIG_IPV6_SIT_6RD
1526 struct ip_tunnel_6rd ip6rd;
1527 #endif
1528
1529 if (dev == sitn->fb_tunnel_dev)
1530 return -EINVAL;
1531
1532 ipip6_netlink_parms(data, &p);
1533
1534 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1535 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1536 return -EINVAL;
1537
1538 t = ipip6_tunnel_locate(net, &p, 0);
1539
1540 if (t) {
1541 if (t->dev != dev)
1542 return -EEXIST;
1543 } else
1544 t = netdev_priv(dev);
1545
1546 ipip6_tunnel_update(t, &p);
1547
1548 #ifdef CONFIG_IPV6_SIT_6RD
1549 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1550 return ipip6_tunnel_update_6rd(t, &ip6rd);
1551 #endif
1552
1553 return 0;
1554 }
1555
1556 static size_t ipip6_get_size(const struct net_device *dev)
1557 {
1558 return
1559 /* IFLA_IPTUN_LINK */
1560 nla_total_size(4) +
1561 /* IFLA_IPTUN_LOCAL */
1562 nla_total_size(4) +
1563 /* IFLA_IPTUN_REMOTE */
1564 nla_total_size(4) +
1565 /* IFLA_IPTUN_TTL */
1566 nla_total_size(1) +
1567 /* IFLA_IPTUN_TOS */
1568 nla_total_size(1) +
1569 /* IFLA_IPTUN_PMTUDISC */
1570 nla_total_size(1) +
1571 /* IFLA_IPTUN_FLAGS */
1572 nla_total_size(2) +
1573 /* IFLA_IPTUN_PROTO */
1574 nla_total_size(1) +
1575 #ifdef CONFIG_IPV6_SIT_6RD
1576 /* IFLA_IPTUN_6RD_PREFIX */
1577 nla_total_size(sizeof(struct in6_addr)) +
1578 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1579 nla_total_size(4) +
1580 /* IFLA_IPTUN_6RD_PREFIXLEN */
1581 nla_total_size(2) +
1582 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1583 nla_total_size(2) +
1584 #endif
1585 0;
1586 }
1587
1588 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1589 {
1590 struct ip_tunnel *tunnel = netdev_priv(dev);
1591 struct ip_tunnel_parm *parm = &tunnel->parms;
1592
1593 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1594 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1595 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1596 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1597 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1598 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1599 !!(parm->iph.frag_off & htons(IP_DF))) ||
1600 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1601 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1602 goto nla_put_failure;
1603
1604 #ifdef CONFIG_IPV6_SIT_6RD
1605 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1606 &tunnel->ip6rd.prefix) ||
1607 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1608 tunnel->ip6rd.relay_prefix) ||
1609 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1610 tunnel->ip6rd.prefixlen) ||
1611 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1612 tunnel->ip6rd.relay_prefixlen))
1613 goto nla_put_failure;
1614 #endif
1615
1616 return 0;
1617
1618 nla_put_failure:
1619 return -EMSGSIZE;
1620 }
1621
1622 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1623 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1624 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1625 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1626 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1627 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1628 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1629 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1630 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1631 #ifdef CONFIG_IPV6_SIT_6RD
1632 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1633 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1634 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1635 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1636 #endif
1637 };
1638
1639 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1640 {
1641 struct net *net = dev_net(dev);
1642 struct sit_net *sitn = net_generic(net, sit_net_id);
1643
1644 if (dev != sitn->fb_tunnel_dev)
1645 unregister_netdevice_queue(dev, head);
1646 }
1647
1648 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1649 .kind = "sit",
1650 .maxtype = IFLA_IPTUN_MAX,
1651 .policy = ipip6_policy,
1652 .priv_size = sizeof(struct ip_tunnel),
1653 .setup = ipip6_tunnel_setup,
1654 .validate = ipip6_validate,
1655 .newlink = ipip6_newlink,
1656 .changelink = ipip6_changelink,
1657 .get_size = ipip6_get_size,
1658 .fill_info = ipip6_fill_info,
1659 .dellink = ipip6_dellink,
1660 };
1661
1662 static struct xfrm_tunnel sit_handler __read_mostly = {
1663 .handler = ipip6_rcv,
1664 .err_handler = ipip6_err,
1665 .priority = 1,
1666 };
1667
1668 static struct xfrm_tunnel ipip_handler __read_mostly = {
1669 .handler = ipip_rcv,
1670 .err_handler = ipip6_err,
1671 .priority = 2,
1672 };
1673
1674 static void __net_exit sit_destroy_tunnels(struct net *net,
1675 struct list_head *head)
1676 {
1677 struct sit_net *sitn = net_generic(net, sit_net_id);
1678 struct net_device *dev, *aux;
1679 int prio;
1680
1681 for_each_netdev_safe(net, dev, aux)
1682 if (dev->rtnl_link_ops == &sit_link_ops)
1683 unregister_netdevice_queue(dev, head);
1684
1685 for (prio = 1; prio < 4; prio++) {
1686 int h;
1687 for (h = 0; h < HASH_SIZE; h++) {
1688 struct ip_tunnel *t;
1689
1690 t = rtnl_dereference(sitn->tunnels[prio][h]);
1691 while (t != NULL) {
1692 /* If dev is in the same netns, it has already
1693 * been added to the list by the previous loop.
1694 */
1695 if (!net_eq(dev_net(t->dev), net))
1696 unregister_netdevice_queue(t->dev,
1697 head);
1698 t = rtnl_dereference(t->next);
1699 }
1700 }
1701 }
1702 }
1703
1704 static int __net_init sit_init_net(struct net *net)
1705 {
1706 struct sit_net *sitn = net_generic(net, sit_net_id);
1707 struct ip_tunnel *t;
1708 int err;
1709
1710 sitn->tunnels[0] = sitn->tunnels_wc;
1711 sitn->tunnels[1] = sitn->tunnels_l;
1712 sitn->tunnels[2] = sitn->tunnels_r;
1713 sitn->tunnels[3] = sitn->tunnels_r_l;
1714
1715 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1716 ipip6_tunnel_setup);
1717 if (!sitn->fb_tunnel_dev) {
1718 err = -ENOMEM;
1719 goto err_alloc_dev;
1720 }
1721 dev_net_set(sitn->fb_tunnel_dev, net);
1722 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1723 /* FB netdevice is special: we have one, and only one per netns.
1724 * Allowing to move it to another netns is clearly unsafe.
1725 */
1726 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1727
1728 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1729 goto err_reg_dev;
1730
1731 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1732 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1733
1734 t = netdev_priv(sitn->fb_tunnel_dev);
1735
1736 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1737 return 0;
1738
1739 err_reg_dev:
1740 ipip6_dev_free(sitn->fb_tunnel_dev);
1741 err_alloc_dev:
1742 return err;
1743 }
1744
1745 static void __net_exit sit_exit_net(struct net *net)
1746 {
1747 LIST_HEAD(list);
1748
1749 rtnl_lock();
1750 sit_destroy_tunnels(net, &list);
1751 unregister_netdevice_many(&list);
1752 rtnl_unlock();
1753 }
1754
1755 static struct pernet_operations sit_net_ops = {
1756 .init = sit_init_net,
1757 .exit = sit_exit_net,
1758 .id = &sit_net_id,
1759 .size = sizeof(struct sit_net),
1760 };
1761
1762 static void __exit sit_cleanup(void)
1763 {
1764 rtnl_link_unregister(&sit_link_ops);
1765 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1766 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1767
1768 unregister_pernet_device(&sit_net_ops);
1769 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1770 }
1771
1772 static int __init sit_init(void)
1773 {
1774 int err;
1775
1776 pr_info("IPv6 over IPv4 tunneling driver\n");
1777
1778 err = register_pernet_device(&sit_net_ops);
1779 if (err < 0)
1780 return err;
1781 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1782 if (err < 0) {
1783 pr_info("%s: can't register ip6ip4\n", __func__);
1784 goto xfrm_tunnel_failed;
1785 }
1786 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1787 if (err < 0) {
1788 pr_info("%s: can't register ip4ip4\n", __func__);
1789 goto xfrm_tunnel4_failed;
1790 }
1791 err = rtnl_link_register(&sit_link_ops);
1792 if (err < 0)
1793 goto rtnl_link_failed;
1794
1795 out:
1796 return err;
1797
1798 rtnl_link_failed:
1799 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1800 xfrm_tunnel4_failed:
1801 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1802 xfrm_tunnel_failed:
1803 unregister_pernet_device(&sit_net_ops);
1804 goto out;
1805 }
1806
1807 module_init(sit_init);
1808 module_exit(sit_cleanup);
1809 MODULE_LICENSE("GPL");
1810 MODULE_ALIAS_RTNL_LINK("sit");
1811 MODULE_ALIAS_NETDEV("sit0");
Linux with added FPC-III support