projects / GitHub / exynos8895 / android_kernel_samsung_universal8895.git / blob
? search:


b7ea5eaa4fd165fa38d00fcbf2dee254c0f23bb3
[GitHub/exynos8895/android_kernel_samsung_universal8895.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/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 && (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, NET_NAME_UNKNOWN,
253 ipip6_tunnel_setup);
254 if (!dev)
255 return NULL;
256
257 dev_net_set(dev, net);
258
259 nt = netdev_priv(dev);
260
261 nt->parms = *parms;
262 if (ipip6_tunnel_create(dev) < 0)
263 goto failed_free;
264
265 return nt;
266
267 failed_free:
268 ipip6_dev_free(dev);
269 failed:
270 return NULL;
271 }
272
273 #define for_each_prl_rcu(start) \
274 for (prl = rcu_dereference(start); \
275 prl; \
276 prl = rcu_dereference(prl->next))
277
278 static struct ip_tunnel_prl_entry *
279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
280 {
281 struct ip_tunnel_prl_entry *prl;
282
283 for_each_prl_rcu(t->prl)
284 if (prl->addr == addr)
285 break;
286 return prl;
287
288 }
289
290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
291 struct ip_tunnel_prl __user *a)
292 {
293 struct ip_tunnel_prl kprl, *kp;
294 struct ip_tunnel_prl_entry *prl;
295 unsigned int cmax, c = 0, ca, len;
296 int ret = 0;
297
298 if (copy_from_user(&kprl, a, sizeof(kprl)))
299 return -EFAULT;
300 cmax = kprl.datalen / sizeof(kprl);
301 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
302 cmax = 1;
303
304 /* For simple GET or for root users,
305 * we try harder to allocate.
306 */
307 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
308 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
309 NULL;
310
311 rcu_read_lock();
312
313 ca = t->prl_count < cmax ? t->prl_count : cmax;
314
315 if (!kp) {
316 /* We don't try hard to allocate much memory for
317 * non-root users.
318 * For root users, retry allocating enough memory for
319 * the answer.
320 */
321 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
322 if (!kp) {
323 ret = -ENOMEM;
324 goto out;
325 }
326 }
327
328 c = 0;
329 for_each_prl_rcu(t->prl) {
330 if (c >= cmax)
331 break;
332 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
333 continue;
334 kp[c].addr = prl->addr;
335 kp[c].flags = prl->flags;
336 c++;
337 if (kprl.addr != htonl(INADDR_ANY))
338 break;
339 }
340 out:
341 rcu_read_unlock();
342
343 len = sizeof(*kp) * c;
344 ret = 0;
345 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
346 ret = -EFAULT;
347
348 kfree(kp);
349
350 return ret;
351 }
352
353 static int
354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
355 {
356 struct ip_tunnel_prl_entry *p;
357 int err = 0;
358
359 if (a->addr == htonl(INADDR_ANY))
360 return -EINVAL;
361
362 ASSERT_RTNL();
363
364 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
365 if (p->addr == a->addr) {
366 if (chg) {
367 p->flags = a->flags;
368 goto out;
369 }
370 err = -EEXIST;
371 goto out;
372 }
373 }
374
375 if (chg) {
376 err = -ENXIO;
377 goto out;
378 }
379
380 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
381 if (!p) {
382 err = -ENOBUFS;
383 goto out;
384 }
385
386 p->next = t->prl;
387 p->addr = a->addr;
388 p->flags = a->flags;
389 t->prl_count++;
390 rcu_assign_pointer(t->prl, p);
391 out:
392 return err;
393 }
394
395 static void prl_list_destroy_rcu(struct rcu_head *head)
396 {
397 struct ip_tunnel_prl_entry *p, *n;
398
399 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
400 do {
401 n = rcu_dereference_protected(p->next, 1);
402 kfree(p);
403 p = n;
404 } while (p);
405 }
406
407 static int
408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
409 {
410 struct ip_tunnel_prl_entry *x;
411 struct ip_tunnel_prl_entry __rcu **p;
412 int err = 0;
413
414 ASSERT_RTNL();
415
416 if (a && a->addr != htonl(INADDR_ANY)) {
417 for (p = &t->prl;
418 (x = rtnl_dereference(*p)) != NULL;
419 p = &x->next) {
420 if (x->addr == a->addr) {
421 *p = x->next;
422 kfree_rcu(x, rcu_head);
423 t->prl_count--;
424 goto out;
425 }
426 }
427 err = -ENXIO;
428 } else {
429 x = rtnl_dereference(t->prl);
430 if (x) {
431 t->prl_count = 0;
432 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
433 t->prl = NULL;
434 }
435 }
436 out:
437 return err;
438 }
439
440 static int
441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
442 {
443 struct ip_tunnel_prl_entry *p;
444 int ok = 1;
445
446 rcu_read_lock();
447 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
448 if (p) {
449 if (p->flags & PRL_DEFAULT)
450 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
451 else
452 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
453 } else {
454 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
455
456 if (ipv6_addr_is_isatap(addr6) &&
457 (addr6->s6_addr32[3] == iph->saddr) &&
458 ipv6_chk_prefix(addr6, t->dev))
459 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
460 else
461 ok = 0;
462 }
463 rcu_read_unlock();
464 return ok;
465 }
466
467 static void ipip6_tunnel_uninit(struct net_device *dev)
468 {
469 struct ip_tunnel *tunnel = netdev_priv(dev);
470 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
471
472 if (dev == sitn->fb_tunnel_dev) {
473 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
474 } else {
475 ipip6_tunnel_unlink(sitn, tunnel);
476 ipip6_tunnel_del_prl(tunnel, NULL);
477 }
478 ip_tunnel_dst_reset_all(tunnel);
479 dev_put(dev);
480 }
481
482 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
483 * if sufficient data bytes are available
484 */
485 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
486 {
487 int ihl = ((const struct iphdr *)skb->data)->ihl*4;
488 struct rt6_info *rt;
489 struct sk_buff *skb2;
490
491 if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8))
492 return 1;
493
494 skb2 = skb_clone(skb, GFP_ATOMIC);
495
496 if (!skb2)
497 return 1;
498
499 skb_dst_drop(skb2);
500 skb_pull(skb2, ihl);
501 skb_reset_network_header(skb2);
502
503 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
504
505 if (rt && rt->dst.dev)
506 skb2->dev = rt->dst.dev;
507
508 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
509
510 if (rt)
511 ip6_rt_put(rt);
512
513 kfree_skb(skb2);
514
515 return 0;
516 }
517
518 static int ipip6_err(struct sk_buff *skb, u32 info)
519 {
520 const struct iphdr *iph = (const struct iphdr *)skb->data;
521 const int type = icmp_hdr(skb)->type;
522 const int code = icmp_hdr(skb)->code;
523 struct ip_tunnel *t;
524 int err;
525
526 switch (type) {
527 default:
528 case ICMP_PARAMETERPROB:
529 return 0;
530
531 case ICMP_DEST_UNREACH:
532 switch (code) {
533 case ICMP_SR_FAILED:
534 /* Impossible event. */
535 return 0;
536 default:
537 /* All others are translated to HOST_UNREACH.
538 rfc2003 contains "deep thoughts" about NET_UNREACH,
539 I believe they are just ether pollution. --ANK
540 */
541 break;
542 }
543 break;
544 case ICMP_TIME_EXCEEDED:
545 if (code != ICMP_EXC_TTL)
546 return 0;
547 break;
548 case ICMP_REDIRECT:
549 break;
550 }
551
552 err = -ENOENT;
553
554 t = ipip6_tunnel_lookup(dev_net(skb->dev),
555 skb->dev,
556 iph->daddr,
557 iph->saddr);
558 if (!t)
559 goto out;
560
561 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
562 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
563 t->parms.link, 0, iph->protocol, 0);
564 err = 0;
565 goto out;
566 }
567 if (type == ICMP_REDIRECT) {
568 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
569 iph->protocol, 0);
570 err = 0;
571 goto out;
572 }
573
574 if (t->parms.iph.daddr == 0)
575 goto out;
576
577 err = 0;
578 if (!ipip6_err_gen_icmpv6_unreach(skb))
579 goto out;
580
581 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
582 goto out;
583
584 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
585 t->err_count++;
586 else
587 t->err_count = 1;
588 t->err_time = jiffies;
589 out:
590 return err;
591 }
592
593 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
594 const struct in6_addr *v6addr)
595 {
596 __be32 v4embed = 0;
597 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
598 return true;
599 return false;
600 }
601
602 /* Checks if an address matches an address on the tunnel interface.
603 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
604 * Long story:
605 * This function is called after we considered the packet as spoofed
606 * in is_spoofed_6rd.
607 * We may have a router that is doing NAT for proto 41 packets
608 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
609 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
610 * function will return true, dropping the packet.
611 * But, we can still check if is spoofed against the IP
612 * addresses associated with the interface.
613 */
614 static bool only_dnatted(const struct ip_tunnel *tunnel,
615 const struct in6_addr *v6dst)
616 {
617 int prefix_len;
618
619 #ifdef CONFIG_IPV6_SIT_6RD
620 prefix_len = tunnel->ip6rd.prefixlen + 32
621 - tunnel->ip6rd.relay_prefixlen;
622 #else
623 prefix_len = 48;
624 #endif
625 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
626 }
627
628 /* Returns true if a packet is spoofed */
629 static bool packet_is_spoofed(struct sk_buff *skb,
630 const struct iphdr *iph,
631 struct ip_tunnel *tunnel)
632 {
633 const struct ipv6hdr *ipv6h;
634
635 if (tunnel->dev->priv_flags & IFF_ISATAP) {
636 if (!isatap_chksrc(skb, iph, tunnel))
637 return true;
638
639 return false;
640 }
641
642 if (tunnel->dev->flags & IFF_POINTOPOINT)
643 return false;
644
645 ipv6h = ipv6_hdr(skb);
646
647 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
648 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
649 &iph->saddr, &ipv6h->saddr,
650 &iph->daddr, &ipv6h->daddr);
651 return true;
652 }
653
654 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
655 return false;
656
657 if (only_dnatted(tunnel, &ipv6h->daddr))
658 return false;
659
660 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661 &iph->saddr, &ipv6h->saddr,
662 &iph->daddr, &ipv6h->daddr);
663 return true;
664 }
665
666 static int ipip6_rcv(struct sk_buff *skb)
667 {
668 const struct iphdr *iph = ip_hdr(skb);
669 struct ip_tunnel *tunnel;
670 int err;
671
672 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
673 iph->saddr, iph->daddr);
674 if (tunnel) {
675 struct pcpu_sw_netstats *tstats;
676
677 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
678 tunnel->parms.iph.protocol != 0)
679 goto out;
680
681 skb->mac_header = skb->network_header;
682 skb_reset_network_header(skb);
683 IPCB(skb)->flags = 0;
684 skb->dev = tunnel->dev;
685
686 if (packet_is_spoofed(skb, iph, tunnel)) {
687 tunnel->dev->stats.rx_errors++;
688 goto out;
689 }
690
691 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6)))
692 goto out;
693
694 err = IP_ECN_decapsulate(iph, skb);
695 if (unlikely(err)) {
696 if (log_ecn_error)
697 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
698 &iph->saddr, iph->tos);
699 if (err > 1) {
700 ++tunnel->dev->stats.rx_frame_errors;
701 ++tunnel->dev->stats.rx_errors;
702 goto out;
703 }
704 }
705
706 tstats = this_cpu_ptr(tunnel->dev->tstats);
707 u64_stats_update_begin(&tstats->syncp);
708 tstats->rx_packets++;
709 tstats->rx_bytes += skb->len;
710 u64_stats_update_end(&tstats->syncp);
711
712 netif_rx(skb);
713
714 return 0;
715 }
716
717 /* no tunnel matched, let upstream know, ipsec may handle it */
718 return 1;
719 out:
720 kfree_skb(skb);
721 return 0;
722 }
723
724 static const struct tnl_ptk_info tpi = {
725 /* no tunnel info required for ipip. */
726 .proto = htons(ETH_P_IP),
727 };
728
729 static int ipip_rcv(struct sk_buff *skb)
730 {
731 const struct iphdr *iph;
732 struct ip_tunnel *tunnel;
733
734 iph = ip_hdr(skb);
735 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
736 iph->saddr, iph->daddr);
737 if (tunnel) {
738 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
739 tunnel->parms.iph.protocol != 0)
740 goto drop;
741
742 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
743 goto drop;
744 if (iptunnel_pull_header(skb, 0, tpi.proto))
745 goto drop;
746 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
747 }
748
749 return 1;
750
751 drop:
752 kfree_skb(skb);
753 return 0;
754 }
755
756 /*
757 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
758 * stores the embedded IPv4 address in v4dst and returns true.
759 */
760 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
761 __be32 *v4dst)
762 {
763 #ifdef CONFIG_IPV6_SIT_6RD
764 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
765 tunnel->ip6rd.prefixlen)) {
766 unsigned int pbw0, pbi0;
767 int pbi1;
768 u32 d;
769
770 pbw0 = tunnel->ip6rd.prefixlen >> 5;
771 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
772
773 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
774 tunnel->ip6rd.relay_prefixlen;
775
776 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
777 if (pbi1 > 0)
778 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
779 (32 - pbi1);
780
781 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
782 return true;
783 }
784 #else
785 if (v6dst->s6_addr16[0] == htons(0x2002)) {
786 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
787 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
788 return true;
789 }
790 #endif
791 return false;
792 }
793
794 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
795 const struct in6_addr *v6dst)
796 {
797 __be32 dst = 0;
798 check_6rd(tunnel, v6dst, &dst);
799 return dst;
800 }
801
802 /*
803 * This function assumes it is being called from dev_queue_xmit()
804 * and that skb is filled properly by that function.
805 */
806
807 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
808 struct net_device *dev)
809 {
810 struct ip_tunnel *tunnel = netdev_priv(dev);
811 const struct iphdr *tiph = &tunnel->parms.iph;
812 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
813 u8 tos = tunnel->parms.iph.tos;
814 __be16 df = tiph->frag_off;
815 struct rtable *rt; /* Route to the other host */
816 struct net_device *tdev; /* Device to other host */
817 unsigned int max_headroom; /* The extra header space needed */
818 __be32 dst = tiph->daddr;
819 struct flowi4 fl4;
820 int mtu;
821 const struct in6_addr *addr6;
822 int addr_type;
823 u8 ttl;
824 int err;
825 u8 protocol = IPPROTO_IPV6;
826 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
827
828 if (skb->protocol != htons(ETH_P_IPV6))
829 goto tx_error;
830
831 if (tos == 1)
832 tos = ipv6_get_dsfield(iph6);
833
834 /* ISATAP (RFC4214) - must come before 6to4 */
835 if (dev->priv_flags & IFF_ISATAP) {
836 struct neighbour *neigh = NULL;
837 bool do_tx_error = false;
838
839 if (skb_dst(skb))
840 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
841
842 if (!neigh) {
843 net_dbg_ratelimited("nexthop == NULL\n");
844 goto tx_error;
845 }
846
847 addr6 = (const struct in6_addr *)&neigh->primary_key;
848 addr_type = ipv6_addr_type(addr6);
849
850 if ((addr_type & IPV6_ADDR_UNICAST) &&
851 ipv6_addr_is_isatap(addr6))
852 dst = addr6->s6_addr32[3];
853 else
854 do_tx_error = true;
855
856 neigh_release(neigh);
857 if (do_tx_error)
858 goto tx_error;
859 }
860
861 if (!dst)
862 dst = try_6rd(tunnel, &iph6->daddr);
863
864 if (!dst) {
865 struct neighbour *neigh = NULL;
866 bool do_tx_error = false;
867
868 if (skb_dst(skb))
869 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
870
871 if (!neigh) {
872 net_dbg_ratelimited("nexthop == NULL\n");
873 goto tx_error;
874 }
875
876 addr6 = (const struct in6_addr *)&neigh->primary_key;
877 addr_type = ipv6_addr_type(addr6);
878
879 if (addr_type == IPV6_ADDR_ANY) {
880 addr6 = &ipv6_hdr(skb)->daddr;
881 addr_type = ipv6_addr_type(addr6);
882 }
883
884 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
885 dst = addr6->s6_addr32[3];
886 else
887 do_tx_error = true;
888
889 neigh_release(neigh);
890 if (do_tx_error)
891 goto tx_error;
892 }
893
894 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
895 dst, tiph->saddr,
896 0, 0,
897 IPPROTO_IPV6, RT_TOS(tos),
898 tunnel->parms.link);
899 if (IS_ERR(rt)) {
900 dev->stats.tx_carrier_errors++;
901 goto tx_error_icmp;
902 }
903 if (rt->rt_type != RTN_UNICAST) {
904 ip_rt_put(rt);
905 dev->stats.tx_carrier_errors++;
906 goto tx_error_icmp;
907 }
908 tdev = rt->dst.dev;
909
910 if (tdev == dev) {
911 ip_rt_put(rt);
912 dev->stats.collisions++;
913 goto tx_error;
914 }
915
916 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
917 if (IS_ERR(skb)) {
918 ip_rt_put(rt);
919 goto out;
920 }
921
922 if (df) {
923 mtu = dst_mtu(&rt->dst) - t_hlen;
924
925 if (mtu < 68) {
926 dev->stats.collisions++;
927 ip_rt_put(rt);
928 goto tx_error;
929 }
930
931 if (mtu < IPV6_MIN_MTU) {
932 mtu = IPV6_MIN_MTU;
933 df = 0;
934 }
935
936 if (tunnel->parms.iph.daddr && skb_dst(skb))
937 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
938
939 if (skb->len > mtu && !skb_is_gso(skb)) {
940 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
941 ip_rt_put(rt);
942 goto tx_error;
943 }
944 }
945
946 if (tunnel->err_count > 0) {
947 if (time_before(jiffies,
948 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
949 tunnel->err_count--;
950 dst_link_failure(skb);
951 } else
952 tunnel->err_count = 0;
953 }
954
955 /*
956 * Okay, now see if we can stuff it in the buffer as-is.
957 */
958 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
959
960 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
961 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
962 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
963 if (!new_skb) {
964 ip_rt_put(rt);
965 dev->stats.tx_dropped++;
966 kfree_skb(skb);
967 return NETDEV_TX_OK;
968 }
969 if (skb->sk)
970 skb_set_owner_w(new_skb, skb->sk);
971 dev_kfree_skb(skb);
972 skb = new_skb;
973 iph6 = ipv6_hdr(skb);
974 }
975 ttl = tiph->ttl;
976 if (ttl == 0)
977 ttl = iph6->hop_limit;
978 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
979
980 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
981 ip_rt_put(rt);
982 goto tx_error;
983 }
984
985 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
986
987 err = iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr,
988 protocol, tos, ttl, df,
989 !net_eq(tunnel->net, dev_net(dev)));
990 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
991 return NETDEV_TX_OK;
992
993 tx_error_icmp:
994 dst_link_failure(skb);
995 tx_error:
996 kfree_skb(skb);
997 out:
998 dev->stats.tx_errors++;
999 return NETDEV_TX_OK;
1000 }
1001
1002 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
1003 {
1004 struct ip_tunnel *tunnel = netdev_priv(dev);
1005 const struct iphdr *tiph = &tunnel->parms.iph;
1006
1007 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
1008 if (IS_ERR(skb))
1009 goto out;
1010
1011 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1012
1013 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1014 return NETDEV_TX_OK;
1015 out:
1016 dev->stats.tx_errors++;
1017 return NETDEV_TX_OK;
1018 }
1019
1020 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1021 struct net_device *dev)
1022 {
1023 switch (skb->protocol) {
1024 case htons(ETH_P_IP):
1025 ipip_tunnel_xmit(skb, dev);
1026 break;
1027 case htons(ETH_P_IPV6):
1028 ipip6_tunnel_xmit(skb, dev);
1029 break;
1030 default:
1031 goto tx_err;
1032 }
1033
1034 return NETDEV_TX_OK;
1035
1036 tx_err:
1037 dev->stats.tx_errors++;
1038 kfree_skb(skb);
1039 return NETDEV_TX_OK;
1040
1041 }
1042
1043 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1044 {
1045 struct net_device *tdev = NULL;
1046 struct ip_tunnel *tunnel;
1047 const struct iphdr *iph;
1048 struct flowi4 fl4;
1049
1050 tunnel = netdev_priv(dev);
1051 iph = &tunnel->parms.iph;
1052
1053 if (iph->daddr) {
1054 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1055 NULL,
1056 iph->daddr, iph->saddr,
1057 0, 0,
1058 IPPROTO_IPV6,
1059 RT_TOS(iph->tos),
1060 tunnel->parms.link);
1061
1062 if (!IS_ERR(rt)) {
1063 tdev = rt->dst.dev;
1064 ip_rt_put(rt);
1065 }
1066 dev->flags |= IFF_POINTOPOINT;
1067 }
1068
1069 if (!tdev && tunnel->parms.link)
1070 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1071
1072 if (tdev) {
1073 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1074
1075 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1076 dev->mtu = tdev->mtu - t_hlen;
1077 if (dev->mtu < IPV6_MIN_MTU)
1078 dev->mtu = IPV6_MIN_MTU;
1079 }
1080 }
1081
1082 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1083 {
1084 struct net *net = t->net;
1085 struct sit_net *sitn = net_generic(net, sit_net_id);
1086
1087 ipip6_tunnel_unlink(sitn, t);
1088 synchronize_net();
1089 t->parms.iph.saddr = p->iph.saddr;
1090 t->parms.iph.daddr = p->iph.daddr;
1091 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1092 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1093 ipip6_tunnel_link(sitn, t);
1094 t->parms.iph.ttl = p->iph.ttl;
1095 t->parms.iph.tos = p->iph.tos;
1096 t->parms.iph.frag_off = p->iph.frag_off;
1097 if (t->parms.link != p->link) {
1098 t->parms.link = p->link;
1099 ipip6_tunnel_bind_dev(t->dev);
1100 }
1101 ip_tunnel_dst_reset_all(t);
1102 netdev_state_change(t->dev);
1103 }
1104
1105 #ifdef CONFIG_IPV6_SIT_6RD
1106 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1107 struct ip_tunnel_6rd *ip6rd)
1108 {
1109 struct in6_addr prefix;
1110 __be32 relay_prefix;
1111
1112 if (ip6rd->relay_prefixlen > 32 ||
1113 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1114 return -EINVAL;
1115
1116 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1117 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1118 return -EINVAL;
1119 if (ip6rd->relay_prefixlen)
1120 relay_prefix = ip6rd->relay_prefix &
1121 htonl(0xffffffffUL <<
1122 (32 - ip6rd->relay_prefixlen));
1123 else
1124 relay_prefix = 0;
1125 if (relay_prefix != ip6rd->relay_prefix)
1126 return -EINVAL;
1127
1128 t->ip6rd.prefix = prefix;
1129 t->ip6rd.relay_prefix = relay_prefix;
1130 t->ip6rd.prefixlen = ip6rd->prefixlen;
1131 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1132 ip_tunnel_dst_reset_all(t);
1133 netdev_state_change(t->dev);
1134 return 0;
1135 }
1136 #endif
1137
1138 static int
1139 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1140 {
1141 int err = 0;
1142 struct ip_tunnel_parm p;
1143 struct ip_tunnel_prl prl;
1144 struct ip_tunnel *t = netdev_priv(dev);
1145 struct net *net = t->net;
1146 struct sit_net *sitn = net_generic(net, sit_net_id);
1147 #ifdef CONFIG_IPV6_SIT_6RD
1148 struct ip_tunnel_6rd ip6rd;
1149 #endif
1150
1151 switch (cmd) {
1152 case SIOCGETTUNNEL:
1153 #ifdef CONFIG_IPV6_SIT_6RD
1154 case SIOCGET6RD:
1155 #endif
1156 if (dev == sitn->fb_tunnel_dev) {
1157 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1158 err = -EFAULT;
1159 break;
1160 }
1161 t = ipip6_tunnel_locate(net, &p, 0);
1162 if (!t)
1163 t = netdev_priv(dev);
1164 }
1165
1166 err = -EFAULT;
1167 if (cmd == SIOCGETTUNNEL) {
1168 memcpy(&p, &t->parms, sizeof(p));
1169 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1170 sizeof(p)))
1171 goto done;
1172 #ifdef CONFIG_IPV6_SIT_6RD
1173 } else {
1174 ip6rd.prefix = t->ip6rd.prefix;
1175 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1176 ip6rd.prefixlen = t->ip6rd.prefixlen;
1177 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1178 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1179 sizeof(ip6rd)))
1180 goto done;
1181 #endif
1182 }
1183 err = 0;
1184 break;
1185
1186 case SIOCADDTUNNEL:
1187 case SIOCCHGTUNNEL:
1188 err = -EPERM;
1189 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1190 goto done;
1191
1192 err = -EFAULT;
1193 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1194 goto done;
1195
1196 err = -EINVAL;
1197 if (p.iph.protocol != IPPROTO_IPV6 &&
1198 p.iph.protocol != IPPROTO_IPIP &&
1199 p.iph.protocol != 0)
1200 goto done;
1201 if (p.iph.version != 4 ||
1202 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1203 goto done;
1204 if (p.iph.ttl)
1205 p.iph.frag_off |= htons(IP_DF);
1206
1207 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1208
1209 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1210 if (t) {
1211 if (t->dev != dev) {
1212 err = -EEXIST;
1213 break;
1214 }
1215 } else {
1216 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1217 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1218 err = -EINVAL;
1219 break;
1220 }
1221 t = netdev_priv(dev);
1222 }
1223
1224 ipip6_tunnel_update(t, &p);
1225 }
1226
1227 if (t) {
1228 err = 0;
1229 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1230 err = -EFAULT;
1231 } else
1232 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1233 break;
1234
1235 case SIOCDELTUNNEL:
1236 err = -EPERM;
1237 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1238 goto done;
1239
1240 if (dev == sitn->fb_tunnel_dev) {
1241 err = -EFAULT;
1242 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1243 goto done;
1244 err = -ENOENT;
1245 t = ipip6_tunnel_locate(net, &p, 0);
1246 if (!t)
1247 goto done;
1248 err = -EPERM;
1249 if (t == netdev_priv(sitn->fb_tunnel_dev))
1250 goto done;
1251 dev = t->dev;
1252 }
1253 unregister_netdevice(dev);
1254 err = 0;
1255 break;
1256
1257 case SIOCGETPRL:
1258 err = -EINVAL;
1259 if (dev == sitn->fb_tunnel_dev)
1260 goto done;
1261 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1262 break;
1263
1264 case SIOCADDPRL:
1265 case SIOCDELPRL:
1266 case SIOCCHGPRL:
1267 err = -EPERM;
1268 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1269 goto done;
1270 err = -EINVAL;
1271 if (dev == sitn->fb_tunnel_dev)
1272 goto done;
1273 err = -EFAULT;
1274 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1275 goto done;
1276
1277 switch (cmd) {
1278 case SIOCDELPRL:
1279 err = ipip6_tunnel_del_prl(t, &prl);
1280 break;
1281 case SIOCADDPRL:
1282 case SIOCCHGPRL:
1283 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1284 break;
1285 }
1286 ip_tunnel_dst_reset_all(t);
1287 netdev_state_change(dev);
1288 break;
1289
1290 #ifdef CONFIG_IPV6_SIT_6RD
1291 case SIOCADD6RD:
1292 case SIOCCHG6RD:
1293 case SIOCDEL6RD:
1294 err = -EPERM;
1295 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1296 goto done;
1297
1298 err = -EFAULT;
1299 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1300 sizeof(ip6rd)))
1301 goto done;
1302
1303 if (cmd != SIOCDEL6RD) {
1304 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1305 if (err < 0)
1306 goto done;
1307 } else
1308 ipip6_tunnel_clone_6rd(dev, sitn);
1309
1310 err = 0;
1311 break;
1312 #endif
1313
1314 default:
1315 err = -EINVAL;
1316 }
1317
1318 done:
1319 return err;
1320 }
1321
1322 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1323 {
1324 struct ip_tunnel *tunnel = netdev_priv(dev);
1325 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1326
1327 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen)
1328 return -EINVAL;
1329 dev->mtu = new_mtu;
1330 return 0;
1331 }
1332
1333 static const struct net_device_ops ipip6_netdev_ops = {
1334 .ndo_init = ipip6_tunnel_init,
1335 .ndo_uninit = ipip6_tunnel_uninit,
1336 .ndo_start_xmit = sit_tunnel_xmit,
1337 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1338 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1339 .ndo_get_stats64 = ip_tunnel_get_stats64,
1340 .ndo_get_iflink = ip_tunnel_get_iflink,
1341 };
1342
1343 static void ipip6_dev_free(struct net_device *dev)
1344 {
1345 struct ip_tunnel *tunnel = netdev_priv(dev);
1346
1347 free_percpu(tunnel->dst_cache);
1348 free_percpu(dev->tstats);
1349 free_netdev(dev);
1350 }
1351
1352 #define SIT_FEATURES (NETIF_F_SG | \
1353 NETIF_F_FRAGLIST | \
1354 NETIF_F_HIGHDMA | \
1355 NETIF_F_GSO_SOFTWARE | \
1356 NETIF_F_HW_CSUM)
1357
1358 static void ipip6_tunnel_setup(struct net_device *dev)
1359 {
1360 struct ip_tunnel *tunnel = netdev_priv(dev);
1361 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1362
1363 dev->netdev_ops = &ipip6_netdev_ops;
1364 dev->destructor = ipip6_dev_free;
1365
1366 dev->type = ARPHRD_SIT;
1367 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1368 dev->mtu = ETH_DATA_LEN - t_hlen;
1369 dev->flags = IFF_NOARP;
1370 netif_keep_dst(dev);
1371 dev->addr_len = 4;
1372 dev->features |= NETIF_F_LLTX;
1373 dev->features |= SIT_FEATURES;
1374 dev->hw_features |= SIT_FEATURES;
1375 }
1376
1377 static int ipip6_tunnel_init(struct net_device *dev)
1378 {
1379 struct ip_tunnel *tunnel = netdev_priv(dev);
1380
1381 tunnel->dev = dev;
1382 tunnel->net = dev_net(dev);
1383 strcpy(tunnel->parms.name, dev->name);
1384
1385 ipip6_tunnel_bind_dev(dev);
1386 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1387 if (!dev->tstats)
1388 return -ENOMEM;
1389
1390 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1391 if (!tunnel->dst_cache) {
1392 free_percpu(dev->tstats);
1393 dev->tstats = NULL;
1394 return -ENOMEM;
1395 }
1396
1397 return 0;
1398 }
1399
1400 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1401 {
1402 struct ip_tunnel *tunnel = netdev_priv(dev);
1403 struct iphdr *iph = &tunnel->parms.iph;
1404 struct net *net = dev_net(dev);
1405 struct sit_net *sitn = net_generic(net, sit_net_id);
1406
1407 iph->version = 4;
1408 iph->protocol = IPPROTO_IPV6;
1409 iph->ihl = 5;
1410 iph->ttl = 64;
1411
1412 dev_hold(dev);
1413 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1414 }
1415
1416 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1417 {
1418 u8 proto;
1419
1420 if (!data || !data[IFLA_IPTUN_PROTO])
1421 return 0;
1422
1423 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1424 if (proto != IPPROTO_IPV6 &&
1425 proto != IPPROTO_IPIP &&
1426 proto != 0)
1427 return -EINVAL;
1428
1429 return 0;
1430 }
1431
1432 static void ipip6_netlink_parms(struct nlattr *data[],
1433 struct ip_tunnel_parm *parms)
1434 {
1435 memset(parms, 0, sizeof(*parms));
1436
1437 parms->iph.version = 4;
1438 parms->iph.protocol = IPPROTO_IPV6;
1439 parms->iph.ihl = 5;
1440 parms->iph.ttl = 64;
1441
1442 if (!data)
1443 return;
1444
1445 if (data[IFLA_IPTUN_LINK])
1446 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1447
1448 if (data[IFLA_IPTUN_LOCAL])
1449 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1450
1451 if (data[IFLA_IPTUN_REMOTE])
1452 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1453
1454 if (data[IFLA_IPTUN_TTL]) {
1455 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1456 if (parms->iph.ttl)
1457 parms->iph.frag_off = htons(IP_DF);
1458 }
1459
1460 if (data[IFLA_IPTUN_TOS])
1461 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1462
1463 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1464 parms->iph.frag_off = htons(IP_DF);
1465
1466 if (data[IFLA_IPTUN_FLAGS])
1467 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1468
1469 if (data[IFLA_IPTUN_PROTO])
1470 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1471
1472 }
1473
1474 /* This function returns true when ENCAP attributes are present in the nl msg */
1475 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1476 struct ip_tunnel_encap *ipencap)
1477 {
1478 bool ret = false;
1479
1480 memset(ipencap, 0, sizeof(*ipencap));
1481
1482 if (!data)
1483 return ret;
1484
1485 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1486 ret = true;
1487 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1488 }
1489
1490 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1491 ret = true;
1492 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1493 }
1494
1495 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1496 ret = true;
1497 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1498 }
1499
1500 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1501 ret = true;
1502 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1503 }
1504
1505 return ret;
1506 }
1507
1508 #ifdef CONFIG_IPV6_SIT_6RD
1509 /* This function returns true when 6RD attributes are present in the nl msg */
1510 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1511 struct ip_tunnel_6rd *ip6rd)
1512 {
1513 bool ret = false;
1514 memset(ip6rd, 0, sizeof(*ip6rd));
1515
1516 if (!data)
1517 return ret;
1518
1519 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1520 ret = true;
1521 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1522 }
1523
1524 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1525 ret = true;
1526 ip6rd->relay_prefix =
1527 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1528 }
1529
1530 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1531 ret = true;
1532 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1533 }
1534
1535 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1536 ret = true;
1537 ip6rd->relay_prefixlen =
1538 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1539 }
1540
1541 return ret;
1542 }
1543 #endif
1544
1545 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1546 struct nlattr *tb[], struct nlattr *data[])
1547 {
1548 struct net *net = dev_net(dev);
1549 struct ip_tunnel *nt;
1550 struct ip_tunnel_encap ipencap;
1551 #ifdef CONFIG_IPV6_SIT_6RD
1552 struct ip_tunnel_6rd ip6rd;
1553 #endif
1554 int err;
1555
1556 nt = netdev_priv(dev);
1557
1558 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1559 err = ip_tunnel_encap_setup(nt, &ipencap);
1560 if (err < 0)
1561 return err;
1562 }
1563
1564 ipip6_netlink_parms(data, &nt->parms);
1565
1566 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1567 return -EEXIST;
1568
1569 err = ipip6_tunnel_create(dev);
1570 if (err < 0)
1571 return err;
1572
1573 #ifdef CONFIG_IPV6_SIT_6RD
1574 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1575 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1576 #endif
1577
1578 return err;
1579 }
1580
1581 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1582 struct nlattr *data[])
1583 {
1584 struct ip_tunnel *t = netdev_priv(dev);
1585 struct ip_tunnel_parm p;
1586 struct ip_tunnel_encap ipencap;
1587 struct net *net = t->net;
1588 struct sit_net *sitn = net_generic(net, sit_net_id);
1589 #ifdef CONFIG_IPV6_SIT_6RD
1590 struct ip_tunnel_6rd ip6rd;
1591 #endif
1592 int err;
1593
1594 if (dev == sitn->fb_tunnel_dev)
1595 return -EINVAL;
1596
1597 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1598 err = ip_tunnel_encap_setup(t, &ipencap);
1599 if (err < 0)
1600 return err;
1601 }
1602
1603 ipip6_netlink_parms(data, &p);
1604
1605 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1606 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1607 return -EINVAL;
1608
1609 t = ipip6_tunnel_locate(net, &p, 0);
1610
1611 if (t) {
1612 if (t->dev != dev)
1613 return -EEXIST;
1614 } else
1615 t = netdev_priv(dev);
1616
1617 ipip6_tunnel_update(t, &p);
1618
1619 #ifdef CONFIG_IPV6_SIT_6RD
1620 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1621 return ipip6_tunnel_update_6rd(t, &ip6rd);
1622 #endif
1623
1624 return 0;
1625 }
1626
1627 static size_t ipip6_get_size(const struct net_device *dev)
1628 {
1629 return
1630 /* IFLA_IPTUN_LINK */
1631 nla_total_size(4) +
1632 /* IFLA_IPTUN_LOCAL */
1633 nla_total_size(4) +
1634 /* IFLA_IPTUN_REMOTE */
1635 nla_total_size(4) +
1636 /* IFLA_IPTUN_TTL */
1637 nla_total_size(1) +
1638 /* IFLA_IPTUN_TOS */
1639 nla_total_size(1) +
1640 /* IFLA_IPTUN_PMTUDISC */
1641 nla_total_size(1) +
1642 /* IFLA_IPTUN_FLAGS */
1643 nla_total_size(2) +
1644 /* IFLA_IPTUN_PROTO */
1645 nla_total_size(1) +
1646 #ifdef CONFIG_IPV6_SIT_6RD
1647 /* IFLA_IPTUN_6RD_PREFIX */
1648 nla_total_size(sizeof(struct in6_addr)) +
1649 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1650 nla_total_size(4) +
1651 /* IFLA_IPTUN_6RD_PREFIXLEN */
1652 nla_total_size(2) +
1653 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1654 nla_total_size(2) +
1655 #endif
1656 /* IFLA_IPTUN_ENCAP_TYPE */
1657 nla_total_size(2) +
1658 /* IFLA_IPTUN_ENCAP_FLAGS */
1659 nla_total_size(2) +
1660 /* IFLA_IPTUN_ENCAP_SPORT */
1661 nla_total_size(2) +
1662 /* IFLA_IPTUN_ENCAP_DPORT */
1663 nla_total_size(2) +
1664 0;
1665 }
1666
1667 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1668 {
1669 struct ip_tunnel *tunnel = netdev_priv(dev);
1670 struct ip_tunnel_parm *parm = &tunnel->parms;
1671
1672 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1673 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1674 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1675 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1676 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1677 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1678 !!(parm->iph.frag_off & htons(IP_DF))) ||
1679 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1680 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1681 goto nla_put_failure;
1682
1683 #ifdef CONFIG_IPV6_SIT_6RD
1684 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1685 &tunnel->ip6rd.prefix) ||
1686 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1687 tunnel->ip6rd.relay_prefix) ||
1688 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1689 tunnel->ip6rd.prefixlen) ||
1690 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1691 tunnel->ip6rd.relay_prefixlen))
1692 goto nla_put_failure;
1693 #endif
1694
1695 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1696 tunnel->encap.type) ||
1697 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1698 tunnel->encap.sport) ||
1699 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1700 tunnel->encap.dport) ||
1701 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1702 tunnel->encap.flags))
1703 goto nla_put_failure;
1704
1705 return 0;
1706
1707 nla_put_failure:
1708 return -EMSGSIZE;
1709 }
1710
1711 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1712 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1713 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1714 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1715 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1716 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1717 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1718 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1719 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1720 #ifdef CONFIG_IPV6_SIT_6RD
1721 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1722 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1723 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1724 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1725 #endif
1726 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1727 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1728 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1729 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1730 };
1731
1732 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1733 {
1734 struct net *net = dev_net(dev);
1735 struct sit_net *sitn = net_generic(net, sit_net_id);
1736
1737 if (dev != sitn->fb_tunnel_dev)
1738 unregister_netdevice_queue(dev, head);
1739 }
1740
1741 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1742 .kind = "sit",
1743 .maxtype = IFLA_IPTUN_MAX,
1744 .policy = ipip6_policy,
1745 .priv_size = sizeof(struct ip_tunnel),
1746 .setup = ipip6_tunnel_setup,
1747 .validate = ipip6_validate,
1748 .newlink = ipip6_newlink,
1749 .changelink = ipip6_changelink,
1750 .get_size = ipip6_get_size,
1751 .fill_info = ipip6_fill_info,
1752 .dellink = ipip6_dellink,
1753 .get_link_net = ip_tunnel_get_link_net,
1754 };
1755
1756 static struct xfrm_tunnel sit_handler __read_mostly = {
1757 .handler = ipip6_rcv,
1758 .err_handler = ipip6_err,
1759 .priority = 1,
1760 };
1761
1762 static struct xfrm_tunnel ipip_handler __read_mostly = {
1763 .handler = ipip_rcv,
1764 .err_handler = ipip6_err,
1765 .priority = 2,
1766 };
1767
1768 static void __net_exit sit_destroy_tunnels(struct net *net,
1769 struct list_head *head)
1770 {
1771 struct sit_net *sitn = net_generic(net, sit_net_id);
1772 struct net_device *dev, *aux;
1773 int prio;
1774
1775 for_each_netdev_safe(net, dev, aux)
1776 if (dev->rtnl_link_ops == &sit_link_ops)
1777 unregister_netdevice_queue(dev, head);
1778
1779 for (prio = 1; prio < 4; prio++) {
1780 int h;
1781 for (h = 0; h < HASH_SIZE; h++) {
1782 struct ip_tunnel *t;
1783
1784 t = rtnl_dereference(sitn->tunnels[prio][h]);
1785 while (t) {
1786 /* If dev is in the same netns, it has already
1787 * been added to the list by the previous loop.
1788 */
1789 if (!net_eq(dev_net(t->dev), net))
1790 unregister_netdevice_queue(t->dev,
1791 head);
1792 t = rtnl_dereference(t->next);
1793 }
1794 }
1795 }
1796 }
1797
1798 static int __net_init sit_init_net(struct net *net)
1799 {
1800 struct sit_net *sitn = net_generic(net, sit_net_id);
1801 struct ip_tunnel *t;
1802 int err;
1803
1804 sitn->tunnels[0] = sitn->tunnels_wc;
1805 sitn->tunnels[1] = sitn->tunnels_l;
1806 sitn->tunnels[2] = sitn->tunnels_r;
1807 sitn->tunnels[3] = sitn->tunnels_r_l;
1808
1809 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1810 NET_NAME_UNKNOWN,
1811 ipip6_tunnel_setup);
1812 if (!sitn->fb_tunnel_dev) {
1813 err = -ENOMEM;
1814 goto err_alloc_dev;
1815 }
1816 dev_net_set(sitn->fb_tunnel_dev, net);
1817 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1818 /* FB netdevice is special: we have one, and only one per netns.
1819 * Allowing to move it to another netns is clearly unsafe.
1820 */
1821 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1822
1823 err = register_netdev(sitn->fb_tunnel_dev);
1824 if (err)
1825 goto err_reg_dev;
1826
1827 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1828 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1829
1830 t = netdev_priv(sitn->fb_tunnel_dev);
1831
1832 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1833 return 0;
1834
1835 err_reg_dev:
1836 ipip6_dev_free(sitn->fb_tunnel_dev);
1837 err_alloc_dev:
1838 return err;
1839 }
1840
1841 static void __net_exit sit_exit_net(struct net *net)
1842 {
1843 LIST_HEAD(list);
1844
1845 rtnl_lock();
1846 sit_destroy_tunnels(net, &list);
1847 unregister_netdevice_many(&list);
1848 rtnl_unlock();
1849 }
1850
1851 static struct pernet_operations sit_net_ops = {
1852 .init = sit_init_net,
1853 .exit = sit_exit_net,
1854 .id = &sit_net_id,
1855 .size = sizeof(struct sit_net),
1856 };
1857
1858 static void __exit sit_cleanup(void)
1859 {
1860 rtnl_link_unregister(&sit_link_ops);
1861 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1862 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1863
1864 unregister_pernet_device(&sit_net_ops);
1865 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1866 }
1867
1868 static int __init sit_init(void)
1869 {
1870 int err;
1871
1872 pr_info("IPv6 over IPv4 tunneling driver\n");
1873
1874 err = register_pernet_device(&sit_net_ops);
1875 if (err < 0)
1876 return err;
1877 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1878 if (err < 0) {
1879 pr_info("%s: can't register ip6ip4\n", __func__);
1880 goto xfrm_tunnel_failed;
1881 }
1882 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1883 if (err < 0) {
1884 pr_info("%s: can't register ip4ip4\n", __func__);
1885 goto xfrm_tunnel4_failed;
1886 }
1887 err = rtnl_link_register(&sit_link_ops);
1888 if (err < 0)
1889 goto rtnl_link_failed;
1890
1891 out:
1892 return err;
1893
1894 rtnl_link_failed:
1895 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1896 xfrm_tunnel4_failed:
1897 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1898 xfrm_tunnel_failed:
1899 unregister_pernet_device(&sit_net_ops);
1900 goto out;
1901 }
1902
1903 module_init(sit_init);
1904 module_exit(sit_cleanup);
1905 MODULE_LICENSE("GPL");
1906 MODULE_ALIAS_RTNL_LINK("sit");
1907 MODULE_ALIAS_NETDEV("sit0");