?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ALSA: usb-audio: Fix an out-of-bound read in create_composite_quirks
[linux/fpc-iii.git] / net / ipv6 / sit.c
blob11282ffca567a2cbff07f7cea1dc1660bda07c5a
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 (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 if (!dev_valid_name(parms->name))
249 goto failed;
250 strlcpy(name, parms->name, IFNAMSIZ);
251 } else {
252 strcpy(name, "sit%d");
253 }
254 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
255 ipip6_tunnel_setup);
256 if (!dev)
257 return NULL;
258
259 dev_net_set(dev, net);
260
261 nt = netdev_priv(dev);
262
263 nt->parms = *parms;
264 if (ipip6_tunnel_create(dev) < 0)
265 goto failed_free;
266
267 return nt;
268
269 failed_free:
270 ipip6_dev_free(dev);
271 failed:
272 return NULL;
273 }
274
275 #define for_each_prl_rcu(start) \
276 for (prl = rcu_dereference(start); \
277 prl; \
278 prl = rcu_dereference(prl->next))
279
280 static struct ip_tunnel_prl_entry *
281 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
282 {
283 struct ip_tunnel_prl_entry *prl;
284
285 for_each_prl_rcu(t->prl)
286 if (prl->addr == addr)
287 break;
288 return prl;
289
290 }
291
292 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
293 struct ip_tunnel_prl __user *a)
294 {
295 struct ip_tunnel_prl kprl, *kp;
296 struct ip_tunnel_prl_entry *prl;
297 unsigned int cmax, c = 0, ca, len;
298 int ret = 0;
299
300 if (copy_from_user(&kprl, a, sizeof(kprl)))
301 return -EFAULT;
302 cmax = kprl.datalen / sizeof(kprl);
303 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
304 cmax = 1;
305
306 /* For simple GET or for root users,
307 * we try harder to allocate.
308 */
309 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
310 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
311 NULL;
312
313 rcu_read_lock();
314
315 ca = t->prl_count < cmax ? t->prl_count : cmax;
316
317 if (!kp) {
318 /* We don't try hard to allocate much memory for
319 * non-root users.
320 * For root users, retry allocating enough memory for
321 * the answer.
322 */
323 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
324 if (!kp) {
325 ret = -ENOMEM;
326 goto out;
327 }
328 }
329
330 c = 0;
331 for_each_prl_rcu(t->prl) {
332 if (c >= cmax)
333 break;
334 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
335 continue;
336 kp[c].addr = prl->addr;
337 kp[c].flags = prl->flags;
338 c++;
339 if (kprl.addr != htonl(INADDR_ANY))
340 break;
341 }
342 out:
343 rcu_read_unlock();
344
345 len = sizeof(*kp) * c;
346 ret = 0;
347 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
348 ret = -EFAULT;
349
350 kfree(kp);
351
352 return ret;
353 }
354
355 static int
356 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
357 {
358 struct ip_tunnel_prl_entry *p;
359 int err = 0;
360
361 if (a->addr == htonl(INADDR_ANY))
362 return -EINVAL;
363
364 ASSERT_RTNL();
365
366 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
367 if (p->addr == a->addr) {
368 if (chg) {
369 p->flags = a->flags;
370 goto out;
371 }
372 err = -EEXIST;
373 goto out;
374 }
375 }
376
377 if (chg) {
378 err = -ENXIO;
379 goto out;
380 }
381
382 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
383 if (!p) {
384 err = -ENOBUFS;
385 goto out;
386 }
387
388 p->next = t->prl;
389 p->addr = a->addr;
390 p->flags = a->flags;
391 t->prl_count++;
392 rcu_assign_pointer(t->prl, p);
393 out:
394 return err;
395 }
396
397 static void prl_list_destroy_rcu(struct rcu_head *head)
398 {
399 struct ip_tunnel_prl_entry *p, *n;
400
401 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
402 do {
403 n = rcu_dereference_protected(p->next, 1);
404 kfree(p);
405 p = n;
406 } while (p);
407 }
408
409 static int
410 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
411 {
412 struct ip_tunnel_prl_entry *x;
413 struct ip_tunnel_prl_entry __rcu **p;
414 int err = 0;
415
416 ASSERT_RTNL();
417
418 if (a && a->addr != htonl(INADDR_ANY)) {
419 for (p = &t->prl;
420 (x = rtnl_dereference(*p)) != NULL;
421 p = &x->next) {
422 if (x->addr == a->addr) {
423 *p = x->next;
424 kfree_rcu(x, rcu_head);
425 t->prl_count--;
426 goto out;
427 }
428 }
429 err = -ENXIO;
430 } else {
431 x = rtnl_dereference(t->prl);
432 if (x) {
433 t->prl_count = 0;
434 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
435 t->prl = NULL;
436 }
437 }
438 out:
439 return err;
440 }
441
442 static int
443 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
444 {
445 struct ip_tunnel_prl_entry *p;
446 int ok = 1;
447
448 rcu_read_lock();
449 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
450 if (p) {
451 if (p->flags & PRL_DEFAULT)
452 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
453 else
454 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
455 } else {
456 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
457
458 if (ipv6_addr_is_isatap(addr6) &&
459 (addr6->s6_addr32[3] == iph->saddr) &&
460 ipv6_chk_prefix(addr6, t->dev))
461 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
462 else
463 ok = 0;
464 }
465 rcu_read_unlock();
466 return ok;
467 }
468
469 static void ipip6_tunnel_uninit(struct net_device *dev)
470 {
471 struct ip_tunnel *tunnel = netdev_priv(dev);
472 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
473
474 if (dev == sitn->fb_tunnel_dev) {
475 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
476 } else {
477 ipip6_tunnel_unlink(sitn, tunnel);
478 ipip6_tunnel_del_prl(tunnel, NULL);
479 }
480 dst_cache_reset(&tunnel->dst_cache);
481 dev_put(dev);
482 }
483
484 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
485 * if sufficient data bytes are available
486 */
487 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
488 {
489 int ihl = ((const struct iphdr *)skb->data)->ihl*4;
490 struct rt6_info *rt;
491 struct sk_buff *skb2;
492
493 if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8))
494 return 1;
495
496 skb2 = skb_clone(skb, GFP_ATOMIC);
497
498 if (!skb2)
499 return 1;
500
501 skb_dst_drop(skb2);
502 skb_pull(skb2, ihl);
503 skb_reset_network_header(skb2);
504
505 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
506
507 if (rt && rt->dst.dev)
508 skb2->dev = rt->dst.dev;
509
510 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
511
512 if (rt)
513 ip6_rt_put(rt);
514
515 kfree_skb(skb2);
516
517 return 0;
518 }
519
520 static int ipip6_err(struct sk_buff *skb, u32 info)
521 {
522 const struct iphdr *iph = (const struct iphdr *)skb->data;
523 const int type = icmp_hdr(skb)->type;
524 const int code = icmp_hdr(skb)->code;
525 struct ip_tunnel *t;
526 int err;
527
528 switch (type) {
529 default:
530 case ICMP_PARAMETERPROB:
531 return 0;
532
533 case ICMP_DEST_UNREACH:
534 switch (code) {
535 case ICMP_SR_FAILED:
536 /* Impossible event. */
537 return 0;
538 default:
539 /* All others are translated to HOST_UNREACH.
540 rfc2003 contains "deep thoughts" about NET_UNREACH,
541 I believe they are just ether pollution. --ANK
542 */
543 break;
544 }
545 break;
546 case ICMP_TIME_EXCEEDED:
547 if (code != ICMP_EXC_TTL)
548 return 0;
549 break;
550 case ICMP_REDIRECT:
551 break;
552 }
553
554 err = -ENOENT;
555
556 t = ipip6_tunnel_lookup(dev_net(skb->dev),
557 skb->dev,
558 iph->daddr,
559 iph->saddr);
560 if (!t)
561 goto out;
562
563 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
564 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
565 t->parms.link, 0, iph->protocol, 0);
566 err = 0;
567 goto out;
568 }
569 if (type == ICMP_REDIRECT) {
570 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
571 iph->protocol, 0);
572 err = 0;
573 goto out;
574 }
575
576 if (t->parms.iph.daddr == 0)
577 goto out;
578
579 err = 0;
580 if (!ipip6_err_gen_icmpv6_unreach(skb))
581 goto out;
582
583 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
584 goto out;
585
586 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
587 t->err_count++;
588 else
589 t->err_count = 1;
590 t->err_time = jiffies;
591 out:
592 return err;
593 }
594
595 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
596 const struct in6_addr *v6addr)
597 {
598 __be32 v4embed = 0;
599 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
600 return true;
601 return false;
602 }
603
604 /* Checks if an address matches an address on the tunnel interface.
605 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
606 * Long story:
607 * This function is called after we considered the packet as spoofed
608 * in is_spoofed_6rd.
609 * We may have a router that is doing NAT for proto 41 packets
610 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
611 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
612 * function will return true, dropping the packet.
613 * But, we can still check if is spoofed against the IP
614 * addresses associated with the interface.
615 */
616 static bool only_dnatted(const struct ip_tunnel *tunnel,
617 const struct in6_addr *v6dst)
618 {
619 int prefix_len;
620
621 #ifdef CONFIG_IPV6_SIT_6RD
622 prefix_len = tunnel->ip6rd.prefixlen + 32
623 - tunnel->ip6rd.relay_prefixlen;
624 #else
625 prefix_len = 48;
626 #endif
627 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
628 }
629
630 /* Returns true if a packet is spoofed */
631 static bool packet_is_spoofed(struct sk_buff *skb,
632 const struct iphdr *iph,
633 struct ip_tunnel *tunnel)
634 {
635 const struct ipv6hdr *ipv6h;
636
637 if (tunnel->dev->priv_flags & IFF_ISATAP) {
638 if (!isatap_chksrc(skb, iph, tunnel))
639 return true;
640
641 return false;
642 }
643
644 if (tunnel->dev->flags & IFF_POINTOPOINT)
645 return false;
646
647 ipv6h = ipv6_hdr(skb);
648
649 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
650 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
651 &iph->saddr, &ipv6h->saddr,
652 &iph->daddr, &ipv6h->daddr);
653 return true;
654 }
655
656 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
657 return false;
658
659 if (only_dnatted(tunnel, &ipv6h->daddr))
660 return false;
661
662 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
663 &iph->saddr, &ipv6h->saddr,
664 &iph->daddr, &ipv6h->daddr);
665 return true;
666 }
667
668 static int ipip6_rcv(struct sk_buff *skb)
669 {
670 const struct iphdr *iph = ip_hdr(skb);
671 struct ip_tunnel *tunnel;
672 int err;
673
674 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
675 iph->saddr, iph->daddr);
676 if (tunnel) {
677 struct pcpu_sw_netstats *tstats;
678
679 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
680 tunnel->parms.iph.protocol != 0)
681 goto out;
682
683 skb->mac_header = skb->network_header;
684 skb_reset_network_header(skb);
685 IPCB(skb)->flags = 0;
686 skb->dev = tunnel->dev;
687
688 if (packet_is_spoofed(skb, iph, tunnel)) {
689 tunnel->dev->stats.rx_errors++;
690 goto out;
691 }
692
693 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6)))
694 goto out;
695
696 err = IP_ECN_decapsulate(iph, skb);
697 if (unlikely(err)) {
698 if (log_ecn_error)
699 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
700 &iph->saddr, iph->tos);
701 if (err > 1) {
702 ++tunnel->dev->stats.rx_frame_errors;
703 ++tunnel->dev->stats.rx_errors;
704 goto out;
705 }
706 }
707
708 tstats = this_cpu_ptr(tunnel->dev->tstats);
709 u64_stats_update_begin(&tstats->syncp);
710 tstats->rx_packets++;
711 tstats->rx_bytes += skb->len;
712 u64_stats_update_end(&tstats->syncp);
713
714 netif_rx(skb);
715
716 return 0;
717 }
718
719 /* no tunnel matched, let upstream know, ipsec may handle it */
720 return 1;
721 out:
722 kfree_skb(skb);
723 return 0;
724 }
725
726 static const struct tnl_ptk_info tpi = {
727 /* no tunnel info required for ipip. */
728 .proto = htons(ETH_P_IP),
729 };
730
731 static int ipip_rcv(struct sk_buff *skb)
732 {
733 const struct iphdr *iph;
734 struct ip_tunnel *tunnel;
735
736 iph = ip_hdr(skb);
737 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
738 iph->saddr, iph->daddr);
739 if (tunnel) {
740 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
741 tunnel->parms.iph.protocol != 0)
742 goto drop;
743
744 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
745 goto drop;
746 if (iptunnel_pull_header(skb, 0, tpi.proto))
747 goto drop;
748 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
749 }
750
751 return 1;
752
753 drop:
754 kfree_skb(skb);
755 return 0;
756 }
757
758 /*
759 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
760 * stores the embedded IPv4 address in v4dst and returns true.
761 */
762 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
763 __be32 *v4dst)
764 {
765 #ifdef CONFIG_IPV6_SIT_6RD
766 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
767 tunnel->ip6rd.prefixlen)) {
768 unsigned int pbw0, pbi0;
769 int pbi1;
770 u32 d;
771
772 pbw0 = tunnel->ip6rd.prefixlen >> 5;
773 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
774
775 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
776 tunnel->ip6rd.relay_prefixlen;
777
778 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
779 if (pbi1 > 0)
780 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
781 (32 - pbi1);
782
783 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
784 return true;
785 }
786 #else
787 if (v6dst->s6_addr16[0] == htons(0x2002)) {
788 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
789 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
790 return true;
791 }
792 #endif
793 return false;
794 }
795
796 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
797 const struct in6_addr *v6dst)
798 {
799 __be32 dst = 0;
800 check_6rd(tunnel, v6dst, &dst);
801 return dst;
802 }
803
804 /*
805 * This function assumes it is being called from dev_queue_xmit()
806 * and that skb is filled properly by that function.
807 */
808
809 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
810 struct net_device *dev)
811 {
812 struct ip_tunnel *tunnel = netdev_priv(dev);
813 const struct iphdr *tiph = &tunnel->parms.iph;
814 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
815 u8 tos = tunnel->parms.iph.tos;
816 __be16 df = tiph->frag_off;
817 struct rtable *rt; /* Route to the other host */
818 struct net_device *tdev; /* Device to other host */
819 unsigned int max_headroom; /* The extra header space needed */
820 __be32 dst = tiph->daddr;
821 struct flowi4 fl4;
822 int mtu;
823 const struct in6_addr *addr6;
824 int addr_type;
825 u8 ttl;
826 int err;
827 u8 protocol = IPPROTO_IPV6;
828 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
829
830 if (skb->protocol != htons(ETH_P_IPV6))
831 goto tx_error;
832
833 if (tos == 1)
834 tos = ipv6_get_dsfield(iph6);
835
836 /* ISATAP (RFC4214) - must come before 6to4 */
837 if (dev->priv_flags & IFF_ISATAP) {
838 struct neighbour *neigh = NULL;
839 bool do_tx_error = false;
840
841 if (skb_dst(skb))
842 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
843
844 if (!neigh) {
845 net_dbg_ratelimited("nexthop == NULL\n");
846 goto tx_error;
847 }
848
849 addr6 = (const struct in6_addr *)&neigh->primary_key;
850 addr_type = ipv6_addr_type(addr6);
851
852 if ((addr_type & IPV6_ADDR_UNICAST) &&
853 ipv6_addr_is_isatap(addr6))
854 dst = addr6->s6_addr32[3];
855 else
856 do_tx_error = true;
857
858 neigh_release(neigh);
859 if (do_tx_error)
860 goto tx_error;
861 }
862
863 if (!dst)
864 dst = try_6rd(tunnel, &iph6->daddr);
865
866 if (!dst) {
867 struct neighbour *neigh = NULL;
868 bool do_tx_error = false;
869
870 if (skb_dst(skb))
871 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
872
873 if (!neigh) {
874 net_dbg_ratelimited("nexthop == NULL\n");
875 goto tx_error;
876 }
877
878 addr6 = (const struct in6_addr *)&neigh->primary_key;
879 addr_type = ipv6_addr_type(addr6);
880
881 if (addr_type == IPV6_ADDR_ANY) {
882 addr6 = &ipv6_hdr(skb)->daddr;
883 addr_type = ipv6_addr_type(addr6);
884 }
885
886 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
887 dst = addr6->s6_addr32[3];
888 else
889 do_tx_error = true;
890
891 neigh_release(neigh);
892 if (do_tx_error)
893 goto tx_error;
894 }
895
896 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
897 dst, tiph->saddr,
898 0, 0,
899 IPPROTO_IPV6, RT_TOS(tos),
900 tunnel->parms.link);
901 if (IS_ERR(rt)) {
902 dev->stats.tx_carrier_errors++;
903 goto tx_error_icmp;
904 }
905 if (rt->rt_type != RTN_UNICAST) {
906 ip_rt_put(rt);
907 dev->stats.tx_carrier_errors++;
908 goto tx_error_icmp;
909 }
910 tdev = rt->dst.dev;
911
912 if (tdev == dev) {
913 ip_rt_put(rt);
914 dev->stats.collisions++;
915 goto tx_error;
916 }
917
918 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
919 if (IS_ERR(skb)) {
920 ip_rt_put(rt);
921 goto out;
922 }
923
924 if (df) {
925 mtu = dst_mtu(&rt->dst) - t_hlen;
926
927 if (mtu < 68) {
928 dev->stats.collisions++;
929 ip_rt_put(rt);
930 goto tx_error;
931 }
932
933 if (mtu < IPV6_MIN_MTU) {
934 mtu = IPV6_MIN_MTU;
935 df = 0;
936 }
937
938 if (tunnel->parms.iph.daddr && skb_dst(skb))
939 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
940
941 if (skb->len > mtu && !skb_is_gso(skb)) {
942 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
943 ip_rt_put(rt);
944 goto tx_error;
945 }
946 }
947
948 if (tunnel->err_count > 0) {
949 if (time_before(jiffies,
950 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
951 tunnel->err_count--;
952 dst_link_failure(skb);
953 } else
954 tunnel->err_count = 0;
955 }
956
957 /*
958 * Okay, now see if we can stuff it in the buffer as-is.
959 */
960 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
961
962 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
963 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
964 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
965 if (!new_skb) {
966 ip_rt_put(rt);
967 dev->stats.tx_dropped++;
968 kfree_skb(skb);
969 return NETDEV_TX_OK;
970 }
971 if (skb->sk)
972 skb_set_owner_w(new_skb, skb->sk);
973 dev_kfree_skb(skb);
974 skb = new_skb;
975 iph6 = ipv6_hdr(skb);
976 }
977 ttl = tiph->ttl;
978 if (ttl == 0)
979 ttl = iph6->hop_limit;
980 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
981
982 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
983 ip_rt_put(rt);
984 goto tx_error;
985 }
986
987 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
988
989 err = iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr,
990 protocol, tos, ttl, df,
991 !net_eq(tunnel->net, dev_net(dev)));
992 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
993 return NETDEV_TX_OK;
994
995 tx_error_icmp:
996 dst_link_failure(skb);
997 tx_error:
998 kfree_skb(skb);
999 out:
1000 dev->stats.tx_errors++;
1001 return NETDEV_TX_OK;
1002 }
1003
1004 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
1005 {
1006 struct ip_tunnel *tunnel = netdev_priv(dev);
1007 const struct iphdr *tiph = &tunnel->parms.iph;
1008
1009 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
1010 if (IS_ERR(skb))
1011 goto out;
1012
1013 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1014
1015 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1016 return NETDEV_TX_OK;
1017 out:
1018 dev->stats.tx_errors++;
1019 return NETDEV_TX_OK;
1020 }
1021
1022 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1023 struct net_device *dev)
1024 {
1025 switch (skb->protocol) {
1026 case htons(ETH_P_IP):
1027 ipip_tunnel_xmit(skb, dev);
1028 break;
1029 case htons(ETH_P_IPV6):
1030 ipip6_tunnel_xmit(skb, dev);
1031 break;
1032 default:
1033 goto tx_err;
1034 }
1035
1036 return NETDEV_TX_OK;
1037
1038 tx_err:
1039 dev->stats.tx_errors++;
1040 kfree_skb(skb);
1041 return NETDEV_TX_OK;
1042
1043 }
1044
1045 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1046 {
1047 struct net_device *tdev = NULL;
1048 struct ip_tunnel *tunnel;
1049 const struct iphdr *iph;
1050 struct flowi4 fl4;
1051
1052 tunnel = netdev_priv(dev);
1053 iph = &tunnel->parms.iph;
1054
1055 if (iph->daddr) {
1056 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1057 NULL,
1058 iph->daddr, iph->saddr,
1059 0, 0,
1060 IPPROTO_IPV6,
1061 RT_TOS(iph->tos),
1062 tunnel->parms.link);
1063
1064 if (!IS_ERR(rt)) {
1065 tdev = rt->dst.dev;
1066 ip_rt_put(rt);
1067 }
1068 dev->flags |= IFF_POINTOPOINT;
1069 }
1070
1071 if (!tdev && tunnel->parms.link)
1072 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1073
1074 if (tdev) {
1075 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1076
1077 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1078 dev->mtu = tdev->mtu - t_hlen;
1079 if (dev->mtu < IPV6_MIN_MTU)
1080 dev->mtu = IPV6_MIN_MTU;
1081 }
1082 }
1083
1084 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1085 {
1086 struct net *net = t->net;
1087 struct sit_net *sitn = net_generic(net, sit_net_id);
1088
1089 ipip6_tunnel_unlink(sitn, t);
1090 synchronize_net();
1091 t->parms.iph.saddr = p->iph.saddr;
1092 t->parms.iph.daddr = p->iph.daddr;
1093 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1094 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1095 ipip6_tunnel_link(sitn, t);
1096 t->parms.iph.ttl = p->iph.ttl;
1097 t->parms.iph.tos = p->iph.tos;
1098 t->parms.iph.frag_off = p->iph.frag_off;
1099 if (t->parms.link != p->link) {
1100 t->parms.link = p->link;
1101 ipip6_tunnel_bind_dev(t->dev);
1102 }
1103 dst_cache_reset(&t->dst_cache);
1104 netdev_state_change(t->dev);
1105 }
1106
1107 #ifdef CONFIG_IPV6_SIT_6RD
1108 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1109 struct ip_tunnel_6rd *ip6rd)
1110 {
1111 struct in6_addr prefix;
1112 __be32 relay_prefix;
1113
1114 if (ip6rd->relay_prefixlen > 32 ||
1115 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1116 return -EINVAL;
1117
1118 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1119 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1120 return -EINVAL;
1121 if (ip6rd->relay_prefixlen)
1122 relay_prefix = ip6rd->relay_prefix &
1123 htonl(0xffffffffUL <<
1124 (32 - ip6rd->relay_prefixlen));
1125 else
1126 relay_prefix = 0;
1127 if (relay_prefix != ip6rd->relay_prefix)
1128 return -EINVAL;
1129
1130 t->ip6rd.prefix = prefix;
1131 t->ip6rd.relay_prefix = relay_prefix;
1132 t->ip6rd.prefixlen = ip6rd->prefixlen;
1133 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1134 dst_cache_reset(&t->dst_cache);
1135 netdev_state_change(t->dev);
1136 return 0;
1137 }
1138 #endif
1139
1140 static int
1141 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1142 {
1143 int err = 0;
1144 struct ip_tunnel_parm p;
1145 struct ip_tunnel_prl prl;
1146 struct ip_tunnel *t = netdev_priv(dev);
1147 struct net *net = t->net;
1148 struct sit_net *sitn = net_generic(net, sit_net_id);
1149 #ifdef CONFIG_IPV6_SIT_6RD
1150 struct ip_tunnel_6rd ip6rd;
1151 #endif
1152
1153 switch (cmd) {
1154 case SIOCGETTUNNEL:
1155 #ifdef CONFIG_IPV6_SIT_6RD
1156 case SIOCGET6RD:
1157 #endif
1158 if (dev == sitn->fb_tunnel_dev) {
1159 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1160 err = -EFAULT;
1161 break;
1162 }
1163 t = ipip6_tunnel_locate(net, &p, 0);
1164 if (!t)
1165 t = netdev_priv(dev);
1166 }
1167
1168 err = -EFAULT;
1169 if (cmd == SIOCGETTUNNEL) {
1170 memcpy(&p, &t->parms, sizeof(p));
1171 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1172 sizeof(p)))
1173 goto done;
1174 #ifdef CONFIG_IPV6_SIT_6RD
1175 } else {
1176 ip6rd.prefix = t->ip6rd.prefix;
1177 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1178 ip6rd.prefixlen = t->ip6rd.prefixlen;
1179 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1180 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1181 sizeof(ip6rd)))
1182 goto done;
1183 #endif
1184 }
1185 err = 0;
1186 break;
1187
1188 case SIOCADDTUNNEL:
1189 case SIOCCHGTUNNEL:
1190 err = -EPERM;
1191 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1192 goto done;
1193
1194 err = -EFAULT;
1195 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1196 goto done;
1197
1198 err = -EINVAL;
1199 if (p.iph.protocol != IPPROTO_IPV6 &&
1200 p.iph.protocol != IPPROTO_IPIP &&
1201 p.iph.protocol != 0)
1202 goto done;
1203 if (p.iph.version != 4 ||
1204 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1205 goto done;
1206 if (p.iph.ttl)
1207 p.iph.frag_off |= htons(IP_DF);
1208
1209 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1210
1211 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1212 if (t) {
1213 if (t->dev != dev) {
1214 err = -EEXIST;
1215 break;
1216 }
1217 } else {
1218 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1219 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1220 err = -EINVAL;
1221 break;
1222 }
1223 t = netdev_priv(dev);
1224 }
1225
1226 ipip6_tunnel_update(t, &p);
1227 }
1228
1229 if (t) {
1230 err = 0;
1231 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1232 err = -EFAULT;
1233 } else
1234 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1235 break;
1236
1237 case SIOCDELTUNNEL:
1238 err = -EPERM;
1239 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1240 goto done;
1241
1242 if (dev == sitn->fb_tunnel_dev) {
1243 err = -EFAULT;
1244 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1245 goto done;
1246 err = -ENOENT;
1247 t = ipip6_tunnel_locate(net, &p, 0);
1248 if (!t)
1249 goto done;
1250 err = -EPERM;
1251 if (t == netdev_priv(sitn->fb_tunnel_dev))
1252 goto done;
1253 dev = t->dev;
1254 }
1255 unregister_netdevice(dev);
1256 err = 0;
1257 break;
1258
1259 case SIOCGETPRL:
1260 err = -EINVAL;
1261 if (dev == sitn->fb_tunnel_dev)
1262 goto done;
1263 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1264 break;
1265
1266 case SIOCADDPRL:
1267 case SIOCDELPRL:
1268 case SIOCCHGPRL:
1269 err = -EPERM;
1270 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1271 goto done;
1272 err = -EINVAL;
1273 if (dev == sitn->fb_tunnel_dev)
1274 goto done;
1275 err = -EFAULT;
1276 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1277 goto done;
1278
1279 switch (cmd) {
1280 case SIOCDELPRL:
1281 err = ipip6_tunnel_del_prl(t, &prl);
1282 break;
1283 case SIOCADDPRL:
1284 case SIOCCHGPRL:
1285 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1286 break;
1287 }
1288 dst_cache_reset(&t->dst_cache);
1289 netdev_state_change(dev);
1290 break;
1291
1292 #ifdef CONFIG_IPV6_SIT_6RD
1293 case SIOCADD6RD:
1294 case SIOCCHG6RD:
1295 case SIOCDEL6RD:
1296 err = -EPERM;
1297 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1298 goto done;
1299
1300 err = -EFAULT;
1301 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1302 sizeof(ip6rd)))
1303 goto done;
1304
1305 if (cmd != SIOCDEL6RD) {
1306 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1307 if (err < 0)
1308 goto done;
1309 } else
1310 ipip6_tunnel_clone_6rd(dev, sitn);
1311
1312 err = 0;
1313 break;
1314 #endif
1315
1316 default:
1317 err = -EINVAL;
1318 }
1319
1320 done:
1321 return err;
1322 }
1323
1324 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1325 {
1326 struct ip_tunnel *tunnel = netdev_priv(dev);
1327 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1328
1329 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen)
1330 return -EINVAL;
1331 dev->mtu = new_mtu;
1332 return 0;
1333 }
1334
1335 static const struct net_device_ops ipip6_netdev_ops = {
1336 .ndo_init = ipip6_tunnel_init,
1337 .ndo_uninit = ipip6_tunnel_uninit,
1338 .ndo_start_xmit = sit_tunnel_xmit,
1339 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1340 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1341 .ndo_get_stats64 = ip_tunnel_get_stats64,
1342 .ndo_get_iflink = ip_tunnel_get_iflink,
1343 };
1344
1345 static void ipip6_dev_free(struct net_device *dev)
1346 {
1347 struct ip_tunnel *tunnel = netdev_priv(dev);
1348
1349 dst_cache_destroy(&tunnel->dst_cache);
1350 free_percpu(dev->tstats);
1351 free_netdev(dev);
1352 }
1353
1354 #define SIT_FEATURES (NETIF_F_SG | \
1355 NETIF_F_FRAGLIST | \
1356 NETIF_F_HIGHDMA | \
1357 NETIF_F_GSO_SOFTWARE | \
1358 NETIF_F_HW_CSUM)
1359
1360 static void ipip6_tunnel_setup(struct net_device *dev)
1361 {
1362 struct ip_tunnel *tunnel = netdev_priv(dev);
1363 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1364
1365 dev->netdev_ops = &ipip6_netdev_ops;
1366 dev->destructor = ipip6_dev_free;
1367
1368 dev->type = ARPHRD_SIT;
1369 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1370 dev->mtu = ETH_DATA_LEN - t_hlen;
1371 dev->flags = IFF_NOARP;
1372 netif_keep_dst(dev);
1373 dev->addr_len = 4;
1374 dev->features |= NETIF_F_LLTX;
1375 dev->features |= SIT_FEATURES;
1376 dev->hw_features |= SIT_FEATURES;
1377 }
1378
1379 static int ipip6_tunnel_init(struct net_device *dev)
1380 {
1381 struct ip_tunnel *tunnel = netdev_priv(dev);
1382 int err;
1383
1384 tunnel->dev = dev;
1385 tunnel->net = dev_net(dev);
1386 strcpy(tunnel->parms.name, dev->name);
1387
1388 ipip6_tunnel_bind_dev(dev);
1389 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1390 if (!dev->tstats)
1391 return -ENOMEM;
1392
1393 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1394 if (err) {
1395 free_percpu(dev->tstats);
1396 dev->tstats = NULL;
1397 return err;
1398 }
1399
1400 return 0;
1401 }
1402
1403 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1404 {
1405 struct ip_tunnel *tunnel = netdev_priv(dev);
1406 struct iphdr *iph = &tunnel->parms.iph;
1407 struct net *net = dev_net(dev);
1408 struct sit_net *sitn = net_generic(net, sit_net_id);
1409
1410 iph->version = 4;
1411 iph->protocol = IPPROTO_IPV6;
1412 iph->ihl = 5;
1413 iph->ttl = 64;
1414
1415 dev_hold(dev);
1416 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1417 }
1418
1419 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1420 {
1421 u8 proto;
1422
1423 if (!data || !data[IFLA_IPTUN_PROTO])
1424 return 0;
1425
1426 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1427 if (proto != IPPROTO_IPV6 &&
1428 proto != IPPROTO_IPIP &&
1429 proto != 0)
1430 return -EINVAL;
1431
1432 return 0;
1433 }
1434
1435 static void ipip6_netlink_parms(struct nlattr *data[],
1436 struct ip_tunnel_parm *parms)
1437 {
1438 memset(parms, 0, sizeof(*parms));
1439
1440 parms->iph.version = 4;
1441 parms->iph.protocol = IPPROTO_IPV6;
1442 parms->iph.ihl = 5;
1443 parms->iph.ttl = 64;
1444
1445 if (!data)
1446 return;
1447
1448 if (data[IFLA_IPTUN_LINK])
1449 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1450
1451 if (data[IFLA_IPTUN_LOCAL])
1452 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1453
1454 if (data[IFLA_IPTUN_REMOTE])
1455 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1456
1457 if (data[IFLA_IPTUN_TTL]) {
1458 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1459 if (parms->iph.ttl)
1460 parms->iph.frag_off = htons(IP_DF);
1461 }
1462
1463 if (data[IFLA_IPTUN_TOS])
1464 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1465
1466 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1467 parms->iph.frag_off = htons(IP_DF);
1468
1469 if (data[IFLA_IPTUN_FLAGS])
1470 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1471
1472 if (data[IFLA_IPTUN_PROTO])
1473 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1474
1475 }
1476
1477 /* This function returns true when ENCAP attributes are present in the nl msg */
1478 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1479 struct ip_tunnel_encap *ipencap)
1480 {
1481 bool ret = false;
1482
1483 memset(ipencap, 0, sizeof(*ipencap));
1484
1485 if (!data)
1486 return ret;
1487
1488 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1489 ret = true;
1490 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1491 }
1492
1493 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1494 ret = true;
1495 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1496 }
1497
1498 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1499 ret = true;
1500 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1501 }
1502
1503 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1504 ret = true;
1505 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1506 }
1507
1508 return ret;
1509 }
1510
1511 #ifdef CONFIG_IPV6_SIT_6RD
1512 /* This function returns true when 6RD attributes are present in the nl msg */
1513 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1514 struct ip_tunnel_6rd *ip6rd)
1515 {
1516 bool ret = false;
1517 memset(ip6rd, 0, sizeof(*ip6rd));
1518
1519 if (!data)
1520 return ret;
1521
1522 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1523 ret = true;
1524 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1525 }
1526
1527 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1528 ret = true;
1529 ip6rd->relay_prefix =
1530 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1531 }
1532
1533 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1534 ret = true;
1535 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1536 }
1537
1538 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1539 ret = true;
1540 ip6rd->relay_prefixlen =
1541 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1542 }
1543
1544 return ret;
1545 }
1546 #endif
1547
1548 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1549 struct nlattr *tb[], struct nlattr *data[])
1550 {
1551 struct net *net = dev_net(dev);
1552 struct ip_tunnel *nt;
1553 struct ip_tunnel_encap ipencap;
1554 #ifdef CONFIG_IPV6_SIT_6RD
1555 struct ip_tunnel_6rd ip6rd;
1556 #endif
1557 int err;
1558
1559 nt = netdev_priv(dev);
1560
1561 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1562 err = ip_tunnel_encap_setup(nt, &ipencap);
1563 if (err < 0)
1564 return err;
1565 }
1566
1567 ipip6_netlink_parms(data, &nt->parms);
1568
1569 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1570 return -EEXIST;
1571
1572 err = ipip6_tunnel_create(dev);
1573 if (err < 0)
1574 return err;
1575
1576 if (tb[IFLA_MTU]) {
1577 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1578
1579 if (mtu >= IPV6_MIN_MTU && mtu <= 0xFFF8 - dev->hard_header_len)
1580 dev->mtu = mtu;
1581 }
1582
1583 #ifdef CONFIG_IPV6_SIT_6RD
1584 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1585 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1586 #endif
1587
1588 return err;
1589 }
1590
1591 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1592 struct nlattr *data[])
1593 {
1594 struct ip_tunnel *t = netdev_priv(dev);
1595 struct ip_tunnel_parm p;
1596 struct ip_tunnel_encap ipencap;
1597 struct net *net = t->net;
1598 struct sit_net *sitn = net_generic(net, sit_net_id);
1599 #ifdef CONFIG_IPV6_SIT_6RD
1600 struct ip_tunnel_6rd ip6rd;
1601 #endif
1602 int err;
1603
1604 if (dev == sitn->fb_tunnel_dev)
1605 return -EINVAL;
1606
1607 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1608 err = ip_tunnel_encap_setup(t, &ipencap);
1609 if (err < 0)
1610 return err;
1611 }
1612
1613 ipip6_netlink_parms(data, &p);
1614
1615 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1616 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1617 return -EINVAL;
1618
1619 t = ipip6_tunnel_locate(net, &p, 0);
1620
1621 if (t) {
1622 if (t->dev != dev)
1623 return -EEXIST;
1624 } else
1625 t = netdev_priv(dev);
1626
1627 ipip6_tunnel_update(t, &p);
1628
1629 #ifdef CONFIG_IPV6_SIT_6RD
1630 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1631 return ipip6_tunnel_update_6rd(t, &ip6rd);
1632 #endif
1633
1634 return 0;
1635 }
1636
1637 static size_t ipip6_get_size(const struct net_device *dev)
1638 {
1639 return
1640 /* IFLA_IPTUN_LINK */
1641 nla_total_size(4) +
1642 /* IFLA_IPTUN_LOCAL */
1643 nla_total_size(4) +
1644 /* IFLA_IPTUN_REMOTE */
1645 nla_total_size(4) +
1646 /* IFLA_IPTUN_TTL */
1647 nla_total_size(1) +
1648 /* IFLA_IPTUN_TOS */
1649 nla_total_size(1) +
1650 /* IFLA_IPTUN_PMTUDISC */
1651 nla_total_size(1) +
1652 /* IFLA_IPTUN_FLAGS */
1653 nla_total_size(2) +
1654 /* IFLA_IPTUN_PROTO */
1655 nla_total_size(1) +
1656 #ifdef CONFIG_IPV6_SIT_6RD
1657 /* IFLA_IPTUN_6RD_PREFIX */
1658 nla_total_size(sizeof(struct in6_addr)) +
1659 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1660 nla_total_size(4) +
1661 /* IFLA_IPTUN_6RD_PREFIXLEN */
1662 nla_total_size(2) +
1663 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1664 nla_total_size(2) +
1665 #endif
1666 /* IFLA_IPTUN_ENCAP_TYPE */
1667 nla_total_size(2) +
1668 /* IFLA_IPTUN_ENCAP_FLAGS */
1669 nla_total_size(2) +
1670 /* IFLA_IPTUN_ENCAP_SPORT */
1671 nla_total_size(2) +
1672 /* IFLA_IPTUN_ENCAP_DPORT */
1673 nla_total_size(2) +
1674 0;
1675 }
1676
1677 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1678 {
1679 struct ip_tunnel *tunnel = netdev_priv(dev);
1680 struct ip_tunnel_parm *parm = &tunnel->parms;
1681
1682 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1683 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1684 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1685 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1686 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1687 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1688 !!(parm->iph.frag_off & htons(IP_DF))) ||
1689 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1690 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1691 goto nla_put_failure;
1692
1693 #ifdef CONFIG_IPV6_SIT_6RD
1694 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1695 &tunnel->ip6rd.prefix) ||
1696 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1697 tunnel->ip6rd.relay_prefix) ||
1698 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1699 tunnel->ip6rd.prefixlen) ||
1700 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1701 tunnel->ip6rd.relay_prefixlen))
1702 goto nla_put_failure;
1703 #endif
1704
1705 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1706 tunnel->encap.type) ||
1707 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1708 tunnel->encap.sport) ||
1709 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1710 tunnel->encap.dport) ||
1711 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1712 tunnel->encap.flags))
1713 goto nla_put_failure;
1714
1715 return 0;
1716
1717 nla_put_failure:
1718 return -EMSGSIZE;
1719 }
1720
1721 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1722 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1723 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1724 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1725 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1726 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1727 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1728 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1729 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1730 #ifdef CONFIG_IPV6_SIT_6RD
1731 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1732 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1733 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1734 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1735 #endif
1736 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1737 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1738 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1739 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1740 };
1741
1742 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1743 {
1744 struct net *net = dev_net(dev);
1745 struct sit_net *sitn = net_generic(net, sit_net_id);
1746
1747 if (dev != sitn->fb_tunnel_dev)
1748 unregister_netdevice_queue(dev, head);
1749 }
1750
1751 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1752 .kind = "sit",
1753 .maxtype = IFLA_IPTUN_MAX,
1754 .policy = ipip6_policy,
1755 .priv_size = sizeof(struct ip_tunnel),
1756 .setup = ipip6_tunnel_setup,
1757 .validate = ipip6_validate,
1758 .newlink = ipip6_newlink,
1759 .changelink = ipip6_changelink,
1760 .get_size = ipip6_get_size,
1761 .fill_info = ipip6_fill_info,
1762 .dellink = ipip6_dellink,
1763 .get_link_net = ip_tunnel_get_link_net,
1764 };
1765
1766 static struct xfrm_tunnel sit_handler __read_mostly = {
1767 .handler = ipip6_rcv,
1768 .err_handler = ipip6_err,
1769 .priority = 1,
1770 };
1771
1772 static struct xfrm_tunnel ipip_handler __read_mostly = {
1773 .handler = ipip_rcv,
1774 .err_handler = ipip6_err,
1775 .priority = 2,
1776 };
1777
1778 static void __net_exit sit_destroy_tunnels(struct net *net,
1779 struct list_head *head)
1780 {
1781 struct sit_net *sitn = net_generic(net, sit_net_id);
1782 struct net_device *dev, *aux;
1783 int prio;
1784
1785 for_each_netdev_safe(net, dev, aux)
1786 if (dev->rtnl_link_ops == &sit_link_ops)
1787 unregister_netdevice_queue(dev, head);
1788
1789 for (prio = 1; prio < 4; prio++) {
1790 int h;
1791 for (h = 0; h < HASH_SIZE; h++) {
1792 struct ip_tunnel *t;
1793
1794 t = rtnl_dereference(sitn->tunnels[prio][h]);
1795 while (t) {
1796 /* If dev is in the same netns, it has already
1797 * been added to the list by the previous loop.
1798 */
1799 if (!net_eq(dev_net(t->dev), net))
1800 unregister_netdevice_queue(t->dev,
1801 head);
1802 t = rtnl_dereference(t->next);
1803 }
1804 }
1805 }
1806 }
1807
1808 static int __net_init sit_init_net(struct net *net)
1809 {
1810 struct sit_net *sitn = net_generic(net, sit_net_id);
1811 struct ip_tunnel *t;
1812 int err;
1813
1814 sitn->tunnels[0] = sitn->tunnels_wc;
1815 sitn->tunnels[1] = sitn->tunnels_l;
1816 sitn->tunnels[2] = sitn->tunnels_r;
1817 sitn->tunnels[3] = sitn->tunnels_r_l;
1818
1819 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1820 NET_NAME_UNKNOWN,
1821 ipip6_tunnel_setup);
1822 if (!sitn->fb_tunnel_dev) {
1823 err = -ENOMEM;
1824 goto err_alloc_dev;
1825 }
1826 dev_net_set(sitn->fb_tunnel_dev, net);
1827 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1828 /* FB netdevice is special: we have one, and only one per netns.
1829 * Allowing to move it to another netns is clearly unsafe.
1830 */
1831 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1832
1833 err = register_netdev(sitn->fb_tunnel_dev);
1834 if (err)
1835 goto err_reg_dev;
1836
1837 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1838 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1839
1840 t = netdev_priv(sitn->fb_tunnel_dev);
1841
1842 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1843 return 0;
1844
1845 err_reg_dev:
1846 ipip6_dev_free(sitn->fb_tunnel_dev);
1847 err_alloc_dev:
1848 return err;
1849 }
1850
1851 static void __net_exit sit_exit_net(struct net *net)
1852 {
1853 LIST_HEAD(list);
1854
1855 rtnl_lock();
1856 sit_destroy_tunnels(net, &list);
1857 unregister_netdevice_many(&list);
1858 rtnl_unlock();
1859 }
1860
1861 static struct pernet_operations sit_net_ops = {
1862 .init = sit_init_net,
1863 .exit = sit_exit_net,
1864 .id = &sit_net_id,
1865 .size = sizeof(struct sit_net),
1866 };
1867
1868 static void __exit sit_cleanup(void)
1869 {
1870 rtnl_link_unregister(&sit_link_ops);
1871 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1872 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1873
1874 unregister_pernet_device(&sit_net_ops);
1875 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1876 }
1877
1878 static int __init sit_init(void)
1879 {
1880 int err;
1881
1882 pr_info("IPv6 over IPv4 tunneling driver\n");
1883
1884 err = register_pernet_device(&sit_net_ops);
1885 if (err < 0)
1886 return err;
1887 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1888 if (err < 0) {
1889 pr_info("%s: can't register ip6ip4\n", __func__);
1890 goto xfrm_tunnel_failed;
1891 }
1892 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1893 if (err < 0) {
1894 pr_info("%s: can't register ip4ip4\n", __func__);
1895 goto xfrm_tunnel4_failed;
1896 }
1897 err = rtnl_link_register(&sit_link_ops);
1898 if (err < 0)
1899 goto rtnl_link_failed;
1900
1901 out:
1902 return err;
1903
1904 rtnl_link_failed:
1905 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1906 xfrm_tunnel4_failed:
1907 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1908 xfrm_tunnel_failed:
1909 unregister_pernet_device(&sit_net_ops);
1910 goto out;
1911 }
1912
1913 module_init(sit_init);
1914 module_exit(sit_cleanup);
1915 MODULE_LICENSE("GPL");
1916 MODULE_ALIAS_RTNL_LINK("sit");
1917 MODULE_ALIAS_NETDEV("sit0");
Linux with added FPC-III support