]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv6/sit.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ip_tunnels: extend iptunnel_xmit()
[mirror_ubuntu-artful-kernel.git] / net / ipv6 / sit.c
1 /*
2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 *
14 * Changes:
15 * Roger Venning <r.venning@telstra.com>: 6to4 support
16 * Nate Thompson <nate@thebog.net>: 6to4 support
17 * Fred Templin <fred.l.templin@boeing.com>: isatap support
18 */
19
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38
39 #include <net/sock.h>
40 #include <net/snmp.h>
41
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/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->iflink == 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->iflink == 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->iflink == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
118 return t;
119 }
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if ((t != NULL) && (t->dev->flags & IFF_UP))
122 return t;
123 return NULL;
124 }
125
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 struct ip_tunnel_parm *parms)
128 {
129 __be32 remote = parms->iph.daddr;
130 __be32 local = parms->iph.saddr;
131 unsigned int h = 0;
132 int prio = 0;
133
134 if (remote) {
135 prio |= 2;
136 h ^= HASH(remote);
137 }
138 if (local) {
139 prio |= 1;
140 h ^= HASH(local);
141 }
142 return &sitn->tunnels[prio][h];
143 }
144
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146 struct ip_tunnel *t)
147 {
148 return __ipip6_bucket(sitn, &t->parms);
149 }
150
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152 {
153 struct ip_tunnel __rcu **tp;
154 struct ip_tunnel *iter;
155
156 for (tp = ipip6_bucket(sitn, t);
157 (iter = rtnl_dereference(*tp)) != NULL;
158 tp = &iter->next) {
159 if (t == iter) {
160 rcu_assign_pointer(*tp, t->next);
161 break;
162 }
163 }
164 }
165
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167 {
168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169
170 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 rcu_assign_pointer(*tp, t);
172 }
173
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175 {
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel *t = netdev_priv(dev);
178
179 if (t->dev == sitn->fb_tunnel_dev) {
180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 t->ip6rd.relay_prefix = 0;
182 t->ip6rd.prefixlen = 16;
183 t->ip6rd.relay_prefixlen = 0;
184 } else {
185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187 }
188 #endif
189 }
190
191 static int ipip6_tunnel_create(struct net_device *dev)
192 {
193 struct ip_tunnel *t = netdev_priv(dev);
194 struct net *net = dev_net(dev);
195 struct sit_net *sitn = net_generic(net, sit_net_id);
196 int err;
197
198 err = ipip6_tunnel_init(dev);
199 if (err < 0)
200 goto out;
201 ipip6_tunnel_clone_6rd(dev, sitn);
202
203 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 dev->priv_flags |= IFF_ISATAP;
205
206 err = register_netdevice(dev);
207 if (err < 0)
208 goto out;
209
210 strcpy(t->parms.name, dev->name);
211 dev->rtnl_link_ops = &sit_link_ops;
212
213 dev_hold(dev);
214
215 ipip6_tunnel_link(sitn, t);
216 return 0;
217
218 out:
219 return err;
220 }
221
222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
223 struct ip_tunnel_parm *parms, int create)
224 {
225 __be32 remote = parms->iph.daddr;
226 __be32 local = parms->iph.saddr;
227 struct ip_tunnel *t, *nt;
228 struct ip_tunnel __rcu **tp;
229 struct net_device *dev;
230 char name[IFNAMSIZ];
231 struct sit_net *sitn = net_generic(net, sit_net_id);
232
233 for (tp = __ipip6_bucket(sitn, parms);
234 (t = rtnl_dereference(*tp)) != NULL;
235 tp = &t->next) {
236 if (local == t->parms.iph.saddr &&
237 remote == t->parms.iph.daddr &&
238 parms->link == t->parms.link) {
239 if (create)
240 return NULL;
241 else
242 return t;
243 }
244 }
245 if (!create)
246 goto failed;
247
248 if (parms->name[0])
249 strlcpy(name, parms->name, IFNAMSIZ);
250 else
251 strcpy(name, "sit%d");
252
253 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
254 if (dev == NULL)
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 net *net = dev_net(dev);
470 struct sit_net *sitn = net_generic(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, netdev_priv(dev));
476 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
477 }
478 dev_put(dev);
479 }
480
481
482 static int ipip6_err(struct sk_buff *skb, u32 info)
483 {
484
485 /* All the routers (except for Linux) return only
486 8 bytes of packet payload. It means, that precise relaying of
487 ICMP in the real Internet is absolutely infeasible.
488 */
489 const struct iphdr *iph = (const struct iphdr *)skb->data;
490 const int type = icmp_hdr(skb)->type;
491 const int code = icmp_hdr(skb)->code;
492 struct ip_tunnel *t;
493 int err;
494
495 switch (type) {
496 default:
497 case ICMP_PARAMETERPROB:
498 return 0;
499
500 case ICMP_DEST_UNREACH:
501 switch (code) {
502 case ICMP_SR_FAILED:
503 case ICMP_PORT_UNREACH:
504 /* Impossible event. */
505 return 0;
506 default:
507 /* All others are translated to HOST_UNREACH.
508 rfc2003 contains "deep thoughts" about NET_UNREACH,
509 I believe they are just ether pollution. --ANK
510 */
511 break;
512 }
513 break;
514 case ICMP_TIME_EXCEEDED:
515 if (code != ICMP_EXC_TTL)
516 return 0;
517 break;
518 case ICMP_REDIRECT:
519 break;
520 }
521
522 err = -ENOENT;
523
524 t = ipip6_tunnel_lookup(dev_net(skb->dev),
525 skb->dev,
526 iph->daddr,
527 iph->saddr);
528 if (t == NULL)
529 goto out;
530
531 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
532 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
533 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
534 err = 0;
535 goto out;
536 }
537 if (type == ICMP_REDIRECT) {
538 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
539 IPPROTO_IPV6, 0);
540 err = 0;
541 goto out;
542 }
543
544 if (t->parms.iph.daddr == 0)
545 goto out;
546
547 err = 0;
548 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
549 goto out;
550
551 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
552 t->err_count++;
553 else
554 t->err_count = 1;
555 t->err_time = jiffies;
556 out:
557 return err;
558 }
559
560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
561 const struct in6_addr *v6addr)
562 {
563 __be32 v4embed = 0;
564 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
565 return true;
566 return false;
567 }
568
569 static int ipip6_rcv(struct sk_buff *skb)
570 {
571 const struct iphdr *iph = ip_hdr(skb);
572 struct ip_tunnel *tunnel;
573 int err;
574
575 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
576 iph->saddr, iph->daddr);
577 if (tunnel != NULL) {
578 struct pcpu_tstats *tstats;
579
580 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
581 tunnel->parms.iph.protocol != 0)
582 goto out;
583
584 secpath_reset(skb);
585 skb->mac_header = skb->network_header;
586 skb_reset_network_header(skb);
587 IPCB(skb)->flags = 0;
588 skb->protocol = htons(ETH_P_IPV6);
589 skb->pkt_type = PACKET_HOST;
590
591 if (tunnel->dev->priv_flags & IFF_ISATAP) {
592 if (!isatap_chksrc(skb, iph, tunnel)) {
593 tunnel->dev->stats.rx_errors++;
594 goto out;
595 }
596 } else {
597 if (is_spoofed_6rd(tunnel, iph->saddr,
598 &ipv6_hdr(skb)->saddr) ||
599 is_spoofed_6rd(tunnel, iph->daddr,
600 &ipv6_hdr(skb)->daddr)) {
601 tunnel->dev->stats.rx_errors++;
602 goto out;
603 }
604 }
605
606 __skb_tunnel_rx(skb, tunnel->dev);
607
608 err = IP_ECN_decapsulate(iph, skb);
609 if (unlikely(err)) {
610 if (log_ecn_error)
611 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
612 &iph->saddr, iph->tos);
613 if (err > 1) {
614 ++tunnel->dev->stats.rx_frame_errors;
615 ++tunnel->dev->stats.rx_errors;
616 goto out;
617 }
618 }
619
620 tstats = this_cpu_ptr(tunnel->dev->tstats);
621 tstats->rx_packets++;
622 tstats->rx_bytes += skb->len;
623
624 netif_rx(skb);
625
626 return 0;
627 }
628
629 /* no tunnel matched, let upstream know, ipsec may handle it */
630 return 1;
631 out:
632 kfree_skb(skb);
633 return 0;
634 }
635
636 static const struct tnl_ptk_info tpi = {
637 /* no tunnel info required for ipip. */
638 .proto = htons(ETH_P_IP),
639 };
640
641 static int ipip_rcv(struct sk_buff *skb)
642 {
643 const struct iphdr *iph = ip_hdr(skb);
644 struct ip_tunnel *tunnel;
645
646 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
647 iph->saddr, iph->daddr);
648 if (tunnel != NULL) {
649 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
650 tunnel->parms.iph.protocol != 0)
651 goto drop;
652
653 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
654 goto drop;
655 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
656 }
657
658 return 1;
659
660 drop:
661 kfree_skb(skb);
662 return 0;
663 }
664
665 /*
666 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
667 * stores the embedded IPv4 address in v4dst and returns true.
668 */
669 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
670 __be32 *v4dst)
671 {
672 #ifdef CONFIG_IPV6_SIT_6RD
673 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
674 tunnel->ip6rd.prefixlen)) {
675 unsigned int pbw0, pbi0;
676 int pbi1;
677 u32 d;
678
679 pbw0 = tunnel->ip6rd.prefixlen >> 5;
680 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
681
682 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
683 tunnel->ip6rd.relay_prefixlen;
684
685 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
686 if (pbi1 > 0)
687 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
688 (32 - pbi1);
689
690 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
691 return true;
692 }
693 #else
694 if (v6dst->s6_addr16[0] == htons(0x2002)) {
695 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
696 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
697 return true;
698 }
699 #endif
700 return false;
701 }
702
703 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
704 const struct in6_addr *v6dst)
705 {
706 __be32 dst = 0;
707 check_6rd(tunnel, v6dst, &dst);
708 return dst;
709 }
710
711 /*
712 * This function assumes it is being called from dev_queue_xmit()
713 * and that skb is filled properly by that function.
714 */
715
716 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
717 struct net_device *dev)
718 {
719 struct ip_tunnel *tunnel = netdev_priv(dev);
720 const struct iphdr *tiph = &tunnel->parms.iph;
721 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
722 u8 tos = tunnel->parms.iph.tos;
723 __be16 df = tiph->frag_off;
724 struct rtable *rt; /* Route to the other host */
725 struct net_device *tdev; /* Device to other host */
726 unsigned int max_headroom; /* The extra header space needed */
727 __be32 dst = tiph->daddr;
728 struct flowi4 fl4;
729 int mtu;
730 const struct in6_addr *addr6;
731 int addr_type;
732 u8 ttl;
733 int err;
734
735 if (skb->protocol != htons(ETH_P_IPV6))
736 goto tx_error;
737
738 if (tos == 1)
739 tos = ipv6_get_dsfield(iph6);
740
741 /* ISATAP (RFC4214) - must come before 6to4 */
742 if (dev->priv_flags & IFF_ISATAP) {
743 struct neighbour *neigh = NULL;
744 bool do_tx_error = false;
745
746 if (skb_dst(skb))
747 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
748
749 if (neigh == NULL) {
750 net_dbg_ratelimited("sit: nexthop == NULL\n");
751 goto tx_error;
752 }
753
754 addr6 = (const struct in6_addr *)&neigh->primary_key;
755 addr_type = ipv6_addr_type(addr6);
756
757 if ((addr_type & IPV6_ADDR_UNICAST) &&
758 ipv6_addr_is_isatap(addr6))
759 dst = addr6->s6_addr32[3];
760 else
761 do_tx_error = true;
762
763 neigh_release(neigh);
764 if (do_tx_error)
765 goto tx_error;
766 }
767
768 if (!dst)
769 dst = try_6rd(tunnel, &iph6->daddr);
770
771 if (!dst) {
772 struct neighbour *neigh = NULL;
773 bool do_tx_error = false;
774
775 if (skb_dst(skb))
776 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
777
778 if (neigh == NULL) {
779 net_dbg_ratelimited("sit: nexthop == NULL\n");
780 goto tx_error;
781 }
782
783 addr6 = (const struct in6_addr *)&neigh->primary_key;
784 addr_type = ipv6_addr_type(addr6);
785
786 if (addr_type == IPV6_ADDR_ANY) {
787 addr6 = &ipv6_hdr(skb)->daddr;
788 addr_type = ipv6_addr_type(addr6);
789 }
790
791 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
792 dst = addr6->s6_addr32[3];
793 else
794 do_tx_error = true;
795
796 neigh_release(neigh);
797 if (do_tx_error)
798 goto tx_error;
799 }
800
801 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
802 dst, tiph->saddr,
803 0, 0,
804 IPPROTO_IPV6, RT_TOS(tos),
805 tunnel->parms.link);
806 if (IS_ERR(rt)) {
807 dev->stats.tx_carrier_errors++;
808 goto tx_error_icmp;
809 }
810 if (rt->rt_type != RTN_UNICAST) {
811 ip_rt_put(rt);
812 dev->stats.tx_carrier_errors++;
813 goto tx_error_icmp;
814 }
815 tdev = rt->dst.dev;
816
817 if (tdev == dev) {
818 ip_rt_put(rt);
819 dev->stats.collisions++;
820 goto tx_error;
821 }
822
823 if (df) {
824 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
825
826 if (mtu < 68) {
827 dev->stats.collisions++;
828 ip_rt_put(rt);
829 goto tx_error;
830 }
831
832 if (mtu < IPV6_MIN_MTU) {
833 mtu = IPV6_MIN_MTU;
834 df = 0;
835 }
836
837 if (tunnel->parms.iph.daddr && skb_dst(skb))
838 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
839
840 if (skb->len > mtu) {
841 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
842 ip_rt_put(rt);
843 goto tx_error;
844 }
845 }
846
847 if (tunnel->err_count > 0) {
848 if (time_before(jiffies,
849 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
850 tunnel->err_count--;
851 dst_link_failure(skb);
852 } else
853 tunnel->err_count = 0;
854 }
855
856 /*
857 * Okay, now see if we can stuff it in the buffer as-is.
858 */
859 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
860
861 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
862 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
863 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
864 if (!new_skb) {
865 ip_rt_put(rt);
866 dev->stats.tx_dropped++;
867 dev_kfree_skb(skb);
868 return NETDEV_TX_OK;
869 }
870 if (skb->sk)
871 skb_set_owner_w(new_skb, skb->sk);
872 dev_kfree_skb(skb);
873 skb = new_skb;
874 iph6 = ipv6_hdr(skb);
875 }
876 ttl = tiph->ttl;
877 if (ttl == 0)
878 ttl = iph6->hop_limit;
879 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
880
881 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, fl4.daddr,
882 IPPROTO_IPV6, tos, ttl, df);
883 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
884 return NETDEV_TX_OK;
885
886 tx_error_icmp:
887 dst_link_failure(skb);
888 tx_error:
889 dev->stats.tx_errors++;
890 dev_kfree_skb(skb);
891 return NETDEV_TX_OK;
892 }
893
894 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
895 {
896 struct ip_tunnel *tunnel = netdev_priv(dev);
897 const struct iphdr *tiph = &tunnel->parms.iph;
898
899 if (likely(!skb->encapsulation)) {
900 skb_reset_inner_headers(skb);
901 skb->encapsulation = 1;
902 }
903
904 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
905 return NETDEV_TX_OK;
906 }
907
908 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
909 struct net_device *dev)
910 {
911 switch (skb->protocol) {
912 case htons(ETH_P_IP):
913 ipip_tunnel_xmit(skb, dev);
914 break;
915 case htons(ETH_P_IPV6):
916 ipip6_tunnel_xmit(skb, dev);
917 break;
918 default:
919 goto tx_err;
920 }
921
922 return NETDEV_TX_OK;
923
924 tx_err:
925 dev->stats.tx_errors++;
926 dev_kfree_skb(skb);
927 return NETDEV_TX_OK;
928
929 }
930
931 static void ipip6_tunnel_bind_dev(struct net_device *dev)
932 {
933 struct net_device *tdev = NULL;
934 struct ip_tunnel *tunnel;
935 const struct iphdr *iph;
936 struct flowi4 fl4;
937
938 tunnel = netdev_priv(dev);
939 iph = &tunnel->parms.iph;
940
941 if (iph->daddr) {
942 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
943 iph->daddr, iph->saddr,
944 0, 0,
945 IPPROTO_IPV6,
946 RT_TOS(iph->tos),
947 tunnel->parms.link);
948
949 if (!IS_ERR(rt)) {
950 tdev = rt->dst.dev;
951 ip_rt_put(rt);
952 }
953 dev->flags |= IFF_POINTOPOINT;
954 }
955
956 if (!tdev && tunnel->parms.link)
957 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
958
959 if (tdev) {
960 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
961 dev->mtu = tdev->mtu - sizeof(struct iphdr);
962 if (dev->mtu < IPV6_MIN_MTU)
963 dev->mtu = IPV6_MIN_MTU;
964 }
965 dev->iflink = tunnel->parms.link;
966 }
967
968 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
969 {
970 struct net *net = dev_net(t->dev);
971 struct sit_net *sitn = net_generic(net, sit_net_id);
972
973 ipip6_tunnel_unlink(sitn, t);
974 synchronize_net();
975 t->parms.iph.saddr = p->iph.saddr;
976 t->parms.iph.daddr = p->iph.daddr;
977 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
978 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
979 ipip6_tunnel_link(sitn, t);
980 t->parms.iph.ttl = p->iph.ttl;
981 t->parms.iph.tos = p->iph.tos;
982 if (t->parms.link != p->link) {
983 t->parms.link = p->link;
984 ipip6_tunnel_bind_dev(t->dev);
985 }
986 netdev_state_change(t->dev);
987 }
988
989 #ifdef CONFIG_IPV6_SIT_6RD
990 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
991 struct ip_tunnel_6rd *ip6rd)
992 {
993 struct in6_addr prefix;
994 __be32 relay_prefix;
995
996 if (ip6rd->relay_prefixlen > 32 ||
997 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
998 return -EINVAL;
999
1000 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1001 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1002 return -EINVAL;
1003 if (ip6rd->relay_prefixlen)
1004 relay_prefix = ip6rd->relay_prefix &
1005 htonl(0xffffffffUL <<
1006 (32 - ip6rd->relay_prefixlen));
1007 else
1008 relay_prefix = 0;
1009 if (relay_prefix != ip6rd->relay_prefix)
1010 return -EINVAL;
1011
1012 t->ip6rd.prefix = prefix;
1013 t->ip6rd.relay_prefix = relay_prefix;
1014 t->ip6rd.prefixlen = ip6rd->prefixlen;
1015 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1016 netdev_state_change(t->dev);
1017 return 0;
1018 }
1019 #endif
1020
1021 static int
1022 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1023 {
1024 int err = 0;
1025 struct ip_tunnel_parm p;
1026 struct ip_tunnel_prl prl;
1027 struct ip_tunnel *t;
1028 struct net *net = dev_net(dev);
1029 struct sit_net *sitn = net_generic(net, sit_net_id);
1030 #ifdef CONFIG_IPV6_SIT_6RD
1031 struct ip_tunnel_6rd ip6rd;
1032 #endif
1033
1034 switch (cmd) {
1035 case SIOCGETTUNNEL:
1036 #ifdef CONFIG_IPV6_SIT_6RD
1037 case SIOCGET6RD:
1038 #endif
1039 t = NULL;
1040 if (dev == sitn->fb_tunnel_dev) {
1041 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1042 err = -EFAULT;
1043 break;
1044 }
1045 t = ipip6_tunnel_locate(net, &p, 0);
1046 }
1047 if (t == NULL)
1048 t = netdev_priv(dev);
1049
1050 err = -EFAULT;
1051 if (cmd == SIOCGETTUNNEL) {
1052 memcpy(&p, &t->parms, sizeof(p));
1053 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1054 sizeof(p)))
1055 goto done;
1056 #ifdef CONFIG_IPV6_SIT_6RD
1057 } else {
1058 ip6rd.prefix = t->ip6rd.prefix;
1059 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1060 ip6rd.prefixlen = t->ip6rd.prefixlen;
1061 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1062 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1063 sizeof(ip6rd)))
1064 goto done;
1065 #endif
1066 }
1067 err = 0;
1068 break;
1069
1070 case SIOCADDTUNNEL:
1071 case SIOCCHGTUNNEL:
1072 err = -EPERM;
1073 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1074 goto done;
1075
1076 err = -EFAULT;
1077 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1078 goto done;
1079
1080 err = -EINVAL;
1081 if (p.iph.protocol != IPPROTO_IPV6 &&
1082 p.iph.protocol != IPPROTO_IPIP &&
1083 p.iph.protocol != 0)
1084 goto done;
1085 if (p.iph.version != 4 ||
1086 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1087 goto done;
1088 if (p.iph.ttl)
1089 p.iph.frag_off |= htons(IP_DF);
1090
1091 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1092
1093 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1094 if (t != NULL) {
1095 if (t->dev != dev) {
1096 err = -EEXIST;
1097 break;
1098 }
1099 } else {
1100 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1101 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1102 err = -EINVAL;
1103 break;
1104 }
1105 t = netdev_priv(dev);
1106 }
1107
1108 ipip6_tunnel_update(t, &p);
1109 }
1110
1111 if (t) {
1112 err = 0;
1113 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1114 err = -EFAULT;
1115 } else
1116 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1117 break;
1118
1119 case SIOCDELTUNNEL:
1120 err = -EPERM;
1121 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1122 goto done;
1123
1124 if (dev == sitn->fb_tunnel_dev) {
1125 err = -EFAULT;
1126 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1127 goto done;
1128 err = -ENOENT;
1129 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1130 goto done;
1131 err = -EPERM;
1132 if (t == netdev_priv(sitn->fb_tunnel_dev))
1133 goto done;
1134 dev = t->dev;
1135 }
1136 unregister_netdevice(dev);
1137 err = 0;
1138 break;
1139
1140 case SIOCGETPRL:
1141 err = -EINVAL;
1142 if (dev == sitn->fb_tunnel_dev)
1143 goto done;
1144 err = -ENOENT;
1145 if (!(t = netdev_priv(dev)))
1146 goto done;
1147 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1148 break;
1149
1150 case SIOCADDPRL:
1151 case SIOCDELPRL:
1152 case SIOCCHGPRL:
1153 err = -EPERM;
1154 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1155 goto done;
1156 err = -EINVAL;
1157 if (dev == sitn->fb_tunnel_dev)
1158 goto done;
1159 err = -EFAULT;
1160 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1161 goto done;
1162 err = -ENOENT;
1163 if (!(t = netdev_priv(dev)))
1164 goto done;
1165
1166 switch (cmd) {
1167 case SIOCDELPRL:
1168 err = ipip6_tunnel_del_prl(t, &prl);
1169 break;
1170 case SIOCADDPRL:
1171 case SIOCCHGPRL:
1172 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1173 break;
1174 }
1175 netdev_state_change(dev);
1176 break;
1177
1178 #ifdef CONFIG_IPV6_SIT_6RD
1179 case SIOCADD6RD:
1180 case SIOCCHG6RD:
1181 case SIOCDEL6RD:
1182 err = -EPERM;
1183 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1184 goto done;
1185
1186 err = -EFAULT;
1187 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1188 sizeof(ip6rd)))
1189 goto done;
1190
1191 t = netdev_priv(dev);
1192
1193 if (cmd != SIOCDEL6RD) {
1194 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1195 if (err < 0)
1196 goto done;
1197 } else
1198 ipip6_tunnel_clone_6rd(dev, sitn);
1199
1200 err = 0;
1201 break;
1202 #endif
1203
1204 default:
1205 err = -EINVAL;
1206 }
1207
1208 done:
1209 return err;
1210 }
1211
1212 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1213 {
1214 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1215 return -EINVAL;
1216 dev->mtu = new_mtu;
1217 return 0;
1218 }
1219
1220 static const struct net_device_ops ipip6_netdev_ops = {
1221 .ndo_uninit = ipip6_tunnel_uninit,
1222 .ndo_start_xmit = sit_tunnel_xmit,
1223 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1224 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1225 .ndo_get_stats64 = ip_tunnel_get_stats64,
1226 };
1227
1228 static void ipip6_dev_free(struct net_device *dev)
1229 {
1230 free_percpu(dev->tstats);
1231 free_netdev(dev);
1232 }
1233
1234 static void ipip6_tunnel_setup(struct net_device *dev)
1235 {
1236 dev->netdev_ops = &ipip6_netdev_ops;
1237 dev->destructor = ipip6_dev_free;
1238
1239 dev->type = ARPHRD_SIT;
1240 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1241 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1242 dev->flags = IFF_NOARP;
1243 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1244 dev->iflink = 0;
1245 dev->addr_len = 4;
1246 dev->features |= NETIF_F_NETNS_LOCAL;
1247 dev->features |= NETIF_F_LLTX;
1248 }
1249
1250 static int ipip6_tunnel_init(struct net_device *dev)
1251 {
1252 struct ip_tunnel *tunnel = netdev_priv(dev);
1253
1254 tunnel->dev = dev;
1255
1256 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1257 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1258
1259 ipip6_tunnel_bind_dev(dev);
1260 dev->tstats = alloc_percpu(struct pcpu_tstats);
1261 if (!dev->tstats)
1262 return -ENOMEM;
1263
1264 return 0;
1265 }
1266
1267 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1268 {
1269 struct ip_tunnel *tunnel = netdev_priv(dev);
1270 struct iphdr *iph = &tunnel->parms.iph;
1271 struct net *net = dev_net(dev);
1272 struct sit_net *sitn = net_generic(net, sit_net_id);
1273
1274 tunnel->dev = dev;
1275 strcpy(tunnel->parms.name, dev->name);
1276
1277 iph->version = 4;
1278 iph->protocol = IPPROTO_IPV6;
1279 iph->ihl = 5;
1280 iph->ttl = 64;
1281
1282 dev->tstats = alloc_percpu(struct pcpu_tstats);
1283 if (!dev->tstats)
1284 return -ENOMEM;
1285 dev_hold(dev);
1286 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1287 return 0;
1288 }
1289
1290 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1291 {
1292 u8 proto;
1293
1294 if (!data)
1295 return 0;
1296
1297 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1298 if (proto != IPPROTO_IPV6 &&
1299 proto != IPPROTO_IPIP &&
1300 proto != 0)
1301 return -EINVAL;
1302
1303 return 0;
1304 }
1305
1306 static void ipip6_netlink_parms(struct nlattr *data[],
1307 struct ip_tunnel_parm *parms)
1308 {
1309 memset(parms, 0, sizeof(*parms));
1310
1311 parms->iph.version = 4;
1312 parms->iph.protocol = IPPROTO_IPV6;
1313 parms->iph.ihl = 5;
1314 parms->iph.ttl = 64;
1315
1316 if (!data)
1317 return;
1318
1319 if (data[IFLA_IPTUN_LINK])
1320 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1321
1322 if (data[IFLA_IPTUN_LOCAL])
1323 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1324
1325 if (data[IFLA_IPTUN_REMOTE])
1326 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1327
1328 if (data[IFLA_IPTUN_TTL]) {
1329 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1330 if (parms->iph.ttl)
1331 parms->iph.frag_off = htons(IP_DF);
1332 }
1333
1334 if (data[IFLA_IPTUN_TOS])
1335 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1336
1337 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1338 parms->iph.frag_off = htons(IP_DF);
1339
1340 if (data[IFLA_IPTUN_FLAGS])
1341 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1342
1343 if (data[IFLA_IPTUN_PROTO])
1344 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1345
1346 }
1347
1348 #ifdef CONFIG_IPV6_SIT_6RD
1349 /* This function returns true when 6RD attributes are present in the nl msg */
1350 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1351 struct ip_tunnel_6rd *ip6rd)
1352 {
1353 bool ret = false;
1354 memset(ip6rd, 0, sizeof(*ip6rd));
1355
1356 if (!data)
1357 return ret;
1358
1359 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1360 ret = true;
1361 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1362 sizeof(struct in6_addr));
1363 }
1364
1365 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1366 ret = true;
1367 ip6rd->relay_prefix =
1368 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1369 }
1370
1371 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1372 ret = true;
1373 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1374 }
1375
1376 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1377 ret = true;
1378 ip6rd->relay_prefixlen =
1379 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1380 }
1381
1382 return ret;
1383 }
1384 #endif
1385
1386 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1387 struct nlattr *tb[], struct nlattr *data[])
1388 {
1389 struct net *net = dev_net(dev);
1390 struct ip_tunnel *nt;
1391 #ifdef CONFIG_IPV6_SIT_6RD
1392 struct ip_tunnel_6rd ip6rd;
1393 #endif
1394 int err;
1395
1396 nt = netdev_priv(dev);
1397 ipip6_netlink_parms(data, &nt->parms);
1398
1399 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1400 return -EEXIST;
1401
1402 err = ipip6_tunnel_create(dev);
1403 if (err < 0)
1404 return err;
1405
1406 #ifdef CONFIG_IPV6_SIT_6RD
1407 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1408 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1409 #endif
1410
1411 return err;
1412 }
1413
1414 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1415 struct nlattr *data[])
1416 {
1417 struct ip_tunnel *t;
1418 struct ip_tunnel_parm p;
1419 struct net *net = dev_net(dev);
1420 struct sit_net *sitn = net_generic(net, sit_net_id);
1421 #ifdef CONFIG_IPV6_SIT_6RD
1422 struct ip_tunnel_6rd ip6rd;
1423 #endif
1424
1425 if (dev == sitn->fb_tunnel_dev)
1426 return -EINVAL;
1427
1428 ipip6_netlink_parms(data, &p);
1429
1430 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1431 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1432 return -EINVAL;
1433
1434 t = ipip6_tunnel_locate(net, &p, 0);
1435
1436 if (t) {
1437 if (t->dev != dev)
1438 return -EEXIST;
1439 } else
1440 t = netdev_priv(dev);
1441
1442 ipip6_tunnel_update(t, &p);
1443
1444 #ifdef CONFIG_IPV6_SIT_6RD
1445 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1446 return ipip6_tunnel_update_6rd(t, &ip6rd);
1447 #endif
1448
1449 return 0;
1450 }
1451
1452 static size_t ipip6_get_size(const struct net_device *dev)
1453 {
1454 return
1455 /* IFLA_IPTUN_LINK */
1456 nla_total_size(4) +
1457 /* IFLA_IPTUN_LOCAL */
1458 nla_total_size(4) +
1459 /* IFLA_IPTUN_REMOTE */
1460 nla_total_size(4) +
1461 /* IFLA_IPTUN_TTL */
1462 nla_total_size(1) +
1463 /* IFLA_IPTUN_TOS */
1464 nla_total_size(1) +
1465 /* IFLA_IPTUN_PMTUDISC */
1466 nla_total_size(1) +
1467 /* IFLA_IPTUN_FLAGS */
1468 nla_total_size(2) +
1469 /* IFLA_IPTUN_PROTO */
1470 nla_total_size(1) +
1471 #ifdef CONFIG_IPV6_SIT_6RD
1472 /* IFLA_IPTUN_6RD_PREFIX */
1473 nla_total_size(sizeof(struct in6_addr)) +
1474 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1475 nla_total_size(4) +
1476 /* IFLA_IPTUN_6RD_PREFIXLEN */
1477 nla_total_size(2) +
1478 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1479 nla_total_size(2) +
1480 #endif
1481 0;
1482 }
1483
1484 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1485 {
1486 struct ip_tunnel *tunnel = netdev_priv(dev);
1487 struct ip_tunnel_parm *parm = &tunnel->parms;
1488
1489 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1490 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1491 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1492 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1493 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1494 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1495 !!(parm->iph.frag_off & htons(IP_DF))) ||
1496 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1497 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1498 goto nla_put_failure;
1499
1500 #ifdef CONFIG_IPV6_SIT_6RD
1501 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1502 &tunnel->ip6rd.prefix) ||
1503 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1504 tunnel->ip6rd.relay_prefix) ||
1505 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1506 tunnel->ip6rd.prefixlen) ||
1507 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1508 tunnel->ip6rd.relay_prefixlen))
1509 goto nla_put_failure;
1510 #endif
1511
1512 return 0;
1513
1514 nla_put_failure:
1515 return -EMSGSIZE;
1516 }
1517
1518 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1519 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1520 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1521 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1522 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1523 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1524 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1525 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1526 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1527 #ifdef CONFIG_IPV6_SIT_6RD
1528 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1529 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1530 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1531 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1532 #endif
1533 };
1534
1535 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1536 .kind = "sit",
1537 .maxtype = IFLA_IPTUN_MAX,
1538 .policy = ipip6_policy,
1539 .priv_size = sizeof(struct ip_tunnel),
1540 .setup = ipip6_tunnel_setup,
1541 .validate = ipip6_validate,
1542 .newlink = ipip6_newlink,
1543 .changelink = ipip6_changelink,
1544 .get_size = ipip6_get_size,
1545 .fill_info = ipip6_fill_info,
1546 };
1547
1548 static struct xfrm_tunnel sit_handler __read_mostly = {
1549 .handler = ipip6_rcv,
1550 .err_handler = ipip6_err,
1551 .priority = 1,
1552 };
1553
1554 static struct xfrm_tunnel ipip_handler __read_mostly = {
1555 .handler = ipip_rcv,
1556 .err_handler = ipip6_err,
1557 .priority = 2,
1558 };
1559
1560 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1561 {
1562 int prio;
1563
1564 for (prio = 1; prio < 4; prio++) {
1565 int h;
1566 for (h = 0; h < HASH_SIZE; h++) {
1567 struct ip_tunnel *t;
1568
1569 t = rtnl_dereference(sitn->tunnels[prio][h]);
1570 while (t != NULL) {
1571 unregister_netdevice_queue(t->dev, head);
1572 t = rtnl_dereference(t->next);
1573 }
1574 }
1575 }
1576 }
1577
1578 static int __net_init sit_init_net(struct net *net)
1579 {
1580 struct sit_net *sitn = net_generic(net, sit_net_id);
1581 struct ip_tunnel *t;
1582 int err;
1583
1584 sitn->tunnels[0] = sitn->tunnels_wc;
1585 sitn->tunnels[1] = sitn->tunnels_l;
1586 sitn->tunnels[2] = sitn->tunnels_r;
1587 sitn->tunnels[3] = sitn->tunnels_r_l;
1588
1589 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1590 ipip6_tunnel_setup);
1591 if (!sitn->fb_tunnel_dev) {
1592 err = -ENOMEM;
1593 goto err_alloc_dev;
1594 }
1595 dev_net_set(sitn->fb_tunnel_dev, net);
1596
1597 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1598 if (err)
1599 goto err_dev_free;
1600
1601 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1602
1603 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1604 goto err_reg_dev;
1605
1606 t = netdev_priv(sitn->fb_tunnel_dev);
1607
1608 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1609 return 0;
1610
1611 err_reg_dev:
1612 dev_put(sitn->fb_tunnel_dev);
1613 err_dev_free:
1614 ipip6_dev_free(sitn->fb_tunnel_dev);
1615 err_alloc_dev:
1616 return err;
1617 }
1618
1619 static void __net_exit sit_exit_net(struct net *net)
1620 {
1621 struct sit_net *sitn = net_generic(net, sit_net_id);
1622 LIST_HEAD(list);
1623
1624 rtnl_lock();
1625 sit_destroy_tunnels(sitn, &list);
1626 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1627 unregister_netdevice_many(&list);
1628 rtnl_unlock();
1629 }
1630
1631 static struct pernet_operations sit_net_ops = {
1632 .init = sit_init_net,
1633 .exit = sit_exit_net,
1634 .id = &sit_net_id,
1635 .size = sizeof(struct sit_net),
1636 };
1637
1638 static void __exit sit_cleanup(void)
1639 {
1640 rtnl_link_unregister(&sit_link_ops);
1641 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1642 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1643
1644 unregister_pernet_device(&sit_net_ops);
1645 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1646 }
1647
1648 static int __init sit_init(void)
1649 {
1650 int err;
1651
1652 pr_info("IPv6 over IPv4 tunneling driver\n");
1653
1654 err = register_pernet_device(&sit_net_ops);
1655 if (err < 0)
1656 return err;
1657 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1658 if (err < 0) {
1659 pr_info("%s: can't register ip6ip4\n", __func__);
1660 goto xfrm_tunnel_failed;
1661 }
1662 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1663 if (err < 0) {
1664 pr_info("%s: can't register ip4ip4\n", __func__);
1665 goto xfrm_tunnel4_failed;
1666 }
1667 err = rtnl_link_register(&sit_link_ops);
1668 if (err < 0)
1669 goto rtnl_link_failed;
1670
1671 out:
1672 return err;
1673
1674 rtnl_link_failed:
1675 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1676 xfrm_tunnel4_failed:
1677 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1678 xfrm_tunnel_failed:
1679 unregister_pernet_device(&sit_net_ops);
1680 goto out;
1681 }
1682
1683 module_init(sit_init);
1684 module_exit(sit_cleanup);
1685 MODULE_LICENSE("GPL");
1686 MODULE_ALIAS_NETDEV("sit0");
Ubuntu Artful kernel mirror