]> 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
add vif using local interface index instead of IP
[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 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <asm/uaccess.h>
32 #include <linux/init.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/if_ether.h>
35
36 #include <net/sock.h>
37 #include <net/snmp.h>
38
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_fib.h>
43 #include <net/ip6_route.h>
44 #include <net/ndisc.h>
45 #include <net/addrconf.h>
46 #include <net/ip.h>
47 #include <net/udp.h>
48 #include <net/icmp.h>
49 #include <net/ipip.h>
50 #include <net/inet_ecn.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55
56 /*
57 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58
59 For comments look at net/ipv4/ip_gre.c --ANK
60 */
61
62 #define HASH_SIZE 16
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64
65 static void ipip6_fb_tunnel_init(struct net_device *dev);
66 static void ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68
69 static int sit_net_id;
70 struct sit_net {
71 struct ip_tunnel *tunnels_r_l[HASH_SIZE];
72 struct ip_tunnel *tunnels_r[HASH_SIZE];
73 struct ip_tunnel *tunnels_l[HASH_SIZE];
74 struct ip_tunnel *tunnels_wc[1];
75 struct ip_tunnel **tunnels[4];
76
77 struct net_device *fb_tunnel_dev;
78 };
79
80 static DEFINE_RWLOCK(ipip6_lock);
81
82 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
83 struct net_device *dev, __be32 remote, __be32 local)
84 {
85 unsigned h0 = HASH(remote);
86 unsigned h1 = HASH(local);
87 struct ip_tunnel *t;
88 struct sit_net *sitn = net_generic(net, sit_net_id);
89
90 for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
91 if (local == t->parms.iph.saddr &&
92 remote == t->parms.iph.daddr &&
93 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
94 (t->dev->flags & IFF_UP))
95 return t;
96 }
97 for (t = sitn->tunnels_r[h0]; t; t = t->next) {
98 if (remote == t->parms.iph.daddr &&
99 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
100 (t->dev->flags & IFF_UP))
101 return t;
102 }
103 for (t = sitn->tunnels_l[h1]; t; t = t->next) {
104 if (local == t->parms.iph.saddr &&
105 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
106 (t->dev->flags & IFF_UP))
107 return t;
108 }
109 t = sitn->tunnels_wc[0];
110 if ((t != NULL) && (t->dev->flags & IFF_UP))
111 return t;
112 return NULL;
113 }
114
115 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
116 struct ip_tunnel_parm *parms)
117 {
118 __be32 remote = parms->iph.daddr;
119 __be32 local = parms->iph.saddr;
120 unsigned h = 0;
121 int prio = 0;
122
123 if (remote) {
124 prio |= 2;
125 h ^= HASH(remote);
126 }
127 if (local) {
128 prio |= 1;
129 h ^= HASH(local);
130 }
131 return &sitn->tunnels[prio][h];
132 }
133
134 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
135 struct ip_tunnel *t)
136 {
137 return __ipip6_bucket(sitn, &t->parms);
138 }
139
140 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
141 {
142 struct ip_tunnel **tp;
143
144 for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
145 if (t == *tp) {
146 write_lock_bh(&ipip6_lock);
147 *tp = t->next;
148 write_unlock_bh(&ipip6_lock);
149 break;
150 }
151 }
152 }
153
154 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
155 {
156 struct ip_tunnel **tp = ipip6_bucket(sitn, t);
157
158 t->next = *tp;
159 write_lock_bh(&ipip6_lock);
160 *tp = t;
161 write_unlock_bh(&ipip6_lock);
162 }
163
164 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
165 struct ip_tunnel_parm *parms, int create)
166 {
167 __be32 remote = parms->iph.daddr;
168 __be32 local = parms->iph.saddr;
169 struct ip_tunnel *t, **tp, *nt;
170 struct net_device *dev;
171 char name[IFNAMSIZ];
172 struct sit_net *sitn = net_generic(net, sit_net_id);
173
174 for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
175 if (local == t->parms.iph.saddr &&
176 remote == t->parms.iph.daddr &&
177 parms->link == t->parms.link) {
178 if (create)
179 return NULL;
180 else
181 return t;
182 }
183 }
184 if (!create)
185 goto failed;
186
187 if (parms->name[0])
188 strlcpy(name, parms->name, IFNAMSIZ);
189 else
190 sprintf(name, "sit%%d");
191
192 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
193 if (dev == NULL)
194 return NULL;
195
196 dev_net_set(dev, net);
197
198 if (strchr(name, '%')) {
199 if (dev_alloc_name(dev, name) < 0)
200 goto failed_free;
201 }
202
203 nt = netdev_priv(dev);
204
205 nt->parms = *parms;
206 ipip6_tunnel_init(dev);
207
208 if (parms->i_flags & SIT_ISATAP)
209 dev->priv_flags |= IFF_ISATAP;
210
211 if (register_netdevice(dev) < 0)
212 goto failed_free;
213
214 dev_hold(dev);
215
216 ipip6_tunnel_link(sitn, nt);
217 return nt;
218
219 failed_free:
220 free_netdev(dev);
221 failed:
222 return NULL;
223 }
224
225 static struct ip_tunnel_prl_entry *
226 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
227 {
228 struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
229
230 for (p = t->prl; p; p = p->next)
231 if (p->addr == addr)
232 break;
233 return p;
234
235 }
236
237 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
238 struct ip_tunnel_prl __user *a)
239 {
240 struct ip_tunnel_prl kprl, *kp;
241 struct ip_tunnel_prl_entry *prl;
242 unsigned int cmax, c = 0, ca, len;
243 int ret = 0;
244
245 if (copy_from_user(&kprl, a, sizeof(kprl)))
246 return -EFAULT;
247 cmax = kprl.datalen / sizeof(kprl);
248 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
249 cmax = 1;
250
251 /* For simple GET or for root users,
252 * we try harder to allocate.
253 */
254 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
255 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
256 NULL;
257
258 read_lock(&ipip6_lock);
259
260 ca = t->prl_count < cmax ? t->prl_count : cmax;
261
262 if (!kp) {
263 /* We don't try hard to allocate much memory for
264 * non-root users.
265 * For root users, retry allocating enough memory for
266 * the answer.
267 */
268 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
269 if (!kp) {
270 ret = -ENOMEM;
271 goto out;
272 }
273 }
274
275 c = 0;
276 for (prl = t->prl; prl; prl = prl->next) {
277 if (c >= cmax)
278 break;
279 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
280 continue;
281 kp[c].addr = prl->addr;
282 kp[c].flags = prl->flags;
283 c++;
284 if (kprl.addr != htonl(INADDR_ANY))
285 break;
286 }
287 out:
288 read_unlock(&ipip6_lock);
289
290 len = sizeof(*kp) * c;
291 ret = 0;
292 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
293 ret = -EFAULT;
294
295 kfree(kp);
296
297 return ret;
298 }
299
300 static int
301 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
302 {
303 struct ip_tunnel_prl_entry *p;
304 int err = 0;
305
306 if (a->addr == htonl(INADDR_ANY))
307 return -EINVAL;
308
309 write_lock(&ipip6_lock);
310
311 for (p = t->prl; p; p = p->next) {
312 if (p->addr == a->addr) {
313 if (chg)
314 goto update;
315 err = -EEXIST;
316 goto out;
317 }
318 }
319
320 if (chg) {
321 err = -ENXIO;
322 goto out;
323 }
324
325 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
326 if (!p) {
327 err = -ENOBUFS;
328 goto out;
329 }
330
331 p->next = t->prl;
332 t->prl = p;
333 t->prl_count++;
334 update:
335 p->addr = a->addr;
336 p->flags = a->flags;
337 out:
338 write_unlock(&ipip6_lock);
339 return err;
340 }
341
342 static int
343 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
344 {
345 struct ip_tunnel_prl_entry *x, **p;
346 int err = 0;
347
348 write_lock(&ipip6_lock);
349
350 if (a && a->addr != htonl(INADDR_ANY)) {
351 for (p = &t->prl; *p; p = &(*p)->next) {
352 if ((*p)->addr == a->addr) {
353 x = *p;
354 *p = x->next;
355 kfree(x);
356 t->prl_count--;
357 goto out;
358 }
359 }
360 err = -ENXIO;
361 } else {
362 while (t->prl) {
363 x = t->prl;
364 t->prl = t->prl->next;
365 kfree(x);
366 t->prl_count--;
367 }
368 }
369 out:
370 write_unlock(&ipip6_lock);
371 return err;
372 }
373
374 static int
375 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
376 {
377 struct ip_tunnel_prl_entry *p;
378 int ok = 1;
379
380 read_lock(&ipip6_lock);
381 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
382 if (p) {
383 if (p->flags & PRL_DEFAULT)
384 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
385 else
386 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
387 } else {
388 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
389 if (ipv6_addr_is_isatap(addr6) &&
390 (addr6->s6_addr32[3] == iph->saddr) &&
391 ipv6_chk_prefix(addr6, t->dev))
392 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
393 else
394 ok = 0;
395 }
396 read_unlock(&ipip6_lock);
397 return ok;
398 }
399
400 static void ipip6_tunnel_uninit(struct net_device *dev)
401 {
402 struct net *net = dev_net(dev);
403 struct sit_net *sitn = net_generic(net, sit_net_id);
404
405 if (dev == sitn->fb_tunnel_dev) {
406 write_lock_bh(&ipip6_lock);
407 sitn->tunnels_wc[0] = NULL;
408 write_unlock_bh(&ipip6_lock);
409 dev_put(dev);
410 } else {
411 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
412 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
413 dev_put(dev);
414 }
415 }
416
417
418 static int ipip6_err(struct sk_buff *skb, u32 info)
419 {
420
421 /* All the routers (except for Linux) return only
422 8 bytes of packet payload. It means, that precise relaying of
423 ICMP in the real Internet is absolutely infeasible.
424 */
425 struct iphdr *iph = (struct iphdr*)skb->data;
426 const int type = icmp_hdr(skb)->type;
427 const int code = icmp_hdr(skb)->code;
428 struct ip_tunnel *t;
429 int err;
430
431 switch (type) {
432 default:
433 case ICMP_PARAMETERPROB:
434 return 0;
435
436 case ICMP_DEST_UNREACH:
437 switch (code) {
438 case ICMP_SR_FAILED:
439 case ICMP_PORT_UNREACH:
440 /* Impossible event. */
441 return 0;
442 case ICMP_FRAG_NEEDED:
443 /* Soft state for pmtu is maintained by IP core. */
444 return 0;
445 default:
446 /* All others are translated to HOST_UNREACH.
447 rfc2003 contains "deep thoughts" about NET_UNREACH,
448 I believe they are just ether pollution. --ANK
449 */
450 break;
451 }
452 break;
453 case ICMP_TIME_EXCEEDED:
454 if (code != ICMP_EXC_TTL)
455 return 0;
456 break;
457 }
458
459 err = -ENOENT;
460
461 read_lock(&ipip6_lock);
462 t = ipip6_tunnel_lookup(dev_net(skb->dev),
463 skb->dev,
464 iph->daddr,
465 iph->saddr);
466 if (t == NULL || t->parms.iph.daddr == 0)
467 goto out;
468
469 err = 0;
470 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
471 goto out;
472
473 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
474 t->err_count++;
475 else
476 t->err_count = 1;
477 t->err_time = jiffies;
478 out:
479 read_unlock(&ipip6_lock);
480 return err;
481 }
482
483 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
484 {
485 if (INET_ECN_is_ce(iph->tos))
486 IP6_ECN_set_ce(ipv6_hdr(skb));
487 }
488
489 static int ipip6_rcv(struct sk_buff *skb)
490 {
491 struct iphdr *iph;
492 struct ip_tunnel *tunnel;
493
494 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
495 goto out;
496
497 iph = ip_hdr(skb);
498
499 read_lock(&ipip6_lock);
500 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
501 iph->saddr, iph->daddr);
502 if (tunnel != NULL) {
503 secpath_reset(skb);
504 skb->mac_header = skb->network_header;
505 skb_reset_network_header(skb);
506 IPCB(skb)->flags = 0;
507 skb->protocol = htons(ETH_P_IPV6);
508 skb->pkt_type = PACKET_HOST;
509
510 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
511 !isatap_chksrc(skb, iph, tunnel)) {
512 tunnel->dev->stats.rx_errors++;
513 read_unlock(&ipip6_lock);
514 kfree_skb(skb);
515 return 0;
516 }
517 tunnel->dev->stats.rx_packets++;
518 tunnel->dev->stats.rx_bytes += skb->len;
519 skb->dev = tunnel->dev;
520 skb_dst_drop(skb);
521 nf_reset(skb);
522 ipip6_ecn_decapsulate(iph, skb);
523 netif_rx(skb);
524 read_unlock(&ipip6_lock);
525 return 0;
526 }
527
528 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
529 read_unlock(&ipip6_lock);
530 out:
531 kfree_skb(skb);
532 return 0;
533 }
534
535 /* Returns the embedded IPv4 address if the IPv6 address
536 comes from 6to4 (RFC 3056) addr space */
537
538 static inline __be32 try_6to4(struct in6_addr *v6dst)
539 {
540 __be32 dst = 0;
541
542 if (v6dst->s6_addr16[0] == htons(0x2002)) {
543 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
544 memcpy(&dst, &v6dst->s6_addr16[1], 4);
545 }
546 return dst;
547 }
548
549 /*
550 * This function assumes it is being called from dev_queue_xmit()
551 * and that skb is filled properly by that function.
552 */
553
554 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
555 struct net_device *dev)
556 {
557 struct ip_tunnel *tunnel = netdev_priv(dev);
558 struct net_device_stats *stats = &dev->stats;
559 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
560 struct iphdr *tiph = &tunnel->parms.iph;
561 struct ipv6hdr *iph6 = ipv6_hdr(skb);
562 u8 tos = tunnel->parms.iph.tos;
563 struct rtable *rt; /* Route to the other host */
564 struct net_device *tdev; /* Device to other host */
565 struct iphdr *iph; /* Our new IP header */
566 unsigned int max_headroom; /* The extra header space needed */
567 __be32 dst = tiph->daddr;
568 int mtu;
569 struct in6_addr *addr6;
570 int addr_type;
571
572 if (skb->protocol != htons(ETH_P_IPV6))
573 goto tx_error;
574
575 /* ISATAP (RFC4214) - must come before 6to4 */
576 if (dev->priv_flags & IFF_ISATAP) {
577 struct neighbour *neigh = NULL;
578
579 if (skb_dst(skb))
580 neigh = skb_dst(skb)->neighbour;
581
582 if (neigh == NULL) {
583 if (net_ratelimit())
584 printk(KERN_DEBUG "sit: nexthop == NULL\n");
585 goto tx_error;
586 }
587
588 addr6 = (struct in6_addr*)&neigh->primary_key;
589 addr_type = ipv6_addr_type(addr6);
590
591 if ((addr_type & IPV6_ADDR_UNICAST) &&
592 ipv6_addr_is_isatap(addr6))
593 dst = addr6->s6_addr32[3];
594 else
595 goto tx_error;
596 }
597
598 if (!dst)
599 dst = try_6to4(&iph6->daddr);
600
601 if (!dst) {
602 struct neighbour *neigh = NULL;
603
604 if (skb_dst(skb))
605 neigh = skb_dst(skb)->neighbour;
606
607 if (neigh == NULL) {
608 if (net_ratelimit())
609 printk(KERN_DEBUG "sit: nexthop == NULL\n");
610 goto tx_error;
611 }
612
613 addr6 = (struct in6_addr*)&neigh->primary_key;
614 addr_type = ipv6_addr_type(addr6);
615
616 if (addr_type == IPV6_ADDR_ANY) {
617 addr6 = &ipv6_hdr(skb)->daddr;
618 addr_type = ipv6_addr_type(addr6);
619 }
620
621 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
622 goto tx_error_icmp;
623
624 dst = addr6->s6_addr32[3];
625 }
626
627 {
628 struct flowi fl = { .nl_u = { .ip4_u =
629 { .daddr = dst,
630 .saddr = tiph->saddr,
631 .tos = RT_TOS(tos) } },
632 .oif = tunnel->parms.link,
633 .proto = IPPROTO_IPV6 };
634 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
635 stats->tx_carrier_errors++;
636 goto tx_error_icmp;
637 }
638 }
639 if (rt->rt_type != RTN_UNICAST) {
640 ip_rt_put(rt);
641 stats->tx_carrier_errors++;
642 goto tx_error_icmp;
643 }
644 tdev = rt->u.dst.dev;
645
646 if (tdev == dev) {
647 ip_rt_put(rt);
648 stats->collisions++;
649 goto tx_error;
650 }
651
652 if (tiph->frag_off)
653 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
654 else
655 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
656
657 if (mtu < 68) {
658 stats->collisions++;
659 ip_rt_put(rt);
660 goto tx_error;
661 }
662 if (mtu < IPV6_MIN_MTU)
663 mtu = IPV6_MIN_MTU;
664 if (tunnel->parms.iph.daddr && skb_dst(skb))
665 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
666
667 if (skb->len > mtu) {
668 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
669 ip_rt_put(rt);
670 goto tx_error;
671 }
672
673 if (tunnel->err_count > 0) {
674 if (time_before(jiffies,
675 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
676 tunnel->err_count--;
677 dst_link_failure(skb);
678 } else
679 tunnel->err_count = 0;
680 }
681
682 /*
683 * Okay, now see if we can stuff it in the buffer as-is.
684 */
685 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
686
687 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
688 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
689 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
690 if (!new_skb) {
691 ip_rt_put(rt);
692 txq->tx_dropped++;
693 dev_kfree_skb(skb);
694 return NETDEV_TX_OK;
695 }
696 if (skb->sk)
697 skb_set_owner_w(new_skb, skb->sk);
698 dev_kfree_skb(skb);
699 skb = new_skb;
700 iph6 = ipv6_hdr(skb);
701 }
702
703 skb->transport_header = skb->network_header;
704 skb_push(skb, sizeof(struct iphdr));
705 skb_reset_network_header(skb);
706 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
707 IPCB(skb)->flags = 0;
708 skb_dst_drop(skb);
709 skb_dst_set(skb, &rt->u.dst);
710
711 /*
712 * Push down and install the IPIP header.
713 */
714
715 iph = ip_hdr(skb);
716 iph->version = 4;
717 iph->ihl = sizeof(struct iphdr)>>2;
718 if (mtu > IPV6_MIN_MTU)
719 iph->frag_off = tiph->frag_off;
720 else
721 iph->frag_off = 0;
722
723 iph->protocol = IPPROTO_IPV6;
724 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
725 iph->daddr = rt->rt_dst;
726 iph->saddr = rt->rt_src;
727
728 if ((iph->ttl = tiph->ttl) == 0)
729 iph->ttl = iph6->hop_limit;
730
731 nf_reset(skb);
732
733 IPTUNNEL_XMIT();
734 return NETDEV_TX_OK;
735
736 tx_error_icmp:
737 dst_link_failure(skb);
738 tx_error:
739 stats->tx_errors++;
740 dev_kfree_skb(skb);
741 return NETDEV_TX_OK;
742 }
743
744 static void ipip6_tunnel_bind_dev(struct net_device *dev)
745 {
746 struct net_device *tdev = NULL;
747 struct ip_tunnel *tunnel;
748 struct iphdr *iph;
749
750 tunnel = netdev_priv(dev);
751 iph = &tunnel->parms.iph;
752
753 if (iph->daddr) {
754 struct flowi fl = { .nl_u = { .ip4_u =
755 { .daddr = iph->daddr,
756 .saddr = iph->saddr,
757 .tos = RT_TOS(iph->tos) } },
758 .oif = tunnel->parms.link,
759 .proto = IPPROTO_IPV6 };
760 struct rtable *rt;
761 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
762 tdev = rt->u.dst.dev;
763 ip_rt_put(rt);
764 }
765 dev->flags |= IFF_POINTOPOINT;
766 }
767
768 if (!tdev && tunnel->parms.link)
769 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
770
771 if (tdev) {
772 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
773 dev->mtu = tdev->mtu - sizeof(struct iphdr);
774 if (dev->mtu < IPV6_MIN_MTU)
775 dev->mtu = IPV6_MIN_MTU;
776 }
777 dev->iflink = tunnel->parms.link;
778 }
779
780 static int
781 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
782 {
783 int err = 0;
784 struct ip_tunnel_parm p;
785 struct ip_tunnel_prl prl;
786 struct ip_tunnel *t;
787 struct net *net = dev_net(dev);
788 struct sit_net *sitn = net_generic(net, sit_net_id);
789
790 switch (cmd) {
791 case SIOCGETTUNNEL:
792 t = NULL;
793 if (dev == sitn->fb_tunnel_dev) {
794 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
795 err = -EFAULT;
796 break;
797 }
798 t = ipip6_tunnel_locate(net, &p, 0);
799 }
800 if (t == NULL)
801 t = netdev_priv(dev);
802 memcpy(&p, &t->parms, sizeof(p));
803 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
804 err = -EFAULT;
805 break;
806
807 case SIOCADDTUNNEL:
808 case SIOCCHGTUNNEL:
809 err = -EPERM;
810 if (!capable(CAP_NET_ADMIN))
811 goto done;
812
813 err = -EFAULT;
814 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
815 goto done;
816
817 err = -EINVAL;
818 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
819 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
820 goto done;
821 if (p.iph.ttl)
822 p.iph.frag_off |= htons(IP_DF);
823
824 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
825
826 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
827 if (t != NULL) {
828 if (t->dev != dev) {
829 err = -EEXIST;
830 break;
831 }
832 } else {
833 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
834 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
835 err = -EINVAL;
836 break;
837 }
838 t = netdev_priv(dev);
839 ipip6_tunnel_unlink(sitn, t);
840 t->parms.iph.saddr = p.iph.saddr;
841 t->parms.iph.daddr = p.iph.daddr;
842 memcpy(dev->dev_addr, &p.iph.saddr, 4);
843 memcpy(dev->broadcast, &p.iph.daddr, 4);
844 ipip6_tunnel_link(sitn, t);
845 netdev_state_change(dev);
846 }
847 }
848
849 if (t) {
850 err = 0;
851 if (cmd == SIOCCHGTUNNEL) {
852 t->parms.iph.ttl = p.iph.ttl;
853 t->parms.iph.tos = p.iph.tos;
854 if (t->parms.link != p.link) {
855 t->parms.link = p.link;
856 ipip6_tunnel_bind_dev(dev);
857 netdev_state_change(dev);
858 }
859 }
860 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
861 err = -EFAULT;
862 } else
863 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
864 break;
865
866 case SIOCDELTUNNEL:
867 err = -EPERM;
868 if (!capable(CAP_NET_ADMIN))
869 goto done;
870
871 if (dev == sitn->fb_tunnel_dev) {
872 err = -EFAULT;
873 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
874 goto done;
875 err = -ENOENT;
876 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
877 goto done;
878 err = -EPERM;
879 if (t == netdev_priv(sitn->fb_tunnel_dev))
880 goto done;
881 dev = t->dev;
882 }
883 unregister_netdevice(dev);
884 err = 0;
885 break;
886
887 case SIOCGETPRL:
888 err = -EINVAL;
889 if (dev == sitn->fb_tunnel_dev)
890 goto done;
891 err = -ENOENT;
892 if (!(t = netdev_priv(dev)))
893 goto done;
894 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
895 break;
896
897 case SIOCADDPRL:
898 case SIOCDELPRL:
899 case SIOCCHGPRL:
900 err = -EPERM;
901 if (!capable(CAP_NET_ADMIN))
902 goto done;
903 err = -EINVAL;
904 if (dev == sitn->fb_tunnel_dev)
905 goto done;
906 err = -EFAULT;
907 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
908 goto done;
909 err = -ENOENT;
910 if (!(t = netdev_priv(dev)))
911 goto done;
912
913 switch (cmd) {
914 case SIOCDELPRL:
915 err = ipip6_tunnel_del_prl(t, &prl);
916 break;
917 case SIOCADDPRL:
918 case SIOCCHGPRL:
919 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
920 break;
921 }
922 netdev_state_change(dev);
923 break;
924
925 default:
926 err = -EINVAL;
927 }
928
929 done:
930 return err;
931 }
932
933 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
934 {
935 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
936 return -EINVAL;
937 dev->mtu = new_mtu;
938 return 0;
939 }
940
941 static const struct net_device_ops ipip6_netdev_ops = {
942 .ndo_uninit = ipip6_tunnel_uninit,
943 .ndo_start_xmit = ipip6_tunnel_xmit,
944 .ndo_do_ioctl = ipip6_tunnel_ioctl,
945 .ndo_change_mtu = ipip6_tunnel_change_mtu,
946 };
947
948 static void ipip6_tunnel_setup(struct net_device *dev)
949 {
950 dev->netdev_ops = &ipip6_netdev_ops;
951 dev->destructor = free_netdev;
952
953 dev->type = ARPHRD_SIT;
954 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
955 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
956 dev->flags = IFF_NOARP;
957 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
958 dev->iflink = 0;
959 dev->addr_len = 4;
960 dev->features |= NETIF_F_NETNS_LOCAL;
961 }
962
963 static void ipip6_tunnel_init(struct net_device *dev)
964 {
965 struct ip_tunnel *tunnel = netdev_priv(dev);
966
967 tunnel->dev = dev;
968 strcpy(tunnel->parms.name, dev->name);
969
970 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
971 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
972
973 ipip6_tunnel_bind_dev(dev);
974 }
975
976 static void ipip6_fb_tunnel_init(struct net_device *dev)
977 {
978 struct ip_tunnel *tunnel = netdev_priv(dev);
979 struct iphdr *iph = &tunnel->parms.iph;
980 struct net *net = dev_net(dev);
981 struct sit_net *sitn = net_generic(net, sit_net_id);
982
983 tunnel->dev = dev;
984 strcpy(tunnel->parms.name, dev->name);
985
986 iph->version = 4;
987 iph->protocol = IPPROTO_IPV6;
988 iph->ihl = 5;
989 iph->ttl = 64;
990
991 dev_hold(dev);
992 sitn->tunnels_wc[0] = tunnel;
993 }
994
995 static struct xfrm_tunnel sit_handler = {
996 .handler = ipip6_rcv,
997 .err_handler = ipip6_err,
998 .priority = 1,
999 };
1000
1001 static void sit_destroy_tunnels(struct sit_net *sitn)
1002 {
1003 int prio;
1004
1005 for (prio = 1; prio < 4; prio++) {
1006 int h;
1007 for (h = 0; h < HASH_SIZE; h++) {
1008 struct ip_tunnel *t;
1009 while ((t = sitn->tunnels[prio][h]) != NULL)
1010 unregister_netdevice(t->dev);
1011 }
1012 }
1013 }
1014
1015 static int sit_init_net(struct net *net)
1016 {
1017 int err;
1018 struct sit_net *sitn;
1019
1020 err = -ENOMEM;
1021 sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1022 if (sitn == NULL)
1023 goto err_alloc;
1024
1025 err = net_assign_generic(net, sit_net_id, sitn);
1026 if (err < 0)
1027 goto err_assign;
1028
1029 sitn->tunnels[0] = sitn->tunnels_wc;
1030 sitn->tunnels[1] = sitn->tunnels_l;
1031 sitn->tunnels[2] = sitn->tunnels_r;
1032 sitn->tunnels[3] = sitn->tunnels_r_l;
1033
1034 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1035 ipip6_tunnel_setup);
1036 if (!sitn->fb_tunnel_dev) {
1037 err = -ENOMEM;
1038 goto err_alloc_dev;
1039 }
1040 dev_net_set(sitn->fb_tunnel_dev, net);
1041
1042 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1043
1044 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1045 goto err_reg_dev;
1046
1047 return 0;
1048
1049 err_reg_dev:
1050 dev_put(sitn->fb_tunnel_dev);
1051 free_netdev(sitn->fb_tunnel_dev);
1052 err_alloc_dev:
1053 /* nothing */
1054 err_assign:
1055 kfree(sitn);
1056 err_alloc:
1057 return err;
1058 }
1059
1060 static void sit_exit_net(struct net *net)
1061 {
1062 struct sit_net *sitn;
1063
1064 sitn = net_generic(net, sit_net_id);
1065 rtnl_lock();
1066 sit_destroy_tunnels(sitn);
1067 unregister_netdevice(sitn->fb_tunnel_dev);
1068 rtnl_unlock();
1069 kfree(sitn);
1070 }
1071
1072 static struct pernet_operations sit_net_ops = {
1073 .init = sit_init_net,
1074 .exit = sit_exit_net,
1075 };
1076
1077 static void __exit sit_cleanup(void)
1078 {
1079 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1080
1081 unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1082 }
1083
1084 static int __init sit_init(void)
1085 {
1086 int err;
1087
1088 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1089
1090 if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1091 printk(KERN_INFO "sit init: Can't add protocol\n");
1092 return -EAGAIN;
1093 }
1094
1095 err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1096 if (err < 0)
1097 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1098
1099 return err;
1100 }
1101
1102 module_init(sit_init);
1103 module_exit(sit_cleanup);
1104 MODULE_LICENSE("GPL");
1105 MODULE_ALIAS("sit0");
Ubuntu Artful kernel mirror