Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 *
10 * Changes:
11 * Roger Venning <r.venning@telstra.com>: 6to4 support
12 * Nate Thompson <nate@thebog.net>: 6to4 support
13 * Fred Templin <fred.l.templin@boeing.com>: isatap support
14 */
15
16#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17
18#include <linux/module.h>
19#include <linux/capability.h>
20#include <linux/errno.h>
21#include <linux/types.h>
22#include <linux/socket.h>
23#include <linux/sockios.h>
24#include <linux/net.h>
25#include <linux/in6.h>
26#include <linux/netdevice.h>
27#include <linux/if_arp.h>
28#include <linux/icmp.h>
29#include <linux/slab.h>
30#include <linux/uaccess.h>
31#include <linux/init.h>
32#include <linux/netfilter_ipv4.h>
33#include <linux/if_ether.h>
34
35#include <net/sock.h>
36#include <net/snmp.h>
37
38#include <net/ipv6.h>
39#include <net/protocol.h>
40#include <net/transp_v6.h>
41#include <net/ip6_fib.h>
42#include <net/ip6_route.h>
43#include <net/ndisc.h>
44#include <net/addrconf.h>
45#include <net/ip.h>
46#include <net/udp.h>
47#include <net/icmp.h>
48#include <net/ip_tunnels.h>
49#include <net/inet_ecn.h>
50#include <net/xfrm.h>
51#include <net/dsfield.h>
52#include <net/net_namespace.h>
53#include <net/netns/generic.h>
54
55/*
56 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
57
58 For comments look at net/ipv4/ip_gre.c --ANK
59 */
60
61#define IP6_SIT_HASH_SIZE 16
62#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
63
64static bool log_ecn_error = true;
65module_param(log_ecn_error, bool, 0644);
66MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
67
68static int ipip6_tunnel_init(struct net_device *dev);
69static void ipip6_tunnel_setup(struct net_device *dev);
70static void ipip6_dev_free(struct net_device *dev);
71static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
72 __be32 *v4dst);
73static struct rtnl_link_ops sit_link_ops __read_mostly;
74
75static unsigned int sit_net_id __read_mostly;
76struct sit_net {
77 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
78 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
79 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
80 struct ip_tunnel __rcu *tunnels_wc[1];
81 struct ip_tunnel __rcu **tunnels[4];
82
83 struct net_device *fb_tunnel_dev;
84};
85
86static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
87{
88 struct ip_tunnel *t = netdev_priv(dev);
89
90 return net_generic(t->net, sit_net_id);
91}
92
93/*
94 * Must be invoked with rcu_read_lock
95 */
96static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
97 struct net_device *dev,
98 __be32 remote, __be32 local,
99 int sifindex)
100{
101 unsigned int h0 = HASH(remote);
102 unsigned int h1 = HASH(local);
103 struct ip_tunnel *t;
104 struct sit_net *sitn = net_generic(net, sit_net_id);
105 int ifindex = dev ? dev->ifindex : 0;
106
107 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
108 if (local == t->parms.iph.saddr &&
109 remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || ifindex == t->parms.link ||
111 sifindex == t->parms.link) &&
112 (t->dev->flags & IFF_UP))
113 return t;
114 }
115 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
116 if (remote == t->parms.iph.daddr &&
117 (!dev || !t->parms.link || ifindex == t->parms.link ||
118 sifindex == t->parms.link) &&
119 (t->dev->flags & IFF_UP))
120 return t;
121 }
122 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
123 if (local == t->parms.iph.saddr &&
124 (!dev || !t->parms.link || ifindex == t->parms.link ||
125 sifindex == t->parms.link) &&
126 (t->dev->flags & IFF_UP))
127 return t;
128 }
129 t = rcu_dereference(sitn->tunnels_wc[0]);
130 if (t && (t->dev->flags & IFF_UP))
131 return t;
132 return NULL;
133}
134
135static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
136 struct ip_tunnel_parm *parms)
137{
138 __be32 remote = parms->iph.daddr;
139 __be32 local = parms->iph.saddr;
140 unsigned int h = 0;
141 int prio = 0;
142
143 if (remote) {
144 prio |= 2;
145 h ^= HASH(remote);
146 }
147 if (local) {
148 prio |= 1;
149 h ^= HASH(local);
150 }
151 return &sitn->tunnels[prio][h];
152}
153
154static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
155 struct ip_tunnel *t)
156{
157 return __ipip6_bucket(sitn, &t->parms);
158}
159
160static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
161{
162 struct ip_tunnel __rcu **tp;
163 struct ip_tunnel *iter;
164
165 for (tp = ipip6_bucket(sitn, t);
166 (iter = rtnl_dereference(*tp)) != NULL;
167 tp = &iter->next) {
168 if (t == iter) {
169 rcu_assign_pointer(*tp, t->next);
170 break;
171 }
172 }
173}
174
175static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
176{
177 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
178
179 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
180 rcu_assign_pointer(*tp, t);
181}
182
183static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
184{
185#ifdef CONFIG_IPV6_SIT_6RD
186 struct ip_tunnel *t = netdev_priv(dev);
187
188 if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
189 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
190 t->ip6rd.relay_prefix = 0;
191 t->ip6rd.prefixlen = 16;
192 t->ip6rd.relay_prefixlen = 0;
193 } else {
194 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
195 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
196 }
197#endif
198}
199
200static int ipip6_tunnel_create(struct net_device *dev)
201{
202 struct ip_tunnel *t = netdev_priv(dev);
203 struct net *net = dev_net(dev);
204 struct sit_net *sitn = net_generic(net, sit_net_id);
205 int err;
206
207 __dev_addr_set(dev, &t->parms.iph.saddr, 4);
208 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
209
210 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
211 dev->priv_flags |= IFF_ISATAP;
212
213 dev->rtnl_link_ops = &sit_link_ops;
214
215 err = register_netdevice(dev);
216 if (err < 0)
217 goto out;
218
219 ipip6_tunnel_clone_6rd(dev, sitn);
220
221 ipip6_tunnel_link(sitn, t);
222 return 0;
223
224out:
225 return err;
226}
227
228static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
229 struct ip_tunnel_parm *parms, int create)
230{
231 __be32 remote = parms->iph.daddr;
232 __be32 local = parms->iph.saddr;
233 struct ip_tunnel *t, *nt;
234 struct ip_tunnel __rcu **tp;
235 struct net_device *dev;
236 char name[IFNAMSIZ];
237 struct sit_net *sitn = net_generic(net, sit_net_id);
238
239 for (tp = __ipip6_bucket(sitn, parms);
240 (t = rtnl_dereference(*tp)) != NULL;
241 tp = &t->next) {
242 if (local == t->parms.iph.saddr &&
243 remote == t->parms.iph.daddr &&
244 parms->link == t->parms.link) {
245 if (create)
246 return NULL;
247 else
248 return t;
249 }
250 }
251 if (!create)
252 goto failed;
253
254 if (parms->name[0]) {
255 if (!dev_valid_name(parms->name))
256 goto failed;
257 strscpy(name, parms->name, IFNAMSIZ);
258 } else {
259 strcpy(name, "sit%d");
260 }
261 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
262 ipip6_tunnel_setup);
263 if (!dev)
264 return NULL;
265
266 dev_net_set(dev, net);
267
268 nt = netdev_priv(dev);
269
270 nt->parms = *parms;
271 if (ipip6_tunnel_create(dev) < 0)
272 goto failed_free;
273
274 if (!parms->name[0])
275 strcpy(parms->name, dev->name);
276
277 return nt;
278
279failed_free:
280 free_netdev(dev);
281failed:
282 return NULL;
283}
284
285#define for_each_prl_rcu(start) \
286 for (prl = rcu_dereference(start); \
287 prl; \
288 prl = rcu_dereference(prl->next))
289
290static struct ip_tunnel_prl_entry *
291__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
292{
293 struct ip_tunnel_prl_entry *prl;
294
295 for_each_prl_rcu(t->prl)
296 if (prl->addr == addr)
297 break;
298 return prl;
299
300}
301
302static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
303{
304 struct ip_tunnel *t = netdev_priv(dev);
305 struct ip_tunnel_prl kprl, *kp;
306 struct ip_tunnel_prl_entry *prl;
307 unsigned int cmax, c = 0, ca, len;
308 int ret = 0;
309
310 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
311 return -EINVAL;
312
313 if (copy_from_user(&kprl, a, sizeof(kprl)))
314 return -EFAULT;
315 cmax = kprl.datalen / sizeof(kprl);
316 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
317 cmax = 1;
318
319 /* For simple GET or for root users,
320 * we try harder to allocate.
321 */
322 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
323 kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
324 NULL;
325
326 ca = min(t->prl_count, cmax);
327
328 if (!kp) {
329 /* We don't try hard to allocate much memory for
330 * non-root users.
331 * For root users, retry allocating enough memory for
332 * the answer.
333 */
334 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
335 __GFP_NOWARN);
336 if (!kp) {
337 ret = -ENOMEM;
338 goto out;
339 }
340 }
341
342 rcu_read_lock();
343 for_each_prl_rcu(t->prl) {
344 if (c >= cmax)
345 break;
346 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
347 continue;
348 kp[c].addr = prl->addr;
349 kp[c].flags = prl->flags;
350 c++;
351 if (kprl.addr != htonl(INADDR_ANY))
352 break;
353 }
354
355 rcu_read_unlock();
356
357 len = sizeof(*kp) * c;
358 ret = 0;
359 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
360 ret = -EFAULT;
361
362 kfree(kp);
363out:
364 return ret;
365}
366
367static int
368ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
369{
370 struct ip_tunnel_prl_entry *p;
371 int err = 0;
372
373 if (a->addr == htonl(INADDR_ANY))
374 return -EINVAL;
375
376 ASSERT_RTNL();
377
378 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
379 if (p->addr == a->addr) {
380 if (chg) {
381 p->flags = a->flags;
382 goto out;
383 }
384 err = -EEXIST;
385 goto out;
386 }
387 }
388
389 if (chg) {
390 err = -ENXIO;
391 goto out;
392 }
393
394 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
395 if (!p) {
396 err = -ENOBUFS;
397 goto out;
398 }
399
400 p->next = t->prl;
401 p->addr = a->addr;
402 p->flags = a->flags;
403 t->prl_count++;
404 rcu_assign_pointer(t->prl, p);
405out:
406 return err;
407}
408
409static void prl_list_destroy_rcu(struct rcu_head *head)
410{
411 struct ip_tunnel_prl_entry *p, *n;
412
413 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
414 do {
415 n = rcu_dereference_protected(p->next, 1);
416 kfree(p);
417 p = n;
418 } while (p);
419}
420
421static int
422ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
423{
424 struct ip_tunnel_prl_entry *x;
425 struct ip_tunnel_prl_entry __rcu **p;
426 int err = 0;
427
428 ASSERT_RTNL();
429
430 if (a && a->addr != htonl(INADDR_ANY)) {
431 for (p = &t->prl;
432 (x = rtnl_dereference(*p)) != NULL;
433 p = &x->next) {
434 if (x->addr == a->addr) {
435 *p = x->next;
436 kfree_rcu(x, rcu_head);
437 t->prl_count--;
438 goto out;
439 }
440 }
441 err = -ENXIO;
442 } else {
443 x = rtnl_dereference(t->prl);
444 if (x) {
445 t->prl_count = 0;
446 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
447 t->prl = NULL;
448 }
449 }
450out:
451 return err;
452}
453
454static int ipip6_tunnel_prl_ctl(struct net_device *dev,
455 struct ip_tunnel_prl __user *data, int cmd)
456{
457 struct ip_tunnel *t = netdev_priv(dev);
458 struct ip_tunnel_prl prl;
459 int err;
460
461 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
462 return -EPERM;
463 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
464 return -EINVAL;
465
466 if (copy_from_user(&prl, data, sizeof(prl)))
467 return -EFAULT;
468
469 switch (cmd) {
470 case SIOCDELPRL:
471 err = ipip6_tunnel_del_prl(t, &prl);
472 break;
473 case SIOCADDPRL:
474 case SIOCCHGPRL:
475 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
476 break;
477 }
478 dst_cache_reset(&t->dst_cache);
479 netdev_state_change(dev);
480 return err;
481}
482
483static int
484isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
485{
486 struct ip_tunnel_prl_entry *p;
487 int ok = 1;
488
489 rcu_read_lock();
490 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
491 if (p) {
492 if (p->flags & PRL_DEFAULT)
493 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
494 else
495 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
496 } else {
497 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
498
499 if (ipv6_addr_is_isatap(addr6) &&
500 (addr6->s6_addr32[3] == iph->saddr) &&
501 ipv6_chk_prefix(addr6, t->dev))
502 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
503 else
504 ok = 0;
505 }
506 rcu_read_unlock();
507 return ok;
508}
509
510static void ipip6_tunnel_uninit(struct net_device *dev)
511{
512 struct ip_tunnel *tunnel = netdev_priv(dev);
513 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
514
515 if (dev == sitn->fb_tunnel_dev) {
516 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
517 } else {
518 ipip6_tunnel_unlink(sitn, tunnel);
519 ipip6_tunnel_del_prl(tunnel, NULL);
520 }
521 dst_cache_reset(&tunnel->dst_cache);
522 netdev_put(dev, &tunnel->dev_tracker);
523}
524
525static int ipip6_err(struct sk_buff *skb, u32 info)
526{
527 const struct iphdr *iph = (const struct iphdr *)skb->data;
528 const int type = icmp_hdr(skb)->type;
529 const int code = icmp_hdr(skb)->code;
530 unsigned int data_len = 0;
531 struct ip_tunnel *t;
532 int sifindex;
533 int err;
534
535 switch (type) {
536 default:
537 case ICMP_PARAMETERPROB:
538 return 0;
539
540 case ICMP_DEST_UNREACH:
541 switch (code) {
542 case ICMP_SR_FAILED:
543 /* Impossible event. */
544 return 0;
545 default:
546 /* All others are translated to HOST_UNREACH.
547 rfc2003 contains "deep thoughts" about NET_UNREACH,
548 I believe they are just ether pollution. --ANK
549 */
550 break;
551 }
552 break;
553 case ICMP_TIME_EXCEEDED:
554 if (code != ICMP_EXC_TTL)
555 return 0;
556 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
557 break;
558 case ICMP_REDIRECT:
559 break;
560 }
561
562 err = -ENOENT;
563
564 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
565 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
566 iph->daddr, iph->saddr, sifindex);
567 if (!t)
568 goto out;
569
570 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
571 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
572 t->parms.link, iph->protocol);
573 err = 0;
574 goto out;
575 }
576 if (type == ICMP_REDIRECT) {
577 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
578 iph->protocol);
579 err = 0;
580 goto out;
581 }
582
583 err = 0;
584 if (__in6_dev_get(skb->dev) &&
585 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
586 goto out;
587
588 if (t->parms.iph.daddr == 0)
589 goto out;
590
591 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
592 goto out;
593
594 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
595 t->err_count++;
596 else
597 t->err_count = 1;
598 t->err_time = jiffies;
599out:
600 return err;
601}
602
603static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
604 const struct in6_addr *v6addr)
605{
606 __be32 v4embed = 0;
607 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
608 return true;
609 return false;
610}
611
612/* Checks if an address matches an address on the tunnel interface.
613 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
614 * Long story:
615 * This function is called after we considered the packet as spoofed
616 * in is_spoofed_6rd.
617 * We may have a router that is doing NAT for proto 41 packets
618 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
619 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
620 * function will return true, dropping the packet.
621 * But, we can still check if is spoofed against the IP
622 * addresses associated with the interface.
623 */
624static bool only_dnatted(const struct ip_tunnel *tunnel,
625 const struct in6_addr *v6dst)
626{
627 int prefix_len;
628
629#ifdef CONFIG_IPV6_SIT_6RD
630 prefix_len = tunnel->ip6rd.prefixlen + 32
631 - tunnel->ip6rd.relay_prefixlen;
632#else
633 prefix_len = 48;
634#endif
635 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
636}
637
638/* Returns true if a packet is spoofed */
639static bool packet_is_spoofed(struct sk_buff *skb,
640 const struct iphdr *iph,
641 struct ip_tunnel *tunnel)
642{
643 const struct ipv6hdr *ipv6h;
644
645 if (tunnel->dev->priv_flags & IFF_ISATAP) {
646 if (!isatap_chksrc(skb, iph, tunnel))
647 return true;
648
649 return false;
650 }
651
652 if (tunnel->dev->flags & IFF_POINTOPOINT)
653 return false;
654
655 ipv6h = ipv6_hdr(skb);
656
657 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
658 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
659 &iph->saddr, &ipv6h->saddr,
660 &iph->daddr, &ipv6h->daddr);
661 return true;
662 }
663
664 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
665 return false;
666
667 if (only_dnatted(tunnel, &ipv6h->daddr))
668 return false;
669
670 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
671 &iph->saddr, &ipv6h->saddr,
672 &iph->daddr, &ipv6h->daddr);
673 return true;
674}
675
676static int ipip6_rcv(struct sk_buff *skb)
677{
678 const struct iphdr *iph = ip_hdr(skb);
679 struct ip_tunnel *tunnel;
680 int sifindex;
681 int err;
682
683 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
684 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
685 iph->saddr, iph->daddr, sifindex);
686 if (tunnel) {
687 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
688 tunnel->parms.iph.protocol != 0)
689 goto out;
690
691 skb->mac_header = skb->network_header;
692 skb_reset_network_header(skb);
693 IPCB(skb)->flags = 0;
694 skb->dev = tunnel->dev;
695
696 if (packet_is_spoofed(skb, iph, tunnel)) {
697 DEV_STATS_INC(tunnel->dev, rx_errors);
698 goto out;
699 }
700
701 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
702 !net_eq(tunnel->net, dev_net(tunnel->dev))))
703 goto out;
704
705 /* skb can be uncloned in iptunnel_pull_header, so
706 * old iph is no longer valid
707 */
708 iph = (const struct iphdr *)skb_mac_header(skb);
709 skb_reset_mac_header(skb);
710
711 err = IP_ECN_decapsulate(iph, skb);
712 if (unlikely(err)) {
713 if (log_ecn_error)
714 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
715 &iph->saddr, iph->tos);
716 if (err > 1) {
717 DEV_STATS_INC(tunnel->dev, rx_frame_errors);
718 DEV_STATS_INC(tunnel->dev, rx_errors);
719 goto out;
720 }
721 }
722
723 dev_sw_netstats_rx_add(tunnel->dev, skb->len);
724
725 netif_rx(skb);
726
727 return 0;
728 }
729
730 /* no tunnel matched, let upstream know, ipsec may handle it */
731 return 1;
732out:
733 kfree_skb(skb);
734 return 0;
735}
736
737static const struct tnl_ptk_info ipip_tpi = {
738 /* no tunnel info required for ipip. */
739 .proto = htons(ETH_P_IP),
740};
741
742#if IS_ENABLED(CONFIG_MPLS)
743static const struct tnl_ptk_info mplsip_tpi = {
744 /* no tunnel info required for mplsip. */
745 .proto = htons(ETH_P_MPLS_UC),
746};
747#endif
748
749static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
750{
751 const struct iphdr *iph;
752 struct ip_tunnel *tunnel;
753 int sifindex;
754
755 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
756
757 iph = ip_hdr(skb);
758 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
759 iph->saddr, iph->daddr, sifindex);
760 if (tunnel) {
761 const struct tnl_ptk_info *tpi;
762
763 if (tunnel->parms.iph.protocol != ipproto &&
764 tunnel->parms.iph.protocol != 0)
765 goto drop;
766
767 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
768 goto drop;
769#if IS_ENABLED(CONFIG_MPLS)
770 if (ipproto == IPPROTO_MPLS)
771 tpi = &mplsip_tpi;
772 else
773#endif
774 tpi = &ipip_tpi;
775 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
776 goto drop;
777 skb_reset_mac_header(skb);
778
779 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
780 }
781
782 return 1;
783
784drop:
785 kfree_skb(skb);
786 return 0;
787}
788
789static int ipip_rcv(struct sk_buff *skb)
790{
791 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
792}
793
794#if IS_ENABLED(CONFIG_MPLS)
795static int mplsip_rcv(struct sk_buff *skb)
796{
797 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
798}
799#endif
800
801/*
802 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
803 * stores the embedded IPv4 address in v4dst and returns true.
804 */
805static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
806 __be32 *v4dst)
807{
808#ifdef CONFIG_IPV6_SIT_6RD
809 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
810 tunnel->ip6rd.prefixlen)) {
811 unsigned int pbw0, pbi0;
812 int pbi1;
813 u32 d;
814
815 pbw0 = tunnel->ip6rd.prefixlen >> 5;
816 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
817
818 d = tunnel->ip6rd.relay_prefixlen < 32 ?
819 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
820 tunnel->ip6rd.relay_prefixlen : 0;
821
822 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
823 if (pbi1 > 0)
824 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
825 (32 - pbi1);
826
827 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
828 return true;
829 }
830#else
831 if (v6dst->s6_addr16[0] == htons(0x2002)) {
832 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
833 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
834 return true;
835 }
836#endif
837 return false;
838}
839
840static inline __be32 try_6rd(struct ip_tunnel *tunnel,
841 const struct in6_addr *v6dst)
842{
843 __be32 dst = 0;
844 check_6rd(tunnel, v6dst, &dst);
845 return dst;
846}
847
848/*
849 * This function assumes it is being called from dev_queue_xmit()
850 * and that skb is filled properly by that function.
851 */
852
853static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
854 struct net_device *dev)
855{
856 struct ip_tunnel *tunnel = netdev_priv(dev);
857 const struct iphdr *tiph = &tunnel->parms.iph;
858 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
859 u8 tos = tunnel->parms.iph.tos;
860 __be16 df = tiph->frag_off;
861 struct rtable *rt; /* Route to the other host */
862 struct net_device *tdev; /* Device to other host */
863 unsigned int max_headroom; /* The extra header space needed */
864 __be32 dst = tiph->daddr;
865 struct flowi4 fl4;
866 int mtu;
867 const struct in6_addr *addr6;
868 int addr_type;
869 u8 ttl;
870 u8 protocol = IPPROTO_IPV6;
871 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
872
873 if (tos == 1)
874 tos = ipv6_get_dsfield(iph6);
875
876 /* ISATAP (RFC4214) - must come before 6to4 */
877 if (dev->priv_flags & IFF_ISATAP) {
878 struct neighbour *neigh = NULL;
879 bool do_tx_error = false;
880
881 if (skb_dst(skb))
882 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
883
884 if (!neigh) {
885 net_dbg_ratelimited("nexthop == NULL\n");
886 goto tx_error;
887 }
888
889 addr6 = (const struct in6_addr *)&neigh->primary_key;
890 addr_type = ipv6_addr_type(addr6);
891
892 if ((addr_type & IPV6_ADDR_UNICAST) &&
893 ipv6_addr_is_isatap(addr6))
894 dst = addr6->s6_addr32[3];
895 else
896 do_tx_error = true;
897
898 neigh_release(neigh);
899 if (do_tx_error)
900 goto tx_error;
901 }
902
903 if (!dst)
904 dst = try_6rd(tunnel, &iph6->daddr);
905
906 if (!dst) {
907 struct neighbour *neigh = NULL;
908 bool do_tx_error = false;
909
910 if (skb_dst(skb))
911 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
912
913 if (!neigh) {
914 net_dbg_ratelimited("nexthop == NULL\n");
915 goto tx_error;
916 }
917
918 addr6 = (const struct in6_addr *)&neigh->primary_key;
919 addr_type = ipv6_addr_type(addr6);
920
921 if (addr_type == IPV6_ADDR_ANY) {
922 addr6 = &ipv6_hdr(skb)->daddr;
923 addr_type = ipv6_addr_type(addr6);
924 }
925
926 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
927 dst = addr6->s6_addr32[3];
928 else
929 do_tx_error = true;
930
931 neigh_release(neigh);
932 if (do_tx_error)
933 goto tx_error;
934 }
935
936 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
937 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
938 0, dst, tiph->saddr, 0, 0,
939 sock_net_uid(tunnel->net, NULL));
940
941 rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
942 if (!rt) {
943 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
944 if (IS_ERR(rt)) {
945 DEV_STATS_INC(dev, tx_carrier_errors);
946 goto tx_error_icmp;
947 }
948 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
949 }
950
951 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
952 ip_rt_put(rt);
953 DEV_STATS_INC(dev, tx_carrier_errors);
954 goto tx_error_icmp;
955 }
956 tdev = rt->dst.dev;
957
958 if (tdev == dev) {
959 ip_rt_put(rt);
960 DEV_STATS_INC(dev, collisions);
961 goto tx_error;
962 }
963
964 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
965 ip_rt_put(rt);
966 goto tx_error;
967 }
968
969 if (df) {
970 mtu = dst_mtu(&rt->dst) - t_hlen;
971
972 if (mtu < IPV4_MIN_MTU) {
973 DEV_STATS_INC(dev, collisions);
974 ip_rt_put(rt);
975 goto tx_error;
976 }
977
978 if (mtu < IPV6_MIN_MTU) {
979 mtu = IPV6_MIN_MTU;
980 df = 0;
981 }
982
983 if (tunnel->parms.iph.daddr)
984 skb_dst_update_pmtu_no_confirm(skb, mtu);
985
986 if (skb->len > mtu && !skb_is_gso(skb)) {
987 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
988 ip_rt_put(rt);
989 goto tx_error;
990 }
991 }
992
993 if (tunnel->err_count > 0) {
994 if (time_before(jiffies,
995 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
996 tunnel->err_count--;
997 dst_link_failure(skb);
998 } else
999 tunnel->err_count = 0;
1000 }
1001
1002 /*
1003 * Okay, now see if we can stuff it in the buffer as-is.
1004 */
1005 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1006
1007 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1008 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1009 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1010 if (!new_skb) {
1011 ip_rt_put(rt);
1012 DEV_STATS_INC(dev, tx_dropped);
1013 kfree_skb(skb);
1014 return NETDEV_TX_OK;
1015 }
1016 if (skb->sk)
1017 skb_set_owner_w(new_skb, skb->sk);
1018 dev_kfree_skb(skb);
1019 skb = new_skb;
1020 iph6 = ipv6_hdr(skb);
1021 }
1022 ttl = tiph->ttl;
1023 if (ttl == 0)
1024 ttl = iph6->hop_limit;
1025 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1026
1027 if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1028 ip_rt_put(rt);
1029 goto tx_error;
1030 }
1031
1032 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1033
1034 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1035 df, !net_eq(tunnel->net, dev_net(dev)));
1036 return NETDEV_TX_OK;
1037
1038tx_error_icmp:
1039 dst_link_failure(skb);
1040tx_error:
1041 kfree_skb(skb);
1042 DEV_STATS_INC(dev, tx_errors);
1043 return NETDEV_TX_OK;
1044}
1045
1046static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1047 struct net_device *dev, u8 ipproto)
1048{
1049 struct ip_tunnel *tunnel = netdev_priv(dev);
1050 const struct iphdr *tiph = &tunnel->parms.iph;
1051
1052 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1053 goto tx_error;
1054
1055 skb_set_inner_ipproto(skb, ipproto);
1056
1057 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1058 return NETDEV_TX_OK;
1059tx_error:
1060 kfree_skb(skb);
1061 DEV_STATS_INC(dev, tx_errors);
1062 return NETDEV_TX_OK;
1063}
1064
1065static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1066 struct net_device *dev)
1067{
1068 if (!pskb_inet_may_pull(skb))
1069 goto tx_err;
1070
1071 switch (skb->protocol) {
1072 case htons(ETH_P_IP):
1073 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1074 break;
1075 case htons(ETH_P_IPV6):
1076 ipip6_tunnel_xmit(skb, dev);
1077 break;
1078#if IS_ENABLED(CONFIG_MPLS)
1079 case htons(ETH_P_MPLS_UC):
1080 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1081 break;
1082#endif
1083 default:
1084 goto tx_err;
1085 }
1086
1087 return NETDEV_TX_OK;
1088
1089tx_err:
1090 DEV_STATS_INC(dev, tx_errors);
1091 kfree_skb(skb);
1092 return NETDEV_TX_OK;
1093
1094}
1095
1096static void ipip6_tunnel_bind_dev(struct net_device *dev)
1097{
1098 struct ip_tunnel *tunnel = netdev_priv(dev);
1099 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1100 struct net_device *tdev = NULL;
1101 int hlen = LL_MAX_HEADER;
1102 const struct iphdr *iph;
1103 struct flowi4 fl4;
1104
1105 iph = &tunnel->parms.iph;
1106
1107 if (iph->daddr) {
1108 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1109 NULL,
1110 iph->daddr, iph->saddr,
1111 0, 0,
1112 IPPROTO_IPV6,
1113 RT_TOS(iph->tos),
1114 tunnel->parms.link);
1115
1116 if (!IS_ERR(rt)) {
1117 tdev = rt->dst.dev;
1118 ip_rt_put(rt);
1119 }
1120 dev->flags |= IFF_POINTOPOINT;
1121 }
1122
1123 if (!tdev && tunnel->parms.link)
1124 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1125
1126 if (tdev && !netif_is_l3_master(tdev)) {
1127 int mtu;
1128
1129 mtu = tdev->mtu - t_hlen;
1130 if (mtu < IPV6_MIN_MTU)
1131 mtu = IPV6_MIN_MTU;
1132 WRITE_ONCE(dev->mtu, mtu);
1133 hlen = tdev->hard_header_len + tdev->needed_headroom;
1134 }
1135 dev->needed_headroom = t_hlen + hlen;
1136}
1137
1138static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1139 __u32 fwmark)
1140{
1141 struct net *net = t->net;
1142 struct sit_net *sitn = net_generic(net, sit_net_id);
1143
1144 ipip6_tunnel_unlink(sitn, t);
1145 synchronize_net();
1146 t->parms.iph.saddr = p->iph.saddr;
1147 t->parms.iph.daddr = p->iph.daddr;
1148 __dev_addr_set(t->dev, &p->iph.saddr, 4);
1149 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1150 ipip6_tunnel_link(sitn, t);
1151 t->parms.iph.ttl = p->iph.ttl;
1152 t->parms.iph.tos = p->iph.tos;
1153 t->parms.iph.frag_off = p->iph.frag_off;
1154 if (t->parms.link != p->link || t->fwmark != fwmark) {
1155 t->parms.link = p->link;
1156 t->fwmark = fwmark;
1157 ipip6_tunnel_bind_dev(t->dev);
1158 }
1159 dst_cache_reset(&t->dst_cache);
1160 netdev_state_change(t->dev);
1161}
1162
1163#ifdef CONFIG_IPV6_SIT_6RD
1164static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1165 struct ip_tunnel_6rd *ip6rd)
1166{
1167 struct in6_addr prefix;
1168 __be32 relay_prefix;
1169
1170 if (ip6rd->relay_prefixlen > 32 ||
1171 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1172 return -EINVAL;
1173
1174 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1175 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1176 return -EINVAL;
1177 if (ip6rd->relay_prefixlen)
1178 relay_prefix = ip6rd->relay_prefix &
1179 htonl(0xffffffffUL <<
1180 (32 - ip6rd->relay_prefixlen));
1181 else
1182 relay_prefix = 0;
1183 if (relay_prefix != ip6rd->relay_prefix)
1184 return -EINVAL;
1185
1186 t->ip6rd.prefix = prefix;
1187 t->ip6rd.relay_prefix = relay_prefix;
1188 t->ip6rd.prefixlen = ip6rd->prefixlen;
1189 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1190 dst_cache_reset(&t->dst_cache);
1191 netdev_state_change(t->dev);
1192 return 0;
1193}
1194
1195static int
1196ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1197{
1198 struct ip_tunnel *t = netdev_priv(dev);
1199 struct ip_tunnel_6rd ip6rd;
1200 struct ip_tunnel_parm p;
1201
1202 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1203 if (copy_from_user(&p, data, sizeof(p)))
1204 return -EFAULT;
1205 t = ipip6_tunnel_locate(t->net, &p, 0);
1206 }
1207 if (!t)
1208 t = netdev_priv(dev);
1209
1210 ip6rd.prefix = t->ip6rd.prefix;
1211 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1212 ip6rd.prefixlen = t->ip6rd.prefixlen;
1213 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1214 if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1215 return -EFAULT;
1216 return 0;
1217}
1218
1219static int
1220ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1221 int cmd)
1222{
1223 struct ip_tunnel *t = netdev_priv(dev);
1224 struct ip_tunnel_6rd ip6rd;
1225 int err;
1226
1227 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1228 return -EPERM;
1229 if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1230 return -EFAULT;
1231
1232 if (cmd != SIOCDEL6RD) {
1233 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1234 if (err < 0)
1235 return err;
1236 } else
1237 ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1238 return 0;
1239}
1240
1241#endif /* CONFIG_IPV6_SIT_6RD */
1242
1243static bool ipip6_valid_ip_proto(u8 ipproto)
1244{
1245 return ipproto == IPPROTO_IPV6 ||
1246 ipproto == IPPROTO_IPIP ||
1247#if IS_ENABLED(CONFIG_MPLS)
1248 ipproto == IPPROTO_MPLS ||
1249#endif
1250 ipproto == 0;
1251}
1252
1253static int
1254__ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1255{
1256 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1257 return -EPERM;
1258
1259 if (!ipip6_valid_ip_proto(p->iph.protocol))
1260 return -EINVAL;
1261 if (p->iph.version != 4 ||
1262 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1263 return -EINVAL;
1264
1265 if (p->iph.ttl)
1266 p->iph.frag_off |= htons(IP_DF);
1267 return 0;
1268}
1269
1270static int
1271ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1272{
1273 struct ip_tunnel *t = netdev_priv(dev);
1274
1275 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1276 t = ipip6_tunnel_locate(t->net, p, 0);
1277 if (!t)
1278 t = netdev_priv(dev);
1279 memcpy(p, &t->parms, sizeof(*p));
1280 return 0;
1281}
1282
1283static int
1284ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1285{
1286 struct ip_tunnel *t = netdev_priv(dev);
1287 int err;
1288
1289 err = __ipip6_tunnel_ioctl_validate(t->net, p);
1290 if (err)
1291 return err;
1292
1293 t = ipip6_tunnel_locate(t->net, p, 1);
1294 if (!t)
1295 return -ENOBUFS;
1296 return 0;
1297}
1298
1299static int
1300ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1301{
1302 struct ip_tunnel *t = netdev_priv(dev);
1303 int err;
1304
1305 err = __ipip6_tunnel_ioctl_validate(t->net, p);
1306 if (err)
1307 return err;
1308
1309 t = ipip6_tunnel_locate(t->net, p, 0);
1310 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1311 if (!t)
1312 return -ENOENT;
1313 } else {
1314 if (t) {
1315 if (t->dev != dev)
1316 return -EEXIST;
1317 } else {
1318 if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1319 (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1320 return -EINVAL;
1321 t = netdev_priv(dev);
1322 }
1323
1324 ipip6_tunnel_update(t, p, t->fwmark);
1325 }
1326
1327 return 0;
1328}
1329
1330static int
1331ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1332{
1333 struct ip_tunnel *t = netdev_priv(dev);
1334
1335 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1336 return -EPERM;
1337
1338 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1339 t = ipip6_tunnel_locate(t->net, p, 0);
1340 if (!t)
1341 return -ENOENT;
1342 if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1343 return -EPERM;
1344 dev = t->dev;
1345 }
1346 unregister_netdevice(dev);
1347 return 0;
1348}
1349
1350static int
1351ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1352{
1353 switch (cmd) {
1354 case SIOCGETTUNNEL:
1355 return ipip6_tunnel_get(dev, p);
1356 case SIOCADDTUNNEL:
1357 return ipip6_tunnel_add(dev, p);
1358 case SIOCCHGTUNNEL:
1359 return ipip6_tunnel_change(dev, p);
1360 case SIOCDELTUNNEL:
1361 return ipip6_tunnel_del(dev, p);
1362 default:
1363 return -EINVAL;
1364 }
1365}
1366
1367static int
1368ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1369 void __user *data, int cmd)
1370{
1371 switch (cmd) {
1372 case SIOCGETTUNNEL:
1373 case SIOCADDTUNNEL:
1374 case SIOCCHGTUNNEL:
1375 case SIOCDELTUNNEL:
1376 return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1377 case SIOCGETPRL:
1378 return ipip6_tunnel_get_prl(dev, data);
1379 case SIOCADDPRL:
1380 case SIOCDELPRL:
1381 case SIOCCHGPRL:
1382 return ipip6_tunnel_prl_ctl(dev, data, cmd);
1383#ifdef CONFIG_IPV6_SIT_6RD
1384 case SIOCGET6RD:
1385 return ipip6_tunnel_get6rd(dev, data);
1386 case SIOCADD6RD:
1387 case SIOCCHG6RD:
1388 case SIOCDEL6RD:
1389 return ipip6_tunnel_6rdctl(dev, data, cmd);
1390#endif
1391 default:
1392 return -EINVAL;
1393 }
1394}
1395
1396static const struct net_device_ops ipip6_netdev_ops = {
1397 .ndo_init = ipip6_tunnel_init,
1398 .ndo_uninit = ipip6_tunnel_uninit,
1399 .ndo_start_xmit = sit_tunnel_xmit,
1400 .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1401 .ndo_get_stats64 = dev_get_tstats64,
1402 .ndo_get_iflink = ip_tunnel_get_iflink,
1403 .ndo_tunnel_ctl = ipip6_tunnel_ctl,
1404};
1405
1406static void ipip6_dev_free(struct net_device *dev)
1407{
1408 struct ip_tunnel *tunnel = netdev_priv(dev);
1409
1410 dst_cache_destroy(&tunnel->dst_cache);
1411 free_percpu(dev->tstats);
1412}
1413
1414#define SIT_FEATURES (NETIF_F_SG | \
1415 NETIF_F_FRAGLIST | \
1416 NETIF_F_HIGHDMA | \
1417 NETIF_F_GSO_SOFTWARE | \
1418 NETIF_F_HW_CSUM)
1419
1420static void ipip6_tunnel_setup(struct net_device *dev)
1421{
1422 struct ip_tunnel *tunnel = netdev_priv(dev);
1423 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1424
1425 dev->netdev_ops = &ipip6_netdev_ops;
1426 dev->header_ops = &ip_tunnel_header_ops;
1427 dev->needs_free_netdev = true;
1428 dev->priv_destructor = ipip6_dev_free;
1429
1430 dev->type = ARPHRD_SIT;
1431 dev->mtu = ETH_DATA_LEN - t_hlen;
1432 dev->min_mtu = IPV6_MIN_MTU;
1433 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1434 dev->flags = IFF_NOARP;
1435 netif_keep_dst(dev);
1436 dev->addr_len = 4;
1437 dev->features |= NETIF_F_LLTX;
1438 dev->features |= SIT_FEATURES;
1439 dev->hw_features |= SIT_FEATURES;
1440}
1441
1442static int ipip6_tunnel_init(struct net_device *dev)
1443{
1444 struct ip_tunnel *tunnel = netdev_priv(dev);
1445 int err;
1446
1447 tunnel->dev = dev;
1448 tunnel->net = dev_net(dev);
1449 strcpy(tunnel->parms.name, dev->name);
1450
1451 ipip6_tunnel_bind_dev(dev);
1452 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1453 if (!dev->tstats)
1454 return -ENOMEM;
1455
1456 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1457 if (err) {
1458 free_percpu(dev->tstats);
1459 dev->tstats = NULL;
1460 return err;
1461 }
1462 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1463 return 0;
1464}
1465
1466static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1467{
1468 struct ip_tunnel *tunnel = netdev_priv(dev);
1469 struct iphdr *iph = &tunnel->parms.iph;
1470 struct net *net = dev_net(dev);
1471 struct sit_net *sitn = net_generic(net, sit_net_id);
1472
1473 iph->version = 4;
1474 iph->protocol = IPPROTO_IPV6;
1475 iph->ihl = 5;
1476 iph->ttl = 64;
1477
1478 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1479}
1480
1481static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1482 struct netlink_ext_ack *extack)
1483{
1484 u8 proto;
1485
1486 if (!data || !data[IFLA_IPTUN_PROTO])
1487 return 0;
1488
1489 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1490 if (!ipip6_valid_ip_proto(proto))
1491 return -EINVAL;
1492
1493 return 0;
1494}
1495
1496static void ipip6_netlink_parms(struct nlattr *data[],
1497 struct ip_tunnel_parm *parms,
1498 __u32 *fwmark)
1499{
1500 memset(parms, 0, sizeof(*parms));
1501
1502 parms->iph.version = 4;
1503 parms->iph.protocol = IPPROTO_IPV6;
1504 parms->iph.ihl = 5;
1505 parms->iph.ttl = 64;
1506
1507 if (!data)
1508 return;
1509
1510 ip_tunnel_netlink_parms(data, parms);
1511
1512 if (data[IFLA_IPTUN_FWMARK])
1513 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1514}
1515
1516#ifdef CONFIG_IPV6_SIT_6RD
1517/* This function returns true when 6RD attributes are present in the nl msg */
1518static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1519 struct ip_tunnel_6rd *ip6rd)
1520{
1521 bool ret = false;
1522 memset(ip6rd, 0, sizeof(*ip6rd));
1523
1524 if (!data)
1525 return ret;
1526
1527 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1528 ret = true;
1529 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1530 }
1531
1532 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1533 ret = true;
1534 ip6rd->relay_prefix =
1535 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1536 }
1537
1538 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1539 ret = true;
1540 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1541 }
1542
1543 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1544 ret = true;
1545 ip6rd->relay_prefixlen =
1546 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1547 }
1548
1549 return ret;
1550}
1551#endif
1552
1553static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1554 struct nlattr *tb[], struct nlattr *data[],
1555 struct netlink_ext_ack *extack)
1556{
1557 struct net *net = dev_net(dev);
1558 struct ip_tunnel *nt;
1559 struct ip_tunnel_encap ipencap;
1560#ifdef CONFIG_IPV6_SIT_6RD
1561 struct ip_tunnel_6rd ip6rd;
1562#endif
1563 int err;
1564
1565 nt = netdev_priv(dev);
1566
1567 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1568 err = ip_tunnel_encap_setup(nt, &ipencap);
1569 if (err < 0)
1570 return err;
1571 }
1572
1573 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1574
1575 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1576 return -EEXIST;
1577
1578 err = ipip6_tunnel_create(dev);
1579 if (err < 0)
1580 return err;
1581
1582 if (tb[IFLA_MTU]) {
1583 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1584
1585 if (mtu >= IPV6_MIN_MTU &&
1586 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1587 dev->mtu = mtu;
1588 }
1589
1590#ifdef CONFIG_IPV6_SIT_6RD
1591 if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1592 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1593 if (err < 0)
1594 unregister_netdevice_queue(dev, NULL);
1595 }
1596#endif
1597
1598 return err;
1599}
1600
1601static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1602 struct nlattr *data[],
1603 struct netlink_ext_ack *extack)
1604{
1605 struct ip_tunnel *t = netdev_priv(dev);
1606 struct ip_tunnel_parm p;
1607 struct ip_tunnel_encap ipencap;
1608 struct net *net = t->net;
1609 struct sit_net *sitn = net_generic(net, sit_net_id);
1610#ifdef CONFIG_IPV6_SIT_6RD
1611 struct ip_tunnel_6rd ip6rd;
1612#endif
1613 __u32 fwmark = t->fwmark;
1614 int err;
1615
1616 if (dev == sitn->fb_tunnel_dev)
1617 return -EINVAL;
1618
1619 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1620 err = ip_tunnel_encap_setup(t, &ipencap);
1621 if (err < 0)
1622 return err;
1623 }
1624
1625 ipip6_netlink_parms(data, &p, &fwmark);
1626
1627 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1628 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1629 return -EINVAL;
1630
1631 t = ipip6_tunnel_locate(net, &p, 0);
1632
1633 if (t) {
1634 if (t->dev != dev)
1635 return -EEXIST;
1636 } else
1637 t = netdev_priv(dev);
1638
1639 ipip6_tunnel_update(t, &p, fwmark);
1640
1641#ifdef CONFIG_IPV6_SIT_6RD
1642 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1643 return ipip6_tunnel_update_6rd(t, &ip6rd);
1644#endif
1645
1646 return 0;
1647}
1648
1649static size_t ipip6_get_size(const struct net_device *dev)
1650{
1651 return
1652 /* IFLA_IPTUN_LINK */
1653 nla_total_size(4) +
1654 /* IFLA_IPTUN_LOCAL */
1655 nla_total_size(4) +
1656 /* IFLA_IPTUN_REMOTE */
1657 nla_total_size(4) +
1658 /* IFLA_IPTUN_TTL */
1659 nla_total_size(1) +
1660 /* IFLA_IPTUN_TOS */
1661 nla_total_size(1) +
1662 /* IFLA_IPTUN_PMTUDISC */
1663 nla_total_size(1) +
1664 /* IFLA_IPTUN_FLAGS */
1665 nla_total_size(2) +
1666 /* IFLA_IPTUN_PROTO */
1667 nla_total_size(1) +
1668#ifdef CONFIG_IPV6_SIT_6RD
1669 /* IFLA_IPTUN_6RD_PREFIX */
1670 nla_total_size(sizeof(struct in6_addr)) +
1671 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1672 nla_total_size(4) +
1673 /* IFLA_IPTUN_6RD_PREFIXLEN */
1674 nla_total_size(2) +
1675 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1676 nla_total_size(2) +
1677#endif
1678 /* IFLA_IPTUN_ENCAP_TYPE */
1679 nla_total_size(2) +
1680 /* IFLA_IPTUN_ENCAP_FLAGS */
1681 nla_total_size(2) +
1682 /* IFLA_IPTUN_ENCAP_SPORT */
1683 nla_total_size(2) +
1684 /* IFLA_IPTUN_ENCAP_DPORT */
1685 nla_total_size(2) +
1686 /* IFLA_IPTUN_FWMARK */
1687 nla_total_size(4) +
1688 0;
1689}
1690
1691static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1692{
1693 struct ip_tunnel *tunnel = netdev_priv(dev);
1694 struct ip_tunnel_parm *parm = &tunnel->parms;
1695
1696 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1697 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1698 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1699 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1700 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1701 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1702 !!(parm->iph.frag_off & htons(IP_DF))) ||
1703 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1704 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1705 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1706 goto nla_put_failure;
1707
1708#ifdef CONFIG_IPV6_SIT_6RD
1709 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1710 &tunnel->ip6rd.prefix) ||
1711 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1712 tunnel->ip6rd.relay_prefix) ||
1713 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1714 tunnel->ip6rd.prefixlen) ||
1715 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1716 tunnel->ip6rd.relay_prefixlen))
1717 goto nla_put_failure;
1718#endif
1719
1720 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1721 tunnel->encap.type) ||
1722 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1723 tunnel->encap.sport) ||
1724 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1725 tunnel->encap.dport) ||
1726 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1727 tunnel->encap.flags))
1728 goto nla_put_failure;
1729
1730 return 0;
1731
1732nla_put_failure:
1733 return -EMSGSIZE;
1734}
1735
1736static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1737 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1738 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1739 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1740 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1741 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1742 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1743 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1744 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1745#ifdef CONFIG_IPV6_SIT_6RD
1746 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1747 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1748 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1749 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1750#endif
1751 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1752 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1753 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1754 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1755 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1756};
1757
1758static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1759{
1760 struct net *net = dev_net(dev);
1761 struct sit_net *sitn = net_generic(net, sit_net_id);
1762
1763 if (dev != sitn->fb_tunnel_dev)
1764 unregister_netdevice_queue(dev, head);
1765}
1766
1767static struct rtnl_link_ops sit_link_ops __read_mostly = {
1768 .kind = "sit",
1769 .maxtype = IFLA_IPTUN_MAX,
1770 .policy = ipip6_policy,
1771 .priv_size = sizeof(struct ip_tunnel),
1772 .setup = ipip6_tunnel_setup,
1773 .validate = ipip6_validate,
1774 .newlink = ipip6_newlink,
1775 .changelink = ipip6_changelink,
1776 .get_size = ipip6_get_size,
1777 .fill_info = ipip6_fill_info,
1778 .dellink = ipip6_dellink,
1779 .get_link_net = ip_tunnel_get_link_net,
1780};
1781
1782static struct xfrm_tunnel sit_handler __read_mostly = {
1783 .handler = ipip6_rcv,
1784 .err_handler = ipip6_err,
1785 .priority = 1,
1786};
1787
1788static struct xfrm_tunnel ipip_handler __read_mostly = {
1789 .handler = ipip_rcv,
1790 .err_handler = ipip6_err,
1791 .priority = 2,
1792};
1793
1794#if IS_ENABLED(CONFIG_MPLS)
1795static struct xfrm_tunnel mplsip_handler __read_mostly = {
1796 .handler = mplsip_rcv,
1797 .err_handler = ipip6_err,
1798 .priority = 2,
1799};
1800#endif
1801
1802static void __net_exit sit_destroy_tunnels(struct net *net,
1803 struct list_head *head)
1804{
1805 struct sit_net *sitn = net_generic(net, sit_net_id);
1806 struct net_device *dev, *aux;
1807 int prio;
1808
1809 for_each_netdev_safe(net, dev, aux)
1810 if (dev->rtnl_link_ops == &sit_link_ops)
1811 unregister_netdevice_queue(dev, head);
1812
1813 for (prio = 0; prio < 4; prio++) {
1814 int h;
1815 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1816 struct ip_tunnel *t;
1817
1818 t = rtnl_dereference(sitn->tunnels[prio][h]);
1819 while (t) {
1820 /* If dev is in the same netns, it has already
1821 * been added to the list by the previous loop.
1822 */
1823 if (!net_eq(dev_net(t->dev), net))
1824 unregister_netdevice_queue(t->dev,
1825 head);
1826 t = rtnl_dereference(t->next);
1827 }
1828 }
1829 }
1830}
1831
1832static int __net_init sit_init_net(struct net *net)
1833{
1834 struct sit_net *sitn = net_generic(net, sit_net_id);
1835 struct ip_tunnel *t;
1836 int err;
1837
1838 sitn->tunnels[0] = sitn->tunnels_wc;
1839 sitn->tunnels[1] = sitn->tunnels_l;
1840 sitn->tunnels[2] = sitn->tunnels_r;
1841 sitn->tunnels[3] = sitn->tunnels_r_l;
1842
1843 if (!net_has_fallback_tunnels(net))
1844 return 0;
1845
1846 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1847 NET_NAME_UNKNOWN,
1848 ipip6_tunnel_setup);
1849 if (!sitn->fb_tunnel_dev) {
1850 err = -ENOMEM;
1851 goto err_alloc_dev;
1852 }
1853 dev_net_set(sitn->fb_tunnel_dev, net);
1854 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1855 /* FB netdevice is special: we have one, and only one per netns.
1856 * Allowing to move it to another netns is clearly unsafe.
1857 */
1858 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1859
1860 err = register_netdev(sitn->fb_tunnel_dev);
1861 if (err)
1862 goto err_reg_dev;
1863
1864 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1865 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1866
1867 t = netdev_priv(sitn->fb_tunnel_dev);
1868
1869 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1870 return 0;
1871
1872err_reg_dev:
1873 free_netdev(sitn->fb_tunnel_dev);
1874err_alloc_dev:
1875 return err;
1876}
1877
1878static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1879{
1880 LIST_HEAD(list);
1881 struct net *net;
1882
1883 rtnl_lock();
1884 list_for_each_entry(net, net_list, exit_list)
1885 sit_destroy_tunnels(net, &list);
1886
1887 unregister_netdevice_many(&list);
1888 rtnl_unlock();
1889}
1890
1891static struct pernet_operations sit_net_ops = {
1892 .init = sit_init_net,
1893 .exit_batch = sit_exit_batch_net,
1894 .id = &sit_net_id,
1895 .size = sizeof(struct sit_net),
1896};
1897
1898static void __exit sit_cleanup(void)
1899{
1900 rtnl_link_unregister(&sit_link_ops);
1901 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1902 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1903#if IS_ENABLED(CONFIG_MPLS)
1904 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1905#endif
1906
1907 unregister_pernet_device(&sit_net_ops);
1908 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1909}
1910
1911static int __init sit_init(void)
1912{
1913 int err;
1914
1915 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1916
1917 err = register_pernet_device(&sit_net_ops);
1918 if (err < 0)
1919 return err;
1920 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1921 if (err < 0) {
1922 pr_info("%s: can't register ip6ip4\n", __func__);
1923 goto xfrm_tunnel_failed;
1924 }
1925 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1926 if (err < 0) {
1927 pr_info("%s: can't register ip4ip4\n", __func__);
1928 goto xfrm_tunnel4_failed;
1929 }
1930#if IS_ENABLED(CONFIG_MPLS)
1931 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1932 if (err < 0) {
1933 pr_info("%s: can't register mplsip\n", __func__);
1934 goto xfrm_tunnel_mpls_failed;
1935 }
1936#endif
1937 err = rtnl_link_register(&sit_link_ops);
1938 if (err < 0)
1939 goto rtnl_link_failed;
1940
1941out:
1942 return err;
1943
1944rtnl_link_failed:
1945#if IS_ENABLED(CONFIG_MPLS)
1946 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1947xfrm_tunnel_mpls_failed:
1948#endif
1949 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1950xfrm_tunnel4_failed:
1951 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1952xfrm_tunnel_failed:
1953 unregister_pernet_device(&sit_net_ops);
1954 goto out;
1955}
1956
1957module_init(sit_init);
1958module_exit(sit_cleanup);
1959MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1960MODULE_LICENSE("GPL");
1961MODULE_ALIAS_RTNL_LINK("sit");
1962MODULE_ALIAS_NETDEV("sit0");
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
68static bool log_ecn_error = true;
69module_param(log_ecn_error, bool, 0644);
70MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71
72static int ipip6_tunnel_init(struct net_device *dev);
73static void ipip6_tunnel_setup(struct net_device *dev);
74static void ipip6_dev_free(struct net_device *dev);
75static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 __be32 *v4dst);
77static struct rtnl_link_ops sit_link_ops __read_mostly;
78
79static int sit_net_id __read_mostly;
80struct 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 */
93static 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
126static 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
145static 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
151static 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
166static 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
174static 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
191static 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
218out:
219 return err;
220}
221
222static 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
267failed_free:
268 ipip6_dev_free(dev);
269failed:
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
278static 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
290static 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 }
340out:
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
353static int
354ipip6_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);
391out:
392 return err;
393}
394
395static 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
407static int
408ipip6_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 }
436out:
437 return err;
438}
439
440static int
441isatap_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
467static void ipip6_tunnel_uninit(struct net_device *dev)
468{
469 struct ip_tunnel *tunnel = netdev_priv(dev);
470 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
471
472 if (dev == sitn->fb_tunnel_dev) {
473 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
474 } else {
475 ipip6_tunnel_unlink(sitn, tunnel);
476 ipip6_tunnel_del_prl(tunnel, NULL);
477 }
478 ip_tunnel_dst_reset_all(tunnel);
479 dev_put(dev);
480}
481
482/* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
483 * if sufficient data bytes are available
484 */
485static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
486{
487 const struct iphdr *iph = (const struct iphdr *) skb->data;
488 struct rt6_info *rt;
489 struct sk_buff *skb2;
490
491 if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
492 return 1;
493
494 skb2 = skb_clone(skb, GFP_ATOMIC);
495
496 if (!skb2)
497 return 1;
498
499 skb_dst_drop(skb2);
500 skb_pull(skb2, iph->ihl * 4);
501 skb_reset_network_header(skb2);
502
503 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
504
505 if (rt && rt->dst.dev)
506 skb2->dev = rt->dst.dev;
507
508 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
509
510 if (rt)
511 ip6_rt_put(rt);
512
513 kfree_skb(skb2);
514
515 return 0;
516}
517
518static int ipip6_err(struct sk_buff *skb, u32 info)
519{
520 const struct iphdr *iph = (const struct iphdr *)skb->data;
521 const int type = icmp_hdr(skb)->type;
522 const int code = icmp_hdr(skb)->code;
523 struct ip_tunnel *t;
524 int err;
525
526 switch (type) {
527 default:
528 case ICMP_PARAMETERPROB:
529 return 0;
530
531 case ICMP_DEST_UNREACH:
532 switch (code) {
533 case ICMP_SR_FAILED:
534 /* Impossible event. */
535 return 0;
536 default:
537 /* All others are translated to HOST_UNREACH.
538 rfc2003 contains "deep thoughts" about NET_UNREACH,
539 I believe they are just ether pollution. --ANK
540 */
541 break;
542 }
543 break;
544 case ICMP_TIME_EXCEEDED:
545 if (code != ICMP_EXC_TTL)
546 return 0;
547 break;
548 case ICMP_REDIRECT:
549 break;
550 }
551
552 err = -ENOENT;
553
554 t = ipip6_tunnel_lookup(dev_net(skb->dev),
555 skb->dev,
556 iph->daddr,
557 iph->saddr);
558 if (t == NULL)
559 goto out;
560
561 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
562 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
563 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
564 err = 0;
565 goto out;
566 }
567 if (type == ICMP_REDIRECT) {
568 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
569 IPPROTO_IPV6, 0);
570 err = 0;
571 goto out;
572 }
573
574 if (t->parms.iph.daddr == 0)
575 goto out;
576
577 err = 0;
578 if (!ipip6_err_gen_icmpv6_unreach(skb))
579 goto out;
580
581 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
582 goto out;
583
584 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
585 t->err_count++;
586 else
587 t->err_count = 1;
588 t->err_time = jiffies;
589out:
590 return err;
591}
592
593static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
594 const struct in6_addr *v6addr)
595{
596 __be32 v4embed = 0;
597 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
598 return true;
599 return false;
600}
601
602/* Checks if an address matches an address on the tunnel interface.
603 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
604 * Long story:
605 * This function is called after we considered the packet as spoofed
606 * in is_spoofed_6rd.
607 * We may have a router that is doing NAT for proto 41 packets
608 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
609 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
610 * function will return true, dropping the packet.
611 * But, we can still check if is spoofed against the IP
612 * addresses associated with the interface.
613 */
614static bool only_dnatted(const struct ip_tunnel *tunnel,
615 const struct in6_addr *v6dst)
616{
617 int prefix_len;
618
619#ifdef CONFIG_IPV6_SIT_6RD
620 prefix_len = tunnel->ip6rd.prefixlen + 32
621 - tunnel->ip6rd.relay_prefixlen;
622#else
623 prefix_len = 48;
624#endif
625 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
626}
627
628/* Returns true if a packet is spoofed */
629static bool packet_is_spoofed(struct sk_buff *skb,
630 const struct iphdr *iph,
631 struct ip_tunnel *tunnel)
632{
633 const struct ipv6hdr *ipv6h;
634
635 if (tunnel->dev->priv_flags & IFF_ISATAP) {
636 if (!isatap_chksrc(skb, iph, tunnel))
637 return true;
638
639 return false;
640 }
641
642 if (tunnel->dev->flags & IFF_POINTOPOINT)
643 return false;
644
645 ipv6h = ipv6_hdr(skb);
646
647 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
648 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
649 &iph->saddr, &ipv6h->saddr,
650 &iph->daddr, &ipv6h->daddr);
651 return true;
652 }
653
654 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
655 return false;
656
657 if (only_dnatted(tunnel, &ipv6h->daddr))
658 return false;
659
660 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661 &iph->saddr, &ipv6h->saddr,
662 &iph->daddr, &ipv6h->daddr);
663 return true;
664}
665
666static int ipip6_rcv(struct sk_buff *skb)
667{
668 const struct iphdr *iph = ip_hdr(skb);
669 struct ip_tunnel *tunnel;
670 int err;
671
672 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
673 iph->saddr, iph->daddr);
674 if (tunnel != NULL) {
675 struct pcpu_sw_netstats *tstats;
676
677 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
678 tunnel->parms.iph.protocol != 0)
679 goto out;
680
681 skb->mac_header = skb->network_header;
682 skb_reset_network_header(skb);
683 IPCB(skb)->flags = 0;
684 skb->protocol = htons(ETH_P_IPV6);
685
686 if (packet_is_spoofed(skb, iph, tunnel)) {
687 tunnel->dev->stats.rx_errors++;
688 goto out;
689 }
690
691 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
692
693 err = IP_ECN_decapsulate(iph, skb);
694 if (unlikely(err)) {
695 if (log_ecn_error)
696 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
697 &iph->saddr, iph->tos);
698 if (err > 1) {
699 ++tunnel->dev->stats.rx_frame_errors;
700 ++tunnel->dev->stats.rx_errors;
701 goto out;
702 }
703 }
704
705 tstats = this_cpu_ptr(tunnel->dev->tstats);
706 u64_stats_update_begin(&tstats->syncp);
707 tstats->rx_packets++;
708 tstats->rx_bytes += skb->len;
709 u64_stats_update_end(&tstats->syncp);
710
711 netif_rx(skb);
712
713 return 0;
714 }
715
716 /* no tunnel matched, let upstream know, ipsec may handle it */
717 return 1;
718out:
719 kfree_skb(skb);
720 return 0;
721}
722
723static const struct tnl_ptk_info tpi = {
724 /* no tunnel info required for ipip. */
725 .proto = htons(ETH_P_IP),
726};
727
728static int ipip_rcv(struct sk_buff *skb)
729{
730 const struct iphdr *iph;
731 struct ip_tunnel *tunnel;
732
733 iph = ip_hdr(skb);
734 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
735 iph->saddr, iph->daddr);
736 if (tunnel != NULL) {
737 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
738 tunnel->parms.iph.protocol != 0)
739 goto drop;
740
741 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
742 goto drop;
743 if (iptunnel_pull_header(skb, 0, tpi.proto))
744 goto drop;
745 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
746 }
747
748 return 1;
749
750drop:
751 kfree_skb(skb);
752 return 0;
753}
754
755/*
756 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
757 * stores the embedded IPv4 address in v4dst and returns true.
758 */
759static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
760 __be32 *v4dst)
761{
762#ifdef CONFIG_IPV6_SIT_6RD
763 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
764 tunnel->ip6rd.prefixlen)) {
765 unsigned int pbw0, pbi0;
766 int pbi1;
767 u32 d;
768
769 pbw0 = tunnel->ip6rd.prefixlen >> 5;
770 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
771
772 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
773 tunnel->ip6rd.relay_prefixlen;
774
775 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
776 if (pbi1 > 0)
777 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
778 (32 - pbi1);
779
780 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
781 return true;
782 }
783#else
784 if (v6dst->s6_addr16[0] == htons(0x2002)) {
785 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
786 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
787 return true;
788 }
789#endif
790 return false;
791}
792
793static inline __be32 try_6rd(struct ip_tunnel *tunnel,
794 const struct in6_addr *v6dst)
795{
796 __be32 dst = 0;
797 check_6rd(tunnel, v6dst, &dst);
798 return dst;
799}
800
801/*
802 * This function assumes it is being called from dev_queue_xmit()
803 * and that skb is filled properly by that function.
804 */
805
806static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
807 struct net_device *dev)
808{
809 struct ip_tunnel *tunnel = netdev_priv(dev);
810 const struct iphdr *tiph = &tunnel->parms.iph;
811 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
812 u8 tos = tunnel->parms.iph.tos;
813 __be16 df = tiph->frag_off;
814 struct rtable *rt; /* Route to the other host */
815 struct net_device *tdev; /* Device to other host */
816 unsigned int max_headroom; /* The extra header space needed */
817 __be32 dst = tiph->daddr;
818 struct flowi4 fl4;
819 int mtu;
820 const struct in6_addr *addr6;
821 int addr_type;
822 u8 ttl;
823 int err;
824
825 if (skb->protocol != htons(ETH_P_IPV6))
826 goto tx_error;
827
828 if (tos == 1)
829 tos = ipv6_get_dsfield(iph6);
830
831 /* ISATAP (RFC4214) - must come before 6to4 */
832 if (dev->priv_flags & IFF_ISATAP) {
833 struct neighbour *neigh = NULL;
834 bool do_tx_error = false;
835
836 if (skb_dst(skb))
837 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
838
839 if (neigh == NULL) {
840 net_dbg_ratelimited("nexthop == NULL\n");
841 goto tx_error;
842 }
843
844 addr6 = (const struct in6_addr *)&neigh->primary_key;
845 addr_type = ipv6_addr_type(addr6);
846
847 if ((addr_type & IPV6_ADDR_UNICAST) &&
848 ipv6_addr_is_isatap(addr6))
849 dst = addr6->s6_addr32[3];
850 else
851 do_tx_error = true;
852
853 neigh_release(neigh);
854 if (do_tx_error)
855 goto tx_error;
856 }
857
858 if (!dst)
859 dst = try_6rd(tunnel, &iph6->daddr);
860
861 if (!dst) {
862 struct neighbour *neigh = NULL;
863 bool do_tx_error = false;
864
865 if (skb_dst(skb))
866 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
867
868 if (neigh == NULL) {
869 net_dbg_ratelimited("nexthop == NULL\n");
870 goto tx_error;
871 }
872
873 addr6 = (const struct in6_addr *)&neigh->primary_key;
874 addr_type = ipv6_addr_type(addr6);
875
876 if (addr_type == IPV6_ADDR_ANY) {
877 addr6 = &ipv6_hdr(skb)->daddr;
878 addr_type = ipv6_addr_type(addr6);
879 }
880
881 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
882 dst = addr6->s6_addr32[3];
883 else
884 do_tx_error = true;
885
886 neigh_release(neigh);
887 if (do_tx_error)
888 goto tx_error;
889 }
890
891 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
892 dst, tiph->saddr,
893 0, 0,
894 IPPROTO_IPV6, RT_TOS(tos),
895 tunnel->parms.link);
896 if (IS_ERR(rt)) {
897 dev->stats.tx_carrier_errors++;
898 goto tx_error_icmp;
899 }
900 if (rt->rt_type != RTN_UNICAST) {
901 ip_rt_put(rt);
902 dev->stats.tx_carrier_errors++;
903 goto tx_error_icmp;
904 }
905 tdev = rt->dst.dev;
906
907 if (tdev == dev) {
908 ip_rt_put(rt);
909 dev->stats.collisions++;
910 goto tx_error;
911 }
912
913 if (df) {
914 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
915
916 if (mtu < 68) {
917 dev->stats.collisions++;
918 ip_rt_put(rt);
919 goto tx_error;
920 }
921
922 if (mtu < IPV6_MIN_MTU) {
923 mtu = IPV6_MIN_MTU;
924 df = 0;
925 }
926
927 if (tunnel->parms.iph.daddr && skb_dst(skb))
928 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
929
930 if (skb->len > mtu && !skb_is_gso(skb)) {
931 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
932 ip_rt_put(rt);
933 goto tx_error;
934 }
935 }
936
937 if (tunnel->err_count > 0) {
938 if (time_before(jiffies,
939 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
940 tunnel->err_count--;
941 dst_link_failure(skb);
942 } else
943 tunnel->err_count = 0;
944 }
945
946 /*
947 * Okay, now see if we can stuff it in the buffer as-is.
948 */
949 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
950
951 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
952 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
953 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
954 if (!new_skb) {
955 ip_rt_put(rt);
956 dev->stats.tx_dropped++;
957 kfree_skb(skb);
958 return NETDEV_TX_OK;
959 }
960 if (skb->sk)
961 skb_set_owner_w(new_skb, skb->sk);
962 dev_kfree_skb(skb);
963 skb = new_skb;
964 iph6 = ipv6_hdr(skb);
965 }
966 ttl = tiph->ttl;
967 if (ttl == 0)
968 ttl = iph6->hop_limit;
969 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
970
971 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
972 if (IS_ERR(skb)) {
973 ip_rt_put(rt);
974 goto out;
975 }
976
977 err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr,
978 IPPROTO_IPV6, tos, ttl, df,
979 !net_eq(tunnel->net, dev_net(dev)));
980 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
981 return NETDEV_TX_OK;
982
983tx_error_icmp:
984 dst_link_failure(skb);
985tx_error:
986 kfree_skb(skb);
987out:
988 dev->stats.tx_errors++;
989 return NETDEV_TX_OK;
990}
991
992static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
993{
994 struct ip_tunnel *tunnel = netdev_priv(dev);
995 const struct iphdr *tiph = &tunnel->parms.iph;
996
997 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
998 if (IS_ERR(skb))
999 goto out;
1000
1001 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1002 return NETDEV_TX_OK;
1003out:
1004 dev->stats.tx_errors++;
1005 return NETDEV_TX_OK;
1006}
1007
1008static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1009 struct net_device *dev)
1010{
1011 switch (skb->protocol) {
1012 case htons(ETH_P_IP):
1013 ipip_tunnel_xmit(skb, dev);
1014 break;
1015 case htons(ETH_P_IPV6):
1016 ipip6_tunnel_xmit(skb, dev);
1017 break;
1018 default:
1019 goto tx_err;
1020 }
1021
1022 return NETDEV_TX_OK;
1023
1024tx_err:
1025 dev->stats.tx_errors++;
1026 kfree_skb(skb);
1027 return NETDEV_TX_OK;
1028
1029}
1030
1031static void ipip6_tunnel_bind_dev(struct net_device *dev)
1032{
1033 struct net_device *tdev = NULL;
1034 struct ip_tunnel *tunnel;
1035 const struct iphdr *iph;
1036 struct flowi4 fl4;
1037
1038 tunnel = netdev_priv(dev);
1039 iph = &tunnel->parms.iph;
1040
1041 if (iph->daddr) {
1042 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1043 NULL,
1044 iph->daddr, iph->saddr,
1045 0, 0,
1046 IPPROTO_IPV6,
1047 RT_TOS(iph->tos),
1048 tunnel->parms.link);
1049
1050 if (!IS_ERR(rt)) {
1051 tdev = rt->dst.dev;
1052 ip_rt_put(rt);
1053 }
1054 dev->flags |= IFF_POINTOPOINT;
1055 }
1056
1057 if (!tdev && tunnel->parms.link)
1058 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1059
1060 if (tdev) {
1061 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1062 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1063 if (dev->mtu < IPV6_MIN_MTU)
1064 dev->mtu = IPV6_MIN_MTU;
1065 }
1066 dev->iflink = tunnel->parms.link;
1067}
1068
1069static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1070{
1071 struct net *net = t->net;
1072 struct sit_net *sitn = net_generic(net, sit_net_id);
1073
1074 ipip6_tunnel_unlink(sitn, t);
1075 synchronize_net();
1076 t->parms.iph.saddr = p->iph.saddr;
1077 t->parms.iph.daddr = p->iph.daddr;
1078 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1079 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1080 ipip6_tunnel_link(sitn, t);
1081 t->parms.iph.ttl = p->iph.ttl;
1082 t->parms.iph.tos = p->iph.tos;
1083 if (t->parms.link != p->link) {
1084 t->parms.link = p->link;
1085 ipip6_tunnel_bind_dev(t->dev);
1086 }
1087 ip_tunnel_dst_reset_all(t);
1088 netdev_state_change(t->dev);
1089}
1090
1091#ifdef CONFIG_IPV6_SIT_6RD
1092static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1093 struct ip_tunnel_6rd *ip6rd)
1094{
1095 struct in6_addr prefix;
1096 __be32 relay_prefix;
1097
1098 if (ip6rd->relay_prefixlen > 32 ||
1099 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1100 return -EINVAL;
1101
1102 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1103 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1104 return -EINVAL;
1105 if (ip6rd->relay_prefixlen)
1106 relay_prefix = ip6rd->relay_prefix &
1107 htonl(0xffffffffUL <<
1108 (32 - ip6rd->relay_prefixlen));
1109 else
1110 relay_prefix = 0;
1111 if (relay_prefix != ip6rd->relay_prefix)
1112 return -EINVAL;
1113
1114 t->ip6rd.prefix = prefix;
1115 t->ip6rd.relay_prefix = relay_prefix;
1116 t->ip6rd.prefixlen = ip6rd->prefixlen;
1117 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1118 ip_tunnel_dst_reset_all(t);
1119 netdev_state_change(t->dev);
1120 return 0;
1121}
1122#endif
1123
1124static int
1125ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1126{
1127 int err = 0;
1128 struct ip_tunnel_parm p;
1129 struct ip_tunnel_prl prl;
1130 struct ip_tunnel *t = netdev_priv(dev);
1131 struct net *net = t->net;
1132 struct sit_net *sitn = net_generic(net, sit_net_id);
1133#ifdef CONFIG_IPV6_SIT_6RD
1134 struct ip_tunnel_6rd ip6rd;
1135#endif
1136
1137 switch (cmd) {
1138 case SIOCGETTUNNEL:
1139#ifdef CONFIG_IPV6_SIT_6RD
1140 case SIOCGET6RD:
1141#endif
1142 if (dev == sitn->fb_tunnel_dev) {
1143 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1144 err = -EFAULT;
1145 break;
1146 }
1147 t = ipip6_tunnel_locate(net, &p, 0);
1148 if (t == NULL)
1149 t = netdev_priv(dev);
1150 }
1151
1152 err = -EFAULT;
1153 if (cmd == SIOCGETTUNNEL) {
1154 memcpy(&p, &t->parms, sizeof(p));
1155 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1156 sizeof(p)))
1157 goto done;
1158#ifdef CONFIG_IPV6_SIT_6RD
1159 } else {
1160 ip6rd.prefix = t->ip6rd.prefix;
1161 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1162 ip6rd.prefixlen = t->ip6rd.prefixlen;
1163 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1164 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1165 sizeof(ip6rd)))
1166 goto done;
1167#endif
1168 }
1169 err = 0;
1170 break;
1171
1172 case SIOCADDTUNNEL:
1173 case SIOCCHGTUNNEL:
1174 err = -EPERM;
1175 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1176 goto done;
1177
1178 err = -EFAULT;
1179 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1180 goto done;
1181
1182 err = -EINVAL;
1183 if (p.iph.protocol != IPPROTO_IPV6 &&
1184 p.iph.protocol != IPPROTO_IPIP &&
1185 p.iph.protocol != 0)
1186 goto done;
1187 if (p.iph.version != 4 ||
1188 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1189 goto done;
1190 if (p.iph.ttl)
1191 p.iph.frag_off |= htons(IP_DF);
1192
1193 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1194
1195 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1196 if (t != NULL) {
1197 if (t->dev != dev) {
1198 err = -EEXIST;
1199 break;
1200 }
1201 } else {
1202 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1203 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1204 err = -EINVAL;
1205 break;
1206 }
1207 t = netdev_priv(dev);
1208 }
1209
1210 ipip6_tunnel_update(t, &p);
1211 }
1212
1213 if (t) {
1214 err = 0;
1215 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1216 err = -EFAULT;
1217 } else
1218 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1219 break;
1220
1221 case SIOCDELTUNNEL:
1222 err = -EPERM;
1223 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1224 goto done;
1225
1226 if (dev == sitn->fb_tunnel_dev) {
1227 err = -EFAULT;
1228 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1229 goto done;
1230 err = -ENOENT;
1231 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1232 goto done;
1233 err = -EPERM;
1234 if (t == netdev_priv(sitn->fb_tunnel_dev))
1235 goto done;
1236 dev = t->dev;
1237 }
1238 unregister_netdevice(dev);
1239 err = 0;
1240 break;
1241
1242 case SIOCGETPRL:
1243 err = -EINVAL;
1244 if (dev == sitn->fb_tunnel_dev)
1245 goto done;
1246 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1247 break;
1248
1249 case SIOCADDPRL:
1250 case SIOCDELPRL:
1251 case SIOCCHGPRL:
1252 err = -EPERM;
1253 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1254 goto done;
1255 err = -EINVAL;
1256 if (dev == sitn->fb_tunnel_dev)
1257 goto done;
1258 err = -EFAULT;
1259 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1260 goto done;
1261
1262 switch (cmd) {
1263 case SIOCDELPRL:
1264 err = ipip6_tunnel_del_prl(t, &prl);
1265 break;
1266 case SIOCADDPRL:
1267 case SIOCCHGPRL:
1268 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1269 break;
1270 }
1271 ip_tunnel_dst_reset_all(t);
1272 netdev_state_change(dev);
1273 break;
1274
1275#ifdef CONFIG_IPV6_SIT_6RD
1276 case SIOCADD6RD:
1277 case SIOCCHG6RD:
1278 case SIOCDEL6RD:
1279 err = -EPERM;
1280 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1281 goto done;
1282
1283 err = -EFAULT;
1284 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1285 sizeof(ip6rd)))
1286 goto done;
1287
1288 if (cmd != SIOCDEL6RD) {
1289 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1290 if (err < 0)
1291 goto done;
1292 } else
1293 ipip6_tunnel_clone_6rd(dev, sitn);
1294
1295 err = 0;
1296 break;
1297#endif
1298
1299 default:
1300 err = -EINVAL;
1301 }
1302
1303done:
1304 return err;
1305}
1306
1307static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1308{
1309 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1310 return -EINVAL;
1311 dev->mtu = new_mtu;
1312 return 0;
1313}
1314
1315static const struct net_device_ops ipip6_netdev_ops = {
1316 .ndo_uninit = ipip6_tunnel_uninit,
1317 .ndo_start_xmit = sit_tunnel_xmit,
1318 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1319 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1320 .ndo_get_stats64 = ip_tunnel_get_stats64,
1321};
1322
1323static void ipip6_dev_free(struct net_device *dev)
1324{
1325 struct ip_tunnel *tunnel = netdev_priv(dev);
1326
1327 free_percpu(tunnel->dst_cache);
1328 free_percpu(dev->tstats);
1329 free_netdev(dev);
1330}
1331
1332#define SIT_FEATURES (NETIF_F_SG | \
1333 NETIF_F_FRAGLIST | \
1334 NETIF_F_HIGHDMA | \
1335 NETIF_F_GSO_SOFTWARE | \
1336 NETIF_F_HW_CSUM)
1337
1338static void ipip6_tunnel_setup(struct net_device *dev)
1339{
1340 dev->netdev_ops = &ipip6_netdev_ops;
1341 dev->destructor = ipip6_dev_free;
1342
1343 dev->type = ARPHRD_SIT;
1344 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1345 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1346 dev->flags = IFF_NOARP;
1347 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1348 dev->iflink = 0;
1349 dev->addr_len = 4;
1350 dev->features |= NETIF_F_LLTX;
1351 dev->features |= SIT_FEATURES;
1352 dev->hw_features |= SIT_FEATURES;
1353}
1354
1355static int ipip6_tunnel_init(struct net_device *dev)
1356{
1357 struct ip_tunnel *tunnel = netdev_priv(dev);
1358
1359 tunnel->dev = dev;
1360 tunnel->net = dev_net(dev);
1361
1362 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1363 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1364
1365 ipip6_tunnel_bind_dev(dev);
1366 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1367 if (!dev->tstats)
1368 return -ENOMEM;
1369
1370 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1371 if (!tunnel->dst_cache) {
1372 free_percpu(dev->tstats);
1373 return -ENOMEM;
1374 }
1375
1376 return 0;
1377}
1378
1379static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1380{
1381 struct ip_tunnel *tunnel = netdev_priv(dev);
1382 struct iphdr *iph = &tunnel->parms.iph;
1383 struct net *net = dev_net(dev);
1384 struct sit_net *sitn = net_generic(net, sit_net_id);
1385
1386 tunnel->dev = dev;
1387 tunnel->net = dev_net(dev);
1388 strcpy(tunnel->parms.name, dev->name);
1389
1390 iph->version = 4;
1391 iph->protocol = IPPROTO_IPV6;
1392 iph->ihl = 5;
1393 iph->ttl = 64;
1394
1395 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1396 if (!dev->tstats)
1397 return -ENOMEM;
1398
1399 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1400 if (!tunnel->dst_cache) {
1401 free_percpu(dev->tstats);
1402 return -ENOMEM;
1403 }
1404
1405 dev_hold(dev);
1406 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1407 return 0;
1408}
1409
1410static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1411{
1412 u8 proto;
1413
1414 if (!data || !data[IFLA_IPTUN_PROTO])
1415 return 0;
1416
1417 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1418 if (proto != IPPROTO_IPV6 &&
1419 proto != IPPROTO_IPIP &&
1420 proto != 0)
1421 return -EINVAL;
1422
1423 return 0;
1424}
1425
1426static void ipip6_netlink_parms(struct nlattr *data[],
1427 struct ip_tunnel_parm *parms)
1428{
1429 memset(parms, 0, sizeof(*parms));
1430
1431 parms->iph.version = 4;
1432 parms->iph.protocol = IPPROTO_IPV6;
1433 parms->iph.ihl = 5;
1434 parms->iph.ttl = 64;
1435
1436 if (!data)
1437 return;
1438
1439 if (data[IFLA_IPTUN_LINK])
1440 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1441
1442 if (data[IFLA_IPTUN_LOCAL])
1443 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1444
1445 if (data[IFLA_IPTUN_REMOTE])
1446 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1447
1448 if (data[IFLA_IPTUN_TTL]) {
1449 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1450 if (parms->iph.ttl)
1451 parms->iph.frag_off = htons(IP_DF);
1452 }
1453
1454 if (data[IFLA_IPTUN_TOS])
1455 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1456
1457 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1458 parms->iph.frag_off = htons(IP_DF);
1459
1460 if (data[IFLA_IPTUN_FLAGS])
1461 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1462
1463 if (data[IFLA_IPTUN_PROTO])
1464 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1465
1466}
1467
1468#ifdef CONFIG_IPV6_SIT_6RD
1469/* This function returns true when 6RD attributes are present in the nl msg */
1470static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1471 struct ip_tunnel_6rd *ip6rd)
1472{
1473 bool ret = false;
1474 memset(ip6rd, 0, sizeof(*ip6rd));
1475
1476 if (!data)
1477 return ret;
1478
1479 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1480 ret = true;
1481 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1482 sizeof(struct in6_addr));
1483 }
1484
1485 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1486 ret = true;
1487 ip6rd->relay_prefix =
1488 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1489 }
1490
1491 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1492 ret = true;
1493 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1494 }
1495
1496 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1497 ret = true;
1498 ip6rd->relay_prefixlen =
1499 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1500 }
1501
1502 return ret;
1503}
1504#endif
1505
1506static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1507 struct nlattr *tb[], struct nlattr *data[])
1508{
1509 struct net *net = dev_net(dev);
1510 struct ip_tunnel *nt;
1511#ifdef CONFIG_IPV6_SIT_6RD
1512 struct ip_tunnel_6rd ip6rd;
1513#endif
1514 int err;
1515
1516 nt = netdev_priv(dev);
1517 ipip6_netlink_parms(data, &nt->parms);
1518
1519 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1520 return -EEXIST;
1521
1522 err = ipip6_tunnel_create(dev);
1523 if (err < 0)
1524 return err;
1525
1526#ifdef CONFIG_IPV6_SIT_6RD
1527 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1528 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1529#endif
1530
1531 return err;
1532}
1533
1534static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1535 struct nlattr *data[])
1536{
1537 struct ip_tunnel *t = netdev_priv(dev);
1538 struct ip_tunnel_parm p;
1539 struct net *net = t->net;
1540 struct sit_net *sitn = net_generic(net, sit_net_id);
1541#ifdef CONFIG_IPV6_SIT_6RD
1542 struct ip_tunnel_6rd ip6rd;
1543#endif
1544
1545 if (dev == sitn->fb_tunnel_dev)
1546 return -EINVAL;
1547
1548 ipip6_netlink_parms(data, &p);
1549
1550 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1551 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1552 return -EINVAL;
1553
1554 t = ipip6_tunnel_locate(net, &p, 0);
1555
1556 if (t) {
1557 if (t->dev != dev)
1558 return -EEXIST;
1559 } else
1560 t = netdev_priv(dev);
1561
1562 ipip6_tunnel_update(t, &p);
1563
1564#ifdef CONFIG_IPV6_SIT_6RD
1565 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1566 return ipip6_tunnel_update_6rd(t, &ip6rd);
1567#endif
1568
1569 return 0;
1570}
1571
1572static size_t ipip6_get_size(const struct net_device *dev)
1573{
1574 return
1575 /* IFLA_IPTUN_LINK */
1576 nla_total_size(4) +
1577 /* IFLA_IPTUN_LOCAL */
1578 nla_total_size(4) +
1579 /* IFLA_IPTUN_REMOTE */
1580 nla_total_size(4) +
1581 /* IFLA_IPTUN_TTL */
1582 nla_total_size(1) +
1583 /* IFLA_IPTUN_TOS */
1584 nla_total_size(1) +
1585 /* IFLA_IPTUN_PMTUDISC */
1586 nla_total_size(1) +
1587 /* IFLA_IPTUN_FLAGS */
1588 nla_total_size(2) +
1589 /* IFLA_IPTUN_PROTO */
1590 nla_total_size(1) +
1591#ifdef CONFIG_IPV6_SIT_6RD
1592 /* IFLA_IPTUN_6RD_PREFIX */
1593 nla_total_size(sizeof(struct in6_addr)) +
1594 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1595 nla_total_size(4) +
1596 /* IFLA_IPTUN_6RD_PREFIXLEN */
1597 nla_total_size(2) +
1598 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1599 nla_total_size(2) +
1600#endif
1601 0;
1602}
1603
1604static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1605{
1606 struct ip_tunnel *tunnel = netdev_priv(dev);
1607 struct ip_tunnel_parm *parm = &tunnel->parms;
1608
1609 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1610 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1611 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1612 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1613 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1614 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1615 !!(parm->iph.frag_off & htons(IP_DF))) ||
1616 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1617 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1618 goto nla_put_failure;
1619
1620#ifdef CONFIG_IPV6_SIT_6RD
1621 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1622 &tunnel->ip6rd.prefix) ||
1623 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1624 tunnel->ip6rd.relay_prefix) ||
1625 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1626 tunnel->ip6rd.prefixlen) ||
1627 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1628 tunnel->ip6rd.relay_prefixlen))
1629 goto nla_put_failure;
1630#endif
1631
1632 return 0;
1633
1634nla_put_failure:
1635 return -EMSGSIZE;
1636}
1637
1638static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1639 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1640 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1641 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1642 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1643 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1644 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1645 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1646 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1647#ifdef CONFIG_IPV6_SIT_6RD
1648 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1649 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1650 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1651 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1652#endif
1653};
1654
1655static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1656{
1657 struct net *net = dev_net(dev);
1658 struct sit_net *sitn = net_generic(net, sit_net_id);
1659
1660 if (dev != sitn->fb_tunnel_dev)
1661 unregister_netdevice_queue(dev, head);
1662}
1663
1664static struct rtnl_link_ops sit_link_ops __read_mostly = {
1665 .kind = "sit",
1666 .maxtype = IFLA_IPTUN_MAX,
1667 .policy = ipip6_policy,
1668 .priv_size = sizeof(struct ip_tunnel),
1669 .setup = ipip6_tunnel_setup,
1670 .validate = ipip6_validate,
1671 .newlink = ipip6_newlink,
1672 .changelink = ipip6_changelink,
1673 .get_size = ipip6_get_size,
1674 .fill_info = ipip6_fill_info,
1675 .dellink = ipip6_dellink,
1676};
1677
1678static struct xfrm_tunnel sit_handler __read_mostly = {
1679 .handler = ipip6_rcv,
1680 .err_handler = ipip6_err,
1681 .priority = 1,
1682};
1683
1684static struct xfrm_tunnel ipip_handler __read_mostly = {
1685 .handler = ipip_rcv,
1686 .err_handler = ipip6_err,
1687 .priority = 2,
1688};
1689
1690static void __net_exit sit_destroy_tunnels(struct net *net,
1691 struct list_head *head)
1692{
1693 struct sit_net *sitn = net_generic(net, sit_net_id);
1694 struct net_device *dev, *aux;
1695 int prio;
1696
1697 for_each_netdev_safe(net, dev, aux)
1698 if (dev->rtnl_link_ops == &sit_link_ops)
1699 unregister_netdevice_queue(dev, head);
1700
1701 for (prio = 1; prio < 4; prio++) {
1702 int h;
1703 for (h = 0; h < HASH_SIZE; h++) {
1704 struct ip_tunnel *t;
1705
1706 t = rtnl_dereference(sitn->tunnels[prio][h]);
1707 while (t != NULL) {
1708 /* If dev is in the same netns, it has already
1709 * been added to the list by the previous loop.
1710 */
1711 if (!net_eq(dev_net(t->dev), net))
1712 unregister_netdevice_queue(t->dev,
1713 head);
1714 t = rtnl_dereference(t->next);
1715 }
1716 }
1717 }
1718}
1719
1720static int __net_init sit_init_net(struct net *net)
1721{
1722 struct sit_net *sitn = net_generic(net, sit_net_id);
1723 struct ip_tunnel *t;
1724 int err;
1725
1726 sitn->tunnels[0] = sitn->tunnels_wc;
1727 sitn->tunnels[1] = sitn->tunnels_l;
1728 sitn->tunnels[2] = sitn->tunnels_r;
1729 sitn->tunnels[3] = sitn->tunnels_r_l;
1730
1731 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1732 ipip6_tunnel_setup);
1733 if (!sitn->fb_tunnel_dev) {
1734 err = -ENOMEM;
1735 goto err_alloc_dev;
1736 }
1737 dev_net_set(sitn->fb_tunnel_dev, net);
1738 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1739 /* FB netdevice is special: we have one, and only one per netns.
1740 * Allowing to move it to another netns is clearly unsafe.
1741 */
1742 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1743
1744 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1745 if (err)
1746 goto err_dev_free;
1747
1748 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1749
1750 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1751 goto err_reg_dev;
1752
1753 t = netdev_priv(sitn->fb_tunnel_dev);
1754
1755 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1756 return 0;
1757
1758err_reg_dev:
1759 dev_put(sitn->fb_tunnel_dev);
1760err_dev_free:
1761 ipip6_dev_free(sitn->fb_tunnel_dev);
1762err_alloc_dev:
1763 return err;
1764}
1765
1766static void __net_exit sit_exit_net(struct net *net)
1767{
1768 LIST_HEAD(list);
1769
1770 rtnl_lock();
1771 sit_destroy_tunnels(net, &list);
1772 unregister_netdevice_many(&list);
1773 rtnl_unlock();
1774}
1775
1776static struct pernet_operations sit_net_ops = {
1777 .init = sit_init_net,
1778 .exit = sit_exit_net,
1779 .id = &sit_net_id,
1780 .size = sizeof(struct sit_net),
1781};
1782
1783static void __exit sit_cleanup(void)
1784{
1785 rtnl_link_unregister(&sit_link_ops);
1786 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1787 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1788
1789 unregister_pernet_device(&sit_net_ops);
1790 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1791}
1792
1793static int __init sit_init(void)
1794{
1795 int err;
1796
1797 pr_info("IPv6 over IPv4 tunneling driver\n");
1798
1799 err = register_pernet_device(&sit_net_ops);
1800 if (err < 0)
1801 return err;
1802 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1803 if (err < 0) {
1804 pr_info("%s: can't register ip6ip4\n", __func__);
1805 goto xfrm_tunnel_failed;
1806 }
1807 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1808 if (err < 0) {
1809 pr_info("%s: can't register ip4ip4\n", __func__);
1810 goto xfrm_tunnel4_failed;
1811 }
1812 err = rtnl_link_register(&sit_link_ops);
1813 if (err < 0)
1814 goto rtnl_link_failed;
1815
1816out:
1817 return err;
1818
1819rtnl_link_failed:
1820 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1821xfrm_tunnel4_failed:
1822 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1823xfrm_tunnel_failed:
1824 unregister_pernet_device(&sit_net_ops);
1825 goto out;
1826}
1827
1828module_init(sit_init);
1829module_exit(sit_cleanup);
1830MODULE_LICENSE("GPL");
1831MODULE_ALIAS_NETDEV("sit0");