1
2
3
4
5
6
7
8
9
10
11
12
13
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
57
58
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
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 strlcpy(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
320
321
322 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
323 kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
324 NULL;
325
326 rcu_read_lock();
327
328 ca = min(t->prl_count, cmax);
329
330 if (!kp) {
331
332
333
334
335
336 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
337 __GFP_NOWARN);
338 if (!kp) {
339 ret = -ENOMEM;
340 goto out;
341 }
342 }
343
344 c = 0;
345 for_each_prl_rcu(t->prl) {
346 if (c >= cmax)
347 break;
348 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
349 continue;
350 kp[c].addr = prl->addr;
351 kp[c].flags = prl->flags;
352 c++;
353 if (kprl.addr != htonl(INADDR_ANY))
354 break;
355 }
356out:
357 rcu_read_unlock();
358
359 len = sizeof(*kp) * c;
360 ret = 0;
361 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
362 ret = -EFAULT;
363
364 kfree(kp);
365
366 return ret;
367}
368
369static int
370ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
371{
372 struct ip_tunnel_prl_entry *p;
373 int err = 0;
374
375 if (a->addr == htonl(INADDR_ANY))
376 return -EINVAL;
377
378 ASSERT_RTNL();
379
380 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
381 if (p->addr == a->addr) {
382 if (chg) {
383 p->flags = a->flags;
384 goto out;
385 }
386 err = -EEXIST;
387 goto out;
388 }
389 }
390
391 if (chg) {
392 err = -ENXIO;
393 goto out;
394 }
395
396 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
397 if (!p) {
398 err = -ENOBUFS;
399 goto out;
400 }
401
402 p->next = t->prl;
403 p->addr = a->addr;
404 p->flags = a->flags;
405 t->prl_count++;
406 rcu_assign_pointer(t->prl, p);
407out:
408 return err;
409}
410
411static void prl_list_destroy_rcu(struct rcu_head *head)
412{
413 struct ip_tunnel_prl_entry *p, *n;
414
415 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
416 do {
417 n = rcu_dereference_protected(p->next, 1);
418 kfree(p);
419 p = n;
420 } while (p);
421}
422
423static int
424ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
425{
426 struct ip_tunnel_prl_entry *x;
427 struct ip_tunnel_prl_entry __rcu **p;
428 int err = 0;
429
430 ASSERT_RTNL();
431
432 if (a && a->addr != htonl(INADDR_ANY)) {
433 for (p = &t->prl;
434 (x = rtnl_dereference(*p)) != NULL;
435 p = &x->next) {
436 if (x->addr == a->addr) {
437 *p = x->next;
438 kfree_rcu(x, rcu_head);
439 t->prl_count--;
440 goto out;
441 }
442 }
443 err = -ENXIO;
444 } else {
445 x = rtnl_dereference(t->prl);
446 if (x) {
447 t->prl_count = 0;
448 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
449 t->prl = NULL;
450 }
451 }
452out:
453 return err;
454}
455
456static int ipip6_tunnel_prl_ctl(struct net_device *dev,
457 struct ip_tunnel_prl __user *data, int cmd)
458{
459 struct ip_tunnel *t = netdev_priv(dev);
460 struct ip_tunnel_prl prl;
461 int err;
462
463 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
464 return -EPERM;
465 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
466 return -EINVAL;
467
468 if (copy_from_user(&prl, data, sizeof(prl)))
469 return -EFAULT;
470
471 switch (cmd) {
472 case SIOCDELPRL:
473 err = ipip6_tunnel_del_prl(t, &prl);
474 break;
475 case SIOCADDPRL:
476 case SIOCCHGPRL:
477 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
478 break;
479 }
480 dst_cache_reset(&t->dst_cache);
481 netdev_state_change(dev);
482 return err;
483}
484
485static int
486isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
487{
488 struct ip_tunnel_prl_entry *p;
489 int ok = 1;
490
491 rcu_read_lock();
492 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
493 if (p) {
494 if (p->flags & PRL_DEFAULT)
495 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
496 else
497 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
498 } else {
499 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
500
501 if (ipv6_addr_is_isatap(addr6) &&
502 (addr6->s6_addr32[3] == iph->saddr) &&
503 ipv6_chk_prefix(addr6, t->dev))
504 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
505 else
506 ok = 0;
507 }
508 rcu_read_unlock();
509 return ok;
510}
511
512static void ipip6_tunnel_uninit(struct net_device *dev)
513{
514 struct ip_tunnel *tunnel = netdev_priv(dev);
515 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
516
517 if (dev == sitn->fb_tunnel_dev) {
518 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
519 } else {
520 ipip6_tunnel_unlink(sitn, tunnel);
521 ipip6_tunnel_del_prl(tunnel, NULL);
522 }
523 dst_cache_reset(&tunnel->dst_cache);
524 dev_put_track(dev, &tunnel->dev_tracker);
525}
526
527static int ipip6_err(struct sk_buff *skb, u32 info)
528{
529 const struct iphdr *iph = (const struct iphdr *)skb->data;
530 const int type = icmp_hdr(skb)->type;
531 const int code = icmp_hdr(skb)->code;
532 unsigned int data_len = 0;
533 struct ip_tunnel *t;
534 int sifindex;
535 int err;
536
537 switch (type) {
538 default:
539 case ICMP_PARAMETERPROB:
540 return 0;
541
542 case ICMP_DEST_UNREACH:
543 switch (code) {
544 case ICMP_SR_FAILED:
545
546 return 0;
547 default:
548
549
550
551
552 break;
553 }
554 break;
555 case ICMP_TIME_EXCEEDED:
556 if (code != ICMP_EXC_TTL)
557 return 0;
558 data_len = icmp_hdr(skb)->un.reserved[1] * 4;
559 break;
560 case ICMP_REDIRECT:
561 break;
562 }
563
564 err = -ENOENT;
565
566 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
567 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
568 iph->daddr, iph->saddr, sifindex);
569 if (!t)
570 goto out;
571
572 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
573 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
574 t->parms.link, iph->protocol);
575 err = 0;
576 goto out;
577 }
578 if (type == ICMP_REDIRECT) {
579 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
580 iph->protocol);
581 err = 0;
582 goto out;
583 }
584
585 err = 0;
586 if (__in6_dev_get(skb->dev) &&
587 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
588 goto out;
589
590 if (t->parms.iph.daddr == 0)
591 goto out;
592
593 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
594 goto out;
595
596 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
597 t->err_count++;
598 else
599 t->err_count = 1;
600 t->err_time = jiffies;
601out:
602 return err;
603}
604
605static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
606 const struct in6_addr *v6addr)
607{
608 __be32 v4embed = 0;
609 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
610 return true;
611 return false;
612}
613
614
615
616
617
618
619
620
621
622
623
624
625
626static bool only_dnatted(const struct ip_tunnel *tunnel,
627 const struct in6_addr *v6dst)
628{
629 int prefix_len;
630
631#ifdef CONFIG_IPV6_SIT_6RD
632 prefix_len = tunnel->ip6rd.prefixlen + 32
633 - tunnel->ip6rd.relay_prefixlen;
634#else
635 prefix_len = 48;
636#endif
637 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
638}
639
640
641static bool packet_is_spoofed(struct sk_buff *skb,
642 const struct iphdr *iph,
643 struct ip_tunnel *tunnel)
644{
645 const struct ipv6hdr *ipv6h;
646
647 if (tunnel->dev->priv_flags & IFF_ISATAP) {
648 if (!isatap_chksrc(skb, iph, tunnel))
649 return true;
650
651 return false;
652 }
653
654 if (tunnel->dev->flags & IFF_POINTOPOINT)
655 return false;
656
657 ipv6h = ipv6_hdr(skb);
658
659 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
660 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661 &iph->saddr, &ipv6h->saddr,
662 &iph->daddr, &ipv6h->daddr);
663 return true;
664 }
665
666 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
667 return false;
668
669 if (only_dnatted(tunnel, &ipv6h->daddr))
670 return false;
671
672 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
673 &iph->saddr, &ipv6h->saddr,
674 &iph->daddr, &ipv6h->daddr);
675 return true;
676}
677
678static int ipip6_rcv(struct sk_buff *skb)
679{
680 const struct iphdr *iph = ip_hdr(skb);
681 struct ip_tunnel *tunnel;
682 int sifindex;
683 int err;
684
685 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
686 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
687 iph->saddr, iph->daddr, sifindex);
688 if (tunnel) {
689 struct pcpu_sw_netstats *tstats;
690
691 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
692 tunnel->parms.iph.protocol != 0)
693 goto out;
694
695 skb->mac_header = skb->network_header;
696 skb_reset_network_header(skb);
697 IPCB(skb)->flags = 0;
698 skb->dev = tunnel->dev;
699
700 if (packet_is_spoofed(skb, iph, tunnel)) {
701 tunnel->dev->stats.rx_errors++;
702 goto out;
703 }
704
705 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
706 !net_eq(tunnel->net, dev_net(tunnel->dev))))
707 goto out;
708
709
710
711
712 iph = (const struct iphdr *)skb_mac_header(skb);
713 skb_reset_mac_header(skb);
714
715 err = IP_ECN_decapsulate(iph, skb);
716 if (unlikely(err)) {
717 if (log_ecn_error)
718 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
719 &iph->saddr, iph->tos);
720 if (err > 1) {
721 ++tunnel->dev->stats.rx_frame_errors;
722 ++tunnel->dev->stats.rx_errors;
723 goto out;
724 }
725 }
726
727 tstats = this_cpu_ptr(tunnel->dev->tstats);
728 u64_stats_update_begin(&tstats->syncp);
729 tstats->rx_packets++;
730 tstats->rx_bytes += skb->len;
731 u64_stats_update_end(&tstats->syncp);
732
733 netif_rx(skb);
734
735 return 0;
736 }
737
738
739 return 1;
740out:
741 kfree_skb(skb);
742 return 0;
743}
744
745static const struct tnl_ptk_info ipip_tpi = {
746
747 .proto = htons(ETH_P_IP),
748};
749
750#if IS_ENABLED(CONFIG_MPLS)
751static const struct tnl_ptk_info mplsip_tpi = {
752
753 .proto = htons(ETH_P_MPLS_UC),
754};
755#endif
756
757static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
758{
759 const struct iphdr *iph;
760 struct ip_tunnel *tunnel;
761 int sifindex;
762
763 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
764
765 iph = ip_hdr(skb);
766 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
767 iph->saddr, iph->daddr, sifindex);
768 if (tunnel) {
769 const struct tnl_ptk_info *tpi;
770
771 if (tunnel->parms.iph.protocol != ipproto &&
772 tunnel->parms.iph.protocol != 0)
773 goto drop;
774
775 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
776 goto drop;
777#if IS_ENABLED(CONFIG_MPLS)
778 if (ipproto == IPPROTO_MPLS)
779 tpi = &mplsip_tpi;
780 else
781#endif
782 tpi = &ipip_tpi;
783 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
784 goto drop;
785 skb_reset_mac_header(skb);
786
787 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
788 }
789
790 return 1;
791
792drop:
793 kfree_skb(skb);
794 return 0;
795}
796
797static int ipip_rcv(struct sk_buff *skb)
798{
799 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
800}
801
802#if IS_ENABLED(CONFIG_MPLS)
803static int mplsip_rcv(struct sk_buff *skb)
804{
805 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
806}
807#endif
808
809
810
811
812
813static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
814 __be32 *v4dst)
815{
816#ifdef CONFIG_IPV6_SIT_6RD
817 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
818 tunnel->ip6rd.prefixlen)) {
819 unsigned int pbw0, pbi0;
820 int pbi1;
821 u32 d;
822
823 pbw0 = tunnel->ip6rd.prefixlen >> 5;
824 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
825
826 d = tunnel->ip6rd.relay_prefixlen < 32 ?
827 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
828 tunnel->ip6rd.relay_prefixlen : 0;
829
830 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
831 if (pbi1 > 0)
832 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
833 (32 - pbi1);
834
835 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
836 return true;
837 }
838#else
839 if (v6dst->s6_addr16[0] == htons(0x2002)) {
840
841 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
842 return true;
843 }
844#endif
845 return false;
846}
847
848static inline __be32 try_6rd(struct ip_tunnel *tunnel,
849 const struct in6_addr *v6dst)
850{
851 __be32 dst = 0;
852 check_6rd(tunnel, v6dst, &dst);
853 return dst;
854}
855
856
857
858
859
860
861static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
862 struct net_device *dev)
863{
864 struct ip_tunnel *tunnel = netdev_priv(dev);
865 const struct iphdr *tiph = &tunnel->parms.iph;
866 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
867 u8 tos = tunnel->parms.iph.tos;
868 __be16 df = tiph->frag_off;
869 struct rtable *rt;
870 struct net_device *tdev;
871 unsigned int max_headroom;
872 __be32 dst = tiph->daddr;
873 struct flowi4 fl4;
874 int mtu;
875 const struct in6_addr *addr6;
876 int addr_type;
877 u8 ttl;
878 u8 protocol = IPPROTO_IPV6;
879 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
880
881 if (tos == 1)
882 tos = ipv6_get_dsfield(iph6);
883
884
885 if (dev->priv_flags & IFF_ISATAP) {
886 struct neighbour *neigh = NULL;
887 bool do_tx_error = false;
888
889 if (skb_dst(skb))
890 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
891
892 if (!neigh) {
893 net_dbg_ratelimited("nexthop == NULL\n");
894 goto tx_error;
895 }
896
897 addr6 = (const struct in6_addr *)&neigh->primary_key;
898 addr_type = ipv6_addr_type(addr6);
899
900 if ((addr_type & IPV6_ADDR_UNICAST) &&
901 ipv6_addr_is_isatap(addr6))
902 dst = addr6->s6_addr32[3];
903 else
904 do_tx_error = true;
905
906 neigh_release(neigh);
907 if (do_tx_error)
908 goto tx_error;
909 }
910
911 if (!dst)
912 dst = try_6rd(tunnel, &iph6->daddr);
913
914 if (!dst) {
915 struct neighbour *neigh = NULL;
916 bool do_tx_error = false;
917
918 if (skb_dst(skb))
919 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
920
921 if (!neigh) {
922 net_dbg_ratelimited("nexthop == NULL\n");
923 goto tx_error;
924 }
925
926 addr6 = (const struct in6_addr *)&neigh->primary_key;
927 addr_type = ipv6_addr_type(addr6);
928
929 if (addr_type == IPV6_ADDR_ANY) {
930 addr6 = &ipv6_hdr(skb)->daddr;
931 addr_type = ipv6_addr_type(addr6);
932 }
933
934 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
935 dst = addr6->s6_addr32[3];
936 else
937 do_tx_error = true;
938
939 neigh_release(neigh);
940 if (do_tx_error)
941 goto tx_error;
942 }
943
944 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
945 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
946 0, dst, tiph->saddr, 0, 0,
947 sock_net_uid(tunnel->net, NULL));
948
949 rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
950 if (!rt) {
951 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
952 if (IS_ERR(rt)) {
953 dev->stats.tx_carrier_errors++;
954 goto tx_error_icmp;
955 }
956 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
957 }
958
959 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
960 ip_rt_put(rt);
961 dev->stats.tx_carrier_errors++;
962 goto tx_error_icmp;
963 }
964 tdev = rt->dst.dev;
965
966 if (tdev == dev) {
967 ip_rt_put(rt);
968 dev->stats.collisions++;
969 goto tx_error;
970 }
971
972 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
973 ip_rt_put(rt);
974 goto tx_error;
975 }
976
977 if (df) {
978 mtu = dst_mtu(&rt->dst) - t_hlen;
979
980 if (mtu < IPV4_MIN_MTU) {
981 dev->stats.collisions++;
982 ip_rt_put(rt);
983 goto tx_error;
984 }
985
986 if (mtu < IPV6_MIN_MTU) {
987 mtu = IPV6_MIN_MTU;
988 df = 0;
989 }
990
991 if (tunnel->parms.iph.daddr)
992 skb_dst_update_pmtu_no_confirm(skb, mtu);
993
994 if (skb->len > mtu && !skb_is_gso(skb)) {
995 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
996 ip_rt_put(rt);
997 goto tx_error;
998 }
999 }
1000
1001 if (tunnel->err_count > 0) {
1002 if (time_before(jiffies,
1003 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
1004 tunnel->err_count--;
1005 dst_link_failure(skb);
1006 } else
1007 tunnel->err_count = 0;
1008 }
1009
1010
1011
1012
1013 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1014
1015 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1016 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1017 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1018 if (!new_skb) {
1019 ip_rt_put(rt);
1020 dev->stats.tx_dropped++;
1021 kfree_skb(skb);
1022 return NETDEV_TX_OK;
1023 }
1024 if (skb->sk)
1025 skb_set_owner_w(new_skb, skb->sk);
1026 dev_kfree_skb(skb);
1027 skb = new_skb;
1028 iph6 = ipv6_hdr(skb);
1029 }
1030 ttl = tiph->ttl;
1031 if (ttl == 0)
1032 ttl = iph6->hop_limit;
1033 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1034
1035 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1036 ip_rt_put(rt);
1037 goto tx_error;
1038 }
1039
1040 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1041
1042 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1043 df, !net_eq(tunnel->net, dev_net(dev)));
1044 return NETDEV_TX_OK;
1045
1046tx_error_icmp:
1047 dst_link_failure(skb);
1048tx_error:
1049 kfree_skb(skb);
1050 dev->stats.tx_errors++;
1051 return NETDEV_TX_OK;
1052}
1053
1054static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1055 struct net_device *dev, u8 ipproto)
1056{
1057 struct ip_tunnel *tunnel = netdev_priv(dev);
1058 const struct iphdr *tiph = &tunnel->parms.iph;
1059
1060 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1061 goto tx_error;
1062
1063 skb_set_inner_ipproto(skb, ipproto);
1064
1065 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1066 return NETDEV_TX_OK;
1067tx_error:
1068 kfree_skb(skb);
1069 dev->stats.tx_errors++;
1070 return NETDEV_TX_OK;
1071}
1072
1073static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1074 struct net_device *dev)
1075{
1076 if (!pskb_inet_may_pull(skb))
1077 goto tx_err;
1078
1079 switch (skb->protocol) {
1080 case htons(ETH_P_IP):
1081 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1082 break;
1083 case htons(ETH_P_IPV6):
1084 ipip6_tunnel_xmit(skb, dev);
1085 break;
1086#if IS_ENABLED(CONFIG_MPLS)
1087 case htons(ETH_P_MPLS_UC):
1088 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1089 break;
1090#endif
1091 default:
1092 goto tx_err;
1093 }
1094
1095 return NETDEV_TX_OK;
1096
1097tx_err:
1098 dev->stats.tx_errors++;
1099 kfree_skb(skb);
1100 return NETDEV_TX_OK;
1101
1102}
1103
1104static void ipip6_tunnel_bind_dev(struct net_device *dev)
1105{
1106 struct net_device *tdev = NULL;
1107 struct ip_tunnel *tunnel;
1108 const struct iphdr *iph;
1109 struct flowi4 fl4;
1110
1111 tunnel = netdev_priv(dev);
1112 iph = &tunnel->parms.iph;
1113
1114 if (iph->daddr) {
1115 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1116 NULL,
1117 iph->daddr, iph->saddr,
1118 0, 0,
1119 IPPROTO_IPV6,
1120 RT_TOS(iph->tos),
1121 tunnel->parms.link);
1122
1123 if (!IS_ERR(rt)) {
1124 tdev = rt->dst.dev;
1125 ip_rt_put(rt);
1126 }
1127 dev->flags |= IFF_POINTOPOINT;
1128 }
1129
1130 if (!tdev && tunnel->parms.link)
1131 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1132
1133 if (tdev && !netif_is_l3_master(tdev)) {
1134 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1135
1136 dev->mtu = tdev->mtu - t_hlen;
1137 if (dev->mtu < IPV6_MIN_MTU)
1138 dev->mtu = IPV6_MIN_MTU;
1139 }
1140}
1141
1142static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1143 __u32 fwmark)
1144{
1145 struct net *net = t->net;
1146 struct sit_net *sitn = net_generic(net, sit_net_id);
1147
1148 ipip6_tunnel_unlink(sitn, t);
1149 synchronize_net();
1150 t->parms.iph.saddr = p->iph.saddr;
1151 t->parms.iph.daddr = p->iph.daddr;
1152 __dev_addr_set(t->dev, &p->iph.saddr, 4);
1153 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1154 ipip6_tunnel_link(sitn, t);
1155 t->parms.iph.ttl = p->iph.ttl;
1156 t->parms.iph.tos = p->iph.tos;
1157 t->parms.iph.frag_off = p->iph.frag_off;
1158 if (t->parms.link != p->link || t->fwmark != fwmark) {
1159 t->parms.link = p->link;
1160 t->fwmark = fwmark;
1161 ipip6_tunnel_bind_dev(t->dev);
1162 }
1163 dst_cache_reset(&t->dst_cache);
1164 netdev_state_change(t->dev);
1165}
1166
1167#ifdef CONFIG_IPV6_SIT_6RD
1168static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1169 struct ip_tunnel_6rd *ip6rd)
1170{
1171 struct in6_addr prefix;
1172 __be32 relay_prefix;
1173
1174 if (ip6rd->relay_prefixlen > 32 ||
1175 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1176 return -EINVAL;
1177
1178 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1179 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1180 return -EINVAL;
1181 if (ip6rd->relay_prefixlen)
1182 relay_prefix = ip6rd->relay_prefix &
1183 htonl(0xffffffffUL <<
1184 (32 - ip6rd->relay_prefixlen));
1185 else
1186 relay_prefix = 0;
1187 if (relay_prefix != ip6rd->relay_prefix)
1188 return -EINVAL;
1189
1190 t->ip6rd.prefix = prefix;
1191 t->ip6rd.relay_prefix = relay_prefix;
1192 t->ip6rd.prefixlen = ip6rd->prefixlen;
1193 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1194 dst_cache_reset(&t->dst_cache);
1195 netdev_state_change(t->dev);
1196 return 0;
1197}
1198
1199static int
1200ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1201{
1202 struct ip_tunnel *t = netdev_priv(dev);
1203 struct ip_tunnel_6rd ip6rd;
1204 struct ip_tunnel_parm p;
1205
1206 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1207 if (copy_from_user(&p, data, sizeof(p)))
1208 return -EFAULT;
1209 t = ipip6_tunnel_locate(t->net, &p, 0);
1210 }
1211 if (!t)
1212 t = netdev_priv(dev);
1213
1214 ip6rd.prefix = t->ip6rd.prefix;
1215 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1216 ip6rd.prefixlen = t->ip6rd.prefixlen;
1217 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1218 if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1219 return -EFAULT;
1220 return 0;
1221}
1222
1223static int
1224ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1225 int cmd)
1226{
1227 struct ip_tunnel *t = netdev_priv(dev);
1228 struct ip_tunnel_6rd ip6rd;
1229 int err;
1230
1231 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1232 return -EPERM;
1233 if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1234 return -EFAULT;
1235
1236 if (cmd != SIOCDEL6RD) {
1237 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1238 if (err < 0)
1239 return err;
1240 } else
1241 ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1242 return 0;
1243}
1244
1245#endif
1246
1247static bool ipip6_valid_ip_proto(u8 ipproto)
1248{
1249 return ipproto == IPPROTO_IPV6 ||
1250 ipproto == IPPROTO_IPIP ||
1251#if IS_ENABLED(CONFIG_MPLS)
1252 ipproto == IPPROTO_MPLS ||
1253#endif
1254 ipproto == 0;
1255}
1256
1257static int
1258__ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1259{
1260 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1261 return -EPERM;
1262
1263 if (!ipip6_valid_ip_proto(p->iph.protocol))
1264 return -EINVAL;
1265 if (p->iph.version != 4 ||
1266 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1267 return -EINVAL;
1268
1269 if (p->iph.ttl)
1270 p->iph.frag_off |= htons(IP_DF);
1271 return 0;
1272}
1273
1274static int
1275ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1276{
1277 struct ip_tunnel *t = netdev_priv(dev);
1278
1279 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1280 t = ipip6_tunnel_locate(t->net, p, 0);
1281 if (!t)
1282 t = netdev_priv(dev);
1283 memcpy(p, &t->parms, sizeof(*p));
1284 return 0;
1285}
1286
1287static int
1288ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1289{
1290 struct ip_tunnel *t = netdev_priv(dev);
1291 int err;
1292
1293 err = __ipip6_tunnel_ioctl_validate(t->net, p);
1294 if (err)
1295 return err;
1296
1297 t = ipip6_tunnel_locate(t->net, p, 1);
1298 if (!t)
1299 return -ENOBUFS;
1300 return 0;
1301}
1302
1303static int
1304ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1305{
1306 struct ip_tunnel *t = netdev_priv(dev);
1307 int err;
1308
1309 err = __ipip6_tunnel_ioctl_validate(t->net, p);
1310 if (err)
1311 return err;
1312
1313 t = ipip6_tunnel_locate(t->net, p, 0);
1314 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1315 if (!t)
1316 return -ENOENT;
1317 } else {
1318 if (t) {
1319 if (t->dev != dev)
1320 return -EEXIST;
1321 } else {
1322 if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1323 (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1324 return -EINVAL;
1325 t = netdev_priv(dev);
1326 }
1327
1328 ipip6_tunnel_update(t, p, t->fwmark);
1329 }
1330
1331 return 0;
1332}
1333
1334static int
1335ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1336{
1337 struct ip_tunnel *t = netdev_priv(dev);
1338
1339 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1340 return -EPERM;
1341
1342 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1343 t = ipip6_tunnel_locate(t->net, p, 0);
1344 if (!t)
1345 return -ENOENT;
1346 if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1347 return -EPERM;
1348 dev = t->dev;
1349 }
1350 unregister_netdevice(dev);
1351 return 0;
1352}
1353
1354static int
1355ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1356{
1357 switch (cmd) {
1358 case SIOCGETTUNNEL:
1359 return ipip6_tunnel_get(dev, p);
1360 case SIOCADDTUNNEL:
1361 return ipip6_tunnel_add(dev, p);
1362 case SIOCCHGTUNNEL:
1363 return ipip6_tunnel_change(dev, p);
1364 case SIOCDELTUNNEL:
1365 return ipip6_tunnel_del(dev, p);
1366 default:
1367 return -EINVAL;
1368 }
1369}
1370
1371static int
1372ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1373 void __user *data, int cmd)
1374{
1375 switch (cmd) {
1376 case SIOCGETTUNNEL:
1377 case SIOCADDTUNNEL:
1378 case SIOCCHGTUNNEL:
1379 case SIOCDELTUNNEL:
1380 return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1381 case SIOCGETPRL:
1382 return ipip6_tunnel_get_prl(dev, data);
1383 case SIOCADDPRL:
1384 case SIOCDELPRL:
1385 case SIOCCHGPRL:
1386 return ipip6_tunnel_prl_ctl(dev, data, cmd);
1387#ifdef CONFIG_IPV6_SIT_6RD
1388 case SIOCGET6RD:
1389 return ipip6_tunnel_get6rd(dev, data);
1390 case SIOCADD6RD:
1391 case SIOCCHG6RD:
1392 case SIOCDEL6RD:
1393 return ipip6_tunnel_6rdctl(dev, data, cmd);
1394#endif
1395 default:
1396 return -EINVAL;
1397 }
1398}
1399
1400static const struct net_device_ops ipip6_netdev_ops = {
1401 .ndo_init = ipip6_tunnel_init,
1402 .ndo_uninit = ipip6_tunnel_uninit,
1403 .ndo_start_xmit = sit_tunnel_xmit,
1404 .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1405 .ndo_get_stats64 = dev_get_tstats64,
1406 .ndo_get_iflink = ip_tunnel_get_iflink,
1407 .ndo_tunnel_ctl = ipip6_tunnel_ctl,
1408};
1409
1410static void ipip6_dev_free(struct net_device *dev)
1411{
1412 struct ip_tunnel *tunnel = netdev_priv(dev);
1413
1414 dst_cache_destroy(&tunnel->dst_cache);
1415 free_percpu(dev->tstats);
1416}
1417
1418#define SIT_FEATURES (NETIF_F_SG | \
1419 NETIF_F_FRAGLIST | \
1420 NETIF_F_HIGHDMA | \
1421 NETIF_F_GSO_SOFTWARE | \
1422 NETIF_F_HW_CSUM)
1423
1424static void ipip6_tunnel_setup(struct net_device *dev)
1425{
1426 struct ip_tunnel *tunnel = netdev_priv(dev);
1427 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1428
1429 dev->netdev_ops = &ipip6_netdev_ops;
1430 dev->header_ops = &ip_tunnel_header_ops;
1431 dev->needs_free_netdev = true;
1432 dev->priv_destructor = ipip6_dev_free;
1433
1434 dev->type = ARPHRD_SIT;
1435 dev->mtu = ETH_DATA_LEN - t_hlen;
1436 dev->min_mtu = IPV6_MIN_MTU;
1437 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1438 dev->flags = IFF_NOARP;
1439 netif_keep_dst(dev);
1440 dev->addr_len = 4;
1441 dev->features |= NETIF_F_LLTX;
1442 dev->features |= SIT_FEATURES;
1443 dev->hw_features |= SIT_FEATURES;
1444}
1445
1446static int ipip6_tunnel_init(struct net_device *dev)
1447{
1448 struct ip_tunnel *tunnel = netdev_priv(dev);
1449 int err;
1450
1451 tunnel->dev = dev;
1452 tunnel->net = dev_net(dev);
1453 strcpy(tunnel->parms.name, dev->name);
1454
1455 ipip6_tunnel_bind_dev(dev);
1456 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1457 if (!dev->tstats)
1458 return -ENOMEM;
1459
1460 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1461 if (err) {
1462 free_percpu(dev->tstats);
1463 dev->tstats = NULL;
1464 return err;
1465 }
1466 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1467 return 0;
1468}
1469
1470static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1471{
1472 struct ip_tunnel *tunnel = netdev_priv(dev);
1473 struct iphdr *iph = &tunnel->parms.iph;
1474 struct net *net = dev_net(dev);
1475 struct sit_net *sitn = net_generic(net, sit_net_id);
1476
1477 iph->version = 4;
1478 iph->protocol = IPPROTO_IPV6;
1479 iph->ihl = 5;
1480 iph->ttl = 64;
1481
1482 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1483}
1484
1485static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1486 struct netlink_ext_ack *extack)
1487{
1488 u8 proto;
1489
1490 if (!data || !data[IFLA_IPTUN_PROTO])
1491 return 0;
1492
1493 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1494 if (!ipip6_valid_ip_proto(proto))
1495 return -EINVAL;
1496
1497 return 0;
1498}
1499
1500static void ipip6_netlink_parms(struct nlattr *data[],
1501 struct ip_tunnel_parm *parms,
1502 __u32 *fwmark)
1503{
1504 memset(parms, 0, sizeof(*parms));
1505
1506 parms->iph.version = 4;
1507 parms->iph.protocol = IPPROTO_IPV6;
1508 parms->iph.ihl = 5;
1509 parms->iph.ttl = 64;
1510
1511 if (!data)
1512 return;
1513
1514 if (data[IFLA_IPTUN_LINK])
1515 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1516
1517 if (data[IFLA_IPTUN_LOCAL])
1518 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1519
1520 if (data[IFLA_IPTUN_REMOTE])
1521 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1522
1523 if (data[IFLA_IPTUN_TTL]) {
1524 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1525 if (parms->iph.ttl)
1526 parms->iph.frag_off = htons(IP_DF);
1527 }
1528
1529 if (data[IFLA_IPTUN_TOS])
1530 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1531
1532 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1533 parms->iph.frag_off = htons(IP_DF);
1534
1535 if (data[IFLA_IPTUN_FLAGS])
1536 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1537
1538 if (data[IFLA_IPTUN_PROTO])
1539 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1540
1541 if (data[IFLA_IPTUN_FWMARK])
1542 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1543}
1544
1545
1546static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1547 struct ip_tunnel_encap *ipencap)
1548{
1549 bool ret = false;
1550
1551 memset(ipencap, 0, sizeof(*ipencap));
1552
1553 if (!data)
1554 return ret;
1555
1556 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1557 ret = true;
1558 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1559 }
1560
1561 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1562 ret = true;
1563 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1564 }
1565
1566 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1567 ret = true;
1568 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1569 }
1570
1571 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1572 ret = true;
1573 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1574 }
1575
1576 return ret;
1577}
1578
1579#ifdef CONFIG_IPV6_SIT_6RD
1580
1581static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1582 struct ip_tunnel_6rd *ip6rd)
1583{
1584 bool ret = false;
1585 memset(ip6rd, 0, sizeof(*ip6rd));
1586
1587 if (!data)
1588 return ret;
1589
1590 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1591 ret = true;
1592 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1593 }
1594
1595 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1596 ret = true;
1597 ip6rd->relay_prefix =
1598 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1599 }
1600
1601 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1602 ret = true;
1603 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1604 }
1605
1606 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1607 ret = true;
1608 ip6rd->relay_prefixlen =
1609 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1610 }
1611
1612 return ret;
1613}
1614#endif
1615
1616static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1617 struct nlattr *tb[], struct nlattr *data[],
1618 struct netlink_ext_ack *extack)
1619{
1620 struct net *net = dev_net(dev);
1621 struct ip_tunnel *nt;
1622 struct ip_tunnel_encap ipencap;
1623#ifdef CONFIG_IPV6_SIT_6RD
1624 struct ip_tunnel_6rd ip6rd;
1625#endif
1626 int err;
1627
1628 nt = netdev_priv(dev);
1629
1630 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1631 err = ip_tunnel_encap_setup(nt, &ipencap);
1632 if (err < 0)
1633 return err;
1634 }
1635
1636 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1637
1638 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1639 return -EEXIST;
1640
1641 err = ipip6_tunnel_create(dev);
1642 if (err < 0)
1643 return err;
1644
1645 if (tb[IFLA_MTU]) {
1646 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1647
1648 if (mtu >= IPV6_MIN_MTU &&
1649 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1650 dev->mtu = mtu;
1651 }
1652
1653#ifdef CONFIG_IPV6_SIT_6RD
1654 if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1655 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1656 if (err < 0)
1657 unregister_netdevice_queue(dev, NULL);
1658 }
1659#endif
1660
1661 return err;
1662}
1663
1664static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1665 struct nlattr *data[],
1666 struct netlink_ext_ack *extack)
1667{
1668 struct ip_tunnel *t = netdev_priv(dev);
1669 struct ip_tunnel_parm p;
1670 struct ip_tunnel_encap ipencap;
1671 struct net *net = t->net;
1672 struct sit_net *sitn = net_generic(net, sit_net_id);
1673#ifdef CONFIG_IPV6_SIT_6RD
1674 struct ip_tunnel_6rd ip6rd;
1675#endif
1676 __u32 fwmark = t->fwmark;
1677 int err;
1678
1679 if (dev == sitn->fb_tunnel_dev)
1680 return -EINVAL;
1681
1682 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1683 err = ip_tunnel_encap_setup(t, &ipencap);
1684 if (err < 0)
1685 return err;
1686 }
1687
1688 ipip6_netlink_parms(data, &p, &fwmark);
1689
1690 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1691 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1692 return -EINVAL;
1693
1694 t = ipip6_tunnel_locate(net, &p, 0);
1695
1696 if (t) {
1697 if (t->dev != dev)
1698 return -EEXIST;
1699 } else
1700 t = netdev_priv(dev);
1701
1702 ipip6_tunnel_update(t, &p, fwmark);
1703
1704#ifdef CONFIG_IPV6_SIT_6RD
1705 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1706 return ipip6_tunnel_update_6rd(t, &ip6rd);
1707#endif
1708
1709 return 0;
1710}
1711
1712static size_t ipip6_get_size(const struct net_device *dev)
1713{
1714 return
1715
1716 nla_total_size(4) +
1717
1718 nla_total_size(4) +
1719
1720 nla_total_size(4) +
1721
1722 nla_total_size(1) +
1723
1724 nla_total_size(1) +
1725
1726 nla_total_size(1) +
1727
1728 nla_total_size(2) +
1729
1730 nla_total_size(1) +
1731#ifdef CONFIG_IPV6_SIT_6RD
1732
1733 nla_total_size(sizeof(struct in6_addr)) +
1734
1735 nla_total_size(4) +
1736
1737 nla_total_size(2) +
1738
1739 nla_total_size(2) +
1740#endif
1741
1742 nla_total_size(2) +
1743
1744 nla_total_size(2) +
1745
1746 nla_total_size(2) +
1747
1748 nla_total_size(2) +
1749
1750 nla_total_size(4) +
1751 0;
1752}
1753
1754static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1755{
1756 struct ip_tunnel *tunnel = netdev_priv(dev);
1757 struct ip_tunnel_parm *parm = &tunnel->parms;
1758
1759 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1760 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1761 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1762 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1763 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1764 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1765 !!(parm->iph.frag_off & htons(IP_DF))) ||
1766 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1767 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1768 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1769 goto nla_put_failure;
1770
1771#ifdef CONFIG_IPV6_SIT_6RD
1772 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1773 &tunnel->ip6rd.prefix) ||
1774 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1775 tunnel->ip6rd.relay_prefix) ||
1776 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1777 tunnel->ip6rd.prefixlen) ||
1778 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1779 tunnel->ip6rd.relay_prefixlen))
1780 goto nla_put_failure;
1781#endif
1782
1783 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1784 tunnel->encap.type) ||
1785 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1786 tunnel->encap.sport) ||
1787 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1788 tunnel->encap.dport) ||
1789 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1790 tunnel->encap.flags))
1791 goto nla_put_failure;
1792
1793 return 0;
1794
1795nla_put_failure:
1796 return -EMSGSIZE;
1797}
1798
1799static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1800 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1801 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1802 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1803 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1804 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1805 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1806 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1807 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1808#ifdef CONFIG_IPV6_SIT_6RD
1809 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1810 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1811 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1812 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1813#endif
1814 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1815 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1816 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1817 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1818 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1819};
1820
1821static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1822{
1823 struct net *net = dev_net(dev);
1824 struct sit_net *sitn = net_generic(net, sit_net_id);
1825
1826 if (dev != sitn->fb_tunnel_dev)
1827 unregister_netdevice_queue(dev, head);
1828}
1829
1830static struct rtnl_link_ops sit_link_ops __read_mostly = {
1831 .kind = "sit",
1832 .maxtype = IFLA_IPTUN_MAX,
1833 .policy = ipip6_policy,
1834 .priv_size = sizeof(struct ip_tunnel),
1835 .setup = ipip6_tunnel_setup,
1836 .validate = ipip6_validate,
1837 .newlink = ipip6_newlink,
1838 .changelink = ipip6_changelink,
1839 .get_size = ipip6_get_size,
1840 .fill_info = ipip6_fill_info,
1841 .dellink = ipip6_dellink,
1842 .get_link_net = ip_tunnel_get_link_net,
1843};
1844
1845static struct xfrm_tunnel sit_handler __read_mostly = {
1846 .handler = ipip6_rcv,
1847 .err_handler = ipip6_err,
1848 .priority = 1,
1849};
1850
1851static struct xfrm_tunnel ipip_handler __read_mostly = {
1852 .handler = ipip_rcv,
1853 .err_handler = ipip6_err,
1854 .priority = 2,
1855};
1856
1857#if IS_ENABLED(CONFIG_MPLS)
1858static struct xfrm_tunnel mplsip_handler __read_mostly = {
1859 .handler = mplsip_rcv,
1860 .err_handler = ipip6_err,
1861 .priority = 2,
1862};
1863#endif
1864
1865static void __net_exit sit_destroy_tunnels(struct net *net,
1866 struct list_head *head)
1867{
1868 struct sit_net *sitn = net_generic(net, sit_net_id);
1869 struct net_device *dev, *aux;
1870 int prio;
1871
1872 for_each_netdev_safe(net, dev, aux)
1873 if (dev->rtnl_link_ops == &sit_link_ops)
1874 unregister_netdevice_queue(dev, head);
1875
1876 for (prio = 0; prio < 4; prio++) {
1877 int h;
1878 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1879 struct ip_tunnel *t;
1880
1881 t = rtnl_dereference(sitn->tunnels[prio][h]);
1882 while (t) {
1883
1884
1885
1886 if (!net_eq(dev_net(t->dev), net))
1887 unregister_netdevice_queue(t->dev,
1888 head);
1889 t = rtnl_dereference(t->next);
1890 }
1891 }
1892 }
1893}
1894
1895static int __net_init sit_init_net(struct net *net)
1896{
1897 struct sit_net *sitn = net_generic(net, sit_net_id);
1898 struct ip_tunnel *t;
1899 int err;
1900
1901 sitn->tunnels[0] = sitn->tunnels_wc;
1902 sitn->tunnels[1] = sitn->tunnels_l;
1903 sitn->tunnels[2] = sitn->tunnels_r;
1904 sitn->tunnels[3] = sitn->tunnels_r_l;
1905
1906 if (!net_has_fallback_tunnels(net))
1907 return 0;
1908
1909 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1910 NET_NAME_UNKNOWN,
1911 ipip6_tunnel_setup);
1912 if (!sitn->fb_tunnel_dev) {
1913 err = -ENOMEM;
1914 goto err_alloc_dev;
1915 }
1916 dev_net_set(sitn->fb_tunnel_dev, net);
1917 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1918
1919
1920
1921 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1922
1923 err = register_netdev(sitn->fb_tunnel_dev);
1924 if (err)
1925 goto err_reg_dev;
1926
1927 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1928 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1929
1930 t = netdev_priv(sitn->fb_tunnel_dev);
1931
1932 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1933 return 0;
1934
1935err_reg_dev:
1936 free_netdev(sitn->fb_tunnel_dev);
1937err_alloc_dev:
1938 return err;
1939}
1940
1941static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1942{
1943 LIST_HEAD(list);
1944 struct net *net;
1945
1946 rtnl_lock();
1947 list_for_each_entry(net, net_list, exit_list)
1948 sit_destroy_tunnels(net, &list);
1949
1950 unregister_netdevice_many(&list);
1951 rtnl_unlock();
1952}
1953
1954static struct pernet_operations sit_net_ops = {
1955 .init = sit_init_net,
1956 .exit_batch = sit_exit_batch_net,
1957 .id = &sit_net_id,
1958 .size = sizeof(struct sit_net),
1959};
1960
1961static void __exit sit_cleanup(void)
1962{
1963 rtnl_link_unregister(&sit_link_ops);
1964 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1965 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1966#if IS_ENABLED(CONFIG_MPLS)
1967 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1968#endif
1969
1970 unregister_pernet_device(&sit_net_ops);
1971 rcu_barrier();
1972}
1973
1974static int __init sit_init(void)
1975{
1976 int err;
1977
1978 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1979
1980 err = register_pernet_device(&sit_net_ops);
1981 if (err < 0)
1982 return err;
1983 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1984 if (err < 0) {
1985 pr_info("%s: can't register ip6ip4\n", __func__);
1986 goto xfrm_tunnel_failed;
1987 }
1988 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1989 if (err < 0) {
1990 pr_info("%s: can't register ip4ip4\n", __func__);
1991 goto xfrm_tunnel4_failed;
1992 }
1993#if IS_ENABLED(CONFIG_MPLS)
1994 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1995 if (err < 0) {
1996 pr_info("%s: can't register mplsip\n", __func__);
1997 goto xfrm_tunnel_mpls_failed;
1998 }
1999#endif
2000 err = rtnl_link_register(&sit_link_ops);
2001 if (err < 0)
2002 goto rtnl_link_failed;
2003
2004out:
2005 return err;
2006
2007rtnl_link_failed:
2008#if IS_ENABLED(CONFIG_MPLS)
2009 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2010xfrm_tunnel_mpls_failed:
2011#endif
2012 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2013xfrm_tunnel4_failed:
2014 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2015xfrm_tunnel_failed:
2016 unregister_pernet_device(&sit_net_ops);
2017 goto out;
2018}
2019
2020module_init(sit_init);
2021module_exit(sit_cleanup);
2022MODULE_LICENSE("GPL");
2023MODULE_ALIAS_RTNL_LINK("sit");
2024MODULE_ALIAS_NETDEV("sit0");
2025
