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 <linux/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 IP6_SIT_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 unsigned int sit_net_id __read_mostly;
80struct sit_net {
81 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
82 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
83 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_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->ifindex == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
106 return t;
107 }
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111 (t->dev->flags & IFF_UP))
112 return t;
113 }
114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 if (local == t->parms.iph.saddr &&
116 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
118 return t;
119 }
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if (t && (t->dev->flags & IFF_UP))
122 return t;
123 return NULL;
124}
125
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 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
199 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
200
201 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
202 dev->priv_flags |= IFF_ISATAP;
203
204 dev->rtnl_link_ops = &sit_link_ops;
205
206 err = register_netdevice(dev);
207 if (err < 0)
208 goto out;
209
210 ipip6_tunnel_clone_6rd(dev, sitn);
211
212 dev_hold(dev);
213
214 ipip6_tunnel_link(sitn, t);
215 return 0;
216
217out:
218 return err;
219}
220
221static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
222 struct ip_tunnel_parm *parms, int create)
223{
224 __be32 remote = parms->iph.daddr;
225 __be32 local = parms->iph.saddr;
226 struct ip_tunnel *t, *nt;
227 struct ip_tunnel __rcu **tp;
228 struct net_device *dev;
229 char name[IFNAMSIZ];
230 struct sit_net *sitn = net_generic(net, sit_net_id);
231
232 for (tp = __ipip6_bucket(sitn, parms);
233 (t = rtnl_dereference(*tp)) != NULL;
234 tp = &t->next) {
235 if (local == t->parms.iph.saddr &&
236 remote == t->parms.iph.daddr &&
237 parms->link == t->parms.link) {
238 if (create)
239 return NULL;
240 else
241 return t;
242 }
243 }
244 if (!create)
245 goto failed;
246
247 if (parms->name[0])
248 strlcpy(name, parms->name, IFNAMSIZ);
249 else
250 strcpy(name, "sit%d");
251
252 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
253 ipip6_tunnel_setup);
254 if (!dev)
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 dst_cache_reset(&tunnel->dst_cache);
479 dev_put(dev);
480}
481
482static int ipip6_err(struct sk_buff *skb, u32 info)
483{
484 const struct iphdr *iph = (const struct iphdr *)skb->data;
485 const int type = icmp_hdr(skb)->type;
486 const int code = icmp_hdr(skb)->code;
487 unsigned int data_len = 0;
488 struct ip_tunnel *t;
489 int err;
490
491 switch (type) {
492 default:
493 case ICMP_PARAMETERPROB:
494 return 0;
495
496 case ICMP_DEST_UNREACH:
497 switch (code) {
498 case ICMP_SR_FAILED:
499 /* Impossible event. */
500 return 0;
501 default:
502 /* All others are translated to HOST_UNREACH.
503 rfc2003 contains "deep thoughts" about NET_UNREACH,
504 I believe they are just ether pollution. --ANK
505 */
506 break;
507 }
508 break;
509 case ICMP_TIME_EXCEEDED:
510 if (code != ICMP_EXC_TTL)
511 return 0;
512 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
513 break;
514 case ICMP_REDIRECT:
515 break;
516 }
517
518 err = -ENOENT;
519
520 t = ipip6_tunnel_lookup(dev_net(skb->dev),
521 skb->dev,
522 iph->daddr,
523 iph->saddr);
524 if (!t)
525 goto out;
526
527 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
528 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
529 t->parms.link, 0, iph->protocol, 0);
530 err = 0;
531 goto out;
532 }
533 if (type == ICMP_REDIRECT) {
534 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
535 iph->protocol, 0);
536 err = 0;
537 goto out;
538 }
539
540 err = 0;
541 if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
542 goto out;
543
544 if (t->parms.iph.daddr == 0)
545 goto out;
546
547 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
548 goto out;
549
550 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
551 t->err_count++;
552 else
553 t->err_count = 1;
554 t->err_time = jiffies;
555out:
556 return err;
557}
558
559static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
560 const struct in6_addr *v6addr)
561{
562 __be32 v4embed = 0;
563 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
564 return true;
565 return false;
566}
567
568/* Checks if an address matches an address on the tunnel interface.
569 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
570 * Long story:
571 * This function is called after we considered the packet as spoofed
572 * in is_spoofed_6rd.
573 * We may have a router that is doing NAT for proto 41 packets
574 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
575 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
576 * function will return true, dropping the packet.
577 * But, we can still check if is spoofed against the IP
578 * addresses associated with the interface.
579 */
580static bool only_dnatted(const struct ip_tunnel *tunnel,
581 const struct in6_addr *v6dst)
582{
583 int prefix_len;
584
585#ifdef CONFIG_IPV6_SIT_6RD
586 prefix_len = tunnel->ip6rd.prefixlen + 32
587 - tunnel->ip6rd.relay_prefixlen;
588#else
589 prefix_len = 48;
590#endif
591 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
592}
593
594/* Returns true if a packet is spoofed */
595static bool packet_is_spoofed(struct sk_buff *skb,
596 const struct iphdr *iph,
597 struct ip_tunnel *tunnel)
598{
599 const struct ipv6hdr *ipv6h;
600
601 if (tunnel->dev->priv_flags & IFF_ISATAP) {
602 if (!isatap_chksrc(skb, iph, tunnel))
603 return true;
604
605 return false;
606 }
607
608 if (tunnel->dev->flags & IFF_POINTOPOINT)
609 return false;
610
611 ipv6h = ipv6_hdr(skb);
612
613 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
614 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
615 &iph->saddr, &ipv6h->saddr,
616 &iph->daddr, &ipv6h->daddr);
617 return true;
618 }
619
620 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
621 return false;
622
623 if (only_dnatted(tunnel, &ipv6h->daddr))
624 return false;
625
626 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
627 &iph->saddr, &ipv6h->saddr,
628 &iph->daddr, &ipv6h->daddr);
629 return true;
630}
631
632static int ipip6_rcv(struct sk_buff *skb)
633{
634 const struct iphdr *iph = ip_hdr(skb);
635 struct ip_tunnel *tunnel;
636 int err;
637
638 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
639 iph->saddr, iph->daddr);
640 if (tunnel) {
641 struct pcpu_sw_netstats *tstats;
642
643 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
644 tunnel->parms.iph.protocol != 0)
645 goto out;
646
647 skb->mac_header = skb->network_header;
648 skb_reset_network_header(skb);
649 IPCB(skb)->flags = 0;
650 skb->dev = tunnel->dev;
651
652 if (packet_is_spoofed(skb, iph, tunnel)) {
653 tunnel->dev->stats.rx_errors++;
654 goto out;
655 }
656
657 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
658 !net_eq(tunnel->net, dev_net(tunnel->dev))))
659 goto out;
660
661 err = IP_ECN_decapsulate(iph, skb);
662 if (unlikely(err)) {
663 if (log_ecn_error)
664 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
665 &iph->saddr, iph->tos);
666 if (err > 1) {
667 ++tunnel->dev->stats.rx_frame_errors;
668 ++tunnel->dev->stats.rx_errors;
669 goto out;
670 }
671 }
672
673 tstats = this_cpu_ptr(tunnel->dev->tstats);
674 u64_stats_update_begin(&tstats->syncp);
675 tstats->rx_packets++;
676 tstats->rx_bytes += skb->len;
677 u64_stats_update_end(&tstats->syncp);
678
679 netif_rx(skb);
680
681 return 0;
682 }
683
684 /* no tunnel matched, let upstream know, ipsec may handle it */
685 return 1;
686out:
687 kfree_skb(skb);
688 return 0;
689}
690
691static const struct tnl_ptk_info ipip_tpi = {
692 /* no tunnel info required for ipip. */
693 .proto = htons(ETH_P_IP),
694};
695
696#if IS_ENABLED(CONFIG_MPLS)
697static const struct tnl_ptk_info mplsip_tpi = {
698 /* no tunnel info required for mplsip. */
699 .proto = htons(ETH_P_MPLS_UC),
700};
701#endif
702
703static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
704{
705 const struct iphdr *iph;
706 struct ip_tunnel *tunnel;
707
708 iph = ip_hdr(skb);
709 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
710 iph->saddr, iph->daddr);
711 if (tunnel) {
712 const struct tnl_ptk_info *tpi;
713
714 if (tunnel->parms.iph.protocol != ipproto &&
715 tunnel->parms.iph.protocol != 0)
716 goto drop;
717
718 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
719 goto drop;
720#if IS_ENABLED(CONFIG_MPLS)
721 if (ipproto == IPPROTO_MPLS)
722 tpi = &mplsip_tpi;
723 else
724#endif
725 tpi = &ipip_tpi;
726 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
727 goto drop;
728 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
729 }
730
731 return 1;
732
733drop:
734 kfree_skb(skb);
735 return 0;
736}
737
738static int ipip_rcv(struct sk_buff *skb)
739{
740 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
741}
742
743#if IS_ENABLED(CONFIG_MPLS)
744static int mplsip_rcv(struct sk_buff *skb)
745{
746 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
747}
748#endif
749
750/*
751 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
752 * stores the embedded IPv4 address in v4dst and returns true.
753 */
754static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
755 __be32 *v4dst)
756{
757#ifdef CONFIG_IPV6_SIT_6RD
758 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
759 tunnel->ip6rd.prefixlen)) {
760 unsigned int pbw0, pbi0;
761 int pbi1;
762 u32 d;
763
764 pbw0 = tunnel->ip6rd.prefixlen >> 5;
765 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
766
767 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
768 tunnel->ip6rd.relay_prefixlen;
769
770 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
771 if (pbi1 > 0)
772 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
773 (32 - pbi1);
774
775 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
776 return true;
777 }
778#else
779 if (v6dst->s6_addr16[0] == htons(0x2002)) {
780 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
781 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
782 return true;
783 }
784#endif
785 return false;
786}
787
788static inline __be32 try_6rd(struct ip_tunnel *tunnel,
789 const struct in6_addr *v6dst)
790{
791 __be32 dst = 0;
792 check_6rd(tunnel, v6dst, &dst);
793 return dst;
794}
795
796/*
797 * This function assumes it is being called from dev_queue_xmit()
798 * and that skb is filled properly by that function.
799 */
800
801static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
802 struct net_device *dev)
803{
804 struct ip_tunnel *tunnel = netdev_priv(dev);
805 const struct iphdr *tiph = &tunnel->parms.iph;
806 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
807 u8 tos = tunnel->parms.iph.tos;
808 __be16 df = tiph->frag_off;
809 struct rtable *rt; /* Route to the other host */
810 struct net_device *tdev; /* Device to other host */
811 unsigned int max_headroom; /* The extra header space needed */
812 __be32 dst = tiph->daddr;
813 struct flowi4 fl4;
814 int mtu;
815 const struct in6_addr *addr6;
816 int addr_type;
817 u8 ttl;
818 u8 protocol = IPPROTO_IPV6;
819 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
820
821 if (tos == 1)
822 tos = ipv6_get_dsfield(iph6);
823
824 /* ISATAP (RFC4214) - must come before 6to4 */
825 if (dev->priv_flags & IFF_ISATAP) {
826 struct neighbour *neigh = NULL;
827 bool do_tx_error = false;
828
829 if (skb_dst(skb))
830 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
831
832 if (!neigh) {
833 net_dbg_ratelimited("nexthop == NULL\n");
834 goto tx_error;
835 }
836
837 addr6 = (const struct in6_addr *)&neigh->primary_key;
838 addr_type = ipv6_addr_type(addr6);
839
840 if ((addr_type & IPV6_ADDR_UNICAST) &&
841 ipv6_addr_is_isatap(addr6))
842 dst = addr6->s6_addr32[3];
843 else
844 do_tx_error = true;
845
846 neigh_release(neigh);
847 if (do_tx_error)
848 goto tx_error;
849 }
850
851 if (!dst)
852 dst = try_6rd(tunnel, &iph6->daddr);
853
854 if (!dst) {
855 struct neighbour *neigh = NULL;
856 bool do_tx_error = false;
857
858 if (skb_dst(skb))
859 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
860
861 if (!neigh) {
862 net_dbg_ratelimited("nexthop == NULL\n");
863 goto tx_error;
864 }
865
866 addr6 = (const struct in6_addr *)&neigh->primary_key;
867 addr_type = ipv6_addr_type(addr6);
868
869 if (addr_type == IPV6_ADDR_ANY) {
870 addr6 = &ipv6_hdr(skb)->daddr;
871 addr_type = ipv6_addr_type(addr6);
872 }
873
874 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
875 dst = addr6->s6_addr32[3];
876 else
877 do_tx_error = true;
878
879 neigh_release(neigh);
880 if (do_tx_error)
881 goto tx_error;
882 }
883
884 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
885 dst, tiph->saddr,
886 0, 0,
887 IPPROTO_IPV6, RT_TOS(tos),
888 tunnel->parms.link);
889 if (IS_ERR(rt)) {
890 dev->stats.tx_carrier_errors++;
891 goto tx_error_icmp;
892 }
893 if (rt->rt_type != RTN_UNICAST) {
894 ip_rt_put(rt);
895 dev->stats.tx_carrier_errors++;
896 goto tx_error_icmp;
897 }
898 tdev = rt->dst.dev;
899
900 if (tdev == dev) {
901 ip_rt_put(rt);
902 dev->stats.collisions++;
903 goto tx_error;
904 }
905
906 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
907 ip_rt_put(rt);
908 goto tx_error;
909 }
910
911 if (df) {
912 mtu = dst_mtu(&rt->dst) - t_hlen;
913
914 if (mtu < 68) {
915 dev->stats.collisions++;
916 ip_rt_put(rt);
917 goto tx_error;
918 }
919
920 if (mtu < IPV6_MIN_MTU) {
921 mtu = IPV6_MIN_MTU;
922 df = 0;
923 }
924
925 if (tunnel->parms.iph.daddr && skb_dst(skb))
926 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
927
928 if (skb->len > mtu && !skb_is_gso(skb)) {
929 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
930 ip_rt_put(rt);
931 goto tx_error;
932 }
933 }
934
935 if (tunnel->err_count > 0) {
936 if (time_before(jiffies,
937 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
938 tunnel->err_count--;
939 dst_link_failure(skb);
940 } else
941 tunnel->err_count = 0;
942 }
943
944 /*
945 * Okay, now see if we can stuff it in the buffer as-is.
946 */
947 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
948
949 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
950 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
951 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
952 if (!new_skb) {
953 ip_rt_put(rt);
954 dev->stats.tx_dropped++;
955 kfree_skb(skb);
956 return NETDEV_TX_OK;
957 }
958 if (skb->sk)
959 skb_set_owner_w(new_skb, skb->sk);
960 dev_kfree_skb(skb);
961 skb = new_skb;
962 iph6 = ipv6_hdr(skb);
963 }
964 ttl = tiph->ttl;
965 if (ttl == 0)
966 ttl = iph6->hop_limit;
967 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
968
969 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
970 ip_rt_put(rt);
971 goto tx_error;
972 }
973
974 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
975
976 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
977 df, !net_eq(tunnel->net, dev_net(dev)));
978 return NETDEV_TX_OK;
979
980tx_error_icmp:
981 dst_link_failure(skb);
982tx_error:
983 kfree_skb(skb);
984 dev->stats.tx_errors++;
985 return NETDEV_TX_OK;
986}
987
988static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
989 struct net_device *dev, u8 ipproto)
990{
991 struct ip_tunnel *tunnel = netdev_priv(dev);
992 const struct iphdr *tiph = &tunnel->parms.iph;
993
994 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
995 goto tx_error;
996
997 skb_set_inner_ipproto(skb, ipproto);
998
999 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1000 return NETDEV_TX_OK;
1001tx_error:
1002 kfree_skb(skb);
1003 dev->stats.tx_errors++;
1004 return NETDEV_TX_OK;
1005}
1006
1007static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1008 struct net_device *dev)
1009{
1010 switch (skb->protocol) {
1011 case htons(ETH_P_IP):
1012 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1013 break;
1014 case htons(ETH_P_IPV6):
1015 ipip6_tunnel_xmit(skb, dev);
1016 break;
1017#if IS_ENABLED(CONFIG_MPLS)
1018 case htons(ETH_P_MPLS_UC):
1019 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1020 break;
1021#endif
1022 default:
1023 goto tx_err;
1024 }
1025
1026 return NETDEV_TX_OK;
1027
1028tx_err:
1029 dev->stats.tx_errors++;
1030 kfree_skb(skb);
1031 return NETDEV_TX_OK;
1032
1033}
1034
1035static void ipip6_tunnel_bind_dev(struct net_device *dev)
1036{
1037 struct net_device *tdev = NULL;
1038 struct ip_tunnel *tunnel;
1039 const struct iphdr *iph;
1040 struct flowi4 fl4;
1041
1042 tunnel = netdev_priv(dev);
1043 iph = &tunnel->parms.iph;
1044
1045 if (iph->daddr) {
1046 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1047 NULL,
1048 iph->daddr, iph->saddr,
1049 0, 0,
1050 IPPROTO_IPV6,
1051 RT_TOS(iph->tos),
1052 tunnel->parms.link);
1053
1054 if (!IS_ERR(rt)) {
1055 tdev = rt->dst.dev;
1056 ip_rt_put(rt);
1057 }
1058 dev->flags |= IFF_POINTOPOINT;
1059 }
1060
1061 if (!tdev && tunnel->parms.link)
1062 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1063
1064 if (tdev) {
1065 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1066
1067 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1068 dev->mtu = tdev->mtu - t_hlen;
1069 if (dev->mtu < IPV6_MIN_MTU)
1070 dev->mtu = IPV6_MIN_MTU;
1071 }
1072}
1073
1074static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1075{
1076 struct net *net = t->net;
1077 struct sit_net *sitn = net_generic(net, sit_net_id);
1078
1079 ipip6_tunnel_unlink(sitn, t);
1080 synchronize_net();
1081 t->parms.iph.saddr = p->iph.saddr;
1082 t->parms.iph.daddr = p->iph.daddr;
1083 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1084 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1085 ipip6_tunnel_link(sitn, t);
1086 t->parms.iph.ttl = p->iph.ttl;
1087 t->parms.iph.tos = p->iph.tos;
1088 if (t->parms.link != p->link) {
1089 t->parms.link = p->link;
1090 ipip6_tunnel_bind_dev(t->dev);
1091 }
1092 dst_cache_reset(&t->dst_cache);
1093 netdev_state_change(t->dev);
1094}
1095
1096#ifdef CONFIG_IPV6_SIT_6RD
1097static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1098 struct ip_tunnel_6rd *ip6rd)
1099{
1100 struct in6_addr prefix;
1101 __be32 relay_prefix;
1102
1103 if (ip6rd->relay_prefixlen > 32 ||
1104 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1105 return -EINVAL;
1106
1107 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1108 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1109 return -EINVAL;
1110 if (ip6rd->relay_prefixlen)
1111 relay_prefix = ip6rd->relay_prefix &
1112 htonl(0xffffffffUL <<
1113 (32 - ip6rd->relay_prefixlen));
1114 else
1115 relay_prefix = 0;
1116 if (relay_prefix != ip6rd->relay_prefix)
1117 return -EINVAL;
1118
1119 t->ip6rd.prefix = prefix;
1120 t->ip6rd.relay_prefix = relay_prefix;
1121 t->ip6rd.prefixlen = ip6rd->prefixlen;
1122 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1123 dst_cache_reset(&t->dst_cache);
1124 netdev_state_change(t->dev);
1125 return 0;
1126}
1127#endif
1128
1129static bool ipip6_valid_ip_proto(u8 ipproto)
1130{
1131 return ipproto == IPPROTO_IPV6 ||
1132 ipproto == IPPROTO_IPIP ||
1133#if IS_ENABLED(CONFIG_MPLS)
1134 ipproto == IPPROTO_MPLS ||
1135#endif
1136 ipproto == 0;
1137}
1138
1139static int
1140ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1141{
1142 int err = 0;
1143 struct ip_tunnel_parm p;
1144 struct ip_tunnel_prl prl;
1145 struct ip_tunnel *t = netdev_priv(dev);
1146 struct net *net = t->net;
1147 struct sit_net *sitn = net_generic(net, sit_net_id);
1148#ifdef CONFIG_IPV6_SIT_6RD
1149 struct ip_tunnel_6rd ip6rd;
1150#endif
1151
1152 switch (cmd) {
1153 case SIOCGETTUNNEL:
1154#ifdef CONFIG_IPV6_SIT_6RD
1155 case SIOCGET6RD:
1156#endif
1157 if (dev == sitn->fb_tunnel_dev) {
1158 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1159 err = -EFAULT;
1160 break;
1161 }
1162 t = ipip6_tunnel_locate(net, &p, 0);
1163 if (!t)
1164 t = netdev_priv(dev);
1165 }
1166
1167 err = -EFAULT;
1168 if (cmd == SIOCGETTUNNEL) {
1169 memcpy(&p, &t->parms, sizeof(p));
1170 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1171 sizeof(p)))
1172 goto done;
1173#ifdef CONFIG_IPV6_SIT_6RD
1174 } else {
1175 ip6rd.prefix = t->ip6rd.prefix;
1176 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1177 ip6rd.prefixlen = t->ip6rd.prefixlen;
1178 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1179 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1180 sizeof(ip6rd)))
1181 goto done;
1182#endif
1183 }
1184 err = 0;
1185 break;
1186
1187 case SIOCADDTUNNEL:
1188 case SIOCCHGTUNNEL:
1189 err = -EPERM;
1190 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1191 goto done;
1192
1193 err = -EFAULT;
1194 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1195 goto done;
1196
1197 err = -EINVAL;
1198 if (!ipip6_valid_ip_proto(p.iph.protocol))
1199 goto done;
1200 if (p.iph.version != 4 ||
1201 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1202 goto done;
1203 if (p.iph.ttl)
1204 p.iph.frag_off |= htons(IP_DF);
1205
1206 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1207
1208 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1209 if (t) {
1210 if (t->dev != dev) {
1211 err = -EEXIST;
1212 break;
1213 }
1214 } else {
1215 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1216 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1217 err = -EINVAL;
1218 break;
1219 }
1220 t = netdev_priv(dev);
1221 }
1222
1223 ipip6_tunnel_update(t, &p);
1224 }
1225
1226 if (t) {
1227 err = 0;
1228 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1229 err = -EFAULT;
1230 } else
1231 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1232 break;
1233
1234 case SIOCDELTUNNEL:
1235 err = -EPERM;
1236 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1237 goto done;
1238
1239 if (dev == sitn->fb_tunnel_dev) {
1240 err = -EFAULT;
1241 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1242 goto done;
1243 err = -ENOENT;
1244 t = ipip6_tunnel_locate(net, &p, 0);
1245 if (!t)
1246 goto done;
1247 err = -EPERM;
1248 if (t == netdev_priv(sitn->fb_tunnel_dev))
1249 goto done;
1250 dev = t->dev;
1251 }
1252 unregister_netdevice(dev);
1253 err = 0;
1254 break;
1255
1256 case SIOCGETPRL:
1257 err = -EINVAL;
1258 if (dev == sitn->fb_tunnel_dev)
1259 goto done;
1260 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1261 break;
1262
1263 case SIOCADDPRL:
1264 case SIOCDELPRL:
1265 case SIOCCHGPRL:
1266 err = -EPERM;
1267 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1268 goto done;
1269 err = -EINVAL;
1270 if (dev == sitn->fb_tunnel_dev)
1271 goto done;
1272 err = -EFAULT;
1273 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1274 goto done;
1275
1276 switch (cmd) {
1277 case SIOCDELPRL:
1278 err = ipip6_tunnel_del_prl(t, &prl);
1279 break;
1280 case SIOCADDPRL:
1281 case SIOCCHGPRL:
1282 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1283 break;
1284 }
1285 dst_cache_reset(&t->dst_cache);
1286 netdev_state_change(dev);
1287 break;
1288
1289#ifdef CONFIG_IPV6_SIT_6RD
1290 case SIOCADD6RD:
1291 case SIOCCHG6RD:
1292 case SIOCDEL6RD:
1293 err = -EPERM;
1294 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1295 goto done;
1296
1297 err = -EFAULT;
1298 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1299 sizeof(ip6rd)))
1300 goto done;
1301
1302 if (cmd != SIOCDEL6RD) {
1303 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1304 if (err < 0)
1305 goto done;
1306 } else
1307 ipip6_tunnel_clone_6rd(dev, sitn);
1308
1309 err = 0;
1310 break;
1311#endif
1312
1313 default:
1314 err = -EINVAL;
1315 }
1316
1317done:
1318 return err;
1319}
1320
1321static const struct net_device_ops ipip6_netdev_ops = {
1322 .ndo_init = ipip6_tunnel_init,
1323 .ndo_uninit = ipip6_tunnel_uninit,
1324 .ndo_start_xmit = sit_tunnel_xmit,
1325 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1326 .ndo_get_stats64 = ip_tunnel_get_stats64,
1327 .ndo_get_iflink = ip_tunnel_get_iflink,
1328};
1329
1330static void ipip6_dev_free(struct net_device *dev)
1331{
1332 struct ip_tunnel *tunnel = netdev_priv(dev);
1333
1334 dst_cache_destroy(&tunnel->dst_cache);
1335 free_percpu(dev->tstats);
1336 free_netdev(dev);
1337}
1338
1339#define SIT_FEATURES (NETIF_F_SG | \
1340 NETIF_F_FRAGLIST | \
1341 NETIF_F_HIGHDMA | \
1342 NETIF_F_GSO_SOFTWARE | \
1343 NETIF_F_HW_CSUM)
1344
1345static void ipip6_tunnel_setup(struct net_device *dev)
1346{
1347 struct ip_tunnel *tunnel = netdev_priv(dev);
1348 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1349
1350 dev->netdev_ops = &ipip6_netdev_ops;
1351 dev->destructor = ipip6_dev_free;
1352
1353 dev->type = ARPHRD_SIT;
1354 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1355 dev->mtu = ETH_DATA_LEN - t_hlen;
1356 dev->min_mtu = IPV6_MIN_MTU;
1357 dev->max_mtu = 0xFFF8 - t_hlen;
1358 dev->flags = IFF_NOARP;
1359 netif_keep_dst(dev);
1360 dev->addr_len = 4;
1361 dev->features |= NETIF_F_LLTX;
1362 dev->features |= SIT_FEATURES;
1363 dev->hw_features |= SIT_FEATURES;
1364}
1365
1366static int ipip6_tunnel_init(struct net_device *dev)
1367{
1368 struct ip_tunnel *tunnel = netdev_priv(dev);
1369 int err;
1370
1371 tunnel->dev = dev;
1372 tunnel->net = dev_net(dev);
1373 strcpy(tunnel->parms.name, dev->name);
1374
1375 ipip6_tunnel_bind_dev(dev);
1376 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1377 if (!dev->tstats)
1378 return -ENOMEM;
1379
1380 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1381 if (err) {
1382 free_percpu(dev->tstats);
1383 dev->tstats = NULL;
1384 return err;
1385 }
1386
1387 return 0;
1388}
1389
1390static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1391{
1392 struct ip_tunnel *tunnel = netdev_priv(dev);
1393 struct iphdr *iph = &tunnel->parms.iph;
1394 struct net *net = dev_net(dev);
1395 struct sit_net *sitn = net_generic(net, sit_net_id);
1396
1397 iph->version = 4;
1398 iph->protocol = IPPROTO_IPV6;
1399 iph->ihl = 5;
1400 iph->ttl = 64;
1401
1402 dev_hold(dev);
1403 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1404}
1405
1406static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1407{
1408 u8 proto;
1409
1410 if (!data || !data[IFLA_IPTUN_PROTO])
1411 return 0;
1412
1413 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1414 if (!ipip6_valid_ip_proto(proto))
1415 return -EINVAL;
1416
1417 return 0;
1418}
1419
1420static void ipip6_netlink_parms(struct nlattr *data[],
1421 struct ip_tunnel_parm *parms)
1422{
1423 memset(parms, 0, sizeof(*parms));
1424
1425 parms->iph.version = 4;
1426 parms->iph.protocol = IPPROTO_IPV6;
1427 parms->iph.ihl = 5;
1428 parms->iph.ttl = 64;
1429
1430 if (!data)
1431 return;
1432
1433 if (data[IFLA_IPTUN_LINK])
1434 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1435
1436 if (data[IFLA_IPTUN_LOCAL])
1437 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1438
1439 if (data[IFLA_IPTUN_REMOTE])
1440 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1441
1442 if (data[IFLA_IPTUN_TTL]) {
1443 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1444 if (parms->iph.ttl)
1445 parms->iph.frag_off = htons(IP_DF);
1446 }
1447
1448 if (data[IFLA_IPTUN_TOS])
1449 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1450
1451 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1452 parms->iph.frag_off = htons(IP_DF);
1453
1454 if (data[IFLA_IPTUN_FLAGS])
1455 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1456
1457 if (data[IFLA_IPTUN_PROTO])
1458 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1459
1460}
1461
1462/* This function returns true when ENCAP attributes are present in the nl msg */
1463static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1464 struct ip_tunnel_encap *ipencap)
1465{
1466 bool ret = false;
1467
1468 memset(ipencap, 0, sizeof(*ipencap));
1469
1470 if (!data)
1471 return ret;
1472
1473 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1474 ret = true;
1475 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1476 }
1477
1478 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1479 ret = true;
1480 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1481 }
1482
1483 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1484 ret = true;
1485 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1486 }
1487
1488 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1489 ret = true;
1490 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1491 }
1492
1493 return ret;
1494}
1495
1496#ifdef CONFIG_IPV6_SIT_6RD
1497/* This function returns true when 6RD attributes are present in the nl msg */
1498static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1499 struct ip_tunnel_6rd *ip6rd)
1500{
1501 bool ret = false;
1502 memset(ip6rd, 0, sizeof(*ip6rd));
1503
1504 if (!data)
1505 return ret;
1506
1507 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1508 ret = true;
1509 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1510 }
1511
1512 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1513 ret = true;
1514 ip6rd->relay_prefix =
1515 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1516 }
1517
1518 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1519 ret = true;
1520 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1521 }
1522
1523 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1524 ret = true;
1525 ip6rd->relay_prefixlen =
1526 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1527 }
1528
1529 return ret;
1530}
1531#endif
1532
1533static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1534 struct nlattr *tb[], struct nlattr *data[])
1535{
1536 struct net *net = dev_net(dev);
1537 struct ip_tunnel *nt;
1538 struct ip_tunnel_encap ipencap;
1539#ifdef CONFIG_IPV6_SIT_6RD
1540 struct ip_tunnel_6rd ip6rd;
1541#endif
1542 int err;
1543
1544 nt = netdev_priv(dev);
1545
1546 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1547 err = ip_tunnel_encap_setup(nt, &ipencap);
1548 if (err < 0)
1549 return err;
1550 }
1551
1552 ipip6_netlink_parms(data, &nt->parms);
1553
1554 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1555 return -EEXIST;
1556
1557 err = ipip6_tunnel_create(dev);
1558 if (err < 0)
1559 return err;
1560
1561#ifdef CONFIG_IPV6_SIT_6RD
1562 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1563 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1564#endif
1565
1566 return err;
1567}
1568
1569static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1570 struct nlattr *data[])
1571{
1572 struct ip_tunnel *t = netdev_priv(dev);
1573 struct ip_tunnel_parm p;
1574 struct ip_tunnel_encap ipencap;
1575 struct net *net = t->net;
1576 struct sit_net *sitn = net_generic(net, sit_net_id);
1577#ifdef CONFIG_IPV6_SIT_6RD
1578 struct ip_tunnel_6rd ip6rd;
1579#endif
1580 int err;
1581
1582 if (dev == sitn->fb_tunnel_dev)
1583 return -EINVAL;
1584
1585 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1586 err = ip_tunnel_encap_setup(t, &ipencap);
1587 if (err < 0)
1588 return err;
1589 }
1590
1591 ipip6_netlink_parms(data, &p);
1592
1593 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1594 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1595 return -EINVAL;
1596
1597 t = ipip6_tunnel_locate(net, &p, 0);
1598
1599 if (t) {
1600 if (t->dev != dev)
1601 return -EEXIST;
1602 } else
1603 t = netdev_priv(dev);
1604
1605 ipip6_tunnel_update(t, &p);
1606
1607#ifdef CONFIG_IPV6_SIT_6RD
1608 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1609 return ipip6_tunnel_update_6rd(t, &ip6rd);
1610#endif
1611
1612 return 0;
1613}
1614
1615static size_t ipip6_get_size(const struct net_device *dev)
1616{
1617 return
1618 /* IFLA_IPTUN_LINK */
1619 nla_total_size(4) +
1620 /* IFLA_IPTUN_LOCAL */
1621 nla_total_size(4) +
1622 /* IFLA_IPTUN_REMOTE */
1623 nla_total_size(4) +
1624 /* IFLA_IPTUN_TTL */
1625 nla_total_size(1) +
1626 /* IFLA_IPTUN_TOS */
1627 nla_total_size(1) +
1628 /* IFLA_IPTUN_PMTUDISC */
1629 nla_total_size(1) +
1630 /* IFLA_IPTUN_FLAGS */
1631 nla_total_size(2) +
1632 /* IFLA_IPTUN_PROTO */
1633 nla_total_size(1) +
1634#ifdef CONFIG_IPV6_SIT_6RD
1635 /* IFLA_IPTUN_6RD_PREFIX */
1636 nla_total_size(sizeof(struct in6_addr)) +
1637 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1638 nla_total_size(4) +
1639 /* IFLA_IPTUN_6RD_PREFIXLEN */
1640 nla_total_size(2) +
1641 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1642 nla_total_size(2) +
1643#endif
1644 /* IFLA_IPTUN_ENCAP_TYPE */
1645 nla_total_size(2) +
1646 /* IFLA_IPTUN_ENCAP_FLAGS */
1647 nla_total_size(2) +
1648 /* IFLA_IPTUN_ENCAP_SPORT */
1649 nla_total_size(2) +
1650 /* IFLA_IPTUN_ENCAP_DPORT */
1651 nla_total_size(2) +
1652 0;
1653}
1654
1655static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1656{
1657 struct ip_tunnel *tunnel = netdev_priv(dev);
1658 struct ip_tunnel_parm *parm = &tunnel->parms;
1659
1660 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1661 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1662 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1663 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1664 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1665 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1666 !!(parm->iph.frag_off & htons(IP_DF))) ||
1667 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1668 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1669 goto nla_put_failure;
1670
1671#ifdef CONFIG_IPV6_SIT_6RD
1672 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1673 &tunnel->ip6rd.prefix) ||
1674 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1675 tunnel->ip6rd.relay_prefix) ||
1676 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1677 tunnel->ip6rd.prefixlen) ||
1678 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1679 tunnel->ip6rd.relay_prefixlen))
1680 goto nla_put_failure;
1681#endif
1682
1683 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1684 tunnel->encap.type) ||
1685 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1686 tunnel->encap.sport) ||
1687 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1688 tunnel->encap.dport) ||
1689 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1690 tunnel->encap.flags))
1691 goto nla_put_failure;
1692
1693 return 0;
1694
1695nla_put_failure:
1696 return -EMSGSIZE;
1697}
1698
1699static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1700 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1701 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1702 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1703 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1704 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1705 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1706 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1707 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1708#ifdef CONFIG_IPV6_SIT_6RD
1709 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1710 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1711 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1712 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1713#endif
1714 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1715 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1716 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1717 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1718};
1719
1720static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1721{
1722 struct net *net = dev_net(dev);
1723 struct sit_net *sitn = net_generic(net, sit_net_id);
1724
1725 if (dev != sitn->fb_tunnel_dev)
1726 unregister_netdevice_queue(dev, head);
1727}
1728
1729static struct rtnl_link_ops sit_link_ops __read_mostly = {
1730 .kind = "sit",
1731 .maxtype = IFLA_IPTUN_MAX,
1732 .policy = ipip6_policy,
1733 .priv_size = sizeof(struct ip_tunnel),
1734 .setup = ipip6_tunnel_setup,
1735 .validate = ipip6_validate,
1736 .newlink = ipip6_newlink,
1737 .changelink = ipip6_changelink,
1738 .get_size = ipip6_get_size,
1739 .fill_info = ipip6_fill_info,
1740 .dellink = ipip6_dellink,
1741 .get_link_net = ip_tunnel_get_link_net,
1742};
1743
1744static struct xfrm_tunnel sit_handler __read_mostly = {
1745 .handler = ipip6_rcv,
1746 .err_handler = ipip6_err,
1747 .priority = 1,
1748};
1749
1750static struct xfrm_tunnel ipip_handler __read_mostly = {
1751 .handler = ipip_rcv,
1752 .err_handler = ipip6_err,
1753 .priority = 2,
1754};
1755
1756#if IS_ENABLED(CONFIG_MPLS)
1757static struct xfrm_tunnel mplsip_handler __read_mostly = {
1758 .handler = mplsip_rcv,
1759 .err_handler = ipip6_err,
1760 .priority = 2,
1761};
1762#endif
1763
1764static void __net_exit sit_destroy_tunnels(struct net *net,
1765 struct list_head *head)
1766{
1767 struct sit_net *sitn = net_generic(net, sit_net_id);
1768 struct net_device *dev, *aux;
1769 int prio;
1770
1771 for_each_netdev_safe(net, dev, aux)
1772 if (dev->rtnl_link_ops == &sit_link_ops)
1773 unregister_netdevice_queue(dev, head);
1774
1775 for (prio = 1; prio < 4; prio++) {
1776 int h;
1777 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1778 struct ip_tunnel *t;
1779
1780 t = rtnl_dereference(sitn->tunnels[prio][h]);
1781 while (t) {
1782 /* If dev is in the same netns, it has already
1783 * been added to the list by the previous loop.
1784 */
1785 if (!net_eq(dev_net(t->dev), net))
1786 unregister_netdevice_queue(t->dev,
1787 head);
1788 t = rtnl_dereference(t->next);
1789 }
1790 }
1791 }
1792}
1793
1794static int __net_init sit_init_net(struct net *net)
1795{
1796 struct sit_net *sitn = net_generic(net, sit_net_id);
1797 struct ip_tunnel *t;
1798 int err;
1799
1800 sitn->tunnels[0] = sitn->tunnels_wc;
1801 sitn->tunnels[1] = sitn->tunnels_l;
1802 sitn->tunnels[2] = sitn->tunnels_r;
1803 sitn->tunnels[3] = sitn->tunnels_r_l;
1804
1805 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1806 NET_NAME_UNKNOWN,
1807 ipip6_tunnel_setup);
1808 if (!sitn->fb_tunnel_dev) {
1809 err = -ENOMEM;
1810 goto err_alloc_dev;
1811 }
1812 dev_net_set(sitn->fb_tunnel_dev, net);
1813 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1814 /* FB netdevice is special: we have one, and only one per netns.
1815 * Allowing to move it to another netns is clearly unsafe.
1816 */
1817 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1818
1819 err = register_netdev(sitn->fb_tunnel_dev);
1820 if (err)
1821 goto err_reg_dev;
1822
1823 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1824 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1825
1826 t = netdev_priv(sitn->fb_tunnel_dev);
1827
1828 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1829 return 0;
1830
1831err_reg_dev:
1832 ipip6_dev_free(sitn->fb_tunnel_dev);
1833err_alloc_dev:
1834 return err;
1835}
1836
1837static void __net_exit sit_exit_net(struct net *net)
1838{
1839 LIST_HEAD(list);
1840
1841 rtnl_lock();
1842 sit_destroy_tunnels(net, &list);
1843 unregister_netdevice_many(&list);
1844 rtnl_unlock();
1845}
1846
1847static struct pernet_operations sit_net_ops = {
1848 .init = sit_init_net,
1849 .exit = sit_exit_net,
1850 .id = &sit_net_id,
1851 .size = sizeof(struct sit_net),
1852};
1853
1854static void __exit sit_cleanup(void)
1855{
1856 rtnl_link_unregister(&sit_link_ops);
1857 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1858 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1859#if IS_ENABLED(CONFIG_MPLS)
1860 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1861#endif
1862
1863 unregister_pernet_device(&sit_net_ops);
1864 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1865}
1866
1867static int __init sit_init(void)
1868{
1869 int err;
1870
1871 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1872
1873 err = register_pernet_device(&sit_net_ops);
1874 if (err < 0)
1875 return err;
1876 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1877 if (err < 0) {
1878 pr_info("%s: can't register ip6ip4\n", __func__);
1879 goto xfrm_tunnel_failed;
1880 }
1881 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1882 if (err < 0) {
1883 pr_info("%s: can't register ip4ip4\n", __func__);
1884 goto xfrm_tunnel4_failed;
1885 }
1886#if IS_ENABLED(CONFIG_MPLS)
1887 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1888 if (err < 0) {
1889 pr_info("%s: can't register mplsip\n", __func__);
1890 goto xfrm_tunnel_mpls_failed;
1891 }
1892#endif
1893 err = rtnl_link_register(&sit_link_ops);
1894 if (err < 0)
1895 goto rtnl_link_failed;
1896
1897out:
1898 return err;
1899
1900rtnl_link_failed:
1901#if IS_ENABLED(CONFIG_MPLS)
1902 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1903xfrm_tunnel_mpls_failed:
1904#endif
1905 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1906xfrm_tunnel4_failed:
1907 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1908xfrm_tunnel_failed:
1909 unregister_pernet_device(&sit_net_ops);
1910 goto out;
1911}
1912
1913module_init(sit_init);
1914module_exit(sit_cleanup);
1915MODULE_LICENSE("GPL");
1916MODULE_ALIAS_RTNL_LINK("sit");
1917MODULE_ALIAS_NETDEV("sit0");