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,
95 __be32 remote, __be32 local,
96 int sifindex)
97{
98 unsigned int h0 = HASH(remote);
99 unsigned int h1 = HASH(local);
100 struct ip_tunnel *t;
101 struct sit_net *sitn = net_generic(net, sit_net_id);
102 int ifindex = dev ? dev->ifindex : 0;
103
104 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
105 if (local == t->parms.iph.saddr &&
106 remote == t->parms.iph.daddr &&
107 (!dev || !t->parms.link || ifindex == t->parms.link ||
108 sifindex == t->parms.link) &&
109 (t->dev->flags & IFF_UP))
110 return t;
111 }
112 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
113 if (remote == t->parms.iph.daddr &&
114 (!dev || !t->parms.link || ifindex == t->parms.link ||
115 sifindex == t->parms.link) &&
116 (t->dev->flags & IFF_UP))
117 return t;
118 }
119 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
120 if (local == t->parms.iph.saddr &&
121 (!dev || !t->parms.link || ifindex == t->parms.link ||
122 sifindex == t->parms.link) &&
123 (t->dev->flags & IFF_UP))
124 return t;
125 }
126 t = rcu_dereference(sitn->tunnels_wc[0]);
127 if (t && (t->dev->flags & IFF_UP))
128 return t;
129 return NULL;
130}
131
132static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
133 struct ip_tunnel_parm *parms)
134{
135 __be32 remote = parms->iph.daddr;
136 __be32 local = parms->iph.saddr;
137 unsigned int h = 0;
138 int prio = 0;
139
140 if (remote) {
141 prio |= 2;
142 h ^= HASH(remote);
143 }
144 if (local) {
145 prio |= 1;
146 h ^= HASH(local);
147 }
148 return &sitn->tunnels[prio][h];
149}
150
151static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
152 struct ip_tunnel *t)
153{
154 return __ipip6_bucket(sitn, &t->parms);
155}
156
157static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
158{
159 struct ip_tunnel __rcu **tp;
160 struct ip_tunnel *iter;
161
162 for (tp = ipip6_bucket(sitn, t);
163 (iter = rtnl_dereference(*tp)) != NULL;
164 tp = &iter->next) {
165 if (t == iter) {
166 rcu_assign_pointer(*tp, t->next);
167 break;
168 }
169 }
170}
171
172static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
173{
174 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
175
176 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
177 rcu_assign_pointer(*tp, t);
178}
179
180static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
181{
182#ifdef CONFIG_IPV6_SIT_6RD
183 struct ip_tunnel *t = netdev_priv(dev);
184
185 if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
186 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
187 t->ip6rd.relay_prefix = 0;
188 t->ip6rd.prefixlen = 16;
189 t->ip6rd.relay_prefixlen = 0;
190 } else {
191 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
192 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
193 }
194#endif
195}
196
197static int ipip6_tunnel_create(struct net_device *dev)
198{
199 struct ip_tunnel *t = netdev_priv(dev);
200 struct net *net = dev_net(dev);
201 struct sit_net *sitn = net_generic(net, sit_net_id);
202 int err;
203
204 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
205 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
206
207 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
208 dev->priv_flags |= IFF_ISATAP;
209
210 dev->rtnl_link_ops = &sit_link_ops;
211
212 err = register_netdevice(dev);
213 if (err < 0)
214 goto out;
215
216 ipip6_tunnel_clone_6rd(dev, sitn);
217
218 dev_hold(dev);
219
220 ipip6_tunnel_link(sitn, t);
221 return 0;
222
223out:
224 return err;
225}
226
227static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
228 struct ip_tunnel_parm *parms, int create)
229{
230 __be32 remote = parms->iph.daddr;
231 __be32 local = parms->iph.saddr;
232 struct ip_tunnel *t, *nt;
233 struct ip_tunnel __rcu **tp;
234 struct net_device *dev;
235 char name[IFNAMSIZ];
236 struct sit_net *sitn = net_generic(net, sit_net_id);
237
238 for (tp = __ipip6_bucket(sitn, parms);
239 (t = rtnl_dereference(*tp)) != NULL;
240 tp = &t->next) {
241 if (local == t->parms.iph.saddr &&
242 remote == t->parms.iph.daddr &&
243 parms->link == t->parms.link) {
244 if (create)
245 return NULL;
246 else
247 return t;
248 }
249 }
250 if (!create)
251 goto failed;
252
253 if (parms->name[0]) {
254 if (!dev_valid_name(parms->name))
255 goto failed;
256 strlcpy(name, parms->name, IFNAMSIZ);
257 } else {
258 strcpy(name, "sit%d");
259 }
260 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
261 ipip6_tunnel_setup);
262 if (!dev)
263 return NULL;
264
265 dev_net_set(dev, net);
266
267 nt = netdev_priv(dev);
268
269 nt->parms = *parms;
270 if (ipip6_tunnel_create(dev) < 0)
271 goto failed_free;
272
273 return nt;
274
275failed_free:
276 free_netdev(dev);
277failed:
278 return NULL;
279}
280
281#define for_each_prl_rcu(start) \
282 for (prl = rcu_dereference(start); \
283 prl; \
284 prl = rcu_dereference(prl->next))
285
286static struct ip_tunnel_prl_entry *
287__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
288{
289 struct ip_tunnel_prl_entry *prl;
290
291 for_each_prl_rcu(t->prl)
292 if (prl->addr == addr)
293 break;
294 return prl;
295
296}
297
298static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
299 struct ip_tunnel_prl __user *a)
300{
301 struct ip_tunnel_prl kprl, *kp;
302 struct ip_tunnel_prl_entry *prl;
303 unsigned int cmax, c = 0, ca, len;
304 int ret = 0;
305
306 if (copy_from_user(&kprl, a, sizeof(kprl)))
307 return -EFAULT;
308 cmax = kprl.datalen / sizeof(kprl);
309 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
310 cmax = 1;
311
312 /* For simple GET or for root users,
313 * we try harder to allocate.
314 */
315 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
316 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
317 NULL;
318
319 rcu_read_lock();
320
321 ca = t->prl_count < cmax ? t->prl_count : cmax;
322
323 if (!kp) {
324 /* We don't try hard to allocate much memory for
325 * non-root users.
326 * For root users, retry allocating enough memory for
327 * the answer.
328 */
329 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
330 if (!kp) {
331 ret = -ENOMEM;
332 goto out;
333 }
334 }
335
336 c = 0;
337 for_each_prl_rcu(t->prl) {
338 if (c >= cmax)
339 break;
340 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
341 continue;
342 kp[c].addr = prl->addr;
343 kp[c].flags = prl->flags;
344 c++;
345 if (kprl.addr != htonl(INADDR_ANY))
346 break;
347 }
348out:
349 rcu_read_unlock();
350
351 len = sizeof(*kp) * c;
352 ret = 0;
353 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
354 ret = -EFAULT;
355
356 kfree(kp);
357
358 return ret;
359}
360
361static int
362ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
363{
364 struct ip_tunnel_prl_entry *p;
365 int err = 0;
366
367 if (a->addr == htonl(INADDR_ANY))
368 return -EINVAL;
369
370 ASSERT_RTNL();
371
372 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
373 if (p->addr == a->addr) {
374 if (chg) {
375 p->flags = a->flags;
376 goto out;
377 }
378 err = -EEXIST;
379 goto out;
380 }
381 }
382
383 if (chg) {
384 err = -ENXIO;
385 goto out;
386 }
387
388 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
389 if (!p) {
390 err = -ENOBUFS;
391 goto out;
392 }
393
394 p->next = t->prl;
395 p->addr = a->addr;
396 p->flags = a->flags;
397 t->prl_count++;
398 rcu_assign_pointer(t->prl, p);
399out:
400 return err;
401}
402
403static void prl_list_destroy_rcu(struct rcu_head *head)
404{
405 struct ip_tunnel_prl_entry *p, *n;
406
407 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
408 do {
409 n = rcu_dereference_protected(p->next, 1);
410 kfree(p);
411 p = n;
412 } while (p);
413}
414
415static int
416ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
417{
418 struct ip_tunnel_prl_entry *x;
419 struct ip_tunnel_prl_entry __rcu **p;
420 int err = 0;
421
422 ASSERT_RTNL();
423
424 if (a && a->addr != htonl(INADDR_ANY)) {
425 for (p = &t->prl;
426 (x = rtnl_dereference(*p)) != NULL;
427 p = &x->next) {
428 if (x->addr == a->addr) {
429 *p = x->next;
430 kfree_rcu(x, rcu_head);
431 t->prl_count--;
432 goto out;
433 }
434 }
435 err = -ENXIO;
436 } else {
437 x = rtnl_dereference(t->prl);
438 if (x) {
439 t->prl_count = 0;
440 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
441 t->prl = NULL;
442 }
443 }
444out:
445 return err;
446}
447
448static int
449isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
450{
451 struct ip_tunnel_prl_entry *p;
452 int ok = 1;
453
454 rcu_read_lock();
455 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
456 if (p) {
457 if (p->flags & PRL_DEFAULT)
458 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
459 else
460 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
461 } else {
462 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
463
464 if (ipv6_addr_is_isatap(addr6) &&
465 (addr6->s6_addr32[3] == iph->saddr) &&
466 ipv6_chk_prefix(addr6, t->dev))
467 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
468 else
469 ok = 0;
470 }
471 rcu_read_unlock();
472 return ok;
473}
474
475static void ipip6_tunnel_uninit(struct net_device *dev)
476{
477 struct ip_tunnel *tunnel = netdev_priv(dev);
478 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
479
480 if (dev == sitn->fb_tunnel_dev) {
481 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
482 } else {
483 ipip6_tunnel_unlink(sitn, tunnel);
484 ipip6_tunnel_del_prl(tunnel, NULL);
485 }
486 dst_cache_reset(&tunnel->dst_cache);
487 dev_put(dev);
488}
489
490static int ipip6_err(struct sk_buff *skb, u32 info)
491{
492 const struct iphdr *iph = (const struct iphdr *)skb->data;
493 const int type = icmp_hdr(skb)->type;
494 const int code = icmp_hdr(skb)->code;
495 unsigned int data_len = 0;
496 struct ip_tunnel *t;
497 int sifindex;
498 int err;
499
500 switch (type) {
501 default:
502 case ICMP_PARAMETERPROB:
503 return 0;
504
505 case ICMP_DEST_UNREACH:
506 switch (code) {
507 case ICMP_SR_FAILED:
508 /* Impossible event. */
509 return 0;
510 default:
511 /* All others are translated to HOST_UNREACH.
512 rfc2003 contains "deep thoughts" about NET_UNREACH,
513 I believe they are just ether pollution. --ANK
514 */
515 break;
516 }
517 break;
518 case ICMP_TIME_EXCEEDED:
519 if (code != ICMP_EXC_TTL)
520 return 0;
521 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
522 break;
523 case ICMP_REDIRECT:
524 break;
525 }
526
527 err = -ENOENT;
528
529 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
530 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
531 iph->daddr, iph->saddr, sifindex);
532 if (!t)
533 goto out;
534
535 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
536 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
537 t->parms.link, 0, iph->protocol, 0);
538 err = 0;
539 goto out;
540 }
541 if (type == ICMP_REDIRECT) {
542 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
543 iph->protocol, 0);
544 err = 0;
545 goto out;
546 }
547
548 err = 0;
549 if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
550 goto out;
551
552 if (t->parms.iph.daddr == 0)
553 goto out;
554
555 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
556 goto out;
557
558 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
559 t->err_count++;
560 else
561 t->err_count = 1;
562 t->err_time = jiffies;
563out:
564 return err;
565}
566
567static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
568 const struct in6_addr *v6addr)
569{
570 __be32 v4embed = 0;
571 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
572 return true;
573 return false;
574}
575
576/* Checks if an address matches an address on the tunnel interface.
577 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
578 * Long story:
579 * This function is called after we considered the packet as spoofed
580 * in is_spoofed_6rd.
581 * We may have a router that is doing NAT for proto 41 packets
582 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
583 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
584 * function will return true, dropping the packet.
585 * But, we can still check if is spoofed against the IP
586 * addresses associated with the interface.
587 */
588static bool only_dnatted(const struct ip_tunnel *tunnel,
589 const struct in6_addr *v6dst)
590{
591 int prefix_len;
592
593#ifdef CONFIG_IPV6_SIT_6RD
594 prefix_len = tunnel->ip6rd.prefixlen + 32
595 - tunnel->ip6rd.relay_prefixlen;
596#else
597 prefix_len = 48;
598#endif
599 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
600}
601
602/* Returns true if a packet is spoofed */
603static bool packet_is_spoofed(struct sk_buff *skb,
604 const struct iphdr *iph,
605 struct ip_tunnel *tunnel)
606{
607 const struct ipv6hdr *ipv6h;
608
609 if (tunnel->dev->priv_flags & IFF_ISATAP) {
610 if (!isatap_chksrc(skb, iph, tunnel))
611 return true;
612
613 return false;
614 }
615
616 if (tunnel->dev->flags & IFF_POINTOPOINT)
617 return false;
618
619 ipv6h = ipv6_hdr(skb);
620
621 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
622 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
623 &iph->saddr, &ipv6h->saddr,
624 &iph->daddr, &ipv6h->daddr);
625 return true;
626 }
627
628 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
629 return false;
630
631 if (only_dnatted(tunnel, &ipv6h->daddr))
632 return false;
633
634 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
635 &iph->saddr, &ipv6h->saddr,
636 &iph->daddr, &ipv6h->daddr);
637 return true;
638}
639
640static int ipip6_rcv(struct sk_buff *skb)
641{
642 const struct iphdr *iph = ip_hdr(skb);
643 struct ip_tunnel *tunnel;
644 int sifindex;
645 int err;
646
647 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
648 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
649 iph->saddr, iph->daddr, sifindex);
650 if (tunnel) {
651 struct pcpu_sw_netstats *tstats;
652
653 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
654 tunnel->parms.iph.protocol != 0)
655 goto out;
656
657 skb->mac_header = skb->network_header;
658 skb_reset_network_header(skb);
659 IPCB(skb)->flags = 0;
660 skb->dev = tunnel->dev;
661
662 if (packet_is_spoofed(skb, iph, tunnel)) {
663 tunnel->dev->stats.rx_errors++;
664 goto out;
665 }
666
667 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
668 !net_eq(tunnel->net, dev_net(tunnel->dev))))
669 goto out;
670
671 err = IP_ECN_decapsulate(iph, skb);
672 if (unlikely(err)) {
673 if (log_ecn_error)
674 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
675 &iph->saddr, iph->tos);
676 if (err > 1) {
677 ++tunnel->dev->stats.rx_frame_errors;
678 ++tunnel->dev->stats.rx_errors;
679 goto out;
680 }
681 }
682
683 tstats = this_cpu_ptr(tunnel->dev->tstats);
684 u64_stats_update_begin(&tstats->syncp);
685 tstats->rx_packets++;
686 tstats->rx_bytes += skb->len;
687 u64_stats_update_end(&tstats->syncp);
688
689 netif_rx(skb);
690
691 return 0;
692 }
693
694 /* no tunnel matched, let upstream know, ipsec may handle it */
695 return 1;
696out:
697 kfree_skb(skb);
698 return 0;
699}
700
701static const struct tnl_ptk_info ipip_tpi = {
702 /* no tunnel info required for ipip. */
703 .proto = htons(ETH_P_IP),
704};
705
706#if IS_ENABLED(CONFIG_MPLS)
707static const struct tnl_ptk_info mplsip_tpi = {
708 /* no tunnel info required for mplsip. */
709 .proto = htons(ETH_P_MPLS_UC),
710};
711#endif
712
713static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
714{
715 const struct iphdr *iph;
716 struct ip_tunnel *tunnel;
717 int sifindex;
718
719 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
720
721 iph = ip_hdr(skb);
722 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
723 iph->saddr, iph->daddr, sifindex);
724 if (tunnel) {
725 const struct tnl_ptk_info *tpi;
726
727 if (tunnel->parms.iph.protocol != ipproto &&
728 tunnel->parms.iph.protocol != 0)
729 goto drop;
730
731 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
732 goto drop;
733#if IS_ENABLED(CONFIG_MPLS)
734 if (ipproto == IPPROTO_MPLS)
735 tpi = &mplsip_tpi;
736 else
737#endif
738 tpi = &ipip_tpi;
739 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
740 goto drop;
741 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
742 }
743
744 return 1;
745
746drop:
747 kfree_skb(skb);
748 return 0;
749}
750
751static int ipip_rcv(struct sk_buff *skb)
752{
753 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
754}
755
756#if IS_ENABLED(CONFIG_MPLS)
757static int mplsip_rcv(struct sk_buff *skb)
758{
759 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
760}
761#endif
762
763/*
764 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
765 * stores the embedded IPv4 address in v4dst and returns true.
766 */
767static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
768 __be32 *v4dst)
769{
770#ifdef CONFIG_IPV6_SIT_6RD
771 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
772 tunnel->ip6rd.prefixlen)) {
773 unsigned int pbw0, pbi0;
774 int pbi1;
775 u32 d;
776
777 pbw0 = tunnel->ip6rd.prefixlen >> 5;
778 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
779
780 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
781 tunnel->ip6rd.relay_prefixlen;
782
783 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
784 if (pbi1 > 0)
785 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
786 (32 - pbi1);
787
788 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
789 return true;
790 }
791#else
792 if (v6dst->s6_addr16[0] == htons(0x2002)) {
793 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
794 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
795 return true;
796 }
797#endif
798 return false;
799}
800
801static inline __be32 try_6rd(struct ip_tunnel *tunnel,
802 const struct in6_addr *v6dst)
803{
804 __be32 dst = 0;
805 check_6rd(tunnel, v6dst, &dst);
806 return dst;
807}
808
809/*
810 * This function assumes it is being called from dev_queue_xmit()
811 * and that skb is filled properly by that function.
812 */
813
814static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
815 struct net_device *dev)
816{
817 struct ip_tunnel *tunnel = netdev_priv(dev);
818 const struct iphdr *tiph = &tunnel->parms.iph;
819 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
820 u8 tos = tunnel->parms.iph.tos;
821 __be16 df = tiph->frag_off;
822 struct rtable *rt; /* Route to the other host */
823 struct net_device *tdev; /* Device to other host */
824 unsigned int max_headroom; /* The extra header space needed */
825 __be32 dst = tiph->daddr;
826 struct flowi4 fl4;
827 int mtu;
828 const struct in6_addr *addr6;
829 int addr_type;
830 u8 ttl;
831 u8 protocol = IPPROTO_IPV6;
832 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
833
834 if (tos == 1)
835 tos = ipv6_get_dsfield(iph6);
836
837 /* ISATAP (RFC4214) - must come before 6to4 */
838 if (dev->priv_flags & IFF_ISATAP) {
839 struct neighbour *neigh = NULL;
840 bool do_tx_error = false;
841
842 if (skb_dst(skb))
843 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
844
845 if (!neigh) {
846 net_dbg_ratelimited("nexthop == NULL\n");
847 goto tx_error;
848 }
849
850 addr6 = (const struct in6_addr *)&neigh->primary_key;
851 addr_type = ipv6_addr_type(addr6);
852
853 if ((addr_type & IPV6_ADDR_UNICAST) &&
854 ipv6_addr_is_isatap(addr6))
855 dst = addr6->s6_addr32[3];
856 else
857 do_tx_error = true;
858
859 neigh_release(neigh);
860 if (do_tx_error)
861 goto tx_error;
862 }
863
864 if (!dst)
865 dst = try_6rd(tunnel, &iph6->daddr);
866
867 if (!dst) {
868 struct neighbour *neigh = NULL;
869 bool do_tx_error = false;
870
871 if (skb_dst(skb))
872 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
873
874 if (!neigh) {
875 net_dbg_ratelimited("nexthop == NULL\n");
876 goto tx_error;
877 }
878
879 addr6 = (const struct in6_addr *)&neigh->primary_key;
880 addr_type = ipv6_addr_type(addr6);
881
882 if (addr_type == IPV6_ADDR_ANY) {
883 addr6 = &ipv6_hdr(skb)->daddr;
884 addr_type = ipv6_addr_type(addr6);
885 }
886
887 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
888 dst = addr6->s6_addr32[3];
889 else
890 do_tx_error = true;
891
892 neigh_release(neigh);
893 if (do_tx_error)
894 goto tx_error;
895 }
896
897 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
898 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
899 0, dst, tiph->saddr, 0, 0,
900 sock_net_uid(tunnel->net, NULL));
901 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
902
903 if (IS_ERR(rt)) {
904 dev->stats.tx_carrier_errors++;
905 goto tx_error_icmp;
906 }
907 if (rt->rt_type != RTN_UNICAST) {
908 ip_rt_put(rt);
909 dev->stats.tx_carrier_errors++;
910 goto tx_error_icmp;
911 }
912 tdev = rt->dst.dev;
913
914 if (tdev == dev) {
915 ip_rt_put(rt);
916 dev->stats.collisions++;
917 goto tx_error;
918 }
919
920 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
921 ip_rt_put(rt);
922 goto tx_error;
923 }
924
925 if (df) {
926 mtu = dst_mtu(&rt->dst) - t_hlen;
927
928 if (mtu < 68) {
929 dev->stats.collisions++;
930 ip_rt_put(rt);
931 goto tx_error;
932 }
933
934 if (mtu < IPV6_MIN_MTU) {
935 mtu = IPV6_MIN_MTU;
936 df = 0;
937 }
938
939 if (tunnel->parms.iph.daddr)
940 skb_dst_update_pmtu(skb, mtu);
941
942 if (skb->len > mtu && !skb_is_gso(skb)) {
943 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
944 ip_rt_put(rt);
945 goto tx_error;
946 }
947 }
948
949 if (tunnel->err_count > 0) {
950 if (time_before(jiffies,
951 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
952 tunnel->err_count--;
953 dst_link_failure(skb);
954 } else
955 tunnel->err_count = 0;
956 }
957
958 /*
959 * Okay, now see if we can stuff it in the buffer as-is.
960 */
961 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
962
963 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
964 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
965 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
966 if (!new_skb) {
967 ip_rt_put(rt);
968 dev->stats.tx_dropped++;
969 kfree_skb(skb);
970 return NETDEV_TX_OK;
971 }
972 if (skb->sk)
973 skb_set_owner_w(new_skb, skb->sk);
974 dev_kfree_skb(skb);
975 skb = new_skb;
976 iph6 = ipv6_hdr(skb);
977 }
978 ttl = tiph->ttl;
979 if (ttl == 0)
980 ttl = iph6->hop_limit;
981 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
982
983 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
984 ip_rt_put(rt);
985 goto tx_error;
986 }
987
988 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
989
990 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
991 df, !net_eq(tunnel->net, dev_net(dev)));
992 return NETDEV_TX_OK;
993
994tx_error_icmp:
995 dst_link_failure(skb);
996tx_error:
997 kfree_skb(skb);
998 dev->stats.tx_errors++;
999 return NETDEV_TX_OK;
1000}
1001
1002static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1003 struct net_device *dev, u8 ipproto)
1004{
1005 struct ip_tunnel *tunnel = netdev_priv(dev);
1006 const struct iphdr *tiph = &tunnel->parms.iph;
1007
1008 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1009 goto tx_error;
1010
1011 skb_set_inner_ipproto(skb, ipproto);
1012
1013 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1014 return NETDEV_TX_OK;
1015tx_error:
1016 kfree_skb(skb);
1017 dev->stats.tx_errors++;
1018 return NETDEV_TX_OK;
1019}
1020
1021static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1022 struct net_device *dev)
1023{
1024 switch (skb->protocol) {
1025 case htons(ETH_P_IP):
1026 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1027 break;
1028 case htons(ETH_P_IPV6):
1029 ipip6_tunnel_xmit(skb, dev);
1030 break;
1031#if IS_ENABLED(CONFIG_MPLS)
1032 case htons(ETH_P_MPLS_UC):
1033 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1034 break;
1035#endif
1036 default:
1037 goto tx_err;
1038 }
1039
1040 return NETDEV_TX_OK;
1041
1042tx_err:
1043 dev->stats.tx_errors++;
1044 kfree_skb(skb);
1045 return NETDEV_TX_OK;
1046
1047}
1048
1049static void ipip6_tunnel_bind_dev(struct net_device *dev)
1050{
1051 struct net_device *tdev = NULL;
1052 struct ip_tunnel *tunnel;
1053 const struct iphdr *iph;
1054 struct flowi4 fl4;
1055
1056 tunnel = netdev_priv(dev);
1057 iph = &tunnel->parms.iph;
1058
1059 if (iph->daddr) {
1060 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1061 NULL,
1062 iph->daddr, iph->saddr,
1063 0, 0,
1064 IPPROTO_IPV6,
1065 RT_TOS(iph->tos),
1066 tunnel->parms.link);
1067
1068 if (!IS_ERR(rt)) {
1069 tdev = rt->dst.dev;
1070 ip_rt_put(rt);
1071 }
1072 dev->flags |= IFF_POINTOPOINT;
1073 }
1074
1075 if (!tdev && tunnel->parms.link)
1076 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1077
1078 if (tdev) {
1079 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1080
1081 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1082 dev->mtu = tdev->mtu - t_hlen;
1083 if (dev->mtu < IPV6_MIN_MTU)
1084 dev->mtu = IPV6_MIN_MTU;
1085 }
1086}
1087
1088static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1089 __u32 fwmark)
1090{
1091 struct net *net = t->net;
1092 struct sit_net *sitn = net_generic(net, sit_net_id);
1093
1094 ipip6_tunnel_unlink(sitn, t);
1095 synchronize_net();
1096 t->parms.iph.saddr = p->iph.saddr;
1097 t->parms.iph.daddr = p->iph.daddr;
1098 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1099 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1100 ipip6_tunnel_link(sitn, t);
1101 t->parms.iph.ttl = p->iph.ttl;
1102 t->parms.iph.tos = p->iph.tos;
1103 t->parms.iph.frag_off = p->iph.frag_off;
1104 if (t->parms.link != p->link || t->fwmark != fwmark) {
1105 t->parms.link = p->link;
1106 t->fwmark = fwmark;
1107 ipip6_tunnel_bind_dev(t->dev);
1108 }
1109 dst_cache_reset(&t->dst_cache);
1110 netdev_state_change(t->dev);
1111}
1112
1113#ifdef CONFIG_IPV6_SIT_6RD
1114static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1115 struct ip_tunnel_6rd *ip6rd)
1116{
1117 struct in6_addr prefix;
1118 __be32 relay_prefix;
1119
1120 if (ip6rd->relay_prefixlen > 32 ||
1121 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1122 return -EINVAL;
1123
1124 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1125 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1126 return -EINVAL;
1127 if (ip6rd->relay_prefixlen)
1128 relay_prefix = ip6rd->relay_prefix &
1129 htonl(0xffffffffUL <<
1130 (32 - ip6rd->relay_prefixlen));
1131 else
1132 relay_prefix = 0;
1133 if (relay_prefix != ip6rd->relay_prefix)
1134 return -EINVAL;
1135
1136 t->ip6rd.prefix = prefix;
1137 t->ip6rd.relay_prefix = relay_prefix;
1138 t->ip6rd.prefixlen = ip6rd->prefixlen;
1139 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1140 dst_cache_reset(&t->dst_cache);
1141 netdev_state_change(t->dev);
1142 return 0;
1143}
1144#endif
1145
1146static bool ipip6_valid_ip_proto(u8 ipproto)
1147{
1148 return ipproto == IPPROTO_IPV6 ||
1149 ipproto == IPPROTO_IPIP ||
1150#if IS_ENABLED(CONFIG_MPLS)
1151 ipproto == IPPROTO_MPLS ||
1152#endif
1153 ipproto == 0;
1154}
1155
1156static int
1157ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1158{
1159 int err = 0;
1160 struct ip_tunnel_parm p;
1161 struct ip_tunnel_prl prl;
1162 struct ip_tunnel *t = netdev_priv(dev);
1163 struct net *net = t->net;
1164 struct sit_net *sitn = net_generic(net, sit_net_id);
1165#ifdef CONFIG_IPV6_SIT_6RD
1166 struct ip_tunnel_6rd ip6rd;
1167#endif
1168
1169 switch (cmd) {
1170 case SIOCGETTUNNEL:
1171#ifdef CONFIG_IPV6_SIT_6RD
1172 case SIOCGET6RD:
1173#endif
1174 if (dev == sitn->fb_tunnel_dev) {
1175 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1176 err = -EFAULT;
1177 break;
1178 }
1179 t = ipip6_tunnel_locate(net, &p, 0);
1180 if (!t)
1181 t = netdev_priv(dev);
1182 }
1183
1184 err = -EFAULT;
1185 if (cmd == SIOCGETTUNNEL) {
1186 memcpy(&p, &t->parms, sizeof(p));
1187 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1188 sizeof(p)))
1189 goto done;
1190#ifdef CONFIG_IPV6_SIT_6RD
1191 } else {
1192 ip6rd.prefix = t->ip6rd.prefix;
1193 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1194 ip6rd.prefixlen = t->ip6rd.prefixlen;
1195 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1196 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1197 sizeof(ip6rd)))
1198 goto done;
1199#endif
1200 }
1201 err = 0;
1202 break;
1203
1204 case SIOCADDTUNNEL:
1205 case SIOCCHGTUNNEL:
1206 err = -EPERM;
1207 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1208 goto done;
1209
1210 err = -EFAULT;
1211 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1212 goto done;
1213
1214 err = -EINVAL;
1215 if (!ipip6_valid_ip_proto(p.iph.protocol))
1216 goto done;
1217 if (p.iph.version != 4 ||
1218 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1219 goto done;
1220 if (p.iph.ttl)
1221 p.iph.frag_off |= htons(IP_DF);
1222
1223 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1224
1225 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1226 if (t) {
1227 if (t->dev != dev) {
1228 err = -EEXIST;
1229 break;
1230 }
1231 } else {
1232 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1233 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1234 err = -EINVAL;
1235 break;
1236 }
1237 t = netdev_priv(dev);
1238 }
1239
1240 ipip6_tunnel_update(t, &p, t->fwmark);
1241 }
1242
1243 if (t) {
1244 err = 0;
1245 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1246 err = -EFAULT;
1247 } else
1248 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1249 break;
1250
1251 case SIOCDELTUNNEL:
1252 err = -EPERM;
1253 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1254 goto done;
1255
1256 if (dev == sitn->fb_tunnel_dev) {
1257 err = -EFAULT;
1258 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1259 goto done;
1260 err = -ENOENT;
1261 t = ipip6_tunnel_locate(net, &p, 0);
1262 if (!t)
1263 goto done;
1264 err = -EPERM;
1265 if (t == netdev_priv(sitn->fb_tunnel_dev))
1266 goto done;
1267 dev = t->dev;
1268 }
1269 unregister_netdevice(dev);
1270 err = 0;
1271 break;
1272
1273 case SIOCGETPRL:
1274 err = -EINVAL;
1275 if (dev == sitn->fb_tunnel_dev)
1276 goto done;
1277 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1278 break;
1279
1280 case SIOCADDPRL:
1281 case SIOCDELPRL:
1282 case SIOCCHGPRL:
1283 err = -EPERM;
1284 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1285 goto done;
1286 err = -EINVAL;
1287 if (dev == sitn->fb_tunnel_dev)
1288 goto done;
1289 err = -EFAULT;
1290 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1291 goto done;
1292
1293 switch (cmd) {
1294 case SIOCDELPRL:
1295 err = ipip6_tunnel_del_prl(t, &prl);
1296 break;
1297 case SIOCADDPRL:
1298 case SIOCCHGPRL:
1299 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1300 break;
1301 }
1302 dst_cache_reset(&t->dst_cache);
1303 netdev_state_change(dev);
1304 break;
1305
1306#ifdef CONFIG_IPV6_SIT_6RD
1307 case SIOCADD6RD:
1308 case SIOCCHG6RD:
1309 case SIOCDEL6RD:
1310 err = -EPERM;
1311 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1312 goto done;
1313
1314 err = -EFAULT;
1315 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1316 sizeof(ip6rd)))
1317 goto done;
1318
1319 if (cmd != SIOCDEL6RD) {
1320 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1321 if (err < 0)
1322 goto done;
1323 } else
1324 ipip6_tunnel_clone_6rd(dev, sitn);
1325
1326 err = 0;
1327 break;
1328#endif
1329
1330 default:
1331 err = -EINVAL;
1332 }
1333
1334done:
1335 return err;
1336}
1337
1338static const struct net_device_ops ipip6_netdev_ops = {
1339 .ndo_init = ipip6_tunnel_init,
1340 .ndo_uninit = ipip6_tunnel_uninit,
1341 .ndo_start_xmit = sit_tunnel_xmit,
1342 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1343 .ndo_get_stats64 = ip_tunnel_get_stats64,
1344 .ndo_get_iflink = ip_tunnel_get_iflink,
1345};
1346
1347static void ipip6_dev_free(struct net_device *dev)
1348{
1349 struct ip_tunnel *tunnel = netdev_priv(dev);
1350
1351 dst_cache_destroy(&tunnel->dst_cache);
1352 free_percpu(dev->tstats);
1353}
1354
1355#define SIT_FEATURES (NETIF_F_SG | \
1356 NETIF_F_FRAGLIST | \
1357 NETIF_F_HIGHDMA | \
1358 NETIF_F_GSO_SOFTWARE | \
1359 NETIF_F_HW_CSUM)
1360
1361static void ipip6_tunnel_setup(struct net_device *dev)
1362{
1363 struct ip_tunnel *tunnel = netdev_priv(dev);
1364 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1365
1366 dev->netdev_ops = &ipip6_netdev_ops;
1367 dev->needs_free_netdev = true;
1368 dev->priv_destructor = ipip6_dev_free;
1369
1370 dev->type = ARPHRD_SIT;
1371 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1372 dev->mtu = ETH_DATA_LEN - t_hlen;
1373 dev->min_mtu = IPV6_MIN_MTU;
1374 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1375 dev->flags = IFF_NOARP;
1376 netif_keep_dst(dev);
1377 dev->addr_len = 4;
1378 dev->features |= NETIF_F_LLTX;
1379 dev->features |= SIT_FEATURES;
1380 dev->hw_features |= SIT_FEATURES;
1381}
1382
1383static int ipip6_tunnel_init(struct net_device *dev)
1384{
1385 struct ip_tunnel *tunnel = netdev_priv(dev);
1386 int err;
1387
1388 tunnel->dev = dev;
1389 tunnel->net = dev_net(dev);
1390 strcpy(tunnel->parms.name, dev->name);
1391
1392 ipip6_tunnel_bind_dev(dev);
1393 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1394 if (!dev->tstats)
1395 return -ENOMEM;
1396
1397 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1398 if (err) {
1399 free_percpu(dev->tstats);
1400 dev->tstats = NULL;
1401 return err;
1402 }
1403
1404 return 0;
1405}
1406
1407static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1408{
1409 struct ip_tunnel *tunnel = netdev_priv(dev);
1410 struct iphdr *iph = &tunnel->parms.iph;
1411 struct net *net = dev_net(dev);
1412 struct sit_net *sitn = net_generic(net, sit_net_id);
1413
1414 iph->version = 4;
1415 iph->protocol = IPPROTO_IPV6;
1416 iph->ihl = 5;
1417 iph->ttl = 64;
1418
1419 dev_hold(dev);
1420 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1421}
1422
1423static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1424 struct netlink_ext_ack *extack)
1425{
1426 u8 proto;
1427
1428 if (!data || !data[IFLA_IPTUN_PROTO])
1429 return 0;
1430
1431 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1432 if (!ipip6_valid_ip_proto(proto))
1433 return -EINVAL;
1434
1435 return 0;
1436}
1437
1438static void ipip6_netlink_parms(struct nlattr *data[],
1439 struct ip_tunnel_parm *parms,
1440 __u32 *fwmark)
1441{
1442 memset(parms, 0, sizeof(*parms));
1443
1444 parms->iph.version = 4;
1445 parms->iph.protocol = IPPROTO_IPV6;
1446 parms->iph.ihl = 5;
1447 parms->iph.ttl = 64;
1448
1449 if (!data)
1450 return;
1451
1452 if (data[IFLA_IPTUN_LINK])
1453 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1454
1455 if (data[IFLA_IPTUN_LOCAL])
1456 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1457
1458 if (data[IFLA_IPTUN_REMOTE])
1459 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1460
1461 if (data[IFLA_IPTUN_TTL]) {
1462 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1463 if (parms->iph.ttl)
1464 parms->iph.frag_off = htons(IP_DF);
1465 }
1466
1467 if (data[IFLA_IPTUN_TOS])
1468 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1469
1470 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1471 parms->iph.frag_off = htons(IP_DF);
1472
1473 if (data[IFLA_IPTUN_FLAGS])
1474 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1475
1476 if (data[IFLA_IPTUN_PROTO])
1477 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1478
1479 if (data[IFLA_IPTUN_FWMARK])
1480 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1481}
1482
1483/* This function returns true when ENCAP attributes are present in the nl msg */
1484static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1485 struct ip_tunnel_encap *ipencap)
1486{
1487 bool ret = false;
1488
1489 memset(ipencap, 0, sizeof(*ipencap));
1490
1491 if (!data)
1492 return ret;
1493
1494 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1495 ret = true;
1496 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1497 }
1498
1499 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1500 ret = true;
1501 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1502 }
1503
1504 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1505 ret = true;
1506 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1507 }
1508
1509 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1510 ret = true;
1511 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1512 }
1513
1514 return ret;
1515}
1516
1517#ifdef CONFIG_IPV6_SIT_6RD
1518/* This function returns true when 6RD attributes are present in the nl msg */
1519static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1520 struct ip_tunnel_6rd *ip6rd)
1521{
1522 bool ret = false;
1523 memset(ip6rd, 0, sizeof(*ip6rd));
1524
1525 if (!data)
1526 return ret;
1527
1528 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1529 ret = true;
1530 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1531 }
1532
1533 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1534 ret = true;
1535 ip6rd->relay_prefix =
1536 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1537 }
1538
1539 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1540 ret = true;
1541 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1542 }
1543
1544 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1545 ret = true;
1546 ip6rd->relay_prefixlen =
1547 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1548 }
1549
1550 return ret;
1551}
1552#endif
1553
1554static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1555 struct nlattr *tb[], struct nlattr *data[],
1556 struct netlink_ext_ack *extack)
1557{
1558 struct net *net = dev_net(dev);
1559 struct ip_tunnel *nt;
1560 struct ip_tunnel_encap ipencap;
1561#ifdef CONFIG_IPV6_SIT_6RD
1562 struct ip_tunnel_6rd ip6rd;
1563#endif
1564 int err;
1565
1566 nt = netdev_priv(dev);
1567
1568 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1569 err = ip_tunnel_encap_setup(nt, &ipencap);
1570 if (err < 0)
1571 return err;
1572 }
1573
1574 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1575
1576 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1577 return -EEXIST;
1578
1579 err = ipip6_tunnel_create(dev);
1580 if (err < 0)
1581 return err;
1582
1583 if (tb[IFLA_MTU]) {
1584 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1585
1586 if (mtu >= IPV6_MIN_MTU &&
1587 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1588 dev->mtu = mtu;
1589 }
1590
1591#ifdef CONFIG_IPV6_SIT_6RD
1592 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1593 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1594#endif
1595
1596 return err;
1597}
1598
1599static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1600 struct nlattr *data[],
1601 struct netlink_ext_ack *extack)
1602{
1603 struct ip_tunnel *t = netdev_priv(dev);
1604 struct ip_tunnel_parm p;
1605 struct ip_tunnel_encap ipencap;
1606 struct net *net = t->net;
1607 struct sit_net *sitn = net_generic(net, sit_net_id);
1608#ifdef CONFIG_IPV6_SIT_6RD
1609 struct ip_tunnel_6rd ip6rd;
1610#endif
1611 __u32 fwmark = t->fwmark;
1612 int err;
1613
1614 if (dev == sitn->fb_tunnel_dev)
1615 return -EINVAL;
1616
1617 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1618 err = ip_tunnel_encap_setup(t, &ipencap);
1619 if (err < 0)
1620 return err;
1621 }
1622
1623 ipip6_netlink_parms(data, &p, &fwmark);
1624
1625 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1626 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1627 return -EINVAL;
1628
1629 t = ipip6_tunnel_locate(net, &p, 0);
1630
1631 if (t) {
1632 if (t->dev != dev)
1633 return -EEXIST;
1634 } else
1635 t = netdev_priv(dev);
1636
1637 ipip6_tunnel_update(t, &p, fwmark);
1638
1639#ifdef CONFIG_IPV6_SIT_6RD
1640 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1641 return ipip6_tunnel_update_6rd(t, &ip6rd);
1642#endif
1643
1644 return 0;
1645}
1646
1647static size_t ipip6_get_size(const struct net_device *dev)
1648{
1649 return
1650 /* IFLA_IPTUN_LINK */
1651 nla_total_size(4) +
1652 /* IFLA_IPTUN_LOCAL */
1653 nla_total_size(4) +
1654 /* IFLA_IPTUN_REMOTE */
1655 nla_total_size(4) +
1656 /* IFLA_IPTUN_TTL */
1657 nla_total_size(1) +
1658 /* IFLA_IPTUN_TOS */
1659 nla_total_size(1) +
1660 /* IFLA_IPTUN_PMTUDISC */
1661 nla_total_size(1) +
1662 /* IFLA_IPTUN_FLAGS */
1663 nla_total_size(2) +
1664 /* IFLA_IPTUN_PROTO */
1665 nla_total_size(1) +
1666#ifdef CONFIG_IPV6_SIT_6RD
1667 /* IFLA_IPTUN_6RD_PREFIX */
1668 nla_total_size(sizeof(struct in6_addr)) +
1669 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1670 nla_total_size(4) +
1671 /* IFLA_IPTUN_6RD_PREFIXLEN */
1672 nla_total_size(2) +
1673 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1674 nla_total_size(2) +
1675#endif
1676 /* IFLA_IPTUN_ENCAP_TYPE */
1677 nla_total_size(2) +
1678 /* IFLA_IPTUN_ENCAP_FLAGS */
1679 nla_total_size(2) +
1680 /* IFLA_IPTUN_ENCAP_SPORT */
1681 nla_total_size(2) +
1682 /* IFLA_IPTUN_ENCAP_DPORT */
1683 nla_total_size(2) +
1684 /* IFLA_IPTUN_FWMARK */
1685 nla_total_size(4) +
1686 0;
1687}
1688
1689static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1690{
1691 struct ip_tunnel *tunnel = netdev_priv(dev);
1692 struct ip_tunnel_parm *parm = &tunnel->parms;
1693
1694 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1695 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1696 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1697 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1698 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1699 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1700 !!(parm->iph.frag_off & htons(IP_DF))) ||
1701 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1702 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1703 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1704 goto nla_put_failure;
1705
1706#ifdef CONFIG_IPV6_SIT_6RD
1707 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1708 &tunnel->ip6rd.prefix) ||
1709 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1710 tunnel->ip6rd.relay_prefix) ||
1711 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1712 tunnel->ip6rd.prefixlen) ||
1713 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1714 tunnel->ip6rd.relay_prefixlen))
1715 goto nla_put_failure;
1716#endif
1717
1718 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1719 tunnel->encap.type) ||
1720 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1721 tunnel->encap.sport) ||
1722 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1723 tunnel->encap.dport) ||
1724 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1725 tunnel->encap.flags))
1726 goto nla_put_failure;
1727
1728 return 0;
1729
1730nla_put_failure:
1731 return -EMSGSIZE;
1732}
1733
1734static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1735 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1736 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1737 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1738 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1739 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1740 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1741 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1742 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1743#ifdef CONFIG_IPV6_SIT_6RD
1744 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1745 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1746 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1747 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1748#endif
1749 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1750 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1751 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1752 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1753 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1754};
1755
1756static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1757{
1758 struct net *net = dev_net(dev);
1759 struct sit_net *sitn = net_generic(net, sit_net_id);
1760
1761 if (dev != sitn->fb_tunnel_dev)
1762 unregister_netdevice_queue(dev, head);
1763}
1764
1765static struct rtnl_link_ops sit_link_ops __read_mostly = {
1766 .kind = "sit",
1767 .maxtype = IFLA_IPTUN_MAX,
1768 .policy = ipip6_policy,
1769 .priv_size = sizeof(struct ip_tunnel),
1770 .setup = ipip6_tunnel_setup,
1771 .validate = ipip6_validate,
1772 .newlink = ipip6_newlink,
1773 .changelink = ipip6_changelink,
1774 .get_size = ipip6_get_size,
1775 .fill_info = ipip6_fill_info,
1776 .dellink = ipip6_dellink,
1777 .get_link_net = ip_tunnel_get_link_net,
1778};
1779
1780static struct xfrm_tunnel sit_handler __read_mostly = {
1781 .handler = ipip6_rcv,
1782 .err_handler = ipip6_err,
1783 .priority = 1,
1784};
1785
1786static struct xfrm_tunnel ipip_handler __read_mostly = {
1787 .handler = ipip_rcv,
1788 .err_handler = ipip6_err,
1789 .priority = 2,
1790};
1791
1792#if IS_ENABLED(CONFIG_MPLS)
1793static struct xfrm_tunnel mplsip_handler __read_mostly = {
1794 .handler = mplsip_rcv,
1795 .err_handler = ipip6_err,
1796 .priority = 2,
1797};
1798#endif
1799
1800static void __net_exit sit_destroy_tunnels(struct net *net,
1801 struct list_head *head)
1802{
1803 struct sit_net *sitn = net_generic(net, sit_net_id);
1804 struct net_device *dev, *aux;
1805 int prio;
1806
1807 for_each_netdev_safe(net, dev, aux)
1808 if (dev->rtnl_link_ops == &sit_link_ops)
1809 unregister_netdevice_queue(dev, head);
1810
1811 for (prio = 1; prio < 4; prio++) {
1812 int h;
1813 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1814 struct ip_tunnel *t;
1815
1816 t = rtnl_dereference(sitn->tunnels[prio][h]);
1817 while (t) {
1818 /* If dev is in the same netns, it has already
1819 * been added to the list by the previous loop.
1820 */
1821 if (!net_eq(dev_net(t->dev), net))
1822 unregister_netdevice_queue(t->dev,
1823 head);
1824 t = rtnl_dereference(t->next);
1825 }
1826 }
1827 }
1828}
1829
1830static int __net_init sit_init_net(struct net *net)
1831{
1832 struct sit_net *sitn = net_generic(net, sit_net_id);
1833 struct ip_tunnel *t;
1834 int err;
1835
1836 sitn->tunnels[0] = sitn->tunnels_wc;
1837 sitn->tunnels[1] = sitn->tunnels_l;
1838 sitn->tunnels[2] = sitn->tunnels_r;
1839 sitn->tunnels[3] = sitn->tunnels_r_l;
1840
1841 if (!net_has_fallback_tunnels(net))
1842 return 0;
1843
1844 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1845 NET_NAME_UNKNOWN,
1846 ipip6_tunnel_setup);
1847 if (!sitn->fb_tunnel_dev) {
1848 err = -ENOMEM;
1849 goto err_alloc_dev;
1850 }
1851 dev_net_set(sitn->fb_tunnel_dev, net);
1852 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1853 /* FB netdevice is special: we have one, and only one per netns.
1854 * Allowing to move it to another netns is clearly unsafe.
1855 */
1856 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1857
1858 err = register_netdev(sitn->fb_tunnel_dev);
1859 if (err)
1860 goto err_reg_dev;
1861
1862 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1863 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1864
1865 t = netdev_priv(sitn->fb_tunnel_dev);
1866
1867 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1868 return 0;
1869
1870err_reg_dev:
1871 ipip6_dev_free(sitn->fb_tunnel_dev);
1872err_alloc_dev:
1873 return err;
1874}
1875
1876static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1877{
1878 LIST_HEAD(list);
1879 struct net *net;
1880
1881 rtnl_lock();
1882 list_for_each_entry(net, net_list, exit_list)
1883 sit_destroy_tunnels(net, &list);
1884
1885 unregister_netdevice_many(&list);
1886 rtnl_unlock();
1887}
1888
1889static struct pernet_operations sit_net_ops = {
1890 .init = sit_init_net,
1891 .exit_batch = sit_exit_batch_net,
1892 .id = &sit_net_id,
1893 .size = sizeof(struct sit_net),
1894};
1895
1896static void __exit sit_cleanup(void)
1897{
1898 rtnl_link_unregister(&sit_link_ops);
1899 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1900 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1901#if IS_ENABLED(CONFIG_MPLS)
1902 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1903#endif
1904
1905 unregister_pernet_device(&sit_net_ops);
1906 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1907}
1908
1909static int __init sit_init(void)
1910{
1911 int err;
1912
1913 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1914
1915 err = register_pernet_device(&sit_net_ops);
1916 if (err < 0)
1917 return err;
1918 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1919 if (err < 0) {
1920 pr_info("%s: can't register ip6ip4\n", __func__);
1921 goto xfrm_tunnel_failed;
1922 }
1923 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1924 if (err < 0) {
1925 pr_info("%s: can't register ip4ip4\n", __func__);
1926 goto xfrm_tunnel4_failed;
1927 }
1928#if IS_ENABLED(CONFIG_MPLS)
1929 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1930 if (err < 0) {
1931 pr_info("%s: can't register mplsip\n", __func__);
1932 goto xfrm_tunnel_mpls_failed;
1933 }
1934#endif
1935 err = rtnl_link_register(&sit_link_ops);
1936 if (err < 0)
1937 goto rtnl_link_failed;
1938
1939out:
1940 return err;
1941
1942rtnl_link_failed:
1943#if IS_ENABLED(CONFIG_MPLS)
1944 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1945xfrm_tunnel_mpls_failed:
1946#endif
1947 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1948xfrm_tunnel4_failed:
1949 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1950xfrm_tunnel_failed:
1951 unregister_pernet_device(&sit_net_ops);
1952 goto out;
1953}
1954
1955module_init(sit_init);
1956module_exit(sit_cleanup);
1957MODULE_LICENSE("GPL");
1958MODULE_ALIAS_RTNL_LINK("sit");
1959MODULE_ALIAS_NETDEV("sit0");