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
86/*
87 * Must be invoked with rcu_read_lock
88 */
89static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
90 struct net_device *dev,
91 __be32 remote, __be32 local,
92 int sifindex)
93{
94 unsigned int h0 = HASH(remote);
95 unsigned int h1 = HASH(local);
96 struct ip_tunnel *t;
97 struct sit_net *sitn = net_generic(net, sit_net_id);
98 int ifindex = dev ? dev->ifindex : 0;
99
100 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
101 if (local == t->parms.iph.saddr &&
102 remote == t->parms.iph.daddr &&
103 (!dev || !t->parms.link || ifindex == t->parms.link ||
104 sifindex == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
106 return t;
107 }
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || 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_l[h1]) {
116 if (local == t->parms.iph.saddr &&
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 t = rcu_dereference(sitn->tunnels_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
124 return t;
125 return NULL;
126}
127
128static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
129 struct ip_tunnel_parm *parms)
130{
131 __be32 remote = parms->iph.daddr;
132 __be32 local = parms->iph.saddr;
133 unsigned int h = 0;
134 int prio = 0;
135
136 if (remote) {
137 prio |= 2;
138 h ^= HASH(remote);
139 }
140 if (local) {
141 prio |= 1;
142 h ^= HASH(local);
143 }
144 return &sitn->tunnels[prio][h];
145}
146
147static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
148 struct ip_tunnel *t)
149{
150 return __ipip6_bucket(sitn, &t->parms);
151}
152
153static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
154{
155 struct ip_tunnel __rcu **tp;
156 struct ip_tunnel *iter;
157
158 for (tp = ipip6_bucket(sitn, t);
159 (iter = rtnl_dereference(*tp)) != NULL;
160 tp = &iter->next) {
161 if (t == iter) {
162 rcu_assign_pointer(*tp, t->next);
163 break;
164 }
165 }
166}
167
168static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
169{
170 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
171
172 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
173 rcu_assign_pointer(*tp, t);
174}
175
176static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
177{
178#ifdef CONFIG_IPV6_SIT_6RD
179 struct ip_tunnel *t = netdev_priv(dev);
180
181 if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
182 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
183 t->ip6rd.relay_prefix = 0;
184 t->ip6rd.prefixlen = 16;
185 t->ip6rd.relay_prefixlen = 0;
186 } else {
187 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
188 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
189 }
190#endif
191}
192
193static int ipip6_tunnel_create(struct net_device *dev)
194{
195 struct ip_tunnel *t = netdev_priv(dev);
196 struct net *net = dev_net(dev);
197 struct sit_net *sitn = net_generic(net, sit_net_id);
198 int err;
199
200 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
201 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
202
203 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 dev->priv_flags |= IFF_ISATAP;
205
206 dev->rtnl_link_ops = &sit_link_ops;
207
208 err = register_netdevice(dev);
209 if (err < 0)
210 goto out;
211
212 ipip6_tunnel_clone_6rd(dev, sitn);
213
214 dev_hold(dev);
215
216 ipip6_tunnel_link(sitn, t);
217 return 0;
218
219out:
220 return err;
221}
222
223static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
224 struct ip_tunnel_parm *parms, int create)
225{
226 __be32 remote = parms->iph.daddr;
227 __be32 local = parms->iph.saddr;
228 struct ip_tunnel *t, *nt;
229 struct ip_tunnel __rcu **tp;
230 struct net_device *dev;
231 char name[IFNAMSIZ];
232 struct sit_net *sitn = net_generic(net, sit_net_id);
233
234 for (tp = __ipip6_bucket(sitn, parms);
235 (t = rtnl_dereference(*tp)) != NULL;
236 tp = &t->next) {
237 if (local == t->parms.iph.saddr &&
238 remote == t->parms.iph.daddr &&
239 parms->link == t->parms.link) {
240 if (create)
241 return NULL;
242 else
243 return t;
244 }
245 }
246 if (!create)
247 goto failed;
248
249 if (parms->name[0]) {
250 if (!dev_valid_name(parms->name))
251 goto failed;
252 strlcpy(name, parms->name, IFNAMSIZ);
253 } else {
254 strcpy(name, "sit%d");
255 }
256 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
257 ipip6_tunnel_setup);
258 if (!dev)
259 return NULL;
260
261 dev_net_set(dev, net);
262
263 nt = netdev_priv(dev);
264
265 nt->parms = *parms;
266 if (ipip6_tunnel_create(dev) < 0)
267 goto failed_free;
268
269 return nt;
270
271failed_free:
272 free_netdev(dev);
273failed:
274 return NULL;
275}
276
277#define for_each_prl_rcu(start) \
278 for (prl = rcu_dereference(start); \
279 prl; \
280 prl = rcu_dereference(prl->next))
281
282static struct ip_tunnel_prl_entry *
283__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
284{
285 struct ip_tunnel_prl_entry *prl;
286
287 for_each_prl_rcu(t->prl)
288 if (prl->addr == addr)
289 break;
290 return prl;
291
292}
293
294static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
295 struct ip_tunnel_prl __user *a)
296{
297 struct ip_tunnel_prl kprl, *kp;
298 struct ip_tunnel_prl_entry *prl;
299 unsigned int cmax, c = 0, ca, len;
300 int ret = 0;
301
302 if (copy_from_user(&kprl, a, sizeof(kprl)))
303 return -EFAULT;
304 cmax = kprl.datalen / sizeof(kprl);
305 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
306 cmax = 1;
307
308 /* For simple GET or for root users,
309 * we try harder to allocate.
310 */
311 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
312 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
313 NULL;
314
315 rcu_read_lock();
316
317 ca = t->prl_count < cmax ? t->prl_count : cmax;
318
319 if (!kp) {
320 /* We don't try hard to allocate much memory for
321 * non-root users.
322 * For root users, retry allocating enough memory for
323 * the answer.
324 */
325 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
326 if (!kp) {
327 ret = -ENOMEM;
328 goto out;
329 }
330 }
331
332 c = 0;
333 for_each_prl_rcu(t->prl) {
334 if (c >= cmax)
335 break;
336 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
337 continue;
338 kp[c].addr = prl->addr;
339 kp[c].flags = prl->flags;
340 c++;
341 if (kprl.addr != htonl(INADDR_ANY))
342 break;
343 }
344out:
345 rcu_read_unlock();
346
347 len = sizeof(*kp) * c;
348 ret = 0;
349 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
350 ret = -EFAULT;
351
352 kfree(kp);
353
354 return ret;
355}
356
357static int
358ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
359{
360 struct ip_tunnel_prl_entry *p;
361 int err = 0;
362
363 if (a->addr == htonl(INADDR_ANY))
364 return -EINVAL;
365
366 ASSERT_RTNL();
367
368 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
369 if (p->addr == a->addr) {
370 if (chg) {
371 p->flags = a->flags;
372 goto out;
373 }
374 err = -EEXIST;
375 goto out;
376 }
377 }
378
379 if (chg) {
380 err = -ENXIO;
381 goto out;
382 }
383
384 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
385 if (!p) {
386 err = -ENOBUFS;
387 goto out;
388 }
389
390 p->next = t->prl;
391 p->addr = a->addr;
392 p->flags = a->flags;
393 t->prl_count++;
394 rcu_assign_pointer(t->prl, p);
395out:
396 return err;
397}
398
399static void prl_list_destroy_rcu(struct rcu_head *head)
400{
401 struct ip_tunnel_prl_entry *p, *n;
402
403 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
404 do {
405 n = rcu_dereference_protected(p->next, 1);
406 kfree(p);
407 p = n;
408 } while (p);
409}
410
411static int
412ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
413{
414 struct ip_tunnel_prl_entry *x;
415 struct ip_tunnel_prl_entry __rcu **p;
416 int err = 0;
417
418 ASSERT_RTNL();
419
420 if (a && a->addr != htonl(INADDR_ANY)) {
421 for (p = &t->prl;
422 (x = rtnl_dereference(*p)) != NULL;
423 p = &x->next) {
424 if (x->addr == a->addr) {
425 *p = x->next;
426 kfree_rcu(x, rcu_head);
427 t->prl_count--;
428 goto out;
429 }
430 }
431 err = -ENXIO;
432 } else {
433 x = rtnl_dereference(t->prl);
434 if (x) {
435 t->prl_count = 0;
436 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
437 t->prl = NULL;
438 }
439 }
440out:
441 return err;
442}
443
444static int
445isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
446{
447 struct ip_tunnel_prl_entry *p;
448 int ok = 1;
449
450 rcu_read_lock();
451 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
452 if (p) {
453 if (p->flags & PRL_DEFAULT)
454 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
455 else
456 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
457 } else {
458 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
459
460 if (ipv6_addr_is_isatap(addr6) &&
461 (addr6->s6_addr32[3] == iph->saddr) &&
462 ipv6_chk_prefix(addr6, t->dev))
463 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
464 else
465 ok = 0;
466 }
467 rcu_read_unlock();
468 return ok;
469}
470
471static void ipip6_tunnel_uninit(struct net_device *dev)
472{
473 struct ip_tunnel *tunnel = netdev_priv(dev);
474 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
475
476 if (dev == sitn->fb_tunnel_dev) {
477 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
478 } else {
479 ipip6_tunnel_unlink(sitn, tunnel);
480 ipip6_tunnel_del_prl(tunnel, NULL);
481 }
482 dst_cache_reset(&tunnel->dst_cache);
483 dev_put(dev);
484}
485
486static int ipip6_err(struct sk_buff *skb, u32 info)
487{
488 const struct iphdr *iph = (const struct iphdr *)skb->data;
489 const int type = icmp_hdr(skb)->type;
490 const int code = icmp_hdr(skb)->code;
491 unsigned int data_len = 0;
492 struct ip_tunnel *t;
493 int sifindex;
494 int err;
495
496 switch (type) {
497 default:
498 case ICMP_PARAMETERPROB:
499 return 0;
500
501 case ICMP_DEST_UNREACH:
502 switch (code) {
503 case ICMP_SR_FAILED:
504 /* Impossible event. */
505 return 0;
506 default:
507 /* All others are translated to HOST_UNREACH.
508 rfc2003 contains "deep thoughts" about NET_UNREACH,
509 I believe they are just ether pollution. --ANK
510 */
511 break;
512 }
513 break;
514 case ICMP_TIME_EXCEEDED:
515 if (code != ICMP_EXC_TTL)
516 return 0;
517 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
518 break;
519 case ICMP_REDIRECT:
520 break;
521 }
522
523 err = -ENOENT;
524
525 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
526 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
527 iph->daddr, iph->saddr, sifindex);
528 if (!t)
529 goto out;
530
531 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
532 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
533 t->parms.link, iph->protocol);
534 err = 0;
535 goto out;
536 }
537 if (type == ICMP_REDIRECT) {
538 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
539 iph->protocol);
540 err = 0;
541 goto out;
542 }
543
544 err = 0;
545 if (__in6_dev_get(skb->dev) &&
546 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
547 goto out;
548
549 if (t->parms.iph.daddr == 0)
550 goto out;
551
552 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
553 goto out;
554
555 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
556 t->err_count++;
557 else
558 t->err_count = 1;
559 t->err_time = jiffies;
560out:
561 return err;
562}
563
564static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
565 const struct in6_addr *v6addr)
566{
567 __be32 v4embed = 0;
568 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
569 return true;
570 return false;
571}
572
573/* Checks if an address matches an address on the tunnel interface.
574 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
575 * Long story:
576 * This function is called after we considered the packet as spoofed
577 * in is_spoofed_6rd.
578 * We may have a router that is doing NAT for proto 41 packets
579 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
580 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
581 * function will return true, dropping the packet.
582 * But, we can still check if is spoofed against the IP
583 * addresses associated with the interface.
584 */
585static bool only_dnatted(const struct ip_tunnel *tunnel,
586 const struct in6_addr *v6dst)
587{
588 int prefix_len;
589
590#ifdef CONFIG_IPV6_SIT_6RD
591 prefix_len = tunnel->ip6rd.prefixlen + 32
592 - tunnel->ip6rd.relay_prefixlen;
593#else
594 prefix_len = 48;
595#endif
596 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
597}
598
599/* Returns true if a packet is spoofed */
600static bool packet_is_spoofed(struct sk_buff *skb,
601 const struct iphdr *iph,
602 struct ip_tunnel *tunnel)
603{
604 const struct ipv6hdr *ipv6h;
605
606 if (tunnel->dev->priv_flags & IFF_ISATAP) {
607 if (!isatap_chksrc(skb, iph, tunnel))
608 return true;
609
610 return false;
611 }
612
613 if (tunnel->dev->flags & IFF_POINTOPOINT)
614 return false;
615
616 ipv6h = ipv6_hdr(skb);
617
618 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
619 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
620 &iph->saddr, &ipv6h->saddr,
621 &iph->daddr, &ipv6h->daddr);
622 return true;
623 }
624
625 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
626 return false;
627
628 if (only_dnatted(tunnel, &ipv6h->daddr))
629 return false;
630
631 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
632 &iph->saddr, &ipv6h->saddr,
633 &iph->daddr, &ipv6h->daddr);
634 return true;
635}
636
637static int ipip6_rcv(struct sk_buff *skb)
638{
639 const struct iphdr *iph = ip_hdr(skb);
640 struct ip_tunnel *tunnel;
641 int sifindex;
642 int err;
643
644 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
645 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
646 iph->saddr, iph->daddr, sifindex);
647 if (tunnel) {
648 struct pcpu_sw_netstats *tstats;
649
650 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
651 tunnel->parms.iph.protocol != 0)
652 goto out;
653
654 skb->mac_header = skb->network_header;
655 skb_reset_network_header(skb);
656 IPCB(skb)->flags = 0;
657 skb->dev = tunnel->dev;
658
659 if (packet_is_spoofed(skb, iph, tunnel)) {
660 tunnel->dev->stats.rx_errors++;
661 goto out;
662 }
663
664 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
665 !net_eq(tunnel->net, dev_net(tunnel->dev))))
666 goto out;
667
668 /* skb can be uncloned in iptunnel_pull_header, so
669 * old iph is no longer valid
670 */
671 iph = (const struct iphdr *)skb_mac_header(skb);
672 err = IP_ECN_decapsulate(iph, skb);
673 if (unlikely(err)) {
674 if (log_ecn_error)
675 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
676 &iph->saddr, iph->tos);
677 if (err > 1) {
678 ++tunnel->dev->stats.rx_frame_errors;
679 ++tunnel->dev->stats.rx_errors;
680 goto out;
681 }
682 }
683
684 tstats = this_cpu_ptr(tunnel->dev->tstats);
685 u64_stats_update_begin(&tstats->syncp);
686 tstats->rx_packets++;
687 tstats->rx_bytes += skb->len;
688 u64_stats_update_end(&tstats->syncp);
689
690 netif_rx(skb);
691
692 return 0;
693 }
694
695 /* no tunnel matched, let upstream know, ipsec may handle it */
696 return 1;
697out:
698 kfree_skb(skb);
699 return 0;
700}
701
702static const struct tnl_ptk_info ipip_tpi = {
703 /* no tunnel info required for ipip. */
704 .proto = htons(ETH_P_IP),
705};
706
707#if IS_ENABLED(CONFIG_MPLS)
708static const struct tnl_ptk_info mplsip_tpi = {
709 /* no tunnel info required for mplsip. */
710 .proto = htons(ETH_P_MPLS_UC),
711};
712#endif
713
714static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
715{
716 const struct iphdr *iph;
717 struct ip_tunnel *tunnel;
718 int sifindex;
719
720 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
721
722 iph = ip_hdr(skb);
723 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
724 iph->saddr, iph->daddr, sifindex);
725 if (tunnel) {
726 const struct tnl_ptk_info *tpi;
727
728 if (tunnel->parms.iph.protocol != ipproto &&
729 tunnel->parms.iph.protocol != 0)
730 goto drop;
731
732 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
733 goto drop;
734#if IS_ENABLED(CONFIG_MPLS)
735 if (ipproto == IPPROTO_MPLS)
736 tpi = &mplsip_tpi;
737 else
738#endif
739 tpi = &ipip_tpi;
740 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
741 goto drop;
742 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
743 }
744
745 return 1;
746
747drop:
748 kfree_skb(skb);
749 return 0;
750}
751
752static int ipip_rcv(struct sk_buff *skb)
753{
754 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
755}
756
757#if IS_ENABLED(CONFIG_MPLS)
758static int mplsip_rcv(struct sk_buff *skb)
759{
760 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
761}
762#endif
763
764/*
765 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
766 * stores the embedded IPv4 address in v4dst and returns true.
767 */
768static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
769 __be32 *v4dst)
770{
771#ifdef CONFIG_IPV6_SIT_6RD
772 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
773 tunnel->ip6rd.prefixlen)) {
774 unsigned int pbw0, pbi0;
775 int pbi1;
776 u32 d;
777
778 pbw0 = tunnel->ip6rd.prefixlen >> 5;
779 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
780
781 d = tunnel->ip6rd.relay_prefixlen < 32 ?
782 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
783 tunnel->ip6rd.relay_prefixlen : 0;
784
785 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
786 if (pbi1 > 0)
787 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
788 (32 - pbi1);
789
790 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
791 return true;
792 }
793#else
794 if (v6dst->s6_addr16[0] == htons(0x2002)) {
795 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
796 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
797 return true;
798 }
799#endif
800 return false;
801}
802
803static inline __be32 try_6rd(struct ip_tunnel *tunnel,
804 const struct in6_addr *v6dst)
805{
806 __be32 dst = 0;
807 check_6rd(tunnel, v6dst, &dst);
808 return dst;
809}
810
811/*
812 * This function assumes it is being called from dev_queue_xmit()
813 * and that skb is filled properly by that function.
814 */
815
816static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
817 struct net_device *dev)
818{
819 struct ip_tunnel *tunnel = netdev_priv(dev);
820 const struct iphdr *tiph = &tunnel->parms.iph;
821 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
822 u8 tos = tunnel->parms.iph.tos;
823 __be16 df = tiph->frag_off;
824 struct rtable *rt; /* Route to the other host */
825 struct net_device *tdev; /* Device to other host */
826 unsigned int max_headroom; /* The extra header space needed */
827 __be32 dst = tiph->daddr;
828 struct flowi4 fl4;
829 int mtu;
830 const struct in6_addr *addr6;
831 int addr_type;
832 u8 ttl;
833 u8 protocol = IPPROTO_IPV6;
834 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
835
836 if (tos == 1)
837 tos = ipv6_get_dsfield(iph6);
838
839 /* ISATAP (RFC4214) - must come before 6to4 */
840 if (dev->priv_flags & IFF_ISATAP) {
841 struct neighbour *neigh = NULL;
842 bool do_tx_error = false;
843
844 if (skb_dst(skb))
845 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
846
847 if (!neigh) {
848 net_dbg_ratelimited("nexthop == NULL\n");
849 goto tx_error;
850 }
851
852 addr6 = (const struct in6_addr *)&neigh->primary_key;
853 addr_type = ipv6_addr_type(addr6);
854
855 if ((addr_type & IPV6_ADDR_UNICAST) &&
856 ipv6_addr_is_isatap(addr6))
857 dst = addr6->s6_addr32[3];
858 else
859 do_tx_error = true;
860
861 neigh_release(neigh);
862 if (do_tx_error)
863 goto tx_error;
864 }
865
866 if (!dst)
867 dst = try_6rd(tunnel, &iph6->daddr);
868
869 if (!dst) {
870 struct neighbour *neigh = NULL;
871 bool do_tx_error = false;
872
873 if (skb_dst(skb))
874 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
875
876 if (!neigh) {
877 net_dbg_ratelimited("nexthop == NULL\n");
878 goto tx_error;
879 }
880
881 addr6 = (const struct in6_addr *)&neigh->primary_key;
882 addr_type = ipv6_addr_type(addr6);
883
884 if (addr_type == IPV6_ADDR_ANY) {
885 addr6 = &ipv6_hdr(skb)->daddr;
886 addr_type = ipv6_addr_type(addr6);
887 }
888
889 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
890 dst = addr6->s6_addr32[3];
891 else
892 do_tx_error = true;
893
894 neigh_release(neigh);
895 if (do_tx_error)
896 goto tx_error;
897 }
898
899 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
900 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
901 0, dst, tiph->saddr, 0, 0,
902 sock_net_uid(tunnel->net, NULL));
903
904 rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
905 if (!rt) {
906 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
907 if (IS_ERR(rt)) {
908 dev->stats.tx_carrier_errors++;
909 goto tx_error_icmp;
910 }
911 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
912 }
913
914 if (rt->rt_type != RTN_UNICAST) {
915 ip_rt_put(rt);
916 dev->stats.tx_carrier_errors++;
917 goto tx_error_icmp;
918 }
919 tdev = rt->dst.dev;
920
921 if (tdev == dev) {
922 ip_rt_put(rt);
923 dev->stats.collisions++;
924 goto tx_error;
925 }
926
927 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
928 ip_rt_put(rt);
929 goto tx_error;
930 }
931
932 if (df) {
933 mtu = dst_mtu(&rt->dst) - t_hlen;
934
935 if (mtu < 68) {
936 dev->stats.collisions++;
937 ip_rt_put(rt);
938 goto tx_error;
939 }
940
941 if (mtu < IPV6_MIN_MTU) {
942 mtu = IPV6_MIN_MTU;
943 df = 0;
944 }
945
946 if (tunnel->parms.iph.daddr)
947 skb_dst_update_pmtu(skb, mtu);
948
949 if (skb->len > mtu && !skb_is_gso(skb)) {
950 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
951 ip_rt_put(rt);
952 goto tx_error;
953 }
954 }
955
956 if (tunnel->err_count > 0) {
957 if (time_before(jiffies,
958 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
959 tunnel->err_count--;
960 dst_link_failure(skb);
961 } else
962 tunnel->err_count = 0;
963 }
964
965 /*
966 * Okay, now see if we can stuff it in the buffer as-is.
967 */
968 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
969
970 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
971 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
972 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
973 if (!new_skb) {
974 ip_rt_put(rt);
975 dev->stats.tx_dropped++;
976 kfree_skb(skb);
977 return NETDEV_TX_OK;
978 }
979 if (skb->sk)
980 skb_set_owner_w(new_skb, skb->sk);
981 dev_kfree_skb(skb);
982 skb = new_skb;
983 iph6 = ipv6_hdr(skb);
984 }
985 ttl = tiph->ttl;
986 if (ttl == 0)
987 ttl = iph6->hop_limit;
988 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
989
990 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
991 ip_rt_put(rt);
992 goto tx_error;
993 }
994
995 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
996
997 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
998 df, !net_eq(tunnel->net, dev_net(dev)));
999 return NETDEV_TX_OK;
1000
1001tx_error_icmp:
1002 dst_link_failure(skb);
1003tx_error:
1004 kfree_skb(skb);
1005 dev->stats.tx_errors++;
1006 return NETDEV_TX_OK;
1007}
1008
1009static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1010 struct net_device *dev, u8 ipproto)
1011{
1012 struct ip_tunnel *tunnel = netdev_priv(dev);
1013 const struct iphdr *tiph = &tunnel->parms.iph;
1014
1015 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1016 goto tx_error;
1017
1018 skb_set_inner_ipproto(skb, ipproto);
1019
1020 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1021 return NETDEV_TX_OK;
1022tx_error:
1023 kfree_skb(skb);
1024 dev->stats.tx_errors++;
1025 return NETDEV_TX_OK;
1026}
1027
1028static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1029 struct net_device *dev)
1030{
1031 if (!pskb_inet_may_pull(skb))
1032 goto tx_err;
1033
1034 switch (skb->protocol) {
1035 case htons(ETH_P_IP):
1036 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1037 break;
1038 case htons(ETH_P_IPV6):
1039 ipip6_tunnel_xmit(skb, dev);
1040 break;
1041#if IS_ENABLED(CONFIG_MPLS)
1042 case htons(ETH_P_MPLS_UC):
1043 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1044 break;
1045#endif
1046 default:
1047 goto tx_err;
1048 }
1049
1050 return NETDEV_TX_OK;
1051
1052tx_err:
1053 dev->stats.tx_errors++;
1054 kfree_skb(skb);
1055 return NETDEV_TX_OK;
1056
1057}
1058
1059static void ipip6_tunnel_bind_dev(struct net_device *dev)
1060{
1061 struct net_device *tdev = NULL;
1062 struct ip_tunnel *tunnel;
1063 const struct iphdr *iph;
1064 struct flowi4 fl4;
1065
1066 tunnel = netdev_priv(dev);
1067 iph = &tunnel->parms.iph;
1068
1069 if (iph->daddr) {
1070 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1071 NULL,
1072 iph->daddr, iph->saddr,
1073 0, 0,
1074 IPPROTO_IPV6,
1075 RT_TOS(iph->tos),
1076 tunnel->parms.link);
1077
1078 if (!IS_ERR(rt)) {
1079 tdev = rt->dst.dev;
1080 ip_rt_put(rt);
1081 }
1082 dev->flags |= IFF_POINTOPOINT;
1083 }
1084
1085 if (!tdev && tunnel->parms.link)
1086 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1087
1088 if (tdev && !netif_is_l3_master(tdev)) {
1089 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1090
1091 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1092 dev->mtu = tdev->mtu - t_hlen;
1093 if (dev->mtu < IPV6_MIN_MTU)
1094 dev->mtu = IPV6_MIN_MTU;
1095 }
1096}
1097
1098static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1099 __u32 fwmark)
1100{
1101 struct net *net = t->net;
1102 struct sit_net *sitn = net_generic(net, sit_net_id);
1103
1104 ipip6_tunnel_unlink(sitn, t);
1105 synchronize_net();
1106 t->parms.iph.saddr = p->iph.saddr;
1107 t->parms.iph.daddr = p->iph.daddr;
1108 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1109 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1110 ipip6_tunnel_link(sitn, t);
1111 t->parms.iph.ttl = p->iph.ttl;
1112 t->parms.iph.tos = p->iph.tos;
1113 t->parms.iph.frag_off = p->iph.frag_off;
1114 if (t->parms.link != p->link || t->fwmark != fwmark) {
1115 t->parms.link = p->link;
1116 t->fwmark = fwmark;
1117 ipip6_tunnel_bind_dev(t->dev);
1118 }
1119 dst_cache_reset(&t->dst_cache);
1120 netdev_state_change(t->dev);
1121}
1122
1123#ifdef CONFIG_IPV6_SIT_6RD
1124static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1125 struct ip_tunnel_6rd *ip6rd)
1126{
1127 struct in6_addr prefix;
1128 __be32 relay_prefix;
1129
1130 if (ip6rd->relay_prefixlen > 32 ||
1131 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1132 return -EINVAL;
1133
1134 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1135 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1136 return -EINVAL;
1137 if (ip6rd->relay_prefixlen)
1138 relay_prefix = ip6rd->relay_prefix &
1139 htonl(0xffffffffUL <<
1140 (32 - ip6rd->relay_prefixlen));
1141 else
1142 relay_prefix = 0;
1143 if (relay_prefix != ip6rd->relay_prefix)
1144 return -EINVAL;
1145
1146 t->ip6rd.prefix = prefix;
1147 t->ip6rd.relay_prefix = relay_prefix;
1148 t->ip6rd.prefixlen = ip6rd->prefixlen;
1149 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1150 dst_cache_reset(&t->dst_cache);
1151 netdev_state_change(t->dev);
1152 return 0;
1153}
1154#endif
1155
1156static bool ipip6_valid_ip_proto(u8 ipproto)
1157{
1158 return ipproto == IPPROTO_IPV6 ||
1159 ipproto == IPPROTO_IPIP ||
1160#if IS_ENABLED(CONFIG_MPLS)
1161 ipproto == IPPROTO_MPLS ||
1162#endif
1163 ipproto == 0;
1164}
1165
1166static int
1167ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1168{
1169 int err = 0;
1170 struct ip_tunnel_parm p;
1171 struct ip_tunnel_prl prl;
1172 struct ip_tunnel *t = netdev_priv(dev);
1173 struct net *net = t->net;
1174 struct sit_net *sitn = net_generic(net, sit_net_id);
1175#ifdef CONFIG_IPV6_SIT_6RD
1176 struct ip_tunnel_6rd ip6rd;
1177#endif
1178
1179 switch (cmd) {
1180 case SIOCGETTUNNEL:
1181#ifdef CONFIG_IPV6_SIT_6RD
1182 case SIOCGET6RD:
1183#endif
1184 if (dev == sitn->fb_tunnel_dev) {
1185 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1186 err = -EFAULT;
1187 break;
1188 }
1189 t = ipip6_tunnel_locate(net, &p, 0);
1190 if (!t)
1191 t = netdev_priv(dev);
1192 }
1193
1194 err = -EFAULT;
1195 if (cmd == SIOCGETTUNNEL) {
1196 memcpy(&p, &t->parms, sizeof(p));
1197 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1198 sizeof(p)))
1199 goto done;
1200#ifdef CONFIG_IPV6_SIT_6RD
1201 } else {
1202 ip6rd.prefix = t->ip6rd.prefix;
1203 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1204 ip6rd.prefixlen = t->ip6rd.prefixlen;
1205 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1206 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1207 sizeof(ip6rd)))
1208 goto done;
1209#endif
1210 }
1211 err = 0;
1212 break;
1213
1214 case SIOCADDTUNNEL:
1215 case SIOCCHGTUNNEL:
1216 err = -EPERM;
1217 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1218 goto done;
1219
1220 err = -EFAULT;
1221 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1222 goto done;
1223
1224 err = -EINVAL;
1225 if (!ipip6_valid_ip_proto(p.iph.protocol))
1226 goto done;
1227 if (p.iph.version != 4 ||
1228 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1229 goto done;
1230 if (p.iph.ttl)
1231 p.iph.frag_off |= htons(IP_DF);
1232
1233 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1234
1235 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1236 if (t) {
1237 if (t->dev != dev) {
1238 err = -EEXIST;
1239 break;
1240 }
1241 } else {
1242 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1243 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1244 err = -EINVAL;
1245 break;
1246 }
1247 t = netdev_priv(dev);
1248 }
1249
1250 ipip6_tunnel_update(t, &p, t->fwmark);
1251 }
1252
1253 if (t) {
1254 err = 0;
1255 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1256 err = -EFAULT;
1257 } else
1258 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1259 break;
1260
1261 case SIOCDELTUNNEL:
1262 err = -EPERM;
1263 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1264 goto done;
1265
1266 if (dev == sitn->fb_tunnel_dev) {
1267 err = -EFAULT;
1268 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1269 goto done;
1270 err = -ENOENT;
1271 t = ipip6_tunnel_locate(net, &p, 0);
1272 if (!t)
1273 goto done;
1274 err = -EPERM;
1275 if (t == netdev_priv(sitn->fb_tunnel_dev))
1276 goto done;
1277 dev = t->dev;
1278 }
1279 unregister_netdevice(dev);
1280 err = 0;
1281 break;
1282
1283 case SIOCGETPRL:
1284 err = -EINVAL;
1285 if (dev == sitn->fb_tunnel_dev)
1286 goto done;
1287 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1288 break;
1289
1290 case SIOCADDPRL:
1291 case SIOCDELPRL:
1292 case SIOCCHGPRL:
1293 err = -EPERM;
1294 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1295 goto done;
1296 err = -EINVAL;
1297 if (dev == sitn->fb_tunnel_dev)
1298 goto done;
1299 err = -EFAULT;
1300 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1301 goto done;
1302
1303 switch (cmd) {
1304 case SIOCDELPRL:
1305 err = ipip6_tunnel_del_prl(t, &prl);
1306 break;
1307 case SIOCADDPRL:
1308 case SIOCCHGPRL:
1309 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1310 break;
1311 }
1312 dst_cache_reset(&t->dst_cache);
1313 netdev_state_change(dev);
1314 break;
1315
1316#ifdef CONFIG_IPV6_SIT_6RD
1317 case SIOCADD6RD:
1318 case SIOCCHG6RD:
1319 case SIOCDEL6RD:
1320 err = -EPERM;
1321 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1322 goto done;
1323
1324 err = -EFAULT;
1325 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1326 sizeof(ip6rd)))
1327 goto done;
1328
1329 if (cmd != SIOCDEL6RD) {
1330 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1331 if (err < 0)
1332 goto done;
1333 } else
1334 ipip6_tunnel_clone_6rd(dev, sitn);
1335
1336 err = 0;
1337 break;
1338#endif
1339
1340 default:
1341 err = -EINVAL;
1342 }
1343
1344done:
1345 return err;
1346}
1347
1348static const struct net_device_ops ipip6_netdev_ops = {
1349 .ndo_init = ipip6_tunnel_init,
1350 .ndo_uninit = ipip6_tunnel_uninit,
1351 .ndo_start_xmit = sit_tunnel_xmit,
1352 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1353 .ndo_get_stats64 = ip_tunnel_get_stats64,
1354 .ndo_get_iflink = ip_tunnel_get_iflink,
1355};
1356
1357static void ipip6_dev_free(struct net_device *dev)
1358{
1359 struct ip_tunnel *tunnel = netdev_priv(dev);
1360
1361 dst_cache_destroy(&tunnel->dst_cache);
1362 free_percpu(dev->tstats);
1363}
1364
1365#define SIT_FEATURES (NETIF_F_SG | \
1366 NETIF_F_FRAGLIST | \
1367 NETIF_F_HIGHDMA | \
1368 NETIF_F_GSO_SOFTWARE | \
1369 NETIF_F_HW_CSUM)
1370
1371static void ipip6_tunnel_setup(struct net_device *dev)
1372{
1373 struct ip_tunnel *tunnel = netdev_priv(dev);
1374 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1375
1376 dev->netdev_ops = &ipip6_netdev_ops;
1377 dev->needs_free_netdev = true;
1378 dev->priv_destructor = ipip6_dev_free;
1379
1380 dev->type = ARPHRD_SIT;
1381 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1382 dev->mtu = ETH_DATA_LEN - t_hlen;
1383 dev->min_mtu = IPV6_MIN_MTU;
1384 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1385 dev->flags = IFF_NOARP;
1386 netif_keep_dst(dev);
1387 dev->addr_len = 4;
1388 dev->features |= NETIF_F_LLTX;
1389 dev->features |= SIT_FEATURES;
1390 dev->hw_features |= SIT_FEATURES;
1391}
1392
1393static int ipip6_tunnel_init(struct net_device *dev)
1394{
1395 struct ip_tunnel *tunnel = netdev_priv(dev);
1396 int err;
1397
1398 tunnel->dev = dev;
1399 tunnel->net = dev_net(dev);
1400 strcpy(tunnel->parms.name, dev->name);
1401
1402 ipip6_tunnel_bind_dev(dev);
1403 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1404 if (!dev->tstats)
1405 return -ENOMEM;
1406
1407 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1408 if (err) {
1409 free_percpu(dev->tstats);
1410 dev->tstats = NULL;
1411 return err;
1412 }
1413
1414 return 0;
1415}
1416
1417static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1418{
1419 struct ip_tunnel *tunnel = netdev_priv(dev);
1420 struct iphdr *iph = &tunnel->parms.iph;
1421 struct net *net = dev_net(dev);
1422 struct sit_net *sitn = net_generic(net, sit_net_id);
1423
1424 iph->version = 4;
1425 iph->protocol = IPPROTO_IPV6;
1426 iph->ihl = 5;
1427 iph->ttl = 64;
1428
1429 dev_hold(dev);
1430 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1431}
1432
1433static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1434 struct netlink_ext_ack *extack)
1435{
1436 u8 proto;
1437
1438 if (!data || !data[IFLA_IPTUN_PROTO])
1439 return 0;
1440
1441 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1442 if (!ipip6_valid_ip_proto(proto))
1443 return -EINVAL;
1444
1445 return 0;
1446}
1447
1448static void ipip6_netlink_parms(struct nlattr *data[],
1449 struct ip_tunnel_parm *parms,
1450 __u32 *fwmark)
1451{
1452 memset(parms, 0, sizeof(*parms));
1453
1454 parms->iph.version = 4;
1455 parms->iph.protocol = IPPROTO_IPV6;
1456 parms->iph.ihl = 5;
1457 parms->iph.ttl = 64;
1458
1459 if (!data)
1460 return;
1461
1462 if (data[IFLA_IPTUN_LINK])
1463 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1464
1465 if (data[IFLA_IPTUN_LOCAL])
1466 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1467
1468 if (data[IFLA_IPTUN_REMOTE])
1469 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1470
1471 if (data[IFLA_IPTUN_TTL]) {
1472 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1473 if (parms->iph.ttl)
1474 parms->iph.frag_off = htons(IP_DF);
1475 }
1476
1477 if (data[IFLA_IPTUN_TOS])
1478 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1479
1480 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1481 parms->iph.frag_off = htons(IP_DF);
1482
1483 if (data[IFLA_IPTUN_FLAGS])
1484 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1485
1486 if (data[IFLA_IPTUN_PROTO])
1487 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1488
1489 if (data[IFLA_IPTUN_FWMARK])
1490 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1491}
1492
1493/* This function returns true when ENCAP attributes are present in the nl msg */
1494static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1495 struct ip_tunnel_encap *ipencap)
1496{
1497 bool ret = false;
1498
1499 memset(ipencap, 0, sizeof(*ipencap));
1500
1501 if (!data)
1502 return ret;
1503
1504 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1505 ret = true;
1506 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1507 }
1508
1509 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1510 ret = true;
1511 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1512 }
1513
1514 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1515 ret = true;
1516 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1517 }
1518
1519 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1520 ret = true;
1521 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1522 }
1523
1524 return ret;
1525}
1526
1527#ifdef CONFIG_IPV6_SIT_6RD
1528/* This function returns true when 6RD attributes are present in the nl msg */
1529static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1530 struct ip_tunnel_6rd *ip6rd)
1531{
1532 bool ret = false;
1533 memset(ip6rd, 0, sizeof(*ip6rd));
1534
1535 if (!data)
1536 return ret;
1537
1538 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1539 ret = true;
1540 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1541 }
1542
1543 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1544 ret = true;
1545 ip6rd->relay_prefix =
1546 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1547 }
1548
1549 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1550 ret = true;
1551 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1552 }
1553
1554 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1555 ret = true;
1556 ip6rd->relay_prefixlen =
1557 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1558 }
1559
1560 return ret;
1561}
1562#endif
1563
1564static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1565 struct nlattr *tb[], struct nlattr *data[],
1566 struct netlink_ext_ack *extack)
1567{
1568 struct net *net = dev_net(dev);
1569 struct ip_tunnel *nt;
1570 struct ip_tunnel_encap ipencap;
1571#ifdef CONFIG_IPV6_SIT_6RD
1572 struct ip_tunnel_6rd ip6rd;
1573#endif
1574 int err;
1575
1576 nt = netdev_priv(dev);
1577
1578 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1579 err = ip_tunnel_encap_setup(nt, &ipencap);
1580 if (err < 0)
1581 return err;
1582 }
1583
1584 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1585
1586 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1587 return -EEXIST;
1588
1589 err = ipip6_tunnel_create(dev);
1590 if (err < 0)
1591 return err;
1592
1593 if (tb[IFLA_MTU]) {
1594 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1595
1596 if (mtu >= IPV6_MIN_MTU &&
1597 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1598 dev->mtu = mtu;
1599 }
1600
1601#ifdef CONFIG_IPV6_SIT_6RD
1602 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1603 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1604#endif
1605
1606 return err;
1607}
1608
1609static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1610 struct nlattr *data[],
1611 struct netlink_ext_ack *extack)
1612{
1613 struct ip_tunnel *t = netdev_priv(dev);
1614 struct ip_tunnel_parm p;
1615 struct ip_tunnel_encap ipencap;
1616 struct net *net = t->net;
1617 struct sit_net *sitn = net_generic(net, sit_net_id);
1618#ifdef CONFIG_IPV6_SIT_6RD
1619 struct ip_tunnel_6rd ip6rd;
1620#endif
1621 __u32 fwmark = t->fwmark;
1622 int err;
1623
1624 if (dev == sitn->fb_tunnel_dev)
1625 return -EINVAL;
1626
1627 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1628 err = ip_tunnel_encap_setup(t, &ipencap);
1629 if (err < 0)
1630 return err;
1631 }
1632
1633 ipip6_netlink_parms(data, &p, &fwmark);
1634
1635 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1636 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1637 return -EINVAL;
1638
1639 t = ipip6_tunnel_locate(net, &p, 0);
1640
1641 if (t) {
1642 if (t->dev != dev)
1643 return -EEXIST;
1644 } else
1645 t = netdev_priv(dev);
1646
1647 ipip6_tunnel_update(t, &p, fwmark);
1648
1649#ifdef CONFIG_IPV6_SIT_6RD
1650 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1651 return ipip6_tunnel_update_6rd(t, &ip6rd);
1652#endif
1653
1654 return 0;
1655}
1656
1657static size_t ipip6_get_size(const struct net_device *dev)
1658{
1659 return
1660 /* IFLA_IPTUN_LINK */
1661 nla_total_size(4) +
1662 /* IFLA_IPTUN_LOCAL */
1663 nla_total_size(4) +
1664 /* IFLA_IPTUN_REMOTE */
1665 nla_total_size(4) +
1666 /* IFLA_IPTUN_TTL */
1667 nla_total_size(1) +
1668 /* IFLA_IPTUN_TOS */
1669 nla_total_size(1) +
1670 /* IFLA_IPTUN_PMTUDISC */
1671 nla_total_size(1) +
1672 /* IFLA_IPTUN_FLAGS */
1673 nla_total_size(2) +
1674 /* IFLA_IPTUN_PROTO */
1675 nla_total_size(1) +
1676#ifdef CONFIG_IPV6_SIT_6RD
1677 /* IFLA_IPTUN_6RD_PREFIX */
1678 nla_total_size(sizeof(struct in6_addr)) +
1679 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1680 nla_total_size(4) +
1681 /* IFLA_IPTUN_6RD_PREFIXLEN */
1682 nla_total_size(2) +
1683 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1684 nla_total_size(2) +
1685#endif
1686 /* IFLA_IPTUN_ENCAP_TYPE */
1687 nla_total_size(2) +
1688 /* IFLA_IPTUN_ENCAP_FLAGS */
1689 nla_total_size(2) +
1690 /* IFLA_IPTUN_ENCAP_SPORT */
1691 nla_total_size(2) +
1692 /* IFLA_IPTUN_ENCAP_DPORT */
1693 nla_total_size(2) +
1694 /* IFLA_IPTUN_FWMARK */
1695 nla_total_size(4) +
1696 0;
1697}
1698
1699static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1700{
1701 struct ip_tunnel *tunnel = netdev_priv(dev);
1702 struct ip_tunnel_parm *parm = &tunnel->parms;
1703
1704 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1705 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1706 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1707 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1708 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1709 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1710 !!(parm->iph.frag_off & htons(IP_DF))) ||
1711 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1712 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1713 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1714 goto nla_put_failure;
1715
1716#ifdef CONFIG_IPV6_SIT_6RD
1717 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1718 &tunnel->ip6rd.prefix) ||
1719 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1720 tunnel->ip6rd.relay_prefix) ||
1721 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1722 tunnel->ip6rd.prefixlen) ||
1723 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1724 tunnel->ip6rd.relay_prefixlen))
1725 goto nla_put_failure;
1726#endif
1727
1728 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1729 tunnel->encap.type) ||
1730 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1731 tunnel->encap.sport) ||
1732 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1733 tunnel->encap.dport) ||
1734 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1735 tunnel->encap.flags))
1736 goto nla_put_failure;
1737
1738 return 0;
1739
1740nla_put_failure:
1741 return -EMSGSIZE;
1742}
1743
1744static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1745 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1746 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1747 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1748 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1749 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1750 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1751 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1752 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1753#ifdef CONFIG_IPV6_SIT_6RD
1754 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1755 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1756 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1757 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1758#endif
1759 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1760 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1761 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1762 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1763 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1764};
1765
1766static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1767{
1768 struct net *net = dev_net(dev);
1769 struct sit_net *sitn = net_generic(net, sit_net_id);
1770
1771 if (dev != sitn->fb_tunnel_dev)
1772 unregister_netdevice_queue(dev, head);
1773}
1774
1775static struct rtnl_link_ops sit_link_ops __read_mostly = {
1776 .kind = "sit",
1777 .maxtype = IFLA_IPTUN_MAX,
1778 .policy = ipip6_policy,
1779 .priv_size = sizeof(struct ip_tunnel),
1780 .setup = ipip6_tunnel_setup,
1781 .validate = ipip6_validate,
1782 .newlink = ipip6_newlink,
1783 .changelink = ipip6_changelink,
1784 .get_size = ipip6_get_size,
1785 .fill_info = ipip6_fill_info,
1786 .dellink = ipip6_dellink,
1787 .get_link_net = ip_tunnel_get_link_net,
1788};
1789
1790static struct xfrm_tunnel sit_handler __read_mostly = {
1791 .handler = ipip6_rcv,
1792 .err_handler = ipip6_err,
1793 .priority = 1,
1794};
1795
1796static struct xfrm_tunnel ipip_handler __read_mostly = {
1797 .handler = ipip_rcv,
1798 .err_handler = ipip6_err,
1799 .priority = 2,
1800};
1801
1802#if IS_ENABLED(CONFIG_MPLS)
1803static struct xfrm_tunnel mplsip_handler __read_mostly = {
1804 .handler = mplsip_rcv,
1805 .err_handler = ipip6_err,
1806 .priority = 2,
1807};
1808#endif
1809
1810static void __net_exit sit_destroy_tunnels(struct net *net,
1811 struct list_head *head)
1812{
1813 struct sit_net *sitn = net_generic(net, sit_net_id);
1814 struct net_device *dev, *aux;
1815 int prio;
1816
1817 for_each_netdev_safe(net, dev, aux)
1818 if (dev->rtnl_link_ops == &sit_link_ops)
1819 unregister_netdevice_queue(dev, head);
1820
1821 for (prio = 1; prio < 4; prio++) {
1822 int h;
1823 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1824 struct ip_tunnel *t;
1825
1826 t = rtnl_dereference(sitn->tunnels[prio][h]);
1827 while (t) {
1828 /* If dev is in the same netns, it has already
1829 * been added to the list by the previous loop.
1830 */
1831 if (!net_eq(dev_net(t->dev), net))
1832 unregister_netdevice_queue(t->dev,
1833 head);
1834 t = rtnl_dereference(t->next);
1835 }
1836 }
1837 }
1838}
1839
1840static int __net_init sit_init_net(struct net *net)
1841{
1842 struct sit_net *sitn = net_generic(net, sit_net_id);
1843 struct ip_tunnel *t;
1844 int err;
1845
1846 sitn->tunnels[0] = sitn->tunnels_wc;
1847 sitn->tunnels[1] = sitn->tunnels_l;
1848 sitn->tunnels[2] = sitn->tunnels_r;
1849 sitn->tunnels[3] = sitn->tunnels_r_l;
1850
1851 if (!net_has_fallback_tunnels(net))
1852 return 0;
1853
1854 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1855 NET_NAME_UNKNOWN,
1856 ipip6_tunnel_setup);
1857 if (!sitn->fb_tunnel_dev) {
1858 err = -ENOMEM;
1859 goto err_alloc_dev;
1860 }
1861 dev_net_set(sitn->fb_tunnel_dev, net);
1862 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1863 /* FB netdevice is special: we have one, and only one per netns.
1864 * Allowing to move it to another netns is clearly unsafe.
1865 */
1866 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1867
1868 err = register_netdev(sitn->fb_tunnel_dev);
1869 if (err)
1870 goto err_reg_dev;
1871
1872 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1873 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1874
1875 t = netdev_priv(sitn->fb_tunnel_dev);
1876
1877 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1878 return 0;
1879
1880err_reg_dev:
1881 ipip6_dev_free(sitn->fb_tunnel_dev);
1882 free_netdev(sitn->fb_tunnel_dev);
1883err_alloc_dev:
1884 return err;
1885}
1886
1887static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1888{
1889 LIST_HEAD(list);
1890 struct net *net;
1891
1892 rtnl_lock();
1893 list_for_each_entry(net, net_list, exit_list)
1894 sit_destroy_tunnels(net, &list);
1895
1896 unregister_netdevice_many(&list);
1897 rtnl_unlock();
1898}
1899
1900static struct pernet_operations sit_net_ops = {
1901 .init = sit_init_net,
1902 .exit_batch = sit_exit_batch_net,
1903 .id = &sit_net_id,
1904 .size = sizeof(struct sit_net),
1905};
1906
1907static void __exit sit_cleanup(void)
1908{
1909 rtnl_link_unregister(&sit_link_ops);
1910 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1911 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1912#if IS_ENABLED(CONFIG_MPLS)
1913 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1914#endif
1915
1916 unregister_pernet_device(&sit_net_ops);
1917 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1918}
1919
1920static int __init sit_init(void)
1921{
1922 int err;
1923
1924 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1925
1926 err = register_pernet_device(&sit_net_ops);
1927 if (err < 0)
1928 return err;
1929 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1930 if (err < 0) {
1931 pr_info("%s: can't register ip6ip4\n", __func__);
1932 goto xfrm_tunnel_failed;
1933 }
1934 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1935 if (err < 0) {
1936 pr_info("%s: can't register ip4ip4\n", __func__);
1937 goto xfrm_tunnel4_failed;
1938 }
1939#if IS_ENABLED(CONFIG_MPLS)
1940 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1941 if (err < 0) {
1942 pr_info("%s: can't register mplsip\n", __func__);
1943 goto xfrm_tunnel_mpls_failed;
1944 }
1945#endif
1946 err = rtnl_link_register(&sit_link_ops);
1947 if (err < 0)
1948 goto rtnl_link_failed;
1949
1950out:
1951 return err;
1952
1953rtnl_link_failed:
1954#if IS_ENABLED(CONFIG_MPLS)
1955 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1956xfrm_tunnel_mpls_failed:
1957#endif
1958 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1959xfrm_tunnel4_failed:
1960 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1961xfrm_tunnel_failed:
1962 unregister_pernet_device(&sit_net_ops);
1963 goto out;
1964}
1965
1966module_init(sit_init);
1967module_exit(sit_cleanup);
1968MODULE_LICENSE("GPL");
1969MODULE_ALIAS_RTNL_LINK("sit");
1970MODULE_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#include <linux/module.h>
21#include <linux/capability.h>
22#include <linux/errno.h>
23#include <linux/types.h>
24#include <linux/socket.h>
25#include <linux/sockios.h>
26#include <linux/net.h>
27#include <linux/in6.h>
28#include <linux/netdevice.h>
29#include <linux/if_arp.h>
30#include <linux/icmp.h>
31#include <linux/slab.h>
32#include <asm/uaccess.h>
33#include <linux/init.h>
34#include <linux/netfilter_ipv4.h>
35#include <linux/if_ether.h>
36
37#include <net/sock.h>
38#include <net/snmp.h>
39
40#include <net/ipv6.h>
41#include <net/protocol.h>
42#include <net/transp_v6.h>
43#include <net/ip6_fib.h>
44#include <net/ip6_route.h>
45#include <net/ndisc.h>
46#include <net/addrconf.h>
47#include <net/ip.h>
48#include <net/udp.h>
49#include <net/icmp.h>
50#include <net/ipip.h>
51#include <net/inet_ecn.h>
52#include <net/xfrm.h>
53#include <net/dsfield.h>
54#include <net/net_namespace.h>
55#include <net/netns/generic.h>
56
57/*
58 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
59
60 For comments look at net/ipv4/ip_gre.c --ANK
61 */
62
63#define HASH_SIZE 16
64#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
65
66static int ipip6_tunnel_init(struct net_device *dev);
67static void ipip6_tunnel_setup(struct net_device *dev);
68static void ipip6_dev_free(struct net_device *dev);
69
70static int sit_net_id __read_mostly;
71struct sit_net {
72 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
73 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
74 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_wc[1];
76 struct ip_tunnel __rcu **tunnels[4];
77
78 struct net_device *fb_tunnel_dev;
79};
80
81/*
82 * Locking : hash tables are protected by RCU and RTNL
83 */
84
85#define for_each_ip_tunnel_rcu(start) \
86 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
87
88/* often modified stats are per cpu, other are shared (netdev->stats) */
89struct pcpu_tstats {
90 unsigned long rx_packets;
91 unsigned long rx_bytes;
92 unsigned long tx_packets;
93 unsigned long tx_bytes;
94};
95
96static struct net_device_stats *ipip6_get_stats(struct net_device *dev)
97{
98 struct pcpu_tstats sum = { 0 };
99 int i;
100
101 for_each_possible_cpu(i) {
102 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
103
104 sum.rx_packets += tstats->rx_packets;
105 sum.rx_bytes += tstats->rx_bytes;
106 sum.tx_packets += tstats->tx_packets;
107 sum.tx_bytes += tstats->tx_bytes;
108 }
109 dev->stats.rx_packets = sum.rx_packets;
110 dev->stats.rx_bytes = sum.rx_bytes;
111 dev->stats.tx_packets = sum.tx_packets;
112 dev->stats.tx_bytes = sum.tx_bytes;
113 return &dev->stats;
114}
115/*
116 * Must be invoked with rcu_read_lock
117 */
118static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
119 struct net_device *dev, __be32 remote, __be32 local)
120{
121 unsigned int h0 = HASH(remote);
122 unsigned int h1 = HASH(local);
123 struct ip_tunnel *t;
124 struct sit_net *sitn = net_generic(net, sit_net_id);
125
126 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
127 if (local == t->parms.iph.saddr &&
128 remote == t->parms.iph.daddr &&
129 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
130 (t->dev->flags & IFF_UP))
131 return t;
132 }
133 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
134 if (remote == t->parms.iph.daddr &&
135 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
136 (t->dev->flags & IFF_UP))
137 return t;
138 }
139 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
140 if (local == t->parms.iph.saddr &&
141 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
142 (t->dev->flags & IFF_UP))
143 return t;
144 }
145 t = rcu_dereference(sitn->tunnels_wc[0]);
146 if ((t != NULL) && (t->dev->flags & IFF_UP))
147 return t;
148 return NULL;
149}
150
151static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
152 struct ip_tunnel_parm *parms)
153{
154 __be32 remote = parms->iph.daddr;
155 __be32 local = parms->iph.saddr;
156 unsigned int h = 0;
157 int prio = 0;
158
159 if (remote) {
160 prio |= 2;
161 h ^= HASH(remote);
162 }
163 if (local) {
164 prio |= 1;
165 h ^= HASH(local);
166 }
167 return &sitn->tunnels[prio][h];
168}
169
170static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
171 struct ip_tunnel *t)
172{
173 return __ipip6_bucket(sitn, &t->parms);
174}
175
176static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
177{
178 struct ip_tunnel __rcu **tp;
179 struct ip_tunnel *iter;
180
181 for (tp = ipip6_bucket(sitn, t);
182 (iter = rtnl_dereference(*tp)) != NULL;
183 tp = &iter->next) {
184 if (t == iter) {
185 rcu_assign_pointer(*tp, t->next);
186 break;
187 }
188 }
189}
190
191static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
192{
193 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
194
195 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
196 rcu_assign_pointer(*tp, t);
197}
198
199static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
200{
201#ifdef CONFIG_IPV6_SIT_6RD
202 struct ip_tunnel *t = netdev_priv(dev);
203
204 if (t->dev == sitn->fb_tunnel_dev) {
205 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
206 t->ip6rd.relay_prefix = 0;
207 t->ip6rd.prefixlen = 16;
208 t->ip6rd.relay_prefixlen = 0;
209 } else {
210 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
211 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
212 }
213#endif
214}
215
216static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
217 struct ip_tunnel_parm *parms, int create)
218{
219 __be32 remote = parms->iph.daddr;
220 __be32 local = parms->iph.saddr;
221 struct ip_tunnel *t, *nt;
222 struct ip_tunnel __rcu **tp;
223 struct net_device *dev;
224 char name[IFNAMSIZ];
225 struct sit_net *sitn = net_generic(net, sit_net_id);
226
227 for (tp = __ipip6_bucket(sitn, parms);
228 (t = rtnl_dereference(*tp)) != NULL;
229 tp = &t->next) {
230 if (local == t->parms.iph.saddr &&
231 remote == t->parms.iph.daddr &&
232 parms->link == t->parms.link) {
233 if (create)
234 return NULL;
235 else
236 return t;
237 }
238 }
239 if (!create)
240 goto failed;
241
242 if (parms->name[0])
243 strlcpy(name, parms->name, IFNAMSIZ);
244 else
245 strcpy(name, "sit%d");
246
247 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
248 if (dev == NULL)
249 return NULL;
250
251 dev_net_set(dev, net);
252
253 nt = netdev_priv(dev);
254
255 nt->parms = *parms;
256 if (ipip6_tunnel_init(dev) < 0)
257 goto failed_free;
258 ipip6_tunnel_clone_6rd(dev, sitn);
259
260 if (parms->i_flags & SIT_ISATAP)
261 dev->priv_flags |= IFF_ISATAP;
262
263 if (register_netdevice(dev) < 0)
264 goto failed_free;
265
266 dev_hold(dev);
267
268 ipip6_tunnel_link(sitn, nt);
269 return nt;
270
271failed_free:
272 ipip6_dev_free(dev);
273failed:
274 return NULL;
275}
276
277#define for_each_prl_rcu(start) \
278 for (prl = rcu_dereference(start); \
279 prl; \
280 prl = rcu_dereference(prl->next))
281
282static struct ip_tunnel_prl_entry *
283__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
284{
285 struct ip_tunnel_prl_entry *prl;
286
287 for_each_prl_rcu(t->prl)
288 if (prl->addr == addr)
289 break;
290 return prl;
291
292}
293
294static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
295 struct ip_tunnel_prl __user *a)
296{
297 struct ip_tunnel_prl kprl, *kp;
298 struct ip_tunnel_prl_entry *prl;
299 unsigned int cmax, c = 0, ca, len;
300 int ret = 0;
301
302 if (copy_from_user(&kprl, a, sizeof(kprl)))
303 return -EFAULT;
304 cmax = kprl.datalen / sizeof(kprl);
305 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
306 cmax = 1;
307
308 /* For simple GET or for root users,
309 * we try harder to allocate.
310 */
311 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
312 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
313 NULL;
314
315 rcu_read_lock();
316
317 ca = t->prl_count < cmax ? t->prl_count : cmax;
318
319 if (!kp) {
320 /* We don't try hard to allocate much memory for
321 * non-root users.
322 * For root users, retry allocating enough memory for
323 * the answer.
324 */
325 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
326 if (!kp) {
327 ret = -ENOMEM;
328 goto out;
329 }
330 }
331
332 c = 0;
333 for_each_prl_rcu(t->prl) {
334 if (c >= cmax)
335 break;
336 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
337 continue;
338 kp[c].addr = prl->addr;
339 kp[c].flags = prl->flags;
340 c++;
341 if (kprl.addr != htonl(INADDR_ANY))
342 break;
343 }
344out:
345 rcu_read_unlock();
346
347 len = sizeof(*kp) * c;
348 ret = 0;
349 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
350 ret = -EFAULT;
351
352 kfree(kp);
353
354 return ret;
355}
356
357static int
358ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
359{
360 struct ip_tunnel_prl_entry *p;
361 int err = 0;
362
363 if (a->addr == htonl(INADDR_ANY))
364 return -EINVAL;
365
366 ASSERT_RTNL();
367
368 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
369 if (p->addr == a->addr) {
370 if (chg) {
371 p->flags = a->flags;
372 goto out;
373 }
374 err = -EEXIST;
375 goto out;
376 }
377 }
378
379 if (chg) {
380 err = -ENXIO;
381 goto out;
382 }
383
384 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
385 if (!p) {
386 err = -ENOBUFS;
387 goto out;
388 }
389
390 p->next = t->prl;
391 p->addr = a->addr;
392 p->flags = a->flags;
393 t->prl_count++;
394 rcu_assign_pointer(t->prl, p);
395out:
396 return err;
397}
398
399static void prl_list_destroy_rcu(struct rcu_head *head)
400{
401 struct ip_tunnel_prl_entry *p, *n;
402
403 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
404 do {
405 n = rcu_dereference_protected(p->next, 1);
406 kfree(p);
407 p = n;
408 } while (p);
409}
410
411static int
412ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
413{
414 struct ip_tunnel_prl_entry *x;
415 struct ip_tunnel_prl_entry __rcu **p;
416 int err = 0;
417
418 ASSERT_RTNL();
419
420 if (a && a->addr != htonl(INADDR_ANY)) {
421 for (p = &t->prl;
422 (x = rtnl_dereference(*p)) != NULL;
423 p = &x->next) {
424 if (x->addr == a->addr) {
425 *p = x->next;
426 kfree_rcu(x, rcu_head);
427 t->prl_count--;
428 goto out;
429 }
430 }
431 err = -ENXIO;
432 } else {
433 x = rtnl_dereference(t->prl);
434 if (x) {
435 t->prl_count = 0;
436 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
437 t->prl = NULL;
438 }
439 }
440out:
441 return err;
442}
443
444static int
445isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
446{
447 struct ip_tunnel_prl_entry *p;
448 int ok = 1;
449
450 rcu_read_lock();
451 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
452 if (p) {
453 if (p->flags & PRL_DEFAULT)
454 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
455 else
456 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
457 } else {
458 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
459
460 if (ipv6_addr_is_isatap(addr6) &&
461 (addr6->s6_addr32[3] == iph->saddr) &&
462 ipv6_chk_prefix(addr6, t->dev))
463 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
464 else
465 ok = 0;
466 }
467 rcu_read_unlock();
468 return ok;
469}
470
471static void ipip6_tunnel_uninit(struct net_device *dev)
472{
473 struct net *net = dev_net(dev);
474 struct sit_net *sitn = net_generic(net, sit_net_id);
475
476 if (dev == sitn->fb_tunnel_dev) {
477 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
478 } else {
479 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
480 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
481 }
482 dev_put(dev);
483}
484
485
486static int ipip6_err(struct sk_buff *skb, u32 info)
487{
488
489/* All the routers (except for Linux) return only
490 8 bytes of packet payload. It means, that precise relaying of
491 ICMP in the real Internet is absolutely infeasible.
492 */
493 const struct iphdr *iph = (const struct iphdr *)skb->data;
494 const int type = icmp_hdr(skb)->type;
495 const int code = icmp_hdr(skb)->code;
496 struct ip_tunnel *t;
497 int err;
498
499 switch (type) {
500 default:
501 case ICMP_PARAMETERPROB:
502 return 0;
503
504 case ICMP_DEST_UNREACH:
505 switch (code) {
506 case ICMP_SR_FAILED:
507 case ICMP_PORT_UNREACH:
508 /* Impossible event. */
509 return 0;
510 case ICMP_FRAG_NEEDED:
511 /* Soft state for pmtu is maintained by IP core. */
512 return 0;
513 default:
514 /* All others are translated to HOST_UNREACH.
515 rfc2003 contains "deep thoughts" about NET_UNREACH,
516 I believe they are just ether pollution. --ANK
517 */
518 break;
519 }
520 break;
521 case ICMP_TIME_EXCEEDED:
522 if (code != ICMP_EXC_TTL)
523 return 0;
524 break;
525 }
526
527 err = -ENOENT;
528
529 rcu_read_lock();
530 t = ipip6_tunnel_lookup(dev_net(skb->dev),
531 skb->dev,
532 iph->daddr,
533 iph->saddr);
534 if (t == NULL || t->parms.iph.daddr == 0)
535 goto out;
536
537 err = 0;
538 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
539 goto out;
540
541 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
542 t->err_count++;
543 else
544 t->err_count = 1;
545 t->err_time = jiffies;
546out:
547 rcu_read_unlock();
548 return err;
549}
550
551static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
552{
553 if (INET_ECN_is_ce(iph->tos))
554 IP6_ECN_set_ce(ipv6_hdr(skb));
555}
556
557static int ipip6_rcv(struct sk_buff *skb)
558{
559 const struct iphdr *iph;
560 struct ip_tunnel *tunnel;
561
562 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
563 goto out;
564
565 iph = ip_hdr(skb);
566
567 rcu_read_lock();
568 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
569 iph->saddr, iph->daddr);
570 if (tunnel != NULL) {
571 struct pcpu_tstats *tstats;
572
573 secpath_reset(skb);
574 skb->mac_header = skb->network_header;
575 skb_reset_network_header(skb);
576 IPCB(skb)->flags = 0;
577 skb->protocol = htons(ETH_P_IPV6);
578 skb->pkt_type = PACKET_HOST;
579
580 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
581 !isatap_chksrc(skb, iph, tunnel)) {
582 tunnel->dev->stats.rx_errors++;
583 rcu_read_unlock();
584 kfree_skb(skb);
585 return 0;
586 }
587
588 tstats = this_cpu_ptr(tunnel->dev->tstats);
589 tstats->rx_packets++;
590 tstats->rx_bytes += skb->len;
591
592 __skb_tunnel_rx(skb, tunnel->dev);
593
594 ipip6_ecn_decapsulate(iph, skb);
595
596 netif_rx(skb);
597
598 rcu_read_unlock();
599 return 0;
600 }
601
602 /* no tunnel matched, let upstream know, ipsec may handle it */
603 rcu_read_unlock();
604 return 1;
605out:
606 kfree_skb(skb);
607 return 0;
608}
609
610/*
611 * Returns the embedded IPv4 address if the IPv6 address
612 * comes from 6rd / 6to4 (RFC 3056) addr space.
613 */
614static inline
615__be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
616{
617 __be32 dst = 0;
618
619#ifdef CONFIG_IPV6_SIT_6RD
620 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
621 tunnel->ip6rd.prefixlen)) {
622 unsigned int pbw0, pbi0;
623 int pbi1;
624 u32 d;
625
626 pbw0 = tunnel->ip6rd.prefixlen >> 5;
627 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
628
629 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
630 tunnel->ip6rd.relay_prefixlen;
631
632 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
633 if (pbi1 > 0)
634 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
635 (32 - pbi1);
636
637 dst = tunnel->ip6rd.relay_prefix | htonl(d);
638 }
639#else
640 if (v6dst->s6_addr16[0] == htons(0x2002)) {
641 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
642 memcpy(&dst, &v6dst->s6_addr16[1], 4);
643 }
644#endif
645 return dst;
646}
647
648/*
649 * This function assumes it is being called from dev_queue_xmit()
650 * and that skb is filled properly by that function.
651 */
652
653static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
654 struct net_device *dev)
655{
656 struct ip_tunnel *tunnel = netdev_priv(dev);
657 struct pcpu_tstats *tstats;
658 const struct iphdr *tiph = &tunnel->parms.iph;
659 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
660 u8 tos = tunnel->parms.iph.tos;
661 __be16 df = tiph->frag_off;
662 struct rtable *rt; /* Route to the other host */
663 struct net_device *tdev; /* Device to other host */
664 struct iphdr *iph; /* Our new IP header */
665 unsigned int max_headroom; /* The extra header space needed */
666 __be32 dst = tiph->daddr;
667 struct flowi4 fl4;
668 int mtu;
669 const struct in6_addr *addr6;
670 int addr_type;
671
672 if (skb->protocol != htons(ETH_P_IPV6))
673 goto tx_error;
674
675 if (tos == 1)
676 tos = ipv6_get_dsfield(iph6);
677
678 /* ISATAP (RFC4214) - must come before 6to4 */
679 if (dev->priv_flags & IFF_ISATAP) {
680 struct neighbour *neigh = NULL;
681
682 if (skb_dst(skb))
683 neigh = dst_get_neighbour(skb_dst(skb));
684
685 if (neigh == NULL) {
686 if (net_ratelimit())
687 printk(KERN_DEBUG "sit: nexthop == NULL\n");
688 goto tx_error;
689 }
690
691 addr6 = (const struct in6_addr*)&neigh->primary_key;
692 addr_type = ipv6_addr_type(addr6);
693
694 if ((addr_type & IPV6_ADDR_UNICAST) &&
695 ipv6_addr_is_isatap(addr6))
696 dst = addr6->s6_addr32[3];
697 else
698 goto tx_error;
699 }
700
701 if (!dst)
702 dst = try_6rd(&iph6->daddr, tunnel);
703
704 if (!dst) {
705 struct neighbour *neigh = NULL;
706
707 if (skb_dst(skb))
708 neigh = dst_get_neighbour(skb_dst(skb));
709
710 if (neigh == NULL) {
711 if (net_ratelimit())
712 printk(KERN_DEBUG "sit: nexthop == NULL\n");
713 goto tx_error;
714 }
715
716 addr6 = (const struct in6_addr*)&neigh->primary_key;
717 addr_type = ipv6_addr_type(addr6);
718
719 if (addr_type == IPV6_ADDR_ANY) {
720 addr6 = &ipv6_hdr(skb)->daddr;
721 addr_type = ipv6_addr_type(addr6);
722 }
723
724 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
725 goto tx_error_icmp;
726
727 dst = addr6->s6_addr32[3];
728 }
729
730 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
731 dst, tiph->saddr,
732 0, 0,
733 IPPROTO_IPV6, RT_TOS(tos),
734 tunnel->parms.link);
735 if (IS_ERR(rt)) {
736 dev->stats.tx_carrier_errors++;
737 goto tx_error_icmp;
738 }
739 if (rt->rt_type != RTN_UNICAST) {
740 ip_rt_put(rt);
741 dev->stats.tx_carrier_errors++;
742 goto tx_error_icmp;
743 }
744 tdev = rt->dst.dev;
745
746 if (tdev == dev) {
747 ip_rt_put(rt);
748 dev->stats.collisions++;
749 goto tx_error;
750 }
751
752 if (df) {
753 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
754
755 if (mtu < 68) {
756 dev->stats.collisions++;
757 ip_rt_put(rt);
758 goto tx_error;
759 }
760
761 if (mtu < IPV6_MIN_MTU) {
762 mtu = IPV6_MIN_MTU;
763 df = 0;
764 }
765
766 if (tunnel->parms.iph.daddr && skb_dst(skb))
767 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
768
769 if (skb->len > mtu) {
770 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
771 ip_rt_put(rt);
772 goto tx_error;
773 }
774 }
775
776 if (tunnel->err_count > 0) {
777 if (time_before(jiffies,
778 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
779 tunnel->err_count--;
780 dst_link_failure(skb);
781 } else
782 tunnel->err_count = 0;
783 }
784
785 /*
786 * Okay, now see if we can stuff it in the buffer as-is.
787 */
788 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
789
790 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
791 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
792 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
793 if (!new_skb) {
794 ip_rt_put(rt);
795 dev->stats.tx_dropped++;
796 dev_kfree_skb(skb);
797 return NETDEV_TX_OK;
798 }
799 if (skb->sk)
800 skb_set_owner_w(new_skb, skb->sk);
801 dev_kfree_skb(skb);
802 skb = new_skb;
803 iph6 = ipv6_hdr(skb);
804 }
805
806 skb->transport_header = skb->network_header;
807 skb_push(skb, sizeof(struct iphdr));
808 skb_reset_network_header(skb);
809 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
810 IPCB(skb)->flags = 0;
811 skb_dst_drop(skb);
812 skb_dst_set(skb, &rt->dst);
813
814 /*
815 * Push down and install the IPIP header.
816 */
817
818 iph = ip_hdr(skb);
819 iph->version = 4;
820 iph->ihl = sizeof(struct iphdr)>>2;
821 iph->frag_off = df;
822 iph->protocol = IPPROTO_IPV6;
823 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
824 iph->daddr = fl4.daddr;
825 iph->saddr = fl4.saddr;
826
827 if ((iph->ttl = tiph->ttl) == 0)
828 iph->ttl = iph6->hop_limit;
829
830 nf_reset(skb);
831 tstats = this_cpu_ptr(dev->tstats);
832 __IPTUNNEL_XMIT(tstats, &dev->stats);
833 return NETDEV_TX_OK;
834
835tx_error_icmp:
836 dst_link_failure(skb);
837tx_error:
838 dev->stats.tx_errors++;
839 dev_kfree_skb(skb);
840 return NETDEV_TX_OK;
841}
842
843static void ipip6_tunnel_bind_dev(struct net_device *dev)
844{
845 struct net_device *tdev = NULL;
846 struct ip_tunnel *tunnel;
847 const struct iphdr *iph;
848 struct flowi4 fl4;
849
850 tunnel = netdev_priv(dev);
851 iph = &tunnel->parms.iph;
852
853 if (iph->daddr) {
854 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
855 iph->daddr, iph->saddr,
856 0, 0,
857 IPPROTO_IPV6,
858 RT_TOS(iph->tos),
859 tunnel->parms.link);
860
861 if (!IS_ERR(rt)) {
862 tdev = rt->dst.dev;
863 ip_rt_put(rt);
864 }
865 dev->flags |= IFF_POINTOPOINT;
866 }
867
868 if (!tdev && tunnel->parms.link)
869 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
870
871 if (tdev) {
872 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
873 dev->mtu = tdev->mtu - sizeof(struct iphdr);
874 if (dev->mtu < IPV6_MIN_MTU)
875 dev->mtu = IPV6_MIN_MTU;
876 }
877 dev->iflink = tunnel->parms.link;
878}
879
880static int
881ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
882{
883 int err = 0;
884 struct ip_tunnel_parm p;
885 struct ip_tunnel_prl prl;
886 struct ip_tunnel *t;
887 struct net *net = dev_net(dev);
888 struct sit_net *sitn = net_generic(net, sit_net_id);
889#ifdef CONFIG_IPV6_SIT_6RD
890 struct ip_tunnel_6rd ip6rd;
891#endif
892
893 switch (cmd) {
894 case SIOCGETTUNNEL:
895#ifdef CONFIG_IPV6_SIT_6RD
896 case SIOCGET6RD:
897#endif
898 t = NULL;
899 if (dev == sitn->fb_tunnel_dev) {
900 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
901 err = -EFAULT;
902 break;
903 }
904 t = ipip6_tunnel_locate(net, &p, 0);
905 }
906 if (t == NULL)
907 t = netdev_priv(dev);
908
909 err = -EFAULT;
910 if (cmd == SIOCGETTUNNEL) {
911 memcpy(&p, &t->parms, sizeof(p));
912 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
913 sizeof(p)))
914 goto done;
915#ifdef CONFIG_IPV6_SIT_6RD
916 } else {
917 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
918 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
919 ip6rd.prefixlen = t->ip6rd.prefixlen;
920 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
921 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
922 sizeof(ip6rd)))
923 goto done;
924#endif
925 }
926 err = 0;
927 break;
928
929 case SIOCADDTUNNEL:
930 case SIOCCHGTUNNEL:
931 err = -EPERM;
932 if (!capable(CAP_NET_ADMIN))
933 goto done;
934
935 err = -EFAULT;
936 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
937 goto done;
938
939 err = -EINVAL;
940 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
941 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
942 goto done;
943 if (p.iph.ttl)
944 p.iph.frag_off |= htons(IP_DF);
945
946 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
947
948 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
949 if (t != NULL) {
950 if (t->dev != dev) {
951 err = -EEXIST;
952 break;
953 }
954 } else {
955 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
956 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
957 err = -EINVAL;
958 break;
959 }
960 t = netdev_priv(dev);
961 ipip6_tunnel_unlink(sitn, t);
962 synchronize_net();
963 t->parms.iph.saddr = p.iph.saddr;
964 t->parms.iph.daddr = p.iph.daddr;
965 memcpy(dev->dev_addr, &p.iph.saddr, 4);
966 memcpy(dev->broadcast, &p.iph.daddr, 4);
967 ipip6_tunnel_link(sitn, t);
968 netdev_state_change(dev);
969 }
970 }
971
972 if (t) {
973 err = 0;
974 if (cmd == SIOCCHGTUNNEL) {
975 t->parms.iph.ttl = p.iph.ttl;
976 t->parms.iph.tos = p.iph.tos;
977 if (t->parms.link != p.link) {
978 t->parms.link = p.link;
979 ipip6_tunnel_bind_dev(dev);
980 netdev_state_change(dev);
981 }
982 }
983 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
984 err = -EFAULT;
985 } else
986 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
987 break;
988
989 case SIOCDELTUNNEL:
990 err = -EPERM;
991 if (!capable(CAP_NET_ADMIN))
992 goto done;
993
994 if (dev == sitn->fb_tunnel_dev) {
995 err = -EFAULT;
996 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
997 goto done;
998 err = -ENOENT;
999 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1000 goto done;
1001 err = -EPERM;
1002 if (t == netdev_priv(sitn->fb_tunnel_dev))
1003 goto done;
1004 dev = t->dev;
1005 }
1006 unregister_netdevice(dev);
1007 err = 0;
1008 break;
1009
1010 case SIOCGETPRL:
1011 err = -EINVAL;
1012 if (dev == sitn->fb_tunnel_dev)
1013 goto done;
1014 err = -ENOENT;
1015 if (!(t = netdev_priv(dev)))
1016 goto done;
1017 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1018 break;
1019
1020 case SIOCADDPRL:
1021 case SIOCDELPRL:
1022 case SIOCCHGPRL:
1023 err = -EPERM;
1024 if (!capable(CAP_NET_ADMIN))
1025 goto done;
1026 err = -EINVAL;
1027 if (dev == sitn->fb_tunnel_dev)
1028 goto done;
1029 err = -EFAULT;
1030 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1031 goto done;
1032 err = -ENOENT;
1033 if (!(t = netdev_priv(dev)))
1034 goto done;
1035
1036 switch (cmd) {
1037 case SIOCDELPRL:
1038 err = ipip6_tunnel_del_prl(t, &prl);
1039 break;
1040 case SIOCADDPRL:
1041 case SIOCCHGPRL:
1042 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1043 break;
1044 }
1045 netdev_state_change(dev);
1046 break;
1047
1048#ifdef CONFIG_IPV6_SIT_6RD
1049 case SIOCADD6RD:
1050 case SIOCCHG6RD:
1051 case SIOCDEL6RD:
1052 err = -EPERM;
1053 if (!capable(CAP_NET_ADMIN))
1054 goto done;
1055
1056 err = -EFAULT;
1057 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1058 sizeof(ip6rd)))
1059 goto done;
1060
1061 t = netdev_priv(dev);
1062
1063 if (cmd != SIOCDEL6RD) {
1064 struct in6_addr prefix;
1065 __be32 relay_prefix;
1066
1067 err = -EINVAL;
1068 if (ip6rd.relay_prefixlen > 32 ||
1069 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1070 goto done;
1071
1072 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1073 ip6rd.prefixlen);
1074 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1075 goto done;
1076 if (ip6rd.relay_prefixlen)
1077 relay_prefix = ip6rd.relay_prefix &
1078 htonl(0xffffffffUL <<
1079 (32 - ip6rd.relay_prefixlen));
1080 else
1081 relay_prefix = 0;
1082 if (relay_prefix != ip6rd.relay_prefix)
1083 goto done;
1084
1085 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1086 t->ip6rd.relay_prefix = relay_prefix;
1087 t->ip6rd.prefixlen = ip6rd.prefixlen;
1088 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1089 } else
1090 ipip6_tunnel_clone_6rd(dev, sitn);
1091
1092 err = 0;
1093 break;
1094#endif
1095
1096 default:
1097 err = -EINVAL;
1098 }
1099
1100done:
1101 return err;
1102}
1103
1104static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1105{
1106 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1107 return -EINVAL;
1108 dev->mtu = new_mtu;
1109 return 0;
1110}
1111
1112static const struct net_device_ops ipip6_netdev_ops = {
1113 .ndo_uninit = ipip6_tunnel_uninit,
1114 .ndo_start_xmit = ipip6_tunnel_xmit,
1115 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1116 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1117 .ndo_get_stats = ipip6_get_stats,
1118};
1119
1120static void ipip6_dev_free(struct net_device *dev)
1121{
1122 free_percpu(dev->tstats);
1123 free_netdev(dev);
1124}
1125
1126static void ipip6_tunnel_setup(struct net_device *dev)
1127{
1128 dev->netdev_ops = &ipip6_netdev_ops;
1129 dev->destructor = ipip6_dev_free;
1130
1131 dev->type = ARPHRD_SIT;
1132 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1133 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1134 dev->flags = IFF_NOARP;
1135 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1136 dev->iflink = 0;
1137 dev->addr_len = 4;
1138 dev->features |= NETIF_F_NETNS_LOCAL;
1139 dev->features |= NETIF_F_LLTX;
1140}
1141
1142static int ipip6_tunnel_init(struct net_device *dev)
1143{
1144 struct ip_tunnel *tunnel = netdev_priv(dev);
1145
1146 tunnel->dev = dev;
1147 strcpy(tunnel->parms.name, dev->name);
1148
1149 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1150 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1151
1152 ipip6_tunnel_bind_dev(dev);
1153 dev->tstats = alloc_percpu(struct pcpu_tstats);
1154 if (!dev->tstats)
1155 return -ENOMEM;
1156
1157 return 0;
1158}
1159
1160static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1161{
1162 struct ip_tunnel *tunnel = netdev_priv(dev);
1163 struct iphdr *iph = &tunnel->parms.iph;
1164 struct net *net = dev_net(dev);
1165 struct sit_net *sitn = net_generic(net, sit_net_id);
1166
1167 tunnel->dev = dev;
1168 strcpy(tunnel->parms.name, dev->name);
1169
1170 iph->version = 4;
1171 iph->protocol = IPPROTO_IPV6;
1172 iph->ihl = 5;
1173 iph->ttl = 64;
1174
1175 dev->tstats = alloc_percpu(struct pcpu_tstats);
1176 if (!dev->tstats)
1177 return -ENOMEM;
1178 dev_hold(dev);
1179 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1180 return 0;
1181}
1182
1183static struct xfrm_tunnel sit_handler __read_mostly = {
1184 .handler = ipip6_rcv,
1185 .err_handler = ipip6_err,
1186 .priority = 1,
1187};
1188
1189static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1190{
1191 int prio;
1192
1193 for (prio = 1; prio < 4; prio++) {
1194 int h;
1195 for (h = 0; h < HASH_SIZE; h++) {
1196 struct ip_tunnel *t;
1197
1198 t = rtnl_dereference(sitn->tunnels[prio][h]);
1199 while (t != NULL) {
1200 unregister_netdevice_queue(t->dev, head);
1201 t = rtnl_dereference(t->next);
1202 }
1203 }
1204 }
1205}
1206
1207static int __net_init sit_init_net(struct net *net)
1208{
1209 struct sit_net *sitn = net_generic(net, sit_net_id);
1210 int err;
1211
1212 sitn->tunnels[0] = sitn->tunnels_wc;
1213 sitn->tunnels[1] = sitn->tunnels_l;
1214 sitn->tunnels[2] = sitn->tunnels_r;
1215 sitn->tunnels[3] = sitn->tunnels_r_l;
1216
1217 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1218 ipip6_tunnel_setup);
1219 if (!sitn->fb_tunnel_dev) {
1220 err = -ENOMEM;
1221 goto err_alloc_dev;
1222 }
1223 dev_net_set(sitn->fb_tunnel_dev, net);
1224
1225 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1226 if (err)
1227 goto err_dev_free;
1228
1229 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1230
1231 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1232 goto err_reg_dev;
1233
1234 return 0;
1235
1236err_reg_dev:
1237 dev_put(sitn->fb_tunnel_dev);
1238err_dev_free:
1239 ipip6_dev_free(sitn->fb_tunnel_dev);
1240err_alloc_dev:
1241 return err;
1242}
1243
1244static void __net_exit sit_exit_net(struct net *net)
1245{
1246 struct sit_net *sitn = net_generic(net, sit_net_id);
1247 LIST_HEAD(list);
1248
1249 rtnl_lock();
1250 sit_destroy_tunnels(sitn, &list);
1251 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1252 unregister_netdevice_many(&list);
1253 rtnl_unlock();
1254}
1255
1256static struct pernet_operations sit_net_ops = {
1257 .init = sit_init_net,
1258 .exit = sit_exit_net,
1259 .id = &sit_net_id,
1260 .size = sizeof(struct sit_net),
1261};
1262
1263static void __exit sit_cleanup(void)
1264{
1265 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1266
1267 unregister_pernet_device(&sit_net_ops);
1268 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1269}
1270
1271static int __init sit_init(void)
1272{
1273 int err;
1274
1275 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1276
1277 err = register_pernet_device(&sit_net_ops);
1278 if (err < 0)
1279 return err;
1280 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1281 if (err < 0) {
1282 unregister_pernet_device(&sit_net_ops);
1283 printk(KERN_INFO "sit init: Can't add protocol\n");
1284 }
1285 return err;
1286}
1287
1288module_init(sit_init);
1289module_exit(sit_cleanup);
1290MODULE_LICENSE("GPL");
1291MODULE_ALIAS_NETDEV("sit0");